International Journal of Hospitality Management 29 (2010) 468–475
Contents lists available at ScienceDirect
International Journal of Hospitality Management journal homepage: www.elsevier.com/locate/ijhosman
Hotel efficiency: A bootstrapped metafrontier approach A. Assaf a, C.P. Barros b,c,*, A. Josiassen d a
Isenberg School of Management, Department of Hospitality and Tourism Management, University of Massachusetts-Amherst, USA Instituto Superior de Economia e Gesta˜o, Technical University of Lisbon, Rua Miguel Lupi 20, 1249-078 Lisbon, Portugal c UECE (Research Unit on Complexity and Economics), Portugal d Centre for Tourism and Services Research, Victoria University, Australia b
A R T I C L E I N F O
A B S T R A C T
Keywords: Efficiency Data envelopment analysis Metafrontier Taiwanese hotels
This paper introduces the metafrontier concept to account for the environmental and technological differences between various hotels groups. The interesting feature of the model is that it ensures that heterogeneous hotels are compared based on one homogenous technology. We test the model using a panel data sample of 78 Taiwanese hotels. The results clearly indicate that the size, ownership, and classification of a particular hotel have a significant impact on its efficiency. More implications of the results are provided. ß 2009 Elsevier Ltd. All rights reserved.
1. Introduction The importance of efficiency in the daily operations of the hotel industry is well supported and evidenced in the literature. Currently, efficiency studies cover several international hospitality industries such as Portugal (Barros and Santos, 2006; Barros, 2005a,b; Barros and Mascarenhas, 2004), Taiwan (Hwang and Chang, 2003), the US (Anderson et al., 1999a,b; Anderson et al., 2000; Reynolds, 2003; Brown, 2002); and the UK (Johns et al., 1997). An important characteristic of the modern efficiency literature on hotels relates to the extensive use of advanced efficiency methods such as DEA (data envelopment analysis) and SFA (stochastic frontier analysis) which are relatively flexible methods and can simply account for the multiple inputs/outputs setting of the industry (Barros and Dieke, 2008). A review of the current literature indicates however that most existing studies tend however to suffer from one common limitation in terms of accounting for the impact of environmental variables on efficiency. For example, environmental factors such as size, location, and type of ownership are all uncontrollable variables that can interfere with the efficiency results and thus should be carefully addressed. A review of the literature clearly indicates that most of the current studies tend to combine hotels belonging to different environmental groups (e.g. small and large hotels) in one data sample prior to the
* Corresponding author. Tel.: +351 213 016115; fax: +351 213 925 912. E-mail addresses:
[email protected] (A. Assaf),
[email protected] (C.P. Barros),
[email protected] (A. Josiassen). 0278-4319/$ – see front matter ß 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijhm.2009.10.020
estimation of efficiency. The impact of environmental variables is then determined by using either a second stage regression (Barros and Dieke, 2008) or analysis of variance (ANOVA) (Chen, 2007). While it is true that these approaches might provide some inferences on the impact of environmental variables on efficiency, the idea of combining heterogeneous hotels in one sample may in the first place distort the efficiency results and subsequently the results of the second stage regression or the ANOVA model. Such issue was recently raised by O’Donnell et al. (2007) who argued that the SFA or DEA methods might sometimes lead to inaccurate results if the sample under consideration includes firms which belong to different environmental characteristics. This is because the efficiency frontiers for these firms might not be identical to provide an unbiased comparison. For example, large and small hotels have different characteristics in terms of economies of scale, market share, and access to advanced technologies and thus should not be treated as one homogenous group. To account for this heterogeneity problem, O’Donnell et al. (2007) have recently provided a methodological solution by introducing the concept of metafrontier which can ensure that all heterogeneous firms or groups are assessed based on their distance from a common and identical frontier. In a simple definition the metafrontier function can be considered as an envelop of all possible frontiers that might arise from the heterogeneity between firms (Rao et al., 2003). In other words it aims to provide a homogenous boundary for all heterogeneous firms. The impact of environmental variables can then be measured more accurately by assessing the distance of a firm belonging to a certain environmental group from the metafrontier.
Method
Focus on restaurant chain efficiency
Role of satisfaction on efficiency
(1) Sales (2) Tips (3) Turnover (1) Server wage (2) Seats (3) Square footage (4) Units in the same State (5) Several other contextual variables 60 brand restaurants from a chain
(2) Profit (3) Guest satisfaction (4) Average purchase frequency (2) Labor costs (3) Employee satisfaction (4) Rent
(5) Taxes and insurances (6) Square footage (7) Number of seats
(1) Revenue (1) Costs of goods sold Data on 32 surveys undertaken in a 36 samebrand restaurants
(4) Operating expenses (5) Depreciation expenses
(1) Occupancy rate (2) Rate of guest satisfaction (3) Number of guests (4) Room revenue (5) Other revenue
DEA
The boundary of the output set is the production possibility frontier and represents technically efficient production. The
Reynolds and Thompson (2007)
(3)
DEA
PðxÞ ¼ f y : ðx; yÞ 2 T g
Reynolds and Biel (2007)
The boundaries for the input sets determine the ‘isoquants’. The output set is defined for a specific vector of input x as the set of all output vectors y which can be produced using x :
Study
(2)
Table 1 Literature survey of frontier models on tourism.
