ECONOMIC ANALYSIS OF TOURISM CONSUMPTION DYNAMICS: A Time-varying Parameter Demand System Approach

Transcription

1 ECONOMIC ANALYSIS OF TOURISM CONSUMPTION DYNAMICS: A Tme-varyng Parameer Demand Sysem Approach Dors Chenguang Wu 1 Sun Ya-sen Unversy, Chna Gang L Unversy of Surrey, Uned Kngdom Hayan Song The Hong Kong Polyechnc Unversy, Hong Kong SAR Dors Chenguang Wu ABSTRACT Ths sudy consders he dynamcs of he consumpon behavour of ourss from an economc perspecve. The evoluon of varous demand elasces s explored usng a me-varyng parameer almos deal demand sysem model. The op four source markes for oursm n Hong Kong are examned, and hree major ours expendure caegores, ncludng shoppng, hoel accommodaon and meals ousde hoels, are nvesgaed for each marke. Elascy analyss reveals dfferen consumpon rends and paerns across he source markes. The fndngs wll serve as a useful reference for Hong Kong oursm-relaed ndusres and he governmen n her effors o enhance he compeveness of Hong Kong as an nernaonal oursm desnaon. Keywords: oursm consumpon, me-varyng parameer, almos deal demand sysem, demand elascy, evoluon, Hong Kong. 1 Dors Wu, PhD s an asssan professor n he Busness School, Sun Ya-sen Unversy, Guangzhou, Chna; Gang L, PhD s a senor lecurer n economcs n he Faculy of Managemen and Law a he Unversy of Surrey, UK (Faculy of Busness, Economcs and Law, Unversy of Surrey, Guldford GU2 7XH, Uned Kngdom. Emal<g.l@surrey.ac.uk>); and Hayan Song, PhD s char professor n he School of Hoel and Toursm Managemen, The Hong Kong Polyechnc Unversy, Hong Kong. 1

2 INTRODUCTION Inernaonal oursm plays an ncreasngly sgnfcan role n naonal and global economes. Tours expendure on shoppng, hoel accommodaon, food servces, ranspor, vsor aracons, eneranmen and so forh conrbues subsanally o developng and developed economes by generang gross domesc produc (GDP), creang employmen and provdng soco-economc developmen opporunes. In 2008, more han 80 counres earned over US$1 bllon from nernaonal oursm, wh worldwde receps reachng $944 bllon (UNWTO, 2009). Gven he mporance of oursm consumpon o a desnaon s economy and s drec lnk o he performance of oursm-relaed busnesses, oursm demand analyss has araced ncreased aenon from boh academcs and praconers. I s of grea mporance o sudy he consumpon behavour of ourss from an economc perspecve. The economc sudy of oursm s se whn a well-esablshed academc dscplne, wh a frm heorecal foundaon and well-esed mehodologes (Sabler, Papaheodorou & Snclar, 2010). In hs dscplne, he modellng of oursm demand s one of he mos developed and rgorous areas of he economc analyss of oursm (Sabler e al., 2010, p. 425). In parcular, demand elascy analyss can be used o assess ourss consumpon behavour scenfcally and o generae useful manageral mplcaons for oursm-relaed ndusres and ours desnaons. Varous demand elasces, such as ncome elascy, own-prce elascy and cross-prce elascy, can be calculaed based on an esmaed demand model o deermne he sensvy of ours demand for oursm goods and servces o changes n a ours s dsposable ncome, he prce of oursm producs and he prce of alernave consumpon producs. Such analyses can help oursm organsaons o manage and promoe a desnaon effecvely n a compeve envronmen, and can gude oursm servce provders n seng approprae prcng sraeges (Corés-Jménez, Durbarry & Pulna, 2009). Ths sudy ams o provde new nsghs no he dynamcs of Hong Kong nbound ourss consumpon behavour regardng a group of oursm goods and servces and o evaluae he compeveness of Hong Kong as an nernaonal ours desnaon. The remander of he paper s organsed as follows. In he nex secon, a bref revew of he relevan leraure on oursm consumpon s provded, wh a parcular focus on pas emprcal sudes usng he almos deal demand sysem (AIDS) model. Secon 3 ncludes he dscusson of varous versons of he AIDS model, followed by an nroducon o he emprcal research desgn and daa collecon. The resuls of he emprcal sudy are presened and dscussed n deal n Secon 4. The concludng remarks and manageral mplcaons of he sudy s fndngs are presened n Secon 5. 2

