lable at ScienceDirect

Tourism Management 41 (2014) 141e147

Contents lists avai

Tourism Management

journal homepage: www.elsevier .com/locate/ tourman

Quality and endogenous tourism: An empirical approachq

Isabel P. Albaladejo, María Isabel González-Martínez, María Pilar Martínez-García*

University of Murcia, 30100 Murcia, Spain

h i g h l i g h t s

� A theoretical model of endogenous growth and tourism is presented with data from Spain (1970e2010).� Both, theoretical and empirical analysis, highlight the role of quality of tourism services in long-run economic growth.� The TLG hypothesis is examined including quality of accommodations as an additional factor.� We conclude that quality of accommodations had a positive long-run effect on Spanish economic growth in 1970e2010.

a r t i c l e i n f o

Article history:Received 6 July 2012Accepted 6 September 2013

JEL classification:O41C61F43C32

Keywords:Quality in tourism servicesEndogenous economic growthTradeCointegrationGranger causality

q The authors have been partially supported by MIC01551/ECON, ECO2011-24352 (María Pilar Martínez-(Isabel Albaladejo). The third author also acknowledtion IS1104 “The EU in the new economic complex gpolicy evaluation”.* Corresponding author. Tel.: þ34 868 88 37 79; fa

E-mail addresses: isalba@um.es (I.P. Albaladejo), mMartínez), pilarmg@um.es (M.P. Martínez-García).

0261-5177/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.tourman.2013.09.006

a b s t r a c t

We propose a theoretical model and an empirical study that highlight the role of quality of tourismservices and endogenous tourism in long-run economic growth. We study a theoretical growth model ofinternational trade where tourism is the growth engine and quality of tourism services has a positiveimpact on long-term growth. We also provide an empirical analysis to test the relation between tourism,quality and economic growth in Spain over the period 1970e2010. Our results show that in the long run,tourist arrivals, quality of tourism accommodations, and foreign GDP have a positive effect on SpanishGDP. In the short term, changes in economic growth appear to lead to growth in tourist arrivals. Ourfindings support a two-way causal relationship between real GDP growth and tourism growth in Spain.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The belief that international tourism can promote economicgrowth is known as the tourism-led growth (TLG) hypothesis.Several studies have analyzed the relationship between tourismand economic growth both from a theoretical and an empiricalperspective. Both research lines, however, have experienceddifferent development levels and have not always evolved in thesame direction. Empirical studies have grown in profusion during

INN under projects ECO2008-García) and ECO2010-19830ges the support by COST Ac-eography: models, tools and

x: þ34 868 88 7905.aribel@um.es (M.I. González-

All rights reserved.

the past ten years and have focused on the causality relationshipbetween tourism and economic growth (Brida and Pulina (2010),Ivanov and Webster (2013) and Pablo-Romero and Molina (2013)provide reviews of published literature). Theoretical research hasrelied on Ramsey type models and study the impact of tourism onlong-run growth (see Lozano, Gómez, and Rey-Maquieira (2008)and Albaladejo andMartínez-García (2013) and references therein).With few exceptions, theoretical models assume that the touristattraction of countries is an exogenous characteristic given by theircultural, historical or natural endowment, where nothing can bechanged. However, the spread of the tourist industry all over theworld leads us to think that tourist attraction of countries is ceasingto be an exogenous characteristic. The tourist appeal is, on thecontrary, being endogenously improved by the economy itself.

In this paper we ask about the endogeneity of tourism and aboutthose factors that boost tourism and economic growth. We proposea theoretical model of endogenous growth and tourism which ischecked with real data. Following Albaladejo and Martínez-García

1 In Albaladejo and Martínez-García (2013) a more complex model with twokinds of productions (for domestic consumption/investment and for tourism) isanalyzed. However, assuming different goods do not change the model’s insights inthis paper. Thus, the simplification of a sole good is justified on the basis of a clearerexposition.

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147142

(2013), we assume that tourism arrivals and the quality of tourismservices can be endogenously enhanced, with the result that thelong-run economic growth rate in the host economy is positivelyaffected. The theoretical model inspires an empirical study toexamine the question on the TLG hypothesis in Spain, over theperiod 1970e2010, taking into account the quality of tourism ser-vices. This allows us to test whether higher quality of tourismservices can have a positive impact on long-term growth, as issuggested in our theoretical model.

The quality of tourism services is one of the main factors posi-tively affecting the success of tourism destinations and manycountries are adopting a policy of quality service in order toconsolidate themselves as an alternative in these highly competitivemarkets (Go & Govers, 2000). For example, Aguiló, Alegre and Sard(2005) shows that the Balearic Islands is a competitive destinationgiven that they have undergone a considerable restructuring pro-cess directed at offering improved quality. There exits an increasinginterest in having knowledge about the mechanisms that boost theappeal of one country versus others and how this appeal willeventually promote economic growth in the host countries.

Quality of tourism services is a broad concept. It is remarkablethe efforts devoted by academics to develops instruments tomeasure the quality of tourism services as perceived by the con-sumers. The most used instruments are SERVQUAL (Parasuraman,Zeithaml, and Berry 1988, 1991) and SERVPERF (Brown, Churchill,& Peter, 1993; Cronin & Taylor, 1992, 1994). However, in ourempirical study we have focused on the supply of tourism services,specifically on the quality of tourism accommodations. Nicolau andSellers (2010), and the references therein, highlight that quality ofaccommodations is strategic for increasing tourism competitive-ness. We have followed this idea in the empirical part of this paper.

