Sales of private firms and the role of CEO compensation

Journal of Corporate Finance 43 (2017) 444–463

Contents lists available at ScienceDirect

Journal of Corporate Finance journal homepage: www.elsevier.com/locate/jcorpfin

Sales of private firms and the role of CEO compensation Natasha Burns a,⁎, Jan Jindra b,1, Kristina Minnick c a b c

The University of Texas at San Antonio, United States Securities and Exchange Commission, United States Department of Finance, Bentley University, United States

a r t i c l e

i n f o

Article history: Received 1 June 2016 Received in revised form 30 January 2017 Accepted 4 February 2017 Available online 7 February 2017 JEL classification: G3 G34

a b s t r a c t We analyze the relation of private firms' CEO compensation with the probability of sale of a firm and its valuation at the time of the sale. Specifically, we study whether equity-based remuneration is consistent with compensating the CEO for effort related to selling the private firm, or with compensating for the illiquidity of the equity-based compensation for private firms. Using a sample of large private firms with public filings, we find that CEOs of IPO and acquired private firms have higher total and equity-based compensation than CEOs of firms that remain private. We also show that CEO compensation is positively related to the valuation premium of IPOs versus acquired firms. © 2017 Elsevier B.V. All rights reserved.

Keywords: Private firms Mergers and acquisitions Initial public offerings Exit CEO compensation IPO versus acquisition valuation IPO valuation premium puzzle

1. Introduction Recent literature provides evidence on how certain firm-specific, industry-wide, and market-wide factors are correlated with a private firm's choice between going public and getting acquired.2 However, the question of how a private firm's CEO compensation relates to the decision to sell the firm is largely unexplored. In this paper, we analyze the relationship between CEO compensation and the probability of a private firm in a particular year: (i) remaining private, (ii) selling via an initial public offering (IPO), or (iii) selling to an acquirer. We also analyze whether the valuation of selling firms is related to CEO compensation, an analysis related to the “IPO valuation premium puzzle” whereby private firms are acquired instead of realizing a higher valuation via IPO (Poulsen and Stegemoller, 2008; Bayar and Chemmanur, 2011, 2012). Ex ante, it is not obvious whether the relation between CEO compensation and the probability of a sale of a private firm is negative or positive. CEOs of firms remaining private are possibly paid more due to the illiquidity of the firm's shares (or because of

⁎ Corresponding author. E-mail addresses: [email protected] (N. Burns), [email protected] (J. Jindra), [email protected] (K. Minnick). 1 The Securities and Exchange Commission, as a matter of policy, disclaims responsibility for any private publication or statement by any of its employees. The views expressed herein are those of the author and do not necessarily reflect the views of the Commission or of the author's colleagues upon the staff of the Commission. 2 Prior literature analyzes the role of market and industry characteristics (Brau et al., 2003), firm-specific operating characteristics (Poulsen and Stegemoller, 2008), and, in a theoretical framework, industry as well as firm-specific operating characteristics (Bayar and Chemmanur, 2012) as the determinants of the decision of private firms to pursue an acquisition versus an IPO.

http://dx.doi.org/10.1016/j.jcorpfin.2017.02.004 0929-1199/© 2017 Elsevier B.V. All rights reserved.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

445

CEO private benefits of control; see Helwege and Packer, 2009). Conversely, CEOs of firms being sold may be paid more because higher CEO compensation reflects payment for additional effort associated with selling the firm. In our research, we develop and analyze hypotheses linking CEO compensation with (i) the probability of a sale of a private firm and (ii) the valuation at the time of the sale. Using a sample of 4727 private firm-years spanning 2003 to 2011 from Capital IQ, supplemented with hand-collected data on CEO ownership and CEO characteristics, we find that private firms that get acquired or pursue an IPO pay their CEOs more in terms of total, bonus, and equity-based compensation when compared to firms that remain private. These results are consistent with the view that the compensation of CEOs at acquired private firms reflects compensation structured to reward the CEOs for the effort associated with selling the firm, as well as serving to compensate the CEO for potentially adverse career outcomes, such as job loss following an acquisition. Consistent with the expectation that shares will be more liquid after an IPO, option and restricted stock compensation play important roles for firms that sell via an IPO but are generally insignificantly related with the likelihood of selling via an acquisition. Bonus compensation plays a more important role for acquisition firms or IPO firms. Given the potential self-selection of private firms involved in selling, it is important to address endogeneity. We attempt to address endogeneity in our study using an instrumental variable approach and propensity score matching. With respect to instrumental variables, we argue that CEO education level, age, tenure, and the education of the local workforce (measured as the fraction of population of the headquarter state with graduate degrees) are appropriate instruments that affect CEO compensation but do not directly affect the probability of sale of the firm. The results of the instrumental variable approach are consistent with the motivation framework of CEO compensation. With respect to propensity score matching, after matching the selling firms with firms remaining private, the results continue to be consistent with the motivation framework of CEO compensation. Hence, after using instrumental variable and propensity matching approaches to address endogeneity, the results continue to corroborate our prior findings. We also assess the impact of alternative modeling techniques on our conclusions using hazard and competing risk models. A hazard model analyzes the conditional probability of an event, such as sale of a firm, given that the event has not occurred up to the end of the sample period. Hence, a hazard model takes into account that private firms that do not sell out by the end of their available data in the sample retain the option to do so. Competing risk models allow researchers to analyze the probability of an outcome, such as a private firm being acquired, while taking into account the probability of a competing outcome, such as a private firm going public via an IPO. The results of hazard and competing risks models are consistent with our findings using logistic regressions and further corroborate our conclusions. We further explore the association of CEO compensation with the quality of the outcome for the shareholders of the selling firm, i.e., the valuation at the time of the IPO/acquisition. Following the methodology in Bayar and Chemmanur (2012), we find that in our sample, the median premium acquired firms would have realized had they pursued an IPO instead is 108%, a result comparable to their finding of about a 75% IPO premium. Using a two-stage treatment regression approach to explain the premium, we find that the inclusion of CEO compensation characteristics in the second stage model implies that acquired private firms no longer sell at a discount relative to the private firms going public. Hence, the difference between how acquired and IPO firms compensate their CEOs helps explain part of the variation in the valuation premium of IPO firms relative to acquired firms documented in the prior literature (Poulsen and Stegemoller, 2008; Bayar and Chemmanur, 2012). Our research contributes to the literature by analyzing the differences in CEO compensation among firms that remain private, are acquired, or pursue an IPO. Earlier research explores factors associated with private firms deciding to either sell or go through the IPO process. Brau et al. (2003) examine market and industry characteristics affecting the choice of going public versus being acquired by a public acquirer and find that takeovers of private targets by publicly traded firms are more likely in higher marketto-book and more leveraged industries. In a comparison of IPO and acquired private firms, Poulsen and Stegemoller (2008) examine firm-specific characteristics and find that IPOs are preferred when the firms have greater growth opportunities and face financial constraints. More recently, Bayar and Chemmanur (2012) provide a theoretical model of a private firm's exit decision and show that the private firm CEO must consider the trade-off between the loss of control from an acquisition and financial constraints associated with remaining private. The model predicts that firms with higher probability of success in the product market are more likely to go public. Our sample includes private firms that remain private which help us determine the characteristics associated with selling or going public. Our work also complements research examining CEO compensation of public target firms and takeover activity.3 Cai and Vijh (2007) find that target CEOs with less liquid equity holdings accept a lower premium. We contribute to the existing literature by examining another covariate of the decision to sell a private firm—CEO compensation, a previously unstudied area. The paper proceeds as follows: We develop hypotheses and discuss endogeneity in Section 2. In Section 3, we describe the data used in our study and provide univariate comparisons. In Section 4, we present empirical findings based on multivariate analysis and additional robustness tests. Section 5 concludes.

3 One strand of the acquisition literature explores the relationship between CEO compensation and acquisition behavior of acquirers. Datta et al. (2001) analyze how equity compensation affects whether managers pursue an acquisition of another company. Minnick et al. (2011) analyze whether equity compensation is related to acquiring bank holding company's CEOs finding accretive deals.

446

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

2. Hypotheses development and endogeneity Ex ante, it is not obvious whether the association between CEO pay and the probability of sale of a private firm should be positive or negative. We therefore discuss frameworks for either a positive or a negative association of CEO compensation and the probability of sale of a private firm. We broadly group these into either the motivation framework or the illiquidity framework. 2.1. Motivation framework of CEO compensation We propose that a positive relation between CEO total or equity-based compensation and the probability of the sale of a firm reflects the goal of motivating the CEO. There are several reasons that lead to the motivation framework of CEO compensation. First, CEOs of private firms undergoing a sale may be paid more to encourage additional effort. In particular, selling the firm—whether to an acquirer or to the public in an IPO—requires effort by and is time consuming for the CEO, above and beyond the management of the firm. For example, Bengtsson and Hand (2011) propose that CEOs in private venture-capital backed firms are “rewarded with higher pay because identifying investors and convincing them to make risky investments … require[s] considerable time, skill, and costly actions on [the CEO's] part”. Hence, additional incentives in form of higher CEO compensation are required. Second, the sale of a firm leads to a change in control that directly affects the CEO of the selling firm, i.e., a potential loss of job. Therefore, higher total and equity-based compensation of the CEO of the selling firm reflects the payment for the potentially adverse effects on the CEO because of the change in control. Finally, after an IPO, firms have a liquid market for their shares, which enables the CEOs to sell their shares following lockup expiration (Aggarwal et al., 2002; Helwege et al., 2007). Similarly, an acquisition allows CEOs to cash out of their option and stock holdings, since the stock and options vest upon a change in control, allowing the CEOs to exercise their options and to sell the stock in the acquisition (Cai and Vijh, 2007). Hence, prior to the IPO or the acquisition, higher equity-based compensation reflects the increase in effort associated with selling the firm. Hence, we propose the following motivation hypothesis: H1A. (Compensation): There is a positive association between total and equity-based CEO compensation and the probability of selling the firm. Prior literature documents that private firms that are acquired have a lower valuation than firms that go public (Poulsen and Stegemoller, 2008; Bayar and Chemmanur, 2012), which Bayar and Chemmanur (2011) refer to as the “IPO valuation premium puzzle”. This “IPO valuation premium puzzle” reflects a higher quality outcome for shareholders of private firms going public than those that are acquired. We propose that the “IPO valuation premium puzzle” reflects the CEO compensation structure and the motivation the compensation provides. Specifically, we propose that: H2. (Exit pricing): There is a positive association between equity-based CEO compensation and the sales price of the firm.

2.2. Illiquidity framework of CEO compensation Because private firms' shares are illiquid, they may not be readily sellable and may have to be sold at a discount. Officer (2007) documents that private takeover target firms are sold at a discount and argues that the discount reflects the illiquid nature of the private firms. Since the illiquidity of shares of firms that remain private is not expected to change dramatically over time, the CEOs may require more equity-based compensation to offset the illiquidity discount.4 This in turn may increase total compensation since total compensation is positively associated with equity-based compensation. Therefore, we propose that a negative relation between CEO total and equity-based compensation with the likelihood of a sale, reflects the illiquid nature of the private firms' shares. Hence, we propose the following illiquidity hypothesis: H1B. (Compensation): There is a negative relation between total and equity-based CEO compensation and the probability of selling the firm.

