Internatonal Conference On Appled Economcs ICOAE 2011 315 MEASURING OPERATION EFFICIENCY OF THAI HOTELS INDUSTRY: EVIDENCE FROM METAFRONTIER ANALYSIS PHANIN KHRUEATHAI 1, AKARAPONG UNTONG 2, MINGSARN KAOSAARD 3,RENATO ANDRIN VILLANO 4 Abstract Ths paper utlzes a unque hotellevel dataset to examne operatonal effcency and technology gap n Thaland s hotels. Ths paper classfes the hotels n Thaland nto fve groups wth dstnctve levels of operatonal technologes. A metafronter analyss s appled to evaluate the operatonal effcency scores of the hotels n same groups and between groups. The result show that, the hotels n the fve groups dffer n the use they make of nput operatonal effcency. Meanwhle, the mean effcency of the hotels wth room rate between 300900 baht per nght and total revenue lower than 1 mllon baht per year s partcularly low. Ths study suggests to transferrng knowledge about operatonal management from the hotels wth hgher operatonal effcency to the hotels that had low operatonal effcency. Ths mght help to mprove operatonal effcency and compettveness n long run. JEL codes: D240 Producton; Cost; Captal, Total Factor, and Multfactor Productvty; Capacty. Key Words: Tha s hotels ndustry, operatonal effcency, stochastc metafronter 1. Introducton The operatonal effcency of the hotel ndustry n Thaland has been extensvely analyzed usng advanced effcency methods such as DEA (Data Envelopment Analyss) and SFA (Stochastc Fronter Analyss) (Akarapong, 2004; Mngsarn and Akarapong, 2005; Akarapong and Mngsarn, 2009). However, these methods assume homogenous technology and the same envronmental characterstcs, makng the results not strctly comparable across dfferent groups of hotels (Assaf, Barros and Josassen, 2009). To assess more accurately the mpact of dfferent technologes and envronmental characterstcs, ths study apples the concept of metafronter analyss developed by Rao, O Donnell and Battese (2003) and O Donnell, Rao and Battese (2007) to estmate the envelope of possble fronters that mght arse from the heterogenety between groups of hotels. Moreover, most of prevous studes of hotel effcency focused on the estmaton of manageral or operatonal effcences by usng a lmted data set and restrctve functonal form. They also assumed that technologes are smlar across hotels and ndusal envronment. But n fact, the dfferent groups of hotel use a dfferenct manageral or operaton technology. Such as the foregn nvestment hotels had to use the standard manageral technology from the hotels chan whle the local hotels ddn t have these and manage the hotel on ther own. In order to examne the patterns and dfferences n performance n these dfferent categores of hotels, the purpose of ths paper s to estmate the operatonal effcences of the Tha s hotel ndustry usng CobbDouglas functonal form, a larger data set and a methodology that would be smlar to the hotel envronment and technology across dfferent groups of hotels. The man objectve of ths study s to use metafronter analyss to assess the operatng effcency of fve dfferent hotel types n Thaland. There are 1) foregn nvestment 2) room rate more than 900 baht per nght (or more than 30 US$ per nght) 3) room rate less than 300 baht per nght (or less than10 US$ per nght) 4) room rate between 300900 baht per nght (or between 1030 US$ per nght) and 5) total revenue less than 1 mllon baht per year (less than 300 thousand US$) and room rate between 300900 baht per nght and total revenue more than 1 mllon baht per year. The study focuses on the potental of dfferent types of ownershp to rase operatng effcency through foregn nvestment. In addton, the queston of whether hgher room rates prce are more productve than lower rates s analyzed. Greater productvty gans are expected at hgher levels of cooperaton at large hotels because they should open up a broader range of opportuntes to mprove operatonal effcency. The paper s organzed as follow. Secton 2 contans method of analyss, and s followed by the results and dscusson n secton 3. In secton 4, concludng comments are presented. 