Assessing the productivity of the Italian hospitality sector:a post-WDEA pooled-truncated and spatial analysis

Claudio Detotto • Manuela Pulina • Juan Gabriel Brida

� Springer Science+Business Media New York 2013

Abstract This paper analyses the productivity of the

hospitality sector (hotel and restaurants) in Italy at a

regional level by using a mix of non-parametric and

parametric approaches. A novel pooled-truncated and

spatial analysis is employed, based upon a window data

envelopment analysis (WDEA), where pure technical

efficiency is computed. The WDEA results show that

Lombardy is the best relative performer. However, overall

Italian regions reveal important sources of inefficiency

mostly related to their inputs. As a post-WDEA, the

pooled-truncated estimation indicates that the rate of util-

isation and regional intrinsic features positively affect

hospitality efficiency. Nevertheless, the spatial analysis

does not support evidence of spill-over effects amongst

Italian regions.

Keywords Regional hospitality sector � Dynamic

window data envelopment analysis � Double

bootstrap � Pooled-truncated regression � Spatial

heterogeneity

JEL Classification C14 � C24 � L83 � R11

1 Introduction

During the second half of 2007, the sub-prime crisis

extended to many other sectors of the economy including

tourism. As the World Travel and Tourism Council

(WTTC 2010) reports, global travel and tourism (T&T)

economy GDP declined by 4.8 % in 2009, causing the loss

of almost 5 million jobs, while investment shrank by over

12 %. Nonetheless, this economic sector still employed

over 235 million people worldwide (8.2 % of all employ-

ment) and generated 9.4 % of the world’s GDP. The more

recent figures released by the World Tourism Organization

(UNWTO 2012) showed a slowdown in international

tourism arrivals with an overall growth of 4 % in 2011.

This setback was due to the on-going economic and

political instability. Nevertheless, in this economic turmoil,

Europe and Italy recorded a relatively positive perfor-

mance, with growth levels above the average figure reg-

istered worldwide.

The contribution of tourism demand as a driver of

economic growth is well established (Bimonte et al.

2012). As pointed by Federalberghi-Mercury (2010), the

Italian hospitality sector plays an important role as a

revenue generator. In terms of number of hotel rooms, it

ranks fourth after the United States, Japan and China

(WTTC 2010) and amongst the European countries, Italy

has the leadership in terms of hotel dimension and

quality (number of stars). Yet, Italian hospitality is

characterised by a strong seasonality, given that its rate

of utilisation (40 %) is much lower than the global

leading countries (e.g. Japan, 74 %; France, 61 %; U.S.,

60 %; Spain, 53 %). Such rate of utilization decreased

by 6.8 % during the period 2000–2004, although tourist

arrivals and nights of stay rose by 7.4 and 1.9 %,

respectively (ISTAT 2011).

C. Detotto (&) � M. Pulina

Researcher at Centro Ricerche Economiche Nord e Sud

(CRENoS) and Department of Economics (DiSEA), Universita

di Sassari, Via Torre Tonda, 34, 07100 Sassari, Italy

e-mail: cdetotto@uniss.it

M. Pulina

e-mail: mpulina@uniss.it

J. G. Brida

Department of Economics at the School of Economics and

Management, Free University of Bolzano, Piazza

dell’Universita, 39100 Bolzano, Italy

e-mail: JuanGabriel.Brida@unibz.it

123

J Prod Anal

DOI 10.1007/s11123-013-0371-x

Overall, Italy has been defined as a pioneer in the hos-

pitality industry alongside other main Mediterranean des-

tinations (namely France and Spain) but while it is

characterised by a very long life cycle, it also displays quite

scattered renewal models within each of the regions (e.g.

Manera Erbina et al. 2010). As emphasised in Federal-

berghi-Mercury (2012), the structural characteristics of

Italian hotels at a regional level show that most of the

infrastructures are located in the North. However, over the

years, there has been a redistribution of the supply within

the country, characterised by an increase in hotel capacity

in the southern regions, where, between 2000 and 2010,

hotel accommodation proved to be particularly dynamic

(e.g. Basilicata, ?87.9 %; Sicily, ?58.5 %; Apulia,

?58.5 %; Calabria, ?57.3 %).

Within this setting, it seems interesting to investigate the

productivity of this rather heterogeneous Italian hospitality

sector on a regional basis. The question is particularly

important in the light of an increasing awareness of sus-

tainability issues that challenge the need for a further

expansion of tourism infrastructure (Bruni et al. 2011). The

question is particularly important as, over the time span

under investigation (2000 and 2004), overall Italian supply

capacity grew by 7.9 %, reaching more than two million

hotel beds in 2004 (ISTAT 2011).

One of the objectives of this paper is to investigate

whether the expansion of this capacity is supported by

productivity performance. As argued by Brown and Dev

(1999, 2000), there is no common agreement on the defi-

nition of productivity, which still remains a rather vague

concept. This is particularly true for the hospitality indus-

try, which is characterised by similarities with the manu-

facturing production process but also by distinctive

features that relate to its heterogeneous and intangible

nature. Within a production function setting, one needs to

take into account a key set of inputs and outputs employed

in the hospitality sector. In this respect, a further distinction

should also be made on the relevant methodological

frameworks, either microeconomic or macroeconomic, that

can help one to better identify the indicators of produc-

tivity. From an empirical point of view, data availability is

an important issue to be addressed.

At a microeconomic level, data are available either from

accounting books or through ad hoc questionnaires

designed to obtain very detailed information on both

demand and supply, as well as on those environmental and

intangible factors that may contribute to influence the

overall performance of the sector. Unfortunately, since

hotels and other types of accommodation produce imper-

fectly substitutable outputs, the firms operate on different

production frontiers, which makes it almost impossible to

compare them. This problem could be overcome by col-

lapsing firms’ heterogeneity into few aggregated indices,

mainly financial data such as value added, total revenues,

total costs and total investments. How to select these

variables is not trivial since it can drive the analysis and the

results could depend on the data. For an extensive study

about misspecification due to variable selection refer to

Melao (2005).

In present study, widely accepted indicators are used in

line with the existing empirical research. Once the objec-

tive measures of performance have been identified, it is

possible to conduct an empirical investigation assuming the

actual productivity of the sector under analysis. Several

approaches concerning data and methodological frame-

work can be used to measure productivity. The simplest

measures of productivity consist of indices and perfor-

mance indicators, such as the ratio of real outputs over real

inputs. However, more sophisticated tools, both parametric

(e.g. Deterministic Frontier Analysis or the Stochastic

Frontier Analysis) and non-parametric (Data Envelopment

Analysis) are also employed in order to calculate the effi-

ciency of a sector that actually incorporates the concept of

the production possibility frontier. The greater the output

for a given input, or the lower the input for a given output,

the more efficient the sector is.

To date, empirical research papers have mainly focused

on the productivity in the manufacturing sector, health

services, educational institutions, the services sector and

private organisations such as banks. On balance, less

attention has been given to analysing hospitality efficiency

and, more in general, tourism activity from a supply side

perspective (see Barros 2005; Wanhill 2011; Pulina et al.

2010; Brida et al. 2012). To this aim, the present paper

employs a novel combined approach involving three dis-

tinctive steps of research, underpinned to a macroeconomic

framework. First, a standard window data envelopment

analysis (WDEA) is run following the seminal papers by

Charnes et al. (1978), Banker et al. (1984), and Charnes

et al. (1985). WDEA is a non-parametric panel approach

that can be viewed as an appropriate tool to analyse the

productivity level of a group of Decision Making Units

(DMUs) with respect to its own performance over time, as

well as the performance of the relatively most productive

decision units within the sample set. Specifically, a macro

investigation is run by analysing Italian hospitality eco-

nomic efficiency at a regional level and its dynamic evo-

lution over time.1 Second, a double bootstrap procedure

and a pooled-truncated regression are run. This parametric

implementation adds new knowledge to policy makers on

the main factors that influence efficiency of the regional

1 As argued by Diewert and Mendoza (1995), DEA is highly

sensitive to data errors and outliers. Notably, since aggregation

alleviates measurement errors at the individual level, using regional

data allows one to reduce such biases.

J Prod Anal

123

hospitality sector. In this respect, the present paper expands

on other non-parametric studies on Italian hospitality (e.g.

Pulina et al. 2010; Brida et al. 2012). The last step of the

research involves a spatial heterogeneity investigation to

test the existence of possible spill-over effects amongst

Italian regions. This latter analysis helps public and private

agents to capture the actual degree of networking and joint

marketing strategies existing amongst the regions.

It is important to highlight that WDEA provides relative

efficiency scores only. This means that enlarging the

sample size can move the production frontier and hence

most efficient regions may be judged as inefficient. Para-

doxically, the WDEA procedure is ideal to detect ineffi-

cient DMUs since they still remain inefficient although

new DMUs are included. Besides, the present study does

not able one to analyse systemic factors that affect Italian

tourism supply and/or the efficiency of Italian sector with

respect to other European countries (that actually is out of

the scope of the present investigation). The proposed

framework allows one to gain insight into Italian regional

differences in tourism productivity proposing a novel post-

DEA technique, that can be considered as a benchmark for

further studies in the field.

The paper is organised as follows. In the next section, an

updated literature review is provided. In the third section,

the methodology adopted is highlighted. The fourth section

provides a DEA analysis for the hotel sector. The fifth and

sixth sections present post-DEA results. Finally, some

policy implications are drawn in the final conclusions.

2 An updated literature review

DEA and post-DEA have been widely used in the litera-

ture. These approaches have been adopted in the empirical

literature in recent years to analyse the efficiency in the

manufacturing sector (e.g. Kravtsova 2008), health services

(e.g. Olesen and Petersen 2002), education and institutions

(Robert et al. 2010), the services sector (Prado Lorenzo and

Garcıa Sanchez 2007) and private organisations such as

banks. Emrouznejad et al. (2008) provide an ample litera-

ture review on the most popular researched areas.

Less attention has been given to the hospitality sector,

which plays an important role in countries highly specia-

lised in tourism. Barros and Alves (2004), Barros (2005),

and more recently Pulina et al. (2010), and Assaf and

Agbola (2011), provide an extensive literature review on

efficiency in the hotel sector. One of the main features of

the reviewed empirical studies is the cross-section dimen-

sion and the relative low number of observations.

