Market noise, investor sentiment, and institutional investors in the ADR market.

Li, DeQing ; Jin, Jongdae

ABSTRACT

This study examines the effects of market noise in the ADR market. We find ADR return affected by noise trader risk and increases (decreases) when investors are irrationally optimistic (pessimistic). Our results also suggest institutional investors have engaged in stealth trading to exploit their information advantage in the noisy ADR market. Through a Granger causality regression, we find the returns on ADR portfolios with high institutional ownership lead the returns of those with low institutional ownership in the low-noise period, confirming that institutional trades reflect market information that is ultimately incorporated into other securities. Finally, we find institutional investors help reduce volatilities of European ADRs. However, for ADRs of Asian and South American firms, magnitude of the stabilizing arbitrage positions taken by rational investors is insignificant.

INTRODUCTION

Fischer Black (1986) suggests that noise is as influential as information in financial markets. Investors who trade on noise are willing to trade even though it is better for them not to trade. They do so because they think the noise on which they base their trading is information.

From existing literature, we can identify three possible effects of noise on securities trading. First, market noise leads to the existence of noise trader risk. De Long et al. (1990) develop a noise trader risk model which argues that when investment decisions are made based on market noise, the decisions are irrational and unpredictable because they are led by investor sentiment in general. Hence, noise traders become a source of risk in the finanical markets. Second, the existence of noise in capital markets provides an opportunity for informed institutional investors to exploit their information advantage. Barclay and Warner (1993) show that informed institutional investors are more likely to engage in "stealth trading" strategies in which the institutions spread their trades gradually over time. Third, the irrational behavior of noise traders in a noisy market may cause asset prices to move away from their fundamental values and destabilize the market. On the other hand, rational institutional investors would take positions opposite to those of the noise traders and help stabilize the market despite De Long et al. (1990) predict that institutional investors would fail to totally encounter the irrational activities of noise traders.

We examine the three possible effects of noise in the ADR market. Our results show that ADR return is affected by investor sentiment in the ADR market. ADR return increases (decreases) when investors are irrationally optimistic (pessimistic). We also find that in the lownoise period, ADRs with high institutional ownership exhibit autocorrelation similar to ADRs with low institutional ownership. However, in the high-noise period, ADRs with high institutional ownership exhibit significant higher autocorrelation than ADRs with low institutional ownership. The result implies institutional investors may have engaged in stealth trading to expolit a noisy market. Through a Granger causality regression, we find returns on ADR portfolios with high institutional ownership lead the returns of those with low institutional ownership in the low-noise period, confirming that institutional trades reflect market information that is ultimately incorporated into other securities. Finally, we find that institutional investors help reduce volatility of European ADR returns. However, for ADRs of Asian and South American firms, the magnitude of the stabilizing arbitrage positions taken by institutional investors is insignificant.

LITERATURE AND MOTIVATION

Financial economists have hypothesized the existence of noise trading in stock markets (for example, Black (1986), Trueman (1988), De Long et al. (1989), (1990), Palomino (1996)). While Black (1986) does not give a reason why investors would rationally want to engage in noise trading, he asserts that it must account for an important fraction of total trading in securities markets. Trueman (1988) suggests that an investment manager has incentive to engage in noise trading because of the positive signal about his ability to collect private information. De Long et al. (1990) develop a noise trader risk model in which irrational noise trader sentiment drives security prices from their fundamental values. The tendency of noise traders to trade according to their sentiment renders their investment behavior totally unpredictable. According to the model, assets subject to unpredictable changes in investor sentiment must be underpriced in the market relative to their fundamental values. An application of this argument is the discounts of closed-end funds. A high level of noise trader risk is associated with large closed-end fund discounts, and a low level of noise trader risk is assoicated with small closed-end fund discounts. Moreover, movements in closed-end fund discounts result primarily from individual investors' irrational, but correlated trading patterns. Though De Long et al. (1990) suggest that rational institutional investors will take positions to offset the irrational tradings of individual investors, they also predict institutional investors would fail to fully offset the irrational behavior of individual investors.

Empirical studies providing direct evidence of noise trading have been very few. Golec (1997) examine bond activities of retailers after the release of weekly retail statistics by Johnson Reebok Service and find direct evidence that bond traders indeed trade on noise. Lee, Shleifer, and Thaler (1991) provide indirect evidence of noise trading by showing a significant link between investor sentiment and discounts of closed-end funds. They show that fluctuations in discounts of closed-end funds reflect changes in investor sentiment. That is, widening(narrowing) discounts reflect the irrational pessimism(optimism) of individual investors. Barclay and Warner (1993) confirm the presence of stealth trading among institutional investors and thus provide indirect evidence of the existence of market noise.

Regarding market destablization, the traditional theoretical view is that asset prices do not deviate significantly from their fundamental values as a result of noise trading. It is argued that incentives exist for skillful, rational speculators to compete against noise traders, and that these speculators are the marginal, price-setting investors (Friedman (1953), and Fama (1965)). However, De Long et al. (1990) suggest that asset prices can be much more volatile than traditioanl models would allow because rational arbitrageurs with short horizons will not offset noisy variations in asset price today given the self-fulling belief that asset prices will vary unpredictably with market noise in future. As a result, the noise trader risk caused by investor sentiment is unpredictable and renders rational arbitrages ineffective. Palomino (1996) echos this suggestion by saying that nosie traders are agents with unpredictable beliefs and that the willingness of arbitrageurs to exploit noise traders' misconceptions is low in a capital market that is less than perfect. Empirical evidence on whether irrational (noise traders) investors destabilize financial markets or rational (institutional investors) traders stabilize markets in a noisy environment is, however, lacking.

