Productivity and Efficiency of Balinese Star-Rated Hotels in the Aftermath of
the Terrorist Attacks: A Malmquist Productivity Index Approach Based on
Data Envelopment Analysis
Abstract. This paper analyzes efficiency and productivity change of Balinese star-rated hotels
during 2004-2010, by employing data envelopment analysis (DEA) and a Malmquist total factor productivity index. A two-stage approach is also employed to see whether environmental variables influenced hotel efficiency. It is found that Balinese hotels technical efficiency level has increased from 64% in 2004 to 75% in 2010. The hotels technical efficiency was driven mainly by pure technical efficiency rather than scale efficiency. It is also found that total productivity change amounted to 7.9% annually, which was mainly due to the numbers of tourist nights increasingly sharply a couple of years after the detrimental effects of the two terrorist bomb attacks on the island of Bali.
Master’s Thesis Economics – August, 2012
I Made Agus Adnyana – 2137976
1. Introduction
The 2010 United Nations World Tourism Organization (UNWTO) annual publication reported that Indonesia was classified as one of the top ten countries in Asia and Pacific regions in terms of the number of visits and the tourism revenue generated. This implies that tourism has been recognized as an important industry in Indonesian economy. At the end of 2010, Badan Pusat Statistik (BPS)1
The promising features of Indonesian tourism sector are partly a consequence of the substantial role of Bali Island as one of the world’s most popular holiday destination
noted that foreign arrivals exceeded 7 million, a 40 percent increased compare to the data in 2005 which was only 5 million arrivals. In 2009, foreign tourist expenditure reached $7.3 billion meaning that per capita spending approached $ 1,085.
2
Bali’s economy is dependent on tourism and tends to specialize on the sector, like other “small islands in tourism industry” (SITE
. The small size of the island (5,633 km2 or just 0.3 % of Indonesian archipelago area) give significant contributions to the national tourism industry. Firstly, the share of the number of foreign visitors arriving directly to Bali through I Gusti Ngurah Rai International Airport has increased for the last five years. The share (expressed as a fraction of all foreign visitors to Indonesia) was 29.08%, 31.64% and 31.72% in 2005, 2008 and 2010 respectively. Secondly, Bali is the province with second largest number ofstar-rated hotels after Jakarta, with 13% of national figure (BPS, 2011). Finally, Bali is a hub for tourists to some other provinces of Indonesia such as Lombok and Yogyakarta. It was estimated that more than 60% of visitors to Indonesia go to Bali before going to Lombok, Yogyakarta or elsewhere (Asiamoney, 10 April, 2006).
3 in Giannoni and Maupertuis, 2007). Consequently,
its development is much more affected by trade, hotels, and restaurants as part of tourism industry. These divisions provide a dominant share of Bali’s GDP. Their share increased from 29 percent in 2004 to 31 percent in 2010 (BPS, 2010). This growth is also reflected in a 23 percent increase of hotel workers from 46.751 people in 2004 to 57.608 people in 20104 However, on 12th October 2002, one year after the 9/11 tragedy in the USA, bombs exploded in Kuta, a tourist destination which is located in the Southern part of Bali. Within a week of the attack, major tour operators swiftly withdrew their holiday programs from Bali and resorts throughout Indonesia. Bali’s hotel occupancy rate plunged from an average of 75% to 14%. Bali’s tourism started on to normal throughout 2004 and the first semester of 2005. In 2004 visitor arrivals to Bali peaked at 1.5 million, 44% higher than in 2003. But then a second suicide bombing in October 2005 in Kuta and its neighboring district Jimbaran killed 23
.
1 Central Bureau of Statistics of Indonesia
people. Following the second terrorist attack, in 2006 foreign tourist arrivals decreased about 15.8% from 2004, and the hotels’ occupancy rate also fell to around 48% (BPS, 2006).
Although Smyth et al. (2009) showed that those tragedies had just a transitory effect on tourist arrival to Bali, the occupancy rate of Bali’s hotels, especially star-rated hotel, were still difficult to increase. The occupancy rate of star-rate hotels in 2006 was only 44.46, and rose slightly to 53.32 in 2007 (BPS, 2008). As a result, oversupply of rooms and strong competition among hotels reduced the room rate, especially in 2009 when the global crisis arose, by offering bonus nights, meals or spa treatments to wealthy guests who are used to lavish lifestyles. Executive Director of Bali Hotels Association, Djinaldi Gosana, as quoted from Australian local newspaper The Age on Monday, Feb. 16, 2009 said that it was a bad idea, because it was difficult to raise the room rate afterward. In addition, the tourism development in other regions such as Lombok which has similar characteristics with Bali has also grown rapidly in the last three years. Moreover, in the last month of 2012 is estimated about 3,242 new rooms entered the market (Knight Frank, 2011) which would make competition very tight.
Facing such situations, the improvement of the hotel’s productivity and efficiency is inevitable. Formulating competition strategies, strengthening corporate operations and upgrading the quality of services are important for hotels in order to improve their competitiveness (Foo et al., 2011). When a hotel tries to improve its efficiency, it tries to maximize its benefit and profit, while minimizing its effort and expenditure. Hotel managers will sacrifice profits and be inefficient if they cannot manage the hotel correctly in order to reduce its costs, waste, and duplication.
From those facts and trends, two main questions emerge: What are the characteristics of Balinese star-rated hotels in term of productivity and efficiency, and how did these patterns change over the period 2004-2010? This thesis aims to answer these questions by employing two-stage approaches. In the first-stage, the components of hotels’ level of efficiency will be estimated by constructing data envelopment analysis (DEA) and growth of productivity will be evaluated using DEA-Malmquist productivity index. In the second-stage, econometric models will be used to examine whether some environmental variables (i.e., hotel properties that managers cannot affect, such as proximity to an airport) affect the hotel efficiency.
2. Literature Review
Over the last few years, measuring an industry performance has gathered major interest amongst researchers as it shows how well the resources are being managed by an organization. A significant effort has been made focusing on hospitality and tourism industry. In this part we will explain some definitions and concepts related to measuring of hotels performance.
2.1. Basic Concepts related to Productivity
A natural measure of performance is productivity. Productivity is commonly defined as ratio of outputs to inputs, where a larger ratio means better performance (Coelli et al, 2005). Actually, performance is a relative concept. For example, on a micro-level, a hotel performance in a year could be measured relative to its performance in a previous year, or could be measured relative to other hotel performance in the same year. The performance could also be evaluated at a higher aggregation level. For example, we can estimate the performance of the hotel industry over time or across geographical regions. In this thesis, we will focus on total factor productivity (TFP). Coelli et al (2005) describe TFP as a productivity measure involving all factors of production, not like partial measures such as labor productivity in a factory or land productivity in farming.
Figure 1. Productivity, Technical Efficiency and Scale Efficiency
Sherman and Zhu (2006) describe efficiency as the ability of a firm to produce the outputs with a minimum resource level required, i.e. “to do the job right”. Following Coelli et al. (2005), we try to illustrate the difference between productivity and efficiency. Suppose we want to produce a single output q using single input x, so output q is a function of input x (q=f(x)). There are four firms A, B, C and D. As depicted in Figure 1, f(x) is the production frontier which represents the maximum output attainable from each input level and hence reflects the current state of technology in the industry. Therefore, firms in this industry could not operate above the production frontier. The value of the production frontier f(x) is finite
D •
and non-negative. We can see in the Figure 1 that for all value of x represented on horizontal axis, the values of q are all non-negative. An additional unit of an input of production frontier will not decrease output. The marginal product of x is positive along the production frontier. Marginal product at a point is a slope of the production frontier at that point (q/x). This condition named monotonicity.
Firms in this industry are called technically efficient if they operate on the production frontier, and are not technically efficient if they operate beneath the production function. Therefore, firm A, B and C represents efficient firms. Firm D is called inefficient because technically, using the same input of x, D can produce output q more, and the same as the amount of firm C produce q.