As a further update to this strand of the literature, similar

analytical features have been encountered for. For exam-

ple, Neves and Lourenco (2009) use a static input-oriented

DEA model to determine the efficiency of a sample of

worldwide hotel companies during the period 2000–2002.

The authors note that the majority of the hotel companies

are characterised by decreasing return-to-scale. Further-

more, specialised hotel companies perform better than

those characterised by a diversification strategy. Moriarty

(2010) applies the DEA method to establish relative tech-

nical and scale efficiency of hospitality divisions in New

Zealand during the period 1999–2003. He observes that the

majority of the hospitality divisions exhibit increasing

return-to-scale. Shuai (2009) analyses the impact of inter-

net marketing on hotel performance. It is found that

internet marketing (e-communication and e-transaction

orientations) is positively associated with hotel efficiency.

A possible rationale is that hotels’ efficiency is driven by

an increase in tourism demand due to a reduction in

‘‘search’’ price. Assaf and Cvelvar (2010) analyse the

performance of Slovenian hotels over the period

2005–2007. Bernini and Guizzardi (2010) employ a sto-

chastic translog production function to explore the effi-

ciency of a balanced panel of 414 Italian hotels over the

period 1998–2005. They find that efficiency is enhanced by

both firms’ location and the tourism vocation of a desti-

nation. A significant contribution to efficiency relates to the

firms’ location in seaside cities but especially in arts cities.

Italian business corporations are mainly characterised by a

relevant use of the labour factor of production and

decreasing return-to-scale. Sirirak et al. (2011) find an

empirical evidence of the positive influence of ICT adop-

tion on hotel performance in Taiwan. Shuai and Wu (2011)

make use of a DEA and a grey entropy approach to eval-

uate whether internet marketing affects the operating per-

formance of 48 international hotels in Taiwan for the years

2006 and 2007. Results suggest that hoteliers have to adopt

a more strategic approach to exploit the e-market oppor-

tunities by first preparing the ground for direct contact with

customers. Assaf and Agbola (2011) via a DEA double

bootstrap method estimate the performance of a total of 31

Australian hotels for the period 2004–2007. The findings

show that larger hotels and those located in cities are more

efficient than those in suburban and regional areas.

The majority of the reviewed studies relate to the per-

formance of a sample of hospitality firms. Less attention is

paid to the overall and comparative efficiency of a specific

geographical area. By applying DEA methods on aggre-

gated data, for example, Cracolici and Nijkamp (2006)

calculated a provincial efficiency score and a ranking of

efficient provinces in Italy while Cracolici (2008) explore

tourism competitiveness of Italian provinces for the years

1998 and 2001 and observe a weak decrease of efficiency

over the years. Suzuki et al. (2011) extend the work by

Cracolici (2008) and employ Euclidean Distance Minimi-

zation to investigate Italian provinces efficiency as tourist

J Prod Anal

123

destinations in 2001. The findings show that the perfor-

mance of many Italian provinces can be improved con-

siderably. Barros et al. (2011), analyse the determinants of

the technical efficiency of leading French tourism regions.

They find that the sea, sun and strategy based on the beach

endowment, along with the presence of museums and

monuments, play a relevant role in explaining efficiency in

French regions. Molina-Azorin et al. (2011) use a multi-

level approach and hierarchical linear models to show that

Spanish hotel performance is highly influenced by its

location.

To date, one of the main shortcomings of recently

published studies relates to the use of a relatively low

number of observations, although there are a few

exceptions (e.g. Cracolici 2008; Neves and Lourenco

2009, Bernini and Guizzardi 2010; Suzuki et al. 2011).

The present study stands as a novel example of a more

robust WDEA in analysing regional hospitality produc-

tivity and the main causes of economic inefficiency,

combined with a spatial analysis investigation to estab-

lish possible spill-over effects. To the authors’ knowl-

edge, so far, no study has been undertaken combining

WDEA with a spatial econometric investigation for the

hospitality sector.

3 Methodology: WDEA and post-WDEA

3.1 Window data envelopment analysis

In the literature two main DEA models are considered: the

DEA-CCR model developed by Charnes et al. (1978),

which assumes that all DMUs are operating at constant

returns to scale (CRS); the DEA-BCC model, developed by

Banker et al. (1984), which hypothesises variable returns to

scale (VRS).

The DEA approach estimates the (unknown) efficient

production frontier that allows one to calculate and com-

pare a firm’s efficiency to its own benchmark. Different

from a parametric approach, which requires an a priori

specification of the functional form of the production

function as well as an a priori hypothesis on the disturbance

term, DEA is a flexible linear technique that reduces a

multiple input–output productive structure into an easier

virtual uni-input–output analysis. As an additional advan-

tage, the economic variables of interest can be used jointly

despite their measure scale (Koksal and Aksu 2007).

Within a given sample of productive units, a subgroup will

achieve a relative efficiency equal to 1 (or 100 %) and the

residual DMU will be considered as inefficient if it has

reached a score of \1 (or less than 100 %). Mathemati-

cally, the linear programming solves the maximisation

problem in the following manner:

Max hi yi; xi; ui; við Þ ¼PN

n¼1 uniyniPK

k¼1 vkixki

ð1Þ

subject to:

XN

n¼1

uniyni ¼ 1 ð2Þ

uni � 0

vki � 0 n ¼ 1; 2; . . .;N outputs and k ¼ 1; 2; . . .;K inputs

ð3Þ

where yni is the quantity of output n produced by the DMU

i; uni is the weight of output n for the DMU i; xki is the

quantity of input k employed by the DMU i; vki is the

weight of input k for the DMU i.2 Furthermore, we employ

an input-oriented model as described in Eqs. (1–3) since it

is reasonable that in the short run tourist industry have

more control over its inputs than its outputs (see Wang

et al. 2006).

In this study, the Banker, Charnes and Cooper (BCC)

model is adopted, since a preliminary investigation depic-

ted VRS for the most of the regions. The WDEA analysis

has been further implemented by calculating the ratio

between CRS and VRS technical efficiency scores, which

gives scale efficiency scores that can be either CRS,

decreasing returns to scale (DRS) or increasing returns to

scale (IRS) (see Charnes et al. 1978; Banker et al. 1984;

Cullinane et al. 2004). Economic theory states that in the

long run the production level can change when inputs,

which are no longer fixed, vary in the same proportion. The

production function depicts IRS (DRS) when inputs are

increased by a factor a and the output increases by more

(less) than a, and CRS when inputs and output increase by

the same factor a.

3.2 Double bootstrap and pooled-truncated

specification

Though DEA has many advantages, it is not possible to

directly evaluate factors that influence DMUs’ efficiency.

2 The solution of Eq. (1) is given by either a maximisation or a

minimisation approach when either one input or one output is used.

However, in the presence of a multivariate input–output framework,

the problem can be solved with either an output-oriented method

(O-OM), by maximising the numerator while keeping the denomina-

tor constant, or an input-oriented (I-OM) method, by minimising the

denominator while keeping the numerator constant. Within the

O-OM, no DMU in the sample, with the same type of inputs, is able to

derive a higher quantity of output. In general, this setting is employed

for planning and strategic objectives. For example, it is used when a

DMU needs to understand whether an expansion of its capacity is

feasible, as long as the existing infrastructure has already been used at

its maximum capacity given the level of the inputs (Cullinane et al.

2004).

J Prod Anal

123

As an extension, a parametric specification can be used

such as a truncated regression since many of the DEA

efficiency scores typically equal to one. However, since

DEA efficiency scores are the result of a linear program-

ming, they are characterised by high correlation that leads

to bias parameter estimates. Recently, Simar and Wilson

(2007) proposed a double bootstrap method that overcomes

possible problems of serial correlation amongst the esti-

mated efficiency scores and approximates their asymptotic

distribution. They argue that conventional bootstrap tech-

niques used in the post-WDEA procedure do not allow for

valid inferences. Only very recently a few studies

employing the double-bootstrap procedure have appeared

in the hospitality literature (e.g. Barros and Dieke 2008;

Assaf et al. 2011; Assaf and Cvelvar 2010; Assaf and

Agbola 2011; Barros et al. 2011).

Following this methodological strand of the literature, in

the present paper, the following pooled-truncated specifi-

cation is used:

hit ¼ Zitbþ eit � 1 i ¼ 1; . . .; n t ¼ 1. . .T ð4Þ

where hit is the ith DMU’s efficiency score at time t (DMUs

are technically efficient or inefficient when hit = 1 or

hit [ 1, respectively3); Zit contains factors that are assumed

to affect the DMUs’ efficiency; b is the vector of param-

eters to be estimated; eit is the residual that is assumed to be

white noise. The b estimates and their confidence intervals

are calculated following a procedure that refers to the

Algorithm #2 proposed by Simar and Wilson (2007). First,

the DEA input-oriented is run for the DMUs under inves-

tigation using xit and yit as inputs and outputs, respectively.

Second, Eq. (4) is estimated by using the maximum like-

lihood method to obtain estimates of the parameters ðbbÞand the deviation standard of the residuals ðbreÞ. Third, for

each DMU the following loop is repeated L1 times (in this

case 10,000 times): (a) for each DMU, eitb is drawn from

the Nð0; breÞ distribution with left truncation at ð1� ZitbbÞ,

with b = 1,…,L1; (b) again, for each DMU, h�itb ¼ Zitbb þ

eitb is computed; (c) a new pseudo data set is defined where

x*it = xit and y�it ¼ yitðbhit þ bh�itbÞ; (d) using the constructed

pseudo data set ðy�it; x�itÞ, the input-oriented DEA is re-run

to compute efficiency estimates for all the DMUs ðbh�itbÞ.Fourth, the bias-corrected efficiency scores are computed

as follows:bbh it ¼ 2bhit �

PL1

b¼1bh�itb. Fifth, the maximum

likelihood method is used to estimate the truncated

regression of the bias-corrected efficiency scores, which

provides with marginal effects of the explanatory variables

ðbbÞ and estimated standard deviation of the residuals ðbreÞ.Sixth, again for each DMU the following bootstrapping loop

is repeated L2 times (again, 10,000 times): (1) for each DMU,

eits is drawn from the Nð0; bbreÞ distribution with left trunca-

tion at ð1� Zits^bÞ;with s = 1,…, L2; (2) for each DMU,

h��its ¼ Zit^bþ eits is computed; (3) the maximum likelihood

method is employed to estimate the truncated regression of

h��its on Zit that provides with a new set of estimated marginal

effects of the explanatory variables ðbbbb sÞ and standard

deviation of residuals (bbbr e;s). Hence, the L2 bootstrap esti-

matesbbbb s and

bbbr e;s are used to construct estimated confidence

intervals for each of the unknown element in (4) (see also

Balcombe et al. 2008; Assaf et al. 2011).