2. Method of analyss 2.1 Analytcal Framework Operatonal effcency s an mportant factor n manageral busness. The estmaton of techncal effcency represents to the ablty of compettveness (Hwang and Chang, 2003). Relatve effcency (Farrell, 1957) has been extended and modfed to Data Envelopment Analyss (DEA) and Stochastc Fronter Analyss (SFA). Both approaches are popular n the effcency lterature, however; DEA has some restrctons such as nablty to take nto account error term n the output and stochastc element of producton, no assumpton about dstrbuton effcency, No sgnfcant test of the techncal effcency (Barros, 2006; Barros and Deke, 2008). On the other hand, the advantage of the stochastc fronter approach s that t allows for random dsturbances, such as the effect of qualty of nputs, and measurement errors n the output varables (Barros, 2006; Barros and Deke, 2008). Accordng to these advantages, ths study used the stochastc fronter (SFA) approach wth emphass on the parametrc model, and then calculated the effcency scores for ndvdual hotel unts. 1 Faculty of Management Scence, Uttaradt Rajabhat Unversty, Thaland 2 Student of Toursm and Envronmental Economcs, The Unversty of the Balearc Islands, Span 3 Publc Polcy Studes Insttute Chang Ma Unversty, Thaland 4 School of Busness, Economcs and Publc Polcy, Unversty of New England, Armdale, NSW Australa 315
Out put Y 316 Internatonal Conference On Appled Economcs ICOAE 2011 2.1.1 Stochastc Fronter Analyss (SFA) The stochastc fronter framework n ths study s a parametrc specfcaton of econometrc models to estmate the producton fronter and measure effcency scores. The basc stochastc fronter producton functon s defned as: Y = (X, ) exp (ε ) (1) where Y s the output of th ( = 1, 2,..., N) frm; X s the correspondng matrx of nputs; s the vector of parameters to be estmated; and ε s the error term that conssts of two ndependent elements, V and U, such that ε V U. The V s are assumed to be symmetrc, dentcally and ndependently dstrbuted errors that represent random varatons n output, as a result of factors outsde the control of the decsonmakng unt, as well as the effects of measurement error n the output varable, varables excluded from the model and statstcal nose. They are assumed to be normally dstrbuted wth mean zero and varance ζ 2 [V N(0, )] v V. The U s are nonnegatve random varables that represent the stochastc shortfall of outputs from the most effcent producton. U s defned by truncaton of the normal dstrbuton wth mean U = δ 0 + J j=1 δ j Z j and varance ζ 2 U, where Z j s the value of the jth explanatory varable assocated wth the techncal neffcency effect of frm ; and δ 0 and δ j are unknown parameters to be estmated (Battese and Coell, 1995). The maxmum lkelhood method s used to estmate the parameters of both the stochastc fronter model and the neffcency effects model. The varance parameter of the lkelhood functon s estmated n terms of ζ 2 ζ 2 V + ζ 2 U and γ ζ 2 U ζ 2. The techncal effcency of a frm can be defned by the rato of the observed output to the correspondng stochastc fronter output by Y TE expu. X ; expv (2) 2.1.2 Metafronter Approach The metafronter producton s a producton functon that covers ndvdual fronter of groups. A graph of the metafronter functon s presented n fgure 1. Several studes are used to estmate techncal effcency n dfferent regons, envronmental, and technologes of agrcultural producton. To begn wth the stochastc metafronter framework was done by Battese and Rao (2002), Battese, Rao and O Donnell (2004), and O'Donnell, Rao and Battese (2008). Then, Vllano, Flemng and Flemng (2008) proposed that other studes, such as latent class model (Greene, 2004), and statecontngent fronter (O'Donnell and Grffths, 2006) stll have based estmators of the parameters of the fronter and techncal