Venture Capital Valuation, Partial Adjustment, and Underpricing: Behavioral Bias or Information Production? *

Transcription

1 This article is forthcoming in The Financial Review. Venture Capital Valuation, Partial Adjustment, and Underpricing: Behavioral Bias or Information Production? * Jan Jindra a and Dima Leshchinskii b November 2014 JEL Codes: D82, G02, G24 Keywords: Initial public offering, IPO underpricing, prospect theory, anchoring, asymmetric information. * We thank Wolfgang Bessler, Zhiwu Chen, Douglas Cumming, Jean Helwege, Sophia Johan, Alexander Ljungqvist, Paul Schultz, Derek Stimel, Robert Van Ness (the editor), Ellie Yin, Jan Zimmerman, two anonymous referees, and participants at Menlo College Seminar Series, the Academy of Behavioral Finance and Economics Conference 2011, Midwest Finance Association Annual Meeting 2013, and Financial Management Association Meeting 2013 for helpful comments and discussions. a Jan Jindra, The U.S. Securities and Exchange Commission, 44 Montgomery Street, San Francisco, CA 94104; jj@janjindra.com; phone: The Securities and Exchange Commission, as a matter of policy, disclaims responsibility for any private publication or statement by any of its employees. The views expressed herein are those of the author and do not necessarily reflect the views of the Commission or of the author s colleagues upon the staff of the Commission. b Dima Leshchinskii, Menlo College, Atherton, CA 94027; dleshchinskii@menlo.edu; phone:

2 Venture Capital Valuation, Partial Adjustment, and Underpricing: Behavioral Bias or Information Production? ABSTRACT Using a sample of venture capital (VC) backed IPOs, we analyze the role played by perceived valuation changes on IPO underpricing. We find that perceived valuation change from the last pre- IPO VC round to the IPO affects IPO underpricing in a non-linear way. Further analysis indicates that information-based theories, not behavioral biases, explain this non-linearity. We also find that the previously documented partial adjustment effect and its non-linear impact on IPO underpricing are related to the trajectory of the perceived valuation changes, which stands in stark contrast to prior evidence of the importance of behavioral biases.

3 1. Introduction There is growing evidence that various corporate decisions are affected by behavioral biases. 1 Specifically, with respect to IPO research, Loughran and Ritter (2002) propose that issuers do not get upset about IPO underpricing after the offering. This is because the issuers look at the net change in their wealth and exhibit much less dissatisfaction with share underpricing. Ljungqvist and Wilhelm (2005) find that managers who experience increases in personal wealth, that exceed the losses due to IPO underpricing, are less likely to switch underwriters for a followon equity offering. However, research also shows that information coming to light during the book building process affects the pricing of IPOs in a manner consistent with theories unrelated to behavioral biases (Cornelli and Goldreich, 2003; Hanley, 1993; Lowry and Schwert, 2004). Therefore, the relevance of behavioral biases for IPO pricing remains an important issue. In our research, we compare the relative importance of behavioral biases and information-based theories on IPO underpricing. Two examples illustrate the focus of our study. First, the LinkedIn IPO took place on May 19, 2011, during a time of otherwise slow IPO activity. Prior to its IPO, LinkedIn raised money from VCs and other strategic investors in several rounds, giving it valuation of about $2 billion based on the last round. LinkedIn s IPO shares were priced at $45 per share, implying a $4 billion valuation. Its stock price closed at $94.25, more that 109% above its IPO price on the first day of trading. A second example involves Zymogenetics Inc. which achieved valuation of $430 million or about $11.70 per common share based on the last pre-ipo round valuation. About a year later, 1 For example, a firm s 52-week high price serves as a reference point for an acquisition premium (Baker, Pan, and Wurgler; 2012) and borrowers and lenders when negotiating terms of a new loan anchor on past loan terms (Dougal, Engelberg, Parsons, and Van Wesep, 2011).

4 Zymogenetics filed for IPO with the initial filing IPO price range of $16 to $18. Five months later, Zymogenetics went public at $12, a price only slightly above its pre-ipo venture capital (VC) valuation and well below the initial filing price range. On the first day of trading, Zymogenetics shares closed 0.4% above the IPO price. We analyze whether the IPO underpricing is explained by behavioral biases of the issuer or by information collected during the book building period. Specifically, is it the irrational anchoring by LinkedIn and Zymogenetics managers on pre-ipo valuations that led to differing bargain efforts and ultimately resulted in dramatically different IPO underpricing for the two firms, or is it the information collected and observed during the book building period that had a dominant effect on IPO pricing? Using a comprehensive sample of firms going public between 1993 and 2013 with available pre-ipo valuations from earlier VC rounds, we calculate the change in firm value from the time of the VC round to the time of the IPO filing. 2 We find that firms with valuation increases between the last VC round and the IPO filing have median IPO underpricing of 20.5% and firms with VC-to-IPO valuation declines experience median IPO underpricing of only 7.2% (Figure 1). While this result may be consistent with predictions based on the prospect theory and anchoring, it may also be explained by the existing information-based theories of IPO underpricing and the partial adjustment phenomenon (Hanley, 1993). Specifically, issuers suffering from behavioral biases anchor on VC valuations, and to the extent that the IPO filing valuation exceeds the anchor, they do not negotiate aggressively to increase the IPO price. This would lead to higher underpricing for firms with perceived valuation increases from VC round to IPO filing. However, during the book building process, additional information may get incorporated into the IPO price. Therefore, 2 We define valuation change as the percentage difference between the valuation of all firm equity at the last VC round including preferred stock and at the filing midpoint IPO price, excluding the number of primary shares issued in the IPO. 2

5 the IPO underpricing difference may reflect this information production. In particular, to the extent that the adjustment to the information produced during the book building is asymmetric as in Hanley (1993), firms for which positive information comes to light experience higher underpricing than firms for which negative information comes to light. To assess the importance of behavioral biases and information-based theories, we examine the effect of perceived valuation changes from the last VC round to the IPO filing and to the actual IPO. Based on the trajectory of the perceived valuation change, we identify six unique groups based on the VC, IPO filing, and actual IPO valuations. While, the predictions of behavioral biases and information-based theories coincide for five of these groups, the predictions of the two theories differ for one group this group includes firms whose IPO valuation is above the filing valuation but is still below their VC valuation. Focusing on this group, behavioral bias theory predicts low IPO underpricing as the reference point is below the IPO price and induces higher level of effort from the issuer to set the IPO price higher, thus reducing underpricing, while the partial adjustment to the information predicts high IPO underpricing. We find that IPO underpricing is best explained by information-based theories and not by behavioral biases. The first contribution of our research is the surprising lack of direct effect of behavioral biases on IPO underpricing. This result stands in stark contrast to prior evidence that shows the importance of behavioral biases in other settings that involve interaction among sophisticated parties. Since our analysis considers the information produced during the book building process, we provide insights into the well-documented partial adjustment phenomenon (Hanley, 1993). Hanley finds that the effect of the partial adjustment phenomenon on IPO underpricing is asymmetric: positive IPO price revisions 3 lead to larger underpricing than in the case for negative 3 We define IPO price revision as in Hanley (1993): the percentage change from the initial filing midpoint to the IPO offer price. 3

