1 Do ADR Investors Herd? Evidence from Advanced and Emerging Markets Rıza Demirer, Ali M. Kutan and Huacheng Zhang † First draft: May 2008 This draft: March 2012 Abstract This paper extends the research on investor herds to American Depository Receipts (ADRs). Using daily price data on 305 ADRs traded in U.S. exchanges issued by corporations from 19 countries, we examine herding behavior in the market for ADRs within country-based portfolios. We also provide evidence from sector-based portfolios. There is significant evidence of herding behavior in the market for ADRs from Chile only regardless of alternative model specifications. On the other hand, we find significant effect of the Asian crisis and the recent credit market crisis on herding behavior in ADR issues from Korea and the U.K., respectively, suggesting a link between market crisis periods and herding behavior. Furthermore, we find no significant effects of currency rates (except Korea) or the performance of the market of origin on herding behavior among ADR issues. In the case of sector-based ADR portfolios, evidence of herding behavior exists in Basic Industries, Capital Goods, Food & Tobacco, and Textile & Trade, but only during periods of large market downturns. We next discuss implications for ADR investors. Key words: American Depository Receipts, Herding Behavior, Return Dispersions, Market Efficiency JEL Classification: G14, G15 † Rıza Demirer. Department of Economics and Finance, School of Business, Southern Illinois University Edwardsville, Edwardsville, IL 62026-1102. E-mail: [email protected]; Ali M. Kutan. Department of Economics and Finance, School of Business, Southern Illinois University Edwardsville; The William Davidson Institute, University of Michigan Business School; and The Emerging Markets Group, Sir Cass Business School, London. E-mail: [email protected]. Huacheng Zhang. Department of Finance, Eller College of Management, University of Arizona, Tucson, AZ, 85721. E-mail: [email protected].
23
Embed
Do ADR Investors Herd? Evidence from Advanced and Emerging ...zhangh/ADR herding.pdf · Our tests of country-based ADR portfolios indicate evidence of herding behavior in the ADRs
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Do ADR Investors Herd?
Evidence from Advanced and Emerging Markets
Rıza Demirer, Ali M. Kutan and Huacheng Zhang †
First draft: May 2008
This draft: March 2012
Abstract
This paper extends the research on investor herds to American Depository Receipts
(ADRs). Using daily price data on 305 ADRs traded in U.S. exchanges issued by
corporations from 19 countries, we examine herding behavior in the market for ADRs within
country-based portfolios. We also provide evidence from sector-based portfolios. There is
significant evidence of herding behavior in the market for ADRs from Chile only regardless
of alternative model specifications. On the other hand, we find significant effect of the Asian
crisis and the recent credit market crisis on herding behavior in ADR issues from Korea and
the U.K., respectively, suggesting a link between market crisis periods and herding behavior.
Furthermore, we find no significant effects of currency rates (except Korea) or the
performance of the market of origin on herding behavior among ADR issues. In the case of
sector-based ADR portfolios, evidence of herding behavior exists in Basic Industries, Capital
Goods, Food & Tobacco, and Textile & Trade, but only during periods of large market
downturns. We next discuss implications for ADR investors.
Key words: American Depository Receipts, Herding Behavior, Return Dispersions, Market
Efficiency
JEL Classification: G14, G15
† Rıza Demirer. Department of Economics and Finance, School of Business, Southern Illinois University
Edwardsville, Edwardsville, IL 62026-1102. E-mail: [email protected];
Ali M. Kutan. Department of Economics and Finance, School of Business, Southern Illinois University
Edwardsville; The William Davidson Institute, University of Michigan Business School; and The Emerging
Markets Group, Sir Cass Business School, London. E-mail: [email protected].
