public intervention on heritage conservation and determinants of heritage authorities’...

16
Int Tax Public Finance (2011) 18: 1–16 DOI 10.1007/s10797-010-9132-1 Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis Massimo Finocchiaro Castro · Calogero Guccio · Ilde Rizzo Published online: 9 March 2010 © Springer Science+Business Media, LLC 2010 Abstract Government plays a significant role in the field of heritage conservation using a mix of different tools such as public spending, tax-expenditures and regu- lation. Surprisingly, the literature on the economics of heritage has not extensively investigated the performance of the public actors involved in the implementation of conservation policies or its determinants. In this paper we address this issue, from a theoretical and an empirical perspective, using Sicily as a case study. More precisely, we analyse the determinants of the differences in the efficiency levels of conservation activity of the nine Sicilian heritage authorities over the period 1993–2005. Eco- nomic, political and managerial variables are used to distinguish nondiscretionary from discretionary causes. The results show that the efficiency scores seem to be af- fected by economic and political factors, whereas the managerial variables do not affect the performance of heritage authorities. Keywords Heritage regulation · Local cultural policy · Efficiency analysis · DEA JEL Classification D24 · D72 · C14 · Z10 Although this work is the result of common reflections, Massimo Finocchiaro Castro has written Sects. 3 and 4.1, Calogero Guccio has written Sects. 4.2 and 4.3, and Ilde Rizzo has written Sects. 1 and 2. M. Finocchiaro Castro ( ) Department of Law and Social Science, Mediterranean University of Reggio Calabria, via Tommaso Campanella 38/A, 85100 Reggio Calabria, Italy e-mail: massimo.fi[email protected] C. Guccio · I. Rizzo Department of Economics and Quantitative Methods, University of Catania, C.so Italia, 55, 95121 Catania, Italy C. Guccio e-mail: [email protected] I. Rizzo e-mail: [email protected]

Upload: ilde

Post on 14-Jul-2016

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Int Tax Public Finance (2011) 18: 1–16DOI 10.1007/s10797-010-9132-1

Public intervention on heritage conservation anddeterminants of heritage authorities’ performance:a semi-parametric analysis

Massimo Finocchiaro Castro · Calogero Guccio ·Ilde Rizzo

Published online: 9 March 2010© Springer Science+Business Media, LLC 2010

Abstract Government plays a significant role in the field of heritage conservationusing a mix of different tools such as public spending, tax-expenditures and regu-lation. Surprisingly, the literature on the economics of heritage has not extensivelyinvestigated the performance of the public actors involved in the implementation ofconservation policies or its determinants. In this paper we address this issue, from atheoretical and an empirical perspective, using Sicily as a case study. More precisely,we analyse the determinants of the differences in the efficiency levels of conservationactivity of the nine Sicilian heritage authorities over the period 1993–2005. Eco-nomic, political and managerial variables are used to distinguish nondiscretionaryfrom discretionary causes. The results show that the efficiency scores seem to be af-fected by economic and political factors, whereas the managerial variables do notaffect the performance of heritage authorities.

Keywords Heritage regulation · Local cultural policy · Efficiency analysis · DEA

JEL Classification D24 · D72 · C14 · Z10

Although this work is the result of common reflections, Massimo Finocchiaro Castro has writtenSects. 3 and 4.1, Calogero Guccio has written Sects. 4.2 and 4.3, and Ilde Rizzo has written Sects. 1and 2.

M. Finocchiaro Castro (�)Department of Law and Social Science, Mediterranean University of Reggio Calabria, via TommasoCampanella 38/A, 85100 Reggio Calabria, Italye-mail: [email protected]

C. Guccio · I. RizzoDepartment of Economics and Quantitative Methods, University of Catania, C.so Italia, 55, 95121Catania, Italy

C. Guccioe-mail: [email protected]

I. Rizzoe-mail: [email protected]

Page 2: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

2 M. Finocchiaro Castro et al.

1 Introduction

It is increasingly stressed in the economic and political debate that cultural heritagecan play a relevant role as “strategic” resource for sustainable local development.A necessary, though not sufficient condition, for such a role is that cultural heritageis properly conserved.

Almost everywhere the public sector plays an important role in the conservationof cultural heritage, even if with different quantitative and qualitative characteristics.In this field Government action can follow various patterns using a mix of differ-ent tools such as public spending, tax-expenditures and regulation. The efficiencyand effectiveness of heritage conservation policies, i.e. their capability to meet citi-zens’ demand and to score the expected results in terms of ‘public interest’, cruciallydepends on the features of the decision-making process and of the actors involved.A crucial role is played by heritage authorities, i.e. the bureaucrats responsible forimplementing conservation policies. Surprisingly, the literature on the economics ofheritage has not extensively investigated their performance, and in this paper we tryto fill this gap. For this purpose, using Sicily as a case study, we analyse the perfor-mance of Heritage Authorities (Soprintendenze). The size of Sicily1 as well as thequalitative and quantitative relevance of Sicilian heritage endowment2 is such thatthe analysis of the conservation activity appears relevant and suitable to offer wideand general suggestions. In fact, political decisions about heritage policy are taken atpolitical level by the Regional government and namely by the Head of the RegionalHeritage Office (Assessore), whereas their implementation is carried out by nine So-printendenze, which are responsible for any decision regarding heritage conservation.Their activity offers an interesting case study to understand the effects of conserva-tion policy, to analyse the role of government in the heritage field and to devise somepolicy implications.

