1-s2.0-s0261560613000405-main.pdf=3f_tid=3d09d085c2-e095-11e3-b7f3-00000aab0f26=26acdnat=3d1400641827_07615d03fac2a3a9cd8434787b9e033a...

8/21/2019 1-s2.0-S0261560613000405-main.pdf=3f_tid=3d09d085c2-e095-11e3-b7f3-00000aab0f26=26acdnat=3d1400641

http:///reader/full/1-s20-s0261560613000405-mainpdf3ftid3d09d085c2-e095-11e3-b7f3-00000aab0f2626acdnat3 1/20

Large foreign ownership and stock price

informativeness around the world

Wen He a,*, Donghui Li b,1, Jianfeng Shen b,2, Bohui Zhang b,3

a The University of New South Wales, School of Accounting, Australian School of Business, Kensington,

Sydney, NSW 2052, Australiab The University of New South Wales, School of Banking and Finance, Australian School of Business,

Kensington, Sydney, NSW 2052, Australia

JEL classications:

G15

O16

Keywords:International nancial market

Large foreign ownership (LFO)

Stock price informativeness

Probability of informed trading (PIN)

Price non-synchronicity (NONSYNC)

a b s t r a c t

This study investigates the relation between large foreign owner-

ship (LFO) and the informativeness of stock prices in 40 markets.

We show that LFO is positively related to price informativeness,

measured by probability of informed trading (PIN) and price non-

synchronicity (NONSYNC) which reects rm-specic variations in

stock returns. We also nd a stronger association between stock

returns and future earnings innovations for rms with higherLFO.

Further analysis reveals that the effect of LFO on price informa-

tiveness is stronger in developed economies and markets with

strong investor protection and a transparent information

environment.

2013 Elsevier Ltd. All rights reserved.

1. Introduction

Foreign investors have emerged as an important group of investors in many nancial markets. Take

our sample of 2002 as an example, foreigners on average owned 28% of shares in 3189 rms in 40

markets. Given the signicant presence of foreign investors in local markets, the specic roles played

by foreign investors in local markets have attracted considerable attention from both academics and

* Corresponding author. Tel.: 61 2 93855813; fax: 61 2 93865925.

E-mail addresses: [email protected](W. He),[email protected](D. Li),[email protected](J. Shen),bohui.

[email protected](B. Zhang).1 Tel.: 61 2 93855873; fax: 61 2 93866347.2 Tel.: 61 2 93854581; fax: 61 2 93866347.3 Tel.: 61 2 93855834; fax: 61 2 93866347.

Contents lists available atSciVerse ScienceDirect

Journal of International Money

and Financej o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / lo c a t e / j i m f

0261-5606/$ see front matter 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.jimonn.2013.04.002

Journal of International Money and Finance 36 (2013) 211230
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]://www.sciencedirect.com/science/journal/02615606http://www.elsevier.com/locate/jimfhttp://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://dx.doi.org/10.1016/j.jimonfin.2013.04.002http://www.elsevier.com/locate/jimfhttp://www.sciencedirect.com/science/journal/02615606http://crossmark.dyndns.org/dialog/?doi=10.1016/j.jimonfin.2013.04.002&domain=pdfmailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]



policy makers. One stream of research has studied the effects of foreign equity investments on the real

side of invested rms and local economy, such as productivity, investment and growth (e.g., Henry,

2000a;Bekaert et al., 2005,2009). Another stream of research examines properties of stock prices

including liquidity (Bekaert et al., 2007;Rhee and Wang, 2009), return volatility (Bae et al., 2004;Li

et al., 2011), and the cost of capital (Bekaert and Harvey, 2000;Henry, 2000b).

This study investigates whether foreign investors, particularly large foreign investors (LFO), affectthe extent to which stock prices incorporate value-relevant information and thus are informative about

a rms fundamental value. In an efcient market, stock prices incorporate all the available information

and reect a rms intrinsic value. However, prices could fail to reveal all the available information

because of various frictions such as information cost (Grossman and Stiglitz, 1980) and limits to

arbitrage (Shleifer and Vishny, 1997). Consequently, the degree of price informativeness varies across

rms and has important implications for both managers and investors. More informative stock prices

can help managers make better decisions in capital investment, resulting in higher efciency in capital

allocation and corporate investment (Wurgler, 2000; Durnev et al., 2004; Chen et al., 2007). Infor-

mative prices also reduce the risk for uninformed investors and thus lower the cost of capital for the

rm (Fernandes and Ferreira, 2009). Therefore, given foreign investorssubstantial share ownership, it

is important to understand the impact of foreign investors on a rms stock price informativeness.The existing literature suggests two potential channels through which large foreign shareholders

could improve the informativeness of stock prices of the invested rm. First, large foreign shareholders

could improve the informativeness of stock prices through their informed trading. Compared with

small shareholders, large shareholders tend to have a stronger incentive and better capability to collect

and process value-relevant information.4 Thus the information-based trading by large shareholders, for

example, selling their stakes upon negative information, facilitates the capitalization of fundamentals

into stock price (Edmans and Manso, 2011;Admati and Peiderer, 2009;Edmans, 2009).5

Second, large foreign shareholders could enhance price informativeness by improving corporate

governance and disclosure quality of the invested rms.6Jin and Myers (2006) show that poor investor

protection and opaque nancial disclosure reduces price informativeness. The reason is that managers

and controlling shareholders tend to engage in self-dealing activities to extract private benets at theexpense of other shareholders (e.g.,Shleifer and Vishny, 1989;Morck, 1996). To camouage their self-

dealing activities, managers may withhold information or manipulate accounting disclosure, which

makes stock prices less informative about a rms fundamentals (Fan and Wong, 2002;Jin and Myers,

2006; Gul et al., 2010). Furthermore, agency problems could deter sophisticated investors from

engaging in costly and risky informed arbitrage, which also reduces the amount of relevant informa-

tion in stock prices (Morck et al., 2000). If large foreign shareholders, due to their signicant interest at

stake, closely monitor managers and constrain agency problems, we expect to nd more informative

stock prices.

However, it remains an empirical question whether large foreign investors can be effective monitors

of local managers and insiders. On the one hand, relative to domestic shareholders, foreign share-

holders may incur a higher cost to collect relevant information to effectively monitor insiders due tothe geographic and cultural distance (Kang and Kim, 2008, 2010). On the other hand, large foreign

shareholders would be less engaged with insiders compared with their domestic counterparties, which

could facilitate their monitoring role. Further, the differences in corporate governance practices

4 A number of studies document mixed evidence on whether foreign investors have better information than local investors

(Grinblatt and Keloharju, 2000;Hau, 2001;Seasholes, 2004;Choe et al., 2005;Dvorak, 2005). We note that foreign investors in

these studies are likely to be foreign mutual funds with relatively small ownership in local companies. In contrast, we focus on

large foreign investors that own at least 5% of shares of a rms stock. The substantial shareholding of large foreign investors

should allow them to be better informed than outside investors.5 In their theoretical model, Edmans and Manso (2011) show that the monitoring role of large shareholders does not

necessarily rely on active trading by large shareholders. The mere threat of exit upon negative information could effectively

convey information about rm value and improve the informativeness of stock prices.6 The vast majority of blockholder literature focuses on how blockholders add value through direct intervention or moni-

toring. See,Shleifer and Vishny (1986),Admati et al. (1994),Kahn and Winton (1998),Maug (1998),Aghion et al. (2004), among

others.

W. He et al. / Journal of International Money and Finance 36 (2013) 211230212



between the home country and the host country could also make large foreign shareholders more

effective monitors of insiders compared with their domestic counterparties, especially for those large

foreign shareholders coming from good governance countries and investing in poor governance

countries (Kho et al., 2009).

We dene large foreign shareholders as those ultimate owners who own more than 5% of

outstanding shares of a rm and are domiciled outside the country of the invested rm.7 LFO iscalculated as the percentage of shares outstanding owned by all large foreign shareholders. Following

the literature, we employ two measures of price informativeness: probability of informed trading (PIN)

and price non-synchronicity (NONSYNC). Developed byEasley et al. (1997),PINis the ratio of informed

trading intensity to total trading intensity and thus measures the probability of trades placed by

informed traders. The market micro-structural model inEasley et al. (1997)indicates that the higher

thePIN, the greater the amount of information in stock prices. 8 Building on this insight, a few studies

have usedPINto measure price informativeness (e.g.,Easley et al., 1996, 1998,2002;Chen et al., 2007;

Brockman and Yan, 2009,Ferreira et al., 2011).

Our second proxy for price informativeness,NONSYNC, is based on rm-specic variations in stock

returns. Specically, we compute NONSYNCas a logistic transformation of the R-square from the

regression of stock returns on returns of the local market index and US market index. Roll (1988) is oneof the rst to suggest that rm-specic return variation might capture the rate of information incor-

poration into stock prices through trading.9 Morck et al. (2000)propose the use of this variable to

measure the informativeness of stock prices. Recent empirical evidence supports the view that large

rm-specic return variation indicates more informative stock prices. For example, Durnev et al.

