A nonlinear dynamic model to international tourism demand in Spanish Mediterranean coasts

65 4, XXI, 2018 Economics DOI: 10.15240/ ul/001/2018-4-005 In oduc ion The Medi e anean is one o he main des ina ions o in e na ional ou ism in Spain. Acco ding o he Spanish S a is ics Ins i u e (INE, 2016), mo e han hal o in e na ional ou is s s aying a ho els chose he Medi e anean coas al p o inces as a des ina ion in 2015. Spain has abou 3,500 km o Medi e anean coas line (INE, 2016). As shown in Fig. 1, hese kilome e s a e dis ibu ed be ween he peninsula coas (2,058 km) and he a chipelago o he Balea ic Islands (1,428 Km). Tou ism is an impo an economic sec o on he Spanish Medi e anean coas s and i has become one o he mos impo an sou ces o employmen . Fo example, in Balea ic Islands he ou ism sec o con ibu ed 44.8% o he g oss domes ic p oduc (GDP) and c ea ed 150,346 jobs (32.0% o o al), in 2014. These da a a e signi i can ly highe han he na ional a e age (11.1% o he GDP and 13% o he employmen ). Acco ding o he Ho el Occupancy Su ey (INE, 2016), in 2015 he p o inces o he Medi e anean coas s accoun o 47.15% o he o al s a employed in ho els in Spain, and 47.36% o ho el beds o e ed. Mo eo e , he numbe o isi o s is inc easing e e y yea . In e na ional ou is s s aying a ho els ha e g own by 26.3% om 2010 o 2015, wi h a mean annual g ow h a e o nea ly 5% du ing his pe iod (INE, 2016). In his con ex o a consolida ed des ina ion wi h a g owing ou ism demand, we conside i e y in e es ing o gain a be e knowledge o he de e minan s o i s demand. A NONLINEAR DYNAMIC MODEL FOR INTERNATIONAL TOURISM DEMAND ON THE SPANISH MEDITERRANEAN COASTS Isabel Albaladejo, Ma ibel González-Ma ínez Fig. 1: The Spanish Medi e anean coas s Sou ce: own EM_4_2018.indd 65EM_4_2018.indd 65 28.11.2018 13:12:5128.11.2018 13:12:51 66 2018, XXI, 4 Ekonomie Since he 1990s, ou ism demand modelling s udies ha e shi ed om s a ic eg ession models o mo e sophis ica ed dynamic speci i ca ions. Dynamic demand models can accoun o habi pe sis ence and wo d-o -mou h ecommenda ions om p e ious isi o s. The mos common way o b ing a dynamic s uc u e in o demand models is o include he lagged demand in a linea ashion as an explana o y a iable (Mo ley, 2009; Ga ín-Muñoz, 2006; 2007; 2009, among o he s). This p oposal implies ha he e ec o p e ious ou is s on he cu en ou ism demand is cons an (Albaladejo, González-Ma ínez, & Ma ínez-Ga cía, 2016). Howe e , se e al s udies (Mo ley, 1998; 2009; Albaladejo, González-Ma ínez, & Ma ínez- Ga cía, 2016) a i m ha a linea model is no enough o accoun o he dynamics o ou ism demand. We a gue ha he cons an e ec assump ion is no app op ia e o analyzing ou ism demand. The in l uence o pas ou is s will be la ge o smalle depending on he epu a ion o a ac i eness o he ou ism des ina ion. The a ac i eness o a des ina ion depends no jus on i s na u al ea u es (wea he , beaches, na u al a eas...) bu also on cha ac e is ics like quali y o accommoda ions, quan i y o ou ism se ices, di e si y o supply, and conges ion o o e c owding, among o he s. The quali y o ou ism se ices is one o he main ac o s a ec ing he success o ou ism des ina ions and many des ina ions a e adop ing a policy o quali y se ice in o de o consolida e hemsel es as an al e na i e op ion in he highly compe i i e ma ke s o ou ism (Albaladejo, González-Ma ínez, & Ma ínez-Ga cía, 2014). Ano he impo an ac o is o e c owding (San ana-Jiménez & He nández, 2011). Today, he e is conce n abou he conges ion su e ed by se e al adi ional des ina ions (Ba celona, London, Pa is...). In a des ina ion like he Spanish Medi e anean coas s, i is impo an o know i hese a ibu es in l uence he e ec o he pas ou is s on cu en demand and how. In his pape , he in e na ional ou ism demand in he Spanish Medi e anean coas s is analyzed using annual da a o he pe iod 2005 o 2015. A dynamic ou ism model is p oposed which allows he