3.3 Spatial analysis

An important limitation when dealing with regional data is

the existence of spatial correlation between observations,

which produces unbiased but inefficient Ordinary Least

Squares (OLS) coefficients (Anselin 1988) due to the non

diagonal structure of the disturbance term. Spatial auto-

correlation means that similar values tend to be close to

each other. Unfortunately, model (4) does not take into

account spatial spill-over effects among Italian regions. In

this paper, we employ the Moran’s I test to the DEA

efficiency scores in order to check for the presence of

spatial autocorrelation among Italian regions. This way we

can identify the existence of clusters of regions character-

ized by similar level of technical efficiency.

The Moran’s I test is calculated as follows (Anselin

1988):

I ¼ n

S0

~h0W ~h~h0~h

ð5Þ

where n is the number of units, the vector ~hindicates the

deviation from the mean of the variable of DEA scores has

in Eqs. (1–3), W represents the spatial weight matrix and S0

is the sum of all components of the matrix W. Under the

null hypothesis, the expected value of Moran’s I test is:

EðIÞ ¼ �ðn� 1Þ�1 ð6Þ

In this study, the inverse of the distance between the cen-

troids is taken as a spatial weight, which signifies that the

higher the distance between two regions the lower is the

magnitude of the spatial effects between them.

By construction, the Moran I test ranges from -1 to ?1.

Specifically, positive values of the spatial autocorrelation

3 Standard DEA scores range between zero and one, as mentioned in

Sect. 3.1. Unfortunately, double bounds of the dependent variable

impose a more complex procedure in solving Eq. (4). In order to

simplify the econometric model, we follow the suggestion of Simar

and Wilson (2007), who propose to use the inverse of standard DEA

scores. This way, the new dependent variable has a single lower

bound equals to one.

J Prod Anal

123

indicate spatial similarity, while negative values are asso-

ciated with spatial dissimilarity (Griffith 2003; pp. 5). A

zero value indicates that the pattern is randomly spatially

distributed.

4 Italian regions economic efficiency

4.1 Hospitality productivity

As previously stated, the definition of productivity is

complex and includes several dimensions of performance

ranging from efficiency to effectiveness, from tangible

measures, such as outputs, rates of turnover and absen-

teeism, to intangible measures such as customer satisfac-

tion, loyalty and job satisfaction.

From an economic point of view, productivity is defined

as the amount of output of goods and services for each unit

of input used in a given time period. The concept of pro-

ductivity is closely linked to the notion of efficiency. If a

DMU is efficient it is said to be operating on the production

frontier. Hence, a rise in efficiency implies an increase in

productivity, while an outward shift in the production

frontier implies productivity growth. As Farrell (1957)

stated, the economic efficiency relates to how far a given

DMU can increase its output without using further

resources.

Mandl et al. (2008) emphasise that effectiveness relating

to the final outcome to be achieved is another important

concept. As far as the hospitality sector is concerned,

effectiveness can be influenced by several endogenous

determinants such as firms’ characteristics, pricing policy,

as well as exogenous factors such as the stage of the life

cycle of product in the destination and socio-economic

characteristics of the segments of demand.

To measure efficiency and effectiveness one needs to

employ observed data. To exemplify this semantic issue

and build a more systematic framework within an empirical

setting, two distinctive lines of research can be identified.

On the one hand, if one is analysing the productivity of a

DMU at a microeconomic level, it seems reasonable to use

a wider range of objective measures of efficiency of per-

formance (e.g. cost per unit; output per employee) and

outcomes of effectiveness (e.g. sales), possibly including

those intangible outcomes, such as customer experience

and satisfaction, that can be only quantified with ad hoc

survey data (Sigala 2004). In this respect, the implemen-

tation of categorical variables makes it possible to pick up

qualitative information, such as the quality of the services

offered and the quality of the working environment, which

can give a contribution in assessing DMUs performance.

On the other hand, if one is analysing the productivity at a

macroeconomic level, it seems reasonable to use standard

objective measures of efficiency and effectiveness such as

overall costs and overall revenues. These are generally

census data that tend to allocate firms to a specific industry

without providing much detailed information.

From a practitioner point of view, productivity, and

hence efficiency and effectiveness, is a strategic tool to

increase national wealth, particularly in mature economies

where insufficient productivity cannot be hidden by an

expansion in market size (Goh 2010). This is particularly

true for a labour intensive productive process, such as the

one which takes place in the hospitality sector. From

research related to the manufacturing sector it emerges

that, while less productive firms are not able to face

competition, an intermediate level of productivity allows

firms to only compete in internal markets and firms char-

acterised by high productivity are commonly more com-

petitive in external markets (e.g. Bernard et al. 2007).

Hospitality, and more in general tourism activity, is

regarded as a non-standard export sector, since consump-

tion happens in situ, and it is the consumer who moves

instead of the product (Cortes-Jimenez and Pulina 2010).

Hospitality can serve local, national and foreign consum-

ers. Hence, one can argue that the more productive a DMU,

the more it is able to equally compete within these three

segments of demand.

Increasingly, the hotel sector tends to operate in a

globalised and highly competitive marketplace, where

making profits becomes more and more challenging, given

the very volatile demand. Thus, monitoring hotel produc-

tivity is a key strategy to identify whether the sector is

profitable and is growing over time, and moreover to set ad

hoc marketing and pricing policies.

4.2 Inputs-outputs and data description

The DEA approach allows one to deal with multi-output

multi-input processes deriving a virtual input and output,

obtained as linear combination of the observed inputs and

outputs. The choice of inputs and outputs is a critical task

because a correct identification allows one to represent

properly the production function. Assaf et al. (2010)

present a comprehensive literature review that sheds light

on the indicators commonly used in the most relevant

studies on tourism hospitality to assess the DMUs perfor-

mance. The majority of these studies employ a microeco-

nomic framework, where efficiency is assessed at a firm

level. In general, inputs and outputs are obtained from

accounting books, which tend to gather data in a more

aggregated manner. Only a very few studies take into

account how intangible outputs, and specifically customer

satisfaction, contribute to the overall firms productivity.

Following previous studies, and based on the data

availability at a macroeconomic level, in the present study,

J Prod Anal

123

sales revenue (SR) and value added (VA) are employed as

outputs. Specifically, sales revenue is defined as the prod-

uct between the price at which goods and services are sold

and the number of units, or amount sold. This indicator has

been used in various studies and more recently in Reynolds

and Thompson (2007) to assess efficiency in brand res-

taurants. In the data provided by the Italian National

Institute of Statistics (ISTAT), value added is defined as the

market value of firms’ product, or service, minus the cost

of inputs purchased from other firms. It uses retail price

rather than market price (i.e. the price at which the man-

ufacturer recommends that the retailer sells the product).

Therefore, the value added in such a specification encom-

passes price-discrimination issues that are going to incor-

porate non-financial factors such as location and quality.

Nonetheless, these measures are recognised to be adequate

indicators of firms efficiency and have been used in other

empirical investigations (e.g. Barros and Santos 2006;

Wang et al. 2006; Min et al. 2008). Referring to the

aggregated business of the tourism sector, SR indicates the

market share in monetary terms of the DMUs. VA repre-

sents the ability to make profit of each region in the sector

under study.

Within the classical framework, two inputs are also

employed: labour costs (LC) and gross fix investments

(K)—used here as physical capital production factor.

Specifically, LC are defined as the total expenditure borne

by employers in order to employ workers; this indicator

includes direct remuneration, bonuses, payments for days

not worked, severance pay, benefits in kind. They also

include indirect costs linked to employees, such as con-

tractual and voluntary social security contributions, direct

social benefits, vocational training costs, other social

expenditure (e.g. medical services), and taxes relating to

employment regarded as labour costs, minus any subsidies

received. K are defined as the acquisition of fix capital,

which also comprises the value of capital goods produced

by the firm. In this context, each DMU combines labour

and capital to produce the outputs, namely sales revenue

and value added. The DEA approach allows us to detect

which DMUs use the inputs in the most efficient way.

The data used in this analysis are collected from ISTAT,

which assembles the variables at regional level. Table 1

shows descriptive statistics of the variables (in thousands

euros) under study in the DEA procedure. The analysis

concerns 19 Regions and the two autonomous Provinces of

Trento and Bozen for the time span 2000–2004.

5 Window-DEA results

At a macro level, a comparison of efficiency is provided

amongst all of the 20 Italian regions. One region, Trentino

Alto Adige, is reported as two provinces by the Italian

National Statistics Office (ISTAT) and this analysis follows

the national classification; thus, 19 regions and two

autonomous provinces (Trento and Bozen) are considered

resulting in 21 DMUs (see Fig. 1). Given the availability of

official statistic data, updated in December 2010, a time

span of 5 years (2000–2004) is considered. Over-fit prob-

lems are avoided as the minimum number of DMUs is

more than twice the total number of inputs and outputs in

the DEA (Min et al. 2008).

To run the analysis, the software package Frontier

Analyst 3.1.5 is used. The selection of the length of the

window is an important issue in window DEA because the

results may depend on the number of windows employed.

In the present analysis, the following formulas adapted

from Sun (1988) are applied (see also Cooper et al. 2000).

Given n DMUS and k periods, the length of window is

given by the following:

p ¼kþ1

2when k is odd

kþ12� 1

2when k is even

�

ð7Þ

In this case, as twenty-one regions and 5 years’ worth of

data are used, for a total of 105 observations, the chosen

window length is 3 years and three separate windows are

constructed as shown in Table 2; hence, the number of

Table 1 Descriptive statistics

Variables Code Model Obs. Mean St.Dev. Min Max

Capital costs K DEA (INPUT) 105 185,048.30 198,983.38 5,796 1,277,027

Labour costs LC DEA (INPUT) 105 482,125.35 482,844.53 14,581 2,157,532

Sales revenue SR DEA (OUTPUT) 105 2,380,467.55 2,367,530.27 100,890 10,770,620

Value added VA DEA (OUTPUT) 105 908,060.98 845,105.92 25,982 3,710,862

Bed places utilization NRU POST-DEA 105 37.93 8.53 21.8 58.8

Coeff. Variation of bed places utilization CV-NRU POST-DEA 105 14.91 4.91 3.65 23.32

Art cities DART POST-DEA 105 0.14 0.35 0 1

High quality hotels HQ-HOTEL POST-DEA 105 10.99 4.74 4.33 22.37

Calculation on data collected from ISTAT

J Prod Anal

123

observations reduces to 63 (21 regions over three separate

windows).