XðyÞ ¼ fx : ðx; yÞ 2 T g
Units
To introduce the metafrontier model, let y and x, respectively, denote non-negative output and input vectors of dimensions (N 1) and (M 1) respectively. We consider the case where there are K(>1) groups and each group operates under a specific technology, Tk(k = 1, 2, . . ., K). Battese et al. (2004) argued that since technology is a state of knowledge related to the transformation of N input into M outputs, it is possible to conceptualize the existence of an over-arching technology or metatechnology, which they represent by T*. The technology of a given group, called technology set, is defined as the set of all feasible input–output vectors. T ¼ ðx; yÞ 2 Rþpþq jx can produce y (1)
(1) Number of employees (2) Surfaced area, (3) guest rooms
Inputs
3. Technical details of the metafrontier model
Data from a questionnaire
Outputs
The efficiency and productivity literature in tourism is old and well established. Table 1 presents a survey of the published literature, as well as the methodologies and data used. It is clear that none of the previous studies have followed the metafrontier approach. Other advanced models such as the Bayesian stochastic frontier, random frontier, and latent frontier models are also yet to reach the hospitality and tourism literature (Lewis and Anderson, 1999; Greene, 2005; Orea and Kumbhakar, 2004). With regards to the policy implications of the studies reviewed, they vary, but in general, they propose policies to overcome the identified inefficiency.
Data envelopment analysis papers Chen (2009) DEA with slacks
2. Literature survey
which describes the amount of some p inputs x that can produce q outputs y. We can define the input and output sets associated with the production technology set T, which provides an equivalent representation of production technology. The input set defined for a specific output vector y is the set of all input vectors x which can produce y.
469
Prices
Contribution
In this paper we aim to extend the application of the metafrontier to provide further evidence on the impact of environmental variables on hotel efficiency. We divide the hotels into three heterogeneous groups according to three environmental characteristics: size, classification, and type of ownership. These variables are the most popular in several efficiency studies, and were found in most cases to be strong determinants of hotel efficiency. In testing the model we use a sample of Taiwanese hotels ranging from years 2004 to 2008. In this way the study also contribute the analysis of efficiency in this country as none of the previous studies has applied the metafrontier concept. The estimation of the metafrontier is further enhanced with the use of the bootstrap approach which can be used to examine the statistical properties of efficiency scores generated through the metafrontier. The paper unfolds as follows: Section 2 provides the literature review, followed by a discussion of the metafrontier model and the DEA bootstrap in Section 3. A description of the data and variables used is provided in Section 4. Section 5 presents the results and Section 6 presents discussions and summary of the main findings.
Role of slacks on efficiency
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
470
Table 1 (Continued ) Study
Method
Units
Inputs
Outputs
Prices
Contribution
Barros and Dieke (2008)
DEA two-stage procedure: 1st stage Malmquist model; 2nd stage a bootstrapped tobit model
25 Portuguese travel agencies 2000-2004
(1) Wages, (2) capital, (3) total costs excluding wages and (3) book value of premises
(1) Sales and (2) profits
Barros and Santos (2006)
DEA allocative model
15 hotels observed from 1998 to 2002
(1) Employees and (2) physical capital
(1) Sales, (2) added value and (3) earnings
(1) Price of labor and (2) price of capital
Role of input prices on efficiency
Barros (2005a,b)
DEA–CCR and DEA–BCC models
42 Enatur hotels (Portugal), observed from 1999 to 2001
(1) Capital and (2) labor
(1) Sales; (2) number of guests; (3) nights spent
–
Analysis of a small hotels network
Barros and Alves (2004)
DEA–Malmquist index
42 Enatur hotels,
(1) Full-time workers; (2) cost of labor; (3) book value of property; (4) operating costs and (5) external costs
(1) Sales; (2) number of guests; (3) nights spent
–
Analysis productivity change in hotel industry
Role of bootstrap on the accuracy of efficiency scores
Hwang and Chang (2003)
CCR DEA model; superefficiency model; Malmquist
45 hotels in Taiwan
(1) Number of full-time employees; (2) number of guestrooms; (3) total dimension of meal department; (4) operating expenses
(1) Room revenue; (2) food and beverage revenue; (3) other revenue
–
Efficiency and productivity in hotel industry
Reynolds (2003)
DEA CCR and BCC model
38 restaurants
(1) Front-of-house hours worked per day during lunchtime; (2) front-of-hours worked during dinner per day; (3) average wages; uncontrollable input (4) number of competitors within a two-mile radius; (5) seating capacity
(1) Sales; (2) customer satisfaction
–
Efficiency of restaurants
Brown (2002)
DEA–CCR model and cluster analysis
46 US hotels rated in consumer report
(1) Median price; (2) problems (defined in a 4-point scale); (3) service; (4) upkeep; (5) hotels and (6) rooms
