tourism demand modelling and forecasting: how …epubs.surrey.ac.uk/7594/69/s5.pdftourism demand...

Tourism Economics 2010 16 (1) 63ndash81

Tourism demand modelling and forecastinghow should demand be measured

HAIYAN SONG

School of Hotel and Tourism Management The Hong Kong Polytechnic UniversityHung Hom Kowloon Hong Kong SAR PR China E-mail hmsongpolyueduhk

GANG LI

Faculty of Management and Law University of Surrey Guildford GU2 7XH UK

STEPHEN F WITT AND BAOGANG FEI

School of Hotel and Tourism Management The Hong Kong Polytechnic UniversityHung Hom Kowloon Hong Kong SAR PR China

Tourist arrivals and tourist expenditure in both aggregate and percapita forms are commonly used measures of tourism demand inempirical research This study compares these two measures in thecontext of econometric modelling and the forecasting of tourismdemand The empirical study focuses on demand for Hong Kongtourism by residents of Australia the UK and the USA Using thegeneral-to-specific modelling approach key determinants of tourismdemand are identified based on different demand measures Inaddition the forecasting accuracy of these demand measures isexamined It is found that tourist arrivals in Hong Kong areinfluenced mainly by touristsrsquo income and lsquoword-of-mouthrsquohabitpersistence effects while the tourism price in Hong Kong relative tothat of the tourist origin country is the most important determinantof tourist expenditure in Hong Kong Moreover the aggregatetourism demand models outperform the per capita models withaggregate expenditure models being the most accurate Theimplications of these findings for tourism decision making are thatthe choice of demand measure for forecasting models should dependon whether the objective of the decision maker is to maximize touristarrivals or expenditure (receipts) and also that the models should bespecified in aggregate form

Keywords tourist arrivals tourist expenditure forecasting accuracyHong Kong

Tourism demand is usually regarded as a measure of visitorsrsquo use of a good orservice (Frechtling 2001) The concept of tourism demand originated from theclassical definition of demand in economics namely the desire to possess a

TOURISM ECONOMICS64

commodity or to make use of a service combined with the ability to purchaseit lsquoThe significance level and repercussions of tourism demand provide a strongcase for better understanding of the nature of the touristsrsquo decision-makingprocessrsquo (Sinclair and Stabler 1997 p 15) Tourism demand is a special formof demand in that a tourism product is a bundle of complementary goods andservices (Morley 1992) Consumers instead of goods and services aretransported and tourism consumption occurs simultaneously with tourismproduction (Schulmeister 1979)

Tourism demand can be measured in a variety of ways Kim (1988 p 25)categorized the measurement criteria for all types of travel and tourism demandinto four groups (i) a doer criterion such as the number of tourist arrivalsthe number of tourist visits and the visit rate (ii) a pecuniary criterion forexample the level of tourist expenditure (receipts) and share of expenditure(receipts) in income (iii) a time-consumed criterion such as tourist-daystourist-nights and (iv) a distance-travelled criterion for instance the distancetravelled in miles or kilometres Among the above four categories the doercriterion and pecuniary criterion dominate international tourism demandstudies Considering statistical availability and consistency between data sourcestourist arrivals (TA) and tourist expenditure (TE) (receipts) are the mostcommonly used tourism demand measures in empirical studies along with theirderivatives such as the tourist participation rate derived from tourist arrivalsdivided by population of the origin countryregion (TA_P) and touristexpenditure per capita derived from total tourist expenditure divided bypopulation (TE_P) Other variations of tourism demand measures can also beidentified in empirical studies Based on published studies over the periods from1961 to the early 1990s Crouch (1994) and Lim (1997) summarized tourismdemand measures into five and six broad categories respectively In Crouchrsquos(1994) review 19 out of 80 published studies used more than one tourismdemand measure This phenomenon was also observed by Lim (1997) Li et al(2005) reviewed the literature on tourism demand modelling and forecastingusing econometric approaches published over the period 1990ndash2004 and foundthat 20 measures of tourism demand were used in those studies In addition18 out of 84 studies included in the review used more than one demandmeasure Based on Crouchrsquos categorization Table 1 summarizes the applicationsof various tourism demand measures in the studies reviewed by Crouch (1994)Lim (1997) and Li et al (2005) respectively It can be seen that the tourismdemand measures TA (and TA_P) and TE (and TE_P) have dominated tourismdemand modelling and forecasting studies over the past four decades Thefollowing discussion will therefore focus on these two sets of measures oftourism demand

Tourist arrivals versus tourist expenditure

Although both tourist arrivals and tourist expenditure are commonly usedmeasures in tourism demand modelling and forecasting the differences betweenthe two should not be ignored

First the fundamental difference between these two measures is the way inwhich these two types of data are collected International tourist arrivals are

65Tourism demand modelling and forecasting

Table 1 Tourism demand measures identified in previous review studies

Crouch (1994) Lim (1997) Li et al (2005)

Tourist arrivals (departures) 51 51 53Tourist expenditure (receipts) 40 49 24Length of stay 3 6 0Nights spent at tourist accommodation 6 4 1Others 5 9 10Total studies reviewed 80 100 84Periods of publications under review 1961ndash1992 1961ndash1994 1990ndash2004

often recorded by frontier counts while tourist expenditure data are normallycollected through visitor surveys (Witt and Witt 1995) Due to the differencesin data compilation the trends and patterns of tourism demand reflected inthe time series can be different and this issue will be discussed below

