What IPO Order Flow Reveals about the Role of the Underwriter

Michael Aitken Capital Markets Technology Chair University of New South Wales Sydney, NSW, Australia mai@cmcrc.com.au Frederick H. deB. Harris McKinnon Professor of Economics and Finance Babcock Graduate School of Management Wake Forest University Winston-Salem, NC 27109 rick.harris@mba.wfu.edu Thomas H. McInish Wunderlich Chair and Professor of Finance Institute for the Study of Security Markets Memphis, TN 38152 tmcinish@memphis.edu Kathryn Wong University of New South Wales Sydney, NSW, Australia k.wong@unsw.edu.au Current Draft July 2007 Abstract IPO underwriter-brokers dominate aftermarket trading but often follow rather than lead in price discovery. This suggests that the underwriter shares a certification, external monitoring, and signaling role with aftermarket brokers, venture capitalists, and founder-owners retaining equity. In this paper we investigate the cross-sectional determinants of the role of the underwriter in aftermarket price discovery. The underwriters role expands with greater issue uncertainty and diminishes with venture capitalist involvement and greater retention. Issue uncertainty is measured in pages of idiosyncratic risk factors and in the delay between the announcement and completion of the issue. Our first novel result is that verifiable facts are not a substitute for, but a complement to, underwriter certification and advice. Specifically, the underwriters contribution to price discovery increases with the number of supplier and customer contracts reported in the prospectus. Secondly, the underwriters role in price discovery declines when the IPO is first in a new technology or product space. These findings indicate that the verification process (not de novo information production) is the key function of the underwriter Keywords: Initial public offering, IPO, underwriter, venture capitalist Classification: G24, G32

We wish to acknowledge the constructive insights and helpful advice of Sara Moeller, Rob Nash, Terry Walter, John Parsons, Ron Guido, Victor Bivell (Editor, Australian Venture Capital Newsletter) and especially, Simon Perrott (Head of Equity Capital Raising at Merrill Lynch-Asia Pacific, and later, Head of Industrials, Chairman of the ABN Amro-Rothschild Joint Venture), as well as a dedicated knowledgeable referee and seminar participants at the University of New South Wales, Security Industry Research Centre-Asia Pacific, and the 2004 FMA Meetings in New Orleans. The research was partially funded by the Capital Markets CRC-Australia, the Australian Stock Exchange, and the Babcock School, Wake Forest University. All errors and omissions are our own.

Electronic copy available at: http://ssrn.com/abstract=1372753

What IPO Order Flow Reveals about the Role of the Underwriter
1. Introduction Underwriter-brokers dominate aftermarket trading of IPOs (Ellis, Michaely, and OHara 2000) with market shares often as high as 70% of the first-day trading volume, declining slowly to 45% - 55% over several weeks. An execution channel with such large segments of order flow would normally attract stealth trading by the most informed participants, whose transactions establish the new issues permanent price trends (i.e., the post-issue price discovery). In addition, of course, underwriter-brokers play the most active role in pre-marketing road shows and book building, activities that greatly influence the issuers offer price (i.e., the pre-issue price discovery). Despite the underwriters extraordinary pre-issue access and concentrated post-issue order flow, it turns out that both pre- and post-issue price discovery are quite diffuse. Aggarwal and Conroy (2000) show that underwriter-brokers follow rather than lead other brokers in Nasdaq pre-opening quotations, suggesting that underwriter-brokers play a secondary role in the learning that takes place to identify the underpricing equilibrium established by the initial trades. And post-issue, we show below that the underwriter-brokers order flow is not the sole source of statistically significant price discovery in 80% of the most actively-traded IPOs. To clarify the role of the underwriter, in this paper we examine cross-sectional variation in price discovery estimated from trade-to-trade data on aftermarket trading. All the order flow executed by the underwriter-broker (both proprietary and agency), we refer to as the underwriter execution channel. We find that in some IPOs, informed traders regularly emerge in the underwriter execution channel, and the underwriter channel then leads in price discovery. In other IPOs, both the underwriter channel and the non-underwriter channel contribute to price discovery. Finally, in still other IPOs, the non-underwriter channel leads the price discovery. To understand these crosssectional differences, we then examine a large sample of issue documents to identify the securityspecific determinants of these price discovery metrics. The logic of using microstructure information revealed in the aftermarket order flow to uncover new insights about the corporate finance role of the underwriter is as follows. Underwriter-

2
Electronic copy available at: http://ssrn.com/abstract=1372753

brokers verify facts claimed in the prospectus and advise liquidity traders about the underpricing premium they should be willing to accept or to pay in the aftermarket. Informed traders (whose actions establish all new permanent price trends) seek to trade stealthily in execution channels populated by large numbers of these liquidity traders. In new issues where the role of the underwriter in providing these advisory services to liquidity traders is extensive, we should expect price discovery to occur in the underwriter-brokers channel. This is exactly what we observe. We find that the underwriter-brokers contribution to price discovery is positively related to several traditional measures of issue uncertainty that underwriters can best explain--namely, the time delay from announcement to issue as well as the number and complexity of idiosyncratic risk factors listed in the prospectus. Two ownership characteristics, external monitoring by VC financiers and equity retention by founder-owners, which substitute for the verification processes and reputational effects of endorsement by the underwriter, are negatively related to the underwriters contribution to price discovery. When institutional investors with large positions in an IPO encounter VC financiers, underwriters are often heavily involved in advising about execution strategy, and price discovery in the underwriter-broker execution channel rises accordingly. One novel finding is that the number of supply chain and forward sales contracts that are listed in the prospectus and need to be verified is positively related to underwriter price discovery. In other words, verifiable facts about the isssuers business model prove to be a complement to, rather than a substitute for, underwriter certification and advice. Our second novel finding is that the role of the underwriter-broker in price discovery diminishes when the issuer is the first IPO in its technology or product market space. Unlike in seasoned issues, the certification, external monitoring, and signaling roles are widely dispersed in IPOs. Specifically, underwriter-brokers appear to share an ongoing certification, external monitoring, and signaling role with other aftermarket brokers, venture capitalists, and founder-owners retaining large amounts of equity. This dispersion of roles (especially in first IPOs in a product or technology space) leads informed traders to seek counterparties outside the dominant, underwriter execution channel.

2. Australian IPOs Our sample comprises the most actively-traded non-privitization IPOs underwritten during 1996-99 in Australia. Examination of Australian IPOs allows a much more continuous analysis of the role of the underwriter in information flows and price discovery because Australia does not impose a post-issue quiet period. Indeed, Australian underwriters are not prohibited from giving opinions about value at any time. Table 1 displays the frequency distribution by issue size for the 176 Australian nonprivitization IPOs underwritten during 1996-1999. Our Top 50 most-actively traded sample (listed in column 3) captures 8 of the 15 largest issues. This highest 9% of the size distribution exhibits issue sizes ranging from $81 to $385 million AUD. In the next size range from $21 to $80 million, the sample includes 20 of the 43 IPOs issued. As many as 90 of the 176 underwritten IPOs have issue sizes less than $11 million AUD, and our Top 50 most-actively-traded sample includes 17 such issues. We test whether issue size is an important conditioning variable in the model.

2.1. The Australian float process All IPOs in Australia are intertwined with the policies and procedures for listing on the Australian Stock Exchange (ASX), a screen-based order-driven electronic trading system since 1987. Unlike NASDAQ dealers who have an affirmative obligation to make markets in the stocks they underwrite, there are no official market makers on the ASX for either seasoned issues or IPOs. Nevertheless, like their NASDAQ counterparts, Australian underwriter-brokers do facilitate large block trading, advise on institutional execution strategy, and provide liquidity themselves in house stocks.1 For example, Australian underwriter-brokers are often consulted for several months postissue about indications of interest regarding lines of allocated stock to buy or sell.2 This facilitation of institutional trades serves to attract future capital raising/corporate finance business from IPO clients and related issuers. Analogous incentives influence NASDAQ dealer-underwriters. The ASX float process itself is quite similar to U.S. procedures, with one exception. After establishing eligibility and selecting syndicate members, the issuing firm and its underwriter prepare road show documents that include an indicative range of issue prices. For the ensuing ten days to two

weeks, the underwriter markets the issue to securities firms and institutional clients who offer informal guidance about the potential demand and sometimes about the draft prospectus itself. During a three-to-five day typical exposure period, the final prospectus (including a fixed price) is examined and registered by the Australian Securities Investment Commission (ASIC), and then the offer period begins.3 Formal marketing of the IPO, acceptance, and subsequent processing of

applications typically take two to three weeks. Once the offer period ends, the underwriter, in consultation with the issuing firm, allocates and distributes shares. Listing and quotation then ensue within several days. As noted earlier, one additional feature of the trading regulations subsequent to the ASX float process is quite distinct from the U.S. experience. Unlike the mandatory 25-day quiet period postissue in the U.S., there are no restrictions whatsoever on the timing of analysts forecasts and recommendations regarding an Australian IPO. Throughout the pre-issue and post-issue events, analysts recommendations emanate from all quarters, including the underwriter-broker. This regulatory difference in Australia poses a very favorable institutional environment for investigating information transmission and price discovery in order to reveal insights about the role of the underwriter in the pre- and post-issue market.

2.2 Similarities between U.S. and Australian IPOs The allocation outcomes of the Australian float process exhibit characteristics similar to those in the U.S.. Table 2, Panel A shows that Australian underwriters allocate only 68% of IPO shares to the top twenty shareholders, each purchasing over 100,000 shares. Like U.S. underwriters, Australian underwriters distribute the remaining shares widely, allocating almost 20% to large individual and institutional investors purchasing 5,000 to 100,000 shares, and the remaining 12% to retail investors purchasing less than 5,000 shares. In contrast, Australian practitioners tell us that a common practice in German and Japanese underwriting is to allocate almost the entire float in large block sales to a few financial intermediaries. Table 2, Panel B displays the commonalities between U.S. and Australian offer statistics and underwriting fees. We begin with the universe of all 214 IPOs that were listed between June 1996

