Payout Policy and the Interaction of Firm- and. Country-level Governance

Transcription

1 Payout Policy and the Interaction of Firm- and Country-level Governance Richard Herron May 25, 2017 Abstract For a panel of 1900 firms across 21 countries from 2004 to 2008, the impact of firm- and country-level governance on payout policy is consistent with La Porta et al. (2000). In weak legal regimes, dividend and repurchase payout ratios increase in firmlevel governance. In strong legal regimes, dividend payout ratios decrease in firm-level governance while repurchase payout ratios increase, suggesting a substitution from rigid dividends to flexible repurchases. These results are robust to an instrumental variable (IV) approach and an alternative firm-level governance measure in a larger panel of 9461 firms from 2002 to JEL classifications: G35, G34, G15 Keywords: Payout Policy, Corporate Governance, International Financial Markets Thank you for helpful comments and discussions to Jay Dahya and Jérôme Taillard, as well as seminar participants at Baruch College. An earlier version of this paper circulated as What matters more in governance? Firm or country?. Finance Division, Babson College. 326 Tomasso Hall, Babson Park, MA , United States. Office number Fax number rherron1@babson.edu.

2 1 Introduction A large corporate governance literature examines the role of firm- and country-level governance around the world, but does not fully connect firm- and country-level governance to the mechanisms through which investors ultimately receive returns: cash dividends and share repurchases. I fill this gap by examining payout policy and the interaction of firm- and country-level governance for 1900 firms across 21 countries from 2004 to For this sample, Aggarwal et al. (2011) provide Gov41, a well-established firm-level governance measure. 1 I find that around the world both cash dividends and share repurchases fit a model in which shareholders use additional rights to achieve payout (e.g., La Porta et al., 2000). In weak legal regimes, shareholders use additional rights to extract both cash dividends and share repurchases. However, in strong legal regimes, shareholders use additional rights to substitute from rigid cash dividend policies to flexible share repurchase plans. This substitution is better for both firms and shareholders, but requires a stronger country-level governance environment to ensure payout and avoid expropriation. The relation between payout policy and firm- and country-level governance is interesting for at least two reasons. First, La Porta et al. (1997, 1998, 2000) and Djankov et al. (2008) investigate country-level governance and conclude that some countries offer investors stronger legal protection than other countries. Further, they find that stronger legal protection at the country-level results in greater financial development, less expropriation of shareholder wealth, and greater return of capital to investors via cash dividends. However, the relation between payout policy and firm-level governance in the context of differing country-level governance remains relatively unexplored. Second, given firms and investors residency constraints, firm-level governance improvements may be the most feasible way for firms to commit to return capital and for minority 1 Gov41 is the sum of 41 individual firm attributes expressed as a percentage. These 41 attributes span four areas: board, audit, anti-takeover, and compensation and ownership. With these same data Aggarwal et al. (2011) find a relation between firm-level governance and international institutional holdings and Aggarwal et al. (2010) find a relation between firm-level governance and valuations. 1

3 shareholders to avoid expropriation. 2 The efficacy of firm-level governance improvements, conditional on country-level governance, is not yet fully understood. The extant literature suggests that firm-level governance has little value when country-level governance is low. For example, Doidge et al. (2007) show that firm-level characteristics matter little relative to country-level characteristics in weak legal regimes and that firm-level characteristics matter only once country-level shareholder protection exceeds some threshold. Aggarwal et al. (2010) show that firm- and country-level governance are complements, and that higher country-level governance makes it optimal to invest more in firm-level governance. Aggarwal et al. (2011) show that foreign institutional investors, particularly U.S. institutional investors, demand improvements in firm-level governance. That institutional investors use limited resources to improve firm-level governance suggests that there may be some observable outcome. Mitton (2004) is the closest study to this paper. Using a one-year, cross-sectional analysis of dividend payout policy and firm-level governance for 365 firms, he finds a positive relation between dividend payout ratios and firm-level governance, but only in strong legal regimes. I extend Mitton (2004) along several dimensions. First, I examine both cash dividends and share repurchases. This extension matters as share repurchases occur in more than half of the observations in my sample. Von Eije and Megginson (2008) confirm the importance of share repurchases and show that share repurchases rapidly increase in the European Union (EU) from 1989 to My comprehensive payout policy analysis shows that the relation between firm- and country-level governance is a continuum and that payout policy responses to firm-level governance depend heavily on the country-level legal environment. Second, I analyze a panel rather than a single cross section. This panel provides enough variation in firm-level governance to implement an instrumental variable (IV) approach, as in Aggarwal et al. (2011), to better address the possible endogeneity between payout policy and firm-level governance. Robustness tests with an alternative firm-level governance 2 Another option is cross-listing to an exchange with better country-level governance (e.g., Doidge et al., 2004, 2009). 2

