Herding by foreign investors and emerging market equity returns: Evidence from Korea

International Review of Economics and Finance 19 (2010) 698–710

Contents lists available at ScienceDirect

International Review of Economics and Finance

j ourna l homepage: www.e lsev ie r.com/ locate / i re f

Herding by foreign investors and emerging market equity returns: Evidencefrom Korea

Jin Q Jeon a,⁎, Clay M. Moffett b

a Dongguk Business School, Dongguk University, 3ga 26, Phil-dong, Jung-gu, Seoul, Koreab Department of Economics and Finance, University of North Carolina Wilmington, NC, United States

a r t i c l e i n f o

⁎ Corresponding author. Dongguk Business School, DE-mail address: [email protected] (J.Q Jeon).

1 Section 2 discusses and compares the various de2 Gompers and Metrick (2001) report that large iquity market from 1980 and 1996. In contrast, Khann

while domestic institutional investors have a negligi

1059-0560/$ – see front matter © 2010 Elsevier Inc.doi:10.1016/j.iref.2010.03.001

a b s t r a c t

Article history:Received 10 May 2009Received in revised form 23 February 2010Accepted 24 February 2010Available online 4 March 2010

This paper studies the effect of herding by foreign investors on stock returns in the Koreanmarket. We conduct both pre and post-liberalization analyses and utilize a three-stage leastsquares analysis in order to control for the simultaneous relationship. We find evidence of asignificant impact of foreign investor herding on stock returns in addition to intra-year positivefeedback trading by foreign investors. However, changes in domestic institutional ownershipdo not have any significant effect on stock returns. In addition, foreign investors tend to buy/sell shares that domestic institutions sell/buy in the herding year.

© 2010 Elsevier Inc. All rights reserved.

JEL classification:G11G14G15

Keywords:HerdingFeedback tradingForeign investmentEmerging markets

1. Introduction

Much has been written about the explanatory potential of herding and the associated feedback trading with variousphenomena including stock price movements, momentum and even volatility.1 However, to the best of our knowledge, there hasbeen little research on the impact of foreign investors in newly liberalized markets. Over the last few decades, one of the mostimportant trends in international markets is the liberalization of financial markets in emerging economies. Financial marketliberalization has provided global investors with new investment opportunities to invest in what were restricted domesticsecurities. We believe this resultant growth of foreign ownership in emerging markets is of great significance to researchersinterested in understanding the impact of trading behaviors of global investors on local markets as well as to investors — foreignand domestic, individual and institutional. One of the most successful emerging markets is the Korean market which has severalunique characteristics that make it of great interest to those curious about investment behavior.

Normally domestic institutional investors are recognized as the most important investment group — particularly in moreestablished markets.2 However, they may not be the most influential class in some emerging markets. In Korea foreign investors,most of whom are U.S. and European institutional investors, hold more than 40% of the total market capitalization while domestic

ongguk University, Seoul, Korea. Tel.: +82 02 2260 8880; fax: +82 02 2260 3684.

finitions of herding and feedback trading detailed in the previous literature as well as our applicable usage.nstitutional investors with at least $100 million under management almost doubled their share of the U.S.a and Palepu (1999) document foreign ownership has a positive and significant effect on firm value in India,
e ble effect on firm value.
All rights reserved.

mailto:[email protected]

http://dx.doi.org/10.1016/j.iref.2010.03.001

http://www.sciencedirect.com/science/journal/10590560

699J.Q Jeon, C.M. Moffett / International Review of Economics and Finance 19 (2010) 698–710

institutional investors hold approximately 17% (as of 2003) with the remainder held by individuals or non-financial companies.This structure is common in emerging markets, particularly in East-Asian markets with the recent innovation of foreigninvestment.

We add to the literature by concentrating on four issues. First, we investigate the cross-sectional relationship between changesin foreign or domestic institutional ownership and stock returns. This seeks to assess the relative importance of herding by foreignor domestic institutional investors in the Korean market. We generally follow Nofsinger and Sias' (1999) ownership changeportfolio approach. We then extend the work by performing a three-stage least squares regression (3SLS) analysis controlling forsimultaneity between changes in foreign or domestic institutional ownership and abnormal stock returns. Further, we considerthe change of economic regime by dividing the sample period into two sub-periods, pre- and post-1998, when the Koreangovernment abolished the limits on foreign equities ownership. Second, we examine whether changes in foreign or domesticinstitutional ownership are related to positive feedback trading. Third, we examine whether changes in foreign or domesticinstitutional ownership are consistent with information cascades.3 Finally, we investigate the possible existence of informationasymmetries between foreign and domestic institutions. Even though a number of papers focus on herding by institutions or retailinvestors, to our knowledge, this is the first study that investigates herding by foreign investors and its impact on emergingmarketstock returns.

We find a strong and positive relation between changes in foreign ownership and stock returns. After 1998, when the foreignownership limit was abolished in the Korean market, the relationship becomes even stronger. Since the ownership data isobserved once at the end of the year, the significant correlation between changes in foreign ownership and abnormal returns maycome from either the positive impact of changes in foreign ownership on stock returns or intra-year positive feedback trading byforeign investors. We further investigate both hypotheses using three-stage least squares (3SLS) analysis and find that the resultsare supportive of both hypotheses. We also find little evidence that changes in domestic institutional ownership have a significanteffect on stock returns. Neither foreign nor domestic institutional herding is consistent with information cascades — ownershipchanges during the herding year are not positively correlatedwith those during the pre-herding year. In addition, we find evidenceof information asymmetries between foreign and domestic institutions. This suggests that foreign institutions tend to buy/sellshares that domestic institutions sell/buy in the herding year.

The study is organized as follows. In Section 2, we introduce the theory and empirical evidence on institutional herding.Section 3 describes the data and methodology used in this study and addresses general findings on foreign and domesticinstitutional herding in Korea. Section 4 examines the effect of herding by foreign investors on stock prices. In Section 5, we discussadditional issues related to pre-herding behavior by foreign investors. In Section 6, we examine informational cascades andinformation asymmetries and Section 7 concludes.

2. Herding and feedback

We address the issue of the price impact of foreign investors in emerging markets by focusing on their herding behaviors. Theliterature both theoretically as well as empirically confirms that herding and feedback trading have the potential to explain thebehaviors of institutional investors (Lakonishok, Shleifer, & Vishny, 1992; Nofsinger & Sias, 1999; Wermers, 1999; and Dennis &Strickland, 2002). Since institutional investors are regarded as sophisticated investors in the capital markets, researchers havestudied extensively whether institutional investors have the ability to identify mispriced stocks and outperform the market. Someauthors consider institutional investors to engage in “herding” or “feedback trading” — the trading of securities withoutappropriate fundamental information. Banerjee (1992) argues that herding is a rational behavior because in following others'decisions, it may “reflect information that they have and we do not.” Bannerjee goes on to point out that as this process isextended, it offers less and less information to those with an increasingly distant view of the leader. He identifies this as the “headexternality” as fewer and fewer use their own information in making a decision, but base it upon others preceding (the head)which serves to actually impede the flow of information as investors act sequentially.

Avery and Zemsky (1998) identify this sequential investing where the group of investors follows the lead of the market whileignoring or failing to gather private information as an “informational cascade.” We avail ourselves of their distinction, examiningherding as the concurrent investing, feedback trading as the reaction to the returns of the risky assets and informational cascadesas the sequential response of agents following the lead of other investors completely independent of private information. We holdthese distinctions as important due to the evidence in the literature that local institutional herding has a significant impact onstock returns in, particularly in non-U.S equitymarkets (Kim&Nofsinger, 2005; Chen &Hong, 2006).We seek to clarify the specificnature of these general effects on stock market returns.