3 RELEVANT LITERATURE A varey of modern economerc approaches have been employed o model he relaonshp beween oursm demand and s economc deermnans. These nclude he auoregressve dsrbued lag model, error-correcon model, vecor auoregressve model and sysem-of-equaons approach (see L, Song & W, 2005, and Song & L, 2008, for comprehensve revews of he applcaon of hese mehods n oursm demand analyss). The sysem-of-equaons approach n parcular has several advanages over s sngle-equaon counerpars. By ncludng a number of oursm goods and servces, such as resauran dnng and hoel accommodaon, n he sysem and esmang hem smulaneously, a sysem-of-equaons model allows he researcher o nvesgae how a ours chooses a bundle of goods and servces o maxmze hs or her uly whn ceran budge consrans. I also allows he nerrelaonshps amongs alernave oursm goods and servces n a demand sysem he subsuon or complemenary effecs o be evaluaed effecvely (L, Song & W, 2004). Toursm Consumpon Analyss Usng he Almos Ideal Demand Sysem Whn he famly of complee sysems of demand equaons, he AIDS model has araced he mos aenon n consumer demand sudes, ncludng hose relaed o oursm demand. Ths s due o s superory over oher mehods such as he lnear expendure sysem, he ndrec addlog model, he Roerdam model and he double logarhmc sysem (Thomas, 1993). The AIDS model s explcly derved from economc heory, and has a srong heorecal underpnnng. I s capable of analysng he demand for dfferen oursm goods and servces and her neracons, and allows he cross-prce elasces amongs varous oursm goods and servces o be esmaed drecly. The basc assumpons of economc demand heory homogeney and symmery can also be consdered n he sysem (Wu, L & Song, 2011). Because of s srong heorecal underpnnng, he AIDS model provdes more accurae resuls for demand elascy compared o sngle-equaon demand models. The AIDS model has been appled wh varous model specfcaons and esmaon mehods o oursm consumpon analyss n wo ways. The frs, whch has a greaer number of emprcal applcaons, concerns he analyss of he demand for oubound oursm across several desnaons. Each desnaon under sudy s regarded as a oursm produc whn he demand sysem, and he naure of he desnaons (.e., luxury or necessy) and her relaonshps (.e., subsuon or complemenary) are denfed hrough demand elascy analyss. Such analyses generae useful fndngs on desnaon markeng and managemen. For example, Whe (1985) employs he AIDS model o esmae he budge allocaons of US ourss amongs wesern European counres, denfyng he luxury desnaons and he complemenary or subsuon relaonshps beween desnaons. 3

4 In he recen oursm leraure, an error-correcon verson of he AIDS model (EC-AIDS) has been nroduced o analyse he shor-run dynamcs of demand for varous ours desnaons. For nsance, Durbarry and Snclar (2003) develop an EC-AIDS model o analyse he demand for oursm o Ialy, Span and he Uned Kngdom by French resdens. Ialy and he Uned Kngdom are denfed as luxury desnaons, and Span s found o be a subsue desnaon for Ialy and he Uned Kngdom as far as French ourss are concerned. L e al. (2004) employ an EC-AIDS model o examne UK resdens oubound oursm demand for several counres n wesern Europe, and show ha ravellng o mos desnaons n hs regon appears o be a luxury for UK ourss n he long run, and ha he subsuon or complemenary effecs vary across desnaons. Corés-Jménez e al. (2009) use an EC-AIDS model o analyse he oubound Ialan oursm demand for four European counres: France, Germany, Span and he Uned Kngdom. They fnd ha France and Germany are subsue desnaons n he shor run for Ialan ourss. The second caegory of AIDS applcaons n oursm s he examnaon of ourss budge allocaon amongs several oursm produc caegores, ncludng accommodaon, food, ranspor, shoppng and eneranmen. Such sudes are very few, manly because of he dffculy of obanng ours expendure and prce daa a he produc caegory level. Fuj, Khaled and Mak (1985), he frs o explore hs applcaon, employ a basc AIDS model o analyse ours demand for food and drnk, lodgng, recreaon and eneranmen, local ranspor, clohng and oher ems n Hawa. Dvsekera (2009) analyses he demand for accommodaon, food, ranspor, shoppng and eneranmen amongs en oursm source markes for Ausrala, and denfes he complemenary relaonshps of he demand for hese expendure caegores by calculang he cross-prce elasces. These wo sudes are based on he sac AIDS model, whch focuses solely on long-run demand elasces. Wu e al. (2011) employ an EC-AIDS model o analyse he shor-run dynamcs of he demand for shoppng, hoel accommodaon, meals ousde hoels and oher oursm goods and servces n Hong Kong. They consder egh major source markes, and fnd subsuon relaonshps amongs he expendure caegores n mos cases. All of hese sudes calculae he average long-run or shor-run demand elasces over he sample perod, bu do no consder he evoluon of ourss consumpon behavour over me. Evoluon of Toursm Consumpon: The Tme-varyng Parameer Approach The esmaon mehods used n he economerc analyss of oursm demand can be caegorsed no wo ypes: consan parameer and me-varyng parameer (TVP) approaches. In consan parameer demand models, he coeffcens of he explanaory varables reman unchanged for he enre esmaon perod. Thus, once he models have been esmaed, he average mpac of deermnans on oursm demand over he sample perod can be evaluaed. The majory of emprcal sudes adop he consan parameer approach o esmae oursm demand models. However, n pracce, he relaonshp beween oursm demand and s deermnans may vary n response o echnologcal developmens, preference shfs amongs ourss and mprovemens n 4

