The Diffusion of Corporate Governance to Emerging Markets: Evaluating Two Dimensions of Investor Heterogeneity

Transcription

1 MPRA Munch Personal RePEc Archve The Dffuson of Corporate Governance to Emergng Markets: Evaluatng Two Dmensons of Investor Heterogeney Woochan Km and Taeyoon Sung and Shang-Jn We Aprl 2014 Onlne at MPRA Paper No , posted 18. June :41 UTC

2 The Dffuson of Corporate Governance to Emergng Markets: Evaluatng Two Dmensons of Investor Heterogeney WOOCHAN KIM Korea Unversy Busness School and ECGI TAEYOON SUNG Yonse Unversy, School of Economcs SHANG-JIN WEI Columba Unversy Busness School, NBER, and CEPR Abstract Ths paper nvestgates whether foregn nstutonal nvestors n emergng markets can enhance shareholder value. We pay specal attenton to two dmensons of nvestor heterogeney: whether an nvestor declares self as an actvst, and whether an nvestor comes from a country wh a strong tradon of nvestor actvsm. Frst, we apply an event study approach to the announcements of block purchases by foregn nstutonal nvestors n Korea. We fnd that stock prces rse on average, but only when foregn nstutonal nvestors declare themselves as actvsts. Source country dentes also matter: The posve stock prce reactons are more pronounced when the actvst nvestors come from source countres wh a strong tradon of nvestor actvsm. Second, we examne corporate fnancal polces and governance practces of target frms one to three years followng block purchases by foregn actvst nvestors. We fnd that target frms are more lkely to reduce cash holdngs, rase leverage ratos, and peg dvdend payouts and stock repurchases more closely to changes n earnngs, but only f foregn actvsts are from countres wh a strong tradon of actvsm. We conclude that openness to foregn actvst nvestors, especally those from countres wh a strong tradon of actvsm, can ndeed put pressures on frms n emergng markets to adopt corporate governance practces that enhance shareholder values. Key words: foregn block nvestors, actvst nvestors, corporate governance, emergng markets JEL classfcaton: F21, G11, G15, G23, G32, G34, G35 Km: Korea Unversy Busness School, 145 Anam-Ro, Seongbuk-Gu, Seoul, , Korea, tel: E-mal: wckm@korea.ac.kr ; Sung: Yonse Unversy, School of Economcs, 50 Yonse-Ro, Seodaemun-Gu, Seoul , tel: E-mal: tsung@yonse.ac.kr; We (correspondng author): Columba Unversy, Graduate School of Busness, 619 Urs Hall, 3022 Broadway, New York, NY 10027, USA, tel: Emal: shangjn.we@columba.edu. We thank Reena Aggarwal, Pedro Matos, We Jang, and other partcpants at the conference, corporate governance around the world, held at Columba Unversy, and semnars at Korea Unversy Busness School and the ESG Research Study Group for helpful comments. We also thank Ellen Ln and Nkhl Patel for edoral assstance

3 1. INTRODUCTION The effects of fnancal globalzaton for overall growth performance of developng countres are controversal (see Stultz, 2005 on the lms of fnancal globalzaton, and a survey of emprcal work by Kose et al., 2010 on the subject). On the one hand, some argue that foregn drect nvestment appears more benefcal and generates less vulnerably to host countres than foregn portfolo nvestment (We, 2001 and 2006). On the other hand, some report evdence that greater openness to foregn portfolo nvestors can reduce the cost of capal for domestc frms (Bekaert and Harvey, 2000, and Char and Henr, 2004). In ths paper, we assess emprcally one partcular potental benef of openness to foregn portfolo nvestment as a mechansm for transmtng better corporate governance practces to host country frms, and thereby enhancng shareholder value. The potental for corporate governance mprovement through nvestor engagement s commensurate to the scale of actvst nvestors, whch could nclude actvst hedge funds, penson funds, and soveregn wealth funds. The level of assets under management by actvst hedge funds alone s estmated to have ncreased by about $10 bllon n 2003 to over $90 bllon by 2013 (Marketwatch.com, 2014) 1. Ths type of nvestor engagement could be especally useful n host countres that are not prevously known for shareholder-responsve corporate governance. The more room for mprovement n corporate governance, the more useful the external pressure for change. Our partcular settng s foregn actvst nvestors n Korea. Korea s known for problematc corporate governance. Research suggests exstence of extensve tunnelng actves by controllng shareholders (Bae, Kang, and Km, 2002; Baek, Kang, and Lee, 2006; Black et al., 2008; and Km, Sung, and We, 2011). Because there s sgnfcant room for mprovement n the corporate governance area, s meanngful to nvestgate possble value-enhancng effects of nvolvement of foregn nstutonal nvestors. At the same tme, Korea has a well-developed dsclosure rule that requres foregn actvst nvestors to dsclose ther ntenton whenever they purchase a 5% block or more. The dsclosure rule s useful for dentfyng sgnfcant block purchases by foregn actvst nvestors, whch s crucal for research on dffuson of cross-border corporate governance. In ths paper, we frst use an event study approach, examnng the stock prce response to foregn nstutonal nvestor s block purchase announcements. The relatve narrow wndow (e.g., fve days before and after the event) allows us to dentfy the prce effect that s lkely attrbutable to the stock purchase actons. Moreover, we separate block purchases wh an explc actvst agenda ( pckng a fght ) and 1 Actvsts here to stay as war chests near $100 bllon,

4 those prompted solely by valuaton judgments ( pckng a stock. ). The dfference n stock prce responses s lkely to capture the value of actvsm. More nterestngly, we nvestgate whether the effect could be stronger f actvst nvestors are comng from countres wh a tradon of nvestor actvsm. Ths s based on our conjecture that home country governance practces become an mportant reference when nvestment managers nvest abroad and set the strength of ther engagement wh corporate managers n host countres. If ths s what the market expects, the stock prce reacton of a target frm should be greater when engaged by actvst from a country wh hgher governance standards. We also analyze changes n frm value, corporate governance, and corporate fnancal polces (e.g. cash holdng, leverage, corporate governance, dvdend payouts, and stock repurchases) of the target over a four-year perod [-1, +3] around the year of block holdng announcement. We check f the results found n our earler event studes translate nto these long-term outcome varables. Lastly, we supplement our study wh a survey of nternatonal nstutonal nvestors wh an exposure to Asan markets. The goal s to narratvely check whether and to what degree foregn nvestors from countres wh stronger shareholder protecton are more effectve actvsts. To our knowledge, ths s done for the frst tme n the lerature. We have two key fndngs. Frst, we fnd that stock prces rse on average, but only when foregn nstutonal nvestors declare themselves as actvsts. We dscover that source country dentes also matter. The posve stock prce reactons are pronounced when the actvst nvestors are comng from source countres wh a strong tradon of nvestor actvsm. Second, we examne corporate fnancal polces and governance practces of target frms. We fnd that target frms are more lkely to reduce cash holdngs, rase leverage ratos, mprove overall corporate governance, and mprove dvdend payout s (and stock repurchases ) responsveness to earnngs f foregn actvsts are from countres wh a strong tradon of nvestor actvsm. The rest of the paper s organzed as follows. Frst, Secton 2 dscusses the related lerature, the hypotheses, and the survey results. Secton 3 brefs the data and Secton 4 presents the results wh robustness checks. Fnally, Secton 5 concludes. 2. Hypothess, Data, and Survey of Market Partcpants A. Related lerature A number of papers have provded evdence that nternatonal drect nvestors may help promote good corporate governance around the world. One obvous channel through whch governance standards can - 3 -