Why would investors herd? Finance theories offer several explanations of why institutional investors might trade together.Wermers (1999) summarizes four popular theories as to why institutional investors herd. First, institutional managers are subject toreputational risk. This is the risk of acting differently from othermanagers, with different results, so thatmanagersmay ignore privateinformation and trade with the crowd (Sharfstein & Stein, 1990) in order to insure more consistent results. Second, institutionalmanagers may receive similar private information because they analyze the same price factors (Froot, Scharfstein, & Stein, 1992;Hirshleifer, Subrahmanyam, & Titman, 1994). Third,Wermers suggests institutional mangers infer private information from trades ofother managers, resulting in informational cascades (Bihkchandani, Hirshleifer, & Welch, 1992; Avery & Zemsky, 1998). Finally,

3 An information cascade arises when decisions are made by each investor sequentially, but investors begin to ignore their private signals in favor of theobserved actions of previous investors (Banerjee (1992), Welch (1992)).

700 J.Q Jeon, C.M. Moffett / International Review of Economics and Finance 19 (2010) 698–710

institutional managers may share similar aversion to stocks with particular characteristics, such as lower liquidity or higher risk(Falkenstein, 1996).

Our study is closely related to the literature which examines the correlations between institutional herding andcontemporaneous stock returns. These studies provide insight on why institutions herd and its impact on stock prices.Lakonishok et al. (1992) first develop a model to examine whether institutions herd. Using a sample of pension funds, they findthat herding and positive feedback trading are two main facets of trading by money managers. Even though this is related to theargument that herding institutions potentially destabilize stock prices, they do not find evidence that the institutions' practices arenecessarily destabilizing to market equilibriums, they find it may in fact promote it. Nofsinger and Sias (1999) examine the priceimpact of institutional herding. They argue that institutional herding and positive feedback trading are more common thanherding by individual investors. They find stocks purchased by institutions outperform stocks sold by institutions and institutionalherding is correlated with lag returns. We utilize this information in focusing our study on the impact of institutional investors,both foreign and domestic for Korea. Dennis and Strickland (2002) utilize an event study with event dates determined by largemarket swings. They find that both abnormal returns and turnover rates are related to the percentage of institutional ownershipand in particular both mutual and pension funds utilize feedback trading and herding. Sias (2003) reexamines why institutionalinvestors herd. He does not support the hypothesis that institutional herding is related to positive feedback trading. Rather, hefinds that institutions herd as a result of inferring information from each other's trades, (i.e., information cascades). We seek toidentify these trading patterns and their significance in the emerging market of Korea.

Recent literature provides international evidence on institutional herding. Kim and Nofsinger (2005) use a sample of Japanesefirms. They argue that characteristics of Japanese institutions are somewhat different from those of U.S. institutions, sinceinstitutions in Japan usually have strong long-term relationship with the firms in which they hold stock. This allows for Japaneseinstitutions to have better private information on the firms. They find that when institutional herding in Japan occurs to a lesserextent than is common in the U.S, the price impact of herding in Japan is greater. Chen and Hong (2006) examine institutionalherding in Taiwan with daily institutional holding data. In particular, they investigate the relationship between institutionalownership changes and returns localized around release events of earnings forecasts by analysts. They find that the relationshipbetween changes in institutional ownership and contemporaneous returns are the results of the price impact of herding and theyfind institutional investors more informed in buying behavior than in their selling behavior.

3. Data, variable descriptions and methodology

3.1. Sample

This study is based on a sample of Korean firms listed on the Korea Exchange (KRX) from 1992 through 2003. Monthly stockreturns, annual foreign ownership and domestic institutional ownership, and financial data come from the KIS-Value databaseprovided by the Korea Investor Service (KIS). Macroeconomic data including the exchange rate and interest rate are obtained fromthe Bank of Korea's online statistical database, BOK Economic Statistic System. The Korean security market was initially opened toforeign investors in 1992. The Korean government, however, maintained a strict limit on equity holdings for foreign investors. Thislimit was set at 10% in 1992, raised to 20% in 1996 and then 55% in 1997. After the financial crisis in 1997–1998, the Koreangovernment abolished the foreign equity holding restriction. This effort was an attempt to attract more foreign investors andprovide greater liquidity for the Koreanmarkets. The net result was a dramatic increase in foreign investment as shown in Table 1.To address this significant structural change, we divide the sample period into two sub-periods, pre-1998 (from 1992 to 1997), the

Table 1Ownership structure in Korea. This table reports annual ownership by foreign investors, domestic institutions, and domestic individuals from 1992 to 2003 inKorea. All firms in the sample are listed on the Korean Stock Exchange (KRS). Each type of ownership shows the market-value weighted mean and median of theratios of the number of shares held by each investor to the number of shares outstanding at the end of the year.

Year N Foreign ownership Local Inst. ownership Local Indiv. ownership

Mean Median Stdev Mean Median Stdev Mean Median Stdev

1992 438 0.079 0.080 0.084 0.239 0.216 0.145 0.315 0.308 0.2531993 444 0.086 0.080 0.073 0.247 0.256 0.148 0.342 0.320 0.2281994 466 0.103 0.089 0.080 0.261 0.274 0.153 0.335 0.308 0.2301995 490 0.115 0.110 0.077 0.247 0.266 0.149 0.304 0.276 0.2231996 527 0.116 0.134 0.084 0.219 0.203 0.147 0.315 0.300 0.2361997 549 0.127 0.106 0.105 0.186 0.204 0.126 0.331 0.307 0.2361998 552 0.195 0.199 0.176 0.163 0.172 0.122 0.283 0.234 0.2501999 567 0.225 0.187 0.149 0.124 0.044 0.142 0.269 0.212 0.2642000 572 0.273 0.261 0.174 0.105 0.044 0.129 0.222 0.151 0.2192001 586 0.371 0.357 0.218 0.081 0.064 0.116 0.226 0.201 0.2042002 611 0.330 0.371 0.226 0.177 0.150 0.131 0.217 0.157 0.1922003 628 0.403 0.435 0.216 0.173 0.162 0.117 0.197 0.124 0.188Total 6430 0.264 0.224 0.209 0.163 0.131 0.142 0.253 0.188 0.227


period of existence of foreign ownership limits, and post-1998 (from 1999 to 2003) when the restrictions were removed.4 Weexclude the year 1998 from analysis, due to the abnormal volatility of stock returns for Korean firms during that year as a result ofthe Asian economic crisis. This results in 6340 firm-year observations.

There are several advantages in using Korean data to investigate the impact of herding by foreign investors. First, foreigninvestment in Korea comprised 40.3% of total market capitalization in 2003 which was fourth in the world behind Hungary(72.6%), Finland (55.7%), and Mexico (46.4%) and first among Asian countries.5 Second, most foreign investors in Korea areinstitutional investors. The Korean Financial Supervisory Service (KFSS) reports that as of 2003 99.8% of foreign investors in theKorea Stock Exchange are institutions such as investment companies (50.9%), investment banks (20.9%) and pension funds(10.4%).6 The characteristics of foreign investors, however, are likely different from those of domestic institutions. These foreigninvestors are predominately from well-established markets such as the U.S or UK, and have well-established global standards andpractices. This paper seeks to understand how these firms act and whether they impact markets differently from domesticinvestors.

3.2. Variable descriptions

3.2.1. Key variables of interestThe key variables of this study are foreign and domestic institutional shareholdings, changes in shareholdings, and abnormal

herding-year returns. Foreign ownership (Local Inst Ownership) is defined as the ratio of number of shares held by foreign investors(domestic institutional investors) to the number of shares outstanding. We define domestic institutions as domestic banks,securities companies and insurance companies headquartered in Korea. Changes in foreign ownership (Δ Foreign Ownership) aredefined as foreign ownership at the end of the yearminus ownership at the beginning of the year with the same formula applied todomestic changes. Abnormal herd-year returns is a firm's excess returns over (or under) monthly compounded market returns(Korea Composite Stock Price Index — KOSPI returns) during the year.