neffcency because the results reveal that lack of success n accountng for envronmental varables. Therefore, metafronter analyss was used to estmate the technology gap rato and estmate parameters of fronter and techncal neffcences. Fgure 1 Metafronter and Indvdual Fronters 2 Metafronter Indvdual fronters 0 Source: (Battese et al., 2004) Input X From fgure 1, the estmaton of the standard stochastc fronter model for R dfferent groups wthn the ndustry defned as: Y ( j) (X ( j), ( j) ) e v ( j) u ( j) (3) = 1,2,, N j, t = 1,2,, T, j = 1,2,, R, Suppose that, for the j th group, there are sample date on N j frms that produce one product from the varous nputs. Where Y (j) s the output for the th frm for the j th group. X (j) s β (j) s v (j) s a vector of values of functons of the nput used by the th frm for the j th group. the parameter vector assocated wth the xvarables for the stochastc fronter for the j th group nvolved. 2 N(0, ) statstcal nose assumed to be ndependently and dentcally dstrbuted as v( j) random varables. 316
Internatonal Conference On Appled Economcs ICOAE 2011 317 u (j) s nonnegatve random varables assumed to account for techncal neffcency n producton and assumed to be 2 N(, ) ndependently dstrbuted as truncatons at zero of the ( j) ( j) dstrbuton, where µ (j) s some approprate neffcency model, defned by Battese and Coell (1992) and (1995). In smplfed verson, the model s presented as: v u ) ( j) u ( j) X ( j) v X, e e ( j) ( j Y ( j) (4) Assumed that exponent of fronter producton functon s lnear n the parameter vector, β (j), so that X s a vector of functon of the nput for the th frm. The metafronter producton functon model s expressed by X X, e, Y = 1,2,, N. (5) Where β s the vector of parameters for the metafronter functon such that: X X( j), j = 1,2,, J. (6) Equaton 6, the metafronter producton functon s solved the optmzaton problem by Battese, Rao and O Donnell (2004). The optmzaton problem s defned as: N Mn 1 ln X, ln X, ( j) ln X, ln X, ( j) s.t. (7) where β (j) s the estmated coeffcent vector assocated wth the groupj stochastc fronter The observed output defned by the stochastc fronter for the j th group n equaton 4 and t s alternatvely expressed n term of the metafronter functon n equaton 5, such that: X, v ) X, e ( j U ( j) Y ( j) e X, (8) where the frst term on the rghthand sde of equaton 10.6 s the same as techncal effcency relatve to stochastc fronter for the j th group (Battese, Rao and Prasado, 2002). Y ( j) U TE ( j) ( j) e v( j) X ( j), ( j) e (9) The second term on the rghthand sde of equaton 9 s the technology gap rato (TGR) (Battese, Rao and Prasado, 2002) or the metatechnology rato (MTRs) (O Donnell et al, 2007) or envronmenttechnology gap rato (ETGR) (Vllano, Flemng and Flemng, 2008), whch s expressed as: X, TGR ETGR ( j) X, (10) The TGR or ETGR measure the rato of the output for the fronter producton functon for j th group relatve to the potental output that s defned by the metafronter functon, gven the observed nput (Battese, Rao and Prasado, 2002) and (Battese, Rao and O'Donnell, 2004). The TGR or MTR or ETGR has values between zero and one. The techncal effcency of th frm, relatve to the metafronter, s denoted by TE, s defned n a smlar way to equaton 9, TE can be expressed as: TE Y v ) X, e ( j (11) From equaton 11, t s the rato of the observed output relatve to the last term on the rghthand sde of equaton 6, whch s the metafronter output, adjusts for the correspondng random error. Equaton 8, 9, 10 and 11 mply that an alternatve expresson for the techncal effcency relatve to the metafronter can be expressed by Y TE X, e v( j) e U( j) X, X, ( j) TE TE TGR (12) O'Donnell, Rao and Battese (2008) presented the extensons to the basc metafronter framework, such as multpleoutput; technologcal change (Coell et al., 2005); tmenvarant neffcency effects can be found n (O'Donnell, Rao and Battese, 2008); alternatve orentatons and dentfyng groups (Orea and Kumbhakar, 2004) and (O'Donnell and Grffths, 2006). 317
