Transformations  in
Business & Economics

ISOLATING LONG CYCLES IN UNEMPLOYMENT RATES OF CROATIA USING SPECTRAL MODELLING
Marinko Skare, Dean Sinkovic

ABSTRACT. This paper is the first to explore unemployment nature and determinants in Croatia in a comprehensive way (88 time series / variables analysed) using nonlinear multivariate singular spectrum modelling. An important empirical finding of this study is the cyclical nature of unemployment. Unemployment in Croatia exhibits quite remarkably precise recurring business cycles of 5-6 years. This cyclical behaviour of unemployment is important both from theoretical and practical point of view. If empirical results of this paper are validated by other studies, a 'partial hysteresis' hypothesis should be recognised in terms of unemployment since both the hysteresis hypothesis and the natural rate of unemployment hypothesis hold in Croatia. Both hysteresis and natural rate of unemployment govern unemployment dynamics in Croatia causing cyclical behaviour and the observed cycles of 5-6 years. Therefore, the empirical evidence of this study calls for researching the so-called 'partial hysteresis', with unemployment having not only some properties of random walk (non-stationarity) but also of stationary series (mean reversion). Spectral forecasting techniques show to be superior to other standard time series models of unemployment .

KEYWORDS:  unemployment, persistence, causality, Croatian disease, partial hysteresis, spectral methods.

JEL classification:  C14, C38, C54, E24, E61.