Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
An assessment of the performance of Italian Public Historical Archives: Preservation vs utilisation
Running Title: An assessment of the performance of Italian PHAs
Calogero Guccio*a, Marco Martoranaa, Isidoro Mazzaa, Giacomo Pignataroa,b, Ilde Rizzoa
a Department of Economics and Business, University of Cataniab Department of Management, Economics and Industrial Engineering, Politecnico di Milano
Abstract
Economists have paid an increasing attention to the estimation of efficiency in the public sector. Archives constitute an exception notwithstanding their cultural and historical relevance. This paper is, to the best of our knowledge, the first attempt to fill this lacuna. The analysis of the performance of archives raises theoretical issues because their outputs have characteristics of a pure public good (preservation) and of a private service (utilisation). Moreover, the trade-off between the long and short-run goals of, respectively, preservation and utilisation is central in most public policies concerning heritage. We address this issue by studying the efficiency of public historical archives (PHAs) in Italy, over the period 2011-2012. We use Data Envelopment Analysis (DEA) and implement one-stage and two-stage approaches to investigate PHAs’ performance in the management of their different and potentially conflicting functions. Our analysis shows that PHAs perform better in the preservation function, considering the objectives and constraints they face.
Keywords: Cultural heritage; Archives; Efficiency; Network DEA; Nondiscretionary and intermediate measure; Non-parametric frontier
JEL Classification: D24; Z10; Z181. Introduction
*a Corresponding author: Corso Italia, 55 – 95129 Catania, [email protected], tel. +39 095 7537744, fax +39 095 7537710
1
Archives perform the function of storing and preserving data for future generations
(e.g. the Arctic World Archive) and, in many cases, operate as cultural institutions that
give access to valuable records and items they protect. Since “archives are preserved
for use by present and future generations” (International Council on Archives, 2012;
p.6), they have two different and potentially contrasting missions, a long-run one – to
preserve - and a short-run one - to supply services on demand for various purposes
(research, education and entertainment) - which, however, may cause the deterioration
of the preserved items
Several projects have developed across the world (see the E-ARK project in the
European Union) to exploit digital technologies and develop a common international
standard able to improve the utilisation and valorisation of archival collections, without
jeopardizing their preservation. However, insofar as the process of digitisation of
records is still far from being completed, the fundamental trade-off between
preservation and utilisation (or valorisation) still exists, and necessarily influences the
policies implemented. This is the case, for instance, of the Italian public historic
archives (Archivi di Stato - PHAs) whose extensive and fragile patrimony highlights
the problem of combining the preservation of valuable items with their utilisation on
demand and broader valorisation to the benefit of the public. The design of policies,
especially those regarding the allocation of resources within the spectrum of activities
and functions of cultural institutions, as well as the managerial organisation face this
general trade-off between preservation and utilisation. Since these two functions serve,
at least partially, different interests, a crucial policy issue is whether the actual
performance of cultural institutions in pursuing these two functions is consistent with
the set of relevant social preferences.
2
The economic investigation of the performance of provision of cultural services is
longstanding. It has developed using different methodologies and approaches and
concerned several fields: museums, theatres, libraries, and heritage authorities. Among
the different aspects of the performance of organisations providing cultural services,
efficiency is probably the mostly explored one with a focus on technical efficiency,
namely the maximisation of the output of a production process for given quantities of
inputs. 1
The measurement of the differential efficiency in performing the functions of
preservation and utilisation is a crucial step in exploring the specific policy issue
highlighted above. This is a quite complex technical problem for the technical
interrelations existing between preservation and utilisation in the identification of the
production function. In this paper we measure the technical efficiency of PHAs in Italy,
employing an official detailed dataset provided by the Italian Ministry for Cultural
Goods (Ministero dei beni e delle attività culturali – MiBAC) and by the State General
Accounting Department (Ragioneria Generale dello Stato – RGS) of the Italian
Ministry of Economy and Finance. We use different non-parametric approaches,
including one- and two-stage models, to account for the peculiar features of PHAs and,
above all, to offer a meaningful picture of Italian PHAs’ performance, in terms of
efficiency of the two functions. This analysis is instrumental to discuss the policy issue
of how managerial decisions impact on the trade-off in performance between
preservation and utilisation. To the best of our knowledge, this paper represents the first 1 Among the others, see: Mairesse and Vanden Eeckaut (2002), Pignataro (2002), Bishop and Brand (2003), Del Barrio et al. (2009), Del Barrio and Herrero (2014) and Del Barrio et al. (2019), for museums; Hammond (2002), Chen et al. (2005), Simon et al. (2011), De Witte and Geys (2011, 2013), Borges et al. (2018) and Guccio et al. (2018a), for libraries; Zieba, (2011), Last and Wetzel (2011), Castiglione et al. (2018) and Fernández-Blanco et al. (2018), for theatres; Finocchiaro Castro and Rizzo (2009) and Finocchiaro Castro et al. (2011), for heritage authorities. Pignataro (2011) and Fernández-Blanco et al. (2013) provide surveys of the use of these techniques in the cultural sector. Frontier methods have been commonly used to deal with the economic impact of culture and cultural related traits; see, for instance, Ventelou and Bry (2006) and Lagravinese et al. (2019).
3
attempt to analyse the technical efficiency of archives, filling a lacuna that contrasts
with the relevant role played by these institutions in Italy and worldwide.
The specific function of PHAs is to preserve primary sources that are available only
upon requests and cannot be borrowed. PHAs, therefore, provide services combining
public (preservation) and private (utilisation) characteristics similarly to museums and
public libraries, which have in common with the former the nature of memory
institutions, though with some specificities. On the one hand, archival records are in
most cases unique, unlike libraries, where publications are in general replaceable. This
feature relates archives to museums. On the other hand, PHAs have services on demand
that are rare in the latter. Moreover, archives are bound to accept and preserve records
and documents of selected offices according to specific disposal procedures2 whereas
museums can have a voice in accruing their collection.
The twofold objective of PHAs - supplying services on demand for various purposes
(research, education, entertainment) and preserving valuable items and records for
future generations - is, as discussed earlier, familiar to cultural management and may
have a striking influence on the outcome of public policies. The different performance
of PHAs with respect to their functions can assume a policy relevance inasmuch it may
contrast with social preferences for the two functions, and it may be explained looking
at some relevant features of the decision-making context characterising these
institutions: the resource constraints and the incentives faced by the institutions’
managers; their institutional goals; their personal characteristics, such as preferences
and educational background.
