Property Rights Protection, Corporate Transparency, and Growth *

Transcription

1 Property Rights Protetion, Corporate Transpareny, and Growth * Art Durnev Vihang Errunza Alexander Molhanov MGill University MGill University Massey University Abstrat In ountries with seure property rights, orporate improves investment effiieny and inreases growth by alleviating information asymmetry. However, in ountries with inseure property rights, greater an inrease the risk of government expropriation and lower the benefits of. Using data from 59 industries in 69 ountries, we find that in ountries with weak property rights protetion, industries that would benefit the most from exhibit worse investment effiieny, grow slower, pratie worse governane and have less informative earnings ompared to industries that an effiiently operate at minimal levels of. * Art Durnev, Desautels Faulty of Management, MGill University, 1001 Sherbrooke Street West, Montreal, Quebe H3A 1G5, Canada. Tel: (514) Fax: (514) art.durnev@mgill.a. Vihang Errunza, Desautels Faulty of Management, MGill University, 1001 Sherbrooke Street West, Montreal, Quebe H3A 1G5, Canada. Tel: (514) Fax: (514) vihang.errunza@mgill.a. Alexander Molhanov, Department of Commere, Massey University, Private Bag , North Shore MSC, Aukland, New Zealand. Tel: ext Fax: a.e.molhanov@massey.a.nz. Art Durnev s researh is supported by the Institut de Finane Mathématique de Montréal (IFM2) and the Soial Sienes & Humanities Researh Counil of Canada (SSHRC). Vihang Errunza s researh is supported by the Bank of Montreal hair in finane and banking, IFM2 and SSHRC. We thank Pat Akey for superb researh assistane. 1

2 I. Introdution Transpareny is thought to play a key role in the well-funtioning of finanial markets and in the effiieny of investment deisions. 1 In transparent markets, apital is direted to profitable projets and withdrawn from unprofitable ones (Wurgler (2000), Verdi (2006)). When insiders possess better information than outside investors, information asymmetry may result in a suboptimal resoure alloation. If the degree of asymmetry is suffiiently high, markets ould even fail due to lemons problem (Akerlof (1970)). The importane of has been widely reognized by both aademis and market regulators, resulting in numerous rules and regulations being introdued over time to ensure timely and reliable dislosure of finanial information reating standards to whih firms must adhere. 2 At the same time, firms may hoose to maintain a level of higher than that required by regulatory authorities. The benefits of suh ations are intuitive and are related to reduing unertainty surrounding the firm whih, in turn, ould lower ost of apital, inrease liquidity, improve value estimates in orporate ontrol ontests (Healy and Palepu (2001)); redue ontrating osts assoiated with managerial ompensation (Core (2001)); signal managerial talent (Trueman (1986)); and derease litigation osts. Another stream of literature argues that the benefits of greater are limited to ompanies that operate in ountries with developed apital markets, strong investor protetion, and seure property rights. The benefits ome from, for example, lower ost of external finaning, more transparent ontrats with suppliers, and lower ost of information for ustomers. With low risk of government interferene, firms an better benefit from and hoose an appropriate level of in order to maximize investment effiieny and growth. On the other hand, the benefits of orporate are lower in markets with greater risks of government intervention, that is, when property rights are less seure (Watts and Zimmerman (1978 and 1986), Leuz and Oberholzer-Gee (2006), Stulz (2005), Durnev and Fauver (2008), Durnev and Guriev (2008)). In ountries with predatory governments, firms that 1 We onsider in a broad sense. In our view, is determined by a set of market mehanisms that failitate information aquisition and proessing by investors; a market struture that enhanes informational ontent for onsumers and suppliers and makes the ontrat between eonomi parties more effiient. 2 Sarbanes-Oxley At of 2002 is perhaps the most well-known reent legislation aimed at inreasing in finanial markets. Canada, Japan, Australia, among others, have enated similar legislations. 2

3 are more -dependent (for example, those that need to raise external apital in the near future) fae greater risk of government expropriation sine it is easier for governments to take profits away from more transparent firms. Therefore, ompanies that would benefit the most from high levels of orporate in ountries with inseure property rights may hoose to at opaquely, in order to mitigate the risk of governmental expropriation or lower the osts that they may fae if their predatory governments hoose to fore them to pay eonomi rents. This may ompliate, among other things, firm s aess to external finaning and make it harder to write enforeable ontrats with suppliers due to moral hazard and adverse seletion problems. Thus, when the risk of expropriation is higher, firms in greater need of would alloate apital less effiiently and grow more slowly than those that an funtion well at minimum levels of. Hene, this paper examines how orporate affets the effiieny of industrial investment and growth onditional on the degree of property rights protetion. We use a differene-in-differene methodology and examine within-ountry variation of investment effiieny and growth aross industries onditional on the industry need for and ountry degree of property rights protetion. By ontrolling for unobserved industry and ountry harateristis, the differene-in-differene approah redues the problems of endogeneity, omitted variables bias, and model misspeifiation that typially affet ross-setional regressions. Moreover, we ontrol for other variables that an affet, apital alloation, and growth inluding quality of legal environment, ountry eonomi performane, degree of finanial development, and ountry trade openness. The foal variable in our analysis is industry measure of. We build upon a line of reasoning established in Rajan and Zingales (1998), who investigate the link between dependene on external finaning and eonomi growth. The authors argue that the U.S. markets an be used as a benhmark, assuming that firms true need for external finaning is observed in the U.S. However, using only the U.S. data to onstrut the benhmark for in every ountry is problemati, as the need for for the same industry an differ aross ountries. 3 Moreover, depending on a ountry, an industries propensity to expropriation varies. Therefore, we extend the Rajan and Zingales (1998) methodology into what we all a gap methodology. The gap measures the distane between industries pratied 3 The osts and benefits of are shown to vary aross industries. Adamati and Pfeiderer (2000) disuss how inentives to dislose information depend on industry ompetitiveness. A large degree of variation in has also been doumented empirially in Durnev, Mork, and Yeung (2004) and Bharat, Pasquariello and Wu (2007). 3

