Securities litigation, withdrawal risk and initial public offerings

Journal of Corporate Finance 17 (2011) 438–456

Contents lists available at ScienceDirect

Journal of Corporate Finance j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j c o r p f i n

Securities litigation, withdrawal risk and initial public offerings☆ (Grace) Qing Hao ⁎ Department of Finance, Robert J. Trulaske, Sr. College of Business, University of Missouri, Columbia, MO 65211, USA

a r t i c l e

i n f o

Article history: Received 1 November 2007 Received in revised form 13 December 2010 Accepted 23 December 2010 Available online 14 January 2011 JEL classification: G24 K22

a b s t r a c t I examine the relations between litigation risk, withdrawal risk, and the costs of going public using a sample of withdrawn and completed initial public offerings (IPOs) filed during 1996– 2005. Firms with a higher probability of offer withdrawal face higher litigation risk if they complete these offers. Firms with higher litigation risk pay slightly higher gross spreads, but do not underprice their IPOs by a greater amount. Withdrawal probability is strongly and positively associated with underwriter gross spreads, consistent with underwriters charging fees that reflect the probability of not getting paid. When the pre-market demand for an IPO is weak, a higher withdrawal probability raises underpricing on completed deals. © 2011 Elsevier B.V. All rights reserved.

Keywords: IPO Litigation risk Withdrawal risk IPO failure Underpricing Underwriter gross spread

1. Introduction Litigation risk poses a significant threat to firms going public in the United States. Once issuers are sued, regardless of the eventual outcome of the lawsuits, just the indirect costs due to lost management time and damaged reputation represent an enormous burden, not to mention the potentially huge legal liabilities for a young firm with limited resources. How does the legal threat affect the initial public offering (IPO) process? This paper tries to shed light on several open questions in the literature. First, is the IPO withdrawal decision affected by an issuer's legal liability concerns? How is the probability of IPO withdrawal related to litigation risk? For firms with high litigation risk, the expected costs of completing an IPO should be high for both underwriters and issuers. Therefore, I expect firms with high litigation risk to have a high probability of IPO withdrawal. However, the literature lacks empirical evidence on this question. Second, do litigation risk and withdrawal risk affect IPO underpricing? The litigation risk hypothesis for IPO underpricing argues that firms with higher litigation risk underprice their IPOs by a greater amount as a form of insurance. While many researchers have contributed to the debate (see, e.g., Ibbotson, 1975; Tinic, 1988; Hughes and Thakor, 1992; Drake and Vetsuypens, 1993; Hensler, 1995; Lowry and Shu, 2002; Zhu, 2009; Pukthuanthong et al., 2009), one important issue has been

☆ This paper is substantially revised from one chapter of my Ph.D. dissertation at the University of Florida. I owe special thanks to an anonymous referee and David Denis (the editor), who have provided many constructive suggestions, and to Jay Ritter (my dissertation committee chair), for his IPO data, valuable comments, and continued support. I sincerely thank Chunrong Ai, Kathleen Hanley, John Howe, Christopher James, Kevin M. LaCroix, James Leigh-Pemberton, Jeffrey Wooldridge, Yun Zhu, and seminar participants at the 2007 European Financial Management Symposium on IPOs at Oxford University for helpful comments. I thank Matt Crook and Anamaria Calincan for providing excellent research assistance. I acknowledge financial support from the Robert J. Trulaske, Sr. College of Business at the University of Missouri. ⁎ 425 Cornell Hall, University of Missouri, Columbia, MO 65211, USA. Tel.: + 1 573 884 1446; fax: + 1 573 884 6296. E-mail address: [email protected]. 0929-1199/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jcorpfin.2010.12.005

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

439

ignored. If IPOs with greater litigation risk are more likely to be withdrawn, then the effect of withdrawal risk also needs to be considered when examining how litigation risk affects underpricing of completed IPOs. Third, do litigation risk and withdrawal risk affect underwriter gross spreads? Firms with fundamentally higher underwriting risks and requiring more extensive underwriter marketing efforts should pay higher underwriting fees. If litigation risk and withdrawal risk are part of the overall underwriting risk, I expect them to positively affect gross spreads. However, this issue has received no attention in the literature. Using a sample of withdrawn and completed IPO filings during the 1996–2005 period, I uncover several key findings. First, firms that are more likely to withdraw their IPOs, are more vulnerable to lawsuits if they complete their offers. At least one in every five IPOs filed with the SEC is withdrawn. The withdrawal decision is significantly affected by the information learned during the registration period. For issuers completing their IPOs, the estimated probability of withdrawal turns out to be a significant predictor of litigation that is brought against the issuers later. A one standard deviation increase in the probability of withdrawal raises the probability of being sued by 1%. Compared with the 4.2% unconditional probability of an IPO firm being sued, the probability of being sued increases by approximately 25%. The result suggests that the withdrawal decision is strongly influenced by litigation risk. Second, I find that firms with higher litigation risk do not underprice their IPOs by a greater amount. The results are robust to controls for withdrawal risk and the potential endogeneity of litigation risk in the underpricing regression. Given the empirical support found for the litigation risk hypothesis during the 1988–1995 sample period by Lowry and Shu (2002), I explore what may cause the lack of support for the hypothesis in the 1996–2005 sample period. Several institutional changes may contribute to the declining use of underpricing as litigation insurance. First, the increasing use of directors and officers (D&O) insurance and errors and omissions (E&O) insurance may reduce the need to insure through underpricing, and these insurance fees are likely to be substantially lower than the cost of additional IPO underpricing. Second, the passage of the Private Securities Litigation Reform Act of 1995 (PSLRA) has increased the hurdles to successful private litigation for securities fraud, thus reducing expected litigation costs faced by issuers and underwriters.1 This reduced legal threat can also reduce the incentives to leave money on the table as an indirect form of litigation insurance. When IPO market conditions are weak, firms with higher withdrawal risk underprice their offers by a greater amount to increase the probability of completing the offer. Specifically, a one standard deviation increase in the predicted probability of withdrawal is associated with an increase of 4.94% in underpricing. This relation is not present when IPO market conditions are strong. Third, firms with higher litigation risk and withdrawal risk pay higher gross spreads to underwriters. Withdrawal risk affects gross spreads more strongly than litigation risk does. The effect of withdrawal risk on gross spreads is not only statistically, but also economically significant, more specifically, a one standard deviation increase in the probability of withdrawal increases the expected gross spread by 9.1 basis points (or 1.3% of the average gross spread). Based on the average proceeds of $104 million (2005 purchasing power), this suggests that a one standard deviation increase in the probability of withdrawal increases the dollar amount of underwriter revenue by $94,640 on completed deals. The findings suggest that underwriters are charging fees that reflect the probability of not getting paid. The remainder of this article is as follows. In Section 2, I develop the hypotheses. In Section 3, I describe the data and report the empirical results. Section 4 concludes the paper. 2. Hypothesis development Litigation risk can be a significant concern in the process of going public. Lowry and Shu (2002) report that 5.8% of the firms that went public during 1988–1995 were sued in class actions and the average settlement-to-IPO proceeds ratio was 10% for those sued. The indirect costs of being sued, such as lost management time and damaged reputation, can be much larger. In this section, I discuss how legal concerns can affect several decisions in the IPO process. I also discuss how withdrawal risk can affect IPO pricing and gross spreads. 2.1. Litigation risk and IPO withdrawal When an IPO is withdrawn, issuers most often refer to “current market conditions” as the reason. Unfavorable market movements not only can reduce the proceeds raised in an IPO, but also can affect the after-market stock performance. Given that a necessary condition for litigation against the issuers is poor after-market stock performance, deciding whether to complete or withdraw an offer when negative information arrives during the registration period is critical in reducing future legal liabilities for issuers and underwriters alike. However, withdrawing an IPO also has associated costs. First, withdrawing an IPO can delay a profitable investment project due to the inability to immediately tap alternative sources of financing. This opportunity cost could be substantial for firms in nascent industries whose early entrants can enjoy significant first-mover advantages. Second, withdrawing an IPO can put cash-strapped firms in financial distress. Guo (1998) finds the survival rate for firms that withdraw to be lower than that of firms that complete their issues. A withdrawn issuer not only loses the chance to raise

1 The PSLRA stimulated heated debate on whether it would benefit or harm shareholders, as this legislation not only may reduce frivolous lawsuits; it also may deter meritorious lawsuits (see, e.g., Spiess and Tkac, 1997; Johnson et al., 2000; Ali and Kallapur, 2001). Nevertheless, litigation risk for issuing firms and underwriters would be reduced by the passage of the PSLRA regardless of its net effect on shareholder wealth.

440

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

public equity, it also loses the registration fees, legal and accounting expenses, and management time devoted to the offering process. The following example illustrates the issuer's stress from a failed IPO. “We were out of money for almost four months after the IPO fell through,” says Cote, 54. After falling four weeks behind on payroll, Cote laid off 24 of his 67 employees. And as part of a last-ditch bid to avoid bankruptcy, he and his two business partners signed over the rights to second mortgages on their homes to a lender. “We were trying to do the IPO to overcome our cash shortage,” broods chairman Andrew Young. “Instead, this just exacerbated it.” (Jerry Useem, “All dressed up and no IPO (initial public offering failures),” February 1998, Inc., Vol. 20, No. 2.) A third cost of withdrawing an IPO is bad publicity and increased uncertainty about firm valuation, which reduces the opportunity to tap public securities markets in the near future. Lerner (1994) states: “A firm that withdraws its IPO may later find it difficult to access the public marketplace. Even if the stated reason for the withdrawal is poor market conditions, the firm may be lumped with other businesses whose offerings did not sell because of questionable accounting practices or gross overpricing.” Consistent with this viewpoint, Dunbar and Foerster (2008) find that only about 9% of withdrawn IPOs are able to return to have a successful IPO. Moreover, Lian and Wang (2009) examine the IPO valuations of issuers that return to the market successfully after withdrawing their first IPO attempt. They find that the negative connotations of the first withdrawal translate into lower valuations for second-time IPOs. The median offer price multiples for second-time IPOs are 15%–49% lower than those of first-time IPOs. Besides issuers, underwriter incentives also affect the withdrawal decision. On the one hand, underwriters have incentives to withdraw offers facing weak demand. If the underwriter struggles to complete an offer with weak pre-market demand, it can incur greater costs later. If the after-market price drops below the offer price, the costs can be in the form of losses on unsold shares and after-market price support, reputation damage, expected legal costs and expected increases in professional liability insurance premiums, if litigation is later brought against the underwriters. On the other hand, an underwriter of a withdrawn IPO not only loses the chance to be paid while continuing to pay for the employee time and expense devoted to the offering process, but the underwriter can also suffer from negative publicity and reduced reputation capital. Although it can be costly to withdraw an IPO, it can be risky to complete a premature IPO too. All else being the same, I expect the risks of completing an IPO with weak demand, including the legal risk, to be positively related to the probability of withdrawal. Hypothesis 1. All else being the same, IPO litigation risk is positively related to IPO withdrawal risk. Given that offers with the greatest litigation risk are most likely to be withdrawn in the first place, there is a selection bias induced since these suits are not observed. Consequently testing for a positive relation between litigation risk and withdrawal probability based on a truncated sample of completed IPOs, could work against finding any significant results. Nevertheless, I find that withdrawal probability is positively and significantly related to litigation risk, even in the case of completed IPOs. 2.2. Litigation risk, withdrawal risk, and IPO underpricing 2.2.1. Litigation risk and IPO underpricing The litigation risk hypothesis posits an explanation for IPO underpricing that focuses on legal liability. It argues that IPO firms underprice their new issues to deter potential lawsuits. However, there is mixed evidence in the literature. For example, Tinic (1988) presents evidence consistent with the hypothesis, while Drake and Vetsuypens (1993) argue against the hypothesis. Lowry and Shu (2002) provide a comprehensive review of the literature. Using a simultaneous equation approach to address the endogeneity between litigation risk and underpricing, Lowry and Shu (2002) find supporting evidence for the litigation risk hypothesis. Their methodology is an important improvement over the prior studies. However, as Ljungqvist (2007, p. 44) points out, “Lowery and Shu's instruments would appear to be weak,” which “may aggravate the effect of simultaneity bias, rather than solving it. To be considered strong, an instrument needs to be highly correlated with the first-stage endogenous variable.” Examining a more recent sample period than Lowery and Shu's, Zhu (2009) and Pukthuanthong et al. (2009) find evidence against the litigation risk hypothesis. The legal underpinning of the litigation risk hypothesis is Section 11 of the Securities Act of 1933, under which damages are capped by the IPO offer price.2 The hypothesis has two dimensions. First, decreasing the offer price could reduce the chance that the after-market price will drop below the offer price and lower the probability of being sued. The reason is that if the after-market price is above the offer price, then litigation under Section 11 would have no standing. Second, decreasing the offer price could reduce the maximum damages that plaintiffs can recover under Section 11 in case the after-market price drops below the offer price. In summary, underpricing reduces the expected damages under Section 11, therefore reducing the likelihood of Section 11 lawsuits. However, there are avenues for lawsuits against issuers beyond Section 11. For example, class action suits pertaining to IPOs almost always include claims under Section 10(b) of the Securities Exchange Act of 1934 and Rule 10b-5 promulgated thereunder in addition to Section 11 of the Securities Act of 1933 (Alexander, 1993). Damages from 1934 Act suits are not capped by the offer price. Rather, they are determined by the difference between the price paid and the price the plaintiffs would have paid if there

