Calendar-Time Portfolio

significance for the equally-weighted approach, which possibly can be attributed to the higher negative BHAR of -6 % compared to our -4%. Collectively, it seems that the study corresponds to the findings in this thesis to a reasonable degree. Other European long-run studies (Andrikopoulos, 2009; Jeanneret, 2005), on the other hand, report findings in close proximity to the American counterparts, and hence only resemble this paper’s findings to a minor degree. Besides, both studies exhibit differences in sample size to our study, since Jeanneret resorts to a substantially smaller sample size (114), while Andrikopoulos’ (2009) sample consists of a far bigger number (1,542), as well as matching approach.

According to the market timing hypothesis, as mentioned in the “Literature Review”, equity issuance will be undertaken when the respective costs are the lowest and hence most favorable for the issuing firm. These times are following high returns according to the market hypothesis. As a consequence of the issue and the connected high returns a priori, the stock price will adjust downwards after the equity issuance. Our BHAR results are for both the equally- and the value-weighted approach negative, and display statistical significance for the latter. As a result, our results support the market timing hypothesis to a certain degree, as a significant negative stock price development compared to the chosen benchmark is observable for the value-weighted approach. Nevertheless, it is important to keep in mind that a different benchmark might have led to a varying outcome.

or more different event company returns, the possibility of one return dominating the entire portfolio return should be reduced essentially.

Fama & French 3-factor Model

Table 7: 3-factor model results considering All Months (left) and 20+ Adjustment (right)

The part of the table concerning the All Month approach (left) demonstrates statistical significance for all coefficients, given the equally-weighted approach. While the market portfolio factor and the book-to-market factor show statistical significance on the 1 % level, the size factor relies on significance on the 5 % and the intercept, or Jensen’s alpha, on the 10 % level. Besides, the market portfolio factor is positive and close to 1, implying that the formed SEO portfolio returns follow the market development quite markedly. In contrast to this, the other three coefficents have negative signs. On the other hand, the value-weighted coefficients draw a slightly different picture. The market portfolio factor still demonstrates significance on the 1 % level, as well as the size factor. By contrast, the book-to-market coefficient presents significance only on the 5 % level in this case. Correspondingly, the size coefficient increased in absolute terms by turning more negative, whereas the opposite is evident for the book-to-market factor, implying a more driving impact of the size factor. Additionally, the intercept no longer displays significance, as it approaches a value of zero in the value-weighted approach. The coefficient of determination is slightly higher for the equally-weighted approach, implying a better description of the models variance. Altogether, it can be concluded that abnormal negative performance cannot be precluded for the equally-weighted approach, while no abnormal performance can be derived given the value-weighted approach. This implies that smaller companies seem to experience more pronounced negative abnormal returns than larger issuing companies.

Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat Intercept -0.004 -1.758* -0.002 -0.683 Intercept -0.006 -2.277** -0.002 -0.595

Rm-Rf 0.935 18.191*** 0.802 16.172*** Rm-Rf 0.959 18.770*** 0.811 16.969***

SMB -0.248 -2.253** -0.649 -6.109*** SMB -0.100 -0.872 -0.506 -4.733***

HML -0.333 -3.255*** -0.232 -2.353** HML -0.321 -3.149*** -0.211 -2.210**

R² 0.555 0.542 R² 0.623 0.608

***, ** and * denote significance on the 1 %, 5 % and 10 % level respectively

Full sample (T = 284) Full sample (T = 220)

Equally-Weighted Value-Weighted Equally-Weighted Value-Weighted

While the left statistics consider all potential months which included return data, the right statistics only consider months which included at least 20 different event to decrease the potential influence of one event's return results on the accumulated monthly return data. The stated R square refers to the adjusted R square of the regression. The T identifies the months that include conforming return data for the sample.

On the other hand, the segment featuring the results for the 20+ adjustment reveals differences as the 20+ adjustment accounts for an essentially higher R2 than for the All Month equivalent. Comparing the two samples further, similar characteristics are recognizable in terms of statistical significance.

Considering the equally-weighted approach, the main difference is the size factor as it changes from significance on the 5 % level to no statistical significance for the 20+ adjustment. By contrast, the intercept increases in terms of statistical significance, as well as in absolute negative value terms, to a 5

% level for the 20+ adjustment. In turn, the value-weighted results are identical in statistical significance and also only change marginally on a value base, except for the size factor which increases by 0.143 (~22%). Consequently, the same reasoning regarding abnormal performance can be followed as for the entire sample before since negative abnormal performance in the equally-weighted approach cannot be precluded, whereas no abnormal performance is prevailing given value-weights. As a result, the larger companies also seem to outperform the smaller companies in this approach.

Carhart 4-factor Model

Likewise to the description of the results using the 3-factor model, this section will consider the entire sample and associated results. Subsequently, the results of the 4-factor model will be compared to the ones from the factor model. Taking a look at the R2 of table below, the pattern established in the 3-factor is maintained as the equally-weighted averages exceed the value-weighted counterparts, while the 20+ adjustment also exhibits a higher R2 level than the All Month approach.

