Upload
arohan-walia
View
221
Download
0
Embed Size (px)
Citation preview
8/7/2019 Currency Premium and Firm Characteristics
1/34
Is the Exchange Risk Premium in Stock Markets
Related to Firm Characteristics?
Hyunchul Chung and Basma Majerbi
*
January 2009
* Chung is at the School of Business, Hanyang University, 17 Haengdang-dong, Seongdong-gu, Seoul, 133-791,Korea. He may be reached at [email protected]. Majerbi is at the Faculty of Business, University of Victoria,PO Box 1700 STN CSC,Victoria, BC, V8W 2Y2, Canada. She may be reached at [email protected]. Majerbi wouldlike to acknowledge the generous financial support from the UVic Business Research Fund and from SSHRC.
8/7/2019 Currency Premium and Firm Characteristics
2/34
1
Is the Exchange Risk Premium in Stock Markets
Related to Firm Characteristics?
Abstract
This paper provides new evidence on the determinants of exchange risk premia in the stock
market using firm level data from Korea. We conduct empirical asset pricing tests using
cross-sectional data at firm level to determine whether exchange risk is priced under
alternative model specifications and to see whether the estimated exchange risk premium
can be related to firm characteristics. Our results support the hypothesis of a significant
unconditional exchange risk premium in the Korean stock market at the individual firm
level. We also find that the exchange risk premium is more significant, statistically and
economically, for firms with larger size and higher percentage of foreign ownership.
However, we find weak evidence suggesting that the exchange risk premium may be lower
for firms with higher trading liquidity.
8/7/2019 Currency Premium and Firm Characteristics
3/34
2
1. IntroductionThe question of whether foreign exchange risk is priced in the stock market has been the
subject of an ongoing debate. This goes back to early theoretical models such as Shapiro
(1974), Dumas (1978) and Choi (1986), which clearly identified the potential impacts ofexchange rate movements on the firm's expected cash flows and hence its value. In the
theoretical asset pricing literature, it has been shown since the seminal work by Adler and
Dumas (1983) that there should be a non-zero exchange risk premium in stock returns because
of deviations from purchasing power parity (PPP)1. Indeed, under PPP deviations, investors
may consider a foreign investment as riskier, perceiving exchange risk as a real currency risk
that is partly non-diversifiable, and hence requiring some compensation in terms of expected
returns when investing in foreign stock markets.
Although early unconditional tests provided mixed results on this question, the bulk of
the empirical evidence based on conditional asset pricing tests points towards the existence of
a significant price of risk related to the currency factor in stock markets of both developed and
emerging countries. These studies, however, only focused on testing whether the exchange
risk factor commands a significant risk premium in equity returns, mainly at the aggregate
market level. They do not try to explain what drives such risk premium, nor do they look at
cross-sectional differences in the estimated exchange risk premia to see if they can be related
to specific characteristics of the underlying assets at a more disaggregated level.
In this paper, we try to address these issues which go beyond the question of whether
exchange risk is a priced factor in explaining equity returns. In particular, given the firm-
specific nature of the foreign exchange exposure, we are interested in 1) estimating the
exchange risk premium at the firm level, and 2) investigating whether there is a link between
the significance and magnitude of the estimated exchange risk premium and firm
characteristics such as firm size, foreign ownership and trading volume (as a measure of
liquidity). The second question is particularly interesting given the considerable number ofprevious studies documenting that cross-sectional variations in stock returns can be related to
firm-specific characteristics, such as firm size, book-to-market value, trading volume, and
dividend yield ( for instance Fama and French, 1992; Chan et al., 1991; Brennan et al., 1998).
1 PPP deviations are well documented in the economic literature for both developed and emerging markets: Roll(1979), Abuaf and Jorion (1990), Salehizadeh and Taylor (1999), Li (1999).
8/7/2019 Currency Premium and Firm Characteristics
4/34
3
Thus, we try to investigate whether these characteristics can also explain the cross-sectional
variations in the exchange risk premium since the latter can be a significant component of the
total equity returns as demonstrated by previous studies.
Since most of the previous studies on the pricing of exchange risk rely on market-level
indices, the documented results represent an average effect of foreign exchange risk on asset
pricing. Therefore, it is important to use firm level data to investigate the above questions. In
this paper, we use firm data from South Korea, with all returns expressed in US dollar, thus
taking the view point of a US investor interested in the Korean stock market. Our focus is on
the potential link between the estimated exchange risk premium and firm characteristics.
Therefore, unlike studies that look at the determinants of exchange risk exposure, we are
interested in the exchange risk premium estimated in the context of an asset pricing model that
includes the exchange risk as a pricing factor2
. In this framework, it is difficult to use a multi-country dataset unless we do it at the aggregate market level as in previous studies. Therefore,
we focus on one country to be able to exploit data on individual firm returns and firm-specific
characteristics.
Like many other emerging markets, Korea had experienced important currency crises
with overwhelming negative impacts on its economy and stock market. After the Asian crisis
of late 1997, Korea switched to a free-floating exchange rate regime resulting in greater
uncertainty in the value of its currency with respect to major world currencies 3. Therefore,
Korea provides a good sample for the purpose of our study particularly given the availability
of individual firm characteristics data on all firms listed on the major Korean exchange.
In this paper, we test international asset pricing models with exchange risk factors using
firm level data to see if the estimated exchange risk premium is related to specific firm
characteristics such as size, liquidity and foreign ownership. Our main results can be
summarized as follows. In line with previous studies using market level data, our firm level
analysis supports the hypothesis of a significant exchange risk premium in the Korean stock
2 In the literature about exchange rate exposure (not pricing), we typically estimate a regression where exchange ratechanges represent one of the explanatory variables and estimate the corresponding betas. Examples of such studiesinclude Choi and Prasad (1995), Faff and Marshall (2005), Dominguez and Tesar (2006) among others. This isdifferent from the international asset pricing literature where the focus is on estimating and testing the significance ofthe price of risk attached to the exchange rate factor.3Since 1990, Korea was under a peg system (Market Average Exchange System (MARS), which allowed inter-bank spot rates tofluctuate within a certain band around each days weighted-average exchange rate. The original band of0.4 increased to 10 in
Nov. 1997. On Dec. 16, 1997, Korea abolished the band on the wons exchange rate against the US dollar and shiftedto a free-floating exchange rate system.
8/7/2019 Currency Premium and Firm Characteristics
5/34
4
market even at the unconditional level. We also find that the exchange risk premium is more
significant, both statistically and economically, for firms with larger size and higher foreign
ownership level. However, we find weak evidence suggesting that the exchange risk premium
may be lower for firms with higher liquidity. These results have important implications for
asset valuation for portfolio managers as well as the determination of the cost of capital for
Korean firms.
The rest of the paper is organized as follows. In section 2, we briefly discuss the
existing empirical literature on the pricing of exchange risk in the stock market and motivate
our approach. Section 3 outlines the empirical model and methodology and section 4
describes the data. The empirical results from tests of alternative asset pricing models are
presented in section 5. Finally, section 6 concludes the paper and suggests some guidelines for
future research.
2. Previous ResearchTheoretically, if PPP holds and if there are no barriers to international investments and
no differences in consumption goods, the single-index capital asset pricing model (CAPM)
should hold internationally and exchange risk should not be priced. Given the wide empirical
evidence against such a perfect world, some early theoretical studies considered the effects of
foreign exchange risk on asset returns and developed International Asset Pricing Models
(IAPM) that include exchange risk factors along with the traditional market risk factor [Solnik
(1974), Sercu (1980), Stulz (1981a), Adler and Dumas (1983)]. More recently, Chaieb and
Errunza (2007) showed that exchange risk should be priced in global equity markets under the
joint assumptions of PPP deviations and partial segmentation. In this framework, exchange
rate changes should yield a non-zero risk premium even in the presence of a broader local
market risk in addition to the world market factor.