4 and industries desired ( norm). The pratied is the observed industry alulated using loal market data. The norm is measured using levels in developed markets, suh as the U.S., Canada, Germany, and the U.K. We onstrut the index using four attributes: informational, insider, aounting, and dislosure. Informational measures the degree to whih stok pries reflet available information. Mork, Yeung, and Yu (2000) argue that if stok returns move asynhronously with market returns, more firm-speifi information is impounded in stok pries. Insider is aptured by a measure developed in Llorente et al. (2002) whih assesses the intensity of trading on private information. Aounting analyzes how informed stok returns are with respet to future hanges in earnings. If aounting numbers reflet appropriate information in a timely manner, urrent stok returns should reflet more information about future earnings (Durnev et al. (2003), Lundholm and Myers, 2002)). Finally, we use the number of dislosed items in firms finanial statements to measure the amount of dislosed information. Our results suggest that in ountries with less seure property rights, industries with higher gap (differene between observed and norm) experiene sub-optimal apital alloation and slower industry growth. This is beause under risk of government intervention, -dependent industries disproportionately redue leading to a higher gap. It is important to note that our study does not imply that orporate by itself is detrimental to a firm, but rather the need for in a weak property rights environment. Our study ontributes to the existing literature in several ways. First, it adds to the literature on institutional development, property rights protetion, and unoffiial eonomy (Johnson, Kaufmann, and Zoido-Lobaton (1998) and Friedman et al. (2000)) by showing that institutional development affets investment effiieny and growth through orporate. Seond, we doument that the risk of government interferene affets industries asymmetrially; only industries that require high levels of exhibit worse investment effiieny and slower growth when property rights are weak. In other words, firms annot fully apture the benefits of when property rights are not proteted. Third, it has been assumed that orporate is largely determined by firms need for external finane as argued by Rajan and Zingales (1998), who doument slower growth in industries with high dependene on external finaning when the finanial system is 4

5 underdeveloped. However, by expliitly ontrolling for external finaning needs, we show that our results go beyond the previously doumented external finaning hannel. Indeed, the effet of external finaning is not bigger than the effets of the other determinants of. Moreover, our results remain robust when we alulate the gap between norm and pratie by removing the need for external finaning omponent from the gap. Thus, our results go signifiantly beyond those of Rajan and Zingales (1998) and unveil the relationship further than that of external finaning need and finanial development. 4 We onsider several alternative theories that an drive our results. First, one an hypothesise that a firm an ompensate for the lak of by adjusting other governane mehanisms. We argue and show empirially that this is not the ase. Indeed, more dependent industries pratie worse governane and engage in more earnings management when property rights are not proteted. Seond, it is possible that orporate insiders ollude with the government to expropriate ompany resoures at the expense of minority shareholders. Suh expropriation requires opaity and may result in ineffiient investment deisions. Under this senario one would expet that the impat of the need for for investment effiieny (ontingent on property rights protetion) is stronger when ollusion is easier to ahieve, for example, in firms with high state ownership. We show empirially that this is not the ase. Third, one an argue that our results are driven by high orrelation between investor protetion and property rights protetion. However, we show that the results are robust when we use the part of property rights protetion that is not explained by investor protetion. Finally, while the differene-in-differene methodology aims at reduing endogeneity problem, our results an still be driven by reverse ausality or errors in variables. We further address endogeneity by running two types of instrumental variables regressions. First, ethnolinguisti frationalization, proportion of Catholis, distane from equator, and settlers mortality rates are used as instruments of property rights protetion. Seond, we estimate all regressions in differenes using past values of level variables as instruments. Our results are onfirmed. The rest of the paper is organized as follows. Setion II ontains bakground and hypotheses development. In Setion III, we desribe the empirial methodology, the data, and the variables. 4 In Setion II, we disuss alternative motives of exeeding minimum standards. 5

