Investor protection and corporate governance in family firms: Evidence from China

Transcription

1 Invesor proecion and corporae governance in family firms: Evidence from China Absrac: This paper examines he impac of invesor proecion on he governance arrangemens of family businesses by using a unique and deailed family-firm daa se in China s sock markes for he period Our findings show ha provincial variaions in he qualiy of law enforcemen in China shapes he differen forms of family governance arrangemens in family firms. We find ha family firms wih srong legal proecion of shareholders have more concenraed ownership and greaer paricipaion of family managemen, wih less conrol-enhancing machanisms. Moreover, such relaions are more pronounced in family firms hiring an ousider-ceo han in family firms managed by a family member. This paper conribues o our undersanding of he relaions beween invesor proecion and ownership concenraion and beween invesor proecion and he preservaion of family managemen in China s conex for he firs ime and also sheds lighs on he relaven sudies in oher emerging counries, especially hose sharing he similar characerisics wih China in family governance and legal developmen. Keywords: invesor proecion, corporae governance, family firms, China 1 Inroducion Emerging economies are ypically characerised by an absence of effecive formal insiuions for invesor proecion, resuling in weak governance environmen (Mion, 2002; Allen e al., 2005). To survive in he adverse poliical and insiuional environmen, family firms locaing in differen regions end o seek alernaive ways o safeguard heir ineress in he absence of formal insiuional proecion, one of which is presumed o be family governance (La Pora e al., 2002; Burkar e al., 2003; Lins, 2003). The law and finance lieraure has provided evidence ha legal proecion can help o shape he inernal governance mechanisms in family firms, such as ownership concenraion and managemen preservaion. Some sudies sugges ha ownership srucure is negaively associaed wih he qualiy of legal proecion for invesors (La Pora e al., 1999; Shleifer and Wolfenzon, 2002; Wu e al., 2009), showing ha ownership concenraion and invesor proecion are subsiues. Burkar, Panunzi and Shleifer (2003) examine he relaionship beween he qualiy of invesor proecion and he decision o 1

2 keep conrol including ownership righs and managerial power in he family. Their findings show a negaive relaionship beween invesor proecion and ownership concenraion, as well as beween invesor proecion and he preservaion of family managemen. Oher sudies find an opposie resul, suggesing ha srong legal proecion of shareholder righs can predic higher shares holding by conrolling shareholders (Casillo and Skaperdas, 2005). Burkar and Panunzi (2006) sae ha he relaions beween legal proecion and ownership concenraion is no necessarily uniform bu depends on how legal proecion ineracs wih monioring. If legal rules peraining o he righs of shareholders are subsiues for monioring, in which he large shareholder s monioring can counerac he manager s exracion, weaker legal proecion would allow large shareholders o reduce heir shares in order o resore he manager s incenives. In his case, legal proecion is considered as a complemen o ownership concenraion and he relaionship beween he wo is posiive. Aganin and Volpin (2003) empirically find similar evidence of a posiive relaion beween legal proecion and ownership concenraion in Ialian corporaions during he period The above sudies sugges ha legal proecion of shareholders affecs organizaional rouines (Boyer and Hollingsworh, 1997) and helps o deermine he sraegic choices facing organizaions in emerging economies (Peng e al., 2003; Young e al., 2008). Since ownership, conrol, and managemen srucure are generally considered o be he aribues of family firms (Villalonga and Ami, 2006; Mroczkowski and Tanewsk 2007), such view raises some quesions of wheher legal proecion for invesors is expeced o shape he conrolling family s decisions on he hree elemens of inernal corporae governance mechanisms in emerging economies. And if so, dose he relaionship beween invesor proecion and family governance end o be posiive or negaive? In his paper, we aemp o address hese quesions by examining he impac of invesor proecion on family governance arrangemens including family ownership, conrol, and managemen of family firms lised in China s sock exchanges during he period China provides an ideal case for his analysis for hree reasons. Firs, China s evoluion of legal rules proecive of invesors follows a similar paern o oher emerging counries. Since he 1990s, China and oher emerging counries, such as Easern Europe and he former Sovie republics, have gone hrough a fundamenal ransiion oward marke-based economies from cenral planning sysems, leading o he similar insiuional evoluion of legal sysems (Peng and Heah, 1996). Thus, China s experience 2

3 may provide lessons and insighs on he way family business develops in oher emerging counries. Second, he governance environmen in China is o a large exen capured by he ineracions beween he local governmens and naional legal insiuions (Xia and Fang, 2005). 1 Therefore, alhough wih he same regulaion of he wrien legal rules a he naional level, family firms in individual regions are exposed o differen degrees in he effeciveness of legal enforcemen and informal alernaive mechanisms in China (Chen e al., 2005; Fan e al., 2007; Wu e al., 2009). Compared o convenional sudies which end o creae omied-variables and aggregaion biases due o he use of cross-counry samples, invesigaing regional (provincial) variaions in he qualiy of law enforcemen wihin one counry is considered more fruiful (Wu e al., 2009). Third, he increasingly imporan role played by Chinese family firms in he economy provides laboraory-like environmen o examine family-firm issues in China s capial marke developmen (Zhou e al., 2010). Addiionally, family managemen is cerainly prevalen among Chinese family firms. The daa from our sample firms show ha family firms in which family members eiher ac as direcors or managers accoun for 78% of all lised family firms and 65% are direcly managed by family members aking in he posiion of Chairman or CEO. The prevalence of family managemen in China herefore allows us o ake an in-deph comparison of he role of family Vs professional managers on family firms. This paper conribues o his growing lieraure in he following ways. We make wo mehodological innovaions. Firs, we improve upon he relaively crude measures of family firms in he lieraure. The convenional measures of Chinese family firms usually defined as consising of a crude classificaion of he non-sae-owned firms (e.g. Chen e al., 2005), or an approach ha does no ake accoun of he muliple circumsances in which he company is conrolled by more han one individual or family (e.g. Su and Zhu, 2003). In his paper, we exend he convenional approaches by proposing finer-grained crieria for he definiion of family firms, hereby lowering he risk ha nonfamily firm feaures may be inclined o disor he resuls. Second, we use an upper quarile score as a breakpoin of he level of invesor proecion, which ends o be a more accurae indicaor of he higher qualiy of legal 1 The effeciveness of law enforcemen is a par of invesor proecion iniially measured by La Pora e al. (1998). In his paper, he level of law enforcemen is defined by a broader concep - he qualiy of governance environmen in which a lised company has is base o operae. The exac definiion of governance environmen will be provided in Secion

