Multiple Bookrunners, Bargaining Power, and the Pricing of IPOs

Transcription

1 Multiple Bookrunners, Bargaining Power, and the Pricing of IPOs Craig Dunbar a * and Michael R. King a a Ivey Business School, Western University, 1255 Western Road, London Ontario, N6G 0N1, Canada This version: January 2, 2018 PRELIMINARY AND SUBJECT TO REVISION Abstract We examine the dynamics of underwriting syndicates for U.S. initial public offerings (IPOs) from 2002 to We find that much of the increase in IPOs with multiple bookrunners is due to the inclusion of passive non-lead bookrunners who receive lower share allocations than active lead bookrunners. Adding lead or non-lead bookrunners improves IPO outcomes for issuers by lowering underwriting spreads, price adjustments and first-day returns. The exception is IPOs in which a sole lead bookrunner is paired with non-lead bookrunners, where first-day returns are significantly higher. These findings suggest issuers are using their bargaining power to push for multiple lead and non-lead bookrunners, which is associated with better IPO outcomes, except for sole-lead IPOs where underwriters have more bargaining power. JEL classification: G24, G32, C22 Keywords: Initial public offerings, underwriters, syndicates, share allocation, underpricing, bargaining, title inflation. Acknowledgements: For helpful comments we are grateful to Walid Busaba, Michael Dambra, Matthew Gustafson, Michelle Lowry, Anya Mkrtchyan, Saurin Patel, Felipe Restrepo, and Jay Ritter and seminar participants at Western University and the 2017 Northern Finance Association. * Corresponding author: Tel: ; Fax: , address: cdunbar@ivey.uwo.ca (C. Dunbar)

2 1. Introduction We examine the composition and dynamics of underwriting syndicates in U.S. initial public offerings (IPOs) from 2002 to 2014 and study the impact of multiple bookrunners on IPO outcomes. In 2002 the average IPO had 1.34 bookrunners but by 2014, this number had increased to 3.28 bookrunners. This increase in the average number of bookrunners is many times greater than the modest increase in average deal size, suggesting some type of structural change in IPO syndicate dynamics around the global financial crisis. We consider whether changes in the relative bargaining power of issuers vs. underwriters can explain these changes. And we examine how the increase in multiple bookrunners has affected price adjustments over the bookbuilding period, underwriting spreads, share allocations to underwriters, first-day returns, and equity analyst coverage following the IPO. We show that most of the increase in average number of bookrunners is due to the inclusion of banks who are allocated significantly fewer shares than other bookrunners. Market participants refer to these banks as passive bookrunners, but we use the term non-lead bookrunners. 1 Nonlead bookrunners receive equal credit for league table purposes but their compensation is much lower as they receive a lower share allocation and therefore lower underwriting fees. If we define lead bookrunners as any underwriter with at least 90% of the highest share allocation in an IPO, by 2014 around three-quarters of IPOs featured non-lead bookrunners, up from 5% in The average number of non-lead bookrunners increased from 0.06 in 2002 to 1.80 in Despite the rise in non-lead bookrunners, the average number of lead bookrunners has remained stable since 2002, with between one to two lead bookrunners allocated around half of 1 Barzel, Habib and Johnsen (2006) discuss the emergence of passive bookrunners. This phenomenon was confirmed through interviews with equity capital markets executives at Citigroup, Morgan Stanley, Bank of Montreal and RBC in New York in November

3 the shares in an IPO. Instead of accepting reduced allocations, boorkunners have allocated a smaller share of IPOs to non-bookrunning managers who share has declined steadily from around half in 2002 to less than 20% in In other words, underwriters who used to be classified as co-managers (or syndicate members) have been able to break into bookrunner roles, even when they are present in title only. Some IPOs even feature non-lead bookrunners who have no equity trading desk, no equity salesforce, and no equity research analysts. We examine how changing syndicate composition and dynamics affects the bargaining position of issuers and ultimately IPO outcomes. This paper tests alternative explanations for the use of multiple lead and non-lead bookrunners in IPOs. We examine this issue from two perspectives: the point of view of the issuer, and the point of view of the lead bookrunner(s). 2 Issuers should prefer to pay lower costs to go public, all else equal, consisting of lower underwriting spreads and lower first-day IPO returns. Issuers should also prefer more equity analyst coverage as extra attention potentially has a positive effect on longer-term share performance (Merton, 1987). The lead bookrunner(s), on the other hand, should prefer better economics for each deal, consisting of a higher underwriting spread and/or greater share allocation. 3 The lead bookrunner(s) should be concerned with how the current IPO impacts future deal flow, and should prefer a syndicate structure and IPO pricing that increases the probability of receiving future equity mandates (Ljungqvist, Marston, and Wilhelm, 2009). Finally, all else equal, the lead bookrunner(s) should prefer an IPO with higher first-day returns as less effort is required to market an underpriced IPO (Baron, 1982) and the lead (s) may benefit from soft dollar commissions (Reuter, 2006; Nimalendran, Ritter, and Zhang, 2007; Goldstein, 2 We focus on the lead bookrunner(s) as they have the largest economic stake in the IPO and likely face the largest reputational consequences of positive or negative IPO outcomes. 3 Ellis, Michaely, and O'Hara (2000) show how the lead bookrunner earns considerable on average from post-ipo price stabilization and market making. 3

4 Irvine, and Puckett, 2011). 4 Our first hypothesis explaining the increased prevalence of IPOs featuring multiple lead and non-lead bookrunners takes the perspective that the choice to add bookrunners is primarily driven by the issuer. Our issuer bargaining power hypothesis (H1) builds on Hu and Ritter (2007) who argue that having multiple bookrunners increases the issuer s leverage in IPO negotiations. The presence of multiple bookrunners increases competition and can lead to lower direct costs of going public. A larger bookrunning team can lead to better monitoring, greater effort, more information production, and higher pricing with lower first-day returns. Multiple bookrunners should also lead to greater analyst coverage, as bookrunning firms typically feel compelled to initiate equity coverage on firms they take public. Given the increased use of multiple bookrunners, the issuer bargaining power hypothesis predicts that IPO outcomes have improved over time. What could be the source of this growing issuer bargaining power? We conjecture that issuers of larger IPOs have increased bargaining power, which may explain pressure on the 7% underwriting spread (Chen and Ritter, 2000). Conversations with bankers highlighted the emergence of IPO consultants (e.g. Solebury Capital) staffed by independent investment bankers who advise issuers on how to structure the offering in ways that benefit the issuer at the expense of the underwriters. While growing issuer bargaining power may reduce underwriter profitability for any given IPO, lead underwriters may still agree to participate if accepting additional non-lead bookrunners is combined with a greater share allocation to the leads and/or an increased probability of wining future equity mandates (i.e. they sacrifice economics today to win more deals in the future). Our alternative hypothesis explaining the use of multiple lead and non-lead bookrunners takes 4 While this description of lead bookrunner preference relies on agency arguments, Benveniste, Busaba and Wilhelm (2002) argue that incentives for underwriters and issuers can be aligned. 4

