Modelos de previsão para arrecadação do IRPS na Província do Niassa

Abstract

The forecast of the personal income tax - irps is fundamental for the planning of government activities, as this is the second main tax of the Mozambican state. This work develops two forecast models for the collection of irps in the Cuamba and Lichinga fiscal area directions whose sum corresponds to the irps of the Niassa province. For data analysis, integrated autoregressive moving average models - arima based on the box-jenkins methodology and simple exponential holt and holt-winters adjustment models were used. The results show that the IRPS collection series in the Cuamba fiscal area direction did not comply with the assumptions for the application of the exponential adjustment methods. And the model that best fit the series was 4fl/m4(l,1,27). While the series of IRPS collection in the Lichinga fiscal area was adequate for the application of the box-jenkins and holt exponential adjustment methods and the arima(l,l,2l) and arima models were estimated. holt(0.4,0.02) respectively. Comparing the forecast performance statistics, it is concluded that the arima(l,1,2l) model is the most adequate and the boxjenkins method produced better results in relation to the exponential adjustment method. Finally, its application is recommended to forecast the irps in the remaining provinces of Mozambique. Keywords: irps, arima models, exponential adjustment, box-jenkins, holt and holt-winters