5 spll over from one country to another s through cross-border mergers. For example, Ross and Volpn (2004) study cross-border mergers among 49 countres and show that targets are often n countres wh poorer nvestor protecton than ther acqurers countres; ths s consstent wh, though not drectly provng the noton that governance typcally splls over from countres wh stronger corporate governance standards to countres wh weaker ones. A number of studes also suggest that cross-border spllovers through drect nvestment or M&As produce a valuaton effect. Brs and Cabols (2008) study full takeovers n 39 countres and fnd that the better the governance practces n the acqurer s country, the hgher the merger premum n cross-border mergers. Brs, Brsley, and Cabols (2008) show that Tobn s q of an ndustry ncreases when frms whn that ndustry are acqured by foregn frms comng from countres wh better corporate governance standards. Martynova and Renneboog (2008) reconfrm ths fndng n samples lmed to partal takeovers. More recently, Albuquerque et al. (2013) show that cross-border mergers and acqusons are assocated wh subsequent mprovements n the governance, valuaton, and productvy of the target frms local rvals. Ths productvy spllover effect s stronger when the acqurer s from a country wh stronger shareholder protecton and f the target s ndustry s more competve. Many developng countres, whle open to drect foregn nvestment, mantan hghly restrctve controls on ndrect or portfolo nvestment by foregn nvestors. Chna s a promnent example for ths. Whle s a leadng host of foregn drect nvestment n the world, places severe restrctons on nward foregn portfolo nvestment. No foregn mutual funds, penson funds, or hedge funds can drectly nvest n Chnese companes on the Shangha or Shenzhen stock exchanges, except through a lmed quota of qualfed foregn nstutonal nvestors (QFII) or through a porton of Chnese shares lsted n Hong Kong or New York. Such asymmetrc treatments of drect and ndrect foregn nvestment n capal controls regmes are common among developng countres. Ths means a proper assessment of the costs and benefs of openness to nternatonal portfolo nvestment has mportant polcy mplcatons, especally for developng countres. In ths context, s of partcular nterest to nvestgate nternatonal nstutonal nvestment as a possble mechansm for dffuson of value enhancng corporate governance practces. In theory, ndrect or portfolo nvestment can also be a channel for dffuson of corporate governance f foregn nstutonal nvestors mprove monorng of corporate management va more shareholder actvsm. For example, foregn nvestors may be better monors of corporate governance than domestc nvestors as the former are not owned, controlled, or have close busness tes wh local frms. Aggarwal et al. (2011) suggest that one reason for ths s that nstutonal nvestors at home may feel more compelled to be loyal to management. They ced the example of Fdely nvestment, reported n BusnessWeek (2006), whch apparently was more aggressve on governance ssues n Europe than n s - 4 -

6 home country, the Uned States. Ferrera and Matos (2008) report evdence that foregn nstutonal ownershp s posvely correlated wh frm value outsde the Uned States, but such correlaton s absent n the Uned States. Km, Km, and Kwon (2009) study domestc and foregn blockh olders that swch ther nvestment purpose from passve to actve n the Korean market and show that the posve valuaton effect around the tme of swch s greater for foregn block holders. The most closely related paper to ths study s Aggarwal et al. (2011). They use a panel regresson approach wh data at annual frequency, lnkng company level corporate governance scores wh foregn nstutonal holdngs from 23 OECD countres durng Ths s an mportant paper that has provded frst-ever evdence that foregn nstutonal nvestors and local corporate governance are lnked. At the same tme, however, panel regressons do not allow for sharp dentfcaton of causaly especally when the data are at a relatvely low frequency. In addon, ther data do not allow them to separate foregn nvestors wh an actvst ntenton versus those whout such an ntenton. In comparson, we focus on a major emergng market economy, Korea, rather than developed markets. (Korea s not n Aggarwal et al. s data set), and take advantage of a Korean legal dsclosure rule that requres block purchasers to declare f they plan to be an actvst or not. The frst part of our paper wll use an event study approach to sharpen the dentfcaton of the effects of block purchases by foregn nvestors. We look nto possbly dfferent effects of purchases by actvst nvestors versus pure portfolo nvestors. Among actvst foregn nvestors, we also examne possble role of home country tradon n nvestor actvsm. We wll report evdence that both dmensons of nvestor heterogeney matter n mportant ways. B. Hypotheses A key hypothess we nvestgate n ths paper s whether block purchases by foregn actvst nvestors can brng posve valuaton responses. We check two dmensons of heterogeney. The frst s between actvst and passve nvestors. An actvst nvestor s one who, at the tme of purchase, declares hs/her ntenton to push for changes n management practces or other corporate governance aspects for the explc purpose of enhancng shareholder value. A passve nvestor s one who does not have such ntenton. Note that a truly passve nvestor does not have an ncentve to declare self as an actvst snce ths mght antagonze management wh no tangble benef. By the Korean law, an nvestor has to declare self as an actvst nvestor before can undertake a number of actons such as askng for a board seat, makng a shareholder proposal at the shareholder meetngs, etc. As we wll note later, an nvestor wh an actvst ntenton would always declare self as an actvst. Otherwse, wll not be able to deploy actvst - 5 -

7 tools legally. On the other hand, a pure portfolo nvestor s unlkely to declare self as an actvst as would otherwse antagonze the management for no apparent benef (See Km, Km, and Kwon, 2009, for a detaled descrpton of the dsclosure rules on block holders n Korea). The second dmenson of nvestor heterogeney s the type of source countres from whch an nvestor s comng. In some source countres (e.g., the Uned States or the Uned Kngdom), s common for nstutonal nvestors to challenge management for ther value destroyng or other nonvalue-maxmzng practces. Inally, they mght engage wh the management n prvate, but f such practces persst, they may take stronger actons aganst the management n publc. In some cases n whch corporate management does not respond or does not do so adequately, one may see hostle takeover attempts. In some other countres (e.g., Japan or Denmark), s relatvely uncommon to see such nvestor actvsm. These dfferences may reflect dfferent legal envronment, nstutonal nvestors tradon, or ther combnatons. Whatsoever the reason, we conjecture that home country governance practces can be an mportant reference when nstutonal fund managers nvest abroad and calbrate ther engagement wh corporate managers n host countres. If ths s what the market expects, the stock prce reacton of a target frm should be greater when s engaged by actvst from a country wh a stronger tradon of nvestor actvsm or hgher governance standards. To summarze, the frst hypothess we test s whether actvst foregn nstutonal nvestors enhance frm value more than non-actvst nvestors; the second hypothess we test s whether actvst foregn nvestors from countres wh a strong tradon of actvsm enhance frm value more than actvst nvestors from countres whout such a tradon. We make use of three approaches. The frst s a survey of a group of nternatonal nstutonal nvestors that are actve n Asan emergng market economes. The second s an event study approach appled to cumulatve abnormal returns at the daly frequency. The thrd s a panel regresson on changes n corporate fnancal practces at the annual frequency followng a block purchase by foregn actvst nvestors. C. Survey We start wh a survey of nstutonal nvestors who are actve n Asa, especally n cross-border portfolo nvestment. Ths s not meant to be a drect test of our hypotheses per se because a gven market practoner may observe only nosy sgnals about other nvestors actons and have only mperfect knowledge about the effectveness of other funds actons. Nonetheless, s useful to gauge the extent to whch market practoners recognze the occurrence of shareholder actvsm by foregn nstutonal - 6 -

8 nvestors n general, ther common tactcs, and ther assessment of effectveness. As far as we know, such a survey has not been mplemented before or at least has not been reported n the exstng lerature. The survey respondents are members of the Asa Corporate Governance Assocaton (ACGA). The ACGA s a Hong Kong based non-prof organzaton establshed to promote good corporate governance n Asa; s 100 members ncludng penson funds, nvestment management companes, nsurance companes, and other fnancal nstutons. 2 In the survey, we ask three substantve questons. Frst, we ask to what extent respondents agree that foregn nstutonal nvestors can add value by shareholder engagement or actvsm. Second, we ask the respondents to ndcate ther vews on the most mportant areas of engagement or actvsm. We ask them to pck up to four out of a lst of nne potental areas. Thrd, we ask whether source country tradon matters. More precsely, we ask respondents the extent to whch they agree that nvestors from countres wh stronger nvestor/shareholder rghts are more lkely to demand actons from the management to enhance value. We attach the actual survey questons n the appendx. We worked wh the Secretarat of the ACGA and posted our survey on the web usng Qualtrcs. The ACGA secretarat promoted the survey n a meetng wh s membershp and urged s members to respond to the survey. Durng a three-week survey perod (August 22 September 13, 2013), we receved 38 responses, about 40% of ACGA members. Note that the survey questons are not specfcally about foregn nvestor actvsm n Korea, and the survey respondents don t necessarly nvest n Korea. The responses should be vewed as subjectve perceptons about foregn nvestor actvsm n general by a group of self-selected fund managers wh a specal nterest n corporate governance n Asa. Nonetheless, f the hypotheses we wsh to test are flatly rejected by market practoners, we wll have to be suspcous about clams regardng the benefs of foregn nvestor actvsm. Regardng the frst queston, 97% of our respondents eher strongly agreed or bascally agreed that foregn nstutonal nvestors can add value by shareholder engagement/actvsm. Only one respondent dsagreed. We conclude that there s a very broad agreement, though not a consensus, that shareholder actvsm can enhance shareholder value n emergng market economes. For the second queston, our respondents consdered ncrease scrutny on related-party transacton (95% of the responses), mprove board ndependence (89%), lnk CEO turnover more to performance (53%) to be the most mportant areas of engagement or actvsm. These are followed by lnk dvdend payouts more to earnngs (47%), and reduce dle cash (34%). Lnk stock buybacks more to earnngs, and ncrease leverage receved relatvely few affrmatve responses. 2 Vs s home page at for detals

