The Role of Foreign Blockholders in Stock Liquidity: A Cross-Country Analysis

Transcription

1 The Role of Foreign Blockholders in Stock Liquidity: A Cross-Country Analysis Lilian Ng, Fei Wu, Jing Yu, and Bohui Zhang This Version: April 2011 Ng is from University of Wisconsin, Milwaukee, Wu is from Jiangxi University of Finance and Economics, Yu is from University of Western Australia and Zhang is from University of New South Wales. Authors Contact Information: Ng: lng@uwm.edu, (1) ; Wu: fwu@jxufe.edu.cn; Yu: jing.yu@uwa.edu.au, (61) ; and Zhang: bohui.zhang@unsw.edu.au, (61)

2 The Role of Foreign Blockholders in Stock Liquidity: A Cross-Country Analysis ABSTRACT This paper investigates whether and how foreign block ownership affects a company s stock liquidity. Using a sample of international firms from 40 countries worldwide, we document strikingly negative liquidity effects of foreign block ownership in the majority of our sample countries. Further analyses indicate that the presence of foreign block ownership decreases the number of free-float shares available to the public, resulting in reduction in the stock turnover. These results support the notion that the inactive trading of foreign blockholders dampens trading activities and renders trading friction in the stock market. Results also show that the negative liquidity effect of foreign blockholders is most pronounced in firms with opaque information environments and firms from countries with poor information disclosure. These findings are consistent with the notion that foreign blockholders are perceived as informed traders and thereby discourage the participation of liquidity traders. Finally, we show that effective spreads have greater positive impact on the implied cost of capital for firms with higher level of foreign block ownership. Keywords: Foreign Blockholders, Stock Liquidity, and Foreign Ownership JEL Classification Number: G11, G12, G23

3 1 Introduction Foreign equity investment has grown dramatically in the last decades. During the period , the annual foreign direct investment (FDI) flows into and out of the U.S. has increased its initial position by more than 10% (Baker, Foley, and Wurgler, 2008). Likewise, foreign institutional investors hold at least US$3.7 trillion in non-us firms, which accounts for 71% of the total institutional holdings across 27 countries. Despite its size and increasing importance, local governments are still cautious about foreign equity investment in part due to the debate on whether foreign investors may destablize local capital markets and reduce stock liquidity (Tesar and Werner, 1995; Choe, Kho, and Stulz, 1999; Boyer, Kumagai, and Yuan, 2006; Dijk and Vagias, 2010). In this study, we attempt to assess the role of foreign investors with a focus on foreign large shareholders. Specifically, we explore the channels through which foreign large shareholders impact local stock liquidity in a cross-country analysis. Traditionally, foreign equity investment is beneficial to firm valuation in several perspectives such as improving scarce resource allocation efficiency, reducing local risk and exporting professional personnel. Yet, these benefits come at a liquidity cost as foreign investors stakes become large. In the theoretical framework of Bolton and von Thadden (1998) and Kahn and Winton (1998), high stakes incentivize foreigners to intervene with company business in a long term (leading to inactive trading effect) and/or collect private information (leading to information asymmetry effect). Both mechanisms can dampen stock market liquidity, thereby counteracting foreign ownership s value enhancing benefits. Although understanding the negative effect of foreign block ownership is crucial to practice, the literature to date has been silent on the issue. This paper therefore fills this gap and studies whether and how foreign block ownership affects stock market liquidity for firms from 40 countries for the period. To test the relation between foreign block ownership and stock market liquidity, we gather from a rich and recent dataset a firm s block ownership constituents on a monthly basis, where one block holding is defined as the disclosed holding exceeding 5% of total shares. The interest of this research 1

4 rests on the role of foreign large shareholders, but we also consider other constituents of block ownership in order to disentangle the differential effects of domestic versus foreign blockholders. To accurately measure stock liquidity, for the first time we adopt monthly effective spread and quoted relative spread computed based on high frequency trading data as our primary liquidity measures for our wide sample of international stocks. With these key variables, our empirical tests proceed in the following two steps. We first show that firms with high foreign block ownership exhibit low stock market liquidity measured by effective spread and relative quoted spread. This result is robust to the use of various alternative liquidity measures (i.e., Amihud s (2002) illiquidity measure and stock turnover). Further, out of the 39 countries in our sample 1, this negative effect of foreign block ownership on stock liquidity is persuasive in at least 28 countries. We attempt several ways to mitigate concerns over joint determination and reverse causation between stock liquidity and foreign block ownership including controlling for firm fixed effects and lagged liquidity measure (i.e., Granger causality test) and performing two-stage least squares (2SLS) regression. Our results continue to hold after controlling for the endogeneity problem. Given the observed causation from foreign block ownership to stock market liquidity, we turn to explore its economic foundations. We conjecture that the negative liquidity consequence of foreign block ownership arises from two mechanisms. The first mechanism comes from foreign block ownership s adverse impact on trading activities. Bolton and von Thadden (1998) model the costs and benefits of block shareholders management monitoring. The authors postulate that the presence of block ownership immediately reduces the number of shares tradable to public and results in inactive trading of the company s stock. Foreign investors should have similar liquidity effects if their equity stakes are large. In support of this notion, we show that reduction in foreign block ownership leads to an immediate decline in free float shares tradable among ordinary investors and such decline is not due to holding changes of domestic large shareholders. Central to the second mechanism is the information asymmetry effect caused by foreign block 1 Luxembourg is excluded due to lack of firm-month observations in this country. 2

