Heterogeneous Beliefs, IPO Valuation, and the Economic Role of the Underwriter in IPOs

Transcription

1 Heterogeneous Beliefs, IPO Valuation, and the Economic Role of the Underwriter in IPOs Thomas J. Chemmanur and Karthik Krishnan We empirically analyze the economic role of the underwriter in initial public offerings (IPOs), distinguishing between the certification and market power hypotheses. We find that equity in high-reputation underwriter backed IPOs is priced higher and further away from intrinsic value than that in low-reputation underwriter backed IPOs. Our results are robust to controlling for the endogenous selection of firms to take public by underwriters. Overall, our results support the market power hypothesis and reject the certification hypothesis, indicating that the role of underwriters is to obtain the highest possible valuation for the IPOs that they back rather than to price the equity close to intrinsic value. Starting with Miller (1977), the effect of heterogeneity in investor beliefs on the valuation of equity has been the subject of considerable debate among financial economists. In particular, formalizing Miller s (1977) intuition, Harrison and Kreps (1978) and Morris (1996) have demonstrated that the equity of firms will reflect the valuation of the most optimistic investors (and will sell at a premium over their fundamental value) when investors subject to short sale constraints have heterogeneous prior beliefs (Chen, Hong, and Stein, 2002; Duffie, Garleanu, and Pederson, 2002). Given that firms going public have a very sparse track record publicly available to investors, the equity of firms conducting an initial public offering (IPO) are likely to be characterized by a higher degree of heterogeneity in investor beliefs and, therefore, provide a particularly appropriate context to test the implications of heterogeneous beliefs and behavioral theories of equity valuation. Existing empirical analyses of such theories have been primarily in non-ipo settings (Chen et al., 2002; Diether, Malloy, and Scherbina, 2002). The objective of this paper is to extend the above literature to the IPO setting by analyzing the role of heterogeneity in investor beliefs under short sale constraints on the valuations of the equity of firms going public. In particular, we empirically analyze the implications of these theories on the role of the underwriter in the valuation of the equity of firms going public, both in the IPO and in the immediate secondary market. This paper was previously circulated under the title, Heterogeneous Beliefs, Short Sale Constraints, and the Role of the Underwriter in IPOs. For helpful comments and discussions, we thank Douglas Cook, Iraj Fooladi, Elena Loutskina, Gang Hu, and seminar participants at Boston College, Babson College, Dalhousie University, the 2008 FIRS Conference, the 2008 European Finance Association Annual Meetings, the 2007 Financial Management Association Annual Meetings, and the 2007 Eastern Finance Association Annual Meetings. Special thanks to the Bill Christie (Editor) and an anonymous referee for helpful comments and suggestions that greatly improved the paper. Thomas Chemmanur acknowledges summer support from the Carroll School of Management at Boston College. All errors and omissions remain are own. Thomas J. Chemmanur is a Professor of Finance in the Carroll School of Management, Boston College in Chestnut Hill, MA. Karthik Krishnan is an Assistant Professor of Finance in the College of Business at Northeastern University in Boston, MA. Financial Management Winter 2012 pages

2 770 Financial Management Winter 2012 Underwriters have a particularly important role to play in IPO valuation. This is because, first, they set the IPO offer price after ascertaining the views of investors on their valuation of the IPO through the book-building process. Second, participation by a highly reputable underwriting syndicate may affect the beliefs of at least some investors about the future prospects of the IPO firm. Therefore, we focus on the economic role of the underwriter in affecting the heterogeneity in investor beliefs faced by IPO firms, which, in turn, will affect the valuation of equity in the IPOs that they underwrite. The economic role of the underwriter in IPOs has become a particularly relevant topic of study in light of the recent empirical evidence that the relationship between underwriter reputation and IPO underpricing has reversed. While earlier evidence documented that IPOs underwritten by high-reputation investment banks were characterized by a smaller extent of underpricing, more recent evidence indicates that IPOs underwritten by high-reputation underwriters are characterized by greater underpricing (Beatty and Welch, 1996; Cooney, Carter, Dark, and Singh, 2001; Loughran and Ritter, 2004) who find that the average underpricing has gone from being smaller for high-reputation underwriters in the 1980s to being larger for highreputation underwriters in the 1990s. This is consistent with a parallel debate in the venture capital literature. For instance, Francis and Hasan (2001) use data from the 1990s and find that IPOs of firms backed by venture capitalists are underpriced more than those not backed by venture capitalists, in contrast with the findings of the prior literature that uses data from the 1980s. In order to incorporate the role of the underwriter in the Miller (1977) setting, we conjecture that underwriters are motivated to obtain high valuations for the equity of firms they take public by making use of their long-term relationships with various participants in the IPO market (institutional investors, analysts, and co-managing underwriters). These relationships enable them to attract greater participation by these market players in the IPOs of firms backed by them, leading to higher valuations for the equity of these firms (both in the IPO and in the secondary market). 1 Participation by high-reputation underwriters and by a larger number of other highquality market participants (i.e., high-reputation co-managers and a larger number of institutional investors and analysts) may make retail investors more optimistic about the IPO firm s prospects. For example, Welch (1992) finds that later investors in a firm s IPO will optimally ignore their private information and buy shares in the IPO after becoming more optimistic about the IPO firm s prospects if they observe strong demand from earlier investors for these shares. 2 Such higher optimism of retail investors increases the heterogeneity in investor beliefs, which, in turn, may lead to higher valuations, both in the IPO and in the immediate aftermarket based on arguments made by Miller (1977) and others discussed earlier. In summary, the greater heterogeneity in investor beliefs regarding firms whose IPOs are underwritten by high-reputation underwriters (relative to firms whose IPOs are underwritten by low-reputation underwriters) will lead to higher 1 Underwriters may be motivated to obtain high valuations for the IPO firm for two reasons. First, their gross spread revenue per share is proportional to the IPO offer price. Moreover, successfully obtaining high IPO proceeds will increase the underwriter s reputation with subsequent issuers leading to higher future revenues. 2 Consistent with this, in their survey of the diffusion of interest among institutional and retail investors, Shiller and Pound (1989) found that 44% of investors who bought shares in a firm s IPO did so because someone they knew of bought stock in the company. Further, various other papers have found evidence consistent with the idea that unlike large and institutional investors, retail investors trades are motivated by analyst recommendations and other sources of information that may grab their attention, even if such sources do not benefit them in the long run. For example, Malmendier and Shanthikumar (2007) find that small investors follow analyst recommendations literally, even if analysts are affiliated with an underwriter; whereas, larger traders discount analyst recommendations. Barber and Odean (2008) find that individual investors are more likely to be net buyers in attention getting stocks, unlike professional investors.

3 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 771 valuations of these firms in the IPO, as well as in the immediate secondary market. 3 We will refer to the above hypothesis as the market power hypothesis. 4 The above hypothesis contrasts with the prevailing view that the underwriter acts as a certifying intermediary who produces information about the intrinsic value of the firm in an IPO market characterized by asymmetric information between firm insiders and outsiders. Since information production is costly (and therefore characterized by moral hazard), underwriters have a short-term incentive to sell the equity of firms they take public at a price higher than their intrinsic value. However, underwriters have their reputation at stake with investors, mitigating this moral hazard problem in information production, thereby inducing them to price equity in IPOs consistent with intrinsic value (Chemmanur and Fulghieri, 1994). This means that the certification hypothesis has implications dramatically different from the market power hypothesis for the pricing of IPOs. While the certification hypothesis implies that high-reputation underwriters price equity in IPOs backed by them closer to the intrinsic value due to their concern for preserving their reputation in the IPO market, the market power hypothesis implies that high-reputation underwriters would price this equity higher and further away from its intrinsic value. In order to make the market power hypothesis more precise and understand the relationship between the IPO and the first-day secondary market share valuations of IPOs underwritten by high- and low-reputation underwriters, we study the following economic environment. Consider an IPO market populated by both retail and institutional investors where a firm s IPO may be underwritten by either a high-reputation or a low-reputation underwriter. Underwriters care about both the initial offer price and the first-day closing secondary market share price of the IPO firm, maximizing a weighted average of the two. Thus, an underwriter chooses the offer price (OP) and (implicitly) the secondary market price (secondary market price) to maximize 5 γ 1 (.)OP + γ 2 (.)SMP, (1) where the weights γ 1 (.) and γ 2 (.) are decreasing functions of offer price and secondary market price, respectively. 6 Furthermore, for simplicity, we assume that the γ 1 (.) and γ 2 (.) functions are the same for high- and low-reputation underwriters. For the reasons discussed earlier, assume 3 We are assuming here that the increase in the heterogeneity in investor beliefs is symmetric about the mean level of investor beliefs about the firm. In such a setting, it can be shown using a formal theoretical model that the beliefs of the marginal investor about the firm going public and, therefore, the valuation of the firm s equity in the IPO and in the immediate secondary market will be increasing in the heterogeneity in investor beliefs. While, due to space limitations, we will not present a formal theoretical analysis here, the above results on equity valuation under heterogeneous beliefs will follow from a theoretical model along the lines of that in Bayar, Chemmanur, and Liu (2011). 4 This paper primarily focuses on the role of the lead underwriter in IPOs. Through the rest of the paper, we refer to lead underwriters in an IPO as underwriters and explicitly refer to co-managing underwriters (co-managers) wherever necessary. 5 In practice, the variables explicitly chosen by the underwriter will be the IPO offer price and the extent of participation by high-quality market players which, in turn, will determine the extent of heterogeneity in investor beliefs and the secondary market stock valuation. As we explain below, the extent of participation by high-quality market players will be determined by the reputation of the underwriter and the extent of underpricing. 6 The rationale for assuming such an objective function is that underwriters care about the IPO, as well as the secondary market price of the firm s equity. Underwriters care about the firm s IPO price since they receive a fraction of gross IPO proceeds as compensation. Underwriters also care about the IPO firm s short run secondary market price as they are committed to engaging in price support activities if this price falls below a certain threshold level. Additionally, their reputation with future issuers will be damaged if the stock prices of firms whose IPOs they previously underwrote languish in the secondary market. Our results will remain qualitatively unchanged even if we allow the weights γ 1 (.) and γ 2 (.) to depend on additional IPO parameters such as the fraction of equity sold by insiders in the IPO. See Chemmanur (1993) for a similar objective function.

