10 1016@j Ribaf 2017 07 124
10 1016@j Ribaf 2017 07 124
10 1016@j Ribaf 2017 07 124
PII: S0275-5319(17)30327-6
DOI: http://dx.doi.org/doi:10.1016/j.ribaf.2017.07.124
Reference: RIBAF 814
Please cite this article as: Peltomäki, Jarkko, Graham, Michael, Hasselgren,
Anton, Investor Attention to Market Categories and Market Volatility: The
Case of Emerging Markets.Research in International Business and Finance
http://dx.doi.org/10.1016/j.ribaf.2017.07.124
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Investor Attention to Market Categories and Market Volatility: The Case of
Emerging Markets
Jarkko Peltomäki*
E-mail: jape@sbs.su.se
Michael Graham
E-mail: mgr@sbs.su.se
Anton Hasselgren
E-mail: anton.hasselgren@sbs.su.se
Abstract
This paper examines the impact of investor attention on stock market and FX market volatility in
emerging economies using newly constructed innovative attention proxies that capture the full
spectrum of the dynamics of the information processing stages. Our results show that the new practical
proxies are better at capturing the complex nature of investor attention to market categories. We find
that investor attention explains stock market volatility and shocks to attention but not FX market
volatility in emerging markets. Thus, the emerging stock market, an important segment of the global
equity market, is particularly sensitive to changes to investor attention.
JEL Classification: 7; 12
1
1. Introduction
In the immediate aftermath of the Brexit vote in the UK in June 2016, various news
outlets reported that the most searched words and phrases by individuals in the UK,
according to Google Trends, were based on terms related to the vote. The reports
individuals’ attention to specific events, thus, validating the use of the Google-based
search volume index (SVI)to investigate some financial market phenomena, e.g.,
cannot be certain that trading decisions are made on the basis of information gathered
In this paper, we estimate three (3) practical innovations of the investor attention
variable that incorporate investors’ trading decisions and apply them to equity and
(EMs). These innovations are as follows: First, based on the Multiple Resource
Theory (Wickens, 1992), we estimate a refined proxy for attention from the SVI and
abnormal trading volumes (ATV) in the market. We do this by taking the first
principal component of the SVI and ATV. Thus, we reinforce an obvious online query
that information from an internet search is used when making trading decisions. This
maximizes the amount of variation captured from both variables and accentuates the
SVI variable most likely to affect EM-specific volatility. To this effect, we take the
first principal component of the SVI and excess ATV, the difference between abnormal
2
trading volume in EMs and the United States. In additionally, we adapt a method by
Bessembinder and Seguin (1993) and split the SVI and the new PCA-based proxy for
investor attention into expected and unexpected attention. This exercise informs on
the component of attention that most likely impacts on price behavior. We expect the
SVI proxy.
An inherent assumption in asset pricing models is that financial markets extract and
cannot be reflected in prices until agents pay attention to and act on the information
(see e.g., Peng and Xiong, 2006; Huberman and Regev, 2001). The Multiple Resource
Theory suggests that investor attention to specific events can be broken down into
three stages: Perception, processing, and action (see Wickens, 1992). As noted above,
attention to a financial asset does not automatically translate into trading decisions,
action (see Vozlyublennaia, 2014). We attempt to capture and apply the full spectrum
for attention in the extant literature have, to date, mainly focused on the individual
stages of investor attention. For example, the SVI captures perception and processing
whilst the ATV focuses on action, with no established links between the two.
We apply our practical innovations of the investor attention to the, emerging markets
implying investors with limited time and effort utilize only a limited set of
information (Grossman and Stiglitz, 1980; Barber and Odean, 2008). Implicitly, to
3
efficiently allocate scarce attention, investors categorize assets or markets at the initial
step of the portfolio allocation decision (see Bernstein 1995; Swensen, 2000). This
large volume of information (Mullainathan and Thaler, 2001; Barberis and Shleifer,
2003). Nevertheless, the literature that looks into the important role investor attention
plays in determining stock price volatility is silent on the impact of attention on the
volatility of returns in important market categories (see e.g. Vlastakis and Markellos,
2012; Smith, 2012; Aouadi et al., 2013; Andrei and Hasler, 2015).
While the emerging market category encompasses a wide variety of countries with
literature has, for instance, established that the EM’s risk profile is,’ inherently,’
different from developed markets (Bekaert and Harvey, 1997; Aggarwal et al., 1999),
and is a segregated part of the global stock market (e.g., Graham et al., 2016). Thus
hypothesis that investor attention explains stock market volatility and apply it
specifically to the emerging market category using our newly developed measures of
attention to EM economies.
1
This market category is momentous and constitutes 10 of the 20 largest economies of the world (See
http://www.dailyreckoning.com.au/emerging-markets-are-still-a-buy/2010/02/26/). In addition,
investments in these markets, primarily by mutual funds based in developed countries, have increased
exponentially since the 1990s (Kaminsky, et al., 2001, IMF, 2013).
2 Huang and Liu (2007), in a related study, show that rational inattention to important news may make
investors over- or underinvest. This impacts on the positive association between EM equity market
volatility and investor attention to EM economies.
4
In this paper, we also examine whether investor attention exerts any measurable
influences on volatility in the EM foreign exchange (FX) market. The focus on the
EM currency category has some merits as some EM currencies are currently included
important risk factor in EM equity returns (see Carrieri et al. 2006; Ladekarl and
Peters, 2013).4 In the market itself, recent reports in the financial press speculate on
countries.5 The Bank for International Settlements has also noted significant and deep
swings in capital flows to and from EMs, resulting in a reappraisal of their FX market
and has not considered the EM currency category (see e.g., Smith, 2012; Goddard et
Our results show that our practical innovations of the investor attention variable have
higher economic significance and improve the explanatory power of the variable in
3 See http://www.euromoney.com/Article/3260551/Emerging-market-currency-volatility-challenges-
multinationals-risk-management.html
4 The literature is not in agreement on common global currency market factors. This makes currency
volatility important in understanding the risk-return profile in currency markets (Nolte et al. 2015). Given
the differing risk-return profiles across market segments, a focus on market category provides an outlet
to understand unique features of the segment.
