does political economy reduce agency costs? some evidence from dividend policies around the world

Journal of Empirical Finance 18 (2011) 16–35

Contents lists available at ScienceDirect

Journal of Empirical Finance

j ourna l homepage: www.e lsev ie r.com/ locate / jempf in

Does political economy reduce agency costs? Some evidence from dividendpolicies around the world

HiuLam Choy a, Ferdinand A. Gul b, Jun Yao c,⁎a Lebow College of Business, Drexel University, United Statesb Accounting and Finance, School of Business, Monash University, Sunway Campus, Malaysiac School of Accounting and Finance, The Hong Kong Polytechnic University, Hong Kong

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +852 3400 3454; faxE-mail addresses: [email protected] (H. Ch

0927-5398/$ – see front matter © 2010 Elsevier B.V.doi:10.1016/j.jempfin.2010.10.001

a b s t r a c t

Article history:Received 12 December 2008Received in revised form 28 March 2010Accepted 1 October 2010Available online 8 October 2010

This study shows that firms in proportional-electoral countries pay out lower dividends andthat the correlation between a firm's growth potential and dividend payout ratio is weaker inproportional-electoral countries. However, firms in proportional-electoral countries that cross-list in majoritarian system countries, tend to pay out higher dividends and the negative relationbetween growth potential and dividend payout tend to be stronger than their peers that do notcross-list. For a few countries that changed their electoral system towards a more proportionalsystem, we observe a decrease in dividend payout ratio and a weaker relation between growthand dividends after the change. Overall these results indicate that a country's political systemaffects the severity of agency problems. Further, the effect of legal origin on dividend policyreverses once we include the political economy variables in the regressions. We also documentthat the electoral system not only affects the amount of dividends paid by a firm but also theform of payment.

© 2010 Elsevier B.V. All rights reserved.

JEL classification:G35G38

Keywords:Political economyDividendInvestor protectionLegal protection

1. Introduction

An important strand of the global markets finance literature provides some insights into the linkage between the degree ofinvestor protection in a country, which depends on its legal origin and enforcement, and the development of financial market, thefinancial and ownership structure of firms, and the severity of agency problems in that country (Modigliani and Perotti (1997);LaPorta et al. (1998, 2000a, b): Levine and Zervos (1998)). However, recent research suggests that political economy variablesmayindeed be a more reliable indicator of investor protection than a country's legal origin. For example Pagano and Volpin (2005),using a sample of 21 OECD countries, show that countries with a proportional electoral system have significantly weakershareholder protection than those with a majoritarian electoral system.

Unlike legal origin, which is static in nature, the type of electoral system a country adopts can change over time. This dynamicnature of the electoral system better matches the change in law and financial regulations over time. By using simply the legalorigin as a proxy for shareholder protection, prior studies do not allow for such a change in shareholder protection over time.Further, by using the historical measure of legal origin, prior studies basically assume that even with all the legal reforms orimprovements civil law countries are willing to undertake, they will always lag behind common law countries in terms of investorprotection and financial market development (Pagano and Volpin (2005)). On the contrary, as the political system also varies overtime, it accommodates the change in shareholder protection over time. Apart from Pagano and Volpin (2005), other studies also

: +852 2330 9845.oy), [email protected] (F.A. Gul), [email protected] (J. Yao).

All rights reserved.

http://dx.doi.org/10.1016/j.jempfin.2010.10.001

mailto:[email protected]



http://dx.doi.org/10.1016/j.jempfin.2010.10.001

http://www.sciencedirect.com/science/journal/09275398

17H. Choy et al. / Journal of Empirical Finance 18 (2011) 16–35

emphasize the role of political economy in explaining differences in financial development and corporate governance (Beck et al.(2001); LaPorta et al. (2002); Rajan and Zingales (2003); Bushman et al. (2004)).

In this paper, we extend this interesting and evolving line of research on the role of political economy in investor protection andfinance by examininghowdividendpolicies can beused by afirm to commit itself to reduce the agencyproblemof free cashflows andhow these policies varywith thepolitical economyacross countries. In addition,we examinehowa change in thepolitical economy, interms of the country's electoral system of a country, can affect the dividend policy of firms. Pagano and Volpin (2005) document thatthe anti-directors' rights index of a country varieswith its electoral system.We expect that the effect of electoral system is not limitedto the legal regulation (e.g. anti-directors' rights) but also extends to its enforcement. A country's law and its enforcement togethershape the agency costs investors face. A country's electoral system determines how a political party and politicians can win in anelection and hence their campaign style and strategies. These campaign strategies, in turn, affect the degree of representation outsideinvestors and minority shareholders have in the government and hence their welfare. Therefore, we anticipate a country's electoralsystem affects its agency costs. In this paper, we focus on one aspect of agency costs: the severity of the free cash flow problem. Morespecifically, we examine whether minority shareholders in countries with proportional electoral system suffer from higher agencycosts of free cash flow. With poorer shareholder protection, proportional-electoral countries are likely to be associated with moresevere agency problems as minority shareholders are less capable of exercising their rights to mitigate such problems.

Our results show that firms incorporated in countries with proportional electoral system pay out a smaller portion of theirearnings as dividends. These results are consistent with the findings of Pagano and Volpin (2005) that countries with proportionalelectoral system have weaker investor protection law. Further, our results suggest that both the legal rule and its enforcement areless favorable to investors in countries with proportional electoral system. The poorer investor protection offered in proportional-electoral countries makes it difficult for minority shareholders in these countries to extract dividends from firms.

Consistent with the theory and findings in prior studies (Jensen (1986); Smith andWatts (1992) and Gaver and Gaver (1993)),we find that low-growth firms (i.e. firms that are likely to have more severe free cash flow problems and higher agency costs)make higher dividend payouts than high-growth firms. More importantly, we find that after controlling for legal origins and anti-director rights, such a relationship between a firm's growth potential and dividend policy is much weaker in a country with theproportional electoral system than in one with a majoritarian system. However, once we control for the electoral system and anti-director rights of a country, the effect of legal origin on a firm's dividend policy reverses. These results suggest that the effect oflegal origin on dividend payout policy documented in LaPorta et al. (2000a) can, in fact, be attributed to the electoral system.

In addition, we examine a group of firms that cross-list their stocks in an overseas exchange where the proportionality score ofthe cross-listing countries differ from their domicile countries.1 These cross-listing firms are subject to the laws in the cross-listingcountries in addition to their home countries' regulations. Our results suggest that firms in proportional countries that cross-list inother exchanges tend to paymore dividends than their non-cross-listing peers. Also, the negative correlation between growth anddividend payout tends to be stronger for these cross-listing firms.2

As a country's electoral system can change over time, we also evaluate the relationship between changes in the electoral systemand agency problem: we investigate how a change in a country's electoral system affects a firm's dividend payout policy. Using asmall sample of countries that changed their electoral systems during our sample period, we find that for firms in countries thatswitch from a pure majoritarian (proportional) system to a mixed one their dividend payouts decrease (increase) and the relationbetween growth and dividend payout is weaker (stronger) for these firms after the change.3

We perform several robustness checks for our results. First, we repeat our analyses for the common and civil law countriesseparately.4 Not surprisingly, the results continue to hold in both sub-samples. This suggests that within both common and civil lawcountries, the type of a country's electoral system affects the dividend policies. Second, we examine the impact of electoral system onthe formof dividendpayment. For this analysis,we expandour classification of dividend-payingfirms to thosepaying stockdividends.Our results suggest that for firms in proportional-electoral countries, even if they choose to pay dividends, they aremore likely to payout stock than cashdividends. Third, our results are robustwhenweexcludeU.S.firms (the countrymostdominant in our sample), useweighted least square for our analyses, and run regression analyses for each year separately. Accordingly, thesefindings confirm that acountry's electoral systemdoes have a significant impact on firms' dividend policies and consequently the agency costs investors face.

We also find that in addition to the electoral system the extent of direct government involvement in the economy and financialsystem, the cost of entry, and the risk of expropriation (political economy proxies used in Bushman et al. (2004)) all have asignificant impact on the severity of agency problems. More specifically, we find that the state ownership of banks and firms, thecost of entry, and the risk of expropriation have a negative impact on the dividend payout ratio. Also, low-growth firms payrelatively low dividends in a country with an extensive state ownership of firms, high cost of entry, and high risk of expropriation.This evidence is consistent with the argument that governments can use their ownership and control over banks and firms to favorconnected parties and expropriate assets from minority shareholders. These findings strongly suggest that the political system5

has a significant impact on the agency costs in a country and the country-level dividend policies.

1 In our sample, we do not observe any firm in a pure majoritarian country (i.e. Prop=0) cross-list in a pure proportional country or any country with aproportionality score greater than zero. Moreover, the cross-listing tends to be unilateral: firms in countries with high proportionality scores cross-list theirstocks in countries with lower scores but not the other way round.

2 We thank an anonymous referee for suggesting this test to us.3 We acknowledge that the sample size for the change analysis is small. However, we are constrained by the number of countries that have changed their

electoral system during our sample period. The cases we documented in the paper are all countries that have made such changes during our sample period.4 The proportionality scores of civil law countries vary from 1 to 3 whereas the proportionality scores of common law countries vary among 0, 1 and 3.5 We find similar results using a factor analysis of all the five political economy variables.

18 H. Choy et al. / Journal of Empirical Finance 18 (2011) 16–35

This paper contributes to the literature in the following ways. First, we document for the first time that political economy doesaffect a firm's dividend policy. More specifically, we show that countries with a proportional system have lower dividend payouts.This result suggests that a country's political economy is a significant determinant of the extentminority shareholders can exercisetheir rights to curtail the agency problems in that country. Agency costs, in particular those attributed to free cash flow problems,are higher in countries with a proportional electoral system than those with a majoritarian electoral system. Second, our findingthat after controlling for a country's legal origin and anti-director rights index (a proxy for investor protection in LaPorta et al.(2000a)), a country's electoral system still has a significant impact on dividend policies suggests that the electoral system itselfcaptures certain aspects of agency problems not reflected in the legal regime and the anti-director rights index. The result that theelectoral system continues to have a significant effect on the relationship between growth and dividend payouts within thecommon and civil law country sub-samples separately further confirms this. After we controlled for the electoral system andinvestor protection of a country, the legal origin no longer has any significant effect on firms' dividend payout. These resultssuggest that the dynamic electoral system dominates the effect of the static legal origin. Since the electoral system of a countrychangeswith time, whereas the legal origin is fixed and determined by historical factors, a country's electoral system ismore likelyto reflect the current investor protection condition of a country. Results of our test with a small sample of countries that changedtheir electoral system again confirm that the change in electoral system leads to a change in dividend payout policies consistentwith our expectation. Thus, future cross-country agency-cost studies should consider not only the effect of legal origin but also theeffect of the electoral system prevailing in the country. The type of electoral system complements the anti-director rights index incapturing the extent of agency problems in a country. The finding of the more dominant role of political economy in explainingcorporate finance policies improves our understanding of the factors that drive dividend policies in different countries. Thedocumentation of how a change in the electoral system affects the dividend policy also adds to the corporate finance literature.Third, our result that firms in proportional countries that cross-list their stocks in majoritarian countries tend to pay out higherdividends and exhibit a stronger negative correlation between potential growth and dividend payout ratio than their peers that donot cross-list, suggests that cross-listing in countries with better investor protection helps to lower agency costs and enhance thewelfare of minority shareholders. Fourth, our finding that low-growth firms tend to have higher dividend payout ratios confirmsthat dividend policy helps to reduce the agency cost of free cash flow not just in U.S. but also around the globe. Fifth, by examiningthe extent to which firms employ dividend policies to curtail agency costs around the globe, this paper provides an indirectmeasure of the agency problem/costs across countries. Finally, our paper provides avenues for future research on the role ofpolitical economy in other aspects of corporate finance such as firm's capital structure across different countries.

The paper proceeds as follows. Section 2 provides a literature review of some studies that examine dividend policies anddiscusses howwe develop our hypotheses. Section 3 describes our sample and Section 4 presents the empirical findings. Section 5shows the results of our sensitivity analyses. Section 6 concludes.