This paper builds on Finocchiaro Castro and Rizzo (2009), where the authors pro-vide a definition of conservation activity considering the literature on the economicsof heritage and perform an efficiency analysis of the activity of Soprintendenze byapplying the Data Envelopment Analysis technique to the peculiar features of her-itage conservation over the period 1993–2000. We extend the results of FinocchiaroCastro and Rizzo (2009) with two major developments in terms of methodology andcontent. More precisely, first, we compute the efficiency of Soprintendenze applyinga more accurate and up-to-date methodology, i.e. the homogeneous bootstrap proce-dure (Simar and Wilson 1998); second, we develop the content with the evaluation ofthe determinants of the Soprintendenze’s performance using both economic and po-litical variables. As far as the second research aim is concerned, we evaluate the de-terminants of Soprintendenze performance using the two-stage analysis: inefficiencyestimates are used as a dependent variable in the second stage of the analysis to inves-tigate the influence of environmental variables on performance. We employ different

1Sicily is the fourth most populated region in Italy and has more inhabitants than some European countriessuch as Ireland, Norway and Croatia.2Sicily has the same number (in many cases more) of Unesco World Heritage listed properties of 32%of the countries reported by Unesco. Nation-wide, it has 9% of all Italian heritage and 30% of Italianarchaeological finds.

Page 3: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 3

techniques in the two-stage analysis to compare parametric and semi-parametric es-timation approaches.

Our results show that Sicilian Heritage Authorities achieve, on average, low ef-ficiency levels and that their performance seems to be affected by economic (suchas the size of the regulated territory, per-capita income, level of education, numberof buildings built before 1919) and political (the coincidence of the political con-stituency of the Assessore and the territory under the competence of Soprintendenza)variables, whereas the managerial variables such as seniority of Soprintendente donot affect the performance.

Tentative policy implications stemming from our analysis would suggest to intro-duce incentives toward efficiency in the decision-making process and to reshape theterritorial design of Soprintendenze to reduce their costs of production.

The analysis develops as follows: we discuss the main economic characteristics ofGovernment intervention for conservation in Sect. 2, and, then, in Sect. 3, we describethe institutional features of our case study, the Soprintendenze in Sicily. In Sect. 4,the methodological issues underlying the measurement of the efficiency of HeritageAuthorities are explored, technical efficiency of Sicilian Heritage Authorities is es-timated, and the empirical analysis of its determinants is presented. Section 5 offerssome concluding remarks.

2 Main features of government intervention for conservation

In all the industrialised countries the public sector plays an important role in theconservation of cultural heritage, even if with different quantitative and qualitativecharacteristics (Peacock and Rizzo 2008). The analysis of the normative rationale forGovernment intervention is outside the scope of this paper, and the related efficiencyand equity arguments are taken for granted;3 in what follows, the attention will beconcentrated on the features of public action and on its effects. In fact, though marketfailure provides a rationale for Government intervention, this is not to say that Gov-ernment action is efficient in providing conservation or that there is only one way tointervene (Mazza 2003).

In the heritage field, Government uses direct monetary tools—such as expenditure—as well as indirectly monetary tools—e.g. tax-expenditures. At the same time, amajor role is played by a nonmonetary instrument such as regulation (Rizzo andThrosby 2006). In what follows, attention will be concentrated on public spendingand regulation.4

Public expenditure can be used for many purposes: to buy tangible and intangiblegoods for the restoration activity (e.g. the salaries for experts and staff involved inheritage conservation, equipment for diagnosis, etc.) as well as buildings of artisticinterest or to provide subsidies and/or loans to cultural institutions (public, private ornonprofit) or to private owners of historic buildings.

3A general survey of the pros and cons of the normative justifications for government intervention in thecultural field is provided by Frey (2003).4Tax expenditures, e.g. tax allowances to incentive private financing are not taken into account becausethey are not relevant for our case study since they are outside the influence of Regional government.

Page 4: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

4 M. Finocchiaro Castro et al.

Regulation is aimed at restricting or modifying the activities of public and privateactors to control the stock of heritage, and penalties are provided for noncompliance.5

Regulation is exerted in many different ways: for instance, listing historical sites andindividual buildings, preventing the demolition of a building or a group of buildings;imposing restrictions on the uses to which the building can be put on its appearanceand the way restoration or reuse is carried out; imposing limitations on the use ofland affecting heritage buildings.

The regulator enjoys many degrees of freedom since the notion of heritage is notwell defined and conservation can be carried out in various forms. In some cases, con-servation aims not only at preserving heritage but also at enhancing it; in other cases,the last component is almost absent. Different forms of conservation impinge uponproperty rights may generate a distributional impact and produce different economicimplications (Rizzo 2003). For instance, restrictions on the use of buildings, their ap-pearance and the way in which restoration and reuse is carried out might underminethe possibility of restoring and revitalising historical centres, which is usually oneobjective on the political agenda of local authorities. If an undue “conservationist”stance is adopted well beyond what is justified by the costs-benefits comparison andheritage is simply preserved, its full enjoyment, in terms of opportunities for reuse,might be prevented, and, therefore, its potential economic benefits cannot be fullygenerated.

The above-mentioned problems are more likely to arise if the conservation deci-sions are mainly driven by the preferences and/or the expertise of the Heritage Au-thorities when they have no incentives to take into account society’s preferences andthe opportunity costs of conservation decisions. Sicily offers a good example of theoccurrence of the above-mentioned problems as it will be outlined in the followingsection.

