Managerial confidence and initial public offerings

Transcription

1 Managerial confidence and initial public offerings Thomas J. Boulton a, T. Colin Campbell b,* May, 2014 Abstract Initial public offering (IPO) underpricing is positively correlated with managerial confidence. We hypothesize that highly overconfident managers, who tend to overvalue their own firm, use underpricing to signal their beliefs to the market in an effort to receive greater value for their shares in follow-on offerings. Evidence from the subsequent capital raising activities of IPO firms supports this conjecture. However, firms with highly overconfident managers do not consistently outperform firms with less confident managers following their IPO, which suggests that overconfidence is not a proxy for firm quality. Running Head: Managerial confidence and initial public offerings JEL classification: D21; D22; G24; G30 Keywords: Initial public offering (IPO); Managerial confidence; Signaling; Underpricing For valuable comments, the authors express their thanks to Joanna Campbell, Alex Petkevich, Heather Rhodes, Jordan Schoenfeld, Scott Smart, and seminar participants at the 2013 Midwest Finance Association Meetings (Chicago) and Miami University. Research funding was provided by the Lindmor Professorship (Boulton) and the ARMCO Alumni Professorship (Campbell). Any errors or omissions are the responsibility of the authors. a Miami University, Farmer School of Business, Oxford, OH 45056, USA b Miami University, Farmer School of Business, Oxford, OH 45056, USA * Corresponding author at: Miami University, Farmer School of Business, Oxford, OH 45056, USA. Tel.: ; Fax: ; address: campbet7@miamioh.edu (T.C. Campbell).

2 Managerial confidence and initial public offerings Abstract Initial public offering (IPO) underpricing is positively correlated with managerial confidence. We hypothesize that highly overconfident managers, who tend to overvalue their own firm, use underpricing to signal their beliefs to the market in an effort to receive greater value for their shares in follow-on offerings. Evidence from the subsequent capital raising activities of IPO firms supports this conjecture. However, firms with highly overconfident managers do not consistently outperform firms with less confident managers following their IPO, which suggests that overconfidence is not a proxy for firm quality. JEL classification: D21; D22; G24; G30 Keywords: Initial public offering (IPO); Managerial confidence; Signaling; Underpricing

3 Managerial confidence and initial public offerings 1. Introduction IPO underpricing, which refers to firms apparent willingness to sell their IPO shares at a discount relative to the market s equilibrium price, has puzzled researchers for decades. One well-known explanation suggests that information asymmetry between issuing firms and potential investors leads IPO firms to attempt to signal their quality to market participants. For example, Welch (1989) proposes a model in which high-quality firms use underpricing to signal their quality to the market and to differentiate themselves from low-quality firms. High-quality firms benefit from the costly signal by receiving better terms in follow-on offerings. Because firm quality is likely to be revealed in the aftermarket, low-quality firms do not mimic the signal, allowing for a separating equilibrium. 3 The empirical evidence regarding underpricing and signaling is mixed. Consistent with the predictions of Welch (1989), Jegadeesh, Weinstein, and Welch (1993) report a positive relation between underpricing and the probability and size of follow-on offerings. However, Michaely and Shaw (1994) reject signaling when using a simultaneous equation framework modeling both underpricing and the decision to issue equity in the future. More recently, Francis, Hasan, Lothian, and Sun (2010) conclude that signaling plays a role in the underpricing of foreign firms from financially segmented capital markets, and Boulton, Smart, and Zutter (2013) find evidence that focused IPO firms signal in order to differentiate themselves from diversified issuers. We revisit the signaling hypothesis and study the impact of managerial overconfidence on IPO underpricing and firms subsequent capital raising activities. Managerial overconfidence 3 Other prominent IPO signaling models include Allen and Faulhaber (1989), Grinblatt and Hwang (1989), and Chemmanur (1993). 1

4 refers to managers biased beliefs in their own ability, which lead them to overvalue their firm and its investments. We use this terminology similarly to Malemendier and Tate s (2005) overconfidence in means rather than optimism which has been used at times interchangeably in the literature, but may be interpreted as representing the manager s external outlook. Prior research demonstrates a link between managerial overconfidence and corporate policies (e.g., Malmendier and Tate, 2005 and 2008; Campbell, Gallmeyer, Johnson, Rutherford, and Stanley, 2011; Gervais, Heaton, and Odean, 2011). However, little is known about how managerial overconfidence impacts IPO outcomes. For example, do highly overconfident managers use underpricing to signal their beliefs about firm value to potential investors? If so, this should lead to predictable relations between managerial overconfidence, IPO underpricing, and firms seasoned equity offering (SEO) activities. We first show that simple adjustments to incorporate managerial overconfidence into Welch s (1989) model do not violate the conditions needed for a signaling equilibrium under reasonable assumptions. The model predicts that a high-quality firm could underprice to signal quality in order to receive a higher total value for the firm s equity through follow-on offerings. Because highly overconfident managers tend to overvalue their firm, we predict that they are more likely to strategically underprice their IPO to signal their valuation in an attempt to raise capital on more favorable terms in follow-on offerings. If highly overconfident managers use underpricing as a signaling strategy, they should also be more likely to perform a follow-on offering, return to the equity market more quickly and more frequently, and raise a larger percentage of their equity capital through SEOs, relative to less overconfident or unbiased managers. Finally, because highly overconfident managers tend to incorrectly value their firm 2