2 The attributions and obligations of Italian PHAs and of the central and peripheral offices of the Italian State, with respect to official records conservation, are ruled by law no. 1409/1963 (art. 1) and by the Italian Code of Cultural Goods (law no. 42/2004, art. 41).
4
The quality of bureaucracy is relevant for reaching a specific output (Adams and
Klobodu, 2017) and the management of PHAs is generally attributed to people with an
archivist educational background. As it often happens with professionals with
managerial roles in cultural institutions, their objectives may be strongly influenced by
their reputation within the profession and by their primary adherence to specific
professional standards. However, the specific educational background and experience
of managers of PHAs may not provide them with the economics and management skills
needed to deal with utilisation activities. According to this reasoning, we would expect
that PHAs would perform better in preserving the stored items rather than facilitating
various utilisations by the public. Within this view, the recent proliferation of public
projects to improve digitisation of PHAs in Italy3 can be interpreted as an attempt by
the national policymaker to overcome the natural managerial risk-aversion for
utilisation and valorisation, improving fruition without risk for conservation.
Very little attention has been so far devoted to PHAs in the cultural economics
literature, with the only exceptions being Borowiecki and Navarrete (2017) and Guccio
et al. (2016), which consider the implications of digitisation. In addition to provide
insights on the specificities of PHAs helpful to define interventions improving their
performance, the present study contributes to the empirical literature on the efficiency
of cultural institutions in several ways. Firstly, because of the peculiar nature of the
preservation function of PHAs, we are able to overcome some of the shortcomings in
the measurement of efficiency of cultural institutions, above all those due to the lack of
homogeneity of outputs across organisations. Secondly, by using both physical and
monetary measures of the size of the collection managed by PHAs, we move further in
3 See Guccio et al. (2016) for a comprehensive view of ICT programs involving Italian PHAs, including digitization. A list current or already completed digitization projects involving Italian PHAs, can be found on the websites of the Central Institute for Archives – ICAR and of the General Archive Directorate – DGA.
5
overcoming the homogeneity problems inherent in the aggregate measure of “number
of pieces” customarily used in other studies. Finally, our empirical strategy will allow a
consistent and separate estimation of efficiency of preservation and utilisation, which
can provide information on how the managers of these institutions weigh these
activities in their managerial choices, enabling us to study their operational relationship
and derive policy implications.
As expected, among the most significant results of our empirical analysis, we show
higher efficiency scores in the preservation function than in the utilisation one, with
evidence of a trade-off between these two functions. Such results have relevant policy
implications and stress the importance of programs aiming at overcoming the
institutional and operational constraints faced by PHAs’ managers and at enhancing
users’ participation.
The study is organised as follows. Section 2 offers a brief overview of the theoretical
issues as well as the institutional features of PHAs in Italy. Section 3 describes the
methodology and the data used. Section 4 provides technical efficiency estimates.
Section 5 presents some concluding remarks.
2. Theoretical and institutional background
2.1 Theoretical issues
The measurement of efficiency of cultural institutions initially developed through the
use of productivity indicators and then evolved in several attempts to employ the
method of efficiency frontiers. The most challenging task faced by this growing
literature on the assessment of the technical efficiency of cultural institutions is perhaps
the characterisation of the production process, which is essential for the measurement
of efficiency and for its relevance for managerial choices. It actually determines the
6
perimeter of benchmarking and therefore it frames the evaluation of the relative
performance of the different institutions.
Cultural institutions are responsible for several functions and, therefore, their efficiency
needs to be evaluated for any production process. A first problem to deal with is
therefore related to the clear identification of appropriate measures of the different
outputs, which should capture the relevant aspects of the production process that are
homogenous across the different institutions, and separate proper outputs from
outcomes (De Witte and Geys, 2011). Several empirical papers investigate the
efficiency of museums and libraries. PHAs differ from museums and libraries because
their obligations to keep all public records make the preservation output easier to define
and measure than for the former. Moreover, the documents preserved are relatively
homogenous across the different units under examination. The homogeneity problem,
which is typical of most analyses of multi-output production, is further alleviated in this
study by the unique data set that includes the monetary value of each item collected in
the archives.4
A related issue, concerning the production function specification, is whether the
different outputs (or some of them) are “vertically integrated”, that is whether some of
them are inputs for other outputs or “independent” of each other. This aspect is
particularly relevant for an appropriate measurement of the efficiency of preservation
and utilisation activities of cultural institutions, and has significant implications in
terms of the type of frontier estimation model, which best suits the characteristics of
their production process and in terms of policy, since different assumptions on the
production process entail different possible interventions to improve PHAs’
performance. Preservation could be regarded as an intermediate output, which serves as 4 Monetary values are also used for libraries by Guccio et al. (2018a), and for movable cultural assets by Guccio et al.(2018b).
7
an input for the production of the (final) utilisation output (see, among others, Guccio
et al., 2018a). However, because of the potentially wide heterogeneity that can
characterise the objects preserved by similar cultural institutions, the contribution of
preservation to utilisation, in terms of their input-output relation, may be very
idiosyncratic within the benchmarking set and highly dependent on the qualitative
characteristics of objects. This problem is less severe for PHAs, since, as noted earlier,
objects preserved in PHAs are relatively homogenous.
Regarding the problem of how to model the relation between preservation and
utilisation, there are several potential approaches that can be taken into account. In fact,
preservation activity can be considered as a non-discretionary input, so as to reflect the
obligation of PHAs in preserving archives, or it can be regarded as an intermediate
output, which allows for studying the true operational relationship among the two
activities. Since nonparametric techniques do not allow for statistical tests to check the
empirical validity of the specification of the production function, our strategy consists
in employing and comparing outcomes from the above approaches.