6 IPO price revisions. We show that this asymmetric partial adjustment effect is further related to past VC valuation. Our analysis shows that for firms for which filing valuation exceeds (is below) the valuation at the last VC round, the partial adjustment effect is amplified (muted) by the VC valuation. Therefore, the second contribution of our research is the finding that behavioral biases contribute to the non-linear effect of the partial adjustment phenomenon. Our work closely relates to a theoretical model of Loughran and Ritter (2002) that considers the effects of behavioral biases for firms going public; and, to Ljungqvist and Wilhelm (2005) who provide test of the model. Our work is different from these two papers in that neither Loughran and Ritter or Ljungqvist and Wilhelm provide direct evidence on the effect of behavioral biases on IPO underpricing. Instead, Loughran and Ritter explain why issuers have less ex post regrets with IPO underpricing and do not address why issuers do not negotiate higher IPO price ex ante. 4 Our research directly studies this issue for one set of companies the ones that used VC financing before the IPO. 2. Literature and Analytical Framework 2.1. VC Financing An important feature of our study is VC financing and the resulting observable pre-ipo valuation of entrepreneurial start-up companies. By the late 1990 s about 80% of VC financing 4 Consider a hypothetical company owner of a firm about to go public. Prior to the IPO, she owns 50% or 500,000 of the firm s shares. The firm raises $2 million in the IPO by issuing 100,000 primary shares at $20 each. On the first trading day the shares close at $30. Therefore, $1,000,000 was left on the table. Despite suffering dilution due to the high underpricing, the owner is content because her wealth increased on the first day of trading by $5 million while the negative effect of the dilution on her wealth is smaller. What this hypothetical example and, similarly, the analysis of Loughran and Ritter (2002) and Ljungqvist and Wilhelm (2005) do not address, is why the owner so easily agreed to $20 per share valuation and did not attempt to negotiate for a higher IPO price during the book building process. Both papers explain, why the owner does not have the ex post regrets, but ex ante the owner should be concerned about valuing at $20 stock that is worth more than that. 4

7 was done by VC funds organized as limited partnerships with two types of partners: several limited partners (LPs) and one general partner (GP, the VC firm itself) (Gompers and Lerner, 1999). The LPs provide money and, in return, receive a share of the fund s profits. The GP manages the fund, that is, identifies and invests in portfolio companies, executes due diligence, monitors the portfolio firms, and assists them with managerial expertise and guidance. The GP s personal investment generally accounts for only 1% of total capital invested, while he/she receives about 20% of the capital gains, usually called carried interest (or carry ). Therefore, the GP has incentives to demonstrate success of the fund often measured by the profitability of the investments and the number of companies brought to IPO. VC funds have a limited life, usually from eight to twelve years; however, GPs very often raise follow on VC funds based on the successful performance of their past funds. 5 VC funds usually provide financing to their portfolio companies in several stages or rounds. During each round, venture capitalists (VCs) provide capital to private companies in return for a certain number of company shares to be converted into common shares at the time of IPO. Often, more than one VC fund makes an investment at each round several funds organize a syndicate and invest money together. 6 Each financing round provides information about the company s valuation agreed to by both the company insiders and the VCs (members of the syndicate). This valuation performed by professionals outside of the firm (i.e., VCs) serves as a 5 For detailed descriptions of VC financing see, e.g., Sahlman (1990), Gompers and Lerner (1999), and Kaplan and Stromberg (2003). 6 Lerner (1994) studied syndication of VC investment and showed that when experienced venture capitalists invest for the first time in later rounds, the firm is usually doing well, although there could be an element of window dressing experienced VCs could invest in companies which are already doing well and more likely to go public. Kandel, Leshchinskii, and Yuklea (2011) conjecture that VC syndicates which include VC funds of different age, will increase the quality of IPOs by preventing older VC funds from bringing unprepared companies to market. 5

8 reference point at a subsequent round of financing or an IPO. 7 Frequently, but not always, such information is publicly available. We note that VC contracts often have an anti-dilution provision, which adjusts upward the number of common shares received by VC, if the company issues equity at a price below the last round valuation (Gompers, 1999), hence, the valuation at each round has important implication for future capital raising. In our study, we focus on the last round of VC financing immediately prior to the IPO, frequently referred to as bridge financing. It is important in our work to consider that VC firms are heterogeneous. For example, they differ in their experience, focus, and success. Exit through IPO has been traditionally considered as a clear indicator of successful investment. Several studies focus on identifying the factors leading to a successful IPO. Nahata (2008) finds that IPOs of firms backed by more reputable VCs are more successful. Nahata reports that one particular VC characteristic emerges as the only consistently robust predictor of VC and portfolio company performance. This measure is the cumulative market capitalization of IPOs backed by the VC firm. Lerner (1994) and Cumming and Dai (2011) use VC size as a proxy for reputation. Using Canadian data Cumming (2006) and Johan (2010) indicate that quality of VC-backed IPOs also depends on the VC firm s affiliation. Most of the VC firms are independent, but some of them are affiliated with banks or other financial institutions, large corporations, and other entities. 8, 9 Since VC valuation and IPO pricing are affected by the type of VCs, we control for VC characteristics in our analysis. 7 Grants of stocks and stock options to employees that take place between the VC valuation rounds or between a VC round and IPO are also awarded based on prior VC round valuation. 8 Gompers (2002) provides an overview of the history of corporate venture investment and argues that corporations have learned to be as successful investors as independent VCs, especially when they have a strategic fit with portfolio companies. Unlike independent VCs, funds of affiliated VCs often have unlimited life span. 9 We also analyze the effect of length of the VC investment from the time of its first investment in the firm (VC duration) on IPO underpricing since Cumming and Johan (2010) show heterogeneity in VC duration. In untabulated 6

9 2.2. IPO Process and IPO Underpricing While there are several IPO mechanisms, the most common mechanism in the U.S. is book building. 10 At the first step in the book building process a company planning its public offering chooses the book-running manager (underwriter) (Ellis, Michaely, and O Hara, 2000). The underwriter assembles a syndicate to assist in the sale of company shares to the public and performs the due diligence process. The company files its registration statement with the SEC and issues a preliminary prospectus. The prospectus, which contains, among other information, IPO price range and number of shares to be issued, is available to investors when the company and the underwriter promote the IPO through a road show. As the road show progresses, the underwriter receives indications of interest as well as other relevant information from investors (Cornelli and Goldreich, 2003). Under Regulation M, no shares can be sold at this stage of the IPO process, no purchase orders can be solicited by the underwriters, and any indications of interest to purchase shares in the offering or thereafter are not legally binding. However, during the road show, the extent of the interest for a firm s shares becomes known to both the underwriter and the issuer. Prior to the issuance, the ultimate IPO offer price and the exact number of shares to be sold are determined. Both the price and the size of the issue can be different from what was detailed in the original prospectus. After the final terms are negotiated and an agreement is reached, the underwriter and the issuer execute the Underwriting Agreement, the final prospectus is produced, and the underwriter files a price amendment on the morning of the chosen effective date. At this point the underwriter begins selling the securities at the agreed upon price. Once approved, the results we note that VC duration does not affect IPO underpricing significantly and its inclusion in the regressions does not affect our conclusions. 10 Derrien and Womack (2002) compare auction and book building mechanisms. Aussenegg, Pichler, and Stomper (2006) study German markets, which permit when-issued trading prior to the IPO. They show that this trading reveals relevant information for pricing IPOs, but it does not fully supplant book building as a source of information. Benveniste and Busaba (1997) compare the fixed price method and book building and show that book building generates higher expected proceeds at the cost of greater uncertainty. 7