5 and its prospects, we do not expect them to outperform other managers over the long-run. We test these predictions using a large sample of IPOs from We construct three measures of managerial overconfidence for our tests. The first measure is based on an IPO firm s investments in the two year period ending with the IPO year. This measure is motivated by Malmendier and Tate (2005) and Campbell et al. (2011), which suggest that high overconfidence should lead managers to invest more aggressively, resulting in greater levels of firm investments. We calculate this measure as the match-adjusted ratio of capital expenditures to beginning of year property, plant, and equipment. The matching procedure controls for the impact of the firm s size, age, and industry. Additionally, this measure is a ratio rather than a dollar amount, helping to address potential concerns that firms with larger cash financing requirements are misclassified as having highly overconfident managers. The second measure is based on managerial decisions around the IPO. Specifically, we focus on three characteristics that are likely to be associated with managerial overconfidence: firm age, share overhang, and recent market returns. The expectation is that, compared to less confident managers, highly overconfident managers will take their firms public sooner, retain a greater fraction of the firm s post-ipo shares, and be less influenced by current market conditions when going public. The third measure is based on an analysis of the wording used in the Risk Factors section of the Form S-1 filed closest to the IPO offer date ( registration statement ). Managers are classified as highly overconfident when the text analysis suggests greater tenacity, insistence, assurance, collectives or plurality, and less hesitation or ambivalence specificity, the avoidance of overstatement, or acknowledgement of the speakers limited vision. Importantly, these measures provide three distinct proxies for managerial overconfidence that are unlikely to be subject to the same empirical concerns. 3

6 We find strong evidence that IPOs backed by highly overconfident managers are underpriced significantly more than other IPOs. On average, highly overconfident managers underprice by 33.8 percent. This compares to underpricing of 18.7 percent and 16.9 percent for IPOs with managers exhibiting moderate and low overconfidence, respectively. The relation between managerial overconfidence and underpricing remains strong in multivariate analysis that controls for other factors thought to affect underpricing. Having established a link between managerial overconfidence and IPO underpricing, our tests turn to Welch s (1989) prediction that signaling through underpricing is designed to improve firms subsequent capital raising outcomes. We follow Francis et al. (2010) and explore the probability, speed, size, and frequency of IPO firms follow-on offerings. Taken together, the evidence is consistent with the notion that highly overconfident managers strategically engage in IPO signaling in an effort to raise capital on more favorable terms in follow-on offerings. Specifically, compared to managers with moderate or low levels of overconfidence, we find that highly overconfident managers are: (i) more likely to raise additional capital through an SEO; (ii) issue SEOs more quickly following their IPO; (iii) raise a larger fraction of their equity capital in SEOs; and (iv) access the equity market for additional capital more often. Our results are evident for each of the three measures of managerial overconfidence and robustness tests reasonably rule out alternative explanations, including those based on misclassification of high-quality or highgrowth firms as having managers with high overconfidence. The remainder of the paper proceeds as follows. In Section 2 we expand the Welch (1989) model to consider managerial overconfidence. In Section 3 we describe our empirical approach, sample selection, and descriptive statistics. In Section 4 we present our empirical results and in Section 5 we offer our conclusions. 4

7 2. Background and signaling model Welch s (1989) signaling model shows that when both high- and low-quality firms exist, and there is a positive cost to a low-quality firm imitating a high-quality firm coupled with a chance that imitation is discovered, a signaling equilibrium can exist where high-quality firms underprice to signal their quality. In the model, three possible equilibria exist: (1) the pooling equilibrium where there is no benefit to signaling quality through underpricing and all firms issue stock, (2) the underpricing equilibrium where high-quality firms underprice as a signal of quality, but low-quality firms do not imitate, and (3) the announcement equilibrium where no signal of quality is needed, high-quality firms do not underprice, and low-quality firms do not issue. Importantly, the equilibrium is determined by the likelihood that investors learn of the firm s quality between the IPO and subsequent follow-on offerings (denoted by r in the model). If this probability is sufficiently low (below approximately 0.2 in Welch s calibration), the cost of imitating a high-quality firm is sufficiently low and low-quality firms will imitate, distorting signals of firm quality through IPO underpricing. As a result, a pooling equilibrium results where firms do not intentionally underprice to signal quality. On the other hand, if this probability is sufficiently high (above approximately 0.6 in Welch s calibration), the cost of underpricing outweighs the expected benefits for both high-quality and low-quality firms. In effect, highquality firms do not need to signal when the probability of investors learning the true firm value is sufficiently high. However, when the probability of reveal (r) is moderate (between 0.2 and 0.6 in Welch s calibration), the high-quality firms can increase aggregate proceeds raised from their IPO and subsequent follow-on offerings by using IPO underpricing as a signal of quality. Low-quality 5

