BENTOUR 2020 Archivage

Public debt and economic growth : a new assessment
El Mostafa Bentour
To cite this version:

El Mostafa Bentour. Public debt and economic growth : a new assessment. Economics and Finance.
Université Grenoble Alpes [2020-..], 2020. English. �NNT : 2020GRALE006�. �tel-03282014�
HAL Id: tel-03282014

https://theses.hal.science/tel-03282014
Submitted on 8 Jul 2021
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
THÈSE
Pour obtenir le grade de
DOCTEUR DE L’UNIVERSITE GRENOBLE ALPES
Spécialité : Sciences économiques
Arrêté ministériel : 25 mai 2016
Présentée par
EL MOSTAFA BENTOUR
Thèse dirigée par Cyriac Guillaumin , Maître de Conférences

(HDR) à l’Université Grenoble Alpes
Préparée au sein du laboratoire Centre de Recherche en
Economie de Grenoble (CREG, EA 4625)
dans l'École Doctorale Sciences Economiques (ED 300)
Dette publique et croissance économique : Une

nouvelle évaluation
Public Debt and Economic Growth : A new

assessment
Thèse soutenue publiquement le 16 octobre 2020,

devant le jury composé de :
M. Jean-Pierre ALLEGRET
Professeur à l’Université Côte d’Azur, président du jury
M. Cyriac GUILLAUMIN
Maître de Conférences (HDR) à l’Université Grenoble Alpes,
directeur de thèse
Mme Florence HUART
Maître de Conférences (HDR) à l’Université de Lille, examinateur
Mme Valérie MIGNON
Professeur à l’Université Paris Nanterre, rapporteur
M. Patrick VILLIEU
Professeur à l’Université d’ Orléans, rapporteur
UNIVERSITE GRENOBLE ALPES
ÉCOLE DOCTORALE DE SCIENCES ÉCONOMIQUES (ED 300)
CENTRE DE RECHERCHE EN ÉCONOMIE DE GRENOBLE (CREG, EA 4625)
THÈSE
Pour obtenir le grade de
Docteur de l’Université Grenoble Alpes
Discipline : Sciences Économiques
Présentée et soutenue par
El Mostafa Bentour
le 16 Octobre 2020
DETTE PUBLIQUE ET CROISSANCE ÉCONOMIQUE :

UNE NOUVELLE ÉVALUATION
Directeur de thèse
M. Cyriac Guillaumin, Maître de Conférences, habilité à diriger des recherches (HDR), à
l’Université Grenoble Alpes
Jury
M. Jean-Pierre ALLEGRET Président du Jury

Professeur à l’Université Côte d’Azur
M. Cyriac GUILLAUMIN Directeur de thèse
Maître de Conférences (HDR) à l’Université Grenoble Alpes
Mme Florence HUART Examinateur
Maître de Conférences (HDR) à l’Université de Lille
Mme Valérie MIGNON Rapporteur
Professeur à l’Université Paris Nanterre
M. Patrick VILLIEU Rapporteur
Professeur à l’Université d’Orléans
Remerciements
J’exprime, tout d’abord, ma profonde gratitude et mon entière reconnaissance à Cyriac Guillaumin,
mon directeur de thèse, pour avoir accepté d’encadrer ce travail. Je le remercie pour sa grande
disponibilité, ses conseils et suggestions utiles et ses encouragements précieux qu’il m’a prodigués
tout au long de ma thèse.
Je présente, ensuite, mes plus sincères remerciements aux membres du jury, Jean-Pierre Allegret,
Florence Huart, Valérie Mignon et Patrick Villieu, qui ont accepté d’évaluer et de donner leur
opinion d’experts sur ma thèse.
Je souhaite également remercier les membres de mon comité de suivi individuel, Jean-Pierre
Allegret, Olivier Bonroy et Faruk Ülgen, pour les discussions que nous avons eues et qui m’ont
permis de mieux comprendre ce que je souhaitais moi-même (dé)montré dans mes travaux.
Mes travaux ont également bénéficié des échanges lors des conférences et séminaires auxquels j’ai
participé. Je remercie particulièrement Christophe Bravard et Nadine Massard pour leurs précieux
conseils.
Je tiens, enfin, à remercier tous les membres de ma famille pour leur soutien et leur patience ainsi
que tous mes amis qui m’ont, de près ou de loin, aidé pour la réalisation de ce travail.
Que toutes ces personnes trouvent ici toute mon estime et toute ma considération.
2
L’Université Grenoble Alpes n’entend donner aucune approbation ou improbation aux opinions
émises dans cette thèse. Ces opinions doivent être considérées comme propre à leur auteur.
3
Table of contents
Résumé de la thèse en français ............................................................................. 8
1.Introduction.................................................................................................................................. 8
2.Contribution de la thèse à la littérature ........................................................................................ 9
2.1. Résumé du débat théorique sur les effets de la dette publique ............................................. 10
2.2. Résumé du débat empirique sur l’effet seuil de la dette publique sur la croissance
économique ........................................................................................................................................ 11
2.3. Résumé du débat sur la valeur des multiplicateurs de dépenses publiques ........................ 12
2.4. Résumé du débat sur les effets de la dette publique dans les modèles économiques .......... 14
3.Contribution empirique de la thèse ............................................................................................ 16
3.1. Présentation et choix de l’échantillon de pays ....................................................................... 16
3.2. Contribution empirique du Chapitre 1 : pas de seuil universel de dette publique ............. 17
3.3. Contribution empirique du Chapitre 2 : sensibilité du multiplicateur budgétaire aux
variations de la dette publique et au cycle économique ................................................................. 19
3.4. Contribution empirique du Chapitre 3 : Le ratio de dette publique potentiel à cibler à long
terme est une fonction endogène de la productivité potentielle du capital public ....................... 20
4.Conclusion ................................................................................................................................. 22
General Introduction........................................................................................... 24
1.Historical facts of public debt and growth in advanced economies .......................................... 25

2.Debates about public debt effects on economic growth ............................................................ 28
2.1. The theoretical debate on the public debt effects................................................................... 28
2.2. The empirical debate on the public debt threshold effects on economic growth ................ 29
2.3. The debate on the fiscal multiplier value ................................................................................ 30
2.4. The debate on the effects of public debt in economic models ............................................... 31
3.Interest of the thesis ................................................................................................................... 33
4.Plan of the thesis ........................................................................................................................ 34
Chapter 1. On the public debt and growth threshold: One size does not
necessarily fit all ................................................................................ 38
1. Introduction ............................................................................................................................... 39
2. Literature review ....................................................................................................................... 41
3. Econometric Methodology........................................................................................................ 44
4
3.1. Country specific methodology ................................................................................................. 44
3.2. Panel methodology .................................................................................................................... 47
4. Data and preliminary analysis ................................................................................................... 48
4.1 Data description ........................................................................................................................ 48
4.2 Correlations and data heterogeneity tests .............................................................................. 49
5. Country specific analysis .......................................................................................................... 50
5.1. Scatter plots analysis ................................................................................................................ 50
5.2. Regression kink results ............................................................................................................ 52
6. Panel specification analysis ...................................................................................................... 54
6.1. Scatter plots analysis ................................................................................................................ 54
6.2 Panel regressions....................................................................................................................... 55
7. Conclusion ................................................................................................................................ 58
Chapter 1. Tables and Figures ...................................................................................................... 59
Appendix A. Tables ...................................................................................................................... 72
Appendix B. Individual scatter plots for public debt and economic growth ................................ 76
Appendix C. Hansen (2017) algorithms for the regression kink model ....................................... 85
Chapter 2. Short-term effects of public debt on growth: The spending

multiplier pass-through ..................................................................... 87
1.Introduction................................................................................................................................ 88
2.Literature review ........................................................................................................................ 90
2.1. The state dependency of fiscal multipliers on the business cycle.......................................... 93
2.2. Fiscal multipliers’ dependency on the fiscal position ............................................................ 96
2.3. Fiscal multipliers in a constrained monetary policy and exchange rate regime ............... 100
2.4. The impact of the assessment methods ................................................................................. 103
2.5. Do we need a consensus about a unique fiscal multiplier size for all countries? .............. 106
3.Methodology ............................................................................................................................ 109
3.1. The general methodology of a VAR/SVAR model ............................................................... 110
3.1.1 VAR formulation ................................................................................................................. 110
3.1.2. Lag selection procedure ...................................................................................................... 111
3.1.3. Identification and analysis of shocks .................................................................................. 111
3.2. Application .............................................................................................................................. 113
4.Data .......................................................................................................................................... 117
4.1. Data source .............................................................................................................................. 117
4.2. Preliminary analysis of some previous fiscal multipliers with relation to public debt ..... 118
5.Results...................................................................................................................................... 120
5
5.1. The effects of the time periods ............................................................................................... 121
5.2. The effect of the business cycle .............................................................................................. 124
5.3. The exogenous effect of the public debt accumulation/reduction ...................................... 125
5.4. The effect of the public debt ratio movements jointly with the business cycle.................. 126
5.5. The endogenous effect of public debt .................................................................................... 127
5.6. Robustness check: the public debt crowding-in/out effects ................................................ 127
6.Conclusion ............................................................................................................................... 136
Appendix A. Tables .................................................................................................................... 138
Appendix C. Data........................................................................................................................ 179
Chapter 3. Public debt effects in theory-based models with an empirical

assessment of the potential public debt ......................................... 187
1. Introduction ............................................................................................................................. 188

2. Literature review ..................................................................................................................... 190
2.1. Government debt effects in the ILA and OLG models ....................................................... 190
2.2. Government debt in endogenous growth models ................................................................. 196
2.3. Government debt in the new Keynesian models and the positive approach ...................... 203
2.3.1. Government debt in the new Keynesian models ................................................................. 203
2.3.2. The positive approach of public debt .................................................................................. 204
2.4. Discussion.................................................................................................................................. 207
3. The theoretical framework ...................................................................................................... 209
3.1. The choice of an endogenous growth model .......................................................................... 209
3.2. Justification of government capital stock and government expenditure flows in the
production function ........................................................................................................................ 210
3.3. The model ................................................................................................................................. 212
3.3.1. The productive sector .......................................................................................................... 212
3.3.2. The human capital sector .................................................................................................... 215
3.3.3. The quality of public institutions ........................................................................................ 215
3.3.4. The government sector ........................................................................................................ 219
3.3.5. The productive potential government capital and potential government debt .................... 220
4. Empirical evidence.................................................................................................................. 221
4.1. Estimation of the Cobb-Douglas production function .......................................................... 222
4.2. Simulations of targeted/potential public debt ratios ............................................................. 229
4.3. Robustness check ..................................................................................................................... 237
4.3.1. The impact of data shortness ............................................................................................... 237
4.3.2. The impact of elasticity ....................................................................................................... 239
4.3.3. The impact of the differential “long-term interest rate – economic growth rate” ............... 240
4.4. Discussion.................................................................................................................................. 242
6
5. Conclusion .............................................................................................................................. 247
Appendix A. Tables .................................................................................................................... 249
Appendix B. Figures ................................................................................................................... 258
General Conclusion ........................................................................................... 265
References........................................................................................................... 270
7
Résumé de la thèse en français
Dette publique et croissance économique : une nouvelle évaluation
1. Introduction
Au cours de la dernière décennie, la dette publique est devenue de plus en plus élevée, atteignant
des niveaux inquiétants, pour de nombreux pays avancés, en particulier dans la zone euro. Ces
pays avaient mis en œuvre des mesures de relance budgétaire massives en 2009 en réponse aux
graves ralentissements suite à la crise financière de 2008-2009. Deux ans après, les pays de la zone
euro ont commencé à inverser le cours des expansions budgétaires, en s’orientant vers des
consolidations fiscales, afin de stabiliser les ratios de déficit et d’endettement publics exacerbés
par les stimuli budgétaires conjugués à une croissance en berne. En effet, les préoccupations quant
aux défauts de paiement de la dette souveraine dans de nombreux pays, notamment de la périphérie
de la zone euro, ont conduit aux mesures de consolidations budgétaires et d’austérité. Face à ces
mesures, les pouvoirs publics et les économistes ont été partagés entre les effets de court terme des
mesures d’assainissement budgétaire et les effets de long terme reliés notamment à l’impact de la
dette publique sur la croissance économique.
Pour ce qui est des effets de long terme en particulier, les économistes semblent être au moins
unanimes en théorie sur l’impact négatif de la dette publique sur la croissance économique à long
terme. Ce constat est devenu tellement ancré dans l’esprit de tout un chacun qu’elle semble
légitimer les affirmations quant à l’éventuelle existence d’un seuil de dette publique au-delà duquel
la croissance économique serait altérée. C’est ainsi que deux économistes (Reinhart et Rogoff,
2010) ont déterminé, pour un échantillon de pays développés, l’existence d’un seuil de dette
publique (rapportée au PIB) de 90%. À la suite de ce résultat, dont l’implication est forte en termes
de politiques économiques, surtout parvenu en temps de crise où les finances publiques sont sous
pression, un afflux incessant de travaux a émergé, examinant la relation entre la dette publique et
la croissance économique et suscitant plusieurs controverses (Caner et al., 2010 ; Kumar et Woo,
2010 ; Reinhart et Rogoff, 2010 ; Lin, 2014 ; Bentour, 2018).
En ce qui concerne les effets de la dette publique à court terme, ils pourraient être appréciés dans
un contexte de politique budgétaire en relation avec les multiplicateurs keynésiens. En effet, la
dette publique est contractée pour faire face aux besoins de financement des déficits publics,
lorsque les recettes publiques ne couvrent pas les dépenses publiques. Ainsi, les effets imminents
de la dette publique sur la croissance sont aussi liés aux multiplicateurs de politique budgétaire,
dont la valeur peut déterminer la capacité à générer la croissance économique aussi bien que les
choix des politiques économiques (programme de stimuli versus austérité). En effet, une valeur
élevée (faible) du multiplicateur de dépenses peut générer une croissance élevée (faible) pour une
politique budgétaire expansionniste et ainsi réduire (augmenter) le poids de la dette. Étant donné
leur importance, les multiplicateurs budgétaires ont aussi fait l’objet d’échanges intenses dans les
milieux académiques à la suite de la crise de 2008-2009. Ainsi, si Reinhart et Rogoff (2010) ont
été les précurseurs de l’intense débat sur le seuil de la dette publique, c’est grâce à Auerbach et
Gorodnichenko (2012) que le débat autour de la valeur des multiplicateurs notamment budgétaires
a été déclenché. Ces auteurs ont révélé la sensibilité des multiplicateurs au cycle conjoncturel,
notamment faibles en temps d’expansion et élevés en temps de récession économique. Dès lors,
d’autres recherches furent publiées, ayant tendance à confirmer ce résultat et débattant davantage
sur la dépendance des multiplicateurs à d’autres déterminants tels que le déficit public,
l’endettement public, le régime de change, la politique monétaire, le degré d’ouverture, etc.
(Romer et Romer, 2011 ; Ramey, 2011, 2018 ; Auerbach et Gorodnichenko, 2012, 2013 ; Delong
et Summers, 2012 ; Farhi et al., 2017 ; Ramey et Zubairy, 2018).
En s’appuyant sur les limites économétriques des études précédentes, mises en évidence dans les
différents débats à propos des effets de la dette publique sur la croissance économique depuis la
crise de 2008-2009, cette thèse se propose d’analyser la pertinence des arguments en faveur et/ou
en défaveur d’un seuil universel présenté supra, que la théorie économique ne semble pas signaler
formellement. Nous montrons notamment, que l’existence d’un tel seuil, surtout commun à tous
les pays, semble créer un clivage plutôt qu’un consensus au niveau des résultats empiriques. Ainsi,
outre la revue de littérature théorique et empirique vigoureusement analysée, nous adoptons une
démarche économétrique basée sur des arguments et des modèles économiques pour montrer que
la relation entre la dette publique et la croissance économique est davantage un lien endogène
émanant des canaux et variables par lesquelles les effets de la dette se transmettent à la croissance
économique. De plus, nous montrons que ces canaux et effets qu’ils induisent diffèrent selon
l’horizon temporel amenant à une distinction de traitement et de modélisation tenant compte des
spécificités propres à chaque pays. Ceci a abouti à une formulation temporelle simulée de la limite
de la dette (potentielle) en relation avec les performances de chaque pays en termes notamment de
taux d’intérêt de long terme et de productivité potentielle du capital publique.
2. Contribution de la thèse à la littérature et débats des effets de la dette

publique sur la croissance économique
Dans le volet consacré à la revue de littérature, la contribution de la thèse est large et diversifiée.
La thèse est constituée de trois chapitres dont chacun assemble et analyse une revue de littérature
autour des débats. Le chapitre 1 évoque les récents débats empiriques quant à l’existence d’un
seuil de dette publique affaiblissant la croissance économique, sans omettre les différents débats
théoriques sur les effets de la dette publique sur la croissance économique distinguant notamment
la vision classique de long terme de la vision keynésienne de court terme. Partant de cette dernière,
le chapitre 2 étale une riche littérature évaluant ainsi les effets de la dette publique sur la croissance
économique à travers l’effet des multiplicateurs de dépenses publiques. La littérature à ce sujet est
9
essentiellement portée sur les différentes nouvelles contributions autour de la taille du

multiplicateur qui s’est révélée sensible à une variété de déterminants dont notamment la position
de l’économie dans le cycle et la situation fiscale (dette et déficit publics). Le chapitre 3 quant à
lui se charge de présenter une revue de littérature très riche et contrastée des modèles évaluant les
effets de la dette sur la croissance particulièrement dans l’optique de long terme, tout en résumant
les avantages et les critiques à l’encontre de ces modèles.
2.1. Résumé du débat théorique sur les effets de la dette publique
En théorie, les effets de la dette sur la croissance économique ont été particulièrement débattus
dans l’après-guerre suite à la forte augmentation de la dette publique des pays avancés. La
littérature distingue particulièrement la vision keynésienne de court terme qui suppose que les
dépenses publiques financées par la dette publique sont susceptibles de stimuler la demande
globale dans un contexte de rigidité des prix et des salaires. Cependant, selon la vision classique
dont la tendance est le long terme, la dette publique est susceptible de réduire le stock de capital
et la productivité, puis de réduire la production.
Divers canaux sont cités pour expliquer les principales raisons de ces effets, comme le résume
Hansen (1959). Une dette publique plus élevée peut déclencher une épargne privée plus élevée
(effet Pigou), moins d’incitations à travailler et à investir en particulier pour les détenteurs des
obligations d’État (effet Kaldor), un effet d’incitation négatif des impôts supplémentaires
nécessaires pour financer le paiement des intérêts et des intérêts plus élevés défavorables pour
contrer l’impact inflationniste de “l’effet Pigou”. En outre, Modigliani (1961) a déclaré que la dette
publique peut évincer l’investissement privé en réduisant le crédit à l’économie ou en augmentant
les taux d’intérêt à long terme sur les emprunts publics. Les points de vue des keynésiens et des
classiques sont résumés dans ce qui est appelée par la suite “l’analyse conventionnelle” des effets
de la dette publique, reflétant le paradigme dominant des chercheurs (Elmendorf et Mankiw,
1999).
Une autre contribution théorique importante alimentant le débat sur les effets du financement par
la dette publique est “l’équivalence ricardienne”. L’idée attribuée à l’origine à Ricardo, affirme
que la croissance économique peut être insensible à la dette publique, du fait que les
consommateurs, supposés prospectifs (à anticipations rationnelles) peuvent réagir en réduisant
leurs dépenses d’un montant équivalent à celui d’une augmentation des dépenses publiques. Ils
anticipent en effet des impôts futurs finançant le déficit généré par les nouvelles dépenses
publiques et par conséquent épargnent et renoncent à la dépense. Dans le même sens, Barro (1974)
montre que les obligations d’État constituent un actif pour leurs propriétaires et un passif pour les
contribuables. Dans l’ensemble, aucune richesse nette ne serait créée et l’effet serait nul. Les
discussions entre économistes sur les effets de la dette publique dans l’optique de l’équivalence
ricardienne, ont aussi invoqué les échanges quant à la redistribution de tels effets sur les
générations successives. Cette redistribution est surtout reconsidérée dans le cadre des transferts
10
intergénérationnels dans des modèles à générations imbriquées (Diamond, 1965 ; Blanchard,

1985).
Quant au débat sur l’analyse à court terme, il se concentre en particulier sur la composition des
dépenses publiques pour lesquelles les économistes (classiques et néoclassiques notamment) ont
tendance à considérer que les dépenses en capital peuvent générer des effets positifs mais pas
nécessairement les dépenses courantes. Ce point de vue est contrasté par les keynésiens qui
soutiennent que les effets des dépenses sont toujours positifs, avec des différences en termes
d’efficacité qui peuvent être plus élevées pour certains types que pour d’autres. Ces opinions
opposées sont résumées dans Aschauer (2000) qui suppose que la relation à court terme entre la
dette et la croissance pourrait être positive en raison de l’impact des dépenses d’investissement sur
la croissance, tandis que la relation à long terme pourrait être inversée en raison de la prime de
risque élevée, à la suite d’une dette publique élevée, augmentant ainsi le coût de la dette. Pour
Blanchard (2006), l’association négative entre la dette et la croissance pourrait être attribuée au
retard des politiques fiscales contracycliques, ce qui signifie que les pays en récession devraient
adopter des politiques budgétaires expansionnistes pour stimuler la croissance et, en période
d’expansion, des politiques restrictives pour réduire les niveaux d’endettement.
2.2. Résumé du débat empirique sur l’effet seuil de la dette publique sur la croissance
économique
Dans les économies avancées, l’augmentation de la dette publique pendant la période précédant la
crise a été accentuée par les mesures de relance budgétaire mises en œuvre en 2008-2009. Une
année plus tard, Reinhart et Rogoff (2010) ont constaté que, pour un échantillon de 20 pays
avancés, il existe un seuil commun de 90% du ratio de la dette publique au PIB au-delà duquel la
croissance économique serait altérée. En particulier, ces deux auteurs ont observé que la
corrélation entre la dette publique et la croissance est faible pour des niveaux d’endettement
modérés et devient forte et négative lorsque le ratio de dette publique, mesuré par le rapport entre
la dette publique et le PIB, dépasse 90%. Ils observent que les taux de croissance médian et moyen
correspondant aux ratios d’endettement supérieurs à ce seuil diminuent respectivement de 1 et 4
points. Ce résultat, basé sur une approche statistique descriptive simple, a déclenché un afflux de
recherches empiriques, utilisant différentes méthodes économétriques, dont la plupart des résultats
ont suscité beaucoup de controverses.
Ainsi, certaines études ont tendance à confirmer l’existence d’un seuil de dette publique sans
nécessairement parvenir à la valeur de 90% supposée par Reinhart et Rogoff (2010) (Caner et al.,
2010 ; Checherita et Rother, 2010 ; Kumar et Woo, 2010 ; Lin, 2014). Néanmoins, le débat autour
du seuil de la dette publique s’est intensifié, après que plusieurs auteurs aient contesté les
conclusions de Reinhart et Rogoff (2010) comme, par exemple, Fergusson et Johnson (2011),
Minea et Parent (2012), Baglan et Yoldas (2013), Pescatori et al. (2014), Egert (2015). Ces
11
contestations ont été notamment alimentées par la révélation de certaines erreurs de codification
et de calcul dans l’article de Reinhart et Rogoff (2010) par Herndon et al. (2013).
Le résumé de ces résultats peut être regroupé en trois groupes :
- un premier groupe de recherches empiriques se concentrant principalement sur l’examen

d’une relation concave non linéaire entre la dette publique et la croissance économique
(Checherita et Rother, 2010 ; Kumar et Woo, 2010 ; Pescatori et al., 2014 ; Eberhardt et
Presbitero, 2015 ; Chudik et al., 2017) ;
- un deuxième groupe intéressé par la causalité entre la dette publique et la croissance

économique (Panizza et Presbitero, 2012, 2013 ; Di Sanzo et Bella, 2015 ; Gomez-Puig et
Sosvilla-Rivero, 2015) ;
- un troisième groupe qui examine les effets d’autres variables institutionnelles et

macroéconomiques interférant avec la relation entre la dette publique et la croissance
économique (Greiner, 2011 ; Kourtellos et al., 2013 ; Pan et Wang, 2013 ; Sharpe, 2013 ;
Marchionne et Parekh, 2015).
Les résultats de toutes ces recherches n’ont abouti à aucun consensus. Spécifiquement, certains
chercheurs ont repris l’étude de la relation dette-croissance pour le même échantillon de 20 pays
avancés utilisé par Reinhart et Rogoff (2010), adoptant différentes approches économétriques, et
ont abouti à des résultats différents indiquant généralement de faibles niveaux de seuils
d’endettement autour de 20 à 30% (Baglan et Yoldas, 2013 ; Egert, 2015 ; Lee et al., 2017).
D’autres chercheurs ont utilisé différents échantillons et différentes périodes rejetant ainsi
l’existence de tout seuil (Pescatori et al., 2014 ; Eberhardt et Presbitero, 2015 ; Chudik et al., 2017 ;
Syssoyeva-Masson et De Sousa Andrade, 2017).
En règle générale, les évaluations des effets des politiques budgétaires sont quantifiées par les
valeurs du multiplicateur budgétaire. L’efficacité des dépenses publiques dans l’analyse
keynésienne/classique (comme mentionné au I.1) dépendant de la composition et du type de
dépenses (courantes, en capital, de transferts, etc.) est empiriquement liée aux valeurs de leurs
multiplicateurs. De ce côté, un volet important de recherches empiriques a été lancé au lendemain
de la crise financière, révélant notamment davantage la sensibilité de ces multiplicateurs à de
nombreux déterminants économiques et institutionnels.
2.3. Résumé du débat sur la valeur des multiplicateurs de dépenses publiques
Bien que la majorité des recherches précédentes (Section I.2) se concentrent sur l’étude du seuil
de la dette publique et considèrent la relation dette-croissance à travers des formulations et modèles
économétriques non linéaires, les canaux par lesquels les effets de la dette publique se matérialisent
sont proprement liés aux effets des multiplicateurs de dépenses financées par une telle dette, en
12
particulier à court terme. L’évaluation des multiplicateurs budgétaires a été largement reconsidérée
à la suite du développement des techniques économétriques. Depuis la crise financière de 2008,
une littérature massive évaluant les multiplicateurs budgétaires est apparue (Romer et Romer,
2011 ; Ramey, 2011, 2018 ; Auerbach et Gorodnichenko, 2012, 2013 ; Delong et Summers, 2012 ;
Farhi et al., 2017 ; Ramey et Zubairy, 2018).
Les principales conclusions de cette littérature révèlent de grandes différences dans la taille des
multiplicateurs. En effet, la taille du multiplicateur s’est montrée sensible à une variété de
déterminants qui n’étaient pas (ou rarement) considérés avant la crise de 2008-2009. Les
multiplicateurs budgétaires sont particulièrement sensibles à la période d’estimation, propres à
chaque pays, et à la méthode d’évaluation (Batini et al., 2012 ; Baum et al., 2012). En outre, les
multiplicateurs budgétaires se sont révélés sensibles au cycle économique, en particulier, les
multiplicateurs de dépenses budgétaires ont tendance à être plus importants dans les récessions
que dans les périodes d’expansions (Auerbach et Gorodnichenko, 2012, 2013 ; Barro et Redlick,
2011 ; Parker, 2011 ; Corsetti et al., 2012 ; Caggiano et al., 2015 ; Fazzari et al., 2015 ; Glocker et
al., 2019).
Par ailleurs, d’autres chercheurs conduisent à découvrir la vulnérabilité des multiplicateurs

budgétaires à d’autres déterminants sans être nécessairement conditionnés par la cyclicité de
l’économie. Ces déterminants sont la situation budgétaire mesurée par le niveau des ratios
d’endettement et/ou des déficits (Corsetti et al., 2013 ; Huidrom et al., 2016), la politique
monétaire contrainte soit par la tendance des taux d’intérêt à leur borne inférieure zéro (trappe à
liquidité) ou par la perte de l’indépendance monétaire comme dans le cas d’un taux de change fixe
ou d’une union monétaire (Hall, 2009 ; Cogan et al., 2010 ; Christiano et al., 2011 ; Delong et
Summers, 2012 ; Farhi et Werning, 2017).1
Néanmoins, les travaux très récents de Ramey (2018) et Ramey et Zubairy (2018) considèrent que
les multiplicateurs de dépenses publiques sont en moyenne inférieurs à l’unité. Cela contraste la
tendance des études conduites après la crise de 2008-2009 et dont les résultats parviennent à des
multiplicateurs plus importants, en temps de récessions, pouvant atteindre la valeur de 2. Cela va
également dans le sens du même consensus sur les multiplicateurs de dépenses avant la récession
de 2008-2009, considéré comme faible et que les effets de la politique budgétaire sont de très
courte durée (Coenen et al., 2012). En particulier, les effets de celle-ci tenant compte de la situation
budgétaire de l’économie mesurée par le niveau de la dette publique et/ou du déficit public sont
fortement débattus au lendemain de la crise financière de 2008-2009 (Boussard et al., 2012 ;
Corsetti et al ., 2012 ; Blot et al., 2014b ; Canzoneri et al., 2015 ; Bi et al., 2016 ; Huidrom et al.,
1
Bentour (2020) a aussi trouvé la sensibilité des multiplicateurs de dépenses aux fluctuations des prix du pétrole pour
un nombre de pays de la région MENA. En particulier, ces multiplicateurs sont positifs et très élevés (dépassant la
valeur de 2) notamment pour les pays exportateurs de pétrole, pour des périodes de baisse du prix de pétrole, alors
qu’ils sont faibles, voire négatifs pour certains pays, en période d’expansions des prix pétroliers.
13
2016 ; Perdichizzi, 2017 ; Poghosyan, 2017 ; Auerbach et Gorodnichenko, 2017 ; Afonso et Leal,
2018 ; Blanchard, 2019 ; Broner et al., 2019).
Très récemment, lors de la conférence présidentielle de l’American Economic Association,

Blanchard (2019) a déclenché une nouvelle vague du débat sur la dette publique et la croissance
économique liée au coût budgétaire d’une dette publique élevée ainsi qu’à son effet sur le bien-
être. Il a particulièrement minimisé les inquiétudes concernant le coût de la dette publique pour
l’économie américaine comme, historiquement, le taux d’intérêt nominal est resté toujours en
moyenne inférieur au taux de croissance nominal, exception faite de certaines petites périodes
autour des années 1980.
2.4. Résumé du débat sur les effets de la dette publique dans les modèles économiques
Pour évaluer les effets de la dette publique, de nombreux modèles de croissance théoriques ont été
conçus. Le modèle de Ramsey (1928) était la base de la première classe des modèles d’agents à
vie infinie (ILA). Plus tard, le débat entre économistes sur l’horizon de vie des agents et le type de
liens de transferts opérationnels intergénérationnels entre eux a abouti au développement des
modèles à générations imbriquées considérant l’existence/l’absence de continuité, entre les
générations, assurée par ces liens (Samuelson, 1958 ; Diamond, 1965). Le débat a porté sur
l’implication de l’altruisme entre les générations et le cycle de vie des agents pour déterminer
l’existence d’effets de la dette publique sur les comportements des agents en termes d’épargne,
d’accumulation du capital, d’utilité des consommateurs et du taux d’intérêt. Les contributions
importantes dans ce sens remontent à Diamond (1965), Yaari (1965), Barro (1974), Blanchard
(1985), Buiter (1988), Aiyagari (1985, 1987), Weil (1989) et Ni (1999).
Les effets redistributifs d’une génération à l’autre ont fait l’objet de nombreux débats dans l’après-
guerre. Buchanan (1958) et Meade (1958), discutant des contributions de leurs prédécesseurs,
notamment Lerner (1943), montrent que la dette publique ne constitue pas un fardeau pour les
contribuables actuels. Ce fardeau est plutôt transféré partiellement ou totalement aux générations
futures qui devront payer des impôts pour rembourser la dette (Bowen et al., 1960 ; Modigliani,
1961). Dans le même sens, les partisans de l’équivalence ricardienne montrent que la neutralité de
la dette peut se produire en fonction notamment de l’existence de liens altruistes opérationnels
entre générations (Barro, 1974). Ainsi, le débat a émergé modélisant les effets de la politique
budgétaire, précisément l’échange (swap) de la dette et son effet sur l’utilité sociale et les taux
d’intérêt. Cela a également déclenché un débat controversant entre les néoclassiques et les
partisans de l’équivalence ricardienne (Barro, 1976 ; Feldstein, 1976 ; Cukierman et Meltzer,
1989).
Les effets sur la dette publique sont également pris en compte dans l’économie politique de la dette
publique. À cet égard, deux approches ont été débattues. D’une part, l’approche normative où le
gouvernement est considéré comme un planificateur social pour qui la priorité est de maximiser le
bien-être social de ses individus (Barro, 1979 ; Lucas et Stockey, 1983 ; Aiyagari et al., 2002).
14
D’autre part, l’approche positive considérant la dette publique comme une variable d’État utilisée
par chaque gouvernement comme une stratégie pour influencer les choix de son successeur ou
comme un moyen de façonner les anticipations des agents économiques privés (Svensson et
Persson, 1989 ; Alesina et Tabellini, 1990). Par conséquent, les idées les plus répandues étaient
liées aux effets de la politique budgétaire sous des gouvernements qui suivent des règles
d’engagement (commitment) contre des politiques discrétionnaires. En particulier, l’incohérence
temporelle du gouvernement a un impact sur la façon dont les agents économiques forment leurs
anticipations affectant ainsi leurs décisions économiques (Kydland et Prescott, 1977).
Les effets de la dette publique sont également examinés dans la classe des modèles de croissance
endogène. Le développement de tels modèles a été essentiellement popularisé par Romer (1986),
Lucas (1988) et Barro (1990). Ces modèles ont été proposés comme une alternative au modèle de
croissance néoclassique de Solow (1956) et Swan (1956). Celui-ci n’a pas été en mesure
d’expliquer les taux de croissance non nuls du PIB par tête persistant dans de nombreuses
économies développées, et a donc été remis en cause pour avoir omis les déterminants de
croissance à long terme. Ainsi, le cadre de croissance endogène englobe d’autres déterminants de
la croissance à long terme, en particulier pour l’évaluation des politiques budgétaires, incorporant
le secteur public dans le secteur productif (Lucas, 1988 ; Romer, 1989 ; Barro, 1990 ; Barro et
Sala-i -Martin, 1992, 1995 ; Futagami, 1993 ; Jones, 1995, 2003 ; Corsetti et Roubini, 1996 ;
Turnovsky, 1997 ; Greiner, 2007, 2012, 2016 ; Futagami et al., 2008 ; Maebayashi et al., 2017).
Récemment, la politique budgétaire, et en particulier la dette publique a également été modélisée

dans la classe des modèles néo-keynésiens. Les travaux pionniers dans ce sens sont principalement
dus à Christiano et al. (2005) et Smets et Wouters (2007). Ces derniers diffèrent des autres modèles
en assumant des taxes générant des distorsions au lieu des taxes forfaitaires comme cela est
particulièrement supposé dans les modèles à générations imbriquées. De plus, ils rejoignent la
littérature de l’économie politique de la dette en discutant des politiques gouvernementales selon
des règles d’engagement ou de discrétion. En particulier, certains auteurs considèrent que la dette
publique optimale suivrait un processus de marche aléatoire chaque fois que le gouvernement peut
atteindre un engagement politique incohérent dans le temps (Benigno et Woodford, 2003 ;
Schmitt-Grohe et Uribe, 2004). D’importantes recherches empiriques sur ces modèles ont
récemment prospéré (Leith et Wren-Lewis, 2013 ; Mayer et al., 2013).
Malgré les contributions importantes des modèles précédents dans l’évaluation de la dette publique
et des effets de la politique budgétaire, les nouveaux modèles keynésiens ont été particulièrement
critiqués par Mankiw (2000). Celui-ci a contribué par la théorie des “Savings-Spenders” qui a
influencé de nombreuses recherches empiriques sur la politique budgétaire en essayant de
considérer les comportements suivant des règles empiriques, en particulier dans les modèles
d’inspiration de la nouvelle économie keynésienne. Par conséquent, Chari et al. (2009) montrent
que cette classe de modèle n’est pas à présent utile pour l’analyse des politiques publiques. Ces
auteurs se basent en particulier sur la critique du modèle de Smets et Wouters (2007) qui constitue
15
le noyau et la référence fondamentale pour de nombreuses contributions récentes adoptant le cadre

d’analyse de la nouvelle économie keynésienne.
3. Contribution empirique de la thèse
Cette partie résume les principaux résultats empiriques et aboutissements de la thèse à la suite de
l’évaluation économétrique des liens entre la dette publique et la croissance économique.
Premièrement, nous présentons la contribution économétrique du chapitre 1 étudiant la forme non
linéaire entre la dette publique et la croissance économique pour un échantillon de 20 pays avancés.
Cette relation fait apparaître l’existence d’un point de retournement dans la relation pour chaque
pays permettant de confirmer ou d’infirmer l’existence d’un seuil de dette à ne pas dépasser.
Deuxièmement, nous détaillons les résultats du chapitre 2 relatifs à l’étude des multiplicateurs de
dépenses publiques en fonction du cycle conjoncturel et le sens d’évolution du ratio d’endettement.
Troisièmement, Nous résumons la contribution empirique du chapitre 3 reliant la dette publique à
la croissance économique dans un cadre de croissance endogène, à travers la productivité du capital
public.
Les principales conclusions de nos évaluations empiriques conduisent à rejeter le taux

d’endettement commun qui s’applique à tous les pays. Nous montrons spécifiquement que la
relation entre la dette publique et la croissance économique est plutôt sensible à la période
d’analyse et surtout propre à chaque pays. En outre, dans un cadre à court terme, l’accumulation
et la contraction de la dette publique ont un impact différent sur la croissance économique via
l’effet des multiplicateurs de dépenses. Ces derniers se révèlent, notamment dans notre thèse,
sensibles aux autres facteurs liés au cycle conjoncturel (expansion, récession), à l’évolution du
taux d’endettement public (accumulation, contraction) et au croisement de ces facteurs. À long
terme, la dette publique est liée, dans le cadre d’un modèle de croissance endogène, aux
fondamentaux macroéconomiques des économies et à la productivité du capital public.
3.1. Présentation et choix de l’échantillon de pays
Compte tenu de l’endettement élevé des économies avancées, nous nous sommes concentrés sur
un échantillon de 20 pays de l’OCDE, dont 11 sont membres de la zone euro, à savoir l’Australie,
l’Autriche, la Belgique, le Canada, le Danemark, la France, l’Allemagne, la Grèce, l’Irlande,
l’Italie, le Japon, les Pays-Bas, la Nouvelle-Zélande, la Norvège, le Portugal, l’Espagne, la Suède,
la Suisse, le Royaume-Uni et les États-Unis2. Le choix de cet échantillon est justifié par de
nombreuses raisons dont :
2
Exception faite au chapitre 2 qui est restreint à 18 pays en omettant l’Australie et la Nouvelle Zélande suite au
manque de données trimestrielles suffisantes pour ces pays.
16
- l’importance économique de ce groupe de pays avancés influençant les économies

mondiales notamment du fait qu’ils pèsent sur le système économique et financier
international (en termes de risque systémique) ;
- le niveau élevé de la dette publique par rapport au PIB accumulé pour la majorité de ces
pays, dont plusieurs d’entre eux ont rencontré des problèmes de dette souveraine lors de la
crise financière de 2008-2009, voire même avant ;
- l’échantillon est constitué des pays qui ont été très étudiés dans la littérature et dont les
résultats sont débattus après la crise financière de 2008-2009 en termes de dette publique
et de politique budgétaire, ce qui nous permet de confronter nos résultats à ceux de la
littérature existante ;
- enfin, la disponibilité de séries de données longues, tant annuelles que trimestrielles pour
la dette publique et la croissance économique.
3.2. Contribution empirique du Chapitre 1 : pas de seuil universel de dette publique
Malgré la diversité des méthodes économétriques utilisées pour étudier la relation croissance-dette,
la plupart des analyses précédentes souffrent d’un biais d’échantillonnage (courtes périodes
d’analyse). Presque toutes les méthodes économétriques avancées utilisées sont appliquées à des
échantillons débutant dans les années 1970. Cette date coïncide avec la fin du système de Bretton
Woods et le début de la libéralisation des marchés dans les pays développés. Pour remédier à ce
biais, l’analyse a été étendue dans le chapitre 1 sur une longue période, 1880-2008, subdivisée en
cinq sous-périodes correspondant aux changements majeurs survenus dans l’ordre monétaire et
politique international (Rodrik, 2011 ; Obstfeld et Taylor, 2002). Ces changements pourraient
affecter la stabilité des agrégats macroéconomiques et leurs interdépendances, dont éventuellement
la relation dette-croissance. Notre analyse distingue les sous-périodes suivantes :
- 1880-1913 : cette période correspond à la fin de la première mondialisation ou la fin de

l’ère du mercantilisme ;
- 1914-1945 : une période avec deux guerres dévastatrices et la grande dépression

économique de 1929. L’ordre économique international a été marqué par le régime de
l’étalon-or et le déclin de l’hégémonie britannique remplacé par la suprématie américaine ;
- 1946-1970 : le monde a connu une croissance et un développement solides au cours de

cette période, notamment sous les accords de Bretton Woods et les régimes de taux de
change fixes (centrés autour du dollar américain). Cette période de près de trois décennies
de prospérité est désignée par le terme des trente-glorieuses ;
- 1971-1990 : ces deux décennies ont connu des événements économiques et politiques
turbulents avec la fin de la convertibilité du dollar en 1971 et l’augmentation des régimes
17
de change flottants, les chocs pétroliers des années 1970 avec la coexistence du chômage
et de l’inflation et la crise de la dette souveraine des années 1980 ;
- 1991-2008 : cette période a connu une large libéralisation financière conformément aux
recommandations du Consensus de Washington, entraînant une instabilité financière et des
crises pour la plupart des pays émergents. Au cours de cette même période, l’organisation
mondiale du commerce (OMC) a été créée, les accords de libre-échange bilatéraux et
multilatéraux ont proliféré et la zone euro a été établie.
D’abord, une analyse préliminaire descriptive variée par pays et en panel a été conduite. Puis, une
forme quadratique de la croissance économique en fonction du ratio d’endettement sur toute la
période 1880-2008 et les sous-périodes susmentionnées a été estimée3. La méthode d’estimation
GMM a été utilisée corrigeant le problème d’endogénéité dû à la causalité inverse allant de la
croissance au taux d’endettement. L’analyse descriptive par pays et par panel ainsi que les
régressions ont révélé la sensibilité de la relation à la période et aux échantillons de pays et est
altérée par l’hétérogénéité des pays. L’hétérogénéité diminue lorsque l’on omet des pays
importants de l’échantillon et se manifeste moins dans l’échantillon des pays de la zone euro.
À la suite de cette hétérogénéité détectée, nous avons jugé primordial le recours à des régressions
par pays, ce qui, à notre connaissance, était rarement mené dans les recherches empiriques
précédentes. Ainsi, le chapitre a également étudié le lien entre la croissance et la dette publique en
utilisant une nouvelle et innovante méthode appelée “regression kink” développée par Hansen
(2017). Cette méthode a le pouvoir de détecter de manière endogène les points de tournure dans
les relations économétriques étudiées. Appliquée à la relation dette et croissance pour chaque pays,
elle met en évidence différentes formes de relations entre la dette publique et la croissance
économique. Par conséquent, certains pays peuvent croître avec des ratios d’endettement public
élevés, pendant que d’autres pourraient voir leur croissance se contracter par rapport à des ratios
d’endettement même faibles, tandis que la croissance dans d’autres pays s’avère insensible à la
dette publique. L’étude révèle également l’instabilité de la relation dans le temps. Chaque pays se
distingue par des relations différentes selon les périodes d’estimation considérées, en particulier
lorsque la transition se fait entre des périodes connues pour certains changements spécifiques du
système monétaire international.
Tous ces résultats rejettent l’existence d’un seuil commun et sa valeur de 90% comme le prétendent
Reinhart et Rogoff (2010). Néanmoins, il faut signaler que les travaux précédents n’adoptent pas
généralement un cadre de modélisation impliquant d’autres équations et variables par lesquelles la
dette publique interagit avec la croissance économique. Par conséquent, il convient d’étudier les
effets de la dette publique sur la croissance économique considérant les caractéristiques propres à
3
Cette méthode a été utilisée par Checherita et Rother (2010) pour un échantillon différent de pays et une période
relativement courte comparée à celle de ce chapitre.
18
chaque pays, sans s’affranchir notamment des modèles théoriques pouvant expliquer les canaux
de transmission des effets de la dette publique à la croissance économique.
3.3. Contribution empirique du Chapitre 2 : sensibilité du multiplicateur budgétaire

aux variations de la dette publique et au cycle économique
Le deuxième chapitre est particulièrement consacré à approfondir la relation entre la dette publique
et la croissance économique à court terme, notamment via l’évaluation des effets de la dette
publique à travers les multiplicateurs de dépense. L’importance de cet exercice s’est inscrite dans
le débat déclenché sur les multiplicateurs budgétaires par Auerbach et Gorodnichenko (2012),
prouvant qu’ils sont plus élevés en période de récession et plus faibles en temps d’expansion. En
conséquence, une grande littérature sur les multiplicateurs budgétaires a émergé en considérant
des déterminants autres que le cycle économique, à savoir, les régimes de change, les politiques
monétaires, le degré d’ouverture, etc. Ceci a conduit à des résultats suscitant plusieurs controverses
entre économistes adoptant des méthodes différentes et/ou des échantillons de pays et des périodes
différents.
L’une des méthodes fréquemment sollicitées pour évaluer les multiplicateurs sont les modèles
VAR structurels (SVAR), également adoptés dans ce chapitre. Nous utilisons une panoplie de ces
modèles pour estimer les multiplicateurs de dépenses à l’aide de données trimestrielles, pour un
échantillon de 18 pays de l’OCDE. Nous avons contrôlé les effets du cycle économique et les effets
des mouvements de ratio de la dette publique pour déduire les multiplicateurs de dépenses dans
ces conditions.
Les résultats montrent que, en contrôlant les effets du cycle économique, les multiplicateurs de
dépenses sont beaucoup plus élevés en période de récession qu’en période d’expansion. Ces
résultats sont en conformité avec ce qui est observé dans la littérature récente sur les
multiplicateurs budgétaires, qui sont importants en période de récession et faibles, voire négatifs
en période d’expansion. Compte tenu de ces résultats, les politiques budgétaires devraient être
contracycliques. Ces résultats vont également à l’encontre de toute consolidation budgétaire
fondée sur une réduction des dépenses en période de récession, ce qui pourrait nuire à l’économie.
En outre, en contrôlant de manière exogène les mouvements de la dette publique, indépendamment

du cycle économique, il a été révélé que les multiplicateurs de dépenses sont plus importants en
période d’accumulation qu’en période de contraction de la dette. En plus, contrôlant conjointement
pour les mouvements de la dette publique et du cycle conjoncturel, les multiplicateurs sont plus
élevés sous l’accumulation de la dette dans les deux cas d’expansion et de récession. Cependant,
l’endogénéisation de la variable dette publique au PIB dans un SVAR conduit à des multiplicateurs
plus élevés en récession qu’en expansion.
De plus, un modèle plus élargi a été établi sur une longue période de données trimestrielles pour
les États-Unis (1966q1-2019q2), pays disposant de longues séries chronologiques trimestrielles
19
d’un ensemble de six variables fiscales et monétaires prises en compte dans ce modèle. L’objectif
de ce modèle était l’examen et l’explication des canaux de transmission des effets de la dette
publique et des dépenses publiques sur la croissance économique en relation avec d’autres
variables économiques en l’occurrence le taux d’intérêt, l’inflation et l’investissement privé. Les
estimations de ce modèle ont été produites en considérant la variation du cycle conjoncturel.
Les principaux résultats de ce modèle montrent que les dépenses publiques ont des effets positifs
mais de courte durée sur la croissance économique. En outre, la dette publique évince
l’investissement privé “crowding out”, entraînant une baisse du taux de croissance en période
d’expansion, tandis qu’en période de récession, les effets de la dette publique sur la croissance
sont positifs. Cet effet d’éviction peut jouer le rôle de relais sur les multiplicateurs de dépenses et
pourrait expliquer, toutes choses égales par ailleurs, la faible taille des multiplicateurs de dépenses
en expansion, alors qu’en période de récession, l’effet positif de la dette “crowding in” conduit à
des multiplicateurs plus élevés.
Dans toute la panoplie des modèles SVAR examinés dans le deuxième chapitre, ceux-ci montrent
clairement que pendant la période de récession, les effets sur les variables ont généralement
tendance à persister. La convergence vers la trajectoire de long terme après les chocs se produit
plus rapidement en période d’expansion qu’en période de récession. Cependant, malgré les
principales conclusions du deuxième chapitre, l’approche SVAR, exigeant la stationnarité des
variables comme condition nécessaire, est restreinte à l’évaluation notamment des effets de court
terme des variables de la dette et des dépenses publiques. Cependant, compte tenu des effets
théoriques de la dette publique à long terme, leur évaluation s’impose.
3.4. Contribution empirique du Chapitre 3 : Le ratio de dette publique potentiel à

cibler à long terme est une fonction endogène de la productivité potentielle du
capital public
En relation avec le sujet du seuil de dette publique discutée dans le chapitre 1, le troisième chapitre
étend l’examen des effets de long terme de la dette publique dans le cadre des modèles théoriques,
tenant compte des fondamentaux macroéconomiques propres à chaque pays. Par conséquent, un
modèle de croissance endogène a été appliqué au même échantillon étudié dans le chapitre 1
constitué de 20 économies développées. Dans ce modèle, une formule paramétrée a été simulée
pour une dette potentielle/limite qu’un pays pourrait cibler pour financer ses investissements
productifs sans pour autant la dépasser. Ce potentiel, lié à la croissance économique et à la
productivité du capital public ainsi qu’au taux d’intérêt, est dynamique, spécifique à chaque pays
et dépendant du temps. De plus, celui-ci a tendance à évoluer à l’opposé de la dette publique
réellement observée. Cela répond en particulier à la principale recommandation formulée dans le
premier chapitre selon laquelle le seuil de la dette publique est propre à chaque pays et à la période
et devrait être étudié compte tenu des fondamentaux économiques du pays. Il apporte notamment
une contribution qui se distingue des limitations des recherches mises en évidence dans le premier
20
chapitre, et qui fixent et testent en particulier des valeurs de seuils particulièrement exogènes et
atemporels.
De plus, les résultats du troisième chapitre montrent que la dette publique simulée est passée en
dessous du niveau observé de la dette en temps de crise pour de nombreux pays avancés,
notamment les plus touchés par la crise. Cela constitue un message clair de recommandation de
politique économique montrant que les pays sont de plus en plus à l’abri du danger de la dette
publique tant que la dette potentielle (simulée) reste supérieure à la dette publique observée. Les
résultats montrent que de nombreux pays sont soumis à une pression de la dette publique, en
particulier après la crise financière de 2008-2009. Pour certains pays comme l’Irlande, ce stress
est de courte durée, et la dette potentielle est rapidement redevenue plus élevée que la dette
observée à la suite du redressement de l’activité économique juste quelques années après la crise.
Cependant, pour de nombreux autres pays, les effets sont plus prolongés. La dette potentielle a
également révélé que des pays comme la Grèce, la Belgique et, dans une certaine mesure, l’Italie,
avaient des problèmes de dette accumulée au cours des périodes même avant la crise de 2008.
Empiriquement, la moyenne de la dette potentielle simulée en pourcentage du PIB sur la période

étudiée (1960-2015) est passée de valeurs plus élevées de 150% à 200% en Suisse, en Suède, au
Danemark, en Norvège, aux États-Unis, en Autriche et aux Pays-Bas, à des valeurs moyennes de
80% à 120% enregistrées en France, en Allemagne, en Italie, en Espagne, au Portugal, au
Royaume-Uni, en Australie et en Nouvelle-Zélande. Ces valeurs sont même réduites à environ
50% à 80% en temps de crise, notamment pour la Grèce (moins de 50%), le Royaume-Uni (60%),
l’Italie et l’Espagne (50%) la France et l’Allemagne (80%) et la Belgique (environ 70%).
Ce chapitre a comme valeur ajoutée la modélisation de la limite d’endettement publique comme

fonction endogène dépendante de la productivité du capital public, du taux d’intérêt de long terme
et de la croissance économique. La série des dépenses publiques potentielles montre une tendance
à la baisse dans le temps mais toujours plus élevée que celle observée dans de nombreux pays de
l’échantillon. Cette évolution est due à la tendance générale à la baisse de la productivité du stock
de capital public, en particulier pendant l’époque de la grande modération (1985-2015). Le
message de politique économique est que l’accroissement de la dette potentielle cible, requière
nécessairement l’amélioration de la productivité du capital public. D’où la nécessité de
sélectionner les types de dépenses publiques susceptibles d’améliorer cette productivité. Pour
reformuler la déclaration de Krugman (1994) sur l’importance de la productivité à long terme, “la
productivité n’est pas tout, mais à long terme, c’est presque tout”. L’amélioration de la
productivité pourrait être réalisable par le choix de dépenses publiques plus productives, ce qui
implique la sélection de projets ayant des multiplicateurs plus élevés.
21
4. Conclusion
Dans les trois chapitres, la thèse s’oppose à tout seuil d’endettement commun s’appliquant à tous
les pays. Le sujet de la croissance économique et de la dette publique reste néanmoins l’un des
sujets les plus débattus de la macroéconomie après la crise financière de 2008-2009. Cette
importance sera maintenue et davantage de recherches à ce sujet sont anticipées à moyen et long
terme pour de nombreux événements d’actualité. Effectivement, la crise sanitaire de 2020, en
raison de mesures d’urgence rigoureuses et de confinement de la population, conduit à
l’hibernation de presque tous les secteurs économiques pendant plusieurs mois. Par conséquent,
tous les experts et institutions de prospection prévoient une inévitable récession, et même une
dépression prolongée, dans presque toutes les économies mondiales. Toutes les mesures
susceptibles d’aggraver les déficits et d’augmenter le niveau de la dette publique, combinées à une
contraction du PIB, conduiraient à une flambée des ratios d’endettement public.
Au regard de ces mesures, certains économistes s’attendent à ce que ces ratios augmentent
d’environ 20 à 40 points de PIB pour de nombreux pays. Selon Kose et al. (2020), des
augmentations similaires ont été observées lors des crises précédentes. En effet, le ratio
d’endettement a flambé de 31 et 35 points de pourcentage respectivement pour l’Indonésie et la
Thaïlande lors de la crise financière de 1997, et de 27 et 38 points de pourcentage respectivement
pour la Lettonie et l’Irlande lors de la crise de 2008-2009. La reprise de la croissance économique,
incertaine, est conditionnée par la durée des mesures de confinement sujet à l’évolution de la
pandémie Covid-19 et au développement d’un vaccin contre celle-ci.
Dans ces conditions, la politique budgétaire, longtemps subordonnée à la politique monétaire dans
la grande modération (1986-2007), est désormais activement sollicitée. La crise sanitaire actuelle
conduit notamment à une plus grande implication de la politique budgétaire, surtout que la
politique monétaire conventionnelle a déjà atteint ses limites, en particulier dans de nombreuses
économies avancées où les taux d’intérêt nominaux atteignent et franchissent pour certains la borne
inférieure zéro. Bien que certains points de vue considèrent que la faiblesse des taux d’intérêt
réduit le coût de la dette publique, ce qui peut amener à une trajectoire de la dette publique
soutenable, la menace du redressement des taux d’intérêt en réponse à une inflation future
anticipée, conjuguée à une croissance morose, pourrait prolonger les pays endettés dans des vagues
de crise de la dette souveraine.
Par conséquent, les recherches sur la politique budgétaire et les problèmes de l’endettement et la
croissance économique resteront désormais très actives. De nombreuses questions méritent
notamment d’être soulevées au sujet de la dette publique et la croissance économique au cours de
la crise économique actuelle à la suite des retombées économiques et sociales de la crise sanitaire
de 2020. Il s’agit notamment d’étudier de nombreux objectifs, notamment conflictuels, de la
politique budgétaire visant à sauver des emplois ainsi que des entreprises et secteurs économiques
en difficultés d’une part en particulier à court terme, et relancer la croissance économique et ainsi
22
restaurer les recettes publiques, d’autre part à moyen terme. Ces deux objectifs sont inséparables
de l’objectif de soutenir la croissance à long terme tout en considérant les effets ultérieurs de
l’augmentation de la dette publique. En outre, la nature de la crise économique actuelle est
susceptible de soulever de nombreux défis dans les politiques économiques. En particulier, cette
crise est caractérisée par la coexistence des chocs négatifs d’offre et de demande, liés notamment
aux mesures de confinement des populations et restrictions de mobilité internationale. Notamment,
du côté offre, ceci a conduit à limiter les capacités de productions et d’exportations des entreprises
et altérer les chaînes de valeurs globales pour certains secteurs. Du côté demande, les mesures du
confinement ont limité la demande des ménages notamment en termes des services de voyages,
transports, restaurations, ainsi que la demande en produits d’importations. Cette dualité du choc
offre-demande pourrait avoir un impact important sur le cadre macroéconomique et
particulièrement sur le cadre de la politique budgétaire. Il s’ensuit que la modélisation des effets
de la politique budgétaire pourrait sortir du cadre traditionnel, notamment que cette crise ne semble
pas ressembler aux crises économiques internationales vécues précédemment (de la grande
dépression à la crise des subprimes). De plus, de nombreux défis nécessitant la coordination des
politiques monétaires et budgétaires peuvent être à l’ordre du jour, dont les effets devront être
évalués, en particulier dans la zone euro.
23
General Introduction
Public Debt and Economic Growth: A New Assessment
To finance their spending, governments generally have three options: first, they can raise taxes,
second, they can create money and, third, they can borrow from domestic and/or foreign markets.
The first option is politically not desirable and can even create economic downturns, harming
economic growth and employment, especially for countries that have high tax burdens, which is
the case for all developed countries. It also takes time to obtain a political consensus in a
democratic system to raise taxes. The second option, the creation of money, whenever it is
possible, can also cause significant damage to the economy by creating inflation and hence
decreasing households’ purchasing power. Furthermore, as the central banks in advanced countries
are determined to gain and maintain their credibility vis-à-vis the private agents, the money-
creation option is likely to alter such credibility. The third option, namely, raising funds through
borrowing, remains the quick option for financing public deficit needs, particularly short-term
needs. The choice of borrowing to finance public deficits also has its disadvantages, with doubtful
negative effects for economic growth, especially when the borrowed money is destined for
expenditure that is not necessarily pro-growth, and the accumulated public debt is higher. The
capacity of a country to pay its debt is generally assessed by the level of its total debt reported to
its national income (GDP).
Over the last decade, public debt has increasingly become a worrying issue for many advanced
countries, especially in the euro area. These countries implemented massive fiscal stimuli in 2009
in response to the severe downturns of the 2008 financial crisis. In around 2011 countries started
to reverse the course from fiscal expansion to fiscal consolidation to reduce deficit and debt ratios,
which were exacerbated by those fiscal stimuli with slow recovery and long-lasting recession. The
fear of a European sovereign debt default by many periphery euro area countries urged them to
turn to fiscal consolidation and austerity measures in Europe. This was also fuelled by the triggered
debate about the presumable existence of a public debt threshold hampering economic growth
(Caner et al., 2010; Kumar and Woo, 2010; Reinhart and Rogoff, 2010; Lin, 2014; Bentour, 2018).
Following the recession and the fiscal stimuli programmes, almost all the euro area countries have
overcome the rule of a 3% and 60% threshold stated in the European Monetary Union (EMU)
Stability and Growth Pact (SGP). Consequently, rigorous austerity measures have been
implemented to restore confidence, especially across the periphery of the euro area, whereas
recessionary effects were more pronounced, and the economic growth rate was still negative. The
negative short-term spillovers undermined the incomplete path of recovery, meaning the euro
faced a dilemma of boosting economic growth while reducing public debt and deficit. The trade-
off between boosting economic growth by fiscal stimuli to combat negative effects on social
welfare and employment, and reducing public debt and deficit through programmes of fiscal
adjustment (austerity, consolidation), depends on the size of the fiscal multipliers in each action.
1. Historical facts of public debt and growth in advanced economies
The history of public debt as a percentage of GDP in advanced economies shows that the period
following the 2008 financial crisis has been, in terms of scale, approaching the period of the 1929
Great Depression and the devastating World War II era, where public debt ratios were abnormally
high (Figure 1). What makes the difference between the two periods is that, in the recent period,
public debt accumulated in peaceful periods of economic expansion and prosperity, which was
worsened by the 2008 economic crisis, while the two aforementioned events (the 1929 Great
Depression and World War II) were clearly the principal factors contributing to accumulating high
public debt to GDP shares in that period.
Following the Great Depression and World War II period (1926–45), the public debt of advanced
economies (the G20 advanced)§ dropped down from a high average approaching 120% to very
moderate levels ranging between 20% and 40%. This stable average was sustained for a period of
around thirty years, labelled “Les Trentes Glorieuses” by the French economist Jean Fourastié.
This prosperous period was characterized by strong economic growth under the stable Bretton
Woods system of exchange rates. However, debt ratios started to rise, albeit not very high, from
the mid-1970s, with the end of this system, but also as a result of the two 1970s oil price shocks.
This period especially put pressure on the public finances of developing and low-income countries,
for which the debt to GDP increased from an average of 24% in 1976 to 100% in 1994. This period
is known as the developing countries’ public debt crises.
Moreover, following a period of low interest rates and financial liberalization that started after the
1989 “Washington consensus”, which contributed to lowering the cost of borrowing and
encouraged access to the international financial market, contributing to appeasing the public debt
of low-income and developing countries, the public debt to GDP of advanced countries once again
started a sustained increase from an average of approximately 40% to more than 90% in 2015. The
latter level of public debt ratio is considered by Reinhart and Rogoff (2010) to be an upper
limit/threshold that, once passed, may reduce economic growth in advanced countries. This
statement triggered an influx of empirical research studying the relationship between public debt
ratios and economic growth, fundamentally based, for the majority, on these two variables alone.
§
The G20 advanced economies here should be distinguished from the G20 group of countries set in 1999 and
becoming active after the 2008 economic crisis. The latter contains the big emerging countries (China, India, Brazil,
etc.) and the European Union countries block, while the former is only for the advanced OECD countries that have an
available long history of public data, as reported in the IMF databases (Abbas et al., 2010; Mbaye et al., 2018).
25
Figure 1: Public debt as a percentage of GDP over the period 1926–2015
140
120
World War II
100 and Great
Depression Developping
80 countries Financial
public debt globalisation
60 crises and the 2008
The "Trentes great
40 Glorieuses" period recession
20
G-20 Advanced G-20 Emerging Low Income
Source: Author’s own construction using the IMF database (Mbaye et al., 2018).
The previously analysed debt to GDP ratio, representing the burden of public debt of a country
normalized by its output, increases when the numerator (the debt in level) grows faster than the
denominator (the GDP). From an accounting perspective, the growth rate of the debt in level is the
nominal interest rate (the cost of borrowing), the latter is compared to the nominal growth rate of
GDP (or to the real growth rate once deflated) from the famous sustainability public debt equation
involving the primary deficit/surplus. Figure 2 shows an analysis of the average real growth rate
over the same long period, as was done previously for the debt to GDP ratio, for the G20 advanced
countries. We notice that the growth rate is weak for the two decades of the Great Depression and
World War II (1926–45), averaging 0.4%. On the contrary, the “Trente Glorieuses” period
recorded a high average of 4.3%, followed by a declining growth rate performance in the third and
fourth periods of 1976–95 and 1996–2015, respectively, to an average of 2.2% and 1.4%.
Apparently, this analysis shows that public debt accumulation in advanced economies is linked to
the growth rate performances in these countries. This fact is endorsed by putting together GDP
growth rate averages by decades and the public debt to GDP ratio in Figure 3. The latter shows
that economic growth following the post-war period is a decreasing path from a high level of 5.4%
in the decade of 1946–55 until reaching a low level of 0.5% in 2006–15. On the other hand, the
public debt to GDP average over decades has increased, from 30.8% in the 1970s to 79% in the
last decade. However, this analysis could not show, at high levels, which variable may cause the
other, which has been highly debated in the recent literature: high public debt levels may cause
26
economic growth to slow down (Reinhart and Rogoff, 2010), especially in a time of peaceful
economic expansion, as is the case for the period 1995–2007.
Figure 2: G20 Real GDP Growth Rate during 1926-2015 (%, Average)
12.0
10.0
8.0
Average
6.0 1946-1975 ≈ 4.3%
Average
4.0 1976-1995 ≈2.2%
Average
2.0 1996-2015≈1.4%
0.0
1926
1929
1932
1935
1938
1941
1944
1947
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
2007
2010
2013
-2.0 Average 1926-
1945 ≈0.4%
-4.0
-6.0
-8.0
-10.0
-12.0
Source: Author’s own construction, calculating averages of GDP growth rates from Maddison database
(https://www.rug.nl/ggdc/historicaldevelopment/maddison/).
Figure 3: Debt to GDP ratios and Real GDP Growth rate over decades for G20 advanced countries
-1 0 1 2 3 4 5 6
2006-2015 0.5 78.8
1996-2005 2.3 61.3
1986-1995 2.1 44.0
1976-1985 40.1 2.3
1966-1975 30.8 3.7
1956-1965 30.3 3.8
1946-1955 59.2 5.4
1936-1945 -0.1 88.8
1926-1935 0.9 73.3
0 10 20 30 40 50 60 70 80 90 100
Debt to GDP (%, Bottom scale) Real GDP growth rate (%, Top scale)
Source: Author’s own construction calculating period averages from the IMF database (Mbaye et al., 2018) for the
public debt to GDP shares, and from the Maddison database for the real GDP growth rate
(https://www.rug.nl/ggdc/historicaldevelopment/maddison/).
27
2. Debates about public debt effects on economic growth

2.1. The theoretical debate on the public debt effects
In theory, the effects of debt on economic growth were debated particularly in the post-war period
following the surge in public debt of advanced countries. The literature distinguishes between the
Keynesian short-term view, which assumes that government expenditure financed by public debt
is more likely to stimulate the aggregate demand in the context of Keynesian prices and wage
rigidities. However, according to the classical vision, which tends to focus on the long-term effects,
government debt is likely to reduce capital stock and lower productivity, and then it reduces the
output. Diverse channels are cited as the main reasons for these effects, especially as summarized
by Hansen (1959). Higher public debt can trigger higher private saving (the Pigou effect), fewer
incentives to work and invest, especially for owners of government bonds (the Kaldor effect),
negative incentive effect of the additional taxes needed to finance the payment of interest and
higher interest rates that are unfavourable to countering the inflationary impact of the “Pigou
effect”. Furthermore, Modigliani (1961) stated that public debt can crowd out private investment
by reducing credit to the economy or by raising long-term interest rates on public borrowing. Both
Keynesian and classical views are summarized in what is known as the “conventional analysis” of
the effects of government debt, which reflects the dominant paradigm among economists and
policy-makers (Elmendorf and Mankiw, 1999).
Another important theoretical contribution that fuels the debate about the public debt financing
effects is the “Ricardian equivalence”. The idea, originally attributed to Ricardo, states that
economic growth can be insensitive to public debt, as consumers, assumed to be forward-looking
(with rational expectations) can react by reducing their expenditure following a public expenditure
increase by an equivalent amount. The reason is that they expect future taxes to finance the deficit
generated by the new public expenditure. The equivalence is implied, as any public expenditure
reducing public saving is assumed to be offset by an equivalent increase in private saving. This
leads to unchanged total savings, and hence no effect is expected on other macroeconomic
variables. Similarly, Barro (1974) shows that government bonds constitute an asset to their owners
and a liability to taxpayers. Overall, no net wealth will be created, and the effect is null. The ideas
of Ricardo-Barro effects of debt somehow invoke the redistribution of such effects across
generations. They are especially discussed in the scope of intergenerational transfers and have
been analysed in the first models of overlapping generations (Diamond, 1965; Blanchard, 1985).
The debate about the short-term analysis between economists (Keynesians, new Keynesians,
classical and new classical) focuses on the composition of public spending, for which classical and
neoclassical economists consider that capital expenditure, but not necessarily current expenditure,
may have positive effects. This view is contrasted by Keynesians arguing that the effects of
spending are always positive, with differences in terms of their efficiency, which may be higher
for capital expenditure than other types of expenditure. These opposing views are gathered by
28
Aschauer (2000), who assumes that the short-term relationship between debt and growth could be
positive as a result of the impact of financed capital expenditure on growth, while the long-term
relationship could be reversed because of the high-risk premium following high public debt,
raising the cost of debt. For Blanchard (2006), the negative association between debt and growth
could be attributed to delayed counter-cyclical tax policies, which means that countries in
recession are adopting expansionary fiscal policies to boost growth and, in times of expansion,
restrictive policies to reduce debt levels. The relationship between debt and growth was debated
further following the 2008 financial crisis in the wake of much recent empirical research.
2.2. The empirical debate on the public debt threshold effects on economic
growth
In advanced economies, the trend of public debt in the period of pre-crisis was exacerbated by the
fiscal stimulus implemented in the 2008–9 period of crisis. At that time, a controversial paper by
Reinhart and Rogoff (2010) reported that, for a sample of 20 advanced countries, there is a 90%
common threshold of the government debt to GDP ratio, over which debt has a negative effect on
economic growth. This triggered an influx of research, creating an empirical debate with many
controversial results about the existence of the public debt threshold, from which the debt–growth
relationship changed its sign.
The empirical debate on the effects of public debt on economic growth is of particular importance,
as its outcome has significant policy actions. In particular, if economic growth is revealed to be
reduced by high levels of public debt, as assumed by Reinhart and Rogoff (2010) and other
research supporting their results (Caner et al., 2010; Checherita and Rother, 2010; Kumar and
Woo, 2010; Lin, 2014), expansionary fiscal policies that may have positive effects in the short
term may reduce long-term economic growth, fully or partially offsetting the fiscal stimulus short-
term effects.
The debate about the public debt threshold has intensified, as several authors have contested the
Reinhart and Rogoff (2010) findings (Fergusson and Johnson, 2011; Minea and Parent, 2012;
Baglan and Yoldas, 2013; Pescatori et al., 2014; Egert, 2015), especially after the revelation of
some codification and calculation errors in the Reinhart and Rogoff (2010) article by Herndon et
al. (2013). The summary of such results could be gathered under three groups: a first group of
empirical research focusing mainly on the examination of a concave, non-linear relationship
between debt and growth (Checherita and Rother, 2010; Kumar and Woo, 2010; Pescatori et al.,
2014; Eberhardt and Presbitero, 2015; Chudik et al., 2017); a second group interested in the
causality between debt and growth (Panizza and Presbitero, 2012, 2013; Di Sanzo and Bella, 2015;
Gomez-Puig and Sosvilla-Rivero, 2015); and a third group of studies that examine the impacts of
other institutional and macroeconomic variables interfering with the relationship between debt and
growth (Greiner, 2011; Kourtellos et al., 2013; Pan and Wang, 2013; Sharpe, 2013; Marchionne
and Parekh, 2015). The results of all this research did not lead to any confirmed consensus.
29
Generally, assessments of the effects of fiscal policies are quantified by the fiscal multiplier values.
The distinction of public expenditure efficiency in Keynesian/classical analysis (as mentioned in
the theoretical debate) by type (current, capital, defence, social, etc.) is empirically linked to their
multipliers’ estimates. In this vein, another important strand of empirical research was triggered in
the aftermath of the financial crisis, revealing more about the sensitivity of such multipliers to
many economic and institutional determinants.
2.3. The debate on the fiscal multiplier value
Despite the fact that the previous debate about the impact of government debt on economic growth
considers the direct relationship between growth and public debt, the channels through which such
effects materialize are related to the fiscal multiplier effects, particularly in the short term. The
fiscal multiplier assessment has been extensively reconsidered following the development of
econometrics and statistics. Since the 2008 financial crisis, a significant amount of literature
assessing fiscal multipliers has emerged (Romer and Romer, 2011; Ramey, 2011, 2018; Auerbach
and Gorodnichenko, 2012, 2013; Delong and Summers, 2012; Farhi et al., 2017; Ramey and
Zubairy, 2018).
The main findings of this literature reveal key differences in the size of the multipliers. In fact, the
multiplier’s value showed sensitivity to a variety of determinants that were not (or rarely)
considered before the 2008 crisis. Fiscal multipliers are especially time- and country-specific, and
even sensitive to the assessment method (Batini et al., 2012; Baum et al., 2012). Furthermore, the
fiscal multipliers are found to be sensitive to the business cycle; in particular, fiscal spending
multipliers tend to be larger in recessions than in periods of expansion (Auerbach and
Gorodnichenko, 2012, 2013; Barro and Redlick, 2011; Parker, 2011; Corsetti et al., 2012;
Caggiano et al., 2015; Fazzari et al., 2015; Glocker et al., 2019). Other research has led to
ascertaining the vulnerability of fiscal multipliers to other determinants and not necessarily
conditioned by the state of the economy in the business cycle. These determinants are fiscal
position measured by the level of debt ratios and/or deficits (Corsetti et al., 2013; Huidrom et al.,
2016), the constrained monetary policy, either by the zero lower bound (ZLB) interest rate
(liquidity trap) or by the loss of monetary independence, as in a pegged exchange rate or a
monetary union (Hall, 2009; Cogan et al., 2010; Christiano et al., 2011; Delong and Summers,
2012; Farhi and Werning, 2017).
Nevertheless, the very recent works of Ramey (2018) and Ramey and Zubairy (2018) consider that
government spending multipliers are, on average, lower than unity. This contrasts with the
tendency of the post-2008 crisis research confirming larger multipliers, especially in recessions,
which could reach the value of 2. This is also in line with the same consensus on spending
multipliers before the 2008 recession, considered to be weak, and that fiscal policy effects are
short-lived (Coenen et al., 2012). In particular, fiscal policy effects, taking into account the fiscal
position of the economy, measured by the level of public debt and/or fiscal deficit, have been
30
highly debated in the aftermath of the 2008 financial crisis (Boussard et al., 2012; Corsetti et al.,
2012; Blot et al., 2014b; Canzoneri et al., 2015; Bi et al., 2016; Huidrom et al., 2016; Perdichizzi,
2017; Poghosyan, 2017; Auerbach and Gorodnichenko, 2017; Afonso and Leal, 2018; Blanchard,
2019; Broner et al., 2019). These contrasting results about spending multipliers still make it
interesting to reconsider studying fiscal multipliers and to contribute to this unsettled debate.
Very recently, in a presidential lecture of the American Economic Association, Blanchard (2019)
triggered another wave of public debt and growth debate related to the fiscal cost of high public
debt, as well as its effect on welfare. He particularly minimized worries about the public debt cost
for the American economy as, in historical records, the nominal interest rate has remained, on
average (except in some small periods around the 1980s), below the nominal growth rate.
2.4. The debate on the effects of public debt in economic models
To assess the effects of government debt, many theoretical growth models have been designed.
Ramsey’s (1928) model was the basis for the first class of the infinitely lived agents’ models.
Later, the debate between economists about agents’ lifetime horizon, and the type of
intergenerational operative transfer linkages between such agents, led to the concept of modes of
overlapping generations (families/dynasties) being built to account for the existence or absence of
continuity between generations provided by such linkages (Samuelson, 1958; Diamond, 1965).
The debate was focused on the implications of the altruism between generations and the lifetime
agents’ horizon in determining the existence of government debt effects on agents’ behaviour in
terms of saving, capital accumulation, consumers’ utility and interest rates. Important
contributions in this sense are those of, for example, Diamond (1965), Yaari (1965), Barro (1974),
Blanchard (1985), Buiter (1988), Aiyagari (1985, 1987), Weil (1989) and Ni (1999). The
redistributive effects across generations have been much debated in the post-war period. Buchanan
(1958) and Meade (1958), discussing the contributions of their predecessors, particularly Lerner
(1943), show that public debt does not constitute any burden for the present taxpayers. The burden
is instead transferred, partially or totally, to future generations that will have to pay taxes to service
the debt (Bowen et al., 1960; Modigliani, 1961). Furthermore, advocates of the Ricardian
equivalence show that debt neutrality could happen, depending particularly on the existence of
operative altruistic links (bequests) between generations (Barro, 1974). In this way, the debate
emerged, especially with useful contributions modelling fiscal policy insights, precisely the debt–
tax swap and their effects on welfare utility and interest rates. This also created an intense debate
between the neoclassicists and proponents of the Ricardian-equivalence conjecture (Barro, 1976;
Feldstein, 1976; Cukierman and Meltzer, 1989).
The public debt effects are also considered in the political economy of public debt. In this regard,
two approaches have been debated: the normative approach, where the government is a social
planner for whom the priority is to maximize the social welfare of its individuals (Barro, 1979;
Lucas and Stockey, 1983; Aiyagari et al., 2002); and the positive approach, which considers public
31
debt as a state variable used by each government as a strategy to influence its successor’s choices
or as a way to shape private economic agents’ expectations (Svensson and Persson, 1989; Alesina
and Tabellini, 1990). Hence, the most disseminated ideas were related to the effects of fiscal policy
under governments following committed rules versus discretionary policies. In particular, the
government time inconsistency has an impact on the way the economic agents form expectations
affecting their economic decisions (Kydland and Prescott, 1977).
The public debt effects are also examined in the class of endogenous growth models. The
development of such models was pioneered by Romer (1986), Lucas (1988) and Barro (1990).
These models were brought as an alternative to the neoclassical growth model of Solow (1956)
and Swan (1956), which was unable to explain the persistent per capita non-zero growth rates in
many developed economies and was therefore highlighted for missing the determinants of long-
term growth. Thus, the endogenous growth framework encompasses other determinants of long-
term growth, particularly in assessing fiscal policies, by incorporating the public sector into the
productive sector (Lucas, 1988; Romer, 1989; Barro, 1990; Barro and Sala-i-Martin, 1992, 1995;
Futagami, 1993; Jones, 1995, 2003; Corsetti and Roubini, 1996; Turnovsky, 1997; Greiner, 2007,
2012, 2016; Futagami et al., 2008; Maebayashi et al., 2017).
Recently, fiscal policy, and particularly government debt, was also modelled under the class of the
new Keynesian models. The pioneering works in the new Keynesian modelling framework are
principally those of Christiano et al. (2005) and Smets and Wouters (2007). The latter differ from
the models of the overlapping generations by assuming distortionary taxes instead of lump-sum
taxes, as particularly assumed in the overlapping generation models. Furthermore, they join the
literature of the political economy of debt by discussing government policies under commitment
or discretion rules. In particular, some authors argue that optimal public debt would follow a
random walk process whenever the government can achieve a time-inconsistent policy
commitment (Benigno and Woodford, 2003; Schmitt-Grohe and Uribe, 2004). The important
empirical research of these models has flourished recently (Leith and Wren-Lewis, 2013; Mayer
et al., 2013).
Despite important contributions of the previous classes of models in assessing the effects of
government debt and fiscal policy, the new Keynesian models have been subject, in particular, to
Mankiw’s criticism. Mankiw (2000) contributed by the “savers–spenders” theory, which has
influenced much of the empirical research on fiscal policy, trying to consider the behaviour
following the rule of thumb, especially in the New Keynesian models. In this way, Chari et al.
(2009) show that this class of models is not yet useful for public policy analysis. In particular, they
address their critics on the model of Smets and Wouters (2007), which constitutes the fundamental
reference for many recent contributions and policy-makers using the new Keynesian analysis
framework.
32
3. Interest of the thesis
Following the previous preliminary analysis showing the soaring levels of public debt in advanced
economies, as well as the debates through the theoretical and empirical literature on the public
debt effects on economic growth, the subject of public debt and economic growth linkages, which
constitute the core of this thesis, deserves to be reconsidered among academicians, as well as
policy-makers. Furthermore, the new reconsideration of fiscal policy effects assessments through
fiscal multipliers, which were revealed to be sensitive to the business cycle, and particularly to the
fiscal position represented by the public debt or deficit situations, as well as other economic and
institutional determinants, adds more enthusiasm to this subject and justifies the choice of thesis
subject.
Therefore, the current thesis builds on the very recent literature findings on the fiscal policy
debates, particularly on the public debt threshold effects and fiscal multipliers, while also paying
attention to the old literature on the public debt effects, which flourished in the post-war era (World
War II) and constituted the theoretical core of many public debt effects and economic growth
models. Moreover, it contributes empirically to the public debt and economic growth literature
using a variety of economic and econometric models across all three chapters of this thesis.
Given the high indebtedness of the advanced economies, our focus was on a sample of advanced
countries. It is constituted by 20 OECD advanced countries, of which 11 are euro area members,
namely, Australia, Austria, Belgium, Canada, Denmark, France, Germany, Greece, Ireland, Italy,
Japan, The Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United
Kingdom and the United States. The focus on this sample was justified for many reasons:
- The economic importance of this group of advanced countries influencing the world
economies, and particularly that they weigh on the international economic and financial system
(systemic risk).
- The high public debt to GDP accumulated for the majority of these countries, and many of
these countries have encountered sovereign debt problems, either in the current financial crisis
or previously.
- The sample is constituted by countries that have been highly studied and for which results have
been debated in the post-2008 financial crisis in terms of public debt and fiscal policy issues,
which allows us to turn our results into important findings in the literature.
- Finally, the availability of a long history of data, annually as well as quarterly, is also an
important reason for the choice of this sample.
33
4. Plan of the thesis
The first chapter of this thesis, entitled “Public debt and economic growth: one size does not
necessarily fit all”, highlights the highly-debated issue in the empirical literature about the
existence of a debt threshold inversing the relationship between public debt and economic growth
from a positive relationship for low and moderate ratios of public debt to a negative one for higher
public debt ratios. Our empirical application in the first chapter, applying panel polynomial
equation regression, as in Checherita and Rother (2010) and many other papers that build on it,
and a new method of kink regression by Hansen (2017), applied to individual countries, which
searches endogenously the public debt ratio threshold, concludes that the turning point in the
relationship between public debt and growth is rather a country- and time-specific one. Following
these findings, we recommend in the first chapter studying the relationship between public debt
and economic growth, taking into account countries’ macroeconomic fundamentals, which should
be captured by an appropriate theoretical modelling framework rather than a simple econometric
relationship implying two main economic variables (public debt ratio and economic growth), as
done by most of the empirical literature following the Reinhart and Rogoff (2010) paper.
As public debt is destined to finance public expenditure, we assess, especially in the short term,
the effects and relationship of public debt on growth through the impact of government expenditure
on the output. This impact is known as Keynesian multipliers. Therefore, the second chapter,
“Short-term effects of public debt on growth: the spending multiplier pass-through”, assesses the
short-term effects of public debt, particularly through fiscal multipliers. As the body of literature
studying the effects of (optimal) fiscal policy is substantial in size, our focus in this chapter, in the
literature review and in our application, is dedicated to the fiscal policy effects through spending
multipliers (no tax/income multipliers are considered).
Applying the methodology of a structural vector autoregressive (SVAR) model for each country,
the results show that the expenditure multipliers are much higher in times of recession than in
times of expansion. Moreover, it is observed that, generally, the idea of spending multipliers being
weak, and even negative in recent decades compared to the 1960s, 1970s and 1980s, is not well
supported by our findings. The previous results are in line with what is observed in the recent
literature about fiscal multipliers, in advanced economies, being large in times of recession and
weak or even negative in times of expansion.
Furthermore, introducing endogenously the public debt to GDP variable in an SVAR model leads
to higher multipliers in recession than expansion for each country of the sample. These results
were checked by a large SVAR model conducted on a long history of quarterly data for the United
States, as the country with long quarterly times series of six fiscal and monetary variables
available, as considered in this model, namely, public debt, GDP, private investment, public
expenditure, interest rates and inflation. The main results of this model show that government
expenditure has positive but short-lived effects on economic growth. Furthermore, public debt
34
crowds out private investment, leading to a decreasing growth rate in times of expansion, while in
times of recession the public debt effects on growth are positive. This crowding-out effect may
play pass-through to the expenditure multipliers and could explain, ceteris paribus, the weak size
of spending multipliers in times of recession, while the crowding-in effect leads to higher
multipliers. Consequently, this chapter concludes on the fiscal policy effects and spending
multipliers and contributes to the current economic literature.
The previous chapter considers a short-term approach, while public debt also has prolonged long-
term effects, especially in the classical models. Therefore, the third chapter, entitled “Public debt
effects in theory-based models with an empirical assessment of the potential public debt”,
undertakes the essence of the relationship between public debt and economic growth in all
theoretical models, citing the over-lapping generations (OLG) models, the infinitely lived agents
(ILA) models, the endogenous growth models and the new Keynesian models. For several
considerations that were justified in this chapter, the endogenous growth model was chosen as a
framework for our application to assess the long-term effects of public debt on economic growth.
The main findings of this chapter are the ability to derive, for each country, an optimal potential
endogenous government debt as a time series derived from potential public investment (the
government gross fixed capital formation). This parametrized public debt, drawn from an
economic model proper to each country, based on a Cobb-Douglas production function augmented
by public and human capital, jointly with a public budget constraint equation, is an important result
against vehiculating any exogenous threshold, therefore endorsing the conjecture of the first
chapter that there is no specific threshold that fits all. Furthermore, for many advanced countries
with high actual public debt, potential debt is overcome by actual debt in many periods, especially
in times of crisis. Consequently, the third chapter closes with the effects of public debt and
contributes to the debate about fiscal policy in the long term.
Figure 4 presents a graphical structure of the thesis, illustrating its main parts and describing its
steps as well as the important findings and consequent recommendations.
35
Figure 4: Thesis Structure
Key Literature Review General Introduction Objectives, Results and

Recommendations
Chapter 1 Aim
Public debt and economic Assess empirically public
- Review of Theoretical debt-growth thresholds
effects of public debt on growth: one size does not
economic growth necessarily fit all
- Survey of empirical results - Sample: 20 OECD countries
of debt-growth thresholds - Period: Annual data from Results
studies following Reinhart 1881 to 2012 Threshold is country and
and Rogoff (2010) time specific
- Methods: Panel and
individual kink regression
(Hansen, 2017)
Recommendations
Assess short run debt effects
by country
Chapter 2 Aim
- Survey of empirical studies Short-term effects of Assess short run effects of
on fiscal multipliers public debt on growth: public debt via fiscal
- Discussion of the debt the spending multiplier multipliers
effects on fiscal multipliers pass-through
- Assessing debt effects on
-Sample: 18 OECD countries Results
economic growth through
Confirmed sensitivity of
spending multipliers, - Period: Quarterly data from spending multipliers to the
business cycles, etc. 1966:1 to 2019:2 business cycle
- Methods: SVAR models
Recommendations
Assess the public debt
effects in the long run
Chapter 3: Public debt
- Survey of theoretical public effects in theory-based
debt models: ILA, OLG, models with an empirical Aim
endogenous growth and new assessment of the potential Determine optimal debt limit
Keynesian models public debt endogenously in a long-term
- Normative versus positive framework
- Sample: 20 OECD
approach of the public debt countries
effects Results
- Period: Annual data, 1960-
- Theory of saving-spending For each country, optimal
(Mankiw, 2000) 2015
public debt is a time series
- Method: endogenous derived from potential public
growth model investment
General Conclusion
and Perspectives
36
Chapter 1
On the public debt and growth threshold:

One size does not necessarily fit all
Chapter 1. On the public debt and growth threshold: One size does not necessarily fit all
Chapter 1
On the public debt and growth threshold: One size does not
necessarily fit all
Abstract: In a time of high debt and sluggish economic growth, the Reinhart and Rogoff (2010)
conjecture of a common 90% debt threshold for advanced economies triggered a controversial
debate among economists and policy-makers. We analyze the accuracy of this result for a sample
of 20 advanced economies over the period of 1880-2010. Using a regression kink model with an
unknown threshold proposed by Hansen (2017), we examine the relationship between public debt
and economic growth. We show that the relationship between public debt and economic growth is
time-varying and state-dependent and subject to data and country heterogeneities. The relationship
is instable either by country, by group of countries or across periods of time and particularly
sensitive to country size, government effectiveness and government expenditures. The kink
regression method shows diverse curves for the debt-growth relationship. For a set of countries,
growth slows starting from low debt levels over the postwar period. However, other countries start
flourishing from low to medium levels of debt, while some countries show flat curves in the debt-
growth relationship, especially over the long period of 1881-2010. These findings reject the
existence of any common threshold fitting all countries and call for more theory-based models that
take into account fundamentals that vary between countries and impact debt–growth interactions.
JEL classification: C13, C15, C23, H63, O57.
Keywords: Public debt, Economic growth, Regression kink, Non-linearity, Threshold effects.
1. Introduction
In 2010, the signs of a sovereign debt crisis in Europe and other advanced economies divided
economists and policy-makers regarding the efficiency of the economic policies necessary to
overcome the recession. Some are in favor of continuing the stimulus packages implemented at
the beginning of the crisis, while others call for urgent fiscal consolidation and austerity policies
to reduce the public deficit and debt levels. The former, believing in the role of Keynesian
multipliers, make growth a priority to stabilize the deficit and debt ratios. The latter hold that high
levels of debt hamper growth, which can be explained by a negative causality running from debt
to economic growth. In both cases, the relationship between public debt and economic growth is
at the forefront implying many controversies.
Thus, in what is an extremely controversial subject, Reinhart and Rogoff (2010) (RR (2010)
hereafter) reported that, for a sample of 20 advanced countries, there is a 90% common threshold
of the government debt-to-GDP ratio over which debt has a negative effect on economic growth.
Several authors5 have contested this result, especially after the revelation of some codification and
calculation errors in RR’s (2010) article.6 This resulted in an influx of research using different
methods to study a set of econometric properties likely to alter the link between economic growth
and public debt.
The trend among emerging empirical literature has been the examination of a concave non-linear
relationship between debt and growth, with some focusing on a variety of econometric issues, such
as endogeneity, causality, and heterogeneity rather than using a proper theoretical modeling
framework. Despite the use of a variety of econometric tools to overcome such issues, no
consensus has been found on the robust existence of a single threshold that fit all and at which debt
starts to alter growth. Moreover, despite most researchers agreeing on the negative correlation
between high public debt and economic growth, it is difficult to agree on the direction of causality
between debt and growth in the long term as suggested by the economic theory. Many economists
warn against the hasty interpretation of this research and call for more investigation on this subject
(Panizza and Presbitero, 2012; Minea and Parent, 2012). According to Minea and Parent (2012),
institutions such as the OECD, the EU Commission and the French Report on Public Finance of
April 2010 have a tendency to be influenced by the 90% threshold of RR (2010) in acting to reduce
their public debt using austerity measures. The authors warn, in particular, against drawing
conclusions based on considering exogenous thresholds instead of searching endogenously for
such thresholds as the relationship between public debt and growth, which could be altered by
complex non-linearities and sudden jumps around high debt thresholds. Their findings also suggest
5
See, for example, Ferguson and Johnson (2011), Herndon et al. (2013), Baglan and Yoldas (2013), Pescatori et al.
(2014), Egert (2015), Eberhardt and Presbitero (2015), and Chudik et al. (2017).
6
See Herndon et al. (2013).
39
that there may be a regime in which very high public debt is positively correlated with economic
growth, similarly to what is observed with low and moderate levels of public debt.
In theory, the effects of debt on economic growth are summarized by the "conventional analysis”
which reflects the dominant views among economists and policymakers (Elmendorf and Mankiw,
1999). This approach adopts the deficit budget Keynesian view in the short term, assuming that a
deficit financed by government bonds boosts economic activity. Government expenditures act as
a stimulus for the aggregate demand in the context of Keynesian prices and wages rigidities.
Nevertheless, for the short term, the debate is rather on the composition of public spending. The
debt intended to finance the capital goods, would have positive short-term effects, which may
become negative in the long run due to the induced risk premium following high debt (Aschauer,
2000). Keynesians report that the spending effect is positive for all expenditure types, and only
differs in efficiency.
In the long term, the economy adheres to the classical vision, for which government debt reduces
capital stock and lower productivity, hence, reducing the output. This goes through diverse
channels as Hansen (1959) reported: higher debt can trigger higher private saving, less incentives
to work and invest especially for the owners of the government bonds and negative incentive effect
due to additional taxes needed to pay the debt service. Likewise, public debt can crowd out private
investment by reducing credit to the economy or by raising long-term interest rates (Modigliani,
1961).
However, according to Barro (1974), economic growth can be insensitive to public debt. Under
the assumption of perfect information, which assumes rational expectations, economic agents
expect future taxes to finance the deficit generated by the new public spending and thus reduce
their expenditures. This effect is known as Ricardian equivalence7: any public expenditure
reducing public saving is assumed offset by an increase in private saving by an equivalent amount.
Consequently, the national saving is unchanged and no effect is expected on other economic
variables.
If theory suggests mainly a causality running from debt to economic growth, few papers examining
empirically this issue have not reached any consensus about the direction of the causality (Panizza
and Presbitero, 2012; Gomez-Puig and Sosvilla-Rivero, 2015 and Di Sanzo and Bella, 2015). The
relationship between public debt and economic growth could also be eclipsed by the interference
of other economic and institutional variables. Some authors criticized previous empirical works
for the omission of such institutional variables in the debt growth analysis (Panizza, 2015;
Kourtellos et al., 2013). Nevertheless, few researches tried to enrich the debt growth relationship
7
The name Ricardian equivalence is due to Buchanan (1976) who found a similarity between the proposal of Barro
(1974) and that of David Ricardo. Moreover, O'Driscoll (1977) notes that Barro's (1974) proposition contradicts
Ricardo's conclusions that there is no equivalence of choice between financing a war by taxation or a debt and decides
to call it “Ricardian non-equivalence”. Barro's (1974) proposal is also referred to as modern Ricardian equivalence
theorem (Ahiakpor, 2013).
40
by other macroeconomic and institutional variables (Sharpe, 2013, Pan and Wang, 2013; Greiner,
2011; Marchionne and Parekh, 2015).
Given the absence of consensus on a tolerable level of debt, questions still need to be asked: Does
public debt that is too high reduce economic growth? Does the turning point in the relationship
between growth and debt exist for all countries and at all times? What is its size? Does it fit all
countries or is it a country-specific one?
The purpose of this article is to investigate the existence of the debt threshold effect, its size and
whether it fits all countries or varies across countries and periods. The previous empirical studies
have several limitations: some, using a long period of analysis as in RR (2010) suffer from
methodological issues. They generally adopt simple descriptive statistical approaches to generalize
for a common threshold that fits all. While those using different econometric tools could have a
short time sample bias. Generally, their samples start after the 1970s. The main papers set
exogenous thresholds to test. Furthermore, countries differ in their economic policies affecting the
debt–growth relationship, and the panel approach adopted by all the previous empirical literature
suffers from a high heterogeneity bias.
This paper adds to the existing literature by adopting a different approach. Unlike previous
research that has examined the debt-growth thresholds across panels of countries, our methodology
gives priority to country specific analysis. Surveying the previous empirical approaches and
starting from their limitations, we use a novel econometric method proposed by Hansen (2017)
that searches thresholds endogenously for individual countries. We undertake estimations using
long time period of 1880-2010 and sub-periods depending on the World major economic and
political events. We also run panel regressions by varying countries sample according to some
sizeable countries, exchange rate regime and type of government expenditures and effectiveness.
The remainder of the paper is organized as follows. Section 2 reviews the existing literature.
Section 3 describes the econometric methodology. Section 4 describes the data and descriptive
statistics. Section 5 provides estimation results for country specific regressions. Section 6 presents
estimation results for panel analysis. Section 7 concludes.
2. Literature review
The issue of the government debt threshold has been extensively studied since the 2010 debt crisis,
provoking several controversies. Initiated by the early work of RR (2010), researchers have
examined public debt thresholds for different panels of countries while correcting for econometric
issues. Table 1 summarizes the main contributions.
Insert Table 1 about here
41
RR (2010) found that the correlation between public debt and growth is low for normal debt levels
and becomes strong and negative when the debt-to-GDP ratio exceeds 90%. They observe that
median and average growth rates corresponding to debt ratios over this threshold shrink by 1 and
4 points respectively. This result, based on a simple descriptive statistical approach, has given rise
to much empirical research examining the relationship between debt and growth using increasingly
econometric methods.
In a subsequent paper, Reinhart et al. (2012) emphasized their previous findings of the 90%
threshold by analyzing periods of public debt overhangs for a sample of 22 advanced economies
going back to the nineteenth century.8 They defined a debt overhang period as a debt-to-GDP ratio
of more than 90% lasting for at least five consecutive years. As a result, 26 periods were detected
and 23 of these are associated with lower growth. On average, an annual growth lower by 1.2
percentage point than in periods of debt ratios less than 90%. However, from the 22 advanced
countries, only 13 have episodes of debt overhang from which two countries (Italy and Greece)
have both 8 periods of debt overhang (4 each). Therefore, the sample of countries, with different
economic policy experiences, used to emphasize the 90% common rule is reduced. Furthermore,
almost all episodes of high debt resulted from costly wars and the Great Depression and, only six
countries have had a debt overhang in peacetime: Belgium, Canada, Greece, Ireland, Italy and
Japan. So, should this small sample of heterogeneous countries in size, time sample and monetary
sovereignty be sufficient for concluding about the common 90% threshold?
In fact, the conjecture of a common threshold does not seem to be accepted by many authors. For
example, Ferguson and Johnson (2011) stated that RR (2009, 2010) “jumble big and small
countries together from different areas and different political choices. This makes induced policy
lessons from such samples a likely misleading exercise”. They argue that “political choices for
smaller countries are frequently influenced by external factors, while big countries like United
States and Japan are the principal players in the international system of which policies affect the
rest of the world”. Consequently, the authors opposed the idea of a common threshold debt arguing
using historical counter-examples. Especially, the United Kingdom constitutes an interesting fact
countering Reinhart and Rogoff claim showing that the British industrial revolution flourished
while debt-to-GDP ratio exceeded 200% for decades.
Empirically, an important controversy came from Herndon et al. (2013). By replicating the
exercise of RR (2010), these authors corrected some data processing errors. Consequently, the
average growth rate of countries with government debt-to-GDP ratios of more than 90% is not
dramatically different from that calculated for countries with moderate to high debt-to-GDP ratios.
This paper triggered an influx of research examining debt thresholds using different econometric
tools (Table 1).
8
The time sample differs between countries depending on data availability. The United States have the largest dataset
(1791-2011) while Ireland has the shortest sample (1924-2011).
42
A significant amount of research has replicated the same sample of 20 advanced countries used by
RR (2010), finding different results that generally point to low levels of debt thresholds. Baglan
and Yoldas (2013) used Bayesian inference and found a debt threshold of 20%. Egert (2015) used
the Hansen (1999) method and found a debt threshold of 20 to 30%. Lee et al. (2017) examining
the relationship between public debt and median GDP growth suggest that the debt threshold may
exist around 30%. Surprisingly, Minea and Parent (2012), using panel smooth threshold regression
(PSTR) models found a convex relationship between debt and growth where the effect is negative
below a high debt threshold of 115% and positive above this threshold. However, these results are
surrounded by more uncertainties and may result in dangerous consequences when translated into
policy decisions.
Other researchers have used different samples and periods to study the long-term effects of debt
on economic growth. For example, Pescatori et al. (2014) analyzed debt and growth data over long
period, considering lead economic growth by 1, 5, 10 and 15 years to be affected by the current
debt. Their analysis rejects any threshold from which economic growth is undermined. However,
they found that high debt increases output volatility. They also found that countries with high but
declining debt grow as fast as countries with lower growth. Eberhardt and Presbitero (2015)
studied nonlinearity by correcting the heterogeneity of the debt-growth relationship across
countries. Their results highlighted the negative non-robust relationship between public debt and
long-term economic growth, but failed to determine a common debt threshold for all countries.
Chudik et al. (2017) developed tests for threshold effects in the context of dynamic heterogeneous
panel data models and found no evidence of a universally applicable threshold effect. Regardless
of the threshold, they found significant negative long-run effects of public debt build-up on output
growth. Furthermore, Syssoyeva-Masson and De Sousa Andrade (2017) highlight the long
memory of public debt series and recommend studying the debt–growth relationship in a long-
term framework.
Nevertheless, other researchers seem to support RR’s (2010) findings, albeit not necessarily the
90% debt threshold. Caner et al. (2010) were the first to review the results of RR (2010) and
confirm the negative link between public debt and economic growth above the 90% threshold.
Kumar and Woo (2010) also highlighted a negative non-linear relationship for higher debt levels
for a sample of advanced and emerging countries over the period of 1970-2007. Particularly, the
per capita growth is 0.2% lower following a 10-percentage point increase in the debt ratio. Lin
(2014) applied a threshold quantile Lasso regression to a sample of 62 cross-sections combining
developing and developed countries over the period 1991-2005, and to individual countries for a
subsample of data of 22 countries over the period 1961-2010. For cross-country analysis, he
confirmed the 90% threshold for the median quintile as defended by RR (2010). However, for
country-specific analysis, he demonstrated that tipping points range widely between 10% and
100% across countries. The widespread values of the threshold were emphasized in particular
when controlling for macroeconomic and demographic changes. The existence and value of these
thresholds increase by quintile, reflecting the asymmetric effects of debt on growth, and are more
43
common in developing than developed countries. Checherita and Rother (2010) found a significant
polynomial term between public debt and per capita economic growth considering a sample of 12
eurozone countries over the period of 1970–2009. Unlike research that has pointed to a defined
threshold, their paper indicates an interval (90–100%) from which public debt starts to hamper
economic growth.
3. Econometric Methodology
Unlike the previous research that essentially examined panel groups despite the aforementioned
heterogeneity,9 we adopt a different approach in which we give more importance to country
specific regressions. For this purpose, we first run an innovative methodology namely a regression
kink recently developed by Hansen (2017). This method is more appropriate as it permit searching
endogenously for unknown thresholds.
To support our results, we also run panel regressions in which we assume that economic growth is
a non-linear10 (quadratic) form of debt. Our aim, by this second approach, is to show that we could
find thresholds in the relationship between debt and growth in heterogeneous panel groups as
shown by partisans of RR (2010) results, but these are instable by time and countries sample as
well as other institutional characteristics.
Our approach for panel regressions is different from the previous researches as it considers long
period of analysis split according to the major events in the international economic and political
order, as well as, varying sample by country size and level of public expenditures and government
effectiveness. This approach complements the first one in results: while the individual regressions
show different relationship curves by countries that are also instable over time, the panel
regressions show that threshold is highly affected by country sample and time period. Both
approaches argue against a unique threshold that fit all countries.
3.1. Country specific methodology
For the individual regressions, we use the kink regression method developed by Hansen (2017)
that searches for endogenous thresholds.11 The regression function is everywhere continuous
9
Only fewer studies weakly investigated country specific regressions such as in Gomez-Puig and Sosvilla-Rivero
(2017) that used time series regressions for 11 eurozone countries between 1961 and 2015 and draw conclusions based
on thresholds that vary from a minimum of 21% in France to a maximum of 61% in Belgium.
10
The example of debt-growth non-linear effect is becoming more used to test new econometric methods. See, for
example, Egert (2015), Henderson et al. (2015) or Hansen (2017).
11
The difference in countries in terms of institutions, governance, and economic policies, among others, contributes
mainly to such endogeneity bias. In general, these variables are difficult to measure, and their effects could be better
assessed in a theoretical model (such as endogenous growth model) rather than in a simple non-linear relationship for
which the main goal here is to detect a turning point in the link between debt and growth simply from the data-
generating process (DGP).
44
except on this threshold where the slope has a discontinuity. Instead of assuming exogenous known
thresholds as in many previous empirical researches and by the traditional regression discontinuity
models, this method considers that the threshold is unknown and should be estimated.
The recent regression kink with an unknown threshold constitutes an important advancement of
the threshold regression models. The first class of such models used the regression discontinuity
design (RDD), introduced early by Thistlethwaite and Campbell (1960) and recently enhanced by
regression kink design (RKD) (Nielsen et al., 2010) and emphasized theoretically by Card et al.
(2012). Both RDD and RKD are especially involved when a policy variable of interest (the
outcome) is totally or partially determined by a known assignment rule of an observed treatment
variable (covariate). Both methods become important for identifying causal effects in observations
settings in many areas such as educational outcomes, election outcomes and unemployment,
among others (Card et al., 2017).12 The only difference is that RDD uses a “discontinuity” or a
“jump” in level of a treatment status at a threshold of an assignment variable, while RKD examines
discontinuities in derivatives (slope discontinuities) rather than the level (Athey and Imbens,
2017). Despite their important use, some authors warn that their results could be biased especially
for smaller population size, and particularly in the presence of confounding nonlinearities between
an assignment variable and an outcome variable (Ando, 2017).
As for Hansen’s (2017) regression kink with an unknown threshold, it is the latest method to
determine thresholds endogenously without the need for a treatment or an assignment variable as
in the previous methods. The conventional regression kink design assumes that the threshold is
known. This is suitable in many policy-oriented applications where the threshold is determined by
policy (Hansen, 2017). Instead, we treat the threshold as an unknown to be estimated. This method
is particularly appropriate when either the threshold is not set by the policy, or when one wishes
to investigate the robustness of this assumption. The features of such regression correspond highly
with our aim of examining endogenous thresholds from a direct relationship between economic
growth and public debt. We aim to confirm or deny the existence of debt thresholds rather than to
measure any policy effects, as the previous methods do. Hansen (2017)13 developed an inference
and estimation toolkit that tests for the presence of the threshold, estimation, and inference on the
regression function and parameters.
Using (𝑎)− = 𝑚𝑖𝑛[𝑎, 0] and (𝑎)+ = 𝑚𝑎𝑥[𝑎, 0] to denote the “negative part” and “positive part”
of a real number a, Hansen's regression kink model takes the following form:
𝑔𝑡 = 𝛽1 (𝑑𝑡 − 𝛾)− + 𝛽2 (𝑑𝑡 − 𝛾)+ + 𝛽3′ ℎ𝑡 + 𝜀𝑡 (1)
12
Other threshold models include a variety of autoregressive time series models with threshold, used particularly in
financial applications (Chen et al., 2011).
13
A theoretical generalization of Hansen’s (2017) method to panel data appeared recently in Zhang (2017).
45
where (𝑔𝑡 , 𝑑𝑡 , ℎ𝑡 ) are, respectively, variables describing economic growth, public debt-to-GDP
ratio, and a k-vector of other explanatory variables which includes an intercept. 𝜀𝑡 is the error term,
which is independent and identically distributed with zero mean and constant variance
(𝜀𝑡 ~𝑖𝑖𝑑(0, 𝜎𝜀2 )). The variables (𝑔𝑡 , 𝑑𝑡 , ℎ𝑡 ) are observed for 𝑡 = 1, … , 𝑛. The parameters to be
estimated are the regression slopes 𝛽𝑖 , with 𝑖 = 1,2,3, and the parameter 𝛾 called the threshold or
“kink point”. In equation (1) the slope with respect to the variable d equals 𝛽1 for values of 𝑑𝑡 less
than 𝛾, and equals 𝛽2 for values of 𝑑𝑡 greater than 𝛾; yet the regression function is continuous in
variables 𝑑 and ℎ, except the slope with respect to 𝑑 which is discontinuous at 𝑑 = 𝛾 (kink point).
Equation (1) has 𝑝 = 3 + 𝑘 parameters. 𝛽 = (𝛽1 , 𝛽2 , 𝛽3 ) are the regression slopes and are
generally unconstrained so that 𝛽 ∈ 𝑅𝑝−1 . However, for the parameter 𝛾, the model only makes
sense if the threshold is in the interior of the support of the threshold variable 𝑑. We thus assume
that 𝛾 ∈ Γwhere Γ is compact and strictly in the interior of the support of 𝑑.
To be applied to the debt growth relationship, we rewrite equation (1) with lagged independent
variable 𝑑𝑡−1 (so that this is plausibly pre-determined) and set ℎ𝑡 = (𝑔𝑡−1 , 1) and then 𝛽3 = (𝛿, 𝑐)
so that the regression contains a lagged dependent variable to account for dynamic effects and
minimize autocorrelations. Equation (1) becomes:
𝑔𝑡 = 𝛽1 (𝑑𝑡−1 − 𝛾)− + 𝛽2 (𝑑𝑡−1 − 𝛾)+ + 𝛿𝑔𝑡−1 + 𝑐 + 𝜀𝑡 (2)
Equation (2) can be written as 𝑔𝑡 = 𝛽′ 𝑥𝑡 (𝛾) + 𝜀𝑡 , where 𝑥𝑡 = ((𝑑𝑡 − 𝛾)− , (𝑑𝑡 − 𝛾)+ , 𝑔𝑡−1 )′ and
the least squares criterion for estimation is:
1 2
𝑆𝑛 (𝛽, 𝛾) = ∑𝑛𝑡=1(𝑔𝑡 − 𝛽′ 𝑥𝑡 (𝛾)) (3)
𝑛
Minimizing (3) yields the least squares estimator:
(𝛽̂, 𝛾̂) = 𝑎𝑟𝑔 min {𝑆𝑛 (𝛽, 𝛾)} (4)

𝛽∈𝑅𝑘−1 ,𝛾∈Γ
The criterion 𝑆𝑛 (𝛽, 𝛾) is quadratic in 𝛽 but non-convex in 𝛾. Hansen (2017) uses a combination of
concentration and grid search. Particularly, by concentration we write:
1 2
𝛾̂ = arg 𝑚𝑖𝑛 𝑚𝑖𝑛 {𝑆𝑛 (𝛽, 𝛾)} = 𝑎𝑟𝑔 𝑚𝑖𝑛{𝑆𝑛 (𝛽̂ (𝛾), 𝛾)} = ∑𝑛𝑡=1(𝑔𝑡 − 𝛽 ′ 𝑥𝑡 (𝛾)) (5)
𝑘−1
𝛾∈𝛤 𝛽∈𝑅 𝛾∈𝛤 𝑛
Where, for a given 𝛾, the parameters 𝛽̂(𝛾) are the least squares coefficients from regressing 𝑔𝑡 on
𝑥𝑡 (𝛾). The kink point 𝛾̂ is determined by a grid search over 𝛾 ∈ Γ, and once found, the parameters
𝛽̂ are determined by standard least squares of 𝑔𝑡 on 𝑥𝑡 (𝛾̂).
46
The deduced regression function is then:
𝑔𝑡 = 𝛽̂′ 𝑥𝑡 (𝛾̂) + 𝑒̂𝑡 (6)
where 𝑒̂𝑡 are residuals with an estimated error variance:

1
𝜎̂ 2 = ∑𝑛𝑡=1 𝑒̂𝑡2 = 𝑆𝑛 (𝛽̂(𝛾), 𝛾) (7)
𝑛
In order to test for the unknown threshold, Hansen (2017) conducted an algorithm14 to test the
nested regression model (2) against the following linear model assuming 𝛽1 = 𝛽2 = 𝛽𝑙 :
𝑔𝑡 = 𝛽𝑙 𝑑𝑡−1 + 𝛿𝑔𝑡−1 + 𝑐 + 𝜀𝑡 (8)
3.2. Panel methodology
To investigate the existence of a threshold at which debt reduces growth, we proceed for the panel
approach by the same way as in Checherita and Rother (2010).15 We estimate the following
equation:
2
𝑔𝑖,𝑡+5 = 𝛼𝑑𝑖,𝑡 + 𝛽𝑑𝑖,𝑡 + 𝛾𝑖 + 𝜀𝑖,𝑡 (9)
where 𝑔𝑖,𝑡+5 is the 5-years lead economic growth for country 𝑖; 𝑑𝑖,𝑡 is the debt to GDP ratio for
country 𝑖 at time 𝑡, 𝛼 and 𝛽 are parameters associated with the debt and its square, 𝛾𝑖 is the constant
term associated with each country 𝑖 called fixed effect, and 𝜀𝑖,𝑡 is the error term with zero mean
and constant variance (𝜀𝑖,𝑡 ~𝑖𝑖𝑑(0, 𝜎𝜀2 )). The time lag of five years is assumed: the current debt
will likely affect growth after 5 years. This is the case where debt is more negatively correlated
with growth (Table 1.A).
This equation is analogous to many modelling curves in the economic literature: Mincer’s (1974)
earning equation, Laffer curve (tax rate, Government revenue) and Kuznets curve (income,
inequality); where the dependent variable is a quadratic polynomial form of the explanatory
variable and thus expected to have a turning point (negative slope of the quadratic term). Similarly,
the non-linear term in (9) assumes that the rhythm by which debt affects growth changes from a
specific turning point. For example, low public debt could have a positive effect on growth and
starting from a certain threshold (high debt), the effect becomes negative.
14
The algorithm is presented in the appendix C (algorithm 1).
15
Our exercise is distinct from Checherita and Rother (2010) by adopting a longer span of time as well as different
sample of countries and different instrumental variables. For example, these authors use Gross Capital Formation as
instrument while it is rejected in our choice. The Pearson’s correlations (Table A.3) reveal that this variable is weakly
correlated with the explanatory variable.
47
Equation (9) changes its trend if it has a derivative null at a certain level of debt ratio. The debt
threshold 𝑑̂ is then deduced by deriving (9) according to the debt ratio:
𝛼
𝑑̂ = − (10)
2𝛽
Equation (9) is to be estimated using Generalized Least Squares (GLS) method. However, to
remedy to the major problems raised by the literature in terms of endogeneity, which may be
caused in particular by the omission of other explanatory variables or by reverse causality, we use
the generalized method of moments (GMM) involving instrumental variables.
Thus, the equation to be estimated by GMM is:

2
𝑔𝑖,𝑡+5 = 𝛼𝑑𝑖,𝑡 + 𝛽𝑑𝑖,𝑡 + 𝜑𝐼𝑛𝑠𝑡 + 𝛾𝑖 + 𝜀𝑖,𝑡 (11)
where, 𝐼𝑛𝑠𝑡 are a set of instrumental variables and 𝜑 the set of their associated parameters. The
threshold is deduced as in (10).
Another interesting method used previously by some authors is the PSTR models standing for
panel smooth transition regression. Gonzalez et al. (2005) developed these models as an
enhancement for the panel threshold regression (PTR) models of Hansen (1999). These models
explain the dependent variable as a linear term of the independent variable augmented with non-
linear terms as a multiplication of the same independent variable with an indicator function. This
latter is modelled in the form of a logistic function which depends on a threshold variable. In case
the threshold variable is the same as the dependent variable, this yields a polynomial form as in
(9). Despite their popularity and advantages of accounting for fixed effects in the panel data, the
PSTR do not allow lagged explained variables to be in the right-hand side of the specification
(Colletaz and Hurlin, 2006). This means they do not allow for dynamic effects.
4. Data and preliminary analysis
In this section, we present the data sources and a preliminary analysis (as the Pearson’s correlations
and statistical heterogeneity tests) for the data generating process.
4.1 Data description
Regarding this work, we consider a sample of 20 developed countries over the period of 1880–
2010 from the IMF database. This is the same sample of countries used by RR (2010), in addition
to a significant amount of subsequent research. The methodology and description of the data are
48
published in Abbas et al. (2010).16 The database reports the public debt-to-GDP ratio. The latter
comes from the Maddison17 data according to the Geary-Khamiss method, in international dollar.
For consistency, the Maddison Project data is also used for GDP growth.
The 20 advanced countries are: Australia, Austria, Belgium, Canada, Denmark, France, Germany,
Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden,
Switzerland, the United Kingdom and the United States.
4.2 Correlations and data heterogeneity tests
Table A.1 (appendix A) shows that Pearson’s correlations tend to be negative, especially between
current debt and 5-years lead growth. 13 out of 20 correlations are significantly negative over the
period of 1950-2008. But, the weak correlations may suggest that the form of the relationship is
not necessarily linear, since the Pearson correlations assume linearity. This seems to be revealed
by scatter plots for individual countries.
An issue related to data and country samples is heterogeneity. The source of the heterogeneity
could be unobserved characteristics as a result of other variables involving economic policies and
institutions. The heterogeneity in country behaviors is known as the fixed effect in econometrics.
Some of the previous works have warned against the high heterogeneity of countries as a result of
differences in fiscal and monetary policies, country size, and quality of institutions.18
However, to the best of our knowledge, no paper has invoked the issue of data heterogeneity. We
take the opportunity to study such heterogeneity. RR (2010), and other researchers, used
descriptive statistics based on conditional means and medians to conclude, for all countries, the
existence of a common debt threshold beyond which GDP growth slows. This assumes that the
data generating process is homogenous for all countries. However, running appropriate statistical
tests for the equality of means, equality of medians and equality of variances across countries, for
the economic growth categorized by public debt, strongly rejects the null hypothesis of equality
for such statistics. This means that the data generating process is highly heterogeneous across
countries. Table A.2 presents the results of the equality tests for different periods. The source of
variation is several times higher between countries than it is within countries. The heterogeneity
seems to decrease as the period shortens. Specifically, over the period of 1991-2008, the null
16
The database and the paper are drawn from www.imf.org/external/pubs/cat/longres.aspx?sk=24332.
17
http://www.ggdc.net/maddison/maddison-project/data.htm.
18
Despite its importance, we prefer not enriching equation (9) for panel data by a set of variables representing
economic and monetary policies for two reasons. First, these variables are not necessarily observed over the same
long period of debt and growth. Second, this could deviate the model from its assumed non-linear form leading to
many ad-hoc relationships between variables. What we have done next is splitting the sample of the 20 countries and
run our regressions on subsamples according to their levels in the main types of government expenditures. The level
of expenditures should reflect to some extend the countries’ economic governance.
49
hypothesis of equality of means and medians is accepted at 5% level and the test indicates
variances homogeneity.
5. Country specific analysis
The aim of this section is to analyze the relationship between public debt and economic growth
and investigate whether a unique debt turning point exists for all countries. Surveyed empirical
studies have not reached any consensus about a clear relationship between debt and growth.
Despite the diversity of methods, they showed sensitivity to different econometric problems and
data samples shortness. Those using a long period of analysis as in Reinhart and Rogoff (2010,
2012) could have methodological issues, especially by adopting simple descriptive statistical
approaches to generalize for a common rule that fits all. However, those using somehow advanced
econometric analysis could have short time sample bias. Generally, their samples start after the
1970s. Our aim is to reconsider an investigation of the debt-growth relationship, starting from a
data descriptive approach and ending to econometric estimations over a long period. We try to
remedy to limitations of samples and econometric methods used in the previous literature by using
both advanced recent econometric tools and rich statistical data analysis over long period for panel
and country specific analyses.
5.1. Scatter plots analysis
Despite the diversity of the econometric methods used to study the growth-debt relationship, they
are generally applied to samples starting from the seventies. This coincides with the end of the
Bretton Woods system and the beginning of market liberalization in developed countries. To
remedy this shortcoming, we extend the analysis to the long period of 1880-2008, split to five sub-
periods corresponding to the main changes in the international economic and political order
(Rodrik, 2011; Obstfeld and Taylor, 2002).19 These world events may affect the stability of the
macroeconomic aggregates and their interdependencies, hence the debt and growth linkages. We
distinguish the following sub-periods:
- 1880-1913. This period fits with the end of the first globalization (mercantilism era);
- 1914-1945. A period with two devastating wars and the 1929 great depression. The
international economic order was marked by the gold standard regime;
19
Breakpoint tests based on Dickey-Fuller unit root test emphasized this partition around these dates. For example,
break dates in debt series are 1918, 1941, 1964 and 2007 for the USA; 1915, 1950, 1969 and 1990 for the UK; 1906,
1944, 1975 and 1996 for Japan; 1896, 1960 and 1991 for France; 1902, 1937, 1974 and 1992 for Germany. Detailed
results for all countries are available upon request from the author.
50
- 1946-1970. The world experienced strong growth and development during this period,
notably with the Bretton Woods agreements and fixed exchange rate regimes;
- 1971-1990. A period of turbulent economic and political events with the end of the
convertibility of the dollar in 1971 and the tendency to adopting floating exchange rates,
as well as the oil shocks of the 1970s with the coexistence of unemployment and inflation,
and the sovereign debt crises of the 1980s;
- 1991-2008. Countries underwent extensive financial liberalization under the Washington

Consensus recommendations, resulting in financial instability for many emerging markets.
The World Trade Organization (WTO) was created, bilateral and multilateral free trade
agreements proliferated, and the Euro currency was launched.
We first conduct a scatter plots analysis for the relationship between public debt and growth for
individual countries (Figures B.1 to B.9). The relationship is plotted over eight periods: the whole
period 1880-2008, the five previous sub-periods, and two other periods (1946-2008 and 1971-
2008). Scatter plots are organized for each period for individual countries. We draw scatter plots
for 5-years20 lead economic growth explained by the current public debt. The chart analysis is
supported by the Pearson’s correlations in Table A.1 for the debt and growth over different eras.
We tend to assume lag effect between debt and growth. We believe that a deficit financed by a
government debt will act with a certain delay on economic growth either on the short-term or in
the long-term. For example, capital expenditures, which assumed to affect economic growth more
than any other type of expenditures are likely to impact growth with a delay. For example, a port
or a road financed by bond issuance will make time to be constructed and begins to benefit to the
community. Furthermore, other macroeconomic channels by which debt affects economic growth,
such as interest rates, openness, population and others, are likely to react with a delay rather than
immediately affecting growth. Our statistical data endorsed such assumed delay in the effect of
debt on growth where current debt-to-GDP ratio is more correlated with lead growth than current
growth (Table A.1).
Individual scatter plot analysis shows that the relationship form changes by countries and periods.
We notably distinguish:
- Flat curves for the case of Austria (1991-2008 and 1880-2008), Denmark (1880-2008) and
Portugal (1946-1970);
- Negative linear as in Germany (1880-2008 and 1880-1913), the Netherlands (1914-1945),

Canada (1971-1990) and, Austria and Italy (1946-1990);
20
The 5 years’ lag is justified by the fact that the debt is supposed to affect growth over medium to long-run.
51
- Positive linear for Belgium (1880-2008), Australia and Portugal (1880-1913), Austria and
Germany (1914-1945), France and Switzerland (1946-1970), Ireland and Portugal (1991-
2008) and France (1946-1990);
- Convex relationship in Denmark and Norway (1880-1913), Germany and Greece (1946-
1970), Italy and USA (1991-2008), Italy, Portugal, Greece, Sweden, the United Kingdom
and Austria (1946-1970) and, Japan, Germany, the United States and Greece (1946-1990);
- Concave relationship as in France, Italy, Switzerland and New Zealand (1880-2013),

Australia, Canada, Italy, Spain and the United Kingdom (1914-1945), the Netherlands,
New Zealand and United Kingdom (1946-1970), Belgium, Denmark, France, Spain,
Sweden and Switzerland (1971-1990), France, Norway and Sweden (1991-2008),
Belgium, France, Spain and Denmark (1946-2008) and, Australia and Sweden (1946-
1990).
Scatter plot analysis highlights that the form of the debt-growth relationship is country and time
specific. The economic explanation is straightforward. We note especially differences across
countries in the economic and political governance, the structure of debt (external or domestic,
currency of denomination, maturity), the aging population (Germany and Japan versus Portugal
and Spain), the degree of openness, the size of the economy (Greece and Ireland versus Japan and
the United States), the structure of public expenditures, changeable economic policies and political
and external environment as well as the level of the development in a country over time. Therefore,
the fact that there is neither a unique economic policy, nor a comparable level of institutional and
demographic variables across countries, makes less defendable the conjecture of “one size debt
threshold that fits all”.
5.2. Regression kink results
We use the kink regression method of Hansen (2017), allowing detecting unknown thresholds.
Country specific regressions emphasized what is reported on country specific scatter plots. The
regression kink with an unknown threshold shows a variety of forms for the growth-debt
relationship. Tables 2 and 3 present the regression kink results for the whole and the postwar
periods, respectively.
Furthermore, for better visualization of the results, we graph the summary of our findings across
more than five landscape pages (Figures 1 to 5), where each page contains four countries’ results
for two periods: 1881–2010 and 1950–2010. For each figure, two panel charts are presented: the
left-hand-side chart for the whole period, and the right-hand side for the post-war period. Each part
contains a chart with three components: the first part (left) simply draws the debt ratio variable
over time. This shows a mixture of high upward and downward sloping trends over time, indicating
possible structural breaks. The second part (in the middle) shows kink regression plots, where the
red point corresponds to the turning point (kink point or threshold), along with the regression line
52
corresponding to equation (2). The third part (on the right) presents the threshold parameter in axis
with the F-statistic showing the minimum of the Fisher test that indicates the threshold. Asymptotic
values (confidence intervals) are displayed in dashed blue lines.
The threshold F-statistic is calculated according to the test of the hypothesis 𝐻0 : 𝛾 = 𝛾0 against
𝐻0 : 𝛾 ≠ 𝛾0 . The criterion test is to reject for high values of the F-type statistic 𝐹𝑛 (𝛾0 ), where
̂ 2 (𝛾)−𝜎
𝑛(𝜎 ̂ 2) 1
𝐹𝑛 (𝛾) = , and 𝜎̂ 2 = ∑𝑛𝑡=1 𝑒̂𝑡2 is from equation (7). The threshold 𝛾̂ corresponds to the
̂2
𝜎 𝑛
minimum of the threshold F-statistic which is also the minimum of the least squares criterion.
Hansen (2017) presents an algorithm for the bootstrap confidence intervals for parameters and the
F-statistic.21
Insert Tables 2 and 3 about here

Insert Figures 1 to 5 about here
According to these results, we can distinguish many cases depending on the sign and magnitude
of parameters 𝛽1 and 𝛽2 :
✓ Case 1: 𝛽2 > 𝛽1 > 0
The growth rate is always positively affected by debt and the effect becomes higher after the
debt threshold. This case is only supported by Australia over the period 1956-2010.
✓ Case 2: 𝛽1 > 𝛽2 > 0
The effect of debt over growth is always positive but is reduced for debt values above the
threshold compared to the effect of debt values below the threshold. This case is reported in
countries such as Ireland (1929-2010) and Norway (1881-2010).
✓ Case 3: 𝛽2 > 0 > 𝛽1
The effect is negative for debt ratios below the kink point and then becomes positive after that
point. This case is presented by Austria (1956-2010), Denmark, France, Italy, Spain, Sweden
and Switzerland over the long period 1881-2010.
✓ Case 4: 𝛽1 > 0 > 𝛽2
The effect is positive and then becomes negative after the turning point. This case is presented
by the United Kingdom (1956-2010), the United States (1791-2010), Austria (1956-2010),
Germany (1881-2010) and the Netherlands (1956-2010).
21
The algorithm is presented in the appendix C (algorithm 2).
53
✓ Case 5: 𝛽1 < 𝛽2 < 0
The effect is always negative but emphasized less after the turning point. This case is observed
in Japan over both the long and short periods and Italy over the recent period.
✓ Case 6: 𝛽1 ≅ 0 (respectively 𝛽2 ≅ 0) and 𝛽2 ≠ 0 (respectively 𝛽1 ≠ 0)
Growth is insensitive to the debt ratio before the threshold (respectively after the threshold)
and the effect has the sign of 𝛽2 (respectively 𝛽1 ). These special cases are shown by Ireland
(1956-2010), Portugal (1881-2010 and 1956-2010) and Norway 1881-2010).
✓ Case 7: Flat curves
Debt neutrality is shown for the case of the United States and the United Kingdom (1881-
2010), Australia (1910-2010), Canada (1881-2010), Denmark (1956-2010), Sweden (1881-
2010) and Greece (1884-2010).
The kink regression method emphasizes the differences in the data-generating process of the debt–
growth relationship across countries. It reveals that the thesis of a debt threshold is case-specific
and changeable over time. It also suggests that there is no simple formula determining a specified
debt threshold or even a range of it, considered a dangerous zone not to reach, just by considering
the direct relationship between economic growth and the public debt-to-GDP ratio. This considers
more theory-based models, taking into account country fundamentals that are different between
countries and impact the debt–growth interactions.
6. Panel specification analysis
We present in this section, scatter plots analysis and estimation results for the panel groups.
6.1. Scatter plots analysis
Figures 6 and 7 present scatter plots for the panel analysis between the 5-year lead economic
growth and the public debt ratio. Figures are presented in panel graphs by period for the whole and
the euro zone sample. These scatter plots show that economic growth is weakly correlated with
the public debt. The trend-line of the relationship is flat in almost all periods for both the whole
sample and the eurozone. The relationship is only apparently negative for the whole sample over
the period of 1950-2008 and the subsequent postwar period (1946-1970). For the euro zone
sample, the analysis is nearly the same except a weak positive correlation over the period of 1991-
2008.
Insert Figures 6 and 7 about here
54
6.2 Panel regressions
We first estimate equation (9) using GLS method. However, for the GMM method, a set of
convenient instrumental variables is needed. The choice of instrumental variables for the GMM
method can be challenging. In practice, these are chosen to be correlated with the explanatory
variables and orthogonal to the error term before introducing instruments (error in equation (9)),
which means weakly correlated to the dependent variable of the initial regression before
considering such instruments. We choose a set of variables that can act as potential instruments
based on the Pearson’s correlations with the explained and explanatory variables (Table A.3). The
following variables and their first and second lags are considered to be candidates: old people
dependency ratio (ODR), shares to GDP of, respectively; exports (EX), imports (IM), government
consumption (GC) and gross capital formation (GCF). The data source of the dependency ratio
series is the World Development Indicators of the World Bank, while all other instruments are
from the Penn World Tables (version 9.0), which provides data back to 1950 for all the countries
of the sample adjusted for purchasing power parity (PPP).
Table A.3 shows first that, gross capital formation is weakly correlated with 5-years lead growth
but also weakly correlated with debt. As a result, it is removed from the list of candidate
instruments. Second, exports, imports, dependency ratio and government consumption are
correlated with the debt variable and weakly correlated with growth. We also used the Sargan test
which excluded the GCF at the estimation stage.22
Equation (9) is estimated over the periods of 1950-1970, 1971-1990, 1991-2008, 1971-2008 and
1950-2008. The periods of 1880-1913 and 1914-1945 were excluded as the debt series experienced
breaks for several countries during these periods as a result of the great depression and the World
wars. Similarly, estimations start from 1950 instead of 1946 as the debt of many advanced
countries stands highly abnormal following World War II. Reinhart and Rogoff (2009) reported
that defaults and restructuring debt in these times are among the highest in history.23 For a similar
reason, the recent financial crisis is not considered in estimation. The main argument is that these
points at the end of the period could statistically distort the results in addition to the ambiguous
character of the crisis on debt and growth.24
Equation (11) is estimated with the fixed effect25 relative to each country 𝛾𝑖 . Some authors remove
the fixed effect by differentiating their models. However, the fixed effect is important to keep in
our case as it considers the heterogeneity of the panel. The fixed effect can be removed if the
22 2
This statistic follows a 𝜒𝑟−𝑘 distribution where 𝑟 is the number of instruments and 𝑘 is the number of estimated
parameters (including the constant term).
23
The GMM instruments are not observed before 1950.
24
Baum et al. (2013) tested this effect for the euro area sample by introducing the years 2009-2010 and found a
considerable upward effect on the threshold, especially in dynamic panel regression.
25
The term fixed effects imply that although the intercept may differ across countries, it is time invariant. The fixed
effects model allows for heterogeneity or individuality among countries.
55
heterogeneity is rejected. Furthermore, differencing the equation, in our case, will modify the
assumed non-linear quadratic form and not allow to easily deduce the concavity of the relation and
thus the value of the threshold according to the formula (10). We also run a variety of tests for no
cross-section dependence26 for the estimated panel model over all the periods. Table A.6
summarizes results for the 20 OECD sample of countries and the 10 Euro sample of countries.
These tests strongly reject the null hypothesis of no cross-section dependence. Therefore,
heterogeneity and cross-section dependence among other problems facing panel methodology
support our approach considering studying growth-debt relationship for countries individually.27
Table 4 presents estimation results by GLS and GMM methods for the five periods of; 1950-1970,
1971-1990, 1991-2008, 1971-2008 and 1950-2008. The results are not significant for the GLS
over all periods except the whole period 1950-2008 over which, the form of the equation is convex
(β > 0).
However, the GMM method leads to significant results. For the fixed effects model, two thresholds
result from the concavity of the relation (β < 0) over the two periods 1971-1990 and 1971-2008,
with respective values of 47.5% and 46.5%. Assuming a model without fixed effects, the estimates
are statistically significant over three periods: 1971-1990, 1991-2008 and 1971-2008, with
respectively 49.4%, 80.1% and 62.8% thresholds. The J statistics shows the efficiency of the
instruments considered in the regressions.
These results emphasized the drawbacks from which the GLS method suffers. It reveals also that
the relationship is affected by high heterogeneity behaviors across countries. We deal with the first
issue by continuing the estimation using the GMM method. For the heterogeneity issue, we
considered the size effect between countries and vary the sample by removing sizable countries
namely Japan, the United Kingdom, and the United States. We also present the results for the
sample of the euro area countries as a group of homogenous monetary and exchange rate regimes.
Similarly, we filter countries according to a number of characteristics (government effectiveness
and expenditure levels reflecting public economic governance).
For the fixed effects model (Table 5), the results of the estimates show a concave relationship
between debt and growth for the periods of; 1950-1970, 1971-1990 and 1971-2008. The results do
26
Testing for cross-sectional dependence is crucial for selecting the appropriate and efficient estimator. We use four
tests: Lagrange multiplier (LM) test from Breusch and Pagan (1980), two tests of Pesaran (2004, 2006), one based on
Lagrange multiplier and the other on pairwise correlation coefficients. The latter has a lower power when the
population average pair-wise correlations are zero (Pesaran et al., 2008). The fourth test is proposed by Pesaran and
al. (2008), which developed a bias-adjusted test that is a modified version of the LM test.
27
For comparison purpose with the regression kink, we also run a dynamic quadratic form
2
(𝑔𝑡 = 𝑐 + 𝛼𝑑𝑡−1 + 𝛽𝑑𝑡−1 + 𝛿𝑔𝑡−1 + 𝜀𝑡 ) estimation for each country using GLS method over 1880-2010 (no observed
instruments on this period) and GMM method for 1950-2010. Results, available upon request from the author, are not
significant.
56
not improve by excluding Japan alone. The two periods of 1971-2008 and 1950-2008 on which
the relationship becomes significant shows, a convexity of the relationship. Excluding Japan and
the United States, the model improves over the period 1950-1970 and the threshold over this period
is around 78.4%. Adding the United Kingdom to the excluded countries, the results of the model
remain substantially the same. The threshold varies from 40% over 1971-1990 to nearly 78% for
the period 1950-1970. For the euro area, the relationship was concave and significant over the sub-
periods of; 1950-1970, 1971-1990, 1991-2008 and 1971-2008. Estimates for the period of 1950-
2008 are rejected. The threshold ranges from a minimum of 45% over the period 1971-2008 to a
maximum of 94% over the period 1991-2008.
Insert table 5 about here
We generate fixed effect for this model in Table A.4. This reports the deviation of each country
(𝛾𝑖 − 𝛾̅ ) from the homogenous constant term (the average constant for the whole panel). These
results emphasize high spreads across countries which makes it possible to conclude that the
heterogeneity greatly affects the growth-debt linkages. This may explain the big differences in
results across periods and samples. The threshold estimates are not tiny to consider as a claim for
a one unique threshold for all countries in all times. On the contrary, this strongly suggests that the
growth debt relationship is rather country specific than a common one. Furthermore, even when
considering no fixed effects results (Table 6), no substantial improvement is found.
For more investigation, we vary the samples by grouping countries according to their level of
government effectiveness and expenditure levels reflecting good public economic governance. We
split the sample into two groups of high and moderate level based on government effectiveness
and shares to GDP of respectively: total final government consumption, military expenditures and,
government transfers and subsidies. The first variable is from Worldwide Governance Indicators
of the World Bank, while other variables are from the World Development Indicators of World
Bank database. Countries are filtered form high to low values on average over time and the frontier
between high and moderate groups is determined according to the average of countries averages.
Therefore, the average government consumption to GDP over 1960-2008 ranges from 10.1% for
Switzerland to 23.6% recorded for Sweden. The average of the sample is 17.7% and this value is
used to split the two groups. Similarly, the government effectiveness index average over 1996-
2016 ranges from 0.53 for Italy to 2.09 for Finland, and its average over the 20 countries is 1.58
which separates the group of countries with moderate governance from the one with high
governance. The same approach was followed for the other variables.
Table A.5 summarizes the estimated thresholds by periods and country samples for the two groups
of countries for each of the four considered variables. The estimation is done by GMM method.
The results are fuzzy, and many non-significant thresholds are reported especially over the period
57
1951-2008. However, the recent period seems to point out to more homogeneity in countries
behavior and significant thresholds, though different by sample, are reported. These results
emphasize the idea of studying the debt-growth relationship on country by country case.
The investigation (both preliminary data analysis and estimation) revealed high heterogeneity in
data and behaviors across countries. Despite that we could sometimes prove the existence of a debt
threshold by advanced econometric methods, the heterogeneity suggests that this one tend to be
rather country specific than common rule for all countries. Furthermore, when dealing with panel
cross-section data, it is usually assumed that cross section errors in panel data models are
independent, especially for large cross-section size. However, the presence of such cross-section
dependence in estimation can result in serious problems of efficiency loss and inappropriate
statistical tests.
7. Conclusion
In this paper, we studied the existence of possible threshold effects in the relationship between
public debt and economic growth. We used two econometric approaches. The first method, applied
individually to each country of the sample of 20 advanced countries, is the kink regression method
developed by Hansen (2017), which searches endogenously for an unknown threshold. The second
method is a technique that previously was explored by some authors especially Checherita and
Rother (2010), which we applied to the panel of the previous countries.
Both methods clearly counter the claim of a common debt threshold that fit all countries. It reveals
that the threshold, whenever it exists, is a country specific rather than a common rule to fit all.
Unlike all the empirical literature examining the existence and values of debt thresholds on a cross-
section data, our analysis undertakes the question on both cross-section and country-specific over
a long data span. Country-specific analysis highlighted, in fact, diverse types of relationship
between growth and public debt. Accordingly, some countries can grow with high debt to GDP
ratios; others could see their growth shrink from even low debt ratios, while growth in some others
is insensitive to public debt. The study also reveals the instability of the relationship over time:
almost every country exhibits a different relationship by period, especially when considering the
whole period (1880–2008) and the post-war period (1950–2008).
The results drawn from our study point to further interesting developments since several economic
and institutional variables such as interest rate and governance could be integrated into the analysis
as they could have some notable effects on the debt-growth nexus beyond the simple model
developed here.
58
Chapter 1. Tables and Figures
59
Table 1. Empirical studies survey of the public debt and economic growth threshold
Authors Samples Econometric methodologies Debt thresholds Other notes on methods and results
1 Reinhart and Rogoff 20 developed countries and 24 Growth mean and the median analysis 90% for advanced Arbitrarily set thresholds: 30%, 60%
(2010) developing countries: 1946– according to pre-established thresholds countries and 90% and conclude that growth is
2009 altered from 90%
2 Caner et al. (2010) 26 developed and 75 Hansen’s (2000) OLS threshold 77% and 64% 77% for the whole sample, 64% for
developing countries: 1980– econometric estimate developing countries
2008
3 Herndon et al. The 20 developed countries of Replication of RR’s (2010) work after No threshold found The relationship is sensitive to country
(2013) RR (2010) correction of data errors and period
4 Kumar and Woo 38 countries, most of which are Econometric estimation by various High level but not Correction of problems of reverse
(2010) OECD countries: 1970–2007 methods such as GMM reported causality, endogeneity and
heterogeneity
5 Checherita and 12 euro area countries: 1970– Estimation of a polynomial form of Interval [70% to 100%] The degree of the polynomial form is
Rother (2010) 2009 growth explained by the debt ratio around 1.2 to 3
6 Checherita et al. 11 euro, 22 OECD and 14 EU Cobb-Douglas optimization 65%: OECD; 63%: the The debt threshold is a non-linear
(2014) countries: 1960–2010 augmented by public spending EU; 50%: euro function of the elasticity of the public
expenditure/private capital ratio
7 Pescatori et al. 34 mainly developed countries: Analysis of growth (t + k) and debt (t); No threshold found High government debt tends to
(2014) 1875–2011 k = {1,5,10,15} using a descriptive increase economic growth volatility
approach
8 Chang and Chiang 15 OECD countries: 1990– Threshold method for non-dynamic 32.3% and 66.25% The relationship is positive over the
(2009) 2004 panel three regimes delimited by the two
thresholds
9 Cecchetti et al. 18 OECD countries: 1980– Threshold method for non-dynamic 85% Debt negatively influences growth
(2011) 2010 panel above 85% threshold
10 Minea and Parent 20 developed countries of RR Estimation of econometric relationship 60%, 90% and 115% Between 90% and 115%, negative
(2012) (2010) and Abbas et al. (2010) with changing thresholds effect, between 60% and 90% and
data above 115% positive
11 Egert (2015) The RR (2010) data and Estimation and detection of 20%: central debt; 50%: Individual estimate confirms threshold
countries sample endogenous thresholds (Hansen, 1999) general debt for some countries around 30%
(United States)
12 Baglan and Yoldas 20 developed countries of RR Inference techniques to remedy Between 18% and 53% No robust threshold and subject to
(2013) (2010): 1954–2008 endogeneity and heterogeneity issues uncertainty
13 Baum et al. (2013) 12 euro countries: 1990–2010 Threshold method for dynamic and 67% Study of short-term impact
non-dynamic panel
14 Kourtellos et al. Sample of 82 countries: 1980– Threshold regressions using the Solow No threshold found Heterogeneity and influence of
(2012) 2009 growth model institutional quality
60
Table 1 (continued)
Authors Samples Econometric Methodologies Debt thresholds Other notes on methods and results
15 Eberhardt and Sample of 105 countries: 1970– Techniques addressing heterogeneity No common Estimation of dynamic and static non-
Presbitero (2015) 2008 and dependence in cross-sections threshold linear models by GMM method
16 Panizza and Presbitero 17 OECD countries: 1981– Estimation by GMM of linear and non- No threshold found No negative effect of the debt on growth
(2012) 2008 linear relationships
17 Sharpe (2013) 12 euro countries: 1998–2011 Estimated relationship of debt and 40% and 133% Negative effect for debt ratios over 40%,
interest rate by GMM and TSLS emphasized above the 133% threshold
18 Pan and Wang (2013) 12 euro countries: 1970–2009 Bayesian analysis using dynamic Negative Common factors and shocks affect
factor models relationship positive growth and negative debt
19 Gomez-Puig and 11 euro countries: 1980–2013 Granger (1969) causality method 56% to 103% Causality dependent on the country, the
Sosvilla-Rivero (2015) threshold found only for four countries
20 Di Sanzo and Bella 12 countries of the euro: 1970– Studies of individual causality by non- Threshold not Causality results vary across countries
(2015) 2012 parametric tests examined
21 Greiner (2011) Long-term simulation for Italy, Simulated endogenous growth model Threshold not The impact of debt on growth is positive
Germany and the eurozone examined if the pace of debt remains lower than
the pace of GDP
22 Marchionne and Sample of 27 countries: 1994– Estimation considering the GINI index Non-linearity; no The results suggest non-linear link that
Parekh (2015) 2010 threshold reported depends on the income distribution
23 Lin (2014) 62 developing and developed Threshold quantile Lasso regression Thresholds ranging Thresholds vary by country and quintile
countries: 1991–2005 from 10% to 100% and more common in developing than in
developed countries
24 Lee et al. (2017) RR (2010) database for 20 Test for threshold effects by regressing Around 30% The median real GDP growth falls
developed countries growth median on public debt abruptly above a debt-to-GDP ratio of
30%
25 Gomez-Puig and 11 euro area countries: 1961– Time-series regressions based on Variable threshold Threshold varies across countries from
Sosvilla-Rivero (2017) 2015 economic growth literature from 21% to 61% 21% in France to 61% in Belgium
26 Chudik (2017) 40 advanced and developing Test for thresholds in dynamic No evidence of any Significant negative long-term effects of
countries: 1965–2010 heterogeneous panel with cross- threshold public debt build-up on output growth
sectionally dependent errors
27 Syssoyeva-Masson and 60 developed and developing Panel regression using time-series Two regimes: 44% Highlight the debt long-memory process
De Sousa Andrade countries, IMF data: 1970– cointegration and Hansen (1999) and 48% and recommend long-term analysis
(2017) 2012
61
Table 2. Regression kink results over long periods

Specification 𝑔𝑡 = 𝛽1 (𝑑𝑡−1 − 𝛾)− + 𝛽2 (𝑑𝑡−1 − 𝛾)+ + 𝛿𝑔𝑡−1 + 𝑐 + 𝑒𝑡
Country β1 β2 δ c γ Period
Australia 0.002 (0.321) 1.719 (0.234) 0.49 (0.503) 1.11 (2.996) 134 (167.9) 1901-2010
Austria 0.070 (0.072) -0.04 (0.039) 0.584 (0.085) 2.389 (1.335) 35.6 (6.4) 1881-2010
Belgium 0.000 (0.000) 0.088 (0.394) 0.214 (0.135) 1.487 (0.333) 173.4 (231.3) 1881-2010
Canada 1.016 (1.133) 0.002 (0.016) 0.314 (0.127) 1.471 (0.978) 22.7 (4.1) 1881-2010
Denmark -0.067 (0.058) 0.048 (0.04) 0.016 (0.174) 0.679 (0.793) 33.4 (6.8) 1881-2010
France -0.029 (0.01) 0.167 (0.064) -0.203 (0.147) -0.306 (1.22) 161 (20.4) 1881-2010
Germany 0.318 (0.127) -0.051 (0.033) 0.459 (0.148) 2.587 (0.904) 20.1 (2.3) 1881-2010
Greece -0.008 (0.016) -2.063 (0.194) 0.027 (0.07) 1.172 (2.576) 215 (1.5) 1884-2010
Ireland 0.177 (0.191) 0.04 (0.016) 0.435 (0.15) 1.07 (0.829) 35.1 (7.9) 1929-2010
Italy -0.058 (0.035) 0.053 (0.053) 0.371 (0.146) 0.001 (1.051) 66 (18.8) 1881-2010
Japan -0.434 (0.159) -0.006 (0.008) -0.11 (0.104) 2.564 (0.671) 22.8 (4.9) 1881-2010
Netherlands 0.04 (1.806) -0.009 (1.182) 0.233 (3.205) 3.235 (26.27) 110.9 (2439) 1881-2010
New Zealand -0.005 (0.01) 0.074 (0.112) -0.048 (0.107) 0.814 (0.973) 163.3 (78.2) 1881-2010
Norway 2.647 (0.87) 0.032 (0.028) -0.123 (0.14) 2.501 (0.705) 14.6 (1.4) 1881-2010
Portugal 125.31 (17.9) -0.069 (0.022) -0.106 (0.137) 4.482 (0.954) 13.6 (0.0) 1881-2010
Spain -0.059 (0.018) 0.017 (0.028) 0.014 (0.098) 0.399 (0.796) 74.6 (16.2) 1881-2010
Sweden 6.189 (0.01) -0.001 (0.048) 0.109 (0.102) 2.092 (1.21) 13.4 (16.3) 1881-2010
Switzerland -0.023 (0.016) 3.977 (0.595) 0.047 (0.119) 0.272 (0.63) 74.3 (0.6) 1899-2010
United Kingdom -30.542 (0.02) 0.00 (0.235) 0.362 (0.067) 0.932 (1.571) 27.5 (55.0) 1881-2010
United States 34.306 (8.17) -0.013 (0.02) 0.292 (0.107) 3.082 (0.936) 7.6 (0.1) 1881-2010
Note: Standard errors are given in parentheses.
Table 3. Regression kink results over the period 1956-2010
Specification 𝑔𝑡 = 𝛽1 (𝑑𝑡−1 − 𝛾)− + 𝛽2 (𝑑𝑡−1 − 𝛾)+ + 𝛿𝑔𝑡−1 + 𝑐 + 𝑒𝑡

Country β1 Β2 δ c γ
Australia 0.034 (0.033) 0.614 (0.155) 0.119 (0.167) 2.259 (0.697) 37.4 (1.1)
Austria -0.086 (0.038) 0.039 (0.025) -0.017 (0.141) 1.555 (0.517) 44.0 (7.7)
Belgium 51.31 (4.898) -0.016 (0.024) 0.157 (0.107) 2.923 (0.694) 38.9 (3.8)
Canada 0.149 (0.238) -0.006 (0.019) 0.328 (0.138) 1.587 (0.634) 52.7 (10.9)
Denmark -36.619 (0.028) -0.007 (0.095) 0.033 (0.024) 2.390 (1.191) 4.4 (15.4)
France 0.521 (0.478) -0.024 (0.014) 0.476 (0.16) 1.705 (0.667) 16.1 (1.3)
Germany -0.132 (0.061) 0.005 (0.033) 0.038 (0.183) 1.326 (0.817) 34.8 (7.7)
Greece -0.058 (0.035) 0.053 (0.053) 0.371 (0.146) 0.001 (1.051) 66 (18.8)
Ireland 3.28 (1.227) 0.037 (0.015) 0.346 (0.128) 1.456 (0.667) 27.3 (1.0)
Italy -0.053 (0.014) 0.091 (0.045) 0.100 (0.197) 0.502 (0.438) 105.4 (4.7)
Japan -0.339 (2.241) -0.011 (0.016) 0.299 (0.2) 2.171 (2.383) 18.1 (75.7)
Netherlands 0.083 (0.039) -0.064 (0.113) 0.162 (0.179) 2.821 (0.962) 61.7 (10.2)
New Zealand 1.81 (0.055) -0.014 (0.023) -0.074 (0.169) 2.058 (0.771) 19.4 (0.4)
Norway -0.037 (0.023) 0.793 (0.007) 0.337 (0.139) 0.926 (0.677) 57.9 (0.7)
Portugal 10.74 (2.203) -0.066 (0.021) 0.377 (0.132) 3.998 (0.95) 14.5 (0.2)
Spain -0.064 (0.078) 0.038 (0.026) 0.549 (0.126) 0.543 (0.841) 34.5 (18.7)
Sweden -0.119 (0.044) 0.056 (0.025) 0.394 (0.156) -0.198 (0.579) 40.2 (5.6)
Switzerland -0.083 (0.036) 0.05 (0.037) 0.177 (0.144) 0.288 (0.492) 39.9 (8.8)
United Kingdom 0.124 (0.072) -0.021 (0.019) 0.154 (0.18) 3.037 (0.873) 59 (8.5)
United States 1.892 (0.617) -0.072 (0.028) 0.028 (0.156) 4.650 (0.800) 34.4 (0.7)
Note: Standard errors are given in parentheses.
62
Table 4. GLS and GMM panel estimation results

2
Regression by Generalized Least Squares method (GLS). Specification: 𝑔𝑖,𝑡+5 = 𝛼𝑑𝑖,𝑡 + 𝛽𝑑𝑖,𝑡 + 𝛾𝑖 + 𝜀𝑖,𝑡
Model with fixed effects
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
Coef. prob. Coef. prob. Coef. prob. Coef. prob. Coef. prob.
 0.010 0.647 -0.017 0.309 0.049** 0.030 -0.003 0.770 -0.045*** 0.000
 0.000 0.689 0.000 0.789 0.000 0.196 0.000 0.801 0.000*** 0.000
 3.296*** 0.000 2.685*** 0.000 -0.470 0.620 2.229*** 0.000 4.206*** 0.000
Threshold NS NS NS NS 123.3a
Model without fixed effects
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 -0.038*** 0.000 -0.014 0.307 0.015 0.234 -0.005 0.576 -0.038*** 0.000
 0.000*** 0.007 0.000 0.546 0.000 0.251 0.000 0.841 0.000*** 0.000
 4.619*** 0.000 2.411*** 0.000 1.448*** 0.002 2.165*** 0.000 3.802*** 0.000
a
Threshold 139.0 NS NS NS 107.0a
2
Regression by Generalized Method of Moments (GMM). Specification: 𝑔𝑖,𝑡+5 = 𝛼𝑑𝑖,𝑡 + 𝛽𝑑𝑖,𝑡 + 𝜑𝐼𝑛𝑠𝑡 + 𝛾𝑖 + 𝜀𝑖,𝑡
Model with fixed effects
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.159 0.671 0.190** 0.014 -0.204 0.592 0.093** 0.013 0.060 0.469
 -0.001 0.733 -0.002*** 0.006 0.002 0.389 -0.001*** 0.004 -0.001 0.169
 -0.444 0.957 -0.707 0.625 5.571 0.690 -0.014 0.989 3.290** 0.019
b
J-statistic 2.755 0.097 3.923 0.141 10.774 0.001 0.334 0.563 3.692 0.055
Threshold NS 47.5 NS 46.5 NS
Model without fixed effects
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.038 0.808 0.234* 0.069 0.161** 0.043 0.106** 0.016 0.566 0.186
 -0.001 0.529 -0.002* 0.081 -0.001** 0.034 -0.001*** 0.005 -0.006 0.157
 4.176* 0.099 -2.845 0.277 -3.684 0.202 -0.682 0.616 -6.376 0.397
J-statisticb 4.526 0.339 2.949 0.399 0.786 0.675 1.143 0.565 0.666 0.717
Threshold NS 49.4 80.1 62.8 NS
Notes: Significant at 1% (***), 5% (**) and 10% (*). NS is Non-Significant.
a: Coefficients are statistically significant, but the function is convex which means that the debt affects negatively
growth for debt ratios below this threshold and positively above this threshold.
b: J-statistic is the value of the GMM objective function.
The threshold is calculated according to the formulae: 𝑑 ̂ = − 𝛼.
2𝛽
63
Table 5. GMM results for varying countries sample (fixed effects model)
Sample of 20 OECD countries
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.159 0.671 0.190** 0.014 -0.204 0.592 0.093** 0.013 0.060 0.469
 -0.001 0.733 -0.002*** 0.006 0.002 0.389 -0.001*** 0.004 -0.001 0.169
γ -0.444 0.957 -0.707 0.625 5.571 0.690 -0.014 0.989 3.290** 0.019
J-statisticb 2.755 0.097 3.923 0.141 10.774 0.001 0.334 0.563 3.692 0.055
Threshold NS 47.5 NS 46.5 NS
Sample of 19 OECD countries (Japan excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.137 0.457 0.184** 0.015 -0.981*** 0.007 0.098** 0.019 -0.172*** 0.000
 -0.001 0.579 -0.002*** 0.008 0.009*** 0.005 -0.001*** 0.008 0.001** 0.013
γ -0.364 0.932 -0.764 0.597 23.638*** 0.007 0.113 0.912 7.298*** 0.000
J-statisticb 2.646 0.104 3.097 0.213 6.874 0.009 0.149 0.700 4.641 0.098
Threshold NS 40.8 56.6a 51.5 75.5a
Sample of 18 OECD countries (Japan and USA excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.344** 0.017 0.186** 0.015 -0.731** 0.018 0.091** 0.024 -0.162*** 0.000
 -0.002** 0.041 -0.002*** 0.008 0.007** 0.012 -0.001*** 0.010 0.001** 0.011
γ -4.910 0.138 -0.731 0.609 16.307** 0.028 0.296 0.766 7.139*** 0.000
J-statisticb 5.375 0.146 2.402 0.301 11.895 0.003 0.372 0.542 3.692 0.055
Threshold 78.4 40.8 54.0a 51.0 79.3a
Sample of 17 OECD countries (Japan, USA and UK excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.296** 0.033 0.162** 0.028 -0.500* 0.052 0.070** 0.047 -0.172** 0.011
 -0.002* 0.074 -0.002** 0.013 0.005** 0.030 -0.001** 0.017 0.001* 0.075
γ -3.428 0.269 -0.256 0.851 10.658* 0.088 0.832 0.333 7.496*** 0.000
J-statisticb 3.912 0.418 1.916 0.384 17.670 0.000 0.389 0.823 7.681 0.021
Threshold 76.5 40.1 51.9a 48.6 80.3a
Sample of 10 Euro countries
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 1.929* 0.096 0.100* 0.057 1.063** 0.013 0.064* 0.071 -0.123** 0.019
 -0.014* 0.095 -0.001** 0.027 -0.006* 0.092 -0.001** 0.021 0.001 0.126
γ -37.593 0.136 0.655 0.557 -41.709*** 0.006 1.443* 0.085 6.689*** 0.000
J-statisticb 2.191 0.534 3.652 0.161 4.996 0.172 0.765 0.382 4.387 0.112
Threshold 69.6 46.1 94.0 45.0 NS
Notes: Significant at 1% (***), 5% (**) and 10% (*). NS is Non-Significant.
a: Coefficients are statistically significant, but the function is convex which means that the debt affects negatively
growth for debt ratios below this threshold and positively above this threshold.
The threshold is calculated according to the formulae: 𝑑 ̂ = − 𝛼.
2𝛽
64
Table 6. GMM estimation results for reduced samples for model without fixed effects
Sample of 20 OECD countries
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
coef. prob. coef. prob. coef. prob. coef. prob. coef. prob.
 0.038 0.808 0.234* 0.069 0.161** 0.043 0.106** 0.016 0.566 0.186
 -0.001 0.529 -0.002* 0.081 -0.001** 0.034 -0.001*** 0.005 -0.006 0.157
γ 4.176* 0.099 -2.845 0.277 -3.684 0.202 -0.682 0.616 -6.376 0.397
J-statisticb 4.526 0.339 2.949 0.399 0.786 0.675 1.143 0.565 0.666 0.717
Sample of 19 OECD countries (Japan excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.054 0.753 0.291** 0.037 0.172* 0.092 0.136*** 0.003 0.711 0.357
 -0.001 0.526 -0.003** 0.042 -0.001* 0.088 -0.001*** 0.001 -0.008 0.322
γ 3.657 0.210 -3.871 0.166 -3.976 0.280 -1.394 0.304 -9.868 0.508
J-statisticb 6.446 0.168 0.764 0.858 4.645 0.326 0.571 0.752 3.765 0.152
Sample of 18 OECD countries (Japan and United States excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.033 0.799 0.254** 0.036 0.162* 0.074 0.109*** 0.002 0.270 0.416
 -0.001 0.502 -0.003** 0.039 -0.001* 0.063 -0.001*** 0.001 -0.004 0.339
γ 4.113** 0.049 -2.873 0.220 -3.160 0.310 -0.524 0.592 -0.656 0.904
J-statisticb 5.507 0.239 0.744 0.863 4.562 0.335 0.242 0.886 5.574 0.062
Sample of 17 OECD countries (Japan, United States and United Kingdom excluded)
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.247* 0.096 0.223* 0.066 0.415** 0.021 0.109*** 0.003 0.249* 0.069
 -0.003* 0.055 -0.002* 0.063 -0.003* 0.051 -0.001*** 0.001 -0.003** 0.045
γ 0.104 0.966 -2.088 0.364 -12.292** 0.021 -0.601 0.569 -1.585 0.541
J-statisticb 2.873 0.412 0.386 0.943 0.319 0.853 0.594 0.743 7.465 0.058
Threshold 36.55 45.31 76.62 59.12 45.09
Sample of 10 Euro zone countries
1950-1970 1971-1990 1991-2008 1971-2008 1950-2008
 0.285* 0.101 0.226* 0.061 0.449** 0.035 0.074* 0.065 0.173 0.214
 -0.004* 0.091 -0.002* 0.053 -0.002** 0.041 -0.001** 0.041 -0.003* 0.098
γ 0.148 0.957 -2.624 0.339 -16.135* 0.057 0.591 0.544 1.741 0.423
J-statisticb 4.306 0.23 0.971 0.615 0.843 0.656 2.066 0.559 12.577 0.002
Threshold 36.6 55.4 91.7 52.6 NS
Notes: significant at 1% (***), 5% (**) and 10% (*). NS is Non-Significant.
The threshold is calculated according to the formulae: 𝑑̂ = − 𝛼.
2𝛽
65
Figure 1. Regression kink results for United States, Japan, United Kingdom and Australia
United States : 1881-2010 United States : 1956-2010

Turning point Asy mptotic v alues
*Turning point Asy mptotic v alues
* *
**
70
8 10 12
* *
5 10 15
6
* *
20 40 60 80 100
15
Threshold F Statistic
* *
* **
GDP Growth Rate

GDP Growth Rate
**
** * ** * *
*** ** * ****
60
4
*** * ** * * *
Debt/GDP
Debt/GDP
***** ** * **
* *** ***** * ** * * *
10
* ** * ***** **
****** ** ****
* *******
******** * * * * ** * * *
2
* * * **
* **
50
** *** ** *** *
6
-10 -5 0
*
** * * ** * *
**** *** *
4
*
0
5
* *
*** *
40
*
* ** *
2
-2
* * *
0
0
30
* *
1880 1940 2000 20 60 100 20 60 100 1960 1980 2000 30 40 50 60 70 0 20 40 60
year Debt/GDP Threshold Parameter year Debt/GDP Threshold Parameter
Japan : 1881-2010 Japan : 1956-2010
6 8 10 12
200
Turning point Asy mptotic v alues Turning point Asy mptotic v alues
* *
15
8 10 12
*
*
** **
20
150
***
* * *
GDP Growth Rate

*
GDP Growth Rate
10
150
******
* **
** **
15
* *
Debt/GDP
***** ** *
Debt/GDP
* * **
** ** ** *
100
* * *
5
**** ** **
100
* ***** *** ** ** ** **
6
*
10
************
*** **
* * * *
* ** * ** **
*
*
** ****** * * * * *
0
**
** ** *
2
* * ** * * * *
4
** * * *
* **
50
**
50
* *
5
**
-2 0
* *
-5
2
* *
** * *
*
**
*
0
-10
0
* * *
0
0
1880 1940 2000 0 50 100 200 0 50 100 200 1960 1980 2000 0 50 100 150 0 50 100 150
United Kingdom : 1881-2010 United Kingdom : 1956-2010
5
80 100 120 140

5
* * *
6
100 150 200 250
* * *
* *
**
4
5
GDP Growth Rate

*
GDP Growth Rate
*
* * * * ** ** *
***** ***** * * * **
4
***
**** *** * * ******* * * * **
Debt/GDP
***
**
Debt/GDP
** * ** * *
*** **** ***
3
**** *
3
* ***** * *
0
*
***** * * * * ** * *
* ** ** * **
* * *
2
***** ** *
*** *
* * * *** *
2
2
* *
** * *
-5
*
*
60
0
** * *
1
*
1
*
50
-10
40
*
* ***
0
0
-2
*
1880 1920 1960 2000 50 150 250 50 150 250 1960 1980 2000 40 60 80 120 40 60 80 120
Austra lia : 1901-2010 Australia : 1956-2010

2.5
20 40 60 80 100 120 140

10
*
* *Turning point Asy mptotic v alues
10 15 20 25 30 35 40
2.0
* *
1.5
** **
*
* * * ** ** **
GDP Growth Rate

* * *
GDP Growth Rate
*
** * *
**
* * * * *
** ** ** ** * * * *
** *
1.5
* * * ** * *
Debt/GDP
** *** *
** * * *
Debt/GDP
*
*** **
*
* * * *
**
1.0
****
*
* * * **
2
*
* *** ** ** ** * *
* * * * * * **
0
** * *
* ** * ***
1.0
*
* * * *
*
*
* ** ** *
* **
0.5
** *
-5
*
0.5
*
*
*
* *
-2
-10
0.0
0.0
* *
1900 1940 1980 20 60 100 140 20 60 100 140 1960 1980 2000 10 20 30 40 10 20 30 40
66
Figure 2. Regression kink results for Austria, Belgium, Canada and Denmark
Austria : 1881-2010 Austria : 1956-2010
10 20 30 40 50 60 70
Turning point Asy mptotic v alues *
*
0.0 0.5 1.0 1.5 2.0

80
4
GDP Growth Rate
20
GDP Growth Rate

* * * *
**
Debt/GDP
Debt/GDP
60
**** *
3
* *
10
4
** * ** * **
* * ** * * * **
*****
40
2
* * *** * * **** * *
* ****
***** ***
* ****** ********* *** **** ***
* **
*** * * * ****
2
**
* ***** **********
0
* * ** * * * * * * * * *
1
*
20
* * * *
-10
* *
0
* *
0
* * *
1880 1940 2000 20 40 60 80 20 40 60 80 1960 1980 2000 10 30 50 70 10 30 50 70
Belgium : 1881-2010 Belgium : 1956-2010
4
***
* Turning point Asy mptotic v alues
6
6
*** *** *
200
* * *
120
GDP Growth Rate

**
8
* **
* * *
GDP Growth Rate
* * **
4
3
* * * *
*
4
* * **
*
Debt/GDP
*
*** * * * ** **
Debt/GDP
150
6
* * * * * ** ** **
* ***** * * ***
2
** * * *** * * * ** * * *
* **
2
****
** * * * *
40 60 80
2
* *
****** * **
4
** ** * * ****
0
100
**** * * * *
* * * * **
* **
2
1
0
-4 -2
* *
* * *
*
50
* *
0
*
0
*
1960 1980 2000 40 60 80 120 40 60 80 120
1880 1940 2000 50 100 200 50 100 200
Canada : 1881-2010 Canada : 1956-2010
1.5
6
50 60 70 80 90 100
*
5 10 15
*
0.0 0.5 1.0 1.5 2.0 2.5
140
* * * *
**** **
4
** * * ***
* *
*** ** ** * *
GDP Growth Rate
GDP Growth Rate

** * * *
1.0
*
* ** ** ** *** * ** * *
**
Debt/GDP
20 40 60 80 100
********* ******* *
Debt/GDP
2
** ********* ** *
** *
*
** ************* *****
*
*** *
* *
*
* * ** * **
*
-15 -10 -5 0
* ** * ** * * * * *
** ** ** * * * *
0
** *** * * *
* *
0.5
*
** *
* ** *
* *
-2
* *
* *
-4
0.0
* *
1880 1940 2000 20 60 100 140 20 60 100 140 1960 1980 2000 50 70 90 50 70 90
Denmark: 1881-2010 Denmark: 1956-2010

1.5
10 15
10 20 30 40 50 60 70
10 20 30 40 50 60 70
8
* *
0.0 0.5 1.0 1.5 2.0 2.5

*
* *
*** * *
*
GDP Growth Rate
GDP Growth Rate

*
1.0
* * ** *
**
5
** * ** * * * *
** * * *
Debt/GDP
Debt/GDP
** *
***
***
****** * ** **
*
* * * **
*
* * *
4
** * *
*
** **
**
***
*
*
***
* **
* *
*
** **** ** *
0
* * * * * * ** *
*
* **
* ** * * ** *
* *
*** *
* * **
0.5
2
**
-15 -10 -5
* * * *
* * *
* *** * ***
* * *
* * *
0
* * *
* *
* *
0.0
* *
*
1880 1940 2000 10 30 50 70 20 40 60 1960 1980 2000 10 30 50 70 0 20 40 60
67
Figure 3. Regression kink results for France, Germany, Italy and Ireland
France : 1881-2010 France : 1956-2010

-1 0 1 2 3 4 5 6
* Turning point Asy mptotic v alues
5 10 15 20 25 30
* *
15
*
20 30 40 50 60
100 150 200
0.6
GDP Growth Rate
** * *
GDP Growth Rate

* *
* *
10
* ** *
** *
Debt/GDP
*
Debt/GDP
* *
0.4
* * ***
** ** ** *
* **
5
**** **
** * ** * * ****
*
* *
**** * * *
* * ** * ** *
**** * * ******* * * *
0.2
* * ** *
******* *** * *
* * *** *
0
** * * * * ***
* * ** * ** ****
50
*
* * * *
-5
0.0
* *
0
* * *
1880 1920 1960 2000 50 100 200 50 100 200 1960 1980 2000 20 40 60 20 40 60
Germany : 1881-2010 Germany : 1956-2010
70
* *
10 20 30 40 50 60 70
6
15
60
*
* ***
3
GDP Growth Rate

GDP Growth Rate
4
****
10
* * * *
50
*
Debt/GDP
* * ** *
Debt/GDP
**** ** *
* * *
***** * *** *
2
**
2
**** ***
5
**** ** * * ** *
40
* * **
* *
* ** **** ***** * ** * * * *
** ***** ***
* ***** *
0
** ****** * *** *
*** * ** *** * **
0
* * ** *
30
1
*
2
* **
*
-2
**
-5
20
-4
0
0
* * *
1880 1940 2000 10 30 50 70 10 30 50 70 1960 1980 2000 20 30 40 50 60 70 20 30 40 50 60 70
Italy : 1881-2010 Italy : 1956-2010
100 120
0.0 0.5 1.0 1.5 2.0 2.5 3.0

10 20 30
0 1 2 3 4 5 6 7
Turning point Asy mptotic v alues *

* *
** * *
6
*
140
** **
*
GDP Growth Rate
***
GDP Growth Rate
* *
4
*
Debt/GDP
Debt/GDP
* *
100
* * **
80
*
****** * ** * ** *
****
*
* * * * ** *
*** * ** ** *** *
**** **** ****
2
* * * **
** ************
**
************ * *
**
-20 -10 0
** * * * **
* ** ****** *** ** * * **
60
* * *
** * *
60
* ***
0
* *
40
*
*
-2
*
20
*
*
1880 1940 2000 20 60 100 140 20 60 100 140 1960 1980 2000 40 60 80 120 0 20 60 100
Ireland : 1929-2010 Ireland : 1956-2010

2.0
** * *
30 40 50 60 70 80 90
* * *
* * *
20
* *
** **
80
* *
* *
** * ** * **
GDP Growth Rate

GDP Growth Rate
1.5
** * ** * *
5
5
* ** * * * *
15
* * * * * * *
* * **
Debt/GDP
Debt/GDP
* ********** ** * ** * *** ** * ** * * * **
60
*** *
** * * ** * * *
1.0
**** ** ** *
0
10
* * *
* ** *
40
*
0.5
5
-5
-5
* *
20
0.0
0
* *
1940 1980 20 40 60 80 20 40 60 80 1960 1980 2000 30 50 70 90 0 20 40 60 80
68
Figure 4. Regression kink results for Portugal, Spain, Sweden and Switzerland
Portugal : 1881-2010 Portugal : 1956-2010
70
50 60 70
15
10
*
10
40
* *
*
60
* **
10
*
* *
GDP Growth Rate

** *
GDP Growth Rate

8
* * *
30
** * **** ** * * * *
5
*
50
* * *
Debt/GDP
** * ** * **
Debt/GDP
******** * * * * * ***
** ****
5
*
6
*** ** * * ** ************ * * * *
* * * *
***** ** ********
* *
20 30 40
40
**
20
* * * *
*** * * ***** ***** *** * *
0
** * ** * * **
0
* * *
4
*
* ** * ** **** * * * *
30
*
-5
10
* * *
2
*
-5
20
* *
-10
*
0
* *
1880 1920 1960 2000 20 30 40 50 60 70 20 30 40 50 60 70 1960 1980 2000 20 30 40 50 60 70 20 30 40 50 60 70
Spain : 1881-2010 Spain : 1881-2010
* *
140
10
10
1.5
* *
10 20 30 40 50 60
4
* **
GDP Growth Rate
GDP Growth Rate

** *
* * * **
*** ** * * *
100
*
5
* *** ******
Debt/GDP
Debt/GDP
3
* * * **
1.0
*
** * ** * ** **** ** ** * *
5
* **
* * ** ***** * * * ** * *
* *
* * *
* * ** * ** * * * **
2
**
0
* *
**** * * * ** ** * * * *
60
* * ** *** *
* * ** * * ** * * *
0.5
* * * * * * *
* ** *
* * * * *
1
* *
0
-5
* *
20
*
*
0.0
0
* *
1880 1920 1960 2000 20 60 100 140 20 60 100 140 1960 1980 2000 10 30 50 10 30 50
Sw eden : 1881-2010
Sweden : 1956-2010
**
6
80
**
8 Turning point Asy mptotic v alues
8 10
** * *
80
* * * * *
*******
5
GDP Growth Rate

**
* * * * * ** *
GDP Growth Rate
****
4
*** ******* * * * ** *
6
** * ** * ** *** *
60
* **
Debt/GDP
******* * * * **
60
*** * **
Debt/GDP
***** ** * * * ** ** *
* * * *
** ** * *
6
* ** *
2
** * ** **
0
** * * ** * *
4
** *
* * * * * *
40
40
4
* * * * *
*
0
*** * * *
** *
2
2
-5
-2
20
* *
20
* *
0
*
0
*
1960 1980 2000 20 40 60 80 20 40 60 80
1880 1920 1960 2000 20 40 60 80 20 40 60 80
Sw itzerland : 1899-2010 Switzerland : 1956-2010
70
80
Turning point Asy mptotic v alues * * *

*
4
* *
10 15
30
4
60
* * ** *
GDP Growth Rate
**
** *
60
* * * *** *
GDP Growth Rate

* *
2
*
** * *
50
*
Debt/GDP
3
** * * **
Debt/GDP
20
0
** * ** * * * *
40
*
40
*
* * * * *** *
5
**
* * *
2
*
* ** ********* ** * ** *
-2
* *** ** ******* * * * *
0 5 10
**
30
0
** * ***** ** *
* * * ***
20
**** *
-4
** *
1
*
20
* *
-5
* **
-6
* *
0
10
0
*
1900 1940 1980 0 20 40 60 80 0 20 40 60 80
1960 1980 2000 10 20 30 40 50 60 70 10 20 30 40 50 60 70
69
Figure 5. Regression kink results for New Zealand, Greece, Netherlands and Norway
100 150 200 250 New Zealand : 1881-2010 New Zealand: 1956-2010
4
2.0
* *
10 15
20 30 40 50 60 70
*
GDP Growth Rate

* * *
1.5
GDP Growth Rate
5
*
* *** * * * * ** * *
*
Debt/GDP
****** * ***
Debt/GDP
* * * *
5
*
**** **
**** * * * ** * * * **
* * * ** * * * * ***
1.0
* *
2
** * **
**********
*** ** ** **** * * *
*** ******* * ***** * * * ** *** * *
0
* *
* * ** * *** ***** *
* ** **
* * * *
0.5
* *
-10 -5
*
1
** * ** * * *
50
-5
* * * * *
0.0
*
0
*
1880 1940 2000 50 150 250 50 150 250 1960 1980 2000 20 40 60 0 20 40 60
Greece: 1884-2010 Greece: 1956-2010
* *
100
20
*
0.0 0.5 1.0 1.5 2.0 2.5

10
200
* *
* * *
**
** ***
4
***
GDP Growth Rate

* *
GDP Growth Rate
*** **
* **
80
*************** *****
*
****
**
150
** *
**** * ** * * **
0
*** * ** ***
*
Debt/GDP
Debt/GDP
* * ***** * ** ** * *
5
*
3
* *
60
* * * ** *
* *** * * *** **
100
-20
* * * ** ***
2
* * *
*
40
*
0
** * *
* * *
50
1
-40
20
*
0
* *
1900 1960 50 150 50 150 1960 1980 2000 20 40 60 80 20 40 60 80
Netherlands : 1881-2010
Netherlands: 1956-2010
40 50
0.0 0.1 0.2 0.3 0.4 0.5 0.6
80
*
0.0 0.5 1.0 1.5 2.0 2.5

*
200
GDP Growth Rate
70
** *
GDP Growth Rate
* *
20 30
Debt/GDP
4
150
* **
Debt/GDP
* ** *
* ***
60
* * * *
* *
* * * * * ** *
2
* *
* * * **
***
100
50
*
* **
10
* * *** ** ** * *
0
* *
***
**** ************** ** *
** *
** ********** ****
*** *
40
0
***** ********* ******** * *

*
* * *
50
* * ** ** *
-2
*
-10
** *
1960 1980 2000 40 50 60 70 80 40 50 60 70 80
1880 1940 2000 50 100 150 200 50 100 150 200
Norway : 1881-2010
15

