bene ts of globalization can be obtained once all capital controls are removed. These bene ts are presumed to include more stable exchange rates and higher growth rates in the economy, not to mention the lowering of risk. At least this has been the conventional thinking up until recently. Bekaert, Harvey, Lunblad Blad and Siegal (2007) propose that capital market openness holds the greatest potential for growth opportunities in liberalizing countries. However, there is an emerging viewpoint that holds that such market liberalizations might in fact lead to nancial instability and impair the ability of the government to deal with negative macroeconomic shocks. For a detailed view of this strand of thinking, one can look at papers by Stiglitz (2000, 2004) and others. Other papers associate such liberalizations with reduced cost of equity capital, increased returns to equity positions and net capital in ows: see papers by Henry (2000), Bekeart and Harvey (2000) and Kim, Landi and Yoo (2009). In essence there is controversy over the perceived bene ts and costs of market liberalization. In February 1980, Korea replaced its xed exchange rate system with a multiple-basket pegged exchange rate system, thus permitting the exchange rate to uctuate against major currencies relevant for its trade. Under this system, the basic exchange rate of the Won against the USD was determined as the weighted average of two baskets.1 In March 1990 the multiple-basket pegged exchange rate system was itself replaced by the Market Average Exchange Rate System (MARS). In December 1997 (during the peak of the Asian nancial crisis period) the daily uctuation limits for the interbank exchange rate were abolished and thus South Korea’s exchange rate system shifted to a totally free oating mechanism at that time.2 In May 1998, the country’s stock and money markets were also opened to foreign investor ownership. The impact of exchange rate uctuation on the stock markets can be summarized by at least two major facts. First, the volatility of the exchange rate in uences rms’ import and export businesses and other businesses sensitive to foreign exchange rates. This in turn affects their stock prices. Second, investors in foreign stock markets are subject to foreign exchange risk. Therefore, in order to avoid facing these risks, they may hold currency market positions or trade nancial derivatives so as to hedge the increased currency risks. Such derivative trading may have spillover effects on the stock market. They may also demand higher interest rates to compensate for the increased volatility, if traders choose not to hedge, and are invested in the spot debt markets.

The two baskets are the SDR basket and a trade-weighted basket composed of major trading currencies, with an adjustment factor which was termed the policy variable: Exchange Rate D SDR basket E TWB P [D E 1] . The exchange rate bands are and free before December 1, 1997, between December 1, 1997 and November 20,, r1997, 10.0 , and after December 16, 1997, respectively. r 2.25

Some papers have examined whether foreigners are at a disadvantage in the Korean market (Choe, Kho and Stulz (RFS), 2005) either because of lack of information available to foreigners or because foreigners face higher trading costs. By examining the cost structure of trades made by foreigners versus domestic traders, they concluded that foreigners face higher costs on their medium and large sized trades in the stock market.3 Kim and Yoo (2009) investigated the different behavior of equity investors in the Korean markets. They concluded that foreign investors are not speculators but long-term value investors. Kim, Land and Yoo (2009) also examined the inter-temporal behavior of foreign investors in the Korean equity markets. They showed that increases (decreases) in the net in ows of foreign funds lead to the appreciation (depreciation) of value of the Won. However, the reverse causation link does not hold. The motivation of this study was to explore the trading behavior of investment trust companies (ITCs), banks, and foreigners, under the free oating exchange rate system and open capital and nancial markets to assess evidence of adverse consequences of liberalization, if any. These groups are the dominant entities in the Korean markets and any material change in the markets will be re ected through their trading activities. Hence this will be a good place to search for any evidence of instability in the markets. Our paper contributes to the current literature on market liberalization by studying its impact on the spot currency, stock and futures markets in a particular country that has just undergone market liberalization. Korea is a good test case as there is reasonable legal protection for foreign investors in her markets.4 The study speci cally tried to nd out whether the patterns of trade in these markets changed drastically, which might give clues of some instability in the markets. The primary mechanism for the investigation is the linkages between the volatility of the daily high and low exchange rates (Won/USD) and the trading behavior of the different categories of the major market participants.5