On the empirical side, the evidence from testing such asset pricing models at the
unconditional level is mixed. Early tests, such as Hamao (1988) and Jorion (1991), wererather inconclusive and generally found no evidence that exchange risk is priced on the
Japanese and US stock markets. On the other hand, Vassalou (2000) finds that exchange risk,
along with foreign inflation risk, can explain part of the cross-sectional variation in equity
returns for ten developed countries. Carrieri and Majerbi (2006) also find that exchange risk is
significantly priced at the unconditional level using data on nine emerging markets regardless
8/7/2019 Currency Premium and Firm Characteristics
6/34
5
of whether we assume full integration or partial segmentation of international capital markets.
This result seems to hold both at the market level and at the more disaggregated industry and
portfolio levels.4
There is less controversy from empirical tests based on conditional asset pricing models
with most evidence strongly supporting the hypothesis that foreign exchange risk is priced in
the stock markets of major developed countries [see for example Dumas and Solnik (1995),
De Santis and Gerard (1998), Choi et al. (1998), Doukas, Hall and Lang (1999), Carrieri
(2001)]. More recent studies by Carrieri, Errunza and Majerbi (2006a, 2006b) confirm the
importance of currency risk premia in both developed and emerging stock markets under
alternative asset pricing specifications and exchange rate measures. Furthermore, based on
their more generalized international asset pricing model, Chaieb and Errunza (2007) provide
strong evidence on the significance of the price of global currency risk using data on eightmajor emerging markets.
These studies, however, do not try to explain what factors drive the estimated currency
risk premia and how do they vary across individual firms. Indeed, the evidence shown at the
market level only provides investors with an estimate of an average exchange risk premium
for a given country or market. It is not sufficient to inform investors about whether specific
types of firms will command different levels of risk premium related to the currency factor.
More importantly, we do not know if this exchange risk premium can be related to individual
firm characteristics. For instance, foreign portfolio investors are known to prefer investments
in large and well-known firms5. Thus, if all firms in a given market are aggregated within a
market index, the results will be averaged out in the sense that the firm specific exchange
risk premium could be underestimated for securities with high foreign demand and
overestimated for securities with low foreign demand or vice versa.6
4 In this study however, the portfolios and industry returns used are diversified across a number of emerging markets.The study also used firm level data on a country by country estimation, but the number of firms covered for each
country was very limited to allow for more general conclusions as explained by the authors.5 The reasoning behind this argument will be explained in detail in section 2.1. Kang and Stulz (1997) investigate theforeign equity ownership in Japanese firms using a disaggregated firm-level data and find that foreign investors holdmore shares of large-size firms, manufacturing firms, and firms with good accounting performance and low leverage.6
For example, though not directly related to FX premium, Christoffersen et al. (2006) illustrate the potential
average-out problem with market-level analysis. They investigate the revaluation effect after market liberalization inemerging markets using both market-level and firm-level data. They find significant cross-sectional differences inrevaluation effect depending on firm size used as a proxy for foreign investors demand. Market-level analysis showsthe revaluation effect of 31.8% points for eight-month-liberalization period whereas the firm-specific conditionalestimates show 11.8% points for the smallest firm and 37.4% points for the largest firm, respectively.
8/7/2019 Currency Premium and Firm Characteristics
7/34
6
Therefore, we investigate cross-sectional differences in the estimated exchange risk
premia by using disaggregated data that can better reflect foreign investors demand. In
particular, we are interested in three main characteristics of firms, namely firm size, foreign
ownership and trading volume, all of which seem to play a role in determining investors
demand for foreign assets as explained below.
2.1. Firm size and FX risk premium
The existence of a potential link between firm size and the exchange risk premium is an
empirical question to be tested with two contrasting views: While one can argue that large
firms, preferred by foreign investors, are more exposed to FX risks than small firms, one can
also argue that large firms are more likely to manage FX risks internally, thus reducing the
need for investors to be concerned with FX hedging.7
Many studies have looked at the impact of firm size in asset pricing, though not directly
including the effects of exchange risk. For instance, Christoffersen et al. (2006) found that the
impact of capital market liberalization differs depending on firm size used as a proxy for
foreign investments. The rationale for using firm size as a proxy for foreign investors demand
is based on the importance of information availability. For example, in the IAPMs of Black
(1974), Stulz (1981b), and Errunza and Losq (1985), the informational barrier can make cross-
border investments costly, or prohibit such investments in the limit. The home bias literature
emphasizes the importance of information asymmetry to explain the predominance of home
assets in international portfolios.8 In her survey of market experts and participants, Chuhan
(1994) reported limited information as one of the major impediments to investing in emerging
markets. On the other hand, firm size has been used in many studies as a proxy for information
richness and found to be a good indicator of information availability.9 Hence it is reasonable to
assume that foreign investors, who generally have limited information, prefer information-
rich, large firms to information-poor, small firms in international investment decisions,especially in emerging markets. 10 Since foreign portfolio investors are known to prefer
7 Nance et al. (1993) suggested that larger firms are more likely to hedge exchange rate risks.8 See for example, French and Poterba (1991), Cooper and Kaplanis (1994) and Lewis (1999).9 See Bailey and Jagtiani (1994), Kang and Stulz (1997) and Bailey, Chung and Kang (1999) among others.10 It is plausible that the cost of information on small-size firms is too high for foreign investors in relation to thepotential diversification benefit. Hence, foreign investors may not invest in small-size firms. These small-size firmsbecome non-traded in the vein of Stulz (1981b) who shows that there could exist non-traded assets that do not
8/7/2019 Currency Premium and Firm Characteristics
8/34
8/7/2019 Currency Premium and Firm Characteristics
9/34
8
countries, Dominguez and Tesar (2006) also find that a firm size dummy is statistically
significant for six of the eight countries, with the sign on the coefficients suggesting that large-
and medium-sized firms are likely to have lower levels of exposure than smaller firms.11
Therefore, in light of this conflicting evidence, the question of how the exchange risk
premium relates to firm size is an important empirical issue to be investigated since the impact
is not so obvious.
2.2.Liquidity and FX riskAn asset is considered liquid when it can be sold at fair market value in a timely fashion.
Since the work of Amihud and Mendelson (1986) liquidity has long been recognized as an
important factor that affects asset pricing (for example Chordia et al. ,2000; Hasbrouck and
Seppi, 2001; Pastor and Stambaugh, 2003; Acharya and Pedersen, 2005; Bekaert et al., 2007
among others).
As mentioned earlier, besides firm size, in her survey of market experts and participants,
Chuhan (1994) also reported lack of liquidity as one of the major impediments for foreign
institutional investors to investing in emerging markets. For this reason foreign portfolio
investors are more likely to prefer companies with higher liquidity and one can expect that
liquid firms with more foreign investors, who may be more sensitive to FX risks, will
command more significant FX risk premium. In the study by Muller and Verschoor (2006),
they found that FX exposure is significantly related to liquidity, in addition to firm size, and
the documented link suggests that firms with weak liquidity tend to show smaller exposure to
exchange risk.
Therefore this is another empirical question to answer. We can test whether liquidity
matters to explain differences in the estimated FX risk premium across firms along the same
lines as firm size. Assets liquidity is notoriously difficult to measure (see OHara, 2003 for a
discussion) and typically measured based on its spreads, depths and volumes. We use the
trading volume as our liquidity measure. Here we simply define the trading volume as the totalnumber of shares traded.
2.3.Foreign Ownership and FX premium
11 Note that this result was obtained after using a transformation of the estimated exchange beta coefficient (by takingthe square root of its absolute value) to focus only on the magnitude, regardless or the direction of the exposure.When the regression was done using the estimated beta, they found no relation between the exposure and firm size.
8/7/2019 Currency Premium and Firm Characteristics
10/34
9
Intuitively, we can think of FX risk premiums as additional expected excess returns
required by investors facing FX risks. Compared to domestic investors, foreign investors are
more likely to be concerned about the exchange risk effects on their portfolio holdings thus
requiring a compensation for taking on such additional risk.