5 the perspective that the choice to add bookrunners is primarily driven by the lead bookrunner(s). Our underwriter bargaining power hypothesis (H2) builds on Barzel, Habib and Johnsen (2006) whose model predicts first-day returns can be increased by adding symmetrically ignorant bookrunners to a syndicate (analogous to non-lead bookrunners in our terminology). By creating and controlling the IPO syndicate, the lead bookrunner can crowd out competition and maintain greater control of information, leading to a lower initial filing price range, higher direct issuer costs (justified based on the larger size of the syndicate) and higher IPO first-day returns. 5 Given the increasing use of multiple bookrunners over time, the underwriter bargaining power hypothesis predicts that the economic rents to lead bookrunners should be increasing over time with worse IPO outcomes for issuers. What could be the source of growing underwriter power? We conjecture that by adding bookrunners with larger share allocations, the size of the underwriting syndicate should decline over time, leading to a greater concentration among the survivors. This growing market concentration should tilt bargaining power towards the underwriters and increase the costs for issuers. We test these alternative hypotheses by examining how the number and structure of bookrunners affects IPO outcomes. We focus on four metrics: (i) price adjustments from the average price of the first filing range to the offering price; (ii) first-day returns; (iii) analyst coverage; and (iv) underwriting spreads. We report univariate and multivariate comparisons across samples of IPOs with different bookrunner structures, namely: (i) all IPOs; (ii) IPOs with a sole lead bookrunner, and (iii) IPOs with multiple lead bookrunners. To mitigate endogeneity 5 Higher first-day returns also can arise based on Prospect Theory (see Loughran and Ritter, 2002; Ljungqvist and Wilhelm, 2005). If less competition leads to a lower initial filing price range, the bookrunner need not adjust prices up given good news during bookbuilding to keep the issuer satisfied, leading to higher first-day returns. 5

6 concerns, we also report comparisons of treated vs. control IPOs created using propensity score matching. Overall, we find evidence consistent with both the issuer bargaining power and the underwriter bargaining power hypotheses. First, IPO outcomes for issuers are significantly better when adding additional lead bookrunners, or when adding non-lead bookrunners to syndicates with two or more leads. Comparing IPOs featuring multiple lead bookrunners to sole lead IPOs, price adjustments are 3.1% lower, first-day returns are 2.6% lower, and the underwriting spreads are 13.5 basis points lower (all differences are significant at the 5% level except price adjustments). For IPOs featuring two or more lead bookrunners, adding non-lead bookrunners is associated with 6.0% lower price adjustments, 4.5% lower first-day returns and 25.6 basis points lower underwriting spreads (with differences significant at the 5% level except for underwriting spreads). Adding multiple nonleads is also associated with 1.5 more equity analysts following the issuer post-ipo. These differences are large and economically significant. For example, our sample features 510 IPOs with multiple lead bookrunners and no non-leads, which raised $171 billion in proceeds (in 2010 dollars). The reduction in first-day returns for multiple-lead IPOs relative to sole-lead translates to $4.5 billion less money left on the table (31% of the actual level). Our sample includes 171 IPOs with multiple lead bookrunners featuring one or more non-leads that raised $93.7 billion in proceeds (in 2010 dollars). The reduction in first-day returns for IPOs with non-leads relative to IPOs without non-leads represents $4.2 billion less money left on the table (56% of the actual level). Second, IPO outcomes for issuers are worse and lead underwriters are better off when adding non-leads to a sole-lead IPO. In this case, price adjustments are 4.7% higher, first-day returns are 9.7% higher, and underwriting spreads are 5.7 basis points higher (all differences are significant 6

7 at the 5% level except spreads). These differences are economically significant. Our sample features 410 IPOs with a sole lead bookrunner plus non-leads that raised $101.7 billion (in 2010 dollars). The increase in first-day returns relative to IPOs without non-leads translates to an additional $9.9 billion left on the table (56% of the actual level). Finally, given the increasing use of syndicates with multiple lead and non-lead bookrunners, we explore the implication of IPO size on bargaining power. From 2002 to 2008 there were 70 IPOs raising $400 million or more (in constant 2010 dollars). For these larger IPOs, approximately 10.1% of the proceeds was raised by syndicates having multiple lead bookrunners with one or more non-leads. From 2009 to 2014 the number of IPOs in this size category grew to over 60% of the capital raised. This finding is consistent with the view that issuers of larger IPOs have been using their bargaining power to push for bookrunner structures that are associated with better IPO outcomes. Overall, our evidence provides a balanced picture of the forces affecting bookrunner choice and structure. Issuers are using their bargaining power to push for syndicates with multiple lead bookrunners and non-leads as this leads to better IPO outcomes. Underwriters (particularly the more reputable ones) have used their bargaining power to increase their share allocation via syndicates with a sole lead bookrunner and multiple non-leads, which allows them to capture better economics. Both sides appear to be gaining something from the rise of multiple bookrunners. In addition to contributing to our understanding of multiple bookrunners, our paper contributes to the debate about the secular decline in the number of U.S. IPOs since the 1990s. Researchers find mixed evidence on the cause of the decline, which may be due to increasing globalization (Doidge, Karolyi and Stulz, 2013), the introduction of new regulations (Dambra, Field, and Gustafson, 2015), the benefits of a strategic sale for speeding a product to market (Gao, Ritter, and 7

8 Zhu, 2013), or perhaps the shock of the global financial crisis. 6 We find that the changing structure of underwriting syndicates has contributed to these trends. While the number of large IPOs has grown, the number of mid-size and small IPOs has stagnated. We believe this stagnation has been partly driven by the higher cost of going public given the rise of sole lead bookrunners with multiple non-leads, particularly for mid-sized IPOs. This finding suggests that solutions to the declining number of IPOs by smaller issuers should consider alternative syndicate structures that increase issuer bargaining power. Our paper also contributes to studies of the breakdown of the 7% solution documented by Chen and Ritter (2000). The number of U.S. IPOs paying an underwriting spread below 7% has increased, from below 20% of IPOs in the 1990s, to 30% in the 2000s, and close to 40% since 2010 (Lowry, Michaely and Volkova, 2017). We find that most of the decline has arisen in that segment of the IPO market where issuers appear to have growing bargaining power, using multiple lead bookrunners with some non-leads. 2. Data We collect data on all U.S. firm-commitment IPOs over the period January 1, 2002 and December 31, 2014 from the Thomson Financial Securities Data Company (TFSDC) New Issues database. Consistent with prior IPO studies, we exclude closed-end funds, real estate investment trusts (REITs), limited partnerships, unit investment trusts, unit offerings (combinations of equity and warrants), IPOs by banks 7, American depositary receipts and shares, global depositary receipts and shares, IPOs with offer prices below $5, and IPOs not listing on the American Stock Exchange, Nasdaq, or NYSE. Offerings are excluded if aftermarket trading data is not available on the Center 6 In terms of regulations, studies mention the 2000 Regulation Fair Disclosure, the 2002 Sarbanes-Oxley Act, the 2003 Global Settlement against investment banks, and the 2010 Jumpstart Our Business Startups (JOBS) Act. 7 We exclude IPOs having Standard Industrial Classification (SIC) codes 6000, 6021, 6022, 6029, 6035, 6036 or

