Populism

Institutional Members: CEPR, NBER and Università Bocconi
WORKING PAPER SERIES
Economic Insecurity and the Demand of

Populism in Europe
Luigi Guiso, Helios Herrera, Massimo Morelli, Tommaso Sonno
Working Paper n. 704
This Version: August, 2023
IGIER – Università Bocconi, Via Guglielmo Röntgen 1, 20136 Milano –Italy

http://www.igier.unibocconi.it
The opinions expressed in the working papers are those of the authors alone, and not those of the Institute,
which takes non institutional policy position, nor those of CEPR, NBER or Università Bocconi.
Economic Insecurity and the Demand of Populism in Europe∗
L. Guiso† H. Herrera ‡
M. Morelli§ T. Sonno ¶
August 2023
Abstract
We document the spiral of populism in Europe and the direct and indirect role of
economic insecurity shocks. Using survey data on individual voting, we make two
contributions to the literature, namely: (1) Economic insecurity shocks have a sig-
nificant impact on the populist vote share, directly as demand for protection, and
indirectly through the induced changes in trust and attitudes; (2) A key consequence
of increased economic insecurity is a drop in turnout. The impact of this largely ne-
glected turnout effect is substantial: conditional on voting, when economic insecurity
increases almost 40% of the induced change in the vote for a populist party comes
from the turnout channel.
Keywords: turnout, trust in politics, voter sentiments

JEL: D72, F68
∗
Luigi Guiso and Massimo Morelli wish to thank the Italian Ministry of Research (MIUR) for the PRIN
funding; Massimo Morelli also wishes to thank the Dondena and Igier research centres and the European
Research Council, advanced grant 694583. We thank Luca Brugnara, Nina Fluegel, and Luca Vitale for
excellent research assistance. The usual disclaimer applies.
†
Einaudi Institute for Economics and Finance, and CEPR
‡
Warwick University, and CEPR
§
Bocconi University, IGIER, and CEPR
¶
University of Bologna, and CEP
1 Introduction
What determined the populism wave in Europe in the 21st century? Are the common
sources related to economic crises or stagnation and, if so, through what channels? This
paper provides an empirical analysis of the channels through which economic insecurity
affected the “demand” of populism. The focus on the common features of populist parties
(rather than a predominance of right-wing orientation) and the focus on a broad notion
of economic insecurity (rather than just globalization shocks) are necessary for a deep
understanding of the phenomenon.
The 21st-century external threats of globalization and migration, as well as the finan-
cial crisis, undermined citizens’ confidence in both leftist (government-based) policies and
rightist (market-based) policies that respect the institutional constraints and functioning
of politics. Global market competition, immigration, and robotization are making some
believers in free markets shake.1 At the same time, the ability of governments to keep
welfare state policies is reduced due to reduced fiscal space and supranational constraints
(particularly in Europe). Facing this two-sided crisis, there is room for new movements
(and existing ones) to promote a radical removal of constraining institutions.2 In this con-
text, a negative economic security shock that affects a citizen at a time when both left
and right traditional recipes are perceived as ineffective, may depress the motivation to
vote for traditional parties on all sides of the political spectrum. This disappointment,
in turn, generates an abstention-based space for populist platforms who thus experience a
massive increase in support. Figure 1 lends support to this rationale. It shows a pattern
familiar to several European countries, namely an economic crisis followed by voter apathy
and disaffection with traditional parties followed, in turn, by more space for entry of new
1
OECD data indeed shows that not only trust in governments and institutions collapsed in the
21st century in Europe (https://www.oecd.org/governance/trust-in-government/), but also trust in mar-
kets collapsed going back to the levels of the 1978 oil-driven crisis (https://www.oecd.org/sdd/leading-
indicators/financialcrisisseescollapseinoecdconsumerconfidence.htm).
2
Various forms of exit, rejection of international treaties previously subscribed, construction of walls,
and so on, are just examples of simple protection commitments that have traction today but would not
have attracted votes in other decades.
1
populist parties or greatly magnified the vote share of existing ones.
Figure 1: Populism, Economics, Electoral Participation and Trust

Populism, economics, turnout and trust
Industrial production (2006=100, Istat) Trust in political parties (2006=100, Eurobarometer)
Voter turnout (2006=100, ParlGovl) M5S (IMG, Piepoli, Index Research, Ixé, Ipsos, Scenari politici)
105 25%
INDUSTRIAL PRODUCTION, TRUST IN POLITICAL PARTIES,
100
95 20%
ELECTORAL PARTICIPATION
90
M5S POOL RESULTS

15%
85
80
10%
75
70 5%
65
60
2005 2006 2007 Populism,
2008 economics,
2009 2010 turnout
2011 and trust
2012 2013 2014 2015 2016
0%
Industrial production ( 2004=100, Eurostat) Trust in political parties (2004=100, Eurobarometer)

Voter turnout (2004=100, ParlGov) Populist parties share (SYRIZA, LAOS, Indep Greek, Golden Dawn)
120 18%
16%
INDUSTRIAL PRODUCTION, M5S SHARE OF
100
14%
VOTES, ELECTORAL PARTICIPATION
POPULIST PARTIES VOTE SHARE

80 12%
10%
60
8%
40 6%
4%
20
2%
0
2003 2005 Populism,
2007 economics,
2009 turnout
2011 and 2013
trust 2015 2017
0%
Industrial production ( 2007=100, Eurostat) Trust in political parties (2007=100, Eurobarometer)

