Cross-Sectional Evidence on the Rationality of the Mean and Variance of Inflation Expectations

Abstract

This paper draws on the results of a specially commissioned survey of consumers in Sweden to test the hypothesis of rational information processing with respect to lay forecasts of inflation. The mean and variance of such forecasts are important inputs to the economic theory of household labour supply, savings, and portfolio choice. Unlike most empirical tests of rational expectations, cross-sectional data on individuals are employed rather than time series data on some group average forecast. This allows testing for the presence of forecasting biases due to socioeconomic factors. “Mean rationality” and “variance rationality” are also examined. That is, we test for the presence of biases both in the mean inflation forecasts and in subjective confidence about the accuracy of the forecasts.

It is found that individuals who face lower information costs, or higher benefits from information collection, do indeed produce the most accurate forecasts, so that the theory of optimal information processing is not rejected. However, there are signs of systematic bias in the mean inflation forecasts of particular socioeconomic groups. Moreover, consumers’ subjective ratings of the likely accuracy of their forecasts are systematically in error. Those individuals who consistently produce more accurate forecasts do not express a consistently higher degree of confidence in their forecasts. Hence the hypotheses of mean rationality and variance rationality are both rejected.