348 B.E. Eckbo et al.
French (1993)size and book-to-market factors. UMD is a momentum factor inspired by
Carhart (1997)and constructed as the returns difference between the one-third highest
and the one-third lowest CRSP performers over the past 12 months. LMH is theEckbo
and Norli (2005)turnover factor, defined as a portfolio long in low-turnover stocks and
short in high-turnover stocks.
The alpha estimates are reported inTable 19for equity issuers, andTable 20for
debt issuers. As first reported byEckbo and Norli (2005), the estimated coefficients
on the turnover factor LMH tend to be both a greater and more significant than the
coefficients on the momentum factor UMD. When the coefficient on LMH is significant,
the extended model increases the regressionR^2 marginally above the Fama–French
model. Moreover, when significant, the estimated coefficients on both UMD and LMH
are typically negative, indicating that issuers tend to be relatively liquid, growth stocks.
When using the Fama–French model, the alphas are significant and negative for pri-
vate placements of equity (panel F ofTable 19) and for private placements of straight
debt (panels D and F inTable 20). However, the alpha estimates are insignificant in all
samples when using the extended model. There is ample evidence that the momentum
factor UMD helps explain the cross-section of expected stock returns. Evidence that
the turnover factor LMH is also priced is found inEckbo and Norli (2002 and 2005).
Assuming UMD and LMH are indeed priced risk factors, then the results inTable 19
and Table 20fail to reject the hypothesis of zero post-issue abnormal performance.
Table 21shows the alpha estimates reported in much of the literature that uses factor
regressions to estimate post-issue abnormal performance. For IPOs, and with the ex-
ception of non-venture-backed IPOs studied byBrav and Gompers (1997), the alphas
are statistically insignificantly different from zero (Brav, Geczy, and Gompers, 2000;
Ritter and Welch, 2002; Eckbo and Norli, 2005). For SEOs, and with the exception
ofJegadeesh (2000), all large-sample studies (3, 000 +SEOs) also report insignificant
alphas. These includeBrav, Geczy, and Gompers (2000), Eckbo, Masulis, and Norli
(2000), andLyandres, Sun, and Zhang (2005). For portfolios of SEOs, the Fama–French
model tend to produce larger (and sometimes significant) alphas than extended models
adding UMD, LMH and, most recently, the investment factor ofLyandres, Sun, and
Zhang (2005). Overall, assuming these factors are priced, the null of zero abnormal
post-SEO performance is not rejected.
Finally, studies of debt issues also find alphas that are indistinguishable from zero.
The largest sample is found inEckbo, Masulis, and Norli (2000), who study a total of
1,329 straight debt issues and 459 convertible debt offerings, report insignificant alpha
estimates for both types of debt issues.Spiess and Affleck-Graves (1999)report sig-
nificantly negative alphas for a constrained sample of debt issuers, where issues by a
given company that take place within five years of each other are excluded. However,
Butler and Wan (2005)show that adding a liquidity factor (much like the turnover fac-
tor ofEckbo and Norli (2005)produces insignificant alpha estimates also for the type
of restricted sample used bySpiess and Affleck-Graves (1999). Thus, again assuming