Tourism: low-cost airlines, climate and economic crisis

UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA Departamento de Análisis Económico Aplicado Doctorado en Economía Programa doctorado: Aplicaciones a las finanzas y seguros, a la economía sectorial, al medio ambiente y las infraestructuras. TESIS DOCTORAL TOURISM: LOW-COST AIRLINES, CLIMATE AND ECONOMIC CRISIS Tesis doctoral presentada por D. Federico Inchausti Sintes Dirigida por Dr. Juan Luis Eugenio Martín El Director, El Doctorando, Las Palmas de Gran Canaria, Junio de 2014 “A mis padres, a mi familia y a mis amigos que también son familia.” i Contents Contents ...................................................................................................................................i List of figures ......................................................................................................................... v List of tables ........................................................................................................................ vii Chapter 1: Does low-cost travelling imply higher tourism expenditure in the destination? 1.1Introduction ...................................................................................................................... 1 1.2 State of the art .................................................................................................................. 4 1.2.1. LCCs characteristics ................................................................................................. 4 1.2.2. Main consequences of the presence of LCCs in a destination ................................. 5 1.2.3. Have LCCs increased the flow of tourism? ............................................................. 5 1.2.4. Have LCC passengers got a different behaviour with respect to traditional carrier passengers? ......................................................................................................................... 6 1.2.5. Have LCCs presence promoted economic growth in the region where they travel to? ....................................................................................................................................... 7 1.2.6. Are LCCs passengers’ savings in the origin transferred to higher tourism expenditure at the destination? ........................................................................................... 8 1.2.7. Modeling tourism expenditure ................................................................................. 8 1.2.7.1. Sample selection models .................................................................................. 9 1.2.7.2. System of equations ......................................................................................... 9 1.2.7.3. Other methodologies ...................................................................................... 10 1.3 Methodology .................................................................................................................. 11 1.3.1. First approach: Descriptive analysis ...................................................................... 11 1.3.2. Second approach: Regression analysis ................................................................... 12 1.3.3. Third approach: Simultaneous system of equations ............................................... 13 1.3.4. Theoretical microeconomic motivation ................................................................. 14 1.3.5. Econometric approach to the microeconomic motivation ...................................... 15 1.4 Case study ...................................................................................................................... 17 1.5 Methodology .................................................................................................................. 19 1.5.1. Endogeneity ............................................................................................................ 20 1.5.2. Contemporary correlations among error terms ...................................................... 20 ii 1.5.3. Heteroskedasticity ................................................................................................. 21 1.5.4. Estimation under the presence of endogeneity, contemporary correlation error terms and heteroskedasticity. ........................................................................................... 21 1.5.5. The reduced form................................................................................................... 22 1.6 Results ........................................................................................................................... 23 1.6.1. Reweighting effect ................................................................................................. 23 1.6.2. Results from the structural form ............................................................................ 27 1.6.3. Results from the reduced form .............................................................................. 35 1.6.4. Savings transfer ratios ........................................................................................... 36 1.7 Conclusions and further research .................................................................................. 37 Chapter 2: Understanding tourists´economising strategies during the global economic crisis 2.1 Introduction ................................................................................................................... 41 2.2 State of the art ................................................................................................................ 45 2.1. Macroeconomic indicators ....................................................................................... 46 2.2. Microeconomic indicators ........................................................................................ 48 2.3 Methodology .................................................................................................................. 49 2.3.1. Econometric modeling ........................................................................................... 49 2.4 Case study ...................................................................................................................... 53 2.5 Results ........................................................................................................................... 54 2.5.1. Economising strategies by country ........................................................................ 54 2.5.2. Estimation .............................................................................................................. 56 2.5.3. Post-estimation analysis ........................................................................................ 59 2.6 Conclusions and further research .................................................................................. 66 Chapter 3: Tourism: economic growth, employment and Dutch Disease. 3.1Introduction .................................................................................................................... 69 3.1.1. GDP and employment in Spain at a glance ........................................................... 72 3.1.2. Inbound Tourism in Spain at a glance ................................................................... 73 3.2 State of the art ................................................................................................................ 76 3.2.1. Economic growth theories ..................................................................................... 76 3.2.2. Tourism and economic growth .............................................................................. 76 3.2.2.1. Tourism and economic growth by countries .................................................. 77 3.2.2.2. Tourism and economic growth by regions ..................................................... 77 3.2.2.3. Tourism and capital accumulation .................................................................. 78 3.2.3. Tourism and employment ...................................................................................... 79 3.2.4. Tourism and CGE models ..................................................................................... 80 iii 3.2.5. Unemployment in CGE models ............................................................................. 83 3.3 Theoretical framework of the impact of tourism on an economy: TNT-T model ......... 86 3.4 Case study ...................................................................................................................... 89 3.5 Methodology .................................................................................................................. 90 3.5.1. Recursive-dynamic CGE model (BLOFIVIS models) ........................................... 90 3.5.1 Main equations ........................................................................................................ 92 3.5.2. Model closure ......................................................................................................... 94 3.5.3. Shock, scenarios and calibration path .................................................................... 95 3.6 Results ............................................................................................................................ 96 3.6.1. Unemployment ....................................................................................................... 96 3.6.2. Gross real added value (GRAV), foreign account deficit and capital accumulation97 3.6.3. Domestic demand ................................................................................................. 101 3.6.4. Winners and losers: “Dutch Disease” .................................................................. 102 3.6.4.1. Expenditure effect ........................................................................................ 104 3.6.4.2. Resource effect ............................................................................................. 105 3.7 Conclusions and further research ................................................................................. 110 Chapter 4: The role of climate and the tourism destination choice 4.1Introduction .................................................................................................................. 113 4.2 State of the art .............................................................................................................. 115 4.2.1. Destination choice: the last step in a complex decision process .......................... 115 4.2.1.1. The role of climate ....................................................................................... 116 4.2.1.2. The role of distance ...................................................................................... 117 4.2.1.3. Destination choice methodologies: all roads guide to mixed logit .............. 118 4.3 Case study .................................................................................................................... 120 4.4 Methodology ................................................................................................................ 121 4.5 Results .......................................................................................................................... 122 4.6 Conclusions and further research ................................................................................. 128 Annexs Annex 3.1. TSA-IOT and SAM ......................................................................................... 131 A.3.1.1 Introduction ........................................................................................................... 131 A.3.1.2 Integration of the IOT and the TSA ...................................................................... 132 A.3.1.2.1. Input-Output Table accounting framework ................................................... 132 A.3.1.2.2. Tourism Satellite Account (TSA) .................................................................. 136 A.3.1.3 TSA and IOT ......................................................................................................... 138 A.3.1.3.1 Supply Table ................................................................................................... 144 iv A.3.1.3.1.1 Production block ..................................................................................... 144 A.3.1.3.1.2 Imports .................................................................................................... 147 A.3.1.3.2 Use Table ....................................................................................................... 148 A.3.1.3.2.1 Intermediate demand .............................................................................. 148 A.3.1.3.2.2 Final demand .......................................................................................... 151 A.3.1.3.2.3 Gross Capital Formation ......................................................................... 159 A.3.1.3.2.4 Exports .................................................................................................... 160 A.3.1.3.2.5 Mathematical adjustment ................................................................................ 161 A.3.1.4 New IOT: comparison and conclusion ................................................................. 168 A.3.1.5 SAM ...................................................................................................................... 169 A.3.1.5.1. Allocation of primary income account: ........................................................ 173 A.3.1.5.2 Secondary distribution of income account: ................................................... 183 A.3.1.5.3 Use of disposable income account: ............................................................... 188 A.3.1.5.4 Capital account .............................................................................................. 189 A.3.1.5.5 Capital ROW account .................................................................................... 195 A.3.1.6 Supply and use table with tourism categories....................................................... 198 Annex 3.2 Recursive-dynamic CGE model ...................................................................... 207 A.3.2.1 Zero profit conditions ....................................................................................... 207 A.3.2.2. Market clearance conditions ............................................................................ 215 A.3.2.3. Income balance (budget constraint) ................................................................. 222 Turismo: Aerolíneas de bajo coste, clima y crisis económica ........................................... 225 Annex. Codes .................................................................................................................... 293 References ......................................................................................................................... 301 v List of figures Figure 1. Effect of the variables p aand p al ..................................................................... 13 Figure 2. Tourism arrivals and expenditure ........................................................................ 43 Figure 3. Probabilities of the economising strategies ......................................................... 61 Figure 4. Final effects on the destination ............................................................................ 63 Figure 5. Moving median probability of economising strategies by climate ...................... 65 Figure 6. Moving median probability of economising strategies by age ............................ 66 Figure 7. Real GDP growth rate (2008=100) and employment growth rate by sectors (%) 72 Figure 8. Unitary labour cost and unemployment rate ........................................................ 73 Figure 9. Inbound tourism in Spain from 2002 to 2012 ...................................................... 75 Figure 10. Unemployment (minimum wage) ...................................................................... 84 Figure 11. Unemployment (wage curve) ............................................................................. 85 Figure 12. TNT-T model ..................................................................................................... 88 Figure 13. TNT-T model with unemployment and capital accumulation ........................... 89 Figure 14. General structure of the blovifis model. ............................................................ 91 Figure 15. Unemployment rate with change in employment elasticity’s respect to the BAU situation (%) ........................................................................................................................ 97 Figure 16. Change in gross real added value, current account deficit and capital accumulation respect to the BAU situation (%) ................................................................ 100 Figure 17. Changes in domestic demand respect to the BAU situation (%) ..................... 102 Figure 18. Change in the real exchange rate (%) .............................................................. 104 Figure 19. Scheme of the datasets .................................................................................... 132 Figure 20. Disentangling AISS values into the SAM ...................................................... 171 xii ellos recuperé la vocación por la economía. Han sido dos grandes amigos y excelentes maestros. Por ultimo, y no por ello menos importante, quiero agradecer a Casiano Manrique por su ideas y desinteresa supervisión en la elaboración de la base de datos del capítulo tres de esta tesis así como sus consejos y supervisión en lo referente al modelo de equilibrio general. En Las Palmas de Gran Canaria a 8 de Junio de 2014. “A continuación te dejo con mi yo, conmigo, con el que fui y ahora está contigo.” Federico Inchausti Sintes. xiii 1 CHAPTER 1 Does low-cost travelling imply higher tourism expenditure in the destination? 1.1 Introduction One key for success of tourism sector as an economic growth generator is the capacity to provide added value. Amongst some other aspects, tourism expenditure is central to measure gross added value of tourism destinations. However, tourism expenditure is not only disbursed in the destination but also in the country of residence. Such decomposition is not trivial in terms of added value. For instance, tour operators located in origin are an open door for channeling tourists, but at the same time, they also detract, as deserved, part of the potential added value of the destination. Generally speaking, the result of the negotiations between tour operators and hoteliers determine the share of the added value between the origin country and the destination. Arguments against the tour operators’ empowerment are usually stated by hoteliers and local government. Nevertheless, tourism market structure has changed and it keeps changing dramatically. Traditionally, most tourists opted for comprehensive packages which were paid in travel agencies. The advent of Internet has shortened the ‘distance’ between origins and destinations. It has opened up new alternatives to the tourists, allowing for more customized services. There has been a shift towards decomposing tourism packages, such that travel, accommodation, meals, or excursions can be booked separately. Under this new market structure, tourism service products can be distributed either by direct sales on the internet or by cheaper internet intermediaries. 2 However, it should be noted that its success depends on tourists’ confidence on the system. Production costs of tourist products arranged on the internet are likely to decrease. Such efficiency gain implies lower prices and or higher profits depending on market competition. In any case, lower prices increase consumer surplus and higher profits increase producer surplus, so that social welfare increases. Additionally, a lower price implies an increase in the number of tourists, even when such price decrease is homogeneous for all destinations because it can generate additional traffic from tourists who, under lower prices, can afford travelling. Hence, added value at the destination is expected to increase either due to a higher number of tourists or due to higher profits. Controversial discussions have arisen concerning the convenience of the new market structure. In particular, special attention has focused on the presence of low-cost carriers (LCCs), which have boosted recently due to the new situation. Tourism destination policymakers wonder about the consequences for the whole market and the best strategy to deal with it. The consequences are multiple. First, the presence of LCCs may attract new tourists to the destination because they may afford travelling at lower prices. Such competitiveness gain is more or less effective depending on how alternative destinations are also dealing with it. Second, airline market is also affected by LCC entrance. Flagship companies are likely to lose market share and they may even stop flying to the destination at all. All this may affect the share of the profile of the tourists at the destination. Such market share redistribution has an impact on tourism expenditure in the destination. Third, tourists usually face two kinds of constraints for holiday taking. On the one hand, the number of available days for tourism is limited. Even if a tourist can afford paying for a three months holidays, he or she also faces a time constraint. On the other hand, tourists may spend a limited amount of money on holidays, which represents a tourism budget constraint. Both constraints are key elements to understand 3 tourism destination choice. The tourism budget constraint is distributed between travelling, accommodation, meals and other expenditure. It is interesting to explore how the presence of LCCs may contribute to a redistribution of such budget. Household savings from cheaper travel tickets may be transferred, fully or partially to higher tourism expenditure in the destination. Testing and quantifying this hypothesis is the purpose of this paper: Hypothesis: Low-cost travelling savings from the origin are transferred, at least partially, to higher tourism expenditure in the destination. Testing this hypothesis is relevant to understand one key impact of the presence of LCCs in tourism markets. Quantifying such impact is relevant for policymaking, especially to understand the degree of support that LCCs should receive by destinations. Current literature has focused on the traffic generated by LCCs and the market share redistribution, but it has not dealt with added value redistribution between origin and destination. This paper explores such relationship. The dichotomy between expenditure in origin and destination and their reciprocal relations and causality permits to analyze these hypothetical situations fostered by LCCs. One methodology that is able to estimate this relationship is a Simultaneous System of Equations. Amongst some alternative models considered, the one estimated by Three Stages Least Squares (3SLS) (Zellner and Theil, 1962) is chosen. 4 1.2 State of the art 1.2.1. LCCs characteristics According to the Civil Aviation Authority in UK (CAA, 2006), the words “low cost” should be kept to charter carriers; and use the words “no-frills” instead of “low cost” to what public opinion define as a “low cost” carrier. However, in this paper, the term “low cost” is used in its popular sense. According to Lawton (2002) and Doganis (2006), the characteristics of LCCs can be described from a flyer and operational point of view. From a flyer perspective, LCCs offer single-class service, with high density seating and food or beverages payable on board. On the ground, LCCs do not provide special check-in services or frequent flyers programs. From an operational point of view, LCCs operate with a single aircraft type which reduces maintenance and pilot training. They also operate with fast turnaround times, so that it increases aircraft utilization. Sorenson (1991) and Caves, Christensen and Tretheway (1984) link this last issue to the achievement of economies of density in LCCs. They usually operate in secondary airports allowing quicker operations and lower airport fees. LCCs are focused on shorter routes to maximize the number of trips in both directions. They sell tickets directly to customers and they do not offer connection flights. All the aspects mentioned above are focused on reducing cost and thus offering lower ticket prices. Literature coincides in supporting this issue. Generally speaking, Malighetti, Paleari and Redondi (2009), Fu, Dresner and Oum (2011), Ben Abda, Belobaba and Swelbar (2012), Rosselló and Riera (2012), Alderighi, Cento and Piga (2011) or Alderighi, Cento, Nijkamp and Rietveld (2012) agree that, for their respective cases, the entry of LCCs has decreased ticket prices. Windle and Dresner (1999) provide a further literature review on this issue for the case of The United States of America during the nineties. 5 1.2.2. Main consequences of the presence of LCCs in a destination The literature in regards to LCCs has been more focused on the transportation sector. Nonetheless, and at least from a tourism sector perspective, some questions mutually related bloom as soon as it gets LCCs into consideration: Have the presence of LCCs increased the flow of tourism to a destination? Have LCCS passengers’ got different preferences/profiles with respect to the traditional carrier passengers? Are LCCs passengers’ savings in the origin transferred to higher tourism expenditure at the destination? Has LCCs’ presence promoted economic growth in the region they travel to? The first two of these research inquiries have been considered in the literature but the last two have not been explored sufficiently yet. 1.2.3. Have LCCs increased the flow of tourism? The presence of LCCs traveling to destinations may increase the flow of tourists. However its success depends on many factors. Some key determinants are the coexistence of similar routes, the connectivity of the destination airport with the tourist destination, the behavior of competing destinations in regards to the presence of LCCs and the sensitivity of the passengers to lower fares together with passengers’ willingness to accept LCCs service quality. Obviously, the answer to this question varies with each case study. For this reason, it is not surprising that the literature shows a wide range of results in this sense. The Civil Aviation Authority (2006) conducted a report concerning LCCs in the UK. It concludes that there is not plausible evidence of an increase in the flow of passengers due to LCCs beyond the natural stationary growth in the sector. However, the report shows little evidence of an increase in the traffic flow of some routes in comparison with the usual traffic flow of these routes. However, in general, the report concludes that LCCs have succeeded in increasing its market share rather than increasing new passengers flow. Young and Whang (2011) differ from the previous report and affirm that LCCs stimulated new demand to the tourist island of Jeju in South Korea. Graham and Dennis (2010) remark that the flow of 6 tourists to Malta has increased due to LCCs. Rey, Myro and Galera (2011) state that, on average, a 10% increase in the number of visitors traveling with LCCs, increases the average number of tourists traveling from EU-15 countries to Spain by a 0.2%. According to Davison and Ryley (2010), LCCs have increased the demand for short breaks from regional airports such as East Midlands region in the UK to cultural destinations such as Prague or Berlin, whereas destinations such as Faro or Alicante remain as week-long holidays. In the case of Australia, Forsyth (2003) concludes that the entrance of LCCs have had little impact in the transport sector and thus, in the flow of tourists. Finally, Pulina and Cortés-Jiménez (2010) conclude that LCC carriers have boosted the tourism demand in Alghero (Italy) in the last decade. 1.2.4. Have LCC passengers got a different behaviour with respect to traditional carrier passengers? After any LCC entrance at a destination, according to their behavior, there may be three sets of tourists: a) new tourists who fly due to the presence of LCCs, b) current tourists who are willing to accept the trade-off between lower prices and new air transport service quality and c) current tourists who keep booking with non-LCCs. Such different behavior may also be correlated with their budget constraint and it may have an impact on their tourism expenditure at the destination. O´Connell and Williams (2005), Mason (2005) and; Graham and Dennis (2010) use a descriptive analysis in their papers. O´Connell and Williams (2005) affirm that LCC passengers focus their decision on price. Whereas, traditional carrier passengers take into account a wider set of attributes to make their decision such as reliability, quality, flight schedules, connections, frequent flyer programmes and comfort. Mason (2005) and MartínezGarcia, Ferrer-Rosell and Coenders (2012) analyse the demand of business travelers and leisure travelers. According to Mason (2005), the advent of internet and low cost airlines are the main factors behind the change in demand of these two travelers profile. On the one hand, 7 leisure travelers are taking holidays more frequently but with shorter stays. On the other hand, business travelers are also shifting towards LCCs, especially in short-haul route. According to Donzelli (2010) LCCs are reducing the seasonality in Southern Italy. In the case of Malta, Graham and Dennis (2010) state that LCCs tourist preferences seem different from non-LCCs ones. Especially, they point out that LCC tourists do not show so much interest in the cultural heritage, which represent the traditional tourist attractions of the islands. They also conclude that LCCs operate on some routes to the island with the same frequency as the preexisting companies. Additionally, LCCs may have provided new opportunities to travel during off-peak periods or to undertake short-breaks holidays. Again, the nature of the destination and its dependence on the climatic conditions for attracting tourists make a difference on this issue. Thus, different answers are expected to be obtained for different destinations. Young and Whang (2011) use a time series regression analysis with tourism demand as a dependent variable. They conclude that LCCs have no influence in changing seasonal pattern. According to them, LCCs have just overtaken the preexisting schedule flights to the island. On the contrary, Pulina and CortésJiménez (2010) analyze the relationship between tourism demand and supply in Alghero (Italy). They state that LCCs have changed the seasonal pattern of foreign tourists whereas national tourists (Italians) have not changed their preferences and they keep travelling to the island in August, mainly. They also conclude that LCCs have boosted the tourism demand to the island in the last decade. 