Are CEOs relevant to capital structure?

Transcription

1 Are CEOs relevant to capital structure? Hursit Selcuk Celil Peking University Daniel Sungyeon Kim Peking University December 19, 2014 Abstract This paper studies how capital structure is affected by CEOs. Using the sample of S&P1,500 firms during 1992 to 2013, capital structure changes substantially in the year of CEO replacement, and stays relatively stable for the rest of the CEO s tenure at the firm. We further find that the change in the capital structure is driven by the outside CEOs, but not by the inside promotion. We find that firms with CEO turnover experience faster speed of adjustment post-turnover, and that the absolute deviation from the target leverage drops significantly. We propose that change in leverage in the initial year is not due to time-invariant CEO characteristics, but by the new CEO s attempt to close the gap between the target leverage and actual leverage during his predecessor. 1

2 1 Introduction What derives firm s capital structure decisions? This research question long has been debated since Modigliani and Miller (1958, 1961) first seminal work appeared in literature. Both empirical and theoretical side of financial economics research tries to understand the main determinants of corporate capital structure. Today time-variant factors, i.e. profitability and tangibility, and time-invariant factors, i.e. firm fixed-effects, are documented to have statistically significant association with firm debt ratio (Rajan and Zingales (1995), Flannery and Ragan (2006)). Recent work of Lemmon, Zender and Roberts (2008) shows that timeinvariant factors, such as firm s initial leverage, manage to capture most of the variation in firm s capital structure dynamics. They argue that there is a huge persistency in firm leverage ratio over time. On the other hand, DeAnglo and Roll (2014) argues in the long-run firm changes their capital structure and such decisions are time-variant. In this paper, we try to reconcile these two arguments by examining corporate executive officer (CEO) turnover events in the sample of US firms. Our motivation relies on the fact that corporate policy choices, i.e. financing and investment, are decided by the management team of a firm. These decisions are not solely based on prospects of certain policies but also depends on expertise levels, characteristics, and incentives of team members. For instance, Chava and Purnandam (2010) analyzes CEO s compensation based incentives on some corporate financial policies, and documents that CEO s incentives are positively associated with firms leverage. Coles, Daniel, and Naveen (2006) also shows that risk-taking incentives are associated positively with risk profile of corporate investment and financing decisions. In a survey, Graham and Harvey (2001) show that firm s financial policies are affected by CFO s or CEO s personal characteristics such as the age and education (MBA degree) of the executives. In this context, when CEO of a firm is replaced by another executive, firm is exposed to different set of characteristics of a 1

3 management team, and hence, its corporate policy choices may reflect this alternation. Figure 1 further motivates our research question by showing how leverage changes around CEO turnover. We plot average of book leverage in the sub-sample of our firms where, 1) CEO turnover takes place and financial data is available from at least five years prior to the CEO turnover and until at least five years after the CEO turnover, and 2) within the 11-year window there is no other CEO turnover event except for the one in Year 0. The solid line represents the firms whose leverage decreased in the year of CEO turnover, and the dashed line represents the firms whose leverage increase in the year of CEO turnover. The striking fall (jump) in leverage in the year of CEO turnover is preceded and followed by years of relatively stable level of leverage. It implies that significant and meaningful changes in leverage happens in the year of CEO turnover. First using the sub-sample of S&P 1,500 firms with CEO turnover and their matched counterparts, we test whether CEO turnover is affects corporate capital structure. We first show that, in the year of CEO turnover, firms with CEO turnover exhibit significantly different pattern of leverage relative to the firms without CEO turnover. Leverage jumps or drops significantly for the firms with CEO turnover, but their matched counterparts doesn t exhibit as distinct change in leverage in the same year. We further show that relative to the firms without CEO turnover, the firms with CEO turnover close the gap between the target leverage and actual leverage more quickly. This finding is further supported by the fact that absolute deviation from target leverage drops significantly after CEO turnover, but it stays relatively stable for the firms without CEO turnover. Finally, we also show that firms with CEO turnover experience faster speed of adjustment relative to firms without CEO turnover after the turnover takes place. We also find that the initial leverage that CEO takes in the first year of his CEO post is the most important factor explaining the level of leverage. Previously documented initial IPO leverage of a firm becomes only marginally significant when considered together with 2

4 CEO initial leverage, suggesting that it may be the time-invariant CEO characteristics not firm characteristics that governs the leverage choices. We also find that the CEOs who are hired from outside are the ones who drive the change in the leverage. However, we do not find support for any time-invariant CEO characteristics that affects corporate leverage choices. Rather, we find that the new incoming CEO adjusts capital structure such that the gap between the target leverage and actual leverage can get narrowed down. Our paper falls within the extensive literature on capital structure. Most notably, Modigliani and Miller (1958, 1961) irrelevance proposition raises stream of research that asks whether capital market imperfections make firm value depend on its capital structure. Along this line main research question argues if firm select target debt-equity ratio by trading-off benefit and cost of leverage. Myers (1984) and Myers and Majluf (1984) are proposed pecking order theory. In brief, three sources of funding available to a firm to pursue its investment projects: internal capital, debt, and equity. There is no adverse selection problem associated with very former one. However, outside investors view debt has relatively less adverse selection problem than equity, and hence equity is associated with higher premium than debt. Within this context, investment generally first financed with firm s internal capital, if it is insufficient safe debt will be used. Equity financing will be used if only if both retained earnings and debt is inadequate. In summary, a firm has no strong preference on achieving to certain target leverage ratio, and hence firm s financial deficit is a main factor that explains the capital structure of a firm. Baker and Wurgler (2002) argues that a firm s observed capital structure is derived by market value of its security, which means share prices, managers tend to issue (repurchase) outstanding shares when its market-to-book ratio is high (low). Therefore, managers time the market and there is no target leverage ratio. Welch (2004) proposes managerial inertia permits stock price changes affect the market value of firm s leverage ratio. In this regard, long-run dynamics of firm s leverage ratio is associated with it s stock price dynamics. 3

