7. CONCLUSIONS AND RECOMMANDATION
4.4 E VENT S TUDY M ETHODOLOGY
The event study was introduced by Ray Ball and Philip Brown (1968) and Eugene Fama et al. (1969) in the late 1960s seminal studies. The methodology they introduced was essentially the same as that which is in use today, and has become the standard method of measuring security price changes in response to an event or announcement (Mackinlay 1997). This research method is widely used by a great number of researchers from different fields; some examples are: Palepu and Ruback (1992) used event study to investigate the post-merger performance for 50 largest U.S.
merged corporate between 1979 to mid-1984; Rau and Vermaelen (1998) used event study to investigate the long-term underperformance of bidding firms in mergers and tender offers listed on the NYSE and AMEX covered by both CRSP and COMPUSTAT in 1980 to 1991; Flouris and Swidler (2004) conducted an event study to analyze the impact of American Airline’s takeover of Trans World Airlines in 2001. Basically, there are two major objectives when conducting an event study: first, to test the efficient market hypothesis, to see how efficiently the market incorporates new information; second, to examine the wealth impact of an event (Sudarsanam 2003). The purpose in this dissertation is the latter of the two.
Usually, when this methodology is used, the focus is on short-horizon studies to measure the effects of an economic event (Sheng & Li 2000). For example, investigators can measure the abnormal return of shares in relation to an announcement of M&A, earnings, or issuing new debts, using short event windows of a few days around the event (Mackinlay 1997). However, in this dissertation, the investigation centers on the long-horizon event studies, where the post-event windows measuring post-merger performance of the collected data is up to five years (twenty quarters).
4.4.1 Conducting an Long Horizon Adjusted Event Study
Sheng and Li (2000) have generalized four steps for conducting an event study with a large amount of data of short-run share returns. These steps are:
Step 1, identifying the event
Step 2, evaluating the abnormal return
Step 3, testing the abnormal returns with statistical hypothesis testing Step 4, analyzing and explaining the results
In this dissertation, the event study is applied to each M&A deals in the long-run;
therefore, in step 2, the evaluation of the excess operating performance will be added in addition to the abnormal returns. The adjusted steps are:
Step 1, identifying the event
Step 2, evaluating the abnormal return and the excess operating performance Strep3, testing the abnormal returns and the excess operating performance with
statistical hypothesis testing using a Z-test Step 4, analyzing and explaining the result
Each step is described as below.
Step 1, identifying the event
The event that this dissertation refers to is the merger activities of the collected data.
The event window is defined by the effective date of legal completion of the merger.
The reason for choosing the completion date instead of the announcement date is that, the investigation focuses on the post-merger performance in the long-term rather than the short-term effect of announcement. The estimation windows are set to be two years (eight quarters) before the mergers, while the post-event windows are set to be five years (twenty quarters) after the mergers, so that the post-merger performance can be observed more completely. The event study windows are presented in figure 4.1.
Figure 4.1 Time Line for Event Studies Source: MacKinlay 1997
Step 2, evaluating abnormal returns and excess operating performance
In this dissertation, ‘abnormal returns’ refers to the difference between expected and actual returns, while ‘excess operating performance’ refers to the difference between expected and actual operating performance. The abnormal returns and excess operating performance will be evaluated with the formula based on the model generalized by Sheng and Li (2000). Figure 4.1 helps to describe the models.
Models used for evaluating abnormal returns and excess operating performance are mean-adjusted returns model and mean-adjusted operating performance model:
1. Mean-Adjusted Return Model (Model 1)
The abnormal return for a single firm is
∑
=where
AR
iE refers to the abnormal returns of a company i in each post-merger timeE,
R
iE refers to the actual returns of company i in each post-merger time E,Ti
refers to the expected returns of company i calculated by averaging the returns of that company in the pre-merger period betweent and
0 t1.The average abnormal return (AAR) of the all data firms in each time E in the post-merger periods is:
where N is the number of the total merging deals in the whole sample.
The cumulated abnormal return for a single firm i from the effective merger time until the time E after mergers is:
)
The average cumulated average abnormal return (ACAR) for all data firms from the effective merger time until the time E after mergers is:
)
where N is the total number of the merging companies, and E refers to each time of the post-merger period.
2. Mean-Adjusted Operating Performance Model (Model 2)
The excess operating performance for a single firm is
∑
=where
AP
iE refers to the excess operating performance of a company i in the post-merger period time E,P
iE refers to the actual performance of a company i intime the post-merger time E,
∑
Ti
refers to the expected performance calculated by averaging the performance of that company in the period betweent
0 and t1 before merger.The average excess operating performance (AAP) over all data firms in time E is:
∑
=where N is the number of securities in the whole sample with a return in event time E.
Step 3, testing the abnormal returns and excess operating performance with statistical hypothesis testing using a Z-test
After the calculation in the previous step, the AR, CAR, and AP of each time period will be calculated. Based on those results, AAR, ACAR, and AAP would also be calculated. In step 3, these calculated data in the post-merger window will be tested through statistical hypothesis testing using a Z-test.
When a Z-test is conducted, the researcher aims to determine if the difference between a sample mean and the population mean is large enough to be statistically significant. In this dissertation, the sample means refer to the AAR and ACAR when conducting the market assessment, AAP when conducting the operating performance assessment. On the other hand, the population mean refers to zero because AAR, ACAR, and AAP for the estimation windows should equal to zero. The null and the
alternative hypotheses for both assessments are as below:
¾ For the market assessment
AAR:
H : The average abnormal returns = 0
0H1: The average abnormal returns≠0
ACAR:
H : The cumulated average abnormal returns = 0
0H1: The cumulated average abnormal returns≠0
¾ For the operating performance assessment
AAP:
H : The average excess operating performance = 0
0H1: The average excess operating performance≠0
The p-value of the two-tailed Z-test is the criterion to judge if the sample means significantly differ from the population mean. When the p-value is small enough and
shows strong evidence, the null hypotheses can be rejected, while the alternative hypotheses can be inferred. On the contrary, if the p-value is large and shows no evidence, the null hypotheses cannot be rejected. The p-value of a test is described as below according to Keller & Warrack (2003):
P-value: 0~0.01 : Highly significant (Overwhelming Evidence) P-value: 0.01~0.05: Significant (Strong Evidence)
P-value: 0.05~0.1: Not Significant (Weak Evidence):
P-value: 0.1~1.0 : Not Significant (No Evidence)
Step 4, analyzing and explaining the result
The conclusion of the investigation is based on the results of the abnormal returns and the excess operating performance. By considering the abnormal returns of the shares and the excess operating performance using accounting data, the post-merger performance of the selected sample firms can be analyzed and explained.