1.2.5. Have LCCs presence promoted economic growth in the region where they travel to? The literature in regard to tourism and economic growth is diverse, but it lays on any of the tourism-led growth hypotheses (Balassa, 1978). The literature about LCCs and economic growth does not follow any of these hypotheses, and the approach is more general and scattered. Moreover, the advent of LCCs is a recent phenomenon and probably, the time 14 1.3.4. Theoretical microeconomic motivation A household minimizes its tourist expenditure per night and person subject to a given utility and according to a given destination, nights and party size. The dual minimizing problem may be represented as: oo dd M in PQ P Q  .:st  1/ oo dd QQ U     (CES utility function) ,, 1 n o o oj oj j PQ P Q   and ,, 1 n d d dj dj j PQ P Q  ; and represent the total tourist expenditure per night and person at origin and the total tourist expenditure at destination, respectively. More precisely, ,oj P and ,oj Qare the prices and quantities of goods/services j at origin o per night and person. In the same manner, ,dj P and ,dj Q are the prices and quantities of the goods/services j at destination d per night and person. 1      and  is the elasticity of substitution between the expenditure in origin and destination per night and person. o  and d  represents the share of utility provided by tourism expenditure in origin and destination per night and person. A higher share in destination implies a higher preference towards expenditure in destination with respect to the origin. The solution to this problem yields the quantities consumed in origin and destination per night and person. The marginal rate of substitution (MRS) is expressed as follows: 1 ooo ddd PQ PQ        15 Rearranging o Qand d Q, multiplying by o P and d Pon both sides and substituting  , the MRS between expenditure in origin and destination (per night and person) is obtained: 1 dd oo dd oo P PQ PQ P        Let 1 dd oo P P         , so that oo dd PQ PQ  . Thus  represents the share between tourism expenditure in origin with respect to tourism expenditure in destination. It is a key parameter for the understanding of this issue, as it is explained below. Finally, it should be noted that for elasticities  lower than one, an increase in the prices in destination decreases the expenditure in origin per night and person. For elasticities  higher than one, an increase in the prices in destination increases the expenditure in origin per night and person. 1.3.5. Econometric approach to the microeconomic motivation Expenditure in origin and destination are functions of exogenous explanatory variables ( n Z ), endogenous explanatory variables ( o E,d E, where ooo EPQ  and ddd EPQ  ) and error terms (o  , d  ) that gather up all the information not included in the explanatory variables. (,,) ondo EfZE   (,,) dnod EfZE   0onndo n EZE       (1) 0dnnod n EZE       (2) Where 0  ,  ,  , 0  ,  ,  are parameters associated to each explanatory variable. 16 Given that od EE and substituting (1) and (2): 00 ii o d on n EZ                (3) 00 ii do dn n EZ                 (4) An important theoretical result is that od       in (3) and do      in (4) show the presence of error correlation between the expenditure in origin and expenditure in destination for a given household. Such error correlation needs to be tackled with simultaneous equations. For the purpose of this research, pa and pal multiplicative dummies also need to be part of the model. Thus, the equations can be specified as follows: ,0 , , , ,io p ip e id q iq s is i pqs EZEPaPalu           (5) ,,,,,id d z iz e io q iq s is i pqs EZEPaPale          (6) where subscript i denotes households or individuals and the assumptions taken into account are the following ones: 2 () i Var u   ; ()0 ij Var u u  (a1) Homokedasticity and non-autocorrelation in equation (5). 2 () i Var e   ; ()0 ij Var ee  (a2) Homokedasticity and non-autocorrelation in equation (6). () ii i Eue   (a3) Contemporary correlations between errors in equations (5) and equations (6). ()0 ii EuZ  (a4) Exogenous explanatory variables in equations (5) are 17 1.4 Case study The methodology is applied to the Canary Islands (Spain), which is an ideal destination for tourism research because arrivals and departures are well documented since air travelling is the main mean of transportation to the islands. Additionally, Instituto Canario de Estadística (ISTAC) provides a very good set of surveys that describe the tourism sector appropriately. Every term, ISTAC conducts a large Tourist Expenditure Survey, which is the basis of the dataset used in this paper. Canary Islands emerged as a tourism destination in the sixties of the last century. Since then, tourism sector has increased its relevance in the economy providing economic development along the way. Nowadays, Canary Islands is a well-known tourism destination that attracts about 12 million tourists every year. Traditionally, most tourists arrived in charter flights. They used to book their holidays in travel agencies in origin. Such bookings comprised the whole package, i.e. flight, accommodation, transits and excursions in some cases. Some other tourists were coming with scheduled flights, but it was not until 2002 when independent traveling with low cost emerged strongly. According to ISTAC time series data, low cost traveling to Canary Islands has increased during the last ten years. For instance, in 2006Q1, non-correlated with the error term. ()0 ii EeZ  (a5) Exogenous explanatory variables in equations (6) are non-correlated with the error term. , ()0 ioi EuE  (a6) Endogenous explanatory variable ( ,oi E) correlated with the error term in equation (5). , ()0 idi EeE  (a7) Endogenous explanatory variable ( ,di E ) correlated with the error term in equation (6). 18 the share of low cost traveling represented 19.88%. In 2012Q1, it reached 29.50% and in 2014Q1, it increased up to 38.53%. These figures reveal that the presence of LCCs represents an important share of the current market that seems to keep growing. It proves that the market structure has changed and it keeps changing. Dataset The period chosen for the dataset starts in 2009 and it finishes in the second term of 2011. The survey is a cross section study that includes questions related with expenditure in origin and destination, socio-economic attributes, motivations in choosing Canary Islands, impression about the holidays, length of stay or previous visits to the islands, among other variables that are explained below. It should be noted that not all the passengers that travel to the Canary Islands are ‘true’ tourists, because some of them are foreigners that reside in the islands. In order to avoid potential biases in terms of the length of stay or expenditure in the destination, only passengers who stay a maximum of thirty one nights are finally considered. Thus, the dataset is comprised of 53,608 observations. Variables During the research of this paper, many variables and alternative specification models were considered. Final endogenous and exogenous variables that are estimated in the model are shown below: Endogenous variables: Exporigin (Expenditure in origin per person and night), expdestination (Expenditure in destination per person and night). Exogenous variables: income (yearly income divided by 12 months), term (term), year (year), p (kind of tourist package: flight, flight + accommodation, flight + accommodation + Breakfast, flight + accommodation + half board, flight + accommodation + full board + flight 19 + accommodation + all inclusive), a (category of accommodation: 5* hotel, 4* hotel, 3*, 2* or 1* hotel, apartment, house of friends or relatives, others (e.g. timesharing), pa (multiplicative dummy between package and category of accommodation), pal (pa multiplied by a low cost dummy), destination (island visited: La Palma, El Hierro, Tenerife, Gran Canaria, Fuerteventura and Lanzarote), party (it has got members with age lower than 2 years, between 2 and 12, between 13 and 65 years, older than 65 years), people (alone, with couple, with family, friends and relative or coworkers), motivation (main reason to travel to Canary Islands: climate, beaches, landscape, environmental quality, quietness, active tourism, health tourism, theme park, golf, other sports, nightlife, shopping, new place, ease of traveling, prices, for kids), and previous visits (from 1 to more than 10 times). 1.5 Methodology Simultaneous equations model A simultaneous equations model provides a suitable framework to model the dichotomy and mutual relationship between expenditure in origin and expenditure in destination. However, some tests have to be taken into account in order to accept this model and its estimation as a suitable characterization of the hypothesis outlined in this paper. The first question to answer is the presence of simultaneity (Gujarati, 2003); additionally, it is required to fulfill the identification condition or full rank, so that the number of exogenous variables (K) in the system minus the number of exogeneous variables in equation (k) need to be greater or equal than the number of endogeneous variables in a equation minus one: K - k > = m - 1. Thus, the presence of simultaneity depends on the acceptance of two of the hypotheses shown above: endogeneity ((a6) and (a7)) and contemporary correlations among error terms (a3). 20 1.5.1. Endogeneity Durbin-Wu-Hausmann test is conducted to test the presence of endogeneity between both endogenous variables, since they also belong to the explanatory variables in the other equation. In both cases, the existence of endogeneity is not rejected. Thus the econometric method needs to deal with endogeneity. Some methods that are able for this purpose are: IOLS (Indirect Ordinary Least Squares (exactly identified equations)), IV (Instrumental Variables) or 2SLS (Two-Stages Ordinary Least Squares (overidentified equations)). These methods belong to the family of the limited information methods because each equation is estimated independently of the others as there is not relationship among them. 1.5.2. Contemporary correlations among error terms The second step is to check contemporary correlation between the error terms of the two equations. Two approaches are calculated. First, equations (5) and (6) are estimated by OLS but considering only all the exogenous variables of their respective equations. Second, the correlation of the residuals of the two equations are calculated (correlation=0.1537). An alternative to this process might be to estimate a SUR model, where the endogenous variables are excluded as explanatory variables from each other equations as in the first approach. After that, a correlation among residuals and Breusch–Pagan test of independence are calculated. The correlation is -0.1537 and Breusch-Pagan test is not rejected. Under the presence of contemporary correlations, SUR estimation is able to deal with this problem. This estimation belongs to the family of the complete information method and it takes into account the simultaneity. The acceptance of endogeneity and error correlations between the two equations support the suitability of this methodology to treat with the hypothesis outlined in this paper. The econometric method able to estimate this kind of model is 3SLS (Three-Stages Last Squares) (Zellner and Theil, 1962). 3SLS gathers up 2SLS (endogeneity) and SUR 21 (contemporary correlations among equations). Nonetheless, there is another issue to deal with, i.e. heteroskedasticity. 1.5.3. Heteroskedasticity Such issue generally affects the efficiency of the estimator and thus the individual significance of the estimates. Firstly, both equations are estimated separately by OLS. Secondly, Breusch-Pagan/Cook-Weisberg test for heteroskedasticity is applied on the residuals of equations (5) and (6). The null hypothesis (homoskedasticity) is rejected. 1.5.4. Estimation under the presence of endogeneity, contemporary correlation error terms and heteroskedasticity. Endogeneity, contemporary correlation and heteroskedasticity can be treated by the generalized method of moments (GMM). GMM (Hayashi, 2000) can be seen as a general framework in which any other estimator is a particular case of it (either of the two methods: limited information or completed information can be modeled within a GMM context). One advantage is that it permits robust estimation under the presence of heteroskedasticity, which implies efficiency gains (Greene, 1997). A disadvantage is that it requires as many instruments as equation moments and each explanatory variable requires one equation moment. However, as Cameron and Trivedi (2005) remark, the exogenous variables can be instrumented by themselves. In regards to the system shown in equations (5) and (6), a robust 3SLS estimation is carried out but it did not achieve a solution due to the non-positive semidefinite residual covariance matrix. The exclusion of some dummy variables permits robust 3SLS GMM estimation but, at the same time, it produces a misspecification problem. At this time, a non-robust 3SLS estimation is chosen. The 3SLS satisfies the requirement for an IV estimator and thus it is consistent. Additionally, for normally distributed disturbances, the 3SLS estimator is asymptotically 22 efficient (Greene, 1997). In 3SLS, the first two steps are equal to 2SLS (deal with endogeneity), the last step calculates the variance and covariance matrix of the residuals from step 2 and re-estimates the equations by SUR (deal with correlations among residuals in both equations). On the one hand, the 2SLS estimation will be less efficient under the presence of contemporary correlation of the error terms of each equation (Schmidt, 2005). On the other hand, the SUR estimation will be biased and inconsistent (Schmidt, 2005), under the presence of endogeneity. The non-resolved question of heteroskedasticity produces an efficiency loss in the 3SLS estimation. In contrast, 3SLS permits a good specification model in regards to the robust 3SLS GMM estimation. The non-presence of some explanatory variables in some equations (excluded variables) such as motivation in the expenditure in origin equation or years in the expenditure in destination equation help to avoid the identification problem. The first equation is just identified and the second one is over identified. Summarizing the hypotheses: (a1) and (a2) are not held. From (a1) to (a5) are necessary conditions for SUR estimations. These assumptions plus assumptions (a6) and (a7) are necessary conditions to apply 3SLS estimation. 1.5.5. The reduced form This form expresses the endogenous variables as a function of exogenous explanatory variables. This form has three important implications in a simultaneous equations model: it allows for the identification of the models (alternatively to the condition: K - k > = m - 1, already explained), estimators such as 3SLS or IOLS, among others, use the reduced form to figure out the system and, it permits to evaluate the direct impact of any exogenous explanatory variable in any endogenous variable. The equations model in reduced form are the following ones: 23 4 12 2 11 13 3 11 1 2011 31 14 11 15 5 11 2009 1 31 16 6 11 17 7 11 1 [( [( ) / (1 )] [( ) / (1 )] [( ) / (1 )] [( ) / (1 )] [( ) / (1 )] )/(1 )] iittit t yiyhhih yh hh ih dd h exporigin income term year pa pal d                                7 1 19 4 18 8 11 19 9 11 11 516 110 10 11 111 11 11 12 11 111 [( ) / (1 )] [( / (1 )] [( / (1 )] [ (1 )] [ ) )/ [( ) / (1 )] id d cc ic rr ir dr rr io m im om i estination country party people motivation previous                            111 11 /(1 )] [1/(1 )] ii eu    (7) 4 12 2 11 13 3 11 1 2011 31 14 11 15 5 11 2009 1 31 16 6 11 17 7 1 1 [( [( ) / (1 )] [( ) / (1 )] [( ) / (1 )] [( ) / (1 )] [( ) / (1 )/(1 )] iittit t yiyhhih yh hh ih dd h income term year pa pal expdestination                               7 1 1 19 4 18 8 11 19 9 11 11 516 110 10 11 11 11 11 11 12 111 )] [( ) / (1 )] [( / (1 )] [( / (1 )] [ (1 )] ) )/ [( ) / (1 )] [ id d dd ic rr ir dr rr io m im om i destination country party people motivation previous                           111 11 [1/(1 )] /(1 )] ii ue    (8) Equations (7) and (8) are the origin and destination equations in reduced form, respectively. 1.6 Results 1.6.1. Reweighting effect This study has considered six different kinds of accommodation, i.e. five* hotel; four* hotel; three, two or one star hotel; apartment; friends or relatives accommodation; and other. It has also taken into account six different kinds of package holidays, i.e. only flight; flight, accommodation and self-catering; flight, accommodation and breakfast; flight, 30 Table 3 (continues). Structural form results of 3SLS estimation (part I) Origin equation Destination equation Parameter Std. desv Parameter Std. desv Party composition Lower than 2 years -9.590*** (1.018) -7.047*** (1.244) 2 and 12 years -2.203*** (0.425) -5.113*** (0.463) 13 and 65 years 1.109*** (0.336) -4.725*** (0.201) Older than 65 years 0.599 (0.560) -7.466*** (0.418) Couple -10.341*** (0.501) 0.202 (0.823) Family -1.947*** (0.632) 1.201*** (0.464) Friends -10.488*** (0.620) 2.227*** (0.760) Coworkers 20.195*** (1.980) 3.939* (2.026) Main motivation Climate - - -3.066*** (0.621) Beaches - - -1.702*** (0.404) Landscape - - -1.345*** (0.375) Environmental quality - - -0.931* (0.507) Quietness - - -2.067*** (0.283) Active tourism - - -4.173*** (0.606) Helth tourism - - 2.825** (1.116) Theme park - - 1.643** (0.765) *** p<0.01, **p<0.05, *p<0.10 31 Table 4 shows the estimates of the dummy variables that are used to identify each tourist profile. These estimates belong to the same estimation of Table 3 but they are shown in a separate table for the ease of presentation. As shown in Figure 1, it is important to distinguish between Non-LCC and LCC estimates. The first ones correspond to a shift from the benchmark and 60 out of 62 estimates are significant, which prove the relevance of such distinction. However, LCC dummies are an additional shift from non-LCC shift. Their significance is critical, because it tests if LCC travelers spend differently than Non-LCC travelers, and hence they test the hypothesis enquired in this paper. The table shows that 35 out of 62 estimates are significant, which means that LCC travelers for these combinations of accommodation and food regime are different than Non-LCC travelers. Nevertheless, the results from the structural form cannot be used to measure the direct impact of each dummy, Table 3 (continues). Structural form results of 3SLS estimation (part I) Origin equation Destination equation Parameter Std. desv Parameter Std. desv Main motivation Theme park - - 1.643** (0.765) Golf - - 9.846*** (1.014) Other sports - - -1.840* (1.023) Nightlife - - 4.274*** (0.633) Shopping - - 5.169*** (0.628) New place - - -1.311*** (0.454) Ease of traveling - - -2.916*** (0.409) Price -4.663*** (0.888) For kids -1.893*** (0.529) Observations 53,608 53,608 R2 0.842 0.718 Chi2 2.89e+05 1.37e+05 *** p<0.01, **p<0.05, *p<0.10 32 but to test direction of the impact and significance. In order to measure the impact, it is necessary to obtain these results by the reduced form as shown in Table 5. Table 4. Structural form results of 3SLS estimation (Part II) Non-LCC Origin equation Destination equation Parameter Std. desv Parameter Std. desv Flight only 5 stars hotel -7.606 (8.862) 130.161*** (6.950) 4 stars hotel 14.800** (6.825) 90.253*** (6.651) 3,2 or 1 stars hotel 17.171*** (6.458) 73.906*** (6.276) Apartment 16.103*** (5.680) 65.531*** (5.709) Friends and family 25.071*** (4.992) 44.614*** (5.398) Other 27.826*** (5.406) 49.915*** (5.872) Flight and Accommodation 5 stars hotel 102.242*** (7.306) 72.965*** (12.194) 4 stars hotel 75.679*** (5.402) 47.179*** (9.028) 3,2 or 1 stars hotel 56.268*** (5.353) 49.695*** (7.786) Apartment 53.837*** (5.216) 48.885*** (7.576) Other 51.411*** (5.431) 46.114*** (7.297) Flight, accommodation and breakfast 5 stars hotel 114.416*** (6.140) 57.306*** (12.255) 4 stars hotel 89.794*** (5.544) 49.033*** (10.086) 3,2 or 1 stars hotel 70.638*** (5.504) 47.437*** (8.704) Apartment 66.021*** (5.388) 46.611*** (8.391) Other 83.573*** (13.895) 30.455** (12.910) *** p<0.01, **p<0.05, *p<0.10 33 Table 4 (continues). Structural form results of 3SLS estimation (Part II) Non-LCC Origin equation Destination equation Parameter Std. desv Parameter Std. desv Flight, accommodation and half board 5 stars hotel 127.901*** (5.373) 35.821*** (12.315) 4 stars hotel 90.219*** (5.014) 35.368*** (9.572) 3,2 or 1 stars hotel 71.111*** (5.075) 38.490*** (8.404) Apartment 72.649*** (5.097) 37.799*** (8.478) Other 77.778*** (9.290) 38.801*** (10.334) Flight, accommodation and full board 5 stars hotel 129.789*** (5.914) 29.751** (12.363) 4 stars hotel 92.318*** (5.056) 31.847*** (9.567) 3,2 or 1 stars hotel 80.284*** (5.354) 33.219*** (8.897) Apartment 71.242*** (5.676) 34.092*** (8.427) Other 129.932*** (5.687) 22.668* (12.160) Flight, accommodation and all inclusive 5 stars hotel 125.542*** (5.194) 21.548* (11.597) 4 stars hotel 99.646*** (4.724) 22.534** (9.727) 3,2 or 1 stars hotel 89.705*** (4.792) 25.560*** 9.171) Apartment 82.958*** (4.923) 29.942*** (8.891) Other 136.551*** (5.551) 14.949 (12.291) *** p<0.01, **p<0.05, *p<0.10 34 Table 4 (continues). Structural form results of 3SLS estimation (Part II) LCC Origin equation Destination equation Parameter Std. desv Parameter Std. desv Flight only 5 stars hotel 1.156 (5.086) -11.154*** (3.652) 4 stars hotel -1.386 (3.192) -12.686*** (2.316) 3,2 or 1 stars hotel 3.858 (3.731) -8.354*** (2.664) Apartment -2.086 (1.908) -1.446 (1.379) Friends and family -4.040*** (1.239) 2.127** (0.907) Other -4.443* (2.559) 1.667 (1.850) Flight and Accommodation 5 stars hotel -0.897 (5.161) -16.616*** (3.726) 4 stars hotel -8.813*** (2.112) 0.878 (1.623) 3,2 or 1 stars hotel -9.360*** (1.782) 2.416* (1.380) Apartment -7.815*** (0.721) 1.704** (0.690) Other -3.971 (2.607) 1.119 (1.888) Flight, accommodation and breakfast 5 stars hotel -10.057*** (2.732) 2.299 (2.054) 4 stars hotel -10.735*** (2.151) 1.717 (1.701) 3,2 or 1 stars hotel -6.020* (3.226) 5.539** (2.318) Apartment -10.409*** (2.749) -0.433 (2.117) Other -24.301 (16.176) 24.828** (11.634) *** p<0.01, **p<0.05, *p<0.10 35 1.6.3. Results from the reduced form Reduced form results are a convenient transformation from the structural form results that deal with the system iterations in order to reveal the “true” direct impact of each exogenous variable. Such transformation is applied to key dummies presented in Table 4 and shown in Table 5. Income elasticities can be obtained from the reduced form. In particular, income elasticity with respect to expenditure in origin is 1.74, whereas in destination such elasticity is 1.98. Table 5 shows how much more or less each LCC profile is spending in origin and destination. This result is weighted by mean nights and mean party size in order to obtain a figure closer Table 4 (continues). Structural form results of 3SLS estimation (Part II) LCC Origin equation Destination equation Parameter Std. desv Parameter Std. desv Flight, accommodation and half board 5 stars hotel -9.156*** (2.639) 1.914 (1.981) 4 stars hotel -8.990*** (1.085) 3.986*** (0.909) 3,2 or 1 stars hotel -2.623 (2.001) 0.427 (1.444) Apartment -9.585*** (2.265) 5.225*** (1.681) Other -3.286 (11.397) -8.374 (8.212) Flight, accommodation and full board 5 stars hotel -14.699** (7.224) 10.954** (5.222) 4 stars hotel -3.556 (2.693) -0.034 (1.953) 3,2 or 1 stars hotel -2.614 (4.199) -0.471 (3.028) Apartment -4.008 (4.300) 7.506** (3.083) Other -8.843 (7.152) 5.843 (5.171) Flight, accommodation and all inclusive 5 stars hotel -10.666*** (3.727) 0.679 (2.770) 4 stars hotel -8.748*** (1.036) 3.123*** (0.873) 3,2 or 1 stars hotel -13.828*** (1.557) 4.150*** (1.370) Apartment -14.490*** (1.773) 5.059*** (1.491) Other -20.593*** (5.716) 7.985* (4.276) *** p<0.01, **p<0.05, *p<0.10 36 to the one faced by each tourist. For the four most relevant profiles the results are similar. All of them save money in origin with respect to Non-LCC tourists. In particular, saving figures vary between 179.97 euros and 70.97 euros per mean party and nights. The key enquiry of this paper is to test if such savings are transferred into higher expenditure in the destination. It proves that for the most popular profiles LCC tourists spend more money than Non-LCC tourists. Such higher expenditure varies between 48.11 euros and 7.52 euros per mean party and nights. This figure proves the hypothesis that LCC tourists’ savings in origin are transferred, at least partially, as higher expenditure in the destination. 1.6.4. Savings transfer ratios It is interesting to calculate the percentage of savings in the origin that is transferred as additional expenditure in the destination. Out of the four most relevant profiles, the one of tourists who book only the flight and stay in friends or relatives accommodation transfer 49.8%, which represents the highest transfer value. On the contrary, tourists who book flight Table 5. Reduced form results: The impact of LCC with respect to non-LCC by mean nights and mean party size (euros) OriginDestination 5 * 4 * 3,2,1* Apartment Family or friends Other Only flight (F) O D -92.36 -188.27* -187.05 -274.78* -16.46 -52.78 -84.81 -52.78 -70.97* 35.38* -99.72* 24.74 (F) + Accommodation (A) O D -196.67 -310.80* -167.03* -14.54 -173.60* 15.64* -179.97* 7.52* -106.30 12.76 (F+A) + Breakfast O D -145.82* 7.91 -159.24* -3.81 -40.22* 64.31* -208.11* -45.35 -195.27 407.73* (F+A)+Half Board O D -169.31* 5.30 -136.13* 48.11* -47.31 -0.99 -146.99* 78.45* -168.91 -185.13 (F+A)+Full Board O D -138.67* 138.56* -66.16 -12.58 -54.47 -17.83 9.38 165.83* -112.71 88.01 (F+A)+ All-inclusive O D 202.91* -24.86 -152.33* 33.78* -248.29* 35.68* -285.26* 60.04* -362.25* 96.59* * Means that the original dummy variable from Table 4 is significant Bold font means top four most relevant profile 37 and apartment with self-catering transfer only 4.1% of their savings. Tourists who stay in four* hotel with half board transfer 35.3%, whereas those who stay in four* hotel with allinclusive transfer 22.1%. Thus, it is clear that despite the hypothesis is true, there is net savings, so that not all the savings are transferred as additional expenditure in the destination. 1.7 Conclusions and further research Canary Islands as many other tourism destinations around the world have faced a relevant market structure change. Tourists are traveling more often with LCC airlines and it has an impact on the destination. On the one hand, LCC tourists’ perception of saving money with cheaper air fares may encourage them to spend more money in the destination. On the other hand, LCC airlines may increase air traffic towards a particular destination. This paper tests if the former hypothesis is true. For that purpose, a system of equations of expenditure in origin and expenditure in the destination is considered. Within all econometric methods that may estimate such system, 3SLS model is chosen because it is able to deal with endogeneity and contemporary error correlation appropriately. Another issue for the destination is related with a redistribution of the relevance of each tourist profile due to the presence of LCCs. LCC travelers may be willing to stay on different kind of accommodation or to enjoy simple meal packages with respect to traditional NonLCC travelers. Hence, it may imply a redistributing effect of tourist profiles within a destination. In Canary Islands, the tourist profiles that experience significant growth are “Only flight + Staying with friends or relatives”, “Flight + Staying in apartment with Selfcatering”, whereas the tourist profiles that reduce its presence are “Flight + Staying in 4* hotel with Half-board” and “Flight + Staying in 4* hotel with All-inclusive”. It is also relevant to note that the average length of stay is also different between LCC tourists and Non-LCC tourists. For instance, for the case of “Only flight + Staying with friends or 38 relatives”, LCC tourists stay, on average, 2 days less than Non-LCC tourists. However, for the rest of relevant tourist profiles, LCC tourists stay, on average, 1 day less than Non-LCC tourists. Finally, it should be noted that the mean party size hardly varies between LCC and Non-LCC tourists. Identifying the role of LCC with an econometric model is not straightforward. It is necessary to compare vis-a-vis the expenditure of LCC tourists and Non-LCC tourists. For that purpose, the set of tourist products (i.e. package and accommodation) needs to be exactly the same and the only difference between these two needs to be just the air company chosen (i.e. if the tourist travels with LCC or not). It makes sense for large samples. The way of distinguishing between tourist products and the kind of air company chosen is employing dummy variables. The significance of the dummy variables is necessary to test the significance of the model specification. Provided that the dummy variables are significant, the focus is the value of the difference between dummies associated with LCC tourists and Non-LCC tourists. Such differences will determine, ceteris paribus, how much or less is every kind of tourist spending at the destination. The results of the econometric model are appropriate because they are significant and they show the expected signs and values. More precisely, the results show that, on average, the hypothesis is true. It means that, ceteris paribus, tourists who travel with LCCs are spending more money at the destination per night and party size than those tourists who did not travel with LCCs. Nevertheless, such transfer ratios are usually lower than fifty per cent, although they differ by tourist profiles. Amongst the most popular tourist profiles, the highest savings transfer rate belongs to “flight + stay with friends or relatives” tourist profile, which reaches 49.8%. For the rest of the relevant tourist profiles the percentages are lower. For instance, “flight + apartment with self-catering” case has got a transfer rate of 4.1%, “flight + 4* hotel with half-board” case has got a savings transfer rate of 35.3% and finally, “flight + 4* with 39 All-inclusive” has got a savings transfer rate of 22.1%. It proves that savings transfer is heterogeneous by tourist profiles and that not all the savings are finally transferred, but some are net savings for the tourist. It should be taken into account that, despite the existence of such LCC savings transfers, LCC tourists also stay less nights in the destination. Thus, it is an interesting issue to explore its consequences in terms of total expenditure and economic growth of the destinations. It should be noted that the same methodology can be applied to test the hypothesis to any origin airport, destination airport or route. The model can also focus on specific airlines, if required. However, it only takes into account a part of the story because the flow of tourists is not measured. Hence, this analysis requires the complementary study of forecasting LCC and Non-LCC tourist arrivals. Both studies together provides light on final added value and hence on GDP growth and employment. Such results are relevant to assess the entrance of LCCs at the destination. Future research may also focus on related issues, such as the role of expenditure in origin and how it is converted into added value in the destination. Additionally, it is interesting to explore the nationality issue of the LCCs and how it affects the control on frequency and air fares, which may be a sensitive issue for a destination. 