Voter turnout (2007=100, ParlGov) Front National (ParlGov)
105 16%
14%
INDUSTRIAL PRODUCTION, TRUST IN POLITICAL
100
PARTIES, ELECTORAL PARTICIPATION
12%
95 FRONT NATIONAL VOTE SHARE
10%
90
8%
85
6%
80
4%
75 2%
70 0%
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Notes: The figures show the evolution of economic activity, trust in political parties, electoral partic-
ipation and consensus to populist parties in Italy, Greece, and France. Economic activity (measured
by the index of industrial production), the share of the vote going to the populist parties and voter
turnout are on the left scale; trust in political parties on the right scale.
In this paper we offer evidence consistent with the above timing of events: economic
insecurity causes faith in traditional parties to wane, inducing voter disillusion; in turn,
this economic insecurity induced disillusion sparks support for populist platforms. We show
2
that endogenous turnout effects (which have been largely neglected in the literature) are
key in order to quantify the relevance of economic insecurity. This is due to the fact that
economic insecurity shocks affect at the same time the willingness to participate in elections
as well as the willingness to switch to a populist party conditional on voting.
Before turning to the description of our research design, data, and empirical strategy,
a few words are in order to define what we mean by economic insecurity and populist
parties. For the former, many studies have used either unemployment data in isolation
(e.g., Algan et al. 2017) or measures based on exposure to globalization and automation
(e.g., Colantone and Stanig, 2018a). For us, not only do we want to include both of these
types of insecurity, but it is important also to include the specific additional insecurity
caused by the financial crisis (see e.g. Guiso et al., 2022; Schafer et al., 2022). Thus, our
measure will be affected by the most relevant concerns coming from the different crises of
the period.3 For the latter, we will simply adopt the classification of parties as populist or
not that is prevalent in political science, mostly based on their anti-elite rhetoric and pro-
people rhetoric (Rooduijn et al., 2019, and the corresponding Popu-List.org).4 As shown
in Bellodi et al. (2023), the rhetoric employed by populist parties derives its strategic
justification from pandering to people’s fears and increased distrust. Hence, the demand
for populist platforms is due to the fact that populist parties are proposing platforms that
rely less on trusting political representatives.
We study the determinants of the demand for populist platforms in the countries cov-
ered by the European Social Survey. Our empirical analysis accounts for selection into
3
Our main source of data for the measures of economic insecurity, changing over time, is the European
Social Survey, which is available starting in 2002. Changes in economic insecurity responses to survey
questions can reflect a number of threats, some of which may be cultural. We intend by no means to
exclude the possibility that the main triggers behind such reported changes may be cultural threats.
Moreover, economic insecurity perceptions may affect the populist vote in several indirect ways, as for
instance fostering a fear of white-status loss in the case of Trump voters, as documented in Mutz (2018).
4
Underlying the Popu-List.org classification of parties that we use, is the definition of populism proposed
by Mudde (2004): widely accepted in the recent political science literature, it characterizes populism as “a
political narrative that antagonizes the people and the corrupt elite, and that aims for policies that reflect
the will of and are understood by the people”(Mudde, 2004).
3
electoral participation. We show that adverse shocks to economic security and trust in po-
litical parties induce people not to vote and, if they do, to choose a populist party. Ignoring
the voter participation margin would bias the estimates substantially, understating the un-
derlying demand for populist parties; but also would obscure the mechanism by which the
disappointment induced by the crisis favors populists. A simultaneous Heckprobit estima-
tion of participation and populist vote shows that economic insecurity has statistically and
economically significant direct effects on both margins: it lowers the chances of turning
out, but when a vote is cast it raises the chances of voting populist.
The data we use, namely the representative repeated cross-sections of the European
Social Survey from 2002 to 2018 allow us to build a pseudo-panel to show that trust in
politics and immigrant attitudes variables are also affected by changes in economic security.
We conduct a mediation analysis that allows us to establish that the direct effect of economic
insecurity shocks on turnout and populist voting is a lower bound for the role of such shocks,
since such shocks also affect trust and attitudes, which in turn have reinforcing effects on
turnout and vote choice in the same direction. Thus, we can document a large total effect
(direct and indirect) of economic insecurity on the demand for populism.
The most important novel contribution of our empirical analysis concerns the so far
neglected turnout mechanism: more than one third of the increase in the propensity to
vote for a populist party relative to other parties after economic security shocks come from
a turnout effect. Ignoring the turnout channel, one could reach the conclusion (see e.g.
Norris and Inglehart, 2019) that economic variables do not matter much in the decision
to vote for a populist party. Indeed, failure to consider that economic security shocks
significantly affects the decision to abstain makes inconsistent any estimate of the impact
of economic insecurity.5
5
The increase in abstention is due to an increased distrust in political parties, namely to the fact that
voters feel less represented by traditional parties, so they lose interest in politics. As for models that analyze
how economic conditions jointly affect vote choice and turnout a seminal contribution is Weschle (2014),
who in a spatial model with abstention both by alienation and indifference illustrates the importance of
the turnout channel when economic shocks arise.
4
In terms of empirical methodology, the problem to be solved is that the votes for populist
parties depend both on the vote choice among those who vote and, simultaneously, on the
decision to participate. Therefore, in order to identify the role of turnout we need an
exogenous variation in the likelihood to vote which is orthogonal to the party preference.
We use mean temperature and total rainfall on the day of the elections in each region-year.
The identification assumption is that meteorological conditions on election day affect the
cost of going to the polls (differentially for countries where it rains infrequently or where
temperatures are frequently low) but not the preference for voting for a specific party.
While this empirical strategy allows us to properly isolate the turnout effects, we cannot
rule out that in the relationship between economic insecurity and voting for a populist party
other confounders might still play a role. However, to alleviate this concern, we perform a
sensitivity analysis that shows how big the omitted variables would need to be in order to
bring the role of economic insecurity in populist votes to insignificance.
For a review of the literature on populism in the social sciences in general, see e.g.
Gidron and Bonikowski (2013) and Mudde and Kaltwasser (2017), and the most compre-
hensive one by Guriev and Papaioannou (2020). Rodrik (2018) traces the origin of today’s
populism to the globalization shock, arguing that history and economic theory imply that
waves of globalization will predictably lead to a populist backlash, and with specific timing
(when the shock hits) and geographical pattern (in the countries most severely affected).
While the shock of globalization generates demand for populist policies when considered in
isolation and for specific events,6 Guiso et al. (2019) show that globalization shocks alone
cannot account for the cross-country evidence of populist outbreak in Europe. They show
that the interaction of globalization with a euro-dummy captures all the explanatory power
and, in presence of such an interaction variable, globalization shocks alone lose relevance. In
contrast, using the broader notion of economic insecurity that we propose, the interaction
6
Dorn et al. (2020), Colantone and Stanig (2018a,b), Jensen and Bang (2017) are clear examples of well
identified effects of the China shock on specific manifestations like Brexit. Pástor and Veronesi (2021) show
that the backlash against globalization is a response to rising income inequality if aversion to inequality is
assumed in voter’s preferences.
5
effects with institutional variables do not eliminate the significance of economic insecurity.
Algan et al. (2017) study the political consequences of the Great Recession in Europe,
showing that in elections after 2008 the regions where unemployment rose saw the sharpest
decline of trust in institutions and establishment politics. Dustmann et al. (2017) reach
similar results showing that in the aftermath of the crisis mistrust of European institu-
tions, largely explained by the poorer economic conditions of the Euro-area countries, is
correlated with the populist vote. Foster and Frieden (2017) nuance this result using indi-
vidual characteristics from the Eurobarometer survey, and also show that the correlation
is stronger in debtor countries. We contribute to this literature by finding that economic
insecurity affects the consensus for populist parties not directly but primarily because it
disappoints the supporters of the traditional parties of both left and right. This induces
abstention and creates a potential electoral basis for a populist platform.7
The paper is organized as follows: in section 2 we describe the data. We illustrate the
conceptual framework and the econometric specification in section 3. Our main results fol-
low in sections 4 and 5 which quantify the direct and indirect effects of economic insecurity
through turnout and voter sentiments. Section 6 concludes.
2 The Data
Our main source of individual data is the European Social Survey (ESS). The ESS
systematically tracks changing situations, values and attitudes. It covers all European
countries, though not every country participates in every wave. Data has been collected
every two years, since September 2002, by face-to-face interviews. We use eight waves.
The questionnaire consists of a core module, constant from round to round, and smaller
rotating modules, repeated at intervals, on selected substantive topics. We will use the
7
Schafer et al. (2022) document that within-individual changes in income significantly impact partic-
ipation, especially among the poor. For other studies that look at populism and participation look at
Leininger and Meijers (2021) and Huber and Ruth (2017).
6
core module, which covers a wide range of social, economic, political, psychological and
demographic variables. Table A1, in Appendix A, shows summary statistics for all the
variables described below.
Turnout and voting. The ESS asks people whether they voted in the last parlia-
mentary election in their country and which party they voted for. From these we obtained
our turnout variable and constructed a dummy that takes value 1 if the voter voted for a
populist party.8
A second set of data is introduced in order to have variables that affect the cost of
participating in an election but not the voter’s choice of party. To this end, we have collected
data on the weather on the day of the national elections considered, at the NUTS3-region
level. In particular, we have obtained data on the average temperature and precipitation
on election day in each region using the E-OBS dataset provided by the European Climate
Assessment & Dataset project.
Economic insecurity. The key explanatory variable that we construct from the ESS
data is economic insecurity. We capture heterogeneity in economic insecurity with three
measures. First, whether the voter has been unemployed for some time in the past five
years, forcing search for a new job; second, as a measure of financial distress, whether the
voter is experiencing income difficulties, i.e. finds it hard to live on her current income;9
and third, an indicator of exposure to globalization, constructed exploiting information
in the ESS on type of employment, industry and skill level – classifying as more exposed
low-skill workers in manufacturing. The indicator takes value of 1 if the individual is a
8
Responses to the ESS do not necessarily correspond to what people actually did in the voting booth.
However, we can study the aggregate accuracy of the ESS data by aggregating it at the country level and
comparing it with real-world data. Data from ParlGov allows us to compare both turnout and populist
vote share of 127 elections in 32 countries from 1999 to 2018. The correlation between turnout in the
ESS and actual turnout is quite high, i.e. 78%, and the correlation between ESS votes for populist parties
conditional on participation and actual voting for populist parties is even higher, at 87%. However, it is
important to stress that, for the sake of our analysis, the availability of individual-level data is crucial, as
we want to study the impact on turnout and vote choice. Individual-level observational turnout data is
rarely available (e.g., Schafer et al., 2022). Moreover, individual voting choices are never observable in the
first place. Hence the ESS represents the best dataset to answer our research question.
9
Answers range from 1 (“Living comfortably on present income”) to 4 ( “Finding it very difficult on
present income”).
7
blue-collar worker in manufacturing; 0 otherwise. We will find it useful to combine these
three objective measures of financial and economic distress in a single composite index of
economic insecurity by taking the first principal component, re-scaled to vary between 0
(least insecure) and 1 (most insecure). With this measure we are agnostic about the specific
factor causing economic insecurity.
Economic insecurity may also be produced by labor market competition due to immi-
gration. Unfortunately, there are no data on immigration inflows by country of origin and
region of destination, which would enable us to obtain intra-country variation in individual
exposure to labor market pressure. To capture fear of displacement in the labor market
due to the possible arrival of cheap labor, we use a measure of sentiments towards im-
migrants: whether the voter would like fewer immigrants from low-wage countries, with
answers ranging from 1 to 4 increasing in degree of support for immigration quotas. The
ESS also collects people’s attitudes towards quotas on immigrants from countries of the
same race/ethnicity and from countries of different race and ethnicity, as well as whether
people agree with the statement that immigrants make their country worse. Our results
do not change when using different measures.
Trust in traditional politics and institutions. The ESS has several proxies for
confidence in institutions, governments and political parties, all on a scale between 0 (no
trust) and 10 (full trust). These indicators tend to be closely correlated and thus hard to
tell apart. In analyzing individual voting behavior we use trust in political parties, which
speaks directly to our model. In studying the link between economic insecurity and trust
we use all the measures.
Other controls. We enrich the set of explanatory variables with two proxies for voters’
ability to foresee the pitfalls of the populist platforms. The first is education, measured by
the number of years of full-time schooling completed. The second is a measure of attention
to politics, captured by two variables: how many hours per week people devote to watching
TV in general and how many of these hours are spent watching news or programs about
8
politics and current affairs.10 Watching TV in general is taken as a proxy for little interest
in politics, and thus as a proxy for poor information. Watching news and programs about
politics, given the time spent watching TV, is used to proxy for information level. Voting
for an anti establishment party may entail some risk and be more appealing for risk prone
voters. Similarly, sensitivity to policies that offer short term protection at the expense of
long term policies may depend on people’s subjective discount. We use age as a proxy
for subjective discounting, on the presumption that older people are less likely to have to
bear for the future cost of current policies. As a proxy for risk tolerance we use the ESS
indicator of whether people consider it important to avoid taking risks. In all regressions
we control for gender and political orientation, measured on a scale from 0 (far left) to
10 (far right). Needless to say, some of the variables can proxy for more than one of the
dimensions of heterogeneity that we have listed. For instance gender may also reflect risk
preferences as may age.
Populist parties. To identify populist parties in Europe, we rely on the PopuList
proposed by Rooduijn et al. (2019) available at www.popu-list.org. The PopuList is a list of
populist European parties that obtained not less than 2% of the vote in at least one national
parliamentary election since 1998. Peer-reviewed by more than 30 academics, the list is
kept up to date and records changes in the classification of individual parties over time.
All of these features make the classification reliable and useful for our analysis. Rooduijn
et al. (2019) base their classification of populist parties on the classic definition provided
by Mudde (2004).11 Using criteria compatible with Mudde (2004) definition, the authors
identify 82 populist parties in 28 of the 31 countries examined. The full list of parties is
available in Appendix B, Table A2.
10
For wave eight of the ESS we use the variables “internet use time” and “time spent watching/listening
to/reading the news” since the questions on media use have been slightly changed.
11
Mudde (2004) defines a party as populist if (a) it endorses the set of ideas that society is ultimately
separated into two homogeneous and antagonistic groups, “the pure people” versus “the corrupt elite,” and
(b) it argues that politics should be an expression of the general will of the people.
9
3 Methodology
In order to empirically model demand of populism accounting for endogenous voter par-
ticipation, we model voting as a two-step decision: a) whether to participate in an election
(the participation decision); and b) conditional on participation, which party to vote for
– in particular, whether or not to vote for a populist party (the voting decision). Esti-
mating the turnout and vote choice decisions simultaneously is important for two related
but distinct reasons: first, to get consistent estimates of the voting decision if unobserved
components of the participation decision are correlated with unobserved components of the
voting decision. Second, to pin down the channels through which voters’ characteristics
impact vote choice.
Denoting by z a variable that affects only the participation decision and by x a variable
that affects both the participation and the party choice, note that our dependent variable
of interest, namely the probability of voting for a populist party conditional on voting,
denoted by π C (x), must be equal to the ratio of the joint distribution and the marginal
probability of turning out, namely π J (x, z)/π V (x, z). π J (x, z) is the joint probability of
voting and preferring a populist party, which is basically what one estimates when ignoring
the turnout incentives. The effect of a change in x, say an increase in economic insecurity,
on the conditional probability of voting is
πxC = (πxJ π V − πxV π J )/(π V )2
or, in percentage terms,

