Socio-Economic Planning Sciences xxx (xxxx) xxx

Contents lists available at ScienceDirect

Socio-Economic Planning Sciences

journal homepage: http://www.elsevier.com/locate/seps

Invited paper

An analysis of the managerial performance of Italian museums using a

generalised conditional efficiency model
Calogero Guccio a, *, Marco Martorana a, Isidoro Mazza a, Giacomo Pignataro a, b, Ilde Rizzo a
a
Department of Economics and Business, University of Catania, Italy
b
Department of Management, Economics and Industrial Engineering, Politecnico di Milano, Italy

A R T I C L E I N F O A B S T R A C T

JEL classification: Museums are important cultural institutions having the mission of conserving, interpreting, researching and
C14 displaying heritage. The evaluation of their efficiency has been extensively investigated in the literature using
C61 different approaches and methods. In this paper, we employ a generalised conditional efficiency model to assess
I21
the true efficiency of Italian museums, i.e. the efficiency in the provision of museums’ service potential,
Z1
consistently dealing with the impact of the socio-demographic and institutional environment in which museums
Keywords:
operate. Results show that the operational environment matters. In particular, conditional estimates suggest that
Managerial efficiency
Conditional estimator
higher income levels and larger hospitality sectors positively influence museums efficiency. Ownership matters:
Robust FDH museums’ service potential is higher for private institutions while, among public institutions, efficiency is
Public goods generally lower for state-owned museums.
Service potential
Cultural heritage
Museums

1. Introduction Spanish museums considering both aspects. Our study aims at contrib­
uting to fill this gap adopting a different approach.
The analysis of the efficiency of production of museums has gained a In order to estimate the technical efficiency of Italian museums, we
rising interest over time. The growing focus is on technical efficiency focus on the production process of outputs that can be interpreted under
using the frontier techniques. A crucial step for ensuring a conceptually the direct control of the managers and take into account the potential
consistent application of this methodology and the provision of correct heterogeneity arising from the relevant context of the production of
information to decision-makers is the definition of the production services. The analysis is based on a large database of Italian museums,
function in terms of inputs and outputs that are under the control of built on by the Italian National Statistical Office (ISTAT), which run
museums’ managers. Moreover, a correct estimation of efficiency often three waves of surveys, in 2011, 2015 and 2017. Each survey contains
faces the additional problem of the potential idiosyncratic impact of over 100 questions.
environmental and contextual factors, not under the control of A line of reasoning that traces back to Bradford et al. [2] and Parks
managers. et al. [3] contributes to characterize museums’ provision as a two-stage
These issues are not specific to the field of museums since they production process where only the first stage is fully under the control of
characterize the production of several public services. There are in fact museums and consists in the use of the available space and workers to
many studies on the measurement of efficiency and performance of provide a service defined in terms of opening days and exhibitions of
different services (health care, local services, public libraries, etc.) that collections (the ‘service potential’, according to the definition of Ham­
deal with the identification of outputs, distinct from outcomes, as well as mond [4], while the second stage involves the use of the output of the
with the impact of exogenous factors on the production process. In first stage for the realization of the outcomes of the service, as directly
contrast, the literature on the efficiency of museums has substantially related to its main goals. Since the objective of our analysis is to assess
overlooked these issues, with the recent exception of Del Barrio-Tellado the efficiency of museums’ management in transforming the inputs of
and Herrero-Prieto [1] who have estimated the efficiency of a sample of the service into the outputs strictly under its control, we will focus on the

* Corresponding author.
E-mail addresses: [email protected], [email protected] (C. Guccio).

https://doi.org/10.1016/j.seps.2020.100891
Received 15 November 2019; Received in revised form 31 May 2020; Accepted 31 May 2020
Available online 17 July 2020
0038-0121/© 2020 Elsevier Ltd. All rights reserved.

Please cite this article as: Calogero Guccio, Socio-Economic Planning Sciences, https://doi.org/10.1016/j.seps.2020.100891
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

first stage. of successive studies have applied the same methodology.3

Following De Witte and Geys [5], the estimation of the technical Basso and Funari [7] make the same assumption regarding inputs but
efficiency frontier and the measurement of the efficiency assessment of consider four different outputs: visitors paying full or reduced admit­
each museum are carried out using a generalised conditional efficiency tance fee, number of exhibitions and ancillary research activities (such
model, which allows to consider inputs, outputs and environmental as seminars). In line with those studies Taheri and Ansari [8], rank the
factors simultaneously, estimating unconditional and conditional effi­ performance (in terms of visitors) of a set of 19 cultural-historical mu­
ciency in just one step. As for contextual or environmental factors, we seums in Tehran, considering the common inputs of space, facilities,
examine some variables related to the potential demand of museums’ external promotion of the museum and staff, but disentangling the latter
services and to the governance of museums, which can exert an variable according to education. Mairesse and Vanden Eeckaut [9] use a
“external” pressure on the service potential’s size, that can influence the related deterministic nonparametric approach, namely Free Disposal
performance. More precisely, we take into consideration the following Hull analysis, which differs from DEA because it relaxes the assumption
factors: the number of beds of hotels and similar structures in the of convex frontier and then generalise the comparison of a museum to
province where the museum is located; the per capita income at regional other museums. They use operational budget and different categories of
level, accounting for the potential demand in the area; the current staff as inputs in three different models having different outputs: pres­
expenditure for cultural and touristic activities at regional level; the ervation, research and communication, impact on the community edu­
public/private nature of the museum governance. cation and conservation.
The contribution of this study to the literature on the efficiency of These variables are incorporated in the complex production function,
museums is twofold. It relies on the novel adoption of a generalised with multiple inputs and outputs, which Del Barrio and Herrero [10] use
conditional efficiency model and on the size of the data set. To the best to perform an efficiency analysis, using DEA, of the regional museum
of our knowledge, it is the first time that such a large dataset is used in a system in Spain. In Del Barrio et al. [11] DEA is associated to the use of
work estimating museums’ efficiency.1 This allows to analyse thor­ multivariate statistical analysis to filter the database and generate ho­
oughly a decentralized and diffused system of museums, like the Italian mogeneous groups of museums. A similar approach is adopted by Wang
one, and to test the specific methodology of efficiency measurement et al. [12] to filter the number of indicators in the evaluation of the
applied in this work, on a much wider basis than in previous studies. In economic performance of provincial museums of 31 provinces in China.
this way we are able to offer a more extensive and insightful analysis Basso et al. [13] jointly use DEA and Balance Scorecard (BSC), a
than previous studies because we disentangle the contribution of man­ measuring tool generally used for profit-oriented organizations, which
agement on museum efficiency. they adapt to museums. In their two-stage approach, DEA provides the
Our results show that the operational environment matters and efficiency scores for the different perspectives elicited in the BSC. Unlike
significantly affects museums efficiency. Conditional estimates suggest the previous studies, Bishop and Brand [14] adopt a parametric (sto­
in fact that higher per capita income levels and larger hospitality sectors chastic) frontier approach using a single stage maximum likelihood
positively influence museums efficiency. In addition, our results procedure of estimation.
contribute to the literature on private supply of public services showing
that private ownership is associated to higher level of efficiency, and to
the literature on devolution, finding lower efficiency levels for state than 2.2. The attempt to identify production processes under the direct control
local museums. of museums’ managers
The remainder of the paper is structured as follows. Section 2 posi­
tions our approach and contribution in the context of the relevant The effort to move the analysis of efficiency of museums towards the
literature. Section 3 describes the institutional setting, while Section 4 use of more advanced approaches has not adequately investigated their
illustrates the model of production function, the data and the efficiency production process, focusing on what is strictly under control of their
estimation method. Our empirical findings are presented in Section 5. managers. This is a crucial aspect for developing a consistent oper­
Finally, the main conclusions are summarized in Section 6. ationalization of the theoretical concept of efficiency of production and
for using its measurement for purposes of evaluation and control of the
2. Literature review museums’ managers.
A first question regards the identification of what has to be consid­
2.1. The measurement of museums’ efficiency, from productivity ered as the output of the production process managed by museums’
indicators to frontiers managers. A crucial issue in this identification is the focus on those re­
sults of the transformation of the inputs employed by the museums’
The measurement of the efficiency of production of museums, as well managers that strictly depend on their efficient management of the
as of other cultural institutions, has developed from an initial stage production process. This implies avoiding the use of measures that are
characterized by the wide use of productivity indicators2 to several at­ more related to the outcomes of the service.4 The final outcomes of
tempts of employing efficiency frontiers. public provision are generally beyond the direct control of providers and
With regards to the application of the latter methodology to mu­ should not be included in the evaluation of providers’ technical effi­
seums, before Del Barrio-Tellado and Herrero-Prieto [1] previously ciency [15]. Their inclusion would in fact bias the efficiency assessment
mentioned Pignataro [6], is one of the first to apply Data Envelopment exercise, penalizing the scores of those providers that are located in
Analysis (DEA) to estimate the efficiency of museum in attracting visi­ areas with relatively low demand, independently of their effort to use
tors, considering staff and size of the exhibition area as inputs. A number resources efficiently. Moreover, it would mix-up issues of efficiency in
producing outputs and of effectiveness in achieving outcomes, since the

