IPO Commissions: Theoretical Predictions and Existings

Transcription

1 Purchasing IPOs with commissions: Theoretical predictions and empirical results Michael A. Goldstein Babson College Paul J. Irvine University of Georgia W. Andy Puckett University of Missouri June, 2008 We thank the Abel/Noser Corporation and Judy Maiorca for providing the data. Also thanks to Dan Bradley, John Griffin, Charles Hadlock, Kathleen Hanley, Eugene Kandel, Laurie Krigman, Jim Linck, Marc Lipson, Alexander Ljungqvist, Jonathan Reuter, Chester Spatt, Kent Womack, Donghang Zhang and seminar participants at the 2007 AFA conference, the 2006 Harvard EVI conference, Clemson University, the University of Missouri, the University of Nebraska and Syracuse University for their helpful comments. Goldstein wishes to acknowledge financial support provided by the Babson Faculty Research Facility. Corresponding author

2 Purchasing IPOs with commissions: Theoretical predictions and empirical results Abstract The prevailing literature suggests that either short-term or long-term investors send commissions to underwriters in return for lucrative IPO allocations. We reconcile these two ideas into a single theoretical framework, generate new and unique hypotheses, and, using a proprietary database of institutional trades, find direct evidence that institutions engage in churning stocks and paying abnormally large commissions to the lead underwriters of upcoming IPOs. Paying excess commissions to underwriters is a particularly effective way for short-term investors to receive lucrative IPO allocations. As predicted by our model, excess commissions are inversely related to the concentration of the underwriter s client base, indicating that underwriters concern for their long-term client relationships disciplines this practice. We estimate that total market-wide abnormal commission payments to be $2.2 million per IPO, and that an additional $1 payment of excess commissions to the lead underwriter leads to $2.76 in investor profits.

3 I. Introduction Institutional clients of investment banks are justifiably interested in receiving initial public offering (IPO) allocations given their historical profitability. The 1,555 firms that went public from 1999 to 2005 left more than $82 billion on the table. 1 Since lead underwriters have significant discretion in allocating shares when bookbuilding is used, much of the lobbying by institutional clients should be directed toward the lead underwriter. 2 Existing academic theories that seek to explain the allocation decisions of underwriters suggest that underwriters receive benefits (tangible or intangible) in return for allocating shares to certain clients. Benveniste and Spindt (1989) suggest that IPO allocations encourage privatelyinformed investors to reveal their information to the lead underwriter. While this intangible benefit may be factor in allocation decisions, it is also possible that underwriters allocate IPO shares to investors who provide them with more tangible benefits. According to the agency view advocated by Loughran and Ritter (2002, 2004), investors will engage in rent-seeking behavior, such as sending trading commissions to the underwriter s brokerage arm, to increase their probability of being allocated profitable IPO shares. Recent survey evidence by Jenkinson and Jones (2007) raises doubts concerning the extent of information production by institutional investors in the IPO process, and instead provides support for the agency view. Both large and small institutions responding to the survey indicated that brokerage commissions paid to the underwriter are the most important determinant in receiving IPO allocations. Similarly, NASD documents report that Robertson Stephens used an index, ranking investors by the amount of commissions paid over the previous eighteen months, in order to decide who would receive profitable IPO allocations, with recently generated commissions receiving greater weight. 3 This evidence is also consistent with empirical findings by Goldstein, Irvine, Kandel, and Wiener (2008) who 1 Money left on the table is defined as the difference between the first day closing price and the offer price times the number of shares offered as in Loughran and Ritter (2002). We obtain information for the total dollar value of money left on the table from Jay Ritter s website: 2 Boehmer, Boehmer, and Fishe (2006) find the lead underwriter is responsible for allocating approximately 75% of the total number of shares offered. 3 According to the SEC, the allocation of IPO shares based on past or expected future commission business is legal. However, lead underwriters are restricted from sharing in any client profits that may result from underpricing. See Letter of Acceptance, Waiver, and Consent sent to the NASD (no. CAF030001). 1

4 suggest that institutions develop relationships with brokerage firms by paying premium commissions. They find that institutions concentrate their trading to become important clients of particular brokers, and in return, receive preferential treatment with regard to brokerage services. Only two empirical papers to date investigate the relationship between trading commissions and IPO allocations. Using semi-annual mutual fund reports from , Reuter (2006) finds a positive correlation between the commissions paid to an underwriter and a mutual-fund family s holdings of recent profitable IPOs from that same underwriter. The infrequency of Reuter s (2006) data suggest that lead underwriters allocate profitable IPOs to long-term clients who provide a regular stream of commission revenue. Alternatively, Nimalendran, Ritter and Zhang (2007) investigate whether aggregate trading volume in the 50 most liquid stocks is related to subsequent money left on the table. Using Trade and Quote (TAQ) data, they find a positive relationship during the internet bubble period. 4 They suggest that short-term investors are churning stocks in order to send commissions to the lead underwriter immediately preceding the IPO offer date. The longterm and short-term views of IPO allocation seem contradictory, as two distinct groups receive preference in IPO allocation. In addition, both studies are constrained by both IPO allocation and commission data limitations. Our paper adds to the existing literature by theoretically and empirically joining the long-term and short-term agency views of Reuter (2006) and Nimalendran, Ritter, and Zhang (2007). We are able to overcome many of the data limitations of these previous papers by using a proprietary database of institutional trades. We first contribute to the existing theoretical literature by presenting a single framework where both long-term and short-term investors coexist. In our model, the lead underwriter faces a tradeoff: if the underwriter allocates shares to short-term investors she will receive the additional commission dollars that these investors provide. However, if the underwriter s malfeasance is detected by longterm investors, these investors will penalize the underwriter by withholding future commission revenues. The model provides two main testable empirical predictions: shortterm investor commission payments are increasing in the expected profitability of the IPO 4 Nimalendran, Ritter and Zhang (2007) do not have information on trading commissions or the brokerage firm involved in each trade, and therefore their inference on the behavior of short-term traders is limited to circumstantial evidence. 2