10 distribution of the stocks to interested investors begins. Subsequently, the company s stock begins to trade for the first time. A firm s readiness to go public invariably plays a role in the IPO pricing. Johan (2010) studies the difference between the actual IPO date and the initially disclosed intended IPO date, the latter being part of filings by Canadian firms seeking to go public. 11 The results indicate that time difference between the intended and actual IPO dates is smaller for firms of higher quality and firms using high quality auditors. Johan (2010) also reports bigger time differences for VC-backed firms and for firms with more reputable underwriters. Hence, we develop a related measure to control for a firm s preparedness to go public as well as for underwriter reputation. Lack of readily available data for our entire sample and lack of cross-sectional variation in the available data preclude us from controlling for auditor quality Relevant Theories and Hypotheses Numerous theoretical explanations for the IPO process and the resulting pricing exist and broadly fall into theories based on information considerations and behavioral biases. 13 Since we are studying the effects of VC valuation on IPO underpricing, we start with the more recent behavioral theories which indicate that VC valuation can play an important role in explaining IPO pricing. 11 We thank the anonymous referee for alerting us to Johan (2010). 12 Specifically, we collect data on identity of auditors from Audit Analytics. We are able to obtain auditor data for only 244 IPOs in our sample. We note that 230 firms out of the 244 firms with auditor data used a top-tier auditor at the time of the IPO. 13 There are other underpricing theories, as well. For example, Tinic (1988) and Hughes and Thakor (1992) propose a model in which IPO underpricing mitigates legal liability. Boehmer and Fishe (2001) provide evidence consistent with such model and show that higher underpricing results in higher trading volume of shares and increased trading revenue to the underwriter who also makes the market. Habib and Ljungqvist (2001) argue that reducing underpricing is costly while the benefit of doing so varies with the managers participation in the sale and the degree of dilution they suffer. However, the role of VC valuation in these theories is not clear. 8

11 Behavioral Theories The classical economics theory focuses on utility of the level of wealth. However, the prospect theory of Kahneman and Tversky (1979) stipulates that changes in status relative to particular reference points affect the perceived value. Therefore, instead of utility function, the prospect theory proposes a value function, which represents an individual s preferences over gains and losses relative to a reference point. The prospect theory also proposes an asymmetric treatment of losses and gains, which manifests loss aversion a loss is disliked more than a gain of an equivalent size, i.e., although the value function is increasing in positive wealth changes, it exhibits a different slope for decreases versus increases in wealth. In addition to the kink at zero, the value function is convex in the domain of losses and concave in gains. A related psychological phenomenon, anchoring and adjustment (Tversky and Kahneman, 1974), is also relevant to our study. This phenomenon describes a belief formation process that begins at a specific initial value, which may be salient or entirely irrelevant, and then adjusts toward a final value, based on other considerations and information. The typical bias observed is that the final value is based on a biased adjustment from the potentially arbitrary initial value, hence its term anchor (for example, Baker, Pan, and Wurgler, 2012; Cooney, Moeller, and Stegemoller, 2009; Dougal, Engelberg, Parsons, and Van Wesep., 2011; Moeller, 2010). According to the prospect theory, to the extent that firm insiders anchor on some value they believe fairly represents the value of the company, such irrationality may lead to different negotiation intensity for wealth increase versus decrease and ultimately to inefficient IPO pricing. Research measuring directly the change in wealth of issuing firms managers-owners is difficult due to limited data; however, since most of the managers wealth is linked to their ownership in the company, we consider specific reference points as to which managers anchor as implied by the 9

12 prospect theory. Since many firms in our sample have raised money from VCs, post-money valuations at the last VC financing round is a reasonable proxy for such anchor value, which managers compare to the filing IPO midpoint price or the actual IPO price at later stages in the offering. Consider an example of IPO by LinkedIn which took place on May 19, Prior to its IPO, LinkedIn raised money from VCs and other strategic investors in several rounds, giving it a valuation of about $2 billion based on the last 2010 pre-ipo VC round. LinkedIn s IPO shares were priced at $45 per share at the IPO, implying a $4 billion valuation. Hence, LinkedIn enjoyed a dramatic increase in value since its last round of VC funding in 2010 and, consequently, many of its pre-ipo owners realized large windfalls. Under the prospect theory and anchoring, such windfalls overshadow IPO underpricing and result in high underpricing. In the case of LinkedIn s IPO, its stock price closed at $94.25, more than 109% above its IPO price on the first day of trading. 14 Usually, in the registration statements, a firm going public specifies a range of prices at which it intends to go public. If the midpoint of the initial IPO filing price range implies a company valuation above the value at the VC financing round, there are two effects on the issuer. First, the issuer is more optimistic and eager to proceed with the IPO than would be the case if its IPO midpoint were below the VC valuation. Second, a higher implied valuation at the time of IPO filing causes the issuer to reset its anchor to the new valuation (Figure 2, Panel A). Note that if the issuer is loss-averse, as conjectured by the prospect theory, then such re-anchoring occurs with positive price revision, not with a negative one. Hence, if the new implied valuation based on the filing midpoint is lower than the one at the last VC round, then the issuer remains anchored on VC valuation, i.e., issuers are reluctant to dismiss higher valuation at VC financing round (Figure 2, 14 We note that SDC does not contain sufficient information about LinkedIn s VC valuation preceding the IPO. 10

13 Panel B). This, in effect, is the underlying argument in Baker, Pan, and Wurgler (2012), where a firm s 52-week high price is used as a reference point for an acquisition premium; in effect, the firm insiders anchor on the highest value during the last year. We hypothesize that firms with positive valuation change between the reference point and the actual IPO price should experience high IPO underpricing. This is because issuers being content with positive valuation change do not insist on a higher IPO price. 15 We also propose that firms with zero or negative valuation change between the reference point and the IPO price should experience low or no underpricing at all. Dissatisfied with perceived wealth decrease, the issuers should exert high effort when negotiating the IPO price with their underwriters. Therefore, if behavioral biases affect IPO pricing, we should observe a non-linear effect of valuation change on IPO underpricing: positive relation for valuation increases and no effect for valuation decreases. Figure 2 describes the six possible scenarios. The scenarios are determined by the relative levels of the valuations of the firms at the time of the last VC round, IPO filing, and actual IPO. We first compare the VC round valuation with the valuation implied by the midpoint of the IPO filing range. We note whether the implied valuation at midpoint of the IPO filing range is above the last VC round valuation (groups 1, 2, and 3) or whether it is below the VC valuation (groups 4, 5, and 6). Next, we compare the actual IPO price with the filing midpoint as well as with the VC valuation. Specifically, the six scenarios are determined in Table 1. We note that to the extent that issuers suffer from behavioral biases, those in groups 1, 2, and 3 re-anchor to the IPO filing range midpoint, while those in groups 4, 5, and 6 remain anchored 15 Bradley, Cooney, Jordan, and Singh (2004) hypothesize that the issuing firm and its underwriter are more likely to negotiate from a set of rounded prices when the anticipated offer price is high and/or when there is a large degree of aftermarket price uncertainty. With a lower stock price and/or less uncertainty, negotiations should revolve around a finer set of prices, resulting in more fractional IPOs. They find that the least undervalued issues had fractional prices at both the filing and IPO dates. 11

14 to the original VC valuation. Figure 2 summarizes the predictions regarding the IPO underpricing for the six groups as follows: Group 1. High IPO underpricing. Group 2. Low IPO underpricing. (Note that without the reference point change, we should observe high IPO underpricing.) Group 3. Low IPO underpricing. Group 4. High IPO underpricing. Group 5. Low IPO underpricing. Group 6. Low IPO underpricing. Specifically, the predictions indicate that that IPO underpricing depends on the trajectory that the valuation of firms follows, starting from the last VC round to IPO filing and to IPO itself. These predictions lead to the following hypotheses: H1: Firms with perceived valuation increase from the reference point to the IPO experience higher IPO underpricing. H2: The relation between IPO underpricing and the VC-to-filing midpoint change matters only for firms in groups 4, 5, and 6, i.e. firms with negative VC-to-filing midpoint change Information-Based Theories Information-based theories include models that describe IPO underpricing as: a payment to investors to induce truthful information sharing during book building (Benveniste and Spindt, 1989), a mechanism to overcome fear of winner s curse (Rock, 1986), a way the issuer can signal its type (Francis, Hasan, Lothian, and Sun, 2010; Welch, 1989), and, an outcome of asymmetric information between the issuer and the underwriter (Baron, 1982). 12

