and Corporate Valuation

Transcription

1 THE JOURNAL OF FINANCE * VOL. LVII, NO. 3 * JUNE 2002 Investor Protection Corporate Valuation RAFAEL LA PORTA, FLORENCIO LOPEZ-DE-SILANES, ANDREI SHLEIFER, ROBERT VISHNY* ABSTRACT We present a model effects legal protection mority shareholders cash-flow ownership by a controllg shareholder on valuation firms. We n test th model usg a sample 539 large firms from 27 wealthy economies. Constent with model, we fd evidence higher valuation firms countries with better protection mority shareholders firms with higher cashflow ownership by controllg shareholder. RECENT RESEARCH SUGGESTS THAT THE EXTENT legal protection vestors a country an important determant development its fancial markets. Where laws are protective outside vestors well enforced, vestors are willg to fance firms, fancial markets are both broader more valuable. In contrast, where laws are unprotective vestors, development fancial markets stunted. Moreover, systematic differences among countries structure laws ir enforcement, such as htorical orig ir laws, account for differences fancial development (La Porta et al. (1997, 1998)). How does better protection outside vestors (both shareholders creditors) promote fancial market development? When ir are better protected by law, outside vestors are willg to pay more for fancial assets such as equity debt. They pay more because y recognize that, with better legal protection, more firm's prits would come back to m as terest or dividends as opposed to beg expropriated by entrepreneur who controls firm. By limitg expropriation, law raes price that securities fetch marketplace. In turn, th enables more entrepreneurs to fance ir vestments externally, leadg to expansion fancial markets. Although ultimate benefit legal vestor protection for fancial development has now been well documented, effect protection on valuation has received less attention. In th paper, we present a oretical empirical analys th effect. * La Porta Shleifer are from Harvard University, Lopez-de-Silanes from Yale University, Vhny from University Chicago. We thank Altan Sert Ekatera Trizlova for research asstance, Malcolm Baker, Simeon Djankov, Edward Glaeser, Simon Johnson, Rene Stulz, Daniel Wolfenzon, Jeff Wurgler, Luigi Zgales, three anonymous referees for comments, NSF for support th research. 1147

2 1148 The Journal Fance In th context, it important to recognize differences structure ownership control among firms both with across countries, sce se differences fluence power as well as centives controllg shareholders to expropriate mority shareholders. In most countries, large publicly traded firms are generally not widely held, but rar have controllg shareholders (La Porta, Lopez-de-Silanes, Shleifer (1999a)), who are entrenched at helm have ability to designate monitor corporate managers. These shareholders have power to expropriate mority shareholders, as well as creditors, with constrats imposed by law. The central agency problem such firms not failure Berle Means (1932) pressional managers to serve mority shareholders, but rar -ten legal-expropriation such morities, as well as creditors, by controllg shareholders (Shleifer Vhny (1997)). The power controllg shareholders to expropriate outside vestors moderated by ir fancial centives not to do so. An important source such centives equity or cash-flow ownership by controllg shareholder. In general, expropriation costly (Burkart, Gromb, Panunzi (1998)), refore higher cash-flow ownership should lead to lower expropriation, or thgs beg equal. Th exactly centive effect managerial cash-flow ownership emphasized by Jensen Mecklg (1976) modeled th paper. Usg company data from 27 wealthy economies, we n evaluate fluence vestor protection ownership by controllg shareholder on corporate valuation. We use Tob's q to measure valuation. We use orig a country's laws dex specific legal rules as dicators shareholder protection. To assess centive effects ownership, we focus on companies which have controllg shareholders, reby hopg to keep power to expropriate relatively constant. We consider cash-flow ownership by controllg shareholder as a measure centives. Th empirical strategy designed to allow us to assess effect vestor protection on corporate valuation holdg both power centives to expropriate constant, as well as to shed light on Jensen- Mecklg effect a new context. Constent with ory, better shareholder protection empirically associated with higher valuation corporate assets. Th fdg provides support for quantitative importance expropriation mority shareholders many countries, as well as for role law limitg such expropriation. We also fd evidence that higher centives from cashflow ownership are associated with higher valuations. Th research contues a number strs corporate fance. First, th paper relates to "law fance" literature, summarized recently La Porta et al. (2000b). In addition to identifyg effects vestor protection on fancial market development, th literature also shows how law fluences corporate ownership structures (La Porta et al. (1998, 1999a), Claessens, Djankov, Lang (2000)), dividend policies (La Porta et al. (2000a)), size firms (Kumar, Rajan, Zgales (1999)), efficiency

3 Investor Protection Corporate Valuation 1149 vestment allocation (Rajan Zgales (1998), Wurgler (2000)), economic growth (Demirguc-Kunt Maksimovic (1998), Beck, Leve, Loayza (2000)), even susceptibility a country's fancial markets to a crash (Johnson et al. (2000a)). Our study valuation also relates to work that exames votg premium different countries, tends to fd higher votg premia countries with ferior shareholder protection (e.g., Lease, McConnell, Mikkelson (1983), DeAngelo DeAngelo (1985), Zgales (1994), Nenova (2000)). Th paper also contues a large literature on effects corporate ownership structures on valuation. Demsetz Lehn (1985), Morck, Shleifer, Vhny (1988), McConnell Servaes (1990), Holderness, Kroszner, Sheehan (1999), among ors, study effect managerial ownership on pritability valuation firms United States. Morck et al. dtguh negative control effects (which y call entrenchment) from positive centive effects higher ownership. These studies U.S. data generally fd that valuation both positively affected by centives, negatively affected by entrenchment. More recently, Gorton Schmid (2000) fd evidence positive effects bank ownership on valuation German firms. In a study closely related to ours, Claessens et al. (2002) separate effects entrepreneurial control cash-flow ownership on valuation firms several East Asian countries. They fd that stronger entrepreneurial control adversely affects valuation, while cash-flow ownership affects it positively. Section I paper presents our model. Section II describes data. Section III presents a prelimary analys data, Section IV more complete regression analys. Section V dcusses robustness results, Section VI concludes. I. A Simple Model In th section, we present a model a firm fully controlled by a sgle shareholder, called entrepreneur. A sizable oretical literature deals with optimal ownership structures firms dependg on levels "private benefits control" (Grossman Hart (1988), Harr Raviv (1988), Bebchuk (1999), Wolfenzon (1999), Bennedsen Wolfenzon (2000)). High private benefits control, which typically accompany low levels shareholder protection, lead to heavy consolidation control equilibrium (Grossman Hart (1988), Zgales (1995), La Porta et al. (1999a), Bebchuk (1999)). Expropriatg outside vestors-even legally-may require secrecy, which mediates agast shared control (La Porta et al. (1999a)). Alternatively, an entrepreneur who gives up control vites hostile takeover bids from raiders who mselves wh to expropriate mority shareholders (Zgales (1995), Bebchuk (1999)). La Porta et al. show that, most countries, control deed heavily concentrated, usually hs a foundg family. Our assumption that re one controllg shareholder thus constent with available ory evidence.

