Friend or Foe? Foreign Investors and the Liquidity of Six Asian Markets

Authors


Corresponding author: Diego A. Agudelo, Kra. 49 7 Sur 50. Finance Department, EAFIT University, Medellin, Colombia. Tel: +57 4 2619510, Fax: +57 4 2619294, email: dagudelo@eafit.edu.co.

Abstract

Studying foreign flows and the liquidity of six Asian markets and the Johannesburg Stock Exchange, we provide evidence of two contrary effects of foreigners on liquidity. On the one hand, foreign trade has a negative but transitory impact on the overall liquidity of the market on a daily basis, consistent with foreign investors demanding liquidity more aggressively than locals and incorporating market-wide information. On the other hand, the overall share of foreign ownership in the market is positively related to improved liquidity, consistent with foreigners improving liquidity provision and sending a positive signal to the market on transparency and monitoring. Overall, the results portray foreign investors as aggressively demanding liquidity in the very short term, but having a lasting positive effect on the liquidity of emerging markets.

1. Introduction

Liquidity is a critical factor to consider in emerging stock markets. From a practitioner’s point of view, liquidity is a first-order factor in evaluating investment opportunities in emerging markets: the substantial returns potentially earned can be substantially reduced after accounting for trading costs. Moreover, institutional investors still regard liquidity as one of the major obstacles to investing in emerging markets, along with poor disclosure standards and weak regulatory regimes, as reported by Bekaert & Harvey (2003) and Freeman & Bartels (2000). From an academic perspective, liquidity is a fundamental characteristic of stock market development. Stock market liquidity is linked to economic growth (Levine, 2003); higher liquidity allows firms to raise capital at a lower cost (Ellul & Pagano, 2005), and increases the incentives for financial analysts to acquire information (Hölmstrom & Tirole, 1993). Furthermore, Bekaert et al. (2005) report that liquidity is a priced factor of returns in emerging markets. Liquidity considerations are fundamental to studying efficiency and informed trading and to evaluating momentum strategies. For all its importance, the liquidity of emerging markets has been little studied, mainly because of the lack of good quality data. As indicated by Lesmond (2005), availability of daily bid-and-ask quotes is limited, and only a few studies have been able to use transaction data.1,2

In the present paper, we explore the relationship between foreign investors and the liquidity of emerging markets. We provide evidence of two empirical regularities. On the one hand, foreign trade is negatively related to the overall liquidity of the market, on a daily frequency, for three of seven emerging markets. Although the effect is economically small, it is statistically significant. On the other hand, increases in the overall share of foreign ownership in the market are positively related to improved liquidity in two of the seven emerging markets. Overall, the results portray foreign investors as aggressively demanding liquidity, but having a positive effect on the liquidity in short horizons.

Allowing foreign investors in emerging markets has been controversial since the instigation of liberalization processes. Krugman (1998) and Stiglitz (2000) express fears of excessive volatilities and inflation, increased boom and bust cycles, and appreciation of exchange rates caused by the instability of foreign investors’ flows and holdings. Unlike foreign direct investment, which is widely regarded as beneficial, foreign portfolio flows are considered potentially damaging for emerging economies. Foreign portfolio investments, sometimes dubbed “hot money,” might flee from a developing country at the first sign of trouble during times of financial stress, further disrupting its capital markets. In particular, foreign flows have been accused of causing contagion, the transmission of financial instability across emerging markets, during the crises in the second half of the 1990s. Furthermore, some economists have called for increasing regulation on foreign flows to emerging markets (Ito & Portes, 1998; Eichengreen, 1999; French-Davis & Griffith-Jones, 2002; Rubin & Weisberg, 2003).3 Although most of the emerging markets identified by Standard & Poors (2004) currently have few or no direct barriers to the entry of foreign investors, countries such as China, Colombia, India, Indonesia, the Philippines, Saudi Arabia, Taiwan, and Thailand have either formal restrictions for foreign outflows or ceilings to foreign ownership.

Hence, the question of whether foreign investors are “friend or foe” of the emerging markets is still relevant. Do they make the markets more efficient, liquid, transparent, and dynamic or does their speculative behavior make markets more volatile, unstable, and prone to external shocks and financial crises?

Empirical answers have mostly focused on the effects of the liberalizations of emerging markets that took place in the late 1980s and early 1990s. Some of the results have been unambiguously positive: the evidence suggests that, upon liberalization, the incoming foreign speculators benefited the emerging markets. The cost of capital decreased, as reported by Miller (1999), Errunza & Miller (2000), Bekaert & Harvey (2000), and Henry (2000), and the efficiency of the markets increased, as reported by Kim & Singal (2000). There is also a positive association between the degree of openness of an emerging market and the overall information environment, as reported by Bae et al. (2005). Bekaert et al. (2002) show a more direct relation between foreign inflows to the emerging markets and the decrease in the cost of capital after the liberalization. Furthermore, Huang & Shiu (2005) report that foreign institutions in Taiwan are regarded as more sophisticated than locals, better at forecasting firm performance, and active at monitoring management and demanding regulatory improvements.

Concerning other aspects, the evidence has been mixed. The hypothesis of excessive volatility upon the liberalization has been mostly unsupported by the data, as in Bekaert & Harvey (1997) and Kim & Singal (2000). Kim & Singal (2000) also offer evidence that the volatilities of inflation and exchange rate in emerging economies have dropped since stock market liberalization. However, Dvorak (2001) shows the opposite: in the post-liberalization era, foreign trading can be related to increasing volatility in emerging markets. Moreover, Bae et al. (2004) find that firms more open to foreign ownership are more volatile, while Domowitz & Coppejans (2000) report similar results for firms that cross-list. Concerning the role of foreign flows on contagion, the evidence is also ambiguous: Kaminsky et al. (2004) find evidence consistent with foreign investors taking part in it, whereas Karolyi (2003) and Choe et al. (1999) fail to find such evidence. Kaminsky & Schmukler (2003) report evidence of emerging countries experiencing enlarged boom and bust cycles upon liberalization but also argue that the liberalization should lead to increasing stability in the long term.

Although the abovementioned research mostly focuses on the liberalization process, there are very few studies that consider the effects of foreign flows on emerging markets since liberalization.4 Therefore, the present paper aims to answer whether the foreign investors have an overall positive or negative effect in the liquidity of a set of emerging markets. Answering this question will add to our understanding of the role of foreign investors in emerging markets, and the associated benefits and costs. This research also intends to test in a wider sample of countries, the country-specific findings of Huang & Shiu (2005) on foreign investors improving the information of the Taiwanese firms, and Choe et al. (2005) and Agudelo (2005) on foreign investors reducing liquidity on the firms and days where they trade the most, in Korea and Indonesia respectively. This question also has a direct implication for the institutional investors that engage actively in herding, as reported by Sias (2004) and Bowe & Domuta (2001), because decreasing liquidity in times of active foreign trading would make herding more costly.

To answer this question we, first, provide a case study on the differential effects of foreign trading and foreign ownership on liquidity on a specific country, Indonesia. Next, we examine the market-wide effects of foreign trading and ownership for a group of seven emerging markets in a longer period of time. As a measure of liquidity, the present paper uses the proportional quoted bid-ask spread.

The effects of foreign investors on the liquidity of emerging markets have been virtually unexplored, to our knowledge, by the extant literature, probably due to the difficulty in obtaining liquidity and foreign investor trading data. The present paper overcomes this difficulty by using a detailed data set on the Indonesian stock market that includes foreign trading and ownership on a firm-day basis, as well as the Bloomberg database for daily data on bid-ask spreads and market-wide foreign activity for a group of emerging markets.5 The papers most related to this research are Froot et al. (2001) and Richards (2005), which find that foreign trading causes increased price pressure in emerging markets but do not investigate the effects on liquidity. Choe et al. (2005) find that foreign investors have a larger price impact than locals on Korea, which is not driven by information but by investment style.6

The present paper obtains two secondary results. First, we are able to explore the relationship between market liquidity and market return, volatility, and turnover, providing an out-of-sample test of the findings of the systematic liquidity literature in the USA (e.g. Chordia et al., 2001). Second, we are able to update the studies on the effects of foreign investors in emerging markets on prices (e.g. Richards, 2005) and volatility (e.g. Bekaert & Harvey, 1997).

The remainder of this paper is organized as follows. Section 2 explains the hypothesis in the context of the empirical and theoretical literature, Section 3 presents the data, Section 4 explains the econometric model and discusses the results, and, finally, Section 5 provides the conclusion.

2. Hypothesis

We analyze the effects of foreign flows and foreign ownership on the liquidity of a group of emerging markets both at the firm and market levels. In principle, we do not expect foreign investors to be any different from locals. In an efficient and transparent market, where both groups have access to the same information and are equally likely to be informed, foreign investors should be indistinguishable from domestic investors. Even in that stylized case, we can expect a minor beneficial effect of foreign flows: the inventory model of Ho & Stoll (1981) indicates that higher trading volumes are associated with higher liquidity, by reducing the inventory cost of liquidity providers, which is supported by the empirical evidence of Stoll (2000) and Lesmond (2005). By contrast, an increased share of the traded volume by foreign investors should not have a significant effect on the liquidity of emerging markets, either at the firm or market level, once we control for the effects of volume.

