Short-window earnings response coefficients estimated from pooled time-series cross-sectional regressions are systematically smaller than corresponding averages of firm-specific coefficients estimated from time-series regressions. The cause is a negative relation between firm-specific earnings response coefficients and unexpected earnings variances. If the hypotheses of equality of firm-specific coefficients and equality of firm-specific unexpected earnings variances are rejected, firm-specific estimation should be used instead of pooled estimation. Using pooled estimation may lead to incorrect inferences about the magnitude of estimated coefficients and/or incorrect inferences about differences in coefficient behavior between groups of firms.

Under Ralston Purina Company's 1986 incentive contract 14 managers would receive 49.1 million in stock if within ten years the stock price closed above 100 for ten consecutive days. While the contract required a 57.8% increase in stock price, it did not motivate managers to create value because the rate of return required to reach 100 in ten years was substantially less than Ralston's cost of equity capital at the time of the contract's adoption. Barring any action by managers that would substantially change the market's expectations about the firm, reaching the 100 hurdle price would be easy. In fact, managers collected the contract's payoffs within five years despite an industry-adjusted loss of $2.1 billion in shareholder value.

The size effect is sensitive to the length of the return interval used in estimating betas. Beta changes with the return interval because an asset's covariance with the market and the market's variance do not change proportionately as the return interval is changed. We document beta sensitivity to the return interval. Evidence from cross-sectional regressions of returns on monthly and annual betas is inconsistent with beta changes stemming only from the higher standard errors of the longer-interval betas. We provide evidence that the size effect becomes statistically insignificant when risk is measured by betas estimated using annual returns.

This study assesses the impact of the firm size effect on test statistics based on market-adjusted and market model abnormal returns. The performance of two alternative abnormal return methods that explicitly control for the effect of firm size on expected returns is also examined. These methods, the size control portfolio and size model approaches, are based upon a companion portfolio approach where companion portfolios are constructed on the basis of firm size. Simulation results indicate that when event dates are clustered in calendar time and the event affects either small or large firms, conventional t-statistics based on market-adjusted or market model abnormal returns are misspecified in that their empirical Type I error rates deviate significantly from those expected under a true null hypothesis. Significance tests based on empirical distributions mitigate these over-rejection tendencies but, in doing so, they sacrifice power. On the other hand, conventional t-statistics based on size control portfolio or size model abnormal returns are well-specified across a variety of event conditions. Accordingly, either size-based method can control for the firm size effect in event study contexts where such control is warranted. Since the size control portfolio approach is simpler to implement, it is the preferred alternative. The study also documents that abnormal performance is detected more often when large firms are affected than when small firms are affected.

SFAS No. 2, R&D, Market reaction, R&D expensing

The capital asset‐pricing model's (CAPM) primary empirical implication is a positively sloped linear relation between a security's expected rate of return and its relative risk (beta). Recent research indicates that inferences about the risk‐return relation are sensitive to the choice of the return measurement interval. We perform multivariate tests of the Sharpe‐Lintner CAPM using monthly and annual returns on market‐value‐ranked portfolios. The CAPM is rejected using monthly returns, a result consistent with previous research. In contrast, we fail to reject the CAPM when annual holding period returns are used.

We examine the specification and power of tests based on performance-matched discretionary accruals, and make comparisons with tests using traditional discretionary accrual measures (e.g., Jones and modified-Jones models). Performance matching on return on assets controls for the effect of performance on measured discretionary accruals. The results suggest that performance-matched discretionary accrual measures enhance the reliability of inferences from earnings management research when the hypothesis being tested does not imply that earnings management will vary with performance, or where the control firms are not expected to have engaged in earnings management.