2

Lowry and Shu (2002, p. 311) explain the guidelines for the calculation of associated damages under Section 11.

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

441

had been no wrongdoing. If damages under the 1934 Act are sufficiently high, even high underpricing can fail to deter litigation.3 Therefore, lowering the offer price can only reduce but not eliminate the expected litigation cost under the 1934 Act. More importantly, two institutional changes may contribute to a declining use of underpricing as litigation insurance. First, directors and officers (D&O) insurance, which in the late 1970s and early 1980s was thought to be “novel,” is now considered a business necessity. Even if underpricing can help to reduce litigation liability, the wide usage of D&O insurance among companies may reduce the incremental level of underpricing needed purely for insurance purposes. Chalmers et al. (2002) find a significant negative relation between the three-year post-IPO stock price performance and the D&O insurance coverage purchased in conjunction with an IPO, which is consistent with the hypothesis that managers of IPO firms have superior inside information and use D&O insurance to insure against potential litigation costs. D&O insurance typically covers management and thus affects management incentives. A D&O policy, however, can extend coverage for claims against the company itself (Bordon et al., 1998). According to the insurance advisory company TillinghastTowers Perrin 2001 Directors and Officers Liability Survey, more than 90% of U.S. companies that buy D&O insurance also buy coverage for the corporate entity itself. Furthermore, there is likely to be a spillover effect of the D&O insurance from management to the firm itself because of lower management risk exposure. Besides issuers, underwriters also have incentives to insure against litigation. Presumably, underpricing can help not only issuers, but also underwriters to insure against litigation under the 1933 Act.4 However, the increasing use of D&O and errors and omissions (E&O) insurance can reduce underwriters' incentives to rely on underpricing as litigation insurance.5 IPO firms used to carry D&O insurance policy with sub-limited coverage for the underwriters. Today, underwriters typically carry their own policies that combine D&O and E&O insurance.6 The second institutional change that may contribute to a declining use of underpricing as litigation insurance is the passage of the PSLRA in December 1995. PSLRA increased the hurdles shareholders would need to overcome to pursue private litigation claims of securities fraud, thus reducing the expected litigation costs to issuers and underwriters. Beatty et al. (2003) find both univariate and multivariate evidence consistent with a reduction in IPO litigation cases subsequent to the passage of PSLRA. Although neither D&O and E&O insurance, nor the PSLRA protects issuers or underwriters from being sued, they can reduce the incentives for leaving money on the table as an indirect form of insurance. How do practitioners think about the importance of litigation risk in IPO pricing? In a survey of the chief financial officers (CFOs) of the nonfinancial U.S. companies that had successfully completed an IPO or attempted and subsequently withdrew an IPO between 2000 and 2002, Brau and Fawcett (2006) find that the litigation risk hypothesis for underpricing receives low support among the CFOs who responded. This raises further doubts about the validity of this hypothesis. Is there international evidence for the litigation risk hypothesis? One significant piece of evidence is that underpricing occurs even in countries where litigation risk is not a concern (Keloharju, 1993). Thus, litigation risk is clearly not the only cause of underpricing. However, it remains an open empirical question as to whether litigation risk is one important factor that significantly affects underpricing in the U.S. 2.2.2. Withdrawal risk and IPO underpricing The decision to withdraw an offer can be made by issuers, underwriters, and occasionally the SEC. When an offer is withdrawn, issuers lose the chance of going public, at least temporarily. How does the risk of withdrawal affect IPO pricing? Intuitively, all else being the same, issuers with a higher probability of withdrawal can underprice more to increase the probability of completion, because investor demand for a stock should be increasing in the stock's value-to-price ratio. Besides issuers, underwriters also have incentives to underprice offers with weaker demand by a greater amount. Although more underpricing means less underwriting fees paid to underwriters, this may be more than recaptured if underwriters can extract the value associated with underpricing from its customers. Moreover, for deals with greater withdrawal risk, underwriters should be compensated more when the deals are successfully completed, since they not only exert more extensive marketing efforts on the completed deals but also lose money on the deals that are withdrawn. The compensation can come either indirectly through more money left on the table (giving underwriters more soft dollar revenue from rent-seeking investors), or directly through higher gross spreads. Hypothesis 2. All else being the same, IPO withdrawal risk is positively related to underpricing.

3 Each Act has different advantages and disadvantages from a plaintiff's point of view. For example, the burden of proof is lower for claims brought under the 1933 Act, while damages tend to be higher for claims brought under the 1934 Act. 4 Underwriters can underprice IPOs for various reasons. For example, more underpricing increases returns to IPO investors and reduces the likelihood that IPO investors are dissatisfied with the poor subsequent issuer performance and stop investing in the underwriters' later IPOs. Alternatively, underpricing can represent indirect underwriter compensation since underwriters can through their allocation powers extract the economic benefits associated with underpricing (Loughran and Ritter, 2004; Reuter, 2006; Nimalendran et al., 2007; Liu and Ritter, 2010). 5 “Errors and omissions insurance is a coverage that protects those people that give advice, make educated recommendations, design solutions or represent the needs of others. As the name suggests, it protects these people when they've done something they shouldn't have (error) or when they neglected to do something they should have (omission). It is also referred to as Professional Liability or Malpractice Insurance” (http://law.freeadvice.com). 6 I thank Kevin M. LaCroix, an attorney and a partner in Oakbridge Insurance Services, for making this comment.

442

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 1 Distribution of completed and withdrawn IPOs filed with the SEC from 1996 to 2008. The sample consists of completed and withdrawn IPOs filed during 1996– 2008, excluding unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). IPO filings with information on underwriter names and the dollar amount of the filing missing from the SDC database are also excluded. Year filed

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 1996–2005 2006 2007 2008 2006–2008

(1)

(2)

(3)

(4)

(5)

Total issues filed

Completed issues

Withdrawn issues

Percentage completed (%)

Percentage withdrawn (%)

696 464 305 521 410 63 63 78 155 123 2878 186 159 61 406

600 382 208 443 264 47 42 65 123 110 2284 140 91 6 237

96 82 97 78 146 16 21 13 32 13 594 46 68 55 169

86.2 82.3 68.2 85.0 64.4 74.6 66.7 83.3 79.3 89.4 79.4 75.3 57.2 9.8 58.4

13.8 17.7 31.8 15.0 35.6 25.4 33.3 16.7 20.7 10.6 20.6 24.7 42.8 90.2 41.6

2.3. Litigation risk, withdrawal risk, and gross spreads 2.3.1. Litigation risk and gross spreads Given that litigation risk is an added cost beyond the underwriting risk borne by underwriters, it is reasonable to expect litigation risk to affect underwriting fees. However, whether the effect is statistically or economically significant remains an empirical issue. Two pieces of empirical evidence support the hypothesis that litigation risk is related to gross spreads. First, IPO gross spreads in other countries are much lower than in the U.S. (Torstila, 2003). One reason frequently advanced to explain the higher underwriting costs in the U.S. is the greater potential for lawsuits.7 Consistent with this hypothesis, Torstila (2001) finds that European issuers generally pay significantly lower gross spreads on European listings than on U.S. listings. Even though issuing companies' D&O insurance can help to cover at least a portion of the settlements in typical class action lawsuits, underwriters also pay for their own professional liability insurance, which represents part of their underwriting expenses. Thus, it is reasonable to expect that litigation risk affects underwriting fees.8 Second, underwriter gross spreads in the U.S. decreased after the passage of the PSLRA in December 1995. Since the PSLRA increases the hurdles to private litigation for securities fraud, it should reduce the compensation underwriters require to bear litigation risk.9 Hypothesis 3. All else being the same, IPO litigation risk is positively related to underwriter gross spreads. However, the entry of commercial banks into the securities underwriting market also reduced gross spreads (Kim et al., 2008). As individual commercial banks were allowed to underwrite IPOs through Section 20 subsidiaries during the 1990–1998 period, they raised IPO underwriting competition, which was further heightened after the passage of the Financial Services Modernization Act in 1999, when commercial banks were generally allowed to compete with investment banks for equity underwriting. However, the effect of commercial bank entry on reducing gross spreads was likely to initially be rather gradual as individual banks received permission to underwrite equity offers on a case-by-case basis over the 1990–1998 period, while the passage of the PSLRA was more likely to affect underwriter expected costs immediately in 1996. Nevertheless, I also control for the effect of commercial bank entry in the empirical test. 2.3.2. Withdrawal risk and gross spreads The relation between withdrawal risk and gross spreads is straightforward. IPOs with a greater risk of withdrawal require more extensive marketing efforts by underwriters, subject the underwriters to greater inventory risk and price support expense as well as producing negative effects on underwriter reputations and client relationships. Thus, such IPOs should exhibit greater gross spreads for completed deals (and even more money left on the table) to compensate underwriters for the forgone revenue on deals that get withdrawn and the higher costs of underwriting weak deals. Hypothesis 4. All else being the same, IPO withdrawal risk is positively related to underwriter gross spreads. In the following section, I empirically examine the relations between litigation risk and the withdrawal decision, offer pricing, and gross spreads. 7 8 9

I thank James Leigh-Pemberton, Chairman of Credit Suisse in the Investment Banking division (based in London) for helpful comments on this issue. However, I do not claim that the difference in the gross spreads between the U.S. and other countries is entirely due to lawsuit potential. This can make it harder to detect the relation between litigation risk and underwriter spread.

