The impact of tourist destination on hotel efficiency : a data envelopment analysis approach

The impact of tourist destination on hotel efficiency: A data envelopment analysis approach Rubén Lado-Sestayo Business Department (Financial Economics and Accounting) University of A Coruña Faculty of Economics & Business, Campus de Elviña s/n, 15071 A Coruña, (Galicia, Spain) [email protected] Ángel Santiago Fernández-Castro Business Department (Financial Economics and Accounting) University of A Coruña Faculty of Economics & Business, Campus de Elviña s/n, 15071 A Coruña, (Galicia, Spain) [email protected] Corresponding Author: Rubén Lado-Sestayo; Business Department (Financial Economics and Accounting Area); University of A Coruña; Faculty of Economics & Business, Campus de Elviña s/n, 15071 A Coruña, (Galicia, Spain); email: [email protected]. ABSTRACT This paper evaluates the impact of location on hotel efficiency using a sample of 400 Spanish hotels, the novel aspect being that location is considered at the tourist destination level. Moreover, for the first time, the location variables are based on the main theoretical models concerning location in the hotel sector, namely geographical positioning models, agglomeration and urbanization economic models and competitive environment models. The methodology consists of a four-stage data envelopment analysis (DEA) model that decomposes superefficiency in the portion attributable to the tourist destination and the portion attributable to hotel management. Then, managerial efficiency is regressed against hotel characteristics, while tourist destination efficiency is explained by the characteristic of each location. The findings highlight the importance of tourist destinations, providing novel empirical support for the propositions of the main location models. Indeed, the tourist destination is the main cause of differences in the level of efficiency among hotels. The occupancy level, degree of seasonality and market concentration are the variables with the greater impact on efficiency. Keywords Data Envelopment Analysis; Efficiency; location; tourist destination; hotel. Highlights • This paper discusses location effects in the analysis of hotel efficiency. • The proposed four-stage model is a novel means of analysing hotel efficiency. • Tourist destination variables are the main cause of differences in efficiency. 1. Introduction The study of the tourist destination is of special importance for the hospitality industry because its product can only be consumed where it has been generated (Bull, 1994). Moreover, the tourist destination has been shown to have a significant impact on hotel profitability and survival (Lado-Sestayo, Vivel-Búa & Otero-González, 2016a). However, despite the importance of the characteristics of the tourist destination, they have been neglected in studies of efficiency in the hotel sector, hindering understanding of the extent to which inefficiencies are due to hotel management or to the characteristics of the location at the tourist destination level (Barros, 2005). Knowledge of the causes of inefficiencies is important for benchmarking purposes as it allows identification of the scope for improvement for both hotel and tourist destination managers (Huang, Mesak, Hsu, & Qu, 2012). This paper analyses the efficiency of 400 Spanish hotels, considering the characteristics of the tourist destination and internal factors related to the firm. From a theoretical perspective, there are three approaches to studying the impact of location on the hotel sector: one line of research is focused on geographical positioning variables; a second line is concerned with the study of the existence of externalities; the third approach analyses the impact of the competitive environment (Yang, Luo, & Law, 2014). These approaches have not yet been considered in the study of hotel efficiency, so filling this gap is the first main objective of this paper. Most studies on efficiency in the hotel sector have focused on firm characteristics, usually dealing with hotels from one single market, with some exceptions (e.g. Assaf, 2012; Neves & Lourenço, 2009), and using small sample sizes (Oliveira, Pedro, & Marques, 2015; Wöber, 2007). In addition, the few recent studies which have considered some external aspects of hotels, mostly environmental variables, highlight the existence of differences in efficiency levels among locations. However, these studies have focused on regional characteristics and have not incorporated factors at the tourist destination level (De Jorge & Suárez, 2014; ParteEsteban & Alberca-Oliver, 2015; Pulina, Detotto, & Paba, 2010). It is common for a region to have several tourist destinations, so the impact of the location might be different within the same region. For example, in Spain there are 17 regions and 97 tourist destinations. Moreover, some of these tourist destinations are in the same region, but they are very different regarding seasonality, accessibility and other characteristics (rural vs. urban, coastal vs. inland, etc.). Therefore, a regional empirical study may not be sufficient to evaluate the impact of location on efficiency. Thus, unlike previous studies which have addressed regional effects, the second