每股盈餘對公司資本結構的影響 - 政大學術集成

全文

(1)國立政治大學財務管理研究所博士論文. 論文題目政治. 大. 立Share and Capital Structure Earnings Per. er. io. sit. y. ‧. ‧ 國. 學. Nat. n. al v i n C h 周冠男U 博士指導教授： engchi 研究生：陳苡文撰. 中華民國一百零四年十二月.

(2) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.

(3) 謝辭 12 月 31 日，2015 年即將邁入尾聲，我的人生來到一個嶄新的里程碑。就讀博士班以來，深刻體驗學海無涯這句話的真諦，前方永遠有挖掘不完的知識。探索新知的過程固然辛苦，然而，學習新知所帶來的喜悅，卻是無可取代。綜合多年所學，終於完成博士論文的撰寫。能夠走到這裡，有太多的人需要感謝。首先，我要感謝我的指導教授周冠男老師。周老師強調博士生必須培養獨立研究的能力，也經常鼓勵我在研究議題及方法上多方嘗試，但當我遇到瓶頸時，周老師總是在. 治政大不厭其煩地與我討論研究想法，不但加深我論文的深度，更拓展我研究上的視野。感謝立. 最關鍵的時刻提出精闢見解，幫助我度過難關。感謝陳聖賢教授，在我就讀博班期間，. 張紹基教授、陳嬿如教授以及王衍智教授，對於我論文的撰寫、研究方法，以及整個文. ‧ 國. 學. 章架構，給予寶貴的建議，幫助我提升論文品質。. ‧. 感謝系上博班期間的任課老師：張元晨教授、湛可南教授、顏錫銘教授、李志. sit. y. Nat. 宏教授，以及盧敬植教授，幫助我在財務的專業領域以及研究方法上打穩基石，讓我之. al. er. io. 後的研究得以建構在這些基礎之上。對於我碩士班的指導教授杜化宇老師，在我選擇以. v. n. 學術研究為志業時，與我討論並給予鼓勵，我由衷地表達感謝之意。. Ch. engchi. i n U. 感謝系辦助教：采彤，雪玲，溫柔以及偉峰，有你們的幫忙，我才能專心致力於學術研究。感謝系上學長姊：偉劭、胤哲、依婷、楚彬，美菁、周燕的帶領，讓我在剛進入博班懵懂之際，順利步上軌道。感謝我的同學富德及品揚，修課期間日以繼夜討論文章的景象，是我一生的回憶。感謝系上學弟妹：維中、曉琳、建霖、晉吉、濰儒、舜芬，你們的加入，豐富了我的博班生活。最後，感謝我的家人，你們的支持給我勇氣，更是我努力的動力來源。畢業，是一個階段的結束，更是另一階段的開始。期許自己，帶著五年多來的美好回憶與師長的諄諄教誨，憑藉著能力與勇氣，開啟我人生的下一篇樂章。.

(4) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.

(5) Abstract Empirical studies have found that managers choose debt rather than equity to avoid EPS dilution and buy back outstanding shares to boost EPS, I thus explore the resulting effect of EPS on leverage. A firm’s leverage is negatively influenced by the level of its EPS. I also find that fluctuations in EPS have large effects on leverage and these effects persist for at least a decade. Besides, the negative impact of EPS on leverage becomes much stronger after the passage of SOX, in which period managers engage in more actions of debt-equity choices or stock repurchases with the sole. 政治大 corporate governance are two 立 economic mechanisms through which EPS negatively purpose of manipulating EPS. Furthermore, managers’ equity incentives and. ‧. ‧ 國. 學. influences leverage.. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.

(6) Table of Contents 1. INTRODUCTION ............................................................................................................... 1 2. DATA AND SUMMARY STATISTICS ............................................................................. 7 3. THE EFFECT OF EPS ON CAPITAL STRUCTURE .................................................... 8 3.1 DETERMINANTS OF ANNUAL CHANGES IN CAPITAL STRUCTURE................. 8 3.2 THE EFFECTS OF HISTORICAL EPS ON LEVERAGE ........................................... 12 3.3 THE EFFECT OF EPS ON CAPITAL STRUCTURE BEFORE AND AFTER THE SARBANES-OXLEY ACT ................................................................................................ 14 4. PERSISTENCE OF THE EFFECT OF EPS ON CAPITAL STRUCTURE ............... 16 4.1 THE PERSISTENT EFFECT OF EPS ON LEVERAGE .............................................. 16 4.2 EFFECT OF EPS ON CUMULATIVE CHANGES IN CAPITAL STRUCTURE ........ 18. 政治大. 4.3 THE SEPARATION OF EPS EFFECT AND MARKING-TIMING EFFECT ............. 19. 立. 5. ECONOMIC MECHANISMS THROUGH WHICH EPS EFFECT ON CAPITAL STRUCTURE ........................................................................................................................ 22. ‧ 國. 學. 5.1. POTENTIAL MECHANISMS OF THE NEGATIVE EFFECT OF EPS ON CAPITAL STRUCTURES ................................................................................................................... 22. ‧. 5.2. EMPIRICAL RESULTS ............................................................................................... 24. sit. y. Nat. 6. CONCLUSION .................................................................................................................. 27. io. er. APPENDIX A: VARIABLE DEFINITIONS....................................................................... 29 APPENDIX B: THE EXPECTED RELATIONSHIP BETWEEN THE CONTROL. n. al. i n U. v. VARIABLES AND CORPORATE CAPITAL STRUCTURES ........................................ 30. Ch. engchi. REFERENCE ........................................................................................................................ 33 TABLE 1: SUMMARY STATISTICS OF CAPITAL STRUCTURE AND FIRM CHARACTERISTICS .......................................................................................................... 40 TABLE 2: THE EFFECT OF EPS ON ANNUAL CHANGES IN LEVERAGE ............ 41 TABLE 3: THE EFFECT OF EPS ON NET EQUITY ISSUES....................................... 43 TABLE 4: THE EFFECT OF HISTORICAL EPS ON LEVERAGE ............................. 44 TABLE 5 THE EFFECT OF EPS ON ANNUAL CHANGES IN LEVERAGE BEFORE AND AFTER SOX................................................................................................................. 45 TABLE 6: PERSISTENCE OF EPS EFFECT ON LEVERAGE .................................... 46 TABLE 7: THE EFFECT OF HISTORICAL EPS ON CUMULATIVE CHANGES IN LEVERAGE........................................................................................................................... 48 i.

(7) TABLE 8: THE SEPARATION OF EPS EFFECT AND MARKING-TIMING EFFECT ON LEVERAGE.................................................................................................................... 49 TABLE 9: THE EFFECT OF EPS ON ANNUAL CHANGES IN LEVERAGE: SORTED BY EXECUTIVE COMPENSATION ............................................................... 51 TABLE 10: THE EFFECT OF EPS ON ANNUAL CHANGES IN LEVERAGE: SORTED BY CORPORATE GOVERNANCE .................................................................. 52. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ii. i n U. v.

(8) 1. Introduction This paper investigates the short- and long-term effects of the level of earnings per share (EPS) on corporate capital structures. I further examine whether the primary channel through which EPS influences leverage is EPS management actions, and which characteristics of a firm have impact on the relationship between EPS and capital structures. To the best of my knowledge, this is the first paper that studies the long-term side effect of EPS management on firms’ real financial policies. Although traditional corporate finance theory suggests that EPS dilution is. 治政大 that EPS dilution affects Brealey, Myers, and Allen, 2007), it has been found 立. irrelevant in firm valuation and corporate policy (Modigliani and Miller 1958;. corporate managers’ decisions of debt-equity financing and of repurchasing their. ‧ 國. 學. companies’ stocks. Corporate managers are likely to consider EPS as an important. ‧. factor when making financing decisions for the following reasons.. sit. y. Nat. First, managers may actively manage reported EPS as investors and researchers. io. er. generally consider EPS to be one of the most important variables in determining the profitability of a company and estimating a share’s fundamental price. Because. al. n. v i n C hconsistent EPS growth, investors reward firms that report e n g c h i U consistently meet analysts’ EPS forecasts, and avoid EPS disappointments, previous empirical studies show that managers are motivated to opportunistically manage EPS to meet analyst expectations (Skinner and Sloan, 2002; Gunny, 2010), avoid losses (Burgstahler and Dichev, 1997), and maximize stock price (Teoh, Welch, and Wong, 1998a, b). Second, a growing stream of empirical literature shows that executives’ annual bonus compensation contracts are frequently based on EPS performance (Ittner, Larcker, and Rajan, 1997; Kim and Yang, 2010; and Huang, Marquardt, and Zhang,. 1.