4 measure expand this panel to 9461 firms over 10 years and find similar results. Third, I also examine payout initiation decisions so that I more fully cover the different aspects of payout policy; including payout initiation, payout level, and choice between cash dividends and share repurchases. My panel regressions show that dividend payout ratios increase in country-level governance, consistent with La Porta et al. (2000). Also, there is a positive relation between dividend payout ratios and firm-level governance, and the economic effect of firm-level governance is like that of the country-level governance, consistent with Mitton (2004). However, with both firm- and country-level governance measures in the panel regression, countrylevel governance dominates and the relation between dividend payout ratios and firm-level governance is neither economically meaningful nor statistically significant. This first finding supports the belief that the importance of country-level governance outweighs that of firm-level governance in corporate payout decisions. The finding that country-level governance dominates firm-level governance does not imply that firm-level governance does not affect payout decisions. Further analysis reveals complementarity between firm- and country-level governance with respect to dividend and repurchase levels, as well as dividend and repurchase initiations. I identify this complementarity by splitting the sample on country-level governance and examining the relation between payout and firm-level governance. In weak legal regimes, both dividend and repurchase payout ratios increase in firm-level governance. These positive relations suggest that in weak legal regimes firm-level governance meaningfully increases shareholder rights and facilitates return of capital, consistent with La Porta et al. (2000). In strong legal regimes, dividend payout ratios decrease in firm-level governance, but repurchase payout ratios increase in firm-level governance. These results are consistent with a more nuanced model in which, conditional on strong country-level governance, increases in firm-level governance facilitate a substitution from dividends to repurchases. If shareholders are confident that firms will return capital when appropriate, then both shareholders and 3

5 firms should prefer share repurchases over cash dividends. For shareholders, the advantages of share repurchases relative to dividends are tax related. Shareholders only owe taxes if they choose to participate in share repurchase plans, which allows shareholders to choose the timing of their capital gains and associated taxes. This timing advantage allows shareholders to either offset capital gains with capital losses or significantly delay capital gain taxes. Ignoring this timing advantage, capital gain tax rates are lower than dividend tax rates for 47 % of sample observations. For another 48 % of observations capital gain and dividend tax rates are the same. Even with equal capital gains and dividend tax rates, delayed capital gain taxes have an edge over annual dividend taxes because of the power of compounding. 3 For firms, the main advantage of share repurchases relative to cash dividends is flexibility. Lintner (1956) provides early evidence that managers treat regular dividends as a strong commitment to return capital to shareholders. Brav et al. (2005) show that this strong commitment continues today, while Denis and Osobov (2008) and von Eije and Megginson (2008) show that this strong commitment exists around the world. Repurchases allow firms to return temporary cash flows to shareholders without committing to a regular dividend. This flexibility is value-enhancing; Brav et al. (2005, Table 2) provide survey evidence that 88.1 % of chief financial officers (CFOs) think there are negative consequences to reducing dividends and that 65.4 % of CFOs would not reduce dividends to fund a new project. 4 However, to reduce the strong commitment to return capital that dividends provides and reap the advantages of share repurchases, the firm must provide sufficiently high firm- and country-level governance. Overall, in strong legal regimes improvements to firm-level governance result in lower cash dividends, but higher share repurchases. This is consistent with a substitution from 3 See DeAngelo et al. (2008, 12) for a complete discussion of dividend taxation. 4 Other advantages of share repurchases over cash dividends include exploiting share undervaluation, removing low valuation shareholders, increasing ownership concentration, removing block holders, and preserving executive stock option values. See DeAngelo et al. (2008, 13) for a complete discussion of the advantages of repurchases relative to dividends. 4

6 dividends to repurchases. In weak legal regimes, improvements to firm-level governance result in higher dividends and repurchases. Both are consistent with the La Porta et al. (2000) model and an improved return of capital to shareholders. Although the specifics differ between strong and weak legal environments, in both regimes increases in firm-level governance move firms and shareholders towards the preferred payout policy. These results are robust to an IV approach to address the simultaneous choice of payout ratios and firmlevel governance. These results are also robust to different payout ratio definitions, Tobit regressions to address left-censored payout ratios, and alternative firm- and country-level governance measures. I next investigate payout initiation decisions and find that they are also consistent with a more nuanced model that considers both dividends and repurchases. In weak legal regimes, the relations between firm-level governance and both dividend and repurchase initiations are negative. This suggests that improved firm-level governance enables firms to postpone the return of capital in the presence of profitable growth opportunities. In strong legal regimes, the relation between dividend initiation decisions and firm-level governance is also negative, but the relation between repurchase initiation and firm-level governance is non-negative. This lends further support for a model with substitution from rigid dividends to flexible repurchases. These results are again robust to an IV approach, logit regressions to address binary payout decisions, and alternative of firm- and country-level governance measures. My robustness tests include an alternative firm-level governance measure: the service of either a U.S.- or U.K.-based director on the board of directors of a non-u.s. or non-u.k. firm. Gov41 is a well understood governance index and provides continuous variation in firm-level governance, but it limits the sample to the largest firms in 21 countries. The U.S.- and U.K.-based director indicator variables offer an out-of-sample test because they are available for more firm-years, in both the cross section and the time series. These alternative firmlevel governance results are qualitatively similar to my main results. As with Gov41, the choice of a U.S. director and choice of payout ratios is possibly endogenous and these results 5