5 and decreasing in the concentration of the underwriter s client base. This second hypothesis is new to the literature and uniquely derives from our model. Unlike previous work, we are able to test our predictions directly using a proprietary institutional trading database. The Abel/Noser database contains trade-level data from 1999 to 2005 for 840 institutional investors, where the data for each trade include both commissions paid and the identity of the brokerage firm involved. This data enables our paper to make four distinct contributions to our current understanding of the relationship between lead underwriter commission revenues and IPOs. First we investigate directly the existence and timing of excess underwriter commission revenues in the period surrounding IPO issuance. We divide all IPOs into quartiles based on the amount of money left on the table, and, using an event study methodology, find significant increases in lead underwriter commission revenue for the two most profitable IPO quartiles. Excess commission revenues appear in the ten-day period immediately preceding the IPO offer date, and are evident in both bubble ( ) and non-bubble ( ) periods. Post-issue commission payments in return for IPO allocations appear to be specific to one subsequently prosecuted brokerage firm rather than a general phenomenon. 5 Our results are consistent with strategic decisions by short-term investors to use commission dollars as a means of obtaining profitable IPO allocations. We confirm that these findings are robust to a variety of alternate specifications and cannot be attributed to market-wide changes in trading volume or the clustering of IPO issuance. For the most profitable quartile of IPOs, we estimate that the lead underwriter receives abnormal commission revenues of approximately $2.2 million per IPO during the ten day period before the offer date. Overall, across the two most profitable IPO quartiles, our estimates suggest that lead underwriters collectively received $1.14 billion in excess commissions. Our second contribution investigates potential trading strategies institutions may use to increase commissions paid to lead underwriters. For example, institutions may simply re- 5 The SEC strictly prohibits ex post profit sharing. In 2002 the SEC charged Credit Suisse First Boston (CSFB) with receiving kickbacks in the form of inflated commissions from clients who received profitable IPO allocations. The SEC claimed that several of CSFB s institutional clients kicked back up to 65% of the IPO profits to the brokerage in the form of excessive commissions. 3

6 allocate normal trading volume to an underwriter with an upcoming IPO. Alternatively, institutions may inflate commission revenues by churning stocks (the rapid purchase and sale of the same stock), increasing the average commission per share paid, or paying unusually high commissions for some trades. These choices are not mutually exclusive. In fact, we find that all of these strategies are used by institutions in our sample: the total number of shares traded, commissions from churn trades, average commission per share charged, and the frequency of trades paying greater than 10 cents per share (an unusually high commission rate) are all significantly elevated in the pre-ipo period. We estimate that commissions from churn trades account for 25% of the total increase in commission revenue for the most profitable IPO quartile. Our third contribution is to investigate empirically a prediction unique to our model: that short-term investor commission payments are decreasing in the concentration of the underwriter s client base. Using a multivariate regression to investigate the relation between underwriter characteristics and abnormal IPO commissions, we find robust evidence that abnormal commissions paid to the lead underwriter are inversely related to the concentration of the underwriter s client base. This finding is consistent with our conjecture that long-term investors can effectively discipline their brokerage firms. This disciplinary mechanism keeps the excess commission practice small relative to the potential IPO profits available. Finally, our data allow us to examine whether institutions are successful in using commissions to capture IPO profits. Specifically, we investigate whether increased commissions sent to the lead underwriter result in larger allocations of profitable IPOs. Since IPO allocations are not available in our data, we proxy for these allocations by examining net selling by each institution in the 30 days after the IPO. We find evidence that short-term and long-term institutions interact differently with lead underwriters. Our findings suggest that profitable IPO allocations to long-term institutions are primarily determined by the long-term commission revenue streams that such institutions provide. Alternatively, excess commissions sent to the lead underwriter in the period immediately preceding the IPO offer date are more important for short-term investor allocations. We estimate that $1 dollar of abnormal commission revenue sent by short-term institutions to the lead underwriter generates $2.76 in IPO profits from allocated shares. It appears that transient institutions are successful in using commissions to capture excess IPO profits. 4