e ec o p e ious ou is s o a y wi h he cha ac e is ics o he des ina ion. Ou model is an ex ension o he s anda d dynamic equa ion o ou ism demand o include in e ac ion e ec s be ween p e ious ou is s and wo ea u es: quali y o ou ism se ices and ou ism conges ion. The aim o he wo k is o s udy how bo h a ibu es a ec he ela ionship be ween p e ious ou is s and cu en ou ism demand on he Spanish Medi e anean coas s. Empi ical i ndings on quali y and conges ion play an impo an ole in policy decisions. A sys em GMM dynamic panel da a analysis (Blundell & Bond, 1998) is ca ied ou o es ima e he model. The da a a e disagg ega ed bo h by p o ince o des ina ion ( he 11 Spanish p o inces ha make up he Medi e anean a ea) and by coun y o o igin. Since he main ou is ma ke is he Eu opean Union (EU), we conside Eu opean ou is s om he ollowing coun ies: Belgium, F ance, Ge many, Holland, I aly, Po ugal and Uni ed Kingdom. In 2015, mo e han 86% o Eu opean Union ou is s came om one o hese coun ies (INE, 2016). The esul s show e idence o a s ong pe sis ence in ou ism demand. P e ious ou is s ha e an impo an posi i e and non-cons an e ec . I is posi i ely in l uenced by he quali y o he ou ism se ices and nega i ely by ou ism conges ion. Ou analysis p o ides e idence ha he p e ious ou is s e ec depends on he a ac ion o epu a ion o he des ina ion. This esul suppo s a nonlinea dynamic ou ism demand model. In addi ion, o he bes o he au ho s´ knowledge, his is he i s ecen s udy o conside he Spanish Medi e anean coas s as a whole. Mos p e ious s udies ha e ied o explain he demand in di e en Medi e anean coun ies (Ga & Falzon, 2014; Papa heodo ou, 1999). O he s udies on Spain analyze he majo ou is cen e s o he Medi e anean, especially he Balea ic Islands (Ga ín-Muñoz & Mon e o- Ma ín, 2007; Rosselló, Aguiló, & Rie a, 2005; Aguiló, Rie a, & Roselló, 2005). The pape is o ganized as ollows. The ollowing sec ion discusses he ole o he p e ious ou is s on ou ism demand. I summa izes he empi ical li e a u e o he panel da a case, and ou dynamic model is p oposed. Sec ion 2 p esen s he da a and some desc ip i e s a is ic o he a iables conside ed in he s udy. I also p o ides he empi ical model and desc ibes he econome ic me hod used o es ima ion. Sec ion 3 con ains he esul s and hei in e p e a ion. Finally, he las sec ion d aws some conclusions. EM_4_2018.indd 66EM_4_2018.indd 66 28.11.2018 13:12:5128.11.2018 13:12:51 67 4, XXI, 2018 Economics 1. The Impac o P e ious Tou is s 1.1. Li e a u e Re iew The e is a widesp ead ag eemen ha ou ism demand is likely o be a ec ed by p e ious ou is s, ei he because o hei in l uence on o he po en ial isi o s, o because hey epea des ina ions (Mo ley, 2009; Ga ín-Muñoz, 2006; 2007; 2009, among o he s). P e ious ou ism l ows inc ease in o ma ion abou a des ina ion, he eby educing unce ain y o po en ial isi o s (wo d-o -mou h ecommenda ions). Besides, since he e is less unce ain y associa ed wi h a des ina ion wi h which you a e al eady amilia , habi s migh induce ou is s no o a y hei des ina ion o e ime (habi pe sis ence). Rega ding ou ism demand modelling, dynamic econome ic models allow us o ake in o accoun he causal ela ionship be ween p e ious isi o s’ l ows and he cu en demand. Mos common dynamic speci i ca ions include he p e ious ou is s as an explana o y a iable in he model. Focusing on he panel da a analysis, he e is an impo an numbe o s udies ha ha e es ima ed a lagged dependen a iable model. As examples, we ha e he wo k by Maloney and Mon es Rojas (2005) o ou is demand a Ca ibbean des ina ions; Naudé and Saayman (2005) o ou is demand in 43 A ican s a es; Ga ín-Muñoz (2006; 2007; 2009) o ou ism demand a di e en Spanish des ina ions; Ga ín-Muñoz and Mon e o-Ma ín (2007) o ou ism demand in he Balea ic Islands (Spain); Massidda and E zo (2012) o domes ic ou ism in I aly; and Rod íguez, Ma ínez-Roge and Pawlowska (2012) o academic