7 are robust to an IV approach. Robustness tests also include a within-industry analysis, a matched sample between weak and strong legal regimes, and tests without Japan and U.K. I proceed as follows. Section 2 reviews the related literature and Section 3 describes my data. Section 4 provides the key results with respect to dividend and repurchase payout ratios, as well as dividend and repurchase initiations. Section 5 provides robustness checks, including an alternative firm-level governance measure that allows a larger panel in Section 5.4. Section 6 concludes. 2 Literature Review I connect three literatures that address the impact of country-level governance on corporate decisions, the relation between firm- and country-level governance, and the determinants of payout policy around the world. My contribution is identifying the impact of firm-level governance on payout policy in the context of country-level governance. The first literature focuses on country-level governance and its impact on corporate decisions. La Porta et al. (1998) begin this literature and La Porta et al. (2000) examine the relation between country-level governance, investment opportunities, and dividend payout ratios. They propose two competing hypotheses of dividend payout policy. Under their outcome hypothesis, dividends are the outcome of good governance so that dividend payout is higher in strong governance regimes because this is where shareholders rights are strong enough to force management to return capital. The outcome model predicts that dividend payout decreases in investment opportunities in strong governance regimes because shareholders forgo return of capital to facilitate investment, confident that they will be able to force dividend payout when investment opportunities decrease. Under their substitute hypothesis, dividends are an adaptation that compensates for weak governance so that dividend payout is higher in poor governance regimes. The substitute model predicts an ambiguous relation between growth opportunities and dividend payout because of the tension between 6

8 the need to return capital to investors to initially raise capital and the need retain capital to invest in growth opportunities. La Porta et al. (2000) find support for the outcome model and observe higher dividend payout ratios in strong governance regimes. As well, they find that dividend payout ratios decrease in investment opportunities in strong governance regimes, but are unrelated to investment opportunities in weak governance regimes. It could be that firm-level governance explains some portion of the unexplained variation in payout ratios in weak and strong governance regimes. As well, the impact of firm-level governance could vary significantly across weak and strong governance regimes. Pinkowitz et al. (2006) use Fama and French (1998) valuation regressions to determine the marginal value of cash and dividends in countries around the world. They hypothesize that because minority shareholders fear expropriation from majority shareholders, minority shareholders in weak legal regimes value cash less and dividends more. They find support for both hypotheses. I contribute to this literature by examining how shareholders in different legal regimes use additional rights to influence payout policy, including both cash dividends and share repurchases. The second literature examines firm-level governance around the world, but typically as an effect, rather than as a cause. Aggarwal et al. (2010) look at the amount of firm-level governance around the world and find complementarity between firm- and country-level governance. They posit that firms with lower firm-level governance choose lower firm-level governance because the costs of higher firm-level governance exceed the benefits, given their country-level governance. Aggarwal et al. (2011) connect international institutional investment flows and firm-level governance changes in 23 countries from 2003 to They conclude that changes in institutional ownership positively affect changes in firm-level governance, but not vice versa. They connect firm-level governance changes to observable outcomes, such as CEO turnovers and firm valuation gains. Their firm-level governance measure is Gov41, which I use and discuss in more detail in Section 3. Aggarwal et al. (2011) high- 7

9 light the demand for firm-level governance, which suggests that there may be additional benefits to firm-level governance, including some of the variation in dividend payout policy that La Porta et al. (2000) are unable to explain. Doidge et al. (2007) use the source data for Gov41, as well as Credit Lyonnais Securities Asia (CLSA) firm-level ratings data, and show that country characteristics matters more than firm characteristics in determining firm-level ratings in weak legal regimes. In strong legal regimes, firm characteristics explain a larger fraction of the variation in firm-level ratings. They propose that the net benefit of firm-level investor protection is lower when country-level corporate governance is weaker. This is because firm-level governance is more costly when the country-level legal framework is weak and less beneficial when the country s capital markets are less developed. Thus, these firms invest less in corporate governance. I contribute to this literature by determining if payout policy changes as firm-level governance changes, and if this response varies with country-level governance. The third literature focuses on dividend and repurchase determinants in the U.S. and around the world. Fama and French (2001) investigate the attributes of firms that pay dividends to answer the fundamental question of who pays dividends. Their objective is to better understand the large decline in dividend payers in the U.S. from 66.5 % in the late 1970s to 20.8 % in the late 1990s. They conclude that large, profitable firms are more likely to pay dividends and that firms with better investment opportunities are less likely to pay dividends. This framework identifies two sources of the decline in dividends. First, there was a decline in the fraction of firms that are likely to pay dividends. Second, there was a decline in the propensity of any given firm to pay dividends. They conclude that these two mechanisms are equally responsible for the disappearing dividends trend. DeAngelo et al. (2006) build upon Fama and French (2001) and propose the life-cycle theory of dividends. The life-cycle theory is that young firms have large investment opportunities but little capital, so retention dominates distribution. As firms mature, their investment opportunities decrease and their cumulative profits increase, so they distribute 8

