Initial public offerings, Underwriting compensation, Underpricing, Regulatory

Transcription

1

2 Underwriter Compensation Structure: Can It Really Bond Underwriters? Jacqueline L. Garner * Mississippi State University Beverly B. Marshall Auburn University Abstract Underwriter compensation can be structured as all cash or a combination of cash and warrants. Using a sample of small IPOs, we find that underwriter compensation contracts that include warrants in exchange for cash can serve as certification for IPO firms by substituting for reputation capital. When underwriters accept warrants when they could have received more cash compensation, the IPOs avoid the well documented long-run underperformance. However, when underwriters receive warrants after maximizing cash compensation, the IPO experiences higher underpricing and poorer long-run performance. The findings are consistent with a motivation by the underwriters to circumvent regulatory constraints. Keywords: guidelines Initial public offerings, Underwriting compensation, Underpricing, Regulatory JEL Classification: G18, G24 * Corresponding author: Department of Finance and Economics, Mississippi State University, 312 McCool Hall, Mississippi State, MS 39762, Phone: (662) ; Fax: (662) ; jacqueline.garner@msstate.edu We thank the editors (Robert and Bonnie Van Ness) and two anonymous referees for their helpful comments and suggestions. We also thank participants at the 2010 Financial Management Association meeting. 2

3 1. Introduction Underwriter compensation in IPOs is regulated by Financial Industry Regulatory Authority (FINRA) Rule 5110, which imposes limits on both the types and the amount of compensation. In general, cash compensation is limited to 13%, comprised of a 10% limit on the spread and a 3% limit on the expense allowance. Non-cash compensation, including warrants, is also subject to regulations, which includes policies on the lock-up period, warrant term, quantity of warrants issued, and the warrant valuation method. Further, Notice to Members provides guidelines on total compensation, which vary with the size of the offering and are suggested to represent the maximum amount of compensation underwriters and related persons may receive in a public offering. For larger offerings, Chen and Ritter (2000) show that total compensation allowed by the regulations is approximately 7% and results in the clustering of spreads for offers over $20 million. 1 However, for offers below $20 million, the regulatory guidelines cap total compensation at levels between 7.52% and 15.80% of proceeds resulting in significant variation in the actual level and types of compensation for small offerings. For example, for offerings below $12 million, the guidelines suggest total compensation at levels over 13% of offer proceeds, which exceeds the limit on cash compensation. Therefore, it is not surprising that warrants are accepted as compensation among small offerings where the dollars generated as a percentage of the offer size could be insufficient to cover expenses and the risk of underwriting the offering. For these underwriters, once cash compensation is maximized, warrants are simply included by default. 1 The compensation guidelines for offers of $50 million or more is 6.89%, whereas the compensation guideline for offers of $20 million is 7.52%. 3

4 However, warrants are not always included in compensation by default. Among our sample of underwriters receiving warrant compensation, cash compensation is not at the maximum allowable level about 40% of the time. This suggests that the underwriters either trade off cash compensation for warrant compensation or are forced by the issuing firm to take contingent compensation. The presence of significant regulations on compensation implies that less reputable underwriters would extract excessive compensation in their absence. This suggests that significant monopsony power over the issuing firm could be present despite what appears to be a relatively crowded market of underwriters. Lesser known underwriters could signal their quality to investors in the compensation contract by setting a low level of cash compensation and accepting contingent compensation in the form of warrants as conjectured by Bae and Ho (2007). On the other hand, the issuing firm could force less proven underwriters to take contingent compensation as a bonding mechanism, in an effort to encourage effective performance of responsibilities surrounding the offering. Either way, evidence of a tradeoff can be useful to investors in determining the quality of the underwriter or issuing firm. Although warrant compensation is observed in a handful of offerings over $20 million, 2 its presence is concentrated primarily in offer sizes of $20 million or lower. While other studies have examined warrant compensation (e.g., Ng and Smith, 1996; Dunbar, 1995), the samples have included both large and small offerings even though warrant compensation is generally only present among smaller offerings. Furthermore, the guidelines suggest significant economies of scale, implying that the linear relation between size and compensation is different for large and small IPOs. Other studies that have examined compensation have specifically excluded small 2 We observed five instances where warrant compensation was included in offerings over $20 million during our sample period. 4

5 offerings for these reasons. Chen and Ritter (2000, p. 1108) exclude offerings below $20 million because the compensation for underwriting smaller offerings is much higher due to the diseconomies of scale, and these deals may be accompanied by underwriter warrants. Therefore, the existing literature s examination of cost differences between offers with and without warrant compensation could be due to 1) size-imposed regulatory limits on compensation, or 2) difficulty in controlling for the relation between size and compensation for such a vast range of offer sizes. We focus this paper on underwriter compensation for smaller IPOs that are usually excluded from IPO compensation studies. While IPOs with proceeds of $20 million or less comprise a small percentage of the IPO market when measured in dollar terms, they represent 30% of all IPOs during our sample period. After examining all active lead underwriters during our sample period, we find that over two-thirds of them are associated with these smaller offerings. Roughly a third of the underwriters in our sample are associated with both all cash and combined compensation IPOs. Given the regulatory limits on compensation, underwriters of smaller offers may be forced into undesired compensation structures. The main contribution of this paper is our examination of the differences in total compensation, total offering costs, and performance among smaller IPOs when the underwriters (i) receive only cash compensation versus (ii) when they receive warrant compensation in lieu of cash versus (iii) when they receive warrants merely by default. 3 We pose three research questions: 1. When underwriters have not reached maximum cash compensation levels, why would they accept warrants over cash? 3 In Ng and Smith (1996), the subsample of firms with warrants have an average offer size of $5 million in proceeds while the non-warrant sample has an average of $39 million (in 1981 dollars). 5