46 2.2 State of the art The effects of tourism on economic and employment growth in destinations have been well documented in the literature. The tourism industry is highly sensitive to economic cycles because, on average, outbound, inbound and domestic tourism flows may be affected more than the consumption of other goods and services. Thus, during an economic crisis, the consumption of luxury goods and services, such as tourism, are expected to significantly decrease, which affects arrivals and tourism receipts in destinations (Lanza, Temple, & Urga, 2003; Smeral, 2003; Eugenio-Martin & Campos-Soria, 2011). Destinations need to anticipate such downward shifts in demand by reducing prices or identifying add value demand strategies in an attempt to maintain or improve their market share (Sheldon & Dwyer, 2010), or by devaluing their currencies in relation to the main countries of origin (Prideux, 1999). To date, most of these decisions by policy makers have been based on macroeconomic indicators to evaluate the impact of economic crisis on destinations. Macroeconomic variables, such as arrivals, receipts or expenditure, are readily available over time and so are more likely to be used in applied studies. Figures related to these variables can be used for forecasting and for making homogeneous comparisons between destinations. However, as pointed out by different researchers, microeconomic approaches are also required to effectively manage crisis (Bronner & de Hoog, 2012; or Smeral, 2009). Ideally, policymakers should combine macro and micro indicators. In fact, they need linkages between changes in GDP and arrivals or receipts in order to manage the crisis, while taking into account consumer responses to the crisis (Eugenio-Martin & Campos-Soria, 2014). 2.1. Macroeconomic indicators In recent decades, tourism demand analysis has addressed the issue of the impact of different kinds of crises, such as economic crisis (Smeral, 2010; Hall, 2010; Page, Song & Wu, 2012), terrorist attacks (Blake & Sinclair, 2003; Araña & León, 2008), or natural disasters, such as 47 epidemics or earthquakes (Eugenio-Martin, Sinclair & Yeoman, 2005; Carlsen & Hughes, 2008; or Mao, Ding & Lee, 2010). However, one of the driving forces of tourism demand is the economy. A review of the literature on tourism and crisis suggests that economic and financial crises receive the most attention, although these crises are often linked to other crises such as terrorism (Wang, 2009; Hall, 2010). The first main economic crisis studied in the tourism literature was the Asian financial crisis of mid-1997. The crisis was analyzed by Henderson (1999), Prideaux (1999) and Law (2001). Okumus, Altinay and Arasli (2005) investigated the impact of the February 2001 economic crisis in Turkey on the tourism sector in Northern Cyprus, while O´Brien (2012) analyzed the tourism policies implemented to address the Irish crisis of 2008, during which the Irish tourism industry collapsed dramatically. Thereafter, the global economy crisis triggered in the United States in 2007 was the most widely studied crisis due to its profound negative impact on the world economy in general and on tourism activity in particular (Song & Lin, 2010; Brent-Ritchie, AmayaMolinar & Frechtling, 2010; Page, Son & Chenguang-Wu, 2012). Different studies have stressed the importance of advance planning and coordination between public and private agents in a context of minimizing the effects of the crisis. However, estimations and predictions that focus on macro indicators only partially help policy makers to evaluate the impact of the tourism crisis. There are to main reasons for this. On the one hand, changes in arrivals do not necessarily mean that tourism receipts also decrease by the same proportion; in fact they can increase. For instance, Bronner and Hoog (2012) stated that the number of holidays increased and expenditure decreased in the Netherlands in 2009. These kinds of studies do not take into account the fact that downward demand shifts could also be an effect of reducing prices or devaluing the exchanges rates between origin and destination countries. Thus, arrivals and receipts need to be simultaneously analysed, otherwise the impact of the economic crisis on tourism leads to biased results. On the other 48 hand, as stated by Sheldon and Dwyer (2010), investment and marketing strategies, the development of new products, and action plans for maintaining business viability are not well understood. A wider microeconomic perspective is needed to obtain insight into tourist behaviour during crises in order to further explore these issues. 2.2. Microeconomic indicators Sheldon and Dwyer (2010) stated that the final impact of an economic crisis cannot be approached from a macroeconomic point of view alone, since the crisis may affect the firms’ strategies and the tourists' behaviour. Thus, a microeconomic approach based on individuals or households is required. This approach often deals with participation decisions, expenditure, or any other experimental observation of tourist behaviour as endogenous variables. However, few studies have investigated how tourists redistribute their tourism expenditure during an economic crisis. Bronner and Hoog (2012) proposed a general framework to investigate the consequences of the global economic crisis on individual tourist behaviour and on tourists' economizing strategies. They addressed the kind of cutback decision by taking into account the geographical range of the crisis and its depth, and characterized different expected responses. They included thirteen different cutback decisions (cheese-slicing strategies) such as "expending fewer days on vacations", "booking cheaper accommodation" or "taking another means of transport". Some of these strategies may affect travel expenditure, whereas others may affect expenditure at the destination. Fleischer, Peleg and Rivlin (2011) suggested that most studies on vacation expenditure do not distinguish between these expenditures. To the best of our knowledge, few studies have analyzed cutback decision strategies. Alegre et al. (2013) studied the consequences of the economic crisis on Spanish households and particularly focussed on the role of employment. Thus, they differentiated between two mutually related decisions: tourism participation and tourism expenditure. Finally, EugenioMartin and Campos-Soria (2014) studied how European tourists reacted during the economic 49 crisis by modelling the tourism expenditure cutback decision. This study reinforces the idea that households that cut back on tourism expenditure in 2009 were more likely to spend their holidays closer to home. However, neither of these studies explicitly investigated how these tourists cut back their budget. Tourists may react by shorting the length of stay, travelling by lower-cost carriers, or staying in cheaper accommodation establishments. Papatheodorou, Rossello and Xiao (2010) stated that travelling closer to home is one of the most important strategies to reduce expenditure, while Harris Interactive (2009) suggested that a shorter length of stay during summer is an important cutback decision. Some studies have investigated these issues in economic crisis scenarios, as we have already shown, but most of them have not done so. For instance, following the ETC (2009) report, in economically difficult times tourists tend to minimize product prices and quality, prefer destinations closer to home than long-distance destinations, scale back their expenditure per night, and economize on the duration of their stay. Olive Research (2009) reported that 64% of visitors from the United States, Spain, Ireland, and France were likely to cut back on holidays by changes in duration and spending. However, these are descriptive approaches that did not deeply analyze how tourists cut back tourism expenditure. This is the main focus of this paper. 2.3 Methodology 2.3.1. Econometric modeling Modelling tourism expenditure and how households adjust their tourism expenditure during economic crisis is problematic. Firstly, cutback decisions do not depend on income variations alone, but also depend on other individual characteristics. Since household tourism expenditure may vary due to circumstances not related to the economic crisis, biased results may be obtained. This paper avoids this potential bias since the decision to make a cutback 50 was directly affirmed by the interviewee. Secondly, the "how-to-cut-back" decision was observed only when the individuals had cut back their tourism expenditure. Thus, there was a sample selection bias: tourists who decided not to cut back were not included in the next stage ("how-to-cut-back"). Finally, the "cut back" and "how-to-cut-back" decisions form a simultaneous decision, because the decision to cut back affects the decision of "how-to-cutback". Based on an econometric point of view, both decisions are modelled using a two-step approach. Simultaneity is captured by assuming correlation between the error terms of both equations. The econometric model used to address this two-step decision is an adaptation of the Heckman model (Heckman, 1976, 1979). The endogenous variable in the first step i cb is a binary response variable that takes value 1 if the tourist decides to cut back and zero otherwise, where i denotes individuals. The endogenous variable of the second step is a multinomial response variable. "How-to-cut-back" is a discrete variable denoted by ij hcb , which takes values between one and six according to different alternatives: "fewer holidays", "reduced length of stay", "cheaper means of transport", "cheaper accommodation", "travel closer to home" or "change the period of travel", where j denotes these alternatives. This model is based on random utility models: Let * i cb be the latent variable of cutback decision and * i hcb the latent variable of "how-to-cut-back", which depend on exogenous variables i z and ij x , respectively. In structural equations format, the Heckman model is conducted using a latent variable i l that represents correlation between both equations (Skrondal & Rabe-Hesketh, 2004). The adapted Heckman model in structural equations format is shown below: 51 The model specification for the cutback decision is expressed according to the following equations: * (0,1) iiii i cb z l N     (9) * * 1 0 0 0 i i i if cb cb if cb      (10) where  denotes a vector of unknown parameters, and i  represents the error term. Taking into account equations (9) and (10) we have: (1)( 0)( )( ) iiii iiiii Pcb Pz l P z l Fz l          (11) Equation (11) is a logit distribution function that obtains the probability of cutting back in tourism expenditure. The model specification for the "how-to-cut-back" decision expresses the utility provided by each alternative j, as shown in the following equation: *2 + (0, ) ij ij j i j ij ij hcb x l e e iid     (12) where j  and j  denote vectors of unknown parameters, and ij e represents the error term. Thus, the probability of the "how-to-cut-back" decision is obtained in equation (12), for each alternative j.   6 1 exp ( ) exp ij j i j ij ij ij j i j j xl Phcb j P xl        1,.......,6j  (13) Equations (11) and (13) are the adapted version of the Heckman model in structural equations. Equation (11) is a probit model and equation (13) is a multinomial logit model, so 52 both are estimated simultaneously. In order to make the identification process of the structural equation model easier, the variance of the latent variable is set equal to one. 3.2. Model specification The exogenous variables i z considered in the specification of the cutback equation (equation (11)) can be divided into socioeconomic variables and regional variables. The set of socioeconomic variables are gender (male = 1), education, employment, age, and age squared; the latter variable captures the non-linear effect of age on the cutback decision. These variables are used as a proxy for personal income. Education is a continuous variable that takes the value of the age at which the individual stopped full-time education. Employment can be any of the following: Farmer, forester or fisherman; shop owner or craftsman; professional, such as a lawyer, medical practitioner, accountant, or architect; manager of a company; professional, such as an employed doctor, lawyer, accountant, or architect; general manager, director or top management; middle management; civil servant; office clerk; employees, such as salesman or nurse; supervisor (foreman) or team manager; manual worker; unskilled manual worker; homemaker; student (full time); retired; unemployed; and sets of other occupations within different professional categories. The set of regional variables were taken into account because tourism preferences are affected by the place of residence of the household. As pointed out by Hung, Shang and Wang (2013), households that belong to the same region have a similar tourism expenditure pattern. Therefore, if tourism expenditure estimations ignore factors related to the geographical location of the tourists, biased results are likely. In particular, climate, per capita GDP in Purchansing Power Standards (PPS), and GDP growth have been included. Climate in the region of origin is one of the most important "push" factors in the outbound tourism demand (Agnew & Palutikof, 2006) and explains asymmetries in the willingness to travel between 53 regions (Madison, 2001). Eugenio-Martín and Campos-Soria (2014) showed that households located in regions with a "good climate" are more likely to cut back on their tourism expenditure than those located in regions with a poorer climate. The definition of climate is based on the double-hurdle climate index introduced by Eugenio-Martin and Campos-Soria (2010). The double-hurdle climate index ranges from 0 to 12 depending on the number of months that pass each hurdle. The climatic variables considered were temperature, rainfall, and days with rainfall. The thresholds that determine each hurdle are based on Mieczkowski´s (1985) tourism climatic comfort conditions. On the other hand, per capita GDP in PPS is taken into account as a proxy for personal income as well as the socioeconomic variables. Finally, GDP growth captures expectations regarding personal income variations. Consumption theories predict that changes in demand may be due not only to changes in current income but also to expectations regarding future income. Hong-bumm, Jung-Ho, Seul and SooCheong (2012) analyzed this effect on international tourism demand. Similarly, the exogenous variables ij x considered in the "how-to-cut-back" equation are socioeconomic and regional variables. Tourist preferences and how households take decisions on tourism expenditure cutbacks depend on age and gender. These variables are defined in the same way as in the cutback model specification. At the regional level, climate, length of the coast, and the presence of airports are considered relevant in household cutback strategies. The variable coast represents an index of how relevant the length of the coast is compared to the size of the region, and airport is a dummy variable that takes unitary value if the region has at least one airport. 2.4 Case study This study used a survey conducted at the household level and macrodata concerning the region of origin. Microdata were obtained from the "Attitudes of Europeans Towards 54 Tourism" survey conducted in 2009 in EU-27 regions and was part of Flash Eurobarometer 281 (European Commission, 2010). It contains information on the socioeconomic characteristics of 23,606 households and information on their decisions on outbound tourism demand, such as destination choice and cutback decisions. The macrodata considered in this study was collected for 165 regions of EU-27 countries. Eurostat was the data source for GDP in PPS and GDP growth and the data on climate index were obtained from the World Meteorological Organization. Average per capita GDP in PPS was 22,942.44$ in 2009 for the whole sample, reaching a maximum value of 62,500$ for Luxembourg and a minimum value of 6,400$ in Severozapaden (Bulgaria). It should be pointed out that 95.7% of the regions had negative GDP growth in 2009. Groningen (The Netherlands) reached the lowest growth rate at -17.01%, while Północny (Poland) was one of the few regions that had a positive value of 1.64%. 2.5 Results 2.5.1. Economising strategies by country The results of the descriptive analysis of the dataset are striking. They show that 46.32% of the interviewees had to cut back on tourism expenditure in 2009. Of these, 26.76% opted for "reduced length of stay", whereas 21.84% chose "cheaper accommodation", 18.87% opted for "closer to home", 16.15% chose "fewer holidays", 8.89% opted for "period of travel" and 7.48% chose "cheaper transport". The literature has shown regional differences in the probability of cutbacks. Eugenio-Martin and Campos-Soria (2014) showed that there were marked differences between NorthEuropean and Mediterranean regions due to differences in climate and GDP. An analysis of the relative frequencies of the economizing strategies by country (Table 7) shows that most of the countries did not conform to the ranking in relative frequencies described in the previous 55 paragraph. Countries such as Austria, France, Greece, Malta, Romania, and Slovenia conformed to the ranking in relative frequency at the aggregated level, but this was not the case for the remaining countries. Thus, there was regional heterogeneity in the pattern of cutbacks. In any case, "reduced length of stay" and "cheaper accommodation" were the most frequently chosen economizing strategies, independently of the cited ranking and the country considered. Table 7. Tourists’ economising strategies during an economic crisis by country (EU-27) Country P(reduced length of stay) P(cheaper accommodation) P(closer to home) P(fewer holidays) P(period of travel) P(cheaper transport) Austria 31.84 22.9 17.31 14.52 11.17 2.23 Belgium 24.65 19.17 13.69 16.43 15.06 10.95 Bulgaria 36.52 24.65 10.04 13.24 7.76 7.76 Cyprus 22.07 19.48 22.72 21.42 5.84 8.44 Czech Republic 22.4 29.46 21.57 14.1 4.14 8.29 Denmark 21.95 23.57 15.44 17.07 8.94 13.01 Estonia 21.55 19.16 20.95 17.36 6.58 14.37 Finland 31.03 14.77 19.21 19.21 8.37 7.38 France 37.78 23.28 12.97 12.21 7.63 6.1 Germany 31.4 18.59 16.94 11.57 14.46 7.02 Greece 35.36 21.73 15.94 15.36 6.08 5.5 Hungary 20.76 25.23 13.41 24.28 8.94 7.34 Italy 33.33 21.28 12.85 14.05 14.45 4.01 Ireland 22.6 20.65 26.73 13.69 10 6.3 Latvia 19.1 24.2 16.56 18.47 3.82 17.83 Lithuania 30.92 18.55 25.25 16.49 5.15 3.6 Luxembourg 19.38 22.44 23.46 14.28 10.2 10.2 Malta 25 23.61 22.22 11.11 6.94 11.11 Poland 27.72 18.18 20 11.36 11.36 11.36 Portugal 20.57 22.01 24.4 18.18 11.48 3.34 Romania 26.05 22.4 18.48 15.4 9.52 8.12 Spain 29.77 26.86 13.26 19.41 6.14 4.53 Slovakia 20.13 24.3 23.61 15.27 8.33 8.33 Slovenia 35.38 25.38 15.38 13.84 7.69 2.3 Sweden 33.33 13.19 18.75 19.44 5.55 9.72 The Netherlands 13.63 18.18 27.84 21.02 10.79 8.52 UK 18.93 21.92 23.58 16.94 10.63 7.97 62 instead of Australia. Finally, "cheaper transport" and "period of travel" have a fuzzy effect on destinations and are considered as "other effects". As shown in Figure 4, the full effects reach the highest probability (47.4%), followed by partial effects (35.4%) and then by other effects (17.2%). This information could be used by policymakers and tourism firms to minimize the effects of the economic crisis on destinations, since tourists are more willing to cutback by using economizing strategies at the destination (full effects). Destinations can anticipate downward demand by suitably adapting their strategies. For instance, suppliers of tourism services need to know if the reduction in tourism expenditure is going manifest as fewer holidays, fewer days of vacation, or as lower quality services, such as tourists booking lodging deals. Gokovali, Bahar and Kozak (2007) stated that lower tourism expenditure does not necessarily mean fewer vacation days, since service quality may also change, and thereby affecting tourism expenditure. Thus, destinations should sometimes reduce prices, but not always. They need to identify added-value demand strategies, such as offering more flexible packages. 63 Figure 4. Final effects on the destination Additionally, a post-estimation analysis shows how the estimated probabilities change according to some key determinants, such as climate index and age of the head of the household. Figure 5 shows the moving median of these probabilities in relation to the climate index. According to this figure, there is a clear effect on the probability of cutback alternatives by climate. Firstly, the probabilities of "reduced length of stay" and "cheaper accommodation" are the highest of the six alternatives and steadily increase in line with increases in the climate index. Secondly, "period of travel" and "cheaper transport" show the lowest probabilities and steadily decrease in line with increases in the climate index. Finally, "closer to home" and "reduced number of trips" remain almost constant. However, if we analyze the rate of change of the probabilities of the alternatives by climate in the country of origin, the changes are significant. On the one hand, households located in regions with the best climatic conditions for tourism (climate index = 12) show a 32% higher probability of cutting back using "reduced length" than households located in regions with the worst climate (climate index = 0). In the case of "cheaper accommodation", the change is less marked. Households with the best climate index are 1.05% more likely to cut back using this option than those with the 64 worst climate index. On the other hand, tourists with the highest climate index are 2% less likely to cut back using the option "period of travel" than tourists with the lowest one. These results indicate that differences in the place of origin have a strong effect on these probabilities. Additionally, post-estimation analysis let analyse how the estimated probabilities change with some key determinants, such as climate index and age of the head of the household. Figure 5 plots the moving median of these probabilities in relation to climate index. According to such figure, there is a clear effect on the probability of cutback alternatives by climate. Firstly, the probabilities of “reduced length” and “cheaper accommodation” are the highest of the six alternatives and grow smoothly with the climate index. Secondly, “period of travel” and “cheaper transport” show the lowest probabilities and decrease steadily with the climate index. Finally, “closer to home” and “reduced number of trips” remain almost constant. However, if we analyse the rate of change of the probabilities of the alternatives by climate in origin, the changes are pretty significant. On the one hand, households located in regions with the best climate conditions for tourism (climate index = 12) show a 32% higher probability to cut back with "reduced length" than those households located in regions with the worst climate (climate index = 0). In the case of “cheaper accommodation”, the change is not so sharp. It is 1.05% more likely to cut back with this cutback option for those households with the best climate index rather than those with the worst climate index. On the other hand, it is 2% less likely to cut back changing “the period of time” for those tourists with the highest climate index than tourists with the lowest one. These results indicate that differences in the place of origin play an important role on those probabilities. 65 Figure 5. Moving median probability of economising strategies by climate 66 Figure 6. Moving median probability of economising strategies by age The age of the head of the family offers some new insights. Figure 6 is constructed according to the median probabilities of each alternative of cutback by age moving bands. An analysis of the rate of change of probabilities by age shows that individuals 65 years old have a 34.78% higher probability of choosing "reduced length" than individuals 20 years old. Regards "cheaper accommodation", individuals 65 years old have a 33.33% lower probability of choosing this option than people 20 years old. 2.6 Conclusions and further research The paper investigated the economizing strategies followed by European tourists during the global economic crisis in 2009. To date, most of the literature has addressed this issue from a macroeconomic perspective. Thus, given a GDP shock, this kind of analysis indicates the sensitivity of outbound or inbound tourism demand. However, some tourist profiles within a population undergo different affects. Some tourist profiles may be more or less sensitive to 67 the GDP shock. This information is relevant to tourism and hospitality management strategies, because agents can anticipate this kind of behaviour within a population and define a suitable set of marketing strategies to deal with the shock. Public policymakers are also interested in absorbing impacts such that any negative effects on employment and tourism added value are minimized. For this reason, this paper combines micro and macro perspectives. This enriches the analysis, although it also increases its complexity. The analysis employed an econometric model, but its specification is not straightforward. The main purpose of the study was to understand the choice of the economizing strategy or strategies used by the tourists. However, the model should take into account a simultaneous decision taken by all the tourists, i.e. a cutback decision as well as a "how-to-cut-back" decision. This kind of decomposition is needed to isolate the determinants of each simultaneous decision. Otherwise, the results of the latter will be biased. To address this issue, an adapted model of the 2-stage Heckman model was applied within a generalized structural equations modelling approach. The results have implications for tourism and hospitality management. Firstly, age matters. In particular, older tourists are more likely to reduce their length of stay under a GDP shock. Second, climate in the country of origin also matters. Tourists who live in regions with a good climate are more likely (32%) to reduce their length of stay than tourists who live in regions with a poorer climate. Third, the preferred economizing strategies for most regions are clearly to reduce the length of stay followed by booking cheaper accommodation. The average probabilities of using these strategies are 27% and 20%, respectively. However, there is heterogeneity among regions. This information could be employed to build flexible packages that could suit the needs of different tourist profiles during periods of economic crises. This kind of flexibility should be based on a combination of micro and macroeconomic variables, 68 i.e. age, gender, climate in the regions of origin, severity of the crisis (GDP shock), and expectations regarding GDP growth. Destination managers should understand which economizing strategies involve full effects, partial effects, or any effect on the destination. On the one hand, economizing strategies that have full effects on the destination are "reduced length of stay" and "cheaper accommodation". On the other hand, "fewer holidays" and "closer to home" may or may not directly affect the destination (partial effects). Finally, "cheaper transport" and "period of travel" are fuzzier than the partial effects regarding their impact on the destination (other effects). According to the results, there is a 47.4% probability of cutting back by using economizing strategies which directly affect destinations (full effects), whereas the probabilities are 35.4% and 17.2% for partial effects and other effects, respectively. Ideally, hospitality sector and tourism destination policymakers should coordinate their marketing campaigns. They should take into account the different sensitivities of tourist profiles during GDP shocks and create flexible packages that could minimize the impact of crisis. Further research on other issues is needed. The results presented in this paper could be improved by using methodologies that can address the issue of the severity of cutbacks. 