4 1150 The Journal Fance We assume that th controllg shareholder has cash flow or equity ownership a firm. Entrepreneurs typically control a higher fraction votes than that cash flow by owng shares with superior votg, constructg ownership pyramids, or controllg board (La Porta et al. (1999a)). We assume that a exogenously determed by htory life-cycle firm, do not consider sale equity by entrepreneur. We also assume that entrepreneur manager. In data, controllg shareholders typically serve as managers (La Porta et al. (1999a)), but re are also stances entrepreneurs or ir families hirg pressional managers. Such separation control from management does not st way many forms expropriation by controllg shareholder mority shareholders. For example, controllg shareholder can set up companies with which firm deals on nonmarket terms, reby benefittg himself personally, without actually servg as chief executive ficer. The firm has amount cash I, which it vests a project with gross rate return R. The firm has no costs, so prits are RI. In th simple model, scale vestment does not matter. Not all prits are dtributed to shareholders on a pro rata bas. As a benefit controllg firm, entrepreneur can divert a share s prits from firm to himself, before he dtributes rest as dividends. Th diversion or tunnelg can take form salary, transfer pricg, subsidized personal loans, non-arms-length asset transactions,, some cases, outright ft. In most countries, much such diversion, short ft, legal, but requires costly transactions, such as settg up termediary companies, takg legal rks, so on (Burkart et al. (1998), Johnson et al. (2000b)). As a consequence costs such legal expropriation, when entrepreneur diverts share s prits, he only receives sri - c (k, s)ri, where c (k, s) share prits that he wastes when s diverted. We call c cost--ft function. Here k denotes quality shareholder protection; better protected are shareholders, more has to be wasted to expropriate a given share prits. Thus if law accommodates somethg close to outright ft, n k low c close to zero, but when law very strgent, n k high significant resources must be wasted to expropriate a given share prits. Formally, we assume that Ck > 0, cs > 0, css > 0, Cks > 0. The first equality means that stealg costlier a more protective legal regime; second means that margal cost stealg positive; third means that margal cost stealg res as more stolen; fal-crucial-equality means that margal cost stealg higher when vestors are better protected. We assume that cost c borne by entrepreneur rar than by all shareholders. Th assumption does not affect our prcipal results. Under se assumptions, entrepreneur maximizes a(1 - s)ri + sri - c(k,s)ri, (1) where first term h share after-ft cash flows (or dividends), remag two terms are h benefits from expropriation. Sce so-

5 Investor Protection Corporate Valuation 1151 lution for optimal s dependent RI, scale firm, we can assume that entrepreneur maximizes U= a(l - s) +s - c(k,s). (2) The first order condition for th problem given by which can be rewritten as Us = -ar+ 1 - c,(k,s) =O0, (3) C, (k, s) = 1 - a. (4) The last expression counterpart Jensen-Mecklg (1976) condition for consumption perquites by entrepreneur. It states that higher cash-flow ownership by entrepreneur, greater are h centives to dtribute dividends a nondtortionary way rar than expropriate mority shareholders a dtortionary way, hence lower equilibrium level expropriation for a given k. High cash-flow ownership reduces mority expropriation. We can now exame th first-order condition to derive several testable implications model. Differentiatg first-order condition with respect to k, we get ds Cks(k, S) + css (k, s) dk = 0. (5) We can rearrange terms recall our assumptions obta on function c to ds* Cks(k, s) < 0. (6) dk - css (k, s) Result 1: In countries with better shareholder protection, re less expropriation mority shareholders. Next, we differentiate first-order condition with respect to a to obta ds css (k, s ) -1. (7) da Under our assumptions on cost--ft function c, condition (7) implies ds* 1 dar da - css(k, S(k)<0. s *) (8)

6 1152 The Journal Fance Th gives us anor important comparative static (Jensen Mecklg (1976)). Result 2: Higher cash-flow ownership by entrepreneur associated with less expropriation mority shareholders. But what about implications th model for valuation? The most natural way to measure valuation th model with Tob's q, which given by q = (1 - s*)r. Tob's q here measures valuation firm from perspective a mority outside shareholder who does not receive any private benefits control, rar than from perspective entrepreneur who expropriates. The comparative statics results are given by dq ds* dk dk ' (9) dq ds* = - ~R > O, 1O dca da( dq (11) dr? We summarize se calculations as hyposes to be tested empirical part paper. Result 3: Or thgs beg equal: Hi. Firms more protective legal regimes should have higher Tob's qs; H2. Firms with higher cash-flow ownership by controllg entrepreneur should have higher Tob's qs; H3. Firms with better vestment opportunities should have higher Tob's qs. The model can be used to address one furr terestg question: Does margal benefit stronger centives from cash-flow ownership decrease as shareholder protection improves? That, it case that d2q <0? (12) dadka In prciple, th would be a plausible result, sce, with good shareholder protection, expropriation might be so costly that cash-flow ownership hardly matters. Unfortunately, general case, th result depends on a number

