The impact of arbitrage costs and earnings forecast accuracy on the predictive power of value-to-price ratios

In this dissertation, I empirically examine how arbitrage costs and earnings forecast accuracy affect the predictive power of value-to-price (V/P) ratios cross-sectionally. Prior studies (Frankel and Lee, 1998; Dechow et al., 1999; Lee et al., 1999) document that V/P ratios, where V is a stock's estimated intrinsic value and P is its price, can be used to predict future stock returns. However, prior studies do not examine whether the documented predictive power of V/P ratios varies cross-sectionally. This topic is important because it can help investors improve the effectiveness of a V/P ratio trading strategy and enhance our understanding on the fundamentals of V/P ratios' predictive power.; Lee et al. (1999) model prices and intrinsic values as a co-integrated system (LMS framework). In their framework, prices can deviate from intrinsic values because of noise trading (e.g. Shiller, 1984; DeLong et al., 1990a) or uninformed trading in a noisy rational expectation setting (Wang, 1993), but such deviations cannot last forever because arbitrage forces prices to converge to intrinsic values in the long run. The LMS framework suggests that the predictive power of V/P ratios depends on (a) the speed of price discovery (the price-to-value convergence processes), and (b) the accuracy of intrinsic value estimates. Based on the LMS framework and other theories, I identify two factors that can affect the predictive power of V/P ratios: arbitrage costs and earnings forecast accuracy. Arbitrage costs affect the price discovery speed, while earnings forecast accuracy affects the accuracy of intrinsic value estimates.; I select five proxies for arbitrage costs: past price movements conditional on extreme V/P rankings, bid-ask spreads, stock return volatility, growth rate, and accrual level. I employ a portfolio test and an OLS regression test to examine the impact of arbitrage costs on the predictive power of V/P ratios. During my sample period (1991–1997), a zero-cash investment strategy of buying firms in the top V/P quintile and short-selling firms in the bottom quintile yields large positive returns for firms with small arbitrage costs. However, such a strategy yields no or negative returns for firms with large arbitrage costs. An OLS regression analysis corroborates the portfolio test results. Such results cannot be explained by the Fama-French three factors.; My primary findings on the impact of earnings forecast accuracy on the predictive power of V/P ratios are as follows. First, I document that replacing analyst forecasts with actual future earnings dramatically increases the predictive power of V/P ratios, and such results cannot be explained by the Fama-French three factors or the link between stock returns and earnings surprises. Second, the predictive power of V/P ratios computed from earnings with less transitory components is slightly higher than that of V/P ratios computed from earnings with more transitory components. However, the impact of excluding transitory earnings on the predictive power of V/P ratios is statistically insignificant. Finally, I find that V/P ratios of firms with relatively more accurate forecasts have more predictive power. Moreover, the positive relationship between relative forecast accuracy and the predictive power of V/P ratios is monotonic.