However, there is mounting evidence that foreign investors are more likely to be informed than the average local.7 In that respect, Grinblatt & Keloharju (2000) provide evidence for Finland, Seasholes (2004) and Huang & Shiu (2005) for Taiwan, and Agudelo (2005) for Indonesia. Consequently, the asymmetric information models of Kyle (1985), Glosten & Milgrom (1985), and Easley & O’Hara (1987) imply that foreign investor trading will reduce liquidity because it increases the probability of informed trading. Alternatively, Bonser-Neal et al. (2005) and Choe et al. (2005) present evidence that foreign investors in Indonesia and Korea, respectively, demand liquidity more aggressively than locals, which is a non-informational explanation for the potential negative effects. Therefore, to the null hypothesis of no effect from foreign investor trading we oppose the following alternative hypotheses:

Hypothesis 1.  Foreign investor trading has a negative effect on the liquidity of individual stocks. This will be measured on a panel data regression of a liquidity measure against measures for the intensity of foreign investor trading and suitable control variables. This hypothesis implies a positive coefficient of the intensity of foreign investor trade.

However, we are also interested in exploring the market-wide effects of foreign investors on liquidity. Foreign institutions are presumably better informed than locals in macrovariables, as suggested by Seasholes (2004) for Taiwan, by Agudelo (2005) for Indonesia, in a cross-country study by Chan & Hameed (2006), and in anecdotal evidence from financial analysts in emerging markets. Hence, the overall level of asymmetric information of firms in a market should increase in those times when foreign investors trade more intensively in a given direction. Moreover, Bae et al. (2005) show a positive relationship between the US portfolio flows and the information environment in 25 emerging markets. All in all, one can expect market-wide foreign trading to be positively associated with market-wide liquidity.8 Market-wide (systematic) liquidity is a notion explored by Chordia et al. (2000, 2001)9 and Hasbrouck & Seppi (2001) in the US stock market. As an alternative explanation, foreign trading might affect market-wide liquidity, not through changing market-wide asymmetric information but by causing increased order imbalance. Therefore, the second hypothesis is:

Hypothesis 2.  Market-wide foreign investor trading has a negative effect on market-wide liquidity. This will be measured with a vector autoregressive (VAR) model on seven emerging markets that includes market-wide return, volatility, and turnover. This hypothesis implies that shocks in foreign investor trading Granger-cause positive shocks on the measures of market-wide liquidity: average proportional spread.

Finally, there are reasons to expect a positive relationship between foreign ownership and liquidity. As a short-term effect, if foreign buying is considered as a positive signal in emerging markets (Huang & Shiu, 2005; Richards, 2005) increasing foreign ownership might lure local uninformed investors and/or liquidity providers, improving the trading activity and liquidity, in the same way that positive returns do, as in Griffin et al. (2006). However, as a long-term effect, the results of Huang & Shiu (2005) in Taiwan suggests that both at the market and firm levels, foreigner ownership is a signal of improved information transparency and improved monitoring, and that increased monitoring by foreigners will limit insider trading. In addition, Bae et al. (2005) report that the fraction of stock available to foreign investors is related to an improved information environment at the market level, for 25 emerging markets. Therefore, given that increased foreign ownership is related to reduced information asymmetry at the firm or country level, it will attract larger numbers of uninformed traders, with a subsequent improvement in liquidity.

However, if foreign investors are more likely to be informed than locals, as mentioned previously, increasing foreign ownership should discourage uninformed investors, and be negatively related to liquidity. In that sense, the evidence shows that foreign investors in emerging markets are predominantly institutions, and it has been established that institutions are, overall, better informed than individuals.10 Alternatively, if foreign ownership increases a firm volatility and exposure to worldwide factors, as suggested by Bae et al. (2004), this might have a harmful effect on liquidity because volatility decreases liquidity in the model of Ho & Stoll (1981), and the empirical papers of Chordia et al. (2005) and Lesmond (2005). Nonetheless, Breen et al. (2002) provide evidence of a cross-sectional positive relationship between institutional ownership and liquidity. For the third hypothesis, we assume that the beneficial effect of foreign ownership on liquidity dominates.

Hypothesis 3.  Foreign ownership is positively related to liquidity, both at the firm and market-wide levels. This will be tested in a panel data model at the firm-day level in Indonesia, and in a daily VAR model for seven emerging countries. The hypothesis implies, on the one hand, a negative coefficient of foreign ownership in the firm-level panel data regression of the liquidity measures, and, on the other hand, that shocks in foreign ownership Granger-cause negative shocks to the measures of market-wide liquidity.

3. Data

3.1. Firm-Level Data from Indonesia

The tests for the firm-day level are conducted using the data from the Jakarta Stock Exchange (JSE), the main exchange of Indonesia, for the period April 2004 to March 2006. The JSE is ideal for use in studying the effects of foreign investors on liquidity. First, the JSE is a fairly transparent market. At any given time, the investors can know not only the best bid-and-ask quotes and respective depths, but also the following five quotes and depths on both sides of the limit order book, in screens provided by different data vendors. Changes to the limit order book are updated in real time. After each transaction, agents in the market can observe not only the price and size of the transaction, but also the brokerage firm and the type, whether local or foreigner, of either party. This way, the market participants can tell whether foreign or local investors are actively trading any given stock, and whether they are net buying or selling. Second, JSE keeps a daily record of the buying and selling by foreign investors on each individual stock. Further market description of the JSE can be found in Bonser-Neal et al., 2005), Dvorak (2005), and Agudelo (2005).

From the JSE we obtain three separate data sets for the period April 2004 to March 2006. The first data set compiles the daily trading statistics for each individual stock in the regular board. It includes open, maximum, minimum, and closing prices, closing bid-and-ask prices, and their respective depths, number of transactions, volume and value traded. The second data set consists of the daily volume of shares sold and the volume of shares bought per stock by foreign investors in all the four boards of the exchange. Finally, the last data set records every day the number of shares owned by foreign investors in each stock, along with the maximum allowed share of ownership for foreigners. Although in the past foreign investors were banned from owning more than 50% in some strategic industries, these limits have been lifted. Since 1999, foreigners have been able to own up to 100% in all types of firms, except banks, where they can still own up to 99%. Therefore, we do not expect that foreign ownership limits constitute an important factor in our analysis.

After merging the three data sets, we eliminate those records pertaining to warrants and rights, and end up with 359 stocks. Furthermore, we eliminate firm-days without valid quotes and stock months with fewer than six trading days in the month, to end up with 88 512 stock days. The summary statistics for the data are presented in Table 1, for the size deciles and for the total sample. Most of the trading value and transactions take place in the top two size deciles, and foreigners actively trade in those, whereas they do not trade much on average in medium and small firms. Table 1 also shows that the ownership of foreign investors tends to be quite uniform across the size deciles, at an average of 16%, but, at the same time, there is considerable variation across firms within the same size decile.

Table 1.   Summary statistics of Jakarta Stock Exchange
The deciles of market capitalization are given by the average market capitalization during 2005. Traded value and number of trades are averages of the daily values for the entire period, including non-trading days. Percentage foreign investor trade is an average of the daily values of (Foreign sales + Foreign buys)/2/Volume. Foreign pressure is a magnitude of directional trade by foreign investors: abs(Foreign buys − Foreign sales)/2/Volume. Turnover is volume divided by market capitalization. Proportional bid-ask spread is the average of (ask − bid)/([ask + bid]/2) whenever quotes are available. Market capitalization and percentage foreign investor ownership are averages for the first quarter of 2006. Percentage foreign investor ownership is the proportion of listed shares owned by investors. Stock months with fewer than five trading days are excluded. Total sample: 359 firms. Period: April 2004 to March 2006. Source: Jakarta Stock Exchange.
Size decileTrade value (billion rupiah)Number of tradesPercentage foreign trade, mean (%)P95 (%)Foreign pressure, mean (%)Turnover, mean (%)Proportional quoted spread (%)Market capitalization, mean (%)Percentage foreign investor ownership
MeanP5P95
  1. P5 and P95 refer to the 5% and 9% percentiles, respectively, in the distribution of percentage of foreign investor ownership.

19023.01.910.40.0100.33212.114 91317.60.150.6
211522.21.47.40.0070.2699.929 39410.40.048.6
36514.32.820.00.0070.0939.360 97716.50.455.9
416221.33.430.20.0040.1298.3117 49313.50.042.1
524825.75.850.00.0060.1485.1151 11215.30.038.8
657442.23.830.30.0100.1905.2290 69221.20.763.0
766939.29.450.00.0100.1324.5523 34913.50.939.9
8196552.011.650.00.0160.2483.41 008 39716.40.346.1
94752102.015.555.20.0190.2173.12 391 29418.32.050.9
1023 262216.734.377.10.0190.1641.519 900 00020.31.657.3
Total449668.811.253.90.0120.1905.42 687 82016.30.050.9

3.2. Daily Market-Wide Data for Seven Emerging Markets

The tests at the daily market level are conducted on the seven emerging stock exchanges for which daily market data on foreign activity is available: India (Bombay Stock Exchange), Indonesia (JSE), Korea (Korean Stock Exchange), the Philippines (Philippines Stock Exchange), South Africa (Johannesburg Stock Exchange), Taiwan (Taiwan Stock Exchange), and Thailand (Stock Exchange of Thailand). The span of time available for each country is indicated in Table 2. The total daily values bought and sold by the foreign investors at the market level were obtained from Bloomberg for India, Indonesia, the Philippines, Taiwan, and Thailand, and directly from the Korean Stock Exchange for Korea. For South Africa, Bloomberg provides the daily net value bought by foreign investors. To our knowledge, excepting KOSDAQ in Korea, the daily foreign investor variables are not available for any other exchange, limiting extending the analysis to other emerging markets.11 Other studies have found restrictions on the availability of foreign trade data and have used similar data: Richards (2005), analyzing the effects of foreign flows in prices and returns for the period 1999–2001, and Griffin et al. (2006), studying the relation between turnover and return.

