A study on Risk Evaluation of Revenue Forecasting Accuracy in Tax Administration in Malawi; A Time Series Approach

摘要

Nobody can accurately predict amount of sales revenue expected. There is need for accuracy to eliminate risk of revenue loss. Surveys have shown that accuracy is the most important criterion in selecting a forecasting strategy. Which criterion then provides the most accurate forecast? Human judgment and software have been used to address this issue but the question is how do we judge the accuracy? Is there revenue leakage due to over or under forecasting, resulting from lack of accuracy on the methods used? This paper evaluates accuracy of revenue forecasted over eleven years in a semi-autonomous Tax administration, a revenue collection body which collects taxes on behalf of government in Malawi. We investigate whether forecasting method which was used was the most accurate, unbiased and efficient, and whether better forecasts could have been used. We compare different Mean Absolute Deviation (MAD) and Mean Square Error (MSE) calculated from different time series methods using software CB predictor to establish the most accurate method for forecasting revenue collection for this organization. Actual data provided by Ministry of Finance in Malawi was used. The study shows that the forecasting method used by the organization was not most accurate than using the Last Value Forecasting Method which had the lowest MAD among the other forecasting time series method that were tested. The current method used by the organization had huge forecasting errors. The paper further recommends the use of Last Value Forecasting Method, and suggests other forecasting objectives the organization's may have other than accuracy.