In this section, we summarize the past researches that assess the efficiency of hotel industry. These can be divided into three categories: data envelopment analysis (DEA) studies of the hotel industry, stochastic frontier approach (SFA) studies in hotel industry and incorporating environmental effects into efficiency assessment.
However, efficiency is measured the distance of an observation from the efficient boundary. The boundary or frontier can be constructed by the non-parametric approach (data envelopment analysis) and the parametric approach (stochastic frontier approach). In previous studies, these are largely used in the hotel industry. In the third subsection is the application of meta-frontier.
2.1 Application of DEA approach
DEA is a linear programming-based technique and a non-parametric model, so it’ not necessary to give the functional form or probability distribution previous analyze. Moreover, DEA can deal with the multiple input and multiple output questions. It also can be used to compare the relatively efficiency of units by a common boundary which envelops all observations.
Tsaur (2001) uses the DEA to measuring the operating efficiency of 53 international tourist hotels in Taiwan during 1996-1998. The result reveals a mean efficiency score of 87.33%, which indicates that managers on average could reduce the input by 12.67% without decreasing the output. Hwang and Chang (2003) employ DEA to evaluate the managerial performance of 45 hotels of Taiwan in 1998.
In addition, they use the Malmquist productivity index to measuring the efficiency change of those hotels over time. The result shows that the cost on average should
be reduced 21% and hotel with different operating environment has different efficiency. The chain-hotel managerial and leisure hotels are better managed than independent managerial and urban hotels. The size of hotels has no significance relative to operating efficiency.
Chiang (2004) utilizes DEA-BCC model to evaluate the overall efficiency, pure technical efficiency and scale efficiency of 25 Taipei Taiwan hotels’ performance in 2000. The major finding indicates the international chain managerial hotels, not all perform more efficiently than the independent ones.
Yang and Lu (2006) investigate the managerial performance of 56 international tourist hotels in Taiwan by DEA-BCC model. In this study, they find that the inefficiency of 56 hotels primarily comes from pure technical inefficiency. On the other hand, they examine the operating environment how to influence the performance of hotels. Those environment variables are including management style, location, and closeness to Touyan International Airport. The result is that the operating characteristics are significance related to managerial efficiency on the statistical.
Wang et al. (2006) uses the five different measures, including overall efficiency, allocative efficiency, technical efficiency, scale efficiency and pure technical efficiency to evaluate the operating performance of 49 Taiwan’s hotel sector.
Moreover, this study utilizes the Tobit regression model to examine the effect of business or environment characteristics on hotel’s performance. The finding reveals that the foreign individual travelers, online transaction function and franchising are related to a better efficiency of Taiwan’s international tourist hotels. But the years of hotel has not significantly related to the efficiency measures.
Wang et al. (2006) apply the four-stage DEA to purge away the effect of external operating environments and calculate the pure managerial efficiency of 54 Taiwan’s
international tourist hotels. The external operating environment variables were selected: market condition, management style and hotel size. There are four hotel size classifies: smallest, next to smallest, next to largest, and largest. They observe that the chain-operated and resort hotels are favorable operating environment.
Shang et al. (2008) employ the three-stage DEA and find the average efficiency score and the number of efficient hotels increase in the final-stage, implies that the penalty to hotels operating under unfavorable circumstances is greater than the benefit to hotels operating under favorable circumstances. Besides, they examine the relationship between ecommerce adoption and managers’ performance after purged the effects of exogenous factors, and reveals that the performance is not affected by increased Internet usage. One of reasons for the result may be related to the unique characteristics of hotel industry. For instance, the “human touch” of service is not easily substituted by ecommerce. In other words, from this result we can find that the staff reduced in hotels is restricted.
2.2 Application of SFA approach
SFA is another way widely used for efficient frontier analysis that is a regression-based model. One of the advantages of SFA is that it provides statistical texts of the effect of independent variables on the dependent variable. Therefore, we can confirm what kind of external environments will impact the performance of hotels’ operating.