main objective of this paper is to evaluate the impact of location factors at the tourist destination level. Methodologically, data envelopment analysis (DEA) is used to provide a synthetic indicator of efficiency, considering sales revenue as the output variable, while the inputs are labour costs, depreciation and operational costs. This paper makes two main methodological contributions. First, it proposes a four-stage DEA model to break down efficiency into the aspects attributable to tourist destination and to hotel management using the concepts of “programme” and “managerial” efficiency (Charnes, Cooper, & Rhodes, 1981). No previous study has applied this methodology to analyse the impact of location on efficiency in the hotel sector. The second contribution is related to the measurement of efficiency through “superefficiencies”, extending the efficiency rankings also to efficient units, unlike ordinary efficiency scores (which assign the same scoring to all efficient units). Super-efficiencies make it possible to explain efficiency through linear models with straightforward coefficients, avoiding the Tobit models imposed by ordinary efficiency scores (as truncated variables). Furthermore, as superefficiencies are particularly helpful in identifying potential outliers, their use becomes more appealing when large samples are used, as is the trend in recent literature. The second stage in the empirical analysis is the study of the determinants of hotel efficiency. In particular, a disaggregated regression model explains the determinants of the two components mentioned above, studying the incidence of tourist destination variables with regard to the efficiency attributable to hotel location and the impact of hotel-specific characteristics on managerial efficiency. This disaggregated model is compared against a traditional synthetic model – the most popular methodology in previous studies – explaining the overall efficiency. To summarize, the contributions of this paper are as follows: first, it is novel in incorporating hotel location models in the study of hotel efficiency, allowing the benchmarking of both hotels and tourist destinations. Location is considered at the tourist destination level, using a sample representative of multiple locations in Spain, the second largest market in the world by number of visitors. Second, it proposes a novel methodology employing a four-stage DEA model to decompose and explain the effects of internal and external factors on hotel efficiency, isolating the impact of tourist destination variables, which have not been examined in previous studies. Furthermore, this paper considers the super-efficiency approach, which is particularly helpful when dealing with large samples. The paper is structured in four sections. Following this introduction, section 2 provides the framework through a literature review and explains the basic DEA methodology. Section 3 presents the four-stage DEA model and its empirical application, analysing hotel efficiency estimated and determining factors. Finally, the main conclusions and implications are provided in section 4. 2. Framework 2.1 Previous literature: location and hotel efficiency The study of efficiency in the hotel sector is quite recent (Wöber, 2007) and has focused on quantifying the level of efficiency, with few investigations that also evaluate its determinants. After pioneering work by Morey and Dittman (1995) on the United States (US) market, the first studies were characterized by small sample sizes and a focus on a single tourist destination within a country, especially the US (Anderson, Fish, Xia & Michello 1999; Barros, 2005), but also in other countries (e.g. Taiwan in Hwang & Chang, 2003). Later on, research was carried out on Asian markets, such as China, Japan and Korea (Honma & Hu, 2012; Min & Joo, 2009), European markets, such as Italy, France and Spain (Parte-Esteban & Alberca-Oliver, 2015), and African countries, such as Angola (Barros & Dieke, 2008). From a multi-country perspective, Neves and Lourenço (2009) analysed two international firms with hotels in different countries and Assaf (2012) studied a sample of Asian hotels and tour operators. Table 1 identifies the most recent studies analysing the efficiency levels in the hotel sector. Three aspects stand out as the main contributions of these works: i) the use of higher sample sizes relative to the first studies, favouring their representativeness; ii) the implicit recognition of differences between locations, highlighting the need for studies that examine the effect of location on efficiency in the hotel sector in depth (Parte-Esteban & Alberca-Oliver, 2015); iii) the incorporation of methodological innovations, evaluating not only the level of efficiency but also its determinants and the temporal dimension. Table 1 Recent studies on the level of hotel efficiency using the DEA methodology. Author(s) Sample Variables No. of hotels/Market/Period Inputs Outputs Oliveira, Pedro, and da Cunha Marques (2015) 28/Portugal/2005–2007 Number of rooms; number of employees; F&B capacity; labour costs; capital costs; other costs Total revenues Parte-Esteban and Alberca-Oliver (2015) 1385/Spain/2001–2010 Number of full-time employees; book value of