4 proecion of shareholder righs han he median score commonly used in convenional sudies. This approach is expeced o miigae he risk ha he relaively crude crierion for invesor proecion migh disor he resuls. Based on hese mehodological innovaions and our empirical analysis, we find ha family governance arrangemens are o some exen shaped by legal proecion for shareholders wihin differen local jurisdicion in China. To illusrae, he sudy shows ha family firms operaing in a more proecive governance environmen will have more concenraed ownership righs, more managemen posiions occupied by family direcors or managers, and less use of conrol-enhancing insrumens such as pyramids. The posiive relaionship beween ownership concenraion and he qualiy of invesor proecion is consisen wih Aganin and Volpin (2003) and Casillo and Skaperdas (2005), bu provides opposie evidence o he commonly acceped view in he lieraure ha ownership concenraion is a subsiue for invesor proecion. We also invesigaes he ineracion effecs of family managers and invesor proecion on family governance mechanisms. Our analysis finds ha companies having a family CEO in place have more significan ownership righs and higher level of family managemen, wih less use of conrol-enhancing mechanisms han family firms led by a hired CEO. However, he posiive relaions beween invesor proecion and family ownership, and beween invesor proecion and family managemen, as well as he negaive relaion beween invesor proecion and family conrol we have found are more significan in family firms having an ousider as CEO. One possible explanaion is ha due o he lower level of conflics beween he family owner and family managers, he family ends o coun on more inernal corporae governance arrangemens raher han o depend on he qualiy of exernal invesor proecion o make sraegic and operaing decisions. Our findings provide new evidence on he posiive relaions beween invesor proecion and ownership concenraion, and beween invesor proecion and he preservaion of family managemen in China. I also sheds lighs on he relevan sudies in oher emerging economies, especially hose sharing similar characerisics wih China in erms of family governance and legal developmen. The reminder of his paper is organised as follows. Secion 2 lays ou he research design, including sample selecion, daa sources, and panel-daa models and variables. In Secion 3 we repor he saisics summary. Secion 4 discusses he main resuls, and is followed by he robusness check presened in Secion 5. Conclusion is provided in Secion 6. 4

5 2 Research design 2.1 Sample and daa Our analysis consiss of a panel of 13,365 firm-year observaions from 1,624 nonfinancial companies lised in he Shanghai and Shenzhen Sock Exchange in China during he period Banks and insurance companies are excluded due o he difficuly in compuing agency coss (ROA). By closely examining he ownership and managemen composiion of each paricular company, he full sample is broken down ino wo subsamples, yielding 2,924 family-firm observaions and 10,441 nonfamily-firm observaions. The year of 2000 is chosen as he sar poin of ime period because he number of privae lised companies has soared since 2000 in China s sock markes. In effec, we find very few observaions before A family firm in his paper is, by definiion, one ulimaely owned by a family or an individual. The crierion o idenify he ulimae owner is based on a shareholder (he sae, an insiuion, a family or an individual) having a 5% ulimae conrol righs greaer han he second larges shareholder in a lised company. The inroducion of he Lised Companies Informaion Disclosure Regulaions issued by he China Securiies Regulaory Commission (CSRC) on 30 h Jan 2007 specifies ha shareholders having more han 5% sakes in a lised company and any ransfer of hese shareholders holdings mus be disclosed in is public announcemens or repors. In his conex, a shareholder reaining 5% sakes greaer han he ohers can hus be expeced o have maerial influence over his company. A deailed descripion of his definiion is provided in a companion paper (Zhou e al., 2010). Daa on family ownership and financial variables are mainly sourced from GTA daabase, and double-checked by annual repors, prospecus, and inerim announcemens of lised companies. All relaed public repors or announcemens are obained from wo official websies he Shanghai Sock Exchange (SSE) and he Shenzhen Sock Exchange (SZSE). 2.2 Governance environmen index We consruc an index of governance environmen (GENVIRON) as an indicaor of he qualiy and effecivenss of law enforcemen in China s provincial jurisdicions. In he lieraure, he level of law enforcemen has been ofen measured by he effeciveness of he naional governmen, credi markes and judicial sysem, ec (La Pora e al., 1998; Pisor e 5

