WHY ARE KOREAN MARKETS SO VOLATILE?

Transcription

1 WHY ARE KOREAN MARKETS SO VOLATILE? Oliver Fratzscher Natalie Yoon-na Oh The World Bank November 2002 This paper has been produced in the context of Korea s Financial Sector Assessment Program. Discussions with Korean officials and market participants revealed concerns about the high volatility and leverage in Korean markets and the related market structure. This analytical paper illustrates additional findings related to Korean derivative markets and the structure of Korean securities markets. Conclusions and policy suggestions are still tentative. The authors gratefully acknowledge the collaboration of the Korean FSS/FSC, KSE, KOFEX, KOFA and KSDA. Comments very much appreciated, please forward to: ofratzscher@worldbank.org

2 EXECUTIVE SUMMARY Korean equity markets are among the most volatile markets in the world, and during the past five years had volatility ratios between 30% and 60%, about twice pre-1998 levels. Meanwhile, equity derivative markets have grown exponentially to become the largest in the world by traded contracts, and leverage has increased significantly as open-interest positions have tripled during This paper asks two related questions: has the rapid growth of Korean derivative markets contributed to volatility and how far has market structure exacerbated leverage and volatility? The analysis is based on a dataset provided by KSE which covers daily trading activity in the KOSPI 200 index and futures during the period 1995 through Trading volumes have increased significantly during this period, driven by technological improvements (online trading), low transaction costs, as well as supportive regulatory and tax policies. Moreover, this period includes the Asian crisis (late 1997) and the correction in worldwide equity markets (since mid-2000). However, the econometric analysis reveals that a major contributor to volatility in Korean equity markets has been the speculative behavior in derivative markets since This finding is in contrast to experience in most industrialized countries where derivative markets have reduced market volatility. The main difference in Korean markets is the weak presence of institutional investors and the mostly speculative behavior of securities firms. Noteworthy is the lack of confidence in the mutual fund industry, given the unresolved situation of investment trust companies and the large savings in bank trust accounts. This has contributed to a market structure that is mainly retail-driven, short-term oriented, and where securities firms seek revenues from high turnover and speculative trading given rapidly declining commission rates. The empirical analysis reveals that securities firms are indeed the main contributors to equity market volatility, although their trading volumes are less than 4% in cash markets. However, they conduct over 90% of their trading in futures and options markets, where on average 10% of their trading reflects open positions. This is again in contrast to markets in most industrialized countries, where open-interest positions are taken by institutional investors which typically contribute to a reduction of market volatility. The role of foreign investment in Korean equity markets has been frequently analyzed. Foreign investors have taken net long positions in equity futures since 1999 and tripled their trading in derivative markets during The empirical analysis shows that foreign investors have also significantly contributed to market volatility, albeit to a lesser extent than securities firms. Their impact on volatility was most pronounced in late- 1997, when foreign banks reduced credit lines and foreign investors exited the stock market, doubling both cash and derivative trading volumes in the last quarter of Policy issues are related to equity market volatility in two areas: Regulatory and prudential issues relate to the calculation of capital adequacy for securities companies, enforcement, transparency, and measures to curb extreme volatility. Tax policy issues relate to the growing leverage and speculative nature of Korean markets and measures that could motivate the growth of institutional investors. November 2002 Volatility of Korean markets page 2

3 WHY ARE KOREAN MARKETS SO VOLATILE? 1. INTRODUCTION Korea s financial system has made great strides since 1997 but widely remains bankdriven as bank deposits still exceed the amount of assets managed by institutional investors (table 1). Especially the weak mutual fund industry (32% of GDP ; less than one third as compared to Australia, Hong Kong SAR, Singapore) remains constrained by large bank trust accounts and the unresolved situation of investment trust companies. The lack of institutional demand has also constrained the size of equity capital markets where capitalization (47% of GDP) remains below one half of those levels achieved in Australia, Spain, and Chile. Despite the positive equity market returns in 2001, KOSPI volatility remains at very high levels of 35% which is about twice the pre-1997 volatility and compares to average volatility in emerging markets of around 25% (table 2, chart 2). In fact, Korean equity market volatility is among the top-three in the world, only comparable to those in Argentina, Brazil, and Russia although it has declined from levels of between 30% to 60% during the past four years. Korean derivative markets have also developed at a rapid speed after KOSPI futures were introduced in 1996, options in 1997, and bond futures in Equity future volumes have doubled in the past two years and in 2001 futures trading was three times the size of cash market trading, and six times the size of the equity market capitalization. By any comparison, the size of Korean derivative markets is huge, only comparable to the three largest derivative markets in the world. For example, Korea s ratio of derivatives to cash trading is 2.5 times larger than that observed at Australia s ASX. Moreover, open interest positions have tripled, according to KSE data, from 10 million contracts in 2000 to over 37 million contracts by October This study is motivated by the following question: could there be any relation between rising leverage in Korean derivative markets and persistently high volatility in underlying equity markets? Moreover, does Korea s particular market structure, with emphasis on retail-trading and active securities firms, contribute to both phenomena? Might there be any policy measures that could alleviate high leverage and volatility? The analysis in this paper attempts to answer these questions. In the following parts, the literature on derivatives and the role of institutional investors is reviewed, the methodology is introduced, and a clear determination is made that Korean derivative markets have indeed increased equity market volatility. Subsequently, it is shown that the retail-oriented market structure and weak presence of institutional investors have allowed securities firms to assume a very active role, which often includes speculative behavior, and that securities firms have indeed contributed the largest part of volatility. One possible explanation of this observation is the speculative behavior of securities firms in derivative markets, where 90% of their trading occurs, of which 10% appears to be in open-interest positions. Finally, some thoughts are presented on whether regulatory, prudential, or tax policy measures could help to alleviate these problems, but any observations must remain tentative and open for discussion with the Korean authorities. November 2002 Volatility of Korean markets page 3

