Take a chance? Implications of auditor going concern opinions for IPO investors

Transcription

1 Take a chance? Implications of auditor going concern opinions for IPO investors Abstract In a marked shift, it has recently become relatively common for ordinary IPOs to contain going concern (GC) opinions in their offering documents. We examine the implications of such opinions for IPO investors in a sample of ordinary IPOs from We find no significant difference in underpricing for GC and non-gc IPOs, while VC-backed GC IPOs experience significantly less underpricing and second year GC IPOs are associated with significantly more underpricing. At the same time, GC opinions provide some evidence of inferior post-ipo stock market performance. A GC opinion is associated with mixed evidence of delisting due to poor performance, but GC IPOs show lower post-ipo operating performance. Overall, when issuing GC opinions on IPOs, auditors seem to be able to very meaningfully distinguish between those companies likely to survive and those that are not. Thus, the information is of significant value to IPO investors. Keywords: Going concern opinions; IPO pricing; Delisting; Venture capital; Partial adjustment; Liquidity JEL Classification: G14; M42; G33; G28; K22

2 Take a chance? Implications of auditor going concern opinions for IPO investors 1. Introduction A going-concern (GC) opinion issued in a company s initial public offering (IPO) registration statement indicates that the auditor has substantial doubts about the company s ability to continue as a going concern for a reasonable period of time, not to exceed one year beyond the date of the financial statements being audited. 1 A decade ago, GC opinions were virtually nonexistent in IPO filings for ordinary firms (as opposed to highly speculative microcap or penny stock offerings). However, such opinions have become more common in recent years. For example, in 2014 almost half of all 530 auditors GC opinions were issued for IPO filings, including prominent firms such as Glaukos Corp. and Sciences Ltd. that listed on the New York Stock Exchange and fifteen firms that listed on the Nasdaq (McKenna, 2016). Prior studies examine GC opinions implications for microcap/penny IPOs (Willenborg and McKeown (2001), Weber and Willenborg (2003)). However, the question still remains as to whether GC opinions matter for ordinary 2 IPOs. Our study addresses this question. The aim of this paper is three fold. First, we explore the relationship between GC opinion and IPO performance (underpricing, delisting, operating and stock return performance). Second, we shed light on whether GC opinion has implications for companies IPO withdrawal. Lastly, we investigate whether institutional investors invest in IPOs with GC opinions, and whether firms with GC opinions are more likely to be sued in class actions lawsuits following an admission. To address these questions we examine 831 non-gc IPOs and 111 IPOs with GC opinion, which had a listing on one of the US stock exchanges between 2001 and See Statement on Auditing Standards (SAS) No In this paper, any IPO which is not a microcap or penny stock offering is referred to as an ordinary IPO. 1

3 The focus on ordinary IPOs is especially warranted due to significant recent trends. First, there were 1,340 ordinary IPOs between 2001 and 2013, compared to only 64 unit and penny stocks (Ritter, 2016). In addition, the proportion of the largest exchange-listed firms (with sales exceeding $200 mil measured one year prior to going public) increased substantially. For example, large firms represented 17.85% of all IPOs during the time period, but increased to 35.09% in the time period (Ritter, 2016). Similarly, IPOs gross proceeds have increased substantially during the last few decades. Ritter (2016) reports a significant growth in gross proceeds from $296,693 mil ( ) to $492,923 mil in recent years ( ). Secondly, microcap and penny stocks differ from ordinary IPOs in a number of dimensions. Liu et al (2011) argue that penny stocks have fewer shareholders and demonstrate lower trading volumes (compared to large cap stocks). Bradley et al (2006) find that, in contrast to ordinary IPOs, penny stocks have higher underpricing, longer lock-ups, and greater long-run under-performance. Further, penny stocks are mainly traded over-the-counter. It is therefore not surprising that researches tend to exclude penny stocks in their analyses. Third, the Sarbanes Oxley Act (SOX) of 2002 clearly calls for and highlights the importance of public firms using independent audit services and states auditors reporting requirements. 3 It is important to examine whether SOX had an impact on auditor GC opinion s informativeness and its predictive power of IPO firms performance, delisting, lawsuits and ownership structure. Hence, further research is warranted in this area as a result of a high number of ordinary IPOs, the increased 3 Since January 1, 1989, Professional Auditing Standards (AU Section 341) also clearly states auditors responsibilities with respect to declaring a GC opinion. 2

4 use of GC opinion for these firms, implementation of SOX in 2002, and fundamental differences between microcap/penny stocks and stocks of ordinary IPOs. The decision to go public is one of the most important corporate events. Due to limited knowledge of private firms operations, there is a significant degree of information asymmetry. GC opinions in IPO filings inform the market and potential investors. To illustrate the language in a GC opinion for an IPO firm, we take an excerpt from the February 1, 2001 prospectus of Briazz s IPO filing, which can be found under the section, Report of Independent Accountants. The accompanying financial statements have been prepared assuming that the Company will continue as a going concern. As discussed in Note 2 to the financial statements, the Company has reported operating losses since inception and needs to raise additional capital to fund future operating losses and planned growth. These are conditions that raise substantial doubt about its ability to continue as a going concern. Management's plans in regard to these matters are also described in Note 2. The financial statements do not include any adjustments that might result from the outcome of this uncertainty. Willenborg and McKeown (2001) document that the auditor GC opinion mitigates ex ante investor uncertainty and therefore reduces underpricing in their sample of micro-cap issues that raised $10 mil or less. We contribute to the literature by investigating whether GC opinion matter for ordinary IPOs, and we investigate whether greater default risk might result in lower offering prices. Our findings suggest that GC IPOs are significantly lower priced than non-gc IPOs. Further, we find that GC IPOs with VC-backing are significantly less underpriced. In addition, we differentiate GC IPOs into those with only one GC opinion and those who received a subsequent auditor GC opinion. Firms that continue with GC opinions in the year after the IPO are significantly positively related to underpricing. Therefore, presented results suggest that the market correctly identifies firms that need to go public due to a financial constraint versus those that have a serious structural/persistent problem. 3