πxC /π C = πxJ /π J − πxV /π V . (1)
Equation (1) clarifies that the effects of a change in economic insecurity on the conditional
probability of voting for a populist (in percentage of the sample mean), which is our variable
of interest on the LHS, is a sum of two effects, where the first one on the RHS is the standard
effect on the joint distribution, whereas the second one comes entirely from the neglected
10
turnout incentives.12
Jointly estimating voting and participation decisions we retrieve consistent estimates of
πxC and πxV and can assess the role of turnout in the voting results.
3.1 Instrumented Heckprobit
To deal with the issues related to the fact that people decide whether to vote or not
and then whom to vote for conditional on voting, we estimate a two-step Heckman probit
model, estimating first the probability of participation, and then the probability of voting
for the populist party adjusting for selection.
Electoral participation depends on the same set of variables as the choice of the party,
possibly with opposite signs. For identification, we need a personal characteristic - an
instrument - that affects the net benefit of voting (benefit less cost), but not the choice of the
party conditional on participation. As instruments, we use here the mean temperature and
total rainfall on the day of the elections in each region-year. The identification assumption
is that meteorological conditions on the election day affects the cost of going to the polls
but not the preference for voting for a specific party, which should reflect less transient
factors. Because the effect of rain or heat on the cost of going to the polls may be stronger
in countries where it rains infrequently (or where temperatures are frequently low) we also
include interactions between rainfall and temperature with a dummy variable for southern
countries.
12
Since the turnout depression effects of an increase in economic insecurity will be shown to be significant
and large, the negative sign in front of the last term determines a positive – and strong – effect on the LHS
variable.
11
3.2 Econometric Specification
Formally, we estimate the following selection equation:
voted irct =γ1 xirct + γ2 rain rct + γ3 rain rct × south c +
γ4 av. temperature rct + γ5 av. temperature rct × south c + (2)
γ6 EIict + fc + ft + ict
and the following second stage equation:
voted populist irct = α1 xirct + α2 EIirct + fc + ft + virct (3)
where voted irct is a dummy variable assuming value one if the person i (living in region
r, belonging to country c) voted in period t; xjct the vector of controls; rain rct is the
daily total rainfall in the day of the election in region r ; south c is a dummy indicating
southern countries (Spain, Portugal, Italy, Greece, Cyprus, Malta); av. temperature rct is
the daily mean temperature in region r the day of the election; EIjct the index of economic
insecurity;13 voted populist ict is a dummy assuming value one if individual i voted for a
populist party; fc and ft are country- and time-specific fixed effects; and uirct and virct are
error terms.
We start estimating our model on the sample of countries that have a populist party
in the ESS waves. Later we extend the estimates to all countries and account for selection
induced by populist party existence/entry. As we will see, results are unaffected, suggest-
ing that the included controls already capture the variables that affect populist parties’
presence. In all specifications we control for gender and political orientation and for the
population of the voter’s region; we also include country-level fixed effects and ESS wave
fixed effects. Importantly, country-fixed effects capture all the time-invariant features of
13
Note that in the first econometric specification, we do not use the synthetic index of economic insecurity,
but we regress its three components to show their differential effects on the selection and second stage
equation.
12
the country that may affect the success of populist platforms: the electoral system, the
responsiveness of the established parties to salient political issues (such as labor market
pressure from immigrants), and the level of corruption.14 For brevity, these controls are
not reported. We run regressions using sampling weights to account for differences in the
national’s sample size. In all regressions, standard errors are clustered at the regional level.
Our final dataset consists of more than 142 thousand observations from 25 countries when
estimating the specification with all controls.
4 Results
Some recent papers on the drivers of populism make the prediction that a sharp de-
crease in economic security and trust leads to increased demand and supply of pandering
commitments and consequent populist platforms – see e.g. Bellodi et al. (2023) and ref-
erences therein. The main contribution of our analysis is to confirm this prediction while
establishing an important and neglected factor, namely that a large component of the eco-
nomic insecurity effect on the conditional probability of voting populist comes via a turnout
depression effect. The importance of taking into account turnout incentives in the estima-
tion of voting choices has been established by Weschle (2014), and we will show that the
endogenous turnout role is particularly strong in our context.
In this section we first display some descriptive statistics on economic insecurity, with
preliminary hints about populist voting and turnout depression predictions; then we show
how important for a correct and consistent estimation of the former is to consider the
endogenous turnout depression effect.
14
These are some of the context variables that studies of populism (e.g. van Kessel, 2015) consider
critical in explaining populists’ success.
13
4.1 Descriptives on Economic Insecurity
To illustrate first the distribution of economic insecurity for our sample, we create a
dummy variable "economic insecure" that is equal to 1 if the respondent fits at least one
of the three criteria for insecurity (she/he was unemployed in the last 5 years, or she/he
is exposed to globalization, or her/his income difficulty variable - ranging from 0 to 3 -
is 3). The percentage of respondents classified as economically insecure is 25%. Figure
2 shows the percentage of economically insecure respondents among the respondents who
voted populist or not. As we expect, among those who did not vote populist, only slightly
more than 20% are economically insecure, while among the people who voted populist,
economically insecure are around 30% - a proportion that is 50% higher.
Figure 2: Economic Insecure by Populist Vote
Notes: The figure shows the share of individuals classified as economically insecure
by populist vote.
Figure 3 breaks down the continuous economic insecurity measure by terciles (in each
country) and then reports the share of populist supporters (left figure) and turnout (right
figure) in each tercile. The average vote share among the individuals belonging to the first
tercile of the economic insecurity distribution is below 10%, it increases to around 13% and
to above 15% in the second and third tercile, respectively. Symmetrically, the turnout share
14
among the individuals in the three terciles is, respectively, above 84%, approximately 76%,
and slightly above 70%. These patterns suggest that both phenomena, namely the positive
impact of economic insecurity on populist voting and the negative impact of economic
insecurity on turnout, are present everywhere along the distribution of economic insecurity
without visible discontinuities.
Figure 3: Populist Vote and Turnout by Terciles of Econ. Insecurity
Notes: The figure shows the populist vote (left figure) and turnout share (right figure) by
terciles of economic insecurity. Consistent with our theory, the first tercile presents higher
turnout and lower populist vote and vice versa for the last tercile.
4.2 Main analysis
Table 1 reports the estimates of several specifications, with a progressively augmented

set of controls. The bottom part shows the parameter estimates of the meteorological
instruments on the participation decision. In general, rainfall on election day discourages
participation. This effect is stronger for southern countries. Additionally, we can observe
that participation increases when the temperature is higher with similar effects in Southern
and Northern countries. Increasing temperature on the day of the election by one standard
deviation increases turnout by 4.9% of the sample mean, while a one standard deviation
15
increase in the daily total rainfall on the day of the election lowers turnout by 1.5%. This
conforms with intuition: higher temperature (relative to the country mean captured by the
country fixed-effects) is a good motivation to go to the polls in all countries while going
to vote in the rain is costly – even more so in southern countries where people are less
equipped for it.15 Conditional on the controls and the instruments there is some sign of
selection bias, as shown by the significant correlation between the residuals in the voting
and the participation regressions in all specifications.
The first two columns show the results of participation and voting decisions controlling
for risk and time preferences, education, political information, and the three proxies for
economic insecurity. The proxy for risk aversion has a significant positive effect on partici-
pation: people who consider it important to avoid taking risks are more likely to vote. This
measure has no effect on the choice to vote for a populist party. Hence, we find no support
in the data for the idea that since the populist choice entails risk, it is more appealing to
risk-tolerant voters. Interestingly, women are less likely to participate, and when they do,
they are also less likely to support populist platforms; while the politically right-leaning
are more likely to participate.16 Education has a positive and precisely measured effect
on voting and, conditional on participation, a negative effect on support for a populist
party. The proxy for political information has a significant impact on turnout - more po-
litically informed citizens are more likely to participate, while its relevance decreases in the
specification with full controls.
Our study confirms the importance of the economic insecurity mechanism.17 Economic
15
A Chi-square test for the joint significance of the instruments in the first stage Heckman model strongly
rejects the null that the instruments are jointly equal to zero with a p-value as low as 0.0046. This suggests
that the instruments do have high predictive powers.
16
As one could be concerned that political orientation might be correlated with the likelihood of voting
for a populist party, in Appendix C we replicate Table 1 excluding this control from the analysis. Results
are robust to the exclusion of this control variable.
17
As already mentioned in the introduction, the economic insecurity variable captures the individual
perception of her economic insecurity (this is obviously always true in survey studies of this kind), and
hence the sources of changes in such perception do not need to be only economic. Even cultural or social
shocks can alter the perception of economic insecurity. Hence our findings by no means reflect the will to
establish the primacy of economic over cultural determinants of populism.
16
Table 1: Heckman probit estimates of populist party vote and participation in voting
(1) (2) (3) (4)

Heckprobit Heckprobit Heckprobit Heckprobit
Populist Vote Populist Vote Populist Vote Populist Vote
Risk aversion -0.00256 0.0142*** -0.00161 0.0137*** -0.00264 0.0155*** -0.00358 0.0169***
(0.00800) (0.00390) (0.00803) (0.00389) (0.00845) (0.00413) (0.00861) (0.00419)
ln(Age) -0.306*** 0.762*** -0.292*** 0.751*** -0.301*** 0.768*** -0.324*** 0.780***
(0.0713) (0.0268) (0.0695) (0.0267) (0.0700) (0.0274) (0.0722) (0.0279)
ln(Education) -0.436*** 0.527*** -0.451*** 0.539*** -0.437*** 0.526*** -0.387*** 0.521***
(0.0536) (0.0360) (0.0531) (0.0358) (0.0521) (0.0371) (0.0544) (0.0377)
TV total 0.0171** -0.0208*** 0.0177** -0.0210*** 0.0186*** -0.0191*** 0.0138* -0.0192***
(0.00722) (0.00454) (0.00716) (0.00451) (0.00711) (0.00442) (0.00714) (0.00434)
TV politics -0.0265*** 0.0557*** -0.0264*** 0.0558*** -0.0205** 0.0489*** -0.0169** 0.0484***
(0.00838) (0.00625) (0.00835) (0.00627) (0.00841) (0.00642) (0.00846) (0.00633)
Unemployment 0.147*** -0.167***
(0.0319) (0.0184)
Income difficulties 0.209*** -0.154***
(0.0144) (0.0107)
Exposure globalization 0.0529 -0.113***
(0.0345) (0.0229)
Economic insecurity (PC) 0.813*** -0.718*** 0.726*** -0.656*** 0.707*** -0.650***
(0.0634) (0.0412) (0.0620) (0.0425) (0.0645) (0.0430)
Trust in pol. parties -0.0783*** 0.0473*** -0.0719*** 0.0461***
(0.00611) (0.00371) (0.00607) (0.00376)
Few immigrants from no-EU 0.144*** -0.0287***
(0.0177) (0.00803)
Controls, Wave FE, Country FE Yes Yes Yes Yes