Their data set ends at the year both the capital and foreign exchange markets were liberated. Ours starts at the beginning of this period and is therefore a post-liberalization analysis of the markets. There is a substantive literature that points the existence of a substantial “home bias” for investors. Hence if the opening of a foreign market will lead to substantial capital in ows, there must not only be the promise of substantial returns but also enough protection in place to preserve their control rights over their assets. Individual investors do not have a major role in the currency, derivatives or stock markets in Korea. Their share of the trading volume and market capitalization is very small and (Choe, Kho and Stulz (RFS 2005), reported that only 4.1% of individuals held stock market positions. Since there is no detailed rm level data, the track positions of rms over time. Hence the duration of individual positions cannot be tracked in the different markets was not examined.

First, the study examined whether the major players have different trading

behavior due to expected and unexpected volatilities in the exchange rate, namely to nd out whether these players exhibit asymmetric trading behaviors.6 Second, we analyze the relations between the major players’ trading behavior and the volatility of USD futures volume. This information was used to determine whether these players go long or short by trading USD futures simultaneously for hedging or speculation purposes. This article is organized as follows. Section 2 provides previous related research ndings on exchange rates and stock returns. Section 3 discusses the characteristics of the research model. Empirical results are presented in Section 4. The paper ends with a brief conclusion in Section 5.

The Korea Futures Exchange (KOFEX), launched on April 23, 1999, has become the largest derivative exchange in the world in 2004, by total annual volume of contracts traded. It registered trading of 2,586,818,602 contracts. It however, does not have the value nor the number of futures instruments in the Chicago

Board of Trade. Two kinds of currency-related products, Won/dollar futures and Won/dollar options, are listed on the KOFEX and they have grown dramatically. The trading volume of USD future contracts now ranks third among listed derivative contracts. However, we do not incorporate USD options contracts into our study due to the fact that were not available at the time of this study. This paper focuses on the South Korea Stock Price Index 200 (KOSPI 200), won exchange rate and USD future contracts. The daily data used in this paper covered the period from April 23, 1999 to February 28, 2005. The data were time-synchronized and drawn from the Korea Stock Exchange (KSE), the spot currency market and Korea Futures Exchange (KOFEX) respectively.7 The data thus covered the period of full market liberalization since the government eliminated the foreign investment ceiling completely on May 25, 1998 and the local bond markets and money markets were completely opened up to foreign traders and investors. USD future trading grew due to an increase in institutions’ longer-term hedging demand and the high volatility of the underlying exchange rate during the fourth quarter of 2004. In Korea’s nancial markets, the three major investor groups with high trading activity in the spot and future foreign exchange rate markets are investment trust companies, banks, and foreigners. Investment trust companies and foreigners have traded actively in the Won/US dollar futures market, while securities rms and other institutional participants have reduced their trading volume.8 In order to capture the impact of trading behavior of the three major players for the range volatility of Won/USD rate. Firstly, derived the base line spot trading volume series data and USD futures volume data together with their rates of change. Then process the rst-order differences of the base line data series to generate stationary - I (1) series. Finally, de ned the volatility of the trading volume as a logarithm of the ratio of the daily trading volume:

Vi ,t for i log(vi ,t / vi ,t 1 ) u 100 ,

ITCs ( I ), Banks( B), Foreigners( F ), USD Futures(U ) , where vi ,t represents the trading volume of the series i at day t , and vi ,t 1 represents the trading volume of the series i at day t 1 , and Vi ,t is the rate of change of volume in the data series i .

The data reported/are from April 23, 1999. Annual reports by Korea Futures Exchange (KOFEX) showed the trading volume by major investor groups to be investment trust companies, banks, and foreigners excluding individuals.