60
*
20
*
10
50
*
GDP Growth Rate
*
15
*
*** * *
Debt/GDP
*
5
** *
40
*** *
* *************** *** * ** *
**** ***** ** ** ** *** *** **
*** ************** * * *
10
**
0
********* ***** *
30
* *
-5
*
20
-10
*
10
**
0
1880 1940 2000 10 30 50 10 30 50
year Debt/GDP Threshold Parameter
70
Figure 6. 5-year lead economic growth and public debt over different periods (panel of 20 advanced countries)
1880-2008 1880-1913 1914-1945 1946-1970
60 60 60 12
40 40
40 8
20 20
20 4
GROWTH(t+5)
GROWTH(t+5)
0 0
GROWTH(t+5)
GROWTH(t+5)
-20 0 -20 0
-40 -40
-20 -4
-60 -60
-40 -8
-80 -80
-100 -60 -100 -12

0 40 80 120 160 200 240 280 0 40 80 120 160 200 240 0 50 100 150 200 250 0 40 80 120 160 200 240 280
DEBT(t) DEBT(t) DEBT(t) DEBT(t)

1971-1990 1991-2008 1950-2008 1971-2008
10 16 16 16
8 12 12 12
6
8 8 8
4
GROWTH(t+5)
GROWTH(t+5)
GROWTH(t+5)
GROWTH(t+5)
4 4 4
2
0 0 0
0
-4 -4 -4
-2
-4 -8 -8 -8
-6 -12 -12 -12

0 20 40 60 80 100 120 140 0 40 80 120 160 200 0 40 80 120 160 200 240 0 40 80 120 160 200
DEBT(t) DEBT(t) DEBT(t) DEBT(t)
Figure 7. 5-year lead economic growth and public debt over different periods (panel of 10-euro countries)
1880-2008 1880-1913 1914-1945 1946-1970
60 60 60 12
40 40
40
8
20 20
20
0 0
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
-20 0 -20
-40 0
-40
-20
-60 -60
-4
-40
-80 -80
-100 -60 -100 -8

0 40 80 120 160 200 240 0 40 80 120 160 200 240 0 40 80 120 160 200 240 0 40 80 120 160 200 240
Debt Debt Debt Debt

1971-1990 1991-2008 1950-2008 1971-2008
10 10.0 12 10.0
8 7.5 7.5
8
5.0 5.0
6
4
2.5 2.5
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
0.0 0 0.0
2
-2.5 -2.5
-4
0
-5.0 -5.0
-8
-2 -7.5 -7.5
-4 -10.0 -12 -10.0

0 20 40 60 80 100 120 140 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
Debt Debt Debt Debt
71
Appendix A. Tables
Table A.1. Public debt and economic growth correlations by countries and periods
Correlation of debt(t) and growth(t) Correlation of debt(t) and growth(t+5)
Countries 1950-2008 1950-1970 1971-1990 1991-2008 1971-2008 1950-2008 1950-1970 1971-1990 1991-2008 1971-2008
Australia 0.08 -0.25 0.26 0.14 0.13 -0.07 -0.11 -0.07 -0.44 -0.22
Austria -0.54*** 0.20 -0.28 0.15 -0.3 -0.51*** 0.12 -0.52 0.39 -0.38**
Belgium -0.40** -0.30 -0.21 -0.19 -0.28* -0.47*** -0.12 -0.72** -0.17 -0.44**
Canada -0.1 -0.02 -0.07 0.14 -0.10 -0.37** -0.27 -0.34 -0.61 -0.44**
Denmark -0.13 -0.26 0.08 0.25 0.08 -0.28** -0.18 -0.06 0.27 -0.06
France -0.44*** -0.25 -0.12 0.21 -0.28* -0.54*** -0.17 -0.49 0.31 -0.25
Germany -0.51*** -0.26 -0.29 -0.13 -0.19 -0.57*** 0.00 -0.21 -0.10 -0.30
Greece -0.30** 0.30 -0.14 0.15 0.04 -0.43** 0.01 -0.11 0.72** 0.17
Ireland 0.27** -0.22 0.13 0.41* 0.32** -0.22 0.16 -0.51 -0.31 -0.36*
Italy -0.70*** 0.09 0.00 0.25 -0.38** -0.66*** -0.22 -0.17 -0.16 -0.28
Japan -0.68*** -0.11 0.06 -0.03 -0.44** -0.75*** -0.05 -0.23 -0.24 -0.55**
Netherlands 0.06 -0.23 0.16 0.00 0.12 -0.21 -0.1 -0.41 -0.23 -0.21
New Zealand 0.11 0.22 -0.05 0.06 -0.08 -0.02 0.07 0.02 -0.40 -0.25
Norway -0.25* -0.18 -0.22 -0.19 -0.30 -0.47*** -0.22 -0.69** -0.74** -0.74***
Portugal -0.35** 0.21 -0.04 -0.28 -0.23 -0.32** 0.18 -0.26 -0.24 -0.21
Spain -0.29** -0.36 -0.14 0.54** -0.16 -0.50*** -0.3 -0.85** -0.26 -0.36*
Sweden -0.23* 0.04 0.16 0.01 0.11 -0.52*** -0.23 -0.28 -0.67* -0.43**
Switzerland -0.28** 0.37* -0.23 0.58** 0.04 -0.19 0.21 -0.03 0.73** -0.02
United Kingdom -0.04 -0.20 0.24 0.55** 0.36** -0.14 -0.3 -0.09 -0.58 -0.42
United States 0.01 0.11 0.06 -0.17 -0.09 -0.1 0.03 -0.18 -0.35 -0.21
Total Sample -0.25* -0.34 -0.02 -0.04 -0.07 -0.29** -0.24 -0.14 -0.23 -0.11
Significant Correlations 13/20 1/20 0/20 4/20 6/20 13/20 0/20 3/20 4/20 8/20
Negatives correlations 15/20 12/20 11/20 6/20 12/20 20/20 12/20 19/20 16/20 19/20
Note: Significant at 1% (***), 5% (**) and 10% (*).
72
Table A.2. Data heterogeneity tests across sections and periods

Test for Equality of means, medians and variances of GROWTH Categorized by values of DEBT
Equality of means Sample: 1881 2008 Sample: 1950 2008 Sample: 1971 2008 Sample: 1991 2008
Method df Value Prob. df Value Prob. df Value Prob. df Value Prob.
Anova F-test (2, 2248) 12.655 0.000 (4, 1119) 10.675 0.000 (3, 2.498 0.059 (3, 353) 2.365 0.071
Analysis of Variance 741)
Sum of Mean Sum of Mean Sum of Mean
Source of Variation df df Mean Sq. df Sum of Sq. df
Sq. Sq. Sq. Sq. Sq. Sq.
Between 2 470.52 235.26 4 281.74 70.44 3 34.26 11.42 3 25.42 8.47
Within 2248 41792.62 18.59 1119 7383.49 6.60 741 3387.23 4.57 353 1264.73 3.58
Total 2250 42263.14 18.78 1123 7665.23 6.83 744 3421.50 4.60 356 1290.15 3.62
Equality of medians Sample: 1881 2008 Sample: 1950 2008 Sample: 1971 2008 Sample: 1991 2008
Med. Chi-square 2 22.29 0.000 4 35.61 0.000 3 13.44 0.004 3 6.76 0.080
Adj. Med. Chi-square 2 21.67 0.000 4 32.55 0.000 3 11.14 0.011 3 5.11 0.164
Kruskal-Wallis 2 26.97 0.000 4 49.99 0.000 3 10.88 0.012 3 7.77 0.051
Kruskal-Wallis (tie-adj.) 2 26.97 0.000 4 49.99 0.000 3 10.88 0.012 3 7.77 0.051
Van der Waerden 2 25.71 0.000 4 45.27 0.000 3 9.10 0.028 3 7.24 0.065
Equality of variances Sample: 1881 2008 Sample: 1950 2008 Sample: 1971 2008 Sample: 1991 2008
Bartlett 2 143.47 0.000 4 34.49 0.000 3 22.05 0.000 3 5.82 0.121
Levene (2, 2248) 36.73 0.000 (4, 1119) 5.66 0.000 (3, 2.95 0.032 (3, 353) 0.36 0.781
Brown-Forsythe (2, 2248) 32.60 0.000 (4, 1119) 5.63 0.000 741)
(3, 2.91 0.034 (3, 353) 0.38 0.771
Bartlett weighted standard deviation 4.31 2.57 741) 2.14 1.89
Category statistics of debt
Sample: 1881 2008 Sample: 1950 2008 Sample: 1971 2008 Sample: 1991 2008
Std. Std. Std. Std.
Debt Std. Debt Std. Std. Std.
Count Mean Err. of Count Mean Err. of Count Mean Err. Count Mean Err. of
Dev. Dev. Dev. Dev.
Mean Mean of Mean
[0, 100) 1928 2.256 4.062 0.093 [0, 50) 612 3.156 2.828 0.114 353 Mean
2.080 2.372 0.126 95 1.678 2.107 0.216
[100, 300 1.023 4.582 0.265 [50, 100) 434 2.278 2.235 0.107 334 2.204 1.946 0.106 211 2.025 1.847 0.127
200)
[200, 23 3.918 12.991 2.709 [100, 150) 64 1.790 2.198 0.275 50 1.453 1.668 0.236 43 1.358 1.700 0.259
300) [150, 250) 13 1.578 1.534 0.425 8 0.993 1.162 0.411 8 0.993 1.162 0.411
All 2251 2.109 4.334 0.091 All 1124 2.718 2.613 0.078 745 2.082 2.144 0.079 357 1.829 1.904 0.101
Notes: df is degrees of freedom. Prob. is the probability. Count is the number of observations. Sum of Sq. is the sum of square. Mean Sq. is the mean of square.
Std. Dev. is the standard deviation. Std. Err. of mean is the standard error of mean.
73
Table A.3. Instruments correlations of public debt ratio and economic growth
1950-2008 1950-1970 1971-1990 1991-2008 1971-2008
Growth Debt Growth Debt Growth Debt Growth Debt Growth Debt
Instrument
(t+5) (t) (t+5) (t) (t+5) (t) (t+5) (t) (t+5) (t)
GC -0.09 0.24 -0.12 0.23 0.05 0.27** 0.14 0.14 0.07 0.14
GC(-1) -0.09 0.27 -0.10 0.24 0.04 0.32** 0.16 0.15 0.07 0.17
GC(-2) -0.09 0.29 -0.11 0.23 0.03 0.35** 0.18 0.15 0.07 0.20
IM 0.08 - 0.10 -0.36 -0.09 -0.40** -0.04 -0.39 -0.08 -0.43**
IM(-1) 0.10 0.48**
- 0.13 -0.43* -0.07 -0.37** -0.03 -0.38 -0.06 -0.41**
IM(-2) 0.12 0.48**
- 0.15 -0.43* -0.04 -0.38** -0.02 -0.37 -0.04 -0.41**
EX -0.14 0.48**
0.54** -0.19 0.55** 0.04 0.55*** -0.02 0.23 0.02 0.46**
EX(-1) -0.16 0.54** -0.22 0.57** 0.01 0.54*** -0.05 0.23 -0.01 0.46**
EX(-2) -0.18 0.54** -0.25 0.57** -0.03 0.54*** -0.06 0.23 -0.03 0.46**
GCF -0.08 -0.11 -0.15 -0.02 0.00 -0.19 0.04 0.08 0.00 -0.19
GCF(-1) -0.07 -0.08 -0.13 0.02 0.04 -0.21 0.06 0.08 0.04 -0.20
GCF(-2) -0.09 -0.05 -0.13 0.04 0.03 -0.21 0.02 0.06 0.02 -0.18
ODR -0.23* 0.32 -0.17 -0.17 -0.01 0.02 -0.08 0.32 -0.02 0.28
ODR(-1) -0.23* 0.32 -0.18 -0.15 0.00 0.02 -0.07 0.30 -0.02 0.28
ODR(-2) -0.23* 0.33 -0.18 -0.13 0.00 0.04 -0.07 0.27 -0.01 0.27
Note: Significant at 1% (***), 5% (**) and 10% (*).
Table A.4. Fixed effects generated across countries and periods
Fixed effects for the quadratic form specification: the sample of 20 OECD countries
Country 1950-2008 1950-1970 1971-2008 1971-1990 1991-2008
Australia -1.78 -2.57 0.39 -0.65 15.10
Austria -0.56 3.74 0.08 -0.05 -1.78
Belgium 2.79 -4.60 1.29 1.98 -4.19
Canada 0.45 -5.51 -0.50 -0.43 -5.90
Denmark -1.22 2.08 -0.09 0.00 -0.93
France -1.18 0.86 -0.31 -0.53 0.68
Germany -1.44 2.14 -0.48 -0.47 0.21
Greece 1.04 4.70 0.59 -0.46 -4.92
Ireland 1.11 -3.46 1.13 2.14 1.11
Italy 1.46 0.29 0.19 1.00 -6.57
Japan 2.03 8.80 0.85 0.88 -2.14
Netherlands 0.40 -4.78 -0.50 -0.27 -1.92
New Zealand -1.10 -6.50 -1.17 -1.10 4.62
Norway -0.84 0.45 0.43 0.49 6.82
Portugal -0.22 4.00 0.25 0.65 0.79
Spain -0.12 4.60 0.37 0.00 0.40
Sweden -0.57 0.88 -0.46 -0.99 -2.77
Swiss -2.03 -0.73 -1.18 -1.45 0.01
United Kingdom 2.05 1.61 -0.25 -0.19 3.27
United States -0.78 -5.06 -0.55 -0.37 -1.83
Fixed effects for the quadratic form specification: the sample of 10 Euro countries
Austria -0.80 8.13 -0.26 -0.23 -0.27
Belgium 1.78 -12.25 1.44 1.02 -6.01
France -1.35 1.05 -0.77 -0.29 5.35
Germany -1.53 4.48 -0.87 -0.58 4.48
Greece 0.60 7.66 0.41 -0.71 -7.76
Ireland 1.22 -9.45 0.97 1.13 4.67
Italy 0.99 -1.88 0.24 -0.12 -9.32
Netherlands 0.44 -12.01 -0.70 -1.10 -0.50
Portugal -0.35 7.03 -0.13 0.43 5.32
Spain -0.40 7.59 -0.08 0.27 4.00
74
Table A.5. Sensitivity of debt thresholds to government expenses and government effectiveness.
High Government effectiveness (13 countries): Finland, Denmark, Sweden, Norway, Netherlands, Canada, New Zealand,
Austria, Australia, United Kingdom, Belgium, Germany and United States.
1951-1970 1971-1990 1991-2008 1971-2008 1951-2008
Model Without Fixed Effect NS NS 80.7*** 73.8*** 61.2***a
Model With Fixed Effect 114.7** NS 95.4* 60.8** 92.8***a
Moderate Government effectiveness (7 countries): Ireland, France, Japan, Spain, Portugal, Greece and Italy.
Model Without Fixed Effect 40.3** 41.6** NS 49.3* NS
Model With Fixed Effect NS NS 102.5*** NS NS
High Government Final Consumption as % of GDP (9 countries): Sweden, Denmark, Netherlands, France, Canada, Belgium,
Germany, United Kingdom and Norway.
Model Without Fixed Effect NS NS NS 72.2* 61.4***a
Model With Fixed Effect NS NS NS 52.3** 101.4***a
Moderate Government Final Consumption as % of GDP (11 countries): New Zealand, Italy, Ireland, Austria, Australia, United
States, Greece, Portugal, Japan, Spain and Switzerland.
Model Without Fixed Effect 60.7**a NS NS 67.5*** NS
Model With Fixed Effect NS NS 91.5*** 58.4* NS
High Military Expenditures as % of GDP (9 countries): Unites States, United Kingdom, France, Greece, Portugal, Norway,
Netherlands, Sweden and Australia.
Model Without Fixed Effect NS NS NS 45.2** 122.0*a
Model With Fixed Effect NS NS 65.3** 51.4** 118.9***a
Moderate Military Expenditures as % of GDP (11 countries): Belgium, Germany, Italy, New Zealand, Denmark, Spain,
Canada, Switzerland, Ireland, Austria and Japan.
Model Without Fixed Effect 46.0*a NS 89.3** NS 51.1***
Model With Fixed Effect NS NS 114.2** NS NS
High Government Transfers and Subsidies as % of GDP (7 countries): Belgium, Switzerland, Netherlands, Canada, Sweden,
United States and Japan.
Model Without Fixed Effect 53.4*** 73.0* 93.0*** 81.5*** NS
Model With Fixed Effect NS NS NS 71.6*** NS
Moderate Government Transfers and Subsidies as % of GDP (13 countries):
1951-1970 1971-1990 1991-2008 1971-2008 1951-2008
Model Without Fixed Effect NS NS NS NS NS
Model With Fixed Effect 126.7*** 31.2* NS NS NS
Notes: significant at 1% (***), 5% (**) and 10% (*). NS is Non-Significant. a: The form of Growth-debt link is
convex meaning that the debt affects negatively growth for debt ratios below this threshold and positively above it.
Table A.6. Tests for cross-section dependence for the panel fixed effect model (Least Squares method)
Residual Cross-Section Dependence Test for the 20 OECD countries sample
Degrees of freedom = 190 1950-2008 1950-1970 1971-2008 1971-1990 1991-2008
Test Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob.
Breusch-Pagan LM 1362.7 0.000 474.0 0.000 855.2 0.000 539.0 0.000 671.4 0.000
Pesaran scaled LM 60.2 0.000 14.6 0.000 34.1 0.000 17.9 0.000 24.7 0.000
Bias-corrected scaled LM 60.0 0.000 14.1 0.000 33.8 0.000 17.4 0.000 23.9 0.000
Pesaran CD 32.3 0.000 15.9 0.000 23.6 0.000 17.2 0.000 22.2 0.000
Residual Cross-Section Dependence Test for the 10 Euro countries sample
Degrees of freedom = 45 1950-2008 1950-1970 1971-2008 1971-1990 1991-2008
Test Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob.
Breusch-Pagan LM 450.3 0.000 145.6 0.000 320.8 0.000 204.9 0.000 211.0 0.000
Pesaran scaled LM 42.7 0.000 10.6 0.000 29.1 0.000 16.9 0.000 17.5 0.000
Bias-corrected scaled LM 42.6 0.000 10.4 0.000 28.9 0.000 16.6 0.000 17.1 0.000
Pesaran CD 18.6 0.000 8.6 0.000 16.6 0.000 13.3 0.000 12.8 0.000
Notes: Null hypothesis: No cross-section dependence (correlation) in residuals. Test employs centered correlations
computed from pairwise samples
75
Appendix B. Individual scatter plots for public debt and economic growth
Figure B.1. 5-year lead economic growth and current public debt, 1880 and 2008
Australia Austria Belgium Canada
20 40 20 20
10
10 0 10
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0
0 -40 0
-10
-10 -80 -10

-20
-20 -120 -30 -20

0 40 80 120 160 0 20 40 60 80 100 0 50 100 150 200 250 0 40 80 120 160
DEBT DEBT DEBT DEBT
Denmark France Germany Greece
20 60 20 80
40 0
10 40
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
20 -20
0 0
0 -40
-10 -40
-20 -60
-20 -40 -80 -80

0 20 40 60 80 0 50 100 150 200 250 0 20 40 60 80 0 50 100 150 200 250
DEBT DEBT DEBT DEBT
Ireland Italy Japan Netherlands
10 30 20 60
20 40
0
5
10 20
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
-20
0 0 0
-40
-10 -20
-5
-20 -60 -40
-10 -30 -80 -60

0 20 40 60 80 100 0 40 80 120 160 200 0 50 100 150 200 250 0 50 100 150 200 250
DEBT DEBT DEBT DEBT
New Zealand Norway Portugal Spain
20 20 20 20
10
10 10 10
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0
0 0 0
-10
-10 -10 -10

-20
-20 -20 -20 -30

0 50 100 150 200 250 0 20 40 60 80 0 20 40 60 80 0 50 100 150 200
DEBT DEBT DEBT DEBT
Sweden Switzerland United Kingdom United States
10 20 10 20
5 10
5 10
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 0
0 0
-5 -10
-5 -10
-10 -20
-10 -20 -15 -30

0 20 40 60 80 100 0 20 40 60 80 0 50 100 150 200 250 300 0 40 80 120 160
DEBT DEBT DEBT DEBT
76
20 10 20 10
5
10 5
10
0
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 0
0 -5
-10 -5
-10
-10
-15 -20 -10
-20 -20 -30 -15

115 120 125 130 135 140 60 70 80 90 100 30 35 40 45 50 55 15 20 25 30 35 40 45 50
DEBT DEBT DEBT DEBT
15 20 5 80
10 10 0
40
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
5 0 -5
0
0 -10 -10
-40
-5 -20 -15
-10 -30 -20 -80

14 16 18 20 22 24 26 60 70 80 90 100 110 120 25 30 35 40 45 50 40 80 120 160 200 240
DEBT DEBT DEBT DEBT
Italy Japan Netherlands New Zealand
12 20 30 12
15 8
8
20
10 4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
5 10 0
0
0 -4
0
-4 -5 -8
-8 -10 -10 -12

70 80 90 100 110 120 130 0 20 40 60 80 60 70 80 90 100 110 90 100 110 120 130 140
DEBT DEBT DEBT DEBT
Norway Portugal Spain Sweden
16 8 10 10
12
4 5 5
8
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
0 0 0
0
-4
-4 -5 -5
-8
-12 -8 -10 -10

10 15 20 25 30 35 48 52 56 60 64 68 60 80 100 120 140 160 180 14 16 18 20 22
DEBT DEBT DEBT DEBT
Switzerland United Kingdom United States
15 8 12
8
10 4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
5 0
0
0 -4
-4
-5 -8 -8
-10 -12 -12

2.4 2.8 3.2 3.6 4.0 4.4 20 30 40 50 60 70 0 4 8 12 16 20
DEBT DEBT DEBT
77
12 40 20 20
8
0 10 10
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 -40 0 0
-4
-80 -10 -10
-8
-12 -120 -20 -20

50 60 70 80 90 100 15 20 25 30 35 40 45 50 100 150 200 250 0 40 80 120 160
DEBT DEBT DEBT DEBT
20 60 20 40
30
40 0
10
20
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
20 -20
0 10
0 -40
0
-10
-20 -60 -10
-20 -40 -80 -20

5 10 15 20 25 120 140 160 180 200 220 240 0 5 10 15 20 25 70 80 90 100 110
DEBT DEBT DEBT DEBT
6 30 20 60
20 40
4 0
10 20
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 -20
0 0
0 -40
-10 -20
-2 -20 -60 -40
-4 -30 -80 -60

10 15 20 25 30 35 40 60 80 100 120 140 160 180 0 50 100 150 200 250 40 60 80 100 120 140
DEBT DEBT DEBT DEBT
20 20 20 10
10 10 10 0
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 0 0 -10
-10 -10 -10 -20
-20 -20 -20 -30

100 150 200 250 0 10 20 30 40 50 20 30 40 50 60 70 20 30 40 50 60 70
DEBT DEBT DEBT DEBT
10 20 10 20
5 10
5 10
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 0
0 0
-5 -10
-5 -10
-10 -20
-10 -20 -15 -30

10 15 20 25 30 20 30 40 50 60 70 80 0 50 100 150 200 250 0 40 80 120
DEBT DEBT DEBT DEBT
78
6 12 8 8
6
6
4 8
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
2 4 2
2
0
0 0
0 -2
-2 -4 -2 -4
20 40 60 80 100 8 12 16 20 24 28 32 36 40 60 80 100 120 140 40 60 80 100 120 140 160
DEBT DEBT DEBT DEBT
10 6 12 15
8
4 8 10
6
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2 4 5
2
0 0 0
0
-2 -2 -4 -5
5 10 15 20 25 30 35 10 20 30 40 50 16 18 20 22 24 8 12 16 20 24 28
DEBT DEBT DEBT DEBT
8 12 12 8
6 6
8 8
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 4
4 4
2 2
0 0
0 0
-2 -4 -4 -2
20 30 40 50 60 70 24 28 32 36 40 0 10 20 30 40 50 60 40 80 120 160 200 240
DEBT DEBT DEBT DEBT
12 6 12 12
8
8
4 8
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
0 2 4
0
-4
0 0
-8 -4
-12 -2 -8 -4
40 60 80 100 120 140 160 10 20 30 40 50 60 14 16 18 20 22 24 10 20 30 40 50
DEBT DEBT DEBT DEBT
6 8 8 6
6 4
4 4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2
2 0
2 0
0 -4
0 -2
-2 -8 -2 -4
15 20 25 30 35 40 45 0 20 40 60 80 50 100 150 200 250 300 20 40 60 80 100 120 140
DEBT DEBT DEBT DEBT
79
6 6 6 6
4
4
4 4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
2 2 0
0
-2
0 0
-2
-4
-4 -2 -2 -6
16 20 24 28 32 10 20 30 40 50 60 20 40 60 80 100 120 140 40 50 60 70 80
DEBT DEBT DEBT DEBT
8 4 6 6
3
6 4 4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2 2
1
2 0 0
0
0 -2 -2
-1
-2 -2 -4 -4
0 20 40 60 80 12 16 20 24 28 32 36 15 20 25 30 35 40 45 20 30 40 50 60 70 80
DEBT DEBT DEBT DEBT
10 8 6 6
8
6
4 4
6
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
4 2 2
2
2
0 0
0
0
-2 -2 -2 -2
20 40 60 80 100 40 50 60 70 80 90 100 0 20 40 60 80 30 40 50 60 70 80
DEBT DEBT DEBT DEBT
6 6 8 6
4 6
4 4
2 4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 2 2 2
-2 0
0 0
-4 -2
-6 -2 -4 -2
40 50 60 70 80 24 28 32 36 40 44 48 52 10 20 30 40 50 60 0 10 20 30 40 50
DEBT DEBT DEBT DEBT
4 6 6 8
6
4
2 4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
0 2 2
0
0
-2 0
-2
-2
-4 -4 -2 -4
20 30 40 50 60 70 80 30 35 40 45 50 35 40 45 50 55 60 65 70 30 40 50 60 70
DEBT DEBT DEBT DEBT
80
4.0 4 4 5
3.6 4
3 3
3.2
3
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2.8
2 2 2
2.4
1
2.0
1 1
1.6 0
1.2 0 0 -1
8 12 16 20 24 28 32 56 60 64 68 72 80 90 100 110 120 130 140 60 70 80 90 100 110
DEBT DEBT DEBT DEBT
4 4 4 6
3
3 3
4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 2
1 2
1 1
0
0
-1 0 0
-2 -1 -1 -2
20 30 40 50 60 70 30 40 50 60 70 30 40 50 60 70 70 80 90 100 110 120
DEBT DEBT DEBT DEBT
10 4 3 4
3 2
3
5
2 1
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
0 1 0
1
0 -1
-5
-1 -2 0
-10 -2 -3 -1
20 40 60 80 100 95 100 105 110 115 120 125 40 80 120 160 200 40 50 60 70 80
DEBT DEBT DEBT DEBT
6 5 6 5
4 4
4
4
3 3
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
2 2 2
0
1 1
0
-2 0 0
-4 -1 -2 -1
10 20 30 40 50 60 70 20 30 40 50 60 70 48 52 56 60 64 68 72 30 40 50 60 70
DEBT DEBT DEBT DEBT
5 4 6 4
4 3
3
4
3 2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
2 2 1
1
1 0
0
0 0 -1
-1 -1 -2 -2
30 40 50 60 70 80 90 30 40 50 60 70 80 35 40 45 50 55 60 50 55 60 65 70 75
DEBT DEBT DEBT DEBT
81
6 6 6 6
4
4
4 4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
2 2 0
0
-2
0 0
-2 -4
-4 -2 -2 -6
5 10 15 20 25 30 35 10 20 30 40 50 60 70 20 40 60 80 100 120 140 40 60 80 100 120
DEBT DEBT DEBT DEBT
8 4 6 6
3
6 4 4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2 2
1
2 0 0
0
0 -1 -2 -2
-2 -2 -4 -4
0 20 40 60 80 10 20 30 40 50 60 70 10 20 30 40 50 60 70 20 40 60 80 100 120
DEBT DEBT DEBT DEBT
10 8 6 6
6
4
5 4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
0 2 2
0
0
-5 0
-2 -2
-10 -4 -4 -2
20 40 60 80 100 20 40 60 80 100 120 140 0 50 100 150 200 30 40 50 60 70 80
DEBT DEBT DEBT DEBT
6 6 8 6
4
4 4 4
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
0 2 0 2
-2
0 -4 0
-4
-6 -2 -8 -2
0 20 40 60 80 20 30 40 50 60 70 0 20 40 60 80 0 20 40 60 80
DEBT DEBT DEBT DEBT
6 8 6 8
6
4
4 4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2
0 2 2
0
0
-4 0
-2 -2
-4 -8 -2 -4
20 40 60 80 100 20 30 40 50 60 70 80 30 40 50 60 70 80 30 40 50 60 70 80
DEBT DEBT DEBT DEBT
82
6 8 8 6
4
4 6 6
2
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 4 4
0
0 2 2
-2
-2 0 0 -4
-4 -2 -2 -6
0 10 20 30 40 50 0 20 40 60 80 20 40 60 80 100 120 140 40 60 80 100 120
DEBT DEBT DEBT DEBT
10 6 8 15
8
4 10
6 4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2 5
2 0
0 0
0
-2 -2 -4 -5
0 20 40 60 80 10 20 30 40 50 60 70 10 20 30 40 50 60 70 0 40 80 120
DEBT DEBT DEBT DEBT
10 8 12 8
6
6
5 8
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
0 2 4
2
0
-5 0
-2 0
-10 -4 -4 -2
20 40 60 80 100 20 40 60 80 100 120 140 0 50 100 150 200 30 40 50 60 70 80 90
DEBT DEBT DEBT DEBT
12 6 12 12
8 8
4 8
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 4
2 4
0 0
0 0
-4 -4
-8 -2 -8 -4
0 20 40 60 80 10 20 30 40 50 60 70 0 20 40 60 80 0 20 40 60 80
DEBT DEBT DEBT DEBT
6 8 8 8
6
4 6
4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 4
0 2
0 2
0
-4
-2 0 -2
-4 -8 -2 -4
0 20 40 60 80 100 0 20 40 60 80 0 40 80 120 160 30 40 50 60 70 80
DEBT DEBT DEBT DEBT
83
6 12 8 8
4 6
8 4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 4
4 0
0 2
0 -4
-2 0
-4 -4 -2 -8
0 20 40 60 80 100 0 10 20 30 40 50 60 20 40 60 80 100 120 140 40 80 120 160
DEBT DEBT DEBT DEBT
10 6 12 15
8
4 8 10
6
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4 2 4 5
2
0 0 0
0
-2 -2 -4 -5
0 20 40 60 80 10 20 30 40 50 15 20 25 30 35 40 45 0 20 40 60 80
DEBT DEBT DEBT DEBT
10 12 12 8
8
6
8 8
6
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
4 4 4
2
2
0 0
0 0
-2 -4 -4 -2
20 40 60 80 100 20 40 60 80 100 0 20 40 60 80 0 50 100 150 200 250
DEBT DEBT DEBT DEBT
12 6 12 12
8
8
4 8
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
4
0 2 4
0
-4
0 0
-8 -4
-12 -2 -8 -4
0 40 80 120 160 10 20 30 40 50 60 10 20 30 40 50 60 0 10 20 30 40 50
DEBT DEBT DEBT DEBT
6 8 8 8
6
4 6
4
4
GROWTH(5)
GROWTH(5)
GROWTH(5)
GROWTH(5)
2 4
0 2
0 2
0
-4
-2 0 -2
-4 -8 -2 -4
0 20 40 60 80 0 20 40 60 80 0 50 100 150 200 250 300 20 40 60 80 100 120 140
DEBT DEBT DEBT DEBT
84
Appendix C. Hansen (2017) algorithms for the regression kink model
Algorithm 1: Testing for a Regression Kink with an Unknown Threshold.

1. Generate n iid draws 𝑢𝑡 from the N(0,1) distribution.
2. Set 𝑦𝑡∗ = 𝑒̃𝑡 𝑢𝑡 where 𝑒̃𝑡 are the OLS residuals from the fitted linear model (8).
3. Estimate the linear regression model (8) and the regression kink model (2), and compute
̃ ∗2 −𝜎
𝑛(𝜎 ̂ ∗2 )
the error variance estimates 𝜎̃ ∗2 and 𝜎̂ ∗2 and the F statistic calculated as: 𝑇𝑛∗ = .
̂ ∗2
𝜎
∗ (𝑏),
4. Repeat this B times to obtain a sample of simulated F statistic {𝑇𝑛 𝑏 = 1. . 𝐵}.
5. Compute the p-value as the percentage of simulated F statistics, which exceed the actual
value: 𝑝𝑛 = 𝐵1 ∑𝐵𝑏=1 1( 𝑇𝑛∗ (𝑏) ≥ 𝐹𝑛 ),
6. Compute the level α critical value 𝑐𝛼 as the empirical (1 − 𝛼) of the simulated F statistics
{𝑇𝑛∗ (𝑖), 𝑖 = 1. . 𝐵}.
7. Reject 𝐻0(𝛽1 =𝛽2 ) in favor of 𝐻1 (𝛽1 ≠𝛽2 ) at significance α if 𝑝𝑛 < α, or equivalently if 𝑇𝑛 >
𝑐𝛼 .
The number of bootstrap replications B should set sufficiently large to ensure accuracy of
the p-value. We keep the number B=10.000 in our case as reported by Hansen (2017). We
also use Γ = [min(𝑑𝑡 ) , max(𝑑𝑡 )] for each country and a grid search with the increments
of 1. The number of grid points is then: λ = max(𝑑𝑡 ) − min(𝑑𝑡 ) + 1.
Algorithm 2: Wild Bootstrap confidence intervals for parameters

1. Generate n iid draws 𝑢𝑡 from the N(0,1) distribution.
2. Set 𝑒𝑡∗ = 𝑒̂𝑡 𝑢𝑡 where 𝑒̂𝑡 are the OLS residuals from the fitted regression kink model (2).
3. Set 𝑔𝑡∗ = 𝛽̂′ 𝑥𝑡 (𝛾̂) + 𝑒𝑡∗ , where (𝛽̂, 𝛾̂) are the Least-Square estimates.
4. Using the observations (𝑔𝑡∗ , 𝑔𝑡−1 , 𝑑𝑡−1 ), estimate the regression kink model (2), parameter
estimates (𝛽 ̂∗ ) and 𝜎̂ ∗2 = 1 ∑𝑛𝑡=1 𝑒̂𝑡∗2 where 𝑒̂𝑡∗ = 𝑔𝑡∗ − 𝛽̂∗′ 𝑥𝑡 (𝛾̂ ∗ ).
̂∗ , 𝛾
𝑛
̂ ∗2 (𝛾
𝑛(𝜎 ̂ ∗2 )
̂ )−𝜎 1
5. Calculate the F-statistic for γ: 𝐹𝑛∗ (𝛾̂) = ; where 𝜎̂ ∗2 = ∑𝑛𝑡=1 𝑒̂𝑡∗2 (𝛾̂) and
̂ ∗2
𝜎 𝑛
∗ ∗ ̂ ∗′
𝑒̂𝑡 (𝛾̂) = 𝑔𝑡 − 𝛽 (𝛾̂)𝑥𝑡 (𝛾̂).
6. Repeat this B times to obtain a sample of simulated coefficient estimates (𝛽 ̂∗ , 𝛾
̂∗ ) and F
statistics 𝐹𝑛∗ .
7. Create (1 − 𝛼) bootstrap confidence intervals for the coefficients 𝛽 = (𝛽1 , 𝛽2 , 𝛽3 ) by the
symmetric percentile method: the coefficients plus and minus the (1 − 𝛼) quantile of the
absolute centered estimate bootstrap: for each coefficient 𝛽𝑖 ; 𝑖 = 1. .3, the interval is 𝛽̂𝑖 ±
𝑞1∗ where 𝑞1∗ is the (1 − 𝛼) quintile of |𝛽𝑖∗ − 𝛽̂𝑖 |.
∗
8. Calculate the (1 − 𝛼) quantile 𝑐1−𝛼 of the simulated F statistics 𝐹𝑛∗ .
9. Create (1 − 𝛼) bootstrap confidence interval for γ as the set of γ for which the empirical F
∗ ∗
statistics 𝐹𝑛 (𝛾) are smaller than the bootstrap critical value 𝑐1−𝛼 : 𝐶𝛾∗ = {𝛾: 𝐹𝑛 (𝛾) ≤ 𝑐1−𝛼 }
85
Chapter 2
Short-term effects of public debt on growth:
The spending multiplier pass-through
86
Chapter 2. Short-term effects of public debt on growth: The spending multiplier pass-through
Chapter 2
Short-term effects of public debt on growth: The spending multiplier

pass-through
Abstract: We examine the relationship between public debt and economic growth in the short
term, through spending multipliers. We study the impact of public debt accumulation on the size
of the expenditure multipliers, as well as the effects of the business cycle. We adopt the structural
vector autoregressive (SVAR) methodology, running a panoply of bi-variate and tri-variate SVAR
models on quarterly data for a sample of 18 OECD countries. Furthermore, we estimate an SVAR
with six fiscal and monetary variables for the United States to highlight the channels through which
public debt affects economic growth. In all the above-mentioned SVAR models, we control for
the business cycle and the public debt movement effects. The results show that expenditure
multipliers are higher in times of recession than times of expansion, which is in line with the recent
findings about fiscal multipliers in advanced economies, being larger in recessions than expansion
periods. Moreover, controlling exogenously for public debt, the estimations revealed larger
spending multipliers in debt accumulation than in debt contraction periods, independent of the
business cycle effects. However, introducing endogenously the public debt ratio leads to higher
multipliers in recessions than in expansions. Moreover, the results do not support any tendency for
weak spending multipliers for the recent periods compared to older ones. Furthermore, the United
States SVAR shows that public debt crowds out private investment, leading to a lower growth rate
in times of expansion, while in times of recession the public debt effects on growth are positive.
This crowding-out effect may play pass-through to the expenditure multipliers, which could
explain, ceteris paribus, the weak size of spending multipliers, while the crowding-in effect in
times of recession leads to higher multipliers. The results also revealed that government
expenditure has a positive but short-lived impact on economic growth. Besides, the recession
period has persistent effects on the responses of variables, for which convergence with the long-
term path following the shock is achieved slower than with the expansion time. The policy
implication is that fiscal stimulus effects could take time to materialize in recessions, while such
effects could be short-lived in expansions, which is something that should be considered by policy-
makers in their spending decisions.
JEL classification: C30, E62, H50, H60.
Keywords: Business cycle, Dynamic multiplier, Fiscal position, Government expenditure, SVAR.
87
1. Introduction
At the height of the 2008 financial crisis, fiscal policy was revived as the main and almost sole
active government policy tool to cope with the financial crisis effects, after monetary policy was
constrained in many advanced countries by the zero lower bound (ZLB) interest rate or by
countries being affiliated to a monetary union, as is the case for eurozone countries (Romer, 2011).
The fiscal policy is solicited in two “distinct” but subsequent events. The first, advanced countries,
were urged to implement massive fiscal stimuli plans and bail-out programmes28 (in 2009–10)
designed to dampen the negative effects of the crisis on the private sectors and households at the
beginning of the crisis, and therefore boosting growth (OECD, 2009; ILO, 2011). Consequently,
this leads to exacerbating fiscal positions, with the deficit widening and a sharp increase in debt to
GDP ratios, which were already high in the pre-crisis year (IMF, 2009; Taylor, 2018). This issue
pushed policy-makers, especially in the eurozone, where the debt ratios climbed from the pre-crisis
level in almost all countries by 20 percentage points (European Commission, 2012), supported by
advisors and economists’ views from international institutions and think thank institutes
(International Monetary Fund, World Bank, European Central Bank, OECD, G20) and other
economists from academia, to reverse the track of the fiscal policy from stimuli programmes to
fiscal consolidation. The argument of those economists stands for the modest results obtained from
the assessment of fiscal stimuli programmes,29 as reported by many studies (Baldacci et al., 2009;
Freedman et al., 2010; Taylor, 2011; Coenen et al., 2012; Cogan and Taylor, 2012; Mian and Sufi,
2012; Phelps, 2018), and the likely negative impact of high public debt on economic growth, which
triggered a large open debate between economists.30
However, the prolonged negative impacts of the financial crisis, under the two consequent
programmes (fiscal stimulus and fiscal consolidation), have put the focus on another issue related
to the role of fiscal multipliers in those programmes, namely, trying to find answers to the weak
GDP recovery despite massive fiscal stimuli from one side and the persistent high debt ratios and
deficits despite fiscal consolidation from the other. This triggered a permanent influx of empirical
studies (re)-assessing the values of the multipliers and reviewing their determinants.
One research strand involves assessing the size of the multipliers due to positive fiscal shocks
corresponding to the stimuli programmes (an important list of sample studies and their main
28
Although these fiscal stimuli seem to be large in absolute size, approximately US$ 2 trillion in the G20 countries
(ILO, 2011), some authors qualified these fiscal stimuli by insufficient and timid programmes (Aizenman and
Pasricha, 2011; Stiglitz, 2018), especially compared to emerging countries in Asia and the Pacific, not including Japan
and the Republic of Korea, where the average stimuli programmes were around 9.1% of GDP compared to just 3.4%
of GDP in advanced economies (ILO, 2011).
29
The two important fiscal stimuli that were reviewed are the ARRA programme (the American Recovery and
Reinvestment Act), implemented in the United States, and the EERP programme (the European Economic Recovery
Plan), implemented in Europe. A very large list of studies, reproduced in Table 1.A. in the appendix, on fiscal stimulus
measures and their main findings from the ILO (2011), shows very mixed results, especially on the ARRA programme.
Among 47 studies and reports of this list, 22 reported mixed effects, 21 positive effects and 4 negative effects.
30
More details about this debate are presented in Chapter 1. See also Bentour (2018).
88
findings about the effects of fiscal stimuli around 2008–10 is reproduced in Table 1.A in the
appendix), while the other strand goes the opposite way, evaluating and designing strategies for
successful fiscal consolidation31 and austerity measures (Abbas et al., 2010; OECD, 2011, 2012;
Molnár, 2012; Cogan et al., 2013; Estevão and Samake, 2013; Blot et al., 2014a; Alesina et al.,
2015). In fiscal stimuli programmes, the fiscal instruments used are either related to the increase
in spending programmes (and fiscal transfers to households and bail-out programmes) or based on
tax cuts for households and investors to boost consumption and investment – or a mixture of the
two. However, in fiscal consolidation, which directly aims, ceteris paribus, to sustain and improve
the government’s fiscal position (reducing debt and deficit) by increasing revenue and
rationalizing expenditure, the course of the fiscal instruments is reversed, as in this case, where
taxes should be increased and expenditure cut.32
Since the 2008 financial crisis, a significant body of literature assessing fiscal multipliers has
flourished (Romer and Romer, 2010; Auerbach and Gorodnichenko, 2012, 2013; Delong and
Summers, 2012; Farhi and Werning, 2017; Ramey, 2011; Ramey and Zubairy, 2018).33 Although
the story of fiscal multipliers was remounted at the edge of the Keynesian era following the 1929
Great Depression, their empirical assessment has recently been extensively revived as a result of
the development of econometrics and statistics. Nevertheless, the renewed importance of fiscal
multipliers showed large discrepancies in their values (whether in absolute values or sometimes
even in algebraic signs), especially in the recent financial crisis (Ramey, 2018). These values are
time- and country-specific and even sensitive to the assessment method (Baum et al., 2012; Batini
et al., 2014).
Indeed, in explaining the values of fiscal multipliers, some authors have found those multipliers to
be sensitive to the business cycle. In particular, fiscal spending multipliers were revealed to be
larger in recessions than expansion periods. Auerbach and Gorodnichenko (2012, 2013) were the
first to emphasize the tendency of fiscal multipliers to be large in recessions (potentially reaching
more than 2) compared to expansions. Consequently, many other researchers confirmed their
results, differentiating between fiscal multipliers in recessions and expansions (Barro and Redlick,
2011; Parker, 2011;34 Corsetti et al., 2012; Caggiano et al., 2015; Fazzari et al., 2015; Glocker et
al., 2019). This also pushed some researchers who were not totally convinced, or with opposing
results, to refine their analysis, leading to finding out the vulnerability of fiscal multipliers to other
determinants and not necessarily conditioned by the state of the economy in the business cycle.
31
See the World of Work Report (2010) for a list of detailed consolidation and austerity measures for advanced
countries.
32
We should be cautious when the values of multipliers deducted from fiscal stimulus, for example, are used to draw
conclusions and policy advice on fiscal consolidations, and vice versa, as there is no revealed symmetry of effects in
the instrument changes: an increase/decrease in expenditure could have an effect that is not necessarily the same in
absolute value for a similar decrease/increase in expenditure. Few authors cared about this issue (see Section 2.4).
33
See, for example, Ramey (2018) for a recent and large literature review.
34
Parker (2011) built on earlier versions of Auerbach and Gorodnichenko (2012, 2013) (as did some other authors),
which are issued, respectively, as NBER working papers in 2010 (No. 16311) and 2011 (No. 17447).
89
Therefore, fiscal multipliers were revealed to be dependent on the fiscal position measured by the
level of debt ratios and deficits (Corsetti et al., 2013; Huidrom et al., 2016), on the monetary policy
stance (Hall, 2009), particularly the constrained monetary policy, either by the ZLB interest rate
(liquidity trap) or by the loss of monetary independence, as in the pegged exchange rate or
monetary union (Cogan et al., 2010; Christiano et al., 2011; Delong and Summers, 2012; Farhi
and Werning, 2017).
However, the recent works of Ramey (2018) and Ramey and Zubairy (2018) consider that
government spending multipliers are, on average, lower than unity. This contrasts with the
tendency of the post-2008 crisis researchers to confirm larger multipliers, especially in recessions.
This also aligns with the consensus on spending multipliers before the last recession, considered
to be weak, and that fiscal policy effects are short-lived (Coenen et al., 2012). Furthermore, the
debate about the impact of government debt on economic growth (studied in Chapter 1) is directly
related and assessed via the multiplier’s effects. In particular, fiscal policy effects, taking into
account the fiscal position of the economy, measured by the level of public debt and/or the fiscal
deficit, have been highly debated in the aftermath of the 2008 financial crisis (Boussard et al.,
2012; Corsetti et al., 2012; Blot et al. 2014b; Canzoneri et al., 2015; Bi et al., 2016; Huidrom et
al., 2016; Perdichizzi, 2017; Poghosyan, 2017; Afonso and Leal, 2018; Auerbach and
Gorodnichenko, 2017; Blanchard, 2019; Broner et al., 2019). Very recently, in a presidential
lecture of the American Economic Association, Blanchard (2019) triggered another wave of public
debt and growth debate related to the fiscal cost of high public debt, as well as its effect on welfare,
minimizing worries about the public debt costs for the American economy, as, in historical records,
the nominal interest rate has remained, on average (except for some small periods around the
1980s), below the nominal growth rate. These contrasting results about spending multipliers make
it interesting to reconsider studying fiscal multipliers and contributing to this unsettled debate.
The remainder of this chapter is as follows. Section 2 presents the literature review on fiscal
multipliers, focusing particularly on the fiscal spending multipliers and their determinants,
especially for highly indebted countries and constrained monetary policies. Section 3 presents a
methodology for assessing spending multipliers according to the way the public debt ratio is
evolving. Section 4 presents the empirical results. Section 5 concludes.
A fiscal multiplier is defined as the output change in response to an (exogenous) change in a fiscal
variable in reference to their baseline levels (Spilimbergo et al., 2009; Coenen et al., 2012).35 The
concept of multiplier is generally associated with the general theory of John Maynard Keynes
35
If 𝑌𝑡 and 𝑍𝑡 denote, respectively, the output and the fiscal instrument at time 𝑡, the fiscal multiplier is simply
𝑑𝑌
expressed as 𝑡 . Or, while the effects come with different lag times, the cumulative fiscal multiplier is expressed as
𝑑𝑍𝑡
𝑗=𝑛
∑𝑗=0 𝑑𝑌𝑡+𝑗
𝑗=𝑛 (Chinn, 2013).
∑𝑗=0 𝑑𝑍𝑡+𝑗
90
(1936).36 The idea behind fiscal stimulus is that the fiscal multiplier, as the measure of the policy
instrument effect, is de facto a Keynesian one, which means that the value of such a fiscal
multiplier is larger than unity, making it rewardable/beneficial to go for such fiscal stimulus. In
the Keynesian structural models, the simplest way to compute a spending multiplier is via the
marginal propensity to consume.37 The spending multiplier in the Keynesian framework decreases
with the marginal propensity to import, as well as the rise in interest rates and increases with the
rise of investment due to the expansion of GDP (the accelerator effects). In a vector autoregression
(VAR) approach, spending multipliers are determined using the impulse response function and the
methodology of identification proposed by Blanchard and Perotti (2002).
Despite a continuously growing body of empirical literature on fiscal multipliers in recent years,
these tend to bring more confusion than forming a consensus about the size of the fiscal multiplier.
There are many reasons why the size of the fiscal multiplier changes. Besides the proper
characteristics of the studied economy, which are obviously due to macroeconomic fundamentals,
as well as the institutional environment, the difference of methods and the accuracy of data make
an important contribution to these differences. This section surveys the main important
contributions of the literature on fiscal multiplier determinants, as well as the challenging issues
presented by the methods used to gauge fiscal multipliers.
Recent researchers have mainly been interested in explaining why the recovery slowed in many
advanced countries and fiscal consolidation is hurting many others. In this line of research, the
frontier is not clear. For some, fiscal consolidation hurts in times of crisis, as fiscal multipliers are
larger in recessions than expansions. If this is the case, a legitimate question to consider is why the
large size of these multipliers in a time of crisis did not help fiscal stimuli to recover, especially in
the eurozone countries. This may be because of the asymmetric effects of the two cases (fiscal
stimulus versus austerity). To our best knowledge, very few studies have undertaken the issue of
asymmetry. Our investigation of the recent empirical literature found only two papers with
contrasting results (Baum and Koester, 2011; Riera-Crichton et al., 2014), while some papers draw
conclusions without paying attention to the asymmetry issue (Ramey, 2018; Ramey and Zubairy,
2018). More details about these papers are provided in Section 2.5.
This also triggered much research exploring the factors determining fiscal multipliers and
concentrating on the economic and institutional fundamentals of advanced economies. Researchers
in this way studied the effects of fiscal position related to the level of debt and deficit ratios, the
exchange rate regime (monetary unions), the degree of openness, agents’ expectations (foresight
fiscal policy), the constrained monetary policy at the ZLB interest rate and hand-to-mouth
consumers, among other things. Some have also found contrasting results while trying to explain
36
Historically, according to Hegeland (1954), the concept of fiscal multiplier goes back at least to the “Tableau
Economique” of François Quesnay (1758), as mentioned by some authors (Mustea, 2015).
37
The government spending multiplier for a closed economy under a fixed interest rate is given by: 1/(1 − 𝑚𝑝𝑐)
with 𝑚𝑝𝑐 the marginal propensity to consume.
91
the reasons for the failure of fiscal stimuli to deliver a fast recovery. Figure 1 presents the debated
determinants impacting the fiscal expenditure multiplier and its relationship with public debt and
economic growth linkages.
Figure 1: Fiscal spending multiplier determinants and the link to government debt and GDP
growth
Primary deficit New borrowing

Government
spending
Debt/GDP Debt level ∆debt

ratio - Assessment’ methods
- Econometric issues
- Non-linearities
- Heterogeneity
- Data
- Spending composition
Spending multiplier - Fiscal shocks timing
Agents’ expectations:
- Ricardian agents
- Crowding effects
GDP
Business cycle:
- Recessions
- Expansions
Nature of the shocks:

GDP - Temporary
Growth - Permanent
Fiscal Position:
- High/Low debt ratio - Interest rate
- High/Low deficit ratio - Exchange rate regime
- Monetary policy (ZLB)
- Debt composition (foreign
debt share)
- Degree of openness
Debt sustainability
Legend:
Economic determinant Policy choice determinant Econometric issues determinant
Accounting relationship Two way/feedback effects Multiplier effects on GDP
Source: Author’s own construction.
92
In what follows, as the literature on fiscal multipliers is somehow very rich and large and still
evolving, with many different results, we try to highlight some important contributions without
pretending to cover all of the literature, particularly the recent works related to the determinants
impacting the size of the spending multipliers.
2.1. The state dependency of fiscal multipliers on the business cycle
Some works studied fiscal policy in the pre-crisis of 2008 and proved the linkages between fiscal
policy and state dependence. The more recent example is the paper of Tagkalakis (2008),38 who
found, for OECD countries, larger effects of fiscal policy in recessions than expansions, with more
emphasized effects in countries with less-developed consumer credit markets.39 However, since
the works of Auerbach and Gorodnichenko (2012), research assessing the effects of fiscal policy,
considering the state dependency of the economy, have flourished, especially in the period of the
2008 financial crisis. Among these are the works of Bachmann and Sims (2012), Batini et al.
(2012), Baum et al. (2012), Riera-Crichton et al. (2014), Caggiano et al. (2015), Canzoneri et al.
(2015). The main result of these papers is confirmation of the dependency of spending multipliers
on the business cycle, which is larger during recessions than expansions.
The contributions of Auerbach and Gorodnichenko (2012), which triggered a series of works
studying fiscal multipliers during recessions and expansions, use a regime-switching structural
vector autoregression (SVAR) methodology to assess fiscal multipliers in relation to the business
cycle. They find large size of spending multipliers in recessions than in expansions, making
expansionary fiscal policy more effective in times of recession than expansion. Moreover, at the
disaggregated level, expenditure shows large differences in fiscal multipliers, with military
spending having the largest multiplier. They also show that multipliers tend to increase once the
real-time predictions of fiscal variables are controlled.
Also, focusing on the United States case, Caggiano et al. (2015) estimate a non-linear VAR model
to assess fiscal multipliers. They show two important results related to fiscal spending multipliers.
First, fiscal spending multipliers are greater than 1 in recessions, and, second, they are not
necessarily different from, or larger than, those in expansions. The second result opposes the main
findings of the previous research, which confirmed that fiscal multipliers are larger in times of
recession than expansion. Another important result raised by the authors is related to the non-
linearity effects on fiscal spending multipliers, which seem to be emphasized in extreme events
manifested by deep recessions or strong expansionary periods. Auerbach and Gorodnichenko
(2013) extend the same methodology of Auerbach and Gorodnichenko (2012) to a panel of OECD
countries and confirm their earlier results on the average of studied countries.
38
The author used a panel of 19 OECD countries over the period 1970–2002 to assess the effects of fiscal policy
changes on private consumption during recessions and expansions.
39
This is explained by the presence of individuals facing binding liquidity constraints in a recession that will consume
the additional income generated by the fiscal stimulus.
93
However, considering the discrepancies between countries in terms of the structure and behaviour
of the economies, the fiscal adjustments and policy responses and private agents’ expectations,
and the impacts of all these factors on the multipliers’ size, the results of the papers studying a
single country or averaging a group of countries could not be transposed to other countries and
remain debatable.
In this regard, Batini et al. (2012) and Baum et al. (2012) also confirm that fiscal multipliers tend
to be larger in recessions than expansions, but importantly they differ substantially across groups
of advanced countries, calling for a fit of fiscal policies and country-by-country assessment of
fiscal multipliers. The two papers employ the same methodology of non-linear threshold VAR
(TVAR) model. The only difference is that the threshold is endogenously determined by the sign
of the output gap in the first paper, and output growth in the second paper. Data is split according
to the threshold that separates expansions (positive output gap/growth) and recessions (negative
output gap/growth), chosen to maximize the fit of the model and hence allowing different
regression slopes for the explanatory fiscal variable. Using the output gap as the threshold variable
is argued by the fact that excess capacities are available in the economy under a negative output
gap, which reduces the crowding-out effects of private investment in response to government
expenditure shocks. Besides, the share of credit-constrained households, adjusting spending in
response to a change in disposable income, is higher in recessions. Other studies use output growth
or its moving average (Auerbach and Gorodnichenko, 2012). Moreover, according to Bachmann
and Sims (2012), the role played by household and firm confidence in the transmission of fiscal
policy shocks into economic activity is significant, which emphasizes the evidence of country-
specific properties that should be considered when studying the effects of fiscal policy.
Canzoneri et al. (2015), using a model of costly financial intermediation based on Curdia and
Woodford (2016),40 provide evidence of strong state-dependent fiscal multipliers that can exceed
the value of 2 in times of recession and may fall below unity during times of expansion.
Furthermore, the size of the fiscal multiplier is inversely dependent on the size of the fiscal
stimulus, with a smaller amount of fiscal intervention leading to a larger size of fiscal multiplier,
and vice versa. According to the author, “The reason large fiscal interventions are less effective
than smaller ones is that the negative marginal wealth effect due to the higher tax liabilities is
increasing with the size of the fiscal intervention while the positive marginal effect on the
borrowers, from the reduction in the finance premium, is decreasing with the size of the fiscal
expansion” (Canzoneri et al., 2015). Using a regime-switching framework, Arin et al. (2015) also
confirmed large spending multipliers for the United States over the period 1949Q1–2006Q4 during
40
The model is a simple extension of the basic-representative-household new Keynesian model (as developed in
Woodford, 2003) of the monetary transmission mechanism, allowing for a time-varying wedge between the interest
rate available to households on their savings and the interest rate available to borrowers. This model introduces credit
frictions and financial intermediation for the allocation of resources due to the introduction of heterogeneity in the
spending opportunities currently available to different households.
.
94
economic slowdown, while tax multipliers seem to be larger during periods of economic
expansion.
The problem of large multipliers is challenging for fiscal consolidation and austerity measures
intending, in times of high public debt, to reduce the latter. In this regard, some papers focus on
studying the effects of fiscal consolidation on the public debt ratios, particularly in the short term.
For example, Eyraud and Weber (2013) analyse the short-term fiscal multipliers considered to be
the key linkages between instruments of fiscal consolidation, economic growth and public debt
reduction. They find that, for many advanced countries, fiscal short-term multipliers in the recent
financial crisis have been close to 1, judged to be larger than the average of the short-term
multipliers observed before the 2008 recession. With the crisis environment of constrained
monetary policy, constrained credit agents and depressed external demand, these are likely to raise
the debt ratio in the short term under fiscal consolidation, and this could be emphasized if financial
markets react negatively to this short-term behaviour of public debt.
Along the same lines, Egron (2018) estimate a threshold VAR for France, confirming the higher
value of spending multipliers in recessions than expansions, and therefore warning about the
dangerous effects of fiscal consolidation, particularly in the short term, leading to an increase,
rather than reduction, in the government debt to GDP ratio. Nevertheless, the above results should
be considered carefully with regard to the likely asymmetric effects of an increase versus a
decrease in fiscal instruments (more details on this point in Section 2.5).
Furthermore, using an SVAR model for several MENA countries, Bentour (2020) assessed
spending multipliers considering the oil price fluctuations. The spending multipliers found to be
sensitive to the oil price movements especially for oil exporting countries, being large (more than
2) in time of oil price decrease and weak in time of oil price increase.
With regard to the researchers tending to confirm larger multipliers in times of economic downturn
compared to economic booms, some researchers contrast these results (Barro and Redlick, 2011;
Ramey, 2011; Owyang et al., 2013). For example, Owyang et al. (2013) find multipliers to be
smaller and less than 1. The authors use a large constructed quarterly data set for the United States
(1980 to 2010) and Canada (1921 to 2011) and the unemployment rate as a measure of slack
considering thresholds of 6.5% and 7% for, respectively, the United States and Canada. For a linear
model (no threshold of unemployment considered), either in the United States or Canada, spending
multipliers are all below unity and slightly larger in the United States than in Canada. However,
in periods of high unemployment (period of slack), multipliers are slightly higher than those in
Canada and lower than those in low unemployment rates. For the United States, spending
multipliers are always less than 1 and comparable across all regimes.
These clashing results about fiscal multipliers have pushed other economists to dig deeper and
control for features related to the economic and institutional regimes of the countries, such as fiscal
position, monetary policy stance and exchange rate regimes.
95
2.2. Fiscal multipliers’ dependency on the fiscal position
Since the first wave of studies triggered by the public debt and economic growth threshold idea of
Reinhart and Rogoff (2010), fiscal policy effects have also been revised distinguishing the
presence of high debt and deficit impacts on fiscal multipliers. Until now, the results of these
studies have continued to fuel the debate about such a subject.
Consequently, using a panel of 17 OECD countries (Australia, Austria, Belgium, Canada,

Denmark, Finland, France, Ireland, Italy, Japan, The Netherlands, Norway, Portugal, Spain,
Sweden, the United Kingdom and the United States), Corsetti et al. (2012) find that output and
consumption multipliers are high during times of financial crisis. In particular, for weak public
finances corresponding to “government debt in excess of 100 percent of GDP or net government
borrowing above 6 percent of GDP (each lagged once)”, they find that government spending
responds negatively to a weak fiscal position, thereby contributing to stabilizing public debt.
Depending on the fiscal position, especially with the presence of high public debt, Bi et al. (2016),
adopting a real business cycle (RBC) framework, find that the fiscal multiplier is generally smaller
in a high-debt than a low-debt state when general income tax rates serve as an adjustment
instrument, but the difference shrinks as the wealth effect on labour becomes strong. Furthermore,
uncertainties involving household reactions to the timing and magnitude of the shock, as well as
the debt target of fiscal consolidation, also matter. Expecting a higher debt target is not always
expansionary, especially when households perceive consolidation to be implemented via adjusting
labour tax rates, and expecting a higher debt target produces a positive wealth effect, which reduces
the current hours worked and thus offsets positive government spending effects (Bi et al., 2016).
The previous results are in accordance with the findings of Huidrom et al. (2016), which confirm
that fiscal multipliers are state-dependent of the fiscal position and tend to be systematically
smaller when government debt and deficit are high (weak fiscal position). The authors also show
that the fiscal multipliers’ dependency on the fiscal position is independent of the business cycle
effects. In particular, while the size of the fiscal multiplier tends to be larger in recessions and
weaker in expansions, the effects of the fiscal position (weak/strong) apply independently of the
economy being in recession or expansion.
In relation to fiscal consolidation under high public debt, Boussard et al. (2012) and Berti et al.
(2013) tend to confirm the large effect of fiscal multipliers in times of crisis, which push the debt
ratio to increase in response to fiscal consolidation, particularly in the short to medium term.
However, these undesired effects on the debt dynamics are judged to be short-lived unless these
large multipliers persist over time, which may be caused by non-credible fiscal adjustments and
the very high (abnormal) impact on interest rates and sovereign yield. These two publications
report “critical” fiscal multipliers, defined as “multipliers that can then be defined as the value of
the multiplier for which a fiscal shock would leave the public debt ratio unchanged (while a
multiplier higher than the critical value would entail a short-term increase in the debt ratio)”.
96
They show that these multipliers are inversely correlated to the change in debt ratio. According to
the authors, the true fiscal multiplier could be higher than the critical multipliers, especially for a
group of highly indebted countries, namely, Belgium, Bulgaria, the Czech Republic, Ireland,
Greece, Spain, France, Italy, Cyprus, Lithuania, Hungary, Portugal, Slovenia and the United
Kingdom.
Moreover, Blot et al. (2014b) simulate the dynamic path of public debt and output, under fiscal
consolidation, using a simple macroeconomic model for 11 eurozone countries and considering
time-varying fiscal multipliers. They analyse the ability of EMU countries to reach public debt
ratios below the threshold of 60% in 2032 to comply with the new fiscal rules of the EMU Stability
and Growth Pact. The revised Stability and Growth Pact, which was signed in 2012, outlines
converging (from an average level of debt ratios of around 80% across the eurozone in 2012) to a
60% debt to GDP ratio, by 1/20th of the adjustment yearly, which allows until 2032 to achieve the
target. In this way, Aussilloux et al. (2018) make an exercise of simulation on public expenditure
reduction for France, expecting that the public debt to GDP ratio could decrease by 25 points from
100% currently to 75% in 2040 for a best-case scenario of fiscal consolidation.
Canzoneri et al. (2015) and Broner et al. (2019) consider studying the type of financing spending
to impact fiscal multipliers in times of economic downturn. Canzoneri et al. (2015) find that either
tax-financed or debt-financed spending leads to multipliers that are higher than those in recessions,
with the multipliers being much larger for debt-financed than tax-financed spending. According to
the authors, “The reason is that while higher government spending sets in motion the financial
accelerator, higher taxes partly counter this by reducing the quantity of funds available to
financially constrained individuals”. Broner et al. (2019) study fiscal multipliers, considering the
portfolio of foreign public debt for a panel of 17 advanced countries from 1980 to 2014. In this
case, they reveal that fiscal multipliers are stronger when the expenditure is financed by foreign
resources. Their size increases, in particular, with the share of foreign public debt and is larger
than unity in periods and countries with a high foreign share of public debt (as in the United States
and Ireland today) and smaller than those in the opposite case (as in the United States in the 1950s
and 1960s, and Japan today).
Moreover, Poghosyan (2017) studies the way that the public debt cycles interact with financial
cycles for 57 advanced and emerging economies over the period 1960–2014. He finds that public
debt cycles are asymmetrically linked to financial cycles in the way that public debt expansions
that are preceded by fuelling in credit and financial markets are longer than any other expansions,
while there is no substantial association between public debt contractions and financial cycles.
Afonso and Leal (2018) compute fiscal multipliers for government expenditure in the eurozone
for quarterly data over the period 2001–16, using a structural VAR model. They consider the state
of the economy, particularly the reaction to the public debt level, the pace of economic growth and
the output gap. Government expenditure multipliers accumulate to less than 1 over a year (0.64
97
yearly and 1.1 after two years), while tax multipliers are negative. Moreover, expenditure
multipliers are larger for countries with higher public debt levels during recessions (compared to
low public debt levels where the multiplier is close to 0) and in countries showing positive output
gaps.
In contrast to the previous findings about the effects of fiscal position on fiscal multipliers, very
recently some new studies have minimized the effects of the fiscal position on fiscal multipliers
and then minimized the effects of high public debt on reducing the benefits of fiscal expansion in
advanced economies (Auerbach and Gorodnichenko, 2017; Perdichizzi, 2017; Blanchard, 2019).
For example, using a non-linear panel VAR model controlling for the macroeconomic properties
of 12 eurozone countries over 1985–2015, Perdichizzi (2017) find that fiscal spending multipliers
are insensitive to the level of government debt. Furthermore, these multipliers are larger in times
of recession for countries with low degrees of trade openness, high deficit and fixed exchange rate
regime, compared to countries with high degrees of trade openness, low deficits and flexible
exchange rate regime.
Auerbach and Gorodnichenko (2017) also produce interesting results on this issue for 17 OECD
countries. First, the government spending shock effects depend on a country’s position in the
business cycle, with the fiscal multipliers being larger in times of bad economic conditions than
good ones. Second, fiscal expansion while the economy’s fiscal position is weak is likely to boost
economic output and reduce the debt to GDP ratio, as well as appeasing interest rates and CDS
spreads on government debt. Consequently, these findings suggest that fiscal stimulus under a
weak fiscal position is likely to boost the economy without worrying about the associated modest
cost. However, these results should be considered with caution, as the authors themselves argue,
based on the asymmetric effects of fiscal policy shocks (see discussion in Section 2.4).41 The
authors call for more detailed research on this issue using more frequent and variable data on public
debt and more disaggregated categories of government spending, as well as structural models for
clearer policy recommendations. Besides, Alichi et al. (2019) consider the size of the country and
focus on estimating government spending effects for 23 small countries across the world. They
conclude that fiscal policy in small countries using government primary spending is ineffective at
stimulating the level of GDP over the medium term compared to government spending. However,
in the short term, multipliers for government current primary spending are higher and sensitive to
the level of government debt, the position of the economy in the business cycle, as well as imports
as a share of GDP, among other factors.
41
These results should be interpreted with caution, as the authors argue themselves, because of the problem of
asymmetric effects “… we recode fiscal shocks series so that the sign of the shocks is negative whenever the shocks
take a nonzero value and thus estimated impulse responses show dynamics after an increase in government spending.
This recording may be problematic since the effects of government spending cuts are not necessarily symmetric to the
effects of government spending increase… thus one should bear in mind the caveat that, although we interpret results
as showing responses to increases in government spending, the estimated responses are only based on cuts in
government spending” (Auerbach and Gorodnichenko, 2017).
98
Along the same lines, in analysing the fiscal and welfare costs of higher debt with reference to the
United States, where the safe interest rate42 is less than the growth rate, Blanchard (2019) argues
that both the fiscal and welfare costs of debt may be smaller than assumed in current policy debates.
Blanchard (2019) seems likely to draw the same conclusions for European economies. His results
have triggered a debate, which have been contained until now in economist blogs and some media.
This new paper about debt cost nevertheless warns that the cost of austerity measures driven by
the fear of high debt is likely to hurt more than the cost of debt build-up, since actual data shows
that the interest rate differential/gap growth rate minus interest rate is positive enough to stabilize
the public debt ratio while maintaining a small primary deficit. In this case, two important points
are worth mentioning. First, the Committee of a Responsible Federal Budget (CRFB), in response
to Krugman’s “misinterpretation” of Blanchard’s (2019)’ conclusions, outlined that Blanchard’s
(2019) conclusions are correct if the primary balance in the United States is small, but the
American economy is running a huge primary deficit.43
The second point highlights some arguments and counter-arguments of potential debt finance. The
arguments that Blanchard (2019) reports in favour of potential debt finance (standing against fiscal
consolidation) are: revised large multipliers, debt hysteresis, higher marginal product of public
capital and necessary budget deficits to stimulate demand in the context of constrained monetary
policy. Alternatively, the counter-arguments about the potential costs of public debt are as follows:
the safe interest rate may be artificially lower than the observed one (which could happen in the
case of liquidity discount); the future may be different to the past because of many factors related
to total factor productivity (TFP) and an ageing population; and the last counter-argument relies
on the existence of multiple equilibria.
However, while some enthusiastic supporters of fiscal stimulus welcome the message of this paper
(Krugman, 2019),44 this has not been commonly agreed by other economists, as historical data
showed the opposite for the most important European economies (Mazza, 2019; Philippon,
2019).45 This paper, while minimizing the effects of high public debt when the safe interest rate is
below the nominal GDP growth (which is the case for many advanced economies), is likely to re-
fuel the debate about public debt effects, and Blanchard (2019) himself argues that this should not
be taken as an invitation for more debt rollover and calls for more investigation on this issue.
42
Blanchard (2019) uses the terminology “safe interest rate” to describe, depending on the situation: the risk-adjusted
rate of return on capital or the interest rate on nominal bonds (assuming no default).
43
See http://www.crfb.org/sites/default/files/CRFB_DoNotMischaracterizeBlanchard.pdf.
44
https://www.nytimes.com/2019/01/09/opinion/melting-snowballs-and-the-winter-of-debt.html.
45
http://bruegel.org/2019/01/is-public-debt-a-cheap-lunch/ and https://www.stern.nyu.edu/experience-stern/faculty-
research/true-cost-public-debt.
99
2.3. Fiscal multipliers in a constrained monetary policy and exchange rate

regime
Just before the financial crisis, economists seemed to agree on the powerful role of monetary policy
as a stabilization instrument, eclipsing, to some extent, the role of fiscal policy. However,
conventional monetary policy effectiveness is impaired when the interest rate hits the ZLB rate or
when it loses its independence because of a currency pegged to another country.46 In these cases,
fiscal policy is more desirable.47 In assessing the effects of fiscal policy in the case where monetary
policy is constrained by the ZLB interest rate (also called the liquidity trap), or by the fixed
exchange rate, as when the country participates in a monetary union, the new Keynesian modelling
framework is more desirable than other forms of modelling in assessing the effects of fiscal
policy.48
Previous important contributions on the effects of zero interest rates have been conducted by many
authors in the context of new Keynesian dynamic stochastic general equilibrium (DSGE) models,
which have their basis in the real business cycle (RBC). Therefore, Eggertsson and Woodford
(2003) and Eggertsson (2011) conduct simulations of an economy with a zero interest rate and find
a large effect of temporarily increasing government spending in this situation, much larger than
under normal circumstances. Numerical evidence conducted by Eggertsson (2011) suggests a
spending multiplier that could attain five times the spending multiplier in normal circumstances.
In the same scope of new Keynesian models49 used in analysing monetary stabilization policy,
Woodford (2011) finds that under the severe conditions of a great depression and a lower interest
rate, near 0, these models are likely to report larger multipliers, higher than 1, which, to some
extent, may also increase welfare. In this situation, an increase in government spending could have
a powerful effect, such as offsetting the negative output gap. However, in less severe
circumstances, the fiscal policy through expenditure tool is less powerful, even in the case of a
binding constraint of a zero interest rate, especially when the disturbance implying it to bind is
expected to be transitory. In such a case, although the spending multiplier could stand above unity
in such circumstances, it is viewed as insufficient to eliminate the negative output gap. Similarly,
Christiano et al. (2011) base their analysis on a new Keynesian DSGE model to argue that the
46
Before the financial crisis of 2008, researchers studying the constraint of the zero interest rate policy were inspired
by the real situation of the Japanese economy, where the interest rate has nearly touched zero since 1995, accompanied
by anaemic economic growth (Eggertsson and Woodford, 2003).
47
Farhi et al. (2013) show that both a uniform increase in import tariff and export subsidy, and a VAT tax increase
and uniform payroll tax reduction, can play a role equivalent to exchange rate devaluation (which is mentioned as
fiscal devaluation). However, when these policies are anticipated by economic agents, they need to be supported by
consumption tax reduction and income tax increases.
48
For example, Correia et al. (2012) use a standard new Keynesian model to assess the fiscal policy when the
conventional monetary policy is constrained by the ZLB nominal interest rate. They show that tax policy (distortionary
taxes) can deliver similar expected benefits to the monetary policy (being constrained in this case).
49
Woodford (2011) reviews the benchmark of the neoclassical models, assuming perfect flexibility of prices and
wages, leading generally to spending multipliers of less than 1. The author also inferred cases under which the new
Keynesian DSGE models could lead to spending multipliers greater than, equal to or below unity.
100
government-spending multiplier can be large when the ZLB on nominal interest rates binds, except
when the central bank is committed to a Taylor rule, in which case the spending multiplier is
smaller than 1. Likewise, in the ZLB, the value of the multiplier is positively related to both the
fraction of government spending, as well as the expected duration of the zero bound constraint.
Consequently, efficient government spending should fall in line with the state of the ZLB interest
rate.
For countries participating in a monetary union, Kilponen et al. (2015) simulate fiscal
consolidation effects for 15 structural models from national central banks (NCBs) in the euro area
and the ECB (of which 14 out of 15 have a new-Keynesian dynamic general equilibrium
framework). They find that short-term multipliers are, in general, negative and smaller than 1 in
absolute value, independent of the fiscal instrument, the studied country and the duration of the
fiscal shock, with tax multipliers typically smaller in absolute value than government consumption
multipliers. However, for permanent fiscal shocks, the short and long-term effects depend on the
fiscal instrument, which acts endogenously to stabilize the long-term public debt path.
Furthermore, the simulated short-term multipliers seem to be insensitive to the ZLB exchange
rates, except when simultaneous fiscal consolidation is implemented across the euro area as a
whole.50 In this case, short-term government consumption multipliers become larger than 1. This
result stays valid for non-euro area countries, for which monetary policy is independent and
domestically determined.
Furthermore, Farhi and Werning (2017) differentiate between the fiscal multipliers resulting from
liquidity traps from the ZLB interest rate and those arising from the currency union, using a
standard new Keynesian model but also considering Ricardian and non-Ricardian (hand-to-mouth)
agents. They particularly highlight the differences in fiscal multiplier values in the ZLB case and
those in the situation of a monetary union. In the first case (the liquidity trap due to the ZLB interest
rate), in the standard Ricardian model (generally assuming Ricardian agents), fiscal multipliers are
always larger than 1 in liquidity traps. According to the authors, the mechanism is insured through
inflation: higher government spending during liquidity traps stimulates inflation, and with a fixed
nominal interest rate the real interest rate is reduced, which increases current household
consumption. However, for a country in a monetary union, in the standard Ricardian model,
assuming price flexibility, the crowding-out effects of government spending on private
consumption, as well as domestic inflation spurred by government spending, which induces a loss
of competitiveness and depresses private investment, drags down the multiplier to less than unity.51
50
As the aim of fiscal devaluation is improving competitiveness when the monetary policy is constrained to do that
via the exchange rate devaluation (as in the case of monetary union), and while the competitiveness of a country is
enhanced at the expense of the other countries, simultaneous fiscal reforms targeting fiscal devaluation are likely to
cancel/compensate the effects of fiscal devaluation between countries (European Commission, 2013).
51
According to Farhi and Werning (2017), “the liquidity trap analysis implicitly combines a shock to government
spending with a one-off devaluation. The positive response of consumption relies entirely on this devaluation. A
currency union rules out such devaluation, explaining the difference in the response of consumption”.
101
In contrast to the previous results, Glocker et al. (2019) do not find the fiscal spending multipliers
to be affected by the ZLB interest rate. They only report, using a time-varying parameter factor
augmented vector autoregressive (TVP-FAVAR) framework for the United Kingdom, that fiscal
multipliers are timely and cyclically variant: in particular, they are larger in times of recession
(more than unity) and lower in times of expansion (less than unity). Moreover, and contrary to
research showing “non-Keynesian multipliers” (i.e. inferior to 1) for countries participating in a
currency union, Combes et al. (2014) use panel VAR (PVAR) on quarterly data over the period
1999–2012 to assess the effect of fiscal multipliers on European countries. They reveal that
spending multipliers can be larger than 1, especially for countries that are members of the
eurozone, or for those expecting to join it. Furthermore, these multipliers tend to be higher in the
group of euro countries affected by the 2008 crisis compared to the benchmark of the eurozone
countries.
As previously revealed in Section 2.2, spending multipliers are also sensitive to the way they are
financed. Spending multipliers are much larger when they are debt-financed compared to when
they are tax-financed (Canzoneri et al., 2015), and they are also much higher when they are
financed by foreign rather than domestic debt (Broner et al., 2019). Along the same lines, and
within a monetary union, (local) fiscal multipliers are also sensitive to the transfers from outside
regions/countries for a country participating in a monetary union. Particularly, fiscal multipliers
seem to be large (and may even be larger than 1) when there are fiscal transfers from outside
regions or countries (Farhi and Werning, 2017; Auerbach et al., 2019).52
The degree of openness also plays an important role in this issue, with more closed economies
having larger fiscal multipliers than more open ones. This happens particularly in the short term
and incomplete financial markets, as prices that are not fully adjusted push up demand for home
goods, which stimulates output. Furthermore, the persistency of the fiscal shocks increasing the
magnitude of the fiscal multipliers is due to the agents’ willingness to save more from foreign
fiscal transfers in temporary shocks expecting future lower periods of transfers. Previous to Farhi
and Werning’s (2017) work on regional multipliers, Nakamura and Steinsson (2014) find that
regional fiscal multipliers in the United States, related to military expenditure, are strongly
dependent on the business cycle. Using data on military procurement spending across US regions,
they estimate a government spending multiplier in a monetary union of approximately 1.5. Erceg
and Linde (2012) use a new Keynesian DSGE model for a relatively small open economy to study
the effects of fiscal consolidation for a monetary policy constrained either by a monetary union
affiliation or by the ZLB on policy rates (for a country with an independent monetary policy). The
52
Farhi and Werning (2017) distinguish between local fiscal multipliers due to local government spending only and
the overall fiscal multipliers, which consider in the model not only local government transfers but also the possibility
of transfers from other countries or regions in the currency area (for example, in the US, federal military spending
allocated to each state is financed by the federal budget). This may explain, in part, the difference between the Unites
States’ relatively fast recovery compared to the eurozone countries, as these fiscal transfers are almost absent between
eurozone countries. The allocated federal budget is roughly 1% in the eurozone, while it is around 20% in the United
States (Feyrer and Sacerdote, 2013).
102
impacts of fiscal consolidation differ between the two cases, with differences depending
particularly on the persistency time of the ZLB (generally more than two years) and on the inflation
responsiveness to output gap. Principally, for a sensitive inflation to output gap, fiscal
consolidation is more contractionary under the ZLB rate than under currency union.
2.4. The impact of the assessment methods
Despite the growing body of literature on fiscal multipliers, their estimation remains very
challenging. Generally, methods mostly employed for assessing fiscal multipliers are of three
types: first, structural macro-econometric models; second, the multivariate time series, especially
the standard vector autoregressive model and its derivatives of linear and non-linear types
(structural VAR, threshold VAR, etc.); and, third, the new Keynesian DSGE models.
Macroeconomic structural models were thoroughly developed by public and government

institutions, as well as academicians, in the post-war period and are still largely used in fiscal
policy analysis.53 These are based on works associated with the Cowles Commission programme,
based on Keynes’ general theory and the relative consensus around the IS-LM framework and
econometric tools to estimate the structural parameters. Large-scale models were developed
following the first attempts to model the United States economy by Tinbergen (1939), Klein (1950)
and Klein and Goldberger (1955).54 These kinds of model, however, were exposed to two
important critiques by Lucas (1976) and Sims (1972, 1980).55 Lucas highlighted in his critique the
inability of econometric models to catch structural breaks in the behaviour of economic agents vis-
à-vis monetary policy changes. He expressed this critique as follows: “Given that the structure of
an econometric model consists of optimal decision rules of economic agents, and that optimal
decision rules vary systematically with change in the structure of series relevant to the decision
maker, it follows that any change in policy will systematically alter the structure of econometric
models” (Lucas, 1976).
Sims’ (1972, 1980) critique is directed at exogeneity, causality issues, parameter identification and
hypotheses testing. The contributions of the VAR approach to economics science are well
53
For example, Coenen et al. (2012) simulate government spending multipliers under seven structural models from
six institutions, namely, the European Commission (QUEST), the International Monetary Fund (GIMF), the Board of
Governors of the Federal Reserve System (with two models, FRB-US and SIGMA), the Bank of Canada (BoC-GEM),
the European Central Bank (NAWM) and the OECD (OECD Fiscal). The models tend to produce sizeable output
multipliers of expenditure and targeted transfers, especially under accommodative monetary policies, although
permanent fiscal stimulus can significantly lower such multipliers.
54
See also, for example, Fox (1956) and Chinn (2013) for a complete literature review.
55
The Lucas critique has revolutionized macroeconomics science by developing many theories such as the theory of
rational expectations, which won him the 1995 Nobel Prize, the development of the “time inconsistency” problem by
Kydland and Prescott (both Nobel Prize winners in 2004) and the important contribution of Thomas Sargent over the
credibility of economic policies. Thomas Sargent also contributed to much empirical research in line with Cristopher
Sims, with whom he shared the 2011 Nobel Prize for their contribution to cause and effect in macroeconomics.
103
summarized and synthesized, for example, in Hoover el al. (2008) or Gossé and Guillaumin
(2013).
Following these critiques, Sims (1980) brought the vector autoregressive (VAR) models and their
utilization to diverse applications as an alternative to the macroeconomic structural models. Their
strength resides in their simplicity and robust application, especially in forecasting, compared to
structural models. However, these models have also attracted many critics, especially the famous
Lucas (1976) critique. The third type of models based on the real business cycle (RBC)
neoclassical model, are the new Keynesian DSGE models (due to the new synthesis theory),
largely developed and used in recent years by central banks and monetary institutions. Kydland
and Prescott (1982) first confronted the theory with data through the application of calibration
methods to real business cycle (RBC) models, avoiding the option of fully econometric estimation,
as did the pioneer of structural models.
All these methods have been criticized for caveats on their technical and conceptual design in
analysing fiscal policy effects. For example, Cogan et al. (2010) highlighted the role played by the
quantitative models in evaluating fiscal policy. In particular, government spending multipliers in
the recently used new Keynesian models tend to be much smaller than in the old Keynesian ones.
This conclusion came after simulating a new Keynesian model based on the core model of Smets
and Wouters (2007) for the United States and comparing its results to those of Romer and Bernstein
(2009), which are based on an old Keynesian model for the same country. They conclude that the
models used to assess fiscal policy effects lack robustness and consensus in their results, in which
there are stark differences. Their new Keynesian model showed fiscal multipliers and job impact
six times smaller than those found by Romer and Bernstein’s (2009) old Keynesian model. The
principal challenges remain in many differences in assumption, economic issues, as well as
econometric and statistical aspects. Generally, the drawbacks and caveats of the new Keynesian
models have been discussed in the recent literature, as reported in Chapter 3 (Section 2.4).56
The difference in assumptions, as well as their structure, greatly impacts the size of the multipliers
between these three classes of models. Changing assumptions in the same class of models may
also lead to a sizeable change in fiscal multipliers. For example, assuming complementarities
between public and private capital in the neoclassical RBC models leads to higher spending
multipliers than the modest values without this assumption (Baxter and King, 1993). Under its
standard framework, the new Keynesian DSGE model leads to smaller spending multipliers, while
relaxing some assumptions, such as the presence of the ZLB interest (Christiano et al., 2018; Galí,
2018), or the relaxation of the Ricardian equivalence assumption in favour of “hand-to-mouth”
agents (Farhi and Werning, 2017), leading to an increase in spending multipliers.
56
In addition, a section of Chapter 3 (in the literature review) also summarizes the critiques of the neoclassical models
(overlapping generation models and infinitely lived agents), as well as the new Keynesian models (DSGE models).
104
From an economic perspective, agents’ expectations (“fiscal foresight”) about future spending and
tax policies, the monetary policy reaction, the exchange rate regime, the degree of openness, the
state of the economy and the fiscal position (high/low public debt and deficit) are examples of the
economic determinants making it difficult to deduce the fiscal multiplier, especially given that,
remembering the definition, the fiscal multiplier measures the effects of an exogenous
distortionary fiscal instrument on the level of the output. From an econometric and statistical
perspective, many problems arise when calculating fiscal multipliers. Important issues are the
endogeneity of government spending and non-linearities created by extreme changes, especially
in times of extreme events (high expansions or severe downturns). The accuracy of the data and
its availability on a disaggregated level is also an important determinant of the correct multipliers’
size (Ramey, 2018).
For example, Parker (2011) warned against the impact of the methods and the lack of accurate data
used to gauge the effects of fiscal policies through multipliers, especially the use of linear dynamic
forms represented by the vector autoregression models and linearized dynamic stochastic general
equilibrium models. Accordingly, ignoring the state of the economy (the business cycle) leads to
multipliers representing a weighted average of the phases of the cycles over the studied period,
which is lower than those estimated in recessions and higher in those emanating from expansions.
These caveats can be resolved, according to the author, by considering partial-equilibrium
response estimations on a microeconomic level rather than macro and aggregated data. Along the
same lines, Canzoneri et al. (2015) highlighted that studies assessing fiscal multipliers, without
distinguishing between recessions and expansions, lead to spending multipliers around unity,
while these multipliers could reach more than 2 during recessions and only around 0.5 during
expansions.
The new Keynesian models have been criticized for assessing the fiscal multipliers (see also the
critiques of this type of model for their assessment of debt effects in Chapter 3, Section 2.4). The
new Keynesian DSGE models rely principally on many assumptions that are hard to realize in
reality, for example, the assumption about the behaviour of agents generally supposed to be
Ricardian, with rational expectations, as well as their poor performance in the recent crisis
(Chatelain and Ralf, 2012; Wieland et al., 2012; Blot et al., 2014b). Furthermore, these models are
unable to deal with non-linearities of fiscal multipliers over the business cycle (Blot et al., 2014b).
Furthermore, modelling behaviour is not straightforward. The heterogeneity of economic agents’

behaviour adds complexity to the adopted modelling approach conditioning the fiscal policy
assessment. For example, Farhi and Werning (2017) distinguish between Ricardian and non-
Ricardian consumers that could affect the size of the fiscal multipliers. Particularly, in a non-
Ricardian model where “hand-to-mouth” consumers are considered, marginal propensities to
consume are different between Ricardian and non-Ricardian groups of agents. This leads to
additional effects of government spending increasing output, whether in a liquidity trap or in a
currency union, due to the incidence of taxes, the timing of taxation (on the non-Ricardian agents)
105
and redistribution from Ricardian agents with low marginal propensity to consume permanent-
income to non-Ricardian agents with high marginal propensity to consume. Accordingly, this may
raise the multipliers above 1 for a country within a currency union (Farhi and Werning, 2017).
Another issue that is mentioned in the literature belongs to the “fiscal foresight” policy shocks
(Forni and Gambetti, 2010; Mertens and Ravn, 2010; Leeper et al., 2012, 2013). The standard
VAR models assess the effects of fiscal policy shocks from current and past information from the
data of fiscal variables involved, and they are unable to embed/incorporate the so-called
information on “fiscal foresight” policy shocks, that is, shocks related to agents’
anticipations/expectations on government fiscal policy spending (Caggiano et al., 2015). For
example, Caggiano et al. (2015) address this issue by considering state-dependent fiscal
multipliers, taking explicit information on such expectations using a measure of anticipated
information (“news”) on fiscal spending shocks and a group of macro-fiscal variables from the
Survey of Professional Forecasters.
For the non-linear effects and behaviour of fiscal relationships, Fotiou (2019) uses a mix of the
two methodologies of Auerbach and Gorodrichenko (2012) of the STVAR57 model and the
methodology of Nickel and Tudyka (2014)58 to study non-linear effects on fiscal multipliers for a
panel of 13 countries during 1980–2014. According to the author, non-linearities arise from, first,
the state of the economy (recessions versus expansions), second, the composition of the fiscal
policy, and, third, the government’s fiscal position (high debt ratio versus low debt ratio). He finds
that the initial level of debt and the composition of the fiscal adjustments are the most relevant
sources of non-linearities. Furthermore, for this author, tax-based consolidation tends to be self-
defeating whenever debt is high. Casalis (2017) also studies the effects of non-linearities
controlling for public debt and financial cycles. He reports that non-linearities produce more
queries than answers for policy advice and that comparing results is likely to be misleading in this
environment.
2.5. Do we need a consensus about a unique fiscal multiplier size for all
countries?
Despite the important flux of studies about fiscal multipliers, these studies show more uncertainties
than certitude about their size. The differences in methods, as well as samples of countries and
time periods, play an important role in this issue. Recently, Ramey (2018) summarized the main
research outcome on fiscal policy since the 2008 financial crisis. According to the author, average
fiscal spending multipliers vary in a narrow range between 0.6 and 1. However, there are economic
circumstances where spending multipliers lie outside this range, as well as the impact of
approaches used to assess those multipliers. In particular, this range may widen if country
57
STVAR stands for smooth transition vector autoregressive, which uses a regime-switching model.
58
Nickel and Tudyka (2014) consider fiscal position in assessing fiscal multipliers by incorporating debt ratio as an
interaction term in the panel VAR, which is indicated as interacted panel VAR (IPVAR).
106
characteristics, such as the exchange rate regime, and the type of government spending are
considered.
Along the same lines, Parraga Rodriguez (2018), following an SVAR framework, finds that
government spending multipliers and income transfers multipliers in the United States are by far
below 1 in the short term (approximatively 0.2). However, compared to government fiscal income
transfers multipliers that can reach more than 1 in the long term, the spending multipliers cumulate
only to 1 in the long term. Ramey and Zubairy (2018), focusing again on the case of the United
States, as the single country case for which multipliers are most evaluated in the literature, examine
whether the government spending multipliers are sensitive to the zero bound interest rate and the
business cycle in the economy.
For Ramey (2018), concluding that the average of fiscal multipliers (spending) reported in the
previous literature is low and ranges between 0.6 and 1, averaging fiscal multipliers without
distinguishing between the sign of the fiscal shocks59 is misleading. Indeed, and surprisingly, to
our best knowledge, this flux of very recent studies does not seem to distinguish between
multipliers drawn from negative shocks and those from positive shocks, leading us to understand
that the effects are similar in the two situations. For example, some studies may draw conclusions
on the effect of fiscal multipliers, from fiscal austerity and fiscal consolidation, as if they were the
same as in fiscal stimulus, thereby admitting the symmetrical effects in the two opposed shocks.
One of these studies, Ramey and Zubairy (2018), draws the following conclusion “... If multipliers
are indeed this low, they suggest that increases in government purchases do not stimulate private
activity and that fiscal consolidations based on reducing government purchases are unlikely to do
much harm to the private sector”. Ramey (2018) also does not seem to distinguish between these
situations (fiscal consolidation and fiscal stimuli) in averaging the fiscal multipliers reported in the
previous studies. Another study, by Blot and al. (2014b), concludes that “... Recent mainstream
literature has emphasized that fiscal multipliers may notably be higher in time of crisis. Then, not
only would fiscal consolidation drag down growth more severely, but it could even be self-
defeating”.
In line with the literature studying fiscal multipliers’ dependency on the business cycle, Ramey
and Zubairy (2018) study the state dependency of fiscal multipliers, involving, at the same time,
the zero lower bound for the United States. They contrast the findings of the previous research in
line with Auerbach and Gorodnichenko (2012; 2013) and report multipliers ranging in a narrow
band between 0.6 and 1. The higher magnitude of the fiscal multipliers in the zero lower bound is
also of little evidence in Ramey and Zubairy (2018).
59
For example, the distinction should be made between positive expenditure shocks (fiscal stimulus) and negative
shocks (austerity measures). The outcome of the fiscal policy is different under the two scenarios and depends on the
economic environment and countries’ economic fundamentals.
107
Ramey and Zubairy’s (2018) results should be considered with caution, especially with regards to
their data set. Indeed, while they have the advantage of pointing to their long history and more
frequent data for the United States (sample between 1889 and 2013), half of the period sample is
constructed for the World War periods and before (1889–1946) and interpolated using different
methods in four different sub-periods.60 This quarterly constructed data could imply serious
problems of accuracy and may have serious problems for the calculated multipliers. This may
explain the difference between results where the multipliers seem to be larger in the post-war
period when omitting the initial period (although they suggest that these multipliers are not
statistically strongly significant). Furthermore, Ramey and Zubairy (2018) use narrative methods
to extend Ramey’s (2011) defence news series in order to identify shocks that are unanticipated
and exogenous to the state of the economy.61
Important exceptions to these papers are Baum and Koester (2011) and Riera-Crichton et al.
(2014), who mention that government spending is not necessarily acting counter-cyclically (going
up in times of recessions); rather, it could be, as is the case in many industrial economies, pro-
cyclical (decreasing). In this paper, the authors control for the sign of fiscal shocks (increase or
decrease in government spending), as well as the size of the fiscal intervention, rather than
distinguishing only between recessions and expansions. Their analysis reveals that fiscal
expansions are much more expansionary in recession periods than in expansion periods. Using a
threshold SVAR (TSVAR) to account for non-linearities, Riera-Crichton et al. (2014) find that the
long-term multiplier for bad times and an increase in government spending is around 2.3 higher
than the value of 1.3 if we control for recession only and expansion is considered. However, using
the same methodology of threshold VAR, Baum and Koester (2011) find that public expenditure
multipliers vary depending on the size of the shock, its sign and the level of the output gap.
Consequently, a positive fiscal shock (increase in government expenditure) in crisis periods leads
to a higher spending multiplier, and the latter increases with the size of the fiscal shock. However,
in good times, multipliers are lower and seem to behave more linearly.
To sum up this section, the wide spread of results about the size of spending multipliers leads us
to conclude that these multipliers, despite their simple definition, reflect:
1- The difference of methods and models used to assess these multipliers: with the same data,
and on a single country (the United States, for example), researchers find different results
whether the used model is a structural model, a new Keynesian DSGE model or a (non-)
60
The time series are the real GDP, the GDP deflator, government purchases, federal government receipts, population,
the unemployment rate, interest rates and defence news. The time series data is interpolated using different techniques
depending on the series over the period 1889–1946 and its sub-periods. They follow Gordon and Krenn (2010) by
using various higher-frequency series to interpolate existing annual series. Generally, the proportional Denton (1971)
procedure, resulting in series that average up to the annual series, is used in the interpolation. This method is robust
(Chen, 2007) and recommended in IMF or Federal Reserve Bank publications (Liu et al., 2011; Woo et al., 2013).
61
The news series is linked to government spending due to political and military events and is likely to be independent
of the business cycle. This is an important difference from and other papers using the output gap and economic growth
as the instrument variable to determine the turning points of the business Auerbach and Gorodnichenko (2019) cycle.
108
linear VAR/SVAR model. The assumptions and features, as well as the methods of solving
the three types of model, vary widely.
2- The difference in macroeconomic fundamentals of the studied countries, although the most
advanced countries share, to some extent, the same level of development and qualified
institutions, fiscal policies effects may differ regarding the difference of monetary policy
and exchange rate regimes, as well as economic conditions (the business and/or the
financial cycle), as represented by the levels of debt and deficit, for example. Other
determinants could play an important role as the degree of openness.
3- Consequently, all the differences in empirical results should not be seen as an incongruity
between economists. It is a fact rather than a general theory that should apply to all
countries. In this regard, a good way to study the effects of fiscal policy is to avoid
considering the empirical results of one country or a group of countries as a universal
benchmark for all countries. Therefore, studies of fiscal multipliers should be undertaken
at country level and avoid drawing conclusions from a single country such as the United
States, Japan, the United Kingdom or Germany. When it comes to assessing things
empirically, each country’s data set represents its own model of economic development
and experience across a period of that country, and if this is not even valid to reproduce the
future of this economy itself, it is hardly transposable to a different country.
Whether in fiscal stimuli or fiscal consolidation, accurately estimating fiscal multipliers by type of
expenditure helps policy-makers to know what categories of spending they should increase (in
fiscal stimuli) or cut (in consolidation). The next section undertakes calculations of government
spending multipliers in a sample of advanced countries.
3. Methodology
In our empirical investigation, we use a sample of 18 advanced countries over different periods of
time, a panoply of structural vector autoregressive (SVAR) models to assess the fiscal expenditure
impacts on the output. We especially test how the business cycle could affect expenditure
multipliers, as well as the way accumulating public debt and reducing public debt are impacting
the size of the expenditure multipliers. In what follows, we display, first, a detailed methodology
of an SVAR model, particularly, its formulation, lag selection and identification of shocks. Second,
we discuss the identification restrictions for the considered VAR models linking government
expenditure with GDP, augmented exogenously by a dummy variable that controls for the business
cycle (expansion/recession) and the public debt evolution (accumulation/decumulation), hence
noted an SVAR-X. We also endogenize the public debt variable instead of its exogeneous effects,
making a tri-variate VAR of government expenditure, debt and GDP. This model is also controlled
for the business cycle impacts.
109
3.1. The general methodology of a VAR/SVAR model
The vector autoregressive (VAR) methodology was popularized by the works of Sims (1972,
1980) following the debate between monetarists and Keynesians on exogeneity and causality
issues. In particular, Sims (1980) points out that the structural models of the Cowles Commission
have too many theoretical hypotheses that have not been empirically tested, and he presents the
VAR models as an alternative to these models. However, the standard VAR models, which are
reduced forms, do not integrate the structural model’s simultaneity where there is instantaneous
feedback between two endogenous variables (Desai, 1981). In particular, conditional correlations
assimilated with a causal order in the case of the standard VAR can only be justified under
determined hypotheses that could not be tested in the absence of a priori restrictions derived from
economic theory (Cooley and LeRoy, 1985). To overcome this issue and interpret canonical
innovations as exogenous economic policy shocks, Sims transforms the VAR model into a
structural VAR (SVAR). Furthermore, to justify the identification restrictions imposed on
innovations, Sims refers to Wold's (1954) notion of causal chain.
3.1.1 VAR formulation
A system of N equations linking linearly N endogenous variables through their past values to a
certain level p, representing the optimal/maximum lag, is called a standard vector autoregressive
with lag 𝑝, 𝑉𝐴𝑅(𝑝). It is expressed by Equation (1):
𝑝
𝑌𝑡 = 𝛷0 + ∑𝑖=1(𝛷𝑖 𝑌𝑡−𝑖 ) + 𝜖𝑡 (1)
𝑦1𝑡 𝑎10 𝜀1𝑡

