How good are banks’ forecasts?

Abstract

We analyse the financial forecasts small and medium-sized German banks provided in several waves of a quantitative survey, called LIRES, and compare them with the results the banks actually realized. Based on this unique data set, we find that the predictions are relevant, especially concerning the net interest income for the next year, and persistent, but neither unbiased nor rational. We also find slight evidence for a positive relationship between planning and performance, i.e. banks whose predictions are more accurate tend to have a higher return on assets. Looking at the forecasts made just before the end of the low-interest rate environment, we observe that the explanatory power of predictions went down.