318 Internatonal Conference On Appled Economcs ICOAE 2011 2.2 Analytcal Framework 2.2.1 The Emprcal Model The stochastc fronter analyss model defned by equaton 1 and 2. They were estmated assumng the CobbDouglas functonal form. The nputs are defned as the number of rooms, room rate per nght, number of employees, operatonal expenses and assets. The output s total revenue. The specfcaton of the functonal form s defned by ln (Y ) (k) = β 0(k) + β 1(k) ln(x 1(k) ) + β 2(k) ln(x 2(k) ) + β 3(k) ln(x 3(k) ) +β 4(k) ln(x 4(k) ) + β 5(k) ln(x 5(k) ) + V (k) + U (k) (13) Where Y s total revenue (n baht); X 1 s the number of rooms (n room); X 2 s room rate per nght (n baht); X 3 s the number of employees (n person); X 4 s operatonal expenses (n baht); X 5 s assets (n baht); β 0 β 5 are unknown parameters to be estmated; k s 5 groups of the hotel groups. 2 The V (k) are assumed to be ndependently and dentcally dstrbuted wth mean zero and varance, ζ V(k) ; and the us are techncal effcency effects that are assumed to be halfnormal and ndependently dstrbuted such that U (k) s defned by the truncaton at zero of the normal dstrbuton wth known varance, ζ 2 U(k). The nputs are mpled nputs n that they are measured as costs, assumng all groups faced the same nput prces and no changes occurred n nput prces durng the perod when the survey was undertaken. Smlarly, outputs are mpled outputs n that they are measured as revenue assumng all groups faced the same output prces. The techncal neffcency model s defned followng Battese and Coell (1995) as: U (k) =δ 0(k) + δ 1(k) Z 1(k) + δ 2(k) Z 2(k) + δ 3 Z 3(k) (14) Where Z 1 s rato of workers per room; Z 2 s perod of operaton; Z 3 s rato of foregn guest; δ 0 δ 3 are unknown parameters to be estmated. Many varables were tested for ncluson n the neffcency model. They are dscussed n ths secton and reasons are gven for the expected drecton of ther relatons wth the level of operatonal effcency of hotel ndustry n Thaland. The coeffcent of the rato of workers per room s expected to be postve because lower number of workers should have lower cost of labour. The other neffcency varables, the sgns on the coeffcents of perod of operaton are expected to be negatve because longer perod of operaton should have accumulated more revenues. Fnally, the coeffcent of rato of foregn guest s expected to have a negatve sgn because a hgher number of foregn guests would help the hotels to manage more effectvely. If frms can control the qualty of servce, they can better control servce prces. 2.2.2 Varables The study uses 1,799 samples of hotels and guesthouses from the 2008 Survey Database of the Natonal Statstcal Offce, Thaland. The statstcs for nput and output varables n the operatng effcency of hotel are reported n Table 1. We dvded the hotels nto fve groups by consderng the mpact of dfferent technologes: (foregn nvestment, room rate more than 900 baht per nght, room rate less than 300 baht per nght, room rate between 300900 baht per nght and total revenue less than 1 mllon baht per year and room rate between 300900 baht per nght and total revenue more than 1 mllon baht per year). Table 1 Summary Statstcs for Data on the hotels of Thaland Varables Unts Mn Max Mean SD Total Total revenues Mllon baht 0.0098 2,148.69 20.49 97.11 Total rooms room 2 760 62 84 Room rate baht/nght 60 54,893 707 1,816 Employees person 1 859 38 89 Operatonal expenses Mllon baht 0.0044 1,444.70 10.86 62.39 Assets Mllon baht 0.0010 5,493.44 54.14 255.85 1. Foregn nvestment Total revenues Mllon baht 0.22 2,148.69 299.76 422.73 Total rooms room 7 734 239 197 Room rate baht/nght 129 19,086 3,470 3,696 Employees person 4 859 246 251 Operatonal expenses Mllon baht 0.06 1,444.70 173.34 281.58 Assets Mllon baht 0.002 5,493.44 629.93 1,245.27 Table 1 Summary Statstcs for Data on the hotels of Thaland Varables Unts Mn Max Mean SD 2. Room rate more than 900 baht per nght 318
