What s in a vote? The short- and long-run impact of dual-class equity on IPO firm values $

Transcription

1 Journal of Accounting and Economics 45 (2008) What s in a vote? The short- and long-run impact of dual-class equity on IPO firm values $ Scott B. Smart a, Ramabhadran S. Thirumalai b, Chad J. Zutter c, a Kelley School of Business, Indiana University, Bloomington, IN 47405, USA b Indian School of Business, Gachibowli, Hyderabad , India c Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260, USA Received 26 September 2006; received in revised form 12 June 2007; accepted 16 July 2007 Available online 22 July 2007 Abstract We find that relative to fundamentals, dual-class firms trade at lower prices than do single-class firms, both at the IPO and for at least the subsequent 5 years. The lower prices attached to duals do not foreshadow abnormally low stock or accounting returns. Moreover, some types of CEO turnover are less frequent among duals, and in general CEO turnover is sensitive to firm performance for singles but not for duals. Finally, when duals unify their share classes, statistically and economically significant value gains occur. Collectively, our results suggest that the governance associated with dual-class equity influences the pricing of duals. r 2007 Elsevier B.V. All rights reserved. JEL classification: G12; G14; G30; G32; G34 Keywords: Initial public offerings; Dual class; Ownership structure; Governance; Firm value; CEO turnover 1. Introduction Prompted by the wave of corporate scandals around the turn of the century, Congress passed the Sarbanes Oxley (SOX) Act of This legislation, ostensibly designed to protect investors, contains provisions which presuppose that firms governance practices affect shareholder value. Several research papers published in the wake of SOX look for possible connections between governance and value, but no professional consensus exists regarding whether or how governance and value are linked. In this paper, we contribute to the debate by comparing several attributes of dual- and single-class firms, following them from their IPO dates forward for 5 years. We focus on IPO firms because it is prior to the IPO that firms establish $ For valuable comments, we thank Utpal Bhattacharya, Amy Dittmar, Robert Dittmar, Craig Holden, Sreenivas Kamma, William Megginson, Frederik Schlingemann, Shawn Thomas, Charles Trzcinka, Greg Udell, and presentation participants at the Haub School of Business Saint Joseph s University, the Kelley School of Business Indiana University, and The Wharton School University of Pennsylvania. We also thank Anil Shivdasani (the Referee) and S.P. Kothari (the Editor) for suggestions that greatly improved the paper. Corresponding author. Tel.: ; fax: address: czutter@pitt.edu (C.J. Zutter) /$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi: /j.jacceco

2 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) the governance rules by which they will abide as public companies. Subsequent changes to governance policies, such as the adoption of poison pills, may be precipitated by imminent takeover threats, making it more difficult to disentangle any relation between governance provisions and firm value. Also, by focusing exclusively on newly public dual-class firms, we avoid issues that arise when firms switch from single-class to dual-class status through a recapitalization plan, which may be tied to other simultaneous value-relevant events. 1 Our research design emphasizes comparisons between dual-class and single-class firms because dualclass equity is a particularly effective method by which managers may entrench themselves. By separating cash flow rights from voting rights, dual-class equity enables corporate insiders to exert voting control over a firm in which they may hold a comparatively small economic interest. Hence, dual-class shares potentially exacerbate problems associated with the separation of ownership and control. The importance of dual-class equity in U.S. markets appears to be rising as new firms adopt this particular ownership structure. Smart and Zutter (2003) report that from January 1990 to May 1994 about 7 percent of all U.S. IPOs adopt dual-class equity structures and that these firms account for about 11 percent of the aggregate market capitalization of IPO firms. However, from June 1994 to October 1998, these figures increase significantly. In the latter period, almost 12 percent of all IPOs were dual-class deals accounting for about 31 percent of IPO market capitalization. The Finish Line Inc. provides a prototypical example of a dual-class IPO. Holders of the Finish Line s Class A shares, originally offered to the public through a 1992 IPO, receive one vote per share, while the Finish Line s Class B shareholders control ten votes per share. All other rights of the two share classes are identical. After its IPO, The Finish Line sold seasoned equity (Class A shares only) on three occasions. With each subsequent equity sale, insider ownership steadily decreased, but the Class B shares enabled insiders to retain voting control. Recent Securities and Exchange Commission (SEC) filings reveal that The Finish Line s three founders collectively own just 205,934 of the firm s 21,208,451 Class A shares, yet they own all 2,865,284 Class B shares outstanding. Combining their holdings of both classes, the founders claim on the firm s cash flows barely reaches 12.8 percent, but they still control almost 57.9 percent of the votes that can be cast on any matter requiring shareholder approval. Does the wedge between insiders voting rights and cash-flow claims that dual-class equity facilitates influence the pricing of dual-class equity at the IPO date and beyond? 2 Some of the largest and most vocal institutional investors subscribe to the view that dual-class stock arrangements entrench managers at shareholders expense. For example, in its proxy voting guidelines, the California Public Employees Retirement System (CalPERS) says that it votes against any proposal to create unequal voting rights across share classes. In response to Google s announcement in 2004 that it would go public with dual-class equity, a high-ranking official at TIAA-CREF remarked that Google s shares should be priced at a substantial discount and that the dual-class stock effectively disenfranchises outside shareholders. 3 More recently, the hedge fund Clinton Group, which owns just over 5 percent of the Finish Line s Class A shares, wrote a public letter asking the firm to create shareholder value by eliminating its dualclass ownership structure, repurchasing shares, or going private. Determining whether investors apply a discount to the shares of dual-class firms or more generally to the shares of firms that adopt governance provisions that strengthen the position of corporate insiders remains a challenging empirical question. There are several channels through which dual-class governance structures could influence firm values. For example, Francis et al. (2005) find that the credibility of earnings information released by dual-class firms is inferior to that provided by single-class firms. Perhaps as a consequence, they also find that the salience of dividends as a performance measure is greater in dual-class firms. In a study of firms in East Asian countries, Fan and Wong (2002) find that earnings informativeness declines as the gap between insiders cash flow rights and voting rights widens. 1 Current NYSE and NASDAQ listing requirements make it exceedingly difficult for existing public companies with a single share class to recapitalize with multiple share classes. As a practical matter, therefore, it is only at the IPO stage when a firm can enact a dual-class structure. 2 DeAngelo and DeAngelo (1985) report that insiders control a median of 56.9 percent of the votes while holding just 24.0 percent of the cash flow rights in dual-class firms. 3 See London (2004). Some high-profile regulators also subscribe to this view. Charlie McCreevy, the internal market commissioner of the European Union, said in a Financial Times interview, It is my goal to get the one-share, one-vote principle accepted across the 25 member states. See Buck (2005).