69 CHAPTER 3 Tourism: economic growth, employment and Dutch Disease. 3.1 Introduction Since 2008, Spain is under a strong economic recession (0.92% reduction in the real GDP from 2008 to 2012 and unemployment rate of 24.3% in 2012). The main causes and effects of the actual downturn situation are: high private debt (fed it after years of low interest rates that spurred a real estate bubble), high unemployment rate, lower wages, low private consumption, credit shrinkage (banking crisis) and higher interest rate for public bond emissions. Beyond this brief diagnosis, the Spanish crisis has two characteristic factors: the membership to the European currency (euro) and, as a consequence, a public control deficit to fulfill the EU deficit commitment. The first factor acts as if Spain had a fix exchange rate forcing interior devaluation through lower salaries to earn exterior competitiveness. The second factor does not permit to fall into persistent public budget deficit and reduce the possibility to carry out demand policies to foster the economy. The interior devaluation is already reducing the deficit and the salaries, but it is also making falter the domestic demand. And Spain, as most of the developed economies, relies strongly on the domestic demand to boost the economy. In parallel with the economic situation described above, Spain has been receiving a high arrival of inbound tourism since 2010 as a consequence of the Arab Spring that began on December of 2010 in Tunisia and rapidly spread to other Arab countries of the region. The United Nations World Tourism Organisation (United Nations World Tourism Organisation, 2013) does not share this point of view and affirms that the rise in tourist arrivals to Spain over the regional average (Mediterranean area) obey to internal improvement such as the 70 modernisation of supply, the human resource training, quality improvement or marketing and promotion. This new situation, together with more optimistic economic forecasts (International Monetary Fund, 2013), has fed the idea that tourist arrivals could substitute the weakened domestic demand and revitalize the economy. The success of some Asian countries in the eighties promoting economic growth through export-oriented industries (World Bank, 1993) and the export orientation of the tourism, has guided the studies about tourism and economic growth around the export-led hypothesis (Balassa, 1978). Authors such as Sequeira and Maças (2008) or Dritsakis (2004) support the capacity of tourism, a non-technology-intensive sector, to promote economic growth and enhance capital accumulation. These conclusions contradict Solow (1956) and some other authors findings such as Aghion and Howitt (1998); or Grossman and Helpman (1991) about the relations between high-technology sector and long term growth. Lanza, Temple and Urga (2003) affirm that the lower growth in productivity in tourism-based economy could be overcome by a progressive specialisation on tourism that could improve the terms of trade and compensate the lost in productivity. Moreover, they also highlight the importance of the high price elasticity and income elasticity of demand for tourism that may compensate the loss in productivity in the long term. Additionally, there are more profound consequences in the relationship between tourism and the economy beyond the growth enhanced or the productivity gained that should not be neglected. The differences in the intensive use of capital and labour have also important implications at sectorial level. Copeland (1991) and Chao, Hazari, Laffargue, Sgro and Yu (2006) underscore the importance of non-tradable goods in tourism-based economies. According to them, tourism enhances the consumption of non-tradable goods and improves the terms of trade, although it could produce capital decumulation from the manufacturing sector (capital intensive) to the non-tradable ones (labour intensives). Moreover, the 71 appreciation of the real exchange rate because of tourist arrivals can also undermine the exterior competitiveness of the traditional exports. Both, the displacement in capital and labour endowment from traditional sectors to the non-tradable ones and the appreciation of the real exchange rate can generate an economic “illness” known as Dutch Disease by which the positive effect of tourism on the economy in the short term could ends up in an economic shrinking in the long term (Corden & Neary, 1982). Tourism-led countries are especially sensitive to the Dutch Disease due to the entrance of foreign money. The use of recursive-dynamic CGE models permit to work at these two levels. On the one hand, they allow quantifying the impact of tourist arrivals on GDP and unemployment. On the other hand, they also permits to analyse the effect of such shock on the resource reallocation (capital and labour) among sectors, the lost in competitiveness and, consequently, to test the existence of the Dutch Disease in the economy at micro level over time. To such aims, a recursive-dynamic model with two scenarios and five periods is developed. The first scenario is based on the International Monetary Fund projections (International Monetary Fund, 2013) and in the Economic bulletin of the Bank of Spain (Bank of Spain, 2013), i.e., post-crisis scenario. The second is a pre-crisis scenario (buoyant situation) based on the Spanish economy performance in the five years previous to the economic crisis. Additionally, the model is based on three datasets: Input-Output Table (IOT), the Tourism Satellite Account (TSA) and National Account for Spain. According to Blake, Durbarry, Sinclair and Sugiyarto (2001), the Input-Output framework overestimates the total GDP effect and underestimates the total effect on tourism sector. The TSA is the dataset able to fill the lack of tourism information in the IOT. Thus, the IOT and the TSA are combined to get a deeper representation of the tourism sector that the IOT is not able to provide it. The two scenarios and the combination of the datasets try to provide a wider perspective of the true potential of tourism to promote economic growth and reduce unemployment. 78 larger countries. This conclusion is in accordance with a more general one by Easterly and Kraay (2000). They conclude that, on average, small countries have higher GPD per capita, although they are more vulnerable to international trade shocks. Lanza et al. (2003) analyse the OCDE countries. According to them, the lower growth in productivity in tourism-based economy could be overcome by a progressive specialisation on tourism that improve the terms of trade and compensate the lost in productivity relative to other sectors. Moreover, they also highlight the importance of the high price elasticity and income elasticity of demand for tourism that may compensate the loss in productivity in the long term. Lee and Chang (2008) expand their analysis to include both OECD and non-OECD countries. They find a long-run relationship between the tourism and real GDP per capita. Eugenio-Martín et al. (2004) conclude that countries on lower and medium income in Latin America benefits from the tourism and the economic growth generated by it. 3.2.2.3. Tourism and capital accumulation In general, the literature has followed a theoretical perspective regarding to the capital accumulation process in tourism. The motivation of such process has been based on the distinction between tradable and non-tradable goods. The tradable goods are more capital intensive and can be exported. On the contrary, the non-tradable goods are more labour intensive and can only be consumed within the country. The existence of tourism in the country increases the consumption of non-traded goods. Thus, tourism produces a reallocation of resources from tradable to non-tradable sectors (capital decumulation). At the same time, the appreciation in the real exchange rate because of the tourist arrivals erodes the exterior competitiveness of the tradable goods. This is the theoretical reasoning followed by Copeland (1991) and Chao et al. (2006), although for Copeland (1991) the de-industrialisation is not necessarily harmful, unless external economies are important in the industry. The consequences of the capital decumulation described by these authors can be considered as a 79 symptom of the Dutch Disease. Hazari and Sgro (1995) and Albadalejo and Martínez-García (2013) develop a dynamic model to explain the capacity of tourism to enhance economic growth and capital accumulation. The latter authors affirm that tourism allows the imports of foreign capital. Moreover, the model can endogenously increase the tourism attraction in reaction to tourism demand. Thus, tourism enhances economic growth and capital accumulation. Nowak, Shali and Cortes-Jiménez (2007) and Poirier (1995) carry out an empirical analysis of the capital accumulation process promoted by the tourism in Spain and Tunisia, respectively. Nowak et al. (2007) combine both theoretical and empirical results. They develop a theoretical proof of the so called EKIG hypothesis (exports-capital importsgrowth) and confirm, with econometrics tools, its positive and significant impact in Spain due to the capital imports. Finally, Poirier (1995) follows a more descriptive analysis, its study points out the positive effect of tourism on capital accumulation, deficit reduction and the balance of trade. 3.2.3. Tourism and employment According to Riley, Ladkin and Szivas (2001) tourism requires lower qualification and pay lower wages rather than other more capital-intensive activities (hypothesis also followed by the authors of tourism and economic growth previously adduced). Such description is common in practically all literature about the topic: Mathieson & Wall (1982) or Jafari, Pizam & Przeclawski (1990). This conclusion is in line with the kind of industry described by Lewis (1954). The assumption of the low technological use in the tourism sector should imply a higher demand of lower-qualified workers. However some authors are more skeptical about the link between tourism and employment. Riley et al. (2001) question the relationship between increasing tourism and increasing employment. Tourism activities are more outputsdriven (outputs rule over inputs) than other activities. “If demand varies in the short term, then supply inputs need to match that variation in the cause of productivity” (Riley et al., 80 2001, p. 30). Thus, according to them, the forecast of tourism demand, the standardisation of services and products and flexible resources are crucial to match increasing tourism and increasing employment. 3.2.4. Tourism and CGE models Most of the state of the art about tourism and CGE models has been focused on the analysis of the effect of increasing tourism demand (new tourist arrivals or higher tourism spending) and /or the effect of indirect taxes on tourism and the economy. But the impact of such shocks over time has not been well explored yet. Copeland (1991) develops a theoretical CGE framework to analyse the effect of tourism on the economy. Many of his conclusions have been tested by others researchers such as Adams and Parmenter (1995) or Narayan (2004) among others. Authors such as Adams and Parmenter (1995), Zhou, Yanagida, Chakravorty and Leung (1997) or Narayan (2004) have guided their studies to quantify the impact of increasing tourism demand on the economy. A quite general conclusion of the effect of a rise in tourism (new arrivals or increase in spending) on the economy is that: a positive tourism shock produces an appreciation of the exchange rate that erodes traditional exports and increase imports. Nonetheless, the tourism shock overcomes the decline in traditional exports and the rise in imports; and thus, it improves the terms of trade. The impact of taxes on tourism has also been a recurrent topic. Gooroochurn and Thea Sinclair (2005) focus their studies on the impact of taxes on tourism and the economy. According to them, tourism taxes can bring welfare gains since international tourists bear most of the taxes. Meng, Siriwardana and Pham (2013) combine both the increase in total tourism demand and change in taxes in their study about Singapore. The CGE modeling can be also combined with other methodologies to provide a more precise insight of the shock considered. In this sense, Blake, Durbarry, Eugenio-Martin, 81 Gooroochurn, Hay, Lennon, Sinclair and Yeoman (2006) combine econometric estimations based on tourism indicators with a CGE model. In a first step a structural equations model is used to forecast the tourism spending. In the second step, a CGE model is carried out to quantify the shock predicted in the first step. In their study applied to Scotland, they calculate that an increase of a 10% in tourism spending will increase GDP between 25.3 million pounds and 42.6 million pounds in the short and long-run, respectively. This boost in GDP will also generate 3,326-4,455 more employments. Gago, Labandeira, Picos and Rodriguez (2009) and Blake (2000) are one of the few that include the tourism sector in a CGE model about Spain. Previous CGE studies of Spain such as Polo and Sancho (1993) or Kehoe, Polo and Sancho (1995) did not include tourism. Gago et al. (2009) and Blake (2000) assess the impact of indirect taxes on tourism and in the economy. Moreover, Gago et al. (2009) also study the impact of such shocks on the employment rate. The negative effect of increasing VAT rate on employment varies from 0.88% to 3.36% depending on the shock considered. The shock assumed in any CGE model has deeper effects on the economy beyond the GDP or the unemployment. It also produces changes in the income distribution within households which, depending on the goal chased, should not be neglected. The impact of tourism on the economy has also been extended to poverty relief analysis, and, under these circumstances, the inclusion of different households is vital. Blake, Arbache, Sinclair, and Teles (2008) conduct an analysis of the consequences of tourism on poverty relief in Brazil. Their main conclusion is that tourism benefits all kinds of households but low income classes benefit most. In the same topic, Wattanakuljarus and Coxhead (2008) examine the effect of tourism on poverty in Thailand. According to them, and assuming full employment, tourism is not a pro-poor activity due to that low income households work on agriculture and other tradable 82 sectors not especially related to tourism activities. In any case, the boost in tourism affects positively all kinds of households but high income classes are the most benefited. Some authors have also followed a wide perspective in regards to the impact of tourism on the economy. Blake and Sinclair (2003) quantify the impact of the terrorist attack of September 11th in USA. They highlight the importance of the US government intervention to relief the negative impact of 11-S crisis on the tourism activity, albeit their decisions lacked of suitable cost-effective analysis. Without the government response the employment lost would have been around 383.000 (full time employment), due to such interventions the employment saved are around 60% of the previous 383.000. In terms of GDP, the government response keeps the fall of GDP in 10 billions of dollar, instead of a fall of 30 billions of dollar without the intervention. Dwyer, Forsyth, Spurr and VanHo (2006) assess the impact of Iraq war and SARS on the tourism and in the economy in Australia. Their main finding is that inbound tourism falls as well as outbound tourism due to the crisis. As a result, savings, domestic tourism and other non-tourism consumption increase, although the net effect on GDP is negative. Blake, Sinclair and Sugiyarto (2003) study the economic impact of the accession of Malta and Cyprus to EU, both rely strongly on tourism to grow. Malta and Cyprus will increase their GDP about 4% and 3.5% in the long term, respectively, after the accession to the EU. In terms of employment, Malta will be able to generate 3,559 more full time employments; and Cyprus 8,543 full time employments. Sugygarto, Blake and Sinclair (2003) assess the effect of globalisation and foreign tourism in Indonesia through different tariff reduction scenarios. Their main conclusion is that tourism eases the globalisation process in Indonesia. In fact, tourism can reduce the domestic prices, improves their terms of trade and, in last term, it can improve their macroeconomic performance especially the government accounts. Precisely, according to Mabugu (2002) the lack of good macroeconomic performance is the root of the turmoil in the Zimbabwean economy that erodes the benefit 83 from tourism. The paper simulates five different scenarios from currency devaluation to fiscal deficit in Zimbaue. The conclusions are wide, but in general, the effects on employment, fiscal deficit and GDP are positive. Finally, the use of CGE models has also been extended to study the relationships between tourism and environment. Alavalapati and Adamowicz (2000) develop a simple two sector (resource and tourism sector) and two factor CGE models to quantify the interaction between tourism and environment. According to them, the effect of an environmental tax varies depending on the sector levied. For instance, an increase in an environmental tax in the resource sector improves the economy if the environmental damage occurs in the resource sector. Finally, some authors have also studied the interaction between climate change and tourism. Berrittella, Bigano, Roson and Tol (2006) study the world impact of climate change on tourism. Dwyer, Forsyth, Spurr and Hoque (2010) focus their study in Australia and examine the impact of greenhouse gas reduction on tourism. The latter study is the only one that uses dynamic CGE models applied to tourism. 3.2.5. Unemployment in CGE models There are several ways to model unemployment in CGE models depending on the aim chased. One of the first approaches includes wage rigidity (minimum wage) in order to avoid that marginal labour productivity equal real wage as classical labor market theory affirms. Figure 10 depicts the unemployment with minimum wage. Assuming an inelastic labour supply, the minimum wage produces a downward shifting of the labour demand from an equilibrium wage of (w/p)0 to (w/p)1, this new equilibrium ensure labour market clears. Nonetheless the minimum wage forces a new equilibrium in (w/p)min in which the labour supply exceeds labour demand producing involuntary unemployment. Minimum wages hypothesis assumes a positive correlation between wages and qualification. 84 Figure 10. Unemployment (minimum wage) A natural extension of the wage rigidity is the wage curve (Figure 11). Such curve have been formulated and tested by Blanchflower and Oswald (1995). Rutherford and Light (2002) show the way to implement it in CGE format. Quoting Küster, Ellersdorfer, and Fahl (2007, p. 15):”A wage curve captures the relationship between the level of unemployment and the level of real wages and describes how the price of labor is affected by the unemployment rate”. A wage curve assumes a negative relationship between wages and unemployment. If the unemployment rate is high, firms offer lower wages. From the standpoint of modeling, any of these unemployment models tries to break with classical postulate about the labour market clearance condition. In the literature, the minimum wage has been more oriented to model unskilled workers, whilst the wage curve has been used to model skilled workers. An alternative is to assume both kinds of workers in one simple framework. Lögfren (2001) assume four different skill categories depending on the educational level. Another possibility is to consider different and persistence wage differences among sectors (Katz and Summers, 1989). 85 The existence of labour union can also be considered. There are two basic approaches in this regards. One is considering a situation where the union has the bargaining power over wages and firms decide over the employment level demanded (monopoly union model). Another is to assume that both parties have bargaining power (efficient bargaining model). Devarajan, Ghanem and Thierfelder (1997) model the unemployment assuming the existence of labour unions. This approach could be useful to analyse inflationary process under indexed wages regime such as Brazil in the eighties or Israel in the nineties. Figure 11. Unemployment (wage curve) There are other alternatives to model unemployment in CGE framework that are far from classical postulates. Harris and Todaro (1970) develop a model to explain the migration between rural and urban areas. According to them, rural workers could be willing to migrate to urban areas, despite a high unemployment rate in such urban area. The explanation to this apparent contradiction comes from the positive expectation that rural workers have about the real wage in the urban area. This new postulate goes again the classical intuition behind the 86 two previous unemployment models by which unemployment is caused by wages above the marginal productivity. Rutherford and Light (2002) implement the Harris-Todaro hypothesis in CGE format. This kind of unemployment model is especially thought for developing countries to explain the transition from the countryside to the city. 3.3 Theoretical framework of the impact of tourism on an economy: TNT-T model In this section a new version of the TNT model (Sachs & Larrain, 1994) is developed to include the influence of tourism on the production and consumption of tradable and nontradable goods as well as the existence of a new possible equilibrium due to the underutilisation of labour and the capital accumulation process (named TNT-T model). There are goods that are tradable (goods that can be exported) and non-tradable goods (goods that cannot be exported such as, accommodation, catering services or a haircut). This simple difference in the kind of goods has important consequences on the economy. For instance, the economies tend to consume more tradable goods than they can produce in a boom situation, so that imports grow. In graphical terms (Figure 12), the economy consumes tradable goods in A (Ct0) and produce tradable goods in B (Qt0), the difference between A and B are the net imports. The consumption of non-tradable goods is Cnt0 and, given the characteristics of such goods, it coincides with the production (Pnt0). AD means aggregate demand, PPF possibilities production frontier, NT are non-tradable goods and T tradable goods. When the economy drifts into a degrowth situation, the economy moves downward along AD0 to D. During this process both the consumption of tradable and non-tradable goods decline. The lower consumption of tradable goods can be compensated by an increase 87 in net exports (difference between D and C). But, the fall in the production of non-tradable goods can only be borne by the domestic demand. Tourism sector is export intensive in the consumption of non-tradable goods such as accommodation, catering services and so on. The inclusion of tourism consumption in the model shifts the aggregate demand from AD0 to AD1 (the rise in the slope represents the higher propensity of tourists to consume non-tradable goods). Under these new circumstances, initially, the consumption of non-tradable goods increases from D to E. As soon as the tourism sector demands non-tradable goods and the economic degrowth passes away, the production and the demand increase. Thus, the total consumption goes from E to G. The difference between G (demand with tourism) and G1 (demand without tourism) represents the exports of non-tradable goods. So now, the necessary equivalence between domestic production and domestic consumption for the non-tradable goods no longer holds. On the other side, as it can be appreciated in Figure 12, the movement from E to G can only be achieved by reducing the production (and exports) of tradable goods (from C to H), although the demand increases from D to G. The difference between G and H represents the net exports. To conclude, the model demonstrates that the inclusion of tourism alleviates the fall in nontradable goods in degrowth scenarios and can act as a substitute of the domestic demand. Secondly, the model also shows that, to increase the production of non-tradable goods, the production of tradable goods has to decrease (win-lose situation). In consequence the exports of tradable goods decline. The TNT-T model here explained assumes full use of labor and capital. Now the TNT-T model will be relaxed to include more realistic assumption such as unemployment and capital accumulation that best described the current situation in Spain. 94 Finally, the capital accumulation process is as shonw in equation 19:    ,,1,11 , 1* * * ; 0; , hh time hh hh time hh time time hh hh hh time Capital Capital gos inv invendow r Capital hh time         (19) In this case, the time subindex is included to appreciate the way the capital accumulation works. The variable ,hh time Capital represents the capital accumulation by institutions (hh) (households, government and enterprises) and year (time). The capital accumulated at any time is formed by the capital of the previous year less the depreciation of capital [( ,1hh time Capital )*(1 hh  )] plus the gross operating surplus ( ,1hh time gos  ) in the first year plus the investment generated in the economic process (  1** time hh hh inv invendow r  ,the investment in the previous year is multiplied by the investment endowment by institutions (hh invendow ) and it is multiplied by the economic growth rate (r) plus the depreciation of capital ( hh  )). All elasticities used in the BLOVIFIS model have been taken from Hertel (1998). 3.5.2. Model closure The equations shown above are a brief summarise of the main equations of the BLOVIFIS model4. Other equations related to market clearance conditions and income balance conditions have been omitted as well as those related to the behaviour of households, government, enterprise and tourists. However, there are other equations whose assumptions have a deep impact on the results and they have to be highlighted. These equations compose the model closure and are related to government balance, investment-savings, unemployment and balance of payment.  Government balance: income equals expenditure (zero deficit). 4 The BLOVIFIS model is explained in detail in Annex 3.2 95  Investment-savings: total savings (domestic and abroad) equals investment. Additionally, any change in total savings in financed through capital flows (changes in capital account).  Unemployment: A wage curve is chosen to model unemployment.  Balance of payment: Current account equals Capital account. Changes in the Current account deficit are allowed. Spain is considered a small open economy. Thus, world prices are fixed exogenously. 3.5.3. Shock, scenarios and calibration path The shocks simulated are:  2% increase in tourist arrivals  10% increase in tourist arrivals The 2% increase in tourist arrivals comes from the United Nations World Tourism Organisation projections (United Nations World Tourism Organisation, 2013). On the other hand, the second shock tries to test the strength of tourism to provide economic growth and to reduce the unemployment rate assuming an unlikely tourist arrivals increase of 10%. A dynamic model is based on future projections of economic growth, interest rate and capital depreciation. These projections are used as stationary state of the economy over time. The scenarios projected are:  post-crisis: 5% interest rate, 0.7% GDP and 5% capital depreciation.  pre-crisis: 4% interest rate, 3% GDP and 5% capital depreciation. Both scenarios form the business as usual situation (BAU). The interest rate and the growth rate of the pre-crisis scenario are the average of the spread of the public debt and the average economic growth between 2008 and 2012. In regards to the post-crisis scenario, the projections are based on International Monetary Fund projection (International Monetary 96 Fund, 2013) and in the Economic Bulletin of the Bank of Spain (Bank of Spain, 2013). None of the scenarios projected match with the datasets. Hence, an “estibration” is applied to calibrate the data base according to the scenario projected (Balistreri & Hillberry, 2003). 3.6 Results 3.6.1. Unemployment A wage curve is the functional form chosen to model unemployment. This curve needs an elasticity of employment as input. Blanchflower and Oswald (1995) estimate an elasticity of employment but it is estimated under a normal economic situation context which is far from the current economic situation. Thus, two elasticities of employment are assumed to provide a better insight into the change in the unemployment rate depending on the elasticity considered: the elasticity obtained from Blanchflower and Oswald (1995) (0.1%) and a much more elastic one of 0.001% (elast) which may be more in accordance with the current unemployment situation in Spain. As it can be seen in Figure 15, elast allows for a higher reduction in unemployment as it was expected. With the 2% shock in tourist arrivals, both elasticities have a very similar effect on the unemployment rate in both scenarios. In a post-crisis situation, the impact on unemployment is very small. In the last year, the unemployment rate reduces 1.30% (25.74%) and 2.03% (25.47%) for Blanchflower´s elasticity and elast, respectively. This means around 76,651 and 119,948 new employments, respectively. In a pre-crisis situation, the unemployment rate reduces 1.15% (25.70%) and 1.92% (25.61%) for the fifth year, respectively. This means around 67,807 and 113,208 new employments, respectively. 