and December 1999 and then provide separate comparisons of the 176 underwritten versus 38 nonunderwritten issues. These 214 Australian IPOs have a mean issue size of 116 million AUD, about three-quarters of the mean issue size of $99 million USD for all U.S. IPOs in this period.4 In general, the non-underwritten IPOs like the privatization Telstra II (which raised $14.6 billion AUD in 1999) are much larger than the underwritten IPOs (mean issue size $27 million AUD). Our Top 50 sample of the most-actively-traded underwritten issues has a mean issue size of $46 million AUD, distributed as follows: $161 to $385 million, 2; $41 to $160 million, 16; $11 to $40 million, 16; and $3 to $10 million, 16. Hence, all the IPOs in the bottom third of our sample are much smaller ($3 to $12 million) than the average for underwritten IPOs as a whole ($27 million). This allows us to split the sample and learn some insights about the differences in the role of the underwriter in large versus small actively-traded IPOs.5 Chen and Ritter (2000) define large IPOs in the U.S. as 325 issues with domestic gross proceeds over $80 million and moderate size as 1,111 issues with proceeds between $20 and $80 million. By these definitions, apart from privitizations, Australia in 1996-1999 had only four large issues and 42 moderate size issues out of 176 ASX-listed issues total. Our samples of the top 34 and top 50 most actively-traded IPOs from this 176 include all 4 large issues and 19 of the 42 moderate size issues for a total of 23 securities above $20 million issue size (AUD30 million), as listed in Table 1. The smallest IPO in our top 34 most-actively-traded sample had only AUD12 million ($8 million) in gross proceeds. However, we wish to emphasize three points suggesting that BGN, while below moderate U.S. size, is nevertheless comparable to Nasdaq small-cap issues: 1) The market capitalization of BGN at issue was $20 million and the publicly-held equity was $8 million. 2) All 50 of the IPOs we study were listed on the premier equity exchange (the ASX). Australian micro caps and best efforts that would not qualify for Nasdaq, also do not qualify for listing on the ASX. They trade on two other exchanges, the BSX and the NSX. 3) Fully 32 of our top 34 and 47 of our top 50 most-actively-traded IPOs remain listed or their acquirer remains listed on the ASX today, seven to ten years after issue (see attached spreadsheet of our top 50 sample). In general, Australian underwriting fees are 4%, which is reduced to 3% for issues over $100 million AUD and rises to 5% for issues under $20 million AUD. Beyond the underwriting fee, a 1%

management fee also applies in most cases (in 140 out of the 176 underwritten IPOs). An additional 1% handling fee arises in one-third of all IPOs (in 15 of the Top 50 most-active underwritten, 55 of the 176 underwritten, and 65 of the 214 total IPOs).6 Overall, then, an Australian IPO often incurs 6% underwriting + management + handling fees relative to the 7% solution reported by Chen and Ritter (2000) for U.S. IPOs. The 176 underwritten Australian IPOs experienced first-day median and mean price appreciation (offer to open) of 16.1% and 32.9% (not shown), respectively. This compares to firstday median and mean returns in the U.S. IPOs 1996-1999 of 16.7% and 26.1% (compiled from Loughran and Ritter, 2004, Table 1).7 However, this was an unusual period for IPOs and may not reflect the long-run level of underpricing in Australia versus the U.S. We wish to explore whether the lack of stabilization in Australia affects first day returns. We do so by comparing our sample to a sample from RUUD (1993). Before presenting the analysis, it may be useful to consider a few limitations of the analysis. One difficulty is that the distribution of first day offer to close returns is affected both by underpricing and stabilization, if any. While there is no legal stabilization in Australia, the substantially higher underpricing should reduce the number of first-day negative returns. Australian underwriters reduce the need for stabilization through greater underpricing. But if there is an absence of stabilization in Australia, we expect that the first day intraday returns from first trade price to closing trade price will be substantially less skewed. Statistics for 463 IPOs taken from Rudd (1993, Table 2, p. 146, column 2) are reproduced in Table Panel C., column 2. Comparable statistics for our sample are presented in column 3.

Examining the return from offering price to first-day close, we see (Table 3, Panel C) that there is substantially more underpricing in the Australian sample. Ruud (1993) also reports statistically significant skewness and kurtosis, which, as we have said, she attributes to a dearth of negative return due to underwriter stabilization. Both skewness and kurtosis are higher for our sample. But examination of first-trade-to-close returns show that the distribution exhibits considerably less skewness and about the same level of kurtosis. We think that this measure of skewness is particularly relevant for our sample because it reflects stabilization only rather than the effects of both

stabilization and underpricing. The effects of underpricing on returns may not be fully dissipated with the first trade. We interpret this evidence as supportive of the lack of stabilization in Australia.

2.3 Structure of the underwriting industry in Australia Most investment banks in Australia have brokerage relationships with the Australian Stock Exchange (e.g., UBS-Warburg, Deutsche Bank, ABN Amro, Macquarie Bank, and Merrill Lynch). When an underwriter has no brokerage relationship with the stock exchange, a sponsoring broker usually joins the syndicate. Out of the 176 underwritten IPOs, there were 150 cases of a single underwriter-broker as sole member of the syndicate, 24 cases of single underwriter-brokers accompanied by sponsoring brokers, and 2 cases of underwriters and lead managers (see Figure 1). In Australia, a few non-underwritten privatizations dominate the league tables. For example, the Telstra III issue is expected to raise $27 billion AUD, more than half of the projected total equity issues for 2006. This privitization activity is highly concentrated with the top three brokers handling 21% (8), 13% (5), and 13% (5) of the 38 non-underwritten issues 1996-99, respectively. In contrast, the underwriting industry in Australia is fragmented. There were 51 different underwriter-brokers involved in various syndicate roles in the 176 underwritten issues 1996-99. The three leading underwriter-brokers accounted for only 7% (13), 7% (12) and 6% (10) of the new issues, respectively; most underwriters had 1-3 issues.

2.4 Descriptive statistics Our IPO information comes from the Securities Data Corporation New Issues Database, the Australian Stock Exchange IPO reports, and individual firm prospectuses. Intraday price and trade data for the first 120 days post-issue are extracted from the Stock Exchange Automated Trading System database obtained from the Securities Industry Research Centre of Asia Pacific (SIRCA). In addition, the ASX provided proprietary broker identifications from their audit trail data. This allowed an important advantage offered by our study. Because we can view the entire order book and trade schedule for each IPO and track orders and trades, we can identify two distinct execution channels: the underwriter channel and the execution channel of all the other brokers. Lead managers and

sponsoring brokers are seldom involved in the underwriter due diligence or allocation processes in Australia, so we defined the underwriter channel as order flow executed through the underwriters broker only. Our detailed analysis is limited to the Top 50 most-actively-traded issues from the 176 underwritten IPOs 1996-99. This focus on actively-traded IPOs occurs in order to estimate accurately CFu, our dependent variable for the cross-sectional analysis, which involves a time series-based price discovery technology that requires synchronous trades from the underwriters and other brokers. Executions too far apart in clock time cannot be used to estimate which channel leads the market prices back to arbitrage-free equilibrium following a shock. That is, all thinly-traded IPOs have statistically insignificant price discovery metrics for our dependent variable. Although thinly-traded IPOs do have smaller capitalizations,8 our research should be understood to apply to both larger and smaller floats; they simply must be actively-traded. Finally, most of our Top 50 IPOs are concentrated in one year: 1999 with 78% as opposed to 1996, (10%), 1997 (6%), and 1998 (6%). Similarly, Morgan-Stanley data show that 79% of the largest U.S. IPOs 1996-1999 also occurred in 1999. Ritter (2006) shows that U.S. new issues of all sizes peaked in 1996 with 599 IPOs,but that the 472 and 476 new issues in 1997 and 1999, respectively, were the next highest years of the last decade. Lowry and Schwert (2002) analyze the when-to-issue decision. Ellis, Michaely, and OHara (2000) report that for Nasdaq stocks, the lead underwriter is always the dominant market maker, taking substantial inventory positions and handling as much as 50 percent of the trading volume during the first few months of the aftermarket. Table 3, Panel A, shows that the underwriter-brokers market share in Australia averages 58.5% on the first day of trading and remains as high as 52.1% even forty-five days after issue. For the Top 50 IPOs, Panel A and Figure 2 display the 75.5% first day trading volume declining to no less than 63.7% after 7 days and 51.2% after 45 days. Hence, underwriter-brokers also dominate aftermarket trading in Australia.9 Some of the underwriter-broker trading volume in the U.S. is due no doubt to stabilization activities.10 In Australia, aftermarket price support is seldom authorized and rarely offered. Indeed, prior to 1992, all IPO price stabilization activity by underwriter-brokers was prohibited by the ASIC.

Exceptions have been granted in recent years primarily for privitizations. Hence, the institutional history in Australia is one in which underwriter-brokers have not been expected to offer price support services. Despite this diminution of their role, underwriter-brokers dominate the aftermarket trading in Australia, just like in the U.S. For selected stocks in our actively-traded sample, Table 3, Panel B, presents several Australian issue characteristics, each exhibiting substantial cross-sectional variation. The 33 underwritten IPOs with the largest issue sizes have institutional ownership from 3% to 92% (mean 62.5%), equity retention from 4% to 87% (mean 54.9%), and pages of risk factors in the prospectus from 48 to 178 pages (mean 102.9).11 The number of contracts with suppliers and customers listed in the prospectus varies from 0 to 14 (mean 6.2). On average, small IPOs have substantially smaller

institution ownership (mean 44.2%), fewer pages of risk factors (mean 73.0), and less contracts (mean 3.2). The standard deviations are comparable with one exception; large IPOs have much greater variation in contracts announced in the prospectus. The industry breakdown for the Top 50 issues is presented in Figure 3. These fifty most actively-traded IPOs come from 13 different industries. Especially prominent sectors were biotech, media, industrials (especially computers and computer services), retail, and telecommunication. Telecoms are over-represented, both in Australia and in U.S. data during this time period. Nonunderwritten privitizations arose primarily in banking, transportation, and telecommunication.

3. Common factor weights for the underwriter channel: A microstructure dependent variable. Despite the dominance of the underwriter-broker in aftermarket trading, prior evidence suggests that the underwriter-broker clearly shares the pre-issue price discovery role with other brokers. Aggarwal and Conroy (2000) examine the role of underwriters in the pre-open period for Nasdaq IPOs. The authors report that the lead underwriter and co-managers account for only 37% of the bid improvements and just 8% of the ask improvements during the pre-open period. In order to understand more fully the cross-sectional differences in the role of the underwriterbrokers, we begin by estimating directly the relative contribution to price discovery of the underwriter-brokers versus other brokers. We estimate the order of integration properties of these

10

series with augmented Dickey-Fuller statistics on the first-order serial correlation coefficients. The two prices series can be cointegrated and common factor components estimated only if they are the same order of integration. Since these are IPOs rather than seasoned securities in fully efficient capital markets, it proves important to test for and condition the order of integration tests on drift and deterministic time trends (Enders, 2005, pp. 185-187). Prior to cointegration testing, we estimate the optimal lag length using TSULMAR, a systems forecasting tool that minimizes the Akaike information criterion for the lag length in the system of lead underwriter price and other broker price equations. These lag lengths must be neither too short nor too long in order to correctly diagnose the presence or absence of cointegration and accurately estimate the common factor component for the lead underwriter in subsequent procedures. At an optimal lag length for each stock, we use Johansens reduced rank regression procedure to test the log price series for one cointegrating vector versus the null hypothesis of zero cointegrating vectors. Using critical values obtained from Enders (1995, Table B), the maximal eigenvalue rejects the null hypothesis in 49 out of 50 cases.12 so we conclude that the continuous return (and by

implication the price) series for underwriter-brokers and other brokers are in fact cointegrated.13 Having confirmed cointegrated underwriter-broker and other broker price series in 49 out of 50 cases, we then employ Gonzalo and Grangers (1995) common factor components procedure (the GG procedure) to estimate and test the proportion of the common stochastic trend

w
t =1

attributable

to the underwriter-broker versus other broker trades (see Table 4). GG restrict these common factors to linear functions of the current observable prices, and they restrict transitory disturbances to not Granger-cause the permanent information arrivals wt,. DeJong (2002), Ballie, Booth, Tse, Zabotina (2002), Harris, McInish, and Wood (2002a and 2002b), Hasbrouck (2002) and Huang (2002) debate these restrictions. However, our main interest in this paper lies elsewhere--i.e., with the corporatefinance, security-specific characteristics that explain the cross-sectional variation in this price discovery metric. In the second and fourth columns of Table 4, we report for each IPO the appropriate eigenvalues normalized as common factor weights [f1, f2], which we interpret as the proportion of