4 At the outset, two caveats need to be mentioned. First, it is very difficult to control for the micro-market structure, especially the weak long-term demand due to the absence of strong institutional investors. Long-term statistics of the Korean equity market (chart 1) show that the index has fluctuated around the same level with zero trend since 1987, which appears to be driven by low underlying profitability and high new share issuance. As a result, the behavior of market participants has become more speculative, because long-term risk-adjusted returns have been dismal. Second, international effects of equity market volatility have also influenced the Korean markets, although the impact has changed over time (chart 2). For example, correlation with the Dow Jones volatility has been high during crisis periods, but it has been negative since January However, it appears that the largest effect has been transmitted from volatility of Hong Kong markets. 2. LITERATURE Several explanations in the past have been offered to explain the perceived increase in volatility in other bourses. One which has received most attention by both academics and by market practitioners in the past is the introduction of a futures market or a derivatives market. The introduction of a derivatives market has consistently been surrounded with controversy. Two competing hypotheses exist with respect to the impact of the introduction of the derivatives market on the underlying market. The first set of arguments generally put forward is based upon the notion that as markets become more complete as a result of the derivatives market, investors are better off. Due to lower transaction costs in the derivatives market, it promotes rapid price discovery in the futures market and allows more informed traders in the market place (Fleming, Ostdiek and Whaley (1996)). Froot and Perold (1991) show that market depth is increased by more rapid dissemination of market information, as might be expected to occur upon introduction of equity index futures trading. Also, Danthine (1978) suggests that futures markets improve market depth and reduce volatility, because the cost to informed traders of responding to mis-pricing is reduced. Conventional wisdom suggests that futures trading will add more informed traders to the cash market, making the cash market more liquid and, therefore, less volatile. This is a decidedly pro-derivative view. One empirical study that is in favor of pro-derivatives is Edwards (1988), which documents a small but statistically significant decline in volatility after the introduction of financial futures trading in the following contracts: S&P500 Index, the Value Line Index, T-Bills, and Eurodollar 90-day Time Deposits. Also, Bessembinder and Seguin (1992) report that estimated coefficients relating equity volatility to spot trading volume (S&P 500) declined significantly subsequent to the introduction of futures trading. Pilar and Rafael (2002), looking at the effect of the introduction of derivatives (futures and options) in the Spanish market on the volatility finds that conditional volatility of the underlying index declines after derivative markets are introduced. Similarly, So and Yu (1999) suggest that volatility in Hong Kong decreased after futures and options trading. Lee and Ohk (1992) look at four exchanges, namely Australia, Hong Kong, Japan and United Kingdom and find that Hong Kong and Australia decreased volatility significantly after the introduction of futures, although it was the opposite in the other two markets. November 2002 Volatility of Korean markets page 4