7 Investor Protection Corporate Valuation 1153 difficult to sign third derivatives. Specifically, differentiation yields followg conditions d2q d 2 S( dadk -R dadk. (13) Differentiatg equation (8) with respect to k, we obta ds d2s* Cssk(k,S + csss (k,s ) dk daodk (css (k, s *))2 *(4 In general, we cannot be sure that numerator last expression positive. However, special case a quadratic cost--ft function, we obta th result. Specifically, let c(k,sk) 'ks2 (15) In th case, all our assumptions yields on function c hold differentiation cssk(k,s*) = 1 > 0, (16) csss (k s=0. (17) In th case, expression (13) negative, we have anor testable prediction. Result 4: H4. For quadratic cost--ft function, effect entrepreneur 's cash-flow ownership on valuation lower countries with good vestor protection. The next several sections evaluate hyposes Hi to H4 empirically. First, however, we note that Shleifer Wolfenzon (2002) consider a more general model which an entrepreneur raes external equity funds to fance h vestment, h cash-flow ownership stake, a, as well as scale firm, are determed endogenously. In ir model, it case that a lower countries with better shareholder protection, but hyposes Hi to H4 still hold a market equilibrium for reasons virtually identical to those operatg our model.

8 1154 The Journal Fance A. Construction Sample II. Data Our 539-firm sample cludes largest 20 firms by market capitalization each 27 countries covered by La Porta et al. (1999a) that also have a shareholder who controls over 10 percent votes firm.' Usg largest firms makes it harder to fd benefits vestor protection for corporate valuation, sce large firms have access to substitute mechanms for limitg ir expropriation mority shareholders, cludg public scruty, reputation-buildg, foreign shareholdgs, ltgs on ternational exchanges. Shares largest firms are also most liquid, undermg concern that differences valuation are due to differences liquidity.2 We generally use richest countries based on 1993 per capita come, but exclude a number m that do not have significant stock markets (e.g., Kuwait, United Arab Emirates, Saudi Arabia). Like La Porta et al., we exclude all affiliates foreign firms. A firm defed as an affiliate a foreign company if at least 50 percent its votes are directly or directly controlled by a sgle foreign corporate owner. Unlike La Porta et al., we here exclude, banks fancial firms (SICs 6,000 through 6,999) because valuation ratios for fancial firms are not comparable to those nonfancial firms. As a rule, our companies come from WorldScope database. For Argenta, WorldScope coverage limited we use or sources to add five firms to sample. We generally rely on annual reports, 20-F filgs for companies with American Depositary Receipts (ADRs), proxy statements, -for several countries-country-specific books that detail ownership structures ir companies. We use Internet because many dividual companies (e.g., Scavia), as well as stitutions (e.g., Par Bourse The Fancial Times) have Web sites that conta formation on ownership structures. Virtually all our data are for , though we have 15 observations where data come from earlier years, a few from Because ownership patterns tend to be relatively stable, fact that ownership data do not all come from same year not a big problem. For several countries, our stard procedures do not work because dclosure so limited. For Greece, we take 20 largest corporations for which we could fd ownership data (mostly Bloomberg). For Mexico, we take 20 largest WorldScope firms that have ADRs. For Korea, different sources fer conflictg formation on corporate ownership structures 1 The only exception to rule 20 firms per country Israel, which has 19 firms sample. There are 21 Israeli nonfancial firms with nonmsg values common equity on WorldScope, one which widely held anor a foreign subsidiary. 2 We dcuss liquidity at greater length Section V.

9 Investor Protection Corporate Valuation 1155 chaebols. We were adved by Korean scholars that best source for chaebols (five cases) contas formation as 1984, so we use more stale but reliable data.3 To describe control companies, we identify all shareholders who control over 10 percent votes. In many cases, prcipal shareholders are mselves corporate entities fancial stitutions. We n fd major shareholders se entities, major shareholders major shareholders, so on, until we fd ultimate controllers votes. We say that a corporation has a controllg shareholder (ultimate owner) if th shareholder's direct direct votg firm exceed 10 percent. A shareholder has x percent direct control over firm A if (1) it controls directly firm B which, turn, directly controls x percent votes firm A; or (2) it controls directly firm C which turn controls x percent votes firm B (or a sequence firms leadg to firm B, each which has control over next one, i.e., y form a control cha), which directly controls x percent votes firm A. Havg 10 percent votes likely to suffice to have effective control a firm.4 When multiple shareholders have over 10 percent votes, we pick one with highest mimum votg stake along control cha. In addition to defg control, we compute cash-flow ownership controllg shareholder (or family), a from model. We measure a as fraction sample firm's cash flow owned directly directly by controllg shareholder. The shareholder may hold cash flow stake a directly. If alternatively a fraction x cash flows sample company owned by anor firm which controllg shareholder controls, if he owns fraction y cash flows th corporation, n a equal to product x y. If re are several chas ownership between controllg shareholder sample company, we add h cash-flow ownership across all se chas. Table I summarizes all variables. We use two rough proxies for protection mority shareholders, oretical k model. The first a dummy equal to one if a country's company law or commercial code common law orig, zero orwe. Because we have data on fewer countries than La Porta et al. (1998), we do not dtguh between French, German, Scavian civil law origs th paper. La Porta et al. show that countries with common law legal orig have better protection mority shareholders than do countries with civil law legal orig. The 3 Our results are robust to exclusion Greece, Korea, Mexico. ' Choosg a threshold below 10 percent not possible practice as many countries do not have matory reportg requirements for ownership below 10 percent. La Porta et al. (1999a) present evidence that shareholders controllg over 20 percent votes are typically mselves managers. Our workg paper (La Porta et al. (1999b)) used a smaller sample 371 firms a 20 percent control cutf. The results were similar to those presented here, but stattically weaker. The prcipal difference here a large expansion sample, not a change control cutf.