97 With the 10% shock in the tourist arrivals, the differences in unemployment are higher with the elasticities and scenarios. In a post-crisis situation, in the last year, the unemployment rate reduces 7.53% (24.04%) and 12.03% (22.87%) for Blanchflower´s elasticity and elast, respectively. This means around 443,991 and 709,324 new employments, respectively. In a pre-crisis situation, the unemployment rate reduces 6.84% (24.22%) and 11.50% (23.01%) for the fifth year, respectively. This means around 403,306 and 678,074 new employments, respectively. Thus, none of both shocks mean a strong reduction in the unemployment rate, though it is a relief. The rest of the results of the paper are based on the elast elasticity. The post-crisis situation achieves better unemployment rates than a pre-crisis one. The apparent contradiction can be explained by the higher appreciation in the real exchange rate in the pre-crisis scenario than in the post-crisis one. Figure 15. Unemployment rate with change in employment elasticity’s respect to the BAU situation (%) Post-crisis Pre-crisis 3.6.2. Gross real added value (GRAV), foreign account deficit and capital accumulation According to Figure 16 and focusing on post-crisis situation, the increase in 2% in tourist flows boosts the GRAV from 0.05% in the first year to 0.44% in the fifth year. The unlikely 10% increase means a strong fostering to the economy from 0.25% in the first year to 2.58% 98 in the fifth year. During the five years, the accumulative growth of the GRAV is 1.21% and 6.81% for the 2% and the 10% case, respectively. On the other hand, the improvement in the current account deficit is similar to the change in GRAV. In the case of the 2% increase, the deficit reduces up to 0.06 % in the first year and up to 0.45% in the last year. In the 10% case, the reduction in the deficit goes from 0.28% to 2.75% in the last year. At the end of the fifth year, the deficit accumulates a reduction of about 1.27% and 7.35% for the 2% and the 10% shock, respectively. Regarding to the pre-crisis situation, the increase in tourist arrivals in 2% boosts the GRAV from 0.06% in the first year up to 0.55% in the last year. As in the post-crisis scenario, the 10% increase means a strong fostering to the economy; it goes from 0.28% in the first year to a 3.28% in the fifth year. During the five years, the accumulative growth of the GRAV is 1.53% and 8.63% for the 2% and 10% case, respectively. On the other hand, the improvement in the current account deficit is similar to the change in GRAV. In the case of the 2% shock, the deficit reduces 0.04% in the first year and 0.48% in the last year. These changes in the current account are slightly lower than in the post-crisis scenario. For the 10% case, the reduction in the deficit goes from 0.23% to 2.91% in the last year. At the end of the fifth year, the deficit accumulates a reduction of 1.32% and 7.66% for the 2% and 10% shock, respectively. The change in GRAV due to the increase in tourist arrivals is significant but modest for the 2% case in both scenarios. Nonetheless, the unlikely 10% increase produces a strong impulse to the economy. The increase in tourist arrivals boosts the real exchange rate which declines traditional exports and raises imports. The overall effect in the terms of trade (deficit reduction) is positive because of tourism. This result is in accordance with Adams and 99 Parmenter (1995), Copeland (1991) or Narayan (2004). This general improvement in the terms of trade is especially useful in the current economic situation in Spain in which the country has to repay the money borrowed during the economic boom. To sum up, the increase in GRAV reinforces the first hypothesis highlighted in the state of the art by which tourism is able to promote economic growth, despite of the lower technological use. On the other hand, tourism also enhances capital accumulation (Figure 16) but in modest rates which reinforces the conclusion of Capó et al. (2007) about the low productivity gains generated by tourism in the long term. 100 Figure 16. Change in gross real added value, current account deficit and capital accumulation respect to the BAU situation (%) Post-crisis Pre-crisis Gross real added value Gross real added value Current account (deficit) Current account (deficit) Capital accumulation Capital accumulation 101 3.6.3. Domestic demand The disentangling in tourism and non-tourism categories allows for a more accurate insight of the tourism effect on the economy as it is depicted on Figure 17. The tourism side -tof the domestic demand (resident consumption, government consumption and investment) declines due to the rise in domestic prices produced by the tourist arrivals. Although, at the same time, the economic boosting and the reduction in employment eased by tourist flows increase the domestic non-tourism -ntdemand. The change in domestic demand is very similar in both scenarios. The lower employment effect of tourist arrivals in the pre-crisis scenario because of the higher appreciation of the real exchange rate is compensated by a higher increase in investment due to a more optimistic economic growth. As a result, domestic demand practically converges in both scenarios. More precisely, in the case of the 2% shock, domestic tourism demand falls 0.48% in the first year and 1.79% in the last year in the post-crisis situation. On the contrary, domestic non-tourism demand increases 0.44% in the first year and 2.64% in the last year. These results begin to show the less positive side of the flow of tourist arrivals which will be explained in more detail in the next section. Up to now, the 10% shock has been used as a hypothetical upper bound scenario to provide a better insight into the strength of tourism to generate economic growth. Hereafter, The rest of sections are focused on the most likely 2% shock. 102 Figure 17. Changes in domestic demand respect to the BAU situation (%) Post-crisis Pre-crisis 3.6.4. Winners and losers: “Dutch Disease” From a macroeconomic perspective, the positive effect of the tourist arrivals in the economy has been already highlighted in the previous sections. However, at sectoral level, the tourism shock has diverse effects depending on the goods produced or the services provided, which, in last term, produces a reallocation of resources among the economic sectors. More precisely, tourism sector has two remarkable effects on the economy: it increases the demand of nontradable goods and, at the same time, it boosts the foreign exchange rate that undermines the exterior competitiveness of traditional exports. These results can be seen as a general consequence of the Dutch Disease. The term Dutch Disease was first used by the magazine “The Economist” in 1977 to explain the effect of the oil discovered in the sixties in the North Sea on the Dutch economy. Corden and Neary (1982) and Corden (1984) are the first that model it in academic terms. These papers differentiate among three sectors: booming sector, lagging sector and non-tradable sector. The booming sector begins to produce and export strongly. Such sector demand workers and capital from other sectors (lagging sectors and non-tradable sectors) to keep producing (resource effect). At the same time, the foreign income generated by the booming sector increases the real exchange rate that erodes the 103 exterior competitiveness of traditional exports; and, together with the increase in the demand of non-tradable goods produced by the income generated in the booming sector, cause the expenditure effect. This rise in non-tradable goods also increases the demand of workers and capital in such sector in disfavor of the lagging sector (another resource effect). As a result, it produces a de-industrialisation (lagging sector), a strong increase in domestic prices and in the real exchange that, eventually, falters the competitiveness and shrinks the economy. The Dutch Disease has been traditionally associated with oil exports countries such as Saudi Arabia, Qatar, Venezuela or Norway. Many of them avoid changing most of the revenues obtained from their oil exports into local currency to prevent the Dutch Disease. However, this economic illness can be generalised to any situation in which a country begins to receive an important amount of foreign money that trigger the consequences explained above. For instance, Laplagne, Treadgold and Baldry (2001), Usui (1996), VanWijnbergen (1986) and White (1992) show how developing countries can suffer from this illness due to the external aid received. Van Wijnbergen (1986) argues that the negative connotation of the term Dutch Disease should not hide an important trade theory behind it by which an economy tend to produce those goods that require an intensive use of the most abundant factor in the country (Heckscher-Ohlin trade theory). According to Roca (1998) the diagnosis of the Dutch Disease can be checked following four hypotheses: appreciation of the real exchange rate, decline in the exports of the lagging sector, decline in the outputs of the lagging sector and a likely increase in the outputs of the non-tradable sector. The first two hypotheses compose the expenditure effect, and the last two, the resource effect. Additionally, a fifth hypothesis should be added in regards to the resource effect: the likely increase in employment in the booming and non-tradable sectors and the respective likely declines in the lagging sectors (win-lose situation). 110 To conclude this section, the tourist arrivals unleash the Dutch Disease although its degree varies depending on the effect considered. On the one hand, the expenditure effect is quite reduced. Traditional exports decline because of the appreciation of the real exchange rate, but, with the exception of the tourism side of the agriculture commodities, the fall is not remarkable. On the other hand, the resource effect has remarkable and positive consequences on the production and the employment in the non-tradable sectors. The most tourism-oriented sectors such as accommodation, air transport or travel agencies are the most benefited from tourist arrivals. By contrast, traditional sectors decline their production. 3.7 Conclusions and further research Both the disentangling of the IOT into tourism and non-tourism categories and the recursivedynamic CGE model provide a detailed insight into the effect of the tourist arrivals on the Spanish economic. Besides, the recursive-dynamic CGE model also allows analysing the linkages among sectors and their consequences in the economic growth over time. In this ever-changing world, the relationship between tourism and economic growth is so fuzzy and complex that any conclusion in this regards should be taken cautiously. For instance, the advent of internet and new technology has converted some non-tradable goods in tradable ones such as accounting services. Countries such as India are already taking advantage of this new situation. These constant and unanticipated changes will make reconsider many theories and conclusions previously settled in the field. Keeping this in mind and based on the assumptions and projections established here, this paper affirms that tourism provides economic growth, reduces the unemployment rate, improves the terms of trade and boosts the domestic demand in the medium term in Spain. The effect on those macroeconomic variables varies in intensity. For highly indebted countries like Spain, the 111 improvement in the current account deficit is very beneficial in the short and medium term. The effect on real gross added value and employment is also positive. Albeit the effect on unemployment is not a solution, it is, however, a relief. The tourist arrivals have a double effect on the domestic demand. On the one hand, the domestic demand grows in its nontourism side due to the increase in the employment demand. On the other hand, the tourism side of the domestic demand declines due to the rise in domestic prices produced by the tourist arrivals. Nonetheless, at microeconomic level, the positive effects of the tourism sector are not as clear and positive as at macroeconomic level. Tourism sector seems to be a zero sum game. The assumption of unemployment and capital accumulation in the TNT-T model allows for the existence of a new equilibrium in which both tradable and non-tradable sectors win with tourism (win-win situation). Nonetheless, the posterior results show that tourism is guided towards a zero sum game rather than a non-zero one (win-lose situation). The tourist arrival fosters strongly the appreciation of the real exchange rate that erodes the potential advantages of the existence of a high unemployment and the capital accumulation generated by tourism. Thus, tourism-led growth undermines traditional sectors such as agriculture, energy and minery, and industry in a clear consequence of the Dutch Disease despite the economy is far from its production possibility frontier. In the medium term, its effects vary and are not especially harmful. Its negative consequences are overcome by the positive ones. However, in the long term, following this path, it could happen that the lasting erosion of traditional sectors may end up in a shrinking economic situation. Moreover, this potential vicious circle can be boosted by the low capital accumulation promoted by the labour intensive sectors in which tourism is based. Additionally, the empirical results support that a rise in tourism demand alleviates the fall in non-tradable goods that occur when an economy goes from a 112 booming situation towards a low growth situation in which the demand of non-tradable goods declines strongly because of its dependency of domestic demand. Further research may be oriented to improve the expectation in dynamic CGE models. For instance, the combination of both forward-looking and backward-looking behaviour in the same framework. Additionally, the forward looking behaviour should base their future decision in suitable economic indicators such as inflation index. Finally, there is a potential cause of economic depletion that should be taken into account in future research. The environment acts as a pull factor and its sustainability in the long term is vital for the economic growth in tourism-led countries. The lack of a suitable management in this regards could be as harmful as the Dutch Disease. Thus, both economic revenues and environmental sustainability are necessary for a steady and balanced economic growth. 113 CHAPTER 4 The role of climate and the tourism destination choice 4.1 Introduction The destination choice is the last step in a complex decision making process in which two kind of explanatory variables has to be combined: those related to individual characteristics (also known as choice invariant variables) and to destination characteristics (choice variant variables). In regards to the last one, the success of a tourist destination depends on many characteristics. Amongst them, climate is one of the most important factors (De Freitas, 2003). For some destinations, like those based on sun and beach, the reliability on the climate is so important that the climate conditions are vital (Capó, Riera & Rosselló, 2007) especially when such destinations rely on outdoor activities (cycling, fishing or walking), whereas for other destinations this is considered as a complement asset instead (Gómez-Martín, 2005). In this regard, the climate change means an increasing concern that should be addressed promptly. According to the Intergovernmental Panel on Climate Change (2007), the temperature is expected to rise up to 1.8ºC on average (scenario B1) during 21st century. This global increase in temperature could affect the current tourism flows toward traditional sun and beach destinations. Additionally, a tourism destination also relies on other attributes beyond climate endowments that cannot be neglected. One of these is the level of social and economic development of the country which is also positively perceived by the tourists (Table 13). For policymakers, the understanding and management of both climate and socioeconomic factors is key to become a successful tourism destination. 114 Table 13.Top ten most visited countries Millions of Ranking human France 83 24º Very high USA 67 6º Very high China 57.7 93º Medium Spain 57.7 29º Very high Italy 46.4 30º Very high Turkey 35.7 68º High Germany 30.4 17º Very high UK 29.3 21º Very high Russia 25.7 58º High Malaisia 25 59º High Source: World Tourism Organisation, International Monetary Fund and United Nations. On the other hand, the way that a tourist perceives the destination attributes is also conditioned by individual characteristics in the region of residence. The distance is one of these variables that is not equally perceived by individuals. Curiously, there is no consensus in the literature whether it affects positively or not to the destination choice. For some authors, the distance is seen as a restriction for the individual (Taylor & Knudson, 1976). On the contrary, for some others is perceived as a source of utility in itself (Baxter, 1979). But perhaps, the most conditioned variables by the region of residence is the climate. To tourists in cold regions, the climate in destination is seen as a push factor (tourists from colder regions tend to travel to warmer regions). On the contrary, climate can also act as a pull factor to tourists in warmer regions who tend to travel domestically (Eugenio-Martín & Campos-Soria, 2010). At the same time, individuals from colder regions may be willing to accept colder optimal climate conditions in destination in contrast to individual from warmer regions. These kinds of asymmetries in climate perceptions have to be suitably addressed in order to explain the destination choice of individuals. So far, literature has been not very prolific on climate and destination choice issues. The destination choice has been mainly based on aggregate data (Bigano, Lize & Tol, 2006a; or Madison, 2001) neglecting the differences in 115 the climate perceptions that occurs at individual levels. On the contrary, those who do it focus on few destinations or regions (Bujosa & Rosselló, 2013; Nicolau & Más, 2006). So, the destination choice process has not been addressed yet taking into account a wider set of destinations and combining both individuals and destinations characteristics. This paper estimates a random parameter logit model (mixed logit) where the assumption of individual random parameters permits a better and more accurate insight into the individual perception of the climate conditions. At the same time, individual and destination attributes are combined in the same framework in order to choose a sun and beach destination. Finally, two scenarios are simulated based on the temperature projections of the Intergovernmental Panel on Climate Change (2007). 4.2 State of the art 4.2.1. Destination choice: the last step in a complex decision process The destination choice is the last step within a complex individual decision process. EugenioMartin (2003) bases the decision process on five stages: participation choice, tourism budget constraint, frequency and length of stay, kind of destination and destination and transportation mode choice. The four first stages mainly depend on socioeconomic characteristics of the individuals whereas the last step relies on attributes of the destination. Despite the fact that two kinds of explanatory variables affect different decision levels and thus, they may be perceived separately, the five stages are part of a mutually related holistic decision. So far, none current methodology is able to combine the five stages in one single simultaneous decision making because of the difficulty of including both kind of explanatory variables in the same framework. So far, the literature has focused the research on each stage independently (Train, 1998; or Alegre & Pou, 2006) or combining several (Eymann & Ronning, 1992; or Eymann, 1995). Eugenio-Martín and Campos-Soria (2010) explain the 116 participation choice (first, whether travel or not and travel domestically or abroad) using socioeconomic attributes such as income, age or climate in the regions of residence. Nicolau and Más (2005b) model three tourism decisions: whether take a vacation or not, travel domestically or abroad and finally, taking single or multi-destination vacations. The literature in destination choice has been mainly based on aggregate data (Lize & Tol, 2002, Madison, 2001, Rosselló & Santana, 2014 or Bigano et al., 2006a), however this aggregate approach lacks of an accurate individual description of key destinations variables in which individuals mainly base their destination choice on (Gösslig & Hall, 2006). Moreover, they have been mainly focused on time series analysis (forecasting) (Song & Li, 2009; Smeral, 2009, 2010 or Bigano, Hamilton, Maddison & Tol, 2006b). At micro level, the destination choice process have been addressed by Haider and Ewing (1990), Morley (1994b), Bujosa and Rosselló (2013), and Nicolau and Más (2006). Among the myriad of attributes that affect the individual destination choice, certain mutually related variables have an outstanding importance and are key for the destination in order to attract tourism flows: Climate and distance. 4.2.1.1. The role of climate For instance, sun and beach destinations rely strongly on climate conditions. From a tourist´s perspective, people are likely to differ on the perceptions of such climate attributes and, moreover, it may be affected by climate conditions in their regions of residence (EugenioMartin & Campos-Soria, 2010). According to them, a warmer climate in the region of residence acts as a “pull factor” to travel domestically whereas a colder one acts as a “push factor” to travel abroad. Bigano et al. (2006a) also highlight the possible existence of asymmetric preferences depending on the climate of residence in the sense that, people from different regions share the same perception about the optimal temperature in destination but 117 they differ in the intensity of such climate sensitivity. The advent of the climate change has brought up a renewed interest in the role of climate in tourism (Rosselló-Nadal, 2014). The climate change may influence current tourism patterns. According to Bujosa and Rosselló (2013), Spain´s colder Northern provinces could benefit from a rising in temperature in detrimental of Southern provinces. The same shift in tourism flows is perceived by Moreno and Amelung (2009) due to the global warming, albeit they point out that Mediterranean regions will maintain an outstanding position over the next decades as tourism destination. The most common way to include the climate in tourism models has been through annual average data (Bigano et al. 2006a; Hamilton et al, 2005a,b; Hamilton and Tol, 2007). But, such dataset lacks of suitable information on individual perceptions about climate attributes which, in last term, influence the destination choice by the individual. Alternatively, other authors have used tourism climate index (TCI) (Mienczkowski, 1985) such as Nicholls and Amelung (2008), Amelung, Nicholls and Viner (2007), Amelung & Viner (2006) or EugenioMartín and Campos-Soria (2010) but they are not destination choice models. These kinds of indices are based on revealed preferences and gives subjective weights to the terms included in the index. The last approach to climate variables consist in micro data based on stated preferences in which climate attributes are included as explanatory variables (Bujosa & Rosselló, 2013 and Nicolau & Más, 2006). 4.2.1.2. The role of distance For some authors, the distance affects negatively the destination choice (Mckercher & Lew, 2003). According to them, there are tourism exclusion zones (ETEZ) where little or no tourism occurs. Greer and Wall (1979) argue that the higher the distance, the higher the supply of recreational activities to compensate it. Moreover, the price has a deterrent effect and it is negatively correlated with the distance (Nicolau & Más, 2006) but its effect varies depending on the purpose of the study. According to them, good climate and cultural 118 attributes alleviate the negative effect of prices and distance. This finding shows that the perception of the distance varies among individuals and the aim pursued. On the contrary, other authors suggest that the distance may be a source of utility. According to Baxter (1979), the travel itself is part of the tourism utility and thus, it provides satisfaction. In the same line, Wolfe (1970) and Wolfe (1972) hold that, after passing a certain threshold, distance becomes a positive attribute. There is also a complementary way to describe the role of distance. The importance of the distance between markets and suppliers was first remarked by von Thünen (1826). According to him, suppliers follow a spatial pattern around a market depending on the characteristic of the goods provided. According to Von Thünen´s findings, perishable goods such as vegetables are closer to the market, whereas grain foods such as wheat are further from it. The factor behind this spatial pattern is known as Von Thünen rent or spatial rent (rents that can only be enjoyed in specific geographical places). In the case of tourism, tourists (market) may be more willing to travel further to enjoy better spatial rent (better climate) which is in accordance with Greer and Wall (1979) or Nicolau and Más (2006). Thus, a negative relationship between distance and climate should be expected. For a sun and beach destination, this means that the warmer the destination, the further the potential target markets that can be achieved. In this sense, the climate change could be also seen as a spatial rent generator. This relationship between climate and distance is also shaped by the influence of socioeconomic variables such as income. In the sense that, the distance that the individual is willing to assume can be boosted by higher incomes. 4.2.1.3. Destination choice methodologies: all roads guide to mixed logit As it has been already highlighted in the beginning of the section, the destination choice is a complex decision process which is mainly based on two kinds of variables that not many current methodologies are able to appropriately deal with. On the one hand, people choose 119 their destination based on destination variables (choice variant variables) such as temperature, precipitation, distance, recreational amenities, natural recreational resources or level of social and economic development. On the other hand, personal variables (choice invariant variables) such as temperature and precipitation in origin or income may also influence the destination choice. Moreover, some of these attributes such as temperature or precipitations are differently perceived by tourists. Quoting Nicolau and Más (2005, p. 56): “…it is highly unlikely that the whole sample (….) has the same set of parameter values which implies the need to consider unobserved heterogeneity of tourist in parameter estimation”. Such climate attributes are intimately linked by the existence of a certain trade-off between both; which, lastly, shape the apprehension of the climate by the individual. So, it reflects the existence of a balance in which temperature and precipitation cannot change independently of each other to achieve higher welfare levels (overriding effect). Other climate attributes such as sunny days, humidity or windy days may be also important in the destination choice but vary randomly during the day and thus, they cannot be accurately anticipated. Among the set of methodologies, discrete choice models (microeconometric modeling) allow for a closer insight into individual behavior decision (Train, 2009). The logit/probit model is the first model within the discrete choice family. This is the methodology chosen by Seddighi and Theocharous (2002) and Morley (1994a) to model the decision of revisiting Cyprus and visiting Sydney, respectively. Nevertheless, this approach is not enough when several destination choices are taken into account. Under these circumstances, the destination choice can be modeled assuming a multinomial logit model in which each destination has a categorical value (Morley, 1994b). On the side of the explanatory variables, the existence of both individuals (choice invariant) and destinations (choice variant) explanatory variables reduces the possibilities to the conditional multinomial logit model. Nonetheless, the conditional multinomial logit cannot suitably deal with many alternatives (destinations) and 126 attributes depending on the climate attributes in origin. Thus, if current cold regions become warmer due to the climate change, people from these regions will be more sensitive to temperature as people from current warm regions. This relationship is in accordance with other authors findings such as Bigano et al. (2006a), although they reduce the asymmetric effect to the temperature squared only. The distance negatively affects the tourism destination choice but the income (logarithm of the income) reduces the negative impact of the distance (0.76293). As it was already explained in the previous section, the standard deviation is no longer independent because of the correlations assumed among the random parameters. Thus, the larger the covariance, the greater the relationship between the parameters. In the case of precipitations, the share of the total standard deviation explained by itself is 0.01632, the rest of the total standard deviation is explained by the interaction with temperature (-0.01920), temperature squared (-0.74488D-4) and distance (-0.00231). In the case of precipitation and temperature, this standard deviation means that the individuals with larger sensitivities to precipitations are likely to have lower sensitivities to temperature, until the optimal temperature is achieved (temperature squared). In the case of temperature and distance, the individuals with larger sensitivities to the temperature are less concerned about the distance in their utility function up. This relationship between distance and climate reinforces the hypothesis of Von Thünen (1826) by which people are willing to assume higher distance in order to enjoy higher spatial rents (warmer sun and beach destinations). The relationship between distance and temperature and precipitations is also in accordance with Nicolau and Más (2006). 127 Variables such as road paved, secondary education, life expectancy and ppp_rate reflect that tourists prefer more developed countries but, at the same time, cheaper than their regions of residence. These results are important for policymakers because, in contrast to climate variables, the destination can act over these variables in order to attract tourism (EugenioMartin, Martin-Morales & Sinclair, 2008). Table 16 shows the changes in the tourism flows for the main tourism destinations produced by the increase in 1.8ºC in the temperature (Intergovernmetal Panel on Climate Change, 2007). According to scenario A, with a 1.8ºC increase in temperature in the main outbound tourism countries (Germany and UK), all current sun and beach destinations decrease their share in total tourism flows. For instance, France and Spain decrease their share 5.76% and 6.70%, respectively. In scenario B, all main destinations increase their share, but UK and Germany benefit most from such general increase in temperature. 