6 Setion IV provides empirial analysis. Setion V disusses aveats and robustness issues. Setion VI onludes. II. Bakground and Hypotheses A. The Role of Transpareny and Transpareny Dependene Corporate is ritial to funtioning of finanial and produt markets. It is important for apital markets beause it redues the ost of apital through the redution of information asymmetry, and is also important for suppliers and ustomers, as parties beome more aware of the nature of ontrats they enter. Finanial markets players fae information asymmetry. Firms attempt to alleviate this asymmetry by providing information through regulated finanial statements, footnotes, management disussion and analysis, voluntary ommuniation, management foreast, analysts presentation, and intermediaries suh as finanial analysts and the press. 5 All ountries have some regulation in plae to ensure that a ertain standard of orporate is upheld, yet many firms hoose level of well above the minimum required level. Motives for exeeding minimum dislosure standards have been widely studied. The main motivation behind setting a higher level of than the required minimum is the need for external finaning. 6 However, there is a number of other important motives. Bruno and Claessens (2007) and Durnev and Kim (2005) disuss how firm governane an serve as a substitute for governane regulations. Another researh stream (Warner, Watts, and Wruk (1998), DeAngelo (1988) among others) examines how managers use orporate dislosure to redue the likelihood of undervaluation and to justify poor earnings performane. In addition, managers may have an inentive to voluntarily dislose information before trading their own shares. Trueman (1986) argues that voluntary dislosure ould be used by talented managers to signal their ability. Healy and Palepu (2001) point out that manager who are ompensated via stok options have inentives to engage in voluntary dislosure in order to inrease stok s liquidity and to redue ontrating osts assoiated with stok ompensation for new employees. 5 See Core (2001) and Verrehia (2001) for a omprehensive summary of the literature on informational asymmetry in finanial markets. 6 There are several empirial studies in support of a apital markets transation hypothesis. Lang and Lundholm (1993) observe higher analysts ratings of dislosure in the periods when firms are issuing, or are about to issue, equity. Lang and Lundholm (2000) doument a signifiant inrease in dislosure (and, in partiular, disretionary dislosure) up to six months prior to the equity issuane. Healey, Hutton, and Palepu (1999) doument a positive relationship between analysts ratings of dislosure and frequeny of subsequent debt offerings. 6

7 Transpareny matters for suppliers and ustomers as well, beause they render ontrats between the parties less subjet to adverse seletion and moral hazard problems. A large body of literature on the subjet has been developed following seminal works of Akerlof (1970), Harris and Holmstrom (1982), and Hart and Tirole (1988). Consumers and suppliers an lower searh osts and improve produt prie effiieny if ompanies are more transparent. In a model of sales with searh ost dispersion, Varian (1980) shows that improves the validity of the law of one prie. Admati and Pfleiderer (2000) onsider a model of dislosure with multiple firms whose osts are orrelated. When through regulation inreases, firms are pried more effiiently. Moreover, improved market intensifies ompetition (Stigler (1964)) whih an improve investment deisions. Given a wide variety of fators that influene the desired level of orporate and dislosure quality, one would expet that some firms and industries are more dependent on than others. These ould be firms that antiipate future market transations, have a large share of exeutive ompensation through stok options, or have poor earnings to explain away. B. Transpareny and Investment Effiieny Reent literature indiates that improves apital alloation. Wurgler (2000) measures investment effiieny as the elastiity of investment with respet to value-added and links it to apital markets development and markets informativeness. Bek and Levine (2002) use a similar measure to evaluate the merits of market-based system versus a bank-based system. Chen, Goldstein, and Jiang (2007) apply investment sensitivity with respet to investment opportunities to measure investors pereption of apital alloation effiieny. They doument that firms whose pries inorporate more firm-speifi information have larger sensitivities. Durnev, Mork, and Yeung (2004), expliitly link Tobin s marginal Qs (as a measure of the quality of apital budgeting deision) to informativeness of stok pries. They show that industries with more firm-speifi information alloate apital better. Verdi (2006) explores the link between finanial reporting quality and investment effiieny. The author argues that the higher quality of finanial reporting an improve investment effiieny through the redution in information asymmetry between the firm and investors (thus lowering the ost of apital) and through reduing information asymmetry between investors and managers (thus lowering the monitoring osts for shareholders). The author laims that there are at least two determinants of investment effiieny. First, apital must be raised to finane future 7

8 investments. Seond, there is no guarantee that the orret investment projets will be undertaken. Therefore, information asymmetry an distort investment effiieny through both the ost of apital and projet seletion. 7 C. The Role of Property Rights Protetion The effet of government intervention on the quality of information, finaning deisions, and governane strutures has reently reeived onsiderable attention. The general notion is that ompanies worried about government intervention manipulate aounting numbers and dislose less information. Watts and Zimmerman (1978 and 1986) put forward a positive aounting theory whih studies management s motives for making aounting hoies and voluntary dislosures when managers are onerned with attrating impliit or expliit taxes or regulatory ations. This positive aounting theory has been onfirmed by Han and Wang (1998) and Hall and Stammerjohan (1997). These papers show that oil produers and oil refiners try to underreport profitability to derease tax liability. This is beause oil refiners were under more politial pressure during the period of high oil pries. Leuz and Oberholzer-Gee (2006) study the role of politial ties for firms' finaning strategies and their long-run finanial performane. They find that firms with politial onnetions are less likely to rely on publily traded seurities. Stulz (2005) analyses the twin-ageny problem under whih managers have stronger inentives to divert orporate resoures when the risk of government expropriation is high. Durnev and Guriev (2008) expliitly analyze the orporate hannel through whih predatory governments an slow the growth of natural resoure-oriented industries. The authors stipulate that in the period when oil pries are high, the risk of government intervention is high, so firms in these industries will lower their levels, whih, in turn, will result in less effiient apital alloation. 8 Durnev and Guriev (2008) base their analysis only on firms from oil industry, while we onsider of all industries. 7 The study extends the work of Wang (2003) who demonstrates a positive relationship between apital alloation effiieny and three earnings attributes (value-relevane, persistene, and preision). Botosan (1997) also finds that greater dislosure is assoiated with lower ost of apital for firms with low analyst following. In addition, there are some earlier studies on the issue. Myers and Majluf (1984) build a model where information asymmetry between a firm and investors leads to underinvestment. Berle and Means (1932), Jensen and Mekling (1976) point out that information asymmetry an prevent effiient investment sine managers maximize their personal welfare, and an hoose projets that are not in shareholders best interest. 8 Additional studies in the field inlude Goriaev and Sonin (2006) and Kolotilin (2007). A number of studies doument that profitable firms limit the amount of information they dislose, or even report false information (Shipper (1989), Shivakumar (2000), and Chaney, Faio and Parsley (2007)). 8