11

permanent price adjustment contributed by each execution channel (see Harris, McInish, and Wood, 2002a). In BNO, for example, the underwriter-broker executed trades that contributed 73% of the permanent changes in the price trends in that stock; other brokers contributed 27%. Both parameters are statistically different from zero. Single or double asterisks on the chi-square statistic in the third and fifth columns indicate, respectively, 90% or 95% significance in likelihood ratio tests of the null hypothesis H0: f = [0,1] against the one-tailed alternative Ha: f1 >0 and f2<1 -- i.e., both common factor weights are statistically significant. The fact that there is no post-issue quiet period in Australia during which the underwriterbroker must stop advising clients about changes in material facts might lead to the conjecture that the underwriter provides the only game in town. However, the parameter magnitudes and test statistics in Table 4 tell a quite different story. Not only is the underwriter not information dominant in all cases, but rather, there are even some IPOs in which the underwriter plays a subsidiary role or no role at all in the post-issue price discovery. For example, in IAS (in Panel A), the underwriter proportion of the price discovery is only 44%, and in BMC and REA (in Pane C), the underwriter plays no role with statistically insignificant 16% and 5% common factor weights, respectively. Most relevant to our motivation for examining the cross-sectional determinants of the role of the underwriter in price discovery, the parameter estimates of these common factor weights in Table 4 exhibit substantial cross-sectional variation. In the top half of the table (Panel A), both the underwriter and the other brokers contribute to price discovery. In AAP and AML and WPO, for example, the common factor weights are nearly equal. In the next five IPOs listed, both channels contribute, but the underwriter dominates so that firms such as BNO, CAB, CDO, CLT, and GTP have common factor weights for the underwriter of 69% to 83%. These price discovery parameters approximate the share of the dollar volume these underwriters control, much like the role of the NYSE in trading listed equities (Hasbrouck 1995). In the ten IPOs in Panel B, the underwriter provides the only execution channel in which there is price discovery. Stoughton, Wong and Zechner (2001) predict such an extensive role for the underwriter in first IPOs in fast-changing high technology industries where an underwriters reputation for certifying successful past issues substitutes for verifiable product and service quality

12

information about the offering firm. When few analysts thoroughly understand the technology, the reputation effect of the underwriter for effective due diligence and information processing can be especially important in attracting counterparties to trades that may involve new information. Seven of these ten IPOs are just such issues in software development (HRD), dot-coms (MLB, PLX), telecommunication/ISP services (OTT, CSB, EIS) and medical therapy services (LCH). In nine IPOs in Panel C (e.g., ANX, BRZ, COM, and REA), the underwriter provides no price discovery while the other brokers do so, a curious result indeed given the usual asymmetry of accurate information available to the underwriters customers. REA was a real estate IPO that exhibited little asymmetry of information relative to the typical issue. Finally, six IPOs--ALL and EBT (Gaming equipment and wagering), BDA and EZE (Electronic equipment), BMC (Media), and LIB (Telecommunications)--exhibit only noise trading in both channels with no statistically significant price discovery. BMC was a media dot.com IPO handled by a small brokerage firm with less than 1/5 of 1% of the Australian market in equity capitalraising. Subscriptions were few, and we conjecture that the weak reputation of the underwriter was insufficient to attract liquidity traders as a counterparty for informed trades.

4. Cross-sectional determinants of price discovery in the IPO aftermarket 4.1. Relevant literature In a study of NASDAQ IPOs, Ellis, Michaely, and OHara (2000) report that the lead underwriter is always the dominant market maker in the aftermarket. Aggarwal and Conroy (2000) show however that during the first day pre-open period, the lead underwriter is the least active of all the brokers in quote submission. Our study extends this literature by investigating the role of the lead underwriter in aftermarket price discovery. Stoughton, Wong, and Zechner (2001) analyze the underwriter certification of the issue, focusing on product quality. The certification role of the underwriter, especially in first IPOs in high technology industries, substitutes for retained equity and/or venture capital involvement in externally monitoring the start-ups business plan. We test these propositions directly with a cross-section of the

13

Australian IPOs using as the dependent variable our time-series measure of the price discovery in the underwriter execution channel.

4.2. An overview of theoretical models The multiple roles of the underwriter in the pre-issue and post-issue market can be clarified by a three-period IPO signaling model. In period 1 (see Figure 4), a new venture secures start-up and mezzanine financing and moves toward an initial public offering. Venture capitalists may or may not choose to invest. If so, VCs help develop the business plan, assist in the formation of the management team, and provide strong external monitoring. These VC involvements typically lower the underpricing at issue (Megginson and Weiss, 1991). VC financing is extremely costly to the ownerfounders with each of the seed round and mezzanine round financings requiring assignment to the VC of 25-33% of the company. Therefore, the earlier the VC invests and the later the VC exits, the stronger the signal of issue quality from any subsequent decision to proceed to an IPO.

14

Stochastic processes (N) complicate the assessment of issue quality in the middle of the sequence of events in Figure 4. Some start-ups become low quality issuers (LQSU) despite venture capitalist involvement. Other start-ups emerge as high quality issuers (HQSU) without VC

involvement. High quality start-ups (HQSU) always decide to do IPOs, while lower quality start-ups (LQSUs) chose between mimicking the high quality issuers or reverting to private equity financing. Underwriters play two key roles in this pre-issue period: they conduct much of the price search (Aggarwal and Conroy, 2000), and high reputation underwriters withhold certification of lower quality issues. At the IPO (displayed as three decisions on the far right of Figure 4), underwriters advise issuers about the proportion of equity to retain Eq0, then set an offer price

P0-

in the registered

prospectus, and soon thereafter allocate the shares to subscribers. By retaining more equity, founderowners of HQSUs seek to discourage founder-owners of LQSUs from attempting to mimic. Equilibrium P0- underprices the issue just enough to offset the uninformed (liquidity) traders future expected losses in buying an issue of certified but still unknown quality. In period 2, informed traders (IT) detect the true issue quality and begin secondary market trading on this information, buying HQSUs and selling LQSUs (in the asymmetrically-informed dashed-line box at the left-hand-side of Figure 5). ITs seek to stealth trade in execution channels populated by large numbers of liquidity traders (LTs) who can serve as counterparties to low price impact trades. The market price that emerges from this stealthy order flow eventually reflects a separating equilibrium between high and low quality firms. But price is not fully revealing

immediately because the informed trader trades strategically.14 Specifically, informed traders attempt to find liquidity traders (LTs) from whom they can buy HQSUs at a discount or to whom they can sell LQSUs at a premium. Knowing this, liquidity traders seek out advice from the underwriter to mitigate the expected losses from being less informed. As a result, as we show in the next section, LTs chose to trade in the aftermarket through execution channels where they have access to underwriter advice.

15

4.3. Implications for execution channel choice In the Non-U (Non-Underwriter) execution channel depicted in the NW cell of Figure 5, informed traders attempt to buy HQSUs from liquidity traders at the issue price P0- and attempt to sell LQSUs to liquidity traders at P0 . In contrast, having obtained underwriter advice, liquidity traders who acquired HQSUs at issue are advised to sell at P0 (not at P0- ), thereby pocketing the equilibrium underpricing premium ( P0 - P0- ) on their allocated shares. These transactions are depicted in the SE cell of Figure 5, where they are referred to as the Underwriter (execution) Channel. No trade occurs in the NE or SW cells of Figure 5 other than through arbitrage activities of intermediaries. In a one-shot simultaneous game, the liquidity trader in Figure 5 has a dominant strategy to execute in the underwriters execution channel. LTs earn a positive cash flow that represents an equilibrium risk premium for bearing the adverse selection risk from offer to issuance. Anticipating this, the informed traders have an iterated dominant strategy equilibrium also to execute in the underwriter-brokers channel. This post-issue equilibrium leads to the testable hypothesis that postissue price discovery will occur predominantly in the underwriters channel. In period 3 (not shown), differences between HQSU and LQSU issuers become apparent, and liquidity traders become fully informed. Only then does market price become fully revealing and impound all public information about future price trends. Price discovery in the distant aftermarket should thereafter occur in all execution channels. However, prior to period 3, price discovery will occur predominantly in the underwriter-broker channel where informed traders will find sufficient numbers of liquidity traders to act as counterparties for medium-sized to larger trades.

4.4. Role of the underwriter The lead bank/underwriter/broker is always informed and knows which startups are LQSU versus HQSU. He (2007) develops a more complete theory to explain the selection of the underwriter in a multi-period game with reputation effects. More reputable underwriters allow HQSUs to signal their higher quality by retaining less equity than they otherwise would need to in the one-period game, by underpricing more deeply at issue, and by allocating to repeat purchase institutional customers of the

16

bank/underwriter/brokers. Lead bank/underwriters/brokers are always in a position to identify fair value in after-market trading. What attracts informed trades to the underwriter execution channel thereby triggering price discovery is the presence of large numbers of uninformed counterparties who seek out the advice of the underwriter to avoid being picked off. Uninformed traders approach the bank/underwriter/broker for advice about the size of a premium to pay (P+) relative to the issue price for an HQSU and the size of a discount (P-) to accept for selling an LQSU. learly, the underwriter is informationally advantaged from due diligence, but uninformed investors worry about the extent to which underwriters are beholden to those that have received allocated shares and seek to profit by flipping the stock. Therefore, we would expect the bank/underwriter/brokers role in price discovery to increase with distance from the immediate post-issue period of intense flipping activity. During this intermediate post-issue time period, the role of the underwriter should vary with the cross-sectional characteristics of the IPO. For example, when verifiable facts about a new

ventures business plan are present, liquidity traders approach the underwriter to assess and interpret those facts. This advisory service establishes the amount of underpricing premium a liquidity trader who receives allocation of an HQSU should expect to realize in post-issue trading. In contrast, when few facts are verifiable as with the first IPO in a new technology or product space, Stoughton, Wong, and Zechner (2001) predict that issue quality will be certified by the high reputations of underwriters entrusted to perform advisory services in other IPOs. Alternatively, in this case the issuers would have to retain more equity in order to establish the credibility of their claims regarding issue quality. So, in theory, IPO issuers have two distinct reasons to employ underwriters who perform their post-issue advisory services well: 1) underwriter verification of complex facts reduces the underpricing required to offset adverse selection risk, and 2) underwriter reputations support claims of a high quality issue involving technological firsts when only extraordinary retention would substitute. In addition to the cross-sectional differences in verifiable facts and technological firsts, advisory services available through the underwriter are more valuable to liquidity traders the greater the issue quality uncertainty, the smaller the proportion of equity retained by the founder-owners, and the less the venture capital involvement.

