Foundations of Investment Management
Abnormal Returns
Professor Christopher C. Geczy, PH.D.
Event Studies
3. Conducting An Event Study
• Step 1: Select many firms i=1,…,n having experienced a takeover
attempt
• Step 2: For each firm i, determine the day when the takeover attempt
(=the event) was made public. Call it ti=0
• Step 3: For each firm, use data prior to ti (t=-250,…,-16) to estimate the
market model parameters by regression: ai and bi
• This is called the estimation window
• Different window sizes are sometimes used, depending on the
situation (like +/- 30 days)
Event Studies
3. Conducting An Event Study (continued)
• Step 4: Use data reit for about 15 days before & after the event date ti
• For each day, define the residual eit to be the difference between the
actual return rit and the value of that return predicted by the market
model (estimated on the pre-event period in Step 3), aI + bireMt
eit = reit – {ai + bireMt}
actual return predicted by
return the market model
3
Event Studies
3. Conducting An Event Study (continued)
• eit is called the abnormal return on day t. It is the component of the
actual return that is not “in line” with the market model (=
unexplained by the market model).
Foundations of Investment Management
Abnormal Returns Graph
Professor Christopher C. Geczy, PH.D.
Abnormal Returns
Foundations of Investment Management
Average Abnormal Returns
Professor Christopher C. Geczy, PH.D.
Event Studies
3. Conducting An Event Study (continued)
• Step 5: Compute the average abnormal return (AR) over all firms:
Question: What do you expect a graph of ARt to look like?
• Expect ARt » 0 (maybe slightly negative) for t= -15,…, -1
• ARt > 0 for t=0 (the event date)
• And ARt » 0 for t=+1,…,+15
Foundations of Investment Management
CARs Graph
Professor Christopher C. Geczy, PH.D.
CARs
Foundations of Investment Management
Cumulative Abnormal Returns
Professor Christopher C. Geczy, PH.D.
Event Studies
3. Conducting An Event Study (continued)
• Step 6: Compute the cumulative abnormal return (CAR):
• This is the return obtained by buying a portfolio of firms 15 days before
a takeover attempt is revealed (t=-15), holding the stocks and selling
them on day T
Foundations of Investment Management
Cumulative Abnormal Returns: Part 2
Professor Christopher C. Geczy, PH.D.
Event Studies
3. Conducting An Event Study (continued)
• What do you expect CARt to look like?
• Expect CARt to increase slightly between t=-15 and t=-1 (the CAR
curve should be about flat)
• Then at t=0 CAR should jump because a large abnormal return
on that day is added to the cumulative abnormal return
• After the news are revealed (t=+2,..,+15), no abnormal returns
should keep appearing (i.e., the CAR curve should flatten out)
Foundations of Investment Management
MacKinlay
Professor Christopher C. Geczy, PH.D.
Event Studies
Example: Quarterly earnings announcements for the 30 firms in the Dow from
January, 1989 – December, 1993 from MacKinlay (JEL, 1997)
• Rationale: Markets should react to news in appropriate ways. If they do not,
then we reject various forms of market efficiency
• 600 announcements
• Data collected
• Announcement date
• Actual earnings
• Average expected earnings from I/B/E/S from last month of previous
quarter
Event Studies
• Categorize the announcement base on ∆ = [actual-forecast]/forecast as
Good News (∆ > 2.5%; 189), Bad News (∆ < -2.5%; 238), or
No News (-2.5% ≤ ∆ ≤ 2.5%; 173)
• Windows
• 41 day event window (20-1-20)
• 250 day estimation window
• Market Model results presented here
Results: AR’s and CAR’s
Results: CAR’s
Foundations of Investment Management
Key Takeaways
Professor Christopher C. Geczy, PH.D.
Event Studies
Key Takeaways:
• Earnings announcements clearly convey information useful for
valuation
• Strongly reject the null hypothesis of no impact on the
announcement date for Good and Bad News but not No News
announcements
• Also reject the null for the day after the announcement date – Why?
• Market gradually learns about the forthcoming announcement – pre-
announcement drift
Foundations of Investment Management
Event Study Issues
Professor Christopher C. Geczy, PH.D.
Event Studies
• Event Study Issues
1. Instead of the market model, we can use any model for stock returns
(e.g., a multi-factor model)
• There is no guarantee that the result of the event study conducted
with the market model and the index model will be similar
• Among index models, the number and choice of the factors is also
likely to matter
• Matters for longer-term studies than shorter-term studies. Why?
• Over very short periods of time, the market model typically is
passable
Event Studies
• Event Study Issues (continued)
2. This creates an important difficulty for event studies or more generally tests
of market efficiency
• The Joint Hypothesis Problem: All tests of market efficiency are joint
tests of:
• Efficiency
• The presumed model of market equilibrium (CAPM, a given APT, etc.)
Any rejection of the first may be due to the latter
• Focus on ‘statistical significance’ may be misguided if the model or
procedure generates powerful tests. Economics must guide
interpretation: consider economic significance.
Foundations of Investment Management
Market Efficiency and Real Time
Professor Christopher C. Geczy, PH.D.
Examples on Information and Markets
• How quickly is information incorporated into prices
• Case Study: CNBC’s Morning and Midday Calls (Busse and Green,
2002) Video
a. Note that this example represents both sides of the EMH: quick
information incorporation and the possibility of making money.
b. Morning Call, 11:05 - 11:10 a.m. EST
• Usually < 2 minutes in duration
• Mack Consuelo hosts
c. Midday call, 2:53 – 2:58 p.m. EST
• Also usually < 2 minutes in duration
• Maria Bartiromo hosts
Examples on Information and Markets
d. Named after analysts call
e. Topics are macro news, firm-specific or group news, initiations of
coverage, information about earnings releases or upcoming analyst
meetings, etc.
27
Examples on Information and Markets
f. Sample: Daily observations, June 12 – October 27, 2000 (20
weeks)
• S&P500 loss 4.5% and NASDAQ lost 15.4%
• 322 stock reports on 84 trading days
• Record information
• Identify of firm and time of mention
• Stock price if disclosed
• Whether price chart is shown
• Changes in DJIA and NASDAQ Composite (shown on screen
concurrently)
• Disposition of report (+/-)
Basics
Foundations of Investment Management
CNBC Morning and Midday Calls: Recap
Professor Christopher C. Geczy, PH.D.
Foundations of Investment Management
Price Response
Professor Christopher C. Geczy, PH.D.
Price Response
Speed of Reaction
Foundations of Investment Management
Profitability Graph
Professor Christopher C. Geczy, PH.D.
Profitability
Foundations of Investment Management
Characteristics of Price Impact of CNBC Reports
Professor Christopher C. Geczy, PH.D.
Table 3 (Cross-X Determinants)