The question
When a US-listed company announces a CEO change, how does its share price respond, and does the response depend on why the CEO left and how much influence they held?
An announcement return measures the market's revaluation on the news, not the value a CEO adds to the business. That news bundles what investors learn about the incoming CEO with what the board reveals by acting: dismissing a CEO signals that the board has seen something wrong. No benchmark separates the two. Results here are read as “how the market prices CEO change announcements”, and the design below aims to make that reading as clean as possible.
Threats to interpretation
A raw price move around an announcement is not an effect. Each threat below has a planned response.
| Threat | Why it matters | Response |
|---|---|---|
| Prior decline | Boards often replace CEOs after the stock has fallen; beaten-down stocks tend to rebound regardless. | Short event windows; pre-event run-up reported; matched comparison firms for longer horizons. |
| Market and sector moves | Prices move with the market and the industry on the day. | Abnormal returns against market and sector benchmarks. |
| Biased benchmark | A market model fitted over a year of decline learns a negative “normal” return, inflating post-event abnormal returns. | Estimation window ending well before the event; market-adjusted model as a robustness check. |
| Concurrent news | CEO changes often coincide with earnings, guidance, restructurings or deals. | Flag events whose filing day includes other 8-K items, e.g. 2.02 (results), 1.01/2.01 (deals), 2.05 (restructuring). |
| Timing and leaks | Filings can come after the market closes or days after the news. | After-close filings dated to the next trading day; pre-event days examined for leakage. |
| Microcaps | Many filers are very small or thinly traded and would dominate an equal-weighted average. | Price and liquidity filters; results by size group and value-weighted. |
Identification strategy
- Short windows first. The main test uses the announcement day and the day after, where drift and mean reversion have little time to act.
- Forced vs planned. Planned successions reveal little negative information about the firm, so their reaction is closer to a pure signal about leadership. Classification will follow established definitions in the literature (Parrino, 1997).
- Unplanned exits as a clean subsample. Sudden deaths and health-related departures are not timed by the board, so they carry no board signal (Johnson et al., 1985; Quigley, Crossland & Campbell, 2017). The subsample will be small but is the closest available to a causal estimate.
- Placebo events. The same method run on random dates at comparable firms should find nothing; if it does, the method rather than the CEO news is driving results.
- Longer horizons. Returns over months are compared with matched firms of similar size, sector and prior returns that had no CEO change (Barber & Lyon, 1997).
CEO influence
Rather than assigning weights, we test whether the reaction is larger when proxies for CEO influence are high (Finkelstein, 1992; Adams, Almeida & Ferreira, 2005). If it is not, that is reported as a finding.
| Proxy | Source | Effort |
|---|---|---|
| Firm size | Price data and shares outstanding | Low |
| CEO also chair | Wording in the 8-K itself | Low |
| Tenure | 8-K wording or the earlier appointment event | Medium |
| Founder | Proxy statements and company descriptions | High |
| Large shareholder | Beneficial ownership table in the proxy statement | High |
- Size of reaction, not only direction. Losing a powerful CEO can be good news (entrenchment ends) or bad news (talent leaves); averaged together these cancel. The absolute abnormal return and abnormal trading volume are tested alongside the signed return.
- Joint, not one at a time. The proxies are correlated with each other, with firm size and with forced vs planned departures, so they enter one regression with controls for size, prior returns, departure type, sector and year. Coefficients are read as associations.
Statistical discipline
- Fixed in advance. The main windows, benchmark and hypotheses on this page are set before any returns are computed. Analyses added later are labelled exploratory.
- Multiple testing. Several proxies, windows and outcomes multiply the chance of a false positive. The primary influence test uses a single pre-specified index (the average of standardised proxies); other tests are corrected for multiple comparisons.
- Robust test statistics. Tests allow for event-induced changes in volatility and for correlation across events (Boehmer, Musumeci & Poulsen, 1991; Kolari & Pynnönen, 2010).
- Power before nulls. With a few thousand events, differences of under one percentage point between halves of the sample should be detectable; small subgroups such as founder departures will be far less precise, and a null there is weak evidence.