17

We therefore hypothesize that discovery of permanent price trends in the IPO aftermarket will occur more frequently in the underwriter channel, 1) the larger the number of customer and supplier contracts to be verified (CONTRACTS), 2) the smaller the proportion of equity retained (RETAIN), 3) the less the involvement of venture capitalists (VC), 4) the greater the number of risk factors enumerated in the prospectus (PAGES), a measure of issue uncertainty, 5) the greater the time delay from announcement of the offering to completion of the issue (DELAY), another measure of issue uncertainty, and 6) the smaller the number of other IPOs by the underwriter (OTHERIPOS), which measure the strength of the reputation effects supporting pre-issue endorsements, a substitute form of underwriter certification. In addition, we investigate two important conditioning variables: the percentage of institutional ownership (INSTOWN) and whether the IPO is the first in a new technology or product space (FIRST).

4.5 Empirical results Table 5 reports the cross-sectional determinants of CFU, the common-factor weight in price discovery attributable to synchronous trades in the underwriter execution channel estimated across the Top 50 most-heavily traded IPOs. All four specifications of the model explain 42 to 52 percent of the variation in the normalized dependent variable CFU with F tests all statistically significant at = 0.01.15 The dependent variable is a normalized factor weight between zero and one, the ratio of an eigenvalue for one execution channel relative to the sum of the eigenvalues for both channels. The right-hand-side variables are classified into four types of measures i.e., verifiable facts, issue uncertainty, reputation effects, and ownership characteristics. We seek to ferret out the primary antecedent effects of these independent variables on the decision by uninformed liquidity traders to seek out the underwriter-broker for counsel as to execution strategy in the post-issue aftermarket. As

18

we have argued, one prominent reason for this advisory and brokerage relationship is that liquidity traders are exposed to picking off risk as counterparties to informed traders whose executions will thereafter establish the new permanent price trends. Verifiable facts Salient facts that liquidity traders may need verified, assessed, and

interpreted are measured by the number of supply chain or customer relationships identified in the prospectus that have been formalized into contractual agreements (CONTRACTS) and by a dummy variable for the first IPO in a new technology or product space (FIRST). The number of contracts is significantly positively related to price discovery through the underwriter execution channel. Interpreting the first row of Table 5, the larger the number of contracts claimed in the prospectus, the greater the value to liquidity traders (and hence, to informed traders seeking counterparties in that channel) of due diligence verifications performed by an underwriter. Here we distinguish between certification of an IPO by an underwriter who attests as to verifiable facts in a prospectus versus certification that arises by allowing an issuer to affiliate with the reputational assets of the underwriter. When there is a delay from the expected issue timetable, the role of the underwriter does expand. In this situation, the due diligence of the underwriter allows the release of factual

interpretations and additional information that assures prospective investors. Other brokers are not in a position to make such assurances. In the case of an IPO that is first in a technology space, however, there often is little or no basis for the verification of product or technology claims. We interpret our results to imply that verification, not information production, is the central function of the underwriter. Stoughtons (2001) theoretical model also implies that the role of the underwriter in assuring investors will be negatively related to IPOs that are first in a technology space. In Table 5, row 2, we find that the role of the underwriter in aftermarket price discovery is significantly inversely related to FIRST. Eighteen of the IPOs in our Top 50 sample 1996-99 were first in their technology or product space. Very few verifiable facts are available in a new technology or product space and, as we have seen, the underwriters due diligence about such facts is essential to the underwriters role as an advisor in the aftermarket. This inverse relationship between the role of the underwriter in price discovery and FIRST holds whether or not one controls separately for the

19

reputation effect of underwriters in certifying other IPOs (se specification 3). FIRST is negative and significant even when one includes (in specifications 2 and 4) the number of other issues underwritten (OTHERIPOS) by this underwriter-broker. However, consistent with Stoughton, Wong, and Zechners (2001) prediction, we find that the underwriter certification of issue quality is particularly important in some IPOs that are first in a new technology or product space. Specifically, delay between announcement and issue (DELAY) exhibits a statistically significant interaction term with FIRST. Interpreting rows 3 and 4 in Table 5, the role of the underwriter in resolving delay-based issue uncertainty is accentuated two-fold by the presence of a first IPO in a new technology or product space. For example, in specification 1, CFU/ DELAY = + 0.0142 + 0.0127 FIRST in specification 1, which equals + 0.027 when FIRST = 1 whereas CFU/ DELAY = + 0.014 when FIRST = 0.16 Notice that these results are almost identical in magnitude across all four specifications of the model. Issue uncertainty

The size of the offer (ISSUESIZE in the bottom row of Table 5) is

sometimes used to measure issue uncertainty indirectly. In all four specifications, ISSUESIZE is positively related to the role of the underwriter in after-market price discovery CFU, though it is very small in magnitude by comparison to CONTRACTS and FIRST. Like Ljungqvist and Wilhelm (2002), we refrain from adopting issue size as our only proxy for issue uncertainty on the grounds that issue size is the endogenous consequence of the equilibrium underpricing decision. Instead, in the spirit of establishing the primary, antecedent effects that correlate with issue size, we control for issue size and then measure issue uncertainty directly with two additional variables: DELAY (the number of days after the IPO announcement until the issue date) and PAGES (the number of pages in the prospectus required to enumerate the applicable risk factors).17 Delay between announcement and issuance suggests business plan complexity, unfinished due diligence, or occasionally signals substantial mispricing. Delaying the issue can result from the uncovering of disruptive technologies (during due diligence) that threaten the issuing firms business plan. Such competing technologies may have been thought to be incremental when the indicative range of prices was set, but later turn out to be a roadblock to the issuing firms business model.

20

Sometimes even complementary technologies prove to be easily morphed into adaptations that substitute for what the issuers propose. Such stumbling blocks result in substantial delays that only the underwriter can resolve by seeking fairness opinions and comfort letters from intellectual property experts and auditors. We expect an expanded role of the underwriter in explaining and assessing such matters for liquidity suppliers in the immediate post-IPO aftermarket. DELAY (in row 3 of Table 5) is positive and significant in all specifications of the model. In Table 5, row 5, we find that the DELAY variable is a negative quadratic, declining in its positive proportional effect on the role of the underwriter in price discovery as the length of delay increases. This result may be necessary for the separating equilibrium between high quality issuers and lower quality private equity financing. If it were not so, and lower quality firms could reduce the underpricing required to attract liquidity traders by simply delaying the issue and offering more underwriter advisory services, said lower quality firms would mimic high quality issuers rather than revert to private equity financing. Thus, speaking cross-sectionally across the Top 50 sample, the role of the underwriter peaks when an issue is off schedule several weeks, and thereafter each additional week results in a smaller increase in the role of underwriter. The second direct measure of issue uncertainty we adopt is the risk factors listed in the prospectus. PAGES captures the issuers attempt to secure a safe harbor against subsequent nondisclosure litigation by listing all the relevant risk factors that could be material to the issuers cash flows and valuation. This enumeration of specific risk factors embedded in the prospectus boilerplate must be accomplished without dampening excessively the issue valuation. The resulting rise in issue uncertainty should be positively related to the value of the underwriter in making assessments for liquidity traders shortly after the issue. We find in row 6 of Table 5 that the PAGES variable is

positively related to aftermarket price discovery in the underwriter execution channel in all specifications of the model. However, PAGES interacts with CONTRACTS to reduce the liquidity trading volume that attracts informed traders into the underwriter channel. Specifically, in Table 5, row 7, we find that higher numbers of enumerated risk factors or more complexity in risk factors necessitating more thorough explanations, reduce the effectiveness of an underwriter in verifying facts that certify an

21

IPO. As a result, large liquidity traders (especially institutional traders) may seek out opinions from lawyers and then execute through independent brokers not directly affiliated with the IPO. This increase in the likelihood of a liquidity traders executing with other brokers, decreases the number of informed traders seeking counterparties in the underwriter-brokers channel. Reputation effects We next discuss the effect of underwriter reputations achieved outside

the present IPOs arrangements. OTHERIPOS (the number of other IPOs by the underwriter within the same year)18 captures the effect of a reputational asset. The larger the number of other concurrent underwritings, the more valuable the hostage conveyed to prospective buyers of the IPO, assuring against casual or fallacious certification of issue quality by this underwriter. A priori, in lieu of verifiable facts about the likely success of the proposed offering, a high-quality issuer would especially seek out an endorsement by a high reputation underwriter, and this costly endorsement would serve as a signal of high quality on its own. Hence, underwriter reputation effects in the preissue period would substitute for the underwriter-brokers verification and assessment advisory services to aftermarket brokerage clients. We therefore expect that the larger the number of other IPOs that have been entrusted to any given underwriter, the smaller the factor weight of that underwriter in post-issue price discovery: CFU/OTHERIPOS < 0. In specifications 2 and 4 of Table 5, OTERIPOs proves insignificant. However, these results should not be misinterpreted as a displacement of the role of reputation effects in underwriting. Instead, in the first column (specification 1), we see that OTHERIPOS and CONTRACTS are correlated with a marginally significant positive interaction term. A higher external reputation of the underwriter assists the issuer in securing supply chain and customer contracts that then must be underwriter verified to confirm the development of an IPOs business plan. Once, we account for this interaction, the predicted negative effect of OTHERIPOs as an endorsement mechanism that substitutes for verifiable facts is statistically significant. The composite effect of reputation directly through endorsements in the pre-issue market and indirectly through more contracts named in the prospectus leaves a net negative effect on underwriter verification and advice in the aftermarket -CFU/OTHERIPOS = -0.0364 + 0.0057 (mean CONTRACTS) = -0.0364 + 0.0057 (5.18) = -0.069.19

22

An underwriters pre-issue endorsements or certifications about verifiable facts are not the only ways to establish issue quality. Issue quality can also be signalled by the presence of a venture capitalist as an external monitor (VC), by substantial buying interest on the part of institutional investors (INSTOWN), or by greater retained equity as a choice decision of the founder-owners (RETAIN). In the Top 50 sample reported in Table 5, these ownership characteristics were generally insignificant. In splitting the sample between large and sample IPOs, however, we detected a substantial influence of all three variables on CFU in large IPOs.