Internatonal Conference On Appled Economcs ICOAE 2011 319 Total revenues Mllon baht 0.10 1,161.35 72,41 126.12 Total rooms room 2 760 145 136 Room rate baht/nght 905 54,893 2,483 4,166 Employees person 2 713 135 145 Operatonal expenses Mllon baht 0.043 956 37.47 85.85 Assets Mllon baht 0.002 2,127.54 172.05 299.71 3. Room rate less than 300 baht per nght Total revenues Mllon baht 0.010 19.32 0.98 1.43 Total rooms room 4 316 29 26 Room rate baht/nght 60 299 206 56 Employees person 1 101 7 8 Operatonal expenses Mllon baht 0.040 8.75 0.37 0.64 Assets Mllon baht 0.001 219.24 9.74 17.40 4. Room rate between 300900 baht per nght and total revenue less than 1 mllon baht per year Total revenues Mllon baht 0.035 0.99 0.52 0.25 Total rooms room 2 72 18 11 Room rate baht/nght 300 889 415 131 Employees person 1 16 5 3 Operatonal expenses Mllon baht 0.0067 1.01 0.22 0.16 Assets Mllon baht 0.0020 68.15 8.42 9.68 5. Room rate between 300900 baht per nght and total revenue more than 1 mllon baht per year Total revenues Mllon baht 1.00 148.43 8.55 14.59 Total rooms room 3 456 73 57 Room rate baht/nght 300 900 493 158 Employees person 2 431 34 45 Operatonal expenses Mllon baht 0.047 56.32 3.91 7.14 Assets Mllon baht 0.001 915.38 31.99 69.38 Source: the Natonal Statstcal Offce 2009. 2.3 The emprcal fndng The stochastc fronter analyssgroup and stochastc fronter analysspool estmates were obtaned usng FRONTIER 4.1 (Coell, 1996) n order to formulate the techncal effcency (TE) effects model (Battese and Coell, 1995). The stochastc fronter analyss /metafronter estmates were obtaned usng SHAZAM. 2.3.1 Hypothess Testng A lkelhoodrato (LR) test, for the group s stochastc fronter model s the same for all the operatonal effcency of the hotel ndustry n Thaland. For testng of the null hypothess, we can decde that t would be a good reason or not for estmatng the effcency level of frms to a metafronter operatonal functon. Followng Battese, Rao and O Donnell (2004), we test the null hypothess by calculatng LR statstc. The LR statstc s defned by: lnlh / LH 2lnLH lnl 2 (15) where ln [L(H 0 )] s the value of the log lkelhood functon for the stochastc fronter estmated by poolng the data for all groups. ln [L(H 1 )] s the sum of the value of the log lkelhood functon for the 5 groups operatonal functon. 0 1 0 H1 2.3.2 The Estmaton of the metafronter functon The operatonal effcency s computed usng three approaches. Frst, a standard operaton stochastc fronter (lke producton) was employed usng pooled crosssecton data. Second, group stochastc fronter functons were estmated. Fnally, metafronter analyss was used gven dfferences n operaton envronments and technologes between the fve groups of hotels studed. The gamma parameters are sgnfcant for the fve groups, suggestng the presence of operatonal neffcency, and the LR test = 134.34, wth a pvalue of 0.00 (usng a Chsquare dstrbuton wth 52 degrees of freedom). Therefore, the null hypothess that dfferent groups have the same stochastc fronter models can be rejected. All nputs are assocated wth total revenues and the hgh rato of foregner guests mproves n operaton effcency (Table 2). The estmates of the parameters of the neffcency effects model are presented n Table 2. Estmates of the coeffcents of the varables explanng dfferences n group effcency provde nterestng results. Frst, the coeffcent of the varable denotng the rato of foregn guest s sgnfcant at the 1 and 5 per cent level and has both negatve and postve coeffcents for all groups of hotels. Ths result ndcates that a hgher number of foregn guests s ssocated wth greater operatonal effcency n large hotels (group1 and 2). It was ntally surprsng to fnd that the number of years of operaton has a postve assocaton wth operatonal neffcency n small hotels (group 3 and 4). On the other hand, the longeroperated hotels tend to be more effcent n only large hotels (group 1). 319