5 holders. Under this mechanism, foreign block investors may dampen stock market liquidity by exploiting their superior information advantage. Several theoretical and empirical studies posit that large shareholders are better informed of the company s activities. In the models of Kahn and Winton (1998), large shareholders have better access to the management and enable to trade at the expense of uninformed liquidity traders. Along this line of argument, we expect the trading by foreign block shareholders induces high transactions costs in compensation for the expected loss by liquidity traders (Kyle, 1985). In support, Heflin and Shaw (2000) and Agarwal (2010), among others, find that larger (institutional) block ownership leads to wider bid-ask spreads and higher adverse selection components of spreads. In an international context, the adverse liquidity effect may be exacerbated when blockholders are foreigners as evidence suggests that foreign investors are sophisticated investors and possess superior information about local stocks (Seasholes, 2004; Froot and Ramadorai, 2008). We test this mechanism by examining the marginal effect of foreign block ownership on stock liquidity conditional on the country- and firm-level information environments. Our results are consistent with an information asymmetry explanation. Specifically, we find that the negative effect of foreign block ownership on stock market liquidity is most pronounced in opaque country and firm information environments, where the information asymmetry problem between informed and uninformed traders is most severe. In the supplementary analysis, we study the marginal effect of foreign block ownership on a firm s implied cost of capital and Tobin s Q. It is likely that the negative relationship between foreign block ownership and stock market liquidity emerges in the cost of capital and firm valuation. Consistent with our expectation, the negative effect of stock illiquidity on implied cost of capital is stronger in firms held by foreign large shareholders, though foreign block ownership has no influence on the relation between Tobin Q and stock illiquidity. This result suggests that foreign block ownership increases implied cost of capital via its adverse effect on stock liquidity. To some extent, it reconciles with prior findings that the decline in cost of capital following financial market liberalization is much less than theory would predict (Bekaert and Harvey, 2000). This study is an important contribution to the literature and furthers our understanding on the 3

6 linkage between foreign block ownership and stock liquidity in an international setting. Substantive empirical work establishes strong support for the negative effect of block/institutional investors on stock market liquidity and indicate two alternative channels that the negative liquidity effect may flow through (Heflin and Shaw, 2000; Brock, Chung, and Yan, 2009). 2 Yet, the majority of these studies base on the US samples and therefore are unable to disentangle the liquidity effect of block investors residency with two exceptions of Li, Moshirian, Pham and Zein (2006) and Rhee and Wang (2009). Li, Moshirian, Pham and Zein (2006) analyze the institutional factors that affect the large shareholdings by financial institutions at the country level. Rhee and Wang (2009) reveal a negative liquidity effect of foreign institutional investors for Indonesian companies. Our study extends this line of research in two dimensions by: (1) examining the role of foreign blockholders on stock liquidity at the firm level across 40 countries and (2) identifying its underlying economic explanations. The findings in this study have significant implications for several strands of literature. In the emerging market liberalization literature, it is puzzling that the benefits of market openness are much less than theory expects. Thus far most studies attempt to explain such limitation from a corporate governance perspective. For example, Stulz (2005) argues a twin agency problems explanation that arise because government and corporate insiders build their own interests at the expense of outside investors. In support, Bae and Goyal (2010) find evidence of higher foreign ownership in Korean firms with better corporate governance after market liberalization. Distinct from these studies, this study takes a very different perspective and looks to the direct impact of foreign block ownership on stock liquidity. Our evidence suggests a negative impact of foreign block ownership on stock liquidity across countries which indicates a liquidity cost associated with foreign equity investments in local markets. We further show that foreign block ownership exacerbates the negative impact of stock illiquidity on the implied cost of capital. These results work in concert 2 Prior literature suggests that block shareholding reduces stock liquidity due to inactive trading and information asymmetry mechanisms. Brockman, Chung and Yan (2009) show that higher block ownership leads to lower stock turnover but little impact on spreads once stock turnover is controlled, in support of inactive trading mechanism. By contrast, another strand of literature (e.g., Heflin and Shaw, 2000) argues that the negative liquidity effect should be attributed to the information asymmetry between block shareholders and other investors. 4