15 For example, Benveniste and Spindt (1989) develop a theory, which suggests that there are two sources of uncertainty: industry-related and firm-specific (idiosyncratic). The IPO issuer solicits private information production from investors. If information is unfavorable, then the issuer withdraws its IPO, while positive information leads to upward revision of the initial filing range. However, investors need to be rewarded for their information production effort. Therefore, the theory predicts that high positive price revision leads to high underpricing (higher reward) to reward for the cost of generating new private information. This effect is asymmetric, because in the Benveniste-Spindt model, negative information is free, in the sense that there is no need to offer underpriced shares to induce truthful revelation of negative signals (Benveniste Ljungqvist, Wilhelm, and Yu, 2003). Lowry and Schwert (2002) suggest that underwriters are trying to avoid losses on overpriced issues while allowing informed investors to share the gains on underpriced issues. However, both the Benveniste-Spindt model and Lowry and Schwert link underpricing to new information produced during the book building process. As Ritter and Welch (2002) note [a]ll theories of underpricing based on asymmetric information share the prediction that underpricing is positively related to the degree of asymmetric information. When the asymmetric information uncertainty approaches zero in these models, underpricing disappears entirely. 16 A subset of information-based theories of IPO underpricing includes agency theories. Baron (1982) explains underpricing as a result of an agency conflict between the issuer and underwriters. In Baron (1982), an issuer who is less informed than the underwriter, allows underpricing to compensate underwriters for their efforts. 17 With respect to VC-backed IPOs, 16 Aussenegg, Pichler, and Stomper (2006) confirm that lesser information asymmetry would lead to lower underpricing. 17 We have to take into account that presence of VC investors can also create agency conflicts of their own. Thus, Gompers (1996) grandstanding theory considers the agency issues arising between various VC firms. Under the grandstanding theory, the underpricing of VC-backed firms is caused by younger VCs interested in signaling their 13

16 information about VC investments, valuations, and stakes in the company is often public and should affect the information uncertainty surrounding a firm s IPO, all else held constant. 18 Specifically, the additional information related to the last round of VC financing should reduce the information asymmetry. The more relevant these data are, the less underpricing should be observed, all else equal. Therefore, the recency of a VC financing round and the deviation of the VC valuation to the IPO price are good indicators of how salient and relevant the data are related to the VC funding. 19 Hanley (1993) documents that the relation of the final offer price relative to the range of initial estimated offer prices disclosed in the preliminary prospectus is a good predictor of initial returns. Specifically, IPOs that have final offer prices, which exceed the upper limit of the initial estimated offer range exhibit greater underpricing than all other initial public offerings and are also more likely to increase the number of shares issued. Hanley argues that the IPO price revision is based on information collected during the book building process. Hanley (1993) and later studies (for example, Bradley and Jordan, 2002; Cornelli and Goldreich, 2003) find the offer price revision during the book building process comes up as one of the most important variables explaining IPO underpricing. From these studies it remains unclear how the price revision interacts with owner's prior information. Studying such interaction is difficult because we do not know what this prior information is. However, for VC-backed companies there exists externally validated company valuation from prior financial rounds. According to the information-based theories, this information should be factored into the estimated quality by showing strong track record, while established VCs do not need to signal their quality and therefore, IPOs backed by them will be underpriced less. 18 The role of VC financing in information production before the IPO is discussed by Brav and Gompers (1997) and Bradley and Jordan (2002). 19 Bradley and Jordan (2002) studying partial adjustments in IPOs look at VC backing, but not the VC round valuations, as one of the explanatory variables, and find that VC-backed IPOs are more underpriced. 14

17 offer price at the IPO filing. Once part of the estimated offer price, this information by itself should become irrelevant during the subsequent book building process. Therefore, it should not affect the subsequent IPO price revisions. Thus, we propose the following hypotheses: H3: The timing of the VC funding round relative to the filing of the IPO does not affect IPO underpricing. H4: The impact of information collected during the book building process on IPO underpricing supersedes the effect of the information from VC rounds. Figure 2 compares the analytical framework underlying the tests of the prospect theory with an alternative framework, which accounts for the book building process and its effect on IPO pricing via information production. As argued by Hanley (1993), the underwriter partially revises the IPO price based on information received from investors during the book building process. Therefore, if the information produced during the book building is only partially incorporated into the IPO price, disentangling the effects of behavioral biases and information collection on IPO underpricing is difficult. However, under the behavioral bias theory, we should observe no underpricing for firms in group 5, while under information-based theories, we should observe high underpricing for firms in this group. Hence, in our analysis, we will focus in particular on group 5 which contains firms with negative VC-to-filing midpoint valuation change and positive IPO price revision (from the filing midpoint) with VC-to-IPO valuation change still being negative. 3. Reference Point Variables and Data 3.1. Reference Point Variables To test our hypotheses we construct several variables. First, to reiterate, we calculate the valuation change from the last pre-ipo VC round to the initial filing midpoint ( Filing Valuation Change ). Specifically, Filing Valuation Change variable is calculated based on the VC round 15

18 total (i.e., common and preferred) equity valuation relative to the valuation of the firm s equity based on the initial filing midpoint times the number of shares outstanding prior to the offering. 20 Note that this calculation implicitly makes two assumptions. First, any preferred shares held by VCs are assumed to get converted into common shares at the time of the offering, a common practice in the U.S. Second, the calculation assumes that no additional shares were issued between the last VC round and the IPO filing. To the extent that this assumption is violated, the additional shares issued (for example to employees) would lead us to underestimate the implied share price at the time of the VC round. Since we actually observe a non-trivial number of firms with valuation decreases, the concern about share issuance is mitigated. Furthermore, given that we condition on data availability in SDC VentureXpert, we would expect that any additional VC rounds resulting in issuance of shares would be captured by the database. We observe multiple funding rounds prior to the last VC round for many companies in our sample. Because the prospect theory implies nonlinear relationship between valuation change since the last VC round and IPO underpricing, we calculate the natural log of one plus the percentage difference between last VC valuation and equity valuation based on IPO price. We also calculate the log of one plus the daily implied valuation change. Finally, we use a Filing Valuation Decrease indicator variable, which equals one if Filing Valuation Change variable is negative and zero otherwise Data We use data from SDC's New Issue and VentureXpert databases, Compustat, CRSP, and Jay Ritter and Kenneth French websites. 21 Since very limited data on VC valuations is available 20 For ease of exposition, we also calculate the implied common equity price per share at the last VC round as the post round valuation divided by the number of shares used in the Filing Valuation Change calculation. 21 We obtain data on underwriter ranking through end of our sample from Jay Ritter s website: Returns for industry portfolios based on the Fama-French 49 industries are downloaded from Kenneth French's website: 16

19 prior to 1990, we focus on firms going public in the U.S. between 1990 and 2013 as reported in the SDC New Issue database. We exclude ADRs, spinoffs, limited partnerships, unit offerings, closed-end funds, and observations with offer price less than $5. For all companies in our sample we require that SDC VentureXpert contains information on post-round company valuation at the last VC financing round prior to the IPO filing and information about possible price range at the time of filing for IPO. This requirement dramatically reduces the size of our sample, especially in , when of almost 200 VC-backed IPOs only 4 had last VC round valuation and price range data available. 22 Our sample includes 995 IPOs. Table 2 provides distribution of the sample over time and shows that the period has the largest number of IPOs in our sample. Our sample of firms going public is comparable to IPO samples used in prior research. Our sample distribution is consistent with existing literature, which shows low IPO activity in early 1990 s, a notable increase in IPO activity in late 1990 s, and a dramatic decrease in IPO activity after 2001 (for example, Loughran and Ritter, 2004; Lowry, Officer, and Schwert, 2010). The average IPO underpricing in our sample is 44.0% which reflects the presence of the hot IPO market of The time-series variation of IPO underpricing is consistent with prior literature (Ritter and Welch, 2002) and suggests that our results are not biased by our sample selection criteria. Table 2 also reports the proportion of firms, which experienced decrease in valuation between last VC round and the IPO filing midpoint. For all IPOs, 90, or about 9.0%, of the firms went public at a price below the one implied by the last VC round valuation. With exception of the period, the period contains the most IPO firms with IPO prices below their VC valuations: 16.5%, respectively. 22 Due to the data required for our study, our sample does not include ALL VC-backed IPOs and even if it did, not all IPOs are VC-backed. Hence, our results should be interpreted with caution when extrapolating to a larger set of IPOs. 17