7 In the next section, we develop a theoretical model that explains the coexistence of both long-term and short-term investors, thereby reconciling the findings of Reuter (2006) and Nimalendran, Ritter and Zhang (2007). Section III presents our data and Section IV presents the main empirical results of our investigation. Section V examines some trading strategies that institutions might employ to increase commission payments. Section VI examines the determinants of abnormal commissions, including client concentration and IPO profitability. Section VII examines the relation between abnormal commissions and IPO allocations more directly using data on post-ipo sales. Section VIII concludes. II. Analytical Model and Hypotheses II.A. Analytical Model Benveniste and Spindt (1989) and others have developed models of the bookbuilding process and its relation to IPO underpricing. While information revelation may be a significant determinant in underwriters allocation decisions, it is also probable that quid-proquo commission arrangements are of principal importance in the allocation process (Jenkinson and Jones (2007), Loughran and Ritter (2002, 2004)). As such, our model focuses on allocation decisions across different types of clients who provide commission business to the lead underwriter. We therefore do not attempt to model the entire IPO pricing process, but instead concentrate on outlining results that shed light on commission routing behavior. A risk-neutral brokerage firm s customers consist of two types of investors L and S. The first type, L, are long-term investors who pay a regular stream of commission dollars, C, to the brokerage firm. As in Reuter (2006), Binay, Gatchev, and Pirinsky (2007), and Goldstein, Irvine, Kandel and Wiener (2008) these investors enter into implicit long-term contracts where they agree to pay premium commissions (relative to ECN execution) and, in return, expect to receive premium services, including allocations of desirable IPOs. The second type of investor, S, are short-term investors who attempt to buy their way into IPO allocations by directing commissions, c, to the brokerage firm. The type S commission stream, c, is short-term in nature. That is, short-term investors have no interest in continuing to pay large commissions once the IPO has been allocated. Once the brokerage firm has won the IPO underwriting mandate, they can be considered a monopolist over the allocation process. The underwriter s problem is to decide 5

8 whether to allocate all of the IPO to its long-term investors, L, and receive C or to cheat on the implicit contract and allocate part of the offering to type S investors and receive c as well. Any cheating by the underwriter will increase the underwriter s profit but reduce the total profits of the long-term investors. Should a long-term investor detect cheating, they will punish the underwriter by withholding future commissions. Therefore, the underwriter will cheat on the implicit contract only if the commissions they receive from type S investors are large enough to offset the potential losses from being caught and punished by long-term investors who discover the malfeasance. We model the brokerage firm s revenues if caught cheating by a type L investor as kc (where k is [0, 1]); which represents the fraction of type L trading commissions that the underwriter continues to receive from long-term investors who do not detect cheating. 6,7 Short-term investor revenues c will be increasing in the expected profitability E(π) of their IPO allocation: ~ E (π ~ ) = E( F) A (1) Where F ~ is the expected first day s closing price less the known IPO offer price, and A is the number of shares expected to be received in the allocation process. While A is also unknown ex ante, it is determined endogenously in our model and thus can be predicted by the short-term investor given the excess commissions they pay and the exogenous elements of the lead underwriter s problem. Short-term investors will offer an extra payment when expected profits are positive: ~ A E( F) c > 0 (2) Clearly, there will be an ongoing bargaining problem between the underwriter and the shortterm investor over the division of potential profits. However, the bargaining problem only adds complexity without adding additional insight into our main results. Therefore, to highlight the main implications, we assume that the broker extracts all possible rents from the 6 Alternatively, we could assume that if the underwriter cheats, all long term investors find out, and all of them lower their order flow sent to that underwriter by k. Our presentation assumes that one, or a subset, of type L investors discovers the underwriter s cheating behavior. Under our assumption, the nature of the underwriter s client base could affect their decision, a contention that is empirically tested in Section VI. 7 Intuitively, C could be considered the present value of all future commissions generated from the type L investors. 6

9 short-term investor and thus the size of the bribe can be defined directly as a function of ~ expected IPO underpricing E (F) : ~ c = A E(F). (3) Faced with the possibility of additional revenue from receiving a bribe, the underwriter s choice is to allocate all IPO shares to long-term investors L and receive C or to choose to allocate A shares to short-term investors S. If they are not caught by their long-term investors, the underwriter will continue to receive C and pocket the bribe, c. If they are caught, then the underwriter will receive the bribe, c, but only kc from their long-term customers. The postulated function р(a) represents the probability that long-term investors L detect cheating and punish the lead underwriter. Hence p(a) determines the underwriter s willingness to allocate shares to short-term investors. The probability of being caught cheating is assumed to be increasing and convex in the size of the allocation, A, to type S investors. Given this assumption, the underwriter s decision on whether to cheat depends on the relative payoffs of the two strategies: [ 1 p( A) ] C p( A)( kc), C < c + + (4) or: p ( A)(1 k) C < c, (5) where (1-k) represents the punishment fraction, or the size - as a percentage of brokerage firm commission revenue - of the type L investor that catches the broker cheating on their long-term agreement. If the underwriter s participation constraint in Equation (5) is met, the underwriter will allocate shares to type S institutions. In this framework a type L investor may be unhappy with their allocation and suspect the broker of reneging on their long-term agreement. However, since the actual demand function for IPO shares is determined in the pre-issue bookbuilding process, only the underwriter sees the true aggregate demand curve. Therefore, verification of cheating is costly and uncertain. The underwriter will be more likely to cheat and allocate IPOs to type S investors, the greater the short-term commission payment c, the lower the probability of being caught cheating p(a), and the less severe the punishment fraction (1-k). 7