Table 8: 4-factor model results considering All Months (left) and 20+ Adjustment (right)

Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat

Intercept -0.001 -0.546 0.000 -0.167 Intercept -0.003 -1.121 0.000 -0.066

Rm-Rf 0.880 16.876*** 0.780 15.164*** Rm-Rf 0.892 17.312*** 0.783 15.709***

SMB -0.232 -2.159** -0.643 -6.060*** SMB -0.081 -0.736 -0.498 -4.682***

HML -0.425 -4.146*** -0.270 -2.669*** HML -0.415 -4.127*** -0.251 -2.581***

WML -0.248 -3.885*** -0.102 -1.611 WML -0.263 -4.267*** -0.112 -1.880*

R² 0.576 0.544 R² 0.651 0.612

Full sample (T = 220)

Equally-Weighted Value-Weighted Full sample (T = 284)

Equally-Weighted Value-Weighted

While the left statistics consider all potential months which included return data, the right statistics only consider months which included at least 20 different event to decrease the potential influence of one event's return results on the accumulated monthly return data. The stated R square refers to the adjusted R square of the regression. The T identifies the months that include conforming return data for the sample.

***, ** and * denote significance on the 1 %, 5 % and 10 % level respectively

Starting once again with a description of the sample results including all potential months, statistical significance is prevailing for all coefficients except for the intercept given equal weights. Significance on the 1 % level is achieved for the remaining coefficients but the size factor, which only demonstrates significance on the 5 % level. The value-weighted approach displays significance on the 1 % level for the market portfolio, size and book-to-market coefficient, while the momentum factor as well the intercept do not demonstrate any significance. Consequently, no abnormal performance is detected for the entire sample given the 4-factor model. The 20+ adjustment sample has only two differences regarding the statistical significance of the respective coefficients. In comparison to the All Month approach, the size factor given equal weights is not significance, while the momentum factor in the value-weighted approach experiences statistical significance on the 10 % level. The most important similarity, however, is that both intercepts do not display any significance and consequently no abnormal performance can be inferred, despite the intercept being more negative in absolute terms using equally-weighting averages.

Comparing these results to the 3-factor model, the non-existent statistical significance for the intercept is the most notable difference. This fact implies that the introduction of a fourth factor has led to a better depiction of the portfolio returns and simultaneously reduced the intercept values. Besides, only the value-weighted book-to-market coefficient discloses a varying statistical significance on the 1 % compared to 5 % level, implying similar impacts of the 3-factors in both factor models.

5.3.4 Comparison of the CTP Results to Previous Research

As described above, our findings differ depending on the factor model employed. Since no abnormal performance can be derived given the 4-factor model, the following comparison to former studies will be undertaken using the results obtained from the 3-factor model and particularly the 20+ adjustment approach. Our findings indicate a monthly abnormal performance of -0.06 % for the equally-weighted approach, while a value-weighted abnormal performance of -0.02 % cannot be statistically proven.

The majority of American reference studies are in line with our thesis by experiencing higher negative abnormal performance for their equally-weighted results (Brav et al, 2000: Loughran & Ritter, 1995;

Mitchell & Stafford, 2000). Moreover, these studies report statistical significance for only their equally-weighted results, which is in accordance with our findings, with the exception for Loughran &

Ritter (1995), who display significance for both approaches. Besides, Jegadeesh (2000) only controls for equally-weighted returns and also finds statistically significant negative returns. Nevertheless, the three studies mentioned before present negative monthly abnormal performances which exceed the results in our findings by a substantial amount in the equally-weighted approach (-0.33 % as lowest example versus -0.06 %). For the value weighted approach, Mitchell & Stafford (2000) report a result of -0.03 %, which is very similar to our finding, while the other two studies find markedly higher negative returns. Once again, Eckbo et al. (2000) poses an exceptional case by finding a negative monthly abnormal experience for the value-weighted case which surpasses the equal-weighted equivalent. Additionally, none of Eckbo et al.’s findings demonstrate statistical significance. However, their equally-weighted result (-0.12 %) comes closest to our finding of -0.06 %. Collectively, the American studies paint a similar picture by mostly relating to statistically significant equally-weighted returns, while also not reporting statistical significance while using value weights.

European examples are rare, but Jeanneret (2005) uses also the 3-factor model and arrives at statistically significant abnormal performances of -0.75 % and -0.95% for the equally- and the value-weighted approach respectively. These results exceed our findings by an enormous margin, while they also demonstrate a higher negative abnormal performance for the value- compared to the equally-weighted approach, contrary to our findings. Since Jeanneret only reverts to a sample size of 114 events, his results might be dominated by particular event companies, which could be one explanation of the substantial differences to our results.

As derived in the comparison of the BHAR approach, the results reported for the CTP approach have a similar relationship to the market timing hypothesis. While both averaging schemes display negative average monthly returns, only the equally-weighted approach demonstrates statistical significance.

Consequently, the market timing hypothesis can be considered as supported by the CTP approach as negative returns over three years following the announcement are depicted.