5 Related studies are motivated to test the implications of these corresponding models with empirical specifications. Harris and Raviv (1991) and Rajan and Zingales (1995) study the determinants of firm s leverage and leverage adjustment behavior. They document time variant characteristics of a typical firm has statistically significant and economically meaningful explanatory power in explaining its capital structure. Frank and Goyal (2003) tests the predictions of pecking order theory of corporate leverage on cross-section of publicly traded firms from 1971 to They document that financing deficit of a typical American firm is associated with firm s net equity issuance. However, in terms of explanatory power of financing deficit is documented to be weak net debt issuance over time, specifically during 1990s. Flannery and Rangan (2006) examines whether a typical firm has target capital structures, and determines speed of such adjustment when it deviates. Their main finding suggests a partial adjustment towards a target ratio with a speed of one-third of the difference between the actual and its target debt level. Variables of pecking order theory and market timing by Baker and Wurgler (2002) have statistically significant explanatory power in explaining target leverage ratio, but their economical power are not strong enough targeting behavior of a mean sample firm. Finally, they share price fluctuations have minor and short-term effect on market debt ratio. On the contrary, Lemmon, Zender and Roberts (2008) argues that variation in leverage ratios is driven by an time-invariant unobserved heterogeneity across firms. Previously identified time-variant factors are performing weak in explaining features of leverage. Their main conclusion is that variation in capital structure of a mean firm is primarily explained by firm characteristics that remain persistent over the long run. DeAngelo and Roll (2014) argues time variation in leverage cross sections increases with the time between cross-sections. They show a majority of the firms in their sample have high and low leverage ratios overtime. Persistency in leverage mostly observed at low leverage. They conclude model that does not rely on the assumption on maintaining leverage ratio that is preset will perform well in capturing 4

6 the cross sectional variation of leverage. Harford, Klasa and Walcott (2009) examines capital structure targeting behavior of a firm in the context of large acquisitions. Specifically, they study how bidding firm adjust its capital structure pre and post acquisition period. They show that when a typical firm is over levered ex-ante chooses a equity financing rather than debt financing to acquire a firm with respect to its target level. In addition, they find a positive relationship between changes in target level that is due to an merger event and actual leverage. Bidder firm actively move back to its target leverage within five years if the merger event is financed by debt. DeAngelo, DeAngelo and Whited (2011) documents that firm issues temporary debt to finance investment and deviate from its target temporary. Their model suggests observed slow speed of adjustment to target is due to firm s movements away from target but not to transactional costs. On the other hand Faulkender, Flannery, Hankins and Smith (2012) suggests leverage target and heterogeneity in adjustment speeds are associated with adjustment costs, which is measured by firm s cash flow. 1 They also show that financial constraints also affect significantly the speed of adjustment toward target. Although effects of market variables on adjustment behavior of a firm are less important than financial constraints, they document under-levered firms (over-levers firms) close less (more) of the leverage gap between actual and target leverage when share prices are high relative to book. In the related literature, Frank and Goyal (2007) find that leverage declines after forced CEO turnover. Furthermore, they find that pay-for-performance is associated with speed of adjustment. Cao and Mauer (2010) empirically examines the effect of CEOs on corporate financial policy. Specifically, the paper focuses on firms that change their leverage from zero to positive and vice versa. Findings of the paper external CEO turnover, which is defined by replacement of the CEO with an outsider, is one of the key determinants of the change in financial policy of a firm. Furthermore, the paper also shows that frequency of CEO turnover 1 See also Fisher, Heinkel, and Zechner (1989), Leary and Roberts (2005), and Strebulaev (2007). 5

7 is negatively associated with the changes in firm s capital structure. Cronqvist, Makhija, and Yonker (2012) finds that CEO s personal leverage preferences are carried over to the firm s leverage decisions. In the related literature, The remaining of the paper is organized as follows. Section II explains data sample and variable construction. Section III provides our empirical methodology. We present our results in Section IV, and Section V concludes the paper. 2 Sample and data description We start our sample with all firm-years in the ExecuComp database between 1992 and Among these firm-years, we only keep the firm-years that can be matched to COMPUSTAT- CRSP merged database, and require that the firms have reported SIC code either in CRSP or COMPUSTAT. We then exclude financial firms (firms with SIC code between 6000 and 6999), and firms from regulated industries (SIC code between 4900 and 4949). We then require that chief executive officer be identified from the database. If we cannot identify any CEO for a sample firm in a fiscal year, we seek the missing CEOs by 1) using the date an executive became CEO, 2) using annual title of an executive, or 3) searching the definitive proxy statement of the firm for the year when a CEO is missing. When a firm-year has more than one CEOs, we use pick a single CEO by 1) searching the definitive proxy statement or biography of the CEOs to eliminate executives with false CEO indicator, 2) using the date an executive leaves the CEO position, and 3) if there are co-ceos in a fiscal year, I eliminate the one that is appointed later as a CEO. In the year of CEO turnover, we identify the CEO that has served more than six months as the CEO during the year. 2 Table 1 provides descriptive statistics of the main variables used in this study. Panel A 2 For example, if a CEO was replaced in the eighth month of the fiscal year, we take the replaced CEO as the CEO during the year, and the replacing CEO will be considered CEO of the firm from the next fiscal year. 6