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

443

Table 2 Distribution of outcomes for withdrawn IPOs. The sample consists of 594 withdrawn IPOs filed during 1996–2005. This table reports what happened to the firms in the three years following an IPO withdrawal. “Another IPO” in Column (1) refers to the firms that were able to return later for a successful offering. “Sale of firm/ merger” in Column (2) refers to the firms that either were acquired by or merged with another firm. “Bankruptcy/financial distress” in Column (3) refers to the firms that filed for bankruptcy or experienced significant financial distress later. “Other” in Column (4) refers to the firms whose business was still operating without incurring the events in Columns (1)–(3). “Unknown” in Column (5) refers to the firms for which no information is found to confirm that their business was still operating. Year Filed

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Total Percent of withdrawn

(1)

(2)

(3)

(4)

(5)

Another IPO

Sale of firm/merger

Bankruptcy/financial distress

Other

Unknown

8 4 13 3 5 1 0 2 2 4 42 7%

31 31 20 24 29 5 6 2 8 3 159 27%

8 2 5 7 8 1 1 1 1 0 34 6%

45 35 30 25 76 7 5 2 15 3 243 41%

4 10 29 19 28 2 9 6 6 3 116 19%

3. Empirical analysis 3.1. Data My data cover withdrawn and completed IPOs of U.S. industrials filed with the SEC during 1996–2005. The sample excludes unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). The sample starts in 1996 because the data source for private securities class action lawsuits – Stanford Law School Securities Class Action Clearinghouse (in cooperation with Cornerstone Research) – starts its coverage in 1996. The sample ends at the end of 2005 because the securities laws limit the time between the IPO and the lawsuit filing to three years (see U.S. Supreme Court decision regarding Lampf, Pleva, Lipkind, Prupis & Petigrow v. Gilbertson, 501 US 350 on 6/30/1991). I manually searched through all of the filings at the Securities Class Action Clearinghouse as well as their related court dockets during 1996–2008. Thus, I was able to identify all the class action lawsuits related to IPOs completed during 1996–2005. Data for IPOs are from Thomson Financial's Securities Data Company (SDC) New Issues database. Numerous corrections have been made to the SDC's IPO data based on Jay Ritter's and Alexander Ljungqvist's SDC IPO data correction files, which are downloadable from their websites. The Carter and Manaster (1990) underwriter reputation measures are downloadable from Jay Ritter's website. To identify if there is a star analyst on the lead underwriter's team, I use Institutional Investor's all-America research team data. I use the Center for Research in Security Prices (CRSP) database for data on share prices, returns, and trading volume. I obtain Moody's indices for seasoned corporate bond yields with an Aaa rating or Baa rating and the 10-year (constant maturity) Treasury yield constructed by the Federal Reserve from the Wharton Research Data Services (WRDS). I also use various sources of information (SDC, EDGAR, Factiva, LEXIS/NEXIS, Google, etc.) to manually identify whether the firms that withdraw their IPOs subsequently complete another IPO, combine business with another firm, or go bankrupt/experience significant financial distress in the three years following the IPO withdrawal.

3.2. The decision to withdraw Table 1 presents the annual breakdown of withdrawn versus completed issues. The percentage of issues withdrawn ranges from a minimum of 10.6% for those filed in 2005 to a maximum of 35.6% for those filed in 2000, with the average being 20.6%. In other words, about one in every five IPOs filed during the sample period of 1996–2005 is withdrawn.10 Table 2 reports what happened to the firms in the three years following an IPO withdrawal. Only 7% of the firms that withdrew an IPO are able to return to the market later for a successful offering, and 6% of the firms that withdrew an IPO filed for bankruptcy or experienced significant financial distress later. Note that this 6% is only a lower bound for the percentage of firms that filed for bankruptcy or experienced significant financial distress, because two other categories are likely to contain withdrawn firms experiencing financial distress. (1) About 27% of the withdrawn firms either were acquired by or merged with another firm. It is not clear how many of them made the sale/merger decision because of financial distress. (2) Approximately 19% of the firms that

10 To provide more up-to-date statistics on IPO withdrawals, Table 1 also reports the annual breakdown of withdrawn versus completed issues filed during 2006–2008.

444

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 3 Descriptive statistics of completed and withdrawn IPOs filed with the SEC from 1996 to 2005. The sample consists of 2878 completed and withdrawn IPOs filed during 1996–2005, excluding unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). IPO filings with information on underwriter names and the dollar amount of the filing missing from the SDC database are also excluded. All variables are defined in Appendix A. Statistics for Bank excess underpricing are based on fewer observations because of missing data.

Issuer and issue characteristics Filing amount Technology dummy Utility dummy Investment bank characteristics Carter–Manaster rank Bank market share Bank excess underpricing Missing_excess underpricing dummy Market conditions during registration period Number of IPO filings Ratio of withdrawn IPOs to completed IPOs and IPOs in active registration Average offer price revision Average underpricing Industry return Industry daily return volatility Change in BAA-AAA yield spread Change in AAA-10 year Treasury yield spread

Completed offerings (N = 2284)

Withdrawn offerings (N = 594)

Mean

Mean

93.5 51.4% 1.0%

Median 55.6 1 0

75.7 38.7% 0.3%

Median 50.7 0 0

7.4 4.5% 8.3% 14.5%

8.0 1.7% 9.2% 0

6.9 3.2% 6.3% 21.5%

8.0 0.5% 5.1% 0

74.5 25.3% 4.3% 34.3% 5.7% 1.1% −0.8% −0.7%

76.0 16.4% 2.6% 17.7% 5.5% 0.9% −1.0% 1.0%

62.1 91.2% −0.2% 30.7% −1.8% 1.4% 1.1% 5.2%

63.5 34.7% −3.3% 16.6% −1.7% 1.1% 0.0% 3.0%

withdrew their IPOs lack any information to confirm that their business was still operating in the three years following the IPO withdrawal, based on information from various sources (SDC, EDGAR, Factiva, LEXIS/NEXIS, Google, etc.). It is likely that some of these private firms were too small to be reported in the financial press, even if they experienced significant financial distress or even went bankrupt. Table 3 reports descriptive statistics for the IPO filing sample by IPO completion or withdrawal status. Three issuer and issue characteristic variables are related to deal riskiness. Filing amount equals the average filing price multiplied by the number of shares to be sold as indicated in the initial filing (2005 purchasing power). The results show that the withdrawn issues tend to be smaller than the completed issues. The technology dummy is set equal to one using the criteria in Loughran and Ritter (2002), and zero otherwise. The utility dummy is set equal to one if the issuer's SIC code starts with 49, and zero otherwise. The results show that fewer offers are withdrawn by technology and utility firms. Instead of interpreting the univariate results here, I will rely on the multivariate regressions later to understand the characteristics associated with the offer withdrawal decision. Several investment bank characteristic variables are intended to capture important underwriter characteristics. The Carter– Manaster rank assigns higher prestige to underwriters that are listed more prominently on tombstone advertisements (Carter and Manaster, 1990). The reputation ranks range from 1 (lowest) to 9 (highest). The bank market share is the lead underwriter's IPO market share (based on proceeds) in the calendar year prior to the filing year (Megginson and Weiss, 1991). The bank excess underpricing is the lead underwriter's average excess IPO underpricing in the calendar year prior to the filing year (Hoberg, 2007).11 Lead underwriter IPO underwriting activity in the previous year is adjusted for bank underwriter mergers and acquisitions based on information in Corwin and Schultz (2005), Ljungqvist et al. (2006), and the SDC Mergers and Acquisitions database. Some lead underwriters do not have completed IPOs in the previous year. Rather than discarding these observations, I include dummy variable missing_excess underpricing dummy that is set equal to one when the lead underwriters do not have completed IPOs in the previous year, and zero otherwise. Market conditions during the registration period are presumably critical to the withdrawal decision. Following Benveniste et al. (2003), Edelen and Kadlec (2005), and Dunbar and Foerster (2008), I use eight variables to reflect market conditions at different levels. If the withdrawal date is not specified in the SDC database, I assume that it is 270 calendar days after filing, which is the time period the SEC allows before automatically declaring the offering abandoned (see Lerner, 1994). All of the eight market condition variables are measured over 60 calendar days prior to offer completion or withdrawal date. Standardizing the measurement period to 60 calendar days is a way to circumvent the potential endogeneity of the length of the pre-market, as it could take longer to presell an issue when investor interest is weak. To capture the spillover information from the filing, withdrawing, and pricing of other IPOs, I use the number of IPO filings, ratio of withdrawn IPOs to completed IPOs and IPOs in active registration, average offer price revision, and average underpricing. Industry return and Industry daily return volatility capture the changes in the stock market environment for the issuer's industry. Specifically, I assign firms from the CRSP universe to the 48 industry classes as defined by Fama and French (1997). I then compute

11

If I use the lead underwriter's average excess IPO underpricing in the prior 2 or 3 years, the main results are qualitatively unchanged.

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

445

Table 4 Probit analysis of the decision to withdraw an IPO. The dependent variable equals one for IPO filings that are withdrawn or postponed and zero for completed offerings. The sample consists of 2878 completed and withdrawn IPOs filed during 1996–2005, excluding unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). IPO filings with information on underwriter names and the dollar amount of the filing missing from the SDC database are also excluded. All variables are defined in Appendix A and measured as decimals. The rightmost column reports the mean marginal effects, which are the changes in the probability of withdrawal associated with a unit increase in each independent variable. The mean marginal effect is computed as the sample mean of the individual marginal effects evaluated at every observation. The Huber–White–Sandwich robust standard errors are used to compute p-values. χ2 statistics with significance at the 1%, 5%, and 10% levels are denoted with ⁎⁎⁎, ⁎⁎, and ⁎, respectively. Coefficient Intercept Issuer and issue characteristics Ln(Filing amount) Technology dummy Utility dummy Investment bank characteristics Carter–Manaster rank Bank market share Bank excess underpricing Missing_excess underpricing dummy Market conditions during registration period Ln(Number of IPO filings) Ratio of withdrawn IPOs to completed IPOs and IPOs in active registration Average offer price revision Average underpricing Industry return Industry daily return volatility Change in BAA-AAA yield spread Change in AAA-10 year Treasury yield spread Year fixed effect Pseudo R2 Percent concordant Observations

Marginal effect

0.03 −0.13⁎⁎⁎ −0.46⁎⁎⁎ −0.85⁎⁎

−0.03 −0.11 −0.20

−0.03 −1.64⁎⁎⁎ 0.14 0.07

−0.01 −0.39 0.03 0.02

−0.13⁎ 0.40⁎⁎⁎ −0.15 −0.43⁎⁎ −1.24⁎⁎⁎ 13.72⁎⁎

−0.03 0.09 −0.04 −0.10 −0.29 3.25 0.11 0.10

0.47 0.41⁎⁎ Yes 15.4% 76.3 2878

the cumulative equal-weighted return with daily rebalancing and the standard deviation of the daily returns for firms in the same industry as the issuer over the 60 calendar days prior to offer or withdrawal. The changes to interest rate environments are captured by the change in BAA-AAA yield spread and change in AAA-10 year Treasury yield spread, where the BAA-AAA yield spread is the difference between the Moody's indices for seasoned corporate bond yields with Baa and Aaa ratings, and the AAA-10 year Treasury yield spread is the difference between the Moody's index for seasoned corporate bond yield with an Aaa rating and the 10-year (constant maturity) Treasury yield constructed by the Federal Reserve. To study the decision to withdraw an IPO, I estimate a probit model in which the dependent variable takes the value of one if the IPO is withdrawn and zero otherwise. Table 4 presents the estimation results, including the mean marginal effects, which are the changes in the probability of withdrawal associated with a unit increase in each independent variable. The mean marginal effect is computed as the sample mean of the individual marginal effects evaluated at every observation.12 Examining the probit estimates, the negative coefficient estimate for Ln(Filing amount) suggests that issuers with larger offerings are less likely to withdraw. Larger issues tend to attract institutional investors, who contribute to information production, thus lowering the issue's risk. The negative coefficient estimate for the technology dummy suggests that technology issuers are less likely to withdraw. The ramifications of withdrawing an IPO could be bad publicity, affecting the way suppliers, clients, and possibly employees interact with the company. Employees and suppliers are also likely to have firm-specific investments, making withdrawal very costly if it raises the likelihood of financial distress (Titman and Wessels, 1988). Offerings underwritten by banks with higher reputations or market shares can yield greater issue certification. This is consistent with the negative coefficient estimates for the lead underwriter's reputation rank and market share. The coefficient estimate for the lead underwriter's market share is statistically significant, however, while the coefficient estimate for the lead underwriter's Carter–Manaster rank is not significant at any conventional level.13 If only one of the two variables is included in the regression, the coefficient estimate is negative and statistically significant in either case, consistent with the idea that the two reputation measures are strongly correlated. In the rest of the paper, I include only an underwriter's market share in regression models, but replacing the underwriter's market share with Carter–Manaster rank does not change the main results. Market conditions during the registration period significantly affect the likelihood of an IPO withdrawal. This is consistent with the observation that when an IPO filing is withdrawn, issuers most often refer to “current market conditions” as the reason. 12 As Greene (2003, p. 668) explains, “for computing marginal effects, one can evaluate the expressions at the sample means of the data or evaluate the marginal effects at every observation and use the sample average of the individual marginal effects. …in large samples these will give the same answer. But that is not so in small or moderate-sized samples. Current practice favors averaging the individual marginal effects when it is possible to do so.” 13 Dropping the Carter–Manaster underwriter rank from the regression model makes almost no difference to the estimated coefficients on the other regressors.