10 1156 The Journal Fance Table I The Variables Th table describes variables collected for 27 countries cluded our study. We present description sources from which each variable collected. Variable Description Common law Equals one if orig Company Law or Commercial Code country Englh Common Law, zero orwe. Source: La Porta et al. (1998). Civil law Equals one if Company Law or Commercial Code country origates Roman Law, zero orwe. Source: La Porta et al. (1998). Anti-director Formed by addg one when: (1) country allows shareholders to mail ir proxy vote, (2) shareholders are not required to deposit ir shares prior to General Shareholders' Meetg, (3) cumulative votg or proportional representation morities on board directors allowed, (4) an oppressed morities mechanm place, (5) mimum percentage share capital that entitles a shareholder to call for an Extraordary Shareholders' Meetg less than or equal to 10 percent ( sample median), or (6) when shareholders have preemptive that can only be waived by a shareholders meetg. The range for dex from zero to six. Source: La Porta et al. (1998). Tob's q The ratio market value assets to ir replacement value at end most recent fcal year. The market value assets proxied by book value assets mus book value equity mus deferred taxes plus market value common stock. The replacement value assets proxied by book value assets. Source: WorldScope (1997). Industry-adjusted Tob's q Industry-adjusted Tob's q computed as difference between Tob's q world median Tob's q for firm's dustry. Industry control groups are defed at three-digit SIC level whenever re are at least five WorldScope firms (excludg sample firms) that group at two-digit SIC level or- we. Source: WorldScope (1997).

11 Investor Protection Corporate Valuation 1157 Growth sales (GS) Geometric average annual percentage growth lagged (net) sales for up to three years dependg on data availability. Sales are expressed (US$) dollars. Source: WorldScope (1997). Industry-adjusted GS Average annual dustry-adjusted growth lagged (net) sales for up to three years dependg on data avail- ability. Industry-adjusted GS computed as difference between GS world median GS for firm's dustry. Industry control groups are defed at three-digit SIC level whenever re are at least five WorldScope firms (excludg sample firms) that group, at two-digit SIC level orwe. Source: WorldScope (1997). Control Fraction firm's votg, if any, owned by its controllg shareholder. To measure control, we combe a shareholder's direct (i.e., through shares regtered her name) direct (i.e., through shares held by entities that, turn, she controls) votg firm. A shareholder has an x percent direct control over firm A if: (1) she controls directly firm B which, turn, directly controls x percent votes firm A; or (2) she controls directly firm C which turn controls firm B (or a sequence firms leadg to firm B, each which has control over next one, i.e., y form a control cha) which, turn, directly controls x percent votes firm A. A group n companies form a cha control if each firm 1 through n - 1 controls consecutive firm. A firm our sample has a controllg shareholder if sum her direct direct votg exceeds 10 percent. When two or more shareholders meet our criteria for control, we assign control to shareholder with largest (direct plus direct) votg stake. CF Fraction firm's ultimate cash-flow, if any, owned by its controllg shareholder. CF are computed as product all equity stakes along control cha (see description control for an explanation "control chas"). Wedge The difference between control cash flow.

12 1158 The Journal Fance reason for th fdg may be that judiciary philosophy common law countries allows judges to broadly terpret certa prciples, such as fiduciary duty, hence authorizes m to prohibit more forms mority expropriation (Johnson et al. (2000b)). Alternatively, common law countries may protect mority vestors better because corporate owners have less political fluence. Recent dcussions political fluence large shareholders shapg corporate governance clude Rajan Zgales (2000) La Porta et al. (2000b). The second measure vestor protection dex antidirector, also from La Porta et al. (1998). Th dex reflects such aspects mority as (1) ease votg for directors, (2) freedom tradg shares durg a shareholders meetg, (3) possibility electg directors through a cumulative votg mechanm or proportional representation morities on board, (4) extence a grievance mechanm for oppressed mority shareholders, such as a class-action lawsuit or appraal for major corporate decions, (5) extence a preemptive right to new security sues by firm, (6) percentage votes needed to call an extraordary shareholder meetg. La Porta et al. (1997) fd that antidirector score predicts stock market development across countries. Our measure valuation Tob's q computed for most recent fcal year available, typically The denomator q book value assets. The numerator book value assets mus book value common equity deferred taxes plus market value common equity. To compute market value equity for firms with multiple classes common, Worldscope multiplies total number outstg shares or than preferred stock by price per share most widely traded class common stock. Sce shares with lower votg tend to have larger floats than those with higher votg (La Porta et al. (1999a)), th procedure typically prices equity usg prices lower-votg shares. Th exactly what we want conceptually, sce model's predictions concern value equity to outside mority shareholders, that, without votg premium that reflects power to divert.5 As a check, we have rerun all our regressions excludg 83 firms with multiple classes shares. The results were very similar. To reduce weight outliers, we censor Tob's q at 5th 95th percentiles by settg extreme values to 5th 95th percentile values, respectively.6 5 In practice, importance votg premia computg market values mor our sample, sce roughly half firms with multiple classes shares are from Scavian countries, where votg premia tend to be very low (Nenova (2000)). 6 In our workg paper (La Porta et al. (1999b)), we also present results for cash-flowto-price ratios as measures valuation. We have computed se results for present sample as well. The cash-flow-to-price results provide equally strong support for positive effect vestor protection on valuation, but weaker result on benefits cash-flow ownership. The terpretation cash-flow-to-price plagued by questions wher cash flow reported before or after expropriation as well as wher rk premium constant across countries. Because se problems, we do not present se results.