3 Institutional features and policy tools in Sicilian heritage conservation

Sicily is an Italian region which enjoys full autonomy in the field of heritage pol-icy. Political decisions about heritage policy are taken by the Regional Government,and their implementation is carried out by nine Heritage Soprintendenze, which areresponsible for any decision regarding heritage conservation in the territory of theircompetence.6 Their activity offers an interesting case study for analysing the role ofGovernment in the heritage field.

Elsewhere (Rizzo 2002) it has been pointed out that Soprintendenze are run byexperts (e.g. archaeologists, art historians, architects, etc.) and enjoy high discre-tion because the choice of instruments and their intensity largely depend on theirautonomous technical evaluation. Their activity has been distinguished into passiveconservation (PC) and active conservation (AC).

5Beside these forms of regulation, which Throsby (1997) defines as « hard regulations », there are alsononenforceable forms of regulation, i.e. « soft regulations », mainly applied at international level, whichare implemented by agreement and do not involve penalties.6The territory under the competence of Soprintendenza coincides with the Provincial level.

Page 5: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 5

The PC consists of regulatory activities and implies many different administrativeacts, enforceable on both private and public owners (such as constraints, demolitionorders, authorisations, permission to import and export) which are subject to judicialreview if those affected dispute the decision in court. In some cases (e.g. authorisa-tions), the above regulatory activity is in response to the owner’s demand; in othercases, it implies measures which are spontaneously adopted to constrain owner’s ac-tivity (e.g. landscape constraints) or to punish for violations (e.g. demolition orders).

The AC refers to spending activity and consists of a wide array of activities suchas taking an inventory, performing scientific research, training staff, updating, exca-vating and restoring. These activities can be constrained by the availability of funds.

The degree of autonomy enjoyed by Soprintendenze is very high at the planninglevel when technical decisions are involved, while it is usually low at the operationallevel7 even as far as the management of personnel is concerned.

The distinction between PC and AC might be questioned, since these activitiesmay be interconnected in some cases;8 while recognising the significance of theselinks, using such a distinction is useful because it recalls the conventional distinctionbetween spending and regulation and, thus, helps to understand the complexity ofconservation activities from an economic point of view. Moreover, the distinctionallows for empirical investigation by introducing the possibility of devising indicatorsfor each activity.

Having briefly described the content of conservation activities, we turn to analysehow such a mix is established, taking into account the constraints and incentivesexperienced by Soprintendenze.

Rizzo (2002) points out that AC measures are partially constrained by the avail-ability and timing of financial resources assigned by the regional Government to eachSoprintendenza (these constraints being less severe for PC activities because, as toolsof regulatory activity, their direct spending is less crucial). At the same time, ACprovides more incentives than PC in terms of reputation and prestige among spe-cialists since the restored building or the archaeological excavation is a testimony tothe expertise of the Soprintendenza’s experts and provides opportunities for scientificresearch and publications.

Different incentives are at work for the PC activity. In fact, the performance ofeach Soprintendenza is likely to be monitored by public and private owners of her-itage because of the divisible benefits and costs generated by PC activities; delays orpoor performance may lead to some form of protest and may even generate a judicialaction. The existence of such a form of accountability may thus provide an incen-tive to concentrate attention on PC activities, allocating available resources (such aspersonnel) to those activities subject to stronger external control.

Finally, PC and AC activities might be affected by the incentives generated by thepolitical sponsor, i.e. the Regional government and, namely, the Head of the Regional

7Once the yearly activity program submitted by each Soprintendenza is approved at the regional level, nodiscretionary variation is allowed.8For instance, research and study activities underlying both can be considered interdependent; a discov-ery resulting from an archaeological excavation might call for imposing constraints; at the same time,expropriation is prerequisite to direct intervention.

Page 6: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

6 M. Finocchiaro Castro et al.

Heritage Office (Assessore). Indeed, there is no evidence of any formal incentivescheme (in terms of the size of budget or of private benefits for bureaucrats, such ascareer and salary) to stimulate the performance of Soprintendenze.9 However, this isnot to say that the political sponsor is not interested in heritage conservation; on thecontrary, there is a political interest, with regard to electoral consensus, since heritagematters for citizens. In other words, a “political” demand for the performance ofSoprintendenze is likely to exist especially if the Soprintendenza operates closer tothe electoral constituency of the Assessore.

4 Empirical analysis

4.1 Methodological issues in efficiency measure

To assess the performance of the Heritage Authorities, we employ the nonparamet-ric technique known as Data Envelopment Analysis (DEA). DEA is a linear pro-gramming formulation used to measure the distance from the production frontier inmultiple input and multiple output production environments.10 As a nonparametricapproach, DEA is able to estimate the production function with minimal prior as-sumptions and derive technical and scale efficiency. In the last ten years, DEA hasbeen applied to measure the efficiency of art and culture organisations, showing agreat degree of flexibility. Luksetich and Nold Hughes (1997) investigate, by meansof DEA and regression analysis, the efficiency and its determinants of funding activ-ities of a sample of symphonic orchestras in the United States. The efficiency of re-ligious organisations has been studied by Zaleski and Zech (1997). They apply DEAmethodology to the U.S. Catholic Church to examine the relative shortage of priests.Finally, two contributions focus on the efficiency analysis of museums. Pignataro andZanola (2001) analyse the efficiency levels of museums located in two very differ-ent Italian regions (Sicily and Piedmont), whereas Basso and Funari (2004) focus onmany public Italian museums computing DEA efficiency levels and decomposing theefficiency scores into pure technical and scale components. Finally, Del Barrio et al.(2009) investigate the efficiency of Spanish museums management.