5
See, for example, http://www.ft.com/intl/cms/s/0/a82d9c14-2ce4-11e4-911b-
00144feabdc0.html#axzz3q2DpWLQ9. Accessed on September 01, 2015.
6
See http://www.bis.org/publ/bppdf/bispap73.pdf Accessed on September 01, 2015
5
the estimated models in contrast to the extant measures. In addition we report
evidence that EM attention exerts any measurable effects on market volatility. Even
though targeted investor attention may have predictive power in relation to individual
currency pairs’ volatility (see e.g., Smith, 2012; Goddard et al. 2015), this cannot be
attention that increases the potential for capturing investor information processing
that leads to trading decisions. Second, Peng and Xiong (2006) and Jame and Tong
(2014) suggest that investors have a proclivity towards processing market and sector-
the emerging market, as a distinct category, has relevance in explaining stock return
volatility in EM stock markets. This is interesting due to the proposed segregated role
of emerging markets (see Graham et al., 2016) and informs on how changes in global
The remainder of this study is organized as follows. Section 2 reviews the relevant
literature related to our study. Section 3 presents the data and methods employed in
6
the empirical analyses. Section 4 presents our empirical results and Section 5
concludes.
2. Literature Review
The theoretical literature on the implications of limited attention for asset pricing
greater risk aversion or longer investment horizons tend to update news less
frequently, but select more precise news updates (Huang and Liu, 2007). Moreover,
2006).
The empirical literature has examined the relationship between investor attention and
asset returns using different proxies for attention. Earlier papers took advantage of the
important role mass media outlets play in disseminating information to investors and
inferred indirect measures of attention such as headline news (Chan, 2003), media
coverage (Fang and Peress, 2009; Engelberg and Parsons, 2011), and advertising
expenses (Grullon et al., 2004). These proxies all show important effects on financial
variables. Another set of studies infer investor attention from abnormal trading
volumes (ATV). Using these proxies, Barber and Odean (2008) provide evidence that
investors’ propensity to buy stocks that catch their attention is greater than the
7
propensity to sell.7 Further, Hou et al. (2009), also inferring investor attention from
trading volume, show that price momentum profits (earnings momentum profits) are
higher among high volume stocks (low volume stocks). They also show variations in
investor attention in that price momentum profits (earnings momentum profits) are
Recognizing the growth and importance of the internet as an information hub for
proxies such Wikipedia updates (Rubin and Rubin, 2010) and blogposts (Hu et al.,
2003) have been used, the literature has lately converged around Google-based search
correlated to, but different from, other proxies for attention such as trading volume
and media coverage.8 Importantly, the SVI is shown to be positively associated with
future stock returns (Da et al. 2011). Vozlyublennaia (2014), the internet based proxy
for investor attention, further documents that an increase in attention improves market
efficiency and leads to a significant change in short term index returns. The paper also
notes a long term change in attention following a shock to index returns. Klemola et
al. (2016) also show strong evidence that internet search volumes regarding market
states include relevant information about investor attention and that they perform well
7
An additional proxy, whether the firm appeared in that day’s news (media attention), was also used in
their paper.
8
Da et al. (2011) suggest that this is a superior measure compared to trading volumes, as an example.
This is because, as institutional investors assumingly use a more sophisticated information environment,
the SVI captures retail investor attention as they are most likely to use Google as a source of information.
It should be noted that without proper definitions of differentiation between trader categories the Barber
and Odean (2008) measure of investor attention is different from the SVI in that it is likely to capture
attention from individual investors as well of that of mutual funds and institutional investors.
8
The SVI variable actively incorporates expressed interests by economic agents.
investors who source their information from the internet use it to make trading
decisions. For that reason we measure investor attention through a using approach.
We combine the proxies through PCA to extract information from several variables.
A strand of the literature on investor attention has focused on its relation to the
volatility of stock market returns. Vlastakis and Markellos (2012), Aouadi et al.
(2013), and Andrei and Hasler (2015) all show a positive association between an
examine this question in the context of the EM category and hypothesize that
volatility literature, Goddard et al. (2015) show two important results. First, attention
market volatility. In addition Carrieri et al. (2006) and Ladekarl and Peters (2013)
note that EM currency return is a significant risk factor of EM equity returns.9 Given
this explanatory role of currency returns for emerging equity markets, we relate our
markets.
9
Goddard et al. (2015) also document that attention is related to currency premium.
9
The data used in the empirical analyses are obtained from varied sources. The data on
the MSCI Emerging Market Index, the S&P 500 Index, and the MSCI Emerging
Market Currency Index10 used to estimate the market returns and return volatilities are
drawn from the Thomson Reuters DataStream. Two currencies are pegged in the
MSCI Emerging Market Currency Index; the Jordanian Dinar and the Chinese Yuan11.
However, following the de-pegging of the Yuan in 2005 and the move towards a
managed float, we see no reason for this to introduce significant bias into our
analysis. The data period is between 16 April 2004 and 12 December 2014.12
Returns are calculated as the first log difference of weekly price indexes. Returns on
the EM equity market and the US equity market are denoted EMRT and USRT,
respectively. Following Andrei and Hasler (2015) and Goddard et al. (2015), we
estimate equity and FX volatility by daily GARCH(1,1), and annualize the average
volatility for each week.14 The choice of using the weekly averages of volatility stems
from the Google data capturing weekly average search volumes. We also include
as the log of MSCI Emerging Market Index volatility, EMVOL, minus the S&P500
10 The MSCI Emerging Markets Currency Index consists of 25 currencies against the U.S. dollar. In July
2008 the main constituents were BRL (16.9%), MXN (4.9%), CLP (1.1%), RUB (10.7%), ZAR (6.6%),
ILS (2.3%), TRY (1.3%), PLN (1.6%),CNY (14.5%), KRW (13%), TWD (10.7%), INR (6.4%), MYR
(2.3%), IDR (1.5%), and THB (1.4%), and currencies with individual weights of less than 1% were PEN,
ARS, COP, HUF, EGP, CZK, MAD, JOD, PHP and PKR.