The lines from origin to the particular point i.e. 0A, 0B, 0C and 0D (the slope of which is q/x) measure the productivity. The higher the slope, the higher the productivity. The line 0B is tangent with the production frontier and hence defines as the point of maximum possible productivity (most productive scale size). Point B has an optimal scale because operations at any other point on the production frontier would give a lower productivity. So, although point C and A are technically efficient, firms represented by these points still can improve their productivity. In this case A and C could be more productive by increasing and decreasing their scale of operation toward point B, respectively.
If we compare productivity over time, then we possibly have other source of changing in productivity, namely technical change. By implementing new advances in technology in time t, the production frontier may shift upward to the higher output. Thus, in time t, all firms can technically produce more output using the same level of input in previous period.
To sum up, improvement in productivity in this year compares to previous year, the source of its improvement may come from improvement in efficiency, technical change, exploitation of scale economies or the combination of these three factors.
2.2. Frontier-based studies in the hotel industry
number of employees, room expenditures, and food and beverage expenditures, and three outputs: revenue, physical and service satisfaction index. They found that general managers were operating with average 89% efficiency and the least efficient score in their sample was 64%.
Anderson et al. (1999) used the stochastic frontier approach in order to measure the managerial efficiency of 48 hotels in the United States in 1994. They considered the following variables as inputs: full-time equivalent employees, the number of rooms, total gaming related expenses, total food and beverage expenses, and other expenses. Total revenue was chosen as an output. They found that the hotel industry operating at an 89.4% efficiency score, whereas the highest and the lowest hotel efficiency score were 92.1% and 84.3%, respectively.
The Taiwanese hotel industry has been evaluated several times. First of all, Tsaur (2000) used DEA to investigate 53 international tourist hotels in Taiwan during 1996-1998 and obtained an 87% average efficiency score. He used seven inputs, i.e. the total operating expenses, the number of employees, the number of guest rooms, the total floor space of the catering division, the number of employees in the room division, the number of employees in the catering division, and the catering cost, and six outputs, i.e., total operating revenues, the number of rooms occupied, average daily rate, the average production value per employee in the catering division, total operating revenues of the room division, and total operating revenues of the catering division.
Secondly, Hwang and Chang (2003) utilized the DEA, and Malmquist TFP index which is a DEA-like linear programming to measure productivity change over time proposed by Fare et al. (1994). They measured efficiency on the basis of data collected in 1998 and the change in efficiency of hotels from 1994-1998. They used four inputs, i.e., number of full-time employees, guest rooms, total area of meal department, and operating expenses, and three outputs, i.e., room revenue, food and beverages revenue, and other revenues. Hotels were operating with 79.16% efficiency and only 20 out of 45 hotels increased their efficiency over the four-year period. The average decrease of efficiency from 1994 to 1998 was 4.2%.
Finally, Yang and Lu (2006) adapted DEA to evaluate efficiency of 46 international tourist hotels in Taiwan during 1997-2002 with four inputs, i.e. total operating expenses, number of employees, number of guest rooms, and total area of catering division, and six outputs, i.e. total operating revenues, average occupancy rate, average room rate, average production, value per employee in the catering division, and average production value of catering division. They found that the average efficiency score during 1997-2002 was 73%.
and number of beds, while used two outputs: total revenue and cost of a double rooms. He found that an average efficiency was 88%.
Data envelopment analysis was used by Neves et al. (2009) to select strategies that improve the performance of hotel companies. They used 83 worldwide hotel companies during the 2000-2002 time periods. For the output side, total revenue and EBITDA (earnings before interest, taxes, depreciation, and amortization) was chosen. Current asset, net fixed assets, shareholder’s equity and cost of goods and services were chosen as inputs. They found that the technical efficiency of hotels were 72.46%, 64.39%and 64.26%in 2000, 2001 and 2002 respectively. They argued that the decrease in hotel efficiency was mainly due to the 2001 economic crisis.
DEA Mamlquist TFP index was also used by Barros (2005) with panel data for 42 hotels of a Portuguese hotel chain during 1994-2001. He found that 38% of 42 hotels experienced productivity growth, while others faced a decline. He also found that efficiency change and productivity change, on average, were -3.15% and -7.5% respectively. He used three outputs: sales, number of guests and night spent in hotels, and five inputs: full-time employees, cost of labor, book value of property, operating cost and external cost.
The hotels efficiency studies above show considerable variation in their results, obtained by using various inputs/outputs variables. Multiple inputs and/or outputs were used in all above studies. Choosing between input and output indicators in hotel industry could be particularly challenging, because hotel is a service industry of which inputs and especially outputs may be mostly intangible. Renaghan (1981) in Avkiran (2002) suggests that it is better to evaluate hotel experience as a whole rather than in distinguishable components, which further compounds the measurement problem. Total revenue is used most as output in above studies. Except Anderson et al (2000), which used only total revenue as output because of number of output limitation of stochastic frontier technique, all other studies which use DEA used more than one variable as output to link to the business objectives.
From the input side, many different variables were used. Among the studies, the number of guest rooms and number of employees were the most frequently used. All of the studies at least used four input variables. DEA and SFA make this possible to calculate. However, independence of variables is also a consideration. For the input variables, this consideration is very important since variables which have very strong correlation may have impact to our result, in econometric named multicollinearity (Pedraja-Chappara et al., 1999). For example,
efficiency score so as to make the industry look good, he/she could reduce the number of observations and increase the number of inputs and outputs (Coelli et al, 2005).
Pedraja-Chappara et al. (1999) also showed the impact of sample size and number of input variables to the DEA efficiency score. Given the number of observations, increasing number of inputs will increase the proportion of units deemed efficient. And given number of inputs, lowering number of observations will increase the proportion of units deemed efficient. However, in literature studies that mention above, the number of observations used is no more than 55 and the highest is 83 hotels studied by Neves et al. (2009). And the number of inputs used is more than four. Although all those above studies have sufficient number of observations as suggested by Banker et al. (1989), i.e., number of observations is must be greater than three times of number of input and outputs, it is better to consider the impact of the use of too many input/output variables in relation to number of observations (Coelli et al., 2005).
Some external factors which can be assumed not to be under control of the manager may influence a firm’s efficiency. Fried et al. (1999) used the term “environmental” to describe those factors. Examples of the environmental variable are firm ownership, location characteristics and government regulation. As suggested by Coelli et al. (2005), some scholars such as Hwang and Chang (2003), Barros (2005), Yang and Lu (2006), and Huang et al. (2012) used a two-step approach to determine the effect of environmental factors on efficiency levels. Details about the two-step approach will be discussed in the next section.
Yang and Lu (2006) found that the management style of hotels (being part of an international chain or not) has a significant effect to the efficiency. They found that international chain hotels have a positive effect to efficiency in Taiwan, and have higher average efficiency score than independent hotels. The rationale is that international hotels have a sounder reputation, better brand image, internet marketing, efficient reservation system, and economies of scale (at the chain level).
3. Data and Methods 3.1. Data
3.1.1. Balinese Star-rated Hotel Overview
Based on Statistics of Indonesia (BPS), hotels are classified into ‘classified hotel’ and ‘non-classified hotel’. Classified hotels include all star-rated hotels, while non-‘non-classified hotels involve jasmine-rated hotels5, youth hotels, homestays, camping sites and other accommodations. A star-rated hotel is an establishment which uses a building or part of building especially prepared to any person to stay, eat and obtain service as well as other facilities against payment, and it has fulfilled the requirements as star hotel which have been stated by the Regional Tourism Agency. The special characteristic of a hotel is having restaurant under hotel management. Based on the Indonesia Hotels and Restaurants Association (IHRA) and Ministry of Tourism, star-rated hotels are classified into 5 classes dependent on the hotel quality, which are 1 star, 2 star, 3 star, 4 star and 5 star. According to IHRA, there are two determinants for classifying a hotel into a star-category. First are basic requirements that include having a license, technical feasibility of installation and equipment use, hygiene and sanitation. Second is technical operation that includes physical, management and service operations. The higher the star category is, the higher the quality of hotel is in terms of these two determinants.