In general terms, a DEA input-oriented efficiency score θi is calculated for eachDecision Making Unit (DMU) solving the following program for i = 1, . . . , n (con-

9As measured by, for example, the length of bureaucratic procedures and complaints from the public or bythe efficiency and effectiveness of the spending activity.10Different techniques have been developed to either calculate or estimate efficiency frontier. The nonpara-metric and the parametric approaches are the two most important methodologies used in this respect. Theparametric approach, also known by the ‘Stochastic Frontier Approach’, involves specifying the functionalform of the frontier and then estimating its unknown parameters using econometric techniques (Battese etal. 2005).

Page 7: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 7

stant return to scale case, CRS):

Minλ,θi

θi

subject to Yλ − yi ≥ 0

θixi − Xλ ≥ 0

λ ≥ 0

(1)

where xi and yi are respectively the input and output of ith DMU; X is the matrix ofinput, and Y is the matrix of output of the sample λ is an n×1 vector of variables. Themodel (1) can be modified to account for variable return to scale (VRS) by addingthe convexity constraint: eλ = 1, where e is a row vector with all elements unity.

In this paper, we use DEA method to estimate input-based technical and scale ef-ficiency. The main focus of the study is on the analysis of input-based technical effi-ciency under VRS. However, we also report the technical efficiency scores computedunder CRS. Simar and Wilson (2000a) clarify that traditional DEA methods yieldbiased estimates of efficiency. Based on homogeneous bootstrap procedure for DEAestimators, proposed by Simar and Wilson (1998), the paper estimates the bias andthe confidence intervals of the input-based technical efficiency with VRS.11 WhereasDEA methods have been widely applied, most researchers have largely turned a blindeye to the statistical properties of the estimators. Ignoring the statistical noise in theestimation can lead to biased DEA estimates and misleading results because all thedeviations from the frontier are considered inefficient. Simar and Wilson (2000a) ar-gue that bootstrap is the most currently feasible method to establish the statisticalproperty of DEA estimators. Thus we apply homogeneous bootstrap procedure tocorrect the bias in DEA estimators and to construct their confidence intervals.

The model (1) incorporates only discretionary inputs and does not take into ac-count the presence of environmental variables or factors, also known as nondiscre-tionary inputs, on performance. To take into account these variables, the most ap-propriate approach is the so-called two-stage analysis. This technique uses the inef-ficiency estimates as a dependent variable in the second stage of the analysis to in-vestigate the influence of environmental variables on performance. Recent literatureon two-stage estimation approach shows that the estimates are biased because of se-rial correlation of efficiency scores and suggests applying semi-parametric two-stagetechnique to perform an estimation on nondiscretionary inputs (Simar and Wilson2007). In the following analysis, we employ different techniques in the two-stageanalysis of the determinants of Soprintendenze’s performance in order to compareparametric and semi-parametric estimation approaches.

4.2 DEA efficiency estimate

The application of DEA to estimate the productivity of a DMU calls for some stepsto follow. Firstly, we define the production set in order to specify all the feasible

11Estimates have been obtained using the package FEAR 1.1, developed by Wilson (2007).

Page 8: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

8 M. Finocchiaro Castro et al.

Table 1 Descriptive statistics on input and output (cross-sectional–time-series distribution)

Variable Mean Std. Dev. Min Max Obs.

Input PERSONNEL Overall 236.34 109.36 62.00 510.00 N = 117

(n◦ of staff) Between 107.79 79.92 437.15 n = 9

Within 39.28 104.19 316.50 T = 13

Output AC Overall 6567.25 3442.81 1407.81 19685.29 N = 117

(thousands of euro) Between 2990.74 2623.13 12566.51 n = 9

Within 1958.00 583.01 13686.03 T = 13

PC Overall 317.09 455.58 11.43 3863.40 N = 117

(n◦ of administrative Between 186.24 103.32 636.78 n = 9

acts) Within 420.06 296.29 3543.71 T = 13

Source: our computation on data of the Official Registry of Regione Siciliana

combinations of input and output. Consequently, we make some assumptions on theproduction set.12 Following the contribution of Finocchiaro Castro and Rizzo (2009),we study a production function of Soprintendenze given by 1 input, personnel, and 2outputs, expenditure (AC) and weighted administrative actions (PC).13 The PC datarefer to the number of administrative actions, produced by each Soprintendenza aslisted in the Official Regional Registry, weighted to take into account the differencesin the technical and the administrative difficulty faced in implementing each type ofthe actions listed.14 The AC data refer to the expenditures (i.e. payments) of Soprint-endenze (at 2000 fixed price).

Data on the nine Sicilian Soprintendenze come from official Regional sources andrefer to the period 1993–2005. Thus, our sample is a balanced panel data with 117observations. Table 1 shows the descriptive statistics of variables employed.

12The standard assumptions for estimation of inefficiency with DEA are: (a) the production set is con-vex and closed; (b) the production requires the use of some inputs; and (c) both inputs and outputs arestrongly disposable (Färe et al. 1985). Moreover, in order to allow statistical analysis and inference onefficiency score, the following assumptions are needed: (d) the observed set of inputs and outputs resultsfrom independent draws from a probability density function with bounded support over the production set;(e) the density function is strictly positive for all points along the frontier; and (f) at any point along thefrontier, the density is continuous in any direction toward the interior of the production set. For a survey ofestimation and inference methods for distance function estimation, see Simar and Wilson (2000b).13Kneip et al. (1998) show that the rate of convergence of Farrell’s estimate efficiency score depends onthe number of input and output. In particular, the choice of a simple estimation model makes it possible toderive more consistent estimates of efficiency scores.14Weights (ranging from 1 to 5) have been assigned on the grounds of a questionnaire submitted to someexperts employed by both the Soprintendenze and the Assessorato to take into account the differencesin the technical and the administrative difficulty to implement the actions listed. For further details, seeFinocchiaro Castro and Rizzo (2009).

