How Important Are Relationships for IPO Underwriters and Institutional Investors? * Murat M. Binay Peter F. Drucker and Masatoshi Ito Graduate School of Management Claremont Graduate University 1021 North Dartmouth Avenue Claremont, CA 91711 murat.binay@cgu.edu (909) 621-8647 Vladimir A. Gatchev Department of Finance College of Business Administration University of Central Florida Orlando, FL 32816-1400 vgatchev@bus.ucf.edu (407) 823-3694 Christo A. Pirinsky Department of Finance and Economics Rutgers Business School 111 Washington Street Newark, NJ 07102 cpirinsky@business.rutgers.edu (973) 353-1147 January 2006 * We thank Ekkehart Boehmer, Stephen Brown (the editor), Bob Comment, Scott Lee, and an anonymous referee for helpful comments. All errors are our own.
How Important Are Relationships for IPO Underwriters and Institutional Investors? Abstract We find that underwriters favor institutions they have previously worked with when allocating Initial Public Offerings (IPOs). Regular investors benefit more than casual investors in IPOs by participating more in underpriced issues. Relationship participations are more important in the distribution of IPOs with stronger demand, IPOs of less liquid firms, and deals by less reputable underwriters. Overall, our results are consistent with book-building theories of IPOs and suggest that regular investors improve the efficiency of the IPO process. Interestingly, our results are weaker for 1999-2000, consistent with the idea that in this period other considerations (e.g., commission revenues from clients) affected the allocations of IPOs. 2
I. Introduction Ongoing relationships between institutional investors and underwriters have for long interested financial economists. Book-building theories of IPOs (e.g., Benveniste and Spindt, 1989; Benveniste and Wilhelm, 1990; Sherman and Titman, 2002) reason that establishing a network of regular investors benefits issuers by maximizing the proceeds from an issue. Consistent with these theories, Sherman (2000, 2005) argues that the possibility of forming regular institutional investor clienteles largely contributes to the growing popularity of book building as an IPO allocation mechanism. Recent studies, however, propose that in the late 1990s underwriters favored regular buy-side investors to benefit at the expense of highly underpriced firms; see, for example, Loughran and Ritter (2002) and Reuter (2005, forthcoming). Evidence in the regulatory settlements and the financial press indeed suggests that IPO underwriters have given favorable allocations of underpriced IPOs to institutional investors that generate large commission revenues. 1 Such evidence casts doubts on the efficiency of book-building as an allocation mechanism and questions the role of regular investors in the IPO process. Despite the importance of underwriter-investor relationships, empirical research provides little evidence on the role of regular investors in the equity issue process. One possible reason is that data on actual IPO allocations are proprietary and investment banks rarely disclose it. In this paper we empirically examine the role of underwriter-investor relationships in the IPO process by using institutional positions in IPOs from quarterly 13F disclosures to the SEC. 2 We develop a measure of 1 For example, Susan Pulliam and Randall Smith report in the December 7, 2000 issue of the Wall Street Journal, that the SEC and the U.S. attorney's office in Manhattan began a joint investigation of whether firms were violating securities laws by exacting unusually large trading commissions from clients as kickbacks in return for allocations of hot IPO stocks. Another article by Susan Pulliam and Randall Smith in the November 30, 2001 issue of the Wall Street Journal claims that Big mutual funds like AIM, a unit of London s Amvescap PLC, and Munder routinely get IPO allocations because they are steady customers of the underwriters, trading millions of shares at commissions of a few cents a share. In 2002 and 2003, Credit Suisse First Boston, Robertson Stephens, and J.P. Morgan agreed to pay fines of $100 million, $28 million, and $6 million, to resolve SEC and NASD investigations that linked IPO allocations to excessive brokerage commissions. 2 We acknowledge that end-of-quarter institutional holdings in IPOs are not necessarily equal to initial allocations; however, we expect that they are a close proxy of actual allocations. The data section of our paper and Reuter (2005, forthcoming) present further discussion on the issue. 3
relationship-based participation of institutional investors in IPOs. The measure is constructed as the difference of two probabilities -- the probability for institutional investors to participate in an IPO conditional on having some past participation in IPOs of the same lead underwriter and the unconditional probability for institutional investors to participate in the IPO. Based on this measure and a comprehensive dataset of virtually all IPOs underwritten between 1980 and 2000 we extend the IPO literature in several directions. First, we find that regular institutional investors (i.e., institutional investors with past participation in IPOs by the same lead underwriter) are more likely to participate in an IPO than are institutions with no prior participation in IPOs by the same lead underwriter. For the average IPO during 1980-2000, an institutional investor has a 1.98 percent unconditional probability of participating in an IPO while this probability for a regular institutional investor is 12.56 percent. Significant relationship participations are evident for IPOs with different levels of underpricing (even for overpriced IPOs), for IPOs by lead underwriters with different Carter-Manaster reputation ranks, and for IPOs in different industries. We further find that more (less) underpriced IPOs have higher (lower) relationship participation. Why are regular institutional investors favored in IPOs? In book-building theories (e.g., Benveniste and Spindt, 1989; Benveniste and Wilhelm, 1990; and Spatt and Srivastava, 1991) investment banks engage in information acquisition during the book-building process where investors provide informative indications of interest that help banks price and market the issue more efficiently. Underpricing then rewards investors for truthfully revealing favorable private information. Building a regular clientele of institutional investors allows underwriters to lower expected IPO underpricing by promising regular investors participation in future issues. The repeated game set-up gives the underwriter leverage over regular investors because investors will be favored in future issues only if the information they provide proves useful in valuing the IPOs. According to book-building theories, therefore, underwriter-investor relationships benefit issuers 4