7 with the weak evidence of cost of capital change before and after stock market liberalization. 3 In addition, this paper provides useful insights into the debate over whether foreign investors provide or consume liquidity in local equity markets. On the one hand, foreign investors are blamed for exacerbating stock market illiquidity, especially in the context of financial crises. Boyer, Kumagai, and Yuan (2006) report evidence that the presence of foreign investors facilitates spreading financial crises and reduces local market liquidity in volatile markets. By contrast, several studies find that the participation of foreign equity investors fosters a more liquid stock trading environment. Tesar and Werner (1995) study the foreign portfolio investment in five OECD countries and document a higher stock turnover rate in foreign equity holdings than in local equity holdings. More recently, Dijk and Vagias (2010) expand the sample to foreign equity portfolio investments in 46 countries and report strong evidence that foreign equity portfolio flows lead to higher local stock market liquidity. In this debate, one missing factor is the size of foreign investor s stake. Our findings reconcile the mixed evidence and suggest that foreign equity investments can be harmful to local market liquidity when their ownership becomes large. We also add evidence on the informational role of foreign investors from the perspective of stock market trading. In spite of the above mentioned evidence supporting the information advantage of foreign investors, in the literature there also exists an opposing view that foreign investors are less informed than domestic investors. Choe, Kho, and Stulz (2005) and Hau (2001) show that foreign professional equity traders underperform their local counterparts in Korea and Germany. Our evidence appears to support the information advantage of foreign investors as our results reveal that the inverse relation between stock market liquidity and foreign block ownership is most pronounced in informationally less transparent environments. Finally, we contribute to the stock liquidity literature by improving the accuracy of estimating stock liquidity in non-us countries and by exploring liquidity determinants from a corporate finance 3 Several empirical work (e.g., Amihud and Mendelson, 1986; Acharya and Pedersen, 2005) documents a negative correlation between stock liquidity and returns. Illiquid stocks exhibit high expected required returns due to high transactions costs and/or a high sensitivity to a liquidity risk factor. If foreign block ownership attributes to stock illiquidity, the presence of foreign large shareholdings may result in high expected required returns (i.e., high cost of equity capital). 5

8 viewpoint. Stock liquidity is challenging to estimate. In non-us countries, this becomes more complicated by the fact that in the past we have little access to high frequency transactions data. Using a rich database, we estimate stock liquidity using four different measures: effective spread, relative quoted spread, Amihud s (2002) illiquidity ratio and stock turnover. Among these four measures, spread-based measures are constructed using high frequency transactions data. Using the more reliable liquidity proxies, we not only verify the stock liquidity determinants that previous literature suggests but also propose a new liquidity driver from a corporate finance perspective, that is, firm ownership structure. The rest of the paper is organized as follows: In Section 2 we outline research questions by surveying the literature on firm ownership structure and stock liquidity; Section 3 describes our data sources, variable construction, and summary statistics; Section 4 examines the relationship between stock market liquidity and foreign block ownership and Section 5 explores the underlying channels through which foreign block ownership impacts stock market liquidity; Section 6 examines the effect of foreign block ownership on the relationship between stock market liquidity and the implied cost of capital/firm valuation; Section 7 concludes the paper. 2 Literature review and discussion Foreign large shareholdings can have mixed effects for local market development. Foreign investments are recognized beneficial to local economies in that the openness to foreign investments lowers the cost of capital and enhances the firm value (e.g., Bekaert and Harvey, 2000; Ferreira and Matos, 2008). However, stock markets with prevailing foreign investors may have to take in the loss of stock market liquidity in exchange for these benefits. Drawn upon prior literature, we first present arguments that predict a negative relationship between stock market liquidity and foreign block ownership, followed by arguments indicating a positive relationship between the two. In searching previous studies, we identify two mechanisms, inactive trading effect and information asymmetry effect, which can lead to an inverse relationship between foreign block ownership and stock market liquidity. 6

9 The inactive trading effect arises when foreign large shareholders implement a buy-and-hold inactive trading strategy, using up the trading resources that would otherwise be available to other investors. In the theoretical work by Kahn and Winton (1998), block shareholders have two options: to become an informed trader or to intervene and they have to make a decision depending on the relative costs/benefits ratio. The cost of transacting a block holding is high due to its direct (i.e., inventory shock) and indirect (information signalling) price impact. In our analysis, such cost may be even higher among foreign large shareholders as they are located far from the trading venue and thus lack knowledge of local stocks (as in Choe, Kho, and Stulz, 2005; Hau, 2001). On the other hand, block ownership makes foreigners corporate insiders (e.g., corporate board members). Foreign block holders can gain a large share of private benefits through their privileged access to senior management and effective management intervention. Balancing the costs of trading and benefits of business intervention, foreign blockholders have incentives to hold their ownership in a long run, resulting in inactive trading of their block of shares. Inactive trading of foreign blockholders has a detrimental effect to stock market liquidity because it soaks up the shares available to the public, thereby reducing the trading activity and subsequently the fraction of investors who are willing to actively trade the shares. 4 In Dahlquist, Pinkowitz, Stulz, and Williamson (2003), the analysis of large shareholder ownership in 51 countries reveals that the presence of block holdings largely constrains available shares to investors who are not controlling shareholders, attributing to home bias phenomenon whereby domestic investors bias their portfolios toward their home country stocks. In our context, the reduction of free float shares is also problematic because it dampens the trading activity and diffuses investor base, causing illiquid stock trading. Empirically, Amihud and Mendelson (1986) document that bid-ask spreads are negatively correlated with trading volume and the number of shareholders. Closer to our analysis, Bolton and von Thadden (1998) assert a negative effect of firm ownership concentration to stock liquidity with the principal mechanism of ownership concentration reducing the number of shareholders who can participate in trading. More recently, Brockman, Chung, and Yan (2009) 4 The rationale for concentrated investor base deteriorating stock market liquidity is that reduced potential investors decreases the competition among liquidity suppliers. 7

