Professional Documents
Culture Documents
To cite this article: William A. Reese Jr. & Russell P. Robins (2017) Performing an event study:
An exercise for finance students, The Journal of Economic Education, 48:3, 206-215, DOI:
10.1080/00220485.2017.1320603
Article views: 12
ECONOMIC INSTRUCTION
ABSTRACT KEYWORDS
This exercise helps instructors teach students how to perform a simple event Event study; experiential
study. The study tests to see if stocks earn abnormal returns when added to learning; S&P
the S&P 500. Students select a random sample of stocks that were added to JEL CODES
the index between January 2000 and July 2015. The accompanying spreadsheet A; A; G
calculates cumulative abnormal returns and cumulative abnormal trading vol-
ume and plots them in separate graphs. Students are asked to analyze the data
and draw conclusions. Through this exercise, students learn how to conduct an
event study and determine if a statistically significant event has occurred.
Motivation
Investors who seek to earn unusual (abnormal) positive returns through their investments in common
stocks are often looking for news and events that will influence a stock’s price. The semi-strong form of
the efficient market hypothesis (Fama 1970) supposes that as a result, this information is incorporated
into a stock’s price almost as soon as it becomes public. Event studies are used as a formal statistical test
to determine if and when new information causes a price change in stocks. Over the years, researchers
have conducted hundreds of event studies, some of which have led us to conclude the following:
r After their first day of trading, the stock price of a typical IPO declines over the next three years
(Ritter 1991).
r There are various types of momentum in stock returns (Jegadeesh and Titman 1993; Gutierrez and
Kelley 2008).
r When an analyst upgrades a stock, its price usually goes up (Womack 1996).
r Investors tend to put more money into mutual funds that have recently done well (Frazzini and
Lamont 2008).
While each of these (and numerous additional) facts of finance have been found by academic
researchers through event studies, we frequently do not show undergraduate and MBA students the
methodology for how they are uncovered. Of course, most event studies are complex enough that the
average undergraduate or MBA student would find it challenging to read the research paper, much less
replicate the study. Additionally, gathering sufficient data for an event study to show statistical signifi-
cance with the results is a daunting task for most of these students.
This article helps faculty to teach students how the scientific method (in this case, an event study)
is used to test a hypothesis related to market efficiency while giving students the experiential learning
opportunity of performing it themselves. It also allows each student in the class to conduct the same
study with a somewhat different set of data. This lesson is easily incorporated into an undergraduate or
MBA investments class, a business economics class at a liberal arts college, or an econometrics class (as
a practical application of linear regression). Importantly, students will not get bogged down with data
CONTACT William A. Reese wreese@tulane.edu Professor of Practice, A. B. Freeman School of Business, Tulane University,
New Orleans, LA , USA.
Color versions of one or more of the figures in this article can be found online at www.tandfonline.com/vece.
© Taylor & Francis
PERFORMING AN EVENT STUDY 207
collection and input, thanks to a spreadsheet that we have developed. This allows students more time to
focus on the methodology behind the event study and a greater opportunity to analyze the results.
Our event study allows students to test if the addition of a stock to the S&P 500 index causes a (sta-
tistically significant) positive abnormal return for that stock and to theorize if investors can profit from
it. It has been our experience that when students randomly select a reasonably large (50 or more) sam-
ple of S&P additions, they will find that there are not only statistically significant positive abnormal
risk-adjusted returns when these stocks are added to the index, but that they can visually observe this
through a standard graph of cumulative abnormal returns (CARs). Because transactions costs can be
difficult to quantify, it is not as clear that investors can profit from this knowledge, but it usually appears
that they might have that ability.
This unique opportunity for students to conduct an event study is made possible through an Excel
spreadsheet we have developed.1 The spreadsheet contains a macro that looks up price and volume data
for each stock the student selects along with index values for the S&P 500 from Yahoo! Finance. The
spreadsheet then calculates expected returns for each stock using the market model and determines
the CARs for the 10 days before, day of, and 10 days after it is added to the index. These CARs are
automatically graphed, allowing students the opportunity to visually see the results.
will be using and the graph gives the instructor an opportunity to discuss movements before the event
day, on the event day, and after the event day. Some additional event studies we show include one that
examines how a stock reacts to being mentioned favorably or unfavorably on CNBC. This allows us to
point out that event windows can be as small as a few minutes. A somewhat humorous (in retrospect)
event study was done by the Wall Street Journal looking at abnormal returns surrounding the addition
of “.com” to a stock’s name between January 1998 and April 1999. We suggest that instructors try to find
a wide variety of event studies to show the students. Including one or two that the instructor has done
through his/her own research will certainly be well-received.2
An excellent resource is Fehrs (1990), who demonstrates how to teach students some of the same
principles we do. He also uses the market model to show students how to calculate abnormal returns.