through lower initial underpricing on average. Regular investors also benefit on the margin by increased profits from participation in multiple offerings. The importance of long-term relationships in book-building theories becomes apparent when information acquisition is endogenous. Sherman and Titman (2002) show that when information is costly underwriter ability to reduce underpricing could be substantially limited. In this scenario, the possibility to discriminate in favor of a particular group of regular investors allows underwriters to achieve lower underpricing by lowering the expected returns of uninformed investors (Sherman, 2000). 3 Agency-based explanations of IPO allocations reason that interests of issuers and underwriters are not always perfectly aligned in the distribution of IPOs. In self interest, underwriters build clienteles of favored buy-side clients to benefit at the expense of issuing firms (e.g., Loughran and Ritter, 2002). Agency-based explanations, however, need to provide a rationale for issuing firms to participate in IPOs that are underpriced more than necessary. Loughran and Ritter (2004) provide one such explanation; they argue that part of the higher underpricing in IPOs, especially in the late 1990s, reflects changes in the objective function of issuers. According to Loughran and Ritter (2004), in many of these highly underpriced IPOs the focus of the issuing firm was less on maximizing IPO proceeds and more on generating influential analyst coverage and allocating hot IPOs to the personal brokerage accounts of issuing firm executives. The potential conflict of interest between investment banks and issuers has attracted a lot of attention in the literature. Part of this attention is a result of the perception that shares have been allocated inappropriately given the large amount of money left on the table in the late 1990 s (e.g., Ritter and Welch, 2002). Reuter (2005, forthcoming) presents empirical support for the agency-based explanations by showing that IPO holdings of mutual fund families are increasing in the yearly brokerage commissions the funds pay to lead underwriters. Reuter further finds that the 3 Existing evidence is consistent with the predictions of book-building theories; see, for example, Hanley (1993), Cornelli and Goldreich (2001), Cornelli and Goldreich (2003), Lee, Taylor, and Walter (1999). Jenkinson and Jones (2004), however, find little evidence that more informative indications of interest lead to larger allocations thus questioning the validity of book-building theories. 5
sensitivity of IPO holdings to yearly brokerage commissions is highest for the IPOs with the highest underpricing. Both book-building and agency-based explanations of IPO underpricing predict that particular groups of investors would be favored in the allocation of new issues. The underlying motives for such favoritism towards regular investors, however, are different in the two views. Book-building theories predict that, even though regular investors receive favorable allocations in underpriced IPOs, they help investment banks price and distribute the issue and ultimately reduce IPO underpricing of the average issuer. Agency-based explanations, on the other hand, predict that when expected IPO underpricing is high investment banks favor regular investors that then trade more and return part of the underpricing through higher trading commissions. 4 In order to understand better the role of regular investors in the IPO process, we explore the cross-sectional determinants of relationship participations in IPOs. We confirm that more (less) underpriced IPOs have higher (lower) relationship participations. Book-building theories propose that IPO underwriters favor regular investors, so that when the demand for the IPO revealed during the book-building process is high, regular investors have preferential access to the IPO. In line with this proposition, we find that IPOs with stronger demand revealed during the book-building process (as proxied by higher price adjustment from the mid-point of the filing range to offer price) have higher relationship participation. We also find that relationship participations are higher in smaller firms. To the extent that smaller firms have higher issue costs and lower stock market liquidity, this result is consistent with book-building theories, where relationship investors help underwriters with the placement of issues that are more costly to place. Interestingly, the IPO price adjustment and issuer size are insignificant determinants of IPO relationship participations in 1999-2000. The results are consistent with existing evidence that in the late 1990s underwriters allocated shares based on other criteria (e.g., commission revenues from clients) rather than book-building 4 Agency-based explanations do not exclude the possibility that competing investment banks provide issuers with other benefits (not only price discovery and distribution through regular investors) that compensate issuers for excess underpricing. 6