5 lvng sandards. The TVP approach, whch s based on a recursve esmaon of he parameers n a model, s capable of capurng hese dynamcs and he evoluon of he demand sysem, because he coeffcens are allowed o vary over he sample perod. In oher words, consan parameer models can esmae only he average mpac of deermnans on oursm demand over he sample perod, whereas her TVP counerpars can race he evoluon of ours behavour over me. Due o hs advanage of TVP models over her consan parameer counerpars, hs sudy apples he TVP approach o examne he hsorcal evoluon of ourss consumpon behavour n relaon o dfferen oursm producs. Despe he developmen of he TVP echnque n economercs and s advanages for consumer demand analyss (Harvey, 1989), s applcaon n oursm research s sll lmed. Song and Wong (2003) employ a TVP model o examne he demand for Hong Kong oursm by resdens of 16 major oursm orgn counres. Rddngon (1999) compares he forecasng performance of TVP models wh ha of her consan parameer counerpars, and provdes emprcal evdence of he greaer accuracy of he former. Song, W and Jensen (2003) employ nernaonal oursm daa o compare he forecasng performance of a TVP model wh seven oher economerc and me-seres approaches, and fnd ha he TVP model provdes he mos accurae one-year-ahead forecass. L, Wong, Song and W (2006) combne he TVP echnque wh an error correcon model (ECM), and show he superor forecasng performance of he TVP-ECM compared o oher economerc and me-seres models. All of he aforemenoned applcaons of he TVP echnque are n he conex of sngle-equaon demand models. Thus far, only one sudy n he oursm leraure has appled he TVP approach o AIDS models. L, Song and W (2006) nroduce he TVP echnque no boh he sac and error-correcon versons of he AIDS model (TVP-AIDS and TVP-EC-AIDS, respecvely) o examne he long-run and shor-run evoluon of Brsh ourss consumpon behavour n relaon o a group of European desnaons. However, due o sample sze consrans, he equaons n he developed TVP-AIDS and TVP-EC-AIDS models are esmaed separaely, raher han smulaneously. Ths means ha he cross-equaon symmery resrcon s no esed, and consequenly fully resrced (homogeneous and symmerc) TVP-AIDS and TVP-EC-AID models are no consdered. To he bes of our knowledge, no research has appled he TVP model o analyse he evoluon of oursm consumpon wh regard o dfferen oursm goods and servces and her neracons n a ours desnaon. The presen sudy ams o brdge hs gap n he leraure by combnng he advanced TVP esmaon echnque wh a heorecally sound sysem-of-equaons demand model he AIDS o analyse how ourss allocae her expendure o varous oursm goods and servces n he conex of Hong Kong as an nernaonal ours desnaon. Ths sudy bulds upon prevous research applyng he AIDS n hs way (e.g., Wu e al., 2011) by ncorporang he 5

6 TVP echnque o examne he evoluon of oursm consumpon behavour. I also exends he work of L, Song and W (2006), as he equaons n he demand sysem are esmaed smulaneously raher han equaon by equaon. All of he heorecal resrcons are verfed, ncludng homogeney, symmery and he jon assumpon of boh. The resulng homogeney and symmery resrced TVP-EC-AIDS model s esmaed and used for he evaluaon of shor-run demand elasces and he evoluon of ourss consumpon behavour. METHODOLOGY Gven he advanages of boh he TVP echnque and he AIDS model, as dscussed, a TVP-EC-AIDS model s proposed n hs sudy o examne he dynamcs and evoluon of ourss consumpon behavour n relaon o several oursm expendure caegores. Ths secon nroduces he basc AIDS model proposed by Deaon and Muellbauer (1980), and dscusses he dynamc verson of he model. The TVP-EC-AIDS model s hen nroduced, whch forms he bass for he emprcal analyss n hs sudy. Basc AIDS Model Deaon and Muellbauer (1980) frs proposed he basc AIDS model o analyse he demand for a number of goods and servces smulaneously. The model specfcaon s shown as follows. x w = α + γ j log p j + β log + ϕkdumk + ε P, (1) j k where w ( = 1,, n) s he budge share of he h caegory of oursm goods and servces (here, n caegores of oursm goods and servces consue a complee demand sysem), p j s he prce of he jh oursm produc, x s he oal expendure on all oursm goods and servces n he sysem, P s an aggregae prce ndex, x/p s he real oal expendure per capa, dum s he kh dummy varable ha capures he k effec of a one-off even, α, γ j, β and ϕ k are parameers o be esmaed and ε s a dsurbance erm. The Tornqvs prce ndex (Tornqvs, 1936) s used as he proxy for he aggregae prce ndex n Equaon 1 due o s advanages over oher aggregae prce ndces (see Wu e al., 2011). As a complee sysem s suded, he budge shares should add up o uny, whch requres ha α = 1, γ = 0 and β = 0 n Equaon 1. To avod he j sngulary problem n he resdual varance-covarance marx when seng he dependen varables (budge shares) o add up o uny, one equaon s normally omed from he sysem n he esmaon, and s coeffcens are hen calculaed laer based on he addng-up rule (Deaon & Muellbauer, 1980). A he same me, wo basc resrcons homogeney and symmery are mposed on he parameers accordng o he heorecal properes of demand heory whn he AIDS framework. These resrcons are also mposed n oher versons of he AIDS model dscussed below. 6

7 EC-AIDS Model The dynamc verson of he AIDS (.e., EC-AIDS), raher han he basc AIDS model, s employed here o model ourss consumpon behavour because he EC-AIDS s capable of reflecng connuous adjusmens n such behavour n relaon o new equlbrums of he demand sysem (see Wu e al., 2011). The EC-AIDS model s specfed as x w, = α + λµ, 1 + γ j log p j, + β log + ϕkdumk + ε, P, (2) j where s he dfference operaor, and µ, 1, he esmaed resdual erm from he basc AIDS model (.e., Equaon 1), s he error-correcon erm ha measures he adjusmen of he decson errors made n he prevous perod. The un roo es suggess ha all of he varables ncluded n Equaon 1 are negraed of order one, and ha a conegraon relaonshp exss amongs hem, whch suppors he adopon of he EC-AIDS model. Ths resul s n lne wh prevous sudes adopng he AIDS, such as hose of Corés-Jménez e al. (2009) and L e al. (2004). TVP-EC-AIDS Model To ncorporae he TVP echnque no he EC-AIDS model, he sysem s specfed n a sae space form ha ncludes wo ypes of equaons. The frs s he sgnal equaon, whch descrbes he relaonshp beween oursm demand and s deermnans (see Equaon 3, whch s dencal o Equaon 2 apar from he subscrps of he parameers). The oher s he sae equaon, whch conrols he varaon n he coeffcens n he sgnal equaon over me (see Equaons 4-7). x w, = log ε, (3) =, 1 α, + λ µ, 1 + γ j, log p j, + β, + ϕkdumk + j P k α, α + ϑ, (4) λ = λ 1 + θ, (5) γ γ + ρ j, = j, 1, (6) β β + τ, =, 1, (7) where ϑ, θ, ρ and k τ are dsurbance erms. An effcen recursve algorhm named he Kalman fler (Kalman, 1960) s used o esmae hs sae space model. Deals of he Kalman fler algorhm are dscussed n Durbn and Koopman (2001) and Harvey (1989). Demand Elasces As he parameers n he TVP-EC-AIDS model are specfed o change over me, he calculaed demand elasces also vary across he sample perod. The expendure, 7