(2003) show that rm-specic return variation is positively related to the extent to which stock

returns reect future earnings information.NONSYNCis also found to be positively associated with the

efciency of capital allocation (Wurgler, 2000;Durnev et al., 2004;Chen et al., 2007). In the literature,

NONSYNC has been employed to measure price informativeness in a number of studies (see, for

example, DeFond and Hung, 2004; Piotroski and Roulstone, 2004;Chan and Hameed, 2006;Daouk

et al., 2006;Bris et al., 2007;Brockman and Yan, 2009).

Using a cross section of 3189 rms in 40 markets in 2002, we nd thatLFO is positively related toboth proxies of stock price informativeness, PINand NONSYNC. This positive relation is robust to the

control of lagged price informativeness, and a number ofrm and country characteristics. To gauge the

economic signicance of the effect ofLFO, we show that ifLFO increases from the rst quartile (9%) to

the third quartile (45%) in the sample, PINwould increase by 1.52.3 percentage points. The effect of

LFO is stronger than that of any other rm characteristic included in the regressions except for rm size

which is typically found to be the most signicant determinant ofPIN(Aslan et al., 2011).10 Analysis

based onNONSYNCgives similar inferences about the economic signicance ofLFO. Overall, the results

suggest that LFO is an important determinant of price informativeness for our sample rms.

To shed further light on the effect ofLFO on price informativeness, we examine whether companies

with high LFO have stock prices that contain more information about future operating performance

measured by earnings. Our results show that forrms with higherLFO, current stock returns are moresignicantly related with contemporaneous and future earnings innovations. This evidence suggests

that stock prices of highLFO rms are more informative about rmsfuture earnings, which provides a

more direct support for the effect ofLFO on price informativeness.

7 We trace upward along the chain of ownership to identify the ultimate owners. More details about large foreign ownership

data will be provided in section II.8 Vega (2006) nds that stocks with higher PIN have smaller post-earnings-announcement drifts. Her result suggests that

stocks with higher PINadjust to fundamentals more quickly, supporting PINas a measure of price informativeness.9 Roll (1988) nds that the R-squares of a market model or multi-factor model of individual stocks returns are very low on

average for U.S. stocks, indicating high rm-specic return variations. He further shows that excluding days with public rm-

specic news does not improveR-squares much, which suggests that high rm-specic return variations could be possibly due

to trading on private information.10 In our sample, an inter-quartile increase in rm size is associated with about a 7% decrease in PIN. For all other major

determinants ofPIN, such as bid-ask spread, share turnover and return volatility, an inter-quartile change is associated with a

less than 1% change in PIN.

W. He et al. / Journal of International Money and Finance 36 (2013) 211230 213



We proceed to investigate whether the association between LFO and price informativeness varies

systematically with country-level corporate governance and information infrastructure. This investi-

gation helps understand whether local institutional infrastructures complement or substitute LFO in

improving price informativeness, and thus deepens our understanding of the mechanisms underlying

the informational role of large foreign shareholders. Prior literature offers two competing views. One

view is that large foreign shareholders may play a more important role in markets with weakergovernance and a poorer information environment. In these markets, monitoring by large foreign

shareholders serves as a substitute for weak institutional infrastructures, and the marginal benet of

improvingrmsgovernance and transparency could be higher. Further, due to the bonding of home-

country laws and the lack of engagement with insiders, foreign investors may serve as more effective

monitors of insiders in markets with poor governance (Kho et al., 2009). The other view suggests a

stronger informational role of large foreign shareholders in markets with better governance and infor-

mation environment, due to a complementary effect between institutional infrastructures andrm-level

governance, as suggested by Doidge et al. (2007). In these markets, the cost of collecting information and

monitoring insiders is lower, which could facilitate the informational role of large foreign shareholders.

Therefore, it is an empirical question as to whether large foreign shareholders play a larger or smaller

informational role in markets with stronger governance and better information infrastructures.Using several measures of the strength of investor protection and information disclosure, we nd

that the association between LFO and measures of price informativeness is stronger in markets with

stronger investor protection and better information disclosure. Furthermore, the association between

stock returns and future earnings innovations increases withLFO more signicantly in these markets.

These results suggest that large foreign shareholders play a larger informational role in markets with

stronger investor protection and better information disclosure, which supports the view of comple-

mentarity between market-level and rm-level governance forces.

Our paper contributes to the literature on the role of foreign equity investment in local markets in

three ways. First, we document an important economic benet associated with foreign equity invest-

ment, i.e., improved informativeness of stock prices. Second, while mostof the existing studies focus only

on the effect of foreign equity investment in emerging markets, we study the effect ofLFO on stock priceinformativeness in both emerging and developed markets. We also show that the role of foreign

shareholders could vary signicantlyacross markets. Third, we employ a measure of actual large foreign

ownership to study the effect of foreign equity investment in local markets, while most existing studies

examine market liberalization events or use a stock investibilityindex to proxy for foreign equity in-

vestment.11 While market openness andinvestibility are related to foreign investorsability to invest in

local markets, they dont necessarily correspond to the exact holdings of foreign investors.

Two recent studies are closely related to ours. Gul et al. (2010)examine the relationship between

foreign equity ownership and the rm-level information environment in China. They compare a

measure of stock price non-synchronicity between rms issuing only A-shares, which are invested

almost exclusively by domestic investors, and rms issuing A-shares along with B-shares or H-shares,

both of which are invested almost exclusively by foreign investors.12

Bae et al. (2012) nd that, in 21emerging markets, a stocks investibility by foreign investors is positively related to the timeliness of

incorporating global market information into stock prices.

Our paper also connects to a large stream of literature on block ownership. In particular, Brockman

and Yan (2009) nd evidence that aggregate block ownership is positively associated with the infor-

mativeness of stock prices in the US market. However it is unclear whether their evidence is applicable

to other markets, since there are signicant differences between large foreign shareholders and do-

mestic block shareholders in their monitoring role and information advantage. Furthermore, the in-

ternational setting allows us to examine whether the roles played by blockholders vary with country-

level governance and information infrastructures. Our results show that strong investor protection and

11 The exception isLi et al. (2011)who use the same measure as ours.12 Since 2002, some qualied foreign institutional investors (QFII) could also invest in A-shares, and some qualied domestic

institutional investors (QDII) could also invest in B and H shares and overseas markets. Please refer to Gul et al. (2010, page 11)

for more institutional details.




transparent information environment facilitates large foreign shareholders role in improving price

informativeness.

Finally, our paper adds to an emerging set of literature on the interaction between market-level and

rm-level governance forces. Bergman and Nicolaievsky (2007) model the choice of how much

investor protection and related mechanisms by rms depends on the legal and investor protection

regime of the market. Doidge et al. (2007) nd that observable rm characteristics explain much less ofthe variance in governance ratings than country characteristics, and especially lack explanatory power

in less developed markets. Li et al. (2011) show the stabilizing role of large foreign shareholders is more

profound in markets with better corporate governance. Our study provides further support on this

complementarity view of market-level and rm-level governance forces by showing that large foreign

shareholders play a larger role in markets with stronger investor protection and better information

disclosure.

The paper proceeds as follows. In Section2we describe the construction of key variables and data.

In Section3we present and discuss our empirical results. And nally we conclude in Section4.

2. Construction of variables, data and sample

2.1. Construction of variables

2.1.1. Large foreign ownership (LFO)

FollowingLi et al. (2006)and Li et al. (2011),LFO is calculated as the sum of foreign block share-

holdings with a block dened as a holding larger than orequalto 5% of a rms issued shares. According

toLi et al. (2011), A foreign shareholder is dened as a citizen of another country, a business entity

registered (or head-quartered) in another country, or an unlisted majority-owned subsidiary of a

foreign company (page 9). More details about the identication and construction ofLFO could be

found inLi et al. (2006)andLi et al. (2011).

2.1.2. Probability of informed trading (PIN)

Everything else equal, the more intensively informed investors trade, the more information is

incorporated into the stock price. Thus we employ the probability of informed trading (PIN) developed

byEasley et al. (1997)as one of the proxies of price informativeness. In the market micro-structural

model of Easley et al. (1997), trades are executed by two groups of investors: informed investors

who trade on their private information and uninformed investors who trade on liquidity needs. At the

beginning of each trading day, a private information event occurs with the probability a, where the

probability that bad (good) news happens isd(1 d). Both informed trades and uninformed trades are

assumed to arrive in the market following independent Poisson processes. Informed trades arrive at

the rate ofm only on the days with a private news event. Informed traders buy when the news is

positive and sell when the news is negative. The arrival rate of uninformed buy (sell) trades is b(s)

which is independent of the occurrence and sign of the private news event. Based on the estimated

parameters, the probability that an order is from an informed investor could be calculated as follows13:

PIN am

am b s(1)

The set of parameters, q (a, d, b, s), is estimated by maximizing the following likelihood function,

LqjB; S YTt 1

LqjBt; St (2)

where Tdenotes the number of trading days in a year, and Bt(S

t) denotes the number of buy (sell)

orders on dayt. For a specic dayt, the likelihood function is

13 SeeEasley et al. (1997)andEasley et al. (2002)for a detailed discussion of the structural model and the estimation ofPIN.




LqjBt; St 1 aes

Sts

St!eb

Btb

Bt! adesm

smSt

St! eb

Btb

Bt!

a1 des

Sts

St!ebm

bmBt

Bt!

