This paper explores the performance of governments’ budget forecasts. We argue that budget forecasts tend to be pessimistic if negative fiscal shocks are perceived to be more costly than positive fiscal shocks. Based on an asymmetric loss function, we show that the observed bias in budget forecasts is related to budget risks. We test these predictions with an empirical analysis of the German states’ forecasts of the budget-balance over a period of fourty years. The results confirm the existence of a robust downward bias for the short-term horizon, which disappears over medium forecast horizons. While we find that the bias is significantly associated with a measure of budget risks, the strong pessimism in short-term forecasts is attributed to larger adjustment cost in case of adverse fiscal shocks.

The classic assignment of functions of government entrusts the central government with the stabilization function. Nevertheless, subnational governments are intimately involved in stabilization policies of their own, often restricted by various rules or balanced budget constraints. This paper reviews this literature and follows and expands on a recent macroeconomic strand which estimates “regional multipliers.” I estimate separately regional multipliers for grants, rainy day funds, and a state’s deficit as well as spillover effects from these revenue sources.

The paper evaluates the forecasting performance of official estimates of the revenue effects of tax legislation, in the absence of dynamic scoring. The analysis is based on a detailed account of 246 German tax reforms, which result in 800 overall changes to individual taxes, at federal level over forty years covering the time period from 1982 until 2022. The preliminary results show that the revenue effects of tax reforms are systematically overestimated. If an increase (decrease) in revenues of 1% is predicted, actual revenues only increase (decreases) by around 0.73%. The overestimation is particularly evident in the case of income taxes, whereas changes in general sales taxes are estimated much more reliably - here the actual revenues vary proportionally with the revenue estimates.