11
The Jordanian Dinar represents 0.1% of the index weight in 2008 and the Chinese Yuan 14.5%. In
2005 the Chinese Yuan was de-pegged from the dollar and allowed to vary around a gradually increasing
band against the USD. It was also moved to a managed float against a basket of international currencies.
12
The span of the data is limited to the availability of the Google Trends data (2004-01-09) and that of
the MSCI Emerging Market ETF (2003-04-14). When creating the variables, a 52 weekly mean is
extracted from the MSCI Emerging Market ETF price series, making it the first observation to be used
in the analysis 2004-04-14, or to specify at the end of the week, 2004-04-16.
13
Friday adjusted closing prices are used to the extent they are available. When the Friday prices are not
available, the closing prices from the first trading day prior to that are used.
14
For robustness, we also estimated volatility using EGARCH(1,1), which did not change the results of
this paper.
10
Index volatility, USVOL. MSCI Emerging Market Currency Index volatility is denoted
as EMFXVOL.
The Multiple Resource Theory, a theory of multiple task performance, asserts that
activities differ from those that underline the selection and execution of responses.
Wickens (1992) represents these stages as perception, processing, and action. From
this distinction, we can decipher three layers of investor attention. The first is the
visual or auditory perception of a subject matter, e.g., EMs based on a news item, a
headline, or a discussion (see e.g., Barber and Odean, 2008; and Yuan, 2015). Da et
al. (2011), however, suggest that proxies for attention based on headlines do not
actively incorporate expressed investor interest. This interest may be expressed when
the investor uses an internet search query using key words to gather more information
for processing (the second layer). This is the basis of the Google-based search volume
index (SVI) proxy for investor attention used in the finance literature (see e.g.,
Mondria et al., 2010; Joseph et al., 2011; Da et al., 2011, 2015; Dzielinski 2012;
Vlastakis and Markellos 2012; Ding and Hou, 2015). Vozlyublennaia (2014),
nevertheless, argues that the measure of investor attention based on an internet search
query negate the possibility that the agents who search for the information do not
utilize it in making trading decisions (action, the third stage). The action stage may be
captured by the attention proxy inferred from abnormal trading volumes (ATV) (see
11
In this paper we include 4 proxies for attention in our empirical analyses. We use the
U.S Google-based Search Volume Index (SVI), employing the search word emerging
markets to represent attention to the EMs category.15 This search term typically
relates to information of a financial nature, which minimizes the level of noise in the
search word.16 The Google data is normalized as the probability of a search on the
particular search word for a specific region and time and measures the probability of a
search during an entire week. We label this variable the Emerging Market Search
Volume Index (EMSVI) and, in line with Da et al. (2011), interpret it as capturing the
attention of retail investors. This measures the overall interest in the asset category
displayed by potential and actual retail investors. Applying the multiple resource
theory, Wickens (1992) separates the information processing stages into three which
on the expected and unexpected trading volume’s effects on volatility. We adapt this
approach to further split the EMSVI variable into expected and unexpected
attention not anticipated by the market. We model EMSVI with an AR(1) process as:
15
We also used a global Google-based Search Volume Index with the same search phrase, which yielded
qualitatively similar results.
16
For instance, if instead we were to use specific country names, the chances are that the majority of the
attention attracted by the proxy would consist of individuals seeking non-financial information, such as
population, politics, etc. On the other hand, if we were to use the actual index names, such as “JSE” for
the Johannesburg Stock Exchange, it would be likely that a representative global investor would not
search for this particular name. Thus, this search word would attract attention from within, or
geographically close to, the particular category.
12
𝐸𝑀𝑆𝑉𝐼𝑡 = 𝜌𝐸𝑀𝑆𝑉𝐼𝑡−1 + 𝜃𝑡 . (1)
where 𝐸𝑀𝑆𝑉𝐼𝑡 is the actual EMSVI value and 𝜃𝑡 is the residual from the AR(1)
𝑈𝑁𝐸𝑋𝑃𝐸𝑀𝑆𝑉𝐼𝑡 = 𝜃𝑡 . (3)
Following Barber and Odean (2008) we also include the abnormal trading volume
(ATV), defined as the ratio of the daily trading volume over the yearly average, as an
alternative proxy for attention in our empirical analyses. From Wickens’ (1992)
information processing stages, this investor attention variable should capture the last
stage, action, in the information processing procedure, where investors have paid
attention complements the SVI. We adapt the ATV definition to weekly basis to match
the SVI variable and calculate it as the weekly trading volume of the MSCI Emerging
ETF and the SPDR S&P 500 ETF over the yearly average:
𝑊𝑒𝑒𝑘𝑙𝑦 𝑉𝑜𝑙𝑢𝑚𝑒𝑡
𝐸𝑀𝐴𝑇𝑉𝑡 = ∑𝑡−1
, (4)
𝑖=𝑡−52 𝑊𝑒𝑒𝑘𝑙𝑦 𝑉𝑜𝑙𝑢𝑚𝑒𝑖 /52
13
Where 𝐸𝑀𝐴𝑇𝑉𝑡 is the EM abnormal trading volume and 𝑊𝑒𝑒𝑘𝑙𝑦 𝑉𝑜𝑙𝑢𝑚𝑒𝑡 is the sum
of each week’s daily trading volume. We state Equation (4) so that the ATV is the
weekly trading volume over the yearly average weekly volume up until, but not
emerging markets and the US, respectively. The trading volumes in emerging markets
and the US are represented by the respective trading volumes of the MSCI Emerging
Market Exchange Traded Fund and the Standard and Poor Depositary Receipt
Exchange Traded Fund, for which the data accessed from Yahoo! Finance.