8 presents the statistics for the full sample of 31,361 firm-years, whereas Panel B presents the statistics for the sub-sample of firm-years around CEO turnover. Book leverage is the total debt scaled by the total asset (AT), where total debt is the sum of the debt in current liabilities and non-current debt (DLC+DLTT), Market leverage is the total debt divided by total asset and market value of equity (MKVALT), Ln(Sales) is the natural logarithmic transformation of revenue, Market-to-book is the sum of market value of equity, total debt, and preferred stock(pstkl) less deferred taxes and investment tax credit (TXDITC) divided by the book value of total asset, Profitability is operating income before depreciation (OIBDP) divided by total asset, Tangibility is net property, plant, and equipment (PPENT) divided by total asset, Dividend payer is one if the company has positive total dividend declared (DVT), and zero otherwise, Cash volatility is standard deviation of historical profitability (with at least three historical values), Financing deficit is increase in total asset less increase in retained earnings (RE) scaled by total asset, CEO equity comp. is the stock awards and option awards to the CEO divided by total compensation to the CEO (TDC1), and CEO tenure is the number of years CEO has been in post since the first year of appointment as the CEO. 3 Most variables in Table 1 are close to the statistics reported in the literature, and the values are not substantially different between Panel A and Panel B, except for Dividend payer. The sub-sample of firm-years with CEO turnover are more likely to payout dividends relative to the full sample. Table 2 presents the time-invariant characteristics of the CEOs in our sample. In the full sample, we have 3,787 unique CEO-firm observations, while in the sub-sample with CEO turnover, we have 646 unique CEO-firm observations. CEO age is the age of the CEO in the year when he was appointed as the CEO of the firm, CEO gender is one if the CEO is male, and zero otherwise, CEO honor is one if the CEO uses prefix other than Mr. or 3 Variable definitions are also provided in Table A1. 7

9 Ms, Outside CEO is one if the CEO was appointed within two years from joining the firm, Number of companies worked is the number of companies the CEO appears as an executive in the ExecuComp database, Number of CEO posts is the number of companies the CEO appears as a CEO in ExecuComp database, Years before replacing is the number of years preceding CEO appears in ExecuComp, Years until being replaced is the number of years the CEO appears as CEO of the firm in ExecuComp, and CEO initial leverage is the book leverage of the firm in the year CEO was appointed as the CEO. The most notable difference between Panel A and Panel B is that the CEOs in the subsample are less likely to be outside CEOs (0.35 vs. 0.43), less likely to have worked for other firm as an executive (1.27 vs 1.34) or as a CEO (1.07 vs 1.12), and serves more years in the firm as a CEO (7.67 vs. 3.49). 4 3 Empirical methodology Our base model for the analysis modifies the conventional models of capital structure to incorporate CEO characteristics as below: Lev it = α+β X it 1 +γ Y it 1 +δ µ+ε it, (1) where Lev it is the book leverage of firm i in time t, X is the vector of lagged CEO characteristics, both time-invariant and time-variant, Y is the vector of lagged firm characteristics, µ is the vector of fixed effect (could be any one of year, firm, and CEO fixed effects), ε i t is the error term from the model. Time-variant CEO characteristics in the model are the characteristics that maybe linked to the risk-taking incentives of CEOs. We first consider proportion of CEO s total compen- 4 These difference may have been caused by our CEO selection method in Panel B, where we require the preceding CEO be in post at least five years prior to being appointed, and also that the CEO has to be in post at least five years after being appointed. 8

10 sation paid in either restricted stocks or stock options (CEO equity compensation). Both equity compensation and leverage offers non-linear, convex payoff to CEOS, incentivizing the CEO to take on more risk. As CEO compensation is determined mostly by mutual agreement between the CEO and the board of directors, the compensation breakdown between salary, cash bonus, and equity compensation should reflect CEO s personal preferences. If a CEO has higher portion of his compensation in equity compensation, it may be that he is likely to take on more risk, which may eventually affect his financing decisions (Chava and Purnandam, 2010; Coles, Daniel, and Naveen, 2006). We next consider CEO tenure as the CEO characteristics that may affect CEO s leverage decisions. A CEO with short tenure still have enough time at the CEO post to benefit from long-term planning and taking risks, so the CEO may be more willing take on more risk. However, as CEO becomes more incumbent or it gets closer to the retirement age, the CEO may be less likely to take on risk (Chakraborty, Sheikh, and Subramanian, 2007). We then consider the time-invariant CEO characteristics that may affect CEOs leverage decisions. Many studies, especially in the behavior area, found that males are more likely to take on risk or seek sensations relative to their female counterpart (Faccio, Marchia, and Mura, 2014; Barber and Odean, 2001; Grinblatt and Keloharju, 2009). Risk-taking and sensation-seeking behaviors both can be linked to the decisions to take on more risks through corporate leverage. As such, we expect CEO s gender should play a role in explaining crosssection of corporate leverage. Next, we consider the name prefix of CEOs. ExecuComp reports the name prefix of the executives, and the vast majority of the executives use Mr. or Ms. as their prefix to the last name. Other name prefix, which we will refer to honorable prefix, that ExecuComp reports are Admiral, Captain, Colonel, Dr., General, Governor, Honorable, Lieutenant, Lord, Major, Professor, and Sir. As is evident from the list, the prefix show either one of educational status, social status, or the military rank. While we cannot exclude the possibility that some CEOs may prefer to be just called Mr. or Ms. 9

