The Role of IPO Underwriting Syndicates



The Role of IPO Underwriting Syndicates:

Underpricing, Certification, and Information Production

Shane A. Corwin

and

Paul Schultz*

February 2003

JEL classification: G14; G32

Keywords: IPOs; Syndicates; Underpricing; Information Production; Liquidity

*Mendoza College of Business, University of Notre Dame. We thank Andrew Blum, Stuart Cable, Michel Habib, Alexander Ljungqvist, Michelle Lowry, Jay Ritter, Ann Sherman, Britt Tunick, Bill Wilhelm, members of the Securities Industry Association Syndicate Committee, and seminar participants at the University of Notre Dame and Pennsylvania State University for comments and suggestions. We are also grateful to IBES for providing data on analyst coverage and to Jay Ritter for providing underwriter rankings.

The Role of IPO Underwriting Syndicates:

Underpricing, Certification, and Information Production

Abstract

We examine syndicates for 1,547 IPOs from 1997 through 2000. We find that IPOs underwritten by large syndicates are underpriced less than other IPOs. In addition, offering prices are more likely to be revised in response to information when the syndicate has more underwriters and especially more co-managers. Relationships between underwriters are critical in determining the composition of syndicates, perhaps because they mitigate free-riding and moral hazard problems. Underwriters with a top-ranked analyst are also more likely to be included in a syndicate. Over our sample period, the average number of syndicate members declined while the number of co-managers increased.

1. Introduction

When a firm goes public, it chooses a book manager and co-managers to underwrite the IPO. The issuing firm and the book manager then pick additional non-managing underwriters to complete the underwriting syndicate. Almost all IPO syndicates include one or more co-managers and several non-managing underwriters. Despite this, there has been almost no academic research on the role and functions of these syndicate members. In this paper, we us a sample of 1,547 IPOs from 1997 through 2000 to examine the role and function of managing and non-managing underwriters in IPO syndicates.

Chen and Ritter (2000) suggest that co-managers are included in syndicates as a way of garnering analyst coverage. Using IPOs from 1985 – 1997, they estimate that each additional co-manager adds between 0.36 and 0.55 additional analysts. Consistent with Chen and Ritter, we find that each additional co-manager results in 0.69 additional analysts issuing reports in the three months after an IPO. In addition, we find that having a top-ranked analyst in the issuer’s industry significantly increases the likelihood that an underwriter will be included in a syndicate. We also find that co-managers provide another aftermarket service: market making. We estimate that each additional co-manager results in 0.70 additional market makers.

One new finding of this paper is that, all else equal, larger syndicates are associated with less underpricing. We estimate that adding an additional non-managing underwriter with a Carter–Manaster ranking below eight reduces underpricing by 0.7 percent, while each additional non-managing underwriter ranked eight or higher reduces underpricing by 1.6 percent. Co-managers have a bigger effect. Each additional co-manager with a Carter–Manaster rank below eight reduces underpricing by 6.0 percent. Each higher ranked co-manager reduces underpricing by 7.5 percent.

There are several possible explanations for the relation between underpricing and syndicate size. One is that issuers typically choose co-managers from among the firms that were in the running to be book manager. Thus, if there is fierce competition to be the book manager, syndicates may be larger at the same time that competition leads to more favorable pricing for the issuer. A second explanation is certification. When an underwriter “puts its name on the cover” of a prospectus, it signals its approval of an offering and accepts the accompanying legal liability. Third, additional underwriters may reduce underpricing by improving information production. A number of theories (e.g. Rock (1986), Beatty and Ritter (1986)) focus on the role of asymmetric information as the cause of IPO underpricing. If information production by additional underwriters reduces information asymmetry, underpricing would also be expected to shrink. Finally, larger syndicates disseminate information about the offering to a broader group of potential shareholders. Increasing the number of people who know about the stock prior to the offer may generate more demand at the IPO and may also reduce information asymmetries.

A second finding that supports the hypothesis that larger syndicates produce more information is that offer prices are more likely to be updated for IPOs with more co-managers. We use the total return from the midpoint of the filing price range to the close on the first day of trading as a proxy for the information learned about the market’s valuation of an IPO. Some of this information may be learned in time to adjust the offering price and some may only be learned when the security begins trading. We find that the more co-managers in a syndicate the more likely that an offer price will be adjusted away from the midpoint of the filing range in response to information. Although this is consistent with additional co-managers generating more information, underwriters we have spoken with offer a second explanation for this result. Co-managers have the ear of the issuing firm and are competing with the book manager for future underwriting business. If the book manager has mispriced an offering, the co-managers are likely to tell the issuing firm management, and the issuing firm is likely to pressure the book manager to revise the offering price. Hence IPOs with more co-managers are more likely to have price revisions.

Finally, this paper provides the first analysis of why specific underwriters are included in particular syndicates. We find that top ranked analyst coverage and the likelihood of being selected book manager are significant determinants of whether a specific underwriter is included in a syndicate as a co-manager or non-managing syndicate member. We also find that underwriters are more likely to be included in a syndicate if they are in the same state as the issuer and not in the same state as the book manager. Consistent with Pichler and Wilhelm (2001), relationships between underwriters are very important in determining who participates in a syndicate. The single strongest determinant of whether an underwriter is included in a syndicate is participation in recent syndicates led by the same book manager. We hypothesize that reducing agency problems in syndicates may be one reason why ongoing relationships are so important. Business relationships between underwriters may also be important.

We note that our results may not apply to IPO syndicates from other periods. Even during our relatively short sample period, the nature of IPO syndicates changed. From 1997 to 2000, syndicate size decreased significantly even as offer proceeds and aftermarket volatility increased. Over the same period, the mean number of co-managers per offer rose. Finally, many of the smaller underwriters disappeared during our sample period, while the proportion of IPOs underwritten by the largest investment banks increased.

The rest of the paper is organized as follows. In Section 2, we discuss syndicate formation and operation. In Section 3, we review explanations for the practice of syndication and related empirical hypotheses. We describe our data in Section 4 and present our empirical results in Section 5. We summarize our findings and draw conclusions in Section 6.

2. The Formation and Operation of Underwriting Syndicates

Syndicate formation begins with the selection of the book manager (or lead underwriter) by the issuing firm. Competition to be the book manager can be fierce for the largest, most desirable IPOs, as underwriters vie for top positions in underwriting industry rankings and seek the higher fees associated with the lead position. For these large IPOs, issuing firms may have their choice of a number of potential book managers, and general reputation, research support, industry knowledge, and prior relationships are likely to be important factors in determining the book manager. For smaller less-desirable IPOs, issuing firms may not have a choice of book managers. Krigman, Shaw, and Womack (2001) report that many issuers choose a book manager because it is the only one that will underwrite their IPO.

If several underwriters participate in the “bake sale” to be the book manager, the issuer is likely to select some to be co-managers. Co-managers may be chosen because of their ability to provide analyst coverage or market making, or because their distribution system complements the book manager’s. A partner in the law firm of Goodwin Proctor tells us that he advises clients that it is better to “have more banks on your cover than fewer” because the total fees are the same regardless of the number of co-managers while adding managers increases analyst coverage.

Book managers on occasion advise issuers on the best co-managers to include in a syndicate, but seldom provide direct input. They are more likely to indirectly affect the choice of co-managers. Book managers usually set a lower bound on the portion of the fees that they must get to agree to be the book manager and this limits the number of co-managers and their allocations. As one investment banker told us, if we’re the lead, the best number of co-managers is zero. Savvy issuers make sure that the allocations and fees of the book manager and co-manager are negotiated up front.

Both the issuer and the book manager choose non-managing syndicate members. These syndicate members do less work than co-managers, but are also relatively cheap to include. Syndicate members may be included because they have loaned money to the issuer, because they have existing relationships with the customers of the issuer, or because of personal ties between people at the issuing firm and the underwriter. Others get a place in the syndicate because they clear through or purchase research from the book manager. Syndicate members may also be chosen because the book manager and issuer want minority-owned firms participating in the IPO. Despite these reasons for inclusion, the underwriters we have talked to that usually participate as non-managing syndicate members assure us that they are expected to provide analyst coverage and participate actively in selling the IPO.

In a typical IPO, the numbers of shares underwritten varies substantially across syndicate members. As an example, Table 1 provides the syndicate roles and underwriting allocations for the August 10, 1999 IPO of Blockbuster Inc., a large but otherwise typical IPO. In this example, the joint book managers underwrite 37 percent of the shares, the nine co-managers underwrite 55 percent, and the remaining 11 syndicate members underwrite eight percent. Underwriting allocations for co-managers are usually decided up front, but allocations for other syndicate members are at the sole discretion of the book manager and are usually assigned at the due diligence meeting about 48 hours before the IPO becomes effective. These allocations represent the number of shares each syndicate member agrees to buy from the issuer and are listed in the final prospectus. In the analysis to follow, we use these underwriting allocations to measure the distribution of shares within the syndicate and the relative importance of various underwriters within the IPO syndicate.

The number of shares each syndicate member is credited with selling is usually very different from the number they underwrite, and is again at the discretion of the book manager. The Capital Markets Handbook states that a standard practice in recent years has been for the book manager to credit each syndicate member with selling 10 percent of its underwriting allocation. However, the actual allocation for a syndicate member may vary if, for instance, the syndicate member provides research coverage for the IPO.

Typically, all of the shares to be sold to institutional investors are allocated to an “institutional pot.” This central source for institutional sales serves several purposes. First, it allows institutions to establish large positions without placing orders with numerous syndicate members. Second, it gives the book manager a clearer picture of institutional demand. Finally, an institutional pot allows the book manager to ensure that shares are distributed broadly. In general, the book manager does most of the marketing to institutions and receives credit for selling most of the shares in the institutional pot. However, for some IPOs a “jump ball” or competitive pot is used. In these cases, institutions apportion credit for the shares they purchase to one or more of the syndicate members (including the book manager). This provision provides incentives for all syndicate members to exert institutional selling effort. There are potential problems though. First, if an institution has a soft dollar agreement with one of the syndicate members, they may be tempted to direct sales to that underwriter even if the underwriter had little to do with their decision to buy. In addition, the book manager has final say in distributing shares to investors and may shift shares to those institutions that will credit his sales force. Regardless of whether the institutional pot includes a jump ball portion, the book manager is typically credited with selling the majority of institutional shares and co-managers receive more institutional selling credits than non-managing syndicate members.

An underwriter’s role within the syndicate and their share allocation have a substantial impact on the fees they receive. Chen and Ritter (2000) provide an example of the fee distribution within an underwriting syndicate. The gross spread, which is typically 7% of offer proceeds, is divided between management fees, underwriting fees, and selling concessions. Management fees are shared between the book manager and co-managers, with the book manager typically receiving a larger share. Underwriting fees, less any underwriting and stabilization expenses, are shared among all syndicate members according to the number of shares underwritten. Finally, the selling concession, which is typically 60% of the gross spread, is divided among syndicate members based on the number of shares each is credited with selling. Since the vast majority of selling credits are assigned to the book manager, and to a lesser extent the co-managers, this breakdown awards most of the fees to the book manager, with non-managing syndicate members earning only a small fraction of the total fees paid by the issuer.

3. The Role of Syndicate Members

Practitioners and academics suggest several potential roles for syndicate members. In this section, we discuss these roles and derive related testable predictions.

3.1 Analyst Coverage

Among the many services provided by underwriters, aftermarket analyst coverage is perhaps the most often cited. Practitioner guides to going public counsel potential issuers to select underwriters that can provide good analyst coverage and who will continue to follow the stock in the aftermarket (e.g., Cable (2001)). Academic studies also note the importance of analyst coverage. For example, Krigman, Shaw, and Womack (2000) report that issuers cite analyst coverage as an important determinant when selecting underwriters. Chen and Ritter (2000) state that “an implicit understanding is that the managing underwriters of an IPO will each assign a securities analyst to cover the company, produce research reports, and issue buy recommendations for the stock.” In addition, the COO of one smaller underwriter tells us that they are expected to provide aftermarket research coverage (and market making) even as a non-managing syndicate member. If underwriters are included in a syndicate to increase aftermarket analyst coverage, we expect the number of analysts covering a stock to increase with the number of syndicate members. Consistent with this prediction, Chen and Ritter estimate that an incremental co-manager results in 0.36 to 0.55 additional analysts. If the quality of analyst coverage is also important, we expect underwriters with a top ranked analyst to be more likely to be included in a syndicate.[1]

3.2 Market Making

The great majority of IPOs begin trading on Nasdaq, where the lead underwriter almost always acts as the dominant market maker (see Schultz and Zaman (1994) and Ellis, Michaely, and O’Hara (2000)). Though Ellis et al find that co-manager play a negligible role in aftermarket trading, conversations with underwriters suggest that co-managers, and even non-managing syndicate members with market making operations, are expected to make a market in the stock once trading begins. If securities firms that participate in a syndicate are more likely to make a market than others, we would expect larger syndicates to be associated with a larger number of market makers for Nasdaq-listed IPOs. We would also expect larger syndicates for Nasdaq-listed IPOs than for IPOs that initiate trading on either the New York or American Stock Exchanges, where stocks are traded by a single specialist and underwriters do not make markets.

3.3 Certification

Another potential reason for forming syndicates is to provide certification about the quality of the issue. Numerous studies have examined the effects of underwriter (or other expert) reputation on the pricing of IPOs. This research follows from theories that predict a positive relation between IPO underpricing and the level of asymmetric information or uncertainty about IPO value (see, for example, Rock (1986) and Beatty and Ritter (1986)). The certification hypothesis suggests that reputable underwriters (or auditors) are associated with reduced uncertainty and therefore less underpricing.[2]

Empirical support for the certification hypothesis is mixed. Carter and Manaster (1990) and Carter, Dark, and Singh (1998), among others, find that offers taken public by highly reputable underwriters are associated with reduced underpricing. Beatty (1989) finds similar results related to auditor quality, and Balvers, McDonald, and Miller (1988) find evidence of significant reductions in underpricing from both underwriter and auditor quality. Beatty and Welch (1996) analyze the interactions of underwriter, auditor, and lawyer quality in the 1990s. Consistent with previous research, they find that the presence of a high-quality auditor results in significantly lower underpricing in the 1990s. However, they find that the effects of underwriter quality reversed in the 1990s, with high quality underwriters associated with higher, rather than lower, underpricing.

Notably, no previous research has examined the role of the underwriting syndicate in providing certification. We argue that participation by highly reputable syndicate members and especially co-managers may provide additional certification beyond that of the book manager. Participation by a high-quality syndicate member, especially as a co-manager, signals that it agrees with the book manager’s valuation. This certification is meaningful because underwriters may be jointly and severally liable for damage resulting from false or misleading information provided in the IPO registration statement (Beatty and Welch (1996)). An underwriter’s appearance “on the cover” may also signal that it competed to be the book manager for the IPO. If underwriting syndicates provide incremental certification, we expect IPOs with high quality syndicate members to be less underpriced than other IPOs, all else equal. This effect should be strongest for IPOs with highly reputable co-managers.

3.4 Broader Dissemination of Information to Potential Shareholders

Different underwriters have different investor clienteles. For example, some firms, like Merrill Lynch are known for their retail investor base. Others, like Goldman Sachs, specialize in institutional clients. Underwriters also specialize by geographic region. Traditionally, Piper, Jaffrey and Hopwood’s customers were based in the Midwest, while Janney Montgomery Scott’s customers came from the east coast. As a result of these different clienteles, the issuer or book manager may include more syndicate members in order to disseminate information to a larger group of potential investors. Merton (1987) argues that this increased investor recognition results in higher firm value. As one underwriter noted, the IPO is the “coming out party” for the issuer and it is important to introduce the issuer to as many potential investors as possible. Broader dissemination of information may increase demand at the IPO and may also reduce information asymmetries, leading to reduced underpricing. The information dissemination hypothesis also suggests that the issuer or book manager will select syndicate members with different client bases.