9 Regardng the thrd queston, 95% of our respondents eher strongly agreed or bascally agreed that nvestors from countres wh stronger nvestor/shareholder rghts are more lkely to demand actons from the management to enhance value. Only two respondents dsagreed. Responses to our supplementary questons reveal that 97% of our respondents have exposure to emergng market eques and 79% of our respondents have engaged n shareholder actvsm n emergng markets. 3. DATA A. Foregn block holdng data Our event study sample covers all block purchasng announcements made by foregn portfolo nvestors n the Korean stock market (KRX) from 2005 to s the frst year when block purchases are requred to dsclose the purpose of ther holdngs. We exclude block buyng announcements made by non-fnancal nstutons as we wsh to focus on portfolo nvestments by nstutonal nvestors. We also exclude announcements where pre- or post-announcement block ownershp s greater than 20%. We manually collect the data from the Data Analyss and Retreval Transfer (DART) system, whch s an electronc dsclosure system admnstered by the Fnancal Supervsory Servce (FSS), the Korean secures authory. Smlar to the EDGAR system operated by the U.S. SEC, the system allows companes or nvestors to subm dsclosures onlne. From DART, we collect the date of block holdng dsclosure, the target frm, foregn block holder s name, s country of ncorporaton, and s equy holdngs n the target frm. As for nvestors ncorporated n tax heaven terrores, we track the resdence of ther nvestor managers by nvestgatng ther annual reports or perodcals. 3 Korea s rule on block holdng dsclosure s very smlar to that n the U.S. Any nvestor who acqures share ownershp of 5% or more of a publc company must fle a report to FSS whn 5 (or 10) days. A subsequent change n share ownershp by one percent or more also trggers a dsclosure requrement. Block holders also have the oblgaton to dsclose whether they have the ntenton to nfluence management at the tme of ther flng, whch s the bass of our classfcaton of actvst purchases vs. non-actvst purchases. If they do, they must fle a detaled long-form whn 5 days after the acquson. If not, they must fle an abrdged short-form whn 10 days. The long-form ncludes a secton that lsts ten potental areas of actvsm. The flng actvst must dsclose s detaled ntenton by checkng whchever area that s relevant. Unfortunately, n practce, most of foregn actve nvestors tend 3 One example s Templeton Asan Growth Fund, whch s ncorporated n Luxembourg, but actually managed n Hong Kong

10 to check all ten areas, apparently to gve themselves the maxmum flexbly n engagng wh the management. A swch n block holdng purpose also trggers a flng whn 5 days. 4 For detaled dscussons on Korea s 5% rule and s comparson to the U.S. rule, we refer to Km, Km, and Kwon (2009). Fgure 1 shows the number of foregn block buyng announcements over tme. Notce that we do not lm the sample to the frst 5% announcements made by block holders. We also nclude ther subsequent announcements that had to be made when they ncreased ther holdngs by one percent or more. Thus, one block holder can make multple announcements n a gven target frm. In prncple, one block holder can make announcements n multple frms, and one frm can be a target of multple block holders; but, n practce, these happen rarely. The frst bar chart (A) shows the number of all foregn block buyng announcements from 1998 to The sample ncludes actve (n darker shade snce 2005) as well as passve block holders (n lghter shade). One can see that there was a clear upward trend untl The number of announcements, whch was just one n 1998, soared to 582 n But, subsequent to the peak year of 2007, the number of block buyng announcements drop sgnfcantly. The relatve retrenchment after 2007 possbly reflects a global capal reversal out of emergng market economes back to major source countres that were facng a fnancal crss at home. Becht, Franks, and Grant (2014) also fnd a smlar result n ther study of nternatonal hedge fund actvsm. From ths bar chart, one can also see that the number of actvsts takes up only a small fracton (less than 10%) of all block holders. The second bar chart (B) lms the sample to announcements made by actvst block holders. Table 1 lsts the orgns (country of ncorporaton or regstraton; for tax heaven terrores, the domcle of the management company) of foregn block holders wh ther home country governance characterstcs: hostle takeover, M&A volume, and board ndependence. Hostle takeover s the percentage of domestc frms that experenced hostle takeover attempts durng M&A volume s the percentage of domestc frms that were targeted n completed deals durng These two measures are from Ross and Volpn (2004). Board ndependence s the mean percentage of ndependent drectors (number of ndependent drectors dvded by board sze), and s from Dahya, Dmrov, and McConnell (2008). For each home country characterstcs, we group countres nto two: above and below the country medan values (see second row parentheses). That s, hgh (H) vs. low (L) countres. In the frst column, 4 Korea s 5% rule went through three major revsons snce s frst adopton n The defnon of share ownershp wdened to nclude ndrect ownershp (1994) and other types of benefcal ownershp, such as convertble bonds, warrants, bonds wh warrants, and exchange bonds (1997). It also requred block holders to dsclosure the purpose of ther holdngs (2005)

11 we also ndcate whch countres are non-entrenched (NE) and whch are entrenched (E). A country s classfed as havng a strong tradon of nvestor actvsm (or non-entrenched for short) f at least two of the home country governance characterstcs are above medan. For Ireland and New Zealand, one of the ndcators s above the medan, a second ndcator s below the medan, and a thrd ndcator s mssng. We classfy both countres as non-entrenched. (No other countres are n the same suaton.) All other countres wh nformaton on actvsm ndcators are classfed as whout a strong tradon of actvsm (or entrenched for short). Countres whout such any nformaton on any of the actvsm ndcators are not classfed. We have also derved the frst prncpal component (PC) of the three ndcators. We can label those countres wh a PC value above the medan as non-entrenched and others as entrenched. Because the ndcators may have mssng nformaton for dfferent countres, ths can be done only for a smaller set of countres. In any case, for these countres, we fnd that the country labels on entrenched and nonentrenched are fully consstent wh the smpler methodology descrbed earler. That s, no country classfed as entrenched under one scheme would be re-classfed as non-entrenched under another scheme, and vce versa. In subsequent analyss, we adopt the country classfcaton scheme from the smpler methodology. Note that some funds whn a larger famly can be managed out of dfferent source countres. For example, whle Frankln Templeton Investments Corp. and Lonhart Investment Lmed are managed out of the Uned States, FTIF Templeton Asa Growth Fund and Lonhart (HK) Lmed are managed out of Hong Kong. Our dscusson wh some nternatonal funds suggests that nvestment and corporate engagement decsons are made by local offces rather than by headquarters. For ths reason, we classfy a fund s source country by the locaton of s management offce rather than by s headquarters locaton. In the last two columns, we show the number of block buyng events by all block holders and actvst block holders. The last two rows show the total number of countres and the total number of buyng events. We drop three events by tax heaven actvsts, whose domcles of management companes are unknown. (Domcles reported as England, Wales, and Scotland are all classfed as UK funds.) B. Frm level data Table 2 reports the summary statstcs of the key varables used n our nvestgaton of frm characterstcs of target frms durng Panel A uses the full sample, whle Panel B spls the sample nto those wh foregn block holders and those whout. Panel C spls the targets nto those wh actvsts and those whout. Panel D spls the targets nto those wh actvsts from non-entrenched countres and those whout. The last columns n Panels B, C, and D report the dfference-n-mean test

12 results. ***, **, and * respectvely ndcate that the dfference s statstcally sgnfcant at 1%, 5%, and 10% levels. Assets are n mllon Korean wons (whch are approxmately a thousand U.S. dollars). Cash ncludes cash equvalents. RPT s the sum of related-party sales and related-party purchases. Dvdend ncludes cash dvdends to common and preferred stockholders. Tobn s q s measured by (market value of equy plus book value of debt) dvded by book value of assets. Insde ownershp s percentage of shares held by the company s controllng shareholder and s related-partes. Accountng data are from TS2000, a database constructed by the Korean Lsted Companes Assocaton (KLCA). The stock market data are from DataGude. The dfference-n-mean t-tests results, shown n the last column of each panel, tell us that frms wh foregn block holders, on average, are more profable (hgher EBITDA/Assets), have a hgher Tobn s q, and are more lkely to have some foregn ownershp to start wh. Among the frms wh foregn block holders, those wh foregn actvsts have hgher profably, lower frm value, and hgher foregn ownershp than those whout them. Among the frms wh foregn actvsts, those wh actvsts from nonentrenched countres have lower cash holdngs, hgher leverage, lower profably, lower dvdend payouts, and lower frm value. 4. RESULTS Before we report our preferred specfcatons, we frst replcate some man specfcatons n Aggarwal et al. (2011) n secton 3.A. Interestngly, for our sample, the effect of foregn ownershp on frm-level corporate governance score s not reproduced here. A. Foregn block nvestor ownershp and corporate governance In ths subsecton, we employ the method used n Aggarwal et al. (2011) and run a set of smlar regressons usng Korean data. Table 3 shows the results. In Columns 1-4, we run frm fxed effects regresson of ln(corporate governance score) over a sample perod of In Columns 5-8, we run frst-dfference regressons of ln(corporate governance score) over a sample perod of Corporate governance scores are from Korea Corporate Governance Servce (CGS), whch has been announcng the score every year snce Our key regressors are foregn ownershp, foregn block ownershp, foregn actvst ownershp, and foregn non-actvst ownershp. Foregn block ownershp data are from Korea Exchange (KRX). The other three, however, had to be estmated usng the block purchase