3.2.2. Firm characteristicsChaebol is a dummy variable which equals one if a firm in the sample is a member of the 30 largest Chaebols, Korean

conglomerate business groups, as defined by the Korea Fair Trade Commission (KFTC) or 0 otherwise.7 In Korea, the Chaebolsystem dominates the market and the controlling shareholders of the Chaebol have full control over its affiliated firms usingpyramid ownership structure. This implies that Chaebol-affiliated firms have weak corporate governance (Joh, 2003; Bae, Kang, &Kim, 2002). On one hand, foreign investors who are active monitors of management would not prefer Chaebol-affiliated firms dueto their weak governance. One the other hand, due to their large firm size, Chaebol-affiliated firms are more visible to foreigninvestors (Kang & Stulz, 1997).

Following previous literature, we include several control variables for firm characteristics. Beta (proxy for systematic risk) iscalculated using daily returns with the market portfolio, (KOSPI). Size is the natural logarithm of annual sales.M/B is the market tobook ratio calculated by dividing the market value of equity by book value of equity and serves as a proxy for growth potentials.Dividend yield is the amount of dividends per share divided by the stock price at the end of the year. Leverage is the ratio of totaldebt to total equity. Finally, ROA is net profit (before interest, tax, and exceptional items) divided by the book value of assets and isthe proxy for profitability.

3.2.3. Macroeconomic variablesWe also consider the effects of political risk and macroeconomic conditions on foreign ownership or on abnormal returns

(Desai, Foley, & Hines, 2008; Hau & Rey, 2006; Cauchie, Hoesli, & Isakov, 2004). We employ two variables corresponding tomeasurable aspect of macroeconomic conditions. Exchange rate volatility is the standard deviation of the dollar-denominatedexchange rate measured at a weekly frequency during the herding year. Interest rate volatility is the standard deviation of 3-yearcorporate bond yields using a weekly frequency during the herding year.

3.2.4. Instrumental variablesIn order to ensure identification in our three-stage least squares analysis, we specify six potential instruments for foreign

ownership. Average_Δ Foreign is defined as the industry-average of changes in foreign ownership during the herding year.Pre_Δ Foreign is the change in foreign ownership during the pre-herding year. We expect that Average_Δ Foreign accounts forindustry characteristics and Pre_Δ Foreign captures the preference of foreign investors. Ln_age is the natural logarithm of thenumber of years since the firm was set up. Capitol is a dummy variable which equals one if a firm headquarters is located in Seoul,the capitol of Korea. Gov_ownership is defined as the ratio of the number of shares held by the government to the number of sharesoutstanding. Cash ratio is the value of cash and marketable securities divided by total assets.

4 Lin (2006) and Pan, Fok and Liu (2007) document changes in the trading behaviors of foreign investors in the Asian market after the Asian financial crisis.5 High foreign ownership in Finland is due mainly to the more than 90% of foreign ownership in Nokia. The major portion of foreign ownership in Hungary is

due to the presence of 45 of the world's top 50 multinational companies. In Mexico, foreign owned assembly plants, referred to as “maquiladora industry”,comprise the major portion of foreign investment.

6 KFSS's Annual report, “The stock transaction trend of foreign investors during 2003,” 2004.7 Each year, the KFTC announces the list of top 30 Chaebol groups based on total assets.


We also use two potential instruments for domestic institutional ownership defined as the following: Chaebol_inst is a dummyvariablewhich equals one if afirm is affiliatedwithChaebolswhich ownfinancial institutions as their subsidiaries and Average_Δ Localis the industry-average of changes in domestic institutional ownership during the herding year.8

3.3. Methodology

We examine the herding behavior of foreign investors by following the ownership change portfolio approach used in Nofsingerand Sias (1999). First, firms are sorted into 10 portfolios based on foreign or domestic institutional ownership at the beginning ofthe year. The firms in each decile are then further sorted into 10 portfolios based on the change in ownership over the year. Thisleaves 10 change portfolios within each of the 10 initial ownership portfolios. Then, firms are re-aggregated based on their changesin ownership decile rank resulting in 10 initial ownership stratified, ownership change portfolios, that have similar foreign ordomestic institutional ownership at the beginning of the year but experience different changes in ownership over the year.

Nofsinger and Sias (1999) point out there is a limitation of this analysis. Since it focuses on changes in the fraction of sharesheld by institutions, the change in institutional ownership may not reflect herding. This is due to the possibility the changes resultfrom one or only a few institutional investors taking a large position in a security. However, they assume that firmswith the largestinstitutional ownership changes occur as a result of herding. Kim and Nofsinger (2005) address the same concerns by arguing thenumber of shares in each decile is sufficiently large casting doubt on the idea that only a few institutions are making these trades.We find in our data themeanmarket values from decile 1 to decile 10 are approximately 132 million to 779 million U.S dollars (notreported in the table). Since equity fundswith a portfolio of more than $10million are classified as large funds in Korea, it stretchescredulity to hold that only a few institutions trade these large amounts. Thus, the assumption by Nofsinger and Sias (1999) isdeemed valid for purposes of this study.

While this analysis is helpful in assessing the relative importance of herding by foreign investors, it is possible that the resultsmay be driven by some other factors such as firm characteristics and industry effects. In other to control for the variables thatmight significantly impact the change of foreign ownership and stock returns, we employ a more rigorous model of regressionanalysis. We also address the potential simultaneous relationship between the change in foreign ownership and abnormal stockreturns. Demsetz (1983) and Demsetz and Lehn (1985) illustrate that ownership structure is endogenously determined to reach atrade-off between several cost advantages and disadvantages in the firm. Their arguments are reinforced by Demsetz andVillalonga (2001), who provide both a review of discussion and empirical evidence showing the ownership structure isdetermined by past performance and firm characteristics.

In our study there are several possible explanations for the positive relationship between the change in foreign ownership andabnormal returns. One could be simply foreign investors are attracted to firms already exhibiting some positive abnormal returns.Another possible explanation is that higher foreign ownership leads firms to perform better. Therefore, we need to control forpossible endogeneity of the relationship. To address a simultaneous relation between foreign ownership and stock returns, weemploy three-stage least squares (3SLS) regressions.9,10 The system of equations to jointly estimate the effect of changes in foreignor domestic institutional ownership on stock returns is:

whereare po

8 The9 To a

study dinvesto10 Thefrequen11 All12 The

Δ foreign local inst:ð Þownershipi = β1abnormal herding−year returnsi + β2xi + β3z1 + υi ð1Þ

abnormal herding−year returnsi = δ1Δforeign local inst:ð Þownershipi + δ2xi + δ2z2 + ϖi

xi is the set of control variables, z1 and z2 are the set of instrumental variables, and υi andϖi are the set of error terms whichssibly contemporaneously correlated (i.e., Cov(υi,ϖi)≠0).11

3.4. Descriptive statistics

Table 1 reports descriptive statistics for the overall period, 1992–2003. In 1992when Korea first opened its financial markets toforeign investors, the mean market capitalization of foreign ownership was 7.9%. After the Korean government abolished theforeign ownership limit, foreign ownership dramatically increased from 19.5% in 1998 to 40.3% by the end of 2003. The meanmarket capitalization of domestic institutional ownership was 23.9% in 1992, which significantly decreases after the economiccrisis in 1997.12 This domestic selling continued after 2000, when Korea had largely recovered from the 1997–1998 financial crisis.This selling by domestic institutions was apparently fueled by the bursting of the hi-tech bubble in 2000–2001. However, after2002, domestic firms again increased holdings from 8.1% in 2001 to 17.3% in 2003. Prior to 1998 individual investor ownershipwasvery high, at 33.1%. After the market crash individual holdings went from 28.3% in 1998 to 19.7% in 2003. Over the life of the study,

validity of instrumental variables is discussed in Section 4.2.ddress the endogenous problem, both two-stage least squares (2SLS) and 3SLS are typically used. However, we believe that 3SLS is better suited for thisue to the error terms in the system of equations are possibly contemporaneously correlated. This is due to many foreign or domestic institutionars holding multiple equity stakes in different firms.typical method utilized to control for simultaneity is to determine the direction of causality. However, due to the relatively short sample period and lowcy (annual) data, we employ the cross-sectional specification (3SLS) that can address the contemporaneous relationship.variables are described in Section 3.2.Korean Composite Stock Price index (KOSPI) was at 700 by the end of 1997, and fell to 298.01 by the end of 1998.