3 96 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Exploiting a unique feature of the banking industry which gives managers the authority to vote shares held in trust, Adams and Santos (2006) explore the performance consequences of the wedge between managers cash flow rights and voting rights. Using both Tobin s Q and return on assets as operating performance measures, they find that performance is positively related to managerial voting control for small stakes, but control has negative consequences for performance when managers control a large fraction of votes. In a recent influential paper, Gompers et al. (2003) measure the protections afforded shareholders through firms corporate governance structures. They construct a corporate governance index (dubbed simply, G) and find that firms with poor governance index scores underperform the market and have lower Tobin s Q values than firms with better governance scores. 4 However, Core et al. (2006) find no evidence that poorly governed firms as defined by Gompers et al. achieve unexpectedly poor operating results. 5 Furthermore, they claim that the sub-par stock performance of badly-governed firms eventually reverses. Core et al. conclude that their evidence does not support the hypothesis that bad governance leads to low stock returns and they propose that the anomalous returns reported in Gompers et al. may simply be a manifestation of the new economy pricing puzzle of the (late) 1990s. In this paper, we conduct a variety of tests to build a case that dual-class equity is associated with lower share prices. Our findings are consistent with Core et al. in the sense that although we find that dual-class firms trade at relatively low price levels, returns on dual-class shares do not fall below standard benchmarks. In other words, the market efficiently prices dual-class companies. In Section 2 we provide a literature review. In Section 3 we describe our data and provide descriptive statistics. In Section 4 we compare the market prices of dual- and single-class companies, starting at the IPO date and continuing for the next 5 years. The data show that firms choosing voting-right structures that favor management face a significant and persistent valuation discount in the market, even after controlling for the endogenous choice to go public with dual-class equity. This valuation gap persists while controlling for cross-sectional differences in industry valuation multiples as well as firm-specific attributes including growth, profitability, and leverage. In this regard, our results align with those of Gompers et al. (2003) and other studies reporting a valuation discount for firms with governance mechanisms which favor insiders over shareholders. In contrast to the results in Gompers et al. (2003), we find no connection between governance and mispricing. In Section 5 we evaluate the long-run stock returns realized by dual- and single-class IPOs and find no evidence that dual-class firms underperform. This result is obtained using Fama French Carhart fourfactor pricing regressions. By this standard, the market s valuation of dual-class IPOs is efficient. If the difference in valuation multiples between duals and singles does not reflect a pricing error, what does it reflect? One possibility is that poorly performing firms adopt governance systems to protect incumbents. In this case, we expect to see low operating performance from poorly governed firms. We also examine this possibility in Section 5, reporting the results from tests for differences in operating performance between single- and dual-class firms. As with stock returns, there is at best only scant evidence suggesting that duals exhibit abnormally low operating performance. We acknowledge, however, that operating performance measures are notoriously noisy and can lead to tests with insufficient power to reject the null. Although singles and duals generate similar accounting returns over time, a valuation gap between singles and duals could still be tied to operating performance if dual-class firms are riskier than singles. That is, holding expected cash flows constant, dual-class firms would trade at lower multiples if they carry more systematic risk than singles. The asset pricing literature does not give us a universally accepted method for evaluating risk differences between portfolios of stocks. Nevertheless, we can offer one observation suggesting 4 Gompers et al. (2003) calculate G by simply adding up the number of provisions that reduce shareholders rights. Therefore, a higher G corresponds to worse corporate governance. Whether each of the component provisions of G has a meaningful impact on either the degree to which managers are entrenched or on firm value is an open question. Comment and Schwert (1995), for example, argue that poison pills do very little to deter takeovers and have negligible wealth effects in most cases. It is interesting to note that Gompers et al. (2006) report that dual-class firms on average have lower G index scores than do single-class firms. Presumably this is because dual-class equity is a rather extreme form of entrenchment and firms adopting this structure do not need to adopt many other provisions to protect insiders interests. 5 Eldenburg and Krishnan (2003) examine the operating performance results of private versus public hospitals. They find that public hospitals underpay chief executives and display weaker operating results compared to private hospitals. They conclude that weak governance leads directly to lower performance.

