A statistical evaluation of GAP's forecasting performance for the Albanian economy

This paper aims at offering a statistical evaluation methodology on the forecasting performance of the GAP model, a semi-structural economic model used to support monetary policy decisions at the Bank of Albania since 2011. In this paper we evaluate the forecasts produced purely by the model, and not those used by the Monetary Policy Department, which also include the expert judgment and are not made public. The analytical approach used in the discussion material combines a statistical diagnostic look-up consisting in statistical measurements as RMSE and BIAS important to understand the forecasting performance of the model as an instrument in one, two and three years ahead time horizons. A VAR model is constructed resembling the economic relations represented in the GAP model as a most commonly used tool to obtain economic projections based solely on the information that the data series provide. Comparing the forecasting performance of the two models on a common statistical diagnostic metrics helps us to create a broader understanding of the forecasting abilities of the GAP model and draw discussion issues for potential improvements of the model that would potentially lead to an improved representation of the Albanian economy and increased accuracy in its forecasting performance.