9 Durnev and Fauver (2008) build a theoretial model (whih is onfirmed by empirial observations), in whih owners have lower inentives to enourage value-maximizing behavior by managers if the probability of government expropriation is high. The logi being that it is the more diffiult to expropriate profits that have already been stolen and/or hidden by managers. 9 Whereas Durnev and Fauver (2008) only doument the governane and dislosure response, our paper douments how sub-optimal (aused by risk of expropriation) affets investment effiieny and orporate growth. Caprio, Faio, and MConnell (2008) demonstrate that orporate holdings of liquid assets are lower in more orrupt ountries. In other words, firms hold more wealth in assets that are harder to expropriate. Therefore, suh asset sheltering represents suboptimal asset alloation, thus resulting in slower growth. Their work builds on earlier study by Myers and Rajan (1998), who observe that assets suh as ash are easier to expropriate than fixed assets. In Stulz s (2005) model, owners struture firms asset holdings in order to redue the likelihood that the state will expropriate them in ountries where the likelihood of suh expropriation is high. 10 Bushman, Piotroski, and Smith (2004) doument that governane is primarily related to a ountry's legal regime, whereas the finanial is primarily related to politial eonomy. Government predation and the risk of expropriation of orporate assets has been subjet to numerous studies. Rose-Akerman (1975) analyzes situations in whih politiians extrat bribes from firms that seek government ontrats. More reent studies on the effets of bribery on firms profitability inlude Svensson (2003) and Clarke and Xu (2003). A number of papers examine the reasons why private businesses move underground, and onsider the effets of the presene of organized rime (Alexeev, Janeba, and Osborne (2004)), orruption (Johnson, Kaufmann, and Zoido-Lobaton (1998)), and disretionary taxation poliies (Friedman et al. (2000)) Some additional studies in this field inlude Desai and Dharmapala (2004), and Desai, Dyk and Zingales (2006). 10 There is extensive literature on firm ash management that ould be traed bak to Miller and Orr (1966). Some theoretial studies on the subjet inlude Eppen and Fama (1968, 1969), Constantinides (1976, 1978), Myers (1977), and Kim, Mauer and Sherman (1998). Reent empirial studies inlude Opler et. al (1999) and Dittmar, Mahrt- Smith and Servaes (2003). 11 An inomplete list of papers in the field of firms interation with the state in the ondition of orruption inludes Ades and Di Tella (1998), Bradhan (1997), Graf Lambsdorff (2006), Hellman, Jones and Kaufmann (2000), Friedman et. al. (2000), and Johnson et al. (2000). 9

10 D. Investment Effiieny and Growth Impliations Our main assertion is that ompanies derease when the risk of government expropriation is high. Sine some industries require greater than others, these industries are more likely to be affeted by government predation. When the risk of government intervention is present, -dependent industries redue disproportionately, resulting in sub-optimal apital alloation (lower investment effiieny) and slower growth. Therefore, we hypothesize that firms in ountries with less seure property rights and in industries with high target levels of exhibit worse investment effiieny and slower growth than firms in industries whih operate well with less. Our hypothesis rests on an assumption that firms will not be able to ompensate redution through other hannels, suh as alternative orporate governane mehanisms. Durnev and Fauver (2008) demonstrate that in predatory environment firms sale down on the overall level of orporate governane. 12 We also onfirm empirially that -dependent industries indeed sale down on orporate governane in ountries with inseure property rights. III. Methodology and Data A. Empirial Speifiations and Sample Our main analysis is based on international industry data, as some variables annot be defined on firm level. Our regressions are similar to those used in Rajan and Zingales (1998), with some modifiations. The base regressions inlude the interation terms between industry measures, property rights protetion, ontrol variables, as well as ountry and industry fixed effets. Essentially, we run differene-in-differene regressions and examine whether more -dependent industries have lower investment effiieny and grow slower in ountries with less seure property rights. The advantage of this setting is that we ompare industries within the same ountries, thus alleviating the problems of endogeneity due to omitted variables or reverse ausality. Speifially, it is not very plausible that within-ountry industry differenes in growth affet ountry s level of property rights protetion. With regard to omitted industry and ountry fators, ountry and industry fixed effets, albeit imperfetly, 12 More speifially, on page 1614 Stulz (2005) writes, Greater and boards dominated by outside diretors are often viewed as hallmarks of good governane. When there are signifiant risks of expropriation by the state, neither of these two good-governane attributes are likely to enhane the wealth of shareholders. While inreases firm value, in that it makes it harder for insiders to expropriate from investors, it also dereases firm value beause it makes expropriation by the state easier 10

