engleski

The Analysis and Forecasting of Unemployment in Croatia

Many factors influence trends in unemployment in Croatia. After proclaiming independence in 1991 Croatia had to battle the war for independence with devastating consequences particularly on unemployment which soared. During and after the war government decided to privatize social-owned companies which on the other side pushed once again unemployment upwards since in socialist economy productivity was much lower than in capitalist one. Foreign and domestic private owners of privatized companies tried to minimize the cost of labor force and they dismissed many workers. The consequence of introducing market-oriented economy in Croatia has been ever increasing percentage of women among unemployed persons which is now 61.6 % (in 1991 it was 52.2%), compared with 51.5% in the total population. Also, the number of unemployment benefit recipients has fallen dramatically because welfare state in market-oriented economy is not important as in socialist one. The educational background of unemployed persons has been also changed due to huge turmoil in Croatian economy. When analyzing trends and forecasting the unemployment in Croatia we must take into account its seasonal character because Croatian economy relies very much on tourism and related industries with very strong seasonal variations. This could be helpful for unemployed persons to find a job for several months per year but it is not good for economy as a whole to rely too much on seasonal industries. Also, calendar variations are also very important. We tried several methods for seasonal adjustment and forecasting time series in order to find the most suitable one. Traditional methods of X-11-ARIMA from Statistics Canada and X-12-ARIMA from Census Bureau were confronted to model-based method TRAMO/SEATS. The newest method called X-13-ARIMA-SEATS tried to overcome some problems of both methods and to integrate it. Also, we used structural time series modeling by Harvey and Durbin to see different approach to modeling of time series. TRAMO/SEATS proved to be better when analyzing particular Croatian unemployment series compared with very popular empirically based methods such as X-11-ARIMA and X-12-ARIMA. Structural time series model could not be compared directly with traditional seasonal adjustment methods because of different assumptions, but gave very good results in adjustment and forecasting.

unemployment; seasonal variations; seasonal adjustment; trend

nije evidentirano