5 Greater default risk should naturally increase the chance of subsequent stock delisting. Willenborg and McKeown (2001) demonstrate that GC microcap IPOs are associated with such higher delisting probabilities. Further, Weber and Willenborg (2003) document that the pre-ipo opinions of large auditors in micro-cap IPOs are more predictive of post-ipo stock delisting, implying a certification effect for more prestigious auditors. However, based on auditors GC opinions of stressed Internet companies filing to go public on Nasdaq, Leone et al (2013) find that Big 5 auditors were less likely to issue GC opinions during the bubble period from 1999 to They argue that these auditors did not properly recognize the IPO bubble. Exploring the impact of GC opinions in general on subsequent delisting likelihood of ordinary IPOs, we present evidence that delisting of ordinary IPOs is positively related to GC opinions. This finding is even stronger for firms with a subsequent second year GC opinion. However, once we control for institutional ownership (IO) and analyst coverage, GC opinions are no longer predictive of delisting. Since GC opinions reflect auditors doubt about a firm s ability to continue operations beyond one year, it is no surprise that we find lower post-ipo operating performance for GC IPOs than for non-gc IPOs. We next turn to analyzing the long-run stock performance for GC IPOs. Consistent with prior IPO literature, we present evidence that GC IPOs underperform in the long-run. In our analysis, neither the three-year value weighted buy-and-hold (BHAR) nor the value-weighted calendar-time returns show any difference due to GC opinions. Our baseline results employ ordinary least squares and logistical regression methods. To control and mitigate the selection bias, we employ a propensity score matching technique and rerun our main results with this method. Our main results and overall conclusions remain the same. Our contribution is fivefold. First, we present evidences that GC opinion is predictive of ordinary IPOs withdrawal, long-run underperformance and post-ipo lawsuits. Secondly, we 4

6 contribute to the literature by analyzing the impact of VC backing on the relationship between GC opinions and post-ipo firms characteristics, as well as the effect of second year GC opinion. Third, in contrast to prior studies which analyze a niche group of firms made of microcap and penny stocks (Weber and Willenbord (2003), Willenborg and McKeown, (2001)), we concentrate on a broader, more representative sample of ordinary IPOs, which represent the vast majority of all US IPOs using the most recent time period available ( ). Fourth, we provide evidence that GC IPOs should not be treated as a homogenous group of companies. We show that IPOs which receive an original GC opinion in the IPO filing as well as a subsequent GC opinion in their first annual report ( second year GC IPOs ) are significantly different from IPOs with GC opinion in IPO prospectus only. Hence, future research should be conducted taking this into account. Lastly, our findings complement other studies, which concentrate on analyzing the effect of voluntary management disclosure, the tone and type of information in IPO filings on underpricing and offer price (Hanley and Hoberg (2010), Loughran and McDonald (2013)). Our paper examines another aspect of IPO filings: GC opinion, which is an auditor s certification (or lack thereof). We examine whether it has any predicting power of recently public companies long-run performance, IPO withdrawal, ownership structure and class action lawsuits. The remainder of the paper proceeds as follows. Section 2 reviews relevant literature and develops a set of hypotheses to be empirically tested. Section 3 introduces data selection, main variables, empirical methods, and descriptive statistics. Section 4 and 5 present multivariate tests and results. Section 6 examines firms that withdraw their IPO plans, and section 7 concludes the paper. 2. Related literature and hypotheses development 5

7 On the role of auditors in IPOs, there is a small body of literature that examines the demand for auditors during the IPO process, the role of auditor quality and reputation on predicting IPO longrun performance, the effect of auditor opinions on IPO delisting rate and underpricing. Willenborg (1999) tests and confirms that auditors serve the dual role of both enhancing resource allocation (an informational role) and providing investors with a claim on the auditor in the event of an audit failure (an insurance role). He demonstrates the importance of IPO proceeds in explaining both the impact of auditor choice on IPO underpricing and the auditor's compensation. 2.1 GC Opinion and Long-Run Performance and Delisting It is well documented that IPOs in general underperform over the long run (Ritter (1991), Loughran and Ritter (1995)). However, certain firm and offering characteristics, such as non-venture capital backing (Brav and Gompers (1997)) and less prestigious investment banks as underwriters (Carter, Dark, and Singh(1998)) are associated with more severe IPO long-run underperformance. Private equity (PE) and VC investors' involvement in firms is found to result in better post-ipo stock returns: financially sponsored IPOs outperform non-backed peers in the long-run (Ritter (2016), Levis (2011)). There is limited research on identifying internal reasons for the underperformance. GC opinions for IPOs rendered by auditors indicate that a private firm has substantial financial constraints such that its continual operation and future viability might be an issue of concern. If the market was able to accurately incorporate GC opinions in the offer price, we should not expect to observe any differences in long-run stock performance between GC IPOs and non-gc IPOs. However, if the risk was not fully incorporated initially, we might observe that GC IPOs display lower long-run stock performance than non-gc IPOs. Therefore, we investigate this question, which provides another piece of the story for IPO long-run performance. 6