6 2. When warrants are included in compensation, what are the implications for total offer costs (including underpricing) to the IPO firm? Are the implications different in the presence of foregone cash compensation? 3. What are the implications of warrant compensation on future performance? Are the implications different in the presence of foregone cash compensation? This research is important for two reasons. First, we examine the rational for the use of warrants and the effect on total compensation, total offer costs, and subsequent performance in smaller IPOs. While other studies have examined the effect of warrants on total offer costs, the focus has been on larger offerings. Not only are the smaller IPOs rarely studied, the differences in compensation structure, including the use of warrants is most prevalent in smaller offers. In order for researchers to reach conclusions about warrant use, we must examine the IPOs that use them. Second, to our knowledge, no study has examined the effect of trading off cash compensation for warrant compensation and the effect of that tradeoff on compensation, offer costs, and subsequent firm performance. We find that when underwriters do not receive warrants, total compensation and total offer costs are higher, consistent with a cost reduction rational for the use of warrants. Our results show that compensation is lower when underwriters trade off cash for warrants, but the lower compensation is simply offset by higher underpricing. That is, the costs that are known at the time of the IPO are lower when warrants are used. However, we find no evidence that issuing warrants to the underwriter lowers total offer costs. This finding is consistent with FINRA s ability to regulate the compensation portion of the underwriting agreement, but its inability to regulate the setting of the offer price. Consequently, if underwriters are not able to increase their compensation sufficiently, they may reduce their underwriting risk by setting a lower offer price. Therefore, our 6

7 finding that compensation is lower in the presence of warrants is more consistent with regulatory constraints than cost reduction. We find that firms taken public by underwriters that trade off warrants for some portion of their cash compensation do not face the well-documented IPO underperformance (e.g., Ritter, 1991; Loughran and Ritter, 1995)). 4 The underwriter s tradeoff at the time of the offering may provide a certification to investors that the offering is not overpriced and that the firm will not underperform. On the other hand, when underwriters receive warrants when cash compensation is already maximized we find evidence consistent with Barry, Muscarella, and Vetsuypens (1991) of a circumvention motive. Our paper is most closely related to Ng and Smith (1996), which examines a sample of seasoned equity warrants and finds that the warrants function as a bond, substituting for reputation capital. 2. The rationale for warrants We explore three non-mutually exclusive hypotheses for the use of warrant compensation in IPOs: (1) certification, (2) cost reduction, and (3) circumvention of regulations Certification The issuing firm could use a contingent compensation structure to act as certification or bonding mechanism for the underwriter, as suggested by Ng and Smith (1996). 5 Since underwriter 4 The IPO performance issue has been debated. Brav and Gompers (1997) find that venture backed IPOs do not experience underperformance, while their non-venture backed counterparts do. Gompers and Lerner (2003) suggest that the underperformance is simply a size issue, while Ang, Gu, and Hochberg (2007) find evidence of underperformance in spite of the peso (size) effect. 5 Ng and Smith (1996) examine the presence of warrant compensation in SEOs, whereas this study examines their inclusion in compensation for IPOs. 7

8 warrants are not exercisable under FINRA regulations for at least one year after the IPO, 6 this contingent compensation could serve as evidence of due diligence or post-ipo performance in the absence of reputation capital. To the extent that underwriters are willing to accept compensation contingent on future firm performance, those that accept warrants when they could have received more cash could have an extra incentive to perform due diligence, ensuring better post-ipo performance. For warrant compensation to serve as a form of certification, i.e., a substitute for reputation capital or for the warrants to be credible as a bonding mechanism, we expect the following: i. The underwriter s cash compensation will be below FINRA guidelines. Investors, observing that the underwriter could have received more, know that a rational underwriter will only forgo cash compensation for contingent compensation if it expects that the firm faces sufficiently good long-run prospects. ii. Offering is not overpriced. While the issuing firm has an incentive to set a higher price to maximize proceeds, the underwriter has a competing incentive to set a lower price to increase the value of their warrants, reduce their selling effort, and minimize litigation risk. The inclusion of warrants in the compensation contract will balance the competing incentives between issuer and underwriter, certifying to potential investors that the offer is less likely to be overpriced. iii. The underwriter is associated with better performing offers than similarly ranked underwriters that receive cash only. 6 The regulations contain a provision where longer underwriter lock-up periods reduce the compensation percentage ascribed to the warrants. We find no evidence in our sample of underwriter lock-up periods beyond the one-year anniversary. 8