6 al., 2000; Wang e al., 2008; Wu e al., 2009). We hus define he qualiy of law enforcemen by a broader concep - he qualiy of governance environmen in which a lised companie has is base o operae- and evaluae i by four specific measures: (1) Governmen corrupion (GOVERN); (2) Developmen of financial marke (FINANCE); (3) Developmen of marke inermediaries (INTERMEDIARY); and (4) Efficiency of he judiciary (JUDICIARY). Daa on he four measures are sourced from NERI INDEX of Markeizaion of China s Provinces 2009 Repor compiled by Fan e al. (2010). The NERI index aims o assess he markeizaion process of individual provinces in mainland of China from 1999 o 2007, 2 and has been widely applied in recen lieraure, such as Chen e al.,(2005), Li e al. (2006), Wang e al.(2008) and Wu e al. (2009). The value of each measure is compued by he hree-year average from 2005 o Appendix A.1 and A.2 presen he definiion of indicaors of GENVIRON and he deailed resuls, respecively. 2.3 Variables Dependen variables Dependen variables are capured by hree ses of variables o indicae family ownership, conrol, and managemen, respecively. Family ownership variable is measured as CASHFLOW, he shares held by he family (including all family members). The calculaion of cash-flow righs and conrol righs held by he family follows he mehod inroduced by La Pora e al.(1999) and developed by Faccio and Lang (2002) and Claessens e al.(2002). Family conrol is characerised by he presence of conrol-enhancing insrumens ha allow he excess of conrol righs over cash-flow righs (Baronini and Caprio, 2006). Hence, conrol variables in his paper conain (1) EXCESSCONTROL, he difference beween he family s conrol righs and cash-flow righs; and (2) PYRAMIDS, equalling one when he family exercises conrol over a lised company hrough a leas one oher company, and zero oherwise. Managemen variable is capured as FAMILYRATIO, he number of family members (including he founder and his or her relaives) serving as direcors or managers 4 divided 2 The sample of he NERI index excludes Hong Kong and Macau. 3 Fan e al. (2010) sae ha he aggregae score of NERI index is eviden o be lile differen beween ha compued by Principal Componen Analysis and arihmeic average. 4 In Chinese family firms, aken our sample in 2008 for example, he execuive-direcors who are acive in he posiions of firm managemen accoun for 30% of family direcors. Therefore i is difficul o separae he 6

7 by he oal number of senior managemen ha refers o direcors, supervisors, he Chief Execuive Officer (CEO), he Vice General Managers, he Assisan General Managers, he Secreary of he Board and he Chief Financial Officer (CFO). Following Gomez-Mejia e al. (2003), he founder s relaives refer o his or her faher, moher, siser, broher, son, daugher, spouse, in-laws, aun, uncle, niece, nephew, cousin. Explanaory variables The exen of invesor proecion wihin differen local regimes is measured by he governance environmen index. This index is used o examine he impac of invesor proecion on family governance in he mulivariae analyses. Addiionally, a dummy variable is applied o indicae he high or low level of invesor proecion. The dummy equals one, indicaing rich invesor proecion, if he value of governance environmen of a paricular province is above he upper quarile score (25%) across 31 provinces, and zero oherwise. This approach differs from he convenional sudies in which he median score is commonly used as a breakpoin for he degree of invesor proecion, such as Maury (2006) and Yu and Pan (2008). Conrol variables Recen sudies show ha business affiliaes susained by heir associaion wih a same family would eiher benefi (Faccio e al., 2001; Granoveer, 2005; Luo and Chung, 2005) or suffer from his inerfirm relaionship (Khanna and Palepu, 2000). We hus use SYSTEM o conrol for his poenial effec, which equals one when he firm is under he same family s conrol as a leas one oher lised firm in he same year, and zero oherwise. Oher conrol variables are specified, including SALE, he raio of annual operaing revenue o oal asses; SIZE, he naural log of annual oal asses; LEVERAGE, he raio of liabiliies o oal asses; AGE, he number of years since he iniial public offerings (IPO) of he firm. We also use indusry dummies and indusry-adjused ROA o conrol for indusry effecs, and year dummies o conrol for ime effecs. impac of family direcors from he managemen. In essence, as suggesed by Caspar e al. (2010), a srong board, in a family firm, is feaured by a deep and acive paricipaion in op-execuive managemen. 7

8 2.3 Empirical models To individually examine he impac of invesor proecion on family ownership, conrol, and managemen, we apply he following regression panel-daa models accordingly by using he above-described variables: CASHFLOW + β LEVERAGE 5 = α + β GENVIORN 1 + β AGE 6 + ε i + β SYSTEM + β SALE β SIZE 4 (1) EXCESSCONTROL + β LEVERAGE 5 = α + β GENVIORN + β AGE ε i + β SYSTEM + β SALE β SIZE 4 (2) PYRAMIDS + β LEVERAGE 5 = α + β GENVIORN 1 + β AGE 6 + ε i + β SYSTEM + β SALE β SIZE 4 (3) FAMILYRATIO + β LEVERAGE 5 = α + β GENVIORN + β AGE ε i + β SYSTEM + β SALE β SIZE 4 (4) As he main explanaory variable- GENVIRON- is ime-invarian, we use he random effecs (RE) models for he esimaes of he governance environmen variable. The RE models include indusry dummies o conrol for indusry effecs. Since he precise form of he ime effec is unknown, ime dummies are no included in he RE models here in order o miigae he disorion ha he unfixed ime effec bring o he esimaes (Peersen, 2009). Time dummies are included in regressions in he robusness es. 3 Saisics summary 3.1 Disribuion of family firms Table 1 repors he number and percenage of family firms in China s sock markes during he period In 2000, here were only 68 lised companies realised as family firms according o our definiion, represening less han 7% of lised companies in China s sock markes. By he end of 2009, he number of family firms has soared o 551, accouning for 33.93% of he full sample. Alhough he percenage of family firms is slighly lower han ha in he U.S. or German sock markes, 5 i represens an 8 imes increase since 2000, indicaing an explosive jump of he developmen of family firms in China. 5 Anderson and Reeb (2003) and Villalonga and Ami (2006) show ha family firms represen 35.0% and 38.0% of he U.S. lised firms, respecively, while Andres (2008) finds ha he share of family firms is 37.5% in German sock markes. 8