The first row (with values of 97.1, 98.0 and 88.3 %)

shows the relative technical efficiency of the Abruzzi

region in 2000, 2001 and 2002, respectively. The second

row (with values of 95.3, 88.3 and 93.2 %) shows the

relative technical efficiency of Abruzzi in 2001, 2002 and

2003, respectively, and so on. The same results, read by

column, represent the stability of efficiency scores both in

absolute terms as well as in terms of the relative perfor-

mance of that region with respect to the other regions in the

sample. It is worthwhile noting that two regions present a

few ‘‘extreme mismatches’’ that constitute a greater than

10 % points annual change, which might be due to mis-

reporting of data or computational errors. Examples of

misreporting are also found in Charnes et al. (1985) and

Cullinane et al. (2004). The last columns show the mean,

standard deviation (SD) and coefficient of variation (CV)

for each region. The latter is calculated as the ratio between

the standard deviation and the mean.

Overall, from Table 2 an increasing efficiency score

emerges for Emilia-Romagna, Lazio, Sardinia, Umbria and

Veneto. Reading the results by column, the best perfor-

mance was achieved consistently by Lombardy; however,

in 2003, the most efficient regions included Liguria, Molise

and Piedmont.

Following Pulina et al. (2010), a further step is to test

the relationship that exists between the mean efficiency and

its volatility measured in terms of the CV. A negative

correlation between the mean and the CV is expected as

high efficiency mean is most likely to be associated with a

low volatility. One reason for this negative correlation is

that the CV tends to zero as the mean approaches 100. High

CVs occur in regions such as Aosta Valley and Sicily,

which also have relatively low means. Spearman’s test is

then carried out in order to test whether the DMUs are

characterised by homoscedasticity (as a null hypothesis

q = 0) or heteroscedasticity (as an alternative hypothesis

q = 0). The calculated q for the Italian regions (Table 2)

equals -0.67. This value, in absolute terms, is higher than

the corresponding critical value (0.43), at the two-tailed

5 % level of significance. Hence, the null hypothesis can be

rejected and the Italian regions are characterised by a non-

constant variance at the cross-sectional level.

A synthesis of the main results achieved from the DEA

analysis is provided in Figs. 1 and 2. The former presents a

static analysis of the regional economic efficiency by

comparing the last year to the trend performance across the

time span under investigation. Clockwise, the top right

quadrant depicts the ‘‘moving ahead’’ regions, which

denote a score efficiency higher than both the trend average

score and the average score for 2004. Results show that

eight regions (namely Lombardy, Veneto, Umbria, Lazio,

Piedmont, Emilia Romagna, Tuscany and Liguria) belong

to this group, although only Lombardy is the peer that

shows the highest level of pure technical efficiency in each

year (Table 1). The bottom right quadrant includes the

‘‘catching up’’ regions, which reach a score efficiency

lower than the trend and higher than the average score for

2004. Sardinia and Marche are the sole regions that belong

to this group. The bottom left quadrant depicts the ‘‘falling

further behind’’ regions, which experience an average score

lower than both the trend and the average score in 2004.

Seven regions belong to this group namely: Aosta Valley,

0 0 0 050 55 60 65 70 75 80 85 90 95 100

70

75

80

85

90

95

100Moving aheadLosing momentum

Catching upFalling further behind

Tre

nd 2

000

-200

4

Score 2004

LO

VE

UMLA

ER

SA

LITO

MOPI

BZ

MA

AB

FVGCA

CAL

BA

TR

AP

AV

SI

Fig. 1 Italian regions: static

efficiency performance 2004

and trend (2000–2004). Notes:

‘‘Moving ahead’’: Molise (MO),

Lombardy (LO), Liguria (LI),

Veneto (VE), Lazio (LA),

Piedmont (PI), Umbria (UM),

Marche (MA), Emilia Romagna

(ER) and Toscany (TO);

‘‘Catching up’’ Sardinia (SA);

‘‘Falling further behind’’ Apulia

(AP), Sicily (SI), Trento (TR),

Calabria (CA), Aosta Valley

(AV), Bozen (BZ) and Campania

(CA); ‘‘Losing Momentum’’

Friuli Venezia Giulia (FVG),

Basilicata (BA) and Abruzzo

(AB)

J Prod Anal

123

Table 2 Regional efficiency: a window DEA approach (VRS)

Regions Years Statistics

2000 2001 2002 2003 2004 Mean SD CV

Abruzzo 97.1 98.0 88.3

95.3 88.3 93.2

97.2a 96.2 62.6 90.7 10.5 0.116

Aosta Valley 84.7 81.5 63.5

71.3a 63.5 52.1

62.0 54.7 69.0 66.9 10.4 0.155

Apulia 80.7 59.3 77.8

59.3 77.8 67.5

90.3a 70.7 68.2 72.4 9.6 0.133

Basilicata 86.3 – 78.6

– 78.6 80.3

78.6 86.3 60.4 78.5 8.0 0.102

Bozen 77.4 85.7 90.7

85.7 90.7 69.9

84.8 70.0 75.1 81.1 7.7 0.095

Calabria 74.6 79.6 96.7

79.6 96.7 68.1

99.8 72.4 58.6 80.6 13.5 0.167

Campania 89.6 94.6 100

94.6 100 75.9

100 72.3 59.6 87.4 13.8 0.158

Emilia Romagna 84.1 96.4 92.8

100 92.1 78.5

90.5 79.0 100 90.4 7.8 0.086

Friuli Venezia Giulia 89.5 82.6 93.7

82.5 93.7 90.3

94.1 100 65.2 88.0 9.6 0.110

Lazio 98.6 96.2 95.8

96.2 95.4 76.1

94.5 76.1 100 92.1 8.7 0.095

Liguria 100 91.8 69.0

89.4 68.9 100

65.6 100 85.3 85.5 13.5 0.157

Lombardy 100 100 100

100 100 100

100 100 100 100 0.0 0.000

Marche 90.1 79.9 69.9

73.8 67.3 88.4

68.6 87.9 82.0 78.6 8.5 0.108

Molise 87.7 100 100

100 100 100

100 100 76.7 96.0 7.8 0.082

Piedmont 87.1 100 100

100 100 100

100 100 73.5 95.6 8.8 0.092

Sardinia 66.5 74.5 70.7

74.5 70.7 62.6

J Prod Anal

123

Sicily, Apulia, Trento, Basilicata, Calabria, and Bozen.

Finally, the top left quadrant includes the ‘‘losing

momentum’’ regions, which denote a score higher than the

trend and lower that the average score in 2004. The regions

of Campania, Friuli Venezia Giulia, Abruzzi, Piedmont and

Molise fall into this category.

Figure 2 provides a dynamic picture of the performance

of each Italian region. The majority of the regions fall in

either the ‘‘moving ahead’’ or the ‘‘falling further behind’’

quadrant. There is a stable peer group that includes

Lombardy (as the peer) and Veneto, and a stable ‘‘falling

further behind’’ group that includes Aosta Valley, Apulia,

Basilicata, Trento and Sicily. All the other regions can be

considered in a transition phase. For example, Bozen and

Calabria, which were catching-up regions within the first

window, have become ‘‘falling further behind’’ DMUs

within the last window. Umbria, which in the first window

was in the ‘‘losing momentum’’ quadrant, has improved its

performance and in the second and third window has fallen

into the ‘‘moving ahead’’ quadrant. Marche and Sardinia,

which were ‘‘falling further behind’’ regions, have become

‘‘catching up’’ DMUs within the last window. Lazio and

Emilia Romagna regained a highest performance after

having lost momentum within the second window of

analysis. Abruzzi, Piedmont and Molise, which were

‘‘moving ahead’’ regions, have fallen into the ‘‘losing

momentum’’ quadrant within the last window of

investigation.

The window DEA analysis has been implemented by

calculating the scale efficiency scores, which can be either

CRS, IRS or DRS. Table 3 depicts the share of regions

denoting IRS and DRS, respectively. It emerges that the

share of regions characterised by IRS has increased over

time, while the share of regions that present DRS has

remarkably decreased. Hence, in the long run these firms

typically are small sized.

In order to compare regions characterised by IRS and

non-IRS, the average net rate of utilisation of bed places is

calculated for both groups (see Table 3). The rationale is

that higher ‘‘excess capacity’’, which corresponds to low

utilisation rate, should be associated with non-increasing

returns to scale, because it might indicate that hotels are

oversized. Although regions with increasing returns to

scale have higher utilisation rates, except for the last year

included in the sample, the differences in the average level

of utilisation rate between the two groups are not statisti-

cally significantly different from zero. As a result, there

does not seem to be any empirical evidence of the

Table 2 continued

Regions Years Statistics

2000 2001 2002 2003 2004 Mean SD CV

75.6 67.3 100 73.6 10.2 0.138

Sicily 87.2 77.8 69.8

79.8 69.7 63.4

65.9 63.0 49.6 69.6 10.4 0.150

Toscany 99.2 84.8 82.2

85.5 81.4 82.1

79.4 84.7 81.8 84.6 5.5 0.065

Trento 80.4 68.8 84.6

68.8 84.6 64.0

82.5 66.0 65.9 74.0 8.3 0.112

Umbria 77.7 100 84.7

100 84.7 98.8

89.1 98.8 100 92.6 8.2 0.088

Veneto 93.9 92.3 91.0

95.5 95.6 97.5

98.5 94.9 100 95.4 2.7 0.029

Italy 87.3 87.2 85.7

86.6 85.7 81.4 84.5 3.1 0.037

86.5 82.9 77.8

SD standard deviation, CV coefficient of variationa In 2001 the Basilicata region has not been included in the analysis since official data are missing, which does not affect the DEA overall

analysis. ‘‘Extreme mismatch’’ (over 10 % points change)