property; operational costs Sales De Jorge and Suárez (2014) 303/Spain/1999–2007 Employment; labour costs; number of rooms; operational costs Sales; market share Alberca-Oliver and Parte-Esteban (2013) 1593/Spain/2001–2008 Number of full-time employees; property book value; operational costs Total revenue Oliveira, Pedro, and Da Cunha Marques (2013a) 28/Portugal/2005–2007 Number of rooms; number of employees; number of seats F&B; other costs; capital expenditure Total revenue; price of rooms; price of F&B Oliveira, Pedro, and Da Cunha Marques (2013b) 28/Portugal/2005–2007 Number of rooms; number of employees; number of seats F&B; other costs Total revenue Assaf (2012) 192 hotels and 65 tour operators/12 Asia Pacific countries/2007–2009 Fixed capital; number of full time employees; other operational costs Total revenues Honma and Hu (2012) 15/Japan/2004–2008 Number of employees; number of temporary staff; number of seats in restaurants and bars; number of guest rooms Total revenues Huang, Mesak, Hsu, and Qu (2012) 31 regions/China/2001–2006 Number of full-time employees; number of guest rooms; total fixed assets Total revenue; average occupancy rate Barros, Botti, Peypoch, Robinot, Solonandrasana, and Assaf (2011) 22 regions/France/2003–2007 Tourist arrivals; accommodation capacity (hotels and camping) Bed nights Shuai and Wu (2011) 48/Taiwan/2006–2007 Total number of rooms; number of full-time employees; operating expenses Room revenues; F&B revenues Wu, Tsai, and Zhou (2011) 23/Taipei/2006 Total number of rooms; total number of employees; F&B capacity; total operating cost Room revenues; F&B revenues; other revenues Hsieh and Lin (2010) 57/Taiwan/2006 Accommodation costs; number of employees in the accommodation department; catering costs; number of employees in the catering department Room revenues; catering revenues Hu, Chiu, Shieh, and Huang (2010) 66/Taiwan/1997–2006 Price of labour; price of F&B; price of other operations Room revenues; F&B revenues; other operating revenues Pulina, Detotto, and Paba (2010) 150 (21 regions)/Sardinia Island/2002–2005 (2000– 2002 at macro level) Labour costs; physical capital (only at hotel level) Sales revenue; value added Wu, Liang, and Song (2010) 23/Taipei/2002–2006 Number of rooms; number of employees; F&B capacity; total operating costs Room revenues; F&B revenues; other revenues Barros, Peypoch, and Solonandrasana (2009) 15/Portugal/1998–2004 Number of employees; physical capital Sales; added value Botti, Briec, and Cliquet (2009) 15 hotel chains/France/1997 Costs; territory coverage; chain duration Sales Min and Joo (2009) 31/Korea/2003 Costs of land property; building capacity; other fixed assets; other current assets. Costs of goods sold; selling, general and administrative expenses; non-operating expenses Room revenues; F&B revenues; other revenues. Operating income; non-operating income Neves and Lourenço (2009) 83/different countries/2000– 2002 Current assets; net fixed assets; shareholders’ equity; cost of goods and services Total revenues; EBITDA Perrigot, Cliquet, and Piot-Lepetit (2009) 15 hotel chains/France/1999 Age of the hotel chain in years; number of rooms in the chain; number of hotel openings during the year; royalties in percentage; chain ranking Occupancy rate; total sales Yu and Lee (2009) 58/Taiwan/2004 Number of full-time employees in the room service department; number of full-time employees in the F&B department; number of rooms; floor area in the F&B department; total expenses for each service sector; number of back office staff Room revenues; F&B revenues; other revenues Barros and Dieke (2008) 12/Angola/2000–2006 Total costs; investment expenditure REVPAR Shang, Hung, and Wang (2008) 57/Taiwan/2005 Number of rooms; F&B capacity; number of full-time employees; operating expenses Room revenues; F&B revenues; other revenues Chen (2007) 55/ Taiwan/2002 Price of labour; price of F&B; price of materials Total revenues Notes: F&B denotes food and beverages. A review of the literature prior to 2007 can be found in Wöber (2007). Regarding the larger sample size, Pulina et al. (2010) pioneered such studies, using a sample of 150 hotels on the island of Sardinia. This effort to increase the sample size was followed by Assaf’s (2012) study of 192 Asian hotels, De Jorge and Suarez’s (2014) work including 303 Spanish hotels and Parte-Esteban and Alberca-Oliver’s (2013, 2015) research with a sample of 1593 Spanish hotels, the largest sample in the analysis of efficiency in the hotel sector so far. However, the greater representativeness of hotel characteristics derived from larger sample sizes is achieved at the expense of the inclusion of hotels from a wider variety of tourist destinations within a country. This heterogeneity enhances the importance of the factors of location and the need to study their impact on hotel efficiency in greater depth. It is important to consider that the hotel sector is influenced to a large extent by the characteristics of the location. Indeed, previous literature has shown that the characteristics of the tourist destination significantly affect hotel profitability and survival (Lado-Sestayo et al., 2016a, 2016b; Yu & Lee, 2009). Despite this, none of the previous studies considered variables