ou ism demand in Galicia (Spain). Mo e ecen ly, Capacci, Sco cu and Vici (2015) includes wo lags o he dependen a iable o analyze he impac o he Blue Flag on o eign ou is s on he I alian coas s; and Pop awe (2015) uses a panel da a se o o e 100 coun ies o es he hypo hesis ha co up ion has a nega i e e ec on ou ism. Since all hese s udies include he p e ious ou is s in a linea ashion, he e ec o his a iable on he cu en ou ism demand is assumed cons an o e ime and o he c oss-sec ion. Thus, i is independen o a iables like he quali y o ou ism se ices and ou ism conges ion, which may a ec he des ina ion’s epu a ion. Mo ley (1998; 2009) and Albaladejo, González-Ma ínez and Ma ínez-Ga cía (2016) ha e ques ioned his way o inco po a ing dynamics in o he model. Mo ley (2009) a gues ha he simple inclusion o he lagged dependen a iable allows epea isi s o be inco po a ed in o a model, bu no wo d-o - mou h ecommenda ions. He shows ha his in o ma ion l ow has gene ally been neglec ed in he li e a u e. Using he di usion model (Bass, 1969; Mahajan, Mulle , & Bass, 1990), Mo ley (1998; 2009) inco po a e ele an ou ism in o ma ion l ows in a adi ional ou ism demand model. The esul is a nonlinea model ha includes quad a ic unc ions o p e ious ou is s as e ms. Mo ley (1998), Rosselló, Aguiló and Rie a (2005), Aguiló, Rie a and Roselló (2005) and Hsu and Wang (2008) ha e es ima ed he model wi h ime se ies da a and i nd e idence o suppo his model. Albaladejo, González-Ma ínez and Ma ínez-Ga cía (2016) a gue ha he cons an e ec o he p e ious ou is s esul ing om he usual dynamic econome ic model is no in acco dance wi h he Tou ism A ea Li e Cycle (TALC) heo y o Bu le (1980), one o he mos accep ed heo ies in ou ism li e a u e. Taking in o accoun bo h his heo y and he adi ional ou ism demand model, Albaladejo, González- Ma ínez and Ma ínez-Ga cía (2016) p opose a new dynamic speci i ca ion ha includes a quad a ic unc ion o p e ious demand. They es he model using panel da a om Spanish egions du ing he pe iod 2000-2013. Thei empi ical esul s show a posi i e and dec easing e ec o he p e ious ou is s. Bo h app oaches (di usion model and TALC model) sugges ha usual cons an elas ici y demand models a e likely misspeci i ed. The e o e, a nonlinea dynamic speci i ca ion is p e e ed. We ag ee wi h his gene al conclusion, bu we add a new a gumen no p esen in he p e ious li e a u e. A linea model does no allow one o link he e ec o pas ou is s o epu a ion o a ac i eness o he ou ism des ina ion, assuming a cons an e ec . Howe e , he in l uence o p e ious ou is s may be a ec ed by changes in he epu a ion o he ou ism des ina ion. In es men in he ou ism indus y can imp o e he des ina ion’s epu a ion (Albaladejo & Ma ínez-Ga cía, 2016), implying a mo e posi i e e ec o p e ious ou is s on he cu en demand. The quali y o ou ism se ices o a des ina ion and i s ou is o e c owding a e wo cha ac e is ics wi h a high impac on i s isi o s (Vajčne o á, Žia an, Ryglo á, & And áško, 2014; Bee li & Ma in, 2004). EM_4_2018.indd 67EM_4_2018.indd 67 28.11.2018 13:12:5128.11.2018 13:12:51 68 2018, XXI, 4 Ekonomie 1.2 A Nonlinea Dynamic Demand Model In his pape , we p opose a nonlinea dynamic model whe e cu en ou ism demand can be in l uenced by p e ious ou is s. Ou model allows us o s udy how he numbe o p e ious ou is s in e ac s wi h wo impo an ea u es o he ou is des ina ion: ou ism se ices quali y, and ou ism conges ion. Fo his pu pose, we add o he s anda d dynamic ou ism model wo in e ac ion e ms. The model can be exp essed as 01121 1     TTTQ    31 1  TO X  , (1) whe e subsc ip s and -1 deno e he ime pe iod. The dependen a iable is T, he numbe o ou is s, Q is he se ices quali y, O is he des ina ion´s o e c owding, and X′=(x¹,x²,...,xK) is he ec o o he emaining k explana o y a iables (p ice, income, e c.), which can also include lagged explana o y a iables and dummy a iables. Thus, T –1∙Q –1 cap u es he in e ac ion be ween p e ious ou is s and quali y o ou ism se ices, and T –1∙O –1 he in e ac ion be ween he p e ious ou is s and he o e c owding o he des ina ion. In his model, he ma ginal e ec o T -1 on T is a ec ed by wo cha ac e is ics o he des ina ion. I is measu ed by he exp ession: 12131 1      TQO T   , (2) This e ec measu es he impac o he p e ious ou is s on he cu en demand. I is no cons an . Bo h p e ious le els o quali y and o e c owding a he des ina ion can modi y his e ec . I , as expec ed, β1 and β2 a e posi i e and β3 is nega i e, i is a ma ginal e ec inc easing wi h Q -1 and dec easing wi h O -1 (No e ha he mos common dynamic speci i ca ion se s β2 = β3 = 0, omi ing in e ac ion e ec s. In he esul ing linea model, he ma ginal e ec is cons an .). No e ha , i T is measu ed wi h loga i hms, as is usual in empi ical ou ism demand s udies, Equa ion (2) means ha he elas ici y o ou ism demand wi h espec o lagged demand is no cons an bu dependen on quali y and conges ion. The g ea e he se ices quali y pe cei ed by he ou is and/o he smalle he massi i ca ion in he ou is des ina ion, he be e he epu a ion ha his des ina ion will ha e. A be e epu a ion implies a highe impac o he p e ious ou is s on he cu en demand. In con as , i he inc eases in he isi o s o a des ina ion a e no accompanied by an adequa e in es men in quali y and quan i y o he ou ism se ices, ou is s’ opinion abou he des ina ion will be wo s , and hei posi i e e ec on he u u e demand will be lowe . Tha is, a des ina ion can cushion he downwa d dynamics o his e ec by in es ing in ou ism. 2. Da a and Me hodology 2.1 Da a and Va iables The Medi e anean coas s a e he main des ina ion o in e na ional ou ism in Spain. Acco ding o he INE (2016), in 2015, abou 25.7 million in e na ional ou is s s ayed a ho els in he Medi e anean p o inces. They ep esen 55% o he in e na ional ou is s a i ing in Spain and s aying in ho els (The in e na ional ou is s who chose ho els and simila es ablishmen s as accommoda ion in Spain ep esen ed 67% o o al a i als in 2015 (INE, 2016)). The e olu ion o hese ou is s om 2005 o 2015 is p esen ed in Fig. 2. Tou ism ose sha ply om 2005 o 2006, bu s agna ion is obse ed in 2007 and 2008, and a sha p d op occu s in 2009, as esul o he i nancial c isis and he economic ecession. Since 2010, he numbe o ou is s seems o be expe iencing a new g ow h phase. To explain he in e na ional ou ism demand o he Spanish Medi e anean coas s, we es ima e he model p oposed in Sec ion 1.2 using annual da a disagg ega ed by p o ince o des ina ion and coun y o o igin. We use a balanced panel da a se consis ing o he 11 p o inces ha make up he Spanish Medi e anean coas s and he 7 Eu opean coun ies which a e he main o igin ma ke s o he pe iod 2005-2015. Panel da a ha e some ad an ages o e c oss sec ional o ime se ies da a. One is ha hey enable us o con ol o unobse able c oss sec ional he e ogenei y, which is common in p o incial da a. Time se ies and c oss sec ion s udies no con olling o his he e ogenei y un he isk o ob aining biased esul s. Mo eo e , panel da a usually gi e a la ge numbe o da a poin s, so inc easing he deg ees o eedom, educing he collinea i y among explana o y a iables and imp o ing he e i ciency o econome ic es ima es (Hsiao, 2003; Bal agi, 2008). EM_4_2018.indd 68EM_4_2018.indd 68 28.11.2018 13:12:5228.11.2018 13:12:52 69 4, XXI, 2018 Economics Ou model includes economic demand a iables, such as income and p ices, a dummy a iable o con olling he e ec s o he economic c isis, and a nonlinea e m o cap u e he e ec o p e ious ou is s. This e m means he e ec o he p e ious ou is s can be no cons an , bu can depend on he quali y o he ou ism se ices and he ou is conges ion a he des ina ions (Equa ion 2). The dependen a iable is he numbe o in e na ional ou is s (T) who choose ho els and simila es ablishmen s as accommoda ion (We a e awa e ha