8 β, elascy of demand s calculaed as ε x, = + 1. The calculaon of he compensaed w own-prce and cross-prce elasces s also based on hese formulae; ha s, γ, j, ε, = + w, 1 and ε j, = γ + w j,, respecvely. As he subscrps of hese w w,, formulae sugges, n he TVP-EC-AIDS model, he esmaed coeffcens ( β, γ and γ j ) have dfferen values a dfferen pons n me. In addon, dsnc from he, consan parameer AIDS models, n whch he average of w, (.e., w ) s used for he elascy calculaon, n he TVP-EC-AIDS model, he enre me seres of w, s employed o generae me-varyng demand elasces. Thus, he calculaed demand elasces vary over he sample span, and he evoluon of ourss consumpon behavour can be analysed over me. For nsance, based on he calculaed demand elasces, ourss responses o changes n he oal expendure budge or he prces of dfferen caegores of oursm goods and servces can be evaluaed. Furhermore, he evoluon of consumpon paerns across dfferen oursm source markes can be denfed and compared. Daa for he Emprcal Sudy Ths sudy employs Hong Kong nbound oursm daa o demonsrae he esmaon of he TVP-EC-AIDS model emprcally. The four larges source markes for Hong Kong oursm manland Chna, Japan, Tawan and he Uned Saes are nvesgaed. Manland Chna has been he leadng source marke for Hong Kong s nbound oursm snce Accordng o he Hong Kong Toursm Board (2009), oal ours expendure n Hong Kong was US$12.1 bllon (HK$94.2 bllon) n 2008, and over half of ha fgure (56.5%) was conrbued by manland Chnese ourss. Vsor arrvals from manland Chna reached abou 16.9 mllon n he same year, accounng for 57.1% of oal arrvals n Hong Kong. The rapd ncrease n personal dsposable ncome and he apprecaon of he Chnese currency (he renmnb) n recen years has fuelled he growh of oursm demand from manland Chna. The counry now makes a vas conrbuon o Hong Kong nbound oursm, and a sudy of he consumpon behavour of manland Chnese ourss s hus of grea praccal mporance. Japan, Tawan and he Uned Saes accouned for 5.3%, 3.7% and 3.5%, respecvely, of he overall ours expendure n Hong Kong n I should be noed ha he presen sudy focuses on he budge allocaon of ourss n Hong Kong afer hey have pad for her ransporaon beween her home counry and Hong Kong. Ths s because he relevan daa on nernaonal ranspor expenses are unavalable. Thus, he mpac of ranspor mode choce and relaed expenses on oal oursm consumpon, and he relaed cross-prce elascy analyss, s beyond 8

9 he scope of he curren sudy. The Hong Kong Toursm Board (HKTB) releases annual oursm expendure daa on he followng caegores of goods and servces by source marke: shoppng, hoel accommodaon, meals ousde hoels, eneranmen, ours and oher expendure. These daa are obaned from vsor surveys regularly conduced by he HKTB. However, as he prce ndces on eneranmen and ours are unavalable, hese wo caegores are merged no he oher expendure caegory. Thus, for each source marke, a oursm demand sysem ha conans four oursm expendure caegores, ncludng shoppng, hoel accommodaon, meals ousde hoels and oher expendure, s esablshed. In each demand sysem, ourss allocae her expendure budge amongs he four produc caegores wh he am of maxmsng her uly gven her oal budge consran. Parcular aenon s pad n he emprcal analyss o he consumpon behavour relaed o he frs hree produc caegores, and he fndngs of he oher expendure caegory are excluded from he dscusson as hey do no have specfc praccal mplcaons. Annual daa from 1984 o 2008 (from 1993 o 2008 for manland Chna due o daa avalably) are employed for he model esmaon. The prce ndex for hoel accommodaon s consruced usng he weghed average of room raes across dfferen hoel ypes n Hong Kong hgh arff A hoels, hgh arff B hoels, medum arff hoels and ours gueshouses as repored by he Hong Kong Hoels Assocaon. The prce ndex for shoppng s consruced usng he weghed average of he prces of he man oursm producs purchased by nbound ourss: clohng and foowear, jewellery, waches, cameras and opcal producs, cosmecs and personal care producs and oher real ems. Ther expendure shares are repored by he HKTB. The prce ndex for oher real ems s approxmaed usng he real prce ndex. The prce ndex for meals ousde hoels by ourss s approxmaed by he prce ndex for meals ousde hoels by Hong Kong resdens. The consumer prce ndex n Hong Kong s used as a proxy for he prce ndces for he oher ypes of expendure. All prces employed o derve he prce ndces of shoppng and meals ousde hoels are obaned from he publcaons of he Hong Kong Census and Sascs Deparmen. Usng hese daa, four demand sysems are esmaed each of whch relaes o one of he four key source markes and he evoluon of he demand elasces (he expendure, own-prce and cross-prce elasces) s examned accordngly. EMPIRICAL RESULTS Descrpve Analyss Fgure 1 shows he ours expendure shares of he varous oursm goods and servces for each source marke under consderaon over he sample perod. Varous rends n spendng behavour can be observed. For ourss from he hree Asan markes Japan, Tawan and manland Chna expendure on shoppng has 9

