Print
E-MailThis post comes to us from Edmund Keung, Zhi-Xing Lin and Michael Shih of the National University of Singapore.
In our paper, Does the Stock Market See a Zero or Small Positive Earnings Surprise as a Red Flag?, which was recently accepted for publication in the Journal of Accounting Research, we investigate the stock market reaction to specific levels of reported earnings. Akerlof’s classic analysis of a market with information asymmetry suggests that if potential buyers cannot distinguish good used cars from bad ones, the prices they will pay for all used cars are lower than what good used cars are worth. In our setting, this analysis suggests an increasing incidence of firms managing earnings and/or analyst expectations will induce investors to discount not only the shares of firms that are confirmed manipulators, but also those of firms that are mere “suspects.” To the extent that some of the suspects are actually innocent, lower valuations for the suspect firms would represent a cost jointly borne by all firms for the “numbers game” that not all of them play.
We investigate whether firms incur the aforementioned cost in two stages. We first document a rising trend in the number of firms meeting or narrowly beating analyst earnings forecasts relative to the number of firms narrowly missing the forecasts in the period 1992-2006. This would suggest to investors a rising prevalence of firms playing the numbers game, given empirical results indicating manipulators prefer to meet or narrowly beat analyst earnings forecasts rather than to beat them by a large margin. The impression of a rising number of “lemons” (manipulators) is likely to increase investors’ skepticism toward firms that “make the numbers,” especially those that meet or narrowly beat analyst earnings forecasts, resulting in zero or small positive earnings surprises.
We test this in the second stage of our investigation. We compare the coefficient in the regression of abnormal stock returns around the earnings announcement on earnings surprise across ranges of earnings surprises. The coefficient is referred to in prior studies as the earnings response coefficient (ERC). We find the ERC is significantly lower for earnings surprises in the range [0, 1¢] than for those in adjacent ranges ([-1¢, 0) and (1¢, 2¢]) for firm-quarters in 2002-2006, but not for those in either the period 1992-1996 or 1997-2001. These results suggest investors’ skepticism toward zero and small positive earnings surprises is a fairly recent event, and its development over time was induced by the rising tide of firms playing the numbers game. Our results are robust to the inclusion of controls. Our results are unchanged after controlling for the sign of estimated discretionary accruals and the trajectory of analyst forecasts before the earnings announcement.
We also find evidence that investors’ skepticism toward zero and small positive earnings surprises is justified. The relation of future earnings surprise with current earnings surprise is more negative for current earnings surprises in [0, 1¢] than for those in any other range throughout the period 1992-2006. It appears that analysts became aware of the problem earlier than investors. We compare the coefficient in the regression of analysts’ revision of next-quarter earnings forecast on earnings surprise across ranges of earnings surprises. We term this coefficient the analyst earnings response coefficient (AERC). The AERC is lower for earnings surprises in [0, 1¢] than for those in any other range throughout the period 1992-2006. There is also evidence that investors and analysts are skeptical about firms that narrowly avoid quarterly losses or quarterly earnings declines throughout the same period.
The full paper is available for download here.