9 for Research in Security Prices (CRSP) database. Our final sample covers 1,394 U.S. IPOs from 2002 to We collect offering terms and underwriter roles from the TFSDC New Issues Database and confirm by reading prospectuses on the SEC s Electronic Data Gathering, Analysis, and Retrieval (EDGAR) database. Data for some variables are taken from CRSP (e.g., shares outstanding, share prices after IPO) and Compustat (e.g., pre-ipo revenue). We illustrate the changing composition and roles in underwriting syndicates using the May 2012 IPO of Facebook Inc. As seen in Figure 1, Facebook s IPO raised gross proceeds of $16.0 billion, making it one of the largest U.S. IPOs ever. Facebook paid $176.1 million in underwriting fees, representing an underwriting spread of only 1.1%. Thomson Financial SDC Platinum classifies nine underwriters as bookrunners, with Morgan Stanley in the top left position, followed by J.P. Morgan and Goldman Sachs. Among the bookrunners was the boutique advisory firm Allen & Co., which had no equity trading floor, no equity salesforce, and no research analysts. [Insert Figure 1 here] The full underwriting syndicate for Facebook s IPO, shown in Figure 2, consisted of 33 financial institutions. The share allocations among the bookrunners was very unequal, with Morgan Stanley allocated 38.5% of the shares while J.P. Morgan received 20.1% and Goldman Sachs 15.0%. We estimate that Morgan Stanley received $65.4 million, or 37% of all fees, not including additional fees earned from the 15% overallotment option or profits from price stabilization post-ipo. 8 For league table purposes, each of the nine bookrunners received equal credit of $1.778 billion, or 1/9 th of the gross proceeds. On this basis, Allen & Co ranked as the 11 th largest underwriter of U.S. IPOs in 2012 a ranking that does not reflect either their risk exposure or contribution to the distribution of Facebook shares. It was the only IPO for which Allen & Co. 8 Thomson Financial SDC Platinum reports that the management fee was 20% of the underwriting spread, the selling concession was 60% and the underwriting fee was 20%. 9

10 was a bookrunner that year. Based on the actual share allocation of 8.4 million shares, Allen & Co. was paid an estimated $3.4 million. [Insert Figure 2 here] Table 1 presents data on the number of IPOs by year and the average number of bookrunners. As shown in Hu and Ritter (2007) and Lowry, Michaely and Volkova (2017), the average number of bookrunners per IPO has been increasing over time, growing from 1.34 in 2002 to 2.00 in 2008 to 3.28 in The average allocation of shares to all bookrunners has also increased, growing from 51.6% in 2002 to 70.7% in 2008 to 81.0% in [Insert Table 1 here] We classify bookrunners as lead or non-lead based on share allocations relative to the top allocation from the underwriting table in the prospectus. When IPOs have multiple bookrunners, we classify bookrunners that receive at least 90% of the maximum allocation of shares as lead bookrunners. Bookrunners receiving an allocation less than 90% of the top allocation are classified as non-lead bookrunners. 9 While this classification may incorrectly classify some underwriters based on their actual role in the IPO, we expect that non-lead bookrunners who receive lower share allocations play a smaller role than the lead bookrunner(s). To the extent our classification system is noisy, our analysis loses power to find evidence of economic effects of bookrunner structure on IPO outcomes. Table 1 reports the average number of lead and non-lead bookrunners per IPO over time. The average number of lead bookrunners has remained relatively constant, growing only slightly from 1.28 in 2002 to 1.60 in 2008 to 1.48 in The percentage share allocation to lead bookrunners has also remained constant over time between 50% and 60%. The fraction of IPOs with non-lead 9 The 90% cutoff is arbitrary. We considered alternative cutoffs between 75 and 95% and the findings in our study are not materially impacted. 10

11 bookrunners has increased from 4.7% in 2002 to 35.0% in 2008 to 75.5% in The average number of non-lead bookrunners has grown from 0.06 in 2002 to 0.40 in 2008 to 1.80 in These trends represent a significant change in the composition of underwriting syndicates and the roles of different bookrunners that have not been shown in other studies. While previous research has sought to address how issuers select bookrunners, our objective is to understand (i) the issuer s choice of multiple active lead bookrunners, and (ii) why passive non-lead bookrunners are added. We acknowledge that there could be idiosyncratic reasons for such choices, such as past social relationships between issuers and underwriters (Cooney, Madureira, Singh and Yang, 2015). Our analysis attempts to identify systematic factors affecting these choices as well as their economic consequences on IPO outcomes. 3. Impact of Bookrunner Structure on IPO Outcomes We begin our empirical analysis by examining the impact of IPO syndicate structure on various IPO outcomes using univariate statistics, before providing a multivariate analysis. Given the theoretical predictions of our two hypotheses, we focus on five measures of IPO outcomes: price adjustment, underwriting spread, first-day return, number of equity analysts, and share allocation to lead underwriter(s). Price Adjustment is defined as one hundred times the difference between the IPO offer price and the average of the high and low prices of the first price range reported in the preliminary IPO prospectus (the S-1 or red herring ), divided by this average filing price. To examine the direct cost of going public, we consider the Underwriting Spread, defined as the dollar value of underwriting fees paid as a percentage of the offering size. To examine the indirect cost of going public, we consider the IPO First-day Return, defined as one hundred times the difference between the first trading day closing price and the IPO offer price, divided by the offer price. We also consider the Number of Analysts, defined as the number of unique equity 11

12 analysts on I/B/E/S reporting earnings estimates or making stock recommendations within three months of the IPO issue date. Finally, to consider the impact of deal structure on underwriters, we examine the Lead Bookrunner % Share Allocation, defined as the number of shares allocated to a lead bookrunner (or the average if multiple lead bookrunners), divided by the number total shares to be offered, excluding overallotments. i. Univariate Statistics Table 2 reports descriptive statistics (sample means) for the five IPO outcome variables for three samples based on bookrunner structure. We also report sample means for a wide range of variables measuring underwriter characteristics, issue and issuer characteristics, and market conditions that are used as controls in our multivariate tests. Appendix A has definitions of these variables. Underwriter characteristics are reputation (as measured by the Carter-Manaster ranking), market share of IPOs over the last twelve months ( LTM ) based on equal credit to each bookrunner (Underwriter market share LTM, equal credit to bookrunner) and based on the actual share allocation (Underwriter market share LTM, actual share allocation), and analyst reputation (as measured by the Number of Institutional Investor Industry All-Star Analysts). 10 All measures are defined for lead bookrunners only. In cases with multiple leads we average these measures across all lead bookrunners. We include the following issue and issuer characteristics: IPO size (Log of issue size), issuer size (Log of revenues from the most recent fiscal year), issuer age (Log of firm age), the ratio of total shares outstanding to total post offering float (Share Overhang), and dummy variables set to 1 if the IPO is backed by a private equity firm (Private Equity Backed) or a venture capital firm (Venture Capital Backed). Since Boeh and Dunbar (2014) report an 10 We use a five industry classification scheme based on the TFSDC Macro Industry Codes (TF Macro Code): hightech (HT), health (HEALTH), finance (FINANCE), energy (ENERGY), and other (all other codes). As noted by Boeh and Dunbar (2016), the TFSDC scheme has the advantage that it is closely aligned with the industry corporate finance groups at many major underwriters. 12