9 Table 1 Number and percenage of family firms in a Year All lised firms Family firms Nonfamily firms Family firms (% of oal) b , , , , , , , , , , , , , , , , , , , Toal 13,365 2,924 10, (average) Noes: a The full sample comprises 13,365 firm-year observaions from 1,624 companies lised in China s sock markes during Afer filering five ouliers, he sample of all firms and family firms comprises 13,360 and 2,920 observaions, respecively. As he number of ouliers is oo small o change he descripive saisics much, he unfilered resuls are presened here. b Family firms (% of oal) are compued as he number of family firms divided by he oal number of firms in each year. On closer examinaion of he disribuion of family firms in Figure 1, i is obvious ha here is a much higher incidence of family-managed firms, as opposed o ousider-managed firms in Chinese sock markes. In paricular, ousider-managed caegory includes 628 observaions, represening only 21.48% of family firms, which demonsraes ha he persisence of family managemen is more popular. In family-managed caegory, family firms in which he founder or his family occupies he posiions of Chairman or CEO are dominan: 84.63% versus 15.37% belonging o family-involved group wih family members only acing as direcors or managers. Furher, he able in Figure 1 shows ha he incidence of family firms in which he founder himself/herself severing as Chairman is almos 8 imes higher han ha of firms led by a descendan- or relaive-chairman. This finding suggess he majoriy of Chinese family firms are experiencing he firs-generaion sage of developmen and are indeed relaively young. Figure 1 Disribuion of family firms 9

10 Table 2 shows he indusry disribuion of family firms and nonfamily firms. A closer look a he CSRC indusry codes reveals ha family firms are no uniformly disribued across all indusries. A one exreme are 11 indusries in which here are no family firms. A oher exreme are oher wholesale rade (H09), real esae broker (J09), professional, scienific research and services (K20), saniaion, healh care, nursing services (K37) as well as oher communicaion and culural (L99), which are composed enirely of family firms. Family firms also end o prevail in a leas eigh indusries, such as exiles (C11), livesock (A05), oher fibre producs manufacuring (C13), insrumenaion and culure, office machinery manufacuring (C78), communicaions and equipmen manufacuring (G81), decoraion (E05), communicaion services (G85) and prining (C35). 6 This variance of disribuion highlighs he imporance of conrolling for indusry in he regression analysis. Table 2 Indusry disribuion of firms by CSRC sandard indusry classificaion code (2001) a CSRC Family Nonfamily Family firms in Indusry descripion All firms Code firms firms Indusry (%) b A01 Agriculure A03 Foresry A05 Livesock This examinaion is compued on indusries wih a leas five firms. 10

11 A07 Fishery A09 Agriculure, foresry, Livesock and fishery services B01 Coal mining B03 Oil and naural gas mining B05 Ferrous meals mining and dressing B07 Nonferrous meals mining and dressing B50 Mining Services C01 Food processing C03 Food manufacuring C05 Beverage manufacuring C11 Texiles C13 Clohing and oher fiber producs manufacuring C14 Leaher, fur, feaher and producs manufacuring C21 Wood processing and bamboo, raan, palm and grass producs C25 Furniure manufacuring C31 Paper and paper producs C35 Prining C37 Culural sporing goods manufacuring C41 Oil processing and coking C43 Chemical maerials and chemical producs manufacuring C47 Chemical fibre manufacuring C48 Rubber manufacuring C49 Plasics manufacuring C51 Elecronic componens manufacuring C55 Household elecronic appliances manufacuring C57 Oher elecronic equipmen manufacuring C61 Non-meallic mineral producs C65 Ferrous meal smeling and rolling processing C67 Non-ferrous meal smeling and rolling processing C69 Fabricaed meal producs C71 Ordinary machinery manufacuring C73 Special equipmen C75 Transpor equipmen C76 Elecrical machinery and equipmen manufacuring C78 Insrumenaion and culure, office machinery manufacuring C81 Pharmaceuical C85 Biological producs C99 Oher manufacuring D01 Elecriciy, seam, ho waer producion and supply D03 Gas producion and supply D05 Tap waer producion and supply E01 Civil engineering consrucion E05 Decoraion F01 Rail ranspor F03 Road ranspor F07 Waer ranspor F09 Air ranspor F11 Auxiliary ranspor

12 F19 Oher ranspor F21 Warehousing G81 Communicaions and equipmen manufacuring G83 Compuers and equipmen manufacuring G85 Communicaion services G87 Compuer applicaion services H01 Food, beverages, obacco and household goods wholesale rade H03 Energy, maerials and machinery and elecronics equipmen, wholesale rade H09 Oher wholesale rade H11 Reail H21 Commercial brokerage and agencies J01 Real esae developmen and managemen J05 Real esae managemen J09 Real esae broker K01 Public faciliies services K20 Professional, scienific research and services K30 Caering K32 Hoel and gueshouse accommodaion K34 Tourism K37 Saniaion, healh care, nursing services K39 Renal services K99 Oher social services L01 Publishing L10 Radio, Film and Television L20 Informaion disseminaion services L99 Oher communicaion and culural M Miscellaneous Toal 1, , Noes: a The full sample comprises 1,624 companies lised in China s sock markes in he year of b Family firms in Indusry is compued as he number of family firms divided by he oal number of firms in each indusry. 3.2 Descripive saisics of variables Table 3 repors he descripive saisics of main variables used in his paper. Panel 3A shows means, sandard deviaions, minimum and maximum values of variables of he family-firm sample excep dummy variables. Panel 3B provides he number and percenage of family firms in which he ulimae conrol involves he presence of pyramids or sysems. Panel 3C furher divides he full family-firm sample ino wo subsamples of family firms having a family CEO and family firms hiring an ousider as CEO, and presens he resuls of Independen-Sample T es of mean differences beween he wo groups. 7 7 Due o he meaningless calculaion of means of dummy variables, i is of no ineres o adop Independen- Sample T ess of mean differences for PYRAMIDS and SYSTEM beween family- and ousider-ceo family firms. 12

