“…For example, if price forecasts are observed for one time unit ahead then one can estimate the following equation (letting P denote the expectation held at time t of P, the price to prevail at time t): (1) for each trader and test the joint hypothesis that (a0, a1) = (0, 1) and e is a zero mean white noise error. This procedure seems to have been first suggested by Theil (1966) and has since been used by Friedman (1980), Bailey et al (1984), de Leeuw andMcKelvey (1984) and Struth (1984), amongst others. To see why this is so, let denote a forecast of the price P, to prevail at time t conditional on the information set 't-n available at t -n. The corresponding forecasting bias t_nBt is defined as the mean forecasting error conditional on I _,2:…”