J Prod Anal

123

C

AV

SAAP

SIMA

BA

TR

LI

CALBZWUMTO FVG

ERVE

AB PICA

LO

LA MO

60.0

65.0

70.0

75.0

80.0

85.0

90.0

95.0

100.0

Score 2002

Tre

nd

200

0-20

02

A

B

BZCAL

TRSI

SA AP

AV

MABA

TOW

LACA ER FVG

LI

ABUM

VE

PI ; LO; MO

50.0

55.0

60.0

65.0

70.0

75.0

80.0

85.0

90.0

95.0

100.0

Score 2003

Tre

nd

200

1-20

03

CA

SIAV

TR

BACAL AP

ABFVG

PI MO

BZ

W

MATO

LI

SA

LAER

UMVELO

55.0

60.0

65.0

70.0

75.0

80.0

85.0

90.0

95.0

100.0

60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0

50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0

40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 90.0 95.0 100.0

Score 2004

Tre

nd

200

2-20

04

Fig. 2 Italian regions: dynamic

efficiency performance (last

window year and window

trend). Notes: ‘‘Moving ahead’’:

top right quadrant; ‘‘Catching

up’’: bottom right quadrant;

‘‘Falling further behind’’:

bottom left quadrant; ‘‘Losing

Momentum’’: top left quadrant;

Italian average (W)

Table 3 Percentage of regions

showing IRS and DRS—

Window DEA (VRS)—

(2000–2004)

2000 2001 2002 2003 2004

Italy (IRS; DRS) 47.6; 42;9 57.1; 19.0 76.2; 4.8

66.7; 4.8 81.0; 0.0 52.4; 28.6

81.0; 0.0 66.7; 9.5 66.7; 4.8

Average net rate of utilization

IRS regions 40.04 41.19 37.35 38.63 35.76

No-IRS regions 37.84 38.00 35.83 34.71 39.63

J Prod Anal

123

existence of a relationship between capacity and returns to

scale in Italian tourist industry.

Although the DEA approach does not allow one to gain

a full understanding of the factors of inefficiency, varia-

tions in outputs and inputs can be calculated in order to

achieve the necessary improvement to obtain the score

within the benchmark efficiency frontier. Table 4 provides

useful information on possible ways to improve efficiency.

On the one hand, in a great part of the Italian regions,

potential improvements can be achieved via a consistent

decrease in terms of capital costs (K) and labour costs (LC).

On the other hand, important improvements have been

achieved in terms of sales revenue (SR) and gross value

added (GVA) in the last year under investigation. This is

especially true for the Bozen and Trento provinces.

Although the window-DEA is a powerful procedure,

particularly for entrepreneurs who rely on input-cost min-

imisation as a measure of organisational efficiency, it holds

important limitations since it does not allow for estimating

a random error in the estimation of efficiency (Reynolds

2003; Assaf and Agbola 2011). However, the statistical

properties of the window-DEA can be retained via a

bootstrap approach that can be further extended to a pooled

parametric estimation through which statistical hypotheses

can be directly tested.

6 Post-DEA results: a pooled-truncated regression

Following the econometric approach showed in Sect. 3.2, a

double bootstrapping method is followed and a pooled-

truncated regression is applied. To this aim the software

library FEAR 1.15, for the statistical package R, by Wilson

(2008) is used. The relevant regression equation is given by

the following:

hit ¼ aw þ axNRUit þ a2CV NRUit þ a3DARTit

þ a4HQ HOTELSit þ eit

i ¼ 1; . . .; n t ¼ 1; . . .; Tð8Þ

where hit is the DMU’s (i) (in)efficiency score at time t; a is

the constant term; a1, a2, a3are the parameters to be esti-

mated. NRU is the net rate of utilisation of bed places,

which proxies the capacity of fully utilised establishments

during the opening period. Usually highest levels coincide

with regions that are characterised by higher performances

in the tourism activity, and a negative sign is expected. In

other words, high levels of NRU are associated with (low)

high levels of (in)efficiency h. The net occupancy rate of

bed places is obtained by dividing total overnight stays in a

given period (in this case, during a year) by the product of

the bed places and the number of opening days. CV_NRU is

the annual coefficient of variation of the net rate of util-

isation to pick up the market volatility. It is calculated for

each region as the ratio of the standard deviation of the

monthly net rate of utilisation and the annual average of

NRU. High values of CV_NRU indicate the presence of

seasonality in the hotel demand. Hence, the relationship

between h and CV_NUR is expected to be positive. Usu-

ally, art cities reach high level of tourism industry spe-

cialisation; DART is a dummy variable that explicitly takes

into account the effects that well-known arts city (namely

Rome, Florence and Venice) have on regional hospitality

sector efficiency. Hence, a negative sign is expected.

HQ_HOTELS is defined as the share of high quality hotels

per region. It is calculated as the ratio between the number

of four and five stars hotels and the total number of hotels,

and is a proxy of the quality of the accommodation supply

in a given region. The expected sign is negative since high

quality hotels are more profitable than low quality hotels.

Finally, eit is the residual that is assumed to be white

noise. Some descriptive statistics are shown in Table 1.

The arts city regions (namely Lazio, Tuscany and Veneto)

account for 14 % of the sample. NRU, CV_NRU and

HQ_HOTELS have been multiplied by 100 in order to

express them in percentage terms.

Main results are reported in Table 5. Models (1) and (3)

contain the results of the pooled-truncated approach. In

Model (1), all the variables seem to explain Italian hospi-

tality sector’s economic efficiency. Specifically, an

increase in the net rate of utilisation causes an increase in

the pure technical efficiency (a1 ¼ �0:05), indicating that

the more the physical capital is used, the greater the effi-

ciency of the region.

Furthermore, an increase in the volatility of the net rate

of utilisation (CVNRU) leads to an increase in regional

inefficiency (a2 ¼ 0:05). Our findings are a further empir-

ical evidence of the negative impact of the seasonality on

tourism industry. Tourism seasonality imposes higher costs

to firms because they have to oversize their capability in

order to cope with the peak of the demand, thus reducing

their efficiency.

On average, regions specialised in cultural tourism, such

as Lazio, Veneto and Tuscany, tend to be more efficient

than the other Italian regions. Usually, cultural tourism is

concentrated in a few famous cities, namely Rome, Venice

and Florence, which allow one to realize scale agglomer-

ation effects. On the contrary, maritime and mountain

touristic regions are characterised by a wild spatial dis-

persion of attractiveness and destinations, which leads to

higher expenditures of transports and public services.

In Model (3), HQ_HOTELS is included into the

regression. Notably, all the estimates are congruent, with

the only exception of NRU which, although it still presents

a negative sign, is not statistically significant. HQ_HO-

TELS coefficient is negative (-0.07) and significant, indi-

cating that an increase in the share of high quality hotels

J Prod Anal

123

Ta

ble

4It

alia

nre

gio

ns:

tota

lp

ote

nti

alim

pro

vem

ents

(%)

inte

rms

of

sale

sre

ven

ue

(SR

),g

ross

val

ue

add

ed(G

VA

),la

bo

ur

cost

s(L

C)

and

fix

edin

ves

tmen

ts(K

);w

ind

ow

DE

A(V

RS

)—

(20

00

–2

00

4)