1 3
A recent paper by Bertacchini et al. [24] draws on the same data (for 2011), For an introduction of the economic interpretation of DEA methods see
but has a different objective, namely to test the impact of the ownership and the Førsund [34], whereas, for a thorough survey of DEA applications see, among
organizational structure of museums on their performance. the others, Emrouznejad and Yang [35].
2 4
For a thorough discussion of the main conceptual limitations of productivity The distinction between the outputs and the outcomes of a service can
indicators see Pignataro [33]. already be found in Bradford et al. [2] and to their well-known differentiation
“between the services directly produced (called “D-output”) and the primary
interests of the citizen-consumer (termed “C-output)” (Bradford et al. [2], p.
186).

2
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

production function would be based on elements under control of pro­ environmental factors for the production of museums’ outputs, one of
duction managers (the inputs) and elements (at least partially) out of the few and most recent examples of works where they are used is again
their control (the outcomes).5 [1]. They consider variables related to the socioeconomic characteristics
The mentioned studies on the assessment of the efficiency of pro­ of the environment where the museum is located, such as hotel capacity,
duction of museums do not attempt to identify a production function alternative cultural attractions, economic conditions and ease of access,
having a clear-cut notion of output, as differentiated from outcome. as well as some specific characteristics of museums like age and the type
They generally use a production function that opposes the inputs of managing institution.
directly controlled by museums to the number of visitors, which is the As for the methodologies employed to take into account the impact of
only output for some studies, or one of the multiple outputs for others. A variables representing contextual and environmental factors on effi­
noticeable recent exception regarding the efficiency of museums is ciency, the general literature on the estimation of efficiency has devel­
represented by Del Barrio-Tellado and Herrero-Prieto [1] who distin­ oped several approaches. A largely used technique, also in the field of
guish between programmed outputs - in direct control of museums’ museums, is the one proposed by Simar and Wilson [16]; in which, at a
managers, such as the number of temporary exhibitions, the number of first stage, efficiency scores are computed through nonparametric
publications and the number of dissemination actions (e.g. educational techniques and, at a second stage, those scores are then regressed on
workshops, concerts, seminars, etc.) - and the only observed output, that exogenous variables.7 The main limitation of this two-stage approach is,
is the number of visitors, which they regard as “the most basic expres­ however, its reliance on the so-called separability condition: the envi­
sion of demand … that depends on the appeal of supply as well as on ronmental and contextual factors do not influence the shape of the
visitor preferences and external factors that shape said demand” (p. production set but the position of each production unit within the set.
494). The shortcomings implied by this condition are overcome by the so
Our approach, however, is different since we are interested, as stated called conditional approach [17,18], which conditions the estimation of
above, into the assessment of the efficiency of museums’ managers in efficiency of a productive unit to a given value of environmental and
transforming inputs into outputs. While Del Barrio-Tellado and Herrero- contextual factors, in such a way that the input requirements of its
Prieto [1] evaluate the overall performance of museums, we will focus production process change according to the favourable or unfavourable
only on the portion of performance for which their managers are strictly external conditions.8 We will provide the technical details of this
and directly accountable.6 Our work, therefore, is more in line with De approach in section 4.3, since it is at the basis of the model for our
Witte and Geys [5]. In their work on libraries, they choose to focus on empirical analysis.
the assessment of technical efficiency and deliberately exclude any
analysis of how outputs are transformed into outcomes, since the latter 3. Institutional setting
may also respond to a set of external constraints different from the ones
relevant for the first stage. Their effort, essentially, is one of selecting the Italy has a wide and heterogeneous set of museums, which differ as
appropriate outputs for the technical efficiency assessment as well as the far as institutional features, type of collection, geographical location,
proper contextual and environmental factors that can constrain the and number of visitors are concerned.
efficient transformation of inputs into the service potential, in the first ISTAT [19] runs a survey on Italian Museums and Cultural In­
stage. This approach is surely less ambitious than the one that aims at stitutions (Indagine sui Musei e le Istituzioni Similari) and identifies 4976
providing an overall evaluation of the activities of museums, since it museums and similar institutions. Among them, 4158 are museums,
overlooks how the service potential transforms into observable outputs galleries or collections. Most of them (64.1%) are public: 43% belongs to
of museums, such as visitors, which are instrumental to reach the goals Municipalities while those belonging to the state are 439, that is only
of these institutions [15]. It, however, undoubtedly allows an appro­ 8.8% of the total.
priately focused evaluation of the performance of museums’ managers. Italian museums offer different types of collections ranging from
ethnographic and anthropologic collections (16.6%) to ancient art
(15.9%) and archeology (14.7%), just to recall the most represented
2.3. The consideration of contextual and environmental factors: the
ones.
different methodologies
Overall, museum supply is widespread all over the country with one
Municipality out of three having at least a museum.9 At the same time,
Our choice does not imply that we overlook how factors related to
however, about half (48.9%) of the total is located in Northern
demand and, therefore, to the number of visitors, may affect the efficient
regions.10
management of the production process. A second question, thus, is
Visitors show an increasing trend (though with a small decrease in
related to the characterization of the production process of museums - in
the most recent years)11 as well as a marked polarization: the top 20
terms of what is strictly under control of their managers – and regards
the consideration of (contextual and environmental) factors that are
exogenous to the management of the production process of museums.
7
Overlooking the potential heterogeneity arising from these factors, for Updated reviews of the different approaches to the problem of introducing
the different museums, can severely bias the assessment of the efficiency the influence of exogenous variable into the analysis of efficiency are provided,
of each institution. The inclusion of contextual and environmental fac­ among the others, in Cordero et al. [39] and Narb� on-Perpin�
a and De Witte [40].
8
It is worth noting that there are other approaches to deal with environ­
tors in the efficiency analysis requires a meaningful selection of the
mental and contextual factors. Following Cordero et al. [39], the one-stage
factors and a technically appropriate methodology to ascertain their
approach developed by Banker and Morey [41] include these factors as
influence on efficiency. non-discretionary inputs, i.e. inputs that are not under the managers’ control, in
As for the selection of potentially relevant contextual and the form of an additional restriction in the standard DEA program.
Narb� on-Perpin�a and De Witte [40] identify a meta-frontier approach: “It eval­
uates separate efficiency performance for different groups according to the
5
[15] suggest that realizing the existence of such a two-stage production environmental characteristics” (p. 1109).
9
process is consistent with the distinction between goods and commodities, Even Municipalities with less than 2000 inhabitants (27.9% of total Mu­
where the first ones are intermediate outcomes supplied by the firms, which nicipalities) have a high percentage of museums (16.7%).
10
consumers transform into commodities satisfying their wishes [36,37]. Archaeological sites show an opposite geographical pattern: more than half
6
For a review of models based on the evaluation of the global performance in (52.8%) are located in the South and the Islands.
11
the production of public services see Førsund [38], who also discusses the Museums visitors were 53.9 million in 2011 and increased up to 59.2
different DEA models employed to deal with it. million in 2015.