13 and sell announcements data. As n Aggarwal et al. (2011), foregn ownershp varables are lagged by one year. Also note that we use the same set of control varables used n Aggarwal et al. (2011). The excepton s the ADR dummy, whch appears n Aggarwal et al. (2011), but omted here snce s fxed over tme durng our sample perod. Coeffcent standard errors are clustered at the frm level. Unlke Aggarwal et al. (2011), foregn ownershp has no explanatory power over corporate governance scores. The coeffcents on each of our foregn ownershp varables are statstcally nsgnfcant. (Snce Korean frms are not part of the sample used by Aggarwal et al., we cannot drectly compare our results to ther results on Korea.) One possble reason s that the corporate governance scores are partcular nosy for Korea, so ther varatons over tme are not correlated wh any observables. However, frm sze, sales growth, leverage, PP&E/Assets, and analyst coverage do seem to matter. So measurement errors of the governance score may not be a satsfactory explanaton. Ths motvates us to use alternatve approaches to nvestgate the role of actvst foregn block purchases. B. Prce reacton to block purchases by foregn nstutonal nvestors In ths subsecton, we turn to an event study method. For each block purchase event by foregn nstutonal nvestors, we compute cumulatve abnormal returns (CARs) estmated from a market model over a 190 tradng day perod [-200, -11] pror to the event wndow. We use KOSPI and KODAQ ndex returns as market returns respectvely for KOSPI and KOSDAQ member frms. Fgure 2 reports the averages of CAR[-10, x], where x = -10, -9,, 0, 1, 2,,10. Returns are n percentage terms. In Panel A of Fgure 2, we keep separate tracks of CARs, one for those nvolvng actvst nvestors (the top lne, 139 such announcements n total), and another for those that do not (the bottom lne, 1696 such announcements n total). Sold dots on the lnes ndcate statstcal sgnfcance at the 10% level (alternatve hypothess of average CAR > 0). There are two noteworthy features of the graph. Frst, the CAR lne nvolvng actvst nvestors les everywhere above the one nvolvng non-actvst nvestors. Panel B of Fgure 2 ndcates that the dfference between the two becomes statstcally sgnfcant at the 10% level startng from the day of announcement, and stays so at least for the next ten days (the dotted lnes ndcate the 90 percent confdence nterval). The dfference on day 10 s about 3-4%. Some of the posve CAR reacton to block purchase announcement reflects market recognon of possble stock pckng sklls of foregn nstutonal nvestors. If we assume that the stock pckng sklls are the same between actvst and non-actvst

14 nvestors, then the posve CAR reacton of actvst nvestors over that of non-actvst nvestors s an ndcaton of market s belef that actvsm wll pay off n the form of hgher frm value. Second, for both types of announcement, CARs become sgnfcantly posve startng fve or sx days before the actual announcement. Ths s probably so because actual purchases typcally take place a number of days before the date of dsclosure (at most 5 days for actvst and at most 10 days for nonactvst), and such purchases are known n the market. Alternatvely, f some of the stocks are not very lqud, large block purchases could push up the prces. Becht, Franks, and Grant (2014) fnd a smlar pattern n ther study on nternatonal hedge fund actvsm. We now explore the heterogeney across the source countres of actvst nvestors. We examne two proxes for nvestor actvsm (the frequency of hostle takeovers and the volume of mergers and acquson actves), and one proxy for the qualy of corporate governance (the ndependence of company boards from management). More precsely, ncdence of hostle takeovers refers to the percentage of domestc frms that experenced hostle takeover attempts durng , as reported by Ross and Volpn (2004). The M&A volume refers to percentage of domestc frms that were targets of completed M&A deals durng (Ross and Volpn, 2004). Board ndependence refers to the mean percentage of ndependent drectors (the number of ndependent drectors dvded by board sze), as reported by Dahya, Dmrov, and McConnell (2008). The calculatons of the frst two measures predate our sample perod. Our mantaned assumpton s that these features reflect the tradon of local nstutonal nvestors, local legal envronment, or ther combnatons, and persst over tme. 5 The nformaton s summarzed n Table 1. The medan values for the three ndcators are 1.37% for hostle takeover frequency, 34.09% for M&A volume, and 53.15% for board ndependence. Some of the measures are not avalable for some source countres n our sample. For those for whch the nformaton s avalable, we use H and L to respectvely denote whether a country scores above the medan or below. For example, for Australa, all three measures score above the medan, suggestng that the country has a hgher than medan amount of nvestor actvsm and a hgher than medan level of board ndependence. For Denmark, on the other hand, all three ndcators score below the medans, ndcatng a lower than medan degree of nvestor actvsm and lower than medan level of board ndependence. We also aggregate the nformaton from the three measures nto a sngle dmensonal bnary measure. A country s classfed as havng a strong tradon of nvestor actvsm (or non-entrenched for short) f at least two of the home country governance characterstcs are above medan. For Ireland and New Zealand, one of the ndcators s above the medan, a second ndcator s below the medan, and a thrd 5 Capro, Laeven and Levne (2007) used the same M&A actvy volumes as an ndcator of nvestor actvsm; Alexandrds, Petmezas, and Travlos (2010) have used the same ndcator of board ndependence n ther study

15 ndcator s mssng. We classfy both countres as non-entrenched. (No other countres are n the same suaton.) All other countres wh nformaton on actvsm ndcators are classfed as whout a strong tradon of actvsm (or entrenched for short). Countres whout such any nformaton on any of the actvsm ndcators are not classfed. Besde country names n Table 1, we ndcate whch countres are non-entrenched (NE) and whch are entrenched (E). In Column 7, we report, for each country, the total number of block purchase announcements by nstutonal nvestors of any type over In the last column, we report, for each country, the number of block purchase announcements from self-declared actvst nvestors snce January Clearly, the latter s a small subset of the former. We now examne how the CAR reactons vary by the type of source countres the actvst nvestors come from. In Table 4, we organze the results n four row panels, correspondng to the four ways of sortng source countres: frequency of hostle takeovers, M&A volumes, degree of board ndependence, and a compose measure of non-entrenchment. For each of them, we compute CARs over three event wndows, [-1, +1], [-5, +5], and [-10, +10]. In the frst two columns, we report CARs for actvst nvestors from non-entrenched countres and those from entrenched countres (measured n four ways). Generally speakng, the CARs are always posve and statstcally sgnfcant for non-entrenched country actvst nvestors and rse from the narrow to wder wndows. On the other hand, the CARs for entrenched country actvst nvestors are only sgnfcant for the narrowest wndow of [-1, +1] but are ndfferent from zero n wder wndows. In Column 3, we report the dfference n CARs between these two groups of nvestors. The dfference s always posve for all three event wndows and are statstcally sgnfcant for [-5, +5] and [- 10, +10]. For example, usng the compose measure of entrenchment (the last panel), the CARs for nonentrenched country actvsts are 4.8 percentage ponts hgher over [-5, +5] and 7.2 percentage ponts hgher over [-10, +10]. Because US nvestors domnate our sample and are sometmes consdered specal n the lerature, we want to make sure that our results are not drven by them. We re-compute CARs for nvestors from non-entrenched countres after excludng US as a source country and report the results n Column 4. Interestngly, the CARs excludng the Uned States as a part of the non-entrenched group tend to be hgher. Indeed, n Column 5, when we compute the excess CARs of non-entrenched countres over those of entrenched countres, we fnd a bgger dfference when US s excluded 6. 6 One possble explanaton s that some of the US nvestors that mght consder themselves as non-actvsts f they had come from other countres are used to consder themselves as actvsts n the Uned States. We thank We Jang for ths suggeston