considerations. We, therefore, find support for the idea that during 1999-2000 agency-based considerations affected the role of regular investors in the IPO process. Underwriter reputation, while not explicitly modeled in the book-building theories, could also be linked to relationship participations. If underwriters with higher reputation produce better information about the IPO (Carter and Manaster, 1990) then IPOs underwritten by such underwriters will have less need for the participation of relationship investors for information production. In support of this prediction, we find that IPOs underwritten by banks with higher Carter-Manaster ranks have lower relationship participation. We further find that IPOs with larger syndicates have lower relationship participation, probably because other banks in the syndicate also use their distribution channels so that the lead bank is not the only one that has to rely on regular investors. Our results are related to Reuter (2005, forthcoming), who argues that existing relationships between investors and underwriters increase investor access to underpriced IPOs. In Reuter (2005, forthcoming), relationships are measured by the magnitude of commissions that investors generate for a given underwriter. The approach in our paper is more closely related to existing book-building theories that do not explicitly analyze how other lines of business (e.g., brokerage services) should influence the IPO allocation decision. Unlike Reuter (2005, forthcoming), we examine the relation between IPO underpricing and holdings of institutional investors with ongoing relationships solely in equity issues. In sum, we present evidence for preference of underwriters towards particular clienteles of institutional investors in the allocation of IPO shares. Consistent with predictions of book-building theories, we find that regular investors during 1990-1998 participate in IPOs when demand is high and contribute to the efficiency of the IPO process by facilitating the distribution of offerings of smaller firms. Furthermore, we find that regular investors are more important for IPOs underwritten by less reputable underwriters and for IPOs with smaller syndicates. While our results are consistent with predictions of book-building theories for IPOs during the 1980s and the early 1990s, the results 7
weaken during 1999-2000. We conclude that agency-based considerations in the late 1990s have affected the role of regular investors in the IPO process. The rest of the paper is organized as follows. Section II discusses the data and summary statistics; Section III defines the measure of relationship participation; Section IV analyzes the determinants of relationship participations in IPOs; and Section V concludes. II. Data and Sample Characteristics A. Data We identify the IPOs in our sample from the Thomson Financial Securities Data Corporation (SDC) New Issues database. The sample includes only IPOs for U.S. common stocks during the 1980-2000 period, excluding unit offerings, spin-offs, Real Estate Investment Trusts (REITs), closed-end funds, and American Depository Receipts (ADRs). We also exclude all IPOs of financial companies, IPOs with offer prices below five dollars, and IPOs with missing first two-day information from CRSP. The resulting sample consists of 4,668 IPOs. We obtain issue-specific data including the offer date, IPO proceeds, offer price, and the syndicate composition for our IPO sample from the SDC database. First-day closing prices and SIC code data come from the CRSP daily and monthly files and accounting information comes from the Compustat annual files. The institutional holdings data for our IPO sample originates from the CDA Spectrum 13F Filings database. Under the 1978 amendment to the Securities and Exchange Act of 1934, all institutional investors managing a portfolio with an investment value of $100 million or more are required to file quarterly 13F reports to the SEC listing their equity positions greater than 10,000 shares or $200,000 in market value as of the last date of each quarter. For each IPO in our sample we obtain the end-of-quarter shareholdings for each institutional investor from the CDA Spectrum database. 8
To measure institutional IPO participations, we use the holdings of all institutional investors as reported in the Spectrum database at the end of the issuing quarter. We acknowledge that end-ofquarter institutional holdings are not exactly equal to the actual initial allocations at the day of the IPO. There are two potential problems with using SDC data to proxy for IPO allocations. First, not all institutional investors are required to disclose their holdings and second, institutions can potentially flip or purchase shares between the IPO date and the institutional ownership filing date (at the end of the quarter). Even though not exact, we expect this proxy to be a highly representative statistic of the actual participations based on evidence presented in several previous studies. Using a proprietary data on 38 IPOs managed by a single underwriter during the 1983-1988 period, Hanley and Wilhelm (1995) find that the correlation between original institutional allocations and post-offer reported holdings on 13F files is 0.91. 5 Aggarwal (2003) further shows that flipping during the first two days of trading accounts for only 15% of the shares offered in her sample; so immediate flipping activity by institutional investors is of relatively low magnitude (see also Reuter, 2005, forthcoming for a discussion of post-ipo institutional holdings as a proxy for IPO allocations). B. Summary Statistics Table 1 reports the number of IPOs, institutional participation in IPOs, offer proceeds, and underpricing for our sample by year. The total number of IPOs with available data is 4,668 and 4,107 of these IPOs have positive institutional ownership at the end of the quarter during which the IPO takes place (hereafter known as the IPO quarter). The percentage of IPOs with institutional ownership increases over the time period. In the 1980s usually less than 80% of the IPOs have positive institutional ownership but by the late 1990s institutions participate in around 97% of the IPOs. Offering proceeds (expressed in 1980 U.S. dollars using the CPI index) from the average IPO 5 This number could be misleadingly high, however, because most institutions that are allocated zero shares subsequently hold zero shares. The more relevant correlation would be computed conditional on an institution receiving a positive allocation. 9
