This paper compares the out-of-sample forecasting accuracy of time series models using the Root Mean Square, Mean Absolute and Mean Absolute Percent Errors. We evaluate the performance of the competing models covering the period January 1971 to December 2002. The forecasting sample (January 1996 – December 2002) is divided into four sub-periods. First, for total forecasting sample, we find that MA(4)-ARCH(1) provides superior forecasts of unemployment rate. On the other hand, two forecasting samples show that the MA(4) model performs well, while both MA(1) and AR(4) prove to be the best forecasting models for the other two forecasting periods. The empirical evidence derived from our investigation suggests a close relationship between forecasting theory and labour market conditions. Our findings bring forecasting methods nearer to the realities of UK labour market.
|Number of pages||16|
|Journal||International Journal of Applied Econometrics and Quantitative Studies|
|Publication status||Published - 2005|