Abstract: In this study, I show investors' responses to news differ sharply depending on investor sentiment and are consistent with confirmation bias. Specifically, I consider a mechanism in which increases in investor sentiment mean investors are more likely to have optimistic priors meaning they overreact to good cashflow news that confirms their priors and underreact to bad cashflow news that disconfirms their priors. The reverse patterns hold when investor sentiment is low, with underreaction to good news and overreaction to bad news. The mechanism provides a unifying explanation for the co-existence of investor underreaction and overreaction. It also delivers novel predictions which are validated using analyst survey and return data. These results imply investors' processing biases are not constant (the dominant assumption thus far) and that asset price movements are best understood through investors' priors. This has far-reaching implications for the expectations formation literature.

Abstract: Stock prices aggregate the beliefs of different investors. Using this insight, we estimate the fraction of stock market investors holding survey beliefs. We find that 42% of investors hold beliefs matching those of equity analysts and 25% hold beliefs as observed in individual investor return surveys. Together with risk aversion proxies and rational cash flow forecasts constructed using machine learning techniques, survey beliefs explain 87% of stock market fluctuations. Because investors likely form their beliefs by extrapolating prices and cash flows, we find stock prices would fluctuate 50% less if all investors held rational beliefs. Allowing for investor heterogeneity and using a price-driven price-to-earnings ratio reconciles prior studies.

Abstract:  To address the excess volatility puzzle—the excessive movement in stock prices—researchers often use variance decompositions of stock price ratios, where stock prices are scaled by fundamental measures. We demonstrate that stock price ratios based on fundamental measures with high transitory volatility, such as commonly used earnings measures, are not informative about movements in stock prices. To overcome this, we propose using Street earnings to construct the price-to-earnings ratio. Street earnings, calculated before transitory items, offer a more informative and persistent measure of future fundamentals. Since the Street price-earnings ratio extracts variation in stock prices and returns, its use is highly informative in asset pricing tests. Accordingly, we show that the Street priceearnings ratio has more in- and out-of-sample explanatory power for predicting returns than other valuation ratios. Additionally, we reconcile conflicting views on which subjective expectations drive stock price movements, finding that expectations of short-term earnings growth, long-term earnings growth, and returns can all help explain the excess volatility puzzle.

Abstract:  Trailing price ratios, such as the price-dividend and the price-earnings ratio, scale prices by trailing cash flow measures. They theoretically contain expected returns, yet, their performance in predicting stock market returns is poor. This is because of an omitted variable problem: trailing price ratios also contain expected cash flow growth. We show that structural changes in cash flow growth have undermined the power of trailing price ratios to predict returns. To construct an optimal valuation ratio, we propose scaling prices by cash flow forecasts (forward price ratios). We implement this idea for the S&P500 index using machine learning forecasts of firms’ cash flows. The out-of-sample explanatory power for one-year stock returns ranges from 7% to 11%, thereby beating all other predictors and helping to resolve the out-of-sample predictability debate (Goyal andWelch, 2008).

Abstract: We show that teachers’ value-added (‘VA’) depend on the quality of the teachers that preceded them. To do so, we use administrative data from North Carolina and find that a one-standard deviation increase in last grade’s mean teacher VA causes a 0.08σ decrease in teacher VA. Controlling for prior teacher assignment eliminates this bias. Under a benchmark policy that releases teachers in the bottom five percent of the VA distribution, thirty-two percent of teachers are incorrectly released using traditional techniques. Our results highlight the importance of incorporating dynamic features of education production into the estimation of teacher quality.