5 In contrast, the second school of thought is based on the assumption that derivative securities, which are typically highly leveraged, encourage market destabilizing speculation. Such destabilization stems from a belief that derivatives markets result in uninformed or irrational speculative trading in both futures and cash markets. Such speculators, it is argued, drive prices up or down hoping for short-run bandwagon profits. Stein (1987) argues that futures trading by poorly informed speculators can destabilize the spot market. Hence, the fluctuations in asset prices are attributed not to change in economic fundamentals but to large, speculative trading. This anti-derivative view has become increasingly popular in the wake of several international financial markets crises caused by abuses of derivative products and or markets. 1 More specifically, Figlewski (1981) finds a positive relation between the volatility of GNMA securities prices and open interest in GNMA futures contracts and concluded that the futures market activity increased volatility in the market place. For S&P 500 stocks, Harris (1989) reports that by comparing S&P 500 stocks to non-matched non-s&p stocks before and after the introduction of S&P 500 futures trading, although the differences are economically small for S&P500 stocks, volatility is greater after the introduction of futures trading. In contrast to Edwards (1988) conclusion that derivative trading does not destabilize the underlying market using the standard F-test for equal variance, Kamara et al. (1992) use nonparametric tests to find that post-derivative trading volatility for the underlying market is significantly higher. Antoniou and Homes (1995) confirm the results of Lee and Ohk (1992) by discovering a significant increase in volatility for FTSE-100 data after the introduction of futures trading. The literature is inconclusive on whether the introduction of derivatives promote or reduce volatility in the marketplace. The outcome is unique for different exchanges and different contracts as well as for different market structures. It is therefore of interest to investigate whether the introduction of the derivatives market has increased or reduced volatility in the rapidly growing Korean derivative market 2. This paper focuses especially on who is responsible for this volatility in the Korean securities market and what factors promoted the volatility. The next section will outline the methodology adopted in this study and then the paper will investigate whether there has been increased volatility after the introduction of KOSPI 200 futures and which relationship exists between volatility and different trader types in the Korean securities market. 3. METHODOLOGY 3.1 Base model The regression model used in this paper follows the work by Schwert (1990, Review of Financial Studies). Following his work, Bessembinder and Seguin (1992/1993, Journal of Finance and Journal of Financial and Quantitative Analysis, respectively), have used the same regression model as Schwert (1990) but added in activity variables to provide evidence on interrelations between spot trading volume, futures trading activity, and equity volatility. The method involves iterating between a pair of the form: 1 Stock Market Crash in 1987 and the Asian Crisis in Futures were first introduced in the Korean market and started trading on 3 rd May November 2002 Volatility of Korean markets page 5

6 Rt n j 1 j Rt j 4 i 1 i Di n j 1 j t j U t....<1> t n j 1 j U t j 4 i 1 i D i m k 1 j Ak n j 1 j t j e t <2> Rt = daily return Di = Daily dummy Variable t-j = Volatility U t = Unexpected return captured by the residuals in equation1 Ak= Activity Variable. Volume for equity market and both volume and open interest for the futures market. Initially, the first equation estimates the conditional return based on lagged returns without using lagged volatility estimates. The residual from equation <1>, U t, is transformed using t Ut / 2 which is the estimated conditional return standard deviation on day t. The estimated conditional return standard deviation is put into equation <2> and then the process is iterated with lagged standard deviation estimates included as regressors in estimating equation <1> to regress the volume-volatility relationship in equation <2>. Bessembinder and Seguin (1993), suggest in their study that in investigating the volatility volume relation in financial markets it will be more complete to add in the type of trader generating the volume in the marketplace. Daigler and Wiley take on this suggestion (1999, Journal of Finance), and analyze the impact of trader-types on the futures volatility-volume relation. Along with the suggestion by Bessembinder and Seguin (1993), the authors modify or improve the methodology adopted from the original model. Their justification for modifying the model is that by deriving volatility from the estimated conditional return, standard deviation does not include the volatility effects occurring within the day, and hence is not the best measure of daily volatility in relation to daily volume. Daigler and Wiley (1999) incorporate, along with the original volatility measure, an alternative volatility measure in their paper which is Garman and Klass (1980) 3 as well as high-low range measures of volatility. These alternative volatility measures make use of daily high and low prices in addition to closing prices. This provides three daily price observations rather than only one. It has been shown previously that a variance estimator using high and low prices in addition to closing prices is substantially more efficient, about seven times greater than an estimator based only on the close-to-close variance 4. Similarly, Daigler and Wiley (1999) find that by using high-low variance estimator the explanatory power of the model improves. In 3 This measure is weighted average of the Parkinson high-low estimator and the traditional standard deviation close-to-close estimator. The efficiency of the latter is not much greater than that of Parkinson s estimator. 4 Beckers (1983), and Garman and Klass (1980) November 2002 Volatility of Korean markets page 6

