Benefits of Taiwan Coffee Festival

Transcription

1 Estmatng the Congeston Cost and Recreatonal Benefts of Tawan Coffee Festval Chn-Huang Huang * Abstract After the 91 Earthquake and landslde dsasters, the Tawanese government had fnshed the reconstructon work n Hua-Shan area. Beng held n Hua-Shan area of Gu-Keng Townshp, Yunln County, the 004 Tawan Coffee Festval attracted over one mllon of toursts and successfully promoted a local culture based on the coffee ndustry. The purpose of our research s to estmate the congeston cost and recreatonal beneft of tourst partcpaton n ths event. Frstly, we estmated the congeston cost of toursts usng double-bounded dchotomous choce contngent valuaton method (CVM). Secondly, we apply travel cost method (TCM) to calculate the consumer surplus based on toursts expendture n the event. In travel cost model, the travel cost wll contan the congeston cost. Ths research wll nnovate the evaluaton of the recreatonal benefts. Keywords: TCM, CVM, congeston cost and recreaton beneft. JEL: Q6 1. Introducton Tawan Coffee Festval was one of reconstructon works after the 91 Earthquake and landslde dsasters n Tawan. Ths event attracted 1. mllon toursts and successfully promoted a local culture based on coffee ndustry. Ths specal event provded an alternatve to regular lesure actvtes, such as Tawanese coffee-tastng, muscale, and exhbton of local agrcultural products etc. The huge number of vstors brought overcrowdng or congeston problems, whch mght have dmnshed * Assstant professor, Department of Sport Management, Natonal Tawan College of Physcal Educaton, Tawan, R. O. C. E-mal: hch0555@ms6.hnet.net. 1

2 effects on ndvdual tourst s enjoyment of the festval. The purpose of ths study s to estmate the economc recreatonal benefts and congeston cost receved by partcpants n Tawan Coffee Festval. Crowdng s a negatve effect of number of encounters wth other toursts (Shelby, Vaske and Donnelly, 1996). The level of congeston at a festval may nfluence ndvdual recreaton demand, resultng n a downward shft n the demand curve. If the congeston were mproved by expandng faclty n the event, t would ncrease ndvdual recreatonal beneft. The overcrowdng effect can be estmated drectly by askng partcpants ther wllngness to pay wth hypothetcal changes n congeston condton (Looms and Walsh, 1997). McConnell and Duff (1976) show that excess demand for recreaton ste may lead based evaluaton benefts. When we estmatng the recreatonal benefts based on consumer surplus (CS) wthout consderng congeston costs, ths can provde a based benefts (Cesaro, 1980). Reduced congeston would ncrease wllngness to pay by those who attend the recreaton ste, and the beneft estmaton procedure should be expanded to consder the effects of mproved congeston (Walsh, Mller and Gllam, 1983). So, t s mportant to access the effects of too many vstors attendng specal events. Contngent valuaton method (CVM) can evaluate congeston costs based on responses to hypothetcal scenaros (McConnell, 1977). Overcrowdng may affect ndvdual s travel costs or tme cost at recreaton ste (Jakus and Shaw, 1997). So, we add congeston cost to travel costs n our recreaton demand model and obtan an unbased estmate of the recreatonal benefts. It also llustrates the mprovement n the qualty of the recreaton ste at the same tme.. Economc theory.1 Congeston cost Jakus and Shaw (1997) ntroduced actual congeston percepton level nto