increase substantially over time. In 1980, for example, the average IPO, excluding the overallotment option, raises around $14 million while by 2000 the average IPO raises $98 million. As documented by numerous previous studies (e.g., Ritter and Welch, 2002), we also observe that the average IPO is substantially underpriced -- the average first-day return in our sample is 20.4%. The underpricing in the late 1990s stands out with average underpricing reaching 77% in 1999. The last two years of our sample (1999 and 2000) represent the peak of the high-tech bubble and are dominated with IPOs of Internet firms. [Insert Table 1 about here] Table 2 reports the distribution of initial institutional ownership for the IPO sample firms. The institutional investor data is free of survivorship bias and covers a total of 3,174 different institutions. Consistent with the growth in the institutional investor universe during the past two decades, the number of institutions in our sample increases from 569 in 1980 to 2,062 by 2000. Institutional interest in the IPO market grows significantly over time, as well. While only 119 institutions (21% of the institutional investor universe) report IPO share ownership at the end of the IPO quarter in 1980, this number increases to 684 (33% of the institutional investor universe) by the end of our sample period. The last four columns of Table 2 present distributional characteristics of the fraction of institutional participation in the IPO. The fraction is calculated as the ratio of the total shares owned by institutions at the end of the IPO quarter to the total number of shares offered at the IPO. We choose to normalize by the number of shares offered at the IPO instead of the number of shares outstanding at the end of the IPO quarter because the latter includes pre-issue owners, such as insiders and venture capitalists. We have also replicated all of the major tests in this paper, normalizing with total number of shares outstanding; our major results are robust with respect to this alternative specification of institutional participation. Table 2 shows that the mean and median institutional participation in IPOs are around 48%. This number appears to be lower than the numbers reported in the studies of Hanley and Wilhelm 10
(1995) and Aggarwal, Prabhala, and Puri (2002). 6 The two papers, however, recognize that their samples are biased toward larger offerings, measured by both proceeds and total assets. Since institutional investors display preference toward larger stocks (e.g., Gompers and Metrick, 2001), the institutional allocation figures reported in the above studies potentially overestimate the averages for the overall IPO population. Indeed, the offer-size-weighted average fraction of institutional ownership in IPOs in our sample is 65.5%, which is very close to the averages in the above papers. There is another possible reason for the lower institutional ownership in our sample as compared to the above two papers: although Spectrum reports holdings of some hedge funds, the majority of hedge funds are not covered by the database while hedge funds should show up in the institutional allocation data of the above two studies. Consistent with the evidence in Table 1, we again find that percentage institutional participation in IPO firms increases over time. The average institutional IPO participation in 1980 is 14% while the average institutional IPO participation by 2000 rises to 74%. [Insert Table 2 about here] III. The Relationship Participation Measure A. Construction of the Measure Relationship implies a joint history. Therefore a statistic designed to quantify the role of relationship between underwriters and institutional investors should account for their joint participation in past IPO deals. We build a relationship participation measure indicating the propensity of institutional investors to participate in an IPO conditional on their involvement in past IPOs of the same lead underwriter. 6 For example, Hanley and Wilhelm (1995) use a sample of 38 IPOs managed by a single underwriter during the 1983-1988 period and find an average institutional allocation with a mean of 67% (median of 72%); Aggarwal, Prabhala, and Puri (2002) use a sample of 174 IPOs managed by nine underwriters between May 1997 and June 1998 and find a mean institutional allocation of 73% (median of 74%). 11
The measure of relationship participation ( R i ) for IPO i is the difference between the probability of institutional participation conditional on past participation and the unconditional probability of institutional participation (expressed in percent). R i R ni n i = 100 R N N (1) The unconditional probability of institutional IPO participation ( ni N ) is the number of institutions participating in IPO i ( n i ) divided by all institutions present at the time of the IPO ( N ). Institutions that never participate in an IPO are excluded from the analysis. The probability of institutional participation in IPO i conditional on past relationships R R ( i ) n N is constructed as follows. First, for every IPO and for every institution with F13 filing at the end of the issuing quarter, we find whether the institution has participated in any of the past ten IPOs underwritten by the same lead bank within five years of the current IPO. If the lead bank has less than ten past IPO deals within the past five years then we use all available past IPO deals. We calculate the conditional probability of participation as the number of institutions participating in the IPO that R also have past IPO participation with the same underwriter ( i ) n divided by the number of all institutions present at the time of the IPO that have past IPO participation with the same lead R underwriter ( N ). This measure has the intuitive interpretation of excess participation probability. A positive value of the relationship participation measure would provide evidence for the existence of relationship participation in the IPO, since institutional investors with stronger past IPO business relationship with the underwriter would be more heavily represented in the current IPO. When constructing our measure, we concentrate on lead underwriters since they are most important in the allocation decision and many offerings, especially in the 1980s, are certified by a 12