There are a number of papers dealing with the impact of foreign investments (for example
Bekaert and Harvey, 2000; Henry, 2000a, 2000b; Kim and Singal, 2000; Errunza, 2001;
Bekaert et al., 2002; Patro and Wald, 2004; Chari and Henry, 2004 and Christoffersen et al.
2006). Their focuses lie on market stability (volatility), performance and economic growth,
etc. None of the previous studies seems to look at the relationship between foreign ownership
and FX risk premium, therefore it would be an interesting question to study. We expect that
stocks with larger foreign ownership would command larger exchange risk premia. This is
because more foreign stockholders are exposed to exchange rate fluctuations in their portfolioinvesting and they probably have more willingness to hedge them. While firm size has been
used as a proxy for foreign investment in some previous studies, we can apply directly real
foreign ownership structure data to investigate whether firms with larger foreign ownership
holdings command higher FX risk premium.
In summary, based on the above review, we use these three major firm characteristics
to investigate whether they have an impact on the magnitude of the exchange risk premium
estimated for Korean firms. We will investigate whether any FX premium differences exist
across firms and if such differences can be explained by specific characteristics such as size,
liquidity and foreign ownership structure.
3. Empirical Model and MethodologyFollowing Carrieri and Majerbi (2006), we start with a standard multi-beta pricing
model where we assume that expected asset excess returns are linear functions of factor risk
premia and their corresponding betas:
( ) ,...,T)t,N;,iFr itk
j
jtijiit 11(1
==++= =
L (1)
and =
=k
j
ijjitrE1
)( (2)
8/7/2019 Currency Premium and Firm Characteristics
11/34
10
where rit is the excess return on asset i at time t, measured in U.S. dollar (the reference
currency)12;ij are the assets sensitivities to the risk factors Fj (j = 1,,k); j are the expected
risk premia associated with the factors and itare random errors.13 The choice of factors in
such models is not formally dictated by a specific theoretical framework. Therefore, we
follow the empirical tradition and predetermine the risk factors to investigate based on the
implications of established asset pricing models. More specifically, we use the same
specifications as in Carrieri and Majerbi (2006) in which they estimate the exchange risk
premia using both two-factor and four-factor models as explained below.
The beta pricing model in (2) and (1) is estimated as a restricted seemingly unrelated
regression model (SURM)14:
=
++=k
j
itjjtijit fr1
)( (i=1,,N) (3)
wherefj are the de-meaned values of the risk factors (fjt =Fjt- mean(Fjt)), and E(it)=0.
This specification allows us to use economic variables, such as the change in the exchange
rate, as factors (see Ferson and Harvey (1994)).15
The parameters to be estimated using Eq. (3) are the unconditional betas (ij) and the expected
risk premia (j) for the models risk factors. We test two-factor and four-factor models, using
firm level data, as described below. Our objective is to determine the size and significance of
the coefficient related to the pricing of the exchange risk factor and to see if the estimatedexchange risk premium is related to individual firm characteristics such as firm size, foreign
ownership level and liquidity (trading volume).
12 The excess returns ritare computed as (Rit- Rft) whereRit refers to the gross return on asset i at time tandRftis theone-month US Treasury bill rate.13
Following Ferson and Harvey (1994), if we assume a general beta pricing model for expected returns Rit, such as
=
+=k
j
ijjit bRE1
0)( , where bij are the betas of theRit on the k risk factors, then this implies an expression for
the expected excess returns as in Eq. (2) above, whereij = bjj - bfj are the betas of the excess returns and bfj are thebetas of risk free rateRft.14 The regression is restricted by assuming that the intercept is equal to zero. This is the same procedure used byFerson and Harvey (1994) and Carrieri and Majerbi (2006).15 When factors are not de-meaned, they must all represent returns and the model estimation may require the use ofmimicking portfolios whose returns are substituted for the factors (see Ferson and Harvey (1995) for a more detailedexplanation).
8/7/2019 Currency Premium and Firm Characteristics
12/34
11
3.1 Two factor modelWe first test an unconditional version of model (3) assuming fully segmented markets, i.e.,
where the risk factors are the local market excess return (rm) and the change in an exchange rate
measure (s)16:
itissimmtismtiwit srr ++++= (4)
where rmt is excess return on the Korean market index (expressed in US dollar), st is the
change in the real exchange rate; im and is are the sensitivities of asset i to the market and
the exchange risk factors; m and s are the risk premia associated with the market and
exchange risk factors respectively.
This specification using local market risk instead of the world market risk as implied by
the original I-CAPM of Adler and Dumas (1983) is similar to the models proposed by Jorion(1991) for the US and Choi et al. (1998) for Japan where only a national market index is used
assuming segmented capital markets. However, we also test a version of Equation (4) by
substituting the world market index for the local market index as reported in Section 5.
As in previous studies, we use a single exchange rate measure, as a proxy for the
exchange risk factor.17 However, given previous evidence on the sensitivity of the results to
the choice of exchange rate measure, we use both a real bilateral exchange rate and a real
effective exchange rate index for Korea as alternative measures. As explained by Carrieri,
Errunza and Majerbi (2006a), the use of real exchange rates is more consistent with the
original model of Adler and Dumas (1983) particularly in countries where inflation can be
high and more volatile than what is typically observed in developed countries.18 Indeed, using
the change in the real exchange rate takes into account both a countrys inflation level as well
as the change in its nominal currency value. Moreover, using changes in the real exchange
rate helps overcome possible complications due to fixed exchange rate regimes or large
16 In estimating this two-factor model, as well as in the four-factor model that follows, all risk factors have been de-meaned as explained in the methodology section above.17 The original specification of Adler and Dumas (1983) includes all the currencies of the countries in the model. Forparsimony, previous studies have used aggregate proxies such as the trade-weighted exchange rate in Jorion (1991)and Choi et al. (1998) or the SDR in Choi and Rajan (1997).18 In Adler and Dumas (1983), excess returns should be related to their covariances with the foreign inflation ratesexpressed in the reference currency as a measure of PPP deviations. These terms have often been replaced inempirical testing by the changes in nominal exchange rates based on the assumption that inflation rates are non-random.
8/7/2019 Currency Premium and Firm Characteristics
13/34
12
discrete changes in nominal exchange rates due to devaluations or peg removals, such as the
case in Korea over our sample period.19
3.2 Four-factor modelSince the work by Errunza and Losq (1985) and later by Bekaert and Harvey (1995),
numerous studies had shown that capital markets are neither completely segmented nor fully
integrated due to the existence of various barriers to international investments and capital
flows. This issue becomes more relevant in the case of emerging markets where exposure to
local risk factors, in addition to currency risk, has been established both empirically and
theoretically in the international asset pricing literature (see Carrieri, Errunza and Majerbi
(2006b) and Chaieb and Errunza (2007)).
Such evidence motivates our choice to estimate the exchange risk premium for Korean
stock returns within the context of a partial integration model where both global and local
risks are priced. Therefore, we estimate the following four-factor model that includes global
market and global currency risks, in addition to the local market and local exchange risk
factors used in the two-factor model20:
it
e
is
e
siss
e
im
e
miww
e
t
e
istis
e
mt
e
imwtiwitssrrr ++++++++= (5)
In this model, emtr is the residual from a regression of the local market excess return on the
world market return rwtand it is used as a measure of local market risk that is not captured by
the common world market factor. Similarly, ets is measured by the residual from a regression
of the local currency real exchange rate on a global exchange rate index st. It is used as the
idiosyncratic exchange risk factor. iw and is are the sensitivities of the assets to the world
market and the global exchange risk factors respectively; eim ande
is are the sensitivities to the
idiosyncratic local market and local exchange risk factors respectively. w , s ,e
m ande
s are
the corresponding risk premium parameters.