l

Table 2Characteristics of ownership change portfolios from 1992 to 2003. This table reports the characteristics of ownership change portfolio from 1992 to 2003. Eachyear, firms are sorted into 10 portfolios based on initial foreign or institutional ownership at the beginning of the year. The firms in each initial ownership decileare then further sorted into 10 portfolios based on the change in foreign or institutional ownership over the year. Then, firms are re-aggregated based on theichanges in ownership deciles rank. The ownership change is ownership at the end of the year minus ownership at the beginning of the year. The symbols a, b, andc represent statistical significance at the 1%, 5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

Panel A: Foreign ownership change portfolios and stock returns

Change decile Initial foreign ownership Δ Foreign ownership Abnormal herd-year returns Market-cap ROA BETA Volatility Dividend yield

1 0.082 −0.072 0.047 17.204 −0.001 0.725 0.521 0.0352 0.078 −0.043 0.063 17.610 0.024 0.803 0.502 0.0383 0.083 −0.028 0.114 17.874 0.000 0.775 0.447 0.0354 0.083 −0.017 0.021 17.877 0.014 0.703 0.446 0.0385 0.077 −0.007 0.081 17.727 0.005 0.717 0.432 0.0406 0.083 0.002 0.126 18.169 0.022 0.770 0.415 0.0417 0.086 0.012 0.140 18.210 0.018 0.758 0.420 0.0508 0.084 0.024 0.209 18.386 0.029 0.765 0.428 0.0459 0.077 0.045 0.207 18.597 0.038 0.808 0.432 0.04010 0.078 0.131 0.453 18.730 0.042 0.797 0.493 0.048F-test 0.83 248.87a 9.87a 52.58a 3.12a 3.01a 3.45a 2.12b

Panel B: Domestic institutional ownership change portfolios and stock returns

Change decile Initial local inst. ownership Δ Local inst. ownership Abnormal herd-year returns Market-cap ROA BETA Volatility Dividend yield

1 0.158 −0.122 0.119 17.115 −0.020 0.689 0.480 0.0332 0.142 −0.079 0.116 17.291 0.000 0.731 0.481 0.0393 0.146 −0.060 0.113 17.469 0.000 0.731 0.445 0.0434 0.142 −0.040 0.081 17.600 0.004 0.718 0.464 0.0445 0.143 −0.023 0.077 17.802 0.016 0.746 0.434 0.0436 0.144 −0.007 0.202 17.949 0.031 0.760 0.429 0.0437 0.146 0.010 0.156 17.889 0.017 0.741 0.426 0.0438 0.141 0.033 0.173 18.033 0.019 0.775 0.416 0.0389 0.148 0.067 0.235 18.134 0.032 0.769 0.441 0.04110 0.144 0.170 0.284 18.057 0.039 0.765 0.472 0.043F-test 0.82 623.45a 4.58a 39.55a 5.87a 2.95a 3.12a 1.56


sr

foreign ownership increased from 7.9% in 1992 to 40.3% in 2003while domestic institutional declined from 23.9% to 17.3%, a rathersignificant change. Overall, the weighted averages of foreign ownership, domestic institutional, and domestic individualownership over the study period are 26.4%, 16.3%, and 25.3%, respectively.

4. Empirical results

4.1. Ownership changes and herding-year returns

Table 2 presents the cross-sectional mean initial ownership by foreign and domestic institutional investors and ownershipchanges for firms in each ownership change portfolio. Note that our procedure is designed to have similar foreign or domesticinstitutional ownership at the beginning of the year but experience various ownership changes over the course of the year for eachof the 10 portfolios. The last row presents a test of the null hypothesis in the form of F-statistics, which test that each variable is notdifferent across the ownership change portfolio. The F-statistics both for initial ownership of foreign investors and domesticinstitutions are 0.83 and 0.82 respectively indicating that initial ownership is not significantly different with respect to the Herd-year Returns. Of particular interest to our study are the F-statistics both for changes in foreign ownership and domestic ownershipwhich are 248.87 and 623.45, both significant at the 1% level. These results strongly indicate the procedure was successful inidentifying a significant contributor to the abnormal returns since initial ownership is not statistically different among theportfolios. This occurs while the change in each ownership decile is statistically different and actually monotonically increasingacross the 10 portfolios.

The cross-sectional mean of abnormal herd-year returns over the herding year is also reported in Table 2. As expected, Panel Ashows changes in foreign ownership and abnormal herd-year returns are positively correlated. Firms in the tenth decileexperience the largest increase in foreign ownership, 13.1%, and enjoy abnormal returns of 45.3%, while firms in the first decilesuffer low abnormal returns of 4.7% and show a change in foreign ownership of−7.2%. It should be noted the overall relationshipis not monotonic. For example, the Abnormal Herd-year returns of firms in the third decile are 11.4%, which are much higher thanthose of firms in the fourth decile with a 2.1% abnormal return. A contributor to this non-monotonic relationship is likely theexistence of the strict foreign ownership limit before 1998. This limit was set at 10% in 1992, which was raised to 20% in 1996 and55% in 1997 and finally abolished in 1998. To address these concerns we later perform Nofsinger and Sias' test and regressionanalysis after excluding the period 1992–1999 from the sample in Table 2.


Firm characteristic variables by decile of foreign ownership changes are reported in Table 2. First, we reject the null hypothesisthat each characteristic variable is not different across the ownership change portfolio, since the F-statistics are significant at the5% level or better. However, onlymarket capitalizationmonotonically increases with changes in foreign ownership. ROA, Volatility,and Dividend yield are significantly different across the ownership change portfolio with no apparent monotonic relationshipbetween those variables and foreign ownership changes. Panel B shows the results of domestic institutional ownership changesand abnormal stock returns. Abnormal stock returns of firms in the tenth decile are the highest at 28.4%, but the magnitude ofdomestic institutional ownership changes is substantially less than that of the portfolios with foreign ownership changes at 45.3%.

Table 3 details the results in changes of foreign and domestic institutional ownership and abnormal herding-year returns afterexcluding the period 1992–1998. The initial foreign ownership, again, does not significantly appear different across the foreignownership change portfolios, since the F-statistics for the null hypothesis is 0.28. The results demonstrate the strong andmonotonic relation between changes in foreign ownership and abnormal herding-returns. Firms in the lowest decile experiencingthe biggest decrease in foreign ownership suffer−1.0% of Abnormal Herding-year returnswhile firms in the highest decile enjoy a56.2% abnormal return. Thus, the range of returns between the first decile and the tenth decile is 57.2% for foreign investors, as

able 3wnership change portfolios and stock returns from 1999 to 2003. This table reports the characteristics of ownership change portfolio from 1999 to 2003. Eachear, firms are sorted into 10 portfolios based on initial foreign or institutional ownership at the beginning of the year. The firms in each initial ownership decilesre then further sorted into 10 portfolios based on the change in foreign or institutional ownership over the year. Then, firms are re-aggregated based on theirhanges in ownership deciles rank. The ownership change is ownership at the end of the year minus ownership at the beginning of the year. The symbols a, b, andrepresent statistical significance at the 1%, 5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

Panel A: Foreign and domestic ownership change portfolio and stock returns, 1999–2003

