This article aims to estimate the dynamic factor model for prediction tax receipts in Brazil using monthly data for the period 2001-2013. The factorial model allows to reduce the dimensionality of the high number of taxes taking into account the information contained in the existing interrelations between them and allowing to identify only the relevant information through the variables named factors. Further, in our model the seasonal component of the series of taxes is treated endogenously. This procedure permits to obtain better data fitting and more reliable predictions - once seasonality is a hallmark of certain series of tributes. We confront the predictions of the factorial model with those generated by linear dynamic model applied to each tribute separately and found that the factor model brings considerable gains in terms of efficiency and prediction.
gross tax burden seasonality dynamic factor model Kalman filter and Gibbs sampling