3.3. Estimation procedure
19 Other studies using real rather than nominal exchange rate measures in empirical asset pricing tests for emergingmarkets include Carrieri, Errunza and Majerbi (2006b), Carrieri and Majerbi (2006) and Chaieb and Errunza (2007).20 The same model specification has been used in Carrieri and Majerbi (2006). This decomposition of the factors intoglobal and local is also similar in spirit to the risk decomposition into common and idiosyncratic components usedin Vassalou (2000).
8/7/2019 Currency Premium and Firm Characteristics
14/34
13
Each system of equations in models (4) and (5) is estimated using Hansens (1982)
generalized method of moments (GMM) and allowing for contemporaneous correlations in
the error terms it as in seemingly unrelated regressions models (SURM). The expected risk
premium parameters and the betas coefficients in each model are jointly estimated in a one-
step procedure to avoid the errors-in-variables problem implied by a two-step estimation
procedure la Fama and MacBeth (1973). We use a vector of ones and the contemporaneous
values of the factors, Fjt as the instruments in the GMM estimation. Therefore, the
orthogonality conditions are E( it Fjt)=0, and E( it )=0, for all i = 1,..,N and j=1,..,k. The
iterated GMM procedure employs the Newey-West (1987) estimator to correct for
heteroskedasticity and autocorrelation in the variance-covariance matrix of the parameters.21
The truncation parameter q in the Newey-West estimator is set equal to six, which reasonably
takes into account the number of significant lags in our dataset. We report estimates of the
unconditional betas and the corresponding factors risk premia as well as Hansens (1982) J-
test on the overidentifying restrictions. 22 The regression models in (4) and (5) are
overidentified because we impose the standard CAPM-type restriction on the parameters by
setting the intercepts equal to zero. Finally, we provide additional diagnostic tests based on
the estimated mean pricing errors (APE), root mean squared errors (RMSE) and adjusted
coefficients of determination (adj-R2).
4. Data descriptionOur dataset covers a total of 246 firms listed on the Korean Stock Exchange (KOSPI)
from March 1988 to December 2006. We use firm-level monthly returns obtained directly
from the Korean commercial database fn-guide. This database also includes detailed data on
individual firm characteristics such as size, number and percentage of shares held by
foreigners, volumes and values of shares traded. The available firms are grouped into three
different sets of 12 portfolios each constructed to reflect differences in the various firm-
specific characteristics of size, trading volume, and foreign ownership. Firm size is defined as
the average value of market capitalization of the company over the sample period. Trading
21 Similar to the studies by Vassalou (2000) and Ferson and Harvey (1994), we use iterated GMM following Fersonand Foerster (1994) who showed that such procedure has superior finite sample properties compared to a one-stepGMM estimation.22 The J-test statistic is the minimized value of the quadratic function in the GMM system, and follows a 2distribution when using an optimal weighting matrix as is the case in our estimation procedure.
8/7/2019 Currency Premium and Firm Characteristics
15/34
14
volume is the companys average trading volume over the sample period and foreign
ownership represents the average percentage of stocks held by foreign investors.
After sorting data by average of characteristics values in descending order we group 20
firms into one portfolio. So, for example, in the size-based portfolios, the largest 20 firms are
included in portfolio 1 and the next 20 largest firms go to portfolio 2 and so on until portfolio
12 which contains the 20 smallest firms. Value-weighted returns are then calculated for the
resulting portfolios. We repeat the same procedure to construct the two other sets of portfolios
based on trading volume and foreign ownership. 23 Therefore we have different portfolio
components for each firm characteristic resulting into 12 size-based portfolios, 12 trading
volume-based portfolios and 12 foreign ownership-based portfolios. We also compute equally-
weighted returns for all three groups of portfolios to perform some robustness checks on our
model estimations as explained in section 5 below.The world market return is computed from MSCI World index adjusted for dividends
and available from DataStream. The U.S. market returns series is computed from the S&P 500
index obtained from CRSP. The Korean market returns used as the local market risk factor is
obtained from fn-guide. All returns described above, including those for the three sets of
portfolios, are logarithmic, expressed in US dollar and computed in excess of the 30-day
eurodollar interest rate (from DataStream) used as a proxy for the risk-free rate.
As a measure of the global currency risk factor, we use the real Broad Exchange Rate
index computed by the Federal Reserve Board. This is a trade-weighted exchange rate index
of the US dollar, the reference currency, with respect to major and other important trading
partners including both developed and emerging countries. We will refer to this index as the
Global currency factor.24 We use the log-change in the inverse of this index to capture the
change in the real value of the foreign currencies with respect to the dollar as it should appear
in the model since we are taking the point of view of a US investor.
We compute the real bilateral exchange rate for the Korean won (KRW) against the US$
using the nominal exchange rate and the consumer price indexes for the US and Korea
available from the International Monetary Fund's International Financial Statistics (IFS) and
DataStream. The bilateral rate is expressed in US dollars by Korean won so that a positive
23 In each case, we discard the last six firms in the ranking (out of the total of 246) to have the same number of firmsin each portfolio.24 For more information on this index, see the Federal Reserve Bulletin, October 1998.
8/7/2019 Currency Premium and Firm Characteristics
16/34
15
(negative) change in the rate represents an appreciation (depreciation) of the Korean won with
respect to the dollar. We also use the real effective exchange rate (REER) index for Korea as
an alternative measure of the exchange risk factor in the two-factor model. As with the broad
exchange rate index, we use the log change in the real bilateral exchange rate as well as the
REER index for Korea.
Panel A of Table 1 summarizes the monthly excess returns for the market and exchange
risk factors as well as the US market returns. As shown by the Jarque-Bera test results, none of
returns series is normally distributed. On average, the Korean stock market excess returns
increased by 0.375% over the sample period while that of US stock market increased by
0.048%. Compared to the US market, we can see that the Korean market was relatively more
volatile during the sample period as shown by the higher standard deviation of returns.
Panels B, C and D show the summary statistics for the monthly excess returns for thesize-based, foreign ownership-based and trading volume-based portfolios respectively. None
of the portfolio returns are normally distributed based on Jarque-Bera test results. Most of the
size-based portfolios show negative excess returns during the whole sample period except
portfolio 1 to 3 with the largest 60 firms. Monthly portfolio excess returns range from -0.57%
for portfolio 12 to 0.45% for portfolio 1. Except for the portfolio 1 and 11, all the foreign
ownership-based portfolio returns are negative for the sample period.
Descriptive statistics of firm characteristics of the whole sample are shown in Panel A
of Table 2. Firm characteristics include firm size, trading volume, and foreign ownership. All
the figures are based on the average of monthly value during the sample period. As we can see,
there are significant differences in these characteristics across firms in our sample. For
example, firm sizes range from 6.75 to 23,858 million US dollars and the average firm size is
328 million US$. For foreign ownership structure, it ranges from almost zero (0.02%) to more
than half of the shares outstanding (52.13%). Trading volume varies very widely, ranging from
592 shares per month to more than 4 million shares and its monthly average is 210,158 shares.
Panel B of Table 2 shows the firm characteristics sorted by portfolio groups. Portfolio 1 has
the largest size and trading volume and the highest foreign ownership ratio whereas portfolio
12 has the smallest and lowest values, respectively.
8/7/2019 Currency Premium and Firm Characteristics
17/34
16
5. Empirical results5.1. Two-factor model
We first estimate the asset pricing model in equation (4) using the Korean market returns
and the change in the real bilateral exchange rate of the Korean won against the US dollar asthe two main risk factors.This is similar to the studies by Doukas, Hall and Lang (1999) for
the Japanese market where the underlying model only included the local market factor in
addition to the exchange rate factor. This setting assumes full segmentation of the Korean
market since no global market factor is included in the model, but this assumption will be
relaxed later when we estimate the partial integration four-factor model. We estimate model
(4) separately for each of the three sets of portfolios (ranked by size, foreign ownership and
trading volume). In each case, the portfolios sensitivities to the market and exchange risk
factors are jointly estimated with the corresponding risk premium parameters using the iterated
GMM procedure as described in Section 3.25
Panels A of Tables 2 to 4 report the results of
this estimation for each portfolio group.
