Assessing the Rationality of Survey Expectations: The Probability Approach

This paper considers popular methods for the quantification of survey expectations. We investigate the asymptotic properties of two variants of the probability approach originally suggested by Carlson and Parkin (1975). It is argued that the traditional method can be interpreted as an instrumental variable estimator that suffers from a severe loss of efficiency if the mean of the target variable is close to zero. The regression approach advocated by Pesaran (1984) can be seen as a special case assuming that expectations of survey participants are uniformly distributed. Applying alternative variants of these approaches to the ZEW survey of expected changes in the inflation rate demonstrates that the regression variant of the Carlson-Parkin methodology yields the best fit to the realized changes of the inflation rate. It turns out, however, that even the best performing quantification method fails to fulfill the requirement of rational expectations, as the variance of the expectation error is significantly reduced by including additional variables such as the lagged inflation rate.