5. Cumulative Average Abnormal Returns Study
5.3. Presentation of Our Results
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
chosen. We decided then to choose not to test our CAAR results a second time for the above reason as the parametric t-test appears to be the most robust one, even in case of skewness of the normal distribution curve of our sample.
5.3. Presentation of Our Results
Our results will be organized into different parts, according to the different factors we want to take into account in our study. The first part will be dedicated to focus on each index we selected for every defined window: CAC 40, CAC Mid 60, TAIEX, FTSE TWSE Taiwan 50 index, and FTSE TWSE Mid-cap 100 index. We will then be able to focus on the impact of the presidential elections for each of the index and extend our analysis making comparison between France and Taiwan, the two countries of our study.
To be able to easily interpret the significant values at different level of confidence, we will use the following notation on the t-test results we got after calculations:
17- * CAAR value significant at 90% of confidence 18- ** CAAR value significant at 95% of confidence 19- *** CAAR value significant at 98% of confidence 20- **** CAAR value significant at 99% of confidence
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
CAC 40 CAC Mid 60
Event Window CAAR t-test CAAR t-test
(+1,28) 0,066% 0,013623959 -3,673% -0,799249425
(-7,-1) -0,638% -0,320534419 -1,189% -0,630604275
(-14,-8) -0,571% -0,319720504 -0,023% -0,01491482
(-28,-15) 1,777% 0,502069069 -2,039% -0,994878891
(-28,28) 1,262% 0,158929617 -6,244% -0,933220342
Figure 6: CAAR Results and Significance Test For France Indices
Our first results focus on the French indices we work with: The CAC 40 and The CAC mid 60.
Regarding the CAC 40, we can say that the CAAR (cumulative average abnormal returns) are positive for the period (-28,-15) and (+1,+28) while they appear to be negative for the two weeks period prior to the presidential elections (-14,-8), (-7,-1). Those results are not significant at 90% and 95% confidence levels; the different levels of confidence that interest us to work with in this study. Indeed, the CAAR t-test results are not above (or under for the case of negative abnormal returns) the critical values of significance for N-1=4 degrees of freedom on the t-distribution table. It is then impossible to reject the null hypothesis H0 stating the CAAR are equal to 0. We cannot show in this case that the CAARs are or significantly different from 0 (positive or negative). Even if those results are not significant, they show a trend of how the CAC 40 reacted around the presidential elections that happened in France since 1988. If we take the period (-28,+28), the period including the month prior and the month after the elections (all the windows defined in our study) , it shows that the CAAR are positive (1,262%).
Focusing on this trend of positive CAARs, it is possible to say that the periods of the 15 first days of the presidential campaign (-28,-15) and the month after the elections (+1,+28) show
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
positive CAAR according to our results. Nevertheless, they are not significant at both level of confidence 90% and 95%. On the other side, the closest windows to the election, (-14,-8) and (-7,-1) do not show a trend of positive abnormal returns as Pantzalis, Stovall and Turtle found for the United States. The CAARs appear to be negative for these two periods, but again, not significantly. Then, on the opposite way to the findings of Pantzalis, Stovall and Turtle, the uncertainty does not seem to be resolved the closest we are from the election day.
For the mid-cap index of France, the CAC-mid 60, our results show negative CAARs for all our periods of study, the most important result being after the month after the presidential elections (+1,+28) with a CAAR of -3,673%. Overall, the period (-28,+28), which includes all the windows we defined, shows a negative CAAR of -6,244%. Nevertheless, as for the CAC 40, none of the results observed for the CAC-mid are significant at the 90% and 95% levels of confidence. The results allow us to observe a trend for the CAARs of the CAC-mid around the presidential elections in France, since 2002, but cannot lead us to significant conclusions. We cannot reject the null hypothesis stating that the CAAR is equal to 0 for each window of our study. The value of the CAAR (-6,244%) on the overall period seems relatively far from 0 compared to the value found for the CAC 40. The non-significance of the value might be explained be a higher volatility of this index. As for the CAC 40, the trend observed on the CAC-mid the three weeks before the election date does not match the finding of Pantzalis, Stovall and Turtle in 2000 as the CAARs results are negative in our case (But not significantly negative).