2.2 Italian Public Historical Archives: activities and organisation
In Italy, PHAs perform archival functions under the oversight of the Italian Ministry for
Cultural Goods (Ministero dei beni e delle attività culturali– MiBAC), specifically its
Archives General Directorate, which coordinates the activities of PHAs from a
technical and scientific point of view. Without engaging in a thorough analysis of the
ministerial organisation, it is worth mentioning that PHAs, as other decentralised Italian
State cultural organisations (such as museums and libraries) have a limited autonomy in
the management of inputs, human resources in particular. They cannot hire directly
their staff nor choose the specific professional profiles since employees are allocated by
the centre. The organisation of work is not flexible because of the high degree of 8
unionisation.5 Overall, a formalistic approach prevails: controversies between
employees and managers are often debated in front of administrative courts.6
Italian PHAs are 101: one National Central Archive (Archivio Centrale di Stato), which
has the main responsibility of managing the archives of the central offices of the Italian
unitary State, and the remaining 100 located in the provincial capital towns. The latter
pursue the objective of conserving pre-unitary archival documents as well as those
produced by the peripheral administrations of the unitary State. PHAs carry out the
main responsibility of conserving the records of procedures concluded at least 30 years
before. PHAs also oversee the management of state archives that are still in use and the
discarding of documents of no historical interest.7 Some PHAs also have branches
(Sezioni), which are overall 33, located in minor cities and established because of the
high cultural relevance of the documents existing within the specific area in which they
operate. Furthermore, 17 PHAs run a School of archive studies for the training of the
technical staff of PHAs as well as of those interested in historical research and in the
management of archives.
PHAs’ collections contain several types of records, ranging from paper documents,
parchments and seals to photos, sound recordings, films, drawings, musical scores, etc.
The consultation of archival records can be done mainly by visiting the archive
building and accessing the required material on site, since the process of full
digitisation of PHAs’ collections is still incomplete.
Several policies have been adopted in Italy (as well as in other countries) to increase
the interoperability of archives’ records, employing the most common international
5 For Italian State museums, Zan et al (2018) stress the lack of a human resource strategy and the predominance of a ‘protectionist approach’ to safeguard longstanding vested interests. 6 Limited resources are an additional trait. In general, public expenditures on culture have been quite stagnant in the last decades (Pistoresi et al, 2017). 7 The protection of private documents pertains to 19 Archive Soprintendenze, which are responsible for issuing 'declarations of cultural interest’ for the documents of historical interest.
9
standards, to coordinate preservation activities and to improve the access to preserved
items. They include the implementation in 2004 of a unique informative system for
PHAs, to gather and harmonize the information on PHAs and their collections, and the
institution of the National Archival System (SAN) in 2010. The SAN coordinates the
MiBAC and Italian local jurisdictions and aims at promoting preservation activities and
developing utilisation services, as well as “to spread innovation and best practices” and
also provides the access to the information on Italian archives and the preserved
heritage. Moreover, programs to extend the digitisation of preserved items have been
developed to enlarge the possibilities of preservation as well as of utilisation. On the
one hand, for very old papers or documents, whose inspection is risky, digitisation
might prevent further decay and reduce the conflict between the objectives of
preservation vs. utilisation. On the other hand, digital access allows users to know the
content of collection and to inspect documents without visiting the respective PHAs,
with positive effects on utilisation. Thus, digitisation may be a strategic policy tool to
promote and enhance cultural participation which is a recognised priority at European
and national level (Ateca-Amestoy, 2018).
3. Methods and data
3.1 The empirical strategy
The measurement of technical efficiency of Italian PHAs (our Decision Making Units -
DMUs) is based on the Data Envelopment Analysis – DEA (Charnes et al. 1978) that,
given a sample of n DMUs and the set of relevant inputs and outputs, uses linear
programming techniques to estimate a frontier envelopment surface, and returns for
10
each DMU a measure of (in)efficiency and the potential improvement available to
inefficient DMUs to reach the best practice frontier8.
The application of the DEA requires the identification of inputs and outputs, i.e. the
specification of the relevant production function. A crucial issue for cultural
institutions, including PHAs, is the way preservation activity is treated. We consider
alternative approaches to take this characteristic of the production process of PHAs into
account.
First of all, we employ a basic one-stage approach that focuses on the utilisation
function and disregards preservation activity. This is the simplest model in terms of
assumptions about the production process of PHAs. However, PHAs differ in terms of
size and value of collections, and, additionally, their preservation can be operationally
relevant. If so, the basic one-stage approach would not evaluate fairly PHAs’
performance and, in turn, the derived policy prescriptions would be incorrect. Thus,
just as a preliminary analysis, we estimate efficiency using the basic one-stage
approach,9 and compare kernel density estimates of the efficiency scores that rely on
the reflection method,10 for groups of PHAs identified on the basis of different levels of
preservation activities, as measured by the volume or the value of their collections to
check whether collections are relevant.
8 In what follows we choose an output oriented approach and assume constant returns of scale (CRS). Both the choices are needed to provide a fair and reasonable comparisons between different estimated models. More specifically, the output orientation is necessary as the alternative, i.e. the input orientation, would requires the reduction of the used input to achieve higher levels of efficiency: in the two-stage setting, this alternative would illogically imply the destruction of the collection in the utilisation stage to improve PHAs’ efficiency. As for the CRS assumption, we also perform the Banker (1996) test for the efficiency estimates. The results, available upon request, show that we cannot reject the null hypothesis of CRS at any conventional level of significance.9 To account for the relative small dimension of our sample we employ a consistent bootstrap estimation procedure (Simar and Wilson, 2000) to obtain the sampling distribution of the efficiency scores and derive bias corrected scores.10 In such a way, we are able to avoid the problems of bias and inconsistency at the boundary of support (Simar and Wilson, 2008).
11
Given that kernel estimates provide evidences of a systematic variability of the
efficiency scores with the size of collections, we introduce preservation in the
efficiency estimate as a nondiscretionary input, following Banker and Morey (1986). In
fact, there could be reasons to consider a PHA’s collection as an “uncontrollable”
factor for its management, since it is the outcome of the fulfilment of an obligation to
maintain the records of government offices (according to the specific rules set in the
law). This approach reasonably leads to a better assessment of PHAs performance but it
would not provide any information on the efficiency of preservation since, “by
definition”, there would be no discretionary choice about preservation outputs to be
evaluated on the management’s side.
To allow for a separate evaluation of PHAs’ performance in carrying out their
preservation and utilisation functions, and to derive conclusions on the operational
relationship between the two activities, we employ multistage DEA models (Färe and
Grosskopf, 1996) that assumes the production process to be made of a sequence of
technologies. In a first stage, inputs produce intermediate outputs, which, then, become
inputs in the second stage, where the final outputs are produced.11
In order to facilitate the interpretation of the results in the next section, it is useful to
consider a generic two-stage network structure (Chen and Zhu, 2004). Given the n
DMUj (j=1,…, n) to be evaluated, the inputs and outputs of the j-th DMUs can be
grouped into a two-stage process. Namely, the DMUj uses m inputs xij (i=1,…, m) in the
first stage, and produces D outputs zjd (d=1,…., D) in that same stage. These D outputs
then become the inputs in the second stage whereas yrj, (r=1, …, s) are the outputs of
that stage. Using the Constant Returns to Scale (CRS) DEA model, it is assumed that
for DMUj we might define the following efficiency measure in each stage:11 The production structure could be also articulated in more stages and thus, the two-stage can be considered as a special case of network DEA models (Halkos et al., 2014; Kao, 2014).