3
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

museums and similar institutions attract 31.9% of visitors while 36.5% ministerial administration. This element hinders the effective autonomy
register no more than one thousand admissions per year. A major role is of museums.
played by state museums and similar institutions, which attract more It is an open question whether the enlarged autonomy of state mu­
than 47 million visitors that is the 42.6% of the total. seums, though with the above limitations, the decentralization and the
Museums, alongside with monuments and archaeological sites, have various forms of public-private partnerships have really affected mu­
always been at the ‘core’ of Italian cultural policy, with responsibilities seums performance. Our empirical analysis attempts to shed a light on
shared between the state - through the Ministry for Heritage and Cul­ the relationship between efficiency and some institutional features of
tural Activities and Tourism (Ministero per i beni e le attivita� culturali e il Italian museums.
turismo - MIBACT) - and the decentralized levels.12 Until the mid-1990s
museums were directly managed by national or local governments 4. Model and data
without any financial or organizational autonomy and, since they had no
budget or spending powers, even the revenues from ticket sales and 4.1. The identification of the production function: inputs, outputs and
sponsorships accrued to the general budget of the responsible level of environmental factors
government.
In the last twenty years, major reforms have been undertaken, According to the discussion in section 2.2, our model focuses on the
mainly oriented toward the enlargement of the autonomy of important production process of the outputs under the direct control of museums’
institutions, the decentralization of competences and functions to the managers, that is the service potential. First of all, therefore, we need to
Regional level of government and the introduction of private actors in identify the inputs and the outputs of the production function and, in
the ownership and in the management of institutions.13 Broadly doing so, we have to refer to the different functions of the museums. In
speaking, similar trends characterize cultural policies in many European general, museum activities are not limited to the preservation and ex­
countries [20]. In Italy, however, reforms appear subject to variability hibitions of items and the provision of ancillary services as they may
through time as cultural policy priorities of the different governments in include, for instance, dissemination, education and even research in
power in the last twenty years have often changed [21]. some cases. However, the latter activities are undertaken by a limited
A detailed analysis of the institutional setting is beyond the scope of share of museums while exhibition is common to all of them. This is the
this paper.14 Here it is sufficient to recall just some significant examples reason why in what follows we focus specifically on exhibition, in the
of the changes occurred. In 1998, four National Museum Poles (Poli definition of the service potential provided by museums. We consider a
Museali Nazionali) including the national art galleries and museums in production process involving the use of inputs such as the number of
Rome, Venice, Florence and Naples were granted an autonomous status workers, and exhibition area (in square meters), which are commonly
and a budget.15 In 2014, a national museum system was created. On the used in the literature (see Basso et al. [13]). Regarding the outputs, our
one hand, twenty top Italian museums, monuments and archaeological selection is restricted to the ones that are under the direct control of
sites, followed by further ten in 2016, became autonomous and fully museums, thus representing the service potential. Precisely, we consider
responsible for the management and enhancement of their own collec­ a set of two outputs: the number of opening days in the year (in range)
tions.16 On the other hand, all the other less important national non- and the number of temporary exhibitions. Differently from several other
autonomous museums and heritage sites have been gathered in seven­ works, we do not use the number of visitors as a further measure of the
teen regional museum poles (Poli museali regionali), under the re­ museums’ outputs. The reason is that while there are actions that the
sponsibility of Regional secretariats.17 museums’ managers can take to affect the number of visitors, this also
Notwithstanding some important aspects of the reform of state mu­ depends on many other factors conditioning their demand that are not
seums, such as, for instance, the increasing degree of their financial under the control of managers.
autonomy and of their capability of raising funds [22,23], it is widely We choose to consider the contextual and environmental factors
agreed that the reform process is still largely incomplete and has severe affecting the efficiency of the production process of museums according
shortcomings. Overall, internal resistance, at managerial as well as at to the conditional efficiency approach discussed in Section 2.3. In
employee level, have contrasted the full implementation of new orga­ studying the influence of the contextual factors on the service potential
nizational models [24]. A major shortcoming of the reform process stays of museums, we focus on those factors potentially affecting the demand
in the dominance of an approach too much focused on the legal aspects, of museums’ services, as well as on the institutional features that can
and in the lack of a management perspective, as it is demonstrated by the affect the use of available inputs. As for the first class of factors we
scarce consideration given to a crucial input of the production process, consider: the number of beds available in hotels and similar structures in
such as human resources [25]. Museum managers are not responsible for the province where the museum is located; the per capita income in the
the organization of labour, which actually is under the control of the region. To account for the institutional context, we consider the current
public expenditure per capita for cultural and touristic activities at
regional level, the public/private nature of the museum, and - for public
12
The protection and enhancement of cultural heritage is regulated by the
institutions - whether they are run by the state or local governments. We
Heritage and Landscape Code (Codice dei Beni Culturali e del Paesaggio). hereafter motivate the inclusion of such environmental factors, high­
13
Ponzini [42] provides an extensive analysis of the various forms of privat­ lighting the expected effect on museums’ efficiency.
ization in the period 1996-2006: alienation of state-owned historic real estate, Regarding the contextual type of factors, it is generally acknowl­
mixed public-private entities to manage and to promote cultural heritage, edged that socio-demographic characteristics may affect the demand
introduction of private actors into policy-making and implementation. and the willingness to pay for cultural goods eventually affecting the
14
A detailed description can be found on the web site of MIBACT http:// provision of museums. We should expect a significant effect of income
musei.beniculturali.it/en/structure and in Bodo and Bodo [21]. level in the area where the museum is located [5]. In fact, it can be
15
Pompei archaeological site is the first example of the enlargement of au­ presumed that the demand for cultural goods rises with income,
tonomy to a national institution.
16 potentially generating a group of high-income citizens having a common
Directors were selected through international open competitions, giving rise
interest in exerting pressure on museums to maximize their service po­
to a heated debate and judicial controversies.
17
An important implication of this reform is the separation between the tential, leading to a positive impact of income level on museum
protection and the management and enhancement of cultural heritage. The efficiency.
former is assigned to the peripheral entities of the Minister, i.e. the Superin­ The same logic can be applied to the accommodation capacity in
tendencies, while the latter belongs to autonomous national museums and hotels and similar structures, as a larger and stronger hospitality sector
Regional poles. would exert pressure to cultural providers. Consistently with this

4
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Table 1 opposite perspective, public transfers to cultural institutions, resulting

Variables - description. in additional inputs for providers, may weaken the internal incentive to
Variable name Description improve efficiency [5].
Finally, public/private ownership and state/local governance may
Inputs
Personnel No. of workers determine different internal incentives to use resources efficiently.
Exhibition_space Available space for exhibitions in square meters Regarding ownership and governance, private institutions may have
stronger internal incentives towards an efficient use of the available
Outputs resources. Finally, among the public institutions, we may expect mu­
Opening_days Opening days in the year, in range, taking values: opened
seums run by local jurisdictions to respond better to local demand side
days �31; 31<days�50; 50<days�100; 100<days�150;
150<days�200; 200<days�250; 250<days�300
pressures, given the higher proximity of the management.
Temporary_exhibitions No. of temporary exhibitions
4.2. The data
Socio-demographic and institutional factors
Public Dichotomous variable taking values: 1 if public owned, 0 if
All the data we use in the following analysis are drawn from ISTAT
private
No_state Dichotomous variable taking values: 1 if owned by local databases and refers to the year 2015. The survey on Italian museums
bodies, 0 if state owned and cultural institutions provides information on the relevant inputs and
Number_beds No. of beds in hotels and similar facilities in the province outputs of museums. Data on accommodation capacity in the hospitality
GDP Per capita GDP at regional level, in Euro at current prices sector are extracted from the annual survey “Capacity of collective ac­
Expenditure_culture Per capita public current expenditure in cultural and
commodation establishments”.
touristic activities at regional level, in Euro at current
prices The ISTAT data warehouse is also the source for data on socio-
demographic variables at provincial and regional levels.
Source: our elaboration on data provided by ISTAT
Table 1 summarizes the full set of variables used in the following
empirical analysis.
argument, Del Barrio-Tellado and Herrero-Prieto [1] find a negative After data cleaning, the final sample used in the following analysis
effect of the accommodation capacity and no significant effect of the includes 2496 museums.18
GDP level on the efficiency of museums. The sample reflects some relevant features in the population,

