The formation of experts' expectations on labour markets: Do they run with the pack?

Expectations regarding the economic development might be correlated due to various reasons: because individuals use the same public information and similar evaluation methods, and because of social learning or herding amongst peers. We analyse to what extent expectations are driven by herd behaviour, and if it contributes to make expectations more realistic. In a novel survey the CEOs of the local departments of the German Federal Employment Agency report their expectations on unemployment in the short run. In this data we can discriminate between close and less-close peers to overcome the reflection problem and to quantitatively assess answers regarding the initial questions. We find strong evidence for herding in expectation formation. The size of effect is robust across various specifications and remains even when controlling for forecasts from experts external to the survey. The social multiplier approximately doubles the effect of information (signals) included in the model. Compared to counterfactual expectations without herding constructed from the estimates, herding seems to improve the accuracy of the expectations.