2.3. Endogeneity Private firms in our sample self-select into being acquired or going public via an IPO. Hence, endogeneity may arise due to the simultaneous determination of the outcome and CEO compensation characteristics. While correlation of CEO compensation with private firm sales may be interesting, we attempt to control for the effects of endogeneity on our results in several ways: instrumental variable approach and propensity score matching. For the instrumental variable approach, we employ a two-stage regression analysis to create a measure of “predicted” compensation based on instruments that are not directly related to the likelihood of outcomes for private firms, but are instead directly 4 Another rationale consistent with the illiquidity framework is offered by Helwege and Packer (2009). Helwege and Packer (2009) point out that the existence of benefits of control serves “as the most significant incentive to stay private”. In private firms, equity-based compensation increases the CEO's ownership and correspondingly affects private benefits of control. Hence, if equity-based compensation increases private benefits of control then higher equity-based compensation is less likely to be associated with seeking a sale of a firm.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

447

related to compensation. To estimate CEO predicted cash and equity-based compensation, we use the following instrumental variables: whether the CEO received a graduate degree, CEO's age, length of CEO's tenure at the firm, and the fraction of the firm's headquarters state population with a graduate degree. We propose that these characteristics directly relate to CEO's human capital (or managerial quality) and, hence, compensation. We also argue that these characteristics do not directly affect whether a private firm goes public, is acquired, or remains private, other than through CEO compensation that is reflective of managerial quality. Palia (2001) uses four instrumental variables in his analysis of CEO compensation: CEO tenure, CEO quality of education (i.e., whether the CEO graduated from a prestigious university), CEO age, and firm volatility. Two of our instrumental variables, CEO tenure and age, are identical to the instrumental variables used by Palia (2001). With respect to CEO's human capital, our instrumental variables capturing whether CEO has a graduate degree is related to the Palia's (2001) measure of CEO quality of education. Similarly, Falato et al. (2015) use the selectivity of a college as a signal of manager skill. With respect to the fraction of a firm's headquarter state population with a graduate degree, we extend Minnick and Raman (2016) analysis of board size and propose that the CEO compensation reflects the level of education of the labor supply in the state while it does not affect the likelihood of a firm pursuing a sale. We do not use CEO's attendance of Ivy League as an instrumental variable because while the attendance of Ivy League college is a proxy for quality of education, it may also reflect strong (personal) connections with and ease of access to investment banking firms. Hence, the attendance of Ivy League college may violate the exclusion restriction.5 Furthermore, unlike Palia (2001), we do not use firm volatility as an instrumental variable since given that the firms in our sample are private we are unable to calculate it. To summarize, we use instrumental variables that capture managerial quality while also meeting the exclusion restriction. Our argument that managerial quality affects outcomes for firms relies on similar logic as in Chemmanur et al. (2015) who propose and find support for the argument that top management quality, as proxied by education, affects the firm's innovation activities. For propensity score matching, we identify one matched firm for each treatment (acquired or IPO) firm by finding the closest match in terms of the propensity score from a sample of firms that neither are acquired nor go public in a particular year (“no outcome”). We match based on ROA, size, VC backing, and industry. We allow for matching with replacement, i.e., a no outcome firm may be matched to more than one treatment firm, and restrict the matching propensity score to be within 1% of a treatment firm.

3. Data and sample characteristics Our data is from Capital IQ (CIQ), an affiliate of standard and Poor's, which has CEO compensation data comparable in detail to that of ExecuComp (also an affiliate of Standard and Poor's). CIQ is a comprehensive database of private and public firms that is widely available to researchers. Gao et al. (2012) and Gao et al. (2013) also use this database which they argue is the most comprehensive database used in studying private firms to date. The data is obtained from mandatory SEC disclosures by private firms, as well as Dunn and Bradstreet (for private firms that issued debt). One such reason for the mandatory SEC disclosures is a decision to pursue a registered public debt offering, which triggers a filing of a registration statement (Form S-1) pursuant to the Securities Act. Another reason for the mandatory SEC disclosures are stated in the Exchange Act: “[private firms] must file an Exchange Act registration statement that contains information on executive compensation if: It has more than $10 million total assets and a class of equity securities, like common stock, with 500 or more shareholders…” Gao et al. (2012) use this data in their comparison of the compensation of public and private firms. They suggest private firms reporting compensation are larger relative to private firms in the overall CIQ population, as well as private firms in studies using data from the Surveys of Small Business Finances (Cavalluzzo and Sankaraguruswamy, 2000; Cole and Mehran, 2013). Hence, given their larger size, it is more likely that the private firms in our sample would have the choice to go public relative to the average private firm. Since private firms with lower information asymmetry are more likely to be acquired (Shen and Reuer, 2005), the firms in our sample are also more likely to be acquired than typical private firms. Therefore, the outcomes of going public or being acquired are reasonable possibilities faced by firms in our sample. Since other smaller private firms may not realistically face such outcomes, we believe that our sample is appropriate for our analysis. Nevertheless, it is important to note the uniqueness of our sample before extrapolating the results to samples of other private firms. We identify a sample of private firms during the period 2003 to 2011 that have accounting and compensation data available on CIQ, and are not publicly listed.6 We also exclude financial firms. Our sample contains 4727 private firm-years. Next, we categorize the private firms into three distinct samples: 1) firms that become acquisition targets, 2) firms that become publicly listed through an IPO, and 3) firms that neither are acquired nor go public in a particular year (“no outcome” firms). We first identify the sample of firms that are acquired or go public using CIQ's event data. We identify a sample of 203 acquired, 766 IPO, and 3758 no outcome firm-year observations. From this sample, 1384 firms have one year of data, 679 have two years of data, and 579 have three or more years of data, for an average of about two years per firm.7 We collect additional information about the acquisition and IPO transactions from SDC. Appendix A lists the variables used in our study, their definitions, and sources. 5

We thank the referee for pointing out this issue. Because CIQ identifies firms as private (or public) based on their most recent status, we classify private firm-years as those that do not have a reported market capitalization in the year relevant for our analysis. Further, we hand collect ownership data for private firms and verify the private status of the firms in our sample. 7 We verify that no outcome firms leaving the sample due to data availability in year t + 1 are not leaving due to being acquired (i.e., potentially misclassified by CIQ). We perform a public press search for a random subsample of such firms and find no references to acquisitions. We also perform public press searches for such firms to see whether they leave our sample due to bankruptcy. Our search does not find evidence of bankruptcy explaining why firms exit out sample. 6

448

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

Table 1 Distribution of sample. Panel A: sample over time Year

No outcome

Acquired

IPO

Total

2003 2004 2005 2006 2007 2008 2009 2010 2011 Total

541 443 425 395 352 423 450 420 309 3758

21 39 31 37 27 14 9 13 12 203

56 119 104 96 92 40 29 94 136 766

618 601 560 528 471 477 488 527 457 4727

Panel B: industry distribution Industry Consumer nondurables Consumer durables Manufacturing Energy oil, gas, and coal extr. and products Chemicals and allied products Business Equipment Telephone and television transm. Utilities Wholesale, Retail, and Some Services Healthcare Other Total

No outcome

Acquired

IPO

Total

133 161 255 298 241 480 86 185 814 219 886 3758

9 4 9 11 5 16 4 2 18 7 118 203

10 9 43 59 76 157 16 4 131 82 179 766

152 174 307 368 322 653 106 191 963 308 1183 4727

This table shows the breakout of the private firm-years in our sample for the 2003–2011 time frame. There are three groups—firms remaining private, firms that are acquired, and firms that go public (IPO). IPO firms in our sample list their shares on non-OTC and OTC exchanges. Panel B shows the break down by industry using the Fama-French 12 industry classification.

The time series distribution of the sample of acquired, IPO, and control firms is shown in Table 1, Panel A.8 During 2008, coinciding with the most recent crisis, the number of IPOs dramatically declines. Starting in 2010, the number of firms going public starts to increase and by 2011 exceeds the pre-crisis number of IPOs. This distribution follows the frequency of IPOs based on Jay Ritter's data during the same period.9 Acquisitions show a declining trend over the sample period. The distribution of the acquired firms is similar to a distribution of an SDC sample of private firms subject to an acquisition, a more commonly used database for research on acquisitions. We verified that both the IPO and acquisition samples we use are a subset of these transactions and reflect the availability of the CEO compensation data on CIQ.10 Table 1, Panel B reports the industry representation in the three subsamples. We use the Fama-French 12 industry group classifications. The top three industry groups for firms remaining private, acquired, and going public are “Other”, “Wholesale, Retail, and Some Services”, and “Business Equipment”.11 Based on untabulated analysis, we note that for the IPO sample, technology firms (“Computer Software” and “Electronic Equipment” industry groups based on Fama-French 48 industry group classification) account for 149 or 19.5% of the sample IPOs. Overall, the cross-sectional variation of CEO compensation likely reflects the industry diversity in our sample. Hence, in our multivariate analyses, we control for industry fixed effects. Panel A of Table 2 presents means and medians of descriptive characteristics for the firm-years in each of the three samples: acquired, IPO, and no outcome firm-years. We identify whether a firm is backed by venture capital firms (VCs) using CIQ. Consistent with Bayar and Chemmanur (2012) who show that VCs prefer to take firms public, 45% of the IPO sample is VC backed in contrast to 8% of the acquired firms and 12% of control firms. The average revenue of acquired firms ($615 million) is significantly greater than that of IPO firms ($493 million). In fact, the revenue of IPO firms is also smaller than the revenue of firms remaining private ($796 million). Acquired and IPO firms are similar in terms of total assets. However, firms remaining private are significantly larger than acquired or IPO firms. IPO firms have the lowest leverage, highest cash holdings, and highest capital expenditures out of the three samples. With respect to operating performance, we calculate ROA as the ratio of net income to total assets. The results in Panel A of Table 2 show that IPO firms have the highest ROA, with a mean of − 22.6% for IPOs and − 96.8 for acquired firms. The large

8 IPO firms in our sample list their shares on NYSE/Amex/Nasdaq as well as on the OTC market. To assess the robustness of our findings, in untabulated results we exclude firms going public on the OTC market. Our conclusions are not materially affected when we exclude OTC firms from our analyses. 9 http://bear.warrington.ufl.edu/ritter/IPOs2011Statistics70512.pdf. 10 In untabulated results we note that for the years 2003 through 2011, before requiring any data availability, CIQ contains 31,289 acquisitions of private firms and 10,240 IPOs. The sample decline reflects the requirement that CEO compensation data is available. 11 Our conclusions are robust to including financial firms in our analyses.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

449

Table 2 Univariate analyses by outcome. Panel A: firm characteristics

Test of difference

Venture backed Revenue (million)

