Comparing the numbers in Tables 2 and 3,the results seem somewhat sensitive to what time period is examined,and what sample selection criteria are used.Evidence shows that SEOs from the heavy-volume period of 1970-1972 did very poorly in the bear market of 1973-1974, and failed to recover in the small stock rally of 1975-1976.The Loughran and Ritter(1995, 2000)studies,as well as this chapter's Table 1,include these SEOs from the early 1970s in the post-issue returns,whereas Jegadeesh(2000),Eckbo,Masulis,and Norli(2000),and Brav, Geczy,and Gompers(2000)exclude them.Thus,in January 1976,the Brav,Geczy,and Gompers issuing firm portfolio only includes issuing firms from 1975,whereas the Loughran and Ritter portfolio includes equity issuers from 1971-1975,including over 500 issuers listed,or subsequently listed,on Nasdaq.Issuing firms also did especially poorly during the bursting of the tech stock bubble in 2000.Thus,the different abnormal return estimates from various studies that are reported in Table 3,where the same methodology is used in every study,are largely due to differing sample periods.The Eckbo,Masulis,and Norli finding of minimal abnormal returns after SEOs can partly be attributable to high returns on a small number of NYSE-listed issuing firms in the 1960s,which have a large impact on their conclusions,for these studies weight each period equally.s This sensitivity of the performance results to the sample period used is not unique to issuing companies.Two of the best-known empirical patterns in finance are that growth stocks tend to underperform value stocks and that small firms outperform large firms(Fama and French (1992),Davis,Fama,and French (2000)).But if 1975-1982 is excluded,the "small firm"effect disappears,and if the sample period includes just 1983-1999,small firms underperform.In the 1990s,large growth firms had higher returns than any other style category.So just as the size and book-to-market effects vary across subperiods,it should be no surprise that the relative performance of issuing firms varies in different subperiods. 2.3 Reasons for underperformance The evidence on the long-run performance of firms conducting SEOs is that issuing firms have relatively low returns in the 3-5 years after issuing.A number of explanations have been 4 Eckbo,Masulis,and Norli (2000)exclude Nasdaq-listed issuers prior to 1974.Mitchell and Stafford(2000) exclude Nasdaq-listed issuers prior to 1973.Nasdaq started in February 1971,but CRSP does not start covering Nasdag stocks until December 1972 5 Eckbo,Masulis,and Norli(2000)find that large firms conducting SEOs in the 1960s subsequently outperformed a multifactor benchmark.Their sample excludes SEOs by Nasdag firms prior to 1974,and includes only a small portion of Amex-NYSE issuers in the 1960s and early 1970s,apparently because their sample for this period required Wall Street Journal announcement dates when it was originally constructed.Their sample period is 32 years long(1964-1995),but over 80%of their sample is from the second half.In 1964-1979,53%of their issuers are utilities,whereas in 1980-1995,only 13%are utilities.Because they do not include many SEOs from the 1960s. the portfolio for their multifactor regressions is frequently tiny when utility firms are excluded(Tables 6,8,9 and 10).For example,in November 1964,they have only four firms in the portfolio.In November 1965,the portfolio has increased to only ten firms,and a year after that,it has only 19 firms.In November 1967,only 23 firms are present.By contrast,some months in the 1980s and 1990s have close to 1,000 firms in their issuer portfolio.While they adjust the standard errors for the resulting heteroscedasticity,the coefficient estimates weight each period equally. 16
16 Comparing the numbers in Tables 2 and 3, the results seem somewhat sensitive to what time period is examined, and what sample selection criteria are used. Evidence shows that SEOs from the heavy-volume period of 1970-1972 did very poorly in the bear market of 1973-1974, and failed to recover in the small stock rally of 1975-1976. The Loughran and Ritter (1995, 2000) studies, as well as this chapter’s Table 1, include these SEOs from the early 1970s in the post-issue returns, whereas Jegadeesh (2000), Eckbo, Masulis, and Norli (2000), and Brav, Geczy, and Gompers (2000) exclude them.4 Thus, in January 1976, the Brav, Geczy, and Gompers issuing firm portfolio only includes issuing firms from 1975, whereas the Loughran and Ritter portfolio includes equity issuers from 1971-1975, including over 500 issuers listed, or subsequently listed, on Nasdaq. Issuing firms also did especially poorly during the bursting of the tech stock bubble in 2000. Thus, the different abnormal return estimates from various studies that are reported in Table 3, where the same methodology is used in every study, are largely due to differing sample periods. The Eckbo, Masulis, and Norli finding of