13 Panel 3A shows ha he conrolling family in he family-firm sample on average holds cash-flow righs of 25.5% wih a minimum value of 0.5% and a maximum value of and 87.1%. This mean value is relaively higher han in mos of Eas Asian corporaions, such as Japan, Korea, Malaysia, Philippines, Singapore and Taiwan (Claessens e al., 2000), 8 and also higher han in he U.S. companies (Anderson and Reeb, 2003). 9 The excess of conrol righs over cash-flow righs has a mean value of 10.1%. The number of family direcors or managers divided by he number of senior managemen is 8.3% on average and 46.2% a he maximum, which suppors he argumen of he prevalence of family managemen in Chinese family firms. As seen in Panel 3B, for 69.6% of he companies in he family-firm sample, he ulimae conrol involves he use of a pyramidal srucure, suggesing ha pyramiding is widely spread in he companies owned by a family or an individual in China. This proporion is he highes among all nine Eas Asian companies repored by Claessens e al.(2000). Addiionally, 14.4% of family firms are under he conrol of he same family or individual, implying ha Chinese companies exhibi a relaively lower incidence of syseming han pyramiding in he ulimae conrol srucure, and he sysem so far has no been buil up on China s sock markes. Panel 3C akes a closer look a he descripive saisics of family- and ousider-ceo firms. The analysis shows ha family firms wih a family CEO reain more concenraed cash-flow righs, 33.5% on average, significanly higher han firms hiring a professional CEO wih 22.8% on average. In erms of conrol-enhancing mechanisms, family-ceo firms have less conrol in he excess of ownership righs and also a lower incidence of pyramiding as opposed o ousider-ceo firms (7.2% versus 11.1% and 54.7% versus 74.6%, respecively) and he mean difference in he excess of conrol righs over cash-flow righs beween he wo groups of firms is saisically significan a 1% level. For family managemen, he mean percenage of family direcors or managers in firm managemen is 14.1% in family-ceo companies. This mean proporion is 7.8% larger han in comparable firms, and he mean difference is saisically significan a 1% level. To recapiulae, family- and ousider-ceo family firms differ significanly in he concenraion of cash-flow righs, he separaion of ownership and conrol, he use of 8 Claessens e al. (2000) explore he ownership srucure of corporaions in nine Eas Asian counries. In heir examinaion, he conrolling shareholders hold he smalles cash-flow righs in Japanese companies, only 6.90% on average, while Thai companies display he mos concenraed cash-flow righs wih 32.84% among he nine counries. 9 Anderson and Reeb (2003) observe ha families own 18% of he companies equiy in he S&P500 companies from 1992 o

14 pyramids in he ulimae conrol, and he involvemen of family members acing in managemen. These differences we idenified indicae a disinc role played by a family or hired CEO in family firms, and i is hus imporan o disinguish beween he wo groups of family firms in our analysis. Table 3 Descripive saisics of main variables Panel 3A: Summary saisics of variables Variable Mean Sd. Dev. Min Max GENVIORN GOVERN FINANCE INTERMEDIARY JUDICIARY CASHFLOW EXCESSCONTROL FAMILYRATIO SALE SIZE LEVERAGE AGE Panel 3B: Summary saisics of dummies 0 1 Number Per cen Number Per cen PYRAMIDS SYSTEM 2, Panel 3C: Summary saisics by family/ousider CEO Family CEO Ousider CEO Differ. of means -saisic Number of observaions CASHFLOW *** EXCESSCONTROL *** PYRAMIDS FAMILYRATIO *** SYSTEM SALE ** SIZE *** LEVERAGE *** AGE *** Noes: Descripive saisics of he governance environmen index are compued on a cross-secional sample of 31 provinces in China. Descripive saisics of oher variables are compued on he family-firm subsample which comprises 2,920 observaions afer filering four ouliers. Dummies in Panel 3B are compued by he number and percenage of observaions equalling he column heading o he observaions of he family-firm sample. Dummies in Panel 3C and 3D are compued by he percenage of observaions equalling one o he observaions of family/ousider CEO subsample. Aserisks denoe saisical significance a he 10% (*), 5% (**), or 1% (***) level, respecively. 4 Resuls and discussion 4.1 Caegorical analysis To examine he effecs of invesor proecion on family ownership, conrol, and managemen, we firs use a dummy variable of invesor proecion o spli he full familyfirm sample ino wo subsamples family firms operaing in regimes wih inferior invesor 14

