Volume 30, Issue 1. On the Relationship between Tourist Flows and Household Expenditure in Barbados: A Dynamic OLS Approach

Transcription

1 Volume 30, Issue 1 On he Relaionship beween Touris Flows and Household Expendiure in Barbados: A Dynamic OLS Approach Mahalia Jackman Cenral Bank of Barbados Troy Lorde Universiy of he Wes Indies Absrac Keynesians propose ha increases in ouris arrivals are associaed wih an expansion in privae spending hrough he muliplier effec. To es his hypohesis, his sudy augmens a simple consumpion funcion wih ouris arrivals and employs he dynamic OLS mehod o compue he shor and long run relaionships of he variables. Resuls sugges ha while ouris arrivals have a posiive correlaion wih household expendiure in he shor run, i does no Granger cause household expendiure consumpion. Ciaion: Mahalia Jackman and Troy Lorde, (2010) ''On he Relaionship beween Touris Flows and Household Expendiure in Barbados: A Dynamic OLS Approach'', Economics Bullein, Vol. 30 no.1 pp Submied: Aug Published: February 08, 2010.

2 1. Inroducion Does he number of ouris affec household consumpion decisions? Keynesians argue ha increases in ouris arrivals are associaed wih an expansion in privae spending hrough he muliplier effec. As oulined by Tse (1998) pp. 233, Tourism spending represens a ne demand for domesic oupu. Such a ne demand provides incomes in he economy. recipiens of his exra income in urn raise heir consumpion, creaing furher producion and income gains in an endless bu rapidly diminishing chain. Indeed, a number of sudies have found evidence of causaliy running from ourism o economic oupu (see for insance Dubarry, 2004; Lee and Chang, 2008; and Brida e al, 2008). Given ha income is ofen he primary deerminan of consumpion, i is reasonable o assume ha ouris arrivals can also impac household consumpion. Tse (1998), o he bes of he auhors knowledge, is he only sudy which empirically invesigaed he relaionship beween ourism and aggregae domesic consumpion. Using he Engle-Granger approach o coinegraion and bivariae vecor auoregression, he auhor found ha he ourism-driven consumpion hypohesis is no held in Hong-Kong. Hence, he hypohesis ha ourism growh booss domesic consumpion is no fully verified. Agains his backdrop, his paper seeks o invesigae he impac of ourism on he consumpion behaviour of households in Barbados. Specifically, i aemps o deermine if greaer volumes of arrivals are capable of boosing household expendiure. I is well documened ha he impac of ourism on he Barbadian economy is profound and exensive (see Archer, 1984; Jackman and Greenidge, 2010). Tourism direcly employs over 10% of he labour force, conribues over 10% o Gross Domesic Produc and is he main earner of foreign exchange, accouning for over 60% of he counry s oal foreign receips. More han his, he muliplier effecs of ourism booss oher secors of he economy, such as whole sale and reail, consrucion, enerainmen and food and drinks secors, which aid in servicing he indusry. Therefore, by sudying he case of Barbados, his paper provides some insigh ino how ourism affecs privae spending in counries specialising in ourism. Indeed several sudies advocae ha he impac of ourism ends o depend on he relaive weigh of ourism in he economy. I has also been documened ha ourism ends o have a more favourable effec on small economies (see Sgro and Hazari, 1995; 1995; Candela and Cellini, 1997). Given ha he economies, economic forunes of Barbados are more closely ied o he ourism indusry han in Hong Kong, 1 and he fac ha Barbados has a relaively smaller economy, 2 i is possible ha he ourism-led consumpion hypohesis can hold rue for Barbados bu no for Hong Kong. To effecively examine he long and shor-run co-movemens beween ourism and domesic consumpion, his sudy augmens a model of household expendiure wih ourism indicaors. Tse s sudy, in our opinion, did no pay adequae aenion o he possibiliy of omied variables bias. There are a number of oher facors ha migh impac on domesic consumpion and he exclusion of hese variables from he simple bivariae analysis may have masked he underlying 1 According o he World Travel and Tourism Council, Barbados ranks eigh wih regards o he conribuion of ourism o naional economy, whereas Hong Kong ranks fory-one. 2 According o he CIA World Fac Book, in 2008 Barbados had a GDP per capia of approximaely US $19,100 while for Hong Kong, GDP per capia was roughly US $43,700. 1

