Abstract
This Selected Issues paper focuses on some of the key stylized facts of Korean business and export cycles over 1960–2001, and calculates a chronology for the classical cycle in these series by applying a variant of the Bry and Boschan (1971) cycle-doling algorithm. It highlights that the Korean classical business cycle and exports cycles are extremely asymmetric, as they exhibit long-lived expansions and much shorter-lived contractions. The results also indicate that the probability of ending a contraction or expansion phase in Korean industrial production and Korean real exports is independent of their duration.
IV. Linkages Between Domestic and International Asset Markets: The Korean Case1
This paper analyzes the role of global, regional and domestic factors in Korean asset markets, focusing on the equity, currency and external debt markets. The statistical analysis indicates a strong impact on Korean asset prices from global and regional financial variables and also from the net flows of foreign investors. The sign of the correlations invariably indicates that Korean assets perform well and inflows occur when asset prices elsewhere are rising, suggesting that Korean investments are viewed as a high-beta or cyclical ones. Yet Korean assets have fared relatively well in the current global slowdown, suggesting that strong linkages with global financial markets need not be destabilizing if the domestic economy and macroeconomic policies are sound.
A. Introduction
1. The increasing openness of the Korean economy has been accompanied by stronger linkages between domestic and international asset markets. This paper explores the nature of those linkages, using high-frequency (i.e., daily and weekly) data to focus on the way that information flows between markets and shocks are transmitted. The aim is not to test particular asset pricing models but to use regression analysis to draw inferences about the correlations between different assets and possible causal relationships. The analysis deals with the current linkages rather than historical ones, so the paper concentrates on the postcrisis period of January 1999 to November 2001.2 Thus the start date corresponds approximately to the date of the upgrade of Korea’s sovereign credit rating back to investment grade, and also to approximately one year after the liberalization of capital inflows into equity and bond markets. The focus is predominantly on the equity market, where data are best and—unlike the bond market—foreign participation is quite high, although there is also analysis of the foreign exchange market and of the pricing of Korea’s sovereign foreign currency debt.
2. The paper is organized as follows. Section B provides a brief description of the structure and recent trends in the Korean equity market. Section C provides econometric analysis of price determination in the equity market, focusing on the links with international markets, while Section D examines the role of foreign investors. In Section E, the currency market is studied. Section F contains some analysis of the pricing of Korea’s sovereign external bonds. Section G concludes.
B. The Korean Equity Market: Structure and Recent Trends3
3. There are two stock markets in Korea—the main-board Korea Stock Exchange, and the KOSDAQ market, which focuses on technology and venture companies. The market capitalization of the two markets were W 256 trillion and W 52 trillion, respectively, at the end of 2001, equivalent to a total market capitalization of S256 billion. In terms of its share in the most widely used emerging markets equity benchmark (the MSCIEMF index), the Korean stock market is the largest emerging market, accounting for 17.8 percent of that index in January 2002 (with a significant increase expected due to index methodology changes scheduled later in 2002).
4. The liberalization of foreign investment in the Korean equity market began in January 1992. Ceilings on aggregate foreign investment were gradually increased through the 1990s and completely removed in May 1998.4 As result of positive net inflows in every year since 1992, foreign ownership had reached about 37 percent of all KSE stocks at the end of 2001, with a large majority of this having occurred via portfolio investment rather than through foreign direct investment (Figure IV.1). Foreign holdings are concentrated in the larger, blue-chip stocks, and foreign holdings of companies such as Samsung Electronics and POSCO are now over 50 percent. These levels of foreign ownership are higher than for many other emerging markets, but are not out of line with foreign ownership levels in many other medium-sized industrial countries.
Foreign Investment in the Korea Stock Exchange (As percent of total market capitalization)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Foreign Investment in the Korea Stock Exchange (As percent of total market capitalization)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Foreign Investment in the Korea Stock Exchange (As percent of total market capitalization)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
5. Trading on the KSE occurs in opening and closing auctions, as well as continually between these auctions. Trading has been fully computerized since September 1997. “Online” trading (mainly via the internet, but also through other systems that transmit orders directly from the investor to the trading system) now accounts for over 60 percent of all trades in Korea, the highest such ratio in the world. The share of on-line trading is highest on the KOSDAQ—at nearly 80 percent, versus about 50 percent on the KSE—where the role of individual investors is greatest. Day trading is also high by international standards. Indeed, market turnover is dominated by household trading. Although households hold only about 20 percent of KSE stocks, they account for over 70 percent of trading. By contrast, foreign investors’ share trading is far smaller (at about 11 percent) than their ownership share.
6. Total annual turnover, at about 230 percent of KSE market capitalization in 2000, is quite high by international standards. Explicit transactions costs are very low in the Korean market, given low brokerage commissions, especially for on-line trading. Implicit transactions costs (which reflect the market impact associated with orders) are also low for the larger stocks, given the high level of turnover and availability of alternative means of transactions via ADRs and GDRs in foreign markets.5 By one estimate, total trading costs are smaller for large Korean equities than in any other emerging market.6
7. Prices in the Korean market saw a healthy recovery in 2001, but on average remain far below precrisis levels (Figure IV.2). Indeed, the Korean market had begun to weaken long before the onset of the Asian crisis. At its post-crisis low in June 1998, the KOSPI index was 75 percent below its late-1994 peak. Prices then picked up significantly in the second half of 1998 and in the technology-led boom of 1999, hitting a peak at the start of January 2000. The global technology shakeout of 2000 then saw the KOSPI fall 51 percent, and the KOSDAQ fall nearly 80 percent. The KOSPI showed little direction through the first eight months of 2001 and fell 12 percent on September 12, the day after the terrorist attacks in the United States. However, a strong rally followed in the last three months of the year, and for the year as a whole the KOSPI and KOSDAQ were each up 37 percent. Both these gains were substantially larger than the average 4 percent gain seen in the MSCI Emerging Asia index.