Change decile Initial ownership Δ Foreign ownership Abnormal herd-year returns Market-cap ROA Volatility Dividend yield

1 0.083 −0.069 −0.010 17.133 0.012 0.730 0.0352 0.094 −0.046 −0.015 17.756 0.070 0.682 0.0393 0.091 −0.026 −0.003 17.953 0.010 0.618 0.0294 0.090 −0.015 −0.063 17.935 0.026 0.625 0.0435 0.087 −0.006 0.016 17.813 0.009 0.617 0.0436 0.091 0.004 0.015 18.173 0.032 0.596 0.0437 0.093 0.013 0.071 18.173 0.017 0.622 0.0468 0.092 0.026 0.194 18.643 0.044 0.605 0.0439 0.082 0.050 0.195 18.820 0.048 0.650 0.03210 0.091 0.148 0.562 19.038 0.054 0.711 0.041F-test 0.28 134.52a 8.25a 24.52a 2.05b 3.1a 2.05b

Panel B: Domestic institutional ownership change portfolios and stock returns, 1999–2003

Change decile Initial ownership Δ Local inst. ownership Abnormal herd-year returns Market-cap ROA Volatility Dividend yield

1 0.123 −0.110 0.038 17.099 −0.008 0.713 0.0332 0.096 −0.075 0.032 17.355 0.004 0.716 0.0343 0.107 −0.065 0.004 17.636 0.011 0.688 0.0474 0.095 −0.041 −0.041 17.605 0.010 0.685 0.0415 0.103 −0.027 −0.015 17.796 0.030 0.650 0.0426 0.100 −0.012 0.161 18.071 0.059 0.653 0.0497 0.103 0.005 0.051 18.036 0.026 0.634 0.0408 0.097 0.027 0.135 18.088 0.018 0.633 0.0419 0.106 0.062 0.224 18.305 0.042 0.692 0.03710 0.103 0.202 0.261 18.045 0.069 0.727 0.029F-test 0.95 250.23a 2.98a 15.58a 3.14a 2.12b 1.85c

Panel C: Institutional ownership change portfolio and stock returns in Japan and U.S

Japan results 1975–2001(Kim and Nofsinger 2005)

U.S. results 1977–1995(Nofsinger and Sias 1999)

Decile Initial inst.ownership

Δ Institutionalownership

Herding-yearreturns

Initial inst.ownership

Δ Institutionalownership

Herding-yearreturns

1 0.636 −0.062 −0.051 0.376 −0.160 −0.1312 0.638 −0.027 −0.061 0.370 −0.071 −0.0823 0.638 −0.016 −0.061 0.367 −0.042 −0.0604 0.638 −0.004 −0.053 0.365 −0.025 −0.0295 0.638 −0.009 −0.043 0.366 −0.011 −0.0146 0.638 0.002 −0.033 0.366 0.002 −0.0087 0.638 0.008 −0.016 0.367 0.017 0.0118 0.638 0.017 0.000 0.367 0.037 0.0439 0.638 0.030 0.032 0.366 0.070 0.08410 0.639 0.074 0.128 0.364 0.183 0.184F-stats 0.12 6948.20a 145.70a 0.02 624.82a 68.94a

TOyacc

13 See Section 3.3 for a detailed discussion of the model.14 Instrumental variables must satisfy two requirements: First, every instrument must be highly correlated with endogenous variables. Second, everyinstrument is orthogonal to the error term of the second stage equation.15 The Sargan test (1958) is one of the over-identifying restriction tests to test whether instruments are orthogonal to the error term. In particular, first, weobtain a IV-residual on all exogenous variables (instruments+control variables). Then, we regress the residual on instrument variables and obtain R2. The tesstatistics is S=nR2, where n is the number of observations.

Table 4Test for the validity of instrumental variables. This table reports validity tests for potential instrumental variables initially employed. Six potential instruments for achange in foreign ownership are considered: Average_Δ Foreign, Pre_Δ Foreign, Ln_Age, Capitol, Gov_ownership, Cash ratio. For a change in domestic institutionaownership, two possible instruments are considered: Chaebol_inst and Average_Δ Local. The symbols a, b, and c represent statistical significance at the 1%, 5%, and10% levels, respectively. The descriptions of variables are provided in Section 3.2.

Endogenous variables Instruments t-test F-test Sargan test

Δ Foreign ownership Average_Δ Foreign 4.83a 8.73a 1.225Pre_Δ Foreign −9.67a

Ln_Age −3.07a

Capitol 0.27Gov_ownership 0.28Cash ratio 0.01

Δ Local inst. ownership Chaebol_inst 1.87b 7.03a 0.099Average_Δ Local 5.18b


l

contrasted to only 22.3% for domestic institutions. The result is consistent with the notion that foreign investors are the mostinfluential ownership class in the Korean market. Given that the range of the effect of foreign institutional herding is even largerthan that of domestic institutional herding in Japan at 18%, and in the U.S at 31.4% (shown in Panel C), foreign institutional herdinghas a stronger effect on stock prices than does domestic institutional herding in Korea, and a stronger effect than institutionalherding in Japan and the U.S. These results evidence foreign investors using well developed techniques and global experiences canobtain higher returns commensurate with the higher risk of global investing.

The relationship between the changes in domestic institutional ownership and abnormal herding-returns is examined in Panel B.The abnormal herding-returns of local institutions in the fourth and fifth deciles experience negative returns of −4.1% and −1.5%respectively, while firms in the third decile enjoy positive returns of 0.4%. Also, abnormal returns in the sixth decile are even greaterthan those in the seventh and eighth deciles. These results are consistentwith the hypothesis that the impact of domestic institutionson emerging markets is not as strong as that of foreign investors from more established markets.

4.2. Regression analysis

In this section, we investigate whether this positive relationship between changes in foreign ownership and abnormal herding-returns comes from (1) intra-year positive feedback trading by foreign investors or (2) positive impact of foreign investors onreturns. Note that Table 3 does not inform as to whether stock returns change before or after foreign investors herd during theyear. It is possible that foreign ownership changes and stock return changes occur contemporaneously. In order to examine this,we employ three-stage least squares (3SLS) regressions to control for this simultaneity of the two variables.13

Before proceeding to 3SLS analysis, we check the validity of the instrumental variables chosen. In order to ensure identification,we initially specify six potential instruments for foreign ownership: Average_Δ Foreign, Pre_Δ Foreign, Ln_Age, Capitol,Gov_ownership, and Cash ratio. We also use two potential instruments for domestic institutional ownership: Chaebol_inst andAverage_Δ Local. The definitions of these variables are discussed in Section 3.2.

The results of validity tests for instrumental variables are shown in Table 4.14 Among six potential instrumental variables forforeign ownership, only the three variables Average_Δ Foreign, Pre_Δ Foreign, and Ln_age are statistically relevant to the change inforeign ownership with significant t-stats (all at the 1% level). The two instruments for domestic institutional ownership,Chaebol_inst and Average_Δ Local, are also significantly related to domestic institutional ownership though the confidence in theChaebol_inst variable is at the 5% level. Next, in an effort to strengthen these results, we perform Sargan tests for the joint test of themodel specification and the validity of the instruments. 15 Excluding weak instruments, Capitol, Gov_ownership, and Cash ratio, weobtain a χ2-statistic of 1.225 for the change in foreign ownership and 0.099 for the change in domestic institutional ownershipand, thus, fail to reject the null that instruments are valid.

Table 5 reports the results of 3SLS regressions. In the first regression of 3SLS (1), abnormal herding-returns are positivelycorrelated to changes in foreign ownership. The results show that foreign investors are more likely to be attracted to firms withgreater abnormal returns which suggest foreign investors engage in feedback trading. In the second regression of 3SLS (1), thecoefficient of a change in foreign ownership is also positive and significant at the 1% level on abnormal herding-year returns, whichsuggests that changes in foreign ownership significantly impact or drive abnormal herd-year returns.