12
e j1=
∑d=1
D
wd zdj
∑i=1
m
v i x ij
and e j2=
∑r=1
s
ur yrj
∑d=1
D~wd zdj
(1)
where e j1 and e j
2 denote respectively the efficiency in the first and in the second stage
and vi, wd, ~w d are unknown non-negative weights.
We first apply the two-stage model developed by Seiford and Zhu (1999), assuming
that PHAs use their inputs to preserve collections in the first stage and, then, they use
these assets to produce utilisation activities in the second stage. This is in line with the
so-called “independent two-stage DEA approach” (Halkos et al., 2014) and implies the
application of the standard DEA methodology separately in the first and in the second
stage to calculate e j1 and e j
2, independently. In such a way, it allows to identify the sub-
process efficiencies and the resulting information can be used to determine the reasons
of poor performance in each stage.12
However, this method does not address the potential conflict between the two stages as
arising from the use of intermediate measures (Kao, 2014). To deal with it, we then
apply the centralized network DEA estimator proposed by Liang et al. (2008) and
refined by Kao and Hwang (2008), which basically involves that the efficiency scores
of the two stages are maximized simultaneously. In addition to the efficiency measures
e j1 and e j
2, it is often useful for purpose of comparison to consider a measure of the
global efficiency, namely e jG, which refers to the overall two-stage production process.
Following Liang et al., (2008) we compute the global efficiency measure e jG as the
product of the single stage efficiency measures.13
12 Using this approach, we follow the more recent literature in the related field (e.g. Hammond, 2002; Simon et al., 2011).
13
3.2 Data and specification of the models
For each PHA, we use data drawn from the Italian MiBAC database collected on yearly
basis by the Ministry’s Statistical office, and data on the monetary value of the
collections estimated by the State General Accounting Department (Ragioneria
Generale dello Stato - RGS). The final sample, after the usual analyses for reporting
errors, outliers and missing values, consists of cross-sectional and time series
observations for 99 Italian PHAs for 2 years (2011-2012), thus resulting in 198
observations.
Data include information on the relevant inputs and outputs of PHAs’ functions of
preservation and utilisation. The lack of available data related to the digitisation and to
the online and digital use of collections does not allow for considering all the aspects of
the PHA’s functions.14
The capital inputs are measured by two variables: the total surface area (SURFACE),
measured in square meters, is used as an indirect measure of the infrastructure, since
most of the space in a PHA will be used for storing the collection, and also gives a
rough idea of the scale and importance of the building, housing the PHA’s collection;
and the total shelves length (SHELF), measured in linear meters, as an indicator of the
equipment and the services, which are deemed essential for a PHA to undertake many
of its activities. Current inputs should be well reflected by the number of personnel
(PERS) and by the current PHA expenditure (EXP), excluding current labour costs.
13 Global efficiency, e jG, can be computed as either the mean – that is, e j
G=12¿ + e j
2¿ - or the product
of individual efficiency measures – that is, e jG=e j
1 ∙ e j2. Both definitions ensure that the overall two-
stage production process is efficient if and only if e j1=¿ e j
2=1 (Cook and Zhu, 2014). 14 Digitisation enlarges the possibilities of preservation as well as of utilisation, as previously mentioned. Even if it may be possible that our partial representation of the utilisation function might affect the level of efficiency, the extent of digital access is currently very limited. See on this point Guccio et al. (2016).
14
The size of collection can be measured, first of all, in physical terms. We consider three
variables represented in terms of number of pieces: manuscripts and documents
(M&D), collections of antiquaries (ANTIQUARIES) and other collections
(O_COLLECTION).15 Since these measures may pose problems of comparability of
collections, because of the heterogeneity across the different PHAs, we also consider
the monetary value of the collections as estimated by the RGS (ASSET_VALUE). 16
The utilisation outputs can be represented by several variables. The most common
measure, used in several studies (see section 1), is the number of visitors (VISIT).
Furthermore, we use the total number of consultations processed by PHAs for both
scientific and non-scientific purposes (RESEARCH), and the total number of
documents inspected (D_RESEARCH). Other possible outputs, such as exhibitions,
that PHA’s might organise to enlarge the number of users, cannot be taken into
consideration because of the lack of data.
Table 2 provides an illustration of inputs, outputs and models employed for the
measurement of efficiency as well as the descriptive statistics. We classify the
employed variables in three groups: a) inputs (SURFACE, SHELF, EXP and PERS); b)
size of collections (M&D, ANTIQUARIES, O_COLLECTION and ASSET VALUE);
c) final (utilisation) outputs (VISIT, RESEARCH and D_RESEARCH).
As for the specification of the models, consistently with the empirical strategy
discussed in section 3.1, in the preliminary one-stage analysis we estimate PHAs
efficiency in the pure utilisation function using the variables in group a as inputs, and
those in group c as outputs and analyse the distribution of scores for groups of PHAs
15 Collections of antiquaries include: parchments and scrolls, maps, seals, rubber-stamps, coins. Other collections include: photographs, negatives, microfilms, microfiches, audio-visuals. 16 The estimation criteria include the length of shelves, the state of conservation and the integrity of collections, the historical period to which the records belong, the environmental conditions of conservation, the relevance of collections for historical research, the rarity of collections. See Guccio et al. (2018b) for a detailed discussion on this estimate and its reliability.
15
characterised by different levels of preservation activities, as measured by the volume
or the value of their collections (group b variables).
Then, following the Banker and Morey (1986) approach, we include group b variables
as nondiscretionary inputs. Finally, in the two-stage analysis group b variables are
treated as an intermediate output (i.e. the output of the first stage analysis). For each
approach, we estimate two models (Mod_1 and Mod_2), which differ only in the
measurement of the size of collections. More specifically, Mod_1 includes the set of
physical measures (M&D, ANTIQUARIES and O_COLLECTION) while in Mod_2
we also consider the ASSET VALUE.