(9) 2013), which creates strong incentives for executives to avoid EPS dilution (Healy, 1985; Bens, Nagar, Skinner and Wong, 2003). Brealey and Myers (1996) thus indicate that there is a common belief among executives that share issuance dilutes EPS, and they try to avoid EPS dilution arising from issuing new equity (Brealey and Myers call this view a “fallacy”). Correspondingly, survey evidence shows that, among the firms that considered issuing common equity, out of a total of 11 factors, EPS dilution is the most important factor affecting their decision on common equity issuance (Graham and Harvey, 2001).1. 治政大debt and equity, and whether EPS affects their decisions regarding the choice between 立. Furthermore, recent empirical work shows that the managers’ concerns on the level of. to repurchase their company's stock.2. ‧ 國. 學. Empirical evidence shows that managers are reluctant to issue equity if it dilutes. ‧. the accounting measures of performance or value. For example, Hovakimian,. sit. y. Nat. Hovakimian, and Tehranian (2004) examine how firms choose the form of financing. io. er. by focusing on firms that issue both debt and equity. They find that for about half. al. (44.5%) of debt issuers, issuing equity would dilute the EPS more than issuing debt. n. v i n C25% would. The same is true for only issuers and 20% of equity issuers. The h eofndual gchi U results show that managers are concerned about EPS dilution when they seek external financing. In addition, Huang, Marquardt, and Zhang (2014) show that avoiding EPS dilution helps to resolve underleveraging (as suggested by Fama, 1980), but firms are. 1. Graham and Harvey (2001) find that more than two-thirds of CFOs agree that “EPS dilution” was an important or very important consideration in issuing equity. In that survey as a whole, EPS dilution is regarded as the most important factor among the 11 factors considered in the decision to issue common stock. 2 See Fama (1980), Graham and Harvey (2001), Hovakimian, Hovakimian, and Tehranian (2004), and Huang, Marquardt, and Zhang (2014) for literature related to the effect of EPS on choice between debt-equity issues, and see Bens, Nagar, Skinner and Wong (2003), Brav, Graham, Harvey and Michaely (2005), Oded and Michel (2008), Skinner (2008), Young and Yang (2011), Farrell, Unlu and Yu (2014), and Almeida, Fos, and Kronlund (2015) for literature related to the effect of EPS on stock repurchases. 2.

(10) reluctant to address an overleveraging problem when an equity issue would reduce reported EPS. For literature related to the effect of EPS on stock repurchase, Brav, Graham, Harvey and Michaely (2005) survey 384 financial executives to determine the factors that drive dividend and share repurchase decisions and report that three-fourths of survey respondents indicate that increasing EPS is an important factor affecting their distribution decisions. A substantial body of empirical literature provides evidence regarding the use of share repurchases as a mechanism to boost EPS (Bens, Nagar,. 治政大 cited reason behind the Almeida, Fos, and Kronlund, 2015), and it is one commonly 立. Skinner and Wong, 2003; Hribar, Jenkins, and Johnson, 2006; Oded and Michel, 2008;. increased use of share repurchase programs (see, for example, Grullon and Ikenberry,. ‧ 國. 學. 2000).. ‧. The effects of EPS on financing decisions and on payout policy both have. sit. y. Nat. implicit implications in corporate capital structures. As such, this paper explicitly. io. er. identifies whether, how, and why the level of EPS matters for corporate capital. al. structures. One expects at least a mechanical, short-run impact. The question, however,. n. v i n is whether fluctuations in EPS C have long-run impact on capital structures. If h ea very ngchi U. firms subsequently rebalance away the influence of EPS on the leverage ratio, as the trade-off theory suggests, then EPS would have no persistent impact on capital structures. The significance of EPS for capital structure is therefore an empirical issue. First, I examine the short-term effect of EPS on capital structure. My results show that, annual changes in firms’ leverage ratios are significantly inversely influenced by the level of EPS, which is consistent with the argument that a lower (higher) EPS correlates with a higher probability that a firm will induce an increase. 3.

(11) (decrease) in its debt ratio. These inferences are robust whether leverage is measured in book or market values or whether various control variables are included. I also find that EPS has strong and persistent effects on firms’ capital structures. I employ the external finance weighted-average EPS to evaluate the effect of past EPS on leverage ratios. The weighted-average EPS is a weighted average of a firm’s past EPS which takes high values for firms that raised their external finance, whether equity or debt, when their EPS were high. Regression results show that a firm’s leverage is strongly negatively related to its measure of historical EPS. The main finding is that managers. 治政大 firms. EPS to some extent explain why leverage ratios differ across 立. do not rebalance to some target capital structures, and thus historical valuations of. I then use the passage of the Sarbanes-Oxley (SOX) Act of 2002 as a natural. ‧ 國. 學. experiment to investigate whether the negative relationship between EPS and capital. ‧. structure is a result of attempts to manage EPS through debt-equity choices or stock. sit. y. Nat. repurchases. Cohen, Dey, and Lys (2008) study the consequences of the regulatory. io. er. changes and show that the level of real earnings management (REM) increased. al. n. significantly. Because the actions of mangers to employ debt-equity choices or stock repurchases to manipulate. v i n Cfall EPS the category of REM, h einto ngchi U. I find that SOX. reinforce the negative relationship between EPS and annual changes in leverage. The result suggests that the primary channel through which EPS may influence leverage is via actions of REM. Next, I undertake two separate analyses to evaluate the persistence and find that the influence of past EPS on capital structures is persistent. First, I examine the effect of the weighted-average EPS on the value of leverage many years later. The results show that the effect of the weighted-average EPS on leverage has a half-life of well over 10 years. That is, the current capital structure depends strongly on variations in. 4.

(12) the EPS over the ten previous years or more, even controlling for the last year value of EPS and other control variables. Second, I regress cumulative changes in leverage from the first year of Compustat data to subsequent dates against the weighted-average EPS. The result shows that the weighted-average EPS not only influence leverage levels, but also influence cumulative changes. As such, EPS have an important impact on leverage that persists and accumulates over time. I consider two economic mechanisms through which EPS negatively affects leverage: managerial equity-based compensation channel and corporate governance. 治政大 is more closely tied to the pronounced when the CEO’s potential total compensation 立. channel. The results show that the effect of EPS on capital structures is more. value of stock and option holdings. Managers with higher stock and option. ‧ 國. 學. compensation are more sensitive to short term stock prices, and pay greater attention. ‧. to EPS when they finance externally and when they repurchase stocks (Bens et al.,. sit. y. Nat. 2003; Bergstresser and Philippon, 2006; Huang et al., 2014). Although corporate. io. al. er. governance could also strength the negative effects of EPS on capital structures. n. according to Stein’s (1989) hypothesis that earnings management behavior is. Ch. concentrated among firms with strong corporate. engchi. iv n governance, U. I only find limited. evidence to support this argument. The strong and persistent effect of EPS on capital structures is essentially different from traditional theories of capital structures, and these phenomena are hard to explain with traditional theories of capital structures. In the trade-off theory, there is a long-term debt ratio that will minimize the cost of capital. The trade-off theory predicts that temporary fluctuations in a firm’s characteristics, such as EPS, should at most have temporary effects. My findings show that EPS has very persistent effects on capital structures, however. In the pecking order theory, firms are financed first. 5.

(13) with internally generated funds, then they issue debt if internal funds prove insufficient, and finally they use equity only as a last resort. In essence, the pecking order theory involves rational investors recognizing adverse selection and discounting equity financing, whereas the effect of EPS on capital structures involves irrational investors. In the EPS effect story, investors are irrational and can be tricked by managers’ EPS management. To be more precise, if investors are rational, managers can no longer benefit from managing EPS through debt-equity choice or stock repurchase, and in such case EPS would not influence companies’ capital structures.. 治政 that investors are irrational. Baker and Wurgler argue大 that a firm’s observed capital 立 The market timing theory proposed by Baker and Wurgler (2002) also suggests. structures reflect its cumulative ability to sell overpriced equity shares and raise. ‧ 國. 學. underpriced equity shares, thus current capital structure is strongly related to historical. ‧. market values. Market timing benefits ongoing shareholders at the expense of entering. sit. y. Nat. and exiting ones. The EPS effect, however, puts more emphasis on contracting and. io. al. er. entrenching explanations. As such, the EPS effect on leverage is more pronounced. n. when managers are concerned about compensation that is dependent on stock performance.. Ch. engchi. i n U. v. My study is built on those documenting EPS as an important factor in determining the choice of external financing and stock repurchase (Grullon and Ikenberry, 2000; Bens et al., 2003; Hovakimian et al., 2004; Hribar et al., 2006; Oded and Michel, 2008; Huang et al., 2014; Almeida et al., 2015). However, these studies at most implicitly suggest the short-term, temporary effect of EPS on capital structure. This is the first study to sift through these implicit meanings, and show EPS as a substantial and persistent component of cross-sectional variation in capital structures. My findings indicate that EPS is a significant concern when managers make external. 6.