4 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) that the valuation discount applied to duals does not reflect a higher systematic risk. In Fama French style pricing regressions, we observe no systematic tendency for the factor betas for dual-class firms to exceed those of singles. This fact leads us to the conclusion that dual-class firms are no more risky than their single-class counterparts. Lacking evidence linking differences in operating performance to the valuation gap between singles and duals, we turn to a governance explanation. Section 6 examines differences in CEO turnover events for singles and duals. We first group all turnover events into two categories. External turnover refers to an event when the CEO departs because the firm is acquired. All other cases of turnover we designate as internal turnover. In the sample that includes both types of turnover events three interesting and suggestive findings emerge. First, the incidence of CEO turnover is slightly lower for duals than for singles, consistent with the hypothesis that dual-class shares entrench incumbents. Second, poor accounting performance precedes turnover events for single-class firms but not for duals. Third, stock returns surrounding turnover events vary between singles and duals. For single-class firms, negative abnormal returns precede internal turnover events, but not external turnover events. This suggests that internal governance mechanisms for single-class firms work well enough to remove the CEO of an underperforming firm without intervention from the market for corporate control. However, for dual-class firms the pattern reverses. Negative abnormal returns precede external turnover events, but no correlation exists between internal turnover and prior returns. We refine our turnover analysis by focusing exclusively on the set of internal events and classifying them as either forced or unforced. As in the larger sample, negative accounting performance precedes turnover events, both forced and unforced, for single-class firms, but not for dual-class firms. Collectively, these results suggest that the dual-class structure substantially weakens the link between an executive s tenure on the job and the performance of the company. Our final test, presented in Section 7, focuses on firms that unwind their dual-class voting structures. We track our sample of dual-class firms during the 5 years following their IPOs and find that 37 eventually unify their share classes. Conducting an event study around the effective date of these unifications, we find positive abnormal returns on the order of 5 6 percent using either the market model or Fama French model. Because unifications often occur gradually as insiders divest their Class B shares, and because firms typically make no formal announcement regarding their plans to unify shares, our event study estimates likely understate the true value increase associated with share unifications. 2. Literature on governance and firm value Empirical research on the link between governance and firm value has a long history. Much of the research in this area uses event-study methods to determine the short-term impact of changes in firms governance practices on share prices. For example, dozens of studies use these methods to study the wealth effects of the adoption of anti-takeover amendments, the passage of state anti-takeover laws, and changes in the composition of corporate boards. 6 Some of the early work on dual-class firms fills a niche in this literature. While our paper focuses on firms that adopt a dual-class structure at the IPO, some firms that currently have dual-class voting arrangements created them in a recapitalization transaction. Partch (1987), Jarrell and Poulsen (1988), and Millon-Cornett and Vetsuypens (1989) all offer event study evidence on dual-class recapitalizations, but they reach no consensus on whether recapitalizations help, harm, or have no impact on shareholder wealth. One reasonable conclusion emerging from these papers is that a sample of firms that recapitalize from a single-class to a dualclass structure does not provide ideal conditions for an analysis of the effects of dual-class equity on firm value. For example, takeover activity concomitant with the recapitalization clouds the interpretation of the announcement effect in the market. This ambiguity in the empirical evidence mirrors conflicting predictions from theory. For example some authors conjecture that dual-class voting arrangements are less than optimal, 6 For event-study evidence, see DeAngelo and Rice (1983), Linn and McConnell (1983), Jarrell and Poulsen (1987), McWilliams (1990), Cotter et al. (1997), Shivdasani and Yermack (1999), Fich and Shivdasani (2005), Perry and Shivdasani (2005), Choi et al. (2007), Dahya and McConnell (2007), Faleye (2007), Paul (2007), and Shivdasani (2006).