10 excess capital as dividends. DeAngelo et al. (2006) proxy firm life-cycle with the earnedto-contributed-equity ratio, which is the ratio of retained earnings to book value of equity. Firms with relatively low retained earnings will be reliant on external capital to fund investment opportunities and will be less likely to pay dividends. Firms with relatively high retained earnings will be largely self-financing and more likely to pay dividends. DeAngelo et al. (2006) find strong empirical support for their life-cycle theory of dividends and conclude that it exaggerates the Fama and French (2001) disappearing dividends trend. Skinner (2008) complements this literature by documenting a substitution from dividends to share repurchases in U.S. firms. He finds that U.S. firms that only pay dividends are increasingly rare and that repurchases are increasingly used in place of dividends, even among firms that pay dividends. The increasing popularity of share repurchases allows firms to delay dividend initiation and explains a large fraction of the declining propensity to pay dividends that Fama and French (2001) and DeAngelo et al. (2006) document. Skinner (2008) proposes that share repurchases are substitutes for dividends and have become the predominant payout mechanism, with even the major dividend payers returning large amounts of capital via repurchases. These three domestic studies provide the framework for two major international studies. Denis and Osobov (2008) extend Fama and French (2001) and DeAngelo et al. (2006) to five additional major economies (U.K., Canada, France, German, and Japan) and find that firms from these countries also exhibit a declining propensity to pay dividends. Von Eije and Megginson (2008) use a larger sample of 15 countries (European Union members from 1989 to 2005) to examine the evolving nature of repurchases relative to dividends and find that international share repurchases follow the U.S. trend with a ten-year lag. These papers are the first large-scale cross-country applications of propensity-to-pay models, but do not address the impact of firm-level governance on payout choice around the world. I extend this literature with a wider of panel of 21 countries and examine the impact of firm-level governance on both payout level and initiation for both dividends and repurchases. 9

11 Mitton (2004) examines the relation between dividend payout level and firm-level governance and finds a positive relation, but only in strong legal environments. I expand Mitton (2004) in several ways. First, I expand the outcome variables to include share repurchases. Second, I use a large panel of data with both time-series and cross-sectional variation. This facilitates an IV approach to address the possible endogeneity between the choice of firmlevel governance and payout policy. Third, I examine every aspect of payout policy, including payout level and payout initiation. In short, I expand in several meaningful ways the examination of firm-level governance in the context of the country-level governance to determine how both firm-level and country-level governance affects payout policies around the world. 3 Data My main data are Worldscope annual data and I follow Denis and Osobov (2008) and von Eije and Megginson (2008) to generate firm fundamentals. Country-level governance data are from Djankov et al. (2008) and firm-level governance data are from Aggarwal et al. (2011). The Aggarwal et al. (2011) firm-level governance measure Gov41 limits my main analysis to large firms in the largest economies from 2004 to Robustness checks in Section 5.4 relax this limitation with Bureau Van Dyck (BVD) data on boards of directors and use U.S.- and U.K.-based directors as an alternative firm-level governance measure. 3.1 Firm-level Governance Data The Aggarwal et al. (2011) firm-level governance measure is the least complete data, so I begin the data discussion with Gov41. Aggarwal et al. (2011) provide Gov41, which they aggregate from Risk Metrics data. 5,6 Gov41 is the sum of indicator variables for 41 governance 5 The Gov41 data are available here: 6 RiskMetrics covers U.S. firms that are members of any of the following indices: Standard & Poors (S&P) 500; S&P Small Cap 600; and Russell RiskMetrics also covers non-u.s. firms that are members of the following major stock indexes: MSCI Europe, Australasia, and Far East Index (MSCI EAFE), which covers 1,000 stocks in 21 developed countries outside North America; FTSE All Share Index, which consists of FTSE 100, FTSE 250, and FTSE SmallCap indices; FTSE AllWorldDeveloped index, which consists of 10