related to the tourist destination in the study of hotel efficiency. On the one hand, considering hotels in a single tourist destination, as was usual in the first works on hotel efficiency, prevents the generalization of results. On the other hand, considering hotels in different tourist destinations using a larger sample size, as in recent studies, may bias the results if the effects of location are not properly isolated. As mentioned above, some recent studies have found important differences in efficiency levels among hotels located in different regions within a country (Parte-Esteban & Alberca-Oliver, 2015). For example, Pulina et al. (2010) found considerable differences in efficiency levels among Italian regions in an analysis of the effect of hotel size on efficiency. Shang, Wang, and Hung (2010) found differences in efficiency levels among hotels located in cities or metropolitan areas and hotel resorts in Taiwan. However, no previous study has considered the tourist destination effect. In this context, the main contributions of this paper are that it breaks down efficiency into the portions attributable to location at the tourist destination level and to hotel management and analyses the determinants of these two components separately. Focusing on the determinants of the efficiency level, only three previous studies considered the effect of location, but these did so at the regional level. Huang et al. (2012) analysed 31 Chinese regions (22 provinces, 5 autonomous regions and 4 autonomous municipalities) and emphasized the need for further studies on other countries to analyse differences in the main determinants. The location variables used by these authors were: the percentage of national A grade tourist attractions, the ratio of inbound arrivals received by a particular region to the total inbound arrivals to China, the educational attainment of the urban employed, the average annual earnings of employees in the Chinese hotel industry, the number of hotels and trade openness. For the Spanish market, the pioneering studies considering hotels from different tourist destinations are very recent. De Jorge and Suárez (2014) used regional dummies and the market concentration of the hotel with respect to its four main competitors. Parte-Esteban and Alberca-Oliver (2015) where ∝ is a constant term, the subscript i represents the hotel and j is the tourist destination; H are the hotel characteristics variables, L are the tourist destination variables and D is a dummy variable of tourist destination effects. To the best of our knowledge, none of previous works dealing with the explanation of efficiency scores has applied the decomposition of efficiency into its managerial and location components. This decomposition has been justified on the basis of the literature about hotel spatial distribution and its usefulness is tested through the empirical model. 3.2 Hotel efficiency at the tourist destination level This paper evaluates the efficiency of 400 hotels located in the 97 Spanish tourist destinations in 2011. The Alimarket database was used to collect data on hotel characteristics and the SABI database was used to collect the accounting information of individual hotels. Data on the 97 Spanish tourist destinations were obtained using the National Institute of Statistics (NSI) database. Finally, information on the position of transport nodes was identified in EuroGeographics. The variables selected to evaluate the efficiency level correspond to those most commonly used in DEA analysis in the hotel sector, as shown in the literature review in Table 1: sales revenue as output and depreciation, labour costs and operational costs as inputs. An extensive review of previous studies using these variables can be found in De Jorge and Suárez (2014). Special attention was paid to the coherence among variables, particularly to avoid mixing variables in absolute values with ratios. A descriptive analysis of the variables used in the DEA analysis can be found in Table 2, in which a high level of correlation between all variables can be observed. This supports the adequacy of the variable selection as a positive relationship between input and output is expected. In addition, the correlations between the inputs seem to indicate that the productive structure does not differ significantly between the hotels. Table 2 Descriptive analysis of inputs and outputs. Variables Correlation matrix Descriptive statistics Sales revenue (O) Labour costs (I) Depreciation (I) Operational costs (I) Mean SD N Sales revenue (O) 1 0.960 0.706 0.857 2,141.605 3,162.565 400 Labour costs (I) 0.960 1 0.672 0.798 847.650 1,129.456 400 Depreciation (I) 0.706 0.672 1 0.776 197.568 301.189 400 Operational costs (I) 0.857 0.798 0.776 1 422.033 645.168 400 Notes: O denotes the output and I the input. SD is the standard deviation. N is the number of observations. Fig. 4 shows the overall efficiency for each hotel calculated in the DEA analysis (y axis) and the tourist destinations (x axis). At first sight, important differences between tourist destinations can be observed. There are also considerable differences in the levels of efficiency between tourist destinations in the same region (NUTS II). For instance, Lloret del Mar has