3.5 Information Production

In addition to certification, syndicate members may affect IPO pricing more directly through their role in information production. IPO underwriters are faced with the difficult task of assigning a price to a stock with no prior trading history. One underwriter offered to us that pricing an IPO is “part art and part science.” The science is using comparable traded companies to assign a price to the IPO. The art is determining the market’s interest in the offering. Since valuation methods and the choice of comparables are likely to be similar across underwriters, syndicate members are unlikely to be important in the “science” part of the pricing process. Because different underwriters have different investor clienteles, co-managers and syndicate members may be more helpful in the art of determining market interest. These different demand channels all provide the book manager with potential sources of information about IPO value. The more people providing feedback on an IPO, the more accurate is the assessment of market interest. If syndication improves information production, we expect larger syndicates to uncover more information between the filing of the preliminary prospectus and the offer date, resulting in more accurately priced IPOs. The information production hypothesis also suggests that issuers or book managers will select syndicate members with different client bases.

We note that the underwriters we talked to stressed that different book managers worked with co-managers in different ways. Some book managers consult with co-managers often, others ignore them. In many cases, information provided by co-managers is informal and may be obtained while pursuing other aspects of the syndicate’s business. The COO of one smaller underwriter we spoke to noted that even as a non-managing syndicate member, their firm conveys information to book managers about demand from smaller institutional investors and is expected to participate actively in selling. When asked how often they speak to book managers when selling IPOs, she said “about twice a day.”

6. Competition Among Underwriters

Numerous underwriters may compete for the book manager position in desirable IPOs. Issuers typically choose co-managers from among these competing underwriters. Thus, if there is fierce competition to be the book manager, syndicates may be larger at the same time that competition leads to more favorable pricing for the issuer.

Competition among underwriters continues even after selection of the book manager. Co-managers are interested in furthering their own interests, not those of the book manager. While participating in the IPO syndicate, they may be working to secure a place as lead manager in follow-on offerings or other future underwriting business. Conversations with underwriters suggest that co-managers may also influence pricing by “whispering in the issuer’s ear.” For example, a co-manager might tell the issuer that the book manager mispriced the IPO and “we would have done better for you.” This indirect influence by co-managers may put pressure on the book manager to revise the offering price. If competition between underwriters results in improved IPO pricing, we would expect IPOs with more co-managers to be underpriced less than other IPOs, all else equal.

3.7 Preventing Excess Search

Barzel, Habib, and Johnsen (2000) suggest a motivation for including passive syndicate members. In their model, competing underwriters are included in a syndicate to avoid the costs associated with excess search. Excess search occurs when securities firms that are not in the syndicate engage in a costly search for information about an IPO’s value that allows them to profit from aftermarket trades, but does not result in more accurate pricing of the IPO or better placement of the shares. Viewed in this light, the formation of syndicates decreases costly excess search by other underwriters and compensates them for remaining passive.

4. Data and Sample Characteristics

4.1 Data

We collect an initial sample of 1,939 IPOs issued between January 1997 and December 2000 from the Securities Data Company’s (SDC) Global New Issues Database. For each IPO, SDC includes data on offer characteristics, book manager and syndicate member identity, underwriter roles within the syndicate, and share allocations across underwriters. We do not include IPOs that took place prior to 1997, because SDC’s earlier syndicate member data is incomplete.

Detailed data on syndicate participation is missing for 269 IPOs. For these IPOs, syndicate data and underwriter roles are collected from the final prospectus. As a check of the underwriting allocation data, we then examine whether total shares underwritten by all syndicate members equals the total shares offered in the issue (including international shares) as listed in SDC. Discrepancies are evident for 408 IPOs. Where possible, syndicate participation and share allocation data for these IPOs are corrected using information in the final prospectus.

From the initial sample, we exclude issues by non-U.S. firms, investment funds, REITs, and Units. These restrictions reduce the sample by 189, 61, 26, and 65 issues, respectively. We also exclude 50 IPOs for which discrepancies in underwriting allocations and total shares could not be reconciled, and one issue with incomplete share data. The final sample includes 1,547 IPOs.

SDC assigns each underwriter within the syndicate one of six role designations: book manager, joint book manager, joint lead manager, co-manager, syndicate member, or global lead.[3] For IPOs that include shares offered outside the U.S., syndicate participation and underwriting allocations for internationally-offered shares are generally listed separately. In many cases, this results in the same underwriter being listed twice within the syndicate. When this occurs, we combine domestic and international underwriting allocations to determine the total allocations received by each underwriter and the total number of underwriters within the syndicate. In most cases, underwriters listed as global leads are also listed as co-managers allowing us to ignore the global lead designation. We also performed several checks of the SDC co-manager designations to remove potential data errors. These checks resulted in underwriter role corrections for 14 IPOs.[4] A detailed description of checks and corrections of the data is provided in Appendix A.

There are a large number of mergers in the securities industry over our sample period (see Appendix B for a list). We treat underwriters who change names following a merger as new firms. So, for example, we examine Morgan Stanley Dean Witter separately from either Morgan Stanley or Dean Witter and we study U.S. Bancorp Piper Jaffrey separately from Piper Jaffrey and Hopwood. Our reasoning is that different clienteles and financial capabilities after mergers may change the motives and characteristics of syndicates. It is likely that some underwriters enter or exit during our sample period for reasons unrelated to mergers. In the analysis to follow we examine the robustness of the results to this possibility.

We collect data on aftermarket analyst coverage from IBES. We first identify the number of analysts issuing reports on each IPO firm within three months of the offer. As a measure of the quality of an underwriter’s analyst coverage, we then identify those underwriters that employ an Institutional Investor (henceforth II) All-Star analyst for the IPO’s industry. Every October, II names the top three analysts in each of several dozen industries. For each of these industries, we obtain the identities of the All-Star analysts and their underwriter affiliations from 1996 through 2000. We then assign each IPO to one or more II industries.

Deciding which II industries correspond to a particular IPO can be difficult, since II industries do not correspond directly to SIC codes and analysts that II places in different industries often follow the same stocks. As a first step in assigning IPOs to industries, we identify 10 stocks that II mentions from each of their industries. If fewer than 10 are found, we use all stocks mentioned in descriptions of the industry. We then output IBES industry descriptions for each of the firms identified in each of the II industries. Finally, we match the IBES industry description for each IPO with the IBES industry descriptions corresponding to the II industries. Using this methodology, we are able to match 1,248 IPOs with II industries. The remaining IPOs are assigned to industries by hand after examining descriptions of their business. IBES industry descriptions often correspond to more than one II industry. As a result, 610 of the IPOs fit into one II industry, 321 are classified in two industries, and 598 fit into three or more industries. Eighteen of the IPOs defy classification into any of the industries considered by II (e.g. soil preparation). An underwriter is classified as having a top analyst for an IPO if it employs any of the top three II analysts in any of the industries into which the IPO is classified.

4.2 Summary Statistics

Panel A of Table 2 provides descriptive statistics for the sample IPOs. On average, sample IPOs raised $96.5 million. The mean underpricing is 42.1 percent. While large by historical standards, this level of underpricing is similar to that found by Loughran and Ritter (2002) for the same period. The median adjustment of the offer price from the filing range midpoint is zero, suggesting that the midpoint of the filing price range is a useful forecast of the offer price. As a measure of risk, we estimate the standard deviation of continuously compounded daily returns from day 21 through day 125 after the offer. The mean aftermarket standard deviation in the sample is 5.94 percent.

Panel B of Table 2 provides descriptive statistics for underwriting syndicates. The mean number of underwriters in a syndicate is 16.0, of which 1.03 are book managers and 1.93 are co-managers. On average 41 percent of the shares are underwritten by book managers and another 38 percent are underwritten by co-managers. This leaves only 21 percent of the shares to be underwritten by the many other syndicate members. As an additional measure of syndicate structure, we calculate a Herfindahl Index of underwriting allocations within the syndicate. This index is defined as the sum of the squared percentages of shares allocated to each underwriter. The mean and median values of the syndicate Herfindahl index are 29.26 and 24.42, respectively. For comparison, if shares were distributed equally among four underwriters, the Herfindahl index would be 25.00.

Although not shown, we also examined syndicate structure by investment bank for those that led at least 20 sample IPOs. These large underwriters account for 63 percent of sample offers. Syndicate structures are very similar across book managers. There are two notable exceptions. While most syndicates average 15 to 20 members, Donaldson, Lufkin and Jenrette (DLJ) syndicates average 32 underwriters. In contrast, Fleet Robertson Stephens averages only 6.9 underwriters in their syndicates while the firm’s other incarnations, BankBoston Robertson Stephens and BankAmerica Robertson Stephens, average 8.9 and 7.3 underwriters per syndicate.

Table 3 provides information on offer and syndicate characteristics by year. Even though our sample period covers only four years, there are significant differences in the IPO and syndicate characteristics over time. Offer proceeds average $58 million in 1997, compared to $125 million in 1999 and $123 million in 2000. In addition, the aftermarket standard deviation increases steadily from 4.09 percent in 1997 to 7.95 percent in 2000. Mean underpricing is 14.1% in 1997 and 19.8% in 1998, as compared to 72.2% in 1999 and 55.9% in 2000.[5] An underwriter’s prestige is measured with its Carter - Manaster ranking. This is obtained by examining where the underwriter is listed in tombstone ads. An underwriter with a ranking of nine, the highest ranking, is never listed below anybody except book managers and co-managers. An underwriter with a ranking of eight is only listed below those with rankings of nine, etc. We identify the Carter-Manaster ranking of each syndicate member using data from Jay Ritter’s website, . The average Carter – Manaster ranking for book managers increases steadily over the sample period from 6.95 in 1997 to 8.34 in 2000. This indicates that prestigious underwriters led a higher proportion of IPOs as the sample period progressed. Panel B shows that there are also significant changes in the composition of syndicates over time. The mean number of co-managers increased steadily from 1.42 in 1997 to 2.43 in 2000. At the same time, the total number of underwriters fell from 18.0 in 1997 to 14.6 in 2000. All co-managers together were allocated an average of 33.2 percent of offered shares in 1997 and 43.1 percent of offered shares in 2000.[6]

Different underwriters may have different clienteles. Some may underwrite only large offers while smaller underwriters may take smaller firms public. To examine offering size clienteles, we divide each year’s offerings into quintiles by the amount that they raise. We then select the five book managers who raise the largest amount of money from offerings in each size quintile. Table 4 lists the top five underwriters by offer size quintile. There are some differences in underwriter identities. Lehman appears as a top underwriter only among the three smallest offering quintiles while Salomon Smith Barney is only listed for the largest IPOs. Nonetheless, there is a great deal of consistency across offering sizes. CS First Boston is among the top five underwriters in each offering size quintile. Goldman Sachs is a top underwriter for size quintiles two, four and five.

To provide a more complete description of syndicate characteristics during our sample period, Table 5 reports regressions of syndicate characteristics on offering proceeds and interaction terms between proceeds and the aftermarket volatility, an exchange listing dummy, and year dummies for 1998, 1999, and 2000. Five measures of syndicate size or concentration are used as dependent variables: the number of underwriters in the syndicate, the number of co-managers, the proportion of shares underwritten by the book manager, the proportion of shares underwritten by the co-managers, and the Herfindahl index for the IPO syndicate. Poisson regressions are run when the dependent variable is a count, that is number of underwriters or number of co-managers. Ordinary least squares regressions are used for the other dependent variables.

As the table shows, offering proceeds is highly significant in each of the regressions. The number of underwriters or co-managers increases with the amount raised while the Herfindahl index and the proportion of shares underwritten by the book manager decrease. The regressions in Table 5 also confirm that syndicates have changed over our sample period. After controlling for offering size and aftermarket volatility, the number of underwriters in syndicates declines steadily over the sample period while the number of co-managers increases. It is also interesting that neither the number of underwriters nor the proportions of shares underwritten by the book manager are affected by aftermarket volatility. Thus, we find no evidence that riskier offers are handled by larger syndicates, a prediction of the traditional risk-sharing explanation of syndicates.

5. Results

5.1 Syndicate Structure and IPO Pricing

To examine whether IPO pricing is related to syndicate characteristics, we regress IPO underpricing on measures of syndicate size and a set of control variables identified from previous studies of IPO underpricing. The control variables include the offer proceeds, the natural log of the offer proceeds, a dummy variable for venture capital backing, the standard deviation of aftermarket returns measured from days 21 through 125 after the IPO, a dummy variable for a book manager with a Carter-Manaster ranking of eight or more, and year dummy variables to capture the change in underpricing over time that Loughran and Ritter (2001) document. We also include dummy variables for industries defined by two-digit SIC codes 28, 35, 36, 38, 48, 73 and 87. These are the industries with at least 40 sample IPOs and more than two-thirds of the IPOs in our sample are in one of these industries.

The regression results are shown in Table 6. In the first regression, the dependent variable is underpricing. Independent variables include the numbers of co-managers with Carter-Manaster rankings above and below eight and the numbers of non-managing syndicate members with Carter-Manaster rankings above and below eight. The coefficient estimates reveal that underpricing falls with additional syndicate members of all types, but that co-managers are associated with a greater reduction in underpricing than non-managing syndicate members, and that syndicate members with high Carter-Manaster ranks are linked with a greater reduction in underpricing than syndicate members with low Carter-Manaster rankings. Specifically, each low-ranked non-managing underwriter reduces underpricing by .7 percent and each high-ranked non-managing underwriter reduces underpricing by 1.6 percent. Each additional co-manager with a Carter-Manaster ranking below 8 is associated by a reduction in underpricing of 6.0 percent, while each additional co-manager with a rank above eight is associated with a reduction in underpricing of 7.5 percent. To save space, industry and calendar time dummies are not shown, but adjusted R2’s are reduced more than 50 percent when they are omitted.

The next regression in the table is identical to the first except that the dependent variable is the natural logarithm of one plus underpricing. Results are qualitatively unchanged, but the influence of outliers is reduced by this specification, so the R2 and most of the t-statistics increase. Hence the rest of the regressions in the table use the natural logarithm of one plus underpricing as the dependent variable.

For our purposes, the most important result in Table 6 is that a large syndicate is associated with lower underpricing. Separate regressions in the table show that underpricing increases as the syndicate’s Herfindahl index increases or as the percentage of shares underwritten by the book manager increases. That is, concentration of shares in the hands of a few underwriters is associated with large underpricing. The last regression of the table shows that underpricing decreases with the number of underwriters participating in the underwriting syndicate.

As we observed earlier, the size and composition of syndicates as well as the characteristics of firms going public changed over our sample period. This raises the question of whether the relation between syndicate size and underpricing also changed. As a robustness check, we reran the regression in Table 6 that included the number of underwriters in the syndicate as an explanatory variable for each calendar year of the sample period (not shown). The coefficient on the number of underwriters has the expected negative sign and is significant at the one percent level for three of the four years: 1997, 1999 and 2000. We also reran the regressions where the syndicate members are broken down into co-managers ranked above and below eight and non-managing members ranked above and below eight. The coefficient on co-managers ranked below eight is negative and significant for 1999, while the coefficient on co-managers ranked above eight is significant for 2000. The number of non-managing underwriters ranked below eight is significant in 1997 and 2000. As a whole, this suggests to us that larger syndicates are associated with lower underpricing throughout the sample period.

5.2 Syndicate Structure and Offering Price Revisions

The smaller underpricing associated with larger syndicates could be a result of competition to manage the offering, additional certification, or improved information production. In this section we test more directly for information production by syndicate members. We use revisions in the expected offer price as an indicator of information production during the filing period. Our proxy for the expected offer price is the midpoint of the initial filing range, which we henceforth refer to as the filing price. As shown in Table 2, the median revision from the filing price to the offer price is zero in our sample, suggesting that the filing price provides a useful estimate of the expected offer price. Our measure of the information revealed during the filing period and on the offer day is the total return from the filing price to the closing price on the first day of trading. If a syndicate successfully uncovers information during the filing period, we would expect part of this total return to be reflected in an offer price revision.[7] We use a probit model to test whether the likelihood of an offer price revision is related to syndicate structure, after controlling for the total information revealed. Note that we are not necessarily saying that the book manager consults with other syndicate members on the IPO’s pricing. The other syndicate members may instead influence pricing through their conversations with the issuer.