5. Discussion of split-sample results Recall that Table 1 displays the frequency distribution of issue sizes for both large and small IPO sub-samples and for the population of all Top 50 actively-traded underwritten IPOs. Finding that ISSUESIZE was a significant scaling variable in all specifications of Table 5, we tested for whether large and small IPOs could be validly pooled. We proceeded with two splits for robustness. In one case we reestimated the model with half of our Top 50 sample (25 IPOs) with issue sizes from $365 to $30 million AUD to represent large IPOs (columns 1-3 of Table 6) and half of our Top 50 sample (25 IPOs) with issue sizes $29 to $3 million AUD to represent small IPOs (in the next to last column of Table 6). In the other case, we split the Top 50 sample into 33 IPOs with issue sizes from $365 to $12 million AUD (columns 4-5 of Table 6) and 17 small IPOs with issue sizes from $12 to $3 million AUD (in the last column of Table 6). The Chow test statistic for the (2/3rd, 1/3rd) split yields F15,20 = ((0.95624 - 0.16927 - 0.02485)/15) / (0.19412/20) = 5.23 relative to a 3.09 critical value at = 0.01 indictating that the large and small IPOs should not be pooled. The (, ) split also indicates the two sub-samples should be analyzed separately. In performing the split-sample analysis, all our previous findings on verifiable facts in the full sample were confirmed for both large and small IPOs. The number of contracts to be verified and first IPOs in a new technology or product space raised and lowered, respectively, the role of the underwriter-broker in the aftermarket. In addition, delays between announcement and issue enhanced the role of the underwriterbroker in providing advisory services for aftermarket participants in large IPOs. However, the role of

23

the underwriter-broker in small IPOs did not increase with delays in the announcement to issue schedule (see DELAY in the first five versus the last two columns of Table 6). Instead, if anything, with issues under $12 million AUD (about $8 million USD), the role of the underwriter declines when the IPO got off schedule. Moreover, while delay had diminishing quadratic effects in large IPOs (in columns 1-5 of Table 6), DELAYSQR proved positive and significant in small IPOs (see columns 6 and 7). Delays between announcement and issue increase the role of the underwriter exponentially whenever the issue size is too small for other brokers to justify learning in depth what complexities have arisen. In addition, we found in large IPOs that the role of the underwriter in verifying CONTRACTS interacts with a wide range of important conditioning variables (FIRST, DELAY, PAGES, and OTHERIPOs) without which the true relationship between price discovery in the underwriter channel and CONTRACTS can not be detected. For example, where customer and supplier contracts do exist in first IPOs, the conditioning effect of FIRST on the role of the underwriter interacts with contracts to be verified. Interpreting the coefficients on

CONTRACTS*FIRST in the first and third rows of Table 6 conveys that the partial derivative effect of CONTRACTS is +0.0812 (i.e., +0.1128 - 0.0316) when FIRST = 1 and 39% larger (+0.1128) when FIRST = 0 -- i.e., when the IPO is not first in a new technology or product space. That is, again, the difficulty in verifying claims made in the prospectus when an IPO is FIRST reduces the value of the underwriter advisory services in the aftermarket. In the last two rows that address issue uncertainty variables in Table 6, we find that pages of risk factors is positively related to the role of the underwriter-broker in providing after-market advisory services that ultimately attract informed trades. We also find that the PAGES variable interacts with CONTRACTS in large IPOs in the same way that FIRST interacts with CONTRACTS. That is, the positive effect of contract verification by the underwriter on price discovery in the underwriter channel is reduced by both new, unproven technology and by page after page of idiosyncratic risk factors listed in the prospectus. From the second column of Table 6, we see that the partial derivative of CONTRACTS is CFU/ CONTRACTS = + 0.1128 0.0316 FIRST 0.0004 PAGES = + 0.0812 0.0004 PAGES. Because of the two negative and statistically significant

24

interaction terms, CONTRACTS can have a negative total effect in FIRST issues if the enumerated risk factors stretch beyond 203 pages. In the 25 largest IPOs, the mean number of PAGES is 108, and the maximum of PAGES is 178.20 So, a positive net composite effect of contracts that need to be verified on the role of the underwriter in the aftermarket price discovery is assured. Ownership Characteristics IPOs monitored, financed, and allocated with venture

capitalists (VC) involved and with a higher percentage of the equity retained by the founder-owners (RETAIN) will have less need to certify issue quality through underwriters. The negative and significant parameter estimate on VC in Table 6, across the first five columns in Table 6 for large IPOs, suggests that the prior external monitoring of an IPOs business plan by a venture capitalist can indeed substitute for the role of the underwriter in providing advisory services in the aftermarket. This effect of VC involvement on the role of the underwriter is influenced, however, by the percentage of institutional ownership at issue (INSTOWN). A positive interaction term VC*INSTOWN conveys that the monitoring and certification role for the venture capitalist as a substitute for underwriters is enhanced by large institutional allocations: CFU/VC = -1.1138 + 0.0163 (mean INSTOWN) = -1.1138 + 0.0163 (62.2) which yields a reduced but still negative net effect (-0.1) of VCs as a substitute for underwriter verification and advice. We interpret the positive and significant interaction term VC*INSTOWN to mean that for institutional owners, VC involvement does not substitute as readily for underwriter certification and advice. Specifically, institutional liquidity traders appear to prefer to acquire underwriter-broker advisory services in the aftermarket especially when faced with the dire prospect of being the counterparty to trades with a presumably well-informed venture capitalist. In large IPOs, a negative and significant partial effect for the percentage institutional ownership itself (INSTOWN) indicates that institutional traders have their own relationships with underwrites and other brokers who provide real-time assessments of the state of the market in any position the institutions undertake. Consequently, in large IPOs, the institutional trader sees no need to transact in the underwriter channel in order to acquire the needed advisory services. If the IPO is small, however, note that INSTOWN is now positive and significant. This finding suggests that only

25

the underwriter can satisfy the detailed requests for assessment in small IPOs that the institutional investor community demands. The negative sign on RETAIN in all specifications of Table 6 for large IPOs confirms the signalling mechanism that founder-owners employ to reduce the underpricing cost of doing an IPO. By retaining more equity, founder-owners can diminish the rational discounting of the issue price and yet sustain a separating equilibrium in which low quality firms do not mimic the decision to issue. Since more retention cuts into issue proceeds and therefore into underwriting fees, the underwriter advisory services in a competitive market would diminish. However, this decline in underwriterexecuted trades that permanently move the aftermarket price happens not because the cost of underwriting services must be economized when competitive fee income declines, but rather because the value of even a monopoly underwriters advisory services to liquidity traders is diminished when high proportions of retained equity provide a credible commitment by the principals as to the IPOs high issue quality. Reputation effects deserve one note here in the split-sample results. As in the estimations for the Top 50 actively-traded IPOs, here in Table 6 for large IPOs greater reputation effects expressed through the pre-issue endorsements of higher reputation underwriters result in a reduction of underwriter advisory services that trigger informed trades in the aftermarket. OTHERIPOs is negative and significant in the first five columns of Table 6. However, since a greater reputation of the underwriter also enhances the ability of the issuers to secure supply chain and advance sales contracts to list in the prospectus, the net effect of greater reputation depends on the magnitude of the positive interaction term CONTRACTS*OTHERIPOS. Across the five specifications for large IPOs in Table 6, the number of contracts needed to reverse the substitute relationship between reputation effects and underwriter advisory services in the aftermarket is 6 to 8. Since in the largest 25 IPOs sub-sample, the mean number of contracts is 6.1 and ranges from 0 to 14, it is quite possible for high reputation underwriters to have an increased role in the aftermarket because of the indirect effect of their reputations on the securing of additional contracts that need to be verified. Interestingly, in the smallest IPOs this is not true. The only significant effect of higher reputation in the last column of Table 6 is a negative interaction with CONTRACTS. Higher

26

reputation underwriters appear to legitimize fewer contracts that need to be verified, which implies less underwriter advisory services triggering informed trades in the aftermarket. On average, there are only half as many (3.2) contracts in these smallest 17 actively-traded IPOs. Issue Size Finally, issue size is insignificant in the large IPO estimations of Table 6

indicating that this scale variable has been effectively controlled by splitting the sample.21 Issue size interacts with two of the independent variables. In the presence of either large numbers of PAGES of risk factors or a FIRST-mover status in a new technology or product space, an attempt to do substantial size in the offering drives liquidity traders in the aftermarket toward the advisory services of the underwriter-broker. underwriter-broker channel. In stark contrast, in small IPOs reported in the last two columns, the presence of many PAGES of enumerated risk factors in the prospectus drives liquidity traders and hence informed traders looking for counterparties away from the underwriter-broker as the size of the offering rises. Future research clarifying the reasons for these differences in small IPOs would be most useful and perhaps develop valuable additional insights. Finally, we have clarified the complex role of the underwriter in IPOs that issue first in a new technology or product space. In our best cross-sectional model for large IPOs (column 3 of Table 6), the single largest magnitude coefficient is a 0.3255 on FIRST. This dichotomous variable also interacts with ISSUESIZE, DELAY and CONTRACTS in determining the aftermarket role of the underwriter. The attempt to do larger issue size offerings increases the role of the underwriter in first IPOs. Longer delays enhance the perception of issue uncertainty, especially in first IPOs, and underwriters are called upon to explain away issue uncertainty. Like more pages of risk factors, longer delays between announcement and issue also increase the role of the underwriter in the aftermarket. Finally, underwriter-brokers prove less effective in attracting liquidity traders, the more contracts they claim to be verifying for first-movers in a new technology or product space. The net composite effect of all four terms involving FIRST, evaluated at the mean ISSUESIZE ($80.2 million), the mean DELAY (45.2 days) and the mean of CONTRACTS (6.1) is (-0.3255 + 0.1283 + 0.3028 0.1983 = -0.09). Hence, in large IPOs, price discovery in the underThereafter, informed traders follow these counterparties into the

27

writer execution channel is, on net, reduced when the IPO is first in a new technology or product space because, we conjecture, there is little information available to be verified. Instead, in first IPOs investors appear to seek out advice on execution strategy, assessments of the state of the market, and fairness opinions from a more disparate set of brokers, scientists, and intellectual property attorneys.

6. Summary and conclusions Our first objective was to ascertain whether new insights about the role of the underwriterbroker in the IPO aftermarket could be obtained from the cross-sectional determinants of a standard microstructure metric of the order flow data. The answer is now clear. Price discovery in the underwriter execution channel is systematically related in anticipated ways to reputation effects, to the availability of verifiable facts (contracts listed in the prospectus and to first-mover timing), to issue uncertainty (the pages of risk factors and the time delay between announcement and issue), and to three measures of ownership structure (percentage institutional ownership, percentage retention by founder-owners, and the involvement of venture capitalists). Many of these explanatory variables exhibit important interaction effects, and some of the results hold only in large IPOs with issue sizes in non-privitizations from $30 to $365 million. Interesting differences arise with $3 to $12mm IPOs. Our sample comprises issue databases, aftermarket audit trails, and six months of trade-totrade data for the Top 50 IPOs in Australia during the years 1996-1999. One unique feature of this data set is the ability to identify execution channels and distinguish underwriter executions versus those of other brokers. Another is the absence of a quiet period; information flow from Australian underwriter-brokers is continuous. Finally, a wide range of issue sizes among these fifty most actively-traded issues allows extensive study of both small and large IPOs. Overall, the contribution to price discovery by underwriter-brokers (as measured by the Gonzalo-Granger common factor weight) exceeds the underwriter-brokers participation in trading for approximately 80% of the IPOs examined. However, given the underwriters on-going access to company officials and the information advantage about the placement of the initial shares distributed, it is insightful that other brokers also provide statistically significant price discovery in as many as 34

28

of the Top 50 IPOs. That is, the processing of new information is substantially more diffuse than simple asymmetric information models of IPO issuance would suggest. Summarizing the cross-sectional determinants of these common factor weights, we find a diminished role for the underwriter in IPOs previously financed (and therefore externally monitored) by venture capitalists and in IPOs for which owners and managers have signaled higher issue quality by retaining more equity. However, these substitute credibility mechanisms of external monitoring and signaling quickly prove more costly, we conjecture, than establishing issue quality through underwriter certification. We find the certification role of the underwriter proves especially important in five circumstances: 1) when large numbers of complex risk factors are enumerated in the prospectus, 2) when abnormally long delays arise between the announcement and the issue, 3) when heavy institutional ownership seeks to avoid being picked off as a counterparty to informed trades of venture capitalist financiers, 4) when first IPOs in a new technology or product space attempt to offer large issue size or experience substantial delays, and 5) when numerous claims about the effect of complex contracts on the issuing firms business model are in need of verification. Of primary significance is the newly-discovered relationship between the role of the underwriter and the number of supply chain and forward sales contracts listed in the prospectus. We hypothesized that a primary function of the underwriter-broker is to verify such contracts for liquidity traders seeking advice about the equilibrium issue premium they should expect to realize or pay when executing trades in the aftermarket. Liquidity traders seek to mitigate the picking off risk that arises especially in IPOs from serving as counterparties to informed investors whose trades establish all new permanent price trends. Hence, we expect and observe more price discovery in the underwriter execution channel when IPOs involve more supply chain and forward sales contracts. And this result holds in small as well as large IPOs. Moreover, as expected, the contracts-underwriter relationship is enhanced by higher reputations and reduced by more risk factors and longer delays. In conclusion, we wish to emphasize that verifiable facts are not a substitute for, but a complement to, the underwriters certification and advisory services. Controlling for any unusual announcement delays or risk factors in the offering, when verifiable facts are few, we find that the contribution of the underwriter to price discovery in the aftermarket is significantly reduced. These

29

results coincide with the view that the verification process (not de novo information production) is the key function of the underwriter.