Ave Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med

Total assets (million) Leverage Total cash/total assets CapEx/Total Assets ROA Industry market-to-book # IPOs in industry # M&As in industry

No outcome (N = 3758)

Acquired (N = 203)

IPO (N = 766)

No outcome vs. acquired

No outcome vs. IPO

Acquired vs. IPO

0.12 $796 $37 $1870 $62 90.0% 43.8% 12.6% 4.9% 6.2% 2.6% −74.2% −0.5% 2.44 2.14 10.11 5.00 6.47 3.00

0.08 $615 $120 $1155 $183 88.2% 50.3% 12.3% 4.6% 5.6% 2.3% −96.8% 0.2% 2.37 2.09 9.02 4.50 6.39 3.00

0.45 $493 $95 $763 $108 41.5% 29.5% 18.7% 9.9% 7.2% 3.7% −22.6% 0.5% 2.81 2.54 12.87 9.00 8.20 5.00

***

*** *** *** *** * *** *** *** *** *** *** *** *** * * ** ** *** ***

*** *** *

Panel B: CEO characteristics

Tenure Founder CEO ownership Graduate degree indicator CEO age State level of education

Ave Med Ave N Ave N Ave Ave Med Ave Med N

Cash compensation ratio Equity-based compensation ratio

RSGs/total Options/total

** *** *** *** * * * *** *** *** ***

Test of difference

Total compensation (million)

Bonus/total

** ** ** **

* *** ***

No outcome

Acquired

IPO

No outcome vs. acquired

No outcome vs. IPO

Acquired vs. IPO

5.22 3.00 0.22 3758 8.9% 1657 31.3% 51.16 50.00 7.4% 6.6% 3758

5.42 4.00 0.19 203 4.1% 98 36.6% 50.76 46.00 7.1% 6.5% 203

4.33 3.00 0.29 766 12.9% 497 42.4% 50.73 49.00 7.7% 7.5% 766

* * *

** * ***

** ** ***

***

***

***

*

*

Panel C: CEO compensation

Salary/total

* *** * *** *** *** *** *** * *** ***

Test of difference

Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med Ave Med

No outcome (N=3758)

Acquired (N=203)

IPO (N=766)

No outcome vs. acquired

No outcome vs. IPO

Acquired vs. IPO

$1.94 $0.40 0.72 0.90 0.14 0.00 0.62 0.67 0.10 0.00 0.03 0.00 0.12 0.00

$2.50 $0.58 0.72 0.93 0.16 0.00 0.57 0.55 0.16 0.00 0.02 0.00 0.14 0.00

$1.57 $0.68 0.67 0.75 0.20 0.00 0.50 0.49 0.17 0.07 0.02 0.00 0.20 0.00

* ***

* *** * *** *** *** *** *** *** *** *** ** ** ***

*** * * *** *** *** *** *** *** *** *** ** *** ***

* * * *** *** *** *** *** ** ** *

This table shows the mean and median values of sample characteristics by one of three outcomes: no outcome, acquired, or IPO. Panel A shows the firm-/industryspecific characteristics, Panel B shows distribution of CEO characteristics, and Panel C shows distribution of CEO compensation. All variables are defined in Appendix A. Test of difference is based on the two-tailed t-tests for means or the Wilcoxon rank-sum test for medians. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively.

negative ROAs are a result of the fact that a non-trivial number of firms report large cumulative losses and as a result have negative equity. The median ROA for IPO and acquired firms are 0.5% and 0.2%, respectively. Furthermore, in untabulated results, we split each of the samples based on whether the total equity is positive or negative. A total of 86% of the firms remaining private,

450

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

68% of the acquired firms, and 95% of the IPO firms have positive equity; the average ROA for these firms is a more reasonable −15.3%, −14.2%, and −9.0%, respectively. In our multivariate analyses, we use an indicator variable to account for these negative equity firms. With two exceptions, our sample has similar characteristics to the sample used by Gao et al. (2012). The two exceptions are the greater number of firms with negative equity and the smaller average ROA in our sample relative to figures reported by Gao et al. (2012). Both of these differences likely arise because we include the recent crisis years in our sample, while Gao et al. (2012) do not. In fact, when we exclude 2007–2008 firms from our sample, both the number of firms with negative equity and the average ROA in our sample becomes comparable to results reported by Gao et al. (2012). Panel A of Table 2 also provides industry-specific characteristics for each of the three samples. For the purposes of these variables, we define industry at the two-digit SIC code level. The industry average market-to-book ratio is highest for the IPO firms, followed by firms remaining private and acquired firms. However, the differences appear marginal. The number of firms in the same industry going public during the relevant calendar year is highest for IPO firms and lowest for acquired firms. The market for corporate control, measured as the number of acquisitions of firms in the relevant industry and calendar year, is also most active for IPO firms and lowest for acquired firms. CEO characteristics are reported in Panel B of Table 2. We collect data on CEO tenure and whether the CEO is a founder from CIQ and supplement and hand-collect the missing data from SEC filings. We find that CEOs of IPO firms are most likely to be founders and have the shortest tenure when compared to the other two samples. CEOs of acquired firms have the longest tenure relative to the other two samples. CEOs of acquired firms are also the least likely to be founders. We also collect CEO ownership from CIQ. We are able to collect ownership data for slightly less than half of the sample or 2250 firm-years (=1657 remaining private firms + 98 acquired firms + 497 IPO firms). We find that on average CEOs of firms going public have the highest average share ownership, followed by CEOs of firms remaining private and CEOs of acquired firms. In untabulated results, we find that the median CEO ownership is zero across all three samples. We search the biographical information of CIQ and BoardEx to identify the degree received (graduate degree) and age; where information in these sources is missing, we hand-collect the information from SEC filings and searches of public information). We are able to collect these data for all CEOs in our sample. The results show that a significantly higher fraction of CEOs of IPO firms earned a graduate degree. The average ages of the CEOs range from 50.73 years to 51.16 years and are not statistically different across the three samples, and are similar to that reported by Falato et al. (2015). We also collect data on the fraction of state population with a graduate degree for the state of state's headquarters as in Minnick and Raman (2016). To the extent that the labor supply in a particular state has higher education, the compensation may reflect the varying level of education. In our sample, the state level of education is not statistically different across the three samples.

4. Empirical analyses 4.1. Univariate analysis of CEO compensation and outcomes for private firms With respect to the CEO compensation structure, we analyze total compensation and its components. Specifically, we collect information on salary, bonus, and the values of restricted stock grants (“RSGs”) and of option grants.12 We scale each of the components by the total compensation. Total compensation includes salary, bonus, value of option grant, RSG value, and other compensation. We also calculate a cash compensation ratio as salary plus bonus. Similarly, we define equity-based compensation as the sum of the value of RSG and option grants. For acquired and IPO firms, we measure CEO compensation at least one year prior to the sale.13 Table 2, Panel C presents descriptive statistics of the CEO compensation of remaining private, acquired, and IPO samples. CEOs of firms that remain private earn on average a total compensation of $1.94 million. The mean CEO total compensation is $2.50 million and $1.57 million for acquired and IPO firms, respectively. The results indicate that CEOs of firms pursuing an IPO earn significantly less than CEOs of acquired firms as well as firms that remain private. These results offer mixed support for the hypotheses tested. While CEOs of acquired firms get paid more than CEOs of firms remaining private, a result consistent with the total compensation motivation hypothesis, CEOs of firms going public get the lowest total compensation, a result consistent with the total compensation illiquidity hypothesis. With respect to the composition of CEO pay, on average, cash compensation of firms remaining private and acquired firms is comparable (72%), while the IPO firms pay their CEOs less cash (67%). We also find that the equity-based compensation is higher in IPO and acquired firms relative to the no outcome firms that suggest that selling firms provide stronger equity-based incentives

12 To measure the value of option compensation for these private firms, we calculate the Black-Scholes value of the option following Gao et al. (2012) who use the median volatility of public firms in the same two-digit industry, the 7-year Treasury bond yield as the risk-free rate, the exercise price as the grant date price, and the time to maturity of 7 years. 13 We note that the process leading to an acquisition or an IPO is likely to take longer than one year. For example, a firm has to accomplish several steps prior to going public: retaining an auditing firm to perform audit of several years of financials, retaining advisors, investment bankers, and attorneys, preparing S-1 registration statement, filing amendments if necessary, waiting for response from the SEC regarding the S-1 statement and the amendments, going on road show, pricing the stock, and making allocations to interested investors. There is some anecdotal evidence that the IPO process can take well over a year and may be even close to two years. For acquisitions, Boone and Mulherin (2007) provide anecdotal evidence that that the time between the initial steps in the acquisition process and the ultimate sale of the firm exceeds one year. Hence, measuring compensation in year prior to the actual outcome may in fact capture payment for effort of the CEO up to that point.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

451

to their CEOs. This result is consistent with the motivation hypothesis, whereby equity-based compensation reflects the intent to motivate the CEO to pursue the sale of the firm. Finally, focusing on individual components of the CEO compensation, salary represents a greater portion of total compensation for firms remaining private than for either acquired firms or IPO firms. Bonus and options represent a greater proportion of CEO compensation for selling firms, while restricted stock grants are highest at no outcome firms. Overall, the univariate analysis does not offer a clear support for the motivation over the illiquidity hypotheses. This is likely because firms and CEOs in the three samples differ in other ways, which makes the univariate analysis of the relation between the CEO compensation structure and exit outcomes difficult to interpret. Hence, we next examine the relation of CEO compensation level and structure relate with the outcomes for private firms in a multivariate setting. 4.2. Multivariate analysis: sellout vs. remain private We begin the multivariate analysis by estimating a logistic regression model with the dependent variable equal to one if the private firm becomes an acquisition target or goes public and zero if it remains private in a particular year. In all models, we control for firm size using log of revenue. Following Poulsen and Stegemoller (2008) who show that firms with higher capital Table 3 Logistic regression: probability of sale. Full sample (1) log(CEO total compensation)

Including CEO ownership (2)

(3)

0.100*** (0.00)

Cash compensation ratio

0.205 (0.26) 0.574*** (0.00)

Equity-based comp. ratio

Bonus/total RSG/total Option/total

Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Constant Observations Pseudo R-squared

0.146*** (0.00) 0.029 (0.23) −0.039 (0.27) 0.938*** (0.00) 1.101*** (0.00) −0.266*** (0.01) 1.227*** (0.00) 0.013 (0.76) −0.030*** (0.00) 0.276*** (0.00) 0.001 (0.74) 0.003 (0.63) −4.862*** (0.00) 4727 0.164

(6)

0.321 (0.20) 0.384** (0.04)

CEO ownership

ROA

(5)

−0.210 (0.27) 1.459*** (0.00) −0.785* (0.10) 0.700*** (0.00)

Salary/total

log(Revenue)

(4) 0.040* (0.05)

0.175*** (0.00) 0.022 (0.35) −0.039 (0.27) 0.956*** (0.00) 1.062*** (0.00) −0.268*** (0.01) 1.225*** (0.00) 0.014 (0.74) −0.029*** (0.00) 0.267*** (0.01) 0.001 (0.72) 0.003 (0.59) −3.884*** (0.00) 4727 0.164