20 In Table 3 we report and compare characteristics used in hypotheses 1 to 4, as well as other characteristics previously shown to be related to IPO underpricing. In the first column we report these characteristics for the entire sample. In columns 2 and 3 we divide the sample based on direction of perceived valuation change at the filing date in comparison with the valuation at the last pre-filing date VC round. If there is anchoring, then VC round valuation should remain the reference point for companies in the subsample reported in column 2, while for subsample reported in column 3 new reference point should become the filing midpoint. In columns 4 and 5 we divide the sample based on the direction of the total trajectory of valuation change from VC round to IPO valuation. The subsample in column 4 includes companies with negative change at the time of the IPO, as compared to the last VC round, while such change is positive for companies in column 5. As we have reported earlier, the average underpricing for the entire sample is 44.0%. In columns 2 and 3, for subsamples with filing price below and above the last VC round valuation, the average underpricing is 29.5% and 45.4%, respectively. The difference between the two is statistically significant at 5% level, while differences in underpricing for subsamples with IPO prices below VC round valuation and above the VC round valuation (columns 4 and 5) are 15.7% and 47.5%, respectively, which are significantly different at 1% level. Finally, the median underpricing for the subsample with IPO price below VC valuation (column 4) is only 2.8%. Hypothesis 3 uses the time lag between VC funding round and the filing of the IPO. We note that for the entire sample this lag is on average days with median being only 214 days. For all four subsamples we observe similar numbers. None of the differences are statistically significant. Johan (2010) uses a measure of Canadian firms preparedness to go public based on the difference between the disclosed intended IPO date and the actual IPO date. For our sample of 18

21 U.S. IPOs, the intended IPO date is not available. Instead, for each time cohort in Table 2 we calculate mean and standard deviation of the number of calendar days between the initial IPO filing and the public issue date and then for each observation in our sample we normalize this variable by subtracting the corresponding mean and dividing by the standard deviation. 23 We expect that the higher this normalized number is, the less prepared the company is for the IPO. We understand that higher number could be explained by other factors, like complexity of IPO and market conditions, which we control for by other factors, such as hot/cold IPO markets indicator, abnormal market and industry returns between filing for IPO and issue date, and price revision between filing and IPO. For all subsamples the preparedness is similar and the values are not significantly different. For hypothesis 4 we use the IPO price revision as a proxy for the information collected during the book building process. The average for the entire sample is 4.35%. The results for subsamples in columns 2 and 3 are 3.24% and 4.46%, respectively. For columns 4 and 5, we note an average IPO price revision of -7.79% for firms going public below VC valuation, and 5.87% for companies going above. The difference is statistically significant at 1% level. This result is not surprising due to the construction of these subsamples. Next, we report and compare characteristics previously shown to be related to IPO underpricing. First, market and industry returns are important factors affecting pricing of firms pursuing initial listing (Loughran and Ritter, 2002). We note that market and industry buy-and hold abnormal returns between the last VC round and filing for IPO and between filing and IPO 23 Johan (2010) uses time between IPO announcement and intended IPO date, between intended data and actual IPO and IPO announcement and actual IPO date as measures of preparedness to go public. Intended IPO date is the feature of Canadian IPO prospectuses which is absent in American prospectuses, and thus the same measures could not be used in our analysis. 19

22 dates are generally higher for samples in columns 3 and 5, i.e., for firms with high valuations relative to VC round valuations. Second, many studies (for example, Habib and Ljungqvist, 2001; Ljungqvist, 2003; Nimalendran, Ritter, and Zhang, 2006) show association of high underwriter s ranking with higher IPO underpricing during 1990 s. 24 We identify underwriter s ranking based on Loughran and Ritter (2004). 25 The underwriter s ranking is lower for firms with lower valuations; however, the difference is not statistically significant. Third, prior literature shows that intensity of IPO issuances is positively related to IPO underpricing. Therefore, we define hot and cold IPO periods based on the monthly volume of IPO issuance as in Helwege and Liang (2004). Specifically, we calculate three-month centered moving averages of the number of IPOs for each month in the sample using data reported in Ibbotson, Sindelar, and Ritter (1994) and updated through March 2014 on Jay Ritter s website. 26 Following Helwege and Liang (2004), hot periods are defined as months for which the number of IPOs is in the top quartile of monthly moving averages. Cold periods are defined as months for which the number of IPOs is in the bottom third of the moving averages. The likelihood of going public in hot market is comparable for subsamples conditioning on valuation levels at filing or IPO dates. In the cold market this likelihood is significantly lower (at 5% level) for companies with higher valuations at filing dates. 24 Underpricing can be a result of conflict of interest for underwriters. Ljungqvist (2003) and Ljungqvist and Wilhelm (2003) study mechanisms to mitigate these conflicts of interest. Ljungqvist (2003) wrote that issuers can monitor the investment banks selling effort or they can use contract design. Ljungqvist (2003) studies the latter. However, he has to use the UK data, because in the U.S. spreads cluster at 7% and there is unobserved heterogeneity of services received. Ljungqvist and Wilhelm (2003) study monitoring of underwriters by owners of companies going public and find that in contrast to the strong effect of investment bank ranking on price revisions, bank reputation does not influence the degree of underpricing, after controlling for other effects Thus, underwriter quality appears to influence initial returns only indirectly by influencing price revisions. 25 Data updated through the end of our sample were obtained from 26 Data updated through the end of our sample were obtained from 20

23 The period of is often referred to as the IPO bubble. Almost 42% of IPOs in our sample took place at that period. Columns 3 and 5 show much higher incidence of IPOs from these years than columns 2 and 4. For the whole sample, we note that the average change between the VC round and filing date valuations is 85.29%. By construction, firms with higher valuations at the filing or IPO dates relative to the valuation at the VC round are characterized by significantly higher perceived valuation increases. Nevertheless, we note that the perceived valuation declines in columns 2 and 4 are economically meaningful. We use median industry Q and net income to assets ratios of industry peers that were public at the time of the IPO to account for industry characteristics. Industry Q is significantly higher for companies with positive valuation change at the time of filing (column 2) than for companies with negative valuation change (column 3). Prior literature shows that offering size is related to IPO underpricing (Aggarwal, Prabhala, and Puri, 2002). The average offering size for the entire sample is $70 million and comparable across subsamples. IPO outcomes could be affected by firm-specific characteristics such as its liquidity needs. Therefore, we included a variable which shows how much cash a company had at the end of fiscal year preceding the IPO as a percentage of IPO proceeds. For the entire sample the average is 61.23% (41.74% median) and is not statistically different between the subsamples. Prior literature shows that characteristics of VCs affect the IPO outcome. These characteristics include VCs experience, reputation, size, possible affiliation with other corporate entities, and heterogeneity of VC funds that invest in the same company together. This heterogeneity can affect IPO outcomes in various ways. For example, older VC funds, which 21