Page 9: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 9

Table 2 Efficiency estimate—mean value for each Soprintendenza

Soprintendenze Pure technical efficiency (VRS—input oriented) Total technical

2 3 4 5 6 efficiency (CRS—

Eff. score— Eff. bias corr— Bias—mean Lower bound— Upper bound— input oriented)

mean value mean value value mean value mean value 7

Eff. score—

mean value

DMU_1 0.378 0.348 0.029 0.320 0.371 0.362

DMU_2 0.717 0.669 0.048 0.626 0.706 0.703

DMU_3 0.467 0.421 0.046 0.383 0.458 0.444

DMU_4 0.495 0.449 0.046 0.405 0.488 0.345

DMU_5 0.492 0.452 0.040 0.413 0.484 0.466

DMU_6 0.568 0.466 0.102 0.395 0.555 0.502

DMU_7 0.807 0.732 0.075 0.667 0.796 0.639

DMU_8 0.602 0.527 0.075 0.474 0.589 0.590

DMU_9 0.445 0.414 0.031 0.384 0.439 0.436

All sample 0.552 0.498 0.055 0.452 0.543 0.499

Mean scale efficiency 0.903

Mean scale inefficiency 0.097

Source: our computation on data of the Official Registry of Regione Siciliana

The use of panel data in DEA model is widely discussed in the literature.15 Amongthe several possible ways to deal with panel data in efficiency DEA models, we havechosen to treat the panel as a single cross-section and pool the data, given the slowconvergence rates of DEA estimator. This choice is also based on the hypothesis thatthe conservation activity in Sicily is not affected by relevant technological changes.Consequently, efficiency estimates reflect relative DMU’s performance to the invari-ant technology.16 Finally, we apply the inspection technique developed by Wilson(1993) to control for possible effect of outliers on efficiency estimate. The resultsshow no evidence of any outlier’s effects on the estimate.

Table 2 reports the estimates of the mean efficiency scores, measured with Farrel(1957) efficiency definition, for each DMU. Following Simar and Wilson (1998), we

15The traditional approach to the analysis of efficiency DEA models with panel data is the so-called“window analysis”, converting a panel into an overlapping sequence of windows, which are then treatedas separate cross-sections. However, the choice of windows can lead to bias in estimates. An alternativeway is to employ the Malmquist index of productivity change. Another possibility is to treat the panelas a single cross-section and pool the observations. In this case, each observation being considered as anindependent one, a single frontier is computed and the relative efficiency of each DMU in each periodis calculated. Estache et al. (2004) observe that pooling data is a special case of “window analysis” withthe advantage of being a nondiscretionary choice of windows. We follow the latter approach in order toincrease the estimation power of the model.16We used the Malmquist index to check for increases in the productivity of DMUs. Our data show thatproductivity has not significantly changed during the observation period.

Page 10: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

10 M. Finocchiaro Castro et al.

implement the homogeneous bootstrap procedure to correct the bias in DEA estima-tors and obtain their confidence intervals.17

Column 2 provides the mean values of DEA efficiency scores, columns 3 and 4provide the bias-corrected efficiency scores and the bootstrap bias estimates, respec-tively. Columns 5 and 6 provide the two boundaries of 95% confidence intervals forthe bias-corrected efficiency scores. Finally, column 7 reports efficiency scores underCRS. Table 2 shows a poor efficiency level for the whole sample. The bias-correctedefficiency estimates range from 0.348 to 0.732, with an average value of 0.498. Thus,our results indicate that, on average, each Soprintendenza can reduce its input pro-portionally by 50.2 percent without reducing output if no technological change hastaken place in the period of analysis. Table 2 shows also that the portion of scale inef-ficiency (i.e. the penalty suffered when assuming CRS instead of VRS) is quite small(0.097).

The choice between CRS and VRS depends crucially on various factors relatedto the context and scope of the analysis. To check for CRS, we estimate the correla-tion between CRS efficiency scores and Soprintendenze size measured with variablesrelated to the operating scale such as personnel (correlation 0.294) and populationserved (correlation 0.074). The low correlation values justify the use of the VRS.18

Figure 1 plots the mean and variance of the bias-corrected efficiency estimates foreach year of observation.19 The mean is measured on the left vertical axis, while thevariance is measured on the right vertical axis. Data show a low variability in themean of the bias-corrected efficiency scores over the sample period and, overall, aquite low performance at year level. The highest efficiency scores are 2001 (0.600),2003 (0.579) and 2004 (0.576). It has to be noted that there is no evidence of a timetrend in the average efficiency levels. The patterns of efficiency levels turned out tobe clearly stagnant.

4.3 The determinants of performance in heritage conservation

In this section to investigate the determinants of the performance of Soprintendenzeconservation activity, we consider economic, managerial and political variables.20

The estimated models can be expressed by the following general formulation:

θi = f (zi) + εi, (2)

where θi is the efficiency score that resulted from previous stage, zi is a set of possiblenondiscretionary inputs, and εi is a vector of error terms.