‧
CAAR t-test CAAR t-test CAAR t-test
(+1,28) 7,053% 1,278805656 13,298% 2,143147765 4,309% 1,853327154 (-7,-1) 0,116% 0,024627426 2,783% 2,010073206 2,641% 1,663453795 (-14,-8) -1,059% -0,432993098 -1,007% -0,377412603 -2,378% -0,944155674 (-28,-15) 1,186% 0,243380165 6,682% 6,052251104 4,507% 7,870503549***
(-28,28) 7,296% 0,568785059 21,757% 10,16627679* 9,079% 1,859640643
Figure 7: CAAR Results and Significance Test for Taiwan Indices
The second part of our results will focus on Taiwan with the three indices of our study: TAIEX, FTSE TWSE Taiwan 50 index and FTSE TWSE Mid-cap 100 index. For the TAIEX first, the most important result appears to be the month after the presidential elections (+1,+28) with a positive CAAR of 7,053%. Overall, the period (-28,+28), which includes all the windows we defined, shows a positive CAAR of 7,296%, mostly because of the result found for the month after the presidential election date. Indeed, other results found show values of CAAR which do not deviate much from 0. This is confirmed by the CAAR t-test, conducted to test the significance of these results. None of them is found significant at 90% and 95% level of confidence. More surprisingly, the outcome is the same for the CAAR found for the period (+1,+28). The value of the t-test conducted for this period 1,27881 does not exceed the critical value 2,132 that shows significance at 90% level of confidence for N-1=4 degrees of freedom.
The null hypothesis cannot be rejected for all the CAARs found for the TAIEX.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The FTSE TWSE Mid-cap 100 index will show us different results. In fact all the CAARs found for each event window of our study are positive; expect the one for the period (-14,-8) which has a negative value of -1,007%. As it is possible to see, all the positive CAAR values seem to deviate strongly from 0. The CAAR t-test results will corroborate our impression: The T-test value found for the window (-28,+28) is greater than the critical value of the t-distribution table for a N-1=1 degree of freedom at 90% level of confidence. It is then possible to reject the null hypothesis stating that the value of the CAAR=0. The value of the CAAR for the period (-28,+28) is then different from 0 with a positive sign. We can then say with 90% of confidence that there is a presence of positive abnormal returns and in particular positive CAAR for the FTSE TWSE Mid-cap 100 index and this for the period covering the month before and the month after the presidential election.
Regarding the FTSE TWSE Taiwan 50 index, we can see the pattern that was observed for the TWSE-mid, that is to say all the CAARs found for each event window of our study are positive;
expect the one for the period (-14,-8) which has a negative value of -2,378%. As it is possible to see, all the positive CAAR values seem to deviate strongly from 0. The CAAR t-test results will again validate this impression: The T-test value found for the window (-28,-15) is greater than the critical value of the t-distribution table for a N-1=2 degrees of freedom at the 90%, 95% and even 98% levels of confidence. It is then possible to reject the null hypothesis stating that the value of the CAAR=0. The value of the CAAR for the period (-28,-15) is then different from 0 with a positive sign. We can then say with 98% of confidence that there is a presence of positive abnormal returns and in particular positive CAAR for the TWSE large index and this for the period covering the two weeks after the beginning of the presidential campaign. Overall, we can also say that the whole period of study (-28,+28) has a positive CAAR of 9,079% but the value
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
is not found statistically significant at the confidence level we work with.
Comparing the results we found for both France and Taiwan, for the different stock indices we included in our study, it is possible to say that the Taiwanese stock market seems to react more strongly than the French one as we did not find any significant value for the CAAR results in France whereas we did for Taiwan. This is the big-caps index and the mid-caps one that reacted significantly on the Taiwanese stock market whereas the TAIEX, main index of the country, did not show any significant CAAR results. As the TAIEX includes all the stocks listed on the stock exchange in Taiwan, big, mid and small-caps, we may set the hypothesis that the small-caps stocks are the one reacting less to the presidential elections so that the TAIEX does not show any significant values. This is a possibility of topic for a future research.