0.130*** (0.00) 0.022 (0.36) −0.036 (0.30) 0.869*** (0.00) 0.983*** (0.00) −0.243** (0.02) 1.204*** (0.00) 0.020 (0.64) −0.030*** (0.00) 0.258*** (0.01) 0.000 (0.99) 0.003 (0.64) −3.672*** (0.00) 4727 0.176

0.566** (0.02) 0.146*** (0.00) 0.032 (0.32) 0.051 (0.16) 0.868*** (0.00) 1.010** (0.03) −0.394*** (0.00) 1.272*** (0.00) 0.097* (0.10) −0.026*** (0.01) 0.288** (0.03) −0.006 (0.19) −0.003 (0.64) −4.155*** (0.00) 2252 0.164

0.592** (0.01) 0.160*** (0.00) 0.028 (0.37) 0.051 (0.16) 0.871*** (0.00) 0.982** (0.03) −0.401*** (0.00) 1.274*** (0.00) 0.098* (0.10) −0.027*** (0.01) 0.284** (0.03) −0.006 (0.18) −0.003 (0.70) −3.956*** (0.00) 2252 0.165

−0.111 (0.67) 1.572*** (0.00) 1.401** (0.02) 0.539** (0.04) 0.553** (0.03) 0.120*** (0.00) 0.024 (0.45) 0.050 (0.16) 0.822*** (0.01) 0.945** (0.04) −0.375*** (0.01) 1.251*** (0.00) 0.111* (0.06) −0.029*** (0.01) 0.276** (0.04) −0.007 (0.13) −0.004 (0.58) −3.795*** (0.00) 2252 0.179

This table shows logit estimation where the dependent variable is equal to one if there is an IPO or acquisition and zero otherwise. All regressions include year and industry fixed effects. Table 2 describes the independent variables. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

452

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

expenditures and with venture capital backing are significantly more likely to pursue an IPO, we include both characteristics in our estimations. A number of empirical studies examine whether the likelihood of becoming a target is related to the profitability of a firm, with differing results (Palepu, 1986; Dickerson et al., 2002). Pagano (1993) suggests that companies may show temporarily high performance immediately before an IPO with the hope that investors will think the profitability is permanent. We use ROA to control for a firm's performance. Highly leveraged companies face stronger borrowing constraints, which can motivate the need to go public or be acquired (Perevozchikov, 2010). Hence, we control for leverage, calculated as the ratio of total debt to assets. We control for the tenure of the CEO to capture whether any firms hired a new CEO in anticipation of being sold. We also control for whether the CEO is a founder as a proxy for private benefits of control. Brau et al. (2003), Harford (2005), Hsieh et al. (2011), and Kummer and Steger (2008) suggest that IPOs and acquisitions come in waves. We therefore control for the contemporaneous market conditions using the average industry market-to-book and the number of IPOs and takeovers in the industry as previously defined. We also include an indicator variable, which equals one if the firm has negative equity, and zero otherwise. The independent variables are measured in the year preceding the outcome (Cavender et al., 1992). Consistent with prior literature, we include industry and year fixed effects in all regressions. Industry indicators are based on FamaFrench 12 industry classification to ensure sufficient number of observations in each industry. All regressions use heteroscedasticity consistent standard errors clustered at the firm level. Table 3 reports the results of the logistic regression analysis modeling the probability of a sale. Regressions (1) to (3) use the full sample with available data. Regressions (4) through (6) also include CEO equity ownership for a subset of firms with available data. The coefficient on total compensation in regression (1) indicates that CEOs of selling firms have higher compensation than those that remain private, even after controlling for other firm characteristics. The coefficient is significant at the 1% level. This result is potentially consistent with the motivation hypothesis. We next analyze the role of the components of the CEO pay on the likelihood of a firm being sold. We note a positive and significant coefficient on equity-based compensation in regression (2), a result consistent with the motivation hypothesis and contrary to the illiquidity hypothesis. In regression (3), bonus and option compensation are significantly greater for selling firms, while salary and restricted stock grants make up a lower fraction of total compensation, albeit, only the coefficient on restricted stock grants is insignificant. The results concerning bonus and option compensation are consistent with the motivation framework of CEO compensation at private firms, while the results concerning restricted stock grants contradict the motivation hypothesis.14 While our central focus is on how compensation is related to the likelihood of a sale of a firm, selling the firm affects the value of the CEO's shares that in turn likely affects the CEO's willingness to sell the firm. We note that while share ownership can motivate the CEO to pursue the sale of the firm, it can also reflect CEO private benefits of control. To assess whether our prior conclusions are affected by the level of CEO ownership, we estimate all regressions controlling for CEO ownership. We report the results in Table 3, regressions (4) to (6). In all regressions, the coefficient on CEO ownership is positive and significant at the 5% level, indicating that ownership further serves to motivate the CEO to pursue the sale of the firm. The coefficient on total compensation in regression (4) remains positive, however, is less than half of the corresponding coefficient in regression (1), and is significant at the 10% level. Results in regressions (5) and (6) are consistent with the motivation hypothesis, even after controlling for CEO ownership. Specifically, we note a positive and significant coefficient on equity-based compensation in regression (4). In regression (6), bonus, restricted stock grants, and option compensation are all significantly greater for selling firms. Thus, incentive-based compensation is greater in firms that sell than those that remain private. These results are consistent with the motivation framework of CEO compensation at private firms. With respect to the control variables, selling firms are larger, have more cash, and have higher capital expenditures than firms that remain private. The negative coefficient on the negative equity indicator suggests that firms with large cumulative losses are less likely to be sold. Not surprisingly, venture capital backing is significantly and positively related to the likelihood of being sold in all regressions. Firms with CEOs who are founders and CEOs with shorter tenures are also more likely to be sold. The analyses in Table 3 distinguish between two outcomes: sale of a firm versus remaining private. Since in our sample we observe three outcomes, we next estimate a multinomial logistic regression to separate the forms of outcomes into firms that are acquired, go public, or remain private (base case). The results of the unordered multinomial regression for the full sample are presented in Table 4. The base outcome in all models is for a firm to remain private. In addition to reporting the coefficients, we also present odds ratios for key variables of interest to aid in the interpretation of the results. In regression (1), we find that acquired and IPO firms have higher total compensation relative to firms that stay private, with the coefficients being comparable. The odds ratios imply that for one (log) unit increase in total compensation, getting acquired or going public is about 1.1 times more likely. As regression (2) shows, the significance of the CEO total compensation is related to equity-based compensation. This is true for CEOs of both acquired and IPO firms. The effects of equity-based compensation for the acquired and IPO firms are positive and significant at the ten and 5% level, respectively. The odds ratios imply a 1.7 increase in the likelihood of an acquisition or an IPO for each unit change in equity-based compensation. Regression (3) shows that both acquired and IPO firms have higher bonus and option compensation than those that stay private. IPO firms also have higher RSG grants in the year prior to going public. The results showing higher bonus pay for acquired and IPO firms is consistent with Bengtsson and Hand (2011) who find that CEOs are rewarded for their efforts in raising new equity capital for the firm. Overall, the importance of

14 There are 348 observations of firms that are private as a result of a prior leverage buyout (LBO). LBOs may be larger and may differ from non-LBOs in other ways. Hence, in untabulated results, we exclude all LBOs from our analyses. Our conclusions are not materially affected by the presence of LBO firms in our sample.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

453

Table 4 Unordered multinomial logit: probability of sale. Acquisition vs. no outcome (1) log(CEO total compensation)

(2)

Salary/total

Bonus/total

RSG/total

Option/total

Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Constant Observations Pseudo R-squared

(1)

0.063* (0.08) −0.016 (0.56) −0.003 (0.93) 0.200 (0.66) −0.333 (0.67) 0.119 (0.50) −0.302 (0.27) −0.252** (0.02) 0.009 (0.49) −0.105 (0.59) 0.002 (0.78) −0.010 (0.16) −5.055*** (0.00)

(2)

(3)

0.086** [1.09] (0.03) 0.197 [1.22] (0.54) 0.550** [1.73] (0.03)

Equity-based comp. ratio

ROA

(3)

0.136** [1.15] (0.02)

Cash compensation ratio

log(Revenue)

IPO vs. no outcome

0.107*** (0.00) −0.025 (0.37) −0.004 (0.90) 0.219 (0.63) −0.380 (0.63) 0.123 (0.49) −0.287 (0.29) −0.253** (0.02) 0.009 (0.46) −0.121 (0.53) 0.002 (0.77) −0.010 (0.16) −3.668*** (0.00)

0.214 [1.24] (0.31) 0.549** [1.73] (0.01) −0.076 [0.93] (0.82) 1.340*** [3.82] (0.00) 0.255 [1.29] (0.77) 0.670* [1.95] (0.07) 0.069** (0.03) −0.023 (0.40) −0.003 (0.93) 0.183 (0.69) −0.508 (0.52) 0.156 (0.38) −0.319 (0.24) −0.242** (0.03) 0.009 (0.45) −0.127 (0.52) 0.001 (0.85) −0.010 (0.17) −3.597*** (0.00)

0.174*** (0.00) 0.098** (0.03) −0.166* (0.06) 1.100*** (0.00) 1.572*** (0.00) −0.299** (0.02) 1.471*** (0.00) −0.008 (0.16) −0.048*** (0.00) 0.406*** (0.00) 0.005 (0.97) 0.006 (0.10) −5.578*** (0.00) 4727 0.170

0.198*** (0.00) 0.090** (0.05) −0.162* (0.06) 1.116*** (0.00) 1.532*** (0.00) −0.307** (0.02) 1.466*** (0.00) −0.006 (0.14) −0.047*** (0.00) 0.399*** (0.00) 0.005 (0.97) 0.006* (0.10) −4.776*** (0.00) 4727 0.170

−0.255 [0.77] (0.24) 1.501*** [4.49] (0.00) 1.039* [2.83] (0.06) 0.669*** [1.95] (0.00) 0.152*** (0.00) 0.085* (0.06) −0.151* (0.08) 1.019*** (0.00) 1.446*** (0.00) −0.292** (0.02) 1.453*** (0.00) −0.006 (0.12) −0.048*** (0.00) 0.388*** (0.00) 0.004 (0.77) 0.006* (0.10) −4.515*** (0.00) 4727 0.180

This table shows unordered multinomial logit estimation where the dependent variable is IPO, acquisition, and no outcome. Panel A uses all firms in the sample. Panel B excludes the firms going public on OTC market. All regressions include year and industry fixed effects. All variables are defined in prior tables. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. Odds ratios are reported in brackets below the coefficients for selected variables. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

total and equity-based compensation at acquired and IPO firms further corroborates the motivation framework of CEO compensation for both acquired and IPO firms.15 With respect to the control variables, size (measured by log(Revenue)) is again positively related to the likelihood of being acquired or going public. Firms with higher profitability, lower debt ratios, higher cash and capital expenditures, are more likely to go public than remain private. IPO firms are also less likely to have negative equity than those that stay private. Thus, firms that go public are healthier than firms that remain private, while acquired firms are more comparable to those firms that remain private. Consistent with other research, we find that firms backed by VCs are more likely to go public. Interestingly, coefficients on 15 We also estimate multinomial logistic regressions controlling for CEO ownership. The untabulated results are consistent with prior results. Specifically, we find that CEO ownership and all forms of incentive-based compensation are significantly positively related with the likelihood of an IPO. These results reinforce our prior conclusions about the relevance of the motivation framework of CEO compensation on outcomes for private firms.