8 firms will not imitate this behavior due to the cost of the signal and a sufficiently high probability that a false signal will be discovered before the SEO. In this equilibrium, underpricing represents a credible signal of quality because low-quality firms will not benefit from imitating high-quality firms, allowing for a separating equilibrium. In this equilibrium, high-quality firms signal their quality by intentionally underpricing their IPO and expect to benefit from their signal through higher issue prices in follow-on offerings. This condition is necessary for the signaling equilibrium to hold. Specifically, in the model highquality firms do not deviate from the signaling equilibrium when the total payoff from signaling (the total amount raised in the IPO and SEO) is greater than the payoff from deviating. Welch (1989) shows that this requirement amounts to: ( ) { ( ) ( ( ) ( ) ( ) ) } (1) where α 1 is the proportion of funds raised at IPO, P 1 is the IPO proceeds, V H is the value of a high-quality firm, V L is the value of a low-quality firm, C is the cost of a low-quality firm imitating a high-quality firm, and r is the probability that the true value of the firm is revealed to the market (irrespective of signaling). Welch (1989) shows that the first part of this condition is equivalent to: ( ) ( )[ ( ) ( ) ] (2) which is true by the requirements in the model that (V L C) > 0, V H > 0, and r is a probability between 0 and 1. The second part of this condition is equivalent to: ( ) [ ( ) ( )] ( ) ( ) (3) which is also true because 2V L 3C > 0. Thus, both conditions will be met, allowing for a signaling equilibrium. 6

9 In the Welch (1989) model, the manager correctly values the firm and has complete information about firm quality. The manager decides whether to underprice to signal quality based on the expected costs and benefits of the signal. Because the potential benefits of signaling increase with the value of the high-quality firm (V H ), a result of the signaling equilibrium is that underpricing is expected to increase with V H. While the Welch (1989) model assumes that the manager correctly values his/her own firm, it is straightforward to predict that, because underpricing will increase as V H increases, underpricing will also increase as the manager s estimate (whether correct or overly confident) of V H increases. This prediction is conditional on the assumption that the signaling equilibrium may hold even if managers have biased valuations of their own firms. However, this will depend on the payoff to a high-quality firm from signaling. We start with the assumption that the market understands that some managers are overconfident. However, due to asymmetric information, the market is not certain whether the underpricing signal of firm value is affected by managerial bias. In this case, the market should adjust the price paid for follow-on offerings to account for the possibility that an underpricing signal represents managerial overconfidence rather than high firm quality. Therefore, we expect the market to pay a weighted average of V H and the value of a low-quality firm, V L C, where each value is weighted based on the proportion of highly overconfident managers, which we denote as HO. This would lead the market to pay V H (1 HO) + (V L C)HO at the SEO rather than V H in response to the underpricing signal. For the equilibrium to hold, managers who believe their firm to be high quality must not have reason to deviate from the signaling equilibrium. Thus, we show that the condition outlined in Eq. (1) is still met under reasonable conditions. We update the first part of this condition as follows: 7

10 2( )( ( ) ( ) ) ( )[ ( ) ( ) ] (4) which can be simplified to:. (5) Thus, this part of the equilibrium condition is satisfied whenever the proportion of highly overconfident managers is less than the probability that the true value of the firm is naturally revealed to the market. As noted above, Welch (1989) finds that, under certain assumptions, an r of between about 0.2 and 0.6 would lead to the signaling equilibrium. As we note in the next section, we find that roughly 10% of firms have highly overconfident managers, which is likely to be less than the probability of reveal for which a signaling equilibrium could exist. Thus, if a signaling equilibrium is possible, it is also unlikely that the observed proportion of highly overconfident managers would cause the equilibrium to fail. The second part of the condition for the signaling equilibrium in the Welch (1989) model also changes when the market updates its estimate of firm value based on the proportion of highly overconfident managers as we have proposed. Similar to the first condition above, we update the second condition as follows: ( )[ ( ) ( ) ][ ( ) ( )] ( ) ( ) (6) which must be true by the requirements set forth in Welch s (1989) model for any HO between 0 and 1. Because the proportion of highly overconfident managers must take a value between 0 and 1 by definition, this second condition is also met, meaning that managers who believe their firm to be high quality would not have a reason to deviate, and the signaling equilibrium could exist. Implicit in our additions to the Welch (1989) model are the assumptions that all participants are aware of the proportion of highly overconfident managers, the market cannot determine managerial overconfidence at the IPO, and no managers believe they are overconfident. If 8