Rho -0.391 -0.400 -0.375 -0.373
Cluster SE Region Region Region Region
Countries With P With P With P With P
Observations 155,506 155,506 145,877 142,849
Censored observations 49,509 49,509 46,421 45,099
Selection
Rain -0.00708** -0.00712** -0.00728** -0.00750**

(0.00292) (0.00288) (0.00316) (0.00324)
Rain × South -0.0166** -0.0164** -0.0150* -0.0145*
(0.00807) (0.00805) (0.00792) (0.00771)
Av. Temperature 0.0178*** 0.0179*** 0.0206*** 0.0208***
(0.00612) (0.00612) (0.00635) (0.00639)
Av. Temperature × South -0.00896 -0.00925 -0.00831 -0.00511
(0.0118) (0.0118) (0.0117) (0.0116)
Notes: The table shows Heckman probit estimates of the decisions to vote (Vote) and to vote for a populist party conditional on participation (Populist). Left-hand side variables: a dummy if a voter has
chosen a populist party in the columns Populist and a dummy if (s)he has participated in the election in the column Vote. The excluded instrument in the populist regression is an indicator of weather
condition on election day. All regressions include country and wave fixed effects. Robust standard errors clustered at the region level are shown in parenthesis. *** significant 1% or less; ** significant at
5%; * significant at 10% confidence level.
17
insecurity changes crucially act on two margins: discouraging participation and increasing
the likelihood of a populist vote among those who do decide to vote. The effect on the
participation margin is precisely estimated and highly responsive to unemployment, income
loss, and exposure to globalization. The populist vote is more likely among those who lost
a job, suffer an income loss and are exposed to globalization, despite the latter being not
significantly estimated under standard thresholds.
To facilitate the interpretation of the magnitude of the effects of economic insecurity,
the second set of regressions replaces the three measures of economic insecurity with their
principal component. The index of economic insecurity significantly affects electoral partic-
ipation and voting for the populist party. At sample means, increasing economic insecurity
by one standard deviation lowers turnout by 6.3% of the sample mean and increases the
populist vote by 17%. For an individual who transits from no economic insecurity to eco-
nomic insecurity, the probability of voting for a populist party increases by 12.7 percentage
points (82% of the unconditional sample mean), while the probability of voting falls by as
much as 24 percentage points, equivalent to 30% of the sample mean. These are substantial
effects.18
The third pair of columns have trust in political parties as an additional explanatory
variable. Consistent with our proposed interpretation of the role of disappointment with
politics in the rise of populism, people with greater confidence in political parties are more
likely to vote and to vote for a non-populist party. Those who have lost faith in political
parties are more likely to abstain, but if they do vote, they are more likely to choose
a populist party. Trust in political parties is on a scale of 0 to 10; a drop of 5 points
increases the probability of voting for a populist party by 10% of the sample mean. The
effect on electoral participation is similarly strong: a drop of 5 points lowers the chance of
18
As already underlined in the introduction, while our empirical strategy allows us to properly isolate
the turnout effects, in the relationship between economic insecurity and voting for a populist party we
cannot rule out that other confounders might still play a role. To alleviate this concern, in Appendix D
we perform a sensitivity analysis that shows how big the omitted variables would need to be in order to
bring the role of economic insecurity in populist votes to insignificance.
18
participating in elections by 6.7 percentage points, almost 44% of the unconditional mean
electoral turnout.
The last pair of columns add, as a control, a measure of attitudes towards immigrants,
used as a proxy for fears of competition in the labor market. Support for policies that
limit immigrants from non-EU countries, support for limiting immigrants of the same
race/ethnicity or immigrants of other race/ethnicity than that of the respondent, or an
average of the three measures, all have the same implications: people who are more averse
to immigrants are less likely to vote and more likely to vote for a populist party if they do.
A 1-standard-deviation increase in hostility to immigrants lowers turnout by 1 percent of
the sample mean; the effect on voting for a populist party is more pronounced: it increases
by 15.8% of the sample mean. The effects of the other variables, particularly economic
insecurity and trust in political parties, are unchanged.
4.3 The magnitude of the turnout effect
In the first column of Table 2 we compute the LHS of equation (1), namely the per-
centage change in the conditional probability of voting populist, changing each of the
independent variables by 1 standard deviation. In the second column, we compute the
second term of the RHS of equation (1), namely the percentage change in the propensity to
vote when changing the same three variables by 1 standard deviation. We are interested in
establishing how large is the neglected turnout effect (the second column) as a component
of the overall effect (the first column).
When considering economic insecurity, our main variable, the neglected turnout effect is
38%, a very large fraction.19 Similarly, if one focuses on another key variable like trust, the
effect of the decrease in turnout incentives amount to roughly 25% of the overall effect, while
for anti-immigrant sentiment this contribution is lower, around 7%. In sum, accounting
for the effects on the decision of whether or not to vote is crucial to understand how the
19
Computations are described in detail in Appendix E.
19
drivers of populist voting operate.
Table 2: Direct effects and effects via turnout
Effect of Contribution via

conditional prob of turnout
1SD over sample mean
Economic insecurity (PC) 0.150 0.057
Trust in pol. parties -0.186 -0.046
Few immigrants from no-EU 0.158 0.011
Notes: The table shows the direct effect on voting for a populist party of a 1-standard-deviation in-
crease in Economic insecurity, Trust in political parties, and attitudes towards immigrants respectively
(first column) and the contribution through the change induced in turnout. Calculations use estimates
in Table 1, column 4.
4.4 Robustness
Table 3 presents several robustness exercises of the results in Table 1. The first two
columns run the estimates of the Heckman probit using all the sample countries, not only
those that have a populist party. That is, the turnout equation is estimated using obser-
vations for countries both with and without populist parties. The endogenous presence
of populist parties is fully captured by the country dummies. The results are unaffected.
Economic insecurity lowers participation and increases the populist vote; the effects are
significant and of the same order of magnitude as those in Table 1. The same holds true for
the effects of trust in parties and the other controls. The next two columns add country-
wave fixed effects, capturing changes in populist manifestos and rhetoric. Again the results
are unchanged. One concern is that the populist vote may actually be capturing voting for
a new party as such. To address this, in the last two columns we run the estimates after
dropping individuals who voted for any new party - i.e. a party present in the election for
the first time. The results are basically unaffected. Table A4 in Appendix F reports the
first stages of Table 3.
Additional robustness of our results can be provided by showing that the main findings
could be qualitatively replicated by looking at linear probability models. In Appendix G we
regress our main variable of interest, together with the same set of controls and fixed effects
of the main specifications, on three different binary outcomes: voted, voted for populist,
20
Table 3: Robustness
(1) (2) (3)

Heckprobit Heckprobit Heckprobit
Populist Vote Populist Vote Populist Vote
Risk aversion -0.00272 0.0177*** 0.00156 0.0170*** -0.000808 0.0160***

(0.00850) (0.00407) (0.00870) (0.00447) (0.00862) (0.00412)
ln(Age) -0.197*** 0.763*** -0.205*** 0.797*** -0.282*** 0.767***
(0.0603) (0.0279) (0.0568) (0.0270) (0.0708) (0.0272)
ln(Education) -0.367*** 0.427*** -0.372*** 0.542*** -0.415*** 0.525***
(0.0527) (0.0399) (0.0522) (0.0365) (0.0535) (0.0374)
TV total 0.0169** -0.0209*** 0.0182** -0.0252*** 0.0183** -0.0187***
(0.00717) (0.00383) (0.00722) (0.00441) (0.00729) (0.00446)
TV politics -0.0150* 0.0506*** -0.0185** 0.0528*** -0.0176** 0.0491***
(0.00866) (0.00537) (0.00895) (0.00627) (0.00845) (0.00646)
(0.0592) (0.0365) (0.0601) (0.0480) (0.0655) (0.0426)
Trust in pol. parties -0.0730*** 0.0456*** -0.0825*** 0.0522*** -0.0789*** 0.0476***
(0.00613) (0.00340) (0.00615) (0.00376) (0.00623) (0.00376)
Controls Yes Yes Yes