8 Willenborg and McKeown (2001) find that microcap GC IPOs are more likely to be delisted within three years after the IPO. Weber and Willenborg (2003) document that for those microcap IPOs, the pre-ipo opinions of large auditors are more predictive of post-ipo negative stock delisting. No research to date, however, has examined the effect of GC opinions for ordinary IPOs using the most current time period, which covers the SOX 2002 Act as well as the recent financial crisis. Finding of the extant literature leads us to the following first hypothesis: Hypothesis 1a: Compared to non-gc IPOs, GC IPOs are no more likely to underperform in the three years after public offerings. Hypothesis 1b: Compared to non-gc IPOs, GC IPOs are no more likely to be delisted in the three years after public offerings. 2.2 GC Opinion and Underpricing Perhaps the most puzzling and researched question in the IPO literature is the issue of first day return (underpricing), which has stimulated a variety of theoretical predictions and empirical assessments. The most plausible explanations that stand the test of time are those based on information asymmetry. Ritter (1984) and Rock (1986) suggest that underpricing is a compensation mechanism for the cost that uninformed investors bear to become informed. Everything else being equal, there should be a positive relationship between pre-ipo uncertainty and IPO first day return. High-risk offerings, such as high tech companies, small firms, and Internet firms, should have higher first day return to compensate uninformed investors. Willenborg and McKeown (2001) find that microcap GC IPOs suffer less underpricing due to reduced uncertainty surrounding the IPOs. Previous studies analyze the impact of management disclosure, tone and type of information on IPO firms underpricing and price revisions (Hanley and Hoberg (2010), Loughran and McDonald 7

9 (2013)). We supplement these existing findings by investigating another aspect of an IPO prospectus, namely the auditors certification/disclosure of a going concern. On the one hand, GC IPOs presumably can be classified as a high risk IPOs since auditors expressed substantial doubts about the company s ability to continue as a going concern. On the other hand, as an important accounting certification/disclosure, GC opinions in IPO filings are public information and should reduce the uncertainty involved in an offering. Therefore, the relationship between GC opinions and underpricing for ordinary IPOs is an empirical question that we will address, and test the following null hypothesis: Hypothesis 2: Compared to non-gc IPOs, GC IPOs are no more or less underpriced during public offerings. In contrast to prior studies, we contribute to the literature by differentiating and dividing GC IPOs into the following two groups: i) firms that received a GC opinion in the IPO filing only, ii) IPOs which received GC opinion in the IPO filing and a subsequent GC opinion in their first annual report. We ll refer to the latter group of firms as second year GC IPOs. These firms are more likely to have a serious structural issue, in contrast to IPOs that only received a GC opinion in the IPO filing as a result of being severely financially constrained. Hence, the following additional expectation emerges: second year GC IPOs are more likely to experience higher underpricing and more likely to be delisted within the first three years following the flotation. 2.3 GC Opinion and Venture Capital (VC) Investors 8

10 VC investors invest in firms, restructure them, add value and realize their returns by bringing firms public. 4 VC spent the last few years restructuring the firm from within. Levis (2011) reports that on average VC investors restructure firms for 4.5 years before bringing them to the stock market. In 2015, financial sponsors (venture capital and private equity investors) drove US IPO market by being responsible for 63% of all US IPOs (Ernst and Young (2015), (2016)). This trend has continued during the first half of 2016: financial sponsored IPOs accounted for 49% IPOs by number and 60% by proceeds. In addition, Ritter (2016) reports that between 2001 and 2015 VC (PE)-backed IPOs represented 45% (28%) of all IPOs in the US. However, the extant literature reports that many VCs prefer to conduct only a partial exit at the IPO date, where VCs sell a part of their holdings while they still retain equity ownership in firms even after the flotation (Lin and Smith (1998), Megginson and Weiss (1991)). This can partly be explained by the existence of lock-up agreements, which limit pre-ipo investors ability to sell shares in the aftermarket. Brav and Gompers (2003) report the average lock-up agreement in the US is 180 days. However, past studies demonstrate that even after the lock-up expiration date VC investors prefer to retain an equity stake in the firm (Lin and Smith (1998), Megginson and Weiss (1991), Krishnan et al (2011)). For example, Krishnan et al (2011) report that even three years after the IPO VC investors retain equity ownership and are represented on the board of directors. Prior studies present evidence that some IPOs enjoy lower underpricing. For example, Megginson and Weiss (1991) find that venture capital investors involvement in firms pre-ipo provide certification, thereby reducing information asymmetries and costs associated with an IPO as well as increasing net proceeds. Levis (2011) analyzes IPOs in the UK and presents evidence of lower 4 VCs have several exit mechanisms available: an IPO, sale to another corporation (trade sale), buyback exit, sale to another VC firm (secondary sale) and write-off. VC investors also have an option on whether to complete a full or partial exit (Cumming and MacIntosh (2003)). 9

