Multivariate Analysis (POS)

power, as opposed to the full sample regressions where we also presented models including all variables. As already motivated, we only examine the POS sample.

The first obvious feature from the POS regressions concerns the Adj R², which is decreasing from the 12- to the 36-month regression. This indicates that the variables for a shorter horizon better explain the variations in BHAR. This however, should be expected to a certain degree, as the possibility that other factors explain the variation increases over time. Accordingly, this has to be taken into account when evaluating long-run results, since the effect of the evaluated event can diminish over time due to other factors.

Similar to the full sample regressions, it seems like the value-weighted models for the POS sample have a higher explanatory power and thus, better explain the variations in BHAR. For the equally-weighted regressions, where emphasis is put on smaller firms, FreqIssuer is positive and significant over all three years. This implies that frequent issuers, as long as they experience positive SEO announcement effects, perform better than single issuers consistently over three years. In addition to the FreqIssuer variable, MarketCap is also significant in the first year. The coefficient however is negative, suggesting that companies with a higher market cap, ceteris paribus, perform worse. Over two years, the RelSize coefficient also reports significance. The RelSize has a positive impact, which can be related to the negative impact of the MarketCap within the first year, as a higher market value implies a lower relative size of the issue, ceteris paribus.

When examining the value-weighted regressions, we find additional statistically significant variables explaining the respective long-run return and thus higher explanatory power. Throughout the first year, several coefficients, according to our regression results, are able to explain the BHAR variation. The FreqIssuer coefficient is positive again, while the RelSize, the MarketCap, the B2M, the MOM and IT appear to negatively impact the BHAR. Two years after the announcement, the value-weighted regression features the same significant variables except for the B2M. This indicates that the same variables influence the BHAR for rather large companies over the first two years. After three years, only the RelSize and the MarketCap remain in displaying a negative impact on the BHAR result compared to the results over 12 and 24 months. Both RelSize and MarketCap have a more pronounced negative impact the longer the time horizon, while the other previously significant factors turn

insignificant as time passes. However, the B2M coefficient in the last year appears to have a positive impact, suggesting a positive influence of time on the B2M coefficient and consequently that a higher B2M ratio increases the BHAR after 3 years.

A comparison of the equally- and value-weighted results demonstrates that a consistent finding is the positive impact of the FreqIssuer coefficient on the BHAR results of the POS sample. Consequently, with the premise that the companies experience positive announcement effects, it seems reasonable to assume that frequent issues by companies will lead to a more positive long-term result, compared to their first issues. This finding is rather surprising, as scholars have argued the opposite, namely that frequent issuers underperform single issuers (e.g. Brav et al., 2000; Billet et al., 2010). An explanation to this could potentially be derived from the fact that these FreqIssuer firms experienced a positive announcement effect. Since they are included in the POS sample, their SEO announcements have been considered good news, which could mean that their first SEO was a success from an investor’s point of view. It could potentially indicate that the firms’ prior SEO experience resulted in a satisfactory utilization of the raised capital and thus, has a positive impact on the long-run performance. If this holds, it is reasonable that consecutive SEO’s are successful as well. With regards to industries, we find no evidence that the long-run returns differ for industrial, utility and financial companies. However, we do find that larger IT firms seem to underperform equivalent industrial firms over 12- and 24- months following a SEO. In the value-weighted regressions, we find evidence that the RelSize, MarketCap and MOM have a negative impact on the long-run returns. Furthermore, since the coefficients become larger over time, the impact seems to be more severe as time passes, ceteris paribus, for RelSize and MarketCap. We also find that MOM has a negative relationship with the BHAR for larger firms over 12 and 24 months. Lastly, we find no evidence that the positive 3-day CAR has a significant impact on the BHAR. Thus, for the POS sample, we cannot conclude that the announcement effect of the SEO is of relevance for the long-run returns.

Collectively, our findings for the POS sample suggest that some predictable patterns can be derived in spite of the insignificant impact of the actual announcement effect. First of all, the results indicate that frequent issuers that experience a positive announcement effect outperform single issuers. Secondly, the regression results indicate that the BHAR is subject to being more negative given a higher level of the relative size of the issue and the market cap. Furthermore, it seems that IT companies compared to

industrials perform comparably worse. Lastly, the momentum factor exhibits a negative coefficient, implying that a positive prior performance will lead to more negative BHAR results, whereas a negative prior performance entails more positive BHARs.