Where; 𝑌𝑡 = [ ⋮ ], Ф0 = [ ⋮ ] and 𝜖𝑡 = [ ⋮ ] are vector columns of respectively; endogenous
𝑦𝑁𝑡 0
𝑎𝑁 𝜀𝑁𝑡
variables 𝑦1𝑡 … 𝑦𝑁𝑡 , constant terms and corresponding errors, also called innovations, shocks, of
1 2 … 𝑎1𝑝𝑁
𝑎1𝑝 𝑎1𝑝
each equation in the VAR, and 𝛷𝑝 = [ ⋮ ⋮ ⋮ ⋮ ] is the matrix of coefficients associated
1 1
𝑎𝑁𝑝 𝑎𝑁𝑝 … 1
𝑎𝑁𝑝
with lagged terms.
Defining lag time operator 𝐿 as 𝐿𝑋𝑡 = 𝑋𝑡−1 , Equation (1) can be written as:
𝑝
𝑌𝑡 (𝐼 − ∑𝑖=1 𝛷𝑖 𝐿𝑖 ) = 𝛷0 + 𝜖𝑡 (2)
𝑝
Or equivalently also, Ф(𝐿)𝑌𝑡 = 𝛷0 + 𝜖𝑡 , with I as the matrix identity and Ф(𝐿) = 𝐼 − ∑𝑖=1 𝛷𝑖 𝐿𝑖
110
3.1.2. Lag selection procedure
To determine the optimal lag 𝑝 to be introduced in the VAR, there is a set of methods commonly
used in the economic literature, which select the optimal lag based on the information criteria. The
most well known in the literature are Akaike information criterion (AIC) and Schwarz information
criterion (SC). However, other tests are also solicited such as sequential modified LR test statistic
(LR test), final prediction error (FPE test) and Hannan-Quinn information criterion (HQ). The
methodology of these criteria is based on selecting the model that minimizes a defined function.
The methodology consists of estimating the 𝑉𝐴𝑅(𝑝) for lags ranging from 0 to a maximum ℎ𝑚𝑎𝑥 ,
where the latter is the maximum number of lags included in the VAR model based on a defined
economic theory, a referenced academic article or even sometimes on a simple economic intuition
(Gossé and Guillaumin, 2013). We then retain the number 𝑝 of lags, which minimizes the 𝐴𝐼𝐶 and
𝑆𝐶 criteria defined by the following parametrized functions:
2
̂ )] + 2 𝑝𝑘
𝐴𝐼𝐶(𝑝) = ln[det(Ω (3)
𝑇
2 ln(𝑇)
̂ )] + 2 𝑝𝑘
𝑆𝐶(𝑝) = ln[det(Ω (4)
𝑇
̂ the
where T is the number of observations, k the number of variables in the VAR model and Ω
estimated variance–covariance of the residuals.
3.1.3. Identification and analysis of shocks
The reduced form is seen as a “black box”, as it is difficult to interpret the reduced form parameters
based on an economic theory. Therefore, the structural VAR (SVAR) was developed in the mid-
1980s to overcome this issue and allow economic agents’ behaviour to be described by the VAR
methodology. Since then, many authors have proceeded by including shocks grounded in the
economic theory in relation to economic policies, such as supply and demand shocks (Shapiro and
Watson, 1988; Blanchard and Quah, 1989), monetary shocks (Sims, 1986) or fiscal shocks
(Blanchard and Perotti, 2002). The pass-through from canonical shocks of the reduced form of the
equation (1), namely, 𝜖𝑡 , to structural shocks (based on economic behaviour), 𝜔𝑡 , requires the
existence of a matrix 𝑃 such as:
𝜖𝑡 = 𝑃𝜔𝑡 (5)
Consequently, identification of structural shocks insured once matrix 𝑃 has been estimated yields:
̂𝑡 = 𝑃−1 𝜖̂𝑡
𝜔 (6)
111
The initial method of identification of shocks, initially suggested by Sims (1980), is based on a
recursive Cholesky decomposition. Meanwhile, the standard VAR as a reduced form (without any
economic theory) requires the variables of the VAR system to be ordered considering the degree
of exogeneity/endogeneity of these variables; the order must go from the most exogenous variable
to the least exogenous (the most endogenous) one.
However, for a structural VAR, SVAR, the identification method is based on restrictions over
structural terms that are not necessarily recursive. The restrictions can be in the short term, as in
Bernanke (1986) and Sims (1986), as well as the long term, as pioneered by Blanchard and Quah
(1989). A combination of the two types (short- and long-term restrictions) was developed by Galí
(1992). Recently, restrictions have also targeted specific structural terms, annulling the coefficients
associated with these or imposing specific signs. This is done regarding the economic theory
(Faust, 1998; Canova and De Nicolo, 2002; Blanchard and Perotti, 2002; Uhlig, 2005).
To analyse the shocks of a VAR model, a vector moving average (VMA) (also called Wold
decomposition) is required. According to Wold’s theorem, every stationary system can be
expressed as a weighted average sum to infinity of white noise terms. Considering a stationary
𝑉𝐴𝑅(𝑝) of Equation (1), and supposing that the variables are centralized around the mean (i.e.
𝛷0 = 0), Equation (1) can be written in the form of vector moving average, noted 𝑉𝑀𝐴(∞) as:
𝑌𝑡 = ∑∞
𝑗=0 ѱ𝑗 𝜖𝑡−𝑗 = Ѱ(𝐿)𝜖𝑡 (7)
with Ѱ(𝐿) = ∑∞ 𝑗
𝑗=0 ѱ𝑗 𝐿 , ѱ0 = 𝐼 and 𝜖𝑡 as the vector of innovations (shocks). The impact of these
shocks (innovations) over endogenous variables is determined by dynamic impacts defined from
Equation (3) as:
𝜕𝑌𝑖,𝑡+𝑠
ѱ𝑖𝑗,𝑠 = (8)
𝜕𝜀𝑗,𝑠
where ѱ𝑖𝑗,𝑠 measures the impact of 𝜀𝑗 in time t over the endogenous variable 𝑌𝑖 following 𝑠 periods
of time. For 𝑠 = 0, this is called immediate impact or instantaneous effect (which can also be
assimilated to short-term effect).
Considering (6) and rewriting (7), we can deduct:
𝑌𝑡 = ∑∞ −1 ∞
𝑗=0(ѱ𝑗 𝑃)(𝑃 𝜖𝑡−𝑗 ) = ∑𝑗=0 Ф𝑗 𝜔𝑡−𝑗 (9)
Equation (9) yields responses to a structural dynamic impact following a structural shock 𝜔𝑡 :
𝜕𝑌𝑖,𝑡+𝑠
Ф𝑖𝑗,𝑠 = (10)
𝜕𝜔𝑗,𝑠
112
Furthermore, the structural VAR form (SVAR) can be derived from the reduced form as:
𝑌𝑡 = ∑∞
𝑗=0 𝐴𝑗 𝑌𝑡−𝑗 + 𝜔𝑡 (11)
where matrices 𝐴𝑗 and the structural residual variances of 𝜔𝑡 are estimated by multiplying the
̂𝑗 for 1 ≤ 𝑗 ≤ 𝑝
estimated reduced form terms of Equation (1) by 𝑃̂−1 ; 𝐼 − 𝐴0 = 𝑃̂−1 ; 𝐴̂𝑗 = 𝑃̂−1 Ф
and 𝑉𝑎𝑟(𝜔𝑡 ) = 𝑃̂−1 𝛺(𝑃̂−1 )′. Consequently, the estimation of SVAR is assured once matrix 𝑃 has
been estimated. This matrix represents 𝑛2 unknown parameters comprising 𝑛(𝑛 − 1)/2
identifying constraints mandatory to estimate the SVAR model. These constraints must be imposed
a priori for the estimation of the model.
3.2. Application
For our application, we run a bivariate SVAR linking government expenditure to the GDP to study
the effects/multipliers of government consumption on the output. In order to control exogenously
for the business cycle (expansion versus recession), debt to GDP evolution (accumulation versus
reduction) and the existence of both debt accumulation/reduction under expansion/recession, the
SVAR is augmented by variable dummies corresponding to each of the previous prescribed states,
hence becoming an SVAR-X (X for exogenous).
Following the previous general methodology, the SVAR, in our case, linking two endogenous
stationary variables describing, respectively, the relationship between government expenditure
(𝑔𝑡 ) and GDP (𝑦𝑡 ) for each country, is formulated as:
𝑔𝑡 + 𝛽1,2 𝑦𝑡 = 𝑐1,0 + 𝑐1,1 𝑔𝑡−1 + 𝑐1,2 𝑦𝑡−1 + 𝜀𝑔,𝑡

{ (12)62
𝑦𝑡 + 𝛽2,1 𝑔𝑡 = 𝑐2,0 + 𝑐2,1 𝑔𝑡−1 + 𝑐2,2 𝑦𝑡−1 + 𝜀𝑦,𝑡
where 𝜀𝑔,𝑡 and 𝜀𝑦,𝑡 are, respectively, the structural shocks/innovations of the first and second
variables in this bivariate SVAR, and could be formulated as:
0 𝜎𝑔 0
(𝜀𝜀𝑔,𝑡) = 𝜀𝑡 ≈ 𝑖. 𝑖. 𝑑. (( ) , ( )) (13)
𝑦,𝑡 0 0 𝜎𝑦
The real government consumption/expenditure and the real output (GDP) are considered in log
differentiated natural logarithm, hence designing the growth rate of the corresponding variables
and allowing direct interpretation of simultaneous parameters as elasticities assigned to these
𝛽
variables in the SVAR equations (i.e. (𝛽1,2 ) = 𝛽). Thus, Equation (1) can be formulated as:
2,1
62
The lag order adopted is 1, which is confirmed in the results section by the appropriate tests discussed in the general
methodology.
113
1 𝛽1,2 𝑔𝑡 𝑐1,0 𝑐1,1 𝑐1,2 𝑔𝑡−1 𝜀𝑔,𝑡

[ ] [𝑦 ] = [𝑐 ] + [𝑐 𝑐2,2 ] [𝑦𝑡−1 ] + [𝜀𝑦,𝑡 ] (14)
𝛽2,1 1 𝑡 2,0 2,1
which could be also expressed in the form of:
𝐵𝑣𝑡 = 𝐶0 + 𝐶𝑣𝑡−1 + 𝜀𝑡 (15)
We deduce the reduced form of the SVAR, called a standard VAR model, by multiplying equation
(15) by the inverted matrix 𝐵−1 , assuming it exists, and solving for 𝑣𝑡 in terms of 𝑣𝑡−1 and 𝜀𝑡 :
𝑣𝑡 = 𝐵−1 𝐶0 + 𝐵−1 𝐶𝑣𝑡−1 + 𝐵−1 𝜀𝑡 = 𝑎0 + 𝐴𝑣𝑡−1 + 𝑢𝑡 (16)
Or equivalently, 𝐴1 (𝐿)𝑣𝑡 = 𝑎0 + 𝑢𝑡 with 𝐴1 (𝐿) = 𝐼 − 𝐴𝐿.
We can easily deduce the residuals 𝑢𝑡 as a linear combination of the structural errors 𝜀𝑡 :
1 𝜀𝑔,𝑡 − 𝛽1,2 𝜀𝑦,𝑡

𝑢𝑡 = 𝐵−1 𝜀𝑡 = [ ] (17)
(1−𝛽1,2 𝛽2,1 ) 𝜀𝑦,𝑡 − 𝛽2,1 𝜀𝑔,𝑡
Thus:
𝑣𝑡 = 𝜆 + Ψ(L)𝑢𝑡 (18)
The structural moving average (SMA) representation of 𝑣𝑡 is based on an infinite moving average
of the structural innovations 𝜀𝑡 , deduced by substituting 𝑢𝑡 = 𝐵−1 𝜀𝑡 into equation (18), which leads
to:
𝑣𝑡 = 𝜆 + 𝛹(𝐿)𝐵−1 𝜀𝑡 = µ + Ф(𝐿)𝜀𝑡 (19)
where Ф(𝐿) = ∑∞ 𝑘
𝑘=0 𝜑𝑘 𝐿 .
In order to solve an SVAR, the parameters must be identified, which requires some restrictions to
be imposed. Typical identifying restrictions include either assuming no simultaneous equations
effects from one variable to another in the SVAR (for example, 𝛽1,2 = 0 or 𝛽2,1 = 0) or linear
restrictions on the elements of the matrix (for example, 𝛽1,2 + 𝛽2,1 = 0). In our case, we follow
the methodology of Blanchard and Perotti (2002) by identifying government spending shocks
using a Cholesky decomposition, ordering government spending first as the variable that is clearly
the most exogenous compared to GDP.63 For our case, we are only interested in government
multipliers, and no tax multipliers are considered in the current chapter. As explained in the general
methodology, the number of restrictions needed is determined by the number 𝑛 of endogenous
63
Contrary to our bi-variate case, the Blanchard and Perotti (2002) is a tri-variate SVAR linking three variables: tax
revenue, government expenditure and GDP.
114
variables of the VAR by the formulae, 𝑛(𝑛 − 1)/2. Thus, for a bi-variate model, the number of
𝑐 0
restrictions is only 1 (2(2 − 1)/2). Then, the matrix of shocks after restrictions is [𝑐1,1 𝑐 ].
2,1 2,2
In our restriction, we especially consider that the response of government expenditure to the output
comes with a lag, which means no contemporaneous effects of GDP to government expenditure.
Thus, the coefficient 𝛽1,2 = 0. This is also interesting, as the reverse instantaneous causality from
GDP to expenditure may alter, deducing the effect, ceteris paribus, of government expenditure on
GDP (fiscal multiplier).
In order to draw fiscal multipliers, the formulae of impulse response functions are required. For
the bivariate SVAR model, taking the structural moving average (SMA) representation in Equation
(18) at the horizon time 𝑡 + ℎ, we have:
𝑔𝑡+ℎ 𝑐1,1 𝑐1,2 𝑔𝑡−1 𝜀𝑔,𝑡

[𝑦 ] = [𝑐 ] [
𝑐2,2 𝑦𝑡−1 ] + [ 𝜀 ] (20)
𝑡+ℎ 2,1 𝑦,𝑡
Fiscal multipliers are drawn from structural shocks assigned to each variable; in particular, we are
interested in the effect of structural fiscal (expenditure) shocks on GDP in this case. For this
purpose, we consider the structural moving average (SMA) representation of the SVAR. At the
horizon time 𝑡 + ℎ, the SMA representation is:
𝑔𝑡+ℎ 𝜑0 0
𝜑1.2 𝜀𝑔,𝑡+ℎ ℎ
𝜑1.1 ℎ
𝜑1.2 𝜀𝑔,𝑡
[𝑦 ] = [ 1.1
0 0 ] [ 𝜀𝑦,𝑡+ℎ ] + ⋯ + [ ℎ ℎ ] [𝜀𝑦,𝑡 ] + ⋯ (21)
𝑡+ℎ 𝜑2.1 𝜑2.2 𝜑2.1 𝜑2.2
Then the structural dynamic multipliers are:
𝜕𝑔𝑡+ℎ ℎ
= 𝜑1.1 (22.a)
𝜕𝜀𝑔,𝑡
𝜕𝑦𝑡+ℎ ℎ
= 𝜑2.2 (22.b)
𝜕𝜀𝑦,𝑡
𝜕𝑔𝑡+ℎ ℎ
= 𝜑1.2 (22.c)
𝜕𝜀𝑦,𝑡
𝜕𝑦𝑡+ℎ ℎ
= 𝜑2.1 (22.d)
𝜕𝜀𝑔,𝑡
The structural dynamic multipliers/impacts measure how a unit impulse of the structural shocks at
time 𝑡 affects the level of endogenous variables at the horizon time 𝑡 + ℎ. In particular, the two
first equations (22.a and 22.b) represent the response of, respectively, government expenditure and
GDP growth rates to their proper innovations. The two other equations (22.c and 22.d) assess the
crossing effects of the structural innovations between the endogenous variables of the SVAR. In
particular, Equation (22.d) represents the response of the GDP growth rate to a structural unit
115
shock of government expenditure, which will be our emphasis in this application. Drawing the
ℎ
structural dynamic impacts 𝜑𝑖.𝑗 for the shocks (𝑖, 𝑗) = (1,2) allows us to visualize such dynamic
impacts in what is referred to as the impulse response functions (IRFs). For the cumulative effects
of the structural shock impacts, since the SVAR is designed to be stationary, which means that the
ℎ ℎ
effects 𝜑𝑖.𝑗 fade away in the long term (i.e. lim 𝜑𝑖.𝑗 = 0), the long-term cumulative impact of the
ℎ→∞
structural shocks is captured by the instant IFRs to infinity, which means:
ℎ
∅ = ∑∞
ℎ=0 𝜑𝑖.𝑗 ; (𝑖, 𝑗) = (1,2) (23)
The structural dynamic multipliers (short-term or long-term cumulative) defined above are
different from the Keynesian concept of the fiscal multiplier, generally associated with the general
theory of John Maynard Keynes (1936). The latter is defined as the output change in response to
a (exogenous) change in a fiscal variable in reference to their baseline levels (Spilimbergo et al.,
2009; Coenen et al., 2012). Hence, for 𝐺𝑡 and 𝑌𝑡 denoting, respectively, the fiscal instrument
(government expenditure here) and the output at time 𝑡, the Keynesian or simply fiscal multiplier
∆𝑌𝑡
is expressed as . Or, while the effects come with different lag times, the cumulative fiscal
∆𝐺𝑡
𝑗=ℎ
∑𝑗=0 ∆𝑌𝑡+𝑗
multiplier to the time horizon ℎ is expressed by: 𝑗=ℎ (Chinn, 2013).
∑𝑗=0 ∆𝐺𝑡+𝑗
To compare our results with the findings in the literature and across countries, an exercise of
mapping the IRF impacts to Keynesian fiscal multipliers is required. In the explicit SVAR, the
government expenditure variable, as well as GDP, are introduced in per cent of first differences of
the natural logarithm for the corresponding levels of the variables (i.e. the growth rates in per cent).
The unit root augmented Dickey-Fuller and Phillips-Perron tests show that these variables are
integrated of order one in levels. Thus, using the first difference of logarithms ensures stationarity
of such variables. Furthermore, introducing the variables in logarithms allows us to draw the
Keynesian multipliers directly from the effects of elasticities. Letting 𝜇𝑌/𝐺 define the elasticity of
GDP to government expenditure,64 we have:
𝑑𝑙𝑜𝑔(𝑌𝑡) ∆𝑌𝑡 𝐺𝑡 𝐺𝑡
𝜇𝑌/𝐺 = = × =𝑘 (24)
𝑑𝑙𝑜𝑔(𝐺𝑡) ∆𝐺𝑡 𝑌𝑡 𝑌𝑡
∆𝑌𝑡
The Keynesian multiplier 𝑘 = measuring government expenditure effect on GDP, is then
∆𝐺𝑡,
deduced as the elasticity of GDP to government expenditure scaled by 𝐺 ̅̅̅̅̅̅̅
𝑡 ⁄𝑌𝑡 representing the
̅̅̅̅̅̅̅
averaged share of government expenditure in GDP (or multiplied by (𝑌𝑡 ⁄𝐺𝑡 )) (Gonzales-Garcia et
al., 2013; Ilzetzki et al., 2013; Barnichon and Matthes, 2017; Priftis and Zimic, 2018; Glocker et
al., 2019). However, with the latter references, even though they scale their impact IRFs by share
64
Razzak and Bentour (2013) use this approach to deduce foreign direct investment return from estimated elasticities
of the Cobb-Douglas production function.
116
of consumption, the results are meaningful in terms of size only if the structural shock is expressed
in percentage units. The reason for this is that the structural innovations, especially when using
Cholesky innovations in an SVAR, are expressed in standard deviation units. Therefore, in
practice, for accuracy of results, the impacts should also be scaled by a standard deviation 𝜎𝑔 of
the fiscal variable (government expenditure), as in Combes et al. (2014). Following this precision,
an adjustment coefficient is defined to deduce the short-term (immediate) fiscal (Keynesian)
multiplier from the corresponding Cholesky impact multiplier, as:
̅
𝑌
𝑘 𝑠𝑟 = 𝐼𝑀 𝑠𝑟 × 𝐺
(25)
𝜎𝑔
𝜕𝑦𝑡 0
where, from Equation (22.d), 𝐼𝑀 𝑠𝑟 = = 𝜑2.1 is the immediate effects of government
𝜕𝜀𝑔,0
expenditure Cholesky innovations. For the accumulated (long-term) expenditure multipliers 𝑘 𝑙𝑟 ,

they are deducted in the same way as:
̅
𝑌
̅̅̅̅̅
𝑘 𝑙𝑟 = 𝐼𝑀𝑙𝑟 ∗ /𝜎𝑔 = (∑∞ ℎ
ℎ=0 𝜑2.1 ) ∗ 𝑌/𝐺 /𝜎𝑔 (26)65
𝐺
In the previous bivariate SVAR, the effect of public debt on the expenditure multipliers is
exogenously tested by a dummy representing the way the public debt ratio is evolving. In the
second case, we endogenize the public debt effects and introduce the public debt to GDP ratio in
a tri-variate SVAR linking government consumption, GDP and government debt ratio. For the
identification in this tri-variate model, two other restrictions are needed. These are simply imposed
by assuming that both government consumption and GDP do not have an immediate
(simultaneous) effect on the public debt ratio. Thus, the only structural coefficients assumed to be
non-null are those capturing the public debt effects on the other variables in the model, while the
opposite effects (feedback effects) are delayed by at least one quarter. For the formulations
(equations, IRFs, etc.), the methodology is the same as for the previous bi-variate, or as described
by the general methodology.
4. Data
4.1. Data source
We first solicited the database of the Federal Reserve Bank of Saint Louis for the quarterly data
on real government consumption, GDP and public debt, displayed on their website free of charge,
and downloaded country by country, where data is seasonally adjusted. We noticed that this data,
65
Other authors used formulae without mentioning any normalization with reference to the volatility of the fiscal
instrument (𝜎𝑔 ) (Ilzetzki et al., 2013; Priftis and Zimic, 2018).
117
which was not available for all the 18 sample countries, has the OECD database as its main source.
We therefore avoided the Federal Reserve Bank of Saint Louis data for the two first models and
downloaded constructed national accounts of the 18 countries displayed in the OECD database.
For the last model applied to the United States and requiring a long history, data was found and
downloaded for all variables from the Federal Reserve Bank of Saint Louis website.
Compared to annual data, high-frequency data, especially quarterly data, is considered to be the
most important for assessing fiscal policy effects (Ilzetzki et al., 2013). However, some issues also
arise when using such data. The availability of the observed quarterly national account in many of
the advanced countries is recent and goes back to the 1990s. Although the data is displayed for the
general government consumption and the GDP back to the 1960s, these are estimations rather than
observations going back to pre-1990s, as indicated in the OECD database comments. The
exception is the United States, the United Kingdom, Canada, Norway and France, where data goes
back to before 1990 (Table 1.C in the appendix). Table 2.C presents descriptive statistics for the
economic growth rate and government expenditure. The same issue of observed sample data is
encountered with government debt, where the observed data starts. for the majority, in the late
1990s (Table 3.C in the appendix), but for the latter variable, there is no estimation back in time.
This constrained our estimations for these countries to the corresponding period (except for the
United States), where the quarterly public debt is available when the latter is considered an
estimation, whether as an exogenous variable or as an endogenous one.
4.2. Preliminary analysis of some previous fiscal multipliers with relation to public
debt
In this sub-section, we undertake an exercise analysing some previous expenditure multipliers in

links with public debt ratios, from previous works for 27 European countries. We especially
investigate the calculations and results of Boussard et al. (2012) and Berti et al. (2013). We draw
scatter plots showing short-term expenditure critical multipliers (first year) and public debt to GDP
ratios in 2011 and 2012 (Figures 2 and 3) for the 27 European Union countries. The calculations
are simulated under fixed interest rates. The 27 countries are Belgium (BE), Bulgaria (BG), Cyprus
(CY), the Czech Republic (CZ), Denmark (DK), Estonia (EE), Finland (FI), France (FR), Germany
(DE), Greece (EL), Hungary (HU), Ireland (IE), Italy (IT), Latvia (LV), Lithuania (LT),
Luxembourg (LU), Malta (MT), The Netherlands (NL), Poland (PL), Portugal (PT), Romania
(RO), Slovakia (SK), Slovenia (SI), Spain (ES), Sweden (SE) and the United Kingdom (UK).
The fiscal multipliers decrease exponentially with public debt to GDP ratios with an
“elasticity/country” of -0.539 and -0.597 in 2011 and 2012, respectively. The trendline of the
scatter plot is compatible with a negative power curve, with a high coefficient of determination of
around 90.8% and 92.6% for the two figures, respectively. We highlight the countries not affiliated
with the eurozone countries (with red dots), as well those that joined the eurozone recently (green
dots) and the 12 euro core member countries that joined before 2007 (blue dots). The latter group
118
of countries is more concerned with high public debt and lower fiscal multipliers of less than 1,
except for the case of Luxembourg. Figures 2 and 3 show clearly that all countries with a public
debt to GDP ratio exceeding 60% have fiscal multipliers lower than 1.
These figures clearly show a quite different message to the conclusion of the two papers, namely,
that large, short-term multipliers are likely to increase debt ratios under initial high public debt. It
shows an picture of which high public debt is associated with low spending multipliers, which also
raises a legitimate question. On the one hand, could high public debt also have led to lower fiscal
multipliers (crowding-out effects, for example)? And it raises a similar debate to the one detailed
in Chapter 1 between public debt and economic growth feedback effects (reverse causality). On
the other hand, according to these results, fiscal expansion, as opposed to fiscal consolidation, is
expected to be less effective under high public debt levels, particularly as the multipliers assumed
to be calculated from an exercise of positive fiscal shocks are lower than unity for highly indebted
countries (thus assuming assymetry –this issue is discussed in Section 2.4).
Figure 2. Critical multipliers for EU member states versus the public debt to GDP ratios – year 2011
3.5
Trendline equation: y = 8.5265x-0.539
3 Coefficient of determination: R² = 0.908
EE
2.5
2 BG
RO LT
LU LV
1.5
SK
SI PL
CZ CY; MT
ES
1 HU UK
SW FI
DK IE; PT
NL AT
DE BE
0.5 FR IT
GR
0
0 20 40 60 80 100 120 140 160 180
• Euro member joined before 2007 • Euro member joined after 2007 • Non euro member
Source: Constructed from results of the European Commission working papers reported in Bouassard et al. (2012) –
Table 3 – and Berti et al. (2013) – Table 2.
119
Figure 3. Critical multipliers for EU member states versus the public debt to GDP ratios – year 2012
3
Trendline equation: y = 10.778x-0.597
Coefficient of determination: R² = 0.9257
EE
2.5
2 BG
LU RO
1.5
LT, LV
SK
CZ PL
1 SI NL ES
SW AT
CY FR
DK FI PT
MT HU
DE BE
0.5 UK IT GR
IE
0
0 20 40 60 80 100 120 140 160 180
• Euro member joined before 2007 • Euro member joined after 2007 • Non euro member
Source: Constructed from results of the European Commission working papers reported in Bouassard et al. (2012) –
Table 3 – and Berti et al. (2013) – Table 2.
5. Results
In this section, we display the results of the models discussed in the methodology section and argue
our results with a robustness check based on the United States data set using a more detailed SVAR
with six endogenous (including monetary and fiscal) variables.
We used unit-root tests (augmented Dicky-Fuller and Phillips-Peron) for the stability of the
variables, which confirmed that all the variables are integrated of order 1. A summary of the
stationarity tests is presented in Table 4.C in the appendix for the three variables of GDP,
government consumption and public debt to GDP ratios. Thus, we introduced all the variables in
the three models in first differences of the natural logarithm of such variables, except for prices
(the interest rate and inflation). The sample of countries that was considered comprises 18
advanced countries, of which the majority are eurozone member countries, namely: Austria,
Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, The
Netherlands, Norway, Portugal, Spain, Sweden, the United Kingdom and the United States.
Furthermore, following Blanchard and Perotti (2002), we chose not to test for any long-term
cointegration relationship, as this might also complicate the exercises of SVAR methodology,
especially the way of resolving identification issues, and might deviate from comparing our results
to the leading literature and the model results adopting the Blanchard and Perotti (2002) approach.
120
Although some pioneer researchers did not test for the number of lags to introduce in their SVAR,
simply adopting an SVAR with one lag, we ran the exercise of the determination of such a lag for
each country and each model (Table 5.C in the appendix for models with no control of exogenous
effects of business cycle or debt movements). We found that 15 out of 18 countries have at least
one criterion that indicates that the optimal lag is 1 (more likely indicated by Schwarz information
criterion, SC). The three countries are Denmark, Greece and Japan. When controlling, for example,
for the business cycle, Greece, Japan and Spain have an order of lags superior to 1 for models in
times of recession, while in times of expansion, Austria, Japan, Portugal and Italy have order lags
of 2 to 3. As a result of the multiplicity of the exercises undertaken here for each country
individually (control for the business cycle, debt movements, etc.), and based on the higher number
of countries pointing to only one lag, we preferred to follow the same approach as other researchers
who fixed the model for all the countries to a unique optimal minimum lag equal to 1.
5.1. The effects of the time periods
Do fiscal multipliers tend to be lower in recent periods than those of the previous decades of the
1960s and 1970s?
There are a set of determinants revealed in the economic literature that may work in reducing the
size of the fiscal multipliers. First, there is the increase in trade openness: more closed economies
tend to have higher multipliers (Barrell et al., 2012; Ilzetzki et al., 2013; Batini et al., 2014).
Second, there is labour market flexibility: the more flexible the labour market is, the larger the
fiscal multiplier. For example, based on this, it is expected that Europe will show higher multipliers
than the United States, as the market in the former tends to be rigid, with the presence of stronger
labour syndicates and unions. Rigidities play against wage flexibility, which tend to reduce the
response of output to demand shocks (Cole and Ohanian, 2004; Gorodnichenko et al., 2012; Batini
et al., 2014). Third, there is the size of automatic stabilizers: larger automatic stabilizers tend to
reduce fiscal multipliers, by offsetting part of the initial fiscal shock (Dolls et al., 2012). Fourth,
the flexibility of the exchange rate regime tends to lower the multiplier size, as the movements of
the exchange rate may cushion the effects of fiscal policy actions (Born et al., 2013; Ilzetzki et al.,
2013). Fifth, the fiscal position, with high public debt and fiscal deficit widening, reduce the size
of the multipliers (Ilzetzki et al., 2013; Bi et al., 2016; Huidrom et al., 2016; Kirchner et al., 2010).
Sixth, there is the effect of the active monetary accommodation to fiscal shocks, where
expansionary monetary policy can offset the impact of fiscal contraction on demand.
Based on this, fiscal multipliers are likely to be smaller in recent decades, known as the “Great
Moderation Era”, especially the period 1986–2007, than the pre-1986 period. The reason for this
is that, in this period, all the factors previously cited have been strengthened. The degree of
openness has increased with the proliferation of the free trade agreements and increased
international financial and economic integration. Many exchange rate systems have been switched
to greater flexibility, except for countries that have adopted monetary unions. The monetary policy
121
has been more active in fulfilling its role in stabilizing economies, among other things. All this
leads us to think that fiscal multipliers may have a smaller size in the recent period than previously.
To examine these facts, we split our sample data for each country, 1966Q1–2019Q2, into two
balanced sub-samples, 1966Q1–1991Q4 and 1992Q1–2019Q2, with, respectively, 104 and 110
observations each. We ran a bivariate stationary SVAR with differentiated logarithm of
government expenditure and GDP (i.e. growth rates, in per cent) for all 18 countries for the 2
periods.
The results of the impulse response functions66 (the structural dynamic impacts) are drawn for this
exercise in Figures 1.B.a to 1.B.c in Appendix B. These results are also summarized in Table 3.A,
which shows that the main sample of countries (12 out of 18) tends to confirm that the expenditure
impacts are weak and substantially smaller in the first period than the second one. However, some
exceptions were recorded, especially for small-sized economies such as Finland, Greece, Ireland
and, to some extent, Italy, Portugal and Spain, particularly in the long term, which reported
opposite results: fiscal multipliers tend to be higher in recent periods than previously. This may be
in contrast to the idea that more openness decreases fiscal multipliers as the propensity to import
increases. However, not only the degree of openness that acts on the size of the multipliers, but
also other determinants, could play against the increase of fiscal multipliers, such as the monetary
policy accommodation and the exchange rate regimes (Batini et al., 2014). For a few other
countries, the impacts are even negative in the second period, namely, Canada, Denmark and
Germany in the long term, and France and the United States in the short and long terms. Overall,
for the first period, the multipliers average for the sample is around 0.96 (the impact is 0.18) for
the first quarter, 1.5 (0.26) for the accumulated fourth quarter (first year) and 1.57 for the
accumulated five years. However, for the second period (1992q1–2019q2), the corresponding
multipliers are reduced by more than half, recording on average in the sample 0.47, 0.54 and 0.66,
respectively for the first, the accumulated 4 and the accumulated 20 quarters (last row in Table
3.A).
Nevertheless, these results should be considered with caution for several reasons. The first, related
to the data construction method, is that the quarterly national accounts data for the government
expenditure variable, as well as for the GDP aggregate, is for many countries an estimation, rather
than an observation, in the first period. OECD data downloaded for the purpose of this exercise
displays a comment on each Excel cell data indicating whether the data cell is an observation or
estimation. We noticed that for all the countries’ samples – except for Canada, the United Kingdom
and the United States, for which data is observed from the first quarter of 1966, and Norway and
66
In all our applications, we reported the accumulated structural Cholesky IRFs, as defined in the methodology section,
0 ℎ
deduced from Equation 8 (the first impact is ∅0 = 𝜑𝑖.𝑗 ; the second accumulated impact ∅2 = ∑2ℎ=0 𝜑𝑖.𝑗 ; …; until the
q ℎ
long-term accumulated impact to time horizon q; ∅𝑞 = ∑ℎ=0 𝜑𝑖.𝑗 ). In all our applications, we considered 𝑞 = 20,
which corresponds to five years. The latter accumulated multiplier defined as the sum of effects to 20 quarters is to be
differentiated from what some authors reported as the maximum multiplier; namely, the peak of the effects attained
in a specific point of time.
122
France, for which the data observed starts in 1978 and 1980, respectively – the observed data starts
after the 1990s (mainly in 1995Q1 for the majority; see Table 1.C in the appendix). The quarterly
national account could be estimated using mechanical/statistical methods without any fiscal policy
feedback or any business cycle impacts on the data, which may deviate any fiscal policy
assessment from accurate outcomes.
The second reason, which is related to the values rather than the method of construction, is that
although the Cholesky impact multipliers are higher in the first period than in the period of
openness and financial globalization, the fiscal multipliers could be reduced or amplified between
the two periods. The trick resides in the coefficient of adjustment enabling the expenditure
multipliers to be obtained from the Cholesky impact multipliers. This coefficient has two
components, the first being the share of government consumption to GDP, and the second the
standard error of the growth rate of government consumption. The common tendency for all
advanced economies is for the shares of government consumption to grow as countries prosper
and the welfare state is enhanced. Direct factors of this are also related to an ageing population,
especially in Japan and many European countries. The increasing/decreasing shares of government
consumption could reduce/amplify the fiscal multipliers. The same is true for the second
component, which is the standard error of government consumption, which seems to be lower in
the second period than the first one (as opposed to the growing of the first component). Variables
are less volatile in the second period (the Great Moderation Era). The product of the two
components, which correct the Cholesky impacts to get spending multipliers, could then be higher,
less or approximately the same for each country between the two periods. Calculus on the
countries’ data shows that the adjusting coefficient (Table 2.A) is higher for all countries except
Ireland, Norway, Portugal and Spain, which may lead to a reduced gap (gap impacts shown by the
Cholesky innovations) in government consumption multipliers between the two periods.
A third issue is related to the method of rescaling by the average of the share of government
consumption to GDP (𝐺 ̅̅̅̅̅̅̅
𝑡 ⁄𝑌𝑡 ). This method, issued from the elasticity of the output of the fiscal
instrument, as explained in Equations (24), (25) and (26) and used by many authors (Ilzetzki et al.,
2013; Gonzales-Garcia et al., 2013; Priftis and Zimic, 2018; Glocker et al., 2019), may lead to
overestimated fiscal multipliers, which is the case here for many countries in periods of recession
and long-term accumulated cases. This fact is valid for our results in the current section and
subsequent sections, where some countries that have higher expenditure multipliers in the long
term under a recession could attain more than five (examples are France, Spain, Portugal). The
issue is because 𝑌/𝐺 can display large movements over the sample period (Ramey and Zubairy,
2018). To dampen this effect, some authors use an ex ante conversion approach (Gordon and
Krenn, 2010; Ramey, 2016; Barnichon and Matthes, 2017), which consists of re-scaling all the
𝑝
variables by an estimated "potential output" 𝑌𝑡 . Thus, the variables reconsidered for these authors
𝑝 𝑝
are 𝑌𝑡 /𝑌𝑡 , 𝐺𝑡 /𝑌𝑡 , and so on; for our case, we did not consider this issue and rather focused our
analysis on comparing changes in both structural impacts and multipliers, among the considered
123
cases (first period versus second period, expansion versus recession, debt accumulation versus
debt reduction, etc.), rather than focusing on the size of the fiscal multipliers.
5.2. The effect of the business cycle
We ran the same SVAR controlling for the business cycle in the current case. Some authors have
used the output gap to control for the business cycle position (Batini et al., 2012). For our case, we
used the growth of GDP, as used by Baum et al. (2012), noted 𝑔𝑔𝑟𝑡 , instead of the output gap, and
1 𝑖𝑓 𝑔𝑔𝑟𝑡 > 0
we defined a dummy variable for the business cycle, as follows: 𝑏𝑐𝑑𝑡 = { . The
0 𝑖𝑓 𝑔𝑔𝑟𝑡 ≤ 0
business cycle dummy indicates expansion, while its complement to unity is a proxy for recession.
The SVAR model is augmented by this variable exogenously (SVAR-X) for considering the
effects of the expansion periods only and its complement to unity to account for recessions.
The results of this exercise are displayed in Table 4.A, while the IRFs are plotted in Figures 2.B.a
to 2.B.c in Appendix B. The table reporting short-term (first quarter) and long-term (five years)
impacts, and their corresponding Keynesian multipliers, shows that these multipliers are either
positive and very low or negative and very low (in absolute value) for many countries in the sample
in times of expansion (exceptions are recorded for Greece and Ireland, where short-term
multipliers are, respectively, 0.92 and 1.02, corresponding to impacts of, respectively, 0.42 and
0.55). However, in times of recession, these impacts are all positive and amplified in size. The
maximum of the short-term Cholesky impact is recorded in Ireland by 1.53, corresponding to a
multiplier of (1.02),67 and in Norway by 0.82 (with a multiplier of 2.4). These higher impacts yield
multipliers greater than 1 for many countries (11 countries) and approaching unity for the rest
(between 0.52 and 0.92). On average, expenditure multipliers in the short term are near 0 (0.09) in
times of expansion, while they are more than 1 in recessions (1.55). The long-term accumulated
are negative in expansions (-0.24) and very high in recessions (4.8). Our results confirm the
conjecture of the higher expenditure multipliers in periods of recession than in those of economic
expansion, as revealed by the aftermath of the 2008 crisis literature, especially by Auerbach and
Gorodnichenko (2012, 2013). Another important point is that, from the IRF plots, the effects are
more persistent in times of recession than expansion, as convergence to the long term is more
quickly achievable in the latter than in the former (Figures 2.B.a to 2.B.c).
For a general view with relation to the effects of government debt, we plot the impact multipliers,
associated with the business cycle, versus debt ratios in Figure 3.B. It seems that these impacts are
slightly negatively associated with higher public debt ratios in recessions, while in expansions, the
opposite is observed, particularly for the long term. This means that higher debt may weigh on
expenditure multipliers in recession periods.
67
Although the Cholesky impact is higher, the multiplier is reduced by, in particular, the second component of the
adjusted parameter used as pass-through to fiscal multipliers in Table 2.A. Indeed, government consumption volatility
is higher (exceeding 2) for this country in the period of estimation.
124
5.3. The exogenous effect of the public debt accumulation/reduction
To control exogenously for the effect of public debt, we ran the previous bivariate SVAR, where
the accumulation is proxied by a dummy variable equal to 1 whenever the growth rate of public
debt to GDP ratio is positive, non-null and zero elsewhere. The public debt reduction case is
controlled in the SVARX by the complement of the debt accumulation dummy to unity. The
dummy variable for the debt ratio accumulation is defined according to the sign of the debt ratio
1 𝑖𝑓 𝑑𝑔𝑟𝑡 > 0
growth rate (𝑑𝑔𝑟𝑡 ): 𝑑𝑎𝑑𝑡 = { , where 𝑑𝑎𝑑𝑡 is the dummy indicating debt increase
0 𝑖𝑓 𝑑𝑔𝑟𝑡 ≤ 0
of the debt to GDP ratio (accumulation), and its complement to unity is a proxy for debt reduction.
We do not care whether the public debt ratio is reduced by the high performance of GDP (the
denominator) or accumulated because of weak GDP. In both situations, it is the ratio that is
important, reflecting the capacity to repay or not based on the performances of the economy.
Table 5.A shows the Cholesky impacts and their corresponding values of expenditure multipliers
under public debt movements: in the case where debt to GDP ratio is consequently accumulated;
or in the opposite case, where debt is reduced. Overall, the multipliers (impacts) tend to be higher
in times of debt accumulation than in times of reduction, except in a few countries, where the two
cases are approximately the same, namely, in Finland, Italy, Japan and Norway. The latter has
even larger multipliers in debt reduction cases than in debt accumulation. Convergence to the long-
term accumulated multiplier varies across countries, where it is fast in more than half of cases,
medium in around a third of cases and slow in a few cases (France, Spain and the UK). The
convergence is defined as fast if the accumulated long-term multiplier is approximately attained
in fewer than five quarters, medium if it is attained in between six and nine quarters, and slow
when it is above ten quarters. The size of the expenditure multipliers varies considerably between
countries.
The effects of the way the debt is moving, on multipliers, tends not to be different from the business
cycle effects, especially if we assume that, generally, the public debt ratio increases in times of
recession, which sounds more realistic, due to the double effects of GDP shrinking and the debt
level accumulation in times of recession. For further examination of this, an exercise combining
the business cycle effects jointly with the public debt movements was run. Table 6.A shows the
results. The main conclusion is that, under expansion, multipliers are very low for some countries
and negative for most countries, almost independently from the way the debt ratio is evolving. By
contrast, under recessions, multipliers are higher and could be larger than unity even in the short
term, as is the case for Belgium, Canada, Denmark, Greece, Ireland, Italy, Norway, Spain, Sweden,
the United Kingdom and the United States.
To obtain an overall view of the public debt effects on multipliers, we display, for the sample of
18 countries, scatter plots for the average public debt and the short- and long-term multiplier
values, under debt accumulation and debt recession cases, as illustrated in four charts presented in
Figure 4.B. It seems that, under debt accumulation, there is no clear association between the
125
average of the debt ratio (over the time period) and the multipliers’ size. However, a positive
correlation (albeit less strong) is observed in the case of debt reduction, which means that
expenditure multipliers are higher for high debt ratios conditioned with the case where the debt is
decumulated.
5.4. The effect of the public debt ratio movements jointly with the business cycle
Controlling for the business cycle effects and the public debt ratio movements is captured by
augmenting the SVAR by the product of the two corresponding dummies 𝑏𝑐𝑑𝑡 and 𝑑𝑎𝑑𝑡 yielding
the following four situations:
1- The effect of public debt accumulation under expansion, captured by (𝑏𝑐𝑑𝑡 .𝑑𝑎𝑑𝑡 ),
2- The effect of public debt accumulation under recession, captured by (1 − 𝑏𝑐𝑑𝑡 ).𝑑𝑎𝑑𝑡 ,
3- The effect of public debt reduction under expansion, captured by (1 − 𝑑𝑎𝑑𝑡 ).𝑏𝑐𝑑𝑡 ,
4- The effect of public debt reduction under recession, captured by (1 − 𝑏𝑐𝑑𝑡 ).(1 − 𝑑𝑎𝑑𝑡 ).
The results of the structural dynamic impacts (Cholesky IRFs), as well as the corresponding
multipliers, are presented in Table 6.A. The latter are deduced from Equation (13) for the short-
term (first) multiplier and Equation (14) for the long-term multiplier, as explained in the
methodology section. For the graphical IRFs, they are presented for each country by case of debt
movements under the business cycle. Therefore, Figures 6.B.a to 6.B.c present the impulse
response functions for the case of debt accumulation and the two business cycle cases. Similarly,
Figures 7.B.a to 7.B.c show the impulse response functions for the case of debt contraction and
the two business cycle cases.
From the results summarized for the short- and long-term impacts and multipliers in Table 6.A,
we observe that the effects are negative or positive but near 0 for many of the countries in times
of expansion, regardless of the debt development. Exceptions are recorded for Greece, Ireland and
Italy. However, in times of recession, all 18 countries have positive important multipliers, whether
under debt accumulation or debt reduction, except for The Netherlands and Portugal, which have
weak negative multipliers only in the case of debt reduction under recession. Under recession and
debt accumulation, many countries have expenditure multipliers higher than 1 in the short term,
namely, Belgium, Canada, Denmark, Greece, Ireland, Italy, Norway, Spain, Sweden, the United
Kingdom and the United States. However, under recession and debt reduction, the number of
countries with multipliers greater than 1 reduced to six countries, namely, Belgium, Italy, Japan,
Norway, Spain and the United States.
In order to visualize the previous results in a global view, we scattered the multipliers obtained for
these cases against the averages of the debt ratio (Figure 8.B). In two cases of joint expansion with
debt accumulation (the first two scatter plots for the short and long terms) and the recession with
126
debt reduction (the last two scatter plots), we observe a positive association of the fiscal multipliers
with the average of the debt to GDP ratio. However, there is no clear association for the cases of
expansion and debt reduction or recession and debt accumulation. This means that multipliers tend
to be higher for countries with high debt to GDP ratios, either in times of expansion, and the debt
being accumulated, or times of recession, and the debt ratio being reduced.
5.5. The endogenous effect of public debt
Controlling for the business cycle effects, and considering the endogenous public debt ratio, by
the tri-variate SVAR model, the results confirm the previous results of the large multipliers under
recession, while reporting weak or even negative multipliers in periods of expansion. Tables 7.A
and 8.A present, respectively, the dynamic structural impacts and the associated expenditure
multipliers for the first quarter (short term), first year, second year and fifth year (long term). The
IRFs of these results are shown in Figures 9.B.a to 9.B.c. These results tend to confirm those
reported for the case of the bi-variate model, while controlling exogenously for government debt.
Furthermore, for the short- and long-term impacts, a positive association with the average of the
debt to GDP ratios is noticed in expansion cases, as reported in the scatter plots in Figure 10.B.
We can conclude generally from the previous results of the bi-variate and tri-variate models that:
in times of recession, multipliers tend to be higher than in times of expansion, but they tend to
decrease with debt reduction rather than debt accumulation in times of recession. This may lead
us to consider the self-defeating effects of austerity aimed at reducing public debt and based on
expenditure cuts, as it tends to reduce the higher multipliers recorded in times of recession and
high public debt.
5.6. Robustness check: the public debt crowding-in/out effects
In order to examine why spending multipliers are lower in times of expansion than in recessions,
a more disaggregated model containing behavioural equations is required. For this purpose, we
chose to study, in particular, whether the crowding-out effect that might be behind lowering
spending multipliers originates from public debt. We particularly consider an SVAR with six fiscal
and monetary variables for the United States.
In this section, we present the SVAR with six endogenous variables applied to the United Sates
only, as the country with a large quarterly data set available (1966Q1–2019Q2). This SVAR is
intended to explain and check the robustness of the previous models’ outputs, where other
variables representing monetary policy aggregates and private-sector investment are introduced,
leading to more interactions catching economic behaviour. The endogenous variables considered
are the interest rate, the public debt ratio, prices, output, government expenditure and private
investment. In this model, we follow nearly the same approach as in Sims (1986). The difference
from Sims’s model is that the latter considered money supply equation (which is an interest rate
127
equation explained by money demand), money demand, output, price, unemployment and
investment demand equations (the model is detailed in Sims (1986)).
In our case, we modified the model to include fiscal variables that are of interest to us for studying
the spending multipliers via the impact of expenditure and public debt. Moreover, in order to assess
what is happening on the private demand side, especially the possibility of crowding-in/out effects
of government spending and/or debt to private agent decisions, we kept the private investment
equation. Private consumption could also be considered but we decided to reduce the size of this
model into six variables to gain more degrees of freedom for the quality purpose estimation. The
monetary policy action is presented by the equation of interest rate, and the dynamic of prices is
captured by the inflation equation.
For the interest rate equation, we used the policy variable, which is the effective federal funds rate.
Assuming that the feds follows a conventional monetary policy based on Taylor’s rule, it seems
suitable to assume that the feds policy rate (𝑟𝑡 ) is determined by innovations (𝑒3 , 𝑒4 ) corresponding,
respectively, to GDP growth rate (𝑦𝑡 ) and inflation (𝑡 ). GDP is a best proxy for the output gap, as
we do not consider the output gap in this model. The second equation is related to public debt (𝑑𝑡 ).
For this equation, three variables are important from the classical debt sustainability rule; we could
realistically assume that the public debt ratio is determined by the innovations (𝑒3 , 𝑒4 , 𝑒1 )
corresponding, respectively, to innovations from the GDP growth rate, the inflation rate and the
interest rate.68 The third equation is the output assumed to rely on innovations, 𝑒5 , 𝑒6 and 𝑒2 , from,
respectively, private investment (𝑖𝑡 ), government expenditure (𝑔𝑡 ) (three options are tested: total,
current and capital expenditure) and public debt. For the fourth equation of the system
corresponding to inflation, the latter is assumed to be determined by the interest rate’s innovations
(𝑒1 ) and the investment private innovations (𝑒5 ). The private investment is assumed to be
determined in the fifth equation as the function of innovations coming from the output (𝑒3 ) and
the interest rate (𝑒1 ) (the accelerator equation). The last equation is an error term corresponding to
government expenditure determined by its proper structural innovations (𝑒6 ). This means that the
government expenditure variable does not react simultaneously to the other endogenous variables
in this model, but its reaction comes with a delay. We considered the following formulations for
the equations of this model, such as:
𝑟𝑡 = 𝑓(𝑦𝑡 , 𝜋𝑡 , 𝑒1𝑡 ); 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑟𝑎𝑡𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (1)

𝑑𝑡 = 𝑓(𝑦𝑡 , 𝜋𝑡 , 𝑟𝑡 , 𝑒2𝑡 ); 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡 𝑟𝑎𝑡𝑖𝑜 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (2)
𝑦𝑡 = 𝑓(𝑖𝑡 , 𝑔𝑡 , 𝑑𝑡 , 𝑒4𝑡 ); 𝑜𝑢𝑡𝑝𝑢𝑡 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (3)
𝜋𝑡 = 𝑓(𝑟𝑡 , 𝑖𝑡 , 𝑒5𝑡 ); 𝑖𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (4)
𝑖𝑡 = 𝑓(𝑦𝑡 , 𝑟𝑡 ); 𝑝𝑟𝑖𝑣𝑎𝑡𝑒 𝑖𝑛𝑣𝑒𝑠𝑚𝑒𝑛𝑡 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (5)
{ 𝑔𝑡 = 𝑓(𝑒3𝑡 ); 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (6)
68
It is possible to provide an SVAR with identities equations such as Taylor’s rule for the interest rate and the debt
sustainability equation. However, the structural shocks associated with those identities would be zero, and the situation
is more complex if the identity is dynamic (Cherif and Hasanov, 2017; Ouliaris et al., 2018). As assessment of the
effects of fiscal and monetary policy structural innovations (shocks) on the other endogenous variables is our
requirement, we keep such identities as functional structural equations.
128
Explicitly, the six simultaneous equations of the current model are formulated in the following
system, with all the variables log-differentiated:
𝑟𝑡 = 𝑐(1)𝑦𝑡 + 𝑐(2)𝜋𝑡 + 𝑒1𝑡 ; 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑟𝑎𝑡𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (1)

𝑑𝑡 = 𝑐(3)𝑦𝑡 + 𝑐(4)𝜋𝑡 + 𝑐(5)𝑟𝑡 + 𝑒2𝑡 ; 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑑𝑒𝑏𝑡 𝑟𝑎𝑡𝑖𝑜 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (2)
𝑦𝑡 = 𝑐(6)𝑖𝑡 + 𝑐(7)𝜋𝑡 + 𝑐(8)𝑟𝑡 + 𝑒4𝑡 ; 𝑜𝑢𝑡𝑝𝑢𝑡 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (3)
𝜋𝑡 = 𝑐(9)𝑟𝑡 + 𝑐(10)𝑖𝑡 + 𝑒5𝑡 ; 𝑖𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (4)
𝑖𝑡 = 𝑐(11)𝑦𝑡 + 𝑐(12)𝑟𝑡 + 𝑒6𝑡 ; 𝑝𝑟𝑖𝑣𝑎𝑡𝑒 𝑖𝑛𝑣𝑒𝑠𝑚𝑒𝑛𝑡 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (5)
{ 𝑔𝑡 = 𝑒3𝑡 ; 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (6)
In this exercise, all the variables are made stationary by differentiated natural logarithms, and
hence the variables are all in growth rates. The variables, which will appear in outputs and figures,
are LGDP for GDP growth rate (𝑦𝑡 ), LGE for government consumption expenditure, total, current
and capital (𝑔𝑡 = 𝐿𝐺𝐸 = 𝑑𝑙𝑜𝑔(𝐺𝑂𝑉𝐸𝑋𝑃)), LPRC for inflation (GDP deflator inflation, 𝜋𝑡 =
𝐿𝑃𝑅𝐶 = 𝑑𝑙𝑜𝑔(𝐺𝐷𝑃𝑑𝑒𝑓𝑙𝑎𝑡𝑜𝑟)), LDR for log differentiated of the debt ratio (𝑑𝑡 =
𝑑𝑙𝑜𝑔(𝑑𝑒𝑏𝑡𝑟𝑎𝑡𝑖𝑜)), LPINV for the real private investment growth rate (𝑖𝑡 = 𝐿𝑃𝐼𝑁𝑉 =
𝑑𝑙𝑜𝑔(𝑟𝑝𝑖𝑛𝑣)) and RINTR for the interest rate (𝑟𝑡 = 𝑅𝐼𝑁𝑇𝑅 = 𝑓𝑓𝑟).
We test the SVAR model with six endogenous variables applied to the United Sates only, as the
country with a large quarterly data set. The data set covers, for many variables of this model, a
long-observed history of quarterly data going back to 1953. However, the public debt quarterly
data starts at 1996. Hence, the model is estimated over the period 1966Q1–2019Q2. This SVAR
is intended to check the robustness of the previous models’ outputs, where other variables
representing monetary policy aggregates and private-sector investment are introduced, leading to
more interactions catching economic behaviour. The model functional equations’ determinants are
discussed in the methodology section. The considered endogenous variables are interest rate,
public debt ratio, prices, output, government expenditure and private investment.
In this SVAR we control exogenously for the business cycle and debt movements by introducing
as exogenous the dummies controlling for expansion/recession and debt accumulation/reduction,
as defined for the previous bivariate model. We also produce the cases where debt movements and
business cycle are jointly considered (four cases).
Figure 8 presents a panel of charts showing quarterly data over 1966Q1–2019Q2, by row order
and column, from left to right, the public debt ratio development and output growth, the long-term
interest rate and the inflation rate for the three first charts. The last chart shows the composition of
the total government expenditure, capital and current expenditure growth rates. The dashed grey
areas in all four charts of Figure 8 indicate the recession periods where the quarterly negative
growth is recorded.
The public debt of the United States stands at around US$ 20.42 trillion at the end of June 2019,
from which: 1) about 66% is long-term liabilities, 2) more than 83% is denominated in domestic
currency and 7% in foreign currency, and the rest is not allocated (see Table 6.C). For evolution
129
over a long history (Figure 8), we observe that the growth rate is more volatile in the 1966–84
period, but with the public debt ratio in a downward trend. Staring around 1985, the real growth
rate becomes less volatile than previously, while the public debt ratio reverses track to generally
increase. For the relationship between public debt and interest rate, although the golden rule of
public debt and economic growth stipulates that public debt is accumulated whenever the real
interest rate is higher than economic growth, the public debt ratio and interest rate are evolving the
opposite way. From the 1960s to early 1980s, the interest rate takes an upward trend, while the
public debt to GDP ratio is on a downward trend. Starting from the 1980s up to 2019 the interest
rate records a sustained decrease, while the public debt ratio reverses its path to a general upward
trend (except 1996 until 2001, where it decreases). The same facts are observed when comparing
the trend of inflation and GDP growth rate. The period starting from 1986 is known by economists
as the “Great Moderation Era”. In this period, we can conclude from the previous analysis that
public debt in the United States has accumulated over this period of sustained growth and moderate
inflation and interest rates, which is the case for many advanced countries in our sample. For the
last chart in Figure 8, capital expenditure is less volatile in the second period (1986–2019) than
the first period (1966–85), compared to current expenditure, which means that current expenditure
is more mobilized in times of crisis in recent periods than in the decades of the 1960s and 1970s.
Figure 4. Evolution of some SVAR variables vis-à-vis public debt for the United States
110 110
100 100
90 90
80 80
70 70
60 60
50 50
40 40
.04 16
30 30
14
.03
12
.02
10
.01
8
.00
6
-.01 4
-.02 2
-.03 0
66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18
Real GDP growth rate (%, left scale) Long term interest rate (%, left scale)
Public debt to GDP ratio (%, right scale) Public debt to GDP ratio (%, right scale)
110 20
100
16
90
80 12
70 8
60
4
50
12
40 0
10
3.0 30 -4
8
2.5
6 -8
2.0
4
1.5
2
1.0
0
0.5 -2
0.0 -4
-0.5 -6
66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18
GDP inflation (%, left scale) Government current expenditures growth rate (%, left scale)
Public debt to GDP ratio (%, right scale) Government capital expenditures growth rate (%, right scale)
Note: The shaded area corresponds to recession periods (two consecutive negative quarterly economic growth).
Source: Author’s own construction
130
The estimation of the structural model (the block of equations described in the methodology by
Equations (1) to (6)), controlling for expansion and recession, yields the following table. Standard
errors of the estimated coefficients are displayed between parentheses below the estimated values
of those coefficients.
Model’s estimation controlling for expansion Model’s estimation controlling for recession
𝑟𝑡 = . 129 𝑦𝑡 +. 311𝜋𝑡 + 𝑒1𝑡 𝑟𝑡 = − . 071 𝑦𝑡 +. 383𝜋𝑡 + 𝑒1𝑡
(.256) (.287) (.192) (.279)
𝑑𝑡 = −1.05𝑦𝑡 −1.12𝜋𝑡 −.308𝑟𝑡 + 𝑒2𝑡 𝑑𝑡 = −.639𝑦𝑡 −1.20𝜋𝑡 −.269𝑟𝑡 + 𝑒2𝑡
(.174) (.355) (.087) (.181) (.354) (.088)
𝑦𝑡 =. 167𝑖𝑡 +. 209𝜋𝑡 −. 017𝑟𝑡 + 𝑒4𝑡 𝑦𝑡 =. 163𝑖𝑡 +. 245𝜋𝑡 −. 016𝑟𝑡 + 𝑒4𝑡
(.018) (.029) (.020) (.021) (.032) (.024)
𝜋𝑡 = −.024𝑟𝑡 −.005𝑖𝑡 + 𝑒5𝑡 𝜋𝑡 = −.022𝑟𝑡 −.009𝑖𝑡 + 𝑒5𝑡
(.030) (.021) (.031) (.021)
𝑖𝑡 = −.636𝑦𝑡 −2.13𝑟𝑡 + 𝑒6𝑡 𝑖𝑡 =. 378𝑦𝑡 −1.15𝑟𝑡 + 𝑒6𝑡
(.605) (2.07) (.368) (1.25)
𝑔𝑡 = 𝑒3𝑡 𝑔𝑡 = 𝑒3𝑡
(…) are standard errors (…) are standard errors
From the previous estimations, we notice significant differences in some estimated elasticities
between the two models (expansion versus recession). We also observe that some signs are
inverted from positive to negative, or vice versa, between two situations in the equation of interest
rate and private investment. This shows that some behaviour is changing over the business cycle,
which could explain the differences in fiscal multipliers between periods of expansion and
recession. For the rest of the application, we produce the impulse response functions to structural
shocks of the interest rate (shock1), public debt variable (shock2), private investment (shock3) and
public expenditure (shock6) for the variables output, investment, inflation, interest rate and debt.
We produce these IRFs for eight cases: two for the business cycle periods (expansion versus
recession), two for debt ratio movements (accumulation versus reduction) and four for the joint
business cycle and debt movements (expansion and debt accumulation/reduction and recession
and debt accumulation/reduction). These outputs are displayed by the eight figures in Appendix B
(Figures 11.B.a to 11.B.h).
To shed more light on the effects of fiscal variables on output, inflation and private investment,
we prefer to focus on the corresponding IRFs, which we reproduce in this section. For the effects
of the innovations of public debt (shock2) on output, private investment and inflation, Figure 9
clearly shows in the first row corresponding to the expansion model’s IRFs that public debt
increase has a deflationary effect on the other variables by reducing economic growth, especially
through crowding out private investment in the United States, and inducing an increase in
government expenditure. However, in times of recession (the second row of Figure 5), an increase
in public debt is likely to increase growth by even stimulating inflation and private investment
while keeping the interest rate reduced in the second quarter and pushing up government
expenditure. We also observe that the effects are generally happening with a delay of one quarter
and are at their peaks in the second or third quarter, while fading away (or stabilizing at their long-
131
term path) at the fourth or sixth quarter, except for inflation, which has a persistent long-term
response.
Figure 5. Responses to a structural shock of the government public debt ratio (shock2) in times of expansion
(first row of charts) and recession (second row of charts)
GDP response to Government expenditure in economic expansion and debt accumulation Response of LPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response ofLGE to Shock2
.0008 .004 .0006 .0020
.0004 .000 .0015

.0004 .002
.0002 .0010
.0000 .000 -.001
.0000 .0005
-.0004 -.002
-.0002 .0000
-.002
-.0008 -.004
-.0004 -.0005
-.0012 -.006 -.0006 -.003 -.0010

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Response ofLGDP to Shock2 Response of LPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response ofLGE to Shock2
.0024 .006 .00100 .001 .0025
.0020
.00075 .0020
.0016 .004
.000
.00050 .0015
.0012
.0008 .002 .0010

.00025
-.001
.0004
.0005
.00000
.0000 .000
-.002 .0000
-.0004 -.00025
-.0008 -.002 -.0005
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
For the effects of the structural innovations of total government expenditure (shock6, in Figure 6)
in periods of expansion and recession, we note that the effect is immediate and high (in the first
quarter), especially for the response of GDP, private investment and the public debt ratio. The
effects of government expenditure are generally independent of the business cycle effects, except
for the inflation variable being reduced in times of expansion and pushed up in times of recession.
For the periods of expansion and recession as well, an increase in government expenditure is likely
to immediately increase the output and then reduce the public debt ratio, while inducing an increase
in the interest rate, especially in the second quarter, to counter the inflationary effects, albeit less
important, in times of recession. However, this reduction of the public debt ratio could be a
consequence of an algebraic computation of the increase of GDP being the denominator of the
debt ratio variable. An important point is that all the responses are very short-lived (the effects
occur and fade way within the first year), except for the reaction of the prices. In concordance with
the public debt and government expenditure effects, we can conclude that expenditure multipliers
are mainly weakened in times of expansion and increased in times of recession (as found in the
previous results) by the effects of the public debt that crowd out the private agent decisions of
investing, while the effects of fiscal policy (by expenditure side) are positive and short-lived,
independent of the business cycle.
132
Figure 6. Responses to a structural shock of government expenditure (shock6) in times of expansion

(first row of charts) and recession (second row of charts)
Response ofLGDP to Shock6 Response of LPINV to Shock6 Response ofLPRC to Shock6 Response ofRIN TR to Shock6 Response ofLDR to Shock6
.008 .04 .0004 .004 .002
.003 .000
.006 .03 .0002
.002
-.002
.004 .02 .0000 .001
-.004
.002 .01 -.0002 .000
-.006
-.001
.000 .00 -.0004
-.002 -.008
-.002 -.01 -.0006 -.003 -.010

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.006 .04 .0008 .002 .002
.005 .0006
.03 .001
.000
.004
.0004
.003 .02 .000
.0002 -.002
.002 .01 -.001
.0000
.001
-.004
.00 -.002
.000 -.0002
-.001 -.01 -.0004 -.003 -.006

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
For the effects of debt movements (Figures 7 and 8), we observe almost the same findings about
the reactions of the variables as those observed for the business cycle, except for prices (inflation
and interest rates). A structural innovation of the public debt ratio is likely to reduce output by
crowding out investment and may have a deflationary effect when debt is accumulated. However,
in times of decumulating public debt, the effects of the public debt increase on output, investment
and prices are positive (Figure 7). For the effects of government expenditure, they are positive on
output and investment, while reducing public debt. The prices’ reactions are slightly different for
debt accumulation and debt reduction cases.
Figure 7. Responses to a structural shock of the public debt ratio (shock2) in periods of debt accumulation
(first row of charts) and debt contraction (second row of charts)
Response ofLGDP to Shock2 Response of LPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response of LGE to Shock2
.008 .0004 .001 .004
.001
.004 .000 .003
.0000
.000 .000 -.001 .002

-.0004
-.004 -.002 .001
-.001
-.0008
-.008 -.003 .000
-.002 -.012 -.0012 -.004 -.001

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Response ofLGDP to Shock2 Response of LPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response of LGE to Shock2
.003 .008 .0012 .001 .004
.000 .003
.002 .0008
.004
-.001 .002
.001 .0004
.000 -.002 .001
.000 .0000
-.003 .000
-.004
-.001 -.0004 -.004 -.001

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
133
Figure 8. Responses to a structural shock of government expenditure (shock6) in periods

of debt accumulation (first row of charts) and debt contraction (second row of charts)
.008 .04 .0006 .003 .002
.0004
.006 .03 .002 .000
.0002
.004 .02 .001 -.002
.0000
.002 .01 .000 -.004
-.0002
.000 .00 -.001 -.006

-.0004
-.002 -.01 -.0006 -.002 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0008 .004 .002
.000
.006 .03 .003
.0004
-.002
.004 .02 .002
-.004
.002 .01 .0000 .001
-.006
.000 .00 .000

-.0004 -.008
-.002 -.01 -.001 -.010

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
For the effects of the business cycle and debt movements, we produce the four cases (in Figures 9
to 12). A structural innovation of the debt in periods of debt accumulation and expansion decreases
simultaneously the GDP, investment and prices (inflation and interest rate), while increasing
government expenditure (first row of the panel in Figure 9). For periods of debt accumulation in
recession periods (second row of Figure 9), the effects are opposite (positive) on the first three
variables, while the reaction of the interest rate and government expenditure have almost the same
shape as in the first case. For the effects of government expenditure (Figure 10), they are short-
lived and almost the same, independent of the two considered cases.
Figure 9. Responses to a structural shock of the government debt ratio (shock2) in periods of debt
accumulation and expansion (first row of charts) and debt accumulation and recession (second row of charts)
Response of LGDP to Shock2 Response of LPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response ofLGE to Shock2
.0005 .0004 .001 .0024
.000 .0020
.0000
.000
.0000 .0016
-.0005
-.001 .0012
-.004
-.0010 -.0004 .0008
-.002 .0004
-.0015
-.008
-.0008 .0000
-.003
-.0020
-.0004
-.0025 -.012 -.0012 -.004 -.0008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.003 .010 .00100 .0008 .0025
.0004
.008 .00075 .0020
.002 .0000
.006 .00050 .0015
-.0004
.001 .004 -.0008 .0010

.00025
-.0012
.002 .0005
.00000
.000 -.0016
.000 .0000
-.00025 -.0020
-.001 -.002 -.0024 -.0005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
134
Figure 10. Responses to a structural shock of government expenditure (shock6) in periods of debt
accumulation and expansion (first row of charts) and debt accumulation and recession (second row of charts)
.008 .04 .0006 .003 .002
.0004
.006 .03 .002 .000
.0002
.004 .02 .001 -.002
.0000
.002 .01 .000 -.004
-.0002
.000 .00 -.001 -.006

-.0004
-.002 -.01 -.0006 -.002 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0008 .003 .004
.002
.006 .03 .002
.0004
.000
.004 .02 .001
-.002
.002 .01 .0000 .000
-.004
.000 .00 -.001

-.0004 -.006
-.002 -.01 -.002 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
For the government debt reduction case, jointly with the business cycle, unlike the case where debt
is accumulated, the responses to the public debt structural shock are slightly different over the
business cycle for private investment and interest rate, while they seem to behave the same way
for the other variables, between the two situations. In particular, the output, investment and
inflation are positively affected in the first year with persistent effects for inflation.
Figure 11. Responses to a structural shock of the government debt ratio (shock2) in periods of debt reduction
and expansion (first row of charts) and debt reduction and recession (second row of charts)
.004 .008 .0012 .002 .003
.003 .006 .001

.0008
.002
.002 .004 .000
.0004
.001 .002 -.001 .001
.0000
.000 .000 -.002
.000
-.001 -.002 -.0004 -.003

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0020 .006 .00100 .001 .0025
.0015 .004 .00075 .0020

.000
.0010 .002 .00050 .0015
.0005 .000 -.001 .0010

.00025
.0000 -.002 .0005

.00000
-.002
-.0005 -.004 .0000
-.00025
-.0010 -.006 -.003 -.0005
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
135
Figure 12. Responses to a structural shock of government expenditure (shock6) in periods of debt reduction
and expansion (first row of charts) and debt reduction and recession (second row of charts)
.008 .04 .0008 .004 .002
.0006 .003 .000

.006 .03
.0004 .002 -.002

.004 .02
.0002 .001 -.004
.002 .01
.0000 .000 -.006
.000 .00
-.0002 -.001 -.008
-.002 -.01 -.0004 -.002 -.010

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0006 .003 .002
.0004 .002
.006 .03 .000
.0002 .001
.004 .02 -.002
.0000 .000
.002 .01 -.004
-.0002 -.001
.000 .00 -.006

-.0004 -.002
-.002 -.01 -.0006 -.003 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
This model, containing fiscal and monetary variables, sought to explain why the fiscal multipliers
are weaker, or even negative, in times of expansion than recession. In times of high public debt,
and particularly expansion, an increase in public debt ratio crowds out private investment, hence
reducing output. By contrast, the government expenditure effects on output are all positive in the
short term, independent of the public debt evolution (accumulation or decumulation) and business
cycle. These results align with what we observed in a preliminary analysis (Section 4.2) of the
works of Bouassard et al. (2012) and Berti et al. (2013) for 27 European countries, in which we
highlighted the apparent decreasing relationship between expenditure multipliers’ size and public
debt ratio.
6. Conclusion
In this chapter, we used the methodology of a structural vector autoregressive model (SVAR),
augmented by exogenous dummies variables controlling for the business cycle
(expansion/recession) and public debt movements (accumulation/reduction). We applied this
approach to assess the expenditure multipliers for a sample of 18 OECD countries (Austria,
Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, The
Netherlands, Norway, Portugal, Spain, Sweden, the United Kingdom and the United States) with
different exchange rate regimes, monetary policies and degrees of openness.
The results show that, controlling for the business cycle effects, the expenditure multipliers are
much higher in times of recession than times of expansion, and could attain, in recessions, more
than 1 for many countries in the sample, in the short term, while going beyond the value of 2 in
the long term. Moreover, it is noted that, generally, the idea of spending multipliers being weak,
136
and even negative in recent decades, compared to the 1960s, 1970s and 1980s, is not well
supported by our findings. The previous results are in line with what was observed in the recent
literature about fiscal multipliers, in advanced economies, being large in times of recession but
weak, or even negative, in times of expansion.
Considering these results, the fiscal policy in advanced countries should be designed according to
business cycle fluctuations. In particular, fiscal policies should be designed to counter the business
cyclicality. In times of recession, it is the role of the government sector to stimulate the economy,
while public intervention in times of expansion seems to alter economic growth (as the multipliers
are weak or negative for many countries) and less beneficial to the economy than in times of
recession. These results also run contrary to any fiscal consolidation based on cutting expenditure
in times of recession, which could harm the economy.
Controlling exogenously for the public debt movements, independent of the business cycle, it is
revealed that spending multipliers are larger in periods of debt accumulation than in debt reduction
periods. Furthermore, controlling jointly for debt movements exogenously and the business cycle
reveals the previous tendency, that is, multipliers are higher under debt accumulation in cases of
expansion and recession. However, introducing endogenously the public debt to GDP variable in
an SVAR leads to higher multipliers in recessions than expansions.
Furthermore, a robustness check of the previous results was conducted on a long history of
quarterly data for the United States, as the country with long quarterly time series of the six fiscal
and monetary variables considered in this model, namely, public debt, GDP, private investment,
public expenditure, interest rate and inflation. The period of estimation covers the range 1966q1–
2019q2. The main results of this model show that government expenditure has positive but short-
lived effects on economic growth. Furthermore, public debt crowds out private investment, leading
to lowering growth rates in times of expansion, while in times of recession, the public debt effects
on growth are positive. This crowding-out effect may play pass-through to the expenditure
multipliers and could explain, ceteris paribus, the weak size of spending multipliers, while in times
of recession the crowing-in effect leads to higher multipliers.
In all our models the recession period generally has a persistent effect on variables for which
convergence to the long-term path following the shock is achieved faster in times of expansion
than recession. The policy implication of this, for highly indebted countries, is that fiscal stimulus
effects could take time to materialize in times of depressed economies, while the effects are short-
lived in times of expansion, which should be considered by policy-makers in their spending
decisions.
137
Appendix A. Tables
138
Table 1.A: Sample of studies on the effectiveness of stimulus measures