minimal abnormal returns after SEOs can partly be attributable to high returns on a small number of NYSE-listed issuing firms in the 1960s, which have a large impact on their conclusions, for these studies weight each period equally.5 This sensitivity of the performance results to the sample period used is not unique to issuing companies. Two of the best-known empirical patterns in finance are that growth stocks tend to underperform value stocks and that small firms outperform large firms (Fama and French (1992), Davis, Fama, and French (2000)). But if 1975-1982 is excluded, the “small firm” effect disappears, and if the sample period includes just 1983-1999, small firms underperform. In the 1990s, large growth firms had higher returns than any other style category. So just as the size and book-to-market effects vary across subperiods, it should be no surprise that the relative performance of issuing firms varies in different subperiods. 2.3 Reasons for underperformance The evidence on the long-run performance of firms conducting SEOs is that issuing firms have relatively low returns in the 3-5 years after issuing. A number of explanations have been 4 Eckbo, Masulis, and Norli (2000) exclude Nasdaq-listed issuers prior to 1974. Mitchell and Stafford (2000) exclude Nasdaq-listed issuers prior to 1973. Nasdaq started in February 1971, but CRSP does not start covering Nasdaq stocks until December 1972. 5 Eckbo, Masulis, and Norli (2000) find that large firms conducting SEOs in the 1960s subsequently outperformed a multifactor benchmark. Their sample excludes SEOs by Nasdaq firms prior to 1974, and includes only a small portion of Amex-NYSE issuers in the 1960s and early 1970s, apparently because their sample for this period required Wall Street Journal announcement dates when it was originally constructed. Their sample period is 32 years long (1964-1995), but over 80% of their sample is from the second half. In 1964-1979, 53% of their issuers are utilities, whereas in 1980-1995, only 13% are utilities. Because they do not include many SEOs from the 1960s, the portfolio for their multifactor regressions is frequently tiny when utility firms are excluded (Tables 6, 8, 9 and 10). For example, in November 1964, they have only four firms in the portfolio. In November 1965, the portfolio has increased to only ten firms, and a year after that, it has only 19 firms. In November 1967, only 23 firms are present. By contrast, some months in the 1980s and 1990s have close to 1,000 firms in their issuer portfolio. While they adjust the standard errors for the resulting heteroscedasticity, the coefficient estimates weight each period equally
advanced for these low returns.These ideas also apply to IPOs,where the empirical evidence on long-run underperformance is discussed in Section 4 of this chapter. One possibility is that the underperformance of issuing firms may be just a manifestation of a misspecified model for what the returns should have been.Fama(1998)refers to this as the "bad model"problem.Eckbo,Masulis,and Norli (2000),for example,present evidence that a 6- factor asset pricing model can explain the performance of issuing firms.They argue that the decreased leverage associated with an equity issue lowers the sensitivity of the stock price to inflation shocks,and the extra shares outstanding make the stock more liquid.In general,they argue that issuing firms have low risk as a result of the equity issue,and therefore should have low returns. This raises the question of just how risky companies conducting SEOs really are.Fama- French(1993)3-factor regressions typically find that SEOs have near-average systematic risk,a high sensitivity to the size factor,but a surprisingly modest sensitivity to the book-to-market factor.This presents a misleading picture,however. Table 4 reports several single-factor and multifactor regressions.None of the factors are purged of issuing firms,so the intercepts underestimate the degree of abnormal performance. Row(1)reports a simple one-factor regression where the intercept is the "Jensen alpha."The intercept is a statistically significant negative 52 basis points per month.Row(2)includes a lagged market excess return,which is highly significant.Summing the contemporaneous and lagged betas(see Fama and French(1992)for another paper using this procedure)gives a systematic risk estimate of 1.50,which shows that SEOs expose investors to a high degree of market risk. Table 4 Multi-factor regressions with an EW portfolio of U.S.SEOs,January 1973-December 2000 rpt-ra a+bt(rmt-rA)+bt-1(rmt-1-rA.1)+StSMBt VtVMGt ept a bu b以 St h R2 (1) -0.52 1.37 78.2% (-2.84) (34.65) (2) -0.60 1.36 0.14 78.9% (-3.28) (35.16) (3.60) (3) -0.52 1.20 0.89 -0.01 94.6% (-5.53) (53.43) (30.87) (-0.36) T-statistics are in parentheses.All regressions use 336 observations where the dependent variable is the monthly percentage return on a portfolio of IPOs that have gone public during the prior 36 months.A coefficient of-0.52 represents underperformance of 52 basis points per month,or-6.24 percent per year before compounding.The explanatory variables are described above Table 3. 17