As we said, we will extend this first study to further ones in order to incorporate different factors, related to the presidential elections that, we believe, may have an impact on the market indices. The first one has been selected as a factor of study for many researches previously done, as we could see in our literature review. Nevertheless it has not been adapted to the stock market in France and Taiwan and has been use to conduct study on the main indices in the US. We will use it, in our own way, to include the market capitalization factor in our study and be able to compare the different indices of our study. This factor is the election outcome, which means whether or not the incumbent in power wins or loses. The table number 3 shows the results obtained taking this factor into account for both elections of France and Taiwan. Nevertheless our study does not stop to this point as we want to test if the influence of this factor is different when we are considering different indices: more specifically when we include the market capitalization of the stocks listed on the indices of our study.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Incumbent wins the election Incumbent loses the election
Event Window CAAR t-test CAAR t-test
(+1,28) 4,495% 0,733241805 -8,852% -9,579735302*
(-7,-1) 0,657% 0,476085372 -3,505% -71,37326707****
(-14,-8) -0,956% -0,484693751 2,175% 98,46056423****
(-28,-15) 4,058% 2,151548573* -4,206% -19,48374717**
(-28,28) 7,892% 0,7565875 -14,387% -18,46164446**
Figure 8: CAAR Results and Significance Test When the Incumbent Wins or Loses the Election
(All Indices of France and Taiwan Included)
The results we obtained for this analysis appear to be, for some of them, significant. We will explain them and try to draw conclusion on what it is possible to observe around presidential election when whether the incumbent wins or loses for the indices we selected for this study.
First of all, let us focus on the results when we do not include in our analysis the market capitalization factor (when we only make the difference between a win and a loss of the incumbent, including all the stocks from France and Taiwan we are working with). Two strong trends seem to appear:
When the incumbent win, the value of the CAAR on the overall period of our study (28,+28), which includes all the windows we defined, seems strongly positive. The value of the CAAR for this period is 7,892% but is not significant at 90%+ level of confidence according to the to the CAAR t-test we conducted. Expect for the window (-14,-8), all the CAAR values are positive when the incumbent wins the election. One of them appears to be significant at 90%
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
level of confidence. This is the window (-28,15) for which the CAAR value observed is 4,058%.
The t-test value associated with this result is 2,15 and allows us to to reject the null hypothesis stating that the value of the CAAR=0 . What is interesting to notice and compare with the previous findings made in the United States, is that the CAARs do not appear to be higher the closer we are to the election date as it was possible to see in the United States with the study of Pantzalis and Al. For the combination of indices from France and Taiwan, the uncertainty does not seem to be resolved the closer we are to the election date. The significant result we found for this study belongs to the period of 15 days after the (average) beginning of presidential election campaign. Nevertheless, the period of 1 month after the presidential election outcome shows positive abnormal returns of 4,495%. Even if those are not significant, this is a trend we could expect as the study of Niederhoffer, Gibbs and Bullock showed the presence of abnormal returns (positive or negative) up to 1 month after the election in their comparison between Republicans and Democrats. Even if we do not compare political party and we only consider the outcome of the election for the incumbent, we could expect the presence of abnormal returns (positive or negative) after the outcome. In the case or re-election, the trend seems positive but not significant.
The results we got for the case when the incumbent loses the election will allow us to conduct a comparison with the results we just explained above. They appear to be completely different than the ones analyzing the re-election of the incumbent. Indeed, all the CAARs are negative except the one for the period (-14,-8). Overall, it is possible to say that a strong negative trend for the CAAR values appears when the incumbent loses the election. In fact, the value of the CAAR for the period (-28,+28) is strongly negative with -14,387% and appear to be significant at 95% level of confidence. We can reach the same conclusion with the value -4.206% of the
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
window (-28,-15). For the two windows the closest from the presidential election dates, the significance level is reached with 1% of confidence which are the strongest values obtained for our all study. It means that even if we only have two cases when the incumbent lost the election, it seems that the presence of abnormal returns the two weeks before the actual election date are significant with a high degree of confidence. It is important to notice that the CAAR for the period (-14,-8) is positive which could be seen as a break in the middle of the presidential campaign and come in favor of uncertainty resolution the closer we are from the election date.
Nevertheless, the next period (-7,-1) is negative and go against this theory. Finally, after the loss of the incumbent, the indices show a negative CAAR.
After considering only the win or lose of the incumbent, we tried to broaden our analysis including in our study the market capitalization factor, still with the outcome of the election to see whether or not we can draw further conclusion in this analysis. A difference has been made between big-caps and mid-caps but we also decided to include another factor: the notion of
‘main’ indices. In fact, in France the main index lists the big-caps stocks while in Taiwan the TAIEX lists big, mid and small market capitalization stocks. It was then necessary to make this distinction to be accurate in our analysis. First of all, it is important to say that none of these results have been found significant at 90+ % of confidence. Nevertheless, some trends can be underlined with the results we obtained. The first one follows our first findings: we observe a negative trend when the incumbent loses and we consider independently the ‘main’, ‘big-caps’
and ‘mid-caps’ indices while it is possible to observe a positive trend when the incumbent wins and we consider independently the same indices.