128 Table 16. Changes in tourism flows produced by changes in temperature in the destination Scenario A: +1.8 ºC in Germany and UK Scenario B: +1.8 ºC in all destinations Base share (%) Change in share (%) Change in share (%) Croatia 1.88 -1.77 0.76 Cyprus 2.67 -1.01 0.93 France 7.56 -5.76 4.96 Greece 10.09 -4.94 0.40 Italy 8.04 -7.78 2.96 Malta 4.39 -3.17 1.15 Middle east 2.59 -1.47 2.52 Portugal 4.31 -2.82 1.84 Spain 9.46 -6.70 1.05 UK 6.89 24.99 7.07 Germany 6.52 36.81 6.55 4.6 Conclusions and further research The modelling of the destination choice is challenging. All methodologies have their pros and cons, but, as far we know, mixed logit is probably the model that best fits the destination choice because it combines both origin and destination attributes. The asymmetric climate perceptions by individuals have been tested and confirmed. People from colder regions are more willing to travel abroad than those from warmer regions. People are expected to change their tourism pattern due to the Climate Change. Current outbound tourism countries will benefit most from such increase in temperature. Attributes related to the level of development (road paved, year of education or life expectancy) are positively perceived by the tourist. These variables are more manageable by 129 policymakers although the "revenues" come in the long term. This result demonstrates that sun and beach destinations are more than environmental endowments. Thus, tourism-led growth may require a certain level of economic development to really take advantage of it. As it was mentioned in the paper, the way that climate is included in the tourism model often lacks of a suitable representation. Neither aggregate data, nor current climate indices properly capture the climate effect. As soon as society takes more concern about the impact of climate and environment, more micro data will be available. Despite that, aggregate data are expected to keep maintaining an outstanding position in tourism research. Under these circumstances, the development of a new tourism climate index based on objective weights and in stated preferences will help to have a closer insight into the response of tourists to the climate. In consequence, the use of this climate index as variable in macro data will also help to better capture the climate change impact. 130 131 Annex 3.1. TSA-IOT and SAM A.3.1.1 Introduction This chapter is oriented to explain the elaboration of the dataset necessary to be used in the computable general equilibrium (CGE) model (chapter 3). The methodology is based on the Social Accounting matrices (SAMs). At the same time, the SAM is formed by the inputoutput tables (IOT) and the national accounts (Annual Accounts by Institutional Sectors (AAISSS)). The IOT gather up the information related to the economic flows in the economy for one single year. While the AAIISS collect information on how the resources are generated and used in the economic process (IOT) and how they are subsequently distributed among institutions. However, the aim pursued in the chapter devoted to the CGE model requires a deeper representation of the tourism sector that the IOT is not able to provide. Quoting Dwyer, Forsyth and Dwyer (2010, p. 239): “the problem with measuring the economic significance of tourism spending is that “tourism” does not exist as a distinct sector in any system economic statistic or of national accounts”. In other words, there is not an industry named “tourism”. According to Blake, Durbarry, Sinclair and Sugiyarto (2001), the IOT framework overestimates the total GDP effect and underestimates the total effect on the tourism sector. The Tourism Satellite account (TSA) is the dataset able to fill the lack of tourism information in the IOT. Thus, one of the aims of this chapter is to combine the IOT and the TSA to form a new IOT. On the other hand, and as it was already mentioned, a SAM requires also information about the income distribution by institutions. The AAIISS provide such information but it has to be managed and organised in a suitable manner to create a SAM. 132 This chapter is divided in two parts: first, the TSA and the IOT are combined to create the new IOT. Second, the information gathered up in the AAISSS is also included with the new IOT to elaborate the SAM. A.3.1.2 Integration of the IOT and the TSA This study is based on two datasets: the Input-Output table (IOT) and the Tourism Satellite Accounts (TSA) for Spain in 2006. These two datasets are combined to construct a new IOT whose main purposes are: first, to differentiate between the tourism sector and the rest of the economy and second, to disentangle all goods and services between tourism and non-tourism categories. The IOT is the framework in which the TSA is included. Figure 19 shows the scheme followed in the elaboration of the data base. Figure 19. Scheme of the datasets A.3.1.2.1. Input-Output Table accounting framework “The IOT are double entry matrices which represent the economic relations or flow of goods of an economy during a period of time” (Requeijo, Iranzo, Martínez de Dios, Pedrosa & Salido, 2007). The Spanish IOT accounting framework is adapted to the European System of Accounts (ESA-95). The IOT can be considered an extension of the national accounts (Muñoz, Iráizoz & Rapún, 2008). The IOT accounting framework provided by the Spanish Statistical Institute (INE) is formed by the following inputs-outputs tables: 133 Supply table at basic and purchase prices: The Supply table gathers up data about production by goods and activities and imports, trade margins, transport margins and tax on products. The sum of production by activities and imports provides the total supply at basic prices. Including trade margins, transport margins and tax on products, so that the supply table at purchase prices is obtained. Use table at basic and purchase prices: The use table gathers up data about intermediate demand by goods and activities and final demand: final consumption demand (households, non-profit private institutions and government), Gross capital formation (GCF) (gross fixed capital formation (GFCF) and inventory changes (CHINV)) and exports by products. The sum of intermediate demand by activities, final demand and exports provides the total uses. If the net taxes on products are included, the use table at purchase prices is obtained, otherwise, the use table is at basic prices. Domestic use table at basic prices: The domestic table is equal to the use table but it includes information about domestic origin goods only. Imported use tables at basic prices: The domestic table is equal to the use table but it includes information about imported goods only. In order to keep the accounts balanced, any IOT keeps certain equivalences: 134 75 2 2 ,, , 11 1 75 3 2 2 ,,,, 11 11 st st st st ai iwld mi i awld m utp utp utp utp ai hi f i iwld ah fwld prod imports margins nettaxes intermediate finalconsumption GCF exports           (iot.1) 118 118 4 ,, , 11 1 s tututut ai ai a a j ii j p rod intermediate othertaxes VA      (iot.2) ,, , ,, , ,, , ,, ut mut dut ai ai ai ut mut dut hi hi hi ut mut dut fi fi fi ut dut iwld iwld intermediate intermediate intermediate finalconsumption finalconsumption finalconsumption GCF GCF GCF exports exports     (iot.3) being: ,,,,st utp ut dut mut Index for supply table, use table at purchased prices, use table, domestic use table and import use table respectively a Index for activities from 1 to 81 i Index for commodities from 1 to 118 j Index for gross wages, social contributions, net taxes on products and gross operating surplus , s t ai p rod Production by activity a and commodity i (supply table) , s t iwld imports Imports from the European Union (EU) and rest of the world (ROW) by commodities (supply table) , s t mi margins Trade and transport margins by commodities (supply 135 table) s t i nettaxes Net taxes on products by commodities (supply table) , ut ai intermediate Intermediate demand by activities and commodities (use table) , mut ai intermediate Intermediate demand by activities and commodities (import use table) , dut ai intermediate Intermediate demand by activities and commodities (domestic use table) , utp hi f inalconsumption Final consumption by households, private non-profit institutions and national government by commodities (use table) , mut hi f inalconsumption Final consumption by households, private non-profit institutions and national government by commodities (import use table) , dut hi f inalconsumption Final consumption by households, private non-profit institutions and national government by commodities (domestic use table) , ut f i GCF Gross capital formation (GFCF) + change of inventories (CHINV) by commodities (use table) , mut f i GCF Gross capital formation (GFCF) + change of inventories (CHINV) by commodities (import use table) , dut f i GCF Gross capital formation (GFCF) + change of 238 adicional respecto a los no CBC. Su importancia es fundamental, porque pone a prueba si los viajeros CBC gastan de manera diferente que los viajeros no CBC, que no es sino la pregunta a responder en este artículo. La tabla muestra que 35 de las 62 estimaciones son significativas, lo que significa que los viajeros CBC, para estas combinaciones de alojamiento y régimen de comidas, son diferentes de los viajeros no CBC. Sin embargo, los resultados de la forma estructural no se pueden utilizar para medir el impacto directo de cada perfil, sino para probar la dirección del impacto y la significatividad. Con el fin de medir el impacto directo, es necesario obtener estos resultados mediante la forma reducida como se muestra en la tabla 1.3. 239 Tabla 1.2: resultados forma estructural (Parte II) No CBC Ecuación en origen Ecuación en destino Parámetros Desv. std Parámetros Desv. std Sólo vuelo Hotel 5 estrellas -7,.606 (8,862) 130,161*** (6,950) Hotel 4 estrellas 14,800** (6,825) 90,253*** (6,651) Hotel de 1.2 o 3 17,171*** (6,458) 73,906*** (6,276) Apartmento 16,103*** (5,680) 65,531*** (5,709) Familia y amigos 25,071*** (4,992) 44,614*** (5,398) Otros 27,826*** (5,406) 49,915*** (5,872) Vuelo y alojamiento Hotel 5 estrellas 102,242*** (7,306) 72,965*** (12,194) Hotel 4 estrellas 75,679*** (5,402) 47,179*** (9,028) Hotel de 1, 2 o 3 56,268*** (5,353) 49,695*** (7,786) Apartmento 53,837*** (5,216) 48,885*** (7,576) Familia y amigos 51,411*** (5,431) 46,114*** (7,297) Vuelo, alojamiento y desayuno Hotel 5 estrellas 114,416*** (6,140) 57,306*** (12,255) Hotel 4 estrellas 89,794*** (5,544) 49,033*** (10,086) Hotel de 1, 2 o 3 70,638*** (5,504) 47,437*** (8,704) Apartmento 66,021*** (5,388) 46,611*** (8,391) Familia y amigos 83,573*** (13,895) 30,455** (12,910) 240 Tabla 1.2 (continuación): resultados forma estructural (Parte II) No CBC Ecuación en origen Ecuación en destino Parámetros Desv. std Parámetros Desv. std Vuelo, alojamiento y media pensión Hotel 5 estrellas 127,901*** (5,373) 35,821*** (12,315) Hotel 4 estrellas 90,219*** (5,014) 35,368*** (9,572) Hotel de 1, 2 o 3 71,111*** (5,075) 38,490*** (8,404) Apartmento 72,649*** (5,097) 37,799*** (8,478) Familia y amigos 77,778*** (9,290) 38,801*** (10,334) Vuelo, alojamiento y pensión completa Hotel 5 estrellas 129,789*** (5,914) 29,751** (12,363) Hotel 4 estrellas 92,318*** (5,056) 31,847*** (9,567) Hotel de 1, 2 o 3 80,284*** (5,354) 33,219*** (8,897) Apartmento 71,242*** (5,676) 34,092*** (8,427) Familia y amigos 129,932*** (5,687) 22,668* (12,160) Vuelo, alojamiento y todo incluído Hotel 5 estrellas 125,542*** (5,194) 21,548* (11,597) Hotel 4 estrellas 99,646*** (4,724) 22,534** (9,727) Hotel de 1, 2 o 3 89,705*** (4,792) 25,560*** (9,171) Apartmento 82,958*** (4,923) 29,942*** (8,891) Familia y amigos 136,551*** (5,551) 14,949 (12,291) 241 Tabla 1.2 (continuación): resultados forma estructural (Parte II) CBC Ecuación en origen Ecuación en destino Parámetros Desv. std Parámetros Desv. std Sólo vuelo Hotel 5 estrellas 1,156 (5,086) -11,154*** (3,652) Hotel 4 estrellas -1,.386 (3,192) -12,686*** (2,316) Hotel de 1, 2 o 3 3,858 (3,731) -8,354*** (2,664) Apartmento -2,086 (1,908) -1,446 (1,379) Familia y amigos -4,040*** (1,239) 2,127** (0,907) Hotel 5 estrellas -4,443* (2,559) 1,667 (1,850) Vuelo y alojamiento Hotel 5 estrellas -0,897 (5,161) -16,616*** (3,726) Hotel 4 estrellas -8,813*** (2,112) 0,878 (1,623) Hotel de 1, 2 o 3 -9,360*** (1,782) 2,416* (1,380) Apartmento -7,815*** (0,721) 1,704** (0,690) Familia y amigos -3,971 (2,607) 1,119 (1,888) Vuelo, alojamiento y desayuno Hotel 5 estrellas -10,057*** (2,732) 2,299 (2,054) Hotel 4 estrellas -10,735*** (2,151) 1,717 (1,701) Hotel de 1, 2 o 3 -6,020* (3,226) 5,539** (2,318) Apartmento -10,409*** (2,749) -0,433 (2,117) Familia y amigos -24,301 (16,176) 24,828** (11,634) 242 Tabla 1.2 (continuación): resultados forma estructural (Parte II) CBC Ecuación en origen Ecuación en destino Parámetros Desv. std Parámetros Desv. std Vuelo. alojamiento y media pensión Hotel 5 estrellas -9,156*** (2,639) 1,914 (1,981) Hotel 4 estrellas -8,990*** (1,085) 3,986*** (0,909) Hotel de 1, 2 o 3 -2,623 (2,001) 0,427 (1,444) Apartmento -9,585*** (2,265) 5,225*** (1,681) Familia y amigos -3,286 (11,397) -8,374 (8,212) Vuelo, alojamiento y pensión completa Hotel 5 estrellas -14,699** (7,224) 10,954** (5,222) Hotel 4 estrellas -3,556 (2,693) -0,034 (1,953) Hotel de 1, 2 o 3 -2,614 (4,199) -0,471 (3,028) Apartmento -4,008 (4,300) 7,506** (3,083) Familia y amigos -8,843 (7,152) 5,843 (5,171) Vuelo, alojamiento y todo incluído Hotel 5 estrellas -10,666*** (3,727) 0,679 (2,770) Hotel 4 estrellas -8,748*** (1,036) 3,123*** (0,873) Hotel de 1, 2 o 3 -13,828*** (1,557) 4,150*** (1,370) Apartmento -14,490*** (1,773) 5,059*** (1,491) Familia y amigos -20,593*** (5,716) 7,985* (4,276) 243 Resultados de la forma reducida Los resultados en forma reducida son una transformación conveniente de los resultados de la forma estructural con el fin de obtener el "verdadero" impacto directo de cada variable exógena. Dicha transformación se aplica a las variables dicotómicas referentes a CBC y no CBC que se mostraron en las tablas anteriores. Las elasticidades ingreso se puede obtener de la forma reducida. En particular. la elasticidad del ingreso con respecto al gasto en origen es de 1.74, mientras que en destino tal elasticidad es de 1.98. La Tabla 1.3 muestra cuánto de más o de menos está gastando cada perfil CBC en origen y destino. Este resultado se pondera por las noches media y la composición media del grupo con el fin de obtener una cifra más cercana a la que se enfrenta cada turista. Para los cuatro perfiles más relevantes los resultados son similares. Todos ellos ahorran dinero en origen con respecto a los turistas no CBC. En particular, las cifras de ahorro varían entre € 179,97 y € 70,97 en media por grupo y noche. Esto demuestra que para los perfiles más populares, los Tabla 1.3: Resultados forma reducida: El impacto de las CBCs con respecto a las no CBC por estancia media y tamaño medio del grupo (euros) Origen - Destino 5 * 4 * 3, 2, 1* Apartmento Familia y amigos Otros Sólo vuelo (F) O D -92,36 -188 , 27* -187,05 -274 , 78* -16,46 -52 , 78 -84,81 -52 , 78 -70,97* 35 , 38* -99,72* 24 , 74 (F) + Alojamiento (A) O D -196,67 -310 , 80* -167,03* -14 , 54 -173,60* 15 , 64* -179,97* 7 , 52* -106,30 12 , 76 (F+A) + desayuno O D -145,82* 7 , 91 -159,24* -3 , 81 -40,22* 64 , 31* -208,11* -45 , 35 -195,27 407 , 73* (F+A)+ media pensión O D -169,31* 5 , 30 -136,13* 48 , 11* -47,31 -0 , 99 -146,99* 78 , 45* -168,91 -185 , 13 (F+A)+ pensión completa O D -138,67* 138 , 56* -66,16 -12 , 58 -54,47 -17 , 83 9,38 165 , 83* -112,71 88 , 01 (F+A)+ todo incluído O D 202,91* -24 , 86 -152,33* 33 , 78* -248,29* 35 , 68* -285,26* 60 , 04* -362,25* 96 , 59* * Significa que la variable dicotómica original de la table 1.2 es significativa. Los valores en negrita representan a los perfiles más significativos 244 turistas CBC gastan más dinero que los turistas no CBC. Este gasto superior varía en media entre € 48,1 y € 7,52 por grupo y noche. Esta cifra confirma la hipótesis de que el ahorro de los turistas CBC en origen es transferido, al menos parcialmente, a un mayor gasto en destino. Las Islas Canarias, como muchos otros destinos turísticos de todo el mundo, se han enfrentado a un cambio de la estructura del mercado relevante. Por una parte, la percepción de ahorrar dinero con las tarifas aéreas más baratas de compañías CBC puede animarles a gastar más dinero en el destino. Por otra parte, las aerolíneas CBC pueden aumentar el tráfico aéreo hacia un destino en particular. Este documento prueba si la primera hipótesis es verdadera. A tal efecto, se considera un sistema de ecuaciones de gasto en origen y en destino. Dentro de todos los métodos econométricos que pueden estimar dicho sistema, se elige el modelo MC3E porque es capaz de hacer frente a la endogeneidad y correlación contemporánea de los errores apropiadamente. Otro tema para el destino está relacionado con una redistribución de la relevancia de cada perfil turístico debido a la presencia de aerolíneas de bajo coste. Los viajeros CBC pueden estar dispuestos a permanecer en diferentes tipos de alojamiento o disfrutar de paquetes turísticos más simples con respecto a los viajeros tradicionales (no CBC). Por lo tanto, puede implicar un efecto de redistribución de perfiles turísticos dentro de un destino. En las Islas Canarias, los perfiles de turistas CBC que experimentan un crecimiento significativo son "Sólo vuelo + alojarse con amigos o familiares", "Vuelo + Estancia en apartamento", mientras que los perfiles de turistas que reducen su presencia son "Vuelo + Alojarse en 4 estrellas con media pensión" y "Vuelo + Alojarse en hotel 4 estrellas con todo incluido". También es importante tener en cuenta que la duración media de la estancia también es diferente entre los turistas CBC no CBC. Por ejemplo, para el caso de " Sólo vuelo + alojarse con amigos o familiares", los turistas CBC se quedan, en promedio, 2 días menos que los turistas no CBC. 245 Sin embargo, para el resto de perfiles turísticos relevantes, los turistas CBC se quedan, en promedio, 1 día menos que los turistas no CBC. Capítulo 2: La decisión de recorte de los turistas en tiempos de crisis económica Introducción y objetivos Desde 2008, la UE-27 se encuentra en una situación económica de recesión. En promedio, el crecimiento del PIB real ha disminuido un 0,16 % entre 2008 y 2012 y la tasa de desempleo ha aumentado del 7,6 % en 2008 hasta el 10,6 % en 2012. El efecto y las consecuencias de la crisis han sido diferentes según los países. Por un lado, un país como Alemania ha crecido un 0,8 % en promedio entre 2008 y 2012 e incluso ha reducido su tasa de desempleo del 7.5 % en 2008 al 5,9% en 2012. Por otro lado, países como España o Grecia han sufrido una crisis aguda (una reducción del -0,92 % y -4,34 % en el PIB real desde 2008 hasta 2012. respectivamente) y han mostrado tasas de desempleo durante los últimos cuatro años de 11,3% en 2008 al 25% en 2012 y del 7,7 % en 2008 al 24,3% en 2012 para España y Grecia, respectivamente). La crisis también ha provocado una crisis de deuda en Grecia y Portugal, y bancaria en España. Irlanda y Chipre. A nivel microeconómico, esta situación de crisis tiene un efecto profundo sobre la renta disponible individual y, por tanto, en el consumo total. Bajo estas circunstancias, el consumo turístico es especialmente sensible al recorte en el gasto turístico debido a su alta elasticidad renta (Lanza, Temple y Urga, 2003). Según Riley, Ladkin y Szivas (2001), la actividad turística necesita de una correcta previsión de la demanda para hacerla coincidir con la oferta. Por tanto, la anticipación es clave para el éxito en la actividad turística. Los gestores turísticos y políticos necesitan más información sobre cómo reaccionar en situaciones de crisis económicas. No obstante, hay una falta de indicadores y de información adecuados sobre el 246 comportamiento del turismo en situaciones de crisis económica (Sheldon y Dwyer, 2010; Smeral, 2010; y Bronner y Hoog, 2012). Las consecuencias de esa falta de conocimiento han sido ya estudiadas en la literatura. Según Okumus y Karamustafa (2005), ni el gobierno ni las empresas turísticas de Turquía fueron capaces de hacer frente a la crisis económica que sufrieron en 2001. O'brien (2012) señala que la falta de interacción entre el gobierno y el sector privado explica que el sector del turismo en Irlanda no esté creciendo todavía, mientras que otros destinos europeos ya han vuelto a crecer a pesar de la crisis económica. Hasta el momento, los gestores turísticos y políticos han basado principalmente su análisis en las llegadas y el gasto. Como se puede observar en la Figura 2.1, en 2008 los turistas ajustan inmediatamente su gasto ante el inicio de la crisis, mientras que el número de llegadas sigue creciendo. A medida que la crisis continúa, los turistas empiezan a reducir las llegadas y aumentar el gasto. Por último, desde 2010, llegadas y gasto caen abruptamente. Si bien las llegadas y el gasto muestran una relación clara, si bien la mayor parte de la literatura no las has analizado conjuntamente. Por lo tanto, una mejor comprensión y análisis de la relación mutua entre la demanda y la oferta podría permitir aclarar qué parte del cambio en el gasto turístico se debe a cambios en las llegadas y qué es debido a cambios en los precios. 247 Figura 2.1: Llegada de turistas y gasto Fuente:Eurostat y Organización Mundial del Turismo Para resolver este problema, en este trabajo se separan las llegadas y el gasto en argumentos más explícitos a nivel microeconómico. Este documento se centra en los fundamentos de la decisión de recorte en gasto turístico de los hogares y cómo se lleva a cabo esta decisión durante la crisis económica mundial surgida en 2009 en la Unión Europea. Para tal fin, la decisión de recorte se divide en dos niveles. En primer lugar, los turistas deciden si recortar o no. En este punto, el gasto de los hogares podría ser la variable natural para ser utilizada (véase, por ejemplo, Melenberg y Van Soest, 1996). Sin embargo, el gasto de los hogares puede variar por varias razones no todas relacionadas con la crisis económica. Para evitar este sesgo potencial, una variable de respuesta binaria se utiliza como variable endógena, en la cual se pregunta si han tenido que recortar en gasto turístico o no a causa de la crisis. En segundo lugar, para los turistas que efectivamente recortaron, se les pregunta por su estrategia de recorte de acuerdo a seis alternativas: "menos vacaciones", "reducción de la estancia", "medio más barato de transporte", "alojamiento más barato", "viajar más cerca de casa" o "cambiar el período de viaje". La Tabla 2.1 muestra cómo estas alternativas podrían afectar a la llegada y al gasto turístico. Por ejemplo, la decisión de recortar teniendo "menos 254 Tabla 2.2: Determinantes de la decision de recorte (odds-ratios) Variable Recorte ¿Cómo se recorta? Reducir estancia Tranporte más barato Alojamiento más barato Viajar cerca de casa Cambiar periodo de viaje PIB pc (PPA) 0,999*** Crecimiento 0,963*** Socieconomic variables Variables de empleo: Agricultor, guardabosque, pescador 0,491*** Dueño de una tienda 0,660*** Autónomo 0,614*** Jefe 0,525*** Otro tipo de autoempleo 0,666*** Profesionales empleados 0,598*** Mando superior 0,484*** Mando intermedio 0,542*** Funcionario 0,544*** Oficinista 0,666*** vendedor 0,636*** Otros empleados 0,551*** Supervisor 0,915 Artesano 0,950 Trabajo artesanal no cualificado 1,002 Otro trabajo artesanal 0,512*** Ama de casa 0,809** Estudiante 0,622*** Jubilado 0,658*** Otros no empleados 0,710** Edad 1,039*** 1,003 0,986*** 0,987*** 0,997 0,992** Edad al cuadrado 0,999*** Género (masculino = 1) 0,871*** 1,214** 1,016 1,088 1,204** 0,872 Nivel de educación 0,986*** Variables regionales Clima 1,079*** 1,024** 0,951*** 1,031** 0,965*** 0,964** Costa 1,049 0,908 0,975 1,091 0,767** Aeropuerto 0,898 0,940 1,003 1,133 0,840 Variable latente: L 1( restringido) 1,440 1,142 2,753*** 1,449 1,349 255 En relación a la ecuación del tipo de recorte, la edad influye negativamente sobre las decisiones de: "transporte más barato", "reducir la estancia" y "cambiar el período de viaje". En este último caso, es un 0,8% menos probable reducir los viajes fuera de temporada alta, cuando aumenta la edad respecto a tomar “menos vacaciones” (categoría de referencia). En otras palabras, las personas más jóvenes están más dispuestas a viajar fuera de temporada alta, utilizar el transporte más barato o alojamientos más económicos que las personas mayores. Para las otras dos alternativas, la edad no tiene ninguna influencia significativa en comparación con la categoría de referencia "menos vacaciones". Un razonamiento similar se podría hacer sobre el género. Éste afecta positivamente a la "reducción de la estancia" y "viajar cerca de casa", pero no tiene una influencia significativa en las otras alternativas. Dado que recorta, es un 20,4 % más probable viajar "cerca de casa" para un hombre que para una mujer. Por otra parte, es interesante observar que en cuanto mejora el clima en el lugar de residencia, la probabilidad de elegir "reducir la estancia" y "alojamiento barato" aumenta sobre la probabilidad de "menos vacaciones". Por ejemplo, con una unidad de incremento en el índice de clima, es un 2,4% (1,024 menos 1) más probable reducir la duración de la estancia en lugar de optar por un menor número de días de vacaciones. Esta reducción puede ser explicada porque los hogares ubicados en las regiones con mejores condiciones climáticas tienen una mayor probabilidad de viajar a nivel nacional (Eugenio-Martín y Campos-Soria, 2010) de modo que pueden reducir el número de días más fácilmente. En el caso de "un alojamiento más barato", es 3,1 % más probable (1,031 menos 1) elegir esta opción en lugar de tomar “menos vacaciones”. En resumen, las personas que viven en regiones con buen clima prefieren la reducción de la duración de la estancia o reservar un alojamiento más barato en lugar de optar por un menor número de días de vacaciones. Para el resto de las alternativas de recorte, las probabilidades disminuyen cuando las condiciones climáticas de origen mejoran. 