10 show that block ownership decreases stock market liquidity as it creates real friction as manifested by the reduced trading activities. Another link between foreign block ownership and stock market liquidity is due to an information asymmetry effect, referring to the fact that foreign blockholders are prone to the collection of superior information about the company either because they are blockholders or because they are foreign investors with investment expertise, or both. As blockholders, foreign investors probably have frequent interactions with senior management and therefore gain superior knowledge about the firm. Bhide (1993) cautions the negative impact of block holdings on stock liquidity via this information channel. Following his work, a number of empirical studies, including Heflin and Shaw (2000), Gaspar and Massa (2007) and Rubin (2007), document evidence in favor of the negative liquidity impact of large shareholders through an information effect. Moreover, it is argued that foreign investors specialize in collecting and trading on private information. The existing empirical studies find evidence that foreign investors are informationally advantaged over domestic investors, especially in emerging markets (Seasholes, 2004; Froot and Ramadorai, 2008). To the extent that foreign blockholders are better informed, the trading of foreign investors creates an adverse selection problem for small investors and market makers. As a result of adverse selection concerns of small investors, stock market liquidity decreases in companies with large foreign holdings as other traders widen the bid-ask spreads and trading activities to accommodate the monetary loss of trading against informed traders (Kyle, 1985; Glosten and Milgrom, 1985). In a related study, Agarwal (2010) examines the relationship between institutional ownership and stock liquidity in the US and shows that for institutional ownership levels beyond 35% - 40%, any additional increase in institutional ownership leads to a decrease in stock liquidity. However, there also exist arguments that could justify a positive association between foreign block ownership and stock market liquidity. We group these arguments into three lines of research. First, a company with foreign ownership has a large investor base than a similar company without that kind of ownership (Stulz, 1999). On the premise that foreign blockholders choose to focus 8

11 on short-term trading profits and profit from price discovery, the negative relation between foreign investors and stock market liquidity has to be reversed as the enlarged investor base effectively promotes competition among liquidity suppliers. Second, foreign blockholders can improve stock liquidity through its beneficial association with corporate governance and information environment, particularly in emerging markets. Foreign blockholders can apply their management expertise to restrain corporate insiders from diverting the private benefits of control as well as deter them from informed trading. Wei, Xie, and Zhang (2005) suggest that foreign ownership serves as an effective corporate governance mechanism and enhances firm value in China. Alternatively, the presence of foreign investors may prompt the management to disclose price-sensitive information in a timely manner (Gul, Kim and Qiu, 2010) and attract the coverage of security analysts (Bae, Bailey, and Mao, 2006). Taken together, the reduced insider trading and improved information environment accentuate the information uncertainty faced by liquidity traders, encourage their trading activities and therefore improve stock market liquidity. Finally, the participation of foreign blockholders, as arguably better informed investors, bolsters the competition among informed traders and thus fastens the information incorporated into stock prices (Subrahmanyam, 1991). Such competition is beneficial from liquidity traders viewpoint because it lowers their expected trading loss by making stock prices more informationally efficient, thereby inducing the participation of uninformed traders and resulting in improved stock liquidity. In summary, we expect foreign ownership to have a negative impact on stock liquidity due to inactive trading effect and information asymmetry effect. However, the negative liquidity effect of foreign block ownership has to be weighed against its benefits in enlarging the investor base, improving corporate governance and stock price information efficiency. The unclear balance between the costs and benefits leaves how foreign block ownership relates to stock liquidity to an empirical test. 9