However, instead of asking students to determine if there are abnormal returns surrounding an event,
he has his students first identify abnormal returns and then try to find the event(s) that caused it.
The final step we take in our classes when teaching students about event studies prior to assigning
them this exercise is to work through the data and calculations of a simple event study in class. For this,
we have prepared an event study of how airline stocks reacted to the terrorist attacks of September 11,
2001. Because the effect is so obvious, we only need three stocks in our sample (Continental, Delta, and
Southwest) to demonstrate statistical significance. We use the market model to adjust for risk (pointing
out that we are looking at negative returns for these airline stocks that are greater than those for the mar-
ket as a whole), calculate CARs, t-statistics, and graph the results. We do not make use of the spreadsheet
with the macro for this demonstration, but instead use this opportunity to clearly show the students each
step in the process and how the results can be both statistically and visually interpreted.
Once the class has been taught what an event study is, has seen some examples from published
research, and has seen a demonstration of a simple three-stock event study dealing with an event familiar
to them, the students are ready to do this exercise with a solid understanding of what is being done, why
it is being done, and how to interpret the results.
At this point, each student is ready to select a random sample from the list of stocks that were added
to the S&P 500 index and “run” the event study. After they do this, it is worthwhile to have them answer a
series of questions that are designed to measure their understanding of the exercise. Appendix 2 contains
the full set of instructions we give to the students along with the questions they are asked to answer.
regression to find statistics that can be used to forecast expected returns. One of the simpler ways to do
this, which can be done automatically by the spreadsheet, is with the market model. For this event study,
our event window is day −10 to day +10, that is, the 10 trading days prior to the stock’s addition to
the index, the day of its addition (which is day 0), and the 10 trading days after its addition. We use the
returns for days −110 to −11 (the 100 trading days just prior to the beginning of the event window) to
find estimates of each stock’s alpha and beta, where the stock is the dependent variable, and the market
(S&P 500) is the independent variable.
We use a 100-day estimation period because Armitage, in his survey of event study literature, sug-
gested, “100 days or more seems safe” (Armitage 1995). Once those estimates are calculated, they are
applied to the market’s returns in each of the 21 days in the event window to calculate the “expected”
returns for the stock according to the following formula: E(R) = α + β(RMkt ). For each of the 21 days
in the event window, we have an actual return for the stock and a risk-adjusted expected return for that
stock. Each day’s risk-adjusted abnormal return is simply the actual return for that trading day minus
the expected return for that trading day. CARs are the sum of each daily abnormal return thus far in the
event window. For example, the CAR on the event day will be the sum of the 11 abnormal returns found
in days −10 through 0.
Statistical significance
For each day in the event window, the CARs are averaged across the sample with both an arithmetic mean
and a sample standard deviation being calculated. Standard errors are calculated as the sample standard
deviation divided by the square root of the sample size (the number of S&P 500 additions the student
selected). The spreadsheet then calculates t-statistics for each day in the event window as the mean CAR
for that day divided by the standard error.
We should emphasize that everything mentioned above is done for the student by the Excel spread-
sheet. All the student does is type the ticker symbols for the stocks he/she selects and click the “Down-
load” button on the spreadsheet. Although the students do not do the calculations themselves, we have
taken great pains to ensure that the students can observe how the calculations are done. This allows the
instructor to use the spreadsheet as a learning tool for how to use linear regression to compute expected
returns with the market model, how to find abnormal returns, and how to test for statistical significance.
The graph
Probably the most important piece of this exercise is the final graph that is produced where the average
CARs are plotted on the vertical axis, and the days in the event window (−10 through +10) are plotted on
the horizontal axis. Although each student’s graph will look somewhat different due to different random
samples of stocks, we have found that almost every student’s graph will show a steady increase in CARs
from day −10 until day 0. After day 0, the CARs will decline somewhat, but by day +10, they will have
declined only about halfway from their peak at day 0. Visually, students will clearly be able to see the
effect of being added to the S&P 500, and why event studies can be such a powerful research tool. Figure 1
shows a screenshot of how the spreadsheet typically looks once the average CARs and t-stats have been
calculated, and the graph has been drawn:
Trading volume
It is not unusual for a stock to have a five-fold increase in trading volume on the day it is added to the
S&P 500. Although it is difficult for an investor to profit from an increase in trading volume, this increase
is worth noting in this exercise as it lends support to the CARs and gives the student another example of
an event study.