Author(s) Countries/Periods Effects Main Findings
Verick and Islam 60 countries (2008- PositiveFiscal stimuli have proven more effective in creating jobs when they focus on capital expenditures for employment-intensive infrastructure and labour
(2010) 2009) market measures. The US in 2009 enacted ARRA to restore employment mostly through tax cuts with a goal to create 3.5 million jobs by Q4 2010. By Q3
2009, 300,000-1.5 million jobs were made; 8.4 million jobs were lost since 2007.
Van Doorn, 20 countries Mixed A nation’s original fiscal stance determines how it can respond during a crisis. Hungary, for example, now has greater foreign exchange risks because its
Suri, and Gooptu (1995-2010) external debt surpassed its domestic debt prior to the crisis. Brazil, India, and Egypt have less risky exchange rates by accumulating more domestic debt
(2010) than external debt before the economic downturn hit.
UNESCAP (2010) 37 countries Positive Between 7 and 11 million jobs were saved due to the G20’s 2009 fiscal policies, representing 29-43 per cent of the G20’s total unemployment in the first
(1990s – 2009) half of 2009. In comparison, the Asian region’s policies were successful overall in offsetting the loss of exports and kept regional 2009 GDP dropping at
4.2 per cent instead of 7.8 per cent, a possible outcome had the fiscal policies not been enacted.
IMF (2010a) 186 countries Mixed Fiscal stimuli have provided an essential impetus for recovery in developed and developing countries alike. Following a decline of 0.5 per cent in 2009, the
(1970-2010) world output is projected to increase by 4.25 per cent in 2010. However, growth and recovery in the job market have varied across countries of different
socio-economic levels, even with similar fiscal responses.
Arpaia and Curci 27 EU countries Positive During the crisis, many EU countries implemented reforms to combat the discouraged worker effect and keep workers in the labour market, such as
(2010) (1980s-2009) decreasing the average hours worked per week. This made the early reduction in economic activity less drastic than originally projected.
EEAG (2010) 27 countries Mixed The EU’s stimulus measures succeeded in curbing deflation and preventing a second depression. As example, Germany see its investment and domestic
(1990s-2009) demand recover by the end of Q2/2009; however, consumption and GDP drop after an initial increase brought about by fiscal stimulus measures.
ILO (2010a) EU (2006-2009) Mixed Mixed As a result of the European Globalization Adjustment Fund (EGAF) component of the European Economic Recovery Plan (EERP), 25,000 people
received support in the form of skills training, job and career search help, promotion of entrepreneurship, and training for self-employment. This allowed
many EU citizens to find jobs after losing work due to outsourcing.
Wyplosz (2010) Eurozone – 16 Negative Fiscal policies played a small role in Eurozone’s recovery, due to the pre-existing Stability and Growth Pact and 3 per cent public deficit ceiling limiting
countries their use. Budget balances as a percentage of GDP were less in France, Germany, and Italy compared to the UK and US, implying that less government
(1996-2009) funds went towards fiscal stimuli in the former three countries than the latter two.
Kandil and Morsy 34 emerging Positive The presence of international reserves allowed emerging economies to be in a good position to weather the economic crisis and have their stimulus measures
(2010) countries (1970- succeed, while at the same time avoiding a crowding-out effect. For example, China and Brazil have managed to weather the crisis well due to a rapid
2009) increase in export prices, robust demand, and constant capital inflows.
World Bank (2010a) 14 emerging Positive Due to China’s economic activity, its monetary and fiscal stimuli, and its inflow of foreign capital, the East Asia and the Pacific region overall has recovered
countries (Mid- from the crisis and had its employment, exports, and output return to pre-crisis levels. The region is doing so well that many national governments are
1980s-2009) beginning to end support policies.
ADB (2010) 32 countries Positive Asia, the first region to recover from the global financial crisis, (1990s-2010) experienced a strong and fast V-shaped recovery due to large monetary and
fiscal stimulus packages. These heavily increased the demand for goods and services and promoted growth, which is projected to reach 7.5 per cent in 2010,
which is up from 5.2 per cent in 2009.
Bhaskaran and Ghosh 3 countries Mixed Indonesia, the Philippines, and Thailand were resilient during the global recession thanks to timely stimulus measures. For example, starting in 2008
(2010) (1981-2009) Indonesia injected RP 1 trillion into its economy, which successfully increased liquidity and prevented a dramatic depreciation of its currency.
IDB (2010a) 4 Latin American Positive Decreased economic activity, employment, tax rates, and export prices brought about due to the downturn ended Latin America’s period of expansion that
countries began in 2002-03 and thus decreased its fiscal surplus. MERCUSOR was still able to respond by implementing measures that showed positive impacts,
such as increased government spending, tax cuts, and infrastructure investment.
IMF (2010b) 28 North and South Mixed The North American and Latin American regions experienced different outcomes due to their fiscal stimuli. The recoveries in the US and Canada are
America States propelled by fiscal stimuli, giving the former a 5.6 per cent GDP increase by Q4 2009 and the latter resilient fiscal credibility, increased demand, and 3 per
(1970-2010) cent growth in 2010. In most of Latin America, government fiscal stimuli had small multipliers that did not contribute much to growth.
Kucera, Roncolato, India and South Negative Fiscal stimuli did not contribute much to GDP or job growth in India or South Africa. India’s employment growth rate was 1.7% in Q4/2009 but only 0.2%
and Uexkull (2010) Africa (19602010) in Q1/2010. South Africa experienced 3 negative quarters of growth in 2009 and observed positive growth in Q4/2009 (0.7%), but growth decreased again
in Q1/2010 (-1.3%). Employment declined by 770,000, 833,000, and 870,000 in Q3/2009, Q4/2009, and Q1/2010 respectively.
139
Table 1.A continued: Sample of studies on the effectiveness of stimulus measures

Author(s) Countries/Periods Effects
Main Findings
Kasekende, 56 African Mixed
Most of Africa had favourable pre-crisis macroeconomic standings. Thus, much of the region was able to deliver well-timed, domestic demand and supply-
Ndikumana and countries side focused countercyclical stimuli that promoted a better business environment. However, due to their lack of resources, they should look internationally
Brixova (2010) (Mid-1980s–2009) for support to coordinate and establish a long-term mutually-beneficial outcome.
Ndikumana, el al. 41 countries (1986-
Positive Even with short supply of national resources, African nations enacted a variety of well-timed and targeted policies. Some nations were successful in
(2010) 2009) stimulating domestic demand while offsetting declining exports and supporting local businesses.
World Bank (2010b) 18 countries Positive Most Middle Eastern nations improved their near and long-term outlooks. Others, such as Egypt and Tunisia, spent stimulus money on infrastructure and
(1980-2010) support for SMEs to create new jobs. Unlike the public sector, the private sector has received less support and it is uncertain whether this sector will continue
to recover when Saudi Arabia’s 400 billion dollar stimulus package expires in 5 years.
ILO (2010b) Germany Positive Germany’s labour market-targeted measures left favourable (2006-2009) impacts on employment. Germany’s Kurzarbeit reduced working hour program
had 1.43 million participants by mid-2009. Working time reduced on average by 30.5 per cent, which counted for 432,000 fulltime jobs. If these job losses
had occurred, Germany’s unemployment rate would have risen by 1 percentage point.
IMF (2010c) Germany Mixed The economic downturn was contained in Germany by the government’s automatic stabilizers and fiscal stimulus. Additionally, Germany’s extensive labour
(mid-1970s2010) market programmes, such as the Kurzarbeit short-term subsidy programme, which compensated employers for retaining workers during the crisis,
successfully encouraged many employers to refrain from layoffs.
ILO (2010c) Russia (1992- Negative Despite the rapid pace at which stimulus packages were adopted, hiring has yet to catch up. In September 2009, Russian public employment services have
2010) placed 61,300 out of 328,300 employees in public sector jobs, a 50 per cent increase from 2009 target. However, the ratio of unemployed to 100 labour
market vacancies stands at 212.
Park and Lommen China Mixed China’s timely and effective fiscal stimuli allowed it to experience a stronger and faster V-shaped recovery than expected. However, pressing national issues
(2010) (2008-2010) remain for China to solve in the medium-run, such as aging population, lack of domestic growth, absence of effective social protection, inflexible exchange
rates, and unbalanced growth sources, all of which need stimuli.
Yongding (2009) China (1978-2008) Mixed In 2008, the Chinese government issued a RMB 4 trillion stimulus package and RMB 200 billion worth of local government bonds. While succeeding in
reviving and stabilizing the economy and keeping growth positive, these actions have not insured China against long-term consequences, such as investment
overdrive, decreased efficiency, and slow infrastructure investment.
Zhu and Orton (2010) China (2003-2010) Positive Chinese fiscal policy responses were effective in curbing the negative impacts of the global financial crisis. The GDP growth maintained high during the
crisis, reaching levels of 9 per cent in 2008, 8.7 per cent in 2009, and 8 per cent in 2010. On the labour side, 97 per cent of 150.7 million migrant workers
from rural China found jobs in urban areas.
Shimizu and Orton Japan (1985-2010) Positive Japan implemented effective labour market stimuli. Japan’s crisis response employment measures created 90,000 new local government jobs and granted
(2010) 2.34 million workers wage subsidies by July 2009. Furthermore, 2.55 million temporary Japanese workers were covered under the unemployment insurance
by April 2010.
ILO (2010d) Republic of Mixed Korea’s stimulus helped retain jobs and allowed new labour market entrants to find employment. It raised SME subsidies from 2/3 to 3/4 of wages and gave
Korea large enterprises a subsidy increase from 1/2 to 2/3 of wages. 30 per cent of workplaces with 100 or more employees participated, which saved at least
(2006-2009) 90,000 jobs. Internship programmes at public institutions and SMEs gave employment to 90,000 young Koreans, 80 per cent of whom went on to work
fulltime.
OECD (2010) Korea (1996-2010) Positive Korea’s fiscal stimulus package from 2009 greatly increased national consumption and decreased unemployment levels. By the end of 2009, this measure
created at least 300,000 jobs in the public sector.
ACTU (2009) Australia Mixed Australia’s fiscal labour market measures had mixed results on employment. Its fiscal stimulus package was successful in decreasing national unemployment
(2008-2009) down to 5.7 per cent by late 2009. However, over 140,000 job-seeking Australians remain out of work. Therefore, the economic stimulus and infrastructure
investment should not end.
Commonwealth of Australia (1980s- Positive By the end of Q2 2009, the Australian stimulus package had succeeded in stabilizing the national market and minimizing (2009) the rise in unemployment.
Australia 2009) The unemployment rate was 5.8%, 1.9 points higher than in Feb 2008. However, without the stimulus, the Treasury estimated that an additional 210,000
jobs would have be lost.
ILO (2010e) Australia (2007- Positive Australia’s fiscal stimulus measures increased national growth levels and strengthened the labour market. It increased GDP by 1% between 2008 and 2009
2010) and unemployment declined by 0.5% from a high level in July 2009 (5.8%) to a lower mark in February 2010 (5.3%), as a result of the stimulus.
ILO (2010f) Indonesia (2006- Positive Of the IDR 12.2 trillion infrastructure package, IDR 6.6 trillion went to the Department of Public Works for the creation of water systems, sanitation
2010) systems, and roads. IDR 1.5 trillion devoted to water systems alone was responsible for creating 250,000 jobs in 2009.
140
Table 1.A continued: Sample of studies on the effectiveness of stimulus measures

Author(s) Countries/Periods Effects Main Findings
ILO (2010g) Argentina (1996- Positive Argentina’s measure to reduce tax credits for capital goods, infrastructure firms, and SMEs in an effort to temporarily reduce their tax and social security
2009) burden succeeded in retaining 330,547 employees of those 169,000 employers that registered for the program.
Zandi (2010) United States Positive The ARRA has dampened the negative impact of the recession on GDP and employment. The most economic activity per federal dollar spent has come
(2008-2009) from the extension of unemployment insurance benefits and social programs like food stamps and work-share as well as infrastructure spending.
Taylor (2009) United States Negative The US government’s early response to the 2007 crisis worsened the situation by focusing on liquidity and bank credit market problems rather than risk,
(2006-2008) providing no help to other than banks and creditors. Measures such as term auction facilities, cash infusions, and interest rate cuts failed to substantially
increase consumption, but it increased dollar inflation and raised oil prices.
Burtless (2010) United States Mixed The ARRA was successful in allowing most state governments to maintain pre-crisis levels of employment in government and educational institutions, in
(2007-2010) spite of the massive decrease in tax revenue. This did come at the cost of decreased wages and working hours though. The bill also gave a 9-month subsidy
for the unemployed that covered 65 per cent of former employers’ healthcare.
US Government United States Mixed Due to US stimulus packages, the decline in national GDP slowed from -6.4% in Q1/2009 to -0.7% in Q2/2009, and then turned around to 2.2 and 5.7% in
(2010) (1940s – 2009) Q3 and Q4 of 2009. The rate of job losses also has been reduced substantially, from a high of 691,000 in Q1/2009 to a low of only 69,000 in Q4/2009.
Council of Economic United States Positive The ARRA was able to slow down the rate of job and output loss while also returning GDP to positive levels. In Q2 2009, 2-3% was added to GDP, and in
Advisors (2010) (2007-2009) Q4/2009, 1.5-3% was added to GDP. Employment significantly increased under the ARRA, specifically between 1.5-2 million jobs by the end of 2009.
Joint Eco. Committee United States Mixed By Q1 2010, the ARRA increased GDP by 1.7 to 4.2 per cent and decreased the national unemployment rate by 0.7 to 1.5 percentage points. This resulted
-US Congress (2010) (1950s-2010) in the creation of 1.2-2.8 million jobs since the ARRA was enacted a year prior.
Mishel and Shierholz United States Positive Unemployment insurance should not be reduced, as it both creates jobs and aids those without one. The unemployment insurance measures in the
(2010) (2007-2010) ARRA retained 1.7 million current jobs and created 1.2 million new jobs by July 2010.
Weller (2010) United States Mixed The ARRA allowed the private sector to start recovering prior losses in investment, production, and hiring. The US in Q2 2010 experienced a 2.4% growth
(1950s-2010) rate and gained 630,000 new private sector jobs in the first 7 months of 2010, but unemployment still high and 7.7 million jobs lost since 2007.
Flaherty (2010) Canada (2000- Positive As a result of Canada’s stimulus package, in 2009 both aggregate employment and GDP levels increased. Since its inception in July 2009, the Canada
2010) Action Plan has created 135,000 jobs out of its target of 225,000. The GDP increased by 0.9 per cent in Q3 of 2009 and by 5 per cent in Q4 of 2009. Canada
also experienced the smallest decline in GDP out of all G7 nations, since the crisis began.
ILO (2010g) Canada Positive The earnings protection and work sharing components of Canada’s Action Plan were successful in saving many Canadian jobs during the crisis. In 2009,
(20072010) 13,500 workers received C$3,323 in financial assistance for unpaid wages due to company closures and termination. Also, in 2009, 6,000 work sharing
agreements were signed by Canadian employees, securing the employment of 167,410.
Asian Development Asia (4 G20 Positive The countries in developing Asia rolled out the stimulus measures quickly and decisively in general, proactively using fiscal policy for countercyclical
Bank, ADB (2010) members and 10 purposes during the global developing crisis. The stimulus packages are tilted toward higher countries) spending, in particular infrastructure investments,
other developing consistent with the region’s long-standing pro-growth orientation. Partly helped by the healthy public finances’ situation in the initial stage, evidences
countries) suggest that the stimulus efforts to boost sagging demand have been effective and contributed to the region’s remarkable recovery.
The Brookings G-20 CountriesPositive The article finds that fiscal stimulus played a crucial role in stabilizing the world economy, especially in the G-20 economies. It suggests that while the
Institution (2009) (2008-2009) effectiveness of fiscal stimulus is questioned, the world economy would have been even worse had there not been stimulus packages.
NBER (2010) 22 high income +24Mixed Using the panel SVAR methods, the research finds that the effect of government consumption is very small on impact, with estimates clustered close to zero.
developing countries It supports the notion that fiscal policy may be rather slow in impacting economic activity, raising questions as to the usefulness of discretionary fiscal policy
(1960:Q1 -2007:Q4) for short-run stabilization purposes. The medium- to long-run effects of increases in government consumption vary considerably depending on the level of
development, exchange rate flexibility, openness to trade, and debt of the central government.
OECD Economic 34 OECD Mixed The report verifies the use of fiscal stimulus in virtually all OECD countries. As for the effectiveness of fiscal stimulus, the report suggests after its own
Outlook, InterimCountries review of relevant literatures that the fiscal multipliers are around unity for government spending under normal circumstances. The report suggests that the
Report (2009) (2008-2010) multipliers may be reduced in the current conjuncture as the propensity of households/businesses to save is likely to increase reducing tax cuts multipliers.
OECD Economic 34 OECD CountriesMixed Noting that most OECD countries introduced fiscal packages during the recent crisis, the report recognizes the potentially important role of discretionary
Outlook, Volume 2010(2008-2010) fiscal policy during a large and protracted shock. Meanwhile, the report warns that implementing fiscal stimulus may be slow and may result in a pro-cyclical
Issue 1 rather than counter-cyclical. Looking at past experience, estimates of discretionary fiscal policy show pronounced counter-cyclicality only in Australia,
Canada, Denmark, and the US, while policy has been generally pro-cyclical in Austria, Belgium, Hungary, the Netherlands, Poland, Portugal, and the UK.
Also, fiscal policy was poorly prepared to deal with the crisis as countries were financially constrained prior to the outbreak of the crisis.
Source: International Labour Organization (2011).
141
Table 2.A: Coefficient of correction to get fiscal multipliers from Cholesky impacts innovations by periods and countries samples
Government consumption growth rate Average of GDP to government ̅̅̅̅̅

̅̅̅̅̅ ) 𝑌/𝐺 /𝜎𝑔
standard error (𝜎𝑔 ) consumption (𝑌/𝐺
1966Q1/ 1966Q1/ 1992Q1/ 1966Q1/ 1966Q1/ 1992Q1/ 1966Q1/ 1966Q1/ 1992Q1/

Countries
2019Q2 1991Q4 2019Q4 2019Q2 1991Q4 2019Q4 2019Q2 1991Q4 2019Q4
Austria 0.721 0.469 0.882 4.796 4.575 5.002 6.65 9.75 5.67
Belgium 0.679 0.700 0.599 3.877 3.643 4.096 5.71 5.20 6.84
Canada 1.056 1.288 0.704 4.091 3.616 4.536 3.88 2.81 6.45
Denmark 0.821 0.850 0.751 4.156 4.215 4.101 5.06 4.96 5.46
Finland 1.800 1.809 1.724 3.818 3.605 4.018 2.12 1.99 2.33
France 0.468 0.498 0.291 4.275 4.298 4.253 9.14 8.62 14.63
Germany 1.178 1.459 0.823 5.179 5.103 5.251 4.40 3.50 6.38
Greece 1.827 1.026 2.272 5.558 5.967 5.174 3.04 5.82 2.28
Ireland 1.898 1.116 2.415 5.002 3.717 6.205 2.64 3.33 2.57
Italy 0.706 0.456 0.713 5.050 4.829 5.257 7.15 10.58 7.38
Japan 0.928 1.119 0.576 5.758 6.067 5.469 6.20 5.42 9.49
Netherlands 1.415 1.821 0.884 4.165 4.163 4.167 2.94 2.29 4.71
Norway 1.423 1.444 1.350 4.590 4.756 4.435 3.22 3.29 3.28
Portugal 1.096 1.010 0.671 6.890 8.262 5.606 6.28 8.18 8.35
Spain 1.035 1.060 0.912 6.842 8.128 5.638 6.61 7.67 6.18
Sweden 0.923 1.013 0.749 3.236 2.994 3.463 3.51 2.96 4.62
United Kingdom 1.101 1.162 1.046 4.727 4.251 5.173 4.29 3.66 4.95
United States 0.830 0.952 0.680 5.298 4.400 6.139 6.38 4.62 9.03
142
Table 3.A. Sensitivity of government expenditure Cholesky impacts and the corresponding multipliers to different periods of time
1966Q1-1991Q4 1992Q1-2019Q2
Fourth Quarter 20th Quarter Fourth Quarter 20th Quarter
first Quarter first Quarter
(one year) (5 years) (one year) (5 years)
Countries Impact Multiplier Impact Multiplier Impact Multiplier Impact Multiplier Impact Multiplier Impact Multiplier
Austria 0.27 2.59 0.34 3.31 0.34 3.35 0.03 0.17 0.03 0.19 0.03 0.20
Belgium 0.02 0.13 0.32 1.68 0.49 2.53 0.07 0.48 0.17 1.17 0.18 1.21
Canada 0.27 0.76 0.13 0.36 0.13 0.35 0.04 0.24 -0.09 -0.57 -0.11 -0.68
Denmark 0.24 1.20 0.47 2.35 0.50 2.46 0.10 0.53 -0.05 -0.28 -0.05 -0.28
Finland -0.08 -0.16 0.01 0.02 0.01 0.01 0.25 0.58 0.44 1.03 0.44 1.03
France -0.07 -0.58 0.18 1.53 0.18 1.53 -0.17 -2.42 -0.46 -6.73 -0.61 -8.85
Germany 0.26 0.91 0.21 0.73 0.21 0.73 0.02 0.14 -0.13 -0.83 -0.13 -0.84
Greece -0.14 -0.82 0.16 0.96 0.70 4.05 0.66 1.50 0.97 2.20 0.97 2.21
Ireland 0.33 1.09 0.85 2.83 0.11 0.36 0.79 2.03 0.77 1.98 0.77 1.97
Italy 0.12 1.27 0.30 3.20 0.33 3.49 0.21 1.55 0.46 3.37 0.51 3.80
Japan 0.11 0.61 0.00 0.03 0.00 0.02 0.01 0.10 0.06 0.60 0.06 0.60
Netherlands 0.24 0.55 0.17 0.38 0.17 0.39 -0.01 -0.02 0.14 0.65 0.15 0.71
Norway 0.27 0.89 0.30 1.00 0.30 0.99 0.23 0.75 -0.04 -0.14 0.01 0.02
Portugal 0.72 5.92 0.79 6.47 0.70 5.72 0.20 1.68 0.61 5.11 0.87 7.30
Spain 0.22 1.71 0.22 1.67 0.22 1.66 0.27 1.65 0.77 4.74 1.09 6.74
Sweden 0.34 1.00 0.29 0.85 0.29 0.85 -0.01 -0.03 -0.06 -0.30 -0.07 -0.31
United Kingdom 0.00 -0.01 -0.04 -0.13 -0.04 -0.13 0.06 0.31 0.03 0.16 0.03 0.13
United States 0.03 0.16 -0.02 -0.10 -0.02 -0.12 -0.08 -0.72 -0.30 -2.72 -0.33 -3.02
Average 0.18 0.96 0.26 1.51 0.26 1.57 0.15 0.47 0.18 0.54 0.21 0.66
Note: Impact multipliers (IM) are adjusted by the corresponding adjustment coefficient from table 2.A to obtain fiscal multipliers (FM) according to the formulae
̅̅̅̅̅ /𝜎𝑔 .
𝐹𝑀 = 𝐼𝑀 ∗ 𝑌/𝐺
143
Table 4.A: Business cycle effects on government expenditure short run and long run multipliers
Long run
Short run impact Long run accumulated Short run multiplier (1st Long run accumulated Short run
1992Q1-2019Q2 accumulated
(1st quarter) impact (5 years) quarter) multiplier (5 years) multiplier
multiplier
(Recession
(Recession
Minus
Country Expansion Recession Expansion Recession Expansion Recession Expansion Recession Minus
Expansion)
Expansion)
Austria -0.029 0.375 -0.085 1.292 -0.137 1.793 -0.404 6.175 1.93 6.58
Belgium -0.014 0.276 0.025 1.196 -0.071 1.429 0.131 6.201 1.50 6.07
Canada -0.027 0.383 -0.239 1.413 -0.137 1.907 -1.190 7.043 2.04 8.23
Denmark 0.009 0.419 -0.158 0.814 0.033 1.506 -0.570 2.929 1.47 3.50
Finland 0.071 0.366 0.108 0.752 0.141 0.728 0.215 1.498 0.59 1.28
France -0.080 0.153 -0.461 1.428 -0.318 0.611 -1.835 5.689 0.93 7.52
Germany -0.104 0.359 -0.307 0.805 -0.500 1.723 -1.474 3.861 2.22 5.33
Greece 0.423 0.514 0.457 0.603 0.917 1.113 0.990 1.308 0.20 0.32
Ireland 0.550 1.533 0.486 2.165 1.022 2.846 0.903 4.019 1.82 3.12
Italy 0.112 0.195 0.213 0.515 0.641 1.120 1.226 2.956 0.48 1.73
Japan -0.085 0.322 -0.240 0.915 -0.461 1.745 -1.302 4.953 2.21 6.25
Netherlands -0.065 0.279 -0.069 1.538 -0.246 1.051 -0.261 5.796 1.30 6.06
Norway 0.145 0.819 -0.004 0.906 0.427 2.405 -0.011 2.659 1.98 2.67
Portugal 0.109 0.234 0.295 1.623 0.488 1.051 1.323 7.280 0.56 5.96
Spain 0.085 0.407 0.087 1.481 0.396 1.893 0.406 6.895 1.50 6.49
Sweden -0.099 0.377 -0.200 1.012 -0.303 1.158 -0.614 3.109 1.46 3.72
UK -0.014 0.346 -0.117 1.414 -0.065 1.588 -0.535 6.488 1.65 7.02
USA -0.024 0.310 -0.181 1.060 -0.174 2.224 -1.298 7.608 2.40 8.91
Average 0.05 0.43 -0.02 1.16 0.09 1.55 -0.24 4.80 1.46 5.04
144
Table 5.A. Short run and long run government expenditure multipliers under debt accumulation and debt reduction
Cholesky impacts Multipliers impacts

Under debt Under debt Under debt Under debt Convergence to long run multiplier
accumulation reduction accumulation reduction
Short Long Short Long Under debt Under debt
Country Short run Long run Short run Long run
run run run run accumulation reduction
Austria 0.19 0.38 0.14 0.24 0.90 1.80 0.65 1.15 Fast Fast
Belgium 0.34 1.23 0.21 0.48 1.76 6.36 1.11 2.48 Medium Medium
Canada 0.31 0.85 0.10 0.03 1.54 4.21 0.51 0.13 Medium Medium
Denmark 0.31 0.34 0.14 0.08 1.12 1.22 0.50 0.28 Fast Fast
Finland 0.14 0.22 0.15 0.20 0.28 0.44 0.30 0.39 Fast Fast
France 0.09 0.63 0.03 0.40 0.35 2.49 0.13 1.60 Medium Slow
Germany 0.18 0.38 0.07 0.10 0.87 1.81 0.32 0.47 Fast Fast
Greece 0.77 1.20 0.67 1.01 1.67 2.61 1.44 2.18 Fast Fast
Ireland 1.48 1.70 0.38 0.14 2.75 3.16 0.70 0.27 Fast Fast
Italy 0.35 0.90 0.31 0.74 1.99 5.19 1.79 4.22 Medium Medium
Japan 0.09 0.16 0.07 0.21 0.48 0.89 0.37 1.12 Fast Fast
Netherlands 0.12 0.88 -0.06 0.09 0.45 3.32 -0.23 0.35 Medium Fast
Norway 0.52 0.30 0.62 0.50 1.54 0.89 1.83 1.48 Fast Fast
Portugal 0.20 0.70 0.04 0.31 0.88 3.16 0.20 1.38 Medium Medium
Spain 0.33 2.95 0.26 1.25 1.55 13.72 1.20 5.81 Slow Slow
Sweden 0.32 0.75 0.04 0.04 1.00 2.30 0.12 0.13 Fast Fast
UK 0.36 1.43 0.21 0.47 1.65 6.58 0.95 2.16 Slow medium
USA 0.15 0.28 0.03 -0.11 1.09 2.02 0.22 -0.76 Medium Fast
Average 0.35 0.85 0.19 0.34 1.22 3.45 0.67 1.38 -- --
Notes: Impact multipliers (IM) are adjusted by the corresponding adjustment coefficient from table 2.A to obtain fiscal multipliers (FM) according to the formulae
𝐹𝑀 = 𝐼𝑀 ∗ 𝑌/𝐺 ̅̅̅̅̅ /𝜎𝑔 . Convergence to the long run is fast if it is approximately attained in less than 5 quarters, medium between 6 and 9 quarters and slow in case
it is reached in more than 10 quarters (this is clearly visible from the GDP responses charts).
145
Table 6.A.: The mutually effects of the business cycle and public debt (accumulation/reduction)
Cholesky impacts Multipliers impacts

Expansion Recession Expansion Recession
Under debt Under debt Under debt Under debt Under debt Under debt Under debt Under debt
accumulation reduction accumulation reduction accumulation reduction accumulation reduction
Long Short Long Short Long Short Long Short Long Short Long Short Long Short Long
Country Short run
run run run run run run run run run run run run run run run
Austria -0.06 -0.12 -0.08 -0.16 0.13 0.23 0.15 0.27 -0.30 -0.59 -0.38 -0.76 0.64 1.09 0.72 1.30
Belgium 0.04 -0.04 0.01 -0.11 0.31 1.08 0.23 0.58 0.21 -0.22 0.06 -0.58 1.60 5.58 1.21 2.98
Canada -0.01 -0.18 -0.05 -0.28 0.28 0.83 0.19 0.36 -0.05 -0.91 -0.26 -1.38 1.41 4.12 0.94 1.79
Denmark 0.01 -0.12 0.00 -0.15 0.31 0.55 0.15 0.22 0.04 -0.43 0.02 -0.52 1.11 1.98 0.52 0.78
Finland -0.01 -0.07 -0.04 -0.08 0.11 0.18 0.15 0.17 -0.03 -0.14 -0.08 -0.17 0.22 0.35 0.30 0.33
France -0.09 -0.31 -0.11 -0.46 0.13 0.73 0.12 0.73 -0.35 -1.22 -0.44 -1.84 0.53 2.90 0.47 2.89
Germany -0.19 -0.38 -0.23 -0.42 0.16 0.46 0.10 0.18 -0.89 -1.84 -1.09 -2.00 0.75 2.21 0.47 0.88
Greece 0.28 0.35 0.59 0.81 0.57 0.74 0.31 0.39 0.62 0.75 1.27 1.75 1.23 1.60 0.68 0.85
Ireland 0.35 0.16 0.25 0.07 1.48 1.91 0.34 0.12 0.64 0.31 0.46 0.12 2.75 3.55 0.64 0.23
Italy 0.21 0.36 0.19 0.40 0.25 0.66 0.25 0.52 1.18 2.09 1.10 2.30 1.45 3.81 1.41 2.97
Japan -0.04 -0.11 -0.18 -0.44 0.12 0.17 0.22 0.54 -0.22 -0.60 -0.97 -2.37 0.63 0.93 1.22 2.93
Netherlands -0.13 -0.12 -0.14 -0.12 0.11 0.78 -0.04 0.13 -0.51 -0.45 -0.54 -0.44 0.40 2.94 -0.15 0.51
Norway 0.15 0.05 0.20 0.09 0.56 0.51 0.60 0.72 0.45 0.14 0.58 0.27 1.63 1.49 1.77 2.10
Portugal -0.02 0.06 0.06 0.12 0.17 0.46 -0.01 0.20 -0.07 0.27 0.26 0.54 0.78 2.06 -0.03 0.89
Spain 0.05 0.03 0.09 0.17 0.30 2.14 0.23 0.95 0.25 0.12 0.44 0.77 1.38 9.95 1.08 4.44
Sweden -0.11 -0.16 -0.14 -0.22 0.34 0.85 0.10 0.23 -0.35 -0.50 -0.44 -0.68 1.03 2.62 0.31 0.70
UK 0.00 -0.07 -0.01 -0.11 0.28 0.86 0.20 0.50 -0.01 -0.33 -0.04 -0.52 1.31 3.95 0.93 2.27
USA 0.02 -0.09 0.00 -0.13 0.25 0.61 0.26 0.60 0.15 -0.66 -0.01 -0.96 1.82 4.35 1.87 4.30
Average 0.03 -0.04 0.02 -0.06 0.33 0.76 0.20 0.41 0.04 -0.23 0.00 -0.36 1.15 3.08 0.80 1.84
̅̅̅̅̅ /𝜎𝑔 .
146
Table 7.A. Short run and long run GDP response to a structural government expenditure innovation from an SVAR with endogenous public debt
Short run Impact (1st Long run impact (5 Convergence to the long
Impact in one year Impact in two years
quarter) years) run
Country Sample Expansion Recession Expansion Recession Expansion Recession Expansion Recession Expansion Recession
Austria 2000:1-2019:2 -0.05 0.29 -0.12 0.63 -0.12 0.69 -0.12 0.70 Fast Medium
Belgium 1995:4-2019:2 0.03 0.17 -0.07 0.62 -0.08 0.84 -0.08 0.95 Fast Slow
Canada 1990:1-2019:2 -0.02 0.29 -0.20 0.89 -0.22 1.16 -0.22 1.24 Fast Medium
Denmark 2000:1-2019:2 0.09 0.30 -0.07 0.56 -0.07 0.58 -0.07 0.58 Fast Fast
Finland 2000:1-2019:2 -0.05 0.23 -0.10 0.38 -0.10 0.38 -0.10 0.38 Fast Fast
France 1998:4-2019:2 0.03 0.21 -0.08 0.70 -0.12 1.02 -0.13 1.22 Medium Slow
Germany 1998:1-2019:2 -0.16 0.20 -0.35 0.68 -0.35 0.72 -0.35 0.72 Fast Fast
Greece 1997:4-2019:2 0.45 0.57 0.60 0.82 0.60 0.83 0.60 0.83 Fast Fast
Ireland 2000:1-2019:2 0.08 0.28 -0.05 0.61 -0.05 0.62 -0.05 0.62 Fast Fast
Italy 1995:4-2019:2 0.11 0.19 0.23 0.54 0.24 0.59 0.24 0.59 Fast Fast
Japan 1997:4-2019:2 0.00 0.24 -0.11 0.41 -0.12 0.42 -0.13 0.42 Fast Fast
Netherlands 1999:4-2019:2 -0.11 -0.02 -0.06 0.33 -0.05 0.45 -0.05 0.48 Fast Medium
Norway 1995:4-2019:2 0.13 0.70 0.02 0.83 0.03 0.82 0.03 0.82 Fast Fast
Portugal 1999:4-2019:2 -0.04 0.12 -0.07 0.33 -0.10 0.39 -0.10 0.40 Medium Medium
Spain 1995:4-2019:2 0.03 0.06 0.04 0.53 0.04 0.96 0.04 1.51 Medium Slow
Sweden 1995:4-2019:2 -0.10 0.36 -0.18 0.97 -0.18 1.10 -0.18 1.11 Fast Medium
UK 1995:1-2019:2 0.04 0.35 -0.02 1.07 -0.03 1.47 -0.03 1.68 Fast Slow
USA 1995:1-2019:2 0.05 0.42 -0.06 1.17 -0.06 1.41 -0.06 1.46 Fast Medium
Average 0.03 0.28 -0.04 0.67 -0.04 0.80 -0.04 0.87 -- --
Note: Convergence to the long run is fast if it is approximately attained in less than 5 quarters, medium between 6 and 9 quarters and slow in case it is reached in
more than 10 quarters (this is clearly visible from the GDP responses charts).
147
Table 8.A. Short run and long run government expenditure multipliers from an SVAR with endogenous public debt
Short run multiplier (1st Long run multiplier (5

1st year multiplier 2nd year multiplier
quarter) years)
Country Sample Expansion Recession Expansion Recession Expansion Recession Expansion Recession
Austria 2000:1-2019:2 -0.25 1.40 -0.58 3.01 -0.57 3.31 -0.57 3.35
Belgium 1995:4-2019:2 0.16 0.90 -0.37 3.21 -0.41 4.37 -0.41 4.91
Canada 1990:1-2019:2 -0.11 1.44 -0.98 4.45 -1.08 5.79 -1.09 6.20
Denmark 2000:1-2019:2 0.33 1.06 -0.25 2.02 -0.25 2.07 -0.25 2.07
Finland 2000:1-2019:2 -0.10 0.47 -0.21 0.75 -0.20 0.76 -0.20 0.76
France 1998:4-2019:2 0.13 0.84 -0.32 2.78 -0.47 4.07 -0.51 4.84
Germany 1998:1-2019:2 -0.77 0.96 -1.69 3.25 -1.68 3.44 -1.68 3.44
Greece 1997:4-2019:2 0.98 1.23 1.31 1.78 1.31 1.80 1.31 1.80
Ireland 2000:1-2019:2 0.15 0.52 -0.10 1.14 -0.09 1.15 -0.09 1.15
Italy 1995:4-2019:2 0.65 1.11 1.31 3.09 1.36 3.39 1.36 3.41
Japan 1997:4-2019:2 0.00 1.30 -0.62 2.23 -0.67 2.25 -0.68 2.25
Netherlands 1999:4-2019:2 -0.42 -0.07 -0.21 1.23 -0.20 1.70 -0.20 1.79
Norway 1995:4-2019:2 0.39 2.04 0.07 2.44 0.08 2.42 0.08 2.42
Portugal 1999:4-2019:2 -0.18 0.54 -0.31 1.46 -0.45 1.74 -0.46 1.78
Spain 1995:4-2019:2 0.15 0.27 0.19 2.47 0.17 4.47 0.17 7.02
Sweden 1995:4-2019:2 -0.31 1.09 -0.55 2.99 -0.55 3.37 -0.55 3.41
United Kingdom 1995:1-2019:2 0.17 1.60 -0.09 4.90 -0.11 6.73 -0.12 7.69
United States 1966:1-2019:2 0.23 1.86 -0.25 5.15 -0.27 6.22 -0.27 6.43
Average -0.25 1.40 -0.58 3.01 -0.57 3.31 -0.57 3.35
̅̅̅̅̅ /𝜎𝑔 .
148
Appendix B. Figures
149
Figure 1.B.a: Sensitivity of GDP responses to expenditures impact multipliers by period of time (1st set of countries)
Austria Austria Belgium Belgium
.35 .035 .5 .20
.34 .034 .18
.33 .4
.033 .16
.32
.032 .3 .14
.31
.30 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2
.031 .2 .12
.29
.030 .10
.28 .1
.27 .029 .08
.26 .028 .0 .06

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Canada Canada Denmark Denmark

.28 .06 .50 .10
.26 .04 .08
.45
.24 .02 .06
.00 .40
.22 .04
-.02
.20 .35 .02
Period: 1966/Q1-1991/Q4 -.04 Period: 1992/Q1-2019/Q2 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2
.18 .00
-.06 .30
.16 -.08 -.02
.25
.14 -.10 -.04
.12 -.12 .20 -.06
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Finland Finland France France

.06 .48 .20 -.1
.04
.44 .15 -.2
.02
.40 .10
.00 -.3
-.02 .36 .05
Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2
-.04 -.4
.32 .00
-.06
-.5
.28 -.05
-.08
-.10 .24 -.10 -.6
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
150
Figure 1B.b: Sensitivity of GDP responses to expenditures impact multipliers by period of time (2nd set of countries)
Germany Germany Greece Greece
.27 .06 .8 .95
.26 .04
.90
.6
.25 .02
.85
.24 .00 .4
.23 -.02 .80
.22 -.04 .2
.75
.21 -.06
.0 .70
.20 -.08
.19 -.10 -.2 .65
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Ireland Ireland Italy Italy

1.0 .788 .36 .52
0.9 .784 .32 .48
0.8 .780 .44
.28
0.7 .776
.40
0.6 .772 .24
.36
0.5 Period: 1966/Q1-1991/Q4 .768 Period: 1992/Q1-2019/Q2 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2
.20
.32
0.4 .764
.16
0.3 .760 .28
0.2 .756 .12 .24

0.1 .752 .08 .20
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Japan Japan Netherland Netherland

.12 .07 .26 .20
.10 .06 .24 .16

.05
.08 .22
.04 .12
.06 .20
.03 .08
.04 .18
.02
.02 .16 .04
.01
.00 .00 .14 .00

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
151
Figure 1.B.c: Sensitivity of GDP responses to expenditures impact multipliers by period of time (3rd set of countries)
Norway Norway Portugal Portugal
.32 .3 .84 .9
.82 .8
.31 .2
.80 .7
.30 .1 .78 .6
Period: 1992/Q1-2019/Q2 .76 .5
Period: 1966/Q1-1991/Q4 Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2
.29 .0 .74 .4
.72 .3
.28 -.1
.70 .2
.27 -.2 .68 .1
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Spain Spain Sweden Sweden

.224 1.2 .34 -.01
.223 .33 -.02
1.0
.222
.32 -.03
.221 0.8 .31 -.04
.220
0.6 .30 -.05
.219
.29 -.06
.218
0.4
.217 .28 -.07
.216 0.2 .27 -.08

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
.00 .070 .04 -.05
.065 .03 -.10
-.01
.060
.02
-.15
-.02 .055
.01
.050 Period: 1966/Q1-1991/Q4 -.20
Period: 1966/Q1-1991/Q4 Period: 1992/Q1-2019/Q2 Period: 1992/Q1-2019/Q2
-.03 .00
.045
-.25
-.01
.040
-.04
-.02 -.30
.035
-.05 .030 -.03 -.35
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
152
Figure 2.B.a: The effects of the business cycle (expansion verusus recession) on GDP responses to government expenditure dynamic structural shock
(1st set of countries)
Au stria Austria Belgium Belgium
-.02 1.4 .03 1.4
-.03 1.2 1.2

.02
-.04
1.0 1.0
-.05 .01
0.8 0.8
EXPANSION RECESSION EXPANSION RECESSIO N
-.06 .00
0.6 0.6
-.07
-.01
-.08 0.4 0.4
-.09 0.2 -.02 0.2

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Can ada Canada Denmark Denmark

.00 1.6 .04 .9
-.04 1.4
.00 .8
-.08 1.2
-.12 1.0 -.04 .7

-.16 0.8 -.08 .6
-.20 0.6
-.12 .5
-.24 0.4
-.28 0.2 -.16 .4

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Fin land Finland France France

.110 .8 .0 1.6
.105 1.4
.7 -.1
.100 1.2
.095 .6 -.2 1.0
.090 0.8
.085 .5 -.3 0.6
.080 0.4
.4 -.4
.075 0.2
.070 .3 -.5 0.0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Note: Shaded area bar corresponds to the fourth quarter.
153
Figure 2.B.b: The effects of the business cycle (expansion verusus recession) on GDP responses to government expenditure dynamic structural shock
(2nd set of countries)
-.08 .9 .47 .62
-.12 .8 .46 .60
-.16 .7 .58
.45
-.20 .6 .56
EXPANSION RECESSION EXPANSION RECESSION
.44
-.24 .5 .54
.43
-.28 .4 .52
-.32 .3 .42 .50

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.56 2.2 .22 .52
.55 2.1 .48
.20
.54 .44
2.0
.53 .18 .40
.52 1.9 .36
.16
.51 EXPANSION RECESSION EXPANSION .32 RECESSION
1.8
.50 .14 .28
1.7
.49 .24
1.6 .12
.48 .20
.47 1.5 .10 .16
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

-.08 1.0 -.065 1.6
0.9 1.4
-.12 -.066
0.8 1.2
-.16 0.7 -.067 1.0
-.20 0.6 -.068 0.8
0.5 0.6
-.24 -.069
0.4 0.4
-.28 0.3 -.070 0.2

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
154
Figure 2.B.c: The effects of the business cycle (expansion verusus recession) on GDP responses to government expenditure dynamic structural shock
(3rd set of countries)
Norway Norway Portugal Portug al
.15 .92 .300 1.8
.275 1.6
.90
.10
.250 1.4
.88
.05 .225 1.2
.86 .200 1.0
.00 .175 0.8
.84
.150 0.6
-.05 .82
.125 0.4
-.10 .80 .100 0.2
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.0880 1.6 -.08 1.1
1.4 -.10 1.0

.0875
0.9
1.2 -.12
.0870
0.8
1.0 -.14
.0865 0.7
0.8 -.16
0.6
.0860
0.6 -.18
0.5
.0855 0.4 -.20 0.4
.0850 0.2 -.22 0.3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
.00 1.6 .00 1.1
1.4 1.0
-.02
-.04
0.9
1.2
-.04
-.08 0.8
1.0
-.06 0.7
0.8 -.12 0.6
-.08
0.6
0.5
-.10 -.16
0.4 0.4
-.12 0.2 -.20 0.3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
155
Figure 3.B. Expenditure multipliers effects under the business cycle versus the public debt ratio average
0.6 Short run impacts in expansion Short run impacts in recession

2.0
0.4 1.5
1.0
0.2
0.5
0.0
0 50 100 150 200 0.0
0 50 100 150 200
-0.2
Long run impacts in expansion Long run impacts in recession
0.6 2.5
0.4 2.0
0.2
1.5
0.0
0 50 100 150 200 1.0
-0.2
0.5
-0.4
0.0
-0.6 0 50 100 150 200
Figure 4.B. Expenditure multipliers under the debt ratio movements versus the public debt ratio average
Short run multipliers under debt accumulation Long run multipliers under debt accumulation
200 200
150 150
100 100
50 50
0 0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 0.00 5.00 10.00 15.00
Short run multipliers under debt reduction Short run multipliers under debt reduction
200 200
150 150
100 100
50 50
0 0
-0.50 0.00 0.50 1.00 1.50 2.00 -2.00 0.00 2.00 4.00 6.00 8.00
156
Figure 5.B.a: The effects of the debt ratio movments (accumulation versus contraction) on GDP responses to government expenditure dynamic
structural shock (1st set of countries)
.40 .26 1.4 .50
.36 .24 1.2 .45

.22
.32 1.0 .40
.20
.28 0.8 .35
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMULATION DEBT REDUCTION
.18
.24 0.6 .30
.16
.20 .14 0.4 .25
.16 .12 0.2 .20

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.9 .11 .340 .15
.10 .14
.8 .335
.09 .13
.7 .08 .330
.12
.07
.6 .325 .11
DEBT ACCUMULATION .06 DEBT REDUCTION DEBT ACCUMULATION DEBT REDUCTION
.10
.5 .05 .320
.04 .09
.4 .315
.03 .08
.3 .02 .310 .07
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.24 .20 .7 .45
.6 .40
.22 .19
.35
.5
.20 .18 .30
.4 .25
.18 .17
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMULATION .20 DEBT REDUCTION
.3
.16 .16 .15
.2
.10
.14 .15 .1 .05
.12 .14 .0 .00
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
157
Figure 5.B.b: The effects of the debt ratio movments (accumulation versus contraction) on GDP responses to government expenditure dynamic
structural shock (2nd set of countries)
.40 .100 1.3 1.05
.095 1.00
.36 1.2
0.95
.090
.32 1.1
0.90
.085
.28 1.0 0.85
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMUL ATION DEBT REDUCTION
.080
0.80
.24 0.9
.075
0.75
.20 .070 0.8
0.70
.16 .065 0.7 0.65
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Ireland Irelan d Italy Italy

1.84 .4 1.0 .8
1.80 0.9
.7
.3
0.8
1.76
0.7 .6
1.72 .2
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMUL ATION DEBT REDUCTION
0.6 .5
1.68
0.5
.1
.4
1.64 0.4
1.60 .0 0.3 .3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Japan Jap an Neth erlan d Netherland

.17 .22 .9 .12
.16 .20 .8
.15 .08
.18 .7
.14
.16 .6 .04
.13
.14 .5
.12 DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMUL ATION DEBT REDUCTION
.12 .4 .00
.11
.10 .10 .3
-.04
.09 .08 .2
.08 .06 .1 -.08
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
158
Figure 5.B.c: The effects of the debt ratio movments (accumulation versus contraction) on GDP responses to government expenditure dynamic
structural shock (3rd set of countries)
.55 .64 .8 .32
.50 .62 .7 .28

.60
.45 .6 .24
.58
.40 .56 .5 .20
DEBT ACCUMULATION .54 DEBT REDUCTION DEBT ACCUMULATION DEBT REDUCTION
.35 .4 .16
.52
.30 .3 .12
.50
.25 .48 .2 .08
.20 .46 .1 .04

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

3.0 1.4 .8 .0428
.0424
2.5 1.2
.7 .0420
2.0 1.0 .0416
.6
.0412
1.5 0.8
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMULATION .0408 DEBT REDUCTION
.5
1.0 0.6 .0404
.4 .0400
0.5 0.4
.0396
0.0 0.2 .3 .0392
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
1.6 .48 .30 .04
1.4 .44 .28 .02

.26 .00
1.2 .40
.24 -.02
1.0 .36
.22 -.04
DEBT ACCUMULATION DEBT REDUCTION DEBT ACCUMULATION DEBT REDUCTION
0.8 .32
.20 -.06
0.6 .28
.18 -.08
0.4 .24 .16 -.10
0.2 .20 .14 -.12
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
159
Figure 6.B.a: The effects of the debt ratio accumulation and the business cycle on GDP responses to government expenditure dynamic structural shock
.104 .65 .9 1.6
.60 .8 1.4
.100
.55
.7 1.2
.096
.50
.6 1.0
.092 .45
DEBT ACCUMULATION and EXPANSION DEBT ACCUMULATION and RECESSION DEBT ACCUMULATION and EXPANSION DEBT ACCUMULATION and RECESSION
.5 0.8
.40
.088
.4 0.6
.35
.084 .3 0.4
.30
.080 .25 .2 0.2
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.60 1.6 .226 .55
.55 .224
1.4
.50
.222
.50
1.2
.220 .45
.45
1.0 .218
.40 .40
.216
0.8
.35
.214
.35
.30 0.6
.212
.25 0.4 .210 .30
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.055 .34 .40 1.4
.32 .35 1.2
.050
.30 .30
1.0
.045
.28 .25
0.8
.040 .26 .20
0.6
.24 .15
.035
0.4
.22 .10
.030 0.2
.20 .05
.025 .18 .00 0.0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
160
Figure 6.B.b: The effects of the debt ratio accumulation and the business cycle on GDP responses to government expenditure dynamic structural shock
.11 .7 1.3 .85
.10 1.2 .80

.6
.09 1.1 .75

.5
.08 1.0 .70
.4
.07 0.9 .65
.3
.06 0.8 .60
.05 .2 0.7 .55

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Ireland Irelan d Italy Italy

1.72 1.80 .7 .8
1.68 1.76 .6 .7
1.64 1.72 .5 .6
1.60 1.68 .4 .5
1.56 1.64 .3 .4
1.52 1.60 .2 .3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Japan Jap an Neth erlan d Netherland

-.10 .60 .7 1.0
.55 .6 0.9
-.15 0.8
.50 .5
0.7
-.20 .45 .4 0.6
DEBT ACCUMULATION and EXPANSION DEBT ACCUMULATION and RECESSION DEBT ACCUMULATION and EXPANSION 0.5 DEBT ACCUMULATION and RECESSION
-.25 .40 .3
0.4
.35 .2
-.30 0.3
.30 .1 0.2
-.35 .25 .0 0.1
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
161
Figure 6.B.c: The effects of the debt ratio accumulation and the business cycle on GDP responses to government expenditure dynamic structural shock
Norway Norway Portu gal Portugal
.48 .80 .50 .50
.44 .45
.79 .45
.40
.40
.78 .40
.35
.36
.77 .35 .30
.32
.25
.76 .30
.28
.20
.24 .75 .25
.15
.20 .74 .20 .10
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

2.4 3.0 .55 1.2
.50 1.1
2.0 2.5
1.0
.45
1.6 2.0 0.9
.40 0.8
1.2 1.5
DEBT ACCUMULATION and EXPANSION DEBT ACCUMULATION and RECESSION DEBT ACCUMULATION and EXPANSION 0.7 DEBT ACCUMULATION and RECESSION
.35
0.8 1.0 0.6
.30
0.5
0.4 0.5 .25 0.4
0.0 0.0 .20 0.3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
1.0 1.6 .32 1.1
0.9 1.4 .30 1.0
0.8 .28 0.9
1.2
0.7 .26 0.8
1.0
0.6 .24 0.7
0.8
0.5 .22 0.6
0.6
0.4 .20 0.5
0.3 0.4 .18 0.4
0.2 0.2 .16 0.3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
162
Figure 7.B.a: The effects of the debt ratio reduction and the business cycle on GDP responses to government expenditure dynamic structural shock
Austria Austria Belgium Belg ium
-.06 .28 .02 .60
.26 .00 .55

-.08
.50
.24 -.02
-.10
.45
.22 -.04
-.12 .40
DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION
.20 -.06
.35
-.14
.18 -.08
.30
-.16 .16 -.10 .25
-.18 .14 -.12 .20
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

-.04 .36 .02 .22
.00 .21
-.08 .32 -.02 .20
-.12 -.04
.28 .19
-.06
-.16 .18
DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION -.08 DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION
.24 .17
-.20 -.10
-.12 .16
.20
-.24
-.14 .15
-.28 .16 -.16 .14
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Finland Finland France Fran ce

-.04 .168 -.10 .8
-.15 .7
-.05 .164
-.20
.6
-.06
.160 -.25
.5
-.07 -.30
.156 .4
-.35
-.08
.3
-.40
.152
-.09 .2
-.45
-.10 .148 -.50 .1
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
163
Figure 7.B.b: The effects of the debt ratio reduction and the business cycle on GDP responses to government expenditure dynamic structural shock
-.20 .20 .84 .40
.80 .39
-.24 .18
.38
.76
-.28 .16 .37
.72 .36
-.32 .14
DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION DEBT REDUCTION and EXPANSION .35 DEBT REDUCTION and RECESSION
.68
-.36 .12 .34
.64
.33
-.40 .10 .60 .32
-.44 .08 .56 .31
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.30 .36 .44 .52
.25 .32 .40 .48

.28
.20 .36 .44
.24
.15 .20 .32 .40
DEBT REDUCTION and EXPANSION .16 DEBT REDUCTION and RECESSION DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION
.10 .28 .36
.12
.05 .24 .32
.08
.00 .04 .20 .28
-.05 .00 .16 .24

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

-.16 .55 -.115 .150
-.20 .50 .125

-.120
.100
-.24 .45
-.125
.075
-.28 .40
-.130 .050
-.32 .35
.025
-.135
-.36 .30
.000
-.40 .25 -.140
-.025
-.44 .20 -.145 -.050
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
164
Figure 7.B.c: The effects of the debt ratio reduction and the business cycle on GDP responses to government expenditure dynamic structural shock
.20 .76 .13 .20
.74 .12
.16 .16
.72 .11
.12
.12 .70 .10
DEBT REDUCTION and EXPANSION .68 .09 .08
DEBT REDUCTION and RECESSION DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION
.08 .66 .08
.04
.64 .07
.04
.00
.62 .06
.00 .60 .05 -.04
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.17 1.0 -.14 .24
.16 0.9 -.15 .22
.15 0.8 -.16 .20
-.17
.14 0.7 .18
-.18
.13 0.6 .16
DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION -.19 DEBT REDUCTION and EXPANSION DEBT REDUCTION and RECESSION
.12 0.5 .14
-.20
.11 0.4 -.21 .12
.10 0.3 -.22 .10
.09 0.2 -.23 .08
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
.00 .50 .00 .16
-.02 .45 -.04 .15
-.04 .40 -.08 .14
-.06 .35 -.12 .13

-.08 .30 -.16 .12
-.10 .25 -.20 .11
-.12 .20 -.24 .10

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
165
Figure 8.B. Eexpenditure multiplier under business cycle and public debt movements versus the public debt
ratio average
Short run multipliers in expansion and Long run multipliers in expansion and
debt accumulation debt accumulation
200 200
150 150
100 100
50 50
0 0
-1.00 -0.50 0.00 0.50 1.00 1.50 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
Short run multipliers in expansion and Long run multipliers in expansion and
debt reduction debt reduction
200 200
150 150
100 100
50 50
0 0
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
Short run multipliers in Recession and Long run multipliers in Recession and
200
debt accumulation debt accumulation
200
150 150
100 100
50 50
0 0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 0.00 5.00 10.00 15.00
Short run multipliers in Recession and debt Long run multipliers in Recession and
200 reduction 200 debt reduction
150 150
100 100
50 50
0 0
-0.50 0.00 0.50 1.00 1.50 2.00 0.00 1.00 2.00 3.00 4.00 5.00
166
Figure 9.B.a: The endogenous effects of the public debt ratio and the business cycle on GDP responses to government expenditure dynamic structural
shock from a trivarite SVAR model (1st set of countries)
Austria Austria Belgium Belg ium
-.05 .8 .04 1.0
-.06 0.9
.7 .02
-.07 0.8
.6 .00 0.7
-.08
0.6
-.09 .5 -.02
EXPANSION RECESSION EXPANSION 0.5 RECESSION
-.10
.4 -.04 0.4
-.11 0.3
.3 -.06
-.12 0.2
-.13 .2 -.08 0.1
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.00 1.4 .12 .60
-.04 1.2 .55

.08
.50
-.08 1.0
.04 .45
-.12 0.8
.00 .40
-.16 0.6
.35
-.04
-.20 0.4 .30
-.24 0.2 -.08 .25

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Finland Finland France Fran ce

-.04 .40 .04 1.4
-.05 .38
.00 1.2
-.06 .36
.34 1.0
-.07 -.04
.32
-.08 0.8
EXPANSION .30 RECESSION EXPANSION RECESSION
-.09 -.08
.28 0.6
-.10 .26 -.12
0.4
-.11 .24
-.12 .22 -.16 0.2
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
167
Figure 9.B.b: The endogenous effects of the public debt ratio and the business cycle on GDP responses to government expenditure dynamic structural
shock from a trivarite SVAR model (2nd set of countries)
-.12 .8 .62 .84
.7 .60 .80
-.16
.58
.6 .76
-.20 .56
.5 .54 .72
-.24
EXPANSION RECESSION .52 EXPANSION RECESSION
.4 .68
-.28 .50
.3 .64
.48
-.32 .2 .60
.46
-.36 .1 .44 .56
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.12 .65 .24 .6
.60 .22
.08 .5
.55
.20
.04
.50 .4
.18
.00 .45
.16 .3
.40
-.04
.14
.35
.2
-.08 .12
.30
-.12 .25 .10 .1
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
Japan Japan Netherlands Netherlands

.00 .44 -.05 .5
-.02 .40 -.06 .4

-.04 -.07
.36 .3
-.06 -.08
.32 .2
-.08 -.09
.28 .1
-.10 -.10
-.12 .24 -.11 .0
-.14 .20 -.12 -.1

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
168
Figure 9.B.c: The endogenous effects of the public debt ratio and the business cycle on GDP responses to government expenditure dynamic structural
shock from a trivarite SVAR model (3rd set of countries)
.16 .88 -.03 .40
-.04
.35
.12 .84
-.05
.30
.08 .80 -.06
-.07 .25
.04 .76 -.08
.20
-.09
.00 .72
.15
-.10
-.04 .68 -.11 .10
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

.042 1.6 -.08 1.2
1.4 1.1
.040 -.10
1.2 1.0
.038 -.12 0.9
1.0
0.8
.036 0.8 -.14
EXPANSION RECESSION EXPANSION 0.7 RECESSION
0.6
.034 -.16 0.6
0.4 0.5
.032 -.18
0.2 0.4
.030 0.0 -.20 0.3
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
UK UK USA USA
.04 1.8 .06 1.4
.03 1.6 .04 1.2

1.4
.02 .02
1.0
1.2
.01 .00
1.0 0.8
.00 -.02
0.8
0.6
-.01 -.04
0.6
-.02 -.06 0.4
0.4
-.03 0.2 -.08 0.2
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
169
Figure 10.B The expenditure multipliers from the trivariate SVAR with endogenous debt varibale and
controlled for the business cycle, scattered with the average level of the public debt ratio
Short run impact in Expansion Long run impact in Expansion

200 200
150 150
100 100
50 50
0 0
-0.20 0.00 0.20 0.40 0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80
Short run impact in Recession Long run impact in Recession

200 200
150 150
100 100
50 50
0 0
-0.20 0.00 0.20 0.40 0.60 0.80 0.00 0.50 1.00 1.50 2.00
170
Figure 11.B.a. Impulse response functions in time of economic expansion

Response ofLGDP to Shock2 Response ofLPINV to Shock2 Response ofLPRC to Shock2 Response ofRINTR to Shock2 Response ofLDR to Shock2
.0004 .002 .0004 .0004 .016
.000 .0002
.0000 .012
.0000
-.002 .0000
-.0004 .008
-.0004 -.004 -.0002
-.0008 .004
-.006 -.0004
-.0008
-.0012 .000
-.008 -.0006
-.0012 -.010 -.0008 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.006 .04 .0004 .0008 .002
.005 .0002
.03 .0004 .000
.004
.0000
.003 .02 .0000 -.002
-.0002
.002 .01 -.0004 -.004
-.0004
.001
.00 -.0008 -.006
.000 -.0006
-.001 -.01 -.0008 -.0012 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0002 .001 .003 .0005 .004
.000 .0004
.0000 .002
-.001 .002 .0003
-.0002 -.002 .0002 .000
-.0004 -.003 .001 .0001 -.002

-.004 .0000
-.0006 -.004
-.005 .000 -.0001
-.0008 -.006 -.0002 -.006

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0015 .012 .0010 .005 .001
.0008 .004 .000

.0010
.008
.0006 .003 -.001
.0005
.004 .0004 .002 -.002
.0000
.0002 .001 -.003
.000
-.0005
.0000 .000 -.004
-.0010 -.004 -.0002 -.001 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
171
Figure 11.B.b. Impulse response functions in time of economic recession

.0025 .006 .0008 .0004 .016
.0020
.004 .0000 .012
.0004
.0015
.002 -.0004 .008
.0010
.000 .0000 -.0008 .004
.0005
-.002 -.0012 .000

.0000 -.0004
-.0005 -.004 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.005 .04 .0004 .0004 .002
.004 .001
.03 .0002
.0000
.000
.003
.02 .0000 -.001
.002 -.0004
.01 -.0002 -.002
.001
-.003
-.0008
.00 -.0004
.000 -.004
-.001 -.01 -.0006 -.0012 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0025 .005 .0030 .0006 .001
.0020 .004 .0025 .0005 .000
.0004
.0015 .003 .0020 -.001
.0003
.0010 .002 .0015 -.002
.0002
.0005 .001 .0010 -.003
.0001
.0000 .000 .0005 .0000 -.004
-.0005 -.001 .0000 -.0001 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0012 .00100 .005 .001
.0008 .008 .004

.00075 .000
.0004 .00050 .003

.004 -.001
.0000 .002
.00025
-.002
-.0004 .001
.000 .00000
-.003
-.0008 .000
-.00025
-.0012 -.004 -.001 -.004
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
172
Figure 11.B.c. Impulse response functions in time of debt accumulation

.0010 .004 .0004 .0005 .016
.0005
.0000 .0000 .012
.000
.0000
-.0004 -.0005 .008
-.0005 -.004
-.0008 -.0010 .004
-.0010
-.008
-.0012 -.0015 .000
-.0015
-.0020 -.012 -.0016 -.0020 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0006 .0008 .002
.0004
.006 .03 .0004 .000
.0002
.004 .02 .0000 -.002
.0000
.002 .01 -.0004 -.004
-.0002
.000 .00 -.0008 -.006

-.0004
-.002 -.01 -.0006 -.0012 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0010 .002 .0028 .0005 .002
.0024 .0004 .001

.0008
.001
.0020
.0003 .000
.0006
.000 .0016
.0002 -.001
.0004 .0012
-.001 .0001 -.002
.0008
.0002
.0000 -.003
.0004
-.002
.0000 -.0001 -.004
.0000
-.0002 -.003 -.0004 -.0002 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0020 .012 .0012 .005 .001
.0015 .004 .000

.008
.0008
.0010 .003 -.001
.0005 .004 .002 -.002

.0004
.0000 .001 -.003
.000
-.0005 .0000 .000 -.004
-.0010 -.004 -.001 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
173
Figure 11.B.d. Impulse response functions in time of debt reduction

Response ofLGDP to Shock2 Response ofLPINV to Shock2 Response ofLPRC to Shock2 Response ofRIN TR to Shock2 Response ofLDR to Shock2
.0028 .008 .00100 .0004 .016
.0024
.00075 .0000
.012
.0020 .004
.00050 -.0004
.0016 .008
.0012 .000 .00025 -.0008
.0008 .004
.00000 -.0012
.0004 -.004
.000
-.00025 -.0016
.0000
-.0004 -.008 -.00050 -.0020 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0006 .0008 .002
.0004
.006 .03 .000
.0004
.0002
.004 .02 -.002
.0000 .0000
.002 .01 -.004
-.0002
-.0004
.000 .00 -.006
-.0004
-.002 -.01 -.0006 -.0008 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0012 .003 .0030 .0008 .002
.002 .0025 .001

.0008 .0006
.000
.001 .0020
.0004 .0004 -.001
.000 .0015
.0000 .0002 -.002
-.001 .0010
-.003
-.0004 .0000
-.002 .0005 -.004
-.0008 -.003 .0000 -.0002 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.002 .0100 .00100 .005 .001
.0075 .00075 .004

.000
.001
.0050 .00050 .003
-.001
.0025 .002
.00025
.000 -.002
.0000 .001
.00000
-.003
-.001 -.0025 .000
-.00025
-.0050 -.001 -.004
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
174
Figure 11.B.e. Impulse response functions in time of debt reduction and expansion
.004 .008 .00100 .0004 .016
.00075
.003 .0000 .012
.004
.00050
.002 -.0004 .008
.00025
.001 .000 -.0008 .004
.00000
.000 -.0012 .000

-.004 -.00025
-.001 -.00050 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0006 .0008 .002
.0004
.006 .03 .0004 .000
.0002
.004 .02 .0000 -.002
.0000
.002 .01 -.0004 -.004
-.0002
.000 .00 -.0008 -.006

-.0004
-.002 -.01 -.0006 -.0012 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0012 .002 .0030 .0005 .002
.001 .0025 .0004 .001

.0008
.0003 .000
.000 .0020
.0004 .0002 -.001
-.001 .0015
.0000 .0001 -.002
-.002 .0010
.0000 -.003
-.0004
-.003 .0005 -.0001 -.004
-.0008 -.004 .0000 -.0002 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.002 .012 .00100 .005 .001
.00075 .004
.000
.008
.001
.00050 .003
-.001
.004 .002
.00025
.000 -.002
.001
.00000
.000
-.003
-.001 .000
-.00025
-.004 -.001 -.004
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
175
Figure 11.B.f. Impulse response functions in time of debt reduction and recession
.0016 .004 .0008 .0004 .016
.002
.0012 .0000 .012
.0004
.000
.0008 -.0004 .008
-.002
.0004 .0000 -.0008 .004
-.004
.0000 -.0012 .000

-.006 -.0004
-.0004 -.008 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.006 .04 .0006 .00050 .002
.005 .0004 .00025

.03 .000
.004
.0002 .00000
.003 .02 -.002
.0000 -.00025
.002 .01 -.004
-.0002 -.00050
.001
.00 -.006
.000 -.0004 -.00075
-.001 -.01 -.0006 -.00100 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0020 .005 .0030 .0008 .001
.004 .0025 .000

.0015 .0006
.003 .0020 -.001

.0010 .0004
.002 .0015 -.002
.0005 .0002
.001 .0010 -.003
.0000 .0000
.000 .0005 -.004
-.0005 -.001 .0000 -.0002 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0015 .0012 .005 .001
.008 .004
.0010 .000
.0008
.003
.0005 .004 -.001
.0004 .002
.0000 -.002
.001
.000
.0000
-.0005 -.003
.000
-.0010 -.004 -.0004 -.001 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
176
Figure 11.B.g. Impulse response functions in time of debt accumulation and expansion
Response ofLGDP to Shock2 R esponse ofLPIN V to Shock 2 Response ofLPRC to Shock2 Response ofRINTR to Shock2 Response ofLDR to Shock2
.0004 .004 .0004 .0004 .016
.0000
.000 .0000 .0000 .012
-.0004
-.0008 -.004 -.0004 -.0004 .008
-.0012 -.008 -.0008 -.0008 .004

-.0016
-.012 -.0012 -.0012 .000
-.0020
-.0024 -.016 -.0016 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.008 .04 .0004 .0008 .002
.0002
.006 .03 .0004 .000
.0000
.004 .02 .0000 -.002
-.0002
.002 .01 -.0004 -.004
-.0004
.000 .00 -.0008 -.006

-.0006
-.002 -.01 -.0008 -.0012 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.00100 .002 .003 .0005 .004
.001 .0004
.00075 .002
.002 .0003
.00050 .000
.0002 .000
-.001
.00025
.001 .0001 -.002
-.002
.00000 .0000
-.004
-.003 .000 -.0001
-.00025
-.004 -.0002 -.006
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0020 .012 .0012 .005 .001
.0015 .004 .000

.008
.0008
.0010 .003 -.001
.0005 .004 .002 -.002

.0004
.0000 .001 -.003
.000
-.0005 .0000 .000 -.004
-.0010 -.004 -.001 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
177
Figure 11.B.h. Impulse response functions in time of debt accumulation and recession
.0028 .008 .0008 .0004 .016
.0024
.006
.0000 .012
.0020
.0004
.004
.0016 -.0004 .008
.0012 .002
.0000 -.0008 .004
.0008
.000
.0004
-.0012 .000
-.002 -.0004
.0000
-.0004 -.004 -.0016 -.004

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.006 .04 .0004 .0008 .002
.005
.03 .0002 .0004 .000
.004
.003 .02 .0000 .0000 -.002
.002 .01 -.0002 -.0004 -.004

.001
.00 -.0004 -.0008 -.006
.000
-.001 -.01 -.0006 -.0012 -.008

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.0020 .004 .0030 .0005 .002
.003 .0025 .0004 .001

.0015
.002 .0003 .000
.0020
.0010 .001 .0002 -.001
.0015
.0005 .000 .0001 -.002
.0010
-.001 .0000 -.003
.0000
-.002 .0005 -.0001 -.004
-.0005 -.003 .0000 -.0002 -.005

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
.002 .012 .00100 .005 .001
.00075 .004
.008 .000
.001
.00050 .003
.004 -.001
.002
.00025
.000 .000 -.002
.001
.00000
-.004 -.003
-.001 .000
-.00025
-.008 -.001 -.004
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
178
Appendix C. Data
Table 1.C: Data availability for the sample of the 18 countries

OECD quarterly national accounts: GDP and general government
consumption
Estimated data
Observed data start after
range
Austria 1996Q1 1966Q1-1995Q4
Belgium 1995Q1 1966Q1-1994Q4
Canada 1966Q1 None
Denmark 1995Q1 1966Q1-1994Q4
Finland 1990Q1 1966Q1-1989Q4
France 1980Q1 1966Q1-1979Q4
Germany 1991Q1 1966Q1-1990Q4
Greece 1995Q1 1966Q1-1994Q4
Ireland 1995Q1 1966Q1-1994Q4
Italy 1996Q1 1966Q1-1995Q4
Japan 1994Q1 1966Q1-1993Q4
Netherlands 1996Q1 1966Q1-1995Q4
Norway 1978Q1 1966Q1-1977Q4
Portugal 1995Q1 1966Q1-1994Q4
Spain 1995Q1 1966Q1-1994Q4
Sweden 1993Q1 1966Q1-1992Q4
United Kingdom 1966Q1 None
United States 1966Q1 None
179
Table 2.C: Descriptive statistics for government consumption growth rate
1966Q1-2019Q2 AUT BEL CAN DEN FIN FRA GER GRC IRE ITA JAP NED NOR POR SPA SWE GBR USA
Mean -0.11 -0.07 -0.09 0.09 -0.04 -0.01 0.04 0.13 -0.38 -0.08 -0.04 -0.07 0.13 0.28 0.20 -0.07 -0.12 -0.28
Median -0.09 -0.09 -0.13 0.08 -0.06 -0.06 0.03 -0.10 -0.15 -0.08 -0.13 -0.09 0.14 0.16 0.10 -0.21 -0.13 -0.39
Maximum 3.92 3.51 3.47 3.49 8.27 8.14 6.25 9.95 8.44 3.56 6.13 6.20 8.24 6.16 3.50 4.93 3.91 3.61
Minimum -3.50 -3.34 -3.41 -3.6 -14.55 -6.76 -4.89 -10.2 -18.48 -5.12 -5.13 -11.08 -4.35 -3.14 -4.66 -4.93 -5.47 -4.47
Std. Dev. 1.04 0.87 1.18 1.19 2.28 0.97 1.36 2.94 2.55 1.04 1.28 1.85 1.56 1.27 1.11 1.45 1.31 1.12
Skewness 0.18 0.39 0.25 -0.1 -1.19 1.45 0.65 0.21 -1.63 -0.15 0.59 -0.82 0.19 1.33 -0.09 0.32 -0.21 0.46
Kurtosis 4.53 5.26 3.32 3.57 13.60 36.03 6.54 4.09 15.40 5.63 6.74 10.12 6.58 7.59 5.19 4.38 4.24 4.87
Jarque-Bera 21.94 50.70 3.08 3.66 1047.14 9755.39 125.99 12.01 1458.86 62.24 136.80 473.17 115.01 249.98 42.71 20.46 15.11 38.48
Probability 0.00 0.00 0.21 0.16 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Mean -0.16 -0.02 0.07 0.22 0.20 0.03 -0.04 0.25 -0.13 -0.07 -0.32 -0.16 0.26 0.61 0.31 0.21 -0.17 -0.20
Median -0.30 0.07 0.05 0.32 0.32 0.06 -0.07 -0.14 0.06 -0.06 -0.36 0.06 0.29 0.46 0.33 0.03 -0.09 -0.32
Maximum 2.76 2.22 3.47 3.49 8.27 8.14 5.43 9.83 2.88 2.80 6.13 5.72 8.24 6.16 3.50 4.93 3.91 3.61
Minimum -3.50 -3.34 -3.41 -3.0 -14.55 -6.76 -4.89 -10.2 -3.36 -5.12 -5.13 -11.08 -4.22 -3.14 -4.66 -4.93 -5.47 -4.47
Std. Dev. 1.03 0.95 1.40 1.23 2.52 1.24 1.60 3.44 1.22 1.17 1.48 2.41 1.61 1.55 1.30 1.68 1.49 1.35
Skewness 0.24 -0.27 0.01 -0.3 -1.70 1.23 0.44 0.05 -0.42 -0.63 0.71 -0.95 0.66 0.95 -0.64 -0.19 -0.38 0.14
Kurtosis 3.64 3.89 2.62 3.40 14.24 27.03 4.48 3.25 3.19 5.28 6.27 6.48 8.56 5.67 4.64 3.90 3.92 3.85
Jarque-Bera 2.72 4.70 0.64 2.75 591.33 2505.17 12.74 0.31 3.16 29.03 54.47 67.45 140.38 46.10 18.64 4.06 6.13 3.44
Probability 0.26 0.10 0.73 0.25 0.00 0.00 0.00 0.86 0.21 0.00 0.00 0.00 0.00 0.00 0.00 0.13 0.05 0.18
Observations 103 103 103 103 103 103 103 103 103 103 103 103 103 103 103 103 103 103
Mean -0.06 -0.11 -0.24 -0.0 -0.27 -0.04 0.11 0.02 -0.61 -0.10 0.23 0.00 0.00 -0.03 0.11 -0.34 -0.08 -0.35
Median -0.08 -0.13 -0.32 -0.0 -0.34 -0.11 0.05 -0.10 -0.49 -0.08 0.19 -0.14 -0.01 -0.10 -0.03 -0.41 -0.20 -0.42
Maximum 3.92 3.51 2.60 3.36 7.58 2.30 6.25 9.95 8.44 3.56 5.17 6.20 3.62 3.48 3.48 3.65 3.21 3.16
Minimum -3.12 -1.53 -2.93 -3.6 -10.96 -1.42 -3.11 -6.32 -18.48 -2.32 -2.10 -2.74 -4.35 -1.90 -1.79 -2.98 -3.14 -2.00
Std. Dev. 1.05 0.79 0.91 1.14 2.02 0.61 1.09 2.39 3.34 0.92 1.01 1.11 1.51 0.83 0.89 1.13 1.13 0.86
Skewness 0.13 1.38 0.34 0.02 -0.44 1.03 1.39 0.44 -1.22 0.76 1.21 2.55 -0.38 0.77 1.18 1.02 0.26 1.12
Kurtosis 5.34 7.87 4.27 3.96 12.11 5.94 11.34 5.26 9.84 5.56 7.30 15.07 3.83 5.10 5.82 5.46 3.81 5.81
Jarque-Bera 25.49 143.62 9.50 4.20 383.94 59.03 354.38 27.09 241.32 40.60 111.52 786.52 5.79 31.14 61.93 47.02 4.27 59.30
Probability 0.00 0.00 0.01 0.12 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.00 0.00 0.00 0.12 0.00
180
Table 3.C: Descriptive statistics for government debt to DGP ratio (in percent): different country samples (number of observations)
AUT BEL CAN DEN FIN FRA GER GRC IRE ITA JAP NED NOR POR SPA SWE GBR USA
Mean 78.7 111.2 110.0 48.4 52.4 90.0 68.0 145.4 65.0 119.4 182.5 62.9 36.5 101.8 72.0 60.0 62.5 92.0
Median 77.9 110.2 108.7 50.3 48.9 91.7 66.5 136.7 62.7 116.8 176.4 62.7 42.5 86.2 66.2 58.7 50.4 83.7
Maximum 94.0 136.0 133.8 62.7 70.5 113.8 82.4 191.6 129.4 138.8 240.8 78.7 57.9 146.7 106.9 77.2 92.6 138.4
Minimum 65.0 91.4 86.1 29.9 33.2 68.2 57.5 106.0 26.4 103.9 78.2 49.6 4.0 64.9 40.6 47.4 38.6 68.9
Std. Dev. 9.4 9.9 11.9 7.2 10.9 16.2 7.4 34.6 33.5 11.3 45.8 7.8 16.4 30.8 22.4 8.0 21.7 20.7
Skewness 0.1 0.6 0.2 -0.4 0.3 0.1 0.5 0.2 0.5 0.5 -0.4 0.2 -1.0 0.2 0.4 0.6 0.4 1.3
Kurtosis 1.5 3.3 2.4 2.5 1.7 1.3 2.0 1.2 1.9 1.8 2.0 1.9 2.8 1.3 1.7 2.5 1.3 3.2
Jarque-Bera 7.4 5.6 2.4 2.8 6.7 9.0 6.4 10.7 7.3 9.7 6.0 4.4 15.9 10.7 9.4 6.6 14.1 60.0
Probability 0.02 0.06 0.30 0.25 0.03 0.01 0.04 0.00 0.03 0.01 0.05 0.11 0.00 0.00 0.01 0.04 0.00 0.00
Observations 78 95 118 78 78 78 83 78 81 95 90 83 95 83 95 95 98 214
Table 4.C. Unit Root Tests Results

GDP in level GDP growth rate
ADF test statistic PP test statistic ADF test statistic PP test statistic
t-Statistic Prob.* Adj. t-Stat Prob.* t-Statistic Prob.* Adj. t-Stat Prob.*
AUSTRIA 0.2565 0.9755 0.1309 0.9674 -6.0462 0.0000 -14.6192 0.0000
BELGIUM -0.0176 0.9551 0.0951 0.9647 -8.8412 0.0000 -8.8459 0.0000
CANADA 1.2132 0.9982 1.3062 0.9987 -9.9102 0.0000 -10.0397 0.0000
DENMARK 0.1751 0.9705 0.1018 0.9652 -13.9734 0.0000 -14.0792 0.0000
FINLAND -0.4258 0.9011 -0.4393 0.8987 -5.4589 0.0000 -15.1845 0.0000
FRANCE -0.9383 0.7745 -0.8584 0.7998 -10.6825 0.0000 -18.6270 0.0000
GERMANY -0.2277 0.9315 -0.2759 0.9249 -12.9975 0.0000 -13.2900 0.0000
GREECE -1.8730 0.3446 -1.9270 0.3194 -5.0038 0.0000 -14.8132 0.0000
IRELAND 4.3334 1.0000 3.6556 1.0000 -6.3815 0.0000 -16.4083 0.0000
ITALY 3.0219 0.9994 3.2846 0.9998 -6.8989 0.0000 -9.8068 0.0000
JAPAN 4.4270 1.0000 4.1450 1.0000 -4.8786 0.0001 -11.7076 0.0000
NETHERLAND 0.2495 0.9751 0.1247 0.9669 -16.6916 0.0000 -16.5688 0.0000
NORWAY 0.3101 0.9784 0.2740 0.9765 -17.2703 0.0000 -17.0500 0.0000
PORTUGAL -1.3390 0.6114 -1.3711 0.5960 -5.8511 0.0000 -11.6654 0.0000
SPAIN -0.0987 0.9469 -0.2549 0.9278 -4.1685 0.0009 -12.1406 0.0000
SWEDEN 1.2149 0.9982 1.6841 0.9996 -10.0465 0.0000 -16.4551 0.0000
UK 0.5707 0.9887 0.9282 0.9958 -8.0347 0.0000 -13.0782 0.0000
USA 1.6819 0.9996 2.0992 0.9999 -10.5931 0.0000 -10.9155 0.0000
181
Table 4.C. (Continued): Unit Root Tests Results

Government spending in level Government spending growth rate
AUSTRIA -0.9438 0.7727 -0.8712 0.7959 -5.4560 0.0000 -18.1330 0.0000
BELGIUM -2.3989 0.1433 -2.2107 0.2032 -9.4541 0.0000 -9.3844 0.0000
CANADA -0.5448 0.8785 -0.5332 0.8808 -3.8224 0.0032 -16.3319 0.0000
DENMARK -2.1121 0.2402 -2.0783 0.2537 -10.9333 0.0000 -11.8262 0.0000
FINLAND -2.2514 0.1890 -2.0378 0.2706 -19.5700 0.0000 -18.9491 0.0000
FRANCE -1.0675 0.7286 -1.1667 0.6889 -5.8770 0.0000 -10.0236 0.0000
GERMANY 0.5797 0.9889 0.3823 0.9818 -5.8287 0.0000 -17.4314 0.0000
GREECE -2.0379 0.2705 -2.0599 0.2613 -4.3691 0.0004 -12.6555 0.0000
IRELAND 0.4036 0.9827 0.1888 0.9714 -4.6500 0.0001 -15.7641 0.0000
ITALY 1.7708 0.9816 2.6518 0.9982 -4.0238 0.0016 -10.9827 0.0000
JAPAN -0.5480 0.8778 -0.5640 0.8745 -13.4690 0.0000 -13.5146 0.0000
NETHERLAND 0.6128 0.9898 0.5888 0.9892 -21.1820 0.0000 -20.5844 0.0000
NORWAY 1.0377 0.9969 0.9746 0.9963 -19.9566 0.0000 -18.9978 0.0000
PORTUGAL -1.5195 0.5219 -1.6815 0.4392 -2.1947 0.0275 -2.5213 0.0116
SPAIN -0.4393 0.8987 0.3556 0.9806 -2.8333 0.0554 -10.3272 0.0000
SWEDEN 6.7077 1.0000 6.1057 1.0000 -14.4874 0.0000 -14.9436 0.0000
UK 1.2851 0.9986 1.0381 0.9969 -18.0062 0.0000 -17.6151 0.0000
USA -0.3641 0.9115 -0.7758 0.8234 -12.5167 0.0000 -13.3163 0.0000
Table 4.C. (Continued): Unit Root Tests Results