12 3 Data and sample description 3.1 Sample selection Our international sample covers 67-month period beginning from June 2002 when firm-level foreign strategic blockholding ownership data become first available in Thomson Reuters Datastream database until December Datastream collects from 11 primary sources (e.g., SEC filings and the UK register) the total free float shares and the constituents of firm block share holdings for over 150,000 stock issues around the world, where block holding refers to a disclosed holding exceeding 5% of total number of shares and therefore effectively excluded from normal dealings. The block holders are classified into the following categories: government and government agency, another company, pension fund, investment company, employee/family, foreign investors, and others. These block holdings data are updated monthly at month end, which enables us to examine the liquidity effects of foreign block holdings with great accuracy. In this study, we focus mainly on the level of foreign block holding in a company on a monthly basis and consider all the firms with available block holdings information. Parallel to the firm block holdings data, we measure two primary stock liquidity variables, monthly percentage effective and quoted relative spreads, using transactions data from Thomson Reuters Tick History (TRTH). This database specializes in providing intraday trading data for international public companies across a wide range of countries. In order to construct other alternative liquidity measures and control variables, we also collect additional daily trading data including stock trading volume (both number-based and dollar value-based), the number of shares outstanding, stock price and return data from Datastream and firm financial characteristics data from Worldscope via Datastream. The definitions of all the variables and their corresponding data sources are detailed in the appendix. Our initial sample starts with firm observations with non-missing foreign block ownership in 40 countries. We then apply several filters to form the final sample. First, we drop firm-month 5 Datastream database started to collect the constituents of firm ownership data including foreign ownership data since May However, the prevalence of zero foreign ownership observations in May 2002 raises our concerns. To avoid any data recording errors, we start our sample from June

13 observations with the sum of all the ownership constituents exceeds one. 6 Following prior literature (e.g., Chordia, Roll, and Subrahmanyam, 2000), we also exclude from our sample the observations with spread-based liquidity measures higher than 0.40 and observations with missing data for control variables. Our final sample consists of firm-month observations covering firms from 40 countries (see column 1 of Table 1) Variable construction Key to our analysis is the measurement of foreign block ownership and stock market liquidity. We measure foreign block ownership using the very recent strategic ownership dataset from Datastream. This dataset divides companies equity holdings into free float and block holding categories on basis of the size of equity holdings. A block holding is defined as the disclosed holding that exceeds 5% of the total shares outstanding. Under block ownership category, the ownership structure is further classified in accordance with the residency of the equity investors. Specifically, we are able to track the percentage of foreign block ownership versus a variety of domestic investors such as government/government agency, another company, pension fund, investment company and employee/family on a monthly basis. This feature allows for disentangling the liquidity effect emerging from foreign investors only. Table 1 reports the sample mean of percentage foreign block ownership of 3% and standard deviation of 10%. The average level of foreign ownership is considered significant, given that many firms have zero foreign ownership in our sample. 8 Additionally, the foreign ownership is relatively higher in developed countries, where barriers to market liberalization are much less. For the first time we utilize high frequency transactions data and construct two spread-based measures as our primary liquidity proxies for a wide range of non-us countries. These spread-based 6 The ownership constituents include the percentage free float shares, the percentage block holdings of government/government agency, another company, pension fund, investment company, employee/family, foreign investors and others. 7 We include financial firms in our analysis for purpose of covering a broad sample. However, the empirical results are qualitatively the same after excluding financial firms from the sample. 8 To avoid any unnecessary loss of information, we include firm-month observations with zero and non-zero foreign block ownership. In untabulated results, we re-estimate all the regression models using a restricted sample of firms with non-zero foreign block ownership. The results based on the restricted sample are qualitatively the same and available upon request. 11

14 variables are direct measures of stock market liquidity and thus markedly improve the accuracy of estimating stock liquidity. Several studies in the microstructure literature, (e.g., Lesmond, 2005; Goyenko, Holden, and Trzcinka, 2009) use spread-based measures as a benchmark to assess the reliability of other stock liquidity estimates. Our first spread-based measure is the round-trip effective spread which is calculated for each trade as twice the absolute value of the difference between trading price and midpoint of bid and ask prices scaled by trading price. The second measure used is the quoted relative spread which is equal to the difference between bid and ask prices scaled by the midpoint of bid and ask prices. To match with the monthly frequency of foreign block ownership data, we apply the following algorithm to convert transaction-specific spread measures into monthly stock liquidity estimates. On a given trading day, the respective daily spread measure is set equal to the dollar-volume-weighted spread measures within that day. We then average the daily spread measure to obtain the monthly stock liquidity proxy. The descriptive statistics of these two liquidity measures are summarized in Table 1. The mean values of both the effective and relative spread measures range from 0.00 to 0.05, with the lowest average in the US. The mean effective and relative spreads of the entire sample are the same and equal to Admittedly, none of the stock liquidity measures can accurately capture stock liquidity (Korajczyk and Sadka, 2008). To ensure that our results are not driven by stock liquidity measurement, we further consider two additional liquidity measures which are widely used in the literature: Amihud s (2002) illiquidity ratio and stock turnover. Amihud s (2002) illiquidity measure is equal to the absolute value of stock return divided by dollar trading volume on a given trading day. In addition, we calculate daily stock turnover as the number of shares traded divided by the number of common shares outstanding. Again, we transform daily liquidity measures into monthly measures by taking a natural log of the simple average of respective daily liquidity measures in each month. From Table 1, we observe that stocks in the China and US, on average, exhibit the lowest logistically transformed Amihud s (2002) illiquidity ratio (-4.47 in China; in Spain; in the US), much lower than the sample average of The sample mean of logged stock turnover is -3.40, ranging from in China to in Luxembourg. The cross-country pattern in stock 12