24 approach maturity, might tend to rush unprepared companies to go public. Kandel, Leshchinskii, and Yuklea (2011) conjecture that VC syndicates, which include younger funds, are more likely to have high-quality companies in their portfolio. To capture heterogeneity of VC funds, we compute the difference in age of the youngest and oldest VC fund at the time of IPO. According to Gompers (1996) grandstanding theory, the underpricing of VC-backed firms is caused by younger VCs interested in signaling their quality by showing strong track record while established VCs do not need to signal their quality and, therefore, IPOs backed by them will be underpriced less. Lerner (1994) shows that when experienced venture capitalists invest for the first time in later rounds, the firm is usually doing well. Both Lerner (1994) and Cumming and Dai (2011) use VC size (or its logarithm) as a proxy for reputation. Following prior research, we identify the oldest VC firm, which invested in the company and calculate the natural logarithm of this VC firm s investment in the company (as a measure of involvement) as well as the natural logarithm of total investments in all companies by this VC firm prior to the IPO filing of a particular firm (as a proxy for reputation). For IPOs filed with valuation midpoint below the valuation at the last VC round (column 2), oldest VC firms appear to be significantly more involved (at 5% level). Oldest VC s reputation, as measured by the cumulative investment, is comparable between the two groups of IPOs. Lerner (1994) shows that VC syndication of first round investment may lead to better outcomes. Therefore, we include the number of first round investors as another control variable. The average size of the first-round syndicate for our sample is 2.7 and it is similar for all subsamples. Using an approach inspired by Nahata (2008) we identify the number of instances in which the current lead VC has previously been a lead VC for firms going public. We follow Nahata and classify lead VC as the VC firm, whose fund made the largest investment in the first round. If 22

25 the first round data is unavailable, we use the largest second round VC investor. We observe similar VC reputation based on this measure across all subsamples. Cumming (2006) and Johan (2010) indicate that quality of VC-backed IPOs depends on VC firm s affiliation with financial institutions or large corporations. We include two indicator variables for the oldest VC firm to capture its affiliation with a corporation or financial institution. We find that VC affiliations are similar across the subsamples. 4. Results We first examine the effects of behavioral biases and information production on the IPO pricing. Next, we address possible endogeneity issues by using Heckman s (1979) approach and perform robustness tests by assessing the sensitivity of our results to the IPO period in our sample IPO Underpricing We first examine the roles of information and behavioral biases on pricing of IPOs. Specifically, we test our predictions regarding six possible scenarios in Figure 2. We look at IPO underpricing of each of the six groups in Table 4. The highest IPO underpricing is observed for group 4, comprising firms with negative VC-to-filing midpoint valuation change and IPO price exceeding both the filing midpoint and the VC implied per share valuation. The second highest underpricing is observed for group 1, i.e., firms with positive VCto-filing midpoint valuation change and positive IPO price revision. Both mean and median values for these two groups are statistically indistinguishable from each other. Both of these groups are characterized by the IPO price exceeding both the VC valuation and the initial filing midpoint. The next highest and statistically significant underpricing is observed for group 5 which consists 23

26 of firms with negative VC-to-filing midpoint valuation change and an IPO price that is above the filing midpoint but below the VC implied per share valuation. The underpricing for group 5 is lower than the underpricing for group 4; however, the difference is not statistically significant. A significant difference would be consistent with the predictions of the behavioral bias theories. Group 5 shows significantly higher underpricing than group 6, i.e., firms with negative VC-tofiling midpoint valuation change and negative IPO price revision. This result is consistent with information-based theories of IPO underpricing rather than with the behavioral bias hypothesis. If the midpoint filing range is the only valuation that matters, group 2 should have similarly low underpricing as groups 3 and 6. The results indicate that the underpricing of group 2 is higher than underpricing of both groups 3 and 6, although the differences are not economically large and the statistical significance is mixed. IPO underpricing for both groups 3 and 6 is lower than for other groups. These are the only groups for which IPO price was set below the lowest valuation point at VC round. For group 6, the median underpricing is not significantly different from zero. Overall, we conclude that the univariate results are more consistent with the predictions of the informationbased theories. Nevertheless, the relevance of anchoring is not entirely dismissed, as indicated by results for group 2. The partial adjustment phenomenon is affected by VC valuation for groups 2 and 5. Next, we perform regression analysis of IPO underpricing. All regression models include factors previously shown to affect IPO underpricing and included in Table 3. Heteroskedasticity consistent standard errors are used to assess statistical significance. Table 5 reports the results of regressions analyzing IPO underpricing. In regression (1), we first analyze the effect of IPO price revision on IPO underpricing. We allow for non-linear effect of IPO price revision contingent on whether it is positive or negative. Specifically, we define a 24

27 variable Positive Revision from Filing to IPO which equals the natural logarithm of one plus the IPO price revision when it is larger than zero and zero otherwise. We do the same for IPO price revision less than or equal to zero. The coefficients of both of the IPO price revision variables are positive and significant at the 0.01 level. Consistent with prior empirical literature, the coefficient of the IPO price revision increase is larger than the coefficient of the IPO price revision decrease. The coefficients imply about a 5.2% higher and 3.5% lower IPO underpricing for a firm with initial filing range of $14 to $16 and IPO price of $16 and $14, respectively. This difference of 8.7% in underpricing translates to about $6 million left on the table for the average firm (i.e., the average IPO size of $70 million in Table 3 times 8.7%). Next, we focus on the role of VC valuation in IPO underpricing by including indicator variables for each of the six groups and estimating the model without an intercept. The definition of the groups takes into account the direction and the size of the IPO price revision relative to the direction and the size VC-to-filing midpoint change. Using the group indicators allows us to gauge the relative effects of the partial adjustment and VC valuation on IPO underpricing. To be able to compare the IPO underpricing between various groups, we estimate regressions omitting one group indicator at a time. Thus, in reported regression (3) group 5 indicator is omitted and coefficients of each group indicator capture the difference in underpricing relative to the underpricing of group 5. Results of regression (2) in Table 5 confirm our univariate results. Specifically, the highest underpricing is observed for groups 1 and 4. The coefficient of group 5 indicator implies a positive underpricing of 17.32% (= exp(15.97%)-1), significant at the 0.01 level. Coefficients of group 2, 3, and 6 indicators are all negative, however, insignificant. Turning to regression (3), we note that group 5 has higher IPO underpricing than group 6 by (= exp(21.96%)-1). This difference is 25

28 significant at the 0.01 level. Higher underpricing for group 5 relative to group 6 is inconsistent with the prospect theory's conjecture about re-anchoring occurring only with positive value revision. However, underpricing of group 5 is also significantly lower than the underpricing of groups 1 and 4, indicating that the partial adjustment phenomenon depends in part on pre-ipo VC valuations. In regressions (4) and (5), we include VC characteristics. IPO underpricing of group 5 is not significant (regression (4). Regression (5) indicates that groups 1 and 5 have statistically indistinguishable IPO underpricing. Overall, the regression results point to the relevance of information-based theories of IPO underpricing and are consistent with the partial adjustment phenomenon. However, we observe a further non-linearity in the partial adjustment phenomenon that is related to VC valuation. Hypothesis 3 suggests that the timing of the VC funding round relative to the filing of the IPO does not affect IPO underpricing. We test Hypothesis 3 by including the number of days between the last VC funding round and the IPO date in our regressions. The results indicate that time elapsed since the last VC round does not affect IPO underpricing significantly. This finding is consistent with Hypothesis 3. With respect to control variables, our results in regressions (1) through (3) are comparable with results for IPO samples in prior literature. Firms going public during the period and firms going public after strong industry performance have significantly higher underpricing by about 24%. Strong market performance reduces underpricing. Specifically, the coefficient implies that an increase in market return from VC round to IPO increases by 23.9%, i.e., by one 26