13 association between IPO pricing (measured using price adjustments and first-day returns) and market conditions over the filing period, we include four measures of market conditions. Ince (2014) and Boeh and Dunbar (2016) find asymmetric effects of market returns on IPO pricing, with negative returns having more significant effects than positive returns. We therefore include two measures: Market return from filing to issue if positive; and Market return from filing to issue if negative. We include other variables that reflect the state of the IPO market: Average IPO underpricing during the filing period, Number of IPOs pre offer. [Insert Table 2 here] We begin by examining outcomes for the full sample of IPOs. The first pair of columns in Table 2 compare IPOs with a sole lead bookrunner vs. IPOs with multiple lead bookrunners. We test for differences in sample means using a parametric t-test. IPOs with multiple lead bookrunners feature a statistically higher number of analysts and a statistically lower underwriting spread, lower first-day return, and lower share allocation to each lead bookrunner. These results support the issuer bargaining power hypothesis (H1). In the second pair of columns, we examine the subsample of sole-lead IPOs and compare IPOs with no non-lead bookrunners vs. IPOs with one or more non-leads. Deals with non-leads exhibit a statistically lower underwriting spread and a lower average allocation to the lead bookrunner. But the price adjustment, first-day returns and number of analysts are significantly greater for deals with one or more non-lead bookrunners. This evidence is more consistent with underwriter bargaining power hypothesis (H2). The final pair of columns focus on IPOs with multiple lead bookrunners, and compare IPOs with no non-lead bookrunners vs. IPOs with one or more non-leads. Having non-lead bookrunners increases the number of analysts, lowers the underwriting spread and lowers the share allocation to the lead bookrunner(s). These findings support the issuer bargaining power hypothesis (H1). 13

14 While highly significant, these differences in IPO outcomes based on bookrunner structure are difficult to interpret given significant differences for most control variables across samples. Table 2 shows that the average IPOs with more lead or non-lead bookrunners tend to be taken public by more reputable underwriters (higher Carter-Manaster Ranking, higher LTM market share) who employ more all-star analysts. On average, these IPOs are larger in size and issued by bigger, older firms who have more backing by private equity firms and less by venture capital firms. The market conditions do not vary significantly based on bookrunner structures. ii. Impact of adding lead bookrunners The multivariate regressions in Table 3 consider the impact of adding multiple lead bookrunners on IPO outcomes. The dependent variables are the five IPO outcome variables defined previously. To give the results more economic meaning we report marginal effects where the estimated regression coefficient for the continuous variables represents the impact of a one standard deviation change in the independent variable on the dependent variable. The main variable of interest is the Multiple lead bookrunner dummy, which captures the impact of bookrunner structure on IPO outcomes. This dummy variable is set to 1 if an IPO has more than one lead bookrunner, and zero otherwise. It tests whether adding multiple lead bookrunners affects IPO outcomes, controlling for other factors that vary across IPOs. Each of the regressions also control for underwriter characteristics, issue and issuer characteristics, market conditions, and year and industry fixed effects. In the regressions explaining first-day returns, we also consider the impact of price adjustments. Following Edelen and Kadlec (2005) we first estimate a model of price adjustments then we calculate Abnormal price adjustment as the actual price adjustment in the IPO minus the predicted price adjustment based on the regression model. Abnormal price adjustment if positive takes the calculated value 14

15 if positive, and zero otherwise. This specification recognizes that price adjustments have been shown to have an asymmetric effects on first-day returns. [Insert Table 3 here] The bookrunner dummy variable indicates a significant positive impact of adding more lead bookrunners on IPO outcomes. Consistent with the issuer bargaining power hypothesis (H1), IPOs with multiple lead bookrunners have significantly lower price adjustments (-3.4%) and lower first-day returns (-2.5%) with each lead bookrunner receiving a lower share allocation (-10.6%). The underwriting spread is not significantly lower, nor is the number of analysts significantly different from sole-lead IPOs, controlling for other factors. These resuls suggest that deal economics are worse for a sole lead bookrunner when adding other leads. The estimated coefficients on the controls in Table 3 are consistent with the evidence found in the literature. IPOs led by more reputable underwriters and by older firms feature lower price adjustments, while larger IPOs and IPOs with greater share overhang tend to have more price adjustments. Market returns significantly affect price adjustments but only when market returns are negative. More reputable underwriters and IPOs backed by private equity firms are associated with a higher underwriting spread, whereas larger IPOs and IPOs with greater share overhang tend to pay a lower spread. First-day returns are higher for larger IPOs, deals with greater share overhang, and IPOs backed by venture capital firms. Abnormal price adjustments (particularly if positive) have very significant positive effects on first-day returns. Analyst coverage post-ipo is greater for larger IPOs, bigger issuers, and IPOs with more venture capital backing, but lower for older issuers on average. Finally, the share allocation to the lead bookrunner is lower for IPOs with more reputable underwriters and larger IPOs, where there are more lead bookrunners and greater competition. 15

16 iii. Impact of adding non-lead bookrunners Table 4 reports estimates from models that use different dummy variables to capture the impact of adding non-lead bookrunners to IPOs with either a sole lead or multiple lead bookrunners. To simplify the presentation, we only report the regression estimates for the bookrunner structure dummy variables. Again we report marginal effects where the estimated regression coefficient for the continuous variables represents the impact of a one standard deviation change in the independent variable on the dependent variable. Panel A examines all 1,394 IPOs where the base case are IPOs with multiple lead bookrunners and multiple non-leads. We add three dummy variables set to 1 to identify IPOs with a Sole-lead, no non-leads, Sole-lead, multiple non-leads, and Multiple leads, no non-leads, and zero otherwise. These dummy variables capture differences relative to the base case. All three dummy variables are positive, and 12 out of 15 estimated coefficients are statistically significant at the 10% level or greater. Not surprisingly, traditional IPOs with a sole lead bookrunner and no non-leads feature the highest price adjustments, the highest first-day returns, and the largest share allocation to the lead bookrunner. IPOs with multiple lead bookrunners and one or more non-lead bookrunners (i.e. the base case) have the lowest price adjustments, lower underwriting spreads, the lowest first-day returns, and the smallest average share allocations to the lead bookrunners. In terms of the number of analysts, all three dummy variables are negative, but only the Multiple leads, no non-leads dummy is negative and significant, indicating that adding non-leads results in more analyst following. In general, this evidence mostly supports the issuer bargaining power hypothesis (H1). [Insert Table 4 here] Panel B of Table 4 considers 884 IPOs with a sole lead bookrunner where the base case are IPOs with no non-leads. The dummy variable Sole lead, multiple non-leads captures the 16

17 impact of adding one or more non-leads. This estimate allows comparison to the univariate results from the second pair of columns in Table 2. Adding non-lead bookrunners does not significantly affect price adjustments, underwriting spreads, first-day returns, and the number of analysts, but the share allocations to the sole lead bookrunner are statistically lower by -10.4%. Panel C considers 510 IPOs with multiple lead bookrunners where the base case are IPOs with no non-leads. The dummy variable Multiple lead, multiple non-leads captures the effect of adding one or more non-leads. This estimate allows comparison to the univariate results from the third pair of columns in Table 2. Adding non-lead bookrunners is associated with statistically lower price adjustments, lower first-day returns, and lower share allocations to the lead bookrunners with no effect on the underwriting spread or number of analysts. iv. Predicting Bookrunner Structure Overall, the evidence in Tables 3 and 4 supports the predictions of the issuer bargaining power hypothesis (H1) but the causality is difficult to interpret due to endogeneity problems. We find significant differences in IPO outcomes based on bookrunner structure. This structure, however, is endogenous and arises from choices made by both issuers and underwriters. The factors affecting underwriter choice have been found to correlate with IPO outcomes. To mitigate the effect of this unobserved endogeneity, we use propensity score matching. This methodology controls for selection based on observable characteristics that we believe may be correlated with both bookrunner choices and IPO outcomes. Our IPO sample is well suited to this approach given the large number of issuers making different choices regarding the choice of lead and non-lead bookrunners. We use nearest neighbor caliper-based matching with replacement (Caliendo and Kopeinig, 2008). The matching begins with probit models reflecting the choice of bookrunner structure. Predicted probabilities or propensity scores from these probit models are then used 17