2. Literature review and hypotheses development

2.1. Free cash flows and dividend policy

Jensen (1986) suggests that for firms with free cash flow problem, managers can minimize such agency costs by increasingdividend payout to shareholders. Since reducing dividend payments can trigger a drop in stock price, by increasing dividends,managers commit themselves to pay out the higher level of dividends to shareholders in both the current and future periods. Thatis, managers bond their promises to disgorge free cash flow in future periods. This reduces the inefficiency ofmarginal investmentsand hence the agency costs of free cash flow. Firms with more growth opportunities have lower free cash flows and hence paylower dividends. Consistent with this theory, Smith and Watts (1992) and Gaver and Gaver (1993) find that high-growth firmshave lower dividend payout ratios than non- or low-growth firms.

Closely tied to this free cash flow theory is the life-cycle explanation for dividend payments proposed by DeAngelo et al.(2006). They find thatwhile young rapidly growingfirms havemostly “contributed” equity capital and pay few or no dividends,moremature, highly profitable firms with equity capitalization made up of retained earnings pay the bulk of dividends each year.

2.2. Dividend policies in non-US firms

While there has been much research on dividend policies in the United States (Eije and Megginson (2006); DeAngelo et al.(2004); Weston and Siu (2003); Fama and French (2001)), relatively little is known about dividend policies outside the U.S. Apartfrom three papers that examine how British tax policies affect dividend policies of U.K. firms (Ang et al. (1991); Lasfer (1997); Belland Jenkinson (2002)), there are other country-specific studies that have focused on the tax effects in Australian (Cannavan et al.(2004)), Taiwanese (Lee et al. (2004)) and Canadian (Christoffersen et al. (2005)) firms. Studies on the cross-sectional patterns ofglobal dividend payouts are even scarcer. To our knowledge, only several studies present some evidence of cross-sectional globaldividend payouts. LaPorta et al. (2000a) and Dennis and Osobov (2006) examine the dividend policies across countries. Theagency model in both studies hinges on Jensen's free cash flow theory (Jensen (1986)).

First, LaPorta et al. (2000a) examine how the legal origin and the anti-director rights (a proxy for minority shareholderprotection) of a country in which a firm is incorporated affect its dividend policy. They test whether dividends are an outcome ofan effective system of legal protection of shareholders (“the outcome model”) or dividends are a substitute for legal protection(“the substitute model”). LaPorta et al. (2000a) find that firms operating in countries with better investor protection pay higher


dividends and high-growth firms in these countries pay lower dividends than low-growth firms. They claim that these results areconsistent with the idea that legally protected shareholders are willing to defer their receipt of dividends when firms have goodinvestment opportunities. On the other hand, poorly protected shareholders are not willing to wait and take whatever dividendthey can get even when the firm has great investment opportunities. They argue that this is “part of the agency cost of poorinvestor protection” (LaPorta et al. (2000a)). Accordingly, their results support “the outcome model”. Second, Dennis and Osobov(2006) show that companies headquartered in United States, United Kingdom, Canada, France, Germany and Japan demonstrate adeclining propensity to pay dividends. They find that the propensity to pay is positively related to growth opportunities in civil lawcountries and is negatively related to growth opportunities in common law countries. On this basis they conclude that their resultssupport the agency model over the catering model.6 Third, Pinkowitz et al. (2006) find that the relation between dividends andfirm value is weaker in countries with strong investor protection.

2.3. Political economy and agency problems

Pagano and Volpin (2005) show that countries with proportional electoral systems have significantly weaker shareholderprotection than those withmajoritarian systems. Differences in themechanism of these two electoral systemsmotivate politiciansto cater for the needs of different sectors of the society. The two systems differ in the electoral formula, district magnitude, and theballot structure (Persson and Tabellini (2006); Rae (1969)). Electoral formula determines how votes are translated into seats.District magnitude refers to the number of seats available in a district/constituency and hence the number of districts in a vote.Ballot structure specifies how citizens cast their ballots (Persson and Tabellini (2006)). On the aspect of electoral formula, amajoritarian electoral system adopts the plurality rule while a proportional electoral system uses the proportional rule. That is, in amajoritarian electoral system, the winner of the highest number of votes is elected in a given district; whereas in a proportionalelectoral system, seats are awarded in proportion to the votes won in each district. As for district magnitude, the majoritariansystem tends to havemultiple districts eachwith single seat while proportional system has a single district withmultiple seats. Forexample, in the United States and United Kingdom, where the single member plurality system is adopted, each district elects justone legislator. In Spain, where the proportional system is used, each district elects on average seven legislators. Under themajoritarian electoral system, voters choose among individual candidates; while under the proportional system, voters elect theparties and then the parties decide which members within the party take the seats (Persson and Tabellini (2006)). In summary,under the majoritarian system, voters choose among individuals for a single seat in a district and the candidate with the highestvote wins the seat; under proportional electoral system, voters choose among political parties for the multiple seats available in adistrict and the allocation of seats is determined by the proportion of votes a party wins.

Given the difference in the mechanism of the two systems, a party's strategy to win in an election and be the controlling partyalso differs. In order to win in a majoritarian system, a party has to win a majority of districtswhereas in a proportional system, theparty has to win a majority of votes. Consequently, political parties tend to cater to the needs of more homogenous social groups(e.g. labor unions) and ignore those groups with vastly different preferences (e.g. unorganized minority shareholders) in aproportional electoral system. By shifting their platform towards the homogenous preference of a social group, the politician canattract the additional mass of voters from that social group. Minority shareholders are more likely to have their voice heard underthe majoritarian system. This is because under a majoritarian system, the strategy is to win the pivotal district. As such, thehomogeneity of a social group's interest is not as critical a determinant of the candidates' political platform as in the proportionalsystem. Accordingly, minority shareholder protection is likely to be stronger under a majoritarian electoral system.

Another distinction between the two electoral systems—candidate-based versus party-based voting—also affects the campaignstyle. In the candidate-based majoritarian system, parties tend to campaign in decentralized fashion and the emphasis is more onindividual candidates. With the individual candidates competing for personal votes, they are more inclined to have regular contactwith individual voters (Farrell (2001), p171). They are alsomore likely to voice the concern of these individual voters and incorporatethese concerns in their policies. Also, the action of these individual representatives is more observable to the voters. Hence, voters canbase their votes on the past performance of these individual representatives. On the other hand, under the party-based proportionalsystem, the voter is choosing between parties and not candidates. In such system, the principal “voting constituency” of the individualpolitician is not the voters, but the “selectorates” who determine which candidates to put on the party list and their orders (Farrell(2001), p. 171). In other words, the selectorates, not the voters, have a significant influence over which individual representative ofthe party can win a seat in the election. Thus, candidates elected are more likely to serve the interest of the party and the legislationthey supported is more likely to be based on the party's preference. Under such system, the candidates elected are “prone to havecloser links with organized interest” (Farrell (2001), p. 171). That is, candidates are more likely to cater to the interests of organizedgroups, such as workers' unions, chamber of commerce, and various business groups than those of individual investors.Consequently, individual voters are better served by representatives elected under the majoritarian electoral system where voterschoose candidates and can hold them individually accountable for their policy (Bowler and Farrell (1993)).

Pagano and Volpin (2005) develop a theoretical model that draws the same conclusion as stated above. Using a sample of 21OECD countries, Pagano and Volpin's empirical analyses support the conclusion of theirmodel. Specifically, Pagano and Volpin findthat employment protection increases and shareholder protection regulation decreases (proxied by the anti-director rights indexas in LaPorta et al. (2000a)) with the degree of proportionality in a country's electoral system.

6 Baker and Wurgler (2004) argue that companies supply dividends to meet (or cater for) investor demand. See also Li and Lie (2006).


2.4. Hypotheses development

With the better investor protection law offered by countries with majoritarian electoral system, investors can exercise theirrights and more efficiently utilize the dividend payout as a monitoring tool to minimize agency costs. Accordingly, firms inmajoritarian-electoral countries should have lower agency costs than those in proportional-electoral countries. In addition, inrespect of dividend policy, the incentives of workers are likely more aligned with those of the managers or controllingshareholders. Managers or controlling shareholders are more inclined to retain the cash inside the firm. This excess cash facilitatesempire-building. Workers benefit from this empire-building because it creates employment opportunities. Furthermore, even thiscash hogging reduces the profitability of a firm, it helps to lower the firm's bankruptcy risk. Both workers and creditors can benefitfrom this less risky environment. Thus, workers, managers, controlling shareholders, and creditors tend to favor retaining the freecash flow than distributing it as dividends. As such, these parties likely form political allies againstminority shareholders regardinga firm's dividend policy. These parties also tend to be in more well-organized groups than minority shareholders. These allies aremore likely to dominate in a proportional-electoral country than a majoritarian-electoral country because representatives electedunder the party-based proportional system are more prone to serve the interest of these organized groups.

The free cash flow theory has been widely tested in the United States where both the regulation and the legal enforcement ofinvestor protection are relatively strong. We investigate whether the theory continues to hold in other countries, especially thosewith weaker investor protection. With different political and legal systems, the extent of agency problems is likely to vary acrosscountries. This study examines whether the political system, more specifically, the electoral system of a country has any impact onthe dividend policy of firms incorporated there. We expect that with the poorer investor protection and the “political alliance” ofmanagers, controlling shareholders and workers in proportional-electoral countries, firms in these countries tend to pay lowerdividend than those in majoritarian countries. Hence, our first hypothesis is as follows:

H1. Firms in countries with proportional electoral system have lower dividend payout ratios than firms in countries withmajoritarian electoral system, ceteris paribus.

We also examine whether high-growth firms pay lower dividends than low-growth firms across countries with differentelectoral systems and whether there is any variation in the extent of the negative correlation between growth and dividend policyacross countries. As suggested by Jensen (1986), firms with more profitable investment opportunities have lower free cash flowand pay lower dividends. Rozeff (1982) and Easterbrook (1984) document that the new issue market provides effectivemonitoring and lowers agency costs. Smith and Watts (1992) argue that with their ample profitable investment opportunities,high-growth firms have to go to the issue market more frequently than the low-growth firms. Thus, high-growth firms are subjectto the monitoring of the new issue market on a frequent basis and suffer less from the agency costs of free cash flow. Accordingly,these firms tend to pay lower dividends than their low-growth peers.7 If investor protection (both in terms of law andenforcement) is better inmajoritarian-electoral countries, minority shareholders will more likely be able to extract dividends fromlow-growth firms and the negative correlation between growth and dividend payout should be stronger. Alternatively, following asimilar argument proposed by LaPorta et al. (2000a), investors in majoritarian-electoral countries are likely to be more willing topostpone the receipt of dividends from firms with significant investment opportunities than investors in countries withproportional electoral system. With better investor protection in these countries, investors are willing to trade current dividendsfor a higher reinvestment return because they know they can extract the higher return as dividends in the future. On the otherhand, investors in proportional-electoral countries do not have such protection and try to get whatever they can out from thefirms. Thus, the negative correlation between a firm's potential growth and its dividend policy should be stronger for firms incountries with majoritarian electoral system. Accordingly, our second hypothesis is as follows:

H2. The negative correlation between a firm's potential growth and its dividend payout ratio is weaker in countries withproportional electoral system than in countries with majoritarian electoral system, ceteris paribus.

In addition to the electoral system of a country, we also examine whether the extent of direct government involvement in theeconomy have any effect on the agency cost of a firm. Following Bushman et al. (2004), we measure the degree of governmentinvolvement by the extent of state ownership of firms and banks, the cost of entry for a new firm and the risk of expropriation of assetsbygovernment. Ahighextentof government involvement in theeconomyprovides the opportunities for politicians andtheir cronies toexpropriate the resources of the economy. Government involvement in the economy likely leads to a relationship-based system asopposed to an arm's length system. For example, government ownership of enterprise leads to politically-favored corporations, whichreceives subsidies andother contract or fundingpriorities fromthegovernment (Gul (2006)). This favoritism is further extendedby thegovernment's ownership of banks,which facilitates theprovisionof externalfinancing for these corporations. To further “protect” thesecorporations, thegovernment can increase thehurdles fornewfirms to enter themarket andcompetewith thesefirms. In return for thefavoritism, these corporations likely adopt strategies that support the government policies (such as an increase in employment) eventhough these strategies are not for the best interest of shareholders. All these factors lead to weaker protection for the minorityshareholders. As such, we expect investors in countries with high government involvement more likely suffer from agency problems.

7 An alternative explanation is that with their ample investment opportunities, high-growth firms do not have as much free cash flows for dividenddistribution as low-growth firms.


