Co-integration analysis of quarterly European tourism demand in Tunisia

Abstract

The purpose of this study is to identify the factors that affect the destination choice process. In addition to prices and income factors, the supply factor is introduced as an explanatory variable in the econometric model. Co-integration analysis and error correction models (ECMs) are used to estimate the long run tourism demand elasticities and to forecast the quarterly European tourism demand for a 1-year-ahead horizon. The main finding of this study is that the behaviour of European tourists varies from one country to another. The co-integrating relationships show that the large elasticity magnitude may be the reflection of the relatively expensive services often sought after by tourists from these countries. The estimated values of the supply elasticity corroborate the supply induced demand hypothesis. Finally, compared to the basic structural model and using the root mean squared error, the ECM provides more precise forecasts.

Introduction

Tourism plays an important role in Tunisia's economic development because of its contribution towards balancing the commercial deficit and to cutting unemployment. Since 1986, the tourism industry has become the second largest foreign currency earner after the textile industry. This sector shows a strong seasonal fluctuation that results in a concentration of the demand during certain months of the year, particularly in July, August and September. It is essentially the seaside character of Tunisian tourism that attracts people during the high season. The under-usage of these tourist sites during the off-season has a negative impact on the financial performances of this sector.

The purpose of this paper is to identify the factors that affect the destination choice process. The cost of tourism and the income level are the tourist demand determinants (Lim, 1999). However, in a difficult international conjuncture, and with increasingly aggressive competition, tourism operators may need attractions other than seaside to bring in more tourists, particularly off-season. To target different types of tourists they have developed other kinds of products and services, such as cultural heritage, and/or tourism for other motives like business, health programs or sports. It would be appropriate to consider using the supply factor to explain the tourism demand. In this study, we will give an empirical justification of the supply induced demand hypothesis.

Seasonally non-adjusted quarterly data of European tourist arrivals are used from 1981 to 2005. It is known that economic data is often non-stationary, in particular, high frequency data. Tourism is a seasonal activity and as such data may exhibit non-stationary trends and seasonality; as a result, the traditional least square regression approach will lead to erroneous results (Franses, Hylleberg, & Lee, 1995; Granger & Newbold, 1974). If we consider the seasonal and non-seasonal stationarity of the series, cointegration analysis is a suitable strategy to model and forecast the tourism demand for the following reasons. First, using the difference filters (i.e. seasonal and non-seasonal) proposed by Box and Jenkins to achieve stationarity leads to a loss of information about the long term relationships between non-stationary economic series (Box & Jenkins, 1976). The error correction models (ECMs) provide a way to avoid this problem because the long run information lost due to differentiating is reinstated in the ECMs. Second, since the existence of unit-roots tests of Hylleberg, Engle, Granger, and Yoo (1990) (HEGY), the empirical studies using this procedure show, in most cases, that the hypothesis of unit root is rejected at least at one frequency, so the systematic application of the seasonal difference may generate an overdifferentiation problem (Bell, 1987; Hylleberg, Engle, Granger, & Yoo, 1990). Third, the ECMs provide a way to combine both dynamics of short run and long run adjustment processes simultaneously (Dritsakis, 2004; Dritsakis & Papanatasious, 1998; Gonzalesz & Moral, 1995; Kim & Song, 1998; Kulendran & Witt, 2001).

In this study, the Johansen (1988) and Johansen and Juselius (1990) methods are used to estimate long run tourism demand elasticities. ECMs are then estimated and used to forecast the quarterly TArr from the 4 most important European countries and compute the estimated statistics. After having shown the place of European tourism in the Tunisian economy and its evolution during the last three decades, Section 2 emphasizes the relevant factors that influence the destination choice process. The theoretical framework of analysing the tourism demand will then be discussed. The empirical results of the cointegration analysis and the commentaries will be carefully dissected in the third section. The ECMs are used in the final section to forecast the quarterly European TArr in Tunisia. Their forecasting performances are compared to those of the basic structural model (BSM) using the root mean squared error (RMSE).