17 advanced for these low returns. These ideas also apply to IPOs, where the empirical evidence on long-run underperformance is discussed in Section 4 of this chapter. One possibility is that the underperformance of issuing firms may be just a manifestation of a misspecified model for what the returns should have been. Fama (1998) refers to this as the “bad model” problem. Eckbo, Masulis, and Norli (2000), for example, present evidence that a 6- factor asset pricing model can explain the performance of issuing firms. They argue that the decreased leverage associated with an equity issue lowers the sensitivity of the stock price to inflation shocks, and the extra shares outstanding make the stock more liquid. In general, they argue that issuing firms have low risk as a result of the equity issue, and therefore should have low returns. This raises the question of just how risky companies conducting SEOs really are. FamaFrench (1993) 3-factor regressions typically find that SEOs have near-average systematic risk, a high sensitivity to the size factor, but a surprisingly modest sensitivity to the book-to-market factor. This presents a misleading picture, however. Table 4 reports several single-factor and multifactor regressions. None of the factors are purged of issuing firms, so the intercepts underestimate the degree of abnormal performance. Row (1) reports a simple one-factor regression where the intercept is the “Jensen alpha.” The intercept is a statistically significant negative 52 basis points per month. Row (2) includes a lagged market excess return, which is highly significant. Summing the contemporaneous and lagged betas (see Fama and French (1992) for another paper using this procedure) gives a systematic risk estimate of 1.50, which shows that SEOs expose investors to a high degree of market risk. Table 4 Multi-factor regressions with an EW portfolio of U.S. SEOs, January 1973-December 2000 rpt - rft = a + bt(rmt - rft) + bt-1(rmt-1 - rft-1) + stSMBt + vtVMGt + ept a bt bt-1 st ht R2 (1) -0.52 1.37 78.2% (-2.84) (34.65) (2) -0.60 1.36 0.14 78.9% (-3.28) (35.16) (3.60) (3) -0.52 1.20 0.89 -0.01 94.6% (-5.53) (53.43) (30.87) (-0.36) T-statistics are in parentheses. All regressions use 336 observations where the dependent variable is the monthly percentage return on a portfolio of IPOs that have gone public during the prior 36 months. A coefficient of –0.52 represents underperformance of 52 basis points per month, or –6.24 percent per year before compounding. The explanatory variables are described above Table 3
Row(3)of Table 4 reports Fama-French 3-factor regression coefficients.The estimate of systematic risk is 1.20,considerably lower than the 1.50 value of the summed betas in row(2). The reason for this difference is that the size factor catches some of the systematic risk.Small stocks tend to have higher betas than big stocks,so SMB tends to have a positive factor return in rising markets.In other words,SMB is not orthogonal to the market excess return.Indeed,if one runs regression(2)with SMB as the dependent variable,the summed beta is 0.33.That is, small stocks expose an investor to a beta that is 0.33 higher than large stocks do.Since issuing firms tend to underrepresent the largest stocks,part of the high level of systematic risk that they expose an investor to is captured by the size effect. Market efficiency requires that,if one uses the appropriate benchmark,there should be no abnormal returns on average after an event.As Loughran and Ritter(2000)point out,tests of market efficiency are always joint tests of a(theoretically motivated)model of market equilibrium and the existence of abnormal returns.Since matching by size and book-to-market is empirically motivated,rather than theoretically motivated,the abnormal returns reported in Tables 1 through 4 are not evidence for or against market efficiency.Still,it is hard to imagine that the relatively low post-issue returns on issuing firms can be attributed to low risk,since row (2)of Table 4 shows that issuing firms expose investors to a very high level of systematic risk. Eckbo,Masulis,and Norli(2000)argue that the decreased leverage after an equity issue lowers the systematic risk of equity issuers.Denis and Kadlec(1994),however,report that once various statistical biases are accounted for,there is no change in the equity beta for issuing firms, even though theoretically there should be a change if operating risk doesn't change.It is, however,entirely conceivable that lower leverage is more than offset by increased operating risk, if issuing companies embark on aggressive expansion plans with the money raised in an SEO. Another possible explanation of the negative abnormal returns on issuers is