Huacheng Zhang. Department of Finance, Eller College of Management, University of Arizona, Tucson, AZ,
Herding behavior in financial markets has attracted much attention over the past
decade. The literature, in general, defines herding behavior as the tendency of investors to
mimic the actions of other investors, moving in and out of particular securities, industries or
markets in general as a group (Bikhchandani and Sharma, 2000). One implication of herding
behavior is that it drives security prices away from equilibrium values supported by
fundamentals, potentially leading to market bubbles and subsequent crashes. An increasing
number of published works in the literature has tested the existence of investor herds in a
number of domestic and global markets. In general, the literature provides market specific
results with the strongest support for herding behavior in mostly Asian markets. Interestingly,
the analysis of investor herds has not yet been extended to American Depository Receipts
(ADRs). Therefore, the main goal of this paper is to extend the research on investor herds to
the market for ADRs. To our best knowledge, this is the initial study testing herd behavior in
the market for foreign stocks traded in the U.S. market.
Several studies including Christie and Huang (1995), Wermers (1999) and Chang et
al. (2000) have examined herding behavior in the U.S. market; however, these studies have
only focused on securities issued by U.S. firms for which information is easily accessible
relative to those issued by foreign firms. Studying herding behavior in the market for ADRs
is different from prior studies on herding that focus on securities traded in a single market
and also interesting for several reasons. First, unlike domestic securities traded in the U.S.,
ADR returns are affected by not only the risk factors specific to the U.S. market where the
ADR is traded, but also potentially driven by additional uncertainties related to exchange rate
movements as well as the developments in the home market where the ADR is based on. One
can argue that compared to investors focusing on domestic securities only, investors in ADRs
are exposed to a wider array of risk factors which may create additional uncertainty, thus
potentially leading to a greater tendency to suppress their own beliefs and act as a herd, in
particular during periods of market stress. Second, focusing on ADR returns allows us to
examine if herding behavior is more prevalent among investors in ADR issues from
particular countries, providing us with clues on what might be driving investors towards such
behavior. Such an analysis could provide valuable insight to fund managers focusing on
country specific portfolios as evidence for herding in a particular market would suggest
greater challenges for diversification due to correlated actions of market participants. Third,
3
in addition to country-based ADRs, we also analyze ADRs within sector-based portfolios in
order to investigate possible sector effects on herding behavior. For example, ADR investors
might be interested in foreign firms in particular industries, say telecommunication,
regardless of the country of origin. This may be an important source of herding behavior as
one would have a more homogeneous group of ADR investors not only facing similar
uncertainties specific to those industries but also facing a significant disadvantage regarding
access to information relative to domestic investors in the corresponding home markets.
Finally, herding behavior among ADR investors may be influenced by a number of factors
including the risk factors in the underlying stock market (in our case, the U.S. market), as
well as uncertainties in the home stock market and the currency market. It is therefore
possible to test whether it is the market stress in the underlying market, in the country of
origin or shocks in the currency market which drives such behavior among investors of that
country’s ADRs.
Overall, this paper contributes to the literature by examining herding behavior in the
market for ADRs which, to our best knowledge, has not been studied in the literature. Using
daily price data on 305 ADRs traded in U.S. exchanges issued by corporations from 19
countries, we examine the cross-sectional ADR behavior with respect to movements in the
U.S. market index and test for possible herding behavior across country-based as well as
sector-based ADR portfolios. In a recent study, Chiang and Zheng (2010) use daily sector
returns from 18 advanced and emerging markets and find evidence on herding in some Asian
and advanced markets (except for the U.S.). Our study examines herding behavior from a
different angle by focusing on the securities of foreign firms from a wide range of advanced
and emerging markets.