Wave FE Yes No Yes
Country FE Yes No Yes
Wave × Country FE No Yes No
Rho -0.150 -0.185 -0.331
Cluster SE Region Region Region
Countries All With P With P
(no new P)
Observations 177,567 145,877 143,581
Censored observations 56618 46421 46421
Notes: The table shows robustness Heckman probit estimates of the decisions to vote and to vote for a populist party. Left-hand side variables: a dummy
if a voter has chosen a populist party in the columns Populist, and a dummy if (s)he has participated in the election in the column Vote. The excluded
instrument in the populist regression is an indicator of weather conditions on election day. The first set of regressions includes all countries, not only those
with a populist party; the second set controls for interacted country-wave fixed effects; the last set runs the regressions dropping observations of individuals
who voted for a new party. The first set of and last set of regressions include country and wave fixed effects. Robust standard errors clustered at the region
level are shown in parenthesis. *** significant 1% or less; ** significant at 5%; * significant at 10% confidence level.
21
and votes for non-populist parties, showing that results are in line with our main findings.
Another instructive robustness check is seeing whether the role of turnout incentives
shows up also when estimating a reduced form of populist voting where we consider the
interaction between economic insecurity and weather-related incentives to vote on the right-
hand side. Indeed, if we run a regression mimicking the second stage of our Heckprobit, but
interacting the economic insecurity variable with a dummy summarizing weather conditions
(details below), we see that broadly consistent with our exclusion restriction, the weather
condition we will now describe does not matter for voting populist, but it magnifies populist
vote. In particular, the dummy in this specification takes value one when on the election
day rainfall level was below the country-specific median and the temperature was higher
than the country-specific median and, therefore, this can be considered a “good voting
condition” dummy. What this reduced form shows, in a nutshell, is that (both in a Probit
and in a linear probability model setting), economic insecurity matters more for populist
voting under good voting conditions. Results and further details on the specification are
reported in Appendix J.
In Appendix H, we test the robustness of our main results, i.e. Table 1, when adding a
specific control to alleviate the concern that the weather on the day of the election may be
indicative of the weather of a longer time period in a specific country.20
5 Trust in politics and attitudes toward immigrants
In our main results, we have treated trust and attitudes toward immigrants as controls.
However, in this section, we want to argue that these two variables may themselves be
affected by changes in economic insecurity. If this is the case, then the above results on
the impact of economic insecurity on populist voting and turnout would be a lower bound
20
Specifically, we compute the average temperature in the country in the wave-year, and divide it by the
“typical” temperature in the country (over more than 100 years), and add this as a control.
22
of the total effect, which is in fact partly mediated by trust and attitudes.21
As a first hint of the existence of such a mediation effect in our context, we present
a quick mediation analysis using OLS, and we then turn to a more sophisticated analysis
through the construction of a pseudo-panel.
5.1 Mediation analysis preliminary hints
We perform a mediation analysis using the populist vote as the dependent variable,
economic insecurity as the independent variable, and either trust or attitudes towards
immigrants as mediators in a “standard” OLS setting. This simple setting allows us to
preliminary study the role of the two channels in mediating the impact of economic insecu-
rity on populist vote share using three different methods, namely Delta, Sobel, and Monte
Carlo.22 The results of the analysis confirm our prior: (i) about 11% of the effect of eco-
nomic insecurity on populist votes is mediated by trust in political parties, and (ii) about
7% of the effect of economic insecurity on the populist vote is mediated by attitudes toward
immigrants. These results are statistically significant using the three different methodolo-
gies and, therefore, indicate partial mediation. In other words, our results suggest that
there are significant “cultural channels” that mediate the effect of economic insecurity on
populist vote share, though small in magnitude.
5.2 Synthetic Panel Model
Economic insecurity and trust in political parties are negatively correlated when gauged
using cross-sectional variation in the pooled ESS. Similarly, economic insecurity is corre-
21
The negative impact of economic insecurity on trust has already been established in other contexts:
Ananyev and Guriev (2019) isolate the causal effect of economic downturns on people’s trust during the
2009 recession in Russia, exploiting regional variations in the industrial structure inherited from the Soviet
Union, and noticing that capital-intensive and oil-related industries are more responsive to shocks to GDP.
They find that a decline in GDP causes a sizeable drop in trust in other people.
22
To perform this analysis, we use the Stata package medsem, which provides a post-estimation command
testing mediational hypotheses using Baron and Kenny (1986) approach modified by Iacobucci et al. (2007)
as well as an alternative approach proposed by Zhao et al. (2010) after estimating the concerned mediational
model with the built-in sem command of Stata.
23
lated positively with hostility to immigrants from non-EU countries. These correlations
hold even controlling for observable and country and wave-fixed effects.23 Of course, the
correlations may just reflect unobserved heterogeneity - i.e. some individual characteristics
that drive both economic insecurity and people’s trust in politics and attitudes towards
immigrants. To partially address this problem, we follow Deaton (1985) and construct a
pseudo-panel from the sequence of ESS waves. Panel B of Table A1, in Appendix A, shows
summary statistics for all the variables used in this section. We group the data into fourteen
5-year age cohorts of men and women in each country, respectively. Figure 4, left panel,
shows a simple bivariate correlation between the change in trust in political parties and
that in economic insecurity among the pseudo-panel cohorts. In all cases, an increase in
the economic insecurity of the cohorts leads to a decrease in trust in political parties. The
right panel shows the bivariate correlation between changes in attitudes towards EU immi-
grants and changes in economic insecurity for the same cohorts. This second correlation is
strongly positive.
We can now introduce an econometric specification aimed at strengthening the identifi-
cation of the economic insecurity effect using cohort fixed effects. Specifically, we estimate
the following model:
yjct = β1 xjct + β2 EIjct + fj + fct + ujct (4)
where yjct denotes the generic belief/attitude of cohort j in country c in year t, xjct the vec-
tor of controls, EIjct the index of economic insecurity, and ujct an error term. Unobserved
heterogeneity is controlled for by the cohort-specific fixed effects fj .24 Country-specific
trends in beliefs/attitudes and economic insecurity are captured by country-year fixed ef-
fects fct . The latter picks up any country aggregate variable that affects changes in beliefs
23
Our interpretation is supported by the results in Algan et al. (2017) who show that in regions of
Europe where unemployment increased more sharply following the 2008 crisis, trust in parties and political
institutions fell more and sentiments towards immigrants deteriorated. An IV analysis suggests that the
causality runs from changes in unemployment to changes in trust and sentiments.
24
Our pseudo-panel consists of 840 age/country/year-of-birth groups. Cohorts are relatively large, with
358 observations on average. This reassures us that measurement error in the cohort means is likely to be
negligible. Dropping cohorts with fewer than 50 observations (4.8% of the total) does not alter the results.
24
Figure 4: Economic insecurity, trust, and sentiments
.4
.1
Change in negative sentiments towards immigrants
.2
Change in trust in political parties
.05
0
0
-.2
-.05
-.4
-.1 -.05 0 .05 .1 -.1 -.05 0 .05 .1

Change in economic insecurity Change in economic insecurity
Notes: The figure shows the binned scatterplot (20 equal-sized bins) and linear regres-
sions of the change in economic insecurity (x-axis) and the change in trust in political
parties (y-axis, left figure, 4,166 observations) and attitudes against immigrants (y-axis,
right figure, 4,726 observations) in the synthetic cohorts panel.
over time, including any effect of populist party rhetoric.

Table A7 in Appendix I shows that the coefficient of economic insecurity on trust and
attitudes is significant even in this model with cohort-fixed effects. The effects are substan-
tial: a 1-standard-deviation increase in economic insecurity lowers trust in political parties
by 7.1% of its sample standard deviation and increases hostility to non-EU immigration by
5% of its sample standard deviation.25
The pseudo-panel regressions with fixed effects should give more confidence in the pos-
sibility that economic insecurity indeed affects trust and attitudes, but without reaching
the standard of full identification. Note, however, that reverse causality – people who lose
trust in parties and because of this are more likely to lose their jobs or suffer income losses
– does not seem to be plausible, particularly in light of the fact that any effect that a
generalized loss of confidence in politics has on the economy is already picked up by the
25
The table in the appendix also shows the robustness of these results with different measures of both
trust and attitudes towards immigrants.
25
time fixed effects and similarly for a change in attitudes towards immigrants.
6 Conclusions
Western countries in the 21st century have experienced an unprecedented sequence of

crises that have affected global markets and sovereign states, leaving many people on unsta-
ble grounds in a way unseen before. The rare combination of the inability of both markets
and governments to provide security has shaken the confidence in traditional political par-
ties and institutions, induced frustration and fears aggravated by growing threats from
mass immigration and globalization. This paper has described how this dual crisis, re-
flected in people’s economic insecurity, has systematically affected the demand for populist
policies. Most important, in line with the insights of Weschle (2014) on the importance
of accounting for endogenous turnout when estimating the voting choices, we have shown
that alienation-induced abstention, largely ignored by previous literature, explains almost
40 percent of the total effect of economic insecurity on the conditional probability of voting
populists. We have also shown that there are also additional indirect effects of economic
insecurity that are non-negligible, but the main contribution of the paper is certainly the
identification and quantitative evaluation of the turnout depression component of populist
voting.
The future will tell us whether the populists who won elections and make policies that
will make economic insecurity ultimately even higher, and what is this going to entail. In
particular, Funke et al. (2020) show that once in office the populist has caused in history
significant negative consequences for economic outcomes, and research should focus on
whether this is going to create a vicious cycle or determine a backlash.
26
References
Algan, Y., S. Guriev, E. Papaioannou, and E. Passari (2017). The european trust crisis
and the rise of populism. Brookings Papers on Economic Activity 2017 (2), 309–400.
Ananyev, M. and S. Guriev (2019). Effect of income on trust: evidence from the 2009
economic crisis in russia. The Economic Journal 129 (619), 1082–1118.
Baron, R. M. and D. A. Kenny (1986). The moderator–mediator variable distinction in so-

cial psychological research: Conceptual, strategic, and statistical considerations. Journal
of personality and social psychology 51 (6), 1173.
Bellodi, L., M. Morelli, A. Nicolo, and P. Roberti (2023). The shift to commitment politics
and populism: Theory and evidence. CEPR Press Discussion Paper No. 18338 .
Cinelli, C. and C. Hazlett (2020). Making sense of sensitivity: Extending omitted variable
bias. Journal of the Royal Statistical Society Series B-Statistical Methodology 82 (1),
39–67.
Colantone, I. and P. Stanig (2018a). Global competition and brexit. American political
science review 112 (2), 201–218.
Colantone, I. and P. Stanig (2018b). The trade origins of economic nationalism: Import
competition and voting behavior in western europe. American Journal of Political Sci-
ence 62 (4), 936–953.
Deaton, A. (1985). Panel data from time series of cross-sections. Journal of economet-
rics 30 (1-2), 109–126.
Dorn, D., G. Hanson, K. Majlesi, et al. (2020). Importing political polarization? the
electoral consequences of rising trade exposure. American Economic Review 110 (10),
3139–83.
27
Dustmann, C., B. Eichengreen, S. Otten, A. Sapir, G. Tabellini, and G. Zoega (2017).
Europe’s trust deficit: Causes and remedies.
Foster, C. and J. Frieden (2017). Crisis of trust: Socio-economic determinants of europeans

confidence in government. European Union Politics 18 (4), 511–535.
Funke, M., M. Schularick, and C. Trebesch (2020). Populist leaders and the economy. AER
(forthcoming).
Gidron, N. and B. Bonikowski (2013). Varieties of populism: Literature review and research
agenda.
Guiso, L., H. Herrera, M. Morelli, and T. Sonno (2019). Global crises and populism: the
role of eurozone institutions. Economic Policy 34 (97), 95–139.
Guiso, L., M. Morelli, T. Sonno, and H. Herrera (2022). The financial drivers of populism
in europe. CEPR Discussion paper No. DP17332 .
Guriev, S. and E. Papaioannou (2020). The political economy of populism. Journal of