3 relaionship beween he wo variables under invesigaion. In addiion, insead of using convenional sysem based or residual based mehods o coinegraion, his sudy uilises a dynamic OLS approach. The main advanage of his mehod is ha i allows for he simulaneous modelling of saionary [I(0)] and non-saionary [I(1)] variables, which is an imporan gain. I is unreasonable o assume ha because a variable is [I 0)], i does no have a long run effec on he dependen variable. The remainder of his paper is organised as follows. The subsequen secion describes he empirical mehodology employed. This is followed by a discussion of he resuls and he final secion provides some concluding remarks. 2. Empirical Mehodology 2.1 Empirical Model Boh he empirical and heoreical lieraure have idenified a number of key variables ha may impac households decision o consume. The sandard variables o include in any consumpion are income, prices and ineres raes. Following Keynes (1936), higher income is expeced o boos domesic consumpion. On he oher hand, a higher price ofen deers demand and is expeced o be negaively relaed o household expendiure. Finally, in line wih classical economic heory, one posulaes ha higher raes of ineres encourage saving and discourages consumpion, and hence should be inversely relaed o household expendiure. In his paper, a simple consumpion funcion is augmened wih ouris flows o explore he impac of ourism on household expendiure in Barbados. Formally, C = f ( T, Y, P, i ) (1) where C, T, Y, P and i represen household expendiure, ouris arrivals, income, prices and ineres raes, respecively. 2.2 Daa The daabase employed consiss of annual daa for he period 1970 o Observaions on income (proxied by real GDP per capia) and household expendiure (deflaed by he GDP deflaor) are aken from World Bank s World Developmen Indicaors Daabase (2008); prices (approximaed by he consumer price index) are obained from he Inernaional Moneary Fund s Inernaional Financial Saisics Daabase (2008); and finally, observaions on he minimum deposi rae 3 and ouris arrivals are obained from he Cenral Bank of Barbados Annual Saisical Diges. All variables, excep he minimum deposi rae, are expressed in logarihms. 2.3 Economeric Approach Given he relaively small size of he sample, he auhors op for a single-esimaion approach o coinegraion. Hence, he dynamic OLS (DOLS) procedure developed by Saikkonen (1991) and generalised by Sock and Wason (1993) is employed. DOLS correcs for possible simulaneiy 3 The minimum deposi rae is se by he Cenral Bank of Barbados and serves as a benchmark for ineres raes movemens in Barbados. 2

4 bias amongs he regressors and is herefore preferable o oher single equaion coinegraion echniques, namely he Engle and Granger (1987) wo-sep esimaor. DOLS esimaes are also asympoically normally disribued; hus, direc saisical inference on he parameers of he coinegraing vecor is possible. In addiion, unlike mos residual based and sysem based ess for coinegraion, DOLS can accommodae varying orders of inegraion, i.e. i allows for direc esimaion of a mixure of I(0) and I (1) variables. I should be noed ha when all he variables of ineres are I (1) and here is a single coinegraing vecor, DOLS is asympoically equivalen o he full informaion maximum likelihood approach developed by Johansen (1988). The poenial simulaneiy bias and small sample bias among he regressors is deal wih by including lags and leads of he firs differenced I (1) regressions. Hence, he esimaion of he long run relaionship of equaion 1 is: k ' I ' j j j=-k C B X X (2) where X = [T, Y, P, i], X I is he subse of I (1) variables and B is he vecor of long-run coefficiens. The equaion is esimaed wih k = 2, hen a general o specific specificaion search echnique (see Hendry and Richard, 1982; Campos e al., 2005;) is employed in order o obain a parsimonious represenaion of he daa generaing process. To invesigae he shor run dynamics of he model, one uses he esimaes from equaion 2 o generae an error correcion model. Formally, k k k ' I ' j j j j j j ( 1 ' 1) j1 j=0 j=0 (3) C C X Y C B X which denoes he changes in household expendiure as a funcion of changes of he lagged firs difference of he non-saionary variables (X I ), saionary variables (Y) and an error correcion erm. As in he case of equaion of 2, equaion 3 is reduced o a more parsimonious form using a general o specific procedure. 3. Resuls As a preliminary sep o he analysis, he order of inegraion of he variables is deermined. A plo of each series shows ha several series appear o undergo a leas one srucural break over he period under sudy. Since sandard uni roo ess can have reduced power if hey are applied o a ime series wih a srucural break, he uni roo es by Saikkonen and Lukepohl (2002) and Lanne e al. (2002) ha akes srucural breaks ino accoun is adoped. The es considers models wih general nonlinear deerminisic shif funcions. In he firs sep of he es, he deerminisic componen is esimaed and subraced from he series. In he second sep, he sandard ADF uni roo es is applied o he ransformed series. Criical values can be found in Lanne e al. (2002, p.678). 3