11 instead of the honorable prefix even if they qualify, it may well be true that CEOs who prefer to have honorable prefix might attach more pride to the status that comes with the honorable prefix. The CEOs with more pride in their prior achievements should be more confident in their decisions, henceforth their leverage decisions can be different from that of other similar firms. In the spirit of LRZ, the final measure we use as time-invariant CEO characteristic is the leverage of the firm in the year the CEO is first identified as influential CEO (we define influential as being a CEO for at least half of a fiscal year). After showing that firm fixed effects adds substantial goodness-of-fit to leverage models, LRZ shows that firm s initial leverage in the year of IPO is the single most significant factor in explaining the level of leverage. Similarly, we first test whether CEO fixed effects improves the leverage model even further beyond firm fixed effects, and then use the firm leverage in the first year of CEO post as the time-invariant CEO characteristic. For firm characteristics, we follow Lemmon et al (2008), which we will denote LRZ from this point, for the base conventional model: we employ lagged values of logarithmic transformation of sales, market-to-book, profitability, and tangibility. We expect that the book leverage is higher as the firm size is larger, as the growth opportunity is smaller, as the profitability is lower, and as asset tangibility is higher. 5 From Flannery and Rangan (2006) and LRZ, more academics have focused in explaining cross-sectional difference in leverage. However, as DeAngelo and Roll (2014) point out, not only the cross-sectional variation in leverage, but also time-series change in leverage is important in understanding corporate financial decisions. As such, we also consider following model for change in leverage: Lev it = α+β 1X 1 it 1 +β 2 X 2 it 1 +γ Y it 1 +δµ+ε it. (2) 5 In Table 3, cash flow volatility, financing deficit, and the dummy for paying dividends, but as these variables offer little improvement over the first four variables, we exclude them in further analysis. However, in untabulated results, even with these three additional variables, the results are qualitatively similar. 10

12 In this model, X 1 is the vector of time-variant CEO characteristics which cannot be differenced (such as CEO tenure), X 2 is the vector of other time-variant CEO characteristics, Y is the vector of firm-characteristics, and δ is the year fixed effect. In this model, we do not include the time-variance CEO characteristics considered in the regression for the level of leverage, however. Increasingly popular models in capital structure are partial adjustment models. In this study, we follow the model in Flannery and Rangan (2006): Lev it = α+(1 λ)lev it 1 +λβ Y it 1 +ε it, (3) where λ is the speed of adjustment, and β Y it 1 is the target leverage of the firm given the lagged structural variables of leverage. For the purpose of our research, as in Faulkender et al (2012), we first estimate the target leverage, L = β Y it 1, from Lev it = α+β Y it 1 +ε it, (4) and then apply the target leverage to Equation 3 to estimate the speed of adjustment. There are a major difference in our model from that in Flannery and Rangan (2006). As we need to evaluate the target leverage and speed of adjustment by year, we run above specification by fiscal year. And because we run the model by fiscal year, we do not use any year or firm fixed effects. Thus the estimation of target leverage and speed of adjustment in our model is from cross-sectional regressions only. As in the most of the corporate finance research, our results may suffer from possible endogeneity as the book leverage and other firm performance or characteristics may be endogenously determined. As such we employ propensity score matching method. For propensity score matching, we match our sample firm-years with CEO turnover with other firm-years without CEO turnover. Our sample firms are the ones 1) that experience 11

13 CEO turnover during 1997 and 2007, 2) that didn t experience other CEO turnover during four years prior to and five years after the said turnover, and 3)that are the S&P 1,500 firms excluding the firms in financial and utilities industry. Our candidates for matched firms are from S&P 1,500 firms excluding the firms in financial and utilities industry, which didn t experience any CEO turnover during four years prior to and five years after the year of CEO turnover in our sample firms. Once we have the sample firms and all possible matched firms, we obtain the estimation results from Turnover it = α+βy it 1 +γz +ε it, (5) where Turnover it is one for our sample firms, and zero for match candidates, Y is the vector of firm characteristics that may be essential in finding a better match of our sample firm. We use the the firm size, market-to-book, profitability, asset tangibility, and deviation from the target leverage as the factors to calculate propensity score. Z is the industry fixed effects based on Fama-French 48 industry classification. We estimate the model by each fiscal year, and obtain the predicted values, which is the propensity score. Finally, we then find the matched firms from the same fiscal year that has the closest propensity score as our sample firms, where the difference in propensity score shall not be larger than 0.1. These matched firms are the ones who might have replaced the CEO in the same year as the ones in our sample firms, but who did not replace the CEO. Once we find our matched firms, we compare the corporate financing decisions of our sample firms and that of the matched firms to exhibit different financial decisions the two groups make. Because our sample firms and matched firms are very similar in many aspects of firm characteristics, but differ in the decision to replace CEO, the results from propensity score matching analysis is less subject to endogeneity concern. 12

