premium becomes negative and mostly insignificant with the exception being a near significant negative risk premium for Bayesian market betas in the presence of the TVPFM predicted returns. On the criteria suggested by Kan and Zhang, the return on the value weighted market return could be considered a useless factor.
The work presented here lends support to previous work laying out the importance of lagged macroeconomic information in determining expected investor returns. Further, we find these same macroeconomic factors useful in forecasting returns directly. The information derived from recently observed macroeconomic fundamentals appears particularly important at business cycle turning points and periods of high economic uncertainty such as the period immediately following the equity market peak in March 2000. Those following the stock market just prior to the peak in 2000 no doubt recall an emphasis on new non-traditional measures of valuation quite distinct from the macroeconomic and financial factors investigated here. As noted in Campbell and Shiller (1989), a rising P/D ratio is indicative of higher expected returns in the future. Therefore, the instruments driving returns in the 1994 to early 2000 sample may have been more forward looking measures that are difficult to quantify, and as such cannot be easily introduced into a macroeconomic model such as this. Despite this possible shortcoming, the model does at least as well as the traditional CAPM at pricing risk during our weekly sample period, even prior to 2000.