While theoretical papers on noise trading are many, empirical literature is rare and indirect. As such, this study examines the effects of noise in the American Depository Receipts (ADRs) market. The ADR market presents an interesting scenerio for studying this topic because of several reasons. First of all, Kim, Szakmary, and Mathur (2000) and Patro (2000) have shown that home-country information has a significant impact on ADR return. Given the difficulty in getting accurate information from foreign countries, investors in the ADR market are likely to subject to a considerable amount of market noise. Second, institutions are major players in the ADR market and they usually have better access to information about foreign companies. Evidence of stealth trading by institutional investors could therefore confirm the presence of a noisy ADR market in which insitutional investors exploit their information advantage. Third, the simultanoues presence of noise and informed investors in the ADRs market allows us to investigate if the interactions between noise traders and rational investors stabilize or destabilize asset prices. In short, the ADR market presents an unique environment in which we can examine the above-mentioned effects of market noise directly and simultaneously, rather than indirectly and separately, in a noisy environment.

DATA AND VARIABLES DEFINITIONS

Data

The sample analyzed in this study contains ADRs from 1995 to 2000.The sample period starts from 1995 because complete information about monthly discounts of closed-end country funds is available from the Standard and Poor's Security Owners' Stock Guide only after 1995. Daily returns of ADRs are obtained from the Center for Research in Security Prices (CRSP) database and converted into monthly returns. The numbers of shares held by institutional investors and shares outstanding are obtained from the Standard and Poor's Security Owners' Stock Guide. The market equity capitalization is determined by multiplying price with number of outstanding shares of the ADR.

The ADRs are grouped into three portfolios based on their continent of origin: Asia, Europe, and South America. Each continent's ADR portfolio is further divided into two groups, those with high (above the median) institutional ownership and those with low (below the median) institutional ownership.

The following table shows the sample distribution by year:

ADR distribution by year

 Number Number Number of
 of Asian of European South American
Year ADRs ADRs ADRs

1995 33 75 56
1996 44 95 60
1997 46 123 72
1998 50 127 71
1999 54 129 73
2000 56 132 74

Variables definitions

Following Lee, Shleifer, and Thaler (1991), we use the change in closed-end fund discount [DELTA] discount) to measure the amount of noise trader risk. For our purpose, we use closed-end country funds. The discount of each closed-end country fund is the difference between the fund's net asset value and its price divided by the net asset value. By grouping all the closed-end country funds in the US into Asian, European, and South American funds, the average change in discount [DELTA] discount) of the funds in each group serves as a proxy for investor sentiment regarding the investment outlook of the continent. According to Lee, Shleifer, and Thaler (1991), a widening of the discounts implies investors are more pessimistic whereas a narrowing of the discounts implies investors are more optimistic. De Long et al. (1990) and Lee, Shleifer, and Thaler (1991) have used the terms 'noise trader risk' and 'investor sentiment' interchangeably. Both noise trader risk and investor sentiment refer to the irrational behavior of investors. Noise trader risk, however, is not exactly the same as the market noise described by Black (1986). In the words of Fisher Black, "I use the word "noise" in several senses. Noise is contrasted with information. Noise is what makes our observations imperfect. Noise is the arbitrary element in expectations." That is, noise is something that is anti-information and thus not investor sentiment per se.

The literature has not yet developed a proxy to measure noise in the investment markets. In this study, we propose to use the level of closed-end country fund discount as a proxy for market noise. Our reason is that in a noisy market, noise trader risk is high because investor sentiment will change more abruptly in such an environment where there is an abundant supply of stimulus. In a less noisy market, noise trader risk is low because there are less stimulus to cause investor sentiment to shift suddenly. Given that the change in closed-end country fund discount [DELTA] discount) would be higher (lower) when the level of closed-end fund discount is high (low), it is therefore reasonable to suggest that the level of closed-end country fund discount could serve as a proxy for market noise of the given continent. A large discount implies the continent's market is noisy, and a small discount implies the continent's market is less noisy. (1) Consequently, a year is classified as either a high-noise year or low-noise year when the discount in that year is larger or smaller than the median. The average discounts in the high-noise and low-noise periods for Asia, Europe and South America are shown in the following table, and the F-statistic is calculated to test the null hypothesis that the average discounts in the high-noise and low-noise periods are equal.

Closed-end country funds average discounts (%) in
high-noise and low-noise periods

 Low-noise High-noise
 Continent period period F-statistic

Asia 3.8154 11.4722 20.44 (a)
Europe 14.9132 16.1234 2.52
South America 9.737 22.4369 46.38 (a)

EFFECTS OF INVESTOR SENTIMENT AND INSTITUTIONAL INVESTORS ON ADR RETURNS

Investing in ADR provides a convenient way for diversifying portfolio risk internationally. As a result, the ADR market has experienced an explosive growth in the last 30 years. In 1970, there were only 18 ADRs traded in the U.S. In the year 2000, the number of listed ADRs had increased to 475. Although the ADR market is dominated by institutional investors, the difficulty of obtaining accurate and complete information from foreign countries suggests that influence of noise can be considerable in this market.