454

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

CEO tenure and founder CEO indicator are only significant for IPO firms versus firms that remain private. The results show that IPOs in the post-2001 period are large firms that are managed by founder CEOs and CEOs with short tenures. Overall, the results in Tables 3 and 4 are inconsistent with the illiquidity hypothesis of CEO compensation. Instead, the results are consistent with the view that CEO pay is structured to compensate for higher effort due to the sale of the firm. Specifically, CEOs of selling firms have higher total, equity-based, bonus, and option compensation than CEOs of firms remaining private. 4.3. Endogeneity To control for the potential endogeneity, we use instrumental variable approach and propensity score matching. We test the validity of the previously described instruments in our sample and in untabulated results note that Hansen J-test indicates that our instruments are valid. Specifically, we find that these models yield F-statistics of 27.2 (Cash) and 30.5 (Equity), and p-values of 0.49 (Cash) and 0.57 (Equity) for the Hansen J-test of over-identification. Therefore, given that the F-statistics exceed the critical threshold value of 10, and that the p-values from the Hansen J-test exceed 0.10, the instruments used to estimate the likelihood of an exit are statistically valid (Hansen, 1982). The results of the first-stage models for each of the seven compensation-related variables are reported in Panels B and C of Table 5. In Panel B, we control for industry and year fixed effects, while in Panel C, we control for firm fixed effects. In Panel B, we find that all four variables have significant coefficients in all models. For example, in regression (1) for total compensation, the coefficients on graduate degree indicator, CEO age, and CEO tenure are positive and significant and the coefficient on the state level of graduate education is negative and significant. In regression (2) for cash-based compensation, all variables with the exception of CEO tenure switch signs. The models explain over 20% of the variation in the compensation variables, with the (untabulated) fit statistics being significant at the 1% level. In Panel C, we note reliably significant coefficients on graduate degree indicator and CEO tenure. Hence, including firm fixed effects eliminates the significance of CEO age and state level of education. The (untabulated) fit statistics are all significant at the 1% level. Panel A of Table 5 uses the predicted values of the compensation variables from the first-stage regressions instead of the actual compensation of the CEO of the private firms in our sample. Regressions (1) through (3) control for industry and year fixed effects and regressions (4) through (6) control for firm fixed effects. As in Table 4, we use the unordered multinomial logit model with the firms remaining private as the base group and use heteroscedasticity consistent standard errors clustered at the firm level. Consistent with prior results, the predicted total compensation has a positive and significant coefficient for both acquired and IPO firms (regressions (1) and (4)). Also consistent with prior results (Table 4), the predicted equity-based compensation in regressions (2) and (5) enters with a positive and significant coefficient for both acquired and IPO firms. The coefficients indicate a stronger effect for IPO firms than for acquired firms. Finally, in regressions (3) and (6), for both acquired and IPO firms, we find positive and significant coefficients for predicted bonus and predicted option compensation. In regression (3), we also find a positive and significant coefficient on predicted restricted stock grants and a negative and significant coefficient on predicted salary for acquired firms. However, these two coefficients are not significant when using firm fixed effects in regression (6). Overall, the results of the instrumental variable approach are consistent with the motivation framework of CEO compensation and corroborate our prior conclusions. Our second approach to investigate whether endogeneity may potentially affect our results relies on comparison of CEO compensation using propensity-matched samples as described previously. Propensity-based matching produced 969 matched pairs with standardized differences in firm characteristics of less than 10%, indicating a high degree of similarity in the distributions of independent variables used to predict the likelihood of exit. Panel A of Table 6 reports the differences in compensation structure for the matched firms that remain private with both the acquired and the IPO firms. CEOs of acquired firms and firms going public receive on average higher bonus, option, and total equity-based compensation. All these differences are not only statistically significant but also economically meaningful. For example, relative to firms remaining private, option compensation accounts for 5.4% more of total compensation for acquired and IPO firms. We also note higher (lower) total compensation (RSG) compensation for CEOs of IPO and acquired firms. Finally, salary and cash-based compensation is higher for CEOs of firms remaining private. In Panels B and C, we report the tests of matched firms remaining private versus either the acquired or the IPO firms, respectively. The results generally mirror the findings in Panel A. These results are consistent with the view that the compensation of CEOs of private firms in our sample is structured to compensate them for higher effort related to the sale of the firm. 4.4. Alternative modeling techniques: hazard and competing risk models The logistic regression analyses performed in prior tables do not take into account that private firms that neither sell out nor go public by the end of our sample retain the option to do so later. We address this issue by estimating a hazard model. A hazard model assesses the conditional probability of an event, given that the event has not occurred up to the present time (i.e. the hazard rate). In our case, the event is either an acquisition or an IPO. We estimate the hazard model on a panel dataset for all firms for which we are able to identify the necessary founding dates. We use the same independent variables in the hazard models as we used in the logistic regressions. As before, we control for industry and year fixed effects and rely on heteroscedasticity consistent standard errors. Table 7 shows the results of the hazard rate model analysis for the subsample of firms for which we have the founding date, resulting in a subset of 1141 firm-years for 659 unique firms.16 The results show that total compensation is no longer significant 16

Including CEO equity ownership further reduces the sample to 601 observations.

Table 5 Two stage multinomial logit. Acquisition vs. no outcome (1) Predicted log(CEO total compensation)

(2)

0.042* (0.10)

Predicted cash compensation ratio

Predicted RSG/total Predicted option/total

Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Constant Industry and year fixed effects Firm fixed effects Observations Pseudo R-squared

Acquisition vs. no outcome (3)

0.022 (0.45) −0.027 (0.29) −0.011 (0.71) 0.149 (0.74) −0.564 (0.46) 0.101 (0.56) 0.403 (0.13) −0.024 (0.75) 0.006 (0.61) −0.076 (0.68) −0.004 (0.50) 0.004 (0.71) −3.232*** (0.00) Yes No

(4)

0.157*** (0.00) 0.127* (0.10) −0.148 (0.12) 1.179*** (0.00) 1.421*** (0.00) −0.209 (0.11) 1.617*** (0.00) 0.190*** (0.00) −0.044*** (0.00) 0.383*** (0.00) −0.001 (0.67) 0.012** (0.02) −7.097*** (0.01) Yes No 4727 0.111

IPO vs. no outcome (6)

0.285* (0.10) 0.053 (0.71) 3.283* (0.05)

−1.509** (0.04) 1.315* (0.06) 0.333** (0.04) 1.076** (0.05) 0.019 (0.48) −0.028 (0.27) −0.012 (0.70) 0.181 (0.68) −0.640 (0.41) 0.121 (0.49) 0.378 (0.16) −0.021 (0.79) −0.009 (0.73) −0.083 (0.65) −0.004 (0.49) 0.004 (0.66) 136.853** (0.05) Yes No

(5)

0.154*** (0.00) 0.128* (0.10) −0.144 (0.13) 1.164*** (0.00) 1.422*** (0.00) −0.213* (0.10) 1.611*** (0.00) 0.190*** (0.00) −0.037*** (0.00) 0.386*** (0.00) −0.001 (0.67) 0.012** (0.02) −3.700*** (0.00) Yes No 4727 0.111

(4)

0.019 (0.46) −0.026 (0.30) −0.011 (0.72) 0.145 (0.74) −0.562 (0.46) 0.102 (0.56) 0.408 (0.12) −0.025 (0.75) −0.009 (0.73) −0.072 (0.70) −0.004 (0.50) 0.004 (0.71) −6.584 (0.37) No Yes

(6)

0.807** (0.02) −0.049 (0.73) 2.212** (0.02)

0.093 (0.73) 0.745* (0.06) 1.866 (0.37) 1.006** (0.04) 1.168 (0.35) 5.112*** (0.00) 0.158*** (0.00) 0.126* (0.10) −0.132 (0.14) 1.117*** (0.00) 1.477*** (0.00) −0.239* (0.07) 1.575*** (0.00) 0.185*** (0.00) −0.044*** (0.00) 0.391*** (0.00) −0.002 (0.64) 0.011** (0.04) −44.411 (0.28) Yes No 4727 0.114

(5)

0.022 (0.44) −0.027 (0.29) −0.011 (0.71) 0.148 (0.74) −0.561 (0.46) 0.101 (0.56) 0.404 (0.13) −0.025 (0.75) −0.007 (0.78) −0.076 (0.68) −0.004 (0.50) 0.003 (0.71) −3.061*** (0.00) No Yes

0.356 (0.86) 2.458* (0.10) 6.236 (0.43) 0.127* (0.10) 0.022 (0.40) −0.028 (0.28) −0.011 (0.72) 0.152 (0.73) −0.558 (0.47) 0.105 (0.55) 0.414 (0.12) −0.026 (0.73) −0.009 (0.73) −0.078 (0.67) −0.004 (0.50) 0.003 (0.75) −3.139* (0.06) No Yes

0.156*** (0.00) 0.127* (0.10) −0.141 (0.13) 1.156*** (0.00) 1.426*** (0.00) −0.219* (0.09) 1.607*** (0.00) 0.189*** (0.00) −0.080*** (0.00) 0.385*** (0.00) −0.001 (0.67) 0.012** (0.03) −13.555*** (0.00) No Yes 1197 0.174

0.154*** (0.00) 0.128* (0.10) −0.141 (0.13) 1.153*** (0.00) 1.426*** (0.00) −0.218* (0.09) 1.606*** (0.00) 0.189*** (0.00) −0.073*** (0.00) 0.387*** (0.00) −0.001 (0.67) 0.012** (0.03) −3.347*** (0.00) No Yes 1197 0.144

1.598 (0.36) 3.210*** (0.01) 6.176 (0.19) 1.132** (0.05) 0.156*** (0.00) 0.127* (0.10) −0.141 (0.13) 1.155*** (0.00) 1.430*** (0.00) −0.219* (0.09) 1.606*** (0.00) 0.189*** (0.00) −0.044*** (0.00) 0.383*** (0.00) −0.001 (0.67) 0.011** (0.03) −1.494* (0.11) No Yes 1197 0.184

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

Predicted bonus/total

0.020 (0.45) −0.027 (0.29) −0.011 (0.71) 0.148 (0.74) −0.558 (0.46) 0.103 (0.55) 0.406 (0.13) −0.024 (0.75) 0.004 (0.71) −0.071 (0.70) −0.004 (0.49) 0.004 (0.70) −3.476 (0.47) Yes No

(2)

0.298*** (0.01)