4 772 Financial Management Winter 2012 that for a given extent of underpricing, the heterogeneity in investor beliefs about the IPO firm s prospects and, therefore, secondary market share valuations, will be greater if the IPO is backed by a high-reputation underwriter rather than by a low-reputation underwriter. Further, assume that either a high- or a low-reputation underwriter can generate greater participation by institutional investors (and other high-quality market players) in the IPO by underpricing to a greater extent, since such underpricing will serve to compensate these market players for their participation in the IPO. 7 Finally, we assume that participation by a larger number of such market players will further increase the heterogeneity in investor beliefs, and, as such, the secondary market valuation, of the firm going public. In the above setting, we can show the following. First, for a given level of IPO underpricing, the secondary market share price of a firm whose IPO is backed by a high-reputation underwriter will be greater than if it were backed by a low-reputation underwriter. The intuition here is that high-reputation underwriters face a steeper demand curve for shares than low-reputation underwriters due to the greater heterogeneity in retail investor beliefs associated with highreputation underwriter backed IPOs. Additionally, high-reputation underwriters will underprice IPOs to a greater extent than low-reputation underwriters. The intuition here is that since a low-reputation underwriter backed IPO receives lower secondary market valuations than a highreputation underwriter backed IPO (for the reasons discussed earlier), the offer price of a lowreputation underwriter backed IPO will be lower than that of a high-reputation underwriter backed IPO for the same extent of underpricing. This means that a low-reputation underwriter will underprice IPO shares to a smaller extent than a high-reputation underwriter, given that the weight the low-reputation underwriter places on the IPO offer price in his objective function will be higher for a lower offer price. Moreover, even after underpricing IPOs to a greater extent, high-reputation underwriters will set the OPs of IPOs backed by them at a higher level than lowreputation underwriters (for reasonable model parameter values). The intuition here is that as the high-reputation underwriter underprices to a greater extent (in order to attract greater institutional investor participation and obtain a higher secondary market price), the offer price chosen by him becomes lower as a consequence. A lower offer price increases the weight placed on the offer price in the underwriter s objective function, whereas a higher secondary market price (arising as a result of greater IPO underpricing) will decrease the weight placed on the secondary market price in this objective function. The equilibrium IPO offer price arising from the above trade-off will be higher for a high-reputation underwriter backed IPO than for a low-reputation underwriter backed IPO. 8,9 We test these predictions of the market power hypothesis below. We now summarize the testable predictions arising from the market power and the certification hypotheses. We develop these testable predictions in terms of IPO valuation ratios (i.e., the ratio of primary and secondary market IPO stock valuations relative to their intrinsic value) rather 7 This assumption that institutional investors are important beneficiaries from IPO underpricing is consistent with the evidence that there is a significant amount of flipping of IPO shares by institutional investors to retail investors both in the first few days of trading in secondary market shares (Aggarwal, 2003) and in subsequent months (Chemmanur, Hu, and Huang, 2010). It is also consistent with the evidence that institutional investors make significant abnormal profits from selling their IPO share allocations soon after the IPO (Chemmanur, Hu, and Huang, 2010). 8 Recall that even for the same γ 1 (.) and γ 2 (.) functions, the weight placed on the secondary market price in the objective function of a low-reputation underwriter will be greater than that in the objective function of a high-reputation underwriter since the secondary market share valuation (for a given level of IPO underpricing) will be lower for the low-reputation underwriter inducing him to optimally set a lower offer price. 9 Throughout the above discussion, we assume that the equilibrium offer price emerging from the above trade-off faced by high and low-reputation underwriters is below the maximum possible IPO valuation that is achievable by the firm. This maximum possible IPO valuation will depend on the heterogeneity in investor beliefs in the IPO market, which, in turn, will be a function of underwriter reputation and participation by other high-quality market players like institutional investors and analysts.

5 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 773 than IPO underpricing. This is because underpricing simply reflects the rise in price of a firm s equity from the IPO offer price to the first-day closing price in the secondary market, so that it is affected not only by the price of the equity in the IPO, but also by the first-day closing price of the equity in the secondary market. 10 We use three sets of measures in our empirical analysis. The first measure we study is the ratio of the valuation placed on the firm in the IPO (valuation at the offer price) to its intrinsic value. 11 The implications of the market power hypothesis discussed earlier suggest that the ratio of the IPO firm value to intrinsic value will be larger for high-reputation underwriter backed IPOs than for low-reputation underwriter backed IPOs (H1A). Alternatively, if the role of the underwriter in IPOs is that of certification, one would expect high-reputation underwriter backed IPOs to be priced closer to the intrinsic value than low-reputation underwriter backed IPOs, so that this ratio will be closer to one for high-reputation underwriter backed IPOs (H1B). The second measure we examine is the ratio of the valuation placed on a firm at the close of the first trading day in the secondary market to its intrinsic value. The implications of the market power hypothesis discussed earlier suggest that this ratio will be higher for IPOs of high-reputation underwriter backed firms than for IPOs of low-reputation underwriter backed firms (H2). Of course, one would expect that the valuation difference between high-reputation underwriter backed and low-reputation underwriter backed IPOs to go down over time as hard information about the operating performance of the IPO firm becomes available to investors (the heterogeneity in investor beliefs associated with the reputation of the IPO underwriter dissipates over time). Therefore, we study the ratio of secondary market valuation to intrinsic value at the end of one year, two years, and three years after the IPO. The market power hypothesis predicts that the price to intrinsic value ratio of IPOs backed by high-reputation underwriters and lowreputation underwriters should converge toward each other and toward unity in the years after the IPO (H3A). In contrast, given that underwriter reputation and other IPO characteristics are observable by equity market investors at the IPO stage itself, and that firm value is determined by the equity market alone once the firm s shares start trading in the secondary market, the certification hypothesis predicts that the secondary market price to intrinsic value ratio will not systematically change over time subsequent to the IPO (H3B). The third set of measures relate to the relationship between underwriter reputation and the heterogeneity in investor beliefs about the IPO firm. Under the market power hypothesis, we expect the heterogeneity in investor beliefs to be greater for IPOs backed by high-reputation 10 For underpricing to be a meaningful measure in any study of the economic role of underwriters in IPOs, one has to make the crucial (and rather strong) assumption that the closing price of a firm s stock on the first day of secondary market trading is not affected by underwriter reputation and always equals the intrinsic value of that stock. It is difficult to assume that in a setting with heterogeneous beliefs and short sale constraints, the price of an IPO firm s equity at the close of the first day of secondary market trading will equal its intrinsic value (little information regarding the future performance of the IPO firm is released between the time the IPO offer price is set and the close of trading on the first day). The empirical literature on the long-run underperformance of IPOs (Ritter, 1991) also indicates that the closing price of an IPO firm s equity is not equal to its intrinsic value. This literature suggests that if investors buy IPO shares at the first trading day closing price and hold them for one to three years, they are likely to earn inferior returns when compared to similar investments in the equity of firms that have been public for some time. Apart from the above evidence provided by academic studies, it is easy to see from casual observation that the opening day secondary market price of IPO shares is significantly different from the intrinsic value during some periods. One example is the internet bubble period of , where a number of IPOs were priced far above their intrinsic value, only to climb much higher on the first day of trading in the secondary market. It seems obvious (at least in hindsight) that while these IPOs were highly underpriced (in the sense that their initial returns were very large), they were also significantly overvalued (relative to intrinsic value). 11 We calculate the intrinsic value using multiple methodologies, such as relative valuation using a comparable firm based on simple matching and propensity score matching (Purnanandam and Swaminathan, 2004), as well as using a discounted cash flow method based on Ohlson s (1990) residual income valuation model.

6 774 Financial Management Winter 2012 underwriters (H4). The greater the heterogeneity in investor beliefs, the higher the valuation of the firm s equity in the secondary market (H5). Greater market participation by institutional investors and more reputable co-managing underwriters will increase the heterogeneity in investor beliefs and should also be positively related to the secondary market price to intrinsic value ratio (H6). We would also expect that any heterogeneity in investor beliefs associated with underwriter reputation would dissipate over time subsequent to the IPO (H7) and such decreases in heterogeneous beliefs will be associated with decreases in the IPO firm s stock market value to intrinsic value ratio (H8). These testable predictions are tabulated in the Appendix. The results of our empirical analysis indicate that equity in high-reputation underwriter backed IPOs are priced higher and further away from the intrinsic value when compared to low-reputation underwriter backed IPOs, both at the offer price and at the first-day closing secondary market price. This is consistent with the market power hypothesis, but inconsistent with the certification hypothesis. One potential alternative explanation to the market power hypothesis for this result is that it reflects endogenous matching between high-reputation underwriters and higher quality issuing firms based on unobservable value-relevant firm characteristics not captured by our valuation methodology. We use three different methodologies to control for any such endogenous matching between underwriters and issuing firms. The first methodology we use is to split our sample into hot and cold IPO markets. Prior literature finds that hot IPO market issuers are of the same or lower quality than cold IPO market issuers (Ritter, 1991; Helwege and Liang, 2004). Given this, the endogenous matching argument would suggest that the valuation difference between IPOs backed by high-reputation underwriters and low-reputation underwriters would be the same in hot IPO markets and in cold IPO markets. Alternatively, under the market power hypothesis, one would expect underwriters with high reputation to be able to manipulate retail investor beliefs to a greater extent in hot markets than in cold markets (since investors are likely to be already optimistic in hot markets) (Helwege and Liang, 2004), so that the correlation between underwriter reputation and IPO valuation can be expected to be stronger in hot markets than in cold markets. Our split sample analysis indicates that the association between underwriter reputation and IPO valuation is stronger in hot IPO markets than in cold IPO markets, consistent with the above result due to underwriters market power rather than due to endogenous matching. The second methodology we use to control for the above potential endogenous matching is to exploit the discontinuous nature of New York Stock Exchange (NYSE) listing requirements for assets to control for any endogeneity between underwriter reputation and IPO valuation in a regression discontinuity framework. The third methodology we employ is to use an instrumental variables analysis to correct for the endogeneity between underwriter reputation and IPO valuation (we discuss the instruments we use in Section intrinsic value). We find that even after controlling for any endogenous matching between high-reputation underwriters and high intrinsic value firms using the three methods discussed earlier, high-reputation underwriter backed firms are associated with higher IPO and secondary market valuations. The results of our empirical analysis also indicates that the above relationship between underwriter reputation and post-ipo secondary market valuation is associated with greater heterogeneity in investor beliefs and greater participation by institutional investors, analysts, and high-reputation co-managing underwriters that characterize high-reputation underwriter backed IPOs. 12 Further, our results indicate that the valuation of IPO stocks converge toward their 12 In unreported tests, we also analyze the relationship between the firm valuation ratio at the IPO and the heterogeneity in investor beliefs, as well as market participants and find results similar to those reported for the post-ipo secondary market. However, our analysis of the secondary market valuation is more appropriate since our independent variables in this analysis are measured ex ante relative to the dependent variable (i.e., post-ipo secondary market valuation).