single underwriter. 7 We exclude 148 IPOs with more than one lead underwriter because for these IPOs the definition of the relationship participation measure is ambiguous. We have also calculated a version of the relationship allocation measure assigning equal weights to all syndicate members in the IPO and all major results in the paper are robust with respect to this alternative specification. Mergers between investment banks and mergers between institutional investors also present a challenge since we do not know which relationships remain within the resulting entity. This problem is present mainly in the first several deals of the merged entities. After the merged entities have existed for several deals, our measure captures the relationships established in those deals so the event of the merger becomes less relevant. We assume that the new entity resulting from a merger has no previously established relationships with institutional investors. As a robustness check, we have also replicated all tests in the paper excluding the first ten deals following merger of underwriters, and all results are qualitatively similar. 8 B. Properties of the Measure One of the main questions that this paper addresses is why different IPOs command different levels of relationship participation. We start by examining the relationship participation for IPOs with different levels of underpricing, IPOs with different lead underwriter reputation, and IPOs in different industries. Existing research documents that IPO underpricing has changed over time (Loughran and Ritter, 2004). In order to control for potential nonstationary behavior in the variables and the relations between variables we perform our analysis in this and the remaining sections of the paper separately for several periods as well as for the whole 1980-2000 period. 9 7 See Chen and Ritter (2000). 8 We obtain the sample of major mergers of underwriters from Corwin and Schultz (2005). 9 As another example, Houston, James, and Karceski (2006) find that for 1999-2000 (but not before) IPO offer prices were set at a discount relative to comparable firm valuations. 13
Finally, it is possible that agency conflicts were stronger in 1999-2000 when IPO underpricing was high and so it is important that we perform separate analysis for this period. We are able to estimate the relationship participation measure for 3,598 (87%) of the IPOs in our sample. The sample is reduced because the estimation of the relationship participation measure requires historical deals at most five years prior to the examined IPO. Therefore, we lose the first IPO of each underwriter in the sample within a five year period. Table 3 presents distributional characteristics of the relationship participation measure for the whole sample and underpricing quintiles. We find strong evidence for the existence of relationships between underwriters and institutional investors in the IPO market -- the mean IPO institutional relationship participation is 10.57%, indicating that underwriters allocate more shares to institutional investors that they have done prior business with. Over our period of 1980 to 2000 average relationship participation is not stable, however. In the 1980s, relationship institutional investors have a chance of participating in an IPO 8.39 percentage points higher than the average institutional investor. In the period of 1999 to 2000, in contrast, institutions with past relations with the lead underwriter are 13.8 percentage points more likely to participate in an IPO than the average institution. An alternative interpretation of our estimates is as follows. The average IPO in our sample is underpriced by 21.32%. An institutional investor has an ex-ante probability of participation in the average IPO of 1.98% so that ex-ante an institutional investor expects a return of 0.42% from the average IPO. Because regular investors have a higher ex-ante probability of participating in an IPO, their ex-ante expected return is 2.68%. Therefore, regular institutional investors expect to receive an additional 2.25% return from the average 1980-2000 IPO as compared to the average institution. The additional expected returns for regular investors from underpricing vary over time. For the 1980s, regular investors expect to get an additional 0.69% return from the average IPO while this number for 1999-2000 is 10.05%. One caveat is in place. The measure of relationship participation could on average be positive due to the tendency of some institutional investors to actively participate in the IPO market, 14
regardless of lead underwriter. Such investors are still regular IPO investors, but they are not associated with a particular underwriter (or a set of underwriters). In the cross-section of IPOs, however, these active IPO investors equally affect the relationship participation measure for all IPOs. Therefore, only institutional investors that participate in IPOs conditional on the lead underwriter drive the cross-sectional dispersion in relationship participation. Since our subsequent results rely on the variation in relationship participation across IPOs they should be unaffected by this concern. To further address this issue, we examine the IPO participation frequencies of institutional investors. The average institution in our sample participates in 44 IPOs while the median institution participates in seven IPOs. The most active institution in our sample is Fidelity that participates in 69.10 percent of the IPOs (1,159 IPOs). Even though there appear to be some institutional investors that actively participate in the IPO market (i.e., not necessarily because of any relations with a particular underwriter) the average as well as the median institutional investor in our sample participate in few enough IPOs to establish relations with a limited set of underwriters. We also examine the participation rates of relationship investors for IPOs with different levels of underpricing. We allocate our IPOs into five portfolios based on underpricing and calculate the average relationship participation for each portfolio. The first portfolio consists of IPOs with first-day closing price lower than the offer price (overpriced IPOs). The other four portfolios are created using the IPO underpricing quartiles for IPOs with first-day closing price higher than the offer price (underpriced IPOs). Our results show that more underpriced IPOs have higher institutional participation overall. For the whole period of 1980-2000, for example, institutional investors have 1.33% chance of participating in overpriced IPOs and 2.92% chance of participating in the most underpriced IPOs on average. More importantly for this study, however, we find a monotonic positive relation between IPO underpricing and relationship participation. We find that the relationship participation of institutional investors is lowest (6.64%) for overpriced IPOs while it is highest (15.61%) for the most underpriced IPOs. The F-test indicates that the average levels of 15