29 standard deviation, the IPO underpricing declines by about 3.9%. 27 Larger offerings and offerings by firms with more cash are associated with higher underpricing. Firms that are more prepared for IPOs (going public faster than the rest of the cohort), have higher underpricing, although the significance of this variable disappears once we add VC-related control variables in regressions (4) and (5). These regressions show higher underpricing in firms for which VC age difference is higher and oldest VCs are larger, while the underpricing is lower if oldest VCs invest more in the company. These results are consistent with limited attention span for larger VC firms as in Cumming and Dai (2011). We examine Hypotheses 2 and 4, the effect of the amount of positive or negative information collected during the book building process on IPO underpricing. We use the IPO price revision as a proxy for the information collected during the book building process. Benveniste and Spindt (1989) predict asymmetric reaction (underpricing) to this new information and Lowry and Schwert (2002) and Benveniste, Ljungqvist, Wilhelm, and Yu (2003) find support to this asymmetric reaction. Since our prior results show underpricing is related to VC round valuation, we analyze whether the asymmetric reaction to the information produced during the book building process is affected by such information. We analyze whether VC round valuation for firms in groups 4, 5, and 6 affects underpricing (Hypothesis 2). In Table 6 we report regressions analyzing the effect of price revision from the filing midpoint to IPO as well as the effect of perceived valuation revision from the VC round to filing price midpoint on IPO underpricing for each of the six groups. Under Hypothesis 2, we should 27 In untabulated results, we explore the sensitivity of Table 5 regressions to alternative specifications of market and industry buy-and-hold abnormal returns for various intervals. We note that only the market buy-and-hold return from last VC round to IPO date shows a significant coefficient in these regressions. With respect to industry buy-and-hold returns, the coefficients measuring returns for shorter periods of time are positive and, generally, significant. These significant coefficients are consistent with the results in regressions in Table 5. 27

30 observe significant and negative effect of VC-to-filing valuation change on IPO underpricing for group 4 and significant and positive effect for groups 5 and If Hypothesis 4 holds, we should observe a significant positive effect of IPO price revision for groups 1, 4, and 5 (i.e., firms with positive IPO price revision from the filing midpoint) and insignificant effect of VC-to-filing valuation change for all groups. The regressions use the same control variables as the regressions in Table 5. Regressions (1) and (3) [(2) and (4)] are estimated with (without) group fixed effects. To conserve space, coefficients of the control variables are not reported. We allow for different slope coefficients by interacting the group indicators with both the IPO price revision and the VC-to-filing valuation change. Consistent with prior literature on partial adjustment phenomenon, we find the largest positive and significant coefficients of IPO price revision for all groups experiencing positive IPO price revision. Specifically, the coefficients for groups 1, 4, and 5 imply that a firm going public at top of its initial IPO range of $14 to $16 experiences higher IPO underpricing by 3.9%, 7.6%, and 3.2%, respectively. For groups 3 and 6, the effect of IPO price revision is closer to zero and has mixed statistical significance at best. Among groups experiencing negative IPO price revisions (i.e., groups 2, 3, and 6) only group 2 shows significantly positive relation between IPO price revision and IPO underpricing. The coefficient implies that a firm going public at bottom of its initial IPO range of $14 to $16 experiences 1.8% lower IPO underpricing. With respect to the relevance of VC valuation, we note a consistently negative (positive) effect of VC-to-filing midpoint valuation change on IPO underpricing for group 2 (group 1). The coefficient implies that 20% higher VC-to-filing midpoint valuation leads to approximately 1% lower (1.6% higher) IPO underpricing for group 2 (group 1). This result is consistent with an explanation that VC valuation continues to serve as an anchor for 28 Note that since the VC-to-filing midpoint valuation change is negative for groups 4, 5, and 6, negative coefficient implies higher underpricing. 28

31 the companies in these particular groups and leads to lower (higher) underpricing. In effect, while firms in group 2 (group 1) re-anchor on the higher filing midpoint, the prior lower VC valuation affects the partial adjustment phenomenon negatively (positively). For groups other than groups 1 and 2, the role of VC valuation is largely insignificant. 29 Overall, the evidence of our regression analysis supports Hypothesis 4 and points to the importance of information produced during the book building process and its role in reducing the effects of behavioral biases as measured by information related to the VC funding round. Therefore, the role of behavioral biases for IPO underpricing appears to be limited at best. However, the partial adjustment phenomenon is affected by prior VC valuation, a surprising and previously unknown result Endogeneity Heckman (1979) shows that statistical analysis based on non-random samples can lead to erroneous conclusions. The endogeneity of a firm s valuation as affected by VC characteristics can affect our conclusions. For example, IPO firms with reputable VCs may by characterized by lower VC round valuation since such VC may be more skilled negotiators. Another potential endogeneity concern may arise if other firms in the same industry are used as valuation benchmarks and the subsequent valuation change just reflects the industry performance. We implement Heckman s (1979) estimation method to address these potential endogeneity concerns. In the first equation, we estimate the probability of a firm experiencing increase in valuation from the last VC round to the filing midpoint and calculate the Heckman lambda, also known as the inverse Mills ratio, for each observation in our sample. Heckman shows that 29 As a sensitivity analysis, we also estimate regressions using groups 1, 2 and 3 as control/omitted group and regressions using either IPO price revision or VC-to-filing valuation change. The unreported results of this analysis do not affect our conclusions. 29

32 including lambda in the second equation controls for the selection bias. In the first stage probit, we include VC characteristics, industry performance, and indicator for IPO bubble period. To satisfy the identification requirement, we exclude the VC characteristics used in the first stage probit from the second stage regression. Models (1) and (2) in Table 7 report the results of the Heckman estimation. The first stage probit regressions indicate that firms going public during IPO bubble period and firms that have VCs with small age differences and the oldest VCs with low level of investment in the firm are more likely to experience valuation increase between the VC round and IPO filing. Most importantly though, in the second stage regressions for both models, the signs and the significance of the coefficients on positive/negative revision and group indicators are generally consistent with our prior results in Table 5. The coefficient on the Heckman lambda is insignificant in model (1) while it is significant in model (2). Therefore, while potential endogeneity issues may affect analysis of IPO underpricing and perceived valuation changes, our conclusions are unchanged when we control for endogeneity The Effect of IPOs in The period from 1998 to 2000 was an exceptionally active IPO market characterized by some of the highest IPO underpricing observed for U.S. firms going public. While we control for this period with an indicator variable in our regression models, we next examine the effect that the presence of this period in our sample has on our results. Specifically, we split the sample into two subsamples based on whether the IPO occurred during the period or not. In Table 8, we report the regression results for models similar to regressions reported in Table 5, but estimated separately for each subsample. Specifically, in Table 8, regressions (1) through (3) analyze the 30

33 subsample of IPOs excluding years IPOs occurring during are analyzed in regressions (4) through (6). Consistent with prior literature, the intercepts of regressions (1) and (4) imply higher mispricing for firms going public during the period. Comparing coefficients of individual groups (regressions (2) and (5)), we note the largest differences in underpricing for groups 1 and 4. IPOs in group 3 are actually characterized by lower IPO underpricing during the IPO bubble period, although the coefficient is not statistically significant. In regression (3), we note that firms in group 5 going public outside of the period have significantly higher underpricing than firms in group 6, a result inconsistent with behavioral explanations of IPO underpricing. However, in regression (6) we note that the difference in underpricing between groups 5 and 6 is not statistically significant, which appears to be due to the fact that underpricing of group 6 is higher during the IPO bubble period. Overall, the results indicate that the active IPO market of does not affect our main conclusions. 5. Conclusion Using a broad sample of VC-backed IPOs, we examine the relation between valuation changes of private firms going public and the effect that such valuation changes have on the IPO underpricing. We find that IPO underpricing is largely driven by information collected during the book building process. Surprisingly, the effects of behavioral biases do not appear to play a significant direct role, except for their contribution to the non-linear partial adjustment phenomenon. We are first to show that this partial adjustment is related to VC valuation preceding the IPO filing date. Our results can be interpreted as an indication that the presence of many sophisticated participants in the book building process creates an environment that limits the impact of 31