9 We conjecture that the compensation tradeoff substitutes for reputation capital that more established underwriters have obtained through a long history of IPO issuance and performance. In our sample, warrant compensation is prevalent among middle and low tier underwriters and, for the most part, absent among higher tier underwriters. These underwriters could credibly certify their due diligence by trading off cash compensation for contingent compensation in a fairly priced IPO. 2.2 Cost reduction The contract choice by the IPO firm of warrants, in lieu of cash compensation, could be motivated by a desire to maximize cash received from the IPO. Dunbar (1995) finds that total offering costs, measured as underpricing plus total compensation, are significantly lower for offerings where the underwriter is granted warrants. Using a sample of seasoned equity issues, Ng and Smith (1996) find that issuers use warrants when net issue proceeds would have been lower (costs would have been higher) than if warrants had not been used, consistent with Dunbar. For warrant compensation to serve as a means of cost reduction, we expect the following: i. Total offer costs, including underpricing, will be lower relative to comparable cash only offerings. ii. IPO firms will exhibit characteristics that are consistent with a stronger bargaining position than the underwriter. We assume that the IPO firm s goal is to maximize net proceeds of the offering consistent with Yoeman (2001). IPO firms with strong bargaining position in this setting would be willing to issue warrants to the underwriter only if their issuance results in 1) lower total offer costs as measured by total underwriter compensation and underpricing, and 2) better due diligence resulting in better future performance. In order for the underwriter to not 9

10 walk away from less cash compensation, the firm must have better prospects, which will increase the value of the underwriter warrants. 2.3 Circumvention The use of warrants could be a mechanism for the underwriter to extract more compensation from the offering, as suggested by Barry, Muscarella, and Vetsuypens (1991), circumventing regulatory guidelines on total compensation. Since the valuation calculation employed by FINRA does not incorporate volatility, warrants issued by smaller, more risky IPO firms are subject to greater undervaluation as shown in Garner and Marshall (2009). Therefore, underwriters could extract additional compensation while staying under FINRA s radar. When warrants constitute a circumvention motivation, we expect the following: i. Cash compensation is used to the fullest extent and total cash underwriter compensation is pushing the bounds of regulatory constraints. ii. The underwriter will set a higher exercise price to reduce compensation ascribed to warrants by FINRA guidelines (prior to IPO issuance) and/or set a lower offer price such that there is a higher level of underpricing, which increases the value of warrant compensation (after IPO issuance), iii. The underwriter is associated with poorer performing IPOs than similar ranked underwriters that receive cash only or combined compensation with tradeoff. iv. IPO firms exhibit characteristics that are consistent with the underwriter having a stronger bargaining position than the IPO firm. IPO firms faced with high informational asymmetry or a low probability of good future performance could be in a reduced bargaining position when underwriters receive warrants by default rather than through negotiation. 10

11 Table 1 summarizes our expected relation between total compensation, total offer costs, and performance and warrant selection under the alternative hypotheses. [Table 1 here] 3. Sample 3.1 Sample selection We collect a sample of firm commitment IPOs issued between 1993 and 2004 from Thomson Financial SDC Platinum (SDC). We eliminate mutual funds, real estate investment trusts (REITs), American Depository Receipts (ADRs), and IPOs with proceeds over $20 million. While the incidence of warrant compensation in offerings over $20 million is rare, warrant compensation is present in almost half of all offerings below $20 million. By limiting proceeds to $20 million, our warrant and all cash compensation samples are more homogenous in offer size by limiting proceeds to $20 million. In contrast to Chen and Ritter (2000), Hansen (2001), and Yeoman (2001), we include share-only and unit offerings. 7 For IPOs including warrant compensation, we obtain the warrant exercise price, term, and number of shares represented by the warrants from SDC when available. Where the SDC data are incomplete, we obtain the warrant characteristics from the original offering prospectus. Since the presence of warrant compensation is not always indicated in SDC data even when it exists, we verify its exclusion by examining all offering prospectuses. We eliminate those offerings where we cannot verify the structure of warrant compensation through the examination of an actual 7 Chen and Ritter (2000) limit their sample to offers with gross proceeds of at least $20 million dollars, while Hansen (2001) examines offers whose proceeds range from $10 million to $1 billion. 11