Predicted salary/total

ROA

(1)

0.093 (0.72) 1.414* (0.06)

Predicted equity-based comp. ratio

log(Revenue)

IPO vs. no outcome (3)

455

456

Panel B: first-stage models, industry and year fixed effects

Graduate degree indicator CEO Age CEO Tenure State Level of Education Constant Observations R2

Cash compensation (2)

Equity-based compensation (3)

Salary/total (4)

Bonus/total (5)

RSG/total (6)

Option/total (7)

0.436*** (0.00) 0.016*** (0.00) 0.004** (0.02) -0.022** (0.03) 12.356*** (0.00) 4727 0.215

-0.066*** (0.00) -0.002*** (0.01) 0.002*** (0.00) 0.009*** (0.00) 0.748*** (0.00) 4727 0.211

0.062*** (0.00) 0.002** (0.01) 0.001* (0.06) -0.002* (0.07) 0.138*** (0.00) 4727 0.207

-0.079*** (0.00) -0.002** (0.01) 0.003*** (0.00) 0.008** (0.01) 0.635*** (0.00) 4727 0.214

0.013* (0.06) -0.001** (0.01) 0.001* (0.06) 0.001* (0.07) 0.113*** (0.00) 4727 0.201

0.009** (0.04) 0.001* (0.06) 0.001*** (0.00) -0.002** (0.01) 0.019 (0.22) 4727 0.205

0.053*** (0.00) 0.002** (0.01) 0.002*** (0.00) -0.001** (0.03) 0.119*** (0.00) 4727 0.205

Total compensation (1)

Cash compensation (2)

Equity-based compensation (3)

Salary/total (4)

Bonus/total (5)

RSG/total (6)

Option/total (7)

0.322** (0.02) −0.001 (0.89) 0.048*** (0.00) 0.001 (0.89) 12.843*** (0.00) 1197 0.712

−0.127*** (0.00) 0.002 (0.50) −0.019*** (0.00) 0.001 (0.89) 0.750** (0.04) 1197 0.069

0.115*** (0.00) −0.001 (0.96) 0.014*** (0.00) 0.001 (0.89) 0.072 (0.84) 1197 0.654

−0.114*** (0.00) 0.003 (0.17) −0.006*** (0.01) 0.001 (0.89) 0.501 (0.12) 1197 0.576

0.013* (0.59) −0.001 (0.35) −0.014*** (0.00) 0.001 (0.89) 0.249 (0.25) 1197 0.701

0.014* (0.10) −0.001 (0.20) 0.006*** (0.00) 0.001 (0.89) 0.062 (0.70) 1197 0.725

0.101*** (0.01) 0.001 (0.56) 0.008*** (0.00) 0.001 (0.89) 0.010 (0.98) 1197 0.711

Panel C: first-stage models, firm fixed effects

Graduate degree indicator CEO age CEO tenure State level of education Constant Observations R2

This table shows unordered multinomial logit estimation where the dependent variable is IPO, acquisition, and no outcome. Panel A uses the predicted values of the compensation variables from the first stage shown in Panels B and C. Models 1–3 control for industry and year fixed effects. Models 4–6 control for firm fixed effect. Panels B and C exclude the firms going public on OTC market. All variables are defined in prior tables. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

Total compensation (1)

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

457

Table 6 Propensity score matching. Panel A: no outcome matched firms vs. sale firms No outcome (N = 969)

IPO and acquired (N = 969)

Test of difference in means

$1.292 0.635 0.100 0.022 0.137 0.735 0.159

$1.344 0.515 0.163 0.018 0.189 0.678 0.207

* *** *** ** *** * ***

Total compensation (million) Salary/total Bonus/total RSGs/total Options/total Cash compensation ratio Equity-based compensation ratio Panel B: no outcome matched firms vs. acquired firms

No outcome matched firms (N = 203) Total compensation (million) Salary/total Bonus/total RSGs/total Options/total Cash compensation ratio Equity-based compensation ratio

$1.301 0.621 0.102 0.034 0.091 0.723 0.125

Acquired (N = 203) $1.324 0.567 0.156 0.017 0.136 0.722 0.153

Test of difference in means * *** *** *** *** *

Panel C: no outcome matched firms vs. IPO firms No outcome matched firms (N = 766) Total compensation (million) Salary/total Bonus/total RSGs/total Options/total Cash compensation ratio Equity-based compensation ratio

$1.302 0.640 0.118 0.012 0.120 0.758 0.132

IPO (N = 766) $1.350 0.501 0.166 0.018 0.203 0.667 0.221

Test of difference in means * *** *** *** ** * ***

This table shows the mean and median values of the compensation variables in our sample. All of the no outcome firms are matched using a propensity score matching where we allow replacement and a one to one match. When matching on propensity score, we use size, VC, ROA, and industry indicators. All variables are defined in prior tables. Test of differences reports the two-tailed tests of whether the means are significantly different between each of the relevant samples. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively.

(regression (1)) which is inconsistent with both the motivation and illiquidity hypotheses of CEO compensation. However, results regarding equity-based compensation mirror the previously discussed results based on logistic analyses. Specifically, firms are more likely to be acquired or go public if their CEOs have higher equity-based compensation, particularly option compensation. The results also show that firms are more likely to be acquired or go public if their CEOs receive higher bonus compensation. Hence, the hazard model results further corroborate our earlier findings, indicating that our conclusions regarding option and bonus compensation are robust to a choice of a modeling approach. The results of the hazard model estimation are consistent with the view that the CEO compensation is structured to compensate for higher effort by the CEO due to the sale of the firm. In Table 8, we report results of the competing risk models. These models, used in the mortgage literature, can be used to identify factors associated with, for example, a mortgage prepayment behavior, while taking into account the competing risk of the default on the mortgage (for example, Deng et al., 2000; Ciochetti et al., 2002; and Pennington-Cross, 2008). Hence, the results of a competing risk model are not directly analogous to a multinomial logit model. In our setting, we estimate the competing risk model to identify factors associated with the likelihood of a firm going public, controlling for the competing risk of a firm being acquired. In regression (1), total compensation has a positive and significant coefficient. Results in regressions (2) and (3) regarding equity-based, option, and bonus compensation mirror our previous results. Specifically, firms are more likely to go public if their CEOs have higher equity-based compensation, particularly option compensation, as well as higher bonus compensation. Hence, the competing risk model results further corroborate our earlier findings, indicating that our conclusions regarding the motivation framework of CEO compensation are robust to a choice of a modeling approach. Overall, the multivariate results obtained from alternative modeling approaches are inconsistent with the illiquidity hypothesis of CEO compensation. Instead, the results are in line with the view that the CEO compensation is structured to compensate for higher effort by the CEO due to the sale of the firm.17

17 We collect additional data for CEO compensation following the IPO. We find that the average dollar increase in total compensation is driven by increase in equitybased compensation, with both increases being statistically significant. With respect to the compensation ratio variables, we find that salary and bonus decline significantly while restricted stock grants and option compensation increases significantly. We interpret this evidence as contrary to the illiquidity hypothesis.

458

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

4.5. Valuation at the time of the sale Prior literature documents an “IPO valuation premium puzzle” (Bayar and Chemmanur, 2012), whereby IPO firms enjoy higher valuations at the time of the offering than do acquired firms. If the type and amount of compensation that CEOs receive before selling the firm is related to the effort involved with the process, then the compensation structure should be also related to the payoff received by the shareholders of the selling firm. We therefore analyze the valuations of selling firms and how such valuations are related to CEO compensation structure. Table 9 shows the valuations for target and IPO firms. Due to limited data availability, we are able to use only a subset of firms in our sample. Specifically, we note that the decline in sample size is explained by availability of pricing data, i.e., deal value for acquisitions and IPO price and shares for firms going public. For acquisitions, the loss of 29% of the sample due to the requirement of deal value appears in line with data availability in other databases. The unavailability of IPO pricing data is due to inclusion of OTC firms which account for approximately half of our sample. Panels A and B report the results for unmatched samples. Panel A shows that the acquisition transaction value is $1.3 billion on average, as reported by SDC. The median is lower at $437 million. Following Bayar and Chemmanur (2012) and Poulsen and Stegemoller (2008), we calculate a revenue-based valuation multiple as the transaction value divided by revenue (“Valuation Multiple—Revenue”). Following Poulsen and Stegemoller (2008), we also calculate the valuation multiple for acquired firms as the acquisition transaction value divided by pre-acquisition total assets (“Valuation Multiple—Total Assets” or “asset-based valuation multiple”). The median revenue-based and asset-based valuation multiple for acquired firms is 2.01 and 1.30, respectively. Panel B shows the results for the IPO firms. We follow Bayar and Chemmanur (2012) and calculate the value of the IPO firms as the product between the IPO price and number of shares

Table 7 Survival analysis: probability of sale. (1) log(CEO total compensation)

(2)

Cash compensation ratio

0.544* (0.08) 0.734** (0.02)

Equity-based comp. ratio Salary/total Bonus/total RSG/total Option/total log(Revenue) ROA Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Constant Observations

(3)

0.082 (0.12)

0.083*** (0.01) −0.017 (0.52) −0.019 (0.61) 0.533* (0.06) 0.689 (0.24) −0.305** (0.04) 0.645*** (0.00) 0.090 (0.18) −0.055*** (0.01) 0.297** (0.02) −0.003 (0.49) 0.004 (0.52) −5.957*** (0.00) 1141

0.104*** (0.00) −0.023 (0.39) −0.024 (0.58) 0.547* (0.05) 0.606 (0.30) −0.297* (0.05) 0.632*** (0.00) 0.087 (0.20) −0.054*** (0.01) 0.267** (0.04) −0.003 (0.53) 0.004 (0.56) −5.494*** (0.00) 1141

0.212 (0.51) 1.436*** (0.00) −0.876 (0.28) 0.793** (0.01) 0.078*** (0.01) −0.021 (0.42) −0.020 (0.61) 0.489* (0.08) 0.474 (0.42) −0.250* (0.10) 0.642*** (0.00) 0.083 (0.22) −0.058*** (0.00) 0.250* (0.06) −0.004 (0.40) 0.006 (0.38) −5.418*** (0.00) 1141

This table reports the results of hazard model estimating the probability of sale conditional on the sale not occurring prior to time T. All variables are defined in prior tables. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

459

Table 8 Competing risk analysis: probability of IPO. (1) log(CEO total compensation)

(2)

0.776** [0.995] (0.04)

Cash compensation

0.748* [2.065] (0.06) 0.793*** [1.963] (0.00)

Equity-based compensation

Salary/total

Bonus/total

RSG/total

Option/total

log(Revenue) ROA Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Observations

(3)

0.133*** (0.00) −0.013 (0.74) −0.056 (0.49) 0.832*** (0.01) 0.665 (0.38) −0.281* (0.06) 0.896*** (0.00) 0.133** (0.04) −0.092*** (0.00) 0.506*** (0.00) −0.006 (0.25) 0.004 (0.52) 1141