10 hsorcally domnaed he spendng budge. However, for US ourss, who are long-haul ravellers makng a relavely long say n Hong Kong, he hghes spendng snce 1989 has been on hoel accommodaon. Lkely because of her smlar culural roos, manland Chnese and Tawanese ourss dsplay he mos smlar consumpon behavour, spendng a much hgher proporon of her ravel budges on shoppng and a much lower proporon on hoel accommodaon and meals. From a dynamc perspecve, he share of expendure on shoppng has been ncreasng over me amongs manland Chnese ourss, whereas he oppose rend s seen n he oher markes. The expendure gap beween shoppng and hoel accommodaon has been narrowng snce he 1980s amongs Japanese ourss bu wdenng n he manland Chnese marke snce he 1990s. Meals ousde hoels accoun for he smalles share of oursm expendure across all four markes (wh he excepon of only a couple of years n he manland Chna case). Ths caegory of spendng also shows he mos sable rend over me, alhough US and Japanese ourss have spen slghly more on n recen years. Fgure 1. Share of ours expendure on dfferen caegores of goods and servces. 10

11 Table 1 Fnal sae esmaes of he TVP-EC-AIDS models Shoppng equaon Consan ln(p_shoppng)-ln(p_oher) ln(p_hoels)-ln(p_oher) ln(p_meals)-ln(p_oher) ln(x/p) Error-correcon erm (-1) Hoel accommodaon equaon Consan ln(p_shoppng)-ln(p_oher) ln(p_hoels)-ln(p_oher) ln(p_meals)-ln(p_oher) ln(x/p) Error-correcon erm (-1) Meals ousde hoels equaon Consan ln(p_shoppng)-ln(p_oher) ln(p_hoels)-ln(p_oher) ln(p_meals)-ln(p_oher) ln(x/p) Error-correcon erm (-1) Enre sysem Manland Chna USA Japan Tawan (1.203) (-1.502) (1.829) (0.313) * (-2.427) ** (-6.409) (-0.384) (1.829) (-0.550) (1.315) 0.081* (1.991) ** (-6.409) (-0.683) (0.313) (1.315) (0.880) (-0.153) ** (-6.409) (0.436) (0.033) ** (-3.079) (-0.875) 0.186** (2.821) (-0.528) (0.804) ** (-3.079) (1.429) (1.112) * (-1.539) (-0.528) (0.071) (-0.875) (1.112) (1.918) (-1.389) (-0.528) (-0.007) (0.441) (-0.251) (0.538) 0.193** (5.181) (-0.720) (0.011) (-0.251) (1.117) 0.096* (2.310) ** (-5.883) (-0.720) (-1.366) (0.538) 0.096* (2.310) 0.502** (6.562) 0.025* (2.439) (-0.720) (-1.275) * (-2.480) (0.001) (-0.645) 0.092* (2.234) ** (-5.830) (0.354) (0.001) 0.040* (1.811) (0.046) * (-1.697) ** (-5.830) (-0.229) (-0.645) (0.046) (0.109) (-0.891) ** (-5.830) Log lkelhood AIC Noe: * and ** denoe sgnfcance a he 5% and 1% levels, respecvely. The fgures n parenheses are z-sascs. The one-aled es apples o he esmaes of he ln(x/p) varable, and he wo-aled es apples o he remander. Model Esmaon Followng he descrpve analyss, economerc analyss s conduced based on he TVP-EC-AIDS model. Several unresrced and heorecally resrced versons of he model are esmaed, and he homogeney and symmery resrcons are esed

12 separaely and jonly usng a sample sze-correced sasc developed by Cour (1968) and Deaon (1974). The resuls for he homogeney, symmery based on homogeney, and symmery and homogeney jon ess show ha boh homogeney and symmery resrcons canno be rejeced a he 1% sgnfcance level, wh wo excepons only. The small sample sze s lkely a reason for hese excepons, as dscussed n Wu e al. (2011). To be conssen across all cases and n lne wh demand heory, boh resrcons are mposed on all models for esmaon, followng prevous sudes (e.g., Corés-Jménez e al., 2009; Wu e al., 2011). Thus, he calculaon of he demand elasces and he subsequen analyss are based on he homogeney and symmery resrced TVP-EC-AIDS model. Table 1 shows he fnal sae esmaes of he TVP-EC-AIDS model for he four source markes. All four models yeld sasfacory resuls. In parcular, all of he esmaes of he error correcon erms fall beween -1 and 0, whch s n lne wh he requremen for an error correcon model, and over half of hem are sascally sgnfcan. Based on hese esmaes, varous demand elasces are calculaed and her evoluon s analysed. To ncrease he robusness of he fndngs, only sascally sgnfcan elasces are dscussed (De Mello, Pack & Thea, 2002). Expendure Elasces Fgure 2 demonsraes he evoluon of he expendure elasces for shoppng, hoel accommodaon and meals ousde hoels n he four source markes. A clear dfference s observed beween he shoppng behavour of manland Chnese ourss and ha of ourss from he oher hree markes. For he former, he expendure elascy of demand for shoppng remans hghly nelasc (fallng beween 0 and 1) over he sample perod, suggesng ha manland Chnese ourss conssenly rea shoppng n Hong Kong as a necessy. In conras, he oher ourss regard shoppng as a luxury, as he expendure elasces of hese hree markes are all greaer han 1. I s hus concluded ha Hong Kong has a comparave advanage as a shoppng desnaon for manland Chnese ourss. Wh regard o he evoluon of expendure elasces, a common rend observed across all of he markes s ha he demand for shoppng has become ncreasngly sensve o changes n he expendure budges of ourss. The calculaed correlaon coeffcens beween hese elascy seres are as hgh as Ths rend s mos evden n he US marke, for whch he elascy value rses from 1.20 n 1996 o 1.88 n 2008, wh an annual growh rae of 3.79%. Ths suggess ha US ourss ncreasngly perceve shoppng n Hong Kong as a luxury, wh he mos recen expendure elascy value beng he hghes amongs all of he source markes under consderaon. The growh of he expendure elascy of demand for shoppng s he seades n he manland Chnese marke and he leas sable n he US marke. These fndngs sugges ha durng economc downurns, such as he one ha has lased for he pas hree years, all of he key source markes for Hong Kong nbound oursm end o cu her spendng on shoppng. The Hong Kong real secor s lkely o suffer mos from he reduced expendure of he US marke and leas from ha of he 12