12 attributes multiplied by 100 then divided by 41, so that Gov41 falls between 0 and 100. These 41 governance attributes fall into four areas: board, audit, anti-takeover, and compensation and ownership. 7 See Aggarwal et al. (2010, 2011) for a complete discussion of Gov41. Aggarwal et al. (2011) document that Gov41 differs between country, with country averages as high as 72.8 % in Canada, but as low as 35.9 % in Greece. Given that La Porta et al. (2000) report a positive relation between country-level governance and dividend payout ratios, this could weaken any finding of a relation between Gov41 and payout policy. My analysis controls for country-level governance and follows the literature with firm random effects (REs) and year fixed effects (FEs) to absorb unobserved heterogeneity Firm-level Fundamental Data Worldscope data from 2003 to 2008 provide firm fundamental variables for sample firms from 2004 to The main dependent variable is dividend payout ratio (D/Y), where dividends (D) is Worldscope s Common Dividends Cash and net income available to common shareholders (Y) is Worldscope s Net Income Before Extraordinary Items. Untabulated robustness checks include other payout ratio denominators: total assets (A) is Total Assets and sales (S) is Net Sales or Revenues. Skinner (2008) shows that share repurchases comprise an increasingly large share of payout in the U.S. and von Eije and Megginson (2008) the largest firms in developed markets; and S&P/TSX index of the Toronto Stock Exchange. 7 Board measures characteristics of the board of directors such as board independence, composition of committees, size, transparency, and how the board conducts its work. Audit measures independence of the audit committee and the role of auditors. Anti-takeover measures dual-class structure, role of shareholders, poison pills, and blank check preferred. Compensation and ownership measures executive and director compensation on issues related to options, stock ownership and loans, and how compensation is set and monitored. RiskMetrics sets each of these 41 indicator variables to one if the firm exceeds some minimum level and zero otherwise. Gov41 is the average of these indicators variables each firm-year, scaled to fall between 0 and 100. Unlike the revised anti-director and anti-self-dealing indices (Djankov et al., 2008), Gov41 has both cross-sectional and time-series variation. Some attributes may matter more than others, but linear combinations of binary governance attributes have a foundation in the literature (e.g., Gompers et al., 2003; Bebchuk et al., 2009). In any case, noise in these measures make it more difficult to find statistical relations. To the extent that de facto governance measures are noisy, a more de jure measure of appointment of U.S. director provides support in Section Von Eije and Megginson (2008) use firm REs and La Porta et al. (2000) use country REs. Section 5.5 provides qualitatively similar results with industry and year FEs. 9 The previous year s data are necessary to generate total asset and sales growth rates. 11

13 show the same trend around the world with a ten-year lag. Following von Eije and Megginson (2008), share repurchases (R) are Common/Preferred Purchased, Retired, Converted, Redeemed. A firm is a dividend payer for a given year if cash dividends are positive and a share repurchaser if share repurchases are positive. In addition to Gov41, the independent variables include the Fama and French (2001) and DeAngelo et al. (2006) payout predictors, which Denis and Osobov (2008) and von Eije and Megginson (2008) show relate to dividend and repurchase decisions around the world. Fama and French (2001) propose that the main determinants of the decision to pay dividends are profitability, growth opportunities, and size, and DeAngelo et al. (2006) add the earnedto-contributed-equity ratio. Following Denis and Osobov (2008), the profitability proxy is return on assets (E/A), which is Earnings Before Interest divided by Total Assets, where Earnings Before Interest is Net Income After Preferred Dividends plus Interest Expense on Debt, if available. The growth opportunities proxies are the market-to-book-assets ratio (V/A is Total Assets minus Common Equity plus Market Capitalization divided by Total Assets ) and total asset growth rate (da/a is change in Total Assets divided by Total Assets ). The size proxy is the size percentile within each country-year based on Total Assets. This approach avoids currency conversion and is likely the best size metric within each country-year. The earned-to-contributed-equity ratio (RE/BE) is the ratio of Retained Earnings to Book Equity. I winsorize all firm-level ratios at 5 % in each tail. Finally, I include two additional controls to be sure that the choice between dividends and repurchases is not due to either capital market development or taxation of dividends relative to capital gains. I use the ratio of stock market capitalization to gross domestic product (GDP) in each country to proxy capital market development (Djankov et al., 2008). There is ample evidence that the relative taxation of dividends and capital gains affects the blend of dividends and repurchases, both in the U.S. (Chetty and Saez, 2005) and around the world (Jacob and Jacob, 2013; Alzahrani and Lasfer, 2012). I follow La Porta et al. (2000) and use the dividend tax preference as the ratio of the after-tax value dividends relative to the 12

14 after-tax value of capital gains as a percent (i.e., Dividend Tax Preference = τ D 1 τ CG ). Gov41 limits my analysis to the largest firms in each market, so there are only modest additional data requirements. Over 2004 to 2008 there are 197,896 firm-years in Worldscope without regard to country. Limiting my main analysis to countries with Gov41 data further limits the analysis to 140,669 firm-years. Dropping firm without Gov41 is the most significant data limitation and reduces the sample to 26,452 firm-years. Requiring the main set of variables (D/Y, R/Y, V/A, E/A, da/a, Size, and RE/BE) reduces the sample 25,383 firmyears. I also drop the U.S. and countries without at least five observations every year from 2004 to This provides the main sample of 7566 firm-years. Table 1 provides the number of observations by country and year. Table 2 provides descriptive statistics for key payout and predictor variables by legal origin, country, and firm-level governance quantile, where quantiles are above and below median Gov41 each country-year. Values in Table 2 are the time-series means of countryyear-gov41-quantile medians. Panel A reports these values for common law countries and Panel B reports these values for civil law countries. 11 The median dividend payout ratio is positive in all country-gov41 quantiles. The median repurchase payout ratio is (near) zero for about one half of the country-gov41 quantiles, but untabulated results so that there are repurchasers in every country. There aren t clear relations between firm fundamentals and Gov41 by country, although high Gov41 firms often have higher profitability, consistent with Mitton (2004). 3.3 Comparing Country-level and Firm-level Governance As a first step to understanding the relation between dividend payout ratios and firm- and country-level governance, Table 3 provides dividend payout ratio models with the dividend predictors (Denis and Osobov, 2008; von Eije and Megginson, 2008) and combinations of 10 Gov41 is available for many U.S. firms, but dropping U.S. firms avoids the horse and rabbit stew problem. In untabulated results the U.S. results are similar to other common law country results. 11 I follow La Porta et al. (2000) and don t differentiate between different civil law origins. Civil law is the remainder of countries without English legal origins. 13