14 4 Empirical results Before we begin our main analysis on CEO and leverage choices, Table 3 provides preliminary results on empirical model of leverage in Lemmon et al (2008). Models (1)-(3) and Models (5)-(7) only uses the factors used in LRZ and year or firm fixed effects, whereas Models (4) and (8) replaces firm-fixed effect with CEO-fixed effect. As in LRZ, in our sample, adding firm-fixed effect to the model significantly improves the fit of the model (adjusted R 2 increasing from to with Ln(Sales), market-to-book, profitability, and asset tangibility, and increasing from to when additional cash flow volatility, financing deficit, and dividend payer are put in the model). This strongly supports LRZ argument that some time-invariant firm characteristics may be in place in determining firm leverage. Interestingly, in Models (4) and (8), we show that CEO-fixed effect improves the fit even further with adjusted R 2 of and 0.729, respectively. These results suggest that at least some portion of time-invariant factor that governs firms financing decisions may be associated with time-invariant CEO characteristics. As adding additional three independent variables do not substantially improve the model, we only consider the first four variables in Models (1) to (4) in further analysis. Also, following LRZ, we mainly use year fixed effects in further analysis, as firm fixed effects or CEO fixed effects cannot be used together with time-invariant firm or CEO characteristics. Figure 1 also supports association of firm leverage and CEOs. The figure shows how leverage changed during 11 years around CEO turnover by normalizing the level of leverage by the leverage five years prior to the CEO turnover (Year -5). We define Year 0 as the first year that the new CEO becomes influential, in other words, the first year that the CEO serves at least half of the fiscal year. The blue solid line represents the firm-years that the leverage decreased in Year 0, and the red dashed line represents the firm-years that the leverage increased in Year 0. While we can identify significant change in the leverage in 13

15 Year 0 for both groups, the leverage doesn t show as significant change in other years around CEO turnover, supporting our argument that certain time-invariant CEO characteristics may affect the corporate financial decisions. 4.1 Level of book leverage Motivated by the Figure 1, we first investigate whether the observed pattern in leverage is unique to our sample of firms with CEO turnover or a universal pattern across all US firms. We use propensity score matching method to find matching firms of our sample firms based on: 1) matching firms should not have any CEO turnover within five years prior to and five years after the fiscal year of the CEO turnover of our sample firms, 2) financial and executive compensation data should be available during the 11-year window, and 3) propensity score must be within 0.1 from that of our sample firm. If there are multiple firms that satisfy above criteria, we pick the one with the closest propensity score. The propensity score is obtained as the predicted value from the Equation 5, where we include lagged values of book leverage, size, market-to-book, profitability, tangibility, and deviation from the target leverage, and the model is estimated with industry fixed effect based on Fama-French 48 industry classification. All variables except for the deviation from the target leverage are obtained from COMPUSTAT, while the deviation from the target leverage is obtained from Equation 4 using the variables proposed in Flannery and Rangan (2006). Table 4 presents the descriptive statistics for our sample firms and the matched firms. As we can observe from the table, two groups are very similar in every aspect except for the percentage of firms that pay dividends (65% vs. 50%). But as dummy for dividend payer was not a significant factor in Table 3, we don t place too much emphasis in this difference. As such we can assume that the conditions for financing decisions should be very similar between the two groups, only differing in the fact that our sample firms have replaced CEO in the middle of the sample period. 14

16 Figure 2 presents the leverage changes of the matched firms and our sample firms around the CEO turnover of our sample firms separately for the group which experienced decrease in leverageinyear0andthegroupwhichexperiencedincreaseinleverageinyear0. Figure2(a) represents the group with sudden drop in leverage in Year 0. Our sample firms experience earlier in the period and stay stable until Year -1. Then in Year 0, there s a sudden drop in leverage, and stays relatively stable at the level. The matched firms, however, drops by lesser magnitude in Year -4 and -3, and then stays stable around the leverage level of Year -5 for the rest of the window. In 2(b), we show the group with increase in leverage in Year 0. The contrast between the firms with CEO turnover and matched firms are even more distinct. Our matched firms leverage actually stays relatively stable over the whole window, while the firms with CEO turnover experience huge jump in Year 0 and stays stable in other years in the window. While the evidence in Table 3 suggests that time-invariant CEO characteristics may influence a firm s leverage decisions, it is a difficult task to suggest which specific CEO characteristics are in action. In this study, we focus on possible CEO characteristics that may be associated with the CEO s risk-taking behavior: gender, honor, equity compensation, and tenure. As taking leverage, on average, will increase the risk-profile of the firm, the decision of CEO whether to take on more debt or not maybe positively linked to his tendency for risk taking. Lev it = α+β 1 Time-invariant CEO characteristics (6) + β 2 Time-variant CEO characteristics it +Firm characteristics it +ε, Table 5 shows the results of our regression with CEO characteristics. Model (1) only includes the time-invariant CEO characteristics, Model (2) only includes the time-variant 15