(14) financial or corporate payout policy, which thus has a large and decades-long impact on corporate structures. The remainder of the paper is organized as follows. Section 2 describes the data and summary statistics. Section 3 presents the empirical results of EPS effect on capital structures, and section 4 investigates its persistence. Section 5 examines cross-sectional heterogeneity in the negative effect of EPS on leverage to better understand the economic mechanisms behind the relationship between EPS and leverage. Section 6 concludes the paper.. 治政 sample consists of all firm-year大 observations 立. 2. Data and Summary Statistics My primary. in the annual. Compustat database between 1965 and 2012. I choose 1965 as the start year to. ‧ 國. 學. mitigate the selection bias toward large, successful firms that exists in the early part of. ‧. the Compustat sample. To maintain consistency with previous empirical studies and. sit. y. Nat. to avoid capital structures dictated by regulatory considerations, I exclude financial. io. er. firms (SIC codes between 6000 and 6999) and utilities (SIC codes between 4900 and. al. 4999), as well as government entities (SIC codes greater than or equal to 9000).. n. v i n C hand market) to lieUin the closed unit interval and I require leverage (both book engchi. restrict the sample to exclude firms with a minimum book value of assets below $10 million. To reduce the effects of a few extreme values and data errors, I winsorize all ratios at the 1st and 99th percentiles. For some of my analysis, I also require the year of the first appearance on Compustat. I restrict this sample to firms for which I could determine the first appearance year between 1965 and 2011. Because of its popularity, I relegate a complete discussion of the data, sample construction, and variable definitions to Appendix A. Table 1 presents summary statistics for my final sample. At this point, I simply note the similarity of many. 7.

(15) statistics to those reported in previous empirical studies of capital structure, such as Leary and Roberts (2014). [Table 1 to be inserted here] 3. The Effect of EPS on Capital Structure 3.1 Determinants of annual changes in capital structure I begin studying the net effect of EPS on the annual change in leverage by estimating the following regression, Δ𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 = 𝛼 + 𝛽𝐸𝑃𝑆𝑖𝑡−1 + 𝜸𝑿𝒊𝒕−𝟏 + 𝛿𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡−1 + 𝜂𝑖 + 𝜈𝑡 + 𝜀𝑖𝑡 ,. (1). where i indexes firms; t indexes years; X is a set of 1-year lagged control variables; η. 治政 is a firm fixed effect (Lemmon, Roberts, and Zender, 大 2008); ν is a year fixed effect; 立. and ε is a random error term assumed to be possibly heteroskedastic within firms.. ‧ 國. 學. There are different definitions of leverage used in the literature. For the purpose of. ‧. this study, I use the definitions of market and book leverage ratios, which are. sit. y. Nat. consistent with previous studies such as Titman and Wessels (1988), Frank and Goyal. io. er. (2009), Lemmon et al. (2008), and Leary and Roberts (2014). Book leverage is. al. defined as book debt to total assets. Market leverage is defined as book debt divided. n. v i n by the sum of total debt and C thehmarket value of equity. e n g c h i U Both are symbolized as Leverage. 𝐸𝑃𝑆𝑖𝑡−1 , the firm i’s 1-year lagged EPS, is the key regressor in equation (1).. Although several studies find that diluted EPS are more strongly associated with stock prices and managers’ decisions than basic EPS, I use both measures of EPS to test. 8.

(16) their respective impact on the results.3 EPSPX and EPSFX denote basic EPS and diluted EPS respectively. I expect the sign of 𝛽 to be negative, in light of the fact that the higher (lower) the EPS, the higher the probability of issuing new equity (repurchasing outstanding shares). Thus higher (lower) EPS would decrease (increase) the debt ratios. The lagged leverage, 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡−1 , is included because leverage is bounded between zero and one. When leverage is near one of these boundaries, the change in leverage can only go in one direction, regardless of the values of the other variables. Lagged leverage therefore enters with a negative sign. The control variables, 𝑿, are. 政治大 referred to in previous academic literature and are well shown as important variables 立. affecting debt ratios.4 They are market-to-book ratio, firm size, fixed asset proportion,. ‧ 國. 學. R&D expenses, industry median debt ratio, cash flow volatility, and profitability. The. ‧. definitions of all of the variables and the conjectures on the relations between the. y. sit. io. er. respectively.. Nat. control variables and leverage are detailed in the Appendix A and Appendix B. al. The results from estimating equation (1) using book and market leverage are. n. v i n C hresults focusing onUbasic EPS in columns (1) to (3) presented in Table 2. I present the engchi and diluted EPS in columns (4) to (6). Column (1) and column (4) present the results of a model consisting solely of EPS. The coefficients on the EPS are significantly. 3. Basic EPS is calculated as earnings available to common shareholders divided by weighted-average common shares outstanding, which does not incorporate the effect of ‘‘potentially dilutive’’ securities such as warrants, convertible debt, and employee stock options. Diluted EPS, on the other hand, uses the treasury stock method to account for the effects of potentially dilutive securities. The denominator of this ratio increases as newly granted employee stock options move into the money and existing employee stock options move further into the money. Thus, diluted EPS expands on basic EPS by including the shares of convertibles or warrants outstanding in the outstanding shares number. Although several studies find that diluted EPS is more highly associated with stock prices and managers’ decision than basic EPS, we follow Jennings, LeClere, and Thompson (1997), Core, Guay, and Kothari (2002), and Bens et al. (2003) and consider both measures to reduce significant impact of the different measures on the results. 4 For studies adopting these control variables, see, among others, Rajan and Zingales (1995), Baker and Wurgler (2002), Lemmon et al. (2008), Frank and Goyal (2009), and Leary and Roberts (2014). 9.

(17) negatively associated with both market and book leverage, indicating that the effect of low (high) EPS is to increase (decrease) leverage. These results are consistent with the idea that when EPS is low, firms tend to choose debt rather than equity when they need external funds, or even buy back outstanding shares to boost EPS. The other columns show the results incorporating control variables into the specification. The coefficient estimates of EPS are largely consistent with previous evidence, in terms of sign and statistical significance. Specifically, the coefficients on EPS remain highly significant and even reveal positive changes in economic significance when. 治政大 [Table 2 to be inserted here] 立. considering the control variables.. The coefficient estimates of the control variables are largely consistent with. ‧ 國. 學. those reported in the literature (Rajan and Zingales, 1995; Baker and Wurgler, 2002;. ‧. Lemmon et al., 2008; Frank and Goyal, 2009; Leary and Roberts 2014). Higher. sit. y. Nat. market-to-book ratio, cash flow volatility, and profitability tend to reduce leverage,. io. er. whereas larger firms and those with a higher fixed asset proportion tend to increase leverage. Industry median debt ratio has positive effect on a specific firm’s leverage,. al. n. v i n C h that firms’ financing which is consistent with the evidence e n g c h i U decisions are responses to the financing decisions of peer firms (Leary and Roberts, 2014). So far I have documented the negative relationship between EPS and leverage changes, but this negative relationship does not indicate whether EPS affects capital structures through equity changes, as the EPS story implies. In addition, EPS may be thought of as an indicator of a company's profitability, and thus the relationship between EPS and leverage change I have found is only the result of profitability. Profitability should have a negative impact on net equity issues according to the traditional theory of capital structure. Under the trade-off theory, more profitable. 10.