(3)

Researchers can estimate PINfrom intra-day data. Recent empirical evidence supports the view that

PINcaptures the amount of (private) information in stock prices. For example,Easley et al. (2002) nd

stocks with high PINearn higher returns, consistent with stock prices reecting risk of private infor-

mation as captured by PIN. Vega (2006) nds that stocks with high PINhave smaller post-earnings-

announcement drifts, suggesting that prices of high PINstocks adjust to earnings information in a

timelier manner.Chen et al. (2007) show that corporate investment is more sensitive to prices for high

PIN stocks, consistent with these stocks having more informative prices. Other studies using PIN to

measure price informativeness includeBrockman and Yan (2009), andFerreira et al. (2011).

2.1.3. Price non-synchronicity (NONSYNC)

Another measure of price informativeness is price non-synchronicity (NONSYNC) which is based on

R2 obtained from the market model of stock returns,

ri;j;t aibirm; j; tgi

rm;US; t ej;t

i; j; t (4)

where ri,j,tis the return on rm is stock in market j at time t, rm,j,tand rm,US,tare the returns on the

market index of market j and the U.S. at timetrespectively, andej,trefers to a change in the exchange

rate per U.S. dollar for the currency of marketjat timet. We estimate the above equation using weekly

data in a year. According to Morck et al. (2000), a higher value of 1 R2 indicates that more rm-

specic information is capitalized into stock price and thus stock price is more informative.

FollowingMorck et al. (2000), we logistically transform 1 R2 to obtain an appropriate distribution,

NONSYNC

1R

2

R2

(5)

The idea that price non-synchronicity, or rm-specic return variation, measures price informa-

tiveness is based on the observation ofRoll (1988) that market returns and public news announce-

ments explain only a small portion of stock return variations. He indicates that private information

incorporated in prices could be one reason for the lowR2 of market models for stock returns. Based on

this idea and the pattern of market model R2 across markets,Morck et al. (2000)propose to use price

non-synchronicity as a measure of price informativeness.Durnev et al. (2003)show that rms with

high price non-synchronicity have returns that are more informative about future earnings, which

supports the use of price non-synchronicity as a measure of price informativeness. Further supporting

evidence comes from studies showing that price non-synchronicity is associated with the efciency of

capital allocation and investment (Wurgler, 2000; Durnev et al., 2004; Chen et al., 2007). In theliterature, NONSYNChas been employed to measure price informativeness in a number of studies (see,

for example,DeFond and Hung, 2004;Piotroski and Roulstone, 2004;Chan and Hameed, 2006;Daouk

et al., 2006;Bris et al., 2007;Brockman and Yan, 2009).

2.2. Data and sample

We obtain data from following sources: a) ownership data from the OSIRIS database and Lexis/

Nexis; b) intraday prices, quotes, and trading data from Reuters Datascope Tick History (RDTH, pre-

viously known as TAQTIC); c) stock returns and exchange rate data from Datastream; d) rm-level

nancial data from Worldscope and Global COMPUSTAT; and e) market-level characteristics from

World Development Indicators, Global Competitiveness Report by the World Economic Forum, theInternational Financial Corporation of the World Bank Group, and other resources.

Our sample covers 3189rms with foreign block ownership data in 40 markets in 2002. The number

of rms differs across model specications due to the availability of various control variables. We

require sample rms to have non-missing LFO data to avoid potential errors associated with the




treatment ofrms without LFO data. Leaving out rms with missing LFO also helps mitigate the po-

tential endogeneity issue about the investment decision by foreign investors. Data availability restricts

our sample to year 2002, but we expect our results to be applicable to other time periods since the

ownership of large shareholders tends to be stable and persistent.

LFOis measured at the end of 2002, while PINand NONSYNCare calculated using the data of 2003

which is one-year ahead of the period to construct all independent variables. To control for the effect oflarge domestic ownership, we include as a control variable the largest domestic ownership (LDO)

which is calculated as the percentage of shares outstanding held by the largest domestic shareholders

at the end of 2002. We also control various rm-level and market-level characteristics which are ex-

pected to relate to both proxies of price informativeness and the preference of foreign investors.

Following the previous literature (Easley et al., 2002; Chan and Hameed, 2006), we control the

following rm-level variables which are expected to relate to PINor NONSYNC: 1) rm size, measured as

the log of market capitalization in US$ at the end of 2002 (LogMV); 2) liquidity, measured as the average

of quoted spread in 2002 (PQSPR); 3) trading activity, measured as the average monthly share turnover,

trading volume in shares divided by the number of shares outstanding, in 2002 (TURNOVER); and 4)

stock return volatility, measured as the standard deviation of monthly stock returns in 2002 (STDRET).

In addition to the above variables, we also control the following variables which could relate toforeign investorsinvestment preferences: 1) rm growth, measured as the log ratio of book equity at

the end of the scal year ending in 2002 to the market capitalization at the end of 2002 ( LogBM); 2)

leverage ratio, measured as the ratio of long-term debt over the common equity at the end of the scal

year ending in 2002 (LEVERAGE); 3) dual-listing in the US via initiating an American Depositary Receipt

program, measured as a dummy variable taking value 1 if a rm has an ADR and 0 otherwise (ADR); and

4) rm protability, measured as return on equity for the scal year ending in 2002 (ROE) and annual

stock return in 2002 (RET12).

The measures of price informativeness such as PINandNONSYNCcould differ systematically across

markets with different degrees of economic and nancial development which could be also related to

the foreign investorschoice of invested markets. We include the following market-level variables to

control the degree of economic and nancial development: 1) the log of per capita GDP measured inUS$ in 2002 (LogGDPPC); 2) the annual GDP growth rate in 2002 (GDPGR); 3) the standard deviation of

annual GDP growth rates over the period 1998 to 2002 (STDGDPGR); 4) the ratio of stock market

capitalization at the end of 2002 to the annual GDP of 2002 ( MVGDP); 5) the ratio of private credit to

GDP in 2002, where private credit refers to nancial resources available to private sectors through

loans, purchases of non-equity securities, trade credits and other accounts receivable (PCRGDP); and 6)

the standard deviation of monthly market returns in 2002 (STDMRET).

Table 1 presents the summary statistics of selectedrm characteristics for rms in our sample. Panel

A shows that, on average, 28% of shares outstanding are held by large foreign shareholders. The level of

LFO differs signicantly acrossrms, with a standard deviation of 23% and an inter-quartile value of 34%.

Similarly, there exists signicant variation across rms in the two proxies of price informativeness,PIN

and NONSYNC. More interestingly, both variables seem to vary systematically with LFO. In Panel B, whenwe divide the sample into three equally-sized groups based on the level ofLFO, we nd bothPINand

NONSYNCto increase monotonically withLFO. Finally, in Panel C and D, we report the number ofrms

and the average value of each variable market by market for developed and emerging markets sepa-

rately. The number ofrms differs across markets, ranging from a minimum of 8 rms for Russia to a

maximum of 460 rms for the U.S. On average a developed market has about twice the number ofrms

represented in our sample as that of an emerging market. Firms in emerging markets tend to have

slightly higher LFO. For price informativeness proxies, rms in emerging markets on average have a

lowerNONSYNC, but about the same level ofPIN, relative to those rms in developed markets.

3. Empirical results

3.1. Cross-sectional relationship betweenLFO and price informativeness

We estimate the cross-sectional relation between LFO and price informativeness from the following

regression,




Table 1

Summary statistics. This table presents the summary statistics for rms in our sample. In Panel A we report some summary

statistics ofrm characteristics for all rms in our sample. In Panel C, we report the average value ofrm characteristics within

three groups classied bythe level ofLFO. Finally, in Panels A and B wepresent the average value ofrm characteristics market by

market for developed and emerging markets respectively. All variables are dened in theAppendix.

Statistics LFO PIN NONSYNC LDO LogMV LogBM PQSPR TURNOVER STDRET LEVERAGE ROE RET12 ADRPanel A: Summary statistics ofrm characteristics for the pool sample

Mean 0.28 0.27 1.95 0.14 5.53 0.23 0.03 0.06 0.14 0.49 0.04 0.00 0.14

Stdev 0.23 0.13 1.52 0.19 2.03 0.82 0.04 0.11 0.10 0.81 0.27 0.05 0.35

Q1 0.09 0.18 0.85 0.00 4.03 0.70 0.01 0.01 0.08 0.00 0.00 0.02 0.00

Median 0.19 0.25 1.78 0.00 5.44 0.22 0.01 0.03 0.11 0.21 0.08 0.00 0.00

Q3 0.43 0.33 2.86 0.23 6.87 0.24 0.03 0.07 0.17 0.60 0.16 0.02 0.00

LFO PIN NONSYNC LDO LogMV LogBM PQSPR TURNOVER STDRET LEVERAGE ROE RET12 ADR

Panel B: The average value ofrm characteristics for groups ofrms with different levels ofLFO

Low 0.09 0.25 1.79 0.24 5.75 0.22 0.02 0.07 0.13 0.51 0.04 0.00 0.15

Medium 0 .22 0 .27 1 .96 0.15 5 .44 0.21 0.03 0.06 0.14 0.50 0.02 0.00 0.16

High 0.53 0.28 2.10 0.04 5.40 0.28 0.03 0.05 0.13 0.45 0.05 0.01 0.12

Market N LFO PIN NONSYNC LDO LogMV LogBM PQSPR TURNOVER STDRET LEVERAGE ROE RET12 ADRPanel C: The average value ofrm characteristics for each developed market

