The non stationarity in seasonal series. Application to Tunisian tourism[La non stationnarité dans les séries saisonnières. Application au tourisme tunisien]

The purpose of the paper is to specify the nature of the seasonal behaviors of tourist stay demand in Tunisia by using recent tools of monthly time series analysis. The available series (tourist expenditures, prices, guest nights, reception capacities,...) are analyzed with the classical theory of demand and the theory of supply induced demand. Empirical results show that these series are generated by non stationary processes which seasonality is stochastic and deterministic. The extent of the maximum likelihood method proposed by Lee (1992) for monthly data emphasizes unit roots and cointegration relationships at some seasonal frequencies. Error correction models are derived for endogenous variables. The use of unadjusted series in logarithm leads to direct estimation of short and long run elasticities ; they provide a particularly interesting description of tourist behaviors in Tunisia and about the response possibility of supply by professionals. They suggest some interpretations in relation to the origin of common sources of seasonality that is probably due to climatic and institutional or economic factors.