Debt to GDP in level Debt to GDP growth rate
AUSTRIA -1.4969 0.5300 -1.4544 0.5513 -3.4298 0.0130 -10.0248 0.0000
BELGIUM -2.3717 0.1527 -2.3332 0.1639 -2.1233 0.0331 -12.9174 0.0000
CANADA -0.2620 0.5896 0.0971 0.7116 -3.9377 0.0001 -6.5118 0.0000
DENMARK -1.8083 0.3740 -1.9389 0.3131 -8.0075 0.0000 -8.1116 0.0000
FINLAND -0.2617 0.9248 -0.0973 0.9454 -2.5921 0.0101 -10.4421 0.0000
FRANCE -0.5568 0.8728 -0.3347 0.9139 -3.8903 0.0034 -7.4426 0.0000
GERMANY -0.9315 0.7735 -1.1590 0.6886 -7.3715 0.0000 -7.5112 0.0000
GREECE -0.5351 0.8776 -0.5351 0.8776 -10.4692 0.0001 -10.4603 0.0001
IRELAND -0.4447 0.5189 -0.2289 0.6007 -2.5790 0.0105 -6.0019 0.0000
ITALY -0.7654 0.8237 -0.0263 0.9532 -2.4075 0.0163 -8.9413 0.0000
JAPAN 1.4708 0.9642 4.0753 1.0000 -1.9364 0.0509 -5.5025 0.0000
NETHERLAND -0.9068 0.7812 -1.2742 0.6379 -7.5301 0.0000 -7.7581 0.0000
NORWAY -1.9184 0.3226 -1.9162 0.3236 -9.8130 0.0000 -9.8130 0.0000
PORTUGAL -2.0002 0.2862 -0.8953 0.7849 -1.3327 0.1676 -6.7636 0.0000
SPAIN -1.5777 0.4898 -0.0771 0.9480 -1.8333 0.3623 -5.7902 0.0000
SWEDEN -1.2298 0.6589 -1.2438 0.6527 -3.5245 0.0094 -9.3791 0.0000
UK -1.2100 0.6675 -0.1698 0.9376 -2.4065 0.1427 -7.7246 0.0000
USA -1.0519 0.7344 0.0249 0.9590 -5.6685 0.0000 -10.2270 0.0000
182
Table 5.C. Lag length criteria selection
AUSTRIA BELGIUM
Lag LogL LR FPE AIC SC HQ Lag LogL LR FPE AIC SC HQ
0 -247.79 NA 0.236454 4.233752 4.280713 4.252819 0 -210.76 NA 0.126236 3.606154 3.653114* 3.625221
1 -224.32 45.75102 0.169989 3.903713 4.044596* 3.960916* 1 -203.43 14.28995 0.119309* 3.549690* 3.690572 3.606892*
2 -222.45 3.582547 0.176251 3.939806 4.17461 4.035143 2 -200.70 5.236902 0.121905 3.571142 3.805946 3.666479
3 -215.37 13.31614 0.167322 3.887638 4.216363 4.02111 3 -198.79 3.584916 0.126333 3.606642 3.935367 3.740114
4 -207.98 13.66067 0.158017 3.830107 4.252754 4.001714 4 -193.19 10.34576 0.122992 3.579523 4.00217 3.751131
5 -202.83 9.331025 0.155056* 3.810698* 4.327266 4.02044 5 -188.63 8.271055 0.121889 3.570021 4.086589 3.779762
CANADA DENMARK
0 -247.78 NA 0.236421 4.233611 4.280572 4.252678 0 -226.30 NA 0.361858* 4.659249* 4.712004* 4.680587*
1 -223.40 47.52766 0.16736 3.888124 4.029006* 3.945326* 1 -224.50 3.497872 0.378457 4.704062 4.862326 4.768076
2 -218.14 10.06771* 0.163846* 3.866825* 4.101629 3.962163 2 -218.62 11.15729* 0.364273 4.665724 4.929496 4.772414
3 -214.55 6.760736 0.165008 3.873715 4.20244 4.007187 3 -216.02 4.834917 0.374921 4.694226 5.063507 4.843592
4 -211.35 5.915857 0.167307 3.887237 4.309884 4.058844 4 -215.40 1.127446 0.401911 4.76319 5.23798 4.955233
5 -208.50 5.164119 0.170692 3.906771 4.423339 4.116513 5 -212.18 5.713769 0.408727 4.779148 5.359446 5.013866
FINLAND FRANCE
0 -420.09 NA 4.385257 7.154002 7.200962 7.173069 0 -101.19 NA 0.019706 1.748905 1.795865 1.767972
1 -405.51 28.4201 3.66539 6.974667 7.115549* 7.031869 1 -59.74 80.78687* 0.010446* 1.114207* 1.255089* 1.171409*
2 -398.62 13.19426* 3.490509* 6.925700* 7.160504 7.021037* 2 -56.44 6.313582 0.010572 1.126131 1.360935 1.221468
3 -395.54 5.781953 3.546396 6.941407 7.270132 7.074879 3 -52.48 7.4582 0.010581 1.126736 1.455462 1.260209
4 -392.73 5.197684 3.61958 6.961519 7.384165 7.133126 4 -49.50 5.507511 0.010768 1.144005 1.566652 1.315612
5 -388.37 7.901872 3.599509 6.955466 7.472034 7.165208 5 -46.95 4.62483 0.011042 1.168579 1.685148 1.378321
GERMANY GREECE
0 -296.76 NA 0.542231 5.06369 5.110651* 5.082757* 0 -497.55 NA 16.29946 8.466885 8.513846* 8.485953*
1 -290.52 12.15391* 0.522081* 5.025800* 5.166683 5.083003 1 -495.30 4.383 16.79089 8.496569 8.637451 8.553771
2 -288.89 3.120009 0.543532 5.065986 5.30079 5.161324 2 -492.66 5.051277 17.18456 8.519664 8.754467 8.615001
3 -286.62 4.279879 0.559759 5.095225 5.423951 5.228698 3 -479.44 24.87722 14.70027 8.363341 8.692066 8.496813
4 -283.72 5.345811 0.570534 5.113978 5.536625 5.285585 4 -472.77 12.31058* 14.05582 8.318197 8.740843 8.489804
5 -280.44 5.956539 0.57778 5.126106 5.642674 5.335848 5 -467.75 9.116041 13.82016* 8.300797* 8.817365 8.510539
183
Table 5.C. (Continued): Lag length criteria selection
IRELAND ITALY
Lag LogL LR FPE AIC SC HQ Lag LogL LR FPE AIC SC HQ
0 -549.40 NA 39.25271 9.345774 9.392735* 9.364841* 0 -244.67 NA 0.224253 4.180771 4.227732 4.199838
1 -543.51 11.47876 38.01660* 9.313755* 9.454637 9.370958 1 -227.17 34.09829 0.17841 3.952061 4.092943* 4.009263
2 -542.01 2.870534 39.66607 9.356149 9.590953 9.451486 2 -220.16 13.43143 0.169542* 3.900995* 4.135799 3.996332*
3 -538.92 5.81424 40.28945 9.371565 9.70029 9.505037 3 -216.58 6.728176 0.170794 3.908178 4.236903 4.04165
4 -532.89 11.14844* 38.9361 9.337082 9.759729 9.508689 4 -213.37 5.927506 0.173155 3.921593 4.34424 4.0932
5 -530.06 5.134488 39.73469 9.356893 9.873461 9.566635 5 -210.73 4.797627 0.177264 3.944552 4.461121 4.154294
JAPAN NETHERLAND
0 -262.14 NA 0.301558 4.476961 4.523922* 4.496028* 0 -265.72 NA 0.320425 4.537645 4.584606 4.556713
1 -259.85 4.466312 0.310425 4.50592 4.646802 4.563122 1 -254.51 21.84813 0.283576 4.415458 4.556341* 4.472661*
2 -255.20 8.900914 0.307062 4.494947 4.729751 4.590285 2 -248.70 11.13729* 0.275007* 4.384695* 4.619499 4.480032
3 -249.00 11.67568* 0.295846* 4.457558* 4.786283 4.59103 3 -245.20 6.587284 0.277391 4.393147 4.721872 4.526619
4 -245.80 5.899851 0.300012 4.471227 4.893874 4.642834 4 -244.62 1.059464 0.29407 4.451223 4.87387 4.62283
5 -245.12 1.240386 0.317513 4.527431 5.044 4.737173 5 -242.85 3.221844 0.305514 4.488909 5.005478 4.698651
NORWAY PORTUGAL
0 -677.57 NA 2.596628 6.629967 6.662387 6.64308 0 -288.47 NA 0.471187 4.923252 4.970213 4.94232
1 -660.19 34.24606 2.278932* 6.499457* 6.596716* 6.538796* 1 -229.42 115.1041 0.185335 3.990144 4.131026* 4.047347
2 -658.33 3.645647 2.326857 6.520253 6.682351 6.585818 2 -221.20 15.73303 0.172572 3.91871 4.153514 4.014048
3 -653.39 9.540204 2.305721 6.511095 6.738032 6.602885 3 -214.88 11.88885 0.165949 3.8794 4.208126 4.012872*
4 -651.38 3.830484 2.351221 6.530576 6.822352 6.648592 4 -212.72 4.004849 0.171237 3.910455 4.333102 4.082062
5 -645.71 10.73068 2.313449 6.514287 6.870903 6.65853 5 -210.59 3.863845 0.176837 3.942141 4.458709 4.151883
SPAIN SWEDEN
0 -206.45 NA 0.241323 4.254134 4.306889 4.275472 0 -287.95 NA 0.467068 4.914474 4.961435 4.933541
1 -142.34 124.3063 0.070762 3.02728 3.185543* 3.091294 1 -274.34 26.53764 0.396839* 4.751508* 4.892391* 4.808711*
2 -135.00 13.92385 0.066113 2.959193 3.222966 3.065884 2 -272.36 3.790346 0.410701 4.785762 5.020566 4.881099
3 -128.06 12.89801* 0.062276* 2.899090* 3.268371 3.048456* 3 -269.04 6.237914 0.415566 4.797361 5.126087 4.930833
4 -125.95 3.820907 0.064769 2.937791 3.412581 3.129834 4 -263.86 9.571927* 0.407458 4.777342 5.199989 4.948949
5 -124.06 3.354557 0.067678 2.980865 3.561164 3.215584 5 -262.06 3.264871 0.423144 4.814626 5.331194 5.024368
UK USA
0 -272.01 NA 0.356473 4.644257 4.691218 4.663324 0 -468.51 NA 0.337747 4.590297 4.622717 4.60341
1 -241.14 60.17334 0.226067* 4.188807* 4.329689* 4.246009* 1 -455.52 25.59416 0.309396 4.502618 4.599877* 4.541957*
2 -239.14 3.82958 0.233882 4.222714 4.457517 4.318051 2 -450.61 9.577723* 0.306671* 4.493754* 4.655852 4.559318
3 -235.92 6.058877 0.237035 4.235926 4.564651 4.369398 3 -447.49 6.029241 0.309322 4.502327 4.729265 4.594118
4 -234.71 2.23206 0.248599 4.283245 4.705891 4.454852 4 -443.83 6.996665 0.310371 4.505654 4.797431 4.623671
5 -232.84 3.391929 0.257863 4.319341 4.835909 4.529083 5 -442.32 2.856712 0.318035 4.529954 4.886569 4.674196
184
Table 6.C. Gross External Debt Position: Foreign Currency and Domestic Currency Denominated Debt,
as of September 30, 2019
US Dollars in Millions
Foreign Currency 1,418,375
Short-term 416,113
Long-term 1,002,262
Domestic Currency 17,098,332
Short-term 5,865,360
Long-term 11,232,972
Unallocated * 1,904,566
Gross External Debt Position 20,421,273
* The currency composition is unallocated for debt in Direct Investment and Trade credit and advances of the
General Government.
Source: https://ticdata.treasury.gov/Publish/debtb2019q3.html
185
Chapter 3
Public debt effects in theory-based models

with an empirical assessment of the potential
public debt
186
Chapter 3. Public debt effects in theory-based models with an empirical assessment of the potential
public debt
Chapter 3
Public debt effects in theory-based models with an empirical

assessment of the potential public debt
Abstract: Using an endogenous growth model, we simulate a parametrized endogenous potential

debt that a country should target to finance its expenditure, derived from potential government
spending. We run individual estimations for a sample of 20 advanced countries over the period of
1960–2015, augmenting the Cobb-Douglas production function by human capital and government
expenditure stocks. The potential public debt is a time series derived from a simulated potential
public expenditure. It is the debt limit curve that the economy could bear depending on its optimal
production capacity and its long-term economic growth and interest rate. The results show that the
public capital stock elasticities differ across countries, with an average of around 0.3. Compared
to actual public debt, the targeted potential debt drops below the actual one for many countries in
times of crisis, especially after the 2008 financial crisis, suggesting that countries should reduce
their debt to keep potential debt above actual debt. For many countries, potential debt decreases
faster in response to a rapid accumulation of actual debt, leading to an intersection between the
two curves that generally materializes in times of crisis. For others, the debt situation is safer, as
potential debt is higher and generally moves in parallel with actual debt. The average potential
debt ratio moves from high values of 150% to 200% in Switzerland, Sweden, Denmark, Norway,
the United States, Austria and The Netherlands, to low values of 50% to 80% in times of crisis,
for Greece, Italy, Spain, the United Kingdom, France, Germany and Belgium. The potential public
expenditure series shows a decreasing trend over time but it is still higher than the one observed
in many countries in the sample. This is driven by the general decreasing trend of public capital
stock productivity, especially in the Great Moderation Era (1985–2015). The policy message to
increase potential public expenditure, and hence the targeted potential debt, is to enhance such
productivity. This might be achievable through the choice of higher productive public capital,
which implies selective public projects with higher multipliers.
JEL classification: B22, B23, C51, H54, H63.
Keywords: Potential debt, Infinitely lived agents, Endogenous growth, Government expenditure,
New Keynesian models, Public capital.
187
public debt
1. Introduction
To assess the effects of government debt, many theoretical growth models have been designed.
The first class of these models is the infinitely lived agents’ models, initiated by Ramsey (1928).
Later, the debate between economists about agents’ lifetime horizon and the type of
intergenerational operative transfer linkages between such agents led to building the concept of
overlapping generations’ (families/dynasties) models to account for the existence or absence of
continuity between generations provided by such linkages (Samuelson, 1958; Diamond, 1965). In
particular, the most debated ideas are the degree of implication of altruism between generations
(such as a motive bequest) and the lifetime agents’ horizon in determining the existence of
government debt effects on agents’ behaviour in terms of saving, capital accumulation, consumers’
utility and interest rate. Important contributions to this field are those of Diamond (1965), Barro
(1974), Blanchard (1985), Buiter (1988), Aiyagari (1985, 1989) and Weil (1989).69
In this regard, two major propositions emerged. The first, according to the neoclassical framework,
is that government debt crowds out private capital by increasing interest rates (Modigliani, 1961).
The second is the Ricardian equivalence for which its advocates show that debt neutrality could
happen depending particularly on the existence of operative altruistic links (bequests) between
generations (Barro, 1974). In this way, the debate in the economic literature emerged especially in
the 1970s and 1980s, with useful contributions modelling fiscal policy insights, precisely the debt–
tax swap and its effects on welfare utility and interest rates.
While the most important property characterizing the majority of neoclassical models considering
government debt is that they build their reasoning on household behaviour towards public
expenditure and government debt, other literature on the political economy of debt assesses the
effects of public debt, studying the behaviour of governments and the influence of economic and
political institutions. In this regard, two approaches are debated. The first is the normative
approach, where the government is considered a social planner (a benevolent social planner), for
whom the priority is to maximize the social welfare of its individuals in society (Barro, 1979;
Lucas and Stokey, 1983; Aiyagari et al., 2002). The second is the positive approach, considering
public debt as a state variable used by each government as a strategy to influence its successor’s
choices or as a way to shape private economic agents’ expectations (Persson and Svensson, 1989;
Alesina and Tabellini, 1990). In this regard, the most disseminated ideas are related to the effects
of fiscal policy (government debt and spending policies) under governments following committed
rules versus discretionary policies. In particular, the government time inconsistent70 actions have
an impact on the way economic agents form their expectations, which affects their economic
decisions.
69
See, for example, De la Croix and Michel (2010) and Weil (2008) for a large literature review.
70
The government time inconsistency issue was raised by Kydland and Prescott (1977).
188
public debt
The public debt effects are also assessed theoretically and empirically in the class of endogenous
growth models. The pioneering contributions in this modelling area flourished with the
development of endogenous growth models, especially by Romer (1986) and Lucas (1988). These
models were brought as an alternative to the neoclassical growth model of Solow (1956) and Swan
(1956) (the Solow–Swan model). The principal characteristics of such models are focusing on the
accumulation of knowledge and its endogenization (whether this is embodied in the form of
technological progress or in the form of human capital). Since then, many sources of growth have
been integrated, particularly to the production function, as inputs such as innovation, human
capital, ideas and government goods, for example (Jones, 2003, 2005, 2019; Jones and Romer,
2010; Bloom et al., 2019). The latter and similar contributions assess the effects of fiscal policy
(taxes, government debt and spending) generally integrating the public sector into the productive
sector (Barro, 1990). However, high government spending (especially unproductive spending),
jointly with assumed distortionary taxes, leads to low per-capita growth rates, according to the
neoclassical growth theory, or to lower growth, according to endogenous growth theory. These
results contrast with the growing empirical evidence that higher government spending and taxes
(relative to the size of the economy) are not negatively correlated with the growth rate (Corsetti
and Roubini, 1996).71
Recently, fiscal policy, and particularly government debt, was also modelled under the class of
new Keynesian models,72 despite these models still actively prioritizing monetary policy analysis
(see, for example, Rupert and Šustek, 2019). These models differ from the overlapping generations
models in many aspects. For example, instead of considering all taxes as a lump sum, as assumed
in the overlapping generations models, recent new Keynesian models have considered fiscal policy
assuming distortionary taxes. Furthermore, they join (intersect with) the literature of the political
economy of debt by considering the scope of government actions and discuss government policies
under commitment or discretion rules. In particular, some authors argue that optimal public debt
would follow a random walk process whenever the government can achieve a time-inconsistent
policy commitment (Benigno and Woodford, 2003; Schmitt-Grohe and Uribe, 2004).
This chapter is structured as follows. Section 2 reviews the existing literature, namely, the
government debt effects in economic growth models, as debated in the old generation models and
the endogenous growth models, as well as the very recent new Keynesian models and in the new
political economy of debt (positive approach).73 Section 3 describes the theoretical framework
used to assess the effects of government debt (government expenditure) on the economy. Section
4 describes the empirical evidence. Section 5 concludes.
71
Corsetti and Roubini (1996) interpret such facts by the imprecise distinction between productive and non-productive
public spending. Many forms of public spending affect the productivity of the economy differently, either directly or
indirectly. The theoretical prediction of a negative tax rate effect on growth is weakened once public spending can be
qualified as productive.
72
See, for example, Leith and von Thadden (2008), Leith and Wren-Lewis (2013) and Rossi (2014).
73
These models are categorized, and their main results summarized, in Table A.1 of the appendix.
189
public debt
This section is dedicated to the assessment of government debt effects in the theoretical models.
Without claiming completeness, we will make an inventory of the most important contributions,
modelling the effect of debt on the behaviour of economic agents and the subsequent consequences
on the macroeconomic aggregates. First, we consider the government effects in the infinitely lived
agents and the overlapping generations models (henceforward, ILA and OLG models). Second, a
summary of the public debt effects in endogenous growth models is presented. Third, the effects
of public debt are also considered and discussed in the economic literature of the new political
economy of fiscal policy, also known as the positive approach of public debt and fiscal policy. We
also present some studies assessing the effects of public spending and debt in the recent class of
new Keynesian models, specifically in an integrated framework of fiscal and monetary policy. The
fourth section concludes with two critiques of Mankiw (2000) for the ILA and OLG models and
the Chari et al. (2009) critique of the new Keynesian models.
2.1. Government debt effects in the ILA and OLG models
The main ideas debated in the ILA and OLG models are mostly related to the way government
bonds affect the steady state equilibrium interest rate (hence, capital accumulation) and
consumption (welfare utility). In this way, two major ideas are contrasted. The proposition that
public debt increases equilibrium interest rates in the steady state was confronted by the debt-
neutrality idea known as Ricardian equivalence.74 The first idea, as illustrated by Modigliani
(1961), is that, in a full employment model, increasing government debt increases the conception
of households’ net wealth, which raises consumption and hence reduces saving, resulting in an
increase in real interest rates. This reduces the output share resulting from the accumulation of
capital. However, for Barro (1974), the idea that the “government debt effect on aggregate demand
depends on the assumed increase of the households’ net wealth” is only true in the non-full
employment framework. In a full employment context, public debt effects could have no wealth
effect if agents (generations) were economically connected by operative intergenerational
transfers.
In this regard, two slightly different versions of the same neoclassical core model should be
distinguished. The first class of model of growth assumes infinitely lived agents (ILA), (Ramsey,
1928; Cass, 1965). The second version is OLG models, which instead, have shifted the debate,
74
Buiter (1988) defines Ricardian equivalence by stating that: “There is debt neutrality if, given a program for public
expenditure on current goods and services over time, the real equilibrium of the economy is not affected by a change
in the pattern over time of lump-sum taxes. If there is debt neutrality, for instance, the substitution of borrowing today
for lump-sum taxation today…does not affect the current and future behavior of private consumption and capital
formation.”
190
public debt
considering the intergenerational linkages between generations rather than their lifetime horizon75
(Samuelson, 1958; Diamond, 1965). The subsequent contributions based on these models show
that the debt effect is due to the degree of altruism existing between young and old generations
rather than the presence of the infinitely lived agents (linkages that could result in agents with
infinite lifetime horizons). This created an intense debate between the neoclassical and proponents
of the Ricardian-equivalence conjecture (Barro, 1976; Feldstein, 1976).
Thereby, Diamond (1965) constructed an OLG model to study the effects of government debt
(domestic and foreign) on the long-term competitive equilibrium. The economy assumes an
infinitely long life and agents living for two periods, working in the first and retiring in the second.
The model particularly assesses the effects of domestic and foreign debt on both utility level and
the equilibrium interest rate. In this model, several key assumptions are made about national debt.
First, governments have a one-period maturity avoiding the issue of the expected capital returns.
Second, the debt pays the current interest rate. Third, the debt–labour ratio is held constant. Fourth,
taxes are assumed as a lump sum on the youth generation.
In this model, the effects of government debt on utility and equilibrium interest rates depend on
the coexistence of external and internal debt in the portfolio of government debt. Internal debt
raises the interest rate and decreases the utility level in the efficient competitive equilibrium, and
may increase or decrease it in the inefficient equilibrium. In particular, in the absence of external
debt, domestic debt may increase utility in the case of inefficient competitive equilibrium. External
debt increases the gap between the equilibrium interest rate and economic growth. Specifically, it
moves the interest rate away from the golden rule solution,76 which, in turn, reduces utility. In the
case of an efficient competitive solution, external debt reduces the utility level of individuals in
the long-term equilibrium. However, in the case of an inefficient solution, the external debt effect
can raise or lower the utility independently of the existence of internal debt.77 Furthermore, the
debt swap (the substitution of internal debt by external debt) positively influences the interest rate
and hence negatively influences utility in the efficient equilibrium, while it could reduce or
increase it in the inefficient case. The author differentiates between four effects of public debt on
utility: the effect of domestic debt following changes in the taxes required to finance it, the debt
effect in the relative factor payments, the effect of external debt, and the debt swap effect.
Nevertheless, Barro (1974) constructed a model of overlapping generations based on Samuelson–

Diamond’s core model (Samuelson, 1958; Diamond, 1965) to argue that government bonds
displace the interest rate, and utility in the steady-state equilibrium does not necessarily hold and
depends, in particular, on the existence of operative altruistic links (bequests) between generations.
75
According to some authors (Barro, 1974; Weil, 1989), having operative linkages in OLG models between some
economic agents with defined finite time horizons leads to infinite lifetime connected generations in the OLG models.
76
According to Phelps (1961): “A golden age means a dynamic equilibrium in which output and capital grow
exponentially at the same rate so that the capital–output ratio is stationary over time.”
77
The principal findings of the Diamond model, for which government deficits raise long-term interest rates, have
been empirically tested in many papers, with a mixture of numerical results (Ni, 1999).
191
public debt
Expressly, in the absence of such linkages, the current living generation does not necessarily
consider the welfare of its dependants (future generations). Therefore, shifting the tax burden to
the future may change the effective lifetime budget of the current living generation. Alternatively,
operative altruistic links may cancel the effect on the lifetime budget of the current living
generation, as the latter is aware of the welfare of the future generation. Consequently, debt
neutrality is guaranteed, as the debt-for-tax swap does not affect the resource allocations and
interest rates of the current generation in this case. For Cukierman and Meltzer (1989), it is
sufficient that the bequest motives do not operate for some economic agents only, to conclude
under majority rule, for the non-neutrality of government debt.
Unlike Diamond’s (1965) model, which considers the infinite lifetime horizon, Blanchard (1985)
studies the effects of debt and accumulation of deficit in a finite lived agent’s horizon.78 This is
conducted by constructing what he called an “index of fiscal policy” supposed to capture the
effects of current and anticipated fiscal policy. This index has two parts, of which one shows the
effect of both changes and levels of government spending on aggregate demand, while the second
highlights the effects of government finance; this is reflected by the effects of both government
debt and the expected sequence of accumulated deficits on aggregate demand.
The main conclusions are that a government debt increase displaces the steady-state level of
foreign assets in agents’ wealth in an open economy79 and decreases the steady-state level of
capital and consumption in a closed economy. Consequently, the government can choose any level
of steady state of consumption (open economy) and capital (closed economy) by simply choosing
its level of debt. Similarly, a decrease in current lump-sum taxes increases human wealth and
consumption. The longer the taxes are differed (i.e. shifted to future generations), the larger the
effect. In summary, the increase in government debt and taxes creates initial wealth effects on
consumption, leading to capital decumulation, which makes capital and consumption lower in the
new steady-state level.
For other authors, the way the tax cut is distributed among economic agents plays an important
role in determining the effects of fiscal policy. In this way, Aiyagari (1985) used a modified version
of the OLG model based on Samuelson (1958) to show that Ricardian equivalence80 depends on
the way the tax cut is distributed among agents. In particular, debt neutrality holds in an OLG
framework where this distribution does not change agents’ wealth allocations. The choice of the
OLG model (instead of an ILA model) was attributed to the fact that this type of model allows for
78
Blanchard (1985) imposes a constant instantaneous probability of death over agents’ lifetime. This makes different
agents with different ages and different levels of wealth have the same horizon and assumes the same propensity to
consume. This easily enables the aggregation of the consumption function. Diamond’s (1965) model adopts, however,
specified population and age structures to avoid the aggregation issue.
79
The larger the horizon of agents, the smaller the effect. Specifically, in an infinite horizon, the level of government
debt has no effect on the steady-state level of foreign assets (Blanchard, 1985).
80
Ricardian equivalence is summarized as “the debt–tax swap for financing government spending, [which] has no
effect on the interest rates and consumption allocations”.
192
public debt
a changeable real interest rate (which may be above or below the real growth rate) to different
deficit policy regimes. Furthermore, the OLG models consider heterogeneous agents, which enable
taxes to be distributed differently among them. On the contrary, ILA models do not allow the real
interest rate to go below economic growth, while, in addition, they assume identical economic
agents yielding a uniform tax distribution.
Aiyagari’s results were addressed in response to Miller and Sargent (1984), for whom “a shift to a
different regime with permanently higher deficits will raise the interest rate and may make it
exceed the growth rate”. Aiyagari (1985) shows that this statement depends on how wealth is
distributed, and this does not hold when the distribution of wealth is kept unchanged. His model
shows that a higher level of government spending (or similarly a cut in total taxes) can be financed
by debt alone at an unchanged (and negative) interest rate and with unchanged total taxes, if the
taxes are distributed in a way that maintains wealth distribution. According to the author, this
implies reducing taxes for the younger generation as savers and increasing taxes on the older
generation of non-savers, while keeping total taxes constant. Thus, any actual effect of higher
government spending on interest rates may arise because distributional impacts are not being
controlled for, and not simply because the deficit is higher. So, for Aiyagari (1985), the validity of
Ricardian debt neutrality depends on the way taxes are distributed among taxpayers and not on the
length of the lifetime horizon.
Taking into account the operative linkages between agents, other contributions consider instead
that the effects of government bonds on economic agents’ wealth are related to the way the tax
bases are foreseen. In this regard, Buiter (1988) builds on the Yaari-Blanchard version81 of the
overlapping generations model to show that the real equilibrium of the economy (private
consumption, capital and relative prices) is independent of the pattern of government debt and
lump-sum taxation over time. For the author, the difference between the expected government tax
base and the future tax base of individuals that are alive today explains the variations over time in
the pattern of lump-sum taxation. The former tax base represents the resources of individuals alive
today, and those yet to be born, while the latter represents the resources owned by individuals alive
today only, and not the resources of individuals yet to be born. In particular, economic agents that
are not linked to their future generations through intergenerational transfers do not integrate the
resources of these successors’ generations into their inter-temporal budget constraint. In this
regard, Buiter (1988) shows that debt neutrality holds if, and only if, the population growth rate
and the individual probability of death equal zero.82 Furthermore, under the latter condition, a non-
zero labour productivity rate will not destroy this debt neutrality.
81
Private consumption behaviour is modelled following the Yaari-Blanchard approach (Yaari, 1965; Blanchard,
1985).
82
Blanchard (1985) considers that only an uncertain lifetime condition (a positive non-zero probability of death) is
sufficient to invalidate the debt-neutrality conjecture.
193
public debt
Contrary to the models that take into account intergenerational transfers and those considering the
infinite lifetime horizon induced by such linkages as prerequisites for debt-neutrality validation,83
Weil (1989) developed a model of “overlapping families of infinitely lived agents”84 to show that
the assumption of the “infinite lifetime horizon”85 is not necessarily a condition that induces
Ricardian debt neutrality, as well as, inversely, finite horizons not necessarily implying the
violation of Ricardian neutrality. The model assumes new cohorts entering the economy over time,
for which operative intergenerational linkages between some, but not all, agents, result in “partial
linkages sufficient to endow any agent alive at any date with an effectively infinite economic
horizon”. For Weil (1989), newly arriving families are not linked to pre-existing dynasties through
operative intergenerational transfers.86 The infinite lifetime of the agents (dynasties) is guaranteed
by the continuous arrival of families who are not linked by operative intergenerational transfers
(no intergenerational altruism). The number of newly arriving families (cohorts), which measures
economic disconnectedness (and heterogeneity of the population), is also the growth rate of the
population. The model is viewed as an extreme version of Blanchard’s (1985) version, as it focuses
on the birth rate of new arrival families while setting the probability of death to zero – “agents are
born but never die”. The model of Blanchard (1985) assumes a constant population by equalling
the birth to death rates. Buitter’s (1988) model, however, based on Yaari (1965), assumes distinct
birth and death rates.
By introducing government bonds in his model, which involves levying lump-sum taxes, Weil
(1989) shows that the equilibrium interest rate hinges on the government financing decision and
the rate (speed) of newly arriving cohorts (families). In particular, government bonds increase the
equilibrium interest rate for a positive non-zero population growth rate. For the author, the
anticipated taxes to pay the issued public debt are expected to be compensated by generations that
have not yet been born. However, generations that are alive today do not consider these taxes in
their consumption. This makes them better off and pushes them to spend more, “not because they
might not be alive when future taxes are levied (they will, as they live forever), but because the
future tax base will include new agents to whom they are not economically connected. The real
interest rate must hence rise to maintain aggregate consumption at its market-clearing level.
Infinite lifetimes are therefore not inconsistent with the violation of the Ricardian debt neutrality
83
The relationship between altruistic intergenerational transfers, infinite time horizons and Ricardian debt neutrality
has been widely debated: “It is widely argued that operative intergenerational transfers between all generations,
because they imply infinite horizons, lead to Ricardian neutrality (cf. the debate between Feldstein (1976) and Barro
(1976)). It is also suggested that finite lifespans lead to the violation of the Ricardian proposition (Blanchard, 1985)”
(Weil, 1989).
84
The assumption of “finite lifetime” adopted by OLG models is considered useless in this model by the author.
85
The effective length of consumers’ planning horizon is infinite.
86
“Consider, for instance, a primogeniture economy in which a parent only loves his first-born heir, enough to leave
him a bequest. Assume that all parents have children, and that children do not love their parents. Each child, whether
first-born or not, is, in this economy, linked through operative bequests to the never-ending chain of his first-born
descendants, and is thus part of an infinitely-lived family. Children who are not first-born, however, do not belong to
any pre-existing dynasty, since they were not loved by their parents: they initiate the dynasty to which they belong.
The rate at which new dynasties enter the economy is a reflection, in such an environment, of the proportion of children
who are not loved, or not loved enough, by their parents” (Weil, 1989).
194
public debt
proposition.” Furthermore, Weil (1989) constructed an example for finite horizons, for which
Ricardian debt neutrality is not necessarily violated.
Besides the operative intergenerational transfers, the lifetime horizon and the way taxes and tax
bases are distributed and perceived among generations, an important contribution adds the degrees
of knowledge spillover across generations, as well as the substitutability between consumption and
leisure, to the factors impacting equilibrium interest rates and Ricardian equivalence validation. In
this way, Ni (1999) extended Diamond's OLG model, where capital is the only variable input, by
allowing for capital and labour as inputs, and assuming a learning-by-doing knowledge-based
growth economy in the spirit of Arrow (1962) and Lucas (1988). Diamond’s (1965) model, and
others like it, imply that a government deficit always reduces savings and raises interest rates.
However, Ni’s (1999) main model results show that the effect of government deficit on interest
rates depends on the spillover of knowledge and the elasticity of labour supply. Precisely, in a
neoclassical growth model with elastic labour supply and intergenerational spillover knowledge,
public deficits may not necessarily raise real interest rates.87
The explanation for this is as follows: while a debt-for-tax swap reduces only the capital stock in
the Diamond (1965) model, it reduces both the supply and demand of capital in Ni’s (1999) model.
Then, with consumption and leisure being good substitutes, shifting the tax burden to the future
reduces current savings, capital stock and the labour supply of the near future. The overall impact
on real interest rates is a result of two opposite effects of reduction in the labour supply and the
accumulation of knowledge88 from one side and capital supply from the opposite side. The
reduction in future labour supply and knowledge lowers the equilibrium real interest rate, while
the reduction in capital increases it (as in Diamond, 1965). The first effect may partially offset and
possibly overcome the second one.
To sum up, this section was devoted to the public debt effects in the neoclassical growth models
of ILA and OLG. The most debated question in these models is the links between the validity of
the Ricardian argument and the lifetime horizon of individuals. Indeed, if the presence of agents
with an infinite lifespan in the ILA models (based on Ramsey, 1928) validates the Ricardian
equivalence theorem, the OLG (deriving their core framework from Samuelson’s (1958) model
and Diamond’s (1965) model) assign the validity of Ricardian equivalence to the links that tie the
people who die and those who will be born (Barro, 1974; Blanchard, 1985; Buiter, 1988; Weil,
1989). Beyond debt neutrality and its prerequisites, the neoclassical models show that public debt
raises interest rates, in turn crowding out capital and reducing welfare and utility. However, despite
important contributions in refining the analysis of government debt effects related to the bequest
motives and intergenerational altruistic transfers, the effects are difficult to assess empirically
87
Furthermore, Ni (1999) suggests, using an empirical investigation, that the dynamics of the interest rate effect are
difficult to assess, and the results of empirical studies of the interest rate effect of government deficits should be
carefully interpreted.
88
Government deficits reduce the current labor supply and slow down the accumulation of knowledge.
195
public debt
under the ILA and OLG models. This is a result, in particular, of the difficulty gauging the degree
of such bequest motives between generations.
2.2. Government debt in endogenous growth models
Models of endogenous growth theory were developed to endogenize the role of externalities and
their contribution to explaining the persistence of the long-term per-capita growth rate, as an
alternative to the rival neoclassical Solow–Swan model (Solow, 1956; Swan, 1956).89 The latter
considers the role of such externalities, or what is assumed to be technical progress, as exogenous.
Indeed, the steady-state growth rate in the Solow–Swan model is determined entirely by exogenous
elements, and macroeconomic aggregates (capital, output and consumption) grow at a constant
exogenous rate of the population growth, which makes the per capita corresponding quantities
constant, and hence they do not grow. Therefore, according to Barro and Sala-i-Martin (2004), the
main substantive conclusions about the long term are that steady-state growth rates are independent
of the saving rate or the level of technology. Specifically, a model without technological change
(like the Solow–Swan model) predicts that economies will converge to a steady state with zero
per-capita growth as a consequence of the diminishing returns to capital. Solow's model also
appeared to be “obsolete”, since the total factor productivity (TFP) measure estimated the share of
growth explained by technical progress to be more than 50%, as reported by Jones and Romer
(2010), or ranging between 50% and 70%, according to Hsieh and Klenow (2010). This constitutes
an “empirical” argument for the emergence of endogenous growth models (Hulten, 2001; Aiyar
and Feyrer, 2002; Fuentes and Morales, 2011).
With its standard framework, the Solow–Swan model was unable to explain the persistent per-
capita non-zero growth rates in many developed economies, and hence was highlighted for missing
the determinants of long-term growth. Thus, the crucial goal of the pioneers of the endogenous
growth theory is to encompass other determinants of long-term growth. This includes broadening
the concept of capital, in which the assumption of diminishing return to scale is avoided, to include
other determinants as inputs in the process of production, such as human capital (Lucas, 1988;
Romer, 1990), innovation, ideas and knowledge90 (Grossman and Helpman, 1991; Jones, 1995,
2003), public goods and service flows (Barro, 1990), public capital and productive public capital
(Futagami et al., 1993), public debt (Greiner, 2007).
To provide explicit contributions involving government spending and public debt in endogenous
growth models, we select a benchmark of important contributions. Thus, Barro (1990) was the first
89
In 1956 Solow and Swan published two distinct papers on the same issue, and their model is referred to as the
Solow–Swan model, or often just the Solow model in reference to the more famous of the two economists.
90
Technological progress is also viewed as a form of generating new ideas by which an economy could escape from
diminishing returns to scale in the long term. Consequently, dealing with technological progress as endogenous within
economic growth models, instead of exogenous, is an important strand of the endogenous growth theory. However, a
technical discussion emerged on how to include ideas and some public goods in the neoclassical production function
because of their non-rivalry characteristics (as for the case of ideas) and non-excludable properties (as for some public
services: national defence, justice, law and order).
196
public debt
to introduce government services as flows in the AK modelling framework. This article triggered
a series of theoretical extensions (Barro and Sala-i-Martin, 1992, 1997; Futagami et al., 1993;
Turnovsky, 1997), as well as a variety of empirical contributions (Greiner, 2007, 2011, 2012, 2015,
2016; Ghosh and Mourmouras, 2004a, 2004b; Futagami et al., 2008; Maebayashi et al., 2017;
Minea and Villieu, 2013; Yakita, 2008). The aim of the contributions was to establish the effect of
public spending and government debt policies on productivity and long-term growth from a
perspective of endogenous growth. Public spending encompasses a variety of expenditure and
subsidies covering diverse sectors of health, education, research activities, research and
development (R&D), public roads and infrastructure, defence and security, justice and law
enforcement, and so on. Besides the direct intra-sector effects, some of this spending may also
have externalities on other sectors producing knowledge, ideas and powering human capital by
affecting their productivity. According to Corsetti and Roubini (1996), in addition to positive
effects either on labor productivity or as rents generated proportionally by a fixed factor, as
assumed in previous works (before their paper), productive public spending may also exert an
external effect on the productivity of physical capital.91 The assessment of the external effects of
productive public spending is rather an empirical issue.
The Barro (1990) model highlights an explicit link between government policy and long-term
economic growth in an endogenous growth model by incorporating government investment
expenditure into the neoclassical production function with constant returns. The model studies a
closed economy with infinite lifetime agents and inter-temporal preferences modelled by a utility
function. The author justifies including government services as a separate input of the production
function by the fact that private input is not a close substitute for public input. It is difficult to
ensure some public activities through private firms as their charges are difficult to implement, as
in the case of non-excludable services (national defence and the maintenance of law and order), or
because the service is non-rival (ideas) or because external effects cause private production to be
too low (as argued for basic education).
The model includes public consumption as an input in the production function, such as:
𝑔
𝑦 = 𝑘𝜑 ( ) (1)
𝑘
where 𝑦, 𝑘 and 𝑔 are, respectively, the output per capita, the capital per capita and the government
consumption of goods and services per capita. φ satisfies the conditions of positive and diminishing
marginal products (𝜑′ > 0 and 𝜑" < 0). Considering 𝜑, a Cobb-Douglas production function,
yields:
91
Internet services is an example of how the same public good might affect the productivity of either final goods
and/or human capital accumulation (Corsetti and Roubini, 1996).
197
public debt
𝑦 𝑔 𝛼
= 𝐴( ) (2)
𝑘 𝑘
where A is a constant net marginal product of capital and 0 < 𝛼 < 1. Furthermore, the model
assumes a balanced government budget (tax-financed public services) by a flat-rate income tax
such as:
𝑔 𝛼
𝑔 = 𝑇 = 𝜏𝑦 = 𝜏𝐴 ( ) (3)
𝑘
with 𝑇 being the per capita amount of taxes.
Important implications under the previous assumptions result in different values of government
size (𝑔/𝑦 or τ) having different effects on the long-term growth rate. An increase in τ reduces the
long-term growth rate, while an increase in 𝑔/𝑦 raises it through an increase in marginal
productivity of capital (𝜕𝑦/𝜕𝑘). The two effects cancel each other out for the optimal government
size equalizing government expenditure to output:
𝑔
=𝜏=𝛼 (4)
𝑦
Equation (4) corresponds to the maximum long-term growth rate.92 The growth rate function of
government size is an inverted U relationship (Figure 1). Hence, for a small government, the effect
of raising expenditure 𝑔/𝑦 dominates the effect of raising tax rate τ, while for a large government
size the negative effect of taxes on growth dominates.
Figure 1. The relationship between per-capita long-term economic growth and government size
Growth per capita 𝐲ሶ

𝒚ሶ = 𝒚ሶ 𝒎𝒂𝒙
𝒈 𝒈
=𝝉=𝜶 Government size ( , 𝝉)
𝒚 𝒚
92
For a non-Cobb-Douglas production function, the maximum growth rate depends on the elasticity of substitution
between per-capita government services and per-capita private capital. Similar conditions and results also apply for
the saving rate.
198
public debt
As for the effects on utility, its maximization corresponds to the same conditions that maximize
the economic growth rate if the elasticity of substitution of 𝑦 with respect to 𝑔 equals unity.
Particularly in the case of the Cobb-Douglas production function and the same previous notations,
government size that maximizes utility corresponds to the condition in Equation (4). If the
production function is not a Cobb-Douglas form, the relative size that maximizes utility exceeds
the one that maximizes growth rate if, and only if, the magnitude of substitution between 𝑔 and 𝑘
is superior to unity.
If Barro (1990) considers the flow of public services in the production function, Futagami et al.
(1993) build on this by considering the accumulated productive public capital as an input in the
production function,93 which generates a sustained per-capita growth rate in the long term.
Futagami et al. (1993) argue that public investment stimulates aggregate production indirectly via
the accumulated stock instead of flows, as in Barro (1990). Furthermore, the introduction of the
productive public stock allows dynamic transitional effects analysis instead of being restricted to
steady-state analysis, as in Barro (1990) and others. As a result, they show that a tax rate that
maximizes the economic growth rate turns out higher than the one that maximizes utility.
However, despite these enhancing elements attributed to Futagami et al.’s (1993) model, the latter
still considers that the government budget is balanced at any time, as assumed in Barro (1990).
Therefore, instead of assuming that government is restricted to running a balanced budget in every
period, as in Barro (1990), Futagami et al. (1993), Barro and Sala-i-Martin (1992, 1997) and many
other previous works, subsequent research has omitted this assumption and tried to study fiscal
policies under an unbalanced government budget.94 In this area, Corsetti and Roubini (1996)
consider productive public spending to assess optimal fiscal policy (public spending, tax and
financial policies) in the same framework as endogenous growth models. They relax the balanced
budget assumption to unbalanced budget constraint, thus allowing for government to borrow and
lend. Furthermore, they incorporate in their model a separate human capital accumulation as a
second sector contrary to many previous papers restricting investigation to one-sector models
where public spending can only affect the productivity of the final goods sector. This allows for
studying of the properties of government policies on both the final human and non-human capital
sectors. In particular, they distinguish optimal tax rates for both types of capital under different
93
This was signalled first by Arrow and Kurz (1970), but in a non-endogenous growth framework. Assuming
diminishing returns to scale in private and public capital given an amount of labor services, the per-capita growth rate
depends on the exogenous rate of technological progress alone.
94
Considering an endogenous growth model where the history of debt affects the primary surplus of the government.
Greiner (2014) shows that an economy with a balanced government budget is characterized by a unique balanced
growth path. Inversely, with a permanent public deficit, the balanced growth path is either non-existent or non-unique
and could be either stable or unstable. Moreover, Greiner (2015) shows that a balanced government budget yields
higher balanced growth and welfare and lower inflation than a situation with permanent deficits, especially (for
welfare) when the government does not put a high weight on stabilizing debt. In the absence of the latter condition,
welfare effects hinge on the initial conditions of public debt.
199
public debt
assumptions on technology and distribution, and they analyse the welfare properties of public debt
and assets.
The government service flows are introduced in the production sectors, and the effects of fiscal
policies and debt are studied by deriving four models depending on whether: public spending is
included as input in the output sector, affecting only the productivity of physical capital (Model
1); or affecting only the productivity of human capital (Model 2); or public spending as input in
the human capital sector, affecting, respectively, the same variables (Models 3 and 4).
For some useful details, especially when government service flows enter as input only in the final
production goods, Corsetti and Roubini (1996) consider an aggregate social production function
in the style of Cobb-Douglas, assuming constant returns to scale to its three inputs, namely,
physical capital, human capital and government flows of services, as follows:
𝑌𝑡 = 𝐴(𝑣𝑡 𝐾𝑡 )𝛼𝜖 (𝑢𝑡 𝐻𝑡 )1−𝛼 (𝐺𝑡 )𝛼(1−𝜖) (5)
where 𝑣 and 𝑢 are, respectively, the fraction of total physical and human capital devoted to the
production of final goods, and the productivity of public spending is decreasing in the parameter
0 < 𝜀 < 1. It follows that the optimal government size is deduced by:
𝑔
= 𝛼(1 − 𝜀) (6)
𝑦
As 𝜀 > 0, the optimal government size in Corsetti and Roubini (1996) is less than the one provided
𝑔
by Barro (1990) in Equation (4): = 𝜏 = 𝛼 superior to 𝛼(1 − 𝜀). According to Corsetti and
𝑦
Roubini (1996), the optimal government size should be properly regarded as a result of the optimal
choice of spending that holds with and without distortionary taxation, while Barro (1990) and
others consider the (second-best) optimal choice to be the tax rate under an instantaneously
balanced budget assumption.
Despite Corsetti and Roubini (1996) assuming an unbalanced government budget, they include
public spending flows as productive input, as in Barro (1990), instead of accumulated public
capital, as in Futagami et al. (1993).
Recently, many works95 have concentrated on the issue of public debt, accumulated capital stock
of public spending and, generally, fiscal policy and its effects on long-term growth and welfare, in
an endogenous growth framework, particularly in the presence of unbalanced budget constraint
with debt dynamics. For example, Ghosh and Mourmouras (2004a) extend Futagami et al. (1993)
to the case of welfare-maximizing fiscal rules, in the presence of government debt. Thus, the public
95
A non-exhaustive list includes among these works Greiner and Semmler (1999, 2000), Ghosh and Mourmouras
(2004a, 2004b), Bräuninger (2005), Greiner (2007, 2012, 2015, 2016), Futagami et al. (2008), Yakita (2008), Minea
and Villieu (2013) and Maebayashi et al. (2017).
200
public debt
to private capital ratio is lower under a golden rule of public finance (than under other fiscal
regimes), minimizing crowding-out effects. However, steady-state welfare may be lowered in a
less strict budgetary rule if public consumption rises, leading to crowding-out effects. In the same
context, Futagami et al. (2008) construct an endogenous growth model with productive
government spending where the government finances expenditure through income tax and
government debt and puts a target level of government debt relative to the size of the economy.
The model distinguishes two steady states: a high-growth steady state in which an increase in
government bonds reduces the growth rate; and a low-growth steady state where an increase in
government bonds raises the growth rate. These results are inverted in the case of an income tax
increase. For the level of welfare, it is lower in the low-growth steady state than in the high-growth
steady state.
Nevertheless, Minea and Villieu (2013) assign the existence of the two steady-state growth rates
to the assumption of the public debt target as a ratio to private capital. Therefore, once the target
has been defined in terms of public debt-to-GDP ratio, the model leads to a unique and determined
balanced growth path. Maebayashi et al. (2017) build on the models of Futagami et al. (2008) and
Minea and Villieu (2013) by considering the stock of capital investment rather than flows, as in
those models. Hence, they derive an optimal target debt ratio that depends on the tax rates on wage
income and consumption, as well as the public investment share in total government spending.
The target debt ratio set by the Stability and Growth Pact (SGP) and Maastricht Treaty, namely
60%, is judged to be higher than the optimal level. Moreover, debt reduction based on expenditure
cuts alone improves welfare. In particular, fiscal consolidation based on a target level of debt-to-
GDP ratio rule (i.e. the well-known 60% rule) improves welfare, and the faster the pace of debt
reduction, the greater this improvement is. Furthermore, fiscal consolidation based on expenditure
cuts, jointly with a tax increase, does not always improve welfare. In this case, the welfare gains
(if any) are lower than those under expenditure cuts only.
In the same context of targeted ratio of public debt, other authors study the sustainability of public
finance. Thus, Bräuninger (2005) uses an endogenous growth model in the form of AK production
function to determine a threshold public deficit ratio. Yakita (2008) builds on Futagami et al.’s
(1993) production function to determine a sustainable threshold of public finance that increases in
public capital stock. Consequently, a larger public capital helps to sustain public finance.
Moreover, keeping the debt finance ratio invariable, the threshold of the debt-to-public-capital
ratio increases with reduced public investment in GDP ratio. Increasing public capital ratios in
Bräuninger (2005) and Yakita (2008) requires higher taxes and additional bond issuance, leading
to higher interest rates and crowding-out effects.
Likewise, Greiner and Semmler (1999, 2000) relax the assumption of government balanced budget
and allow for capital market borrowing by the government. Thereby, they analyse the effect of a
deficit-financed increase in productive government spending following some predefined
budgetary regimes. Hence, fiscal policy effects are significantly determined by such budgetary
201
public debt
rules. Likewise, Greiner (2007) analyses an endogenous growth model with public capital and
sustainable public debt.96 The model is used to derive the necessary conditions for the existence of
a sustainable balanced growth path and to analyse the growth effects of deficit-financed increases
in public investment in the sustainable balanced growth path, as well as along the transition path.
Additionally, in a model with elastic labour supply and a government sector in which government
levies a distortionary income tax and issues bonds to finance lump-sum transfers and non-
distortionary public spending, Greiner (2012) shows that the higher the debt ratio, the smaller the
long-term growth rate whenever public spending is adjusted to fulfil the government inter-
temporal budget constraint. However, the public debt ratio has no effect on the balanced growth
rate if the adjustment is on lump-sum transfers.
Analysing the effects of public debt in an endogenous growth model with productive97 public
spending, Greiner (2015) shows that higher debt accompanies smaller long-term growth.
Moreover, discretionary policy, in general, violates the inter-temporal government budget
constraint along a balanced growth path. A balanced government budget gives a unique saddle-
point stable growth path, while a rule-based policy can lead to two saddle-points stable balanced
growth, depending on the inter-temporal elasticity of substitution of consumption and on the
primary surplus policy. For Greiner (2016), an endogenous growth model with public educational
spending shows that the balanced budget policy98 and the policy with a slight deficit yield higher
growth than a debt policy where public debt and GDP grow at the same rate. Furthermore, for high
initial debt ratios and low inter-temporal elasticity of substitution, a strong deficit policy yields
lower welfare than a balanced budget and a slight deficit policy.
In summary, if the neoclassical ILA and OLG models have been interested in the long-term effects
of government fiscal policies (debt and taxation) on the saving–spending behaviour of individuals
and generations, through operative linkages and transfers, the endogenous growth models have
endogenized such policies to assess their effects on the steady-state growth path. Furthermore,
including debt dynamics and/or public capital stock, endogenous models eventually allow for
tracing of the transitional dynamics effects of fiscal and debt policies that could not be ensured by
the ILA and OLG models. The extensions and development of such models triggered a prolific
discussion about the composition of public expenditure (productive as input versus non-productive
as utility), the non-rivalry and non-excludable goods, and the associated effects on physical and
human sectors for each type of public goods. Accordingly, the differentiated effects between the
types of expenditure induce different policies for the government. Despite these important
96
Public debt sustainability is assured by assuming the ratio of the primary surplus to gross domestic income to be a
positive linear function of the debt to income ratio.
97
“The productive public spending can be thought of as encompassing very different types of publicly provided goods
and services, such as justice, enforcement of law and contracts, police services, educational services and government
research activities” (Corsetti and Roubini, 1996).
98
Greiner (2011) compares the outcome of three budgetary rules: the balanced budget rule, a budgetary rule where
debt grows in the long term but at a rate lower than the balanced growth rate, and a rule where public debt grows at
the same rate as all other economic variables and where the inter-temporal budget constraint is fulfilled.
202
public debt
contributions, further enhancement remains, especially in considering the heterogeneity of agents’
behaviour and welfare, which was also an important drawback of the ILA and OLG models
(Mankiw, 2000; Maebayashi et al., 2017). Explicitly, a population could embrace a part of the
agents with Ricardian behaviour, while the other part behaves following a rule of thumb.
2.3. Government debt in the new Keynesian models and the positive approach
This section presents a summary of some very recent new Keynesian models involving public debt
and fiscal policy, as well as the public debt effects in the models, considering a positive approach
where the political regime impacts the trajectory of public debt.
2.3.1. Government debt in the new Keynesian models
Instead of assuming that all taxes are a lump sum, especially in the assumptions about overlapping
generations models, recent works have considered optimal fiscal policy, assuming distortionary
taxes in the class of new Keynesian models in which social welfare is implied from a consumer
utility function. The pioneering works in the new Keynesian modelling framework are principally
those of Christiano et al. (2005) and Smets and Wouters (2007). Government debt effects,
especially in the neoclassical models, are studied in the long term and steady-state equilibrium,
especially in the ILA and OLG models and some endogenous growth models lacking transitional
dynamics. However, the recent generation of the new standard Keynesian models or dynamic
stochastic general equilibrium (DSGE) models have studied optimal government policies
(monetary and fiscal policies) and fiscal consolidation issues considering economic fluctuations
and shocks (in the short and medium term), in which public debt is set in many of these models to
zero.
Nevertheless, some authors have studied the trajectory of public debt in relation to committed
government actions. Particularly, optimal public debt would follow a random walk process
whenever the government can achieve a time-inconsistent policy commitment (Benigno and
Woodford, 2003; Schmitt-Grohe and Uribe, 2004). Specifically, the latter study the implications
of price stickiness for the optimal degree of price volatility. The model considers a government
issuing non-distortionary taxation and can only issue nominally risk-free debt. Specifically, under
the assumption of price stickiness in this class of models, the government (social planner) chooses
to rely more heavily on changes in income tax rates rather than using surprises as a shock absorber
of unexpected innovations in the fiscal budget. The distortions introduced by tax changes are
diminished by spreading them over time, which induces a near random walk property in tax rates
and public debt.
In the context of new Keynesian models augmented by the government’s budget constraint, where
public spending is financed by distortionary taxes and/or debt, Leith and Wren-Lewis (2013)
analyse the optimal response of government debt to shocks, focusing on the type of involved time-
inconsistency policy and its implications for discretionary policies. Like the previous research,
203
public debt
they find that the optimal pre-commitment policy allows debt to follow a random walk path in the
steady state. However, they show that, under a sticky prices framework, governments are tempted
to use their monetary and fiscal policy instruments to change the steady-state level of debt in the
initial period. The debt will be curved to its initial efficient steady state to encounter this temptation
and therefore deter public debt from following the random walk path if following shocks. The new
steady-state debt equates the original (efficient) debt level even though there is no explicit debt
target in the government’s objective function. Analytically, they show that the debt stabilization
instrument crucially depends on the degree of nominal inertia. Furthermore, the size of the debt
stock, and the welfare consequences of introducing debt, are negligible for pre-commitment
policies but can be significant for discretionary policies.
Furthermore, a few studies examine the effects of high debt in fiscal consolidation actions related
to the impact on the magnitude and/or sign of the fiscal multipliers. For example, Mayer et al.
(2013) use a new Keynesian model to analyse whether, and how, the presence of positive levels of
government debt in the steady state influences the responses of macroeconomic variables to a
government spending shock. The model assumes that a fraction of the household sector is
characterized by rule-of-thumb behaviour, as in Galí et al. (2007). They show that large levels of
government debt in the steady state significantly influence the sign and size of short- and medium-
term fiscal multipliers, which therefore depend substantially on the horizon at which the multiplier
is evaluated. Furthermore, there is an interaction between the dynamics of the inflation rate and
the debt level in real terms, which is absent in standard new Keynesian models in which
government debt is assumed to be equal to zero in the steady state. Overall, in the presence of
permanent government debt, the effect of fiscal policy on macroeconomic variables becomes
difficult to predict over time.
2.3.2. The positive approach of public debt
In the previous section the ILA and OLG models were generally developed by economists in an
environment where governments, being benevolent social planners, maximize the utility of their
population. These models are classified under what is referred to in the economic literature as “the
normative approach” (Alesina and Tabellini, 1990), the “tax smoothing” theory of the government
budget (Alesina and Perotti, 1994) or “the equilibrium approach to fiscal policy” (Roubini and
Sachs, 1989). The normative theory of debt and fiscal policy considers public debt as a means of
smoothing consumption by distributing tax distortions over time (Barro, 1979; Turnovsky and
Brock, 1980; Lucas and Stokey, 1983). The core models of this theory assume, in general, a closed
economy without capital where the government is a “benevolent social planner” that maximizes
the utility of a representative agent who consumes, works and saves with the same infinite
(simplified) time horizon of both government and representative agent (Alesina and Perotti, 1994).
However, for advocates of the positive approach (Roubini and Sachs, 1989; Alesina and Tabellini,
1990; Grilli et al., 1991), normative theory, despite explaining the behaviour of debt in many
204
public debt
advanced OECD countries, has been challenged by the rapid accumulation over time of debt in
almost all developed countries. Therefore, it is unable to provide a complete explanation for such
a phenomenon or explain the differences in policies pursued by different countries with
comparable economic conditions.
Alternatively, a positive approach has been the subject of modelling government fiscal policy,
particularly debt policy, in the field of the new political economy of public debt.99 This approach
is particularly interested in the implied impacts of political process on shaping the path of
government debt. The positive approach considers public debt as a state variable used by each
government as a strategy to influence its successor’s choices or to shape private economic agents’
expectations (Persson and Svensson, 1989; Alesina and Tabellini, 1990).
In this way, Roubini and Sachs (1989) notice that, in several industrialized countries, issues of
political management in coalition governments fall behind the slow rate at which fiscal deficits
were reduced during the 1975–85 period. In particular, during this period weaker governments100
had a clear preference for larger deficits. Similarly, Grilli et al. (1991) focus on the role played by
public institutions in offering constraints and incentives that determine the actions of governments.
Governments’ ability to handle growing deficits and debt issues is influenced by the electoral
practice and political process. Grilli et al. (1991), following Roubini and Sachs (1989), note that,
in countries with an electoral system favouring many small political parties, governments generally
have short horizons and therefore act myopically to avoid tackling the hard choices.
Persson and Svensson (1989) consider the level of public debt as the state variable that gives the
current government an instrument to control a rival future government. They compare, in a two-
period perfect-foresight framework, the policy of a conservative government (one that prefers less
debt and deficit), which is certain to be succeeded by a liberal government (a more expansionary
government), with the policy where it is certain that it will remain in power. As a result, a
conservative government may borrow more if it knows it will be succeeded by a liberal
government than it would once certain of remaining in power in the future. Obviously, a
conservative government will collect less tax and leave more public debt than the successor would
prefer. This increases the conservative government’s consumption more than if it remained in
government, while the liberal government (successor) with high debt and constrained resources
reduces consumption more than it would if it ran alone. Thus, the time-consistent level of
99
Alesina and Perotti (1994) provide a survey of the political economy models of budget deficit organized into six
groups: 1- models based upon opportunistic policy-makers and naïve voters with “fiscal illusion”; 2- models of
intergenerational redistribution; 3- models of debt as a strategic variable, linking the current government with the next
one; 4- models of distributional conflicts within social groups and/or political parties in coalition governments; 5-
models of geographically dispersed interests; and 6- models emphasizing the effects of budgetary institutions. In our
case, we are especially interested in the third group, particularly the contributions of Persson and Svensson (1989) and
Alesina and Tabellini (1990).
100
Weaker government is characterized by a short average tenure and by the presence of many political parties in the
ruling coalition.
205
public debt
government consumption is somewhere between the two outcomes that the two governments
would prefer if ruling on their own.
Meanwhile, the Persson and Svensson (1989) approach, known as a principal-agent problem, with
the conservative government being the principal and the liberal successor government being the
agent, simplifies the reality by assuming that the ruling government knows with certainty that it
will be succeeded by a more liberal government. They also assume the homogeneity of
governments’ preferences towards all public goods but different preferences for different levels of
the same public good. In this regard, Alesina and Tabellini (1990) constitute an advancement in
introducing uncertainty about the nature and spending behaviour of successive governments. They
also consider different preferences for different items of public expenditure, while Persson and
Svensson (1989) focus on different levels of the same public good.
Therefore, to properly understand the debt build-up and deficits in several industrialized
economies, Alesina and Tabellini (1990) adopt a positive theory by removing the assumption that
fiscal policy is set by a benevolent social planner who maximizes the welfare of a representative
consumer. Specifically, their findings suggest that differences in political institutions, leading to
different debt policies in different countries, or in the same country at different points in time, help
to explain the debt trajectories over time and across countries. Their model101 is derived from
Lucas and Stokey (1983). In particular, Alesina and Tabellini (1990) show that debt accumulation
and deficit are accentuated by the alternation of elected governments.
Explicitly, they compare the outcome of debt accumulation and deficit in a situation where
governments alternate versus an outcome resulting from a social planning102 government
supposedly elected forever. Specifically, a disagreement between different governments on the
composition of spending between public goods results in a deficit bias and hence an accumulation
of debt higher than would be the case in the situation of a social planner. As explained by Alesina
and Tabellini (1990): “The level of debt left to the last period is larger in a democracy than with a
social planner; namely the social planner would choose to balance the budget in both periods,
while either one of the two parties choose to run a budget deficit in the first period leaving a
positive amount of debt to be repaid in the last period. In this sense, the electoral uncertainty
creates a sub-optimal deficit bias. This bias is stronger for the party with the smaller probability
of reappointment.”
101
The model assumes mainly a constant population of identical individuals with the same time horizon, acting as
consumers, workers and voters. Individuals differ only by their preferences for public goods, supplied by the
government and financed by means of distortionary taxes on labor. The government is elected democratically and is
chosen among two political parties, each maximizing the utility function of its electorate. Disagreements between the
governments are viewed as differences about social welfare.
102
“A social planner: 1) do[es] not face elections; thus, she is reappointed with probability 1 in each period, 2) her
preferences are a weighted average of the preferences of the citizens” (Alesina and Tabellini, 1990).
206
public debt
Furthermore, the equilibrium level of government debt is higher with: 1/ a higher degree of
polarization between alternating governments; and also 2/ with more likelihood of the current
government not being re-elected. Moreover, as the ruling government is unable to curve the
taxation and expenditure policies of its successors (whether the successor belongs to the same or
the opposing party), the law of motion of public debt is the only way in which the fiscal policy of
the ruling government can impact the policies of its successors.
Another important result is related to the probability of re-election. In such an uncertain

environment, both governments have the same incentive for increasing debt, not certain of being
re-elected, and they restrict the next period’s public consumption by increasing borrowing for the
current ruling period. This leads to a bias towards a larger deficit and higher debt for both
governments, which helps to explain the accumulation of public debt in advanced democracies.
2.4. Discussion
The first section reviewing the effects of public debt through the ILA and OLG models leads to
two principal results. The infinitely lived dynasties models, derived from Ramsey (1928), and
adopted by Barro (1974) and others, validate the Ricardian equivalence proposition for which
public debt is neutral. It is worth mentioning that the Ricardian equivalence proposition works
assuming economic agents’ rational expectations. Accordingly, the previous models show that
economic agents (some, but not necessarily all) react to fiscal policy redistribution of the tax
burden among generations through their bequest motives inducing operative transfers to smooth
the pattern of consumption over time. However, because the OLG models of Diamond (1965) and
others lack such bequest motives, a government debt issuance affects the wealth of generations by
raising real interest rates, hence crowding out capital and reducing the steady-state utility.
Despite these important contributions for economic theory in assessing government debt and fiscal
policy effects, these two types of model have been subject to criticism. Mankiw (2000) criticized
the two modelling approaches, arguing against their adequacy and satisfactory role for analysing
fiscal policy. Accordingly, the author is, first, sceptical regarding the assumption that “households
smooth their consumption over time” that is adopted by both versions of the model. This
assumption is far from perfect, according to Mankiw (2000). In particular, current income
significantly impacts consumer spending, as many consumers are far from following completely
rational expectations, instead adopting rule-of-thumb103 behaviour (non-Ricardian behaviour) in
their spending. Second, some individuals may enjoy long lifetime horizons (due to bequest
motives), while others with short time horizons fail to smooth their consumption and accumulate
wealth. Third, many households have net wealth near zero (a striking fact reported in the data),
103
For example, Galí et al. (2007) use a new Keynesian model to empirically test government spending effects on
consumption, involving rule-of-thumb agents that only have access to contemporaneous labour income for
consumption, and Ricardian agents that can smooth consumption by accumulating capital.
207
public debt
and hence no savings, which makes them unable to follow inter-temporal consumption smoothing,
as reported by the Barro-Ramsey or Diamond-Samuelson models.
Thus, for Mankiw (2000), a better model would allow for such heterogeneous behaviour that is
apparent in the data. In this regard, Mankiw (2000) formulated an alternative theory mentioned as
“savers–spenders theory of fiscal policy” to address the neoclassical shortcomings in public
finance policies. This theory shows, in particular, that, even though government debt does not
affect steady-state capital stock and income, it disrupts income distribution and consumption in
the steady-state path and, in turn, raises inequality between spenders and savers. Specifically, a
higher level of debt yields higher taxation to compensate for the interest payments on the debt.
However, the taxes are on both savers and spenders, while the interest payments on debt fall on
the savers’ side. Therefore, a higher level of debt increases the steady-state income and
consumption for savers (with already higher initial income) and lowers it for spenders (with
initially a lower income), which raises inequality between the two groups.
The “savers–spenders” theory of Mankiw has influenced many empirical researchers on fiscal
policy trying to consider the behaviour following the rule of thumb, especially in the new
Keynesian models. However, this type of modelling has also not been immune to criticism. In this
way, Chari et al. (2009) show that this class of model is not yet useful for public policy analysis.
Chari et al. (2009) base their critiques on the model of Smets and Wouters (2007), which
constitutes a fundamental reference for many recent contributions and policy-makers using the
new Keynesian analysis framework. Accordingly, the Smets and Wouters (2007) state-of-the-art
model adds many free parameters to these models, yielding to shocks that are dubiously structural,
as well as many features that are not consistent with the microeconomic evidence.
The drawbacks of the new Keynesian modelling framework were recognized especially after the
2008 financial crisis and are frequently cited in several publications. In this way conferences with
different slogans (rethinking macroeconomics, rebuilding macroeconomic theory, etc.) have
gathered economists from around the world in an attempt to discover why these models failed to
warn about the disaster of the financial crisis. In this way, an important project (the Rebuilding
Macroeconomic Theory Project) asked a number of leading economists to describe how the
benchmark new Keynesian model might be designed after the financial crisis. Approximately
fifteen important articles and contributions (Blanchard, 2018; Carlin and Soskice, 2018; Ghironi,
2018; Haldane and Turrel, 2018; Hendry and Muellbauer, 2018; Krugman, 2018; Linde, 2018;
Mckibbin and Stoeckel, 2018; Reis, 2018; Stiglitz, 2018; Vines and Wills, 2018; Wren-Lewis,
2018; Wright, 2018) by these economists on this question were published in the Oxford Review of
Economic Policy (2018, vol. 34 (1–2)). The authors disagreed that the new Keynesian models
benchmark of Smets and Wouters (2007) should not constitute the starting point for the newly
designed model. Nevertheless, they agree that the core model should consider four elements, as
described in Vines and Wills (2018). The core model should, in particular: i) incorporate financial
frictions rather than assuming that financial intermediation is costless; ii) relax the requirement of
208
public debt
rational expectations; iii) introduce heterogenous agents; and iv) underpin the model with more
appropriate micro-foundations.
3. The theoretical framework
3.1. The choice of an endogenous growth model
We showed in Section 2 that the ILA models, derived from Ramsey (1928), and adopted by Barro
(1974) and others, validate the Ricardian equivalence proposition for which public debt is neutral.
The latter proposition holds in an environment of perfect information assuming economic agents’
rational expectations. Therefore, such models show that economic agents (some, but not
necessarily all) react to fiscal policy redistribution of the tax burden among generations through
their bequest motives inducing operative transfers to smooth the pattern of consumption over time.
However, because the OLG models of Diamond (1965) and others lack such bequest motives, a
government debt issuance affects the wealth of generations by raising real interest rates, in turn
crowding out capital and reducing steady-state utility.
Despite their important contributions to the economic theory assessing government debt and fiscal
policy effects, the ILA and OLG models have been subject to criticism, to some extent diminishing
their contribution. Mankiw (2000) criticized the two modelling approaches, arguing against their
adequacy and satisfactory role in analysing fiscal policy for several reasons:
- First, according to Mankiw (2000), the assumption that “households smooth their
consumption over time” adopted by both versions of the model does not seem convincing.
Specifically, many consumers are far from following completely rational expectations and
instead adopt the rule-of-thumb behaviour (non-Ricardian behaviour) in their spending;
hence, current income significantly impacts consumer spending. For example, Galí et al.
(2007) use a new Keynesian model to empirically test the government spending effects on
consumption, involving rule-of-thumb agents that only have access to contemporaneous
labour income for consumption, and Ricardian agents that can smooth consumption by
accumulating capital.
- Second, some individuals may enjoy long lifetime horizons (due to bequest motives), while
others with short time horizons fail to smooth their consumption and accumulate wealth.
- Third, many households have net wealth near zero (a striking fact reported in the data),
and hence no savings, which makes them unable to follow inter-temporal consumption-
smoothing, as reported by the Barro-Ramsey or Diamond-Samuelson models.
209
public debt
For Mankiw (2000), a better model would allow for such heterogeneous behaviour as is apparent
in the data. He therefore formulated an alternative theory mentioned as “savers–spenders theory
of fiscal policy” to address the neoclassical shortcomings in the public finance policies. This theory
shows that, even though government debt does not affect the steady-state capital stock and income,
it disrupts income distribution and consumption in the steady-state path and, in turn, raises
inequality between spenders and savers.
The “savers–spenders” theory of Mankiw has influenced many empirical researchers on fiscal
policy trying to consider the behaviour of following the rule of thumb, especially in the new
Keynesian models. However, the new Keynesian type of modelling has also not been immune to
criticism. In this way, Chari et al. (2009) show that this class of model is not yet useful for public
policy analysis. Chari et al. (2009) base their critiques on the model of Smets and Wouters (2007),
which constitutes the fundamental reference for many recent contributions and policy-makers
using the new Keynesian analysis framework. Accordingly, the Smets and Wouters (2007) state-
of-the-art model adds many free parameters to these models, yielding to shocks that are dubiously
structural, as well as many features that are not consistent with the microeconomic evidence.
Based on the previous critiques, the endogenous growth modelling framework seems to be a
suitable candidate for our approach. The endogenous growth models via their productive sector
(generally the Cobb-Douglas production function) are extremely flexible in encompassing many
other factors to explain the per-capita long-term growth rate. Therefore, fiscal policy variables
(productive government expenditure or accumulated government capital) are easily integrated to
the production function to assess the effects of such variables in the steady-state (long-term) path
and transitional dynamics.
3.2. Justification of government capital stock and government expenditure flows

in the production function
Many studies have considered the issue of productive government spending, debt and fiscal
policies, and their effects on long-term economic growth and social welfare in the context of
endogenous growth models. We follow this line of research for several reasons and try to avoid
many important issues. We mainly follow the approach of Barro (1990) and Barro and Sala-i-
Martin (2003), but instead consider the public accumulated capital in the production sector rather
than the flow of goods and services provided by the public sector. We also consider human capital,
as in Corsetti and Roubini (1996), which could be an interesting measurable input. Our approach
differs from Barro (1990) and Corsetti and Roubini (1996), as the latter use productive government
flows in the production function, while our study uses accumulated public capital, as in Futagami
et al. (1993) and Maebayashi et al. (2017).
In fact, we judge that models (either theoretical or empirical) using government expenditure flows
instead of public capital stock lack some logic and compatibility with the framework of the Cobb-
Douglas production function, especially when we consider the government as the production
210
public debt
sector, which leads to considering productive government expenditure. This results in missing an
important contribution of the earlier accumulated stock of public expenditure. As the model of
production function, and others considered in the endogenous growth framework, are non-dynamic
(no lag or inertia of the endogenous variable is present in the explanatory variable), considering
the stock of capital ensures the dynamic effects of public expenditure. This means, economically,
that the earlier flow contributions in the output are considered. However, considering only current
flows in the production function implies that one considers the earlier accumulated flows (which
builds the public stock) fully consumed or as if depreciated with a 100% depreciation rate.
Similar earlier critiques have been addressed to the Scully (1996, 1999, 2003) model calculating
growth-maximizing of the government size. The Scully model formulae follow a form similar to
the Cobb-Douglas framework, linking the current output to the first lagged government flows of
expenditure and the first lag of output. Furthermore, assuming that the government budget is
balanced and, with the government budget constraint equalizing government expenditure flows to
the taxes represented as a share of the output, this leads to the following model: 𝑌𝑖𝑡 =
𝛽0 (𝜏𝑌𝑖𝑡−1 )𝛽1 [(1 − 𝜏)𝑌𝑖𝑡−1 ]𝛽2 , with 𝑌 as output, 𝜏 as the lump sum tax rate, and 𝑡 as the time
period. This model is used to derive an optimal government size (equivalent to optimal taxation
𝛽2
rate, as the balanced budget constrained is assumed) represented by the taxation rate as 𝜏 = .
𝛽1 +𝛽2
The Scully model form is highlighted as having many drawbacks and producing spurious
regressions by many authors (Chapple, 1997; Easton, 1999; Kennedy, 2000 and Hill, 2008).
Kennedy’s (2000) critique is related to the unfounded relationship of this particular production
function, which he says as if the public capital were totally used up every year. Moreover, Hill
(2008) shows that this model should use the lagged tax rate in the previous equation instead of a
current tax rate. When correcting this error, the growth-maximizing size of the state varies between
9% and 29% for the United States data, while Scully (1996) reports a government size of 19% for
the same country.
Beyond that, despite many studies enriching the production function with other variables without
paying attention to the nature of these variables, the introduction of stocks rather than flows is
more reasonable and in conformity with the logic standard form of the neoclassical production
function, which has microeconomic foundations.
Furthermore, this issue is emphasized further when we go through the details of the data of such
flows. Indeed, a large part of aggregate government expenditure flows, introduced in the
production function, are generally public employees’ wages and salaries, direct transfers and
subsidies to families and other public and private institutions. Except for some transfers to the
latter (which are generally small compared to the total) that could help in the production process
of these institutions, wages and salaries and transfers to families and price subsidies go directly to
the households’ income and constitute an input to their behavioral consumption function (or utility
211
public debt
function). Thus, it is logically more suitable to consider it in relation to the latter instead of
including it in the production function.104
An interesting issue that is difficult to measure, and which was mainly invoked by what is known
as the positive approach of public debt in the domain of the political economy, is the crucial role
of institutions in economic growth that uses debt and/or deficit as an instrument variable to curve
future governments’ (opposing) decisions and economic agents (Alesina and Tabellini, 1990).
Considering institutions as a sort of game theory between succeeding governments, or between
governments and their citizens in a normative modelling approach, is difficult. However, we could
also take the role of institutions as endogenous rather than exogenous. One way is to augment the
production function by a parameter of the government constraint representing the quality of the
institutions. However, the role of institutions could also be considered by encompassing a formula
or a parameter describing the inherent discretionary policies of the government, as in Maebayashi
et al. (2017).
In the following section, we present the model.
3.3. The model
3.3.1. The productive sector
We use the production function to describe the relationship between accumulated public capital
and real GDP. The production function can take different specifications, such as the constant
elasticity of substitution (CES) or a trans-log production function, which under some specific
restrictions can be reduced to the Cobb-Douglas production function. The latter is a special case
of the former; in other words, the former are more flexible forms. The CES, for example, does not
require the assumptions of perfect competition and profit maximization.105
We consider the Cobb-Douglas production function, which is well grounded in economic theory
and mainly used in practice, to be easy to estimate, and it has good empirical properties (Razzak
and Bentour, 2013). The equation106 is as follows:
𝛽
𝑌𝑡 = 𝐴. 𝐾𝑝𝑡 𝛼 𝐾𝑔𝑡 (7)
104
Government consumption posts and public investments are defined by the system of national accounts (SNA, 1993
and SNA, 2008) and the classification of expenditure by functions of the government, are categorized by the OECD
classification (COFOG).
105
“Kmenta (1967) shows that estimating these flexible forms is not really difficult, except that they require [a] large
number of observations because there have more parameters to estimate than in the Cobb-Douglas. Furthermore, an
omitted variable problem may also be present. The omitted variable problem results in biased and inconsistent OLS
parameter estimates” (Razzak and Bentour, 2013).
106
The time in all our equations is labelled by the subscripted lowercase letter “𝑡” and could be omitted in case no
confusion arises.
212
public debt
where 𝑌𝑡 , 𝐾𝑝𝑡 and 𝐾𝑔𝑡 are, respectively, the per capita of output, and the per capita of private and
public capital. 𝐴 is a constant designating technical progress, and 𝛼 and 𝛽 are shares in the national
income of, respectively, private and public capital stock (also corresponding to elasticities). Private
and public capital are evolving according to the following inventory equations:
𝐾𝑝,𝑡 = (1 − 𝛿𝑝,𝑡−1 )𝐾𝑝,𝑡−1 + 𝐼𝑝,𝑡 (8)
𝐾𝑔,𝑡 = (1 − 𝛿𝑔,𝑡−1 )𝐾𝑔,𝑡−1 + 𝐺𝑡 (9)
where 𝐼𝑝,𝑡 and 𝐺𝑡 are, respectively, the private and public flows of investments and 𝛿𝑝,𝑡−1 and
𝛿𝑔,𝑡−1 are, respectively, the depreciation rate of private and public capital. To simplify, we assume
the same constant depreciation rate for the private and public sectors (𝛿𝑝,𝑡−1 = 𝛿𝑔,𝑡−1 = 𝛿)
Defining 𝑟𝑔 as the productivity of public capital, the first-order condition derivation yields for
public capital:
𝜕𝑌 𝑌 𝑌
𝑟𝑔,𝑡 = = =𝛽 (10)
𝜕𝐾𝑔 ∆𝐾𝑔 𝐾𝑔
𝑌
For a production function with constant returns to scale, 𝛽 = 1 − , we have: 𝑟𝑔 = (1 − ) .
𝐾𝑔
Equation (10) is equal to unity in the optimal steady state path (𝑟𝑔 = 1), which directly yields a
constant optimal government size, as in Barro (1990) and Corsetti and Roubini (1996), when
considering the flows of government services and not public capital. In our case, we prefer to study
the economies considering that not all are in the steady state. This leads to a government
expenditure size that depends on the return to public capital and its elasticity, which may differ
between countries as a result of the differences in public governance and public expenditure
productivities.
From Equation (9) we have:
𝐾𝑔,𝑡 −𝛿𝐾𝑔,𝑡−1 𝐺𝑡 −𝛿𝐾𝑔,𝑡−1 𝑌𝑡−1 𝐺𝑡 𝑌𝑡

= + =[ ] + (11)
∆𝑌𝑡 ∆𝑌𝑡 ∆𝑌𝑡 𝑌𝑡−1 ∆𝑌𝑡 𝑌𝑡 ∆𝑌𝑡
𝐾𝑔,𝑡 1 −𝛿𝐾𝑔,𝑡−1 −𝛿𝛽 ∆𝑌𝑡

From Equation (10) we have the term = and [ ]= . Putting 𝛾 = (which
∆𝑌𝑡 𝑟𝑔,𝑡 𝑌𝑡−1 𝑟𝑔,𝑡−1 𝑌𝑡−1
1+𝛾 ∆𝑌𝑡 𝐺𝑡
yields, = ) as the nominal economic growth rate, and = 𝑔𝑡 = 𝑔̃𝑡 the “potential”107
𝛾 𝑌𝑡 𝑌𝑡
government productive expenditure share to GDP, we obtain:
107
We used the name “potential”, as it is derived from a first-order condition.
213
public debt
1 −𝛿𝛽 1 1+𝛾
= + 𝑔̃𝑡 (12)
𝑟𝑔,𝑡 𝛾 𝑟𝑔,𝑡−1 𝛾
Or equivalently:
𝛾 1 𝛿𝛽 1
𝑔̃𝑡 = [ ] +[ ] (13)
1+𝛾 𝑟𝑔,𝑡 1+𝛾 𝑟𝑔,𝑡−1
Neglecting the depreciation rate yields:
𝛾 1
𝑔̃𝑡 = [ ] (14)
1+𝛾 𝑟𝑔,𝑡
Equation (14) implies that the potential productive government investment depends on the
economic growth rate, as increasing the function of growth as the derivation according to the
growth rate is positive, and inversely on public capital expenditure productivity. Hence, the higher
the growth rate (higher potential of GDP also), the higher the potential public investment.
Furthermore, the higher the productivity (the return 𝑟𝑔,𝑡 ) of public capital, the lower the potential
public investment.
However, once the depreciation rate has been accounted for, the effect of economic growth on the
size of the government depends on the current and past productivity of public capital, as well as
the elasticity and the depreciation rate parameters. The derivation of the productive share on the
𝜕𝑔 1 1 𝛿𝛽 𝑟𝑔,𝑡
growth rate from Equation (13) yields: = [1 − ], and the algebraic sign of this
𝜕𝛾 𝑟𝑔,𝑡 (1+𝛾)2 𝑟𝑔,𝑡−1
𝛿𝛽 𝑟𝑔,𝑡
quantity depends on the sign of the term [1 − ].
𝑟𝑔,𝑡−1
Furthermore, the optimal government expenditure is an endogenous parameter here and not a
constant one, as raised by Barro (1990) or Corsetti and Roubini (1996) (named the size of the
government for these authors). The assumption made by these authors is that the marginal product
of public capital is equal to one in the optimum, leading to a “constant government size”. This is
also due to their consideration of government flows instead of public capital, which ensures this
relationship.
Proposition: In an endogenous growth framework with the Cobb-Douglas production

function encompassing public capital instead of government flows, the potential productive
government investment size is an endogenous parameter of economic growth and public
capital productivity.
214
public debt
3.3.2. The human capital sector
There is a thorough body of literature on technology diffusion where human capital is an essential
element. The theoretical approaches are grounded in Nelson and Phelps (1966), Grossman and
Helpman (1991) and Barro and Sala-i-Martin (1997), while the empirical literature widely cites,
as examples, the works of Benhabib and Spiegel (1994) and Borensztein et al. (1998). Human
capital can either be an additional factor of production (Mankiw et al., 1992) or a factor influencing
technical progress in the production function (Razzak and Bentour, 2013). Either way, it will
appear as an additional repressor. Therefore, we consider having a measure of the stock of human
capital as an additional regressor in the production function. The production function takes the
form of Corsetti and Roubini (1996), except that here we consider the stock of public capital
instead of flows of government services:
𝛼(1−𝜀)
𝑌 = 𝐴. 𝐾𝑝𝑡 𝛼𝜀 𝐻𝑡 1−𝛼 𝐾𝑔𝑡 (15)
where Ht is human capital and the other variables are as previously defined.
3.3.3. The quality of public institutions
North (1990) defines institutions as “the rules of the game in a society or, more formally, [they]
are the humanly devised constraints that shape human interaction”. This characteristic of devised
human constraints emphasizes the role of the endogenized character of institutions compared to
external/exogenous factors outside human control (geography, for example). Their effects (shape
human interactions) are mainly to shape the behaviour of humankind, thus directly impacting
economic agents’ incentives to invest and consumer choices. Following this effect on the
incentives of economic agents, which affects the economic output of their actions, many authors
have raised and emphasized the role of institutions in long-term economic growth and economic
development (Barro, 1996; Acemoglu et al., 2001; Rodrik et al., 2004; Acemoglu and Robinson,
2010). Such a role is embedded (and hence endogenous) in different forms of dynamic interaction
between economic, organizational, political and social factors (Aoki, 2007). The fiscal policy
sector, in general, and the public debt/deficit management issues, in particular, are very important
domains of public policy that are heavily determined and impacted by the institutional framework
of the country.
The institutional framework plays its role in two dimensions. The first is related to the quantitative
effects of government fiscal variables, which is summarized as the dimension size via the amount
of expenditure and investments allocated by the government. The second is related to the
qualitative aspects of government actions, which could be summarized by the quality of
institutions encompassing the effects of many areas of government reform, programmes and
actions (economic freedom, justice, rule of law, order and security, enforcing contracts, protecting
investors, property rights, etc.). The dimension size, proxied by the stock or flows of government
215
public debt
expenditure, has been embedded in models of endogenous theory, as described in Section 2.2
(Barro, 1990; Futagami et al., 1993, 2008; Corsetti and Roubini, 1996). However, the second
aspect of the institutions relates to their quality in its broad sense and is not easy to observe,
although this plays a crucial role in shaping the effects of the institutional framework on fiscal
policy input (government expenditure, government debt, deficits, etc.) and also output variables
(economic growth, society welfare, income distribution, etc.).
Modelling the quality of public institutions is a challenging and complex task for economists. So,
while the economic models and the endogenous growth theory struggle to clearly include the
institutions’ role in the production process, the task is hardly emphasized, as the political side of
the institutions plays an important role that is difficult to consider. The political character of the
institutions implies removing the frontiers between the two disciplines: economic science and
political science. Nevertheless, some economists have attempted to include empirically the role of
institutions considering the data produced on governance indicators from international institutions,
such as economic and political freedom of the heritage foundation, worldwide governance
indicators of the World Bank, country risk profile of the international risk country group (IRCG),
business indicators of the World Bank, and so on, with fuzzy results.108 Instead, some economists
have calibrated the role of institutions to some reduced parameters.
In our case, we also limit modelling the role of institutions by considering the effect of the
institutions in the production process, represented by a parameter 𝜃 ≥ 1 representing the quality
of institutions. For 𝜃 = 1, the quality of institutions has no effect on the production function, which
turns out to be the same as in the previous section. The production function then becomes:
𝑌𝑡 = 𝐴. 𝐾𝑝𝑡 𝛼𝜃𝜀 𝐻𝑡 (𝜃−1) 𝐾𝑔𝑡

𝛼(1−𝜃𝜀) (1−𝛼𝜃)
𝐿𝑡 (16)
with 𝐿𝑡 as the quantity of labour and the other variables as previously defined.
We notice that this production function exhibits constant returns to scale109 in all its inputs as:
(𝛼𝜃𝜀 + (𝜃 − 1) + 𝛼(1 − 𝜃𝜀) + (1 − 𝛼𝜃) = 1). For 𝜃 = 1 and 𝜀 = 0, the production function is
a function of only public capital and labour.
108
As a result of the non-convincing measures and fuzzy results of considering such measures representing the
qualitative dimension of the institutions, we limit our next empirical application (Section 4) to the dimension size
represented by the public capital stock, for which we add human capital, as discussed in Section 3.3.4. The quality of
human capital is also questioned, as the measures considered are mainly based on the educational attainments
represented generally by the average years of schooling for adults. In order to compare countries based on this
measure, the main assumption is that: one year acquired in a society is the same across all other societies, neglecting
by this the differences between educational systems between countries. Nevertheless, we consider that those
differences are reduced, as the sample of countries enjoys a nearly equivalent level of development.
109
The condition of constant returns to scale is crucial for an endogenous growth model.
216
public debt
Normalizing by the labour quantity 𝐿 and labelling using lowercase, Equation (21) yields:
𝑦𝑡 = 𝐴. 𝑘𝑝𝑡 𝛼𝜀𝜃 ℎ𝑡 (𝜃−1) 𝑘𝑔𝑡

𝛼(1−𝜀𝜃)
(17)
The lowercase variables describe the per capita of the respective higher-case quantities, as defined
previously. While the original production exhibits constant returns to scale, the normalized
production function’s returns to scale depend on the quantity 𝛼𝜃 as the sum of the corresponding
elasticities:
- 𝛼𝜃 = 1: constant returns to scale. The higher the quality of institutions, the lower the input
share 𝛼 needed for producing
- 𝛼𝜃 > 1: increasing return to scale
- 𝛼𝜃 < 1: decreasing return to scale
We define 𝜌𝑥 as the rate of return of the input 𝑥 to the output 𝑦 as the marginal rate of that input,
𝜕𝑦
so as: 𝜌𝑥 = . Accordingly, for public capital we have:
𝜕𝑥
𝜕𝑦 𝜕𝑌 𝑌
𝜌𝑘𝑔 = = = 𝛼(1 − 𝜀𝜃) (18)
𝜕𝑘𝑔 𝜕𝐾𝑔 𝐾𝑔
Similarly:
𝑌
𝜌ℎ = 𝛼(𝜃 − 1) (19)
𝐻
And:
𝑌
𝜌𝑘𝑝 = 𝛼𝜀𝜃 (20)
𝑘𝑝
From the inventory public capital accumulation, the equation assuming the inventory stock is
calculated in the beginning of the period and not in the last part of the period:110
𝐾𝑔,𝑡 = (1 − 𝛿𝑔,𝑡−1 )𝐾𝑔,𝑡−1 + 𝐺𝑡 (21)
where 𝛿𝑔,𝑡−1 is the depreciation rate of public capital accumulation 𝐾𝑔,𝑡−1 .
110
The other alternative equation is to assume that the inventory stock is assessed at the end of the period (31 December
instead of 1 January), and thus the equation changes slightly to: 𝐾𝑔,𝑡+1 = (1 − 𝛿𝑔,𝑡 )𝐾𝑔,𝑡 + 𝐺𝑡+1 .
217
public debt
Dividing the equation of public capital by the first lagged output and assuming as negligible the
public capital depreciation rate, we obtain:
∆𝐾𝑔,𝑡 𝛿𝑔,𝑡−1 𝐺𝑡 𝐺𝑡
=− + ≅ =𝑔 (22)
𝑌𝑡 𝑌𝑡 𝑌𝑡 𝑌𝑡
where 𝑔 is government expenditure, representing the dimension size of the institutions (the
𝛼(1−𝜀𝜃)
state/government size). From Equation (23), we have 𝐾𝑔,𝑡 = 𝑌𝑡 , which yields:
𝜌𝑘𝑔
𝛼(1−𝜀𝜃)
𝐾𝑔,𝑡 = 𝜌𝑘𝑔
𝑌𝑡 (23)
∆𝐾𝑔,𝑡 𝛼(1−𝜀𝜃) 𝑌𝑡 𝐺

Considering (28), Equation (27) is now written as = . = = 𝑔.
𝑌𝑡 𝜌𝑘𝑔 𝑌𝑡 𝑌
𝑌𝑡 𝑌𝑡 𝑌𝑡−1 𝛾