In the first set of probit regressions, shown in Panel A of Table 7, the dependent variable equals one if the offer price is greater than the filing price and zero otherwise. Independent variables include the total return from the filing price to the first day’s close if positive, a dummy variable that equals one if the total return is zero or negative, the natural log of expected offer proceeds, and several measures of syndicate structure and underwriter quality. Using separate variables for positive and negative first day returns allows for asymmetric effects on the likelihood of an upward price revision. We use a dummy for a total return that is zero or negative because, as Ruud (1993) shows, aftermarket support by underwriters constrains prices from falling below the offer price. Expected offer proceeds are estimated by multiplying the filing price by the actual number of shares issued.

The first probit regression in Panel A includes only the return variables and the log of the expected proceeds. As expected, the coefficient on positive total return is positive and highly significant, while the coefficient on the negative return dummy is negative and significant. That is, if the total return from the filing price is large and positive, at least part of the return is likely to be reflected in an upward revision of the offer price. If the total return is negative, an upward revision in the offer price is much less likely.

The probit regressions reported in the following columns of the table include interactions between positive total returns and measures of syndicate structure. In the second regression of the panel, the coefficient on the interaction term between the number of underwriters in the syndicate and the positive total return is negative and insignificant. Thus increasing the number of underwriters in the syndicate does not make it more likely that a given positive total return will lead to an upward revision in the offer price.

The next regression includes separate interaction variables for high and low ranked co-managers, and high and low ranked non-managing underwriters. This regression reveals that an offer price is more likely to be revised upward when additional co-managers of either high or low ranking are included, but that non-managing syndicate members do not affect the likelihood of revision. It is interesting that the dummy variable for a high ranked book manager is negative and significant. Perhaps co-managers influence the pricing of IPOs through their conversations with issuers, but issuers are less likely to pay attention to co-managers if the book manager is highly ranked.

The next probit regression in the table includes an interaction between positive total returns and the IPO’s Herfindahl index. Both of these terms are negative and significant. Concentrating the shares in the hands of the book manager or a few syndicate members makes it less likely that an offer price will be adjusted. The last probit regression includes interactions between a positive total return and the proportion of shares underwritten by co-managers and the proportion underwritten by book managers. Neither term is significant, but the coefficient on the book manager’s share is negative while the coefficient on the proportion underwritten by the co-manager is positive.

Panel B replicates the probit model of Panel A except that the dependent variable is now set to one if the IPO offer price is revised downward from the midpoint of the filing range. Here the variables of primary interest are the interactions between a dummy variable for a total return that is zero or negative and measures of syndicate size. The results are consistent with those in Panel A. A positive return from the filing price to the aftermarket close makes a downward revision much less likely. A negative or zero return is more likely to result in a downward price revision if there are a lot of syndicate members, and particularly if there are a lot of co-managers. It is less likely to lead to a downward revision if the book manager is underwriting a large percentage of the shares or if the Herfindahl index for the IPO is high.

Given the changes in syndicates and IPOs over 1997 – 2000, it is reasonable to ask whether syndicate size affects price discovery in the same way over our entire sample period. As a robustness test, we re-estimate the probit models for positive and negative price revisions separately for each year (not shown). In the probit model with a positive price revision as the dependent variable, the interaction term between a positive total return and the number of co-managers ranked eight or higher is positive each year. It is statistically significant at the ten percent level in 2000 and at the five percent level in 1998 and 1999. The interaction between a positive total return and the number of co-managers ranked below eight is also positive each year, and is statistically significant for 1998. In the probit regressions with a negative price revision as the dependent variable, the interaction between the negative return dummy and the IPO’s Herfindahl index is negative each year. It is significant at the ten percent level in 2000, and at the five percent level for other years.

To summarize, the likelihood that an offering price will be revised in response to new information increases when a syndicate is larger and shares are more broadly distributed across underwriters. This is true throughout our sample period, despite changes in the composition of syndicates and the type of firms going public. These results are consistent with syndicates generating information about IPO values prior to the pricing of offers. Prices may be revised because book managers consult with co-managers, but that need not be the case. It is also possible that co-managers may influence prices through their conversations with issuing firms.

5.3 Syndicates and Aftermarket Services

A Nasdaq dealer may be more likely to make a market in a stock if its brokerage customers are familiar with the company and own shares in the stock. Likewise, an investment banker’s analysts are more likely to issue reports on a stock if its clients own shares. Participating in an underwriting syndicate ensures that the firm’s clients own shares in the IPO and may therefore provide incentives for underwriters to act as market makers and to provide analyst coverage.

To test whether large underwriting syndicates are associated with more market makers, we obtain the first number of market makers on the CRSP tape following Nasdaq IPOs. We then regress the number of market makers on the number of underwriters or co-managers and several control variables: offer proceeds, the natural log of offer proceeds, industry and date dummy variables, underpricing and the standard deviation of aftermarket returns. Table 8 reports results from two types of regressions. First, we report ordinary least squares (OLS) regression estimates. The advantage of these regressions is that coefficients are easy to interpret. We also report Poisson regression estimates. These regressions are less intuitive but better specified, since a Poisson regression explicitly incorporates the discrete nature of the dependent variables as well as the higher residual variance that accompanies a higher predicted number of market makers. Nevertheless, the results reported in Panel A of Table 8 are essentially the same for the Poisson and OLS regressions

Not surprisingly, the number of market makers in the aftermarket increases with the size of the offer and the amount of underpricing. The number of market makers also increases with the aftermarket standard deviation. After controlling for these effects, the number of underwriters in a syndicate does not appear to affect the number of market makers. However, the number of co-managers is significantly related to the number of market makers. Our estimates suggest that every additional co-manager increases the number of market makers by 0.7.

In Panel B, we examine how syndicate size affects the number of analysts issuing reports on a company. Our dependent variable in these regressions is the number of analysts listed on IBES that issued reports on an IPO within three months of the offer date. As in Panel A, we report results from both OLS and Poisson regressions that control for the offering proceeds, the log of the offering proceeds, the standard deviation of aftermarket returns, underpricing, and date and industry dummies. After adjusting for these factors, the number of analysts, like the number of market makers, is unaffected by the total number of underwriters but increases significantly with the number of co-managers. Our OLS estimates suggest that, all else equal, each additional co-manager increases the number of analysts following a stock by 0.69, an estimate similar to that of Chen and Ritter (2000). The coefficient on the number of co-managers is highly significant with a z-statistic of 9.88 in the Poisson regression. Overall, the results in Table 8 provide strong support for the hypothesis that including co-managers in a syndicate increases aftermarket analyst coverage and market making. In the next section, we incorporate the quality of analyst coverage and test whether having a top-ranked analyst increases the likelihood of being included in a syndicate.

5.4 Syndicate Participation

To this point, we have focused on the relation between IPO characteristics and the size and structure of syndicates. In this section we attempt to shed additional light on the function of syndicates by examining which underwriters are included in a particular deal. Naturally, we would expect underwriter characteristics to play a big part in syndicate inclusion. For example, we expect that having a top ranked analyst will increase the likelihood that a firm will be included in a syndicate. Likewise, we expect that underwriters who compete to be the book manager will be to be more likely included in the syndicate at the issuer’s request.

We also expect relationships between underwriters to play a role in syndicate inclusion. One Underwriter told us that they are included in syndicates managed by specific underwriters who “know what we can do.” Relationships may also serve to mitigate agency problems. Underwriters complain that co-managers do not work with them and may undermine their efforts with issuing firms[8]. The possibility of future syndicate work may provide the incentives for members of the syndicate to work together.

We estimate probit models to determine how underwriter characteristics affect the probability that they will be included in an IPO syndicate. In estimating the probit model, we use all 1,547 IPOs and include one observation for every eligible underwriter for each IPO. This results in approximately 900,000 observations. We examine two alternative dependent variables. The first is set to one if a particular underwriter is included in the IPO syndicate and zero otherwise. The second is set to one if a particular underwriter is included as a co-manager in the IPO syndicate and zero otherwise. The independent variables include offer characteristics, underwriter size and quality measures, geographic characteristics, and measures of relationships between underwriters. These variables and the related results are discussed in more detail below.

Results for the probit model are presented in Table 9. Columns 1 and 2 list results for syndicate inclusion and columns 3 and 4 list results for co-manager inclusion. Standard errors for the coefficients are listed in parentheses below the coefficient estimates. Given the large sample size, it is not surprising that nearly all of the coefficient estimates are statistically significant. Never the less, the relative significance differs markedly across variables as illustrated by the standard errors.

The results for offer characteristics and year dummy variables are consistent with results from the previous sections. The year dummy variables indicate that the likelihood of being included in a syndicate declines steadily from 1997 to 2000, while the likelihood of being included as a co-manager increases steadily during this period. These results confirm our earlier finding that syndicates have shrunk over the sample period, while the average number of co-managers has increased. The coefficient on offer proceeds is positive and significant in both the syndicate and co-manager models, suggesting that underwriters are more likely to be included in a syndicate for larger offers. This result is consistent with the finding that larger offers are associated with more syndicate members and more co-managers. The coefficient on aftermarket standard deviation is negative and significant in the syndicate model and insignificant in the co-manager model. These results are consistent with earlier findings and provide little evidence of risk sharing as an important determinant of syndicates.

The next three variables reflect underwriter characteristics. We find that underwriters with high levels of capitalization are more likely to be included in a syndicate and regional underwriters are less likely to be included. Not surprisingly, the regional underwriter variable appears to be highly correlated with underwriter capitalization and the sign on the regional dummy is actually reversed when both variables are included simultaneously. Unfortunately, capitalization data is available for less than half of our sample underwriters.

If issuers choose syndicate members from the investment bankers who were in competition to manage the IPO, we would expect a book manager’s main competitors to appear in an IPO syndicate. To test this hypothesis, we include in the probit model an estimate of each underwriter’s ex-ante probability of being selected as book manager. For each underwriter that leads five or more IPOs, we estimate a probit model for the probability of being book manager using the full sample of 1,547 IPOs. The independent variables are offer proceeds, the log of the offer proceeds, aftermarket standard deviation, a dummy variable for whether the underwriter is in the same state as the issuer, and quarterly dummy variables. This estimation results in an estimated probability of being book manager for each underwriter for each IPO. Underwriters who were book manager for fewer than five IPOs during the sample period are assigned a book manager probability of zero.[9]

The probability of being selected as the lead is a highly significant determinant of whether an underwriter is included in a syndicate. One interpretation of this result is that issuing firms often ask the lead underwriter to include in the syndicate other underwriters who had vied for the lead position. In addition, lead probability is highly correlated with other measures of underwriter size that have already been shown to be important determinants of syndicate inclusion.

The results from the previous section suggest that aftermarket analyst coverage is an important determinant of syndicate structure. As an additional test of this hypothesis, we include in the probit model an indicator of whether an underwriter employs one of the top-rated analysts for the IPO’s industry. For each IPO industry, we identify the top-rated analysts using Institutional Investor’s (II) All-Star analyst ratings. The dummy variable Top 3 Analyst is set to one for any underwriter who employs one of the top rated analysts in the IPO’s industry. We also interact this variable with a dummy variable that equals one if the book manager also has a top-rated analyst. As expected, we find that an underwriter with an analyst on II’s first, second or third team is more likely to be included in an underwriting syndicate. However, the coefficient on this variable is reversed when underwriter capitalization is included in the model. This reflects the fact that underwriters with top-rated analysts also tend to be the largest underwriters in the sample. We provide additional evidence related to analyst rankings below.

Our earlier tests provide evidence that syndicate members, or at least co-managers, may play a role in information production. As an additional test of the information production hypothesis, we include in the probit model several geographic variables to reflect differences in customer bases between underwriters. The first two dummy variables identify underwriters who are in the same state as the book manager or an adjacent state. The next two variables identify underwriters who are in the same state as the issuer or an adjacent state. Finally, the fifth variable identifies underwriters who are in the same state as the issuer and are also in a different state from the book manager. Alternatively, underwriter location may proxy for an underwriter’s ability to disseminate information to investors in different geographic areas, and thus create greater overall demand for an issue.

The results suggest that being in the same state as the book manager decreases the likelihood of being included in a syndicate. This is consistent with the information production hypothesis and suggests that underwriters are more likely to be included in a syndicate if they do not duplicate the customer base of the book manager. However, a much more significant determinant of syndicate inclusion is whether an underwriter is located in the same state as the issuer or an adjacent state. Underwriters located near an issuer are far more likely to be included in the syndicate. This is especially true if the book manager is not based in the issuer’s state. These results support the hypothesis that underwriters are added to syndicates to gather information about demand for an IPO, as a disproportionate amount of the demand is expected to come from investors located near the issuer. Alternatively, the importance of underwriter location may reflect the book manager’s desire to include syndicate members who can distribute shares to investors with a preference for investing in nearby companies.

Although the results are not shown, we also examined whether the customer bases of the book manager and potential syndicate member affect the likelihood of syndicate inclusion. We obtained data on the number of institutional and retail representatives for each underwriter from the Securities Industry Yearbook. We classified underwriters as institutional if they have at least as many institutional as retail representatives. We classified underwriters as retail if they have at least four times as many retail as institutional representatives. All other underwriters are classified as mixed. After controlling for other factors, we find no evidence that book managers attempt to include underwriters with different client bases. This lack of results may reflect the inherent difficulty in classifying underwriters as retail or institutional. For example, Merrill Lynch and Morgan Stanley Dean Witter are classified as retail, but both have large institutional as well as retail businesses.

The final four variables in the table provide proxies for the strength of relationships between the book manager and potential syndicate members. As noted above, ongoing relationships may serve to mitigate agency problems within syndicates when syndicate members are expected to actively participate in acquiring information and distributing shares.[10] We define four relationship variables. The first two variables are the proportion of the book manager’s last ten syndicates in which the underwriter participated and a dummy variable that takes a value of one if the underwriter was included in the book manager’s most recent IPO syndicate. The remaining two variables are the proportion of the underwriter’s last ten syndicates in which the book manager participated as a syndicate member (reciprocal participation) and a dummy variable for whether the book manager was included in the underwriter’s most recent IPO syndicate. If relationships are an important determinant of syndicate participation, underwriters who have previously worked with the book manager should have a higher likelihood of syndicate inclusion.

We find that the single most important determinant of whether an underwriter is included in the syndicate is the proportion of the book manager’s previous ten syndicates in which the underwriter participated. Reciprocal relationships, or whether the underwriter included the book manager in its recent syndicates, are also very important. Notably, excluding the relationship variables from the model results in a decrease in the pseudo-R2 from 0.2895 to 0.1424. This is consistent with ongoing relationships mitigating agency problems within the syndicate. It is also consistent with syndicate inclusion being part of a broader set of links between underwriters. An underwriter may be included in a syndicate because it buys research from the book manager or clears through the book manager.

The importance of relationships in determining syndicate inclusion is not surprising. In contrast, the role of relationships in picking co-managers is somewhat puzzling. Underwriters tell us that issuers select co-managers, and thus relationships would seem irrelevant. It is possible that our variables for relationships are correlated with measurement errors in, for example, the probability of being selected as book manager. On the other hand, underwriters have told us that they may discuss potential co-managers with issuers and may provide advice on which would contribute the most to the syndicate. In this case, relationships would be an important determinant of co-manager roles. In addition, both book managers and co-managers have a sort of veto power – each specifies the fee and share structure needed for their participation.