- Table 1 around here –
4. Efficiency results and discussion
4.1 One-stage approach
Figure 1 presents the kernel density estimates of the efficiency scores derived from the
basic one-stage approach for groups of PHAs identified by different levels of
preservation activities, as measured by the volume or the value of their collections.17
Namely, we consider three possible levels (high, intermediate and low levels according
to their sampling distributions) of group b variables.18 One would expect that the
distribution of efficiency scores should not show systematic differences across the
groups, if the size and relevance of collections do not matter. Clearly, this is not the
case: Figure 1 shows that, in general, differences in efficiency of PHAs are inversely
related with levels of collections. Evidently, the distributions of the groups are rather
different and the high-level group is generally more right-skewed, for all the measures
17 Estimates from the one-stage approach are available upon request.18 We also perform the kernel density estimates of the efficiency scores before and after bootstrap correction. The kernel density functions show that the efficiency estimates are robust with respect to sampling variation since there are only small differences between biased and biased corrected efficiency estimates. Results are again available upon request.
16
of collection but O_COLLECTION. Thus, collections matter and their inclusion is
needed to ensure a fair measurement of PHAs’ efficiency.
- Figure 1 around here -
4.2 One-stage approach with nondiscretionary input
In fact, the systematic effect of the size of collections disappears when including the
size of collections as a nondiscretionary input (Figure 2), supporting the Banker and
Morey (1986) approach.
- Figure 2 around here –
- Table 2 around here -
Looking at the summary statistics of the efficiency scores shown in Table 2, the first
result relates to the comparison between the two models (Mod_1 and Mod_2). Average
values are rather close, implying that the physical and monetary measures are similar in
capturing the effect of the size of collection on the efficiency of PHAs utilisation
activities. Second, the dynamics of efficiency in the two periods is slightly decreasing.
Other interesting aspects emerge with respect to some differences in the organisation of
PHAs. Apparently, the presence of a branch is favourable for the utilisation efficiency
while the reverse is true for the presence of a School. Finally, efficiency is higher on
average in Southern regions.
4.3 Two-stage approach
Results from the independent two-stage approach (Seiford and Zhu, 1999), which allow
for disentangling the two functions, are shown in Table 3.
- Table 3 around here -
17
Comparing the results across the two models, we notice that they are quite similar for
the first stage (average scores in the first stage are 0.64 and 0.66 for Mod_1 and
Mod_2, respectively), while average scores in the second stage are slightly lower for
the first model (0.42) than in the second model (0.49). We can then focus on the results
of Mod_2 for further comparisons, since it is inclusive of an additional dimension for
characterizing the size of the intermediate output.
Secondly, comparing the results of the two stages, we notice that the average efficiency
in the second stage, i.e. the utilisation, is remarkably lower than the average efficiency
of preservation. Such a difference can be due to several reasons. PHAs face a severe
institutional constrain since the law compels them to be fully committed to guarantee
preservation, which has a priority in the allocation of resources. Also the specific
educational background of the managers, more humanistic than managerial, may hint at
a bias of theirs towards preservation rather than utilisation. This result has also
important implications in terms of policy, as it indicates that the implementation of
digital technologies for archival storage and access easing the impact of utilisation on
preservation, may indeed be effective in overcoming the trade-off between the two
activities. Finally, we consider the differences with respect to the organization of
PHAs. The average efficiency of PHAs with a branch is higher than the corresponding
value for the overall sample only in the second stage, while it is lower in the first stage.
This result suggests that the existence of a branch enhances the accessibility of archives
and improves efficiency in the utilisation, lowering the conservation effort. The reverse
is true for the presence of a School because the training of archivists naturally increases
the conservation effort. The geographical divide (among the country’s three main
macro areas, North, Centre and South) offers different outcomes in the two stages:
Northern PHAs appear less efficient in conservation than the Southern ones while the 18
opposite holds for the utilisation function. Generally, there are clues of a potential
trade-off between the preservation and the utilisation functions.
Results are generally robust with respect to the use of the network DEA, which allows
for dealing with the potential conflict between the two activities (Table 4). While these
further results confirm the better efficiency in preservation than in utilisation, the
network DEA estimation penalizes the latter efficiency scores remarkably. A policy
suggestion to improve the utilisation efficiency might be to foster networks with other
cultural institutions, for instance using a well-established tool, such as tourist cards,
which so far have been extensively used in Italy but without taking archives into
consideration. Such a policy however implies cooperation among different levels of
government, provided local governments are responsible for cultural policies at local
level as well as for the management of several cultural institutions.
- Table 4 around here -
We finally present Pearson correlations coefficients in Tables 519 to provide an overall
comparison of the different approaches and model specifications and offer further
empirical support to some of the above results. The analysis of correlations confirms
that the scores of the two models (Mod_1 and Mod_2) are highly correlated, especially
within the two-stage approaches.
Regarding the two-stage approaches, the negative and significant correlation between
the scores from the two stages supports the hypothesis of a trade-off between
preservation and utilisation. Moreover, we can consider the correlation between the
results of the one-stage approach with nondiscretionary input with the results of the
two-stage approaches. This analysis is meaningful only when one contrasts the one-
stage estimates with the second stage scores in the two-stage approaches, because both
19 Spearman rank correlations offers analogous results. The relative table is available upon request.19
examine the utilisation function. The correlation coefficients are positive and
significant, and quite high. Therefore, a different definition of the production process
does not seem to affect the assessment of the efficiency of PHAs but does not allow for
discriminating PHAs’ behaviour in the management of the preservation and utilisation
functions.
- Table 5 around here –
5. Discussion and concluding remarks
In this study we estimate the technical efficiency of PHAs, taking into account their
two main functions, namely preservation and utilisation, using one-stage and two-stage
approaches.
We believe that the results of this paper can make a contribution, not only to explore
the efficiency of PHAs, still uncovered in the economic literature, but also to overcome
some of the shortcomings in the analysis of the efficiency of cultural institutions and to
provide insights to policy-makers for the design of policies aiming at improving PHAs’
performance in preservation and utilisation and to managers for their implementation.