(1) Satisfaction value (defined on a 100-point scale); (2) value (defined in a 5-point scale)
–
Analysis the role of satisfaction in efficiency
Anderson et al. (2000)
DEA (technical and allocative)
48 hotels
(1) Full-time equivalent employees; (2) the number of rooms; (3) total gaming-related expenses; (4) total food and leverage expenses; (5) other expenses
(1) Total revenue; (2) other revenue
(1) Wages proxied by the hotel revenue per fulltime employee; (2) rooms price proxied by hotel revenue divided by the product of rooms times occupancy rate and day per-year)
Analysis the role of price and quantity effect in hotel efficiency
Johns et al. (1997)
DEA
15 UK hotels over a 12month period
(1) Number of room nights available; (2) total labor hours; (3) total food and beverage costs and (4) total utilities cost
(1) Number of roomnights sold; (2) total covers served; and (3) total beverage revenue
–
The first paper on tourism efficiency at European level
Bell and Morey (1995)
DEA
31 units of Corporate Travel Departments
(1) Actual level of travel expenditure; (2) nominal level of other expenditure; (3) level of environment factors (ease of negotiating discounts, percentage of legs with commuters, flights required; (4) actual level of labor costs
(1) Level of service provided, qualified as excellent and average
–
Focus on travel department
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
Observed from 1999 to 2001
54 hotels
(1) Room division expenditure; (2) energy costs; (3) Salaries; (4) non-salary expenditure for property; (5) salaries and related expenditure for advertising; (6) non-salary expenses for advertising; (7) fixed marked expenditure for administrative work
(1) Total revenue; (2) level of service delivered; (3) market share; (4) rate of growth
–
Methods
Units
Outputs
Endogenous variable
Prices
DEA and stochastic frontier
31 Corporate travel departments
(1) Total air expenses; (2) hotel expenses; (3) car expenses; (4) labor expenses; (5) hourly labor; (6) part-time labor; (7) fee expenses; (8) technology costs; (9) building and occupancy expenses
(1) Number of trips
(1) Price of labor, estimated by dividing the labor expenses by the number of trips; (2) price of travel obtained dividing the travel expenses by the number of trips; (3) price of capital obtained by dividing the capital expense by the number of trips
Comparing DEA and stochastic methods
Anderson et al. (1999a)
Stochastic translog production frontier
48 Hotels
(1) Number of full-time equivalent employees; (2) number of rooms; (3) total gaming-related expenditure; (4) total food and beverage expenses; (5) other expenses
(1) Total revenue
(1) Price of labor, proxied by the hotel revenue per full-time equivalent employee; (2) room price proxied by hotel revenue by the product of number of rooms times the occupancy rate and day per year. (3) Price of gaming, food, beverage and other expenses proxied as the percentage of total revenue
The first paper adopting stochastic frontier models in tourism
Barros (2004)
Stochastic translog frontier
42 Enatur hotels (Portugal), observed from 1999 to 2001
(1) Sales, (2) market share, (3) trend
(1) Operational cost
(1) Price of labor; (2) price of capital
The first paper at European level adopting stochastic frontier models
Barros (2006)
Stochastic translog cost frontier with technological change
15 Portuguese hotels, 1998–2002
(1) Sales, (2) market share, (3) trend
(1) Operational cost
(1) Price of labor; (2) price of capital
Focus on technological change in defining tourism efficiency
Assaf and Matawie (2008)
Stochastic Battese and Coelli (1988) frontier model
Web questionnaire on 90 health care foodservices
(1) Production space/area; (2) degree of readiness of raw material; (3) age of equipment; (4) skill level of employees; (5) number of meals; (6) dummies
(1) Total cost
(1) Labor price
Focus on food services
Study Stochastic frontier papers Anderson et al. (1999b)
The first USA paper on hotel efficiency
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
DEA
Morey and Dittman (1995)
(2) Energy price Stochastic translog random frontier
12 Angola Hotels, 1990– 2007
Revpar, profit, market share
Operational cost
(1) Price of labor; (2) price of capital-premises and (3) price of capital-finance
Focus on African hotels
471
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
472
metafrontier can be described as a function that envelops separate group frontiers, each having their specific state of technology and environmental factors. In other words, the metafrontier model is considered as an envelope of all the possible group technologies. For example, if a particular output, y, can be produced using input vector, x, in one of the groups, then (x,y) are considered as part of the metatechnology, T*, which is defined by O’Donnell et al. (2007) as: T ¼ fðx; yÞ : x 0 and y 0; such that x can produce y in at least one regional technology; T 1 ; T 2 ; :::; T k g
(4)
The convexity property of this metatechnology was ensured by O’Donnell et al. (2007) by defining the metatechnology as the convex hull of the union of group specific technologies, denoted by: T Convex HullfT 1 [ T 2 [ ::: [ T k g D0 ðx; yÞ
(5)
Di ðx; yÞ