Our tests of country-based ADR portfolios indicate evidence of herding behavior in
the ADRs from Chile, Korea and the U.K. The robustness tests suggest that the results are
robust to alternative model specifications and that shocks in the currency market and the
market of origin have no significant impact on herding behavior in the ADR market. Overall,
we find that herding behavior is more prevalent for ADRs from Chile and the U.K whereas,
we observe asymmetry in herding behavior in the case of Korea where such behavior occurs
during periods of large market losses only. We also run similar tests after classifying ADRs
into different sectors based on the North American Industry Classification System (NAICS)
codes and test whether herd behavior exists within sector-based ADR groups. We find
4
evidence consistent with herding behavior in Basic Industries, Capital Goods, Food &
Tobacco, and Textile & Trade during periods of large market downturns only. The finding of
herding behavior during large negative market shocks is consistent with several studies
including Kahneman and Tversky (1979) and Kahneman et al. (1990) which suggest that
investors are more concerned with potential losses than gains, further suggesting
asymmetries in utility functions or assessment of risk by investors.
This paper is organized as follows. Section 2 provides a brief review of the literature
on ADRs and tests of herding behavior. Section 3 provides the methodological details.
Section 4 presents empirical results on country and sector based ADR portfolios. Finally,
Section 5 concludes the paper.
2. Related literature review
ADRs have been studied extensively from different angles in the literature. A major
research area has been to identify the explanatory variables for the price premium on ADRs
relative to the corresponding securities traded in their home markets [e.g. Kadiyala and
Kadiyala (2004), Grossman et al. (2007), Arquette et al. (2008), Chan et al. (2008)]. Another
line of research has focused on the different risk factors, including those related to the
underlying securities at the home market, exchange rate movements, home market index,
driving ADR returns [Kim et al. (2000), Patro (2000), Bin and Morris (2003), Kutan and
Zhou (2006)]. Other studies examine why investors would prefer ADRs over other
investments [Alaganar and Bhar (2001), Arnold et al. (2004), Aggarwal et al. (2007)]. Given
the findings in these studies, one may conclude that ADR investors behave differently from
investors who focus only on domestic securities. Although a number of studies in the
literature have proposed rational or irrational explanations to herding behavior, most of them
have been explained in the context of domestic markets and therefore it will be interesting to
see whether such behavior applies to ADR investors who may be exposed to a wider variety
of uncertainties.1
Tests to detect herding behavior have been applied to a number of advanced and
emerging markets. Some of the prior studies provide support for rational asset pricing models
whereas a number of studies find significant evidence of herding behavior. Christie and
Huang (1995) study U.S. stock returns classified into sectors and find no evidence of herding
1 For further discussion, see Levy (2004), Devenow and Welch (1996), and Scharfstein and Stein (1990).
5
in the U.S. market. Wermers (1999) uses trade data to examine the mutual fund industry in
the U.S. and finds that herd behavior exists in small cap and growth oriented funds. Chang et
al. (2000) use firm level data to examine stock returns in the U.S., Hong Kong, Japan, South
Korea, and Taiwan and find significant evidence of herding in Taiwan and South Korea.
Demirer and Kutan (2006) investigate Chinese stock returns and report no herding behavior.
However, in a later study, Tan et al. (2008) examine Chinese A and B shares separately and
find evidence of herding in this market. In another study on Asian markets, Demirer et al.
(2010) study sector-based portfolios and find evidence of herding in the Taiwanese stock
market. Chiang and Zheng (2010) offer a more comprehensive study of investor herds and
examine sector returns in a number of advanced and emerging markets. Their tests yield
evidence of herding in Asian and advanced markets (except for the U.S. market). Using a
different testing methodology, Carpenter (2007) finds evidence of herding propensity among
non-bank financial institutions (NBFIs) in the Australian foreign exchange market. Uchida
and Nakagawa (2007) report similar findings when they examine the Japanese loan market
where they find irrational herding among Japanese banks.
As noted earlier, however, the tests on herding have not yet been extended to
securities of firms that are traded in foreign markets, in this case the U.S. market. As
suggested in prior studies in the ADR literature, ADR returns should be explained by a wider
array of risk factors, including international interest rates, exchange rate movements, risk
premiums in the home market, as well as the risk premiums in the market where these
securities are traded. Therefore, challenged with greater uncertainties, investors in the ADR
market might have a greater tendency to follow each others’ trades or the market consensus.
