Using point forecasts to anchor probabilistic survey scales

We present the results of an experiment where a random subset of the participants in the Bundesbank's household panel receive personalized response scales, centered at each participant's point forecast. Personalized response scales offer two advantages over the standard scale which is centered at zero inflation: First, they mitigate the impact of the central tendency bias which leads respondents to assign greater probability mass to the center of the scale at zero. Second, they eliminate the need to adjust the scale when actual inflation falls outside the range for which the response scale was designed. Our results show that the personalized survey responses are of higher quality in three dimensions: (i) higher internal consistency, (ii) more uni-modal responses, and (iii) a signiftcant reduction in the use of the (minimally informative) unbounded intervals of the response scale.