Range based volatility estimators are highly ef cient relative to other returns based estimators of volatility such as daily squared returns using opening and closing prices. Furthermore, they are robust to noise generated by market frictions: see Brandt and Diebold (2006), Alizadeh, Barandt and Diebold (2002) and Rogers, Christopher and Satchell (1991). Hence, along the lines of Chaboud and LeBaron (2001), this paper measures the volatility of the foreign exchange rate (Won/USD) with a scaled measure of the daily high-low range. This creates the X variable:

Xt

>log(hight ) log(lowt )@u 100 , where hight and lowt represent the day t high and low exchange rate of the Won against the USD, respectively. 3.2.

This study explored if the uctuation of the foreign exchange rate affects the spot and USD futures trading behavior among the three major investor groups and the relationship between spot and USD future volume. Employed the autoregressive integrated moving average (ARIMA) model to classify the range of the daily high and low of the foreign exchange rate (Won/USD) into expected and unexpected foreign exchange rate range volatilities. Some scholars have argued that there are asymmetric effects in nancial markets due to the expected and unexpected impacts. For example, Bessembinder and Seguin (1993) found that unexpected volume shocks have a larger effect on volatility than expected volume shocks do. The relation is asymmetric and the impact of positive unexpected volume shocks on volatility is larger than the impact of negative shocks.9 It was assumed that players use their expectations of the range volatility of the exchange rate to adjust their spot or USD future positions in advance. The two expected and unexpected variables are then added into a multivariate GARCH (1, 1) model. These models are described below. a.

The investigation tried to determine whether the major players have asymmetric trading behavior in response to expected and unexpected volatility components. It was de ned that the daily range of the high-low exchange rate is the volatility variable within the model. This variable, derived from the ARIMA model, is then divided into two variables, the expected and unexpected components of

Several related studies, such as Warther (1995), Chen et al. (1999), Chang et al. (2000), Naranjo and Nimalendran (2000), Bekeart and Wu (2000), and Wu (2001) also mention that the expected and unexpected innovation have asymmetric effects.

the daily high-low ranges.10 It was then observed that the unexpected high-low range component impacts the trading behavior of these major players more than the expected high-low range component does. In order to use the ARIMA methodology, it is rst necessary to identify whether each series is stationary. The ARIMA model can only be used on a stationary series. If it is determined that a series is non-stationary, then it could be differentiated repeatedly until a stationary series occurs. The original X variable series is transformed into a stationary series by applying the ADF and PP tests, as denoted by I (0) . Then employ the general ARIMA (p, q) model to derive the expected and unexpected component variables from the high-low range in the X variable. The ARIMA (p, q) model can be written as follows:

Xt where, X is composed of a logarithm of the daily high-low range of the Won/ USD rate, and

EX t UEX t

D 0 E (¦ D i X t i ) E (¦ E i H t i ) , X t EX t , where, EX t and UEX t represent the expected and the unexpected components of the daily high-low range of the Won/USD rate at day t , respectively. The optimal ARIMA model is de ned. The optimal criteria require that the estimated coef cients be signi cant, the model generates a minimum AIC value, and the residual terms have no series correlation. It was found that the optimal ARIMA model of the high-low Won/USD rate series is ARIMA (5,0). The expected range of the high and low exchange rate can be evaluated by the actual high-low range subtracted by the optimal residual terms. The residuals are estimated from the ARIMA model. The residual values are interpreted as the unexpected range of the high and low exchange rate. By creating these two variables, the model may account for whether asymmetric trading behavior does exist. It was demonstrated that the volatility transmission mechanism is asymmetric in effect. Negative innovations (when the high-low range expands) in South Korea’s exchange rate market increase the volatility in the spot market and this results in increased trading. Positive innovations (when the high-low range narrows), do not result in increased volatility. Furthermore, trading is not as robust as is the case for negative innovations.

There is another example which employs the general ARIMA (p,0,q) model to divide the expected and unexpected variables from the change in original variables, for example, Lee and Chen (2005).

Braun et al. (1995), Koutmos and Booth (1995) and Geoffrey, Martikainen and Tse (1997), all employed the EGARCH model to investigate that there are volatility asymmetric effects between the futures and spot markets. They assessed whether negative innovations generate a larger impact than positive innovations in these two nancial markets. The ndings were that most investors will initiate more hedging strategies in the face of negative impacts than in the face of positive impacts. b.