We contribute to this literature by providing the very first evidence from the market for
ADRs. In the next section we explain our methodology.
3. Methodological Considerations
Two broad groups of methodologies have been proposed in the literature to test the
existence of investor herds. In the first group, herding behavior is measured using
trading/holding data (e.g. buy/sell order activity) whereas the methodologies in the second
group are based on herding measures that are estimated using price data. Within the first
group of herding tests, Lakonishok, Shleifer and Vishny (1992) propose a methodology
where they use an adjusted ratio of net buyers in a security over the sum of net buyers and
6
net sellers and examine the probability distribution of this ratio in order to make inferences
on herding.2 Methodologies in the second group utilize price data and examine cross
sectional behavior of returns within a group of securities to test for the existence of investor
herds. In this paper, we employ the testing methodology originally proposed by Chang et al.
(2000) and applied in a number of studies to developed and emerging stock markets.3 This
methodology uses the conditional form of CAPM and focuses on the relation between return
dispersion and market return in order to come up with inferences on herding in a market. For
this purpose, they first define return dispersion as the cross sectional absolute deviation
(CSAD) of security returns within a portfolio. Let tir , be the return on ADR i on day t. The
return dispersion is then formulated as
||1
,
1
, tm
n
i
tit rrN
CSAD
(1)
where tmr , is the return on the equally-weighted ADR portfolio and on day t, and N is the
number of securities in the ADR portfolio. Following the conditional CAPM specification,
Chang et al. (2000) derive a positive and linear relation between return dispersion and the
market return. Considering a portfolio of stocks with different sensitivities to the market risk
factor, one expects the cross-sectional variation in stock betas to lead to a greater variation in
stock returns as each stock would differ in their sensitivity to the market. Therefore, from a
rational asset pricing framework, one would expect a positive and linear relation between
dispersion and the absolute value of the market return. However, Chang et al. (2000) argue
that, in a market where herding behavior exists, the positive and linear relation would no
longer hold. Therefore, they propose the following model which is based on a general
quadratic relationship between return dispersion and the market return
ttmtmt rrCSAD 2
,2,1 || (2)
where tmr , is the equally weighted average of the ADR returns in the portfolio on day
t. In this specification, rational assets pricing models predict a significantly positive value for
α1 (due to the cross-sectional variation in security betas) and an insignificant value for the
coefficient of the non-linear term (α2). However, considering a market where herding
2 Several applications of this methodology include Wermers (1999), Bow and Domuta (2004), Carpenter and
Wang (2007), Uchida and Nakagawa (2007), and more recently, Lin (2008). 3 See Chang, Cheng, and Khorana (2000), Lin and Swanson (2003), Gleason, et al. (2004), Demirer and Kutan
(2006), Tan et al. (2008), and more recently Demirer et al. (2010), and Chiang and Zheng (2010).
7
behavior exists, investors’ correlated actions of moving in and out of markets as a herd would
lead to greater directional similarity in stock returns across the portfolio, leading to lower
return dispersions. Therefore, a significant and negative estimate for α2 is used as support for
the presence of herding behavior. In this study, we follow an alternative specification
originally suggested by Duffee (2000) and estimate
ttmtmtmt rrrCSAD 2
,3,2,10 (3)
where is the equally weighted realized return of all ADRs in the portfolio on day
t. In this alternative specification, (2+1) and (2-1) capture the relation between the
CSAD term and the market return for positive and negative realizations of rm,t, respectively.
Similarly, a significant and negative value for the non-linear term (3) will be consistent with
herding behavior.
In order to check the robustness of the findings, we perform two additional tests.
First, we include additional variables in the model in order to control for the effects of the
market of origin where the ADR is based on and the currency rates. This is based on the
literature on ADRs suggesting that ADR returns can be explained by a wider array of risk
factors, including risk premiums in the home market, risk premiums in the market where
these securities are traded as well as uncertainties regarding exchange rate movements (e.g.