Table 2.   Summary statistics of daily market data for seven emerging markets
Summary statistics of daily market data for seven emerging markets. Number of trading days is the total number of days included in the sample, excluding days with fewer than 30 trading stocks and Saturdays for Korea and Taiwan. Turnover estimated as Total trading value/Total market capitalization. MK_FITRADE: proportion of the market total trading value by foreigners, calculated as: [Market foreign sales ($) + Market foreign buys ($)]/2/Market_Trading volume ($). MK_FIPRESSURE: abs[Market foreign sales t ($) – Market foreign buys t ($)]]/Total_market_capitalization ($)). Vwbaspread is the daily value-weighted proportional spread, calculated from daily quotes and using Datastream market capitalization as weights.
CountryPeriodNumber of trading daysTurnoverForeign investor trade (%)Foreign pressureVwbaspread (%)Source for quotes
StartEndMean (%)MeanMinimumMaximumMean (%)MeanMinimumMaximum
India18 April 200015 December 200514170.7617.00.233.40.0160.80.34.9Bloomberg
Indonesia11 February 199815 December 200519140.15529.30.067.60.0134.70.612.5Datastream
Korea16 February 199612 April 200522440.82411.21.045.20.0320.50.22.4Bloomberg
Philippines20 July 199926 July 200514900.06748.92.077.80.0082.40.813.9Bloomberg
South Africa27 June 19979 December 200521150.158   0.0141.20.34.2Datastream
Taiwan23 January 199626 July 200523290.9187.00.336.20.0210.50.10.8Bloomberg
Thailand4 January 199926 July 200516100.41825.93.757.80.0213.20.714.4Bloomberg

The market-wide liquidity is estimated as the value-weighted average of the proportional bid-ask spreads calculated from daily-stock quotes as in Chordia et al. (2005). The quotes are obtained from Bloomberg, for India, Taiwan, Thailand, and Korea, and from Datastream, for Indonesia and South Africa. We take five steps to reduce the estimation error on this variable. First, we drop invalid quotes, namely stock days with missing bid or ask, or ask price equal or less than the bid. Second, we fill up missing quotes with the quotes from the former date if available. Third, in each market, we drop dates with fewer than 30 trading stocks. Fourth, after the three previous steps, we discard stocks with missing quotes on any of the trading days of a given year. Finally, we drop the top 1% proportional spreads in the distribution of spreads of a given country in a given day. The weights used in the average come from the market capitalization firm data from Datastream.12 The time-series plots of the value-weighted proportional spreads are presented in Figure 1.

Figure 1.

 Daily value-weighted proportional spreads for seven emerging markets: (a) India, (b) Indonesia, (c) Korea, (d) the Philippines, (e) South Africa, (f) Taiwan, and (g) Thailand.

LOGVWBASPREAD (dashed line): log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes from Bloomberg for all countries but Indonesia and South Africa that come from Datastream. ADJUSTED LOGVWBASPREAD (solid line): adjusted value averaged spread, after detrending and filtering out deterministic trends, as explained in Section 4.

Other market-wide variables are calculated as usual in the literature: returns as the change on the log of the main index in local currency units, volatility as the absolute value of the return, and turnover as the total trading value divided by the total market capitalization. These data were obtained from Bloomberg and complemented or corrected using Datastream as required.13

4. Results

4.1. Firm-Level Models for Jakarta Stock Exchange

We start by analyzing the effect of foreign trading at the firm level in Indonesia. As mentioned above, from JSE we obtain daily buys and sells of foreign investors at the firm level, for the period April 2004 to March 2006. To estimate the effects of foreign trading on stock liquidity we set up the following panel data model, similar to the one in Grullon et al. (2004):

image(1)

where a stands for the firm, and t for the trading day. We use the log of proportional bid-ask spread as the liquidity measure. The model includes firm fixed effects to filter out any firm-specific effect and focus on the time-series effects as well as day-of-the-week dummies to filter any weekly seasonality. We control for the usual determinants of liquidity, namely, the stock return, volatility, and volume. Column A of Table 3 presents the results of the basic time-series model (Equation 1). The firm-specific variables have the predicted sign: positive for volatility and negative for return and volume, the first two being highly significant.

Table 3.   Effects of foreign trading on liquidity at firm-day level
Regressand: Daily change on the log of the proportional bid-ask spread at the daily level, for stocks from Jakarta Stock Exchange (JSE), Indonesia, from April 2004 to March 2005. Stock months with fewer than six trading days were eliminated. Includes omitted constant, firm-specific and day-of-the-week dummy variables. RETURN: change on the log of price of the stock, adjusted for splits. VOLAT: square residual of the return. TURNOVER: log of the trading value scaled down by the market capitalization. FITRADE: proportion of the trading volume due to foreigners, calculated as: [Foreigners sales (shares) + Foreign buys (shares)]/2/Trading volume (shares). FIPRESSURE: absolute value of net buys by foreigners, calculated as: abs[Foreigners sales − Foreign buys]/Shares outstanding. *** and ** represent the significance levels at 1% and 5% respectively.
 ABCDEFGH
RETURN−0.1630.1590.159−0.158−0.233***−0.044−0.285***0.052
VOLAT1.637***1.638***1.635***1.637***1.053***1.545***1.474***1.878***
TURNOVER−0.103***−0.105***−0.104***−0.105***−0.099***−0.090***−0.100***−0.109***
FITRADE 0.173*** 0.121***0.138***0.045**−0.1120.141***
FIPRESSURE  0.220***0.120***0.061**0.099***0.326***0.114***
Size decilesAllAllAllAllAllAll 1–5 6–10
PeriodApril 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006
N 88 512 88 512 88 512 88 512 46 304 42 208 30 195 58 317
R20.7080.7090.7090.7090.7380.7810.6060.611

Next, we test Hypothesis 1, which asserts a negative relationship between foreign trading and the individual stock liquidity. For this, we include two measures of the intensity of foreign trading on a particular stock day. The first is FITRADE, defined as the proportion of trade by foreigners in a given stock day, and calculated as:14

image

Assuming that foreign investors are more likely to be informed than locals, we should expect a positive effect of FITRADE on the spreads. However, high FITRADE does not necessarily indicate high informed trading. Agudelo (2005) shows in an extension to the PIN model of Easley et al. (1997) that high trading volume days by foreigners can be explained by large uninformed or informed foreign trading or both. On the contrary, in the family of PIN models, high directional volume is identified with informed trading. Therefore, we use FIPRESSURE, a proxy of the information content of foreign trades,15 defined as follows:

image

In days when foreign informed traders have good information in a given stock, they will tend to actively buy the stock, and conversely, bad information will result in a strong tendency to sell. Therefore, FIPRESSURE approximates the probability of informed trading by foreigners in a given day. As a consequence, and following Agudelo (2005), we expect a positive effect of FIPRESSURE on the liquidity measures. Moreover, the results of Choe et al. (2005) in Korea and Bonser-Neal et al. (2005) in Indonesia suggest that foreigners are more aggressive liquidity demanders, without implying any information advantage. Although the two effects overlap to some degree, we expect FITRADE to be more a measure of liquidity demand by foreigners per se, whereas FIPRESSURE to be more a measure of informed trading.

Columns B and C of Table 3 present the results of model 1 (Equation 1), including alternatively the two measures of foreign trading. Both the coefficients of FITRADE and FIPRESSURE are positive and significant at the 1% level, and remain so when we include both variables in column D. This means that on stock days when foreign investors are either trading a lot or aggressively trading in one direction, the liquidity of the stock decreases on average, as measured by the bid-ask spread. Columns E–H test the robustness of this finding, by constraining the model, respectively, to the first half of the sample, the second half, the smaller firms, and the larger firms. We observe that, overall, the effects of FIPRESSURE are very robust, and tend to be larger in the size deciles one to five, hereas the effects of FITRADE are absent only for the smaller firms.

These results are supportive of Hypothesis 1: foreign trading activity per se has a harmful effect on the liquidity of individual firms, consistent with aggressive liquidity demand, as in Bonser-Neal et al. (2005). Moreover, when foreign trading has a strong buying or selling trend it further reduces the liquidity, consistent with foreign investors trading on information at the firm level, as in Seasholes (2004), Huang & Shiu (2005), and Agudelo (2005). The liquidity demanding effects tend to be larger in larger firms, while the information effects tend to be larger in smaller firms.

Next, we investigate whether foreign investor trading is related to market-wide effects on the liquidity of the JSE. For this, we propose two measures of market-wide trading activity of foreign investors: MKT_FITRADE and MKT_FIPRESSURE, analogous to the already defined measures of firm-specific foreign trading activity, and defined as follows:

image
image

MK_FITRADE is simply the proportion of the total trading value of the market due to foreign investors. It measures how actively foreigners trade in the market irrespective of their net buying or selling. MK_FIPRESSURE, by contrast, will be high whenever foreign investors are actively net selling or net buying in the market, and low in days when foreign sales are roughly as high as foreign buys. Therefore, analogous to FIPRESSURE, we consider that MKT_FIPRESSURE serves as a proxy for the overall level of information of foreigners in the market.

