future cash flow model than much of what is currently in the anomaly literature. A broad competing class in the predictability literature has focused on non-fundamental momentum effects that provide some predictability over sorted portfolios. Jegadeesh and Titman (1993) and Lee and Swaminathan (2000) report to find consistent profit opportunities on the order of 1% per month employing a basic strategy of buying past winners and selling past losers.

Another question of interest in testing a factor model based on fundamental information is the extent to which tangible fundamentals actually drive the stock market. The late 1990’s is a notable period in which the valuation of broad classes of equities de-coupled from traditional pricing measures. The classic dividend discount model of Gordon (1962) extended by Campbell and Shiller (1989) posits that the value of an individual equity or broad market index is a function of future anticipated cash payments. Changes in price should move in tandem with growth in dividends and expectations of higher dividends in the future. Pairing this model with the experience of the late 1990’s required extraordinary future dividend income growth to explain the rapid increase in equity valuations. Even an extension of the Campbell and Shiller model using earnings as a proxy for dividends struggles to account for the run up in equity valuations without allowing for substantial future earnings growth significantly greater than trend GDP growth. Despite record 7.5% real earnings growth in the 1990’s, as documented in Fama and French (2002), the non-fundamental equity premium, calculated here as real capital gains net of real earnings growth, was a substantial 5.22% per year. The dividend yield by the end of the 1990’s fell to as low as 1.1%, indicating a very high level of expected dividend growth in the future or a new regime of near zero discount rates. Therefore, a further question of interest is the degree to which observable fundamentals even matter in determining equity returns over different stages of the business cycle, or in the case of the late 1990’s, during a possible speculative boom.

One of the primary motivating factors in the development of the conditional pricing models of Ferson and Harvey (1999), Wu (2001) and Cochrane (1996) is the strong empirical evidence that equity market risk premia are time varying. Each of these conditional beta representations use lagged macroeconomic factors to capture time variation. Another approach to addressing time variation in risk factors is to allow the evolution of risk sensitivities to evolve in a Bayesian

2