15 proecion and family firms wih superior invesor proecion. Then, he hree ses of variables indicaing family ownership, conrol, and managemen, as well as conrol variables are compared beween he wo groups of family firms, respecively. Panel 4A displays he resuls. Overall, he analysis shows ha family firms locaing in regimes wih a more proecive governance environmen are predominae: 58.12% versus 41.88% of family firms wihin a less proecive governance environmen. Then he analysis finds ha he wo groups of family firms differ significanly in family ownership, conrol, and managemen. Specifically, i shows ha family firms wih srong invesor proecion hold more concenraed cash-flow righs, 28.3% on average, compared o family firms wih weak proecion having 21.7% of cash-flow righs on average. For 74.2% of he laer group of companies, he ulimae conrol is capured by he use of a pyramidal srucure, in which he number is lower in provinces wih sronger invesor proecion (66.3% on average). Following his, he excess of conrol righs over cash-flow righs is significanly higher in family firms ha are beer proeced han comparable companies. Furher, 17.7% of family firms locaing in less proecive regimes are under he conrol by he same family or individual, while only 12.3% in more proecive regimes. For family managemen, beer proeced family firms have larger involvemen of family members in managemen a 9.5%, as opposed o comparable firms a 6.6%. The above findings firs demonsrae a negaive relaion beween he qualiy of invesor proecion and he use of conrol-enhancing mechanisms. As he use of hese mechanisms are assumed o reflec he family s incenives and abiliies o appropriae privae benefis of conrol (Bebchuk e al., 1999; La Pora e al., 1999), he expropriaion, exposed o weak legal proecion of shareholders, is less likely o be deeced and consrained by laws. In his regard, for family firms wihin a less proecive governance environmen, he family would hus magnify boh he incenive and power o exrac a he expense of minoriy shareholders by exercising conrol-enhancing mechanisms. On he oher hand, he analysis finds ha boh ownership concenraion and involvemen of family managemen are posiively relaed o invesor proecion, which are inconsisen wih he convenional predicion in he lieraure, likewise La Pora e.al (1998), Shleifer and Wolfenzon (2002) and Burkar e al. (2003). Such relaions can be explained as follows. The regimes wih beer invesor proecion are likely o have more efficien governmen, easier access o finance, more efficien cours and judiciary. A more 15

16 proecive governance environmen would predic boh a reducion o he operaion coss of family firms and also a rise in he expropriaion coss for he conrolling shareholders. To his end, he family would have greaer inenion o make effors o run he business well raher han o unnel i. I would provide he company wih more cash-flow invesmen and involvemen of family managemen, and meanwhile limi he use of conrol-enhancing mechanisms such as pyramids. Alernaive explanaion for his finding is given by Casillo and Skaperdas (2005) who propose ha beer proecion of shareholders can inensify he compeiion beween owners and managers, hereby giving rise o he appropriaive coss associaed wih he owner-manager agency conflic. To counerac he negaive effecs of law and is changes, shareholders would commi o a lower exen of appropriaion by holding more shares of he company. Thus, his finding can also reflec he posiive relaionship beween legal proecion and ownership concenraion as for insance Aganin and Volpin (2003) noe in Ialian corporaions. In addiion, in family firms operaing in more proecive regimes, he larger preservaion of family managemen also indicae ha o achieve he same exen of exracion he family would ake more cover mechanisms such as in he form of delegaing family managers o replace he use of conrol-enhancing mechanisms. This is expeced o be a response o he relaively high expropriaion coss in regimes wih beer legal proecion of shareholders. Panel 4B furher divides he family-firm sample ino wo subsamples family firms having a family CEO and family firms wih a hired CEO o invesigae he ineracion effecs of family managers and invesor proecion on family governance arrangemens. As seen in his panel, boh subsamples show ha family firms in a more proecive governance environmen have significanly higher cash-flow righs and more paricipaion of family managemen, as opposed o family firms wih inferior proecion of shareholders. This is consisen wih he resuls repored in Panel 4A. Furher, he analysis shows ha, alhough he mean values of cash-flow righs and family managemen raio in family-ceo firms are on average higher han hose in ousider-ceo group, he mean difference of ownership concenraion and family managemen beween regimes wih beer and poorer invesor proecion is more significan in ousider-ceo family firms. To illusrae, our analysis shows ha, for ownership concenraion, he mean difference beween he regimes is significan a 1% in ousider-led family firms, while a 5% in family-led firms. The ousider-ceo group also has a much larger -saisics of mean difference of family managemen: versus

17 in family-ceo group. This finding suggess ha he arrangemens of he conrolling family on ownership concenraion and managemen composiion are o some exen subjeced o he changes of invesor proecion, bu such effecs migh no be uniform beween familyand ousider-ceo family firms. Anoher noable resul is ha, he posiive mean difference in he separaion of ownership and conrol beween regimes wih weak and srong legal proecion only survives in family firms hiring an ousider CEO: he mean difference is and significan a 5% level in ousider-ceo firms; i is however and insignifican in family-ceo firms. This finding suggess ha, for family firms having a hired CEO in place, he conrolling shareholder would adjus his/her incenives of exracion of he privae benefis of conrol hrough he form of pyramid srucures in he ligh of he srengh of invesor proecion. In oher words, in ousider-ceo family firms, he conrolling family would inensify he exracion a he expense of minoriy shareholders when legal proecion of shareholders is relaively weak. In conras, he incidence of he conrolenhancing mechanisms is of no difference in family-ceo firms regardless of he qualiy of legal proecion of shareholders. Table 4 Impac of invesor proecion family ownership, conrol, and managemen Panel 4A: Full family-firm sample Governance environmen Variable Low (75%) High (25%) Diff. of means -saisic Number of observaions CASHFLOW *** EXCESSCONTROL *** PYRAMIDS FAMILYRATIO *** SYSTEM SALE *** SIZE *** LEVERAGE AGE *** Panel 4B: Family/ousider CEO subsample Family CEO Ousider CEO Governance environmen Diff. of means Governance environmen Diff. of means Variable Low (75%) High (25%) Low (75%) High (25%) Number of observaions CASHFLOW (2.132)** (8.791)*** EXCESSCONTROL (-1.087) ** (2.279) PYRAMIDS FAMILYRATIO (-2.651)*** (-7.314)*** 17