Almost all of star-rated hotels are resort hotels which are spread along the Southern part of Bali Island. Kuta, Nusa Dua (both are located in Badung regency) and Sanur (located in Denpasar) with its beaches are the main site. Almost 70% of total hotels are located in that area (see Figure 2). The Northern part of Bali including Lovina which is located in Buleleng regency has few star-rated hotels. This part also depends on its beaches to attract the tourists. Meanwhile, the Eastern part of Bali such as Ubud which is located in Gianyar regency is famous with its “terassering”, a traditional irrigation system, and its art markets. The Nusa Lembongan islands, which are located on Bali’s East coast (Klungkung regency), are renowned for their marine life, snorkeling and diving locations.
5An establishment which uses a building or part of building especially prepared to any person to stay,
Figure 2.Map of Bali Islands.
Source :www.mapsbali.com
Table 1.Descriptive statistics for star-rated hotels in the period 2004-2010. Year 2004 2005 2006 2007 2008 2009 2010 1-Star Hotels 12 9 11 9 11 9 9 Rooms 340 1435 331 283 568 277 446 OCR 24.99 27.36 39.68 36.47 38.29 42.53 52.66 2-Star Hotels 22 36 30 34 35 27 26 Rooms 1228 2352 1871 2079 2449 1782 1676 OCR 39.34 39.46 38.08 49.57 51.08 47.41 52.79 3-Star Hotels 33 36 38 38 39 35 35 Rooms 3104 3537 3540 3354 3405 2299 2485 OCR 37.81 42.11 36.62 48.34 61.19 59.55 58.77 4-Star Hotels 31 29 30 28 28 41 48 Rooms 3689 4562 4218 3923 3938 5349 6064 OCR 50.21 47.46 43.22 52.12 62.55 59.22 66.32 5-Star Hotels 35 36 38 36 37 37 37 Rooms 9578 8224 10044 9306 9880 8977 10462 OCR 53.84 49.8 49.14 57.31 66.94 60.92 60.13 All-Star Hotels 133 146 147 145 150 149 155 Rooms 17939 20110 20004 18945 20240 18684 21133 OCR 48.64 46.4 44.46 53.32 62.77 59.00 60.16
Note : Hotels=number of hotels ; Rooms=number of rooms ; OCR = Hotel occupancy rate Source : BPS
1-star hotels has the lowest and 5-star hotels had the highest average for number of hotels and number of rooms during the period 2004 to 2010. Meanwhile, 2-star, 3-star and 4-star hotels had about the same the number of hotels. However, for the average number of rooms available, 4-star hotels had a higher figure than 3-star hotels, and 3-star hotels had a higher figure than 2-star hotels, which reflects that higher-quality hotels were generally bigger. Interestingly, although 1-star hotels have the lowest occupancy rate, however, it seems that these hotels did not suffer from the impact of the Bali bomb 2. While other hotels suffered from this tragedy in terms of decreasing occupancy rates in 2006, 1-star hotels enjoyed a significant increase. The main reason for this phenomenon is that most of the customers of 1-star hotels are domestic guests, and hotels likely have a lower room rate than its competitors. 5-star hotels dominated the occupancy rate during the period of 2004-2009, but in 2010, it was taken over by 4-star hotels.
3.1.2. Database Construction
exclude three hotels in 2004 and six hotels in 2010 because those hotels have incomplete data for some of variables used in this study. Thus, we have 130 hotels and 149 hotels in 2004 and 2010 respectively.
In order to keep the homogeneity of these hotels for sensible comparison, I only choose the star-rated hotels. This paper differs from previous study such as Min et al. (2008) who used four-diamond above and also considering the number of annual room capacity to estimate the hotel productivity in Taiwan, while Barros (2005) used only small chain hotels.
3.1.3. Input and Output Selection
The main activity of the hotels is to offer guests a place to sleep and to serve them with food and beverages. However, the hospitality industry has made significant progress in the past decade. The basic fulfillment of customers’ needs is no longer enough. In order to survive, hotels are implementing various services to impress their guests. Services such as convention venues, social activities, entertainment, shopping facilities, spa and wellness facilities etc. are becoming ever more important besides accommodation and catering. Nevertheless, the main process of a hospitality unit can still be represented as the conversion of inputs of various resources into output, like for any other firm. Output is a concrete measurement showing the extent to which an organization has reached its objectives (Yang and Lu, 2006). In this study, output is represented by total revenue. Total revenue refers to revenue from lease of rooms and other revenues such as from food and beverages, conferences etc. which significantly influences the financial efficiency of hotels.
Considering the Pedraja-Chappara et al. (1999) study and Coelli et al.’s (2005) suggestion as discussed in the previous section, we considered to choose only three inputs to produce the output. This number is much lower in comparison to number of observations than in previous studies in the literature. All of these had fewer observations than we have, but more in number of input/output variables. As a consequence, we are much less likely to find artificially high efficiency levels.
of the hotel in terms of room numbers, but also reflect other services offered by hotels, such as entertainment, swimming pools, etc. Both financial data, total revenue and other input, are deflated to constant prices for 2000, using the implicit gross domestic product (GDP) deflator. Only for the data 2010, I have to estimate total revenue and other input because there was a change in the VHT-L questionnaire6. Table 2 shows descriptive statistics for all the
variables in the model.
Tabel 2. Descriptive statistics of inputs/output variables
2004 2010
TR1 NSTDR1 NSUIR1 OTHI1 TR1 NSTDR1 NSUIR1 OTHI1
1-star # Obs 12 12 12 12 8 8 8 8 Mean 2213.79 40 3 544.05 6169.79 24 13 1505.01 Std. Dev 3699.08 22 1 840.27 6457.65 12 12 2012.64 Min 72.00 12 1 19.03 607.21 9 1 72.16 Max 13649.88 97 6 3117.97 20342.34 48 32 6147.66 2-star # Obs 22 22 22 22 22 22 22 22 Mean 3245.73 43 9 798.28 6699.44 52 10 1166.75 Std. Dev 3510.93 28 10 816.53 9876.24 53 10 1660.20 Min 48.38 12 1 6.84 198.67 6 1 22.79 Max 15408.92 115 35 3658.17 44224.52 271 35 7666.66 3-star # Obs 32 32 32 32 35 35 35 35 Mean 7423.20 91 10 2154.26 10505.56 65 8 2161.79 Std. Dev 12384.27 76 18 3505.57 20941.12 42 9 4455.54 Min 437.61 8 1 102.98 746.06 9 1 96.17 Max 70155.48 395 100 16648.09 124315.00 170 33 25330.48 4-star # Obs 30 30 30 30 47 47 47 47 Mean 14591.26 111 18 3393.32 31833.52 110 18 6309.61 Std. Dev 20202.83 79 18 4502.70 48464.44 91 27 13374.51 Min 507.18 16 1 80.95 538.07 7 1 86.30 Max 79486.44 329 84 19583.57 248388.40 397 138 86431.22 5-star # Obs 34 34 34 34 37 37 37 37 Mean 56189.10 242 31 12446.48 83141.21 218 29 13997.00 Std. Dev 51198.27 159 37 10811.46 65584.38 148 38 12737.20 Min 2848.45 13 2 579.84 3245.22 13 2 358.48 Max 202388.90 561 189 41082.57 240367.10 561 189 64281.70 all-star # Obs 130 130 130 130 149 149 149 149 Mean 20643.70 122 17 4753.90 34475.44 113 17 6226.93 Std. Dev 35643.28 123 25 7720.23 52735.44 114 26 11145.33 Min 48.38 8 1 6.84 198.67 6 1 22.79 Max 202388.90 561 189 41082.57 248388.40 561 189 86431.22
1 TR=total revenue (in million rupiah) ; NSTDR=number of standard rooms ; NSUIR=number of suite rooms ;
OTHI=other input (in million rupiah)
6For the revenue, I estimated it using number of guest, average length of stay both for foreign and domestic
Statistics of all input and output variables of the hotels appear to increase with the numbers of stars of hotels. We also can see that the average total revenues and other inputs of all star-rated hotels in 2004 were lower than in 2010. The growth of output, however, is higher than growth of average other input which is one input factors in our production function. Moreover, the average number of standard rooms also experienced a decreasing trend, except for 2-star hotels which showed an increase from 43 in 2004 to 52 in 2010. 3-, 4- and 5-star hotels had a lower number of suite rooms in 2010 than those in 2004, which is different for 1-star and 2-star classed hotels, which have a higher average in 2010 than that in 2004.
