A real time evaluation of national government real GDP growth forecasts : an application for Malta

Abstract

The aim of this dissertation is to evaluate the accuracy of Malta’s national government real GDP growth forecasts through absolute and relative accuracy assessments of the MFEI macroeconometric model-based real GDP growth forecasts, presented annually in the SCPs. In line with the theoretical framework underpinning the literature in the field of forecast accuracy, this dissertation analyses the absolute accuracy in forecasts through OLS regression-based tests for unbiasedness and efficiency. The relative accuracy in forecasts is assessed through the adoption of benchmark forecasts including forecasts produced by the EC and forecasts produced by a three-variable VAR(1) model. The results suggest that MFEI forecasts of real GDP growth do not display statistically significant bias for all forecast horizons considered. Furthermore, current-year and three-yearahead forecasts manage to fully incorporate both the information contained in the past forecast errors and the data set used in forecasts. However, a mere indication of weak inefficiency is apparent in one-year-ahead and two-year-ahead forecasts. Moreover, varied findings are obtained when using benchmark forecasts to assess the relative accuracy of MFEI real GDP growth forecasts. The MFEI forecasts manage to outperform the EC forecasts for the one-year-ahead forecast horizon but are found inferior to EC forecasts for the current-year forecast horizon. Nevertheless, the difference in accuracy is rather small. In fact, for all forecast horizons the MDM test results point towards the equality of predictive power for both sets of forecasts. The three-variable VAR(1) model outperforms the MFEI forecasts for all the forecast horizons considered, although the relative difference is not large particularly at the three-year-ahead forecast horizon.