explicitly state that the agents who search for the information act on it to make trading
decisions (Vozlyublennaia 2014). In this paper we also utilize a refined measure for
investor attention, which increases the probability that internet search results are
incorporated into investor decisions, thus taking into consideration the final layer in
combined measure from the EMSVI and EMATV variables to capture a wider
spectrum of investor attention. This was done by taking the first principal component
of the EMSVI and EMATV variables, which we label ENHASVI. This is significant
given that the EMSVI and ATVEM contain information from different stages on the
information accumulation procedure (see e.g., Wickens 1992). The first principal
and processing information. As with the EMSVI variable, we separated the ENHASVI
variable into two parts, the expected and the unexpected components following the
14
Furthering the approach of extracting enhanced measures from combined attention
proxies, we also take the first principal component of the EMSVI and the excess
EMATV, i.e. the difference between EMATV and USATV, labeled ENHASVI2. In
US market. Thus ENHASVI2 disentangles the variation in the variables that are most
the US Federal Reserve Bank (FED) during the recent global financial crisis
substantially affected fund flows to EMs (IMF, 2013). To control for this effect, we
also gather search queries data from Google for the search words fed, monetary
policy, qe, central bank, boe, boj, and ecb. We sum up the standardized values
derived from the search and divide the result by the number of variables to create a
measure that captures attention towards global Central Banks actions (CBSVI). This is
∑𝑛
𝑖=1 𝑆𝑉𝐼𝑖,𝑡
𝐶𝐵𝑆𝑉𝐼𝑡 = , (5)
𝑛
where n equals 7 and 𝑆𝑉𝐼𝑖,𝑡 represents the values of the different search words for
week t. We use standardized values so that any one search phrase does not drive the
variation of the CBSVI variable. Our results do not change if we use non-standardized
values.
15
In addition the literature suggests that bad news increases volatility (see e.g., Fostel
and Geanakoplous, 2012). Thus we gather search volumes data on the search phrase
market crash to control for investors’ demand for negative information and label this
media attention during the last decades. Therefore we employ a control variable,
emerging market news, labeled EMdNEWS, in our study (see e.g. Vlastakis and
Markellos, 2012; Barber and Odean, 2008; Andrei and Hasler, 2013). We estimated
the EMdNEWS variable by counting the total number of news stories using the phrase
emerging market in the LexisNexis database and taking the first log difference of the
Furthermore we take into account possible spillovers from mature markets to EMs
and control for market sentiment (see e.g., proxied by the CBOE Put-Call ratio,
labeled PUT-CALL), the level of expected market volatility in the US, proxied by the
VIX index (VIX), and the level of US-based financial distress, proxied by the
Cleveland Financial Stress Index (CFSI). Data on the CBOE Put-Call and the VIX
index are also sourced from DataStream. The CFSI, sourced from the Federal Reserve
Bank of St. Louis (FRED), is divided into several types of stress component
originating from different financial markets in the US. In this study, we utilize the
stress index of stock markets and that of weighted dollar crashes, labeled
16
Descriptive statistics on the variables utilized in this study are presented in Table 1.
We note that the mean return for emerging markets for the stated sample period,
0.1%, is identical to that of the US market. However, consistent with the literature, we
detect greater stock market volatility in EMs (a mean value of 18.4%) relative to both
the US stock market volatility (16.8%) and FX emerging market volatility (11.5%).
The abnormal trading volumes are also shown to be higher for EMs. We also note that
the EMSVI variable has a mean of 41.77. Market sentiment over the sample period is
largely bullish, despite the inclusion of the period encompassing the financial crisis,
as shown by the mean of the PUT-CALL ratio of 0.652, indicating that more call than
Table 2 presents the pairwise correlations for all pairs of variables included in our
study. We note that the correlation between the return on the MSCI Emerging Market
Index and S&P500 Index is 0.754. Additionally all measures of volatility, apart from
excess volatility (VOLEXSS), exhibit high and positive correlations (over 0.8) with
each other. The VIX index is also highly correlated with EMVOL (0.855) and
EMFXVOL (0.793). Furthermore the VIX index shows a correlation of 0.933 with
USVOL. The attention variables ENHASVI and ENHASVI2 are highly correlated with
each other (0.857) and with the EMSVI attention proxy (0.829 and 0.786,
addition, the enhanced attention variables both show correlations in excess of 0.5 with
17
relatively high correlation with abnormal trading volume in the US, USATV, (0.574)
compared to that between ENHASVI2 and USATV (0.080). VIX and the equity
volatility measures are negatively correlated with all the variables measuring returns.
We estimate equation (6) to examine the relation between investor attention to EMs
and volatility for emerging equity and currency markets. Previous studies presume
that changes in investor attention should have an impact on price volatility (see e.g.,
Vlastakis and Markellos, 2012; and Aouadi et al., 2013). So far, this proposition has
global financial markets. We examine this hypothesis for the emerging market
where 𝑉𝑂𝐿 is the natural logarithm of a volatility variable: (1) EM stock volatility
(EMVOL), (2) excess volatility (VOLEXSS), and (3) EM currency index volatility
volume in EMs (EMATV), (2) the EM Search Volume Index (EMSVI), (3) the
18
component of the EMSVI and EMATV variables, and (4) the EM-specific attention
variable (ENHASVI2) obtained from the first principal component of the EMSVI and
the excess EMATV, i.e., the difference between EMATV and USATV. In Equation (6),
a positive and statistically significant value for the attention coefficient would support
our hypothesis that investor attention to the EM category has an impact on EM price
volatility. In relation to the studies by Peng and Xiong (2006) and Jame and Tong
(2014), it would suggest that investors' proclivity towards processing market and
information.
The control variables included in the model are CBSVI and CRASHSVI, attention to
the activities of central banks and negative news, respectively; EMdNEWS is the
volume of news on EMs; EMVOL and EMFXVOL are emerging market stock and
stock market and dollar stress, respectively; 𝑈𝑆𝑅𝑇 and 𝑆𝑃𝑅𝐷𝑅𝑇 are returns on the
equity market and the spread of return between EMs and the US equity market,
respectively; and PUT-CALL and VIX capture market stress and the level of risk in the
US, respectively.