11 managers fail to adjust their expectations for SEO payoffs to account for the proportion of overconfident managers, they would continue to (incorrectly) expect total proceeds from IPO and subsequent follow-on offerings to be V H, the payoff from the original Welch (1989) model. In this case, the conditions for the signaling equilibrium in Eqs. (5) and (6) would revert to the conditions in the Welch (1989) model. The second assumption, that the market cannot determine the manager s overconfidence at the IPO, is similar to Welch s (1989) assumption that the market cannot perfectly determine firm value. Asymmetric information prevents the market from fully realizing the firm s true quality, and thus the market cannot determine whether the manager s estimate deviates from the firm s true value. Finally, the third assumption is typically implicit in the models used in the literature on managerial biases (Malmendier and Tate, 2005; Goel and Thakor, 2008; Campbell et al., 2011). If managers realize that they are overconfident, they would lower their estimate of firm value accordingly, correcting their overconfident beliefs. Thus, the possibility that managers may be overconfident requires the assumption that they do not realize their overconfidence level. To the extent that underpricing can be used to signal firm quality, we predict that highly overconfident managers are more likely to believe their firm is high-quality and, therefore, more likely to use underpricing as a signal, all else equal. This leads us to a number of specific predictions. First, underpricing should increase as managerial overconfidence increases. Formally: H1: IPO underpricing will increase with managerial overconfidence If highly overconfident managers underprice to signal their belief about firm quality, with the goal of raising capital on more favorable terms in follow-on offerings, they should take additional actions consistent with IPO signaling. Based on prior works, this means that 9

12 managerial overconfidence should be positively correlated with the likelihood of an SEO, the number of SEOs, and the fraction of total funds raised in SEOs, and negatively correlated with the time elapsed between the IPO and SEO (e.g., Jegadeesh et al., 1993; Francis et al., 2010) Formally stated: H2: The likelihood of raising capital through a follow-on offering will increase with managerial overconfidence H3: The time between IPO and subsequent follow-on offerings will decrease with managerial overconfidence H4: The number of follow-on offerings by the firm will increase with managerial overconfidence H5: The percentage of total equity capital (IPO plus SEO proceeds) raised through SEOs will increase with managerial overconfidence We use the empirical measures of managerial overconfidence described in the following section to test these predictions on a large sample of IPOs issued between 1982 and Empirical approach 3.1. Managerial overconfidence The empirical literature on managerial overconfidence considers several proxies for this bias, which cannot generally be measured directly. For example, Malmendier and Tate (2005, 2008) and Campbell et al. (2011) use proxies for CEO overconfidence based on the CEO s decisions regarding the exercise of stock options granted as a part of compensation and personal trading in the firm s stock. Malmendier and Tate (2008) also measure CEO overconfidence based on the media s portrayal of the CEO. The main shortcoming of these measures is the data required for 10

13 estimation. Malmendier and Tate (2005, 2008) use proprietary and hand collected data, while Campbell et al. (2011) use Execucomp data for CEOs at S&P 1500 companies. Because similar data is not readily available for newly public firms, and because CEO option and stock transaction decisions are likely to be highly endogenous with IPO performance, we consider two alternative measures of overconfidence that are available for a large sample of IPO firms. These measures are described in detail below Managerial overconfidence - firm investment Our first measure of managerial overconfidence is motivated by Campbell et al. (2011), who theoretically link overconfidence to firm investment and use firm investment levels as an empirical proxy for managerial overconfidence. Specifically, the authors classify managers as highly overconfident if they chose very high levels of industry-adjusted investment for their firm (as measured by capital expenditures) for two consecutive years. We adapt their measure for the study of IPO firms. Specifically, we measure overconfidence at the IPO based on a firm s level of investment in the two years ending with the IPO year. Similar to the more general setting studied by Campbell et al. (2011), high managerial overconfidence in our sample firms should lead the managers to invest more aggressively, resulting in greater levels of firm investments. One possible concern is that firms in particular industries or during particular periods of time may be expected to invest more heavily, and thus add noise to our measure of managerial overconfidence. Many IPO firms are young and growing firms that require more investment to optimize firm value compared to older and more established firms. We take a number of steps to address this concern. First, we control for industry and firm size by matching each IPO firm to a non-ipo firm in the same three-digit SIC industry based on market capitalization and IPO issue 11

14 year, using a similar procedure to Ritter (1991). We require that each match candidate is publicly traded for at least 3 years to prevent matching IPO firms with other firms pursuing IPOs during the same period. Additionally, all matching firms are required to have non-missing data for capital expenditures and plant, property, and equipment (PP&E) in Compustat. By matching our sample firms to control firms during the same time period on the basis of size and industry, we hope to control for other factors that could influence a firm s capital expenditure needs. We also scale capital expenditures by beginning of year PP&E, which helps to control for the nature of the firm (i.e., whether the firm has largely tangible or intangible assets; whether the firm is likely to require small or large amounts of capital expenditures). We then calculate the match-adjusted firm investment, measured as the IPO firm s capital expenditures to PP&E ratio minus the matching firm s capital expenditures to PP&E ratio for the IPO year and the year immediately prior to IPO. We classify a firm as having highly overconfident managers if the firm has match-adjusted firm investment above the 80 th percentile in-sample during the IPO year and the year prior to IPO. We classify managers as having low overconfidence if the firm has match-adjusted firm investment below the 20 th percentile insample for the IPO year and the year prior to IPO. The remaining firms are classified as having moderately overconfident managers. 4 One potential concern is that this measure of overconfidence is based on a single firm-level variable, while prior studies have used multiple measures of managerial overconfidence. As an alternative, we consider a second measure that 4 The 20 th and 80 th percentiles break-points for low and high overconfidence, respectively, and the use of two consecutive years of firm investment data to measure managerial confidence are similar to Campbell et al. (2011). Our results are substantively unchanged if we use the firm s match-adjusted investment only during the IPO year. 12