Similarly, the risk of expropriationmeasure the extent of investor protection a country offers its investors. The inclusion of these otherpolitical economymeasures inour analysesprovides additional insights on the important role of political economyondividendpolicies.

We further include the legal regime measures as proposed by LaPorta et al. (2000a) in our analysis to control for any effectthese variables have on dividend policies.

3. Method

3.1. Sample selection

Our sample is based on all firm-year observations with the required financial data in the period of 1994–2002 in Global Vantage.8

Firms without consolidated financial statements are excluded. We further require the observations to have the electoral system datacreated following Pagano and Volpin (2005). Pagano and Volpin collected the electoral system data on 45 countries comprising bothOECD and developing countries.We thenmatch the samplewith the other political economy (i.e. state ownership of firms and banks,cost of entry, and risk of expropriation) data. The government ownership data are retrieved fromBushman et al. (2004). Data on stateownership of banks are retrieved from LaPorta et al. (2002). The cost of entry data are constructed as in Bushman et al. (2004) usingthe information provided by LaPorta et al. (2002). The risk of expropriation of assets by the government data are obtained fromLaPorta et al. (1998). Appendix A provides a detailed description of our variables and the sources of data.

Table 1 provides the distribution of our sample by the proportionality score. Our sample is dominated by firms in U.S., whichconstitute about 34% of the sample. Japan (19.86%) and United Kingdom (10.81%) also have significant representation in thesample. Our sample has limited observations on firms in Pakistan (0.03%), India (0.03%), South Korea (0.04%), and Turkey (0.07%).The final sample includes 35,356 observations covering 34 countries.

3.2. Variables and descriptive statistics

Table 2 Panel A describes the various measures of the political system, legal origin and legal protection of countries in oursample. Panel B presents the mean of contextual variables used in this study, grouped by the proportionality score.

3.2.1. Country-specific variables

3.2.1.1. Electoral system. We use the proportionality construct (Prop) of a country's voting system proposed by Pagano and Volpin(2005) in our analyses. This proportionality construct comprised of three components: (i) an indicator variable of whether at leastsome candidates are elected via a proportional rule (Proportional Representation); (ii) an indicator variable of whether somelegislators are elected via a majoritarian rule (Plurality); and (iii) whether most seats are allocated via a majoritarian rule(Housesys). Data on these indicator variables are obtained from the World Bank Database of Political Institutions. The degree ofproportionality is then computed as (Proportional Representation−Plurality−Housesys+2). The value of proportionality thusmeasured varies from 0 to 3. It takes on a value of 0 if no seats are assigned proportionally, a value of 1 if a minority of seats isassigned proportionally, a value of 2 if the majority of seats are assigned this way, and a value of 3 if all seats are assigned via aproportional rule (Pagano and Volpin (2005)). Canada, Malaysia, Pakistan, Singapore, Thailand, United Kingdom, and UnitedStates all have a proportionality score of 0, indicating that these countries adopt mostly the majoritarian electoral system. On theother hand, countries like Argentina and Austria have proportionality scores of 3, suggesting that the proportional electoral systemis the prevailing system in those countries (Tables 1 and 2 Panel A). In total, there are seven countries with a proportionality scoreof 0 (i.e. countries with pure majoritarian electoral system) and fourteen countries with a score of 3 (i.e. countries with pureproportional electoral system) in our sample. However, the pure majoritarian group makes up more than 55% of the sample aseach country has a larger number of observations; whereas the pure proportional group constitutes only 8.98% of the wholesample (Table 1). The average proportionality score of our sample is 0.692 (untabulated).

3.2.1.2. Other political variables. The other political variables we examine are state ownership of enterprise (SOE), state ownershipof bank (StBank), cost of entry (COE), and the risk of expropriation (Ex_risk). We follow the procedures in Bushman et al. (2004) incomputing these proxies. State ownership of enterprise is computed as the share of country-level output supplied by state-ownedenterprises. Countries with more state-owned enterprise investment receive higher ratings. The ratings range from 0 to 10.Column 5 of Table 2 Panel A shows that the maximum rating in our sample is 8 (Israel, Pakistan, and Taiwan) and the minimum is0 (Ireland, Netherlands, and New Zealand), with an average of 2.542 (Table 2 Panel B Last column).

State ownership of banks is measured as the share of assets of the top 10 banks in a given country owned by the government ofthat country. The state ownership of banks ranges from a minimum of 0% (Canada, Japan, New Zealand, South Africa, UnitedKingdom, and United States) to a maximum of 85.96% (Pakistan) in our sample (Column 6 of Table 2 Panel A). The average state

8 A significant portion of data required for our analysis is missing from Global Vantage prior to 1988; therefore we start our analysis from 1988. To compute thedecile ranking of growth of each firm, we use the average sales growth in the prior five years. In other words, we need the sales data in the prior six years tocompute the average sales growth, Therefore, 1994 is the first year with all the data items required for our analysis and marks the beginning of our sampleperiod.

able 1istribution of sample by countries. This table presents the number of firm-year observations used in our analyses by country and the degree of proportionalityProp) as constructed by Pagano and Volpin (2005).

Prop=0 Prop=1 Prop=2 Prop=3

Country Obs Percentage Country Obs Percentage Country Obs Percentage Country Obs Percentage

Canada 1648 4.66% Australia 808 2.29% Germany 1334 3.77% Argentina 36 0.10%Malaysia 962 2.72% France 1670 4.72% Spain 420 1.19% Austria 198 0.56%Pakistan 11 0.03% India 11 0.03% Switzerland 618 1.75% Belgium 288 0.81%Singapore 691 1.95% Italy 301 0.85% Denmark 386 1.09%Thailand 357 1.01% Japan 7023 19.86% Finland 282 0.80%UK 3822 10.81% Korea 13 0.04% Indonesia 160 0.45%USA 12,116 34.27% Mexico 143 0.40% Ireland 198 0.56%

New Zealand 106 0.30% Israel 46 0.13%Philippines 54 0.15% Netherlands 663 1.88%Taiwan 73 0.21% Norway 249 0.70%

Portugal 47 0.13%South Africa 215 0.61%Sweden 383 1.08%Turkey 24 0.07%

Total 19,607 55.46% Total 10,202 28.86% Total 2372 6.70% Total 3175 8.98%


TD(

ownership of banks is 6.162% (Table 2 Panel B Last column). This statistic indicates that most of our observations are fromcountries with relatively low government control of banks.

Cost of entry is constructed using principal components factor analysis of three components: (i) number of steps with which astart-up has to comply to obtain legal status; (ii) time it takes to become operational; and (iii) cost of becoming operational as ashare of per capita GNP. Table 2 Panel A shows that the cost of entry measure varies from a low of−2.19 (Canada) to a high of 3.81(Indonesia) in our sample. Risk of expropriation (Ex_risk) measures the risk of outright confiscation or forced nationalization bythe state. This information is retrieved from LaPorta et al. (1998). The higher the score of this measure, the lower the risk ofexpropriation. This risk of expropriation score ranges from a minimum of 5.22 (Philippines) to a maximum of 9.98 (Netherlands,Switzerland and United States; Table 2 Panel A). We use an indicator variable that assumes a value of one when the score ofexpropriation is higher than the median (i.e. risk of expropriation lower than the sample median).9

3.2.1.3. Legal regime. Two variables are used to capture the legal regime of a country: legal origin and legal protection. Legal origindescribes whether the country has a common or civil law tradition (Civil). Legal origin of each country is presented in Column 1 ofTable 2 Panel A. About 41% of our observations are from countries with civil law tradition (Table 2 Panel B Last Column). Legalprotection is proxied by the index of anti-director rights as in Djankov et al. (2006). This is a revised version of the anti-directorrights index used in LaPorta et al. (2000a). The index is computed by adding one when: (i) the country allows shareholders to mailtheir proxy vote; (ii) shareholders are not required to deposit their shares prior to the General Shareholders' Meeting;(iii) cumulative voting or proportional representation of minorities on the board of directors is allowed; (iv) an oppressedminorities mechanism is in place; (v) the minimum percentage of share capital that entitles a shareholder to call for anExtraordinary Shareholders' Meeting is less than or equal to 10%; (vi) shareholders have preemptive rights that can only bewaivedby a shareholders meeting. The index ranges from zero to six. The higher the index, the better the investor protection is. The anti-director rights index of our sample ranges from 2 to 5 (Column 2 of Table 2 Panel A), with an average of 3.790 (Table 2 Panel B LastColumn). It indicates that most of our sample firms are in countries with relatively good investor protection regulation. We use anindicator variable Lowprotection, which takes on a value of 1 if a country has an anti-director rights index below or equal to themedian10 and 0 otherwise, to proxy for countries with low investor protection.

Djankov et al. (2006) suggest an alternative measure of investor protection—anti-self-dealing index. Unlike the anti-directorrights index, this measure focuses explicitly on the protection of minority shareholders against self-dealing by controllingshareholder(s). This index incorporates both the ex-ante private control of self-dealing and the ex-post private control of self-dealing. The index ranges from a minimum of 0 to a maximum of 1. The higher the index, the better the protection against self-dealing the country offers its investors. Data used to construct this index are gathered from the responses to a questionnairedesigned by the authors. Interested readers are referred to a detailed description of the survey provided in Djankov et al. (2006).The anti-self-dealing index ranges from 0.17 (Mexico) to 1 (Singapore) in our sample. The average index value of our sample is0.611 (Table 2 Panel B Last Column). We report the results of the analyses using anti-director rights index as a proxy for investorprotection in Section 4 and those using anti-self-dealing index as a proxy for investor protection in Section 5.

9 Due to data limitation, we have only two year's measures of state ownership of enterprise, 1995 and 2000. We use the 1995 score for the 1994–1997 analysesand the 2000 score for the 1998–2002 analyses. Further, the 1995 scores of state ownership of banks are used for all our analyses. Similarly, one measure of thecost of entry and one measure of the risk of expropriation for each country is used throughout the analyses.10 We compute the median of anti-director rights index for each year. Lowprotection takes on a value of 1 if its value is below or equal to the median of that year.

Table 2Descriptive statistics. Panel A presents values of the political economy variables, legal origin, and legal protection used in our analyses by country. Civil is anindicator variable which takes on a value of 1 if the country has civil law tradition, 0 otherwise. Index of anti-director rights is an index constructed by LaPorta et al(1998) and revised by Djankov et al. (2006) measure the anti-director rights of shareholders and acts as a proxy for legal protection of each country. Index of antiself-dealing is an index constructed by Djankov et al. (2006) to measure the legal protection of minority shareholders against expropriation by corporate insidersProp is the degree of proportionality of a country's voting system as constructed by Pagano and Volpin (2005). SOE is the score for state ownership of enterprise acomputed in Bushman et al. (2004). The 1995 scores are used for the 1994–1997 analyses and the 2000 scores are employed for the 1998–2002 analyses. StBank ithe proportion of the top 10 banks owned by the government in 1995. COE is the index developed by Bushman et al. (2004) to measure the cost for a foreigncompany to enter a country's market. Risk of expropriation is International Country Risk Guide's assessment of the risk of ‘outright confiscation’ or ‘forcednationalization’. The higher the score, the lower the risk of expropriation. A detailed description of the construction and sources of data for each variable is alsoprovided in Appendix A. Panel B presents the mean of variables grouped by proportionality scores during our sample period 1994–2002. Dividend/sale is total cashdividends paid to common and preferred shareholders divided by net sales. Dividend/earning is total cash dividends paid to common and preferred shareholderdivided by net income before extraordinary items. Dividend/cashflow is total cash dividends paid to common and preferred shareholders divided by operating cashflows. Operating cash flow is total funds from operations net of non-cash items from discontinued operations. Sales growth is the average growth rate of net sales inprevious 5 years (Salest/Salest−5)1/5−1. Growth decile is the rank decile for Sales growth. Firms are ranked by legal origin into 10 equal-size groups. It rangefrom 0 to 9 in ascending order. Total Assets is the book value of total assets. ROAt−1 is operating income divided by total book assets at the end of previous fiscayear (t−1). RETAIN t−1 is retained earnings divided by total common equity at the end of previous fiscal year (t−1). Ln (GDP) is the natural logarithm GDP pecapita in a country. TaxAdv is the index developed by LaPorta et al. (2000a) to measure the relative tax advantage of dividend versus capital gain in a country.