(18) firms use more debt (less equity) to shield taxable income and to control the agency cost of free cash flow (Jensen, 1986). Under the pecking order theory, holding investment fixed, more profitable firms keep the level of internal capital higher, and thus have less need to finance new projects with external capital, thus reducing new equity issues. I use the following regression to examine the effect of EPS on net equity issues: 𝑒 𝑒 + ∆ ( ) = 𝛼 + 𝛽𝐸𝑃𝑆𝑖𝑡−1 + 𝛾𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑖𝑡−1 + 𝜃𝑋𝑖𝑡−1 + 𝛿∆ ( ) 𝐴 𝑡−1 𝐴 𝑡 𝜂𝑖 + 𝜈𝑡 + 𝜀𝑖𝑡 .. (2). 治政 common and preferred stock minus the purchase of大 common and preferred stock 立. The dependent variable in equation (2) is net equity issues, defined as the sale of. divided by total assets. I include EPS and Profitability (measured as operating income. ‧ 國. 學. before depreciation divided by total assets) in the same equation. If EPS is merely a. ‧. measure of profitability, I expect the sign of the coefficients on EPS and profitability. sit. y. Nat. to be equal and negative based on both the trade-off theory and the pecking order theory. Table 3 presents the results of estimating equation (2).. er. io. n. 3 to be inserted here] a[Table iv l C n Table 3 shows that the coefficient is significantly positive, whereas the U h e nongEPS i h c coefficient on profitability is significantly negative. The results show the distinguishable effects of EPS and profitability on net equity issues. Firms’ managers are less concerned about EPS dilution when EPS is relatively high, in which situation they are willing to issue new shares. By contrast, when EPS is low, managers seem less inclined to increase the outstanding shares, which would further depress the reported EPS. They may even buy back outstanding shares to increase EPS (Bens et al., 2003; Hribar et al., 2006; Oded and Michel, 2008; Almeida et al., 2015). Most importantly, I find that EPS affects capital structure through net equity issues.. 11.

(19) Profitability is significantly negatively related to net equity issues, which is consistent with both the trade-off theory and the pecking order theory, as well as the empirical results found in Baker and Wurgler (2002). 3.2 The effects of historical EPS on leverage I have shown the short-term negative effect of EPS on leverage ratios, which is implied by previous literature. What I do not know is whether the influence is long-term and persistent. There are two distinguishable outcomes. The effect of EPS on capital structure could be the result of temporary opportunism, and this effect is. 治政 target leverage ratio and thus大 EPS 立. reduced when managers rebalance soon after. Alternatively, managers do not rebalance to some. has persistent effects.. Consequently, historical valuations of EPS will help to explain why leverage ratios. ‧ 國. 學. differ across firms. To address the long-term effect of EPS on leverage, I examine the. ‧. following regression:. (3). sit. y. Nat. 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 = 𝛼 + 𝛽𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡−1 + 𝛾𝐸𝑃𝑆𝑖𝑡−1 + 𝜹𝑿𝒊𝒕−𝟏 + 𝜂𝑖 + 𝜈𝑡 + 𝜀𝑖𝑡 .. io. er. where the indices i and t correspond to firm and year, respectively. The dependent variable, 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒, is either market or book leverage ratios. The control variables are. n. al. Ch. engchi. as defined in Appendix A. In equation (3), I. iv n focus on U. the external finance. weighted-average EPS (i.e. historical EPS), 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 . For a given firm-year, it is defined as 𝑒 +𝑑𝑠. 𝑠 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡−1 = ∑𝑡−1 𝑠=0 ∑𝑡−1 𝑒. 𝑟=0 𝑟 +𝑑𝑟. ∗ 𝐸𝑃𝑆𝑠 ,. (4). where the summations are taken starting from the first year of Compustat data, and 𝑒 and 𝑑 denote net equity and net debt issues, respectively. The definition of net equity. 12.

(20) issues and net debt issues are described in Appendix A. Following Baker and Wurgler (2002), I hold the minimum weight at zero.5 By definition, EPSefwa is the average of the past years’ EPS starting from the first year of Compustat data. Instead of each year’s EPS contributing equally to the final average, for each firm, I give more weight to the EPS for the year in which the firm raised more external finance, whether by issue debt or equity. The rationale is similar to that of Baker and Wurgler (2002). Because external financing events represent opportunities to change leverage, I therefore put more emphasize on those years’ EPS. 治政大 I also include the 1-year identifies precisely which lag EPS are the most relevant. 立. when significant external financing decisions were being made. This measure. lagged EPS in equation (3) to control for the nearest cross-sectional variation in the. ‧ 國. 學. level of EPS. As such, the EPSefwa measures the residual influence of past, within-firm. ‧. variations in EPS. A significant negative estimated coefficient on EPSefwa,t-1 would. sit. y. Nat. indicate that the historical within-firm variation in EPS, not the current cross-firm. io. er. variation, is important in explaining the cross section of leverage. Table 4 reports results for regressions of book and market leverage on the weighted average EPS.6. n. al. i n C [Tableh4 to be inserted here] engchi U. v. In Table (4), columns (1) and (4) show that the coefficients on EPSefwa are significantly negative. The results are robust even after controlling for the traditional determinants of capital structure. Actually, the magnitudes of the coefficients on EPSefwa become larger after considering the control variables. For example, in panel A,. 5. Baker and Wurgler (2002) point out that the purpose of not allowing negative weights is to ensure the formation of weighted average. Otherwise the weights might not be increasing in the total amount of external finance raised in each period, which would eliminate the intuition that the weights correspond to times when capital structure was most likely to be changed. A zero weight means that the variable contains no information about the EPS in that year. 6 Following Baker and Wurgler (2002), we shows the results of Fama-MacBeth calendar-time regressions. The results of OLS regressions, however, yield no new conclusions and are omitted to save space. 13.

(21) the magnitude of the coefficient on EPSefwa increase by more than 60 (70) percent from the model consisting solely of EPSefwa to that consisting of all the control variables; that is from −1.230 (−1.180) to −1.990 (−2.040) when I consider basic EPS (diluted EPS). An interesting result is that the effect of historical EPS is stronger than once-lagged EPS, which is especially true in book leverage (panel A). The results suggest that historical within-firm variation in EPS is an important and strong determinant of capital structure. More precisely, the capital structure of a specific firm is the outcome of the cumulative effect of EPS. That is, firms with higher. 政治大. leverage are those that avoid equity financing and/or repurchase outstanding shares when EPS are low.. 立. 3.3 The effect of EPS on capital structure before and after the Sarbanes-Oxley Act. ‧ 國. 學. In this subsection, I use the passage of the Sarbanes-Oxley (SOX) Act of 2002 as. ‧. a natural experiment to investigate whether the negative relationship between EPS and. io. er. EPS through debt-equity choices or stock repurchases.. sit. y. Nat. capital structure is a result of attempts by EPS-focused decision-makers to manage. al. The literature has long recognized that managers can take accounting actions or. n. v i n C hcalled accrual-basedU earnings management (AEM) real economic actions, which are engchi and real earnings management (REM) respectively, to manage EPS. More specifically,. in AEM managers use accrual-based earnings management techniques to provide flexibility within the accounting rules to boost an EPS number, and in REM managers depart from normal operational practices to achieve certain financial reporting goals and thereby mislead stakeholders (Schipper, 1989; Healy and Wahlen, 1999; Ewert and Wagenhofer, 2005; Roychowdhury, 2006). When managers employ debt-equity choices or stock repurchases with the sole purpose of manipulating EPS numbers, these actions fall into the category of REM (Hribar et al. 2006; Huang et al., 2014).. 14.

(22) The U.S. Congress enacted the Sarbanes-Oxley Act (SOX) in July 2002 in an attempt to protect shareholders and the general public from accounting errors and fraudulent practices, as well as to improve the accuracy of corporate disclosures. Cohen, Dey, and Lys (2008) study the consequences of the resulting regulatory changes and show that the level of AEM declined after the passage of SOX, whereas the level of REM increased significantly. These findings echo the logic that firms switch to managing EPS using real methods instead of accrual-based methods because REM is relatively harder to detect after SOX (Graham, Harvey, and Rajgopal, 2005).. 政治大. Cohen and Zarowin (2010) and Zang (2012) also show the tradeoffs between AEM and REM.. 立. In light of the fact that managers engage in more REM activities after SOX, I. ‧ 國. 學. expect SOX to reinforce the negative relationship between EPS and annual changes in. ‧. debt ratios. I then focus my analysis on equation (1) across two time periods – the. sit. y. Nat. pre-SOX period (from 1965 to 2001) and the post-SOX period (from 2002 to 2012). I. io. er. also test an alternative specification in which I include a SOX dummy and an. al. interaction term between EPS and the SOX dummy (pool regression). The SOX. n. v i n dummy takes the value of 1 in C thehpost-SOX period U e n g c h i and 0 otherwise. Table 5 reports the results of the EPS effect on book leverage incorporating the effect of SOX, whereas the results of using market leverage are suppressed due to space considerations and the similarity of the findings. [Table 5 to be inserted here] The estimated coefficients on EPS are all significantly negative, indicating that, either before or after the passage of SOX, the level of EPS has a significantly negative impact on leverage. However, the magnitude (absolute value) of the coefficient on EPS in the post-SOX period is approximately two times larger than that in the. 15.