11 first-day returns for VC and private equity (PE) backed IPOs in comparison to non-sponsored companies. In contrast, there are studies which report that underpricing of backed IPOs is not always lower. For example, Ritter (2016) reports that the average first day return for VC-backed IPOs is 29.3%, compared to 13.6% for non-financially-sponsored IPOs. It s important to note that an IPO is the most visible exit route available to VCs (Krishnan and Masulis (2010)), the most profitable exit route (Brau et al (2003)), and it also allows VCs to build a track record. Having a successful track record is vital since VCs are repeat players in the market, where every few years they will be approaching potential limited partners asking them to commit additional capital, who in turn will gauge VC investors success by their past deals. Given the importance of good reputation in the VC industry and VC investors insider knowledge of the firm (as a results of restructuring), one could argue that VC investors would not choose the most visible exit mechanism (i.e. IPO) unless they are certain about the firms future prospects, despite the auditors going concern opinions. Hence, prior studies lead us to expect that having venture capitalists pre-flotation would act as an additional certification mechanism for GC IPOs. Having venture capitalists as equity investors in a firm, about which auditors expressed significant doubts, is likely to reduce information asymmetry and lead to lower underpricing. Hence, the following third hypothesis emerges: Hypothesis 3: GC IPOs with VC-backing are likely to experience lower underpricing than the GC IPOs without such backing. 2.4 GC Opinion and Institutional Ownership 10

12 A relatively large body of literature has demonstrated that institutional investors are more sophisticated and more important players in the financial markets (e.g., Sias, Starks, and Titman (2006)). In the IPO arena, Field and Lowry (2009) document that firms with the lowest levels of institutional investment significantly underperform, suggesting that institutional investors are able to avoid the worst performing firms. Their paper raises another, deeper question: How are institutions able to identify and avoid the worst performing firms? Using a large sample of transaction-level institutional trading data, Chemmanur, Hu and Huang (2010) provide evidence that institutional investors possess significant private information, and they realize considerable above-average returns in the first few months after IPOs. Our paper on GC opinions in IPOs provides a promising setting for testing whether institutional investors also employ publicly available information to avoid the worst performing IPOs. More specifically, we test the following hypothesis: investment. Hypothesis 4: Compared to non-gc IPOs, GC IPOs attract the same level of institutional 2.5 GC Opinion and Lawsuits The existing literature has not reached a definite conclusion regarding the complex relationship between disclosure and subsequent litigation risk following IPOs. Field, Lowry and Shu (2005) document that disclosure deters some types of litigation after the enactment of Private Securities Litigation Reform Act of 1995 (PSLRA). Using word content analysis on the time-series of IPO prospectuses, Hanley and Hoberg (2012) find evidence that strong disclosure serves as a substitute for underpricing and provides an effective hedge for all types of lawsuits. On the other hand, Rogers and Van Buskirk (2009) provide evidence that firms decrease the precision and magnitude of disclosure after being sued and suggest that managers of sued firms perceive that disclosure might trigger lawsuits. 11

13 As one of the most important sources of accounting disclosure, auditors GC opinions in IPOs and the subsequent probability of lawsuits offer us an opportunity for testing an unsettled relationship and contribute to the ongoing debate. Hence, we test the following hypothesis: Hypothesis 5: Compared to non-gc IPOs, GC IPOs are no less likely to be sued in class action lawsuits after initial public offerings. 2.6 GC Opinion and IPO Withdrawal As probably the most important event in the life of a corporation, going public is not an easy transition for most companies, and some companies withdraw their plans due to various reasons. A few papers have examined the causes and consequences of this special group of companies who filed initial registration(s) with the SEC, but discontinue their effort to go public (Busaba, Benveniste, and Guo(2001), Dunban and Forster (2008), Cooney, Moeller, and Stegemoller (2009), Hao (2011)). Among other factors, GC opinions from independent auditors could possibly be the determinant for a company to withdraw their IPO plans. In our paper, we shed light on this question by testing the following hypothesis: Hypothesis 6: Compared to non-gc companies, GC companies are no more likely to withdraw their plans for public offerings. 3. Data and methods We collect data for all the variables employed in this study from a variety of sources. In this section, we introduce the data sources, the main data selection criteria, and descriptive statistics for the main variables of interest. 12

14 3.1. Data and main variables Our primary IPO sample spans over more than a decade from 2001 to For certain analyses we use data only up to 2010, specifically when we examine post-ipo long-run returns, delisting frequencies, and litigation probability. We identify IPOs using Thomson Financial s Securities Data Company (SDC) database. After eliminating closed-end funds, spin-offs, unit issues, real estate investment trusts, limited partnerships, financials (SIC ), IPOs with offer prices less than $5, and IPOs with critical missing information from their SEC filings or CRSP, we arrive at a total of 1019 IPOs. After further deleting IPOs that are not listed on CRSP within 6 months of issuing, following Liu and Ritter (2011), our final sample consists of 942 IPOs. Of these 831 are non-gc IPOs and 111 are GC IPOs. To avoid delisting bias, we include only IPO firms in this analysis for which we have the required accounting data available for all three years post listing. We also collect a sample of 591 withdrawn IPOs from SDC, using the same criteria and time period. SDC provides a variety of additional IPO characteristics such as lead underwriters, proceeds, offer price, VC backing, share overhang, auditors, and high tech industry classification. We obtain accounting data from Compustat, analyst and earnings forecasts from the Institutional Brokers Estimate System (I/B/E/S), and 13f institutional holdings data from the Thomson Reuters Ownership Database. For each company in our sample, we gather information on class action lawsuits for three years after the IPO date from Stanford Law School s Securities Class Action Clearinghouse. Stock price data needed to compute underpricing, long-run stock performance, and delisting frequencies are from CRSP. Finally, we hand-collect GC opinions from IPO firm filings downloaded from EDGAR. We examine both Form S-1 (the initial filing) and Form 424B3 (the final prospectus). ***Table 1 about here*** 13