13 Investor Protection Corporate Valuation 1159 For each firm, we also compute its annual sales growth rate over most recent three fcal years. Th our rough proxy for value growth opportunities. We cap growth sales at both 5th 95th percentiles to avoid problems with outliers. We use sales rar than earngs growth to avoid dealg with volatility manipulability earngs. In Section V, we dcuss or measures vestment opportunities. We also compute dustry adjusted Tob's q. For each company a given dustry, we make th adjustment relative to world-wide rar than country-wide average for that dustry (i.e., take out world-wide dustry effects rar than country-wide dustry effects). Consider computation dustry-adjusted growth sales. We first fd world-wide median growth real sales for each dustry usg all WorldScope (nonsample) firms sample countries. The dustry-adjusted growth sales for a company difference between its own sales growth world median sales growth its dustry.7 The idea that different dustries might be at different stages maturity growth that determe ir valuations. One fal sue differences consolidation rules fancial statements among countries, which can, prciple, dtort our measures Tob's q. Accountg procedures can result excessive consolidation both sales balance sheet items when partially owned subsidiaries are treated as if y are fully owned. To address th problem, we collect data on consolidation procedures used by sample firms for ir subsidiaries with asset values at least U.S. $10 million. We also collect data on equity values excessively consolidated subsidiaries (Vsub) usg market values for publicly traded subsidiaries book values for privately held ones. We n recompute Tob's q as follows: Debtpar + Vpar + (1- f)vsub (18) Assetspar where Debtpar consolidated book value debt parent company, Vpar market value equity parent company,,b fraction equity that parent company owns subsidiary, Assetspar consolidated book value assets parent company. As it turns out, excessive consolidation limited practical importance: The correlation between adjusted unadjusted Tob's q Accordgly, we only report results usg unadjusted Tob's q.8 7Industry defed at three-digit SIC level whenever re are at least five WorldScope nonsample firms control group at two-digit SIC level when previous condition not met. In 13 cases, we have a two-digit SIC dustry defition. 8 In our workg paper (La Porta et al. (1999b)), we have verified that results are unaffected by th adjustment computation Tob's q.

14 1160 The Journal Fance III. Prelimary Results on Investor Protection Valuation Table II presents relationship between legal orig (civil versus common law) valuation across 27 countries. For each country, we present median Tob's q sample firms, anti-director score, median sales growth rate firms from that country. We also compute median medians each variable among civil law common law countries separately. Table II confirms that common law countries have sharply higher anti-director scores than civil law countries do. The median antidirector score two for civil law countries four for common law countries. The prcipal result Table II that companies with controllg shareholders countries have higher valuations common law than civil law countries. The median medians (MOM) Tob's q for common law, for civil law countries (t = -2.16). However, growth rate sales also higher (though not stattically significantly) common law countries, suggestg that vestment opportunities ir companies face may be better. The result that better vestor protection associated with higher valuation also obtas if we divide countries accordg to wher ir antidirector score above or below median, although difference MOMs no longer stattically significant. The results from sortg by legal orig also hold for sample all WorldScope firms, as reported our workg paper (La Porta et al. (1999b)). Th prelimary evidence constent with ma prediction our model. At same time, model generates additional predictions, which may also mean that a simple comparon medians omits important confoundg effects. In next section, we turn to regression analys to exame all predictions. IV. Regression Analys Table III presents relationship between valuation, vestor protection, ownership. We estimate all regressions usg country rom effects. The natural alternative specification fixed effects. However, fixed effects are not feasible our setup given that re no with-country variation legal variables.9 The rom effects specification supported by Breusch Pagan (1980) Lagrange multiplier test which strongly rejects null hypos that errors are dependent with countries. The rom effects specification uses both with between country variation cash-flow ownership to estimate its effects on valuation, but does not treat firms a given country as dependent observations. Instead, stard errors are adjusted to reflect cross-correlation between observations due 9 It feasible, however, to estimate cash-flow coefficient usg only withcountry variation that variable, as we do Table V.

15 Investor Protection Corporate Valuation 1161 Table II Data Panel A classifies countries by legal orig presents medians by country for both sample 539 firms that have a controllg shareholder. Panel B reports tests medians for civil versus common law legal orig. Variables are defed Table I. Anti-director Growth Sales Country Rights Tob's q (%) Panel A: Medians Argenta Austria Belgium Denmark Fl France Germany Greece Italy Japan Korea Mexico Nerls Norway Portugal Spa Sweden Switzerl Civil law median Australia Canada Hong Kong Irel Israel New Zeal Sgapore United Kgdom United States Common law median Sample median Panel B: Test Medians (z-stattic) Civil versus common law -3.53* -2.16** *Significant at 1 percent level. "*Significant at 5 percent level.

16 1162 The Journal Fance Table III Rom-Effects Regressions for Raw Data The table presents results rom-effect regressions for sample 539 firms with a controllg shareholder. The dependent variable Tob's q. The dependent variables are: (1) growth sales, three-year geometric average annual growth rate sales; (2) common law, a dummy variable that equals one if legal orig Company Law or Commercial Code country which firm corporated Common Law zero orwe; (3) anti-director, dex anti-director country which firm corporated; (4) CF Rights, fraction cash-flow held by firm's controllg shareholder; (5) teraction between CF common law; (6) teraction between CF anti-director. Table I provides defitions for variables. Stard errors are shown parenses. (1) (2) (3) (4) Constant * * * * (0.0836) (0.0900) (0.1649) (0.1635) Growth sales * * * * (0.1403) (0.1408) (0.1403) (0.1411) Common law *** ** (0.1400) (0.1472) Anti-director ** (0.0490) (0.0478) CF *** ** (0.1334) (0.2680) CF * common law (0.2367) CF * anti-director (0.0828) Overall R *Significant at 1 percent level. "*Significant at 5 percent level. ***Significant at 10 percent level. to common country components. In all regressions, we control for past growth rate sales as a measure vestment opportunities for each firm. We report four regressions. In first two, we use common law dummy as measure shareholder protection, second two, antidirector score. For each measure shareholder protection, we present two specifications. First, we use shareholder protection as only dependent variable, besides sales growth rate. From pot view model, th corresponds to regressg Tob's q on k R. Second, we also clude regression cash-flow controllg shareholder as well as an teraction term between that measure vestor protection variable. Th corresponds to testg full model, sce we are regressg Tob's q on k, R, a, k a. o Recall that Hyposes 2 4 predict that centives from cash-flow ownership should exert a positive fluence on valuation, that th fluence should be greater countries with ferior protection shareholders.

