Event study

An event study is a powerful tool used by researchers to assess the financial impact of changes in corporate policy or other events on the stock prices of firms. It is frequently used in management and finance research.

The basic idea is to determine whether there is an “abnormal” stock price effect associated with an unanticipated event. This method helps researchers infer the significance of the event by measuring its impact on firm value, often using stock price changes as a proxy for financial performance or firm performance. Stock prices are assumed to reflect the true value of firms because they are assumed to be driven by discounted value of all future cash flows and incorporate all relevant information.

Here is a step-by-step approach based on the process outlined in the sources:

1. Define an Event that Provides New Information to the Market:
   • The event study method requires an event that delivers new information to the market.
   • The market is expected to react very quickly to the release of relevant information.
   • Crucially, the event should be unanticipated by the market. If the event is anticipated, the information may already be incorporated into stock prices, and no significant abnormal return will be observed at the time of the announcement.
   • Examples of events studied using this method include corporate control changes, corporate refocusing, CEO turnover, divestiture, acquisitions, joint ventures, product recalls, customer service changes, and major legislation.
2. Outline a Theory that Justifies a Financial Response to This New Information:
   • Before conducting the study, you need a clear theoretical basis to explain why the event is expected to elicit a financial response from the market. This theory justifies using stock price performance as a measure of the event’s impact.
   • The theory should predict the direction and significance of the abnormal return. For example, does your theory predict a positive, negative, or neutral stock price reaction to the event?.
3. Identify a Set of Firms that Experience This Event and Identify the Event Dates:
   • Select the specific companies that were affected by the event you are studying.
   • For each firm in your sample, determine the precise event date. This is the day the information about the event became public or became known to the market. News items are often released the day before they appear in print, so the previous trading day might be considered the event day.
4. Choose an Appropriate Event Window and Justify Its Length:
   • The event window is the period around the event date during which you will examine stock returns.
   • The choice of event window length is one of the most crucial design issues.
   • For events that are released and processed very quickly, a short event window, such as 1 to 2 days around the announcement, is usually sufficient and can help capture the significant effect of the event while minimizing the influence of other factors. For example, the stock market reaction to firm-specific information has been found to adjust within 15 minutes of the release.
   • For less anticipated events, or when information leakage might occur prior to the announcement, a longer event window might be necessary to capture the full effect.
   • However, using long event windows greatly exacerbates the difficulty of controlling for confounding effects.
   • Regardless of the length, you must justify your choice of event window.
5. Eliminate or Adjust for Firms that Experience Other Relevant Events During the Event Window:
   • A major challenge is dealing with confounding effects – other events unrelated to your primary event that occur during the event window and could also affect stock prices.
   • If a firm in your sample experiences another significant event (e.g., an earnings announcement, a major lawsuit filing, a debt restructuring) during your chosen event window, this could contaminate your results.
   • Methods to address confounding effects include:
     • Eliminating firms that experience confounding events during the event window.
     • Adjusting for or controlling for the effects of these other events.
   • Failure to adequately control for confounding events can cast serious doubt on the validity of your empirical results.
6. Compute Normal Returns During the Event Window and Test Their Significance:
   • To determine if an event had an “abnormal” impact, you first need to estimate what the stock’s return would have been without the event. This is called the normal return.
   • The standard approach uses a market model based on historical stock prices to estimate normal returns. This involves regressing the firm’s stock return on the market’s return over an estimation period prior to the event. The market model is given by: Rit = αi + βiRmt + εit, where Rit is the return on firm i on day t, Rmt is the market return on day t, αi and βi are estimated parameters, and εit is the error term.
   • The abnormal return (AR) for firm i on day t is the difference between the actual return (Rit) and the estimated normal return from the market model: ARit = Rit - (αi + βiRmt).
   • Abnormal returns are typically aggregated over the event window to get cumulative abnormal returns (CAR). CARi = Σ ARit over the event window.
   • The significance of the abnormal returns or cumulative abnormal returns is then tested. The test statistic Z = ACAR * n^0.5 is commonly used, where ACAR is the average cumulative abnormal return across all firms and n is the number of firms. A statistically significant Z value suggests that the event had a significant impact on the values of the firms.
7. Report the Percentage of Negative Returns and the Binomial Z or Wilcoxon Test Statistics:
   • For certain situations, especially with small sample sizes, it’s important to report non-parametric tests.
   • The binomial Z test or the Wilcoxon signed rank test can be used to assess the significance of the number of positive or negative abnormal returns observed.
8. For Small Samples, Use Bootstrap Methods and Discuss the Impact of Outliers:
   • Small sample sizes are common in management literature using event studies. Statistical tests often rely on normality assumptions, which may not hold, especially with small samples.
   • For small samples, it is prudent to use “bootstrap” methods, which do not rely on normality assumptions. Bootstrap methods involve repeatedly resampling the data to create an empirical distribution of abnormal returns and test for significance based on this distribution.
   • You should also identify and discuss the impact of outliers (extreme abnormal returns). Outliers can be caused by noise or measurement error and can significantly influence results. Researchers should carefully assess whether the results are driven by outliers. While some researchers simply eliminate outliers, discussing their impact is crucial.
9. Outline a Theory that Explains the Cross-Sectional Variation in Abnormal Returns and Test This Theory Econometrically:
   • After determining the significance of the abnormal returns, the next stage often involves explaining why the abnormal returns vary across different firms.
   • This involves using a theory to identify firm-specific characteristics (e.g., size, diversification, industry) that might correlate with the size or significance of the abnormal return.
   • This theory is then tested econometrically, often using regression analysis, where the abnormal return (or cumulative abnormal return) is the dependent variable, and the firm-specific characteristics are the independent variables.
10. Report Firm Names and Event Dates in Data Appendix:
    • For transparency and to facilitate replication by other researchers, you should list the names of the firms included in your sample and the corresponding event dates in an appendix.

Conducting an event study properly requires careful attention to these design and implementation issues. Readers can be confident that the conclusions from an event study are valid only if they are confident that the researcher has truly isolated the abnormal returns associated with the event. This relies on the key assumptions that markets are efficient, the event was unanticipated, and there were no confounding effects.