‧
CAAR t-test CAAR t-test CAAR t-test
(+1,28) 3,770% 0,698015954 3,611% 1,088890945 6,353% 0,713203269 (-7,-1) -0,250% -0,131982617 0,137% 0,098607668 0,163% 0,131640225 (-14,-8) -1,344% -0,768020502 -1,699% -0,738439773 0,085% 0,054160244 (-28,-15) 4,337% 1,084915305 4,366% 1,042640163 3,091% 1,242721719 (-28,28) 6,513% 0,849778135 5,993% 0,6727383 9,692% 0,682244147
Figure 9: CAAR Results and Significance Test When the Incumbent Wins for Every Indices Category
(All Indices of France and Taiwan Included) Incumbent loses + main
CAAR t-test CAAR t-test CAAR t-test
(+1,28) -7,928% -1,200987444 -7,928% -1,200987444 -9,776% -1,532600855 (-7,-1) -3,456% -1,306794686 -3,456% -1,306794686 -3,554% -1,39080905 (-14,-8) 2,197% 0,738436433 2,197% 0,738436433 2,153% 0,748780748 (-28,-15) -4,422% -1,111006131 -4,422% 1,111006131 -3,990% -1,008308483 (-28,28) -13,608% -1,322661293 -13,608% -1,322661293 -15,167% -1,466893925
Figure 10: CAAR Results and Significance Test When the Incumbent Loses for Every Indices Category
(All Indices of France and Taiwan Included)
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
When we look closer to the case when the incumbent loses, we find similarities between the three types of indices we are working with. A positive trend for the CAARs can be observed for the period (-14,-8). Nevertheless, it is possible to see that all the other periods show negative CAARs, the strongest values being after the election date. We can set up an explanation based on the hypothesis that the newly elected president brings uncertainty on the stock market for investors. This might be the reason why a negative trend can be observed overall, but more strongly after the election outcome. We cannot observe uncertainty resolution as we come closer to the election date and has already been shown in the United States. Finally, they are not noticeable differences in values between the windows of the different indices of our study. On the other way, when we look closer to the case when the incumbent wins, it is possible to observe a general positive trend on the overall period of study (including all the event windows).
Nevertheless, this trend seems stronger for the mid-caps indices than the main and big-caps indices. In fact, all the period appears to have positive CAARs when we consider the mid-caps.
Regarding the main and big-caps indices, it is possible to see that the CAARs are negatives when we approach the elections dates. Again, as a conclusion for this focus on the time when we consider the win of the incumbent, it is possible to say that the uncertainty does not appear to be reduced the closer we are from the presidential election day. As for the case when the incumbent lose, the strongest values of CAARs are after the election outcome. The mid-caps indices seem to be more reactive than the other indices as the CAAR value observed is more important than the two others.
After having exposed the results obtained in case of win or loss of the incumbent and linked them with the different indices of our study, we will broaden our study, incorporating another variable that has not been studied before. We have seen in our literature review that analysis
‧
have been conducted in the US to see whether the election of Democrats or Republicans have had a greater influence on the stock market and compare them. One possibility for us would have been to compare for France the impact of elections of “right wing” presidents against “left wing” ones and for Taiwan to compare the impact of elections of KMT presidents against DPP ones on the indices of our study. Nevertheless, we decided to define another variable of comparison, not studied before, that we believe very relevant in our study: whether the political party in power (for which the president belongs to) is re-elected or not. In fact, in our previous study (whether the incumbent win or loss), a criteria cannot be taken into account. This criteria is the fact for the incumbent not to be candidate again (for any reason: health, restriction on the
have been conducted in the US to see whether the election of Democrats or Republicans have had a greater influence on the stock market and compare them. One possibility for us would have been to compare for France the impact of elections of “right wing” presidents against “left wing” ones and for Taiwan to compare the impact of elections of KMT presidents against DPP ones on the indices of our study. Nevertheless, we decided to define another variable of comparison, not studied before, that we believe very relevant in our study: whether the political party in power (for which the president belongs to) is re-elected or not. In fact, in our previous study (whether the incumbent win or loss), a criteria cannot be taken into account. This criteria is the fact for the incumbent not to be candidate again (for any reason: health, restriction on the