7 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 775 intrinsic value over the three years after the IPO, with the heterogeneity in investor beliefs associated with the IPO firm declining over this time period. We confirm that the above decline in IPO stock valuation over the long run is associated with a decline in the heterogeneity in investor beliefs over the same time period. Overall, our results support the market power hypothesis and reject the certification hypothesis. Further, unlike results based on IPO underpricing, the evidence based on our measures indicates that the economic role of the underwriter remains the same across the 1980s and the 1990s. What do our empirical results tell us about the effect of heterogeneous beliefs among investors, in the presence of short sale constraints on IPO valuation and the role of underwriters in IPOs? First, consistent with the heterogeneous beliefs theory of Miller (1977) and Morris (1996), a greater extent of heterogeneity in investor beliefs is associated with higher IPO stock valuations. Second, high-reputation underwriters are able to build reputation with issuers by obtaining higher equity valuations for shares sold in the IPO and in the immediate secondary market. Third, high-reputation underwriters are able to obtain higher IPO valuations by generating greater participation by higher quality market players such as high-reputation co-managing underwriters and institutional investors, and by obtaining greater analyst coverage for IPOs backed by them. 13 This, in turn, induces retail investors to become more optimistic about the future prospects of these IPO firms, thus increasing the heterogeneity in investor beliefs associated with them and resulting in higher firm valuations. Fourth, the above pricing behavior by high-reputation underwriters may not damage their reputation with institutional investors since the valuation of equity in firms backed by highreputation underwriters becomes even higher (and commands an even higher premium over shares in IPOs backed by low-reputation underwriters) in the first few days of secondary market trading. Since there is considerable evidence indicating that there is a significant amount of flipping of IPO shares by institutional investors to retail investors, both in the first few days of trading in the secondary market shares (Aggarwal, 2003) and in subsequent months (Chemmanur, Hu, and Huang, 2010), institutional investors may be able to make higher profits by buying shares in IPOs backed by high-reputation underwriters (despite their higher IPO valuations) by selling them at even higher valuations to retail investors. Finally, the above indicates that the investors who are least well-off from participating in IPOs backed by high-reputation underwriters are retail investors who, consistent with the well-documented long-term underperformance of IPOs (Ritter, 1991), may be purchasing heavily overvalued shares of equity when they invest in IPOs in the secondary market. Further, such retail investors are worse off investing in shares backed by high-reputation underwriters relative to investing in IPOs backed by low-reputation underwriters since the former are priced higher and further away from intrinsic value. The rest of the paper is organized as follows. Section I discusses how our paper relates to the existing literature. Section II describes the data and the sample selection methodology. Section III outlines the empirical methodology used in the paper. Section intrinsic value describes our empirical results on IPO stock valuation and underpricing. Section V presents our results about the heterogeneity in investor beliefs regarding firms making IPO. Section VI describes our multivariate analysis of the mechanisms that drive the equity market valuation effects of IPO underwriter reputation, while Section VII reports our conclusions. 13 In unreported tests and consistent with our expectations, we also find that IPOs backed by higher reputation underwriters are associated with greater participation by high-quality market players such as institutional investors, analysts, and more reputable co-managing underwriters.

8 776 Financial Management Winter 2012 I. Relation to the Existing Literature The evidence provided by the prior literature (Beatty and Welch, 1996; Cooney et al., 2001; Loughran and Ritter, 2004) indicates that high-reputation underwriters are also associated with higher IPO underpricing (i.e., a larger amount of money left on the table) during the 1990s. We reconcile our results that high-reputation underwriters obtain higher IPO and secondary market valuations with those in the above-mentioned literature by noting the following. First, as documented by Purnanandam and Swaminathan (2004), IPOs can be underpriced and overvalued relative to intrinsic value at the same time provided that the immediate secondary market price can deviate from the intrinsic value (since underpricing merely captures the initial return from the IPO offer price to the first-day closing price in the secondary market). Additionally, the money left on the table by underwriters may be a mechanism through which they share value with market players such as institutional investors and analysts, thus attracting participation by them to the IPOs they underwrite. Consequently, high-reputation underwriters may leave a greater extent of money on the table to attract greater participation by these market players, thereby enabling IPO firms to obtain higher market valuations both in the IPO and the secondary market. Consistent with this argument, Chemmanur, Hu, and Huang (2010) document that institutional investors sell a large fraction of their IPO allocations by the end of the first month after the IPO and capture much of the money left on the table on these shares. In a similar vein, Aggarwal, Krigman, and Womack (2002) and Cliff and Denis (2004) have found that IPOs that leave more money on the table are associated with greater analyst coverage. In addition to the theoretical and empirical literature discussed earlier on the effect of heterogeneous beliefs among investors on equity valuation, this paper is also related to the broader literature on the role of financial intermediaries in IPOs. For instance, Chemmanur, Krishnan, and Loutskina (2008) study the role of venture backing in IPOs and conclude that venture capitalists perform a marketing rather than a certifying role. However, they do not examine the role of underwriter reputation in IPOs. Our paper is also related to Cook, Kieschnick, and Van Ness (2006) who argue that investment bankers try to promote IPOs to induce sentiment investors to buy the stock. 14 While they indicate that pre-offer publicity is related to investment bank compensation and IPO underpricing, they do not study the role of underwriter reputation in IPOs or its impact on the heterogeneity in investor beliefs about the firm s prospects. Further, our paper is related to Aggarwal, Krigman, and Womack (2002) who analyze the role of underpricing and analyst research in determining insider sales around lockup periods. While their paper relates underpricing to analyst coverage and insider sales at the lockup period, our paper analyzes how underwriter reputation affects IPO and first-day closing secondary market valuations. Our paper is also related to the broader theory regarding IPOs (see Allen and Faulhaber, 1989; Grinblatt and Hwang, 1989; Chemmanur, 1993; and Welch, 1989 for theoretical models of IPO underpricing) and the going public decision (see, e.g., Chemmanur and Fulghieri, 1999), as well as the empirical literature on the going public decision (see, e.g., Chemmanur, He, and Nandy, 2010) See also Ho, Huang, Lin, and Lin (2010) who document that prior to their offerings, IPO firms tend to report higher earnings and disclose inflated earnings forecasts in an attempt to manage news coverage around the IPO. Furthermore, Chan (2010) examines individual versus institutional trades in IPO stocks and documents that IPO stocks open-to-close returns are positively related to small trade participation, small trade purchases, and small trade order imbalance, but only in the case of hot IPO samples. 15 Houge, Loughran, Suchanek, and Yan (2001) study the relationship between the long-term performance of IPOs and investor uncertainty and the divergence of investor opinion about the firm s IPO. In contrast, our focus here is on the effect of heterogeneous beliefs on IPO valuation and the impact of the underwriter on the above variables.

9 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 777 II. Data and Sample Selection The initial sample is obtained from the Securities Data Company (SDC) Platinum New Issues database and consists of 7,780 IPOs issued from 1980 to Of this sample, only 5,276 firms are present on the Center for Research in Security Prices (CRSP) and Compustat data sets. As is common in the IPO literature, we remove IPOs of American Depository Receipts (ADRs), nonordinary shares, Real Estate Investment Trusts (REITs), closed-end fund shares, or unit offerings. We do not rely only on SDC classifications to identify ADRs, nonordinary shares, REITs, closedend funds, and unit offerings. Instead, we use share codes from CRSP to implement this filter. We also remove firms with an offer price of the IPO less than $5, as well as firms with missing data for assets leaving us with 3,737 IPOs in the sample period. Panel A of Table I presents the descriptive statistics for our IPO sample. The IPOs in our sample have a median asset size of $21.07 million, median earnings before interest, taxes, depreciation, and amortization (EBITDA) of $3.01 million, and median proceeds from the IPO of about $29 million. Venture capital firms back about 45% of all the IPO firms in the sample. The mean underpricing (21%) and the median underpricing (7%) for our sample are consistent with the numbers reported in Loughran and Ritter (2004) (mean underpricing of 18.7% and median underpricing of 6.3%). We also provide summary statistics for our measures of heterogeneous beliefs and underwriter market share used to measure underwriter reputation. For additional analysis, we need to impose additional filters depending on the intrinsic value calculation methodology. This further decreases the size of the sample for which we can calculate intrinsic value ratios. The next section describes these valuation methodologies in detail. III. Methodology A. Intrinsic Value Calculations To calculate intrinsic value, we follow a methodology similar to the one in Chemmanur, Krishnan, and Loutskina (2008) and Purnanandam and Swaminathan (2004). We compute three measures of intrinsic value. The first measure uses the market valuation of a comparable firm based on similar industry, size, and EBITDA/sales ratio to that of the IPO firm. The second method uses the market value of a comparable firm based on a propensity score-based matching approach. The final method is a discounted cash flow approach using Ohlson s (1990) residual income valuation method. A detailed overview of the three methods follows. 1. The Basic Comparable Firm Approach The first approach that we use to estimate the intrinsic value of IPO companies is a matching technique based on finding an industry peer with comparable Sales and EBITDA profit margins (EBITDA/Sales). This approach is similar to the one followed in Chemmanur et al. (2008), Purnanandam and Swaminathan (2004), and Bhojraj and Lee (2002). We first remove the IPO firms that have negative or missing sales and EBITDA values in Compustat leaving us with 2,724 firms. To obtain candidate firms to match, we use all the Compustat firms that have been public for at least three years prior to the IPO date and exclude ADRs, foreign company stocks, REITS, nonordinary shares, closed-end funds, and firms with a stock price less than $5 at the IPO date. The remaining Compustat firms are divided into 48 industry portfolios based on Fama and French (1997) industry classifications. For each year, we divide each industry portfolio into three portfolios based on sales and then separate each sales portfolio into three portfolios based