an average efficiency level of around 50%, while Barcelona has 80%, but both are in the Catalonia region. Moran's I test (with a value of 5.988) confirms that spatial autocorrelation is statistically significant (p-value < 0.001). The null hypothesis, which is the existence of a random spatial distribution of the variable studied, is rejected at the 99% significance level using an inverse distance matrix which incorporates all hotels in the sample. The test was run for different neighbourhood matrices, rejecting the null hypothesis in all cases. These results support the existence of location effects on hotel efficiency, which should be analysed at the tourist destination level. The next section addresses this issue. Hotel efficiency level (∅𝑖𝑗) Average efficiency level in tourist destination j Fig. 4. Hotel efficiency and average efficiency level in tourist destination Figure 5 shows the differences between the results of the disaggregated (four-stage) DEA model and the previous aggregate model. As shown in the figure, decomposition increases the median value of hotel super-efficiencies from 52% to 90%, as could be expected because the location effect is expected to be completely removed. Moran’s I test for the hotel superefficiencies in the disaggregated model gets a value of 0.612, with a P-value of 0.270. This shows that managerial efficiency is not spatially correlated, i.e., that the effect of the tourist destination has been removed through the disaggregated model. Thus, hotel characteristics should be used to explain managerial efficiency, while the characteristic of each tourist destination should be used to explain tourist destination efficiency. Efficiency level (%) Tourist destinations Methodology Moran I P-value Aggregated model 0.612 0.270 Disaggregated model 5.988 <0.001 Fig. 5. Median and interquartile range of hotel efficiency in the aggregated and disaggregated models. 3.3 Determinants of hotel efficiency at the tourist destination level The determinants of efficiency are classified in two groups, those corresponding to hotel characteristics and those corresponding to tourist destination (Table 3). It should be noted that there is no unanimity in the selection of determinants of hotel efficiency. Many researchers have selected them solely based on previous studies. This paper not only uses traditional determinants, but also tourist destination variables which have not been evaluated previously (Table 3). These have been selected using the theoretical framework of the three location models developed in the hotel sector. First, geographical positioning models emphasize centrality and proximity to transport nodes as relevant aspects of hotel service. Thus, those tourist destinations with better accessibility will enjoy a competitive advantage over other locations, generating higher incomes for incumbent hotels and consequently increasing efficiency. To include this aspect, the distance between the nearest international airport and the tourist destination is considered (Ashworth & Tunbridge, 1990; Egan & Nield, 2000). This measure was considered by Lee and Jang (2011) to relate hotel distance to international airport and price. Regarding efficiency analysis, Honma and Hu (2012) found that hotel distance to international airports has a negative effect on hotel global efficiency using DEA methodology in their analysis of Japan’s major hotel companies. Hence, the first hypothesis is formulated as follows: Hypothesis 1. The accessibility of the tourist destination has a positive impact on efficiency. Second, agglomeration models highlight the importance of externalities arising from agglomeration and urbanization economies. The concentration of economic activity and especially of hotel activity favours knowledge exchange, as well as the development of the leisure sector. From this perspective, those hotels located in tourist destinations with high agglomeration will take advantage of positive externalities, increasing their efficiency level. The existence of agglomeration economies is quantified using population density as an indicator of economic activity concentration in the tourist destination (Canina et al., 2005). Luo and Jang (2016) found a positive relation between this variable and hotel profitability, so a similar effect may be expected regarding efficiency analysis. Also, Peiró-Signes et al. (2015) found a positive effect of agglomeration on revenue per available room. So, the corresponding hypothesis is: Hypothesis 2. Agglomeration in the tourist destination has a positive impact on efficiency. Third, competency models pay special attention to differentiation and to the competitive environment. From this perspective, the relative position of the hotel plays a key role in explaining the profit margin. Thus, those tourist destinations with high levels of competition, associated with low market concentration, will have lower profitability (Porter, 2008). On the other hand, tourist destinations with a high market concentration favour collusive practices, increasing the profit margin (Lado-Sestayo et al., 2016b). Consequently, a higher level of market concentration has a positive impact on the level of hotel efficiency. Market concentration was considered at the