446

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 5 Distribution of sued IPO firms. The sample consists of 2276 IPOs filed and completed during 1996–2005, excluding unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). IPO filings with information on underwriter names and the dollar amount of the filing missing from the SDC database are also excluded. The sample size is slightly smaller than the completed IPO sample size in Table 1, because some IPOs in Table 1 that are filed during 1996–2005 are completed after 2005. A sued IPO firm is defined as a firm sued in class actions for violations relating to the IPO due to misleading information about the issuing firms' business (under the Securities Act of 1933 or the Securities Exchange Act of 1934). IPO firms that are involved in laddering cases because their lead underwriters engaged in the price manipulative laddering agreement are excluded from the sued issuer sample. Year issued

Total number of IPOs

Number of IPO firms sued for issuers' wrongdoing

Percentage sued (%)

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Total

491 396 233 411 331 60 48 47 136 123 2276

15 12 11 12 15 6 4 6 8 6 95

3.1 3.0 4.7 2.9 4.5 10.0 8.3 12.8 5.9 4.9 4.2

Unfavorable market movements not only can reduce the proceeds raised in an IPO, but also may negatively affect the after-market stock performance. Very unfavorable market conditions can lead underwriters to refuse to underwrite many issues. Several coefficient estimates are consistent with a strong effect of information spillovers. The coefficient estimate for the Ln (number of IPO filings) is significantly negative. This suggests that information spillovers become more significant in markets with more filings, resulting in enhanced precision in IPO valuation. Thus, withdrawals are less likely. The ratio of withdrawn IPOs to completed IPOs and IPOs in active registration is strongly, positively related to the probability of withdrawal. A one standard deviation increase in this ratio raises the withdrawal probability by 8%. In addition, a one standard deviation increase in the average underpricing of other IPOs reduces the withdrawal probability by 3.2%; a one standard deviation increase in industry return reduces the withdrawal probability by 4.4%; a one standard deviation increase in industry daily return volatility raises the withdrawal probability by 2.3%. The yield spread can often predict the risk of default in the economy. Default probabilities are usually higher when the yield spread is substantial. If firms are more likely to have negative news in such market environments, withdrawals are more likely when spreads are higher. The results are consistent with this viewpoint. 3.3. Litigation risk The typical allegation against IPO firms in class action suits is that material information about the issuing firm's business was false or misleading. Table 5 shows that at least one in every twenty-five firms that went public during 1996–2005 was sued with this as a major allegation. However, in year 2001, hundreds of class actions were filed against firms that went public during 1999– 2000. These lawsuits alleged laddering (Hao, 2007), analyst conflicts of interest, and undisclosed underwriter compensation from soft dollar commissions (Reuter, 2006; Nimalendran et al., 2007). Although these lawsuits have several allegations, they are often referred to as laddering cases for convenience. I will refer to them as laddering cases here after. Because laddering lawsuits focus more on underwriters rather than on issuers, they are less comparable to IPO lawsuits during the periods before and after the bubble period of 1999–2000. Thus, the focus of this study is the litigation risk arising from false or misleading information about the issuing firm's business disclosed in connection with its IPO. I do not consider issuers that are involved in lawsuits where the sole reason for the suit is underwriter laddering as “sued firms.” However, a few of the issuers involved in laddering cases are also sued for false or misleading information about the firm's business; these firms are treated as “sued firms.” Later in the robustness section, I address possible concerns about this classification. Table 6 provides descriptive statistics, broken down by whether or not an issuer is sued for false or misleading information about its business. The sued and non-sued issuers exhibit several characteristics. For example, sued issuers are bigger in size and raised more IPO proceeds; the lead underwriters of sued issuers have higher reputations and larger market shares; furthermore, a higher percentage of the sued issuers are venture capital (VC)-backed firms. These findings are consistent with the “deep pocket” theory of lawsuits. In addition, sued firms have lower post-IPO one-year returns than nonsued firms, consistent with the necessary (but not sufficient) condition for a lawsuit filing under a claim of false or misleading information. Although Table 6 demonstrates some univariate differences between the sued and nonsued firms, these differences could be correlated. Therefore, I do not want to over-interpret the univariate differences here. Later I will rely on multivariate regressions to investigate the relations. One objective of this study is to investigate the relations between litigation risk and key IPO decisions. One of these decisions is IPO underpricing. A complication in estimating the litigation risk and underpricing relation is that they can affect each other. Not only can firms with a higher probability of being sued underprice their IPOs to a greater extent as insurance against litigation, but also firms that underprice more may lower their probability of being sued. When investigating whether litigation risk affects underpricing, I attempt to mitigate the endogeneity issue by using a two-stage regression estimation procedure as in Lowry and

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

447

Table 6 Descriptive statistics of sued versus non-sued firms. The sample consists of 2276 IPOs filed and completed during 1996–2005, excluding unit offers, ADRs, carveouts/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). Some of the variables are based on fewer observations because of missing data. A sued IPO firm is defined as a firm sued in class actions for violations relating to the IPO due to misleading information about the issuing firms' business (under the Securities Act of 1933 or the Securities Exchange Act of 1934). IPO firms that are involved in laddering cases because their lead underwriters fail to disclose the price manipulative laddering agreement are excluded from the sued issuer sample. All variables are defined in Appendix A. Sued IPO firms (N = 95)

Issuer and issue characteristics Gross spread (%) Underpricing (%) Proceeds Assets Age Technology dummy Utility dummy VC-backed dummy Pure primary dummy NYSE/Amex dummy Offer price revision (%) Probability of withdrawal (%) One-year return (%) Investment bank characteristics Carter–Manaster rank Bank market share (%) Bank excess underpricing (%) Missing_excess underpricing dummy All-star analyst dummy Market conditions prior to offer Nasdaq return (%) Comparable firms' return volatility (%) Comparable firms' share turnover (%)

Non-sued IPO firms (N = 2181)

Mean

Median

Mean

Median

7.0 25.8 166.7 294.1 12.4 53.7% 1.1% 52.6% 66.3% 17.9% 7.8 17.4 −44.0

7.0 11.0 72.9 47.5 8.0 1 0 1 1 0 6.3 12.2 −63.8

7.1 32.8 101.5 259.1 12.5 51.5% 1.0% 47.9% 72.5% 14.5% 3.5 16.7 4.4

7.0 12.5 56.5 32.4 7.0 1 0 0 1 0 0.0 13.7 −19.7

7.9 6.3 16.4 9.5% 29.5%

8.5 3.1 13.1 0 0

7.4 4.4 8.1 14.6% 21.7%

8.0 1.7 9.2 0 0

1.2 4.6 55.5

1.1 4.3 55.9

1.4 4.8 54.7

1.9 4.6 55.8

Shu (2002). The variables in the first stage probit regression model of litigation include all the exogenous variables that can affect underpricing and litigation. In addition, an instrumental variable that is uniquely related to litigation is also included. Instrumental variables (IVs) do not have to be part of the economic model, because the essential purpose of the IVs is not economic prediction. The IV method is an econometric technique that can extract variation in the endogenous explanatory variable that is unrelated to the error term in the regression equation, and we can use this variation to estimate the causal effect on the dependent variable (Kennedy, 2003, pp. 167–168). Thus, it is not necessary (although attractive) for the IVs to be observable ex ante.14 This has important implications for the choice of IVs. I use the IPO stock's post-issue one-year return as an IV for litigation risk. Conceptually, an IPO stock's post-issue return is driven by information that is not available to the market at the time of the IPO. If the market is surprised by negative news not long after the IPO, issuers are likely to be alleged as knowing the negative information without disclosing it in the prospectus. For this practical reason, I use the issuing firm's post-IPO one-year stock return to proxy for undisclosed information. A negative post-IPO return is related to allegedly undisclosed negative information, such as risky features of its products, lack of compliance with certain regulations and statutory requirements, unstable relationships with its key customers or suppliers, or a declining demand for its products and services. If issuers fail to disclose such material information, it is likely that underwriters are unable to factor it into their IPO pricing decisions. Thus, the post-issue return could be unrelated to IPO underpricing, although underwriters are likely to be accused of failing to undertake adequate due diligence efforts. Next, I provide statistical evidence to justify the postissue return as a valid IV for IPO litigation risk. Specifically, I focus on the relevance and exogeneity requirements for an IV (Wooldridge, 2009). The key issue is whether the post-issue one-year stock return is an econometrically acceptable IV. Empirically it is strongly related to litigation, but not related to underpricing. For IPOs occurring during 1996–2005, the correlation coefficient between the post-IPO one-year stock return and the lawsuit dummy is −0.11 and is highly significant with a p-value less than 0.0001, while the correlation coefficient between the post-IPO one-year stock return and underpricing is 0.04 and is not significantly different from zero. Later I show that the one-year stock return is also highly correlated with litigation outcomes in a multivariate regression, where the IV's relevance requirement is met. In addition, it also passes the exogeneity test as an IV, which I will explain later. The selected control variables are shown in the IPO literature to be significant determinants of underpricing and shareholder suits. These control variables are: the logarithm of proceeds; the logarithm of assets; the logarithm of one plus firm age; a VC-

14

I greatly appreciate helpful comments from professors Jeffrey Wooldridge and Chunrong Ai.