First of all, we study the effects of investor sentiment and institutional ownership in the ADRs market. The following regression is performed:

Rt = [a.sub.0] + [a.sub.1] [R.sub.t-1] + [a.sub.2] [DELTA] [Discount.sub.t] + [a.sub.3] [DELTA] Institutional [Ownership.sub.t] + [[epsilon].sub.t]

where [R.sub.t] is the compounded monthly ADR portfolio return at time t for each continent and [R.sub.t-1] is the ADR portfolio return at time t-1. [DELTA] Discount is the change in the average discount of close-end country fund from period t to t-1 for each continent. According to Lee, shleifer, and Thaler (1991), when the change in average discount ([DELTA] Discount) is positive (i.e., the average discount widens), individual investors are more pessimistic and asset returns would be affected negatively. Conversely, when) Discount is negative, the individual investors are more optimistic and asset returns would be affected positively. Thus, if investor sentiment is priced in the ADR market, the coefficient of) Discount should be negative and significant. Lee, Shleifer, and Thaler (1991) report a significant negative relation between the returns of NYSE stocks and the average [DELTA] Discount of a basket of domestic closed-end funds.

[DELTA] Institutional Ownership is the change in the ratio of institutional ownership from month t to month t-1 for each continent's ADR portfolio. A priori, we expect ADR return to be positively correlated with [DELTA] Institutional ownership. That is, ADR return would be higher or lower when institutions increase or decrease their holdings. The [R.sub.t-1] is for controlling the effect for serial correlation in ADR return.

The regression results for each continent are shown in Table I.

In Table I, it is shown that the coefficients of [R.sub.t-1] are 0.2470, 0.3190, and 0.3870, for Asia, Europe, and South America respectively. The t-statistics are 2.29, 2.74, and 3.68 and all are significant at the 5% level, implying that there is positive autocorrelation in ADR portfolios returns. The coefficients of [DELTA] Discount have the expected negative signs and are -0.0056 for Asia, -0.0060 for Europe, and -0.0113 for South America respectively. All their t-statistics are significant at the 1% level. That is, ADR return is affected by investor sentiment in the ADR market. When investor sentiment becomes irrationally optimistic or pessimistic, as reflected by a narrowing or widening of the discount of closed-end country funds, ADR return of the same continent moves higher or lower correspondingly. The result is consistent with that of Lee, Shleifer, and Thaler (1991). For Asian and South American ADRs, the coefficients of [DELTA] Institutional Ownership are positive and significant, that is, there is a positive relation between changes in institutional ownership and ADR portfolio returns. The coefficient of [DELTA] Institutional Ownership is also positive for Europe, though insignificant. It is possible that the information about European countries is more accessible than that of Asian and South American countries, the role of institutional ownership of European ADRs is therefore less influential. This conjecture is consistent with our earlier observation that the noise levels of the high-noise and low-noise periods are similar for Europe.

In the noise trader risk model of DeLong et al. (1990), they suggest that rational institutional investors may exploit irrational behavior of noise traders by taking positions opposite to those of the noise traders. However, the model also predicts that institutional investors would not be completely successful because the unpredictable noise trading will render the arbitrage activities of institutional investors futile. The significantly negative coefficients of [DELTA] Discount in Table I support the postulations of the noise trader risk model of DeLong et al. (1990). That is, investor sentiment has a significant effect even in the presence of rational institutional investors. In other words, institutional investors are unable to neutralize the effect of trading led by irrational investor sentiment.

Table I shows that noise trader risk is important even in the presence of institutional investors. It would be of interest to know then if the impacts of investor sentiment and institutional ownership on the ADR return are different in the high-noise and low-noise periods. To study this, we perform the previous regression on high-noise years and low-noise years separately. Regression results are shown in Table II.

Table II shows that investor sentiment is important in determining ADR return in both the high-noise and low-noise periods. However, change in institutional ownership has a significant impact on the returns of Asian and South American ADRs only during the high-noise period. Institutional ownership is not significant at all in the low-noise period. Conceivably, when the market is noisy (such as Asia and South American), the information possessed by institutional investors becomes more important. During low-noise period, the information advantage of institutional investors may be less significant. This is probably why institutional ownership does not play a significant role in the pricing of European ADRs in both the high-noise and low-noise periods because information about European markets is more accurate and readily available to investors.

MARKET NOISE AND ADR RETURN AUTOCORRELATION

Table I and II confirm that noise trader risk is present in the ADR market. If the ADR market is noisy, then the private information of institutional investors would be valuable and it is logical that institutional investors will exploit their informational advantage. One possible way to do so is the use of "stealth trading" strategies in which institutional investors spread their trades gradually over time. According to Barclay and Warner (1993), stealth trading would induce ADR return autocorrelation. While insitutional investors may stealth trade frequently in the ADR market, we expect the likelihood to be higher in the high-noise period than the low-noise period. Thus, we expect that in the high-noise period, ADRs with high institutional ownership would exhibit significant higher autocorrelation than ADRs with low institutional ownership. In the low-noise period, we expect ADRs with high institutional ownership to exhibit similar or higher autocorrelation than ADRs with low institutional ownership. The return autocorrelations of all the individual ADRs in the high-noise and low-noise periods are shown in Table III.