The multivariate GARCH model is not only is able to test for the time-varying variance or volatility in the spot and USD futures markets, but it can also be used to investigate volatility transmission among the three markets: exchange rate, spot, and USD futures. This paper adopted a multivariate GARCH (1, 1) model to investigate the dynamic relationships in the three markets to examine the phenomenon of feedback in uence arising from USD futures. For the three major players, the USD futures can be traded to hedge their spot position risk or to speculate under the free- oating rate. After computing the expected and unexpected components of the variables from the ARIMA model in the high-low range series, the conditional mean equation was changed so that both variables are included. With the multivariate GARCH (1,1) model, equations are written as follows:

VI ,t D I ¦ D I ,1,iVI ,t i ¦ D I , 2,i EX t i ¦ D I ,3,iUEX t i ¦ D I , 4,i Rt i H I ,t ,

VB ,t D B ¦ D B ,1,iVB ,t i ¦ D B , 2,i EX t i ¦ D B ,3,iUEX t i ¦ D B , 4,i Rt i H B ,t ,

VF ,t D F ¦D F ,1,iVF ,t i ¦ D F , 2,i EX t i ¦ D F ,3,iUEX t i ¦D F , 4,i Rt i H F ,t , (8) VU ,t D U ¦ D U ,1,iVU ,t i ¦ D U , 2,iVI ,t i ¦ D U ,3,iVB ,t i ¦ D U , 4,iVF ,t i H U ,t ,

Produced by The Berkeley Electronic Press, 2010 ª hII ,t «h « IB ,t « hIF ,t « ¬hIU ,t hIB ,t hIF ,t hBB ,t hBF ,t hBF ,t hFF ,t hBU ,t hFU ,t hIU ,t º hBU ,t »» . hFU ,t » » hUU ,t ¼

Here Vi ,t is the logarithm of the ratio of trading volume in the I , B , F , or U series at day t . Term Rt is the logarithm of the ratio of the KOSPI 200 Index at day t . Terms EXt and UEXt are respectively the expected and the unexpectenge components of the daily high-low price of Won/USD at day t . Term Rt is the index return of KOSPI 200. Because shocks of the mean equation are the main drivers in the multivariate model, it is important that the mean equation is not misspeci ed. The var models up to eight lags were estimated and test the individual and joint signi cance of the coef cients were tested. A lag length of two was chosen by the Akaike information criteria (AIC) (See Table 3).11 Thus, the time lag length of this model is two days per series. To model the time-varying covariance matrix H t , a multivariate GARCH model.12 This time-dependent, conditional parameterization is justi ed by our nding of heteroscedasticity in the volatility of trading volume. For this methodology, Engle and Mezrich (1996) provided a concise survey. This case the four-dimensional time-varying covariance matrix contains four variance series and six covariance series. In the diagonalized parameterization without spillovers the Positive De nite and the Vech models need 30 parameters. Given that we have to include the VAR and the terms for the transmission of volatility shocks, the number of parameters makes the estimation process intractable. On the contrary, the approach proposed by Bollerslev (1990) is less complex, because it restricts the correlation to be constant.13 The parameterization for conditional variances is shown in Equation (11) and for covariances in Equation (12):

hi ,t ci ai H i2,t 1 bi hi ,t 1 for i hij ,t

U ij hi ,t h j ,t for i z j and i, j

In Equation (11), hi ,t are the variances of I , B , F , or U series. Here, the spillovers of these four volatilities are included as lagged squared innovations. The coef cient bi accounts for the identical shock to volatility from the previous day, whereas the coef cients a i ( i I , B, F , U ) measure the impact of the sectoral volatility shocks. To preserve a non-negative variance we estimate the coef cients in the variance equations in absolute value were estimated. In Equation (12), the co-variances hij are driven by the variances pij and correlation coef cients.

This speci cation is also employed in Karolyi (1995). See, for example, Univariate (G)ARCH introduced by Engle (1982) and Bollerslev (1986). A similar model is employed by Koutmos and Booth (1995).