3 observaton behavor model to analyze the crowdng condton of recreaton ste. They mentoned that the key problem n determnng the economc costs of congeston s how to measure congeston cost, and suggested that later researcher can obtan monetary values of congeston from CVM. The contngent valuaton method s the stated preference approach to evaluate economc benefts ganed by toursts n recreaton actvtes. The bass of economc valuaton s welfare change measurement. We can measure the welfare change of ndvdual s decson from ther wllngness to pay (WTP) for benefts, and ther wllngness to accept (WTA) costs. In general, WTA s greater than or equal to WPT (Randall, 00). CVM s carred out by askng respondents about ther WTP for a specfc good. The questonnare of respondents may be desgned n open-ended or close-ended formats. We must descrbe the hypothetcal scenaro to be valued n detal. It s also mportant to pre-test the questonnare and examne the questons theoretcally and emprcally to determne ther effects on ndvdual bd behavor to avod varous sources of bases. The 004 Tawan Coffee Festval took place over a perod of ffty-two days from October 8 to November 8, 004, n Hua-Shan Vllage and Green Tunnel Park of Gu-Keng Townshp, Yunln County and attracted over 1. mllon of toursts to ths event. The vstor densty was calculated to be about 3,077 vsts per day. Hua-Shan s a rural vllage on a hll covered by arecas. The densty of toursts was hgh and the lack of parkng spaces lead to dffcultes n parkng and drvng, especally n Hua-Shan Vllage (where parkng s free). Ths reduced the ndvdual s satsfacton, and recreatonal benefts. The frst objectve of our research s to estmate a bd functon of contngent approach that reflects the vstor s wllng to pay to mtgate crowdng condton. Toursts were asked to report ther wllngness to pay for expanded park area to reduce the tme spent on watng or lookng for parkng. 3

4 Suppose that survey respondents possess a utlty functon, u( x, q, g), where x s tmes of trps, q s congeston qualty of ths event, g represents all other good. The expendture functon, e( p, q, u) s obtaned by mnmzng expendtures subject to constrant of ncome, such as, m = px + g, where m s ncome, p s prce of trps. The wllngness to pay for an mprovement of congeston qualty n Hua-Shan vllage s shown below: * WTP = e( p, q, u) e( p, q, u) (1) Where * q represents reduce congeston condton after expanded park faclty, for the sake of mtgatng the congeston of traffc and parkng. We wll take WTP nto recreaton demand as one of the travel costs components to evaluate recreatonal benefts more precsely.. Recreatonal Benefts The travel cost method (TCM) s a popular non-market valuaton technque, whch commonly measures the use value of envronmental goods, such as recreatonal benefts of toursts (Eom and Larson, 006). It s the revealed preference approach that measures use value, and s reflected by changes n demands for related market goods. TCM attempts to derve benefts from the demand functon of observed behavor, and calculates the consumer surplus based on toursts expendture on the trp. The basc assumpton of the TCM s that travel cost s proportonal to the vstor s dstance from the trp ste, and estmates consumer surplus of travelers wllngness to pay for travel. The travel costs components nclude the expendtures and travel tme wth the trp, beng proxes for the prces of the recreatonal trp. The value of tme s an opportunty cost trade-off between work and recreaton. The model contans the costs of substtute stes, as our excluson may overstate the estmatons of consumer surplus (Rosenthal, 1987). But, most researches seldom consder congeston cost when estmatng recreatonal benefts. 4

5 Jakus and Shaw (1997) suggested that congeston could affect ndvdual s cost n recreaton ste. They ntroduced congeston cost nto demand functon, measured prces as a functon of congeston. Such that x = f ( P ( C ), Pk ( Ck ), Y ) () where x s trps demand for recreaton ste, P s trps prces for ste, P s trps prces of substtute ste k, Y s vstors ncome. So, the travel costs of recreaton demand model wll contan congeston cost that toursts are wllng to pay to mtgate crowdng condton for expanded park area to reduce the tme spent on watng or lookng for parkng. 3. Data and Emprcal Model We desgned a face-to-face vstor survey n ths study. In order to meet the assumptons of the TCM, ntervewers had screened vstors for sngle-destnaton or sngle-purpose trps. An on-ste sample was conducted from October 8 to November 8, 004. Durng ths perod 150 people were asked to fll out a survey and 30 refused n Hua-Shan Vllage and Green Tunnel Park, respectvely. Total sample sze were 40, resultng n a response rate of 80%. 3.1 The double bound of dchotomous choce and survval functon We choce the double bounded dscrete choce elctaton method for ths study. Because double bounded dchotomous choce CV model s more effcent and mproved statstcal nformaton than sngle bounded model (Hanemann, Looms, and Kannnen, 1991; Scarpa & Bateman, 000; Bateman et al., 001). Ths method asks a tourst whether he (or she) s wllng to pay a pre-chosen randomly assgned amount. Respondents wll be asked face to face n ntervews wth open-end queston n pre-test, Durng the Tawan Coffee Festval, t s dffcult for vstors to fnd parkng spaces due to the large crowd. If the management agency were to buld a new parkng k lot to provde more parkng spaces, are you wllng to pay NT$ per entry for the 5