10 778 Financial Management Winter 2012 Table I. Summary Statistics and Underpricing Trend in the 1980s and the 1990s This table reports the descriptive statistics for a sample of IPOs. IPOs with an offer price below $5.00 per share, unit offers, REITs, closed-end funds, banks and S&Ls, and ADRs are excluded. Total Proceeds (in $ millions) and venture backed dummy are as reported in SDC. Assets (in $ millions) and EBITDA (in $ millions) are as reported in Compustat for the fiscal year prior to the IPO date. Underpricing is the percentage change in stock price from the offer price to the closing value at the end of the first trading day. Offer price data are from SDC, whereas first-day closing prices are obtained from CRSP. Daily volume is calculated as number of shares traded during the first trading day multiplied by price. Daily turnover calculated as the first-day turnover. Daily number of trades is calculated as the first-day number of trades. The adjusted values of the volume, turnover, and number of trades variables are calculated by subtracting them from the average value of that statistic calculated three years after the IPO date. Panel A reports descriptive statistics for the full IPO sample; Panel B reports the time trend of underpricing for the sample. 80s time period corresponds to issues that are placed from 1980 to 1989, 90s time period corresponds to issues placed from 1990 to 1998, and 90s incl. bubble time period corresponds to issues placed from 1990 to High-reputation underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank higher than the median for the sample. Low-reputation underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank lower than the median. The ranking is calculated as the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). Data are from SDC Platinum, CRSP, and Compustat. The statistical significances are for the t-test for the differences for the equality of means and the Wilcoxon-Mann-Whitney rank sum test for the equality of medians of two subsamples. Panel A. Summary Statistics Assets EBITDA Total Venture Underpricing Adj. Adj. Daily Adj. Daily Underwriter Proceeds Backed Daily Turnover Number of Market Volume trades Share High Mean , reputation Median underwriter Std. Dev. 1, , Count 1,842 1,842 1,831 1,842 1,842 1,385 1, ,842 Low Mean , reputation Median underwriter Std. Dev , Count 1,895 1,895 1,894 1,895 1,895 1,340 1, Mean diff , Median diff Total Mean Median Std. Dev , Count 3,737 3,737 3,725 3,737 3,737 2,725 2,726 1,825 3,737 (Continued)

11 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 779 Table I. Summary Statistics and Underpricing Trend in the 1980s and the 1990s (Continued) Panel B. Underpricing Trend Over Time Time High Low Difference Period Reputation Reputation Underwriter Underwriter 80s Mean Median Count s Mean Median Count 1,148 1,027 90s incl. bubble Mean Median Count 1,538 1,213 Significant at the 0.01 level. Significant at the 0.10 level.

12 780 Financial Management Winter 2012 on the EBITDA profit margin (EBITDA/Sales). This procedure gives us nine portfolios for each industry-year. If there are less than three firms in any of the industry-year Sales-EBITDA margin portfolios, we then do a 2 2 sort on sales and EBITDA/sales. If this sort is not sufficient, we then consider only one portfolio per industry-year. For each IPO firm, an appropriate comparable firm is then chosen from the corresponding year-industry Sales-EBITDA margin portfolio having the closest sales value to the sales of the IPO firm. We then estimate the offer price to intrinsic value and secondary market price to intrinsic value ratios of the IPO firms based on the price multiples of their comparable firms as follows: ( ) ( ) OP Offer PriceIPO Shares Outstanding = IPO IV Sales Prior Year Fiscal Sales ( IPO ) (2) Prior Year Fiscal Sales Comparable, Price Comparable Shares Outstanding Comparable ( ) SMP IV Sales ( ) First Day Closing PriceIPO Shares Outstanding = IPO Prior Year Fiscal Sales ( IPO ) Prior Year Fiscal Sales Comparable, Price Comparable Shares Outstanding Comparable (3) ( ) OP = IV EBITDA ( ) Offer PriceIPO Shares Outstanding IPO Prior Year Fiscal EBITDA ( IPO ) Prior Year Fiscal EBITDA Comparable, Price Comparable Shares Outstanding Comparable (4) ( ) SMP IV EBITDA ( ) First Day Closing PriceIPO Shares Outstanding = IPO Prior Year Fiscal EBITDA ( IPO ) Prior Year Fiscal EBITDA Comparable, Price Comparable Shares Outstanding Comparable (5) ( ) OP = IV Earnings ( ) Offer PriceIPO Shares Outstanding IPO Prior Year Fiscal Earnings IPO ( ) Prior Year Fiscal Earnings Comparable, Price Comparable Shares Outstanding Comparable (6) ( ) SMP IV Earnings ( ) First Day Closing PriceIPO Shares Outstanding = IPO Prior Year Fiscal Earnings IPO ( ) Prior Year Fiscal Earnings Comparable. Price Comparable Shares Outstanding Comparable (7)

13 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 781 Here, OP/IV is the offer price to intrinsic value ratio and SMP/IV is the first-day closing secondary market price to intrinsic value ratio for the IPO stock. Offer price of the IPO stock is obtained from the SDC Platinum database. Shares outstanding is the number of shares outstanding immediately after the IPO for IPO stocks and the number of shares outstanding on the day closest to the IPO day for the comparable firm. The shares outstanding data and the market prices are obtained from the CRSP daily files. Earnings, Sales, and EBITDA data are obtained from the annual Compustat files. Based on this procedure, we are able to successfully find matching firms for 2,655 IPO firms. 2. Propensity Score-Based Comparable Firm Approach The propensity score-based approach of obtaining comparable firms has significant advantages over the basic comparable firm approach. First, the propensity score-based approach allows us to match on a larger set of parameters than the basic matching method. Additionally, it allows us to control for a possible endogeneity bias in our intrinsic value calculations. To expand a bit more on the endogeneity issue, it is possible that high-reputation underwriters have information about the future performance of firms and select firms that they expect to perform better in the future. This could potentially bias our intrinsic value calculations. To correct for this, we include three-year ex post average growth in sales, average growth in cost of goods sold, and average growth in selling and general expenses in the propensity score model. These variables allow us to control for the underwriter s private information regarding improvements in future sales, as well as the cost efficiency of the firm. We explicitly control for endogenous matching between underwriters and firms in Section IV.C using instrumental variables and regression discontinuity analyses. Following Chemmanur, Krishnan, and Loutskina (2008), we make use of the nearest match version of the propensity score-based matching approach similar to Dehejia and Wahba (1999, 2002). We run the same filters for the set of candidate and IPO firms as we did in the basic comparables approach. For each fiscal year, we then run a probit model for the set of all IPO and non-ipo firms where the dependent variable takes a value of one if the firm is an IPO firm and zero if it is not. The independent variables in the regression are sales, EBITDA to sales ratio, net income to sales ratio, the average three-year ex post growth in sales, the average three-year ex post growth in cost of goods sold, and the average three-year ex post growth rate in selling and general expenses. We then obtain the predicted value of the probabilities from the probit model and match the IPO firm with a non-ipo firm having the closest predicted probability to that of the IPO firm, while keeping the restriction that the comparable firm must be in the same industry as the IPO firm. Finally, offer price and secondary market price to intrinsic value ratios are calculated as before using Equations (2) (7). Based on this procedure, we are able to successfully find matching firms for 2,277 IPO firms. 3. Discounted Cash Flow Valuation Method We also use the discounted cash flow method to compute the intrinsic value of IPO firms. The most significant benefit of the discounted cash flow method is that we do not have to restrict the sample of firms to have positive earnings or net income. However, we do require that firms have earnings data for at least two years after the IPO date. Intrinsic values computed using this method allows us to check the robustness of the results from the comparables approach.