It seems likely that the effects of analyst coverage, capitalization, underwriter location and other variables will differ for offerings of different size. To examine this possibility, we estimate the probit model separately by offer size quintile. We list results for the first, third, and fifth quintiles in Table 10. Panel A lists results for syndicate participation, while Panel B lists results for co-manager participation. In general, the conclusions are similar across IPO size categories although the model has better explanatory power for large IPOs than for small IPOs. Several specific results deserve mention, however. First, in both the syndicate and co-manager models for large IPOs, the coefficient on a high Carter-Manaster rank is a positive and significant while the coefficient on the regional underwriter dummy is negative and significant. Thus, low-ranked underwriters and regional underwriters are less likely than other underwriters to be included in the syndicates of large IPOs. Secondly, while the probability that an underwriter will be the book manager has a positive effect on syndicate inclusion for small and medium sized IPOs, this effect is reduced or eliminated for large IPOs. Finally, having an All-Star analyst has little effect on the likelihood of being included in a syndicate for a small offer, but significantly increases the likelihood of being included in a large IPO’s syndicate even after controlling for underwriter rank and capitalization. The effects of other variables are the same across offer size quintiles. In particular, prior relationships and geographic location are important and the probability of being included in a syndicate declines over time for all offer size categories.

One potential problem with these probit models is that some of the smaller underwriters may have entered or disappeared during the sample period for reasons other than mergers. As a result, an underwriter could be included in the probit regression for a specific IPO although they were not in business at the time. As a robustness check, we reestimated the probit models (not shown) including only those underwriters who appeared in syndicates during both the first half of 1997 and the last half of 2000. The coefficients are almost unchanged, suggesting that our results are not seriously affected by entry and exit of underwriters. Of particular interest is that the likelihood of being included in a syndicate declines over time even for underwriters who participated during the entire sample period. Finally, although not shown, we estimated the model separately for each sample year. While there are minor differences year by year, the main results are confirmed. Perhaps the most interesting difference across years is that the coefficient on participation in the previous ten syndicates increases steadily from 1.83 in 1997 to 3.21 in 2000. Thus, the importance of relationships appears to have increased over our sample period. [11]

5.5 Specific Underwriter Relationships

We have shown that one of the most important predictors of whether an underwriter is included in a syndicate is a prior working relationship with the book manager. In this section, we look more closely at the role of relationships by examining how often specific pairs of underwriters work together.

Table 11 reports the percentage of a book manager’s syndicates that included other specific underwriters. Panel A lists pairings for the ten underwriters who led the largest number of syndicates. Each column corresponds to a book manager and each row corresponds to a syndicate member. So, for example, in the DLJ (fifth) column and the Goldman Sachs (first) row, we see 53.41. This means that Goldman Sachs was included in 53.41 percent of the IPO syndicates led by DLJ. For comparison, the last column gives the percentage of all IPOs that included the underwriter as a syndicate member. As an example, Goldman Sachs is included in 18.62 percent of all syndicates during the sample period.

A percentage appears in bold characters in the table when a Chi-square test indicates that the probability that an underwriter appears in a book manager’s syndicates is significantly different from the overall percentage of syndicates in which the underwriter appears. The large number of book manager - syndicate member pairs that occur much more or much less frequently than expected confirms the importance of relationships. Specific examples are illuminating. Goldman Sachs appears in 53.4 percent of DLJ’s syndicates but less than 9 percent of Hambrecht and Quist’s and none of Fleet Robertson’s syndicates. CS First Boston appears in 22.6 percent of Goldman’s IPO syndicates and 46.3 percent of Salomon’s. For Hambrecht and Quist the pattern is just the opposite. They are included in 46.4 percent of Goldman’s IPO syndicates, but only 17.7 percent of Salomon’s.

If disseminating information across different clienteles is an important reason for forming syndicates, we would expect the issuers who employ institutional investment banks as book managers to seek out retail underwriters for their syndicates and vice versa. In constructing Table 11, we categorized the top ten book managers by their brokerage clienteles. Goldman Sachs, CS First Boston, Lehman Brothers and Fleet Robertson are classified as institutional. Retail underwriters include Morgan Stanley Dean Witter, Merrill Lynch, and Salomon Smith Barney. The remaining firms, DLJ, Bear Stearns, and Hambrecht and Quist are classified as mixed. If issuers (or book managers) select syndicate members with different clienteles, we would expect to see high frequencies in the “northeast” and “southwest” corners of the table. Instead, Table 11 provides no evidence that brokerage clienteles are an important determinant of underwriting relationships. In fact, relationships seem strongest between mixed clientele firms and others. As before, the lack of results may reflect the difficulty associated with classifying underwriters as institutional or retail.

Panel B of Table 11 repeats the analysis of Panel A, but for combinations of underwriters who led the most offers (shown in the columns) and underwriters who participated in the most syndicates but were never the book manager. Here evidence of relationships seems, if anything, stronger. For example, Edward Jones participated in 73.3 percent of Morgan Stanley Dean Witter’s IPO syndicates, but only 14.7 percent of Merrill Lynch’s and 5 percent of IPOs led by CS First Boston. E*Offering participated in 40.3 percent of CS First Boston’s IPOs and 26.1 percent of Fleet Robertson’s, but only 2.3 percent of IPOs brought to market by DLJ. Again, client base does not seem to be an important determinant of syndicate inclusion. Edward Jones, a retail broker, has strong relationships with both retail and institutional book managers. Wasserstein Parella, an institutional broker, has strong relationships with institutional, retail, and mixed brokerage clientele firms

As an additional illustration of relationships, Table 12 lists book manager and syndicate member pairs in which the book manager underwrote at least ten IPOs and the syndicate member participated in at least 85 percent of the book manager’s IPO syndicates. Again, underwriters are categorized as institutional, retail or mixed. Significant relationships are observed for all three categories of book manager, but in most cases the syndicate member is a retail broker. Montgomery Securities, and its’ successor NationsBanc Montgomery, included H.C. Wainwright, Cruttendon Roth, and Alex Brown in almost all of their syndicates. Morgan Keegan included Stephens, Interstate Johnson Lane, J.C. Bradford, J.B. Hilliard and Robinson Humphrey in almost all of their syndicates during the sample period. A particularly interesting case is CS First Boston and Invemed Associates. Invemed was included in 90 percent of the 119 IPO syndicates led by CS First Boston. They appeared in less than 0.4 percent of syndicates led by other investment bankers. An underwriter tells us that Invemed and Credit Suisse First Boston have an agreement whereby Invemed brings their underwriting business to CSFB in exchange for the right to participate in CSFB’s syndicates.

These results serve to emphasize the importance of relationships. Book managers appear with the same syndicate members again and again. These relationships in some cases reflect other business linkages between underwriters. For example, an underwriter may appear in a syndicate of a particular book manager if it clears trades or purchases research through them. In addition though, these continuing relationships may serve to limit opportunistic behavior on the part of the book manager. For example, syndicate members may avoid participation in a book manager’s future offers if the book manager cuts back on selling credits after the syndicate members have expended time and effort to sell the offer. Likewise, a syndicate member who fails to provide market making or analyst coverage following an IPO or disseminate it may not be included in the book manager’s future syndicates.

6. Conclusions

Almost all U.S. IPOs are sold by underwriting syndicates that consist of a book manager, one or more co-managers, and several non-managing underwriters. Despite their importance, there has been almost no empirical research on the role and composition of IPO syndicates. In this paper, we use data on more than 1,500 IPOs from 1997 to 2000 to test hypotheses about the role of IPO syndicates.

Performing aftermarket services seems to be one important role of syndicate members. We find that adding co-managers to an IPO syndicate increases both the number of market makers and the number of analysts issuing reports in the aftermarket. In contrast, syndicate members other than co-managers seem to have little effect on either analyst coverage or market making after controlling for other offering characteristics. We also find that underwriters who can provide coverage by a top-ranked analyst are more likely to be included in an IPO syndicate.

We find that syndicate members, and particularly co-manager, have statistically and economically significant effects on IPO pricing. Each additional non-managing syndicate member decreases underpricing by 0.7 to 1.6 percent, while each additional co-manager decreases underpricing by 6.0 to 7.5 percent. We also find that prices are far more likely to be revised in response to information when there are more syndicate members, and particularly more co-managers.

In forming syndicates, relationships are critical. An underwriter is much more likely to be included in a syndicate if it has appeared in past syndicates put together by the same book manager. In fact, some underwriters appear in virtually every IPO syndicate led by a particular underwriter. An underwriter is also more likely to be included in a syndicate if the book manager received shares in the underwriter’s recent deals. While some relationships appear to reflect business ties between underwriters, we argue that these ongoing relationships serve to minimize free-riding and moral hazard problems in syndicates when members are expected to actively participate in information production and in marketing the IPO.

Finally, we find that even over our short four-year sample period, syndicates have changed dramatically. Despite the fact that IPO proceeds and aftermarket volatility increased over 1997 through 2000, syndicates grew smaller. At the same time, the number of co-managers increased. One explanation for this change is that the importance of co-managers in producing information or providing aftermarket services has increased over time. This could reflect a change in the nature of the firms going public over our sample period. IPOs in 2000 were younger, riskier, and more volatile firms than 1997 IPOs. Alternatively, underwriters have told us that issuers have become more sophisticated and now demand more co-managers. This change in syndicate structure over time is an interesting area for future research.

References

Aggarwal, Reena, 2000, Stabilization activities by underwriters after initial public offerings, Journal of Finance 55, 1075-1103.

Balvers, R. J., B. McDonald, and R. E. Miller, 1988, Underpricing new issues and the choice of auditor as a signal of investment banker reputation, Accounting Review 63, 605-622.

Barzel, Yoram, Michel Habib, and D. Bruce Johnsen, 2000, IPO syndicates, private foreknowledge, and the economics of excess search, Working paper, George Mason University School of Law.

Beatty, Randolph P., 1989, Auditor reputation and the pricing of initial public offerings, Accounting Review 64, 693-709.

Beatty, Randolph P., and Jay Ritter, 1986, Investment banking, reputation, and the underpricing of initial public offerings, Journal of Financial Economics 15, 213-232.

Beatty, Randolph P., and Ivo Welch, 1996, Issuer expenses and legal liability in initial public offerings, Journal of Law and Economics 39, 545-601.

Benveniste, Lawrence, and Paul Spindt, 1989, How investment bankers determine the offer price and allocation of new issues, Journal of Financial Economics 24, 343-361.

Booth, J. R., and R. Smith, 1986, Capital raising, underwriting and the certification hypothesis, Journal of Financial Economics 15, 261-281.

Bradley, Daniel, Bradford Jordan, and Jay Ritter, 2002, The quiet period goes out with a bang, Forthcoming, Journal of Finance.

Burch, John, and Bruce Foerster, 2002, Securities Industry Association Capital Markets Handbook, Aspen Law and Business, New York.

Cable, Stuart, 2001, Top Ten To-Do’s for CEOs of the Next New Thing, IPO Journal 4(11), 11-13.

Carter, Richard, and Steven Manaster, 1990, Initial public offerings and underwriter reputation, Journal of Finance 45, 1045-1067.

Carter, Richard, Frederick H. Dark, and Ajai K. Singh, 1998, Underwriter reputation, initial returns, and the long-run performance of IPO stocks, Journal of Finance 53, 285-311.

Chen, Hsuan-Chi, and Jay Ritter, 2000, The seven percent solution, Journal of Finance 55, 1105-1131.

Eccles, Robert, and Dwight Crane, 1988, Doing Deals (Harvard Business School Press, Boston, MA).

Ellis, Katrina, Roni Michaely, and Maureen O’Hara, 2000, When the underwriter is the market maker: An examination of trading in the IPO aftermarket, Journal of Finance 55, 1039-1074.

Fama, Eugene F., and Kenneth French, 1997, Industry costs of equity, Journal of Financial Economics 43, 153-193.

Hanley, Kathleen, 1993, The underpricing of initial public offerings and the partial adjustment phenomenon, Journal of Financial Economics 34, 231-150

Krigman, Laurie, Wayne Shaw, and Kent Womack, 2001, Why do firms switch underwriters?, Journal of Financial Economics 60, 245-284.

Logue, Dennis, Richard Rogalski, James Seward, and Lynn Foster-Johnson, What is special about the roles of underwriter reputation and market activities in initial public offerings?, Journal of Business 75, 213-243.

Loughran, Tim, and Jay Ritter, 2002, Why has IPO Underpricing Changed Over Time?, Working paper, University of Florida and University of Notre Dame.

Merton, Robert, 1987, A simple model of capital market equilibrium with incomplete information, Journal of Finance 42, 483-510.

Pichler, Pegaret, and William Wilhelm, 2001, A theory of the syndicate: form follows function, Journal of Finance 56, 2237-2264.

Rock, Kevin, 1986, Why new issues are underpriced, Journal of Financial Economics 15, 187-212.

Ruud, Judith, 1993, Underwriter price support and the IPO underpricing puzzle, Journal of Financial Economics 34, 135-151.

Schultz, Paul, and Mir Zaman, 1994, Aftermarket support and underpricing of initial public offerings, Journal of Financial Economics 35, 199-219.

Securities Industry Association, Securities Industry Yearbook, 1995 – 2002, New York, NY.

Standard and Poor’s Guide to Securities Dealers of North America, 1998, McGraw-Hill, Charlottesville, VA, 22902.

Titman, S., and B. Trueman, 1986, Information quality and the valuation of new issues, Journal of Accounting and Economics 8, 159-172.

Table 1: Underwriting Syndicate for Blockbuster Inc

The table lists underwriting share allocations for the IPO of Blockbuster, Inc. (BBI) on August 10, 1999 as listed in the final IPO prospectus. Underwriter roles within the syndicate are from SDC. The syndicate involves 22 underwriters including 2 joint book managers, 9 co-managers, and 11 other syndicate members. The Herfindahl Index of underwriting allocations is 11.36, with 37% of shares underwritten by the book managers, 54.9% of shares underwritten by the co-managers, and 8.1% of shares underwritten by the remaining syndicate members.

|Underwriter |Underwritten Shares |SDC Underwriter Role |

|Domestic Shares: | | |

|Salomon Smith Barney |4,433,500 |Joint Book Manager |

|Bear Stearns & Co Inc |4,433,500 |Joint Book Manager |

|CS First Boston Corp |2,675,000 |Co-Manager |

|Goldman Sachs & Co |2,675,000 |Co-Manager |

|JP Morgan |2,675,000 |Co-Manager |

|Banc of America Securities LLC |900,000 |Co-Manager |

|ING Barings |900,000 |Co-Manager |

|PaineWebber |900,000 |Co-Manager |

|Schroder & Co Inc |900,000 |Co-Manager |

|SG Cowen Securities Corp |900,000 |Co-Manager |

|Wit Capital Group Inc |900,000 |Co-Manager |

|BancBoston Robertson Stephens |228,000 |Syndicate Member |

|Barrington Associates |228,000 |Syndicate Member |

|Cazenove & Co |228,000 |Syndicate Member |

|First Union Capital Markets |228,000 |Syndicate Member |

|Gabelli & Co Inc |228,000 |Syndicate Member |

|Gerard Klauer Mattison & Co |228,000 |Syndicate Member |

|Guzman & Co |228,000 |Syndicate Member |

|Loop Capital Markets |228,000 |Syndicate Member |

|Utendahl Capital Partners |228,000 |Syndicate Member |

|Wasserstein Perella Securities |228,000 |Syndicate Member |

|Williams Capital Group LP |228,000 |Syndicate Member |

| | | |

|International Shares: | | |

|Salomon Brothers International Ltd |1,302,000 |Global Lead |

|Bear, Stearns International Ltd |1,302,000 |Global Lead |

|Credit Suisse First Boston (Europe) |744,000 |Syndicate Member |

|Goldman Sachs International |744,000 |Syndicate Member |

|J.P. Morgan Securities Ltd |744,000 |Syndicate Member |

|Bank of America International Ltd |272,800 |Syndicate Member |

|ING Barings Ltd, for ING Bank |272,800 |Syndicate Member |

|PaineWebber International (U.K.) Ltd |272,800 |Syndicate Member |

|J. Henry Schroder & Co. Ltd |272,800 |Syndicate Member |

|Societe Generale |272,800 |Syndicate Member |

|Total U.S. Shares |24,800,000 | |

|Total U.S. and International Shares |31,000,000 | |

Table 2: Summary Statistics for Offer and Syndicate Characteristics

The table lists summary statistics for offer characteristics (Panel A) and syndicate characteristics (Panel B). The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. Underpricing is defined as the percentage return from the offer price to the first day’s closing price. Offer proceeds and offered shares are based on global proceeds. The adjustment from the filing price is the percentage return from the midpoint of the original filing range to the offer price. Aftermarket standard deviation is estimated using continuously-compounded daily returns from days 21 through 125 after the IPO and is missing for 31 IPOs. Adjusted Carter-Manaster Ranks for book managers are taken from Jay Ritter’s web page at http:// bear.cba.ufl.edu/ritter/Rank.HTM. The Megginson-Weiss rank for the book manager is defined as the proportion of all IPO proceeds during the sample period for which that underwriter served as book manager. Syndicate characteristics are based on underwriting allocations as listed in the final prospectus or in SDC. The Herfindahl Index is the sum of the squared percentages of shares underwritten by each investment bank.