15 turnover variable shows some discrepancy with that in spread-based measures, which prompts us to use all these measures in order to capture all the aspects of stock market liquidity. In the baseline regression tests, we also include a host of control variables which have been shown to be correlated with a stock s liquidity in prior studies: past 12-month stock return (Hameed, Kang, and Viswanathan, 2010) and stock return volatility (Amihud and Mendelson, 1980), an American Depository Receipt (ADR) dummy variable for the US cross-listing (Domowitz, Glen, and Madhavan, 1998; Foerster and Karolyi, 1999; Lang, Lins and Miller, 2003) 9, a dummy variable for the inclusion in major market index (Beneish and Whaley, 1996), stock dividend yield, bookto-market ratio and firm size measured by the log of market capitalization in the US dollars. Several interesting findings emerge from Table 1. Across all the 40 countries, stocks, on average, experience positive market returns in the past 12 months (Ret 12m), with several emerging markets exhibiting the highest average 12-month past returns (e.g., in Brazil, average Ret 12m = 0.05; in Poland, average Ret 12m = 0.06). Several countries display strong preference for US stock exchange cross-listings. In Argentina, Canada, Ireland, Israel, and Mexico, more than 15% local stocks are cross listed on the US stock exchanges. Finally, it appears that large companies are predominant in our sample based on two noticeable observations: (1) out of the entire sample, 33% of companies are included in the MSCI s country index; (2) the average firm size measured by market capitalization is 5.31 which amounts to US$ 213 millions. Table 2 presents Pearson correlation coefficients between variables in our baseline model regression. First, we observe very strong correlation between two spread-based measures (correlation coefficient = 0.907), which suggests the validity of both liquidity measures. Amihud s illiquidity seems to be a more appropriate liquidity measure than stock turnover as the correlation with spreadbased measures is about 0.60, higher than the correlation of stock turnover with two spread-based liquidity measures. More important, we observe a negative and significant correlation coefficient 9 We control for this dummy variable for a variation of effects following the ADR listing on the US exchanges: (1) Trading orders being diverted from home market to the US stock market (Domowitz, Glen, and Madhavan, 1998); (2) Increased investor recognition and expanded investor base (Foerster and Karolyi, 1999); (3) Enhanced information environments for cross-listed firms (Lang, Lins, and Miller, 2003). Due to the mixed effects of US market cross-listing, the sign of the ADR dummy is unpredictable in our baseline specification. 13

16 between foreign block ownership variable and spread-based stock illiquidity measures which sheds some light on the multivariate analysis on the impact of foreign block ownership on stock market liquidity. Finally, the correlations between the variables are only moderate, which rules out concerns about multicollinearity issues in the following regression estimations. 4 Empirical results Following the preliminary analysis in the previous section, we conduct a series of multivariate analyses to understand the impact of foreign ownership on stock market liquidity in a cross-country sample. We first present the cross-country firm-level regressions in Tables 3-4, and then verify the robustness of the empirical results in Tables Stock liquidity and foreign block ownership In the following empirical analysis, we model the level of foreign ownership as a determining factor to stock market liquidity and control for other firm characteristics that may have an impact on stock liquidity. Formally, we state our baseline ordinary least squares (OLS) regression model as below: Liquidity i,t = α 0 + α 1 F O i,t 1 + α 2 Ret 12m i,t + α 3 RV ol i,t + α 4 ADR i,t + α 5 Index i,t +α 6 DY i,t + α 7 BM i,t + α 8 Size i,t + Other controls + ε t, (1) where Liquidity i,t, alternatively, represents the average percentage effective spread and percentage relative spread for firm i in month t, and F O i,t 1 measures the percentage foreign block holding for firm i at month t 1. The list of contemporaneous control variables includes average monthly stock return over previous 12 months (Ret 12m), stock return volatility over previous 12 months (RV ol), an ADR dummy (ADR), the inclusion in MSCI country index dummy (Index), stock dividend yield (DY ), the log of book-to-market ratio (BM), the log of market capitalization (Size). Other controls are country, month, and industry fixed effects in panel regression model estimations or country and industry fixed effects in Fama-MacBeth regression estimations. 14