Country (1) (2) (3) (4) (5) (6) (7) (8)

Civil Index ofanti-directorrights

Index ofanti-self-dealing

Prop SOE StBank COE Risk ofexpropriation

Panel A: Values of political economy variablesArgentina 1 2 0.34 3 0.29 60.50 0.79 5.91Australia 0 4 0.76 1 1.26 12.33 −2.15 9.27Austria 1 2.5 0.21 3 6.55 50.36 0.53 9.69Belgium 1 3 0.54 3 4.00 27.59 −0.36 9.63Canada 0 4 0.64 0 0.90 0.00 −2.19 9.67Denmark 1 4 0.46 3 3.00 8.87 −1.70 9.67Finland 1 3.5 0.46 3 4.00 30.65 −1.31 9.67France 1 3.5 0.38 1 6.00 17.26 1.19 9.65Germany 1 3.5 0.28 2 4.00 36.36 0.32 9.90India 0 5 0.58 1 6.62 84.94 2.71 7.75Indonesia 1 4 0.65 3 3.00 42.90 3.81 7.16Ireland 0 5 0.79 3 0.00 4.48 −1.35 9.67Israel 0 4 0.73 3 8.00 64.64 −0.36 8.25Italy 1 2 0.42 1 4.00 35.95 1.74 9.35Japan 1 4.5 0.50 1 4.00 0.00 −0.05 9.67Korea 1 4.5 0.47 1 3.13 25.41 0.43 8.31Malaysia 0 5 0.95 0 6.00 9.93 0.33 7.95Mexico 1 3 0.17 1 2.25 35.62 3.13 7.29Netherlands 1 2.5 0.20 3 0.00 9.20 −0.08 9.98New Zealand 0 4 0.95 1 0.00 0.00 −2.06 9.69Norway 1 3.5 0.42 3 6.00 43.68 −1.44 9.88Pakistan 0 4 0.41 0 8.00 85.96 0.97 5.62Philippines 1 4 0.22 1 2.85 27.23 1.08 5.22Portugal 1 2.5 0.44 3 4.00 25.66 1.44 8.90Singapore 0 5 1.00 0 2.00 13.53 −0.67 9.30South Africa 0 5 0.81 3 5.17 0.00 −0.44 6.88Spain 1 5 0.37 2 6.00 1.98 1.40 9.52Sweden 1 3.5 0.33 3 4.00 23.20 −1.36 9.40Switzerland 1 3 0.27 2 2.00 13.35 −0.59 9.98Taiwan 1 3 0.56 1 8.00 76.51 −0.41 9.12Thailand 0 4 0.81 0 4.91 17.09 −0.32 7.42Turkey 1 3 0.43 3 2.85 56.46 0.92 7.00UK 0 5 0.95 0 2.00 0.00 −1.73 9.71USA 0 3 0.65 0 0.92 0.00 −1.91 9.98

Prop=0 Prop=1 Prop=2 Prop=3 All

Panel B: Average values of variables grouped by proportionality scoresDividend/earning 0.353 0.448 0.354 0.340 0.379Dividend/sale 0.022 0.011 0.016 0.022 0.019Dividend/cashflow 0.241 0.202 0.313 0.397 0.249Sales growth 0.099 0.020 0.048 0.061 0.070Growth decile 4.790 4.713 5.179 5.412 4.850Total assets (US$ Millions) 2902 2420 4376 2110 2790ROAt−1 0.108 0.055 0.073 0.086 0.088RETAIN t−1 0.646 0.398 0.259 0.288 0.517SOE 1.493 4.069 3.833 3.149 2.542StBank (%) 1.323 6.177 24.277 22.460 6.162COE −1.713 0.072 0.276 −0.507 −0.956Risk of expropriation 9.729 9.561 9.854 9.306 9.651

(continued on next page


.-.ss

s

slr

)

Table 2 (continued)

Panel B: Average values of variables grouped by proportionality scoresCivil 0.000 0.909 1.000 0.855 0.406Index of anti-director rights 3.661 4.184 3.635 3.436 3.790Index of anti-self-dealing rights 0.737 0.498 0.293 0.434 0.611Ln(GDP) 10.151 9.950 10.186 9.525 10.039TaxAdv 0.693 0.717 0.757 0.773 0.711Number of observations 19,607 10,202 2372 3175 35,356

Prop=0 Prop=1 Prop=2 Prop=3 All


3.2.1.4. Other country-specific measures. In addition to the political and legal variation across countries, we include the log of realGDP per capita [Ln(GDP)] in our analysis. Real GDP per capita of each country is computed by deflating each country's GDP to aconstant 1996 U.S. price level. This variable captures the “wealth” of a country. We include this variable in our analyses to ensurethat our political variables are not just capturing the effect of “rich” versus “poor” countries. The average log of real GDP per capitais 10.039 (Table 2 Panel B), corresponding to an average real GDP per capita of US$22,902.

As the dividend policy of a firm is influenced by the relative tax advantage/disadvantage of dividend versus capital gain in acountry, we include a tax advantage variable (TaxAdv) to control for such an effect. This dividend tax advantage variable isretrieved from LaPorta et al. (2000a) and is computed as the ratio of the value of US$1 distributed as dividend income (to anoutside investor) to the value of US$1 received in the form of capital gains when kept inside the firm as retained earnings. Themean dividend tax advantage is 0.711 (Table 2 Panel B Last Column).

3.2.2. Firm-specific variables

3.2.2.1. Growth. We use two different measures to proxy for a firm's potential growth (Growth). First, we measure a firm's growthin sales during the prior five years and then rank them into deciles by legal origin (i.e. Civil law origin or Common law origin).Firms in the top decile are those with the largest growth in sales during the last five years whereas those in the bottom decile arethose with the smallest growth. We assume that fast growers in the past will continue their momentum in the following year.Table 2 Panel B shows that our sample has an average sales growth of 0.070.

In addition to sales growth, we also use factor analysis to identify commonalities underlying: sales growth in the prior fiveyears, market-to-book of equity, market-to-book of assets, and price-to-earnings (P/E) ratio. Using the criterion of retainingfactors with eigenvalues greater than 1, the analysis reveals one common factor. We denote this factor as Growth Score. Firms arethen ranked by Growth Score into deciles by legal origin. We perform our analyses using both proxies of growth to examinewhether our results are robust to these different proxies. Results using the past sales growth as a proxy for potential growth arepresented in Section 4 while those using the common factor are presented in Section 5 as part of the sensitivity analyses.

3.2.2.2. Retained earnings. We include the proportion of a firm's equity that is made up of retained earnings (RETAIN) in ouranalysis. This variable measures the proportion of earned capital to total capital and proxies for a firm's life-cycle stage (DeAngeloet al. (2006)). DeAngelo et al. (2006) argue that firmswith a high proportion of total equity from retained earnings tend to bemoremature firms with large cumulative profits that make them self-financing; whereas firms with low retained earnings/total equitytend to be firms in the capital infusion stage and rely significantly on external financing. With ample free cash flow retained in thefirm, mature firms face higher agency cost or more severe overinvestment problem than younger firms. Table 2 Panel B shows thatour sample firms have on average 51.7% of total equity made up of retained earnings.

3.2.2.3. Profitability. We also control for a firm's profitability in our analyses. Fama and French (2001) document that dividendpayers have higher profitability than non-payers. A loss firm is less likely to have a high payout ratio than a profit firm (excludingliquidation dividend), ceteris paribus. As such, we include a firm's return on assets (ROA) in the prior period to control for thisperformance effect. The average ROA of our sample is 0.088 (Table 2 Panel B Last Column).

3.2.2.4. Firm size. Fama and French (2001) also points out that dividend payers are more than 13 times larger, as measured byassets, than non-payers. To control for this size effect, we include the log of assets (SIZE) in our analysis and our sample has anaverage of US$2790.41 million invested in assets.

4. Empirical analysis

4.1. Univariate analysis

Table 3 shows the average dividend payout of firms grouped by legal origin, electoral system and growth. Results show thathigh-growth firms (i.e. firms with growth rate above the sample median) in general have lower dividend payout ratio. However,the difference between the high and low-growth firms is significant only for firms in countries with low proportionality scores,regardless of the country's legal origin. That is, the negative correlation between growth and dividend payout is significant only for

Table 3Median dividend payout ratio by legal origin, growth, and electoral system. This table presents the median dividend/earnings ratio for our sample over theperiod 1994–2002. We first compute themedian ratio for each firm over the sample period. Thenwe calculate themedian value for all the firms in a specific group⁎⁎⁎, ⁎⁎, ⁎ indicate significance at the 1-, 5-, and 10-percent level, respectively.

GrowthNWorld Median Growth GrowthbWorld Median Growth Z-statistic

Common law countriesLow degree of proportionality (i.e. Prop=0,1) 0.117 0.338 −14.316⁎⁎⁎

High degree of proportionality (i.e. Prop=2,3) 0.222 0.218 0.093Civil law countries

Low degree of proportionality (i.e. Prop=0,1) 0.154 0.313 −11.185⁎⁎⁎

High degree of proportionality (i.e. Prop=2,3) 0.248 0.270 −1.085


.

either pure majoritarian-electoral countries (proportionality score=0) or countries with mostly majoritarian electoral system(proportionality score=1). Thus, these univariate results provide support for our second hypothesis. Interestingly, the differencebetween the high and low-growth firms is not always significant for common law countries or insignificant for civil law countries,as documented by LaPorta et al. (2000a).

4.2. Model specification

To test the first and second hypotheses regarding the impact of electoral system on a firm's dividend policy, we employ thefollowing model specification:

Dividi;t = α + β1Growthi;t + β2Civilc;t + β3Civilc;t⁎Growthi;t + β4Lowprotectionc;t+ β5Lowprotectionc;t⁎Growthi;t + β6Propc;t + β7Propc;t⁎Growthi;t+ β8Sizei;t + β9ROAi;t−1 + β10RETAINi;t−1 + β11Ln GDPc;t

� �

+ β12TaxAdvc;t + β13SOEc;t + β14SOEc;t⁎Growthi;t + β15StBankc;t+ β16StBankc;t⁎Growthi;t + β17COEc;t + β18COEc;t⁎Growthi;t+ β19Ex riskc;t + β20Ex riskc;t⁎Growthi;t + εi;t

ð1Þ

i is an index of firm; t is an index for year; and c is an index for country. Divid is the industry-adjusted dividend payout
whereratio. Dividend payout ratio is computed as total cash dividend paid to common and preferred shareholders deflated by earnings.Next, we compute the median dividend payout ratio for each industry in each country. We then calculate the world median foreach industry as themedian of dividend payout ratio across countries. The industry-adjusted dividend payout ratio is computed asthe firm's dividend payout ratio minus the world median of its industry. The stand-alone term of the degree of proportionality of acountry's electoral system (Prop) tells us whether countries with a high degree of proportionality tend to pay out a higher or lowerlevel of dividends. The interaction termwith growth tells us whether there is any difference in the correlation between growth andthe dividend payout ratio for firms in proportional-electoral versus those in majoritarian-electoral countries. In other words, thestand-alone term of political variables is for testing H1whereas the interaction term is for testing H2. All variables are as defined inSection 3. Since dividend payout ratio is non-negative, we use the Tobit regression for our analyses.
4.3. Regression analyses

4.3.1. Electoral systemTable 4 Panel A presents the results for the multivariate analysis. Column 1 presents the results for the effect of a country's

legal origin on a firm's dividend policy. Similar to the findings in LaPorta et al. (2000a), we observe a significantly negative coefficienton Civil (coefficient=−0.038, significant at the 5% level). Consistent with prior research, high-growth firms, on average, have lowerpayout ratio than other firms (coefficient of Growth=−0.060, significant at the 1% level). Further, this negative correlation betweengrowth and dividend payout ratio is less significant in civil law countries (coefficient of Civil⁎Growth=0.022, significant at the 1%level).

Second, we run the analysis with the anti-director rights index (Column 2). Firms in countries with low anti-director rightsindex tend to pay out lower dividends (coefficient of Lowprotection=−0.200, significant at the 1% level). As documented inLaPorta et al. (2000a), the negative correlation between growth and payout ratio also tends to be less significant in these countries(coefficient of Lowprotection⁎Growth=0.053, significant at the 1% level). Results in Column 3 show that when legalorigin and anti-director rights index are included in the analysis together, they both have incremental impact on a firm'sdividend payouts. Again, these are similar to the results reported in LaPorta et al. (2000a). Overall, the results in columns 1, 2and 3 are consistent with those in LaPorta et al. (2000a). These analyses suggest that our sample is similar to that inLaPorta et al. (2000a).

