Print
E-MailAlexander Vedrashko is Associate Professor of Finance at the Beedie School of Business at Simon Fraser University. This post is based on a forthcoming article by Professor Vedrashko; Roni Michaely, Rudd Family Professor of Management at the Johnson Graduate School of Management at Cornell University; and Amir Rubin, Associate Professor of Finance at the Beedie School of Business at Simon Fraser University.
One striking behavioral regularity often cited as evidence of behavioral biases in the market is investors’ inattention on Fridays. This regularity is explained using the intuition that on Fridays, investors and traders could be preoccupied with the upcoming weekend and, thus, pay less attention to corporate news announcements on that day. Studies investigating this issue reported reduced response to earnings announcements (DellaVigna and Pollet, 2009) and merger announcements (Louis and Sun, 2010) on Fridays.
In our article we show that this pattern in investor behavior on Fridays appears to extend to corporate news events other than earnings and merger announcements. We find a reduced reaction to announcements of dividend changes, repurchases, and seasoned equity offerings (SEOs) on Fridays. Taken at face value, these combined results could present comprehensive and persuasive evidence that investors underreact to events occurring in the market on Fridays, which is consistent with inattention on these days. However, we show it is not investors’ inattention but rather selection bias that is the reason for the findings of reduced reaction on Fridays. We conclude that while various examples of cognitive constraints are reported in the literature, reduced investor reaction to Friday news is not a manifestation of this phenomenon.
Our method of addressing selection bias borrows from empirical research methodology in the medical and natural sciences. Because the relevant firm characteristics influencing the firm’s decision to announce on Friday are unknown, techniques such as matching firms, instrumental variables, and Heckman (1976) selection or treatment effect models can reduce the extent of the bias but do not fully eliminate it. We employ a two-step procedure that first tests whether the selection bias problem is present. We partition firms into two groups based on whether they have made at least one announcement on a Friday during the sample period (which we call the Friday announcer firms) and then compare the announcement reaction of the Friday announcer firms to that of non-Friday announcer firms on Mondays through Thursdays. This test is analogous to exposing all subjects to a placebo (a Monday-Thursday announcement day) in that market response to Monday-Thursday announcements should not be different between the two firm types, if inattention is associated with Friday rather than firm characteristics.
For all five announcement types, we find that the Friday announcer firms experience a lower market response compared to the non-Friday announcer firms on all weekdays, not only on Fridays. This suggests that Friday announcer firms differ from non-Friday announcer firms and there is nothing special about Friday as an announcement day in terms of its effect on market reaction. Thus, a study that overlooks non-random differences between firms will mistakenly attribute a differential response on Fridays to the announcement day rather than to confounding factors—firm characteristics. In particular, we find that Friday announcer firms of five different announcement types have common unobserved characteristics.
The second step of our method addresses the selection bias problem by testing the Friday inattention effect exclusively using the relatively homogeneous sample of the Friday announcer firms. This allows us to avoid the sample selection problem even when the source of the difference between the two groups of firms is unknown. In the full sample, reaction to Friday announcements mechanically appears smaller because it is benchmarked against the sample that unjustifiably includes the non-Friday announcer firms, which induce on average a greater announcement reaction than the Friday announcer firms. An analogy of this benchmarking problem is a study of cancer rates caused by the spread of prostate cancer in the body that mistakenly includes both men and women. Based only on the sample of Friday announcer firms (i.e., “men only”), the market reaction to announcements on Fridays is not different from that on other weekdays.
Overall, the methodological approach and empirical findings illustrate that in studies on announcement timing, firms whose announcements exhibit the “special timing” should be analyzed as a distinct group, due to their unobserved firm characteristics, to avoid selection bias. The method introduced here is applicable to other studies testing the effects of causal economic variables that appear to be external to the firm but can actually be associated with firm characteristics, such as in studies testing the effect of CEO or director attributes (tenure, age, gender, education, “star” status, overconfidence, or power) on firm performance or market reaction to various corporate announcements and events. The differential market reaction to corporate events could appear because CEOs and directors with different attributes may also tend to be employed by firms with certain characteristics that influence market reaction. For example, consider a study that initially finds that firms underperform if their CEO also chairs the board of directors. Selection bias can arise because the (possibly unobserved) characteristics of firms that tend to have CEO/chair dual appointments (“weak governance” firms) can be negatively associated with firm performance. The first step of the proposed method would be to test whether the “weak governance” firms underperform during periods when their CEO is not chairing the board. If the test result is affirmative, it means these firms have different performance from firms that have never had a dual appointment CEO, and the initial finding about the effect of the dual appointment on firm performance is influenced by selection bias. Thus, the second step in analyzing the effect of dual appointments would be to base the analysis on the bias-free sample of the “weak governance” firms. One advantage of the method we use is that it allows for quantifying the severity of a potential selection bias. The method also allows for eliminating the bias at the study design stage without identifying the influential firm characteristics and requiring data on them.
The full article is available for download here.
References
DellaVigna, S., Pollet, J., 2009. Investor inattention and Friday earnings announcements. Journal of Finance 74, 709–749.
Heckman, J., 1976. The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Annals of Economic and Social Measurement 5, 475–492.
Louis, H., Sun, A., 2010. Investor inattention and the market reaction to merger announcements. Management Science 56, 1781–1793.