12 prospectus. After we exclude offers that are not on CRSP or have missing data, our preliminary sample is comprised of 964 IPOs. We describe our variables in the Appendix. 3.2 Descriptive statistics We separate our sample into two subsamples: (1) All cash and (2) Combined compensation. The All cash subsample includes IPOs where all underwriter compensation is in the form of cash. The Combined compensation subsample includes IPOs where the underwriter receives some portion of compensation in the form of underwriter warrants. In our sample, underwriters of 517 IPOs received all cash compensation with no underwriter warrants. For the remaining 447 IPOs, the warrant value represents approximately 12% of total compensation on average. For most years, the two types of compensation contracts are approximately evenly split. Exceptions include 1996 where 71% of the firms had cash only contracts, and 2001, 2003 and 2004 where over two-thirds of the firms had combined compensation contracts. We provide summary statistics for the two subsamples in Table 2. Table 3 provides a summary of IPO volume and warrant use by year. [Insert Table 2 and Table 3] Our univariate results in Table 2 indicate that All cash IPOs (Panel A) are larger, exhibit higher offer prices and are underwritten by more prestigious underwriters than their Combined compensation counterparts. Surprisingly, eighteen offerings were underwritten by investment banks with the highest ranking (9.1). 8,9 These IPOs are also more likely to have venture backing 8 These investment banks include JP Morgan, Lehman Brothers, Goldman Sachs, Salomon Smith Barney, and Deutsche-Alex Brown. 9 While the average rank of the combined group seems quite low (2.76), the most frequent underwriters in this subsample include Cruttendon-Roth, Meyers, Paulson, and Whale Securities that have ranks ranging from 1.1 to

13 and a high reputation auditor. Total compensation is significantly lower for the All cash group, while long-run performance is higher. Based on offering characteristics, the Combined compensation subsample (Panel B) appears to be more risky at the time of the IPO. IPO firms that include warrant compensation sell a larger portion of the firm at the time of the offering as indicated by the shares ratio. Yeoman (2001) theorizes that larger share ratios will be associated with larger spreads since price uncertainty increases as the portion of the firm being sold increases. Regarding underwriting costs, we find that the Combined compensation firms exhibit significantly higher underwriter compensation. The median firm in the Combined compensation subsample pays 14.25% of IPO proceeds in total compensation, which includes warrant value representing 2% of IPO proceeds. The median warrant value of 2% suggests that the exercise price is 125% of the offer price, and the underwriter receives warrants representing 10% of the offering securities. As shown in Table 2, the Combined compensation subsample exhibits a minimum warrant value of 0% even though the compensation structure includes warrants. This result is because underwriters could receive warrants with a zero value prior to 2001 by setting the exercise price at 165% of the public offer price. In 2000, FINRA closed this loophole by requiring a minimum level of compensation for warrants. For each percentage of public offering securities where the underwriter receives warrants, a minimum compensation level of.2% is assigned. If the underwriter receives warrants equal to 10% of the public offering securities, as allowed by guidelines, this compensation has a minimum value of 2%, regardless of the exercise price. Given this new minimum, underwriters now have no incentive to set an exercise price higher than 125% of the public offer price. 13

14 In addition to underwriter compensation, costs of the offering also include the level of underpricing, which is significantly higher among the Combined compensation subsample. Ex ante, these offerings are also significantly more risky using standard deviation as a measure of stock return volatility. Since both underwriter compensation and underpricing are higher for the Combined compensation subsample, total offer costs, which is the sum of them, are also significantly higher for this subsample. We compare actual compensation to that suggested by the FINRA guidelines across offer size categories in Table 4 for both the All cash and Combined compensation subsamples. The median underwriter receiving all cash compensation is well within the guidelines. Conversely, the median underwriter receiving a combination of cash and warrants is above the guidelines. We find that more than half of the IPOs with underwriter warrants are at the limits of cash compensation, as evidenced by the spread and the expense allowance medians at the maximum levels. The highest levels of warrant value occur when cash compensation levels are maximized, with a 10% spread and a 3% expense allowance. This is consistent with a setting where the median underwriters maximize cash compensation relative to warrant value. Although not evident from this table, some underwriters receive warrant compensation even when total compensation is below the guidelines. [Insert Table 4 here] Table 4 also shows that the exercise to IPO price rises as proceeds increase and cash compensation levels decline. Most warrants represent 10% of the offering securities, carry five year terms (the maximum allowed), and are restricted from exercise or transfer until the first anniversary of the IPO. The FINRA formula makes no provision in the calculation for the warrant term. Therefore, to adjust the level of compensation under the guidelines, underwriters tend to 14

15 raise/lower the exercise price, while leaving the percentage of offering securities at 10% and the term at 5 years. Since the degree of cash compensation forfeited for warrant value should play a critical role in determining the motive for the use of warrants, we subdivide the combination compensation subsample further into 3 groups: Full Tradeoff, Partial Tradeoff and No Tradeoff. The table below displays total compensation and total cash compensation relative to FINRA guidelines and limits, respectively, for each of the three groups. The table also indicates the FINRA warrant value (relative to zero) in each group. Full Tradeoff Partial Tradeoff No Tradeoff Total compensation level relative to Below Above Above guidelines for offer size and type Cash compensation relative to limits Below Below Equal (13%) FINRA warrant value >0 >0 0 Full tradeoff underwriters receive warrants in lieu of cash compensation. This choice could serve as certification of the underwriter s faith in the future performance of the IPO firm. Partial tradeoff underwriters also substitute warrants for some cash compensation, but since their level of compensation is restricted, this choice could be a means to circumvent binding compensation regulations. Since No tradeoff underwriters are already receiving the maximum level of cash, their only means to receive additional compensation is via warrants. We present characteristics of the combined compensation IPO firms in Table 5 and test for significant differences across the subsamples. [Insert Table 5 here] No tradeoff IPOs, where cash compensation is maximized, clearly represent the most risky IPOs based on offering characteristics. No tradeoff firms are smaller, exhibit lower offer prices, have higher underpricing, and are more likely to be unit IPOs. When underwriters choose warrants 15