15 We examine both the final prospectus (SEC Form 424B3) and the original registration statement (SEC Form S-1) for evidence of GC opinions. If a firms has a CG opinion in either of these documents, they are included in our GC sample. 5 Panel A of Table 1 presents the distribution of total IPOs and GC IPOs in our sample by year. There are more than 100 IPOs per year in our sample from , but significantly fewer in the 3 years before and 3 years following this period. This lack of IPO deal flow is not surprising since 2001 follows the bursting of the Internet bubble of , and corresponds to the recent financial crisis. The percentage of IPOs that receive a GC opinion from their auditors shows considerable variation. For instance, in 2009, none of the 34 IPOs received a GC opinion whereas, in 2003, close to 20% of IPOs received one. Possibly due to the JOBS Act in April 2012 favoring smaller firms (Dambra, Field, and Gustafson (2014)), both overall IPO volume and percentage of GC IPOs (29%) climbed up significantly in Overall, 11.4 percent of our sample firms have GC opinions, which is a huge increase for ordinary IPOs compared to previous decades. Panel B reports descriptive statistics for non-gc versus GC IPOs. Several differences are immediately apparent. First, GC IPOs are smaller and have lower offer prices. The average non-gc IPO raises an average of $211 mil with an offer price of $14, and the average company has a market capitalization of $831 mil. In contrast, the average GC IPO raises $77 mil and has an offer price of $10 and a market capitalization of $352 mil. Both mean and median differences, using independent group t-tests and Kruskal-Wallis tests respectively, are economically and statistically significant. GC IPOs are less likely to use higher quality agents (investment banks and auditors). Non-GC IPOs have a Carter-Manaster underwriter rank of 8.3 on a 9-point scale compared to 6.8 for GC IPOs (Carter and Manaster (1990)). 88% of non-gc IPOs use a big 5 auditor compared to 72% for GC 5 In four cases the GC opinion is expunged in the final prospectus. We include these four cases in our GC sample, but we also rerun our analyses with them in the non-gc sample, and our results are not materially impacted. 14

16 IPOs. Institutional ownership is approximately 1/3 higher for non-gc IPOs, and these companies receive significantly more earnings forecasts from analysts. Finally, 40% of non-gc IPOs are hightech compared to 18% of GC IPOs. Internet-related IPOs are somewhat less likely to be GC IPOs, with a mean difference of 7% and statistical significance at the 5% level. However, there is no significant difference between the proportions of VC-backed IPOs for the two groups. Panel C provides an industry breakdown of our sample. The majority of GC opinions in our sample (69%) are for firms in the manufacturing sectors, 20% for firms in the services industry, and 7% for firms in the transportation, communications, electric, etc. industry. The remaining industries are sparsely populated. 6 In Panel D, we compare IPOs from the 1990s to our sample. A potential reason for the sharp increase in GC opinions is that IPOs became riskier in our sample relative to previous periods. However, the evidence in Panel D (and in more detailed studies such as Gao, Ritter, and Zhu (2014)) suggests that, if anything, the opposite is true. IPOs in our sample are substantially larger (in terms of both proceeds and post-ipo market cap), much more likely to have a big 5 auditor, and significantly older. Not surprisingly, underpricing (about 9.5% for our overall sample) is much smaller during than that during (about 15%) and that during (about 65%) documented by Loughran and Ritter (2004). These characteristics are associated with less risky deals. The offer prices are about the same, and the only potentially negative factor is the percentage of companies with positive EPS at the time of the offer is 5 percent smaller in our sample, 50% versus 55%. Overall, it does not appear that the rise in GC opinions is due to a sharp increase in the riskiness of firms going public. Whether regulatory changes such as Sarbanes-Oxley and Regulation FD or events such as the Global Settlement of 2003 are an explanation is an open question for future 6 In our sample there are 154 biotech IPOs of which 36% received a GC opinion. Our main results are robust excluding these IPOs. 15

17 research. For the remainder of our paper, we focus on the implications of a GC opinion for IPO companies and their investors Empirical Methods Long-run stock performance This paper employs three different methodologies to compare the long-run performance of the GC IPOs to non-gc IPOs, since these different methods assume different investment strategies that could be used by different kinds of investors. We first report both three year cumulative abnormal returns (CARs) and three year buy-andhold abnormal returns (BHARs) for GC IPOs and non-gc IPOs, respectively, and test their differences in means. To obtain CARs, we sum abnormal returns as follows: CCCCCC iiii = TT tt=1( RR iiii RR mmmm ), (1) where RR iiii is the simple monthly return for sample firm i, and RR mmmm is the simple monthly return for the corresponding value-weighted size/book-to-market (BM) portfolio constructed by Fama and French (1993). We match each firm to one of the 25 corresponding size/bm portfolios at the beginning of the announcement year using the size/bm breakpoints from Ken French s website 7. We report both equal-weighted (EW) and value-weighted (VW) average CARs for GC IPOs and non- GC IPOs. For VW returns, we construct the weights using the IPO s first available market capitalization (shares outstanding multiplied by closing price) from the CRSP daily data, scaled by the level of the CRSP VW index at that date