7 view of this finding, this paper seeks to measure volatility using the high-low variance estimator for both regressions <1> and <2> instead of deriving volatility measure using the original method. The rest of the regression model remains the same. A more efficient Parkinson s high-low variance estimator 5 [ln(ht) ln(lt)] 2 /4ln(2)] using intra-day high(h) and low (L) prices are used in the model as a proxy for volatility. Also, the intraday high and low volatility measure [(Ht Lt)/Ct] are also examined but are not reported as the results are remaining very similar. 3.2 Specification As a preliminary step in performing the regression model, augmented Dickey Fuller tests (5 lags) were conducted for all variables in testing for series containing a unit root and for trends in the time series data. Mackinnon critical values for rejection of hypothesis of a unit root for all series were tested at a 0.1 level. Both volume and open series showed significant trends in our time series data. To control for both secular growth and trend in volume and open interest the series were de-trended by subtracting moving average of 50 days from the original series. Partial autocorrelation was examined (correlogram and Q- statistics), specifying lags up to 36. There is at least one positive and significant partial autocorrelation at higher lags but lag of one showed no sign of autocorrelation. The presence of significant partial autocorrelation reinforces the need to accommodate volatility persistence in the regression model. To separate the expected and unexpected level of volume and open interest, an ARIMA (0,1,10) model was used, which has been previously established in the literature. For robustness, other ARIMA specifications were examined, which were based on no serial correlation in the residuals using the Breusch- Godfrey LM serial correlation test at a significance level of at least 0.1. To test for multicollinearity, the Pearson correlation coefficient was computed. The test statistics for individual coefficients used are t-statistics for the hypothesis that the coefficient is zero, computed using White s (1980) heteroskedasticity consistent standard errors. Lagged standard deviation estimates are included to accommodate the persistence of volatility shocks (see French, Schwert and Stambaugh, 1987, and Jain and Joh, 1988). We use the Schwartz (1978) criterion to set the lag limits to 10 in order to incorporate the range of significant lags identified for each variable. Moreover, daily dummies allow for a day of week difference in mean volatilities. Empirical studies of the volatility-volume relation for various financial instruments show that a significant positive relation exists between volatility and volume, and R 2 values from regressions of volatility on volume typically range from percent when using volatility measured as the absolute price change or the squared price change as in the case of Bessembinder and Seguin (1992, 1993), but when using better volatility measures the R 2 increases up to 72 percent as reported by Daigler and Wiley (1999). The explanatory power of regression models in this study ranged from 17 to 49 percent. The DW statistics for all the regressions were closely around 2, showing no sign of autocorrelation. 5 This High Low Variance estimator was developed by Parkinson (1980). Beckers (1983) compared between Parkinson s High-Low Variance and traditional close to close variance and found that, in general, Parkinson s estimator incorporated new information and was a more accurate measure of volatility. November 2002 Volatility of Korean markets page 7

8 3.3 Regression analysis Using equation <1>, the unexpected return is computed, which is the residual in regression model <1>. The return measure is computed as ln(pt/pt-1), where Pt and Pt-1 are closing prices on successive days. The R 2 from equation <1> is less than five percent for our study which appears to be low, but Bessembinder and Seguin (1993) confirm that the highest R 2 was three percent, which is consistent with weak-form efficiency, as exante known regressors have little predictive power for realized returns. Past unexpected returns in the specification (equation <2>) are included first because that allows for possible effects of recent returns on volatility in the model. Second, many studies of equity spot market volatility (incl. Schwert, 1990), found that these lags have some explanatory power. Past studies, including several cited by Karpoff (1987), have also documented a correlation between volume and contemporaneous signed price changes. Finally, according to Bessembinder and Seguin (1993), lagged signed forecast errors and lagged ts, which are proportional to unsigned forecast errors, allow the relation between unexpected return and volatility to vary depending on the sign of the unexpected return. The effects of trading activity on volatilities are evaluated by including trading activity variables Ak in equation <2>. It has been documented that the activity variables are serially correlated, indicating that activity variables are highly forecastable, hence, recent literature on volatility-volume relationships has differentiated the effects of expected and unexpected activity variables in the regression model. This partitioning allows us to examine the extent to which surprises versus trend activity variables affect the volatilityvolume relationship. Activity variables in our study showed a significant trend in the series according to the ADF test. Further, there is at least one positive and significant partial autocorrelation at higher lags for the series according to the correlogram and Q statistics. To mitigate any effects of secular volume growth and also to remove any trend in the series, we first construct a de-trended activity series by applying a 50 day moving average to the original series. We then partition the de-trended series into expected and unexpected components using the ARIMA (0,1,10) 6 specification. This specification calculates the expected value using the previous day s volume and the 10-day moving average of the change in volume. The unexpected component of the de-trended series is interpreted as the daily activity shock. The expected component of the de-trended series which is roughly equal to the prior day s level of the de-trended series reflects activity that is forecastable 2ut highly variable 1cross days. The unexpected component of each series is defined as the estimated residual from the second regression, while the expected component is defined as the difference between the actual series and the unexpected component. 6 Bessembinder and Seguin (1993) document that their study does not attempt to find an ideal model for each series, but instead chooses an arbitrarily long set of autoregressive coefficients. ARIMA (0,1,10) has been used by Bessembinder and Seguin (1992, 1993) and Daigler and Wiley (1999) and our study looked at both ARIMA (0,1,10) and other ARIMA models to deal with serial correlation, but failed to detect any significant differences in the results. Hence, out study reports only results from ARIMA (0,1,10) in order to be consistent with other studies. November 2002 Volatility of Korean markets page 8