Table 2
Variables - descriptive statistics.
Variables Obs. Mean St. Dev Min Max
Inputs
Personnel 2496 11.58 21.56 1.00 411.00
Exhibition_space 2496 1088.84 3130.56 100.00 65,000.00

Outputs
Opening_days 2496 223.81 92.12 30.00 300.00
Temporary_exhibitions 2496 2.03 7.98 0.00 50.00

Socio-demographic and institutional factors

Public 2496 0.65 0.48 0.00 1.00
No_state 2496 0.93 0.25 0.00 1.00
Number_beds 2496 66,620.44 70,451.84 2324.00 366,341.00
GDP 2496 28,199.43 6356.25 16,301.22 37,949.43
Expenditure_culture 2496 93.82 32.17 11.70 292.61

Source: our elaboration on data provided by ISTAT

Table 3
Means of inputs and outputs - by geographical area, ownership and governance.
Observations Personnel Exhibition space Opening days Temporary exhibitions

Full sample 2496 11.58 1088.84 223.81 2.03

By geographical area
North-West 526 14.71 1284.33 204.60 2.14
North-East 657 10.97 1085.32 211.98 1.59
Centre 721 10.38 1158.70 231.55 1.82
South 363 10.47 854.89 237.49 2.96
Islands 229 11.64 800.84 255.81 2.26
By institutional factors
Private 866 9.99 985.10 208.71 2.07
Public 1630 12.42 1143.96 231.83 2.01
State-owned 171 32.62 2852.81 292.28 1.78
Non-state-owned 1459 10.05 943.68 224.75 2.04

Source: our elaboration on data provided by ISTAT

Finally, regarding the institutional factors, we expect that public

expenditures in culture and tourism at regional level may result in a
18
more favourable environment for cultural institutions such as museums, Observations are reduced because of lack of data for some variables, and to
potentially affecting the demand, in line with the above points. On an eliminate outliers, such as, for instance, museums included in parks because of
their extension.

5
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

including the share of public owned museums (65%) and state-owned of accommodating for the potential impact of environmental factors on
museums (7%), as well as the unequal geographical distribution of the efficient behaviour of museums.
museums among areas, with half of the total of museums (47.3%) To overcome the potential drawback arising from outliers, Cazals
located in Northern regions. By and large, museums have comparable et al. [17] propose to estimate a partial frontier, instead of the full
size with the exception of state-owned museums that are remarkably frontier, considering the expected value of the maximum output effi­
larger. ciency score of the single DMU compared to m < n randomly drawn
Table 2 summarizes the sample statistics of the variables employed in DMUs, which use a level of inputs lower than x. This is the order-m ef­
our model and Table 3 reports the mean values by geographical area and ficiency measure λm ðx; yÞ. Cazals et al. [17] show that λm ðx; yÞ depends
ownership. on SY ðyjxÞ, such that
Z ∞
λm ðx; yÞ ¼ ½1 ð1 SY ðuyjxÞÞm �du (5)
4.3. The model for estimating efficiency 0

Given the inputs, outputs and environmental variables identified in The efficiency score b λ m;n ðx; yÞ can be then estimated by substituting
section 4.1 and measured by the data collected in the datasets indicated the empirical b
S Y;n ðyjxÞ in equation (5), as follows:
in section 4.2, we use them to estimate technical efficiency of museums Z ∞
according to the conditional approach, already mentioned in section 2.3. b
λ m;n ðx; yÞ ¼ ½1 ð1 S Y ðuyjxÞÞm �du
b (6)
Let us consider a ‘production’ process carried out by decision-making
0

units (DMUs, museums in our case), which use x ðx 2 Rpþ Þinputs to A DMU that performs better than the average m DMUs in its reference
produce y ðy 2 Rqþ Þ outputs. The production technology is the set of all subsample may obtain a b λ m;n < 1, thus being super-efficient [26,27]. As
feasible combinations of inputs and outputs, Ψ ¼ fðx; an efficient observation is located on the best practice frontier, it obtains
�
yÞ 2 Rpþq
þ
� x can produce yg. Being Ψ unknown, it has to be estimated an efficiency score b
λ m;n equal to 1. An inefficient observation obtains an
�
from a sample of DMUs denoted by χ ¼ fðxi ; yi Þ� i ¼ 1; …; ng. In what
n efficiency score b
λ m;n higher than 1. The inefficiency (1 - b
λ m ) indicates the
follows, we use an output-oriented model assuming that museums potential percentage increase in output if the observation would pro­
maximize their outputs for given inputs. duce as efficiently as its references m. Following Daraio and Simar [27],
Following the probabilistic representation of the production function De Witte and Geys [5,15] the value of m should be chosen as the value
proposed by Cazals et al. [17], we define the probability that a combi­ for which the number of super-efficient DMUs (i.e. b λ m;n < 1Þ declines
nation (x, y) is dominated as the joint probability function19: only marginally with m.22 The order-m estimator requires the same as­
HXY ðx; yÞ ¼ ProbðX � x; Y � yÞ (1) sumptions of its full frontier counterpart (the FDH, namely) and
asymptotically converges to it, but for finite m it does not envelope all
where X and Y are the sets of inputs and outputs, respectively. Equation the observed data and is thus less sensitive to extreme points and outliers
(1) can be decomposed as: [27].
To account for the impact of the environmental or contextual factors
HXY ðx; yÞ ¼ Prð Y � yjX � xÞPrðX � xÞ ¼ SYjX ð Y � yjX � xÞFX ðX � xÞ on museums’ efficiency, we need to move to an estimation of efficiency
(2)
¼ SY ðyjxÞFX ðxÞ conditioned to their values, as opposed to the “unconditional” estima­
tion in (6). As it will be shown, we will also make use of the information
where SY ðyjxÞ is the conditional survivor function of Y and FX ðxÞ is the in (6).
cumulative distribution of X. Under free disposability of inputs and Let Z be the vector of environmental variables. The conditionrefal
outputs, the upper boundary of SY ðyjxÞ is the Debreu-Farrell output- approach, proposed by Daraio and Simar [18,26], consists in condi­
oriented efficiency measure20 tioning the production process to a given value Z ¼ z, ðz 2 Rkþ Þ. The joint
λðx; yÞ ¼ supfλjSY ðλyjxÞ > 0g ¼ supfλjHXY ðx; λyÞ > 0g (3) probability function in (2) becomes:

The estimation of (3) requires the substitution of HXY and SY with HXYjZ ðx; yjzÞ ¼ ProbðX � x; Y � yjZ ¼ zÞ (7)
their empirical versions H b XY ðx; yÞ and b
S Y ðyjxÞ. The estimator of the ef­ In (7) the function HXYjZ ðx; yjzÞ is simply the probability, for a DMU
ficiency is then: operating at level ðx; yÞ; to be dominated by DMUs facing the same
� � � environmental conditions Z, i.e. there exist DMUs that produce more
λ n ðx; yÞ ¼ sup λ�b
b S Y;n ðλyjxÞ > 0 (4)
outputs using less inputs with comparable levels of environmental var­
iables Z ¼ z. Equation (7) can be decomposed as follows:
which, coincides with the Free Disposal Hull (FDH) estimator as shown
in Cazals et al. [17] and Daraio and Simar [18].21 The efficiency esti­ HXYjZ ðx; yjzÞ ¼ Prð Y � yjX � x; Z ¼ zÞPrðX � xjZ ¼ zÞ (8)
mation model we use, however, departs from this general formulation
since it takes into account the potential sensitivity of the efficiency Analogously to the unconditional case, the partial order-m frontier
estimator in (4) to the presence of outliers and, moreover, our objective can be used to evaluate DMUs’ efficiency in the following way:
Z ∞
λm ðx; yjzÞ ¼ ½1 ð1 SY ðuyjx; zÞÞm �du (9)
0
19
See Daraio and Simar [18,26] for a detailed formal description of the model.
20
The Debreu-Farrell output-oriented (sometimes referred to as the Farrell The conditional efficiency score b λ m;n ðx; yjzÞ can be calculated by
output-oriented) efficiency measure is the extension to the output-oriented case substituting the empirical S Y;n ðuyjx; zÞ in equation (9) as follows:
b
of the Farrell [43] efficiency measure. It is defined as the increase of output the Z ∞
DMU should achieve to be considered as being output efficient [18,27,44]. b
λ m;n ðx; yjzÞ ¼ S Y;n ðuyjx; zÞÞm �du
½1 ð1 b (10)
21
Such estimator assumes free disposability of inputs, meaning that if ðx; yÞ 2 0
Ψ, then ðx’ ; y’ 2 ΨÞ, for x’ � x and y’ � y. The alternative non-parametric esti­
mator, the DEA assumes the convexity of the production set. Although DEA is
commonly used, its consistency strictly depends on whether such assumption
22
holds. Conversely, even if the true production set is convex, the more general As shown in Cazals et al. [17]; the number of superefficient DMUs decreases
FDH estimator is still consistent [27], and thus preferable in the absence of as m increases, but at a decreasing pace, and it tends to become stable beyond a
strong theoretical justifications in support of convexity. certain value of m.

6
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Table 4 Kortelainen [28]; this procedure can remove irrelevant covariates by

Unconditional efficiency estimates - by region, geographical area, ownership over smoothing them and it allows us to obtain statistical inference (in
and public governance. particular, p-values) on the role of the environmental variables.23
Sample

Mean St. Dev Min Max 5. Empirical findings

Full sample 1.068 0.797 0.350 5.673
By region 5.1. The results of the unconditional efficiency estimates
Abruzzo 1.055 0.408 0.350 2.711
Basilicata 1.052 0.343 0.350 1.504 We start by presenting and discussing the estimates from the un­
Calabria 1.156 0.409 0.350 3.048 conditional order-m, which do not account for the heterogeneity arising
Campania 1.494 3.542 0.350 4.768
from the operational environment in which museums operate. Summary
Emilia Romagna 1.016 0.414 0.350 4.984
Friuli-Venezia Giulia 0.948 0.368 0.350 2.460 statistics of the unconditional estimates are shown in Table 4, with the m
Lazio 1.119 0.320 0.350 2.910 parameter set at m ¼ 300. For statistical inference, we use 500 bootstrap
Liguria 0.982 0.328 0.350 2.455 replications.24
Lombardia 1.054 0.410 0.350 3.114
Recalling that a score equal to one corresponds to full efficiency and
Marche 1.012 0.488 0.350 5.673
Molise 1.014 0.322 0.350 1.275 that higher (lower) scores imply lower efficiency (super-efficiency), the
Piemonte 0.921 0.389 0.350 2.392 national average score of 1.068 indicates that Italian museums could
Puglia 1.160 0.521 0.350 5.111 improve their efficiency, on average, of about 6.8%. However,
Sardegna 1.124 0.305 0.350 1.995 remarkable differences across regions, areas, and ownership also emerge
Sicilia 1.234 0.900 0.350 4.603
and, additionally, we observe a remarkable variation around the mean
Toscana 1.078 0.319 0.350 2.388
Trentino Alto-Adige 0.985 0.297 0.350 1.757 (standard deviation ¼ 0.797). Regional average efficiency levels range
Umbria 1.090 0.370 0.350 3.126 from 0.921 (Piemonte), to 1.494 (Campania). In four regions, all located
Valle D’Aosta 1.011 0.310 0.350 1.324 in northern Italy, the average value is lower than 1 (Friuli-Venezia
Veneto 1.004 0.331 0.350 1.986
Giulia, Liguria, Piemonte, Trentino-Alto Adige), i.e. museums in such
By geographical area
North-West 0.996 0.390 0.350 3.114
regions are on average superefficient, while the lower levels of efficiency
North-East 0.997 0.366 0.350 4.984 (scores >1.15) are displayed by southern regions (Campania, Sicilia,
Centre 1.075 0.374 0.350 5.673 Puglia, Calabria). More in general, museums in northern Italy show
South 1.211 1.814 0.350 5.111 higher efficiency levels, on average (about 0.996), while museums in the
Islands 1.183 0.690 0.350 4.603
South present an average value of 1.211.
By ownership and public governance
Private 1.040 1.222 0.350 5.673 Private museums (1.040) perform generally better than public ones
Public 1.082 0.422 0.350 4.984 (1.082). Among the latter, lower levels of efficiency are displayed by
state-owned 1.243 0.170 0.350 2.711 State museums.
Non-state-owned 1.063 0.439 0.350 4.984

Source: our elaboration on data provided by ISTAT. Note: m ¼ 300; no. of

bootstrap replications ¼ 500 5.2. The results of the conditional efficiency estimates and the evaluation
of the impact of the environmental factors

First of all, a comparison of Figs. 1 and 2 suggests the existence of a

connection between the regional distribution of the unconditional scores
and the regional distribution of some of the environmental variables
Smoothing techniques are required to estimate SY;n ðyjx; zÞ because of
defined in Table 1, especially regarding per capita GDP and the number
the equality constraint Z ¼ z.
of beds available in hotels and similar structures.
To evaluate the influence of z on the production process, Daraio and
In Fig. 3, to determine the densities of the museums’ efficiency es­
λ m;n ðx; yjzÞ in (10)
Simar [18] suggest to compare the conditional scores b timates, we report the univariate kernel smoothing distribution using
with the unconditional ones b
λ m;n ðx; yÞ in (6). Therefore, Daraio and Simar the reflection method.25 In particular, in the left hand side of Fig. 3 we
[18] employ a smoothed non-parametric kernel regression of the ratio of report the univariate kernel smoothing distribution for the full sample
conditional scores to unconditional scores Qz ¼ b λ m;n ðx; yjzÞ=b
λ m;n ðx; yÞ on estimates and the conditional univariate kernel smoothing distribution
m
z. This method allows to detect the positive, negative or neutral influ­ for private (solid line) and public (dash line) museums, in the right hand
ence of z, according to the slope of the non-parametric kernel regression. side. The latter allows to shed a light on the differences among public
The main drawback of this approach is that z should be a continuous and private institutions. Incidentally, it can be noted that the employed
variable. To overcome this problem, we follow the approach proposed methodology works well in controlling for the presence of outliers, as
by De Witte and Kortelainen [28] and extend the conditional efficiency shown by the relatively low number of observations on the tails.
in (10) so as to include discrete variables. Furthermore, while the As previously discussed, therefore, the socio-demographic and the
original contribution of Daraio and Simar [18] allows only for a institutional context may exert an “external” pressure on the service
graphical inference about the impact of the variables z (i.e., favourable potential, implying that the unconditional scores may be biased,
or unfavourable), this approach leads to estimating a non-parametric
p-value that allows for carrying out statistical inference on the role of
23
environmental factors in DMUs’ efficiency. Actually, it can be inter­ We would like to thank Kristof De Witte for providing us the R codes routine
preted as the non-parametric equivalent of the statistic t used in OLS. to perform the estimates.
24
Our choice of the value of B is essentially due to computational needs. In
More precisely, we estimate a mixed kernel, which treats continuous,
fact, choosing B ¼ 500 and given our sample size, the employed R code takes
discrete ordered and discrete unordered variables differently, using
about 6 h for each estimate with a value of m ¼ 300. Higher values of B increase
different bandwidths for each type of variable. The bandwidth vector is calculation times exponentially. Thus, we believe that our choice represents a
then estimated by the least squares cross-validation (LSCV) method, as good balance between computational needs and accuracy in the estimates. In
suggested by Li and Racine [29]; and the significance of the effect of the the same line of reasoning, see Asatryan and De Witte [45].
conditioning variable is evaluated by applying the nonparametric 25
In doing so, we avoid the problems of bias and inconsistency at the
bootstrap test proposed by Racine [30]. As argued in De Witte and boundary of support [32].