In addition, we need to control for the factors known to affect the level of market-wide liquidity. The literature on systematic liquidity has reported commonality patterns among the liquidity of individual stocks in the USA, and has explained part of that commonality as caused by market-specific variables.16 Accordingly, we include the following variables identified by the literature: the daily return of the market index, (MK_RETURN), the volatility of the market return (MK_VOLAT), and the log of the turnover of the market (MK_TURNOVER). The resulting model is presented in column B of Table 4 The sign of the coefficients of the market control variables is found as given by the empirical literature on systematic liquidity: negative coefficients of MK_RETURN, MK_TURNOVER, and positive coefficients of MK_VOLAT, all of them significant at the 1% level.

Table 4.   Effects of market foreign trading on liquidity at firm-day level
Regressand: Daily change on the log of the proportional bid-ask spread at the daily level, for stocks from Jakarta Stock Exchange (JSE), Indonesia, from April 2004 to March 2005. Stock months with fewer than six trading days were eliminated. Includes omitted constant, firm-specific and day-of-the-week dummy variables. RETURN: change on the log of price of the stock, adjusted for splits. VOLAT: square residual of the return. TURNOVER: log of the trading value scaled down by the market capitalization. FITRADE: proportion of the trading volume due to foreigners, calculated as: [Foreigners sales (shares) + Foreign buys (shares)]/2/Trading volume (shares). FIPRESSURE: absolute value of net buys by foreigners, calculated as: abs[Foreigners sales − Foreign buys]/Shares outstanding. Market-level data taken from Bloomberg: MK_RETURN: change on the log of the JSE index. MK_VOLAT: absolute value of MK_RETURN. MK_TURNOVER, log of the market turnover estimated as total trading value in JSE/Total market capitalization of JSE. MK_FITRADE: proportion of the market total trading value by foreigners, calculated as: [Market foreign sales t ($) + Market foreign buys t ($)]/2/Market_trading volume t ($). MK_FIPRESSURE: abs[Market Foreign sales t ($) − Market foreign buys t ($)]]/Total_market_capitalization t. EWBASPREAD: log of the equally weighted average of the proportional spread in the market. VWBASPREAD: log of the value-weighted average of the proportional spread in the market, using market capitalization as weights. *** represents the significance level at 1%.
VariableABCDEFG
RETURN−0.158−0.047−0.043−0.045−0.042−0.070−0.062
VOLAT1.637***1.468***1.458***1.471***1.459***1.374***1.370***
TURNOVER−0.105***−0.102***−0.102***−0.102***−0.102***−0.103***−0.103***
FITRADE0.121***0.117***0.082***0.117***0.083***0.075***0.075***
FIPRESSURE0.120***0.114***0.121***0.111***0.120***0.109***0.107***
MK_RETURN −0.641***−0.167−0.631***−0.1740.178−0.040
MK_VOLAT 6.610***6.022***6.466***5.988***3.946***4.139***
MKT_TURNOVER −0.066***−0.045***−0.075***−0.048***−0.008−0.046***
MK_FITRADE  0.625*** 0.612***0.0410.103***
MK_FIPRESSURE   0.343***0.110***−0.0440.129***
VWBASPREAD     0.627*** 
EWBASPREAD      0.627***
N88 51288 35488 35488 35488 35486 87387 065
R20.7090.7150.7180.7150.7180.7330.732

Next, we include alternatively and together the variables of interest, MK_FITRADE and MK_FIPRESSURE, as presented in columns C–E of Table 4. Both variables have significant and positive coefficients. This suggests that increasing foreign trading activity in the market reduces to some extent the liquidity of the stocks in the market. Over and above that effect, there is an incremental reduction on liquidity if foreigners are net buyers or net sellers in the market. As we build the market-wide model over the firm-specific variable, the market-wide effect of foreigners goes beyond the firm-specific effect previously reported. In unreported tests we found the MK_FITRADE effect is robust after running it separately in the first and second halves of the sample, and in the lower and upper halves by size decile. The MK_FIPRESSURE effects are significant only in the first half of the sample and for larger firms. The economic effects are also important: an increase of 1 SD in MK_FITRADE or MK_FIPRESSURE is related to an average rise of 5% and 1%, respectively, on the proportional spread.

Overall, the results support Hypothesis 2, in the sense that foreign trading is related to negative market-wide effects on the liquidity of an emerging market such as Indonesia. Although causality from foreign trade to liquidity cannot be inferred from those results, it is difficult to come up with an alternative explanation.17 Causality will be tested later in a more appropriate econometric model.

Columns E and G of Table 4 incorporate alternatively two measures of market-wide liquidity in the empirical model: EWBASPREAD, the equally weighted average of the proportional spreads across the stocks in the market, and VWBASPREAD, the value-weighted average of the proportional spreads. Clearly, the value-weighted measure subsumes most of the effects of the market-wide variables, including the market foreign trading variables, while leaving unaffected the coefficients of the firm-specific variables.18 Therefore, we can argue that the market-wide effect of the foreign investors on liquidity is incorporated in the systematic liquidity of the market, as measured by a value-weighted average of proportional spreads. In the next section, we will study the relation between market foreign activity and market-wide liquidity in the context of a VAR model, for Indonesia and six other emerging markets.

Table 5.   Effects of foreign ownership on liquidity at stock-month level
Regressand: log of the average proportional bid-ask spread for each stock month. Data from Jakarta Stock Exchange (JSE), Indonesia, from April 2004 to March 2006. Regressions include omitted constant, and month dummy variable. PRICE: log of the last closing price of the month. MARKET_CAP: log of the average market capitalization in rupiah. VOLAT: average of the absolute value of the returns. VOLUME: log of the average trading value in Rupiahs. No_TRADES: log of the average daily number of transactions. FI_OWNERSHIP: average share of the firm owned by foreign investors. FITRADE: average proportion of the trading volume due to foreigners, calculated as: [Foreigners sales (shares) + Foreign buys (shares)]/2/Trading volume (shares)*** and ** represent the significance levels at 1% and 5% respectively.
 ABCDEFG
PRICE−0.312***−0.310***−0.315***−0.344***−0.280***−0.422***−0.278***
MARKET CAP−0.009−0.011−0.019**−0.018−0.0210.002−0.005
VOLUME0.0120.0140.0080.050***−0.022−0.0330.028**
VOLAT9.558***9.546***9.404***8.349***11.814***7.215***11.016***
No_TRADES−0.252***−0.254***−0.246***−0.290***−0.220***−0.144***−0.290***
FI_OWNERSHIP −0.113***−0.125***−0.149**−0.109**−0.119**−0.165***
FITRADE  0.230***0.1280.301**0.3310.043
Size decilesAllAllAllAllAll1–56–10
PeriodApril 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006April 2004 to March 2006
N4483448344832360212315382945
R20.8090.8100.8110.7910.8340.7310.747

Finally, we investigate the relation between foreign ownership and liquidity. Table 5 presents the results of a cross sectional regression of the liquidity on firm characteristics, based on Stoll (2000) and Lesmond (2005):

image(2)

where a stands for the firm and m for the month. Liquidityam is the average proportional spread of the stock month. The model includes month fixed effects, to control for any time-dependent effects. The resulting coefficients are presented in column A of Table 5. The R2 of the regression is quite high (81%) and the coefficients of price, number of trades, and volatility appear significant and with the expected sign. By contrast, volume and market capitalization turn out to be insignificant, probably because their effect is captured by number of trades and price.

Table 6.   Granger causality results for a base 4-vector autoregressive (VAR) model daily country model with liquidity
The results of a base 4-VAR model with daily endogenous variables VWBASPREAD, RETURN, VOLAT, and TURNOVER are presented. The base 4-VAR is estimated with a constant, and two lags for all cases, except the Philippines and South Africa with only one lag. Panel A presents the correlations between the contemporaneous innovations of each endogenous variable. Panel B presents the chi-squares statistics for the pairwise Granger causality test. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes from Bloomberg for all countries but Indonesia and South Africa come from Datastream. Value weights from individual market capitalization are from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed: RETURN: change on the log of the main market index. VOLAT: absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. The variables were detrended and filtered out of deterministic trends, as explained in Section 4. ** indicates correlations and chi-squared significant at the 5% level.
  Panel A: Correlations between VAR innovationsPanel B: χ2-statistics from Granger causality tests. Null hypothesis: row variables do not Granger-cause column variables
VWBASPREADRETURNVOLATTURNOVERVWBASPREADRETURNVOLATTURNOVER
IndiaVWBASPREAD1.00    6.83**6.11**7.60**
RETURN−0.39**1.00  12.66** 56.92**16.13**
VOLAT0.14**−0.10**1.00 14.45**3.11 40.93**
TURNOVER0.17**−0.020.14**1.000.052.750.70 
IndonesiaVWBASPREAD1.00    2.736.42**8.11**
RETURN−0.24**1.00  24.18** 24.23**18.22**
VOLAT0.040.011.00 2.442.89 21.60**
TURNOVER−0.08**0.22**0.45** 9.84**0.265.49 
KoreaVWBASPREAD1.00    2.7016.32**4.17
RETURN−0.29**1.00  29.28** 20.03**92.55**
VOLAT0.08**−0.04**1.00 24.47**4.69 16.75**
TURNOVER0.06**0.21**0.28**1.0016.20**1.356.08** 
PhilippinesVWBASPREAD1.00    1.270.620.88
RETURN−0.10**1.00  3.50 0.1910.12**
VOLAT0.07**−0.031.00 0.421.14 0.17
TURNOVER−0.020.030.23**1.005.32**0.218.45** 
South AfricaVWBASPREAD1.00    1.6940.52**3.80
RETURN−0.13**   1.65 3.571.31
VOLAT0.20**1.001.00 34.52**1.49 6.30**
TURNOVER0.020.030.15**1.002.470.252.10 
TaiwanVWBASPREAD1.00    2.084.414.87
RETURN−0.18**1.00  17.40** 27.64**45.07**
VOLAT−0.17**0.05**1.00 4.2811.56** 6.04**
TURNOVER−0.08**0.30**0.25**1.003.917.32**2.77 
ThailandVWBASPREAD1.00    4.4710.88**7.02**
RETURN−0.19**1.00  7.49** 1.7692.34**
VOLAT0.12**0.001.00 8.64**1.62 4.75
TURNOVER−0.010.23**0.38**1.0011.01**3.442.49 