9 3.4 Open-interest regression The open interest data provide an additional measure of trading activity. Open interest is partitioned into expected and unexpected components using the ARIMA (0,1,10) model. This allows analysis of open interest along two dimensions. First, the expected portion reflects open interest as of the beginning of the trading day. The second dimension, unexpected open interest, captures unanticipated changes in net contract formation. Hence, expected open interest is a proxy for yesterday s level of open interest whilst unexpected component of open interest approximates the change in open interest in a day. Bessembinder and Seguin (1993) note the importance of including open interest as an activity variable. Firstly, by combining volume and open interest variable in the model it provides insights into the price effects of market activity generated by informed versus uninformed traders or hedgers versus speculators. This is because many speculators are day traders who do not hold open positions overnight. Closed positions as of the close of trading likely reflects primarily hedging activity and, thus proxies for the amount of uninformed trading. Secondly, the open interest is a good proxy for depth in the market because open interest captures the willingness and ability of traders to risk capital and take positions in response to a perceived deviation of price from intrinsic value. 4. KOSPI-200 FUTURES Two different measures are used to investigate whether volatility of the Korean equity market increased or decreased after the introduction of futures trading: The first measure follows the study of Edwards (1988) by using the F-test and Brown-Forsythe for equal variance of returns for inter-day 7 and the T-test for equal mean of volatility using Parkinson and standard high-low intra-day volatility measures 8. Table 5 shows that two intra-day (Parkinson and standard High-low) measures and one inter-day volatility measure are consistent in reporting that the volatility has increased significantly since the introduction of futures market in May Both T-test and F-test reject the null hypothesis that pre futures and post futures mean volatility is the same at the 1% significance level. The positive coefficients further illustrate that the volatility after the introduction of futures is greater than before, indicating that the volatility has increased significantly since the introduction of futures trading. During the sample period, the Asian Crisis (late 1997) was considered to have the most significant impact in increasing volatility. To avoid extreme periods which may have been the major factor in increasing volatility, an additional post-futures sample is constructed which discards dates during the Asian Crisis period (July 1997 to December 1998). The result for post-futures trading excluding the Asian Crisis does not alter the result that volatility has increased since the futures trading. Table 6 reports a significant increase in volatility post-futures trading, period by period. The first period analyzes pre-futures trading up to the Asian Crisis. The second period 7 Return is computed using ln(pt/pt-1) where Pt represents closing price for today and Pt-1 is closing price for yesterday. 8 Parkinson volatility measure is [ln(ht) ln(lt)] 2 /4ln(2)] and standard high-low is (Ht-Lt/Pt) where H is high price and L is Low price for the day. November 2002 Volatility of Korean markets page 9

10 covers futures trading during the Asian Crisis. The third period starts after the Asian crisis and the fourth period includes the recession period until the end of the sample. As expected, the difference in volatility is greatest during the Asian Crisis and post crisis or restructuring period (period 2 and 3) and falls during the fourth period, but the difference is still significant at the 1% level. The results confirm that the volatility post-derivative trading has indeed increased. The second measure follows the regression model adopted by Bessembinder and Seguin (1992). Table 7 reports expected and unexpected volume and volatility relationships for the underlying KOSPI (columns 1 and 2).To investigate the effect of equity index futures trading on stock market volatility, we include a dummy variable equal to one for the trading days subsequent to the introduction of KOSPI200 futures trading on May 6 th, 1996, and equal to zero for the trading days in our sample prior to this date. We allow the regression intercept and slope coefficients on volume variables to shift subsequent to the introduction of equity index futures trading. The results are reported in third and fourth columns of table 7. The key result of this analysis is that the estimated change in the slope coefficient associated with both expected and unexpected spot trading volume remain positive and the magnitude of the coefficient decreases only marginally after the introduction of futures which indicates that the futures trading had little impact in reducing the volatility in the underlying market, according to this measure. The estimated coefficient for the shift in the regression intercept subsequent to the introduction of futures trading was reduced by a factor of ten but it is not significant, hence we cannot conclude that the volatility level has decreased as a result of the introduction of futures. These findings are not consistent with the notion that equity volatility has been reduced and market depth (measured as the share volume required to move prices) has been increased by the introduction of futures trading. This suggests that the pro-derivative view does not hold in the case of Korea. 5. MARKET STRUCTURE Following the work of Daigler and Wiley (1999), this study looks at the impact of the trader-type on the volatility-volume relation for the Korean Stock Exchange. The period of investigation is between January 1995 to December 2001, based on a dataset which was kindly provided by KSE. The study partitions the period of investigation into five sub-periods. This is done to separate different phases of the economy cycle during the period of examination, such as the Asian Crisis during 1997, restructuring period during 1998, and recession in late It is therefore considered that dividing the sample period would give more meaningful results of our investigation. The first sub-period looks at the underlying market for the period where no derivatives market existed and the period is labeled period 1. The second-sub-period covers from the inception of derivatives trading to the start of the Asian crisis. The third sub-period is during the Asian Crisis. The fourth sub-period is the reconstruction period after the Asian crisis and before the world recession. The fifth period is during the world recession to the end of the sample period, December Each of the sub-periods is 18 months long, with exception of the first sub-period. November 2002 Volatility of Korean markets page 10