Korean Stock Prices
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Korean Stock Prices
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Korean Stock Prices
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
8. The average profitability of Korean companies remains fairly weak. Although a few companies (e.g., Samsung Electronics, Hyundai Motor and POSCO) are sometimes mentioned by analysts as globally competitive companies, average profitability remains low, due in part to the high interest burden of many Korean companies. Despite substantial corporate and financial restructuring after the 1997 crisis, corporate governance is still perceived to be poor.7 In particular, there is investor mistrust of the complex crossholdings of Korean chaebol, the poor corporate governance practices of many companies, and the meager dividend payments. This has shown up in a “Korea discount” whereby a higher discount rate (i.e., a lower price-earning ratio) is applied to Korean companies than companies in other advanced Asian economies (Table IV. 1).8
Price-Earnings Ratios
(Price/Forecast 2002 Earnings, January 2002)
Price-Earnings Ratios
(Price/Forecast 2002 Earnings, January 2002)
| Korea | 10.8 |
| Hong Kong | 14.9 |
| Singapore | 22.9 |
| MSCI Emerging Markets Asia | 18.1 |
| MSCI All Countries | 20.1 |
Price-Earnings Ratios
(Price/Forecast 2002 Earnings, January 2002)
| Korea | 10.8 |
| Hong Kong | 14.9 |
| Singapore | 22.9 |
| MSCI Emerging Markets Asia | 18.1 |
| MSCI All Countries | 20.1 |
C. The Korean Equity Market—Statistical Analysis
Introductory Analysis
9. In this section, the focus is on examining the “where”, “when” and “how” of price discovery in the Korean equity market, and the factors that appear to explain short-term price movements. In particular, Korean price changes are regressed on price changes in other national stock markets to explore the extent to which Korean prices appear to be explained by developments in other countries. The analysis begins with regressions of overnight returns, intraday returns, and overall daily returns (i.e., the sum of overnight and intraday returns) on returns in U.S. stock markets and regional stock markets.9
10. Regressions of overnight Korean returns on overnight U.S. stock returns illustrate the importance of U.S. markets for the Korean market (Table IV.2). A simple regression of the overnight KOSPI return on the previous day’s returns on the Dow Jones Industrial index, the S&P500 index, the Nasdaq composite index, and the Philadelphia Semiconductor index10 shows an adjusted R-squared of about 0.56 (i.e., 56 percent of the variance in the overnight KOSPI return is explained by U.S. returns). But overnight returns on other markets in the Asian timezone also have significant explanatory power, with an R-squared nearly as high. However, this correlation with other Asian markets appears to largely represent a common response to the previous day’s returns in U.S. (and other international) markets. In particular, when both U.S. and Asian timezone returns are included as explanatory variables, the adjusted R-squared rises only slightly above the level obtained with only U.S. returns. Still with an R-squared of 0.59 from a regression of the overnight KOSPI return on returns in other national markets, it is clear that the overnight return in Korea is very substantially affected by international markets—especially the U.S. equity market.
Regressions of Korean Returns on Regional and U.S. Returns
Regressions of Korean Returns on Regional and U.S. Returns
| Adjusted R-squared | |
|---|---|
| Kospi overnight change on previous day’s US returns | 0.555 |
| Kospi overnight change on overnight Asian returns | 0.525 |
| Kospi overnight change on US and Asian returns | 0.590 |
| Kospi daytime change on previous day’s US returns | 0.006 |
| Kospi daytime change on daytime Asian returns | 0.185 |
| Kospi daytime change on US and Asian returns | 0.208 |
| Kospi daily change on previous day’s US returns | 0.150 |
| Kospi daily change on same day Asian returns | 0.283 |
| Kospi daily change on US and Asian returns | 0.302 |
Regressions of Korean Returns on Regional and U.S. Returns
| Adjusted R-squared | |
|---|---|
| Kospi overnight change on previous day’s US returns | 0.555 |
| Kospi overnight change on overnight Asian returns | 0.525 |
| Kospi overnight change on US and Asian returns | 0.590 |
| Kospi daytime change on previous day’s US returns | 0.006 |
| Kospi daytime change on daytime Asian returns | 0.185 |
| Kospi daytime change on US and Asian returns | 0.208 |
| Kospi daily change on previous day’s US returns | 0.150 |
| Kospi daily change on same day Asian returns | 0.283 |
| Kospi daily change on US and Asian returns | 0.302 |
11. By contrast, external factors—either U.S. or Asian in origin—explain a far smaller fraction of intraday Korean equity returns. As would be expected, the previous day’s U.S. returns have essentially zero explanatory power for intraday Korean returns, with their full impact already felt in opening prices. Intraday returns in other Asian countries have some modest explanatory power for intraday Korean returns, but the adjusted R-squared of 0.17 suggests that domestic Korean factors and company-specific news instead explains most of the variation in intraday returns.
12. Regressions of daily returns—which are the sum of overnight and intraday returns—provide an overall measure of the importance of external factors. The previous day’s returns in the U.S. explain about 15 percent of the variance in daily Korean returns, whereas Asian returns appear to explain a higher proportion (28 percent) of the variance. However, nearly half of this impact would appear to be the indirect impact of U.S. returns. Overall, the previous day’s returns in the U.S. and the daily returns in five other Asian countries explain about 30 percent of the variance in daily returns. While this proportion is significant, it still implies a major role for domestic influences (or for global factors that are not captured in the returns indices used as proxies for external factors).11
13. The role of domestic factors can be assessed by including some domestic variables in a regression of daily KOSPI returns. Given the large number of possible explanatory variables, a stepwise procedure was followed to determine which variables are most correlated with Korean returns.12 Variables were added to the equation based on their marginal significance level, with the goal of finding a parsimonious equation with plausible parameter estimates that explained the daily return on the KOSPI.13 The resulting equation—which should be thought of as a statistical model rather than a behavioral one—was as follows:
Adjusted R-squared= 0.377, Number of observations = 529 where d Variable represents the log differenced change in the variable, with the exception of the domestic bond rate where it represents the basis point change. T-statistics are shown in parentheses, and—by the design at the stepwise regression procedure—all variables are highly significant.
