Abstract:
Central bank surveys frequently elicit households' probabilistic beliefs about future inflation. However, most household surveys use a response scale that is tailored towards low-inflation regimes. Using data from a randomized controlled trial included in the Bundesbank Online Panel Households, we show (i) that keeping the original scale in high-inflation regimes distorts estimates of histogram moments and forces households to provide probabilistic expectations that are inconsistent with the point forecasts and (ii) how shifting the scale improves the consistency of predictions by allowing respondents to state more detailed beliefs about higher inflation ranges. We also explore potential disadvantages of adjusting the response scale.