Secondly the two demand measures serve different purposes Tourismproductservice suppliers are more interested in tourist volumes because theyhave direct impacts on their supply capacity For example the decisions oninvestment in new hotels and new aircraft rely largely on accurate forecasts oftourist arrivals (Sheldon 1993 p 18) However the tourist volume measurecannot meet the forecasting needs of economic planners because it does not takeaccount of the economic impact of tourism on the related sectorsactivitiesTourist expenditure (that is the receipts of the destination) is the main concernof governments and central banks However as a foundation on which theeconomic impact of tourism activities is assessed they very often do not provideadequate information and usually suffer from biases due to problems in the datacollection process as well as linkages and leakages in the economy (Frechtling1987)

Thirdly when the evolution of the data series over time is considereddifferent patterns in the arrivals and expenditures series emerge for mostdestinationorigin country pairs Sheldon (1993) studied the percentage changesin international tourist expenditure and arrivals for 15 OECD countries andconcluded that international tourist arrivals fluctuated to different degrees thandid international tourist expenditures This study further explained that thedifferent fluctuation patterns were likely to do with the length of tourist stayand fluctuations in exchange rates as well as other economic factors such asincome in the tourist origin country and relative tourism prices HoweverSheldon (1993) only investigated the accuracy of tourist expenditure forecastsgenerated by various models and did not compare the forecasting accuracy ofthe demand models using different measures of tourism demand Garciacutea-Ferrerand Queralt (1997) noticed the different annual growth rates between touristreceipts and arrivals in Spain and speculated that these differences might bea result of distinctive tourist behaviours and raised the question about thechoice of the relevant variable to measure tourism demand They also suspectedthat the explanatory variables employed added little to the explanatory powerof the evolution of international tourism demand But they did not explore thisissue further in their econometric analyses by examining the differences in the

TOURISM ECONOMICS66

key determinants of the two demand measures Moreover they reported theforecasts of tourist expenditure only but not those of tourist arrivals Hencethere was no comparison of forecasting accuracy between the two measures Qiuand Zhang (1995) attempted to make an empirical comparison between touristarrivals and tourist expenditure in tourism demand analysis but their focus wasonly on the choice of proper functional form (linear or log-linear) The estimatedmodels were the general autoregressive distributed lag models (ADLMs) whichincluded both statistically significant and insignificant variables Thus it is notpossible to compare the different determinants of the two demand variablesAdditionally the forecasting accuracy of the alternative models with differenttourism demand measures was not examined The applications of ADLMs torecent tourism demand modelling and forecasting studies were summarized bySong and Li (2008)

Fourthly with regard to tourism demand forecasting the literature hasfocused mainly on comparing the forecasting performance of alternative modelsbased on the same tourism demand measure No attempt has been made toexamine the forecasting accuracy of different demand measures systematically

In light of the above gaps in the literature it is necessary to investigatecomprehensively the underlying factors that contribute to the different demandpatterns measured by the two variables and the impacts of the selection of thedemand measure on the accuracy of tourism demand forecasts This study aimsto achieve this objective and a particular focus is to examine whether thedifferent trendspatterns in tourism demand are in fact caused by differentfactors using the econometric approach Both dependent variables in total andper capita forms are considered and the general-to-specific modelling approachis used to decide which factors influence the demand for tourism as definedby the two types of measure Furthermore this study examines the impact ofusing different measures in the demand models on the forecasting performanceof these models

Demand for Hong Kong tourism

Achieving an average annual growth rate of 111 which was higher than mostof the destinations in the Asia Pacific region during the same period inter-national visitor arrivals in Hong Kong increased from 093 million in 1975 to2525 million in 2006 Hong Kong was ranked 16th in the World TourismOrganizationrsquos (UNWTO) list of top destinations in 2006 in terms of thenumber of visitor arrivals The growth rate of visitor arrivals in Hong Kongin 2006 was 81 while the world and regional average growth rates in thesame year were 45 and 76 respectively Meanwhile the growth of tourismreceipts rose from HK$2975 million in 1975 to HK$89398 million in 2006giving an annual growth rate of 116 and 58 in nominal and real termsrespectively Tourism receipts grew rapidly during the 1970s and 1980s inparticular and this trend continued until the mid-1990s Tourism has becomethe second largest foreign currency earner since 1995 and the income generatedfrom tourism has contributed around 6 to Hong Kongrsquos gross domesticproduct (GDP) over the past decade (Zhang et al 2001) The Asian financialcrisis caused a significant decline in tourism receipts in 1997 and 1998 with


a huge drop in visitor numbers from the affected countries Meanwhile pricereduction also played a significant role in the reduction of tourism revenue