Our paper contributes to the literature in two ways. First,building on the paper by Albaladejo and Martínez-García (2013), itprovides a simplified theoretical model of endogenous growth andtourism which can be contrasted with real data. Secondly, on theempirical side, it is the first attempt to examine the role of inter-national tourism in economic growth, including quality of tourismaccommodations as an additional factor. The paper is organized asfollows. Section 2 presents the theoretical model where threedifferent agents in the economic interact simultaneously: firms,consumers and tourists. The quality of tourism services and tourismarrivals promote the imports of foreign capital and economicgrowth. In Section 3 we contrast this model with real data. We usethe cointegration test by Johansen, and the multivariate Grangercausality tests based on an error correction model (ECM). Section 4concludes.

2. A theoretical framework

One of the first attempts to provide a theoretical model ofgrowth and tourism was the paper by Hazari and Sgro (1995), whoprovide a dynamic model of trade where the aggregate demand fornon-traded goods by tourists and domestic residents is taken intoaccount. Following this seminal paper, Hazari and Sgro (2004, ch.12), Chao, Hazari, and Sgro (2005), Nowak, Sahli, and Cortés-Jiménez (2007) and Schubert and Brida (2011) have also devel-oped Ramsey type models providing a theoretical link betweentourism and growth. Other papers like Cerina (2007), Giannoni(2009), Rigall-i-Torrent (2008), Gómez, Lozano, and Rey-Maquieira (2008), Lozano et al. (2008) and Rey-Maquiera, Lozano,and Gómez (2009) study the connection between environmentand tourism led economic growth.

The mainstream in this literature have considered that touristsarrive in the host country at a given exogenous rate, which is aparameter that is independent of the country’s characteristics. In

contrast, Albaladejo and Martínez-García (2013) endogenize therate at which tourists arrive. They assume that tourists arrive at arate which depends on the quality of the tourist services, which canbe endogenously improved by the country. To our knowledge thiswas the first attempt to link quality and endogenous growth oftourism countries in the literature. In this section we present asimplified version of this model which can be used as inspirationfor the empirical study.

Let us assume a market economy with three different types ofagents: domestic consumers; tourists and producers of a tradablegood which can be used either by domestic consumption or in-vestment and as tourism services.1 Production of this good, Y, re-quires labor, domestic and foreign capital. Since foreign capital is anecessary input for production, it must be imported by tradingdomestic production with nonresidents, that is, tourists.

2.1. Output production and firms’ behavior

Technology is described by a Cobb-Douglas production functionwith constant returns to scale and, in per capita terms, it can bewritten as

y ¼ kadkbf ; 0 < aþ b < 1; (1)

where kd and kf are the total per capita domestic and importedcapital in the economy.

Assuming perfect competition, instantaneous profit maximiza-tion leads to the following demand functions for inputs:

ð1� a� bÞy ¼ w; (2)

aykd

¼ r; (3)

bykf

¼ rrp: (4)

where w is wage paid to labor and r and rr are respectively the netrate of return to households that own both domestic and importedcapitals. The price p denotes the terms of trade.

2.2. Domestic consumers

Households own financial assets and labor. They hold assets inthe form of ownership claims on domestic and imported capital.Thus per capita households’ assets reads

assets ¼ kd þkfp: (5)

Assets deliver a rate of return and labor is paid a wage. The totalincome received by households is the sum of asset and labor in-comes. Households use the income that they do not consume toaccumulate more assets, that is,

dðassetsÞdt

¼ rkd þ�rr � _p

p

�kfpþw� c� n$assets; (6)

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147 143

where _p=p is the depreciation of foreign capital by increases in theterms of trade, c is per capita consumption and n> 0 is the constantpopulation growth rate. We omit time subscripts in Eq. (6) and inthe subsequent analysis whenever no ambiguity results.

Accumulation of per capita foreign and domestic capitals is theresult of decisions taken by households on consumption and oninvestment. Thus, per capita investment in foreign capital, if, drivesper capita foreign capital accumulation:

_kf ¼ if � nkf : (7)

From Equations (5)e(7) the evolution of per capita domesticcapital is obtained

_kd ¼ rkd þ rrkfpþw� c� if

p� nkd: (8)

It is assumed, for simplicity, that both capitals (domestic andforeign) do not depreciate.

Therefore, domestic household decides upon per capita con-sumption, c, and per capita investment in foreign capital, if, in orderto maximize overall utility, U, given by

U ¼ZN

0

c1�q

1� qeðn�rÞtdt (9)

subject to (7) and (8). Positive parameter q is the inverse of theelasticity of intertemporal substitution. The higher q is themore theconsumers prefer a uniform, smooth consumption path. For q ¼ 1 alogarithmic utility function is considered. Parameter r > 0 is theconstant rate of time preference.