30

REFERENCES

Aggarwal, Reena, 2000, Stabilization activities by underwriters after IPOs, Journal of Finance 55, 1075-1103. Aggarwal, Reena, 2003, Allocation of initial public offerings and flipping activity, Journal of Financial Economics 68, 111-135. Aggarwal, Reena and Pat Conroy, 2000, Price discovery in initial public offerings and the role of the lead underwriter, Journal of Finance 55, 2903-2922. Aggarwal, Reena, Nagpurnanand Prabhala, and Manju Puri, 2002, Institutional allocation in initial public offerings: Empirical evidence, Journal of Finance 57, 1421-1442. Aitken, Michael, Garvey, Gerald, and Peter Swan, 1995, How brokers facilitate trade for long-term clients in competitive securities markets, Journal of Business 68, 1-15. Baillie, Richard., G.Geoffrey Booth, Yiuman Tse, and Tatyana Zabotina, 2002, Price discovery and common factor models, Journal of Financial Markets 5, 309-321. Chen, H.and Jay Ritter, 2000, The seven percent solution, Journal of Finance 55, 1105-1131. DeJong, Frank, 2002, Measures of the contribution to price discovery: A comparison, Journal of Financial Markets 5, 323-327. Ellis, Katrina, Roni Michaely, and Maureen OHara, 2000, When the underwriter is the market maker: An examination of trading in the IPO aftermarket, Journal of Finance 55, 1039-1074. Enders, Walter, 1995, Applied Time-Series Analysis, Wiley, New York, NY. Engle, Robert, and Clive Granger, 1987, Co-integration and error correction: representation, estimation, and testing, Econometrica 55, 251-276. Gonzalo, Jesus, and Clive Granger, 1995, Estimation of common long-memory components in cointegrated systems, Journal of Business and Economic Statistics 13, 27-35. Harris, Frederick, Thomas McInish, Gary Shoesmith, and Robert Wood, 1995, Cointegraton, error correction and price discovery on informationally-linked security markets, Journal of Financial and Quantitative Analysis 30, 563-579. Harris, Frederick, Thomas McInish, and Robert Wood, 2002a, Security price adjustment across exchanges: An investigation of common factor components for Dow stocks, Journal of Financial Markets 5, 277-308. Harris, Frederick, Thomas McInish , and Robert Wood, 2002b, Common factor components versus information shares: alternative approaches to price discovery research, Journal of Financial Markets 5, 341-348. Hasbrouck, Joel, 2002, Stalking the "efficient price" in market microstructure specifications: an overview, Journal of Financial Markets 5, 329-339. He, Ping, 2007, A theory of IPO waves, Review of Financial Studies, 20, 983-1020.

31

Hess, Alan, and Peter Frost, 1982, Tests for price effects of new issues of seasoned securities, Journal of Finance 37, 11-25. Huang, Roger, 2002, The quality of ECN and NASDAQ market maker quotes, Journal of Finance 57, 1285-1319. Kleidon, Allan, 1992, Arbitrage, non-trading, and stale prices: October 1987, Journal of Business 65, 483-508. Koh, Francis, and Terry Walter, 1989, A direct test of Rocks model of the pricing of unseasoned equity issues, Journal of Financial Economics 23, 251-272. Lewellen, Katharina, 2004, Risk, reputation, and IPO price support, Journal of Finance, forth. Ljungqvist, Alex and William Wilhelm, 2002, IPO allocations: discriminatory or discretionary?, Journal of Financial Economics 65, 167-201. Ljungqvist, Alex, Tim Jenkinson and William Wilhelm, 2003, Global integration in primary equity markets: The role of U.S. banks and U.S. investors, Review of Financial Studies 16, 63-99.. Loughran,Tim and Jay Ritter, 2004, Why has IPO underpricing changed over time, Financial Management 33, 5-38.. Loughran, Tim., Jay Ritter, and Kristian Rydqvist, 1994, Initial public offerings: International insights, Pacific-Basin Finance Journal 2, 165-199. Lowry, Michelle and G. William Schwert, 2002, IPO market cycles: Bubbles or sequential learning? Journal of Finance 57, 1171-1200. Megginson, William and Kathleen Weiss, 1991, Venture capitalist certification in initial public offerings, Journal of Finance 46, 879-903. Page, Ellis, 1963, Ordered hypotheses for multiple treatments: A significance test for linear ranks, Journal of the American Statistical Association 58, 216-230.
Ritter, Jay R., 2006, Some factoids about the 2006 IPO market, University of Florida, working paper.

Rudd, Judith S., 1993, Underwriter price support and the IPO underpricing puzzle, Journal of Financial Economics 34, 135-151. Stoughton, Neal, Kit Pong Wong, and Josef Zechner, 2001, IPOs and product quality, Journal of Business 74, 375-408.

32

Table 1. Frequency Distribution of Underwritten IPOs By Issue Size. These data reflect the issue size (equity raised) of all non-privitization IPOs underwritten in Australia from 1996 to 1999 in AUD. Top 50 refers to the fifty most-actively traded IPOs. Two split-sample analyses were performed on IPOs with large issue size ($12 to $385mm or $29 to $385mm) versus small issue size ($3 to $12mm or $3 to $29mm). Underwritten IPOs Cumulative Percent 44 0.25 46 18 10 12 16 15 13 2 176 0.51 0.61 0.67 0.74 0.83 0.91 0.99 1.00 Top 50 IPOs Large

Issue Size $0-5MM $6-10MM $11-15MM $16-20MM $21-30MM $31-40MM $41-80MM $81-160MM $161-$385MM Total

Count

Full Sample 5 11 4 2 5 5 10 6 2 50

Small 5 11 5 11 4 2

3 2 5 5 10 6 2 33 2 5 10 6 2 25

17

25

33

Table 2. Descriptive statistics. Panel A displays the percentage of the total shares outstanding at issue held by shareholders with holdings of a given size for all 214 Australian IPOs June 1996-December 1999. Panel B presents the issue size and underwriting fees for all 214 IPOs, for the 176 underwritten IPOs, and for the Top 50 most frequently-traded underwritten IPOs. Panel C presents the returns from the offering price (Poffer) to the first trade price (Popen) and to the closing price (Pclose) on the first trading day for the 176 underwritten IPOs. Results obtained by Rudd (1993) Are presented for comparison with our sample. Panel A: Distribution of allocations by lot size Number of Shares Mean (%) Size of Allocation 5,00110,0011,001-5,000 10,000 100,000 10.8% 8.5% 12.6% Panel B: Issue size and fees Sample N Issue size (millions AUD) Underwriting Fee (%) All 214 IPOs 176 Underwritten Top 50 All 214 176 Top 50 All 214 140 of 176 Top 50 All 214 55 of 176 Top 50 Mean 116.4 26.9 45.6 3.92 3.84 3.60 1.18 1.14 0.98 1.16 1.03 1.0 Median 9.9 10.0 28.6 4.0 4.0 3.88 1.0 1.0 1 1.0 1.0 1.0 Standard Deviation 1008.7 43.6 62.8 1.24 1.23 1.25 0.87 0.49 0.69 0.46 0.36 0.30 100,001 and over 65.9% Percentage held by top 20 shareholders 68.0%

1-1,000 2.2%

Management Fee (%)

Handling Fee (%)

Panel C: First-day returns Log (Pclose/Poffer) Ruud 0.064 Mean 0.020 Median -0.288 Minimum 0.626 Maximum 0.121 Std. dev. 1.541* Skewness 6.020* Kurtosis Number of IPOs with 1st day negative returns Number of IPOs with 1st day positive returns *Significantly different from 0 at the 0.01 level **Significantly different from 3 at the 0.01 level. Log(Pclose)/Popen) Current Study -0.0107 0.2049 0.0095 0.1236 -0.2513 -0.2513 0.3365 0.3365 0.0921 0.3445 0.8813* 3.739* 5.761* 8.2409* 48 34 128 132

34

Table 3. Trading and Issue Characteristics for Underwritten IPOs. Panel A presents the market share for the underwriter and the syndicate of the trading dollar volume and of the number of trades. On the first trading day, the underwriter accounts for 58.5% of the trading volume and 53% of the trades. The syndicate is only slightly larger, accounting for 60.1% of the volume and 55.2% of the trades. In the Top 50 underwritten IPOs, the underwriter alone accounts for 75.5% of the trading and 47.9% of the trades on Day 1, declining slowly to 51.2% of the volume and 39.8% of the trades on Day 45. Panel B presents the number of contracts, the number of pages of risk factors listed in the prospectus, the percentage of shares retained by pre-IPO owners, and the percentage of shares held by institutions for select stocks and the Top 50 sub-samples.

