Relevance of Input Data Time Series for Tax Revenue Forecasting

The paper deals with the question whether for the tax is advantageous to use the longest possible time series. This issue arose from two opposing general econometric claims. The first general contention is that the longer the time series, the better and more accurate results. The second claim requires that the estimated system to be not unnecessarily volatile. To test this issue was chosen long time series of selected tax revenues of Denmark. This time series were artificially reduced and the quality of the forecast estimated by regression analysis was tested using in terms of predictive power. The result is that for the same explanatory variables for a longer or shorter time series estimations based on shorter time series are slightly accurate.