Rational Forecasts from Non-Rational Models

determined schemes are considered because of their overwelming popularity. An unbiased and optimal (in the sense of minimum forecast error variance) extrapolative predictor is also described and used in the analysis because it provides a useful benchmark as the 'best' extrapolative proxy available. Section 4 examines the implications for estimation of using these three extrapolative proxies in the context of a simple two equation macro model due to Wallis (1980) and section 5 extends the model to include dynamics. The simple two equation model used bears a very close resemblance to the two main equations of the condensed St. Louis model described by Anderson and this, with its linearity and simplicity make it an ideal structure in which to house the analysis. Section 6 gives some numerical comparisions of multiplier error using the method under feasible values for the parameters.