Differential performance impacts of outsider and insider interim CEO successions

ABSTRACT

This study finds differential performance impacts of outsider and insider interim CEO successions. Stock markets react incrementally positively to interim CEO successions when the interim CEOs are from outside rather than inside firms. Relative to insider interim CEOs, outsider interim CEOs produce worse earnings performance and greater restructuring charges during their services but lead more frequently to outsider permanent CEO appointments followed by better long-term earnings performance. Our findings suggest that market reactions to interim CEO successions rationally impound the investors’ expectation that outsider relative to insider interim CEO successions will generate greater short-term disruption but pave the way for greater long-term performance improvement.

KEYWORDS:

• CEO turnover
• interim CEO succession
• interim CEO origin
• disruption
• turnaround
• stock market reaction
• firm performance

Data availability

The data used are available from public sources indicated in the paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. In our sample between 1997 and 2016, interim CEO successions constitute 12.2% of total CEO turnovers (excluding death, illness, and new job appointments) and stock market reactions to these interim CEO successions suggest that they generate an economically significant, $31 billion loss of total market capitalization.

2. See for example Behn et al. (2006), Ballinger and Marcel (2010), or Intintoli (2013).

3. The case study by Mooney, Semadeni, and Kesner (2012) discusses anecdotes suggesting that all interim CEOs are not the same and the tasks of interim CEOs mandated by boards of directors are associated with their origins. However, they do not formally examine differential performance impacts of insider versus outsider CEO elections.

4. A recent example is Edward Breen of DuPont (Chakravorty 2015). In 2015, suffering a decline in profitability, DuPont needed to make a dramatic change in operations. In October 2015, DuPont CEO Ellen Kullman stepped down and the board of directors named as interim CEO Edward Breen, who lead restructurings in Tyco International from 2002 to 2012. In November 2015, Breen assumed the role of CEO and board chairman permanently.

5. Mooney, Semadeni, and Kesner (2012) provide Andrew Miller of Guardian Media Group as an example of a ‘contender’ interim CEO. Guardian Media Group CFO Miller was appointed interim CEO in April 2010 and appointed permanent CEO in July 2010. They also mention Paul Allaire of Xerox as an example of a ‘groomer’ interim CEO. Allaire groomed Anne M. Mulcahy for the position of permanent CEO and his firm-specific knowledge and experience as the former Xerox CEO brought certainty and stability to the organization.

6. We employ two measures of quarterly earnings performance produced by interim CEOs: (1) The difference between the quarterly earnings reported by an interim CEO and the same quarter earnings forecasted by the analyst group before the appointment of the interim CEO (UE_QTR) and (2) the industry-adjusted quarterly return on assets (ADJROA_QTR).

7. In our sample, most heirs apparent are not promoted to interim or permanent CEO, suggesting the abandonment of an internal succession plan. This is more prominent in outsider interim succession than in insider interim succession.

8. To our knowledge, the prior literature on interim CEO succession does not use propensity score matching to control for potential bias inherent in these covariates.

9. The negative short-term performance impact of disruption arising from interim leadership can be regarded as a part of CEO turnover costs necessary to make long-term performance turnaround.

10. Exceptionally, Rivolta (2018) documents that delay in CEO successions is associated with positive change in stock and accounting performance, particularly in firms without succession plans in place.

11. In both interim succession and normal leadership transition, a firm also has to spend resources on searching for and grooming a qualified permanent successor. The cost of searching for an interim CEO is incremental but likely to be immaterial because an interim CEO search generally does not require as thorough an evaluation of candidates as a permanent CEO search.

12. Intintoli (2013) reports firm performance after CEO departure announcements without concurrent appointment of new permanent successors. His sample includes (1) interim CEO appointments as well as (2) CEO departure announcements after which outgoing CEOs remain in office until a new permanent CEO is named. In the latter case, the post-announcement firm performance reflects departing CEOs’ performance.

13. In case of CEO death or illness investigated by Behn et al. (2006) and Rivolta (2018), the desirable match-quality for a new permanent CEO is largely random and the interim CEO’s role is largely that of seat-warmer. In their setting, we cannot test the performance impacts of outsider versus insider interim CEOs who may play more or less active roles in operation, permanent successor search, and restructuring to achieve ultimate performance turnaround.

14. In our sample, 30% of outsider interim CEOs are promoted to new permanent CEO, while 22% of insider interim CEOs become new permanent CEOs, suggesting that a considerable number of interim CEOs play the role of contender rather than seat-warmer and this is more prominent for outsider than insider interim CEOs.

15. See the Xerox case in footnote 5 for an example of a ‘groomer’ interim CEO.

16. See the DuPont case in footnote 4 for an example of an outsider interim CEO who undertakes fixing urgent problems and later be promoted to a permanent CEO.

17. If the data is unavailable or unclear in Factiva, we search the BoardEx database and the executive profile and biography section of the Bloomberg website (https://www.bloomberg.com) to find reliable data.

18. We allow for a maximum of 30 days from the departure announcement date in defining immediate departure because we occasionally find that the actual departure date of a CEO in the ExecuComp and BoardEx databases are coded as of the beginning or ending date of the departure month, although the press release in Factiva reports the same departure month but a different departure date.

19. If the interim CEO office is made up of a group of officers among whom at least one has been employed for more than one year prior to the CEO departure date, this is also classified as insider interim CEO (N = 11). Six interim cases do not assign the official title of interim CEO to the new interim leader. Instead, he or she is given a non-CEO title such as interim president or interim chairman of the board. According to press releases, despite the absence of an official CEO title, these non-CEO officers also perform the role of interim CEO.