6 convenence of spendng less tme on lookng and watng for parkng? Pre-test ndcated the lkely range of bds, we had found top fve frequences ntal cost amounts were NT10, 0,50,100,00 to fnal questonnares. An ntercept survey was conducted on-ste. If the frst close-end queston was answered yes, the amounts were ncreased (approxmately double) n follow-up queston to NT$0,40,100,00,400. If the frst queston was answered wth no, the amounts were decreased (approxmately halved) n the follow-up queston to NT$5,10,5,50,100. Double bounded model generated nterval censored data on wllng to pay, whch can be analyzed usng survval analyss (Nelson, 198; Imber, Stevenson and Wlks, 1991). The survval functon s more flexble for t allows a varety of assumptons about parametrc dstrbutons, and no extra costs wth t (Le n, 1996). We follow Carson, Wlks and Imber (1994) and Le n s (1996) analyss approach, and let M be the ndvdual s maxmum wllngness to pay, M = α + σε, where s a locaton parameter, s a scale parameter, and s a random dsturbance. Suppose E s the bd prce to ndvdual, and then the probablty of rejectng the offer s defned as follows: E α E α P ( M < E) = P( ε < ) = F( ) (3) σ σ where F( ) s the cumulatve probablty functon of. The survval functon can be wrtten as E α S ( E) = 1 G( E) = 1 P( M < E) = P( M E) = Sε ( ) (4) σ Equaton (4) gves the probablty of an answer to yes. If the frst and second answers are both yes, the result s a rght censored observaton. If both answer are no then t s left censored. An nterval observaton wll be found of alternatve answers n sgn. Let be frst prce of bd value and E be second prce, I =1 E1 6

7 denote ndvdual answer bd prce E s yes, and I =0 denote the answer s no, where =1,. The log lkelhood functon of observatons s E α E1 α E1 α E α log L = I1(1 I ) log F( ) F( ) + I (1 I1) log F( ) F( ) σ σ σ σ E α E α) + I1 I log 1 F( ) + (1 I1)(1 I ) log F( σ σ (5) 3. Count data model of recreaton demand functon Count data model have been used n travel cost models of recreaton demand ncreasngly. The on-ste surveys sample s truncated for short of non-users and endogenously stratfed of frequency tourst. Shaw (1988) developed a Posson model to correct the problems of truncated and endogenous stratfcaton from on-ste surveys. The Posson probablty densty functon s λ e P( x = n) = λ n, n = 1,, (6) n! The parameter λ s mean and varance of the dstrbuton, t s common to specfy as an exponental functon, λ = exp( β ), where are ndependent varables and z z are parameters of the functon. The log-lkelhood functon of Posson s ln L = T = 1 [ zβ z β x e ln( x!)] (7) Because on-ste surveys sample s truncated for short of non-users, we modfed equaton (6) as below: h ( x x > 0) = x e λ λ x (1 exp( λ ))! The log-lkelhood functon of truncated Posson s ln L = m [ = 1 8 λ z β x ln( e 1) ln( x!)] 9 Shaw s (1988) on-ste Posson model corrected both truncaton and endogenous stratfcaton of on-ste samples, the functon s: h( x x e > 0) = ( x λ λ x 1 1)! 7