that the findings are just due to chance,possibly because a few industries that had heavy issuance activity failed to live up to expectations.Although the sample sizes involve thousands of firms,the number of independent observations is considerably smaller.In Table 5 of Section 3,however,it is shown that over a wide variety of corporate financing-related events,there is a persistent pattern of underreaction.Thus,the chance explanation requires not only that SEOs just happened to underperform,but that underreaction just happened to occur in a large variety of other events. Jung,Kim and Stulz (1996)examine a number of explanations for the decision to issue equity or debt.They interpret their evidence on debt and equity issues as consistent with an agency model.In particular,firms that issue equity when they could apparently issue debt have more negative announcement returns,the lower is their market-to-book ratio.The agency explanation is that managers will tend to squander corporate resources if given the opportunity, although this may not be intentional. Another possibility is that investors and managers are systematically overoptimistic at the time of issue.After all,for most of the issuers,good things have been happening to the stock price in the year prior to issue.Issuers tend to be firms that have recently outperformed other 18
18 Row (3) of Table 4 reports Fama-French 3-factor regression coefficients. The estimate of systematic risk is 1.20, considerably lower than the 1.50 value of the summed betas in row (2). The reason for this difference is that the size factor catches some of the systematic risk. Small stocks tend to have higher betas than big stocks, so SMB tends to have a positive factor return in rising markets. In other words, SMB is not orthogonal to the market excess return. Indeed, if one runs regression (2) with SMB as the dependent variable, the summed beta is 0.33. That is, small stocks expose an investor to a beta that is 0.33 higher than large stocks do. Since issuing firms tend to underrepresent the largest stocks, part of the high level of systematic risk that they expose an investor to is captured by the size effect. Market efficiency requires that, if one uses the appropriate benchmark, there should be no abnormal returns on average after an event. As Loughran and Ritter (2000) point out, tests of market efficiency are always joint tests of a (theoretically motivated) model of market equilibrium and the existence of abnormal returns. Since matching by size and book-to-market is empirically motivated, rather than theoretically motivated, the abnormal returns reported in Tables 1 through 4 are not evidence for or against market efficiency. Still, it is hard to imagine that the relatively low post-issue returns on issuing firms can be attributed to low risk, since row (2) of Table 4 shows that issuing firms expose investors to a very high level of systematic risk. Eckbo, Masulis, and Norli (2000) argue that the decreased leverage after an equity issue lowers the systematic risk of equity issuers. Denis and Kadlec (1994), however, report that once various statistical biases are accounted for, there is no change in the equity beta for issuing firms, even though theoretically there should be a change if operating risk doesn’t change. It is, however, entirely conceivable that lower leverage is more than offset by increased operating risk, if issuing companies embark on aggressive expansion plans with the money raised in an SEO. Another possible explanation of the negative abnormal returns on issuers is that the findings are just due to chance, possibly because a few industries that had heavy issuance activity failed to live up to expectations. Although the sample sizes involve thousands of firms, the number of independent observations is considerably smaller. In Table 5 of Section 3, however, it is shown that over a wide variety of corporate financing-related events, there is a persistent pattern of underreaction. Thus, the chance explanation requires not only that SEOs just happened to underperform, but that underreaction just happened to occur in a large variety of other events. Jung, Kim and Stulz (1996) examine a number of explanations for the decision to issue equity or debt. They interpret their evidence on debt and equity issues as consistent with an agency model. In particular, firms that issue equity when they could apparently issue debt have more negative announcement returns, the lower is their market-to-book ratio. The agency explanation is that managers will tend to squander corporate resources if given the opportunity, although this may not be intentional. Another possibility is that investors and managers are systematically overoptimistic at the time of issue. After all, for most of the issuers, good things have been happening to the stock price in the year prior to issue. Issuers tend to be firms that have recently outperformed other