14 782 Financial Management Winter 2012 The specific model we follow to compute the intrinsic value of the firm is similar to Ohlson (1990). The intrinsic value of the firm is computed as follows: IV = B 0 + EPS 1 r B 0 + (EPS 1 r B 0 ) + (EPS 2 r B 1 ) (1 + r) 2 ( ) 1. (1 + r) (r g) (8) The last term is the discounted terminal value of the stock. B i is the book value of the issuer in year i after the IPO, EPS i is the earnings per share in year i after the IPO. EPS and book value data are obtained from Compustat. We assume a constant required return of 13%. We perform our calculations with two values of terminal growth rates, g = 0 and g = 5%, to check the robustness of our results. Using intrinsic value calculations, we set the terminal value equal to zero if it is negative under the assumption that managers are unlikely to continue negative net present value projects to perpetuity. Finally, we only use firms that have positive intrinsic values for our valuation ratio calculations. Based on this procedure, we are able to compute intrinsic value and valuation ratios for 2,112 firms using the 0% terminal growth rate and 2,113 firms using the 5% terminal growth rate. 16,17,18 B. Underwriter Reputation Measurement Our main test variable is underwriter reputation. We use two different market share-based measures for underwriter reputation. The first measure is based on Megginson and Weiss (1991) and is calculated as the total market share of the lead underwriter in the IPO market over the entire sample period. The second measure is based on the market share of the lead underwriter over a rolling 12-month period prior to the IPO issue date (Cooney et al., 2001). The market share for a given time period is calculated as the ratio of total proceeds issued by the underwriter in that period to the total size of the IPO market in that period. If more than one underwriter acts as a lead underwriter for an issue, we split the proceeds from that offering equally between each of the lead underwriters to calculate their individual market share. We then take the average of the market shares of all the lead underwriters of the issue to calculate the lead underwriter syndicate rank for that issue. We only report results using the Megginson and Weiss (1991) ranking measure in this paper for brevity (MW ranking from here on). However, all our tests are conducted using 16 One firm that had a positive intrinsic value under a 5% growth rate assumption has a negative intrinsic value under a 0% growth rate assumption. 17 There is a potential survivorship bias induced by this methodology since we are using actual future EPS of an IPO firm. However, surviving firms are likely to have higher intrinsic values. Therefore, keeping the IPO valuation constant, a survivorship bias should reduce the difference between offer price and the intrinsic value, thus biasing our results against finding overvaluation of IPOs. More importantly, while a survivorship bias can bias intrinsic value, it is unlikely to bias the IPO and secondary market valuation of a firm for a given intrinsic value. That is, a survivorship bias requires that underwriters value IPOs without any regard to intrinsic value, which is unlikely. Since our analysis focuses on the valuation of a firm in the IPO and the immediate post-ipo aftermarket relative to intrinsic value, a survivorship issue is less likely to affect our results. In addition, since we also conduct our analysis with the comparable firm methodology, which is less likely to be affected by a survivorship bias, we believe that our results are not driven by such biases. 18 We also conduct our tests using an expanded specification: IV = B o + EPS 1 r B 0 (1 + r) + EPS 2 r B 1 + (EPS ( 2 r B 1 ) + (EPS 3 r B 2 ) (1 + r) (1 + r) 2 (r g) The results obtained with this measure of intrinsic value are qualitatively similar as the ones reported in the paper. ).

15 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 783 both ranking methods and the results are qualitatively similar when we use the 12-month rolling rankings. We also measure the reputation of co-managing underwriters in IPOs in a similar manner. We calculate the average market share of the co-managing underwriter group for each IPO as a measure of reputation. The market share for each underwriter who acts as a co-managing underwriter is calculated as the total proceeds in all IPOs for which the underwriter acts as a co-managing underwriter (IPO proceeds are split equally among all co-managing underwriters if there is more than one) divided by the total IPO proceeds within the sample period. As with lead underwriters, we also calculate 12-month rolling rankings for co-managing underwriters and conduct all tests using this ranking method. We use both binary and continuous specifications of underwriter reputation for our tests. The binary high-reputation dummy takes a value of one if underwriter reputation is higher than the median of the sample and zero otherwise. We report results based only on the binary version of the underwriter reputation variable for many of our tests to ensure continuity with univariate comparisons. However, all our multivariate tests are replicated using the continuous market share variables. The results obtained using the continuous variables are similar to the results reported in this paper. C. Proxies for Heterogeneous Beliefs We argue that high-reputation underwriters elicit more market participant interest and this, in turn, may induce higher levels of heterogeneity in retail investor beliefs. This, in conjunction with short sale constraints, is able to obtain higher valuations for IPOs. To examine the relationship between underwriter reputation and heterogeneous beliefs, we identify proxies for heterogeneous beliefs. We motivate our proxy using the model of Harris and Raviv (1993) who argue that when heterogeneous beliefs are higher in the market, there is more trading between agents and this should be empirically measurable. A number of empirical papers in the finance literature (Kandel and Pearson, 1995), as well as the accounting literature (Bamber, 1987; Bamber, Barron, and Stober, 1997) have used trading activity as a proxy for heterogeneous beliefs among investors. The three market-based proxies that we use are daily volume (defined as the log of the price of the stock multiplied by the number of shares traded in that day), daily turnover (defined as the ratio of the number of shares of the IPO stock traded in that day to the number of shares of that stock outstanding on that day), and the number of trades in a given day. The first two proxies are also used in Chemmanur, Krishnan, and Loutskina (2008). We add the number of trades as it is also representative of trading activity for the stock. We obtain the weekly and monthly values of these variables by averaging the daily values over time. To control for possible liquidity effects, we use adjusted values of the statistics described earlier. The adjusted values are calculated as the value of the variable minus the average monthly value of the same variable after three years. The data for calculating the proxies of heterogeneous beliefs are obtained from the CRSP database. IV. IPO Underpricing and Valuation A. IPO Underpricing in the 1980s and 1990s Panel B of Table I demonstrates how underpricing and underpricing differences between highand low-reputation underwriters have evolved over time. In the table, the 80s period includes IPOs from 1980 to 1989, the 90s period includes IPOs from 1990 to 1998, and the 90s including bubble period includes IPOs issued from 1990 to The above results indicate

16 784 Financial Management Winter 2012 that the underpricing difference between high- and low-reputation underwriter backed IPOs has clearly flipped from the 1980s to the 1990s. The difference in underpricing between high- and low-reputation underwriter backed IPOs increases from an average of 1.5% in the 1980s to 7.6% in the 1990s. The median underpricing difference increases from 0.14% (not statistically significant) in the 1980s to about 4.2% in the 1990s. This effect is also documented in Loughran and Ritter (2004) who find that the average underpricing has gone from being lower for highreputation underwriter backed IPOs in the 1980s to being higher for high-reputation underwriter backed IPOs in the 1990s. The above trend highlights our motivating question of the economic role of the underwriter in IPOs. One of the secondary objectives of this paper is to test whether IPO valuation and the economic role of the underwriter in an IPO was fundamentally different in the 1980s and 1990s using alternative measures of the economic role of the underwriter (such as IPO valuation relative to intrinsic value). B. Univariate Results on Valuation The univariate tests for primary and secondary market valuation ratios are reported in Tables II IV. The tables list valuations over the same time periods as the underpricing trend reported in Table I for ease of comparison. Table II presents the comparisons using the basic comparable firm approach. Based on the sales multiple, high-reputation underwriters obtain a median offer price of about 1.57 times the intrinsic value for the stock of the IPO firm, whereas low-reputation underwriters get a statistically significantly lower offer price of about 1.52 times the intrinsic value in the 1980s. The results are similar for first-day closing secondary market prices where highreputation underwriters obtain a median secondary market price that is 1.65 times the intrinsic market value, while low-reputation underwriters obtain a median secondary market price that is 1.58 times the intrinsic value in the 1980s. The differences are even more stark for the 1990s period. Here, the median offer price valuation ratio for IPOs backed by high-reputation underwriters is 0.36 higher than that for IPOs backed by low-reputation underwriters. Further, the median first-day closing secondary market price valuation ratio for IPOs backed by high-reputation underwriters is 0.46 higher than the median valuation ratios for IPOs backed by low-reputation underwriter in the 1990s. These differences are highly statistically and economically significant. The results are also similar for the 1990s including the bubble period, where high-reputation underwriter backed IPOs have higher median offer price valuation ratios and first-day closing secondary market price valuation ratios than low-reputation underwriter backed IPOs. The results obtained above hold when we use valuation ratios based on EBITDA and earnings multiples. To test whether clustering around industry and year can affect the standard errors in our univariate analysis, we calculate the Somers D clustered statistic and the 90% confidence intervals for the Hodges-Lehmann clustered median difference test for all the univariate results presented here (unreported). Our results do not change qualitatively. Similarly, all subsequent multivariate tests on valuations are conducted with and without clustering on industry as well as the year of issue. Our results are robust to clustering the standard errors. Table III reports that the results are similar when we use the propensity score-based comparable firm approach. This method controls for the possibility that high-reputation underwriters could potentially select better firms thereby biasing the basic comparable method results. For the sales multiple-based valuation in the 1980s, there is a statistically significant difference of 0.28 between the median offer price valuation ratios of high- and low-reputation underwriter backed IPOs and 0.32 between the median first-day closing secondary market price ratios of high- and low-reputation underwriter backed IPOs (our results are similar in other time periods as well). The results based on valuation ratios using other multiples are qualitatively similar although the statistical significance using the EBITDA multiple is weaker. Finally, we use the discounted cash

17 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 785 Table II. The Valuation of IPOs Backed by High- and Low-Reputation Underwriters Using the Basic Comparable Firm Approach This table reports the cross-sectional distribution of the ratio of offer price to intrinsic value (OP/ IV) and the first trading day closing secondary market price to intrinsic value (SMP/IV) for IPOs. The intrinsic value is the fair value of the IPO firm computed based on the market price-to-sales, market price-to-ebitda, or market price-to-earnings ratios of an industry peer. The industry peer is a comparable publicly traded firm in the same Fama and French (1997) industry as the IPO firm and has the closest sales and EBITDA profit margin (EBITDA/Sales) in the pre-ipo fiscal year. High-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank higher than the median for the sample. Low- Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank lower than the median. The rankings are calculated as the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). 80s time period corresponds to issues that are placed from 1980 to 1989, 90s time period corresponds to issues placed from 1990 to 1998, and 90s incl. bubble time period corresponds to issues placed from 1990 to For median ratios, the statistical significances correspond to the sign test for median OP/IV (or SMP/IV) equal to one. For differences, the statistical significances are for a Wilcoxon-Mann-Whitney rank sum test for the equality of medians of the two subsamples. The IPOs are from SDC Platinum and all other data are from CRSP and Compustat. OP/IV SMP/IV High Low Difference High Low Difference Reputation Reputation Reputation Reputation Underwriter Underwriter Underwriter Underwriter Sales Multiple 80s Median Count s Median Count s incl. Median bubble Count 1, , EBITDA Multiple 80s Median Count s Median Count s incl. Median bubble Count Earnings Multiple 80s Median Count s Median Count s incl. Median bubble Count Significant at the 0.01 level. Significant at the 0.05 level.