320 Internatonal Conference On Appled Economcs ICOAE 2011 Fnally, the rato of workers per room has postve assocaton wth operatonal neffcency. Ths result suggests that the hgher the number of workers, the lower the level of effcency n only large hotels (group 1). Estmated operatonal effcences wth respect to the group fronters and the metafronter, together wth estmated MTRs, are presented n Table 3. Hotels dffer n operatonal effcency, MTRs, and the use they make of nputs. The value of MTRs ranges from 0.56 to 0.86, whch explans that on average, hotels n Thaland operate between 56 86 percent of the potental total revenue gven the technology avalable to the ndustry as a whole. As expected, estmated operatonal effcences are lower and dspersed n the metafronter model. The average MTR were found to be sgnfcantly dfferent for fve groups 5. However, the metafronter analyss provdes a more consstent and homogenous effcency comparson. Mean MTRs vary consderably between hotels and across groups whereas mean operatonal effcency wth respect to the pooled fronter are reasonably smlar across groups but dffer n the operatonal effcency wth respect to group fronters. Hotels wth the lowest total revenue and room rate per nght have the lowest (Group 4) MTR (0.56) due to a lack of operatng technology, few foregners, and ther small sze that precludes laboursavng technologes. In terms of the relatonshp between effcency and hotel classfcaton, the effcency of foregn nvestment hotels s hgher than domestc nvestment hotels (0.83) and they can earn revenue from the other sources of ncome, such as entertanment actvtes, food and beverage. Meanwhle, the MTRs of groups 1, 2, 3 and 5 are lower than group 4, and group 4 has the lowest MTRs. Group 4 has the lowest average MTR rato hence ts average effcency s reduced from 37.21 percent when compared relatve to the fronter wthn group to 10.66 percent when compared to the metafronter. Table 2 Estmates for parameters of the stochastc fronter model. Varables Group 1 Group 2 Group 3 Group 4 Group 5 Pooled fronter Metafronter Fronter model Constant 5.196 (0.980) 4.956 (0.480) 5.327 (0.382) 8.588 (0.683) 6.679 (0.993) 4.994 (0.118) 5.421 Total rooms (rooms) 0.220 (0.217) 0.192 (0.732) 0.272 (0.045) 0.076 (0.052) 0.034 (0.213) 0.149 (0.027) 0.074 Room rate (baht per nght) 0.169 (0.128) 0.193 (0.072) 0.117 (0.066) 0.140 (0.089) 0.170 (0.366) 0.163 (0.033) 0.124 Employees (persons) 0.218 (0.227) 0.285 (0.072) 0.308 (0.043) 0.410 (0.059) 0.406 (0.294) 0.403 (0.014) 0.429 Operatonal (baht) expenses 0.561 (0.103) 0.530 (0.029) 0.504 (0.025) 0.362 (0.033) 0.424 (0.218) 0.517 (0.045) 0.548 Assets (baht) 0.003 (0.023) 0.015 (0.009) 0.004 (0.008) 0.022 (0.017) 0.015 (0.028) 0.008 (0.006) 0.017 Ineffcency model effect Constant 0.968 (0.766) 0.577 (0.149) 9.992 (3.558) 11.035 (7.281) 0.090 (0.039) 0.101 (0.036) Rato of workers per room (%) 1.115 (0.701) 0.121 (0.116) 1.864 (1.198) 1.254 (1.744) 0.016 (0.113) 0.034 (0.032) Perod of operaton (day) 0.127 (0.097) 0.004 (0.004) 0.076 (0.030) 0.102 (0.069) 0.002 (0.006) 0.0008 (0.0015) 5 We test the samplng dstrbuton of the dfference means by usng a t test. The value of the test statstc s 3.56, whch falls n the rejecton regon, thus, we reject H 0. 320
Internatonal Conference On Appled Economcs ICOAE 2011 321 Rato of foregn guest (%) 0.033 (0.022) 0.083 (0.001) 0.014 (0.008) 0.045 (0.026) 0.002 (0.001) 0.0018 (0.0004) Varance parameter Sgmasquared Gamma 0.895 (0.593) 0.802 (0.179) 0.190 (0.017) 0.302 (0.083) 2.080 (0.612) 0.902 (0.029) 2.459 (1.545) 0.957 (0.030) 0.200 (0.021) 0.00004 (0.0000 1) LogL 34.47 116.19 494.34 131.60 363.38 0.243 (0.008) 0.00000 7 (0.0000 03) 1274.32 Note : denote sgnfcance at the 1% level. denote sgnfcance at the 5% level. denote sgnfcance at the 10% level. : The numbers n parentheses are standard errors. Source: Author's calculaton. Table 3 Estmates of Techncal