using only he da a o ou is s s aying a ho els is a limi a ion. Howe e , in his pape we ha e chosen o be syn he ic and o p esen only ho el demand da a. In any case, in Spain ho el demand da a a e qui e signi i can because hey ep esen a ound 70% o he in e na ional demand egis e ed in Spain h oughou he pe iod (INE, 2016)). Da a a e aken om he Ho el Occupancy Su ey (INE, 2016). Two adi ional economic ac o s a e included among he explana o y a iables: o igin income and p ice. To measu e o igin income, we use he eal pe capi a GDP o each o igin coun y (GDP). This a iable was aken om OCDE (2016). The p ice a iable included in ou model e l ec s he cos o li ing o ou is s a he di e en des ina ions ela i e o he cos o li ing in he coun y o o igin (IP): / /  des ina ion des ina ion o igin o igin Spain o igin CPI IP CPI EX (3) whe e CPIdes ina ion and CPIo igin a e he consume p ice indices (CPIs) o each o he 11 des ina ions conside ed and each o igin coun y, espec i ely; EXSpain/o igin is he nominal e ec i e exchange a e o Spain s each coun y (The nominal exchange a e be ween Spain and Eu ozone coun ies is equal o 1. The e o e, we only need o mul iply he CPI o he o igin coun y by he nominal exchange a e in he case o he Uni ed Kingdom). Da a on exchange a es and CPIs o each coun y we e collec ed om Eu os a (2016). Da a on CPI o he 11 Spanish p o inces we e collec ed om he INE (2016). Addi ionally, based on Fig. 2, we conside a dummy a iable o cap u e he in l uence on ou ism o he i nancial and economic c isis. Fig. 2: In e na ional ou is s lodged a ho els in Spanish Medi e anean p o inces Sou ce: own om he da a p o ided by he INE (2016) EM_4_2018.indd 69EM_4_2018.indd 69 28.11.2018 13:12:5228.11.2018 13:12:52 70 2018, XXI, 4 Ekonomie This a iable, D2009, akes he alue 1 om 2009 onwa d and 0 in o he yea s. To build he in e ac ions included in he nonlinea e m o ou model, a measu e o quali y o ou ism se ices (Q) and a measu e o ou is conges ion (O) ha e o be de i ned. Measu ing bo h ea u es is no an easy ask. The e is no uni e sal de i ni ion o ei he o hese a ibu es in ou ism. Wi h ega d o quali y, we ocus on he supply o ou ism se ices, speci i cally on he quali y o ou ism accommoda ions. Nicolau and Selle s (2010), and he e e ences he ein, highligh ha quali y o accommoda ions is s a egic o inc easing ou ism compe i i eness. We ha e ollowed his idea in ou pape . The pe cen age o luxu y ho els is u ilized as a p oxy o quali y o ou ism accommoda ions. We use he o i cial classi i ca ion sys em in Spain om 1 o 5 golden s a s and om 1 o 3 sil e s a s o de i ne he ca ego y o ho els. Conside ing as luxu y ho els he ou and i e golden s a s ho els, he quali y o ou ism se ices is de i ned as 100 ou and i e golden s a s ho els o al numbe o ho els Q (4) Da a on ho els we e collec ed om he INE (2016). The a io (4) has also been used as quali y measu e in he pape o Albaladejo, González-Ma ínez and Ma ínez-Ga cía (2014). Conges ion o o e c owding o a des ina ion inc eases when he numbe o isi o s is excessi e in ela ion o he space o capaci y o he des ina ion o accommoda e hose ou is s, especially du ing peak pe iods. In o de o ob ain an idea abou he conges ion in each p o ince, we conside he ela ionship be ween i s ou ism demand and i s ou ism supply, speci i cally we use he a io be ween he o al numbe o ou is s lodged a ho els and he o al numbe o ho el beds as p oxy o ou is conges ion a each des ina ion p o ince domes ic and in e na ional ou is s o al numbe o beds O (5) Da a on ho el beds and ou is s we e collec ed om he INE (2016). The a io (5) can be conside ed as a measu e o he densi y o ou is s lodged a ho els. I epo s on he ela ionship be ween ou ism demand and ou ism supply in each p o ince. The g ea e he numbe o ho el beds in a des ina ion, he highe he chance o accommoda ing isi o s sui ably. A ele an ad an age o using bo h measu es (Q and O) is ha hei homogeneous cha ac e allows compa isons among se e al p o inces. 