18 to create matches. We match a given treated IPO with non treated IPOs having propensity scores within +/- 10%. We limit the possible set of matched IPOs to those within the same common support. 11 While we report results for one specific set of probit models, our findings are robust to alternative probit model specifications (e.g. we consider models that drop insignificant variables and also consider models with variable interactions). We present the models with the best matching properties. 12 We also check the robustness of our results to different propensity score approaches including nearest neighbor (with and without replacement) and kernel-based matching. Our results are not significantly different using these alternative approaches. Table 5 reports the estimates from the probit models predicting bookrunner structure that are used to generate propensity scores for matching. The first probit estimates the probability of adding multiple lead bookrunners, while the second and third probits estimates the probability of adding non-leads to sole-lead and multiple-lead IPOs, respectively. We report marginal effects that capture the impact of a one standard deviation change in the independent variable on the probability of a bookrunner structure. [Insert Table 5 here] The first model in Table 5 includes all 1,394 IPOs and predicts the probability of multiple lead bookrunners. The dependent variable is a dummy set to one if the IPO has multiple lead 11 This screen eliminates potential matching of IPOs with propensity scores above (below) the maximum (minimum) score for the sample of treated firms. We also eliminate treated firms with propensity scores above (below) the maximum (minimum) score for the sample of non-treated firms. 12 We test the appropriateness of the probit models through an examination of the balancing properties of the independent variables. For different ranges of propensity scores, we use an F-test to check whether there are significant differences between the means of independent variables in treated and matched observations in that range. For the models we report, we find no significant differences indicating the treated and matched sample are balanced (i.e. based on these independent variables, firms in the treated and control groups look very similar). This balancing enhances our confidence that differences in outcomes are due to the treatment (bookrunner structure selection) and not due to other IPO differences. 18

19 bookrunners and zero otherwise. The independent variables are the same as Tables 3 and 4, namely underwriter characteristics, issue and issuer characteristic, and market conditions. While significant at the 10% level, Carter-Manaster ranking positively affects the likelihood of using multiple lead bookrunners. Larger IPOs and private equity backed IPOs also tend to use multiple bookrunners. The direction of the coefficients on the market return variables indicate that multiple lead bookrunners are associated with less extreme market movements, both smaller for positive and negative market returns from filing to pricing. Finally, the likelihood of using multiple lead bookrunners is lower when a higher fraction of IPOs over the past 12 months used multiple lead bookrunners, perhaps due to underwriter capacity constraints. The negative sign indicates mean reversion in the time series (e.g. the likelihood of using a multiple bookrunner decreases after an active period). The second model in Table 5 considers the 884 IPOs having a sole lead bookrunner, and predicts the probability of adding non-leads. The dependent variable is a dummy set to one if the IPO has non-lead bookrunners and zero otherwise. Carter-Manaster ranking, Log of issue size and Private Equity backed significantly positively affect the likelihood of non-lead bookrunner use. Unlike the first model, market returns do not have significant effects. Like the first model, the significantly negative sign on Fraction of IPOs with non-lead bookrunners if sole lead LTM indicates mean reversion in the time series. Finally, the third model of Table 5 considers the 510 IPOs having multiple lead bookrunners. The dependent variable equals one if the IPO has non-lead bookrunners and zero otherwise. IPO size, firm size and a larger share overhang significantly affect the likelihood of using non-lead bookrunners in this sample. Like the previous models, the significantly negative 19

20 sign on Fraction of IPOs with non-lead bookrunners if multiple lead LTM indicates mean reversion in the time series. v. Impact of Bookrunner Structure on IPO outcomes using matched sample Table 6 reports differences in IPO outcomes for treated vs. control IPOs matched using the propensity scores generated in Table 5. The t-statistics are based on bootstrapped standard errors. Using the full sample of IPOs, the first pair of columns show that adding lead bookrunners reduces the underwriting spread by 13.5 basis points, and reduces first-day returns by -2.6%. Adding lead bookrunners has no significant effect on price adjustments and the number of analysts. Finally, adding lead bookrunners significantly reduces the lead bookrunner share allocation by 10.7%. Overall, this evidence is consistent with the predictions of the issuer bargaining power hypothesis (H1) and confirms the findings from Tables 2 through 4. [Insert Table 6 here] The second pair of columns in Table 6 examine how adding non-lead bookrunners to a sole-lead managed deal affects IPO outcomes. The results when controlling for endogeneity are very different from the evidence in Tables 3 and 4 and more consistent with the univariate evidence in Table 2. Adding non-leads results in significantly higher price adjustments (4.7%), and higher first-day returns (9.7%). The impact on the underwriting spread is positive but not statistically significant. The only finding that does not change relative to Tables 3 and 4 is the significantly negative effect of adding non-leads on the lead bookrunner s share allocation (-13.7%). Overall, these findings are more consistent with the underwriter bargaining power hypothesis (H2). The third pair of columns in Table 6 examine the impact of adding non-lead bookrunners for IPOs with multiple leads. Consistent with the evidence in Tables 2 to 4, adding non-leads significantly reduces price adjustments (-6.0%), first-day returns (-4.5%) and lead bookrunner allocations (-4.9%). The number of analysts covering is also significantly higher (+1.5 analysts). 20

21 These effects are economically important. The reduction in first-day returns, for example, is almost 56% of the total amount of money left on the table for this sample. This evidence is consistent with the issuer bargaining power hypothesis (H1). Overall, the findings from the propensity score analysis show that moving to multiple lead bookrunners and then adding non-leads results in better outcomes from the issuer s perspective (H1). When sole-lead bookrunners add non-leads, however, IPO outcomes are more favorable for the sole lead bookrunner and less favorable for the issuer (H2). 4. Impact of IPO size on Bargaining Power Thus far, the evidence suggests that moving to multiple bookrunners both lead and nonlead -- appears to be driven by issuer bargaining power and leads to better IPO outcomes. The decision to add non-leads when there is a sole lead appears to be driven by underwriter bargaining power, as this combination is more favorable to underwriters. Anecdotal evidence suggests that the bargaining power of the issuer and underwriters is affected by the size of the IPO. We explore this relationship in this section. Table 7 reports the cumulative IPO proceeds raised (in constant 2010 dollars) broken down by time period, IPO size and bookrunner structure. From 2002 to 2008 there were 787 IPOs, 70 (9%) of which were larger than $400 million, although those large IPOs raised 44% of the gross proceeds. Of the large IPOs, 62% of the capital raised was in syndicates with multiple lead bookrunners but only 10% of the capital raised was in syndicates with multiple lead bookrunners and non-leads. From 2009 to 2014 there were 607 IPOs, 86 (14%) of which were larger than $400 million. The large IPOs raised 63% of the gross proceeds in this period. Of the large IPOs, 66% of the capital raised was in syndicates with multiple lead bookrunners and 60% of the capital raised was in syndicates with multiple lead bookrunners and non-leads. While underwriter concentration 21