Panel A: Underwriter and syndicate market share

Underwriter Alone % Volume % Trades 176 Underwritten IPOs Day 1 Day 2 Day 3 Day 7 Day 45 Day 90 Top 50 Underwritten IPOs Day 1 Day 2 Day 3 Day 7 Day 45 Day 90 58.5 52.7 54.0 53.3 52.1 41.3 75.5 75.9 65.5 63.7 51.2 39.0 53.0 47.9 49.3 50.3 49.1 39.3 47.9 46.0 44.5 43.8 39.8 27.9 Syndicate % Volume 60.1 55.0 55.6 55.2 51.4 48.1 78.1 78.1 67.7 65.4 54.0 41.4 % Trades 55.2 50.7 51.2 52.6 49.2 45.0 50.8 48.6 47.2 46.0 42.5 30.4

Panel B: Issue characteristics of selected actively-traded underwritten IPOs

Symbol No. of Contracts Pages of Risk Factors Percent Equity Retained Percent Institutional Ownership Symbol No. of Contracts Pages of Risk Factors 90 100 80 108 90 80 132 113 155 100 52 68 101 145 136 Percent Equity Retained Percent Institutional Ownership

91.6 83.4 100 6 AAP 77.8 60.7 106 1 ALL 39.1 5.0 48 1 BDA 36.7 75.6 88 2 BMC 26.2 37.3 104 4 BNO 42.1 30.0 104 4 BRS 71.5 31.5 64 7 CAB 44.9 3.9 72 5 CDO 63.2 34.6 86 3 CLT 73.9 62.1 124 14 EIS 73.1 60.0 71 1 GTP 87.4 47.2 115 7 HOY 52.6 70.4 108 7 HRD 70.4 51.0 80 12 IIN 62.0 53.5 112 7 ISC 33 Large and 17 Small IPOs by Issue Size (Top 50 Sample) No. of Contracts Pages of Risk Factors Percent Equity Retained Percent Institutional Ownership Delay (Announcement to Issue) Venture Capital Financing Other IPOs First IPO CF Lead Underwriter

12 LIB 5 MLB 14 MYO 7 OTT 1 PRI 9 REA 14 ROC 4 SPK 2 SSX 10 TAP 7 TMN 13 TNE 6 TVL 7 UEL 8 VXS Means Large Small 6.2 3.2 102.9 73.0 54.9 55.6 62.5 44.2 43.2 40.3 0.30 0.24 2.79 1.47 0.33 0.41 0.67 0.47

47.4 74.0 50.9 15.0 41.5 70.7 92.0 81.9 79.6 61.0 72.5 63.3 38.6 36.0 79.5 73.6 88.7 61.2 59.3 9.6 83.8 73.2 79.7 72.2 43.0 55.1 66.8 58.5 4.2 74.6 Standard deviations Large Small 4.4 2.5 28.0 19.9 23.4 21.8 20.4 24.6 22.1 7.97 0.46 0.44 2.03 2.24 0.48 0.51 0.19 (.16 to .95) (.5 to .73)

35

Table 4. Gonzalo and Granger common factor weights representing the proportion of price discovery attributable to each channel. For each of the execution channels, we present the normalized common factor weights (in decimal format). We test the elements of this common factor vector for significance using the methodology developed by Gonzalo and Granger (1995). In each case the null hypothesis is that the factor weight for the indicated channel is 0. The test statistic is distributed chi-squared with one degree of freedom and rejects the null with the indicated level of significance. Panel A presents the results for the IPOs for which both the underwriter and the other brokers attract informed trades. Panel B presents IPOs for which the underwriter alone attracts informed trades. Panel C presents IPOs for which the underwriter does not attract informed trades. In column six of each panel we present the return from offering price to the closing price on the first day of trading. The coefficient of correlation between the common factor weights in Panels A, B, and C, column 2 and the offer price to first day closing price returns is 0.1387, which is not statistically significant at the 0.05 level. Return offering Underwriter Other Brokers
Issuing Firm Common Factor Weight Chi-Square Test Common Factor Weight Chi-Square Test

price to first day closing price

Panel A: Underwriter and other brokers both attract informed trades

AAP ALL TAP BNO CAB CDO CLT GTP HRR IFA IIN IAS ISC MYO PRI ROC SPK SSX TMN TNE TVL UEL VXS WPO TLO 0.47 0.50 0.54 0.73 0.82 0.69 0.83 0.78 0.58 0.52 0.79 0.44 0.73 0.79 0.82 0.72 0.73 0.65 0.73 0.85 0.76 0.72 0.83 0.48 0.58 5.36** 4.47** 12.15** 16.99** 27.36** 12.49** 24.35** 26.78** 11.71** 3.16* 32.89** 2.59* 24.66** 28.69** 17.30** 24.54** 18.50** 10.61** 16.57** 10.64** 20.30** 11.57** 13.27** 5.67** 5.46** 0.53 0.50 0.46 0.27 0.18 0.31 0.17 0.22 0.42 0.48 0.21 0.56 0.27 0.21 0.18 0.28 0.27 0.35 0.27 0.15 0.24 0.28 0.17 0.52 0.42 7.33** 4.95** 18.58** 24.55** 15.45** 10.78** 11.51** 15.67** 6.73** 2.76* 14.44** 4.27* 25.44** 6.21** 10.51** 17.36** 11.80** 12.73** 8.21** 6.52** 12.44** 7.69** 5.13** 7.45** 2.92* 0.18378 0.13793 0.34000 0.67500 1.24167 0.01000 0.74000 0.83000 0.16000 0.02308 0.11000 0.09500 -0.29000 1.20000 0.52000 0.01000 -0.22069 0.32749 0.63000 1.65000 1.36000 0.06667 0.20000 0.21000 0.75000

36

Table 4--continued Underwriter

Issuing Firm Common Factor Weight Chi-Square Test

Other Brokers
Common Factor Weight Chi-Square Test

Panel B: Underwriter alone attracts informed trades

ADZ BRS CSB EIS HOY HRD LCH MLB OTT PLX 0.95 0.66 0.68 0.88 0.93 0.64 0.58 0.90 0.77 0.55 5.44* 4.53* 3.30* 29.81** 15.1** 4.72* 3.44* 11.80** 5.71* 0.05 0.34 0.32 0.12 0.07 0.36 0.42 0.10 0.23 0.45 0.02 1.09 0.79 2.78 0.10 1.71 1.93 0.60 0.60 1.71 0.35500 0.04118 0.92500 -0.11000 0.27500 0.00000 0.36000 2.61364 3.45000 0.68333

2.59*

Panel C: Underwriter does not attract informed trades

0.46 ALL 0.28 ANX 0.62 BDA 0.42 BGN 0.16 BMC 0.23 BRZ 0.37 COM 0.56 EBT 0.53 EZE 0.22 HOT 0.76 LIB 0.23 OTO 0.05 REA 0.31 SWG 0.44 VGL *Significant at the 0.1 level. **Significant at the 0.5 level. 1.10 1.81 2.53 2.43 0.09 0.46 2.49 1.93 2.23 0.54 1.12 0.67 0.04 2.13 2.27 0.54 0.72 0.38 0.58 0.84 0.77 0.63 0.44 0.47 0.78 0.24 0.77 0.95 0.69 0.56 1.71 13.64** 1.00 4.72* 2.55 6.06** 8.39** 1.29 1.86 10.96** 0.25 6.71** 15.25** 11.81** 3.71* 0.13793 0.25000 0.82000 0.10000 1.00000 0.40000 0.54000 -0.13000 0.82000 0.16667 0.52000 0.24000 1.66000 0.70000 0.17647

37

Table 5. Cross-sectional analysis of Top 50 IPOs. We report the results of common factor weights in the underwriter channel for the 50 most-actively-traded IPOS regressed on: CONTRACTS, the number of supply chain and forward sales contracts listed in the prospectus; FIRST, a dummy variable equal to one if the firm is the first IPO in a new technology or product market space; DELAY, the number of days from the filing of the offering documents to the completion of the issue; RETAIN, the percentage of equity retained by owner-founders; PAGES, the pages of idiosyncratic risk factors enumerated in the prospectus; VC, a dummy variable equal to one if a venture capital firm provided equity financing; INSTOWN, the proportion of the offering purchased by institutions; OTHERIPOS, the number of other IPOs by the underwriter during that year; and ISSUESIZE in AUD raised.
Independent Variables Verifiable Facts CONTRACTS FIRST Issue Uncertainty DELAY DELAY*FIRST DELAYSQR PAGES CONTRACTS* PAGES Ownership Chars. VC INSTOWN VC*INSTOWN RETAIN Reputation Effects OTHERIPOS CONTRACTS* OTHERIPOS ISSUESIZE 0.0045 (1.51) 0.0009 (0.61) -0.0364 (-1.90*) 0.0057 (1.67) 0.0021 (2.69**) 0.474 3.12** 50 0.0021 (2.44**) 0.477 2.82** 50 0.0004 (0.27) -0.0129 (-0.96) -0.0002 (-0.14) Specification 1 0.0458 (1.84*) -0.5723 (-2.19*) 0.0142 (2.85**) 0.0127 (2.0 4*) -0.0002 (-3.22**) 0.0036 (2.31*) -0.0005 (-2.66**) -0.3435 (-1.75*) Coefficients (t-statistics in parentheses) Specification 2 Specification 3 0.0684 (3.27**) -0.5623 (-2.17*) 0.0141 (2.96**) 0.0124 (1.99*) -0.0002 (-3.29**) 0.0029 (1.89*) -0.0006 (-2.88**) -0.0519 (-0.82) 0.0506 (2.31*) -0.5594 (-2.20*) 0.0141 (3.01**) 0.0121 (1.99*) -0.0002 (-3.28**) 0.0020 (1.28) -0.0004 (-1.96*) -0.0619 (-0.98) 0.0018 (1.52) 0.0038 (1.24) 0.0004 (0.26) -0.0129 (-0.97) Specification 4 0.0698 (3.36**) -0.4937 (-1.87*) 0.0123 ( 2.49**) 0.0109 (1.73*) -0.0002 (-2.83**) 0.0034 (2.12*) -0.0006 (-2.92**) -0.2844 (-1.44)

0.0023 0.0024 (2.71**) (2.82**) R2 0.518 0.457 F Test 2.98** 2.90** N 50 50 *Significant at the 0.05 level **Significant at the 0.01 level

38

Table 6. Cross-sectional analysis of split sample. We report split-sample results of the cross-sectional determinants of common factor weights measuring price discovery in the underwriter channel. The first three columns report 50% and 66% sub-samples of the Top 50 most actively-traded IPOs with issue sizes above $30 million AUD (25 IPOs) and above $12 million (33 IPOs), respectively. The last two columns report the 50% and 33% sub-samples with issue sizes $3 to $29 (25 IPOs) and $3 to $12 million AUD (17 IPOs), respectively. Independent Coefficients (t-statistics in parentheses) variables Large Issue Size Small Issue Size

39

Verifiable Facts CONTRACTS FIRST CONTRACTS* FIRST Issue Uncertainty DELAY CONTRACTS* DELAY FIRST*DELAY DELAYSQR PAGES CONTRACTS* PAGES Ownership Chars. VC INSTOWN VC*INSTOWN RETAIN Reputation effects OTHERIPOS CONTRACTS* OTHERIPOS ISSUESIZE FIRST*ISSUESIZE PAGES*ISSUESIZE

0.1379 (6.01**) -0.0840 (-2.25*)

0.1128 (6.72**) -0.3284 (-1.94*) -0.0316 (-2.78*)

0.1140 (8.08**) -0.3255 (-2.05*) -0.0324 (-3.33**) 0.0240 (10.31**) -0.0021 (-6.69**) 0.0067 (1.88*) -0.0001 (-2.98**) 0.0026 (3.17**) -0.0004 (-4.91**) -0.9690 (-6.09**) -0.0015 (-1.97*) 0.0144 (6.06**) -0.0023 (-3.63**) -0.0588 (-6.85**) 0.0090 (6.01**)

0.1009 (4.59**) -0.1329 (-0.72) -0.0190 (-1.08) 0.0222 (6.03**) -0.0018 (-3.71**)

0.1027 (3.98**) -0.1061 (-2.14*)

0.3860 (2.61*) -0.1849 (-1.88*)

