Determining macroeconomic indicators to implement a short-term forecasting model for VAT revenue

Resumen

Macroeconomic indicators are a good source of information for short-term forecasting due to several reasons: they cover different areas of the economy and provide faster modes of dissemination. In this study we use a set of indicators to obtain a valid forecast for VAT revenue using a blend of statistical methods such as transfer functions and principal component analysis. The objective is to enforce parsimony and avoid multicollinearity problems with minimum information loss. In a previous paper we combine several macroeconomic indicators obtaining good predictive performance in terms of prediction error and RMSE for VAT revenue. In this study we add some leading indicators related to consumption for maximizing prediction accuracy. The evaluations involved splitting the data into a training sample and a validation sample using one-step ahead forecast during the validation period and comparing results with the observed values of VAT revenue. The results in the validation sample account for the improvement in model prediction. We apply the proposed method to quarterly data beginning in 1995 and ending in the last quarter of 2017.