Australia 157 0.24 0.31 2.74 0.13 3.76 0.41 0.04 0.04 0.18 0.22 0.26 0.01 0.01

Austria 26 0.38 0.29 1.19 0.20 6.26 0.26 0.01 0.02 0.12 0.56 0.07 0.02 0.06

Belgium 50 0.38 0.40 2.85 0.13 5.06 0.21 0.03 0.01 0.10 0.52 0.06 0.01 0.02

Canada 146 0.27 0.24 2.08 0.09 5.55 0.43 0.02 0.05 0.18 0.44 0.01 0.05 0.32

Denmark 38 0.22 0.29 2.41 0.15 5.41 0.19 0.03 0.04 0.13 0.63 0.11 0.01 0.08

Finland 21 0.24 0.33 3.13 0.13 5.74 0.56 0.02 0.04 0.09 0.41 0.12 0.02 0.10

France 196 0.26 0.32 2.53 0.25 5.46 0.42 0.03 0.05 0.15 0.69 0.02 0.00 0.10

Germany 138 0.35 0.25 2.96 0.18 4.61 0.01 0.05 0.01 0.19 0.47 0.07 0.02 0.05

Hong Kong 110 0.36 0.30 2.00 0.17 5.22 0.12 0.03 0.05 0.13 0.25 0.02 0.01 0.05

Ireland 24 0.22 0.32 3.06 0.13 5.28 0.37 0.05 0.04 0.16 0.48 0.03 0.02 0.13

Italy 59 0.35 0.22 1.57 0.17 6.06 0.10 0.01 0.06 0.12 1.09 0.00 0.00 0.07

Japan 163 0.22 0.23 2.05 0.10 5.21

0.07 0.01 0.05 0.12 0.38 0.05 0.01 0.02Netherlands 57 0.21 0.31 2.51 0.14 5.56 0.24 0.03 0.05 0.12 0.45 0.02 0.01 0.16

New Zealand 30 0.26 0.25 1.94 0.10 5.16 0.45 0.02 0.03 0.10 0.66 0.10 0.02 0.03

Norway 38 0.23 0.29 2.13 0.16 5.44 0.17 0.02 0.15 0.16 0.74 0.08 0.01 0.05

Singapore 61 0.37 0.32 1.95 0.13 4.75 0.16 0.03 0.06 0.12 0.26 0.07 0.01 0.00

Spain 65 0.22 0.19 2.14 0.19 6.97 0.63 0.01 0.05 0.10 0.79 0.08 0.01 0.07

Sweden 52 0.20 0.24 1.84 0.10 5.45 0.48 0.02 0.09 0.21 0.48 0.10 0.00 0.08

Switzerland 74 0.25 0.29 2.33 0.17 6.02 0.36 0.02 0.04 0.14 0.50 0.05 0.01 0.05

United Kingdom 256 0.19 0.23 2.82 0.09 5.35 0.50 0.05 0.08 0.14 0.43 0.01 0.00 0.09

United States 460 0.21 0.19 3.26 0.10 4.73 0.61 0.02 0.12 0.25 0.62 0.12 0.01 1.00

Average 106 0.27 0.28 2.36 0.14 5.38 0.30 0.03 0.05 0.14 0.53 0.01 0.00 0.12

Panel D: The average value ofrm characteristics for each emerging market

Argentina 29 0.51 0.30 1.50 0.12 5.04 0.42 0.02 0.12 0.29 0.47 0.00 0.02 0.24

Brazil 44 0.57 0.32 0.84 0.10 5.10 0.18 0.07 0.02 0.19 0.68 0.12 0.02 0.16

Chile 50 0.39 0.42 1.48 0.17 5.44 0.06 0.06 0.01 0.09 0.43 0.10 0.01 0.27

China 36 0.21 0.20 2.07 0.25 5.14 0.91 0.00 0.07 0.10 0.15 0.03 0.02 0.00

Greece 38 0.29 0.19 0.50 0.22 5.26 0.35 0.01 0.04 0.12 0.41 0.09 0.03 0.03

India 108 0.40 0.25 1.51 0.07 4.86 0.51 0.01 0.03 0.13 0.46 0.20 0.02 0.04

Indonesia 83 0.36 0.40 1.96 0.18 3.62 0.03 0.06 0.04 0.25 0.67 0.07 0.07 0.02

Israel 33 0.25 0.32 1.92 0.13 5.13 0.57 0.04 0.05 0.17 0.43 0.02 0.03 0.39

Korea 88 0.22 0.19 0.57 0.11 4.70 0.40 0.01 0.24 0.20 0.57 0.06 0.01 0.02

Malaysia 93 0.30 0.32 1.58 0.14 4.49 0.06 0.02 0.02 0.09 0.24 0.05 0.00 0.00

Mexico 24 0.30 0.31 0.77 0.27 6.28 0.06 0.03 0.02 0.11 0.52 0.08 0.00 0.35

Philippines 54 0.39 0.40 2.28 0.11 3.26 0.43 0.10 0.04 0.16 0.44 0.01 0.01 0.00

Poland 39 0.53 0.24 1.32 0.10 5.15 0.22 0.03 0.02 0.14 0.54 0.03 0.00 0.00

Portugal 21 0.35 0.35 2.29 0.22 4.78 0.02 0.14 0.01 0.13 1.85 0.02 0.02 0.05

Russian 8 0.25 0.18 2.92 0.23 6.34 0.30 0.00 0.00 0.15 0.27 0.19 0.01 0.25

South Africa 54 0.32 0.34 1.54 0.15 5.13 0.06 0.04 0.03 0.14 0.47 0.10 0.05 0.07Taiwan 38 0.21 0.29 0.83 0.10 6.23 0.19 0.01 0.20 0.17 0.21 0.11 0.02 0.05

Thailand 97 0.30 0.34 2.89 0.12 4.18 0.08 0.03 0.01 0.12 0.41 0.14 0.03 0.00

Turkey 31 0.54 0.19 0.19 0.10 4.32 0.06 0.01 0.25 0.23 0.22 0.13 0.00 0.00

Average 51 0.35 0.29 1.50 0.15 4.97 0.13 0.04 0.06 0.16 0.50 0.08 0.00 0.10




Price informativenessm; i a0a1LFOm; i a2LDOm;i X

j

bjFirm controljm; i

X

k

gkMarket controlkm m; i : (6)

Price informativeness is measured by either PINor NONSYNC. To isolate the effect ofLFO from its

domestic counterparty, we control for the effect of the holding of the largest domestic shareholder

(LDO). We also include the lagged value of price informativeness proxy to control for its persistence. We

further control various rm-level and market-level variables which are expected to relate to both the

proxies of price informativeness and the investment preference of large foreign shareholders, as dis-

cussed in the above section. To minimize the potential endogeneity issue, the dependent variable is

constructed based on data in 2003 which is one-year ahead of the period, 2002, used to construct all

independent variables.

Tables 2and 3 present the regression results usingPINand NONSYNCas the dependent variable

respectively. We consider various model specications to test the relation between LFO and price

informativeness proxies: model (1) tests the bivariate relation without including control variables;

models (2)(4) sequentially control the holdings of the largest domestic shareholders, the laggedvalue of price informativeness proxy, and other rm-specic variables; and nally model (5) con-

trols both rm-level and market-level variables. The sample size varies with the addition of control

variables. The t-statistics (reported in parentheses) are computed based on robust standard errors

clustered by market. Our discussion focuses on the results using PIN as the dependent variable

(Table 2), while the key results are qualitatively similar if NONSYNC is used as the dependent

variable.

The results inTable 2show that the coefcient ofLFO is positive and signicant across all model

specications. Along with country-xed effects, LFO explains about 20% of the cross-sectional variation

Table 2

Relation betweenLFO and PIN. This table presents the results of the regression specied in the following equation,

PINm; i a0 a1 LFOm; i a2LDOm; i a3 PIN L1m; i a4LOGMVm; i a5LOGBMm; i a6 PQSPRm; i a7TURNOVERm; i

a8STDRETm; i a9LEVERAGEm; i a10ROEm; i a11RET12m; i a12ADRm; i b1LOGGDPPCm b2 MVGDPm

b3PCRGDPm b4 GDPGRm b5STDGDPGRm b6STDMRETm k; i

For each explanatory variable, we report its regression coefcient and the robustt-statistic (in parentheses) calculated from the

standard error with the market clustering. All variables are dened in theAppendix.