2.2 Me hodology and Model Speci i ca ion Following he model p oposed in Sec ion 2 and conside ing he a iables de i ned abo e, he econome ic model is ep esen ed as ,1,12,1,1     ij ij ij ij i TTTQ   3,1 ,1 4 , 5 ,   ij i j ij TO GDP IP  6, 2009 ij D  (6) whe e he subsc ip i deno es he des ina ion p o ince (Alican e, Alme ia, Balea ic Islands, Ba celona, Cas ellon, Gi ona, G anada, Malaga, Mu cia, Ta agona and Valencia); j deno es he o igin coun y (Belgium, F ance, Ge many, Holland, I aly, Po ugal and Uni ed Kingdom), and indica es he ime pe iod ( = 2005-2015). ηij is he unobse ed p o incial-speci i c a iable (o i xed e ec s) ha a ies ac oss p o inces and o igin coun ies, bu is in a iable o e ime, and εij, is a dis u bance e m. A key assump ion h oughou his pape is ha he dis u bance εij, is unco ela ed ac oss p o inces, bu p o incial he e oscedas ici y and se ial co ela ion a e allowed o . All a iables a e exp essed in na u al loga i hms so ha he coe i cien s may be in e p e ed as elas ici ies. As discussed in Sec ion 1.2, he e ec o he p e ious ou is (Tij, -1) depends on β1, β2, β3, and he p e ious quali y and conges ion o he des ina ion (see Equa ion 2). Since a log-log model is used, his e ec ep esen s he elas ici y o cu en ou ism demand wi h espec o p e ious demand. A posi i e sign is expec ed o β1, hus a posi i e β2 would imply ha his elas ici y inc eases wi h he p e ious le el o quali y o he ou ism supply, gi en a le el o ou is conges ion. Likewise, a nega i e β3 would imply ha he elas ici y dec eases wi h he p e ious ou is conges ion, gi en a le el o quali y. I β2 and β3 a e ze o, he elas ici y is cons an ac oss he des ina ions and h ough ime. As usual in demand models, we expec a posi i e sign o β4 and a nega i e sign o β5 and β6. EM_4_2018.indd 70EM_4_2018.indd 70 28.11.2018 13:12:5228.11.2018 13:12:52 71 4, XXI, 2018 Economics A gene alized me hod o momen s (GMM) panel da a es ima ion (A ellano & Bond, 1991; A ellano & Bo e , 1995; Blundell & Bond, 1998) was applied o conduc ou empi ical analysis. O dina y Leas Squa es (OLS) is no app op ia e o es ima e dynamic panel models wi h he lagged dependen a iable among he eg esso s. The lagged dependen a iable is co ela ed wi h he unobse able p o incial e ec (ηi), which gi es ise o “dynamic panel bias” (Nickell, 1981). The wi hin g oups and andom e ec s es ima o s do no elimina e he “dynamic panel bias” and a e also biased and inconsis en . To sol e his p oblem, A ellano and Bond (1991) sugges i s di e encing he model o emo e he unobse ed i xed e ec s (ηi). As he di e enced lagged dependen a iable is s ill po en ially endogenous, i is ins umen ed wi h lagged le els o he endogenous a iable o sol e he p oblem o au oco ela ion. I he εij, a e no se ially co ela ed, we can use lags 2 and upwa ds o he endogenous a iable as ins umen s. Blundell and Bond (1998) ex ended his es ima o by building a sys em o equa ions o med by he equa ion in i s di e ences and he equa ion in le els. The ex ended GMM es ima o , called sys em GMM, uses lagged i s -di e ences as ins umen s o equa ion in le els, in addi ion o he usual lagged le els as ins umen s o equa ion in i s -di e ences. In his pape , we apply he sys em GMM (Blundell & Bond, 1998) p ocedu e o es ima e he model (6). We use he one-s ep obus o he e oscedas ici y es ima o and he wo-s ep es ima o o compa ison (One-s ep GMM es ima o is based on he assump ion ha he εij, a e i.i.d. In his pape , we use one-s ep obus es ima o s, whe e he esul ing s anda d e o s a e consis en wi h panel-speci i c au oco ela ion and he e oscedas ici y). Al hough he wo-s ep es ima o is heo e ically p e e ed, i is app op ia e o conside he one- s ep esul s when making in e ences, since he asymp o ic s anda d e o s o one-s ep GMM es ima o s a e i ually unbiased (A ellano & Bond, 1991). A c ucial assump ion o he alidi y o GMM is ha he ins umen s a e exogenous. We conduc wo diagnos ic es s: Hansen (1982) J- es s o he o e iden i ying es ic ions o he GMM es ima o s, and he A ellano and Bond (1991) es o au oco ela ion in he dis u bance e m, εij, (The Hansen s a is ics is a chi-squa ed es o de e mine i he esiduals a e co ela ed wi h he ins umen a iables. I nonsphe ici y is suspec ed in he e o s, he Hansen o e iden i i ca ion es is heo e ically supe io o he Sa gan (1958) es ). 