As indicated by theJ-testin Tables 3, 4 and 5, the two-factor model cannot be rejected at
any statistical significance level. The estimated market premium parameter m is not
statistically significant on average over the sample period except when we use the trading
volume-based portfolios. This is not surprising since previous evidence showed that it is often
hard to detect significant prices of risk in an unconditional asset pricing model given the time
varying nature of these prices. The exchange risk premium parameter s in this two factor
model is generally negative and statistically significant across all three groups of portfolio
estimations, even at the unconditional level. This is consistent with the recent empirical
evidence such as Carrieri et al. (2006a and 2006b) who documented a significant negative
price of exchange risk in a conditional setting.
The exchange risk betas are generally negative and larger for firms with larger size and
those with higher foreign ownership percentage. For the size-based portfolios, the betas varyfrom -0.997 for portfolio#1 (including the 20 largest size firms) to 0.054 for portfolio#12
(including the 20 smallest size firms). Similarly, for the foreign ownership-based portfolios,
25 All exchange rate changes used in testing models (4) and (5) are computed such that a positive value means anappreciation of the foreign currencies against the US dollar (the reference currency) in real terms. Thus a positiveexposure to this factor means that the asset returns increase with a real depreciation of the dollar.
8/7/2019 Currency Premium and Firm Characteristics
18/34
17
the betas vary from 0.714 for portfolio#1 (highest foreign ownership percentage) and 0.034
for portfolio#12 (lowest foreign ownership percentage).
The local market betas are all highly significant (at the 1% level) for all portfolios in this
two-factor model. However, the exchange risk betas are only significant for the largest size
firms (portfolios 1-5) and for firms with higher foreign ownership percentage (portfolios 1-7).
When firms are grouped into portfolios based on trading volume levels, the estimated
exchange risk betas, which are also negative and highly significant, do not seem to vary too
much across assets compared to the size-based and the foreign ownership-based portfolios.
We plot the average exchange risk premia for each asset (computed from the product of
the estimated price of risks and the assets exposure coefficienti) in Figure 1. On average,
we can see that the exchange risk premium is larger for firms with larger sizes and higher
percentage of foreign ownership. The positive relationship between the exchange risk
premium and firm size documented here is consistent with some of the evidence suggested by
Carrieri and Majerbi (2006)26. Our results are also consistent with some of the evidence found
in the literature about exchange rate exposure where studies such as De Jong et al (2006) and
Muller and Verschoore (2006) found that the exposure coefficients seem to increase with firm
size. . The relationship between the estimated premium and trading volume is less obvious
although the premium seems to be the smallest for firms with the highest level of trading
volume (pf01).
5.2. Four-factor model
To further investigate the relationship between the exchange risk premium and firm
characteristics, we estimate equation (5) which assumes that the Korean stock market is
partially integrated by taking into account the impact of global risk factors such as the world
market risk on the pricing of local assets. This is important because in the previous two-factor
model, we may be overestimating the exchange risk premium because of missing pricing
factors. Similar to Carrieri and Majerbi (2006), our partial integration model includes theworld market risk, the residual local market risk, a global exchange risk factor measured by
the change in the real exchange rate index of the dollar, and a residual local exchange risk
26 However, as mentioned before, this study used size-based portfolios from nine emerging markets, including Korea,in the same cross-section when estimating the model.
8/7/2019 Currency Premium and Firm Characteristics
19/34
18
factor measured by the change in the real bilateral $/w exchange rate. Panels B in Tables 3 to 5
summarize the results of this estimation for each portfolio group.
First, based on theJ-testfor overidentifying restrictions, the four-factor model can not be
rejected at any statistical significance level. Furthermore, this model seems to provide a
slightly better specification than the two-factor model as shown by the small reduction in the
average pricing errors (APE) and the root mean squared errors (RMSE) for all portfolio
groups. The adjusted R-squared are also slightly higher than those observed in the two-factor
model for all portfolio groups..
The risk premium parameters w and me for the global and local market risks respectively
are both insignificant in this unconditional model, however, the exchange risk premium
parameters s and se are significant only when we use foreign ownership-based portfolios
(Table 4) and trading volume-based portfolios (Table 5). Similar to the previous two-factor
model, the risk premium parameter for the bilateral exchange risk factor is negative in all
estimations. The exposure coefficients for this factor are also negative and highly significant
for the larger size firms (portfolios 1-5) and those with higher levels of foreign ownership
(portfolios 1-7). We also find a similar pattern in terms of the size of the estimated risk
premium for the local currency factor which seems, on average, to be an increasing function of
both firm size and foreign ownership level. Figure 2 presents the estimated risk premia for the
local and the global exchange risk factors. When firms are grouped according to their size, the
total exchange risk premium estimated for the resulting portfolios is mainly driven by the
bilateral exchange risk factor which yields a positive premium that seems to be increasing with
the firm size, while the global currency risk premium is, on average over the sample period,
almost zero. When firms are grouped according to their foreign ownership structure, we also
find that the risk premium associated with the bilateral exchange rate factor is mostly positive
and larger for firms with higher level of foreign ownership. However, the global exchange risk
premium, which is negative for all portfolios, is now substantially larger than that obtained for
the size-based portfolios. Finally, consistent with the results of the two-factor model, we do
not find a clear relationship between the exchange risk premium and trading volume, although
here again we find some suggestion that the total exchange risk premium seems to be
decreasing and becoming negative for more liquid firms.
8/7/2019 Currency Premium and Firm Characteristics
20/34
19
5.3. Robustness checks
We perform a number of robustness checks at the two-factor model level. First, we
substitute the Korean real effective exchange rate (REER) index for the bilateral $/won rate as
a measure of the exchange rate risk factor. This aggregate measure captures the changes of the
Korean won value against the currencies of all its major trading partners. Previous studies such
as Choi et al.(1998) showed that the price of exchange risk, particularly at the unconditional
level, is sensitive to the choice of the exchange rate measure.27 Moreover, Dominguez and
Tesar (2006) studied firm level exposure to different exchange rates and found that the number
of significant exposure coefficients varies widely depending on the measure use.
Based on the REER measure, we find that the estimated exchange rate premium
parameter (s) remain significant at the 5% or 1% levels for all portfolio groups. The exposure
coefficientsis also remain highly significant for the largest size firms (portfolios 1-5) as well
as for firms with higher foreign ownership percentage (portfolios 1-4 &6) with the magnitude
of the exchange rate premium positively related to both firm size and foreign ownership. For
instance, for the size-based portfolios with significant exposures, the estimated premium varies
from 0.83% for portfolio 1 (largest size) to 0.26% for portfolio 5 and keeps decreasing until
reaching -0.13% for the smallest size portfolio. Similar results are obtained for the foreign
ownership-based portfolios using the REER. In this case the estimated premium varies from
0.46% for portfolio 1 to -0.14% for portfolio 12.We also find similar results for the trading
volume-based portfolios as with the bilateral rate, with no clear relationship between the
magnitude of the premium and our trading liquidity measure, except that portfolio 1 with the
20 most liquid firms having a relatively lower exchange risk premium.28
The next step in our robustness checks is to estimate all models above using equally
weighted portfolio returns for the same portfolio groups. Indeed, since the size factor seems
quite strong in both models, by computing equally weighted returns for each portfolio where
firms are grouped by foreign ownership level or trading volume, we try to avoid compounding
27 Choi et al. (1998) found that when using the bilateral JPY/US$ exchange rate, the test results support thehypothesis that exchange risk is priced in both the unconditional and conditional versions of the model. However,when the trade-weighted exchange rate is used as a measure for the exchange risk factor, the results are mixed andthe significance of the exchange risk factor is only confirmed in the conditional setting. In a similar vein, Dominguezand Tesar (2001) found that many firms in their sample, including some emerging markets, are exposed to one ormore bilateral exchange rates included in the world exchange rate index but not to the index itself.