18 786 Financial Management Winter 2012 Table III. The Valuation of IPOs Backed by High and Low-Reputation Underwriters Using the Propensity Score-Based Comparable Firm Approach This table reports the cross-sectional distribution of the ratio of offer price to intrinsic value (OP/ IV) and the first trading day closing secondary market price to intrinsic value (SMP/IV) for IPOs. The intrinsic value is the fair value of the IPO firm computed based on the market price-to-sales, market price-to-ebitda, or market price-to-earnings ratio of an industry peer. The industry peer is a comparable publicly traded firm in the same Fama and French (1997) industry as the IPO firm that has the closest propensity score value based on sales, operating margin (EBITDA/Sales), profit margin (Net Income/Sales), sales growth, cost of goods sold growth, and selling and general expenses growth. High-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank higher than the median for the sample. Low-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has reputation rank lower than the median. The rankings are calculated as the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). 80s time period corresponds to issues that are placed from 1980 to 1989, 90s time period corresponds to issues placed from 1990 to 1998, and 90s incl. bubble time period corresponds to issues placed from 1990 to For median ratios, the statistical significances correspond to the sign test for median OP/IV (or SMP/IV) equal to one. For differences, the statistical significances are for a Wilcoxon-Mann-Whitney rank sum test for the equality of medians of the two subsamples. The IPO data are from SDC Platinum and all other data are from CRSP and Compustat. OP/IV SMP/IV High Low Difference High Low Difference Reputation Reputation Reputation Reputation Underwriter Underwriter Underwriter Underwriter Sales Multiple 80s Median Count s Median Count s incl. Median bubble Count EBITDA Multiple 80s Median Count s Median Count s incl. Median bubble Count Earnings Multiple 80s Median Count s Median Count s incl. Median bubble Count Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level.

19 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 787 Table IV. The Valuation of IPOs Backed by High and Low-Reputation Underwriters Using the Discounted Cash Flow Approach This table reports the cross-sectional distribution of the ratio of offer price to intrinsic value (OP/IV) and the first trading day closing secondary market price to intrinsic value (SMP/IV) for IPOs. The intrinsic value is the fair value of the IPO firm estimated using the residual income model of Ohlson (1990) with a constant discount rate of 13%. In the table, Growth = 5% represents the aggregate sample of IPOs across years where intrinsic values are calculated under the assumption of 5% indefinite earnings growth after Year 3, Growth = 0% growth represents the aggregate sample of IPOs across years where intrinsic values are calculated under the assumption of no earnings growth after Year 3. High-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank higher than the median for the sample. Low-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank lower than the median. The rankings are calculated as the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). In this table, 80s time period corresponds to issues that are placed from 1980 to 1989, 90s time period corresponds to issues placed from 1990 to 1998, and 90s incl. bubble time period corresponds to issues placed from 1990 to For median ratios, statistical significances correspond to the sign test for median OP/IV (or SMP/IV) equal to one. For differences, statistical significances are for a Wilcoxon-Mann-Whitney rank sum test for the equality of medians of the two subsamples. The IPOs are from SDC Platinum and all other data are from CRSP and Compustat. OP/IV SMP/IV High Low Difference High Low Difference Reputation Reputation Reputation Reputation Underwriter Underwriter Underwriter Underwriter Growth = 0% 80s Median Count s Median Count s incl. Median bubble Count Growth = 5% 80s Median Count s Median Count s incl. Median bubble Count Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level. flow approach to value the IPO stocks in Table IV. The reported results are generally consistent with the results above and corroborate our hypothesis that high-reputation underwriters obtain higher primary and secondary market valuations for their clients. We find, in subsequent sections, that our valuation ratios are higher for high-reputation underwriters in multivariate tests, as well as in tests that control for endogenous matching between underwriters and firms. Overall, our empirical results indicate that high-reputation underwriters

20 788 Financial Management Winter 2012 are able to obtain higher valuations in both the primary and the secondary markets for their clients, consistent with Hypotheses (H1A) and (H2). The results in this section are consistent with the idea that the underwriter s primary goal is to get the highest possible valuation for their clients and that high-reputation underwriters are better at obtaining these higher valuations. Moreover, these valuation results are consistent across the 1980s and 1990s in the sense that high-reputation underwriters obtain higher valuations than low-reputation underwriters in both these time periods. In contrast, the underpricing results are not consistent across the 1980s and 1990s. C. Multivariate Results and Endogenous Matching between Underwriter Reputation and Firms 1. Ordinary Least Squares (OLS) Regressions in the Full Sample Panel A of Table V reports the results from the OLS regression of the log of our offer price to intrinsic value ratio and the log of the first-day closing secondary market price to intrinsic value ratio. We control for firm size, whether or not the firm is backed by a venture capitalist, and the fraction of the firm sold in the IPO. In addition, we control for industry and year fixed effects. In these regressions, we use both the dummy and the log of the continuous version of underwriter reputation. The OLS model confirms our univariate results that high-reputation underwriters are associated with higher valuations, both in the IPO and the immediate post-ipo secondary market. In particular, we find that IPOs backed by high-reputation underwriters (based on the high-reputation dummy) are associated with a 37.5% increase in the offer price valuation ratio and a 44.6% increase in the secondary market price valuation ratio. Similarly, a one standard deviation increase in the continuous version of the underwriter reputation variable is associated with a 30% increase in the offer price valuation ratio and with a 35% increase in the secondary market price to valuation ratio. In terms of the control variables, we find that IPOs of larger firms are associated with lower valuations. Further, IPOs of firms backed by venture capitalists are associated with higher valuations, consistent with our expectations. IPOs of firms selling a larger fraction of equity are associated with lower valuations, consistent with investors viewing sales of higher stakes in the firm with greater apprehension. 2. IPO Valuation by Underwriters in Hot and Cold IPO Markets In this section, we test whether the association between underwriter reputation and the valuation of IPO stocks changes depending on whether the IPO market is hot or cold. If high-reputation underwriters systematically pick better firms based on underlying and unobservable quality, then the association between underwriter reputation and the valuation of IPO stocks in hot IPO markets should not be different from that in cold IPO markets (see below for further explanation). Alternatively, if high-reputation underwriters have a greater ability to obtain higher valuation by making investors optimistic, then they should obtain even higher valuations than low-reputation underwriters in hot IPO markets, when investors can be more easily manipulated. We rely on an important assumption for this test. We assume that higher quality firms are equally or less likely to conduct IPOs in hot markets than in cold markets. Various empirical papers have demonstrated this assumption. Helwege and Liang (2004) find no difference either in characteristics such as size, growth potential, and industry or in the post-ipo operating performance of firms whose IPOs are issued in hot markets relative to those whose IPOs are issued in cold markets. Ritter (1991) reports that long-run IPO stock return underperformance is actually higher in high volume years (consistent with IPOs issued in hot markets underperforming relative to those issued in cold markets).

21 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 789 Table V. IPO Valuation for High and Low-Reputation Underwriter Backed IPOs This table reports the result of OLS regressions where the dependent variables are the log of the ratio of offer price to intrinsic value and the log of the ratio of the first trading day closing secondary market price to the intrinsic value for IPOs. The intrinsic value is the fair value of the IPO firm computed based on the market price-to-sales ratio of an industry peer. The industry peer is a comparable publicly traded firm in the same Fama and French (1997) industry as the IPO firm and has the closest sales and EBITDA profit margin (EBITDA/Sales) in the pre-ipo fiscal year. The independent variables are: Underwriter reputation, based on the log of the market share of the IPO underwriter; Underwriter reputation dummy, which takes a value of one if the market share of the lead underwriter is greater than the sample median, and zero otherwise; Size, which is the log of assets of the IPO firm; VC backing dummy; the fraction of firm sold, calculated as the number of shares sold in an IPO as fraction of the number of shares outstanding; industry fixed effects; and year fixed effects. Panel A reports the regression results for the overall sample. Panel B presents the regression results for the sample split by hot and cold IPO markets. The IPOs are from SDC Platinum and all other data are from CRSP and Compustat. Robust t-statistics and z-statistics are reported in parentheses in Panels A and B, respectively. Panel A. IPO Valuation (Full Sample) Offer Offer Secondary Secondary Price to Price to Market Price Market Price Intrinsic Value Intrinsic Value to Intrinsic to Intrinsic Ratio Ratio Value Ratio Value Ratio Underwriter reputation (continuous) (10.682) (11.438) Underwriter reputation dummy (7.281) (8.078) Size ( ) ( ) ( ) ( ) VC backing (3.856) (4.803) (3.937) (4.939) Fraction of firm sold ( 7.742) ( 7.817) ( 9.626) ( 9.789) Constant (8.774) (4.601) (10.464) (5.591) Adj. R Observations 2,655 2,655 2,655 2,655 (Continued)

22 790 Financial Management Winter 2012 Table V. IPO Valuation for High and Low-Reputation Underwriter Backed IPOs (Continued) Panel B. IPO Valuation in Hot and Cold IPO Markets Offer Price to Offer Price to Secondary Market Price Secondary Market Price Intrinsic Value Ratio Intrinsic Value Ratio Intrinsic Value Ratio to Intrinsic Value Ratio Hot Cold Diff Hot Cold Diff Hot Cold Diff Hot Cold Diff Market Market Market Market Market Market Market Market Underwriter reputation (continuous) (11.058) (3.875) (2.500) (10.150) (3.757) (2.360) Underwriter reputation dummy (6.868) (2.206) (1.790) (7.592) (2.596) (1.860) Size ( 9.102) ( 6.445) ( 8.705) ( 5.725) ( 8.805) ( 7.238) ( 9.545) ( 6.560) VC backing (3.808) (1.460) (4.041) (2.720) (3.843) (1.439) (4.181) (2.738) Fraction of firm sold ( ) ( 4.747) ( ) ( 4.212) ( ) ( 6.412) ( ) ( 6.330) Constant (8.594) (3.069) (3.655) (2.471) (10.338) (3.517) (5.840) (2.949) Adj. R Observations 1, , , , Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level.