Households are competitive in that each takes as given the rateof returns, r(t) and rr(t), the wage rate, w(t), and the terms of tradep, and their evolution, _p=p:

Optimally conditions for an interior solution lead to the usualRamsey rule which drives the dynamics of per capita consumption:

gch_cc¼ 1

qfr � rg ¼ 1

q

�aykd

� r

�: (10)

Here, and henceforth, gx denotes the growth rate of variable x:_x=x.

An optimal solution must also satisfy that

rr � _pp

¼ r; (11)

which says that both types of capitals are considered perfect sub-stitute assets which yield the same return.

2.3. A representative tourist‘s demand function

Eugenio-Martin (2003) postulates that tourists face a multi-stage decision problem in which the decisions about the destina-tion, budget and the length of stay are taken at different stages.Following this idea, we assume that, prior to traveling, the decisionabout the destination and budget has to some extent been taken.Once these decisions have been taken and our country is thechosen destination, the demand function of a tourist (it can beunderstood as a bundle of services like meals in restaurants, nightsin hotels, entertainments, etc.) is described by the followingfunction

dT ¼ bp; (12)

where b is the budget of a representative tourist. Both, terms oftrade, p, and budget, b, will evolve as time goes.

2.4. International market on equilibrium

In a perfect-foresight equilibrium, all agents take as given thetime paths of variables outside their control, the markets clear andequilibrium is then characterized by equating supply and demandfor all relevant quantities.

Taking this definition into account, note that there exists a flowof tourists arriving in the country to purchase and consume do-mestic tourism services. Let T(t) be the number of tourists in thehost economy at date t. On equilibrium, at any time t, internationalmarket clears, so the aggregate tourism demand for tourism ser-vices must equal aggregate tourism consumption, CT, that is

bpT ¼ CT : (13)

Additionally, on equilibrium, international trade is balanced atany time. Therefore, aggregate imports of foreign capital carried outfor domestic population must be equal to exports of tourismservices:

ifpL ¼ CT ; (14)

where L(t) is the size of domestic population.Equations (13) and (14) put forward that imports of foreign

capital, needed for production and growth, are financed by tourism,that is

if L ¼ bT (15)

2.5. Long-run growth rate

As has been known since Kaldor (1963), there are someempirical regularities on long-run growth rates of countries. In thelong run, per capita output and physical capital grow over time, therate of return of capital remains constant and the ratio of physicalcapital to final output is nearly constant. In order to obtain the long-run growth rate in the model we have just proposed, we focus onthose solutions that satisfy these regularities together with con-stant growth rates of final output (which in economic growththeory has been called balanced growth path).

Note that a constant rate of return of domestic capital, r, implies,from (10), a constant growth rate for per capita consumption. From(11), the constancy of the rate of return of foreign capital, rr, impliesthat the terms of trade grow at constant rate. Therefore, the con-stancy of the rate of returns implies, taking into account (3) and (1),that

gy ¼ gkd ¼ b

1� agkf (16)

Taking into account (15) and (7), the growth rate of foreigncapital on the balanced path equilibrium will be given by

gkf ¼ bTkf L

� n: (17)

Then, the growth rate of per capita foreign capital is constant ifand only if (bT)/(kfL) is constant, that is

gkf ¼ gb þ gT � n: (18)

Then, from (16),

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147144

gy ¼ b

1� aðgb þ gT � nÞ: (19)

Therefore, on a balanced growth path, the output growth ratewill be proportional to the addition of the growth rate of thenumber of tourists, gT, and the exogenously given growth rate oftourists’ budget, gb. Hence, those economic or marketing strategiesaffecting the growth rate of tourists will have a permanent effect onthe equilibrium growth rate.

Based on the diffusionmodel given byMorley (1998), Albaladejoand Martínez-García (2013) supposed that the number of touristsvisiting a certain destination is a function of the cumulative numberof past visitors, that is

TðtÞ ¼ s

Zt

0

TðsÞds: (20)

where s is the intrinsic rate of tourism attraction of the country.Although the tourism attraction of a country may depend on

several characteristics such as environmental quality, culturalattractiveness, destination image, etc., Albaladejo and Martínez-García (2013) considered that the intrinsic rate of tourism attrac-tion (s) is a measure of the quality of tourism services, which isincreasingly recognized as a key determinant in the success oftourist destinations.

Therefore, on differentiating Equation (20), it is obtained thatthe growth rate of T is influenced by the quality of tourism servicesof the country and by its own evolution, that is,

gT ¼ sþ gs: (21)

This equation says that the growth rate of the number of touristsis the addition of the growth rate of those tourist arriving thecountry because the diffusion effect (they have news from pastvisitors) and the growth rate of the intrinsic rate of tourismattraction, which is assumed a function of quality in this paper.

Taking into account (19) and (21),

gy ¼ b

1� aðgb þ sþ gs � nÞ; (22)

a long-term relationship between per capita output growth andquality of tourism services is put forward.2

In the following section we provide an empirical justification ofthe assumption that quality of tourism services is a relevant factorin economic growth for the Spanish economy.