Table 3 shows the matrix with correlations of inputs and output. Our data show a lower correlation among inputs. This will lessen redundancy and lower the tendency to incur multicollinearity in our analysis. This condition is appropriate with our discussion in Section 2 about impact of multicollinearity. This matrix correlation also shows that all inputs have positive correlation to output. So, additional one input would generally not reduce the value of output, thus the monotonicity property of production function is satisfied.
Table 3.Correlation coefficients among inputs and output.
2004 2010
TR* OTHI* NSTDR* NSUIR* TR* OTHI* NSTDR* NSUIR*
TR* 1 1
OTHI* 0.9808 1 0.9125 1
NSTDR* 0.6841 0.6677 1 0.6919 0.4602 1
NSUIR* 0.5389 0.4992 0.5354 1 0.3732 0.2284 0.4118 1
*TR=total revenue; OTHI=other input ; NSTDR=number of standard rooms ; NSUIR=number of suite room
3.2. Methods
3.2.1. Data Envelopment Analysis (DEA)
DEA identifies the inefficiency in a particular hotel by comparing it to similar hotels regarded as efficient. There are two orientations when measuring technical efficiency i.e. input and output orientation. Input-oriented technical efficiency comes up from the question: “by how much can input quantities be proportionally reduced without changing the output quantities produced?”. In contrary, output-oriented technical efficiency comes up from the question: “by how much can output quantities be proportionally expanded without altering the input quantities used?”. The difference between these two orientations is easier to illustrate using one-input x and one-output y as depict in Figure 3 (Coelli et al., 2005).
Figure 3. Input- and output-oriented efficiency
As depicted in Figure 3a, f(x) is the production function and an inefficient hotel is operating at point P. If a hotel operates at point P with C amount of input and A amount of output, then the technical efficiency (TE) of that hotel for input-oriented is the ratio AB/AP, while TE for output-oriented is the ratio CP/CD. If we assume our production function has the property of constant returns to scale (outputs changes induce proportional changes in input) as depict in Figure 3b, TE input-orientation will be the same with TE output-orientation, AB/AP=CP/CD. Now, associated with our input and output variables that we have, assume that we have the data for three inputs and one output for each of n hotels. For the i-th hotel, these are represented by column vectors xi and scalar yi. Let X denote the (nx3)-input matrix and y the
(nx1)-output vector with observation for all hotels. Assume our model is constant return to scale, then output-oriented DEA solves the following linear programming problem for firm in each year (Coelli et al, 2005)7
7Tim Coelli also provides computer software, named DEAP, to solve the linear programming problem. We use
this software in our study.
: maxθI ……….(1) s.t. -θiyi + y’λ≥ 0, xi – Xλ≥ 0, nI’λ=1, λ≥ 0,
where I is (nx1)-summation vector containing ones, λ is an (nx1)-vector of constants, θi are
scalars, 1≤θi≤∞ and prime denote transposed vector. Given input quantity held constant, 1-θi
denotes the proportional increase in output that could be achieved. 1/θidefine a TE score
Banker et al. (1984) suggest an extension of the constant return to scale (CRS) DEA model, which is known as the BCC model, to deal with situations of variable returns to scale (VRS). Since not all firms are operating at the optimal scale, they may further decompose the overall technical efficiency into pure technical efficiency (PTE) times scale efficiency (SE). Back to Figure 1, A, B and C are efficient in VRS, however only B is efficient in CRS because A and C does not operating at the optimal scale.
In the VRS model, there is one differentiation from CRS by adding the convexity constraint, nI’λ, to equation 1. Moreover, the relation among the TE, PTE, and SE scores is:
TE = PTE × SE
TE under constant returns to scale is also called as (global) technical efficiency since it takes no account of scale effect as distinguished from PTE. On the other hand, BCC expresses the (local) pure technical efficiency (PTE) under variable returns-to-scale circumstances (Cooper et al., 2007). Figure 5 shows that A is operating locally efficient (PTE=1). Under input oriented, globally technical efficiency (TE) is measured by its failure to achieve scale efficiency (SE) as represented by LM/LA. Point A in Figure 1 depicts the same condition. B and C have SE=1 i.e. they are operating in most productive scale size (MPSS). G also has SE=1, thus it has TE=PTE which is lower than one. SE for point E can be measured by (PQ/PE)x(PE/PR) = PQ/PR.
Figure 5. Scale efficiency
3.2.2. Two-stage approach
double-censored Tobit will be the best choice. Coelli et al. (2005) suggested of using two-stage approach because of the following advantages:
1. It can accommodate more than one variable.
2. It can accommodate both continuous and categorical variables.
3. It does not make prior assumption regarding the direction of the influence of the environmental variables.
4. One can conduct hypothesis tests to see if the variables have a significant influence upon efficiencies.
5. It is easy to calculate, and
6. The method is simple and therefore transparent.
This method can also be used to assess the influence of various management factors upon efficiency such as age, education, training etc. The one disadvantage of this method is that the results are likely to be biased if the variables used in the first-stage are highly correlated with the second-stage variables (Coelli et al., 2005). Fortunately, in this study we used variables which are independent with variables in the first-stage. All the environmental variables and input variables have low correlations (see Table A and B in the Appendix).
A Tobit regression model will be used to identify potential antecedents of efficiency of Balinese hotel industry.
𝑇𝑇𝑇𝑇𝑖𝑖 = α + β1(𝐼𝐼𝐼𝐼𝑇𝑇𝐼𝐼𝑖𝑖) + β2(𝑆𝑆𝑇𝑇𝑆𝑆𝑆𝑆𝑖𝑖) + ε𝑖𝑖
In estimating above model, this study uses technical efficiency (TE) as dependent variable. The independent variables are dummy variable of international chain hotels (INTC) as a proxy of managerial style, and the star-rating of the hotel (STAR). INTC is a dummy variable for international chain hotel (1=international chain and franchise, 0=independent).
4. Static Analysis
4.1. Hotels Efficiency in 2010
Table 4 shows the basic result of our frontier estimation in 2010, when the Balinese tourism condition was relative stable. The hotels are ranked by number of standard rooms, in ascending order.8
Hotel
For all hotels efficiency result can be seen in the Appendix-Table C. Hotels which have technical efficiency equal to one (these appear on the frontier), are globally efficient hotels relative to the hotels in the sample. These hotels are the reference for inefficient hotels because they are not only superior in management and organization of hotel but also appropriate in scaling their activities. Management of inefficient hotels can benchmark the operation of these hotels and compare their source allocation and output. Table 4. Frontier hotels in 2010
Star1 MS2 Year3 Ratio of Other input4 (%) Proportion of suite room5 (%)
Hotel Star1 MS2 Year3
Ratio of Other input4 (%) Proportion of suite room5 (%) H0402 5 1 0 36.2 70.0 H9007 5 1 1 12.1 5.6 H0359 2 0 0 30.2 78.0 H9008 3 0 0 20.4 1.7 H0702 5 1 0 43.3 53.6 H0338 4 0 0 10.6 15.7 H0401 5 1 1 18.7 6.3 H0354 4 1 0 34.8 3.1 H0409 4 0 1 11.5 2.9 H0330 5 1 0 26.7 5.4 H0406 4 1 1 17.2 2.9 H0379 5 1 0 13.2 6.3 H9003 3 0 0 10.4 9.8 H0334 5 1 0 10.3 7.2 H0403 4 0 1 11.6 12.5 H0310 4 0 1 11.3 3.8 H9014 5 1 0 28.6 3.3 H7103 5 1 0 11.3 2.6 H9017 5 1 0 15.1 57.1 H0317 5 1 0 12.3 8.1
1Number of star of the hotel
2Managerial Style: 1= International chain hotel ; 0= Independent hotel 3Year : 1=operational year > 2004 ; 0=operational year <= 2004. 4Ratio of other input=Other input/Total Revenue.