The demand for Hong Kong tourism by tourists from three long-haulmarkets ndash Australia the UK and the USA ndash is chosen as the focus of thisresearch because they are relatively mature tourism source markets for HongKong and the results produced should be more representative Figures 1ndash4 showtotal tourist arrivals total tourist expenditure (in real terms) tourist arrivalsper capita and real tourist expenditure per capita respectively from the threesource markets Figures 1 and 3 show that tourist arrivals from all the threemarkets in both aggregate and per capita terms have grown steadily duringthe period 1981ndash2006 On the contrary Figures 2 and 4 suggest that realtourist expenditure decreased in the 1990s across all these markets and mostsignificantly in the Australian market This is most likely due to the globaleconomic downturns in many industrial nations over this period It was alsoobserved that some mega events such as the outbreak of the Severe AcuteRespiratory Syndrome (SARS) in 2003 influenced tourism demand to HongKong dramatically With regard to the growth rates of both visitor arrivals andtheir expenditures in Hong Kong from the three main source markets Figures5ndash7 suggest that their evolution has differed substantially For instance thedemand for Hong Kong tourism by Australian residents experienced remarkablegrowth in 1984 against the previous year with an annual growth rate of 219in terms of tourist arrivals However the demand grew only modestly with anannual rate of 17 in terms of tourist expenditure over the same period (seeFigure 5) More interestingly arrivals grew year by year over the period 1992ndash1994 while the opposite trend was observed with tourist expenditure Overallthere were 18 out of 24 cases where the gaps between the growth rates of touristarrivals and expenditures were greater than 5 percentage points As far as theUK and the USA were concerned the cases were 16 and 15 out of 24respectively With respect to the direction of the demand changes oppositesigns (positivenegative) were found in 7 6 and 8 out of 24 cases for Australiathe UK and the USA respectively

The model

Determinants of tourism demand

Tourism demand could be affected by a wide range of factors such as economicattitudinal and political factors but the majority of the econometric studiestend to examine the demand for tourism by focusing predominantly oneconomic factors Income and prices play important roles in determiningtourism demand Crouch (1994) reveals that income is the most importantexplanatory variable and the income elasticity generally exceeds unity but isbelow two which implies that international travel is still regarded as luxuryconsumption Economic theory also indicates that the price of tourism productsservices is related negatively to tourism demand The price variables shouldinclude the prices of the goods and services related to both the destination andsubstitute destinations Additionally marketing expenditure consumer tastesconsumer expectations habit persistence origin population and one-off events

TOURISM ECONOMICS68

Figure 1 Tourist arrivals in Hong Kong from Australia the UK and the USA(1981ndash2006)

Figure 2 Real tourist expenditure in Hong Kong by visitors from Australiathe UK and the USA (1981ndash2006)

0

200000

400000

600000

800000

1000000

1200000

82 84 86 88 90 92 94 96 98 00 02 04 06

Australia UK USA


Figure 3 Tourist arrivals per capita in Hong Kong from Australia the UKand the USA (1981ndash2006)

Figure 4 Real tourist expenditure per capita in Hong Kong by visitors fromAustralia the UK and the USA (1981ndash2006)

TOURISM ECONOMICS70

Figure 5 Annual growth rates () of tourist arrivals and tourist expenditurein Hong Kong by visitors from Australia (1982ndash2006)

are all potentially important factors that could be incorporated into the tourismdemand model (Song and Witt 2000) Witt and Witt (1995) suggested thatthe time trend could represent changing touristsrsquo tastes The lagged dependentvariable describes touristsrsquo expectations habit persistence the lsquoword-of-mouthrsquoeffect and supply constraints Lagged explanatory variables are also oftenincluded in demand models to capture the dynamic effects of variousinfluencing factors on tourism demand (Lim 1997) Fewer studies haveincluded the marketing expenditure variable in their empirical analyses becauseof the unavailability of marketing expenditure data even though promotion isgenerally recognized as an important strategy to enhance the awareness of adestination and to highlight its attractiveness so as to increase the number oftourists and tourism receipts Therefore the most commonly consideredinfluencing factors for tourism demand in econometric models are originincome the own price of a destination substitute prices of alternativedestinations and dummy variables to capture the effects of one-off events Thesevariables are also considered in the current study

Real GDP instead of household disposable income is chosen in this studyto measure the income level of an origin country The reason for this is becauseof the high proportion of business visitors in overall inbound tourist arrivalsin Hong Kong Tourism price and exchange rates are jointly taken into account

82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40

Tourist arrivals Tourist expenditure


Figure 6 Annual growth rates () of tourist arrivals and tourist expenditurein Hong Kong by visitors from the UK (1982ndash2006)

in calculating the two relative price variables in this study that is the ownprice and substitute prices With regard to the own price of tourism in HongKong it is defined as

CPIHKtEXHKtPit = ndashndashndashndashndashndashndashndashndashndashndash CPIitEXit

where CPIHKt and CPIit are the consumer price indices (CPIs) for Hong Kongand origin country i respectively EXHKt and EXit are the exchange ratesbetween the Hong Kong dollar and the US dollar and between the currencyof origin country i and the US dollar respectively The exchange rates aremeasured by the annual average market rates of the local currency against theUS dollar The own-price variable defined above is a lsquorelativersquo or lsquoeffective priceof tourismrsquo (Durbarry and Sinclair 2005 p 981) It reflects the costs of tourismactivities in Hong Kong relative to those in the touristrsquos origin country Itembodies a plausible decision-making process by a tourist that is a decisionbetween domestic and international tourism In other words domestic tourismis regarded as a substitute for international tourism or at least used as abenchmark when a tourist plans his or her international travel

After taking account of both geographic and cultural characteristics

ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72

Figure 7 Annual growth rates () of tourist arrivals and tourist expenditurein Hong Kong by visitors from the USA (1982ndash2006)

mainland China Singapore South Korea Taiwan and Thailand are selected tobe the substitute destinations for Hong Kong in this study Following Songet al (2003) and Gallet and Braun (2001) the substitute price variable is definedas a weighted average index of the selected countriesrsquoregionsrsquo tourism pricesnamely