17 CEO characteristics, Model (3) only includes main firm firm characteristics, and Model (4) includes all variables in Models (1)-(3). We also consider the initial leverage in the year of new CEO as well as the IPO initial leverage from LRZ. Model (1) supports our prediction as firms with CEOs using name prefix other than Mr./Ms. have lower book leverage. The initial leverage in the year the CEO takes the post is also highly significant in explaining the level of book leverage, however, the CEO gender doesn t seem to play a role. In Model (2), CEO tenure takes significant negative loading, suggesting that as CEO becomes more incumbent, the CEO is more likely to take lower book leverage. In Model (3), Ln(Sales), market-to-book, and tangibility significantly explains the level of book leverage. As in LRZ, initial leverage of the firm in the year of IPO is also highly significant in explaining the book leverage. But noting the substantial difference in adjusted R 2 between Model (1) and (3), the association of IPO initial leverage maybe dominated by the CEO initial leverage. In Model (4), we document significance of CEO initial leverage in explaining the level of book leverage. Considered together with CEO initial leverage, IPO initial leverage is only marginally significant, while other firm characteristics are still highly significant. The result suggests that the time-invariant firm characteristic maybe the time-invariant CEO characteristic rather than any firm characteristics. CEO tenure also loses explanatory power in Model (4) while still being marginally significant. Comparing Model (4) to Model (2) from Table 3, CEO characteristics, mostly the CEO initial leverage, improves the explanatory power of the leverage model substantially (adjusted R 2 of vs ). 16

18 4.2 Change in book leverage Next we consider possible factors that may cause the change in leverage ( Lev it ). Lev it = α+β 1 Time-variant CEO characteristic it +β 2 CEO change dummy + β 3 Outside CEO*CEO change dummy+γ Firm characteristics it +δyear+ε, where Time-variant CEO characteristics include CEO tenure and equity compensation, CEO change dummy is one if a new CEO became influential in the year, and zero otherwise, and Outside*CEO change dummy is the interaction term of the indicator for the new CEO being hired from outside and the indicator for the CEO s first influential year. 6 We do not include the indicator for the CEO being hired from outside, as we only intend to include time-variant CEO characteristics, whereas the indicator for outside will be fixed for the tenure of a CEO. Firm characteristics include Ln(Sales), market-to-book, profitability, asset tangibility, the previous year s deviation from the target leverage, and the interaction term of the previous year s deviation from the target leverage and the CEO change dummy. Table 6 presents the results of change in book leverage on CEO and firm characteristics. Model (1) suggests that change in leverage over time is influenced by change in equity compensation and CEO tenure. When CEO is given more equity compensation, it is highly likely that the CEO will increase the firm leverage. And as in Table 5, as CEO becomes more incumbent, leverage decreases more over time. However, we don t find any significant association between the change in leverage and CEO turnover. However, the dummy for CEO change and its interaction term with the new CEO coming from outside doesn t seem to have any explanatory power. In Model (2), increase in growth opportunities will lead to decrease in leverage, and increase in tangibility will lead to increase in leverage, as in Table 6 In Table 6, Change in equity compensation is the change in the proportion of equity compensation in total CEO compensation ( Equity t Equity Total Compensation t 1 t Total Compensation ). t 1 17

19 5. Interestingly, the change in leverage is significantly negatively associated with lagged deviation from the target leverage. It suggests that firms do seek to close the gap between the target leverage and actual leverage. In Model (3), we combine all variables in Models (1) and (2) and add the interaction term of the dummy for CEO change and the lagged deviation from the target leverage. In this specification, all variables that were significant in explaining change in leverage remain significant, and the interaction term of outside CEO and the CEO change dummy is positively significantly associated with CEO change. Thus the results show that in the year of CEO change, if the CEO is from outside, on average he will increase the leverage. This result supports our hypothesis that corporate leverage should change significantly in the year of CEO turnover. 4.3 Deviation from target leverage and speed of adjustment In the previous section, we present supporting evidence that the corporate leverage may change when the CEO turnover takes place, especially when the new CEO comes from outside the firm. In this section, we investigate compare our sample firms and their matched counterparts in other aspects of the book leverage: deviation from the target leverage and speed of adjustment. We first test whether the convergence of leverage documented in LRZ is different across our sample firms and the matched firms. LRZ use the book leverage in the beginning year of the sample to sort firms into four groups, but we use deviation from the target leverage in the year prior to the CEO turnover to sort the companies into four groups. We modify the method because, unlike LRZ, we are more interested in how corporate leverage changes around the event of CEO turnover, and we believe the financing policies should be more sensitive to the degree of deviation from target leverage rather than the level of leverage. Figure 3 presents how book leverage moves around the year of CEO turnover. Q1 is the group of firms with most negative deviation from the target, and Q4 is the group with 18