Consistent with our expectation, panel A of Table III shows that in the low-noise period, for both Asia and South America, ADRs with high institutional ownership exhibit autocorrelations similar to ADRs with low institutional ownership. For Asia, the mean daily autocorrelation for individual ADRs with low institutional ownership and high institutional ownership are 0.0040 and 0.0164 respectively. The t-statistic is 0.46 and not significant. For South America, the mean daily autocorrelation for individual ADRs with low institutional ownership and high institutional ownership are 0.0185 and 0.0391 respectively. The t-statistic is 1.06 and not significant. For Europe, ADRs with high institutional ownership exhibit higher autocorrelation than ADRs with low institutional ownership.

For the high-noise period, panel B of Table III shows that ADRs with high institutional ownership exhibit significant higher autocorrelation than ADRs with low institutional ownership for Asia, Europe, and South America. For Asia, the mean daily autocorrelation for individual ADRs with low institutional ownership and high institutional ownership are -0.0030 and 0.0504 respectively. The t-statistic is 10.6, significant at the 1% level. For Europe, the mean daily autocorrelation for individual ADRs with low institutional ownership and high institutional ownership are -0.0169 and 0.0311 respectively. The t-statistic is 16.26 and significant at the 1% level. For South America, similar result is obtained.

In sum, the results in table III support our earlier conjecture that institutional investors exploit their information advantage in the noisy ADR market.

CROSS-PREDICTABILITY OF ADR PORTFOLIO RETURNS IN HIGH-NOISE AND LOW-NOISE PERIODS

From the above, we find that noise is present in the ADR market and institutional investors react differently in high-noise and low-noise environments. In order to confirm that institutional trades contain information not found in non-institutional trades, a Granger causality regression model is used. For each continent's ADR portfolio the following regressions are performed for the high-noise and low-noise periods separately:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

where [R.sub.high, t] and [R.sub.low, t] are the returns at time t for ADR portfolios with high and low institutional ownership, the [d.sub.i, t] are dummy variables for each day of the week i, k is the lag in days and u is the error term.

According to Brennan et al. (1993), portfolios that are first to reflect market-wide information have a better ability to predict the returns of portfolios that are late to reflect marketwide information than the ability of the latter to predict the former. That is, if institutional investors trade on information, returns on portfolios with high institutional ownership should lead the returns of those portfolios that have low institutional ownership. For both the low-noise and the high-noise periods, we therefore expect returns on ADR portfolios with high institutional ownership to lead the returns of those with low institutional ownership if institutions trade on information. That is, we expect [R.sub.high, t-k] to predict [R.sub.low, t] better than [R.sub.low, t-k] to predict [R.sub.high, t]. In the Granger causality regressions, we therefore expect [b.sub.high, k] to be larger than [a.sub.low, k].

Panel A of Table IV shows that in the low-noise period, returns on ADR portfolios with high institutional ownership lead the returns of those with low institutional ownership for all three continents. For Asia, [a.sub.low] is 0.0172, and [b.sub.high] is 0.0770. For Europe, [a.sub.low] is -0.0104, and [b.sub.high] is 0.0351. For South America, [a.sub.low] is -0.0290, and [b.sub.high is 0.0825. That is, for all the three continents, [a.sub.low] is less than [b.sub.high] . The F-statistics, Wilcoxon Z - values, and Kruskal-Wallis Chi-squares are all significant at the 5% level. These results show that the ability of [R.sub.high, t-k] to predict [R.sub.low, t] is much greater than the ability of [R.sub.low, t-k] to predict [R.sub.high, t]. That is, even though the market noise is low (relatively speaking) in the low-noise period, ADR portfolios with high institutional ownership still reflect market-wide information sooner than ADR portfolios with less institutional ownership.

In the high-noise period, we observe unexpected results. The returns of high institutional ownership ADR portfolios do not lead the returns of those with low institutional ownership for Asia and South America. For Asia, [a.sub.low] is 0.0031, and [b.sub.high] is 0.0213. For South America, [a.sub.low] is 0.0385, and [b.sub.high] is 0.0807. Despite in both cases, the size of [b.sub.high] is larger than the size of [a.sub.low], the F-statistics, Wilcoxon Z values, and Kruskal-Wallis Chi-squares are all insignificant. These results mean that we cannot reject the null hypothesis that [b.sub.high] = [a.sub.low], that is, the ability of [R.sub.high, t-k] to predict [R.sub.low, t] is not much greater than the ability of [R.sub.low, t-k] to predict [R.sub.high, t]. We think there are two possible reasons for these results. One reason may be that in the high-noise period, institutions deliberately divulge their information very slowly over time through stealth trading, making their information advantage less useful for others to predict returns. This is consistent with our earlier results in Table III that insitutions stealth trade particularly in the high-noise period. The other possible reason is that in the high-noise period risk exposure is conceivably higher for investments in Asian and South American ADRs, institutional investors may be affected by their risk concern such that their ability to impound information in ADR prices is affected. Sias and Stark (1997) suggest that if institutional investors are motivated to trade for reasons not associated with information, then there is no reason to expect the returns on portfolios with high institutional ownership to lead the returns on portfolios with low institutional ownership. For European ADRs, the risk is conceivably lower than those of Asian and South American ADRs, returns on portfolios with high institutional ownership lead the returns on portfolios with low institutional ownership because institutional investors' ability to impound information in ADR prices is less affected by risk concern. This conjecture regarding the concern of risk by institutional investors is consistent with the results in the following section.