3. An empirical analysis

The TLG hypothesis puts forward the existence of several argu-ments for tourism being a major factor of overall long-run economicgrowth. Tourism brings in foreign currencies which can be used toimport capital goods and services, favoring economic growth(McKinnon, 1964). A lot of empirical studies on TLG hypothesisemploy cointegration tests and Granger causality tests to investigatethe relationship between tourism and economic growth, and severalresults have been obtained. The paper of Balaguer and Cantavella-Jorda (2002), a starting point for research on TLG hypothesis, sup-ported this hypothesis in Spain for the period 1975e1997. They founda long-run relationship, and unidirectional causality from tourism

2 Note that Equation (22) expresses a relation between growth rates of differentvariables. By definition, gy ¼ _y=y, so the integral of gy is lny þ constant (the samehappens with the remaining growth rates in Equation (22)). Therefore, on inte-grating (22) we obtain a relation between Neperian logarithms, among variables inlevels, which is the equationwe are going to test in the following empirical analysis.

activity to economic growth. However, Oh (2005) disagreed with theTLG hypothesis in the case of South Korea, for the period 1975e2001.He did not find a long-run equilibrium relation, and causality testsindicated that economic growth led tourism expansion instead oftourism led economic growth. Dritsakis (2004), Kim, Chen, and Jang(2006) and Nowak et al. (2007) supported both the TLG hypothesisand the economic growth led tourism expansion. Kim et al. (2006)found two cointegration equations and a bidirectional causality be-tween tourism and economic development with two differentTaiwanese datasets. The results of Dritsakis (2004) and Nowak et al.(2007) showed the existence of a long-run relationship, and a two-way causal relation between tourism growth and economic devel-opment inGreece (1960e2000) andSpain (1960e2003), respectively.Gunduz and Hatemi-J (2005) and Katircioglu (2009) investigated theTLG hypothesis in the case of Turkey but reached different conclu-sions. Gunduz and Hatemi-J (2005) provided empirical support forthe TLG hypothesis for the period 1963e2002. They found unidirec-tional causality from international tourism to economic growth.However, Katircioglu (2009) did not find any cointegration relation-ship and rejected the TLGhypothesis for the period1960e2006. Up toourknowledge, in the case of Spain, the last published study is theoneby Cortés-Jiménez and Pulina (2010). These authors proved the val-idity of the TLG hypothesis in Spain and Italy for the period 1954e2000. Moreover, in the case of Spain the authors identified a bidi-rectional relationship between tourism and economic growth.

In the above-mentioned empirical research, tourism has beenmeasured either by tourism income or by the number of touristarrivals. The quality of tourism services offered by the host countryhas never been taken into account when analyzing the impact oftourism on economic growth. Since our focus is on quality,following the previous theoretical model, we have chosen thetourists arrivals as the measure of international tourism.

In this sectionweanalyze the roleof tourism foreconomic growthin Spain, over the period 1970e2010, taking into account that thequality offered by tourism accommodations can affect the long-rungrowth rate. Our growth model includes four variables: SpanishrealGDP, tourist arrivals, thequalityof tourismaccommodations, andforeign real GDP. The latter variable measures the tourists budget. Ifforeign real GDP grows, it seems reasonable to think that the budgetavailable for tourism of their citizens will be higher. The goal istwofold. First, to contribute to resolve the questions on the TLG hy-pothesis in Spain. Second, to determine if long-term economicgrowth in Spain is positively affected not only by the number oftourists but also by the quality of accommodations offered.

3.1. Methodology

Our empirical analysis consists of three stages: unit root tests,cointegration analysis, and Granger causality tests. In order to applythe correct econometric methodology, it is necessary to determinetheorder of integrationof the variablesweuse prior to estimating themodel (see Engle and Granger (1987)). Therefore, the first stage is totest for the presence of unit roots for the individual time series usingthe Augmented Dickey-Fuller (ADF) test (Dickey and Fuller, 1979,1981). Since the variables considered are integrated of order one,I(1), in the second stagewe analyze if they are cointegrated.We applythe multivariate cointegration procedure developed by Johansen(1988, 1991) and Johansen and Juselius (1990) to test the number ofcointegrating vectors and to estimate the coefficients of these coin-tegrating vectors. We provide a complete information on the long-run relationship, because the significance of each variable in thecointegrating vector has been tested, something which has not al-ways been studied. Since there is a cointegrating vector, the consid-ered variables are causally related at least in one direction (Engle andGranger, 1987), but cointegration by itself does not indicate the

Table 1Dickey Fuller unit root tests.

Level LY LT LQ LM

ss �3.11 (1) �1.66 (0) 0.72 (0) �2.85 (1)

First difference DLY DLT DLQ DLM

sm �3.14** (0) �5.67** (0) �3.65** (0) �5.08** (1)

Note: D denotes the first difference of variable under consideration, ss is the ADFstatistic in the most general model with an intercept and a time trend, sm is the ADFstatistic in the model with an intercept but no time trend. Numbers in brackets arethe number of lags for the ADF test based on the Akaike information criterion (AIC).Critical values are taken fromMacKinnon (1996). The symbol *(**) indicates that thenull hypothesis can be rejected at the 10%(5%) level.

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147 145

direction of the causal relationship. In the third stage we use themultivariate Granger causality test based on the ECM to investigatethe causal relationships between them. The ECM is a VAR model infirst differences augmented with the relevant error correction term,obtained from the cointegrating regression. This is the suitablemodelto test for Granger causality with cointegrated variables (otherwisespurious results may arise, Granger (1988)). Additionally, our short-run analysis includes the interpretation of the error correction termcoefficients, which have not always been shown in the literature. Thesign and size of these coefficients are important because they showwhether each variable is set in the right direction to achieve its long-run value, and the speed of this adjustment.