5 98 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) while others argue that they can serve to maximize firm value by solving various types of underinvestment problems. 7 Cox and Roden (2002) examine the relative prices of high-vote and low-vote shares in U.S. dual-class firms. They find that the voting premium on high-vote shares can be reduced when holders of low-vote shares receive higher dividend payments. However, for the vast majority of dual-class firms in the U.S. the cash flow rights are identical for the two (or more) share classes and only one share class trades in the open market. This is always the case for firms that go public with dual-class shares. Furthermore, evidence of a voting premium between the two publicly traded share classes speaks to the distribution of firm value between different shareholder groups, but it does not imply that dual-class structures reduce firm value on the whole. Though we supplement our primary results with event-study evidence, our main emphasis is on the long-run impact of dual-class stock on valuation ratios such as the earnings-to-price (E/P) and EBITDA-to-price ratios. Our approach aligns closely with that of Yermack (1996), Gompers et al. (2003), Bebchuk and Cohen (2005), and Gompers et al. (2006). In a study designed to assess the relation between the size and effectiveness of corporate boards, Yermack reports that firms with smaller boards have higher Q values in the period Yermack also finds that smaller boards are more likely to dismiss CEOs following a period of poor performance and that the sensitivity of CEO turnover to performance is greater among firms with small boards. We report similar results when we compare circumstances surrounding CEO turnover among dualclass versus single-class firms. Bebchuk and Cohen (2005) focus on firms with staggered boards, arguing that a staggered board provides incumbents with a high degree of insulation from outside monitoring. They find that firms with staggered boards have significantly lower Tobin s Q values. Bebchuk and Cohen argue that staggered boards represent, ythe key arrangement that protects incumbents from removal in U.S. publicly traded companies. However, they exclude dual-class firms from their analysis, commenting that, yin such firms, the holding of superior voting rights is likely to be the key for entrenching incumbents. In a similar vein, Gompers et al. (2006) examine the relation between firm value, cash flow rights, and voting rights. 8 They acknowledge that financial economists widely agree that insider stock ownership can have both positive and negative effects on firm value. 9 Their most interesting result is that the value of dual-class firms depends positively on insiders cash flow rights and negatively on insiders voting rights. 10 Our analysis differs from that in extant literature in several important ways. First, Gompers et al. (2003) as well as Bebchuk and Cohen (2005) exclude dual-class firms from their analysis and focus instead on crosssectional variation in governance provisions among single-class firms. Although dual-class firms constitute less than 10 percent of the universe of firms covered by the Investor Responsibility Research Center (IRRC), the primary source of data on firm-specific governance characteristics, in the late 1990s the percentage of new firms coming to market with dual-class equity was greater than 10 percent. Because they are typically larger than single-class IPOs, dual-class companies account for a larger percentage of aggregate equity value than their numbers alone indicate. Second, while our paper offers evidence on the relation between dual-class equity and firm value in the spirit of Gompers et al. (2006), our analysis departs from theirs in several interesting and important ways. Their sample includes a broader cross-section of dual-class firms, but our analysis uses a longer time series. Moreover, their sample includes much older and larger dual-class firms than the IPO firms which we study See DeAngelo and DeAngelo (1985), Fischel (1987), Ruback (1988), Grossman and Hart (1988), Harris and Raviv (1988), and Denis and Denis (1994) for very different theoretical perspectives on the merits of dual-class equity. 8 See Villalonga and Amit (2006) for a similar analysis of large, family-owned U.S. corporations. 9 See Stulz (1988) for theoretical work on the offsetting effects of managerial ownership. Morck et al. (1988) and McConnell and Servaes (1990) offer empirical evidence that higher insider ownership at first increases, then decreases firm value. 10 Evidence that the separation of cash flow and voting rights leads to lower firm value is not restricted to U.S. markets. In a study of emerging market firms, Lins (2003) finds lower firm value when managers voting ownership exceeds their cash flow ownership. Similarly, in a study of firms in Asian countries, Claessens et al. (2002) report lower firm value when the largest shareholder s voting ownership exceeds their cash flow ownership. The legal environments and the protections they offer investors vary dramatically across countries, so it is not clear whether the results of these studies can be extended to the United States. 11 They report a mean (median) age, as of 2000, for dual-class firms in their sample of (7.21) years. All of the dual-class firms in our sample went public between 1990 and 1998, so by the end of 2000, the maximum age of a dual-class firm in our study is just 11 years. The definition of age here is years since the CRSP listing date.

6 This difference in sampling choices allows us to address an interesting question does the market apply a discount to dual-class IPOs from their inception, or does the discount emerge gradually over time? Immediately after the IPO, it is common for insiders to own a large fraction of the outstanding shares in both single-class and dual-class companies. However, as firms grow and return to the market to raise equity through seasoned offerings, the voting power of single-class insiders declines at the same rate as their cash flow rights, while dual-class insiders voting rights change at a much slower rate than their economic ownership does. Megginson et al. (2007) report that more than 40 percent of IPO firms in the 1990s issued seasoned equity within 5 years of the IPO date, and the probability of an IPO firm conducting a seasoned offering is much higher if that firm is a dual-class company. Therefore, over time the gap between the economic incentives and voting power of dual-class firms widens. Our evidence suggests that investors discount dual-class equity starting at the IPO date and that the discount is persistent for at least 5 years following the IPO. Third, we also provide two important previously undocumented pieces of evidence supporting the hypothesis that investors discount dual-class firms. Similar to the evidence on large versus small corporate boards presented by Yermack (1996), we find that CEO turnover in single-class firms follows a period of subpar accounting performance, a pattern that dual-class turnover events do not mimic. Intuitively, one might guess that the valuation discount we document exists because investors expect dual-class managers to be firmly entrenched. Using a broad definition of CEO turnover, we find that turnover events are slightly less common among dual-class firms compared to single-class companies. We also show that the sign and magnitude of abnormal returns surrounding turnover events depend on the type of turnover (e.g., acquisition related or not) and whether the firm is a single or a dual. 3. Data Our primary data source is the Disclosure New Issues database from Disclosure Inc. The data set provides a wealth of information about debt and equity offerings on an issue-by-issue basis for any original or subsequent registration or prospectus filed with the SEC. Compared to another well-known IPO data provider, Securities Data Corp., Disclosure under-samples small IPOs, though it does not exclude them. Types of firms excluded from our data set include closed-end funds, unit offers, investment companies, real-estate investment trusts, and limited partnerships. We extract issues from Disclosure by selecting records for firm-commitment IPOs of common stock from 1990 through We stop collecting data in 1998 because we want to track each IPO firm for several years after the IPO date. The search yields 3,628 issues. We eliminate duplicate records, reducing our sample to 2,787 issues. We further eliminate 165 issues because we cannot match data with Standard and Poors COMPUSTAT. Our final sample covers 2,622 IPOs (including 253 dual-class issues) with offer prices ranging from $5 to $35. Table 1 presents some descriptive statistics for our sample. Dual-class firms raise more money at the IPO and have a higher average market capitalization than single-class firms. Duals are less likely to use venture financing, but more likely to employ a high-reputation investment bank. Dual-class IPOs are significantly more likely to be equity carveouts or firms that anticipate paying dividends than singles, but not reverse LBOs or firms with anti-takeover provisions. Finally, duals list less often on NASDAQ, list fewer uses for their IPO proceeds, and have higher institutional ownership following the IPO than do single-class IPO firms. In the next section we control for these characteristics when examining the market s pricing of duals and singles. 4. Market pricing of dual- and single-class IPOs S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Our primary interest is comparing the market values of singles and duals. In this section we examine the market s pricing of dual- and single-class IPOs by testing for systematic differences in valuation ratios across the different firm types. We offer two measures: the E/P ratio and the EBITDA-to-price ratio. 12 If the market 12 In both ratios a measure of market value appears in the denominator, so a higher market value, relative to fundamentals, results in a lower ratio.