11 KSE has also provided us with data on four types of traders in the securities market, which are classified as foreigners, institutions, retailers and securities firms. This section aims to provide an analysis of which type of trader contributed most to the volatility in the Korean market during the specific sub-periods. The results of the regression are shown in table 8. The main finding is that securities companies and foreign investors contributed most to the volatility during the sample period 1995 to The coefficients show that the unexpected shock in volume by securities firms contributed most significantly to volatility, approximately with twice the magnitude of the foreigners. But looking at the each sub-period, it is revealed that different investor types contributed to volatility over different periods: Before the Asian crisis (periods 1 and 2), volatility was mostly driven by retail investors, for both expected and unexpected levels. During this period, retail investors had a positive and significant contribution to the volatility at both expected and unexpected levels. During the Asian crisis (period 3) most trader types had a significant impact to the volatility at the unexpected level 9. This is expected during this period as the market was turbulent and any unexpected volume by any trader type would have created volatility. An interesting observation during this period is that the dominant investor type that contributed to the volatility was the foreign investors, unlike in pre-crisis periods when retailers had the most significant impact. In the post-asian crisis periods (periods 4 and 5), there is another change in dominant investor types: Now, securities companies dominate the volatility in the marketplace for both periods 4 and 5. Foreign investors still have a significant impact on the volatility but the magnitude of the coefficient confirms that the securities companies contribute most to the volatility in the Korean securities market during the period However, Granger-causality tests have been inconclusive on whether volumes of particular trader types cause volatility or whether the reverse relationship dominates. From our observations, there are two offsetting factors present: first, the historically low returns in Korean equity markets combined with fast technological changes have clearly introduced higher market volatility as a structural phenomenon. On the other hand, it appears that trading behavior of particular investor types, especially of securities firms, have further increased volatility, mainly through speculative trading in equity derivatives. 5.1 Retail versus institutional investors Table 10 presents the investor-type composition for all sub-periods. In terms of trading volumes, the Korean market appears to be largely retail-driven. The retail investors dominate the market in terms of volume (65% average, up to 89% in period 5). This is similar for the futures market where retail investors dominate the market in terms of both trading volume and value. While the retailers are the dominant players in the Korean market for both underlying and the derivatives market, the institutional players are relatively small 10. This can be explained by still relatively inactive insurance companies trading, limited engagement of pension funds, and low credibility in the mutual fund 9 Securities companies does not show significance in the table reported but for volatility measure 2 it does show significance at 0.1 level. 10 Institutional investors trade approximately 10 percent in both equity and derivatives markets. November 2002 Volatility of Korean markets page 11