14. The results can be interpreted as follows:
Two domestic variables are strongly associated with the daily return on the KOSPI. In the case of the exchange rate (defined as won per dollar), the coefficient indicates that a one percentage point appreciation of the won is associated with a 0.81 percentage point increase in die KOSPI. This suggests that flow effects (e.g., foreign purchases leading to increases in stock prices and won appreciation) or common “sentiment effects” are more important in the short-run than any competitiveness effects (whereby export stocks would weaken as the exchange rate appreciated). In the case of the bond market, the coefficient indicates that a 10 basis point fall in yields is associated with a 0.3 percent increase in stock prices. Since foreign investors are not particularly active in the bond market, this is unlikely to be due to flow effects. Instead, it is likely due to common sentiment or required return effects (falls in required rates of return on Korean assets boosting both bond and equity markets). Further, this effect must be more important than any substitution effects (negative correlations from investors switching from one asset to the other) or any effect from monetary policy expectations (whereby increases in stock prices lead to expectations of higher interest rates).
The four foreign equity indices that are jointly most correlated with the Korean market are the benchmark indices for Singapore, Hong Kong and Japan, as well as the prices of U.S. semiconductor stocks. The fact that broader U.S. indices are not included does not mean that these are not important influences on Korean stock prices, rather that the impact of movements in these broader indices shows up (along with some Asia-specific factors) in correlations with Singapore, Hong Kong and Japanese stock prices. However, given the presence of two major semiconductor manufacturers (Samsung Electronics and Hynix) in the KOSPI, and many other computer related stocks, it is not surprising that U.S. semiconductor stocks also have some additional explanatory power.
D. The Role of Foreign Investors in the Equity Market
Possible Channels of Causation Between Foreign Inflows and Prices
15. In considering the influence of external factors on Korean equity prices, the question arises as to whether net purchases and sales by foreign investors also have an independent impact on equity prices. This might not be surprising in light of evidence (e.g., Tesar and Warner (1994)) from monthly or quarterly data that foreign inflows and domestic equity returns are positively correlated for a range of countries. Indeed, a regression of monthly KOSPI returns on monthly net purchases of KSE stocks by foreigners also indicates significant correlation:14
Adjusted R-squared= 0.248, Number of observations = 36
16. The apparent very strong positive correlation between net inflows and Korean returns might reflect a number of different factors:15
• Feedback trading. Foreigners might increase their holdings following price increases. In this case, market returns could be driving inflows rather than vice versa, but it might not be possible to identify the exact causality without high frequency data.
• Price pressures from permanent changes in demand. If the demand curve for stocks is downward sloping (rather than flat as traditionally assumed—with prices purely determined by fundamentals and not demand and supply), then foreign inflows represent an outward shift in the demand curve summed over all investors and should result in higher prices. This may be related to what Clark and Berko (1997) call the “base-broadening effect” of higher stock from increasing foreign investor participation in emerging stock markets.16
• Temporary price pressures. If there is temporary illiquidity in the market that results in a temporarily-downward-sloping demand curve, purchases by foreign investors may drive up prices in the short run. However, once portfolios of other investors have readjusted, initial price effects might be reversed. In this case, the correlation between flows and returns would decline as the horizon of the return measurement period increased.
• Information revelation. If foreigners have more information relevant to the pricing of domestic assets than domestic investors, this information may be revealed through their trading and contribute to price determination.17 In this case the correlation between inflows and returns would reflect the market reacting to the information held by foreign investors.
• Omitted variables. Net inflows and price increases could both be responding to some other variables, and the positive correlation might be due to a failure to control for these variables.
17. Fortunately, detailed Korean trading data are available and allow analysis of the relationship between net inflows and returns to shed light on the relative importance of the above factors. In particular, foreign investors must register with the Financial Supervisory Service, a requirement that originated due to the earlier limits on foreign holdings of Korean equities. Indeed, for each equity trade, the investor group (foreign or domestic, with domestic divided up into seven categories) of the buyer and seller is recorded by the Korea Stock Exchange. Hence Korea has a rich database for exploring the impact of trading by different groups on market prices, and data are available at daily (or even higher) frequency, allowing quite precise tests of some of the possible explanations above.18 Accordingly, daily data are used in the remainder of this section to examine both the factors that influence foreign inflows and their impact on Korean asset prices. It is noteworthy that this data captures the trading of all foreign investors, as opposed to the data used in some previous studies which includes only one class of investors (e.g., only U.S. investors, or only mutual funds).19
18. A simple way to begin the analysis of the role of foreign investors would be to compare average KOSPI returns on days when foreigners were net buyers and on days when they were net sellers. Sorting the data from January 1999 to November 2000 reveals large differences in returns in these two groups of days. In particular, days when foreigners are net purchasers have average daily returns of 0.56 percent, and days of net foreign sales have average returns of -0.81 percent. The difference between these is strongly statistically significant.20 Furthermore, a regression of returns on net inflows (as a fraction of market capitalization) yields a regression coefficient with a t-statistic of 9.8. This correlation would appear to be far stronger than any previous empirical evidence on the relationship between returns and the net purchases of any particular investor group.
19. However, net inflows into Korea are also strongly correlated with returns in other Asian markets. For example, regressions of returns in Tokyo, Hong Kong, Singapore, Sydney, and Taipei on Korean net inflows with regression t-statistics ranging from 4.8 to 7.5. The reason is presumably not due to any causal influence from net flows into Korea, but instead because Korean inflows are correlated with the previous night’s return on benchmark U.S. indices (with t-statistics around 10) and Asian markets also respond to the previous day’s U.S. return. This suggests that a simple correlation between net foreign purchases and Korean returns may overstate the true impact of flows on prices. This highlights the need to (i) understand what drives net inflows into Korea; and (ii) control for other influences on Korean stock prices.