The coefficients of three instrumental variables, Average_Δ Foreign, Pre_Δ Foreign, and Ln_age are significant in the firstregression and are validated as good instruments. While most of the control variables are generally not significantly related tochanges in foreign ownership changes, Ln_Sales (proxy for firm size),M/B, and Exchange rate volatilitymatter. The positive effect of

t

Table 5Simultaneous equation analysis of ownership changes and abnormal returns. This table reports estimation results for simultaneous equation analysis using the 3-stage least squares (3SLS) regressions. The z-statistics are within parentheses below the estimated coefficients. a, b, and c represent statistical significance at the 1%5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

3SLS (1) 3SLS (2)

Δ Foreign ownership returns Abnormal returns Δ Local inst. ownership Abnormal returns

Abnormal returns 0.021a

[8.03]Abnormal returns 0.002

[0.03]Δ Foreign ownership 1.702a

[8.66]Δ Local inst. ownership 0.162

[0.12]Average_Δ Foreign 0.061a

[0.40]Average Δ Local 1.008a

[5.01]Pre_Δ Foreign −0.201a

[−9.35]Chaebol_inst −0.002

[−0.28]Ln_Age −0.007a

[−2.60]Average_abnormal returns 0.779a

[10.59]Average_abnormal returns 0.956a

[13.99]Pre_abnormal returns −0.049a

[−2.33]Pre_abnormal returns −0.050

[−1.90]b

Chaebol −0.003[−0.45]

0.002[0.08]

Chaebol 0.001[0.03]

0.044[0.98]

Beta −0.001[−0.04]

−0.421a

[−7.28]Beta −0.005

[0.06]−0.288a

[−5.98]Ln_Sales 0.012a

[6.12]0.092a

[6.44]Ln_Sales −0.001

[−0.02]0.085a

[5.77]M/B −0.010c

[−1.92]0.526a

[9.02]M/B −0.005

[−0.42]0.522a

[11.45]Dividend yield −0.030

[−1.01]0.112[0.98]

Dividend yield −0.015[−0.33]

0.285[1.33]

Leverage −0.001[−1.08]

−0.001[−0.13]

Leverage −0.001a

[−2.88]0.001[0.04]

ROA 0.004[0.42]

0.005[0.05]

ROA 0.042b

[2.33]0.029[0.34]

Exchange rate volatility −0.080a

[−2.67]−0.099[−1.06]

Exchange rate volatility 0.071[0.58]

−0.079[−0.21]

Bond rate volatility −0.007[−1.23]

−0.253b

[−1.99]Bond rate volatility −0.021

[−1.03]−0.311b

[−2.12]Year dummy Yes Yes Year dummy Yes YesIntercept 1.105b

[2.06]−1.001a

[4.89]Intercept 0.009

[0.22]−1.557a

[−5.78]Observations 1782 1782 Observations 1756 1756X2 311.58a 992.20a X2 172.50a 910.33a


,

firm size is consistent with a notion that larger firms are more visible and, so, more attractive to foreign investors (Kang & Stulz,1997; Dahlquist & Robertsson, 2001). The negative effect of the market to book ratio suggests that foreign investors are less likelyto choose a growth firm. In addition, any exchange rate fluctuation significantly reduces foreign investment, consistent with anotion that exchange rate volatility is a disincentive for foreign investors because it increases the dimension of risk inherent inforeign investing.

In comparison, 3SLS (2) shows that there is no evidence that domestic institutional herding is related to positive feedbacktrading or that it impacts abnormal herding-year returns. The coefficients of abnormal herding-returns in the first regression(0.002) and of changes in local ownership (0.162) in the second regression of 3SLS (2) are not significant. This indicates that manyinstitutions in the Koreanmarket have a long-term relationshipwith the firms in which they invest, they usually neither engage infeedback trading, nor have a short-term focus. Kim and Nofsinger (2005) discuss this issue using their sample of institutions inJapan. Similar to the case of Japan, numerous financial institutions in Korea are Chaebol-affiliated and have reasons for a strongrelationship with other Chaebol-affiliated firms in their portfolio that extends beyond current returns. In addition, as distinct fromforeign investors, domestic institutions significantly reduce their ownership in highly leveraged firms, while their ownershipincreases for highly profitable firms with high ROA.

In summation, the evidence supports both hypotheses that foreign investors engage in positive feedback trading and thatchanges in foreign ownership significantly impact abnormal returns in the Korean market. We find no evidence that domesticinstitutional ownership changes have a significant effect on stock returns.

5. Pre-herding-year returns and ownership changes

In this section, we consider whether foreign or domestic institutional investors engage in positive feedback trading in a moredirect way. The portfolios are sorted based on previous herding-year returns, then examined as to the subsequent changes in

Table 6Pre-herding-year returns and ownership change. This table shows the relation between pre-herding-year returns and ownership changes. Each year, firms aresorted into 10 portfolios based on previous herding-year returns. The ownership change is ownership at the end of the year minus ownership at the beginning othe year. Pre-herding abnormal returns are monthly compounded market-adjusted returns during the previous herding year. a, b, and c represent statisticasignificance at the 1%, 5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

Panel A: Foreign ownership changes by decile of pre-herding annual returns

Whole period: 1999–2003 Sub-period: 1999–2003

Decile Pre_herding abnormal returns Δ Foreign ownership Decile Pre_herding abnormal returns Δ Foreign ownership

Loser −0.4097 −0.0148 Loser −0.5689 −0.01662 −0.2486 −0.0156 2 −0.3199 −0.01613 −0.1791 −0.0057 3 −0.2247 −0.00474 −0.1173 −0.0050 4 −0.1884 0.00295 −0.0422 0.0022 5 −0.0730 −0.00096 0.0838 0.0056 6 0.2201 0.01137 0.2045 0.0074 7 0.2488 0.01448 0.3655 0.0077 8 0.3753 0.01549 0.5466 0.0059 9 0.6382 0.0119Winner 1.3211 0.0121 Winner 1.8685 0.0164F-stats 434.18a 4.64a F-stats 238.96a 3.27a

Panel B: Local institutional ownership changes by decile of pre-herding annual returns

Whole period: 1999–2003 Sub-period: 1999–2003

Decile Pre_herding abnormal returns Δ Local inst. ownership Decile Pre_herding abnormal returns Δ Local inst. ownership

Loser −0.4189 −0.0049 Loser −0.5882 0.00912 −0.2607 0.0000 2 −0.3388 0.00693 −0.1864 −0.0201 3 −0.2398 −0.02504 −0.1264 −0.0118 4 −0.2118 −0.01905 −0.0358 −0.0117 5 −0.0746 −0.00936 0.0918 −0.0061 6 0.2310 0.01097 0.2013 −0.0164 7 0.2418 −0.01358 0.3636 −0.0090 8 0.3655 −0.00159 0.5365 −0.0019 9 0.6441 −0.0029Winner 1.3719 −0.0015 Winner 1.9532 −0.0182F-stats 571.15a 1.89b F-stats 296.19a 2.52b

16 We also control for the initial level of foreign ownership in year t. While not reported in the paper, the results are qualitatively the same as those reported inthe table.


fl

foreign or domestic institutional ownership. Foreign or domestic institutions exhibit feedback trading patterns if they increaseownership in firms with higher previous returns and decrease ownership in firms with lower previous returns.

The results are reported in Table 6. Panel A shows that firms with previous higher performance experience a subsequentincrease in foreign ownership with an approximately monotonic change in ownership through the progression of the deciles. Thatis, foreign investors tend to buy shares with higher previous performance (winners) and sell shares with lower previousperformance (losers). This evidence looksmore apparent in the full sample period, but is still consistent, though somewhat weakerin the sub-period post-1998. Panel A generally supports the hypothesis that foreign investors exhibit patterns of positive feedbacktrading. Panel B suggests that domestic institutions are not likely to buy previous winners or sell previous losers. As discussed inthe previous section, the fact that many domestic institutions are Chaebol-affiliated and have a strong relationship with otherChaebol-affiliated firms in their portfolios is likely a significant contributor to these results.