16 over cash compensation (Full and Partial tradeoff), the IPOs are of higher quality as evidenced by a greater presence of venture capital backing and prestigious auditors. More reputable underwriters, as measured by underwriter rank, are also associated with the Full tradeoff firms. No tradeoff firms appear to have limited bargaining position, inconsistent with a cost reduction motivation. However, the significantly lower market share of the underwriters associated with no tradeoff firms suggests that they also are in an inferior bargaining position. Underwriters that receive warrants by default because they have reached all cash compensation limits (No tradeoff), structure the warrants with the highest exercise prices, suggesting that they are minimizing the value assigned to the warrants by FINRA. Depending on the bargaining position of the IPO firm, underwriters that receive warrants with higher exercise prices could set a lower offer price, which will increase the warrants value through underpricing. Such behavior is consistent with a circumvention motivation. 4. Methods and results Our univariate results in Tables 2 and 5 show significant differences between the All cash and Combined compensation subsamples and among Combined compensation groups on various firm and offer characteristics, including the level of underpricing and compensation. Our univariate results also show that tradeoff firms IPOs have lower underpricing and total compensation but this could simply reflect differences in firm and offer characteristics rather than the efficacy of the compensation contract. Therefore, we wish to examine the simultaneous effect of multiple factors on total compensation and total costs. 16

17 4.1. Effect on total compensation To determine the effect of the warrant compensation decision on underwriter compensation, we model the following OLS regression with Total compensation as the dependent variable: Total compensation = β0 + β1tradeoff + β2size +β3revision + β4ipo volume + β5 Shares ratio + β6secondary shares + β7unit + β8rank + β9venture backing + β10auditor + β11 Age of firm + β12 Warrant compensation (1/0) + βjindustry dummies (1) Due to the clustering of gross spreads by proceeds size all regressions use standard errors clustered on FINRA compensation guidelines. For the Combined compensation groups, we estimate the underwriter warrants using two methods, the FINRA valuation method and the Black-Scholes option-pricing model. To determine the effect of the tradeoff on underwriter compensation, we include the variable Tradeoff in Equation (1) for those IPOs with warrants. Tradeoff takes the value of one if the IPO is in the Full Tradeoff subsample. Warrant compensation takes the value of one if the firm uses warrants. All other variables in equation (1) are previously defined. 4.2 Effect on total offer costs Underwriter compensation is not the only cost to the IPO firm. Therefore, we include underpricing with compensation costs to examine the total offer costs to the IPO firm. We estimate the following regression with Total offer costs as the dependent variable for the All cash and Combined compensation groups: Total offer costs = β0 + β1size + β2revision+ β3ave. underpricing + β4shares ratio + β5warrant to shares + β6rank +β7venture backing + β8auditor + β9age + β10 IPO volume + β11 Warrant compensation (1/0)+ βj Industry dummies (2) We substitute Warrants to shares for Unit in Equation (2) since it better captures the compound nature of underpricing in the presence of multiple warrants within the unit. To 17

18 determine the effect of the compensation tradeoff on total costs, we modify Equation (2) for the Combined compensation group by including two additional variables: Tradeoff and Exercise price/offer price. Underpricing is windsorized at the 1% and 99% levels. We recognize that we could have a potential self-selection problem in Equations (1) and (2). For example, as Dunbar (1995) notes, firms could choose a specific compensation contract when costs are low and avoid the other contract when costs are high. Failure to account for endogenous self-selection creates an omitted variables bias, which results in biased and inconsistent parameter estimates. Therefore, to control for the endogenous choice of warrant use, we use a self-selection model to allow for differences in the use of warrants on compensation. 10 We discuss this in the next section. 4.3 Modeling the selection of warrant compensation We expect the decision to include warrant compensation is a function of various firm and offer characteristics and IPO market conditions. We estimate a probit model where the dependent variable, Warrant compensation takes the value of one if the compensation structure includes warrants. Warrant compensation (1/0) = β0 + β1size + β2revision+ β3ave. underpricing + β4 IPO volume + β5shares ratio + β6secondary shares + β7dual + β8unit + β9rank + β10venture backing + β11auditor + β12firm age + βj Industry dummies (3) Critical to using this approach is our identification of an exclusion variable, which will identify our self-selection model. The exclusion variable must be (1) relevant, namely correlated with 10 Li and Prabhala (2007) outline the increased use of these models in finance. Ang and Nagel (2012) use this procedure to control for the endogenous choice of inside versus outside CEO hires, and Campa and Kedia (2002) use this procedure to control for the endogenous choice to diversify. 18