448

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 7 Probit analysis of litigation risk. The dependent variable is the lawsuit dummy. All variables are defined in Appendix A and measured as decimals. The sample consists of IPOs filed and completed during 1996–2005, excluding unit offers, ADRs, carve-outs/spin-offs, reverse LBOs, partnerships, and financial firms (SIC code 6000–6999). The rightmost column reports the mean marginal effects, which are the changes in the probability of litigation associated with a unit increase in each independent variable. The mean marginal effect is computed as the sample mean of the individual marginal effects evaluated at every observation. The Huber– White–Sandwich robust standard errors are used to compute p-values. χ2 statistics with significance at the 1%, 5%, and 10% levels are denoted with ⁎⁎⁎, ⁎⁎, and ⁎, respectively. Model 1 Probit coefficient Intercept Issuer and issue characteristics Ln(Proceeds) Ln(Assets) Ln(1 + Age) Technology dummy Utility dummy VC-backed dummy Pure primary dummy NYSE/Amex dummy Offer price revision Probability of withdrawal One-year return Investment bank characteristics Bank market share Bank excess underpricing Missing_excess underpricing dummy All-star analyst dummy Market conditions prior to offer Nasdaq return Comparable firms' return volatility Comparable firms' share turnover Pseudo R2 Percent concordant Observations

Model 2 Marginal effect

−1.87⁎⁎⁎

Probit coefficient

Marginal effect

−2.10⁎⁎⁎

0.09 0.01 0.03 0.14 −0.18 0.03 −0.16 −0.02 0.00 0.88⁎⁎ −0.49⁎⁎⁎

0.01 0.00 0.00 0.01 −0.02 0.00 −0.01 −0.00 0.00 0.08 −0.04

0.15⁎ −0.02 0.04 0.15 −0.07 0.06 −0.08 −0.04 0.00 0.75⁎

0.01 0.00 0.00 0.01 −0.01 0.01 −0.01 −0.00 0.00 0.07

0.94 0.27 −0.15 0.01

0.08 0.02 −0.01 0.00

0.71 0.32⁎ −0.05 0.01

0.06 0.03 −0.01 0.00

−0.85 −4.89 −0.67 7.25% 74.1% 2187

−0.07 −0.42 −0.06

−0.33 −3.57 −0.61 2.44% 61.0% 2187

−0.03 −0.32 −0.05

backed dummy; a technology stock dummy; a utility stock dummy; a pure primary offering dummy; a NYSE/Amex listed dummy; the offer price revision; the probability of offer withdrawal (calculated based on the estimation results in Table 4); the lead underwriter's Carter–Manaster rank, and its prior IPO market share and excess underpricing in the preceding year; an all-star analyst dummy (Cliff and Denis, 2004); the cumulative return of the Nasdaq composite index over 15 days prior to the offer; and the average share turnover and standard deviation of returns for firms similar to the issuing firm over the one year prior to the offer. Appendix A contains the definition of these variables. The probit regression estimation results are presented in Table 7. The post-IPO one-year return is highly correlated with litigation in Model 1. For comparison, Model 2 excludes the post-IPO one-year return. The pseudo R2 drops by more than a half, and the concordant percentage drops from 74.1% to 61.0%. In the untabulated incremental chi-square test when comparing Model 1 with Model 2, the chi-square statistic is highly significant (p b 0.0001). The evidence suggests that the post-IPO one-year return meets the relevance requirement for an IV. The post-issue one-year return also passes the exogeneity test as an IV. If the model is exactly identified, it is not possible to test whether an IV for IPO litigation is correlated with the error term in the underpricing equation. However, for the overidentified model, the joint validity of IVs can be examined by the Sargan test. Therefore, I try to create another IV using the generated instrument approach introduced by Wooldridge (2002). To gain more comfort about the reliability of applying the Sargan test, I generate three alternative IVs using the following three models: (1) Model 1 in Table 7, (2) Model 2 in Table 7, and (3) extended Model 2 in Table 7 with the following two added variables: the log of the number of days when the firm is in registration, and the primary shares offered as a percentage of the total shares outstanding at the IPO completion date. Although the two variables are not significantly related to litigation, the litigation model fit is slightly improved. I use both the post-IPO one-year return and one of the three generated IVs to identify IPO litigation. Then I perform the Sargan test three times. The Sargan test requires the residuals from the underpricing equation to be regressed on all the exogenous variables, including the IVs. The Sargan test statistics are 0.0001×2187=0.2187, 0.001×2187=2.187, and 0.0004×2187=0.8748 for the above three tests, respectively. In other words, the null hypothesis that the IVs are uncorrelated with the structural error cannot be rejected at a conventional significance level in any of the three tests. For simplicity, I only report the post-IPO one-year return results for identifying litigation in this study. Several variables can potentially capture the effects of issuer and lead underwriter deep pockets on the probability of IPO litigation. For example, the issuer's proceeds and total assets, whether the IPO is VC-backed, and the underwriter's prior IPO

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

449

market share. However, none of these variables turns out to be a significant predictor of litigation.15 The results are generally consistent with the findings in Lowry and Shu (2002). Although share turnover activity for firms similar to the issuing firm is a marginally significant predictor of litigation during 1988–1995 (Lowry and Shu, 2002), it is not significantly related to litigation in my sample period. It is unclear why the relation should change. One conjecture is that the passage of the PSLRA reduces frivolous lawsuits, thus making share turnover less related to litigation.16 Interestingly, the predicted probability of withdrawal is significantly and positively correlated with litigation. A one standard deviation increase in the probability of withdrawal (13%) raises the probability of being sued by about 1%. Compared with the 4.2% unconditional probability of an IPO firm being sued, the economic impact of a firm having a higher probability of withdrawal is significant. The finding suggests that if an IPO that is likely to be withdrawn is instead completed, then the issuer's odds of being sued increase significantly compared to other IPOs. This is consistent with the notion that withdrawal reflects the riskiness of an IPO. For example, Lian and Wang (2009) find that the negative information conveyed by the withdrawal event is incorporated into lower valuations for second-time IPOs. My finding provides additional evidence that for issuers completing their IPOs, the estimated probability of withdrawal is a significant predictor of shareholder litigation that is brought against the issuer later.

3.4. Litigation risk, withdrawal risk, and IPO underpricing To investigate whether litigation risk significantly affects underpricing, I use the fitted values of the litigation likelihood from the probit model results in Table 7 as instruments in the second-stage OLS underpricing regression. The standard errors are corrected following Maddala (1983, p. 245). The results are reported in Table 8, Panel A, Column (1). Of primary interest is the coefficient on the fitted value of litigation risk. Contrary to the results based on the earlier Lowry and Shu's 1988–1995 sample period, litigation risk does not significantly affect underpricing in my sample period. Columns (2)–(5) show that this result is robust to various alternative regression specifications. In Panel A, Column (1) shows that the coefficient on the withdrawal probability is significantly positive, suggesting that all else being the same, issuers with a greater probability of withdrawal chose to underprice their issues by a greater amount. As a robustness check, I omit the investment bank characteristic variables and offer proceeds from the regression model. Column (3) demonstrates robustness of the results. However, if I further omit the offer price revision from the regression model, then the coefficient on the withdrawal probability becomes negative, as shown in Column (4). Note that omitting the offer price revision also reduces the adjusted R2 by more than a half, consistent with the evidence in the literature that offer price revision is the single most important determinant of underpricing (Ince, 2010). I conjecture that omitting the offer price revision from the regression model may introduce omitted variable bias to the results in Column (4). Presumably, the offer price revision partly reflects the strength of investor response during the pre-market. On the one hand, underpricing is empirically found to be increasing in pre-market demand. On the other hand, the withdrawal probability tends to be higher when pre-market demand is weak. Therefore, without controlling for pre-market demand, a high probability of withdrawal can appear to be associated with low underpricing.17 Given that the strong positive relation between underpricing and offer price revisions is mainly driven by IPOs with strong pre-market demand (Loughran and Ritter, 2002), I further examine if the effect of the withdrawal probability on underpricing is more pronounced for IPOs with weak pre-market demand. Presumably, when the positive relation between underpricing and offer price revisions is weaker, the relation between the withdrawal probability and underpricing can be more clearly detected. To examine whether the positive relation between the withdrawal probability and underpricing is more pronounced when the pre-market demand is weak, I create a dummy variable (Up) that equals one if an offering is priced above the midpoint of the initial offer price range. This dummy variable is then multiplied by the withdrawal probability and added to the underpricing regression specification. Column (5) shows that both the coefficients on the Probability of withdrawal and the interaction term Probability of withdrawal × Up are statistically significant, but with opposite signs. When the pre-market demand is weak, the effect of withdrawal risk on underpricing is economically significant — a one standard deviation increase in the predicted probability of withdrawal (13%) is associated with a 4.94% increase in underpricing. The average company in the sample raised approximately $104 million in total proceeds (2005 purchasing power). Based on the average proceeds, an increase in the probability of withdrawal by one standard deviation increases the money left on the table by about $5.1 million. However, withdrawal risk does not seem to affect underpricing when pre-market demand is strong, as suggested by the coefficient on the interaction term Probability of withdrawal × Up. 15 I replace underwriter's market share with the Carter–Manaster reputation rank, but the coefficient on the Carter–Manaster rank is not statistically significant either. 16 In untabulated analysis, I find that none of the following variables is a significant predictor of IPO litigation: the length of the registration period, the primary shares offered as a percentage of the total shares outstanding, median industry stock price runup prior to the IPO date, median industry stock return volatility prior to the IPO date, median industry ROA (return on asset) volatility prior to the IPO date, recent IPOs' underpricing and after-market returns, prior IPO litigation in the industry, prior IPO litigation for the same underwriter, a variable indicating whether the underwriter is a commercial bank, and officer and director shareholdings. I conjecture two reasons why IPO litigation is hard to predict. First, IPOs with a high ex ante probability of litigation would have been withdrawn in the first place. Second, the main reason for IPO litigation is undisclosed negative information, which may be hard to capture by variables that are publicly observable at the time of the IPO. 17 Statistical tests based on variance inflation factors (VIFs) or tolerances show that multicollinearity is not a problem when both offer price revision and withdrawal probability are included in the regression model.

450

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 8 Ordinary least-squares regressions of underpricing (%) for IPOs during 1996–2005. In Panel A, the dependent variable is percentage underpricing. In Panel B, fitted offer price revision in Column (2) is the estimated value of offer price revision based on the regression coefficient estimates in Column (1). All other variables are defined in Appendix A. Variables that are measured as percentages are marked with (%). For OLS regressions, White (1980) standard errors are used to compute p-values. For the second stage OLS regressions, p-values are based on the standard errors corrected by the methodology in Maddala (1983, p. 245). Statistics with significance at the 1%, 5%, and 10% levels are denoted with ⁎⁎⁎, ⁎⁎, and ⁎, respectively.

Panel A Intercept Issuer and issue characteristics Fitted litigation risk Ln(Proceeds) Ln(Assets) Ln(1 + Age) Technology dummy Utility dummy VC-backed dummy Pure primary dummy NYSE/Amex dummy Offer price revision (%) Probability of withdrawal (%) Probability of withdrawal (%) × Up Investment bank characteristics Bank market share (%) Bank excess underpricing (%) Missing_excess underpricing dummy All-star analyst dummy Market conditions prior to offer Nasdaq return (%) Comparable firms' return volatility (%) Comparable firms' share turnover (%) Adjusted R2 Observations

Dependent variable = Panel B Intercept Issuer and issue characteristics Fitted litigation risk Ln(Assets) Ln(1 + Age) Technology dummy Utility dummy VC-backed dummy Pure primary dummy NYSE/Amex dummy Fitted offer price revision (%) Probability of withdrawal (%) Probability of withdrawal (%) × Up Feedback from contemporaneous IPOs Average offer price revision (%) Market conditions prior to offer Industry return (%) Nasdaq return (%) Comparable firms' return volatility (%) Comparable firms' share turnover (%) Adjusted R2 Observations

(1)

(2)

(3)

(4)

(5)

(6)