In Column 4, we introduce our political measure (Prop). The stand-alone term of Prop has a significantly negative effect(coefficient=−0.090, significant at the 1% level) on a firm's dividend payout ratio. Thus, consistent with our first hypothesis,firms in countries with proportional electoral system tend to have lower payout ratio than firms in majoritarian electoral system.An examination of the interaction term between the degree of proportionality and growth (Prop⁎Growth) indicates that the


negative correlation between growth and dividend payout tend to be less significant in proportional-electoral countries(coefficient=0.019, significant at the 1% level), consistent with our second hypothesis.

Next, and more importantly, we assess whether the type of electoral system has any incremental impact on a firm's dividendpolicy after controlling for the legal origin and anti-director rights index. To do this we have to include the electoral system, legalorigin and anti-director rights index concurrently in the model. Because the type of electoral system is highly correlated with thelegal origin (a correlation of 0.88), we run the analysis using orthogonalized variables of proportionality scores (Column 5) andlegal origin (Column 6). Without the orthogonalization process, it becomes difficult to disentangle the separate effects of theindependent variables (i.e. proportionality scores and legal origin) on the dependent variable (Maddala (1992), Gorsuch(1973)).11 We derive the orthogonalized variables using the modified Gram–Schmidt procedure (Golub and van Loan (1996)).This procedure generates orthogonalized proportionality scores (legal origin) such that the effect of legal origin (proportionalityscores) is removed from the scores (legal origin). In other words, the coefficient of the orthogonalized proportionality scores (legalorigin) captures the independent contribution of the proportionality scores (legal origin) to the dividend payout policy.

Results of the analysis using orthogonalized proportionality scores (Column 5) indicate that a country's electoral system has asignificantly negative effect (coefficient of Prop=−0.106, significant at the 1% level) on a firm's payout after controlling for thelegal regime. Further, its effect on the correlation between growth and dividend payout remains significantly positive (coefficientof Prop⁎Growth=0.014, significant at the 1% level). Thus, these results suggest that firms in proportional-electoral countries tendto pay out a smaller portion of their earnings as dividend and the negative correlation between a firm's growth and its dividendpayout is weaker in these proportional-electoral countries, providing support for H1 and H2. The coefficient of Civil12 continues tobe significantly negative (coefficient=−0.034, significant at the 5% level) and its interaction with growth (Civil⁎Growth) has asignificantly positive coefficient (coefficient=0.019, significant at the 1% level). Accordingly, the legal origin, anti-director rights,and the type of electoral system all have significant impact on a firm's dividend policy.

Column 6 shows that when the orthogonalized legal origin variable is used in the analysis, the degree of proportionality of acountry's electoral system continues to have a significantly negative effect on dividend payout ratio (coefficient of Prop=−0.088,significant at the 1% level). The correlation between growth and dividend payout ratio is weaker in the proportional-electoralcountries (coefficient of Prop⁎Growth=0.017, significant at the 1% level). However, the coefficient of Civil becomes significantlypositive while that of the interaction term Civil⁎Growth becomes significantly negative. That is, the effect of Civil is exactlyopposite to that documented in LaPorta et al. (2000a). These results suggest that the effect of legal origin on dividend payoutdocumented in LaPorta et al. (2000a) can actually be attributed to the proportionality of a country's electoral system. When theindependent effect of legal origin of a country (i.e. independent of that country's electoral system) is considered, civil law countriesno longer have a lower payout ratio nor is the negative relation between growth and payout less significant in these countries.

An examination of the control variables show that firm size (coefficient of Size=0.046, significant at the 1% level) has asignificantly positive effect on a firm's payout ratio (Columns 5 and 6). Firms in countries with high GDP per capita also tend to payout more of its earnings as dividends (coefficient of Ln(GDP)=0.028, significant at the 1% level). Similarly, firms in countries withhigh dividend tax advantage over capital gains also tend to pay out a larger portion of their earnings as dividends (coefficient ofTaxAdv=0.832, significant at the 1% level).

4.3.1.1. Cross-listing. Prior studies document that firms cross-list in other countries in order to grow their shareholder base,increase market valuation (Doidge et al. (2004) and Lang et al. (2003)), increase liquidity and the ability to raise equity (Reese andWeisbach (2002) and Lins et al. (2002)), and lower the cost of capital (Hail and Leuz (2009)). Cross-listing in countries with bettershareholder protection is also associated with increased transparency and improved governance. By cross-listing their stocks in aforeign exchange, firms subject themselves to the rules and regulations of the foreign country and exchange in addition to those oftheir domicile countries. Further, these firms are subject to an additional layer of monitoring by the foreign regulatory agencies,analysts and possibly journalists (Doidge et al. (2006)).

Given the potential stronger governance of cross-listing firms suggested by prior literature, we examine whether cross-listinghelps to lower the agency costs of free cash flows for firms in proportional-electoral countries. For our sample, we do not observeany firm in the pure majoritarian countries (i.e. Prop=0) cross-list in countries with higher proportionality scores. Cross-listingtends to be unilateral: firms in countries with high proportionality scores tend to cross-list in countries with lower proportionalityscores but not vice versa. Table 4 Panel B presents the results of the cross-listing analyses. We use the orthogonalized proportionalscores in our analyses. Column 1 shows that when firms in civil law countries cross-list in common law countries, they tend to paylower dividends. There is, however, no significant difference in the relationship between growth and dividend payout for thesecross-listing firms and their non-cross-listing peers. In Column 2, we examine the impact of cross-listing for firms in countries withlow protection (i.e. countries with belowmedian anti-director rights index). We observe that these cross-listing firms tend to payout a larger portion of their earnings as dividends (coefficient of Lowprotection⁎Cross=0.388. significant at the 5% level) and thenegative relation between growth and dividend payout tends to be stronger (coefficient of Lowprotection⁎Growth⁎Cross=−0.097. significant at the 1% level). Column 3 provides the results when firms in countries with high proportionality scores cross-

11 When we run the analysis using the raw proportionality scores and legal origin (i.e. measures without orthogonalization), the result is similar to that incolumn 6.12 In this analysis with orthogonalized proportionality scores, Civil captures two effects: the common effect of Civil and proportionality on dividend payout andthe independent effect of Civil. In Column 6, where the orthogonalized Civil is used in the analysis, Civil captures only the effect of Civil on dividend payoutindependent of proportionality.

Table 4Regression results: using 1994–2002 panel data. Tobit regression analysis of industry-adjusted dividend-to-earnings ratios on legal origins, anti-directors rights andproportionality levels using 1994–2002 panel data is performed. The dependent variable, industry-adjusted dividend-to-earnings ratio is calculated as in LaPorta et al(2000a). Themedian of Dividend/Earnings for each industry in each country (C_D/E) is first computed. Theworldmedian of C_D/E for each industry is then calculatedFinally, the industry-adjusted dividend-to-earnings ratio is computed as the difference between a firm's Dividend/Earnings and the world median of C_D/E for theindustry the firm belongs to. Eight industries are defined according to the firm's primary SIC. 1) agriculture; 2) mining; 3) construction; 4) manufacturing5) communications and transportation; 6)wholesale; 7) retail; 8) services.Growth is the rank decile for the average sales growth in the prior five years. Size is the log ototal assets. ROA is lagged return on assets, computed as operating income divided by total assets at the end of last year. RETAIN is retained earnings divided by totaequity evaluated at the end of prior year. Civil is an indicator variable that takes on a value of 1 if the country has civil legal origin. Civil⁎Growth is the interaction termbetween Civil andGrowth. Lowprotection is an indicator variable that equals one if Index of Anti-director Rights is smaller than or equal to themedian value for that yeaand zero otherwise. Prop is the proportionality score of the country. Its value ranges from0 to 3, with puremajoritarian-electoral countries take on a value of 0 and pureproportional-electoral countries takeonavalueof3. Ln(GDP) is logof realGDPper capita.TaxAdv is the relative taxadvantage ofdividendversus capital gain ina countryIt is calculated as the ratio of the value of US$1 distributed as dividend income to the value of US$1 received in the form of capital gains when kept inside the firm aretained earnings. Cross is the cross-listing indicator variable. A detailed description of the variables is provided in Appendix A. Because legal origin (Civil) andproportionality scores (Prop) have a high correlation (0.88), we used the orthogonalized variables in our analysiswhen both variables are present. Column 5 in panel Aand column 2 in panel B presents the results when Prop is orthogonalized on Civil while Column 6 in panel A and column 3 in panel B presents those when Civil iorthogonalized on Prop. We use the modified Gram–Schmidt procedure (Golub and Van Loan (1996)) to orthogonalize the variables. We also include year dummiein all our analysis. Standard errors are shown in parenthesis. ⁎⁎⁎, ⁎⁎, ⁎ indicate significance at the 1-, 5-, and 10-percent level, respectively.

(1) (2) (3) (4) (5) (6)Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.(Std error) (Std error) (Std error) (Std error) (Std error) (Std error)

Panel A Effect of electoral systemIntercept −0.764*** −0.792*** −0.871*** −0.638*** −0.929*** −0.882***

(0.081) (0.087) (0.088) (0.080) (0.088) (0.087)Firm characteristics Growth −0.060*** −0.052*** −0.060*** −0.064*** −0.061*** −0.065***

(0.002) (0.001) (0.002) (0.002) (0.002) (0.002)Size 0.046*** 0.046*** 0.045*** 0.046*** 0.046*** 0.046***

(0.002) (0.002) (0.002) (0.002) (0.002) (0.002)ROAt−1 −0.183*** −0.310*** −0.184*** −0.300*** −0.132** −0.132**

(0.052) (0.049) (0.052) (0.050) (0.052) (0.052)RETAIN t−1 0.018*** 0.017*** 0.018*** 0.016*** 0.017*** 0.017***

(0.003) (0.003) (0.003) (0.003) (0.003) (0.003)Legal regime Civil −0.038** −0.019 −0.034** 0.068***

(0.015) (0.015) (0.016) (0.007)Civil⁎Growth 0.022*** 0.017*** 0.019*** −0.004***

(0.003) (0.003) (0.003) (0.001)Lowprotection −0.200*** −0.184*** −0.016 −0.016

(0.043) (0.044) (0.045) (0.045)Lowprotection⁎Growth 0.053*** 0.044*** 0.024*** 0.024***

(0.007) (0.007) (0.007) (0.007)Political Economy Prop −0.090*** −0.106*** −0.088***

(0.008) (0.008) (0.008)Prop⁎Growth 0.019*** 0.014*** 0.017***

(0.001) (0.001) (0.001)Country specific Ln(GDP) 0.018** 0.017** 0.028*** 0.010 0.028*** 0.028***

(0.007) (0.008) (0.008) (0.007) (0.008) (0.008)TaxAdv 0.737*** 0.772*** 0.760*** 0.737*** 0.832*** 0.832***

(0.025) (0.026) (0.026) (0.025) (0.027) (0.027)χ2 3627.97 3573.85 3675.09 3688.90 3878.78 3878.78Number of observations 35,356 35,356 35,356 35,356 35,356 35,356

(1) (2) (3) (4)Coeff. Coeff. Coeff. Coeff.(Std error) (Std error) (Std error) (Std error)

Panel B Effect of cross-listingIntercept −0.953*** −0.925*** −0.931*** −0.927***

(0.088) (0.088) (0.088) (0.088)Firm characteristics Growth −0.061*** −0.061*** −0.060*** −0.061***

(0.002) (0.002) (0.002) (0.002)Size 0.051*** 0.046*** 0.046*** 0.046***

(0.002) (0.002) (0.002) (0.002)ROAt−1 −0.107** −0.131** −0.132** −0.132**

(0.052) (0.052) (0.052) (0.052)RETAIN t−1 0.017*** 0.017*** 0.017*** 0.017***

(0.003) (0.003) (0.003) (0.003)Legal regime Civil −0.023 −0.035** −0.034** −0.035**

(0.016) (0.016) (0.016) (0.016)Civil⁎Growth 0.019*** 0.019*** 0.019*** 0.019***

(0.003) (0.003) (0.003) (0.003)Civil⁎Cross −0.409***

(0.064)Civil⁎Growth⁎Cross 0.007

(0.010)



.

.