19 warrant use; and (2) valid, not correlated with our outcomes (compensation and total costs). That is, the only role that our exclusion variable will play in influencing the outcome (total compensation and total offer costs) is its effect on the use of warrants. We use a dummy variable, Dual, as an exclusion variable that predicts warrant use. Our use of Dual as an exclusion variable derives from several findings. First, we do not find significant differences in the percent of IPOs with dual class shares between the Cash and Combined compensation groups in Table 2. Second, following Booker, Sass, Gill, and Zimmer (2011) and Ang and Nagel (2012), when we include the exclusion variable, Dual in our outcome equations (Equations (1) and (2)) along with all other independent variables, it is insignificant. We use Equation (3) to compute the Inverse Mills ratio and use it in lieu of the dummy variable, Warrant compensation, in Equations (1) and (2). Following Heckman (1979), we assume that unobservable variables influencing the choice of warrant use can be accounted for by the Inverse Mills ratio. The Inverse Mills ratio displays negative values for the non-selection of warrant compensation and positive values for warrant selection. 4.4 Effect on compensation Our results of the estimation of Equation (1), the determinants of compensation, are reported in Table 6. In Columns 1a and 1b, we show results for the full sample. In the other columns we display results for the All cash subsample (Column 2) and the Combined compensation subsample, using both the FINRA valuation of warrants (Column 3a) and the Black- Scholes valuation (Column 3b). In Column 4, we report the results for the 231 IPOs where the underwriter accepted warrants and was associated with both types of compensation contracts. [Insert Table 6 here] 19

20 For the full sample, our results in Columns 1a and 1b indicate that the use of warrant compensation increases total compensation costs, consistent with a circumvention rationale for the use of warrants. However, remember that we have not controlled for self-selection bias in these two specifications. The inclusion of Tradeoff in Column 1b indicates that the presence of a tradeoff reduces compensation costs, consistent with a cost reduction rationale. For the full sample, we also find that compensation costs are inversely related to offer size as indicated by the positive coefficient on the inverse of proceeds. Greater offer price revisions, higher share ratios, and unit offerings are associated with higher compensation costs, consistent with being more difficult to value. The presence of higher reputation underwriters and more prestigious auditors are both associated with lower compensation costs, consistent with their certification role. Lower compensation costs are also associated with older firms, which suggest that an operating track record makes the underwriting process easier for the underwriter. For the All cash subsample (column 2), we find that the non-selection of warrants results in higher compensation costs as indicated by the negative coefficient on the Inverse Mills ratio since the underlying variable has negative values for this subsample. This result is consistent with the cost reduction explanation for the use of warrants. 11 We also find that more established firms with reputable underwriters and venture backing have lower compensation costs, consistent with a risk reduction story. The insignificance of offer size is consistent with the clustering of spreads 11 Results from estimation of Equation (3) suggest that Unit offerings are more likely to include warrant compensation. Firms selling a larger proportion of shares and those with multiple share classes are more likely to include warrants. Warrant compensation is decreasing in proceeds and less likely to be included in the presence of reputable underwriters, in IPOs with secondary shares, and in periods of high IPO volume. Results are available on request. 20

21 around 7%, shown by Chen and Ritter (2000) and evident even among smaller offerings as seen in Table 4. For the Combined compensation subsample, we find that the exchange of warrant compensation for cash compensation (Tradeoff) significantly reduces total compensation costs. This finding is invariant to the method of option valuation. Since warrants received by the underwriter in unit offerings are typically exotic options, Garner and Marshall (2009) show that their undervaluation under the FINRA formula is more severe. While the warrant valuation method does not alter the effect of Tradeoff on total compensation when the Black-Scholes valuation is used, we find that total compensation is higher for unit offerings, and this finding appears to be directly attributable to the valuation difference between the FINRA and Black- Scholes models. The negative coefficient on the Inverse Mills ratio, which has positive values for this subsample, implies that the selection of warrants leads to lower total compensation. Coupled with the negative coefficient on Tradeoff, there is a significant compensation savings for the IPO firm when the underwriter replaces cash compensation with warrants. The results are consistent with a cost reduction rationale for warrant use. For the combined compensation sample, we also see that compensation costs are inversely related to offer size and increasing with offer price revision. We also find that rank is not associated with compensation, perhaps due to the fact that two-thirds of the underwriters in this group are ranked at 3.1 or less, and none are ranked at 8.1 or higher. The lack of variation in rank suggests a need for underwriter revelation of type, perhaps via a tradeoff. As a robustness check we limit our sample to the subsample of 413 IPOs where the underwriters are associated with both compensation structures. We compute the Inverse Mills ratio by estimating Equation (3) just for this subsample, and re-estimate Equation (1). In Column 4, we 21

