Forecasting GDP Growth of Brunei Darussalam Using Factor Models

Abstract

This paper evaluates the relative performance of factor models in forecasting GDP growth using a large quarterly panel dataset compiled for the Brunei economy. The common factors are extracted through the estimation of both static and dynamic principal components, and are used to compute pseudo out-of-sample forecasts in a recursive scheme. These factor-based forecasts are then compared to a standard benchmark univariate autoregressive model. The forecasting results show that the forecast errors of the benchmark model increase with the prediction horizon but the forecast errors of factor models remain relatively unchanged. In spite of poorer forecasting performance in one-and two-quarter ahead forecasts, factor models significantly outperform the benchmark in three-and four-quarter ahead forecasts. This implies that the information conveyed by the large dataset provides predictive power at longer horizons, illustrating the usefulness of factor models as a macroeconomic forecasting tool for Brunei.