17 To begin with, we estimate Equation (1) using an OLS panel regression in a pooled sample of time-series cross-sectional firm-month observations with two alternative stock liquidity measures. Models (1) - (2) of Panel A in Table 3 present the regression estimates of the baseline specification with effective spread as dependent variable. Model (1) reports a positive association between stock market illiquidity and foreign block ownership after controlling for other firm characteristics. The coefficient of F O t 1 (0.005) is positive and statistically significant (t-statistic = 3.67), which indicates that the presence of foreign block ownership, on average, widens the effective spread and reduces stock market liquidity. From an economic perspective, a one-standard-deviation increase in foreign block ownership leads to a five basis point decrease in percentage effective spread. As noted in Agarwal (2010), the magnitude of block holdings matters to stock market liquidity as it alters the information impact of block holdings on the participation of liquidity traders. To gain further insights into the relationship between stock liquidity and foreign ownership, we disaggregate foreign ownership into low and high ownership levels using 15% as a cutoff. 10 In Model (2), we observe that the coefficient of F O t 1 (0.000) loses its significance (t-statistic = 1.23) if foreign block ownership falls below 15%. In other words, the negative liquidity effect of foreign block ownership only exists at high level of foreign block holdings. Furthermore, the regression results in Models (1) and (2) indicate the reliable selection of control variables. As Model (1) shows, the coefficients of most firm characteristic variables are statistically significant with an exception of 12-month stock return variable (Ret 12m t ) and their signs are consistent with expectation. The coefficient of RV ol t is statistically significant and positive at 1% level, which suggests the uncertainty among traders weakens stock liquidity. The coefficient of ADR t has a positive coefficient, suggesting that companies that are cross-listed in the US have lower stock liquidity in local market. This is consistent with a trading migration effect following the US cross-listings as addressed in Domowitz, Glen, and Madhaven (1998) and Levine and Schmukler (2007), whereby international cross-listing diverts investors of domestic shares abroad. Surprisingly, stocks of companies included in the MSCI country index display lower stock liquidity, even though 10 The selection of foreign ownership cutoffs is arbitrary. Yet, our results remain if we use 10% or 20% as the new thresholds. 15

18 these companies are argued to have high visibility to market investors. 11 Finally, we observe the negative coefficients on DY t and Size t and a positive coefficient on BM t, which means that stock market liquidity tends to be higher for dividend-paying firms, large sized firms and value firms. Our results are robust to another spread-based liquidity measure - quoted relative spread. In Models (3) - (4), we repeat the panel regression analysis with the dependent variable of quoted relative spread, RSpread. The coefficient of F O t 1 continues to be significant at 1% level and the result in Model (4) suggests that its statistical significance is mainly driven by companies with foreign block ownership above 15%. In addition to the preliminary results, we further consider two additional liquidity measures, Amihud s illiquidity ratio (Illiquidity) and stock turnover (T urnover), computed based on daily stock trading data for two reasons. First, the measurement of stock liquidity is a rather challenging subject. The use of two alternative liquidity measures allows for a check of robustness of our preliminary results. Second, as discussed in Section 2, foreign block ownership can worsen stock market liquidity through inactive trading mechanism or information asymmetry mechanism. Since they measure total transactions costs, effective spread and relative spread cannot disentangle the sources through which foreign block ownership reduces stock market liquidity. However, daily Amihud s illiquidity measure is equal to the absolute value of stock return over dollar trading volume. By definition, this measure captures the price impact of trading and thus can act as a rough estimate for the information asymmetry component of stock illiquidity. Stock turnover, on the other hand, is a straightforward measure for trading activity and the result based on this measure will shed some light on the inactive trading mechanism. Models (5)-(8) in Panel A contain results of alternatively regressing Amihud s illiquidity ratio and stock turnover on foreign block ownership using the pooled panel regression. We find preliminary evidence in support of both inactive trading effect and information asymmetry effect. Consistent with the information asymmetry effect, Amihud s illiquidity ratio increases with foreign block ownership. To the extent that Amihud s illiquidity measure proxies for the information 11 It should be noted that the sign of Index t variable reverses if we use Amihud s (2002) illiquidity ratio and stock turnover to proxy for stock market liquidity (see Models (5)-(8) in Table 3). 16

19 asymmetry faced with liquidity traders, the existence of foreign block ownership exacerbates the information asymmetry problem faced by liquidity traders. Interestingly, the information asymmetry effect is driven by companies with high level of foreign block ownership (see Model (6)). In addition, stock turnover decreases in foreign block ownership as the coefficient of F O t 1 (-0.691) is statistically negative (t-statistic = -8.01). Again, the negative effect to stock turnover mainly comes from firms with high level of foreign block ownership. In fact, foreign block ownership increases stock turnover when the foreign block ownership falls under 15% (see Model (8)). Since we measure foreign ownership and various stock liquidity measures on a monthly basis, panel regression results might not be statistically reliable. Thus, in Panel B of Table 3, we apply Fama-MacBeth regression procedure to analyze the relationship between foreign block ownership and stock liquidity. We find that the coefficient of F O t 1 remains significant at 1% level, no matter how the stock liquidity is measured. The only slight difference from previous findings is that the coefficient of F O t 1 turns significant at 5% if the dependent variable used in the model estimation is effective spread (ESpread) and if foreign ownership falls below 15%. Due to the long panel of 67-month observations in our sample, we only report the Fama-MacBeth regression results in the remaining tests as this regression procedure is more appropriate to identify cross-sectional correlations over time. Thus far, all the regression results are based on a pooled sample. It is likely that the results are dominated by observations from certain large countries such as the US and the UK. We employ two approaches to mitigate this concern. First, we conduct tests on the relationship between foreign block ownership and stock liquidity in each individual country. Specifically, we estimate the baseline specification (Equation (1)) with four different liquidity measures (i.e., effective spread, relative spread, Amihud s illiquidity ratio and stock turnover) in each of the 39 sample countries (Luxembourg is excluded due to insufficient firm-month observations.). To conserve space, Table 4 presents only the estimated coefficients of F O t 1 in each country. In Panel A with panel regression results, out of the 39 countries, the coefficient of F O t 1 is positive in countries and significantly positive in countries at the 5% level depending on the measure of stock liquid- 17