;fl

r

.s

ss

)

Table 4 (continued)

Panel B Effect of cross-listingLowprotection −0.007 −0.037 −0.014 −0.042

(0.045) (0.046) (0.045) (0.046)Lowprotection⁎Growth 0.024*** 0.031*** 0.023*** 0.029***

(0.007) (0.007) (0.007) (0.007)Lowprotection⁎Cross 0.388**

(0.180)Lowprotection⁎Growth⁎Cross −0.097***

(0.027)Political Economy Prop −0.103*** −0.106*** −0.110*** −0.107***

(0.008) (0.008) (0.008) (0.008)Prop⁎Growth 0.014*** 0.014*** 0.015*** 0.014***

(0.001) (0.001) (0.001) (0.001)Prop⁎Cross 0.049* 0.272***

(0.026) (0.085)Prop⁎Growth⁎Cross −0.012*** −0.047***

(0.004) (0.013)Country specific Ln(GDP) 0.026*** 0.027*** 0.028 0.028***

(0.008) (0.008) (0.008) (0.008)TaxAdv 0.851*** 0.833*** 0.832*** 0.833***

(0.027) (0.027) (0.027) (0.027)χ2 4044.77 3901.41 3888.67 3893.17Number of Observations 35,356 35,356 35,356 35,356

Legal regime



list in countries with lower scores. Results are similar to those of anti-director rights. When firms in countries with highproportionality scores cross-list in countries with lower scores, they tend to pay out a larger portion of their earnings as dividendsand the negative relationship between growth and dividend payout also tends to be stronger. Finally, we examine how cross-listing of firms domiciled in civil law, low protection and high proportionality score countries in common law, higher protectionand lower proportionality score countries affects their dividend policy. We observe that these firms tend to pay out a significantlylarger portion of their earnings as dividends. Further, the negative correlation between growth and dividend payout also tends tobe stronger. Overall, our results suggest that cross-listing helps to lower the agency costs of free cash flows for firms in countrieswith poor investor protection.

able 5ffect of change in electoral system on dividend payout ratio. Tobit regression analysis of the effect of a change in electoral system on the industry-adjustedividend payout is performed. Only countries with a change in the electoral system in the sample period are included in the analysis: New Zealand, Japan,hilippines, and Italy. Further, we limit our analysis to observations in these countries in the one year before the electoral system change and one year after thehange. Observations in the year of change are excluded from this analysis. The dependent variable, industry-adjusted dividend-to-earnings ratio is calculated asLaPorta et al. (2000a). The median of Dividend/Earnings for each industry in each country (C_D/E) is first computed. The world median of C_D/E for eachdustry is then calculated. Finally, the industry-adjusted dividend-to-earnings ratio is computed as the difference between a firm's Dividend/Earnings and theorld median of C_D/E for the industry the firm belongs to. Eight industries are defined according to the firm's primary SIC. 1) agriculture; 2) mining; 3)onstruction; 4) manufacturing; 5) communications and transportation; 6) wholesale; 7) retail; 8) services. Growth is the rank decile for the average sales growththe prior 3 years. Size is the log of total assets. ROA is lagged return on assets, computed as operating income divided by total assets at the end of last year. RETAINretained earnings divided by total equity evaluated at the end of prior year. PProp is a dummy variable that takes on a value of 1 if the observation is in a periodhen the country's electoral system is closer to the proportional system. For example, when Italy changes its electoral system from a pure proportional one to aystem dominated by the majoritarian rule in 1994, PProp takes on a value of 1 for Italian firms in 1993 and a value of 0 in 1995. A detailed description of theariables is provided in Appendix A. We also include country dummies in all our analysis. Standard errors are shown in parenthesis. ⁎⁎⁎, ⁎⁎, ⁎ indicate significancet the 1-, 5-, and 10-percent level, respectively.

Coefficient Std error

Intercept −0.010 0.174Firm characteristics Growth −0.097*** 0.018

Size 0.027** 0.012ROAt−1 −9.9525*** 1.627RETAIN t−1 −0.229 0.150

Political economy PProp −0.478*** 0.107PProp⁎Growth 0.083*** 0.022

Country dummies Philippines 0.819** 0.417Japan 0.969*** 0.164New Zealand 1.285*** 0.242χ2 143.92Number of observations 1216

TEdPcininwciniswsva

Table 6Regression results: relationship between dividend payout and other political economy variables. Tobit regression analysis of industry-adjusted dividend-toearnings ratios on legal origins, anti-directors rights, proportionality scores and other political economy variables (i.e. state ownership of banks and firms, cosof entry, risk of expropriation) using 1994–2002 panel data is performed. The dependent variable, industry-adjusted dividend-to-earnings ratio is calculatedfollowing LaPorta et al. (2000a). First, the median of Dividend/Earnings for each industry in each country (C_D/E) is computed. The world median of C_D/E foeach industry is then calculated. Finally, the industry-adjusted dividend-to-earnings ratio is computed as the difference between a firm's Dividend/Earnings andthe world median of C_D/E for the industry the firm belongs to. Growth is the rank decile for the average sales growth in the prior five years. Size is measured alog of total assets. ROA is lagged return on assets, computed as operating income divided by total assets at the end of last year. RETAIN is retained earningdivided by total equity evaluated at the end of prior year. Civil is an indicator variable that takes on a value of 1 if the country has civil origin. Civil⁎Growth ithe interaction term between Civil and Growth. Lowprotection is an indicator variable that equals one if Index of Anti-director Rights is smaller than or equal tothe median value for that year and zero otherwise. Prop is the proportionality score of the country. Its value ranges from 0 to 3, with pure majoritarian-electoracountries take on a value of 0 and pure proportional-electoral countries take on a value of 3. Ln(GDP) is log of real GDP per capita. TaxAdv is the relative taxadvantage of dividend versus capital gain in a country. It is calculated as the ratio of the value of US$1 distributed as dividend income to the value of US$1received in the form of capital gains when kept inside the firm as retained earnings. SOE is the score for state ownership of enterprise as computed in Bushmanet al. (2004). The 1995 scores are used for the 1994–1997 analyses and the 2000 scores are employed for the 1998–2002 analyses. StBank is the proportion othe top 10 banks owned by the government in 1995. COE is the index developed by Bushman et al. (2004) to measure the cost for a foreign company to enter acountry's market. Ex_risk is an indicator variable that takes on a value of 1 if the risk of expropriation score is above the sample median and 0 otherwise. Therisk of expropriation score is computed according to International Country Risk Guide's assessment of the risk of ‘outright confiscation’ or ‘forcednationalization’. The higher the score, the lower the risk of expropriation. A detailed description of the variables is provided in Appendix A. Because Civil andproportionality scores (and other political economy variables) have a high correlation, we used the orthogonalized variables in our analysis when the variableare present in the same analysis. Panel A presents the results when Prop (and other political economy variables) is orthogonalized on Civil while Panel Bpresents those when Civil is orthogonalized on Prop (and other political economy variables). We use the modified Gram–Schmidt procedure (Golub and VanLoan (1996)) to orthogonalize the variables. We also include year dummies in all our analysis. Standard errors are shown in parenthesis. ⁎⁎⁎, ⁎⁎, ⁎ indicatesignificance at the 1-, 5-, and 10-percent level, respectively.


Panel A Orthogonalized political economy variablesIntercept −0.462*** −0.976*** 0.866*** 0.169


(0.002) (0.002) (0.002) (0.002)Size 0.042*** 0.048*** 0.044*** 0.045***

(0.002) (0.002) (0.002) (0.002)ROAt−1 −0.150*** −0.044 −0.098** −0.147***

(0.052) (0.051) (0.051) (0.052)RETAIN t−1 0.015*** 0.012*** 0.013*** 0.016***

(0.003) (0.003) (0.003) (0.003)Legal regime Civil −0.055*** −0.101*** −0.108*** −0.066***

(0.015) (0.015) (0.016) (0.016)Civil⁎Growth 0.022*** 0.025*** 0.023*** 0.020***

(0.003) (0.003) (0.003) (0.003)Lowprotection −0.194*** 0.230*** 0.041 −0.132***

(0.046) (0.045) (0.045) (0.046)Lowprotection⁎Growth 0.039*** 0.006 0.010 0.026***

(0.007) (0.007) (0.007) (0.007)Political economy Prop −0.111*** −0.017** −0.129*** −0.077***

(0.007) (0.008) (0.008) (0.008)Prop⁎Growth 0.014*** 0.009*** 0.013*** 0.010***

(0.001) (0.001) (0.001) (0.001)SOE −0.133***

(0.007)SOE⁎Growth 0.013***

(0.001)StBank −0.222***

(0.009)StBank⁎Growth 0.008***

(0.001)COE −0.185***

(0.008)COE⁎Growth 0.008***

(0.001)Ex_risk 0.118***

(0.008)Ex_risk⁎Growth −0.006***

(0.001)Country specific Ln(GDP) 0.041*** 0.005 −0.144*** −0.085***

(0.016) (0.008) (0.010) (0.010)TaxAdv 0.217* 1.208*** 0.760*** 0.905***

(0.113) (0.029) (0.027) (0.027)χ2 4277.79 5568.79 4925.49 4214.45Number of Observations 35,356 35,356 35,356 35,356



-t

r

sss

l

f

s

)

able 6

Panel B Orthogonalized legal origin variableIntercept −0.330*** −0.981*** 0.754*** 0.198**


(0.002) (0.002) (0.002) (0.002)Size 0.042*** 0.048*** 0.044*** 0.045***

(0.002) (0.002) (0.002) (0.002)ROAt−1 −0.150*** −0.044 −0.098** −0.147***

(0.052) (0.051) (0.051) (0.052)RETAIN t−1 0.015*** 0.012*** 0.013*** 0.016***

(0.003) (0.003) (0.003) (0.003)Legal regime Civil 0.117*** −0.020*** 0.158*** 0.036***

(0.007) (0.008) (0.007) (0.008)Civil⁎Growth −0.009*** −0.001 −0.007*** −0.001

(0.001) (0.001) (0.001) (0.001)Lowprotection −0.194*** 0.230*** 0.041 −0.132***

(0.046) (0.045) (0.045) (0.046)Lowprotection⁎Growth 0.039*** 0.006 0.010 0.026***

(0.007) (0.007) (0.007) (0.007)Political economy Prop −0.049*** −0.052*** −0.047*** −0.079***

(0.009) (0.008) (0.009) (0.008)Prop⁎Growth 0.012*** 0.013*** 0.014*** 0.015***

(0.002) (0.002) (0.002) (0.001)SOE −0.047***

(0.004)SOE⁎Growth 0.006***

(0.001)StBank −0.221***

(0.008)StBank⁎Growth 0.001***

(0.000)COE −0.105***

(0.008)COE⁎Growth 0.005***

(0.001)Ex_risk 0.122***

(0.008)Ex_risk⁎Growth −0.006***

(0.001)Country specific Ln(GDP) −0.017*** 0.005 −0.144*** −0.085***

(0.008) (0.008) (0.010) (0.010)TaxAdv 0.903*** 1.208*** 0.760*** 0.905***

(0.028) (0.029) (0.027) (0.027)χ2 4277.79 5568.79 4925.49 4214.45Number of observations 35,356 35,356 35,356 35,356

(continued)



T

4.3.1.2. Change in electoral system. The above results suggest that the impact of electoral system dominates that of legal origin. Onepotential explanation is that unlike the legal origin, which is static, the electoral system varies over time and hence better matchesthe legal protection offered to investors. To further explore this issue, we examine whether a change in the electoral system of acountry leads to a change in the dividend policy of firms. Four countries changed their electoral system during the early 1990s:New Zealand, Italy, Japan, and Philippines. New Zealand changed their electoral system from a pure majoritarian one(proportionality score=0) to a mixed system (proportionality score=1) in 1993. Japan and Philippines also switched theirsystems from a pure majoritarian system (i.e. 0) to a more proportional one (i.e.1) in 1994 and 1996, respectively. Italy, on theother hand, switched from a pure proportional system (proportionality score=3) to a mixed system with mostly majoritarianelection rules (proportionality score=1) in 1994. The change in electoral system in these four countries provides us with analternative setting to test our hypothesis. Specifically, it allows us to test a strong form of our hypothesis: a change in electoralsystem leads to a change in the dividend policy.