11 aount for them. Nevertheless, later in the draft, we provide a battery of additional robustness heks to further address endogeneity onerns. We start with a Rajan and Zingales (1998) benhmark regression, ( INVESTMENT_ EFF CONTROLS j + ε j j or GROWTH j ) = α + δ + βtrans _ NORM j j RISK +, (1) where j indexes industries and indexes ountries. In (1), α i and δ are industry and ountry fixed effets, whih ontrol for ountry and industry unobserved fators, respetively. The standard errors in the above regression are lustered by ountry to aount for the errors orrelation within ountries. Clustered standard errors are also robust to heteroskedastiity (Petersen (2008)). In (1), the dependent variable is either industry investment effiieny (INVETMENT_EFF) or industry growth (GROWH) alulated over time period from 2000 through The independent variables inlude interation terms of the need for measures (TRANS_NORM) with the risk of expropriation (RISK). In order to redue endogeneity, and ontrol variables (CONTROLS) are measured over preeding period, After ontrolling for fixed effets, the oeffiient of interest ( β ) measures the inremental inrease in investment effiieny or growth given a unit inrease in the need for, onditional on the risk of expropriation. 13 Following Rajan and Zingales (1998), the need for is alulated using U.S. industry data. This is a strong assumption whih may bias our results. Relying on U.S. data to rank industries in terms of their needs suffers from multiple problems. First, this method assumes that similar industries in different ountries have the same or at least a similar need for. Seond, same industries in different ountries fae different degrees of expropriation risk. The following example illustrates this point. Consider two industries, agriulture and petroleum, in two ountries with weak property rights protetion, Russia and Zimbabwe. While the need for (measured using U.S. data) an be higher for the petroleum industry, the risk of expropriation is higher for the agriulture industry in Zimbabwe 13 We do not inlude the norm level in regression (1) beause, for the same industry, it does not vary aross ountries. 11

12 and for the petroleum industry in Russia. Therefore, using the norm based on U.S. data would not reflet how industries reat to the risk of expropriation. We deal with the first onern (industries may have different target levels of aross ountries) by re-defining the norm measure using data from other developed ountries, namely Germany, Canada, and the U.K. To address the seond onern (same industries may have different degrees of expropriation risk aross ountries) we introdue a new gap methodology. We define gap as the differene between norm (desired level of measured using U.S. data) and pratie (observed level of alulated using loal markets data). Returning to our example, if agriulture industry is more sensitive to expropriation in Zimbabwe, firms in that industry are more likely to hide profits. Therefore, while the norm is higher for the petroleum industry for every ountry, the gap (differene better norm and pratie) would be larger for the agriulture industry in Zimbabwe. We make two preditions with respet to the gap. First, we expet that industries with larger gap invest apital less effiiently and grow slower. Seond, the negative impat of the gap on investment effiieny is likely to be stronger in ountries with greater expropriation risk. This is beause under a greater risk of government intervention, -dependent industries disproportionately redue leading to higher gap. For these tests, we estimate ( INVESTMENT_ EFF γ * TRANS _ GAP j j or GROWTH + CONTROLS j j + ε ) = α + δ + βtrans _ GAP j j j RISK +, (2) where TRANS_GAP is the differene between norm and pratie. We predit that oeffiients on TRANS_GAP and interation term TRANS_GAP RISK will be negative, that is, β < 0 and γ < 0. Next, we disuss the hoie of ontrol variables. Interation of industry external finaning need with ountry finanial development: This is our main ontrol, aimed at taking out the effet of external finaning (doumented in Rajan and Zingales (1998)) on investment effiieny investment and growth. 12

13 Interation of intangibles intensity with ountry expropriation risk: This ontrol is inluded beause Claessens and Laeven (2003) show that in ountries with more seure property rights, intangibles-intensive industries growth faster. Interation of industry norm with GDP per apita: We are onerned that ountry expropriation risk is highly (negatively) orrelated with the level of eonomi development. Our results remain robust if we use GDP growth instead of the level of GDP. Interation of industry norm with ountry finanial development: More dependent industries presumably have easier aess to external finaning in ountries with larger finanial markets. Interation of industry norm with trade openness: Even under the risk of expropriation, more dependent industries an tap international apital markets to raise external finaning. Industry fration: We use this ontrol when the dependent variable is industry growth, in order to aount for initial growth ondition. Larger industries are expeted to have lower growth rates. Our sample omes from multiple firm and ountry level data sets. For U.S., we use COMPUSTAT and CRSP. International firm data ome from Worldsope, OSIRIS, and loal stok exhanges. Our objetive is to over as many ountries as possible. Worldsope ontain aounting and market data for about 30,000 firms from 50 ountries with ative stok markets. We supplement it with a newly released firm data from OSIRIS. OSIRIS overs a larger set of publily traded ompanies (around 55,000) from 70 ountries (see Caprio, Faio, and MConnell (2008) for details). 14 For some ountries (e.g., Saudi Arabia, Jordan, Oman), we hand-ollet firm data using information from loal stok exhanges. After data leaning, our final sample onsists of 42,438 firms from 69 ountries. Firm level data is aggregated to 2-digit SIC industries. The final sample onsists of 59 industries from 69 ountries. 14 When a firm is present in both OSIRIS and Wordsope and there is a disrepany in aounting numbers (less than 1% of the sample of firms), we rely on a better known data set, Wordsope. 13

