Univariate results – SEO discount

The average SEO discount in the large sample is 4.75 percent. This again is higher than that found in previous empirical studies, which we ascribe to the same reasons as discussed earlier.

In an attempt to address the issues relating to a comparatively small sample size, the larger sample (with information on the SEO discount only) is employed. This sample consists, as noted, of 2.065 observations. Employing a larger sample has several advantages. Firstly it reduces the impact of potential outliers that may induce substantial noise in analyses conducted on smaller samples. Secondly, the aforementioned issues of multicollinearity, are ceteris paribus lessened when a larger sample is employed.

Table 25

Obs.

Least

liquid 2

Most

liquid % Δ

All (2,065) 8.11% 4.30% 1.83% 343.2% < 0.0001 ***

SEO discount by Liquidity index Liquidity Tertile

P-value

Not surprisingly a substantial difference in average SEO discount is found across the liquidity tertiles.

As before, it is relevant to control for various confounding effects. Table 26 reports the average SEO discount for the portfolios formed by deal value and liquidity tertile. Table 26 strongly indicates that average SEO discount is substantially and significantly lower in the lost liquid tertile than in the least liquid fraction.

Table 26

Deal value Tertile

Least

liquid 2

Most

liquid % Δ

Smallest 9.49% 6.96% 4.73% 100.8% < 0.0001 ***

2 6.91% 4.10% 1.94% 256.1% < 0.0001 ***

Largest 4.96% 2.37% 1.25% 296.9% < 0.0001 ***

SEO discount by Deal value-Liquidity index Liquidity Tertile

P-value

98

We observe that the larger sample produces a relation between liquidity and deal value, which is substantial and significant across all tertiles. Table 26 indicates the SEO discount to be most severe in the case of small issuances with low liquidity. It generally finds the largest deal value tertile to be the least underpriced. This is somewhat surprising finding is, as noted a consequence of the tendency toward a positive correlation between firm size and size of principal offering.

Table 27 addresses this, analyzing the effect of illiquidity on the SEO discount, controlling for the relative deal size.

Table 27

Again, the finding of Corwin (2003) that larger deals (measured as a percentage of outstanding equity) requite a larger discount. While the tertile of the smallest deals required an average discount of 2.18 percent across all liquidity groups, the groups of the largest deals were on average discounted by a staggering 7.97 percent. Liquidity retains a substantial effect on SEO discount, with a very high level of statistical significance. The larger dataset reveals that as in table 22 on the small dataset, the effect of illiquidity seems more pronounced in smaller ‘% Co.

sold’ tertiles.

This seems to confirm that the importance of liquidity in determining SEO discounts diminishes as deals become larger in a relative sense. While in table 22 the effect vanished altogether on the tertile of the largest deals, in table 27 however, the relation remains statistically significant across all portfolios including the portfolio containing the largest deals.

This could be seen to confirm the hypothesis that the absence of difference in average SEO discount for the largest ‘% Co. sold’ deals of the small sample is caused by a low frequency of larger and liquid firms issuing large proportions of equity. In a sense this thus confirms the anecdotal evidence of the Thomas Cook Group.

% Co. sold Tertile

Least

liquid 2

Most

liquid % Δ

Smallest 3.91% 1.68% 0.93% 319.7% < 0.0001 ***

2 5.58% 4.12% 2.61% 113.8% < 0.0001 ***

Largest 10.60% 7.84% 5.46% 94.2% < 0.0001 ***

SEO discount by % of Company sold-Liquidity index Liquidity Tertile

P-value

99

While a significant effect of liquidity is retained, the relative effect of illiquidity on the SEO discount in the large ‘% Co. sold’ tertile is still substantially lower when compared to the smallest

‘% Co. sold’ tertile (94.2 vs. 319.7 percent). However, this again is a question of how one chooses to interpret the figures. Expressed in absolute terms, deals in the largest ‘% Co. sold’ tertile has an average difference across the liquidity tertiles of 514 basis points versus 298 basis points for the smaller ‘% Co. sold’ tertile. This, contrarily to the first interpretation, entails a larger ‘real’ effect.

Again, controlling for the effect of risk, table 28 confirms the insights from the analysis of the smaller dataset, while obtaining substantially more consistent levels of statistical significance.

Table 28

Table 28 again confirms the insight of Corwin (2003) and Mola and Loughran (2004) that issuances of more risky firms require larger discounts. While the average SEO discount across all liquidity tertiles of the least risky group is 2.62 percent, that of the most risky group stands at 7.64 percent.

Table 28 further confirms that the effect of illiquidity seems larger in the case of risky firms than in the case of firms with relatively stable returns. While the difference in SEO discount from belonging to the most liquid tertile to the least liquid tertile is staggering 404.1 percent in the least risky group, the ‘impact’ is reduced to 148.6 percent in the most risky group. The absolute difference is 631 versus 369 basis points when looking across the liquidity tertiles for the highest and lowest volatility tertiles respectively.

The larger sample size enables one to further scrutinize the significance of these relations, as it is possible to split the dataset in both dimensions in quintiles rather than tertiles. The results of this analysis are available in appendix 5. Generally they confirm the insights form the analysis based on tertiles. All results are still significant at a 1 percent confidence level.

Volatility Tertile

Least

liquid 2

Most

liquid % Δ

Lowest 4.61% 2.32% 0.92% 404.1% < 0.0001 ***

2 6.14% 3.85% 1.98% 210.2% < 0.0001 ***

Highest 10.56% 8.12% 4.25% 148.6% < 0.0001 ***

Liquidity Tertile

SEO discount by Volatility-Liquidity index

P-value

100

In summary, the larger sample generally corroborates the insights from the analysis of the SEO discount on the small sample, finding secondary market liquidity a statistically significant predictor of the SEO discount.

As previously noted, the above discussed effects may interact in a variety of ways. To reliably establish that there is indeed a relation between secondary market liquidity and SEO discount, one must again control for the relevant variables simultaneously utilizing a multiple regression framework. As before this analysis is done via an OLS regression.