7
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Fig. 1. Geographical distribution of unconditional efficiency estimates, average value by region.

Source: our elaboration on data provided by ISTAT

favouring those museums that operate in more favourable contexts. To show with respect to museum effectiveness, differences arise across
deal with it, we estimated the efficiency of museums using the gener­ public museums depending on their degree of autonomy.
alised conditional efficiency model represented in section 4.3. The re­ In fact, among public museums, state-owned ones display on average
sults of this estimation are reported in Table 5. lower levels of efficiency and the impact of the variable No_state (¼ 1, if
The three columns differ in the set of the conditioning variables. the museum is local) is favourable and significant. A possible explana­
More specifically, column (1) considers the public/private ownership tion relies on the fact that, as we said before, state managers lack real
only; column (2) is run on the subsample of public museums and is autonomy in ‘core’ aspects such as human resources organization. Such
conditioned to the state- or local-governance; column (3) includes an aspect is enhanced by the fact that they are substantially larger than
ownership and the socio-demographic factors. the rest in terms of personnel, generating further rigidities on the
Compared with the unconditional scores, the conditional estimates management. Another element of rigidity refers to the space, since they
shown in the second part of Table 5 present a lower mean value (ranging are mainly located in huge historical buildings. A further intuition could
from 1.057 to 1.036), implying a lower level of inefficiency on average, be that public local museums are more reactive to local interests to
and a lower standard deviation. exploit the service potential.
As for the specific institutional factors, the analysis of the p-values, in Among the socio-demographic factors, hospitality sector size and
the first part of Table 5, reveals the significance and the nature of the income level are favourable and significant, in line with the idea that a
effect of the different environmental variables on efficiency. Public stronger demand for cultural goods induces providers towards a more
ownership is generally unfavourable, implying that private institutions efficient use of available resources. This result is also in line with the
are relatively more efficient, even if its significance is weak. Such a positive influence of income level on library service potential reported
result seems in line with the common wisdom that private ownership by De Witte and Geys [5]. Finally, the amount of public expenditures on
fosters incentives eliciting efficiency. However, as Bertacchini et al. [24] cultural and touristic activities is not significant.

8
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Fig. 2. Geographical distribution of socio-demographic factors, by region.

Source: our elaboration on data provided by ISTAT

Fig. 3. Kernel density estimates for full sample and for subsamples of public and private museums.
Source: our elaboration on data provided by ISTAT

By and large, conditional estimates show that the operational envi­ entire collections. In fact, very often only a small share of the collection
ronment matters and that estimating the efficiency frontier without its is on display. In the literature this phenomenon is known as ‘Prado
consideration would lead to incorrect conclusions on the efficiency of disease’ and is usually considered in discussing the behaviour of mu­
museums. seums directors [31] in terms of resources allocation. Collecting arte­
facts if they are not displayed raises, among the other things, a cost
5.3. Robustness checks opportunity question which appears more severe for museums which
heavily rely on public funds. Thus, from the point of view of the service
In this Section we provide several robustness checks of our results. A potential of museums, the extent of exhibited collections is a measure of
potential weakness in the way we assess efficiency is related to the wide the ability of the management to efficiently use the available space and,
heterogeneity of museums in our sample, above all in terms of the for these reasons, they should be included among the outputs.27 More­
different types of their collections. Although the order-m estimator is over, since the number of pieces in the exhibited collections is probably
robust to the existence of outliers, one could argue that a qualitative correlated with the different types of museum (i.e. the different typol­
heterogeneity may not be adequately taken into account. ogies of their collections), its use in the model for estimating efficiency
For this reason, we first estimate an alternative model, adding as a can provide a test of the robustness of our results. If the use of this
further output the extent of exhibited collections (i.e. the number of additional output leads to efficiency estimates that are consistent with
pieces actually exhibited).26 The choice of this additional output is based the ones previously obtained, we can be reassured about the capability
on the following considerations. Museums do not generally exhibit their of our model to adequately control for the qualitative heterogeneity of

26 27
The sample statistics for this additional variable are: mean 1202.26; stan­ Consistently with this logic [5,15], include the number of items among
dard deviation 2920.83; min 100.00; max 50,000.00. outputs to assess the efficiency of libraries in maximizing their service potential.

9
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Table 5
Conditional efficiency estimates.
Institutional and Environmental Variables Baseline efficiency model

Unconditional Conditional

(1) (2)a (3)

Influence P-Value Influence P-Value Influence P-Value

Public – Unfavourable * Unfavourable Not significant

No_state – – – Favourable *** – –
Number_beds – – – – – Favourable **
GDP – – – – – Favourable ***
Expenditure_culture – – – – – Favourable Not significant
Mean efficiency score 1.068 1.052 1.057 1.036
St. dev. efficiency score 0.797 0.681 0.701 0.625
Observations 2496 2496 1630 2496
R-square – 0.0361 0.0163 0.1017

Note: Table presents estimates of the conditional FDH model with m ¼ 300 and no. of bootstrap replications ¼ 500; (2) favourable indicates an efficiency-enhancing
association; (3) asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels respectively; a: estimated on the relative subsample of public museums.
Source: our elaboration on data provided by ISTAT.

Table 6 distributions in Fig. 5 show a good result in terms of the consistency of

Unconditional estimates for the efficiency model in 5.3 (robustness check) - by the efficiency estimates, since there does not appear to exist any
region, geographical area, ownership and public governance. remarkable difference in the assessment of managerial efficiency ob­
Sample tained from the two models.
Mean St. Dev Min Max
Furthermore, in order to test the influence of the environmental
factors in this alternative model, in Table 7 we report the results of the
Full sample 1.137 0.343 0.450 3.984
generalised conditional efficiency model. Once again, the estimates in
Abruzzo 1.085 0.329 0.450 2.017 Table 7 seem to confirm previous results.
Basilicata 1.173 0.311 0.517 1.653 We made a further check, by an explicit consideration of the different
Calabria 1.178 0.294 0.517 1.487 categories of museums. Following a classification in the ISTAT survey,
Campania 1.290 0.687 0.450 3.984 we identified three different categories. A first group includes the arts
Emilia Romagna 1.111 0.343 0.450 2.561
Friuli-Venezia Giulia 1.055 0.313 0.450 1.597
museums, a second one the natural and scientific museums while the last
Lazio 1.197 0.293 0.450 2.842 one is a residual category.28
Liguria 1.050 0.262 0.450 1.619 The results of these additional estimates are reported in Table 8 and
Lombardia 1.130 0.334 0.450 2.556 are generally consistent with the ones in Tables 5 and 7, as far as the
Marche 1.077 0.341 0.450 2.623
impact of environmental variables as well as the efficiency scores are
Molise 1.113 0.321 0.475 1.355
Piemonte 1.033 0.355 0.450 1.926 concerned. Furthermore, we can also notice that with respect to the
Puglia 1.201 0.243 0.450 1.481 group of “Other” museums the “Art” museums are more efficient in
Sardegna 1.210 0.283 0.450 1.771 providing the service potential, whereas the “Hist_Science&technology”
Sicilia 1.272 0.357 0.450 2.545 museums are less efficient, although the dummy is significant only in the
Toscana 1.163 0.321 0.450 2.900
Trentino Alto-Adige 1.0903 0.289 0.450 2.054
baseline efficiency model.
Umbria 1.1388 0.265 0.517 1.501 Summing up, these further estimates provide evidence of the
Valle D’Aosta 1.1154 0.298 0.450 1.351 robustness of our efficiency assessment and of the capability of the
Veneto 1.1079 0.314 0.450 1.732 generalised conditional efficiency model to properly evaluate the impact
of environmental factors.
North-West 1.083 0.330 0.450 2.556
North-East 1.098 0.322 0.450 2.561
Centre 1.149 0.315 0.450 2.900 6. Conclusions
South 1.192 0.430 0.450 3.984
Islands 1.243 0.326 0.450 2.545 In this paper we estimate the technical efficiency of museums with a
Private 1.100 0.392 0.450 3.984 more focused identification of their production function, including only
Public 1.156 0.313 0.450 2.900
state-owned 1.348 0.171 0.450 2.496
the outputs whose realization is under the direct control of their man­
Non-state-owned 1.133 0.318 0.450 2.900 agers, what we have been referring to as the service potential. In other
words, we exclude, in our estimation of technical efficiency, some of the
Source: our elaboration on data provided by ISTAT. Note: m ¼ 300; no. of
traditional outputs used in most works on the efficiency of museums,
bootstrap replications ¼ 500
like the number of visitors, which seem to belong more to the realm of
outcomes and surely are the result, at least partially, of factors out of the
our sample of museums.
control of the museums’ managers. At the same time, however, we are
In Table 6, we report the summary statistics of unconditional effi­
aware that the efficient operation of museums can be conditioned by
ciency estimates for this alternative model, which employ the number of
workers and the exhibition area as inputs and opening days, temporary
exhibitions and the number of exhibited pieces as outputs. The results in
28
Table 6 are substantially in line with the ones in Table 4. The More precisely, according to the nomenclature of the ISTAT survey, the first
geographical distribution of the average efficiency values at the regional group (Art) includes the art museums (art collections from the 5th up to the end
level, represented in Fig. 4, seems to confirm this evaluation. of the 19th century) and the modern and contemporary art museums (art col­
A further check of the consistency of the efficiency estimates of the lections from the 20th century up to now). The second group (Hist_Scien­
ce&technology) includes all the history, natural history and natural sciences,
two models can be carried out through the use of the univariate kernel
science and technology, industrial museums. The third group (Other) includes
smoothing distribution and its appropriate bandwidth [32]. The kernel
religion, archeology and other museums.