14 B. Transpareny Measures The norm measures are alulated using a sample of U.S. firms in COMPUSTAT and CRSP tapes. The pratie measures are based on the sample of international firms from Worldsope, OSIRIS, and hand-olleted data from loal stok exhanges. All firm data are aggregated to 2-digit SIC industries. The urrent measures are alulated during the period of (note that it is lagged with respet to investment effiieny measures and growth whih relate to ). 15 One onern is that the rankings of industries hange from one period to another. Our results are robust to alternative time periods for measures, namely or We also show that is not a peuliarity of U.S. markets by repliating measures using data from U.K., German, and Canadian data, other developed markets. The rank order orrelation between the measures alulated with U.S. U.K., German, and Canadian data range from to (p-value of 0.00), that is, transparent industries in the U.S. are likely to be transparent in other developed markets. We onsider four types of : informational, aounting, insider, and diret dislosure. For informational, we use a measure of firm-speifi information in stok pries alulated as the degree of stok prie asynhroniity (Mork, Yeung, and Yu, 2000). Intuitively, if a firm s stok return is highly orrelated with the market and industry fators then the stok return is less likely to ontain firm-speifi information. On the other hand, if the stok return moves asynhronously with the market return, it is indiative of more firmspeifi information being imputed into stok pries. We onstrut it for every firm i in CRSP by running: r i, t = α i + β, i rind, t + β 2, irm, t + ε i, t 1, (3) where r i, t is firm i s weekly return, r, is a 2-digit SIC industry value-weighted return, and ind t is a market value-weighted return. 16 The above regression is run using data. Informational is then defined as the logarithmi transformation of one minus the r m, t 15 We exlude 1998 beause of the Asian finanial rises whih an ontaminate our measures. 16 Industry and market indexes exlude the firm in question to avoid spurious orrelation between individual returns and indies for markets with few firms. 14

15 2 oeffiient of determination of the above regression, ln( (1 R 2 i ) / R ) i. Low values of informational mean that individual stok returns move more synhronously with industry and market indexes, refleting less firm speifi information affeting stok pries. Firm observations are aggregated to 2-digit SIC ode industries by taking industry medians. The seond measure, aounting, is based on the idea that a firm s stok return inorporates information about future earnings (Collins et. al (1994), Lundholm and Myers (2002), and Durnev, et al. (2003)). It is defined as the magnitude of oeffiients on future hanges in earnings in the regression: 3 3 r i, t = i + b0 Ei, t + bτ Ei, t+ τ + τ ri, t+ τ + ui, t τ = 1 τ = 1 α. (4) In (4), r i, t is firm i s annual return and E t+τ is the earnings per share hange τ years ahead (τ=1, 2, 3), saled by the prie at the beginning of the urrent quarter. Future returns ( r i, t+ τ ) are used to mitigate errors-in-variables problem in measuring expeted returns. The above regression is run during the time period. Aounting is then the sum of the oeffiients on future hanges in earnings, b 1 + b2 + b3. All our results are robust if we define informativeness to be an inrease in the oeffiient of determination of regression (4) relative to a benhmark regression whih inludes only ontemporaneous hange in earnings. All else equal, the greater the assoiation between urrent stok returns and future earnings, the more informative the urrent stok pries are, the result of higher aounting. The third measure, insider, is based on stok return autoorrelation onditional on trading volume, and it reflets the degree of informational asymmetry assoiated with a ompany. Llorente et al. (2002) show that higher information asymmetry between different groups of traders is likely to result in returns being positively autoorrelated (onditional on trading volume). Insider is the oeffiient C 2 (multiplied by negative one) in the time-series regression: R i, t+ = Ai + C1, iri, t + C2, iri, tvi, t + λi, t 1, (5) 15

16 run for eah firm i with weekly returns (R i,t ) and trading volume data (V i,t ) from 1995 through Trading volume is alulated as de-trended volume: VOL VOL i, t 20 i, t j V =, log = log, (6) i t j 1 N i, t 20 Ni, t j 1 where VOL is the number of shares traded, and N is the number of shares outstanding. We aggregate this variable by taking 2-digit SIC ode industry medians. Finally, we use Standard & Poor s dislosure sores as a measure of. S&P onduted a survey of 1,600 ompanies around the globe onerning firms' and dislosure. These ompanies omprise one of S&P s global indexes. Transpareny and dislosure are evaluated by searhing for the inlusion of 91 possible information items. These 91 items were seleted after examining the annual reports and other aounts of leading ompanies around the world and identifying the most ommon dislosure items. The inlusion of eah item is sored on a binary basis ( yes denotes inluded and no denotes not inluded) to ensure objetivity. Eah yes answer is equal to one point. These items are then grouped into three sub-ategories: (i) ownership struture and investor relations (22 items); (ii) finanial and information dislosure (34 items); and (iii) board and management struture and proess (35 items). We define an aggregate dislosure index as the sum of these three ategories. The index ranges from 0 to 91 with a higher sore representing more and dislosure. The sample inludes 1,494 firms from 40 ountries. This measure is alulated as the industry average over available years, The four finanial measures (informational, aounting, insider, and dislosure) are aggregated into a single index using Prinipal Component Analysis (PCA). We all this index, aggregate finanial. 17 The first prinipal omponent aptures 63% of the orresponding ross-setional variane of the three variables above. Moreover, only the first eigenvalue is signifiantly larger than one; thus one fator is suffiient to apture muh of the ommon variation among the three variables. The loadings for the aggregate finanial index (based on the PCA) are: 0.68 for the aounting, 0.62 for the informational, 0.08 for the insider 17 PCA is a statistial method to redue multidimensional data sets to lower dimensions. The PCA an be viewed as an orthogonal linear transformation that alters the data to a new oordinate system suh that the greatest variane by any projetion of the data omes to lie on the first oordinate (alled the first prinipal omponent), the seond greatest variane on the seond oordinate, and so on. See Stevens (1986) for details. 16

