Forecasting the enterprise tax base through regression of one-dimensional time series

Abstract

© 2019, Research Trend. All rights reserved. The authors obtained a forecast of the enterprise tax base. In this paper, the authors evaluated ARIMA models according to the Box-Jenkins method and regression models with dummy variables to account for additive and multiplicative seasonality. On a sample of 48 observations on the tax basis of the estimated model ARMA (1;0), ARMA (1;1), SARMA (1, 1) x (0,1) 6, a model with seasonal dummy variables. The authors focused on the method of selecting the most valid model according to the criteria RMSE and AIC used the method of selection of the designated circle of the most simple model with the fewest parameters. The reliability of the results is confirmed by the information criterion of Akaike, the mean square error of the forecast, the diagnosis of residues on the normal distribution using a special test and the absence of autocorrelation using the Ljung-Box test. The statistical significance of regression models with dummy variables for seasonality was not confirmed. A promising direction of development of this study can be a combination of forecasts, as well as the use of polynomial trends.