8 Let w = x 1 we can re-wrte the above equaton as below: h( x λ w e > 0) = λ 10 w! The equaton of (10) s a parameter truncated Posson s ln L = T = 1 [ w of Posson, and the log-lkelhood functon of [( x 1)! ] zβ z β ( x 1) e ln ] 11 The consumer surplus of toursts equals the area under expected demand functon, E ( x )= λ, and the wllngness to pay for access s WTP = β0 + β1c C β0 + β e x 1C e dc = = (1) 0 β1 0 β C C= C 1 where C s travel cost, β 1 s coeffcent of C, and β 1 <0. 4. Results 4.1 Estmatng of congeston cost Two hundred and forty respondents responded to dfferent bd amounts n the ntal and follow-up WTP questons. 0.4% of the respondents answered, yes, yes, 6.3% of the respondents answered yes, no, 30.0% of the respondents answered no, yes, and 3.3% of the respondents answered no, no. We followed Stacy (196) who defned the generalzed gamma dstrbuton densty functon, ηr( λe) f ( E) = rk 1 exp Γ( k) r [ ( ηe) ] where η = exp( α ) s locaton parameter, r = 1/ σ s the scale parameter, (13) k = 1/ δ s the shape parameter and Γ s the gamma functon. When r=k=1 (or σ = δ = 1), the dstrbuton s the exponental dstrbuton. If k=1 (or δ = 1), t s a Webull dstrbuton, and k (or δ ) tends to nfnty, t results n the lognormal dstrbuton. Our valuaton functon adopt lognormal, Webull, and exponental dstrbutons to explan respondents wllngness to pay by double bounded data. The explanatory 8

9 varables ncluded GENDER, a dummy varable of gender (code 1 for male, 0 for female); LINCOME s the log of the ndvdual s monthly ncome, TIMES s the number of tmes the respondent vsted ths event, and LOCAL s a dummy varable of event locaton (code 1 f respondent vst Green Tunnel Park; 0 f respondent vst Hua-Shan Vllage). We were usng a lkelhood rato to test goodness-of-ft of three survval evaluaton models. In table 1 only lognormal dstrbuton reject the null hypothess wth all parameters equal to zero wth p<0.01. As a result, the generalzed gamma dstrbuton gave the lognormal dstrbuton as the best representaton of the emprcal data. In the lognormal dstrbuton evaluaton model, GENDER and LOCAL varables are sgnfcantly dfferent from 0 at =0.1. The coeffcent of GENDER s postve to evaluated model, showng that male had hgher ntenton to offer bd of wllngness to pay than female. The coeffcent of LOCAL s negatve to estmated model, demonstratng that respondents who vsted Hua-Shan Vllage would pay more than those that vsted Green Tunnel Park to reduce the tme spent on watng or lookng for parkng. The advantages of measurng welfare by medan s that t not only rule out extreme observatons, but t s also better than mean senstve to specfc dstrbuton (Le n, 1996). In addton, Cooper, Hanemann and Sgnorello (00) had compared the confdence nterval for double bounded model, and showed that mean WTP s more based than medan WTP assocated wth the follow-up responses. So, our research adopted medan WTP, and calculated by χ χ log( WTP ), = 0. 5 = Xβ + σε χ (14) The medan for lognormal dstrbuton was 68 NT$ n table 1. Ths was a respondent s wllngness to pay to mtgate crowdng condton for expanded 9

10 parkng area to reduce the tme spent on watng or lookng for parkng. We wll add mprovement qualty cost to toursts travel costs to estmate recreatonal benefts. Table 1. Survval valuaton functon of wllngness to pay Varable Lognormal Webull Exponental INT 5.9 (5.44) 5.8 (5.8) 5.9 (4.14) GENDER 0.9 (.11)* 0.6 (.07)* 0.7 (1.47) LINCOME (1.00) (0.63) (0.48) TIMES -0.0 (0.41) (1.9) (0.65) LOCAL (3.33)** (.68)** (.09)* SCALE 0.81 (15.96)*** 0.33 (14.88)*** 1 (8.61)*** Log Lkelhood Log-lkelhood rato 13.59*** 1.65** 6.61 WTP (NT$) Log-lkelhood rato=(-)*(restrcted Log-lkelhood- Log Lkelhood), χ ( 4,0.99 ) = *p<0.1, **p<0.05, ***p<0.01, t values n parentheses. 4. Recreatonal benefts Recreaton demand functon at a gven ste has a relatonshp between number of trps by an ndvdual and trp prce or personal preference. The non-negatve and nteger nature of trp demand sutes the count data model to estmate recreatonal benefts. The applcaton of count data model to assess recreaton ste demand by adoptng on-ste survey always encounters the problems of askng the frequency of vsts and truncated non-user samples. So, we follow Shaw s (1988) on-ste Posson model to correct these two evaluaton problems. The demand functon of ndependent varables ncludes TRACOST, SUBCOST, STAY, MARRIAGE and EDUYEAR. TRACOST s travel cost contanng 10