When our macro obtains price data from Yahoo! Finance for each stock the student selects, it also
records the daily trading volume. Our spreadsheet calculates the average daily trading volume for each
210 W. A. REESE AND R. P. ROBINS
stock from days −110 through −11. Once the event window begins on day −10, that day’s trad-
ing volume is measured against the average daily trading volume in days −110 through −11 to reg-
ister the stock’s abnormal trading volume. Cumulative abnormal trading volumes (CAVs) are calcu-
lated in the same manner as CARs. CAVs are averaged across the sample, and standard errors are
calculated. Calculations of t-statistics for trading volume are done the same way they are for returns.
The average CAVs are also graphed against the days of the event window just as the average CARs
were.
Typically, students will be able to observe a slight increase in trading volume about five days prior to
a stock’s inclusion in the S&P 500. This is probably due to the trading activity of frontrunners, but the
big story occurs on day 0 when the index funds must purchase shares of the stock. Average CAVs will
often shoot up to well over 1,000 percent. They typically maintain that (cumulative) level for days +1
through +10, indicating that daily trading volume quickly returns to normal levels. Figure 2 shows a
screenshot of how the spreadsheet looks once the average CAVs and t-stats have been calculated, and the
graph has been drawn.
Notes
1. The spreadsheet for this assignment can be found at https://breese7160.tulane.edu/event-study/.
2. There are a number of resources on the Internet that can be used along with this assignment to help instructors
teach students how to conduct an event study. www.eventstudytools.com/event-study-methodology demonstrates
how to do an event study in a YouTube video. www.eventstudymetrics.com/index.php/event-study-methodology
gives students some details on the statistics that are used. http://bauer.uh.edu/rsusmel/phd/lecture%206.pdf
includes a lecture that instructors can modify to teach students how to do an event study. http://web.mit.edu/
doncram/www/eventstudy.html contains some excellent sources for event study methodology, although each was
written prior to 1998.
212 W. A. REESE AND R. P. ROBINS
References
Armitage, S. 1995. Event study methods and evidence of their performance. Journal of Economic Surveys 8 (4): 25–52.
Fama, E. 1970. Efficient capital markets: A review of theory and empirical work. Journal of Finance 25 (2): 383–417.
Fehrs, D. 1990. Management decisions, market events, and stock price changes: A student project for finance courses.
Journal of Financial Education 19: 5–9.
Frazzini, A., and O. Lamont. 2008. Dumb money: Mutual fund flows and the cross-section of stock returns. Journal of
Financial Economics 88 (2): 299–322.
Gutierrez, R., and E. Kelley. 2008. The long-lasting momentum in weekly returns. Journal of Finance 63 (1): 415–47.
Jegadeesh, N., and S. Titman. 1993. Returns to buying winners and selling losers: Implications for stock market efficiency.
Journal of Finance 48 (1): 65–91.
Keown, A., and J. Pinkerton. 1981. Merger announcements and insider trading activity: An empirical investigation. Journal
of Finance 36 (4): 855–69.
Ritter, J. 1991. The long-run performance of initial public offerings. Journal of Finance 46 (1): 3–27.
Womack, K. 1996. Do brokerage analysts’ recommendations have investment value? Journal of Finance 51 (1): 137–61.
should give you returns for the 10 days prior to the event date, the event date, and the 10 days
after the event date. This is our event window.
(8) For each day in the event window, in Excel, multiply the beta estimate you calculated above times
the return for the S&P 500 for that day and then add your alpha estimate. This is the “risk-adjusted
expected return” for the stock for that day. Based on that stock’s sensitivity to the S&P 500 during
the prior 100 days, this is the return we would expect the stock to have on that day if nothing
unusual happened to it. Calculate the expected return for your stock for each of the 21 days in
the event window.
(9) For each day in the event window, subtract your stock’s risk-adjusted expected return from its
actual return. This is its “abnormal” return for that day. We interpret this as the return that (per-
haps) was influenced by the event we are looking at.
(10) For each of the 21 days in the event window, calculate the “cumulative abnormal return” (CAR)
for your stock. For day 1, the CAR is the day 1 abnormal return. For day 2, the CAR is the day 1
abnormal return plus the day 2 abnormal return. For day 3, the CAR is the sum of the abnormal
returns for days 1–3. For day 4, the CAR is the sum of the abnormal returns for days 1–4, etc.
(11) Select additional stocks that were added to the S&P 500 Index and repeat each of the above steps
with each of them. There is no specific number of stocks that you should select. The number of
stocks that you select is your “sample size.” Generally, a larger sample size is better, but you must
consider how much time it takes you to collect the data for a larger sample size.
(12) Once you have CARs for each of the event days for each stock in your sample, find the average
(arithmetic mean) CAR for each day in the event window for your sample of stocks. Calculate
the standard deviation of the CARs for each day as well.
(13) A t-statistic tells you how many standard errors a test statistic is from the null hypothesis that
you want to test. In our case, the null hypothesis is that nothing unusual is happening to these
stocks during the event window (meaning that the CARs are zero). So for each day in the event
window, you should divide the average CARs you calculated by the standard error for that day.