7 100 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Table 1 Descriptive statistics Means Dual class Single class P-value Number 253 2,369 Offer value in millions $ $ *** Market capitalization in millions $ $ *** Venture-backed *** High-reputation I-bank *** Equity-carveout ** Reverse-LBO Anti-takeover provisions No dividends anticipated ** NASDAQ-listed *** Number of uses of proceeds *** Lagged market return Fraction of institutional ownership *** Number of institutional owners *** This table presents sample means by offer type. Offer value is the total number of shares offered times the CPI-adjusted final offer price. Market capitalization is the total number of shares outstanding after the offering times the CPI-adjusted first-day closing price. Indicator variables are set equal to one, respectively, for venture-backed, high-reputation I-bank, equity-carveout, reverse-lbo, anti-takeover provisions, and NASDAQ-listed deals. Number of uses of proceeds is the number of uses of proceeds listed in the final prospectus. Lagged market return is the compounded daily CRSP value-weighted return over the 22 trading days preceding the initial public offering. Institutional ownership is end-of-quarter 13f institutional ownership of publicly traded shares for the quarter in which the IPO took place. P-values refer to tests of equal means across offer types. Respectively, ***, **, and * denote significant difference at 1, 5, and 10 percent. imposes a cost on companies that insulate managers through dual-class ownership, then we expect to find that these companies trade at lower prices, relative to earnings or EBITDA, than firms with the more typical oneshare, one-vote structure and we expect this valuation discount to persist Univariate analysis We begin by calculating the E/P and EBITDA-to-price ratios. In the ratio using earnings, we divide by price because the more conventional approach of putting earnings in the denominator results in a highly skewed distribution. 13 For example, many IPO firms have very low earnings, so the distribution of price-to-earnings (P/E) ratios has a very long right tail. For consistency in reporting and interpretation, we divide EBITDA by price. If the market believes that dual-class equity entrenches insiders to the detriment of outside shareholders then dual-class firms should have higher E/P and EBITDA-to-price ratios relative to singles. In Table 2, we calculate mean and median values for each of these ratios at the time of the IPO and at 1-year intervals for the first 5 years after the IPO. In each year, we collect data at the fiscal year end to calculate ratios. For example, each year we take the earnings and market price of the stock at the end of the fiscal year to calculate the E/P ratio. Time zero in our tests refers to the IPO year, not the IPO date. In other words, the E/P ratio at time zero is the first fiscal year end after the IPO. In Table 2, the mean (median) E/P value for dual-class firms in the initial year is (0.046), and for singles the mean (median) value is (0.040). If we invert the means to obtain P/E ratios we obtain for duals and for singles, a difference of $1.96 per dollar of earnings. Correspondingly, the mean (median) EBITDA-to-price value for dual-class firms at the IPO date is (0.119), and for singles the mean (median) value is (0.095). In addition to the mean and median values for each ratio for the 5 years following the IPO year, Table 2 also shows whether the difference between singles and duals is significant for each year. For the E/P ratio, the difference in means is significant every year, and the difference in medians is significant in three out of 6 years. 13 We also exclude firms with negative earnings, though including these firms does not change our fundamental conclusions.