23 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 791 The assumption that the quality of firms is not different between hot and cold markets allows us to keep the effect of any endogenous matching between underwriters and firms on valuation constant across hot and cold IPO markets. Further, the possibility that hot IPO market firms quality may be worse than cold IPO market firms quality, on average, predicts that the incremental effect of underwriter reputation on IPO valuation will be lower in hot IPO markets as compared to cold IPO markets (based on the endogenous matching conjecture). In contrast, if high-reputation underwriters can obtain higher valuation for the IPOs that they back by manipulating investor sentiment, then we expect that the effect of underwriter reputation on IPO valuation will be higher in hot IPO markets when compared to cold IPO markets. This is because high-reputation underwriters will find it easier to manipulate retail investor beliefs in hot markets when investors tend to be more optimistic. Evidence from Helwege and Liang (2004) is consistent with this argument. We define hot IPO markets by using the three-month moving averages of the number of IPOs prior to the sample IPO. Specifically, if the average number of IPOs in the three months prior to the sample IPO is greater than the sample median, we categorize the IPO to be issued in a hot market, and vice versa. Panel B of Table V reports the results of the OLS regressions of the IPO valuation ratios split by hot and cold IPO markets. We find that while underwriter reputation has positive coefficient estimates in both hot and cold IPO markets, the coefficient estimate on underwriter reputation is significantly higher in hot markets than in cold markets. Economically, a one standard deviation increase in underwriter reputation results in an 18% higher increase in IPO offer time valuation ratio in hot markets as compared to that in cold markets. Similarly, based on the high-reputation underwriter dummy, we find a 22% greater association between underwriter reputation and valuation in hot IPO markets than in cold IPO markets. In summary, we find evidence that the positive association between underwriter reputation and IPO valuations is consistent with the manipulation of investor beliefs by IPO underwriters, consistent with Hypotheses (H1A) and (H2). 3. Endogenous Matching between Underwriter Reputation and Firms: Regression Discontinuity Analysis In this section, we analyze the causal effect of underwriter reputation on IPO valuation using a regression discontinuity (RD) analysis. In particular, we use the fuzzy RD technique discussed in Van der Klaauw (2002) and Angrist and Lavy (1999). Identification in a fuzzy RD technique comes from a discontinuous jump in the expected probability of treatment. Van der Klaauw (2002) and Angrist and Lavy (1999) analyze the effect of jumps in continuous treatment variables on their variables of interest. We use the two-stage approach discussed in the papers above to implement our RD analysis. We estimate the following model: y j = x j β + δz j + ε j, (9) where z j is the potentially endogenous treatment variable. The fuzzy RD approach requires a variable (we will call it the assignment variable) S j such that z j experiences a jump at a known point S 0. Note that unlike an instrumental variables approach, S j can be correlated with the error variable, ε j. The restriction required for identification is that no other effect causes a jump in y j at S j = S 0. The RD can then be implemented with a two-stage approach. We first estimate z j = f (S j ) + γ.1{s j S 0 }+η j, (10) where f (S j ) is a continuous function of S j, 1{S j S 0 } is an indicator function that is equal to one if S j S 0, and η j is an error term. We augment the second-stage equation with the conditional

24 792 Financial Management Winter 2012 expectation E(z j S j ), which is calculated as the predicted value from Equation (10). The second stage model is then y j = x j β + δ.e(z j S j ) + λ.s j + u j. (11) We use a piecewise quadratic functional form for f (S). Specifically, f (S) = φ 1 S + φ 2 S 2 + φ 11 (S S 0 )1{S j S 0 } + φ 12 (S S 0 ) 2 1{S j S 0 }. The standard errors in the second stage are bootstrapped to account for generated regressors. This approach is analogous to an instrumental variables approach, with the discontinuity 1{S j S 0 } serving as the identifying instrument. We exploit the discontinuity arising from NYSE listing requirements for net tangible assets from the sample period NYSE required $16 million in assets for listing until 1983, $18 million from 1984 to 1994, and $40 million from 1995 to 1997 for a company to list on the exchange. After that, NYSE listings did not have restrictions based on assets (see, e.g., Anderson and Dyl, 2008 and Cowan, Carter, Dark, and Singh 1992). Therefore, to account for the changes in the NYSE listing requirements over time (using the sample from 1980 to 1997), our assignment variable, S j, is defined as the log-normalized level of assets ( S j = log Firm assets NYSE asset listing requirements ). (12) Our identification strategy depends on the idea that firms that are larger than the listing requirement for NYSE (i.e., S j 0) will be more likely to list on the NYSE. 19 Evidence in prior literature indicates that NYSE listings are related to higher liquidity and more visibility and investor recognition for the firm (Grammatikos and Papaioannou, 1986; Kadlec and McConnell, 1994), thus reducing the necessity of a high-reputation underwriter. Consequently, while the endogeneity argument would predict that underwriter reputation is positively related to firm size (to the extent that firm quality is correlated to size), this relation would reverse in a discontinuous manner at the point where the firm becomes eligible to list on the NYSE since high-reputation underwriters services, such as creating liquidity and investor recognition are, to some extent, substituted by the NYSE listing. We provide a graphical analysis of the discontinuity. Figure 1 graphs kernel-weighted mean smoothing estimations between underwriter reputation (measured as the log of underwriter market share) and the normalized size variable. The graph clearly indicates that there is a sudden decline in the underwriter reputation at zero, as expected, supporting our identification strategy. The results of the RD analysis are reported in Table VI. Consistent with our expectations, the coefficient estimate on the jump point indicator (1{Norm. Size > 0}) is statistically significant and negative in the first-stage regression indicating that there is a downward jump in the underwriter reputation at the point where a firm becomes eligible to list on the NYSE based on assets. In the second stage, we find that the coefficient estimate on the predicted underwriter reputation from the first stage is positive and statistically significant. Thus, after controlling for self-selection between firms and reputed underwriters using the RD approach, we find that high-reputation underwriters are linked to higher valuations, both in the IPO and the first-day closing secondary market, which is consistent with (H1A) and (H2). 19 NYSE also has other restrictions for firms to be eligible for listing. Some listing requirements, such as total market value of shares, can be directly influenced by underwriters. As such, these variables are unsuitable as candidates to obtain a discontinuity for our analysis. In addition, income requirements are more complicated and are based on different levels of required income levels in one, two, and three years prior to the IPO making the analysis less tractable.

25 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 793 Figure 1. Discontinuity of the Treatment (Underwriter Reputation) When Normalized Size =0 One concern with the assignment variable is that firms may be able to manipulate their size levels to comply with NYSE requirements. McCrary (2008) argues that such active sorting may undermine the identification requirements of the RD design. To rule out this concern, we graphically and statistically analyze the density function of the normalized size variable around the cutoff point (at S j = 0). Visual analysis of the graph indicates no upward jump of S j at the cutoff point. We then conduct the statistical test recommended in McCrary (2008) to ensure that our identification strategy is valid. The test is implemented as a Wald test of the null hypothesis that the discontinuity in the density function of the assignment variable is zero at the cutoff point. The estimator for this test is calculated in two steps. First, we obtain a finely gridded histogram. Then, we smooth the histogram using local linear regressions separately on either side of the cutoff point. The test statistic is based on the estimates of the density function from local linear regressions from the two sides of the cutoff point. The test does not find a statistically significant jump in the density function of the assignment variable S j at the cutoff point S j = 0 validating our identification strategy Endogenous Matching between Underwriter Reputation and Firms: Instrumental Variables Analysis We use instrumental variables analysis to demonstrate the causal effect of underwriter reputation on IPO valuation. We use two instruments to satisfy exclusion restrictions. The first instrument 20 Another concern with our identification strategy is that NYSE listed firms may have higher valuations if the NYSE listing itself provides greater investor recognition and reduces information asymmetry for the firm going public. However, since the eligibility for NYSE listing is negatively related to the propensity of using a higher reputation underwriter in Table VI, such an effect would diminish the effect of the predicted underwriter reputation on IPO valuation, biasing the results against finding a positive relation between underwriter reputation and IPO valuation.

26 794 Financial Management Winter 2012 Table VI. Endogenous Matching between Underwriter Reputation and Firm Quality: Regression Discontinuity Analysis This table reports the results of a regression discontinuity analysis. The dependent variables are: underwriter reputation in the first stage and the ratio of offer price and the immediate post-ipo secondary market price to intrinsic value ratio in the second stage. The independent variables in the first stage are: 1{Norm. Size > cutoff} which is a dummy variable equal to one if the normalized assets of the IPO firm is greater than the cutoff required for NYSE listing, where the normalized assets is the log of the ratio of assets of the firm divided by the NYSE listing cutoff for a given year; Norm. Size; Norm. Size 2 ;1{Norm. Size > 0} (Norm. Size 0); 1{Norm. Size > 0} (Norm. Size 0) 2 ; the fraction of the firm sold in the IPO firm; and the VC backing dummy. Underwriter reputation is calculated as the log of the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). The independent variables in the second stage are the predicted underwriter reputation from the first stage; Norm. Size; the fraction of the firm sold in the IPO firm; and the VC backing dummy. Year and industry fixed effects are included in every specification. Robust t-statistics are reported in parentheses for the first-stage regression and bootstrapped z-statistics are reported in parentheses in the second-stage regressions. First Stage Second Stage Underwriter Offer Price Secondary Reputation to Intrinsic Market Price (Continuous) Value Ratio to Intrinsic Value Ratio 1{Norm. Size > 0} Underwriter reputation ( 2.196) (2.575) (2.293) Norm. Size Norm. Size (4.800) ( 4.203) ( 4.018) Norm. Size ( 0.150) 1{Norm. Size > 0} (Norm. Size 0) (0.908) 1{Norm. Size > 0} (Norm. Size 0) 2 ( 1.880) Fraction of firm sold Fraction of firm sold ( 2.640) ( 5.435) ( 6.279) VC backing VC backing (9.015) (0.175) (0.406) Constant Constant ( ) (3.647) (3.451) Observations 2,154 Observations 2,154 2,154 Adj. R Adj. R Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level. is the location concentration of the underwriter used in the IPO. This is defined as the Herfindahl index of the amount of IPOs underwritten by the lead underwriter across different states measured over five years prior to the sample IPO. IPO underwriters that concentrate on fewer geographic locations are less likely to gain large market shares. We expect a negative relationship between