Economic Literature (forthcoming).
Huber, R. A. and S. P. Ruth (2017). Mind the gap! populism, participation and represen-
tation in europe. Swiss Political Science Review 23 (4), 462–484.
Iacobucci, D., N. Saldanha, and X. Deng (2007). A meditation on mediation: Evidence

that structural equations models perform better than regressions. Journal of consumer
psychology 17 (2), 139–153.
Jensen, M. J. and H. P. Bang (2017). Populism and connectivism: An analysis of the

sanders and trump nomination campaigns. Journal of Political Marketing 16 (3-4), 343–
364.
28
Leininger, A. and M. J. Meijers (2021). Do populist parties increase voter turnout? evidence
from over 40 years of electoral history in 31 european democracies. Political Studies 69 (3),
665–685.
Mudde, C. (2004). The populist zeitgeist. Government and opposition 39 (4), 541–563.
Mudde, C. and C. R. Kaltwasser (2017). Populism: A very short introduction. Oxford

University Press.
Mutz, D. C. (2018). Status threat, not economic hardship, explains the 2016 presidential
vote. Proceedings of the National Academy of Sciences 115 (19), E4330–E4339.
Norris, P. and R. Inglehart (2019). Cultural backlash: Trump, Brexit, and authoritarian
populism. Cambridge University Press.
Pástor, L. and P. Veronesi (2021). Inequality aversion, populism, and the backlash against
globalization. The Journal of Finance 76 (6), 2857–2906.
Rodrik, D. (2018). Populism and the economics of globalization. Journal of international

business policy 1 (1), 12–33.
Rooduijn, M., S. Van Kessel, C. Froio, A. Pirro, S. De Lange, D. Halikiopoulou, P. Lewis,

C. Mudde, and P. Taggart (2019). The populist: An overview of populist, far right, far
left and eurosceptic parties in europe.
Schafer, J., E. Cantoni, G. Bellettini, and C. Berti Ceroni (2022). Making unequal democ-
racy work? the effects of income on voter turnout in northern italy. American Journal
of Political Science 66 (3), 745–761.
Weschle, S. (2014). Two types of economic voting: How economic conditions jointly affect
vote choice and turnout. Electoral Studies 34, 39–53.
Zhao, X., J. G. Lynch Jr, and Q. Chen (2010). Reconsidering baron and kenny: Myths
and truths about mediation analysis. Journal of consumer research 37 (2), 197–206.
29
Appendix
A Descriptive statistics
Panel A of Table A1 presents descriptive statistics of the ESS data at the individual
level, therefore, used for the Heckprobit estimations. Panel B, instead presents descriptive
statistics of the ESS data aggregated at the cohort-level, for the pseudo-panel analysis.
Table A1: Descriptive statistics
Variable Obs. Mean St. Dev. Min Max
A. Demand analysis
Voted 270,777 0.78 0.41 0 1
Vote for populist party 175,521 0.15 0.36 0 1
Risk aversion 279,621 3.94 1.43 1 6
Age 287,968 49.43 17.86 18 100
Education 289,218 12.84 3.99 0 25
TV total 267,308 4.33 2.06 0 7
TV politics 280,590 2.13 1.45 0 7
Female 288,947 0.53 0.50 0 1
Right wing 255,451 5.12 2.17 0 10
Regional population (1000) 254,568 2561 3504 28 18075
Unemployment 287,944 0.13 0.34 0 1
Income difficulties 283,463 1.00 0.86 0 3
Exposure to globalization 262,146 0.30 0.46 0 1
Economic insecurity (PC) 256,807 0.22 0.21 0 1
Trust in political parties 256,871 3.59 2.36 0 10
Want less immigrants from outside EU 279,413 2.55 0.90 1 4
Daily total rain fall 255,235 2.84 4.87 0 35
Daily mean temperature 255,118 10.05 6.81 -12 27
B. Pseudo panel analysis

Risk aversion 6,013 4.14 0.55 2 6
Age 6,071 56.58 16.84 22 90
Education 6,071 11.56 2.32 3 19
TV total 5,978 4.36 0.82 1 7
TV politics 6,071 2.29 0.63 0 7
Female 6,071 0.50 0.50 0 1
Right wing 6,069 5.17 0.66 0 10
Regional population (1000) 5,109 2397 2544 118 14375
Economic insecurity (PC) 6,013 0.22 0.09 0 1
Trust in political parties 5,455 3.49 1.12 0 8
Want less immigrants from outside EU 6,071 2.65 0.39 1 4
Trust politicians 6,071 3.55 1.11 1 8
Trust national parliament 6,070 4.40 1.23 0 9
Trust European parliament 6,070 4.35 0.84 0 9
Government satisfaction 6,043 4.27 1.17 0 9
Want less immigrants different race/ethnicity from majority 6,071 2.56 0.38 1 4
Want less immigrants same race/ethnicity from majority 6,071 2.21 0.34 1 4
Immigrants make country worse 6,071 5.23 0.91 2 9
Notes: The table shows summary statistics of the variables used in the analysis from the ESS data. The construction of the single variables is
discussed in the text.
30
B Populist parties
Table A2 lists parties that are defined as populist by the PopuList on the one hand and
by van Kessel (2015) and Inglehart & Norris (2016) on the other.
Table A2: Comparison PopuList, van Kessel and Norris & Inglehart (N&I)
CountryParty PopuList Kessel N&I
AT Freedom Party (FPA) 1 1 1
AT Alliance for the Future of Austria (BZS) 1 1 0
AT Team Stronach (TS) 1 1 0
AT Liste Dr. Martin 1 0 0
BE Flemish Interest (VB) 1 1 1
BE National Front (FN) 1 1 0
BE List Dedecker (LDD) 1 1 0
BG National Movement Simeon the Second (NDSV) 1 1 0
BG Attack Party (Ataka) 1 1 1
BG Law, Order and Justice (RZS) 1 1 0
BG Reload Bulgaria/Bulgaria Without Censorship (BBZ/BBT) 1 - -
BG Citizens for European Development of Bulgaria (GERB) 1 1 0
BG VMRO-BND Bulgarian National Movement 1 0 1
BG NFSB National Front for the Salvation of Bulgaria 1 0 1
CH Swiss People?s Party (SVP) 1 1 1
CH Swiss Democrats (SD) 0 1 0
CH League of Ticinesians (LdTi) 0 1 0
CH Geneva Citizens’ Movement (MCG) 0 1 0
CY Citizens’ Alliance (SYM/SYPOL) 1 - -
CZ ANO 2011 (ANO) 1 1 0
CZ Freedom and Direct Democracy (SPD) 1 - -
CZ Public Affairs (VV) 1 1 0
CZ Dawn of Direct Democracy (Ãssvit) 1 1 1
CZ Rally for the Republic-Republican Party of Czechoslovakia (SPR-RSC) 1 0 0
CZ Sovereignty–Jana Bobosikova Bloc 1 - -
DE Party of Democratic Socialism/ The Left (PDS/Linke) 1 1 0
DE NPD National Democratic Party 0 0 1
DE AfD Alternative for Germany 1 0 1
DK Danish People’s Party (DF) 1 1 1
DK Progress Party (FrP) 1 0 0
EE Res Publica (ERP) 1 - -
EE Conservative People’s Party (EKRE) 1 - -
ES Podemos 1 - 1
FI True Finns (PS) 1 1 1
FI Blue Reform (SIN) 1 - -
FR National Front (FN) 1 1 1
FR MPF Popular Republican Movement 0 0 1
FR La France Insoumise 1 - -
GB British National Party 0 1 1
GB UK Independence Party 1 1 0
GB NF National Front 0 0 1
GR Coalition of the Radical Left (SYRIZA) 1 1 1
GR Independent Greeks (ANEL) 1 1 1
GR Popular Orthodox Rally (LAOS) 1 1 1
GR XA Golden Dawn 0 0 1
GR ND New Democracy 0 0 1
GR Democratic Social Movement (DIKKI) 1 0 0
HR Croatian Party of Rights dr. Ante Starcevic (HSP-AS) 0 1 1
HR Croatian Labourists’ Labour Party (HL-SR) 1 1 0
HR HSS Croatian Peasants Party 0 0 1
HR HDSSB Croatian Democratic Alliance of Slavonia and Baranja 1 (until 2015) 0 1
HR HSP Croatian Party of Rights 1 0 1
HR Human Shield 1 0 0
HR Bridge of Independent Lists (MOST) 1 0 0
HR HDZ Croatian Democratic Union 0 0 1
31
CountryParty PopuList Kessel N&I
HU FIDESZ-Hungarian Civic Alliance (FIDESZ-MPSZ) 1 (since 2002) 1 1
HU Movement for a Better Hungary (Jobbik) 1 1 1
HU Hungarian Justice and Life Party (MIEP) 1 0 0
IE Sinn Fain (SF) 1 1 -
IS Centre Party (M) 1 - -
IS People’s Party (FIF) 1 - -
IS Citizens’ Movement (BF) 1 1 -
IT Forza Italia (FI) / People for Freedom (PdL) 1 1 0
IT Northern League (LN) 1 1 1
IT 5 Star Movement (M5S) 1 1 1
IT Brothers of Italy (Fdl) 1 0 1
LT Labour Party (DP) 1 (only in 2004) 1 0
LT Order and Justice Party (TT) 1 1 0
LT DK The Way of Courage 1 0 1
LT National Resurrection Party (TPP) 1 0 0
LT Lithuanian Centre Party (LCP) 1 (since 2016) 0 0
LU Alternative Democratic Reform Party (ADR) 1 1 1
LV All for Latvia (VL) 0 1 1
LV New Era Party (JL) 1 0 0
LV Zatler’s Reform Party 1 0 0
NL List Pim Fortuyn (LPF) 1 1 0
NL Liveable Netherlands (LN) 0 1 0
NL Freedom Party (PVV) 1 1 1
NL SGP Political Reformed Party 0 0 1
NL Socialist Party (SP) 1 0 0
NO Progress Party (FrP) 1 1 1
PL Self Defence (SO) 1 1 0
PL Law and Justice (PiS) 1 (since 2005) 1 1
PL SP United Poland 0 0 1
PL KNP Congress of the New Right 0 0 1
PL Kukiz’15 1 - -
PL League of Polish Families (LPR) 1 0 0
RO Greater Romania Party (PRM) 1 1 0
RO United Romania Partry (PRU) 1 - -
RO People’s Party - Dan Diaconescu (PP-DD) 1 1 1
SE Sweden Democrats (SD) 1 1 1
SI Slovenian National Party (SNS) 1 1 0
SI SDS Slovenian Democratic Party 0 0 1
SI The Left (L) 1 0 0
SI List of Marjan Sarec 1 - -
SK Movement for a Democratic Slovakia (HZDS) 0 1 0
SK Direction (Smer) 1 (until 2006) 1 0
SK Slovak National Party (SNS) 1 1 1
SK Ordinary People and Independent Personalities (OLaNO) 1 1 0
SK KDH Christian Democratic Movement 0 0 1
SK Real Slovak National Party (PSNS) 1 0 0
SK Alliance of the New Citizen 1 0 0
SK We are family (SR) 1 - -
TR MHP National Action Party - - 1
Notes: The table compares the classification of populist parties according to the PopuList with that in van Kessel as well as with that in Inglehart
and Norris. The sign "-" indicates that the country and/or time period is not covered.
32
C Excluding Partisanship
Table A3: Main Table Excluding Partisanship
(1) (2) (3) (4)