17 Investor Protection Corporate Valuation 1163 In Table III, growth sales has a positive coefficient all specifications. When common law dummy cluded alone, it significant at 10 percent level. But when it cluded along with cash-flow teraction term, its coefficient significant at 5 percent level, implies that Tob's q res by an impressive 0.28 as one moves from civil to common law orig, or thgs beg equal. The coefficient on cashflow 0.26 significant at 10 percent level, although coefficient on teraction term not. These parameter estimates imply that, as cash-flow ownership res from 20 percent to 30 percent, Tob's q res by civil law countries, common law countries. When cluded alone, anti-director score significant. But when cash-flow teraction term are added to regression, coefficient on anti-director becomes significant at 5 percent level suggests that an improvement score by two pots (from civil law to common law median) raes Tob's q by about 0.2. The coefficient on cash-flow variable 0.52 significant at 5 percent level. The coefficient on teraction term significant, although its sign constent with prediction ory. These results imply that as cash-flow ownership res from, say 20 percent to 30 percent, Tob's q creases by about 0.05 when anti-director score two, 0.03 when anti-director score four. The centive effect small even civil law countries.10 Table IV presents results with dustry-adjusted data. The result that vestor protection associated with higher valuation about as significant as it Table III. The result that centives are associated with higher valuation when vestor protection poor also hold, as do results that benefits cash-flow ownership for valuation are higher low vestor protection countries. The results are thus similar to those without dustry adjustment supportg hyposes presented Section I. These results are constent with predictions ory concerng effects vestor protection entrepreneurial cash-flow ownership on firm valuation.1" They provide direct evidence expropriation mority shareholders by controllg shareholders. Although our data do not provide direct evidence how expropriation works, papers by Johnson et al. (2000b) usg case studies, by Bertr, Mehta, Mullaathan (2002) usg flow funds side pyramidal groups, show that non-arms-length transactions among firms are an important tunnelg channel. 10 We have also reestimated se regressions usg Tob's q from previous year from next year as alternative dependent variables, omittg utilities (because y are regulated companies), omittg firms with large government ownership. We also tried removg all firms with Tob's q below 5th percentile above 95th percentile. The results are robust to se changes specification. " In an effort to tell a more prece story, we have cluded both antidirector score legal orig regression (both become significant), added a proxy for quality law enforcement regression (significant), added a measure difference between control cash flow controllg shareholder (significant).

18 1164 The Journal Fance Table IV Rom-Effects Regressions for Industry-Adjusted Data The table presents results rom-effects regressions for sample 539 firms with a controllg shareholder. The dependent variable dustry-adjusted Tob's q. The dependent variables are: (1) dustry-adjusted growth sales, three-year geometric average annual growth rate dustry-adjusted sales; (2) common law, a dummy variable that equals one if legal orig Company Law or Commercial Code country which firm corporated Common Law zero orwe; (3) anti-director, dex antidirector country which firm corporated; (4) CF, fraction cash flow held by firm's controllg shareholder; (5) teraction between CF common law; (6) teraction between CF anti-director. Table I provides defitions for variables. Stard errors are shown parenses. (1) (2) (3) (4) Constant (0.0704) (0.0758) (0.1389) (0.1318) Industry-adjusted growth sales * * * (0.1336) (0.1337) (0.1336) (0.1336) Common law *** ** ( ) (0.1196) Anti-director ** (0.0415) (0.0377) CF ** * (0.1158) (0.1583) CF * common law (0.2058) CF * anti-director (0.0481) Overall R *Significant at 1 percent level. **Significant at 5 percent level. ***Significant at 10 percent level. V. Robustness Results In th section, we address five sues robustness. (1) Can differences market liquidity among countries account for our results? (2) Do we have good measures vestment opportunities? (3) Are our results driven by selection most valuable firms each country? (4) Are results somehow driven by more complex ownership structures, such as teractions between multiple large shareholders? (5) What can be done about endogeneity ownership? It might be argued that valuation levels are low when capital markets are small, as y are low vestor protection countries. Firms may fd it costly to rae external fancg countries with small capital markets for agency reasons we emphasize or, alternatively, because vestors require a premium to compensate for lower liquidity small fancial markets (Pagano (1989)). A liquidity premium may expla lower ratios cash flow to