15 firm- and country-level governance measures. I omit profitability (E/A) from payout ratio regressions throughout because earnings-based measures in the dependent variable denominator and an independent variable numerator mechanically create a negative coefficient. The country-level governance measure is legal origin (Djankov et al., 2008) and the firm-level governance measure is Gov41 (Aggarwal et al., 2011). Consistent with von Eije and Megginson (2008) all models include firm REs and year FEs, and cluster standard errors by firm. In untabulated results my results are qualitatively similar if I replace firm REs with industry FEs and cluster standard errors by industry, where industry is based on two-digit SIC codes. I drop observations with non-positive net income to avoid negative dividend payout ratios and winsorize firm-level ratios at 5% in each tail. Column 1 includes the fundamental dividend predictors from Denis and Osobov (2008), plus the ratio of stock market capitalization to GDP and the dividend tax preference. Dividend payout ratios decrease in growth opportunities (da/a) and increase in both capital market development and dividend tax preference. There is no relation between dividend payout ratios and size, but size is the percentile ranking by total assets within each country-year and has little variation in this sample (the interquartile range of size in this table is 14 %). Counter to the literature, there is a weak negative relation between dividend payout ratios and retained earnings (RE/BE) of % = 2.5 %. Column 2 adds a common law indicator variable to control for country-level governance. This reduces the economic and statistical significance of the capital market development variable, but all other coefficients are qualitatively the same as in Column 1. This result is consistent with La Porta et al. (2000) who show that dividend payout ratios increase in country-level governance and decrease in growth opportunities. I proxy growth opportunities with total asset growth rates (da/a) to avoid reverse causality interpretations between market-to-book-assets ratios (V/A) and dividend payout ratios. My results are qualitatively similar if I proxy growth opportunities with either market-to-book-assets ratios or sales growth rates (ds/s). Column 3 replaces the common law indicator variable with Gov41 and reports similar results. The Gov41 coeffi- 14

16 cient is 50 times smaller than the common law coefficient, but the economic significance is nominally the same. This is because Gov41 falls between 0 and 100 and has a mean and standard deviation that are that are 120 times and 20 times larger than the common law indicator mean and standard deviation, respectively. Without the country-level control the economic and statistical significance on the capital markets development variable return. The inferences from the remaining predictors remain the same. When the dividend payout ratio regressions include both firm- and country-level governance measures, I find that country-level governance dominates firm-level governance. Column 4 include both firm- and country-level governance measures and the common law indicator coefficient is unchanged, but the Gov41 coefficient is smaller in magnitude and not significantly different from zero. This result suggests that both firm- and country-level governance matter, but that country-level dominates. Given the importance of firm-level governance in other studies, as well as its dependence on country-level governance, it could be that there is more nuance in the relation between payout policy and firm- and country-level governance. This is my focus going forward as I examine payout policy and the interaction of firm- and country-level governance. 3.4 Firm-level Governance Instrumental Variable An implicit assumption so far is that firm-level governance is randomly assigned. However, this assumption is unlikely to be correct. It is more likely that firm-level governance is chosen each year alongside payout policy. This simultaneous choice makes Gov41 possibly endogenous, but an IV for Gov41 can address this concern. The Gov41 IV must be correlated with payout policy (relevance), but only through Gov41 (exclusion restriction). Aggarwal et al. (2011) show that Gov41 varies within countries in both the cross section and the time series, which makes the average Gov41 for all other firms in the same country-year a possible IV. I refer to the average of Gov41 for all other firms in the same country-year as the leave-out average of Gov41. For the leave-out average to be a 15