3.2. Data

The data used in this article are annual series on Spanish real GDP(Yt), number of tourists (Tt), ratio of luxury hotels and the totalnumber of hotels in Spain (Qt), and foreign real GDP (Mt) for theperiod 1970e2010. The ratio of luxury hotels (hotels of four and fivestars) and the total number of hotels is utilized as a proxyof qualityoftourism accommodations.3 Spanish real GDP series is from INE(SpanishNational Accounts). It isGDPat1986 constantprices and it ismeasured inmillions of local currency units. Data on tourist arrivals,measured in thousands of visitors, is from theEncuesta deOcupaciónHotelera, which is the Hotel Occupancy Survey published in Spain. Itis amonthly surveycompiledby the INEwith informationonall hotelestablishments registered as such in the corresponding register ofthe Tourist Boards of each Spanish autonomous community. Data onhotels are also taken from the Hotel Occupancy Survey.4 Finally,foreign real GDP is calculated as the weighted average of real GDP ofthecountries fromwhichmostof the touristswhocome toSpain:UK,Germany, France, Italy, the Netherlands, Belgium and Portugal. RealGDP data for these countries come from OCDE. On average, duringtheperiod2000e2010 (the only data available) nearly 80%of touristscame fromone of these countries: UnitedKingdom (27.6%), Germany(18.03%), France (15.22%), Italy (5.62%), Netherlands (4.36%), Belgium(3.19%) and Portugal (3.63%). To calculate this variable we haveweighted the realGDP of each country by theproportion of tourists itrepresents over these seven countries. Additionally, we haveconsidered the weighted average of real GDP of the three countriesfromwhichmost of the tourists come: UK, Germany and France. Theresults hardly differ, and are available on request.

All variables are expressed in natural logarithms so that thecoefficients of the model indicate elasticities.

3.3. Unit root test

The first step of our analysis is to test the order of integration ofthe natural logarithm of all the variables. The ADF test was carriedout to examine the presence of a unit root for both levels and firstdifferences of variables. After a graphical diagnosis, the unit roothypothesis was tested including a constant and a time trend in the

3 Although we have focused on the quality of tourism accommodations, alter-native variables can be used to define the tourism services quality of a destination(Spain). As a measure of quality we have also used the ratio between regular in-ternational flights and total number of regular flights. The results are robusts withthat presented in the paper and are available on request.

4 In Spain, a five-category system using stars is employed to classify the hotels.Hotels receive up tofive stars according to facilities and services offered. However, thisclassification depends on each Spanish region (see as examples, Order 7/1988 of theBalearic Islands, Order 77/2006 of the Autonomous Community of Madrid, Order 29/1987 of the Autonomous Community of Murcia, etc). Moreover, since the 1970s theminimum requirements have changed (see Real Decreto 3093/1982 and Real Decreto1634/1983). Ouranalysis suffers from thesedrawbacks. It shouldbe taken into accountthat the results could be affected by these changes between regions and over time.

ADF equation for levels of all variables and a constant in the ADFequation for first differences of variables. Table 1 shows the results.Based on MacKinnon’s (1996) critical values, the null hypothesis ofone unit root against the alternative of stationarity cannot berejected in levels of variables, but is rejected in their first differ-ences. Accordingly, all the variables under consideration are I(1).

3.4. Cointegration

The Johansen procedure chooses the number of cointegratingrelationships in the system, r, by two likelihood ratio test statistics:the trace test and maximum eigenvalue test. The trace test checksthe null hypothesis of r cointegrating vectors against the alternativehypothesis of n cointegrating vectors, where n is the number of I(1)variables. The maximum eigenvalue test, on the other hand, teststhe null hypothesis of r cointegrating vectors against the alternativehypothesis of r þ 1 cointegrating vectors. Neither of these teststatistics follows a standard chi square distribution. Asymptoticcritical values have been tabulated by Osterwald-Lenum (1992) fora range of values of n. Table 2 shows the cointegration analysisresults. According to the Akaike information criterion (AIC), wehave included three lags in the level VAR system to assure white-noise residual. Both trace and maximum eigenvalue tests suggestthat there is a unique cointegrating vector, at 5% significance level.This vector indicates the long-run relationship between the fourvariables included in our model.

Normalized cointegrating vector (b), reported in Table 3, in-dicates that tourist arrivals, the ratio between luxury hotels and thetotal number of hotels, and foreign GDP have a positive influence onthe Spanish GDP. LR tests show that in the cointegrating vector allvariables are statistically significant, at 5% significance level.5 Theinterpretation of the coefficients in the cointegrating equation is asfollows: an increase of 1% of tourist arrivals would imply an esti-mated increase of 0.08% in Spanish real GDP in the long run; anincrease of 1% in Spanish ratio of luxury hotels would cause, ingeneral, an increase of 0.11% in Spanish real GDP in the long run, soshowing the influence of the quality of tourism accommodations inthe economic growth. The foreign GDP has also a relevant rolewhen analyzing the long-run relationship. Our results imply thatlong-term economic growth in Spain is positively affected not onlyby tourist arrivals but also by the quality of accommodationsoffered in Spain, as we wanted to prove.