5Proportion of suite rooms=Number of suite rooms/Total rooms available.
First of all, 5-star hotels dominated the frontier. 11 out of 20 frontier hotels are 5-star hotels, spreading from small to big hotels. Table 4 also shows that 3- and 4-star hotels appear on the frontier 2 and 6 times respectively and most of them are medium size hotel. 1-star hotel did never appear on the frontier, and only one 2-star hotel appears on the frontier.
On hotels size comparison, small size hotels which are on the frontier tend to have higher proportion number of suite rooms to total rooms available. For example, H0402, H0359 and H0702 have proportion of suite rooms of more than 50%. Since the rates of suite rooms are usually much higher than for standard ones, they may generate higher revenue while incurring cost for maintenance this room is not much different than standard ones. However, the big categorized hotels have different combinations. They have much lower proportion
number of suite rooms, with the exception of H9017. The last seven frontier hotels, on the average, have no more than 10% proportion of suite rooms.
For managerial style, international chain hotels dominated the frontier with 65% appearance. Small, medium and big size hotel are all represented on the frontier. Compared to number of all international chain hotels in the sample, 30% of them are on the frontier.
There are six hotels on the frontier which officially started to operate after the first year of our study. Three of them are international chain hotels, and four for them are 4-star hotels. So, although some hotels can be categorized as new entrants because they officially started to operate after 2004, these hotels could also able to be a benchmark for other hotels in 2010.
It is interesting then to evaluate how new entrant hotels performance compare to continuing hotels. Table 5 shows the performance of two groups of hotels, the new entrants which are hotels that operated after 2004, and continuing hotels which are hotels that operated in 2004 (or before). New entrant hotels have lower average technical efficiency than continuing hotels but the different is relative small which is only 2.39%. However, the lowest TE score reached by new entrant hotels is 38.20% while continuing hotel is 45.30%, showing a larger dispersion for the former.
Table 5. TE, PTE and SE of new entrant and continuing hotels
TE PTE SE
New Entrant Average (%) 73.27 77.45 94.57
# obs :38 Std. Dev (%) 19.02 18.50 8.62 Min (%) 38.20 42.30 63.90 Max (%) 100.00 100.00 100.00 Continuing Average (%) 75.66 79.43 95.75 # obs :111 Std. Dev (%) 16.66 16.87 9.28 Min (%) 45.30 45.90 48.40 Max (%) 100.00 100.00 100.00
4- and 5-star hotels TE are above the overall average (75.1%). 4- and 5-star hotels also have pure technical efficiency and scale efficiency above the overall average. These show that, on average, 5- and 4-star hotels are superior compare to other star-numbered hotels. The quality of service may influence inefficiency since 5-star hotels have the best quality. Service quality is related to customers’ satisfaction, which is likely to lead to more output.
Comparing the scale efficiency of the hotel, it can be concluded that the 5 star hotels which have more guest rooms tend to exhibit decreasing return to scale (DRS). On the contrary, most of the 1 and 2 star hotels experienced increasing return to scale (IRS).
There are five inefficient hotels that operated on an efficient scale. These hotels are efficient in scale but still technically inefficient (see point G in Figure 5). In order to be efficient hotels, they should focus on PTE. However, some hotels can be locally efficient, i.e. hotels that have PTE equal to one but failure to achieve scale efficiency. 14 hotels are in this circumstance. These hotels have good hotel organizational factor associated with hotel management, but are globally inefficient hotels because they are too large or too small.. Thus, in order to reduce its inefficiency, they have to optimize the size of operations by considering the return to scale. However, it is more difficult to reduce scale inefficiency than pure technical efficiency, for example hotels have difficulties to reduce their numbers of guest rooms. It is not easy to sell the rooms to other parties. Therefore, it makes sense to deal with scale inefficiency only when a hotel turns out to be technically efficient (Avkiran, 2002).
Following the two-stage approach, after obtaining hotels efficiency levels using DEA in the first-stage, we now attempt to evaluate the environmental variable that influence those efficiency in the second stage. We regresses the efficiency of all hotels on environmental variables that we have discussed in Section 3, using the double-censored regression Tobit model. Table D in the appendix shows that the correlations between independent variables are below 0.5, which shows that multicollinearity is probably not a problem. The Tobit results are showed in Table 6.
Table 6.Tobit regression results, 2010.
Variable Coefficient Std. Err. t- statistics P>|t|
INTCH 0.161 0.033 4.820 0.000
STAR 0.043 0.013 3.310 0.001
Constant 0.564 0.044 12.790 0.000
Log likelihood = 32.12; Prob> chi2 = 0.0000
centralized reservation systems, among other benefits.
In line with INTCH, the star-number (STAR) variable has a positive significant coefficient at 1%. It means that the higher the number of stars, the more efficient the hotels are, on average. This seems to make sense, since the quality of technical operations such as management and service operations are a criterion by IHRA to classify a hotel into classes with a specific number of stars, as discussed before.
4.2. Hotel Efficiency in 2004
In this section we try to evaluate how Balinese star-rated hotels performed in terms of efficiency six year before our previous analysis. The Balinese tourism condition in this year was different from 2010. As mentioned in Section 1, this year was the time in which Balinese tourism attempted to recover from the 1st Bali bombing. First of all, as our previous analysis, we show the result of our frontier estimation in 2004.
Table 7. Frontiers hotel in 2004
Hotel Star1 MS2 Ratio of Other input3
(%) Proportion of suite room 4 (%) H0402* 3 1 13.3 71.4 H0702* 5 1 31.4 60.6 H0501 3 0 14.1 10.0 H0403* 3 0 17.8 12.5 H0360 5 0 21.6 11.3 H0354* 4 1 18.2 3.1 H0330* 5 1 23.1 5.4 H0334* 5 1 9.6 7.7 H0310* 4 0 17.2 5.8 H0386 5 1 21.8 4.1
1Numbers of stars of the hotel
2Managerial Style: 1= International chain hotels ; 0= Independent hotels 3Ratio of other input=Other input/Total Revenue.
4 Proportion of suite rooms=Number of suite rooms/Total rooms available *Efficient hotel in 2010.
international chain management style, while the rest are independent 4-star hotels. The efficient hotels in 2004 that were inefficient in 2010 represent small, medium and big size hotels. The change of the frontier was mainly for the medium size hotels, since more hotels appeared efficient in 2010 in this hotel size.
For the size of efficient hotels, the frontier in 2004 had a similar pattern to the frontier in 2010. Small size efficient hotels tended to have higher proportions of suite room in comparison to the big size hotels. Smalls hotel with high proportions of suite rooms are usually small luxury hotels.
Table E in the Appendix shows the technical efficiency (TE) of all hotels, which we decomposed into pure technical efficiency (PTE), and scale efficiency (SE) scores. In 2004, average TE, PTE, and SE of all hotels is 63.8%, 70.9%, and 91.0%, respectively. The pattern is similar to 2010, but the scores are lower. In summary, in 2004 star-rated hotels in Bali operate less efficient than in 2010. The 1st Bali bombing influenced the tourism condition, and resulted in lower hotel productivity.