15 incorporates multiple characteristics that may proxy for managerial overconfidence, and offers a valuable robustness check for our first overconfidence measure Managerial overconfidence - IPO characteristics As a second measure of managerial overconfidence, we consider an alternative based on multiple managerial decisions related to the IPO. A common measure of overconfidence previously used in the literature is the manager s personal portfolio investment decisions, specifically the purchases and sales of stock in his or her own firm (Malmendier and Tate, 2005, 2008; Campbell et al., 2011). However, constructing a similar measure for IPO firms is difficult because post-ipo stock sales could be largely endogenous to post-ipo stock performance. In addition, most managers commit to post-ipo lock-up periods during which they agree not to sell stock. A possible alternative is share overhang, which represents the ratio of shares retained to shares sold at the IPO. An overconfident manager should overvalue his/her firm s stock, leading to the prediction in prior works that such a manager will hold too large a personal position in the firm s stock. It is straightforward to argue that such a manager would also seek to retain a larger portion of the firm s post-ipo stock as the manager would expect the stock to be worth more in the future when the market realizes its true value. As this is a less direct measure of overconfidence and may also represent other factors, we include it as one part of a multi-faceted overconfidence measure. The second variable used to calculate the multi-faceted overconfidence measure is the state of the broader market when the firm goes public. Baker and Wurgler (2002) find that many firms time their capital raising activities to coincide with favorable stock market conditions. A rational manager should avoid selling stock during a poor market when the firm would receive a lower 13

16 value for newly issued shares, while an overconfident manager might ignore market conditions and issue in a bad market believing that their firm s prospects and their ability as a manager can offset any shortcomings that the firm will suffer due to market conditions. To capture market conditions, we include the return on the CRSP value-weighted index over the 22 trading days immediately prior to the IPO issue date as the second facet of managerial overconfidence. The third variable in our multi-faceted measure is the age of the IPO firm. Fink, Fink, Grullon, and Weston (2010) link younger IPO firms to higher idiosyncratic risk. Thus, we include the difference between the IPO year and the firm founding year as the third facet of our second managerial overconfidence measure. The intuition for including firm age is that highly overconfident managers will tend to take their firms public sooner than managers with moderate or low overconfidence. We use these three IPO characteristics to form our multi-faceted measure of managerial overconfidence. To construct the measure, we sort all sample firms into deciles based on each of the three measures so that the highest decile represents the most overconfident behavior (greatest percentage of shares retained, lowest recent market performance, youngest firms), and calculate the average decile ranking across the three measures for each IPO firm. We then classify a firm as having highly overconfident managers if the average decile ranking is above the 80 th percentile in-sample, and classify managers as having low overconfidence if the average decile ranking is below the 20 th percentile in-sample. The remaining firms are classified as having moderately overconfident managers. An added benefit of this multi-faceted measure is that the data required to construct the measure is available for a larger sample of IPOs than the investment-based measure. 14

17 A potential shortcoming of this measure is that managers might hesitate to go public during a poor market if they expect their firm to suffer from greater undervaluation, or early in the firm s existence if they are less concerned with diversification. However, it is important to note that overconfidence reflects not simply the manager s valuation of the firm, but their view of their ability to control firm outcomes. Based on this, we argue that a highly overconfident manager, despite the belief that the firm is undervalued by the market, will believe that they can overcome this shortcoming by underpricing to signal firm quality. Additionally, our measure requires that a manager must choose to issue public equity early in the firm s existence, in poor market conditions, and retain a large percentage of equity (share overhang) to be classified as highly overconfident. Thus, highly confident managers not only choose to issue public equity during a poor market and early in the firm s existence, but also retain a large fraction of the firm s equity post-ipo. Recognizing that each of these measures are imperfect proxies for managerial overconfidence, we consider a third measure of confidence that does not rely on characteristics of the firm or IPO Managerial overconfidence text-based measure As an alternative to the two measures above, we consider a third measure of managerial confidence based on a textual analysis of the wording used in the Risk Factors section of the registration statement. 5 If managers are more overconfident, this should be reflected in their word choice. Moreover, this should not represent conscious signaling of firm quality through word choice, as the cost of imitation (using more confident wording) would essentially be zero. 5 Malmendier and Tate (2008) and Hirshleifer, Low, and Teoh (2012) also use textual analysis to measure CEO overconfidence. 15