12 industry. The latter has been hurt by the still unresolved situation of investment trust companies, as well as by relatively large bank trust accounts. Moreover, the absence of any regulatory or tax measures does not encourage further development of institutional investors. In contrast, in the Tokyo stock exchange, it appears that retail investors are not a dominant player, as they account for only 15 percent in the equity market and are negligible in the futures market. The dominant players are institutional or securities companies, which appears to be a common trend in other developed markets. Falling transaction costs have further encouraged retail trading, especially as brokerage rates are undercut by very competitive brokerage firms 11. The rapid acceptance of on-line trading had reduced transaction costs even further 12. Not only low costs, but easy availability through on-line trading made it possible for retail investors to access investment services more efficiently. On-line trading appears to be one of the main determinants of volume growth and, in-turn, volume growth provided more liquidity which induced further activity. The on-line trading ratio (online trading value/total stock market value) has risen substantially for both stock and futures trading. Low brokerage fees and quality information available on line lowers barriers for individual investors to enter into the market through on-line trading, providing liquidity in the market by increasing trading volumes, but to some extent it is also providing an opportunity for individual investors to engage in day-trading and to speculate on the information available which may contribute to the high volatility in the marketplace. 5.2 Foreign investors The role of foreign investors in developing countries has received mixed reviews, especially during the Asian crisis. Dornbusch and Park (1995) claim that foreigners engage in trading strategies that make stock prices overreact to changes in fundamentals; and Radelet and Sachs (1998) agree with this claim and believe this to be especially true during the Asian crisis 13. On the other hand, Choe, Kho and Stulz (CKS) argue that the foreign investors did not play a destabilizing role in the equity markets during the Asian crisis. Hence, the literature is inconclusive on whether foreigners destabilized markets or contributed to the Asian crisis. From the results of our regression, during the Asian crisis (period 2) the foreigners did have a significant impact on the volatility in the equity market. Whether the volatility destabilized the market is not our focus. However, we analyze which trader-type was contributing most volatility during this period. The difference between CKS study and ours is that the CKS study looks from the period Dec 1996 to Dec The second period extends from June 1997 to December To confirm that foreign investors were the leaders in the volatility during the Asian Crisis (period3), we formulate a lead-lag relationship between foreigners and the rest. This is done to show that the foreign investors were ahead of all other trader-types in selling the 11 The barrier in setting up a brokerage firm has lowered which would attract more brokerage firms to the market and add more competition amongst the brokerage firms.. 12 The commission fees for traditional securities trading are around 0.5 percent of trading amount while commissions for online trading are around 0.1 to percent of trading value. 13 Kim and Wei (1999) and Park and Song (1999) further support the view that the Asian crisis was further exacerbated by foreign investors. November 2002 Volatility of Korean markets page 12

13 market and in increasing volatility during the crisis. When we regress one day lead for foreigners, (table 9, column 2), foreigners expected volume has a significant impact on the volatility and the unexpected component significance disappears. This confirms our hypothesis that yesterday s surprise volume by foreigners is carried over to the next day in the marketplace as the expected volume related volatility. The magnitude of the impact on the volatility is the greatest for foreign investors (42.1) as compared to institutional (16.3) and retailers (4.44) who also have a significant impact on the volatility during that period. When two day lead is regressed for the foreigners the result is similar to one day lead. But from the third day, the significance disappears, and it is conjectured that foreigners were leading the volatility during the Asian crisis for up to two days, which were then followed by institutional investors. The foreigners do have a significant impact on volatility post-asian crisis, but the impact is not as great as that of securities companies. The trading activities by foreigners are closely monitored by most of the market participants in Korea since the Asian crisis and it is therefore believed that the foreigners are no longer the main contributor to volatility in the Korean securities market. 5.3 Securities firms Securities firms have most actively contributed to volatility in Korean securities markets since the Asian crisis. As table 8 reveals, only two investor types have significant and positive contributions to volatility since 1999: securities firms and foreign investors. The magnitude of the coefficients, however, indicates that the impact of securities firms is about twice as large as that of foreign investors. However, securities firms contribute only around 3 per cent by trading volume and 5 percent by trading value in the underlying market (table 10). Several explanations have been given for the powerful impact of securities firms, among which are program trading, managed accounts (heavy turnover of client money, possibly churning or front-running), and speculation with uncovered positions in futures and options markets. First, fluctuations in asset prices can be attributed not only to changes in economic fundamentals (or to changes in expectations about them) but to large professionallymanaged, speculative trading programs 14. This hypothesis is reinforced by the knowledge that a larger share of financial asset trading is done by a relatively few professionally managed institutions and that these institutions often pursue computer-guided trading strategies that have them all selling or buying at the same time. Possible explanations for securities companies having such a strong impact on volatility could be due to their speculative activity using program-guided trading. This observation would be supported by the rapid increase in program trading since 1998 (graphs 2 and 3) in both value of program trading and relative to the total equity trading value of the underlying market. The program trading relative to total equity trading value reached up to 25% during the sample period but averages around 5%. Unfortunately, program trading data for KOSPI 200 futures were not complete, hence a similar analysis could not be carried out, but graph 4 shows a similar pattern for the available 4ata. Most program trading occurs around the expiration date of Futures and Options the Triple Witching Hour. 14 Interest Rate Worries and Program Trading Send Stocks Plunging September 12,1986 The Wall Street Journal, November 2002 Volatility of Korean markets page 13