Signal test
The announcement reaction describes how markets price CEO news, but it cannot be traded: by the time a filing appears, the price has usually moved. A separate question is whether returns after the first tradable moment are predictable from what was known at the filing. If they are, the market has under- or over-reacted. The rules below are fixed before any post-event returns are examined.
| Rule | Requirement |
|---|---|
| No look-ahead | Every feature, including forced vs planned and successor type, is derived from the filing itself, not from later news. |
| Realistic entry | Returns start at the next market open after the filing’s acceptance time. |
| Out of sample | Rules are built on events from 2021–2024 and tested once, unchanged, on 2025–2026. |
| Factor adjustment | Returns are adjusted for market, size, value and momentum, since firms changing CEO are often recent losers. |
| Costs and liquidity | Results are reported after estimated trading costs and for liquid stocks separately. |
| Delisted firms included | Prices for firms that later delisted or were acquired are kept, so the test does not see only survivors. |
Horizons are 5 trading days and 1, 3 and 6 months. The expected result, given how public and well studied CEO changes are, is no edge after costs; a clean null is reported as a finding. Nothing on this site is investment advice.
Holdout test, 2016–2020
Patterns on the Statistics page were found by exploring events from 2021–2026, so they could partly be chance. The dataset is being extended back to 2016, and the five years 2016–2020 are kept as a holdout: the hypotheses below were written on 29 October 2026, before any returns for those years were computed, and will be tested once, unchanged. Results will be reported whatever they show. H7–H9, about the CEOs themselves, were added the same day, before those measures had been computed for any years, and are tested on both periods.
| Hypothesis | Test |
|---|---|
| H1. CEO changes are followed by a negative reaction on average. | Mean abnormal return, days −1 to +1, below zero. |
| H2. Naming an interim CEO is worse than naming an insider. | Negative coefficient on interim successor. |
| H3. Hiring an outsider is better than naming an insider. | Positive coefficient on outside successor. |
| H4. Results announced alongside are associated with a worse reaction. | Negative coefficient on results in the same release. |
| H5. An abrupt departure has no effect of its own once the successor is known. | Abrupt coefficient not distinguishable from zero. |
| H6. The successor effects are priced at the announcement, with no later drift. | Interim and outside-hire coefficients for days +2 to +60 not distinguishable from zero. |
| H7. A departure after poor performance is received better than one after good performance. | Positive coefficient on underperformance: the stock's return minus the market's over the three years before the announcement, sign reversed. |
| H8. The departure of a long-serving or founder CEO is received worse. | Negative coefficient on outgoing CEO tenure (years) and on founder status. |
| H9. A successor who has been a CEO before is received better. | Positive coefficient on successor with prior CEO experience. |
All factors are estimated together in one regression, with sector and size held fixed, on abnormal returns standardised by each stock's normal volatility, with standard errors clustered by week. H1–H4 are one-sided tests at the 5% level; H5 and H6 are judged by whether the 95% interval includes zero. Only genuine CEO changes (as judged by the language-model check) at listed companies priced at least $1 are used.
Status
| Component | State |
|---|---|
| Item 5.02 filing index | Active |
| CEO change identification | Active |
| Price data and abnormal returns | Active |
| Confound flags | Active |
| Forced vs planned classification | Active |
| Extension to 2016 and holdout test | In progress |
| Size and sector benchmarks | Planned |
| CEO influence proxies | Planned |
| Post-event signal test | Planned |
References
- Adams, R., Almeida, H. & Ferreira, D. (2005). Powerful CEOs and their impact on corporate performance. Review of Financial Studies.
- Barber, B. & Lyon, J. (1997). Detecting long-run abnormal stock returns. Journal of Financial Economics.
- Boehmer, E., Musumeci, J. & Poulsen, A. (1991). Event-study methodology under conditions of event-induced variance. Journal of Financial Economics.
- Finkelstein, S. (1992). Power in top management teams: Dimensions, measurement, and validation. Academy of Management Journal.
- Johnson, W. B., Magee, R., Nagarajan, N. & Newman, H. (1985). An analysis of the stock price reaction to sudden executive deaths. Journal of Accounting and Economics.
- Kolari, J. & Pynnönen, S. (2010). Event study testing with cross-sectional correlation of abnormal returns. Review of Financial Studies.
- Parrino, R. (1997). CEO turnover and outside succession: A cross-sectional analysis. Journal of Financial Economics.
- Quigley, T., Crossland, C. & Campbell, R. (2017). Shareholder perceptions of the changing impact of CEOs. Strategic Management Journal.