Anderson et al. (1999) apply the stochastic frontier model to measure the efficiency of 48 hotels in the United States. The input and output variables separately are the number of full-time equivalent employees, the number of rooms, total gaming related expenses, total food and beverage expenses, other expenses, and
total revenue. The input prices separately are the average wage, the average price of rooms, the average price of food and beverage operations, the average price of casino operations, the average price of hotel operations, and the average price of other expenses. They found the high efficiency scores in hotel industry. However, this work does not describe the resource of the inefficiency.
Barros (2004) analyzes the technical efficiency of Portuguese state-owned hotel chain by the stochastic frontier function in order to investigate the chain’s performance. The Cobb–Douglas cost functional form is comprised with three input prices (price of labor, price of capital, and price of food) and two outputs (sales and nights occupied). Chen (2007) also applies the stochastic cost frontier function to analyze 55 international tourist hotels’ cost efficiency in 2002. The average efficiency is 80.3%, implied that the almost 20% costs could be reduced without decreasing output. In addition, this paper uses the one-way ANOVA to examine the relationship between the factors of management type and hotels’ performance.
However, this method in Chen’s study has a disadvantage. Because of the first stage assumed that the parameter comes from the same distribution, but this assumption is violated in the second stage. The two-step procedure will give biased results (Wang and Schmidt, 2002).
Although the stochastic frontier approach overcomes some of the statistical limitations of DEA, but the multi-product cost functions must have the information of input price. In particular, the information of input price is difficult to collect.
These past studies employ the input-oriented model to analyze hotels’ operations, but this model is not suitable for the hotel industry. These studies neglect the quasi-fixed inputs, so that they may overstate firms’ capacity to adjust, misleading results, and weaken the DEA approach as a decision tool for managers (Ouellette and Vierstraete, 2004). The characteristics of the hotel industry are such that the
construction of international tourist hotels requires huge capital, various facilities and luxurious décor; these structures are not easy to change. Hence, the inputs, such as the number of rooms and the floor space of hotels are not easy to reduce or expand in the short term. In view of the problem, we employ the directional distance function and incorporate quasi-fixed inputs in the model to measure the efficiency by expanding outputs, contracting inputs, and fixed quasi-fixed inputs simultaneously.
In addition, most of those works estimate a frontier for all hotels from different management types, which implicitly assumes that hotels share a common production technology. However, because of different operational philosophy, managerial mode, human resource, financial capital and characteristics of the physical the different groups should have their own production technology. The meta-frontier approach allows for efficiency comparisons across groups which have different production frontiers. However, this problem can be solved by meta-frontier approach.
2.3 Meta-frontier
Battese and Rao (2002) combine the stochastic frontier approach with meta-frontier function to investigate the technical efficiencies of firms in different groups that may not have the same technology. There are there are two different data generation mechanisms in this study: one explains deviations between output of observations and group frontiers, and the other explains deviations between output of observations and meta-frontiers. This method has a problem that the meta-frontier may lie below on the group frontiers. Therefore, Battese et al. (2004) modify the model by using a single data generation processes and defining the meta-frontier that envelops the group frontiers.
O’Donnell et al. (2008) present the basic analytical framework for the meta-frontier, shows how a meta-frontier can be estimated using non-parametric (DEA) and parametric methods (SFA). They think that firms in different in regions face different production opportunities, so that they have different technology sets.
For example, firms product in different available stocks of physical, human and financial capital, resource and characteristics of the physical, social and economic environment in reality. In view of the opinion, the meta-frontier is as the boundary of an unrestricted technology set and group frontier is as the boundary of a restricted technology set.
Portela and Thanassoulis (2008) propose a new Malmquist-type index that is defined in relation to a meta-frontier and can be decomposed into efficiency change and boundary shift. The frontier is assumed to be convex, which implies true technology cannot be uninvented. The true technology is in reality feasible input output combinations, but which may not have been fully revealed