34 behavioral biases. While this may not be surprising, the lack of behavioral bias effect stands in stark contrast to prior evidence that includes many sophisticated players (acquisitions, loan pricing, retention of underwriters for follow on offerings, etc.). Our results are consistent with Cornelli and Goldreich (2003) who show that the book building process extracts value-relevant information from investors and incorporates it in pricing of the IPO. 32

35 References Aggarwal, Reena, N.R. Prabhala and M. Puri, Institutional allocation in initial public offerings: Empirical evidence, Journal of Finance 57, Aussenegg, Wolfgang, P. Pichler and A. Stomper, IPO pricing with bookbuilding and a when-issued market, Journal of Financial and Quantitative Analysis 41, Baker, Malcolm, X. Pan and J. Wurgler, The effect of reference point on prices on mergers and acquisitions, Journal of Financial Economics 106, Baron, David P., A model of the demand for investment banking advising and distribution services for new issues, Journal of Finance 37, Benveniste, Lawrence M., and P.A. Spindt, How investment bankers determine the offer price and allocation of new issues, Journal of Financial Economics 24, Benveniste, Lawrence M., and W.Y. Busaba, Bookbuilding vs. fixed price: An analysis of competing strategies for marketing IPO, Journal of Financial and Quantitative Analysis 32, Benveniste, Lawrence M., Al. Ljungqvist, W.J. Wilhelm and X. Yu, Evidence of Information Spillovers in the Production of Investment Banking Services, Journal of Finance 58, Boehmer, Ekkehart, and P.R. Fishe, Equilibrium rationing in initial public offerings of equity, Unpublished working paper, University of Richmond. Bradley, Daniel J., and B.D. Jordan, Partial adjustment to public information and IPO underpricing, Journal of Financial and Quantitative Analysis 37, Bradley, Daniel J., J.W. Cooney, Bradford D. Jordan and A.K. Singh, Negotiation and the IPO offer price: A comparison of integer vs. non-integer IPOs, Journal of Financial and Quantitative Analysis 39, Brav, Alon, and P.A. Gompers, Myth or reality? The long-run underperformance of initial public offerings: evidence from venture and nonventure capital-backed companies, Journal of Finance 52, Cooney, John W., T. Moeller and M. Stegemoller, The underpricing of private targets, Journal of Financial Economics 93, Cornelli, Francesca, and D. Goldreich, Bookbuilding: How informative is the order book? Journal of Finance 58, Cumming, Douglas J., The Determinants of Venture Capital Portfolio Size: Empirical Evidence, Journal of Business 79,

36 Cumming, Douglas J., and N. Dai, Fund size, limited attention and valuation of venture capital backed firms, Journal of Empirical Finance 18, Cumming, Douglas J., and S.A. Johan, Venture capital investment duration, Journal of Small Business Management 48, Derrien, Francois, and K.L. Womack., Auctions vs. bookbuilding and the control of underpricing in hot IPO markets, Review of Financial Studies 16, Dougal, Casey, J. Engelberg, Christopher A. Parsons, and Edward D. Van Wesep, Anchoring and the cost of capital, Unpublished working paper, University of North Carolina. Ellis, Katrina, R. Michaely and M. O'Hara, When the underwriter is the market maker: An examination of trading in the IPO aftermarket, Journal of Finance 55, Francis, Bill B., I. Hasan, J.R. Lothian and X. Sun, The signaling hypothesis revisited: Evidence from foreign IPOs, Journal of Financial and Quantitative Analysis 45, Gompers, Paul A., Grandstanding in the venture capital industry, Journal of Financial Economics 42, Gompers, Paul, and J. Lerner An analysis of compensation in the U.S. venture capital partnership, Journal of Financial Economics 51, Gompers, Paul A., Ownership and control in entrepreneurial firms: An examination of convertible securities in venture capital investments, Unpublished working paper, Harvard Business School. Gompers, Paul A., Corporations and the Financing of Innovation: The Corporate Venturing Experience, Federal Reserve Bank of Atlanta Economic Review 87, Habib, Michel A., and A.P. Ljungqvist, Underpricing and entrepreneurial wealth losses in IPOs: theory and evidence, Review of Financial Studies 14, Hanley, Kathleen Weiss, The underwriting of initial public offerings and the partial adjustment phenomenon, Journal of Financial Economics 34, Heckman, James J., Sample selection bias as a specification error. Econometrica 47, Helwege, Jean, and N. Liang, Initial public offerings in hot and cold markets, Journal of Financial and Quantitative Analysis 39, Hughes, Patricia J., and A.V. Thakor, Litigation risk, intermediation, and the underpricing of initial public offerings, Review of Financial Studies 5, Ibbotson, Roger G., J.L. Sindelar, and J.R. Ritter, (1994). The market's problems with the pricing of initial public offerings, Journal of Applied Corporate Finance 7,

37 Johan, Sophia A., Listing standards as a signal of IPO preparedness and quality, International Review of Law and Economics 30, Kahneman, Daniel and A. Tversky, Prospect theory: An analysis of decision under risk, Econometrica 47, Kandel, Eugene, D. Leshchinskii, and H. Yuklea, VC Funds: Aging Brings Myopia, Journal of Financial and Quantitative Analysis 46, 2011, Kaplan, Steven N. and P. Stromberg Financial contracting theory meets the real world: An empirical analysis of venture capital contracts, Review of Economic Studies 70, Lerner, Josh, The Syndication of Venture Capital Investments, Financial Management 23, Ljungqvist, Alexander, Conflicts of interest and efficient contracting in IPOs, unpublished. Ljungqvist, Alexander and W.J. Wilhelm, IPO Pricing in Dot-com Bubble, Journal of Finance 58, Ljungqvist, Alexander and W.J. Wilhelm, Does prospect theory explain IPO market behavior? Journal of Finance 60, Loughran, Tim and J.R. Ritter, Why don t issuers get upset about leaving money on the table in IPOs? Review of Financial Studies 15, Loughran, Tim and J.R. Ritter, Why has IPO underpricing changed over time? Financial Management 33, Lowry, Michelle, and G.W. Schwert, IPO market cycles: Bubbles or sequential learning?, Journal of Finance 57, Lowry, Michelle, and G. W. Schwert, Is the IPO pricing process efficient?, Journal of Financial Economics 71, 3-26, Lowry, Michelle, M.S. Officer and G.W. Schwert, The variability of IPO initial returns, Journal of Finance 65, Moeller, Thomas, Prior target valuations and acquirer returns: risk or perception? Unpublished working paper, Texas Christian University. Nahata, Rajarishi, Venture Capital reputation and investment performance, Journal of Financial Economics 90, Nimalendran, Mahendrarajah, J.R. Ritter and D. Zhang, Do today's trades affect tomorrow's IPO allocations? Journal of Financial Economics 84, Ritter, Jay R., and I. Welch, A review of IPO activity, pricing, and allocations, Journal of Finance 57,

38 Rock, Kevin, Why new issues are underpriced, Journal of Financial Economics 15, Sahlman, William A., The structure and governance of venture-capital organizations, Journal of Financial Economics 27, Tinic, Seha M., Anatomy of initial public offerings of common stock, Journal of Finance 43, Tversky, Amos, and D. Kahneman, Judgment under uncertainty: Heuristics and biases, Science 185, Welch, Ivo, Seasoned offerings, imitation costs, and the underpricing of initial public offerings, Journal of Finance 44,

39 Figure 1 IPO Underpricing 37