Reg

ions

Var

iab

les-

yea

rs

SR

00

GV

A0

0L

C0

0K

00

SR

01

GV

A0

1L

C0

1K

01

SR

02

GV

A0

2L

C0

2K

02

SR

03

GV

A0

3L

C0

3K

03

SR

04

GV

A0

4L

C0

4K

04

Ab

ruzz

o0

.02

6.6

-2

.9-

2.9

0.0

1.4

-2

.0-

2.0

0.0

0.0

-1

1.7

-5

3.9

0.0

0.0

-4

.7-

4.7

0.0

0.0

-1

1.7

-5

3.9

0.0

4.3

-6

.8-

6.8

0.0

0.0

-2

.8-

64

.30

.01

2.4

-3

.8-

3.8

0.0

0.0

-3

7.4

-4

0.5

Ao

sta

Val

ley

13

.50

.0-

15

.3-

15

.34

.80

.0-

32

.7-

18

.57

.10

.0-

36

.5-

38

.7

6.8

0.0

-2

8.7

-2

8.7

7.1

0.0

-3

6.5

-3

8.7

0.0

0.0

-4

7.9

-4

7.9

0.0

0.0

-3

8.0

-3

8.0

0.0

4.3

-4

5.3

-4

5.3

0.0

0.0

-3

1.1

-4

4.0

Ap

uli

a0

.00

.9-

19

.3-

19

.30

.00

.0-

40

.7-

58

.50

.00

.0-

22

.2-

70

.8

0.0

0.0

-4

0.7

-5

8.5

0.0

0.0

-2

2.2

-7

0.8

0.0

0.0

-3

2.5

-3

2.5

0.0

0.0

-9

.7-

66

.40

.00

.0-

29

.3-

40

.80

.00

.0-

31

.9-

57

.6

Bas

ilic

ata

0.2

23

.2-

13

.7-

13

.7-

--

–0

.80

.0-

21

.4-

65

.8

––

––

0.8

0.0

-2

1.4

-6

5.8

0.0

4.2

-1

9.7

-7

7.2

0.0

0.0

-2

1.4

-6

3.4

0.0

0.0

-1

3.7

-6

8.3

0.0

0.0

-3

9.6

-8

3.5

Bo

zen

42

.50

.0-

22

.6-

22

.66

4.0

0.0

-1

4.3

-2

3.1

54

.40

.0-

9.3

-2

1.3

64

.00

.0-

14

.3-

23

.15

4.4

0.0

-9

.3-

21

.32

6.2

0.0

-3

0.1

-9

0.9

0.0

0.0

-1

5.3

-6

2.7

0.0

0.0

-3

0.0

-9

4.6

0.0

0.0

-2

5.0

-5

2.6

Cal

abri

a0

.00

.0-

25

.4-

63

.60

.00

.0-

20

.4-

86

.37

.60

.0-

3.3

-6

1.2

0.0

0.0

-2

0.4

-8

6.3

7.6

0.0

-3

.3-

61

.20

.00

.0-

31

.9-

68

.6

0.0

0.0

-0

.3-

68

.70

.00

.0-

27

.6-

72

.50

.00

.0-

41

.4-

46

.6

Cam

pan

ia0.0

1.3

-1

0.4

-1

0.4

12

.10

.0-

5.4

-5

.40

.00

.00

.00

.0

12

.10

.0-

5.4

-5

.40

.00

.00

.00

.01

1.0

0.0

-2

4.2

-2

4.2

0.0

0.0

0.0

0.0

0.0

0.0

-2

7.7

-3

0.9

0.0

0.0

-4

0.4

-4

0.4

Em

ilia

Rom

agn

a2

.40

.0-

15

.9-

28

.43

.80

.0-

3.6

-3

.67

.80

.0-

7.2

-7

.2

0.0

0.0

0.0

0.0

11

.20

.0-

7.9

-7

.90

.00

.0-

21

.6-

21

.6

0.0

0.0

-9

.6-

9.6

0.0

0.0

-2

1.0

-2

4.9

0.0

0.0

0.0

0.0

Fri

uli

Ven

ezia

Giu

lia

0.0

10

.4-

10

.5-

10

.50

.00

.0-

17

.4-

17

.42

6.1

0.0

-6

.3-

57

.3

0.0

0.0

-1

7.5

-1

7.5

26

.10

.0-

6.3

-5

7.3

0.0

0.0

-2

1.6

-2

1.6

0.0

0.0

-5

.9-

74

.50

.00

.00

.00

.00

.00

.0-

34

.8-

34

.8

Laz

io0

.01

3.5

-1

.4-

42

.39

.30

.0-

3.8

-3

.81

.30

.0-

4.2

-4

.2

9.3

0.0

-3

.8-

3.8

3.2

0.0

-4

.6-

4.6

0.0

19

.6-

9.7

-5

3.2

0.0

0.0

-5

.5-

5.5

0.0

1.2

-2

3.9

-2

3.9

0.0

0.0

0.0

0.0

Lig

uri

a0

.00

.00

.00

.00

.02

.1-

8.2

-8

.27

.20

.0-

31

.0-

31

.0

0.0

0.0

-1

0.6

-1

0.6

7.6

0.0

-3

1.1

-3

1.1

0.0

0.0

0.0

0.0

0.0

0.0

-3

4.4

-3

4.4

0.0

0.0

0.0

0.0

0.0

0.0

-1

4.7

-4

0.7

Lo

mb

ard

y0

.00

.00

.00

.00

.00

.00

.00

.00

.00

.00

.00

.0

J Prod Anal

123

Ta

ble

4co

nti

nu

ed

Reg

ions

Var

iab

les-

yea

rs

SR

00

GV

A0

0L

C0

0K

00

SR

01

GV

A0

1L

C0

1K

01

SR

02

GV

A0

2L

C0

2K

02

SR

03

GV

A0

3L

C0

3K

03

SR

04

GV

A0

4L

C0

4K

04

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

Mar

che

0.0

0.0

-9

.9-

9.9

0.0

11

.0-

20

.1-

20

.10

.03

.0-

30

.1-

30

.1

0.0

0.0

-2

6.2

-2

6.2

0.0

0.0

-3

2.7

-3

2.7

0.0

0.0

0.0

0.0

0.0

0.8

-3

1.4

-3

1.4

0.0

7.9

-1

2.1

-1

2.1

0.0

0.0

-1

8.1

-1

9.5

Mo

lise

1.3

0.0

-1

2.3

-1

2.3

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

10

.9-

11

.7-

11

.7

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

2.3

0.0

-2

3.3

-5

1.2

Pie

dm

on

t9

.10

.0-

12

.9-

12

.90

.00

.00

.00

.00

.00

.00

.00

.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

1.7

-2

6.5

-2

6.5

Sar

din

ia0

.01

.0-

33

.5-

34

.87

.30

.0-

25

.5-

43

.93

.30

.0-

29

.3-

51

.8

7.3

0.0

-2

5.5

-4

3.9

3.3

0.0

-2

9.3

-5

1.8

0.0

0.0

0.0

0.0

0.0

0.0

-2

4.4

-6

2.3

0.0

0.0

-3

2.7

-4

7.8

0.0

0.0

0.0

0.0

Sic

ily

0.0

5.4

-1

2.8

-1

2.8

16

.50

.0-

22

.2-

22

.27

.50

.0-

30

.2-

30

.2

8.6

0.1

-0

.2-

0.5

7.7

0.0

-3

0.3

-3

0.3

0.0

0.0

-3

7.4

-3

7.4

0.0

0.0

-3

4.1

-3

4.1

0.0

0.0

-3

7.0

-3

7.1

0.0

2.5

-5

0.4

-5

0.4

To

scan

y0

.00

.0-

0.8

-0

.81

1.2

0.0

-1

5.2

-1

5.2

14

.60

.0-

17

.8-

17

.8

12

.10

.0-

14

.5-

14

.51

9.6

0.0

-1

8.6

-1

8.6

1.8

0.0

-3

6.6

-3

6.6

0.0

0.0

-2

0.7

-2

0.7

4.9

0.0

-1

5.3

-3

8.3

0.0

0.0

-1

8.2

-1

8.2

Tre

nto

21

.90

.0-

19

.6-

19

.60

.30

.0-

31

.2-

31

.22

8.1

0.0

-1

5.4

-4

4.6

0.4

0.0

-3

1.2

-3

1.2

28

.10

.0-

15

.4-

44

.62

.00

.0-

17

.9-

17

.9

0.0

0.0

-1

7.5

-6

4.6

0.0

0.0

-3

4.0

-9

1.6

0.0

0.0

-3

4.1

-7

9.3

Um

bri

a0

.00

.0-

22

.3-

56

.50

.00

.00

.00

.00

.00

.0-

15

.3-

66

.6

0.0

0.0

0.0

0.0

0.0

0.0

-1

5.3

-6

6.6

2.5

0.0

-3

6.0

-9

0.9

0.0

0.0

-1

0.9

-6

9.6

0.0

0.9

-1

.2-

1.2

0.0

0.0

0.0

0.0

Ven

eto

9.5

0.0

-6

.1-

6.1

4.8

0.0

-7

.7-

7.7

0.0

0.0

-9

.0-

9.0

4.1

0.0

-4

.5-

4.5

0.0

0.0

-4

.4-

19

.20

.00

.6-

1.2

-1

.2

0.0

0.0

-1

.5-

45

.80

.90

.0-

5.1

-5

.10

.00

.00

.00

.0

Ital

y4

.83

.9-

12

.7-

18

.86

.70

.7-

13

.5-

18

.47

.90

.2-

13

.9-

29

.8

6.2

0.0

-1

2.4

-1

8.0

8.4

0.0

-1

3.9

-3

0.5

2.1

1.9

-1

8.4

-3

0.5

0.0

0.0

-1

3.5

-3

8.9

0.3

1.3

-1

7.1

-3

0.4

0.1

0.2

-2

2.2

-3

2.7

J Prod Anal

123

(represented by four and five-stars hotels) improves the

economic efficiency of the hospitality sector.

Finally, the Ordinary Least Squares (OLS) estimates of

Model (1) and (3) are presented in Table 5, Column (2) and

(4), respectively. None of the coefficients are statistically

significant except for CV_NRU (0.004). These empirical

results highlight that post-DEA analysis is very sensitive to

the approach used (see Simar and Wilson 2011, for an

extensive discussion). By ignoring the property of the multi-

step procedure by Simar and Wilson (2007), one could obtain

biased estimates that lead to wrong conclusions.

7 Spatial econometric analysis

Spill-over effect investigation of aggregated, or individual,

efficiency is a new branch of research in the empirical

analysis. In a post-DEA framework, Helfand and Levine

(2004) and Sampaio de Souza et al. (2005) find spatial

effects due to the existence of some functional relationship

between DMUs’ efficiency in two distinct points in space.

In a similar way, in the present study, the existence of

spatial autocorrelation between regional DEA scores is

tested by applying the Moran’s I test. Spill-over effects can

operate through different channels. For instance, tourism

hot spots can lead to positive effects for neighbouring

places, increasing their technical efficiency. Furthermore,

technical efficiency can be spread through ‘‘learning by

watching’’ or best practice imitation. In this sense, the

presence of a cluster of homogeneous regions, character-

ized by similar level of technical efficiency, is expected.

The analysis consists of two phases: first, a DEA

approach is run for each year considered; second, the

Moran’s I test is performed in each year.

Table 6 reports the test statistics and the respective

associated p values. Empirical evidence of the presence of

spatial effects is rather mixed. In four cases out of five, the

null hypothesis of absence of spatial autocorrelation cannot

be rejected; only in the last year of the sample, spatial

effects between DMUs’ scores efficiency are detected. The

Moran’s I test is also performed on the average DEA scores

of the period, and still the null hypothesis of absence of

spatial autocorrelation cannot be rejected (see last row of

Table 6).

8 Discussion and conclusions

This paper has added to the literature by examining the

productivity performance in the hospitality sector in a more

sophisticated manner through the use of a novel mix of non-

parametric and parametric approaches. Based on a macro-

economic framework, a dynamic WDEA has been employed

to evaluate and monitor the productivity of the hospitality

sector in Italy, over the period 2000–2004, based on the

official data availability. Assessing pure technical efficiency

in this economic sector has an important role in regional

planning and policy evaluation. Tourism activity and, hence,

hospitality has a key role in the Italian economy, though still

little is known on its performance.

Specifically, this paper has investigated the economic

efficiency of the Italian hospitality sector at a regional

level. Although Italy may be regarded as one single tour-

ism destination, given its geographical, cultural and his-

torical features, it also offers differentiated tourism

products and services that require a disaggregated

investigation.