18 Second, following Lyon, Barber and Tsai (1999), we also compute buy-and-hold abnormal returns (BHARs) for three years after the offering date by compounding, using the returns of the size and book-to-market ratio matched portfolios as benchmarks. We obtain the size/bm breakpoints from Ken French s website and match each firm to one of the 25 corresponding size/bm portfolios at the end of the offering year. The delisting returns are included, and the proceeds in the matching size/bm portfolio are re-invested if a stock stops trading prior to the end of the event window. Therefore, BHAR is computed as follows: BBBBBBBB iiii = TT tt=1(1 + RR iiii ) TT tt=1 (1 + RR mmmm ), (2) where RR mmmm is the return of the corresponding value-weighted size/book-to-market (BM) portfolio constructed by Fama and French (1993). Both BHARs and CARs are known to suffer from a lack of independence and crosscorrelation in event-time returns, which may lead to biased test statistics (Fama,1998). Therefore, we also apply the calendar-time regression approach to estimate the average abnormal monthly returns. We first form portfolios in calendar time based upon proposed hypothesis. The time-series variation in portfolio returns mitigates the effect of any cross-correlation on the variance of abnormal returns that could plague the BHARs and CARs methods. Thus, we form portfolios in every calendar month over the sample period of GC and non-gc IPOs. We implement the following regression model to examine the effect of long-run stock returns of IPOs with GC or without GC opinions: RR pppp RR ffff = aa + bb pp RR mmmm RR ffff + ss pp SSSSSS tt + h pp HHHHHH tt + ee pppp, (3) 17

19 where RR pppp is the simple monthly return on the calendar-time portfolio (for both equal-weighted and value-weighted), RR ffff is the monthly return on three-month Treasury bills, and RR mmmm is the return on the corresponding value-weighted size/book-to-market (BM) portfolio constructed by Fama and French (1993). SSSSSS tt is the difference in the returns of value-weighted portfolios of small stocks and big stocks, HHHHHH tt is the difference in the returns of value-weighted portfolios of high book-tomarket stocks and low book-to-market stocks. The parameter estimates aa, bb pp, ss pp, and h pp are generated as regression outputs. The error term of the regression is denoted by ee pppp. The intercept estimate aa tests the null hypothesis that the mean monthly excess return on the calendar-time portfolio is not significantly different from zero. We also form a zero-investment hedge portfolio that goes long in GC IPOs and short in non- GCs IPOs. To compare the short-term differences in performance between GC and non-gc IPOs, we incorporate those GC and non-gc IPOs with issue dates within six months prior to the calendar date in the portfolios. To compare the long-term performance differences, we incorporate those GC and non-gc IPOs with issue dates within 12 and 36 months prior to the calendar date in the portfolios. After creating portfolios, we regress the time series of portfolio returns on the three stock market factors: market, size and book-to-market, as suggested by Fama and French (1993), and examine the intercepts as abnormal monthly returns. We require at least five observations per month to form portfolios of each category. The reported p-values are based on heteroskedasticity-adjusted standard errors (White, 1980) to account for the changing portfolio constituents Multivariate Regression Analyses Our first OLS regression tests the determinants of IPO initial return using the following regression model: 18

20 INITIAL RETURN = ββ 0 + ββ 1 * (GC IPO) + ββ 2 * (LOG(IPO MCAP)) + ββ 3 * (PARTIAL) + ββ 4 *(SHARE OVERHANG) + ββ 5 *( MULTI-CLASS) + ββ 6 *( MARKET RETURN) + ββ 7 *( HIGH TECH) + ββ 8 *( VC- BACKING) + ββ 9 * (INTERNET IPO) + ββ 10 *(TOP TIER) + ββ 11 ( BIG5 AUDITOR) + ββ 12 *(EARNINGS NVE) + ββ 13 *( GC*VC-backing) + ββ 14 *( INSTI. OWN.) + ββ 15 *( LOG(ANALYSTS)) + εε ii Our first logistic regression examines the factors that influence the probability of being delisted by the stock exchanges of the newly listed companies within three years of their IPOs. The logistic regression model is specified as follows, where PP ii is the probability of being delisted: LOG (PP ii 1 PP ii ) = ββ 0 + ββ 1 * (GC IPO) + ββ 2 * (LOG(IPO MCAP)) + ββ 3 * (INITIAL RETURN) + ββ 4 *( PARTIAL) + ββ 5 *( MULTI-CLASS) + ββ 6 *( MARKET RETURN) + ββ 7 *( HIGH TECH) + ββ 8 *( VC- BACKING) + ββ 9 *(INTERNET IPO) + ββ 10 *(TOP TIER) + ββ 11 *( BIG5 AUDITOR) + ββ 12 *(EARNINGS NVE) + ββ 13 *( GC*VC-backing) + ββ 14 *( INSTI. OWN.) + ββ 15 *( LOG(ANALYSTS)) + εε ii All variables are explained in the Appendix Univariate results To begin our analyses, we examine univariate differences between key variables in our GC and non-gc IPO samples. 8 Table 2 presents the results. Underpricing for non-gc IPOs averages 14.1% compared to 4.9% for GC IPOs, and the difference is highly significant. Thus, despite the presumed higher risk of GC IPOs, they are significantly less underpriced, at least on a univariate level. Subsequent analyses will explore whether this difference is due to factors other than GC status. ***Insert Table 2 about here*** We explore several measures of aftermarket performance. First, we examine delisting rates within 3 years of the IPO. To identify delistings due to performance reasons, we rely on CRSP delisting codes and We find that 20% of GC IPOs delist within 3 years of the IPO 8 The appendix provides detailed information on each variable used in this paper. 19