4. Empirical Results
The results of examining whether investor attention to emerging market has an impact
on the volatility of EMs stock returns, equation (6), are presented in Table 3. Panel A
19
in Table 3 estimates 4 models using our different investor attention proxies. First we
detect from the models estimated, (1) to (4), that the coefficients of all the attention
proxies are positive and statistically significant at the 1% level. Thus a higher degree
investment category, confirming our hypothesis as well as results from the extant
studies (see e.g., Vlastakis and Markellos, 2012; Smith, 2012; Aouadi et al., 2013;
We expect that the newly constructed and enhanced investor attention variable would
index (SVI) proxy. Given that our dependent variables are log values, we examine the
economic significance of the attention proxies by estimating the ratio of the product
standard deviation of the dependent variable, i.e., the log of volatility. The results
indicate that a standard deviation increase in the EMATV and EMSVI leads to a 0.1153
volatility. The estimated economic impact is, however, generally greater for our
enhanced variables relative to the EMSVI. Confirming our expectation, we show that
ratios for the ENHASVI and EHANSVI2 are 0.1085 and 0.0794, respectively, which
implies that the signaling value of different measures of investor attention can be
Resource Theory.
20
Panel A in Table 3 also shows that lagged volatility, the VIX index, and the SPRDRT
On the other hand, the coefficient for the return on the S&P 500 Index is consistently
negative and statistically significant for all estimated models. The PUT-CALL and the
significance.
We re-estimate equation (6) using excess EM volatility, the difference between log
volatility over the US market and report the results in Panel B of Table 3. The
reported results indicate that excess EM volatility increases with investor attention to
EMs. This relation is robust as all proxies of attention show positive and statistically
increase in the attention proxies and the regression coefficient to the standard
deviation of excess volatility. The results show that a standard deviation increase in
our enhanced attention variables, ENHASVI and EHANSVI2, generates a 0.0764 and
volatility. We find that the estimated economic impact of the EMATV and EMSVI are
21
explaining EM currency volatility. This result is contrary to that reported by Smith
(2012) and Goddard et al. (2015) who find that investor attention has predictive
models excluding a variety of control variables, the VIX variable being the most
noteworthy. As the results were largely unchanged, we do not report the results of this
exercise.
Our results show that investor attention explains volatility. However, this effect may
examining whether investor attention explains volatility, with the expected and
attention (see Equations (2) and (3)). The unexpected component represents shocks to
attention that are not anticipated by the market. Table 4 presents the results of this
estimation for the EMSVI (Panel A) and ENHASVI (Panel B). Panel A shows that the
22
volatility, and also for the excess EM equity volatility, EMVOLEXSS. However, the
unexpected part is only significant for EM equity volatility (model 1). This result may
ex-post rational stock prices (see Shiller, 1981). In examining the economic
in the log annualized EM volatility. Regarding excess volatility, the estimated ratios
excess EM volatility. However, similar to the earlier results reported in Table 3, there
volatility.
In Panel B of Table 4 we can see that the expected component of the enhanced
attention variable, the EXPENHASVI, is positive and significant for both EM equity
volatility and excess EM equity volatility. The economic significance of this variable
on the outcome variable is 0.0672 and 0.0428, respectively. For the unexpected part of
the enhanced variable, the UNEXPENHASVI, we find that it is positive and significant
for EM equity volatility, as was the case with the UNEXPEMSVI. However, for the
attention variable is 0.0855 for EM equity volatility and 0.0603 for excess EM equity
enhanced variable consistently improves the measure compared to the Google search-
based EMSVI variable. Similarly, we find higher R-squared values for all models in
23
Panel B compared to those in Panel A, highlighting the potential of enhancing the
maintains its significance for a wider range of quantiles than the EMSVI, most
importantly for the 10th% and 90th% quantiles, in a quantile regression framework.
Furthermore, creating a dummy variable for negative returns in the MSCI Emerging
Market Index and interacting with the attention variables and re-estimating equation
(6) indicates that the effect of attention on volatility is generally more pronounced
5. Conclusion
This paper examines whether investor attention has explanatory power regarding
emerging market equity and currency volatility. Attention is a scare resource and
17 Both the coefficient for the attention variables and the coefficient for the interaction of the dummy
variable with the attention variables are positive and highly significant.
24
In this paper we estimate and apply three (3) attention innovations to examine the
search volume index (EMSVI) and abnormal trading volumes (EMATV) in the market.
First, we take the first principal component of the EMSVI and the EMATV to form a
new attention variable, based on Multiple Resource Theory. We argue that this new
attention variable is better in capturing the potential that retail investors who see
headlines and search for information (proxied by SVI) use that information in their
trading decisions (ATV). Second, we disentangle the variation in the SVI variable that
is most likely to affect EM-specific volatility by taking the first principal component
of SVI and excess ATV. Thus the enhanced investor attention variables potentially
index (SVI) proxy. Third, we split attention into an expected and an unexpected
component to inform on the component attention that most likely impacts on price
behavior. Our results indicate that the economic significance of the enhanced
attention variables is higher than the traditionally used Google-based SVI proxy.
We find that investor attention exerts positive and significant influence on stock
market volatility as well as excess volatility in EMs. This finding has three
implications. First, in line with Peng and Xiong (2006) and Jame and Tong (2014),
processing. Second, investor attention to the EM category of the global stock market
is a relevant determinant of stock market volatility in EMs. Third, the finding suggests
that EM equities have a segregated role in the global market, which is in line the
evidence in Graham et al. (2016). Our results also show that both expected and
unexpected (shocks to) attention are significant when explaining changes in excess
25
EM equity volatility. However, investor attention to EMs does not show any
information processing about the EM category does not affect FX similar to equities.
Overall, our results show that investors in EM equities should be aware that the
also make it evident that the measuring of investor attention in research can benefit
from adopting a multi-dimensional approach, taking into account several aspects and
levels of attention. It will be interesting to see how future research further develop this
approach.