CPIjtPst = 5Σj=1

ndashndashndashndash wijt EXjt

where j = 1 2 3 4 and 5 represent the five substitute destinations respectivelyand wijt is the share of international tourist arrivals to countryregion j and iscalculated from

wijt = TAijt5Σj=1

TAijt

where TAijt is the inbound tourist arrivals to substitute destination j from theorigin countryregion i at time t The above formula indicates that the weightsfor calculating the substitute price variable change over time in order to reflectthe dynamics of the substitution effect It should be noted that to ensure thecomparability of the model estimations using the two sets of demand measuresthe same definition of the substitute price variable is adopted across all models

ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06



with different measures of tourism demand In other words the weights for thesubstitute price calculation are always tourist arrivals regardless of thedependent variable Although the patterns of historical evolutions of touristarrivals and tourist expenditure are different as far as the same destination isconcerned as discussed above the proportions of market shares among a fewdestinations measured by tourist arrivals are highly consistent with thosemeasured by tourist expenditure at each point of time Therefore the choiceof the weighing scheme does not have a significant effect on the variations ofthe calculated substitute price variable Real GDP CPI and exchange rates heretake the forms of indices and their values in 2000 are specified as 100 Suchrescaling does not affect the variables concerned and has no effect on theestimated coefficients of these variables in a log-linear model

The data

The Hong Kong Tourism Board (HKTB) publishes various up-to-date keytourism data Monthly or cumulative visitor arrivals are published by countryterritory of residence and by the mode of transport Visitor profiles along withtheir expenditures are also surveyed every year Strictly speaking internationalvisitors consist of international tourists and international same-day visitors asdefined by UNWTO The actual arrivals data used in this study areinternational visitor arrivals by country of residence provided by the HKTBTourist expenditure (receipts) is defined by UNWTO1 as expenditure ofoutbound (inbound) visitors plus their payments to foreign (national) carriersfor international transport In this research however data on the inboundovernight tourist expenditure in Hong Kong are used The expenditure of same-day visitors and all the payments made to carriers registered in Hong Kongare excluded due to incomplete statistics of same-day visitor expenditure fromspecific countries

The real GDP CPI and exchange rate data were collected from theInternational Financial Statistics Yearbook published by the International MonetaryFund (IMF) Where per capita demand was concerned population data wereneeded and these were collected from the IMF publications The sample periodof this empirical study is from 1981 to 2006 and the observations of tourismdemand are at the annual frequency

Model specification

In this study the following demand function is proposed to model the demandfor tourism in Hong Kong by residents from each of the three origin countries

TDit = AYitβ2Pit

β3Pstβ4eit (1)

where TDit is the proxy of tourism demand that is TA TA_P TE and TE_Pin Hong Kong from origin country i at time t Yit is the income level (or incomeper capita Y_Pit when TA_P and TE_P are concerned) of origin country iPit is the own price of tourism in Hong Kong at time t Pst is the substituteprice of tourism at time t and eit is the error term which is used to capturethe influence of all other factors that are not included in the demand model

According to Song and Witt (2000) one major feature of the power function

TOURISM ECONOMICS74

[Equation (1)] is that it can be transformed into a log-linear specification whichcan be estimated easily using the ordinary least squares (OLS) method Aftertaking logarithms of Equation (1) we have

lnTDit = β1 + β2 lnYit + β3 lnPit + β4 lnPst + εit (2)

where β1 = lnA εit = lneit and β2 β3 and β4 are income own-price and cross-price elasticities respectively To capture the influence of various one-off eventson the demand for tourism in Hong Kong a number of dummy variables areincluded in the above model These dummy variables include D97 (D97 = 1 in1997 and 0 otherwise) to capture the effect of the Asian financial crisis in 1997D911 (D911 = 1 in 2001 and 0 otherwise) to capture the effect of the September11 terrorist attack on the USA in 2001 DSARS (DSARS = 1 in 2003 and 0otherwise) to reflect the influence of SARS in Hong Kong in 2003 and DFLU

(DFLU = 1 in 2004 and 0 otherwise) to reflect the impact of avian fluThe above equation is a static model and does not take into account the

dynamics of the touristrsquos decision-making process In order to capture thedynamics of tourism demand the general specification of the ADLM is usedin this study The exact lag length of either 1 or 2 in each general ADLM isdetermined by the Akaike information criterion (AIC) and the Schwarz criterion(SC) The general ADLM initially takes the following form

lnTDit = α1 + 2Σj=1

γj lnTDitndashj + 2Σj=0

λj lnYitndashj + 2Σj=0

θj lnPitndashj

+ 2Σj=0

φj lnPstndashj + α2D97 + α3D911 + α4DSARS + α5DFLU + εit(3)

Model estimation

The general-to-specific modelling approach is applied to reduce the number ofexplanatory variables in the initial equation keeping only the underlyinginfluencing factors based on both statistical significance and the sensibleeconomic interpretation of the estimated parameters associated with thesefactors The general-to-specific approach was proposed originally by Davidsonet al (1978) and modified subsequently by Hendry and von Ungern-Sternberg(1981) and Mizon and Richard (1986) but full development of the general-to-specific modelling approach has taken place only recently (see Hendry 1995)This approach starts with a general dynamic ADLM which includes allpotentially influential factors with a sufficient lag structure By removingstatistically insignificant variables from the model one by one starting with theleast significant the general ADLM is reduced to a more specific one with allremaining variables being significant Following such a scientific procedure thekey determinants of tourism demand can be identified The final specific modelwould be most appropriate for forecasting and policy evaluation purposes

