Dynamic risk adjustment in long-run event study tests

ABSTRACT

The existence of long-run abnormal returns after major corporate events has become a controversial subject of debate. We contribute new evidence by implementing a daily rolling prediction error (RPE) approach using popular asset pricing models to adjust for time-varying risk parameters in asset pricing models when estimating long-run abnormal returns. Using this simple approach, we find initial significant return responses in the month or two after SEOs and M&As but none thereafter. Robustness checks with different asset pricing models, corporate events, and subperiods corroborate our results. Also, simulation tests confirm the robustness of the RPE method to potential risk shifts. We conclude that, after dynamic risk adjustment, long-run abnormal returns do not occur after the major corporate actions under study.

KEYWORDS:

• Abnormal return
• long-run event study
• merger and acquisition
• seasoned equity offering
• stock repurchase
• stock split’

JEL CLASSIFICATION:

• C10
• G14
• G32
• G34
• G35

Acknowledgements

We are thankful for helpful comments and suggestions from two anonymous referees, seminar participants at the 2020 Southwestern Finance Association Conference, 2020 Financial Management Association Conference, and Seminar in the Department of Finance, Texas A&M University. Also, comments are appreciated from Ihsan Badshah, Boone Bowles, Leo Chan, Yong Chen, Bilal Ertuk, Mohammadali Fallah, Tristan Fitzgerald, Jianhua Huang, Shane Johnson, Ankit Kapur, Adam Kolasinski, Dipali Krishnakumar, Huan Qiu, and Wei Wu. Remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ See studies by Loughran and Ritter (1995), Brav, Geczy, and Gompers (2000), Billett, Flannery, and Garfinkel (2011), Malmendier, Moretti, and Peters (2018), Kolari, Pynnonen, and Tuncez (2022), and others.

² See behavioural theories by Kahneman and Tversky (1982), Barberis, Shleifer, and Vishny (1998), and others.

³ For example, see Eckbo, Masulis, and Norli (2000), Boehme and Sorescu (2002), Buyn and Rozeff (2003), Eckbo, Masulis, and Norli (2007), Bessembinder and Zhang (2013), Fu and Huang (2015), and others.

⁴ Eckbo, Masulis, and Norli (2000) argues that underperformance after SEOs is likely explained by risk changes. More generally, Kothari and Warner (2007, p. 21) point out that, in contrast to short-run studies, risk-adjustment is crucial in long-run studies to avoid false abnormal return measurements.

⁵ As mentioned by Scholes and Williams (1977) and Dimson (1979), nonsynchronous trading could be a issue using daily data. We repeated our tests with Dimson aggregated coefficients, but the results are unchanged. Methodological details and empirical results are available upon request from the authors.

⁶ In forthcoming empirical analyses, we utilize Fama-French factor models in place of the market model also.

⁷ See, for example, Eckbo (1983). As pointed out by an anonymous referee, some recent examples are Boudaker et al. (2022), Pandey and Kumari (2021), and Yousaf, Riaz, and Goodell (2023).

⁸ Because compounded returns accurately match multi-period returns, we can define $CA{R}_{iT}$ as $(1+CA{R}_{iT-1})\times (1+A{R}_{iT})-1$ instead of the sum of $A{R}_{iT}$. However, the results turn out to be virtually identical for these two definitions of abnormal returns.

⁹ In unreported results, we fixed the before estimation window at 2 months and used different post-event windows to test risk shifts. Our findings are similar to Table 6.

¹⁰ We also estimated abnormal returns using only the post-event estimation window. As before, no significant long-run abnormal returns are found. Moreover, short-run abnormal returns are positive which indicates risk shifting.

¹¹ Similar SPLT results are available upon request.

¹² It is worth noting that, in our sampling procedures, we dropped small M&A deals. Also, SRs occurred relatively infrequently among smaller firms compared to larger firms. Hence, compared to SEO firms, M&A and SR firms tend to be relatively larger in our samples. These size differences help to explain the relatively higher beta risk of SEO firms compared to M&A and SR firms in Figure 3.

¹³ Similar SR and SPLT results are available upon request.

¹⁴ Similar results for SPLTs are available upon request. No long-run abnormal returns are detected.

¹⁵ Beta changed more than 10% in our empirical tests reported above.

¹⁶ Since abnormal returns only exist in the first 6 months after events, given the use of a 6-month estimation window, any test after the second year is used to test our model specification.