26
Appendix A. Variable Descriptions
27
References
Aggarwal, R., Inclan, C., & Leal, R. (1999). Volatility in emerging stock markets.
Andrei, D., & Hasler, M. (2015). Investor attention and stock market volatility.
Aouadi, A., Arouri, M., & Teulon, F. (2013). Investor attention and stock market
Barber, B. M., & Odean, T. (2008). All that glitters: The effect of attention and news
Barberis, N., & Shleifer, A. (2003). Style investing. Journal of Financial Economics,
68, 161–199.
Bekaert, G., & Harvey, C. R. (1997). Emerging equity market volatility. Journal of
Bessembinder, H., & Seguin, P. J. (1993). Price volatility, trading volume, and market
Carrieri, F., Errunza, V., & Majerbi, B. (2006). Does emerging market exchange risk
affect global equity prices? Journal of Financial & Quantitative Analysis, 41,
511–540.
28
Chan, W. S. (2003). Stock price reaction to news and no-news: drift and reversal after
Da, Z., Engelberg, J., & Gao, P. (2011).In search of attention. The Journal of Finance,
66, 1461–1499.
Da, Z., Engelberg, J., & Gao, P. (2015). The sum of all FEARS: Investor sentiment
Ding, R., & Hou, W. (2015). Retail investor attention and stock liquidity. Journal of
Dzielinski, M. (2012). Measuring economic uncertainty and its impact on the stock
Engelberg, J. E., & Parsons, C. A. (2011). The causal impact of media in financial
Fang, L., & Peress, J. (2009). Media coverage and the cross-section of stock returns.
Fostel, A., & Geanakoplous, J. (2012). Why does bad news increase volatility and
Goddard, J., Kita, A., & Wang, Q. (2015). Investor attention and FX market volatility.
Graham, M., Peltomäki, J., & Piljak, V. (2016). Global economic activity as an
Grullon, G., Kanatas, G., & Weston, J. P. (2004). Advertising, breadth of ownership,
29
Hou, K., Xiong, W., & Peng, L. (2009). A tale of two anomalies: the implications of
investor attention for price and earnings momentum. Working Paper, Ohio State
University.
Huang, L., & Liu, H. (2007). Rational inattention and portfolio selection. The Journal
Huberman, G., & Regev, T. (2001). Contagious speculation and a cure for cancer: A
nonevent that made stock prices soar. The Journal of Finance, 56, 387–396.
Hu, N., Dong, Y., Liu, L., & Yao, L. J. (2013). Not all that glitters is gold: The effect
Jame, R., & Tong, Q. (2014). Industry-based style investing. Journal of Financial
Joseph, K., Wintoki, M. B., & Zhang, Z. (2011), Forecasting abnormal stock returns
and trading volume using investor sentiment: Evidence from online search.
Kaminsky, G., Lyons, R., & Schmukler, S. (2001). Mutual fund investment in
Klemola, A., Nikkinen, J., & Peltomäki, J. (2016). Changes in investors’ attention and
Ladekarl, J., & Peters, E. E. (2013). Emerging market currency: The common risk
30
Mondria, J., Wu, T., & Zhang, Y. (2010). The determinants of international
investment and attention allocation: Using internet search query data. Journal of
Nolte, Ingmar and Taylor, Mark P. and Xu, Qi, (2015). Dissecting Volatility Risks in
Peng, L., & Xiong, W. (2006). Investor attention, overconfidence and category
Rubin, A., & Rubin, E. (2010). Informed investors and the internet. Journal of
Smith, G. P. (2012). Google Internet search activity and volatility prediction in the
Swensen, D., 2000. Pioneering Portfolio Management. The Free Press, New York.
Vlastakis, N., & Markellos, R. N. (2012). Information demand and stock market
31
Yuan, Y. (2015). Market-wide attention, trading, and stock returns. Journal of
32
Table 1. Descriptive Statistics
This table contains descriptive statistics for all variables used in this study. EMRT and the USRT are end-of-the-
week returns for the MSCI Emerging Market Index and S&P500 Index, respectively. RTSPRD is the difference
between weekly returns on the MSCI Emerging Market Index and the S&P500 Index. Volatility of return for
emerging stock markets (EMVOL), the US stock market (USVOL), and emerging market foreign exchange index
(EMFXVOL) are weekly averages of annualized daily volatility estimated through a GARCH(1,1) based on the
daily returns of the ETFs mentioned above. The excess volatility (EMVOLEXSS) is the log difference between
MSCI Emerging Markets Index volatility, EMVOL, and the S&P500 Index volatility, USVOL. EMSVI, CBSVI,
and CRASHSVI denote investor attention to emerging markets, central banks, and market crash, respectively, and
are the average probability of a search on Google on a specific search phrase during a given week. EXPEMSVI
and UNEXPEMSVI represent expected and unexpected attention towards the emerging markets, respectively.
EMATV and USATV are abnormal trading volumes for the SPDR ETF and the MCSI Emerging Market ETF,
respectively. ENHASVI is the first principal component of EMSVI and EMATV. ENHASVI2 is the first principal
component of EMSVI and the difference between EMATV and USATV. EXPENHASVI and UNEXPENHASVI
represents an unexpected and expected components of the PCA variable. EMdNEWS is the first log difference of
news on emerging markets. VIX and CBOE PUT-CALL ratio proxy for market risk in the US and the US market
sentiment, respectively. STOCKSTRESS and USDSTRESS proxy for stress in the stock and foreign exchange
market (USD) originating from the US. The data is in weekly form and is between 2004-04-16 and 2014-12-12.
33
Table 2. Pearson Correlation
This table contains Pearson correlation values for all variables used in this study. EMRT and the USRT are end-of-the-week returns for the MSCI Emerging Market Index and the S&P500 Index, respectively.