19 Investor Protection Corporate Valuation 1165 price countries with small capital markets, but does not suffice to expla lower qs. Regardless required rate return, we expect firms a non-agency-cost world to vest until margal Tob's q equal to one. We are measurg average rar than margal q, but re no reason to expect difference between margal average q to be higher common law than civil law countries. The differences required rates return thus cannot account for our results, but private component cash flows can.12 Past sales growth may be a poor measure vestment opportunities, which might conceivably bias our results. We have tried three alternative measures vestment opportunities: ratio capital expenditure to sales (Berger Ofek (1995)), past growth assets, more prece but less econometrically appropriate actual future sales growth. The conclusions we draw are robust to se changes specification. Anor possible bias our analys may come from fact that firms common law countries are larger (Kumar et al. (1999)), larger firms might have higher valuations, perhaps because y have better vestment opportunities. We use two strategies to address th concern. First, we have redone our analys controllg for logarithm sales. When we do that, anti-director dummy significant when combed with cash-flow, but common law dummy significant. Size always significant, but its coefficient actually negative, not positive, as objection suggests. Second, we reestimate our results usg a broader sample, which cludes a large number smaller firms. Th sample adds to our basic sample widely held firms that we come across process constructg 539-firm sample, as well as sample medium-size firms (those with capitalization around $500 million) from La Porta et al. (1999a). The results are robust to th expansion sample. As anor sensitivity check, we have focused on firms with only one shareholder with a stake above 10 percent. The idea to make sure that our results are driven by effects described model rar than by teractions between multiple large shareholders. The results also hold th sample 422 firms where re only one large shareholder. Interestgly, centive effect larger sample firms with a sgle large shareholder. In empirical analys Section IV, we have assumed that a exogenous. Our defense th assumption that, generally speakg, ownership patterns are extremely stable, especially outside United States, are shaped largely by htories companies ir foundg families. 12 Some sample firms have ADRs traded United States, which generally require better dclosure corporate formation. We have vestigated effect havg an ADR on valuation, found a small positive effect for firms common law countries no effect for firms civil law countries. Th result also constent with view that liquidity drives our results, sce, on that ory, benefit an ADR for valuation should be higher less liquid markets ( civil law countries).

20 1166 The Journal Fance Still, we next consider some ways to get around th assumption, some empirical implications endogenous ownership. We also dcuss centives results more broadly. Accordg to Shleifer Wolfenzon (2002) or studies, cash-flow ownership may vary systematically across countries, dependg on ir legal systems. The centive effect we are pickg up may n be a crosscountry not just a cross-firm effect. Our terpretation ownership coefficient may n be problematic due to th endogeneity. Lackg struments, we can address th problem by focusg solely on with-country variation cash-flow ownership (fixed effects estimation), which arguably more exogenous to legal regime. In Table V, we use as our cash-flow variable for each firm measure relative to country mean. Th adjustment reduces magnitude significance vestor protection results for both raw dustry-adjusted Tob's q to 10 to 15 percent significance level. The centive effects also show up at about 10 percent significance level, dicatg that our earlier fdgs are not driven solely by differences among countries. A possible reason for weakness (despite stattical significance) our centive results comes from assumg that degree control by controllg shareholder constant (effectively nearly absolute) as long as he has over 10 percent votes. If degree control res as votg crease, moreover if cash-flow are correlated with votg, n our centive measure may be capturg greater control by domant shareholder rar than greater centives. And if greater control associated with greater expropriation, n greater control would fset beneficial centive effect. We might be fdg a stronger centive effect if we could dentangle centives from power. The same problems that plague U.S. data separatg centives from control may also plague ternational data. Table VI presents country means control, cash-flow, ir difference, which we call wedge. A wedge close to zero pots to small deviations from one-share-one-vote (through pyramids or multiple classes stock). The clear message Table VI that, th sample, deviations from one-share-one-vote are small. Although some countries mean wedge exceeds 15 percent, world-wide mean means about 10 percent. Th mean 11 percent for civil law countries, a stattically significantly lower 8 percent for common law countries. Th evidence pots to problem separatg econometrically cash-flow ownership from control. VI. Conclusion In th paper, we presented a simple ory consequences corporate ownership for corporate valuation different legal regimes. We have also tested th ory usg data on companies from 27 wealthy countries around world. The results generally confirm crucial predictions

21 Investor Protection Corporate Valuation 1167 Table V Rom Effects Demeaned Ownership The table presents rom-effects regressions for cross section 539 firms 27 countries that have a controllg shareholder. The depen- dent variables are: (1) Tob's q Panel A; (2) dustry-adjusted Tob's q Panel B. The dependent variables are: (1) growth sales, three-year geometric average annual growth rate sales; (2) dustry-adjusted GS, three-year geometric average annual growth rate dustry-adjusted sales; (3) common law, a dummy variable that equals one if legal orig Company Law or Commercial Code country which firm corporated common law zero orwe; (4) anti-director, dex anti-director country which firm corporated; (5) CF, deviations from country mean fraction cash-flow held by firm's 10 percent controllg shareholder (defed Table I); (6) teraction between CF common law; (7) teraction between CF anti-director. Table I contas defitions for variables. Stard errors are shown parenses. Independent Variables CF Rights * Growth Industry-adjusted Common Anti-director CF CF Rights * Anti-director Overall Constant Sales GS Law Rights Rights Common Law Rights R2 Panel A: Tob's q * l "l'* (0.0836) (0.1403) (0.1400) * * ***l 'll* (0.0865) (0.1403) (0.1452) (0.1345) (0.2372) ' * (0.1649) (0.1403) (0.0490) * * ! (0.1698) (0.1400) (0.0505) (0.2675) (0.0826) Panel B: Industry-adjusted Tob's q * ** (0.0704) (0.1336) (0.1199) * ** (0.0728) (0.1337) (0.1242) (0.1275) (0.2250) * (0.1389) (0.1336) (0.0415) * l l-! (0.1440) (0.1333) (0.0430) (0.2536) (0.0783) *Significant at 1 percent level. *Significant at 5 percent level. **Significant at 10 percent level.