Empirically, we examine whether there is any change in the dividend policy in the post-electoral-system-change periods bycomparing the dividend payout ratios in the year prior to the change in electoral system and the year after the change. Wediscard observations in the year of change in this analysis. The dividend payout ratio is then regressed on an indicator variable,PProp, that takes on a value of 1 in a period when the country's electoral system is closer to the proportional electoral system. Forexample, PProp takes on a value of 1 for Italy in 1993 and a value of 0 in 1995. We also include other firm characteristics in our


analysis.13 As the results in Table 5 indicate, firms tend to have lower dividend payout in the period when their electoral systemsare dominated by the proportional rule (coefficient of PProp=−0.478, significant at the 1% level). Also, the negative relationbetween growth and dividend payout tends to be weaker in this period (coefficient of PProp⁎Growth=0.083, significant at the1% level). When the country's electoral system is closer to the majoritarian system, firms pay out higher dividend and therelation between growth and payout tends to be stronger as well. These results provide further support for our hypotheses H1and H2.14

4.3.2. Other political economy measuresTable 6 presents the results when measures of the government's involvement in the economy—state ownership of enterprise

(SOE), state ownership of bank (StBank), cost of entry (COE), and the risk of expropriation (Ex_risk)—are each included in theanalyses after controlling for the legal origin (i.e. civil vs. common), the anti-director rights index, and the proportionality score ofeach country. As these political economy measures are also correlated with the legal origin of a country,15 we employ theorthogonalized variables approach as in the previous sections. That is, we first construct a set of orthogonalized variable of each ofthese political economy measure and proportionality scores (i.e. the portion of each political economy measure that isorthogonalized to legal origin). We then use this set of orthogonalized variables in our analysis of the model as specified in Eq. (1).Panel A presents the results of these analyses. Panel B presents the results when the legal origin is orthogonalized on theproportionality scores and each individual political economy variable in the respective analysis. In both panels, each individualpolitical economy variable is included in the analysis after controlling for the legal origin (i.e. civil vs. common), the anti-directorrights index, and the type of electoral system of each country.

Panel A shows that firms in countries with high state ownership in firms (Column 1) and banks (Column 2) tend to paylower dividends than their peers in countries with low state ownership. Cost of entry (Column 3) has a similar effect on afirm's payout policy. Firms in countries with high cost of entry pay out a lower portion of their earnings as dividends. Column4 shows that firms in countries with low risk of expropriation (i.e. high risk of expropriation scores) pay higher dividendsthan those in countries with high risk. Further, coefficients of the interaction terms between these political economymeasures and growth are also consistent with agency theory. The negative correlation between growth and dividends isstronger in countries with low state ownership, low cost of entry, and low risk of expropriation. Overall, the effects of thesepolitical economy measures on a firm's dividend policy are consistent with our expectation: firms in countries with betterinvestor protection (i.e. low state ownership, low cost of entry, and low risk of expropriation) tend to pay out a higherproportion of their earnings as dividends and the negative correlation between growth and dividend payout ratio is strongerin these countries.

Next, we observe that even with the inclusion of all these other political economy variables, the orthogonalized electoralsystem variable continues to have a significant impact on a firm's dividend policy. The coefficient of Prop is significantlynegative. This confirms our H1 that firms in countries with a proportional electoral system tend to have a lower payout ratiothan their peers in countries with a majoritarian electoral system. The significantly positive coefficient of the interaction termbetween Prop and Growth is consistent with our second hypothesis that the negative correlation between a firm's growthpotential and its dividend payout ratio is weaker in proportional-electoral countries. This result suggests that the politicaleconomy variables—the state ownership, cost of entry, risk of expropriation, and a country's electoral system—complementone another in capturing the protection a country offers its investors. As such, they play a significant role in determining theagency costs investors face. Also, these results confirm that the proportionality of a country's electoral system and otherpolitical economy variables have a significant independent effect (i.e. independent of the legal origin) on a firm's dividendpayout policy.

Panel B shows that when the orthogonalized legal origin is used in the analysis, its effect on dividend policy reverses inmost cases. We observe that firms in civil countries tend to pay higher dividends in all but one (the case of state-ownedbanks) cases when orthogonalized legal origin is used. The impact of its interaction term with Growth on the payout ratio isinconsistent with the overinvestment theory. The coefficients of the proportionality scores and other political economyvariables continue to have a significant impact on the payout ratio as documented in Panel A. These results further confirmour Hypotheses 1 and 2.

5. Sensitivity analyses

In this section, we report the results (some of them are untabulated) of several robustness checks. First, we investigate whethera country's electoral system affects the form of dividend payment, in addition to the amount paid out. To investigate this issue, we

13 Since New Zealand changes its electoral system in 1993, we do not have enough data to compute the average sales growth in the prior 5 sales. Therefore, weuse the average sales growth in the prior 3 years to rank our observations into deciles for this analysis.14 We acknowledge that the sample of our change analysis is small and may not be generalized to other sample periods. However, we are constrained by thenumber of countries that have changed their electoral sample in our sample period.15 The correlations between Civil and SOE, StBank, COE, and Ex_risk are 0.684, 0.426, 0.794, and −0.445 respectively. All correlations are significant at the 1%level.

able 7egression results: the choice between cash dividend and stock dividend. Two-stage regression analysis results on the impact of electoral system on dividendolicy using 1994–2002 panel data are reported. In the first stage, a logistic regression analysis of the determinants of whether a firm pays dividends (either cashividends or stock dividends) or not is performed. In the second stage, a Tobit regression analysis of industry-adjusted stock dividend-to-total-dividend ratio iserformed. We include the inverse Mill's ratio computed from the first-stage analysis in our second-stage analysis to control for self-selection bias. Growth is thenk decile for the average sales growth in the prior five years. Size is the log of total assets. ROA is lagged return on assets, computed as operating income dividedy total assets at the end of last year. RETAIN is retained earnings divided by total equity evaluated at the end of prior year. Civil is an indicator variable that takes onvalue of 1 if the country has civil legal origin. Civil⁎Growth is the interaction term between Civil and Growth. Lowprotection is an indicator variable that equals oneIndex of Anti-director Rights is smaller than or equal to the median value for that year and zero otherwise. Prop is the proportionality score of the country. Itsalue ranges from 0 to 3, with pure majoritarian-electoral countries take on a value of 0 and pure proportional-electoral countries take on a value of 3. Ln(GDP) isg of real GDP per capita. TaxAdv is the relative tax advantage of dividend versus capital gain in a country. It is calculated as the ratio of the value of US$1istributed as dividend income to the value of US$1 received in the form of capital gains when kept inside the firm as retained earnings. Cross is the cross-listingdicator variable. A detailed description of the variables is provided in Appendix A. Because legal origin (Civil) and proportionality scores (Prop) have a highorrelation (0.88), we used the orthogonalized proportionality score in our analysis. We use the modified Gram–Schmidt procedure (Golub and Van Loan (1996))orthogonalize the variables. We also include year dummies in all our analysis. Standard errors are shown in parenthesis. ⁎⁎⁎, ⁎⁎, ⁎ indicate significance at the 1-,

-, and 10-percent level, respectively.

(1) (2)Coeff. Coeff.

(Std error) (Std error)

Intercept −2.619*** −0.061*(0.313) (0.036)

Firm characteristics Growth −0.152*** 0.000(0.007) (0.001)

Size 0.204*** 0.004***(0.008) (0.001)

ROAt−1 0.060 0.112***(0.190) (0.015)

RETAIN t−1 2.056*** 0.011***(0.049) (0.003)

Legal regime Civil 0.517*** −0.073***(0.083) (0.004)

Civil⁎Growth 0.033** 0.002***(0.015) (0.001)

Lowprotection 0.770*** −0.010(0.167) (0.014)

Lowprotection⁎Growth 0.002 −0.005**(0.026) (0.002)

Political economy Prop −0.386*** 0.006***(0.044) (0.002)

Prop⁎Growth 0.059*** −0.000(0.008) (0.000)

Country specific Ln(GDP) −0.006 0.029***(0.028) (0.003)

TaxAdv 3.479*** −0.284***(0.108) (0.009)

Inverse Mill's ratio 0.012(0.014)

Number of observations 35,356 35,356


TRpdprabaifvlodincto5

employ a two-stage analysis. In the first stage, we examine the impact of the electoral system on whether or not a firm will pay adividend. In the second stage, we investigate if the electoral system affects the form of dividend payment. Results of this two-stageanalysis are reported in Table 7. Again, because proportionality score and legal origin are highly correlated, we use theorthogonalized proportionality score in our analyses. Column 1 presents the results of the first-stage analysis, which confirms thatfirms in proportional-electoral countries are less likely to make a dividend payment (coefficient of Prop=−0.386, significant atthe 1% level) and the negative correlation between potential growth and dividend payout is weaker for proportional-electoralcountries (coefficient of Prop⁎Growth=0.059, significant at the 1% level). Column 2 shows the results of the second stage. As afirm self-selects into paying dividends, we control for this selection bias in the second stage by including the inverse Mill's ratiocomputed from first stage. We observe that firms in proportional-electoral countries prefer stock to cash dividends (coefficient ofProp=0.006, significant at the 1% level). The growth status of a firm does not affect the choice between the different forms ofdividend payment.

Second, to examine whether our results are robust to the different measures of shareholder protection, we rerun our analysesusing the anti-self-dealing index as introduced in Djankov et al. (2006) instead of the anti-director rights index. The higher theindex, the better the protection against self-dealing a country offers. Results (untabulated) indicate that firms in countries withhigh anti-self-dealing index (a continuous measure) tend to pay out a larger portion of their earnings as dividends. This result isconsistent with the argument that it is easier to extract dividends in countries with better minority shareholder protection. The


significantly negative coefficient of the interaction term between the anti-self-dealing index and growth again confirms thatagency costs are lower in countries with good investor protection. More important still, results show that the degree ofproportionality of electoral systems continues to have a significant impact on a firm's dividend policy: the stand-alone term of Propremains significantly negative; while its interaction term with Growth continues to have a significantly positive coefficient.Accordingly, the effect of the degree of proportionality on a firm's dividend policy is robust to the different shareholder protectionproxies used.

Third, in addition to the proportion of earnings distributed as dividends, LaPorta et al. (2000a) also use sales and operating cashflows as deflators to measure dividend payout. We repeat the analyses in Section 4 using these two deflators and our results aresimilar to those reported above. Fourth, we repeat our analyses using the common factor from the factor analysis as a proxy for afirm's potential growth and the results are qualitatively the same as those reported in Table 4. Fifth, we also repeat our analysesusing a common factor from a factor analysis of all the five political economy variables and obtain similar results. Sixth, we run theanalyses year by year. The results continue to hold in each year. Seventh, as U.S. firms dominate our sample (about 34% of oursample), to ensure that our results are not driven solely by the U.S. firms, we first repeat all our analyses on a sample excluding U.S.firms. The results are robust to this sample change. We then employ the weighted least square to rerun our analyses. Our resultsremain unchanged.

Eighth, we rerun the above analyses for common and civil law countries separately.16 The degree of proportionality of acountry's electoral system continues to have a significantly positive impact on the relationship between dividend payouts andgrowth in both common and civil law countries. In addition, whenwe examine dividend payers only, we find that in proportional-electoral countries, firms have lower payouts and the correlation between growth and dividend payouts also tend to be weaker.These results suggest that the degree of proportionality of a country's electoral system influences a firm's payouts within bothcommon and civil law countries. Finally, we rerun the analyses using dividend payers only and our results are robust to such achange.

6. Conclusion

This paper documents evidence on how the political economy plays an important role in a country's agency costs anddividend policies. Our results show that a country's electoral system is a significant determinant of the degree of investorprotection—both legal regulation and its enforcement—a country can offer. In a proportional-electoral country where investorprotection is poor, minority investors have difficulties in enforcing their rights. Therefore, agency costs of free cash flows tendto be higher in these proportional-electoral countries. Specifically, we show that in a proportional-electoral country, firmstend to retain a significant portion of their earnings and only distribute a small portion as dividends to their shareholders.More importantly, our results suggest that minority shareholders have more difficulties in extracting dividends from low-growth firms in these proportional-electoral countries than investors in majoritarian-electoral countries. However, for firmsin proportional-electoral countries that cross-list in other countries, this agency cost tends to be lower than their peers thatdo not cross-list their stocks. We also find that when a country changes its electoral system from a pure majoritarian(proportional) system to a mixed one (i.e. the proportionality score of its electoral system increases (decreases)), firms'dividend payout ratios decrease (increase) and the negative relation between growth and dividend payout weakens(strengthens).