10 The underwriter s problem is to maximize revenue (R) from Equation (4) conditional on the allocation size, A, to the short-term investors: Max A [ p( A) ] C + p( A)( kc + c R = 1 ) (6) Substituting Equation (3) for c produces the first order condition: or: R ~ = p' ( A)(1 k) C + E( F) = 0 A (7) p '( A)(1 k) C = E( F ~ ) (8) II.B. Hypotheses The solution to the underwriter s problem in Equation (8) trades off the marginal ~ benefit from sending one more share to type S investors, E (F), against the marginal cost, p '( A)(1 k) C, which is the change in the probability of detection multiplied by the cost of the punishment. Figure 1 illustrates the potential solutions to the problem. Under our assumption that the underwriter extracts the entire surplus from type S investors, the marginal ~ gain from the bribe is E (F). The total size of the bribe is then calculated from (3): the higher the expected underpricing, the steeper the bribe size line in Figure 1. Using our earlier assumption that p ''( A) > 0, underwriter costs are represented in Figure 1 as a convex function, where (1-k)C is a constant. This result leads directly to our first hypothesis: Hypothesis 1: Short-term investors commission payments are increasing in the expected profitability of the IPO. Empirically, Hypothesis 1 predicts that we should observe larger increases in underwriter commissions, the larger the expected profitability of the IPO. When expected underpricing is relatively low, only small bribes will be forthcoming. Even if we relax the restrictive bargaining assumption that the underwriter extracts all of the short-term investor s profits, Hypothesis 1 still holds; any point below the Bribe Size line and to the left of the Underwriter Cost curve is a potential equilibrium. 8

11 From Equation (8) we can also hypothesize about the size of the bribe conditional on the underwriter s client base. The optimal number of shares allocated to type S investors is a function of the cost of the punishment (1-k)C, which represents the amount of commission business the underwriter loses should their malfeasance be detected by a type L investor. The more important any one particular type L client is to the underwriter, the greater the punishment fraction (1-k). Thus, the concentration of an underwriter s client base may affect the size of the optimal allocation to short-term investors. Hypothesis 2: The likelihood that an underwriter will cheat the long-term contract and allocate IPO shares to short-term investors is a decreasing function of the concentration of the underwriter s client base. If our assumptions regarding the cost of punishment are correct, then the more important a particular long-term client is to the underwriting broker, the greater the penalty incurred should the underwriter be caught allocating shares to short-term investors. 8 Effectively, the probability of detection and the cost of the punishment will be greater for underwriters with a concentrated client base. Empirically, we expect fewer excess commission payments to this type of underwriter. II.C. Price uncertainty and timing The timing of excess commission payments can tell us about the nature and behavior of the IPO allocation market. At first glance, a risk-neutral type S investor could potentially offer the extra payment at any time. However, Dixit (1989) outlines the value waiting to invest can have, even for the risk-neutral investor. Nimalendran, Ritter and Zhang (2007) focus on the ten days prior to the IPO, but only find significant results in the week prior to the IPO issue date. Any degree of risk aversion would suggest that type S investors would wait to offer excess payments until just before the issue date if the uncertainty over IPO profitability declines as the offer date approaches. Academic research on the predictability of IPO underpricing and institutional experience with road-show presentations indicates that this is likely the case. Institutional clients are privy to the broker s pre-ipo road show 8 Larger clients may also monitor the lead underwriter more closely. Expecting to receive substantial IPO allocations, large long-term clients can spread their monitoring costs over larger potential profits. Further, comprising a significant fraction of bookbuilding demand, large clients may have better information about total demand for the issue. 9