5 Table 1: Uni Roo Tess Variable Level Firs Break Decision Difference Dae Household Expendiure ** 2001 I(1) Touris Arrivals *** 1991 I(1) Income *** 1992 I(1) Price *** 1979 I(1) Policy Rae *** N.A I(0) Noes: ***, ** and* indicae significance a he 1%, 5% and 10% levels, respecively. Table 1 presens he uni roo es resuls, which imply ha all variables exceping he minimum deposi rae, are I (1). Hence, equaion 2 is esimaed inclusive of he leads and lags of Y, T, and P in firs differences. The parsimonious specificaion, obained by sequenially deleing he insignifican variables, is presened in Table 2. The resuls suggess ha only prices and income form he equilibrium subspace, i.e. only prices and income seem o impac household expendiure in he long run, and heir coefficiens conform o he a priori expecaions oulined in Secion 2.1. These findings are somewha similar o hose of Tse (1998), who also repor he absence of a long-run relaionship beween ouris arrivals and domesic consumpion. Table 2: Long Run Esimaes Dependen Variable: C Consan (16.00) [0.000] P (-5.70) [0.000] Y (11.1) [0.000] R Long Run Diagnosic Tes Saisics Auoregressive condiional heeroscedasiciy F (2, 22) = [0.551] Serial Correlaion (LM) F (1, 22) = [0.504] 2 Normaliy (2) = [0.897] Ramsey Rese Tes F (1, 27) = [0.674] Heeroscedasiciy F (10, 17) = [0.976] Noe: T-saisics are in parenheses ( ) and he associaed p values in square brackes [ ]. The resuls of he shor-run dynamic coefficiens are shown in Table 3. As required, he coefficien on he error correcion erm is negaive and significan, herefore confirming he exisence of a coinegraing relaionship. In fac, he value of he coefficien implies ha abou 4

6 29% of he discrepancy beween he acual and equilibrium household expendiure is correced in each year. I should be noed ha here is some difference beween he variables enering he shor run model and hose in he long-run. Though he income variable reains is posiive and significan coefficien, he findings indicae ha in he shor run, prices have no effec on he consumpion decisions of households. In addiion, he minimum deposi rae reduces household consumpion in he shor-run bu has no effec in he long-run. Furhermore, his impac is lagged one year. This is in line wih he findings of Mamingi e al (2009) who repor ha on average, i akes abou four o six quarers for he full effec of changes in he minimum deposi rae o be ransmied o he economy via adjusmens. Of paricular ineres, ouris arrivals seem o be posiively relaed o household consumpion. Table 3: Parsimonious Error-Correcion Model Dependen Variable: C Consan (16.00) [0.000] (-3.35) [0.002] T (1.85) [0.076] Y (3.10) [0.004] i (-2.93) [0.007] R Shor Run Diagnosic Tes Saisics Auoregressive condiional heeroscedasiciy F (1, 26) = [0.349] Serial Correlaion (LM) F (2, 26) = [0.736] 2 Normaliy (2) = [0.464] Ramsey Rese Tes F (1, 27) = [0.698] Heeroscedasiciy F (10, 17) = [0.126] Noe: T-saisics are in parenheses ( ) and he associaed p values in square brackes [ ]. 5

7 To es he adequacy of he resuls, boh he long run and shor run models are subjeced o a baery of diagnosic ess (see he boom panel of Tables 2 and 3). These ess imply ha he models are well specified 4. Moreover, he models appear o effecively explain household consumpion paerns in Barbados evidenced by Figures 1 and 2. Figure 1: Fied Long Run Model 8.1 HCONS Fied I should be noed ha he general error correcion model had a non-normaliy problem. A look a he residual series revealed very large values for observaions in he early 1980s, early 1990s and 2001, which corresponds o periods of recessions in he Barbadian economy. To correc his issue, he model was re-esimaed wih observaion specific dummy variables. 6