| |Mean |Min |25th Percentile |Median |75th Percentile |Max |

|Panel A – Offer and Underwriter Characteristics |

|Offer Price |13.46 |3.33 |10.00 |13.00 |16.00 |97.00 |

|Offer Proceeds ($mil) |96.48 |2.32 |29.70 |50.05 |88.35 |5470.00 |

|Total Shares Offered (mil) |5.93 |0.37 |2.60 |4.00 |6.00 |173.91 |

|Underpricing (%) |42.06 |-43.23 |1.27 |15.67 |49.20 |697.20 |

|Adjustment from Filing Price (%) |7.33 |-65.91 |-9.09 |0.00 |18.18 |344.44 |

|Aftermarket Standard Deviation (%) |5.94 |0.52 |3.88 |5.83 |7.68 |16.81 |

|Book Manager’s Carter-Manaster Rank |7.61 |1.10 |7.10 |8.10 |9.10 |9.10 |

|Book Manager’s Megginson-Weiss Rank (%) |5.22 |0.00 |0.32 |3.69 |7.91 |17.90 |

|Panel B – Syndicate Characteristics |

|# Book Managers per Issue |1.03 |1.00 |1.00 |1.00 |1.00 |2.00 |

|# Co-Managers per Issue |1.93 |0.00 |1.00 |2.00 |3.00 |12.00 |

|Total Underwriters per Issue |15.96 |1.00 |10.00 |15.00 |21.00 |90.00 |

|Herfindahl Index (%) |29.26 |3.89 |19.50 |24.42 |30.83 |100.00 |

|Book Manager Allocation (%) |41.00 |6.00 |30.14 |38.29 |45.00 |100.00 |

|Allocation per Book Manager (%) |40.22 |6.00 |29.50 |37.65 |44.00 |100.00 |

|Co-Manager Allocation (%) |38.05 |0.00 |32.23 |40.87 |48.57 |82.50 |

|Allocation per Co-Mgr (%) |20.25 |0.00 |15.09 |20.25 |25.42 |52.78 |

|Book Manager Proceeds ($mil) |28.47 |0.38 |10.20 |17.43 |29.40 |718.26 |

|Proceeds per Book Manager ($ mil) |26.99 |0.38 |10.17 |17.40 |28.80 |718.26 |

|Co-Manager Proceeds ($mil) |41.13 |0.00 |8.99 |18.78 |35.76 |2513.74 |

|Proceeds per Co-Manager ($mil) |15.14 |0.00 |6.14 |10.17 |16.67 |502.75 |

|Proceeds for Rest of Syndicate ($mil) |16.68 |0.00 |4.92 |9.41 |16.56 |1144.00 |

|Proceeds per Syndicate Member ($mil) |1.10 |0.00 |0.43 |0.80 |1.28 |25.42 |

Table 3: Summary Statistics by Year

The table lists mean values for offer characteristics (Panel A) and syndicate characteristics (Panel B) by year. The last column lists a p-value from a test of the restriction that means are equal across years based on analysis of variance. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. The number of active underwriters is a count of all underwriters that participate in at least one syndicate during a given year. The number of active book managers and co-managers is calculated similarly. All other variables are defined as in Table 2.

| |1997 |1998 |1999 |2000 |p-value |

|Panel A – Offer and Book manager Characteristics |

|Number of IPOs |454 |294 |459 |340 |- |

|Offer Price |11.83 |12.38 |14.84 |14.69 |0.000 |

|Offer Proceeds ($mil) |58.09 |80.72 |124.53 |123.49 |0.000 |

|Total Shares (mil) |4.03 |5.31 |6.71 |7.93 |0.000 |

|Underpricing (%) |14.13 |19.81 |72.19 |55.88 |0.000 |

|Adjustment from Filing Price (%) |-1.53 |-0.38 |17.78 |11.71 |0.000 |

|Aftermarket Standard Deviation (%) |4.09 |5.40 |6.63 |7.95 |0.000 |

|Book Manager’s Carter-Manaster Rank |6.95 |7.11 |8.05 |8.34 |0.000 |

|Book Manager’s Megginson-Weiss Rank (%) |3.55 |4.25 |6.36 |6.75 |0.000 |

|Panel B – Syndicate Characteristics |

|# of Active Underwriters |401 |344 |370 |265 |- |

|# of Active Book Managers |130 |102 |72 |50 |- |

|# of Active Co-managers |137 |104 |116 |80 |- |

|# Book Managers per Issue |1.00 |1.01 |1.04 |1.06 |0.000 |

|# Co-Managers per Issue |1.42 |1.60 |2.28 |2.43 |0.000 |

|Total Underwriters per Issue |18.01 |14.66 |15.78 |14.59 |0.000 |

|Herfindahl Index (%) |29.66 |32.57 |27.57 |28.15 |0.002 |

|Book Manager Allocation (%) |39.39 |42.59 |40.51 |42.44 |0.040 |

|Allocation per Book Manager (%) |39.35 |42.31 |39.25 |40.87 |0.077 |

|Co-Manager Allocation (%) |33.24 |34.18 |41.53 |43.14 |0.000 |

|Allocation per Co-Mgr (%) |22.75 |20.03 |19.19 |18.52 |0.000 |

|Book Manager Proceeds ($mil) |16.35 |21.60 |35.37 |41.26 |0.000 |

|Proceeds per Book Manager ($ mil) |16.33 |21.02 |33.61 |37.45 |0.000 |

|Co-Manager Proceeds ($mil) |22.47 |35.09 |55.21 |52.27 |0.000 |

|Proceeds per Co-Manager ($mil) |11.78 |13.51 |17.77 |17.48 |0.000 |

|Proceeds for Rest of Syndicate ($mil) |15.45 |15.82 |19.96 |14.65 |0.357 |

|Proceeds per Syndicate Member ($mil) |0.83 |0.99 |1.25 |1.35 |0.000 |

Table 4: Book Manager Participation by Offer Size Quintile

The table lists the top five underwriters in each offer size quintile. Underwriters are ranked based on the total offer proceeds for which they acted as book manager. Offer size quintiles are defined by year. Book manager allocation is the proportion of shares underwritten by the book manager. This number is stated per book manager if there are joint book managers. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs.

|Book Manager |Number of Offers |Total IPO Proceeds |Book Manager Allocation (%) |

| | |($ mil) | |

|Panel A – Offer Size Quintile 1 (Small) |

|CS First Boston |15 |618.00 |41.86 |

|Lehman |9 |381.45 |40.96 |

|Deutsche Bank Alex Brown |9 |377.10 |35.43 |

|DLJ |7 |240.24 |35.92 |

|Chase Hambrecht & Quist |7 |231.50 |36.89 |

|Panel B – Offer Size Quintile 2 |

|CS First Boston |27 |1558.23 |39.79 |

|Goldman Sachs |13 |693.85 |41.23 |

|Morgan Stanley DW |11 |531.02 |40.55 |

|Deutsche Bank Alex Brown |8 |513.95 |37.54 |

|Lehman |9 |437.35 |37.81 |

|Panel C – Offer Size Quintile 3 |

|CS First Boston |26 |1814.15 |37.78 |

|Fleet Robertson Stephens |17 |1106.27 |43.98 |

|Deutsche Bank Alex Brown |13 |1032.60 |33.51 |

|Lehman |14 |906.72 |35.90 |

|Hambrecht |19 |859.95 |34.89 |

|Panel D – Offer Size Quintile 4 |

|Goldman Sachs |37 |3274.25 |36.18 |

|CS First Boston |26 |2565.21 |36.75 |

|Morgan Stanley DW |26 |2431.84 |33.87 |

|Merrill Lynch |23 |2145.90 |33.19 |

|DLJ |24 |1708.42 |37.54 |

|Panel D – Offer Size Quintile 5 (Large) |

|Morgan Stanley DW |49 |25474.29 |28.03 |

|Goldman Sachs |60 |21010.24 |28.58 |

|Merrill Lynch |27 |9828.16 |26.50 |

|CS First Boston |25 |8641.46 |29.80 |

|Salomon Smith Barney |25 |8061.52 |27.79 |

Table 5: Determinants of Syndicate Size and Concentration

The table lists coefficient estimates from OLS regressions of syndicate size and concentration measures on various offer characteristics. The sample consists of 1,547 IPOs from 1997 through 2000, excluding units, rights, investment funds, and REITs. Aftermarket standard deviation is estimated using continuously-compounded daily returns from days 21 through 125 after the IPO and is missing for 31 IPOs. Offer proceeds are global proceeds expressed in millions of dollars. The NYSE/Amex dummy takes a value of one if the IPO is initially listed on the NYSE or Amex, and zero otherwise. t-statistics and z-statistics based on robust standard errors are listed in parentheses below the coefficients.

| | | |Log Proceeds x |Log Proceeds x |Log Proceeds x |Log Proceeds x |Log Proceeds x | |

| | |Log of Proceeds |Standard Deviation |NYSE/Amex Dummy |1998 Dummy |1999 Dummy |2000 Dummy | |

|Dependent Variable |Intercept | | | | | | |R2 |

|Poisson Regressions |

|Number of Underwriters |1.6796 |0.3335 |-0.1540 |0.0033 |-0.0553 |-0.0714 |-0.0991 |0.1458 |

| |(25.45) |(17.39) |(-0.86) |(0.34) |(-5.48) |(-7.24) |(-8.51) | |

|Number of Co-Managers |-0.8771 |0.3221 |0.4116 |0.0121 |0.0115 |0.0398 |0.0383 |0.1192 |

| |(-12.17) |(15.02) |(3.21) |(1.37) |(1.11) |(4.01) |(3.87) | |

|Ordinary Least Squares Regressions |

|Proportion of Shares Book Manager |0.7721 |-0.1080 |-0.0583 |0.0029 |0.0124 |0.0237 |0.0327 |0.2631 |

|Underwrites |(27.38) |(-12.82) |(-1.22) |(0.91) |(3.92) |(7.98) |(10.25) | |

|Proportion of Shares Co-Managers |0.0221 |0.0859 |0.0922 |0.0008 |-0.0001 |0.0015 |0.0002 |0.3341 |

|Underwrite |(0.98) |(12.26) |(2.08) |(0.26) |(-0.04) |(0.55) |(0.05) | |

|Herfindahl Index |0.6810 |-0.1090 |-0.1083 |0.0084 |0.0115 |0.0181 |0.0255 |0.2635 |

| |(22.17) |(-12.45) |(-2.30) |(2.89) |(3.63) |(6.30) |(8.17) | |

Table 6: Syndicate Structure and IPO Underpricing

The table lists coefficient estimates from OLS regressions of IPO underpricing on syndicate characteristics and a set of control variables. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. Control variables include offer proceeds, the natural log of offer proceeds, aftermarket standard deviation, the book manager’s Megginson-Weiss rank, a dummy variable for venture capital backing, and dummy variables for two-digit SIC codes 28, 35, 36, 38, 48, 60, 73, and 87. A dummy for Internet stocks is also included. The Megginson-Weiss rank is defined as the proportion of all IPO proceeds during the sample period for which an underwriter served as book manager. Adjusted Carter-Manaster Ranks for underwriters as well as identities of Internet stocks are taken from Jay Ritter’s web page at http:// bear.cba.ufl.edu/ritter/Rank.HTM. All other variables are defined as in Table 2. t-statistics based on White’s heteroskedasticity consistent standard errors are listed in parentheses below the coefficients.

|Dependent Variable |Underpricing |Log of 1+ |Log of 1+ |Log of 1+ |Log of 1+ |

| | |Underpricing |Underpricing |Underpricing |Underpricing |

|Intercept |-0.748 |-0.355 |-0.395 |-0.440 |-0.298 |

| |(-7.16) |(-7.15) |(-6.76) |(-6.09) |(-6.44) |

|# Co-managers Ranked < 8 |-0.060 |-0.032 | | | |

| |(-2.48) |(-2.56) | | | |

|# Co-managers Ranked ( 8 |-0.075 |-0.045 | | | |

| |(-3.47) |(-3.82) | | | |

|# Non-Managing Underwriters Ranked < 8 |-0.007 |-0.004 | | | |

| |(-2.74) |(-2.64) | | | |

|# Non-Managing Underwriters Ranked ( 8 |-0.016 |-0.008 | | | |

| |(-2.68) |(-2.74) | | | |

|Herfindahl Index | | |0.137 | | |

| | | |(2.91) | | |

|% Shares Underwritten by | | | |0.185 | |

|Book Manager | | | |(2.87) | |

|% Shares Underwritten by | | | |0.028 | |

|Co-Managers | | | |(0.34) | |

|Number of Underwriters | | | | |-0.005 |

| | | | | |(-4.64) |

|Dummy for Book Manager Ranked ( 8 |0.064 |0.057 |0.041 |0.041 |0.028 |

| |(1.42) |(2.43) |(1.91) |(1.80) |(1.29) |

|Venture Capital Backing |0.132 |0.078 |0.082 |0.082 |0.074 |

| |(3.29) |(3.89) |(4.04) |(4.05) |(3.70) |

|Proceeds x .01 |-0.011 |-0.004 |-0.011 |-0.011 |-0.009 |

| |(-1.85) |(-1.36) |(-2.77) |(-2.72) |(-2.17) |

|Log of Proceeds |0.233 |0.121 |0.084 |0.085 |0.094 |

| |(7.04) |(7.70) |(6.17) |(6.21) |(6.52) |

|Aftermarket ( |4.574 |2.357 |2.397 |2.367 |2.420 |

| |(5.01) |(4.79) |(4.82) |(4.72) |(4.92) |

|Industry and Date Dummies |Yes |Yes |Yes |Yes |Yes |

|R2 |0.2727 |0.3100 |0.2944 |0.2959 |0.3017 |

Table 7: Syndicate Structure and Offer Price Revisions

The table lists coefficient estimates from Probit regressions for the likelihood of offer price revisions. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. In Panel A, the dependent variable takes a value of one if the offer price is greater than the midpoint of the initial filing range and zero otherwise. In Panel B, the dependent variable takes a value of one if the offer price is less than the midpoint of the initial filing range and zero otherwise. Positive Total Return is defined as the return from the midpoint of the filing range to the first day’s closing price, if positive, and zero otherwise. The Negative Total Return dummy is defined as one if the return from the midpoint of the filing range to the first day’s closing price is zero or negative. Expected offer proceeds is defined as the number of shares issued globally times the midpoint of the initial filing price range. Aftermarket standard deviation is estimated using continuously-compounded daily returns from days 21 through 125 after the IPO and is missing for 31 IPOs. The book manager’s Megginson-Weiss rank is defined as the proportion of all IPO proceeds during the sample period for which the underwriter served as book manager. Adjusted Carter-Manaster Ranks for underwriters are taken from Jay Ritter’s web page at Rank.HTM. The model also includes dummy variables for two-digit SIC codes 28, 35, 36, 38, 48, 60, 73, and 87. The filing period market return is the return on the CRSP equally-weighted index compounded from the filing date through the offer date. z-statistics are reported in parentheses below the coefficients.