What Explains Foreign Inflows?
20. Daily data allow a precise analysis of the determinants of foreign inflows. In particular, if net purchases by foreigners (or any other group of investors) respond systematically to recent returns, daily data should be able to capture these linkages.21 Of course, nonresidents are not the only participants in the Korean market and if one finds that foreigners typically are buyers following a certain type of information then it follows that domestic residents in aggregate must be sellers in response to the same information. Accordingly, it will only be possible to identify which of the two groups has the dominant role in responding to the information if the trading behavior is accompanied by price effects.
21. Correlations between net purchases of different investor groups may give some preliminary information about the trading behavior of different groups. Although the “adding-up constraint” implies a perfect negative correlation between any group’s net purchases and the net purchases of the rest of the market as a group, it is possible that the net purchases of individual subgroups could actually be positively correlated if they share similar trading patterns. In the case of Korea, the net purchases of retail investors (which account for 74 percent of all trading in the sample period) are indeed negatively correlated with net purchases by foreigners and by domestic institutions. However, the net purchases of the latter two groups are also negatively correlated, indicating that there is no close correspondence between their trading patterns.
22. To better understand the trading patterns of foreigners and the two domestic investor groups, data on daily net purchases as a percent of market capitalization were regressed on returns on a range of different assets over the previous two days. Since net purchases are highly autocorrelated, lagged value of net purchases were also included as potential regressors. Again, a stepwise regression procedure was used to arrive at parsimonious regressions that appear to characterize trading behavior. The results are shown in Table IV.3.
What Drives Net Purchases?
What Drives Net Purchases?
| Net Purchases of | |||
|---|---|---|---|
| Explanatory Variables | Foreigners | Domestic Institutions | Domestic Households |
| Lagged net purchases (t-1) | 0.232 | 0.157 | 0.226 |
| (6.3) | (3.9) | (5.8) | |
| Lagged net purchases (t-2) | 0.120 | 0.098 | |
| (3.7) | (2.5) | ||
| Lagged KOSPI return (t-1) | 0.281 | −0.23 | |
| (3.8) | (2.6) | ||
| Lagged U.S. returns | dNasdaq(−l) | dNasdaq (−1) | dPhil. Semic. (−1) |
| 0.321 | −0.27 | −0.247 | |
| (3.1) | (3.6) | (3.7) | |
| dPhil. Semic. (−1) | dDow (−2) | dS&P500 (−1) | |
| 0.348 | 0.451 | −0.407 | |
| (5.1) | (3.0) | (2.1) | |
| dS&P500 (−2) | |||
| −0.741 | |||
| (4.7) | |||
| Lagged regional returns | dHang Seng(−l) | ||
| 0.334 | |||
| (3.4) | |||
| Other variables | |||
| Adjusted R-squared | 0.412 | 0.065 | 0.224 |
| Number of observations | 580 | 587 | 587 |
What Drives Net Purchases?
| Net Purchases of | |||
|---|---|---|---|
| Explanatory Variables | Foreigners | Domestic Institutions | Domestic Households |
| Lagged net purchases (t-1) | 0.232 | 0.157 | 0.226 |
| (6.3) | (3.9) | (5.8) | |
| Lagged net purchases (t-2) | 0.120 | 0.098 | |
| (3.7) | (2.5) | ||
| Lagged KOSPI return (t-1) | 0.281 | −0.23 | |
| (3.8) | (2.6) | ||
| Lagged U.S. returns | dNasdaq(−l) | dNasdaq (−1) | dPhil. Semic. (−1) |
| 0.321 | −0.27 | −0.247 | |
| (3.1) | (3.6) | (3.7) | |
| dPhil. Semic. (−1) | dDow (−2) | dS&P500 (−1) | |
| 0.348 | 0.451 | −0.407 | |
| (5.1) | (3.0) | (2.1) | |
| dS&P500 (−2) | |||
| −0.741 | |||
| (4.7) | |||
| Lagged regional returns | dHang Seng(−l) | ||
| 0.334 | |||
| (3.4) | |||
| Other variables | |||
| Adjusted R-squared | 0.412 | 0.065 | 0.224 |
| Number of observations | 580 | 587 | 587 |
23. The results of the flows regressions indicate that the net purchases of foreign investors can be well explained by just a few variables. A regression on net purchases on the two previous days yields an adjusted R-squared of about 0.17, illustrating the (positive) autocorrelation in net flows. However, flows also appear to respond positively to overnight returns in U.S. equity markets (especially in the Nasdaq and Philadelphia Semiconductor indices) and to the previous day’s returns in domestic and regional markets (proxied by the KOSPI and Hang Seng indices). The adjusted R-squared of this augmented equation is a remarkable 0.41, and the correlations with all the returns variables are positive. Hence there is strong evidence that foreign investors have been “positive feedback” or momentum investors at the aggregate level, tending to buy immediately following good news in Korean, regional, and U.S. markets.22
24. The regressions for net purchases of Korean households and institutions also show some significant responses to the previous day’s returns. Not surprisingly—given the adding up constraint—net purchases by households show the opposite tendency to foreign flows. Households tend to be net sellers in response to price increases on the previous day in Korea or the United States. There is also evidence that they tend to continue to sell on the second day after U.S. price increases. Thus, their trading pattern can be characterized as contrarian with respect to short-run returns. The equation for net purchases by institutional investors shows a much lower degree of explanatory power than the other two groups, but suggests that institutions tend to sell (by implication to foreigners) following price increases in the U.S. market on the previous day, but to buy (by implication from households) following U.S. increases two days earlier.