20 most positive deviation from the target. On average, the firms with more positive deviation from the target will have higher book leverage, and the firms with more negative deviation from the target will have lower book leverage. When sorted by the deviation from the target leverage in the year prior to CEO turnover, our sample firms with CEO turnover, depicted in Figure 3(a), close the gap quite rapidly over the next five years. It is interesting to note that the leverage of Q1 and Q4 has been deviating away from the center until Year -1. The middle two groups leverage stays relative stable over the 11-year window. Figure 3(b) depicts similar graph for the matched firms. Relative to our sample firms with CEO turnover, the convergence is rather slow after Year 0. This is especially true for Q1. This suggests that when leverage is too high, both new CEOs and continuing CEOs try to lower down the leverage. But when leverage is too low, it is more likely that the new CEOs take on more leverage, while continuing CEOs only do so in very slow pace. We next consider how the deviation from the target leverage changes over time for the twogroupsoffirms. Forthefair comparison, wetake theabsolutevalueofdeviation from the target leverage, and plot the value for our sample firms and the matched firms over 11-year window. Figure 4(a) depicts the value for the two groups. The firms with CEO turnover exhibit significantly different behavior than the firms without CEO turnover. While the absolute deviation from the target leverage is quite similar between the two groups before Year 0, the firms with CEO turnover significantly reduces the deviation after the turnover, but the deviation of firms without CEO turnover doesn t improve over the same period. This results further supports above finding that new CEOs may work to close the gap between the book leverage and target leverage while continuing CEOs may be reluctant to do so. We then calculate speed of adjustment for each of the group. We plug in the target book leverage, Lev in: Lev it = (1 λ)lev it 1 +Lev it +ε it, (7) 19

21 where λ is the speed of adjustment. As in the case of target leverage, we estimate Equation 7 by year to assess the change in speed of adjustment over time. Figure 4(b) depicts the speed of adjustment for the two groups within the 11-year window around CEO turnover. Although the results are somewhat noisy, the firms experiences faster speed of adjustment after the CEO turnover relative to the firms without CEO turnover. Prior to CEO turnover, the speed of adjustment between the two groups are comparable, and especially in the year of CEO turnover, the speed of adjustment in the group with CEO turnover is slower than their matched counterparts. This figure also supports the our story that firms become more flexible in changing leverage after the event of CEO turnover. 4.4 Debt maturity So far, we have documented the association between CEO characteristics and total book leverage, change in book leverage, deviation from the target, and the speed of adjustment. We further investigate whether the CEO characteristics affect the type of debt the firm is taking, specifically the maturity structure of the debt. In Table 7, we repeat the analysis in Tables 5 and 6 with dependent variable now being leverage in three different maturity types: short-term, mid-term, and long-term. We define short-term debt as the debt maturing in less than a year (DLC), mid-term debt as the debt maturing in two to five years (DD2, DD3, DD4, and DD5), and long-term debt as the non-current debt less the mid-term debt (DLTT - Σ 5 T=2 DD T). The first three columns use the level of leverage as the dependent variable, and the last three columns use the change in leverage as the dependent variable. As in Table 6, the time-varying factors of the last three columns are the changes from the previous year. As in previous tables, CEO initial leverage is highly significant in explaining the level of leverage of all maturities. The level of CEO equity compensation is significantly positively associated with long-term leverage, which is consistent with our intuition that equity compensation should be linked to CEO s long-term risk-taking decisions. Furthermore, the 20

22 equity compensation is significantly negatively associated with short-term leverage, indicating the less a CEO is compensated with long-term compensation, the CEO is more likely to increase the short-term risk-taking behavior. CEO tenure is significantly negatively associated with short-term leverage and long-term leverage, which indicates that the more incumbent a CEO is, the less amount of leverage the CEO will take. In the last three columns of table, change in book leverage in all maturities is significantly negatively associated with initial CEO leverage. If a CEO starts off with higher leverage, the CEO will be decreasing leverage over time, and vice versa. As before, change in equity compensation is not significant in explaining change in book leverage. Interestingly, if CEO tenure is higher, the CEO will lower the long-term leverage, while the CEO will increase the short-term leverage. This result shows that the CEO s career plan also affects the firm s leverage choices. As CEO s get closer to the retirement, the CEO will consider less about the long-term aspect of the firm, but the CEO will take short-term risks. ButcomparedtoTable5, themodelsintables6and7showsubstantiallylowergoodnessof-fit. Recalling Figure 1, where leverage trend of the two groups move mostly in the opposite direction, the weaker goodness-of-fit maybe comes from positive changes and negative changes canceling out each other. In Table 8 provides the results when we divide the sample into two cases: 1) firm-years when the leverage increases, and 2) firm-years when the leverage decreases. CEO initial leverage is again significantly associated with change in leverage in all models except when short-term leverage is decreasing. However, the direction is opposite from the previous results: when the leverage in the first year of CEO is high, the leverage tends to increase, andwhentheleverageinthefirstyearofceoislow, theleveragetendstodecrease. We believe this is coming from the years of CEO turnover as the leverage change in the year of CEO turnover is so large in magnitude. CEO tenure is again mostly negatively associated with the change in book leverage. When there is CEO turnover, the magnitude of increase 21