22 report the results for the 231 IPOs where the underwriter accepted warrants. For this subgroup, total compensation is higher for unit offerings and lower in the presence of a tradeoff. The selection of warrants suggests a reduction in total compensation, consistent with a cost reduction rationale. In summary, the results of Table 6 suggest that the selection of warrants reduces underwriter compensation. However, underwriter compensation is only part of the story. While FINRA regulates compensation, it does not regulate pricing of the IPO (Sjostrom, 2010). The issuing firm is interested in total offer costs, which include underpricing, according to the net proceeds maximization view (Yeoman, 2001). We now turn our attention to total offer costs, which captures underwriter compensation and underpricing Effect on total offer costs Our results from the estimation of Equation (2), on the determinants of Total offer costs, are presented in Table 7. [Insert Table 7] For the full sample, we find that the inclusion of warrants in the compensation structure reduces total offer costs; however, this result is before we control for self-selection. In Column 1b, we include two additional independent variables, Tradeoff and Relative Exercise price/offer price. Tradeoff is as defined before. Relative Exercise price/offer price is the ratio of the exercise price divided by the offer price less 1.2, which is the mode. If the warrants have the normal exercise price of 120% of the offer price, the variable takes the value of zero. We find some evidence of higher total offer costs in Column 1b when the underwriter sets an exercise price above the standard 120% of the IPO offer price. Since warrants are more valuable when underpricing increases, we take this finding as evidence that underwriters that receive warrants with higher 22

23 exercise prices increase their value through underpricing. Our results present evidence consistent with an agency conflict. Our finding of significant underpricing on warrants with relatively high exercise prices suggests that underwriters set a lower offer price and increase the value of warrants through underpricing. The result demonstrates that warrants with high exercise prices were used as a way to circumvent regulatory constraints on compensation levels, at least prior to the closing of the loophole in We also find in Columns 1a and 1b that total offer costs are decreasing in proceed size as shown by the positive coefficient on the inverse of proceeds. We find a positive and significant coefficient on Revision, indicating that higher underpricing occurs when the offer price is adjusted upwards, consistent with Hanley (1993) and suggestive of a demand story. Unit IPOs are associated with greater total offer costs, and this result holds when we replace the indicator variable, Unit, with the Warrant to shares ratio. This measure better reflects the speculative nature of the unit and allows us to capture the compound nature of underpricing in the presence of multiple warrants. For the full sample we also show lower costs in the presence of more reputable underwriters, venture backing, and the use of a prestigious auditor consistent with their certification role. For the All cash subsample, the negative coefficient on the Inverse Mills ratio indicates that the non-selection of warrants leads to higher total offer costs since the values of the underlying variable are negative for this subsample. The result is consistent with a cost reduction justification for the inclusion of warrants in the compensation structure. We also find that total offer costs increase with proceed size, as evidenced by the negative relation with 1/Proceeds for this subsample, which is opposite the finding for the full sample. As shown in Table 6, proceed size and compensation are not related for the All Cash subsample; therefore, the relation between Total 23

24 offer costs and offer size for this subsample shown in Table 7 is due entirely to higher underpricing among larger IPOs. Suggestive of a demand story are the positive and significant coefficients on Revision and Average underpricing. While the presence of a reputable auditor is associated with lower total offer costs for the All cash subsample, consistent with Beatty (1989), the presence of venture backing is not significant. The positive relation with Rank is consistent with the realignment of incentives hypothesis introduced by Ljungqvist and Wilhelm (2003) during the bubble period and results reported by Loughran and Ritter (2004). Also suggestive of higher underpricing levels during the bubble period is the positive and significant relation between IPO volume and Total offer costs. Older firms in the All cash subsample exhibit lower total cost consistent with the lower compensation levels already shown in Table 6. Among Combined compensation IPOs, we find no evidence that the inclusion of warrant compensation is motivated by cost reduction since the coefficient on the Inverse Mills ratio is insignificant. We find that the exchange of warrant value for cash compensation (Tradeoff) does not significantly affect total offer costs even after controlling for self-selection. The result is evidence that lower compensation among tradeoff firms is simply substituted with a lower offer price. On the other hand, total offer costs are increasing in the ratio of the underwriter warrant exercise price to the offer price as we found with the full sample before controlling for selfselection, consistent with regulatory circumvention. For the Combined compensation IPOs, we see that smaller offerings exhibit higher total offer costs, which can reflect their more speculative nature and the need by the underwriter to make sure there is adequate demand for the IPO. We find that offerings associated with multiple warrants incur greater total offer costs, consistent with their risky nature and the compounding of underpricing in the multiple option structure. Unlike the All cash subsample, the Combined 24

25 Compensation subsample shows no relation between Total offer costs and either Revision or IPO volume. Although underwriter compensation is positively related to these variables (Table 6), they are no longer significant when Underpricing is included. Also, unlike the All cash subsample, Rank, Venture backing, and Auditor all have the negative and significant relation suggested by a certification role reducing asymmetrical information. We have already shown that Combined compensation IPOs are more risky and could benefit the most from the monitoring role of these institutions (Chemmanur and Fulghieri, 1994; Jain and Kini, 1999; Fang, 2005; Beatty, 1989). We re-estimate our total offer costs results for the subsample of underwriters that varied their compensation structure and chose to receive warrants in Column 4. For this subgroup, total offer costs are also not significantly different in the presence of a tradeoff. Among underwriters that vary their compensation structure, neither the presence of a tradeoff nor the selection of warrants has a significant effect on total offer costs since any compensation cost savings in Table 7 appear to be offset by higher underpricing. While we find similar results for this subsample regarding the relation with offer size and Warrant to shares ratio, we find no evidence of the certification role of the underwriter, venture backing, or prestigious auditor for the more limited set of underwriters. 4.6 Effect on performance Test of underwriter s compensation ex ante If more reputable underwriters consciously trade off cash compensation, we expect them to do so when that decision has a high internal rate of return (IRR). To test the efficacy of the underwriter s choice, we value the warrants ex ante and compare this value to the foregone cash compensation. We value the warrants for share-only IPOs using the Black-Scholes (1972) optionpricing formula and warrants issued in unit IPOs using exotic option valuation procedures as outlined in Garner and Marshall (2009). We value the warrants at the IPO date and on the one- 25

