Non-Linearities, Large Forecasters And Evidential Reasoning Under Rational Expectations

Rational expectations is typically taken to mean that, conditional on the information set and the relevant economic theory, the expectation formed by an economic agent should be equal to its mathematical expectation. This is correct only when actual inflation is “linear” in the aggregate inflationary expectation or if it is non-linear then forecasters are “small” and use “causal reasoning”. We show that if actual in- flation is non-linear in expected inflation and (1) there are “large” forecasters, or, (2) small/ large forecasters who use “evidential reasoning”, then the optimal forecast does not equal the mathematical expectation of the variable being forecast. We also show that when actual inflation is non-linear in aggregate inflation there might be no solution if one identifies rational expectations with equating the expectations to the mathematical average, while there is a solution using the “correct” forecasting rule under rational expectations. Furthermore, results suggest that published forecasts of inflation may be systematically different from the statistical averages of actual inflation and output, on average, need not equal the natural rate. The paper has fundamental implications for macroeconomic forecasting and policy, testing the assumptions and implications of market efficiency and for rational expectations in general.