When we include in the model the average proportion of foreign ownership, its coefficient in column B of Table 5 appears significant and negative. This result is robust to the inclusion of the average proportion of foreign trade, and to restrict the model to the first or second half of the period, and to the lower half or upper half of firms by size, as shown in columns 4–7 of Table 5. The negative relation between foreign ownership and the proportional spread is also economically significant: a 1 SD increase in the distribution of foreign ownership (18%) is associated with a proportional reduction of 2.2% in the proportional spread.

Although these results are consistent with Hypothesis 3, they are not enough to presume that foreign ownership improves the liquidity of the firms. The causality, if it exists at all, might run in the opposite direction: other things being equal, foreigners might well prefer liquid firms, as suggested by Breen et al. (2002).

4.2. Market-Wide Models in Seven Emerging Markets at Daily Frequency

The results of the previous section for Indonesia suggest that foreign investor activity has negative effects on liquidity, not only at the stock level but also at the market level, as posed by Hypothesis 2. In this section, we further explore this question using a VAR specification for six emerging countries: India, Indonesia, Korea, the Philippines, Taiwan, and Thailand. In some specifications a seventh market, South Africa, will be included, for which only total daily net buys are available. VAR models have been used in the study of market-wide liquidity by Chordia et al. (2005) and Fujimoto (2004), and to study the inter-relations of foreign flows with both cost of capital by Bekaert et al. (2002), and with returns by Richards (2005) and Griffin et al. (2006).

We start with a base 4-VAR to model the dynamic behavior of four market-wide variables on a daily frequency: liquidity, measured as the log of the value-weighted average of the proportional bid-ask spread, return as the change on the log of the main local country index, volatility as the absolute value of the return, and turnover as the total market traded value divided by the market capitalization. This set of variables is used by Fujimoto (2004) in her monthly study of US liquidity. Chordia et al. (2005) use a similar set of variables in their daily VAR models of bond and stock market liquidity, but instead of turnover, they include a measure of market-wide order imbalance, which has been found to have negative effects on liquidity. Unfortunately, the unavailability of transaction data impedes us in incorporating the order imbalance in our VAR model. The importance of controlling for other market variables that potentially affect market-wide liquidity cannot be understated. Foreign flows are hardly an exogenous variable in this context. For example, there is evidence of foreign flows both chasing returns and generating price pressure on emerging markets (Richards, 2005; Choe et al., 1999, 2005).

Next, we discuss the expected effects on liquidity of the other three market variables. From an empirical point of view, Fujimoto (2004) and Chordia et al. (2005) provide strong evidence that in the USA negative (positive) market returns Granger-cause decreases (increases) in systematic liquidity. This has been explained as follows: liquidity providers find it more difficult to keep a balanced inventory in falling markets than in rising ones (Chordia et al., 2001). Volatility is a factor that reduces the utility of risk-averse liquidity providers, hence reducing liquidity, in the inventory models of both Grossman and Miller (1988) and Ho & Stoll (1981). In the USA, portfolio volatility has been found to be positively related to proportional spread by Huberman & Halka (2001), and positive shocks in stock market volatility induce negative shocks in systematic liquidity, as reported by Chordia et al. (2005) and Fujimoto (2004). The increase in trading reduces the inventory cost in both the models of Ho & Stoll (1981) and Grossman and Miller (1988). Using a VAR model, Fujimoto (2004) finds that a positive innovation to aggregate turnover impacted negatively on both the proportional spread and the price impact. However, Chordia et al., 2002) argue and provide evidence that, beyond turnover, an increase in the absolute value of order imbalance hurts liquidity because it increases the inventory risk for the liquidity provider, or might signal private information. Therefore, to the extent that turnover is correlated with order imbalance, the positive effect of turnover on liquidity might be decreased or even distorted in a model where order imbalance is omitted.

Before running the VAR model, we adjust the market variables to account for deterministic trends both in the mean and the variance, in a fashion similar to that ni Chordia et al. (2005). Namely, if w is either the vector of liquidity, returns, volatility or any foreign investor variable, for each country c we obtain the vector of residuals u after controlling for a matrix of adjustment variables, x′, in the following mean equation:

image(3)

The vector of residuals is then regressed in the following equation of the variance:

image(4)

The resulting predicted values of variance xγc are then used to standardize the residuals of the mean equation as follows:

image(5)

where ac and bc are calculated so that the means and variances of the adjusted variables are the same as those of the unadjusted series.

Following Chordia et al. (2005) the adjustment variables are chosen as follows:

  • 1A dummy variables for each of 4 days of the week, to filter out any predictable weekly patterns.
  • 2Two crisis dummies, one for the Asian crises (2 July 1997 to 31 December 1997), the other for the Russian crises (6 July 1998 to 31 December 1998).
  • 3Eleven dummy variables for the month of the year.
  • 4A dummy variable for holiday effects; this is equal to 1 for a day preceding or following a holiday, unless it is in a weekend.
  • 5Linear and quadratic trend terms.
  • 6Individual dummies for days with too little trading activity, or with an outlier in any variable, as detected by visual inspection of the residuals.19

With respect to turnover, following Griffin et al. (2006), we detrend it by, first, taking the logarithm, and then subtracting its 100-day moving average, a method that was originally proposed by Lo & Wang (2000).

Next, we run an augmented Dickey–Fuller test (ADF) to check for the stationarity of each adjusted variable, allowing for an intercept, starting from 10 lags and stopping when the coefficient of the last lag turns out to be significant, as suggested by Enders (1995). For every adjusted variable in each country we reject the null hypothesis of not having a unit root at the 5% significance level, and confirm this result using a Philipp–Perron Test (PP).

Initially, we set up a base 4-VAR model per country, with the adjusted liquidity, return, volatility, and turnover variables, for two reasons: first, as an out-of-sample test of the implications of the theory and of the findings of the systematic liquidity literature in the US stock market, and, second, as the results of these country-specific VAR models will give us an idea of the heterogeneity of the inter-relations between variables before we incorporate the foreign investor variables. The base 4-VAR model we propose can be expressed as the following system of four equations:

image

where Xt is the vector of market variables log of the value-weighted spread, index return, index volatility, and turnover, and ut is the vector of innovations. We chose K based on the Schwarz information criterion (SIC). Of the three most used criteria (the others being AIC and Hannan and Quinn criterion (HQC)) the SIC has been found to consistently approximate the true number of lags and render the most parsimonious specification (Lütkephol, 2004). The lags found by this criterion are two for India, Indonesia, Korea, Taiwan, and Thailand, and one for the Philippines and South Africa. We first investigate the cross-correlations of the innovations from each one of the four equations of the VAR system. Significant correlations between innovations of two market variables suggest that they are subject to common influences or that there is unidirectional or bidirectional immediate causality between the two (Lütkephol, 2005).

Panel A of Table 6 reports, for each country, the cross-correlations of the innovations of each four equations. Focusing on the liquidity measure, we find an expected negative correlation with the innovations of the returns significant for the seven countries, and a positive relation with innovations of volatility, significant for five. These two empirical regularities have been reported for the US markets by Chordia et al. (2001). The expected negative relationship with innovations in turnover is only significant in two countries, whereas it is significantly positive in the other two.20 Other than that, the seven countries present a significantly positive correlation between innovations of volatility and innovations of turnover that has been well established in the literature.

Panel B of Table 6, in turn, presents the results of pairwise Granger causality tests between the variables of the country-specific base 4-VAR models. The intersection of row i and column j reports the chi-squared value for the Wald test of the null hypothesis that the coefficients of the K lags of the variable i are jointly zero in the equation of variable j. Focusing on the first column we find that, overall, there is strong causality from returns, volatility and turnover on liquidity. As usual in the VAR methodology, we obtain the impulse response functions (IRF) to estimate the size and sign of the causation.21 The IRF of the orthogonal innovations of the market variables on the liquidity measure are presented in Figure 2. For all the cases where a market variable Granger-causes the liquidity measure, the IRF indicates a significant response and it is robust to the reordering of the variables (not reported). Taken together, the Granger causality results and the IRF in the seven countries reinforce the findings of the cross-correlation of innovations. First, innovations on returns Granger-cause opposite sign shocks in liquidity in four countries. Second, innovations in volatility Granger-cause shocks of the same sign in spreads in four countries. Finally, innovations on turnover Granger-cause shocks of the opposite sign in spreads, also in four countries. Similar results in VAR models have been reported for US markets in the systematic liquidity literature (e.g. Chordia et al. (2005) and Fujimoto (2004). We also found evidence of four well-known empirical regularities: negative returns cause positive volatility shocks (four countries), positive spreads cause volatility increases (five countries), positive returns cause positive turnover shocks (six countries), and finally, positive volatilities cause increasing turnover (four countries). Finally, we note the heterogeneity in the results across countries: whereas there are three countries, India, Indonesia and Korea, where the four effects are robust, in the Philippines and South Africa only one of the four effects shows up, and in Taiwan the only significant effect, returns on liquidity, appears with an unexpected sign.