26 year anniversary of the IPO since they are transferrable as of that date. Since the underwriter is not required to hold the warrants beyond this date, we do not assume that they are held to expiration. We compute the IRRs of the tradeoff decision using the cash foregone as the initial investment and the value of the warrants one year after the IPO as the year one cash flow. The mean and median IRRs as of the IPO date of the Full tradeoff choice are 654% and 189%, respectively. 93% of the IRRs are positive. The mean and median IRRs for the partial tradeoff group are 257% and 130%, respectively, and 84% of the IRRs are positive. While we do not know each underwriter s cost of capital, it is unlikely to be higher than the mean/median values we report here. Under the assumption that due diligence would increase the likelihood of a high IRR choice, the finding that the average warrants have a high IRR suggests that underwriters, through the compensation structure, are incentivized to perform the necessary due diligence Tests of post-ipo stock performance We wish to determine if the choice of warrant compensation in lieu of cash leads to better performance. Better performance would be consistent with certification, and inconsistent with circumvention explanations. Therefore, we examine the relation between BHAR and the choice of the compensation contract by the underwriter. We also control for the rank of the underwriter since the existing literature suggests that the presence of a reputable underwriter (Jain and Kini, 1999; Carter, Dark and Singh, 1998) leads to better performance. We hypothesize that the composition of underwriter compensation is another firm characteristic that shows firm quality. We estimate the following regression: BHAR = β0 + β1full tradeoff + β2some tradeoff+ β3no tradeoff + β4 Rank (4) BHAR, Year 1 BHAR and Year 2 BHAR are net buy-and-hold returns (Lyon, Barber, and Tsai, 1999) and are defined in the Appendix. Full tradeoff takes the value of one if the underwriter 26

27 fully traded off warrants for cash compensation. Partial tradeoff takes the value of one if total compensation is above the FINRA guidelines, but the gross spread and non-accountable expense allowance are below their maximums and warrant valuation under FINRA is greater than zero. No tradeoff takes the value of one if the underwriter maximized cash compensation and received warrants. We present the results of the estimation of Equation (4) using Year 1 BHAR in Table 8. We segregate the results into underwriters that were associated with both types of contracts (All cash and combined) as well as for the full sample of firms. Using equally weighted returns, we find evidence of one year underperformance when underwriters do not tradeoff, regardless of the contract type. We also see that full tradeoff IPOs do not show evidence of either under- or overperformance. For the full sample, we show that ranked underwriters are associated with better performing IPOs. In unreported results, we see a similar performance pattern with the Year 2 BHAR. [Insert Table 8 here] We find that the absence of a tradeoff by underwriters of cash compensation for warrant value is associated with lower IPO firm performance, and the presence of a tradeoff is associated with no underperformance. The result suggests that, to some degree, in the presence of a tradeoff, warrants can bond underwriter behavior and provide a certification role for the IPO. 5. Conclusion and summary We reconcile competing justifications for the use of warrant compensation. Barry, Muscarella, and Vetsuypens (1991) suggests that warrants are a means for underwriters to circumvent regulatory guidelines on compensation, while Dunbar (1995) finds that the use of 27

28 warrants actually lowers total offer costs. Our results show that when underwriters receive only cash as compensation, total offer costs are higher. We also find that when underwriters tradeoff cash for warrants, compensation is lower, but the lower compensation is simply offset by higher underpricing. Consistent with Barry, Muscarella, and Vetsuypens, we find some evidence consistent with a circumvention motive, particularly in the case of underwriters that take warrants with a high exercise price. At the other end of the spectrum, the inclusion of warrants reduces the risk of an overpriced offer. We find that the degree of tradeoff is a signal of long-run performance. When underwriters fully trade off cash compensation for warrant value, the IPO firm neither under nor over performs the market. Underwriters that charge the maximum levels of cash compensation and also do not trade off are associated with significantly underperforming IPO firms. The result is further evidence that the compensation structure shows information about firm quality. Our work can be viewed as evidence that underwriter compensation contracts mitigate the information asymmetry surrounding small IPO firms and lower reputation underwriters. Our results also suggest that efforts by FINRA to modify the valuation of warrants in 2001, to provide a minimum valuation regardless of the exercise price, serve to reduce the potential agency conflict that contributed to higher underpricing and therefore higher total offer costs. 28