effcency (TEs) and Technology Gap Ratos (MTRs). Groups Mn Max Mean SD Total Pool fronter 0.6464 0.9999 0.9074 0.0473 Group fronter 0.1742 0.9999 0.8376 0.0995 Technology gap rato (MTR) 0.3526 1.0000 0.6417 0.1066 Metafronter 0.1109 0.9966 0.5354 0.1016 1. Foregn nvestment (group 1) Pool fronter 0.8295 0.9999 0.9722 0.0463 Group fronter 0.2372 0.9300 0.7822 0.1408 Technology gap rato (MTR) 0.6353 1.0000 0.8371 0.0969 Metafronter 0.1660 0.9109 0.6543 0.1379 2. Room rate more than 900 baht per nght (group 2) Pool fronter 0.6464 0.9999 0.9381 0.0585 Group fronter 0.4116 0.9719 0.7634 0.1304 Technology gap rato (MTR) 0.5041 1.0000 0.7149 0.0908 Metafronter 0.3554 0.8537 0.5415 0.0966 3. Room rate less than 300 baht per nght (group 3) Pool fronter 0.8312 0.9999 0.8952 0.0392 Group fronter 0.1742 0.9406 0.8208 0.0743 Technology gap rato (MTR) 0.4365 0.8621 0.6543 0.0605 Metafronter 0.1109 0.7496 0.5367 0.0671 4. Room rate between 300900 baht per nght and total revenue less than 1 mllon baht per year (group 4) Pool fronter 0.8490 0.9999 0.9027 0.0370 Group fronter 0.2109 0.9315 0.7988 0.1106 Technology gap rato (MTR) 0.3721 0.9600 0.5620 0.0979 Metafronter 0.1260 0.7977 0.4475 0.0948 5. Room rate between 300900 baht per nght and total revenue more than 1 mllon baht per year (group 5) Pool fronter 0.7945 0.9999 0.9062 0.0457 Group fronter 0.7622 0.9999 0.9061 0.0489 Technology gap rato (MTR) 0.3526 1.0000 0.6173 0.1169 Metafronter 0.3126 0.9966 0.5592 0.1107 Source: Author's calculaton. 2.4 Concluson Ths paper has provded some nterestng results on the operatonal effcency of the hotel ndustry n Thaland. The metafronter analyss s used to develop the tradtonal fronter analyss because ths model enables the calculaton of comparable operatonal effcency for frms operatng under dfferent technologes or locatons. The meatfronter analyss dvdes the operatonal effcency nto two parts: 1) operatonal effcency respect to the subgroup; and 2) operatonal effcency respect to the metafronter by consderng the technology gap rato. Paper shows how group fronter and the 321
322 Internatonal Conference On Appled Economcs ICOAE 2011 metafronter can be estmated usng a CobbDouglas functonal form. An emprcal example used crosssectonal data of statstcs for nput and output varables n the operatng effcency of 1,799 hotels. We dvde the hotel nto fve groups. The fndng of the study s that, hotels n the fve groups dffer n the use they make of nput operatonal effcency and technology gap rato (MTRs). Mean MTRs vary substantally between hotels and across groups whereas mean operatonal effcency are reasonably smlar across groups but dffer n the extent of varaton among hotels wthn each group. The mean value of operatonal effcency for the pooled fronter, group fronter and metafronter models across all groups are 0. 90, 0.83 and 0.53 respectvely. Group fronters show that the mean value of MTR vares from 0.56 n hotels wth room rate between 300900 baht per nght and total revenue less than 1 mllon baht per year to 0.83 n hotels wth foregn nvestment. The low MTR s attrbutable to a lack of operaton management. The results suggest that transferrng knowledge and knowledge management about operaton management from hgher operatonal effcency of hotels to lower operatonal effcency of hotels needs to be organzed. For example, qualty standards from foregn nvestment would be to mprove operatonal effcency n smallszed hotels. Furthermore, specfc polcy ntatves desgned to assst hotels groups could be mplemented through the dfference n technologes. For example, foregn nvestment hotels should focus on allocate labour effcency that should be replaced by modern technologes whereas domestc nvestment hotels or hotels whch earn revenue from only one source of ncome (room rate) could ntend to acheve effcency n asset management. The polces towards small hotels mght need to be dfferent from large hotels that enable the government to establsh approprate polces for several types of Thaland hotels. 