Improving Short-Term Forecasting of Macedonian GDP: Comparing the Factor Model with the Macroeconomic Structural Equation Model

This paper evaluates two different models for short-term forecasting of the Macedonian GDP : (a) the medium-scale static factor model, based on the static principal components analysis, and (b) the small-scale macroeconomic structural equation model. Recursive dynamic pseudo out-of-sample forecasts, based on a panel of quarterly time series, indicate that forecast errors of the factor model are smaller overall in comparison to errors of the structural equation model at all forecast horizons. In line with the existing short-term GDP forecasting practice, our medium-scale factor model (that extracts common factors from a data set of 52 variables) diversifies and strengthens the current macroeconomic forecasting strategy in Macedonia