From a methodological point of view, the case of PHAs seems to show that the two-
stage approach could be the most suitable to grasp the peculiar features of cultural
institutions activities. It allows disentangling the efficiency evaluation of preservation
and utilisation functions, otherwise included in a sort of ‘black box’ of the production
process and to explore the potential trade-offs between the two stages. The PHAs case
is particularly significant for the separate estimation of the efficiency of the two
functions, since the preservation activity of PHAs depends on the fulfilment of
obligations of records’ conservation (by law), which is not a matter of managerial
discretion as utilisation.
20
The comparison of the efficiency scores computed for the two stages outlines a better
performance of PHAs’ managers in preservation than utilisation. There are at least two,
somehow related, issues that can be raised here, which can be of general interest for the
management of cultural institutions. The different performance of PHAs with respect to
the two functions can be explained by taking into account the objectives, the incentives
and the constraints faced by the institutions’ managers. There are at least three
constraints that may be relevant for PHAs. The first one refers to the scarcity of
resources that limits the available choices in view of the goals. Second, the goals
themselves are constrained, as PHAs have conservation obligations. Third, a further
constraint may come from the archivist educational background of PHAs managers,
which may not provide them with the managerial skills needed to deal effectively with
utilisation activities, an issue that is common to many cultural institutions. Therefore,
the better performance in preservation activities that we find for PHAs is consistent
with an objective function “dominated” by professional background, in addition to the
constraints actually faced by the institutions’ management. Should the balance in the
performance of different functions be considered unsatisfactory, our remarks would
suggest to re-think the organisational structure of PHAs and the requirements for
accessing the managerial positions.
To enhance the utilisation function, more autonomy should be assigned to managers
combined with clear objectives and a well-defined set of incentives to pursue them. To
enlarge the audiences of PHAs, managers should have some degrees of freedom in
managing the composition and amount of inputs (since, currently, they cannot hire
skilled personnel according to the specific needs of each archive) as well as the range
of services to supply and their organisation. In addition, the potential of ICT and digital
technologies should be exploited as much as possible to overcome the trade-off 21
between the two activities. Fostering the various forms of digital communication might
enlarge audiences, favour crowdsourcing and other forms of users’ communication,
strengthening the links between archives and other cultural institutions to make clear
that archives are an important part of the cultural supply of a territory. Unless these
changes occur, increases in the amount of resources devoted to PHAs would hardly
guarantee an improvement of their performances.
Acknowledgements
We wish to thank the anonymous referees for their careful review, and the managing
editor, Professor Sabah Cavallo, for the helpful advice. The usual disclaimers apply.
References
Adams, S., & Klobodu, E. K. M. (2017). Urbanization, democracy, bureaucratic quality, and environmental degradation. Journal of Policy Modeling, 39(6), 1035-1051.
Ateca-Amestoy, V. (2018) Cultural Heritage Participation. Engagement models, evidence for the EU. Economia della Cultura, 28(4), 419-432
Banker, R.D., & Morey, R.C. (1986). Efficiency analysis for exogenously fixed inputs and outputs. Operation Research, 34 (4), 513-521.
Banker, R. D. (1996). Hypothesis tests using data envelopment analysis. Journal of productivity analysis, 7(2-3), 139-159.
Bishop, P., & Brand, S. (2003). The efficiency of museums: a stochastic frontier production function approach. Applied Economics, 35 (17), 1853-1858.
Borges Ferreira Neto, A., & Hall, J. (2018). Economies of Scale and Governance of Library Systems: Evidence from West Virginia, Economics of Governance, published on line.
Borowiecki, K. J., & Navarrete, T. (2017). Digitization of heritage collections as indicator of innovation. Economics of Innovation and New Technology, 26 (3), 227-246.
Castiglione, C., Infante, D., & Zieba, M. (2018). Technical efficiency in the Italian performing arts companies. Small Business Economics, 51(3), 609-638.
Charnes, A., Cooper, W.W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2 (6), 429–444.
22
Chen, Y., Morita, H., & Zhu, J. (2005). Context-dependent DEA with an application to Tokyo public libraries. International Journal of Information Technology & Decision Making, 4 (3), 385–394.
Chen, Y., & Zhu, J. (2004). Measuring information technology's indirect impact on firm performance. Information Technology and Management, 5 (1-2), 9-22.
Cook, W.D., & Zhu, J. (2014). DEA for two-stage networks: efficiency decompositions and modelling techniques. In W.D. Cook & J. Zhu (Eds.), Data Envelopment Analysis. A handbook on the modelling of internal structures and networks (pp. 1-29). International Series in Operations Research and Management Science, 208. Boston: Springer.
Del Barrio, M.J., Herrero, L.C., & Sanz, J.A. (2009). Measuring the efficiency of heritage institutions: a case study of a regional system of museums in Spain. Journal of Cultural Heritage, 10 (2), 258-268.
Del Barrio, M.J., & Herrero, L.C. (2014). Evaluating the efficiency of museums using multiple outputs: evidence from a regional system of museums in Spain. International Journal of Cultural Policy, 20 (2), 221-238.
Del Barrio, M.J., & Herrero, L.C. (2019). Modelling museum efficiency in producing inter-reliant outputs. Journal of Cultural Economics. https://doi.org/10.1007/s10824-019-09347-2
De Witte, K., & Geys, B. (2011). Evaluating efficient public good provision: theory and evidence from a generalised conditional efficiency model for public libraries. Journal of Urban Economics, 69 (3), 319-327.
De Witte, K., & Geys, B. (2013). Citizen coproduction and efficient public good provision: Theory and evidence from local public libraries. European Journal of Operational Research, 224 (3), 592–602.
Färe, R., & Grosskopf, S. (1996). Productivity and intermediate products: A frontier approach. Economics Letters, 50 (1), 65–70.
Fernández-Blanco, V., Herrero, L.C., & Prieto-Rodríguez, J. (2013). Performance of cultural heritage institutions. In I. Rizzo & A. Mignosa (Eds.), Handbook on Economics of Cultural Heritage (pp. 470-489). Cheltenham (UK) and Northampton, MA (USA): Edward Elgar Publishing Ltd.
Fernández-Blanco, V., Rodríguez-Álvarez, A., & Wiśniewska, A. (2018). Measuring technical efficiency and marginal costs in the performing arts: the case of the municipal theatres of Warsaw. Journal of Cultural Economics, 1-23.
Finocchiaro Castro, M., Guccio, C., & Rizzo, I. (2011). Public intervention on heritage conservation and determinants of heritage authorities performance: a semi-parametric analysis. International Tax and Public Finance, 18 (1), 1-16.