(23) pre-SOX period, indicating that the negative impact of EPS on leverage becomes much stronger after the passage of SOX. The results of the pool regression (columns (3) and (6)) provide similar illustrations of these findings, indicated by the negative and significant coefficient on the interaction term of EPS and SOX. Overall, the above analysis indicates that, while the level of EPS consistently exhibits negative impact on leverage, the magnitude increases significantly from the pre-SOX period to the post-SOX period. The results are consistent with prior studies showing that the level of REM increased in the post-SOX period. Moreover, this finding suggests that the. 治政大 4. Persistence of the Effect of EPS on Capital Structure 立. primary channel through which EPS may influence leverage is REM actions.. A prominent feature in section 3 is that, in the short-term, firms’ EPS have a. ‧ 國. 學. negative influence on changes in leverage ratios. However, the long-term effects of. ‧. EPS are ambiguous. If firm values are highly sensitive to deviations from their target. sit. y. Nat. debt ratios, then these variations in EPS over time should have only a fleeting. io. er. influence on observed capital structures. In contrast, if the function mapping capital. al. structure to firm value is very flat, then these variations in EPS over time will have a. n. v i n C hslowly. I address U persistent effect that reverses quite e n g c h i the long-term effects of EPS on capital structure in this section. 4.1 The persistent effect of EPS on leverage. Following Baker and Wurgler (2002), I employ a system of three regressions to test the persistence of the effect of past EPS variations on firms’ leverage: Leverage𝑖𝑡+1 = 𝛼1 + 𝛽1 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 + 𝛾1 𝐸𝑃𝑆𝑖𝑡 + 𝜹𝟏 𝑿𝒊𝑡 + 𝜂𝑖 + 𝜈𝑡 + 𝜀1,𝑖𝑡+1. (5). Leverage𝑖𝑡+𝜏 = 𝛼2 + 𝛽2 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 + 𝛾2 𝐸𝑃𝑆𝑖𝑡 + 𝜹𝟐 𝑿𝒊𝑡 + 𝜂𝑖 + 𝜈𝑡 + 𝜀2,𝑖𝑡+𝜏 .. (6). Leverage𝑖𝑡+𝜏 = 𝛼3 + 𝛽3 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 + 𝛾3 𝐸𝑃𝑆𝑖𝑡+𝜏−1 + 𝜹𝟑 𝑿𝒊𝑡+𝜏−1 + 𝜂𝑖 + 𝜈𝑡 + 𝜀3,𝑖𝑡+𝜏 .. (7). 16.

(24) In equation (5), I examine the effect of past EPS, 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 , on leverage at time t+1, controlling for other firm characteristics measured at time t.7 Equation (6) looks at the effect of 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 on leverage at time 𝑡 + 𝜏, still using controls from time 𝑡, whereas equation (7) looks at the effect of 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 on leverage at time 𝑡 + 𝜏, controlling for firm characteristics measured at time 𝑡 + 𝜏 − 1 . The difference between equation (6) and equation (7) is the time when the controlling characteristics are measured. The significance of the coefficient on 𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡 will document the persistence of the EPS effect, suggesting that EPS variation from t and before remains. 政治大 indicate the persistent effect 立of the EPS variation from 𝑡 and before on leverage at. a strong determinant of leverage as of t + τ. Specifically, the strength of 𝛽3 will. ‧ 國. 學. time 𝑡 + 𝜏 , even controlling for the 𝑡 + 𝜏 − 1 values of EPS and other characteristics. I require that the set of firms in all three regressions is identical. In. ‧. other words, a firm must survive at least 𝜏 years to be included as a sample firm in. Nat. sit er. io. system regressions.. y. regression (5). Table 6 reports the Fama-MacBeth estimates and t-statistics of the. n. a[Table l C 6 to be inserted here]n i v U 𝛽 is significantly negative, h e n(1) In book leverage (panel A), column h i that g cshows 1 which is consistent with the results in Table 4. In addition, the narrow range of 𝛽1 between -1.28 to -1.99 indicates that the survival effect is slight. 𝛽2 and 𝛽3 both remain strongly significant for at least 10 years, which shows the persistence of the EPS effect on capital structures. That is, for example, a firm’s capital structure as of 2015 depends strongly on variations in the EPS from 2005 and before, even controlling for the value of EPS and other control variables in 2014. In book leverage,. 7. Equation (5) is simply a repetition of equation (3), in which the dependent variable is once-ahead relative to the independent variables. 17.

(25) the ratios of 𝛽2 to 𝛽1 and 𝛽3 to 𝛽1 show only a slight decay in the initial effect of past EPS variations. 𝛽2 divided by 𝛽1 (𝛽3 divided by 𝛽1) is 0.62 (0.87) after 10 years, which indicates that more than half of the initial effect is still apparent 10 years later. The more interesting results are shown in the last column: the ratio of 𝛽3 to 𝛾3 is greater than 1. Thus the influence of the historical variation in EPS, even calculated with data over 10 years old, is much more substantial than that of once-lagged EPS. Panel B reports the results of market leverage and the results are qualitatively unchanged.. 政治大 theory, past, temporary fluctuations in EPS, captured by 𝐸𝑃𝑆 立. My findings run counter to the predictions of trade-off theory. Under trade-off 𝑒𝑓𝑤𝑎,𝑡 ,. should no longer. ‧ 國. 學. matter and, in terms of the coefficients, 𝛽2 and 𝛽3 should be insignificant. But I find that 𝛽2 and 𝛽3 are both significant. In sum, I find that over long horizons the. ‧. EPS have strong and persistent effects on firms’ capital structures.. y. sit. Nat. 4.2 Effect of EPS on cumulative changes in capital structure. io. al. er. One implication of the persistent effect of EPS on leverage in Table 6 is that. v. n. firms' cumulative changes in capital structures are strongly affected by their variations in EPS over time. I next. i n C U h ecumulative regress n g c h ichanges. in leverage against the. weighted-average EPS to examine this argument: 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 − 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,0 = 𝛼 + 𝛽𝐸𝑃𝑆𝑖𝑡−1 + 𝛾𝐸𝑃𝑆𝑒𝑓𝑤𝑎,𝑡−1 + 𝛿𝑋𝑖𝑡−1 + 𝜂𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,0 + 𝜂𝑖 + 𝜈𝑡 + 𝜀𝑖𝑡 .. (8). The dependent variable is the cumulative changes in leverage from the first year of Compustat data to year t. The leverage in the first year of Compustat data, 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,0 , is also included to control the possible effect of the initial leverage level (Lemmon et al., 2008). The OLS regression results from estimating equation (8) using book and market leverage are presented in Table 7.. 18.

(26) [Table 7 to be inserted here] Table 7 shows that the coefficients on the once-lagged EPS and the weighted-average EPS have a similar sign and magnitude to those of Table 4, indicating that these variables not only influence leverage levels but also influence cumulative changes in leverage. More specifically, the coefficient on 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 is significantly negative, indicating that over time the variations in EPS are a relevant and strong determinant of cumulative changes in leverage. I thus provide more evidence that EPS has persistent effects on capital structure and these effects. 政治大 conclusion, thus I suppress these results due to space considerations and the similarity 立 accumulate over time. Fama-MacBeth calendar-time regressions yields no new. 4.3 The separation of EPS effect and marking-timing effect. 學. ‧ 國. of the findings.. ‧. Baker and Wurgler (2002) show that firms tend to raise capital from the equity. sit. y. Nat. market when share prices are perceived to be more favorable. The negative relation. io. er. between EPS and capital structure can be viewed through the lens of market timing. al. theory alone because high EPS tends to induce enthusiasm for the stock price. There. n. v i n is evidence that market timing C relates and equity issues: Analysis of earnings h etonEPS gchi U. forecasts and realizations around equity issues suggest that firms tend to issue equity at times when investors are rather too enthusiastic about earnings prospects (Loughran and Ritter, 1997; Rajan and Servaes, 1997; Teoh et al., 1998a, 1998b; and Denis and Sarin, 2001). I have controlled, however, for these effects by including firms’ market-to-book ratios. In addition, I perform two separate analyses to further address this issue. First, I divide firms into five categories based on market-to-book ratio. That is, for every year of data, I sort all the firms into quintiles according to the value of their 19.