10
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Fig. 4. Geographical distribution of unconditional efficiency estimates for the efficiency model in 5.3 (robustness check), average value by region.
Source: our elaboration on data provided by ISTAT

what are generally known as environmental and contextual factors. By considering the contextual factors, the average potential increase in
focusing on proper outputs (i.e. the service potential) and by considering output reduces remarkably. Our results show that the operational
the potential impact of environmental factors on efficiency, we believe environment matters: neglecting its impact would lead to incorrect
to have been able to provide a more accurate estimate of efficiency and a conclusions on museums’ (in)efficiency. More specifically, income
more appropriate information for policy decisions. levels and the size of the hospitality sector have a favourable and sig­
We used a recent development in the class of non-parametric frontier nificant influence on efficiency, in line with the idea that a stronger
estimators, namely a generalised conditional efficiency model - the demand for cultural goods makes pressure on museums towards a more
extended version developed by De Witte and Kortalainen [28] - to efficient use of resources. This brings about an interesting policy sug­
consistently assess the efficiency of Italian museums, using a large gestion regarding the design of incentives to foster technical efficiency
dataset based on a survey conducted by the ISTAT. The size of the of museums. Visitors are by and large considered the main indicator for
sample and the wideness of the data in the survey allow a technically museums evaluation. This indicator might be misleading as it in­
rigorous estimation of efficiency. corporates the effects of environmental features. This is very much so in
Results from the unconditional estimate show that Italian museums countries like Italy where we observe marked differences across regions.
could increase their service potential by 6.8% and that strong differences Our analysis warns that the attention of policy makers has to concentrate
exist among regions and geographical areas. However, when on outputs which are under the managerial control when evaluating

11
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

museum efficiency.
Finally, the ownership and the nature of the public governance of
museums are also relevant. Private institutions are relatively more
efficient and, among public museums, state-owned ones are relatively
less efficient. The implications of this result may be varied. First of all,
there is room for the improvement of efficiency of state institutions and
it may be useful, for each of these institutions, to consider operational
models borrowed by comparable private and local institutions, when
applicable, given that state-owned museums have on average a larger
size. Second, for selected institutions, the state could consider the pos­
sibility of different governance models, in terms of a decentralization of
their management, at the private or local level, even maintaining their
ownership.

CRediT authorship contribution statement

Calogero Guccio: Conceptualization, Formal analysis, Writing -

original draft, Writing - review & editing. Marco Martorana: Concep­
Fig. 5. Kernel density estimates for the baseline model and for the efficiency tualization, Data curation, Methodology, Writing - original draft,
model in 5.3 (robustness check). Writing - review & editing. Isidoro Mazza: Conceptualization, Writing -
Source: our elaboration on data provided by ISTAT original draft, Writing - review & editing. Ilde Rizzo: Conceptualization,
Writing - original draft, Writing - review & editing, Conceptualization,
Writing - original draft, Writing - review & editing.

Table 7
Conditional efficiency estimates for the efficiency model in 5.3 (robustness check).
Institutional and Environmental Variables

Unconditional Conditional

(1) (2)a (3)

Influence P-Value Influence P-Value Influence P-Value

Public – Unfavourable *** – – Unfavourable ***

No_state – – – Favourable *** – –
Number_beds – – – – – Favourable ***
GDP – – – – – Favourable *
Expenditure_culture – – – – – Favourable Not significant
Mean efficiency score 1.137 1.103 1.117 1.080
St. dev. efficiency score 0.343 0.331 0.288 0.279
Observations 2496 2496 1630 2496
R-square – 0.048 0.044 0.102

Note: Table presents estimates of the conditional FDH model with m ¼ 300 and no. of bootstrap replications ¼ 500; (2) favourable indicates an efficiency-enhancing
association; (3) asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels respectively; a estimated on the relative subsample of public museums.
Source: our elaboration on data provided by ISTAT.

Table 8
Controlling for museum specialization.
Institutional and Environmental Variables Baseline model Efficiency model (robustness check)

(1) (2)

Influence P-Value Influence P-Value

Public Unfavourable Not significant Unfavourable ***

Number_beds Favourable ** Favourable ***
GDP Favourable ** Favourable *
Expenditure_culture Favourable Not significant Favourable Not significant
Hist_Science&technology Unfavourable * Unfavourable Not significant
Art Favourable *** Favourable Not significant
Mean efficiency score 1.019 1.073
St. dev. efficiency score 0.590 0.243
Observations 2496 2496
R-square 0.141 0.105

Note: Table presents estimates of the conditional FDH models with m ¼ 300 and no. of bootstrap replications ¼ 500; (2) favourable indicates an efficiency-enhancing
association; (3) asterisks indicate significance at the 1% (***), 5% (**), and 10% (*) levels respectively.
Source: our elaboration on data provided by ISTAT.

