In this paper was verified the possibility of improving the monthly forecasts of the Value Added Tax on Merchandise and Services (ICMS) collected by the State of Santa Catarina, Brazil. Dynamic regression will be used based on the concepts of cointegration and error correction utilizing the general to specific approach suggested by the London School of Economics (LSE). Different data series were selected and analyzed for the final model industry profit, consumption of electric energy and other energy sources, and cement, and business consultations to the Credit Service Protection Agency (SPC). In the process of the choice of the variables, Granger’s tests of causality and the analysis of long-run equations were used. The results obtained were very satisfactory for forecasts both inside and outside the sample period, indicating that the use of this model by the Budget Department of the State of Santa Catarina will provide more suitable values for the decision making process and improvement in budget planning.