OLS is employed to estimate the models with the data series from 1981 to2004 and the data for 2005 and 2006 are reserved to evaluate forecastingaccuracy Following the above general-to-specific model reduction procedurethe statistically insignificant or incorrectly signed coefficients (that is the signsare contradictory to economic theory in Equation (3)) are removed from the


Table 2 Estimates of Australian models

TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)

lnTAitndash1 0321 lnTEitndash1 lnTA_Pitndash1 0365 lnTE_Pitndash1

(1813) (2092)lnTAitndash2 lnTEitndash2 lnTA_Pitndash2 lnTE_Pitndash2

lnYit 1784 lnYit lnY_Pit 1613 lnY_Pit

(2359) (2232)lnYitndash1 ndash1504 lnYitndash1 lnY_Pitndash1 ndash1581 lnY_Pitndash1

(ndash2043) (ndash2160)lnYitndash2 lnYitndash2 lnY_Pitndash2 lnY_Pitndash2

lnPit lnPit lnPit lnPit

lnPitndash1 lnPitndash1 ndash1514 lnPitndash1 lnPitndash1 ndash1798(ndash12203) (ndash12527)

lnPitndash2 lnPitndash2 lnPitndash2 lnPitndash2

lnPst lnPst lnPst ndash0829 lnPst

(ndash3666)lnPstndash1 lnPstndash1 lnPstndash1 lnPstndash1

lnPstndash2 lnPstndash2 lnPstndash2 lnPstndash2

D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911

DSARS ndash0262 DSARS ndash0339 DSARS ndash0260 DSARS ndash0379(ndash3927) (ndash2215) (ndash3800) (ndash2138)

DFLU ndash0294 DFLU DFLU ndash0310 DFLU

(ndash3466) (ndash3603)R2 0848 R2 0900 R2 0687 R2 0893AC ndash2554 AC ndash0905 AC ndash2505 AC ndash0613SC ndash2209 SC ndash0757 SC ndash2159 SC ndash0465CSQN 2246 CSQN 2419 CSQN 0148 CSQN 0102CSQA 0475 CSQA 0649 CSQA 1551 CSQA 1687CSQH 0638 CSQH 0620 CSQH 0723 CSQH 1173CSQAH 1998 CSQAH 2272 CSQAH 0020 CSQAH 0219CSQF 1034 CSQF 0001 CSQF 0010 CSQF 0108

Note The values in parentheses are t-statistics indicates the estimate is significant at the 10 levelAll other coefficients are significant at the 5 level CSQA Lagrange multiplier chi-square test forserial correlation CSQH the White chi-square test for heteroskedasticity CSQN the Jarque and Berachi-square test for non-normality CSQAH Engle test for autoregressive conditional heteroskedasticityCSQF the Ramsey misspecification test

model The final models that have gone through the reduction procedure areused for demand elasticity analysis and forecasting Tables 2ndash4 report theestimates of the final specific models

In general all the models fit the data well with relatively high adjusted R2sThe diagnostic statistics in the lower parts of Tables 2ndash4 show that most modelspass all five tests with only two exceptions that is the UK TA (TA_P) andTE (TE_P) models fail the CSQF test Overall given the small sample size (only23 observations for all the variables) the estimated demand models can beregarded as well specified

TOURISM ECONOMICS76

Table 3 Estimates of UK Models

TA TE TA_P TE_P

α 2125 α 3478 α ndash2913 α ndash0550(3035) (1500) (ndash1733) (ndash0584)

lnTAitndash1 0725 lnTEitndash1 0778 lnTA_Pitndash1 0718 lnTE_Pitndash1 078(6300) (7714) (6241) (7758)

lnTAitndash2 lnTEitndash2 lnTA_Pitndash2 lnTE_Pitndash2

lnYit 3107 lnYit 4600 lnY_Pit 3126 lnY_Pit 4587(2306) (2383) (2315) (2368)

lnYitndash1 ndash5896 lnYitndash1 ndash4356 lnY_Pitndash1 ndash6016 lnY_Pitndash1 ndash4328(ndash2504) (ndash2276) (ndash2520) (ndash2268)

lnYitndash2 3192 lnYitndash2 lnY_Pitndash2 3219 lnY_Pitndash2

(2388) (2413)lnPit lnPit ndash0641 lnPit lnPit ndash0638

(ndash3849) (ndash3848)lnPitndash1 lnPitndash1 lnPitndash1 lnPitndash1


lnPst lnPst lnPst lnPst



D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911




Note The values in parentheses are t-statistics All coefficients are significant at the 5 level CSQALagrange multiplier chi-square test for serial correlation CSQH the White chi-square test forheteroskedasticity CSQN the Jarque and Bera chi-square test for non-normality CSQAH Engle testfor autoregressive conditional heteroskedasticity CSQF the Ramsey misspecification test

The results in Tables 2ndash4 show that the explanatory variables in the finalTA and TA_P models are consistent suggesting that these variables havesignificant influences on the demand for tourism in Hong Kong measured bythe two arrival variables This suggests that there is little difference betweenthe determinants of total tourist arrivals and the arrivals per capita that is thetourism participation rate Similarly the same determinants were identified inthe two models when TE and TE_P were used to measure the demand for HongKong tourism