INSTITUTIONAL INVESTORS IN THE ADR MARKET: DESTABILIZING OR STABILIZING?

Noise traders move ADR prices away from their fundamental values as investment decisions are led by investor sentiment. One observable consequence is that the ADR return volatility would be higher in the high-noise period. The numbers in the following table confirms this; implying noise traders destabilize financial market.

ADR Return Volatility

 Low-noise High-noise T-statistic

Asia 0.0230 0.0308 -13.96 (a)
Europe 0.0232 0.0269 -14.66 (a)
South America 0.0282 0.0347 -10.81 (a)

On the other hand, De Long et al. (1990) suggest that rational investors such as institutions will offset, though incomplete, the irrational activities of the noise traders. Given such postulation, the next logical question is whether institutional investors help destabilize or stabilize volatility of the ADR market. The following regression is performed to answer this question.

[[delta].sup.2.sub.it] = a0 + a1 [[delta].sup.2.sub.i,t-1] + a2 ([DELTA] Discount)t + a3 ([DELTA] Institutional Ownership) t + [epsilon]t

where [[delta].sup.2.sub.it] is the volatility of the ADR return in time period t for each portfolio, and [[delta].sup.2.sub.i,t-1] is the volatility of the ADR return in time period t-1.

Table V shows that for Europe, the coefficient of the change in institutional ownership is -0.8370 and it is significant at the 5% level, that is, institutional ownership is negatively related to the volatility of the stock return. The result means that institutional investors help stabilize the market of European ADRs. That is, institutions have helped offset the irrational behavior of noise traders. The coefficients of the change in institutional ownership for Asia and South America, on the other hand, are negative but insignificant. That is, institutional investors have no significant effect on the volatilities of the ADRs from these two continents. This result deserves some explanations.

In finance literature, it is well known that rational investors arbitrage and bring prices closer to fundamental values. The effectiveness of arbitrageurs however relies crucially on the stabilizing powers of rational speculation. Some studies have questioned the effectiveness of such speculation in the presence of risk aversion. For example, DeLong et al. (1987) show that the unpredictability of noise traders' beliefs creates a risk that deters rational arbitrageurs from aggressively betting against them, and rational speculation is thus less effective. Figlewski (1979) also shows that it might take a very long time for noise traders to lose most of their money if rational investors must bear fundamental risk in betting against them, and such fundamental risk deters rational speculation. Both of these two papers suggest that the magnitude of the stabilizing arbitrage positions taken by rational investors might be limited. Investors may regard Asia and South America as more risky when compared with Europe, and rational investors are therefore less likely to counter the unpredictable noise trader risk in Asia and South America. Thus, the magnitude of the stabilizing arbitrage positions taken by rational investors might be small and insignificant for both Asia and South America.

The coefficients of [DELTA] Discount are all negative, though only significant for Asia and South America. That is, noise trader risk affects ADRs volatility. This is consistent with DeLong et al. (1990) that noise trading is a source of risk, particularly in Asian and South American financial markets.

We also perform the above regression for the high-noise period and the low-noise period separately, and the results are shown in Table VI.

Results similar to those of Table V are found. Table VI shows that for Europe, the coefficients of the change in institutional ownership are negative and significant in both the highnoise and low-noise periods. Again, this may be due to the lesser degree of risk aversion among arbitragers in this market. For Asia and South America, the coefficients of the change in institutional ownership are not significant in either the high-noise period or the low-noise period. For Asia and South America, the aversion to risk greatly limits rational investors' willingness to bet against noise traders in both the high-noise and low-noise periods.

SUMMARY

This study examines the effects of market noise in the American Depository Receipts (ADRs) market. From existing literature, we can identify three possible effects of noise on securities trading. First, market noise leads to the existence of noise trader risk. Second, the existence of noise in capital markets provides an opportunity for informed institutional investors to exploit their information advantage through stealth trading. Third, the irrational behavior of noise traders in a noisy market may cause the market to destabilize, though rational institutional investors would take positions opposite to those of the noise traders and help stabilize the market. We examine the three possible effects of noise in the ADR market. The ADRs market presents an unique environment in which we can examine the above-mentioned effects of noise directly and simultaneously in a noisy environment.

Our results show that the ADR return is affected by investor sentiment (noise trader risk) in the ADR market. ADR return increases (decreases) when investors are irrationally optimistic (pessimistic). We also find that in the low-noise period, ADRs with high institutional ownership exhibit autocorrelation similar to ADRs with low institutional ownership. However, in the highnoise period, ADRs with high institutional ownership exhibit significant higher autocorrelation than ADRs with low institutional ownership. The result implies institutional investors may have engaged in stealth trading. Through a Granger causality regression, we find returns on ADR portfolios with high institutional ownership lead the returns of those with low institutional ownership in the low-noise period, confirming that institutional trades reflect market information ultimately incorporated into other stocks. Finally, we find that rational investors help stabilize ADRs market in Europe. However, for Asia and South America, the magnitude of the stabilizing arbitrage positions taken by rational investors is insignificant.