Finocchiaro Castro, M., & Rizzo, I. (2009). Performance measurement of heritage conservation activity in Sicily. International Journal of Arts Management, 11 (2), 29-41.
Guccio, C., Martorana, M., Mazza, I., & Rizzo, I. (2016). Technology and public access to cultural heritage: the Italian experience on IT for public historical archives, in K.J. Borowiecki, N. Forbe, A. Fresa (eds) Cultural Heritage in a Changing World, Springer, Heidelberg, pp. 55-76
23
Guccio, C., Mignosa, A., & Rizzo, I. (2018a). Are public state libraries efficient? An empirical assessment using network Data Envelopment Analysis, Socio-Economic Planning Sciences, forthcoming.
Guccio, C., Lisi, D., Mignosa, A., & Rizzo, I. (2018b). Does cultural heritage monetary value have an impact on visits? An assessment using Italian official data. Tourism Economics, 24(3), 297–318
Halkos, G.E., Tzeremes, N.G., & Kourtzidis S.A. (2014). A unified classification of two-stage DEA models. Surveys in Operations Research and Management Science, 19 (1) 1–16.
Hammond, C. J. (2002). Efficiency in the provision of public services: a data envelopment analysis of UK public library systems. Applied Economics, 34 (5), 649-657.
International Council on Archives (ICA) (2012). Principles of Access to Archives. Available on the ICA website, at the url: https://www.ica.org/sites/default/files/ICA_Access-principles_EN.pdf
Kao, C. (2014). Network data envelopment analysis: A review. European Journal of Operational Research, 239 (1), 1–16.
Kao, C., & Hwang, S.-N. (2008). Efficiency decomposition in two-stage data envelopment analysis: an application to non-life insurance companies in Taiwan. European Journal of Operational Research, 185 (1), 418-429.
Lagravinese, R., Liberati, P., & Resce, G. (2019). The impact of economic, social and cultural conditions on educational attainments. Journal of Policy Modeling, https://doi.org/10.1016/j.jpolmod.2019.03.007.
Last, A. K., & Wetzel, H. (2011). Baumol’s cost disease, efficiency, and productivity in the performing arts: an analysis of german public theaters. Journal of Cultural Economics, 35(3), 185.
Liang, L., Cook, W.D., & Zhu, J. (2008). DEA models for two-stage processes: game approach and efficiency decomposition. Naval Research Logistics, 55 (7), 643-653.
Mairesse, F., & Vanden Eeckaut, P. (2002). Museum assessment and FDH technology: towards a global approach. Journal of Cultural Economics, 26 (4), 261-286.
Pignataro, G. (2002). Measuring the efficiency of museums: a case study in Sicily. In I. Rizzo & R. Towse (Eds.), The economics of heritage. A study in the political economy of culture in Sicily (pp. 65-78). Cheltenham (UK) and Northampton, MA (USA): Edward Elgar Publishing Ltd.
Pignataro, G. (2011). Performance indicators. In R. Towse (Ed.), A handbook of cultural economics, second edition (pp. 332-338). Cheltenham (UK) and Northampton, MA (USA): Edward Elgar Publishing Ltd.
Pistoresi, B., Rinaldi, A., & Salsano, F. (2017). Government spending and its components in Italy, 1862–2009: Drivers and policy implications. Journal of Policy Modeling, 39(6), 1117-1140.
Ray, S. C. (1991). Resource-use efficiency in public schools: a study of Connecticut data. Management Science, 37 (12), 1620-1628.
Seiford, L.M., & Zhu, J. (1999). Profitability and marketability of the top 55 US commercial banks. Management Science, 45 (9), 1270–1288.
24
Simar, L., & Wilson, P.W. (2000). Statistical inference in nonparametric frontier models: the state of the art. Journal of Productivity Analysis, 13 (1), 49-78.
Simar, L., & Wilson P.W. (2008). Statistical inference in nonparametric frontier models: recent developments and perspectives. In H.O. Fried, C.A. Knox Lovell, & S.S. Schmidt (Eds.), The measurement of productive efficiency and productivity growth (pp. 421-521). New York: Oxford University Press.
Simon, J., Simon, C., & Arias, A. (2011). Changes in productivity of Spanish university libraries. Omega, 39 (5), 578–588.
Ventelou, B., & Bry, X. (2006). The role of public spending in economic growth: Envelopment methods. Journal of Policy Modeling, 28(4), 403-413.
Zan, L., Bonini Baraldi, S., & Santagati M.E. (2018). Missing HRM: the original sin of museums reform in Italy. Museum Management and Curatorship, 33 (6), 530-545.
Zieba, M. (2011). An analysis of technical efficiency and efficiency factors for Austrian and Swiss non-profit theatres. Swiss Journal of Economics and Statistics, 147 (2), 233-274.
25
TABLES AND FIGURES
Table 1 – The estimated models and descriptive statistics
Variable Mod_1 Mod_2 Obs. Mean St. Dev.
Inputs
SURFACE ♦ ♦ 198 5,380.89 6,177.63
SHELF ♦ ♦ 198 15,187.05 16,957.75
PERS ♦ ♦ 198 25.87 19.83
EXP ♦ ♦ 198 192,103.80 240,869.50
Size of collection: Nondiscretionary inputs (one-stage model) or Intermediate outputs (two-stage models)
M&D ♦ ♦ 198 128,458.10 135,121.90
ANTIQUARIES ♦ ♦ 198 22,095.72 43,525.97
O_COLLECTION ♦ ♦ 198 74,973.65 357,614.90
ASSET_VALUE ♦ 198 1,210,016.00 2,418,765.24
Final outputs – Utilisation
VISIT ♦ ♦ 198 2,775.66 3,361.18
RESEARCH ♦ ♦ 198 1,266.59 2,363.60
D_RESEARCH ♦ ♦ 198 8,803.84 8,958.19
Source: our elaboration on data provided by MiBAC Statistical office. a monetary values in 1,000 euros at current prices.
Figure 1 – Conditional kernel density functions by levels of volume or value of collections
01
23
Den
sity
0 .2 .4 .6 .8 1Efficiency estimates under CRS
High M&D M&D Low M&D
0.5
11.
52
2.5
Den
sity
0 .2 .4 .6 .8 1
Efficiency estimates under CRS
High ANTIQUARIES Intermediate ANTIQUARIES Low ANTIQUARIES
0.5
11.