17 , and 0.39 for dislosure. All loadings are positive meaning that the four proxies of apture albeit eah loading apturing different omponents. When the above measures are alulated using U.S. data, we all them norms. We assume that norm proxies for the desired level of in an industry. When they are alulated using loal data, we all them praties. In most of the analysis, we employ gap measures. Transpareny gaps are omputed by subtrating praties from norms. Thus, higher values of gaps mean that industries exhibit levels lower than desired. Our arguments rest on the assumptions that governments annot unover firms aounting numbers. While it is plausible that the government an expose a part of firms profits, it is unlikely they an do it perfetly. A large part of orporate governane and earnings management literature assumes that neither investors nor regulators an perfetly reveal firms true performane. Moreover, if states an always find out whih ompanies hide their profits, an equilibrium response for firms would be to always reveal performane truthfully (see Watts and Zimmerman (1986) for a more detailed disussion). C. Investment Effiieny and Industrial Growth Aording to Wurgler (2000), effiient apital alloation involves inrease in investment in growing industries and derease in investment in delining industries. He measures investment effiieny as the elastiity of investment with respet to value added. We define investment effiieny as the ountry-speifi, industry-speifi elastiity ( Ω ) of investment (measured by apital expenditures) (I) with respet to firm value (Q), for whih we estimate a panel regression (with firm and year fixed effets), j I ln I i, t i, t 1 = firm effets + time effets + Ω j Q ln Q i, t i, t 1 + ϕ i, i (7) for every industry j and ountry using all firm annual data from 2000 through A similar measure was developed by Wurgler (2000) and was also used in Bushman, Piotroski, and Smith (2004). Firm value is defined as market apitalization plus total assets minus book equity divided by total assets. Our results are robust if we use sales instead of firm value in (7). Industry growth 17

18 (GROWTH) is measured as the growth in real sales alulated from 2000 through Holding everything else equal, larger values of Ω j mean better investment effiieny. D. Property Rights Protetion We use two alternative variables to measure the degree to whih private property is proteted. The first is based on a ountry s assessment of expropriation risk (expropriation risk). The seond measure, politial risk, assesses onstraints imposed on major politial players that would limit government interferene into firms affairs. Expropriation risk is extrated from the International Country Risk Guide (ICRG). The variable we use is ICRG Investment Profile whih is desribed as the assessment of risk of investment due to ontrat viability/expropriation and profits repatriation. We use this index beause it is available for every ountry in the sample. The index ranges from 0 (high expropriation risk) to 12 (low expropriation risk). We hange it (subtrat the index values from 12) so that larger numbers orrespond to larger risk of expropriation. We take quarterly averages from 1995 through All our results do not hange if we use other indexes. For example, we also use expropriation risk and risk of ontrat repudiation from earlier data in the ICRG (ICRG hanged its methodology in 1995) and desribed in Knak and Keefer (1999). We do not use it in the main analysis beause the latest year these indexes are available for is The seond measure of the risk of government interferene is based on Henisz (2000 and 2002) index of politial onstrains (what he alls POLCON_V). POLCON_V is a omprehensive measure of politial onstraints within a ountry whih aims to measure the redibility of poliy ommitment. First, it identifies the number of independent branhes of government (exeutive, lower and upper legislative hambers, judiiary and sub-federal politial entities) with veto power over poliy hange. The preferenes of eah of these branhes and the status quo poliy are then assumed to be independently and identially drawn from a uniform, one-dimensional poliy spae. This initial measure is then modified to take into aount the extent of alignment aross branhes of government using data on the party omposition of the government branhes, as the alignment inreases poliy hange probability. The measure is then modified to aount for preferene heterogeneity within eah legislative branh whih inreases deision osts of overturning poliy for aligned exeutive branhes. The original series ranges from 0 to 1 with larger numbers indiating fewer politial onstraints. We again modify the original series by 18