13 manland Chnese marke. The corollary of hs s ha durng boom perods, he Hong Kong real secor should parcularly arge he US marke. Fgure 2. Evoluon of expendure elasces. Wh respec o he expendure elasces of demand for hoel accommodaon, he calculaed resuls are sascally non-sgnfcan n he frs few years of he sample for he US and Japanese markes, and hus are omed from he analyss. Fgure 2 shows ha manland Chnese ourss always rea hoel accommodaon n Hong Kong as a luxury (.e., he elascy value s greaer han 1), whereas he oher hree markes vew as a necessy (.e., he elascy value s less han 1). In erms of he evoluon of he expendure elasces, oppose rends can be observed beween he par of manland Chna and Tawan on he one hand, and he par of Japan and he Uned Saes on he oher hand. The rend of he former par of markes s a general decrease n expendure elasces, whereas ha of he laer par shows a gradual ncrease. The expendure elasces n he manland Chnese marke decrease from 2.40 n 2002 o 1.70 n 2008, equvalen o an annual declne of 5.55%, whch ndcaes ha hoel accommodaon s vewed less and less as a luxury by manland Chnese ourss. Ths can be explaned by he rapd growh n personal dsposable ncome of resdens of manland Chna and he concoman apprecaon of he renmnb n recen years. Ths rend s lkely o connue, and a smlar level of expendure elascy of demand for hoel accommodaon (.e., close o bu less han 1) may emerge amongs all four markes n he fuure. As for meals ousde hoels, he resuls are sascally sgnfcan only for he manland 13

14 Chnese and Japanese markes. Despe he rapd growh of he Chnese economy and ncrease n personal ncome levels, manland Chnese ourss sll regard meals ousde hoels as a luxury, alhough less so snce Ths s probably because food prces n manland Chna are sll low compared wh hose n Hong Kong. However, nflaon has been relavely hgh n recen years and he prce gap s narrowng; hus, manland Chnese ourss do no perceve dnng ou n Hong Kong o be as much of a luxury as hey dd before. The elascy values decreased by 2.79% annually beween 2004 and For ourss from Japan, a maure marke, he expendure elascy of demand for meals ousde hoels s very sable across he sample perod, wh an average annual declne of only 0.58%, and he values are above, alhough close o, 1. Own-Prce Elasces The sgnfcan own-prce elasces of demand for he hree produc caegores n he four source markes are ploed n Fgure 3. I can be observed ha he own-prce elasces of demand for shoppng n Hong Kong have remaned very sable over me for manland Chnese and Japanese ourss n recen years, wh an average growh of only 1.26% and 0.73%, respecvely, ndcang ha he consumpon behavour of hese ourss has no changed much n response o prce changes n he real secor n Hong Kong. Tawanese ourss appear o be much more sensve o real prce changes compared wh manland Chnese and Japanese ourss. The own-prce elasces for shoppng are and for Tawan and manland Chna, respecvely, n 2008 and for Japan n Fgure 3. Evoluon of own-prce elasces. The own-prce elasces of demand for hoel accommodaon are relavely sable, decreasng slghly (n erms of absolue value, wh an annual declne beween 2.09% 14

15 and 1.20%) over me across all source markes. Ths means ha he demand for hoel accommodaon n hese markes s decreasngly sensve o he varaon n hoel prces n Hong Kong. The demand for accommodaon of US ourss s he mos prce nelasc (wh an elascy value of n 2008) compared wh ha of he hree shor-haul markes. Japanese and Tawanese ourss dsplay very smlar consumpon behavour owards hoel accommodaon, wh smlar elascy evoluon paerns and conssen levels of prce sensvy (wh elascy values n 2008 of and -0.60, respecvely). In comparson, manland Chnese ourss are more sensve o prce changes n accommodaon n Hong Kong, and her demand for accommodaon s prce elasc, wh values rangng beween and over he perod from 2002 o 2008, equvalen o an average annual declne of 1.20% n erms of absolue values. Posve own-prce elasces of he demand for meals ousde hoels are denfed n he Japanese marke, bu he value decreases sharply over he sudy perod, declnng by 3.64% per year on average. The posve own-prce elasces can be explaned by he fac ha Japanese ourss are more neresed n Hong Kong cusne, evdenced by he relavely hgh proporon of her expendure on hs caegory of goods/servces (16.0% n 2008) compared o ha of he oher Asan markes, and hs expendure share has connued o ncrease over recen years. Heung, Qu and Chu (2001) also provde emprcal evdence ha Japanese ourss rae food as he second mos mporan move among 25 for vsng Hong Kong. In addon, meals n Hong Kong are relavely cheap compared o hose n Japan. Thus, despe he ncrease n he prce of such meals, Japanese ourss spen more on food n Hong Kong. The own-prce elasces of meals ousde hoels for he oher hree source markes are no dscussed due o her sascal non-sgnfcance. Cross-Prce Elasces As shown n Fgure 4, hree pars of subsuon relaonshps are denfed based on he calculaed posve cross-prce elasces: shoppng and hoel accommodaon n he manland Chnese marke, accommodaon and meals ousde hoels n he US marke and shoppng and accommodaon n he Tawanese marke. I s noed ha he subsuon effecs beween he wo produc caegores concerned dffer n each par. For example, he demand of manland Chnese ourss for hoel accommodaon s much more sensve o prce varaons n he shoppng caegory han he oher way round (wh cross-prce elasces of 1.04 and 0.17, respecvely, n 2008). In oher words, hoel accommodaon s more subsuable for shoppng han shoppng s for hoel accommodaon amongs hese ourss. Ths fndng furher confrms he mporance of shoppng for manland Chnese ourss. The evoluon of he cross-prce elasces n Fgure 4 ndcaes ha hs consumpon behavour has remaned relavely sable n recen years. The average annual growh raes are lower han one percen over he sample perod n mos cases. 15