17 valid instrument it must induce variation in firm-level governance Gov41 and affect payout decisions only through this relation with Gov41. Aggarwal et al. (2010) use a similar IV with valuation as the dependent variable. Aggarwal et al. (2011) show that Gov41 improves in waves within each country over the sample period, so it could be that some portion of a given firm s Gov41 improvements are due to country-level trends to improve firm-level governance. This same IV motivation appears elsewhere in the literature. In the setting of U.S. governance, John and Kadyrzhanova (2008) and John and Litov (2010) show that good governance leads to good governance in peer firms. In a broader setting, Leary and Roberts (2014) show that peer firms determine capital structure and financial policies, and that these peer effects are more important than many other fundamental determinants. John and Kadyrzhanova (2008) define peer firms as within the same state and Leary and Roberts (2014) define peer firms as within same three-digit SIC code. John and Litov (2010) instrument the Gompers et al. (2003) firm-level governance measure with the leave-out average at the industry-year level. Given my international setting the appropriate peers are within country and year. 12 Throughout, the leave-out average of Gov41 passes weak instrument tests with first-stage F-tests greater than 10, as well as the more demanding Stock and Yogo (2002) minimal F-test scores well beyond conventional significance levels. The exclusion restriction that the leave-out average of Gov41 does not affect payout ratios, except through its impact on firm-level governance must be maintained. 4 Results La Porta et al. (2000) relate dividend payout ratios and country-level governance, but the dividend decision has three aspects. The firm must first decide to initiate payout, second 12 My main results are qualitatively similar if I use the leave-out average at the country-industry-year level, where industry is either Fama and French (1997) 12-industry or one-digit SIC code. However, my robustness tests use either U.S.- or U.K.-based directors as an alternative firm-level governance measure. These indicator variables have more discrete cross-sectional variation and require country-year leave-out averages to identify regressions. For simplicity, I use the same IV approach throughout and use leave-out averages at the country-year level. Both country-year and country-industry-year leave-out averages exceed all weak IV thresholds. 16

18 the level of payout, and third the division of payout between cash dividends and share repurchases. Section 4.1 assesses the relation between firm-level governance and payout level and Section 4.2 assesses the relation between firm-level governance and payout initiation decisions. 4.1 Payout Level Table 4 presents my main analysis of the relation between payout ratios and firm-level governance conditional on country-level governance, where firm-level governance is Gov41 instrumented with the country-year leave-out average of Gov41. I present only panel linear models because they allow a conventional IV approach, but my results are qualitatively similar with panel Tobit models. 13 Columns 1 and 3 mirror Column 3 in Table 3 and control for country-level governance by splitting the sample on legal origin. For weak legal regimes in Column 1, there is a positive relation between Gov41 and dividend payout ratios, consistent with the outcome model. However, for strong legal regimes in Column 3, there is a negative relation between Gov41 and dividend payout ratios. This is more consistent with the substitute model, but further analysis shows that this is consistent with a more nuanced outcome model that substitutes from rigid dividends to flexible share repurchases as governance improves. The economic magnitude of both changes is significant. For the civil law sample, the standard deviation of Gov41 is 6.7, so a one standard deviation rise in firmlevel governance results in a 5.7 % point rise in dividend payout ratios. This is sizable given the average civil law dividend payout ratio of 34 %. The magnitude is similar in common law countries where the standard deviation of Gov41 is 11, so a one standard deviation rise in firm-level governance leads to a 8.5 % point fall in dividend payout ratios. This is sizable given the average common law dividend payout ratio of 44 %. Columns 2 and 4 repeat this analysis with share repurchase payout ratios. There is a 13 My main results in Sections 4.1 and 4.2 use panel linear regressions, but panel Tobit and logit models are more appropriate given the left-censored and binary dependent variables, respectively. However, the Gov41 IV complicates these models. Section 5.2 addresses this concern with the control function approach. Untabulated results for these models are qualitatively similar. 17

19 positive relation between firm-level governance and payout via repurchases for both civil and common law countries, but neither is statistically significant at conventional levels. Considering both Columns 3 and 4 suggests there is a substitution from dividends to repurchases for common law countries. In strong legal regimes, improved firm-level governance allows firms and shareholders to substitute from dividends to the preferred payout method of repurchases. Payout ratios decreases in growth opportunities in both legal regimes, although consistent with the outcome model dividend payout ratios are more sensitive to growth opportunities in common law countries. Dividend payout ratios increase in dividend tax preferences in both civil law and common law countries. Consistent with Alzahrani and Lasfer (2012), the dividend tax preference coefficient is larger in strong legal regimes. All other predictors are consistent with Table 3. Columns 5 and 6 pool the sample and repeat the analyses in Columns 1 to 4 as interactions with common law indicator variables. The same results appear. Civil law firms pay out more dividends and repurchases as firm-level governance improves. Common law firms unconditionally pay out more via dividends and substitute from dividends to repurchases as firm-level governance improves. The results for the level of payout via dividends and repurchases support an outcome model with a continuous relation between firm- and country-level governance. In weak legal regimes, higher firm-level governance relates to higher dividend and repurchase payout. Higher dividend payout is expected to be persistent and is a necessary commitment mechanism in weak legal regimes. In strong legal regimes, higher firm-level governance facilitates a substitution from dividends to repurchases, which provides greater flexibility to both firms and shareholders. These results are robust to an IV approach to address the simultaneous choice of payout policy and firm-level governance. With a better understanding of firm-level governance and payout levels, Section 4.2 moves to payout initiation decisions. 18