12 presentations; simply observing the crowd at these presentations can tell experienced institutional investors something about the latent demand for the issue. Extra payments could conceivably come after the IPO. Brennan and Chordia (1993) suggest that ex post settling up is the preferred way for risk-averse investors to deal with uncertain brokerage firm services. If this is the case for IPO allocations, we would expect to see an increase in commission payments sent to the underwriter after the IPO. However, as the 2003 SEC settlement with CSFB reveals (SEC litigation release 17327), the regulatory authorities have taken the view that this behavior amounts to sharing IPO profits with clients and is therefore actionable. It is unknown whether ex post excess commission payments generally prevailed or were specific to CSFB. As our data is more precise than Nimalendran, Ritter and Zhang (2007), we extend the test period to incorporate twenty trading days both before and after the IPO. If investors offer extra payments throughout the pre-offer period, we should observe elevated lead underwriter commissions for several weeks prior to the IPO issue date. Alternatively, investors could wait and gather information on the expected profitability of the IPO, offering extra payments just prior to the IPO issue date. In this case, we should observe elevated lead underwriter commissions only in the week or two prior to the IPO issue date. Finally, investors may offer ex post payments of elevated lead underwriter commissions after the issue date. These strategies are not mutually exclusive, for example, the IPO allocation market could consist of a payment offered ex ante and then ex post settling up for particularly profitable IPOs. Ultimately, both the existence and timing of excess commission payments is an empirical issue. We address these questions in the next section. III. Data and methodology III.A. Trading data We obtain institutional trading data from the Abel/Noser Corporation, a widely recognized consulting firm that works with institutional investors to monitor their equity trading costs. 9 Abel/Noser clients include pension plan sponsors such as the California Public Employees' Retirement System (CalPERS), the Commonwealth of Virginia, and the YMCA 9 Abel/Noser provides consulting services for equity trading costs in a manner similar to the Plexus Group, whose data has been used extensively in academic studies. Other studies that have used Abel/Noser data include Chemmanur and Hu (2007), Lipson and Puckett (2007), and Goldstein, Irvine, Kandel, and Wiener (2008). 10

13 retirement fund, as well as money managers such as MFS (Massachusetts Financial Services), Putman Investments, Lazard Asset Management, and Fidelity. The Abel/Noser sample contains trades from 840 institutions and covers the period from January 1, 1999 until December 31, Summary statistics for the Abel/Noser trade data are presented in Panel A of Table 1. Institutional investors in the database are responsible for more than 87 million trade executions during the sample period. For each execution, Abel/Noser provide 107 different identifying variables. Our study uses eight of these identifying variables including the institution identity code, identity of the broker/dealer handling the execution, date of execution, stock traded, number of shares executed, execution price, whether the execution is a buy or sell, and commissions paid. While the identity of the institution is not provided to us, the unique identity codes allow us to distinguish between different institutions trades both in the cross-section and through time. 10 The average number of shares per execution varies from 6,669 in 2005 to 11,159 in 2001, while commissions per execution range from $428 in 2002 to $176 in Over the entire sample period, Abel/Noser institutional clients traded more than 755 billion shares ($22.9 trillion), and paid more than $24.6 billion in commissions to broker/dealers. On average, institutions in our sample are responsible for at least 7.97% of total CRSP daily dollar volume during the 1999 to 2005 sample period. 11 In untabulated results, we aggregate trading by the brokerage firm (or ECN) responsible for each execution and investigate commission revenues. We find that all ten of the largest brokers (ranked by average commission revenues per day) also underwrite IPOs during our sample period. Merrill Lynch is the largest broker, earning an average of $873,388 in commissions each trading day. The tenth largest broker is J.P. Morgan, who earns an average of $362,881 per day in commission revenues. 10 Identifying variables also include summary execution costs for ticket orders which often include multiple executions. These variables include the share-weighted execution cost and total number of shares executed in the ticket. Abel/Noser receives trading data directly from the Order Delivery System (ODS) of all money manger clients, and therefore includes all trades executed by managers. The method of data delivery for pension plan sponsors also includes all executed trades. 11 We calculate the ratio of Abel/Noser trading volume to CRSP trading volume during each day of the sample period. We include only stocks with sharecode equal to 10 or 11 in our calculation. In addition, we divide all Abel/Noser trading volume by two, since each individual Abel/Noser client constitutes only one side of a trade. We believe this estimate represents an approximate lower bound for the size of the Abel/Noser database. 11

14 III.B. IPO data We use the Security Data Company s (SDC) new issues database to identify IPOs from March 31, 1999 to October 1, 2005 and the CRSP database to obtain first day closing prices for each IPO firm. 12 We exclude all ADRs, REITs, UITs, closed-end funds, and IPOs with an offer price less than $5. Our filters leave us with a sample of 1,183 IPOs. Finally, we require that the lead underwriter is listed as a broker in the Abel/Noser database, which eliminates 27 firms and leaves us with a final sample of 1,156 IPOs involving 88 different lead underwriters. Following Loughran and Ritter (2002) we calculate money left on the table (Tablemoney) for each IPO in our sample as the first day closing price minus the offer price multiplied by the number of shares offered. As such, Tablemoney represents the total first day IPO profits available to investors. From Hypothesis 1 of our model, we expect excess brokerage commissions around the IPO offer date to be positively related to Tablemoney. To examine this prediction, we rank all IPOs by Tablemoney and separate the sample into quartiles. Summary statistics for each Tablemoney quartile are presented in Panel B of Table 1. IPOs in the highest Tablemoney quartile (quartile=4) present investors with an average of $174.4 million in potential first day profits, which is more than six times the average first day profits of the third quartile ($26.7 million). The second and first quartiles have Tablemoney averages of $6.1 and -$16.7 million respectively. IPO profitability is highly correlated with underpricing, which ranges from an average of percent in the highest Tablemoney quartile to percent in the lowest quartile. Examining our measure of lead underwriter size (commission revenues per day) indicates that lead underwriters of the most profitable IPOs are slightly larger brokerage firms, although this difference is not statistically significant. Average daily lead underwriter commission revenues during the [-60, +60] trading day period are $375,300 for the most profitable IPOs and $337,687 for the least profitable IPOs. 12 Although Abel/Noser trading data spans January 1, 1999 to December 31, 2005, we require all IPOs to have sixty days of trading data before and after the offer date in order to calculate expected levels of commission revenue for each broker. We also check Jay Ritter s IPO website: for possible SDC data errors in our sample and for the SDC data errors mentioned in footnote 4 of Ljungqvist and Wilhelm (2003); see 12