16 Fgure 4. Evoluon of cross-prce elasces. CONCLUSION Tradonal consan parameer models can reveal average consumpon paerns over he me span under consderaon bu canno capure changes n consumpon paerns over a specfc perod. To overcome hs lmaon, he presen sudy combnes he TVP esmaon echnque wh a sysem-of-equaons demand model o examne he evoluon of ourss consumpon behavour paerns over me. Ths sudy enrches he leraure by employng he TVP-EC-AIDS model for he frs me o analyse how ourss allocae her budge amongs dfferen oursm consumpon caegores. Based on he esmaon resuls generaed by he developed model, he expendure, own-prce and cross-prce elasces of he demand for shoppng, hoel accommodaon and meals ousde hoels n Hong Kong are examned from a dynamc perspecve. The emprcal resuls show dverse expendure paerns amongs ourss from four source markes: manland Chna, Japan, Tawan, and he Uned Saes. The fndngs provde new nsghs no he undersandng of he dynamcs of ourss consumpon behavour, and have some useful mplcaons ha may asss he relevan sakeholders n boh he publc and prvae secors n Hong Kong o make 16

17 sraegc decsons regardng oursm plannng and managemen. Ths sudy fnds ha he demand for shoppng by manland Chnese ourss s nsensve o her expendure budge (or personal ncome), bu has become relavely more sensve n recen years. As Hong Kong s perceved o be a shoppng paradse by manland Chnese ourss, he real secor has benefed grealy from hs source marke n recen years. I s hus mporan ha boh he busness secor and he governmen make an acve effor o susan he aracveness of Hong Kong s real busnesses. For real shops ha arge manland Chnese ourss as her man cusomers, he delvery of hgh-qualy goods and servces s val. Invesmen o mprove he shoppng envronmen and esablsh a far and rusng relaonshp beween cusomers and real shops would furher boos he demand for real producs by manland Chnese ourss. Realers are lkely o be grealy affeced by economc crses, as real sales from oursm are closely relaed o he economc condons n source markes. Thus, durng economc downurns, s mporan for he real ndusry o adop proper markeng and promoonal sraeges o arac ourss, parcularly hose from markes wh low expendure elascy values. For example, long-haul markes for Hong Kong oursm, such as he Uned Saes, have suffered grealy from he recen economc recesson, whch has negavely affeced he expendure of nbound ourss from such markes. Tours arrvals from manland Chna, however, have connued o grow over he pas wo years. I s hus crucal for he real secor n Hong Kong o arge manland Chnese ourss n her promoonal and markeng effors. The Hong Kong governmen has already played a par by launchng several successful naves over he pas wo years o encourage shoppng oursm from manland Chna, such as he exenson of he ndvdual vs scheme o non-guangdong resdens who lve n Shenzhen. Hoel accommodaon s vewed by manland Chnese ourss as a luxury, alhough less so n recen years. Ths may be due o gradual ncreases n personal ncome and he apprecaon of he Chnese renmnb. Accordng o he prce elascy esmaes of he hoel ndusry, prcng sraeges are more effecve n aracng ourss from manland Chna han hose from he oher hree source markes under consderaon. However, he mplemenaon of prce dscrmnaon should be appled wh cauon, as applyng dfferen prces o dfferen source markes can be perceved negavely. A beer sraegy would be o mprove servce qualy wh a vew o ncreasng he level of ours sasfacon, whch ends o be more effecve han prcng sraeges n aracng ourss. Ths sudy also fnds ha meals ousde hoels n Hong Kong are sll perceved by manland Chnese ourss o be que expensve compared wh meal prces n manland Chna. To manan he compeve advanages of Hong Kong as a popular nernaonal ours desnaon, he governmen needs o formulae a broader range of effecve 17