The highest impact of uncertainty about tranquility in Bali in this year was experienced by 4-star and lower hotels. 4-4-star hotel’s TE is 62.8%, which is less efficient than in 2010 which TE is 77.8%. Lower-numbered star hotels also experienced the same pattern. For the scale efficiency, there was only one inefficient hotel that operated on the most productive scale. Compared to 2010, this number is much lower.
For 2004, we also ran the censored the Tobit model using the same environmental variables as exogenous variable as in 2010, and technical efficiency as endogenous variable. Low correlation between independent variable is also shown in Appendix-Table F. Again, the log likelihood ratio rejected the null hypothesis that the all coefficients equal to zero at the 1% level of significance. The result can be seen in Table 8. For the dummy variable international chain hotel (INTCH), and number of stars (STAR), both have a positive significant coefficient at level 1%. This is similar to result in 2010. In 2004, both INTCH and STAR variables also influence hotel efficiency positively.
Table 8.Tobit regression result, 2004.
Variable Coefficient Std. Err. t P>|t|
INTCH 0.132 0.043 3.050 0.003
STAR 0.051 0.015 3.330 0.001
Constant 0.433 0.051 8.430 0.000
Log likelihood = 8.69 ; Prob> chi2 = 0.0000
hotel minimum TE score is 17.5%. The maximum TE that was reached by an exiting hotel is 47.10% which is much lower than overall TE (63.8%).
Table 9. TE, PTE and SE of exited and continuing hotels
TE PTE SEC Exited Average 31.09 37.62 89.30 # obs :19 Std. Dev 9.69 19.90 18.52 Min 17.50 18.80 36.50 Max 47.10 100.00 99.70 Continuing Average 68.05 75.27 91.27 # obs :111 Std. Dev 18.85 19.55 12.23 Min 30.90 31.20 32.50 Max 100.00 100.00 100.00 5. Dynamic Analysis
In this section we try to evaluate the change in hotels efficiency and productivity over the period 2004-2010. We use a DEA-Malmquist productivity index to measure the change in hotel productivity. It can be broken down into the effects of changes in technical efficiency and in technological change. First of all, a brief explanation of the DEA-Malmquist productivity index is presented.
5.1. DEA-Malmquist TFP Index
The Malmquist productivity index can be utilized to compare productivity performance between two time periods, in our study between 2004 and 2010. The index was introduced by Caves et al. (1982), based on the idea of Malmquist (1953). The decomposition of the index into its components was developed by Fare et al. (1992, 1994). The concept of a distance function (Shepard, 1953) applied in the index can be estimated by data envelopment analysis (DEA) technique.
Malmquist based his index on the output distance function D (subscript “0” use to indication of output-oriented), defined as:
𝐷𝐷0𝑇𝑇(𝑥𝑥𝑡𝑡, 𝑦𝑦𝑡𝑡) ≡ 𝑖𝑖𝑖𝑖𝑖𝑖 �𝜃𝜃: �𝑥𝑥𝑡𝑡,1𝜃𝜃 𝑦𝑦𝑡𝑡� ∈ 𝑆𝑆𝑡𝑡�
Where x denotes a vector of inputs; y, a vector of outputs; St, the technology set; superscript
T, the technology reference period, usually T =t or T = t + 1; and 1/ϴ, the amount by which
2 / 1 1 1 1 1 1 1 1 1 ) , ( ) , ( ) , ( ) , ( ) , , , ( = ++ + + + + + + t t t O t t t O t t t O t t t O t t t t O y x D y x D y x D y x D y x y x M (2) ) , ( ) , ( 1 1 1 1 t t t O t t t O y x D y x D + + + + (2a) ) , ( ) , ( 1 1 t t t O t t t O y x D y x D + + (2b)
Equation 2a refers to the technologies of period t+1, while considering the amount of production factors in t and t+1, while equation 2b is expressed in terms of technologies of period t, while considering the amount of production factors in t and t+1.
The Malmquist index involves two components: efficiency change (ECH) and technical change (TECH). The formulation of the two can be shown as :
) , ( ) , ( 1 1 1 t t t O t t t O y x D y x D ECH + + + = (3) 2 / 1 1 1 1 1 1 1 ) , ( ) , ( ) , ( ) , ( = + ++ ++ t+ t t O t t t O t t t O t t t O y x D y x D y x D y x D TECH (4)
Improvement in total factor productivity yields Malmquist index values greater than unity. Deterioration in performance over time is associated with a Malmquist index less than unity. The same interpretation applies to the values taken by the components of the overall TFP index. Improvements in the efficiency component yield index values greater than one and are considered to be evidence of catching up (to the frontier). Values of the technical change component greater than one are considered to be evidence of technical progress. While the product of the efficiency and technical change components must, by definition, equal the Malmquist index, and those components may be moving in opposite directions.
5.2. Result and Analysis
Table 10. TFP, Technical Change and Efficiency Change of Bali’s hotels, 2004-2010
# obs Effch1 Techch1 Pech1 Sech1 Tfpch1
1-star 4 1.243 1.186 1.217 1.021 1.473 2-star 12 1.203 1.195 1.051 1.145 1.438 3-star 18 1.088 1.307 1.138 0.956 1.423 4-star 26 1.229 1.275 1.135 1.083 1.568 5-star 32 1.108 1.297 1.061 1.045 1.438 All-star 92 1.155 1.274 1.102 1.049 1.472
1Effch=Efficiency change ;Techch=Technological change ; Pech=Pure efficiency change ; Sech=Scale efficiency change ; Tfpch=Total productivity change.
From 2004-2010, the average Malmquist productivity index change was 47.2%, which corresponds to an annual growth of 7.9%. This high TFP growth can be associated with the change of Balinese tourism conditions, from a very bad condition when Bali had been bombed by terrorists to a good condition where occupancy rates increased about 25% and the average length of hotel stays increase by more than 10%. On the average, total productivity growth was due to improvement in technological change (techch) rather than improvement in efficiency (effch). Improvement in efficiency was 15.5% (2.6% annually) which means that on the average inefficient hotels were getting closer (catch-up) to the global frontier hotels. Apparently, inefficient hotels were able to increase their level of technical efficiency by implementing their potential technology. Improvement in technological change was 27.4% (4.6% annually), an indication that the frontier shifted upwards at a quick pace. In the previous section, we identified the type of innovator hotels that made the technology shift upward (the hotels that were in frontier in 2010).
The technological change is associated with innovation which is not only the adoption of the new technology but also the new management system adopted in the process of hotel operation. It also includes the benefit of using the advance of information and communication technology (ICTs), i.e. by joining the worldwide hotel community such as TripAdvisor and Agoda to gain the acceleration of system of advertisement. This new technology and new management system improved the time of communication between hotel and customer, reducing the searching time to find hotels that use this new system. At the end, such hotels will experience higher occupancy rates compared to hotels that cannot innovate to capture the change of consumer preferences in the world high technology advancement. These hotels therefore managed to benefit more from the improved tourism climate after the fear of more bombs had slowly faded away.
This could be an indication that the potential for five-star hotels efficiency growth was exhausted during the long time period these hotels had been on the frontier. As a consequence more innovation is needed to increase productivity, especially innovations in procedure or methodology of activities to optimizing the potential of capital. On the average, 4-star hotels were also the best at moving toward the frontier, followed by 1-, 2-, 5- and 3-star hotels. However, on technological change, 1-3-star hotels were the lowest but still enjoyed a positive effect. Adopting and implementing new technology that was applied by frontier hotels provided opportunities for this catch-up.
Individually, 87 out of 92 hotels (95 percent) demonstrated an improvement in total productivity, indicated by TFP change >1 (see Appendix-Table G). The rest experienced a regression in total factor productivity. 62 hotels (67%) exhibited improvement in efficiency change, 6.5% of hotels have no change and the rest (26%) was in decline. Improvement in technological change was experienced by 96% hotels, while the rest, 4 hotels, were experienced regression.