22 is growing, issuers of larger IPOs appear to have been able to increasingly use their bargaining power to push for syndicate structures that benefit them (H1). We suspect that part of the reason for this growing issuer power is the increasing importance of larger IPOs (in the later period far more capital is raised in this segment than in the earlier period). [Insert Table 7 here] For smaller IPOs, the patterns are quite different. For mid-sized IPOs (raising between $25 and $400 million), the use of multiple lead and non-lead bookrunners has increased (from 5% to 25% of the capital raised), but the overall capital raised using multiple bookrunners has declined, albeit slightly (from 42 to 41% of the capital raised). The capital raised in this size category using a sole lead and one or more non-leads has grown dramatically, from 18% of the capital raised from 2002 to 2008 to 53% from 2009 to It is within the segment of mid-sized IPOs that underwriters appear to have increasingly exploited their bargaining power. One consequence is that the market for smaller IPOs has declined in number and dollars raised. From 2002 to 2008, IPOs between $25 and $400 million raised 56% of the total capital in IPOs but the capital raised by that size category of deals has dropped to only 37% more recently. [ADD ANALYSIS OF IPO OUTCOMES FOR DIFFERENT SIZE IPOs] 5. Conclusion This study revisits the composition and dynamics of underwriting syndicates in U.S. initial public offerings (IPOs) and the phenomenon of multiple bookrunners. We note the emergence of non-lead bookrunners over the past decade, particularly post We introduce two hypotheses to explain the emergence of multiple lead and non-lead bookrunners in IPOs: issuer bargaining power, and underwriter bargaining power. 22

23 We find evidence consistent with both hypotheses. IPO outcomes are better from the issuer s perspective when additional lead bookrunners are added to sole-lead IPOs or when nonlead boorkunners are added to multiple lead IPOs. In particular, adding bookrunners is associated with lower price adjustments, lower first-day returns, lower underwriting spreads, and greater analyst following. Outcomes are better from the underwriter s perspective, however, when a sole lead bookrunner adds non-leads. These IPOs feature higher price adjustment and higher first-day returns with no decline in underwriting spreads. Finally, we find multiple lead bookrunner structure to be most prevalent among larger offerings, consistent with issuer bargaining power. This evidence may help to explain the growth in number of large IPOs, where issuers have more power, and the decline of small and midsize IPOs where issuer bargaining power is lower. Larger issuers may be using their increased bargaining power to push for a bookrunner structure that results in lower issue cost whereas in the latter category, underwriters are using their greater bargaining power to push for a bookrunner structure that increases issue costs. While we believe our research provides a more complete explanation of IPO syndicate dynamics by considering both issuer and underwriter bargaining power, a number of questions remain. How does syndicate status on an IPO affect underwriter selection for follow-on equity offerings? Since so many IPO firms are acquired, does IPO syndicate status affect an underwriter s ability to win M&A advisory business? We find that sole lead bookrunners may be taking advantage of less sophisticated issuers to drive outcomes more favorable to underwriters. This sub-sample should be explored to see if the issuer sophistication story is supported by the evidence. Finally, while we consider the effects from the perspective of issuers and underwriters, we do not consider the impacts for investors. We leave these topics to future research. 23

24 References Baron, D., A model of the demand for investment banking advising and distribution services for new issues. Journal of Finance 37, Barzel, Y, Habib, M, Johnsen, D.B., 2006, Prevention is Better than Cure: The Role of IPO Syndicates in Precluding Information Acquisition, Journal of Business, 79, Benveniste, L., Busaba, W., Wilhelm, W., Information externalities and the role of underwriters in primary equity markets. Journal of Financial Intermediation 11, Bolton, P., Faure-Grimaud, A., Thinking Ahead: The Decision Problem. Review of Economic Studies 76, Boeh, K., Dunbar, C., IPO waves and the issuance process. Journal of Corporate Finance 25, Boeh, K., Dunbar, C., Underwriter deal pipeline and the pricing of IPOs. Journal of Financial Economics 120, Caliendo, M., Kopeinig, S., 2008, Some Practical Guidance for the Implementation of Propensity Score Matching, Journal of Economic Surveys, 22, Carter, R., Manaster, S., Initial public offerings and underwriter reputation. Journal of Finance 45, Chen, H-C., Ritter, J.R., The Seven Percent Solution. Journal of Finance 55, Cooney, J.W., Madureira, L., Singh A.J., and Yang, K., Social Ties and IPO Outcomes. Journal of Corporate Finance 33, Corwin, S., Schultz, P., The role of IPO underwriting syndicates: pricing, information production, and underwriter competition. Journal of Finance 60, Dambra, M., Field, L., Gustafson, M., The JOBS Act and IPO volume: evidence that disclosure costs affect the IPO decision. Journal of Financial Economics 116, Doidge, C., Karolyi, A., and Stulz, R The U.S. left behind? Financial globalization and the rise of IPOs outside the U.S. Journal of Financial Economics 110, Dong, M., Michel J-S., and Pandes, A., Underwriter Quality and Long-Run IPO Performance. Financial Management 40, Dunbar, C., Factors affecting investment bank initial public offering market share. Journal of Financial Economics 55, Edelen, R., Kadlec, G., Issuer surplus and the partial adjustment of IPO prices to public information. Journal of Financial Economics 77, Ellis, K., Michaely, R., and O'Hara, M., When the Underwriter Is the Market Maker: An Examination of Trading in the IPO Aftermarket. Journal of Finance 55, Gao, X., Ritter, J., Zhu, Z., Where have all the IPOs gone? Journal of Financial and Quantitative Analysis 48, Goldstein, M., Irvine, P., Puckett, A., Purchasing IPOs with commissions. Journal of Financial and Quantitative Analysis 46,

25 Hu, W., and Ritter, J., Multiple Bookrunners in IPOs. University of Florida, working paper. Ince, O., Why do IPO offer prices only partially adjust? Quarterly Journal of Finance 4, Ljungqvist, A., Marston F., and Wilhelm, W., Scaling the Hierarchy: How and Why Investment Banks Compete for Syndicate Co-management Appointments. Review of Financial Studies 22, Ljungqvist, A., Wilhelm, W., Does Prospect Theory Explain IPO Market Behavior? Journal of Finance, 60, Loughran, T., Ritter, J., Why Don t Issuers Get Upset About Leaving Money on the Table in IPOs? Review of Financial Studies, 15, Lowry, M., Michaely, R., and Volkova, E., Initial Public Offerings: A Synthesis of the Literature and Directions for Future Research. Available at SSRN: Merton, R.C., A Simple Model of Capital Market Equilibrium with Incomplete Information. Journal of Finance 42, Nimalendran, M., Ritter, J., Zhang, D., Do today s trades affect tomorrow s IPO allocation? Journal of Financial Economics 84, Popescu, M. and Xu, Z., Co-managers, Information, and the Secondary Market Liquidity of Initial Public Offerings. Financial Management 40, Reuter, J., Are IPO allocations for sale? Evidence from mutual funds. Journal of Finance 61,