14 According to Lee (2002), the major group of index arbitrageurs are securities firms that are the KSE members, but institutional investors are not active in index arbitrage. From that analysis, over half of program trading originates from index arbitrage. For futures data, the program trading data is derived from either index arbitrage or hedging. On average, program trading by value for index arbitrage surpass the hedging activity by 91 times for sales and 112 for purchase of futures. Index arbitrage activity can be speculative depending on how one defines or computes the true price. The amount of arbitrage activity happening compared to hedging in the futures market suggests that the main activity for the futures market is not for hedging. If its core activity is not hedging, the bulk would be speculative. Since main index arbitrageurs are securities companies, one can deduce that they engage in significant program trading which may possibly explain their high contribution to volatility in the underlying market. It would be interesting to investigate further regarding the impact of program trading on the volatility of the Korean market, especially to investigate whether these induce speculative activity. The second prominent activity which may explain the volatility for securities companies is the extent of market penetration in the futures market. While securities companies trading in cash markets is around 4 percent, it is around 40 percent in the futures market 15 (table 10). Most proprietary profit making activities of securities companies occur in the derivatives market rather than in the cash market because of higher leverage 16. Most cash market revenues for the securities companies are obtained through commissions from clients. However, the derivatives market which started in 1996 is still not fully familiar for the retailers, and securities companies who are more professional in derivatives markets appear to take advantage and generate most of their revenues from derivative markets. It is also interesting to note that the underlying volatility is more pronounced with futures volume than the underlying volume (table 11). The magnitude of the coefficient shows the extent of the unexpected volume of futures significantly outweighs the unexpected volume of the underlying market. Furthermore, the expected futures volume has a negative coefficient. This implies that the expected futures volume does not add to volatility in the underlying market, but when some unexpected volume in futures markets is observed, the underlying market becomes more volatile. This may explain how securities companies, who have low presence in underlying market but high presence in the futures market, can have such a big impact on underlying volatility. 5.4 Open-interest positions Bessembinder and Seguin (1993) find that open interest positions in futures markets tend to decrease volatility in cash markets. Their volatility-volume relationships in the futures market for expected open interest as a part of activity variable is always negative and significant for all derivative products examined in their paper. A significant negative coefficient for the effect of expected open interest on volatility (conditional on trading volume) is consistent with the theory that expected open interest is related to the number of traders or amount of capital affiliated with a market. Secondly, these factors enhance 15 For the options market, securities companies contribute 20 percent by trading value and they are the second largest traders after retail investors in the options market. 16 Leverage is defined as value-at-risk over equity. November 2002 Volatility of Korean markets page 14

15 market depth and lower volatility shocks associated with a given volume in deeper markets. However, in our empirical analysis, the relationship between expected and unexpected open interest for the Korean market shows that the coefficient for expected open interest is in fact positive. This is contrary to previous literature on the relationship between open interest and volatility. This finding leads us to question the motive of traders that take open-interest positions. We find that securities companies are the major players in terms of transaction volumes and values and that they engage in about 10% of their trading in open-interest positions. During 2000 and 2001, securities companies have engaged in massive net short positions in futures and options trading. Table 12 shows that open interest of security houses increased from 3.5 million contracts net short position in 2000 to 7.9 million contracts net short in October The motivation may have been to generate commission revenues from retail investors, which want to build net long positions with maximum leverage. As securities companies have low trading activity in the cash market as compared to massive short-selling positions in derivative markets, it appears that they do not hedge positions but engage in arbitrage or speculation through uncovered positions. It would be interesting to clarify how securities firms can support such trades with their limited capital and whether margin requirements or other prudential requirements could be enhanced (i.e. daily reporting of capital adequacy, more rigorous calculation of market risk without using trade collars, etc.). In the literature, Lee finds that by looking at KOSPI 200 futures and cash markets during the period, that these markets are extremely inefficient. Both the degree of mis-pricing and the frequency of mis-pricing are significantly and do not diminish even when the markets have grown substantially. This suggests that the arbitrageurs exploiting the price difference to keep the market efficient are not doing their job well or arbitrageurs are minimal in the Korean market. Hence, it leaves one strong hypothesis that derivative markets are mainly driven by speculative motives. This would also explain why the expected open interest has a positive relationship with volatility and why open interest does not reflect the depth of the market but rather speculative activity. Titman and Wei (1999) find that so called excessive speculation of stock markets in Taiwan reflect the nature of volatile underlying fundamentals. Also, Chung and Lee (1998) find that the Korean market deviates substantially and persistently from fundamentals and interpret this as being a bubble. These findings all point to concerns regarding the structure and operation of Korean derivative markets. 6. CONCLUSIONS AND POLICY ISSUES Korean equity markets are among the most volatile markets in the world, and during the past five years had volatility ratios between 30% and 60%, about twice pre-1998 levels. Technological improvements (online trading), low transaction costs, as well as supportive regulatory policies were contributing factors to fast rising trading volumes which are mostly driven by retail investors (about 70%). However, empirical analysis reveals that securities houses have been the main contributors to market volatility. They conduct over 90% of their trading in futures and options markets, where on average 10% of their trading is providing open short positions, mostly to retailers. However, during periods of extreme market volatility (esp. in late 1997), it appears that foreign investors are leading November 2002 Volatility of Korean markets page 15