8 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Table 2 Pricing multipliers Event year Full sample Dual class Single class P-value Earnings-to-price means Year * Year ** Year *** Year *** Year *** Year * EBITDA-to-price means Year *** Year ** Year *** Year *** Year * Year ** Earnings-to-price medians Year Year Year *** Year *** Year * Year EBITDA-to-price medians Year *** Year *** Year *** Year *** Year * Year ** This table presents mean and median ratio values by event year and offer type. E/P is earnings per share divided by price per share. EBITDA-to-price is earnings before interest, tax, depreciation, and amortization per share divided by price per share. Year 0 refers to the first fiscal year end following the IPO, Year 1 to the second fiscal year end following the IPO, and so on. P-values refer to tests of equal means or medians across offer types. Respectively, ***, **, and * denote significant difference at 1, 5, and 10 percent. For the EBITDA-to-price ratio, the differences in means and medians are significant in every year, with the values for dual-class firms consistently exceeding those of singles. Thus, based on yearly average pricing multiples, dual-class firms have lower relative valuations than do single-class companies and the valuation gap persists for at least 5 years after the IPO. These differences are economically significant as well. For example, if we invert the E/P ratio to the more familiar P/E ratio, the median dual-class P/E equals 25 for singles and 21.7 for duals, a difference of 15.2 percent. The median time zero market capitalization for a dual-class firm equals $193 million, so the difference in P/E ratios implies an economically significant dollar difference of about $29.4 million. Pooling results across all years, the median E/P ratio equals for duals and for singles, which translates into P/E ratios of 17.2 for duals and 20.4 for singles, a difference exceeding 18 percent. Though the dual-class firms in our sample come from a wide range of industries, it is possible that a disproportionate share of them compete in industries where we might expect low valuation ratios. Similarly, we know that dual-class firms tend to be larger than their single-class counterparts, so perhaps the valuation differences we are seeing merely represent differences in other characteristics and are not tied to the dual-class structure per se. In the next section, we develop an empirical methodology that controls for the endogenous decision to issue dual-class equity and controls for various firm-specific characteristics to address these concerns.

9 102 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Multivariate analysis To establish a connection between the valuation differences between singles and duals and the dual-class voting structure, we must account for the endogenous aspect of the decision that firms make when choosing to go public with dual- or single-class equity as well as control for other firm characteristics that can affect market values. In this section we describe a two-stage estimation process, which first models the decision to adopt dual-class shares and then estimates the impact of that decision on firm value at the IPO date and beyond. To address the endogeneity issue, we first estimate a probit model where the dependent variable equals one for dual-class firms and zero for single-class firms. On the right-hand side are variables which may influence firms decisions to adopt dual-class equity. Included in the list of control variables are the natural logarithm of the IPO offer size and dummy variables equal to one for IPO firms that receive venture backing, use a highreputation investment bank, are equity carveouts, are reverse LBOs, have anti-takeover devices in their charters, do not anticipate paying dividends, and list on NASDAQ. We also control for the number of uses of proceeds listed in the IPO prospectus and the market return leading up to the IPO date. We use this model to estimate the probability of going public with dual-class shares for each firm in our sample. Next we use the fitted probability in our second stage regressions on valuation multiples. Our reasons for selecting the variables included in the probit are as follows. Holding all else constant, firms that want to raise more money in the IPO have to sell a larger fraction to outside investors, which implies a greater loss of control. The dual-class structure may have more appeal to insiders who value control, but whose firms have high financing requirements. It is well established in the literature that venture capitalists (VCs) not only provide financing to entrepreneurial firms, but also exercise some pre-ipo control through board seats and participation in the selection of top management. However, the IPO presents VCs with an exit opportunity, and presumably they have incentives to maximize the value of their claims at exit. Doing so enhances the returns that VCs provide to their clients and makes it easier to raise additional capital from institutional investors to fund new investments. Given the concerns expressed by institutions about investing in dual-class IPOs, we expect the presence of a VC investor to discourage the choice of this particular ownership form. To the extent that selling dual-class equity in an IPO is more difficult than selling single-class shares, firms that nevertheless want to go public with multiple share classes may need the additional certification of a high reputation investment banker to ensure the success of the offering. Consequently, we include in the probit a measure of the underwriter s reputation. An investment bank is considered to have a high reputation if its market share of dollars raised is greater than the median bank s market share for the sample. A positive coefficient on this variable is consistent with our conjecture about the need for investment banker certification to market dual-class equity to investors. Zenner et al. (2005) report that tax laws create an incentive for parent companies conducting equity carveouts to do so using dual-class equity. Accordingly, we include a dummy variable equal to one for equity carveouts in our sample. We also include a dummy for reverse LBOs. We do so because when a public firm goes private, control is typically concentrated in the hands of LBO investors and top management. Having concentrated control of the firm, we conjecture that LBO insiders may be more likely than others to attempt to maintain that control by going public as a dual-class entity. Of course, insiders have multiple devices at their disposal that help them maintain control, including anti-takeover provisions such as staggered boards, poison pills, etc. Viewing these devices as substitutes for dual-class equity, we include a dummy equal to one when firms have at least one anti-takeover measure in their corporate charter other than dual-class shares. If insiders place a high value on control, then why do they choose to go public at all? One reason is that over time, a private firm simply becomes too large to manage with private sources of financing. Manager owners of large, mature firms may reach a point at which access to the public equity markets becomes necessary, and for these firms dual-class equity offers a way to continue to exercise majority voting control. To control for the maturity of firms, we include a dummy variable equal to one for those firms who indicate in their prospectus that they do not intend to pay dividends following the IPO. We also use a dummy variable to distinguish between NASDAQ and NYSE/AMEX listed firms.