(27) market-to-book ratio at the beginning of the year, and then estimate equation (1) for each group. By doing so, I rule out the explanation of the variation in market-to-book for the cross section variance of leverage changes. In other words, for each regression, I emphasize the explanatory power of the variations in EPS on firms’ leverage changes. Focusing on the book leverage results, I find that most of the coefficients on both the basic EPS and diluted EPS are significantly negative, while all of the coefficients on the market-to-book ratio are insignificant.8 These findings reveal that the effect of EPS on annual leverage changes is different from that predicted by Baker. 治政大 reluctance to issue new between EPS and leverage change is caused by the managers’ 立. and Wurgler’s marking-timing theory. More specifically, the negative relation. shares when EPS is low and vice versa, which is distinguishable from the negative. ‧ 國. 學. relation between market-to-book ratio and leverage in Baker and Wurgler (2002). The. ‧. results of the market leverage regression lead to a similar conclusion.. sit. y. Nat. Second, I also include the external finance weighted-average market-to-book. io. al. n. this is defined as. er. ratio suggested by Baker and Wurgler (2002) in equation (3). For a given firm-year,. 𝑀𝐵𝑒𝑓𝑤𝑎,𝑡 = ∑𝑡𝑠=0 ∑𝑡. 𝑒𝑠 +𝑑𝑠. 𝑟=0 𝑒𝑟 +𝑑𝑟. Ch. ∗ 𝑀𝐵𝑠 ,. engchi. i n U. v. (9). where 𝑀𝐵 is market-to-book ratio, and 𝑀𝐵𝑒𝑓𝑤𝑎 is the summation of 𝑀𝐵 starting from the first year of Compustat data. Symbols 𝑒 and 𝑑 denote net equity and net debt issues, respectively, as defined in Appendix A. I included 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 and 𝑀𝐵𝑒𝑓𝑤𝑎 in the same regression to race the EPS effect against the marking-timing effect. If the EPS effect dominates, the coefficient on 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 will be highly. 8. There are 10 subsample regressions analyzing the effect of EPS on annual change in book leverage; five are on basic EPS and five are on diluted EPS. All coefficients on EPS are negative, but three coefficients on the EPS are insignificant among the 10 coefficients. 20.

(28) significant even when controlling for 𝑀𝐵𝑒𝑓𝑤𝑎 in the same regression. On the other hand, if EPS just captures the market timing effect, then 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 will no longer matter. The empirical goal is to disentangle the various determinants of capital structure by statistically identifying the significance of the different variables. Table 8 reports the Fama-MacBeth estimates and t-statistic of the regressions considering both the persistent EPS effect and the marking-timing effect. Column (1) presents the results of multivariate regressions that use the control variables suggested by Baker and Wurgler (2002), such as once-lagged market-to-book ratio, firm size,. 政治大. asset tangibility, and profitability. The goal of this model is to duplicate the results of. 立. Baker and Wurgler (2002) that show the determinants of leverage. My findings,. ‧ 國. 學. similar to that of Baker and Wurgler, indicate that the historical within-firm variation in market-to-book is important in explaining the cross section of leverage. Column (2). ‧. shows that the past variation in EPS explain a great deal of the cross section of. Nat. er. io. which is consistent with my findings in Table 4.. sit. y. leverage when using the control variables suggested by Baker and Wurgler (2002),. n. a l [Table 8 to be inserted here]i v n Ch U n gvariations c h i in EPS and market values as a Column (3) considers both the e past horse race between these two variables. The coefficients on 𝐸𝑃𝑆𝑒𝑓𝑤𝑎 and 𝑀𝐵𝑒𝑓𝑤𝑎 are both significant, which indicates that past variations in EPS and market values are two of the important determinants of cross-sectional variation in capital structures. Column (4) includes the control variables used earlier in this paper, some of which overlap with the variables used in Baker and Wurgler. Three extra variables are R&D expenses, cash flow volatility, and industry median debt ratio. The results show that the coefficient on the weighted average EPS is not sensitive to this change in control variables. Moreover, column (4) shows that the explanatory power of the variation in 21.

(29) EPS is much stronger than that of the market-to-book ratio, measured either by the statistical significance or the magnitude of the coefficients. I find that the managers’ decisions incorporating the level of EPS have a persistent effect on firms’ leverage, and the EPS effects are separate from various effects documented in prior literature. 5. Economic Mechanisms through Which EPS Effect on Capital Structure Thus far, this study has shown the important effect of EPS level on firms’ capital structures. I now conduct additional analyses to enrich my findings. I begin with a brief discussion of the potential mechanisms behind the estimated effect of EPS on. 治政 5.1. Potential mechanisms of the negative effect of EPS大 on capital structures 立 capital structures, which I use to guide my subsequent empirical analysis.. Jensen and Meckling (1976) and Smith and Watts (1982) suggest that managers'. ‧ 國. 學. equity incentives − arising from stock-based compensation and stock ownership −. ‧. better align the interests between managers and shareholders. 9 These equity. sit. y. Nat. incentives, however, motivate these managers to increase the value of the shares. io. er. because managers with high equity incentives are more likely to sell shares in the. al. future. Given that the capital market uses current earnings to predict future earnings. n. v i n when pricing firm equity, these C managers expected to manage earnings to keep the h e nare gchi U short-term stock price high (Stein 1989). Indeed, Cheng and Warfield (2005) find evidence supporting the positive effect of equity incentives on earnings management. They define CEO equity incentives as the various stocks and options owned by the CEOs as a percent of total shares outstanding, and find that CEO equity incentives increase the likelihood of earnings management. Other studies also show the relationship between earnings management. 9. Lambert and Larcker (1987), Morck, Shleifer, and Vishny (1988), Hanlon, Rajgopal, and Shevlin (2003), among others examine the association of managerial ownership and stock-based compensation with future firm performance and find evidence consistent with the incentive-alignment effect of these equity incentive elements. 22.

(30) and stock compensation through stock options (Baker et al., 2003; Bartov and Mohanram, 2004; Kwon and Yin, 2006). Because the negative relationship between EPS and capital structures is the result of EPS management via actions that have the opposite effect of equity offerings, such as share repurchases, and given prior studies’ findings that equity incentives induce managers to engage more in earnings management, I expect equity incentives to strengthen the negative impact of EPS on debt ratios. That is, the effect of EPS on capital structures is more pronounced when the CEO’s potential total compensation is more closely tied to the value of stock and. 治政大 can be a function of Stein (1989) also suggested that earnings management 立. option holdings.. governance. Stein develops a model of inefficient managerial behavior and suggests. ‧ 國. 學. that managers engage in costly behaviors to improve short-term accounting earnings. ‧. to thereby induce the market to predict higher future earnings. Furthermore, the. sit. y. Nat. weight the manager places on short-term accounting earnings increases when the. io. al. er. threat of a takeover (which threat leads to better external corporate governance). n. becomes stronger. In sum, corporate governance mechanisms can improve managerial. Ch. performance, but may also encourage managers. engchi. iv n to U manipulate. current financial. statement earnings. A limited empirical literature has tested theories related to the Stein (1989) hypothesis that corporate governance increases focus on short-run accounting earnings. Actually, the limited number of papers discussing the influence of corporate governance on earnings management contradicts Stein's hypothesis.10 Leuz, Nanda, and Wysocki (2003) examine systematic differences in earnings management across 31 countries and find that strong corporate governance, marked by shareholder rights, limits managers’ acquisition of private control benefits, and thus reduces the incentives for managers to engage 10. We focus on the literature employing shareholder right as an indicator of corporate governance. 23.

(31) in earnings management activities. In addition, Farrell, Yu, and Zhang (2013) examine the factors associated with firms that use share repurchases to manage EPS and find that strong corporate governance discourages repurchase-based earnings management.. A recent paper by Ohrn (2014), however, finds that earnings management behavior is concentrated among firms with strong corporate governance, which is consistent with the hypothesis of Stein (1989). Ohrn indicates that one of the reasons why empirical analyses have failed to confirm Stein’s hypothesis is that the levels of corporate governance and earnings management behavior are potentially determined simultaneously and he thus relies on a corporate tax policy, “bonus depreciation”, to. 政治大. address the issue.11 Ohrn finds that the investment behavior of strongly governed. 立. firms is less responsive to bonus depreciation, which is interpreted as evidence that. ‧ 國. 學. corporate governance mechanisms encourage managers to focus on current financial. ‧. statement earnings at the expense of long-run profits. Given the ambiguous findings on the influence of corporate governance on earnings management in previous. y. Nat. capital structures is therefore an empirical issue.. n. al. 5.2. Empirical results. Ch. engchi. er. io. sit. literature, how corporate governance impacts the relationship between EPS and. i n U. v. To shed light on the potential mechanisms behind the negative influence of EPS on leverage, I examine the heterogeneity in the coefficient on EPS, β, from equation (1). Specifically, I rank three groups of firms based on relevant variables and focus on samples within the lower and upper thirds of each relevant variables’ distribution.. 11. While bonus depreciation effectively increases the economic value of investment projects, it leaves the accounting earnings associated with any potential project unchanged. For managers that seek to maximize the economic value of the firm, bonus depreciation provides strong incentive for increased investment. In contrast, for managers that seek to maximize only accounting earnings, then bonus depreciation has no effect on their investment behavior. Therefore, the absence of response (or under-response) to the policy is evidence of earnings management. Ohrn’s (2014) research design avoids the simultaneity complications under the plausible assumption that corporate governance decisions are not made based on an investment response to the tax policy. 24.