the relationship participation measure are significantly different across the underpricing portfolios. The results are robust for the three time periods we examine. Now we interpret our results in terms of ex-ante expected returns from an IPO. Our estimates suggest that, given the positive association between IPO underpricing and the participation of regular investors, ex-ante regular investors expect an additional 3.18% return from underpricing as compared to the average institution (compare with an expected return of 2.25% when we do not take into account the positive link between underpricing and participation of regular investors). We again find that the additional ex-ante expected return for regular investors increases over time; it is lowest in the 1980s (1.07%) and highest in 1999-2000 (13.20%). Overall, the results show that regular investors receive an economically significant return from disproportionately higher participation in more underpriced IPOs. [Insert Table 3 about here] The initial results indicate that IPO underpricing is related to the allocation decision of underwriters -- more regular investors participate in more underpriced issues. These results are consistent both with book-building theories and with agency-cost explanations of IPO underpricing. One of the predictions of book-building theories, not shared by the agency cost theories, is that regular investors will be willing to occasionally participate in overpriced IPOs. We, therefore, expect that the link between underpricing and relationship allocations will be non-monotonic. Our findings in Table 3, however, do not support this conjecture. Existing literature recognizes the important role of the underwriting bank reputation for the IPO process. We, therefore, examine whether relationship participations in IPOs are primarily driven by the reputation of the lead underwriter. As a measure of underwriter reputation we use the Carter-Manaster rank of the lead underwriter. 10 We create five portfolios based on the reputation of the lead underwriter. IPOs with lead banks Carter-Manaster reputation ranks below four form the 10 See Carter and Manaster (1990), Carter, Dark and Singh (1998), and Loughran and Ritter (2004). We obtain the reputation ranks from the website of Jay Ritter, (http://bear.cba.ufl.edu/ritter/ipodata.htm). 16
first portfolio. The cut-off points for the other four portfolios are Carter-Manaster ranks of six, seven, and eight. We examine the distribution of relationship participation for IPOs underwritten by underwriters with different reputations in Table 4. In Panel A of Table 4 we find that average relationship participations in IPOs are similar across IPOs underwritten by underwriters with different reputations. For example, the average relationship participation for underwriters with Carter-Manaster reputation rank between four and six equals 9.55 while the relationship participation for underwriters with reputation rank between seven and eight is 10.42. The two numbers are statistically indistinguishable from each other. If top-tier underwriters are more likely to allocate shares to Fidelity, or other large mutual fund, we should expect to find that top-tier underwriters have higher relationship participations. As noted above, we do not find that in the overall sample. However, for the last two years of our sample we do find that relationship participations are higher for the top-tier underwriters when examining medians. Further examination reveals that relationship participations (even with similar underwriter reputations) vary substantially across IPOs. For IPOs with lead underwriter Carter-Manaster ranks between four and six the standard deviation of IPO relationship participations equals to 10.99. The 25 th percentile of relationship participations for the same group of IPOs is 2.62 while the 75 th percentile is 13.26. The results for other Carter-Manaster reputation ranks also show a substantial variation in relationship participations. The variation in relationship participation, however, differs across reputation ranks. The variation in relationship participation is lower for high reputation underwriters (i.e., Carter-Manaster reputation ranks close to nine) and higher for low reputation underwriters (i.e., Carter-Manaster reputation ranks further away from nine). So while more reputable underwriters on average do not have higher relationship participation, their relationship participation is more stable. 11 11 In unreported tests we also estimate the correlation between underwriter reputation and relationship participation in an IPO for every year in our sample. In most of the years the correlation between the two variables is indistinguishable from zero. 17
When we examine the relationship participation across Carter-Manaster underwriter reputations for three sub-periods, we find similar results. Panel B of Table 4 examines relationship participations between 1980 and 1989, Panel C looks at the 1990-1998 period, while Panel D looks at the last two years of our sample. [Insert Table 4 about here] Overall, the evidence in Table 4 provides the following insights. Relationship participations have nontrivial variation across IPOs. This variation is largely unexplained by the variation in lead underwriter reputation. However, more reputable underwriters appear to have more stable relationship participations than less reputable underwriters. It is possible that the dispersion in relationship participation is mostly driven by the industry of the IPO firm. In order to address this possibility we examine the relationship participations for two portfolios of firms. Internet and technology firms form the first portfolio while all other firms form the second portfolio. 