RTSPRD is the difference between weekly returns on the MSCI Emerging Market Index and the S&P500 Index. Volatility of return for emerging stock markets (EMVOL), the US stock market (USVOL), and
emerging market foreign exchange index (EMFXVOL) are weekly averages of annualized daily volatility estimated through a GARCH(1,1) based on the daily returns of the ETFs mentioned above. The excess
volatility (VOLEXSS) is the log difference between MSCI Emerging Markets Index volatility, EMVOL, and the S&P500 Index volatility, USVOL. EMSVI, CBSVI, and CRASHSVI denote investor attention to
emerging markets, central banks, and market crash, respectively, and are the average probability of a search on Google on a specific search phrase during a given week. EXPEMSVI and UNEXPEMSVI represent
expected and unexpected attention towards the emerging markets, respectively. EMATV and USATV are abnormal trading volumes for the SPDR ETF and the MCSI Emerging Market ETF, respectively. ENHASVI
is the first principal component of EMSVI and EMATV. ENHASVI2 is the first principal component of EMSVI and the difference between EMATV and USATV. EXPENHASVI and UNEXPENHASVI represents an
unexpected and expected components of the PCA variable. EMdNEWS is the first log difference of news on emerging markets. VIX and CBOE PUT-CALL ratio proxy for market risk in the US and the US market
sentiment, respectively. STOCKSTRESS and USDSTRESS proxy for stress in the stock and foreign exchange market (USD) originating from the US. The data is in weekly form and is between 2004-04-16 and
2014-12-12.
UNEXPEMS STOCKSTRE EXPENH UNEXPE
VOLEXSS VOLEMFX VOLEM VOLUS RTUS RTEM RTSPRD EMSVI CBSVI CRASHSVI EXPEMSVI ATVEM ATVUS ENHASVI ENHASVI2 EMdNEWS VIX PUT-CALL USDSTRESS
VI SS ASVI NHASVI
VOLEXSS 1
EMFXVOL -0.148 1
34
Table 3. Investor attention and volatility
This table presents the results from estimating equation (7): 𝑉𝑂𝐿𝑡+1 = 𝛼 + 𝛽1 𝑉𝑂𝐿𝑡 + 𝛽2 𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛𝑡 +
𝛽3 𝐶𝐵𝑆𝑉𝐼𝑡 + 𝛽4 𝐶𝑅𝐴𝑆𝐻𝑆𝑉𝐼𝑡 +𝛽5 𝐸𝑀𝑑𝑁𝐸𝑊𝑆𝑡 + 𝛽6 𝑆𝑇𝑂𝐶𝐾𝑆𝑇𝑅𝐸𝑆𝑆𝑡 + 𝛽7 𝑈𝑆𝐷𝑆𝑇𝑅𝐸𝑆𝑆𝑡 + 𝛽8 𝑈𝑆𝑅𝑇𝑡 +
𝛽9 𝑆𝑃𝑅𝐷𝑅𝑇𝑡 + 𝛽10 𝑃𝑈𝑇 − 𝐶𝐴𝐿𝐿𝑡 + 𝛽11 𝑉𝐼𝑋𝑡 + 𝜖𝑡 , where 𝑉𝑂𝐿 is the natural logarithm of a volatility
variable: (1) EM stock volatility (EMVOL), (2) excess volatility (EMVOLEXSS), and (3) EM currency
index volatility (EMFXVOL); ATTENTION is an investor attention proxy: (1) abnormal trading volume
in EMs (EMATV), (2) the EM Search Volume Index (EMSVI), (3) the enhanced attention variable
(ENHASVI) estimated by taking the first principal component of the EMSVI and EMATV variables, and
(4) the EM-specific attention variable (ENHASVI2) obtained from the first principal component of the
EMSVI and the excess EMATV, i.e. the difference between EMATV and USATV. The control variables
are included in the model are CBSVI and CRASHSVI, attention to activities of central banks and
negative news, respectively; EMdNEWS is the volume of news on EM; 𝑆𝑇𝑂𝐶𝐾𝑆𝑇𝑅𝐸𝑆𝑆 and
𝑈𝑆𝐷𝑆𝑇𝑅𝐸𝑆𝑆 are US stock market and dollar stress, respectively; 𝑈𝑆𝑅𝑇 and 𝑆𝑃𝑅𝐷𝑅𝑇 are the returns
on the US equity market and the spread of return between EMs and the US equity market, respectively;
and PUT-CALL and VIX capture market stress and level of risk in the US, respectively. Finally, lagged
values of the volatility measure is included. Newey-West corrected standard errors are used in the
estimations. The data is in weekly form and is between 2004-04-16 and 2014-12-12. Standard errors
are in parentheses and ***, **, * indicate significance at 1%, 5% and 10% levels, respectively.