25. The above results appear consistent with research into other markets. For example, Grinblatt and Keloharju (2000) find that foreign investors and sophisticated domestic institutional investors tend to be momentum investors in the Finnish market, whereas households and less sophisticated institutions tend to contrarians. Similarly, Bae, Ito and Yamada (2001) show using weekly data that foreign investors in the Japanese market can be characterized as momentum players while domestic investors are contrarians. And Goetzmann and Massa (1999) provide evidence on U.S. mutual find investors that contrarian traders tend to trade more frequently than momentum traders (consistent with evidence that Korean households trade far more actively than foreigners).
The Price Impact of Foreign Flows
26. To assess the price impact—if any—of foreign flows, net purchases by foreigners (as a percent of total market capitalization) are included as an additional regressor to the KOSPI equation shown above in paragraph 13. The resulting equation is as follows:
Adjusted R-squared = 0.393, Number of observations = 529.
27. The foreign flows variable is highly significant, and its coefficient implies that net purchases equivalent to one percent of market capitalization would be associated with an increase of 7 percent to the KOSPI. It is noteworthy that this is a substantially smaller increase than the 18 percent increase that would be suggested by a regression of daily KOSPI returns on net purchases with no other explanatory variables.23 Given that flows are driven partly by foreign returns, this indicates that foreign stock returns have both a direct impact on domestic returns and an indirect one (via flows).
28. The price impact of net purchases by Korean households and institutions can also be examined by including their net purchases in separate regressions of KOSPI returns. The regression coefficient for institutions indicates that net purchases equivalent to 1 percent of market capitalization are associated with a 5 percent increase in the KOSPI (with a t-statistic of 3.0). By contrast, the regression coefficient for households indicates that net purchases equivalent to 1 percent of market capitalization are associated with a 10 percent fall in the KOSPI (t-statistic of 6.4).
29. The adding-up constraint implies that if net purchases of some groups are associated with price increases, then net sales by other groups must not have a countervailing impact. As discussed by Zheng (1999), if net purchases by one group and net sales by another are associated with price increases, then it is reasonable to conclude that the former group is tending to initiate the trades by shifting its demand curve, whereas the latter group is more passively responding by shifting along its demand curve. In the Korean case, it appears that net purchases by foreigners (and institutions to a lesser extent) are associated with price increases, but—somewhat counterintuitively—that net sales by households are associated with price increases. Hence, from the price activity that accompanies their flows, we might conclude that it is the foreign investors rather than the household investors that have an impact on prices in Korea through their trading. Further, based on the earlier regressions looking at the determinants of flows, it would seem that the feedback tendencies observed with respect to the previous day’s price movements are more a reflection of active momentum investing by foreigners and that the apparent contrarian investing by Korean households is somewhat more passive.24
30. However, it is possible that price changes seen on the day of changes in foreign flows might be temporary. Alternatively, Froot et al. (2001) suggest that foreign investor flows may have further ongoing impacts on prices for a month or more. A number of different regressions were run to shed light on this, with little evidence from daily data to suggest either price reversals or continuations. Further, regressions using weekly data also show no impact on prices beyond the week of the flows, suggesting that price impacts are reasonably permanent but that flows do not have ongoing impact on prices.
Summing up and a Comparison with Results for Other Asian Countries
31. Based on the above results, some tentative conclusions might be reached about the factors that might possibly explain the correlation of returns and inflows:
Given the strong evidence that daily flows are positively correlated with the previous day’s returns in Korea and regional and U.S. markets, part of the correlation that is observed in lower frequency (e.g., monthly data) is presumably from feedback trading. Such feedback trading is consistent with the notion that investment in Korea over this period has tended to occur when risk tolerance is increasing and foreign investors are feeling optimistic, perhaps due to wealth effects. In some senses, foreign investment in Korea may have been viewed as a cyclical or high-beta play.25 The possibility that the same-day correlation between returns and inflows could partly reflect intraday feedback trading has not been explicitly tested. However, this seems to be unlikely as many of the trading decisions of foreign investors are likely to have been made in their home markets the previous day. Further, if there is intraday feedback trading it presumably occurs mostly after the opening batch auction. Yet regressions of overnight returns (not shown here) indicate that daily net purchases by foreigners are indeed significant explanators of returns in the opening batch auction. This indicates that at least a significant portion of the daily correlation between returns and flows is unlikely to be due to intraday feedback trading.26
The analysis with daily data (and indeed with monthly data) indicates that the simple bivariate correlation is substantially reduced after controlling for information in other equity markets or in other Korean asset classes. Indeed, the possibility that the price impact of flows might be further reduced if additional control variables were available—e.g., for specific corporate news—cannot be ruled out.
Analysis with daily and weekly data provides no evidence to support the notion that the observed price impact is temporary due to price pressures that are subsequently unwound. However, if reversals occur very slowly over an extended period of time, they are unlikely to be captured by regressions using daily or even weekly data, so some unwinding of price pressures cannot be ruled out.
The regressions above have provided no particular test of the notion that the positive correlation between inflows and returns reflects superior information of foreigners that is revealed through the trading of foreign investors and reflected in prices. However, this seems unlikely, given the strong evidence in Choe et al. (2001) that the stock-level trading of foreign investors yields no evidence of them having an informational advantage over domestic investors.
