Browsing by Subject "Tax estimation"

Now showing 1 - 1 of 1

Steuerschätzung und Analyse der Prognosegüte für die Bundesrepublik Deutschland
(2012) Berberich, Ulrike; Wagenhals, Gerhard
Tax revenue forecasts are the basis for the budgeting and intermediate-term financial planning of the federal government, state governments, and local authorities. In Germany, the so-called working party on tax revenue estimates (Arbeitskreis ?Steuerschätzungen?/ AKS) undertakes the task of forecasting tax revenue since 1955. The participation of many independent experts is supposed to yield political independence. However, the AKS has to account for the basic assumptions about the macroeconomic development provided by the federal government of Germany. This implies the possibility of implicit political influence and poses the risk of a systematic overestimation of the tax revenues, which would be reflected in biased forecasts conducted by the AKS. Therefore, the question arises whether the official estimations by the AKS are rational, namely unbiased and information-efficient. Previous empirical studies imply that the forecasts conducted by the AKS are unbiased and in most analyzed cases also efficient. However, all authors conducted separate tests for selected forecast horizons following the traditional approach. The study at hand provides the first investigation whether the official forecasts by the AKS are still unbiased and information-efficient when the tests for forecast rationality are conducted simultaneously for all forecast horizons according to the pooled approach of Clements, Joutz and Stekler (2007). Additionally, so-called Seasonal Autoregressive Integrated Moving Average (SARIMA) models are proposed as an alternative direct estimation technique and their forecast quality is analyzed in comparison to the forecasting power of the forecasts conducted by the AKS. Therefor, a comparative horizontal, vertical and pooled accuracy analysis of the SARIMA and AKS forecasts between 2001 and 2010 is carried out using descriptive quality measures. An analysis of the official May forecasts conducted by the AKS for the current and four following years and for the target years 1981 to 2010 shows that the official tax revenue forecasts exhibit a significant (negative) bias according to the model with the most reasonable assumptions. On the one hand, this result contradicts previous empirical studies. On the other hand, it confirms the trend for an overestimation of tax revenue by the AKS as suspected in the literature. Testing for information-efficiency implies in most cases that the forecasts conducted by the AKS are efficient, but the test results are not very robust with respect to the significance level, the sample range and the different model assumptions. However, the significant bias in the model with the most plausible assumptions already leads to the conclusion that the official forecasts by the AKS are not rational applying the pooled approach. In contrast to the forecasts conducted by the AKS, the alternative SARIMA forecasts are under all assumptions unbiased and information-efficient if only the one and two years ahead forecasts are considered. Moreover, the latter achieve a higher forecast quality than the official tax revenue estimations, which leads to the recommendation to apply direct stochastic time series models for a short term horizon up to two years. However, the three and four years ahead SARIMA forecasts perform worse than the official forecasts conducted by the AKS, which implies that the SARIMA models are not suitable for the middle-term tax revenue estimation. Furthermore, the forecasting power of the direct stochastic time series models diminishes seriously in comparison to the forecasts conducted by the AKS when an exogenous shock like the financial crisis occurs at the end of the estimation period.