18 oursm polces and connue o nves n publc facles for ourss. Gven he sharp ncrease n he number of ourss from manland Chna, new mmgraon pors for manland Chnese ourss should be nroduced. Addonal effors should also be made o smplfy he enry procedure and o furher exend he ndvdual vs scheme o oher ces n Chna. Dsnc from ourss from oher source markes, who usually vs Hong Kong by ar, a large proporon of manland Chnese ourss ravel o Hong Kong by ran and bus. The Hong Kong governmen should hus make furher nvesmens n land ransporaon o dversfy he means of ranspor, shoren he ravel me and reduce he ransporaon coss of manland Chnese ourss. The fndngs ndcae ha US ourss ncreasngly vew shoppng n Hong Kong as a luxury. Ths mples ha he compeve advanage of he real secor n Hong Kong has weakened wh respec o he US marke. Clearly, he decrease n he dsposable ncome of US ourss because of he recen fnancal crss has led o a reducon n her expendure on shoppng n Hong Kong. The fndngs furher reveal ha Tawanese ourss also vew shoppng more as a luxury han hey dd before. The expendure elascy n he Tawanese hoel model declnes over he sample perod, whereas ha of he Japanese marke ncreases. Overall, he fndngs of hs sudy sugges ha dfferen source markes for Hong Kong oursm dsplay dfferen ypes of consumpon behavour and dfferen evoluon paerns. Ths sudy provdes emprcal confrmaon of he argumen n he leraure ha oursm consumpon s socally nfluenced and culurally framed (Sharpley, 1999). Toursm praconers n Hong Kong need o closely monor he dynamcs of he consumpon behavour of ourss from dfferen markes, and adop flexble, marke-orened promoonal plans and prcng sraeges. I s also mporan ha hey revew busness sraeges regularly and make he necessary adjusmens n response o he dynamc naure of ourss consumpon behavour. Acknowledgemen The auhors would lke o acknowledge he fnancal suppor of he Hong Kong Research Grans Commee (Gran No. PolyU 5003-PPR-6) and he Naonal Naural Scence Foundaon of Chna (Gran No ). REFERENCES Corés-Jménez, I., Durbarry, R., & Pulna, M. (2009). Esmaon of oubound Ialan oursm demand: A monhly dynamc EC-LAIDS model. Toursm Economcs, 15, Cour, R. H. (1968). An applcaon of demand heory o projecng New Zealand real consumpon. Economc Revew, De Mello, M., Pack, A., & Thea, M. (2002). A sysem of equaons model of UK oursm demand n neghbourng counres. Appled Economcs, 34, Deaon, A. S. (1974). The analyss of consumer demand n he Uned Kngdom, 18

19 Economerca, 42, Deaon, A., & Muellbauer, J. (1980). An almos deal demand sysem. Amercan Economc Revew, 70, Dvsekera, S. (2009). Ex pos demand for Ausralan oursm goods and servces. Toursm Economcs, 15, Durbarry, R., & Snclar, M. T. (2003). Marke shares analyss: The case of French oursm demand. Annals of Toursm Research, 30, Durbn, J., & Koopman, S. J. (2001). Tme Seres Analyss by Sae Space Mehods. Oxford: Oxford Unversy Press. Fuj, E. T., Khaled, M., & Mak, J. (1985). An almos deal demand sysem for vsor expendures. Journal of Transpor Economcs and Polcy, 19, Harvey, A. C. (1989). Forecasng, Srucural Tme Seres Models and he Kalman Fler. Cambrdge: Cambrdge Unversy Press. Vncen C.S. Heung, V. C. S., Qu, H., & Chu, R. (2001). The relaonshp beween vacaon facors and soco-demographc and ravellng characerscs: The case of Japanese lesure ravellers. Toursm Managemen, 22, Hong Kong Toursm Board (HKTB). (2009). A Sascal Revew of Hong Kong Toursm Hong Kong: Hong Kong Toursm Board. Kalman, R. E. (1960). A new approach o lnear flerng and predcon problems. Transacons of he ASME-Journal of Basc Engneerng, 82, L, G., Song, H., & W, S. F. (2004). Modelng oursm demand: A dynamc lnear AIDS approach. Journal of Travel Research, 43, L, G., Song, H., & W, S. F. (2005). Recen developmen n economerc modelng and forecasng. Journal of Travel Research, 44, L, G., Song, H., & W, S. F. (2006). Tme varyng parameer and fxed parameer lnear AIDS: An applcaon o oursm demand forecasng. Inernaonal Journal of Forecasng, 22, L, G., Wong, K. K. F., Song, H., & W, S. F. (2006). Toursm demand forecasng: A me varyng parameer error correcon model. Journal of Travel Research, 45, Rddngon, G. (1999). Forecasng sk demand: Comparng learnng curve and me varyng parameer approaches. Journal of Forecasng, 18, Sharpley, R. (1999). Toursm, Tourss and Socey, second edon. Cambrdge: ELM Publcaons. Song, H., & L, G. (2008). Toursm demand modellng and forecasng: A revew of recen research. Toursm Managemen, 29, Song, H., & Wong, K. (2003). Toursm demand modelng: A me-varyng parameer approach. Journal of Travel Research, 42, Song, H., W, S. F., & Jensen T. C. (2003). Toursm forecasng: Accuracy of alernave economerc models. Inernaonal Journal of Forecasng, 19, Sabler, M. J., Papaheodorou, A., & Snclar, M. T. (2010). The Economcs of Toursm, second edon. London: Rouledge. Thomas, R. L. (1993). Inroducory Economercs: Theory and Applcaons, second 19

20 edon. London: Longman. Tornqvs, L. (1936). The bank of Fnland s consumpon prce ndex. Bank of Fnland Monhly Bullen, 10, 1-8. Whe, K. J. (1985). An nernaonal ravel demand model: US ravel o Wesern Europe. Annals of Toursm Research, 12, World Toursm Organzaon (UNWTO). (2009). Toursm Hghlghs (2009 edon). Madrd: World Toursm Organzaon. Wu, D. C., L, G., & Song, H. (2011). Analysng oursm consumpon: A dynamc sysem of equaons approach. Journal of Travel Research, 50,