22 1168 The Journal Fance Table VI Ownership, Control, Wedge Panel A classifies countries by legal orig presents average cash-flow, control, wedge (defed as difference between control cash flow ). Table I defes variables. Panel B reports tests means for civil versus common law legal orig. Country N CF Rights Control Rights Wedge Panel A: Means Argenta Austria Belgium Denmark Fl France Germany Greece Italy Japan Korea Mexico Nerls Norway Portugal Spa Sweden Switzerl Civil law mean Australia Canada Hong Kong Irel Israel New Zeal Sgapore United Kgdom United States Common law mean Sample mean Panel B: Test Means (t-stats) Civil versus common law ** * 0.90 *Significant at 5 percent level. **Significant at 10 percent level. ory, namely that poor shareholder protection penalized with lower valuations, that higher cash-flow ownership by controllg shareholder improves valuation, especially countries with poor vestor protection. The result on centives also constent with fdgs Claessens

23 Investor Protection Corporate Valuation 1169 et al. (2000) on a larger sample companies from Asia. Th evidence directly supports importance expropriation mority shareholders by controllg shareholders many countries, for role law limitg such expropriation. As such, it adds an important lk to explanation consequences vestor protection for fancial market development. The evidence exps our understg role vestor protection shapg corporate fance, by clarifyg roles which both centives law play deliverg value to outside shareholders. REFERENCES Bebchuk, Lucian, 1999, The rent protection ory corporate ownership control, Manuscript, Harvard Law School. Beck, Thorsten, Ross Leve, Norman Loayza, 2000, Fance sources growth, Journal Fancial Economics 58, Bennedsen, Morten, Daniel Wolfenzon, 2000, The balance power close corporations, Journal Fancial Economics 58, Berger, Philip, Eli Ofek, 1995, Diversification's effect on firm value, Journal Fancial Economics 37, Berle, Adolf, Garder Means, 1932, The Modern Corporation Private Property (Mc- Millan, New York). Bertr, Marianne, Paras Mehta, Sendhil Mullaathan, 2002, Ferretg out tunnelg: An application to Indian busess groups, Quarterly Journal Economics 117, Breusch, Trevor S., Adrian Pagan, 1980, The lagrange multiplier test its applications to model specification econometrics, The Review Economic Studies 47, Burkart, Mike, Den Gromb, Fausto Panunzi, 1998, Why higher takeover premia protect mority shareholders, Journal Political Economy 106, Claessens, Stj, Simeon Djankov, Joseph Fan, Larry Lang, 2002, Expropriation mority shareholders East Asia, Journal Fance 57. Claessens, Stj, Simeon Djankov, Larry Lang, 2000, The separation ownership control East Asian corporations, Journal Fancial Economics 58, DeAngelo, Harry, Lda DeAngelo, 1985, Managerial ownership votg : A study public corporations with dual classes common stock, Journal Fancial Economics 14, Demirguc-Kunt, Asli, Vojlav Maksimovic, 1998, Law, fance, firm growth, Journal Fance 53, Demsetz, Harold, Kenneth Lehn, 1985, The structure ownership: Causes consequences, Journal Political Economy 93, Gorton, Gary, Frank Schmid, 2000, Universal bankg performance German firms, Journal Fancial Economics 58, Grossman, Sanford, Oliver Hart, 1988, One-share-one-vote market for corporate control, Journal Fancial Economics 20, Harr, Milton, Artur Raviv, 1988, Corporate governance: Votg majority rules, Journal Fancial Economics 20, Holderness, Clifford, Rall Kroszner, Denn Sheehan, 1999, Were good old days that good? Changes managerial stock ownership sce great depression, Journal Fance 54, Jensen, Michael, William Mecklg, 1976, Theory firm: Managerial behavior, agency costs, ownership structure, Journal Fancial Economics 3, Johnson, Simon, Peter Boone, Alasdair Breach, Eric Friedman, 2000a, Corporate governance Asian fancial cr, Journal Fancial Economics 58, Johnson, Simon, Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Shleifer, 2000b, Tunnelg, American Economic Review Papers Proceedgs 90, Kumar, Krhna, Raghuram Rajan, Luigi Zgales, 1999, What determes firm size? NBER Workg paper 7208.

24 1170 The Journal Fance La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, 1999a, Corporate ownership around world, Journal Fance 54, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, Robert Vhny, 1997, Legal determants external fance, Journal Fance 52, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, Robert W. Vhny, 1998, Law fance, Journal Political Economy 106, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, Robert W. Vhny, 1999b, Investor protection corporate valuation, NBER Workg paper La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, Robert W. Vhny, 2000a, Agency problems dividend policies around world, Journal Fance 55, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, Robert W. Vhny, 2000b, Investor protection corporate governance, Journal Fancial Economics 58, Lease, Ronald C., John McConnell, Wayne Mikkelson, 1983, The market value control publicly traded corporations, Journal Fancial Economics 11, McConnell, John, Henri Servaes, 1990, Additional evidence on equity ownership corporate value, Journal Fancial Economics 27, Morck, Rall, Andrei Shleifer, Robert W. Vhny, 1988, Management ownership market valuation: An empirical analys, Journal Fancial Economics 20, Nenova, Tatiana, 2000, The value a corporate vote private benefits: A cross-country analys, Manuscript, Harvard University. Pagano, Marco, 1989, Tradg volume asset liquidity, Quarterly Journal Economics 104, Rajan, Raghuram, Luigi Zgales, 1998, Fancial dependence growth, American Economic Review 88, Rajan, Raghuram, Luigi Zgales, 2000, The great reversals: The politics fancial development 20th century, Mimeo, University Chicago. Shleifer, Andrei, Robert Vhny, 1997, A survey corporate governance, Journal Fance 52, Shleifer, Andrei, Daniel Wolfenzon, 2002, Investor protection equity markets, Journal Fancial Economics, forthcomg. Wolfenzon, Daniel, 1999, A ory pyramidal structures, Manuscript, Harvard University. Worldscope, 1997, Worldscope Database (Dclosure, Inc., New York). Wurgler, Jeffrey, 2000, Fancial markets allocation capital, Journal Fancial Economics 58, Zgales, Luigi, 1994, The value votg right: A study Milan stock exchange, The Review Fancial Studies 7, Zgales, Luigi, 1995, Inside ownership decion to go public, Review Economic Studies 62,