|Panel A: Upward Price Revisions |

|Intercept |-1.163 |-1.206 |-0.939 |-1.041 |-0.993 |

| |(-6.50) |(-6.47) |(-4.65) |(-5.61) |(-5.43) |

|Positive Total Return |1.559 |1.710 |0.983 |1.977 |1.418 |

| |(12.20) |(7.68) |(2.91) |(9.56) |(2.26) |

|Positive Total Return x # Underwriters | |-0.010 | | | |

| | |(-0.99) | | | |

|Positive Total Return x # Non-Managers Ranked < 8 | | |-0.020 | | |

| | | |(-1.10) | | |

|Positive Total Return x # Non-Managers Ranked ( 8 | | |0.046 | | |

| | | |(1.09) | | |

|Positive Total Return x # Co-managers Ranked < 8 | | |0.698 | | |

| | | |(4.50) | | |

|Positive Total Return x # Co-managers Ranked ( 8 | | |0.759 | | |

| | | |(4.97) | | |

|Positive Total Return x Herfindahl Index | | | |-1.413 | |

| | | | |(-2.66) | |

|Positive Total Return x Book Manager’s Offering Share | | | | |-0.966 |

| | | | | |(-1.26) |

|Positive Total Return x Co-Managers’ Offer Share | | | | |1.488 |

| | | | | |(1.60) |

|Negative Total Return Dummy |-1.637 |-1.643 |-1.536 |-1.626 |-1.612 |

| |(-12.68) |(-12.71) |(-11.75) |(-12.63) |(-12.48) |

|Positive Total Return x Book Manager Ranked ( 8 | | |-0.736 | | |

| | | |(-2.38) | | |

|Natural Logarithm of Expected Proceeds |0.243 |0.2500 |0.163 |0.211 |0.195 |

| |(5.62) |(5.45) |(3.26) |(4.66) |(4.33) |

|Pseudo R2 |0.504 |0.503 |0.517 |0.509 |0.508 |

Table 7 (continued)

|Panel B: Downward Price Revisions |

|Intercept |-0.996 |-0.785 |-0.561 |-0.638 |-0.681 |

| |(-5.48) |(-4.17) |(-2.71) |(-3.31) |(-3.45) |

|Negative Total Return Dummy |2.005 |1.438 |1.196 |2.691 |3.355 |

| |(19.29) |(9.07) |(6.25) |(16.97) |(7.22) |

|Negative Total Return Dummy x # Underwriters | |0.039 | | | |

| | |(4.46) | | | |

|Negative Total Return Dummy x # Non-Managers Ranked < 8 | | |0.027 | | |

| | | |(1.99) | | |

|Negative Total Return Dummy x # Non-Managers Ranked ( 8 | | |0.059 | | |

| | | |(1.96) | | |

|Negative Total Return Dummy x # Co-managers Ranked < 8 | | |0.214 | | |

| | | |(2.24) | | |

|Negative Total Return Dummy x # Co-managers Ranked ( 8 | | |0.206 | | |

| | | |(2.21) | | |

|Negative Total Return Dummy x Herfindahl Index | | | |-1.972 | |

| | | | |(-6.12) | |

|Negative Total Return Dummy x Book Manager’s Offering Share | | | | |-2.497 |

| | | | | |(-4.51) |

|Negative Total Return Dummy x Co-Managers’ Offer Share | | | | |-0.595 |

| | | | | |(-0.90) |

|Positive Total Return |-1.013 |-1.002 |-0.992 |-0.996 |-0.998 |

| |(-6.79) |(-6.73) |(-6.66) |(-6.69) |(-6.70) |

|Negative Total Return x Book Manager Ranked ( 8 | | |0.001 | | |

| | | |(0.00) | | |

|Natural Logarithm of Expected Proceeds |0.046 |-0.009 |-0.068 |-0.048 |-0.037 |

| |(1.06) |(-0.20) |(-1.32) |(-1.01) |(-0.75) |

|Pseudo R2 |0.523 |0.534 |0.537 |0.541 |0.541 |

Table 8: Syndicate Size and Aftermarket Services

The table lists coefficient estimates from regressions of the number of analysts and market makers on measures of syndicate size and a set of control variables. In Panel A, the dependent variable is the initial number of market makers as obtained from CRSP. This variable is defined only for the 1,302 IPOs listed on Nasdaq. In Panel B, the dependent variable is the number of analysts who issued reports on the company during the first three months of trading as obtained from IBES. Aftermarket standard deviation is estimated using continuously-compounded daily returns from days 21 through 125 after the IPO and is missing for 31 IPOs. Underpricing is the return from the offer price to the closing price on the first day of trading. Dummy variables are created for offer dates for each year from 1998 through 2000. Industry dummies are created for two-digit SIC codes 28, 35, 36, 38, 48, 60, 73, and 87. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. For OLS regressions (columns 1 and 2), t-statistics are reported in parentheses below the coefficients. For Poisson regressions (columns 3 and 4), z-statistics are reported.

|Panel A: Determinants of Number of Market Makers |

| |OLS Regressions |Poisson Regressions |

|Intercept |0.8272 |1.4530 |1.7361 |1.7670 |

| |(0.94) |(1.71) |(31.05) |(32.56) |

|Log of Proceeds |2.5181 |2.1311 |0.1713 |0.1526 |

| |(11.44) |(9.45) |(12.96) |(11.09) |

|Proceeds |0.0051 |0.0043 |-0.0000 |-0.0004 |

| |(5.36) |(3.75) |(-0.03) |(-0.76) |

|# Underwriters |0.0061 | |0.0005 | |

| |(0.40) | |(0.64) | |

|# Co-managers | |0.7013 | |0.0359 |

| | |(3.57) | |(3.53) |

|Aftermarket Std. Dev. |24.017 |23.327 |1.1238 |1.1141 |

| |(3.13) |(3.10) |(2.68) |(2.68) |

|Underpricing |2.6112 |2.6284 |0.0912 |0.0920 |

| |(13.24) |(13.60) |(11.54) |(11.72) |

|Year Dummies |Yes |Yes |Yes |Yes |

|Industry Dummies |Yes |Yes |Yes |Yes |

|Adjusted R2 |0.7162 |0.7201 | | |

|Pseudo R2 | | |0.2952 |0.2967 |

Table 8 (continued)

|Panel B: Determinants of Number of Analysts |

| |OLS Regressions |Poisson Regressions |

|Intercept |-1.6146 |-1.0203 |-1.0399 |-0.8994 |

| |(-8.04) |(-5.18) |(-9.00) |(-7.23) |

|Log of Proceeds |0.9750 |0.5597 |0.4512 |0.3281 |

| |(15.59) |(8.75) |(15.00) |(9.63) |

|Proceeds | 0.0003 |-0.0004 |-0.0004 |-0.0008 |

| |(0.48) |(-1.06) |(-2.45) |(-3.07) |

|# Underwriters |-0.0017 | |0.0011 | |

| |(-0.37) | |(0.56) | |

|# Co-managers | |0.6886 | |0.2416 |

| | |(10.78) | |(9.88) |

|Aftermarket Std. Dev. |3.3494 |2.8407 |1.0189 |1.0880 |

| |(1.79) |(1.65) |(1.31) |(1.47) |

|Underpricing |0.2105 |0.2412 |0.0421 |0.0546 |

| |(3.77) |(4.53) |(2.85) |(3.73) |

|Year Dummies |Yes |Yes |Yes |Yes |

|Industry Dummies |Yes |Yes |Yes |Yes |

|Adjusted R2 |0.4202 |0.5165 | | |

|Pseudo R2 | | |0.1133 |0.1367 |

Table 9: Probit Models for Syndicate and Co-Manager Participation

The table lists coefficient estimates from Probit models of syndicate and co-manager participation. The sample includes 639 underwriters and 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. The models include one observation for each eligible underwriter for each IPO, where the set of eligible underwriters is adjusted for mergers and acquisitions among underwriters. In columns 1 and 2, the dependent variable equals one if the underwriter is included in the IPO syndicate and zero otherwise. In columns 3 and 4, the dependent variable equals one if the underwriter is included as a co-manager and zero otherwise. Aftermarket standard deviation is estimated using continuously-compounded daily returns from days 21 through 125 after the IPO and is missing for 31 IPOs. Adjusted Carter-Manaster Ranks for underwriters are taken from Jay Ritter’s web page at . Top 3 Analyst is a dummy variable equal to one if the underwriter employs one of the top three analysts in the IPO’s industry, as ranked by Institutional Investor. Book Manager Analyst is a dummy variable equal to one if the book manager underwriter employs one of the top three analysts in the IPO’s industry. Dummy variables are created to identify underwriters in the same state or an adjacent state to the book manager underwriter and in the same state or an adjacent state to the issuer. Participation in the Previous Ten Syndicates is defined as the proportion of the book manager’s last ten syndicates in which the underwriter was included. Participation in the Most Recent Syndicate is a dummy variable equal to one if the underwriter was included in the book manager’s most recent syndicate. Reciprocal Participation in the Previous Ten Syndicates is defined as the proportion of the underwriter’s last ten syndicates in which the book manager was included. Reciprocal Participation in the Most Recent Syndicate is a dummy variable equal to one if the book manager was included in the underwriter’s most recent syndicate. Standard errors are listed in parentheses below the coefficients.

Table 9 (continued)

| |Syndicate Participation |Co-Manager Participation |

| |1 |2 |3 |4 |

|Intercept |-2.2502 |-2.7767 |-3.1833 |-3.7023 |

| |(0.0199) |(0.0271) |(0.0452) |(0.0598) |

|Year 1998 Dummy |-0.1167 |-0.1240 |0.0323 |0.0272 |

| |(0.0103) |(0.0129) |(0.0248) |(0.0292) |

|Year 1999 Dummy |-0.1636 |-0.1661 |0.0856 |0.1033 |

| |(0.0102) |(0.0128) |(0.0235) |(0.0275) |

|Year 2000 Dummy |-0.1897 |-0.2194 |0.1326 |0.0981 |

| |(0.0125) |(0.0157) |(0.0274) |(0.0326) |

|Ln(Offer Proceeds) |0.1210 |0.1480 |0.0929 |0.1345 |

| |(0.0036) |(0.0046) |(0.0082) |(0.0097) |

|Aftermarket Standard Deviation |-1.0936 |-1.3507 |-0.1923 |0.1255 |

| |(0.1738) |(0.2178) |(0.3709) |(0.4367) |

|Carter-Manaster Rank ≥ 8 |0.1397 |- |0.4179 |- |

| |(0.0121) | |(0.0224) | |

|Ln(Underwriter Capitalization) |- |0.1177 |- |0.1107 |

| | |(0.0025) | |(0.0058) |

|Regional Underwriter |-0.4512 |0.0376 |-0.4210 |-0.1510 |

| |(0.0098) |(0.0114) |(0.0211) |(0.0243) |

|Probability of Being Book Manager |3.6821 |2.9690 |3.7121 |3.8493 |

| |(0.1506) |(0.1690) |(0.1870) |(0.2143) |

|Top 3 Analyst |0.0304 |0.0014 |-0.0864 |-0.0347 |

| |(0.0241) |(0.0252) |(0.0367) |(0.0390) |

|Top 3 Analyst* Book Manager Analyst |-0.0919 |-0.0250 |0.2899 |0.2869 |

| |(0.0396) |(0.0420) |(0.0522) |(0.0553) |

|Book Manager’s State |-0.0556 |-0.0892 |-0.0887 |-0.0417 |

| |(0.0088) |(0.0108) |(0.0191) |(0.0216) |

|Adjacent State to Book Manager |0.0362 |0.0803 |-0.0792 |-0.0584 |

| |(0.0121) |(0.0150) |(0.0311) |(0.0358) |

|Issuer State |0.2808 |0.2616 |0.1981 |0.1470 |

| |(0.0184) |(0.0231) |(0.0370) |(0.0425) |

|Adjacent State to Issuer |0.1667 |0.1920 |0.0923 |0.1184 |

| |(0.0113) |(0.0136) |(0.0249) |(0.0274) |

|Issuer State * Not Book Manager State |0.2106 |0.2732 |0.2122 |0.3518 |

| |(0.0227) |(0.0289) |(0.0457) |(0.0532) |

|Participation in Previous Ten Syndicates |2.3721 |1.9292 |1.1775 |0.9283 |

| |(0.0229) |(0.0254) |(0.0409) |(0.0448) |

|Participation in the Most Recent Syndicate |0.1737 |0.1410 |0.0944 |0.0594 |

| |(0.0145) |(0.0156) |(0.0261) |(0.0275) |

|Reciprocal Participation in Previous Ten |0.7726 |0.6385 |0.5014 |0.4355 |

|Syndicates |(0.0256) |(0.0272) |(0.0427) |(0.0450) |

|Reciprocal Participation in the Most Recent |0.1012 |0.0675 |0.1406 |0.1139 |

|Syndicate |(0.0192) |(0.0195) |(0.0296) |(0.0303) |

|N |898,797 |312,654 |898,797 |312,654 |

|Pseudo-R2 |0.2899 |0.2600 |0.2873 |0.2410 |

Table 10: Probit Models for Syndicate and Co-Manager Participation by Size Quintile

The table lists coefficient estimates from Probit models of syndicate and co-manager participation estimated separately for the first, third, and fifth quintiles of offer proceeds, where quintiles are defined by year. The sample includes 639 underwriters and 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs. The models include one observation for each eligible underwriter for each IPO, where the set of eligible underwriters is adjusted for mergers and acquisitions among underwriters. In Panel A, the dependent variable equals one if the underwriter is included in the IPO syndicate and zero otherwise. In Panel B, the dependent variable equals one if the underwriter is included as a co-manager and zero otherwise. Explanatory variables are defined as in Table 9. Standard errors are listed in parentheses below the coefficients.