28 Detailed results of the estimation using the REER can be obtained from the authors upon request.
8/7/2019 Currency Premium and Firm Characteristics
21/34
20
the size effect on these portfolios so that the resulting portfolio returns in each group only
reflect differences in foreign ownership or trading volume.
Finally, we would like to consider other liquidity measures such as turnover ratio to
further investigate the link between the exchange risk premium and this variable given the
weak results obtained with our current portfolio construction method based on trading volume.
This work is currently underway and will be reported in the final version of the paper.
6. Conclusion and future researchIn this paper, we provide preliminary evidence on the determinants of the exchange risk
premium in the stock market. Previous studies testing international asset pricing models
showed that exchange risk is a significant pricing factor that commands an important risk
premium in expected equity returns. However, most of these studies are based on market level
data and do not provide any insight as to how the estimated exchange risk premium may vary
across firms and what drives such cross-sectional differences . Based on detailed firm level
data from Korea, we try to address this question by looking at the relationship between the
exchange risk premium and some individual firm characteristics such as firm size, foreign
ownership, and liquidity. To our knowledge, this is the first study that looks into this important
issue. We construct our portfolios to reflect differences in these firm characteristics and
estimate the exchange risk premium using unconditional asset pricing models both in the
context of full segmentation and partial integration.
Consistent with previous evidence obtained at the market level, we find that, on average
over the long run, exchange risk is significantly priced in stock returns. More importantly, we
find that the estimated exchange risk premium for Korean firms is larger and more significant
for larger size firms as well as for firms with higher foreign ownership levels. The relationship
between the exchange risk premium and firm liquidity is less strong, although it seems that the
estimated premium is larger for firms with lower liquidity. Further investigation of this issueis needed as mentioned in the previous section before making any robust conclusions. For
instance, it would be interesting to see if similar results obtain for firms traded in other
countries. Moreover, given the time varying nature of the price of exchange risk and the
exposure to this factor, a conditional asset pricing model would be more interesting to
investigate the link between firm characteristics and an estimated time varying FX premium.
8/7/2019 Currency Premium and Firm Characteristics
22/34
21
REFERENCES
Abuaf, N. and P. Jorion, 1990, Purchasing power Parity in the Long Run, The Journal ofFinance, 45 (1), 157-174.
Acharya, V.V. and L.H. Pedersen, 2005, Asset pricing with Liquidity Risk, Journal of
Financial Economics 77, 375-410.
Adler, M. and B. Dumas, 1983, International Portfolio Choice and Corporation Finance: ASynthesis, The Journal of Finance, 38, 925-984.
Amihud, Yakov and Haim Mendelson, 1986, Asset Pricing and the Bid-Ask Spread, Journal ofFinancial Economics 17, 223-249.
Bailey, Warren and Julapa Jagtiani, 1994, Foreign Ownership Restrictions and Stock Prices inthe Thai Capital Market, Journal of Financial Economics 36, 57-87.
Bailey, Warren, Peter Chung and Jun-koo Kang, 1999, Foreign Ownership Restrictions andEquity Price Premiums: What Drives the Demand for Cross-Border Investments?, Journal ofFinancial and Quantitative Analysis, 34, 489-511.
Bekaert, G. and C.R. Harvey, 1995, Time Varying World Market Integration, Journal ofFinance, 50, 403-444.
Bekaert, G., Harvey, C., 2000. Foreign speculators and emerging equity markets, Journal ofFinance, 55, 565-613.
Bekaert, G., Harvey, C., Lundblad, C., 2005. Did Financial liberalization spur economicgrowth? Journal of Financial Economics, 77(1), 3-55.
Bekaert, G., C. R. Harvey and C. Lundblad, 2007, Liquidity and Expected Returns: Lessonsfrom Emerging Markets, Review of Financial Studies, 20, 1783-1825.
Brennan, M., T. Chordia and A. Subrahmanyam, 1998, Alternative Factor Specifications,Security Characteristics, and the Cross-Section of Expected Stock Returns; Journal ofFinancial Economics, 49, 345-373
Carrieri, F., 2001, The Effects of Liberalization on Market and Currency Risk in the EU,European Financial Management, 7, 259-290.
Chaieb, I. and V. Errunza, 2007, International Asset Pricing Under Segmentation and PPPDeviatins, Journal of Financial Economics, 86, 543-578.
Chan, L.K.C.; Hamao, Y.; Lakonishok, L., 1991, Fundamentals and Stock Returns in Japan,Journal of Finance, 46, 1739-1764.
8/7/2019 Currency Premium and Firm Characteristics
23/34
22
Chari A., Henry, P.B., 2004. Risk Sharing and Asset Prices: Evidence From a NaturalExperiment, Journal of Finance 59, 1295-1324.
Carrieri, F and B. Majerbi, 2006, The Pricing of Exchange Risk in Emerging Stock Markets,
Journal of International Business Studies, 37 (3), 372-391.
Carrieri, F, V. Errunza and B. Majerbi, 2006a, Does Emerging Markets Exchange Risk AffectGlobal Equity Prices, Journal of Financial and Quantitative Analysis, 41 (3), 511-540.
Carrieri, F, V. Errunza and B. Majerbi, 2006b, Local Risk Factors in Emerging Markets: Arethey Separately Priced?, Journal of Empirical Finance, 13(4-5), 444-461.
Choi. J.J., 1986, A Model of Firm Valuation With Exchange Exposure, Journal ofInternational Business Studies, 17, 153-159.
Choi, J.J., T. Hiraki, and N. Takezawa, 1998, Is Foreign Exchange Risk Priced in theJapanese Stock Market? Journal of Financial and Quantitative Analysis, 33, 361-382.
Choi, J.J. and M. Rajan, 1997, A Joint Test of Market Segmentation and Exchange RiskFactor in International Capital Markets, Journal of International Business Studies, 28, 29-49.
Christoffersen, P. H. Chung and V. Errunza, 2006, Size matters: The impact of capital marketliberalization on individual firms, Journal of International Money and Finance, 25, 1296-1318.
Chordia, T. R. Roll and A. Subrahmanyam, 2000, Commonality in Liquidity, Journal ofFinancial Economics, 56, 501-530.
Chuhan, P., 1994. Are institutional investors an important source of portfolio investment inemerging markets? Policy Research Working Paper, 1243, World Bank.
Cooper, Ian and Evi Kaplanis, 1994, Home Bias in Equity Portfolios, Inflation Hedging andInternational Capital Market Equilibrium, Review of Financial Studies 7, 45-60.
De Jong, A. Ligterink, J., and Macrae, V., 2006. A Firm-specific analysis of the exchange-rateexposure of Dutch firms, Journal of International Financial Management & Accounting, 17
(1), 1-28.
De Santis, G. and B. Gerard, 1998, How Big is the Premium for Currency Risk, Journal ofFinancial Economics, 49, 375-412.
Dominguez, C., and Tesar, L.L., 2001. A re-examination of exchange rate exposure,American Economic Review, 91, 396-399.
8/7/2019 Currency Premium and Firm Characteristics
24/34
23
Dominguez, C., and Tesar, L.L., 2006. Exchange rate exposure, Journal of InternationalEconomics, 68, 188-218.
Doukas, J., P. Hall and L. Lang, 1999, The Pricing of Currency Risk in Japan, Journal ofBanking and Finance, 23, 1-20.
Doukas, J.A., Hall, P.H., and Lang, L.H.P., 2003. Exchange rate exposure at the firm andindustry level, Financial Markets and Institutions, 12(5), 291-346.
Dumas, B., 1978, The Theory of the Trading Firm Revisited, The Journal of Finance, 33 (3),1019-1029.