3. Resul s We show wo di e en GMM es ima es: one- s ep and wo-s ep e sions o he sys em GMM (GMM-SYS). In bo h es ima es he lagged dependen a iable and he wo lagged in e ac ion e ms a e ea ed as endogenous. Since he usual o mulas o coe i cien s anda d e o s in wo-s ep GMM end o be downwa d biased when he ins umen coun is high, we use he Windmeije (2005) s anda d e o s co ec ion. The empi ical esul s om he es ima ion o he model a e shown in Tab. 1. The es ima ed coe i cien o he lagged dependen a iable is signi i can and posi i e, and he es ima ed coe i cien s o he in e ac ion e ms a e bo h signi i can and ha e opposing signs. As expec ed, he e ec o he lagged dependen a iable depends posi i ely on he pe cen age o luxu y ho els, and nega i ely on he a io be ween ou is s and ho el beds. Thus, he e is a non-cons an e ec o he p e ious numbe o ou is s o e cu en ou is s. Addi ionally, he esul s e eal a gene ally sa is ac o y pe o mance o he econome ic models. The au oco ela ion es s (A ellano & Bond, 1991) do no de ec any se ial co ela ion p oblem in he esiduals. As expec ed, he esiduals in di e ences a e au oco ela ed o o de 1, while he e is no au oco ela ion o second o de . In addi ion, he Hansen (1982) J- es does no ejec he null o join alidi y o he ins umen s. Bo h es ima es (one-s ep and wo-s ep) yield simila esul s. All a iables a e s a is ically signi i can . Es ima ed β1 (0.9626, 0.9674) and β2 (0.0149, 0.0147) a e posi i e, and es ima ed β3 (-0.0159, -0.0182) is nega i e. Thus, he elas ici y o ou ism demand wi h espec o he lagged demand is posi i e, inc easing wi h he pe cen age o luxu y ho els, and dec easing wi h he a io be ween ou is s and ho el beds a he des ina ion. This means ha quali y o ou ism se ices and ou is conges ion a e ele an o explaining in e na ional ou ism demand in he Medi e anean coas s. The implica ion o his esul is ha , in o de o a ac mo e ou is s, he supplie s o ou ism p oduc s should imp o e hei se ice quali y EM_4_2018.indd 71EM_4_2018.indd 71 28.11.2018 13:12:5328.11.2018 13:12:53 72 2018, XXI, 4 Ekonomie and adap ho el capaci y. The es ima ed income elas ici y (0.0301, 0.0294) is posi i e, showing ha he a i al o Eu opean ou is s o Spanish Medi e anean coas s depend posi i ely on he weal h o i s o igin coun y. As expec ed, nega i e p ice elas ici y is es ima ed wi h alues o -0.1752 and -0.2148, sugges ing ha ou is a i als a e also sensi i e o p ice changes. Finally, he dummy a iable ep esen ing he impac o he global c isis has he expec ed nega i e sign (-0.0989 and -0.0971). One o he mos impo an de e minan s o he in e na ional ou ism on he Spanish Medi e anean coas s seems o be he lagged dependen a iable, which con ols bo h he e ec o he wo d-o -mou h ecommenda ions and he e ec o habi pe sis ence. Since β2 and β3 a e signi i can , he e ec o his a iable is non-cons an , bu a ies ac oss he des ina ions and o e ime. In any gi en yea , i a ies be ween he di e en Medi e anean p o inces depending on i s pe cen age o luxu y ho els and he ela ionship be ween i s ou ism demand ( ou is s) and ou ism supply (ho el beds). Mo eo e , gi en a pa icula des ina ion, i inc eases wi h he quali y o se ices o e ed by ho els, and dec eases when he inc ease in he ou is s is no accompanied by a p opo ional inc ease in he numbe o ho el beds. In o de o show hese esul s, he elas ici y o ou ism demand wi h espec o he p e ious ou is s has been calcula ed in each o he p o