256 Por ejemplo, tan pronto como mejora el clima es un 3,6% menos probable recortar viajando fuera de temporada alta en vez de tener un menor número de días de vacaciones. La presencia de la costa en el lugar de residencia sólo tiene influencia en la decisión de cambiar el "período de viaje", mientras que para las otras alternativas no existen diferencias significativas en comparación con tomar menos vacaciones. La presencia de la costa hace un 23,3 % (1 menos 0.767) más probable recortar a través de la duración del viaje que optando por un menor número de vacaciones. El análisis posterior de la estimación permite analizar cómo las probabilidades estimadas cambian con algunos determinantes clave como el índice de clima o la edad del cabeza de familia. La figura 2.2 muestra la media móvil de la probabilidad del tipo de recorte en relación al índice de clima. De acuerdo con dicha figura, existe un claro efecto del clima sobre la probabilidad de las alternativas de recorte. En primer lugar, las probabilidades de "longitud reducida de la estancia" y " alojamiento más barato" son los más altas de las seis alternativas y crecen con el índice de clima. En segundo lugar, "cambiar el período de viaje" y elegir un "transporte más barato" muestran las probabilidades más bajas y disminuye de manera constante con el índice de clima. Por último, viajar "cerca de casa" y "reducir el número de viajes"(menos vacaciones) permanecen casi constantes. No obstante, si se analiza la tasa de cambio de las probabilidades de las alternativas por el clima en origen, los cambios son bastante significativos. Por un lado, los hogares ubicados en las regiones con las mejores condiciones climáticas para el turismo (índice de clima = 12) muestran una probabilidad de recortar un 32% a través de "reducir la estancia" que los hogares ubicados en las regiones con peor clima (índice de clima = 0). En el caso de "un alojamiento más barato", el cambio no es tan agudo. Es un 1,05% más probable recortar con esta opción para los hogares con el mejor índice de clima en lugar de los que tienen el índice de clima peor. Por otro lado, es un 2% menos probable reducir "el período de viaje" para aquellos turistas con el índice de clima más 257 alto que los turistas con el más bajo. Estos resultados indican que las diferencias en el lugar de origen juegan un papel importante en las probabilidades. Figura 2.2: Medias moviles de la probabilidad por clima 258 Figura 2.3: Medias móviles de la probabilidad por edad La figura 2.3 muestra el cambio de la media móvil de la probabilidad de cada estrategia de recorte en función de la edad. Por ejemplo, para personas de 65 años, la probabilidad de optar por "reducir la estancia" es 34,78% superior que para los de 20. En lo que respecta a la alternativa "alojamiento barato", la probabilidad disminuye un 33,33% para personas de 65 años, en comparación con las de 20. Las principales conclusiones de este capítulo pueden resumirse como sigue. En primer lugar, la edad importa. En particular, los turistas de mayor edad tienen más probabilidades de reducir la duración de su estancia bajo un cambio del PIB. En segundo lugar, el clima de origen también influye en el tipo de recorte a realizar. Los turistas que viven en regiones con buen clima son más propensos a reducir el tiempo de estancia (32%) que los turistas que viven en regiones con peor clima. En tercer lugar, para la mayoría de las regiones, las estrategias de recorte preferidas son claras: la forma favorita es a través de la reducción de la 259 duración de la estancia seguida por la reserva de un alojamiento más barato. El promedio de las probabilidades de tales estrategias son 27% y 20%, respectivamente. Sin embargo, coviene recordar que existe una gran heterogeneidad entre las regiones en lo que al tipo de estrategia de recorte se refiere. Todo este conocimiento se puede emplear para construir paquetes más flexibles que pueden adaptarse a las necesidades de los diferentes perfiles de turistas durante los períodos de crisis económicas. Esta flexibilidad debe basarse en una combinación de variables micro y macroeconómicas, como pueden ser la edad, el género, el clima en las regiones de origen, la severidad de la crisis (reducción del PIB) y las expectativas sobre el crecimiento del PIB. Desde el punto de vista de la gestión del destino, es importante entender que las estrategias de recorte implican efectos totales, efectos parciales o ningún efecto sobre el destino. Por un lado, las estrategias de recorte que tienen efectos completos en el destino son "reducción de la longitud de la estancia" y "alojamiento barato". Por otra parte, "menos vacaciones" y "viajar cerca de casa" pueden o no afectar directamente el destino (efectos parciales). Por último. "el transporte más barato" y "cambiar el período de viaje" son estrategias con efectos más difusos en cuanto a su impacto en el destino (otros efectos). Según los resultados, la probabilidad de recortar a través de estrategias que afectan directamente a los destinos (efectos totales) es del 47,4%, mientras que el resto de las probabilidades, los efectos parciales y otros efectos, representan el 35,4 % y 17,2 %, respectivamente. Capítulo 3: Turismo: crecimiento económico, empleo y la “enfermedad holandesa” Introducción y objetivos Desde 2008. España se encuentra bajo una fuerte recesión económica (reducción de 0,92 % en el PIB real desde 2008 hasta 2012 y la tasa de desempleo del 24,3 % en 2012). Las principales 260 consecuencias de la situación actual de crisis son: la elevada deuda privada (hogares y empresas) alimentado después de años de bajas tasas de interés, alta tasa de desempleo, menores salarios, bajada del consumo privado, contracción del crédito (crisis bancaria), mayor tasa de interés para las emisiones de renta fija pública y decrecimiento económico. Más allá de este breve diagnóstico, la crisis española tiene dos factores adicionales: la pertenencia a la moneda europea (euro) y, debido a ello, un control del déficit público para cumplir con el compromiso de déficit de la UE. El primer factor actúa como si España tuviera un tipo de cambio fijo forzando a una devaluación interior a través de salarios más bajos para ganar competitividad exterior. El segundo factor no permite caer en un déficit público persistente y reduce la posibilidad de llevar a cabo políticas de demanda para impulsar la economía. La devaluación interior ya está reduciendo el déficit y los salarios, lo que está contrayendo aún más la demanda interior, y España, como la mayoría de las economías desarrolladas, se basa en gran medida de la demanda interior para impulsar la economía. En paralelo con la situación económica descrita anteriormente. España ha estado recibiendo una gran llegada de turistas desde 2010 como consecuencia de la primavera árabe que comenzó en diciembre de 2010 en Túnez y se extendió rápidamente a otros países árabes de la región. Este flujo positivo e inesperado de turistas se produjo después de años de disminución. Esta nueva situación, junto con las previsiones económicas más optimistas (World Economic Outlook. Fondo Monetario Internacional, 2013), ha alimentado la idea de que las llegadas de turistas podrían sustituir la débil demanda interior, impulsar la economía y reducir la tasa de desempleo. Bajo este contexto, la existencia de bienes transables y no transables es clave para entender el efecto anteriormente aducido del turismo en la economía. Esta hipótesis y la transición de un auge económico hacia una crisis (o hacia un contexto de menor crecimiento) pueden ser teóricamente respaldadas por una versión adaptada del modelo TNT (Larraín y Sachs, 1994). 261 El éxito de algunos países asiáticos en los años ochenta promoviendo el crecimiento económico a través de las industrias orientadas a la exportación (Banco Mundial, 1993) y la orientación exportadora del turismo ha guiado los estudios sobre el turismo y el crecimiento económico en torno a la hipótesis de las exportaciones (Balassa, 1978) . Autores como Neves y Maças (2008) o Dritsakis (2004) apoyan la capacidad del turismo. un sector no intensivo en tecnología, para promover el crecimiento económico y aumentar la acumulación de capital. Estas conclusiones contradicen los resultados de Solow (1956) y otros autores como Aghion y Howitt (1998) o Grossman y Helpman (1991) sobre la relación entre sectores más intensivos en capital y el crecimiento a largo plazo. Lanza, Temple y Urga (2003) afirman que el menor crecimiento de la productividad en la economía basada en el turismo podría ser superado por una progresiva especialización en el turismo que podría mejorar los términos de intercambio y compensar la pérdida de productividad. Por otra parte, también destacan la importancia de la alta elasticidad precio e ingreso de la demanda de turismo que puede compensar la pérdida de productividad en el largo plazo. Además, existen consecuencias más profundas en la relación entre el turismo y la economía que no deben ser descuidadas. Las diferencias en el uso de capital y trabajo tienen también implicaciones importantes a nivel sectorial. Copeland (1991) y Chao, Hazari, Laffargue, Sgro y Yu (2006) ponen de manifiesto la importancia de los bienes no transables en la economía basada en el turismo. Según ellos, el turismo aumenta el consumo de bienes no transables y mejora los términos de intercambio, aunque podría producir desacumulación de capital de los sectores manufactureros (intensivos en capital) a los no transables (intensivos en mano de obra). Por otra parte, la apreciación del tipo de cambio real debido a las llegadas de turistas también puede socavar la competitividad exterior de las exportaciones tradicionales. El desplazamiento de capital y mano de obra de los sectores tradicionales hacia los no transables y la apreciación del tipo de cambio real puede generar una "enfermedad" económica conocida 262 como "enfermedad holandesa " por el cual el efecto positivo del turismo en la economía en el corto plazo podría terminar en una contracción económica en el largo (Corden y Neary, 1982). El uso de modelos de equilibrio general recursivo-dinámico permite trabajar en estos dos niveles. Por un lado, permite cuantificar el impacto de la llegada de turistas en el PIB y el desempleo. Por otro lado, también permite analizar el efecto de tal choque sobre la reasignación de recursos (capital y trabajo) entre los sectores, la pérdida de competitividad y. en definitiva, comprobar la existencia de la "enfermedad holandesa" en la economía a nivel más sectorial. Dos escenarios y cinco períodos son proyectados para tal fin. El primero (escenario post-crisis) se basa en la proyección del Fondo Monetario Internacional (Fondo Monetario Internacional, 2013) y en el Boletín Económico del Banco de España (Banco de España, 2013). El segundo es un escenario previo a la crisis basada en el desempeño de la economía española en los cinco años anteriores a la crisis económica. Los dos escenarios tratan de proporcionar una perspectiva más amplia del verdadero potencial del turismo para dar respuesta a los objetivos perseguidos en este trabajo bajo diferentes trayectorias económicas. Marco teórico del impacto del turismo en una economía: modelo TNT -T En esta sección se desarrolla una nueva versión del modelo TNT (Sachs y Larrain, 1994) para incluir la influencia del turismo en la producción y consumo de bienes transables y no transables, así como la existencia de un nuevo equilibrio debido a la infrautilización del trabajo y el capital (llamado modelo TNT-T). Hay bienes que son transables (bienes que se pueden exportar) y bienes no transables (bienes que no se pueden exportar como alojamiento. servicios de restauración o un corte de pelo). Esta simple diferencia en el tipo de productos tiene importantes consecuencias sobre la 263 economía. Por ejemplo, en una situación de auge económico, las economías tienden a consumir más bienes transables que los que pueden llegar a producir, por lo que las importaciones crecen. En términos gráficos (ver Figura 3.1), la economía consume bienes transables en A (Ct0) y los produce en B (Qt0); la diferencia entre A y B son las importaciones. El consumo de bienes no transables es Cnt0 y. dadas las características de este tipo de bienes coincide con su producción (Pnt0). DA hace referencia a la demanda agregada. FPP significa frontera de posibilidades de producción. NT son bienes no transables y T bienes transables. Cuando la economía se desplaza de una situación de crisis (o a una de decrecimiento económico) la economía se mueve hacia abajo a lo largo de DA0 hasta D. Durante este proceso, tanto el consumo de bienes transables como de no transables decae. El menor consumo de bienes transables puede ser compensado por un aumento de las exportaciones (diferencia entre D y C). Pero la caída en la producción de bienes no transables sólo puede ser soportada por la demanda interior. El turismo es una exportación intensiva en el consumo de bienes no transables tales como alojamiento o servicios de restauración. La inclusión del consumo turístico en el modelo desplaza la demanda agregada de DA0 a DA1. En esta nueva situación, en un principio. el consumo de bienes no transables se incrementa desde D a E. Tan pronto el turismo comienza a demandar bienes no transables y la crisis económica queda atrás, la producción y la demanda agregada aumentan. Por lo tanto, el consumo total pasa de E a G. La diferencia entre G (demanda agregada con el efecto turístico) y G1 (demanda agregada sin turismo) representa las exportaciones de bienes no transables. Esto significa que la equivalencia entre la producción y el consumo de los bienes no transables ya no se cumple debido a la inclusión del turismo en el modelo. Por otro lado, el movimiento de E a G sólo se puede lograr mediante la reducción de la producción (y exportaciones) de mercancías transables (de C a H). a pesar del aumento de la demanda de D a G. 270 Según la figura 3.5 y centrados en la situación posterior a la crisis, el aumento del 2% en la llegada de turistas aumenta la VABR de 0,05 % en el primer año a 0,44 % en el quinto año. El impacto probable del 10 % en la llegada de turistas significa un fuerte fomento a la economía de un 0,25 % en el primer año a un 2,58 % en el quinto año. Durante los cinco años, el crecimiento acumulativo de la VABR es un 1,21 % y un 6,81 % para el impacto del 2 % y 10 %, respectivamente. En el caso del 2 % el déficit se reduce hasta el 0,06% en el primer año y hasta el 0,45 % en el último año. En el caso del 10 %, la reducción del déficit pasa de un 0,28% a un 2,75 % en el último año. Al final del quinto año, el déficit se acumula una reducción de alrededor de 1,27 % y 7,35 % para el choque de 2 % y 10 %, respectivamente. Con respecto a la situación anterior a la crisis, el aumento de las llegadas de turistas en un 2% aumenta el VABR desde un 0,06 % en el primer año hasta un 0,55 % en el último año. Al igual que en el escenario posterior a la crisis, el impacto del 10% significa un fuerte fomento a la economía, que va desde 0,28 % en el primer año a un 3,28 % en el quinto año. Durante los cinco años, el crecimiento acumulativo de la VABR es del 1,53 % y del 8,63 % para los choques del 2 % y 10 %, respectivamente. Por otra parte, la mejora en el déficit por cuenta corriente es similar al cambio en VABR. En el caso del choque de 2 %, el déficit se reduce un 0,04% el primer año y un 0,48 % el último. Estos cambios en la cuenta corriente son un poco más bajos que en el escenario de post-crisis. En el caso del 10 %, la reducción del déficit pasa de 0,23 % a un 2,91 % en el último año. Al final del quinto año, el déficit acumula una reducción del 1,32 % y del 7,66 % para el choque de 2 % y 10 %, respectivamente. El cambio en VABR debido al aumento de la llegada de turistas es significativo pero modesto para el choque del 2 % en ambos escenarios. Sin embargo, el choque poco probable de 10% produce un fuerte impulso a la economía. El aumento en las llegadas de turistas aumenta el tipo de cambio real que hace disminuir las exportaciones tradicionales y eleva las importaciones. El efecto general es una reducción del déficit. Este resultado está de acuerdo 271 con Adams y Parmenter (1995), Copeland (1991) or Narayan (2004) . Esta mejora general en los términos de intercambio es especialmente útil en la situación económica actual en España en el que el país tiene un alto endeudamiento exterior generado durante el boom económico. En resumen, el aumento de VABR prueba que el turismo es capaz de promover el crecimiento económico a pesar de la menor utilización tecnológica. Por otra parte, el turismo también aumenta la acumulación de capital (gráfico 3.5), pero en tasas moderadas que refuerza la conclusión de Capó et al. (2007) acerca de las bajas ganancias de productividad generados por el turismo en el largo plazo. 272 Figura 3.5: Cambio en el valor añadido bruto real, deficit exterior y acumulación de capital (%) Post-crisis Pre-crisis Valor añadido bruto real Valor añadido bruto real Déficit por cuenta corriente Déficit por cuenta corriente Acumulación de Capital Acumulación de Capital 273 Ganadores y perdedores: “la enfermedad holandesa" Desde una perspectiva macroeconómica, el efecto positivo de las llegadas de turistas en la economía ya se ha puesto de manifiesto en los apartados anteriores. Sin embargo, a nivel sectorial, el impacto del turismo tiene efectos diversos en función de los bienes producidos o los servicios prestados, lo que en último término produce una reasignación de recursos entre los sectores económicos. Más concretamente, el turismo tiene dos impactos notables sobre la economía: aumenta la demanda de bienes no transables y. al mismo tiempo, aumenta la tasa de cambio que socava la competitividad exterior de las exportaciones tradicionales. Estos resultados pueden ser vistos como una consecuencia general de la "enfermedad holandesa". El término "enfermedad holandesa" fue utilizado por primera vez por la revista "The Economist" en 1977 para explicar el efecto del petróleo descubierto en los años sesenta en el Mar del Norte en la economía holandesa. Corden y Neary (1982) y Corden (1984) son los primeros que modelan dicho enfermedad en términos académicos. Estos trabajos diferencian tres tipos de sectores: sector en auge, sector rezagado y sector no transable. El sector en auge comienza a producir y exportar con fuerza con lo cual comienza a demandar trabajadores de otros sectores (sectores rezagados y los sectores no transables) para mantener la producción (efecto de recursos). Al mismo tiempo, la renta generada por el sector en auge aumenta el tipo de cambio real que erosiona la competitividad exterior de las exportaciones tradicionales, y, junto con el aumento en la demanda de bienes no transables producidos por los ingresos generados, produce el efecto gasto. Este aumento de los bienes no transables también aumenta la demanda de trabajadores y capital a costa del sector rezagado (otro efecto recurso). Como resultado, se produce una desindustrialización del sector rezagado, un fuerte aumento de los precios internos y en el intercambio real que, en último término, hace tambalear la competitividad y se contrae la economía. 274 La enfermedad holandesa ha sido tradicionalmente asociada a las exportaciones de petróleo de países como Arabia Saudita. Qatar. Venezuela o Noruega. Gran parte de ellos evitan cambiar la mayor parte de los ingresos obtenidos por sus exportaciones de petróleo a la moneda local para prevenirse de la enfermedad holandesa. Sin embargo, esta enfermedad económica puede generalizarse a cualquier situación en la que un país comienza a recibir una cantidad importante de dinero extranjero que desencadenan las consecuencias explicadas en el párrafo anterior. Por ejemplo. Laplagne et al. (2001), Usui (1996), VanWijnbergen (1986) y White (1992) muestran cómo los países en desarrollo pueden sufrir de esta enfermedad gracias a la ayuda externa recibida. Wijnbergen (1984) argumenta que la connotación negativa del término "enfermedad holandesa” no debe ocultar una teoría del comercio importante detrás de dicho efecto por el cual una economía tiende a producir aquellos bienes que requieren un uso intensivo del factor más abundante en el país (teoría del comercio de Heckscher-Ohlin). Según Roca (1998), el diagnóstico de la enfermedad holandesa se puede comprobar atendiendo a cuatro hipótesis: la apreciación del tipo de cambio real, el descenso de las exportaciones del sector rezagado, la disminución de la producción del sector rezagado y un probable aumento de la producción de los sectores no transables. Las dos primeras hipótesis forman el efecto gasto y las dos últimas, el efecto recurso. Además, una quinta hipótesis debe ser añadido en lo que respecta a los objetivos de este artículo: el probable aumento del empleo en los sectores en auge y no transables y los probables respectivos descensos en los sectores rezagados (situación en la que unos ganan y otros pierden). 275 Efecto gasto  Apreciación del tipo de cambio real Como se muestra en la figura 3.6, el incremento del 2% en la llegada de turistas aprecia el tipo de cambio real casi 0,4% y un 0,45% en el quinto año en ambos escenarios, respectivamente. Figura 3.6: cambio en el tipo de cambio real (%)  Disminución de las exportaciones de los sectores rezagados Los sectores rezagados están asociados con la agricultura, la energía y la minería y la industria. Como se muestra en la tabla 3.1, las exportaciones en los sectores rezagados disminuyen en ambos escenarios, a pesar de la alta tasa de desempleo. La apreciación del tipo de cambio real y la menor intensidad en el uso de mano de obra de estas actividades, supera el efecto positivo de los salarios más bajos debido a la alta tasa de desempleo. La disminución de las exportaciones tradicionales es especialmente considerable en el lado turístico de estos sectores y, más precisamente, en los productos básicos agrícolas con una caída de 7,52% y 7,93% en el último año en los escenarios de post-crisis y pre-crisis, respectivamente. Las 276 exportaciones de los sectores rezagados representan el 70% aproximadamente de las exportaciones totales en España. Por lo tanto, cualquier impacto negativo en ellos afectará gravemente a la cuenta corriente. Tabla 3.1: cambio en las exportaciones en los sectores rezagados (%) Post-crisis bien\año 1 2 3 4 5 Agricultura t -1,68 -3,08 -4,58 -6,06 -7,52 nt -0,50 -0,80 -1,19 -1,58 -1,95 Energía and minería t - - - - - nt -0,45 -0,70 -1,04 -1,37 -1,70 Industria t -0,65 -1,11 -1,65 -2,19 -2,71 nt -0,24 -0,31 -0,44 -0,57 -0,70 Pre-crisis 1 2 3 4 5 Agricultura t -1,87 -3,27 -4,86 -6,41 -7,93 nt -0,65 -0,95 -1,41 -1,86 -2,29 Energía and minería t - - - - - nt -0,56 -0,77 -1,14 -1,49 -1,83 Industria t -0,70 -1,10 -1,62 -2,13 -2,63 nt -0,29 -0,29 -0,42 -0,53 -0,62 Efecto recurso  Disminución de la producción en el sector rezagado La producción en el sector rezagado (la agricultura. la energía y la minería y la industria) caen ligeramente durante los cinco años en ambos escenarios como se muestra en la tabla 3.2. El mayor descenso se produce en "energía y minería" en ambos escenarios. Por lo tanto, no se logra alcanzar el equilibrio en el que todos ganan (sectores transables y no transables) como se mostraba en el modelo TNT-T. 277 Tabla 3.2: Cambio en la producción del sector rezagado (%) Post-crisis Sector\año 1 2 3 4 5 Agricultura -0,24 -0,35 -0,53 -0,69 -0.85 Energía and minería -0,31 -0,48 -0,72 -0,95 -1,18 Industria -0,16 -0,20 -0,29 -0,38 -0,46 Pre-crisis 1 2 3 4 5 Agricultura -0,31 -0,40 -0,58 -0,76 -0,93 Energía and minería -0,37 -0,52 -0,77 -1,01 -1,25 Industria -0,19 -0,18 -0,26 -0,33 -0,39  Un aumento de la producción en los sectores no transables Como puede apreciarse en la tabla 3.3, las salidas por actividades aumentan en el sector en auge (B) y en los sectores no transables (N) en ambos escenarios. El sector del alojamiento, el sector del transporte aéreo y el sector de agencias de viajes son los más beneficiados de las llegadas de turistas, con un aumento de 5,29%, 3,01% y 14,67% en el escenario posterior a la crisis y del 5,26%, 3,14% y 13,04% en el previo a la crisis para el quinto año, respectivamente. Este resultado está en línea con el modelo TNT-T por el cual la demanda turística aumenta la demanda agregada hacia los bienes no transables. 278 Tabla 3.3: Cambio en la producción en los sectores en auge ay no-transables (%) Post-crisis Sector\año 1 2 3 4 5 Construcción(N) 0,07 0,19 0,30 0,42 0,55 Comercio(B) 0,01 0,13 0,20 0,28 0,36 Alojamiento(ByN) 0,99 2,06 3,12 4,20 5,29 Restauración(ByN) 0,22 0,56 0,86 1,17 1,49 Transporte por ferrocarril ( B ) 0,49 1,08 1,63 2,20 2,78 Transporte por carretera ( B ) 0,04 0,19 0,29 0,41 0,53 Transporte marítimo(B) 0,07 0,27 0,41 0,56 0,72 Transporte aéreo(B) 0,54 1,18 1,78 2,39 3,01 Otros servicios de -0,01 0,09 0,14 0,20 0,27 Agencias de viaje(ByN) 2,67 5,35 8,08 10,85 13,65 Inmobiliaria(ByN) 0,03 0,17 0,26 0,36 0,47 Alquiler de coche(ByN) 0,13 0,36 0,55 0,75 0,96 Entretenimiento(ByN) 0,08 0,25 0,38 0,52 0,67 Pre-crisis 1 2 3 4 5 Construcción(N) 0,07 0,23 0,37 0,51 0,67 Comercio(B) -0,01 0,16 0,26 0,36 0,48 Alojamiento(ByN) 0,96 2,04 3,09 4,17 5,26 Restauración(ByN) 0,20 0,59 0,91 1,24 1,59 Transporte por ferrocarril(B) 0,47 1,09 1,65 2,23 2,83 Transporte por carretera(B) 0,01 0,20 0,32 0,44 0,59 Transporte marítimo(B) 0,02 0,24 0,38 0,52 0,68 Transporte aéreo(B) 0,53 1,22 1,85 2,49 3,14 Otros servicios de -0,05 0,09 0,16 0,23 0,32 Agencias de viaje(ByN) 2,56 5,12 7,72 10,36 13,04 Inmobiliaria(ByN) 0,01 0,21 0,32 0,45 0,59 Alquiler de coche(ByN) 0,11 0,37 0,57 0,78 1,00 Entretenimiento(ByN) 0,07 0,29 0,45 0,62 0,81 279  Un probable incremento / disminución del empleo en el sector en auge y no-transable / sector rezagado. Los sectores rezagados (L) reducen la tasa de empleo a pesar de la alta tasa de desempleo y el aumento de las llegadas de turistas, con la excepción de la industria que aumenta el empleo del segundo al quinto año en ambos escenarios. Por otro parte, los sectores en auge y no transables aumentan su demanda de mano de obra. Los sectores más beneficiados de la alta tasa de desempleo son el alojamiento, los servicios de restauración, el transporte ferroviario, el transporte aéreo y las agencias de viajes que aumentan su demanda de trabajadores un 6,12%, 2,33%, 3,45%, 3,8% y el 14,67% en la situación posterior a la crisis y un 6,28%, 2,64%, 3,65%, 4,12% y 14.31% en la situación anterior a la crisis, respectivamente. Tal vez, y dando por sentada la apreciación del tipo de cambio real, un aumento mayor en el proceso de acumulación de capital podría permitir a los sectores rezagados absorber dicho capital y. en consecuencia, también aumentar su demanda de empleo. Desafortunadamente, el sector turístico no parece proporcionar tal aumento en la acumulación de capital. Para concluir, en este mundo en constante cambio, la relación entre el turismo y el crecimiento económico es tan difusa y compleja que cualquier conclusión en este respecto debe tomarse con cautela. Por ejemplo, la llegada de Internet y las nuevas tecnologías ha convertido algunos bienes no transables en transables; tales como servicios de contabilidad. Países como la India ya están tomando ventaja de esta nueva situación. Estos cambios constantes e imprevistos harán reconsiderar muchas teorías y conclusiones asentadas previamente en la disciplina. Teniendo esto en cuenta y con base en los supuestos y proyecciones establecidos en este documento, se puede afirma que el turismo proporciona crecimiento económico, reduce la tasa de desempleo, mejora de los términos de intercambio y aumenta la demanda interna en el medio plazo en España. El efecto sobre las variables 286 Tabla 4.3. Variables determinants en la elección del destino Modelo Parámetro Dev. típica Parámetros aleatorios: Precipitaciones -0.02484*** 0.00110 Temperatura 2.43831*** 0.09074 Temperatura^2 -0.09977*** 0.00308 Distancia -7.12421*** 0.20619 Parámetros no aleatorios: Carreteras pavimentadas 0.01732*** 0.00095 Educación secundaria 0.01832*** 0.00229 Esperanza de vida 0.11228*** 0.00847 Ratio ppp -0.06550*** 0.00202 Heterogeneidad en media: Temperatura: temperatura en origen 0.14276*** 0.00403 Temperatura^2: (temperatura en origen)^2 -0.45548D-4*** 0.2911D-05 Distancia: Log(ingreso) 0.76293*** 0.02648 287 Tabla 4.3(continuación). Variables determinants en la elección del destino Modelo Parámetros Dev. Típica Valores diagonal principal matriz de cholesky: Dev. Típica precipitaciones 0.01632*** 0.00073 Dev. Típica temperature 0.62423*** 0.06736 Dev. Típica temperatura^2 0.01162*** 0.00093 Dev. Típica distancia 0.92091*** 0.02336 Valores fuera de la diagonal matriz de cholesky : Cov(temperatura, precipitaciones) -0.01920*** 0.00122 Cov(temperatura^2,precipitaciones) -0.74488D-4*** 0.1896D-04 Cov(temperatura^2, temperatura) 0.00459*** 0.00152 Cov(distancia, precipitaciones) -0.00231*** 0.00030 Cov(distancia, temperatura) 0.20065*** 0.2440 Cov(distancia, temperatura^2) 0.98583*** 0.00059 Dev. Típica de los parámetros aleatorios Precipitaciones 0.01632*** 0.00073 Temperatura 1.33174*** 0.05757 Temperatura^2 0.02025*** 0.00147 Distancia 0.98583*** 0.2200 Mc fadden pseudo R-cuadrado: 0.2966 AIC 32,312 Log function de verosimilitud -16,135.267 La tabla 4.3 muestra el resultado del modelo. Todas las variables explicativas tienen el signo esperado y son altamente significativas. Por otra parte, las desviaciones estándar de los parámetros aleatorios son también altamente significativas lo que refuerza la suposición de parámetro aleatorio para estas variables. Las precipitaciones en destino afectan negativamente a la utilidad individual (-0.02484). Por otro lado, la temperatura en destino afecta positivamente la utilidad individual (2.43831) y este efecto se incrementa por la temperatura en origen (0.14276). En otras palabras, los turistas de regiones cálidas prefieren temperatura cálida en destino. No obstante, la temperatura al cuadrado supone un punto de cambio en el efecto creciente de la utilidad antes aumentos de la temperature en el destino. Bajo estas circunstancias, los turistas tenderán a viajar domésticamente cuando la temperatura en el 288 destino esté por encima de la temperatura optima, a partir de dicho punto la utilidad en el destino decrecerá. Si las regiones frías actuales se calientan debido al cambio climático, los habitantes de estas regiones serán más sensibles a la temperatura como lo son los de las regiones cálidas actuales. Esta relación coincide con otros hallazgos de autores tales como Bigano et al. (2006a), aunque ellos reducen el efecto asimétrico a la temperatura al cuadrado. Como ya se explicó en la sección anterior, la desviación estándar ya no es independiente debido a las correlaciones asumidas entre los parámetros aleatorios. Por lo tanto, cuanto mayor sea la covarianza, mayor será la relación entre los parámetros. En el caso de las precipitaciones, la parte de la desviación estándar total explicada por sí misma es 0,01632, el resto de la desviación estándar total se explica por la interacción con la temperatura (- 0,01920), la temperatura al cuadrado (-0,074488D-4) y la distancia (-0,00231). En el caso de la precipitación y la temperatura, esta desviación estándar significa que los individuos con sensibilidades más grandes a las precipitaciones son menos sensibles a la temperatura. En el caso de la temperatura y la distancia, los turistas con sensibilidades más grandes a la temperatura no están tan preocupados por la distancia en su función de utilidad, hasta un punto de inflexión (temperatura óptima) en el que a mayor distancia la utilidad decrece a pesar de temperaturas más altas. Precipitaciones y distancia muestran una relación negativa. Los individuos que le dan mucha importancia a las precipitaciones no suelen estar tan preocupados por la distancia. La relación negativa entre la distancia y temperatura refuerza la hipótesis de Von Thünen (1826) por la cual la gente está dispuesta a asumir mayores distancias con el fin de disfrutar de mayores rentas de situación (mejores condiciones climáticas en destino). La relación entre la distancia, temperatura y las precipitaciones también está en conformidad con Nicolau y Más (2006). 289 Como puede verse también en ambos modelos, variables como las carreteras pavimentadas, la educación secundaria, la esperanza de vida y ppp_rate reflejan que los turistas prefieren los países más desarrollados, pero, al mismo tiempo, más barato que sus regiones de residencia. Estos resultados son importantes para los políticos porque, a diferencia de las variables climáticas, el destino puede actuar sobre estas variables con el fin de atraer turismo (EugenioMartín. Martín Morales y Sinclair, 2008). La tabla 4.4, muestra los cambios que se producirían en la elección del destino ante cambios en la temperatura porjectados por el panel intergubernamental sobre cambio climático (2007). De acuerdo al escenario A, ante un aumento de 1.8ºC en los dos principales países emisores de turistas (Reino Unido y Gran Bretaña), todos los principales destinos turísticos actuales verían decrecer su entrada de turistas. Por ejemplo, Francia y España verían decrecer su flujo de turistas en un 5.76% y un 6.70% respectivamente. Analizando el escenario B, esto es, suponiendo una aumento general de 1.8ºC, todos los principales destinos turísticos verían aumentar su flujo de turistas, si bien el Reino Unido y Alemania serían los más beneficiados. 