16 In Fgure 3, we present a graphc summary of the results usng the compose measure of nonentrenchment. As one can see, whle the CARs for nvestors from entrenched countres are generally nsgnfcantly dfferent from zero, except for the wndow of [-10, +1], those for nvestors from nonentrenched countres are sgnfcant and rse wh the wdth of event wndows. Indeed, the dfference between the two s clearly sgnfcant from day zero onwards. If we exclude the Uned States as a source country, the dfference n CARs between the two groups becomes even greater. In unreported analyses, we also examne the share prce reactons around the tme of block purchase made by domestc nvestors. Frst, we fnd that share prce reacts more strongly for domestc actvsts than for domestc non-actvsts. But the dfference between the two loses statstcal sgnfcance as we move from a short to longer event wndows, makng hard to determne true value creaton by domestc actvsm. Second, the share prce reacton prompted by domestc actvsts s smaller than that by foregn actvsts, 3.1% and 4.2%, respectvely, over an event wndow of [-10, +10]. If we compare the two after nettng out the share prce reactons prompted by ther respectve non-actvst block holders, the dfference wdens from 1.1% to 2.3% (3.67% for foregn actvst versus 1.37% for domestc actvsts). Ths dfference s statstcally sgnfcant. Overall, these patterns are consstent wh the noton that foregn actvst nvestors are bolder and more aggressve than domestc nvestors n pressng frm management to change fnancal and governance practces that could enhance shareholder values. We have pursued a number of addonal extensons that explore target frm heterogeney. Frst, when we separate target frms nto large versus small ones based on market capalzaton, we fnd that the CAR response s stronger for smaller frms than for larger frms (when block purchases are made by actvst nvestors from non-entrenched countres.) Ths may reflect the possbly that smaller frms corporate governance and fnancal polces are more lkely to be nfluenced by foregn actvst nvestors. However, the dfference n the CAR reactons between the two types of target frms s not statstcally sgnfcant. Second, we look nto potentally dfferent CAR reactons for Chaebol and non-chaebol frms as targets of block purchases. Chaebol frms are members of large conglomerates that are typcally controlled by famles through elaborate cross shareholdng or pyramds (see Km, Sung, and We, 2011, for examples and ther effects on foregn nvestments). We fnd that the CAR response s generally stronger when target frms are not afflated wh a Chaebol. Ths may reflect the possbly that foregn nvestor actvsm s more lkely to be effectve when targets are non-chaebol frms. Unfortunately, the dfference n the CARs s not statstcally sgnfcantly dfferent from zero between these two types of target frms. These extensons are not reported n table format to save space

17 C. Do foregn actvst purchases rase frm values? Beyond the prce reactons to announcements of actvst purchases over relatvely short event wndows, we now check f these prce reactons are confrmed by changes n frm value over the subsequent years. Ths check s useful for another reason: sometmes nvestor actvsm s crczed as actons that only lead to a temporary ncrease n the stock prce that may be reversed n the long run (see, for example, Kahan and Rock, 2007). Our analyss wll shed lght on whether typcal nvestor engagement produces only a short-term change or a longer-lastng mprovement n frm value. We examne ths queston by lookng at how a frm s Tobn s q may be affected by the share purchase of foregn actvst nvestors from non-entrenched countres. We start wh a naïve specfcaton that gnores endogenous selecton of frms targeted by foregn actvst nvestors. We then address ths endogeney va two ways: propensy score matchng and Heckman selecton equaton. Inal naïve specfcaton We frst estmate Equaton (1) for two sub-samples, one consstng of frms targeted by actvsts from non-entrenched countres durng (treatment group) and another consstng of frms targeted by entrenched-country actvsts durng (control group). If a frm s targeted by a non-entrenched country actvst and also by an entrenched country actvst, we assgn these frms to be n the treatment group n = α + α0t0 + α1t1 + α2t 2+ α3t3 + λt + t= 2004 j= 1 ln(tq) δ + ε j (1) ln( TQ) s Tobn s q n natural logarhm of frm n year t. T 0 s a bnary varable takng a value of 1 f year t s the year of block buyng announcement by non-entrenched country actvst for frm, and 0 otherwse. T 1 s bnary varable takng a value of 1 f year t s one year after the block buyng announcement by non-entrenched country actvst for frm, and 0 otherwse. T 2 and T 3 are smlarly defned. Coeffcentα captures the frm value before the treatment, and coeffcentsα 0, α 1, α 2, andα 3 respectvely captures the ncrease n frm value durng [-1, 0], [-1, +1], [-1, +2], and [-1, +3]. λ t s a calendar year dummy, takng a value of 1 for year t, and 0 otherwse. Snce we use a sample perod from 2004 to 2012, we nclude eght year dummes. δ j s an ndustry dummy (11 ndustres from 2-dg Korea

18 SIC code), takng a value of 1 for ndustry j, and 0 otherwse. When estmatng the coeffcents, we use standard errors clustered at the frm level. If a frm s subject to multple buyng announcements by the same non-entrenched country actvst n dfferent years (or buyng announcements by dfferent actvsts n dfferent years), T 0 equals 1 n the year when there was the frst buyng announcement. Ths rule s also appled to targets of entrenched country actvsts. To preserve sample sze, we keep the frms even f there were sellng announcements made by foregn block holders durng our sample perod. We also estmate Equaton (2), whch s estmated usng the full sample that pools the two subsamples together and use nteracton terms between the number of years followng the block purchases and a dummy for treatment group frms. ln(tq) = α + α T βd + β T 0 + α T α T 1 2 D + β T α T D + β T 2 D + β T 3 3 D λ t t = λ D t t = n δ j j = 1 + ε ` (2) D s a bnary treatment dummy takng a value of 1 f frm s targeted by a non-entrenched country actvst, and 0 otherwse. 7 Notce that we also nclude the nteractons between the treatment dummy ( D ) and calendar year dummes. Snce Equaton (2) s bascally a dfference-n-dfferences (DD) equaton, coeffcents β 0, β 1, β 2, and β 3 are DD estmators, and they capture how the dfference n frm value between the treatment group and the control group change before and after the treatment. Coeffcent β captures the dfference n frm value between the treatment group and the control group before the treatment. Agan, the coeffcent standard errors are clustered at the frm level. Columns 1-3 n Table 5 report our results. We see that frm value n the treatment group rses on the year of block purchase and then rses further n subsequent years (Column 1), whereas the frm value n the control group actually experences a declne on the year of block purchase and then falls further n subsequent years (Column 2). The DD regresson n Column 3 confrms the prevous result: the posve frm valuaton effect comes exclusvely from block purchases by non-entrenched country actvsts. The coeffcent of 0.34 (specfcally, ) on the nteracton term between the treatment group dummy ( D ) 7 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We dd not fnd any statstcally sgnfcant dfference between the treatment and the control groups

19 and the treatment year dummy ( T 0 of control group frms, ncreases by 34.35% on the year of treatment. ) suggests that the frm value of treatment group frms, relatve to that Accountng for endogenous selectons We recognze that share purchases by non-entrenched country actvsts are not random. The purposeful purchases have to be taken nto account. We address ths endogeney of treatment n two ways. Frst, we use propensy score matchng (PSM), where matchng frms are dentfed by the nearest neghbor method among the targets of entrenched actvst block holders n the same year block buyng announcement took place by non-entrenched country actvsts. Second, we apply a Heckman selecton specfcaton to account for endogenous treatment. Whle the propensy score method produces a control group that s as smlar as possble to the treatment group based on observable frm characterstcs, the Heckman method also takes nto account potentally unobservable frm characterstcs. The dentfcaton of the Heckman method, however, depends on eher havng exogenous determnants of frm selecton or the dstrbutonal assumpton on the error term of the selecton equaton. We vew the two methods as provdng complementary nformaton for our nference. Propensy scores are estmated by a prob regresson where dependent varable s the probably of a frm beng targeted by a non-entrenched country foregn actvst and rght-hand sde varables (lagged by one year) ncludng frm sze, cash holdngs, leverage, equy holdngs, related-party transactons, cash flows, dvdend payout rato, frm value, nternal ownershp, and foregn ownershp. Column 3 n Table 6 shows the result. Cash holdngs, equy holdngs, and frm value seem to matter. The lkelhood of beng a target ncreases wh cash holdngs or equy holdngs. Ths s consstent wh the noton that holdng excess cash or havng engaged n excessve acqusons n the past s a sgn of poor corporate governance. We also fnd that the lkelhood of beng a target drops wh frm value. Ths may reflect foregn actvsts belef that lower-than-average Tobn s q ndcates suboptmal corporate governance that can be corrected by actvsm. We use the same set of rght-hand varables when estmatng the frst-stage of Heckman s (1979) two-stage procedure. For detals of ths procedure, see Campa, Manuel, and Keda (2002). Columns 4-6 n Table 5 report our PSM results. Column 4 reports the result for the treatment group, Column 5 for the matched control group, and Column 6 for the combned sample of treated and nontreated matched frms. 8 From Column 5, one can see that the negatve valuaton effect of the control 8 The results n Column 1 and Column 4 are slghtly dfferent because we use data n year t-1 to dentfy matched frms. If any varable n the frst stage prob s mssng n year t-1, ths observaton s dropped