Table 10 (continued)

|Panel A: Syndicate Participation |

| |Small Quintile |Size Quintile 3 |Large Quintile |

|Intercept |-2.7294 |-3.4478 |-1.4066 |-1.4004 |-2.5278 |-3.0454 |

| |(0.0627) |(0.0905) |(0.1835) |(0.2305) |(0.0536) |(0.0730) |

|Year 1998 Dummy |-0.0251 |-0.0299 |-0.0822 |-0.0669 |-0.1227 |-0.1160 |

| |(0.0309) |(0.0430) |(0.0235) |(0.0290) |(0.0211) |(0.0265) |

|Year 1999 Dummy |-0.1662 |-0.2185 |-0.0513 |0.0273 |-0.1445 |-0.1309 |

| |(0.0323) |(0.0441) |(0.0401) |(0.0501) |(0.0210) |(0.0265) |

|Year 2000 Dummy |-0.2030 |-0.2909 |-0.0012 |0.0723 |-0.2181 |-0.2484 |

| |(0.0382) |(0.0519) |(0.0512) |(0.0642) |(0.0267) |(0.0341) |

|Ln(Offer Proceeds) |0.2578 |0.4381 |-0.1148 |-0.2273 |0.1820 |0.1172 |

| |(0.0224) |(0.0320) |(0.0521) |(0.0654) |(0.0087) |(0.0114) |

|Aftermarket Standard Deviation |-1.0925 |-2.0275 |-1.1993 |-1.4665 |-1.6856 |-2.4226 |

| |(0.4265) |(0.5733) |(0.4242) |(0.5253) |(0.3820) |(0.4806) |

|Carter-Manaster Rank ≥ 8 |-0.0755 |- |0.0783 |- |0.3507 |- |

| |(0.0373) | |(0.0273) | |(0.0227) | |

|Ln(Underwriter Capitalization) |- |0.0656 |- |0.1076 |- |0.1675 |

| | |(0.0067) | |(0.0055) | |(0.0053) |

|Regional Underwriter |-0.3330 |0.0799 |-0.4590 |0.0228 |-0.5178 |0.0070 |

| |(0.0285) |(0.0332) |(0.0218) |(0.0252) |(0.0195) |(0.0229) |

|Probability of Being Book Manager |4.5392 |3.4385 |5.1287 |4.3298 |1.2058 |1.0184 |

| |(0.5309) |(0.6498) |(0.3471) |(0.4038) |(0.2529) |(0.2768) |

|Top 3 Analyst |-0.0239 |-0.0952 |0.0455 |-0.0062 |0.3007 |0.2790 |

| |(0.0719) |(0.0739) |(0.0481) |(0.0505) |(0.0568) |(0.0602) |

|Top 3 Analyst* Book Manager Analyst |-0.1836 |-0.1686 |-0.2239 |-0.2058 |-0.0995 |-0.0220 |

| |(0.1377) |(0.1468) |(0.0963) |(0.1035) |(0.0776) |(0.0827) |

|Book Manager’s State |0.0053 |-0.0666 |-0.0233 |-0.0426 |-0.1025 |-0.1421 |

| |(0.0258) |(0.0331) |(0.0199) |(0.0241) |(0.0172) |(0.0212) |

|Adjacent State to Book Manager |0.0644 |0.0859 |0.1078 |0.1472 |-0.0608 |-0.0001 |

| |(0.0319) |(0.0408) |(0.0275) |(0.0342) |(0.0248) |(0.0309) |

|Issuer State |0.3652 |0.3086 |0.3119 |0.2956 |0.2543 |0.2439 |

| |(0.0493) |(0.0688) |(0.0479) |(0.0608) |(0.0323) |(0.0399) |

|Adjacent State to Issuer |0.1070 |0.1218 |0.1788 |0.2158 |0.1544 |0.1684 |

| |(0.0329) |(0.0414) |(0.0262) |(0.0311) |(0.0221) |(0.0267) |

|Issuer State * Not Book Manager State |0.0409 |0.0894 |0.2381 |0.3346 |0.2314 |0.2980 |

| |(0.0599) |(0.0824) |(0.0558) |(0.0713) |(0.0437) |(0.0560) |

|Participation in Previous Ten Syndicates |2.3464 |2.1323 |2.2119 |1.7566 |2.5703 |1.9343 |

| |(0.0656) |(0.0759) |(0.0500) |(0.0550) |(0.0471) |(0.0517) |

|Participation in the Most Recent Syndicate|0.2472 |0.1741 |0.2087 |0.1642 |0.1237 |0.1106 |

| |(0.0454) |(0.0493) |(0.0319) |(0.0343) |(0.0277) |(0.0296) |

|Reciprocal Participation in Previous Ten |0.8608 |0.6945 |0.6667 |0.5312 |0.7665 |0.6245 |

|Syndicates |(0.0753) |(0.0815) |(0.0554) |(0.0586) |(0.0516) |(0.0549) |

|Reciprocal Participation in the Most |0.2315 |0.1946 |0.1451 |0.1141 |0.0509 |0.0269 |

|Recent Syndicate |(0.0588) |(0.0603) |(0.0412) |(0.0419) |(0.0379) |(0.0384) |

|N |170,188 |59,264 |182,603 |63,495 |181,383 |63,076 |

|Pseudo-R2 |0.2574 |0.2558 |0.2875 |0.2440 |0.3115 |0.2825 |

Table 10 (continued)

|Panel B: Co-Manager Participation |

| |Small Quintile |Size Quintile 3 |Large Size Quintile |

|Intercept |-3.3962 |-4.0665 |-3.9942 |-4.0504 |-3.4871 |-4.2869 |

| |(0.1352) |(0.2066) |(0.3905) |(0.4635) |(0.1260) |(0.1588) |

|Year 1998 Dummy |0.0314 |-0.1061 |0.0018 |0.0432 |0.0068 |0.0401 |

| |(0.0797) |(0.1328) |(0.0551) |(0.0644) |(0.0517) |(0.0568) |

|Year 1999 Dummy |0.1113 |0.0324 |-0.1063 |0.0132 |0.0420 |0.0840 |

| |(0.0725) |(0.1052) |(0.0859) |(0.1010) |(0.0500) |(0.0554) |

|Year 2000 Dummy |0.0984 |0.0154 |-0.1390 |-0.0733 |0.1402 |0.0763 |

| |(0.0838) |(0.1189) |(0.1078) |(0.1278) |(0.0598) |(0.0685) |

|Ln(Offer Proceeds) |0.1481 |0.3728 |0.3223 |0.2455 |0.1457 |0.1584 |

| |(0.0470) |(0.0731) |(0.1104) |(0.1305) |(0.0203) |(0.0227) |

|Aftermarket Standard Deviation |-0.1646 |-0.6637 |0.8221 |1.3119 |-0.8102 |-1.0636 |

| |(0.9294) |(1.2713) |(0.8761) |(1.0236) |(0.8411) |(0.9496) |

|Carter-Manaster Rank ≥ 8 |0.0905 |- |0.3414 |- |0.7558 |- |

| |(0.0702) | |(0.0506) | |(0.0436) | |

|Ln(Underwriter Capitalization) |- |0.0382 |- |0.0940 |- |0.2103 |

| | |(0.0150) | |(0.0122) | |(0.0145) |

|Regional Underwriter |-0.2973 |-0.1534 |-0.3767 |-0.0996 |-0.5785 |-0.1975 |

| |(0.0595) |(0.0711) |(0.0462) |(0.0515) |(0.0448) |(0.0525) |

|Probability of Being Book Manager |5.4000 |5.6008 |4.6326 |5.1528 |2.1568 |2.2959 |

| |(0.6545) |(0.8352) |(0.4659) |(0.5542) |(0.2850) |(0.3212) |

|Top 3 Analyst |-0.2308 |-0.1538 |-0.3115 |-0.2546 |0.3238 |0.3140 |

| |(0.1343) |(0.1362) |(0.0863) |(0.0903) |(0.0664) |(0.0719) |

|Top 3 Analyst* Book Manager Analyst |0.2655 |0.2367 |0.0789 |-0.0320 |0.1320 |0.1789 |

| |(0.2117) |(0.2174) |(0.1579) |(0.1713) |(0.0855) |(0.0921) |

|Book Manager’s State |-0.0569 |-0.1017 |-0.1753 |-0.1515 |-0.0949 |0.0313 |

| |(0.0573) |(0.0710) |(0.0445) |(0.0501) |(0.0399) |(0.0440) |

|Adjacent State to Book Manager |0.0966 |0.0864 |-0.1062 |-0.0818 |-0.1754 |-0.1331 |

| |(0.0664) |(0.0866) |(0.0719) |(0.0816) |(0.0761) |(0.0833) |

|Issuer State |0.4227 |0.3678 |0.1943 |0.1555 |0.1355 |0.0691 |

| |(0.1012) |(0.1424) |(0.1073) |(0.1252) |(0.0633) |(0.0695) |

|Adjacent State to Issuer |0.0880 |0.1098 |0.0192 |0.0613 |0.0747 |0.0610 |

| |(0.0721) |(0.0886) |(0.0620) |(0.0663) |(0.0486) |(0.0522) |

|Issuer State * Not Book Manager State |0.0162 |0.1096 |0.2340 |0.3175 |0.3007 |0.4768 |

| |(0.1190) |(0.1633) |(0.1205) |(0.1415) |(0.0916) |(0.1052) |

|Participation in Previous Ten Syndicates |1.3069 |1.0914 |1.2070 |0.9731 |1.1126 |0.7921 |

| |(0.1187) |(0.1421) |(0.0893) |(0.0964) |(0.0832) |(0.0884) |

|Participation in the Most Recent Syndicate|0.1307 |0.0585 |0.0600 |0.0435 |0.0099 |0.0139 |

| |(0.0842) |(0.0922) |(0.0591) |(0.0620) |(0.0484) |(0.0501) |

|Reciprocal Participation in Previous Ten |0.7217 |0.6357 |0.5268 |0.4347 |0.5014 |0.4054 |

|Syndicates |(0.1264) |(0.1406) |(0.0937) |(0.0979) |(0.0828) |(0.0860) |

|Reciprocal Participation in the Most |0.1276 |0.0515 |0.1773 |0.1542 |0.1639 |0.1480 |

|Recent Syndicate |(0.0969) |(0.1021) |(0.0668) |(0.0679) |(0.0538) |(0.0545) |

|N |170,188 |59,264 |182,603 |63,495 |181,383 |63,076 |

|Pseudo-R2 |0.2198 |0.2176 |0.2586 |0.2059 |0.4019 |0.3313 |

Table 11: Book Manager – Syndicate Member Frequencies

The table lists frequencies of book manager – syndicate member combinations involving the top 10 book managers in the sample (based on number of IPOs underwritten). Panel A lists frequencies for syndicate participation by the top ten underwriters. Panel B lists frequencies for syndicate participation for underwriters that were involved in at least 60 IPOs, but never acted as a book manager. Frequencies are stated as a percentage of IPOs taken public by the listed book manager. The last column lists the percentage (number) of all IPO syndicates in which the underwriter participated. Proportions are adjusted to account for underwriters that were not eligible for all IPO syndicates due to mergers and acquisitions. The sample includes 1,547 initial public offers issued in the U.S. between 1997 and 2000, excluding units, rights, investment funds, and REITs.

| |Institutional Book Managers |Mixed Book Managers |Retail Book Managers |Total |

| | | | |Syndicates (% |

|Syndicate Member | | | |of eligible) |

| |

|Goldman Sachs |- |10.92** |36.07*** |0.00*** |

| |

|Edward Jones |46.77*** |

|Name |Instl/ |Regnl/ |# Eligible |Name |Instl/ |Regnl/ |% of Book |% of Other |

| |Retail |Natl |Offers | |Retail |Natl |Manager’s |Syndicates |

| | | | | | | |Syndicates | |

|A.G. Edwards & Sons |Retail |Natl |11 |Legg Mason |Retail |Natl |90.91 |21.96*** |

|Banc of America |Inst |Natl |15 |H.C. Wainwright |Retail |Regl |93.33 |4.99*** |

|Banc of America |Inst |Natl |15 |John G. Kinnard & Co. |Retail |Regl |93.33 |2.98*** |

|Cowen & Co. |Mix |Natl |10 |Prudential Securities |Retail |Natl |90.00 |19.41*** |

|Cowen & Co. |Mix |Natl |10 |Raymond James & Associates |Retail |Regl |90.00 |26.76*** |

|Credit Suisse First Boston |Inst |Natl |119 |Invemed Associates |- |Regl |89.92 |0.39*** |

|Donaldson, Lufkin & Jenrette |Mix |Natl |10 |Alex Brown & Sons |Retail |Regl |90.00 |36.29*** |

|Donaldson, Lufkin & Jenrette |Mix |Natl |10 |Furman Selz |Inst |Natl |90.00 |25.32*** |

|Friedman Billings Ramsey |Inst |Natl |14 |Advest |Retail |Regl |92.86 |11.02*** |

|Friedman Billings Ramsey |Inst |Natl |14 |Tucker Anthony |Retail |Natl |92.86 |18.85*** |

|Montgomery Securities |Inst |Regl |21 |H.C. Wainwright |Retail |Regl |95.24 |4.60*** |

|Montgomery Securities |Inst |Regl |21 |Cruttenden Roth |Retail |Regl |90.48 |5.30*** |

|Morgan Keegan |Mix |Regl |10 |Interstate Johnson Lane |Retail |Regl |100.00 |11.15*** |

|Morgan Keegan |Mix |Regl |11 |Stephens |Inst |Regl |100.00 |10.43*** |

|Morgan Keegan |Mix |Regl |11 |J.C. Bradford & Co. |Retail |Regl |90.91 |14.31*** |

|Morgan Keegan |Mix |Regl |11 |J.J.B. Hilliard |Retail |Regl |90.91 |3.10*** |

|Morgan Keegan |Mix |Regl |11 |Robinson Humphrey |Retail |Regl |90.91 |18.49*** |

|NationsBanc Montgomery Sec. |Inst |Regl |20 |H.C. Wainwright |Retail |Regl |100.00 |4.60*** |

|NationsBanc Montgomery Sec. |Inst |Regl |20 |Cruttenden Roth |Retail |Regl |95.00 |5.30*** |

|NationsBanc Montgomery Sec. |Inst |Regl |20 |BT Alex Brown |Retail |Natl |90.00 |38.84*** |

|U.S. Bancorp Piper Jaffray |Retail |Regl |13 |Banc of America |Inst |Natl |92.31 |31.33*** |

Appendix A – SDC Data Corrections

The table describes corrections that were made to the SDC data. SDC provides underwriting allocation data for each underwriter in the syndicate as well as codes to define the role of each underwriter in the syndicate. The codes include Book Manager, Joint Book Manager, Joint Lead, Co-Manager, Syndicate Member, and Global Lead. All corrections to SDC data are verified using the IPO prospectus.

|Ticker |Date |Explanation |

|We gathered underwriting allocation data from prospectuses for all offers with missing SDC underwriting allocation data. There were approximately 269 such cases (13.9% of the sample). |

| |

|We collected additional data from the prospectuses for all remaining offers for which share allocations did not match the total shares offered as stated in SDC. 408 offers had non-matching |

|data. In most cases, these discrepancies are corrected using data from the final prospectus. An additional 59 offers are manually coded as matches: 30 offers differed only by the amount of |

|international share allocations that were not described in the prospectus, 14 offers differ only by an amount of shares offered directly by the firm rather than through the underwriter, and 15|

|offers differed due to incorrect SDC share data. These 59 offers are included in the final sample as matches using total allocations to determine offer size and offered shares. In six of the|

|remaining cases, the discrepancies were due to units and these offers are excluded from the sample. Finally, discrepancies could not be corrected for 50 offers. These are excluded from the |

|final sample. |

| |

|Forty-three offers have two underwriters coded as joint book managers. Joint book codes are verified using the IPO prospectus. 34 of the 43 are listed as Joint Book Managers in the |

|prospectus. The remaining nine are assumed to be corrected are used as listed by SDC. Ninety-four offers have one book manager and one additional underwriter coded as joint lead. Five |

|offers have one book manager and two additional underwriters coded as joint lead. All underwriters coded as joint lead, but not joint book manager, are considered co-managers. |

| |

|Global Lead Codes: Underwriter roles were verified in 66 cases where the underwriter was coded as a global lead, but was not also listed as a co-manager. This check resulted in three cases |

|where the global lead was coded as a co-manager. In all other cases, the global lead indicator is ignored. |

|CTIC |19970321 |SDC underwriter roles in incorrect order. Order is corrected. |

|SCMM |19971006 |Westdeutsche Landesbank Girozentrale coded as co-manager. |

|IMMT |19990426 |SDC underwriter roles in incorrect order. New China Hong Kong Securities and China Everbright Securities H.K. coded as co-managers. |

| | | |

|Co-Manager Codes: Underwriter roles were verified in seven cases where an underwriter received the same share allocation as the book manager, but was not coded as either a co-manager or a |

|joint lead manager. This check resulted in four corrections. |

|ERTH |19980323 |SDC lists four joint lead managers and no co-managers. The prospectus lists two joint book managers and no co-manager. |

|ABNH |19980714 |SDC lists two joint lead manager and no co-managers. The prospectus lists one lead and three co-managers. |

|ZIPL |19990526 |SDC lists one book manager and no co-managers. The prospectus lists one book manager and one co-manager. |

|ERMS |20000802 |SDC lists two joint lead manager and no co-managers. The prospectus lists one book manager and no co-managers. |

| | | |

|Co-Manager Codes: Underwriter roles were verified in 53 cases where the syndicate included more than five underwriters or the syndicate was made up of at least 75% co-managers. This check |

|resulted in ten corrections. |

|LHC |19981027 |SDC lists 10 underwriters, including 1 book manager and 9 co-managers. The prospectus lists 1 book manager and no co-managers. |

|DBSI |19981112 |SDC lists 7 underwriters, including 1 book manager and 6 co-managers. The prospectus lists 1 book manager and 1 co-manager. |

|IDSY |19990630 |SDC lists 17 underwriters, including 1 book manager and 16 co-managers. The prospectus lists 1 book manager and no co-managers. |