6. Additional issues

6.1. Informational cascades

In this section, we examinewhether foreign or domestic institution herding is consistent with informational cascades. Banerjee(1992) and Bihkchandani et al. (1992) suggest that herding arises from informational externalities. Institutional managers ignoretheir noisy information while they infer private information from other managers' trades. If herding is related to informationalcascades, there should be a positive relationship between ownership changes in the pre-herding year and those in the herdingyear.

Table 7 shows the results of OLS regressions of ownership changes in the pre-herding year on ownership changes in theherding year. It is modestly surprising to see that the coefficients of foreign ownership change in the pre-herding year aresignificantly negative in OLS (1).16 The result is similar for the case of local institutional ownership in OLS (2), but the negativecoefficients are not statistically significant. It is possible that a year may be too long of a horizon to detect actual trading activities

wherepossib

17 Chen and Hong (2006) discuss that using infrequent data may not reveal institutional herding if it occurs over a shorter time interval. For this reason, theyuse the unique dataset from the Taiwan Stock Exchange which provides daily institutional trading information.

Table 7Ownership changes during herding year and pre-herding year. This table reports estimation results for OLS regressions. The dependent variable in OLS (1) is achange in foreign ownership, while it is a change in local institutional ownership in OLS (2). The t-statistics are within parentheses below the estimatedcoefficients. a, b, and c represent statistical significance at the 1%, 5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

OLS (1) OLS (2)

Dependent variable Δ Foreign inst. ownership Dependent variable Δ Local inst. ownership

Period 1999–2003 1992–2003 Period 1999–2003 1999–2003

Pre_Δ Foreign −0.198a

[−2.98]−0.192a

[−2.83]Pre_Δ Local −0.008

[−0.34]−0.007[−0.25]

Average_Δ Foreign 0.793a

[6.01]0.721a

[4.01]Average_Δ Local 1.233a

[8.01]1.023a

[4.99]Ln_Age −0.003

[−1.43]−0.003c

[1.69]Chaebol_inst 0.001

[0.01]−0.002[−0.24]

Chaebol −0.001[−0.14]

−0.002[−0.04]

Chaebol 0.001[0.09]

0.001[0.02]

Beta −0.011c

[−1.92]−0.006[−1.03]

Beta −0.012[1.63]

−0.011[−0.12]

Ln_Sales 0.009a

[6.78]0.013a

[6.88]Ln_Sales −0.001

[−0.68]0.001[0.01]

M/B −0.022[−1.12]

0.008[0.89]

M/B 0.002[0.46]

−0.002[−0.38]

Dividend yield 0.014[0.76]

−0.021[−0.11]

Dividend yield 0.033[1.13]

−0.014[−0.33]

Leverage 0.001[0.48]

−0.001[−0.05]

Leverage −0.001b

[−2.11]−0.001a

[−3.01]ROA 0.007

[0.29]0.003[0.17]

ROA 0.028b

[2.44]0.021c

[1.93]Exchange rate volatility −0.010b

[−2.01]−0.992a

[−2.69]Exchange rate volatility 0.003

[0.78]0.025[1.12]


−0.010[−1.06]


−0.058c

[−1.77]Year dummy Yes Yes Year dummy Yes YesIntercept −0.155a

[−5.20]−0.165a

[−4.12]Intercept −0.009

[0.10]0.001[0.01]

Observations 2516 1782 Observations 2516 1782R2 0.123 0.118 R2 0.115 0.103


by institutional managers, particularly if many of their trades are short-term positions.17 The results from our data, however,suggest that neither foreign nor domestic institutions increase nor decrease shares that they have increased or decreased duringthe previous year.

6.2. Informational asymmetries between foreign and domestic institutions

It is intuitive that there is information asymmetry between foreign investors and domestic investors. Literature finds thatdomestic investors have an advantage of languages and short distances, while foreign investors have well developed technologiesandmay bemore practiced at analyzing information. For example, Dvorak (2005) finds that domestic investors have an advantageon short-lived information, but foreign investors have an advantage on long-term information. Grinblatt and Keloharju (2000)argue domestic investors are less sophisticated and take the opposite position of that of more sophisticated foreign investors.

In this section, we examine whether foreign investor herd differently from domestic institutional investors. If there isasymmetry in terms of firm value, we expect changes of foreign ownership are not positively correlated with changes of domesticinstitutional ownership. The more severe the information asymmetry that exists between foreign investors and domesticinstitutions, the more negative the correlation. To control for a simultaneity bias, we again employ 3-stage least squares (3SLS)regressions by estimating the following system of simultaneous equations:

Δforeign local inst:ð Þownershipi = β1Δ local inst:ownershipi + β2xi + β3z1 + υi ð2Þ

Δlocal inst:ownershipi = δ1Δ foreignownershipi + δ2xi + δ2z2 + ϖi

xi is the set of control variables, z1 and z2 are the set of instrumental variables, and υi and ϖi are error terms which arely contemporaneously correlated.

Table 8Simultaneous equation analysis of foreign and domestic ownership changes. This table reports estimation results for simultaneous equation analysis using the 3stage least squares (3SLS) regressions. The z-statistics are within parentheses below the estimated coefficients. a, b, and c represent statistical significance at the 1%5%, and 10% levels, respectively. The descriptions of variables are provided in Section 3.2.

Dependent variables Δ Foreign ownership Dependent variables Δ Local inst. ownership

Δ Local Inst ownership −0.076a

[−3.99]Δ Foreign ownership −0.120a

[−4.34]Average_Δ Foreign 0.855a

[6.09]Average_Δ Local 1.003a

[4.76]Pre_Δ Foreign −0.212a

[9.87]Chaebol_inst 0.020c

[1.76]Ln_Age −0.004b

[−1.99]Chaebol −0.001

[−0.43]Chaebol 0.001

[0.02]Beta −0.004

[−0.77]Beta −0.006

[−0.89]Ln_Sales 0.010a

[5.92]Ln_Sales 0.001

[0.48]M/B −0.042

[−1.04]M/B −0.002

[−0.25]Dividend yield −0.028

[−0.88]Dividend yield −0.017

[−0.45]Leverage −0.001

[−0.69]Leverage −0.001a

[−2.83]ROA 0.005

[0.55]ROA 0.033b

[2.34]Exchange rate volatility −0.091a

[−2.78]Exchange rate volatility 0.069

[0.49]Bond rate volatility −0.006

[−1.18]Bond rate volatility −0.019

[−0.99]Year dummy Yes Year dummy YesIntercept −0.165a

[−4.83]Intercept −0.178a

[−3.90]Observations 1756 Observations 1756X2 278.04a X2 221.43a


-,

Table 8 shows the results of 3SLS regressions. In the first stage, the coefficient of changes in local institutional ownership in issignificant and negative (−0.076) based on the z-statistic. This is consistent with the argument of informational asymmetrybetween foreign and domestic investors. Foreign investors would buy or sell stocks with domestic institutions taking thereciprocal position of selling or buying the stocks in the previous year. The same story appears in the second stage. The negativecoefficient of changes in foreign ownership indicates that domestic institutions subsequently buy what foreign investors sold.Three instruments for changes in foreign ownership, Average_Δ Foreign, Pre _Δ Foreign, and Ln_age, are statistically significant atthe 5% level or greater, implying that they are valid instruments.