12 The results are presented in Table 5. Overall, we find that relationship participations for Internet and technology firms are 12.07 percent while for other firms they are 9.48 percent suggesting that relationship participations are more important for Internet and technology firms. When we examine the difference in relationship participations between our two portfolios for different sub-periods we still find that Internet and technology firms command higher relationship participation than other firms. In the last period of 1999-2000 Internet and technology IPOs have average relationship participation of 14.26 percent while the remaining IPOs have an average relationship participation of 11.92 percent. One reason for our finding that Internet and technology firms have higher relationship participations could be that these firms have higher 12 We obtain the list of Internet IPOs from the website of Jay Ritter, (http://bear.cba.ufl.edu/ritter/ipodata.htm). Technology IPOs are those in SIC codes 3571, 3572, 3575, 3577, 3578 (computer hardware), 3661, 3663, 3669 (communications equipment), 3671, 3672, 3674, 3675, 3677, 3678, 3679 (electronics), 3812 (navigation equipment), 3823, 3825, 3826, 3827, 3829 (measuring and controlling devices), 3841, 3845 (medical instruments), 4812, 4813 (telephone equipment), 4899 (communications services), and 7371, 7372, 7373, 7374, 7375, 7378, and 7379 (software). 18
uncertainty and therefore are more difficult to value (Loughran and Ritter, 2004). As a result, regular investors are invited to facilitate the pricing of the issue. [Insert Table 5 about here] In Table 6 we present the average relationship participations for several investment banks for the years of 1985, 1990, 1995, and 1999. For all of the examined underwriters and for all years we find that relationship participations are positive. We also find that within a given year, there is a difference in relationship participations across different underwriters. Furthermore, there is a significant variation in relationship participation across years for a given underwriter. Similar to the general trend, we find for example that relationship participations for IPOs lead by Goldman Sachs have increased from 5.24 percent in 1985 to 14.25 percent in 1999. This trend is less evident for the rest of the underwriters when examined individually. While the primary objective of Table 6 is to provide a more transparent view at our relationship participation measure, it also suggests that our previous results are unlikely to be limited to a specific underwriter. [Insert Table 6 about here] Thus far, we find that past relationships between underwriters and institutional investors are an important factor in the participation of initial public offerings. The preliminary results also indicate that IPO underpricing is related to the allocation decision -- more regular investors participate in more underpriced issues. In the next section we study in greater detail the determinants of relationship-based participations and IPO-underpricing. IV. The Role of Ongoing Relationships in the IPO Process In this section we examine the determinants of relationship participation in IPOs. At the same time, we also examine the link between relationship participations and IPO underpricing. 19
Since IPO underpricing and relationship participations are simultaneously determined during the IPO process, we use a system of simultaneous equations to perform the analysis. A. Determinants of Relationship Participation We begin by outlining the variables used to explain IPO relationship participations. One of the main predictions of book-building as well as agency-based theories is that regular investors participate disproportionately more in highly underpriced IPOs. IPO underpricing is the return from the IPO offer price to the first day closing price. We also examine how firm size and issue size affect relationship participation. Smaller firms are generally considered to have lower stock market liquidity and if regular investors support the liquidity of the issue we expect to find that smaller firms have higher relationship participations. Furthermore, larger offerings, ceteris paribus, require more effort to sell and it is possible that investment banks use regular investors to help them with the flotation of larger issues. Additionally, Barry (1989) and Habib and Ljungqvist (2001) argue that owners incentives to reduce underpricing are positively related to the size of the offering. If the participation of regular investors reduces IPO underpricing, we again expect larger issues to have higher relationship participations. In general, however, larger offers should have more investors, so that could also lead to a positive relation between relationship participations and offer size. We measure firm size by the assets of the firm in millions of 1980 U.S. dollars (Compustat item 6) and issue size by the IPO proceeds, excluding the overallotment option, in millions of 1980 U.S. dollars from SDC. Benveniste and Spindt (1989) propose that, ceteris paribus, proceeds-maximizing investment banks should give priority to regular investors when allocating IPO shares. A direct implication of book-building theories, therefore, is that when IPO demand is low casual investors have a higher chance of receiving IPO shares than when demand is high. Benveniste and Spindt (1989) further argue that high (low) demand reveals positive (negative) information so that the offer price is adjusted upward (downward). Following this argument, we use the price adjustment from 20