This leaves increased demand associated with foreign inflows as the most likely candidate for explaining the price impact of trading that remains after controlling for other information. This is consistent with the substantial literature that prices rise due to increased demand when stocks are included in indices that are widely used as benchmarks for index funds (see, e.g., Morck and Yang (2001)). Interestingly, the estimated price impact for Korea (7 percent for inflows equivalent to one percent of market capitalization, or 11 percent for unexpected inflows) is quite similar to the results of Clark and Berko (1997) who find an impact of 8 percent for surprise inflows into Mexico using monthly data which they attribute also to increased demand (or base broadening). By contrast, both estimates are far lower than the average 40 percent impact for emerging markets suggested by Froot et al. (2001), whose surprisingly high estimate may be due in part to the failure to account for other variables (e.g., U.S. or regional stock market returns) that are correlated with flows and domestic returns.27
32. However, the overall return performance of the Korean equity market over the sample period would suggest only a very modest price impact from foreign inflows. In particular, total daily net inflows in this period were equivalent to about 8 percentage points of KSE market capitalization. Yet Korean stock prices grew by only about 10 percent over the sample period, little different to or only modestly stronger than price growth in major U.S. and regional indices. Hence, the overall price performance of the Korean market is difficult to reconcile with the observed net foreign inflows and global market trends, unless the price impact of net inflows is no more than about one or two percent per percentage point of market capitalization. The price impact that is obtained from monthly regressions is admitted closer to this than the price impact from daily or weekly data (5 percent versus 7–9 percent). Still, the magnitude of the price impact remains something of an open question, and is addressed further using vector autoregression analysis in Richards (2002).
33. Since detailed investor data are available for a number of other regional markets, it is of interest to ask if the results from Korea are also seen in those markets. Accordingly, similar analysis was applied to daily foreign inflows data for Thailand, the Philippines, Indonesia and Taiwan Province of China. It is noteworthy, however, that net foreign purchases and sales are typically somewhat smaller in these four other markets than in Korea.
34. The analysis for these other countries indicates that most of the Korean results are indeed also observed in these other markets.
Net inflows are positively autocorrelated, and also positively correlated with the previous day’s local market return and with recent returns in either the U.S. market or major regional markets (notably Singapore). In the case of Thailand, where data for domestic investors are separated into household and institutional investors, household investors are contrarian with respect to recent returns—as in Korea—while institutions are difficult to categorize.
Similarly, the correlation between daily net inflows and daily returns is significant, but falls once control variables such as returns on regional or U.S. markets are added. Nonetheless, net inflows (both total and unexpected) are significantly correlated with returns, with regression coefficients that are similar or larger than the estimates for Korea.28 There is again no evidence of reversals, but some evidence for flows to be associated with continued (modest) increases in prices in the week after they occur. As is the case with Korea, estimates of the price impact tend to increase when moving from daily to weekly data, but fall when monthly data are used. Overall, increased demand seems to remain the most likely candidate for the observed correlations.29
E. The Currency Market
Background
35. The exchange rate for the won is freely floating, with minimal intervention by the Bank of Korea. Until late 1997, there were limits on the daily movement in the value of the currency, but these were abolished as part of the liberalization of financial markets in response to the crisis. At its lowest point of W 1,810 per dollar at the start of early 1998, the won had depreciated more than 50 percent from its precrisis level. However, as the economy recovered the won appreciated fairly steadily against the dollar, reaching a peak at W 1,104 per dollar in early September 2000. Since then the exchange rate has weakened somewhat. The currency traded around W 1,300 for much of 2001, with periods of weakness coinciding with depreciations of the Japanese yen and periods of strength corresponding to strong inflows into the equity market.
36. Currency trading in the Seoul market is primarily through two interbank brokers between the hours of 9:30 a.m. and 4:30 p.m., although there is also some OTC trading during this period, as well as before and after the period in which the brokers operate. The vast majority of trading occurs in the Seoul market, with little trading elsewhere in the region or overnight in London or New York. The modest amount of trading that does occur in these latter markets is mainly in the form of nondeliverable forward contracts (i.e., contracts with payouts based on the future value of the won, settled in dollars).
37. The 2001 triennial Survey of Foreign Exchange Activity trading indicated that the won was the 15th most traded currency in April 2001, accounting for 0.4 percent of global turnover, up from 0.2 percent in 1998. After a sharp fall in trading volumes during the 1997–98 crisis, trading has now recovered to above its precrisis level, even though global foreign exchange trading has contracted over the same period (Table IV.4). Despite its recent growth, the market for the won is still relatively undeveloped—certainly compared with the size of the economy, the 13th largest in the world.
Foreign Exchange Turnover (Average of daily spot transactions, billions of U.S. dollars)
Foreign Exchange Turnover (Average of daily spot transactions, billions of U.S. dollars)
| Won | Global Trading, all Currencies | |
|---|---|---|
| 1995 | 4.3 | 494 |
| 1996 | 4.1 | n.a. |
| 1997 | 5.6 | n.a. |
| 1998 | 2.8 | 568 |
| 1999 | 4.3 | n.a. |
| 2000 | 5.5 | n.a. |
| Est. 2001 | 6.3 | 387 |
Foreign Exchange Turnover (Average of daily spot transactions, billions of U.S. dollars)
| Won | Global Trading, all Currencies | |
|---|---|---|
| 1995 | 4.3 | 494 |
| 1996 | 4.1 | n.a. |
| 1997 | 5.6 | n.a. |
| 1998 | 2.8 | 568 |
| 1999 | 4.3 | n.a. |
| 2000 | 5.5 | n.a. |
| Est. 2001 | 6.3 | 387 |
Statistical Analysis
38. The volatility of the won is relatively low by international standards. The standard deviation of daily price changes has been about 0.42 percent over the last three years, with no obvious trend. A preliminary sense of where price determination takes place in the won can be obtained by examining the relative variability of intraday and overnight changes. These indicate that the majority of price discovery takes place in the seven hours of Seoul trading (9:30 a.m. to 4:30 p.m.) rather than in the 17-hour “offshore” period. (A closer examination of the data also reveals some significant negative correlation between the overnight change and the subsequent intraday change—this will be addressed below). That is, most price discovery takes place during the hours that the home market is open. By contrast, price discovery in more globally traded currencies occurs at a fairly constant rate around the clock and a deep liquid market exists throughout the 24-hour trading day.