19 subtrating it from 1 so that larger values of the index orrespond to fewer onstraints, and, therefore, larger politial risk. Therefore, higher values of politial risk orrespond to governments that an easily hange the ``rules of the game`` and deviate from property rights protetion. As a robustness hek, we use the autoray index as an additional proxy for the risk of government intervention. In pursuit of self-enrihment, autorati rulers are more likely to set up extortion regimes and are less subjet to heks and balanes from demorati institutions. The index is alulated by subtrating POLITY s demorati government index from its autorati government index. The autorati (demorati) government measures general losedness (openness) of politial institutions. The two variables assess a number of fators, suh as (i) ompetitiveness of politial partiipation; (ii) regulation of partiipation; (iii) the openness and ompetitiveness of exeutive reruitment; and (iv) onstraints on the hief exeutive. E. Other Measures Regressions (1) and (2) use the following ontrols: interations of external finaning needs with finanial development, intangibles intensity with expropriation risk, with investor protetion, with GDP per apita, with finanial development, and with trade openness. We desribe their onstrution in this setion. External finaning needs: industry median value of apital expenditures minus ash flows from operations divided by apital expenditures. This variable is onstruted using COMPUSTAT. Finanial development: the sum of stok market apitalization and private redit relative to GDP. This variable is onstruted using the World Bank s World Development Indiators. Intangibles intensity: the ratio of R&D and Advertising spendings to net property, plant, and equipment. This variable is onstruted using COMPUSTAT. Investor protetion: First, we use the anti-self dealing index from Djankov, La Porta, and Shleifer (2006). This index is an aggregate measure of legal rules and private enforement mehanisms, suh as dislosure, approval, and litigation, governing a speifi self-dealing transation based on ex-ante and ex-post ontrol of self-dealing. As a robustness hek, we also inlude an updated investor protetion index from Djankov, La Porta, and Shleifer (2006) and a 19

20 de fato measure of law enforement, ICRG s rule of law. The rule of law variable is a quantitative assessment of the strength of a ountry s tradition of law and order. GDP per apita: Real dollar GDP per apita. This variable is onstruted using the World Bank s World Development Indiators. Trade openness: Sum of exports and imports over GDP. This variable is onstruted using the World Bank s World Development Indiators. Industry fration: Ratio of industry s total assets to GDP. For some of the tests, we use the measure of earnings management (earnings quality). The measure is based on the quality of earnings reported in firms finanial statements. Following Dehow, Sloan, and Sweeney (1995) and Chaney, Faio, and Parsley (2007) and measure firm earnings management as a deviation of reported aruals from a benhmark of aounting aruals. We use a ountry benhmark as in Chaney, Faio, and Parsley (2007) and estimate a panel time-series, ross-ountry regression using data, TCA A i, t Sales i, t PP & E i, t = + β + D j J j + [2000,2005] i, t Ai, t Ai, t α D η τ τ +, (8) where is the differene operator, indexes ountries, i indexes firms, and t indexes years. Total urrent aruals, TCA, are defined as (Current Assets) (Current Liabilities) (Cash) + (Short-term and Current Long-term Debt); A is total assets, Sales is total sales, PP&E is the sum of net property, plant, and equipment, and aumulated reserves for depreiation, depletion and amortization. D j are two-digit SIC industry fixed effets and D j are year fixed effets. All variables are expressed in U.S. dollars. 18 The earnings management for firm i in ountry j is defined as the standard deviation of the error term of the above regression alulated over We assign a 2-digit SIC industry ode to every ompany and take industry medians for industry equivalents of the firm-level measures. i, t 18 The results remain robust if we ontrol for past performane (measured as ROA per total assets) in (8) as suggested by Ashbaugh, LaFond, and Mayhew (2005). 20

21 IV. Results A. Summary Statistis This setion desribes our estimation results. Table I presents the desriptive statistis grouped by industry. It is important to note that two types of measures are presented in the table. Norm measures (aounting norm, information norm, insider norm, dislosure norm, and aggregate norm) are omputed based on the sample of U.S. firms and represent industries target measures. Earnings quality, investment effiieny, and industry growth are omputed using a sample of international firms and represent observed numbers aross all ountries. With regard to aounting norm, Eletri, Gas, and Sanitary Servies (SIC 4900) has the lowest sore (0.233), whereas Printing, Publishing, and Allied Industries (SIC 2700) has the highest sore of It also has the highest sore of in informational norm (Petroleum Refining, SIC 2900 sores lowest in that ategory). The highest insider norm is displayed by Apparel and Aessory Stores (SIC 5600), the lowest one is displayed by Railroad Transportation (SIC 4000). Apparel and Other Finished Produts (SIC 2300) have the highest dislosure norm (Water Transportation (SIC 4400) has the lowest). Personal Servies (SIC 7200) have lowest aggregate norm, Printing, Publishing, and Allied Industries have the highest one. Water Transportation (SIC 4400) has the lowest earnings quality, Metal Mining (SIC 1000) has the highest. Railroad Transportation (SIC 4000) has the worst investment effiieny and Textile Mill Produts (SIC 2200) has the lowest growth. Building Constrution (SIC 1500) sores the highest in both investment effiieny and growth. Table II presents desriptive statistis grouped by ountry. Pakistan exhibits the worst investment effiieny, and Zambia has the lowest average industry growth. Sweden has the best investment effiieny, while Ireland exhibits the highest average industry growth. Russia exhibits the worst aggregate, whereas Sweden has the best. As for the overall politial risk, Belgium`s politial risk is the lowest, while ountries like Oman, Saudi Arabia, and China have high politial risk. 19 Table III presents orrelations among various omponents of the norm measures (based on U.S. data). Even though not all measures are signifiantly 19 Is finanial development just a proxy for the level of property rights protetion? While the orrelation between finanial development and expropriation risk is negative (-0.43), there are some exeptions. Several ountries (e.g. South Afria, Malaysia, Jordan) have well developed finanial markets but weak property rights. On the other hand, ountries like Finland, Denmark, and New Zealand respet property rights but have less developed finanial markets. 21