As previously discussed, the analysis of the effects on AR and CAR are divided into two separate sections.
First, the total sample is reviewed, discussing Hypothesis 1. Following the initial results, we will present the results from Hypothesis 2, potentially unveiling similarities and differences across regions, enabling us to explain the aggregated effect of the total sample.
6.2.1 Hypothesis 1: Testing the significance of AR and CAR for the total sample
Hypothesis 1: The AR and CAR are not significantly different from zero in the symmetric event windows, [-1, 1],[-5,5] and [-10,10]
First, the parametric t-test of the cumulative abnormal returns shows the same patterns through all three event windows. The reaction in cumulative abnormal returns is significantly negative at a 1% level in the event windows [-1, 1] and [-5, 5], and at a 10% significance level for the event window [-10, 10]. Hence, when
including the total sample in the analysis, we find that announcements of mergers have a significantly negative impact on the stock price.
Event window CAR T-test: CAR = 0
CAR[-1,1] -0.39% -12.2377***
CAR[-5,5] -0.23% -3.5667***
CAR[-10,10] -0.19% -1.6768*
*p < 0.1 ; **p < 0.05 ; ***p < 0.01
Table 6.2: CAR reactions and parametric results on CAR for the total sample Source: Complied by authors
Second, from looking at the graphs in Figure 6.1, it can be seen that abnormal returns in the two largest event windows increase prior to the announcement, while all event windows show a drop in AR on the day of the announcement (Day 0), which is significant at a 95% confidence level. The reduction in AR moves CAR down to a new level before it continues to fluctuate.
Figure 6.1: Graphical illustration of the AR and CAR development for total sample Source: Complied by authors
In Appendix 7, you will find all average figures on AR and CAR for the total sample. The results portrayed in Appendix 7 show the interesting observation that the abnormal returns experience a significant decrease at the day following the announcement day, something that could be a result of delayed information flow regarding the merger announcement towards the market. From Appendix 7, there is also reported
statistics form the non-parametric sign test, showing a similar pattern in terms of increasing/decreasing returns as well as the pattern of significant abnormal returns. Hence, this shows further validation of the results obtained from the parametric t-test. Regardless of the movements within the three event windows, the main finding in the total sample implies announcement of mergers to produce significant adverse returns.
The overall findings in the total sample show the same consensus in all event windows, where the null hypothesis of AR and CAR not being significantly different from zero is rejected. Following, we will use the sample datasets of the different regions to examine whether there are differences in stock price reactions across geographical areas.
6.2.2 Hypothesis 2: Testing the significance of AR and CAR for each region
By dividing our total sample into three different region-specific samples, we hope to reveal similarities and differences in the stock price reaction across regions around the announcement date. Hence, we want to answer our second hypothesis:
Hypothesis 2: The significance or signs of AR and CAR is not dependent on geographical location
6.2.2.1 North America
When running the parametric test on CAR for the North American subset, we find similar patterns as those found for the total dataset. The reaction on CAR is significantly negative at the 95% confidence level for all event windows. Thus, isolating the effect of North American deals will not change the interpretation of CAR.
When announcing a merger in the US or Canada, the stock prices of the companies will, on average, have a significantly negative impact.
Event window CAR T-test: CAR = 0
CAR[-1,1] -0.88% -7.9269***
CAR[-5,5] -0.81% -3.6493***
CAR[-10,10] -1.62% -4.5699***
*p < 0.1 ; **p < 0.05 ; ***p < 0.01
Table 6.3: CAR reactions and parametric results on CAR for North America Source: Complied by authors
Turning to the effects on AR, it can be observed from the graphs in Figure 6.2 that the abnormal returns start dropping prior to the actual event, showing possible signs of information leakage towards the market.
Besides, by analyzing the day-by-day output from Appendix 8, we can observe a similar pattern as for the total sample, where we see a significant decrease in AR on the day after the announcement. The fact that the results are consistent with the total sample can be an indication that North America drives this effect in the total sample. This is supported by the fact that the North American dataset is the subset with the most observations (N=133). Thus, the results could have a great impact on the total sample. Furthermore, the results from the non-parametric test support the findings estimated by the parametric tests, further validating our results.
Figure 6.2: Graphical illustration of the AR and CAR development for North America Source: Complied by authors
The sample of North America shows significant abnormal returns at the 99% confidence level for the announcement day as well as the proceeding day, with the exception of the announcement day in the [-10,10] event window where it is significant at the 95% confidence level. Nevertheless, the announcements of North American deals have an overall significant adverse effect on the abnormal returns for companies engaging in M&A transactions.