0.130*** (0.00) −0.013 (0.74) −0.053 (0.50) 0.821*** (0.01) 0.563 (0.47) −0.277* (0.06) 0.899*** (0.00) 0.126* (0.06) −0.092*** (0.00) 0.478*** (0.00) −0.006 (0.27) 0.004 (0.49) 1141

0.435 [1.518] (0.28) 1.463*** [4.133] (0.00) 0.946 [0.365] (0.29) 0.747** [2.060] (0.05) 0.107*** (0.00) −0.015 (0.69) −0.050 (0.52) 0.725** (0.02) 0.325 (0.69) −0.236* (0.11) 0.914*** (0.00) 0.120* (0.09) −0.096*** (0.00) 0.452*** (0.00) −0.008 (0.16) 0.007 (0.25) 1141

This table reports the results of competing risk models estimating the probability of an IPO, controlling for competing risk of being acquired. All variables are defined in prior tables. ***, **, * denote significance at the 1%, 5% and 10% levels, respectively. Odds ratios are reported in brackets below the coefficients for selected independent variables. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

outstanding. The value of IPO firms is on average $927 million. We note that, on average, the IPO firms are significantly smaller than the acquired firms are. The median valuation multiple for IPO firms, calculated as the implied value divided by pre-IPO revenue, is 3.35. The median asset-based valuation multiple is 2.74. The means (medians) of both valuation multiples for IPO firms are higher than the multiples for acquired firms, with p-values smaller than 5%. The result for the unmatched samples indicates that valuations of IPO firms are higher than valuations of acquired firms.18 In Panel C of Table 9, we report the results for a propensity score-match sample of acquired firms. Specifically, as in Bayar and Chemmanur (2012), we note that empirical analysis of IPO valuation premium has to account for self-selection by private firms to either get acquired or go public via IPO. Hence, we use propensity score matching along the following dimensions: log(Revenue), ROA, leverage, total cash/total assets, Capex/Total Assets, whether the firm has negative equity, whether the firm is backed by VCs, industry market-to-book, CEO tenure, whether CEO is the founder, and number of IPOs (acquisitions) in the industry in that year. We allow for 18 To assess the economic significance of the differential valuation, we calculate the intrinsic value of the options held by the CEOs of acquired and IPO firms at the offer price and the IPO price, respectively. Based on untabulated results, we find that the average calculated intrinsic value for the CEOs of acquired and IPO firms is $4.70 million and $7.96 million, respectively. Hence, the average economic benefit to CEOs of IPO firms exceeds the economic benefit to CEOs of acquired firms by $3.26 million.

460

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463 Table 9 Valuation by outcome. Panel A: acquisition

Mean

Median

Obs

Transaction value (million) Valuation Multiple—Revenue Valuation Multiple—Total Assets

$1312.43 6.12 2.85

$437.01 2.01 1.30

141 141 141

Panel B: IPO

Mean

Median

Obs

Value at IPO price (million) Valuation Multiple—Revenue Valuation Multiple—Total Assets

927.47** 12.89** 4.85**

367.58** 3.35** 2.74**

342 342 342

Panel C: propensity score-matched analysis using Valuation Multiple—Revenue

Mean

Median

Obs

Imputed IPO Value of Acquired Firm (million) Imputed IPO Premium (Revenue)

2428.85*** 1.85***

473.14*** 1.08***

141 141

Panel A shows the mean and median values of the acquisition deal terms for the acquired firms in our sample. Transaction value is the total consideration of the acquisition. Valuation multiple is the transaction value/Revenue or total assets. Panel B shows the value for IPO firms, calculated as the IPO price × shares outstanding. The test of difference in means and medians is versus the corresponding variables in Panel A. Panel C reports the results of propensity score-match sample where each acquired firm is matched to an IPO firm. Imputed IPO Value of Acquired Firm is calculated following Bayar and Chemmanur (2012) as the transaction value × the Valuation Multiple—Revenue of the matched IPO firm. Premium is the log(Imputed IPO Value of Acquired Firm/transaction value) as in Bayar and Chemmanur (2012). The test of difference in means and medians for the Imputed IPO Value of Acquired Firms is versus the transaction value in Panel A. For Premium, the test of difference from zero is reported. ***, **, * indicate 1%, 5%, and 10% levels of significance, respectively.

matching with replacement, i.e. a non-treatment (no outcome) firm may be matched to more than one treatment firm, and restrict the matching propensity score to be within 1% of a treatment firm. We first calculate the imputed IPO value of the acquired firms as the transaction value × the revenue multiple of the IPO matched firm (Bayar and Chemmanur, 2012). The results indicate that both the imputed mean and median IPO value of acquired firm exceed the transaction value at 1% level of significance. The imputed IPO premium is calculated following Bayar and Chemmanur (2012) as the log of the imputed IPO value of the acquired firm divided by the transaction value. The median imputed IPO premium for our sample is 108% which is slightly larger than the 75.5% reported by Bayar and Chemmanur (2012) (Table 7, Panel A, All acquisitions). Overall, based on the propensity score-matched sample, we conclude that IPO firms in our sample go public at a premium relative to the acquired firms. We next analyze the “IPO valuation premium puzzle” in a multivariate setting and test whether the IPO valuation premium puzzle is related to the CEO compensation structure. Consistent with Bayar and Chemmanur (2012), we estimate treatment-effect regression models. In the first stage, we estimate a probit regression to analyze the probability of going public via IPO versus getting acquired. In the second stage models, the dependent variable is the log of transaction value for acquired firms and log of value of the IPO firm at the IPO price. In both the first and second stage models we control for these firm-specific characteristics: log(Revenue), ROA, CapEx/Total Assets, leverage, VC indicator, negative equity indicator.19 Since we focus on CEO compensation, we include CEO-specific characteristics related to the CEO's tenure and whether the CEO is a founder. In addition, we also include CEO compensation characteristics in our expanded models. To control for industry and market conditions we also use industry average market-to-book ratio and the number of IPOs and takeovers. As before, each of these variables is measured in the year immediately preceding the event year. In all second stage regressions, we include an indicator equal to one if a firm undertakes an IPO and to zero if it gets acquired as well as the inverse Mills ratio from the first stage model. All regressions include year and industry indicators and use heteroscedasticity consistent standard errors. Table 10 reports the results of the treatment-effect models. The first (second) column for each model reports the results of the first (second) stage model. In the second stage model of regression (1), we note that the coefficient on the IPO indicator is positive and significant at the 1% level. The size of the coefficient implies that IPO firms have multiples that are higher by about 59% (e0.465–1) compared to the valuation of acquired private firms. In regression (2), we include cash and equity-based compensation of the CEO. In the second stage model, the coefficient on the IPO indicator becomes insignificant while the coefficient on equitybased compensation enters with positive coefficient that is significant at the 1% level. The second stage model of regression (3) shows that the significance of the equity-based compensation is mainly due to the option compensation. The IPO indicator remains insignificant in the second stage model of regression (3). However, we note that the size of the IPO indicator coefficient does not substantially decline when comparing the base regression (1) and the extended regression (3) which suggest that controlling for CEO compensation characteristics affects the precision of the IPO indicator not its size. We interpret these results as being consistent with the motivation hypothesis as related to exit pricing and as suggesting that the “IPO valuation premium puzzle” is at least in part explained by the varying CEO compensation structure. As an additional robustness check, we estimate the treatment model wherein the second stage dependent variable is either the revenue or asset based valuation multiple. In untabulated results, we find that before including CEO compensation characteristics, 19 Unlike Bayar and Chemmanur (2012), we do not use sales growth and tangible assets in our models because doing so would result in substantially reduced sample. Furthermore, to be consistent with our prior specifications, we use log(Revenue) in lieu of log(Total Assets) used by Bayar and Chemmanur (2012). When we use log(Total Assets) our conclusions are not affected.

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

461

Table 10 Treatment effect regression analysis of valuation. 1st stage: exit = IPO indicator (1) IMR

2nd stage: log (value)

1st stage: exit = IPO indicator (2)

0.399 (0.44) 0.465** (0.03)

Exit = IPO indicator Cash compensation

0.275 (0.54) 0.225* (0.09)

Equity-based compensation Salary/total

2nd stage: log (value) 0.297 (0.56) 0.469 (0.12) 0.039 (0.92) 0.741*** (0.00)

Bonus/total RSG/total Option/total log(Revenue) ROA Leverage Total cash/total assets CapEx/Total Assets Negative equity indicator VC indicator Industry market-to-book Tenure Founder # IPOs in industry # M&As in industry Constant Observations Pseudo R-squared Adjusted R-squared

0.189*** (0.00) 0.037 (0.46) −0.011 (0.89) 0.781 (0.18) 2.073** (0.03) −0.370* (0.09) 2.068*** (0.00) 0.370*** (0.00) −0.076*** (0.00) 0.742*** (0.00) 0.001 (0.93) 0.005 (0.69) −0.934** (0.03) 483 0.196

0.283*** (0.00) 0.045 (0.58) −0.162 (0.51) 0.069 (0.90) 1.641** (0.01) −0.943*** (0.00) 0.217 (0.40) 0.232** (0.04) −0.023 (0.28) 0.083 (0.69) −0.014* (0.06) −0.022** (0.02) 4.142*** (0.00) 483 0.480

0.190*** (0.00) 0.035 (0.49) −0.014 (0.86) 0.765 (0.19) 2.063** (0.03) −0.365* (0.10) 2.060*** (0.00) 0.368*** (0.00) −0.076*** (0.00) 0.734*** (0.00) 0.001 (0.94) 0.005 (0.68) −1.145* (0.05) 483 0.211

1st stage: exit = IPO indicator (3)

0.272*** (0.00) 0.045 (0.57) −0.186 (0.44) 0.135 (0.80) 1.714*** (0.01) −0.935*** (0.00) 0.203 (0.42) 0.228** (0.05) −0.019 (0.37) 0.075 (0.71) −0.012 (0.12) −0.021** (0.03) 4.845*** (0.00) 483 0.480

2nd stage: log (value) 0.453 (0.35) 0.425 (0.14)

−0.186 (0.68) 0.270* (0.06) 1.075 (0.34) 0.348* (0.08) 0.192*** (0.00) 0.035 (0.50) −0.013 (0.87) 0.747 (0.21) 2.067** (0.03) −0.355* (0.11) 2.059*** (0.00) 0.365*** (0.00) −0.077*** (0.00) 0.746*** (0.00) 0.001 (0.94) 0.006 (0.63) −1.147* (0.05) 483 0.254

1.293*** (0.00) 0.559 (0.29) 0.710 (0.39) 0.159*** (0.00) 0.238*** (0.00) 0.027 (0.73) −0.142 (0.55) 0.223 (0.67) 1.417** (0.03) −0.947*** (0.00) 0.260 (0.29) 0.294*** (0.01) −0.024 (0.26) 0.065 (0.75) −0.010 (0.19) −0.022** (0.02) 4.869*** (0.00) 483 0.472

This table shows the treatment effect regressions where the dependent variable is the log of the total consideration from the acquisition for the acquisition sample and the log of value of the firm at the IPO price for IPO sample. Implied value of the firm at the IPO is estimated following Bayar and Chemmanur (2012) as the pre-IPO shares outstanding × the IPO price. IMR is the inverse Mill's ratio derived from a first stage logit regression modeling the probability of IPO which is shown in columns (1–3). Exit = IPO is an indicator variable that is one if the exit was an IPO and 0 if the exit was an acquisition. We control for year and industry fixed effects. The independent variables are described in prior tables. ***, **, * denote significance at the 1%, 5% and 10% level, respectively. p-Values, reported in parenthesis below the coefficients, are derived using heteroskedasticity consistent standard errors clustered at the firm level.

the IPO indicator enters with a positive coefficient, significant at the 5% level. After including CEO compensation characteristics, the coefficient of the IPO indicator declines slightly and becomes insignificant. Hence, our conclusions hold even when analyzing valuation multiples. Overall, we find that higher CEO salary and option compensation of IPO firms is associated with higher shareholder value at the time of the sale, a result consistent with the motivation hypothesis. The varying CEO compensation is associated with the previously documented “IPO valuation premium puzzle”. Hence, our results indicate that CEO compensation is not only related with the decision to sell and the type of the sale, but is also related with the valuation of the selling firms. 5. Conclusion At some point in the life of a private company, the owners may consider a sale of the firm. The choices available to the owners include taking the company public through an IPO or selling the company to an interested acquirer. We are first to analyze the relation between CEO compensation and the decision of firms to sell.