2nd stage OLS

2nd stage OLS

2nd stage OLS

2nd stage OLS

2nd stage OLS

OLS

−26.36⁎⁎⁎

−22.11⁎⁎⁎

−37.52⁎⁎⁎

−71.57⁎⁎⁎

−28.09⁎⁎⁎

−27.61⁎⁎⁎

−0.17 −2.93⁎⁎⁎ −0.82 −1.11 8.27⁎⁎⁎ 10.32 6.34⁎⁎⁎ 8.25⁎⁎⁎

−0.37 −3.60⁎ −0.74 −1.20 5.47⁎⁎ 8.40 6.53⁎⁎⁎ 8.29⁎⁎⁎

−0.46

0.35

0.06 −1.29 9.72⁎⁎⁎ 11.60 7.43⁎⁎⁎ 8.76⁎⁎⁎

0.19 −3.75⁎⁎⁎ 12.85⁎⁎⁎ 22.21 8.92⁎⁎⁎ 6.67⁎⁎

−0.22 −2.33⁎⁎ −1.03 −1.00 8.15⁎⁎⁎ 10.90 6.49⁎⁎⁎ 7.80⁎⁎⁎

−2.37 −1.02 −1.01 8.13⁎⁎⁎ 10.93⁎⁎ 6.48⁎⁎⁎ 7.81⁎⁎⁎

−2.53 1.21⁎⁎⁎ 0.26⁎⁎⁎

−2.39 1.20⁎⁎⁎

−1.23 1.25⁎⁎⁎ 0.30⁎⁎⁎

−2.28

−2.76 1.27⁎⁎⁎ 0.38⁎⁎⁎ −0.36⁎⁎⁎

−2.75 1.27⁎⁎⁎ 0.38⁎⁎ −0.36⁎

0.81⁎⁎⁎ 0.03 1.50 11.05⁎⁎⁎

0.76⁎⁎⁎ 0.03 2.29 10.97⁎⁎⁎

0.80⁎⁎⁎ 0.03 1.10 10.81⁎⁎⁎

0.79⁎⁎⁎ 0.03 1.11 10.81⁎⁎⁎

1.03⁎⁎⁎ 3.67⁎⁎⁎ 0.53⁎⁎⁎

1.03⁎⁎⁎ 4.22⁎⁎⁎ 0.55⁎⁎⁎

1.04⁎⁎⁎ 3.50⁎⁎⁎ 0.54⁎⁎⁎

1.04⁎⁎⁎ 3.51⁎⁎⁎ 0.55⁎⁎⁎

46.46% 2187

46.31% 2187

−0.19⁎

0.93⁎⁎⁎ 3.65⁎⁎⁎ 0.52⁎⁎⁎

2.07⁎⁎⁎ 5.79⁎⁎⁎ 1.26⁎⁎⁎

45.49% 2187

20.57% 2187

46.65% 2187

47.11% 2187

(1)

(2)

(3)

1st stage OLS

2nd stage OLS

2nd stage OLS

Offer price revision (%)

Underpricing (%)

Underpricing (%)

−8.61⁎⁎⁎

−53.89⁎⁎⁎

−51.42⁎⁎⁎

−2.52 −2.03⁎⁎ −0.37 4.33 18.93 5.24⁎⁎

−2.18 −2.04⁎⁎ −0.89 0.74 16.52 6.39⁎⁎ 5.06⁎

1.62⁎⁎⁎ −1.89⁎⁎⁎ 8.75⁎⁎⁎ 0.64 2.21⁎⁎

4.13 0.18 1.77⁎⁎⁎ 0.73⁎⁎⁎ −0.55⁎⁎⁎

−0.53 1.27⁎⁎⁎

0.63⁎⁎⁎ 0.33⁎⁎⁎

23.81% 2187

1.78⁎⁎⁎ 1.34 1.06⁎⁎⁎ 26.15% 2187

1.87⁎⁎⁎ 3.07⁎⁎⁎ 1.20⁎⁎⁎ 25.06% 2187

As a robustness check for the independent effect of withdrawal probability on underpricing, I drop the fitted litigation risk variable and run an OLS regression of underpricing without controlling for endogeneity. White (1980) standard errors are used for calculating p-values. Table 8, Panel A, Column (6) shows the results, which are consistent with the two-stage regression results. Without the fitted litigation risk variable, the effect of withdrawal risk on underpricing remains the same.

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

451

Among other explanatory variables, the lead underwriter's Carter–Manaster rank and IPO market share, all-star analyst dummy, offer proceeds, offer price revision, technology dummy, pure primary dummy, pre-offer Nasdaq return, and share turnover and return volatility of comparable firms are strong predictors of underpricing, consistent with the prior literature. To address the concern that the offer price revision is likely to be endogenous, I follow Benveniste et al. (2003) and use twostage regressions. Specifically, I instrument the offer price revision using the predicted value from the regression model in Eq. (1), which is similar to the regression model in Benveniste et al. (2003). Offer price revision = f1 ðrecent IPO reception; industry condition; issuer characteristicsÞ:

ð1Þ

Recent IPO reception is the average prior offer price revision. Industry condition is proxied by the same industry average firm returns. Both variables are measured over the 60 calendar days prior to offer or withdrawal. Issuing firm characteristics, including firm assets and firm age, are used as control variables. I then use the predicted value of the offer price revision in the second-stage OLS regression of underpricing. For simplicity, I also omit investment bank characteristics and offer proceeds from the regression models. Table 8, Panel B shows the regression results for offer price revision and underpricing. The main results remain robust. Because the model is overidentified, I can perform the Sargan test to examine the validity of the IVs for offer price revision. The Sargan test statistic is 0.0001 × 2187 = 0.2187. The null hypothesis that the IVs are uncorrelated with structural error cannot be rejected at any conventional significance level. 3.5. Litigation risk, withdrawal risk, and IPO gross spreads Before conducting the main test for the effect of litigation risk on IPO gross spreads over my main sample, I examine whether IPO gross spreads decreased immediately and significantly after the passage of the PSLRA in December 1995. Because the entry of commercial banks into the securities underwriting market was around the time of the passage of PSLRA, the coincidence of the two events makes it difficult to test if there is an independent effect of the PSLRA on IPO gross spreads. The speed of the effect of the two events, however, can be different. If underwriting spreads are related to litigation risk, then a sudden reduction in litigation risk due to regulatory change should be accompanied by an immediate decrease in gross spreads. In contrast, the effect of commercial bank entry on reducing gross spreads is likely to be more gradual over time because individual banks slowly received permission to underwrite on a case-by-case basis over the 1990–1998 period. Therefore, I test for the year-by-year change in IPO underwriting spreads over 1989–1999, which covers the period of 1990–1998, when commercial banks were allowed to underwrite IPOs through Section 20 subsidiaries. The sample for this test consists of 4356 IPOs, excluding unit offers, ADRs, closed-end funds, and REITs. I run the regression in Eq. (2) below based on two-year rolling periods. All variables in the equation are defined in Appendix A. Because Kim et al. (2008) show that the entry of commercial banks into the IPO underwriting market reduces gross spreads, I try to use their control variables in the regression equation for comparison purpose. The control variables include various determinants of gross spreads documented in the IPO literature, for example, underwriter reputation (Carter and Manaster, 1990), VC-backing (Megginson and Weiss, 1991), proceeds (Denis, 1993; Lee et al., 1996; Chen and Ritter, 2000; Altinkilic and Hansen, 2000), firm age (James, 1992), and return volatility (Hansen, 2001). In particular, the commercial bank dummy controls for the direct effect of commercial bank entry into the IPO underwriting market. To determine if the underwriter is a commercial bank, I obtain data from Table 1 of Grande et al. (1999), the Appendix of Cornett et al. (2002), Fig. 1 of Ljungqvist et al. (2006), and the Federal Reserve's website. Because I am interested in the year-by-year change in gross spreads to enable me to test for the immediate effect of the PSLRA of 1995, the coefficient on the year dummy is of primary interest. Year dummy equals one if the IPO is completed in the second year of a two-year period, and zero otherwise. Therefore, the coefficient on the year dummy captures the change in gross spreads from the previous calendar year after controlling for other determinants of gross spreads. Gross spread = α + β1 Year dummy + β2 Commercial bank dummy + β3 Carter−Manaster rank + β4 VCbacked dummy + β5 LnðproceedsÞ + β6 Lnð1 + ageÞ + β7 LnðassetsÞ + β8 Technology dummy + β9 Utility dummy + β10 Financial dummy + β11 Pure primary dummy + β12 NYSE = Amex dummy + β13 Comparable firms0 return volatility + :

ð2Þ

I run the regression ten times on ten overlapping two-year rolling periods over 1989–1999. I report the ten estimated coefficients on the year dummy in Column (3) of Table 9. At the beginning of the period when commercial banks started to be allowed to underwrite IPOs, gross spreads dropped by 2, 22, and 8 basis points each year for the first three years. Afterwards, gross spreads decreased by 4 basis points each year for two years and then stopped decreasing in the following year. Then suddenly in 1996, when the PSLRA became effective, gross spreads dropped significantly again by 21 basis points. The evidence is consistent with the argument that the PSLRA reduces IPO gross spreads. Although the evidence still cannot rule out the alternative explanation that the sudden drop in spreads in 1996 is due to a spillover effect from the increased competition measured by IPO market share by commercial banks, it is consistent with the drop in litigation risk. The inconclusive evidence in Table 9 helps to motivate further investigation into the relation between litigation risk and gross spreads. Next, I conduct more direct tests and provide further evidence.

452

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Table 9 Year-by-year change in IPO underwriting spreads (%) during 1989–1999. The sample consists of 4356 IPOs during 1989–1999, excluding unit offers, ADRs, closedend funds, REITs, and IPOs with missing information on the variables in Eq. (2). IPO market shares of commercial banks in Column (6) are computed based on proceeds. The mean spreads in Column (7) are computed on an equal-weighted basis. The regression model is: Gross spread = α + β1 Year dummy + β2 Commercial bank dummy + β3 Carter−Manaster rank + β4 VCbacked dummy + β5 LnðproceedsÞ + β6 Lnð1 + ageÞ + β7 LnðassetsÞ + β8 Technology dummy + β9 Utility dummy + β10 Financial dummy + β11 Pure primary dummy + β12 NYSE = Amex dummy 0 + β13 Comparable firms return volatility + : All of the variables are defined in Appendix A. Statistics with significance at the 1%, 5%, and 10% levels are denoted with ⁎⁎⁎, ⁎⁎, and ⁎, respectively. (1)

(2)

(3)

(4)

Sample period

Year dummy

Coefficient on year dummy

Coefficient on commercial bank dummy

1989–1990 1990–1991 1991–1992 1992–1993 1993–1994 1994–1995 1995–1996 1996–1997 1997–1998 1998–1999

1990 dummy 1991 dummy 1992 dummy 1993 dummy 1994 dummy 1995 dummy 1996 dummy 1997 dummy 1998 dummy 1999 dummy

−0.02 −0.22⁎ −0.08⁎

NA NA −0.37⁎⁎⁎ 0.06 0.37 0.27 −0.06 −0.05 −0.02 −0.11⁎⁎⁎

−0.04 −0.04 0.12⁎⁎ −0.21⁎⁎⁎ 0.04 0.07 −0.01

Number of observations

289 447 744 956 973 943 1132 1085 743 763

(5)

(6)

(7)

Year

IPO market shares of commercial banks (%)

Mean spread (%)