15 Several summary statistics presented in Panel B merit further discussion. First, IPOs in quartiles 3 and 4 are extremely profitable. The profits from receiving these IPO allocations appear large enough to entice short-term institutions to attempt to purchase IPO allocations with excess commission payments. Second, IPO profitability is positively correlated with the dollar size of the offering (ρ=0.31). A higher offer size means a larger allocation pie is available and may allow short-term investors greater opportunities to receive allocations. In the following section we investigate whether empirical evidence supports the hypothesis that short-term investors submit commission bribes in order to receive lucrative IPO allocations. IV. Main Results IV.A. Event Study We first investigate the existence and timing of abnormal lead underwriter commission revenues using an event study methodology. For each IPO in our sample, we collect all trades executed by the lead underwriter during the [-60, +60] trading day period surrounding the IPO offer date and calculate the total commission revenue earned for each day. We separate the sample by Tablemoney quartile, and for each quartile compute the average commission revenue for each day in the time series. 13 As a basis for our statistical tests, we create a benchmark level of mean daily lead-underwriter commission revenue during the [-60, -21] and [+21, +60] non-event period. We then compare the average daily event-period commission revenue to the benchmark level using the standard deviation of commission revenues in both the benchmark and event periods to construct our test statistic. 14 Since prior literature provides little guidance regarding the timing of abnormal commission payments, our initial investigation examines four ten-day event periods surrounding the offer date: [-20, 11], [-10, -1], [1, 10], and [11, 20]. Because of the welldocumented increase in trading activity that occurs on the offer date, the offer date itself is omitted in tests for abnormal commissions In robustness tests, we also separate the IPO sample into quartiles based on Underpricing, and then repeat the event study. Results are similar to those reported. 14 Our methodology is identical to that used in Corwin and Lipson (2004) and Irvine, Lipson, and Puckett (2007). 15 We also examine separately whether commission revenue on the offer date (excluding IPO trading) is significantly different than non-event period commission revenue. We find that abnormal commission revenues are positive and statistically significant for quartile 4 (high Tablemoney) and quartile 3, but are insignificantly 13

16 Table 2 presents our event study results for abnormal commissions. We reiterate that our tests evaluate commissions paid to the lead underwriter only. Our findings suggest that some investors increase commissions sent to the lead underwriter in the period immediately preceding the most profitable IPOs. We also find some evidence of decreases in lead underwriter commission revenue around the least profitable IPOs. For the most profitable quartile, the average lead underwriter receives $374,772 in commissions per day during the non-event benchmark period, and this amount increases by $18,150 per day (t-statistic=4.54) during the [-10, -1] event period. This 5% increase per day cumulates to $181,500 over the [- 10, -1] period, or about another half-day s worth of commissions. For quartile 3, the increase in lead underwriter commissions during the [-10, -1] event period is $13,342 per day (tstatistic=1.95). Our initial results therefore support Hypothesis 1: that excess commission payments exist and that commission payments are increasing in the expected profitability of the IPO. 16 Nimalendran, Ritter, and Zhang (2007) present circumstantial evidence that these abnormal commission payments are observable only during the bubble period. To investigate this possibility, we divide our sample into bubble (1999 to 2001) and non-bubble (2002 to 2005) periods and repeat our analysis for both subperiods. For expositional convenience, we report only abnormal commissions during the [-10, -1] event period for bubble and nonbubble periods in Table 2. Overall, we find that results are qualitatively similar for both bubble and non-bubble periods. We first note that abnormal lead underwriter commissions are $14,608 per day for quartile 4 IPOs during the bubble period, while they are $37,150 in the non-bubble period, and both estimates are statistically significant at the 1% level. Interestingly, due to the increase in overall commissions paid during the non-event periods, in both the bubble and non-bubble periods the abnormal lead underwriter commissions are about 5% more than the different from zero for quartiles 1 and 2. As these results are similar to and supportive of those found in our main results, we do not include them in our tabulated results. 16 We find no evidence of abnormal ex post commission payments for either of these quartiles. In separate tests, we analyze Credit Suisse First Boston (CSFB) alone since CSFB is specifically cited in SEC documents alleging ex-post settling-up behavior. In the CSFB-only sample, we find evidence of significant abnormal commission payments in the [+1, +5] period after the IPO offer date for both quartile 3 and quartile 4 IPOs during the bubble period. By extrapolating our volume data to CRSP total volume levels, we estimate over $46.7 million in excess commissions were received by CSFB in this period. Our estimates are consistent with the magnitude of SEC litigation release claiming $70 million in improper gains that CSFB was ordered to disgorge. 14