8 Figure 2: Fied Error-Correcion Model 0.15 DHCONS Fied A small se back in he findings is ha resuls only indicae ha here is posiive co-movemen beween ouris arrivals and domesic consumpion in he shor-run. However, correlaion does no imply causaion. Hence, we now seek o deermine wheher here is a causal relaionship beween ourism and household expendiure. To es he exisence of causaliy in he shor run, he Granger causaliy es developed from he seminal paper of Granger (1969) is employed. Basically, his es seeks o ascerain wheher or no he inclusion of pas values of a variable x do or do no help in he predicion of presen values of anoher variable y. If variable y is beer prediced by including pas values of x han by no including hem, hen, x is said o Granger-cause y. A es of join significance of he lagged values (in his case ourism in firs differences) consiues he Granger causaliy es. The resuls are provided in Table 4. Table 4: Granger Causaliy Tes Causaliy 2 -Saisic P-Value T C C T Noe: The noaion T C represens he null: Growh in Tourism does no Granger-cause growh in household expendiure. A similar inerpreaion follows for he reverse es. 7

9 From he es saisics, i is clear ha he null hypohesis Growh in Tourism does no Grangercause growh in household expendiure canno be rejeced. In addiion, here is no evidence o sugges ha higher consumpion booss ourism. Overall, our resuls are in line wih hose of Tse (1998). 4. Concluding Remarks The primary purpose of his sudy is o empirically examine he impac of ouris flows on household expendiure in Barbados. Hence, a simple model of consumpion is augmened wih ourism indicaors and DOLS applied o invesigae he long run and shor run relaionships beween he variables of ineres. Despie he imporance of ourism o he Barbadian economy, we are unable o find any evidence of a long-run relaionship beween ouris arrivals and household expendiure. Furhermore, while ouris arrivals have a posiive correlaion wih expendiure in he shor run, i does no Granger cause i. Taken ogeher, our resuls sugges ha, like Hong Kong, he hypohesis ha ourism booss domesic consumpion is no verified for Barbados. References Archer, E (1984) Esimaing he relaionship beween ourism and economic growh in Barbados Journal of Travel Research 22, Brida, J., E. Carrera and W. Risso (2008) Tourism s Impac on Long-Run Mexican Economic Growh Economics Bullein 3(21) pp1-8 Drisakis, N (2004) Tourism as a long-run growh facor: An empirical invesigaion for Greece using causaliy analysis Tourism Economics 10, Dubarry, R (2004) Tourism and economic growh: The case of Mauriius Tourism Economics Campos, J., N.R. Ericsson and D.F. Hendry (2005) General o specific modelling: an overview and seleced bibliography Inernaional Finance Discussion Paper No. 838, Board of Governors of he Federal Reserve Sysem: USA. Candela, G and R. Cellini. (1997) Counries size, consumer s preferences and specializaion in ourism: A noe Rivisa Inernazionale di Scienze Economiche e Commerciali 44, Engle, R and C.W.J. Granger (1987) Coinegraion and error-correcion: represenaion, esimaion and esing Economerica 55, Hendry, D.F. and J.F. Richard (1982) On he formulaion of empirical models in dynamic economerics Journal of Economerics 20, 3-33 Jackman, M., and K Greenidge (2010) Modelling and forecasing ouris flows o Barbados using srucural ime series models Tourism and Hospialiy Research 10, 1-13 Johansen, S. (1988) Saisical analysis of coinegraing vecors Journal of Economic Dynamics and Conrol 12, No 2/3, Keynes, J.M. (1936), The General Theory of Employmen, Ineres and Money, Macmillan Cambridge Universiy Press, London, UK. Lanne, M., H. Lukepohl and P. Saikkonen (2002) Comparison of uni roo ess for ime series wih level shifs Journal of Time Series Analysis 23,

10 Lee, C and C. Chang (2008) Tourism developmen and economic growh: A closer look a panels Tourism Managemen 29, Mamingi, N., D. Boamah and M. Jackman (2009) Ineres rae pass-hrough: evidence for The Bahamas and Barbados Cenral Bank of Barbados Mimeo. Saikkinen, P (1991) Asympoically efficien esimaion of coinegraion regressions Economic Theory 7, 1-21 Saikkonen, P and H. Lukepohl (2002) Tesing for a uni roo in a ime series wih a level shif a unknown ime Economeric Theory 18, Sgro, P and B. Hazari (1995) Tourism and growh in a dynamic model of rade The Journal of Inernaional Trade and Economic Developmen 4, Sock, J.H and M.W. Wason (1993) A simple esimaor of coinegraing vecors in higher order inegraed sysems Economerica 61, Tse, R (1998) Do more ouriss lead o higher levels of consumpion? Tourism Economics 4,