and denote the output and input If we let distance function defined using the metatechnology T*, where for a given group k, an output distance function is defined as n o Dk0 ðx; yÞ ¼ inf u u > 0 : ðy=uÞ 2 Pk ðxÞ (6) and it shows the maximum degree to which a given output vector can be increased and still within the production feasibility set, while an input distance function is defined as: n o Dki ðx; yÞ ¼ sup l > 0 : ðx=l 2 X k ðyÞ (7) l
and it shows the maximum degree to which a given input vector can be radically contracted and yet produce the same level of output, y. From the definition of metatechnology it can be easily shown that the input ‘‘Di ðx; yÞ’’and output distance functions ‘‘D0 ðx; yÞ’’ defined using the metatechnology T*, satisfy the following requirements: Requirement 1 : For any given group k; Dk0 ðx; yÞ D0 ðx; yÞ;
k
¼ 1; 2; :::; k Requirement 2 : For any given group k; Dki ðx; yÞ Di ðx; yÞ; ¼ 1; 2; :::; k
(8) k (9)
Using the conditions in (8) and (9) we can also obtain measures of the gap between the group k technology and the metatechnology. For example, Battese et al. (2004) formulated a technology gap ratio that takes a value between zero and one and measures the ratio of the output for the frontier production function for the k-th group relative the potential output defined by the metafrontier function, given the observed inputs. To illustrate, the outputorientated technology gap ratio can be defined using the output distances functions from technologies Tk and T* as: TGRki ðx; yÞ ¼
Dki ðx; yÞ TEi ðx; yÞ ¼ Di ðx; yÞ GTEki ðx; yÞ
for the construction of the metafrontier as it can easily accommodate for multiple inputs and outputs. The use of the DEA methodology to construct the metafrontier model is possible as long as we can identify separate frontiers for different groups in the data set (O’Donnell et al., 2007). The group frontiers are constructed by estimating a DEA model for each group. The estimation of the metafrotnier then follows by applying the same DEA model to the data set obtained through pooling all observations for firms from all groups. A comprehensive overview of the early literature on DEA models has appeared in several studies, and thus will not be repeated here. For a detailed review refer to Coelli et al. (1998). The DEA model is relatively simple to estimate but is deterministic and has no account for measurement error. The bootstrap approach must generally be combined with DEA to obtain statistical properties of the efficiency scores. The bootstrap is a computer intensive technique basic on the idea of mimicking the unknown distribution of interest through the concept of resampling from the original sample. For more technical details on the DEA bootstrap method refer to Simar and Wilson (2007). 4. Data and group formulation The data used in the empirical application of this study were obtained from the Tourism Bureau in Taiwan which conducts regular surveys of the hotel industry in Taiwan. The variables selected follow the literature (Barros and Dieke, 2008; Barros, 2005a,b; Anderson et al., 1999a,b) and consist of four inputs and five outputs. Inputs are defined as the number of rooms (proxy of capital cost), number of full time equivalent employees in the room division, number of full time equivalent employees in the food and beverage division, and number of full time equivalent employees in other departments. On the output side we use the total room revenues, total food and beverage revenues, total of other revenues (revenues from lease of store spaces, laundry, swimming pool, ball courts, barber shop, beauty salons and bookstores), market share for each hotel (percentage of total hotel guest out of the total guests received by all Taiwanese hotels) and employees performance (measured as the number of guest per employee). In total, data on the above variables were collected from 78 key Taiwanese hotels for the period 2004–2008 (78 5 = 390 observations). The next step was to identify the separate hotel groups for the estimation of the metafrontier model. The division of groups was also based on the environmental differences between the various hotels in the sample. Specifically, we divide the hotels based on their size, type of ownership and classification. The impact of size creates heterogeneity between hotels, and it is generally hypothesized that the economies of scale associated with large hotels enables them to quickly expand their operations and
(10)
where TEi denotes the technical efficiency with respect to metafrontier and GTEki the technical efficiency with respect to a group k. This implies that the technical efficiency of a firm relative to the metafrontier is simple the product of the technical efficiency of that firm relative to the frontier for a particular group and the technology gap for that group.As it is clear from above, the first step in estimating these different efficiency measures is to estimate the metafrontier technology and the group frontier technology. These can be constructed using either a non-parametric technique such as data envelopment analysis (DEA) or a parametric technique such as stochastic frontier. In this paper we use the DEA approach
Fig. 1. Graphical representations of the various hotel groups.
