NBER Reporter: Winter 2000/2001
Barberis and Shleifer study asset prices in an economy in which some investors classify risky assets into different styles and move funds back and forth between these styles depending on relative performance. News about one style can affect the prices of other apparently unrelated styles; assets in the same style will move together too much, while assets in different styles co-move together too little; and high average returns on a style will be associated with common factors unrelated to risk. These assumptions imply that style momentum strategies will be very profitable. The authors use their model to shed light on a number of puzzling features of the data.
Scherbina investigates how differences in opinions regarding stock valuations influence prices. She finds that stock prices are driven by investors with an optimistic outlook whenever market and institutional frictions prevent pessimistic investors from expressing their opinion. As a result, market prices are more likely to be higher than consensus valuations when there are substantial differences of opinion. Using data on analysts' forecasts, Scherbina divides stocks into portfolios based on the dispersion in earnings forecasts and finds that portfolios containing stocks with highly dispersed forecasts on average earn 0.82 percent lower returns per month than portfolios with low-dispersion stocks. The difference in returns is more prominent for the low book-to-market and small stocks. She also documents that consensus forecasts are more upwardly biased the higher the dispersion in the underlying forecasts. This bias arises because the more pessimistic analysts choose not to issue a forecast for fear of jeopardizing their relationship with the management.
Chen, Hong, and Stein develop a model of stock prices in which there are differences of opinion among investors and constraints on short sales. Breadth of ownership is a valuation indicator in the model. When breadth is low--that is, when few investors have long positions in the stock -- this is a signal that the short-sales constraint is tightly binding. It implies that prices are high relative to fundamentals and that expected returns therefore are low. Reductions (increases) in breadth thus should forecast lower (higher) returns. Another prediction of the model is that changes in breadth should be positively correlated with other variables that forecast increased risk-adjusted returns. Using quarterly data on mutual fund holdings over the period 1979-98, the authors find evidence supportive of both of these predictions.
Barber, Odean, and Zheng analyze the mutual fund purchase and sale decisions of over 30,000 households with accounts at a large U.S. discount broker for the six years ending in 1996. They document three primary results. First, investors buy funds with strong past performance; over half of all fund purchases occur in funds ranked in the top quintile of past annual returns. Second, investors sell funds with strong past performance and are reluctant to sell their losing fund investments. Investors are twice as likely to sell a winning mutual fund than a losing mutual fund. Thus, nearly 40 percent of fund sales occur in funds ranked in the top quintile of past annual returns. Third, investors are sensitive to the form in which fund expenses are charged. Although investors are less likely to buy funds with high transaction fees (for example, broker commissions or front-end load fees), their purchases are relatively insensitive to a fund's operating expense ratio. Given evidence on the persistence of mutual fund performance, the purchase of last year's winning funds seems rational. However, the authors argue that selling winning fund investments and neglecting a fund's operating expense ratio when purchasing a fund is clearly counterproductive.
Chan, Karceski, and Lakonishok analyze historical long-term growth rates across a broad cross-section of stocks using a variety of indicators of operating performance. They ask whether it is possible to predict which firms will achieve high future growth using attributes such as past growth, industry affiliation (technology versus nontechnology), book-to-market ratio, past return, and security analysts' long-term forecasts. Historically, some firms have attained very high growth rates, but this is relatively rare. Only about 5 percent of surviving firms do better than a growth rate of 29 percent per year over ten years. Moreover, there is very limited ability to identify beforehand which firms will be able to generate such high long-term growth in the future. The historical patterns thus raise strong doubts about the sustainability of many stocks' valuations. Looking forward, the past growth record does not suggest a high expected return on stocks in general.
Abarbanell and Lehavy demonstrate that relatively small numbers of large optimistic and small pessimistic errors in analysts' forecasts have a disproportional impact on the statistical measure relied on in the earlier literature for drawing inferences about analysts' incentives and their proclivity to issued biased findings. The authors indicate that there is a common empirical source that underlies evidence of bias and inefficiency in distributions of analysts' forecasts, two phenomena that previously have been analyzed as separate manifestations of analyst irrationality. Also, the authors find that analysts do not account completely for systematic forms of earnings management intended to create accounting reserves or to beat market earnings expectations slightly. Taken together, these findings provide a challenge to researchers to refine the existing judgment and the incentive-based explanations for systematic analyst forecast errors in order to account for the role of unusual reported earnings realizations. The results also raise the possibility that systematic "errors" characterize equilibriums in which analysts are completely rational and face symmetric incentives.