12
C. Guccio et al. Socio-Economic Planning Sciences xxx (xxxx) xxx

Acknowledgements [26] Daraio C, Simar L. Conditional nonparametric frontier models for convex and
nonconvex technologies: a unifying approach. J Prod Anal 2007;28:13–32.
[27] Daraio C, Simar L. Advanced robust and nonparametric methods in efficiency
We wish to thank two anonymous referees for their careful review, analysis — methodology and applications. New York: Springer Science; 2007.
an anonymous associate editor, and the Editor, Professor Vedat Verter, [28] De Witte K, Kortelainen M. What explains the performance of students in a
for his advice. The usual disclaimer applies. heterogeneous environment? Conditional efficiency estimation with continuous
and discrete environmental variables. Appl Econ 2013;45(17):2401–12.
[29] Li Q, Racine JS. Nonparametric estimation of conditional CDF and quantile
Appendix A. Supplementary data functions with mixed categorical and continuous data. J Bus Econ Stat 2008;26(4):
423–34.
[30] Racine J. Consistent significance testing for nonparametric regression. J Bus Econ
Supplementary data to this article can be found online at https://doi. Stat 1997;15(3):369–78.
org/10.1016/j.seps.2020.100891. [31] Peacock A. A future for the past, Keynes Lecture. Proc Br Acad 1994;87:198–226.
[32] Simar L, Wilson P. Statistical inference in nonparametric frontier models: recent
developments and perspectives. In: Fried HO, Knox Lovell CA, Schmidt SS, editors.
References The measurement of productive efficiency and productivity growth. New York:
Oxford University Press; 2008. p. 421–521.
[1] Del Barrio-Tellado MJ, Herrero-Prieto LC. Modelling museum efficiency in [33] Pignataro G. Performance indicators. In: Towse R, editor. A handbook of cultural
producing inter-reliant outputs. J Cult Econ 2019;43:485–512. economics. second ed. Cheltenham, UK and Northampton, MA, USA: Edward Elgar;
[2] Bradford DF, Malt RA, Oates WE. The rising cost of local public services: some 2011. p. 332–8.
evidence and reflections. Natl Tax J 1969;22:185–202. [34] Førsund FR. Economic interpretations of DEA. Soc Econ Plann Sci 2018;61:9–15.
[3] Parks RB, Baker PC, Kiser L, Oakerson R, Ostrom E, Ostrom V, Wilson R. Consumers [35] Emrouznejad A, Yang GL. A survey and analysis of the first 40 years of scholarly
as coproducers of public services: some economic and institutional considerations. literature in DEA: 1978–2016. Socio Economic Planning Science 2018;61:4–8.
Pol Stud J 1981;9(7):1001–11. [36] Becker GS. A theory of the allocation of time. Econ J 1965;75:493–517.
[4] Hammond CJ. Efficiency in the provision of public services: a data envelopment [37] Lancaster KJ. A new approach to consumer theory. J Polit Econ 1966;74(2):
analysis of UK public library systems. Appl Econ 2002;34(5):649–57. 132–57.
[5] De Witte K, Geys B. Evaluating efficient public good provision: theory and evidence [38] Førsund FR. Measuring effectiveness of production in the public sector. Omega
from a generalised conditional efficiency model for public libraries. J Urban Econ 2017;73:93–103.
2011;69(3):319–27. [39] Cordero JM, Polo C, Santin D, Sicilia G. Monte-Carlo comparison of conditional
[6] Pignataro G. Measuring the efficiency of museums: a case study in Sicily. In: nonparametric methods and traditional approaches to include exogenous variables.
Rizzo I, Towse R, editors. The economics of heritage: a study in the political Pac Econ Rev 2016;21(4):483–97.
economy of culture in Sicily. Cheltenham, UK and Northampton, MA, USA: Edward [40] Narb� on-Perpin�a I, De Witte K. Local governments’ efficiency: a systematic
Elgar; 2002. p. 65–78. literature review—part II. Int Trans Oper Res 2018;25:1107–36.
[7] Basso A, Funari S. A quantitative approach to evaluate the relative efficiency of [41] Banker RD, Morey RC. Efficiency analysis for exogenously fixed inputs and outputs.
museums. J Cult Econ 2004;28:195–216. Oper Res 1986;34:513–21.
[8] Taheri H, Ansari S. Measuring the relative efficiency of cultural-historical museums [42] Ponzini D. The process of privatisation of cultural heritage and the arts in Italy:
in Teheran: DEA approach. J Cult Herit 2013;14:431–8. analysis and perspectives. Int J Herit Stud 2010;16(6):508–21.
[9] Mairesse F, Vanden Eeckaut P. Museum assessment and FDH technology: towards a [43] Farrell M. The measurement of productive efficiency. J Roy Stat Soc 1957;120:
global approach. J Cult Econ 2002;26:261–86. 253–90.
[10] Del Barrio MJ, Herrero LC. Evaluating the efficiency of museums using multiple [44] Fried HO, Lovell CK, Schmidt SS, Schmidt SS, editors. The measurement of
outputs: evidence from a regional system of museums in Spain. Int J Cult Pol 2014; productive efficiency and productivity growth. New York: Oxford University Press;
20:221–38. 2008.
[11] Del Barrio MJ, Herrero LC, Sanz JA. Measuring the efficiency of heritage [45] Asatryan Z, De Witte K. Direct democracy and local government efficiency. Eur J
institutions: a case study of a regional system of museums in Spain. J Cult Herit Polit Econ 2015;39:58–66.
2009;10:258–68.
[12] Wang S, Cheng E, Zhu J, Fu C, Wang W. Using DEA models to measure the
performance of public culture services in China. Proceedings - 2016 International Calogero Guccio (Ph.D., Catania) is professor of public economics at the Department of
Conference on Computational Science and Computational Intelligence, CSCI 2016 Economics and Business - University of Catania. His research interests include theoretical
2017:447–52. art. no. 7881384. and empirical aspects of the evaluation of efficiency of private and public subjects. His
[13] Basso A, Casarin F, Funari S. How well is the museum performing? A joint use of research output has been published in leading outlets in the field of economics and public
DEA and BSC to measure the performance of museums. Omega 2018;81:67–84. finance.
[14] Bishop P, Brand S. The efficiency of museums: a stochastic frontier production
function approach. Appl Econ 2003;35:1853–8. Marco Ferdinando Martorana (Ph.D., Catania) is research fellow at the Department of
[15] De Witte K, Geys B. Citizen coproduction and efficient public good provision: Economics and Business of the University of Catania. His research areas of interest include
theory and evidence from local public libraries. Eur J Oper Res 2013;224(3): econometrics, public economics, cultural economics and voting participation. His recent
592–602. works focus on the evaluation of public sector efficiency in the fields of education and
[16] Simar L, Wilson P. Estimation and inference in two-stage, semi-parametric models cultural heritage.
of production processes. J Econom 2007;136:31–64.
[17] Cazals C, Florens JP, Simar L. Nonparametric frontier estimation: a robust
Isidoro Mazza (Ph.D., Amsterdam) is professor of public finance at the University of
approach. J Econom 2002;106:1–25.
Catania. His main research interests include migration, lobbying, fiscal federalism,
[18] Daraio C, Simar L. Introducing environmental variables in nonparametric frontier
patronage and voluntary provision of public goods, voting participation, administration of
models: a probabilistic approach. J Prod Anal 2005;24:93–121.
cultural heritage, art market and education.
[19] ISTAT. Indagine sui musei e le istituzioni similari: microdati ad uso pubblico.
Roma: ISTAT; 2015. https://www.istat.it/it/archivio/167566.
[20] Mignosa A. Theory and practice of cultural heritage. In: Rizzo I, Towse R, editors. Giacomo Pignataro (Ph.D, York) is professor of Public Finance at the Department of
The artful economist. Springer International Publishing; 2016. p. 227–44. Economics and Business of the University of Catania. He was Rector of the University of
[21] Bodo C, Bodo S. Country profile: Italy. In: Council of Europe/ERICarts, Compendium Catania. His research interests include several fields, ranging from cultural economics,
of cultural Policies and Trends in Europe. eighteenth ed. 2016 Available under:, efficiency of public expenditure, public procurement to health economics. His research
2222–7334 http://www.culturalpolicies.net. output has been published in leading outlets in the field of economics and public finance.
[22] Federculture. Impresa Cultura. Comunit� a, territori, sviluppo. Roma: Gangemi
Editore; 2018.
Ilde Rizzo (DPhil, Buckingham.) is professor of public finance at the University of Catania.
[23] Unioncamere. Io sono cultura. Rome: Quaderni di Symbola; 2018.
She was Pro Vice-Chancellor of the University of Catania. She received a degree of Doctor
[24] Bertacchini E, Dalle Nogare C, Scuderi R. Ownership, organization structure and
of Science honoris causa from the University of Buckingham, UK. She has published
public service provision: the case of museums. J Cult Econ 2018;42:619–43.
monographs, edited books and referred articles in professional journals. Her research is
[25] Zan L, Bonini Baraldi S, Santagati ME. Missing HRM: the original sin of museum
interdisciplinary and covers several fields, ranging from cultural economics, efficiency of
reforms in Italy. Mus Manag Curatorship 2018;33(6):530–45.
public expenditure, public procurement to health economics.

13