https://orcid.org/0000-0002-1053-4968
![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
This paper investigates the influence of information and communication technologies (ICT) on the efficiency in attracting visitors of Italian museums. Notwithstanding the extensive literature on museum performance measurement, the analysis of the role of technological innovation is relatively neglected. As a first attempt to fill this lacuna, this study presents a two-stage analysis of a novel sample of Italian state-owned museums built by merging information drawn from different sources. In the first stage, we use bootstrapped Data Envelopment Analysis (DEA) to measure the efficiency of museums. In the second stage, we use a bootstrap truncated regression approach to test the extent to which different forms of ICT affect museum efficiency. We distinguish the ICT investments into ‘in situ’ and online services, since the former improve the visitors’ experience on site, while the latter can prepare for the visit or, even, be a substitute of the visit. The results reveal that the use of ICT is generally associated with better performances but ‘in situ’ services show to play a major role.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 The different options for managing the access to and the re-use of digital images of museum collections are explored by Bertacchini and Morando (Citation2013).
2 Contributing content raises questions of authority, as far as the responsibility of the information provided by the public is concerned. Different approaches can be adopted by museums such as, for instance, reviewing the information before incorporating it or presenting the different sources of information (public, curators, etc.) separately on the website (Navarrete, Citation2013).
3 Enumerate Core Survey 4 is the fourth edition of a survey monitoring the status of cultural heritage in Europe. 983 institutions belonging to 28 European countries participated to this fourth round. The dataset includes information for each institution in 2017 on: the state of digitisation activity, the dimension and characteristics of collections, digital access, preservation strategy and expenditure. For more information, see Nauta et al. (Citation2017).
4 These figures are not weighted. Therefore, the actual percentage of the digitization for cultural heritage in Europe is likely to be even lower than what is shown, since institutions with small collections have the same weight as institutions with large collections.
5 Online access of metadata is higher but with similar differences: 76% of libraries and 33% of museums have metadata available online for general use.
6 The issue is quite important for countries like Italy with outstanding heritage distributed across a huge number of sites and museums/institutions.
7 Digital content from museums can be found at the museum websites, but also in the Google Art Project, Wikipedia, image banks, iTunes and Europeana, as well as in a number of video games, blogs and software applications.
8 How to deal with this issue is still an open question in the literature. See, for instance, Simar & Wilson (Citation2007; Citation2011) Banker & Natarajan (Citation2008), McDonald (Citation2009), Daraio et al. (Citation2018) and Banker et al. (Citation2019).
9 The statistical office of MIBACT provides detailed information regarding museums only for those which are state-owned.
10 This procedure makes use of the order-m partial frontier estimator by Cazals et al. (Citation2002) that allows for identifying superefficient units and is based on an iterative procedure. In a few words, at each iteration, the order-m score for each observation is computed (for a set of values of m) leaving out the observation from the reference set. Potential outliers are identified as those superefficient units reporting an output-oriented order-m score lower than 1–α and an input-oriented score higher than 1+α (where α is a parameter and 1±α are thresholds defined for both the orientations) and then further investigated and possibly removed. Once that m and α have been chosen through sensitivity analysis, a potential outlier is removed if the number of observations outperforming it is relatively low. Then the process is iterated until outliers are no longer identified (see Simar, Citation2003 for details).
11 The list of museums in the final sample is available in the online .
12 For descriptive statistics of the variable in the subsample see .
13 The values of the ICT variables reflect the museums’ answers to the relevant questions of the ISTAT survey. Since all these questions are about the implementation of each service, we build up a set of dummy variables, consistently with the possible answers: YES (we attribute a value of 1) or NO (we attribute a value of 0). A museum could also choose not to answer a question and, therefore, the differences in the number of observations for each variable reflect the different response rates for the different questions. details the distribution of answers for each question.
14 The estimates reported in this Section are obtained using the Stata package developed by Badunenko and Mozharovskyi (Citation2016).
15 A recent study (del Barrio-Tellado and Herrero-Prieto, Citation2019) on some Spanish museums finds similar average scores.
16 See and .
17 Results are available from the authors upon request.
18 The descriptive statistics of the employed second stage variables are reported in .
19 Their features and distribution have been described in detail supra, at the end of 3.3.
20 For the online services the number of observations for which the composite index is computed drops to 92, since this is the number of museums that have filed a YES/NO answer to each of the 9 relevant questions.
21 The other estimated models with results largely overlapping those reported here. The results of these additional exercises are available upon request.
22 All estimates are obtained with the bootstrap truncated algorithm proposed by Simar and Wilson (Citation2007). We computed 2000 bootstrap iterations, showing here the mean, the standard deviation and the significance of the coefficients for each variable. The estimates are obtained using the Stata package developed by Badunenko and Tauchmann (Citation2019).
23 The different specifications employed in the estimates shown in and and the use of a robust semiparametric estimator (Simar and Wilson, Citation2007) make us confident about the impact of the variables used on the boundary of frontier efficiency. Moreover, Simar and Wilson (Citation2011) show that the estimator developed by Simar and Wilson (Citation2007) used here is less sensitive to the problem of the omitted variable bias. However, given the limited availability of data and, in particular, the unavailability of panel data, the reported estimates should be taken with some caution.