11 transportaton costs of vstors, convertng round-trp dstance from home to destnaton ste nto dollars, and costs ncurred on-ste. It also contans opportunty cost of tme, usng one-thrd wage rate to convert travel tme to dollars. Wthout estmatng travel tme for a recreaton ste, the consumer surplus of benefts wll underestmate. SUBCOST s substtute ste travel cost of vstors. We calculated the cost of reduced congeston added to travel costs and compared t wth no congeston cost varables, such as N TRACOST and SUBCOST. The substtuton prce has been measured as the dstance from home to alternatve ste offerng smlar to study ste. In ths research the substtute ste s Tongshan Townshp n Tanan County, whch cultvates coffee trees and holds Tongshan Coffee Festval the same as Hua-Shan Vllage. If recreaton demands functon does not nclude the substtuton prce, the consumer surplus wll be overestmated. STAY denotes stay tmes of respondents. We follow Chavas, Stoll, and Sellar (1989) where the researches treat ndvduals on-ste tme as a commodty, and postve to recreaton demand. Marrage denotes ndvdual s martal status (marred or sngle). EDUYEAR represents partcpant s years of educaton been educated years. Table provdes the summary statstcs for the varables ncluded n the econometrc models. Table Varable summary statstcs Varable Mean Standard TRIP TRACOST (NT$) SUBCOST (NT$) 1, NTRACOST (NT$) NSUBCOST (NT$) STAY (hours) MARRIAGE (marred=1, sngle=0) EDUYEAR

12 The emprcal results of on-ste Posson equatons ncluded wth and wthout congeston cost models are lst n Table 3. The lkelhood rato statstc test models goodness-of ft under the null hypothess that all parameters are zero can be rejected. Wth respect to the coeffcent estmates, all ndependent varables are sgnfcantly dfferent from 0 at the level p<0.05. The prce or travel cost and substtuton prce coeffcents were consstent wth demand theory for both on-ste Posson equatons. The on-ste tme s postvely lnked to trps demand, where the longer the tme of stay at Tawan Coffee Festval, the more they attend the event. In other words, Tawan Coffee Festval actvtes have attracted travelers to enjoy the programs. MARRIAGE s postvely related to trps demand, meanng that marred couples generally jon n Tawan Coffee Festval actvtes more than sngle people. But, EDUYEAR s negatvely related to dependent varable, so the hgher the educaton of vstors, the less lkely they are to partcpate n event actvtes. Table 3. Parameter estmates for the travel cost model Varable Wth congeston cost Wthout congeston cost INT (1.080) TRACOST ( )*** SUBCOST (1.9655)** STAY ( )*** MARRIAGE (.354)** EDUYEAR ( )** ( ) ( )*** (.00831)** ( )*** (.3536)** ( )** Restrcted log lkelhood Log-lkelhood rato Log-lkelhood rato=(-)*(restrcted Log-lkelhood- Log Lkelhood), χ (5,0.99 ) = *p<0.1, **p<0.05, ***p<0.01, t values n parentheses. 1

13 Consumer surplus s obtaned by equaton (1), and evaluated at mean of the trps dvded by travel cost (prce) coeffcent. We estmated the mean CS of every tourst n both travel cost models. The consequence of whch wth and wthout congeston cost models CS s NT$ 551,540, respectvely. The results demonstrated that, respondents were wllng to pay more to releve the overcrowdng condtons of the event by expanded parkng. The net of mprovng congeston cost s NT$ 11. Accordng to Gu-Keng Townshp, ths annual event attracted 1. mllons of toursts. Gven ths number of toursts, the total recreatonal beneft contanng mproved congeston cost of the 004 Tawan Coffee Festval s approxmately NT$ 661. mllons and the congeston cost s approxmately NT$ 13. mllons. 5. Concluson Double bound dchotomous choce model of CVM has been used n ths study to estmate the toursts medan congeston cost of mproved qualty of parkng. The estmated wllngness-to-pay by survval functon CS was NT$68, and the total congeston cost s approxmately NT$ 13. mllons, the ncremental recreatonal effects. In addton, an on-ste Posson model of TCM, truncated and endogenously stratfed s adopted to evaluate the use value by calculatng the consumer surplus. We found the mean CS wth congeston cost of average partcpant was NT$551, as estmated by count data model, and the total recreatonal beneft of the 004 Tawan Coffee Festval s approxmately NT$ 661. mllons. It s reasonable to suggest that the vstors could ncrease ther recreatonal benefts by REDUCING the congeston condton of ths event and the qualtes of resources should be consdered. The management could expand the parkng faclty especally n Hua-Shan Vllage to offer opportunty to extract some of the large consumer surplus enjoyed by toursts. 13