3.References Akarapong Untong. (2004), Effcency of hotel and guesthouse n Upper North of Thaland, Documentaton of specal lectures, Faculty of Management and Informaton Scences, Naresuan Unversty. (In Tha) Akarapong Untong. and Mngsarn Kaosaard. (2009), The manageral effcency change of hotels n Chang Ma. Thammsat Economc Journal, 27(3): 126. (In Tha) Assaf, A., Barros, C.P. and Josassen, A. (2009), Hotel effcency: A bootstrapped metafronter approach. Internatonal Journal of Hosptalty Management (Artcle n Press). Barros, C.P. (2006), Analysng the rate of techncal change n the Portuguese hotel ndustry. Toursm Economcs, 12(3): 325 346. Barros, C.P. and Deke, P.U.C. (2008), Techncal effcency of Afrcan hotels. Internatonal Journal of Hosptalty Management, 27(3): 438447. Battese, G. E. (1992), Fronter producton functons and techncal effcency: a survey of emprcal applcatons n agrcultural economcs. Agrcultural Economcs, 7(34): 185208. Battese, G. E. and Coell, T. J. (1995), A model for techncal neffcency effects n a stochastc fronter producton functon for panel data. Emprcal Economcs, 20: 325332. Battese, G.E., Rao, D.S.P. (2002), Technology gap, effcency, and a stochastc metafronter functon. Internatonal Journal of Busness and Economcs, 1(2): 8793. Battese, G.E., Rao, D.S.P. and O Donnell, C.J. (2004), A metafronter producton functon for estmaton of techncal effcences and technology gaps for frms operatng under dfferent technologes. Journal of Productvty Analyss, 21(1): 91103. Coell, T. J., Rao, D.S.P., O'Donnell, C.J. and Battese, G.E. (2005), An Introducton to Effcency and Productvty Analyss. Sprnger, New York. Coell, T.J. (1996), A gude to FRONTIER Verson 4.1: A computer program for stochastc fronter producton and cost functon estmaton. CEPA Workng Papers, No. 7/96, School of Economcs, Unversty of New England, Armdale. Farrell, M.J. (1957), The measurement of productve effcency. Journal of the Royal Statstcal Socety, Seres A (General), 120(3): 253290. Greene, W. (2004), Reconsderng heterogenety n panel data estmators of the stochastc fronter model. Journal of Econometrcs, 126(2): 269 303. Hwang S.N. and Chang T.Y. (2003), Usng data envelopment analyss to measure hotel manageral effcency change n Tawan. Toursm Management, 24(3): 357369. Mngsarn Kaosaard and Akarapong Untong (2005), Benchmarkng the hotel ndustry of Thaland. Proceedng of Asa Pacfc Toursm Assocaton 11th Annual Conference New Toursm for AsaPacfc, July 710, 2005, at Korea Internatonal Exhbton Center Goyang, Gyeongg Provnce, Korea. O Donnell C.J. and Grffths W.E. (2006), Estmatng statecontngent producton fronters. Amercan Journal of Agrcultural Economcs, 88(1): 249266. O Donnell, C.J., Rao, D.S.P. and Battese, G.E. (2007), Metafronter frameworks for the study of frm level effcences and technology ratos. Emprcal Economcs, 34(2): 231255. O Donnell, C.J., Rao, D.S.P. and Battese, G.E. (2008), Metafronter frameworks for the study of frmlevel effcences and technology ratos. Emprcal Economcs, 34(2): 231255. Orea, L. and Kumbhakar, S.C. (2004), Effcency measurement usng a latent class stochastc fronter model. Emprcal Economcs, 29(1): 16983. Rao, D.S.P., O Donnell, C.J. and Battese, G.E. (2003), Metafronter functons for the study of ntergroup productvty dfferences. CEPA Workng Paper Seres No. 01/2003, School of Economcs, Unversty of New England, Armdale. Vllano, R., Flemng, E. and Flemng, P. (2008), Measurng regonal productvty dfferences n the Australan Wool ndustry: A metafronter approach. AARES 52nd Annual Conference. February 58, 2008, Canberra, Australa. 322
Internatonal Conference On Appled Economcs ICOAE 2011 323 ACKNOWLEDGEMENT Ths artcle s a part of Thaland Toursm: From Polcy to Grassroots (Prof. Dr. Mngsarn Kaosaard) whch supported by The Thaland Research Fund (TRF) under TRF ResearchTeam Promoton Grant (TRF Senor Research Scholar). 323