Table 4 Estimates of USA models

TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)


(5770) (5708)lnTAitndash2 ndash0535 lnTEitndash2 lnTA_Pitndash2 ndash0530 lnTE_Pitndash2

(ndash3108) (ndash3032)lnYit lnYit lnY_Pit lnY_Pit

lnYitndash1 lnYitndash1 lnY_Pitndash1 lnY_Pitndash1


(2960) (2331)lnPit lnPit 1815 lnPit lnPit 1783

(2815) (2662)lnPitndash1 lnPitndash1 ndash2996 lnPitndash1 lnPitndash1 ndash3249





D97 D97 D97 D97

D911 D911 D911 D911





When comparing the two pairs of demand measures that is TA (TA_P)versus TE (TE_P) different findings are obtained The most importantdeterminants of TA and TA_P are income and the lagged dependent variabletogether with one-off events However the own-price variable and the SARSoutbreak are the main determinants of TE and TE_P This indicates that thedifferent fluctuations of TA (TA_P) and TE (TE_P) can be explained best bydifferent influencing factors It seems that the lsquoword-of-mouthrsquohabit persistenceeffects (measured by the lagged dependent variable) and the origin income levelstimulate touristsrsquo actual visits to Hong Kong but this does not translate into

TOURISM ECONOMICS78

the expected dollar income for the Hong Kong economy On the other handthe price of Hong Kong tourism relative to that of domestic tourism in thesource markets is the major determinant of touristsrsquo expenditure In additionthe responses of TE and TA to one-off events are found to be different D97DSARS and DFLU are all significant influencing factors in the TA (TA_P) modelswhile only DSARS shows a significant effect on TE (TE_P) This indicates thattourist arrivals are more likely to be subject to the influences of one-off eventsthan tourist expenditure

Interestingly substitute prices are excluded from almost all the models Thisimplies that substitute prices do not play a major role in determining thedemand for Hong Kong tourism by tourists from the three origin countriesThis low substitutability reflects the uniqueness of the tourism offerings inHong Kong and the strong comparative advantage of Hong Kong tourism inthe region

It should be noted that the differences in sample periods lag structures ofthe general ADLM and the inclusion of different explanatory variables allcontribute to the discrepancies of the model estimation results between thecurrent study and the other relevant studies However some consistencies havealso been observed For instance with regard to UK outbound tourism demandinsignificant effects of destination prices on tourist arrivals have been reportedby Song et al (2003) in the case of UK aggregate demand for Thai tourismand by Kulendran and Witt (2001) in the case of UK per capita demand fortourism to Greece the Netherlands and the USA

The finding that TA (TA_P) and TE (TE_P) are affected by differenteconomic factors provides useful information for practitioners Assuming thatthere was a deficit in the balance of payments more foreign exchange receiptsmight be desired rather than increasing the number of tourist arrivals In thissituation an appropriate pricing strategy that encourages tourist expenditurewould be desirable This is also essential as far as tourism businesses (such ashotels) are concerned in terms of their effective yield management Accordingto the model estimation results it seems that although price adjustments (suchas in promotions or price discounts) may not attract a greater number of gueststo a hotel they are likely to increase a hotelrsquos revenue by extending the lengthof the guestsrsquo stay in the hotel Since the model estimation results have shownthat various crises affect tourist expenditure significantly proper pricingstrategies and timely price adjustments are likely to be effective in crisissituations

Forecasting evaluation

To examine the forecasting accuracy of the alternative models one-step-aheadex post forecasts are generated for 2005 and 2006 by each model As with mosttourism forecasting studies the mean absolute percentage error (MAPE) androot mean square percentage error (RMSPE) are used to measure accuracy Thevalues and mean ranks of the models for each country are presented in Table 5Generally speaking most of the models generate relatively accurate forecastswith MAPEs being less than 6 The performance of each of the fourmodels is highly consistent across the three origin countries The TE model is


Table 5 Forecasting performance measured by MAPE and RMSPE

Australia UK USA OverallMAPE RMSPE MAPE RMSPE MAPE RMSPE MAPE RMSPE

TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)

Note The figures in parentheses refer to the ranks of forecasting performance of the models

the best-performing model across all the three source markets followed by theTA TA_P and TE_P models Table 5 shows that the forecasts of the aggregatedemand models are more accurate than the forecasts of the per capita demandmodels which implies that the inclusion of the origin country population intourism demand models is not beneficial in generating accurate forecasts

Conclusion

Tourist arrivals and tourist expenditure are both regarded as plausible measuresof tourism demand which have been used frequently in tourism demandmodelling and forecasting However little empirical research has been under-taken on comparing the appropriateness of these two alternative measures inidentifying the key determinants of tourism demand and forecastingperformance when different measures of demand are used in the modellingprocess This study uses data on the demand for Hong Kong tourism by threekey source markets and tests the suitability of the two alternative tourismdemand measures in tourism demand modelling and forecasting The empiricalresults show that the different patterns of TA (TA_P) and TE (TE_P) fluctuationare likely to be driven by different influencing factors Tourist arrivals (bothTA and TA_P) are more likely to be affected by origin country income andlsquoword-of-mouthrsquohabit persistence effects while tourist expenditure is drivenmainly by destination prices relative to those in the origin country Theforecasting performance of the four models estimated using different dependentvariables is also investigated and the results show that aggregate expenditurecan be predicted most accurately followed by total visits and per capita visitswhile forecasts of the per capita expenditure are the poorest The practicalimplications of these findings are that firstly econometric models of tourismdemand used for forecasting purposes should be estimated in aggregate forminstead of per capita form and secondly in order to devise effective tourismpolicies and strategies to increase tourism demand attention should be paid tothe particular measure of tourism demand of interest and its correspondingeconomic determinants

