Improving Predictive Ability of Macroeconometric Models Using Coefficient Averaging
KGEMM (KAPSARC Global Energy Macroeconometric Model) is a policy analysis tool that was designed to evaluate the impact of domestic policy decisions and global shocks on economic-energy-environmental indicators in Saudi Arabia. It also provides projections for energy-environmental-economic interactions at both national and sectoral levels. Having a desired level of predictive accuracy is crucial for macroeconomic models to effectively inform policy decision-making processes. This work develops a methodological framework employing the Model Averaging approach to enhance KGEMM’s predictive ability. The framework includes the consideration of theoretically consistent estimates from single equation cointegration methods and Vector Autoregressive models for averaging after a set of prediction accuracy metrics widely used in empirical studies – namely, Mean Squared Forecast Error, Forecast Error Bias test, and Forecast Encompassing test – is performed. In this work, the methodological framework is applied to representations of the petrochemical sector and sectoral employment in KGEMM. However, it can be extended to any other representations in the model or other macroeconometric models.
22nd May 2025