18 To analyze the text, we employ a standard text-analysis software Diction that measures the word choice in the text on a number of facets. Specifically, we use Diction s measure of certainty as a proxy for managerial confidence. Texts with more words indicative of tenacity, insistence, assurance, collectives or plurality, and fewer words that suggest hesitation or ambivalence specificity, the avoidance of overstatement, or acknowledging the speakers limited vision receive a higher value of this measure of confidence. 6 We use Diction s measure of certainty standardized by the number of different words used in the text to represent the managerial confidence displayed by word choice. We calculate this for all observations in the sample with the full text registration statement available on the SEC s web database (EDGAR). Similar to the other measures presented above, we then calculate the 80 th and 20 th percentiles of this variable within our sample, and classify managers as highly overconfident if their value is above the 80 th percentile in sample. We classify managers as having low overconfidence if their value is below the 20 th percentile in sample. The remaining managers are classified as moderately overconfident Data and descriptive statistics To calculate our measures of overconfidence and conduct our tests, we use data from a number of sources. Thomson Financial s SDC New Issues database is the basis for our IPO sample. From SDC, we extract all firm-commitment new issues by U.S. firms that are listed in the U.S. between January 1982 and December 2010 with an offer price of at least five dollars. As 6 Our results are similar if we use text measures that place greater weight on insistence and tenacity, exclude measures of plurality and/or specificity, or allow the number of words used to reflect the confidence of the manager. 16

19 is customary in the literature, we exclude limited partnerships, closed-end funds, units, real estate investment trusts, and firms in financial and regulated industries. We retrieve aftermarket prices from the CRSP database and eliminate firms that do not have stock price data available within two weeks of the issue date. Additional accounting data is retrieved from Compustat. These data requirements result in a final sample of 6,243 IPOs. In Table 1, we report the number of IPOs by issue year that fall into the high, moderate, and low overconfidence classifications based on the investment-based measure. The final two columns report the percentage of IPOs in a given calendar year that are brought to market by high and low overconfidence managers, respectively. By construction, the majority of our sample managers are classified as moderately overconfident. The largest annual concentration of highly overconfident managers is found at the height of the technology bubble. Specifically, and percent of IPO firms are taken public by highly overconfident managers in 1999 and 2000, respectively. This pattern quickly reverses when, in 2002, percent of IPO firms are brought to market by low overconfidence managers. Across the entire sample, 7.69 percent (7.32 percent) of the IPO-firm managers are classified as highly overconfident (low overconfidence). << TABLE 1 ABOUT HERE >> In Table 2, we report descriptive statistics (mean values) for the high, moderate, and low overconfidence subsamples. In the final two columns we report differences between the high and moderate overconfidence samples and the high and low overconfidence subsamples, respectively. Underpricing is the percentage difference between the IPO offer price and the closing price on the first day of trading. Underpricing for the moderate and low overconfidence subsamples, 18.7 percent and 16.9 percent, respectively, compares favorably to the 18.5 percent 17

20 average underpricing reported by Boulton, Smart, and Zutter (2010). The average high overconfidence IPO, however, experiences underpricing that is substantially greater (33.8 percent). << TABLE 2 ABOUT HERE >> Table 2 reports other significant differences between IPO firms taken public by highly overconfident managers compared to those brought to market by managers with moderate or low overconfidence. Age indicates that highly overconfident managers tend to take their companies public much sooner than other managers. Highly overconfident managers typically wait about 7 years before going public. This compares to nearly 16 years for moderately overconfident managers and 14 years for managers with low overconfidence. Heaton (2002) suggests that overconfident CEOs, who believe the market undervalues their firm s stock, may be reluctant to issue risky securities. Partially consistent with that notion, share overhang indicates that highly overconfident managers tend to retain a greater fraction of the firm post-ipo than do managers with moderate or low overconfidence. Hanley (1993) documents a tendency for IPO firms to revise their offer price from the initial filing range prior to going public. We find that the typical manager with moderate or low overconfidence tends to revise the offer price very little from the midpoint of the initial filing range (0.2 percent and 0.6 percent, respectively). However, IPOs brought to market by highly overconfident managers tend to experience large positive price revisions prior to the issue date (6.6 percent, on average). Loughran and Ritter (2004) report that just over 39 percent of firms receive venture capital backing for their sample of IPOs issued from We find that venture capital backing is more common for our sample, especially among firms taken public by highly overconfident managers, where 61.6 percent of firms receive venture funding. Highly overconfident managers 18

21 are less likely than managers with moderate or low overconfidence to be associated with equity carve-outs, and more likely to work in a high tech industry. Prior evidence suggests that firms go public when overall stock market conditions are favorable (e.g., Baker and Wurgler, 2002). However, recent market conditions appear to be less important to highly overconfident managers. On average, returns on the CRSP value-weighted index over the month leading up to the IPO are 1.0, 1.5, and 1.6 percent for managers with high, moderate, and low overconfidence, respectively. Debt ratios are significantly lower for high overconfidence firms (11.3 percent) compared to moderate and low overconfidence firms (27.2 percent and 28.8 percent, respectively). 4. Results In this section we provide a motive for highly overconfident managers to engage in IPO signaling. Subsequent tests establish a link between managerial overconfidence and underpricing and explore the relation between managerial overconfidence, underpricing, and the subsequent capital raising activities of IPO firms. The weight of the evidence is consistent with the conjecture that highly overconfident managers use underpricing to signal their beliefs to outside investors. Robustness tests reasonably rule out alternative explanations Managerial overconfidence and offer price revisions IPO signaling models are based on the premise that signaling helps to alleviate information asymmetry between issuing firms and potential investors. This suggests that the value of signaling, and consequently the likelihood that an issuer signals, is positively correlated with the level of information asymmetry. One potential proxy for the level of information asymmetry 19