It should be noted that the conclusions drawn from this study are based ona particular data set related to the demand for Hong Kong tourism Thereforeany attempt to generalize the findings should be made with caution althoughthe methodology and procedure are readily applicable to the investigation of

TOURISM ECONOMICS80

other destinationndashorigin countryregion pairs across different time horizonsPossible extensions of this research are to employ larger data sets and alsoto use panel data analysis to verify the empirical results obtained in thisstudy

Endnotes

1 UNWTO also uses the term lsquotourism expenditurersquo to refer to aggregate spending of touristsin a destination Tourist expenditure and tourism expenditure are used interchangeably in thisstudy

References

Crouch GI (1994) lsquoThe study of international tourism demand a survey of practicersquo Journal ofTravel Research Vol 32 pp 41ndash54

Davidson J Hendry DF Saba F and Yeo S (1978) lsquoEconometric modelling of the aggregatetime series relationships between consumers expenditure and income in the United KingdomrsquoEconomic Journal Vol 88 pp 661ndash692

Durbarry R and Sinclair MT (2005) lsquoTourism competitiveness price and qualityrsquo TourismEconomics Vol 11 pp 45ndash68

Frechtling D (1987) lsquoAssessing the impact of travel and tourism ndash measuring economic benefitsrsquoin Ritchie JB and Goeldner CR eds Travel Tourism and Hospitality Research Handbook JohnWiley New York pp 325ndash331

Frechtling D (2001) Forecasting Tourism Demand Methods and Strategies Butterworth-HeinemannOxford

Gallet CA and Braun BM (2001) lsquoGradual switching regression estimates of tourism demandrsquoAnnals of Tourism Research Vol 28 pp 503ndash507

Garciacutea-Ferrer A and Queralt R (1997) lsquoA note on forecasting international tourism demand inSpainrsquo International Journal of Forecasting Vol 13 pp 539-549

Hendry DF (1995) Dynamic Econometrics An Advanced Text in Econometrics Oxford University PressOxford

Hendry DF and von Ungern-Sternberg T (1981) lsquoLiquidity and inflation effects on consumersrsquoexpenditurersquo in Deaton A ed Essays in the Theory and Measurement of Consumersrsquo BehaviourCambridge University Press Cambridge pp 237ndash261

Kim SH (1988) lsquoThe demand for international travel and tourism to South Korea an econometricevaluation of major economic factorsrsquo PhD thesis University of Santo Tomas Manila

Kulendran N and Witt SF (2001) lsquoCointegration versus least squares regressionrsquo Annals ofTourism Research Vol 28 pp 291ndash311

Li G Song H and Witt SF (2005) lsquoRecent developments in econometric modeling andforecastingrsquo Journal of Travel Research Vol 44 pp 82ndash99

Lim C (1997) lsquoReview of international tourism demand modelsrsquo Annals of Tourism Research Vol 24pp 835ndash849

Mizon GE and Richard JF (1986) lsquoThe encompassing principle and its application to testingnon-nested hypothesesrsquo Econometrica Vol 54 pp 657ndash678

Morley C (1992) lsquoA microeconomic theory of international tourism demandrsquo Annals of TourismResearch Vol 19 pp 250ndash267

Qiu H and Zhang J (1995) lsquoDeterminants of tourist arrivals and expenditures in Canadarsquo Journalof Travel Research Vol 34 pp 43ndash49

Schulmeister S (1979) Tourism and the Business Cycle Econometric Models for the Purpose of Analysisand Forecasting of Short-Term Changes in the Demand for Tourism Austrian Institute for EconomicResearch Vienna

Sheldon PJ (1993) lsquoForecasting tourism expenditure versus arrivalsrsquo Journal of Travel ResearchVol 32 pp 13ndash20

Sinclair MT and Stabler M (1997) The Economics of Tourism Routledge LondonSong H and Li G (2008) lsquoTourism demand modelling and forecasting ndash a review of recent

researchrsquo Tourism Management Vol 29 pp 203ndash220Song H and Witt SF (2000) Tourism Demand Modelling and Forecasting Modern Econometric

Approaches Pergamon Oxford


Song H Witt SF and Li G (2003) lsquoModelling and forecasting the demand for Thai tourismrsquoTourism Economics Vol 9 pp 363ndash387

Witt SF and Witt CA (1995) lsquoForecasting tourism demand a review of empirical researchrsquoInternational Journal of Forecasting Vol 11 pp 447ndash475

Zhang QH Wong KF and Or L (2001) lsquoAn analysis of historical tourism developmentand its implications for the tourism industry in Hong Kongrsquo Pacific Tourism Review Vol 5 pp15ndash21

TOURISM ECONOMICS64

commodity or to make use of a service combined with the ability to purchaseit lsquoThe significance level and repercussions of tourism demand provide a strongcase for better understanding of the nature of the touristsrsquo decision-makingprocessrsquo (Sinclair and Stabler 1997 p 15) Tourism demand is a special formof demand in that a tourism product is a bundle of complementary goods andservices (Morley 1992) Consumers instead of goods and services aretransported and tourism consumption occurs simultaneously with tourismproduction (Schulmeister 1979)

