007-paper 1
TRANSCRIPT
-
8/2/2019 007-Paper 1
1/22
-
8/2/2019 007-Paper 1
2/22
Introduction
This paper focuses on daily stock returns andexcess returns and how the particularcharacteristics of these data affect event study
methodologies for assessing the share priceimpact of firm-specific events.
This paper is actually an extension of eventstudy methodologies used with monthlyreturns by the same writers BROWN andWARNER in 1980.
-
8/2/2019 007-Paper 1
3/22
Introduction(Continue..)
Previously for monthly returns a simple
methodology based on the market model is
both well-specified and relatively powerful
under a wide variety of conditions, and in
special cases even simpler methods also
perform well.
-
8/2/2019 007-Paper 1
4/22
Introduction(Continue..)
Daily and monthly data differ in potentially
important respects.
Non-normality
Non-synchronous trading
Variance estimation
-
8/2/2019 007-Paper 1
5/22
1.Non-Normality
Daily returns are non-normal depart frommonthly data as it was not observed in monthly.
The evidence generally suggests that
distributions of daily returns are fat-tailedrelative to a normal distribution [Fama (1976, p.21)].
Since event studies generally focus on the cross-
sectional sample mean of security excessreturns, this paper examined the small sampleproperties of the mean excess return.
-
8/2/2019 007-Paper 1
6/22
Non-Normality(Conti..)
The Central Limit Theorem[Billingsley (1979,
pp. 308-319)] guarantees that if the excess
returns in the cross-section of securities are
independent and identically distributed
drawings from finite variance distributions,
the distribution of the sample mean excess
return converges to normality as the numberof securities increases.
-
8/2/2019 007-Paper 1
7/22
2.Non-synchronous trading
When the return on a security and the return
on the market index are each measured over a
different trading interval, ordinary least
squares (OLS) estimates of market model
parameters are biased and inconsistent.
With daily data, the bias can be severe
(Scholes and Williams (1977, p. 324), Dimson(1979, p. 197)].
-
8/2/2019 007-Paper 1
8/22
Non-synchronous trading(Conti)
Concerned with that problem, authors of
event studies with daily data have used a
variety of alternative techniques for
parameter estimation [e.g., Gheyara and
Boatsman (1980, p. ill), Holthausen (1981, p.
88)].
-
8/2/2019 007-Paper 1
9/22
3.Variance Estimation
With both daily and monthly data, estimation
of the variance of the sample mean excess
return is important for tests of statistical
significance.
Time-series properties of daily data.
Cross-sectional dependence of the security-
specific excess returns.
stationary of daily variances.
-
8/2/2019 007-Paper 1
10/22
Experimental Design
Sample construction
250 samples of 50 securities.
Simple random sampling and somehow
convenient sampling is used.
Hypothetical dates of events are assigned to
selected securities.
Time period is of July 02, 1962 to December 31,1979.
-
8/2/2019 007-Paper 1
11/22
Experimental design(Conti.)
Sample Construction..
Maximum of 250 daily observations are taken.
For a security to be included in a sample, it must
have at least 30 daily returns in the entire 250 day
period, and no missing return data in the last 20
days.
Estimation Period = 239 days Event Period = 11 days
-
8/2/2019 007-Paper 1
12/22
Experimental design(Conti.)
Excess return measures
Let Ri,t designate the observed arithmeticreturn for security i at day t. Define A, as the
excess return for security iat day t.
Mean adjusted returns
A i,t= Ri,ti,t
-
8/2/2019 007-Paper 1
13/22
Experimental design(Conti.)
Excess return measures
While Rm,tis return on CRSP equaly wieghtedindex for day t.
A i,t= Ri,t Rm,t
OLS market model.
A i,t= Ri,t -Rm,tWhere and are OLS values from the estimation periods.
-
8/2/2019 007-Paper 1
14/22
Summary and Conclusions
Using simulation procedures with actual dailydata, the paper investigates the impact of a
number of potential problems of concern in
the literature. These include Non-normality and the properties of tests
Alternative procedures for market model
parameter estimation Variance estimation: Autocorrelation, cross-
sectional dependence, and variance increases
-
8/2/2019 007-Paper 1
15/22
Non-normality and the properties of tests
The non-normality of daily returns has no
obvious impact on event study
methodologies.
Although daily excess returns are also highly
non-normal, there is evidence that the mean
excess return in a cross-section of securities
converges to normality as the number ofsample securities increases.
-
8/2/2019 007-Paper 1
16/22
Non-normality and the properties oftests(Conti.)
Standard parametric tests for significance of
the mean excess return are well-specified.
In samples of only 5 securities, and even when
event days are clustered, the tests typically
have the appropriate probability of Type I
error.
-
8/2/2019 007-Paper 1
17/22
Non-normality and the properties of tests(Conti.)
As in the case of monthly data, the conclusionthat the methodologies are well-specifiedapplies to excess returns measured in a variety
of ways, including Market Adjusted Returnsand the OLS market model.
With daily data, these two methodologieshave similar power, and, as expected, thepower of each is much greater with daily thanwith monthly data.
-
8/2/2019 007-Paper 1
18/22
Non-normality and the properties of tests(Conti.)
Market Adjusted Returns and the OLS marketmodel also outperform a simpler MeanAdjusted Returns procedure, which has low
power in cases involving event-date clustering. In addition, exchange-listing is an important
correlate of the power of the various tests,with samples of NYSE securities exhibitingdramatically higher power than AMEXsecurities.
-
8/2/2019 007-Paper 1
19/22
Alternative procedures for market
model parameter estimation
Procedures other than OLS for estimating the
market model in the presence of non-
synchronous trading convey no clear-cut
benefit in detecting abnormal performance.
The specification and power of the actual tests
for abnormal performance is similar to that
obtained with the OLS market model.
-
8/2/2019 007-Paper 1
20/22
-
8/2/2019 007-Paper 1
21/22
Variance estimation (Conti.)
Tests which assume non-zero cross-sectionaldependence are only about half as powerful
and usually no better specified than those
employed assuming independence. Finally, we illustrated how variance increases
can cause hypothesis tests using standard
event study procedures to becomemisspecified.
-
8/2/2019 007-Paper 1
22/22
Variance estimation (Conti.)
Several procedures to deal with the possibilityof variance increases were outlined.
However, further research is necessary to fully
understand the properties of alternativeprocedures for measuring abnormal
performance in such situations.