Survey evidence on forecast accuracy of U.S. term spreads

Successful portfolio management strategies partly require accurate forecasts of term spreads. Such forecasts may also be useful for policymaking since the yield curve may contain predictive information for economic growth. This study asks whether experts accurately predict term spreads. We show that the consensus forecasts from two separate panels, while superior to alternative benchmark forecasts, are free of systematic bias but unable to replicate the degree of variability in the actual change. Moreover, these forecasts are directionally accurate under symmetric loss, implying that they are of value to a market participant who assigns similar costs to incorrect upward and downward moves.