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
473
Table 2 Descriptive Statistics of the data. Independent hotels
Chain hotels
Small hotels
Large hotels
International tourist hotels
Tourist hotels
Room revenues Mean St. Dev
113851761 93458596
315580750 287470371
113887103 87857430
432261473 302543515
242695491 234282398
60970889 46467752
F&B revenues Mean St. Dev
122190807 149340894
338385124 362346698
108507233 116743094
506041353 383776759
267704815 298702770
46601946 64515741
Other revenues Mean St. Dev
33691004 84835751
97359207 236786966
34310756 62649576
132295253 102552543
71282872 189864099
24822889 44973661
Number of rooms Mean St. Dev Mean St. Dev
2376 1349 2376 1349
3681 2181 3681 2181
2083 919 2083 919
5345 1747 5345 1747
3436 1826 3436 1826
1455 742 1455 742
1.103 0.777
1.924 1.217
1.213 0.921
2.204 1.235
1.712 1.222
0.925 0.881
42.03 32.07 2376 648 511
54.07 49.28 3681 1242 852
43.52 33.21 2083 592 400
60.12 44.32 5345 1742 851
57.23 40.80 3436 1150 790
39.75 33.50 1455 350 222
FTE in F&B department Mean St. Dev
853 960
1972 1559
801 750
2892 1602
1690 1390
325 379
FTE in other departments Mean St. Dev Sample size (over 5 years)
282 418 240
545 823 150
262 291 295
782 1061 95
509 706 110
163 160 280
Market share (%) Mean St. Dev No. of guests/employees Mean St. Dev Mean Mean St. Dev
FTE refer to the number of full time equivalent employees.
achieve operational savings. In classifying the hotels as large or small we follow Shang et al. (2009) and define small hotels as those having less than 300 rooms and large hotels as those having more than 300 rooms. The impact of ownership (chain or independent hotels) was also tested in several related studies (Barros and Dieke, 2008; Barros, 2006) and it is generally hypothesized that chain ownership might allow hotels to improve their management abilities, to have easier access to new technologies, and to raise capital at lower cost. Finally, in terms of classification we follow the classification of the Taiwanese tourism bureau and divide the hotels into ‘‘international tourist class’’ or ‘‘tourist class’’ hotels. International tourist hotels are those that have higher star rating and better service facilities, and thus by dividing the hotels into these two groups we aim to reflect the impact of quality standards on efficiency. In Fig. 1 we provide a graphical representation of each group.
Table 3 Average efficiency results.
Chain hotels Independent hotels Difference test Large hotels Small hotels Difference test International tourist hotels Tourist hotels Difference test
Average GTE
Average MTE
Average TGR
0.7721 0.7334 5.4220** 0.7321 0.6821 5.7921** 0.7122 0.6832 2.9230
0.6821 0.6122 4.6113** 0.7014 0.6321 4.1125* 0.9422 0.9135 4.3101*
0.8748 0.7722 4.9325** 0.6921 0.6223 6.5511** 0.9182 0.8621 4.9677**
As shown in Fig. 1, the metafrontier envelops each group frontier as thus provide a more consistent and homogenous efficiency comparison. Note that the group frontier might sometimes touch the metafrontier if a hotel is equally efficient with respect to both the groups and metafrontier model. In this study we are estimating three separate metafrontier models, and six group frontiers, one for each environmental category. For more details on the sample size of each group as well as the descriptive statistics of the data refer to Table 2. 5. Results The technical efficiency estimates associated with the group and metafrotnier models were obtained using 2000 bootstrap iterations.1 The average efficiency results are reported in Table 3. We also report the individual efficiency results of selected Taiwanese hotels in Table 4. The impact of ownership was tested with two groups, namely independent hotels and chain hotels. The results for these two groups as well as the associated metafrontier are reported in the first two columns of Table 3. An analysis of variance comparison between the different efficiency results is also reported in Table 3. As it is clear, chain hotels are performing significantly better than independent hotels both in terms of the group and metafrontier models. For instance, the average group efficiency for chain hotels is 77.21%, whereas the average group efficiency for independent hotels is 73.34%. The metafrontier comparison is however expected to be more accurate as it is based on one common homogenous technology. For example, according to the metafrontier results,
*
Coefficients significant at 5%. Coefficients significant 1%. The difference test employed in the table is one-way ANOVA test with F-statistic. **
