Measuring Disagreement in Qualitative Expectations

We assess how well measures of disagreement in qualitative survey expectations reflect disagreement in corresponding quantitative expectations. We consider a variety of measures, belonging to two categories: measures of dispersion in nominal and ordinal variables and measures based on the probability approach of Carlson and Parkin (Economica, 1975; 42, 123-138). Using data from two household surveys that collect both qualitative and quantitative inflation expectations, we find that the probability approaches with time-varying categorization thresholds and either a piecewise uniform or t distribution perform best and the resulting disagreement estimates are highly correlated with the benchmark. Copyright (c) 2015 John Wiley & Sons, Ltd.