18 SYSTEM SALE (-2.595)*** (6.563)*** SIZE (3.567)*** (-3.213)*** LEVERAGE (-0.881) (-0.858) AGE (3.260)*** (6.193)*** Noes: Dummies are compued by he percenage of observaions equalling one o he observaions equalling he column heading. Aserisks denoe saisical significance a he 10% (*), 5% (**), or 1% (***) level, respecively. 4.2 Random effecs models Table 5 displays hree panels of he RE esimaion for he equaions presened before. The sample of Panel 5A, 5B and 5C comprises all family firms, family- and ousider- CEO family firms, respecively. Column 1 and 4 of Panel 5A esimae he effecs of invesor proecion on family ownership and managemen, while Column 2 and 3 esimaes he effecs on he excess of family conrol over cash-flow righs and he use of pyramids. Column 1 and 4 show ha he coefficien of invesor proecion is in he cash-flow righs regression and in he family managemen regression; i is boh saisically significan a 1% level. This finding suggess ha boh he concenraion of he family s ownership righs and preservaion of family managemen are posiively associaed wih local legal proecion of shareholders, which is consisen wih he caegorical analysis in his paper. Addiionally, invesor proecion has a negaive and significan effec on pyramid srucures (-0.113) and hus on he separaion of ownership and conrol righs (-0.008). This resul confirms ha legal proecion would limi he use of conrol-enhancing mechanisms, and also suppors he argumen ha pyramiding is he principal insrumen o enhance he family s conrol righs in he ulimae conrol srucure. Panel 5B and 5C provide furher quanificaion of he ineracion effecs of family managers and invesor proecion on family governance arrangemens. Panel B presens he resuls for family firms having a family CEO, while Panel 5C for family firms hiring a professional CEO. In Panel B, he coefficien of invesor proecion is recorded a 0.011, , and in he ownership righs, he separaion of ownership and conrol, he pyramids, and he family managemen regression, respecively; he absolue value of he coefficien is relaively smaller han ha recorded in Panel 5C (0.017, , and 0.006, respecively). Noably, none of hese coefficiens is saisically significan in Panel B, while all are saisically significan in Panel 5C. This finding suggess ha he effecs of invesor proecion on family ownership, conrol, and managemen are more 18

19 pronounced in family firms hiring a professional CEO as opposed o family firms having a family CEO. I also helps o inerpre he similar resuls repored in he caegorical analysis, which shows ha family-ceo firms have smaller significan mean difference of ownership concenraion and family managemen, and nonsignifican mean difference of he excess of family conrol over ownership righs beween he regimes wih beer and poorer invesor proecion. One possible explanaion is ha as family managers are more likely o say wih he conrolling family a he innermos circle o share high rus and muual affecion (Fe 1992), family-ceo firms are expeced o have a smaller conflic of ineres beween he family and he manager. In his regard, wheher he choice is o run he business well or o unnel i, he family end o coun on inernal corporae governance mechanisms, such as reaining concenraed ownership righs and preserving more managerial power in he family, raher han o depend on he qualiy of exernal invesor proecion. In conras, given a hired CEO in place, family firms are exposed o a more inense divergence of ineress beween he conrolling family and he professional manager and also a conflic beween his manager and oher family managers. Therefore, o miigae he appropriaive coss associaed wih boh kinds of conflics, he family would adjus he decision-making on family governance subjeced o he exernal proecion of shareholders. Our findings also uncover he possible dual effecs of family managers. On one hand, he high rus and loyaly among family members would help o miigae he ownermanager agency conflic ha migh be oherwise inensified in family firms hiring an ousider as CEO. On he oher hand, due o he high rus among family members, family managers are inclined o have consisen ineress wih he conrolling family oher han wih all shareholders. Because he conrolling family has boh he incenives and abiliies o expropriae privae benefis of conrol, he larger involvemen of family managemen especially wih a family CEO would faciliae and even magnify such expropriaion. To his end, he preservaion of family managemen also suggess a higher incidence of collusion beween he family and family managers han beween he family and heir professional counerpars. Table 5 RE esimaion of governance environmen on family ownership, conrol, and managemen Panel 5A: Full family-firm sample CASHFLOW EXCESSCONTROL PYRAMIDS FAMILYRATIO GENVIRON 0.020*** (4.43) *** (-3.58) ** (-2.54) 0.008*** (5.12) SYSTEM (0.64) (0.99) 0.329* (1.87) *** (-2.80) 19