3.5. Granger causality

Once the number of cointegrating vectors has been established,we can test Granger causality. If the cointegration does not exist,

5 Conditional on a given value of r, inference about the elements of cointegratingvectors canbeperformedbymeans of LR statistics,whichwill thenhave their standardchi-squared asymptotic distributions under the null hypothesis being tested.

Table 3Normalized cointegrating vector & LR test for exclusion restrictions.

LY LT LQ LM

b 1 �0.08 �0.11 �1.04LRTb 19.10** 5.33** 3.69** 18.19**

Note: LRT is an LR statistic to test the significance of each variable in the cointe-grating vector. Critical values are taken from c2(1). The symbol *(**) indicates thatthe null hypothesis can be rejected at the 10% (5%) level.

Table 2Johansen cointegration tests.

H0: number of cointegrating vectors Trace test Max-eigenvalue test

r ¼ 0 53.34** 34.57**r � 1 18.77 11.79r � 2 6.98 4.82r � 3 2.16 2.16

Note: r denotes the number of cointegrating vectors. Optimal lag length for the VARwas selected by AIC criterion. Critical values are taken from Osterwald-Lenum(1992). The symbol *(**) indicates that the null hypothesis can be rejected at the10% (5%) level.

Table 4Granger causality test.

Dependentvariable

F-test t-test

DLY DLT DLQ DLM Zt�1 t-sta

DLY 2.43 1.36 17.01** �0.41 �2.94**DLT 9.67** 2.07 5.15* 0.59 0.66DLQ 3.44 0.70 5.43* �0.05 �0.17DLM 2.91 0.86 3.88 0.03 0.19

Note: The symbol *(**) indicates that the null hypothesis can be rejected at the 10%(5%) level.

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147146

causality tests should be conducted in a VAR model in first differ-ences. Because the series are cointegrated, theGranger causality testinvolves specifying a multivariate ECM as follows (Granger, 1988):

DLYt ¼ a1 þ d1Zt�1 þXpi¼1

f1iDLYt�i þXpi¼1

j1iDLTt�i

þXpi¼1

41iDLQt�i þXpi¼1

g1iDLMt�i þ 31t

DLTt ¼ a2 þ d2Zt�1 þXpi¼1

f2iDLYt�i þXpi¼1

j2iDLTt�i

þXpi¼1

42iDLQt�i þXpi¼1

g2iDLMt�i þ 32t

DLQt ¼ a3 þ d3Zt�1 þXpi¼1

f3iDLYt�i þXpi¼1

j3iDLTt�i

þXpi¼1

43iDLQt�i þXpi¼1

g3iDLMt�i þ 33t

DLMt ¼ a4 þ d4Zt�1 þXpi¼1

f4iDLYt�i þXpi¼1

j4iDLTt�i

þXpi¼1

44iDLQt�i þXpi¼1

g4iDLMt�i þ 34t

where Zt�1 is the lagged error correction term. The coefficients ofZt�1 (d), known as the adjustment parameters, measure the speedof adjustment of each dependent variable towards the long-runequilibrium. 31, 32, 33 y 34 are serially independent random errorswith mean zero and finite covariance matrix. Based on AIC crite-rion, the optimal lag length is two (p ¼ 2).

The ECM representation has the advantage of describing bothshort-run dynamics and long-run equilibrium simultaneously. Zt�1shows the long-run relation between the variables, and d providesthe short-run dynamics adjustment towards the long-run equilib-rium, revealing the importance of the cointegrating relationship ineach equation. The coefficients of lagged variables show the short-run changes occurring due to earlier changes in the variables.

Causality can be inferred either from the joint significance oflagged independent variables or of the lagged error correction term(Granger, 1988). The F-test of joint significance of lagged variablesconstitutes the short-run Granger causality. The t-test for the co-efficient of the Zt�1 provides the long-run Granger causality. Table 4examines short-run and long-run Granger causality. The second tothe fifth columns show the F-statistics for the lagged independentvariables and the last two columns show the estimated adjustmentcoefficient and their t-statistics in each of the four equations.

Beginning with the results for the t-test, the coefficient on thelagged error correction term, Zt�1, is statistically significant only inthe DLY equation. Besides, this coefficient (d1¼�0.41) has the

expected negative sign, which means that the Spanish real GDPadjusts in the correct direction to achieve its equilibrium value. IfLYt�1 is above its equilibriumvalue, Zt�1>0, then in order to restoreequilibrium,DLYtmust be negative. Specifically, a deviation from thelong-run equilibrium level during the current period will be cor-rected by 41% in the next period. Our result indicates that a long-runcausal relationship exists, running from T, Q and M to Y, and showsevidence in favor of the TLG hypothesis for the Spanish economy.

The F-statistic on the lagged variables suggests that there isGranger causality running from Spanish real GDP to tourist arrivals.This implies that short-term changes in Spanish real GDP are sig-nificant in influencing changes in number of tourists. Besides F-statistic shows Granger causality running from foreign real GDP toSpanish real GDP, tourist arrivals, and ratio of luxury hotels.