27 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 795 this variable and underwriter reputation. 21 The second instrument that we use is the availability of underwriters that have experience in the industry of the firm based on Fama-French industries. This variable is the unique count of underwriters that have underwritten IPOs in the same industry as the IPO firm in the past five years. More underwriters with experience in the industry of the firm increase the bargaining power of the firm that may reduce the cost of hiring a high-reputation underwriter. 22 Thus, we expect this instrument to be positively correlated with underwriter reputation. The results of our instrumental variables analyses, implemented as a two-stage least squares model (using the continuous version of underwriter reputation) and a treatment effects model (using a binary version of underwriter reputation) are reported in Table VII. We present the results of the first and the second stage models separately. The standard errors in the second stage are corrected for the estimation error from the first stage (Maddala, 1983). In the first stage of the twostage least squares model, we find that underwriter location concentration is negatively associated with underwriter reputation and industry underwriter availability is positively associated with underwriter reputation. The first-stage F-statistic (for the test of the joint significance of the coefficient estimates on the two instruments in the first stage) is and highly statistically significant at the 1% level, and the partial R 2 measuring the correlation between the endogenous variable and instruments after partialing out the effect of the other exogenous variables is 23.2%. These measures suggest that our instruments are significant predictors of underwriter reputation (Staiger and Stock, 1997; Stock and Yogo, 2005). Since we have two instruments, we are able to conduct the test of overidentifying restrictions and are unable to rule out the null hypothesis that our instruments are exogenous. We find that after controlling for the endogenous selection of reputable underwriters using twostage least squares analysis, high-reputation underwriters are associated with higher valuations, both in the IPO and the immediate post-ipo secondary market. Our results for the treatment effects model remain similar to those from the two-stage least squares model. Thus, using instrumental variables analyses, we find the same results as in the previous two subsections, namely IPO underwriter reputation has a positive causal impact on the valuation of the stock in IPOs that they back, consistent with Hypotheses (H1A) and (H2). D. Dynamics of Valuation Ratios We analyze how valuation ratios of IPO stocks evolve over time over a period of one, two, and three years after the IPO date. To do this, we calculate the intrinsic value of IPO stocks using the valuation methodologies discussed earlier for one, two, and three years after the IPO and calculate the ratios in each of these periods. For both the basic and propensity score-based comparable firm approaches, we find a new matched firm every year for one, two, and three years after the IPO date. Table VIII reports the dynamics for the sales and earnings multiple-based valuation ratios for both the basic comparable firm and the propensity score-based comparable firm approaches. The 21 An argument against such an instrument is that by localizing, underwriters are able to specialize in their local areas and thus can obtain higher valuation through better certification. Therefore, we split our sample into quartiles based on the location concentration variable and analyze the difference in the valuation of IPOs relative to the intrinsic value of the highest and the lowest quartile. We do not find a statistically significant difference between these two groups. In addition, to control for the possibility that any other firm related state-specific effect may be contained in this variable, we control for the fixed effects of the state in which the sample IPO firms are located. 22 As before, the instruments for multiple lead underwriters in an IPO underwriting syndicate are the average values of the instrument across all the lead underwriters.

28 796 Financial Management Winter 2012 Table VII. Endogenous Matching between Underwriter Reputation and Firm Quality: Instrumental Variables Analysis This table reports the results of the two-stage least squares model and treatment effects estimations. The dependent variables in the second stage of the two-stage least squares models and the treatment effects models are the log of the ratio of offer price to intrinsic value and the log of the ratio of the secondary market price to intrinsic value. The intrinsic value is the fair value of the IPO firm computed based on the market price-to-sales ratio of an industry peer. The industry peer is a comparable publicly traded firm in the same Fama and French (1997) industry as the IPO firm and has the closest sales and EBITDA profit margin (EBITDA/Sales) in the pre-ipo fiscal year. The 2SLS and the treatment effects models estimate two stages. In the first stage (selection stage), the dependent variables are the underwriter reputation dummy (in the treatment effects model) and the log of the underwriter IPO market share (in the two-stage least squares model). The independent variables are: Underwriter Location Concentration, which is the state level of concentration of the underwriters IPO business and is measured as the Herfindahl index of the amount of IPOs underwritten by the lead underwriter in various states, calculated over a five-year period prior to the sample IPO; Industry Underwriter Availability, which is the number of unique underwriters that conduct business in the same industry as the IPO firm, measured over five years prior to the sample IPO; Size, which is the log of assets of the IPO firm; VC Backing Dummy; Fraction of Firm Sold, calculated as the number of shares sold in IPO as a fraction of the number of shares outstanding; industry fixed effects; and year fixed effects. Firm state effects are included in the 2SLS and the treatment effects models. The IPOs are from SDC Platinum and all other data are fromcrspand Compustat. Robust t-statistics are reported in parentheses. Two-Stage Least Squares Treatment Effects First Stage: Second Second First Stage: Second Second Underwriter Stage: Stage: Underwriter Stage: Stage: Reputation Offer Price Secondary Reputation Offer Price Secondary to Intrinsic Market Price Dummy to Intrinsic Market Price Value Ratio to Intrinsic Value Ratio to Intrinsic Value Ratio Value Ratio Underwriter reputation (continuous) (3.057) (3.472) Underwriter reputation dummy (3.204) (3.852) Size (23.010) ( 8.011) ( 8.852) (13.201) ( 9.093) ( ) VC backing (8.620) (3.881) (4.108) (6.695) (3.476) (3.612) Fraction of firm sold ( 3.700) ( 7.052) ( 8.947) ( 2.885) ( ) ( ) Underwriter location concentration ( ) ( 13.76) Industry underwriter availability (2.040) (1.940) Constant ( 5.170) (3.204) (4.936) (0.008) (0.310) (0.742) Adj. R Wald Chi. Sq. 1, , Observations 2,318 2,318 2,318 2,318 2,318 2,318 Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level.

29 Chemmanur & Krishnan Economic Role of the Underwriter in IPOs 797 Table VIII. Dynamics of Market Valuation for IPOs Backed by High and Low-Reputation Underwriters This table presents the ratio of offer price (OP/IV) and secondary market price to intrinsic value (SMP/IV) for IPOs over time. The data set contains IPOs. The intrinsic value is the fair value of the IPO firm computed based on the market price-to-sales, market price-to-ebitda ratio, or the market price-to-earnings ratio of an industry peer. For the basic matching approach, the industry peer is a comparable publicly traded firm generated by the basic comparable firm approach. For the propensity score approach, the industry peer is a comparable publicly traded firm in the same Fama and French (1997) industry as the IPO firm that has the closest propensity score value based on sales, operating margin (EBITDA/Sales), profit margin (Net Income/Sales), sales growth, cost of goods sold growth, and selling and general expenses growth. OP/IV is the ratio of the offer price to the estimated intrinsic value of the IPO stock. SMP/IV t is the ratio of the closing price on the secondary market in year t after IPO to the estimated intrinsic value of the IPO stock at year t. High-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank higher than the median for the sample. Low-Reputation Underwriter stands for IPOs whose lead underwriting syndicate has a reputation rank lower than the median. The rankings are calculated as the average market share of the lead underwriting syndicate over the sample period based on Megginson & Weiss (1991). For median ratios, statistical significances correspond to the sign test for median OP/IV (or SMP/IV) equal to one. For differences, the statistical significances are for the Wilcoxon-Mann-Whitney rank sum test for the equality of medians of the two subsamples. The IPOs are from SDC Platinum and all other data are from CRSP and Compustat. Sales Multiple Earnings Multiple High Low Difference High Low Difference Reputation Reputation Reputation Reputation Underwriter Underwriter Underwriter Underwriter Panel A. Basic Comparable Firm Valuation Dynamics OP/IV SMP/IV SMP/IV SMP/IV SMP/IV Panel B. Propensity Score-Based Comparable Firm Valuation Dynamics OP/IV SMP/IV SMP/IV SMP/IV SMP/IV Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level. table lists the valuation ratios at offer time, the first-day closing, and one, two, and three years after the IPO. The dynamics for the propensity score-based price ratios are graphed in Figure 2. We find two interesting patterns of valuation ratios over time that significantly support our hypotheses and our intrinsic valuation methodology. First, the median valuation ratios decrease over time for both high- and low-reputation underwriter backed IPOs and move toward unity. Consider, for example, the sales multiple-based ratio dynamics for the propensity score-based comparable firm

30 798 Financial Management Winter 2012 Figure 2. Dynamics of Sales- and Earnings-Based Valuation Ratios Using Propensity Score-Based Comparable Firm Approach approach in Panel B. The ratio decreases from at offer time to at three years after the IPO date for high-reputation underwriter backed IPOs. For the low-reputation underwrite backed IPOs, the ratio decreases from at the offer time to at three years after the IPO date. We find similar declines in valuation ratios for the earnings multiple-based ratios. Further, the analysis obtains similar results even when we use the basic comparable firm approach. Given that the effect of the underwriter diminishes over time and that our valuation measures are not biased, this is the convergence pattern one would expect to see over time. This gives credibility to our valuation methodology since one would expect that the effect of underwriters on valuations fades over time and the valuations of IPOs converge toward their intrinsic values. Another aspect of the valuation dynamics is that the difference in valuation ratios between high and low-reputation underwriter backed IPOs is decreasing over time after the IPO. For example, for the sales multiple-based ratio dynamics using the propensity score-based approach, the difference in valuation between high and low-reputation underwriter backed IPOs starts from