Abstract
The forecasting performance of widely accessible surveys of expected inflation are evaluated against naive and financial market benchmark forecasts. In the period of rising inflation (1960-80), the Michigan household consensus forecasts exhibited smaller errors than the Livingston Survey of economists and the benchmark forecasts. Unlike Livingston, the Michigan forecasts were unbiased. Since 1980, Livingston forecasts have been more accurate than Michigan, though unbiasedness is rejected in both cases. SPF forecasts, available since 1981, have been slightly superior to Livingston. In forecasting inflation, respondents generally failed to take account of cyclical conditions, and strong-form rationality is not supported.