Inflation forecasts, the expected real rate and information efficiency

This paper examines the inflation forecast accuracy and rationality of a time series predictor, an expected inflation series constructed from surveys undertaken by the University of Michigan's Institute for Social Research, and the yield on U.S. Treasury bills. A quadratic loss function is assumed, and Theil's inequality ratios are used to decompose the mean squared foreast error (MSE). Although it is the only unbiased forecast, the time series predictor does not appear to be optimal during the recent period of turbulent inflation. As for bias in the other series, the surveys are inconsistent with weak-form rational expectations and the Treasury bill is reflecting variations in more than one factor. In an examination of the latter under the assumption of no liquidity premium, the results suggest that the forecast error is in part attribute to moderate variation in the expected real rate. Given the minimal cost of observing the Treasury bill yield, the findings of this study suggest that the market, although imperfect, has performed reasonably well in forecasting inflation.