52
2.5
Den
sity
0 .2 .4 .6 .8 1Efficiency estimates under CRS
High O_COLLECTION Intermediate O_COLLECTION Low O_COLLECTION
01
23
Den
sity
0 .2 .4 .6 .8 1Efficiency estimates under CRS
High ASSET_VALUE Intermediate ASSET_VALUE Low ASSET_VALUE
Note: kernel density functions of DEA CRS bias corrected efficiency estimated. Source: our elaboration on data provided by MiBAC Statistical office and RGS
26
Table 2 – One-stage models with nondiscretionary input – efficiency scores
SampleMod_1 Mod_2
Average St. dev. Average St. dev.
All sample 0.5691 0.2804 0.5746 0.2782
Year 2011 0.5854 0.2789 0.5856 0.2731
Year 2012 0.5528 0.2824 0.5635 0.2842
With branches 0.6405 0.2565 0.5791 0.2937
With School 0.4801 0.2639 0.5371 0.2512
North 0.5544 0.2953 0.5552 0.2912
Centre 0.5274 0.2566 0.5320 0.2781
South 0.6283 0.2918 0.6129 0.2653
Source: our elaboration on data provided by MiBACT Statistical office and RGS
Figure 2 – Conditional kernel density functions – one-stage with non-discretionary inputs
01
23
45
Den
sity
0 .2 .4 .6 .8 1Mod_1 - efficiency estimates under CRS
High M&D Intermediate M&D Low M&D
01
23
4D
ensi
ty
0 .2 .4 .6 .8 1Mod_1 - efficiency estimates under CRS
High ANTIQUARIES Intermediate ANTIQUARIES Low ANTIQUARIES
02
46
Den
sity
0 .2 .4 .6 .8 1Mod_1 - efficiency estimates under CRS
High O_COLLECTION Intermediate O_COLLECTION Low O_COLLECTION
0.5
11.
52
Den
sity
0 .2 .4 .6 .8 1Mod_2 - efficiency estimates under CRS
High ASSET_VALUE Intermediate ASSET_VALUE Low ASSET_VALUE
Note: CRS bias corrected efficiency estimated. Source: our elaboration on data provided by MiBACT Statistical office and RGS
27
Table 3 – Independent two-stage models
Mod_1 Mod_2
SampleStage 1 Stage 2 Global Stage 1 Stage 2 Global
Average St. dev. Average St. dev. Average Average St. dev. Average St. dev. Average
All sample 0.6435 0.2512 0.4245 0.2739 0.2550 0.6580 0.2560 0.4879 0.2854 0.3040
Year 2011 0.6346 0.2557 0.4398 0.2791 0.2570 0.6498 0.2601 0.5048 0.2913 0.3064
Year 2012 0.6524 0.2476 0.4092 0.2691 0.2529 0.6662 0.2528 0.4710 0.2798 0.3016
With branches 0.5690 0.2124 0.5172 0.2624 0.3028 0.5849 0.2188 0.5710 0.2682 0.3408
With School 0.7168 0.2128 0.3411 0.2736 0.2382 0.7510 0.2097 0.3601 0.2665 0.2682
North 0.7385 0.2187 0.3500 0.2167 0.2462 0.7561 0.2182 0.4150 0.2459 0.3043
Centre 0.6681 0.2575 0.4187 0.3008 0.2774 0.6838 0.2688 0.4616 0.3116 0.3159
South 0.5141 0.2302 0.5177 0.2927 0.2515 0.5242 0.2333 0.5919 0.2854 0.2962
Source: our elaboration on data provided by MiBAC Statistical office and RGS
Table 4 – Centralized NDEA models
Mod_1 Mod_2
SampleStage 1 Stage 2 Global Stage 1 Stage 2 Global
Average St. dev. Average St. dev. Average Average St. dev. Average St. dev. Average
All sample 0.6063 0.2366 0.3176 0.2060 0.1806 0.6153 0.2395 0.3202 0.2072 0.1848
Year 2011 0.5971 0.2423 0.3284 0.2069 0.1808 0.6062 0.2442 0.3310 0.2076 0.1848
Year 2012 0.6154 0.2316 0.3067 0.2056 0.1805 0.6244 0.2356 0.3093 0.2072 0.1848
With branches 0.5524 0.2084 0.3444 0.1499 0.1925 0.5611 0.2132 0.3488 0.1519 0.1982
With School 0.6578 0.1882 0.2363 0.1129 0.1431 0.6919 0.1864 0.2311 0.1088 0.1483
North 0.6806 0.2089 0.3043 0.2028 0.1949 0.6963 0.2049 0.3051 0.2013 0.2006
Centre 0.6446 0.2527 0.3176 0.2060 0.1792 0.6545 0.2579 0.3202 0.2072 0.1820
South 0.4930 0.2165 0.3284 0.2069 0.1645 0.4935 0.2186 0.3310 0.2076 0.1677
Source: our elaboration on data provided by MiBAC Statistical office and RGS
Table 5 – Pearson correlation coefficients among approaches and model specificationsApproach/Models (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(1)B&M_1986
Mod_1 1.0000
(2) Mod_2 0.6098* 1.0000
(3) S&Z_1999First stage
Mod_1 0.0950 0.3369* 1.0000
(4) Mod_2 0.0983 0.3168* 0.9928* 1.0000
(5) S&Z_1999Second stage
Mod_1 0.7821* 0.3915* -0.2659* -0.2627* 1.0000
(6) Mod_2 0.6802* 0.5289* -0.2248* -0.2349* 0.8884* 1.0000
(7) C_NDEAFirst stage
Mod_1 0.1603* 0.3278* 0.9565* 0.9487* -0.2311* -0.2046* 1.0000
(8) Mod_2 0.1446* 0.3092* 0.9596* 0.9610* -0.2366* -0.2164* 0.9888* 1.0000
(9) C_NDEASecond stage
Mod_1 0.5831* 0.5615* -0.2634* -0.2629* 0.7362* 0.6617* -0.2455* -0.2468* 1.0000
(10) Mod_2 0.5934* 0.5726* -0.2618* -0.2637* 0.7383* 0.6646* -0.2381* -0.2466* 0.9979* 1.0000
Note: B&M_1986 refer to Banker and Morey (1986) one-stage-production approach; S&Z_1999 refer to Seiford and Zhu (1999) independent two-stage approach; C_NDEA refer to Liang et al., (2008).
Source: our elaboration on data provided by MiBAC Statistical office.
28