20 4.2 Payout Initiation La Porta et al. (2000) frame the relation between country-level governance and payout in terms of dividend payout ratios, and Mitton (2004) adds firm-level governance. In the previous section I add share repurchases, but firm-level governance should also affect investor ability to extract payout at any level and in any form. Firm-level governance should affect payout initiation, continuation, and omission, not just level conditional on payment. In this section I address the dividend and repurchase initiation decision and its relation to firm- and country-level governance. I present only panel linear probability models because they allow a more conventional IV approach, but my results are qualitatively similar with panel logit models. 14 The dependent variables in these regressions are indicator variables for dividend payers and share repurchasers. The same possible endogeneity between payout policy and firm-level governance exists in these propensity-to-pay models, so in Table 5 I instrument Gov41 with the leave-out average of Gov41. These models include the same set of independent variables as before, as well as profitability because the dependent variable no longer conditions on net income. I also include all firm-years, regardless of net income. All models control for unobserved heterogeneity with firm REs and year FEs, and cluster standard errors by firm. Columns 1 and 2 in Table 5 show that in civil law countries, both cash dividends and share repurchases are less likely as firm-level governance improves. The coefficients are small, but the economic magnitudes are meaningful. For cash dividends, a coefficient of and a Gov41 standard deviation of 6.8 translate to a 2.2 % point decrease in propensity to pay a dividend on a sample mean of 91 %. For share repurchases, a coefficient of translates to a 6.3 % point decrease in propensity to repurchase on a sample mean of 62 %. This is consistent with the outcome model in which increases to shareholder rights facilitate a delay in dividend and repurchase initiation. However, Table 4 shows that once initiated, 14 Panel logit models with an IV use the control function approach and block bootstrap standard errors. See Footnote 13 and Section

21 both dividend and repurchase payout ratios increase in firm-level governance for firms in civil law countries. Columns 3 and 4 in Table 5 show that in common law countries cash dividends are less likely, but share repurchases are no less likely as firm-level governance improves. These results are again consistent with a more nuanced outcome model where firms and shareholders can delay dividends and substitute from dividends to repurchases by improving shareholder rights. Again, the coefficients are small, but the economic magnitudes are meaningful. For cash dividends, a coefficient of and a Gov41 standard deviation of 11 translate to a 7.0 % point increase in propensity to pay a dividend on a sample mean of 85 %. For share repurchases, the Gov41 coefficient is neither economically nor statistically significant. The coefficient on the standard payout predictors are consistent with the literature. Both cash dividends and share repurchases are more likely as profitability, size, and life-cycle increase. Both cash dividends and share repurchases are less likely as growth opportunities increase, although this relation is only statistically significant in Column Dividend initiation is more likely and repurchase initiation is less likely as the dividend tax preference increases, although this relation is not economically or statistically significant for civil law repurchasers in Column 2. In conclusion, firm-level governance affects payout decisions in both level and initiation, for both dividends and repurchases, but the relations vary across governance regimes. In weak legal regimes increases in firm-level governance relate to increases in levels of both dividends and repurchases. Further, because investors are confident that they will be able to extract payout in the future, they allow firms to delay dividend and repurchase initiation. In weak legal regimes, the relation between firm-level governance and initiation of dividends and repurchases is negative. However, in strong legal regimes increases in firm-level governance lead to a substitution from dividends to repurchases in both levels and initiation decisions. 15 In untabulated results the relation between payout policy and growth opportunities does not depend on firm-level governance. In regressions with interactions between growth opportunities and payout policy the interaction coefficients are not significantly different from zero. 20

22 If both country- and firm-level governance are high, then shareholders have mechanisms to control expropriation and repurchases provide firms with flexibility and shareholders with tax advantages. These results are robust to the standard payout predictors, an IV approach to address possible endogeneity, unobserved firm and year heterogeneity, and standard errors clustered by firm. 5 Robustness This section tests robustness to several different scenarios. These scenarios include alternative payout ratio definitions and alternative regression specifications that accommodate left-censored payout ratios and binary payout initiation. These robustness tests also include a more sophisticated measure of country-level governance and an alternative firm-level governance measure of either U.S.- or U.K.-based directors. This alternative firm-level governance measure provides an out-of-sample test that supports my core results in a much larger panel of data. 5.1 Alternative Payout Ratios Scaling cash dividends and share repurchases by net income is consistent with the literature, but not the only option. Alternative denominators include total assets and total sales. Untabulated results are qualitatively similar with total assets and total sales as payout ratio denominators. 5.2 Alternative Specifications Linear panel regressions are commonly used in the literature and easily interpreted, particularly with an IV approach. However, it is not possible for a firm to pay out less than zero percent of net income. This is less of a concern for dividend payout ratio models given that 89 % of sample firm-years pay dividends, but more of a concern for share repurchase payout 21