17 benchmark level. 17 For quartile 3 IPOs, there are some differences between the time periods. Event period commissions for quartile 3 are not significantly different than benchmark levels during the bubble period, but during the non-bubble period abnormal commissions are $32,712 (t-statistic=2.55) per day for quartile 3 IPOs. Our results suggest that abnormal commission payments exist in both the bubble and non-bubble periods, and that in each case institutions in aggregate seem to pay about 5% more to receive profitable IPO allocations. As noted previously, our estimates suggest that ten-day lead underwriter excess commission revenue per IPO is $181,500 for IPOs in quartile 4 and $133,420 for IPOs in quartile 3. However, our data represents only 7.97 percent of CRSP daily share volume, if we gross up our average abnormal commission per IPO (by 1/0.0797), we estimate total marketwide abnormal commissions of $2,277,000 ($1,674,000) per IPO for lead underwriters of quartile 4 (quartile 3) IPOs. With 577 IPOs in quartiles 3 and 4, aggregate abnormal commissions received by lead underwriters of these IPOs is $1.14 billion, which is both economically and statistically significant. The economic magnitude of this revenue is still small relative to the $58 billion left on the table by these IPOs. It is puzzling why even higher abnormal commissions are not observed given the large profits available. Profit maximizing underwriters would be better served by raising the offer price of the IPO and capturing 7% of any additional proceeds in the form of underwriting fees (Chen and Ritter (2000)). However, this argument ignores externalities surrounding underpricing and allocation decisions including extracting valuable information from informed investors (Benveniste and Spindt (1989)) or managing litigation risks (Tinic (1988), Lowry and Shu (2002)). Our contribution to this puzzle centers on commission revenues. Long-term investors expect to receive IPO allocations in return for the stream of commissions they pay, while underwriters concern for retribution from these investors limits the total amount of shares allocated to short-term investors. It is also interesting that we do not observe excess commissions for quartile 2 IPOs, which are profitable ex post. Consistent with our model, these results suggest that expected 17 There are two primary reasons why average daily non-event period commissions increased significantly between the bubble and non-bubble periods. First, the aggregate level of trading activity is increasing over our sample period for both the overall market and for our sample of Abel/Noser institutions. Second, and perhaps more importantly, trades on Nasdaq-listed stocks were generally not charged explicit commissions prior to 2002, but as a result of decimalization, commissions were charged on more than 90% of the trades for Nasdaqlisted stocks from 2002 onward. 15

18 profits must be large to engender short-term investors to offer commission payments to obtain allocations. The left-hand side of the broker s participation constraint (Equation 5) must be large if only the most profitable IPOs produce abnormal commission payments meeting the constraint. If the probability of long-term investors detecting cheating, p(a), rises quickly in A, then the Underwriter Cost curve in Figure 1 rises quickly in A. In this case, the Underwriter Cost curve and the Low Underpricing Bribe Size line would intersect close to the origin, resulting in low optimal allocations to short-term investors that may be empirically undetectable. Yet, if expected underpricing is high enough, and hence the Bribe Size line is steep enough, an empirically detectable bribe may still exist for the most profitable IPOs. Thus, short-term commissions for IPO allocations are constrained to remain small relative to the potential IPO profits when the probability of detecting cheating rises quickly with the size of the allocation to short-term investors. Only when IPO profits are extremely large can the short-term investors justify a large enough extra payment to induce lead underwriters to allocate shares away from their long-term clients. IV.B. Robustness tests The event study results in Table 2 indicate that for the most profitable IPOs, lead underwriters receive increased commission revenues during the ten-day period immediately preceding the offer date. However, prior research reports that IPO activity is both clustered in calendar time and is related to aggregate market activity (Lowry (2003)). We investigate the potential effects of these trends on our findings in two robustness tests. IV.B.1. Calendar-time regressions If IPO events are clustered in calendar time, our event study may suffer from correlated errors and a tendency to over-reject the null. Although many prior studies such as Ritter and Welch (2002) document IPO clustering in hot markets, we reiterate that our analysis investigates lead underwriter commission revenues only, thus mitigating the clustering problem. We present the following illustration to clarify this issue: 16

19 IPOs issued in April 1999 Goldman Sachs IPOs Merrill Lynch IPOs 4/16 4/5 4/6 4/21 Clustering Credit Suisse IPOs 4/8 4/26 4/28 Clustering The above illustration presents timelines of IPOs issued by three of the largest investment banks in our sample for April Viewing all seven of these IPOs on a single (market-wide) timeline, six fall in the [-10, -1] event period of another IPO and would appear to be clustered. However, when IPOs are separated by lead underwriter as shown above, we find only two IPOs where another IPO from the same lead underwriter exists in the [-10, -1] event period. Although our unit of analysis clearly reduces the clustering problem, it is not eliminated completely. To address this issue, we employ a calendar-time regression approach. We proceed as follows: we aggregate commissions and money left on the table separately for each underwriter on each day in our sample period. We then specify the underwriters daily commission revenue as a function of future money left on the table (Tablemoney): 5,10 4 Commissiont, j = α + β1 Tablemoneyt + n, j + γ kcommissiont n, j + n= 1 n= δ m t m m Month + ε = 2 2 Mktrett The dependent variable in Equation (9) is the commissions received on day t by lead underwriter j. The Tablemoney variable is the aggregate amount of money left on the table by underwriter j, which we sum over days t+1 through t+5 or over t+1 through t+10. In this 17 β (9)