Variable (1) (2) (3) (4) (5)

INTERCEPT 0.184 (19.61) 0.374 (23.22) 0.363 (22.31) 0.279 (12.33) 0.447 (8.82)

LFO 0.066 (4.79) 0.044 (3.13) 0.068 (4.38) 0.058 (3.58) 0.045 (2.78)

LDO 0.056 (3.51) 0.043 (2.61) 0.047 (2.70)

PIN_L1 0.241 (5.94) 0.306 (7.72)

LOGMV 0.025 (11.70) 0.026 (11.77) 0.020 (8.64) 0.019 (8.31)

LOGBM 0.002 (0.43) 0.002 (0.44) 0.005 (1.42) 0.002 (0.53)

PQSPR 0.433 (2.79) 0.401 (2.60) 0.184 (1.25) 0.233 (1.69)

TURNOVER 0.018 (0.69) 0.029 (1.12) 0.024 (1.18) 0.008 (0.34)

STDRET 0.120 (3.49) 0.117 (3.39) 0.100 (2.84) 0.109 (3.12)

LEVERAGE 0.002 (0.51) 0.002 (0.61) 0.002 (0.78) 0.001 (0.33)

ROE 0.018 (1.75) 0.018 (1.68) 0.007 (0.74) 0.005 (0.59)

RET12 0.092 (1.35) 0.081 (1.22) 0.018 (0.28) 0.051 (0.79)

ADR 0.037 (4.37) 0.035 (4.14) 0.026 (3.10) 0.011 (1.62)

LOGGDPPC 0.007 (1.58)

MVGDP 0.005 (1.70)

PCRGDP 0.037 (4.97)

GDPGR 0.900 (4.42)

STDGDPGR 0.543 (3.17)STDMRET 0.276(5.44)

Country dummy Yes Yes Yes Yes No

Nobs 2170 1848 1848 1744 1744

Adj-Rsq 19.76% 38.92% 39.35% 44.27% 39.13%



http:///reader/full/1-s20-s0261560613000405-mainpdf3ftid3d09d085c2-e095-11e3-b7f3-00000aab0f2626acdnat 10/20

in PIN. After controlling various rm-level determinants of PIN, the coefcient on LFO remains

signicantly positive. For example, the coefcient ofLFO in model (4) is 0.043, which is statistically

signicant at 1% level. In model (5), we further replace the country-xed effects with country char-

acteristics and nd similar results.

The coefcient ofLFO is also economically signicant given the sample distribution ofLFO. When LFO

increases from its 25 percentile (9%) to 75 percentile (45%) values, everything else equal, PINwould

increase by 1.52.3 percentage points, depending on the model specication. This suggests that relative

tormswith9% LFO, rmswith45% LFO have about 2% more of alltrades initiated from informedtraders.

To put this effect in comparison, we do a similar calculation for all other rm characteristics included in

our regressions, based on the quartile values reported in Panel A, Table 1and the estimated regressioncoefcients inTable 2. For other rm characteristics except for rm size (LOGMV), an inter-quartile

change is typically associated with a marginal change in PINaround 1% or even less. This comparison

suggests that, although the effectofLFO appears small, LFO is a key determinant of price informativeness.

Turning to control variables, PINseems to vary signicantly with some rm characteristics.PINis

very persistent, as indicated by the signicant coefcient on the lagged value of PIN. LDO is also

positively related toPIN, suggesting that large domestic shareholders also help improve price infor-

mativeness. Further, PIN is higher in stocks that are smaller in size, more illiquid, and less volatile in

returns, consistent with ndings inAslan et al. (2011). Finally,PINalso varies systematically with some

market-level characteristics.PINis lower in markets with a more developed private nancial system, a

higher GDP growth rate and more volatile stock market returns.

Table 3reports regression results usingNONSYNCas the dependent variable. The coef

cient onLFOis positive and statistically signicant across all model specications. After calculating the economic

signicance of the coefcients, we nd LFO has larger impact on NONSYNCthan most rm charac-

teristics except for rm size. For control variables, NONSYNCis lower in rms with lowLDO, a bigger

market capitalization, a higher book-to-market ratio, and higher bid-ask spreads. Further, the

Table 3

Relation betweenLFO and NONSYNC. This table presents the results of the regression specied in the following equation,

NONSYNCm; i a0 a1LFOm; i a2LDOm; i a3NONSYNC L1m; i a4LOGMVm; i a5LOGBMm; i a6PQSPRm; i

a7TURNOVERm; i a8 STDRETm; i a9 LEVERAGEm; i a10ROEm; i a11RET12m; i a12ADRm; i

b1LOGGDPPCm b2MVGDPm b3PCRGDPm b4 GDPGRm b5STDGDPGRm b6STDMRETm k; i

For each explanatory variable, we report its regression coefcient and the robustt-statistic (in parentheses) calculated from thestandard error with the market clustering. All variables are dened in theAppendix.

Variable (1) (2) (3) (4) (5)

INTERCEPT 3.047 (29.75) 4.467 (21.96) 4.364 (21.31) 3.434 (16.68) 1.484 (3.33)

LFO 0.520 (4.75) 0.340 (2.82) 0.547 (3.86) 0.426 (3.12) 0.323 (2.40)

LDO 0.480 (2.96) 0.295 (1.90) 0.342 (2.24)

NONSYNC_L1 0.256 (9.54) 0.309 (12.09)

LOGMV 0.357 (17.87) 0.358 (17.99) 0.281 (14.16) 0.260 (13.37)

LOGBM 0.089 (2.45) 0.088 (2.46) 0.087 (2.55) 0.134 (3.87)

PQSPR 3.618 (3.60) 3.340 (3.35) 1.866 (1.99) 2.500 (2.75)

TURNOVER 0.299 (1.07) 0.207 (0.74) 0.026 (0.12) 0.191 (0.78)

STDRET 0.133 (0.42) 0.105 (0.33) 0.171 (0.56) 0.035 (0.12)

LEVERAGE 0.026 (0.84) 0.023 (0.74) 0.036 (1.29) 0.025 (0.82)

ROE 0.212 (2.17) 0.204 (2.10) 0.165 (1.80) 0.175 (1.90)RET12 1.179 (1.66) 1.088 (1.56) 0.195 (0.30) 0.290 (0.46)

ADR 0.075 (0.72) 0.059 (0.56) 0.065 (0.69) 0.023 (0.26)

LOGGDPPC 0.118 (2.83)

MVGDP 0.004 (0.13)

PCRGDP 0.291 (3.65)

GDPGR 2.820 (1.43)

STDGDPGR 10.140 (6.63)

STDMRET 0.267 (0.44)

Country dummy Yes Yes Yes Yes No

Nobs 3189 1849 1849 1849 1849

Adj-Rsq 19.90% 44.03% 44.28% 48.92% 44.85%




estimated coefcients on both ownership variables, 0.426 forLFO and 0.295 for LDO, are comparable in

the scale to that on the ownership variable reported inFernandes and Ferreira (2008).14

Overall, the results reveal a positive relationship between LFOand measures of price informative-

ness, consistent with our hypothesis that large foreign shareholders improve the informativeness of

stock prices. However, this relationship might be spurious if some rm characteristics are omitted from

in the regression but are related to both LFO and the degree of price informativeness. One possiblyomitted variable is earnings quality, because higher quality of earnings could be associated with both

larger shareholders and more informative prices. We use three measures of earnings quality as sug-

gested by Leuz et al. (2003): the magnitude of accruals (ACCRUAL), earnings smoothing (SMOOTH), and

earnings correlation (CORR).ACCRUAL is widely used in the accounting literature to measure the degree

of earnings management. SMOOTHmeasures the extent to which reported earnings are smoothed

across scal years to reduce the variability of earnings. And CORRmeasures the correlation between

changes in accruals and changes in operating cash ows, and thus the lower the correlation, the more

likely it is that accruals are manipulated to conceal economic shocks to a rms operating cash ows.

Thus earnings quality decreases in ACCRUALand SMOOTH, and increases in CORR. The construction of

these variables is detailed in theAppendix. When we include these measures of earnings quality into

regressions, wend that the coefcient ofLFO remains almost unchanged and statistically signicant inall model specications except one in whichNONSYNCis used as the dependent variable and ACCRUAL

is used as the proxy for earnings quality. We conclude that the positive relation betweenLFOand price

informativeness is robust to the control of earnings quality.

3.2. LFOand the association between stock returns and earnings innovations

At this point our empirical analyses nd a positive relation between LFO and two proxies of price

informativeness: PIN and NONSYNC. Though both measures have been shown theoretically and

empirically in previous studies to capture the extent to which stock prices incorporate value-relevant

information, it is possible that they rely on some impractical assumptions and serve as very noisy

proxies of price informativeness. In this section, we employ a more direct, albeit more narrow, measure

of price informativeness the extent to which earnings information is reected in stock prices, and test

its relation to LFO. Prior studies have shown that stock prices contain information about future earnings

(Ayers and Freeman, 2003;Durnev et al., 2004). IfLFO is positively related to the extent to which stock

prices incorporate relevant information, including earnings information, we would expect this alter-

native price informativeness measure to increase in LFO .15

We run the following regression to test whether stock prices ofrms with higherLFOcontain more

information about contemporaneous and future earnings innovations,

Ri a0a1DEi a2DE1ib1DEiLFOib2DE1i LFOia3LFOia4DEi LDOia5DE1i

LDOi

a6LDOi

a7DEi

LOGMVi

a8DE1i

LOGMVi

a9LOGMVi

X

m

gmMarket dummymi i;

(7)where R is the annual stock return in 2003, DEandDE1 are the annual changes in earnings in 2003 and2004 scaled by the market capitalization at the beginning of each year respectively, and other variables

are dened in section II and measured in 2002. We expect the coefcients ofDEand DE1 to be positive,

suggesting prices contain information about current and future earnings (Ayers and Freeman, 2003). If

LFOis positively associated with the pricing of earnings information, we would expect the coefcients

14 When regressingNONSYNCon OWNERSHIP, which is the percentage of closely held shares, Fernandes and Ferreira (2008)

found a coefcient of 0.007 for developed markets and 0.004 for emerging markets (columns (3) and (7) in their Table 7,page

235) based on a sample of international stocks. Please note that they measured OWNERSHIP in percentage.15 As will be introduced shortly, empirically we measure the informativeness of stock prices about earnings by the association

between stock returns and earnings innovations. We also use the term pricing of earnings informationin later discussions to

refer to this measure, in the sense that the association measures the extent to which earnings information is incorporated in

stock prices.




on the interaction terms ofLFO and earnings changes to be signicantly positive. We control for the

effect ofLDO and rm size, as well as country-xed effect.