Bin and Morris, 2003, Kutan and Zhou, 2006). This suggests that herding behavior among
ADR investors can be driven by additional risk factors including the stock market
performance in the home country as well as shocks in the currency market. For this purpose,
we perform robustness tests by estimating an augmented model specified as
ttFXtHtmtmtUSt rrrrrCSAD 2
,5
2
,4
2
,3,2,10 (4)
where tHr , and tFXr , are the return on the home market index and the exchange rate on
day t, respectively. This specification allows us to control for effects of the home market
index and exchange rate on the non-linear relation between return dispersion and market
return beyond what is explained by the market factor in the conditional CAPM specification.
A similar specification is used by Chiang et al. (2010) in order to examine the effect of the
U.S. market on herding behavior in global markets.
Following a number of studies reporting asymmetries in equity returns and return
dispersions with respect to market conditions (e.g. Duffee, 2000, Longin and Solnik, 2000,
Ang and Chen, 2002, among others), we next perform additional tests by conditioning our
8
estimations on market returns. For this purpose, following the applications of asymmetries to
herding tests in Tan et al. (2008) and Chiang and Zheng (2010), we examine the effect of
market gains and losses on herding behavior by dividing the data into two groups using a
market dummy variable and modify Equation (3) as
ttmtmtmtmtmtmt rrrrCSAD 2
,,4
2
,,3,2,10 )1( (5)
where tm, is a dummy variable that equals one if the market return on day t is
negative and zero otherwise. This specification allows us to test whether herding behavior is
more prevalent during periods of market gains or losses. Next, we present the empirical
findings.
4. Data and Empirical Results
4.1. Data
The data set used in this study contains daily prices for on 305 ADRs issued by
corporations from 19 countries which have at least five ADR issues traded in the U.S. Daily
ADR data and stock market index data are obtained from CRSP and Datastream,
respectively. The data period analyzed is from January 1995 to January 2011, totaling 4,030
listing days for most ADR issues. As suggested by Bikhchandani and Sharma (2000), one
would be more likely to observe herding behavior within sufficiently homogeneous groups of
investors in which they face similar uncertainties and decision problems. For this purpose,
we classify ADRs into two categories: based on the country of origin and industry
classification. Table 1 provides the summary statistics for the cross-sectional absolute
deviations (CSAD) of ADR returns across ADR portfolios as formulated in Equation (1).
Panels A and B in the table report the summary statistics for country and sector based ADR
portfolios, respectively. Note that the number of ADRs for a given country or sector
classification changes over time. The average number of ADRs for each classification over
the sample period is listed in the second column of the table.
Examining country-based ADR portfolios in Panel A, we observe that ADR issues
from India and Ireland have the highest mean level of return dispersion, suggesting higher
variability across ADR returns for these countries. This may also suggest that returns on
ADR issues from these countries had unusual cross-sectional variations due to unexpected
news or shocks, either in their markets or in the U.S. market that they are listed on. On the
9
other hand, the lowest level of cross-sectional dispersion is observed in Chilean ADRs,
suggesting relatively greater directional similarity in the ADR issues from this country.
Following the suggestion by Bikhchandani and Sharma (2000) that it is more likely to
observe herding behavior within homogeneous groups of investors, we also classify ADRs
into sector-based portfolios based on their NAICS classification regardless of their country of
origins. The rationale behind this classification is that ADR investors, in particular emerging
market investors, might be especially interested in foreign firms in specific industries
regardless of the country of origin and this might be another source of herding behavior as
this would lead to homogeneous investor groups facing similar uncertainties specific to those
industries. For this purpose, we assign each ADR to one of fifteen industries including
Agriculture & Forestry, Basic Industries, Capital Goods, Construction, Consumer Durables,