Figure 2.

 Impulse response functions (IRF) of orthogonalized innovations from market variables on liquidity for seven emerging markets: (a) India, (b) Indonesia, (c) Korea, (d) the Philippines, (e) South Africa, (f) Thailand, and (g) Taiwan.

The IRF are calculated from the results of a base 4-vector autoregressive (VAR) model with daily endogenous variables VWBASPREAD, RETURN, VOLAT, and TURNOVER. The base 4-VAR is estimated with a constant, and two lags for all cases, except the Philippines and South Africa with only one lag. The IRF are obtained after a Cholesky decomposition, assuming an ordering VWBASPREAD, RETURN, VOLAT, and TURNOVER. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes are from Bloomberg for all countries but Indonesia and South Africa, which come from Datastream. Value weights from individual market capitalization are from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed: RETURN: change on the log of the main market index. VOLAT: Absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. The variables were detrended and filtered out of deterministic trends, as explained in Section 4. Confidence intervals of 95% in dashed lines.

Next, we include the foreign investor variable MK_FITRADE in the vector Xt of market variables and run a base 5-VAR model for each country. We define this measure as in the former section, adjust it for deterministic trends as indicated before, and test its stationarity with the ADF and PP tests. As before, we select the number of lags by the SIC criteria, rendering two lags in every cases except for Korea and Taiwan, which require only one lag. Table 7 presents the results of the base 5-VAR model: Panel A the correlation matrix between the residuals of each equation and Panel B the results of the test for Granger causality.

Table 7.   Granger causality results for a base 5 vector autoregressive (VAR) daily country model with liquidity and foreign trade
The results of a base 5-VAR model with daily endogenous variables FITRADE, VWBASPREAD, RETURN, VOLAT, and TURNOVER are presented. The base 5-VAR is estimated with a constant, and two lags for all cases, except Korea and Taiwan with three lags. Panel A presents the correlations between the contemporaneous innovations of each endogenous variable. Panel B presents the chi-squared statistics for the pairwise Granger causality test. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes from Bloomberg for all countries but Indonesia and South Africa come from Datastream. Value weights from individual market capitalization come from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed: RETURN: change on the log of the main market index. VOLAT: absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. FITRADE: average proportion of the trading volume due to foreigners, calculated as: [Foreigners sales (LCU) + Foreign buys (LCU)]/2/Trading volume (LCU). The variables were detrended and filtered out of deterministic trends, as explained in Section 4. ** indicates correlations and chi-squared significant at the 5% level.
 Panel A: Correlations between VAR innovationsPanel B: χ2-statistics from Granger causality tests. Null hypothesis: Row variables do not Granger-cause column variables
FITRADEVWBASPREADRETURNVOLATTURNOVERFITRADEVWBASPREADRETURNVOLATTURNOVER
India
 FITRADE1.00     1.040.860.070.93
 VWBASPREAD0.09**1.00   2.71 6.26**6.08**6.90**
 RETURN−0.01−0.39**1.00  1.9712.79** 56.63**16.50**
 VOLAT0.11**0.14**−0.10**1.00 4.3413.54**2.79 40.96**
 TURNOVER0.11**0.18**−0.020.14**1.005.730.082.960.75 
Indonesia
 FITRADE1.00     1.662.544.1711.61**
 VWBASPREAD0.041.00   1.22 2.896.34**8.46**
 RETURN−0.11**−0.23**1.00  3.7024.38** 25.54**15.54**
 VOLAT−0.05**0.040.011.00 0.732.402.99 20.99**
 TURNOVER−0.09**−0.08**0.23**0.45**1.001.6210.45**0.185.12 
Korea
 FITRADE1.00     4.000.781.221.73
 VWBASPREAD0.11**1.00   1.16 1.9234.99**4.21
 RETURN−0.10**−0.29**1.00  25.90**47.21** 24.15**86.05**
 VOLAT0.07**0.06**−0.05**1.00 8.16**22.41**6.41** 15.30**
 TURNOVER0.08**0.06**0.21**0.28**1.0019.00**10.41**2.026.01** 
Philippines
 FITRADE      0.361.380.460.15
 VWBASPREAD1.001.00   2.69 4.241.281.35
 RETURN−0.07**−0.09**1.00  6.78**5.00 1.9413.13**
 VOLAT−0.010.07**−0.031.00 2.273.021.31 0.25
 TURNOVER0.00−0.010.030.24**1.003.275.200.188.82** 
Taiwan
 FITRADE1.00     2.940.981.2613.78**
 VWBASPREAD0.12**1.00   2.91 3.624.605.85
 RETURN−0.17**−0.18**1.00  37.03**13.49** 30.69**59.93**
 VOLAT−0.01−0.17**0.05**1.00 1.134.689.48** 24.25**
 TURNOVER−0.23**−0.08**0.31**0.26**1.0017.96**1.589.12**3.03 
Thailand
 FITRADE1.00     3.081.643.494.46
 VWBASPREAD0.041.00   1.34 4.4510.77**6.79**
 RETURN−0.15**−0.19**1.00  42.18**7.70** 2.7986.94**
 VOLAT−0.030.11**0.001.00 4.278.76**1.57 4.71
 TURNOVER−0.06**−0.010.23**0.38**1.0030.83**13.87**3.961.59 

We focus on the relation between MK_FITRADE and the liquidity variable. Table 7 indicates a positive correlation between the residuals of the spreads and the residual of the foreign investor variable in India, Korea, and Taiwan, significant at the 5% level.22 This suggests that spreads tend to be high in days of strong foreign investor activity, as posed by Hypothesis 2. We examine two alternative explanations. First, this might be a spurious relationship, due to an omitted third market variable. To control for this, we run, for the three countries, a regression of the residual of the liquidity equation against the other residuals, which confirms the significant positive effect (not reported). Second, the causality might run in the opposite direction: higher spreads might induce higher foreign trading, but this seems less realistic an explanation.

By contrast, Panel B shows that we fail to find any Granger causality from MK_FITRADE to the spreads. Taken together, the evidence of Table 7 is partially supportive of Hypothesis 2. Similar results are obtained if we use MK_FIPRESSURE instead (unreported). Restricting the models to the period after January 2000, to avoid the Asian and Russian crises, renders qualitatively the same results (unreported). Summarizing, there is some evidence of market-wide effects from foreign trading in three countries, consistent with short-lived causality (within a day) but not with Granger causality.23

The difference in results between countries can be explained following Bekaert et al. (2003). Whereas all the seven countries had liberalized their markets before 1996, the “intensity” of liberalization varies, due to three group of factors: (i) legal barriers; (ii) indirect barriers (information asymmetry, accounting standards, and investor protection); and (iii) risks particularly important in emerging markets (liquidity, political, economic policy, and currency). They devise a measure of financial liberalization that goes from 0.0 to 1.0. For the period 2000–2002, India had a measure of approximately 0.2, Indonesia 0.35, Korea 0.2, the Philippines 0.35, Taiwan 0.05, and Thailand 0.35. There is no measure available for South Africa.

Therefore, a possible explanation of the differences of the results is given by the degree of financial liberalization. Less liberalized countries (India, Korea, and Taiwan) appear to be more “sensitive” to liquidity demand (in the very short term). Countries more open to foreign trading, such as the Philippines and Thailand, that indeed opened their markets as early as 1991 and 1987, respectively, are not so affected by foreign trading.

As a collateral result, Panel A shows a negative relationship between returns and MK_FITRADE for five countries, whereas Panel B and the corresponding IRF (omitted) show that higher returns Granger-cause lower trading by foreigners. This basically replicates the results of Griffin et al. (2006) in the sense that locals more than foreigners drive the positive return–turnover relation. The authors comment that “Under the assumption that foreign investors are more sophisticated than domestic investors, this evidence could also be consistent with the view that the (return–turnover) relation is stronger among unsophisticated investors” (p. 24).

Next, we explore Hypothesis 3 with a different base 5-VAR daily model. Whereas in the previously we measured the effect due to increasing foreign trading, here we are interested in the effects of increasing ownership by foreigners. Accordingly, we define the variable MK_FINETBUY as a measure of the daily increase in the market foreign ownership, as follows:

image

Note that this variable can be estimated for South Africa, unlike MK_FITRADE, which makes a total of seven emerging markets. We adjust this variable for deterministic trends and seasonality in the mean and variance as described above, and test it with the ADF and PP tests of stationarity. We add this variable to the vector Xt, to obtain a base 5-VAR system, which is, again, estimated country by country. As before, the lags of the VAR are obtained using the SIC criteria, with two lags for all the countries, except Taiwan which requires three.