(32) Analysis of the subsamples allows us to check the difference in each coefficient between these two groups.12 In table 9, I examine whether equity incentives strengthen the negative impact of EPS on leverage. The sample here includes all firm-years with data on CEOs' stock-based compensation and ownership available from the Standard & Poor's ExecuComp database for the period from 1992 to 2012, aligning with the start year of the database. 13 Following Berger, Ofek, and Yermack (1997) and Cheng and Warfield (2005), I use the share ownership (SHR_OWN), the unexercised options exercisable options ownership 治政大 unexercised unvested (OPT_EXER_OWN), and the estimated value of in-the-money 立. ownership. (OPT_OWN),. the. unexercised. options (OPT_ITM_V) by the CEO of a firm as the level of equity incentives. I run. ‧ 國. 學. regressions on the low and high thirds of the distribution of these variables,. sit. y. Nat. [Table 9 to be inserted here]. ‧. respectively.. io. er. The results in table 9 show that all the coefficients on EPS are negative, but only. al. those estimated based on subsamples in firms with high equity incentives are. n. v i n C(with significantly different from zero for the subsample divided by h e nonegexception chi U OPT_ITM_V. In addition, the coefficients on EPS based on the high equity incentives subsamples are many times in magnitude as compared to the low equity incentives subsamples, and the t-test for the difference in the coefficients on EPS based on high and low thirds subsamples respectively is negative and significant at the 0.01 level. Overall, I find that the equity incentives strengthen the negative influence of EPS on. 12. In this subsection, we show the results of using the changes in book leverage as dependent variable and basic EPS (EPSPX) as independent variables. We also examine the robustness of my results to changes in the measure of leverage and EPS, such as market leverage and diluted EPS (EPSFX) respectively. 13 We have checked the significantly negative effect of EPS on leverage, which is consistent with the results shown in Table 2, during the period from 1992 to 2012. 25.

(33) the annual changes in debt ratios, which is consistent with the findings in prior literature that when managers’ compensation links more to stock price, they are intent on managing EPS through means that reduce outstanding shares which in turn affect capital structures. In table 10, I examine the effect of EPS on annual changes in leverage across firms with different levels of corporate governance. I use the G-index and E-index as proxies for corporate governance, and the G-index and E-index are contained in the RiskMetrics Governance Legacy Database. In particular, Gompers, Ishii, and Metrick. 治政大 rights. The score of added for every antitakeover provision that restricts shareholder 立 (2003) construct the G-index in a straightforward manner: for every firm, a point is. the G-index is positively related to managerial power and inversely related to both the. ‧ 國. 學. strength of shareholder rights and corporate governance; that is, an increase (decrease). ‧. in the G-index indicates a decrease (increase) in the strength of corporate governance.. sit. y. Nat. I also use the E-index of Bebchuk, Cohen, and Ferrell (2009) to measure the strength of. io. er. corporate governance, which consists of 6 of the 24 provisions used in the G-index of. al. Gompers et al. (2003). Similar to the G-index, a high (low) E-index indicates weak. n. v i n C h weak corporateUgovernance in the firm. I begin (strong) shareholder rights, implying engchi the sample period from 1996 to 2009, as I require available data from RiskMetrics.14 [Table 10 to be inserted here]. Table 10 shows the regression results of equation (1) based on the bottom and top third subsamples of the G-index and E-index. I find that coefficients on EPS are both significantly negative for the estimation based on the subsamples of low and high strength of corporate governance, measured either by the G-index or E-index. A slight difference between the estimated results in these two subgroups is indicated by the 14. We have checked the significantly negative effect of EPS on leverage, which is consistent with the results shown in Table 2, during the period from 1996 to 2009. 26.

(34) t-test of the difference in the magnitude of the coefficients on EPS for low and high G-index (E-index) subsamples: the coefficient on EPS based on the subsample in firms with high G-index (E-index) is significantly larger than that based on the subsample in firms with low G-index (E-index). The results provide weak evidence that EPS have more significant resulting effects on capital structures in firms with higher strength of corporate governance (lower G-Index or E-index), which to some extent supports the Stein (1989) hypothesis and Ohrn’s (2014) findings that corporate governance mechanisms encourage managers to focus on current financial statement. 治政大influences leverage primarily Overall, the results reinforce my argument that EPS 立. earnings.. through the actions that managers take to influence EPS. Thus the level of equity. ‧ 國. 學. incentives reinforce the impact of EPS on capital structures because managers with. ‧. high equity incentives are more likely to sell shares in the future and this motivates. sit. y. Nat. these managers to engage in EPS management to increase the value of the shares to be. io. er. sold (Cheng and Warfield, 2005). The impact of corporate governance on the. al. relationship between EPS and leverage is unclear; I find only limited evidence in. n. v i n C h governance mechanisms support of the hypothesis that corporate encourage managers engchi U to focus on current financial statement earnings. 6. Conclusion A variety of evidence suggests that EPS affects decisions regarding the choice between debt and equity, and whether to repurchase company stock. This evidence implicitly suggests the short-term and temporary effect of EPS on capital structures. This study, however, has explicitly shown not only the short-term effect but also the long-term and persistent effect of EPS on corporate capital structures. I find that high-leverage firms tend to be those that raised funds through issuing debt or even. 27.

(35) buying back outstanding shares when their EPS was low. The fluctuations in EPS have a significant effect on capital structures and this effect persists for at least a decade. In addition, I find that the negative impact of EPS on leverage becomes much stronger after the passage of SOX, in which period managers engage in more actions of debt-equity choices or stock repurchases with the sole purpose of manipulating EPS. I also identify two economic mechanisms through EPS negatively influences capital structures. I find that the negative impact of EPS on capital structures is only. 治政大 on the capital structure in evidence that EPS has more significant resulting effects 立. significant for firms with high equity incentives for managers. I also find limited. firms with stronger corporate governance.. ‧ 國. 學. These results are hard to explain through traditional theories of capital structure.. ‧. The most realistic explanation for the results is that capital structure is largely the. sit. y. Nat. cumulative outcome of managers’ attempts to manage the reported EPS, whether by. io. al. er. debt-equity choices or stock repurchases. In this story, the influence of EPS on the. n. leverage ratio is not rebalanced away, but instead accumulates over time and affects the capital structure outcome.. Ch. engchi. 28. i n U. v.

(36) Appendix A: Variable definitions Appendix A details the variable construction for analysis. Compustat variable names are denoted by their Xpressfeed pneumonic in bold. Time periods are denoted by (t) or (t - 1) suffixes. Dependent variables Book Leverage. Short-term debt plus long-term debt divided by total assets. [(dltt + dlc)/ at]. Market Leverage. Short-term debt plus long-term debt divided by total assets minus book equity plus market equity [(dltt + dlc)/( at – seq + prcc_f * csho)]. 政治大 Basic earnings per share [epspx] 立. Independent variables EPSPX EPSFX. ‧ 國. Natural logarithm of total assets [log(at)]. y. sit. io. R&D. Property, plant, and equipment divided by total assets [ppent/at]. er. Nat. Fixed Asset. Market value of assets (total assets minus book equity plus market equity) divided by total assets [(at – seq + prcc_f * csho)/at]. ‧. Size. 學. MB ratio. Diluted earnings per share [epsfx]. Research and development expense divided by firm size [xrd/log(at)]. al. n. v i n C hincome before extraordinary U Operating items plus i e h n c g depreciation divided by total assets [(ib + dp)/at]. Cash Flow (CF) Std (CF). Cash Flow Volatility is computed each year as the historical standard deviation of Cash Flow, requiring at least 3 years of data.. Median of industry leverage (Med. of ind. lev). Median of industry leverage is computed each year as the median leverage for a particular industry, which is defined by three-digit SIC code.. Profitability. Operating income before depreciation divided by total assets [oibdp/at]. Net external financing Net Equity Issuances. Sale of common and preferred stock minus purchase of common and preferred stock divided by total assets [(sstk(t) − prstkc(t))/at(t − 1)] 29.