20 group dsappeared. That s, there s a posve ncrease n frm value for the treatment group, but no changes n frm valuaton for the control group. The coeffcents on the nteracton terms n Column 6 are weaker than those n Column 3. But they are all posve and margnally sgnfcant n year t+3. That s, the change n frm value durng [-1, +3] s greater n frms purchased by non-entrenched country actvsts than those purchased by entrenched country actvsts. The economc magnude of the nteracton effect s more sensble under the PSM method than under OLS. In the OLS regresson reported n Column 3, Tobn s q ncreases more for actvst nvestors from non-entrenched countres than those from entrenched ones by 34% n the year of purchase, followed by addonal ncreases of 61%, 87%, and 93% n subsequent years. These seem to be too large. Wh propensy score matchng, the nteracton terms produce more modest results n economc terms. In partcular, accordng to the nteracton terms n Column 6 of Table 5, Tobn s q stll ncreases more for actvst nvestors from non-entrenched countres than for those from entrenched countres, but not sgnfcantly so n the year of purchase, by 25% n the year after the purchase (sgnfcant at the 10% level), 24% two years later (not sgnfcant), and 32% three years later (sgnfcant at the 10%). These estmates are plausble relatve to related estmates n the lerature. For example, n a study of the Lazard Korea Corporate Governance Fund s engagement wh two Korean companes (Taekwang Industral and Daehan Synthetc Fber), Lee and Park (2009) fnd that the CAR responses over [-1 day, +5 days] are 53.5% and 80.3%, respectvely. In a study of the 1999 corporate governance reform n Korea, Black, Jang, and Km (2006) report that an mprovement n the corporate governance score from the worst to the best scenaro over a seven-month perod leads to an ncrease n Tobn s q by In Column 7 of Table 5, wh the Heckman selecton specfcaton, we agan confrm that the posve valuaton effect s mostly a feature for block purchases by non-entrenched country actvst nvestors. The nteracton terms are all posve and statstcally sgnfcant. To summarze, followng the evoluton of a frm s Tobn s q n the years after block purchases, we fnd that frm valuaton s enhanced by non-entrenched country foregn actvsts. We also confrm that these results are unlkely to be drven by the endogenous choce of targets by foregn actvsts. D. Corporate fnancal polces and governance practces We now nvestgate whether target frms of non-entrenched country actvsts experence real changes n corporate fnancal polces or n governance practces n subsequent years that can justfy the jump n

21 share prce n our event studes and also the jump n frm value n subsequent years. Tables 7, 8, 9, 10, and 11 respectvely report the results for () cash holdngs, () leverage, () dvdend payouts, (v) stock repurchases, and (v) corporate governance. As for cash holdngs, leverage, and corporate governance, we use the same regresson specfcatons used n the prevous subsecton for frm value, except that we now replace our dependent varable to cash holdngs, leverage, or corporate governance. In case of dvdend payouts and stock repurchases, we modfy the specfcaton to capture how ther responsveness to earnngs changed wh the treatment. In Tables 7 and 8, we nvestgate f the composon of target frm s balance sheets s nfluenced by actvsts from non-entrenched countres. We look nto cash holdngs n Table 7 and leverage n Table 8. From Columns 1 and 2 of Table 7, one can see that the cash holdngs (cash plus cash equvalents scaled by total assets) of the frms n the treatment group fall n the year of block purchase (7.21% of total assets) and then fall further n subsequent years (23.67% of total assets durng -1 and +3 years), whereas the cash holdngs of the frms n the control group show no change n the year of block purchase nor n subsequent years. Ths fndng suggests that the jump n share prce n our events and the jump n frm value n subsequent years can be related, n part, to target frms reducng dle cash. Column 3 shows that the fall n cash holdngs n treatment group frms relatve to that n control group frms are statstcally sgnfcant n the year of block purchase and also n all subsequent years. The results from both propensy score matchng (Columns 4-6) and Heckman s two-stage procedure (Column 7) confrm that our fndng s robust to endogeney concerns. 9 From Columns 1 and 2 of Table 8, we see that the leverage (debt/assets) of treatment group frms rses n the year of block purchase (11.81% of total assets) and then rses further n subsequent years (34.38% of total assets durng -1 and +3 years), whereas the leverage of control group frms show no change n the year of block purchase nor n subsequent years. Ths fndng suggests that the jump n share prce n our event studes and the jump n frm value n subsequent years can be related, n part, to target frms ncreasng leverage, possbly by dscplnng mangers from wastng free cash flows (Jensen, 1988). Column 3 shows that the rse n leverage n treatment group frms relatve to that n control group frms are statstcally sgnfcant n the year of block purchase and also on all subsequent years. The results from 9 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We fnd smlar but somewhat weaker results

22 both propensy score matchng (Columns 4-6) and Heckman s two-stage procedure (Column 7) confrm that our fndng s robust to endogeney concerns. 10 In Tables 9 and 10, we nvestgate f target frms dvdend dstrbuton polcy s nfluenced by actvsts from non-entrenched countres. More precsely, we expect dvdend payout to be lnked to a frm s profably n general, and we examne ths lnkage s strengthened followng block purchases by actvst foregn nvestors. We do so by augmentng the prevous specfcaton wh some nteracton terms. Specfcally, we estmate Equaton (3) below usng Tob, wh the constrant that dvdend payout rato has to be non-negatve. 11 DIV * = α + α T + γe γ T 0 + α T 0 + πe D + π T 0 E γ T + α T E 2 E D + π T 1 + α T γ T 2 + βd + β T E 2 + γ T 3 E D + π T E 0 D + β T E D + π T E D D + β T 2 2 D + β T 3 3 D λ t t = λ D t t = n j = 1 δ + ε j (3) DIV DIV = DIV = 0 * for DIV for DIV * * > 0 0 (4) measures the dvdend payout rato (cash dvdend/net ncome) of frm n year t. For frms where cash dvdend s greater than net ncome and for frms where net ncome s negatve, but cash dvdends are posve, we let ther dvdend payout rato to have a value of 1. E measures EBITDA over total assets for frm n year t. All other varables are same as n Equaton (2). The coeffcents of our nterest are π 0, π 1, π 2, and π 3, whch tell us whether the dvdend payout s responsveness to earnngs n treatment group frms ncrease relatve to that n control group frms n the year of block purchase and also n all subsequent years. We expect these coeffcents to be posve and statstcally sgnfcant. Table 9 reports our results. From Columns 1 and 2, one can see that the dvdend payouts of treatment group frms become more responsve to earnngs n the year of block purchase and then stay at a smlar level n subsequent years, whereas the dvdend payouts of control group frms show no change n the year of block purchase nor n subsequent years. Ths fndng suggests that the jump n share prce n our 10 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We fnd smlar but somewhat weaker results. 11 We dd not fnd any sgnfcant changes n the level of dvdend payout per se after the treatment

23 event studes and the jump n frm value n subsequent years can be related to target frms ncreased responsveness of dvdend payouts to earnngs. Column 3 shows that the ncrease n responsveness n treatment group frms relatve to that n control group frms are statstcally sgnfcant n the year of block purchase and also n all subsequent years. The results from propensy score matchng (Columns 4-6) confrm that our fndng s robust to endogeney concerns. 12 We next look nto the probably of stock repurchase and see f ths becomes more responsve to earnngs after the treatment. Specfcally, we estmate the followng prob regresson. 13 Z = α + α T + γe γ T 0 + α T 0 + πe D + π T 0 1 E γ T + α T E 2 + α T E D + π T γ T 2 + βd + β T E + γ T 2 3 E D + π T E 0 D + β T E D + π T E D + β T D 2 2 D + β T 3 3 D (5) λ t t = λ D t t = n j = 1 δ + ε j P = F(Z ) Engage n stock repurchase f Do not engage n stock repurchase f Z Z > Z * Z * (6) Z s the latent ndex varable that determnes the lkelhood of engagng n stock repurchase and Z * s the crcal cutoff value, whch translates the underlyng ndex nto a stock repurchase decson. All other varables are same as n Equaton (3). The coeffcents of our nterest are π 0, π 1, π 2, and π 3, whch tell us whether the stock repurchases responsveness to earnngs n treatment group frms ncrease relatve to that n control group frms n the year of block purchase and also n all subsequent years. We expect these coeffcents to be posve and statstcally sgnfcant. Table 10 reports our results. Coeffcents are for the latent ndex varable Z. From Columns 1 and 2, one can see that the stock repurchase of the treatment group frms becomes more responsve to earnngs n the year of block purchase and then grows further n subsequent years, whereas the stock repurchase of control group frms show no sgnfcant change n the year of block purchase nor n subsequent years. Ths fndng suggests that the jump n share prce n our event studes and the jump n frm value n subsequent years can be related to target frms ncreased responsveness of stock repurchase to earnngs. 12 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We fnd smlar but somewhat weaker results. 13 We dd not fnd any sgnfcant changes n the probably of stock repurchase per se after the treatment