Also, instruments for changes in domestic institutional ownership, Average_Δ Local and Chaebol_inst, appear to be significant.Consistent with results in Table 6, changes in foreign ownership in the previous year are negatively correlated with changes in theherding year with a coefficient of −0.212 and a z-statistic of −9.87 which is significant at the 1% level. The coefficient indicatesfirms that experienced higher changes in foreign ownership in the previous year are more likely to have lower (or more negative)changes in foreign ownership. This argument is related to our previous discussion of whether foreign investors possessinformational cascade behaviors discussed in Table 7. This negative relationship indicates that they do not increase shareholdingsof stocks that they have increased in the previous year.

The significance of these results dovetails into Choe et al. (2001) who use Korean intraday data (from 1996 to 1998) to showthat the domestic investors in Korea have a private information advantage over foreign investors, which creates a “substantialdisadvantage for foreign investors” with regard to buying/selling securities which then have large positive/large negativeabnormal returns. While due to our data infrequency, we can't be as specific with regard to results on individual trades, we canconfirm that this opposite pattern of trading behavior is clear and for the duration of our data (1999 through 2003) continues toexist.

7. Conclusion

The liberalization of financial markets in emerging economies usually results in the growth of foreign ownership of domesticsecurities. In Korea, foreign investors, most of whom are institutional investors from the U.S and Europe, holdmore than 40% of thetotal market capitalization while domestic institutional investors hold only about 17%. This has attracted the attention ofresearchers who seek to understand the impact of trading behaviors of global investors on emerging markets.


In this paper, we extend the literature on institutional herding and feedback trading and explore how foreign ownershipchanges are related to stock returns. More specifically, this study focuses on four issues. First, it examines the cross-sectionalrelation between changes in foreign and domestic institutional ownership and stock returns. This serves to assess the importanceof herding by foreign and domestic institutional investors in the Korean market. In addition, we consider the change of economicregimes by dividing the sample period into two sub-periods, pre- and post-1998 when the Korean government abolished the limiton foreign ownership. Secondly, we examine whether changes in foreign and domestic institutional ownership are related topositive feedback trading. Third, we examine whether ownership changes are consistent with information cascades and furthertest for information asymmetries between foreign and domestic institutions.

We find a strong and positive relation between changes in foreign ownership and stock returns. After 1998, when the foreignownership limit was abolished in the Koreanmarket, the relationship became even stronger. Since the ownership data is observedonce at the end of the year, the significant correlation between changes in foreign ownership and abnormal returns may comefrom either the positive impact of changes in foreign ownership on stock returns or intra-year positive feedback trading by foreigninvestors. We further investigate both hypotheses using three-stage least squares (3SLS) analysis and find that the results aresupportive of both hypotheses. We also find little evidence that changes in domestic institutional ownership have a significanteffect on stock returns. Neither foreign nor domestic institutional herding is consistent with information cascades — ownershipchanges during the herding year are not positively correlatedwith those during the pre-herding year. In addition, we find evidenceof information asymmetries between foreign and domestic institutions. This suggests that foreign institutions tend to buy/sellshares that domestic institutions sell/buy in the herding year.

Acknowledgments

The authors would like to express their appreciation to James A. Ligon, Junsoo Lee, seminar participants at the 2007 FMA andSFA conferences, and two anonymous referees for their invaluable suggestions and comments. All errors and omissions are thesole responsibility of the authors.

References

Avery, C., & Zemsky, P. (1998). Multidimensional uncertainty and herd behavior in financial markets. American Economic Review, 88, 724−748.Banerjee, A. (1992). A simple model of herd behavior. American Economic Review, 88, 724−748.Bae, K., Kang, J., & Kim, J. (2002). Tunneling or value added? Evidence from merger byKorean business group, Journal of Finance, 57, 2695−2740.Bihkchandani, S., Hirshleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom and cultural change as informational cascades. Journal of Political Economy,

100, 992−1026.Cauchie, S., Hoesli, M., & Isakov, D. (2004). The determinants of stock returns in a small open economy. International Review of Economics and Finance, 13, 167−185.Chen, A., & Hong, B. (2006). Institutional ownership changes and returns around analysts' earnings forecast release event. Evidence from Taiwan, Journal of Banking

and Finance, 30, 2471−2488.Choe, H., B. Kho, & R. Stulz, 2001, Do domestic investors have more valuable information about individual stocks than foreign investors?, NBER working paper #8073.Dahlquist, M., & Robertsson, G. (2001). Direct foreign ownership, institutional investors, and firm characteristics. Journal of Financial Economics, 59, 413−440.Demsetz, H. (1983). The structure of ownership and the theory of the firm. Journal of Law and Economics, 26, 375−390.Demsetz, H., & Lehn, K. (1985). The structure of corporate ownership. Causes and consequences, Journal of Political Economy, 93, 1155−1177.Demsetz, H., & Villalonga, B. (2001). Ownership structure and corporate performance. Journal of Corporate Finance, 7, 209−233.Dennis, P., & Strickland, D. (2002). Who blinks in volatile markets, individuals or institutions. Journal of Finance, 57, 1923−1949.Desai, M., Foley, F., & Hines, J. (2008). Capital structure with risky foreign investment. Journal of Financial Economics, 88, 534−553.Dvorak, T. (2005). Do domestic investors have an informational advantage? Evidence from Indonesia, Journal of Finance, 60, 817−839.Falkenstein, E. (1996). Preferences for stock characteristics as revealed by mutual fund portfolio holdings. Journal of Finance, 51, 111−135.Froot, K., Scharfstein, D., & Stein, J. (1992). Herd on the street: Informational inefficiencies in amarketwith short-term speculations. Journal of Finance, 47, 1461−1484.Gompers, P., & Metrick, A. (2001). Institutional investors and equity prices. Quarterly Journal of Economics, 116, 229−259.Grinblatt, M., & Keloharju, M. (2000). The investment behavior and performance of various investor-types: A study of Finland's unique data set, 2000. Journal of

Financial Economics, 55, 43−67.Hau, H., & Rey, H. (2006). Exchange rate, equity prices, and capital flows. Review of Financial Studies, 19, 273−317.Hirshleifer, D., Subrahmanyam, A., & Titman, S. (1994). Security analysis and trading patterns when some investors receive information before others. Journal of

Finance, 49, 1665−1698.Joh, S. W. (2003). Corporate governance and firm profitability: Evidence from Korea before the economic crisis. Journal of Financial Economics, 68, 287−322.Kang, J., & Stulz, R. M. (1997). Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan, Journal of Financial Economics, 46, 3−28.Khanna, T. & K.G. Palepu, 1999, Emerging market business groups, foreign investors, and corporate governance, NBER working paper 6955.Kim, K. A., & Nofsinger, J. R. (2005). Institutional herding, business groups, and economic regimes: Evidence from Japan. Journal of Business, 78, 213−242.Lakonishok, J., Shleifer, A., & Vishny, R. W. (1992). The impact of institutional trading on stock prices. Journal of Financial Economics, 32, 23−43.Lin, A. Y. (2006). Has the Asian crisis changed the role of foreign investors in emerging equity markets: Taiwan's experience. International Review of Economics and

Finance, 15, 364−382.Nofsinger, J. R., & Sias, R. W. (1999). Herding and feedback trading by institutional and individual investors. Journal of Finance, 54, 2263−2295.Pan, M., Fok, R., & Liu, Y. (2007). Dynamic linkages between exchange rates and stock prices: Evidence from East Asian markets. International Review of Economics

and Finance, 16, 503−520.Sharfstein, D., & Stein, J. (1990). Herd behavior and investment. American Economic Review, 80, 465−479.Sias, R. W. (2003). Institutional herding. Review of Financial Studies, 17, 165−206.Welch, I. (1992). Sequential sales. Learning, and Cascades, Journal of Finance, 47, 695−732.Wermers, R. (1999). Mutual fund herding and the impact on stock prices. Journal of Finance, 54, 581−622.

Documents

Herding by foreign investors and emerging market equity returns: Evidence from Korea