Determining the most appropriate demand forecasting method for tourism policy and planing

m ABSTRACT In order to reach the objectives determined by the tourism policies, effective planning studies are required. As a result of developments and increased competition seen in tourism industry in the recent years throughout the world, planning has become one of the most important elements for national and international management of tourism industry. To succeed in achieving the correct results of the decisions related to future depends on true forecasting. In the study carried out by considering the importance of planning and forecasting for the tourism sector, the processes of tourism policies and planning were analyzed and the need for demand forecasts was brought to light. Information derived from literature related to the methods of forecast and the processes of application that can be used to produce demand forecasts was reviewed. The forecasting methods taken into consideration are methods called as "univariate methods". The methods conducted for analysis are classical decomposition, exponential smoothing and Box- Jenkins methods to forecast the short run, and trend analysis method to forecast the long run. The univariate methods are first used to determine the most appropriate data set, and then to generate forecasts for Turkish tourism demand. As a result of the analysis, Box- Jenkins method is the best forecasting method for the given series in the short run. At the end of the analysis carried out, it was concluded that Box- Jenkins method is the best forecasting method for the given series in the short run. Also, it was seen that the trend analysis method used to forecast in long run is not enough when employed alone. But it was determined that the trend analysis is capable of producing meaningful forecasts if it is regulated by including the data set of short run obtained by Box- Jenkins method. The variations of results compared by the realizations for the first six months of 2004 are 8% for 1985-2003 data set and 4 % for 1995-2003 data set. This shows that, using the most appropriate data set reduces the variations of forecasts 4 %. Trend analysis method, used to generate medium term forecasts is not adequate by itself even in the short run, because it does not give realistic forecasts according to the last years' increasing trend. By including the forecasts of Box- Jenkins method for 2004-2005 to the data set, the results of trend analysis become more harmonious with the last years' trend and become more realistic.