20 SALE 0.012* (1.71) (-0.53) ** (-2.44) (0.46) SIZE 0.025*** (3.25) 0.008** (2.16) ** (-2.17) (1.54) LEVERAGE 0.002** (2.29) 0.001* (1.86) (0.57) (-0.10) AGE *** (-5.49) *** (-5.49) *** (-2.98) *** (-3.35) Indusry dummies Yes Yes Yes Yes Inercep * (-1.89) (0.03) 5.060*** (3.45) (-0.31) Observaions Number of clusers (firm) R-squared Panel 5B: Family-CEO subsample CASHFLOW EXCESSCONTROL PYRAMIDS FAMILYRATIO GENVIRON (1.03) (-0.74) (-0.28) (0.78) SYSTEM 0.036* (1.66) (1.47) (-1.04) (0.74) SALE 0.042*** (2.62) (0.04) (-0.52) (1.14) SIZE (1.23) (1.19) (0.95) (0.60) LEVERAGE 0.006*** (2.94) (-0.45) (-0.86) 0.003*** (2.73) AGE *** (-6.54) *** (-2.77) (0.64) (-1.51) Indusry dummies Yes Yes Yes Yes Inercep (-0.42) (-0.64) (0.20) Observaions Number of clusers (firm) R-squared Panel 5C: Ousider-CEO subsample CASHFLOW EXCESSCONTROL PYRAMIDS FAMILYRATIO GENVIRON 0.017*** (4.31) *** (-3.57) * (-1.95) 0.006*** (3.86) SYSTEM (0.60) (0.87) (1.35) ** (-2.20) SALE (0.93) (-0.49) *** (-2.81) (0.48) SIZE 0.026*** (3.16) 0.010** (2.32) *** (-2.60) (1.32) LEVERAGE 0.002** (2.20) 0.002** (2.47) (0.62) (-1.25) AGE *** (-3.01) *** (-5.35) *** (-4.70) *** (-3.45) Indusry dummies Yes Yes Yes Yes Inercep ** (-1.98) (-0.11) 8.030*** (4.40) (-0.22) Observaions Number of clusers (firm) R-squared Noes: Aserisks denoe saisical significance a he 10% (*), 5% (**), or 1% (***) level, respecively. 20

21 5 Robusness es Table 6 examines he robusness of main resuls repored in he caegorical analysis o he use of alernaive specificaion of he cu-off poin for he level of invesor proecion. To be consisen wih he convenional sudies, he cu-off poin for he degree of invesor proecion is considered as he median score of invesor proecion index. Tha is, for he high level of invesor proecion, a dummy variable equals one if he value of governance environmen of a paricular province is above he median score of he governance environmen index across 31 provinces in China, and zero oherwise. In Panel 6A, we reinvesigae he mean differences of family ownership, conrol, and managemen beween he regimes wih beer and poorer proecion of shareholders. Panel 6B furher explores he mean differences in family firms managed by a family or a hired CEO. The main resuls remain qualiaively unchanged. Table 7 han addes ime dummies o he use of he RE esimaion o conrol for he ime effecs. The analysis finds ha afer conrolling for he ime effecs, he main resuls hardly change. Panel 7A shows ha for he full family-firm sample he coefficien of invesor proecion is signicanly posiive in he cash-flow righs and he family managemen regression: and 0.006, respecively. As for he family conrol vairables, he coefficien of invesor proecion is negaive a and significan a 5% level in he separaion of ownership and conrol regression, and his coefficien is negaive a , alhough no saisically significan, in he pyramids regression. In Panel 7B ha comprises family firms in which a family member acs as CEO, none coefficien of invesor proecion is saisically significan in all five models, bu his coefficien is significan in he ownership righs, he separaion of ownership and conrol, and he family managemen regression in ousider-led family firms (see Panel 7C). This finding confirms ha he effecs of invesor proecion on family governance arrangemens is mainly driven by professional managers. Table 6 Impac of invesor proecion on family ownership, conrol, and managemen Panel 6A: Full family-firm sample Governance environmen Variable Low (0) High (1) Diff. of means -saisic Number of observaions CASHFLOW *** EXCESSCONTROL *** PYRAMIDS FAMILYRATIO *** SYSTEM SALE *** SIZE * 21

22 LEVERAGE AGE *** Panel 6B: Family/ousider CEO Family CEO Ousider CEO Governance environmen Diff. of means Governance environmen Diff. of means Variable Low (0) High (1) Low (0) High (1) Number of observaions CASHFLOW (-3.377)** (-6.751)*** EXCESSCONTROL (1.059) (3.413)*** PYRAMIDS FAMILYRATIO (-2.453)** (-2.664)*** SYSTEM SALE ( (-6.955)*** 3.619)*** SIZE (-0.428) (-1.226) LEVERAGE (-0.165) (-1.916)* AGE (2.573)*** (3.969)*** Noes: Aserisks denoe saisical significance a he 10% (*), 5% (**), or 1% (***) level, respecively. Table 7 RE esimaion of invesor proecion on family ownership, conrol, and managemen Panel 7A: Full family-firm sample CASHFLOW EXCESSCONTROL PYRAMIDS FAMILYRATIO GENVIRON 0.012*** (2.92) ** (-2.27) (-1.09) 0.006*** (3.75) SYSTEM (0.73) (0.77) (0.49) ** (-2.49) SALE (0.98) (-0.06) ** (-2.06) (0.17) SIZE 0.020*** (2.60) 0.011*** (3.05) (1.06) (0.87) LEVERAGE 0.002** (2.14) 0.001** (2.14) (0.69) (-0.20) AGE *** (-12.77) 0.002** (2.32) 0.049*** (3.01) *** (-8.39) Indusry dummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Inercep (-0.98) (-1.22) (0.69) (0.75) Observaions Number of clusers (firm) R-squared Panel 7B: Family CEO sample CASHFLOW EXCESSCONTROL PYRAMIDS FAMILYRATIO GENVIRON (0.54) (-0.19) (-0.04) (1.08) SYSTEM 0.040* ( (1.10) (-1.31) (1.05) SALE 0.038** (2.25) (0.32) (-0.84) (1.40) SIZE ** 0.229*