Dumas, B. and B. Solnik, 1995, The World Price of Foreign Exchange Risk, The Journal ofFinance, 50, 445-479.
Errunza, V. and E. Losq, 1985, International Asset Pricing under Mild Segmentation: Theoryand Tests, Journal of Finance, 40, 105-124
Errunza, V., 2001. Foreign Portfolio Equity Investments, Financial Liberalization andEconomic Development, Special issue of Review of International Economics, InternationalFinancial Liberalization, Capital Flows and Exchange Rate Regimes: Essays in Honor ofRobert A. Mundell. Volume 9, 703-726.
Faff, R.W., and A. Marshall, 2005, International evidence on the determinants of foreignexchange rate exposure of multinational corporations, Journal of International BusinessStudies, 36, 539-558.
Fama, E. and K. French, 1992, The cross-section of expected stock returns; Journal ofFinance; Vol. 47, 427-465.
Fama, E. and J. MacBeth, 1973, Risk, Return and Equilibrium: Empirical Tests, Journal ofPolitical Economy, 71, 607-636.
Ferson, W.E. and S.R. Foerster, 1994, Finite Sample Properties of the Generalized Method ofMoments in Tests of Conditional Asset Pricing Models, Journal of Financial Economics, 36,29-35.
Ferson, W.E. and C.R. Harvey, 1994, Sources of Risk and Expected Returns in Global EquityMarkets, Journal of Banking and Finance, 18, 775-803.
French, Kenneth and James Poterba, 1991, International Diversification and InternationalEquity Markets, American Economic Review 81, 222-226.
Hamao, Y., 1988, An Empirical Examination of Arbitrage Pricing Theory: Using JapaneseData, Japan and the World Economy, 1, 45-61.
8/7/2019 Currency Premium and Firm Characteristics
25/34
24
Hansen, L.P., 1982, Large Sample Properties of Generalized Mathod of Moments Estimators,Econometrica, 50, 1029-1053.
Hasbrouck, J. and D.J. Seppi, 2001, Common Factors in Prices, Order Flows and Liquidity,
Journal of Financial Economics 59, 383-412.
Henry, P.B., 2000a. Stock market liberalization, economic reform, and emerging marketequity prices, Journal of Finance 55, 529-564.
Henry, P.B., 2000b. Do Stock market liberalizations cause investment booms? Journal ofFinancial Economics 58, 301-334.
Jorion, P., 1991, The Pricing of Exchange Rate Risk in the Stock Market, Journal of Financialand Quantitative Analysis, 26, 361-376.
Kang, Jun-Koo and Rene M. Stulz, 1997, Why is there a home bias? An analysis of foreignportfolio equity ownership in Japan, Journal of Financial Economics 46, 2-28.
Kim, E.H., Singal, V., 2000. Stock market opening: Experience of emerging economies,Journal of Business 73, 25-66.
Kim, W. and T. Sung, 2005, What makes firms manage FX risk?, Emerging Markets Review,6, 263-288.
Lewis, Karen K., 1999, Trying to Explain Home Bias in Equities and Consumption, Journal of
Economic Literature 37, 571-608.
Li, K., 1991, Testing Symmetry and Proportionality in PPP: A Panel-Data Approach, Journalof Business and Economic Statistics, 17, 409-418.
Merton, Robert C., 1987, A Simple Model of Capital Market Equilibrium with IncompleteInformation, Journal of Finance 42, 483-510.
Muller, A. and Verschoor, W.F.C, 2006. European foreign exchange risk exposure, EuropeanFinancial Management, 12(2), 195-220.
Nance, D., Smith, C., Smithson, C., 1993. On the determinants of corporate hedging. Journalof Finance, 1, 267 284.
Newey, W. and K. West, 1987, A Simple Positive-definite Heteroskedasticity andAutocorrelation Consistent Covariance Matrix, Econometrica, 55, 703-708.
OHara, M., 2003, Liquidity and Price Discovery, Journal of Finance 58, 1335-1354.
8/7/2019 Currency Premium and Firm Characteristics
26/34
25
Pastor, L. and R. Stambaugh, 2003, Liquidity Risk and Expected Stock Returns, Journal ofPolitical Economy 111, 642-685.
Patro, D. and J. Wald, 2004, Firm characteristics and the impact of emerging market
liberalizations, Journal of Banking and Finance, 29, 1671-1695.
Roll, R., 1979, Violations of Purchasing Power Parity and Their Implications for EfficientInternational Commodity Markets, in International Finance and Trade, Marchall Sarnat andGiorgio P. Szego (eds.). Cambridge, MA: Ballinger, PP.133-76.
Salehizadeh, M. and R. Taylor, 1999, A Test of Purchasing Power Parity for EmergingEconomies, Journal of International Financial Markets, Institutions and Money, 9, 183-193.
Sercu, P., 1980, A Generalization of the International Asset Pricing Model, Revue de
lAssociation Francaise de Finance, 1, 91-135.
Shapiro, A.C., 1974, Exchange Rate Changes, Inflation and the Value of the MultinationalCorporation, Journal of Finance, 30, 485-502.
Solnik, B., 1974, An Equilibrium Model of the International Capital Market, Journal ofEconomic Theory, 8, 500-524.
Stulz, R.M., 1981a, A Model of International Asset Pricing, Journal of Financial Economics,9, 383-406.
Stulz, R.M., 1981b, On the Effects of Barriers to International Investment, Journal of Finance,36, 923-934.
Vassalou, M., 2000, Exchange Rate and Foreign Inflation Risk Premiums in Global EquityReturns, Journal of International Money and Finance, 19, 433-470.