Our findings support a two-way causal relationship betweenreal GDP growth and tourism growth. In the long run both tourismdemand and quality of tourism accommodations boost Spanish realGDP. Over the short term, economic growth affects the growth inthe number of tourists.

4. Concluding remarks

In this paper a dynamicmodel of economic growth and tourism isstudied. Froma theoretical point of view, themain contribution of thepaper is that it considers tourists as consumers along with domesticpopulation. The exogenous growth of the rest of theworld boosts thearrival of tourists and increases their expenditures in the country, but,in addition, the economy can endogenously increase its rate oftourism attraction through the quality of tourism services. Tourismallows the imports of foreign capital to be financed and tourists ar-rivals increase thanks to quality improvements. Thus, the result is anendogenous growthmodelwhere tourism is thegrowthengine of theeconomy and quality has a positive impact on long-run growth rate.

We have checked the model with data from Spain over theperiod 1970e2010. Our empirical study is the first attempt toinvestigate the relevance of the quality of tourism accommodationson economic growth. In the long run, we can infer that tourist ar-rivals, quality of tourism accommodations and foreign GDP have apositive effect on Spanish real GDP. In the short term, changes ineconomic growth appear to cause tourist arrivals growth.

Our paper identifies some of the key variables for the TLG hy-pothesis: foreign GDP, which cannot be influenced by the politicalmeasures in the host country, tourist arrivals and quality (oftourism accommodations), which can be enhanced devoting re-sources to the tourism sector.

I.P. Albaladejo et al. / Tourism Management 41 (2014) 141e147 147

This paper highlights the role of quality of tourism services inlong-run economic growth for countries receiving tourists. Fromthe theoretical model we have proposed, and also from data of theSpanish economy, we can conclude that investment on quality havea positive effect on economic growth, which should be taken intoaccount in the design of sustainable tourism policies.

Acknowledgments

The authors are very grateful to the anonymous referees forhelpful comments on a previous version of this paper. Their com-ments have greatly improved the final version.

References

Aguiló, E., Alegre, J., & Sard, M. (2005). The persistence of the sun and sand tourismmodel. Tourism Management, 26(2), 219e231.

Albaladejo, I. P., & Martínez-García, M. P. (2013). An endogenous growth model ofinternational tourism. Tourism Economics, 19(3), 509e529.

Balaguer, L., & Cantavella-Jordá, M. (2002). Tourism as a long-run economic growthfactor: the Spanish case. Applied Economics, 34(7), 877e884.

Brida, J. G., & Pulina, M. (2010). A literature review on the tourism-led-growth hy-pothesis. Working paper CRENoS, 201017. Sardinia: Centre for North SouthEconomic Research, University of Cagliari and Sassari.

Brown, T. J., Churchill, G. A., & Peter, J. P. (1993). Improving the measurement ofservice quality. Journal of Retailing, 69(1), 127e139.

Cerina, F. (2007). Tourism specialization and environmental sustainability in a dy-namic economy. Tourism Economics, 13(4), 553e582.

Chao, C. C., Hazari, B. R., & Sgro, P. M. (2005). Tourism and economic development ina cash-in-advance economy. Research in International Business and Finance,19(3), 365e373.

Cortés-Jiménez, I., & Pulina, M. (2010). Inbound tourism and long-run economicgrowth. Current Issues in Tourism, 13(1), 61e74.

Cronin, J. J., & Taylor, S. A. (1992). Measuring service quality: a reexamination andextension. Journal of Marketing, 56(3), 55e68.

Cronin, J. J., & Taylor, S. A. (1994). SERVPERF versus SERVQUAL: reconcilingperformance-based and perceptions-minus expectations measurement of ser-vice quality. Journal of Marketing, 58(1), 125e131.

Dickey,D.,&Fuller,W. (1979).Distributionof theestimators forautoregressive timeserieswith a unit root. Journal of the American Statistical Association, 74(366), 427e431.

Dickey, D., & Fuller, W. (1981). Likelihood ratio statistics for autoregressive timeseries with a unit root. Econometrica, 49(4), 1057e1072.

Dritsakis, N. (2004). Tourism as a long-run economics growth factor: anempirical investigation for Greece using causality analysis. Tourism Economics,10(3), 305e316.

Engle, R. F., & Granger, C. W. J. (1987). Cointegration and error correction: repre-sentation, estimation and testing. Econometrica, 55(2), 251e276.

Eugenio-Martin, J. L. (2003). Modelling determinants of tourism demand as a five-stage process: a discrete choice methodological approach. Tourism and Hospi-tality Research, 4(4), 341e354.

Giannoni, S. (2009). Tourism, growth and residents’ welfare with pollution. Tourismand Hospitality Research, 9(1), 50e60.

Go, F. M., & Govers, R. (2000). Integrated quality management for tourist destina-tions: a European perspective on achieving competitiveness. Tourism Manage-ment, 21(1), 79e88.

Gómez, C. M., Lozano, J., & Rey-Maquieira, J. (2008). Environmental policy and long-term welfare in a tourism economy. Spanish Economic Review, 10(1), 41e62.

Granger, C. W. J. (1988). Some recent developments in a concept of causality. Journalof Econometrics, 39(1e2), 199e211.

Gunduz, L., & Hatemi-, J. A. (2005). Is the tourism-led growth hypothesis valid forTurkey? Applied Economics Letters, 12(8), 499e504.