As = = , where 𝛾 is the growth rate of the economy, the dimension size yields:
𝑌𝑡 𝑌𝑡−1 𝑌𝑡 1+𝛾
𝛼(1−𝜀𝜃) 𝛾
𝑔= ( ) (24)
𝜌𝑘𝑔 1+𝛾
The dimension size of the state is an endogenous function of quantitative quantities related to the
growth of the economy and the productivity of public capital, and to a qualitative variable
representing the efficiency of human capital and the quality of government institutions. In detail:
- The dimension size of the government is a function of the economic growth rate via the
𝛾 𝛼(1−𝜀𝜃)
term : the impact of the latter depends on the sign of quantity . In particular, the
1+𝛾 𝜌𝑘𝑔
size is a growing function of the economic growth rate if 𝜀𝜃 < 1 and the efficiency of the
public capital is positive (the derivative is positive in this case).
- The dimension size is inversely dependent on the efficiency of public capital: the higher
the productivity of public expenditure, the lower the size of the government. Alternatively,
lower efficiency of public capital leads to a higher government size.
- The dimension size of the government is negatively related to the human capital efficiency
and the quality of institutions. A high efficiency of human capital, as well as good quality
of institutions, should lead to a small size of government.
Alternative equations for the dimension size:
𝜕𝑦 𝜕𝑌 𝑌 𝑌
𝜌𝑘𝑔 = = = = 𝛼(1 − 𝜀𝜃) (25)
𝜕𝑘𝑔 𝜕𝐾𝑔 𝐾𝑔 𝐾𝑔
218
public debt
𝑌 𝛼(1−𝜀𝜃) 𝛾
𝜌𝑘𝑔 = 𝛼(1 − 𝜀𝜃) and 𝑔 = ( ) yields:
𝐾𝑔 𝜌𝑘𝑔 1+𝛾
𝛾 𝐾𝑔
𝑔=( ) (26)
1+𝛾 𝑌
Or:
𝛼(1−𝜀𝜃) 𝛾 𝛾 𝐾𝑔
𝑔= 𝑌 ( ) = 𝛼(1 − 𝜀𝜃) ( ) (27)
1+𝛾 1+𝛾 𝑌
𝐾 𝑔
3.3.4. The government sector
The budget constraint is defined as:
𝐵𝑡 = (1 + 𝑖𝑡 )𝐵𝑡−1 + 𝐺𝐼𝑡 + 𝑆𝐺𝑡 − 𝑇𝑡 (28)
where 𝐵𝑡 is government bonds, 𝑖𝑡 is the nominal interest rate, and, 𝐺𝐼𝑡 , 𝑆𝐺𝑡 and 𝑇𝑡 are, respectively,
government flows of expenditure on productive capital, social and security government spending,
and tax revenue. The quantity 𝐺𝐼𝑡 + 𝑆𝐺𝑡 − 𝑇𝑡 is the primary balance (primary surplus or deficit,
depending on its sign). Some authors (Obstfeld, 1997; Haslag and Young, 1998; Terra, 2015)111
add a term of seigniorage revenue, which we assume to be non-existent or negligible112 in our case.
We assume that state dependency on seigniorage revenue is likely to happen when the government
fails to fulfil its financing needs for expenditure through conventional taxes or bond sales.
Furthermore, the option of seigniorage (fiat money) may lead to hyperinflation, which may be self-
defeating. These conditions are of less importance in advanced economies (which are the focus of
the case study of this chapter).
Dividing both sides of this equation by the nominal output production 𝑝𝑡 𝑌𝑡 with 𝑝𝑡 the price
deflator of real output 𝑌𝑡 , and manipulating the left-hand-side equation term (LHS) to raise the real
growth rate and inflation terms yields:
1+𝑖𝑡
𝑏𝑡 = 𝑏 + 𝑔𝑖 + 𝑠𝑔𝑡 − 𝜏𝑡 (29)
1+𝛾𝑡 𝑡−1
where 𝑏𝑡 , 𝑔𝑖𝑡 , 𝑠𝑔𝑡 and 𝜏𝑡 are the ratios to the output of public debt, government productive
investment, government social and security spending and tax revenue, respectively, while 𝛾𝑡 is the
111
“Seigniorage is government revenue resultant from the emission and maintenance of the stock of currency in
circulation” (Terra, 2015). “The government resources constraint establishes that the change in government debt
should be equal to the interest payment on the existing debt added to the resources necessary to supply public goods,
minus the seigniorage revenue” (Terra, 2015).
112
Generally, money-creation revenue accounts for less than 2% of GDP (Haslag and Young, 1998).
219
public debt
current nominal economic growth. This equation describes the law of motion in an inter-temporal
balanced government budget.
3.3.5. The productive potential government capital and potential government debt
In this section, we link the results of the productive sector (Section 3.3.1) to the government sector
(Section 3.3.2). Equation (14) leads to an endogenous potential government investment (which
will be simulated in the first step).
Equations (14) and (17) could be jointly used to derive a potential endogenous (limit) of public
debt in relation to economic and monetary conditions (economic growth, public capital
productivity, interest rate). For example, Maebayashi et al. (2017) consider that the government
could have a targeted potential level of debt in the long term (𝑏𝑡−1 = 𝑏𝑡 = 𝑏̃𝑡 ). We assume that
this potential debt limit will be designated only to financing the potential government investment
needs (−𝑔̃𝑡 ), while we can assume that social and security spending are financed by tax revenue;
this means that 𝑠𝑔𝑡 − 𝜏𝑡 = 0 and 𝑔𝑖𝑡 = −𝑔̃𝑡 (𝑔𝑖𝑡 is negative, as it is a financing need to be filled
by the new issuance of debt; otherwise, there is a positive primary surplus and accumulating new
debt is not necessary). Considering this case, Equation (17) yields:
(1+𝛾𝑡 ) (1+𝛾𝑡 )
𝑏̃𝑡 = (−𝑔̃𝑡 ) = 𝑔̃𝑡 (30)
(𝛾𝑡 −𝑖𝑡 ) (𝑖𝑡 −𝛾𝑡 )
Replacing 𝑔̃𝑡 , as described in (14), leads to:
(1+𝛾𝑡 ) 𝛾 1 𝛿𝛽 1
𝑏̃𝑡 = ([ 𝑡 ] +[ ] ) (31)
(𝑖𝑡 −𝛾 ) 1+𝛾 𝑟
𝑡 𝑡 1+𝛾 𝑟
𝑔,𝑡 𝑡 𝑔,𝑡−1
Equation (30) (or 31) delivers an endogenous potential (optimal/maximal/limit) debt that a country
could target in the long term based on its economic fundamentals related to its potential productive
expenditure, economic growth rate and interest rate. As we are considering the long term, the
interest rate considered is the long-term interest rate. Note that the denominator is the difference
between interest rates and economic growth. The higher this denominator, which means a higher
interest rate than economic growth (unfavourable economic conditions), the lower the potential
debt that a country could issue. This denominator constitutes an inertia bringing down potential
debt in bad economic conditions, where interest rate spreads are higher. Inversely, the higher the
growth rate, the higher the potential debt (the derivation of the potential debt to growth rate is
strictly positive).
220
public debt
4. Empirical evidence
Considering all the previous elements and discussions, this section augments the productive sector
by two additional endogenized inputs representing the government and human capital sectors. The
public sector is included through its dimension size of only productive investments (gross capital
formation flows and stocks).
The set of countries studied is composed of 20 advanced countries (the same sample as considered
in the first chapter), most of which are parts of the European Union, of which 12 adhere to the
European Monetary Union, while the rest of the countries generally adopt a floating exchange rate
system. These countries are, respectively: Australia, Austria, Belgium, Canada, Denmark, France,
Germany, Greece, Ireland, Italy, Japan, The Netherlands, Norway, New Zealand, Portugal, Spain,
Sweden, Switzerland, the United Kingdom and the United States of America.
Data for the accumulated private and public capital stocks is drawn from the IMF database,113 for
which the data is made available for 170 countries. The data range covers the period 1965–2015,
and the database was last updated in January 2017. Either public or private stocks are constructed
according to the inventory method following the equation:
𝐾𝑥,𝑖𝑡+1 = (1 − 𝛿𝑥,𝑖𝑡 ) 𝐾𝑥,𝑖𝑡 + (1 − 𝛿𝑥,𝑖𝑡 /2) 𝐼𝑥,𝑖𝑡 (32)
where 𝐾𝑥,𝑖𝑡 is the stock of capital (𝑥 = {𝑝𝑢𝑏𝑙𝑖𝑐; 𝑝𝑟𝑖𝑣𝑎𝑡𝑒}), 𝛿𝑥,𝑖𝑡 is the corresponding depreciation
rate and 𝐼𝑥,𝑖𝑡 is the investment flows (gross capital formation for the sector x).
To construct the time series of the stocks using the perpetual inventory stock equation (32),
assumptions made about the initial values of stock and data on investment flows are described
fully in the IMF (2015).114 Tables A.2 and A.3 in Appendix A give a summary of the data coverage,
sources, description of the variables and the corresponding descriptive statistics.
Other variables such as labour and population are drawn from World Development Indicators of
the World Bank database,115 while the human capital is from the Barro and Lee website, displaying
educational attainment and average years of schooling for a large set of countries and a wide range
of time periods.116 The human capital data is produced by an interval of five years from 1950 to
2010 (i.e. 1950; 1955; 1960; …; 2005 and 2010), which we extrapolated by simply moving
averages to fill the gaps between the extremities of the five-year intervals. The whole methodology
113
Accessible database on excel file is through the hyperlink in the PDF document “Investment and Capital Stock
Dataset”: https://www.imf.org/external/np/fad/publicinvestment/pdf/csupdate_jan17.pdf.
114
http://www.imf.org/external/np/fad/publicinvestment/data/info122216.pdf.
115
https://databank.worldbank.org/home.aspx
116
http://www.barrolee.com/
221
public debt
and the sources used to produce such data are fully explained in Barro (1993) and revised in Barro
(2013).
4.1. Estimation of the Cobb-Douglas production function
The Cobb-Douglas production function encompassing human capital is modified and log-
linearized to take the form of the following specification:
𝑙𝑛𝑌𝑡 = . 𝑙𝑛𝐾𝑔,𝑡 + 𝛽. 𝑙𝑛𝐾𝑝,𝑡 + (1 − 𝛽). ln(ℎ𝑡 . 𝐿𝑡 ) + 𝐶 + 𝜖𝑡 (33)
where 𝑌𝑡 is the output, 𝐾𝑔,𝑡 represents the stock of public capital and 𝐾𝑝,𝑡 the private capital stock,
and 𝐿𝑡 is the labour, adjusted for human capital by average years of schooling ℎ𝑡 . The private
capital stock and the labour variables ensure constant returns to scale. C is a constant term and 𝜖𝑡
is the error terms assumed to behave independently and identically distributed.
We produce estimations of the previous specification using public capital stock in Table 1, over
two periods of data history: the period 1960–2015; and the sub-period 1960–2007, which excludes
the economic financial crisis and its subsequent impacts. Table 1 shows the estimated elasticities
for all the variables of Equation (33), along with their statistical significance. Overall, the
estimations have a high significant coefficient for all countries, with high adjusted R-squared going
beyond 95% for all countries and a high Fisher (F-test) global significance.117
The country results show that 15 out of 20 countries have significant public capital stock elastities
for the two periods, with one negative significant elasticty for Greece in the first period and three
negative elasticities for Austria, Greece and Japan in the second period. Over the two periods,
Australia, Denmark and New Zealand have non-significant public capital elasticties. The average
of accepted postive elasticties is around 0.32 and 0.36 for the two periods, respectively, and this
average drops, when taking into account negative accepted elastitcities to 0.30 and 0.25,
respectively, over the two periods. Higher elastictities (over 0.5) are recorded for the two periods,
respectively, by Ireland (1.04 and 1.14), followed by Germany (0.59 and 0.51) and Sweden (0.57
and 0.51). Menawhile, many countries have elasticities ranging between 0.15 (the United States)
and 0.25 (Canada and Belgium, for example), except France, which showed a lower accepted
eslasticy value of around 0.07. Figure 2 summarizes the significant elasticities for government
117
The Durbin-Watson statistics are also low for all estimations, which indicates the presumption of a cointegration
relationship in the data. However, in designing the rest of the simulations for the public debt potential, our model is
intended to consider only the long-term effects (classical effects), while the short-term effects (Keynesian effects) are
examined in Chapter 2, which links the effects of the public deficit financed by government bonds to fiscal policy
effects via the multiplier effect (the Keynesian multiplier). Therefore, cointegration and error correction model
formulations are not considered here, where we limit ourselves to long-term estimations. Besides, considering such
formulations will completely modify the form of the Cobb-Douglas production function, which is theory grounded
rather than ad hoc compilations, and complicate obtaining the overall elasticity effects.
222
public debt
capital stocks, for which we assume an average of 0.30 as elasticity for the few non-significant
elasticities (Australia, Austria, Denmark and New Zealand).
223
public debt
Table 1. Specification estimation results using government investment stocks
Specification: 𝑙𝑛𝑌𝑡 = . 𝑙𝑛𝐾𝑔,𝑡 + 𝛽. 𝑙𝑛𝐾𝑝,𝑡 + (1 − 𝛽). ln(ℎ𝑡 . 𝐿𝑡 ) + 𝐶 + 𝜖𝑡
Sample 1960-2015 Sample 1960-2007
Public capital Kg Private capital Kp Intercept Public capital Kg Private capital Kp Intercept
α  C α  C
Coefficient 0.045 0.884*** -1.944* 0.007 1.090*** 0.757
Australia
Probability 0.167 0.000 0.085 0.821 0.000 0.530
Coefficient 0.207 1.239*** 0.763 -0.184** 1.966*** 11.002***
Austria
Probability 0.128 0.000 0.848 0.028 0.000 0.000
Coefficient 0.222*** 0.995*** -2.097* 0.208*** 0.963*** -2.415**
Belgium
Probability 0.000 0.000 0.093 0.000 0.000 0.033
Coefficient 0.243*** 0.171 -11.808*** 0.338*** -0.020 -14.650***
Canada
Probability 0.000 0.113 0.000 0.000 0.860 0.000
Coefficient 0.166*** 0.396*** -8.458*** 0.145*** 0.485** -7.366***
Switzerland
Probability 0.000 0.000 0.000 0.003 0.022 0.005
Coefficient -0.012 0.413*** -7.263*** -0.029 0.478*** -6.413***
Denmark
Probability 0.732 0.000 0.000 0.400 0.000 0.000
Coefficient 0.073*** 0.621*** -5.514*** 0.065*** 0.646*** -5.173***
France
Probability 0.000 0.000 0.000 0.000 0.000 0.000
United Coefficient 0.435*** 0.390*** -10.758*** 0.077 0.728*** -4.330*
Kingdom Probability 0.000 0.000 0.000 0.562 0.000 0.082
Coefficient 0.587*** 0.351*** -12.470*** 0.515*** 0.498*** -10.256***
Germany
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Coefficient -0.251*** 2.186*** 15.217*** -0.207*** 2.157*** 14.679***
Greece
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Coefficient 1.038*** -1.394*** -32.801*** 1.144*** -1.627*** -35.925***
Ireland
Probability 0.000 0.005 0.000 0.000 0.002 0.000
Coefficient 0.408*** 1.224*** -1.167 0.428*** 1.135*** -2.299***
Italy
Probability 0.000 0.000 0.104 0.000 0.000 0.007
Coefficient -0.215 1.075*** 2.146 -0.269* 1.138*** 3.336
Japan
Probability 0.135 0.000 0.562 0.076 0.000 0.387
Coefficient 0.153*** 1.861*** 8.420** 0.107** 1.842*** 8.466**
Netherlands
Probability 0.003 0.000 0.012 0.040 0.000 0.015
Coefficient 0.405*** 0.243*** -11.014*** 0.448*** 0.143*** -12.318***
Norway
Probability 0.000 0.000 0.000 0.000 0.001 0.000
New Coefficient -0.009 1.552*** 6.691*** -0.043 1.865*** 10.733***
Zealand Probability 0.814 0.000 0.001 0.349 0.000 0.001
Coefficient 0.185*** 0.637*** -5.795*** 0.241*** 0.757*** -4.674***
Portugal
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Coefficient 0.195*** -0.500 -19.316*** 0.185*** -0.075 -14.301***
Spain
Probability 0.000 0.158 0.000 0.000 0.849 0.005
Coefficient 0.568*** -0.625*** -22.670*** 0.508*** -0.637*** -22.500***
Sweden
Probability 0.000 0.008 0.000 0.000 0.001 0.000
United Coefficient 0.155* 0.735*** -4.866* 0.083 0.863*** -2.691
States Probability 0.085 0.000 0.061 0.464 0.000 0.391
Notes: Coefficient and probability are, respectively, the estimated elasticities of the Cobb-Douglas production function
(since the equation is log-linearized) and their corresponding p-values. *, ** and *** means significance at 10%, 5%
and 1%, respectively.
224
public debt
Figure 2. Government capital stock elasticities
1.20 1.0
1.00
0.80 0.6 0.6
0.60 0.4 0.4 0.4
0.3 0.3 0.2 0.2 0.3 0.3
0.40 0.2 0.2 0.2 0.2 0.2
0.20 0.1
0.00
-0.20
-0.40 -0.3 -0.2
France
Australia
Denmark
Ireland
Japan
United Kingdom
United States
Austria
Belgium
Canada
Germany
Italy
Portugal
Spain
Sweden
Greece
Netherlands
Switzerland
New Zealand
Norway
Source: Author’s own calculations
For estimations of Equation (33), using government capital flows instead of government capital
stock, Table 2 shows that 16 out of 20 countries have significant government investment elastities
for the period 1960–2015, with one negative elasticty recorded for Greece, and 17 countries for
the period of 1960–2007, with two negative significant elasticties shown for Greece and Germany
in this period. The average of accepted positive elasticities is around 0.24 and 0.26 for the two
respective periods, while it is reduced to around 0.21 and 0.19 over the two periods, respectively,
when accounting for negative accepted elasticties (Greece for the two periods, and Germany for
the period 1960–2007).
By country, higher elasticities of between 0.30 and 0.50 are recorded for Sweden, Norway, Ireland,
The Netherlands and Italy. Belgium and New Zealand’s elasticities are, respectively, negative and
non-significant and positive non-accepted for the two periods of estimations. The values of
positive elasticities range between:
- Relatively low values of around 5% to 15% recorded for four countries, namely, Australia,
Austria, Canada, Denmark, France, Germany and Japan;
- Medium values of around 15% to 25% obtained for countries such as Switzerland, the
United Kingdom, Portugal, Spain and the United States;
- Relatively high values of public capital elasticities shown for Ireland, Italy, The
Netherlands, Norway and Sweden.
225
public debt
Some countries tend to crowd out private capital with public capital where the elasticity of private
capital is either low positive or negative accepted, as with Spain, or positive and rejected, as with
Ireland (1965–2007).
Table 2. Specification estimation results using government investment flows

Specification: 𝑙𝑛𝑌𝑡 = . 𝑙𝑛𝐼𝑔,𝑡 + 𝛽. 𝑙𝑛𝐾𝑝,𝑡 + (1 − 𝛽). ln(ℎ𝑡 . 𝐿𝑡 ) + 𝐶 + 𝜖𝑡
1960-2015 1960-2007
Ig Kp Intercept Ig Kp Intercept
α  C α  C
Australia Coefficient 0.115*** 0.667*** -4.604*** 0.119*** 0.778*** -3.258**
Probability 0.002 0.000 0.001 0.001 0.000 0.011
Coefficient 0.143** 1.543*** 4.953*** 0.004 1.575*** 5.615***
Austria
Probability 0.013 0.000 0.000 0.918 0.000 0.000
Coefficient -0.009 1.978*** 10.167*** -0.030 1.931*** 9.671***
Belgium
Probability 0.894 0.000 0.003 0.639 0.000 0.005
Coefficient 0.120*** 0.500*** -6.801*** 0.210*** 0.340*** -8.999***
Canada
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Coefficient 0.236*** 0.396*** -8.165*** 0.181*** 0.567*** -6.149***
Switzerland
Probability 0.000 0.000 0.000 0.003 0.003 0.005
Coefficient 0.019 0.421*** -7.279*** 0.044** 0.528*** -6.080***
Denmark
Probability 0.233 0.000 0.000 0.012 0.000 0.000
Coefficient 0.106*** 0.634*** -5.300*** 0.099*** 0.656*** -5.017***
France
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Coefficient 0.183*** 0.805*** -3.567*** 0.102*** 0.810*** -3.215***
Great Britain
Probability 0.000 0.000 0.000 0.010 0.000 0.000
Coefficient 0.142 0.853*** -3.005 -0.273** 1.401*** 5.042**
Germany
Probability 0.366 0.000 0.210 0.048 0.000 0.025
Coefficient -0.109*** 1.776*** 9.299*** -0.130*** 2.018*** 12.317***
Greece
Probability 0.004 0.000 0.000 0.000 0.000 0.000
Coefficient 0.302*** 0.591*** -5.965** 0.419*** 0.196 -10.743***
Ireland
Probability 0.000 0.004 0.015 0.000 0.387 0.000
Coefficient 0.472*** 0.943*** -3.415*** 0.479*** 1.246*** -0.047
Italy
Probability 0.000 0.000 0.008 0.000 0.000 0.965
Coefficient 0.077** 0.684*** -4.721*** 0.211*** 0.513*** -7.527***
Japan
Probability 0.049 0.000 0.000 0.000 0.000 0.000
Coefficient 0.321*** 1.082*** -0.668 0.292*** 1.136*** 0.045
Netherlands
Probability 0.000 0.000 0.791 0.000 0.000 0.985
Coefficient 0.398*** 0.213*** -10.281*** 0.467*** 0.044 -12.289***
Norway
Probability 0.000 0.006 0.000 0.000 0.516 0.000
Coefficient 0.042 1.363*** 4.256* 0.028 1.562*** 6.754*
New Zealand
Probability 0.335 0.000 0.074 0.552 0.000 0.052
Coefficient 0.180*** 0.624*** -5.421*** 0.212*** 0.744*** -4.128***
Portugal
Probability 0.000 0.000 0.000 0.000 0.000 0.001
Coefficient 0.163*** -0.239 -15.649*** 0.159*** -0.027 -13.159***
Spain
Probability 0.000 0.384 0.000 0.001 0.948 0.010
Coefficient 0.520*** 0.337 -9.695*** 0.427*** 0.383 -8.944***
Sweden
Probability 0.000 0.158 0.001 0.000 0.110 0.003
Coefficient 0.193*** 0.624*** -5.974*** 0.209*** 0.602*** -6.340***
United States
Probability 0.000 0.000 0.000 0.000 0.000 0.000
Notes: Coefficient and probability are, respectively, the estimated elasticities of Cobb-Douglas production function
(since the equation is log-linearized) and their corresponding p-values. *, ** and *** means significance at 10%, 5%
and 1%, respectively.
226
public debt
The overall average of accepted elasticities over the two periods is around 20%. This value is
somewhat below the average value, which we estimated using panel data for the whole sample of
the same countries. The estimated value for the panel group is around 25% and corresponds exactly
to what is found by Barro’s (1993) panel estimations for government investment flows. Focusing
on the United States’ data, the elasticity is around 0.19 and 0.21 over the two periods and is very
robust to data sample variations in our estimations.
A conducted robustness check generally shows that the elasticity of public capital flows for the
United States varies between 17% and 0.23% following sample time variations. Furthermore, Hill
(2008) shows that the growth-maximizing size of the state varies between 9% and 29% for the
United States (large interval), while Scully (1996), with a different specification to ours, reports a
government size of 19% for the same country. For the countries for which government investment
flow elasticities are rejected over the two periods of estimation, we can consider the average of all
elasticities (0.20) as a calibrated elasticity for all these countries in the subsequent calculations.
We can also adopt calibrated parameters based on similarities between countries. For the case of
Japan, however, an estimation over earlier samples gives highly accepted positive but decreasing
elasticities over time. Belgium also has a sensitive elasticity to sample changes. The elasticity
becomes accepted starting from 1980, where we find a positive accepted elasticity of 7.6% over
1980–2015.
Inspecting the elasticities of public and private capital, we notice that higher government capital
elastictities (and also higher government flow elasticities) are accompanied by lower or negative
private capital (flow) elasticities or rejected private capital elasticties. This is particularly the case
for countries such as Ireland, Norway, Spain and Sweden. This might be attributed to the crowding-
out effect of private investment by government investment. Drawing scatter plots for private and
public capital elasticities for the two periods clearly shows this tendancy. Figures 3 and 4 present
a negative relatiship between public and private capital elasticties for the periods 1960–2015 and
1960–2007, respectively, with Ireland and Greece representing two opposite extreme points in
these figures (the red-coloured dot in the two figures indicates the simple average of elasticties of
the sample). The same trend is shown when considering investment flows instead of stocks
(Figures B.1 and B.2 in the appendix).
227
public debt
Figure 3 : Scatter plot elastictities of Public Investment stocks versus Private
capital stocks (Kp, Ig): Sample 1960-2007
1.4
1.2
IRL
1.0
0.8
0.6
SWE GER
NOR ITA
0.4
CAN
CHE
0.2 SPA AVE BEL
FRA USA NDL
0.0 GBR AUS NZL
DNK
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
-0.2 AUT
JPN GRC
-0.4
Trendline equation: y = -0.2846x + 0.3933
R² = 0.65; Sample 1960-2007
Figure 4 : Scatter plot elastictities of Public Investment Stocks versus Private

1.2
IRL
1.0
0.8
SWE 0.6 GER

NOR
GBR ITA
0.4
AVE
CAN BEL AUT
SPA0.2 CHE PRT
USA NLD
FRA AUS
DNK
0.0 NZL
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
-0.2 JPN
GRC
-0.4
Trendline Equation: y = -0.2527x + 0.3873
R² = 0.53; Sample 1960-2015
228
public debt
4.2. Simulations of targeted/potential public debt ratios
Based on the results of Table 1 for capital stock elasticities, we note that there are no big differences
between elasticities for the two periods (1960–2007 and 1960–2015). Therefore, to save space, we
pursue our calculations considering the results of elasticities over the whole period of 1960–2015.
To conduct simulations and calculus, we decided to produce two simulated exercises. The first one
is based on the estimated elasticity of each country, and the second is based on the average of all
countries’ elasticities (equal to 0.3).
However, for the countries for which elasticities are not accepted across the simulated period, we
decided to consider only the average of elasticities as the benchmark for their elasticities. These
countries are Australia, Austria, Denmark and New Zealand. Greece, over both periods, and Japan,
over the period 1960–2007, although they have negative accepted elasticities, are analysed
assuming these are the true elasticities. Hence, the simulation is done considering their negative
elasticities118 and the average of the elasticities (0.30). For the rest of the countries, we simulate
the rest of our calculations considering the countries’ elasticities and the average of countries’
elasticities (0.3).
Simulation steps
- First, we use the previous government capital values’ elasticities (Table 1), estimated for
the period 1965–2015, to generate the “return” of the government productive investment
𝜕𝑌 𝑌
using the formulae 𝑟𝑔,𝑡 = =𝛼 (equation (10)).
𝜕𝐾𝑔 𝐾𝑔
- Second, we generate the public potential investment based on Equation (13)

𝛾
𝑡 1 𝛿𝛽 1
(𝑔̃𝑡 = [1+𝛾 ] +[ ] ).
𝑟
𝑡 1+𝛾 𝑟
𝑔,𝑡 𝑡 𝑔,𝑡−1
- To avoid high fluctuations generated by data, we use the Hodrick-Prescott filter119 to

smooth the generated potential public investment, before introducing it in the next step.
118
Having negative elasticities does not mean that this should lead to negative simulated investment. These two
countries have positive simulated investment (see Table 2 for elasticities and Figure B.3 for simulated public
investment flows).
119
To obtain a smooth estimate of the long-term trend component of a series, the Hodrick-Prescott filter (HP filter
hereafter) is a widely used smoothing method among researchers. The method first appeared in a working paper in the
early 1980s, was applied to analyse the post-war US business cycles and published later in 1997 (Hodrick and Prescott,
1997). A time series 𝑌𝑡 could be decomposed to its long-term trend 𝐺𝑡 (a sum of growth component) and cyclical
component 𝐶𝑡 : 𝑌𝑡 = 𝐺𝑡 + 𝐶𝑡 . The HP filter algorithm works to smooth the original series by estimating its trend
component, while the cyclical component results as the difference between the original series and its trend. The trend
component is the one that minimizes ∑𝑇1(𝐶𝑡 )2 + 𝜆 ∑𝑇1[(𝐺𝑡 − 𝐺𝑡−1 ) − (𝐺𝑡−1 − 𝐺𝑡−2 )]2 , where T is the number of
observations and λ is a positive parameter of smoothing that depends on the frequency of the time series. The higher
the data frequencies, the larger the value of λ, and the larger λ, the higher the penalty of changes in the trend’s growth
rate (represented by the second term of the previous equation) and the smoother the trend component. In practice, λ is
set empirically to be 1,600 for quarterly data, as suggested by Hodrick and Prescott (1997), while for annual data, λ is
set to 100 in many applications, which we also consider in ours.
229
public debt
Focusing on the long-term tendency and avoiding fluctuations seems to be in line with our
focus on the long-term analysis.
(1+𝛾𝑡 )
- Third, we simulate the potential public debt using Equation (30) (𝑏̃𝑡 = 𝑔̃𝑡 ).
(𝑖𝑡 −𝛾𝑡 )
First, generating public capital returns/productivity shows, on average, big differences between
countries due, mainly, to differences in elasticities. The public capital productivity averages are
widely different across countries, ranging from low values of 5% to 15% in Austria, Denmark,
Portugal and Spain, and very high values of 80% recorded in countries with high values of
elasticities, namely, Ireland and Switzerland (Figure 5).
Figure 5. Scatter plot of public capital productivity versus elasticity

0.9
CHE
0.8 IRL
0.7
AUS
0.6
0.5 GRC
NLD
0.4 NOR
BEL
USA ITANZD
0.3 CAN SWE
FRA
GBR
0.2 SPA GER JPN
Trendline equation: y = 1.54x + 0.02
0.1 AUT PRT R² = 0.70
DNK
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Second, we simulate the potential government investment based on the public capital marginal
productivity (derived from the marginal product of capital, Equation (13)). For comparison, we
produce the descriptive statistics for both observed and simulated government productive
investment (gross capital formation) and draw the averages in Figure 6. This shows that, overall,
the simulated variable overcomes the actual one in all countries by about one to three percentage
points, except for Japan and Greece, where the actual one is, on average, slightly above the
simulated one (by 0.4 percentage points). In particular, the simulated debt ratio is double the actual
one in Australia and Ireland. Figures B.3 and B.4 in Appendix B show the tendency for all
countries towards actual and simulated potential government investment, respectively.120 In all
countries, government investments (actual and simulated) tend to decrease over time (Figure B.5).
120
For all figures and tables when it applies, we point to variables in the form of “Y_XXX”, where Y is the variable
presented (simulated) and XXX is the three-character country code. Table A.4 (in the appendix) presents these codes.
230
public debt
Figure 6. Actual and simulated productive public investment flows (gross capital formation) as % of GDP
-1
Simulated Actual Simulated minus Actual
Third, we simulate the results for the debt limit that a country could target, considering the
previously simulated potential government investment and the conditions of economic and
monetary performance, reflected by the average long-term interest rates and economic growth,
respectively. Data on long-term interest rates is extracted from the OECD database.121 We call this
simulated debt “potential debt”. We use the term “potential”, as this is related to “potential”
government investment (potential, as it is derived from the first condition of maximizing output,
that is, marginal productivity of capital). It is, in fact, the optimal (maximal) public debt that a
country can issue to finance its potential government productive investment and is directly related
to public capital productivity (efficiency). This potential public debt is like a mirror (an opposite
picture) of the actual debt. This is due to the differential between interest rates and economic
growth that appears in the denominator of the potential debt with an opposite sign of the actual
accumulated debt formula. This constitutes an inertia lowering potential debt when interest rates
are higher than economic growth, which pushes down potential debt under actual debt, attracting
the attention of policy-makers to the danger of the debt situation. The danger of public debt could
be measured by the distance between simulated and actual debt. The higher this distance, the safer
the debt is.
First, we consider the simulation using the growth and long-term interest rate differential average
over the period 1960–2015. We produce a set of figures that show the tendencies of simulated
121
https://stats.oecd.org/
231
public debt
public debt for each country. Figures 7, 8, 9 and 10 produce such simulated debt results, along
with actual public debt, over the period 1960–2015 for the 20 sample countries.
The first set of graphs (Figure 7) presents six countries where the debt is not threatening, as we
could still target potential debt that is higher than actual debt over history. These countries are
Australia, New Zealand, Denmark, Switzerland, Norway and Sweden. The average potential debt
to GDP is approximately 96.9% for Australia, and 100.5% for New Zealand, with lower values of
around 80% for both countries. The remaining countries have higher potential average of public
debt based on their performances, which are around 230% in Denmark, 250% in Switzerland,
150% in Norway and 170% in Sweden, respectively.
Figure 7. Simulated potential paths of public debt versus the actual path based on potential simulated
government investment and economic and monetary performances for Australia, New Zealand, Denmark,
Norway, Sweden and Switzerland*
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_AUS D30_AUS ACTUAL_NZL D30_NZL
400 500
350
400
300
250
300
200
200
150
100
100
50
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_DNK D30_DNK ACTUAL_CHE D_CHE D30_CHE
200 250
160 200
120 150
80 100
40 50
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_NOR D_NOR D30_NOR ACTUAL_SWE D_SWE D30_SWE
*: countries having only two curves are those for which elasticities aren’t significant, and thus we limit the exercise
of simulated public debt to their elasticities as the average of all significant elasticities (0.3).
232
public debt
Figure 8 shows a group of four non-euro area countries with floating exchange rates and larger-
sized economies, namely, the United States, Japan, the United Kingdom and Canada. The
sustainable path of the first three countries is undermined during the financial crisis. Actual public
debt remains with a long history under targeted public debt until around the period of crisis (2008–
10), although the two paths (simulated and observed) converge over time, showing that these
countries accumulate public debt over time during the prosperity periods of the 1960s and the
Great Moderation Era of the 1990s. For Japan, the formula shows negative explosive potential
debt based on the average long-term interest rate and growth rate over the whole period. However,
for this country, the long-term interest rate is observed over 1989–2015 only. A robustness check
for countries is conducted based on their performances over certain specific periods and not the
whole period. However, simulation is always produced over the whole period to visualize the track
of simulated debt over the entire period.
For the other countries, the United States and Canada, although potential debt has a tendency to
decrease over time to converge towards actual increasing debt, it still does not constitute a threat
to public finance sustainability. However, the United Kingdom presents a different picture
compared to the United States and Canada. Simulated debt is exceeded by actual debt in the year
2008 and stays under actual debt for the period 2008–15, despite an upward shift. The average
potential debt ratios to target to finance potential expenditure is around 195.5% for the United
States, 100.2% for the United Kingdom and 145.5% for Canada.
233
public debt
government investment and economic and monetary performances for the United States, Japan, the United
Kingdom and Canada
300 400
200
250
0
200
-200
150 -400
-600
100
-800
50
-1,000
0 -1,200
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_USA D_USA D30_USA ACTUAL_JPN D_JPN D30_JPN
140 240
120
200
100
160
80
120
60
80
40
20 40
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_GBR D_GBR D30_GBR ACTUAL_CAN D_CAN D30_CAN
Figure 9 presents the results for a set of euro area countries, namely, Austria, Germany, France,
Italy, The Netherlands and Spain.122 For Austria and The Netherlands, potential debt remains
higher than actual debt over the whole period. However, for Italy, Spain and, to some extent,
France, potential debt falls below actual debt in 2002 for Italy (with potential equalling actual
around 100%), in 2012 for France (with equality in 91%) and in 2010 for Spain (with equality in
around 60%). Germany also has a negative trend of potential debt approaching actual debt in
around 2010 (at nearly 90%) without crossing it. Potential debt stays below actual debt at around
55% in Spain, 58% in Italy and 88% in France. Therefore, these could be the safer limits of debt
(not altering growth) for these countries, and they could keep their public debt under control at
lower ratios than these values. In conclusion, for this set of countries, the potential debt to target
122
As a reminder, note that Austria’s estimated elasticity is not statistically significant, and the graph shows the
simulated debt for Austria based on the average of elasticities equalling 0.3.
234
public debt
in bad times is around 60% to 90%, based on an average of their economic and monetary
performances. However, in normal economic conditions, potential debt could reach more than
double these values.
government investment and economic and monetary performances for Austria, Germany, France, Italy, The
Netherlands and Spain*
500 240
200
400
160
300
120
200
80
100
40
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_AUT D30_AUT ACTUAL_GER D_GER D30_GER
240 300
200 250
160 200
120 150
80 100
40 50
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_FRA D_FRA D30_FRA ACTUAL_ITA D_ITA D30_ITA
300 300
250 250
200 200
150 150
100 100
50 50
0 0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
ACTUAL_NLD D_NLD D30_NLD ACTUAL_SPA D_SPA D30_SPA
*: Countries with only two curves are those for which elasticities are not significant, and thus we limit the exercise of
simulated public debt to their elasticities as the average of all significant elasticities (0.3)
235
public debt
The last group of countries (Figure 10) focuses on the rest of the eurozone countries, namely,
Greece, Portugal, Ireland and Belgium. Except for Portugal, which has the same tendency for
potential debt as Spain and Italy (Spain and Italy are described in Figure 9), the three other
countries have different stories of their own. Belgium’s potential debt goes below its actual debt
in the early years of the 1980s and follows the same tendency as actual debt until 2008, where it
drops and the two curves disconnect and evolve oppositely. Potential debt continues to decrease
until it reaches a lower value of 68% at the end of the period.
For Greece, potential debt crosses actual debt in 1985 at a lower rate of public debt around 47%
and remains with the trend of actual debt for a decade, where potential debt once again surpasses
actual debt in 1995 and stays until 2002, where it drops again below actual debt (as in Italy).
Potential debt continues to decrease under actual debt until the end of the period and even becomes
negative starting in 2010, reaching around -10% in 2011–15. The negative number should be
interpreted in the sense that Greece in this period should not hold any public debt based on its
economic performance at this time.
Nevertheless, some results create serious questions about the validity of the data of interest rates,
inflation, growth, and so on, used to simulate the formulae for those countries and their sensitivity
to the actual data.123 For Ireland, there are two periods where debt is unsustainable (targeted debt
is below actual debt), in 1983–93 and 2008–14. However, Belgium shows an apparently
unsustainable path from the 1980s.
123
Data distribution is relatively heterogenous between countries, where some variables are normally distributed for
some countries and others are not, as shown by Jarque-Berra, Skewness and Kurtosis for this data in Table A.3
(Appendix A).
236
public debt
government investment and economic and monetary performances for Belgium, Greece, Ireland and
Portugal
160 250
140
200
120
150
100
100
80
50
60
0
40
20 -50
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_BEL D_BEL D30_BEL ACTUAL_GRC D_GRC D30_GRC
300 400
350
250
300
200
250
150 200
150
100
100
50
50
0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_IRL D_IRL D30_IRL ACTUAL_PRT D_PRT D30_PRT
4.3. Robustness check
4.3.1. The impact of data shortness
The data on long-term interest rates from the OECD database is unfortunately short samples for
some countries such as Japan, where the data time series starts in 1989, and for Greece in 1998.124
Therefore, for robustness check and to avoid the problem of data shortness, we produce estimations
for potential debt, calibrating the differential of growth and interest rate in the denominator of
Equation (30) by the data of the United States (but we leave the growth rate in the numerator
unchangeable and proper to the country itself). This is also a test for our formula against some
irregular observations in the data. In fact, countries such as Japan, Greece and Ireland are
124
For this purpose we checked many other international sources reporting data on interest rates for government bonds,
securities and treasury bills, such as the International Financial Statistics (IFS) of the IMF, the Bank for International
Settlements (BIS) and the Federal Reserve Bank of Saint Louis
(https://fred.stlouisfed.org/tags/series?t=interest+rate%3Blong-term). These sources report data on long-term interest
rates for different periods, depending on the country, which is the same data as reported by the OECD database.
237
public debt
interesting cases to study, with higher accuracy and deep investigation of the data, and even
specifications to be re-estimated. Along the same lines, Barta (2018) analyses and compares the
cases of Belgium, Canada, Denmark, Greece, Ireland, Italy and Japan, since the 1970s, to identify
factors that differentiate countries accumulating threatening debt from those that keep their debt
under control. He notes that practices in fiscal policy management make a difference rather than
the political bias impact of debt accumulation, as proposed by some authors in reference to the
positive approach of public debt (Persson and Svensson, 1989; Alesina and Tabellini, 1990).
Figure 11 shows the results for Greece and Japan for this exercise. According to the new simulated
debt for Japan, this becomes positive and high with a downward tendency until crossing the
increasing accumulated debt in 2008 at the ratio of 192%. The results produced for Greece are also
improved compared to its own data on interest rates, as previously explained. The intersection
between actual and potential debt is materialized in 2005 at an average ratio of debt equivalent to
100%. Indeed, an IMF note on fiscal space calculating the debt limit based on the fiscal reaction
function, and the interest rate growth differential, shows that the public debt dynamics are not on
a sustainable path to converge to a finite steady-state ratio for the following countries: Greece,
Italy, Japan and Portugal (Ostry et al., 2010).
Figure 11. Simulated potential public debt for Greece and Japan using the US long-term interest rate and
economic growth differential
500 350
300
400
250
300 200
150
200
100
50
100
0
0 -50
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_JPN D_JPN D30_JPN ACTUAL_GRC D_GRC D30_GRC
An exercise of simulation is also conducted for all the other countries calibrating the gap (long-
term interest rate – growth) by the United States one, but although the tendency changes slightly
for many countries, the years of intersections between actual and simulated debt are delayed for
some countries, such as Belgium, until the year of the financial crisis (the intersection using its
238
public debt
own data is around 1982). For the United Kingdom, potential debt stays higher even in times of
crisis when calibrating with the United States’ interest rates, and for many other countries the
average of the simulated debt increases substantially, especially for Australia, Austria, New
Zealand, Belgium, France, Italy, Spain, Greece and The Netherlands. However, for some
countries, such as Denmark, Sweden, Norway and The Netherlands, the average is almost stable
between the two exercises, while it decreases substantially for Switzerland. The results for all the
sample countries are stacked in Figure B.6 in Appendix B.
4.3.2. The impact of elasticity
To gauge the effects of elasticity changes, we assess the effects of three values – the estimated one
from the model, then the one calibrated to 0.15 and 0.30 values – on potential government
investment and debt (Table 3). First, we see that the impact of variation of elasticities is higher for
small countries than sizeable countries. Doubling the value of elasticity (for example, from 0.15
to 0.30) leads to a decrease in simulated public debt, on average, by around three points of GDP
for small countries such as Greece and Ireland. This impact is, however, contained, on average, at
around 1 to 2 points of GDP for sizeable countries (the United States, Japan, Germany, France, the
United Kingdom).
Table 3. The effects of elasticity changes on the potential public debt limit
D D15 D30 D15-D30 D-D30
Mean Median Mean Median Mean Median Mean Median Mean Median
Australia 99.3 96.2 98.3 95.2 96.9 93.9 1.3 1.4 2.3 2.4
Austria 300.7 303.0 301.1 303.3 300.2 302.6 0.9 0.7 0.6 0.4
Belgium 111.4 114.3 111.7 114.5 111.0 114.0 0.7 0.4 0.4 0.2
Canada 138.9 132.3 139.4 132.8 138.5 132.0 0.9 0.9 0.3 0.3
Denmark 233.3 237.7 233.0 237.5 232.7 237.2 0.3 0.2 0.6 0.4
France 131.1 125.8 130.9 125.7 130.4 125.4 0.4 0.3 0.7 0.4
Germany 130.9 140.6 132.6 141.8 132.0 141.4 0.6 0.4 -1.1 -0.8
Greece 90.1 82.9 84.5 80.7 82.6 79.8 1.9 0.9 7.6 3.1
Ireland 163.5 167.3 177.9 176.2 175.1 174.8 2.8 1.4 -11.6 -7.5
Italy 143.3 136.8 145.0 137.5 144.0 137.1 1.0 0.4 -0.7 -0.3
Japan -591.1 -592.7 -585.1 -589.2 -582.8 -587.8 -2.3 -1.4 -8.2 -5.0
Netherlands 178.9 181.5 178.9 181.5 178.3 181.1 0.6 0.4 0.6 0.4
New Zealand 101.7 100.8 101.0 100.4 100.4 100.0 0.6 0.4 1.3 0.8
Norway 150.5 154.8 152.4 156.8 151.3 155.6 1.2 1.2 -0.8 -0.8
Portugal 209.8 209.8 210.2 210.1 208.5 208.8 1.8 1.3 1.4 1.0
Spain 105.1 105.7 105.4 105.8 104.6 105.2 0.8 0.6 0.6 0.4
Sweden 170.3 170.5 171.7 171.2 171.2 171.0 0.5 0.2 -0.9 -0.4
Switzerland 254.7 259.6 254.8 259.6 253.4 258.9 1.4 0.8 1.3 0.7
United Kingdom 98.1 101.6 98.9 102.4 98.5 102.0 0.4 0.4 -0.4 -0.4
United States 187.4 194.9 187.5 194.9 186.7 194.0 0.8 0.9 0.7 0.9
239
public debt
4.3.3. The impact of the differential “long-term interest rate – economic growth rate”
Since the elasticity impact is not highly determinant on the results, as previously shown, we keep
constant the elasticity of public capital at the average of 0.30125 for all countries and simulate the
impacts of the gap between interest rate and growth. In fact, as shown in the previous paragraph,
the elasticity impact is very low on the results compared to what we will show for the differential
in interest rates and economic growth. Furthermore, we focus on some countries where long-term
interest rate data is available over the whole history. We choose, for example, sizeable countries,
namely, the United States of America, France, the United Kingdom and Canada (Japan and
Germany were excluded for reasons of data shortness on interest rates). We simulate results
considering the whole period, and the two sub-periods 1960–84, as well as in the so-called Great
Moderation Era,126 1985–2015, to see the effects of the gap between interest rates and growth (the
denominator in Equations 30 or 31). Descriptive statistics of long-term interest rates and economic
growth for these countries are displayed in Table 4. We note that the difference between long-term
interest rates and economic growth is higher in the period 1960–84 than in the period 1985–2015.
Table 4. Real long-term interest rates and real economic growth differential for selected countries
Long Term Interest Rates (% per GDP Growth rate Gap [interest rate-
annum) (%) growth]
Sample: 1960-2015 Mean Median Mean Median Mean Median
France 7.3 6.7 2.9 2.4 4.5 4.3
United States 6.3 6.1 3.1 3.3 3.2 2.8
United Kingdom 7.8 7.5 2.4 2.6 5.4 4.9
Canada 7.0 6.8 3.2 3.1 3.8 3.7
France 9.4 8.6 4.2 4.5 5.2 4.0
United States 7.3 6.8 3.6 4.5 3.7 2.4
United Kingdom 9.7 8.8 2.5 2.7 7.2 6.1
Canada 8.3 7.6 4.2 4.0 4.1 3.5
France 5.6 4.9 1.8 2.0 3.8 2.9
United States 5.5 5.3 2.7 2.7 2.8 2.5
United Kingdom 6.2 5.1 2.3 2.5 3.9 2.5
Canada 6.0 5.5 2.4 2.6 3.5 2.9
125
We should remember that this is the average of the significant estimated public capital elasticities over the sample
of countries. Fixing the elasticity for all countries to a common value allows comparison of the countries’ results based
on the impact of the interest rate and economic growth gaps alone.
126
The Great Moderation Era is first described by Stock and Watson (2003) analysing, over the period 1960–2002,
the United States quarterly GDP volatility, shown to be highly reduced after 1985 compared to the previous period.
This tendency of output and price stability is mainly attributed to the macroeconomic stabilization policies followed
by independent central banks (Bernanke, 2004). Some economists argue that this period came to an end in 2007 with
the 2008 financial crisis, while others argue that the GDP volatility averages are still lower than pre-1985, assuring
the continuity of the Great Moderation Era (Clark, 2009).
240
public debt
The effects of interest rates and economic growth differences are produced in Table 5 and Figure
11. Overall, there are sizeable and substantial effects on potential debt for the four selected
countries. The averages of simulated public debt, over the period 1960–84 (Scenario 1), is around
168% for the United States, 118% for France, 78% for the United Kingdom and 135% for Canada.
These averages are, respectively, higher in the Great Moderation Era, 1985–2015 (Scenario 2), by
around 38 points of GDP in France, 52 points in the United States, 57 points in the United Kingdom
and 20 points in Canada. For the whole period, the averages of simulated debt are 137% in France,
195% in the United Sates, 100% for the United Kingdom and 145% for Canada.
Table 5. Simulated effect of differences between long-term interest rates and economic growth
Simulating potential debt over the period 1960-2015, using constant elasticity of public capital = 0.30
France United States United Kingdom Canada
Mean 137.4 194.7 100.5 145.1
Median 126.2 194.6 106.6 134.3
Maximum 212.1 279.7 123.7 213.0
Minimum 88.8 127.4 59.9 102.9
Scenario 1: simulating potential debt using interest rates and growth of the period 1960-1984
France United States United Kingdom Canada
Mean 122.2 172.2 77.6 137.1
Median 112.3 172.1 82.3 126.9
Maximum 188.6 247.3 95.5 201.2
Minimum 79.0 112.6 46.3 97.2
Scenario 2: simulating potential debt using interest rates and growth of the period 1985-2015
1985-2015 France United States United Kingdom Canada
Mean 156.0 220.1 134.7 154.4
Median 143.2 220.0 142.9 142.8
Maximum 240.7 316.1 165.8 226.5
Minimum 100.8 144.0 80.3 109.5
Scenario 2 - Scenario 1
1985-2015 France United States United Kingdom Canada
Mean 33.7 47.9 57.1 17.3
Median 31.0 47.9 60.5 16.0
Maximum 52.0 68.8 70.2 25.3
Minimum 21.8 31.4 34.0 12.2
241
public debt
Figure 12. Simulated potential debt sensitivity to the differential of interest rates and economic growth for
selected countries
Using differential [interest rate-economic growth] of the period 1960-1984 Using average differential [interet rates - economic growth] of the period 1985-2015
280 350
240 300
200 250
160 200
120 150
80 100
40 50
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
D30_FRA D30_USA D30_FRA D30_USA

D30_GBR D30_CAN D30_GBR D30_CAN
4.4. Discussion
Contrary to the second chapter, where we adopted a short-term approach assessing the impact of
the presence of public debt on GDP via the effects on the fiscal multipliers, as well as the
interaction with the business cycle and other economic variables, this chapter has adopted a long-
term approach based on an endogenous growth model to assess the relationship between public
debt and economic growth. Economists (classical and new classical) adopting this approach
generally try to avoid the role of empirical estimations and data in their calculations (adhering to
the Lucas (1976) and Sims (1972) critiques). They prefer calibration to estimation and argue as if
all economies are in the steady state in the long term, which is not necessarily true. Furthermore,
calibration is another way of deriving parameters that are under the control of economic theory
and do not necessarily reflect the behaviour and data-generating process proper to each country.
Furthermore, the results of the calibrated models are theoretical and not so easy to interpret by the
policy-makers. In our case, we judge it useful that countries are not necessarily in the steady-state
path and hence considered in transitional dynamics, even in the long term; we prefer to run
estimations in the long term, being suitable for the approach of endogenous theory, and we stay
away from calibrated elasticities. The only calibrations adopted in some variables are based on
data averages.
Chapter 1 of this thesis reported detailed literature on the public debt threshold, which reported
different numerical thresholds revealing sensitivity to the countries’ sample (for panel
regressions), to the period and the country on individual levels (for single-country regressions).
The importance of the debt limits has already been taken into consideration before the recent
financial crisis, first by the Europeans in the establishment of the European Monetary Union
(Maastricht criterion of 60%), and by multilateral institutions in designing their loan programmes,
especially the International Monetary Fund and the World Bank institutions. These Breton Woods
242
public debt
Institutions designed an approach called debt sustainability analysis (DSA) for low-income
countries (LIC) and market access economies (i.e. emerging economies and advanced economies).
This practical framework became operational and was applied especially to LIC in 2002 (IMF and
the World Bank, 2005, 2013). However, since the 2008 crisis, this approach is permanently revised
and has been updated to include other highly indebted countries, which is the case for advanced
countries (IMF, 2013). The DSA is a kind of stress test for debt sustainability based on limits of
some important financial indicators, namely, bond yield spreads, external financing requirements
(as a percentage of GDP), public debt held by non-residents (share of total), public debt in foreign
currency (share of total) and annual change in the share of short-term public debt at original
maturity. A safer benchmark gross government debt of 60% and 70% of GDP was calculated for
the groups of emerging countries and advanced countries, respectively. The 70% limit was able to
reach 85% for the group of advanced countries. This threshold is less than the 90% threshold that
fits all countries suggested by Reinhart and Rogoff (2010).
In this section, we compare our results with some results produced in the recent literature that have
adopted modelling techniques derived from economic theory, although the approaches differ.
Some papers that conclude debt limits, and then the fiscal space defined as these debt limits minus
actual debt, are influential and have important policy implications (Ostry et al., 2010, 2015; Ghosh
et al., 2013; Pienkowski, 2017). For many of these researchers, the debt limit is assessed around
the period of the financial crisis of 2008/2009. Therefore, to compare our results with those raised
by some authors, we limit this comparison to a very short period or a specific year of comparison
(for example, around the 2008 financial crisis).
Returning to our earlier results, Table 6 shows the average limits over the crisis period, 2009–12,
for the 20 countries in the sample. The second column shows observed average debt over the period
2009–12, the third, fifth and seventh columns show, respectively, the averages of simulated
potential debt over the same period considering the economic performances proper to each country
(Potential1), then simulation based on the calibration of economic growth and long-term interest
rates to the Great Moderation Era (1985–2015) performances (Potential2), and the simulation
calibrating long-term interest rates and economic growth to those of the United States for all
countries (Potential3). The other columns, Gap1, Gap2 and Gap3, present the difference between
the previous simulations (Potential1, Potential2 and Potential3) and observed debt (actual),
respectively. When actual debt overcomes simulated debt, the gap is negative, and the country
should curve the accumulation of debt by rapid de-leveraging.
A negative gap over the years of the financial crisis (2008–12) is recorded for Belgium, the United
Kingdom and Spain, and highly negative for Japan, Greece and Italy (Gap1). Portugal, Germany
and France also recorded very low but positive gaps. Considering the Great Moderation Era, Gap2
is negative for Portugal, in addition to the previous countries cited for Gap1, except the United
Kingdom and Japan. The latter recorded high potential debt in contrast to high negative potential
debt for Gap1. Considering the third simulation, Greece, Italy, Japan and Portugal are countries
243
public debt
with worrying debt, even considering the United States’ benchmark for economic growth and
interest rate performances. However, some countries’ potential debt has significantly increased in
this simulation. This is the case for Australia, New Zealand, France, the United Kingdom, Spain
and, to a lesser extent, Germany. Other countries, however, have seen their potential debt decrease
in this last simulation, namely, Austria and Switzerland.
Table 6. Simulated potential debt following the 2009–2012 crisis
Sample: 2009-2012 Actual Potential1 Gap1 Potential2 Gap2 Potential3 Gap3

Switzerland 48.4 255 206.6 283.8 235.4 131.8 83.4
Austria 71.9 179.8 107.8 133.2 61.3 112.1 40.2
Sweden 39.5 138.2 98.7 121.9 82.4 148.8 109.3
Denmark 41.6 119.4 77.8 100.9 59.3 126.8 85.2
Australia 20.3 84.4 64.2 95.6 75.4 126.6 106.4
Norway 43.5 107 63.4 86.1 42.6 111.7 68.2
New Zealand 30.2 87.9 57.8 91.1 60.9 153.1 123
Netherlands 62.8 107.8 45.1 126.7 64 110.8 48.1
United States 91.5 132 40.5 149.2 57.7 132 40.5
Ireland 84.5 111.8 27.3 310.1 225.6 141 56.5
Canada 80.8 104.2 23.4 110.8 30 122.5 41.7
France 81.2 90.9 9.7 103.1 21.9 124.7 43.4
Germany 76.8 85.2 8.4 102 25.2 95.8 19
Portugal 96.8 102.5 5.7 70.5 -26.3 86.8 -10
Spain 62.4 58.9 -3.5 59.5 -2.9 92.7 30.3
United Kingdom 73.2 62.7 -10.5 84 10.8 103.4 30.3
Belgium 96.5 76 -20.5 84.3 -12.2 98.3 1.8
Italy 117.2 60.9 -56.3 41.1 -76.1 63 -54.2
Greece 144.2 1.5 -142.7 1.1 -143.1 2.1 -142
Japan 215.5 -432.1 -647.5 876.5 661 185 -30.4
Note: Countries are filtered, according to the column Gap1, from high values to lower values.
In the same way, Ghosh et al. (2013) use the approach of Ostry et al. (2010) to calculate the debt
limits for advanced countries and to produce a fiscal space defined as the difference between that
debt limit and the actual debt-to-GDP ratio, based, in particular, on the interest rate and economic
growth differential and considering the primary balance. They find that Greece, Italy, Japan and
Portugal have no fiscal space as their debt is unsustainable. However, the other countries have
enough fiscal space, especially given that the assessed debt limits range from minimum values of
152% and 154% for Canada and Germany, respectively, to high values of 246% and 263% for
Ireland and Norway, respectively. For the set of countries with positive fiscal space, the latter
values are around low values of 50% to 70% recorded for the United States, Ireland and Belgium,
medium values from 75% to 100% recorded for the United Kingdom, France and Germany, and
higher values of higher than 100%, especially in northern countries (Norway, Sweden and
Denmark) and Australia and New Zealand. These results are generally in line with our results in
Table (6).
244
public debt
The same exercise was updated by the international rating agency, Moody’s, adopting the same
approach as Ostry et al. (2010) and Ghosh et al. (2013), to calculate a distance to debt limit for a
sample of advanced and emerging economies. The sample of countries also covers the 20 advanced
countries that we studied in this chapter, in addition to other countries. The results are reproduced
in Ostry et al. (2015) in a figure summarizing these distances to debt limit, which are reproduced
in Figure 13. The latter indicates that Japan, Italy and Greece have zero fiscal space, which was
also confirmed by our results in Table 6 (Gap1). Belgium, Spain and the United Kingdom also
report negative differences between potential debt and actual debt in Table (6), concordant to some
extent with Moody’s results, where fiscal space is very low, except for the United Kingdom. Our
results (Table 6) also show that Switzerland, Austria, Sweden, Denmark, Australia, Norway, New
Zealand, The Netherlands, The United States and, to some extent, Ireland and Canada have safer
potential debt higher than actual debt, while France, Germany and Portugal have average potential
debt approaching actual debt. These results are also generally in line with Moody’s calculations,
as reported in Ostry et al. (2015), which are reproduced in Figure 13.
Figure 13. Fiscal space: distance to debt limit (percentage points)

0 50 100 150 200 250
Switzerland 202
206.6
Austria 156.6
107.8
Sweden 188.8
98.7
Denmark 195.7
77.8
Australia 214.5
64.2
Norway 248
63.4
New Zealand 228.1
57.8
Netherlands 158.1
45.1
United States 165.1
40.5
Ireland 105.5
27.3
Canada 149.8
23.4
France 116.9
9.7
Germany 167.9
8.4
Portugal 58.8
5.7
Spain 115.2
0
United Kingdom 132.6
0
Belgium 124.3
0
Italy 0
0
Greece 0
0
Japan 0
0
Fiscal Space as reported in Ostrey et al. (2015) Distance to debt limit (Gap1 from table 6)
Note: For comparison purposes, distance to debt limit, as defined in Ostry et al. (2015), is reproduced and reported
for our sample jointly with the distance from actual to potential debt calculated by our method (Gap1 in Table 6). For
negative Gap1 (potential tolerable debt below observed debt), we report zero instead of negative numbers.
245
public debt
In another IMF paper (Pienkowski, 2017), the debt limit for advanced countries is found to be
137% of GDP. This debt tolerance can be enhanced by issuing GDP-linked bonds, which can
increase this limit to 238% of GDP for all advanced countries (when the bonds issuance is 100%
linked to GDP). The author concludes that this linked debt to GDP clearly shows that there is no
one-size-fits all, which endorses our results.
Another important point revealed in this chapter is the effects of the interest rates and economic
growth differential (gap) in shaping the curve of potential public debt over time. The effects are
higher on potential public debt in the Great Moderation Era (1986–2015) than in 1960–85. This is
due to the gap being higher, on average, in the 1960–85 period and narrowing in the Great
Moderation Era (Table 4). This latter period is characterized, on average, by lower interest rates,
but also lower economic growth compared to 1960–85, which makes the differential between long-
term interest rates and economic growth lower than the one in 1960–85, as revealed in Table 4. As
a consequence of lower interest rates, the Great Moderation Era has played a bigger role in the
accumulation of public debt that countries nowadays struggle to decrease (Cecchetti el al., 2011).
A recent debate was triggered by Blanchard (2019) assuming that, on average (opposite to the case
here for long-term interest rates), interest rates are lower along history than economic growth,
which makes debt safer in advanced economies and means that there is nothing to be concerned
about. Some economists interpreted this as an appeal for fiscal expansion and avoiding fiscal
austerity. More details of this debate are presented in the Chapter 2, as well as a recent discussion
about the type of interest rates by Blanchard (2019) in his analysis.
The third result is that potential government investment to GDP ratios are, on average, higher than
the observed government investment ratios, and they tend to decrease over time. This may indicate
that, first, countries are not reaching their potential productive investments, and, second, that the
issued public debt may drift from financing productive capital to other government expenses as an
enhancement of the welfare state needs and ageing population expenditure. In fact, the size of
social spending in OECD countries increased from 18% in 1980 to 26% in 2014 (Alesina and
Passalacqua, 2015).
Despite the importance of such results, some caveats should be considered, and future development
could enhance the results of our approach. First, we assumed that potential debt is only issued to
finance potential public capital derived by the model, while the finance of the other current
government expenditure is supposedly filled by collecting taxes. This may have led to generating
simulated higher potential public debt, which stays high, above actual debt, especially in good
times for many countries. However, tax revenue is not sufficient to compensate the financing needs
of total expenditure. For example, according to the OECD database website, tax revenue in France
was around 45% of GDP in 2015 (above an average of 34% for OECD countries), while general
expenditure stood at around 57% (social expenditure 43% plus social protection 14%).
Government investment was around 3.5%. Hence, a deficit of approximately 11 points of GDP
was not compensated by taxes (around 20% of additional taxes are needed to finance such needs).
246
public debt
If we adjust the potential debt to consider such a fact (let us reduce potential government
expenditure for all countries by an extra 20% of their actual taxes),127 the potential public debt is
significantly reduced, on average, for example, to 35% over the period 1960–2015, 72% in
Germany, 21% in Italy, 50% in Ireland, 38% in Spain, 62% in Canada and 115% in the USA.
Without considering this fact, averages are higher and are, respectively, 131% in France and
Germany, 143% in Italy, 163% in Ireland, 105% in Spain, 139% in Canada and 187% in the United
States. Some other countries, such as the United Kingdom, saw potential debt significantly reduced
to very low levels.
In this chapter, the adopted theoretical framework of the Cobb-Douglas production function has
focused on the long-term approach to simulate the potential debt that a country could target without
undermining its public finances. However, this long-term approach omits the short-term effects of
fiscal policies, which are highly determinant in public debt management. The framework of
endogenous growth models could be suitable for long-term analysis, while debt related to fiscal-
policy short-term effects is omitted. Furthermore, we derived potential debt from potential public
productive investment, neglecting the effects of other public expenditure such as social spending
in education, justice, health, and so on. In fact, as stated by Corsetti and Roubini (1996), “The
productive public spending can be thought of as encompassing very different types of publicly
provided goods and services, such as justice, enforcement of law and contracts, police services,
educational services and government research activities”. Social spending in education, health,
justice, and others, is to enhance human capital, and to create qualified political and economic
institutions, which develop the business environment. The rule of law and order, as well as
security, are also crucial determinants ensuring stability for the business environment. Taking this
into consideration, it is highly recommended to consider the effects of such categories of
expenditure and to analyse their impacts on economic growth in both the short and long terms.
5. Conclusion
Contrary to the first chapter, which examined the purely statistical–econometric relationship
between economic growth and public debt, albeit with advanced econometric tools (Hansen’s
regression kink method), this chapter has extended the investigation to the role of theory-based
models, taking into consideration country-specific fundamentals. Therefore, the chapter first
surveyed the effects of public debt on economic growth in different classes of theoretical and
empirical public debt models (OLG, ILA, endogenous growth models, new Keynesian DSGE
models, and the normative versus the positive approach). In the second part, an endogenous model
was applied to the sample studied in the first chapter, made up of 20 developed economies.
127
The results of this exercise for all countries are reproduced in Figure B.6 in the appendix.
247
public debt
In this model, we simulated a parameterized formula for a potential/limit debt that a country could
target to finance its productive investments. These limits, linked to economic growth and public
capital productivity, as well as the interest rate, are dynamic, country- and time-specific and tend
to evolve contrary to actual accumulated public debt. Simulated public debt, in particular, drops
under actual levels of debt in times of crisis, especially for many advanced countries severely
affected by the crisis. This sends a clear message of policy recommendation that countries are safe
from the danger of public debt as long as potential debt (simulated/targeted) stays higher than
actual public debt. The results show that many countries are under the stress of public debt,
especially after the financial crisis of 2008. For countries such as Ireland, this stress is short-lived,
and potential debt quickly remains higher than actual debt following the redressing in economic
activity a few years after the crisis. However, for many other countries, the effects are prolonged.
The results also revealed that countries such as Greece, Belgium and, to some extent, Italy, had
problems of accumulated debt, in the early 1980s for the two first countries, and the early 2000s
for Italy.
The potential (limit) public debt is country-specific and evolves contrary to actual accumulated
debt over time. Indeed, it moves in line with the public capital productivity trend, which
historically tends to decline while actual public debt is rolling over. The story of the debt limit was
intensely debated at the forefront of the 2008 financial crisis. The different results found in the
literature supporting the debt limit existence, whether for all countries (or at least for countries of
the same level of development), or case-specific countries, tend to discuss the threshold debt
generally, as a one-size-fits-all for countries, whether determined endogenously or set
exogenously. However, the added value of the present work is that the debt limit is an endogenous
parametrized function linked to the economic returns (productivity) of public capital financed by
such issued debt. Another point to note is that, although we found that the endogenous debt limit,
whenever it exists, is country- and time-specific, using Hansen’s (2017) regression kink method
in the first chapter, this chapter added the finding that potential debt limits are determined yearly
based on the country’s principal macroeconomic fundamentals.
248
public debt
Appendix A. Tables
249
Chapter 3. Public debt effects in theory-based models with an empirical assessment of the potential public debt
Table A.1. Summary and features of the models of public debt and fiscal policy
Authors Type Main Features Main hypotheses Results
- Domestic debt raises interest rate and lowers utility in the efficient case.
- Two periods of infinite long-life agents
- Governments have a one period maturity. - Results for the inefficient case depend on the existence of the foreign debt with
living for 2: working in the 1st and
Diamond - Debt pays the current interest rate. domestic debt.
OLG retiring in the 2nd.
(1965) - Taxes are lump-sum. - Foreign debt widen the gap between the interest rate and growth, lowers utility in
- Study the impact of domestic/foreign
- Debt labor ratio is kept constant. the efficient case.
debt on the interest rate and utility level.
- Results for utility are mixed for the inefficient case. Debt Swap has the same effects.
- The model’s features are mainly based - Government bonds effects on the interest rate and utility does not necessarily hold
on the Samuelson-Diamond properties. - The model’s hypotheses are mainly based and depends especially on the existence of operative altruistic links (bequest motives)
Barro
OLG - Study the effects of public debt on on the Samuelson (1958) and Diamond between generations.
(1974)
interest rate and utility in steady state (1965) models’ assumptions. - Intergenerational transfers guarantee debt-neutrality as the debt-for-tax swap does
equilibrium. not affect the resource allocations and interest rates of the current generation.
- Debt increase (in the steady state): changes the foreign assets in agents’ wealth in an
- A finite lived agents’ horizon.
open economy, and, decreases the level of capital and consumption in a closed
- Studies the effects of debt and deficit
economy.
Blanchard accumulation. - Taxes are lump-sum.
ILA - Taxes decrease: raises wealth and consumption. Effect is larger the longer taxes are
(1985) - Design an “index of fiscal policy” - Constant probability of death.
shifted to future generations. Debt and taxes increases create an initial wealth effects
capturing the effects of current and
on consumption, leading to capital decumulation which makes capital and
expected fiscal policy.
consumption lower in the new steady-state level.
- Ricardian equivalence validation depends on the way taxes are distributed among
- Use a modified version of the OLG - Enabling changeable interest rate to
agents and not on the length of the lifetime horizon. The debt-neutrality holds if the
Aiyagari model of Samuelsson (1958). deficit policy regimes.
OLG tax distribution does not change agents’ wealth allocations. A higher level of spending
(1985) - Assess the Ricardian equivalence - Taxes distributed differently across
can be financed by debt at an unchanged interest rate, if taxes’ distribution maintains
validation. heterogeneous agents.
wealth distributions.
- The equilibrium is independent of the pattern of the public debt and lump-sum
taxation over time.
- The difference between expected government tax base and the future tax base of the
- The model is based on the Yaari-
- Private consumption behavior is current individuals explains the lump-sum taxation variations over time. Agents not
Buiter Blanchard models.
OLG modelled following Yaari (1965) and linked to the future generations by bequest transfers omit their successors’ resources
(1988) - Evaluates the effects of government
Blanchard (1985) approaches. from their budget constraint.
bonds on economic agents’ wealth.
- Debt-neutrality holds if and only if the population growth and the probability of
death sums to zero. Under this condition, a non-zero labor productivity rate will not
destroy this debt-neutrality.
250
- Assumes new cohorts over time, and new - The equilibrium interest rate hinges on the government financing decision
families aren’t linked to the existing by (government bonds by levying lump-sum taxes in this case) and speed of new arrival
- A model of “overlapping families of
transfers. cohorts/families.
infinitely lived agents”, aims to show
- Operative transfers between some but not - Government bonds increase the equilibrium interest rate for a positive non-zero
that: 1st, the “infinite lifetime horizon”
Weil (1989) ILA all agents result in partial linkages leading population growth rate: expected taxes to pay the issued debt are compensated by
assumption do not necessarily imply
to infinite economic horizon. generations that are not yet born and are not considered by the current generations.
debt-neutrality; 2nd, the “finite horizons”
- Agents’ infinite lifetime is assured by the The real interest rate must hence rise to maintain aggregate consumption at its market-
don’t necessarily violate debt-neutrality.
arrival of the families who are not linked clearing level. Infinite lifetimes are so not inconsistent with the violation of the debt-
by transfers. neutrality assumption. Also, finite horizons do not necessarily violate the latter.
𝑦 𝑔 𝛼
- In case of a Cobb-Douglas production function, = 𝐴 ( ) : an increase in taxes
𝑘 𝑘
- Adopts an AK modeling framework reduces the long-term growth while an increase in expenditures raises it by raising
assuming constant returns to scale and marginal productivity of capital.
- Aims at establishing the effect of public - The two effects cancel for the optimal government size equalizing government
including the flows of public services as
spending and government debt policies 𝑔
expenditure to tax rate and the share of the capital ( = 𝜏 = 𝛼). The growth rate
input in the production function. 𝑦
on utility and long-term growth in a
- Argues that private inputs are not a close function of the government size is an inverted U curve: for a government small size,
perspective of endogenous growth
Barro substitute for public inputs as some public raising expenditures’ effect dominates raising tax rate’s effect, while for a large
EGT theory.
(1990) activities are difficult to be insured by government size, the negative effect of taxes on growth dominates.
- The model studies a closed economy
private firms (defense, law and order), or - Maximum utility corresponds to same conditions maximizing growth if the elasticity
with infinite lifetime agents and
the service is non-rival (ideas), or as of substitution of 𝑦 to 𝑔 = 1.
intertemporal preferences modelled by a
external effects cause lower private output - In non-Cobb-Douglas case: maximum growth depends on the elasticity of
utility function.
(basic education). substitution between public services and private capital and, the relative size
- Assumes timely balanced-budget. maximizing utility exceeds the one maximizing growth rate if and only if the
magnitude this elasticity substitution is superior to 1.
- Same approach as in Barro (1990).

- Adopts an AK modeling framework. - Public investment stimulates aggregate production indirectly via the accumulated
Study the effect of public capital stock
Futagami et - The model includes the stock of public stock instead of flows as does Barro (1990). The introduction of the productive public
(instead of government services flows)
al. (1993) EGT capital as an input in the production stock allows dynamic transitional effects analysis instead of being restricted to the
and public debt policies on utility and
function. steady state analysis as in Barro (1990).
growth.
- Assumes timely balanced-budget. - Tax rate maximizing economic growth rate is higher than the one maximizing utility.
- Distinguish an optimal tax rates on both types of capital under different assumptions
- Assume unbalanced budget constraint on technology and distribution and analyze the welfare properties of public debt and
- Multi-sector model encompassing
allowing government borrowing/lending. assets.
productive public spending to assess
- 4 models are derived depending whether: - If the government spending is an input in the production function only (not in the
optimal fiscal policy in the approach of
public spending as input in the output human sector), the optimal choice of the government spending leads to a constant ratio
Corsetti and endogenous growth models.
sector, affects the productivity of capital of expenditures to output in every time independently of the tax policy. This ratio is
Roubini EGT - Incorporate a separate human capital
(model 1); affects the productivity of 𝑔
(1996) accumulation as a second sector. less than the one found by Barro (1990) ( = 𝛼(1 − 𝜀) < 𝛼 𝑎𝑠 𝜀 > 0). In model 1,
human capital (model 2); or public 𝑦
- The production function is a Cobb- the optimal tax on human capital is zero and the one on physical capital is positive.
spending as input in the human capital
Douglas with physical capital, human This result is inverted in model 2. In both models (1 and 2), a government optimal
sector, affects respectively the same
capital and flows of public goods. choice of both government spending and tax rates leads to an instantaneous balanced
previous variables (models 3 and 4).
budget and the optimal public debt is zero.
251
Benigno - New Keynesian framework based on - Optimal public debt would follow a random walk path if the government can achieve
and - New Keynesian models augmented by
Christiano et al. (2005) and Smets and a time-inconsistent policy commitment. Under the assumption of price stickiness, the
Woodford the government’s budget constraint.
Wouters (2007). government (social planner) chooses to rely more heavily on changes in income tax
(2003); - Aim at analyzing the optimal response
NKT - Assume sticky prices in the short run. rates rather than using surprises as a shock absorber of unexpected innovations in the
Schmitt- of government debt to shocks focusing
- Expenditures are financed by non- fiscal budget.
Grohe and on the type of the involved time-
distortionary taxes and or by issuing only - The distortions introduced by tax changes are diminished by spreading them over
Uribe inconsistency policy.
nominal risk-free debt. time which induces a near random walk property in tax rates and public debt.
(2004)
- Optimal pre-commitment policy allows debt to follow a random walk path in the
steady state.
- New Keynesian (DSGE) framework - However, under sticky prices framework, government is tempted to use its policy
- New Keynesian models augmented by
based on Christiano et al. (2005) and Smets instruments to change the steady state level debt in the initial period. Debt is curved
Leith and the government’s budget constraint.
and Wouters (2007). to initial efficient steady state to encounter this temptation and thus deter the public
Wren- - Aim at analyzing the optimal response
NKT - The model assumes sticky prices in the debt to follow random walk path if following shocks; the new steady state debt equates
Lewis of government debt to shocks involving
short run. the original (efficient) debt level despite there is no explicit debt target in the
(2013) time-inconsistency policy and its
- Public spending is financed by government’s objective function. Debt stabilization instruments depend on the degree
implications for discretionary policies.
distortionary taxes and/or by debt. of nominal inertia. The size of the debt stock and welfare consequences of introducing
debt are negligible for pre-commitment policies but can be significant for
discretionary policies.
- Large government debt in steady state impact the sign and size of short/medium run
- Derived assumptions of New Keynesian
- New Keynesian model analyzing the fiscal multipliers, depending on the horizon’s evaluation of these multipliers.
models.
responses of macroeconomic variables, - Presence of dynamic interactions between inflation and debt level in real terms
Mayer et al. - The model assumes a fraction of the
NKT in the steady state, to a government (absent in standard New Keynesian models where debt is set to zero in the steady
(2013) household sector to follow a rule-of-thumb
spending shock in the presence of state).
behavior as in Gali et al. (2007) (non-
positive levels of government debt. - In the case of permanent debt, the fiscal policy effect becomes difficult to predict
Ricardian agents).
over time.
- A conservative government may borrow more if he knew that he will be succeeded

- A principal-agent problem, aiming at
- Two-period perfect-foresight framework, by a liberal government than it would do once certain of remaining in power in the
comparing the policy of a conservative
- Assume the ruling government is certain future; A conservative government will collect less tax and leave more public debt
government, certain of been succeeded
Persson and that he will be succeeded by a liberal than what the successor would prefer.
by a liberal government, to the policy
Svensson PAD government. Assume homogeneity of - This raises the ruler government consumption than it would be if he will stay in the
where he is certain of his reelection. The
(1989) governments’ preferences towards all government, while the successor with high debt and constrained resources reduces
level of public debt is an instrument of
public goods but different preferences for consumption than it would be if he runs solely. Thus, the time-consistent level of
the current government to control the
different levels of the same public good. government consumption is somewhere in between the two outcomes of what each of
rival future government.
the two governments would prefer if ruling on his own.
252
- Assume uncertainty about the nature and - Differences in political institutions, leading to different debt policies, help explain
spending behavior of succeeding debt paths over time and across countries. Debt accumulation and deficit are
A political/positive theory of debt governments. emphasized by the alternation of the elected governments. Disagreement of different
aiming at understanding the debt build- - Assume different preferences for governments on the structure of the spending results in a deficit bias and an
up and deficits in several industrialized different items of public spending and a accumulation of debt higher than it would be in case of a social planner.
economies constant population of identical - The debt left to the last period is larger in a democracy than with a social planner.
Alesina and
Explicitly, they compare the outcome of individuals with the same time horizon. - The electoral uncertainty creates a sub-optimal deficit bias. This bias is stronger for
Tabellini PAD
debt accumulation and deficit in - Individuals differ only by their the party with the smaller probability of reappointment”. The equilibrium level of
(1990)
situation where governments alternate preferences for public goods, supplied by government debt is higher; 1/ the higher the degree of polarization between alternating
versus an outcome resulting from a the government and financed by means of governments and 2/ the more likely the current government will not be reelected.
social planner government supposed distortionary taxes on labor. Moreover, as the ruling government is unable to curve the taxation and expenditure
elected forever. -Government is elected among two parties policies of its successors, the law of motion of the public debt is the only way by
maximizing each the electorate utility which the fiscal policy of the ruling government can impact the policies of its
function. successors.
- 2007: Assume the ratio of the primary
- 2007: The model is used to derive necessary conditions for the existence of a
surplus to gross domestic income to be a
sustainable balanced growth path and analyze growth effects of deficit financed
- 2007: Endogenous growth model with positive linear function of the debt ratio.
increases in public investment.
public capital and sustainable debt. - 2012: Government levies distortionary
- 2012: The long-run growth rate is smaller the higher the debt ratio whenever public
Greiner - 2012: Endogenous growth with elastic income tax and issues bonds to finance
spending is adjusted to fulfill the government inter-temporal budget constraint.
(2007; labor supply and a government sector. lump-sum transfers and non-distortionary
EGT - 2015: Higher debt goes along with smaller long-run growth. Moreover, discretionary
2012; 2015; - 2015: Endogenous growth model with public spending.
policy violates the intertemporal government budget constraint along a balanced path.
2016) productive public spending. - 2015: Rational identical households with
- 2016: Balanced budget policy and a slight deficit policy yield higher growth than a
- 2016: An endogenous growth model perfect foresight maximizing their utility.
policy where debt and GDP grow at the same rate. For high debt and low elasticity of
with public educational spending - 2016: Same assumptions as in Greiner
substitution, a high deficit policy yields lower welfare than a balanced budget and a
(2015).
slight deficit policy.
- Derive an optimal target debt ratio that depends on the tax rates on wage income and
- Consider a stock of capital investment
- Builds on Futagami et al. (2008) consumption, and the ratio of public investment to total spending. Fiscal consolidation
in endogenous framework.
assumptions but instead consider public based on a debt ratio target rule improves welfare and this improvement is more the
Maebayashi - Examine how reducing public debts in
EGT capital not flows as input. fastest the pace of debt reduction is.
et al. (2017) the economy with large public debts
- Assume the depreciation rate of public - Fiscal consolidation based on expenditure cuts jointly with a tax increase does not
affects transition of the economy and
capital is zero. always improve welfare. In this case, the welfare gains (if any) are lower than those
welfare.
under expenditure cuts only.
- Even if debt does not affect the steady state capital stock and income, it disrupts the
- A critic of ILA and OLG models and
income distribution and consumption leading to raise inequality between spenders and
alternative theory mentioned as “Savers- - Assume the role-of-thumb behavior (non-
savers. A higher level of debt yields higher taxation to compensate for the interest
Spenders Theory of fiscal policy” to Ricardian) in the consumers spending.
payments on the debt. But taxes are on both savers and spenders while interest
address the neoclassical shortcomings - Some individuals may enjoy long lifetime
Mankiw payments on debt fall on the savers side. Thus, a higher level of debt increases the
SST considering that “households smooth horizons (due to bequest motives) while
(2000) steady sate income and consumption for savers and lowers it for the spenders which
their consumption over time” is others having short time horizons fail to
raise inequality between the two groups.
inaccurate as many consumers are far smooth their consumption and accumulate
- Many agents have no saving (data fact), so unable to smooth intertemporal
from following complete rational wealth.
consumption as reported by the ILA or OLG models. A better model would allow for
expectations.
such heterogeneous behaviors.
- A critic to the New Keynesian models - Critics based on the examination of Smets - New Keynesian models are not yet useful for public policy analysis. These models
Chari et al.
CNK based on the study of Smets and Wooters and Wooters (2007) properties and include many free parameters yielding to shocks that are dubiously structural as well
(2009)
(2007) core model. assumptions. as many features that are not consistent with microeconomic evidence.
253
Table A.2. description of variables and data sources

The IMF dataset on investment and capital stock, July 2017.
Public Investment General government investment (gross fixed capital formation), in billions of constant 2011 international dollars.
Public Capital Government capital stock (constructed based on government investment flows), in billions of constant 2011 international dollars.
Private Investment Private investment (gross fixed capital formation), in billions of constant 2011 international dollars.
Private Capital Private capital stock (constructed based on private investment flows), in billions of constant 2011 international dollars.
GDP Gross domestic product, in billions of constant 2011 international dollars.
The accompanying 2017 Update of the Manual "Estimating Public, Private, and PPP Capital Stocks"
(http://www.imf.org/external/np/fad/publicinvestment/data/info122216.pdf) and
Background
(http://www.imf.org/external/pp/longres.aspx?id=4959) describes in great detail the series' definitions, the investment series' data
material on data
sources, as well as the methodology in constructing the stock series. The methodology follows the standard perpetual inventory
construction
equation and largely builds on Gupta et al. (2014) "Efficiency-Adjusted Public Capital and Growth" and Kamps (2006) "New
Estimates of Government Net Capital Stocks for 22 OECD Countries, 1960–2001".
Information on public and private investment and GDP comes from three main sources: the OECD Analytical Database (August 2016
version) for OECD countries, and a combination of the National Accounts of the Penn World Tables (PWT, version 9.0) and the IMF
Information sources
World Economic Outlook (WEO, April 2016 ) for non-OECD countries. Information on country income groupings used in
depreciation rates' assumptions is from the World Bank World Development Indicators.
Note that all data series (public investment and capital stock, private investment and capital stock, GDP, etc.) are expressed in billions
Additional notes of constant 2011 international dollars (purchasing power parity adjusted), using the corresponding component-specific deflators from
OECD, PWT, and WB databases mentioned above.
Source of the above
"IMF Investment and Capital Stock Dataset, 2017”, drawn from http://www.imf.org/external/pp/longres.aspx?id=4959.
part of this Table
Other data from other sources
Interest rates We consider long term interest rates from OECD database statistics
Inflation GDP deflator percent change, GDP deflators are from OECD database statistics
Tax revenues Tax revenues as percent of GDP are from
Government
General government expenditures as percent of GDP are from OECD database website
expenditures
Social expenditure Social spending as percent of GDP are from OECD database website
We consider a working age population [15, 64] as a proxy for labor data from the World Development Indicators database
Labor
of the World Bank.
Human capital Measured by average years of schooling of population [15, 64] from Barro-Lee dataset (www.barrolee.com)
254
Table A.3. Descriptive statistics

Public investment (Public Gross Capital Formation) as percent of GDP
Sample: 1960
AUS AUT BEL CAN CHE DNK FRA GBR GER GRC IRL ITA JPN NLD NOR NZL PRT SPA SWE USA
2015
Mean 2.3 4.7 3.4 3.6 3.7 6.3 4.8 3.6 3.5 3.2 4.5 3.6 8.3 4.7 3.6 4.5 4.3 3.9 5.0 4.5
Median 2.2 4.4 2.9 3.5 3.7 4.6 4.5 3.2 3.1 2.8 4.6 3.5 8.8 4.1 3.5 4.6 4.4 3.9 4.7 4.2
Maximum 3.5 6.9 6.4 4.9 4.6 14.0 6.6 6.6 6.1 5.5 6.7 5.0 11.9 8.0 4.8 6.7 6.1 5.7 7.6 7.1
Minimum 1.7 2.5 2.4 2.8 2.2 3.3 3.7 1.6 1.9 1.9 2.1 2.5 4.2 3.3 2.6 2.4 2.4 2.5 3.9 3.4
Std. Dev. 0.4 1.5 1.1 0.6 0.5 3.4 0.7 1.4 1.4 1.1 1.2 0.6 2.3 1.4 0.5 0.9 0.9 0.7 1.1 1.0
Skewness 1.0 0.1 1.3 0.4 -0.6 0.9 0.9 0.7 0.4 0.8 -0.1 0.5 -0.5 1.0 0.4 -0.5 -0.3 0.3 1.0 1.4
Kurtosis 3.7 1.5 3.4 1.9 3.7 2.2 2.9 2.1 1.7 2.3 2.0 2.8 2.2 2.6 3.1 3.0 2.5 3.2 3.0 3.8
Jarque-Bera 10.7 5.5 15.6 4.5 4.0 8.7 7.6 5.9 5.8 6.6 2.5 2.8 3.7 10.3 1.6 2.1 1.3 1.1 8.7 19.9
Probability 0.0 0.1 0.0 0.1 0.1 0.0 0.0 0.1 0.1 0.0 0.3 0.3 0.2 0.0 0.5 0.4 0.5 0.6 0.0 0.0
Private investment (Private Gross Capital Formation) as percent of GDP
Mean 14.9 22.2 23.7 16.3 21.9 16.8 18.9 16.4 17.8 13.2 25.1 22.5 17.9 18.1 19.3 11.7 21.0 20.1 17.5 14.3
Median 14.3 22.4 23.9 16.0 22.0 17.1 18.9 16.4 17.4 13.4 24.5 21.7 18.0 18.0 19.6 11.3 21.0 19.5 17.5 13.8
Maximum 22.2 24.3 30.0 20.8 24.2 23.9 21.9 21.4 21.5 19.0 34.2 31.0 22.9 22.1 27.0 15.2 27.6 28.4 21.0 17.9
Minimum 11.1 19.1 16.2 12.7 17.2 10.7 16.9 13.1 16.0 6.7 16.5 17.3 12.4 15.3 12.8 8.8 14.9 14.9 13.3 11.1
Std. Dev. 3.0 1.5 3.1 2.1 1.6 3.5 1.2 1.9 1.5 2.6 3.9 3.4 2.3 1.6 3.5 1.8 3.0 3.5 1.7 1.9
Skewness 0.9 -0.4 -0.5 0.5 -0.9 0.2 0.4 0.4 0.9 -0.4 0.3 0.9 0.1 0.4 0.0 0.4 0.0 0.8 -0.3 0.4
Kurtosis 2.7 2.0 3.3 2.4 3.7 1.9 2.8 2.9 2.7 3.1 2.7 3.0 3.0 2.7 2.3 2.2 2.8 3.0 2.8 2.1
Jarque-Bera 7.5 3.8 2.4 2.9 9.0 3.0 1.3 1.7 6.9 1.5 0.9 7.8 0.0 1.5 1.2 3.2 0.2 5.2 1.0 3.5
Probability 0.0 0.2 0.3 0.2 0.0 0.2 0.5 0.4 0.0 0.5 0.6 0.0 1.0 0.5 0.6 0.2 0.9 0.1 0.6 0.2
Government capital stock in billions of 2011 constant US Dollars (ppp)
KG AUS AUT BEL CAN CHE DNK FRA GBR GER GRC IRL ITA JPN NLD NOR NZL PRT SPA SWE USA
Mean 163 188 157 464 174 221 1237 980 1490 100 54 904 3622 376 94 62 118 525 221 6752
Median 147 209 192 435 178 228 1233 1025 1671 80 50 907 3779 382 85 66 102 460 232 6229
Maximum 364 237 235 870 275 261 1921 1254 1777 216 124 1386 5991 553 198 105 219 1056 314 10889
Minimum 54 88 44 184 49 132 463 536 726 29 13 376 716 178 26 28 40 168 91 3276
Std. Dev. 82 48 66 189 74 34 459 182 333 61 34 319 1904 103 51 20 59 281 66 2257
Skewness 0.76 -0.77 -0.55 0.46 -0.25 -1.17 -0.12 -1.04 -1.11 0.74 0.81 -0.05 -0.19 -0.16 0.39 0.16 0.42 0.53 -0.51 0.37
Kurtosis 2.82 2.11 1.68 2.31 1.73 3.50 1.75 3.23 2.76 2.18 2.57 1.75 1.52 2.34 1.95 2.42 1.76 2.02 2.12 1.96
Jarque-Bera 5.41 7.38 6.96 3.08 4.32 13.4 3.76 10.17 11.68 6.69 6.56 3.67 5.45 1.26 3.97 1.02 5.20 4.86 4.24 3.85
Probability 0.07 0.02 0.03 0.21 0.12 0.00 0.15 0.01 0.00 0.04 0.04 0.16 0.07 0.53 0.14 0.60 0.07 0.09 0.12 0.15
255
Table A.3. (continued): Descriptive statistics