1 Simar and Wilson (2007) recommended the use of 2000 bootstrap iterations to obtain reliable bootstrap estimates.
474
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
Table 4 Individual efficiency results of selected Taiwanese hotels. Hotel name
GTE
MTE
TGR
Hotel name
GTE
MTE
TGR
Independent hotels Alison house Astar Hotel1 Bao Hwa Hotel El Dorado Hotel Empress Hotel Evergreen Laurel Hotel (Keelung) Gala Hotel Kilin Hotel Taipei Lake Hotel Lion Hotel
0.7932 0.7372 0.5501 0.6523 0.6647 0.8578 0.9092 0.8590 0.7949 0.7221
0.6299 0.4642 0.5472 0.5783 0.4830 0.7410 0.7190 0.8384 0.7772 0.6492
0.7813 0.6310 0.9950 0.8711 0.7262 0.8522 0.7918 0.9795 0.9812 0.9021
Chain hotels Caesar Park Hotel Kenting Caesar Park Taipei Chinatrust Hotel Hualien Gloria Prince Hotel Grand Formosa Hotel, Taroko Grand Formosa Regent Taipei Grand Hyatt Taipei Hotel Royal Chihpen Spa Hotel Royal Hsinchu Hotel Royal Taipei
0.6432 0.8223 0.9310 0.3921 0.8617 0.9013 0.7909 0.9500 0.4446 0.7411
0.5568 0.5672 0.9204 0.3021 0.7946 0.8795 0.5800 0.9216 0.2518 0.5226
0.8504 0.6953 0.9946 0.9815 0.9233 0.9757 0.7341 0.9733 0.5610 0.7132
Small hotels Astar Hotel1 Bao Hwa Hotel Cosmos Hotel Dragon Valley Hotel Empress Hotel Evergreen Laurel Hotel (Keelung) Kilin Hotel Taipei Lake Hotel Lion Hotel Hotel Kuva Chateau
0.5112 0.6132 0.3662 0.7712 0.5310 0.8472 0.8639 0.7712 0.7051 0.9232
0.4613 0.5957 0.3295 0.7523 0.4823 0.7673 0.8576 0.7721 0.7033 0.9230
0.8999 0.9766 0.8723 0.9888 0.9159 0.9062 0.9935 0.9993 0.9911 0.9998
Large hotels The Grand Hotel Parkview Hotel Taoyuan Holiday Hotel Caesar Park Taipei Grand Formosa Regent Taipei Grand Hyatt Taipei Sheraton Taipei Hotel Tayih Landis Tainan The Ambassador Hotel The Sherwood Hotel Taipei
0.6123 0.5275 0.8311 0.8238 0.8482 0.8227 0.8015 0.8010 0.7995 0.7610
0.6051 0.5212 0.8205 0.5772 0.8404 0.5908 0.8002 0.7846 0.7966 0.7578
0.9915 0.9896 0.9810 0.6921 0.9912 0.7321 0.9984 0.9896 0.9920 0.9991
Tourist class hotels Kilin Hotel Taipei Lake Hotel Lion Hotel First Hotel Hotel Kuva Chateau The Leofoo Hotel The Monarch Plaza Hotel The Rivera Hotel South Formosa Hotel Sun Spring Resort
0.8172 0.4911 0.6953 0.7732 0.8810 0.6311 0.5411 0.6062 0.7310 0.6011
0.7697 0.48101 0.6832 0.7381 0.8304 0.6310 0.4194 0.5747 0.7221 0.5827
0.9460 0.9749 0.9883 0.9621 0.9447 0.9959 0.7850 0.9551 0.9842 0.9721
International tourist class hotels Hotel Royal Chihpen Spa Hotel Kingdom Hotel National Hotel Riverview Taipei Hotel Tainan The Grand Hotel The Hibiscus Resort The Lalu Sun Moon Lake Imperial Hotel Taipei Marshal Hotel
0.9378 0.7621 0.4906 0.8201 0.8035 0.7707 0.8724 0.8112 0.5011 0.7542
0.9212 0.6381 0.4838 0.8050 0.7280 0.7152 0.8082 0.7756 0.4622 0.7410
0.9823 0.8392 0.9862 0.9811 0.9160 0.9242 0.9264 0.9620 0.9167 0.9849
chain hotels are only operating at 68.21% efficiency level, but are still higher than independent hotels which are operating at an efficiency level of 61.22%, validating previous research in the area (Barros, 2006). An interesting measure is also the technology gap ratio (TGR) which indicates that independent hotels have achieved only 77.22% of their potential outputs while chain hotels have achieved 87.48% of their potential output. The results for the two groups representing small and large hotels are presented in the third and fourth columns of Table 3. It is clear from the ANOVA results that large hotels are performing better than small hotels both in terms of the group and metafrontier models, validating previous research in the area (Barros, 2005a). For instance, large hotels have an average group efficiency of 73.21% and average metafrontier efficiency of 70.14%, while the average efficiency of small hotels is 68.21% and 63.21% for the group and metafrontier models respectively. The TGR measures also illustrate the advantage of size, with large hotels are achieving around 7% higher of their potential outputs than small hotels. The last two columns Table 2 present the results of the last two groups, namely international tourist and tourist class hotels. As mentioned before, the aim of this grouping is to reflect the impact of quality standards on efficiency. It is evident from the ANOVA results that international tourist hotels operate at a similar group efficiency level in comparison to tourist class hotels; however, their average metafrontier efficiency is significantly higher (close to 4%). The average technology gap ratio is also in line with the efficiency results and is close to 5% higher for international tourist hotels. 6. Discussion and conclusions The major aim of this study was to provide further validation on the impact of environmental variables on hotel efficiency. The use