8/7/2019 Currency Premium and Firm Characteristics
27/34
Mean Std-Err. ewness ur os s arque- era 1 2 3Panel A. Country Returns
KOSPI 0.375 9.176 0.906 6.646 156 (0.00) 0.085 -0.03 -0.039S&P500 0.048 3.940 -0.459 3.899 16 (0.00) -0.056 -0.018 0.030MSCI World index 0.426 4.117 -0.460 3.661 12 (0.00) -0.011 -0.039 0.006Real Broad currency index -0.008 1.167 -0.011 3.314 0.934 (0.63) 0.348 0.006 -0.093Real bilateral $/W rate 0.039 3.624 -3.998 45.357 17497 (0.00) 0.079 0.048 -0.09Real Effective Exchange Rate 0.153 2.767 -3.823 39.976 13425 (0.00) 0.45 -0.037 -0.195
Panel B. Size-Based Portfolio Returnsportfolio # 1 0.45 12.01 0.28 6.05 90 (0.00) 0.07 -0.007 -0.001portfolio # 2 0.05 11.29 0.13 6.67 127 (0.00) 0.169 0.051 0.046portfolio # 3 0.10 11.12 -0.21 5.62 66 (0.00) 0.121 -0.009 0.016portfolio # 4 -0.25 11.73 -0.07 5.87 78 (0.00) 0.17 0.029 0.021portfolio # 5 -0.10 11.01 -0.29 6.06 91 (0.00) 0.186 0.047 0.018portfolio # 6 -0.43 11.16 -0.64 7.57 211 (0.00) 0.167 0.051 0.025
portfolio # 7 -0.53 12.07 -0.70 7.48 206 (0.00) 0.212 0.098 -0.059portfolio # 8 -0.50 11.53 -0.59 6.93 157 (0.00) 0.135 0.039 -0.009portfolio # 9 -0.13 10.94 -0.91 7.96 262 (0.00) 0.102 0.116 -0.055portfolio # 10 -0.24 11.04 -0.83 7.12 186 (0.00) -0.656 0.209 -0.126portfolio # 11 -0.65 11.98 -0.90 6.80 166 (0.00) 0.114 0.029 -0.011portfolio # 12 -0.57 11.39 -1.29 9.72 486 (0.00) 0.126 0.115 -0.062
Panel C. Foreign ownership-based Portfolio Returnsportfolio # 1 0.82 12.43 0.14 7.39 181 (0.00) 0.027 0.015 0.017portfolio # 2 -0.22 11.01 0.05 4.63 25 (0.00) 0.126 0.076 -0.033portfolio # 3 -0.07 11.73 -0.08 5.39 54 (0.00) 0.165 0.010 0.013portfolio # 4 -0.34 13.15 0.15 5.69 68 (0.00) 0.099 -0.058 0.049portfolio # 5 -0.71 13.79 -0.14 5.99 85 (0.00) 0.125 -0.059 0.073
portfolio # 6 -0.15 10.58 -0.19 5.36 54 (0.00) 0.199 0.099 0.065portfolio # 7 -0.48 12.87 0.30 7.83 222 (0.00) 0.088 -0.027 0.042portfolio # 8 -0.80 12.08 -0.33 6.31 107 (0.00) 0.151 0.012 -0.014portfolio # 9 -0.49 11.74 -0.48 7.47 196 (0.00) 0.217 0.079 -0.027portfolio # 10 -0.77 13.44 0.05 6.75 132 (0.00) 0.084 0.069 -0.032portfolio # 11 0.30 12.54 0.16 12.88 916 (0.00) -0.032 -0.022 -0.079portfolio # 12 -0.43 11.29 -0.86 6.93 172 (0.00) 0.172 0.105 0.014
26
Autocorrelations
Table 1. Summary Statistics for Excess Returns
8/7/2019 Currency Premium and Firm Characteristics
28/34
Table 1. Continued
Mean Std-Err. ewness urtos s arque- era 1 2 3Panel D. Trading volume-based Portfolio Returns
portfolio # 1 -0.66 12.46 0.24 4.24 17 (0.00) 0.085 0.014 0.002portfolio # 2 -0.07 12.38 0.22 5.46 58 (0.00) 0.134 -0.035 0.066portfolio # 3 0.84 13.34 0.09 7.13 160 (0.00) 0.059 -0.002 0.009portfolio # 4 -0.13 11.60 0.12 6.42 110 (0.00) 0.012 -0.070 0.035portfolio # 5 0.52 11.63 0.45 9.16 364 (0.00) 0.010 0.015 -0.082portfolio # 6 0.18 10.99 -0.31 4.85 36 (0.00) 0.180 0.012 0.030portfolio # 7 0.40 11.97 -0.17 6.36 107 (0.00) 0.138 0.088 -0.007portfolio # 8 -0.08 10.76 -0.64 6.05 103 (0.00) 0.133 0.076 -0.012portfolio # 9 0.38 10.53 -0.37 7.01 156 (0.00) 0.136 0.064 -0.049portfolio # 10 0.32 9.82 -0.75 6.65 146 (0.00) 0.130 0.090 -0.072portfolio # 11 0.23 10.69 -0.86 8.32 293 (0.00) 0.088 0.128 -0.031portfolio # 12 0.69 9.45 -0.01 5.64 65 (0.04) 0.164 0.144 -0.001
All returns are expressed in US dollar and are computed in excess of the one-month eurodollar deposit rate. KOSPI, S&P500 and M
monthly returns for Korea, USA, and the world, respectively. Portfolio construction is in a decesending order. For example, portfoincludes the 20 largest firms while portfolio #12 includes the 20 smallest firms. The P-value for a Jarque-Bera statistic in paranthe
leads to the rejection of the null hypothesis of a normal distribution. Bold numbers indicate significant autocorrelation.
27
Autocorrelations
8/7/2019 Currency Premium and Firm Characteristics
29/34
Table 2. Summary Statistics of Firm Characteristics
Panel A: Whole dataset (246 individual firms)
Mean Max Min Median St.Dev.Size(mil. US$) 328 23,858.83 6.75 69.04 1,574
Trading volume (#) 210,158 4,171,589 592 62,875 424,873
Trading Value (000s $) 2,178 94,184 20 599 6,781
Turnover ratio (%) 1.13% 5.60% 0.06% 0.86% 0.96%
Foreign ownership (#) 3,741,840 71,289,001 33 255,665 9,673,279
Foreign ownership (%) 8.93 52.13 0.02 4.49 10.58
Panel B: Summary of firm characteristics by portfolio group
foreignownership
basedportfolios
size(mil. US$) trading volume Turnover ratio foreign ownership(%)
Portfolio #1 mean 26,515.90 1,267,984 3.66% 21.17max 238,588.34 4,171,589 5.60% 23.14min 6,655.41 672,105 2.48% 19.8
Portfolio #2 mean 5,083.12 510,011 2.11% 17.63max 6,292.72 626,758 2.44% 19.8min 3,484.36 396,719 1.65% 15.39
Portfolio #3 mean 2,798.68 304,428 1.50% 13.49max 3,469.89 386,912 1.65% 15.11min 2,140.52 224,137 1.35% 11.74
Portfolio #4 mean 1,716.21 163,555 1.26% 8.9max 2,246.89 223,403 1.34% 11.57min 1,355.99 131,048 1.19% 7.56
Portfolio #5 mean 1,184.72 113,202 1.08% 6.83max 1,436.12 130,709 1.16% 7.43min 979.53 93,867 1.00% 5.46
Portfolio #6 mean 855.71 76,309 0.92% 4.9max 1,051.82 91,796 0.98% 5.39min 661.24 66,286 0.87% 4.4
Portfolio #7 mean 627.09 55,170 0.83% 3.4max 733.99 65,114 0.86% 4.39min 545.80 44,984 0.77% 2.54
Portfolio #8 mean 490.77 34,954 0.71% 2.2max 594.61 44,721 0.77% 2.54min 436.02 29,120 0.66% 1.97
Portfolio #9 mean 393.52 25,904 0.62% 1.45max 442.58 28,987 0.66% 1.91min 341.16 23,664 0.57% 1.05
Portfolio #10 mean 292.37 19,143 0.52% 0.87max 344.14 23,611 0.57% 1.03min 255.52 15,370 0.47% 0.66
Portfolio #11 mean 209.80 11,036 0.40% 0.51max 245.39 15,370 0.47% 0.65min 173.15 6,227 0.30% 0.42
Portfolio #12 mean 148.65 3,041 0.22% 0.22max 182.77 5,701 0.29% 0.4min 109.29 1,018 0.13% 0.04
28
liquiditybasedportfoliossizebasedportolios
8/7/2019 Currency Premium and Firm Characteristics
30/34
8/7/2019 Currency Premium and Firm Characteristics
31/34
8/7/2019 Currency Premium and Firm Characteristics
32/34
8/7/2019 Currency Premium and Firm Characteristics
33/34
Figure1.ExchangeRiskPremiaandFirmCharacteristics Twofactormodelusingbilateralexchangerates
0.40
0.20
0.00
0.20
0.40
0.60
0.80
1.00
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
Size
0.60
0.80
ForeignOwnership
Note:pf01referstoportfolio#1whichincludesfirmswiththelargestsize
orhighestforeignownershipstakeorthehighesttradingvolume.
0.40
0.20
0.00
0.20
0.40
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
0.00
0.50
1.00
1.50
2.00
2.50
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
TradingVolume
8/7/2019 Currency Premium and Firm Characteristics
34/34
Figure2.ExchangeRiskPremiaandFirmCharacteristics- Four-factor model -
0.20
0.00
0.20
0.40
0.60
0.80
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
GlobalXRpremium
bilateralXRpremium
totalFXpremium
Size
0.40
0.60
0.80
GlobalXRpremium
bilateralXRpremium
ForeignOwnership
0.60
0.40
0.20
0.00
0.20
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
totalFXpremium
1.20
0.80
0.40
0.00
0.40
0.80
1.20
pf12 pf11 pf10 pf09 pf08 pf07 pf06 pf05 pf04 pf03 pf02 pf01
GlobalXRpremium bilateralXRpremium totalFXpremium
TradingVolume