Private capital stock in billions of 2011 constant US Dollars (ppp)
KP AUS AUT BEL CAN CHE DNK FRA GBR GER GRC IRL ITA JPN NLD NOR NZL PRT SPA SWE USA
Mean 710 534 721 1326 709 282 3036 2597 4560 294 205 3639 5034 895 353 103 380 1781 487 12793
Median 636 533 687 1315 731 266 3227 2718 4670 293 180 3879 5163 905 395 98 364 1608 502 12126
Maximum 1575 748 992 2420 909 469 4331 3670 5742 491 433 4686 7997 1275 515 168 585 3298 669 22156
Minimum 234 243 359 434 374 115 1144 1113 2402 118 54 1839 941 383 155 48 161 649 254 4427
Std. Dev. 375 164 189 596 158 114 1027 844 1046 113 122 868 2622 276 105 34 136 849 115 5845
Skewness 0.70 -0.21 -0.19 0.16 -0.57 0.26 -0.45 -0.33 -0.52 0.12 0.69 -0.64 -0.26 -0.29 -0.46 0.36 0.04 0.47 -0.36 0.28
Kurtosis 2.51 1.74 2.01 1.88 2.17 1.73 1.91 1.63 2.02 1.81 2.27 2.11 1.50 1.94 2.00 2.14 1.71 1.94 2.18 1.77
Jarque-Bera 5.16 4.15 2.63 3.16 4.64 4.36 4.64 5.41 4.76 3.46 5.63 5.73 5.87 3.42 4.30 2.94 3.89 4.72 2.77 4.24
Probability 0.08 0.13 0.27 0.21 0.10 0.11 0.10 0.07 0.09 0.18 0.06 0.06 0.05 0.18 0.12 0.23 0.14 0.09 0.25 0.12
GDP in billions of 2011 constant US Dollars (ppp)
Mean 519 233 290 850 292 174 1588 1514 2339 212 100 1495 2945 481 188 86 179 908 261 9258
Median 461 218 278 831 283 169 1565 1459 2260 207 66 1595 3186 437 179 76 174 839 252 8579
Maximum 1074 381 465 1538 454 246 2503 2538 3618 365 286 2192 4568 789 335 157 289 1590 446 16940
Minimum 165 85 111 275 139 87 540 696 973 57 23 501 575 177 60 38 49 241 112 3212
Std. Dev. 269 93 108 380 86 50 608 570 803 83 74 532 1331 195 88 34 80 412 94 4272
Skewness 0.54 0.10 0.06 0.25 0.22 -0.02 -0.07 0.30 -0.07 -0.06 0.78 -0.41 -0.36 0.18 0.19 0.59 -0.08 0.21 0.36 0.31
Kurtosis 2.08 1.77 1.83 1.86 2.11 1.65 1.79 1.75 1.72 2.29 2.20 1.83 1.67 1.73 1.67 2.11 1.60 1.81 2.02 1.72
Jarque-Bera 4.72 3.61 3.25 3.61 2.32 3.88 3.47 4.52 3.84 1.20 7.17 4.76 5.33 4.07 4.48 5.09 4.62 3.68 3.43 4.70
Probability 0.09 0.16 0.20 0.16 0.31 0.14 0.18 0.10 0.15 0.55 0.03 0.09 0.07 0.13 0.11 0.08 0.10 0.16 0.18 0.10
Working age population (15-65) in millions
Mean 10.88 5.12 6.56 17.95 4.51 3.39 36.96 37.33 52.68 6.55 2.26 37.08 79.11 9.70 2.74 2.20 6.33 25.47 5.52 162.00
Median 10.91 5.17 6.67 18.25 4.50 3.44 38.24 37.28 53.51 6.61 2.12 38.35 81.06 10.15 2.71 2.15 6.51 25.45 5.43 161.00
Maximum 15.79 5.79 7.31 24.32 5.57 3.64 42.13 41.87 55.88 7.42 3.11 39.04 87.13 11.15 3.41 3.00 7.07 31.76 6.18 212.00
Minimum 6.29 4.59 5.90 10.49 3.50 2.94 29.09 34.11 48.84 5.47 1.64 32.85 59.31 7.01 2.26 1.39 5.34 19.62 4.93 109.00
Std. Dev. 2.78 0.41 0.41 4.19 0.55 0.21 3.96 2.31 2.61 0.73 0.50 2.05 7.59 1.32 0.32 0.48 0.61 3.90 0.35 31.69
Skewness 0.05 0.00 0.03 -0.21 0.15 -0.52 -0.39 0.57 -0.30 -0.17 0.46 -0.69 -0.91 -0.59 0.47 0.06 -0.35 0.13 0.44 0.00
Kurtosis 1.88 1.52 2.15 1.90 2.07 2.06 1.91 2.19 1.46 1.38 1.88 1.87 2.85 1.95 2.31 1.84 1.59 1.78 2.20 1.82
Jarque-Bera 2.95 5.14 1.69 3.25 2.23 4.56 4.21 4.52 6.35 6.42 4.95 7.38 7.74 5.85 3.14 3.18 5.82 3.63 3.34 3.23
Probability 0.23 0.08 0.43 0.20 0.33 0.10 0.12 0.10 0.04 0.04 0.08 0.02 0.02 0.05 0.21 0.20 0.05 0.16 0.19 0.20
256
public debt
Table A.4. Countries sample and their corresponding alpha-3 codes used in all Tables
and figures whenever it applies
Country Country code

Australia AUS
Austria AUT
Belgium BEL
Canada CAN
Switzerland CHE
Denmark DNK
France FRA
Germany GER
Greece GRC
Ireland IRL
Italy ITA
Japan JPN
Netherlands NLD
New Zealand NZL
Norway NOR
Portugal PRT
Spain SPA
Sweden SWE
United Kingdom GBR
United States USA
257
public debt
Appendix B. Figures
258
public debt
Figure B.1. Scatter plot elastictities of public investment fLows versus private
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.5 0.0 0.5 1.0 1.5 2.0 2.5
-0.1
-0.2
-0.3
-0.4 Trendline equation: y = -0.2013x + 0.3334

R² = 0.38
Figure B.2. Scatter plot elastictities of public investment flows versus private
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.5 0.0 0.5 1.0 1.5 2.0 2.5
-0.1
-0.2 Trendline equation: y = -0.1335x + 0.2863

R² = 0.21
259
public debt
Figure B.3. Simulated public investment flows as percent of GDP
G_CYC_AUS G_CYC_AUT G_CYC_BEL G_CYC_CAN

6.5 10 7 8
6.0
6 7
8
5.5
5 6
5.0 6
4 5
4.5
4
3 4
4.0
3.5 2 2 3
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
G_CYC_CHE G_CYC_DNK G_CYC_FRA G_CYC_GBR

8 14 10 7
7 12
6
8
6 10
5
5 8 6
4
4 6
4
3 4 3
2 2 2 2
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
G_CYC_GER G_CYC_GRC G_CYC_IRL G_CYC_ITA

8 12 9 10
7 8
8
8
6 7
6
5 4 6
4
4 5
0
2
3 4
2 -4 3 0
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
G_CYC_JPN G_CYC_NLD G_CYC_NOR G_CYC_NZL

16 9 6.0 7.0
14 8 5.6 6.5
5.2
12 7 6.0
4.8
10 6 5.5
4.4
8 5 5.0
4.0
6 4 3.6 4.5
4 3 3.2 4.0
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
G_CYC_PRT G_CYC_SPA G_CYC_SWE G_CYC_USA

10 12 9 9
8
10 8
8
7
8 7
6 6
6 6
5
4
4 5
4
2 2 4 3
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
260
public debt
Figure B.4. Actual public investment flows as percent of GDP
IG_AUS/Y_AUS*100 IG_AUT/Y_AUT*100 IG_BEL/Y_BEL*100 IG_CAN/Y_CAN*100

3.6 7 7 5.0
3.2 6 6 4.5
2.8 5 5 4.0
2.4 4 4 3.5
2.0 3 3 3.0
1.6 2 2 2.5
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
IG_CHE/Y_CHE*100 IG_DNK/Y_DNK*100 IG_FRA/Y_FRA*100 IG_GBR/Y_GBR*100

5.0 16 7 7
4.5 6
12 6
4.0 5
3.5 8 5 4
3.0 3
4 4
2.5 2
2.0 0 3 1
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
IG_GER/Y_GER*100 IG_GRC/Y_GRC*100 IG_IRL/Y_IRL*100 IG_ITA/Y_ITA*100

7 6 7 5.0
6 4.5
5 6
5 4.0
4 5
4 3.5
3 4
3 3.0
2 3
2 2.5
1 1 2 2.0
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
IG_JPN/Y_JPN*100 IG_NLD/Y_NLD*100 IG_NOR/Y_NOR*100 IG_NZL/Y_NZL*100

12 9 5.0 7
8
4.5 6
10
7
4.0 5
8 6
3.5 4
5
6
4 3.0 3
4 3 2.5 2
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
IG_PRT/Y_PRT*100 IG_SPA/Y_SPA*100 IG_SWE/Y_SWE*100 IG_USA/Y_USA*100

7 6 8 8
6 7 7
5
5 6 6
4
4 5 5
3
3 4 4
2 2 3 3
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
261
public debt
Figure B.5. Simulated and actual public investment flows as percent of GDP
7 9 7 8
6 8 7
6
7
5 6
5
6
4 5
5
4
3 4
4
3
2 3
3
1 2 2 2
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
G_ACT_AUS G_CYC_AUS G_ACT_AUT G_CYC_AUT G_ACT_BEL G_CYC_BEL G_ACT_CAN G_CYC_CAN
8 16 10 7
7 14 9 6
12 8
6 5
10 7
5 4
8 6
4 3
6 5
3 4 4 2
2 2 3 1
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
G_ACT_CHE G_CYC_CHE G_ACT_DNK G_CYC_DNK G_ACT_FRA G_CYC_FRA G_ACT_GBR G_CYC_GBR
8 12 9 9
7 10 8 8
7
6 8 7
6
5 6 6
5
4 4 5
4
3 2 4
3
2 0 3 2
1 -2 2 1
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
G_ACT_GER G_CYC_GER G_ACT_GRC G_CYC_GRC G_ACT_IRL G_CYC_IRL G_ACT_ITA G_CYC_ITA
16 9 6.0 7
14 8 5.5
6
5.0
12 7
5
4.5
10 6
4.0
4
8 5
3.5
3
6 4 3.0
4 3 2.5 2
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
G_ACT_JPN G_CYC_JPN G_ACT_NLD G_CYC_NLD G_ACT_NOR G_CYC_NOR G_ACT_NZL G_CYC_NZL
10 12 9 9
9
8 8
10
8
7 7
7
8
6 6 6
6
5
5 5
4
4
4 4
3
2 2 3 3
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
G_ACT_PRT G_CYC_PRT G_ACT_SPA G_CYC_SPA G_ACT_SWE G_CYC_SWE G_ACT_USA G_CYC_USA
262
Figure B.6. simulated potential public debt calibrating differential interest rate and economic growth to the ones of the United States
240 350 360 320 280
300 320 280 240

200
280
250 240
200
160 240
200 200
200 160
120 150 160
160 120
100 120
80 120
80
50 80
80
40
0 40 40
40
0 -50 0 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_BEL D_BEL D30_BEL ACTUAL_ GRC D_ GRC D3 0 _GRC ACTUAL_IRL D_IRL D30_IRL ACTUAL_PRT D_ PRT D30_PRT ACTUAL_ AUT D_AUT D30_AUT
240 300 300 300 500
200 250 250 250

400
160 200 200 200

300
120 150 150 150
200
80 100 100 100
100
40 50 50 50
0 0 0 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_ GER D_ GER D3 0_ GER ACTUAL_FRA D_FRA D30_FRA ACTUAL_ITA D_ITA D30_ITA ACTUAL _ NL D D_ NL D D30 _ NL D ACTUAL_SPA D_SPA D30_SPA
300 500 240 280 240
250 200 240 200

400
200 160 200 160

300
150 120 160 120
200
100 80 120 80
100
50 40 80 40
0 0 0 40 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_ USA D_ USA D30_USA ACTUAL_J PN D_J PN D30_J PN ACTUAL _GBR D_ GBR D3 0 _GBR ACTUAL _ CAN D_ CAN D30 _ CAN ACTUAL_ AUS D_AUS D30_AUS
240 500 240 200 300
200 200 250

400 160
160 160 200

300 120
120 120 150
200 80
80 80 100
100 40
40 40 50
0 0 0 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_ NZL D_ NZL D30_NZL ACTUAL _DNK D_ DNK D30 _DNK ACTUAL _CHE D_ CHE D3 0 _CHE ACTUAL_ NOR D_ NOR D3 0 _ NOR ACTUAL_SWE D_SWE D30_SWE
263
Figure B.7. Actual and simulated potential public debt corresponding to financing the potential public capital and an extra 20% of actual tax revenues
140 250 240 300 350
120 200 300

200 250
100 150 250

160 200
80 100 200
120 150
60 50 150
80 100
40 0 100
40 50
20 -50 50
0 -100 0 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_BEL D_BEL D30_BEL ACT UAL _ GRC D_ GRC D3 0 _ GRC ACTUAL_IRL D_IRL D30_IRL ACTUAL_ PRT D_ PRT D30_PRT ACTUAL_ AUT D_ AUT D30_AUT
200 180 240 240 240
160
200
200 200
160
140
160
120 160 160
120
100 120
120 120
80 80
80
60 80 80
40
40 40
40 40
0
20
0 0 -40 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL _ GER D_ GER D3 0 _ GER ACTUAL_ FRA D_ FRA D30_FRA ACTUAL_ITA D_ITA D30_ITA ACTUAL _ NL D D_ NL D D3 0 _ NL D ACTUAL_SPA D_SPA D30_SPA
280 400 120 180 100
240 200 160

100
80
0
200 140
80
-200
60
160 120
-400 60
120 100
40
-600
40
80 80
-800
20
20
40 -1,000 60
0 -1,200 0 40 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_ USA D_ USA D30_USA ACTUAL_ J PN D_ J PN D30_J PN ACTUAL _ GBR D_ GBR D3 0 _ GBR ACTUAL _ CAN D_ CAN D3 0 _ CAN ACTUAL_ AUS D_ AUS D30_AUS
80 300 400 120 200
70 350
250 100
160
300
60
200 80
250
120
50
150 200 60
40
80
150
100 40
30
100
40
20 50 20
50
10 0 0 0 0
60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15 60 65 70 75 80 85 90 95 00 05 10 15
ACTUAL_ NZL D_ NZL D30_NZL ACTUAL _ DNK D_ DNK D3 0 _ DNK ACTUAL _ CHE D_ CHE D3 0 _ CHE ACT UAL _ NOR D_ NOR D3 0 _ NOR ACTUAL_SWE D_SWE D30_SWE
264
General Conclusion
General Conclusion
The aim of this thesis was to examine the relationship between public debt and economic growth,
particularly the impact of the former on the latter. Although this issue was mainly debated in the
post-war era, leading to a variety of theoretical models involving the long-term effects of public
debt, as well as the short-term effects, the debate about this subject was revived for the advanced
countries following the 2008 crisis. This thesis re-examined, first, the validity of Reinhart and
Rogoff’s (2010) conjecture of a common debt threshold, which stated that debt to GDP ratios
above 90% reduce economic growth for a sample of advanced countries, thus creating a
controversial debate among economists and policy-makers. Second, it assessed the short-term and,
third, long-term effects of public debt on economic growth.
At this level, the methodology used the most recent advanced econometrics methods, focusing on
the relationship between the two variables, based purely on the data-generating process and
comparing it to a large body of empirical literature on this issue that flourished after the 2008 crisis
(Chapter 1). Then, the relationship was analysed considering the economic channels through which
public debt acts on economic growth, first, in the short term (Chapter 2) and, second, in the long
term (Chapter 3). In the short term, as public debt is generated to finance government expenditure
(including government investments), the public debt and economic growth linkages were
examined by adopting a fiscal policy assessment framework invoking the expenditure multipliers.
In the long term, public debt and economic growth are linked by adopting a theoretical framework
of endogenous growth models.
The principal conclusions of our thesis lead us to reject the common debt ratio that applies to all
countries for a sample of 20 advanced countries. It shows, specifically, that the relationship
between public debt and economic growth is rather time- and country-specific. Furthermore, in a
short-term framework, public debt accumulation and contraction have different impacts on
economic growth through expenditure multipliers. The latter are also revealed in our thesis to be
sensitive to other factors related to the business cycle (expansion, recession), the public debt ratio
development (accumulation, contraction) and the coexistence of cross-cutting of these factors. In
the long term, public debt is linked, in the context of the endogenous growth framework, to the
macroeconomic fundamentals and public expenditure’s productivity.
Specifically, Chapter 1 made important contributions to the empirical debate about a common
public debt threshold that applies to all countries. Without claiming exhaustivity, this chapter
summarized a large survey covering the main body of important empirical literature
econometrically studying the relationship between public debt and economic growth, following
the controversial paper of Reinhart and Rogoff (2010). Furthermore, the empirical contribution of
Chapter 1 focused particularly on analysing a sample of 20 advanced economies, including 10 euro
countries, considering a long period of analysis that goes back to 1880. In the first step, we ran a
variety of panel regressions for an economic growth function as a quadratic polynomial of the
265
General Conclusion
public debt ratio following Checherita and Rother (2010). In the second step Chapter 1 addressed
the debt–growth link with a novel and innovative approach for individual countries using the
Hansen (2017) “regression kink” method. This method has the power to detect endogenously the
turning point between explained and explanatory variables compared to models fixing the
exogenous threshold.
For the panel regressions, we used appropriate methods of estimations and tests to deal with several
econometric issues such as endogeneity, sample bias, heterogeneity, non-linearity and reverse
causality. Panel estimations were conducted varying the country sample, as well as considering
sub-periods of time, by break points following world events occurring in the international system.
All these analyses show that the debt–growth nexus changes widely over time and by sample
countries. Furthermore, the revealed relationship is altered by data and cross-country
heterogeneity. The heterogeneity decreases when omitting sizeable countries from the sample and
is less manifested in the sample of 10 euro countries.
Applied to the debt and growth relationship by country, the regression kink method sorted out a
variety of different curves depending on the country. Therefore, some countries could grow with
high debt to GDP ratios, others could see their growth shrink from low debt ratios, while growth
in others was insensitive to public debt. The study also reveals the instability of the relationship
over time. Every country exhibits a different relationship by period, especially when the transition
is between periods known for specific changes in the international monetary system.
All of the results in Chapter 1, whether panel regressions or Hansen (2017) regression kink, stand
in contrast to the existence of a common threshold and its 90% value, as claimed by Reinhart and
Rogoff (2010). It not only argues against the 90% threshold value but also clearly rejects any
common threshold. The threshold, whenever it exists, is country-specific rather than following a
common rule to fit all. Consequently, as the public debt and economic growth relationship is
country-specific (one size does not fit all), the main recommendation of Chapter 1 is for more
investigation using models based on theory to study the effects of public debt on economic growth
considering the economic properties and fundamentals.
Chapter 2 was dedicated, in particular, to examining the relationship between public debt and
economic growth in the short term, through Keynesian multipliers. Specifically, we studied the
impact of public debt accumulation on the size of the expenditure multipliers, as well as the effects
of the business cycle. We adopted, in this chapter, the structural vector autoregressive (SVAR)
methodology, running a panoply of bi-variates and tri-variate SVAR models on quarterly data for
a sample of 18 OECD countries with different exchange rate regimes, monetary policies and
degrees of openness. Besides, we estimated an SVAR with six fiscal and monetary variables for
the United States to highlight the channels by which public debt affects economic growth. In all
the above-mentioned SVAR models, we controlled for business cycle and public debt movement
effects to deduce spending multipliers for these cases.
266
General Conclusion
The results show that, controlling for business cycle effects, the expenditure multipliers are much
higher in times of recession than times of expansion. Moreover, our results do not support the
general tendency of spending multipliers being weak, and even negative, as revealed by some
researchers, in recent decades compared to the 1960s, 1970s and 1980s. The previous results are
in line with what was observed in the recent literature regarding fiscal multipliers, in advanced
economies, namely, being large in times of recession and weak, or even negative, in times of
expansion. Furthermore, controlling exogenously for public debt movements, independent of the
business cycle, it was revealed that spending multipliers are larger in periods of debt accumulation
than in debt contraction periods. Also, jointly controlling exogenously for debt movements and
the business cycle reveals that multipliers are higher under debt accumulation in cases of expansion
and recession. However, introducing endogenously the public debt to GDP variable in an SVAR
leads to higher multipliers in recessions than expansions.
Furthermore, the main results of the United States model show that government expenditure has
positive but short-lived effects on economic growth. Furthermore, the public debt crowds out
private investment, leading to a lowering growth rate in times of expansion, while in times of
recession the public debt effects on growth are positive. This crowding-out effect may play pass-
through to the expenditure multipliers and could explain, ceteris paribus, the weak size of
spending multipliers, while in times of recession the crowing-in effect leads to higher multipliers.
In the panoply of SVAR models considered in Chapter 2, it was noted that recession periods
generally have a persistent effect on variables for which convergence to the long-term path
following the shock is achieved more slowly compared to expansion time. The policy implication
of this is that fiscal stimulus effects could take time to materialize in depressed economies, while
the effects could be short-lived in times of expansion, which should be considered by policy-
makers in their spending decisions.
As the public debt and growth relationship also involves long-term effects, Chapter 3 extends the
investigation of the public debt effects in relation to the role of theory-based models of public debt
on long-term economic growth, considering country-specific fundamentals. Therefore, an
endogenous growth model was adopted for the same sample studied in Chapter 1, constituted by
20 developed countries. Specifically, a Cobb-Douglas production function was augmented by the
stock of public capital and linked to a public debt equation through the productivity of public
capital. Therefore, theoretical formulae and empirical simulations were derived based on this
framework. For the theoretical results, we concluded that, in an endogenous growth framework,
with the Cobb-Douglas production function encompassing public capital instead of government
flows, the potential productive government investment is an endogenous parameter of economic
growth and public capital productivity.
Furthermore, we simulated a parameterized formula for a potential/limit debt that a country could
target to finance its productive investments. This potential, linked to economic growth and public
267
General Conclusion
capital productivity, as well as the interest rate, is dynamic, country- and time-specific and has a
tendency to evolve in contrast to the actual accumulated public debt. This, in particular, responds
to the main recommendation made in Chapter 1, that the public debt threshold is time- and country-
specific, which should be studied considering a country’s economic fundamentals. It also responds
to the caveats of the research highlighted in Chapter 1, which set and tested, in particular, the
exogenous values of thresholds.
Moreover, the results of Chapter 3 show that the simulated public debt drops under the actual
levels of debt in times of crisis for many advanced countries that are severely affected by the crisis.
This has a clear message regarding policy recommendation that countries are safer from threats of
public debt as long as the potential debt (simulated/targeted) stays higher than the actual public
debt. The results show that many countries are under stress of public debt, especially after the
financial crisis of 2008. For some countries, such as Ireland, this stress is short-lived, and the
potential debt quickly remains higher than the actual one following economic recovery. However,
for many other countries, the effects are prolonged. The potential debt also revealed that some
countries (Greece, Belgium and Italy) had debt issues in periods before the 2008 crisis.
The added value of Chapter 3 is that the debt limit is an endogenous parametrized function linked
to the economic returns (productivity) of public capital financed by such issued debt. The potential
public expenditure series shows a decreasing trend over time, albeit still higher than the observed
one in many countries of the sample. This is driven by the general decreasing trend of public capital
stock productivity, especially in the great moderation era (1986–2007). The policy message to
increase potential public expenditure, and hence the targeted potential debt, is to enhance such
productivity. To rephrase Krugman’s (1994) statement about the importance of productivity in the
long term, “productivity isn’t everything, but in the long run it is almost everything”. Productivity
enhancement could be achievable through the choice of higher productive public capital, which
implies selective public capital projects with higher output multipliers.
Across the three chapters, the thesis counters any common debt ratio threshold that applies to all
countries. The subject of growth and public debt remains one of the most debated subjects in fiscal
policy after the 2008 financial crisis. This importance will continue, and we expect more research
about it in the medium and long term in relation to many current events. Indeed, the current sanitary
crisis of 2020, qualified by many politicians as a time of war because of the harsh measures of
emergencies and confining and locking down populations, leads to hibernation of almost all
economic sectors for months. Consequently, all the experts and national and international
forecasting institutions are at least unanimous about the inevitable recessions, if not prolonged
depressions, in the world and in almost all economies of advanced, emergent and developing
countries. All government financial measures (monetary and fiscal), which are likely to deepen the
deficits and increase public debt levels, once combined with a shrinkage in GDP, lead to soaring
ratios of public debt. In this way, some economists expect public debt ratios to jump for many
countries by 20 to 40 percentage points of GDP. Historical examples, according to Kose et al.
268
General Conclusion
(2020), show similar differences in accumulated debt to GDP ratios, comparing before and after
the crises, where, for example, it reached approximately 31% and 35%, respectively, for Indonesia
and Thailand in the 1997 financial crisis, and around 27% and 38%, respectively, for Latvia and
Ireland in the 2008 financial crisis. The recovery and economic growth levels, which are highly
uncertain, are conditioned by the prolonged measures, depending on the evolution of the Covid-
19 pandemic and the development of a vaccine.
In the current trend, fiscal policy, which was subordinated to monetary policy in the Great
Moderation (1986–2007) has, since the 2008 crisis, become active and is leading the scene. The
Covid-19 crisis will lead to greater involvement of fiscal policy; in particular, the conventional
monetary policy has already reached its limit in using the conventional framework, particularly in
many advanced economies, where nominal interest rates are reaching, or near to reaching, the zero
lower bound. Although some views postulate that the trend of low interest rates is in favour of
reduced public debt costs, which seems to point to a sustainable public debt path around higher
levels, the threat of reverting the interest rates path in response to expected inflation in the future
in an environment of low economic growth would plunge many highly-indebted countries into
other waves of unsustainable and sovereign debt crisis.
Therefore, research on fiscal policy, and particularly the public debt overhang issues and economic
growth, is likely to remain hotly debated for a long time to come. Many questions are, hence, worth
raising regarding public debt and economic growth in the post-Covid-19 economic crisis. These
include studying many conflictual objectives of fiscal policy aimed at saving jobs, as well as
rescuing firms from bankruptcy, and raising growth, hence, countries’ revenues in the short term,
while dealing with the subsequent effects of raising public debt, particularly in the medium to long
term. Moreover, the nature of the current economic crisis is likely to raise many issues in economic
policies. An example revealed by the 2008 crisis is that fiscal multipliers are sensitive to the
business cycle and many other determinants related to countries’ economic fundamentals. The
current crisis is qualified as a dual shock of demand and supply shocks, which may have an
important impact on the global macroeconomic framework, and particularly on the fiscal policy
framework. Therefore, modelling public debt and fiscal policy effects may fall outside the regular
traditional framework. At the country level, many challenges and coordination issues may seem to
be highly solicited, and their effects should be assessed, especially across the eurozone.
269
References
References
Abbas A., Basdevant O., Eble S., Everaert G., Gottschalk J., Hasanov F., Park J., Sancak C.,
Velloso R. and Villafuerte M. 2010. “Strategies for Fiscal Consolidation in the Post-Crisis World”,
IMF Fiscal Affairs Department, International Monetary Fund.
Abbas S. A., Belhocine N., ElGanainy A. and Horton M. 2010, “A Historical Public Debt
Database”, IMF working paper 10-245, International Monetary Fund.
Acemoglu D. and Robinson J. A. 2010. “The Role of Institutions in Growth and Development”,
Review of Economics and Institutions, vol. 1(2), pp. 1-33.
Acemoglu D., Johnson S. and Robinson J. A. 2001. “The Colonial Origins of Comparative
Development: An Empirical Investigation”, American Economic Review, vol. 91(5), pp. 1369-
1401.
Afonso A. and Leal S. 2018. “Fiscal Multipliers in the Eurozone: A SVAR analysis”, REM
working paper 047.
Ahiakpor J. C. W. 2013. “the Modern Ricardian Equivalence Theorem: Drawing the Wrong
Conclusions from David Ricardo’s Analysis”, Journal of the History of Economic Thought, vol.
35(1), pp. 77-92.
Aiyagari S. R. 1985. “Deficits, interest rates, and the tax distribution”, Quarterly Review, Federal
Reserve Bank of Minneapolis, vol. 9(1), pp. 5-14.
Aiyagari S. R. 1989. “Equilibrium existence in an overlapping generations model with altruistic

preferences”, Journal of Economic Theory, vol. 47(1), pp. 130-152.
Aiyagari S. R., Marcet A., Sargent, T. and Seppala J. 2002. “Optimal Taxation without State-
Contingent Debt”, Journal of Political Economy, vol. 110(6), pp. 1220-1254.
Aiyar S. and Feyrer J. 2002. “A Contribution to the Empirics of Total Factor Productivity”,
Dartmouth College working paper.
Aizenman J. and Pasricha G. K. 2011. “Net Fiscal Stimulus During the Great Recession”, NBER
working paper 16779, National Bureau of Economic Research.
Alesina A. and Passalacqua A. 2015. “The Political Economy of Government Debt”, NBER
Alesina A. and Perotti R. 1994. “The Political Economy of Budget Deficits”, NBER working paper
4637, National Bureau of Economic Research.
270
References
Alesina A. and Tabellini G. 1990. “A positive theory of fiscal deficits and government debt”,
Review of Economic Studies, vol. 57(3), pp. 403-414.
Alesina A., Favero C. and Giavazzi F. 2015. “The output effect of fiscal consolidation plans”,
Journal of International Economics, vol. 96, pp. 19-42.
Alichi A., Shibata I. and Tanyeri K. 2019. “Fiscal Policy Multipliers in Small States”, IMF
working paper 19/72, International Monetary Fund.
Ando M. 2017. “How much should we trust regression-kink-design estimates?”, Empirical

Economics, vol. 53, pp. 1287-1322.
Aoki M. 2007. “Endogenizing institutions and institutional changes”, Journal of Institutional

Economics, vol. 3(1), pp. 1-31.
Arin K. P., Koray F. and Spagnolo N. 2015. “Fiscal multipliers in good times and bad times”,
Journal of Macroeconomics, vol. 44, pp. 303-311.
Arrow K. J. 1962. “The Economic Implications of Learning by Doing”, Review of Economic

Studies, vol. 29, pp. 155-173.
Arrow K. J. and Kurz M. A. 1970. “Public Investment, the Rate of Return, and Optimal Fiscal
Policy”, Baltimore, Johns Hopkins University Press.
Aschauer D. 2000. “Do states optimize? Public capital and economic growth”, The Annals of
Regional Science, vol. 34(3), pp. 343-363.
Athey S. and Imbens G. W. 2017. “The State of Applied Econometrics: Causality and Policy
Evaluation”, The Journal of Economic Perspectives, vol. 31(2), pp. 3-32.
Auerbach A. J. and Gorodnichenko Y. 2012. “Measuring the Output Responses to Fiscal Policy”,
American Economic Journal: Economic Policy, vol. 4, pp. 1-27.
Auerbach A. J. and Gorodnichenko Y. 2013. “Fiscal Multipliers in Recession and Expansion” in

Fiscal Policy after the Financial Crisis, Alberto Alesina and Francesco Giavazzi, eds., University
of Chicago Press.
Auerbach A. J. and Gorodnichenko Y. 2017. “Fiscal Stimulus and Fiscal Sustainability”, NBER
Auerbach A. J., Gorodnichenko Y. and Murphy D. 2019. “Local Fiscal Multipliers and Fiscal
Spillovers in the United States”, IMF Economic Review, vol. 68, pp. 195-229.
271
References
Aussilloux V., Gouardo C. and Lenglart F. 2018. “Baisser le poids des dépenses publiques : les
leçons de l’expérience des pays européens”, Note d’analyse 74, France Stratégie.
Bachmann R. and Sims E. R. 2012. “Confidence and the transmission of government spending
shocks”, Journal of Monetary Economics, vol. 59, pp. 235-249.
Baglan D. and Yoldas E. 2013. “Government debt and macroeconomic activity: a predictive
analysis for advanced economies”, Finance and Economics Discussion Series, 05, Federal Reserve
Board.
Baldacci E., Gupta S. and Mulas-Granados C. 2009. “How effective is fiscal policy response in
systemic banking crises?”, IMF working paper 09/160, International Monetary Fund.
Barnichon R. and Matthes C. 2017. “Understanding the Size of the Government Spending
Multiplier: It’s in the Sign”, Federal Reserve Bank of Richmond, Working Paper No. 17-15.
Barrell R., Holland D. and Hurst I. 2012. “Fiscal Consolidation: Part 2. Fiscal Multipliers and
Fiscal Consolidations,” OECD Economics Department Working Paper No. 933, OECD, Paris.
Barro R. J. 1974. “Are Government Bonds Net Wealth?”, Journal of Political Economy, vol. 82(6),
pp. 1095-1117.
Barro R. J. 1976. “Perceived wealth in bonds and social security and the Ricardian equivalence
theorem: Reply to Feldstein and Buchanan”, Journal of Political Economy, vol. 84(2), pp. 343-
350.
Barro R. J. 1979. “On the determination of the public debt”, Journal of Political Economy, vol.
87(5), pp. 940-971.
Barro R. J. 1990. “Government spending in a simple model of endogeneous growth”, Journal of

Political Economy, vol. 98(S5), pp. 103-125.
Barro R. J. 1996. “Democracy and Growth”, Journal of Economic Growth, vol. 1(1), pp. 1-27.
Barro R. J. and Lee J. W. 1993. “International comparisons of educational attainment”, Journal of

Monetary Economics, vol. 32(3), pp. 363-394.
Barro R. J. and Lee J. W. 2013. “A new data set of educational attainment in the world, 1950–
2010”, Journal of Development Economics, vol. 104, pp. 184-198.
Barro R. J. and Redlick C. J. 2011. “Macroeconomic Effects from Government Purchases and
Taxes”, Quarterly Journal of Economics, vol. 126(1), pp. 51-102.
272
References
Barro R. J. and Sala-i-Martin X. 1992. “Convergence”, Journal of Political Economy, vol. 100(2),
pp. 223-251.
Barro R. J. and Sala-i-Martin X. 1997. “Technological Diffusion, Convergence, and Growth”,

Journal of Economic Growth, vol. 2(1), pp. 1-26
Barro R. J. and Sala-i-Martin X. 2003. Economic Growth, 2nd edition, MIT Press.
Barta Z. 2018. “In the Red: The Politics of Public Debt Accumulation in Developed Countries”,
Ann Arbor: University of Michigan Press.
Batini N., Callegari G. and Melina G. 2012. “Successful Austerity in the United States, Europe
and Japan”, IMF working paper 12/190, International Monetary Fund.
Batini N., Eyraud L., Forni L. and Weber A. 2014. “Fiscal Multipliers: Size, Determinants, and
Use in Macroeconomic Projections”, Technical Notes and Manuals 14/04.
Baum A. and Koester G.B. 2011. “The Impact of Fiscal Policy on Economic Activity over the
Business Cycle: Evidence from a Threshold VAR Analysis”, Bundesbank Discussion Paper,
03/2011.
Baum A., Poplawski-Ribeiro M. and Weber A. 2012. “Fiscal Multipliers and the State of the
Economy”, IMF working paper 12/286, International Monetary Fund.
Baum A., Checherita-Westphal C. and Rother P. 2013. “Debt and growth: New evidence for the
euro area”, Journal of International Money and Finance, vol. 32(1), pp. 809-821.
Benhabib J. and Spiegel M. 1994. “The roles of human capital in economic development: evidence
from aggregate cross-country data”, Journal of Monetary Economics, vol. 34, pp. 143-173.
Benigno P. and Woodford M. 2003. “Optimal Monetary and Fiscal Policy: A Linear Quadratic
Approach”, NBER working paper 9905, National Bureau of Economic Research.
Bentour E. M. 2018. “On the public debt and growth threshold: one size does not necessarily fit
all”, Cahier de recherche du Creg, n° 2018.01, Université Grenoble Alpes.
Bentour E. M. 2020. “Government Expenditure Multipliers Under Oil Price Swings”, AMF
working paper series, Arab Monetary Fund.
Bentour E. M. and Razzak W. A. 2013. “Do developing countries benefit from foreign direct
investments? An analysis of some Arab and Asian countries”, Review of Middle East Economics
and Finance, vol. 9(3), pp. 357-388.
273
References
Bernanke B. 1986. “Alternative Explanations of the Money–Income Correlation”, Carnegie

Rochester Conference Series on Public Policy, vol. 25, pp. 49–99.
Bernanke B. 2004. “The Great Moderation”, Remarks before the Meetings of the Eastern
Economic Association, Washington, D.C., February 20.
Berti K., De Castro F. and Salto M. 2013. “Effects of fiscal consolidation envisaged in the 2013
Stability and Convergence Programmes on public debt dynamics in EU Member States”,
Economic Papers 504.
Bi H., Shen W. and Yang S. S. 2016. “Debt-dependent effects of fiscal expansions”, European
Economic Review, vol. 88, pp. 142-157.
Blanchard O. 1985. “Debt, Deficits, and Finite Horizons”, Journal of Political Economy, vol.
93(2), pp. 223-247.
Blanchard O. 2006. “Comments on Alan Blinder’s ‘The Case Against the Case Against
Discretionary Policy’”, in the Macroeconomics of Fiscal Policy, ed. Richard W. Kopcke, Geoffrey
M.B. Tootell, and Robert K. Triest, pp. 63-67. Cambridge, MA: The MIT Press.
Blanchard O. 2018. “On the future of macroeconomic models”, Oxford Review of Economic
Policy, vol. 34(1-2), Spring-Summer, pp. 43-54.
Blanchard O. 2019. “Public Debt and Low Interest Rates”, American Economic Review, vol.
109(4), pp. 1197-1229.
Blanchard O. and Perotti R. 2002. “An Empirical Characterization of the Dynamic Effects of
Changes in Government Spending and Taxes on Output”, Quarterly Journal of Economics, vol.
117(4), pp. 1329-1368.
Blanchard O. and Quah D. 1989. “The Dynamic Effects of Aggregate Demand and Supply
Disturbances”, American Economic Review, vol. 79, pp. 655-673.
Bloom N., Jones C. I., Van Reenen J. and Webb M. 2019. “Are Ideas Getting Harder to Find?”,
NBER working paper 23782, National Bureau of Economic Research.
Blot C., Cochard M., Creel J., Ducoudré B., Schweisguth D. and Timbeau X. 2014a. “Fiscal
Consolidation in Times of Crisis: is the Sonner Really the Better?”, Revue de l’OFCE, vol. 132,
pp. 159-192.
Blot C., Cochard M., Creel J., Ducoudré B., Schweisguth D. and Timbeau X. 2014b. “Fiscal
Consolidation, Public Debt and Output Dynamics in the Euro Area: lessons from a simple model
with time-varying fiscal multipliers”, Revue d’Économie Politique, vol. 124, pp. 953-989.
274
References
Borensztein E., De Gregorio J. and Lee J-W. 1998. “How does foreign direct investment affect
economic growth?”, Journal of International Economics, vol. 45(1), pp. 115-135.
Boussard J., De Castro F. and Salto M. 2012. “Fiscal Multipliers and Public Debt Dynamics in
Consolidations”, European Commission Economic Papers 460, European Commission.
Bräuninger M. 2005. “The Budget Deficit, Public Debt, and Endogenous Growth”, The Journal of
Public Economic Theory, vol. 7(5), pp. 827-840.
Breusch T. S. and Pagan A. R. 1980. “The Lagrange Multiplier Test and its Application to Model
Specifications in Econometrics”, Review of Economic Studies, vol. 47(1), pp. 239-253.
Broner F., Clancy D., Martin A. and Erce A. 2019. “Fiscal Multipliers and Foreign Holdings of
Public Debt”, ECB working paper 2255, European Central Bank.
Buchanan J. 1976. “Barro on the Ricardian Equivalence Theorem”, Journal of Political Economy,
vol. 84(2), pp. 337-42.
Buiter W. H. 1988. “Death, Birth, Productivity Growth and Debt Neutrality”, The Economic
Journal, vol. 98(391), pp. 279-93.
Caggiano G., Nodari G. and Castelnuovo E. 2015. “Estimating Fiscal Multipliers: News from a
Non‐linear World”, The Economic Journal, vol. 125(584), pp. 746-776.
Caner M., Grennes T. and Koehler-Geib F. 2010. “Finding the tipping point when sovereign debt
turns bad”, World Bank Policy Research working paper 5391, World Bank.
Canova F. and De Nicolò G. 2002. “Monetary Disturbances Matter for Business Fluctuations in
the G-7”, Journal of Monetary Economics, vol. 49(6), pp. 1131-1159.
Canzoneri M., Collard F., Dellas H. and Diba B., 2015. “Fiscal multipliers in recessions”, The
Economic Journal, vol. 126, pp. 75-108.
Card D., Lee D. S., Pei Z. and Weber A. 2012. “Nonlinear Policy Rules and the Identification and
Estimation of Causal Effects in a Generalized Regression Kink Design”, NBER working paper
Card D., Lee D. S., Pei Z. and Weber A. 2017. “Regression Kink Design: Theory and Practice”,
in Matias D. Cattaneo, Juan Carlos Escanciano (ed.), Regression Discontinuity Designs (Advances
in Econometrics, volume 38), Emerald Publishing Limited, pp. 341-382.
Carlin W. and Soskice D. 2018. “Stagnant productivity and low unemployment: stuck in a
Keynesian equilibrium”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 169-194.
275
References
Casalis A., 2017. “Non-linear effects of the financial cycle on fiscal multipliers”, nonpublished,
University of York.
Cass D. 1965. “Optimum Growth in an Aggregate Model of Capital Accumulation”, Review of

Economic Studies, vol. 32(3), pp. 233-240.
Cecchetti S. G., Mohanty M. S. and Zampolli F. 2011. “The real effects of debt”, BIS working
paper 352, Bank for International Settlements.
Chapple S. 1997. “One for the X-files: A critical assessment of Professor Scully’s ‘Taxation and
economic growth in New Zealand’”, Report to the Periodic Report Group, New Zealand Institute
of Economic Research, Wellington, November.
Chari V. V., McGrattan E. R. and Kehoe P. J. 2009. “The New Keynesian Models: Not Yet Useful
for Policy Analysis”, American Economic Journal: Macroeconomics, vol. 1(1), pp. 242-266.
Chatelain J-B. and Ralf K. 2012. “The Failure of financial macroeconomics and what to do about
it”, The Manchester School, vol. 80, pp. 21-53.
Checherita C. and Rother P. 2010. “The impact of high and growing government debt on economic
growth: an empirical investigation for the Euro area”, ECB working paper 1237, European Central
Bank.
Chen F. 2007. “Sensitivity of Goodness of Fit Indexes to Lack of Measurement Invariance”,

Structural Equation Modeling: A Multidisciplinary Journal, vol. 14(3), pp. 464-504.
Chen C., So M. and Liu F. 2011. “A review of threshold time series models in finance”, Statistics
and Its Interface, vol. 4, pp. 167-181.
Cherif R. and Hasanov F. 2017. “Public debt dynamics: the effects of austerity, inflation, and
growth shocks”, Empirical Economics, vol. 54(1), pp. 1087-1105.
Chinn M. 2013. “Fiscal Multipliers”, The New Palgrave Dictionary of Economics, Steven N.
Durlauf and Lawrence E. Blume, eds. Palgrave Macmillan.
Christiano L. J., Eichenbaum M. and Evans C. L. 2005. “Nominal Rigidities and the Dynamic
Effects of a Shock to Monetary Policy”, Journal of Political Economy, vol. 113(1), pp. 1-45.
Christiano L., Eichenbaum M. and Rebelo S. 2011. “When Is the Government Spending Multiplier
Large?”, Journal of Political Economy, vol. 119(1), pp. 78-121.
Christiano L. J., Eichenbaum M. S. and Trabandt M. 2018. “On DSGE Models”, Journal of
Economic Perspectives, vol. 32(3), pp. 113-140.
276
References
Chudik A., Mohaddes K., Pesaran M. H. and Raissi M. 2017. “Is there a debt-threshold effect on
output growth?”, Review of Economics and Statistics, vol. 99(1), pp. 135-150.
Clark T. 2009. “Is the Great Moderation Over? An Empirical Investigation”, Economic Review,
Federal Reserve Bank of Kansas City, pp. 5-42.
Coenen G., Erceg C. J., Freedman C., Furceri D., Kumhof M., Lalonde R., Laxton D., Lindé J.,
Mourougane A., Muir D., Mursula S., de Resende C., Roberts J., Roeger W, Snudden S, Trabandt
M. and Veld J. 2012. “Effects of Fiscal Stimulus in Structural Models”, American Economic
Journal: Macroeconomics, vol. 4(1), pp. 22-68.
Cogan J. F. and Taylor J. B. 2012. “What the Government Purchases Multiplier Actually
Multiplied in the 2009 Stimulus Package”, in Government Policies and the Delayed Economic
Recovery, Lee Ohanian, John B. Taylor, Ian Wright (eds.) Hoover Press, Stanford, pp. 85-114.
Cogan J. F., Cwik T., Taylor J. B. and Wieland V. 2010. “New Keynesian versus old Keynesian
government spending multipliers”, Journal of Economic Dynamics and Control, vol. 34(10), pp.
281-295.
Cogan J. F., Taylor J. B., Wielend V. and Wolters M. 2013. “Fiscal Consolidation Strategy”,
Journal of Economic Dynamics and Control, vol. 37(2), pp. 404-421.
Cole H. and Ohanian L. 2004. “New Deal Policies and the Persistence of the Great Depression: A
General Equilibrium Analysis”, Journal of Political Economy, vol. 112(4), pp. 779–816.
Colletaz G. and Hurlin, C. 2006, “Threshold effects of the public capital productivity: An
international panel smooth transition approach”, working paper 2006-01, Laboratoire d’économie
d’Orleans.
Combes J. L., Minea A., Mustea L. and Sow N. M. 2014. “The Euro and the Crisis: Evidence on
Recent Fiscal Multipliers”, Revue d’Économie Politique, vol. 124(6), pp. 1013-1038.
Cooley T. and LeRoy S. 1985. “A-theoretical macroeconomics: A critique”, Journal of Monetary

Economics. vol. 16, pp. 283-308.
Correia I., Farhi E., Nicolini J.P. and Teles P. 2012. “Unconventional Fiscal Policy at the Zero
Bound”.
Corsetti G. and Roubini N. 1996. “Optimal Government Spending and Taxation in Endogenous
Growth Models”, NBER working paper 5851, National Bureau of Economic Research.
Corsetti G., Meier A. and Müller G. J. 2012. “What Determines Government Spending
Multipliers?”, IMF Working Paper 12/150, International Monetary Fund.
277
References
Corsetti G., Kuester K., Meier A. and Müller G. J. 2013. “Sovereign risk, fiscal policy, and
macroeconomic stability”, Economic Journal, vol. 123, pp. 99-132.
Cukierman A. and Meltzer A. H. 1989. “A Political Theory of Government Debt and Deficits in a
Neo-Ricardian Framework”, American Economic Review, vol. 79(4), pp. 713-32.
Curdia V. and Woodford M. 2016. “Credit Frictions and Optimal Monetary Policy”, Journal of
Monetary Economics, vol. 84, pp. 30-65.
De la Croix D. and Michel P. 2010. A Theory of Economic Growth: Dynamics and Policy in
Overlapping Generations, Cambridge University Press.
DeLong J. B. and Summers L. 2012. “Fiscal Policy in a Depressed Economy”, Brookings Papers
on Economic Activity, vol. 43(1), pp. 233-297.
Denton F. T. 1971. “Adjustment of Monthly or Quarterly Series to Annual Totals: An Approach

Based on Quadratic Minimization”, Journal of the American Statistical Association, vol. 66, pp.
99–102.
Desai M. 1981. “Testing Monetarism”, Frances Pinter (Publishers) Ltd., London.
Di Sanzo S. and Bella M. 2015, “Public debt and growth in the Euro area: evidence from
parametric and nonparametric Granger causality”, The B.E. Journal of Macroeconomics, vol.
15(2), pp. 631-648.
Diamond P. A. 1965. “National Debt in a Neoclassical Growth Model”, American Economic

Review, vol. 55(5), pp. 1126-1150.
Easton B. 1999. “The whimpering of the state. Policy after MMP”, Auckland, University of
Auckland Press.
Eberhardt M. and Presbitero A. F. 2015. “Public debt and growth: Heterogeneity and non-
linearity”, Journal of International Economics, vol. 97(1), pp. 45-58.
Egert B. 2015. “The 90% public debt threshold: the rise and fall of a stylized fact”, Applied
Economics, vol. 47(34-35), pp. 3756-3770.
Eggertsson G. B. 2011. “What ﬁscal policy is effective at zero interest rates?”, NBER
Macroeconomics Annual, vol. 25, pp. 59-112.
Eggertsson G. B. and Woodford M. 2003. “The Zero Bound on Interest Rates and Optimal
Monetary Policy”, Brookings Papers on Economic Activity, vol. 1, pp. 139-211.
278
References
Egron B. 2018. “Réduction du ratio de dette publique : quels instruments pour quels effets ?”,
working paper 2018-1, EconomiX.
Elmendorf D. and Mankiw N. 1999. “Government Debt”, in Taylor, J. and Woodford, M. (eds.),
Handbook of Macroeconomics, 1C, pp. 1615-1669, North-Holland.
Erceg C. J. and Lindé J. 2012. “Fiscal Consolidation in an Open Economy”, American Economic
Review, vol. 102(3), pp. 186-191.
Estevão M. and Samake I. 2013. “The Economic Effects of Fiscal Consolidation with Debt
Feedback”, IMF working paper 13/136, International Monetary Fund.
European Commission. 2012. “Report on Public finances in European Monetary Union -2012”,
European Economy Number 4/2012 (the report could be accessed online from the link:
https://ec.europa.eu/economy_finance/publications/european_economy/2012/pdf/ee-2012-4.pdf).
European Commission. 2013. “Study on the impacts of fiscal devaluation”, Taxation papers n. 36.
Eyraud L. and Weber A. 2013. “The Challenge of Debt Reduction during Fiscal Consolidation”,
IMF working paper 13/67, International Monetary Fund.
Farhi E. and Werning I. 2017. “Fiscal Multipliers: Liquidity Traps and Currency Unions”, in
Handbook of Macroeconomics 2, pp. 2417-2492.
Farhi E., Gopinath G. and Itskhoki O. 2013. “Fiscal Devaluations”.
Faust J. 1998. “The Robustness of Identified VAR Conclusions about Money”, Carnegie Rochester
Conference Series on Public Policy, vol. 49(1), pp. 207-244.
Fazzari S., Morley J. and Panovska I. 2015. “State dependent effects of fiscal policy”, Studies in
Nonlinear Dynamics and Econometrics, vol. 19, pp. 285-315.
Feldstein M. S. 1974. “Perceived Wealth in Bonds and Social Security: A Comment”, Journal of
Political Economy, vol. 84(2), pp. 331-36.
Ferguson T. and Johnson R. 2011. “A world upside down? Deficit fantasies in the great recession”,
International Journal of Political Economy, vol. 40(1), pp. 3-47.
Feyrer J. and Sacerdote B. 2013. “How Much Would US Style Fiscal Integration Buffer European
Unemployment and Income Shocks? (A Comparative Empirical Analysis)”, American Economic
Review, vol. 103(3), pp. 125-128.
Forni M. and Gambetti L. 2010. “Fiscal Foresight and the Effects of Government Spending”,
CEPR Discussion Paper 7840.
279
References
Fotiou A. 2019. “Non-linearities and fiscal policy”, European University Institute Working Papers,
MWP (2019/11), pp. 1-55.
Fox K. A. 1956. “Econometric Models of the United States”, Journal of Political Economy, vol.
64(2), pp. 128-142.
Freedman C., Kumhof M., Laxton D., Muir D. and Mursula S. 2010. “Global effects of fiscal
stimulus during the crisis”, Journal of Monetary Economics, vol. 57, pp. 506-526.
Fuentes R. J. and Morales M. 2011. “On the measurement of Total Factor Productivity: A Latent
Variable Approach”, Macroeconomic Dynamics, vol. 15, pp. 145-159.
Futagami K., Iwaisako T. and Ohdoi R. 2008. “Debt policy rule, productive government spending,
and multiple growth paths”, Macroeconomic Dynamics, vol. 12, pp. 445-462.
Futagami K., Morita Y. and Shibata A. 1993. “Dynamic analysis of an endogenous growth model
with public capital”, Scandinavian Journal of Economics, vol. 95, pp. 607-625.
Galí J. 1992. “How Well Does The IS-LM Model Fit Postwar U. S. Data?”, Quarterly Journal of
Economics, vol. 107(2), pp. 709-738.
Galí J. 2018. “The State of New Keynesian Economics: A Partial Assessment”, Journal of
Economic Perspectives, vol. 32(3), pp. 87-112.
Galí J., López-Salido J. D. and Vallés J. 2007. “Understanding the effects of Government Spending
on Consumption”, European Economic Association Journal, vol. 5(1), pp. 227-270.
Ghironi F. 2018. “Macro needs micro”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 195-
218.
Ghosh A., Kim J., Mendoza E., Ostry J. and Qureshi M. 2013. “Fiscal Fatigue, Fiscal Space, and
Debt Sustainability in Advanced Economies”, The Economic Journal, vol. 123(566), pp. F4-F30.
Ghosh S. and Mourmouras I. A. 2004a. “Endogenous growth, welfare and budgetary regimes”,
Journal of Macroeconomics, vol. 26, pp. 623-635.
Ghosh S. and Mourmouras I. A. 2004b. “Debt, growth and budgetary regimes”, Bulletin of
Economic Research, vol. 56, pp. 241 250.
Glocker C., Sestieri G. and Towbin P. 2019. “Time-varying fiscal spending multipliers in the UK”,
working paper 643, Banque de France.
280
References
Gomez-Puig M. and Sosvilla-Rivero S. 2015. “On the bi-directional causal relationship between
public debt and economic growth in EMU countries”, IREA working paper 2015-12, Research
Institute of Applied Economics.
Gomez-Puig M. and Sosvilla-Rivero S. 2017. “Heterogeneity in the debt-growth nexus: Evidence

from EMU countries”, IREA working paper 2017-06, Research Institute of Applied Economics.
Gonzales-Garcia J., Lemus A. and Mrkaic M. 2013. “Fiscal Multipliers in the ECCU”, IMF
Gonzalez A., Teräsvirta T. and van Dijk D. 2005. “Panel smooth transition regression models”,
SEE/EFI working paper series in Economics and Finance, No. 604.
Gordon R. J. and Krenn R. 2010. “The End of the Great Depression 1939-41: Policy Contributions
and Fiscal Multipliers”, NBER working paper 16380, National Bureau of Economic Research.
Gorodnichenko Y., Mendoza E. and Tesar L. 2012. “The Finnish Great Depression: From Russia
with Love”, American Economic Review, vol. 102(4), pp. 1619-44.
Gossé J.-B. and Guillaumin C. 2013. “L’apport de la représentation VAR de Christopher A. Sims
à la science économique”, L’Actualité économique, Revue d’analyse économique, vol. 89(4), pp.
305-319.
Greiner A. 2007. “An endogenous growth model with public capital and sustainable government
debt”, Japanese Economic Review, vol. 58(3), pp. 345-361.
Greiner A. 2011. “Economic Growth, Public Debt and Welfare: Comparing Three Budgetary
Rules”, German Economic Review, vol. 12(2), pp. 205-222.
Greiner A. 2012. “Public debt in a basic endogenous growth model”, Economic Modelling, vol.
29(4), pp. 1344-1348.
Greiner A. 2014. “Fiscal and Monetary Policy in a Basic Endogenous Growth Model”,
Computational Economics, vol. 45(2), pp. 285-301.
Greiner A. 2015. “Public Debt, Productive Public Spending and Endogenous Growth”, Japanese
Economic Review, vol. 66(4), pp. 520-535.
Greiner A. 2016. “Human capital formation and public debt: growth and welfare effects of three
different deficit policies”, Journal of Economic Policy Reform, vol. 19(4), pp. 369-385.
Greiner A. and Semmler W. 1999. “An endogenous growth model with public capital and
government borrowing”, Annals of Operations Research, vol. 88, pp. 65-79.
281
References
Greiner A. and Semmler W. 2000. “Endogenous growth, government debt and budgetary
regimes”, Journal of Macroeconomics, vol. 22(3), pp. 363-384.
Grilli V., Masciandaro D., Tabellini G., Malinvaud E. and Pagano M. 1991. “Political and
Monetary Institutions and Public Financial Policies in the Industrial Countries”, Economic Policy,
vol. 6(13), pp. 341-392.
Grossman G. M. and Helpman E. 1991. “Quality Ladders in the Theory of Growth”, The Review
of Economic Studies, vol. 58(1), pp. 43-61.
Haldane A. G. and Turrell A. E. 2018. “An interdisciplinary model for macroeconomics”, Oxford
Review of Economic Policy, vol. 34(1-2), pp. 219-251.
Hall R. 2009. “By how much does GDP rise if the Government buys more output?”, Brookings
Papers on Economic Activity, vol. 2, pp. 183-231.
Hansen B. E. 2017. “Regression Kink with an Unknown Threshold”, Journal of Business and
Economic Statistics, vol. 35(2), pp. 228-240.
Hansen A. H. 1959. “The Public Debt Reconsidered: A Review Article”, Review of Economics
and Statistics, vol. 41(4), pp. 377-78.
Hansen B. E. 1999. “Threshold effects in non-dynamic panels: estimation, testing, and inference”,
Journal of Econometrics, vol. 93(2), pp. 334-368.
Hansen B. E. 2000. “Sample splitting and threshold estimation”, Econometrica, vol. 68, 575-603.
Haslag J. H. and Young E. R. 1998. “Money creation, reserve requirements, and seigniorage”,
Review of Economic Dynamics, vol. 1(3), pp. 677-698.
Hegeland H. 1954. “The Multiplier Theory”, Lund Social Science 9, p. 261.
Henderson D. J., Parmeter C. F. and Su L. 2015. “Nonparametric Threshold Regression:

Estimation and Inference”, mimeo.
Hendry D. F. and Muellbauer J. N. 2018. “The future of macroeconomics: macro theory and
models at the Bank of England”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 287-328.
Herndon T., Ash M. and Pollin R. 2013. “Does high public debt consistently stifle economic
growth? A critique of Reinhart and Rogoff”, PERI working paper 322.
Hill R. 2008. “Optimal taxation and economic growth: a comment”, Public Choice, vol. 134, pp.
419-427.
282
References
Hodrick R. J. and Prescott E. C. 1997. “Postwar U.S. Business Cycles: An Empirical

Investigation”, Journal of Money, Credit and Banking, vol. 29(1), pp. 1-16.
Hoover K. D., Johansen S. and Juselius K. 2008. “Allowing the Data to Speak Freely: The
Macroeconometrics of the Cointegrated Vector Autoregression”, American Economic Review,
vol. 98(2), pp. 251-255.
Horvath M. 2009. “The effects of government spending shocks on consumption under optimal
stabilization”, European Economic Review, vol. 53(7), pp. 815-829.
Hsieh C. T. and Klenow P. J. 2010. “Development Accounting”, American Economic Journal:

Macroeconomics, vol. 2(1), pp. 207-223.
Huidrom R., Kose M. A., Lim J. J. and Ohnsorge F. L. 2016. “Do Fiscal Multipliers Depend on
Fiscal Positions?”, Policy Research working paper 7724. World Bank Group.
Hulten C. 2001. “Total Factor Productivity: A Short Biography”, in New Developments in

Productivity Analysis, Charles R. Hulten, Edwin R. Dean, and Michael J. Harper, eds., Studies in
Income and Wealth, vol. 63, The University of Chicago Press for the National Bureau of Economic
Research, Chicago, pp. 1-47.
ILO. 2011. “A review of global fiscal stimulus”, European Commission-International Institute for
Labour Studies (EC-IILS) Joint Discussion Paper Series No. 5.
Ilzetzki E., Mendoza E. G. and Végh C. A. 2013. “How big (small?) are fiscal multipliers?”,
Journal of Monetary Economics, vol. 60(2), pp. 239-254.
IMF. 2009. Group of Twenty, Meeting of the Deputies, January 31–February 1, London, U.K.,
Note by the Staff of the IMF. https://www.imf.org/external/np/g20/pdf/020509.pdf.
IMF. 2013. “Staff Guidance Note for Public Debt Sustainability Analysis in Market-Access
Countries”, International Monetary Fund.
IMF. 2015. “Making Public Investment more Efficient”, Policy Report, June, Washington DC.
IMF and The World Bank. 2005. “Operational Framework for Debt-Sustainability Assessments in
Low-Income Countries—Further Considerations”.
IMF and The World Bank. 2013. “Staff Guidance Note on the Application of the Joint Bank-Fund
Debt Sustainability Framework for Low Income Countries”.
Jones C. I. 1995. “R&D-Based Models of Economic Growth”, Journal of Political Economy, vol.
103(4), pp. 759-784.
283
References
Jones C. I. 2003. “Population and Ideas: A Theory of Endogenous Growth”, in P. Aghion, R.

Frydman, J. Stiglitz, and M. Woodford, eds., Knowledge, Information, and Expectations in
Modern Macroeconomics: In Honor of Edmund S. Phelps, Princeton University Press.
Jones C. I. 2005. “Growth and Ideas”, in P. Aghion and Durlauf S., eds, Handbook of Economic
Growth, edition 1, vol. 1, chapter 16, pp. 1063-1111.
Jones C. I. 2019. “Paul Romer: Ideas, Nonrivalry, and Endogenous Growth”, Scandinavian Journal
of Economics, vol. 121(3), pp. 859-883.
Jones C. I. and Romer P. M. 2010. “The New Kaldor Facts: Ideas, Institutions, Population, and
Human Capital”, American Economic Journal: Macroeconomics, vol. 2(1), pp. 224-245.
Kennedy P. E. 2000. “On measuring the growth-maximizing tax rate”, Pacific Economic Review,
vol. 5(1), pp. 89-91.
Keynes J. M. 1936. The General Theory of Employment, Interest and Money, Palgrave Macmillan.
Kilponen J., Pisani M., Schmidt S., Corbo V., Hledik T., Hollmayr J., Hurtado S., Júlio P., Kulikov
D., Lemoine M., Lozej M., Lundvall H., Maria J. R., Micallef B., Papageorgiou D., Rysanek J.,
Sideris D., Thomas C. and De Walque G. 2015. “Comparing fiscal multipliers across models and
countries in Europe”, ECB working paper 1760, European Central Bank.
Kirchner M., Cimadomo J. and Hauptmeier S. 2010. “Transmission of Government Spending

Shocks in The Euro Area: Time Variation and Driving Forces”, ECB working paper 1219,
European Central Bank.
Klein L. R. and Goldberger A. 1955. An Econometric Model of the United States, 1929-1952,
Amsterdam: North-Holland.
Klein L.R. 1950. “Economic Fluctuations in the United States, 1921–1941”, New-York: John
Wiley.
Kmenta J. 1967. “On Estimation of the CES Production Function”, International Economic
Review, vol. 8, pp. 180-189.
Kose M. A., Nagle P., Ohnsorge F. and Sugawara N. 2020. “Global Waves of Debt: Causes and
Consequences”, Advanced Edition. World Bank.
Kourtellos A., Stengos T. and Tan C. M. 2013. “The effect of public debt on growth in multiple
regimes”, Journal of Macroeconomics, vol. 38(Part A), pp. 35-43.
Krugman P. 1994. The Age of Diminished Expectations: U.S. Economic Policy in the 1990s, MIT
Press.
284
References
Krugman P. 2018. “Good enough for government work? Macroeconomics since the
crisis”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 156-168.
Krugman P. 2019. Melting Snowballs and the Winter of Debt. The New York Times, Jan. 9,
https://www.nytimes.com/2019/01/09/opinion/melting-snowballs-and-the-winter-of-debt.html.
Kumar M. and Woo J. 2010. “Public Debt and Growth”, IMF working paper 10-74, International
Monetary Fund.
Kydland F. and Prescott E. 1982. “Time to Build and Aggregate Fluctuations”, Econometrica, vol.
50(6), pp. 1345-1370.
Kydland F. E. and Prescott E. C. 1977. “Rules Rather than Discretion: The Inconsistency of
Optimal Plans”, Journal of Political Economy, vol. 85(3), pp. 473-492.
Lee S., Park H., Seo M. H. and Shin Y. 2017. “Testing for a Debt-threshold Effect on Output
Growth”, Accepted version in Fiscal Studies.
Leeper E. M., Richter A. W. and Walker T. B. 2012. “Quantitative Effects of Fiscal Foresight”,
American Economic Journal: Economic Policy, vol. 4(2), pp. 115-144.
Leeper E. M., Walker T. B. and Yang S-C. S. 2013. “Fiscal Foresight and Information Flows”,
Econometrica, vol. 81(3), pp. 1115-1145.
Leith C. and von Thadden L. 2008. “Monetary and fiscal policy interactions in a New Keynesian
model with capital accumulation and non-Ricardian consumers”, Journal of Economic Theory,
vol. 140(1), pp. 279-313.
Leith C. and Wren-Lewis S. 2013. “Fiscal sustainability in a new Keynesian model”, Journal of
Money, Credit and Banking, vol. 45(8), pp. 1477-1516.
Lin T. 2014. “High dimensional threshold quantile regression with an application to debt overhang
and economic growth”, Job market paper, University of Wisconsin.
Lindé J. 2018. “DSGE models: still useful in policy analysis?”, Oxford Review of Economic
Policy, vol. 34(1-2), pp. 269-286.
Lindé J. and Trabandt M. 2013. Fiscal Multipliers in a Nonlinear World.
Liu Z., Wang P. and Zha T. 2011. “Land-Price Dynamics and Macroeconomic Fluctuations”,
Federal Reserve Bank of San Francisco, working paper 2011-16.
Lucas R. E. 1976. “Econometric policy evaluation: A critique”, Carnegie Rochester Conference

Series on Public Policy, vol. 1(1), pp 19-46.
285
References
Lucas R. E. 1988. “On the Mechanics of Economic Development”, Journal of Monetary

Lucas R. E. and Stokey N. L. 1983. “Optimal Fiscal and Monetary Policy in an Economy Without
Capital”, Journal of Monetary Economics, vol. 12, pp. 55-93.
Maebayashi N., Hori T. and Futagami K. 2017. “Dynamic analysis of reductions in public debt in
an endogenous growth model with public capital”, Macroeconomic Dynamics, vol. 21, pp. 1454-
1483.
Mankiw N. G. 2000. “The Savers-Spenders Theory of Fiscal Policy”, American Economic

Review, vol. 90(2), pp. 120-125.
Marchionne F. and Parekh S. 2015. “Growth, debt and inequality”, Economic Issues, vol. 20(2),
pp. 67-94.
Mayer E., Rüth S. and Scharler J. 2013. “Government debt, inflation dynamics and the
transmission of fiscal policy shocks”, Economic Modelling, vol. 33, pp. 762-771.
Mazza J. 2019. “Is public debt a cheap lunch?”, Jan, 21, 2019.
https://www.bruegel.org/2019/01/is-public-debt-a-cheap-lunch/.
Mbaye S., Badia M. and Chae K. 2018. “Global Debt Database: Methodology and Sources”, IMF
McKibbin W. J. and Stoeckel A. 2018. “Modelling a complex world: improving macro-models”,

Oxford Review of Economic Policy, vol. 34(1-2), pp. 329-347.
Mertens K. and Ravn M. O. 2010. “Measuring the Impact of Fiscal Policy in the Face of
Anticipation: A Structural VAR Approach”, The Economic Journal, vol. 120(544), pp. 393-413.
Mian A. and Sufi A. 2012. “The Effects of Fiscal Stimulus: Evidence from the 2009 ‘Cash for
Clunkers’ Program”, Quarterly Journal of Economics, vol. 127(3), pp. 1107-1142.
Miller P. J. and Sargent T. J. 1984. “A reply to Darby”, Federal Reserve Bank of Minneapolis
Quarterly Review, vol. 8(Spring), pp. 21-26.
Mincer J. 1974. “The Human Capital Earnings Function”, NBER Chapters, in: Schooling,
Experience, and Earnings, pages 83-96, National Bureau of Economic Research.
Minea, A. and Parent, A. 2012, “Is High Public Debt Always Harmful to Economic Growth?
Reinhart and Rogoff and Some Complex Nonlinearities”, CERDI working paper 2012-18.
286
References
Minea A. and Villieu P. 2013. “Debt policy rule, productive government spending, and multiple
growth paths: A note”, Macroeconomic Dynamics, vol. 17, pp. 947-954.
Modigliani F. 1961. “Long Run Implications of Alternative Fiscal Policies and the Burden of the
National Debt”, Economic Journal, vol. 71(284), pp. 730-755.
Molnár M. 2012. “Fiscal consolidation: What factors determine the success of consolidation
efforts?”, OECD Journal: Economic Studies, vol. 2012/1.
Mustea L. 2015. “How large are fiscal multipliers in the US?”, Procedia Economics and Finance,
vol. 20, pp. 423-427.
Nakamura E. and Steinsson J. 2014. “Fiscal Stimulus in a Monetary Union: Evidence from US
Regions”, American Economic Review, vol. 104(3), pp. 753-92.
Nelson R. and Phelps E. 1966. “Investment in Humans, Technological Diffusion, and Economic
Growth”, American Economic Review, vol. 56(1/2), pp. 69-75.
Ni S. 1999. “National Debt, Savings, and Real Interest Rates in a Neoclassical Model with
Endogenous Labour Supply and Knowledge-Based Growth”, The Canadian Journal of Economics,
vol. 32(5), pp. 1227-1244.
Nickel C. and Tudyka A. 2014. “Fiscal stimulus in times of high debt: reconsidering multipliers
and twin deficits”, Journal of Money, Credit and Banking, vol. 46(7), pp. 1313-1344.
Nielsen H. S., Torben S. and Taber C. R. 2010. “Estimating the Effect of Student Aid on College
Enrollment: Evidence from a Government Grant Policy Reform”, American Economic Journal:
Economic Policy, vol. 2(2), pp. 185-215.
North D. C. 1990. Institutional Change, and Economic Performance, Cambridge University Press.
Obstfeld M. 1997. “Dynamic Seigniorage Theory”, Macroeconomic Dynamics, vol. 1(3), pp. 588-
614.
Obstfeld M. and Taylor A. M. 2002. “Globalization and Capital Markets”, NBER working paper
O’Driscoll G. P. 1977. “The Ricardian Nonequivalence Theorem”, Journal of Political Economy,

vol. 85(1), pp. 207-210.
OECD. 2009. “Policy Responses to the Economic Crisis: Investing in Innovation for Long-Term
Growth”.
287
References
OECD. 2011. “Fiscal consolidation: targets, plans and measures”, OECD Journal on Budgeting,
vol. 11/2.
OECD. 2012. “Fiscal Consolidation: How Much is Needed to Reduce Debt to a Prudent Level?”,
OECD Economics Department Policy Notes, No. 11, April.
Ostry J. D., Ghosh A. R., Kim J. I. and Qureshi M.S. 2010. “Fiscal Space”, IMF Staff Position
Note 10/11, International Monetary Fund.
Ostry J. D., Ghosh A.R. and Espinoza R. 2015. “When Should Public Debt Be Reduced?”, IMF
Staff Discussion Note SDN/15/10, International Monetary Fund.
Ouliaris S., Pagan A. and Restrepo J. 2018. “Quantitative Macroeconomic Modeling with
Structural Vector Autoregressions – An EViews Implementation”, unpublished e-book available
online at: https://www.eviews.com/StructVAR/structvar.pdf.
Owyang M. T., Ramey V. A. and Zubairy S. 2013. “Are Government Spending Multipliers Greater
during Periods of Slack? Evidence from Twentieth-Century Historical Data”, American Economic
Review, vol. 103(3), pp. 129-134.
Pan H. and Wang C. 2013. “Co-movement of Government Debt and Economic Growth in the
Euro-area: A Bayesian Dynamic Factor Model Analysis”, International Economic Journal, vol.
27(4), pp. 625-643.
Panizza U. and Presbitero A. F. 2012. “Public Debt and Economic Growth Is There a Causal
Effect?”, Money and Finance Research Group (MoFiR) working paper 65.
Panizza U. 2015. “Public debt and long-term economic growth: the research and policy agenda”,
Lecture delivered at the 43rd Economics Conference of the Oesterreichische National Bank,
Vienna.
Parker J. A., 2011. “On Measuring the Effects of Fiscal Policy in Recessions”, Journal of
Economic Literature, vol. 49(3), pp. 703-718.
Parraga Rodriguez S. 2018. “The dynamic effects of public expenditure shocks in the United
States”, Journal of Macroeconomics, vol. 56, pp. 340-360.
Perdichizzi S. 2017. “Estimating Fiscal multipliers in the Eurozone: A Nonlinear Panel Data
Approach”, working paper 58, Università Cattolica del Sacro Cuore, Dipartimento di Economia e
Finanza (DISCE), Milano.
Perotti R. 2005. “Estimating the Effects of Fiscal Policy in OECD Countries”, CEPR Discussion
Paper 4842.
288
References
Perotti R. 2011. “The Effects of Tax Shocks on Output: Not So Large, But Not Small Either”,
NBER working paper 16786, National Bureau of Economic Research.
Persson T. and Svensson L. E. O. 1989. “Why a Stubborn Conservative would Run a Deficit:
Policy with Time-Inconsistent Preferences”, Quarterly Journal of Economics, vol. 104(2), pp. 325-
345.
Pesaran H. M., Ullah A. and Yamagata T. 2008. “A bias-adjusted LM test of error cross-section
independence”, Econometrics Journal, vol. 11(1), pp. 105-127.
Pesaran M. H. 2004. “General diagnostic tests for cross section dependence in panels”, CESifo
working paper 1233.
Pesaran M. H. 2006. “Estimation and Inference in Large Heterogeneous Panels with a Multifactor
Error Structure”, Econometrica, vol. 74(4), pp. 967-1012.
Pescatori A., Sandri D. and Simon J. 2014. “Debt and growth: Is there a magic threshold?”, IMF
Phelps E. S. 1961. “The Golden Rule of Accumulation: A Fable for Growthmen”, American
Economic Review, vol. 51, pp. 638-643.
Phelps E. S. 2018. “The Fantasy of Fiscal Stimulus”, Wall Street Journal, Oct 19.
Philippon T. 2019. “The True Cost of Public Debt”, https://www.stern.nyu.edu/experience-

stern/faculty-research/true-cost-public-debt
Pienkowski A. 2017. “Debt Limits and the Structure of Public Debt”, IMF working paper 17/117,
International Monetary Fund.
Poghosyan T. 2017. “How Do Public Debt Cycles Interact with Financial Cycles?”, Empirical
Economics, vol. 54(71), pp. 425-460.
Priftis R. and Zimic S. 2018. “Sources of borrowing and fiscal multipliers”, Bank of Canada Staff
working paper 2018-32.
Quesnay F. 1758. Tableau Economique. 3d ed. Reprint, Edited by M. Kuczynski and R. Meek.
London: Macmillan, 1972.
Ramey V. 2011. “Identifying Government Spending Shocks: It’s All in the Timing”, Quarterly
Journal of Economics, vol. 126(1), pp. 1-50.
Ramey V. 2016. “Macroeconomic Shocks and their Propagation”, chapter 2 in Handbook of

Macroeconomics (Pages 71-162). Edited by John B. Taylor and Harald Uhlig, Volume 2.
289
References
Ramey V. 2018. “Ten years after the Financial Crisis: What have we learned from the renaissance
of fiscal research?”, Journal of Economic Perspectives. Forthcoming.
Ramey V. A. and Zubairy S. 2018. “Government Spending Multipliers in Good Times and in Bad:
Evidence from US Historical Data”, Journal of Political Economy. vol. 126(2), pp. 850-901.
Ramsey F. P. 1928. “A Mathematical Theory of Saving”, The Economic Journal, vol. 38(152), pp.
543-559.
Reinhart C. M. and Rogoff K. S. 2009. This Time Is Different: Eight Centuries of Financial Folly,
Princeton University Press.
Reinhart C. M. and Rogoff K. S. 2010. “Growth in a time of debt”, American Economic Review,
vol. 100(2), pp. 573-578.
Reinhart C. M., Reinhart V. R. and Rogoff K. S. 2012. “Public debt overhangs: Advanced
economy episodes since 1800”, Journal of Economic Perspectives, vol. 26(3), pp. 69-86.
Reis R. 2018. “Is something really wrong with macroeconomics?”, Oxford Review of Economic
Policy, vol. 34(1-2), pp. 132-155.
Riera‐Crichton D., Vegh C. A. and Vuletin G. 2014. “Fiscal Multipliers in Recessions and
Expansions: Does It Matter Whether Government Spending Is Increasing or Decreasing?”, Policy
Research working paper 6993, World Bank.
Rodrik D., Subramanian A. and Trebbi F. 2004. “Institutions Rule: The Primacy of Institutions
over Geography and Integration in Economic Development”, Journal of Economic Growth, vol.
9(2), p.131-165.
Rodrik D. 2011. The Globalization Paradox: Democracy and the Future of the World Economy,
W. W. Norton.
Romer C. and Bernstein J. 2009. “The Job Impact of the American Recovery and Reinvestment
Plan”, January 8, 2009.
Romer C. D. and Romer D. H. 2010. “The Macroeconomic Effects of Tax Changes: Estimates
Based on a New Measure of Fiscal Shocks”, American Economic Review, vol. 100(3), pp. 763-
801.
Romer D. 2011. “What Have We Learned about Fiscal Policy from the Crisis?”, IMF Conference
on Macro and Growth Policies in the Wake of the Crisis, March.
Romer P.M. 1986. “Increasing Returns and Long-Run Growth”, Journal of Political Economy,
vol. 94(5), pp. 1002-1037.
290
References
Romer P.M. 1990. “Human capital and growth: Theory and evidence”, Carnegie-Rochester
Conference Series on Public Policy, Elsevier, vol. 32(1), pages 251-286.
Rossi R. 2014. “Designing monetary and fiscal policy rules in a new Keynesian model with rule-
if-thumb consumers”, Macroeconomic Dynamics, vol. 18(2), pp. 395-417.
Roubini N. and Sachs J. D. 1989. “Political and economic determinants of budget deficits in the
industrial democracies”, European Economic Review, vol. 33(5), pp. 903-938.
Rupert P. and Šustek R. 2019. “On the mechanics of New-Keynesian models”, Journal of
Monetary Economics, vol. 102, pp. 53-69.
Samuelson P. 1958. “An Exact Consumption-Loan Model of Interest with or without the Social
Contrivance of Money”, Journal of Political Economy, vol. 66(6), pp. 467-482.
Schmitt-Grohe S. and Uribe M. 2004. “Optimal Monetary and Fiscal Policy under Sticky Prices”,
Journal of Economic Theory, vol. 114, pp. 198-230.
Scully G. W. 1996. “Taxation and economic growth in New Zealand”, Pacific Economic Review,
vol. 1(2), pp. 169-177.
Scully G. W. 1999. “Unfinished reform: taxation and economic growth in New Zealand”, Journal
of Private Enterprise, vol. 14(2), pp. 92-114.
Scully G. W. 2003. “Optimal taxation, economic growth, and income inequality”, Public Choice,
vol. 115, pp. 299-312.
Shapiro M. and Watson M. 1988. “Sources of Business Cycles Fluctuations”, in NBER

Macroeconomics Annual 1988, vol. 3.
Sharpe T. P. 2013. “Institutional arrangements and public debt threshold limits”, International
Review of Applied Economics, vol. 27(6), pp. 707-728.
Sims C. 1972. “Money, Income, and Causality”, American Economic Review, vol. 62(4), pp. 540-
552.
Sims C. 1980. “Macroeconomics and Reality”, Econometrica, vol. 48(1), pp. 1-48.
Sims C. 1986. “Are Forecasting Models Usable for Policy Analysis?”, Federal Reserve Bank of
Minneapolis, Quarterly Review (Winter), pp. 2-16.
Smets F. and Wouters R. 2007. “Shocks and Frictions in U.S. Business Cycles: A Bayesian DSGE
Approach”, American Economic Review. vol. 97(3), pp. 506-606.
291
References
Solow R. 1956. “A Contribution to the Theory of Economic Growth”, Quarterly Journal of

Spilimbergo A., Symansky S. and Schindler M. 2009. “Fiscal Multipliers”, IMF Staff Position
Note, SPN/09/11, International Monetary Fund.
Stiglitz J. 2018. “The Myth of Secular Stagnation”, Project Syndicate, August 28.
Stiglitz J. 2018. “Where modern macroeconomics went wrong”, Oxford Review of Economic
Policy, vol. 34(1-2), pp. 70-106.
Stock J. H. and Watson M. W. 2003. “Forecasting Output and Inflation: The Role of Asset Prices”,
Journal of Economic Literature, vol. 41(3), pp. 788-829.
Swan T. W. 1956. “Economic growth and capital accumulation”, Economic Record, vol. 32(2),
pp. 334-361.
Syssoyeva-Masson I. and De Sousa Andrade J. 2017. “The effect of public debt on growth in
multiple regimes in the presence of long-memory and non-stationary debt series”, CeBER working
paper 7.
Tagkalakis A. 2008. “The effects of fiscal policy on consumption in recessions and expansions”,
Journal of Public Economics, vol. 92(5-6), pp. 1486-1508.
Taylor J. B. 2011. “An Empirical Analysis of the Revival of Fiscal Activism in the 2000s”, Journal
of Economic Literature, vol. 49(3), pp. 686-702.
Taylor J. B. 2018. “Fiscal Stimulus Programs During the Great Recession”, Economics Working
Paper 18117. Hoover Institution, Stanford.
Terra C. 2015. “Principles of International Finance and Open Economy Macroeconomics:

Theories, Applications and Policies”, Academic Press, 1st edition.
Thistlethwaite D. and Campbell D. 1960. “Regression-Discontinuity Analysis: An alternative to

the ex post facto experiment”, Journal of Educational Psychology, vol. 51(6), pp. 309-317.
Tinbergen J. 1939. “Statistical Testing of Business-Cycle Theories. II. Business Cycles in the
United States of America 1919-1932”, Geneva: League of Nations.
Turnovsky S. J. 1997. “Fiscal policy in a growing economy with public capital”, Macroeconomic
Dynamics, vol. 1, pp. 615-639.
Turnovsky S. J. and Brock W. A. 1980. “Time consistency and optimal government policies in
perfect foresight equilibrium”, Journal of Public Economics, vol. 13(2), pp. 183-212.
292
References
Uhlig H. 2005. “What Are the Effects of Monetary Policy on Output? Results from an Agnostic
Identication Procedure”, Journal of Monetary Economics, vol. 52(2), pp. 381-419.
Vines D. and Wills S. 2018. “The rebuilding macroeconomic theory project: an analytical
assessment”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 1-42.
Vines D. and Wills S. 2018. “The financial system and the natural real interest rate: towards a ‘new
benchmark theory model’”, Oxford Review of Economic Policy, vol. 34(1-2), pp. 252-268.
Weil P. 1989. “Overlapping families of infinitely-lived agents”, Journal of Public Economics, vol.
38(2), pp. 183-198.
Weil P. 2008. “Overlapping Generations: The First Jubilee”, Journal of Economic Perspectives,
vol. 22(4), pp.115-134.
Wieland V., Cwik T., Muller G. J., Schmidt S. and Wolters M. 2012. “A new comparative
approach to macroeconomic modeling and policy analysis”, Journal of Economic Behavior and
Organization, vol. 83(3), pp. 523-541.
Wold, H. 1954. “A Study in the Analysis of Stationary Time Series”, Second revised edition,
Almqvist and Wiksell Book Co., Uppsala.
Woo J., Kinda T. and Poplawski-Ribeiro M. 2013. “Economic Recovery from Recessions: The
Role of Fiscal Policy, and Structural Reforms”, IMF working paper, International Monetary Fund.
Woodford M. 2003. Interest and Prices: Foundations of a Theory of Monetary Policy, Princeton
University Press.
Woodford, M. 2011. “Simple Analytics of the Government Expenditure Multiplier”, American

Economic Journal: Macroeconomics, vol. 3(1), pp. 1-35.
World of Work Report. 2010. From one crisis to the next? International Institute for Labour
Studies, ILO (republished under Willey Online Library, July 3, 2012).
Wren-Lewis S. 2018. “Ending the microfoundations hegemony”, Oxford Review of Economic

Policy, vol. 34(1-2), pp. 55-69.
Wright R. 2018. “On the future of macroeconomics: a New Monetarist perspective”, Oxford
Review of Economic Policy, vol. 34(1-2), pp. 107-131.
Yaari M. 1965. “Uncertain lifetime, life insurance, and the theory of the consumer”, Review of
Economic Studies, vol. 32, pp. 137-150.
293
References
Yakita A. 2008. “Sustainability of public debt, public capital formation, and endogenous growth
in an overlapping generations setting”, Journal of Public Economics, vol. 92, pp. 897-914.
Zhang Y., Zhou Q. and Jiang L. 2017. “Panel kink regression with an unknown threshold”,
Economics Letters, vol. 157(2017), pp. 116-121.
294

BENTOUR 2020 Archivage

Transféré par

Droits d'auteur :

Formats disponibles

BENTOUR 2020 Archivage

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

BENTOUR 2020 Archivage

Transféré par

Droits d'auteur :

Formats disponibles

Public debt and economic growth : a new assessment

To cite this version:

HAL Id: tel-03282014

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

Thèse dirigée par Cyriac Guillaumin , Maître de Conférences

Dette publique et croissance économique : Une

Public Debt and Economic Growth : A new

Thèse soutenue publiquement le 16 octobre 2020,

Présentée et soutenue par

DETTE PUBLIQUE ET CROISSANCE ÉCONOMIQUE :

M. Jean-Pierre ALLEGRET Président du Jury

Résumé de la thèse en français ............................................................................. 8

1.Historical facts of public debt and growth in advanced economies .......................................... 25

Chapter 2. Short-term effects of public debt on growth: The spending

Chapter 3. Public debt effects in theory-based models with an empirical

1. Introduction ............................................................................................................................. 188

General Conclusion ........................................................................................... 265

Résumé de la thèse en français

Dette publique et croissance économique : une nouvelle évaluation

2. Contribution de la thèse à la littérature et débats des effets de la dette

essentiellement portée sur les différentes nouvelles contributions autour de la taille du

2.1. Résumé du débat théorique sur les effets de la dette publique

intergénérationnels dans des modèles à générations imbriquées (Diamond, 1965 ; Blanchard,

Le résumé de ces résultats peut être regroupé en trois groupes :

- un premier groupe de recherches empiriques se concentrant principalement sur l’examen

- un deuxième groupe intéressé par la causalité entre la dette publique et la croissance

- un troisième groupe qui examine les effets d’autres variables institutionnelles et

2.3. Résumé du débat sur la valeur des multiplicateurs de dépenses publiques

Par ailleurs, d’autres chercheurs conduisent à découvrir la vulnérabilité des multiplicateurs

Très récemment, lors de la conférence présidentielle de l’American Economic Association,

Récemment, la politique budgétaire, et en particulier la dette publique a également été modélisée

le noyau et la référence fondamentale pour de nombreuses contributions récentes adoptant le cadre

3. Contribution empirique de la thèse

Les principales conclusions de nos évaluations empiriques conduisent à rejeter le taux

3.1. Présentation et choix de l’échantillon de pays

- l’importance économique de ce groupe de pays avancés influençant les économies

3.2. Contribution empirique du Chapitre 1 : pas de seuil universel de dette publique

- 1880-1913 : cette période correspond à la fin de la première mondialisation ou la fin de

- 1914-1945 : une période avec deux guerres dévastatrices et la grande dépression

- 1946-1970 : le monde a connu une croissance et un développement solides au cours de

3.3. Contribution empirique du Chapitre 2 : sensibilité du multiplicateur budgétaire

En outre, en contrôlant de manière exogène les mouvements de la dette publique, indépendamment

3.4. Contribution empirique du Chapitre 3 : Le ratio de dette publique potentiel à

Empiriquement, la moyenne de la dette potentielle simulée en pourcentage du PIB sur la période

Ce chapitre a comme valeur ajoutée la modélisation de la limite d’endettement publique comme

Public Debt and Economic Growth: A New Assessment

1. Historical facts of public debt and growth in advanced economies

Figure 1: Public debt as a percentage of GDP over the period 1926–2015

G-20 Advanced G-20 Emerging Low Income

2. Debates about public debt effects on economic growth

2.3. The debate on the fiscal multiplier value

2.4. The debate on the effects of public debt in economic models

3. Interest of the thesis

4. Plan of the thesis

Figure 4: Thesis Structure

Key Literature Review General Introduction Objectives, Results and

On the public debt and growth threshold:

JEL classification: C13, C15, C23, H63, O57.