Appendix A – continued

|PRFT |19990729 |SDC lists 7 underwriters, including 1 book manager and 6 co-managers. The prospectus lists 1 book manager and no co-managers. |

|AIRO |19990729 |SDC lists 11 underwriters, including 1 book manager and 10 co-managers. The prospectus lists 1 book manager and 2 co-managers. |

|QSFT |19990812 |SDC lists 7 underwriters, including 1 book manager and 6 co-managers. The prospectus lists 1 book manager and 2 co-managers. |

|FSHP |19990927 |SDC lists 19 underwriters, including 1 book manager and 18 co-managers. The prospectus lists 1 book manager and 3 co-managers. |

|BTBC |20000215 |SDC lists 7 underwriters, including 1 book manager and 6 co-managers. The prospectus lists 1 book manager and 1 co-manager. |

|SFTY |20000218 |SDC lists 4 underwriters, including 1 book manager and 3 co-managers. The prospectus lists 1 book manager and no co-managers. |

|PMBC |20000614 |SDC lists 5 underwriters, including 1 book manager and 4 co-managers. The prospectus lists 1 book manager and no co-managers. |

Appendix B – Underwriter Name Adjustments Related to Mergers & Acquisitions

|Ann. Date |Eff. Date |Target Name |Acquirer Name |Name Adjustments |

|19970220 |19970502 |Equity Securities Trading Co |Southwest Securities Group Inc |Equity Securities Trading becomes Southwest Securities Group |

| | | | |after eff. date. |

|  |  |  |  |  |

|19970205 |19970531 |Morgan Stanley Group Inc |Dean Witter Discover & Co |Both prior names are replaced with Morgan Stanley Dean Witter |

| | | | |after eff. date. |

|  |  |  |  |  |

|19970407 |19970902 |Alex Brown Inc |Bankers Trust New York Corp |Both prior names are replaced with BT Alex Brown after eff. |

| | | | |date. |

|19981130 |19990604 |Bankers Trust New York Corp |Deutsche Bank AG |Both prior names are replaced with Deutsche Alex Brown after |

| | | | |eff. date. |

|  |  |  |  |  |

|19970515 |19970902 |Dillon Read & Co(UBS AG) |SBC Warburg (Swiss Bank Corp) |Prior to eff. date all Swiss Bank, SBC, and SBC Warburg names |

| | | | |are combined as SBC. Both this name and Dillon read are |

| | | | |replaced with Warburg Dillon Read after eff. date. |

|19980828 |19980828 |SBC Warburg Premier Sec Plc |Union Bank of Switzerland |Prior to eff. date, all UBS names are combined as UBS. Both |

| | | | |names are replaced with UBS Warburg after eff. date. Unit was |

| | | | |actually called Warburg Dillon Read until mid-2000, but the |

| | | | |name was later changed to UBS Warburg. |

|20000712 |20001103 |PaineWebber Group Inc |UBS AG |PaineWebber not in sample after eff. date so no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19970828 |19971008 |Furman Selz LLC |ING Barings (ING Groep NV) |Baring Brothers, ING Bank, and ING Barings are combined prior |

| | | | |to eff. date. All names replaced with ING Barings Furman Selz |

| | | | |after eff. date. |

|  |  |  |  |  |

Appendix B – continued

|Ann. Date |Eff. Date |Target Name |Acquirer Name |Name Adjustments |

|19970630 |19971001 |Montgomery Securities, CA |NationsBank Corp, Charlotte, NC |Montgomery Securities is replaced with NationsBank Montgomery |

| | | | |Securities after eff. date. |

|19970609 |19971001 |Robertson Stephens & Co |BankAmerica Corp |Robertson Stephens replaced with BA Robertson Stephens after |

| | | | |eff. date. |

|19970916 |19980202 |Quick & Reilly Group Inc |Fleet Financial Group Inc, MA |Quick and Reilly replaced with Fleet Financial after eff. date |

| | | | |(Q&R only shows up in 2000). |

|19980529 |19980901 |Robertson Stephens & Co |BankBoston Corp, Boston, MA |BA Robertson Stephens and BankAmerica Robertson Stephens |

| | | | |replaced with Bank Boston Robertson Stephens after eff. date. |

|19980413 |19980930 |BankAmerica Corp |NationsBank Corp, Charlotte, NC |BankAmerica, BA Securities, NationsBank replaced with Bank of |

| | | | |America after eff. date. |

|19990314 |19991001 |BankBoston Corp, Boston, MA |Fleet Financial Group Inc, MA |Both names replaced with Fleet Robertson Stephens after eff. |

| | | | |date. |

|  |  |  |  |  |

|19970722 |19971103 |Oppenheimer (Oppenheimer Group) |CIBC Wood Gundy Securities Inc |CIBC Wood Gundy is the U.S. securities operation of CIBC World |

| | | | |Markets. Oppenheimer replaced with CIBC World after eff. date.|

|  |  |  |  |  |

|19971119 |19971120 |Hampshire Securities Corp |Gruntal & Co Inc |Hampshire Securities replaced with Gruntal after eff. date. |

|  |  |  |  |  |

|19970924 |19971128 |Salomon Inc |Travelers Group Inc |Smith Barney, a subsidiary of Travelers, is combined with |

| | | | |Salomon to form Salomon Smith Barney. Both names replaced |

| | | | |after eff. date. |

|19980406 |19981008 |Citicorp |Travelers Group Inc |Citicorp not in sample after 1998 so no name adjustments |

| | | | |required. |

|20000118 |20000501 |Schroders - Worldwide Investment |Salomon Smith Barney Holdings |Schroder and Schroder Wertheim replaced with Salomon Smith |

| | | | |Barney after eff. date. |

|  |  |  |  |  |

Appendix B – continued

|Ann. Date |Eff. Date |Target Name |Acquirer Name |Name Adjustments |

|19970925 |19980102 |Equitable Securities Corp |SunTrust Banks Inc, Atlanta, GA |Equitable Securities replaced with Sun Trust Equitable |

| | | | |Securities after eff. date. |

| | | | | |

|19970819 |19980202 |Wheat First Butcher Singer |First Union Corp, Charlotte, NC |Wheat First becomes First Union after eff. date. First Union |

| | | | |not in sample prior. |

|19971212 |19980109 |Principal Financial Securities |EVEREN Capital Corp |Principal Financial replaced with Everen after eff. date. |

|19990426 |19991001 |EVEREN Capital Corp |First Union Corp, Charlotte, NC |Everen not in sample after eff. date so no name adjustments |

| | | | |necessary. |

|  |  |  |  |  |

|19971218 |19980121 |Barclays de Zoete Wedd AU Ltd |ABN-AMRO Holding NV |Barclays not in sample after eff. date so no name adjustments |

| | | | |necessary. |

|  |  |  |  |  |

|- |19980102 |Rauscher Pierce Refsnes |Dain Bosworth |Both Names replaced with Dain Rauscher after eff. date. |

|19980209 |19980406 |Wessels Arnold & Henderson LLC |Dain Rauscher Corp |Both names replaced with Dain Rauscher Wessels after eff. date.|

|  |  |  |  |  |

|19971215 |19980501 |Piper Jaffray Companies |US Bancorp, Minneapolis, MN |Piper Jaffray becomes US Bancorp Piper Jaffrey after eff. date.|

| | | | |US Bancorp not in sample prior. |

|19980903 |19990104 |Libra Investments Inc |US Bancorp, Minneapolis, MN |Libra not in sample after eff. date so no name adjustments |

| | | | |necessary. |

|  |  |  |  |  |

|19971117 |19980511 |Roney & Co, Detroit, Michigan |First Chicago NBD Corp |Roney replaced with First Chicago after eff. date. |

|19980413 |19981002 |First Chicago NBD Corp |BANC ONE Corp, Columbus, Ohio |First Chicago replaced with Roney Capital Markets (Banc One) |

| | | | |after eff. date. |

|19990414 |19990614 |Roney & Co, Detroit, Michigan |Raymond James Financial Inc |Roney Capital Markets (Banc One) replaced by Raymond James |

| | | | |after eff. date. |

|  |  |  |  |  |

|19971223 |19980612 |Ohio Co |Fifth Third Bancorp, Cincinnati |Ohio Co. replaced with Fifth Third after eff. date. |

|  |  |  |  |  |

|19980223 |19980630 |Cowen & Co |Societe Generale Securities |Both names replaced with SG Cowen after eff. date. |

|  |  |  |  |  |

Appendix B – continued

|Ann. Date |Eff. Date |Target Name |Acquirer Name |Name Adjustments |

|19980622 |19980827 |Midland Walwyn Inc |Merrill Lynch & Co Inc |Midland Walwyn not in sample after eff. date so no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19980629 |19980908 |Essex Capital Markets Inc |McDonald & Co Investments Inc |Essex not in sample after eff. date so no name adjustments |

| | | | |necessary. |

|19980612 |19981026 |McDonald & Co Investments Inc |KeyCorp, Cleveland, Ohio |McDonald & Co. replaced with KeyCorp after eff. date. |

|  |  |  |  |  |

|19980923 |19990216 |Van Kasper & Co |First Security Corp, Utah |Van Kasper replaced by First Security Van Kasper after eff. |

| | | | |date. |

|20000125 |20000428 |Black & Co Inc |First Security Van Kasper & Co |Black & Co. not in sample after eff. date so no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19980810 |19990326 |Scott & Stringfellow Financial |BB&T Corp, Winston-Salem, NC |Scott & Stringfellow replaced by BB&T after eff. date. |

|  |  |  |  |  |

|19981027 |19990401 |Interstate/Johnson Lane Inc |Wachovia Corp, Winston-Salem, NC |Interstate Johnson Lane becomes Wachovia after eff. date. No |

| | | | |name adjustments necessary. |

|  |  |  |  |  |

|19990609 |19990731 |Vector Securities Intl Inc |Prudential Securities Inc |New unit becomes Prudential Vector Healthcare - no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19990416 |19990813 |Butler Wick Corp |United Community Financial Corp |Butler Wick not in sample after effective date so no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19990325 |19990830 |Fechtor, Detwiler & Co Inc |JMC Group |Fechtor not in sample after effective date so no name |

| | | | |adjustments necessary. |

|  |  |  |  |  |

|19990928 |19991210 |Hambrecht & Quist Group Inc |Chase Manhattan Corp, NY |Hambrecht & Quist, Chase Manhattan, and Chase Securities |

| | | | |replaced with Chase H&Q after effective date. |

|  |  |  |  |  |

Appendix B – continued

|Ann. Date |Eff. Date |Target Name |Acquirer Name |Name Adjustments |

|19991101 |20000131 |Soundview Technology Group |Wit Capital Group Inc |Soundview not in sample after eff. date. Wit Capital changed |

| | | | |to Wit Soundview after eff. date. |

|20000515 |20001016 |Wit Soundview Group Inc |E*Trade Group Inc |E*Trade acquired brokerage business of Wit and created an |

| | | | |alliance between E*Trade/E*Offering Investment banking and Wit.|

| | | | |This is not an actual merger of the underwriting businesses, so|

| | | | |no name adjustments are needed. |

|  |  |  |  |  |

|19990928 |20000316 |Ragen MacKenzie Group Inc |Wells Fargo & Co, California |Ragen Mackenzie changed to Wells Fargo after eff. date. |

|  |  |  |  |  |

|20000830 |20001103 |Donaldson Lufkin & Jenrette |Credit Suisse First Boston |DLJ changed to CSFB after eff. date. |

|  |  |  |  |  |

|20000913 |20001231 |JP Morgan & Co Inc |Chase Manhattan Corp, NY |Eff. date is end of sample period so no name adjustments |

| | | | |necessary. |

|  |  |  |  |  |

|20000918 |20010105 |Wasserstein Perella Group Inc |Dresdner Bank AG |Eff. date is past end of sample period so no name adjustments |

| | | | |necessary. |

-----------------------

[1] Analyst coverage may be important before the offering date as well as in the aftermarket. Though quiet period restrictions precluded issuing public reports prior to and immediately after the offer during our sample period, co-manager analysts often shared projections and pricing formulas with institutional clients.

[2] For a more formal presentation of these arguments, see Booth and Smith (1986), Carter and Manaster (1990), Titman and Trueman (1986), and Balvers, McDonald, and Miller (1988).

[3] Forty-three of the sample’s IPOs include a joint book manager. In addition, 94 IPOs include one underwriter coded as joint lead and five IPOs have two underwriters coded as joint lead. For each of these cases, we verify the underwriter roles using the final IPO prospectus. Of the 43 joint book managers listed in the data, 34 are explicitly referred to as joint book managers in the final prospectus. The remaining nine cases could not be verified, but are not changed. Of the managers designated as joint leads, we found no cases in which this underwriter was explicitly referred to as a joint book-running manager. Throughout the paper, we therefore treat joint leads as co-managers rather than joint book managers.

[4] Underwriter roles were verified using the final prospectus in seven cases where an underwriter received the same underwriting allocation as the book manager but was not coded as a co-manager. This check resulted in corrections for four IPOs. We also verified underwriter roles in 53 cases where the syndicate included more than five co-managers or the syndicate was made up of more than 75% co-managers. This check resulted in ten corrections. For most of these cases, it appears that SDC mistakenly coded all syndicate members as co-managers.

[5] Loughran and Ritter (2002) document increased underpricing of IPOs in 1999 and 2000 and discuss possible explanations.

[6] Although not shown, we also examined whether changes in syndicate characteristics over time are caused by changes in the type of firms going public or in the identities of the book managers. These preliminary tests suggest that the changes in syndicate structure over our sample period are not explained by an increase in the number of or technology stocks or by different book managers.

[7] Hanley (1993) compares underpricing for IPOs that are priced above, below, and within the filing price range and documents that underwriters only partially adjust to information learned before the offering. The result is that underpricing is particularly severe for IPOs whose offer prices are adjusted upward from the filing range. Our analysis tests whether the extent of partial adjustment is related to syndicate structure and underwriter quality.

[8] Co-managers may also be hurt by self-serving behavior of book managers. The December 14, 1998 edition of Investment Dealers Digest provides an example. NationsBanc Montgomery was the book manager for a secondary stock offering for Flextronics International. Donaldson Lufkin and Jenrette, Merrill Lynch, and Morgan Stanley Dean Witter were all slated to serve as co-managers. After the deal was priced, NationsBanc decided instead to sell it as a block trade, and all co-managers were cut out of their fees. Syndicate members were described as “fuming”, “angry” and “shocked.”

[9] We also estimated the probability of being lead including industry and state dummy variables; where industries are defined using the four-digit SIC classification of Fama and French (1997). The conclusions based on this alternative estimation are similar, but the convergence of the lead probability models is problematic.

[10] Pichler and Wilhelm (2001) observe that relationships and reputation promote syndicate stability and may allow members to earn quasi-rents.

[11] As an alternative specification, we estimate OLS regressions for the proportional underwriting allocation received by each eligible underwriter. In comparison to the probit model that treats all syndicate members equally, this model has the advantage of giving more weight to those syndicate members who receive the largest underwriting allocations. In general, the results from this model confirm the findings from the probit models discussed above, but year dummies are insignificant.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download