Tourism demand can be measured in a variety of ways Kim (1988 p 25)categorized the measurement criteria for all types of travel and tourism demandinto four groups (i) a doer criterion such as the number of tourist arrivalsthe number of tourist visits and the visit rate (ii) a pecuniary criterion forexample the level of tourist expenditure (receipts) and share of expenditure(receipts) in income (iii) a time-consumed criterion such as tourist-daystourist-nights and (iv) a distance-travelled criterion for instance the distancetravelled in miles or kilometres Among the above four categories the doercriterion and pecuniary criterion dominate international tourism demandstudies Considering statistical availability and consistency between data sourcestourist arrivals (TA) and tourist expenditure (TE) (receipts) are the mostcommonly used tourism demand measures in empirical studies along with theirderivatives such as the tourist participation rate derived from tourist arrivalsdivided by population of the origin countryregion (TA_P) and touristexpenditure per capita derived from total tourist expenditure divided bypopulation (TE_P) Other variations of tourism demand measures can also beidentified in empirical studies Based on published studies over the periods from1961 to the early 1990s Crouch (1994) and Lim (1997) summarized tourismdemand measures into five and six broad categories respectively In Crouchrsquos(1994) review 19 out of 80 published studies used more than one tourismdemand measure This phenomenon was also observed by Lim (1997) Li et al(2005) reviewed the literature on tourism demand modelling and forecastingusing econometric approaches published over the period 1990ndash2004 and foundthat 20 measures of tourism demand were used in those studies In addition18 out of 84 studies included in the review used more than one demandmeasure Based on Crouchrsquos categorization Table 1 summarizes the applicationsof various tourism demand measures in the studies reviewed by Crouch (1994)Lim (1997) and Li et al (2005) respectively It can be seen that the tourismdemand measures TA (and TA_P) and TE (and TE_P) have dominated tourismdemand modelling and forecasting studies over the past four decades Thefollowing discussion will therefore focus on these two sets of measures oftourism demand

Tourist arrivals versus tourist expenditure

Although both tourist arrivals and tourist expenditure are commonly usedmeasures in tourism demand modelling and forecasting the differences betweenthe two should not be ignored

First the fundamental difference between these two measures is the way inwhich these two types of data are collected International tourist arrivals are








TOURISM ECONOMICS66









The model



TOURISM ECONOMICS68



0

200000

400000

600000

800000

1000000

1200000

82 84 86 88 90 92 94 96 98 00 02 04 06

Australia UK USA




TOURISM ECONOMICS70




82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40








ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References

































TOURISM ECONOMICS66









The model



TOURISM ECONOMICS68



0

200000

400000

600000

800000

1000000

1200000

82 84 86 88 90 92 94 96 98 00 02 04 06

Australia UK USA




TOURISM ECONOMICS70




82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40








ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References





























The model



TOURISM ECONOMICS68



0

200000

400000

600000

800000

1000000

1200000

82 84 86 88 90 92 94 96 98 00 02 04 06

Australia UK USA




TOURISM ECONOMICS70




82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40








ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS68



0

200000

400000

600000

800000

1000000

1200000

82 84 86 88 90 92 94 96 98 00 02 04 06

Australia UK USA




TOURISM ECONOMICS70




82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40








ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References





























TOURISM ECONOMICS70




82 84 86 88 90 92 94 96 98 00 02 04 06

80

60

40

20

0

ndash20

ndash40








ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References
































ndash30

ndash20

ndash10

0

10

20

30

40

50

60

82 84 86 88 90 92 94 96 98 00 02 04 06


TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS72



CPIjtPst = 5Σj=1



wijt = TAijt5Σj=1

TAijt


ndash40

ndash20

0

20

40

60

82 84 86 88 90 92 94 96 98 00 02 04 06




The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References




























The data



Model specification


TDit = AYitβ2Pit

β3Pstβ4eit (1)



TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS74









θj lnPitndashj

+ 2Σj=0


Model estimation





TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References




























TA TE TA_P TE_P

α 7256 α 21007 α ndash2792 α 4233(3980) (381459) (ndash2670) (66449)












D97 0165 D97 D97 0161 D97

(2634) (2511)D911 D911 D911 D911







TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS76


TA TE TA_P TE_P













D97 0232 D97 D97 0232 D97

(3254) (3267)D911 D911 D911 D911








TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References




























TA TE TA_P TE_P

α 5002 α 22343 α ndash4935 α 2913(4527) (52556) (ndash0004) (6599)












D97 D97 D97 D97

D911 D911 D911 D911






TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS78










TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References





























TA 2978(2) 2979(3) 0788(2) 0813(2) 0507(1) 0684(1) 1424(2) 1480(2)TA_P 10580(4) 2557(2) 2108(3) 2168(3) 1259(3) 1680(3) 4625(3) 2135(3)TE 0316(1) 0365(1) 0425(1) 0455(1) 0676(2) 0686(2) 0472(1) 0502(1)TE_P 5203(3) 5242(4) 2526(4) 2692(4) 10185(4) 10259(4) 5996(4) 6060(4)



Conclusion



TOURISM ECONOMICS80


Endnotes


References


























TOURISM ECONOMICS80


Endnotes


References


























tourism demand modelling and forecasting: how …epubs.surrey.ac.uk/7594/69/s5.pdftourism demand...

Documents