(37) Net Debt Issuances. The change in total debt divided by total assets [((dltt(t) + dlc(t)) − (dltt(t − 1) + dlc(t − 1)))/at(t − 1)]. Managers’ equity incentives SHR_OWN. Share ownership by the CEO of a firm [SHROWN_EXCL_OPTS/(SHRSOUT)]. OPT_OWN. Unexercised option ownership by the CEO of a firm [(OPT_UNEX_EXER_NUM+OPT_UNEX_UNEXER_ NUM)/(SHRSOUT)]. OPT_EXER_OWN. Unexercised exercisable option ownership by the CEO of a firm [OPT_UNEX_EXER_NUM/SHROUT/SHRSOUT]. OPT_ITM_V. Estimated value of in-the-money unexercised unvested option ownership by the CEO of a firm [OPT_UNEX_UNEXER_EST_VAL/(SHROUT*SPRC)]. Corporate government measures. 立. 政治大. The governance index of Gompers et al. (2003). E-Index. The entrenchment index of Bebchuck et al. (2009). ‧ 國. 學. G-Index. ‧. Appendix B: The expected relationship between the control variables and. sit. y. Nat. corporate capital structures. io. er. My empirical model of capital structure is a generalization of that used. al. throughout the empirical capital structure literature, and includes seven control. n. v i n variables that were found to be C correlated (e.g., Rajan and Zingales, 1995; h e n togleverage chi U. Baker and Wurgler, 2002; Lemmon et al., 2008; Frank and Goyal, 2009; Leary and Roberts 2014). The control variables and the expected relation between the control variables and leverage are: (1) Market-to-book ratio (MB ratio): Baker and Wurgler (2002) show that firms are more likely to issue equity when their market values are high and to repurchase equity when their market values are low. In addition, a firm with high market-to-book ratio of assets which is generally taken as a sigh of more attractive future growth options, suggested by Myers’ (1984) pecking order theory, tends to protect by limiting its 30.

(38) leverage. From these various perspectives, I expect there is a negative relationship between market-to-book ratio of assets and firm’s leverage. (2) Firm size (Size): Previous researches find that leverage is positively correlated with firm size (e.g., Titman and Wessels, 1988; Rajan and Zingales, 1995; Fama and French, 2002). A number of intuitive explanations can be put forward to account for this stylized fact like higher transparency, lower operating volatility, and cheaper access to public debt market for larger firms allowing them much easier to issue debt. (3) Fixed asset proportion (Fixed asset): In the presence of contracting frictions, fixed. 政治大. assets which are with high tangibility are more desirable from the point of view of. 立. creditors because it allows creditors to more easily repossess a firm’s assets (see. ‧ 國. 學. Shleifer and Vishny, 1992; Campello and Giambona, 2013). Referring to the theory suggestions and previous empirical results that tangibility increases borrowing. ‧. capacity, I expected that firm’s leverage will increase with fixed asset proportion.. sit. y. Nat. io. al. er. (4) R&D expenses (R&D): The theoretical documentations suggest that firms with. n. more R&D expenses will prefer to have more equity based on the economic. Ch. characteristics of intangible investments, mainly. engchi. iv n risk U and. uncertainty of future. benefits, which are then confirmed in most part of the empirical research results like Bhagat and Welch (1995). I therefore expect an inverse relation between R&D expenses and debt ratios. (5) Industry median debt ratio (Med. of ind. lev.): It is widely held that industry average leverage ratios are an economically important determinant of firms’ capital structures (Bradley, Jarrell, and Kim, 1984; Welch, 2004; MacKay and Phillips, 2005, and Frank and Goyal, 2009). Furthermore, recent empirical work by Leary and Roberts (2014) show evidence for a complementary explanation of the importance of 31.

(39) industry leverage, namely that firms are directly influenced by the financing choices of their peers. I therefore include the industry median debt ratio in the regression to account for the industry influence on firms’ leverage. Three-digit SIC codes are used to identify each industry. (6) Cash flow volatility (Std (CF)): Firms with higher cash flow volatility often face greater uncertainty, and thus increase in the probability that a future state of the world will take place in which a firm’s cash flows are insufficient to service its debt obligations (Bradley et al., 1984). Bradley et al., among others, show that the. 政治大. relationship between the debt ratio and volatility is negative.. 立. (7) Profitability: Under the trade-off theory, more profitable firms face lower expected. ‧ 國. 學. costs of financial distress. They also use more debt to shield taxable income and to control the agency cost of free cash flow (Jensen, 1986). Thus, increases (decreases). ‧. in profitability cause firm leverage to increase (decrease) under the trade-off theory.. y. Nat. io. sit. Under the pecking order theory, holding investment fixed, more profitable firms keep. n. al. er. the level of internal capital higher, and thus have less need to finance new projects. Ch. i n U. v. with external debt capital. Therefore, given investment outlays, leverage is lower for more profitable firms.. engchi. 32.

(40) Reference Almeida, Heitor, Vyacheslav Fos, and Mathias Kronlund, 2015, The real effects of share repurchases, Forthcoming, Journal of Financial Economics. Bartov, Eli, and Partha Mohanram, 2004, Private information, earnings manipulations, and executive stock-option exercises, The Accounting Review 79, 889-920. Baker, Malcolm, and Jeffrey Wurgler, 2002, Market timing and capital structure, Journal of Finance 57, 1-32. Bebchuk, Lucian, Alma Cohen, and Allen Ferrell, 2009, What matters in corporate. 治政大M. H. Franco Wong, 2003, Bens, Daniel A., Venky Nagar, Douglas J. Skinner, and 立 governance? Review of Financial Studies 22, 783-827.. Employee stock options, EPS dilution, and stock repurchases, Journal of. ‧ 國. 學. Accounting and Economics 36, 51-90.. ‧. Bergstresser, Daniel, and Thomas Philippon, 2006, CEO incentives and earnings. sit. y. Nat. management, Journal of Financial Economics 80, 511-529.. io. al. er. Bhagat, Sanjai, and Ivo Welch, 1995, Corporate research and development. n. investments international comparisons, Journal of Accounting and Economics 19, 443-470.. Ch. engchi. i n U. v. Bradley, Michael, Gregg A. Jarrell, and E. Han Kim, 1984, On the existence of an optimal capital structure: Theory and evidence, Journal of Finance 39, 857-878. Brav, Alon, John R. Graham, Campbell R. Harvey, and Roni Michaely, 2005, Payout policy in the 21st century, Journal of Financial Economics 77, 483-527. Brealey, Richard A., and Stewart C. Myers, 1996, Principles of Corporate Finance, New York, NY: McGraw-Hill. Burgstahler, David, and Ilia Dichev, 1997, Earnings management to avoid earnings decreases and losses, Journal of Accounting and Economics 24, 99-126.. 33.

(41) Campello, Murillo, and Erasmo Giambona, 2013, Real assets and capital structure, Journal of Financial and Quantitative Analysis 48, 1333-1370. Cheng, Qiang, and Terry D. Warfield, 2005, Equity incentives and earnings management, The Accounting Review 80, 441-476. Cohen, Daniel A., Aiyesha Dey, and Thomas Z. Lys, 2008, Real and accrual‐based earnings management in the pre ‐ and post ‐ Sarbanes ‐ Oxley periods, The Accounting Review 83, 757-787. Cohen, Daniel A., and Paul Zarowin, 2010, Accrual-based and real earnings. 政治大. management activities around seasoned equity offerings, Journal of Accounting. 立. and Economics 50, 2-19.. ‧ 國. 學. Core, John E., Wayne R. Guay, and S. P. Kothari, 2002, The economic dilution of employee stock options: Diluted EPS for valuation and financial reporting, The. ‧. Accounting Review 77, 627-652.. sit. y. Nat. Denis, David J., and Atulya Sarin, 2001, Is the market surprised by poor earnings. n. al. er. io. realizations following seasoned equity offerings? Journal of Financial and Quantitative Analysis 36, 169-193.. Ch. engchi. i n U. v. Ewert, Ralf, and Alfred Wagenhofer, 2005, Economic effects of tightening accounting standards to restrict earnings management, The Accounting Review 80, 1101-1124. Fama, Eugene F., 1980, Agency problems and the theory of the firm, Journal of Political Economy 88, 288-307. Fama, Eugene F., and Kenneth R. French, 2002, Testing trade‐off and pecking order predictions about dividends and debt, Review of Financial Studies 15, 1-33.. 34.