Following the two-stage approach, here we evaluate the environmental variables discussed in Section 2. However, here we use OLS regression model rather than the Tobit because now our dependent variable is efficiency change which values are not limited in zero and one as efficiency at level. We also include initial technical efficiency (ITE) which is TE in 2004 as our independent variable in order to capture possible inherent efficiency change performance in a hotel. Low correlation among independent variables is shown in Appendix-Table H. The OLS result can be seen in Table 11.
Table 11. OLS regression result
Variable Coefficient Std. Err. t P>|t|
ITE -1.344 0.169 -7.980 0.000
INTCH 0.178 0.070 2.550 0.013
STAR 0.022 0.029 0.770 0.445
Constant 1.976 0.127 15.580 0.000
R2 = 0.43 ;Prob>F = 0.000
6. Conclusion
This paper provides a DEA analysis and DEA-Malmquist productivity index to investigate the characteristics of Balinese hotels in term of efficiency and productivity in two time periods. First is in 2004 when the Balinese tourism was in a difficult situation after the first Bali bombing tragedy. Second is in 2010 when the tourism situation was relatively stable. This analysis offered three insights.
First, although Balinese hotels efficiency level have increased in 2010 compared to 2004, the industry is still rather inefficient. The mean efficiency levels were 64% and 75% in 2004 and 2010 respectively. All the time period, on average, pure technical efficiency is much lower than scale efficiency. The implication for this result is twofold. First, in order to increase the efficiency level, hotel management focus should be on the pure technical efficiency and not on the scale, because the first focus is characterized by higher potential percentage increases. Moreover, it is more difficult to reduce scale inefficiency than pure inefficiency, thus it makes sense to deal with scale inefficiency only when a hotel has become technically efficient. For the big size hotels which most operating in decreasing returns to scale i.e. a decrease in the inputs will have a smaller proportional impact on the output, thus increasing efficiency should be done by reducing the size of the operations. For this implication, management should identify which hotel units have lower efficiency than average hotel efficiency which then can be sold if the net present value is equal or lower than selling price. However, for the small size hotels which in region of increasing returns to scale i.e. an increase in the inputs will have a higher than proportional increase on the output, thus enlarging the size of the operation could be the option. Hotels on the frontier, can be seen in Table 4 and 7, can be used as benchmarks for small, medium and big size hotels. Second, tranquility makes an important role on tourism industry since tourism is associated with leisure. The result that inefficiencies had been reduced in 2010 is evidence for this, after the tragedies that had feared tourists to come to Bali. This suggests that the government role on keeping the tranquility should be kept.
Second, Balinese hotels experienced high productivity growth over the period 2004-2010. All of the components of total factor productivity growth also experienced increases. Productivity growth was more driven by technological change rather than by efficiency change. However, some of these innovator hotels were also efficient hotels in 2004. Thus, some inefficient hotels found difficulty to catch-up to the new technology implemented by frontier hotels. Investment in organizational factors associated with hotel management such as marketing initiatives and improvement in quality should be a focus of Balinese hotel managers.
number of stars of a hotel, the higher its efficiency, and the international chain hotel management style also leads to higher efficiency. This finding can be related to the previous analysis when frontiers are dominated by 5-star hotels and international chain hotels.
In summary, the results must be seen as the first attempt to shed new light of on efficiency and productivity of hotel industry in Bali. However, the availability of the data is a limitation, which did not only affect the selection of input and output variables, but also on the time period that could be studied. Further researches are needed to come up with long time periods to capture the golden era of Balinese tourism situation before 1st Bali bombing. The
REFFERENCES
Anderson, R.I., Fisher, M., Xia Y., and Michello, F. (1999),“Measuring efficiency in the hotel industry: a stochastic frontier approach”, Hospitality Management, Vol.18, No.1, pp.45–57.
Avkiran, N. K. (2002), “Monitoring Hotel Performance”, Journal of Asia-Pacific Business, Vol. 4, No. 1, pp. 51-56.
Badan Pusat Statistik (BPS) (various), Statistik Pariwisata Bali, Bali, Indonesia. Badan Pusat Statistik (BPS) (various), Bali Dalam Angka, Bali, Indonesia. Bali Tourism Board (2007), StatistikPariwisata Bali, 2007, Bali, Indonesia.
Banker, R.D., Charnes, A., and Cooper, W.W., (1984), “Some models for estimating technical and scale inefficiencies in data envelopment analysis”, Management Science, Vol. 30, No.9, pp. 1078– 1092.
Barros, C. P. (2005),“Evaluating the Efficiency of a Small Hotel Chain with a Malmquist Productivity Index”, International Journal Of Tourism Research, Vol.7, No.3, pp. 173–184.
Caves, D.W., Christensen, L.R. and Diewert, W.E. (1982)."The Economic Theory of Index Numbers and the Measurement of Input, Output, and Productivity", Econometrica, Vol. 50, No. 6, pp. 1393-1414.
Cooper, W.W., Seiford, L. M. and Tone, K. (2007), Data Envelopment Analysis: A Comprehensive Text with Models, Applications, References and DEA-Solver Software, Springer, New York, USA. Coelli, T.J. (1996), "A Guide to DEAP 2.1: A Data Envelopment Analysis (Computer)
Program."University of New England, Armidale, NSW.
Coelli, T.J., Rao,D.S.P, O’Donnell,C.J. and Battese, G.E. (2005), An Introduction to Efficiency and Productivity Analysis, Springer, New York, USA.
Fare, R.,Grosskopf, S., Lindgren, B. and Roos, P., (1992), "Productivity Change in Swedish Pharmacies 1980-1990: A Non Parametric Malmquist Approach, "Journal of Productivity Analysis, Vol. 3, No. 1, pp.85-102.
Fare, R. and Grosskopf, S. (1994),"Productivity Growth, Technical Progress, and Efficiency Change in Industrialized Countries", The American Economic Review, Vol. 84, No. 1, pp.66-83.
Fare, R. and Grosskopf, S. (1996), Intertemporal Production Frontiers: with Dynamic DEA,Kluwer Academic Publishers, Boston, USA.
Foo L.Y. and Othman, M. (2011), “Data Envelopment Analysis to Measure Efficiency of Hotels in Malaysia”, SEGi Review ISSN 1985-5672 Vol. 4, No. 1, July 2011, 25-36
Fried, H.O., Schmidt, S.S. and Yaisawarng, S. (1999), "Incorporating the Operating Environment into a Nonparametric Measure of Technical Efficiency", Journal of Productivity Analysis, Vol.12, No. 3, pp. 249-267.
Giannoni S., and Maupertuis M.A. (2007), “Is Tourism Specialization Sustainable for a Small Island
Economy? A Cyclical Perspective”, Advances in Modern Tourism Research, pp. 87-105
Huang Y., Mesak, H.I., Hsu, M.K., and Qu, H. (2012), “Dynamic efficiency assessment of the Chinese hotel industry”, Journal of Business Research, Vol. 6, pp.59–67.
Hwang, S.N., and Chang, T.Y. (2003), “Using data envelopment analysis to measure hotel managerial efficiency change in Taiwan”, Tourism Management, Vol. 69, pp.324-357.
Indonesia: Holiday blues, Asiamoney, 10 April, 2006.
Min, H. and Joo, S.(2008), “A data envelopment analysis-based balanced scorecard for measuring the comparative efficiency of Korean luxury hotels”, International Journal of Quality & Reliability Management, Vol. 25, No. 4, pp. 349-365.
Morey, R. and Dittman, D. (1995),“Evaluating a hotel GM's performance”, Cornell Hotel & Restaurant Administration Quarterly, Vol. 36, No.5, pp.30-35.
Neves, J. C., and Lourenco, S. (2009), “Using Data Envelopment Analysis to Select Strategies That Improve the Performance of Hotel Companies”, International Journal of Contemporary Hospitality Management, Vol. 21, No.6, pp.698-712.