39. As was done for the equity returns data, stepwise regressions are used to examine the factors that appear to be associated with changes in the value of the won.30 The regression that is obtained is as follows:
Adjusted R-squared = 0.247, Number of observations = 603.
40. The regression indicates that about one quarter of the variance of the daily exchange rate can be explained by just a few domestic and foreign financial variables. Indeed the explanatory power is perhaps surprisingly high given the notoriously poor fit of monthly or quarterly exchange rate equations using macroeconomic variables. The coefficients can be interpreted as follows:
The daily movement of the won against the dollar in Seoul is significantly correlated with the daily movement of the yen against the dollar in Tokyo. (By contrast, the correlation with the euro is much smaller and not statistically significant.) Over the full 1999–2001 sample, a one percent appreciation of the yen is associated with an appreciation of the won against the dollar of 0.18 percent. A slightly larger elasticity (0.23) is obtained from regressions using weekly changes. There is also some evidence that this elasticity has risen over time, to about 0.45 based on data just for 2001. Still, it is also worth noting that correlations of less than 0.5 imply that the won moves more in line with the dollar than with the yen. Indeed, the standard deviation of the won-dollar rate is lower than that of the won-yen rate.31
Appreciations of the won are associated with positive returns on Korean, regional and (especially technology-related) U.S. markets. This is something of a puzzle, since theory provides no guidance why a currency’s movement should be correlated with global stock returns, and clearly not all currencies can be positively correlated with global returns. Still, the finding that the Korean currency appreciates at times when global stock prices are rising suggests some link between investment in the won and the risk appetite of international investors. The won in some ways may be viewed as a high-beta, or cyclical rather than defensive asset.
The won also tends to appreciate on days of net purchases by foreigners on the KSE. This correlation presumably reflects both flow effects (foreign investors buying won for equity purchases) as well as more general sentiment effects involving both domestic and foreign investors.
Appreciations of the won are also associated with falls in bond yields. Given that nonresidents are not particularly active in the domestic bond market, this is likely to reflect common sentiment effects rather than flow effects.
41. The coefficient on the yen variable is of interest in light of the weakening of the yen in late 2001 and market expectations that it may weaken further. The elasticities obtained above—0.20 to 0.45—are generally larger than the direct share of Japan in Korea’s imports and exports (about 0.16). Thus, they are somewhat larger than would be implied if the Korean exchange rate moved simply to hold constant the nominal trade-weighted effective rate following movements in the yen exchange rate. However, other Asian countries may also tend to depreciate in cases of yen weakness, and an elasticity in the estimated range from market driven movements would not seem implausible as a rough measure of how much the Korean exchange rate would have to move to maintain an unchanged effective exchange rate in the face of yen weakness (against the dollar and euro) that was also accompanied by some degree of depreciation of other Asian currencies.
42. Regressions of overnight and intraday changes in the won provide some further information on its response to information. Not surprisingly—and providing some confirmation that the correlations are not spurious—the correlations with the yen and the KOSPI are also seen with respect to their respective overnight and intraday movements. By contrast, the impact of U.S. stock prices is—as expected—fully reflected in the overnight won movement. Interestingly, the impact of foreign inflows is seen mainly on the overnight return, which suggests that orders for won to fund purchases of Korean stocks are placed overnight, or within the first half hour of KSE trading, which is further evidence that the correlation between foreign inflows and the KOSPI is due in large part to price pressures and not intraday feedback trading. Interestingly, the residual from the overnight equation adds significant explanatory power (with a t-statistic of 8.7) to the regression of the intraday exchange rate change, with a negative sign. This may be consistent with exchange rate movements in the illiquid overnight ndf market being partly reversed once liquidity is restored in the more liquid Seoul daytime market.
F. External Debt
Background
43. The April 2003 and April 2008 global bonds issued by the Republic of Korea in April 1998 are the benchmark external securities for Korea. In addition, there is also substantial external issuance by quasi-government institutions (especially the Korea Development Bank), and some issuance by banks and corporations. The sovereign global bonds are included in JPMorgan Chase’s EMBI Plus index and represent about 3 percent of the index. An additional six KDB issues, two Export Import Bank issues, and one Hanvit Bank issue qualify for inclusion in the broader EMBI Global index, and Korean issues account for 5.6 percent of the market capitalization of that index.
44. Korea’s sovereign rating has been gradually upgraded since the crisis. After savage cuts to its precrisis investment grade rating (AA from Standard and Poor’s and Al from Moody’s), Korea was upgraded back to investment grade in February 1999 by the two major agencies. Subsequent upgrades (most recently by Standard and Poor’s in November 2001) have seen its sovereign rating rise to BBB+ (Standard and Poor’s) and Baa2 (Moody’s). As such, Korea is one of the highest rated emerging market countries. The analysis that follows focuses on period following the upgrade back to investment grade status in February 1999, during which there has been a substantial fall in Korea’s sovereign spread.
Statistical Analysis
45. As in the earlier sections, a stepwise regression technique was used to identify factors that are correlated with the change in Korea’s sovereign spread, as measured by the spread on the Korean EMBI+ index subcomponent in basis points.32 33 Since one of the explanatory variables—the change in the yield spread on U.S. corporate debt may be subject to measurement error due to illiquidity, the regressions use returns measured over a five-day interval.34 The resulting regression was as follows:
Adjusted R-squared=0.430 Number of observations = 571
46. The equation explains nearly half of the variance in weekly changes in Korea’s sovereign spreads. The coefficients can be interpreted as follows:
Not surprisingly, the Korean sovereign spread is substantially correlated with spreads on other BBB-rated countries, which alone explain nearly 30 percent of the variance in Korea’s spread (see also Figure IV.3). The spread on the overall EMBI is less correlated with Korean spreads but adds some marginal explanatory power to the equation. The much higher correlation with other higher-rated emerging markets is not surprising, and indeed Korea has been viewed as something of a safe-haven in emerging bond markets at times of weakness in countries like Turkey and Argentina, which have few economic links with Korea.