26 Appendix A Independent Variable Definitions Variable Definition Underwriter Characteristics Carter- Manaster ranking Ranking for each underwriter (from 0 to 9.001) at the time of the IPO is obtained from Carter and Manaster (1990) as updated by Loughran and Ritter (2004), noting that ranking for some underwriters change over the sample period. For analyses conducted at the deal level (i.e. in Tables 2, 3 and 5), this variable is defined as the average ranking among lead bookrunners. Source: Jay Ritter s website for Carter-Manaster rankings [see link to IPO underwriter reputation rankings ( ) at Thomson Financial SDC (TFSDC) New Issues database is used to identify underwriters and their roles and this is confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. Underwriter market share LTM, equal credit to each bookrunner Underwriter market share LTM, actual allocations Number of Institutional Investor Industry All- Star Analysts This variable is defined similarly to the above except proceeds allocated to underwriters are based on the rule that bookrunners share the proceeds on an equal basis. Sources: Same as for "Underwriter market share, actual allocations" CPI adjusted proceeds of all IPOs over the 12 months prior to the IPO issue date managed by the underwriter based on actual deal level allocations divided by the total dollars raised (CPI adjusted proceeds) over that period. CPI adjusted proceeds are defined as the gross proceeds on an IPO multiplied by the consumer price index as of the end of the month before the offering date using the CPI series that is scaled to equal one in January 2010 ( If an underwriter merged during the 12 month period, the value of IPOs managed by pre-merger underwriters is also included in the numerator. For analyses conducted at the deal level (i.e. in Tables 2, 3, and 5), this variable is defined as the average market share among lead bookrunners. Sources: IPO proceeds are from TFSDC New Issues database confirmed through searches of IPO prospectuses (on EDGAR). Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. For each underwriter we count the number of Institutional Investor All-Star Analysts (ranked one to three) in the same industry as the IPO issuing firm in the most recent ranking of analysts. If an underwriter merged during the 12 month period, analysts at pre-merger underwriters are included in this count. The rankings are assumed to be released on October 1 each year so, for example, IPOs between October and September 30, 2004 use the 2003 ranking. IPOs are classified into five industry categories using TFSDC Macro Industry Code (TF Macro Code): high-tech (HT), health (HEALTH), finance (FINANCE), energy (ENERGY), and other (all other codes). Institutional investor industries are then assigned to one of these five classification (the industry matching table is provided on an on-line Appendix). For analyses conducted at the deal level (i.e. in Tables 2, 3, and 5), this variable is defined as the average ranking among lead bookrunners. Sources: Institutional Investor analyst ranking data. Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. 26

27 Appendix A (Continued) Variable Definition Underwriter Characteristics, Continued Underwriter headquarters in same state as issuer Underwriter headquarters in same state as issuer & lead bookrunner headquarters not in state Underwriter headquarter location (US state) is first identified from TFSDC New Issues Database, and confirmed through website searches. Each IPO issuer headquarter is also identified from the TFSDC New Issues Database and confirmed through prospectus searches. This variable take the value one if the underwriter's headquarter is in the same state as the IPO issuer, and zero otherwise. Sources: Underwriters and issuer headquarters are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. Underwriter headquarter (HQ) location (US state) is first identified from TFSDC New Issues Database, and confirmed through website searches. Each IPO issuer headquarter is also identified from the TFSDC New Issues Database and confirmed through prospectus searches. This variable take the value one if the underwriter's headquarter is in the same state as the IPO issuer and the headquarter of none of the lead bookrunners on the IPO are the same as the issuer, and zero otherwise. Sources: Underwriters and issuer headquarters are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. Underwriter participation in past lead bookrunner deals Underwriter reciprocal participation in deals This variable is defined for underwriters who are not lead bookrunners on an IPO. We begin by counting, for each lead bookrunner in an IPO, the number of IPOs over the past 12 months where it was a lead bookrunner. If a lead bookrunner merged during the 12 month period, IPOs managed by pre-merger organizations are included in this count. For each lead bookrunner having more than three such deals we compute the fraction of those deals where the non-lead book running underwriter is a non-lead bookrunner (if the underwriter merged during the 12 month period, IPOs managed by pre-merger organization are included in this count). This variable is the maximum of those fractions (equal to zero if no lead bookrunner had more than 3 past year deals). Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. This variable is defined for underwriters who are not lead bookrunners on an IPO. We begin by counting the number of IPOs over the past 12 months where the underwriter acted as a lead bookrunner (if the underwriter merged during the 12 month period, IPOs managed by pre-merger organization are included in this count). If this number is less than 4 this variable is set equal to zero. Otherwise we then count, for each lead bookrunner in the current IPO, the number of IPOs over the past 12 months where the underwriter was lead bookrunner and the current lead bookrunner was a non-lead bookrunner. If the current lead bookrunner merged during the 12 month period, IPOs managed by pre-merger organizations are included in this count. We then compute the fraction of those deals by the underwriter where it was lead bookrunner and the current lead bookrunner was a non-lead bookrunner. This variable is the maximum of those fractions. Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. 27

28 Appendix A (Continued) Variable Definition Issue and Issuer Characteristics Log of issue size Issue size equals the number of shares to be sold (primary and secondary), excluding overallotments, in millions multiplied by the offering price. This amount is then multiplied by the consumer price index as of the end of the month before the offering date (using the CPI series that is scaled to equal one in January 2010). This variable is the natural log of issue size. Source: Share and price information from the TFSDC New Issues Database and confirmed by examining IPO prospectuses (EDGAR). Inflation (CPI) data are from the Federal Reserve Bank of St. Louis website ( Log of revenues Log of firm age Share Overhang Private Equity backed We first identify the last 12 months revenue in millions for each IPO issuer (prior to offering date) multiplied by the US consumer price index as of the end of the month before the offering date (using the CPI series that is scaled to equal one in January 2010). To account for some issuers having zero sales, this variable is the natural logarithm of one plus CPI-adjusted revenue. Source: Compustat. When data are not available, sales are obtained from the initial IPO prospectus (EDGAR). Inflation (CPI) data are from the Federal Reserve Bank of St. Louis website ( Corporate founding date (year) is obtained from Jay Ritter's website. Firm age at the time of the IPO is the year of the IPO subtract the founding year. This variable is the logarithm of one plus firm age. Source: Jay Ritter's website ( Ratio of shares retained to shares offered at the time of IPO. Share retained equals shares pre-ipo (for issuers with multiple share classes, this is the sum of shares across all share classes) less secondary shares to be sold. Shares to be offered is the sum of primary and secondary shares. Source: Primary and secondary shares filed from TFSD New Issues database confirmed through search of IPO prospectuses (EDGAR). Number of shares pre-ipo is the number of shares post-ipo less primary shares in the IPO from the final prospectus (EDGAR). Number of shares post-ipo is from the Center for Research in Security Prices (CRSP) with adjustments to account for multiple share classes, from Jay Ritter s website (see link to A list of IPOs from 1980 April 2014 with multiple share classes outstanding at This variable is a dummy variable that equals one if the IPO had prior private equity backing, and zero otherwise. Source: TFSD New Issues database. Venture Capital backed Market Conditions Market return filing to issue if positive Market return filing to issue if negative Average IPO underpricing during filing period This variable is a dummy variable that equals one if the IPO had prior venture capital backing, and zero otherwise. Source: TFSD New Issues database. Cumulative return in percent between the IPO first filing date and the issue date for the CRSP value-weighted index (including dividends) if positive and zero otherwise. Source: CRSP variable VWRETD Cumulative return in percent between the IPO first filing date and the issue date for the CRSP value-weighted index (including dividends) if negative and zero otherwise. Source: CRSP variable VWRETD Underpricing is 100 times the difference between the first trading day opening price for the issuer, divided by the offering price, for each IPO. This variable is the average underpricing for all IPOs offered on dates between the IPO first filing date and the issue date. Source: IPO offering terms are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. The opening trading price for each IPO is obtained from CRSP. 28