The results of the 5-VAR model with MK_FINETBUY are presented in Table 8. We focus on the relations with the foreign investor variable. Starting with Panel B we note that an increasing ownership in emerging markets Granger-causes the spreads in India and Korea at the 5% level.24 The IRF, in Figure 3, show that for those two countries the effects are unambiguously negative, and last approximately 1–2 weeks. Moreover, reversing the order of the variables for the Cholesky decomposition confirms that the negative effect is robust. In addition, the IRF for South Africa and Thailand also show a negative effect from MK_FINETBUY to the spreads, significant at the 10% level, when the foreign variable precedes the spreads in the Cholesky decomposition but not when the order of variables is reversed.25 Panel A reports a negative and significant relationship between the residuals of the liquidity and the foreign investor equations. However, this correlation is of very small magnitude and disappears once we control for the other residuals (unreported). The overall result is robust and even improved when we examine the results of an unreported base 2-VAR model with only the spreads and the foreign net buy variable. Overall, the results of Table 8 support Hypothesis 3, in the sense that an increasing foreign ownership benefits liquidity, even in the short term. As indicated in Section 2 this could be explained by foreign buying being regarded as a positive signal in the emerging markets attracting uninformed traders and liquidity providers, and improving the market-wide liquidity, similarly to positive returns.

Table 8.   Granger causality results for a base 5-vector autoregressive (VAR) daily country model with liquidity and foreign netbuy
The results of a 5-VAR model with daily endogenous variables FINETBUY, VWBASPREAD, RETURN, VOLAT, and TURNOVER are presented. The base 5-VAR is estimated with a constant, and two lags for all cases, except Taiwan with three lags. Panel A presents the correlations between the contemporaneous innovations of each endogenous variable. Panel B presents the chi-squared statistics for the pairwise Granger causality test. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes from Bloomberg for all countries but Indonesia and South Africa come from Datastream. Value weights from individual market capitalization come from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed. RETURN: change on the log of the main market index. VOLAT: absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. FINETBUY: total net buys by foreigners in the market, calculated as: [Foreigners buys (LCU) − Foreign sales (LCU)]/Market capitalization (LCU). The variables were detrended and filtered out of deterministic trends, as explained in Section 4. ** indicates correlations and chi-squared significant at the 5% level.
 Panel A: Correlations between VAR innovationsPanel B: χ2-statistics from Granger causality tests. Null hypothesis: Row variables do not Granger-cause column variables
FINETBUYVWBASPREADRETURNVOLATTURNOVERFINETBUYVWBASPREADRETURNVOLATTURNOVER
India
 FINETBUY1.00     17.20**8.05**1.793.46
 VWBASPREAD−0.07**1.00   6.12** 8.90**5.158.78**
 RETURN0.20**−0.39**1.00  29.94**16.30** 50.20**18.62**
 VOLAT0.050.14**−0.10**1.00 5.5515.18**2.68 41.45**
 TURNOVER0.08**0.17**−0.020.14**1.003.330.032.810.67 
Indonesia
 FINETBUY1.00     2.590.841.7711.29**
 VWBASPREAD−0.07**1.00   1.18 2.696.34**8.54**
 RETURN0.23**−0.23**1.00  54.06**25.77** 21.06**21.81**
 VOLAT0.05**0.040.011.00 5.482.252.87 20.31**
 TURNOVER0.26**−0.08**0.22**0.45**1.000.518.79**0.315.59 
Korea
 FINETBUY1.00     7.89**3.760.5931.14**
 VWBASPREAD−0.041.00   3.40 3.0316.46**4.86
 RETURN0.23**−0.28**1.00  142.90**21.48** 17.81**99.45**
 VOLAT−0.010.08**−0.04**1.00 0.3123.90**5.11 15.86**
 TURNOVER0.23**0.06**0.21**0.29**1.000.9513.24**1.915.88 
Philippines
 FINETBUY1.00     0.284.765.241.59
 VWBASPREAD−0.011.00   1.07 4.311.291.31
 RETURN0.11**−0.09**1.00  29.78**5.38 2.8913.62**
 VOLAT0.020.07**−0.031.00 0.142.991.37 0.25
 TURNOVER0.06**−0.010.040.24**1.001.315.350.278.50** 
South Africa
 FINETBUY1.00     2.832.890.052.09
 VWBASPREAD−0.031.00   2.75 1.3337.93**2.38
 RETURN−0.07**−0.13**1.00  2.003.99 9.90**1.80
 VOLAT−0.010.19**0.031.00 7.07**26.63**5.03 2.08
 TURNOVER0.030.040.06**0.21**1.000.813.340.922.19 
Taiwan
 FINETBUY1.00     5.5335.24**13.77**4.14
 VWBASPREAD−0.06**1.00   3.52 4.124.284.65
 RETURN0.33**−0.18**1.00  50.49**11.99** 16.13**51.66**
 VOLAT0.02−0.17**0.05**1.00 4.584.9510.78** 22.15**
 TURNOVER0.19**−0.08**0.31**0.26**1.007.89**1.476.893.95 
Thailand
 FINETBUY1.00     0.0811.31**0.530.18
 VWBASPREAD−0.06**1.00   4.52 4.5311.05**7.04**
 RETURN0.32**−0.19**1.00  167.98**7.03** 2.2787.14**
 VOLAT−0.040.12**0.001.00 1.698.66**1.29 4.57
  TURNOVER0.06**−0.010.23**0.38**1.009.92**10.33**3.262.01 
Figure 3.

 Impulse response functions (IRF) of orthogonalized innovations from Foreign net buy on liquidity for seven emerging markets: (a) India, (b) Indonesia, (c) Korea, (d) the Philippines, (e) South Africa, (f) Taiwan, and (g) Thailand.

The IRF are calculated from the results of base 5-VAR model with daily endogenous variables VWBASPREAD, RETURN, VOLAT, TURNOVER and FINETBUY. The base 5-VAR is estimated with a constant, and two lags for all cases, except Korea and Taiwan with three lags. The IRF1 are obtained after a Cholesky decomposition, assuming an ordering VWBASPREAD, RETURN, VOLAT, TURNOVER, and FINETBUY, whereas The IRF2 assume the reverse ordering. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes are from Bloomberg for all countries but Indonesia and South Africa, which come from Datastream. Value weights from individual market capitalization are from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed: RETURN: change on the log of the main market index. VOLAT: absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. FINETBUY: total net buys by foreigners in the market, calculated as: [Foreigners buys (LCU) − Foreign sales (LCU)]/Market capitalization (LCU). The variables were detrended and filtered out of deterministic trends, as explained in Section 4. Confidence intervals of 95% in dashed lines.

Table 8 also shows a strong bidirectional relation between foreign net buy and returns: Panel A shows a significant positive correlation between the two for all the countries except South Africa. Furthermore, Panel B shows bidirectional Granger causality for India, Taiwan, and Thailand, while the causality runs only from the returns to the foreign investor trade for Indonesia, Korea, and the Philippines. Figure 4 presents the IRF of returns to orthogonal innovations of the foreign variable, and vice versa. For the cases where Granger causality appears, the direction of the responses is always positive and robust to the reordering of the variables (not reported). This is strong evidence of two effects that have already been reported in the literature: on the one hand, positive returns causing positive foreign net buys are evidence of trend following and/or positive feedback trading by foreign investors, as reported by Froot et al. (2001), Bekaert et al. (2002), and Richards (2005); on the other hand, positive foreign net buys causing positive returns is evidence of price pressure, as reported by Richards (2005), and Choe et al. (1999, 2005), or information content, as in Froot & Ramadorai (2002). However, if information were the cause of this return increase, the classical market microstructure models would imply a contemporaneous or lagged positive relation between the foreign investor variable and the spreads, which is not observed for any of the seven countries.26 By contrast, using transaction data from Korea, Choe et al. (2005) provide evidence consistent with foreigner trading causing price pressure but not evidence of superior information. All in all, the evidence suggests that price pressure is a better explanation of this relation between foreign net buys and returns.

Figure 4.

 Impulse response functions (IRF) of orthogonalized innovations from Return on Foreign net buy and viceversa for seven emerging markets: (a) India, (b) Indonesia, (c) Korea, (d) the Philippines, (e) South Africa, (f) Taiwan, and (g) Thailand.

The IRF are calculated from the results of base 5-VAR model with daily endogenous variables VWBASPREAD, RETURN, VOLAT, TURNOVER, and FINETBUY. The base 5-VAR is estimated with a constant, and two lags for all cases, except Korea and Taiwan with three lags. The IRF are obtained after a Cholesky decomposition, assuming an ordering VWBASPREAD, RETURN, VOLAT, TURNOVER, and FINETBUY. VWBASPREAD: log of the value-weighted average of the proportional quoted spread in the market, calculated from individual quotes for the day. Quotes from Bloomberg for all countries but Indonesia and South Africa, which come from Datastream. Value weights from individual market capitalization are from Datastream. Remaining market variables come mostly from Bloomberg, and are complemented with Datastream as needed: RETURN: change on the log of the main market index. VOLAT: absolute value of RETURN. TURNOVER, log of the market turnover estimated as Total trading value/Total market capitalization. FINETBUY: total net buys by foreigners in the market, calculated as: [Foreigners buys (LCU) − Foreign sales (LCU)]/Market capitalization (LCU). The variables were detrended and filtered out of deterministic trends, as explained in Section 4. Confidence intervals of 95% in dashed lines.

