Previous estimates of a ‘size effect’ based on daily returns data are biased. The use of quoted closing prices in computing returns on individual stocks imparts an upward bias. Returns computed for buy-and-hold portfolios largely avoid the bias induced by closing prices. Based on such buy-and-hold returns, the full-year size effect is half as large as previously reported, and all of the full-year effect is, on average, due to the month of January.

We develop a framework for estimating expected returns—a predictive system—that allows predictors to be imperfectly correlated with the conditional expected return. When predictors are imperfect, the estimated expected return depends on past returns in a manner that hinges on the correlation between unexpected returns and innovations in expected returns. We find empirically that prior beliefs about this correlation, which is most likely negative, substantially affect estimates of expected returns as well as various inferences about predictability, including assessments of a predictor's usefulness. Compared to standard predictive regressions, predictive systems deliver different expected returns with higher estimated precision.

Costs of equity for individual firms are estimated in a Bayesian framework using several factor‐based pricing models. Substantial prior uncertainty about mispricing often produces an estimated cost of equity close to that obtained with mispricing precluded, even for a stock whose average return departs significantly from the pricing model's prediction. Uncertainty about which pricing model to use is less important, on average, than within‐model parameter uncertainty. In the absence of mispricing uncertainty, uncertainty about factor premiums is generally the largest source of overall uncertainty about a firm's cost of equity, although uncertainty about betas is nearly as important.

Sample evidence about the predictability of monthly stock returns is considered from the perspective of a risk-averse Bayesian investor who must allocate funds between stocks and cash. The investor uses the sample evidence to update prior beliefs about the parameters in a regression of stock returns on a set of predictive variables. The regression relation can seem weak when described by usual statistical measures but the current values of the predictive variables can exert a substantial influence on the investor's portfolio decision, even when the investor's prior beliefs are weighted against predictability.

Sample evidence about the predictability of monthly stock returns is considered from the perspective of a risk‐averse Bayesian investor who must allocate funds between stocks and cash. The investor uses the sample evidence to update prior beliefs about the parameters in a regression of stock returns on a set of predictive variables. The regression relation can seem weak when described by usual statistical measures, but the current values of the predictive variables can exert a substantial influence on the investor's portfolio decision, even when the investor's prior beliefs are weighted against predictability.

Sample evidence about the predictability of monthly stock returns is considered from the perspective of a risk-averse Bayesian investor who must allocate funds between stocks and cash. The investor uses the sample evidence to update prior beliefs about the parameters in a regression of stock returns on a set of predictive variables. The regression relation can seem weak when described by usual statistical measures, but the current values of the predictive variables can exert a substantial influence on the investor's portfolio decision, even when the investor's prior beliefs are weighted against predictability.

The Capital Asset Pricing Model implies that (i) the market portfolio is efficient and (ii) expected returns are linearly related to betas. Many do not view these implications as separate, since either implies the other, but we demonstrate that either can hold nearly perfectly while the other fails grossly. If the index portfolio is inefficient, then the coefficients and from an ordinary least squares regression of expected returns on betas can equal essentially any values and bear no relation to the index portfolio's mean‐variance location. That location does determine the outcome of a mean‐beta regression fitted by generalized least squares.

The capital asset pricing model implies that the market portfolio is efficient and expected returns are linearly related to betas. Many do not view these implications as separate, since either implies the other, but the authors demonstrate that either can hold nearly perfectly while the other fails grossly. If the index portfolio is inefficient, then the coefficient and R[squared] from an ordinary least squares regression of expected returns on betas can equal essentially any values and bear no relation to the index portfolio's mean-variance location. That location does determine the outcome of a mean-beta regression fitted by generalized least squares.

We argue that active management’s popularity is not puzzling despite the industry’s poor track record. Our explanation features decreasing returns to scale: As the industry’s size increases, every manager’s ability to outperform passive benchmarks declines. The poor track record occurred before the growth of indexing modestly reduced the share of active management to its current size. At this size, better performance is expected by investors who believe in decreasing returns to scale. Such beliefs persist because persistence in industry size causes learning about returns to scale to be slow. The industry should shrink only moderately if its underperformance continues.