6.2.2.2 Europe
When running the test on CAR on the European subset of the total sample, we find some interesting results, where the tests show opposite results compared to those found for the total sample. Thus, the reaction on CAR is significantly positive at a 99% confidence level for all event windows. As the European subset shows the opposite effects of the total sample and North America, we have to reject Hypothesis 2 that the significance and/or sign of the abnormal returns does not depend on the geographical location of the acquirer of the transaction. The implications of these findings will be further evaluated and explained in Section 7, Discussion.
Event window CAR T-test: CAR = 0
CAR[-1,1] 0.28% 10.4607***
CAR[-5,5] 0.71% 4.1488***
CAR[-10,10] 1.54% 5.4159***
*p < 0.1 ; **p < 0.05 ; ***p < 0.01
Table 6.4: CAR reactions and parametric results on CAR for Europe Source: Complied by authors
By examining the graphs in Figure 6.3, we can draw a consensus that changing the length of the event window will not have a significant impact on the results. By looking at the development of AR and CAR, it could seem like the companies engaging in merger activity tend to experience mostly positive returns around the day of the announcement, regardless of the length of the event window. When studying Appendix 9, this development is confirmed. The trend in Figure 6.3 shows that abnormal returns are significant on the day of the announcement and become even more significantly positive on the day following the announcement (with the exception of the [-10,10] event window which is only significant at a 90% confidence level the day after).
If we study the graphs and look at the day-by-day AR-figures in Appendix 9, we can observe signs of information leakage as the stock price experienced a steady increase already two days before the announcement date. Even though the AR-values prior to the event are insignificant, we observe a slight pre-announcement upwards drift in the security price, potentially resulting from leakage leading investors to drive the stock price upwards. This is consistent with Munk’s (2015) findings on event studies regarding mergers and acquisitions.
Furthermore, the non-parametric sign-test mostly confirms the results generated by our parametric t-test, which further confirms our results. Based on these findings, we can conclude that the overall results of the European subset indicate that a merger announcement on average has significantly positive abnormal
returns. As this finding deviates considerably from the overall findings from the total sample, the results will be carefully considered in Section 7, Discussion.
Figure 6.3: Graphical illustration of the AR and CAR development for Europe Source: Complied by authors
6.2.2.3 Japan
Due to a low sample size of the Japanese subset (N=29), the results yielded from these tests should be carefully interpreted. Thus, we will not add too much weight on the results generated by the Japanese transactions. Regardless of the lack of significant abnormal returns, Japan shows much of the same negative trend as the total sample and the North American subset. In the smallest event window [-1, 1] we find a significant negative effect at the 1% level.
Event window CAR T-test: CAR = 0
CAR[-1,1] -0.57% -3.8349***
CAR[-5,5] -0.13% -1.2752
CAR[-10,10] -0.93% -0.334
*p < 0.1 ; **p < 0.05 ; ***p < 0.01
Table 6.5: CAR reactions and parametric results on CAR for Japan Source: Complied by authors
When analyzing the graphs in Figure 6.4 as well as results in Appendix 10, it is hard to find a pattern of the movement in abnormal returns around the announcement date. The AR looks like it is fluctuating greatly in both the [-10,10] and the [-5,5] event windows. It does not show any specific reaction around the actual announcement date. However, even though the sample size is relatively small, it will together with the North
American sample, drive the abnormal returns of the total sample down at the day after the announcement, offsetting the positive effect of the European transactions. Furthermore, the findings of merger announcements having an insignificant effect on abnormal returns for the two longest event windows, could imply that Japanese investors within Telecom correctly expects the event to happen. Hence, according to the longest event windows, the Japanese market could be defined as efficient. However, the low sample size makes this particular evidence relatively weak. Further research should include a larger sample size to further validate these findings and draw robust conclusions.
Figure 6.4: Graphical illustration of the AR and CAR development for Japan
6.2.2.4 Partial Conclusion
Based on the parametric and non-parametric test results on AR and CAR for different event windows in the various regions, we have found some interesting results. At par with the total sample, the North American sample shows a significantly negative abnormal returns for all event windows, while Europe shows the opposite effect. Even though the results of the Japanese subset are in line with those of the total sample, we will not emphasize these results due to their weak significance and low sample size.