of the metafrontier methodology is in line with the resource-based theory of Barney (1986,1991) and Teece et al. (1997) which justifies that hotels are heterogeneous in terms of the resources and capabilities on which they base their managerial practices, and thus heterogeneity is expected to interfere with efficiency. Given that this methodology is used for the first time in this area, it is difficult to have a direct comparison between the efficiency results of this study and other related studies in the area. However, it is possible to discuss whether related studies have converged to similar conclusions in terms of the impact of the selected environmental variables on efficiency. The impact of size for example was tested in this study and it was clear that it is strong determinant of hotel efficiency. For instance, large hotels had higher efficiency in terms of the group and the metafrontier models, as well as a higher TGR. The idea of linking the variable of firm size to profit and performance is traced back to pioneering research in the area of performance studies (Lall, 1980; Wolf, 1977; Pugh et al., 1969). Different researchers have tested whether larger firms increase market share and achieve economies of scale and as result experience better performance. More recently, Taymaz (2005) has found that larger manufacturing firm size has a positive effect on technical efficiency, and on the ability of the firm to expand its operations. Therefore, in general, it is hypothesized that larger size has a positive relationship with firm profits and firm success. In the hotel sector, studies which have compared small and large hotels have not converged to the same conclusion. For example, two recent studies on Taiwanese hotels (Chen, 2007; Hwang and Chang, 2003) have indicated that no difference in efficiency exist between large scale and small scale hotels. This finding was later contradicted by Barros and Dieke (2008) who showed that large African hotels are more efficient than small
A. Assaf et al. / International Journal of Hospitality Management 29 (2010) 468–475
African hotels. The same conclusion was also reached by Barros (2006, 2005a) on his study on Portuguese hotels. Thus, in general, the relationship between size and efficiency is an area of contradiction in the literature. The results of this study should provide a clearer reflection on the impact of size, given that it also controls for the heterogeneity between small and large hotel groups. In terms of the relationship between efficiency and the other two selected environmental variables, namely, hotel type and hotel classification, the issue appears to be less contradictory. For example, most studies (Chen, 2007; Wang et al., 2006; Hwang and Chang, 2003; Barros and Dieke, 2008) on Taiwan and other international hotels have found that the efficiency of chain operations is higher than independent operations. Such agreement in the literature might be attributed to the fact that chain hotels have better marketing strategies, more sound management policies, and stronger economies of scale (Wang et al., 2006). A chain related hotel has also usually a well-known name, established operating policies, sometimes decor standards, and common reservation systems. The large chain hotels, in particular, are more likely to be in highly accessible, central locations and aim to attract key market targets (i.e. business travelers and package tourists) by means of discounts loyalty inducements and rates negotiated with carriers and tour operators. Similar characteristics can also be attributed to the international tourist hotels in Taiwan, which are mostly parts of chains, and have better branding. International tourist hotels have also higher star rating, which is usually associated with higher efficiency through better service quality. As an illustration, Barros and Dieke (2008) have for example recently found that higher star rating leads to a higher technical efficiency. To sum up, what are the main benefits of the results of this study? The findings are based on a more accurate methodology and thus can be used as a starting point for future investigations into the best practices adopted by the best performing hotels. Particularly, future studies might select some efficient and inefficient properties across several Taiwanese areas, and conduct case studies to determine the difference in practices adopted by these different properties. Such investigation might also be extended to develop a series of best practices that can be adopted by the low performing Taiwanese hotels. The findings could also of interest to the Taiwanese government, especially in the process of adopting improvement strategies to the whole industry. Government officials can also know formulate a better understanding of the difference in efficiency between several types of Taiwanese hotels. Specifically, it is crucial that the Taiwanese government adopts policies that suit each individual hotel group. For example, policies towards small hotels might need to be different from large hotels as these hotels have traditional resource limitations and thus might need further assistance. Future strategies might also focus on encouraging group ownership of new hotels. The need for higher quality standards should also be embedded in any new strategy. References Anderson, R.I., Fish, M., Xia, Y., Michello, F., 1999a. Measuring efficiency in the hotel industry: a stochastic frontier approach. International Journal of Hospitality Management 18, 45–57. Anderson, R.I., Lewis, D., Parker, M.E., 1999b. Another look at the efficiency of corporate travel management departments. Journal of Travel Research 37, 267– 272. Anderson, R.I., Fok, R., Scott, J., 2000. Hotel industry efficiency: an advanced linear programming examination. American Business Review 18, 40–48. Assaf, A., Matawie, K.M., 2008. Cost efficiency modeling in health care foodservice operations. International Journal of Hospitality Management 27, 604–613.