22 between issuing firms and investors is the offer price revision, which reflects information generated during the bookbuilding process. Prior to bookbuilding, issuing firms indicate a range of prices indicative of where they expect to price their IPO. During the bookbuilding process, issuing firms exchange information with potential investors, which assists the firm in setting a final offer price that is often above or below the firm s initial expectations. For example, Hanley (1993) reports that the typical firm revises their final price downward by 4.3% during the period, while Cliff and Denis (2004) report that the average firm revises their final price upward by 3.1% between 1993 and We posit that issues with greater information asymmetry will exhibit larger offer price revisions. In Table 3, we follow Hanley (1993) and report ordinary least squares (OLS) regressions where the dependent variable is the offer price revision, measured as the percent change from the midpoint of the initial filing range to the actual offer price. In addition to our proxies for managerial overconfidence, we control for the width of the preliminary filing range relative to the midpoint of the filing range, the natural log of the inflation-adjusted offer size, the percent change in the Nasdaq index from the IPO filing date to the offer date, the ratio of overallotment shares offered to the total shares offered, underwriter reputation, and controls for issue year and industry. If firms with highly overconfident managers suffer from greater information asymmetry and thus are likely to benefit more from IPO signaling, we expect to observe a positive relation between managerial overconfidence and offer price revisions. << TABLE 3 ABOUT HERE >> The Table 3 results indicate that offer price revisions tend to be percentage points greater for IPO firms with highly overconfident managers. These results suggest that information asymmetry and the amount of information generated during the bookbuilding process are greater 20

23 for IPOs associated with highly overconfident managers. The implication is that these firms stand to benefit more from signaling due to greater information asymmetry between the firm and potential investors Managerial overconfidence and IPO underpricing We argue that highly overconfident managers underprice more when going public to signal their valuation of their firm to the broader market. Our main hypothesis predicts that underpricing will be positively correlated with managerial overconfidence. Table 4 details the results of six regressions that examine the relation between managerial overconfidence and IPO underpricing. Models 1 and 2, 3 and 4, and 5 and 6 report the results using the investment-based, multi-faceted, and text-based overconfidence measures, respectively. << TABLE 4 ABOUT HERE >> In addition to managerial overconfidence, we control for other factors thought to affect underpricing. Prior research finds a positive relation between uncertainty and underpricing. Common proxies for uncertainty include firm size and age (e.g., Ritter, 1984; Beatty and Ritter, 1986). We include the log of the inflation-adjusted offer value to control for size. Our control for age is the log of 1 plus the difference between the IPO issue year and the firm founding year. Studies find that underpricing is negatively correlated with the fraction of the firm that managers sell at the IPO (e.g., Bradley and Jordan, 2002). Share overhang is included to control for this effect. Hanley (1993) finds that underpricing is positively correlated with the offer price revision. To control for this relation, we include the variable offer price revision. The presence of financial intermediaries is also thought to impact underpricing. For example, Barry, Muscarella, Peavy, and Vetsuypens (1990) and Loughran and Ritter (2004) find a relation 21

24 between venture capital backing and underpricing, while Carter and Manaster (1990) and Megginson and Weiss (1991) report that underwriter reputation impacts underpricing. To control for these effects, we include indicator variables set equal to 1 for venture capital backed firms and firms brought to market by top tier investment banks, and zero otherwise. Top tier investment banks are those with Carter and Manaster (1990) rankings equal to 8 or 9. 7 We also include indicator variables for equity carve-outs, high tech firms, and Nasdaq-listed firms. We include the return on the CRSP value-weighted index over the 22 trading days leading up to a firm s IPO to control for hot markets effects. 8 The gross underwriter spread controls for the documented positive relation between underpricing and underwriter spreads (e.g., Cliff and Denis, 2004). Debt ratio is included to control for firms with capital needs and/or financing constraints. All regressions also include industry and issue year indicator variables. In four of the six Table 4 regressions, including all three regressions (1, 3, 5) when comparing high-overconfidence to low-overconfidence managers, we find a positive and significant relation between high managerial overconfidence and underpricing. The coefficient on the high overconfidence measure in the first results column implies that a highly overconfident manager is associated with an additional 5.7 percentage points of underpricing. Given that the mean underpricing across the entire IPO sample is approximately 19.7 percentage points, a highly overconfident manager is associated with an economically significant 29 percent greater underpricing. Furthermore, we find evidence of a significant positive relation between 7 The authors thank Jay Ritter for providing data on IPO firm founding dates and updated Carter and Manaster (1990) rankings. 8 We exclude the controls for age, share overhang, and recent market returns in the regressions that use the multifaceted overconfidence measure. 22