17The confidence intervals and the bias-corrected efficiency scores have been estimated using the homo-geneous bootstrap procedure with 2,000 bootstrap draws as described by Simar and Wilson (1998). Wealso assume the independence between technical inefficiency and output levels as well as the mix of inputsthat are produced.18In addition, we calculated the correlation between CRS and VRS that turned out to be very high (0.918).19Solid line shows mean efficiency measured on the left vertical axis, and dashed line shows variance ofestimated efficiency measured on the right vertical axis.20We do not include the stock of heritage as explanatory variable because the official available data on theamounts of heritage under the supervision of each Soprintendenza are rather obsolete and partial.

Page 11: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 11

Fig. 1 Mean and variance of efficiency estimate across Soprintendenze by year. Source: our computationon data of the Official Registry of Regione Siciliana

As far as the economic variables are concerned, supply and demand variables areused. Looking at the supply-side and considering that Soprintendenze’s outputs areonly partially affected by environmental factors and by the geographical features ofthe area under control, we believe that the only variable affecting the cost of pro-duction is the size of each province. Thus, being heritage scattered in the Provincialterritory, the size of the area, expressed in squared Kms (SIZE), ceteris paribus, islikely to affect negatively the cost of producing both AC and PC activities.

Looking at the demand, it would be useful to use per capita cultural spending as aproxy for the demand for cultural activities. This variable would give also a measureof cultural environment, and, therefore, it might be able to represent the interest ofthe local community for heritage conservation. Reliable data of per capita culturalspending at provincial level for the entire period of observation are not available.For this reason, income per capita is used as a proxy for the demand of conservation(INCOME) on the assumption that higher income per capita implies higher levels ofeconomic activity and, ceteris paribus, a more dynamic private construction sector.As a consequence, increases in building and restoration activities call for higher de-mand of services supplied by Soprintendenze. However, income per capita is also aproxy for the socio-economic status of population such as, for instance, education,which, in turn, positively affects the demand for heritage. To control for the effect ofeducation, we add the variable EDU obtained computing the number of graduates onthousand of inhabitants.

Furthermore, it has to be noted that the areas under the competence of Soprinten-denze in which the concentration of historical buildings is higher represent one of themost relevant source of demand of services provided by Soprintendenza. Thus, weuse the registered estimate of the number of properties built before 1919 sited in the

Page 12: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

12 M. Finocchiaro Castro et al.

Table 3 Variables employed

Dependent variable

EFFICIENCY Efficiency scores (VRS)

Explanatory variables

SIZE Size of the area of each Province i = 1, . . . ,9 (in millions of squared Kms)

EDU Number of graduates on 1,000 inhabitants in each Province i = 1, . . . ,9 in each yearj = 1, . . . ,13

INCOME Per capita income in each Province i = 1, . . . ,9 (in thousands)

OLD_B Number of properties built before 1919 in Province i = 1, . . . ,9 (in thousands)

SENIORITY Length of appointment of each Soprintendente measured in months

AS Dummy for the constituency of the Head of the Regional Heritage Office

Source: our computation of data of the Official Registry of Regione Siciliana

area of each Soprintendenza’s competence (OLD_B) to capture the above-mentionedeffect.

As well as the environmental factors, also the characteristics of the Head of theProvincial Heritage Office (Soprintendente) are likely to affect the efficiency of So-printendenze. In particular, we include into the analysis a measure of the length of theappointment of Soprintendenti in monthly terms21 (SENIORITY), to capture such amanagerial effect.

The interpretation of the impact of this variable is not straightforward. On the onehand, a long tenure implies more experience and, thus, a positive effect on efficiency;on the other hand, from a public choice perspective, longer tenure would imply morepowerful bureaucrats, who would have a greater bargaining power to extract resourcesfrom the political decision-maker and would be less accountable. In this case, wemay observe a negative impact on efficiency. In other studies on the determinants oftechnical efficiency, another managerial variable is “level of managers’ education”;in our case, however, such a variable is not suitable since Soprintendenti are civilservants and the education requirements are established by law. They differ in termsof expertise—e.g. architect, archaeologists, art historians—but there are no a priorireasons to believe that such a difference affects technical efficiency.

Finally, we evaluate the effect of political factors. Soprintendenti are appointedby the Head of Regional Heritage Office, who is the political sponsor of their activ-ity; in a public choice perspective, it is likely that the Soprintendente accountabilityis greater if the territory under his/her competence coincides with the electoral con-stituency of the Head of Regional Heritage Office. To investigate the relevance of this“political” demand, following Guccio and Mazza (2009) analysis,22 we use as vari-able AS, a dummy for the constituency of the Head of the Regional Heritage Office,

21To take into account for possible endogeneity, we introduce a lagged variable.22Guccio and Mazza (2009) show that the allocation of funds to Soprintendenze is affected by politicalfactors.

Page 13: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 13

Table 4 Descriptive statistics

Variable Mean St. Dev. Minimum Maximum

EFFICIENCY 0.55 0.21 0.23 1

SIZE 2.86 0.95 1.61 4.99

EDU 53.86 13.79 28.41 90.25

INCOME 11.68 2.47 7.08 19.69

OLD_B 23.37 14.14 9.88 52.33

SENIORITY 80.54 86.76 12.00 324.00

AS 0.11 0.32 0 1

Source: our computation of data of the Official Registry of Regione Siciliana

with value 1 when the constituency coincides with the territory under the Soprinten-denza’s competence. Table 3 describes the variables employed in the analysis, andTable 4 shows the descriptive statistics.

In two-stage approach, researchers usually adopt censored regression techniques(Tobit) or, in a few cases, OLS estimates to take into account the censored natureof dependent variable. The most recent literature shows that the estimates are bi-ased because of serial correlation of efficiency scores and suggests applying semi-parametric two-stage technique to estimate efficiency scores using nondiscretionaryinputs (Simar and Wilson 2007).