23 in overall leverage and mid-term leverage tend to be smaller when the CEO is promoted from the inside. However, when the CEO is hired from outside, the change in leverage is significantly affected by the CEO. The results support our hypothesis that change in leverage may be associated with CEO turnover. 5 Concluding remarks Why and how firms make their corporate financing decisions has always been a hot topic in finance. While extensive theories and empirical studies were suggested explaining how various economic factors (taxes, bankruptcy risk, corporate governance, etc) may affect corporate leverage decisions, still vast majority of variation in corporate leverage remain unexplained (adjusted R 2 less than 20% without firm fixed effect). In this paper, we propose that not only firm specific conditions matter in explaining the corporate leverage, but also the personal characteristics of the CEO plays significant role in determining corporate leverage. Specifically, incumbent CEOs are less likely to change the leverage, while the incoming CEOs are more likely to change leverage. Furthermore, we show that as CEOs are compensated more with equity-based components, the firms are more likely to take on leverage. Finally, we show that once the CEO is replaced, the firms close the gap between the target leverage and actual leverage more quick with higher speed of adjustment. 22

24 Table A1: Formula Value versus Reported Value versus Heuristic Value Variable Short-term debt Mid-term debt Long-term debt Total debt Book leverage Market leverage IPO leverage Market-to-book Profitability Tangibility Dividend payer Cash volatility Financing deficit CEO equity compensation CEO tenure CEO honor Outside CEO Number of companies worked Number of CEO posts Years before replacing Years until being replaced CEO initial leverage Description Debt maturing in less than a year (DLC) Debt maturing in 2-5 years (DD2, DD3, DD4, DD5) Non-current debt (DLTT) less debt maturing in 2-5 years (DD2, DD3, DD4, DD5) Short-term debt (DLC) plus non-current debt (DLTT) Total debt (DLC + DLTT) divided by total asset (AT) Total debt divided by total asset and market value of equity (MKVALT) Book leverage of the firm in the year it first appears in COMPUSTAT Sum of market value of equity, total debt, and preferred stock (PSTKL) less deferred taxes and investment tax credit (TXDITC) divided by book value of total asset Operating income before depreciation (OIBDP) divided by total asset Net property, plant, and equipment (PPENT) divided by total asset One if the company has positive total dividend declared (DVT), and zero otherwise Standard deviation of historical profitability (with at least three historical values) Increase in total asset less increase in retained earnings (RE) scaled by total asset Sum of stock awards and option awards divided by total compensation (TDC1) Difference between current fiscal year and the year CEO first took the post If the CEO uses prefix other than Mr., Ms, Miss If the CEO was appointed within two years from joining the firm Number of companies the CEO appears as an executive Number of companies the CEO appears as a CEO Number of years previous CEO appears in ExecuComp Number of years the CEO appears as CEO in ExecuComp The book leverage of the firm in the year CEO took the post 23

25 References [1] Baker, M., and Wurgler, J. Market timing and capital structure. Journal of Finance 57 (2002), [2] Barber, B., and Odean, T. Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics 116 (2001), [3] Cao, C., and Mauer, D. Ceo turnover and debt policy change. Working paper, [4] Chakraborty, A., Sheikh, S., and Subramanian, N. Termination risk and managerial risk taking. Journal of Corporate Finance 13 (2007), [5] Chava, S., and Purnandam, A. CEOs versus CFOs: Incentives and corporate policies. Journal of Financial Economics 97 (2010), [6] Coles, J., Daniel, N., and Naveen, L. Managerial incentives and risk-taking. Journal of Financial Economics 79 (2006), [7] Cronqvist, H., Makhija, A., and Yonker, S. Behavioral consistency in corporate finance: Ceo personal and corporate leverage. Journal of Financial Economics 103 (2012), [8] DeAngelo, H., DeAngelo, L., and Whited, T. Capital structure dynamics and transitory debt. Journal of Financial Economics 99 (2011), [9] DeAngelo, L., and Roll, B. How stable are corporate capital structures? Journal of Finance (Forthcoming), [10] Faccio, M., Marchia, M., and Mura, R. CEO gender and corporate risk-taking. Working paper,

26 [11] Faulkender, M., Flannery, M., Hankins, K., and Smith, J. Cash flows and leverage adjustments. Journal of Financial Economics 103 (2012), [12] Fischer, E., Heinkel, R., and Zechner, J. Dynamic capital structure choice: Theory and tests. Journal of Finance 44 (1989), [13] Flannery, M., and Rangan, K. Partial adjustment toward target capital structures. Journal of Financial and Economics 79 (2006), [14] Frank, M., and Goyal, V. Testing the pecking order theory of capital structure. Journal of Financial Economics 67 (2003), [15] Frank, M., and Goyal, V. The effect of market conditions on capital structure adjustment. Finance Research Letters 1 (2004), [16] Frank, M., and Goyal, V. Corporate leverage: How much do managers really matter? Working paper, [17] Frank, M., and Goyal, V. Capital structure decisions: Which factors are reliably important? Financial Management 38 (2009), [18] Graham, J., and Harvey, C. The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics 60 (2001), [19] Grinblatt, M., and Keloharju, M. Sensation seeking, overconfidence, and trading activity. Journal of Finance 64 (2009), [20] Harforda, J., Klasab, S., and Walcottc, N. Do firms have leverage targets? evidence from acquisitions. Journal of Financial Economics 93 (2009), [21] Harris, M., and Raviv, A. The theory of capital structure. Journal of Finance 46 (1991),