14 References Bateman, I. J., Langford, I. H., Jones, A. P., and Kerr, G. N. (001). Bound and path effects n double and trple bounded dchotomous choce contngent valuaton. Resource and Energy Economcs, 3, Carson, R. D., Wlks, L., & Imber, D. (1994). Valung the preservaton of Australa s Kakadu Conservaton Zone. Oxford Economc Papers, 46, Cesaro, F. J. (1980). Congeston and the valuaton of recreaton benefts. Land economcs, 56, Chavas, J., Stoll, J., and Sellar, C. (1989). On the commodty value of travel tme n recreatonal actvtes. Appled Economcs, 1, Cooper, J. C., Hanemann, M. and Sgnorello, G. (00). One-and-one-half-bound dchotomous choce contngent valuaton. The Revew of Economcs and Statstcs, 84, Eom, Young-Sook and Larson, D. M. (006). Improvng envronmental valuaton estmates through consstent use of revealed and stated preference nformaton. Journal of Envronmental Economcs and Management, 59, Hanemann, W., Looms J. & Kannnen, B. (1991). Statstcal effcency of double-bounded dchotomous choce contngent valuaton. Amercan Journal of Envronmental Economcs, 73, Imber, D., Stevenson, G. and Wlks, L. (1991). A contngent valuaton survey of the Kakadu Conservaton Zone, Canberra. Australan Government Publshng for the Resource Assessment Commsson. Jakus, P. & Shaw, D. (1997). Congeston at recreaton area: emprcal evdence on perceptons, mtgatng behavour and management preferences. Journal of Envronmental Management, 50, Le n, C. J. (1996). Double bounded survval values for preservng the landscape of natural parks. Journal of Envronmental Management, 46, Looms, J. B. and Walsh, R. G. (1997). Recreaton Economc Decson: Comparng Benefts and Costs. Pennsylvana: Venture Publshng, Inc. McConnell, K. E. (1977). Congeston and wllngness to pay: a study of beach use. Land Economcs, 53, McConnell, K. E. and Duff, V. A. (1976). Estmatng net benefts of outdoor recreaton under condtons of excess demand. Journal of Envronmental Economcs and Management,, Nelson, W. (198). Appled Lfe Data Analyss. Now York: John Wley. Randall, A. (00). Valung the Outputs of Multfunctonal Agrculture. European Revew of Agrcultural Economcs, 9, Rosenthal, D. H. (1987). The necessty for substtute prces n recreaton demand 14

15 analyss. Amercan Journal of Agrcultural Economcs, 69, Scarpa, R. & Bateman, I. (000). Effcency gans afforded by mproved bd desgn versus follow-up valuaton questons n dscrete-choce CV studes. Land Economcs, 76, Shaw, D. (1988). On-ste samples regresson: Problem of non-negatve ntegers, truncaton and endogenous stratfcaton. Journal of Econometrcs, 37, Shelby, B., Vaske, J.J., and Donnelly, M. P. (1996). Norms, standards, and natural resource. Lesure Scence, 18, Stacy, E. W. (196). A generalzaton of the gamma dstrbuton. Annals of Mathematcal Statstcs, 33, Walsh, R. G., Mller, N. P., & Gllam, L. O. (1983). Congeston and wllngness to pay for expanson skng capacty. Land Economcs, 59,