The log-likelihood function is de ned as follows:

¦ ¨© 2 ln(2S ) 0.5 ln t 1

· | H t | 0.5H tcH t 1H t ¸ ¹.

For the numerical optimization of this function, the investigation started with the simplex algorithm to reduce dependence on the starting values. Then it was switched to the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm.

D I , 4, 2 are all positively signi cant. The evidence that the trading behavior of foreigners is not signi cantly affected by the daily movement of the KOSPI 200 index thus seems to lend weight to the fact that foreign investors are long term investors. It would be expected the trading activities of long-term investors not to be sensitive to daily variation in the index. Another explanation could be that of the higher trading costs that they face as suggested by Choe, Kho and Stulz (2005). The estimated results of Equation (9) uncover whether these three major categories of investors will hold long or short spot positions by trading USD futures for hedging and speculation purposes under the free- oating system. According to the results of Equation (9), the coef cients D j , 2 are all positive and signi cant. This indicates that only ITCs will buy or sell stocks while trading USD futures. It is presumed that the ITCs will rearrange their spot positions against the unexpected range volatility of the exchange rate and then hold long or short USD futures as needed at the same time. This is in order to avoid any loss under the exchange rate uctuation. The banks and foreigners, however, have no obvious trading behavior like the ITCs, relatively speaking. This table reports the maximum-likelihood estimation results of equations (11) to (12), which c is the constant in the variance and covariance equations, a is the coef cient for the lagged squared residuals, and b is the conditional variance and covariance coef cient. Table 2 shows the results of the conditional variance equations. The coef cients b are all signi cant, implying that there are volatility-clustering phenomena among the trading volatilities of these three major groups and USD futures. In addition, it is more like the contemporaneous correlation of volatilities across trader types. The LR1, LR2, and LR3 of Table 2 test whether unexpected innovations (the case where the unexpected high-low range of Won/USD expands) cause a larger current volatility than expected innovations. The results for ITCs and banks are positive and the result of foreigners is negative. However, only LR1, the case for investment trusts, is signi cantly positive and it ts the assumption of the information asymmetric theory on nancial securities (such as Braun et al. (1995)). If a market participant ignores the asymmetric characteristics of unexpected innovations, in anticipating the volatility of the exchange rate market, then exchange rate exposure will increase and pro ts may be impaired. It may be concluded from this evidence that the ITCs should hold long or short spot positions in USD futures for purposes of hedging their asset positions under the free- oating exchange rate system. The banks and foreigners, however, would reduce investment in the stock market under an uncertain and volatile exchange rate. So in summary, if market liberalization results in more stable exchange rates this would be bene cial if the country wanted to attract foreign capital in ows.

Note: LR is the likelihood ratio test. 2. LR1~LR3 are all F 2 (1) distributions. Signi cance levels of 10%, 5%, and 1% are represented by *, **, and ***, respectively

There is some evidence suggesting that the trading behavior of foreigners in the stock market is not signi cantly impacted by movements of the index nor does their trading affect the market index returns. This evidence is consistent with two channels of reasoning. Either foreigners in Korea are long-term value investors who by de nition will not be sensitive to short-term (daily) movements in the index or they are subject to higher trading costs than are domestic participants. It is hoped that the availability of better data will help sort out which of these explanations is appropriate for the market. However, the was no evidence of irrationality found on the part of the participants or that there is any element of instability in the markets that were analyzed. Author information: Submitting author is Lloyd P. Blenman, Department of Finance and Business Law, University of North Carolina – Charlotte, Belk College of Business, Charlotte, NC 28223, U.S.A. Email: lblenman@uncc.edu. Tel: 704-687-7725. Fax: 704-687-6967. Dar-Hsin Chen is a Professor at the Graduate Institute of Finance, National Chiao-Tung University, Taiwan and he acknowledges nancial support from Taiwan’s National Science Council (Grant No.: NSC 94-2416-H212-008). Chun-Da Chen is an Assistant Professor at the Graduate Institute of International Business Management, Da-Yeh University, Taiwan.