20 ity considered. The results become even stronger using Fama-MacBeth regression methods. In Panel B, a total of countries exhibit significant and expected coefficients on F O t 1. Another way to alleviate the sample representation problem is to run a country-level regression analysis. In unreported results, we regress the average effective spread measure in one country on average foreign ownership ratio controlling for country fixed effects. Consistent with firm-level results, the coefficient of F O t 1 is positive and statistically significant (coefficient = 0.046; t-statistic = 8.54) in the country-level regression estimation. To summarize, we conclude that foreign large share holdings induce low stock liquidity, particularly when foreign block ownership exceeds 15%. This finding is persuasive across the majority of the 39 countries and is robust to the choice of stock liquidity measurement and regression estimation method. 4.2 Additional tests We carry out further robustness checks to substantiate the preliminary findings that foreign block holding has a negative impact on stock market liquidity. In particular, we examine the liquidity effect of foreign investors using an alternative database, address an endogeneity issue in Table 5, and finally jointly estimate the impacts of foreign versus domestic block ownership in Table Alternative foreign ownership database Though it has a wide coverage of firms all over the world, foreign block ownership data from Datastream may be limited due to the wide variations in the reporting requirements and accuracy of firm disclosures across countries. In order to alleviate concerns over the reliability of foreign ownership database, we repeat the previous analyses using foreign ownership data from FactSet/Lionshare database. The FactSet/Lionshare database tracks the institutional holdings for international stocks on quarterly, semiannual, or annual basis and this data has been widely used in recent studies including Ferreira and Matos (2008) and Ferreira, Massa and Matos (2010). The FactSet/Lionshare institutional holdings data identify the domicile country of financial institution, allowing us to 18

21 compute the percentage of foreign institutional ownership. For the period where FactSet/Lionshare data are available, we estimate the baseline regression specification on an annual basis using effective spread and quoted relative spread as dependent variables. In unreported regression results, we observe that the coefficient of F O t 1 is 0.014/0.028 and t-statistic is 4.97/7.01 with effective spread/relative spread as dependent variable, respectively. 12 These new results not only confirm the validity of foreign block ownership data from Datastream but also corroborate previous studies that foreign institutional ownership leads to higher effective and relative spreads and lower stock market liquidity The endogeneity problem The potential endogeneity problem may also mask the actual relationship between foreign block ownership and stock liquidity. In particular, we are concerned that foreign ownership and stock market liquidity are jointly determined by certain firm characteristics. However, our previous findings might be an artifact of missing out these firm characteristics. A more serious issue concerning the endogeneity problem is the reverse causality from stock liquidity to foreign block ownership. On basis of our results so far, the reverse causality seems unlikely. We measure foreign block share holdings one period lag of stock market liquidity measures in all the regressions. The time difference between liquidity and foreign ownership measures to some extent mitigates concerns about the reverse causality problem. Admittedly, the level of foreign ownership is highly persistent over time. Simply analyzing the lead-lag relationship between stock liquidity and foreign block ownership might not completely resolve the endogeneity problem. In the complementary analysis, we implement three regression 12 Using the FactSet/Lionshare database, Wei (2010) finds a negative relation between logistically transformed Amihud s measure and foreign institutional ownership for the period , seemingly contradicting our results. To rule out any estimation errors, we replicate Wei (2010) in our sample period by regressing the logistically transformed Amihud s illiquidity measure on foreign institutional ownership and other control variables in Equation (1). The further analysis aligns our conclusion with that in Wei (2010). The full sample result shows a negative and significant coefficient on F O (coefficient = ; t-statistic = ), consistent with Wei (2010). However, in the subsample of stocks with foreign ownership higher than 5%, the regression result suggests a positive coefficient on F O (coefficient = 1.434; t-statistic = 2.78). Again, this set of results implies that the relation between foreign equity investment and stock market liquidity is not always linear and varies with the size of foreign ownership. 19