摘要：The aim of the article is to discuss and demonstrate the problem of nonlinearity in tourism trips in the European Union. Tourism is a growing branch of economy but it is very sensitive for any instabilities of economic and non-economic nature. These can cause a nonlinear structure of the time series which has serious consequences for econometric modelling. An irregular course of the process over time may not be a sufficient reason to use the simplest linear models. Therefore,this paper applies a series of tests and models to describe phenomena with non-linear or partly linear structures. The main class of the analysed econometric models is threshold models. For the purpose of identification of non-linear character of dependency for tourist departures and arrivals in the European Union,the following tests were used: Enders and Siklos test,a modification of Kapetanio set al. test,a newly proposed test for threshold cointegration,Tsay test and Hansen and Seo test. After testing for nonlinearity the forecasts of tourist departures were computed. As the results of testing indicated a stationary threshold structure or a linear cointegration two models were used in prediction,i.e. a stationary threshold model and a linear error correction model (ECM).A comparison of forecasting errors indicates that all forecasts are accepted,but ECM gave the smallest values of errors. The forecast can be utilized in economic practice by tourist industry.
关键词：Threshold cointegration;econometric model;forecasting;tourism industry