21 compared to 6% of non-gc IPOs. This difference is economically and statistically significant, and it suggests that auditor GC opinions meaningfully predict eventual delisting. Next, we compare the frequencies of lawsuits between GC and non-gc IPOs. We find that 20.9% of non-gc IPOs are sued after their IPOs compared to 18.0% of GC IPOs. This difference is not statistically significant. If we focus on Section 11 lawsuits, which are for damages specific to IPO purchases brought under Section 11 or 12 of the Securities Act of 1933, the difference is about the same, but in the opposite direction (6.9% versus 9.9%). We also examine various measures of stock price and accounting performance. We first calculate 3-year buy-and-hold abnormal returns (BHARs) and cumulative abnormal returns (CARs), which are matched on size and book-to-market portfolios. In both cases, we use equal- and valueweighted returns. With the exception of the VW BHARs, GC IPOs underperform non-gc IPOs and the differences are economically large and statistically significant. Following Loughran and Ritter (1997), we also examine accounting performance using EBITDA/TA and CFO/TA in year 1 and year 3. EBITDA/TA is defined as operating income before depreciation, divided by beginning of period total assets. This is a standard measure of firm performance. operations, divided by beginning-of-period total assets. CFO/TA is defined as cash flow from CFO is the amount of cash generated from/used by a business enterprise s normal, ongoing operations during an accounting period scaled by total assets. Since the company s ongoing operations is regarded as its core business, this ratio provides additional information in addition to the traditional profit ratio such as EBITDA/TA. In every case, GC IPOs significantly underperform. Summarizing our univariate results, we find indications that GC IPOs underperform in the long-run and experience a higher performance-related delisting rate. At the same time, GC IPOs are significantly less underpriced than their non-gc IPO counterparts. However, there are no differences 20

22 between the two types of IPOs in terms of lawsuits following the IPO. Since we have shown that GC IPOs are different along many dimensions in Table 1, we next consider a series of multivariate analyses. 4. Multivariate results We begin by first modeling underpricing as a function of IPO, firm, and market characteristics. Our main independent variables of interest are GC IPO, Year2 GC and GC*VC-backing. GC IPO is a dummy variable that equals one if the IPO has a GC opinion in its initial public offering registration documents, zero otherwise. Year2 GC reflects whether the GC IPO gets an auditor s GC opinion for the second time on its second year annual report. GC*VC-backing is the interaction term between GC IPO and VC-backing status. Model (1) presents significantly negative coefficients for GC IPO, which replicates our univariate result and support prior literature. ***Insert Table 3 about here*** Starting with model (3) we also include a dummy variable VC to indicate whether or not the IPO was VC-backed. Consistent with Lee and Wahal (2004) and others, VC-backing is significantly positively related to underpricing. Our findings for ordinary IPOs deviate clearly from prior studies once we examine the importance of GC IPOs that were VC-backed. Model (4) illustrates that GC IPOs do not experience any different underpricing behavior compared to non-gc IPOs, confirming Hypothesis 2. Rather, we find that CG IPOs with VC-backing are significantly negatively related to underpricing, thereby we also confirm Hypothesis 3. We conclude that VC backing presents a certification role in GC IPOs, which is likely to reduce information asymmetry and lead to lower underpricing. On the other hand, we find that second year GC IPOs are associated with significantly greater underpricing, suggesting that the market can distinguish between firms that go public due to financial constraints and those that suffer from structural problems. 21

23 We also include other commonly used control variables such as the natural log of firms IPO market capitalization (Log(IPO mcap.)), the partial adjustment term (percentage difference between the offer price and the midpoint of the initial file range), and share overhang (pre-ipo shares scaled by shares offered). The latter two terms have been shown to be significant predictors of IPO underpricing (e.g., Hanley (1993), Bradley and Jordan (2002), Loughran and Ritter (2004)). We also include an indicator for multiclass IPOs (Smart and Zutter (2003)) and the cumulative market return one month prior to the offering. Partial and Log(IPO mcap.) are positive and highly significant, consistent with the literature. High Tech and Internet IPO are dummy variables that equal one if an IPO is classified as hightech and internet firm, respectively. High-tech is not significant, but Internet companies is positive and significant. We further include variables for third party certifiers of value. Top Tier and Big5 audit are dummy variables indicating whether an IPO retains a top tier underwriter (a Carter-Manaster ranking of 8 or higher) or a Big 5 auditing firm, respectively. As shown, Top Tier is negative and significant indicating that hiring a top tier underwriter is associated with about 4% less underpricing. Lastly, we consider institutional ownership and the natural log of one plus the number of analysts following the firm one year after the IPO. As in Aggarwal, Prabhala, and Puri (2002), institutional ownership is positively and significantly related to underpricing. The coefficient for analyst is positive, a result consistent with Cliff and Denis (2004), who suggest that issuing firms buy analyst coverage with initial returns. However, this coefficient is not statistically significant, possibly due to the changing regulatory environment after Regulation Fair Disclosure of 2000 or events such as the Global Settlement of Taken together, Table 3 suggests that VC-backed GC IPOs are less underpriced while second year GC IPOs are more underpriced. Next we study the impact of GC opinions on delisting. In Table 4, the dependent variable is a dummy variable identifying delisting within three years of the IPO, where delisting is performance-related (CRSP delist codes and ), so we 22