10 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Finally, we include several variables designed to capture time series and cross-sectional variation in the incidence of dual-class IPOs. These include a lagged market return, IPO year dummies, and SIC code dummies. Table 3, Panel A reports our probit estimates. As expected, larger firms and those going public with more reputable underwriters show a higher likelihood of using dual-class equity, while firms with venture backing are less likely to do so. None of the other variables in the model (with the exception of some of the SIC Table 3 Two-stage regressions of pricing multipliers Panel A. Dual-class IPO first-stage probit regression Estimate Std. error Lower limit Upper limit w 2 P-value Intercept 2.045*** LN offer value in millions 0.163*** Venture-backed 0.420*** High-reputation I-bank 0.319*** Equity-carveout Reverse-LBO Anti-takeover provisions No dividends anticipated NASDAQ-listed Number of uses of proceeds Lagged market return Log likelihood Number of observations 2,555 Panel B. Earnings-to-price second-stage regressions Year 0 Year 1 Year 2 Year 3 Year 4 Year 5 Intercept 0.079*** 0.087*** 0.096*** 0.104*** 0.096*** 0.095*** Dual-class fitted value 0.018** 0.027*** 0.039*** 0.053*** 0.070*** 0.063*** LN mkt. Cap. In millions 0.010*** 0.011*** 0.013*** 0.014*** 0.012*** 0.012*** Leverage 0.026*** 0.027*** 0.025*** 0.022*** 0.025*** 0.028*** R&D over total assets 0.062*** 0.068*** 0.074*** 0.077*** *** EBITDA-to-assets 0.072*** 0.079*** 0.061*** 0.041*** 0.033** 0.059*** Two-year sales growth 0.019*** 0.012*** * CAPEX-to-sales 0.011*** 0.004* Industry earnings-to-price 0.001* 0.001* Dividend paying firm 0.027*** 0.021*** 0.024*** 0.020*** 0.010** 0.012** S&P 500 firm Adj. R 2 (%) Number of observations 1,502 1,436 1,313 1, Panel C. EBITDA-to-price second-stage regressions Year 0 Year 1 Year 2 Year 3 Year 4 Year 5 Intercept 0.165*** 0.202*** 0.237*** 0.275*** 0.308*** 0.324*** Dual-class fitted value 0.073** 0.120*** 0.276*** 0.390*** 0.274*** 0.274*** LN mkt. cap. in millions 0.021*** 0.035*** 0.051*** 0.061*** 0.063*** 0.065*** Leverage 0.276*** 0.283*** 0.378*** 0.376*** 0.411*** 0.427*** R&D over total assets 0.139*** 0.119** ** 0.300** EBITDA-to-assets 0.077*** 0.218*** 0.286*** 0.288*** 0.262*** 0.307*** Two-year sales growth 0.054*** 0.045*** CAPEX-to-sales 0.026*** 0.087*** 0.048*** 0.097*** Industry EBITDA-to-price

11 104 S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Table 3 (continued ) Panel C. EBITDA-to-price second-stage regressions Year 0 Year 1 Year 2 Year 3 Year 4 Year 5 Dividend paying firm 0.052*** 0.049*** 0.049*** 0.054*** 0.031* S&P 500 firm * 0.099** Adj. R 2 (%) Number of observations 1,499 1,427 1,306 1, This table presents two-stage regression analysis of pricing multipliers by event year on offer type. In Panel A the probability that an IPO firm is a dual-class firm is estimated using a probit regression. The dependent variable is a dual-class (offer type) indicator variable equal to one for dual-class IPOs and zero otherwise. Market capitalization is the total number of shares outstanding after the offering times the S&P 500-adjusted first-day closing price. Indicator variables are set equal to one, respectively, for venture-backed, high-reputation I-bank, equity-carveout, reverse-lbo, anti-takeover provisions, and NASDAQ-listed deals. Number of uses of proceeds is the number of uses of proceeds listed in the final prospectus. Lagged market return is the compounded daily CRSP value-weighted return over the 22 trading days preceding the initial public offering. The probit regression includes SIC and IPO year dummies. In Panels B and C, pricing multipliers are estimated using OLS regression. The dependent variable in Panel B is earnings-to-price, which is earnings per share divided by price per share. The dependent variable in Panel C is EBITDA-to-price, which is earnings before interest, tax, depreciation, and amortization per share divided by price per share. Dual-class fitted value is the predicted probability that an IPO firm is a dual-class firm from the first-stage probit regression. LN mkt. cap. in millions is the natural logarithm of the S&P 500 index-adjusted (to the beginning of 1990) fiscal yearend market capitalization. Leverage is long-term debt plus short-term debt over total assets. R&D over total assets is research and development expense divided by total assets. EBITDA-to-assets is earnings before interest, tax, depreciation, and amortization divided by total assets. Two-year sales growth in year t is the compound annual growth rate in sales from year t 2 to year t. For years 0 and 1, the 2-year sales growth is the compound annual growth rate in sales from year 0 to year 2. CAPEX-to-sales is the total capital expenditures divided by total sales. Industry multiplier is the respective multiplier for a group of at least five industry-comparable firms. Dividend paying deal equals one for dividend paying firms. S&P 500 deal equals one if the firm is in the S&P 500 index. Respectively, ***, **, and * denote significant difference from zero at 1, 5, and 10 percent. dummies which we have suppressed in the table) have a significant effect on the dual-class decision, though the equity carveout and NASDAQ dummies fall just short of conventional significance levels. Next, we use the fitted probability values from the probit model on the right hand side of our OLS regressions on pricing multipliers. That is, the variable dual in the regressions reported in Panels B and C of Table 3 is not a dummy variable, but instead is the fitted probability of going public with dual-class equity for each sample firm. This approach allows us to control for systematic differences between firms choosing singleclass versus dual-class stock. We include additional control variables to account for other value-relevant differences across firms. On average, duals are larger than singles, so the pricing differences we observe in the previous section might be attributable