1.0974 (4.04*) -0.4847 (-2.64)

0.0247 (9.38**) -0.0028 (-7.04**)

0.0240 (9.69**) -0.0020 (-6.17**) 0.0062 (1.59)

0.0225 (5.96**) -0.0020 (-4.49**)

-0.0543 (-1.68) -0.0101 (-3.21**)

-0.2304 (-2.65) -0.0261 (-4.54*)

-0.0001 (-2.86**) 0.0011 (1.17) -0.0004 (-4.27**) -1.1138 (-5.06**) -0.0017 (-2.04*) 0.0163 (5.10**) -0.0014 (-1.84*) -0.0581 (-4.69**) 0.0102 (5.67**) -0.0015 (-0.91)

-0.0001 (-2.74*) 0.0026 (2.94**) -0.0004 (-4.64**) -0.9599 (-5.37**) -0.0015 (-1.86*) 0.0142 (5.34**) -0.0022 (-2.63**) -0.0598 (-5.39**) 0.0091 (4.87**) 0.0001 (0.15) 0.0015 (2.13*)

-0.0001 (-1.68) 0.0022 (1.74*) -0.0003 (-2.49*) -0.6052 (-3.98**) -0.0010 (-0.79) 0.0088 (3.61**) -0.0019 (-1.75*) -0.0381 (-2.51*) 0.0054 (2.17*)

-0.00008 (-2.94**) 0.0018 (1.41) -0.0004 (-2.74**) -0.6469 (-4.19**) -0.0013 (-1.08) 0.0093 (3.78**) -0.0011 (-1.00) -0.0457 (-2.67**) 0.0054 (2.17*) 0.0002 (0.12)

0.0011 (2.06*) 0.0059 (0.99) 0.0007 (0.79) 0.1363 (.043) 0.0046 (1.87*) -0.0052 (-0.93) 0.0034 (1.18) 0.0257 (0.79) -0.0105 (-1.35) 0.0852 (1.93*)

0.0040 (3.32*) 0.0206 (0.67) 0.0013 (0.50) -1.1663 (-0.92) (0.0053) (1.59) 0.0057 (0.36) 0.0126 (2.99*) 0.0576 (1.33) -0.0416 (-2.98*) 0.3877 (2.15)

0.0016 (4.02**)

0.0015 (2.39*) 0.0001 (0.061) 0.770 8.16** 33 0.863 7.14** 33 -0.0010 (-1.83*) 0.798 2.37* 25 -0.0037 (-1.51) 0.976 2.73* 17

0.00002 (1.85*)

R2 0.960 0.974 0.974 F Test 14.40** 15.54** 18.85** N 25 25 25 *Significant at the 0.05 level **Significant at the 0.01 level

40

Figure 1. Composition of Australian IPO Syndicates (1996-1999) Panel A: Underwritten IPOs (N = 176)
176 Underwriter Lead-manager 2

Underwriter Sponsoring broker 24

132 No. of IPOs 88

Underwriter only 150

44

Panel B: Non-underwritten IPOs

41

Panel B: Non-underwritten IPOs

21

18

15 Others 7

No. of IPOs

12

Sponsoring Broker 21

Lead-Manager Co-lead manager Co-manager 4

6 Lead-manager only 6

Non-Unde rw ritte n (Sponsoring Broker) IPOs N=21

Non-Underw ritte n (Lea d-m a na ge r) IPOs N=17

42

43

Figure 2. Underwriter-brokers market share of trading volume and number of trades (N=50)
100% 90% 80% Market Share (%) 70% 60% 50% 40% 30% 20% 10% 0% 1 7 13 19 25 31 Day Trading Volume Number of Trades 37 43 49 55 61

44

Figure 3. Industry classification of the 50 most actively traded IPOs 1996-1999

Paper & Packaging Retail Building Materials Energy Infrastructure & Utilities Gold Tourism & Leisure Miscellaneous Industrials Investment & Financial Services Healthcare & Biotechnology Diversified Industrials
Telecommunications

Media

Industry GOLD BUILDING & MATERIALS DIVERSIFIED INDUSTRIALS ENERGY HEALTHCARE & BIOTECHNOLOGY INFRASTRUCTURE & UTILITIES INVESTMENT& FINANCIAL SERVICES MEDIA MISCELLANEOUS INDUSTRIALS PAPER & PACKAGING RETAIL TELECOMMUNICATIONMUNICATION S TOURISM & LEISURE

Sample IPOs 1 1 1 2 5 2 3 5 8 1 5 12 4

45

Figure 4: The pre-issue role of VCs and the underwriter in IPOs (Period 1)
Private Equity

LQSU
Mimic Behavior

The IPO
Eq+ 1. Retain

N Invests

Equity
Underwriter & Issuer

Eqo Eq-

Start-Up and Mezzanine Financing

VC
Refuses N

HQSU

2. Allocate Shares

3. Set Offer Price

+ P o Po

Po

Mimic Behavior

LQSU
VC = Venture Capitalist HQSU = High Quality Start-Up LQSU = Low Quality Start-Up Eq = Equity Retained P = Offer Price N = Random Process

Private Equity

46

Figure 5: The post-issue role of the underwriter in IPOs (Period 2) An Iterated Dominant Strategy Equilibrium {Underwriter Channel LT, Underwriter Channel IT}

Information Asymmetry In Post-Offering Market

Choice of Execution Channel

Informed Trader (IT)

Non-Underwriter Channel Underwriter Channel

Trades Thru Non-U Buys HQSU at `

Po or P o
LT Sells HQSU

Sells HQSU
Trades Thru Underwriter
Non-U Channel

Buys HQSU at
P , Sells o

IT
Sells LQSU at
Po or P + o

at P , o Buys LQSU at P+ o

No Trade

LQSU at

LT
LT Buys

Trades Thru Non-U

P+ o

No Trade

LQSU

Liquidity Trader (LT)

LU Trades Thru Underwriter
Underwriter Channel

No Trade

Buys HQSUs at P o Sells HQSUs for P o to

No Trade

earn (P o - P o )

IT = Informed Trader LT = Liquidity Trader

47

APPENDIX FOR THE REFEREE

Figure 1. The ASX Flotation Process

Decision to Float

Discuss Lisiting Eligibility with ASX

Select Team of Advisors

Develop Timetable Select Syndicate Members

Draft Prospectus Due-diligence

Marketing to Securities Firms & Institutions Due-diligence Sign-off

Prospectus Lodged at ASIC Apply to List on ASX Exposure Period ASIC Examines Prospectus

Prospectus registered

Conditional Listing Approval

Offer Period

Marketing of Offer Acceptance of Applications

Allocation & Distribution of Shares

Delivery of CHESS Statements

ASX Check that Listing Conditions Met

Listing Ceremony & Quotation

48

1 2

See Aitken, Garvey and Swan (1995).

Underwriters are consulted first because other syndicate members are seldom told the location of the stock allocations. Australias adoption of the fixed price method is a result of the historical influence of the British Commonwealth. During the early 1990s, less than 10% of the IPOs in Australia employed the full-blown book building method so prevalent in the U.S. However, many Australian floats have now replaced fixed prices with an indicative price range much like U.S. book building. In cross-national comparative studies, Loughran, Ritter and Rydqvist (1994) and Ljungqvist, Jenkinson and Wilhelm (2003) document that fixed price methods lead to a greater probability of the issue failing than the book building method. The greater failure rates are attributed to increased uncertainty during the time delay between fixed price offer and issuance. With full book-building, the U.S. time delay is 36 hours or less. In Australia, this time delay often lasts two to three weeks. The median daily exchange rate in 1999 when most of these IPOs occurred was 0.65 AUD/USD.

We are indebted to the referee for suggesting we expand the sample from Top 30 to Top 50 most-actively-traded IPOs in order to examine the split sample.
6

In addition, there is a very occasional 4% brokerage fee for selling concessions, etc. This occurs in 4 of our Top 50 most-actively traded IPOs. 1999 first-day returns were much higher than in the previous three years in both Australia and the U.S.. Loughran and Ritter (2004) report median first-day returns 1996-1998 of 10.3%, 9.4%, and 9.0% followed by a 1999 first-day return of 37.5%. Similarly, median first-day returns 1996-1998 for Australia were 9.2%, 12.2%, 6.7%, followed by a 1999 first-day return of 33.7%.

As noted earlier, the Top 50 most actively-traded underwritten issues have mean issue size of 45.6 million AUD versus 19.6 million for the remaining 126 underwritten issues.
9

Also note that the percentage of the trades is much smaller than the percentage of the volume indicating that underwriter-brokers get the medium to larger executions, where one would expect to find informed stealth trading. Aggarwal (2000) and Lewellen (2004) examine issues related to price stabilization.

10 11

In measuring risk factors as total page lengths of idiosyncratic risk factors plus boilerplate, we adopt the approach of Koh and Walter (1989). Issuers attempt to mitigate the effect of the enumerated risk factors on issue proceeds by employing language that intertwines the two; hence the measurement choice of total pages.
12

Interestingly, the IPO price series of business management software developer MYOB Limited (MYO) was not cointegrated across execution channels. Instead, the prices of the underwriter-broker and other brokers diverged for substantial periods, allowing persistent arbitrage opportunities. The MYOB Limited issue exhibited 1) a first day pop, 2) a positive deterministic time trend thereafter, and 3) a positive drift in the price expectations. To our knowledge, financial market microstructure research has never reported the non-cointegration of prices across alternative execution channels in any secondary market. MYOB illustrates a folklore that IPO trading at times generates abnormally persistent arbitrage returns.

49

Earlier versions of the paper incorporated several tables of these time-series results, which are available upon request from the authors. Market price may also not be fully revealing empirically because, off the equilibrium path, some low quality firms mimic the high quality issuers if expected high quality issuers set either their retained equity or the degree of underpricing too low to assure a separating equilibrium.
15

13

14

These R2 and F statistics are all calculated with an intercept in the model. The qualitative results with and without an intercept are identical.

16

Here we are ignoring the unchanging effect on the partial derivative of the quadratic term DELAYSQR listed in row 6-- namely, (2) 0.0002 DELAY.

17

Ultimately, we use issue size to split the sample between large and small IPOs (less than $12 million AUD).

Other measures of underwriter reputation, like IPOs in the previous one, three or five years, have similar effects.
19

18

At any number above 6 contracts, it is quite possible for high reputation underwriters to have an increased role in the aftermarket because of the indirect effect of their reputations on the securing of additional contracts to list in the prospectus. Across the sample, the number of contracts ranges from 0 to 14.

178 pages of risk factors is also the maximum value for our entire Top 50 most-heavily traded IPO sample.
21

20

Issue size remains a significant scale factor in our small IPO sub-samples.

ALL BDA BMC BNO BRS CAB CDO CLT EIS GTP HOY HRD IIN ISC LIB MLB MYO OTT PRI REA ROC SPK

0.138 0.82 1.00 0.675 0.041 1.242 0.01 0.74 -0.11 0.83 0.275 0 0.11 -0.29 0.52 2.614 1.2 3.45 0.52 1.66 0.01 -0.221

50

SSX TAP TMN TNE TVL UEL VXS

0.327 0.34 0.63 1.65 1.36 0.067 0.2

51