Pedraja-Chappara, F., Salinas-Jimenez, J., Smith, P. (1999), “On the Quality of the Data Envelopment Analysis Model”, The Journal of Operational Research Society, Vol. 50, No.6, pp. 636-644. Sherman, H. D. and Zhu, J. (2006), Service Productivity Management: Improving service performance
using Data Envelopment Analysis (DEA), Springer, NY, USA.
Shephard, R.W. (1953), Cost and production functions. Princeton University Press, Princeton.
Smyth R., Nielsen, I. and Mishra, V. (2009), “‘I’ve been to Bali too’ (and I will be going back): are terrorist shocks to Bali's tourist arrivals permanent or transitory?”, Applied Economics, 41, 1367–1378.
The Knight Frank Global Network (2011), http://knightfrank.co.id/en.
Tsaur, S.H. (2000), “The operating efficiency of international tourist hotels in Taiwan”, Asia Pacific Journal of Tourism research, Vol.6, No.1, pp. 29-37.
Verbeek, M (2004), A guide to modern econometrics, 2nd ed. John Wiley & Sons Ltd, Chichester, England.
APPENDIX
Table A. Inputs and environmental variables correlations 2004
OTHI NSTDR NSUIR
INTCH 0.431 0.271 0.214
STAR 0.521 0.560 0.370
OTHI=Other input ; NSTDR=number of standard rooms; NSUIR=number of suite rooms INTCH=international chain ; STAR=Number of star of hotel
Table B. Inputs and environmental variables correlations 2010
OTHI NSTDR NSUIR
INTCH 0.282 0.185 0.421
STAR 0.528 0.258 0.397
OTHI=Other input ; NSTDR=number of standard rooms; NSUIR=number of suite rooms INTCH=international chain ; STAR= Number of star of hotel
Table C. DEA results 2010
TE1 PTE1 SE1 # IRS1 # DRS1 # MPSS1 SE=1 &
PTE<1 SE<1 & PTE=1
1-star Mean 0.718 0.762 0.942 6 0 2 1 1 #obs : 12 Std.Dev 0.173 0.181 0.050 Min. 0.520 0.571 0.860 Max. 1.000 1.000 1.000 2-star Mean 0.640 0.706 0.925 20 2 0 0 4 #obs : 22 Std.Dev 0.159 0.190 0.135 Min. 0.418 0.423 0.525 Max. 0.962 1.000 0.995 3-star Mean 0.666 0.730 0.923 27 4 4 1 4 #obs : 32 Std.Dev 0.164 0.179 0.127 Min. 0.382 0.486 0.484 Max. 1.000 1.000 1.000 4-star Mean 0.778 0.801 0.973 25 14 8 1 4 #obs : 30 Std.Dev 0.150 0.154 0.059 Min. 0.453 0.459 0.639 Max. 1.000 1.000 1.000 5-star Mean 0.872 0.888 0.983 8 17 12 2 1 #obs : 34 Std.Dev 0.125 0.125 0.025 Min. 0.524 0.533 0.888 Max. 1.000 1.000 1.000 All-star Mean 0.751 0.790 0.955 86 37 26 5 14 #obs : 130 Std.Dev 0.172 0.172 0.091 Min. 0.382 0.423 0.484 Max. 1.000 1.000 1.000
Table D. Correlation between dependent and independent variables 2004
TE INTCH STAR
TE 1.000
INTCH 0.378 1.000
STAR 0.396 0.410 1.000
TE=initial technical change ; INTCH=international chain ; STAR= Number of star of hotel
Table E. DEA results 2004
TE1 PTE1 SE1 # IRS1 # DRS1 # MPSS1 SE=1 &
PTE<1 SE<1 & PTE=1
1-star Mean 0.525 0.662 0.805 11 1 0 0 1 #obs : 12 Std.Dev 0.183 0.200 0.190 Min. 0.175 0.326 0.365 Max. 0.916 1.000 0.990 2-star Mean 0.561 0.665 0.874 20 2 0 0 4 #obs : 22 Std.Dev 0.205 0.252 0.166 Min. 0.177 0.188 0.325 Max. 0.914 1.000 0.991 3-star Mean 0.584 0.654 0.919 19 8 5 0 4 #obs : 32 Std.Dev 0.212 0.248 0.146 Min. 0.202 0.206 0.392 Max. 1.000 1.000 1.000 4-star Mean 0.628 0.680 0.930 13 13 4 1 1 #obs : 30 Std.Dev 0.200 0.212 0.083 Min. 0.270 0.273 0.689 Max. 1.000 1.000 1.000 5-star Mean 0.787 0.832 0.946 5 24 5 0 5 #obs : 34 Std.Dev 0.179 0.183 0.062 Min. 0.338 0.343 0.784 Max. 1.000 1.000 1.000 All-star Mean 0.638 0.709 0.910 68 48 14 1 15 #obs : 130 Std.Dev 0.216 0.229 0.130 Min. 0.175 0.188 0.325 Max. 1.000 1.000 1.000
1TE=Technical efficiency ; PTE=Pure technical efficiency ; SE=Scale efficiency ; IRS=Increasing return to scale ; DRS=Decreasing return to scale ; MPSS=Most productive scale size
Table F. Correlation between dependent and independent variables 2004
TE INTCH STAR
TE 1.000
INTCH 0.496 1.000
STAR 0.443 0.473 1.000
Table G. Malmquist productivity index 2004-2010 Effch1 Techch1 Tfpch1 1-star Mean 1.243 1.186 1.473 #obs : 4 Std.Dev 0.618 0.159 0.515 Min. 0.931 1.013 1.150 Max. 2.250 1.388 2.280 >1 3 (75%) 4 (100%) 4 (100%) =1 0 (0%) 0 (0%) 0 (0%) <1 1 (25%) 0 (0%) 0 (0%) 2-star Mean 1.203 1.195 1.438 #obs : 12 Std.Dev 0.474 0.149 0.545 Min. 0.728 0.941 1.052 Max. 2.326 1.444 2.889 >1 8 (67%) 11 (92%) 12 (100%) =1 0 (0%) 0 (0%) 0 (0%) <1 4 (33%) 1 (8%) 0 (0%) 3-star Mean 1.088 1.307 1.423 #obs : 18 Std.Dev 0.239 0.182 0.382 Min. 0.726 0.798 0.798 Max. 1.615 1.618 2.248 >1 10 (56%) 17 (94%) 15 (83%) =1 1 (6%) 0 (0%) 0 (0%) <1 7 (39%) 1 (6%) 3 (17%) 4-star Mean 1.229 1.275 1.568 #obs : 26 Std.Dev 0.409 0.240 0.472 Min. 0.900 0.994 0.919 Max. 2.676 2.208 3.042 >1 20 (77%) 25 (96%) 25 (96%) =1 2 (8%) 0 (0%) 0 (0%) <1 4 (15%) 1 (4%) 1 (4%) 5-star Mean 1.108 1.297 1.438 #obs : 32 Std.Dev 0.258 0.184 0.390 Min. 0.806 0.918 0.958 Max. 2.077 1.593 2.600 >1 21 (66%) 31 (97%) 31 (97%) =1 3 (9%) 0 (0%) 0 (0%) <1 8 (25%) 1 (3%) 1 (3%) All-star Mean 1.155 1.274 1.472 #obs : 92 Std.Dev 0.354 0.197 0.435 Min. 0.726 0.798 0.798 Max. 2.676 2.208 3.042 >1 62 (67%) 88 (96%) 87 (95%) =1 6 (7%) 0 (0%) 0 (0%) <1 24 (26%) 4 (4%) 5 (5%)
1Effch=Efficiency change ; Techch=Technological change ; Tfpch=Total productivity change
Table H. Correlation between dependent and independent variables
Effch ITE INTCH STAR
Effch 1.000
ITE -0.601 1.000
INTCH 0.006 0.378 1.000
STAR -0.105 0.414 0.496 1.000