ENDNOTES

Since noise and .Discount may be correlated and cause selection bias, we perform tests for difference in means of [DELTA] Discount between the high-noise and low-noise periods for each portfolio. All the test statistics are insignificant, showing no selection bias.

REFERENCES

Black, Fischer, (1986), Noise, Journal of finance 41, 529-543.

Barclay, M.J, Warner, J.B, (1993). Stealth trading and volatility: which trades move price? Journal of Financial Economics 34, 281-305.

Brennan, M.J., Jegadeesh, N. & Swaminathan, B., (1993), Investment Analysis and The Adjustment Of Stock Prices To Common Information, Review Of Financial Studies 6, 799-824. Brett Trueman, (1988), A theory of noise trading in securities markets, Journal of Finance 43, 83-95.

De Long, J. Bradford, A. Shleifer, L. H. Summers & R.J. Waldmann, (1989), Positive Feedback Investment Strategies And Destabilizing Rational Speculation, Journal of Finance 45, 379-395. De Long, J. Bradford, A. Shleifer, L. H. Summers & R. J. Waldmann, (1990), Noise trader risk in financial markets, Journal of political economy, 703-38.

Fama, E.,(1965), "The behavior of stock market prices,' Journal of business, 34-105.

Figlewski, Stephen, (1979), Subjective information and market efficiency in a betting market, Journal of political economy 87, 75-88.

Friedman, Milton, (1953), The case for flexible exchange rates, in Essays in positive economics, Chicago: University of Chicago Press.

Golec, J., (1997), Herding on noise: the case of Johnson Redbook's weekly retail sales data, Journal of Financial and Quantitative Analysis 32, 367-381.

Kim, Minho, Szakmary, A.C. & Ike Mathur, (2000), Price transmission dynamics between ADRs and their underlying foreign securities, Journal of Banking and Finance 24, 1359-1382.

Lee, Charles, Andrei Shleifer & Richard H. Thaler, (1991), Investor sentiment and the closed end funds puzzle, Journal of finance, 75-109.

Palomino, F., (1996), Noise trading in small markets, Journal of finance, 1537-50.

Patro, Dilip Kumar, (2000_. Return behavior and pricing of American depository receipts. Journal of international financial markets, institutions and money 10, 43-67.

Shleifer, Andrei & Lawrence H. Summers, (1990),The noise trader approaches to finance, Journal of economic perspectives, Vol 4, Spring, 19-33.

DeQing Diane Li, University of Maryland Eastern Shore

Jongdae Jin, University of Maryland Eastern Shore

TABLE I: Effects of Investor Sentiment and Institutional
Investor on ADR return

Each year, all ADRs are grouped into three portfolios
based on their country of origin: Asia, Europe, and
South America. Rt is the ADRs portfolio return at time
t for each continent and Rt-1 is the ADR portfolio
return at time t-1 for each continent. We also group
all the closed-end country funds in US into Asian,
European, and South American funds. The discount
is the difference between the fund's net asset value
and its price divided by the net asset value. The
discount of each continent is the average discount of
the funds in each group, and [DELTA] Discount is the
difference of discount between month t and month t-1
for each continent. [DELTA] Institutional Ratio is
the change of the average institutional ownership
between month t and month t-1 for each continent.

 Model: [R.sub.t] = [a.sub.0] + [a.sub.1][R.sub.t-1]
 + [a.sub.2][DELTA] [Discount.sub.t] +
 [a.sub.3] [DELTA] Institutional [Ownership.sub.t] + [[epsilon].sub.t]

 [DELTA]
 Intercept [R.sub.t-1] Discount

All 0.0043 0.3110 -0.0075
ADRs (0.92) (5.02 (a)) (-9.19 (a))

Asia -0.0166 0.2470 -0.0056
 (-1.23) (2.29 (b)) (-4.93 (a))

Europe 0.0094 0.3190 -0.0060
 (1.79) (2.74 (a)) (-4.48 (a))

S. 0.0058 0.3870 -0.0113
America (0.67) (3.68 (a)) (-6.46 (a))

 [DELTA]
 Institutional Adjusted
 Ownership R-square

All 1.2690 0.3020
ADRs (2.01 (b))

Asia 3.5100 0.2910
 (2.10 (b))

Europe 0.7130 0.2410
 (1.28)

S. 4.4090 0.3820
America (1.67 (c))

(a) Significant at the 1% level.

(b) Significant at the 5% level.

(c) Significant at the 10% level.

