Measuring inflation expectations: a survey data approach

Abstract

Measures of expected inflation from consumer surveys are derived using a modification of the Carlson-Parkin probability approach, which does not assume unbiasedness of expectations, makes use of survey data on expected future as well as perceived past price developments and allows for time varying response thresholds. We apply this method to the normal, central-t and noncentral-t distributions, thereby investigating the effects of nonnormal peakedness and asymmetry. We find that the effects on expected inflation of the former are small and of the latter are substantial, without increasing the accuracy of the expectations, however. Expected and actual inflation show substantial persistence, and, for most of our expectations measures, are cointegrated. Furthermore, the forecast error is stationary, implying weak-form rationality. The co-movement of currently observed expected inflation measures and the unobserved 12-months-ahead inflation rate is of interest for policy makers, for example in the direct inflation targeting strategy. Notwithstanding this, caution is warranted in using them as information variables because the inflation expected by consumers is no causal determinant of actual future inflation.