regional level by De Jorge and Suárez (2016) in the analysis of efficiency. Also, Lado-Sestayo et al., (2016b) found a positive relation between market concentration and profit margin at the tourist destination level. Therefore, the hypothesis is: Hypothesis 3. Market concentration in the tourist destination has a positive impact on efficiency. In addition to hotel location models, the level of demand, the occupancy rate and the level of seasonality at the tourist destination level are included as control variables because they are also expected to affect the level of efficiency. For example, Parte-Esteban and Alberca-Oliver (2015) found that occupancy, demand level and coastal character have an impact at the regional level on hotel global efficiency. The determinants of hotel characteristics reflect the main aspects of hotel activity. These include hotel size, which Honma and Hu (2012) found to have positive impact on efficiency; space dedicated to conference rooms, according to the higher efficiency of resort hotels in contrast with metropolitan ones found by Parte-Estaban and Alberca-Oliver (2015); market share, since Barros and Dieke (2008) found a positive effect of market share on efficiency; quality (number of stars), following De Jorge and Suárez (2014); management agreements with hotel chains, to confirm the results of Wang et al. (2006) and Chen (2007); and the distance of the hotel to the centre, according to the positive effect of centrality on TrevPAR found by Sainaghi (2011). Table 3 Determinants of hotel efficiency considering hotel characteristics and tourist destination characteristics. Variable [label] Definition Hotel characteristics Size Assets [size] 𝒕𝒐𝒕𝒂𝒍 𝒂𝒔𝒔𝒆𝒕𝒔 𝒊𝒏 𝒕𝒉𝒐𝒖𝒔𝒂𝒏𝒅 𝒆𝒖𝒓𝒐𝒔𝒊 Market orientation Meeting space capacity [conference] 𝒔𝒑𝒂𝒄𝒆 𝒄𝒂𝒑𝒂𝒄𝒊𝒕𝒚 𝒎𝟐𝒊 Share Market share [share] 𝑹𝒆𝒗𝒆𝒏𝒖𝒆𝒔 𝒐𝒇 𝒉𝒐𝒕𝒆𝒍𝒊 𝑻𝒐𝒕𝒂𝒍 𝒓𝒆𝒗𝒆𝒏𝒖𝒆𝒔 𝒊𝒏 𝒕𝒐𝒖𝒓𝒊𝒔𝒕 𝒅𝒆𝒔𝒕𝒊𝒏𝒂𝒕𝒊𝒐𝒏𝒋 Quality Star rating [stars] 𝑵𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒔𝒕𝒂𝒓𝒔 𝒊 Management Management agreements [chain] 𝑫𝒖𝒎𝒎𝒚 𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆 𝒐𝒇 𝒆𝒙𝒊𝒔𝒕𝒆𝒏𝒄𝒆 𝒐𝒇 𝒎𝒂𝒏𝒂𝒈𝒆𝒎𝒆𝒏𝒕 𝒂𝒈𝒓𝒆𝒆𝒎𝒆𝒏𝒕𝒊 Centrality Centrality [dist_CBD] 𝐃𝐢𝐬𝐭𝐚𝐧𝐜𝐞 𝐭𝐨 𝐭𝐡𝐞 𝐭𝐨𝐮𝐫𝐢𝐬𝐭𝐢𝐜 𝐝𝐞𝐬𝐭𝐢𝐧𝐚𝐭𝐢𝐨𝐧 𝐢𝐧 𝐤𝐦 𝒊 Tourist destination Accessibility Accessibility [dist_airport] ∑𝑫𝒊𝒔𝒕𝒂𝒏𝒄𝒆 𝒇𝒓𝒐𝒎 𝒕𝒉𝒆 𝒄𝒍𝒐𝒔𝒆𝒔𝒕 𝒂𝒊𝒓𝒑𝒐𝒓𝒕 𝒊𝒏 𝒌𝒎 𝒊𝒋 𝐈𝒊=𝟏 𝐈 Agglomeration Population density [urban] 𝑷𝒐𝒑𝒖𝒍𝒂𝒕𝒊𝒐𝒏 𝒅𝒆𝒏𝒔𝒊𝒕𝒚 𝒋 Competition Market concentration [hhi] 𝑳𝒏(𝑯𝒆𝒓𝒇𝒊𝒏𝒅𝒂𝒉𝒍 𝑰𝒏𝒅𝒆𝒙)𝒋 Occupancy Occupancy level [occu] ∑𝑴𝒐𝒏𝒕𝒉𝒍𝒚 𝒂𝒗𝒆𝒓𝒂𝒈𝒆 𝒐𝒄𝒄𝒖𝒑𝒂𝒏𝒄𝒚𝒋𝒎 𝟏𝟐 𝒎=𝟏 𝟏𝟐 Demand Demand level [visitors] 𝑵𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒐𝒖𝒓𝒊𝒔𝒕𝒔 𝒊𝒏 𝒕𝒉𝒆 𝒕𝒐𝒖𝒓𝒊𝒔𝒕 𝒅𝒆𝒔𝒕𝒊𝒏𝒂𝒕𝒊𝒐𝒏 𝒋 Seasonality Degree of seasonality [season] 𝛔𝟐𝐣(∑𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒐𝒄𝒄𝒖𝒑𝒂𝒕𝒊𝒐𝒏𝒎 𝟏𝟐 𝒎=𝟏 𝟏𝟐 ) Notes: i represents each firm; t represents the time period; j represents each tourist destination; m represents each month of the year (used when monthly data are available). Table 4 contains a descriptive analysis of the variables selected as potential determinants of efficiency. Focusing on variables related to tourist destinations, the results show important differences among these areas regarding population density, demand level, market concentration and accessibility. At the hotel level, the main differences are observed in meeting space capacity and the existence of management agreements. According to the variance inflation factor (VIF), there are no multicollinearity problems between the variables, so they can be incorporated in the model together. Finally, the correlations between tourist destination efficiency scores and the tourist destination variables are much higher than the correlations of efficiency scores and explanatory variables at the hotel level, reinforcing the importance of considering spatial effects. Table 4 Descriptive analysis of efficiency determinants Dimension Variables Correlation Descriptive statistics Score hotel Score tourist destination Mean SD VIF N Hotel size -0.031 - 5.104 12.018 1.82 400 conference -0.022 - 78.745 287.389 1.16 400 share 0.141 - 2.085 3.252 2.32 400 stars -0.024 - 3.105 0.843 1.23 400 chain 0.043 - 0.085 0.279 1.14 400 dis_cbd 0.112 - 21.947 28.440 1.18 400 Tourist destination dist_airport - -0.372 42.242 35.917 1.47 400 urban - 0.432 2.143 3.402 1.75 400 hhi - 0.527 0.035 0.031 2.88 400 visitors - 0.401 1.502 2.218 2.43 400 occu - -0.097 54.516 13.212 2.08 400 season - -0.329 0.258 0.089 1.58 400 Notes: SD is the standard deviation. VIF is the variance inflation factor. N is the number of observations. Table 5 presents the results of the proposed disaggregated model and the traditional synthetic model. Both models strongly confirm the impact of tourist destination variables on efficiency. Even the results of the aggregated model justify this assertion: four internal variables and three tourist destination variables have a statistically significant effect on overall efficiency. The proposed disaggregated model improves the explanatory power remarkably, both in terms of goodness of fit and in terms of the significance of the coefficients. It is worthy of note that in the analysis of the tourist destination component of efficiency all the location variables have a statistically significant impact. The variables associated