Table 5 Post-DEA results (n = 105)

Variables Pooled-truncated estimation (1) OLS (2) Pooled-truncated estimation (3) OLS (4)

Coefficients Confidence intervals Coefficients Coefficients Confidence intervals Coefficients

Constant -28.73*** (99 %) = [-62.64, -27.69] 0.88*** -9.16*** (99 %) = [-23.96, -9.13] 0.86***

NRU -0.05** (95 %) = [-0.12, -0.03] 0.002 -0.01 (95 %) = [-0.06, 0.01] 0.002

CV_NRU 0.05*** (99 %) = [0.01, 0.10] 0.004*** 0.03** (99 %) = [0.02, 0.07] 0.004***

DART§ -1.56*** (99 %) = [-3.18, -1.56] -0.02 -1.26*** (99 %) = [-2.54, -0.93] -0.02

HQ_HOTELS -0.07** (95 %) = [-0.18, -0.05] 0.002

(1) Number of iteration = 10,000; (2) § reference group: no well-known art cities; (3) ** and *** indicate significance at the 5 and 1 %,

respectively

Table 6 Results of Moran’s I test

Year Moran I statistics p value

2000 0.078 0.130

2001 -0.031 0.633

2002 -0.083 0.427

2003 -0.051 0.505

2004 0.219 0.015

Average (00–04) 0.083 0.117

DEA scores are calculated with variable returns to scale and output

oriented

J Prod Anal

123

Empirically, the non-parametric WDEA has provided a

comparison of a DMU with respect to its own past per-

formance as well as the performance of the peer group.

Overall, the Italian regions show a relevant economic

inefficient performance throughout the period under

investigation. These findings are in line with other empir-

ical investigations showing that the hotel sector is charac-

terised by a lower productivity growth compared to other

businesses. This outcome is due, amongst other factors, to

the distinctive features of the hotel sector, which employs a

highly intensive labour productive process and is charac-

terised by excessive fixed costs, as well as a volatile

demand (Brown and Dev 2000; Kilic and Okumus 2005;

Goh 2010). On balance, the relatively most efficient region

is Lombardy, whose capital is Milan, which is one of the

most important centres of Italian finance and business

activity and is able to attract a relevant share of tourism

arrivals, particular foreigners with a high willingness to

pay, as reported in Federalberghi-Mercury (2005).

The key potential improvements that the hospitality sector

in Italian regions can achieve are based on the decrease of

labour and capital costs, though over time efficiency

improvements have been gained in terms of sales revenue

and added value. However, from the pooled-truncated ana-

lysis, it has emerged that various factors also influence the

level of Italian hospitality inefficiency. Poor utilisation of

infrastructure and high seasonal volatility have appeared to

be the main sources of economic inefficiency, while the share

of high quality hotels positively affects the sector efficiency.

This outcome has important implications. On the one hand,

the hotel quality variable can be thought as a proxy of the

overall environment, since it is plausible that 3–5 star hotel

accommodation are located in areas with a high level of

natural amenities; on the other hand, this finding links to

another empirical investigation run for the region of Sardinia

that shows how tourism demand is supply quality-driven in

the long run (Biagi and Pulina 2009). If one restricts the

attention to the Southern Italian regions, these results also

seem to support the work by Cracolici and Nijkamp (2009),

where Sardinia appears to be the region that has achieved the

best performance in comparison with its key competitors.

Environmental quality and a sustainable expansion of tour-

ism infrastructures turn out to be an important mix of

determinants which are not only able to drive tourism

demand, but are also likely to influence firms’ productivity.

The econometric investigation has further highlighted

that only well-known art cities, namely Florence, Rome

and Venice, drive the economic performance of the hos-

pitality sector in their regions, which are likely to be able to

compete more in local, internal and external markets.

Interestingly, the Moran’s I test has shown the lack of spill-

over effects amongst Italian regions. However, it is possi-

ble that spatial disaggregation matters. In fact, this outcome

is likely due to the loss of information during the process of

spatial data aggregation. Furthermore, the use of the

inverse of the distance between the centroids as a spatial

weight does not allow one to account for cross border

effects among tourist hot spots placed in different regions.

Future research may involve the use of provincial data,

characterised by a higher level of spatial disaggregation,

although it has to be said that such a data is rather difficult

to gather for a wider span of time.

Overall, the present findings have underlined important

features of the Italian hospitality sector. While the supply

of new infrastructure has continued to grow, Italian regions

have shown to be economically inefficient. Such an out-

come has also confirmed the empirical findings by Suzuki

et al. (2011) for the Italian provinces, although in that case

a statistic DEA is employed for the year 2001. The over-

investment, mainly designed to satisfy the high seasonal

demand, is one of the main sources of inefficiency. This

may turn into a large source of inefficiency and lack of

competitiveness. Additionally, the work force has less

incentive to perform efficiently since it is employed for

short periods of time throughout the year.

Although most of the Italian regions are characterised by

cities with an outstanding arts and historical heritage (e.g.

Bari, Bologna, Genoa, Palermo, Turin), this unique cultural

capital is not fully exploited as the driver of tourism supply

productivity. One of the reasons may be the low-profile

marketing campaigns, which are mostly run at a regional

level rather than by a central body able to activate a

comprehensive path of growth in decentralised areas and

less-known arts cities. Such a hypothesis has been further

confirmed by the spatial heterogeneity test, which has

shown no spill-over effects amongst the Italian regions,

confirming that Italian regions lack adequate networking

and joint marketing strategies. Ultimately, such a disen-

tangled development is likely to undermine the perfor-

mance of the less-known regions.

The present results are especially relevant for policy

makers and marketing managers, who should consider

synergic and complementary tourism policy for Italy as a

whole. This paper has drawn attention to the need to

reconsider an all-nation strategic approach to the tourism

activity. Although it is facing a long run life cycle, Italy is

endowed with an outstanding environmental and cultural

heritage that makes it inimitable, enabling it to operate in

the market place as one of the big players. Despite this

comparative advantage, Italy records a disharmonic growth

and a very heterogeneous performance at a regional level.

In a dynamic and globalised economy, characterised by

emerging destinations, there is the need to establish a

strong central body with a clear vision of the market

challenges, which may provide a substantive boost to

regain and maintain its actual position.

J Prod Anal

123

However, globalisation can not erase regional historical

and cultural roots. A comprehensive marketing strategy

should take into account the heterogeneous offer within the

Italian regions. One and a half century after the Italian

unification, the country is still characterised by ‘‘States’’ in

a ‘‘State’’ with their own arts, architecture, history, dialects

and enogastronomy and even secessionism ideologies. Yet,

this cultural diversity represents an asset for what is called

localnomics (or glocalisation). As argued by Salazar

(2010), localnomics is increasingly recognised to be a

winning strategy. In fact, the globalisation process of

tourism leads to an inevitably degree of worldwide inte-

gration and homogenisation in training, service and hos-

pitality benchmarks. However, tourism destinations seek to

diversify from the others in order to be more competitive

(Chang 1999). Such local differentiation, for instance

based on locally distinctive cultural identity and geo-

graphical peculiarity, can help to devise a solid strategy.

In this respect, Italian tourism, while developed and

distributed in a global market as a ‘‘unique product’’, still

needs to be tailored to accommodate heterogeneous con-

sumers’ preferences, allowing each region, and even pro-

vincial areas, to experience a renewal phase thanks to their

distinctive assets. Indeed, Italian tourism as an economic

activity, which has revealed to be more resilient to the

ongoing crisis than other economic sectors, offers an extra

advantage to competition if properly managed.

The dynamic non-parametric framework that has been

adopted in this study, combined with a parametric

approach, has provided insight into tourism industry dif-

ferences among the Italian regions. Unfortunately, this

powerful technique can not mitigate some relevant WDEA

limitations, such as variables selection and transferability

of the results, although it does provide an empirical

framework that can be used as a benchmark for further

investigations in the field.

Furthermore, since the efficiency of a given DMU is

measured with respect to other DMUs in the sample, the

obtained efficiency is always relative. The present paper

has two important implications. Firstly, it is not possible to

compare the efficiency of Italian regions sector with other

European competitor regions. Secondly, although the pro-

posed econometric approach captures the factors driving

the differences in efficiency among Italian destinations, it

does not allow one to detect the strengths and weakness of

Italian tourism industry in face of world competition.

From a methodological point of view, an interesting

result relates to the sensitivity of the post-DEA approach

used. Empirically, it has emerged that by ignoring the

property of the multi-step procedure by Simar and Wilson

(2007) one obtains biased results. Hence, the methodology

presented in this paper, although it refers to a single

country, can be applied to other destinations to find

common features and a more comprehensive assessment of

hospitality productivity. Future research can include the

collection and analysis of ad hoc questionnaires in each of

the Italian regions at a microeconomic level to obtain rel-

evant information on different segments of hospitality

supply. This information can be used to measure the spill-

over effects for the different categories of hotels and to

estimate a plant capacity or optimal size of hotels given the

state of travel and tourism in Italy.

Acknowledgments Manuela Pulina and Claudio Detotto acknowl-

edge the financial support provided by the Banco di Sardegna

Foundation (Prot. 1713/2010.0163). Juan Gabriel Brida acknowledges

the financial support provided by the Free University of Bolzano,

projects: ‘‘L’efficienza delle imprese turistiche in Italia’’ and ‘‘The

Contribution of Tourism to Economic Growth’’. Manuela Pulina

acknowledges the financial support provided by the Free University of

Bolzano (SECS-P/01—Economia Politica). The views expressed here

are those of the authors.

References

Anselin L (1988) Spatial econometrics: methods and models. Kluwer,

Dordrecht

Assaf AG, Agbola FW (2011) Modelling the performance of

Australian hotels: a DEA double bootstrap approach. Tour Econ

17(1):73–89

Assaf AG, Cvelvar LK (2010) The performance of the Slovenian

hotel industry: evaluation post-privatisation. Int J Tour Res

12(5):462–471

Assaf A, Barros CP, Josiassen A (2010) Hotel efficiency: a

bootstrapped metafrontier approach. Int J Hosp Man

29(3):468–475

Assaf AG, Deery M, Jago L (2011) Evaluating the performance and

scale characteristics of the Australian restaurant industry. J Hosp

Tour Res 35(4):419–436

Balcombe K, Fraser I, Latruffe L, Rahman M, Smith L (2008) An

application of the DEA double bootstrap to examine sources of

efficiency in Bangladesh rice farming. App Econ

40(15):1919–1925

Banker PC, Charnes A, Cooper WW (1984) Some models for

estimating technical and scale inefficiencies in data envelopment

analysis. Man Sci 30(9):1078–1092

Barros CP (2005) Measuring efficiency in the hotel sector. Ann Our

Res 32(2):456–477

Barros CP, Alves FP (2004) Productivity in the tourism industry.

IAER, August, 10(3). Technical University of Lisbon, Portugal

Barros CP, Dieke PUC (2008) Technical efficiency of African hotels.