35
Panel B: Excess EM Equity Volatility
EMVOLEXSSt+1
EMATV 0.041***
(0.013)
EMSVI 0.001***
(0.0004)
ENHASVI 0.020***
(0.005)
ENHASVI2 0.027***
(0.007)
EMdNEWS 0.029 0.036 0.029 0.026
(0.026) (0.027) (0.027) (0.026)
PUT-CALL -0.054 0.012 -0.019 -0.008
(0.050) (0.054) (0.054) (0.050)
EMVOLEXSS 0.814*** 0.839*** 0.813*** 0.791***
(0.024) (0.021) (0.022) (0.024)
VIX -0.053** -0.054*** -0.053*** -0.048**
(0.022) (0.016) (0.018) (0.021)
CBSVI 0.012 0.015 0.014 0.012
(0.019) (0.020) (0.019) (0.019)
CRASHSVI -0.0005 0.0002 -0.0004 -0.0003
(0.001) (0.001) (0.001) (0.001)
STOCKSTRESS 0.0001 0.0001 0.00003 0.0003
(0.001) (0.001) (0.001) (0.001)
USDSTRESS 0.001 -0.001 -0.001 -0.001
(0.002) (0.001) (0.001) (0.002)
USRT -0.144 0.048 -0.094 -0.163
(0.200) (0.184) (0.189) (0.203)
RTSPRD 0.958** 1.124*** 1.045*** 1.085***
(0.200) (0.261) (0.226) (0.224)
CONSTANT 0.160*** 0.124*** 0.198*** 0.176***
(0.053) (0.042) (0.046) (0.053)
Observations 531 531 531 531
Adjusted R2 0.802 0.798 0.802 0.807
36
Panel C: EM FX Volatility
EMFXVOL+1
EMATV 0.013
(0.011)
EMSVI 0.0001
(0.0004)
ENHASVI 0.004
(0.005)
ENHASVI2 0.003
(0.004)
EMdNEWS 0.042 0.047 0.044 0.045
(0.037) (0.038) (0.037) (0.037)
PUT-CALL -0.085 -0.064 -0.070 -0.067
(0.054) (0.053) (0.052) (0.052)
EMFXVOL 0.853*** 0.858*** 0.856*** 0.856***
(0.027) (0.026) (0.026) (0.027)
VIX 0.130*** 0.123*** 0.127*** 0.127***
(0.034) (0.032) (0.033) (0.034)
CBSVI 0.011 0.012 0.011 0.011
(0.015) (0.015) (0.015) (0.015)
CRASHSVI 0.001 0.002 0.001 0.001
(0.001) (0.001) (0.001) (0.001)
STOCKSTRESS -0.001 -0.001 -0.001 -0.001
(0.001) (0.001) (0.001) (0.001)
USDSTRESS -0.003** -0.003* -0.003** -0.003**
(0.001) (0.002) (0.002) (0.002)
USRT -0.604* -0.594* -0.591* -0.597*
(0.314) (0.314) (0.313) (0.311)
RTSPRD 0.877** 0.932*** 0.915*** 0.926**
(0.339) (0.345) (0.341) (0.343)
CONSTANT -0.620*** -0.597*** -0.599*** -0.606***
(0.128) (0.126) (0.123) (0.127)
Observations 531 531 531 531
Adjusted R2 0.920 0.919 0.920 0.919
37
Table 4. Shocks to Attention
This table presents the results from estimating an extended version of equation (7): 𝑉𝑂𝐿𝑡+1 =
𝛼 + 𝛽1 𝑉𝑂𝐿𝑡 + 𝛽2 𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛𝑡 + 𝛽3 𝐶𝐵𝑆𝑉𝐼𝑡 + 𝛽4 𝐶𝑅𝐴𝑆𝐻𝑆𝑉𝐼𝑡 +𝛽5 𝐸𝑀𝑑𝑁𝐸𝑊𝑆𝑡 + 𝛽6 𝑆𝑇𝑂𝐶𝐾𝑆𝑇𝑅𝐸𝑆𝑆𝑡 +
𝛽7 𝑈𝑆𝐷𝑆𝑇𝑅𝐸𝑆𝑆𝑡 + 𝛽8 𝑈𝑆𝑅𝑇𝑡 + 𝛽9 𝑆𝑃𝑅𝐷𝑅𝑇𝑡 + 𝛽10 𝑃𝑈𝑇 − 𝐶𝐴𝐿𝐿𝑡 + 𝛽11 𝑉𝐼𝑋𝑡 + 𝜖𝑡 , where 𝑉𝑂𝐿 is the
natural logarithm of a volatility variable: (1) EM stock volatility (EMVOL), (2) excess volatility
(EMVOLEXSS), and (3) EM currency index volatility (EMFXVOL); ATTENTION is EXPEMSVI
(equation (2)) and UNEXPEMSVI (equation (3)), separating the expected and unexpected evolution of
the EMSVI through AR(1), respectively. The same procedure is applied to ENHASVI to achieve a
similar seperation, labeled EXPENHASVI and UNEXPENHASVI (included in Panel B). The control
variables are included in the model are CBSVI and CRASHSVI, attention to activities of central banks
and negative news, respectively; EMdNEWS is the volume of news on EM; 𝑆𝑇𝑂𝐶𝐾𝑆𝑇𝑅𝐸𝑆𝑆 and
𝑈𝑆𝐷𝑆𝑇𝑅𝐸𝑆𝑆 are US stock market and dollar stress, respectively; 𝑈𝑆𝑅𝑇 and 𝑆𝑃𝑅𝐷𝑅𝑇 are the returns
on the US equity market and the spread of return between EMs and the US equity market, respectively;
and PUT-CALL and VIX capture market stress and level of risk in the US, respectively. Finally, lagged
values of the volatility measure is included. Newey-West corrected standard errors are used in the
estimations. The data is in weekly form and is between 2004-04-16 and 2014-12-12. Standard errors
are in parentheses and ***, **, * indicate significance at 1%, 5% and 10% levels, respectively.
38
Panel B: Shocks in ENHASVI
EMVOLt+1 EMVOLEXSSt+1 EMFXVOLt+1
EXPENHASVI 0.029*** 0.015** 0.001
(0.008) (0.007) (0.006)
UNEXPENHASVI 0.042*** 0.024*** 0.009
(0.006) (0.007) (0.006)
EMdNEWS 0.011 0.020 0.035
(0.027) (0.028) (0.040)
PUT-CALL 0.094* -0.016 -0.067
(0.050) (0.052) (0.052)
VIX 0.175*** -0.053*** 0.126***
(0.028) (0.017) (0.033)
CBSVI 0.009 0.014 0.010
(0.020) (0.019) (0.015)
CRASHSVI 0.001 -0.0003 0.001
(0.001) (0.001) (0.001)
STOCKSTRESS -0.0003 0.0001 -0.001
(0.001) (0.001) (0.001)
USDSTRESS -0.003* -0.0004 -0.003**
(0.002) (0.001) (0.002)
USRT -0.585*** -0.063 -0.570*
(0.234) (0.183) (0.310)
RTSPRD 1.028*** 1.034*** 0.899***
(0.220) (0.224) (0.339)
EMVOL 0.780***
(0.025)
EMEXSSVOL 0.820***
(0.023)
EMFXVOL 0.858***
(0.026)
CONSTANT -0.948*** 0.192*** -0.596***
(0.109) (0.043) (0.122)
Observations 530 530 530
Adjusted R2 0.940 0.802 0.920
39