with the theoretical framework of location models in the hotel sector are the core of this research and the proposed model confirms the three hypotheses. First, the negative effect of distance from transport nodes on efficiency is consistent with geographical positioning models (Hypothesis 1). Second, the positive effect of population density on efficiency is consistent with agglomeration models (Hypothesis 2). Third, the positive effect of market concentration is consistent with competency models (Hypothesis 3). The results also point out some other relevant effects. There is a positive effect of seasonality, which may be due to the positive effect on prices and the possibility of cost savings resulting from the concentration of activity in specific periods. The positive effect of management agreements with hotel chains is in line with the previous studies (Such-Devesa & Mendieta-Penalver, 2013). Regarding size, its complex effect may demand further study as absolute size has a negative influence (Parte-Esteban & Alberca-Oliver, 2015), while market share (relative size) has a positive impact. Table 5 Regression models Disaggregated model Aggregated model Model 1a Model 1b Model 2 Dependent variable Score Hotel Score Destination Score Global size -1.305*** -0.152* 0.445 (0.078) conference -0.001 0.002 0.008 (0.004) share 6.179*** 0.620** 2.370 (0.295) stars 0.730 -0.592 5.558 (0.957) dist_cbd 0.022 0.050** 0.184 (0.025) chain 40.797** 7.998** 18.211 (3.823) dist_airport -0.058* -0.017 (0.030) (0.021) hhi 3.513** -0.040 (1.733) (1.143) urban 0.562*** 1.476*** (0.203) (0.429) visitors 3.797*** 1.400*** (0.467) (0.493) occu 0.567*** 0.173** (0.092) (0.072) season 52.536*** 13.820 (11.084) (8.911) C 112.747*** 32.502*** 36.543*** 22.681 (8.121) (7.662) Dummy tourist destination Included N 282 282 400 R2 0.254 0.438 0.228 Log-likelihood -1505 -1116 F-test - 91.53*** 4.84*** AIC 3076.068 2245.827 3291.378 BIC 3196.251 2271.195 3343.267 VIF 2.46 1.94 1.75 Notes: Coeff. represents the beta coefficients of the independent variables, Std. Error represents the standard errors robust to heteroscedasticity, following the method proposed by Huber and White (Huber, 1967; White 1980, 1982). The disaggregated model is run for areas with at least 5 DMUs. R2 is a measure of the goodness-of-fit of the model. Log-likelihood is the value of the log likelihood function. F-test is a joint test of the nullity of the estimated parameters. AIC is the Akaike information criterion and BIC is the Bayesian information criterion, which can be used to compare the models. VIF is the mean of the variance inflation factor, which can be used to measure multicollinearity. ***, **, * indicate significance at 1%, 5% and 10% respectively. Fig. 6 shows the joint effect of location variables on the tourist destination component of efficiency for each of the 97 Spanish tourist destinations identified by the National Statistics Institute. To provide this graphic representation, the area of influence of each tourist destination was evaluated using Thiessen polygons. This is the most appropriate methodology if the distance determines the level of influence, as is the case in the hotel sector (Lado-Sestayo et al., 2016a). The darker colour means a higher contribution of the location factor in increasing hotel efficiency through variables outside the control of hotel managers. The spatial analysis indicates that the tourist destinations situated on the coasts, especially the (warmer) islands and Mediterranean and south-western coasts, have conditions that contribute to increasing hotel efficiency. In contrast, a low contribution is found in the centre, except for the capital, Madrid. Fig. 6. Joint effect of tourist destination variables on hotel efficiency The quantitative impact of the significant explanatory variables on efficiency is shown, differentiating between the impact of the tourist destination (Fig. 7) and the impact of the characteristics of the hotel (Fig. 8). The results indicate that the impact of the tourist destination on efficiency is much higher than the effect of hotel characteristics. In particular, occupancy level, degree of seasonality and market concentration are the variables with the greater average impact, with values of 30.93, 13.55 and -13.12 percentage points respectively. The fourth variable in the impact ranking – demand level – has a significantly lower average impact (5.70), Lado-Sestayo, R., Otero-González, L., Vivel-Búa, M., & Martorell-Cunill, O. (2016b). Impact of location on profitability in the Spanish hotel sector. Tourism Management, 52, 405–415. http://dx.doi.org/10.1016/j.tourman.2015.07.011 Lee, S. K., & Jang, S. (2011). Room rates of US airport hotels: examining the dual effects of proximities. Journal of Travel Research, 50 (2), 186-197. Luo, H., & Yang, Y. (2016). Intra-metropolitan location choice of star-rated and non-rated budget hotels: The role of agglomeration economies. International Journal of Hospitality Management, 59, 72-83. https://doi.org/10.1016/j.ijhm.2016.09.007 Marshall, A. (1890). Principles of economics. London: Macmillan. Min, H., & Joo, S. J. (2009). A data envelopment analysis on assessing the competitiveness of Korean hotels. Service Industries Journal, 29(3), 