to differences in firm size rather than to the dual-class equity structure. In our second-stage regressions, we include the logged S&P 500 adjusted market capitalization as a control for firm size. Firms that investors expect to grow rapidly can have lower valuation ratios. To control for this possibility, we include a sales growth variable in our regressions. Specifically, we include the prior 2-year sales growth rate in the regressions. Measuring sales growth prior to an IPO is sometimes impossible because data prior to the IPO is unavailable. Therefore, in event years zero, one, and two, we measure sales growth from year 0 to year 2. This means that the sales growth control variable varies across firms but not across time until event year From event year 3 and beyond, the sales growth rate is simply the compound annual growth rate over the previous 2 event years. In addition, we control for growth opportunities by including in the regression the ratio of R&D to assets and the ratio of capital expenditures to sales. Table 1 indicates that dual-class firms use more leverage, so we include the ratio of long-term plus shortterm debt divided by total assets as a control variable. To address the possibility that dual-class firms may go public at a later stage in their life cycles, we include a dummy variable equal to one for firms which pay 14 Regressions that use pre-ipo sales growth figures when available produce similar results, as do regressions which exclude the sales growth measure.

12 dividends. We also add to the regression model a dummy variable equal to one for firms included in the S&P 500. We control for differences in profitability across firms by including the ROA ratio, which equals earnings before interest, taxes, depreciation, and amortization divided by total assets. Finally, for each firm in our sample we identify a set of comparable firms that come from the same industry and we use these firms to calculate an industry-level valuation ratio in each year. This approach controls for inter-industry valuation differences that vary over time. 15 We run our regression models on each valuation ratio and on each year. In a regression that uses E/P or EBITDA-to-price as the dependent variable, a positive sign on a dual-class fitted value indicates lower relative valuations for duals. Panel B of Table 3 shows the results when our dependent variable is the E/P ratio, and Panel C reports results when we use EBITDA-to-price as the dependent variable. For both valuation ratios, the regressions confirm the results in Table 2. Dual-class firms have higher E/P ratios in each year. In most years the point estimates indicate an E/P difference greater than 0.02, which is large relative to the difference in mean E/P ratios reported in Table 1. Similarly, the regressions in Panel C indicate that dual-class firms have larger EBITDA-to-price ratios than do singles. The difference is significant in all 6 years, and the point estimates suggest an economically significant difference in value between the two firm types. 16 The coefficients on the control variables, though somewhat mixed in terms of significance, generally indicate the following. First, larger firms have lower valuations ratios, a somewhat counterintuitive result. In most of our sample years, larger IPOs (as measured by the IPO offer value) earn higher initial returns than do smaller deals. This leads to the negative coefficient on size in our regressions. Second, firms with more leverage have higher valuation ratios. Third, R&D intensive firms sell at lower multiples, consistent with greater growth opportunities. More profitable firms have higher valuation multiples, as do firms that pay dividends. Both of these estimates likely captures the tendency of more mature firms with higher cash flows and lower growth options to trade at lower prices relative to fundamentals. The signs and significance levels of other controls vary across time. With two dependent variables and 6 year-by-year regressions, we have 12 distinct estimates of the effect of dual-class equity on firm value. All of the point estimates are positive and significant. These findings provide evidence that the market discounts the shares of companies with dual-class equity. The significance and magnitude of this discount is striking and, indirectly, it provides some evidence of the value that managers place on having control. It is hard to imagine that firms going through the IPO process fail to hear, either from their investment bankers or from institutional investors on the road show, that issuers pay a price for insulating managers through a dual-class equity offering. 5. Performance of dual-class firms If the market discounts dual-class shares, then that discount may reflect investors concerns about management s ability to deliver acceptable financial performance in the future. In this section, we try to assess whether the market s discount of dual-class shares is consistent with the future stock and operating performance of these firms. That is, we ask whether dual- or single-class firms exhibit abnormal performance after the IPO Long-run stock returns S.B. Smart et al. / Journal of Accounting and Economics 45 (2008) Our interpretation of the pricing gap between singles and duals is that the market imposes a penalty on firms that go public with dual-class equity. Examining post-ipo stock returns provides a way to assess whether the market s penalty is consistent with ex post performance. If dual-class shares exhibit abnormal positive returns, then we would conclude that over time investors decide that the initial penalty was too severe. Of course, entrenched managers might perform even worse than expected, in which case dual-class stocks would earn negative abnormal returns. 15 Our approach here is similar in spirit to that used to value IPOs in Kim and Ritter (1999). 16 We obtain similar results if we use alternative valuation ratios on the left hand side such as the inverse of Tobin s Q, book-to-market, or sales-to-price.