Arrecadação tributária: modelos de previsão para o estado do Ceará

Abstract

This work uses the methodology of Box and Jenkins to perform monthly forecasts of tax collection of ICMS and IPVA for Ceará State. Modeling the presence of non-stationarity, lags in the autoregressive and moving average components, as well as seasonality, the information criteria indicated two SARIMA forecast models for each tribute. The residual analysis indicated the absence of serial autocorrelation in any of the estimated models, confirming a hypothesis that they behave as white noise processes. The exercise of forecasting derived for the period from January to December of 2016 indicated a good accuracy of forecasts to ICMS and an irregular predictive behavior regarding IPVA.