the midpoint of the filing price range to the offer price as a proxy for IPO demand. We expect that IPOs with higher price adjustments (higher demand) will have higher relationship participations. In addition, Benveniste and Spindt (1989) predict that high demand revealed in the book-building process is not fully reflected in the offer price so that offer prices do not fully adjust to positive information. Hanley (1993) finds evidence consistent with this prediction. In our paper we use both upward price adjustments and downward price adjustments to explain IPO relationship participations. IPO firms substantially differ in their characteristics and some IPOs are considered to be of higher quality than others. For example, VC backed (non-backed) IPOs, IPOs with higher (lower) cash-flows, and IPOs in industries with higher (lower) valuation ratios are on average considered to have higher (lower) quality. Whether relationship investors should participate more in higher quality IPOs or more in lower quality IPOs is, however, unclear. One hypothesis is that investment banks can easily place higher quality firms with non-relationship investors and so we should expect that relationship investors participate relatively more in lower quality IPOs. The alternative hypothesis predicts that relationship investors would benefit from their relations with the underwriter by having higher participations in IPOs of higher quality firms. We obtain IPO VC backing from SDC while the cash flow-to-assets of the firm after the IPO we compute form Compustat as net income (item 172) plus depreciation (item 14) over book value of assets (item 6). We calculate the book-tomarket equity ratio of the IPO industry in the year prior to the offering using Compustat (item 60 divided by item 199 times item 25). Underwriter reputation plays an important role in the IPO process. The link between underwriter reputation and the participation of regular investors in the IPO has not been examined in previous literature. On the one hand, it is possible that more reputable underwriters have achieved their reputation through establishing stronger networks of regular investors. If this is the case then we would expect to find that IPOs underwritten by more reputed investment banks also have higher participation of regular investors. Alternatively, more reputed underwriters may have wider access 21
to investors and consequently higher bargaining power than less reputed underwriters. In that case, highly reputed underwriters do not need to use the services of regular investors as frequently as less reputed underwriters do. This alternative hypothesis suggests that underwriter reputation is negatively related to the participation of regular investors in the IPO. As a measure of underwriter reputation we again use the Carter-Manaster reputation ranks discussed in the previous section. We also use several control variables that could have an effect on our results. We construct our measure of relationship participation based on the lead underwriter only. It is possible that for larger syndicates the relationship participation of the lead underwriter is lower, not because the lead underwriter does not allocate shares to regular investors but because a significant part of the shares are allocated by syndicate members to their own clients. We, therefore, control for the number of syndicate members of an IPO in our regressions. The results presented in Table 3 suggest that the measure of relationship participation is positively related to the unconditional probability of institutional participation in an IPO. In order to ensure that our regression results are not driven by such covariation, we also use the probability of institutional participation in an IPO as another control variable. Different investment banks may establish different levels of regular investor participation in their IPOs. In order to control for this possibility, we construct a measure of historic IPO relationship participation for the lead underwriter of the current IPO. We use the past ten deals over the past five years for the lead underwriter to construct this variable. On average, Internet and technology IPOs involve higher uncertainty than other IPOs. As control variables we include an Internet IPO indicator variable and a technology IPO indicator variable. Internet and technology firms are defined in the previous section. And finally we control for the year and the industry of the IPO, by including year and industry fixed effects. We use the 12- industry-classification from Kenneth French s web site. 13 13 Professor French s website is at http://mba.tuck.dartmouth.edu/pages/faculty/ken.french. We thank him for generously providing these data. The definitions of the industries are provided on the website. 22
One of the potential determinants of IPO relationship participation, as discussed above, is IPO underpricing. But IPO underpricing itself may be driven by the need for participation of regular investors in the IPO. In order to allow for this possibility we also model the determinants of IPO underpricing. As determinants of IPO underpricing we again use the VC backing indicator variable, firm size, IPO syndicate size, cash flow-to-assets after the IPO, the average book-to-market ratio of the IPO industry in the previous year, the IPO price adjustment from the midpoint of filing price range to offer price, the reputation of the lead underwriter, the probability of institutional participation, and Internet and technology indicator variables. We again include year and industry fixed effects. We also use the average underpricing of all IPOs three months prior to the current offering as a determinant of IPO underpricing. Ibbotson, Ritter, and Sindelar (1994) and Lowry and Schwert (2002) show that IPO activity evolves in cycles, so this variable is designed to control for the general level of activity in IPO-markets at the time of the issue. In order to model the dependence of underpricing on the participation of regular investors, we include two additional determinants of underpricing. The first determinant is the relationship participation in the current IPO. We expect that higher relationship participation should lead to higher underpricing. The second determinant of underpricing is the historic level of relationship participation for the lead underwriter of the IPO. We average the relationship participation measure of the lead underwriter for the past ten deals over the past five years to construct the measure of historic participation of regular investors. Book-building theories of IPOs propose that investment banks can establish a regular investor base to reduce the overall IPO underpricing. In this case historic levels of relationship participations of the lead underwriter would be negatively related to the underpricing in the current IPO. However, all else equal, underwriters that always favor the same investors will have smaller networks relative to underwriters that allocate shares to different investors. If larger networks can help reduce underpricing (as suggested by Sherman and Titman, 2002) then we should actually expect to find that higher relationship participations lead to higher 23