Int J Hosp Man 27(3):438–447

Barros CP, Santos CA (2006) The measurement of efficiency in

Portuguese hotels using data envelopment analysis. J Hosp Tour

Res 30(3):378–400

Barros CP, Botti L, Peypoch N, Robinot E, Solonandrasana B, Assaf

G (2011) Performance of French destinations: tourism attraction

perspectives. Tour Man 32(1):141–146

Bernard AB, Jensen JB, Redding SJ, Schott PK (2007) Firms in

international trade. Working Paper 13054, http://www.nber.org/

papers/w13054. National Bureau of Economic Research,

Cambridge

Bernini C, Guizzardi A (2010) Internal and location factors affecting

hotel industry efficiency: evidence from Italian business corpo-

rations. Tour Econ 16(4):883–913

J Prod Anal

123

Biagi B, Pulina M (2009) Bivariate VAR models to test granger

causality between tourist demand and supply: implications for

regional sustainable growth. Pap Reg Sci 88(1):1–14

Bimonte S, Brida JG, Pulina M, Punzo L (2012) Tourism and growth:

stories of the two continents, Chapter 16. In: Punzo L, Feijo C,

Puchet Anyul M (eds) Beyond the global crisis: structural

adjustments and regional integration in Latin America and

Europe. Routledge Studies in the Modern World Economy,

USA, pp 252–268

Brida JG, Deidda M, Pulina M (2012) Investigating economic

efficiency in Italy: a regional comparison. Int J Rev Manag

6(3/4):175–198

Brown JR, Dev CS (1999) Looking beyond RevPAR; productivity

consequences of hotel strategies. Corn Hot Rest Admin Quart

40(2):23–33

Brown JR, Dev CS (2000) Improving productivity in a service

business: evidence from the hotel industry. J Serv Res

2(4):339–354

Bruni ME, Guerriero F, Patitucci V (2011) Benchmarking sustainable

development via data envelopment analysis: an Italian case

study. Int J Environ Res 5(1):47–56

Chang TC (1999) Local uniqueness in the global village: heritage

tourism in Singapore. Prof Geogr 51(1):91–103

Charnes A, Cooper WW, Rhodes E (1978) Measuring the efficiency

of decision making units. Eur J Oper Res 2(6):429–444

Charnes A, Clark CT, Cooper WW, Golany B (1985) A develop-

mental study of data envelopment analysis in measuring the

efficiency of maintenance units in the US air forces. Ann Oper

Res 2(1):95–112

Cooper WW, Seiford LM, Tone K (2000) Data envelopment analysis:

a comprehensive text with models, applications, references and

DEA-Solver software. Kluwer, Boston

Cortes-Jimenez I, Pulina M (2010) Inbound tourism and long-run

economic growth of Spain and Italy. Curr Issue Tour 13(1):61–74

Cracolici MF (2008) Assessment of tourism competitiveness by

analysing destination efficiency. Tour Econ 14(2):325–342

Cracolici MF, Nijkamp P (2006) Competition among tourist destina-

tion. An application of data envelopment analysis to Italian

provinces. In: Giaoutzi M, Nijkamp P (eds) Tourism and

regional development: new pathways. Ashgate, Aldershot, UK,

pp 133–152

Cracolici MF, Nijkamp P (2009) The attractiveness and competitive-

ness of tourist destinations: a study of Southern Italian regions.

Tour Man 30(3):336–344

Cullinane K, Song D-W, Wang T-F (2004) An application of DEA

windows analysis to container port production efficiency. Rev

Net Econ 32:184–206

Diewert WE, Mendoza MNF (1995) Data envelopment analysis: a

practical alternative. UBC Departmental Archives 95-30, UBC

Department of Economics

Emrouznejad A, Barnett RP, Tavares G (2008) Evaluation of research

in efficiency and productivity: a survey and analysis of the first

30 years of scholarly literature in DEA. Soc-Econ Plan Sci

42(3):151–157

Farrell J (1957) The Measurement of productive efficiency. J R Stat

Soc 120(3):253–290

Federalberghi-Mercury (2005) Rapporto 2005 sul sistema alberghiero

in Italia. www.federalberghi.it. Accessed on 31 Jan 2013

Federalberghi-Mercury (2010) Sesto rapporto sul sistema alberghiero

in Italia. www.federalberghiit. Accessed on 10 Jan. 2013

Federalberghi-Mercury (2012) DATATUR Trend e statistiche

sull’economia del turismo. Federalberghi & Format, Roma

Goh HN (2010) Recommending a productivity model for Singapore

hotels: A critical review of productivity models adopted by

researchers and hotel operators. UNLV Theses/Dissertations/

Professional Papers/Capstones. Paper 685

Griffith DA (2003) Spatial autocorrelation and spatial filtering.

Springer, Berlin

Helfand SM, Levine ES (2004) Farm size and the determinants of

productive efficiency in the Brazilian Center-West. Agric Econ

31(2/3):241–249

ISTAT (2011) Sistema di indicatori territoriali. http://sitis.istat.it/sitis/

html/index.htm. Accessed 12 Aug 2011

Kilic H, Okumus F (2005) Factors influencing productivity in small

island hotels: evidence from Northern Cyprus. Int J Cont Hosp

Man 17(4):315–331

Koksal CD and Aksu AA (2007) Efficiency evaluation of A-group

travel agencies with data envelopment analysis (DEA): a case

study in the Antalya region, Turkey. Tour Manag 28(3): 830–834

Kravtsova V (2008) Foreign presence and efficiency in transition

economies. J Prod Anal 29(2):91–102

Mandl U, Dierx A, Ilzkovitz F (2008) The effectiveness and

efficiency of public spending. Eur Comm Econ Pap 301:1–36

Manera Erbina C, Garau Taberner J, Molina de Dios R (2010) The

tourism revolution in the Mediterranean, 1950–2005. Documen-

tos de Trabajo, DT-AEHE 1014:1–17

Melao N (2005) Data envelopment analysis revisited: a neophyte’s

perspective. Int J Manag Dec Mak 6(2):158–179

Min H, Min H, Joo SJ (2008) A data envelopment analysis-based

balanced scorecard for measuring the comparative efficiency of

Korean luxury hotels. Int J Qual Rel Man 25(4):349–365

Molina-Azorin JF, Pereira-Moliner J, Claver-Cortes E (2011) The

importance of the firm and destination effects to explain firmperformance. Tour Man 32(1):22–28

Moriarty JP (2010) Have structural issues placed New Zealand’s

hospitality industry beyond price? Tour Econ 16(3):695–713

Neves JC, Lourenco S (2009) Using data envelopment analysis to

select strategies that improve the performance of hotel compa-

nies. Inter J Cont Hosp Man 21(6):698–712

Olesen OB, Petersen NC (2002) The use of data envelopment analysis

with probabilistic assurance regions for measuring hospital

efficiency. J Prod Anal 17(1/2):83–109

Prado Lorenzo JM, Garcıa Sanchez IM (2007) Efficiency evaluation

in municipal services: an application to the street lighting service

in Spain. J Prod Anal 27(3):149–162

Pulina M, Detotto C, Paba A (2010) An investigation into the

relationship between size and efficiency of the Italian hospitality

sector: a window DEA approach. Eur J Oper Res 20(4):613–620

Reynolds D (2003) Hospitality-productivity assessment using data-

envelopment analysis. Corn Hot Rest Admin Quart 44(2):429–449

Reynolds D, Thompson GM (2007) Multiunit restaurant productivity

assessment using three-phase data envelopment analysis. Int J

Hosp Man 26(1):20–32

Robert WJA, Haug AA, Jaforullah M (2010) A two-stage double-

bootstrap data envelopment analysis of efficiency differences of

New Zealand secondary schools. J Prod Anal 34(2):99–110

Salazar NB (2010) The glocalisation of heritage through tourism:

Balancing standardisation and differentiation. In: Labadi S, Long

C (eds) Heritage and globalisation. Routledge, London,

pp 130–147

Sampaio De Souza MC, Cribari-Neto F, Stosic BD (2005) Explaining

DEA technical efficiency scores in an outlier corrected environ-

ment: the case of public services in Brazilian municipalities.

Braz Rev Economet 25(2):287–313

Shuai JJ (2009) Web content and its influence on operational

performance-case of the hotel industry. Industrial engineering

and engineering management, IEEM, international conference,

8–11 December 2009, pp 885–889. doi:10.1109/

IEEM20095372884

Shuai JJ, Wu WW (2011) Evaluating the influence of E-marketing on

hotel performance by DEA and grey entropy. Exp Syst Appl

38(7):8763–8769

J Prod Anal

123

Sigala M (2004) Using data envelopment analysis for measuring and

benchmarking productivity in the hotel sector. J Trav Tour Mark

16(2):39–60

Simar L, Wilson PW (2007) Estimation and inference in two-stage,

semi-parametric models of production processes. J Econom

13(6):31–64

Simar L, Wilson PW (2011) Two-stage DEA: caveat emptor. J Prod

Anal 36(2):205–218

Sirirak S, Islam N, Khang DB (2011) Does ICT adoption enhance

hotel performance? J Hosp Tour Tech 2(1):34–49

Sun DB (1988) Evaluation of managerial performance in large

commercial banks by data envelopment analysis. IC2 Institute,

Austin

Suzuki S, Nijkamp P, Rietveld P (2011) Regional efficiency

improvement by means of data envelopment analysis through

Euclidean distance minimization including fixed input factors:

an application to tourist regions in Italy. Pap Reg Sci

90(1):67–89

UNWTO (2012) World tourism barometer, vol. 10, February 2012.

http://www.unwto.org/. Accessed on 28 May 2012

Wang FC, Hung WT, Shang JK (2006) Measuring the cost efficiency

of international tourist hotels in Taiwan. Tour Econ 12(1):65–85

Wanhill S (2011) What do economists do? Their contribution to

understanding tourism. Estudios de Economia Aplicada

29(3):679–692

Wilson PW (2008) FEAR: a software package for frontier efficiency

analysis with R. Soc-Econ Plan Sci 42(4):247–254

WTTC (2010) Progress and priorities 2009–10 http://www.

confederacaoturismoportugues.pt/downloads/get/id/192. Acces-

sed on 14 Oct 2013

J Prod Anal

123