inces using he es ima ed coe i cien s om he second column o Tab. 1. Fig. 3 shows hese es ima ed elas ici ies o he yea s 2006 and 2015 (Since elas ici y depends on cha ac e is ics o he ou is des ina ion, o each p o ince he elas ici y o p e ious ou is s is he same o ou is s a i ing om di e en o igin coun ies). The es ima es o bo h yea s show a e y simila anking. G anada is he p o ince wi h he lowes es ima ed e ec o p e ious ou is s, while he Balea ic Islands ha e he highes es ima ed e ec e e y yea . In his p o ince, he in l uence o p e ious ou is is e y simila o bo h yea s showing ha i s a ac i eness as a ou is des ina ion has emained s able o he las en yea s. The same happens in Malaga, which is he hi d p o ince wi h g ea es elas ici y a e he Balea ic Islands and Alme ia. The es o he p o inces show a la ge inc ease in he es ima ed elas ici y du ing he pe iod o analysis. This esul indica es ha imp o emen s ca ied ou in he Medi e anean coas s ho els ha e been app op ia e o sa is ying he g owing ou ism demand ha his a ea is expe iencing. Dependen a iable: Tij, GMM-SYS Explana o y a iables one-s ep wo-s ep Tij, -1 0.9626*** 0.9674*** Tij, -1⋅Qi, -1 0.0149*** 0.0147*** Tij, -1⋅Oi, -1 -0.0159** -0.0182** GDPj 0.0301*** 0.0294** IPij, -0.1752** -0.2148** D2009 -0.0989*** -0.0971*** Hansen es (p- alue) 0.076 0.076 AR(1) (p- alue) 0.000 0.000 AR(2) (p- alue) 0.208 0.222 Numbe o obse a ions 770 770 Numbe o g oups 77 77 Sou ce: own using he x abond2 command in STATA10 (Roodman, 2009) No e: *, **, *** deno e signi i can a he 10%, 5% and 1% le el espec i ely. Tab. 1: Es ima ion esul s o in e na ional ou ism demand model, 2005-2015 EM_4_2018.indd 72EM_4_2018.indd 72 28.11.2018 13:12:5328.11.2018 13:12:53 73 4, XXI, 2018 Economics Fig. 4 shows he sca e plo and eg ession line be ween es ima ed elas ici y o he lagged dependen a iable and pe cen age o luxu y ho els in 2015, and Fig. 5 be ween he es ima ed elas ici y and ou is s-ho el beds a io. Bo h i gu es g aphically illus a e why he e ec o p e ious ou is s is g ea e o less in each p o ince. As expec ed, Fig. 4 shows a posi i e ela ionship and Fig. 5 a nega i e one. On a e age, he highe he pe cen age o luxu y ho els he g ea e he posi i e e ec o he p e ious ou is s on he cu en demand (see Fig. 4). The Balea ic Islands a e an example. Howe e , he e a e p o inces mo ing away om his a e age beha io . Alme ia has a high es ima ed elas ici y in ela ion o i s pe cen age o luxu y ho els. On he o he hand, Ba celona has a low es ima ed elas ici y in ela ion o i s ho el quali y, p obably due o ou is conges ion. Fig. 5 indica es ha he lowe he ou is conges ion, he mo e bene i cial he in l uence o p e ious ou is s on he cu en demand. This is he eason why Alme ia has a high elas ici y. I is he p o ince wi h he lowes ou is s-ho el beds a io. Balea ic Islands and Ba celona mo ing away om his a e age beha io showing a la ge elas ici y, because hey a e he p o inces ha o e he highes ho el quali y. In con as , Cas ellon, due o i s low pe cen age o luxu y ho els, has a lowe es ima ed elas ici y han ha co esponding o i s le el o ou is conges ion. Finally, looking a Fig. 4 and 5, i is easy o unde s and why he Balea ic Islands and G anada a e he p o inces wi h g ea es and lowes elas ici y, espec i ely. The Balea ic Islands ha e he highes quali y ho els, as well as he second lowes a io be ween ou is s and ho el beds. In con as , G anada (along wi h Cas ellon) has he lowes pe cen age o high-quali y ho els, and he g ea es ou is conges ion (along wi h Ba celona). To sum up, esul s om he GMM es ima es imply ha p e ious ou is s ha e played an ac i e ole in he g ow h o ou ism in he Spanish Medi e anean coas s. The signi i can posi i e e ec o he p e ious ou is s e eals Fig. 3: Es ima ed elas ici y o ou ism demand wi h espec o p e ious ou is s Sou ce: own EM_4_2018.indd 73EM_4_2018.indd 73 28.11.2018 13:12:5328.11.2018 13:12:53