29 Appendix A (Continued) Variable Definition Market Conditions, Continued Number of IPOs pre offer A count of the IPOs that are issued over the year prior to the issue date in the current IPO. We exclude closed end funds, real estate investment trusts (REITs), limited partnerships, unit investment trusts, unit offerings (combinations of equity and warrants), issues by banks, American depositary receipts, American depositary shares, global depositary receipts, or global depositary shares, issues with offer prices below $5 and issues not listing on the American Stock Exchange, NYSE or Nasdaq. Offerings are excluded if aftermarket trading data is not available on the Center for Research in Security Prices (CRSP). Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify IPOs (confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval, or EDGAR, system database). Bookrunner Structure Fraction of The number of IPOs over the year prior to the issue date in the current IPO having more than IPOs with one lead bookrunner divided by the total number of IPOs over that period. multiple lead Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify bookrunners underwriters and their roles and this is confirmed through prospectus searches on the SEC LTM Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (nonlead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. Fraction of IPOs with non-lead bookrunners if sole lead LTM Fraction of IPOs with non-lead bookrunners if multiple leads LTM Proportion of IPOs where underwriter is lead bookrunner The number of IPOs over the year prior to the issue date in the current IPO having one lead bookrunner and greater than zero non-lead bookrunners divided by the total number of IPOs having one lead bookrunner over that period. Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify underwriters and their roles and this is confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (nonlead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. The number of IPOs over the year prior to the issue date in the current IPO having more than one lead bookrunner and greater than zero non-lead bookrunners divided by the total number of IPOs having more than one lead bookrunner over that period. Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify underwriters and their roles and this is confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (nonlead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. If the total number of IPOs over the year prior to the current offering where the underwriter participated in some role is less than 4, this variable is set to 0. Otherwise, this variable is the total number of IPOs over this period where the underwriter is a lead bookrunner divided by the total IPOs where the underwriter played some role subtract the sum over all IPOs over this period of NLBR i divided by the sum over all IPOs over this period of NUW i where NLBRi is the number of lead bookrunners on deal i and NUWi is the number of underwriters on deal i. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in its counts. Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify underwriters and their roles and this is confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (nonlead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. 29

30 Appendix A (Continued) Variable Definition Bookrunner Structure, Continued Proportion of IPOs where underwriter is non lead bookrunner Underwriter likelihood to work with multiple lead bookrunners, when lead If the total number of IPOs over the year prior to the current offering where the underwriter participated in some role is less than 4, this variable is set to 0. Otherwise, this variable is the total number of IPOs over this period where the underwriter is a non-lead bookrunner divided by the total IPOs where the underwriter played some role subtract the sum over all IPOs over this period of NNLBR i divided by the sum over all IPOs over this period of NUW i where NNLBRi is the number of non-lead bookrunners on deal i and NUWi is the number of underwriters on deal i. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in its counts. Source: Thomson Financial SDC (TFSDC) New Issues database is used to identify underwriters and their roles and this is confirmed through prospectus searches on the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system database. A lead (nonlead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. For each underwriter at the time of the IPO, we count the number of deals over the 12 months prior to the IPO where the underwriter acted as a lead bookrunner and the number of IPOs where the underwriter acted as lead bookrunner and there was more than one lead bookrunner. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in both counts. If the number of IPOs where the underwriter acted as a lead bookrunner is less than four, this variable is set to zero. Otherwise this variable is the fraction of the underwriter's past deals as a lead bookrunner where there were more than one lead bookrunner less the average fraction of multiple lead bookrunner IPOs across all offerings (i.e. the number of IPOs with more than one lead bookrunner over the past 12 months divided by the total number of IPOs). Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. Underwriter likelihood to include nonlead bookrunner, if sole lead For each underwriter on an IPO, we count the number of deals over the 12 months prior to the IPO where the underwriter acted as a sole lead bookrunner and the number of IPOs where the underwriter acted as sole lead bookrunner and there were more than zero non-lead bookrunners. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in both counts. If the number of IPOs where the underwriter acted as a sole lead bookrunner is less than four, this variable is set to zero. Otherwise this variable is the fraction of the underwriter's past deals as a sole lead bookrunner where there were more than zero non-lead bookrunners less the average fraction of non-zero non-lead bookrunner IPOs across all offerings having sole lead bookrunners (i.e. the number of IPOs with more than zero non-lead bookrunners on deals with one lead bookrunner over the past 12 months divided by the total number of IPOs having a single lead bookrunner). Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. 30

31 Appendix A (Continued) Variable Definition Bookrunner Structure, Continued Underwriter likelihood to include nonlead bookrunner, if multiple lead Underwriter likelihood to work with sole lead bookrunner, when non-lead For each underwriter on an IPO, we count the number of deals over the 12 months prior to the IPO where the underwriter acted as a lead bookrunner on deals with other lead bookrunners and the number of IPOs where the underwriter acted as a lead bookrunner on deals with other lead bookrunners and there were more than zero non-lead bookrunners. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in both counts. If the number of IPOs where the underwriter acted as a lead bookrunner on deals with other lead bookrunners is less than four, this variable is set to zero. Otherwise this variable is the fraction of the underwriter's past deals as a multiple lead bookrunner where there were more than zero nonlead bookrunner less the average fraction of non-zero non-lead bookrunner IPOs across all offerings having multiple lead bookrunners (i.e. the number of IPOs with more than zero non-lead bookrunners on deals with more than a sole lead bookrunner over the past 12 months divided by the total number of IPOs having multiple lead bookrunner). Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. For each underwriter on an IPO, we count the number of deals over the 12 months prior to the IPO where the underwriter acted as a non-lead bookrunner on deals with a sole lead bookrunner and the number of IPOs where the underwriter acted as a non-lead bookrunner. If an underwriter merged during the 12 month period, IPOs managed by pre-merger underwriters are included in both counts. If the number of IPOs where the underwriter acted as a non-lead bookrunner is less than four, this variable is set to zero. Otherwise this variable is the fraction of the underwriter's past deals as non-lead bookrunner where there was a sole lead bookrunner less the average fraction of deals with non-lead bookrunners having only one lead across all offerings having non-lead bookrunners (i.e. the number of IPOs with non-lead bookrunners and a sole lead over the past 12 months divided by the total number of IPOs having non-lead bookrunners). Sources: Underwriters and their roles and allocations are identified using TFSDC New Issues Database and confirmed through prospectus searches on the SEC EDGAR system database. TFSDC s merger and acquisitions database is used to account for underwriter mergers. A lead (non-lead) bookrunner is defined as a bookrunner receiving a share allocation at least (less than) 90% of the maximum allocation given to any bookrunner on the IPO. 31

32 Figure 1: Facebook Inc. IPO Prospectus Source: Facebook Inc. Form 424B4 Filed May 18, 2012, EDGAR Online, Inc. 32

33 Figure 2: Facebook Inc. IPO Prospectus Underwriting Table Source: Facebook Inc. Form 424B4 Filed May 18, 2012, EDGAR online, Inc. 33