TABLE II: Effects of Investor Sentiment and Institutional
Investor on ADR return

Model: [R.sub.t] = [a.sub.0] + [a.sub.1][R.sub.t-1]
+ [a.sub.2] [DELTA] [Discount.sub.t] + [[alpha].sub.3]
Institutional [Ownership.sub.t] + [[epsilon].sub.t]

A: Low-noise period:

 [DELTA]
 [Discount.
 Intercept [R.sub.t-1] sub.t]

All ADRs -0.0003 0.2990 -0.0056
 (-0.01) (3.41 (a)) (-5.33 (a))

Asia -0.0022 0.4710 -0.0029
 (-0.15) (3.20 (a)) (-2.38 (a))

Europe 0.0159 0.4050 -0.0037
 (0.39) (2.69 (a)) (-2.14 (a))

S.America 0.0132 0.4320 -0.0110
 (0.98) (2.88 (a)) (-4.25 (a))

B: High-noise period:

 [DELTA]
 [Discount.
 Intercept [R.sub.t-1] sub.t]

All ADRs -0.0092 0.1880 -0.0080
 (-1.24) (2.15) (b) (-5.79 (a))

Asia -0.0343 0.0456 -0.0072
 (-1.59) (0.72) (-3.61 (a))

Europe 0.0039 0.2500 -0.0060
 (0.50) (1.52) (-3.22a)

S.America -0.0242 0.0686 -0.0094
 (-1.54) (0.36) (-3.05 (a))

A: Low-noise period:

 [DELTA]
 Institutional Adjusted
 Ownership R-square

All ADRs 1.5230 0.25
 (1.60)

Asia 2.1500 0.22
 (1.16)

Europe 0.7800 0.28
 (1.09)

S.America 7.6920 0.39
 (1.36)

B: High-noise period:

 [DELTA]
 Institutional Adjusted
 Ownership R-square

All ADRs 1.8460 0.25
 (1.98 (b))

Asia 5.5660 0.30
 (2.01 (b))

Europe 0.4590 0.19
 (0.61)

S.America 8.0250 0.24
 (2.02 (b))

(a) Significant at the 1% level.

(b) Significant at the 5% level.

Table III: Return Autocorrelations of ADRs

The mean daily return autocorrelations for individual ADRs
in both the high-noise period and the low-noise period are
reported. The t-statistic is calculated to test the null
hypothesis that the mean daily return autocorrelation of
individual ADRs with high institutional ownership is equal
with the mean daily return autocorrelation of individual
ADRs with low institutional ownership.

A: Autocorrelation of individual ADRs in the low noise period

 Low High
 institutional institutional
 ownership ratio ownership ratio t-statistic

Asia 0.0040 0.0164 0.46
Europe -0.0215 0.0419 20.42 (a)
South
 America 0.0185 0.0391 1.06

B. Autocorrelation of individual ADRs in the high noise period:

 High
 Low institutional institutional
 Ownership ratio ownership ratio t-statistic

Asia -0.0030 0.0504 10.61 (a)
Europe -0.0169 0.0311 16.26 (a)
South
 America 0.0383 0.0767 5.68 (a)

(a) Significant at the 1% level.

TABLE IV: Cross-Predictability of ADR Portfolio Return

Each year, all ADRs are grouped into three portfolios based
on their country of origin: Asia, Europe, and South America.
Each continent's ADR portfolio is further divided into those
with high institutional ownership and those with low
institutional ownership. We also group all the closed-end
country funds in US into Asian, European, and South American
funds. The discount is the difference between the fund's net
asset value and its price divided by the net asset value. The
discount of each continent is the average discount of the
funds in each group. We further classify the years in which
the discount is larger than the median as high-noise years
and those years in which the discount is smaller than the
median as low-noise years for each continent. The daily
return of each continent portfolio with high (low) institutional
ownerships is regressed on its own previous five returns and the
previous five returns for the same continent portfolio with low
(high) institutional ownerships in both the high-noise period
and the low-noise period. The sums of the coefficients are reported
below. The F-statistic is calculated to test the null hypothesis
that the ability of the lagged return on the high institutional
portfolio to predict the return on the same continent portfolio
with low institutional ownership is the same as the ability of
the lagged return on the low institutional portfolio to predict
the return on the high institutional portfolio of the same
continent in both the high-noise period and the low-noise period.
Wilcoxon Z value and Kruskal-Wallis Chi-square are also shown in
Table IV.

TABLE IV: Cross-Predictability of ADR Portfolio Return

[R.sub.hight] =[5.summation over (i=1)] [a.sub.i][d.sub.i,t]
+ [5.summation over (k=1)] ([a.sub.hightk][R.sub.hight-k] +
[a.sub.lowk] [R.sub.lowt-k]) + [u.sub.hight], (1)

[R.sub.hight] =[5.summation over (i=1)] [b.sub.i][d.sub.i,t]
+ [5.summation over (k=1)] ([b.sub.hightk][R.sub.hight-k] +
[b.sub.lowk] [R.sub.lowt-k]) + [u.sub.hight], (2)

A: Low-noise period:

 [R.sub. [R.sub.
Dependent Variable high,t-k] low,t-k]

 (Independent Variable)

Asia [R.sub.high,t] 0.0500 0.0172
Asia [R.sub.low,t] 0.0770 -0.0247
Europe [R.sub.high,t] 0.0458 -0.0104
Europe [R.sub.low,t] 0.0351 -0.0316
S.Am [R.sub.high,t] 0.0518 -0.0290
S.Am [R.sub.low,t] 0.0825 -0.0555

B: High-noise period:

 [R.sub. [R.sub.
Dependent Variable high,t-k] low,t]-

 (Independent Variable)