Finally, we examine one last issue the effects of foreign investors on volatility. Neither Panel B of Table 7, nor Panel B of Table 8 show foreign investor activity Granger causing volatility at the market level. The only exception is Taiwan in Table 8, but the causality is of the unexpected sign: increasing foreign ownership Granger-causes reduced volatility. Contemporaneous correlations between the residuals of the foreign variables and volatility are non-existent in Table 8 and ambiguous in Table 7.27 This result supports the findings of Bekaert & Harvey (1997) and Kim & Singal (2000) of no significant effects on volatility from foreign investors.

5. Conclusion

The international finance literature has recently advanced in its understanding of the role of foreign investors in emerging markets. More is known about the effects of foreigners on information, price formation, volatility, and returns in those markets. We contribute to that literature by estimating the effects of foreign trading and foreign ownership on liquidity.

Overall, we have found that foreign trading has a harmful, although short lived, effect on liquidity, at the firm level as well as at the market level. The literature offers two possible explanations for this effect. Although this could be explained in terms of foreign investors being, on average, better informed than locals, the evidence presented here tilts to a rather mechanic reason: foreigners seem to be per se more aggressive liquidity demanders than locals.

However, we have found some evidence that firm ownership has beneficial effects on the liquidity of emerging markets, not only in a cross-sectional regression for individual stock in Indonesia, but also at the market level in India and Korea, among three other exchanges. At the firm level, the finding that foreign ownership is related to increased liquidity seems consistent with foreigners either preferring more liquid or transparent firms, or causing firms to become more informationally efficient. By contrast, at the market level, increasing foreign ownership in India and Korea appears to be causing an important and permanent improvement in liquidity in a horizon of approximately 2 weeks. Although the classical market microstructure models might suggest that foreigners improve the information environment, we do not find this explanation compelling at such a short horizon. Instead, we offer as an alternative explanation that foreign buying (selling) attracts noise traders and liquidity providers to the market, very much in the same manner that positive returns seem to do.

All in all, in the very short term, foreigners are “foes” of the liquidity of emerging markets, because of their liquidity demand and information effects. However, these effects are very short lived, and markets seems resilient to them. By contrast, the evidence portrays foreign ownership as a positive signal for liquidity in stocks and markets in a span of days and weeks. Therefore, disruptive very short-term effects from foreigners on markets should be taken into consideration against the more permanent benefits they cause.

Footnotes

  • 1

    Bonser-Neal et al. (2005) and Choe et al. (2005) study foreign trading on the stock markets of Indonesia and Korea, respectively, using transaction data, but not directly testing effects of foreigners on liquidity.

  • 2

    Some cross-country studies, such as Domowitz & Madhavan (2001), Levine & Schmukler (2005) and Bortolotti et al. (2004) have opted to use turnover as a proxy for liquidity. For a criticism of the use of turnover as a proxy for liquidity, see Lesmond (2005).

  • 3

    By contrast, Edwards (1999) argues against capital controls in emerging countries, stating that they are costly and ineffective in avoiding crises, and that they foster corruption.

  • 4

    However, the focus of those studies have been mainly to understand the characteristics and dynamics of institutional flows rather than their market-wide effects (see, e.g. Froot et al., 2001; Froot & Ramadorai, 2002; Richards, 2005).

  • 5

    The Indonesian database is used by Agudelo (2005) to measure the probability of informed trading by locals and foreigners. The Bloomberg databases of international bid-ask prices and foreign buys and sales have only been used recently. For example, Lesmond (2005) uses it for his cross-sectional study of liquidity of emerging markets, Richards (2005) to study effects of foreign flows on returns and volatilities in six emerging markets, and Griffin et al. (2005) to investigate the role of different types of investors in the relation between volume and return.

  • 6

    In a related topic, Karolyi (2004) and Levine & Schmukler (2005) present evidence that the cross-listing of stocks of emerging market firms as American Depositary Receipts hurts the turnover of the home market. The second study provides evidence consistent with a story of diversion of trading from the home market to the US market upon the cross-listing.

  • 7

    This is not contradictory to locals still having most of the information in the market as posed by the home bias literature, and Brennan & Cao (1997) and Griffin et al. (2004), especially for medium and small size stocks, as shown in Agudelo (2005).

  • 8

    Sadka (2006) discusses how the market-wide level of asymmetric information is priced in the USA and can explain the momentum and post-earning drift anomalies.

  • 9

    Chordia et al. (2000, p. 5) fail to relate the commonality on liquidity to common asymmetric information, consistent with their notion that “little covariation in liquidity would be induced by asymmetric information because few traders possess privileged information about broad market movements.” Although this assertion seems reasonable for the USA, it is not necessarily true for smaller and less diversified markets as the ones we study here, as shown by Chan & Hameed (2006).

  • 10

    In Korea and Taiwan, most foreign trading is carried out by institutions, as reported by Griffin et al. (2005) and Huang & Shiu (2005). Dennis & Weston (2001) provide evidence and cite references in support of the hypothesis that institutions have better information than individuals.

  • 11

    KOSDAQ, the second largest market from Korea, is not included in this analysis for the reduced participation of foreign investors. They traded only 1.1% during 1999–2001 (Richards, 2005, p. 5), and reached only 4.6% on average in 2004.

  • 12

    They are estimated as the number of outstanding shares (NOSH) multiplied by the adjusted price. As NOSH data from Bloomberg is incomplete and unreliable, we use the Datastream NOSH. This requires that, for some countries, we merge the two databases using the SEDOL numbers, and manually complement this using the ticker and name of the stock. The percentage of matching firms by country was 54% for India, 99% for Korea, 97% for the Philippines, 97% for Taiwan and 83% for Thailand. Both databases have quotes for more than 5 years for all the seven emerging markets, except India, and tend to agree quite well. Datastream quotes were preferred for the cases of Indonesia and South Africa for larger span and coverage.

  • 13

    Having two or more alternative sources for each market variable allows us to check for entering data mistakes and to fill up missing values.

  • 14

    As the JSE provides the total buying and selling by foreigners in the four boards (Regular, Negotiated, Crossing, and Cash), for consistency, we use the total volume of the four boards, per stock and day, using the definitions of FITRADE and FIPRESSURE.

  • 15

    A better proxy of the information content, namely the order imbalance by foreigners, would have required transaction data.

  • 16

    See, for example, Chordia et al. (2000) and Fujimoto (2004).

  • 17

    For example, it is hard to think either that the causality runs from higher liquidity to higher foreign trade or of any omitted factor that might be causing the relation between liquidity and foreign trading.

  • 18

    Again, this result is robust after running it separately in the first and second halves of the sample, and in the lower and upper halves by size decile (not reported).

  • 19

    There are two dummy variables for India (23 July 2002 and 30 September 2002 to 30 July 2003), four for Indonesia (23 March 1998, 24 March 1998, 20 May 1998 and 22 May 1998), one for Korea (31 March 1997), two for the Philippines (6 November 2000 and 22 January 2001), four for South Africa (28 October 1997, 29 October 1997, 24 November 1997 and 28 December 1999), one for Taiwan (3 January 2005 to 25 February 2005), and one for Thailand (25 April 2000).

  • 20

    This ambiguous result might be explained by the fact that we are not controlling for order imbalance, which should have a negative effect on spreads, and might have a confounding effect on the turnover variable.

  • 21

    This requires that we first orthogonalize the innovations using the Cholesky decomposition, which, in turn, requires posing an ordering of the variables. The IRF are initially calculated in the following ordering: logVWspread, return, volatility and turnover, which conservatively understate the effects of the other variables on liquidity. However, for robustness, we also verify the IRF obtained reversing this ordering and indicate in the text whenever the significance or direction of the IRF qualitatively changes with the ordering.

  • 22

    Having three significant correlations out of six is highly significant: the probability of having three or more significant correlations out of six by chance is 0.2%, assuming independence and having a 5% individual significance.

  • 23

    Foreigner effects on market liquidity might not show up in the VAR models for three countries but probably occur at the firm level, as in the case of Indonesia. For this country, the panel data result clearly shows a negative relationship between foreign trading and liquidity at the firm level but not in the VAR model at the market level. Indeed, VAR models, as any other econometric model, because of adjusting and averaging data across the market, can obscure effects that do occur at the firm level.

  • 24

    Having two significant relations out of seven is significant at the customary level: the probability of having two significant correlations out of six by chance is 4.4 %, assuming independence and having a 5% individual significance.

  • 25

    The negative effects on the spreads appear for all the countries, and Granger causality is significantly obtained at the 5% level for India, Korea, and South Africa, and marginally at the 10% level for Taiwan.

  • 26

    No significant lagged or contemporaneous effect on spreads is observed either, using, as the foreign variable, the absolute value of market foreign net buys.

  • 27

    Panel A of Table 8 shows two countries with positive correlation between foreign activity and volatility, India and Korea, and one with negative correlation, Indonesia. Although one might suppose that this effect is stronger in times of financial crises, note that for all the countries, except India and the Philippines, the sample includes the Russian crises, and that for Indonesia, Korea, South Africa, and Taiwan, the sample includes the Asian crisis.

Acknowledgments

This is the second essay of my dissertation on the Finance PhD program at Indiana University. I am greatly indebted to my advisor Craig Holden for his guidance and support. I also thank Utpal Bhattacharya, Sam Henkel, Pab Jotikastshira, Christian Lundblad, Chuck Trzcinka and participants at three Indiana University seminars for useful comments. I acknowledge the economic support of EAFIT University, Colombia and Indiana University CIBER. All remaining errors are mine.

Ancillary