20. To mitigate the effects of influential observations on our inferences, we winsorize all continuous variables used in this study at the top and bottom one percent of the sample distributions.

21. If a fiscal quarter is managed only partially by a CEO, we do not count it as a performance quarter of the CEO in Equations (3)(3). $EARN\mathrm{\_}QT{R}_t={\beta }_0+{\beta }_{1}OutsiderInterim\mathrm{\_}PFQT{R}_t+\sum {\gamma }_{j}Control{s}_{jt}+{\epsilon }_t$(3). and (4(4) $RESTRUCT\mathrm{\_}DU{M}_t={\beta }_0+{\beta }_{1}OutsiderInterim\mathrm{\_}PFQT{R}_t+\sum {\gamma }_{j}Control{s}_{jt}+{\epsilon }_t$(4) ). Unlike our study, the prior interim CEO studies (e.g., Ballinger and Marcel 2010) do not clearly identify earnings performance during a full quarter under the interim CEO leadership.

22. Chen et al. (2015) report that interim CEOs who engage more in income-increasing earnings management are more likely to be promoted to the permanent position. We include MEETBEAT1CENT to control for interim CEOs’ incentive to inflate earnings to meet or beat the analyst forecast by one cent. We also measure abnormal balance-sheet working capital accruals using a cross-sectional quarterly Jones model to control for managers’ earnings management incentive more generically. After deleting non-missing abnormal accruals, only 63 observations remain in the sample, making it difficult to interpret test results based on the accrual measure.

23. We measure ROA as operating profit after depreciation and amortization deflated by the beginning-of-the-year total assets because operating income excludes the effects of income from asset sales, financing (interest expense), tax position, recurring foreign exchange translation gains and losses, and extraordinary items.

24. Stock price changes at the announcement of interim CEO succession would rationally reflect the difference between (1) the present value of expected future cash flows arising from restructuring and long-term performance turnaround as a result of the leadership change and (2) the present value of expected future cash flows in the absence of such leadership change. Ceteris paribus, investors’ concern over the increase in future cash-flow uncertainty due to the leadership change would lead to a higher discount rate and therefore reduce the present value of future cash flows after the leadership change, possibly turning the market reaction into a negative value.

25. For example, Panel C of Table 1 shows that the mean of industry-adjusted stock returns 12 months prior to interim CEO appointments (ADJRTN_P1) is −22.93%. Untabulated results indicate that the mean of unadjusted stock returns prior to interim CEO appointments is −7.09% and industry mean return is 15.84%, suggesting that the difference between raw returns and industry mean returns produce the large negative number of industry-adjusted stock returns. Our main test results remain qualitatively unchanged after excluding the observations that have extreme past industry-adjusted returns (i.e., ADJRTN_P1 lower than −80% or ADJRTN_P3 lower than −80%).

26. In contrast, the negative (statistically insignificant) coefficient on SCANDAL may reflect potential new adverse information about the quality of the CEO that unexpected dismissal associated with scandal conveys.

27. In our sample, all cases with POLICYDIF equal to one perfectly predict no restructuring. We drop these cases (N = 15) from the probit estimation in Column (2) of Table 3.

28. In our sample, the trend in pre-turnover industry-adjusted ROA is more meaningful than the absolute level of pre-turnover industry-adjusted ROA for the following reason. Our sample of CEO turnover is a subset of S&P1500 firms because we draw CEO turnover observations from the intersection of ExecuComp and RiskMetrics. On the other hand, we measure the industry average of ROAs using the universe of I/B/E/S firms per industry and year. In general, S&P1500 firms outperform the average I/B/E/S firms with regard to ROAs because a firm enters into S&P1500 index if the firm satisfies a set of criteria including firm performance. Therefore, it is unclear whether a higher industry-adjusted ROA of our sample firm is due to the membership of S&P1500 firms or due to good performance of current management. In this regard, changes in pre-turnover industry-adjusted ROA may capture management performance more meaningfully than level of pre-turnover industry-adjusted ROA.

29. If we restrict the sample period over year t-2 to t + 1, the balanced sample would consist of 64 outsider interim successions and 141 insider interim successions. The larger sample size is due to the availability of Compustat data. The sign and magnitude of the mean and median of changes in ADJROA, ADJATO, and ADJOM over year t-2 to year t + 1 using the two balanced samples are similar and our inference on performance changes are qualitatively identical between the two samples. These results indicate that our findings based on the period year t-2 to t + 2 are not driven by the survivorship bias.

30. Untabulated results indicate that pre-turnover industry-adjusted ROA does not differ between the outsider and insider interim CEO groups within the matched sample as well.

31. After the propensity score matching, we tabulate the long-term performance results in Table 7 using all the observations of outsider and insider interim succession that have non-missing performance data from year t-2 to t + 2. If we require that both of the matched pairs for outsider and insider interim succession observations should have non-missing performance data from year t-2 to t + 2, the number of the pairs available for test reduces to 42 (i.e., 84 observations). The untabulated results using these 42 pairs are qualitatively identical to those in Table 7 except for the following: the mean of change in ADJROA[−1,+2] and the median change in ADJOM[−1,+2] are significantly greater for outsider interim succession compared to insider interim succession. These results using the 42 pairs further confirm the robustness of our main results.