Risk aversion -0.00498 0.0118*** -0.00392 0.0110*** -0.00836 0.0135*** -0.00972 0.0146***
(0.00633) (0.00374) (0.00645) (0.00375) (0.00683) (0.00419) (0.00640) (0.00423)
ln(Age) -0.557*** 0.742*** -0.536*** 0.727*** -0.615*** 0.741*** -0.629*** 0.760***
(0.0721) (0.0265) (0.0749) (0.0263) (0.102) (0.0268) (0.0752) (0.0268)
ln(Education) -0.575*** 0.553*** -0.590*** 0.569*** -0.579*** 0.546*** -0.538*** 0.539***
(0.0396) (0.0365) (0.0409) (0.0362) (0.0374) (0.0368) (0.0377) (0.0373)
TV total 0.0223*** -0.0232*** 0.0228*** -0.0235*** 0.0222*** -0.0211*** 0.0185*** -0.0216***
(0.00586) (0.00424) (0.00586) (0.00425) (0.00482) (0.00405) (0.00514) (0.00404)
TV politics -0.0516*** 0.0670*** -0.0512*** 0.0674*** -0.0455*** 0.0575*** -0.0406*** 0.0568***
(0.00815) (0.00623) (0.00829) (0.00626) (0.00852) (0.00649) (0.00798) (0.00635)
Unemployment 0.161*** -0.151***
(0.0242) (0.0172)
(0.0133) (0.0104)
Exposure globalization 0.0902*** -0.118***
(0.0292) (0.0217)
(0.0498) (0.0401) (0.0531) (0.0418) (0.0492) (0.0426)
(0.00878) (0.00371) (0.00584) (0.00371)
(0.0220) (0.00756)

Rho -0.880 -0.876 -0.955 -0.938
Observations 172,179 172,179 160,462 156,146
Censored observations 61148 61148 56560 54282
Selection
Rain -0.00988*** -0.00987*** -0.0106*** -0.0110***

(0.00229) (0.00229) (0.00262) (0.00242)
Rain × South -0.00305 -0.00300 -0.000180 0.000542
(0.00711) (0.00719) (0.00569) (0.00584)
Av. Temperature 0.0132** 0.0133** 0.0139** 0.0152***
(0.00538) (0.00546) (0.00664) (0.00582)
Av. Temperature × South -0.0146 -0.0150 -0.0146* -0.0137
(0.0103) (0.0103) (0.00878) (0.00901)
condition on election day. All regressions include country and wave fixed effects. Robust standard errors clustered at the region level are shown in parenthesis. *** significant 1% or less; ** significant at 5%;
* significant at 10% confidence level.
33
D Sensitivity Analysis
To alleviate concerns about omitted variable problems we performed a sensitivity anal-

ysis on the OLS regression of economic insecurity on populist vote. The analysis tells us
how important any omitted variables should be to bring the economic insecurity coefficient
to zero. The graph below summarizes the main conclusion: a possible omitted variable
should explain about 6% of the variability of both economic insecurity and populist vote.
To give an idea of the magnitude, a possible omitted variable that had an impact equal to
3 times the variable on political orientation would not be sufficient to bring the economic
insecurity coefficient to 0. For details of the technique refer to Cinelli and Hazlett (2020).
Figure A1: Sensitivity Analysis
Notes: Sensitivity Analysis on the OLS regression of economic insecurity on vote

populist.
34
E Magnitudes’ computations
Table 2 presents the LHS of equation (1) in its first column, while the second term of
the RHS is reported in the second column. Letting σx denote the standard deviation of
any independent variable x, when evaluating the effects of a 1-standard-deviation change
in x equation (1) becomes:
σx πxC /π C = σx πxJ /π J − σx πxV /π V .
Therefore, we need estimates for σx , πxC , and π C to estimate the LHS, and σx , πxV , and π V
to estimate the second term of the RHS, i.e. the neglected role of turnout. Remember that
x indicates either economic insecurity (EI), trust (T), or attitudes toward immigrants (AI).
We have π C = 0.154, π V = 0.783, σEI = 0.209, σT = 2.362, and σAI = 0.901, taken
from the descriptive statistics (Table A1).
πxC and πxV represent the marginal effects of our three x variables for the second stage
(populist vote conditioning on voting, C) and, respectively, the selection equation (voting,
V), estimated from our most comprehensive specification, which includes all of them to-
C
gether (column 4 of Table 1). Obtaining them directly from Stata, these are: πEI = 0.111,
C
πTC = −0.012, πAI V
= 0.027, πEI V
= −0.214, πTV = 0.015, and πAI = −0.009.
C
The first entry of the table (0.15) is σx πEI /π C (0.209×0.111/0.154) and the second
V
entry (0.057) is -σEI πEI /π V (-0.209×(-0.214)/0.783). The ratio of the two gives us the 38%
described in the text.
The same procedure can be repeated for all the other components of the table.
35
F Robustness
Table A4 presents the estimates of the instruments relative to the robustness regressions.
Table A4: First stage Robustness
(1) (2) (3)

Vote Vote Vote
Rain -0.00164 -0.000783 -0.00678**

(0.00248) (0.00258) (0.00332)
Rain × South -0.00594 -0.0160* -0.0165**
(0.00509) (0.00824) (0.00804)
Av. Temperature 0.0166*** -0.00456 0.0193***
(0.00563) (0.00623) (0.00644)
Av. Temperature × South 0.00773 0.0216 -0.00727
(0.00904) (0.0157) (0.0118)
Wave FE Yes No Yes

Country FE Yes No Yes
Wave × Country FE No Yes No
Cluster SE Region Region Region
Countries All With P With P
(no new P)
Notes: The table shows the instruments in the voter turnout regressions in Table 3 in the text.
36
G Linear Probability Model
Table A5 presents results of three different linear probability models. In other words,
we run three OLS regressions with a binary outcomes: votes for each party yes/no, votes
for populist parties yes/no and votes for non-populist parties yes/no. The coefficients are
negatives, with magnitudes in line with what theory predicts: the economic insecurity
coefficient when we consider only the vote for populist parties is the lowest of the three
(column three), followed by the coefficient when we consider all parties (column one) and
the coefficient when we consider only non-populist parties (column 2).
Table A5: Linear Probability Models
(1) (2) (3)

Vote Vote Vote
Economic Insecurity -0.104*** -0.0367*** -0.122***

(0.01) (0.01) (0.01)
Observations 198,144 53,076 161,186

R-squared 0.088 0.224 0.120
Parties All Populists Non-Populists
Controls, Wave FE, Country FE Yes Yes Yes
Notes: Robust standard errors are shown in parenthesis. *** significant 1% or less; ** significant at 5%; * sig-
nificant at 10% confidence level.
37
H Weather Control
In table A6 we replicate the main table by adding a new control to alleviate concerns
about our identification strategy. The concern is that the weather on the day of the election
may be indicative of the weather of a longer time period in a country. Hence, the fact that
it was particularly cold on the day of the election in a given country could proxy for the fact
that, more generally, this country experienced an unusually cold period of time. Particularly
dry or cold periods, however, would affect various industries differently, and therefore the
economy as a whole, which would violate the exclusion restriction that the instrument
only affects turnout, but not voting conditional on turnout. We collected data on average
temperature by country for the years 1901 to 2021 from the official World Bank website
(Climate Change Knowledge Portal) and built a new control as the average temperature in
the country in the wave year, divided by the long-term average temperature in the country.
The coefficients of interest remain almost identical to the main Table of the paper (1) while
the new control impacts on the probability of voting (i.e., a higher temperature than the
"typical" temperature of the country increases the propensity of people to vote, as the
referee suggests and expects), but not on the probability of voting populist.
38
Table A6: Heckman probit estimates of populist party vote and participation in voting
(1) (2) (3) (4)

Risk aversion -0.00229 0.0145*** -0.00136 0.0140*** -0.00232 0.0156*** -0.00319 0.0170***
(0.00809) (0.00391) (0.00811) (0.00390) (0.00853) (0.00414) (0.00869) (0.00421)
ln(Age) -0.289*** 0.762*** -0.276*** 0.751*** -0.284*** 0.768*** -0.305*** 0.781***
(0.0764) (0.0268) (0.0735) (0.0267) (0.0725) (0.0274) (0.0748) (0.0278)
Average Temp./Typical Temp -0.0928 0.652*** -0.106 0.655*** -0.116 0.612*** -0.124 0.623***
(0.114) (0.153) (0.114) (0.154) (0.112) (0.146) (0.114) (0.145)
ln(Education) -0.427*** 0.528*** -0.442*** 0.541*** -0.428*** 0.526*** -0.376*** 0.521***
(0.0563) (0.0361) (0.0552) (0.0360) (0.0537) (0.0372) (0.0562) (0.0378)
TV total 0.0168** -0.0214*** 0.0175** -0.0216*** 0.0184** -0.0197*** 0.0135* -0.0198***
(0.00730) (0.00456) (0.00724) (0.00453) (0.00717) (0.00443) (0.00719) (0.00435)
TV politics -0.0255*** 0.0565*** -0.0254*** 0.0567*** -0.0196** 0.0495*** -0.0159* 0.0490***
(0.00857) (0.00633) (0.00850) (0.00636) (0.00851) (0.00650) (0.00856) (0.00640)
Unemployment 0.143*** -0.166***
(0.0326) (0.0185)
(0.0149) (0.0107)
Exposure globalization 0.0512 -0.118***
(0.0347) (0.0228)
(0.0657) (0.0414) (0.0636) (0.0427) (0.0664) (0.0432)
(0.00623) (0.00372) (0.00621) (0.00377)
(0.0178) (0.00799)