24 use a logit model. For better economic interpretation, we also report marginal effects (measured at the mean). IPO delisting frequencies have not been studied as extensively as IPO underpricing, so the determinants are less well understood empirically, and we restrict our analyses in Table 4 to the controls considered in Table 3. ***Insert Table 4 about here*** Models (1) through (5) suggest that GC IPOs are substantially more likely to delist. The marginal effect initially increases from 0.04 in model (2) to 0.07 in model (4), but loses statistical significance in subsequent models with additional control variables. Models (1), (3), and (5) further suggest that second year GC IPOs are significantly more likely to delist. However, once we control for institutional investors and the natural log of one plus the number of analysts following the firm one year after the IPO in models (7) and (8), neither GC IPOs nor second year GC IPOs are more likely to delist than non-gc IPOs. While models (6), (7), and (8) confirm Hypothesis 1b, we cautiously suggest that there might be no difference in likelihood of delisting for the first three pears post offering for GC and non-gc IPOs. However, this conclusion seems to depend significantly on the choice of control variables. The control for IPO market capitalization is associated with a significantly lower likelihood of delisting in models (1) through (4), while the market return variable is significantly positively related with delisting for all model specifications which is consistent with the notion that IPOs that timed the markets and issued after the hot markets are of lower quality and are more likely to be delisted after IPOs (Lowery and Schwert (2002), Loughran and Ritter (2004), and Demers and Joos (2007)). High-tech firms are more likely to delist, whereas a big 5 auditor in model (5) leads to a decreased likelihood of delisting. Negative EPS (Earnings nve) is positively related to delisting in all models this control variable is used (models (5) through (8)). In the next section we will examine the long-run returns in the three years post IPO. *** Insert Table 5 about here *** 23

25 In Table 5, we consider post-ipo returns in a calendar time setting. Specifically, each month over our sample period, we form portfolios containing all GC and non-gc IPOs, respectively, from the most recent 36 months. We consider both equal and value weighting. We require at least five IPOs in a portfolio for a month to be included. We also calculate the returns on a hedge portfolio that is short GC IPOs and long non-gc IPOs (Mitchell and Stafford (2000), Boehme and Sorescu (2002)). We regress the time-series of portfolio returns on the three Fama-French (1993) factors and the Carhart (1997) momentum factor, focusing on the intercept or alpha. Looking at the results for equal-weighting first, we see that the non-gc IPOs have a positive alpha of about 24 basis points per month, but the estimate is not statistically different from zero. In contrast, the GC IPOs have an alpha of -164 basis points per month, which is economically large and statistically significant (p =.039). The hedge return has an alpha of 188 basis points which again is highly significant, both economically and statistically. The hedge portfolio also shows that the factor exposures for the two portfolios are not substantially different. Turning to the value-weighted returns, the story is similar economically, but statistical significance drops. In particular, the non-gc portfolio has an alpha of essentially zero, whereas the GC portfolio has an alpha of -128 basis points per month, which is economically large, but not reliably distinguishable from zero (p =.141). The hedge portfolio has an alpha of 127 basis points, but, again, the point estimate is not statistically significant. Unlike the equal-weighted results, there is a noticeable difference in factor exposures in the value-weighted portfolios. In particular, the non-gc IPOs have a much smaller market beta, 1.125, compared to for the GC portfolio. The difference of.40 is economically and statistically significant (p =.057). Overall, Table 5 shows that the GC IPOs significantly underperform. When comparing the equal- versus value-weighted results, it is useful to keep in mind that the value-weighted portfolios are often dominated by a few IPOs. Further, value weighting means that as an IPO underperforms, it 24

26 receives a smaller weight going forward. In other words, value-weighting means that, all else the same, greater weight is being given to more successful deals as time passes. With equal-weighting, the reverse is true. The implicit rebalancing means that winners are sold off and losers are purchased. Equal weighting therefore tells us about the average IPO, whereas value weighting tells us about the average dollar invested in IPOs. Of course, small firms are most likely to be susceptible to misspecified model problems and, hence, equally weighted models lack robustness (Fama, 1998). The results in Table 5 provide mixed results and we cannot clearly reject Hypothesis 1a. *** Insert Table 6 about here *** In Table 6, we turn to operating performance. We first look at EBITDA at the time of the IPO. We then examine EBITDA in the subsequent three years, scaled by assets at the time of the IPO 9. Our inferences are consistent with the previous equally-weighted calendar time results; namely, GC IPOs have a lower operating performance. The underperformance is economically large and significant statistically. We also examine the change in EBITDA over the first three years (relative to assets at the beginning of the first year) and find much greater improvement in performance for the non-gc group. The GC IPOs in fact suffer degraded operating performance whereas the non-gc companies improve. In the second part of the table, we look at CFO instead of EBITDA. The results are essentially the same: The GC companies have a lower operating performance overall. 5. Additional analyses Are institutional investors more sophisticated than individual investors in recognizing the informational content of auditors GC opinions? Do analysts make larger errors when forecasting earnings for these riskier GC IPOs? Does the disclosure of a GC opinion immunize a newly listed 9 The scaling of EBITDA and CFO by assets at the beginning of the first year compensates the mechanical effect of retained earnings on future asset growth by deflating with a constant denominator measured at the beginning of period t. See Loughran and Ritter (1997) for more explanations on the rationale of this method. 25