462

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

We document higher equity-based compensation in the year prior to a private firm's sale as compared to firms that remain private. This result is consistent with both the sellout and IPO firms offering CEOs compensation for higher effort related to the sale of the firm. The results are inconsistent with the view that CEO compensation at private firms reflects the illiquidity of the shares. We also document that CEO compensation structure for selling firms is positively related to the valuation of selling firms and explains the “IPO valuation premium puzzle” documented in the prior literature.

Acknowledgements We thank Jeffry Netter, the editor, and the anonymous referee for helpful comments. We also thank Brian Anderson, Onur Bayar, Sergey Chernenko, Bill Francis, Jean Helwege, Annette Poulsen, John Wald, seminar participants at The Ohio State University, North Carolina State University, The University of Texas at San Antonio, and conference participants at the 2014 Financial Management Association Annual Meeting for helpful comments and discussions. Burns thanks UTSA College of Business for financial support.

Appendix A. Appendix

Variable

Definition

Firm characteristics CapEx/Total Assets Leverage Negative equity indicator

CapEx/Total Assets is the ratio of capital expenditures to total assets Ratio of total debt to total assets An indicator variable that is equal to one if the firm has negative book value of equity and zero otherwise Revenue (million) Revenue is the total reported revenue ROA ROA is the ratio of net income divided by total assets Total assets (million) Total assets is the reported total assets Total cash/total assets Ratio of total cash to total assets Venture backed An indicator variable that is equal to one if the firm has venture capital backing and zero otherwise Compensation and CEO characteristics Bonus/Total Ratio of the bonus over total compensation Cash compensation ratio Ratio of the total salary, bonus, and other cash compensation received by CEOs over total compensation CEO age The age of the CEO CEO ownership Equity-based compensation ratio Exit = IPO Founder Graduate degree indicator Options/Total RSGs/Total Salary/Total State level of education Tenure Total compensation (million)

The percent of shares held by the CEO Ratio of the sum of option and RSG compensation over total compensation Indicator variable that is equal to one if the firm pursued an IPO and zero otherwise An indicator variable that is equal to one if the CEO is the founder and zero otherwise An indicator variable that is equal to one if the CEO has a graduate degree and zero otherwise Ratio of the options over total compensation Ratio of the RSGs over total compensation Ratio of the salary over total compensation Percent of population in firm's headquarters state with graduate degree The tenure of the CEO

Total Compensation is the total CEO compensation, including salary, bonus, other cash pay, option and restricted stock grants Transaction valuation characteristics Value at IPO Price (million) The IPO price × shares outstanding after the IPO Imputed IPO Value of Acquired =Transaction value × (Valuation Multiple Revenue of a matched IPO firm) Firm (million) Imputed IPO Premium =log(Imputed IPO Value of Acquired Firm/transaction value) IPO price The actual price at which shares sold in the IPO Offer share price The per share value of total consideration offered Transaction value (million) The total value of all considerations offered in the acquisition Valuation Multiple—Revenue Either the IPO implied value or the transaction value divided by revenue Valuation Multiple—Total Assets Either the IPO implied value or the transaction value divided by total assets Other controls # IPOs in industry The count of the number of IPOs in the industry in a particular year # M&As in industry The count of the number of M&As in the same 2-digit SIC industry and year Industry market-to-book The average ratio of market value of equity to book value of equity for firms in the same 2-digit SIC industry and year

Source Cap IQ Cap IQ Cap IQ Cap IQ Cap IQ Cap IQ Cap IQ Cap IQ

Cap IQ Cap IQ Cap IQ, BoardEx and hand collected from SEC filings Cap IQ Cap IQ Cap IQ Cap IQ and hand collected from SEC filings Cap IQ, BoardEx, and hand collected from SEC filings Cap IQ Cap IQ Cap IQ Census Cap IQ and hand collected from SEC filings Cap IQ

Cap IQ and SDC

Cap IQ and SDC Cap IQ and SDC Cap IQ and SDC Cap IQ and SDC Cap IQ and SDC SDC SDC Cap IQ

N. Burns et al. / Journal of Corporate Finance 43 (2017) 444–463

463

References Aggarwal, R., Krigman, L., Womack, K., 2002. Strategic IPO underpricing, information momentum, and lockup expiration selling. J. Financ. Econ. 66, 105–137. Bayar, O., Chemmanur, T., 2011. IPO versus acquisitions and the valuation premium puzzle: a theory of exit choice by entrepreneurs and venture capitalists. J. Financ. Quant. Anal. 46, 1755–1793. Bayar, O., Chemmanur, T., 2012. What drives the valuation premium in IPOs versus acquisitions? An empirical analysis. J. Corp. Finan. 18, 451–475. Bengtsson, O., Hand, J., 2011. CEO compensation in venture capital markets. J. Bus. Ventur. 26, 391–411. Boone, A., Mulherin, J., 2007. How are firms sold? J. Financ. 62 (2007), 847–875. Brau, J., Francis, B., Kohers, N., 2003. The choice of IPO versus takeover: empirical evidence. J. Bus. 76, 583–612. Cai, J., Vijh, A., 2007. Incentive effects of stock and option holdings of target and acquirer CEOs. J. Financ. 62, 1891–1933. Cavalluzzo, K., Sankaraguruswamy, S., 2000. Pay-to-accounting Performance and Ownership Structure in Privately-held Small Corporations (Georgetown University working paper). Cavender, J., Rogers, W., Fisher, L., Gersh, B., Coggin, J., Meyers, W., 1992. Effects of smoking on survival and morbidity in patients randomized to medical or surgical therapy in the coronary artery surgery study: 10 year follow up. J. Am. Coll. Cardiol. 20, 287–294. Chemmanur, T., Kong, L., Krishnan, K., Yu, Q., 2015. Top Management Human Capital, Inventory Mobility, and Corporate Innovation (Boston College working paper). Ciochetti, B.A., Deng, Y., Gao, B., Yao, R., 2002. The termination of commercial mortgage contracts through prepayment and default: a proportional hazard approach with competing risks. Real Estate Econ. 30, 595–633. Cole, R., Mehran, H., 2013. What Do We Know About Executive Compensation at Privately Held Firms? FRB of New York Staff Report No. 314 http://dx.doi.org/10.2139/ ssrn.1548750 Available at SSRN: https://ssrn.com/abstract=1548750. Datta, S., Iskandar-Datta, M., Raman, K., 2001. Executive compensation and corporate acquisition decisions. J. Financ. 56, 2299–2336. Deng, Y., Quigley, J.M., Van Order, R., 2000. Mortgage terminations, heterogeneity and the exercise of mortgage options. Econometrica 68, 275–307. Dickerson, A.P., Gibson, H.D., Tsakalotos, E., 2002. Takeover risk and the market for corporate control: the experience of British firms in the 1970s and 1980s. Int. J. Ind. Organ. 20, 1167–1195. Falato, A., Li, Dan, Milbourn, T., 2015. Which skills matter in the market for CEOs? Evidence from pay for CEO credentials. Manag. Sci. 61, 2845–2869. Gao, H., Lemmon, M., Li, K., 2012. Is CEO Pay in U.S. Public Firms Efficient? New Evidence From Private Firms (University of Utah working paper). Gao, H., Harford, J., Li, K., 2013. Determinants of corporate cash policy: insights from private firms. J. Financ. Econ. 109, 623–639. Harford, J., 2005. What drives merger waves? J. Financ. Econ. 77, 529–560. Hansen, L., 1982. Large sample properties of generalized method of moments estimators. Econometrica 50, 1029–1054. Helwege, J., Packer, F., 2009. Private matters. J. Financ. Intermed. 18, 362–383. Helwege, J., Pirinsky, C., Stulz, R., 2007. Why do firms become widely held? An analysis of the dynamics of corporate ownership. J. Financ. 62, 905–951. Hsieh, J., Lyandres, E., Zhdanov, A., 2011. A theory of merger-driven IPOs. J. Financ. Quant. Anal. 46, 1367–1405. Kummer, C., Steger, U., 2008. Why merger and acquisition (M&A) waves reoccur: the vicious circle from pressure to failure. Strategic Management Rev. 2, pp. 44–63 Minnick, K., Raman, K., 2016. Board composition and relationship-specific investments by customers and suppliers. Financ. Manag. (forthcoming). Minnick, K., Unal, H., Yang, L., 2011. Pay for performance? CEO compensation and acquirer returns in BHCs. Rev. Financ. Stud. 24, 439–472. Officer, M., 2007. The price of corporate liquidity: acquisition discounts for unlisted targets. J. Financ. Econ. 83, 571–598. Palepu, K., 1986. Predicting takeover targets: a methodological and empirical analysis. J. Account. Econ. 8, 3–36. Palia, D., 2001. The endogeneity of managerial compensation and firm valuation: a solution. Rev. Financ. Stud. 14, 735–764. Pagano, M., 1993. The flotation of companies on the stock market: a coordination failure model. Eur. Econ. Rev. 37, 1101–1125. Pennington-Cross, A., 2008. The duration of foreclosures in the subprime mortgage market: a competing risks model with mixing. J. Real Estate Financ. Econ. 40, 109–129. Perevozchikov, V., 2010. Why Companies Go Public in the Ukraine (National University working paper). Poulsen, A., Stegemoller, M., 2008. Moving from private to public ownership: selling out to public firms vs. initial public offerings. Financ. Manag. 37, 81–101. Shen, J., Reuer, J., 2005. Adverse selection in acquisitions of small manufacturing firms: a comparison of private and public targets. Small Bus. Econ. 24, 393–407.