Rho -0.352 -0.363 -0.335 -0.329
Observations 155,506 155,506 145,877 142,849
Censored observations 49,509 49,509 46,421 45,099
Selection
Rain -0.00581** -0.00586** -0.00567* -0.00586*

(0.00294) (0.00289) (0.00320) (0.00326)
Rain $\times$ South -0.0199** -0.0196** -0.0186** -0.0182**
(0.00839) (0.00838) (0.00834) (0.00812)
Av. Temperature 0.0177*** 0.0178*** 0.0201*** 0.0202***
(0.00578) (0.00578) (0.00599) (0.00603)
Av. Temperature $\times$ South -0.00787 -0.00820 -0.00674 -0.00344
(0.0118) (0.0118) (0.0117) (0.0116)
condition on election day. All regressions include country and wave fixed effects. Robust standard errors clustered at the region level are shown in parenthesis. *** significant 1% or less; ** significant at
5%; * significant at 10% confidence level.
39
I Pseudo panel regressions
The first two columns of Table A7 report controlled fixed-effect pseudo-panel regressions
of trust in political parties and attitudes to non-EU immigrants on our summary measure
of economic insecurity and individual time-varying controls (risk aversion, age, exposure
to the media) as well as country-specific time effects common to all cohorts. Economic
insecurity has a negative and highly significant effect on trust in political parties and a
positive and highly significant effect on hostility towards immigrants. The estimates show
that a 1-standard-deviation increase in economic insecurity lowers trust in political parties
by 7.1% of its sample standard deviation and, on the other hand, increases hostility to
non-EU immigration by 5% of its sample standard deviation.
The rest of the table expands the evidence by regressing several measures of trust
(in politicians, in the national parliament, in the European parliament, and an index of
satisfaction with the government) and attitudes towards immigrants (preference for fewer
immigrants of different race/ethnicity; for fewer immigrants of same race/ethnicity; agree-
ment that immigrants make the country worse). Economic insecurity causes people to lose
confidence in politics, institutions and governments, and to increase aversion to immigrants
across the board.
40
Table A7: Trust and attitude towards immigrants - Pseudo panel
(1) (2) (3) (4) (5) (6) (7) (8) (9)

Trust parties Few immi- Trust politi- Trust na- Trust Euro- Government Few immi- Few im- Immigrants
grants from cians tional par- pean parlia- satisfaction grants from migrants make coun-
no-EU liament ment different from same try worse
race/ethnicity race/ethnicity
from major- from major-
ity ity
Risk aversion -0.0421 0.00444 0.0283 0.0623 -0.0482 0.00601 -0.00888 -0.00765 0.0214
(0.0341) (0.0127) (0.0365) (0.0446) (0.0443) (0.0257) (0.0171) (0.0158) (0.0310)
ln(Age) 0.0400 -0.0707 0.117 0.00704 -0.473 -0.774*** 0.200*** 0.301*** -0.130
(0.208) (0.0561) (0.198) (0.357) (0.331) (0.162) (0.0720) (0.0755) (0.285)
ln(Education) 0.326*** -0.182*** 0.446*** 0.643*** 0.518** 0.386* -0.208*** -0.264*** -0.778***
(0.0843) (0.0500) (0.0712) (0.0635) (0.198) (0.214) (0.0522) (0.0388) (0.205)
TV total -0.0457*** 0.0182*** -0.0504*** -0.0284 -0.0554** -0.0419*** 0.0116** 0.00822 0.0527***
(0.0108) (0.00551) (0.0119) (0.0168) (0.0230) (0.0128) (0.00462) (0.00504) (0.0154)
TV politics 0.101*** -0.0172 0.0655*** 0.0719** -0.00259 0.0152 0.00789 -0.0220 -0.0711**
(0.0226) (0.0122) (0.0229) (0.0283) (0.0409) (0.0278) (0.0136) (0.0146) (0.0313)
Economic insecurity (PC) -0.901*** 0.198** -0.950*** -1.144*** -0.472** -1.490*** 0.293*** 0.293*** 0.533***
(0.171) (0.0791) (0.162) (0.243) (0.201) (0.162) (0.0831) (0.0811) (0.161)
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Cohort FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Wave*Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Number of cohorts 840 840 840 840 840 840 840 840 840
Countries All All All All All All All All All
Observations 4,591 4,955 4,955 4,955 4,955 4,927 4,955 4,955 4,955
Notes: The table shows pseudo-panel fixed effect regressions of trust and attitudes towards immigrants on economic insecurity and controls. Left-hand side variables: several measures of trust (towards national and European)
institutions, and attitudes toward immigration (more details in the text). All regressions include country×wave fixed effects. Robust standard errors clustered at the cohort level are shown in parentheses. *** significant 1% or
less; ** significant at 5%; * significant at 10% confidence level.
41
J Weather Interaction
In this section, we run a regression mimicking the second stage of our Heckprobit, but
interacting the economic insecurity variable with a dummy summarizing weather condi-
tions. The dummy is set equal to one if precipitation on the election day is below the
median precipitation within the country and at the same time temperature on the election
day is above the median of temperature within the country. These two conditions repre-
sent favorable weather conditions that, according to our theory, increase turnout. Table A8
shows that broadly consistent with our exclusion restriction, favorable weather condition
does not matter for voting populist, but it magnifies the populist vote.
Table A8: Weather Interaction
(1) (2)
Probit LPM
Populist Vote Populist Vote
Favorable Weather 0.0533 0.00726

(0.0467) (0.00838)
Economic Insecurity 0.489*** 0.0986***
(0.0508) (0.0111)
Favorable Weather × Economic Insecurity 0.432*** 0.122***
(0.108) (0.0311)
Observations 111,318 112,037

Controls, Wave FE, Country FE Yes Yes
Notes: Robust standard errors clustered at the region level are shown in parenthesis. *** significant 1% or
less; ** significant at 5%; * significant at 10% confidence level.
42

Populism

Uploaded by

Copyright:

Available Formats

Populism

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Populism

Uploaded by

Copyright:

Available Formats

Institutional Members: CEPR, NBER and Università Bocconi

WORKING PAPER SERIES

Economic Insecurity and the Demand of

Working Paper n. 704

This Version: August, 2023

IGIER – Università Bocconi, Via Guglielmo Röntgen 1, 20136 Milano –Italy

glected turnout effect is substantial: conditional on voting, when economic insecurity

from the turnout channel.

Keywords: turnout, trust in politics, voter sentiments

Figure 1: Populism, Economics, Electoral Participation and Trust

M5S POOL RESULTS

Industrial production ( 2004=100, Eurostat) Trust in political parties (2004=100, Eurobarometer)

POPULIST PARTIES VOTE SHARE

Industrial production ( 2007=100, Eurostat) Trust in political parties (2007=100, Eurobarometer)

πxC = (πxJ π V − πxV π J )/(π V )2

or, in percentage terms,

3.1 Instrumented Heckprobit

Formally, we estimate the following selection equation:

voted irct =γ1 xirct + γ2 rain rct + γ3 rain rct × south c +

γ4 av. temperature rct + γ5 av. temperature rct × south c + (2)

and the following second stage equation:

voted populist irct = α1 xirct + α2 EIirct + fc + ft + virct (3)

Figure 2: Economic Insecure by Populist Vote

Figure 3: Populist Vote and Turnout by Terciles of Econ. Insecurity

4.2 Main analysis

Table 1 reports the estimates of several specifications, with a progressively augmented

(1) (2) (3) (4)

Controls, Wave FE, Country FE Yes Yes Yes Yes

Rain -0.00708** -0.00712** -0.00728** -0.00750**

4.3 The magnitude of the turnout effect

Table 2: Direct effects and effects via turnout

Effect of Contribution via

(1) (2) (3)

Risk aversion -0.00272 0.0177*** 0.00156 0.0170*** -0.000808 0.0160***

Controls Yes Yes Yes

5 Trust in politics and attitudes toward immigrants

5.1 Mediation analysis preliminary hints

5.2 Synthetic Panel Model

-.1 -.05 0 .05 .1 -.1 -.05 0 .05 .1

over time, including any effect of populist party rhetoric.

Western countries in the 21st century have experienced an unprecedented sequence of

Baron, R. M. and D. A. Kenny (1986). The moderator–mediator variable distinction in so-

Foster, C. and J. Frieden (2017). Crisis of trust: Socio-economic determinants of europeans

Guriev, S. and E. Papaioannou (2020). The political economy of populism. Journal of

Iacobucci, D., N. Saldanha, and X. Deng (2007). A meditation on mediation: Evidence

Jensen, M. J. and H. P. Bang (2017). Populism and connectivism: An analysis of the

Mudde, C. and C. R. Kaltwasser (2017). Populism: A very short introduction. Oxford

Rodrik, D. (2018). Populism and the economics of globalization. Journal of international

Rooduijn, M., S. Van Kessel, C. Froio, A. Pirro, S. De Lange, D. Halikiopoulou, P. Lewis,

Table A1: Descriptive statistics

Variable Obs. Mean St. Dev. Min Max

B. Pseudo panel analysis

Table A3: Main Table Excluding Partisanship

(1) (2) (3) (4)

Controls, Wave FE, Country FE Yes Yes Yes Yes

Rain -0.00988*** -0.00987*** -0.0106*** -0.0110***

To alleviate concerns about omitted variable problems we performed a sensitivity anal-

Figure A1: Sensitivity Analysis

Notes: Sensitivity Analysis on the OLS regression of economic insecurity on vote

σx πxC /π C = σx πxJ /π J − σx πxV /π V .

Rain -0.00708 -0.00712 -0.00728 -0.00750

Risk aversion -0.00272 0.0177* 0.00156 0.0170* -0.000808 0.0160***

Rain -0.00988* -0.00987* -0.0106* -0.0110*

Economic Insecurity -0.104* -0.0367* -0.122***

Rain -0.00581 -0.00586 -0.00567* -0.00586*