367–385. http://doi.org/10.1080/02642060701849865 Morey, R. C., & Dittman, D. A. (1995). Evaluating a hotel GM's performance: A case study in benchmarking. Cornell Hospitality Quarterly, 36(5), 30. Neves, J. C., & Lourenço, S. (2009). Using data envelopment analysis to select strategies that improve the performance of hotel companies. International Journal of Contemporary Hospitality Management, 21(6), 698–712. http://doi.org/10.1108/09596110910975963 Oliveira, R., Pedro, M. I., & Marques, R. C. (2013a). Efficiency performance of the Algarve hotels using a revenue function. International Journal of Hospitality Management, 35, 59– 67. http://doi.org/10.1016/j.ijhm.2013.05.005 Oliveira, R., Pedro, M. I., & Marques, R. C. (2013b). Efficiency and its determinants in Portuguese hotels in the Algarve. Tourism Management, 36, 641–649. http://doi.org/10.1016/j.tourman.2012.06.009 Oliveira, R., Pedro, M. I., & Marques, R. C. (2015). Efficiency evaluation of Portuguese hotels in the Algarve using data envelopment analysis (DEA). Revista Brasileira de Gestão de Negócios, 17(54), 788. Pan, C.M. (2005). Market structure and profitability in the international tourist hotel industry. Tourism Management, 26, 845–850. Parte-Esteban, L., & Alberca-Oliver, P. (2015). Determinants of technical efficiency in the Spanish hotel industry: Regional and corporate performance factors. Current Issues in Tourism, 18(4), 391–411. http://doi.org/10.1080/13683500.2013.800029 Peiró-Signes, A., Segarra-Oña, M. D. V., Miret-Pastor, L., & Verma, R. (2015). The effect of tourism clusters on US hotel performance. Cornell Hospitality Quarterly 56(2), 155-167. Perrigot, R., Cliquet, G., & Piot-Lepetit, I. (2009). Plural form chain and efficiency: Insights from the French hotel chains and the DEA methodology. European Management Journal, 27(4), 268–280. http://doi.org/10.1016/j.emj.2008.11.001 Porter, M. E. (2008). The five competitive forces that shape strategy. Harvard business review, 86(1), 25-40. 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. European Journal of Operational Research, 204(3), 613–620. http://doi.org/10.1016/j.ejor.2009.11.006 Sainaghi, R. (2011). RevPAR determinants of individual hotels: evidences from Milan. International Journal of Contemporary Hospitality Management, 23(3), 297-311. https://doi.org/10.1108/09596111111122497 Seiford, L. M., & Zhu, J. (1999). Infeasibility of super-efficiency data envelopment analysis models. Information Systems and Operational Research, 37(2), 174-187. Shang, J. K., Hung, W. T., & Wang, F. C. (2008). Service outsourcing and hotel performance: three-stage DEA analysis. Applied Economics Letters, 15(13), 1053–1057. http://doi.org/10.1080/13504850600993523 Shang, J. K., Wang, F. C., & Hung, W. T. (2010). A stochastic DEA study of hotel efficiency. Applied Economics, 42(19), 2505–2518. Shuai, J. J., & Wu, W. W. (2011). Evaluating the influence of e-marketing on hotel performance by DEA and grey entropy. Expert Systems with Applications, 38(7), 8763–8769. http://doi.org/10.1016/j.eswa.2011.01.086 Simar, L. & Wilson, P. W. (2007), Estimation and inference in two-stage, semi-parametric models of productive efficiency. Journal of Econometrics 136 (1), 31–64. https://doi.org/10.1016/j.jeconom.2005.07.009 Such-Devesa, M. J., & Mendieta-Peñalver, L. F. (2013). Size, Efficiency and Productivity in the Spanish Hotel Industry—Independent Properties versus Chain-Affiliated Hotels. Tourism Economics, 19(4), 801-809. Thrall, R. M. (1996). Duality, classification and slacks in DEA. Annals of Operations Research 66 (2), 109–138. https://doi.org/10.1007/BF02187297 Urtasun, A., & Gutiérrez, I. (2006). Hotel location in tourism cities: Madrid 1936–1998. Annals of Tourism Research, 33(2), 382–402. Vivel-Búa, M., Lado-Sestayo, R., & Otero-González, L. (2016). Impact of location on the probability of default in the Spanish lodging industry: A study of MSMEs. Tourism Economics, 22(3), 593–607. http://doi.org/10.5367/te.2015.0461 White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838. White, H. (1982). Maximum likelihood estimation of misspecified models. Econometrica, 50, 1–25. Wöber, K. W. (2007). Data envelopment analysis. Journal of Travel & Tourism Marketing, 21(4), 91–108. Wu, J., Liang, L., & Song, H. (2010). Measuring hotel performance using the integer DEA model. Tourism Economics, 16(4), 867–882. http://doi.org/10.5367/te.2010.0015 Wu, J., Tsai, H., & Zhou, Z. (2011). Improving efficiency in international tourist hotels in Taipei using a non-radial DEA model. International Journal of Contemporary Hospitality Management, 23(1), 66–83. http://doi.org/10.1108/09596111111101670 Yang, Y. (2012). Agglomeration density and tourism development in China: An empirical research based on dynamic panel data model. Tourism Management, 33(6), 1347–1359. Yang, Y., Luo, H., & Law, R. (2014). Theoretical, empirical, and operational models in hotel location research. International Journal of Hospitality Management, 36, 209–220. http://doi.org/10.1016/j.ijhm.2013.09.004 Yu, M. M., & Lee, B. C. Y. (2009). Efficiency and effectiveness of service business: Evidence from international tourist hotels in Taiwan. Tourism Management, 30(4), 571–580. http://doi.org/10.1016/j.tourman.2008.09.005