Table 4 presents the regression results. Column 1 conrms that contemporaneous and future

earnings innovations are positively associated with stock returns, validating the use of the return-

earnings association to examine the informativeness of stock prices. More importantly, the result in

Column 2 shows a signicantly positive coefcient on the interaction terms of LFO and earningschanges, suggesting that stock prices of rms with a higher LFO contain more information about

current and future earnings innovations. The adjusted R-square also increases from 9.1% to 9.8% by

adding the two interaction terms. In Columns 3 to 5, we progressively add controls for LDOand rm

size. The coefcients for the interaction term between LFO and earnings changes remain statistically

signicant. These results suggest that LFO is positively associated with the extent to which stock

returns reect earnings information. This evidence corroborates the ndings in the regressions with

PINand NONSYNCas the proxies of price informativeness.

3.3. Does large foreign shareholders contribute to the informativeness of stock prices?

Our empirical analyses so far indicate a positive relation betweenLFO and the informativeness ofstock prices, after controlling for various rm-level and market-level variables. We interpret our

ndings as consistent with the view that large foreign shareholders contribute to the price discovery

and the informativeness of stock prices. However, we are aware of two potential issues with the above

interpretation.

Therst issue is the possible spurious relation between LFO and price informativeness. It could be

the case that bothLFO and price informativeness are driven by unknown rm characteristics which are

not controlled in our analyses. We think it unlikely though we cannot totally rule out this possibility.

We measure price informativeness using data in 2003, one-year ahead of the period used to construct

LFO, and include the lagged value of the dependent variable as a regressor. IfLFO in 2002 and price

informativeness measure in 2003 are correlated due to the unobserved rm nature and the persistence

in price informativeness measure, we would expect this relation to be trivial once the lagged priceinformativeness measure is controlled. Our regression results indicate a signicant relation between

LFO in 2002 and price informativeness measures in 2003, even after controlling for the contempora-

neous relation betweenLFO and price informativeness.

The second concern with our interpretation is the potential reverse causality. The positive relation

between LFO and price informativeness could be due to large foreign shareholders preference for

Table 4

Relation betweenLFOand the association between stock returns and earnings innovations. This table presents the results of the

regression specied in the following equation,

Ri a0 a1DEi a2DE1i b1DEi LFOi b2DE1i LFOi a3 LFOi a4DEi LDOi a5DE1i LDOi a6LDOi a7DEi LOGMVi

a8DE1i LOGMVi a9 LOGMVi X

m

gmMarket dummymi i

For each explanatory variable, we report its regression coefcient and the robustt-statistic (in parentheses) calculated from the

standard error with the market clustering. All variables are dened in theAppendix.

Variable (1) (2) (3) (4) (5)

DE 0.720 (7.46) 0.361 (2.39) 0.308 (1.06) 0.205 (1.10) 0.092 (0.28)

DE1 0.541 (5.51) 0.303 (1.99) 0.111 (0.40) 0.264 (1.35) 0.204 (0.64)

DE*LFO 1.395 (3.09) 1.435 (3.17) 1.615 (3.38) 1.690 (3.51)

DE1*LFO 0.951 (2.05) 0.935 (2.02) 1.008 (2.02) 1.049 (2.09)

LFO 0.004 (0.07) 0.009 (0.15) 0.039 (0.57) 0.025 (0.36)

DE*LDO 0.876 (1.42) 0.982 (1.58)

DE1*LDO 0.183 (0.30) 0.346 (0.56)

LDO 0.108 (1.27) 0.082 (0.96)

DE*LOGMV 0.011 (0.23) 0.019 (0.39)

DE1*LOGMV 0.089 (1.85) 0.093 (1.92)LOGMV 0.017 (2.45) 0.017 (2.43)

Country dummy Yes Yes Yes Yes Yes

Nobs 1609 1609 1609 1609 1609

Adj-Rsq 9.10% 9.80% 10.30% 10.15% 10.58%




stocks with more informative prices. While this reverse causality is quite possible, we believe it un-

likely to drive our results and thus invalidate our inferences, based on both empirical evidence and

theoretical arguments.

First, a common approach to inferring causality is to run a Granger causality test using the difference

of variables. Due to the availability of only one-year ofLFO data, we are unable to run a formal Granger

causality test. However, as emphasized above, we run a predictive regression of the price informa-tiveness measure in 2003 on LFO in 2002, controlling for the lagged dependent variable. To some

extent, it is similar to the spirit of a simple Granger causality test based on a vector auto-regression

including only one lag of the variables, i.e., VAR(1) model.16 We nd a signicant coefcient on LFO

in our predictive regression after controlling the lag of dependent variable, which seems to be

consistent withLFO Granger-causing price informativeness.

Second, for the reverse causality to drive our results, it must be that large foreign shareholders base

their investment decision on their expectation of the future price informativeness which is not related

to the current price informativeness. Though conceptually possible, this reverse causality due to

forward-looking has not been established in any theoretical or empirical work according to the best of

our knowledge. Instead, ourndings seem to be consistent with a signicant body of literature on the

roles played by large shareholders and foreign investors in improving governance and price discovery,as we discussed in Section1.

Finally, in the cross-market analysis to be presented in Section 3.4, we nd that LFO is more

signicantly related to price informativeness in markets with a more developed economy, better

investor protection, and a more transparent information environment. If our results were driven by a

reverse causality from price informativeness to LFO, it would be difcult to explain why price infor-

mativeness is less attractive to foreign investors in markets with worse investor protection and a

poorer information environment. In these markets where information is lacking, stocks with infor-

mative prices should be more attractive to investors who base their investment decision on price

informativeness, compared to those markets where information is abundant. On the contrary, as we

will discuss in detail in Section3.4, if large foreign shareholders indeed help improve price informa-

tiveness, better investor protection and a more transparent information environment at the marketlevel could facilitate this informational role of large foreign shareholders, which could explain the

cross-market differences in our ndings.

To summarize, we interpret our results as consistent with large foreign shareholders improving price

informativeness. Though a reverse causality is completely possible, it is less likely to drive our results

based on the above discussions. Nevertheless, due to the empirical difculty of establishing a causeeffect

relation, we are still open to alternative explanations and call for cautions in interpreting our results.

3.4. Macro governance forces and the relation betweenLFOand price informativeness

Based on a large cross-section ofrms in 40 markets, we nd a positive relation between LFOand

price informativeness. Our results suggest that large foreign shareholders play an important information

role in local equity markets by improving the informativeness of stock prices. We now proceed to

examine whether this information role of large foreign shareholders vary across countries in a sys-

tematic way. Specically, we examine whether country-level economic development, investor protec-

tion and information quality affect the association between LFO and price informativeness. The results

from this analysis should sharpen our understanding of the informational role of foreign investors.

We rst investigate the effect of economic development on the relation between LFO and price

informativeness. According to the classication by the International Financial Corporation of the World

Bank Group, we divide our sample into developed-market and emerging-market sub-samples. There

16 The equation in a bivariate VAR(1) model, Dyt1 c aDyt bDxt t1, could be expressed as follows after re-arranging

variables: yt1 c (a 1)ytayt1 bxt bxt1 t1. The two variables, y and x here, correspond to price informativeness

andLFOin our analyses respectively. Thus our regression model parallels VAR(1) model in a simple Granger causality test, in the

sense of an unrestricted model without including the 2nd lag of the dependent and the independent variables. We conrm that

including the 2nd lag of the dependent variable price informativeness measures in 2001 does not alter our results.




are signicant differences between these two sub-samples in terms of economic growth, development

and openness ofnancial markets, and presence of foreign investors. We run the regression of price

informativeness on LFO, as specied in Equation(6), for both sub-samples and report the results in

Table 5. To save space, we only report the coefcients ofLFO andLDO, although all the control variables

are included in regressions. The results show that LFO is more related to price informativeness mea-

sures in developed markets than in developing markets. For example, in the regression withPINas the

dependent variable, the coefcient on LFO is 0.060 (t-statistic 2.90) in the developed-market sample,

while the same coefcient becomes 0.028 and insignicant (t-statistic 1.01) in the emerging-market

sample.17 WhenNONSYNCis used as the dependent variable, the coefcient onLFO is also larger in the

developed-market sample, though the difference between the two samples is smaller.

