Forecasting unemployment in Finland: A flow approach

In this paper we evaluate whether the accuracy of Finnish unemployment rate forecasts can be improved by utilising the information in the flows into and out of unemployment. We compare and contrast different methodologies for constructing the flows. Our results indicate that Bayesian vector autoregressive models improve forecasts over a simple autoregressive model. Labour market flows improve forecasts over very short forecasting horizons. Additional labour market variables can improve forecast accuracy. The time-series models struggle to improve upon professional forecasts, but a combination of these forecasts proves advantageous especially when forecasting two quarters ahead.