475
Barney, J., 1986. Strategic factor markets: expectations, luck and business strategy. Management Science 32, 1231–1241. Barney, J., 1991. Firm resources and sustained competitive advantage. Journal of Management 17, 99–120. Barros, C.P., 2004. A stochastic cost frontier in the Portuguese hotel industry. Tourism Economics 10, 177–192. Barros, C.P., Alves, P., 2004. Productivity in tourism industry. International Advances in Economic Research 10, 215–225. Barros, C.P., Mascarenhas, M.J., 2004. Technical and allocative efficiency in a chain of small hotels. International Journal of Hospitality Management 24, 415–436. Barros, C.P., 2005a. Measuring efficiency in the hotels: an illustrative example. Annals of Tourism Research 32, 456–477. Barros, C.P., 2005b. Evaluating the efficiency of small hotel chain with a malmquist productivity index. International Journal of Tourism Research 7, 173–184. Barros, C.P., 2006. Analysing the rate of technical change in tourism industry. Tourism Economics 12, 325–346. Barros, C.P., Santos, C.A., 2006. The measurement of efficiency in Portuguese hotels with DEA. Journal of Hospitality & Tourism Research 30, 378–400. Barros, C.P., Dieke, P.U.C., 2008. Technical efficiency of African hotels. International Journal of Hospitality Management 27, 438–447. Battese, G.E., Coelli, T.J., 1988. Prediction of firm-level technical efficiencies with a generalised frontier production function and panel data. Journal of Econometrics 38, 387–399. Battese, G.E., Rao, P., O’Donnell, C., 2004. A Metafrontier Production Function for Estimation of Technical Efficiencies and Technology Gaps for Firms Operating Under Different Technologies. Journal of Productivity Analysis 21, 91–103. Bell, R.A., Morey, R.C., 1995. Increasing the efficiency of corporate travel management through macro-benchmarking. Journal of Travel Research 33, 11–20. Brown, J.R., Ragsdale, C.T., 2002. The competitive market efficiency of hotel brands: an application of data envelopment analysis. Journal of Hospitality & Tourism Research 26, 260–332. Chen, C., 2007. Applying to stochastic frontier approach to measure hotel managerial efficiency in Taiwan. Tourism Management 28, 696–702. Chen, T.H., 2009. Performance measurement of an enterprise and business units with an application to a Taiwanese hotel chain. International Journal of Hospitality Management 28, 415–422. Coelli, T.J., Prasada, R., Battese, G.E., 1998. An Introduction to Efficiency and Productivity Analysis. Kluwer Academic Press. Greene, W., 2005. Fixed and random effects in stochastic frontier models. Journal of Productivity Analysis 23, 7–32. Hwang, S.N., Chang, T.Y., 2003. Using data envelopment analysis to measure hotel managerial efficiency change in Taiwan. Tourism Management 24, 357–369. Johns, N., Howcroft, B., Drake, L., 1997. The use of data envelopment analysis to monitor hotel productivity. Progress in Tourism and Hospitality Research 3, 119–127. Lall, S., 1980. Monopolistic advantages and foreign involvement by U.S. manufacturing industry. Oxford Economic Papers 32, 102–122. Lewis, D., Anderson, R.I., 1999. Residential real estate brokerage efficiency and the implications of franchising: a Bayesian approach. Real Estate Economics 27, 543–560. Morey, R.C., Dittman, D.A., 1995. Evaluating a hotel GM’s performance: a case study in benchmarking. Cornell Hotel Restaurant & Administration Quarterly 36, 30–35. O’Donnell, C., Rao, D., Battese, G., 2007. Metafrontier frameworks for the study of firm-level efficiencies and technology ratios. Empirical Economics 34, 231–255. Orea, L., Kumbhakar, S., 2004. Efficiency measurement using stochastic frontier latent class model. Empirical Economics 29, 169–183. Pugh, D.S., Hickson, D.J., Hinings, C., Urner, R.C., 1969. The context of organization structure. Administrative Science Quarterly 14, 91–114. Rao, P., O’Donnell, C., Battese, G., 2003. Metafrontier Functions for the Study of Intergroup Productivity Differences. CEPA Working Paper Series No. 01/2003, School of Economics, University of New England, Armidale. Reynolds, D., 2003. Hospitality–productivity assessment using data envelopment analysis. Cornell Hotel and Restaurant Administration Quarterly 44, 130–137. Reynolds, D., Thompson, G.M., 2007. Multiunit restaurant productivity assessment using three-phase data envelopment analysis. International Journal of Hospitality Management 26, 20–32. Reynolds, D., Biel, D., 2007. Incorporating satisfaction measures into a restaurant production index. International Journal of Hospitality Management 26, 352– 361. Shang, J.K., Hung, W.T., Wang, F.C., 2009. Service Outsourcing and Hotel Performance: three stage DEA analysis. Applied Economics Letters 15, 1053–1057. Simar, L., Wilson, P.W., 2007. Estimation and inference in two stage, semi-parametric models of productive efficiency. Journal of Econometrics 136, 31–64. Taymaz, E, 2005. Are small firms really less productive? Small Business Economics 25 (10), 429–445. Teece, D., Pisano, G., Shuen, A., 1997. Dynamic capabilities and strategic management. Strategic Management Journal 18, 509–533. Wang, F., Hung, W.T., Shang, J.K., 2006. Measuring the cost efficiency of international tourist hotels in Taiwan. Tourism Economics 12, 65–85. Wolf, B., 1977. Industrial diversification and internationalization: some empirical evidence. Journal of Industrial Economics 26, 177–191.