24 Column 3 shows that the ncrease n responsveness n treatment group frms relatve to that n control group frms are margnally sgnfcant over a four-year perod [-1, +3]. The results from propensy score matchng (Columns 4-6) confrm that our fndng s robust to endogeney concerns. 14 In Table 11, we nvestgate f an overall measure of corporate governance n target frms s nfluenced by actvsts from non-entrenched countres. Here we restrct our sample to KOSPI frms, of whch we have ther corporate governance scores (CGS). These scores, from Korea Corporate Governance Servce (CGS), takes a value from 0 to 100 and are composed of fve sub-ndces (each takng a value between 0 and 20), ncludng shareholder rghts, board, dsclosure, aud, and dstrbuton. From Columns 1 and 2 of Table 11, one can see that the CGS (n natural logarhm) of treatment group frms and control group frms do not show sgnfcant changes n the year of block purchase and n subsequent years. But, Column 3 shows that the rse n CGS n treatment group frms relatve to that n control group frms are margnally sgnfcant n the year of block purchase and statstcally sgnfcant n all subsequent years. The coeffcent of on the nteracton term between the treatment group dummy ( D ) and the +3 treatment year dummy ( T 3 ) suggests that CGS of treatment group frms, relatve to that of control group frms, ncreases by 92.49%. The results from propensy score matchng (Columns 4-6), however, are weaker. The coeffcent on the nteracton terms s posve but not statstcally sgnfcant. The results from Heckman s two-stage procedure (Column 7), on the other hand, confrm that our fndngs n Columns 1-3 are robust to endogeney concerns. 15 These fndngs suggest that the jump n share prce n our event studes and the jump n frm value n subsequent years can be related to target frms mprovng ther corporate governance practces. Recall that, n our survey of nternatonal funds that are members of the ACGA, many respondents perceve ncreasng scrutny of related party transactons, mprovng board ndependence, and lnkng CEO turnover more to performance, as some of the most mportant tools of engagement. We have checked for evdence of these but faled to fnd statstcally sgnfcant results. Ths could be so because eher these varables are not as well measured, these tools are used relatvely less frequently n Korea than n other countres, or nvestor percepton on ths queston s not on the mark. It wll be an 14 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We fnd smlar results. 15 In unreported analyss, we run a smlar set of regressons usng a treatment group dummy, whch takes a value of 1 f frm s targeted by an actvst (regardless of s source country), and 0 otherwse. The control group ncludes frms purchased by non-actvst foregn block holders. We do not fnd any dfference between these two groups

25 nterestng topc for future research to ascertan the reasons for the dscrepancy between nvestor percepton and statstcal tests on ths queston. E. Robustness checks We perform a number of addonal robustness checks. Frst, nstead of usng the entre set of block buyng announcements, we lm the sample to a subset of frst announcements. That s, for a gven frm and a gven block holder, we lm the sample to the frst 5% block buyng announcement and exclude all subsequent announcements. Ths s to allow for the possbly that the frst announcement contans more news about nvestor actvsm than subsequent ones. In unreported analyss, we fnd that all of our results eher are strengthened or reman ntact. For example, the CAR over [-10, +10] s 8% for actvsts and 12% for non-entrenched country actvsts. Second, followng Km, Km, and Kwon (2009), we lm our sample to swchng events, where passve block holders change the declared purpose of ther block holdngs from passve to actve engagement whout changng the number of shares they hold. Ths approach has the advantage of rulng out the effects of market pressure or stock pckng sklls. In unreported analyss, we fnd that the CAR over [-10, +10] s 6-7%. Unfortunately, gven the small number of such events (only 8), CARs are not sgnfcantly dfferent from zero. Thrd, to rule out the possbly that our results may be drven by a country effect that has nothng to do wh actvsm or governance, we conduct a falsfcaton test by comparng non-actvsts from nonentrenched countres and non-actvsts from entrenched countres. If stock prces also respond more strongly to the announcements fled by the former than the latter, one mght wonder f our key results reported earler are spurous. In unreported analyss, we fnd that none of the two exhb CARs that are sgnfcantly greater than zero. Thus, our results about actvst nvestors from non-entrenched countres are unlkely to reflect only a pure country effect. Fourth, engagement or actvsm polces may be set at the company-level by the headquarters and not at the ndvdual fund level. A tougher test s to see whether fund locaton stll matters even whn the same company. Two nvestment management companes n our sample have funds located across two dfferent types of source countres. In the case of Frankln Templeton (FT), we fnd one fund (Frankln Templeton Investments Corp.) located n Canada (a non-entrenched country) and four funds (Templeton Asset Management Ltd., Frankln Mutual Advsors, L.L.C., FTIF Templeton Asa Growth Fund, and FTIF Templeton Korea Fund) are located eher n Sngapore or n Hong Kong (entrenched countres). In

26 the case of Lonhart, two funds (Lonhart Asa Master Fund and Lonhart Investment Lmed) are located n U.K. (a non-entrenched country) and one fund (Lonhart Hong Kong Lmed) s located n Hong Kong (entrenched country). In unreported analyss, we fnd that the announcements fled by funds located n non-entrenched countres trgger a stronger posve stock prce reacton than those fled by funds that are supervsed by the same headquarters but located n entrenched countres. However, the small sample sze yelds relatvely large standard errors; so we do not fnd statstcal dfference between the two stock prce reactons. Lastly, we conduct event studes around block buyng announcements fled by proxy actvsts and non-proxy actvsts. The former nclude foregn actvst block holders who nate a proxy fght durng our sample perod. 16 In unreported analyss, we fnd that stock prce reacton s stronger for announcements (n = 30) fled by proxy actvsts (6-7% CAR over [-10, +10]) than those (n = 103) by nonproxy actvsts (2-3% CAR over [-10, +10]). The dfference s statstcally dfferent n narrow wndows. Even f we lm the events to those fled by actvsts from non-entrenched countres, the dfference between proxy actvsts and non-proxy actvsts exsts. Gven the small sample sze, however, the dfference loses statstcal sgnfcance. 5. CONCLUSION In ths paper, we nvestgate the possbly that openness to foregn portfolo nvestors could be a mechansm for transmtng better corporate governance practces to host country frms, and thereby enhancng shareholder value. We frst apply an event study approach to examne the announcements of block purchases by foregn nstutonal nvestors n Korea. We fnd that stock prces rse on average, but only when foregn nstutonal nvestors declare themselves as actvsts. Ths means pure portfolo nvestment does not generate sgnfcant enhancement of frm value n emergng market economes. It s foregn actvst nvestors, or a subset of portfolo nvestors that are closest to foregn drect nvestors, that are regarded by the market as potentally enhancng frm value. We fnd that source country dentes also matter n an mportant way. The posve stock prce reactons are especally pronounced when the actvst nvestors are comng from source countres wh a strong tradon of nvestor actvsm. 16 We found 16 events durng our sample perod (eght foregn nvestors aganst 16 target companes). Eght nvestors nclude Lazard Asset Management (9 targets), Cavendsh Square Holdng B.V. (1 target), Consoldated Scence Corp. (1 target), Deccan Value Advsors (1 target), Icahn Partners Master Fund (wh Steel Partners II, 1 target), SC Asan Opportuny Fund (1 target), Soveregn Asset Management (wh Crest Secures, 1 target), and Teton Capal Partners (1 target)