Asia [R.sub.high,t] 0.0474 0.0031
Asia [R.sub.low,t] 0.0213 0.0088
Europe [R.sub.high,t] 0.0369 -0.0030
Europe [R.sub.low,t] 0.0561 -0.0092
S.Am [R.sub.high,t] 0.0713 0.0385
S.Am [R.sub.low,t] 0.0807 0.0183

A: Low-noise period:

 Wilcoxon
Dependent Variable F-statistic Z-value

Asia [R.sub.high,t] 4.46 (b) 1.87 (b)
Asia [R.sub.low,t]
Europe [R.sub.high,t] 5.20 (b) 2.10 (b)
Europe [R.sub.low,t]
S.Am [R.sub.high,t] 5.47 (b) 1.74 (b)
S.Am [R.sub.low,t]

B: High-noise period:

 Wilcoxon
Dependent Variable F-statistic Z-value

Asia [R.sub.high,t] 0.44 0.71
Asia [R.sub.low,t]
Europe [R.sub.high,t] 3.81 (b) 1.52 (c)
Europe [R.sub.low,t]
S.Am [R.sub.high,t] 0.62 0.21
S.Am [R.sub.low,t]

A: Low-noise period:

 Kruskal-Wallis
Dependent Variable Chi-square

Asia [R.sub.high,t] 3.58 (b)
Asia [R.sub.low,t]
Europe [R.sub.high,t] 4.50 (b)
Europe [R.sub.low,t]
S.Am [R.sub.high,t] 3.11 (c)
S.Am [R.sub.low,t]

B: High-noise period:

 Kruskal-Wallis
Dependent Variable Chi-square

Asia [R.sub.high,t] 0.53
Asia [R.sub.low,t]
Europe [R.sub.high,t] 2.36 (c)
Europe [R.sub.low,t]
S.Am [R.sub.high,t] 0.05
S.Am [R.sub.low,t]

(a) Significant at the1% level.

(b) Significant at the 5% level.

(c) Significant at the 10% level.

Table V: Effect of Institutional Investors on ADR Return Volatilities

The volatility of the return for time t for each continent's ADR
portfolio is then regressed on the volatility of the return for time
t-1, the difference of discount between month t and month t-1
([DELTA] Discount), and the change of the institutional ownership
between month t and month t-1 ([DELTA] Institutional Ratio).

[[delta].sup.2.sub.it] = a0 + a1 [[delta].sup.2.sub.i,t-1] + a2
([DELTA] Discount)t + a3 ([DELTA] Institutional Ownership)t +
[[epsilon].sub.t]

 Intercept [[delta].sup. [DELTA]
 2.sub.i,t-1] Discount

Asia 0.0518 0.4750 -0.0033
 (3.48) (4.72 (b)) (-3.76 (a))
Europe 0.0569 0.5090 -0.0011
 (4.43) (4.71 (a)) (-1.40)
S.America 0.1210 0.3960 -0.0027
 (1.82) (3.60 (a)) (-2.53 (a))

 [DELTA] Adjusted
 Institutional R-square
 Ownership

Asia 1.8900 0.3270
 (1.43)
Europe -0.8370 0.3870
 (-1.98 (b))
S.America -1.1090 0.1790
 (-0.58)

(a) Significant at the 1% level.

(b) Significant at the 5% level.

(c) Significant at the 10% level.

Table VI: Effect of Institutional Investors on ADR Return Volatility

The volatility of the return for time t for each continent's ADR
portfolio is regressed on the volatility of the return for time t -1,
the difference of discount between month t and month t-1 ([DELTA]
Discount), and the change of the institutional ownership between month
t and month t-1.

[[delta].sup.2.sub.it] = a0 + a1 [[delta].sup.2.sub.i,t-1] + a2
([DELTA] Discount) t + a3 ([DELTA] Institutional Ownership)t +
[[epsilon].sub.t]

A: Low-noise period:

 Intercept [[delta].sup. [DELTA]
 2.sub.i,t-1] Discount

Asia 0.0270 0.5650 -0.0037
 (1.24) (4.07 (b)) (-3.17 (a))
Europe 0.0593 0.5460 -0.0005
 (3.06) (3.35 (a)) (-0.68)
S.America 0.0342 0.7200 -0.0022
 (1.97) (5.50 (a)) (-2.18 (a))

 [DELTA] Adjusted
 Institutional R-square
 Ownership

Asia 2.5890 0.42
 (1.46)
Europe -1.4290 0.33
 (-2.12 (b))
S.America 1.9590 0.49
 (0.78)

B: High-noise period:

 Intercept [[delta].sup. [DELTA]
 2.sub.i,t-1] Discount

Asia 0.0804 0.3290 -0.0035
 (3.93) (2.06 (b)) (-2.31 (b))
Europe 0.0788 0.3100 -0.0014
 (4.01) (1.83 (c)) (-1.37)
S.America 0.1180 0.2240 -0.0032
 (4.46) (1.36) (-1.83 (c))

 [DELTA] Adjusted
 Institutional R-square
 Ownership

Asia 1.9100 0.19
 (0.87)
Europe -1.7020 0.46
 (-2.58 (a))
S.America 1.6620 0.13
 (0.63)

(a) Significant at the 1% level.

(b) Significant at the 5% level.

(c) Significant at the 10% level.