Our results continue to hold when other political economy variables, specifically, the government's involvement in theeconomy—state ownership of firms and banks, cost of entry, and risk of expropriation—are included in the analyses. Moreinteresting still, when we use an orthogonalized set of legal origin, electoral system, and other political economy variables in ouranalysis, the effect of legal origin on dividend payout is opposite to that reported in LaPorta et al. (2000a). Our results suggest thatthe effect of legal origin on a firm's dividend policy documented in LaPorta et al. (2000a) can actually be attributed to the electoralsystem and other political economy variables that we examine. Accordingly, future studies on the cross-country differences inagency problems should include the type of a country's electoral system and other dimensions of political economy in the modelspecification.

In conclusion, we find that a country's political economy, in particular the type of its electoral system, is a significantdeterminant of the minority investor protection it offers, which in turn affects the agency costs its investors face. With itsdynamic nature, a country's electoral system and other political economy variables are likely to have more impact on theprotection a country currently offers its investors. Althoughwe have examined only the agency problem of free cash flows, politicaleconomy likely has an impact on other agency costs as well. Future studies can examine its effect on these other agency problems.Also, we believe that in addition to the political economy variables we considered in this study, other dimensions of a country'spolitical economy can play a significant role in defining the investor environment a country offers (e.g. political connection asdocumented by Gul (2006)). Future studies can explore these other variables of interest and extend our understanding of therelationship between political economy and the development of a country's financial market and the agency problem its investorsface.

16 In these sub-sample regressions, we use the raw proportionality scores in our analyses of the model as specified in Eq. (1).


Acknowledgments

We would like to thank Shimen Chen, Timothy Chue, Richard Chung, Li Jiang, Karen Lai, David Mayers, Gordian Ndubizu,Bin Srinidhi, Wilson Tong, Steven Wang, Steven Wei, Yong Wang, Wayne Yu and other participants at the workshop at theschool of accounting and finance of the Polytechnic University of Hong Kong, University of California, Riverside, and DrexelUniversity, Accounting and Finance Association of Australia and New Zealand conference for their helpful discussions andcomments.

Appendix A. Variable definitions

Variables Definition Source

Legal regime Civil Indicator variable which takes on a value of 1 if the country has civil law tradition, 0 otherwise. LaPorta et al.(1998)

Index of anti-director rights

Index constructed by LaPorta et al. (1998) and revised by Djankov et al. (2006) to measure theanti-director rights of shareholders and acts as a proxy for legal protection of each country.

Djankov et al.(2006)

Lowprotection Indicator variable which equals 1 if the Index of anti-director rights of the country is smallerthan or equal to the median value for that year, 0 otherwise

Index of anti-self-dealing

Index constructed to measure the legal protection of minority shareholders againstexpropriation by corporate insiders.


Political economy Prop Degree of proportionality of a country's voting system as constructed by Pagano and Volpin(2005). It is computed as Proportional Representation−Plurality−Housesys+2.

World BankDatabase ofPoliticalInstitutions 2000

PProp Defined for countries that change their electoral systems during the sample period only. Anindicator variable that takes on a value of 1 in the period when the country's electoral system iscloser to the proportional system, 0 otherwise.

World BankDatabase ofPoliticalInstitutions 2000

SOE Score for state ownership of enterprise as computed in Bushman, et al. (2004). The 1995 scoresare used for the 1994–1997 analyses and the 2000 scores are employed for the 1998–2002analyses. The higher the SOE investment, the higher the rating.

EconomicFreedom of theWorld: 2005Annual Report

StBank Shares of the top 10 banks owned by the government in 1995. LaPorta et al.(2002)

COE A linear combination of three measures of the cost of entry into a country's market: (i) numberof steps with which a start-up has to comply to obtain legal status; (ii) time it takes to becomeoperational; and (iii) cost of becoming operational as a share of per capita GNP, constructedwith principal components factor analysis as in Bushman, Piotroski, and Smith (2004).


Ex_risk Ex_risk is an indicator variable that takes on a value of 1 if the risk of expropriation score isabove the sample median. The risk of expropriation score is computed according toInternational Country Risk Guide's assessment of the risk of ‘outright confiscation’ or ‘forcednationalization’. The higher the score, the lower the risk.

LaPorta et al.(1998)

Country specific Ln(GDP) Natural logarithm of real GDP per capita in 1996 constant US dollar. WorldDevelopmentindicators

TaxAdv Relative tax advantage of dividend versus capital gain in a country. Computed as the ratio ofthe value of US$1 distributed as dividend income (to an outside investor) to the value of US$1received in the form of capital gains when kept inside the firm as retained earnings.

LaPorta, et al.(2000a)

Firm specific Divid Divid is the industry-adjusted dividend payout ratio. Dividend payout ratio is computed astotal cash dividend paid to common and preferred shareholders deflated by earnings. Next, wecompute the median dividend payout ratio for each industry in each country. Eight industriesare defined according to the firm's primary SIC. 1) agriculture; 2) mining; 3) construction;4) manufacturing; 5) communications and transportation; 6) wholesale; 7) retail; 8) services.We then calculate the world median for each industry as the median of the dividend payoutratio across countries. The industry-adjusted dividend payout ratio equals a firm's dividendpayout ratio minus the world median of its industry.

Global Vantage

Sales growth Average growth rate of net sales in previous 5 years (Salest/Salest−5)1/5−1.Growth Except for Table 5, Growth is the rank decile for Sales growth. Firms are ranked by legal origin

into 10 equal-size groups. Ranges from 0 to 9 in ascending order. In Table 5, due to thelimitation of data, we rank the firms into deciles by the average sales growth in the prior threeyears.

Size Natural logarithm of total assets.ROAt−1 Operating income divided by total book assets at the end of previous fiscal year (t−1).RETAINt− 1 Retained earnings divided by total common equity at the end of previous fiscal year (t−1).Cross Indicator variable that takes on a value of 1 if a firm from a country with lower investor

protection cross-list in a country with better investor protection. Investor protection is definedwith regard to the legal origin, anti-director rights index or proportionality scores, 0 otherwise.


References

Ang, J., Blackwell, D., Megginson, W., 1991. The effect of taxes on the relative valuation of dividends and capital gains: evidence from dual-class British investmenttrusts. The Journal of Finance 46, 383–399.

Baker, M., Wurgler, J., 2004. A catering theory of dividends. The Journal of Finance 59, 1125–1165.Beck, T., Demirguc-Kunt, A., Levine, R., 2001. Law, politics and finance. Working Paper. The World Bank.Bell, L., Jenkinson, T., 2002. New evidence of the impact of dividend taxation and on the identity of the marginal investor. The Journal of Finance 57, 1321–1346.Bowler, S., Farrell, D.M., 1993. Legislator shirking and voter monitoring: Impacts of European parliament electoral systems upon legislator-voter relationships.

Journal of Common Market Studies 31, 45–70.Bushman, R., Piotroski, J., Smith, A., 2004. What determines corporate transparency. Journal of Accounting Research 42, 207–252.Cannavan, D., Finn, F., Grey, S., 2004. The value of dividend imputation tax credits in Australia. Journal of Financial Economics 73, 167–197.Christoffersen, S., Geczy, C., Musto, D., Reed, A., 2005. Crossborder dividend taxation and the preferences of taxable and nontaxable investors: evidence from

Canada. Journal of Financial Economics 78, 121–144.DeAngelo, H., DeAngelo, L., Skinner, D., 2004. Are dividends disappearing? Dividend concentration and the consolidation of earnings. Journal of Financial

Economics 72, 425–456.DeAngelo, H., DeAngelo, L., Stulz, R., 2006. Dividend policy and the earned/contributed capital mix: a test of the life-cycle theory. Journal of Financial Economics 81,

227–254.Dennis, D., Osobov, I., 2006. Why do firms pay dividends? International evidence on the determinants of dividend policy. Working Paper. Purdue University.Djankov, S., LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., 2002. The regulation of entry. Quarterly Journal of Economics 117, 1–37.Djankov, S., LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., 2006. The law and economics of self-dealing. Working Paper. NBER.Doidge, C., Karolyi, K., Lins, K., Stulz, R., Miller, D., 2006. Private benefits of control, ownership, and the cross-listing decisions. Working Paper. ECGI and Dice Center.Doidge, C., Karolyi, K., Stulz, R., 2004. Why are foreign firms listed in the U.S. worth more? Journal of Financial Economics 71, 205–238.Easterbrook, F., 1984. Two agency-cost explanations of dividends. American Economic Review 74, 650–659.Eije, H., Megginson, W., 2006. Dividend policy in the European Union. Working Paper. University of Oklahoma.Fama, E., French, K., 2001. Disappearing dividends: changing firm characteristics or lower propensity to pay? Journal of Financial Economics 60, 3–43.Farrell, D., 2001. Electoral Systems: A Comparative Introduction. Palgrave MacMillan, New York.Gaver, J., Gaver, K., 1993. Additional evidence on the association between investment opportunity set and corporate financing, dividend, and compensation

policies. Journal of Accounting & Economics 16, 125–160.Golub, G.H., Van Loan, C.F., 1996. Matrix computations, In: Hopkins, Johns (Ed.), 3rd ed. University Press, Baltimore.Gorsuch, R.L., 1973. Data analysis of correlated independent variables. Multivariate Behavioral Research 8, 89–107.Gul, F.A., 2006. Auditors' response to political connections and cronyism in Malaysia. Journal of Accounting Research 44, 931–963.Hail, L., Leuz, C., 2009. Cost of capital effects and changes in growth expectations around U.S. cross-listings. Journal of Financial Economics 93, 428–454.Jensen, M., 1986. Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review 76, 323–329.Lang, M., Lins, K., Miller, D., 2003. ADRs, analysts, and accuracy: does cross listing in the U.S. improve a firm's information environment and increase market value?

Journal of Accounting Research 41, 317–345.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and finance. Journal of Political Economy 106, 1113–1155.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000a. Agency problems and dividend policies around the world. The Journal of Finance 55, 1–33.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000b. Investor protection and corporate governance. Journal of Financial Economics 58, 3–27.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2002. Investor protection and corporate evaluation. The Journal of Finance 57, 1147–1170.Lasfer, M.A., 1997. On the motivation for paying scrip dividends. Financial Management 26, 62–80.Lee, Y.T., Liu, Y.J., Roll, R., Subramanyam, A., 2004. Taxes and dividend clientele: evidence from trading and ownership structure. Working paper. University of

California, Los Angeles.Levine, R., Zervos, S., 1998. Stock markets, banks, and economic growth. American Economic Review 88, 537–558.Li, W., Lie, E., 2006. Dividend changes and catering incentives. Journal of Financial Economics 80, 293–308.Lins, K., Strickland, D., Zenner, M., 2002. Do foreign firms issue equity on U.S. stock exchanges to relax capital constraints? Working Paper. University of Utah, Salt

Lake City.Maddala, G.S., 1992. Introduction to Econometrics, 2nd ed. Macmillan Publishing Company, New York.Modigliani, F., Perotti, E., 1997. Protection of minority interest and the development of security markets. Managerial and Decision Economics 18, 519–528.Pagano, M., Volpin, P., 2005. The political economy of corporate governance. American Economic Review 95, 1005–1030.Persson, T., Tabellini, G., 2006. Democracy and development: The devil in the details. American Economic Review 96, 319–324.Rae, D.W., 1969. Decision-rules and individual values in constitutional choice. The American Political Science Review 63 (1), 40–56.Rajan, R., Zingales, L., 2003. The great reversals: the politics of financial development in the 20th century. Journal of Financial Economics 69, 5–50.Reese,W.,Weisbach, M., 2002. Protection ofminority shareholder interests, cross-listings in the United States, and subsequent equity offerings. Journal of Financial

Economics 66, 65–104.Pinkowitz, L., Stulz, R., Williamson, R., 2006. Does the contribution of corporate cash holdings and dividends to firm value depend on governance? A cross-country

analysis. Journal of Finance 61, 2725–2751.Rozeff, M., 1982. Growth, beta and agency costs as determinants of dividend payout ratios. Journal of Financial Research 5, 249–260.Smith, C., Watts, R., 1992. The investment opportunity set and corporate financing, dividend, and compensation policies. Journal of Financial Economics 32,

263–292.Weston, F., Siu, J., 2003. Changing motives for share repurchases. Finance paper 3. Anderson Graduate School of Management, UCLA.

does political economy reduce agency costs? some evidence from dividend policies around the world

Documents