25 managerial overconfidence and underpricing for all three measures of overconfidence, suggesting that this result is not specific to any particular measure. The coefficients on the control variables in Table 4 are generally consistent with expectations. Consistent with the notion that larger offerings pose fewer information asymmetry problems, underpricing is negatively related to the offer size. We also find that underpricing is positively correlated with share overhang, offer price revision, venture capital backing, underwriter reputation, recent market return, and Nasdaq-listing. The negative coefficient on the debt ratio suggests that IPO firms with greater immediate capital needs or debt financing constraints typically underprice less Managerial overconfidence and seasoned equity offerings In the previous section, we document a positive relation between managerial overconfidence and IPO underpricing. In this section, we explore the possibility that this association is motivated by highly overconfident managers desire to signal their beliefs to the broader market to enable their firms to raise subsequent equity capital on more favorable terms. To do so, we explore the relation between managerial overconfidence and the SEO activities of IPO firms. If the observed positive relation between managerial overconfidence and underpricing is the result of signaling, we should also find that the likelihood of an SEO, the speed at which the firm returns to the equity market, and the percentage of the total equity capital raised through SEOs should increase with managerial overconfidence. In Table 5, we report multivariate analyses that examine the relation between managerial overconfidence and the probability, speed, size, and frequency of IPO firms follow-on offerings. The primary variables of interest in these tests are our proxies for managerial overconfidence. Panels A, B, and C report the results for the 23

26 investment-based, multi-faceted, and text-based confidence measures, respectively. Control variables are motivated by Jegadeesh et al. (1993) and Francis et al. (2010). Underpricing is the first day return on the IPO. Aftermarket return 1 (aftermarket return 2) is the market model abnormal return over the window [+1, +20] ([+21, +40]) following the IPO. Assets is the book value of assets. Fraction sold is the number of shares sold at the IPO divided by the total number of shares outstanding after the IPO. All regressions also include unreported controls for industry and IPO issue year. << TABLE 5 ABOUT HERE >> We first test the hypothesis that managerial overconfidence is correlated with the probability that a firm holds a follow-on offering. In the first two columns of Table 5, we report logistic regressions where the dependent variable is set equal to 1 for firms that issue an SEO within 3 years of the IPO and zero otherwise. Because our SEO sample extends through 2010, in these and subsequent SEO tests we exclude IPOs issued after 2007 to allow for a full 3-year post-ipo window. 9 The results are consistent with signaling by highly overconfident managers. Specifically, firms led by high-overconfidence managers are significantly more likely to issue an SEO within 3 years of their IPO than firms with less overconfident managers. In the next two columns of Table 5, we consider the impact of managerial overconfidence on the time that elapses between a firm s IPO and its first follow-on offering. The intuition for this test is that, if high-overconfidence managers underprice because they plan to raise additional capital on more favorable terms in an SEO, they should return to the capital markets sooner than less overconfident managers. We report Tobit regressions where the dependent variable is the natural log of the number of days between the IPO and the first SEO issued within 3 years. For 9 The results reported in prior tables are similar when we exclude IPOs issued after 2007 from the analysis. 24

27 firms that do not issue an SEO within 3 years of their IPO, the dependent variable equals the natural log of 1,095 days (3 years). The results indicate that high-overconfidence managers tend to return to the equity markets sooner than moderate- and low-overconfidence managers. In columns 5 and 6, we consider the possibility that managerial confidence is correlated with the size of firms follow-on offerings. Firms that expect to raise capital on better terms in followon offerings should raise a greater portion of their capital at the more favorable SEO price, rather than the less favorable IPO price. We report OLS regressions where the dependent variable is the natural log of the ratio of the proceeds from a firm s first SEO to the total equity capital raised (IPO plus first SEO). 10 The results indicate a positive relation between managerial overconfidence and the amount of capital raised through follow-on offerings, which is consistent with the conjecture that high-overconfidence managers raise less capital at the IPO, instead choosing to wait to raise capital during the SEO on more favorable terms. In the final two columns of Table 5, we study the relation between managerial overconfidence and the frequency of a firm s follow-on offerings. Although prior studies do not address SEO frequency in a signaling context, there may be reason to expect a positive relation between managerial overconfidence and the number of times a firm returns to the equity markets. If high-overconfidence managers underprice because they plan to raise additional capital at a more favorable price later, they may return to the capital markets more often than less overconfident managers. We report OLS regressions where the dependent variable is equal to the log of 1 plus the number of SEOs that a firm issues within 3 years of the IPO. The results suggest that high-overconfidence managers tend to issue follow-on offerings more frequently than lowoverconfidence managers. 10 The results are qualitatively similar when we use the aggregate proceeds from all SEOs within 3 years of the IPO. 25