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

0.00 0.00 0.00 0.82 1.07 0.39 1.80 2.88 11.64 9.03 14.02

8.18 8.04 7.43 7.49 7.48 7.66 7.50 7.09 7.16 7.12 6.94

The relation between litigation risk and underwriter gross spreads is not subject to an endogeneity problem. As hypothesized in Section 2, litigation risk can affect gross spreads. However, it is hard to imagine why gross spreads should affect litigation risk. Without an endogeneity problem, a two-stage regression model is less efficient. Thus, I use OLS regressions to examine to what extent litigation risk affects gross spreads. Following the literature, I use the occurrence of litigation to proxy for litigation risk. White (1980) standard errors are used for calculating p-values. As robustness checks, I also run the two-staged regression, but the results are not tabulated. Table 10 reports the results. Column (1) shows that the lawsuit dummy is positively and significantly related to gross spreads. However, the economic impact of litigation risk on gross spreads is modest. A one standard deviation increase in the probability of being sued (20%) raises gross spreads, on average, by 2.2 basis points. Compared with the modest effect of litigation risk on gross spreads, the relation between withdrawal risk and gross spreads is much stronger. The withdrawal probability positively affects gross spreads with a p-value less than 0.01. The effect of withdrawal risk on gross spreads is not only statistically, but also economically, significant. Based on the result in Column (1), an increase in the probability of withdrawal by one standard deviation (13%) increases the gross spread, on average, by 9.1 basis points. Based on the average proceeds of $104 million (2005 purchasing power), this suggests that an increase in the probability of IPO withdrawal by one standard deviation increases the dollar amount of underwriter revenue by $94,640 on completed deals. In addition, the insignificant coefficient on the interaction term suggests that the effect of the withdrawal probability on gross spreads does not depend on pre-market demand. In Column (2), the withdrawal probability and the interaction term are excluded from the regression model, and the coefficient on the litigation dummy is slightly increased. In Column (3), the lawsuit dummy is excluded from the regression model, and the coefficient on the withdrawal probability remains the same. As a robustness check, I also estimate the effect of litigation risk on underwriter gross spreads by the two-stage estimation procedure used for litigation risk and underpricing. In untabulated results, the litigation IV passes the exogeneity test following the same procedure as in Section 3.3; the fitted litigation risk and the probability of withdrawal remain statistically significant predictors of gross spreads. To address the concern that the investment bank characteristic variables, offer price revision, and offer size are likely to be endogenous, I report robustness checks in Columns (4)–(6) of Table 10. In addition to omitting the investment bank characteristic variables and offer price revision from the regression model, I follow Lee and Masulis (2009) to use two-stage regressions. Specifically, I instrument the offer size using the predicted value from the regression model in Eq. (3). Offer size = f2 ðcapital expenditure; issuer characteristicsÞ:

ð3Þ

The capital expenditure of the issuing firm in the year prior to the IPO is used to identify offer size. However, about 3.6% of the sample firms have to be omitted from the regression because of missing information for their capital expenditure in the year prior to the IPO. Issuer characteristics, including firm assets and firm age, are used as control variables. Column (4) shows the estimation results for Eq. (3). Both capital expenditure and firm assets are strong predictors of the offer size. I then use the predicted offer size

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

453

Table 10 Ordinary least-squares regressions of gross spread (%) for IPOs during 1996–2005. In Columns (5)–(6), the second stage regressions are estimated separately on the sub-samples with negative and positive residuals from the regression in Column (4). All of the explanatory variables are defined in Appendix A and measured as decimals. White (1980) standard errors are used to compute p-values. Statistics with significance at the 1%, 5%, and 10% levels are denoted with ⁎⁎⁎, ⁎⁎, and ⁎, respectively. (1)

Dependent variable = Intercept Issuer and issue characteristics Lawsuit dummy Ln(Proceeds) Fitted Ln(Filing amount) Ln(Assets) Ln(1 + Age) Technology dummy Utility dummy VC-backed dummy Pure primary dummy NYSE/Amex dummy Offer price revision Probability of withdrawal Probability of withdrawal × Up Capital expenditure Investment bank characteristics Bank market share Bank excess underpricing Missing_excess underpricing dummy All-star analyst dummy Market conditions prior to offer Nasdaq return Comparable firms' return volatility Comparable firms' share turnover Adjusted R2 Observations

(2)

(3)

(4)

(5)

(6)

Negative residual

Positive residual

OLS

OLS

OLS

1st stage OLS

2nd stage OLS

2nd stage OLS

Spread (%)

Spread (%)

Spread (%)

Ln(Filing amount)

Spread (%)

Spread (%)

9.30⁎⁎⁎

9.43⁎⁎⁎

9.30⁎⁎⁎

10.35⁎⁎⁎

8.81⁎⁎⁎

0.11⁎⁎ −0.57⁎⁎⁎

0.12⁎⁎ −0.59⁎⁎⁎

−0.57⁎⁎⁎

−0.03⁎⁎ −0.02 −0.02 0.18 −0.12⁎⁎⁎ 0.07⁎⁎⁎ 0.13⁎⁎⁎ 0.61⁎⁎⁎ 0.70⁎⁎⁎

−0.02⁎ −0.02⁎ −0.09⁎⁎⁎ 0.13 −0.11⁎⁎⁎ 0.07⁎⁎⁎ 0.14⁎⁎⁎ 0.56⁎⁎⁎

−0.12 1.45⁎⁎⁎ −0.36⁎⁎⁎ 0.05 0.06⁎⁎⁎ 0.14 2.42 −0.13 54.26% 2187

−0.03⁎⁎ −0.02 −0.02 0.18 −0.12⁎⁎⁎ 0.07⁎⁎⁎ 0.13⁎⁎⁎ 0.61⁎⁎⁎ 0.70⁎⁎⁎ −0.11

1.33⁎⁎⁎ −0.35⁎⁎⁎ 0.08 0.06⁎⁎⁎ 0.14 3.90⁎⁎⁎ −0.08 53.52% 2187

2.21⁎⁎⁎

0.17⁎⁎

0.43⁎⁎⁎ 0.01 0.20⁎⁎⁎ 0.05 0.17⁎⁎⁎

−0.82⁎ 0.02 −0.06⁎⁎ −0.02 0.27 −0.19⁎⁎ 0.28⁎⁎⁎ 0.24⁎⁎⁎ 0.90⁎⁎⁎

0.00 −0.60⁎⁎ 0.07 0.02 0.10 −0.18 0.10⁎ 0.03 −0.13⁎⁎ 0.49⁎⁎⁎

0.40⁎⁎⁎

1.46⁎⁎⁎ −0.36⁎⁎⁎ 0.05 0.06⁎⁎⁎ 0.14 2.40 −0.13 54.18% 2187

57.97% 2108

39.93% 1062

42.02% 1046

in the second-stage OLS regression of underwriter gross spread. In untabulated results, the coefficient estimate on the withdrawal probability is positive and statistically significant, while the coefficient estimate on the fitted litigation risk is positive, but statistically insignificant. To gain further insight into how underwriters charge gross spreads conditional on offer size, I divide the sample into two subsamples based on whether the residual from the regression in Eq. (3) is positive or negative. The results are reported in Columns (5) and (6). The coefficient on the lawsuit dummy is significantly positive for the sub-sample with negative residuals, while it is zero for the sub-sample with positive residuals. This suggests that the relation between litigation risk and spread is sensitive to the abnormal issue size. Specifically, when the issue size is smaller than normal, where normal refers to the expected issue size based on firm characteristics, underwriters charge additional underwriting fees for incremental litigation risk. But when the issue size is bigger than normal, litigation risk does not affect gross spreads significantly. Furthermore, the coefficient on the probability of withdrawal is 0.90 for the sub-sample with negative residuals, but 0.49 for the sub-sample with positive residuals. The results increase our understanding of how underwriters set gross spreads. When the issue size is smaller than expected, underwriter gross spreads are more sensitive to litigation risk and withdrawal risk. 3.6. Additional robustness checks The 1999–2000 period exhibits dramatically higher underpricing, and almost all the laddering cases involve IPOs completed during these two years. Thus, I exclude the two years from the sample and repeat the analysis. The main results remain robust to excluding the 1999–2000 period. Alternatively, I keep the 1999–2000 period, but exclude IPOs involved in laddering cases from the sample. The main results again remain robust to excluding laddering cases. 4. Conclusions While the relation between litigation risk and IPO underpricing has received considerable attention, we know little about how litigation risk is related to other decisions in the IPO process. The main objective of this study is to conduct a comprehensive examination of the relation of litigation risk to IPO withdrawal, underpricing, and underwriter gross spreads. In addition, this study tests the relative importance of withdrawal risk and litigation risk as determinants of underpricing and gross spreads.

454

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Several findings emerge from the analysis. First, firms with a higher likelihood of IPO withdrawal face higher litigation risk if they complete their offers, compared to other completed IPOs. Given that the offers with the greatest litigation risk would be withdrawn or never registered in the first place, the finding that the withdrawal probability is significantly related to litigation risk, even for completed IPOs, is noteworthy. Second, litigation risk does not significantly affect underpricing during 1996–2005. The finding is robust to controls for the endogeneity of litigation risk in the underpricing regression. The result suggests that the effect of litigation risk on underpricing may have changed from the earlier period, as Lowry and Shu (2002) find that firms with higher litigation risk underpriced their IPOs by a greater amount during 1988–1995. I propose several explanations for the change in the relation. Furthermore, when the pre-market demand for IPOs is weak, firms with higher withdrawal risk underprice their offers by a greater amount to increase the probability of completing the offer. In particular, if the offer price is not revised up, then each additional 13% of withdrawal probability results in underpricing that is 4.94% higher, so that a $104 million deal will leave $5.1 million more on the table. This relation is not present when the pre-market demand is strong. Last but not the least, issuers with higher litigation risk and withdrawal risk pay higher gross spreads. Furthermore, withdrawal risk affects gross spreads more strongly than litigation risk does. Each additional 13% of withdrawal probability results in gross spreads that are 9.1 basis points higher, so that a $104 million deal will pay $94,640 more to the underwriters. For deals with greater withdrawal risk, it makes sense that underwriters should be compensated more on the completed deals, since they lose money on the deals that are withdrawn. Although litigation risk represents a significant threat to firms going public in the United States, the effects of litigation risk on underpricing and gross spreads appear to be modest at best. In contrast, the effects of withdrawal risk on underpricing and gross spreads are strong. The results are consistent with the view that concerns about litigation risk affect the IPO withdrawal decision much more than the underpricing and gross spread decisions. However, it should be recognized that litigation risk may affect withdrawal probability, which in turn affects IPO pricing decisions.

Appendix A. Definition of variables Age is the number of years between the founding year and the IPO year. Founding years are downloadable from Jay Ritter's website. All-star analyst dummy equals one if the lead underwriter has an all-star analyst in the same industry as the issuer in the year prior to the IPO, and zero otherwise. An all-star is defined as any research team mentioned on Institutional Investor's all-star analyst team list in the October issue. Assets is the total assets before offering (in millions of 2005 dollars purchasing power). Average offer price revision is the equal-weighted average of the offer price revision of other IPOs over the 60 calendar days prior to offer or withdrawal. Average underpricing is the equal-weighted average of the underpricing of other IPOs over the 60 calendar days prior to offer or withdrawal. Bank excess underpricing is the lead underwriter's average excess IPO underpricing in the calendar year prior to the offer. Bank market share is the lead underwriter's IPO market share (based on proceeds) in the calendar year prior to the filing year. Capital expenditure is the capital expenditure (data item 128) scaled by assets (data item 6) based on the Compustat annual files in the year prior to IPO. Carter–Manaster rank is the integer part of the IPO lead underwriter reputation ranks that are downloadable from Jay Ritter's website at http://bear.cba.ufl.edu/ritter/Rank.htm. It assigns higher prestige to underwriters that are listed more prominently on tombstone advertisements. The reputation ranks range from 1 (lowest) to 9 (highest). Change in AAA-10 year Treasury yield spread is measured over the 60 calendar days prior to offer or withdrawal, where the AAA10 year Treasury yield spread is the difference between the Moody's index for seasoned corporate bond yield with an Aaa rating and the 10-year (constant maturity) Treasury yield constructed by the Federal Reserve. Change in BAA-AAA yield spread is measured over the 60 calendar days prior to offer or withdrawal, where the BAA-AAA yield spread is the difference between the Moody's indices for seasoned corporate bond yields with Baa and Aaa ratings. Commercial bank dummy equals one if the lead underwriter is a commercial bank, and zero otherwise. Comparable firms' return volatility is the average standard deviation of the daily returns over one year prior to the IPO date for firms that are similar to the issuing firm. To find firms similar to the issuing firm, I select all firms from the same industry as defined by Fama and French (1997) that have market capitalization within 80–120% of the issuing firm, where market capitalization of both the issuing firm and the control firms is measured at the close of the IPO firm's first trading day. Comparable firms' share turnover is average share turnover over one year prior to the IPO date for firms that   are similar to the

volume tradedt ; where t is the day issuing firm. The following formula is used to compute share turnover: turnover = 1−Π−366 t = −1 1− total sharest relative to the IPO date. Trading volume for stocks listed on NASDAQ is divided by two to account for the differences of tradereporting conventions between NASDAQ and NYSE/Amex. Firms similar to the issuing firm are defined under Comparable firms' return volatility above. Filing amount equals the average filing price multiplied by the number of shares to be sold as indicated in the initial filing (in millions of 2005 dollars purchasing power). Financial dummy equals one if the issuer's SIC code starts with 6, and zero otherwise.