We next investigate the effect of country-level governance forces such as investor protection andinformation transparency. We consider two measures of the market-level investor protection regime:

the legal origin and the anti-self-dealing index. There is evidence that markets with common law

provide better protection for investor rights (La Porta et al., 1998). The anti-self-dealing index is

developed byDjankov et al. (2008)and focuses on a markets litigation governing self-dealing trans-

actions. The higher the value of the anti-self-dealing index, the better the protection of shareholders

interests. Two sub-samples are formed based on whether a markets legal system is common civil law

origins and whether its anti-self-dealing index is above the medium value of the index across all

markets in the sample. We run the same regression as specied in Equation(6)for both sub-samples

and report the coefcients ofLFO andLDO in Table 6. The results show thatLFO is more related to both

proxies of price informativeness in markets with the common law legal origin compared to markets

with the civil law legal origin. When we measure investor protection by the anti-self-dealing index, wendLFO is more related with PINin high-index markets than in low-index markets.

Turning to information quality, we consider two market-level proxies: the disclosure score and the

nancial transparency index. The disclosure score is based on survey results about the level and

availability ofnancial disclosures and is reported in the annual Global Competitiveness Report issued

by the World Economic Forum. FollowingGelos and Wei (2005), we average the disclosure scores of

1999 and 2000 and then divide the average score by 10 to obtain a range from 0 to 1. The nancial

transparency index is developed by Bushman et al. (2004) and measures the intensity and timeliness of

nancial disclosures by rms. For both the disclosure score and the nancial transparency index, a

higher value means better disclosure and a more transparent information environment. We then

divide markets into two groups based on the median value of either the disclosure score or the nancial

Table 5

Relation between LFO and price informativeness: Developed versus emerging markets. This table presents the results of the


Price informativenessm;i a0 a1LFOm; i a2LDOm; i a3Price informativeness L1m; i a4LOGMVm; i a5LOGBMm; i

a6PQSPRm; i a7TURNOVERm; i a8STDRETm; i a9LEVERAGEm; i a10ROEm; i a11RET12m; i

a12ADRm;

i b1LOGGDPPCm b2MVGDPm b3PCRGDPm b4GDPGRm b5 STDGDPGRm b6STDMRETm k; i

The price informativeness is measured as eitherPINor NONSYNC. The regression is run separately for developed-market and

emerging-market sub-samples. For the main variables of interests, we report their regression coefcients and the robust t-

statistics (in parentheses) calculated from the standard errors with the market clustering. All variables are dened in the

Appendix.

Variable Dep. Var. PIN Dep. Var. NONSYNC

(1) Developed markets (2) Emerging markets (3) Developed markets (4) Emerging markets

LFO 0.060 (2.90) 0.028 (1.01) 0.462 (2.72) 0.424 (1.93)

LDO 0.042 (2.26) 0.069 (1.67) 0.307 (1.68) 0.454 (1.58)

Nobs 1219 525 1303 546

Adj-Rsq 42.66% 34.76% 42.94% 39.16%

17 We test the signicance of the difference by running a pool regression in which interaction terms between developed

market dummy and rm characteristics are added. The coefcient on the interaction between LFO and developed market

dummy is positive and signicant.




transparency index, and run the same regression in Equation(6)for each sub-sample.Table 7reports

the coefcients ofLFO andLDO. Results show that the relation between LFO and price informativeness

is stronger in markets with a high disclosure score or a high nancial transparency index, regardless of

whetherPINor NONSYNCis used to proxy for price informativeness.Finally, to corroborate the above ndings in our cross-market analyses, we study how the effect of

LFO on the earningsreturns relation varies systematically across markets with different macro

Table 6

Market-level investor protection and the relation betweenLFO and price informativeness. This table presents the results of the


Price informativenessm; i a0 a1LFOm; i a2 LDOm; i a3Price informativeness L1m; i a4LOGMVm; i a5LOGBMm; i

a6 PQSPRm; i a7 TURNOVERm; i a8STDRETm; i a9LEVERAGEm; i a10ROEm; i a11RET12m; i

a12ADRm;

i b1LOGGDPPCm b2MVGDPm b3 PCRGDPm b4GDPGRm b5STDGDPGRm b6STDMRETm k; i

The price informativeness is measured as eitherPINorNONSYNC. The sample is split into two sub-samples based on whether a

markets legal system originates from common law or civil law and whether its anti-self-dealing index is above the medium

value of the index across all markets in the sample. The regression is run separately for two sub-samples. For the main variables

of interests, we report their regression coefcients and the robust t-statistics (in parentheses) calculated from the standard errors

with the market clustering. All variables are dened in theAppendix.

Variable Legal origin Anti-self-dealing index

Common law Civil law High Low

Panel A: Dependent variable PIN

LFO 0.064 (2.48) 0.044 (2.52) 0.069 (3.94) 0.036 (1.18)

LDO 0.045 (1.79) 0.063 (2.77) 0.080 (3.33) 0.022 (0.87)

Nobs 894 850 1120 624Adj-Rsq 35.71% 49.02% 47.91% 32.11%

Panel B: Dependent variable NONSYNC

LFO 0.517 (2.93) 0.140 (0.64) 0.274 (1.57) 0.395 (1.83)

LDO 0.393 (1.95) 0.300 (1.23) 0.421 (2.01) 0.157 (0.65)

Nobs 978 871 1159 690

Adj-Rsq 47.30% 43.34% 44.28% 47.75%

Table 7

Market-level information transparency and the relation between LFO and price informativeness. This table presents the results of

the regression specied in the following equation,

Price informativenessm; i a0 a1LFOm; i a2 LDOm; i a3Price informativeness L1m; i a4LOGMVm; i a5LOGBMm; i

a6 PQSPRm; i a7 TURNOVERm; i a8STDRETm; i a9LEVERAGEm; i a10ROEm; i a11RET12m; i

a12ADRm; i b1LOGGDPPCm b2MVGDPm b3 PCRGDPm b4GDPGRm b5STDGDPGRm

b6STDMRETm k; i

The price informativeness is measured as eitherPINorNONSYNC. The sample is split into two sub-samples based on whether a

markets disclosure score or nancial transparency index is above the medium value of the variable across all markets in the

sample. The regression is run separately for two sub-samples. For the main variables of interests, we report their regression

coefcients and the robust t-statistics (in parentheses) calculated from the standard errors with the market clustering. All

variables are dened in theAppendix.

Variable Disclosure score Financial transparency

High Low High Low

Panel A: Dependent variable PIN

LFO 0.082 (3.65) 0.004 (0.18) 0.052 (2.43) 0.047 (1.85)

LDO 0.064 (2.94) 0.018 (0.69) 0.033 (1.70) 0.085 (2.28)

Nobs 1118 626 1161 583

Adj-Rsq 42.77% 38.01% 42.96% 34.63%

Panel B: Dependent variable NONSYNCLFO 0.513 (2.95) 0.358 (1.76) 0.423 (2.42) 0.363 (1.78)

LDO 0.379 (2.00) 0.402 (1.59) 0.271 (1.45) 0.304 (1.16)

Nobs 1197 652 1244 605

Adj-Rsq 43.68% 39.66% 43.45% 37.14%




governance regimes. We classify markets into two groups along the same dimensions as above, and

estimate separately for each sub-sample the regression of stock returns on earnings innovations and

their interactions with rm characteristics, includingLFO, as specied in Equation(7).Table 8reports

the coefcients of the interaction terms between LFO and current and future earnings changes. The

results indicate that stock prices ofrms with a higher LFO are more informative about contempo-

raneous and future earnings innovations in developed markets and markets with the common law

legal origin, a high anti-self-dealing index, a high disclosure score and a high

nancial transparencyindex. The coefcients on the interactions between LFO and earnings innovations are consistently

positive and statistically signicant in almost all these markets. In contrast, none of the interaction

terms is signicant in emerging markets and markets with a civil law legal origin, a low anti-self-

dealing index, a low disclosure score and a low nancial transparency index.

To summarize, we nd the association between LFO and price informativeness is stronger in

markets with better investor protection and a more transparent informational environment. The re-

sults suggest that large foreign shareholders play a more important informational role in markets with

better investor protection and a more transparent information environment. Ourndings highlight the

importance of macro governance forces in facilitating the information role of large foreign share-

holders. This is consistent with the nding inDoidge et al. (2007)that observable rm characteristics

explain much less of the variation in rmsgovernance ratings than market characteristics, and have

little explanatory power in less developed markets. They suggest that market-level and rm-level

governance mechanisms could be complementary in shaping a rms governance outcome. Viewing

the informational role of large foreign shareholders via monitoring and informed trading as a speci c

rm-level governance force, we argue that its potential effect on the informativeness of stock prices

also depends on the macro governance forces such as investor protection and transparency, and nd

evidence consistent with this argument.

