Revisiting the convergence hypothesis for tourism markets: Evidence from Turkey using the pairwise approach Andrew Abbott a, 1 , Glauco De Vita b, 2 , Levent Altinay c, * a Business School, University of Hull, Cottingham Road, Hull, Hull HU6 7RX, UK b Business School, Oxford Brookes University, Business School, Wheatley Campus, Oxford OX33 1HX, UK c Business School, Oxford Brookes University, Headington Campus, Oxford OX3 0BP, UK article info Article history: Received 31 October 2010 Accepted 5 June 2011 Keywords: Tourism markets Convergence hypothesis Pairwise approach Turkey abstract This paper empirically revisits the tourism markets convergence hypothesis (Narayan, 2006) for the case of Turkey by employing the newly developed pairwise approach to the analysis of stochastic convergence (Pesaran, 2007; Pesaran, Smith, Yamagata, & Hvozdyk, 2009). This new approach allows for unit root tests to be conducted on all possible pairs of tourist arrival differentials across Turkeys 20 major tourist source markets, thus avoiding the need to choose a base source market or total tourist arrivals gure as the benchmark. Using monthly data over the period January 1996 to December 2009, the main nding is that despite considerable co-movementof international tourist arrivals in Turkey, there is no evidence of long-run convergenceamong Turkeys major tourism markets. Cluster-based club convergence analyses alongside bivariate pairwise estimations conrm the lack of convergence and highlight specic tourist source markets with considerable untapped potential. These ndings have profound policy implications for Turkeys inbound tourism planning and promotion. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction A relatively new and still under researched debate in the eld of tourism management concerns the convergence hypothesis for tourism markets. Originally developed by Narayan (2006), the hypothesis states that the effectiveness of policies aimed at attracting visitor arrivals from any one tourist source market can be appraised by investigating the convergence properties of total visitor arrivals in a country and visitor arrivals from any tourist source market of that country. In his seminal contribution, Narayan (2006) investigated whether or not Australias thirteen major tourist source markets were converging over the period January 1991 to September 2003. Using univariate and panel Lagrange multiplier unit root tests, he examined whether the difference between total visitor arrivals and visitor arrivals from each tourist source market to Australia fol- lowed a stationary process. 3 He found strong evidence of conver- gence of Australias tourism markets. Narayan (2007) draws similar conclusions for the case of Fiji, though here the convergence of tourism markets is examined by means of an extended methodological framework that entails testing for cointegration (i.e. the long-run co-movement) between total visitor arrivals to Fiji and visitor arrivals from each of the eight tourist source markets considered. Broadly consistent with the ndings above, the convergence hypothesis appears to hold also in application to Malaysias tourism markets (see Lean & Smyth, 2008), and for the case of Singapore (see Lee, 2009). However, in the latter contribution, given that approximately 70% of international tourist arrivals to Singapore are from Asia, the test for convergence hinges on establishing the stationarity of the difference between international visitor arrivals in Singapore from each specic source continent (Africa, the Americas, Europe and Oceania) and visitor arrivals in Singapore from the key source market (continent), namely Asia. 4 Finally, Lorde and Moore (2008) investigated whether tourist arrivals in the Caribbean are converging over time using a panel of 22 countries and monthly data from 1977 to 2004. Their empirical results suggest that there is no convergence in tourism penetration * Corresponding author. Tel.: þ44 (0) 1865 483832; fax: þ44 (0) 1865 483878. E-mail addresses: a.abbott@hull.ac.uk (A. Abbott), gde-vita@brookes.ac.uk (G. De Vita), laltinay@brookes.ac.uk (L. Altinay). 1 Tel.: þ44 (0) 1482 463570; fax: þ44 (0) 1482 347500. 2 Tel./fax: þ44 (0) 1865 485798. 3 In time series analysis, a stationary process is a random process whose prop- erties such as its mean or variance, do not vary over time. 4 Lee (2009) also examines convergence as a catching upprocess by testing for stationarity of tourist arrival differentials on two subsamples (from May 1993 to January 1997, and from January 2004 to September 2007) constructed around the period most likely to contain structural breaks due, for example, to the occurrences of the 9/11 terrorist attacks, the war in Afghanistan and SARS. Contents lists available at ScienceDirect Tourism Management journal homepage: www.elsevier.com/locate/tourman 0261-5177/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.tourman.2011.06.003 Tourism Management 33 (2012) 537e544