A multivariate analysis of forecast disagreement : confronting models of disagreement with SPF data

This paper documents multivariate forecast disagreement among professional forecasters of the Euro area economy and discusses implications for models of heterogeneous expectation formation. Disagreement varies over time and is strongly counter-cyclical. Disagreement is positively correlated with general (economic) uncertainty. Aggregate supply shocks drive disagreement about the long-run state of the economy while aggregate demand shocks have an impact on the level of disagreement about the short-run outlook for the economy. Forecasters disagree about the structure of the economy and the degree to which individual forecasters disagree with the average forecast tends to persist over time. This suggests that models of heterogeneous expectation formation, which are currently not able to generate those last two features, need to be modified. Introducing learning mechanisms and heterogeneous signal-to-noise ratios could reconcile the benchmark model for disagreement with the observed facts.