4. Conclusion

With the increase in cross-border equity investments due to nancial globalization, the economic

roles played by foreign equity investors attract signi

cant attention from both regulators and aca-demics. In terms of asset pricing implications, most existing studies focus on the benets of risk-

sharing among investors across borders and the reduction in cost of capital associated with the lift-

ing of the cross-border investment barriers. In this study, we investigate another potential economic

benet of foreign equity investment, the improved informativeness of asset prices, which has not been

Table 8

Market-level governance forces and the relation between LFO and the pricing of earnings information. This table presents the

results of the regression specied in the following equation,

Ri a0 a1DEi a2DE1i b1DEi LFOi b2DE1i LFOi a3 LFOi a4DEi LDOi a5DE1i LDOi a6LDOi a7DEi LOGMVi

a8DE1i LOGMVi a9 LOGMVi X

m

gmMarket dummymi i

The sample is split into two sub-samples according to the following market-level variables: The developed market dummy, thecommon law market dummy, the anti-self-dealing index, the disclosure score, and the nancial transparency index. Then the

regression is run separately for two sub-samples. For the main variables of interests, we report their regression coefcients and

the robust t-statistics (inparentheses) calculated from the standard errors with the market clustering. All variables are dened in

theAppendix.

Variable Market development Legal origin Anti-self-dealing

index

Disclosure score Financial

transparency

DVP EMG Common

law

Civil law High Low High Low High Low

DE*LFO 1.820

(3.33)

0.735 (0.63) 2.105

(3.39)

0.322

(0.42)

1.820

(3.33)

0.735 (0.63) 2.046

(3.56)

0.164

(0.16)

1.922

(3.43)

0.393

(0.42)

DE1*LFO 1.118

(1.98)

0.023

(0.02)

1.059

(1.59)

1.097

(1.45)

1.118

(1.98)

0.023

(0.02)

1.078

(1.86)

0.264 (0.24) 1.331

(2.18)

0.624

(0.71)Nobs 1253 356 1029 579 1253 356 1118 491 1283 326

Adj-Rsq 7.26% 24.68% 9.35% 17.71% 7.26% 24.68% 6.78% 21.72% 11.10% 12.41%




sufciently studied in existing literature. Specically, we study the informational role of large foreign

shareholders who are expected to matter most among foreign equity investors.

Using a cross section of 3189 rms from 40 markets in 2002, we nd evidence that rms with a

higher LFO tend to have more informative stock prices, measured by the probability of informed

trading and price non-synchronicity. Further, stock prices of these rms also tend to be more infor-

mative about contemporaneous and future earnings innovations. These results suggest that largeforeign shareholders contribute to the informativeness of stock prices. Using several measures of the

strength of macro governance infrastructure, we nd thatLFO seems to matter more in markets with

stronger macro governance forces, such as those with a more developed economy, better investor

protection, and a more transparent information environment. These ndings are consistent with macro

governance forces facilitating the effect ofLFOon price informativeness, and echo recent evidence of

the complementarity between market-level and rm-level governance forces inDoidge et al. (2007).

Overall, ourndings shed some light on the roles of foreign investorsin local markets andthe effectsof

capitalmarketopening. Wend evidence consistent with the positive role of large foreign shareholders in

improving the informativeness of stock prices in local markets. Further, macro governance infrastructure

seems to affect the extent to which large foreign shareholders could play this informational role.

Our study is subject to several limitations. First, price informativeness is unobservable and we haveto make inferences based on some empirical proxies. Both proxies of price informativeness, the prob-

ability of informed trading (PIN) and price non-synchronicity (NONSYNC), have been shown to capture

the information content of prices theoretically and empirically in the previous literature (e.g., Easley

et al., 1997;Morck et al., 2000). We also substantiate the ndings with further evidence based on a

more direct and intuitive measure of price informativeness the extent to which stock returns reect

earnings news. However, we are aware that the validity of inferences that we draw from our results

depends on whether our measures of price informativeness adequately capture the extent to which

information is incorporated into stock prices. We caution readers of this important assumption when

interpreting our results. Second, our inferences are based on the cross-sectional regression results,

which reveal correlations rather than causality. We address the endogeneity issue partially by running

regressions of price informativeness measures in year 2003 onLFO and control variables in year 2002.Further, our ndings from cross-market analyses are difcult to be explained by the reverse causality.

However, due to data limitations, we are unable to conduct more sophisticated analyses such as Granger

causality test and (semi) natural experiment, to infer the causality more directly. Given these limitations,

we are unable to completely exclude alternative explanations, and we again caution readers of this issue.

Acknowledgments

We acknowledge nancial support from the University of New South Wales. We thank Fariborz

Moshirian, Peter Pham, and Jason Zein for sharing their data. All errors are ours.

Appendix. Variable Denitions

Ownership variables:

LFO: large foreign ownership in 2002, calculated as the sum of foreign block shareholdings, where a

block is dened as a holding larger than or equal to 5% of a rms issued shares.

LDO: the largest domestic ownership in 2002, calculated as the largest shareholdings by domestic

shareholders.

Price informativeness measures:

PIN: the probability of informed trading in 2003, estimated based on the structural model in Easley

et al. (1997).

NONSYNC: the degree of price non-synchronicity in 2003, calculated as the logistic transformation of

(1 R2) obtained from the market model of stock returns.




Firm-level earnings quality measures:

ACCRUAL: accruals in 2002 scaled by last years total assets.

SMOOTH: smoothness of reported earnings, calculated as the ratio of the standard deviation of oper-

ating income to the standard deviation of cash ows over the 5-year window from 1998 to 2002.

CORR: correlation coefcient between accruals and operating cash ows over the 5-year window from

1998 to 2002.

Otherrm-level variables:

LogMV: the log of the market capitalization measured in US$ at the end of 2002.

LogBM: the log of the ratio of book equity at the end of scal year ending in 2002 to the market

capitalization at the end of 2002.

PQSPR: the average of quoted spreads in 2002.

TURNOVER: the average monthly trading volume in shares divided by the number of shares

outstanding in 2002.

STDRET: the standard deviation of monthly stock returns in 2002/LEVERAGE: the ratio of long-term debt over the common equity at the end of scal year ending in

2002.

ROE: the return on equity for the scal year ending in 2002.

RET12: the annual stock return in 2002.

ADR: the dummy variable taking value 1 if a rm has an ADR and 0 otherwise at the end of 2002.

R: the annual stock return in 2003.

DE: the changes in earnings from 2002 to 2003, scaled by the market capitalization at the beginning of

2003.

DE1: the changes in earnings from 2003 to 2004, scaled by the market capitalization at the beginning of

2004.

Market-level governance measures:

DVPand EMG: dummy variables for developed and emerging markets respectively according to the

classication by the International Financial corporation of the World Bank Group.

Common Lawand Civil Law: dummy variables for markets with a common law legal origin and those

with a civil law legal origin.

Anti-self-dealing index: the index measuring a markets litigation governing self-dealing transactions

and developed byDjankov et al. (2008).

Disclosure score: the score measuring the level and availability ofnancial disclosures based on survey

results reported in the annual Global Competitiveness Report issued by the World Economic Forum.

FollowingGelos and Wei (2005), we average the disclosure scores of 1999 and 2000 and then divide

the average score by 10 to obtain a range from 0 to 1.

Financial transparency index: the index measuring the intensity and timeliness ofnancial disclosures

by rms and developed byBushman et al. (2004).

Other market-level variables:

LogGDPPC: the log of per capita GDP measured in US$ in 2002.

MVGDP: the ratio of stock market capitalization at the end of 2002 to the annual GDP of 2002.

PCRGDP: the ratio of private credit to GDP in 2002 where private credit refers to nancial resources

available to the private sector through loans, purchases of non-equity securities, trade credits and otheraccounts receivables.

GDPGR: the annual GDP growth rate in 2002.

STDGDPGR: the standard deviation of annual GDP growth rates over the period 1998 to 2002.

STDMRET: the standard deviation of monthly market returns in 2002.




References

Admati, A.R., Peiderer, P., 2009. The wall street walkand shareholder activism: exit as a form of voice. Review of FinancialStudies 22, 26452485.

Admati, A.R., Peiderer, P., Zechner, J., 1994. Large shareholder activism, risk sharing, and nancial market equilibrium. Journalof Political Economy 102, 10971130.

Aghion, P., Bolton, P., Tirole, J., 2004. Exit options in corporate governance: liquidity versus incentives. Review of Finance 8,327353.

Aslan, H., Easley, D., Hvidkjaer, S., OHara, M., 2011. The characteristics of informed trading: implications for asset pricing.Journal of Empirical Finance 18, 782801.

Ayers, B.C., Freeman, R.N., 2003. Evidence that analyst following and institutional ownership accelerate the pricing of futureearnings. Review of Accounting Studies 8, 4767.

Bae, K.H., Chan, K., Ng, A., 2004. Investability and return volatility in emerging markets. Journal of Financial Economics 71,239263.

Bae, K., Ozoguz, A., Tan, H., Wirjanto, T., 2012. Do foreigners facilitate information transmission in emerging markets? Journal ofFinancial Economics 105, 209227.

Bekaert, G., Harvey, C.R., 2000. Foreign speculators and emerging equity markets. Journal of Finance 55, 565613.Bekaert, G., Harvey, C.R., Lundblad, C., 2005. Does nancial liberalization spur growth? Journal of Financial Economics 77, 356.Bekaert, G., Harvey, C.R., Lundblad, C., 2007. Liquidity and expected return

1-s2.0-s0261560613000405-main.pdf=3f_tid=3d09d085c2-e095-11e3-b7f3-00000aab0f26=26acdnat=3d1400641827_07615d03fac2a3a9cd8434787b9e033a...

Documents