Hazari, B. R., & Sgro, P. M. (1995). Tourism and growth in a dynamic model of trade.Journal of International Trade and Economic Development, 4(2), 243e252.

Hazari, B. R., & Sgro, P. M. (2004). Tourism, trade and national welfare. Contributionsto economic analysis. Amsterdam: Elsevier.

Ivanov, S.H., &Webster, C. (2013). Tourism’s contribution to economic growth: a globalanalysis for thefirst decade of themillennium. Tourism Economics,19(3), 477e508.

Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of EconomicDynamics and Control, 12(2e3), 231e254.

Johansen, S. (1991). Estimation and hypothesis testing of cointegrating vectors inGaussian vector autoregressive models. Econometrica, 59(6), 1551e1580.

Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference oncointegration with applications to the demand for money. Oxford Bulletin ofEconomics and Statistics, 52(2), 169e210.

Kaldor, N. (1963). Capital accumulation and economic growth. In Friedrich A. Lutz, &Douglas C. Hague (Eds.), Proceedings of a conference held by InternationalEconomics Association. London: Macmillan.

Katircioglu, S. T. (2009). Revisiting the tourism-led-growth hypothesis for Turkeyusing the bounds test and Johansen approach for cointegration. Tourism Man-agement, 30(1), 17e20.

Kim, H. J., Chen, M., & Jang, S. (2006). Tourism expansion and economic develop-ment: the case of Taiwan. Tourism Management, 27(5), 925e933.

Lozano, J., Gómez, C. M., & Rey-Maquieira, J. (2008). The TALC hypothesis andeconomic growth theory. Tourism Economics, 14(4), 727e749.

MacKinnon, J. G. (1996). Numerical distribution functions for unit root and coin-tegration tests. Journal of Applied Econometrics, 11(6), 601e618.

McKinnon, R. (1964). Foreign exchange constrain in economic development andefficient aid allocation. Economic Journal, 74(294), 388e409.

Morley, C. L. (1998). A dynamic internationalmodel. Annals of TourismResearch, 25(1),70e84.

Nicolau, J. L., & Sellers, R. (2010). The quality of quality awards: diminishing infor-mation asymmetries in a hotel chain. Journal of Business Research, 63, 832e839.

Nowak, J. J., Sahli, M., & Cortés-Jiménez, I. (2007). Tourism, capital imports and eco-nomic growth: theoryand evidence for Spain. TourismEconomics,13(4), 515e536.

Oh, C. (2005). The contribution of tourism development to economic growth in theKorean economy. Tourism Management, 26(1), 39e44.

Osterwald-Lenum, M. (1992). A note with quintiles of the asymptotic distribution ofthe maximum likelihood cointegration rank test statistics: four cases. OxfordBulletin of Economics and Statistics, 54(3), 461e472.

Pablo-Romero, M., & Molina, J. A. (2013). Tourism and economic growth: a review ofempirical literature. Tourism Management Perspectives, 8, 28e41.

Parasuraman,A., Zeithaml, V. A., &Berry, L. L. (1988). SERVQUAL: amultiple-itemscale formeasuring consumerperceptions of service quality. Journal of Retailing, 64(1),12e40.

Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1991). Refinement and reassessmentof the SERVQUAL scale. Journal of Retailing, 67(4), 420e450.

Rey-Maquiera, J., Lozano, J., & Gómez, C. M. (2009). Quality standards versustaxation in a dynamic environmental model of a tourism economy. Environ-mental Modelling & Software, 24(12), 1483e1490.

Rigall-i-Torrent, R. (2008). Sustainable development in tourism municipalities: therole of public goods. Tourism Management, 29(5), 883e897.

Schubert, S. F., & Brida, J. G. (2011). The impacts of international tourism demand oneconomic growth of small economies dependent on tourism. Tourism Man-agement, 32(2), 377e385.

Isabel P. Albaladejo is Associate Professor in QuantitativeMethods at the University of Murcia (Spain). She teachesMathematics for Economic Analysis at the Faculty of Eco-nomics and Business. She has worked in statisticalmethods related to tourism analysis, in particular discretechoice modeling. This research was directed towards ruraltourism, with the emphasis on tourist behavior. Her con-tributions have been published in Tourism Management,Tourism Economics and in Spanish tourism researchjournals. Her current research focuses on endogenousgrowth models and tourism.

María Isabel González-Martínez is Associate Professor inQuantitative Methods at the University of Murcia (Spain).She teaches Econometrics at the Faculty of Economics andBusiness. She has worked in Time Series EconometricMethods applied to International Economics. Her contri-butions have been published in Spanish and internationaljournal like Revista de Economía Aplicada, Moneda yCredito, and Économie Appliquée, among others. Currentlyshe is interested in tourism and economic growth long-runrelationship.

María Pilar Martínez-García is Associate Professor inMathematics for Economic Analysis at the University ofMurcia (Spain). She has contributed to Economic Dy-namics and Economic Growth Theory with publications injournals like Journal of Economic Dynamics and Control,International Game Theory Review, The Journal of Inter-national Trade & Economic Development, Decisions inEconomics and Finance, Tourism Economics, amongothers. Currently she is interested in the Tourism LedGrowth Hypothesis.