Thus, we run an OLS, Tobit and truncated regression with double bootstrap es-timation on efficiency scores.23 Table 5 reports the estimates. The estimated coeffi-cients show quite similar values confirming the robustness of our empirical analysis.However, it has to be noted that our investigation does not aim at providing an accu-rate estimate of the marginal effects on performance of nondiscretionary inputs butonly average effects. Looking at the supply side, we notice that the variable SIZE issignificant, showing that the DMU’s performance is affected by specific dimension ofthe area in which the DMU operates. The negative sign seems to indicate that, beingheritage scattered in the Provincial territory, the dimension of the area under control,ceteris paribus, affects negatively the cost of producing both AC and PC activities.Moving to the demand-side of the production process, our results show that the sizeof the demand positively affects the efficiency scores. The variables INCOME andOLD_B are always significant with positive sign. This confirms that the demand ex-erts a positive effect on efficiency because of the stimulus of the heritage owners onDMU’s performance (Finocchiaro Castro and Rizzo 2009). In contrast, the variableEDU does not significantly affect the productivity level of Soprintendenze.

On the one hand, the managerial variable SENIORITY is not significant in anyestimated model. A possible explanation lies on the above-mentioned limited op-erational autonomy of Soprintendenti as far as the personnel and management offinancial resources are concerned. On the other hand, the political variable AS is sig-nificant with the expected positive sign showing that the performance is affected also

23Borge and Haraldsvik (2009) have obtained similar results by investigating the efficiency of care for theelderly sector in Norway.

Page 14: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

14 M. Finocchiaro Castro et al.

Table 5 Determinants of Heritage Authorities’ performance

Independent variable: Efficiency scores (VRS input oriented)

Estimation range: 1993–2005

Variable (1) (2) (3)

Truncated regressionOLS Tobit

with double bootstrap

Efficiency scores (VRS) Efficiency scores (VRS) Efficiency scores (VRS)

0.5558∗∗∗ 0.5378∗∗∗ 0.6178∗∗∗Constant

(0.1420) (0.1468) (0.1395)

−0.1560∗∗ −0.1589∗∗ −0.1612∗∗∗SIZE

(0.0609) (0.0627) (0.0584)

−0.0086∗∗∗ −0.0094∗∗∗ −0.0069∗∗EDU

(0.0032) (0.0033) (0.0031)

0.0575∗∗∗ 0.0632∗∗∗ 0.0457∗∗∗INCOME

(0.0164) (0.0171) (0.0172)

0.0106∗∗ 0.0109∗∗ 0.0101∗∗∗OLD_B

(0.0041) (0.0042) (0.0039)

−0.0003 −0.0004 −0.0003SENIORITY

(0.0002) (0.0002) (0.0002)

0.1125∗∗ 0.1253∗∗ 0.1142∗∗AS

(0.0559) (0.0578) (0.0550)

R-squared 0.3638 – –

F-test 6.97∗∗∗ – –

Log likelihood – 46.22 –

Notes: standard errors are reported in parentheses. ∗∗∗ , ∗∗ and ∗ denote significance at 1, 5 and 10 per centlevels, respectively

by the stimulus exerted by the political pressure. Hence, the efficiency scores of theSicilian Soprintendenze seem to be affected by the demand—both exerted by citizensand by the political sponsor—and by the supply variables.

5 Concluding remarks

This paper contributes to the current literature on the empirical analysis in the fieldof heritage conservation in several directions. First, we aim at filling the existinggap in the literature on Heritage Authority performance through the analysis of theefficiency of conservation activity. For this purpose, we apply smoothed bootstrapprocedure to compute the efficiency of decision-making units to correct the bias inDEA estimators and establish their confidence interval. This new procedure provides

Page 15: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

Public intervention on heritage conservation and determinants 15

information about the effects of statistical noise on DEA estimates, often ignored bymost of the researchers in the field of efficiency analysis.

Second, we investigate the determinants of efficiency scores using both economicand managerial variables to understand what variables exert the stronger effect onconservation activity and should be taken into account by heritage policy-makers. Tofurther validate the study of efficiency determinants, we obtained robust results usingdifferent techniques in the two-stage analysis to allow comparisons between para-metric and semi-parametric estimation approaches. The empirical results show thatSicilian Heritage Authorities achieve, on average, low efficiency levels that seem tobe affected by economic and political variables only, whereas the managerial vari-ables do not affect the performance.

Tentative policy implications stemming from our analysis stress the positive roleon efficiency exerted by incentives. Given the institutional features of the Sicilian her-itage organisational structure, the only stimulus depends mainly on demand, whereasalmost no incentives are built in the decision-making process. Thus, a greater op-erational autonomy of Soprintendenti, combined with a systematic assessment oftheir performance, might introduce positive incentives toward efficiency. Finally, theanalysis shows some room for reshaping the territorial design of Soprintendenze sincethe coincidence with the provincial area seems not justified by any sound economicreason and bears negative effects on the costs of production.

However, our findings can be applied to any system of heritage administrationcharacterised by hierarchical multijurisdictions. More generally, such complex mixof heritage authorities’ activity appears to be sensitive to the presence of externalincentives especially those depending on the demand-side. For instance, conserva-tion activity could be made more demand oriented, for example, by improving thedistribution of information to increase public participation that may become a usefulcomplement to expert judgement. Another way to improve public scrutiny of publicdecisions that could be applied to different heritage administrations is given by theadoption of a systematic assessment of the economic impact of regulation, such asCode of Practice or Guidelines agreed between the regulator and those involved inconservation activities. Concluding, the fact that the major stimulus toward efficientperformance depends on demand shows that the empirical results do not rely uponthe specific features of the case study but are likely to be applicable to developedcountries endowed with heritage. Further hints, therefore, can be derived extendingthe research on a comparative perspective.