20 specification commissions received by the lead underwriter are a function of the cumulative future profitability of IPOs issued by underwriter j. To follow Nimalendran, Ritter, and Zhang (2007), we also include four lags of the Commission variable to control for daily autocorrelation in the level of underwriter commissions. Following Dennis and Strickland (2002) we include the absolute value of the CRSP equally-weighted return ( Mktret t ) on day t, since institutional volume is higher during days with large market movements. Finally, we include monthly fixed effects (Month) to control for any time-series changes in the frequency of trading in our data. We adjust all t-statistics using Newey-West standard errors. Table 3 presents the results of these regressions for the full sample as well as the bubble and non-bubble subperiods. The key variables of interest are Tablemoney[1,5] (Tablemoney[1,10]), which represent underwriter j s IPO profitability over the next five (ten) trading days. Results are consistent with our event study findings; today s lead-underwriter commissions are positively related to future money left on the table by that lead underwriter. For the full sample, the coefficient on Tablemoney[1,5] is (t-statistic=2.59), suggesting that for every $1 left on the table in the subsequent five days, the lead underwriter will receive 4.4 cents in abnormal commission revenues. 18 Our findings are similar when investigating Tablemoney[1,10] where we find a coefficient estimate of (tstatistic=2.45). We present identical regressions for bubble and non-bubble period subsamples. Overall, regression coefficients are similar to the full sample results. All of the Tablemoney coefficients in Table 3 are positive and significant, except for Tablemoney[1,5] in the non-bubble period, which has a positive but insignificant coefficient estimate. IV.B.2. Difference-of-differences Test Lowry and Schwert (2002) suggest that IPO activity is related to lagged market activity. Since lead underwriters have some discretion concerning the timing of the offer date, they may be more likely to bring IPOs to market when conditions are particularly favorable. Such conditions could be reflected in higher market volumes across all stocks. Thus, it is possible that aggregate commission payments to all brokerage firms, and not just the lead underwriter, increase just prior to hot IPO offer dates. To address this concern, we 18 This estimate is significantly greater than Nimalendran, Ritter and Zhang s (2007) estimate of the broker s payback ratio (0.52%). Thus, using lead underwriter data reveals a stronger connection between IPO allocations and commission revenue than is apparent from aggregate trading data. 18

21 construct a difference-of-differences test that directly compares the event-period commission revenues received by lead underwriters to the commissions received by other brokerage firms in the database. For each IPO in the sample, we construct a comparative sample of non-lead underwriting brokers by requiring that a broker cannot act as a lead underwriter for any IPO during the [-10, +10] day period surrounding a sample IPO s offer date. By comparing the time-series of commission revenues for lead underwriters to that of non-lead brokers, we control for external market conditions affecting all brokers. For each IPO, we calculate the average daily event period commission revenue [-10, - 1] minus the average daily non-event commission revenue from [-60, -21] and [21, 60], and divide this difference by the average daily non-event period commission revenue. As such, our measure can be interpreted as the percentage change in commission revenue experienced by the brokerage firm in the ten-day period immediately preceding the IPO offer date. This normalizations controls for cross-sectional differences across brokers in each group, and is similar to the methodology employed by Goldstein, Hotchkiss and Sirri (2007). We construct this measure for three groups: the lead underwriter, a matched-pair control sample (where the lead underwriter is matched with one non-lead brokerage firm that is closest in average daily non-event period commission revenue), and a control portfolio of all non-lead brokerage firms. For clarity we limit these tests to the [-10, -1] event period, since this is the period where we find a significant increase in lead underwriter commission revenue. For each group we test whether the percentage change in commission revenue is different from zero. Our results are reported in Table 4, and are consistent with all earlier results. We first examine the ratio of abnormal commissions for the lead underwriters in each quartile. 19 For the highest two Tablemoney quartiles, we find that lead-underwriter commissions increase by 10.62% (quartile 4) and 7.07% (quartile 3). Results not shown here demonstrate that these numbers are similar across the bubble and non-bubble sub-periods, so the magnitude of these increases are relatively constant across time. For quartile 2 and quartile 1, lead underwriter commission revenues increase by 0.69% and decrease by -1.10% respectively. These 19 The results in Table 2 are aggregates of commission dollars to lead underwriters across all IPOs, while the statistics n Table 4 are constructed from ratios of each lead underwriters event and non-event periods. As a result, the ratio of the aggregate numbers in Table 2 will be different than the average of the ratios shown in Table 4. 19