27 We also examne corporate fnancal polces and governance practces of target frms. We fnd that target frms are more lkely to reduce cash holdngs, rase leverage ratos, mprove overall corporate governance, and mprove dvdend payout s (and stock repurchases ) responsveness to earnngs f foregn actvsts are from countres wh a strong tradon of nvestor actvsm. These fndngs are consstent wh the nterpretaton that foregn actvst nvestors from non-entrenched countres are regarded as enhancng value precsely because they engage frm management effectvely to cause the latter to modfy governance practces or fnancal polces. The dffuson of corporate governance s one potental benef to developng countres of capal account openness to cross border portfolo nvestment. Snce ths benef s not usually featured n the dscussons on the costs and benefs of capal account openness (see Kose et al., 2010, as an example), developng a deep understandng of ths pont s mportant. Whle both our econometrc results and market partcpants agree on the value enhancng potental of shareholder actvsm by nternatonal nvestors of certan types, there s some dvergence on the exact mechansms that are most effectve. We do not have defnve nsght on the reasons for the dvergence, and leave for future research. REFERENCES Aggarwal, Reena, Isl Erel, Mguel Ferrera, and Pedro Matos (2011), Does governance travel around the world? Evdence from nstutonal nvestors, Journal of Fnancal Economcs 100 (1), pp A, Chunrong and Edward C. Norton (2003), Interacton terms n log and prob models, Economc Letters 80, pp Albuquerque, Ru, Lus Brando-Marques, Mguel Ferrera, and Pedro Matos (2013), Internatonal Corporate Governance Spllovers: Evdence from Cross-Border Mergers and Acqusons, IMF Workng Paper No. WP/13/2013 Alexandrds, George, Dmrs Petmezas, and Nckolaos G. Travlos (2010), "Gans from mergers and acqusons around the world: New evdence," Fnancal Management 39 (4), pp Bae, Kee-Hong, Jun-Koo Kang, and Jn-Mo Km (2002), Tunnelng or value added? Evdence from mergers by Korean busness groups, Journal of Fnance 57 (6), pp Baek, Jae-Seung, Jun-Koo Kang, and Inmoo Lee (2006), Busness groups and tunnelng: Evdence from prvate secures offerngs by Korean Chaebols, Journal of Fnance 61 (5), pp Bekaert, Geert, and Campbell Harvey (2000), Capal flows and the behavor of emergng market equy returns, Capal Flows and the Emergng Economes: Theory, Evdence, and Controverses, eded by Sebastan Edwards, Natonal Bureau of Economc Research, pp

28 Becht, Marco, Julan Franks, and Jeremy Grant (2014), "The Returns to Hedge Fund Actvsm: An Internatonal Study." European Corporate Governance Instute (ECGI)-Fnance Workng Paper 402 Black, Bernard S., Hasung Jang, and Woochan Km (2006), Does corporate governance predct frms market values? Evdence from Korea, Journal of Law, Economcs, and Organzaton 22 (2), pp Black, Bernard S., Woochan Km, Hasung Jang, and Kyung-Suh Park (2008), How corporate governance affects frm value: Evdence from Korea on a related-party transactons channel, KDI School Research Paper No. 08/19 Brs, Artuaro, Nel Brsley, and Chrstos Cabols (2008), Adoptng Better Corporate Governance: Evdence from Cross-Border Mergers, Journal of Corporate Fnance 14 (3), pp Brs, Artuaro and Chrstos Cabols (2008), The Value of Investor Protecton: Frm Evdence from Cross- Border Mergers, Revew of Fnancal Studes 21 (2), pp BusnessWeek (2006), Specal Report: Fdely s Dvded Loyaltes, October 16. Campa, Jose Manuel, and Sm Keda. (2002), "Explanng the dversfcaton dscount," Journal of Fnance 57 (4), pp Capro, Gerard, Luc Laeven, and Ross Levne (2007), Governance and Bank Valuaton, Journal of Fnancal Intermedaton 16 (4), pp Char, Anusha, and Peter Henry (2004), "Rsk sharng and asset prces: Evdence from a natural experment," Journal of Fnance 59 (3), pp Dahya, Jay, Orln Dmrov, and John J. McConnell (2008), "Domnant shareholders, corporate boards, and corporate value: A cross-country analyss," Journal of Fnancal Economcs 87 (1), pp Ferrera, Mguel A., and Pedro Matos (2008), "The colors of nvestors money: The role of nstutonal nvestors around the world," Journal of Fnancal Economcs 88 (3), pp Heckman, James J. (1979), "Sample selecton bas as a specfcaton error," Econometrca 47 (1), pp Jensen, Mchael C. (1988), Agency costs of free cash flow, corporate fnance, and the market for takeovers. Amercan Economc Revew 76 (2), pp Kahan, Marcel, and Edward B. Rock (2007), Hedge Funds n Corporate Governance and Corporate Control, Unversy of Pennsylvana Law Revew 155 (5), pp Km, Woochan, Woojn Km, and Kap-Seok Kwon (2009), Value of outsde blockholder actvsm: Evdence from the swchers, Journal of Corporate Fnance 15 (4), pp Km, Woochan, Taeyoon Sung, and Shang-Jn We (2011), Does corporate governance rsk at home affect nvestment choces abroad? Journal of Internatonal Economcs 85 (1), pp

29 Kose, Ayhan, Eswar Prasad, Kenneth Rogoff, and Shang-Jn We, 2010, Fnancal Globalzaton and Economc Polces, Handbook of Development Economcs, Vol. 5, eded by Dan Rodrk and Mark Rosenzweg, North-Holland, 2010, pp Lee, Dong Wook and Kyung-Suh Park (2009), Does nstutonal actvsm ncrease shareholder wealth? Evdence from spllovers on non-target companes, Journal of Corporate Fnance, 15 (4), pp Martynova, Marna and Luc Renneboog (2008), Spllover of Corporate Governance Standards n Cross- Border Mergers and Acqusons, Journal of Corporate Fnance 14 (3), pp Ross, Stefano, and Paolo F. Volpn (2004), "Cross-country determnants of mergers and acqusons," Journal of Fnancal Economcs 74 (2), pp Stulz, René M. (2005), "The lms of fnancal globalzaton." Journal of Fnance 60 (4), pp We, Shang-Jn (2001), Domestc Crony Capalsm and Internatonal Fckle Capal: Is There a Connecton? Internatonal Fnance, 4 (1), pp We, Shang-Jn (2006), "Connectng two vews on fnancal globalzaton: Can we make further progress?" Journal of the Japanese and Internatonal Economes, 20 (4), pp

30 Fgure 1: Dstrbuton of foregn block buyng announcements Dstrbuton of foregn block buyng announcements (not lmed frst 5% block holdng announcement) from 1998 to 2009 (12 years). Samples nclude block buyng announcements of target frms lsted eher on the Korea Stock Exchange (KSE) or on KOSDAQ (lsted on KRX after the two exchanges merged nto KRX n 2005). As a way of capturng foregn portfolo nvestment, but not foregn drect nvestment, we exclude block buyng announcements by non-fnancal nstutons. We also exclude announcements where pre- or post-announcement block ownershp s greater than 20%. Fgure A covers all announcements (snce 2005, each bar s spl between actvst vs. non-actvst), whle Fgure B lms the sample to announcements made snce 2005 by actvsts (each bar spl between entrenched vs. non-entrenched home countres). A. All announcements B. Announcements by actvsts

31 Fgure 2: Cumulatve abnormal returns (CAR) around block buyng announcements by foregners Averages of CAR [-10, x] over x, where x = -10,, 10. Returns are n percentage terms. Abnormal returns are estmated from market model usng past 190 tradng days from day 200 to 11. KOSPI return s used as the market return for stocks lsted on the Korea Stock Exchange (KSE), whle KODAQ compose return s used as the market return for stocks lsted on the KOSDAQ market. From the sample n Fgure 1, we exclude announcements that are eher followed by or follows another foregn block holdng announcement whn 10 tradng days. Fgure A shows the averages of CAR for foregn actvsts and non-actvsts. Foregn actvsts are foregn block holders that declared ther actvst ntent. Sold symbols (as opposed to empty symbols) ndcate that average CAR s sgnfcantly dfferent from 0 at the 10% level. No. of announcements by actvsts = 139, no. of announcement by non-actvsts = 1,696. Fgure B shows the dfferencee n market reactons. The dotted lnes report the 90 percent confdence ntervals. A. Actvst vs. non-actvsts B. Dfference n market reactons

32 Fgure 3: Cumulatve abnormal returns (CARs) around foregn actvsts block buyng announcements Market reactons on block buyng announcements by foregn actvsts from non-entrenched vs. entrenched countres. Averages of CAR are calculated n the same manner as n Fgure 2. Foregn actvsts from non-entrenched countres are foregn block holders that declared ther actvst ntent and from countres wh eher a hostle takeover market, actve mergers & acqusons, or ndependent boards. Fgure A compares targets of actvsts from non-entrenched countres (75 f ncludng US actvsts and 24 f not) aganst those of actvsts from entrenched countres (60). Fgure B and C report dfference n market reactons. (Dark/sold dots ndcate statcally sgnfcant at the 10% level.) A. Non-entrenched vs. Entrenched Actvsts B. Dfference n market reactons (ncludes US actvsts) C. Dfference n market reactons (excludes US actvsts)