Acknowledgements We thank two anonymous referees for suggestions regarding the improvement ofthe paper and participants at the 3rd European Workshop on Applied Cultural Economics 2007, the 14thACEI Conference 2008 and the 65th IIPF Congress 2009 for helpful comments and criticisms.

References

Basso, A., & Funari, S. (2004). A quantitative approach to evaluate the relative efficiency of museums.Journal of Cultural Economics, 28, 195–216.

Battese, G. E., Coelli, T. J., & Rao, D. S. (2005). An introduction to efficiency and productivity analysis(2nd ed.). New York: Springer.

Page 16: Public intervention on heritage conservation and determinants of heritage authorities’ performance: a semi-parametric analysis

16 M. Finocchiaro Castro et al.

Borge, L.-E., & Haraldsvik, M. (2009). Efficiency potential and determinants of efficiency: ananalysis of the care for the elderly sector in Norway. International Tax and Public Finance.doi:10.1007/s10797-009-9110-7.

Del Barrio, M. J., Herrero, L. C., & Sanz, J. A. (2009). Measuring the efficiency of heritage Institutions: acase study of a regional system of museums in Spain. Journal of Cultural Heritage, 10, 258–268.

Estache, A. R., Martin, A., & Ruzzier, C. A. (2004). The case for international coordination of electricityregulation: evidence from the measurement of efficiency in South America. Journal of RegulatoryEconomics, 25(3), 271–295.

Färe, R., Grosskopf, S., & Lovell, C. A. K. (1985). The measurement of efficiency of production. Hingham:Kluwer Academic.

Farrel, M. J. (1957). The measurement of productive efficiency. Journal of the Royal Statistical Society,Series A, 120, 253–281.

Finocchiaro Castro, M., & Rizzo, I. (2009). Performance measurement of heritage conservation activity inSicily. International Journal of Arts Management, 11(2), 29–41.

Frey, B. (2003). Public support. In R. Towse (Ed.), A handbook of cultural economics (pp. 389–398).Cheltenham: Edward Elgar.

Guccio, C., & Mazza, I. (2009). Pork barrel and public expenditure for cultural heritage: evidence fromregional allocation in Sicily. Mimeo.

Kneip, A., Park, B. U., & Simar, L. (1998). A note on the convergence of nonparametric DEA estimatesfor production efficiency scores. Econometric Theory, 14, 783–793.

Luksetich, W., & Nold Hughes, P. (1997). Efficiency of fund-raising activities: an application of dataenvelopment analysis. Nonprofit and Voluntary Sector Quarterly, 26, 73–84.

Mazza, I. (2003). Public choice. In R. Towse (Ed.), A handbook of cultural economics (pp. 379–388).Cheltenham: Edward Elgar.

Peacock, A. T., & Rizzo, I. (2008). The heritage game. economics, policy and practice. Oxford: OxfordUniversity Press.

Pignataro, G., & Zanola, R. (2001). Analisi dell’efficienza dei musei. In P. A. Valentino, & G. Mossetto(Eds.), Museo Contro Museo: le Strategie, Gli Strumenti, i Risultati (pp. 139–152). Firenze: Giunti.

Rizzo, I. (2002). Heritage conservation: the role of heritage authorities. In I. Rizzo, & R. Towse (Eds.),The economics of the heritage: a study in the political economy of culture in sicily (pp. 31–47).Cheltenham: Edward Elgar.

Rizzo, I. (2003). Regulation. In R. Towse (Ed.), A handbook of cultural economics (pp. 408–414). Chel-tenham: Edward Elgar.

Rizzo, I., & Throsby, D. (2006). Cultural heritage: economic analysis and public policy. In V. Ginsburgh,& D. Throsby (Eds.), Handbook of the economics of the arts and culture (pp. 983–1016). Amsterdam:North-Holland/Elsevier.

Simar, L., & Wilson, P. W. (1998). Sensitivity analysis of efficiency scores: how to bootstrap in nonpara-metric frontier models. Management Science, 44, 49–61.

Simar, L., & Wilson, P. W. (2000a). A general methodology for bootstrapping in non-parametric frontiermodels. Journal of Applied Statistics, 27, 779–802.

Simar, L., & Wilson, P. W. (2000b). Statistics inference in nonparametric frontier models: the state of theart. Journal of Productivity Analysis, 13, 49–78.

Simar, L., & Wilson, P. W. (2007). Estimation and inference in two-stage semi-parametric models ofproductive efficiency. Journal of Econometrics, 136, 31–64.

Throsby, D. (1997). Seven questions in the economics of cultural heritage. In Hutter, M., & Rizzo, I. (Eds.)Economic perspectives of cultural heritage (pp. 13–30). Basingstoke: Macmillan.

Wilson, P. W. (1993). Detecting outliers in deterministic nonparametric frontier models with multipleoutputs. Journal of Business and Economic Statistics, 11, 319–323.

Wilson, P. W. (2007). FEAR: a software package for frontier efficiency analysis with R. Socio-EconomicPlanning Sciences, 42, 247–254.

Zaleski, P. A., & Zech, C. E. (1997). Efficiency in religious organization. Nonprofit Management andLeadership, 8, 3–18.