290 Tabla 4.4. Cambios en la elección de los principales destinos ante cambios en la temperatura Escenario A: +1.8 ºC en Reino Unido y Alemania Escenario B: +1.8 ºC en todos los destinos Flujos iniciales (%) Cambio en los flujos (%) Cambio en los flujos (%) Croacia 1.88 -1.77 0.76 Chipre 2.67 -1.01 0.93 Francia 7.56 -5.76 4.96 Grecia 10.09 -4.94 0.40 Italia 8.04 -7.78 2.96 Malta 4.39 -3.17 1.15 Oriente medio 2.59 -1.47 2.52 Portugal 4.31 -2.82 1.84 España 9.46 -6.70 1.05 Reino Unido 6.89 24.99 7.07 Alemania 6.52 36.81 6.55 El modelado de la elección de destino es un reto. Todas las metodologías tienen sus pros y sus contras, pero, por lo que sabemos, los modelos logit mixtos son probablemente los que mejor se ajusten al modelado de la elección de destino ya que combina atributos tanto de origen como de destino. Las percepciones asimétricas del clima por parte de los turistas han sido probadas y confirmadas. Los individuos de regiones frías están más dispuestos a viajar a regiones más cálidas. El cambio climático puede suponer un cambio en este patrón turístico ya que las regiones actualmente más frías verán aumentar su flujo turístico al aumentar su temperatura media por el impacto del calentamiento global. Atributos relacionados con el nivel de desarrollo (carreteras pavimentadas, año de la educación o la esperanza de vida) son percibidos positivamente por el turista. Estas variables son más manejables por los políticos a pesar de los "ingresos" vienen en el largo plazo. Este 291 resultado demuestra que los destinos de sol y playa son más que las dotaciones ambientales. Por lo tanto, el crecimiento impulsado por el turismo puede requerir un cierto nivel de desarrollo económico para aprovecharse realmente de ella. Como se mencionó en el documento, la forma en que el clima está incluido en los modelos de demanda turísticos carece a menudo de una representación adecuada. Ni los datos agregados, ni los índices climáticos actuales captan adecuadamente el efecto climático. Tan pronto como la sociedad tome más conciencia por el impacto del clima y el medio ambiente, más micro datos estarán disponibles. A pesar de ello, se espera que los datos agregados sigan manteniendo una posición destacada en la investigación turística. Bajo estas circunstancias, el desarrollo de un índice de clima turístico basado en ponderaciones objetivas y de preferencias declaradas ayudará a tener una visión más precisa de la respuesta de los turistas al clima. En consecuencia, el uso de este índice climático como variable en los datos macro también ayudará a capturar mejor el impacto del cambio climático. 292 293 Annex. Codes Stata code (chapter 1) *endogeneity: *haussman reg exporigin incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total $why2 if nights<31, noconstant predict prexorigin predict resorigin, residual reg expisland prexorigin resorigin incomemonth $trime $pais $pa $pal $destino $people $party $why2 veces_total if nights<31, noconstant reg expisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total $why2 if nights<31, noconstant predict prexpisland predict resisland, residual reg exporigin prexpisland resisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total if nights<31, noconstant *residuals correlations: corr resisland resorigin corr resorigin resisland 294 reg exporigin incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total if nights<31, noconstant predict resorigin1, residual reg expisland incomemonth $trime $pais $pa $pal $destino $people $party veces_total $why2 if nights<31, noconstant predict resisland1, residual corr resorigin1 resisland1 sureg (exporigin expisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total, noconstant) (expisland exporigin incomemonth $trime $pais $pa $pal $destino $people $party $why2 veces_total, noconstant) if nights<=31, corr *heterokedasticity: * separated models: reg exporigin expisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total if nights<31 estat hett reg expisland exporigin incomemonth $trime $pais $pa $pal $destino $people $party $why2 veces_total if nights<31 estat hett * reg3 : "the model" reg3 (exporigin expisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total, noconstant) (expisland exporigin incomemonth $trime $pais $pa $pal $destino $people $party $why2 veces_total, noconstant) if nights<=31 * Residuals: heterokedasticity test 295 lmhreg3 *the model * reg3 : "El modelo con todas las variables" set more off reg3 (exporigin expisland incomemonth $trime $ano $pais $pa $pal $destino $people $party veces_total, noconstant) (expisland exporigin incomemonth $trime $pais $pa $pal $destino $people $party $why2 veces_total, noconstant) if nights<=31 * Estimation controling by heteroskedasticity (GMM) gmm(eq1: exporigin -{xb: expisland incomemonth $pa $pal $trime $pais $ano $destino $people $party veces_total}) (eq2:expisland -{xc:exporigin incomemonth $trime $pais $pa $pal $people $party $why2 $destino veces_total}) if nights<=31, instruments(eq1: incomemonth $pa $pal $trime $pais $ano $destino $people $party veces_total) instruments(eq2:incomemonth $trime $pais $pa $pal $people $party $why2 $destino veces_total) winitial(unadjusted,independent) wmatrix(robust) twostep Stata code (chapter 2) gsem /// ($employmenth education gender age age2 gdppcpps2009 growth climate L@1 -> cutback,probit) /// (2.howcutback1 <- age climate coast airport L@X , mlogit) /// Reduced length (3.howcutback1 <- age climate coast airport L@X , mlogit) /// Cheaper transport (4.howcutback1 <- age climate coast airport L@X , mlogit) /// Cheaper accommodation (5.howcutback1 <- age climate coast airport L@X , mlogit) /// Closer to home (6.howcutback1 <- age climate coast airport L@X , mlogit), /// Period of travel startvalues(iv) startgrid(.1 1 10) var(L@1) noconstant 302 Alegre, J. and Pou, Ll. (2006). The length of stay in the demand for tourism. Tourism Management, 27(6), 1343-1355. Alegre, J. and Pou, Ll. (2008). Research note: Tourism expenditure and all-inclusive packagesthe case of a mature Mediterranean destination. Tourism Economics, 14(3), 645655. Alegre, J. and Pou. L. (2004). Micro-economic determinants of the probability of tourism consumption. Tourism Economics, 10(2): 125–144. Alegre, J., Cladera, M. and Sard, M. (2011). Analysing the influence of tourist motivations on tourist expenditure at a sun-and-sand destination. Tourism Economics, 17(4), 813-832. Alegre. J., Mateo, S. and Pou. Ll. (2013). Tourism participation and expenditure by Spanish households: the effect of the economic crisis and unemployment. Tourism Management, 39: 37–49. Amelung, B. and Viner, D. (2006). Mediterranean tourism: exploring the future with the tourism climatic index. Journal of Sustainable Tourism, 14(4), 349-366. Amelung, B., Nicholls, S. and Viner, D. (2007). Implications of global climate change for tourism flows and seasonality. Journal of Travel research, 45(3), 285-296. Armington, P.S. (1969). A theory of demand for products distinguished by place of production. IMF Staff Papers, 16(1), 170-201. Balaguer, J. and Cantavella-Jordá, M. (2002). Tourism as a long-run economic growth factor: the Spanish case. Applied Economics, 34(7), 877-884. Balassa, B. (1978). Exports and economic growth: further evidence. Journal of Development Economics, 5(2), 181-189. 303 Balistreri, E.J. and Hillberry, R.H. (2003). The gravity model: an illustration of structural estimation as calibration. Economic Inquiry, 46 (2008), pp. 511–527. Bank of Spain. (2013). Economic bulletin: Spanish economic projections report. Madrid, Spain: Bank of Spain. Baxter, M.J. (1979). The interpretation of the distance and attractiveness components in models of recreational trips. Geographical Analysis, 11(3), 311–315. Ben Abda, M., Belobaba, P.P. and Swelbar, S.W. (2012). Impacts of LCC growth on domestic traffic and fares at largest US airports. Journal of Air Transport Management, 18, 21-25. Berrittella, M., Bigano, A., Roson, R. and Tol, R.S.J. (2006). A general equilibrium analysis of climate change impacts on tourism. Tourism Management, 27, 913-924. Bigano, A., Bosello, F., Roson, R., and Tol, R.S.J. (2008). Economy-wide impacts of climate change: a joint analysis for sea level rise and tourism. Mitigation and Adaptation Strategies for Global Change, 13(8), 765-791. Bigano, A., Hamilton, J.M. and Tol, R.S.J. (2006a). The impact of climate on holiday destination choice. Climatic Change, 76(3-4), 389-406. Bigano, A., Hamilton, J.M., Maddison, D.J. and Tol, R.S.J. (2006b). Predicting tourism flows under climate change. Climatic Change, 79(3), 175-180. Blake, A. (2000). The economic effects of tourism in Spain. Christel DeHaan Tourism and Travel Research Institute Discussion Paper 2000/2. Retrieved February 1, 2011, from http://www.nottingham.ac.uk/ttri/series.html 304 Blake, A. and Sinclair, M.T. (2003). Tourism crisis management: US response to September 11. Annals of Tourism Research, 30(4), 813-832. Blake, A., Arbache, J.S., Sinclair, M.T. and Teles, V. (2008). Tourism and poverty relief. Annals of Tourism Research, 35(1), 107-126. Blake, A., Durbarry, R., Eugenio-Martin, J.L., Gooroochurn, N., Hay, B., Lennon, J., Sinclair, M.T. and Yeoman, I. (2006). Integrating forecasting and CGE models: The case of tourism in Scotland. Tourism Management, 27(2), 292-305. Blake, A., Durbarry, R., M. Sinclair, T. and Sugiyarto G. (2001). Modelling tourism and travel using tourism satellite accounts and tourism policy and forecasting models. Christel DeHaan Tourism and Travel Research Institute Discussion Paper 2001/4. Retrieved February 1, 2011, from http://www.nottingham.ac.uk/ttri/series.html. Blake, A., Sinclair, M.T. and Sugiyarto, G. (2003). Tourism and EU accession in Malta and Cyprus. Tourism and Travel Research Institute Discussion Paper, 2003/7. Retrieved February 1, 2011, from http://www.nottingham.ac.uk/ttri/pdf/2003_7.pdf Blanchflower, D. and Oswald, A.J. (1995). An introduction to the wage curve. Journal of Economic Perspectives, 9(3), 153-167. Boehringer, C., Rutherford, T.F. and Wiegard, W. (2003). Computable General Equilibrium Analysis: Opening a Black Box, ZEW Discussion Paper No. 03-56, Mannheim, Germany. Brau, R., Lanza, A. and Pigliaru, F. (2007). How fast are small tourism countries growing? Evidence from the data for 1980-2003. Tourism Economics, 13(4), 603-614. 305 Brent-Ritchie. J.R., Amaya-Molinar. C.M. and Frechtling. D.C. (2010). Impacts of the world recessions and economic crisis on tourism: North America. Journal of Travel Research, 49: 15–45. Bronner, F., and Hoog. R. (2012). Economizing strategies during an economic crisis. Annals of Tourism Research. 39(2). 1048–1069. Brooke, A., Kendrick, D. and Meeraus, A. (1988). GAMS: a user’s guide. San Francisco: The Scientific Press. Bujosa, A. and Rosselló, J. (2013). Climate change and summer mass tourism: the case of Spanish domestic tourism. Climatic change, 117(1-2), 363-375. Burfisher, M.E. (2011). Introduction to computable general equilibrium models. Cambridge: Cambridge University Press. Cameron, A.C. and Trivedi, P.K. (2005). Microeconometrics: methods and applications. Cambridge: Cambridge University Press. Capó, J., Riera, A. and Rosselló, J. (2007). Accommodation determinants of seasonal patterns. Annals of Tourism Research, 34(2), 422-436. Capó, J., Riera, A. and Rosselló, J. (2007). Tourism and long-term growth. A Spanish perspective. Annals of Tourism Research. 34(3), 709-726. Carlsen, J.C. and Hughes. M. (2008). Tourism market recovery in the Maldives after the 2004 indian ocean tsunami. Journal of Travel & Tourism Marketing, 23(2–4), 139-149. Castillo-Manzano, J.L., López-Valpuesta, L. and González-Laxe, F. (2011). The effects of LCC boom on urban tourism fabric: the viewpoint of tourism managers. Tourism Management, 32, 1085-1095. 306 Caves, W.D., Christensen, L.R. and Tretheway, M.W. (1984). Economies of density versus economies of scale: why trunk and local service airline costs differ. The RAND Journal of Economics, 15, 471-489. Chao, C.C., Hazari, B.R., Laffargue, J.P., Sgro, P.M. and Yu, E.S. (2006). Tourism, Dutch Disease and welfare in a n open dynamic economy. Japanese economic review, 57(4), 501515. Chen, C.M. and Chang, K.L. (2012). The influence of travel agents on travel expenditures. Annals of Tourism Research, 39(2), 1258-1263. Civil Aviation Authority, (2006). CAP 770, no-frills carriers: Revolution or Evolution? A study by the Civil Aviation Authority. London, UK. Retrieved 3 October, 2011, from http://www.caa.co.uk/application.aspx?catid=33&pagetype=65&appid=11&mode=detail&id= 2528 Coenen, M. and Van Eekeren, L. (2003). A study of the demand for domestic tourism by Swedish households using a two-staged budgeting model. Scandinavian Journal of Hospitality and Tourism, 3(2), 114-133. Copeland, B.R. (1991). Tourism, welfare and de-industrialization in a small open economy. Economica, 58(232), 515-529. Corden, W. M., and Neary, J. P. (1982). Booming sector and de-industrialisation in a small open economy. The Economic Journal, 92(368), 825-848. Corden, W.M. (1984). Booming sector and Dutch disease economics: survey and consolidation. Oxford Economic Papers, 36(3), 359-380. 307 Davison, L. and Ryley, T. (2010). Tourism destinations preferences of low-cost airline users in the East Midlands. Journal of Transport Geography, 18, 458-465. De Freitas, C.R. (2003). Tourism climatology: evaluating environmental information for decisión making and business planning in the recreation and tourism sector. International Journal of Biometeorology, 48, 45-54. De Mello, M., Pack, A. and Sinclair, T. (2002). A system of equations model of UK tourism demand in neighbouring countries. Applied Economics, 34(4), 509-521. Devarajan, S., Ghanem, H. and Thierfelder, K. (1997). Economic reform and labor unions: a general-equilibrium analysis applied to Bangladesh and Indonesia. The World Bank Economic Review, 11(1), 145-170. Di Liberto, A. (2013). High skills, high growth: Is tourism an exception? The Journal of International Trade and Economic Development, 22(5), 749-785. Díaz-Pérez, F.M., Bethencourt-Cejas, M. and Álvarez-González, J.A. (2005). Research note: The segmentation of canary island tourism markets by expenditure: implications for tourism policy. Tourism Management, 26(6), 961-964. Divisekera, S. (2003). A model of demand for international tourism. Annals of Tourism Research 30(1), 31-49. Divisekera, S. (2010). Economics of tourist´s consumption behaviour: some evidence from Australia. Tourism Management, 31(5), 629-639. Doganis, R. (2006). The airline business. London: Routledge. 308 Dolnicar, S, Crouch, G.I, Devinney, T., Huybers, T., Louviere, J.J. and Oppewal, H. (2008). Tourism and discretionary income allocation. Heterogeneity among households. Tourism Management, 29, 44-52. Domar, E. D. (1946). Capital expansion, rate of growth, and employment. Econometrica, 14(2), 137-147. Donzelli, M. (2010). The effect of low-cost air transportation on the local economy: Evidence from Southern Italy. Journal of Air Transport Management, 16(3), 121-126. Dritsakis, N. (2004). Tourism as a long-run economic growth factor: an empirical investigation for Greece using causality analysis. Tourism Economics, 10(3), 305-316. Durbarry, R. (2004). Tourism and economic growth: the case of Mauritius.Tourism Economics, 10(4), 389-401. Dwyer, L., Forsyth, P. and Dwyer, W. (2010). Tourism Economic Policy. Clevedon: Channel View Publications. Dwyer, L., Forsyth, P., Spurr, R. and VanHo, T. (2006). Economic effects of the world tourism crisis on Australia. Tourism Economics, 12(2), 171-186. Dwyer, L., Forsyth, P., Spurr, R., and Hoque, S. (2010). Estimating the carbon footprint of Australian tourism. Journal of Sustainable Tourism, 18(3), 355-376. Easterly, W. and Kraay, A. (2000). Small states, small problems? Income, growth, and volatility in small states. World Development, 28(11), 2013-2027. Eugenio-Martin, J L., Martin-Morales, N. and Sinclair, M. T. (2008). The role of economic development in tourism demand. Tourism Economics, 14(4), 673-690. 309 Eugenio-Martin, J.,L., Martín-Morales, M.N. and Scarpa, R. (2004). Tourism and economic growth in Latin American countries: a panel data approach. Fondazione Eni Enrico Mattei Discussion Papers 26/2004. Retrieved 12 January 2012, from http://www.feem.it/userfiles/attach/Publication/NDL2004/NDL2004-026.pdf Eugenio-Martin, J.L. (2003). Modelling determinants of tourism demand as a 5-stage process: a discrete choice methodological approach. Tourism and Hospitality Research, 4(4), 341-354. Eugenio-Martin, J.L. (2008). Economic and econometric analysis of the determinants of participation and destination choice in tourism demand.Unpublished doctoral dissertation, University of Las Palmas of Gran Canaria. Eugenio-Martin, J.L. and Campos-Soria, J.A. (2010). Climate in the region of origin and destination choice in outbound tourism demand. Tourism Management, 31(6), 744-753. Eugenio-Martin, J.L. and Campos-Soria, J.A. (2011). The role of income and the substitution pattern between domestic and international tourism demand. Applied Economics, 43(20): 2519–2531. Eugenio-Martin, J.L. and Campos-Soria, J.A. (2014). Economic crisis and tourism expenditure cutback decision. Annals of Tourism Research, 44: 53–73. Eugenio-Martin, J.L., Sinclair. M.T. and Yeoman. I. (2005). Quantifying the effects of tourism crises: An application to Scotland. Journal of Travel and Tourism Marketing, 19(2/3): 23–36. European Commission (EC). (2013). VAT rates applied in the Member States of the European Union. Brussels, Belgium: European Commission. Retrieved 12 February, 2013, from http://ec.europa.eu/taxation_customs/resources/documents/taxation/vat/how_vat_works/rates/vat_rates_en.pdf 310 Eymann, A. (1995). Consumers’ Spatial Choice Behavior. Heidelberg: Physica-Verlag. Eymann, A. and Ronning, G. (1992). Discrete choice analysis of foreign travel demand. H. J. Vosgerau (Ed.), European integration in the world economy. Studies in international economics and institutions. Springer, Berlin. Fernández, F.M. and Manrique de Lara, C. (2006). La matriz de contabilidad nacional. Un método alternativo de presentación de las cuentas nacionales. Revista Asturiana de Economía, 37, 85-106. Fernández-Kranz, D. and Hon, M.T. (2006). A cross-section analysis of the income elasticity of housing demand in Spain: Is there a real estate bubble? The Journal of Real Estate Finance and Economics, 32(4), 449-470. Fiebig, D.G., Keane, M.P., Louviere, J. and Wasi, N. (2010). The generalized multinomial logit model: accounting for scale and coefficient heterogeneity. Marketing Science, 29(3), 393-421. Fisher, I. (1933). The debt-deflation theory of great depressions. Econometrica, 1(4), 337-357. Fleischer, A. Peleg, G. and Rivlin, J. (2011). The impact of changes in household vacation expenditures on the travel and hospitality industries. Tourism Management, 32(4), 815–821. Forsyth, P. (2003). Low-cost carriers in Australia: experiences and impacts. Journal of Air Transport Management, 9(5), 277-284. Frechtling, D.C. (1982). Tourism trends and the business cycle. Tourism Management, 3(4): 285-290. Fu, X., Dresner, M. and Oum, T.H. (2011). Effects of transport service differentiation in the US domestic airline market. Transportation Research Part E, 47(3), 297-305 311 Fujii, E.T., Khaled, M. and Mak, J. (1987). An empirical comparison of systems of demand equations for tourist expenditures in resort destinations. Philippine Review of Economics, 24, 79-102. Gago, A., Labandeira, X., Picos, F., and Rodríguez, M. (2009). Specific and general taxation of tourism activities. Evidence from Spain. Tourism Management, 30(3), 381-392. Ghali, M.A. (1976). Tourism and economic growth: an empirical study. Economic Development and Cultural Change, 24(3), 527-538. Gokovali, U., Bahar, O. and Kozak, M. (2007). Determinants of length of stay: a practical use of survival analysis. Tourism Management, 28(3), 736–746. Gómez Martín, M.B. (2005). Weather, climate, and tourism. A geographical perspective. Annals of Tourism Research, 32(3), 571-591. Gooroochurn, N. and Thea-Sinclair, M. (2005). Economics of tourism taxation: Evidence from Mauritius. Annals of Tourism Research, 32(2), 478-498. Gössling, S., and Hall, C. M. (2006). Uncertainties in predicting tourist flows under scenarios of climate change. Climatic change, 79(3-4), 163-173. Graham, A. and Dennis, N. (2010). The impact of low-cost airline operations to Malta. Journal of Air Transport Management, 16, 127-136. Greene, W.H. (1997). Econometric Analysis (3rd editions). New Yersey: Prentice-Hall International. Greer, T. and Wall, G. (1979). Recreational hinterlands: a theoretical and empirical analysis. Publication Series, Department of Geography, University of Waterloo, 14, 227-246. 318 Nicolau, J.L. and Más, F.J. (2005b). Stochastic modeling: a three-stage tourist choice process. Annals of Tourism Research, 32(1), 49-69. Nicolau, J.L. and Más, F.J. (2006). The influence of distance and prices on the choice of tourist destinations: The moderating role of motivations. Tourism Management, 27(5), 982 996. Nowak, J.J., Sahli, M. and Cortes-Jimenez, I. (2007). Tourism, capital good imports and economic growth: theory and evidence for Spain. Tourism Economics, 13(4), 515-536. O´Brien, A. (2012). Wasting a good crisis. Developmental failure and irish tourism since 2008 Annals of Tourism Research, 39(2): 1138–1155. O´Hagan, J.W. and Harrison, M.J. (1984). Market shares of US tourist expenditure in Europe: an econometric analysis. Applied Economics, 16, 919-931. O’Connell, J. F., & Williams, G. (2005). Passengers’ perceptions of low cost airlines and full service carriers: A case study involving Ryanair, Aer Lingus, Air Asia and Malaysia Airlines. Journal of Air Transport Management, 11(4), 259-272. Oh, C.O. (2005). The contribution of tourism development to economic growth in the Korean economy. Tourism Management, 26(1), 39-44. Okumus, F. and Karamustafa. K. (2005). Impact of an economic crisis. Evidence from turkey. Annals of Tourism Research, 32(4): 942–961. Okumus, F., Altinay, M. and Arasli, H. (2005). The impact of Turkeys´s economic crisis of February 2001 on the tourism industry in Northern Cyprus. Tourism Management, 26: 95– 104. 319 Olive Research. (2009). Impact of economic downturn on attitudes and behaviour of international tourists to Britain. London: VisitBritain. Retrieved 12 April 2012, from http://www.visitengland.org/Images/International%20research-%20full%20report_tcm3019717.pdf Page, S., Song, H. and Wu, D.C. (2012). Assessing the impact of the global economic crisis and swine flu on inbound economic crisis in the United Kingdom. Journal of Travel Research, 51(2): 142–153. Palmer, A., Beltrán, M. and Cortiñas, P. (2006). Robust estimators and bootstrap confidence intervals applied to tourism spending. Tourism Management, 27, 42-50. Papatheodorou, A. (1999). The demand for international tourism in the Mediterranean region. Applied Economics, 31(5), 619-630. Papatheodorou, A., Rosselló, J. and Xiao, H. (2010). Global economic crisis and tourism: Consequences and perspectives. Journal of Travel Research, 49(1), 39-45. Parrilla, J.C., Font, A.R. and Nadal, J.R. (2007). Tourism and long-term growth a Spanish perspective. Annals of Tourism Research, 34(3), 709-726. Poirier, R.A. (1995). Tourism and development in Tunisia. Annals of Tourism Research, 22(1), 157-171. Polo, C. and Sancho, F. (1993). Insights or forecasts? An evaluation of a computable general equilibrium model of Spain. Journal of forecasting, 12(5), 437-448. Prideaux, B. (1999). Tourism perspectives of the Asian financial crisis: lessons for the future. Current Issues in Tourism, 2(4): 279–293. 320 Pulina, M. and Cortés-Jiménez, I. (2010). Have low-cost carriers influenced tourism demand and supply? The case of Alghero, Italy. Tourism Analysis, 15, 617-635. Pyatt, G. and Round, J.I. (1985). Social accounting matrices: A basis for planning. Washington D.C: The World Bank,. Requeijo, G.J., Iranzo, M.J., Martínez de Dios, J., Pedrosa, R. M. and Salido, H.J.(2007). Técnicas Básicas de Estructura Económica (1st edición). Madrid: Delta publicaciones. Rey, B., Myro, R.L. and Galera, A. (2011). Effect of low-cost airlines on tourism in Spain. A dynamic panel data model. Journal of Air Transport Management, 17, 163-167. Riley, M., Ladkin, A. and Szivas, E. (2001). Tourism employment. Analysis and planning. Clevedon, Channel view. Aspect of Tourism. Roca, A.M. (1998). Dutch Disease and banana exports in the Colombian Caribbean, 19101950. ” paper presented at the meeting of the Latin American Cliometric Society (LACLIO) in Cartagena, Colombia, August 27-28. Retrieved 20 October, 2013, from http://www.stanford.edu/group/sshi/lacsc_papers.html. Rosselló, J. and Riera, A. (2012). Pricing European package tours: the impact of new distribution channels and low-cost airlines. Tourism Economics, 18(2), 265-279. Rosselló, J. and Santana-Gallego, M. (2014). Recent trends in international tourist climate preferences: a revised picture for climatic change scenarios. Climatic Change, 124, 119-132. Rosselló-Nadal, J. (2014). How to evaluate the effects of climate change on tourism. Tourism Management, 42, 334-340. Rostow, W.W. (1960). The process of economic growth: A Non-Communist Manifesto. Cambridge: Cambridge University Press. 321 Rutherford, T.F. (1994). The GAMS/MPSGE and GAMS/MILES user notes. Washington D.C: GAMS Development Corporation. Rutherford, T.F. (1999). Applied general equilibrium modeling with MPSGE as a GAMS subsystem: An overview of the modeling framework and syntax. Computational Economics, 14(1-2), 1-46. Rutherford, T.F. and Light, M.K. (2002). A general equilibrium model for tax policy analysis in Colombia. Department of National Planning Discussion Paper 22,. Sachs, J.D. and Larraín, B.F. (1994): Macroeconomía en la economía global. Mexico D.F, México: Prentice Hall Hispanoamérica (3rd edition). Schmidt, S.J. (2005). Econometría. Mexico: Mc Graw-Hill Interamericana. Seddighi, H. R. and Theocharous, A.L. (2002). A model of tourism destination choice: a theoretical and empirical analysis. Tourism management, 23(5), 475-487. Sequeira, T.N. and Maças, N,P. (2008). Does tourism influence economic growth? A dynamic panel data approach. Applied Economics, 40(18), 2431-2441. Sheldon, P. and Dwyer, L. (2010). The global financial crisis and tourism: Perspectives of the academy. Journal of Travel Research, 49(1), 3–4. Sinclair, M.T. and Stabler, M. (1997). The economics of tourism. Routledge Skrondal, A. and Rabe-Hesketh, S. (2004). Generalized latent variable modeling. Chapman & Hall/CRC. Smeral, E. (2003). A structural view of tourism growth. Tourism Economics, 9(1), 77–93. 322 Smeral, E. (2009). The impact of the financial and economic crisis on European tourism. Journal of Travel Research, 48(1), 3-13. Smeral, E. (2010). Impacts of the world recession and economic crisis on tourism: Forecasts and potential risks. Journal of Travel Research, 49(1), 31-38. Smith, A. (1776). La Riqueza de las naciones. Economía, Madrid: Alianza editorial. Solow, R.M. (1956). A contribution to the theory of economic growth. The Quarterly Journal of Economics, 70(1), 65-94. Song, H. and Li, G. (2008). Tourism Demand Modeling and Forecasting—A Review of Recent Research. Tourism Management, 29(2), 203-220. Song, H. and Lin, S. (2010). Impacts of financial and economic crisis on tourism in Asia. Journal of Travel Research, 49(1), 16–30. Sorenson, N. (1991). The impact of geographic scale and traffic density on airline production costs: the decline of the no-frills airlines. Economic Geography, 67(4), 335–345. Spanish Statistical Office (INE). (2004). Cuenta Satélite del turismo de España: nota metodológica. Instituto Nacional de Estadística, Madrid. Retrieved 10 January, 2009, from http://www.ine.es/metodologia/t35/metosateln.pdf Sugiyarto, G., Blake, A. and Sinclair, M.T. (2003). Tourism and globalization: economic impact in Indonesia. Annals of Tourism Research, 30(3), 683-701. Tang, C.F. and Abosedra, S. (2012). Small sample evidence on the tourism-led growth hypothesis in Lebanon. Current Issues in Tourism, 17(3), 1-13. Taylor, Ch.E. and Knudson, D.M. (1976). Area preferences of Midwestern campers. Journal of Leisure Research, 12(Spring), 39–48. 323 Train, K. (1998). Recreation demand models with taste differences over people. Land economics, 74(2), 230-239. Train, K. (2009). Discrete choice methods with simulation. Cambridge: Cambridge University Press. United Nations World Tourism Organisation. (2008). Tourism satellite account: Recommended methodological framework, 2008. Washington D.C: United Nations World Tourism Organisation . Retrieved, 10 September, 2011, from http://unstats.un.org/unsd/statcom/doc08/BG-TSA.pdf United Nations World Tourism Organisation. (2013). Tourism highlights, 2013 edition. Madrid, Spain: United Nations World Tourism Organisation. Retrieved 10 September, 2011, from http://mkt.unwto.org/publication/unwto-tourism-highlights-2013-edition Usui, N. (1996). Aid induced structural change in developing countries: An extension of the two-gap model. Singapore Economic Review, 41, 53–66. Van Soest, A. and Kooreman, P. (1987). A micro-econometric analysis of vacation behaviour. Journal of Applied Econometrics, 2, 215-226. VanWijnbergen, S. (1986). Exchange rate management and stabilization policies in developing countries. Journal of Development Economics, 23, 227–247. Vera, J.F. and Ivars, J.A. (2009). Spread of low-cost carriers: tourism and regional policy effect in Spain. Regional Studies, 43(4), 559-570. Von Thünen, J.H. (1826). Der Isolierte Staat in Beziehung auf landwrirtschaft und Nationalekonomie (Hamburg). 324 Wang, Y.S. (2009). The impact of crisis events and macroeconomic activity on Taiwan´s international inbound tourism demand. Tourism Management, 30, 75–82. Wattanakuljarus, A. and Coxhead, I. (2008). Is tourism-based development good for the poor? A general equilibrium analysis for Thailand. Journal of Policy Modeling, 30(6), 929-955. White, H. (1992). The macroeconomic impact of development aid: A critical survey. Journal of Development Studies, 28, 163–240. Windle, R. and Dresner, M. (1999). Competitive responses to low cost carrier entry. Transportation Research Part E, 35, 59-75. Wolfe, R.I. (1970). Communication. Journal of Leisure Research, 2(1), 85–87. Wolfe, R.I. (1972). The inertia model. Journal of Leisure Research, 4, 73–76. World Bank. (1993). The east Asian miracle. Oxford: Oxford University Press. Wu, D.C., Li, G. and Song, H. (2012). Economic analysis of tourism consumption dynamics: A time-varying parameter demand system approach. Annals of tourism research, 39, 2, 667685. Young, J. and Whang, T. (2011). The impact of low cost carriers on korean island tourism. Journal of Transport Geography, 19(6), 1335-1340. Zellner, A. and Theil, H. (1962). Three-least squares: simultaneous estimation of simultaneous equations. Econometrica, 30(1), 54-78. Zhou, D., Yanagida, J.F., Chakravorty, U. and Leung, P. (1997). Estimating economic impacts from tourism. Annals of Tourism Research, 24(1), 76.