The standard error is the standard deviation (for that day) for your sample divided by the square
root of the sample size. Thus, a larger sample (all else equal) results in a smaller standard error
and a larger t-statistic.
(14) Researchers generally consider a t-statistic with an absolute value greater than 2.0 to indicate that
there is only a small chance that there is nothing going on here (that the CARs are actually zero
for the entire population of stocks that were added to the S&P 500). If the t-stat is greater than 2.0,
we fail to reject our hypothesis that being added to the S&P 500 is (on average) causing a stock’s
return to be different from what it otherwise would have been.
(1) Lookup the data on the IndexChange_500 tab for each of the stocks you selected and note the
date it was added to the index (row A), the date for 113 trading days before it was added to the
index (row D), and the date for 10 trading days after it was added to the index (row E).
(2) A macro embedded in the spreadsheet will go to Yahoo! Finance’s Web site and find the adjusted
closing prices and trading volume for each of the 113 trading days before the event (the day the
stock was added to the index), the day of the event, and the 10 trading days after the event. It will
do this for each stock you selected as well as for the S&P 500 (our proxy for the market).
(3) The Excel spreadsheet will use these adjusted closing prices to calculate daily returns for each
stock for the 112 days prior to the event, the day of the event, and the 10 days after the event.
(4) Using the daily returns from day −112 to day −11, the spreadsheet will run separate regressions
for each stock where the dependent variable is the returns for the stock and the independent vari-
able is the returns for the S&P 500. Estimates of alpha and beta for each stock will be calculated.
(5) Using those estimates of alpha and beta along with the actual returns for the S&P 500, the
spreadsheet will calculate expected daily returns for each stock for each of the 21 days in
the event window (10 days before the event, the day of the event, and 10 days after the
event).
(6) Actual returns for the stock that are greater than its expected return for any day in the event win-
dow are considered to be positive abnormal returns. Actual returns that are less than the expected
return are negative abnormal returns. Cumulative abnormal returns (CARs) are calculated across
the event window for each stock.
(7) The CARs are averaged over the all the stocks you randomly selected for each day in the event
window. Standard errors and t-stats are calculated. The average CARs and the t-stats can be seen
on the “Final” tab. The average CARs are then displayed in a graph as a function of the day in the
event widow, giving you a nice visual picture of what happened to these stocks surrounding the
date they were added to the S&P 500.
(8) Additionally, the daily trading volume for each stock during the 100 trading days prior to the
event window are averaged together to create an “expected volume” for each stock. For each day
in the event window, the expected volume for each stock is subtracted from the stock’s actual
trading volume on that day and scaled by the expected volume to give us a measure of the stock’s
abnormal volume for that day. Cumulative abnormal volumes are calculated just like CARs. They
are then averaged across all the stocks you selected, standard errors are calculated, and t-stats are
found—just as they were for the CARs. The CAVs are also graphed as a function of the day in the
event window—just like the CARs.
The spreadsheet will take a couple of minutes to do all this. When it has completed all its calculations,
you should take a few minutes to look at the cumulative data and the graphs on the “Final” tab as well as
the data for each stock you selected in the “RESULTS” tab.
If you run into any problems, simply close out the spreadsheet and reopen a fresh one from the Web
site. If you found that the spreadsheet seemed to have problems with the data from a particular stock,
choose a different one this time.
You should answer the following questions and turn them in as instructed:
(1) Prior to using the spreadsheet in this exercise, what did you expect to happen to the returns and
trading volume for these stocks on and around the date that they were added to the S&P 500?
(2) In 2–3 paragraphs, summarize what the spreadsheet has done.
(3) Does your event study show that there are statistically significant abnormal returns associated
with a stock being added to the S&P 500? If so, how would you convince someone that these
abnormal returns came from the stock’s addition to the index and not from something else?
(4) Answer question #3, for trading volume instead of returns.
(5) If you heard today that stock XYZ is to be added to the S&P 500 Index next week, what sort of a
market reaction do you expect to see? Explain what you expect to see between now and the date
of the addition, on the date of the addition, and during the ten days after it is added to the index.
(6) How certain are you that the reactions you “expect” to see from the previous question will happen
for stock XYZ?
PERFORMING AN EVENT STUDY 215
(7) Online trading can be done for as little as $8.00 per trade. Considering transactions costs (com-
missions and bid-ask spreads), do you think you can make money from what you have found in
this exercise? If yes, describe how you would do it and how much you think you can make. If not,
explain why not.
(8) Based on the method we used in this exercise, describe how an “abnormal return” is measured.
(9) Come up with an original idea for an event study and describe how you would do it.