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

455

Gross spread is the gross spread paid to the underwriter as a percentage of the offer price. Industry daily return volatility is the standard deviation of the daily returns for firms in the same Fama and French (1997) industry as the IPO firm over the 60 calendar days prior to offer or withdrawal. Industry return is the cumulative equal-weighted return with daily rebalancing for firms in the same Fama and French (1997) industry as the IPO firm over the 60 calendar days prior to offer or withdrawal. Lawsuit dummy equals one if the IPO firm has been sued under the Securities Act of 1933 or the Securities Exchange Act of 1934 within three years following the IPO, and zero otherwise. Missing_excess underpricing dummy equals one if the lead underwriters do not have completed IPOs in the previous year, and zero otherwise. If the lead underwriters do not have completed IPOs in the previous year, I also set the variable bank excess underpricing equal to zero (for regressions only, not for descriptive statistics). Nasdaq return is value-weighted Nasdaq Composite's compounded return (including distributions) over the 15 trading days prior to IPO. Number of IPO filings is the number of IPOs filed over the 60 calendar days prior to offer or withdrawal. NYSE/Amex dummy equals one if the IPO stock is listed on NYSE or Amex, and zero otherwise. Offer price revision = (offer price − midpoint of the original file price range) / midpoint of the original file price range. One-year return is the return from the first-day closing price to the closing price on the one year anniversary after the IPO. Probability of withdrawal is the estimated probability of withdrawal based on the withdrawal probit regression coefficient estimates. Proceeds is the global gross proceeds (in millions of 2005 dollars purchasing power), exclusive of overallotment options. Pure primary dummy equals one if the offering is 100% primary (i.e., no secondary shares sold), and zero otherwise. Ratio of withdrawn IPOs to completed IPOs and IPOs in active registration = number of IPOs withdrawn / (number of IPOs completed + number of IPOs in active registration) over the 60 calendar days prior to offer or withdrawal. Technology dummy equals one if the issuer's SIC code was equal to either 3570, 3571, 3572, 3575, 3577, 3578, 3660, 3661, 3663, 3669, 3674, 3810, 3812, 3820, 3823, 3825, 3826, 3827, 3829, 3840, 3841, 3845, 4812, 4813, 4899, 7370, 7371, 7372, 7373, 7374, 7375, 7378, or 7379, or the issuer is classified as an internet firm by the SDC database, and zero otherwise (Loughran and Ritter, 2002). Underpricing = (Closing price on the first trading day of IPO − offer price) / offer price. Up equals one if the offer price is higher than the midpoint of the original file price range, and zero otherwise. Utility dummy equals one if the issuer's SIC code starts with 49, and zero otherwise. VC-backed dummy equals one if the firm is venture capitalist backed, and zero otherwise. Year dummy equals one if the IPO is completed in the second year of a two-year period, and zero otherwise. References Alexander, J.C., 1993. The lawsuit avoidance theory of why initial public offerings are underpriced. UCLA Law Rev. 41, 17–73. Ali, A., Kallapur, S., 2001. Securities price consequences of the private securities litigation reform act of 1995 and related events. Acc. Rev. 76, 431–460. Altinkilic, O., Hansen, R.S., 2000. Are there economies of scale in underwriting fees? Evidence of rising external financing costs. Rev. Financ. Stud. 13, 191–218. Beatty, R.P., Drake, P.P., Hogan, C.E., 2003. The impact of the 1995 Private Securities Litigation Reform Act on litigation risk and auditor compensation in the IPO market. Working paper. Southern Methodist University. Benveniste, L.M., Ljungqvist, A., Wilhelm Jr., W.J., Yu, X., 2003. Evidence of information spillovers in the production of investment banking services. J. Finance 58, 577–608. Bordon, D.E., Kerns, C., Davisson, M.R., Van Vechten, E.B., 1998. Directors and Officers Liability Insurance Deskbook. American Bar Association. Brau, J.C., Fawcett, S.E., 2006. Initial public offerings: an analysis of theory and practice. J. Finance 61, 399–436. Carter, R., Manaster, S., 1990. Initial public offerings and underwriter reputation. J. Finance 45, 1045–1067. Chalmers, J.M.R., Dann, L.Y., Harford, J., 2002. Managerial opportunism? Evidence from directors' and officers' insurance purchases. J. Finance 57, 609–636. Chen, H., Ritter, J., 2000. The seven percent solution. J. Finance 55, 1105–1131. Cliff, M.T., Denis, D.J., 2004. Do initial public offering firms purchase analyst coverage with underpricing? J. Finance 59, 2871–2901. Cornett, M.M., Ors, E., Tehranian, H., 2002. Bank performance around the introduction of a Section 20 subsidiary. J. Finance 57, 501–521. Corwin, S.A., Schultz, P., 2005. The role of IPO underwriting syndicate: pricing, information production, and underwriter competition. J. Finance 60, 443–486. Denis, D.J., 1993. The costs of equity issues since rule 415: a closer look. J. Financ. Res. 16, 77–88. Drake, P.D., Vetsuypens, M.R., 1993. IPO underpricing and insurance against legal liability. Financ. Manage. 22, 64–73. Dunbar, C.G., Foerster, S.R., 2008. Second time lucky? Withdrawn IPOs that return to the market. J. Financ. Econ. 87, 610–635. Edelen, R., Kadlec, G., 2005. Issuer surplus and the partial adjustment of IPO prices to public information. J. Financ. Econ. 77, 347–373. Fama, E., French, K., 1997. Industry costs of equity. J. Financ. Econ. 43, 153–193. Grande, A., Puri, M., Saunders, A., 1999. Bank entry, competition, and the market for corporate securities underwriting. J. Financ. Econ. 54, 165–195. Guo, R., 1998. On the cost of withdrawn initial public offerings. Working paper, University of Minnesota. Greene, W., 2003. Econometric Analysis, 5th edition. Prentice Hall, New Jersey. Hansen, R.S., 2001. Do investment banks compete in IPOs? The advent of the ‘7% plus contract’. J. Financ. Econ. 59, 313–346. Hao, Q., 2007. Laddering in initial public offerings. J. Financ. Econ. 85, 102–122. Hensler, D.A., 1995. Litigation costs and the underpricing of initial public offerings. Managerial Decis. Econ. 16, 111–128. Hoberg, G., 2007. The underwriter persistence phenomenon. J. Finance 62, 1169–1206. Hughes, P.J., Thakor, A.V., 1992. Litigation risk, intermediation, and the underpricing of initial public offerings. Rev. Financ. Stud. 5, 709–742. Ibbotson, R.R., 1975. Price performance of common stock new issues. J. Financ. Econ. 2, 235–272. Ince, O., 2010. Why do IPO offer prices only partially adjust? Working paper, Virginia Tech. Liu, X., Ritter, J.R., 2010. The economic consequences of IPO spinning. Rev. Financ. Stud. 23, 2024–2059. James, C., 1992. Relationship-specific assets and the pricing of underwriter services. J. Finance 47, 1865–1885. Johnson, M.F., Kasznik, R., Nelson, K.K., 2000. Shareholder wealth effects of the private securities litigation reform act of 1995. Rev. Acc. Stud. 5, 217–233. Keloharju, M., 1993. The winner's curse, legal liability, and the long-run price performance of initial public offerings in Finland. J. Financ. Econ. 34, 251–277. Kennedy, P., 2003. A Guide to Econometrics, 5th ed. The MIT Press, Cambridge, Massachusetts.

456

(G.)Q. Hao / Journal of Corporate Finance 17 (2011) 438–456

Kim, D., Palia, D., Saunders, A., 2008. The impact of commercial banks on underwriting spreads: evidence from three decades. J. Financ. Quant. Anal. 43, 975–1000. Lee, I., Lochhead, S., Ritter, J., Zhao, Q., 1996. The costs of raising capital. J. Financ. Res. 19, 59–74. Lee, G., Masulis, R.W., 2009. Seasoned equity offerings: quality of accounting information and expected flotation costs. J. Financ. Econ. 92, 443–469. Lerner, J., 1994. Venture capitalists and the decision to go public. J. Financ. Econ. 35, 293–316. Lian, Q., Wang, Q., 2009. Does the market incorporate previous IPO withdrawals when pricing second-time IPOs? Financ. Manage. 38, 357–380. Ljungqvist, A., 2007. IPO underpricing. In: Eckbo, B.E. (Ed.), Handbook of Corporate Finance. Elsevier Inc., North-Holland, pp. 1–74. Ljungqvist, A., Marston, F.C., Wilhelm, W.J., 2006. Competing for securities underwriting mandates: banking relationships and analyst recommendations. J. Finance 61, 301–340. Loughran, T., Ritter, J.R., 2002. Why don't issuers get upset about leaving money on the table in IPOs? Rev. Financ. Stud. 15, 413–443. Loughran, T., Ritter, J.R., 2004. Why has IPO underpricing changed over time? Financ. Manage. 33, 5–37. Lowry, M., Shu, S., 2002. Litigation risk and IPO underpricing. J. Financ. Econ. 65, 309–335. Maddala, G., 1983. Limited-dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge, England. Megginson, W.L., Weiss, K.A., 1991. Venture capitalist certification in initial public offerings. J. Finance 46, 879–903. Nimalendran, M., Ritter, J.R., Zhang, D., 2007. Do today's trades affect tomorrow's IPO allocation? J. Financ. Econ. 84, 87–109. Pukthuanthong, K., Walker, T.J., Turtle, H.J., 2009. Legal opportunism, litigation risk, and IPO underpricing. Working paper, Washington State University. Reuter, J., 2006. Are IPO allocations for sale? Evidence from mutual funds. J. Finance 61, 2289–2324. Spiess, D.K., Tkac, P.A., 1997. The private securities litigation reform act of 1995: the stock market casts its vote…. Managerial Decis. Econ. 18, 545–561. Tinic, S.M., 1988. Anatomy of initial public offerings of common stock. J. Finance 43, 789–822. Titman, S., Wessels, R., 1988. The determinants of capital structure choice. J. Finance 53, 1935–1974. Torstila, S., 2001. What determines IPO gross spreads in Europe? Eur. Financ. Manage. 7, 523–541. Torstila, S., 2003. The clustering of IPO gross spreads: international evidence. J. Financ. Quant. Anal. 38, 673–694. White, H.L., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817–838. Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA. Wooldridge, J.M., 2009. Introductory Econometrics: A Modern Approach, fourth ed. South-Western Cengage Learning, Mason, OH. Zhu, Y., 2009. The relation between IPO underpricing and litigation risk revisited: changes between 1990 and 2002. Financ. Manage. 38, 323–355.