Forecasting Tax Revenues of the Protestant Church in Germany (EKD) with Time-series and Panel-data Models

In order to forecast church tax revenues of the Protestant Church in Germany (EKD) and its member churches within an 18-month horizon, seasonal ARIMA, ARIMAX and VAR-models for monthly data are set up and their forecast accuracy is evaluated using ex post out-of-sample forecasts. This approach is appealing because it allows the possibility to establish a simpler forecasting method than the existing procedures that are more expansive.

In comparison to pure ARIMA-models, forecast accuracy may be improved by additionally allowing for exogenous business-cycle indicators. Moreover, multiple reforms of German income-tax laws, regarding both the tax base and the tax rate, are supposed to have considerable influence on the development of tax revenues. Consequently, in order to prevent omitted-variable bias, these different effects have to be measured and implemented in the forecasting models.

Although general improvements through time-series models are being observed, results vary among the member churches. A better understanding of this heterogeneity can be achieved by adding regional information to German Federal States (Laender) within a panel-data setting.