Changes in the external spread are positively correlated with changes in the yield on the Korean three-year domestic bond. This might reflect common sentiment effects yields in the two related asset classes. Alternatively, the correlation may reflect a linkage between the two yields through the use of asset swaps by some investors.
Falls in the Korean sovereign spread are associated with increases in U.S. and regional equity prices. This linkage is not entirely surprising given the strong correlation seen—especially in 2000—between prices on the Nasdaq (and other equity markets) and prices of emerging markets debt. Again, this is further evidence that Korean assets have been viewed as cyclical assets, tending to do better at times when the risk appetite of global investors is rising.
Spreads on Korean and Other BBB-rated Countries (In basis points; 5-day moving averages)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Spreads on Korean and Other BBB-rated Countries (In basis points; 5-day moving averages)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
Spreads on Korean and Other BBB-rated Countries (In basis points; 5-day moving averages)
Citation: IMF Staff Country Reports 2002, 020; 10.5089/9781451822069.002.A004
G. Conclusion
47. This paper has provided a broadly consistent picture of Korean assets being strongly influenced by regional and global factors. For equities this is hardly surprising. What is more surprising, however, is that movements in equity prices in the United States and in regional financial centres appear to affect the value of the Korean won and Korea’s external dollar-denominated bonds. Indeed, although the regressions for each of the three asset classes—equities, the won, and the dollar bonds—reveal significant correlations with other Korean assets, most of the explanatory power comes from external variables. In part this may reflect the limited number of proxies included for domestic factors, and no doubt a substantial amount of the remaining unexplained variance could be explained by domestic factors if one could identify and include high frequency measures of corporate profitability and domestic economic and political conditions. Still it seems clear that external factors play a very substantial role in the determination of Korean asset prices. In addition, the net flows of foreign investors appear also to have a significant impact on the equity and currency markets.
48. One interpretation would be that the Korean assets examined here—equities, the won, and Korea’s external sovereign debt—are all viewed as cyclical or high beta plays, with prices that tend to rise with increases in regional or U.S. asset prices, which are proxies for global and regional economic conditions and the risk appetite of international investors. These correlations work in Korea’s favor at times of global expansions, but would be less favorable at times of global contractions or global falls in asset prices. Yet the Korean economy and Korean asset prices have performed relatively well in the current global slowdown. The equity market was one of the strongest in the world in 2001, the exchange rate strengthened modestly in effective terms, the sovereign spread fell significantly, and net inflows from foreign investors have remained positive. This suggests that the structural reform since the crisis and the strong macroeconomic fundamentals have partly offset or even dominated any impact from weakness in global or regional asset prices. They provide evidence to other middle income and emerging market economies that strong linkages with global financial markets need not be destabilizing if the domestic economy and macroeconomic policies are sound.
References
Bae, Kee-Hong, Keiichi Ito, and Takeshi Yamada, 2001, “How do Foreigners Profit From Local Investors? Evidence from Trades in the Japanese Stock Market,” mimeo.
Brennan, Michael J., and Henry Cao, 1997, “International Portfolio Investment Flows,” Journal of Finance 52, pp. 1851–1880.
Borensztein, Eduardo R., and R. Gaston Gelos, 2000, “A Panic-Prone Pack? The Behavior of Emerging Market Mutual Funds,” Working Paper 00/198(Washington, D.C.: International Monetary Fund).
Choe, Hyuk, Bong-Chan Kho, and René M. Stulz, 1999, “Do Foreign Investors Destabilize Stock Markets? The Korean Experience in 1997,” Journal of Financial Economics 54, pp. 227–264.
Choe, Hyuk, Bong-Chan Kho, and René M. Stulz, 2001, “Do Domestic Investors Have More Valuable Information About Individual Stocks than Foreign Investors?” NBER Working Paper Series, Working Paper 8073 (http://www.nber.org/papers/w8073).
Chopra, Ajai, Kenneth Kang, Meral Karasulu, Hong Liang, Henry Ma, and Anthony J. Richards, 2001, “From Crisis to Recovery in Korea: Strategy, Achievements, and Lessons,” Working Paper 01/154 (Washington, D.C.: International Monetary Fund).
Clark, John, and Elizabeth Berko, 1997, “Foreign Investment Fluctuations and Emerging Market Stock Returns: The Case of Mexico,” Staff Reports Number 24, Reserve Bank of New York (May).
Dahlquist, Magnus, and Göran Robertson, 2001, “Foreigners’ Trading and Price Effects Across Firms,” mimeo, Duke University.
Engen, Eric M., and Andreas Lehnert, 2000, “Mutual Funds and the U.S. Equity Market,” Federal Reserve Bulletin, pp. 797–812 (December).
Froot, Kenneth A., Paul G. J. O’Connell, and Mark S. Seasholes, 2001, “The Portfolio Flows of International Investors,” Journal of Financial Economics 59, pp. 151–193.
Goetzmann, William N., and Massimo Massa, 1999, “Daily Momentum and Contrarian Behavior of Index Fund Investors,” Yale University, Working Paper No. ICF-99.13.
Grinblatt, Mark, and Matti Keloharju, 2000, “The Investment Behavior and Performance of Various Investor Types: A Study of Finland’s Unique Data Set,” Journal of Financial Economics 55, pp. 43–67.
Hahm, Joon-Ho, and Frederic S. Mishkin, 2000, “The Korean Financial Crisis: An Asymmetric Information Perspective,” Emerging Markets Review 1, pp. 25–52.
Kaminsky, Graciela, Richard Lyons, and Sergio Schmukler, 2000, “Managers, Investors, and Crises: Mutual Fund Strategies in Emerging Markets,” mimeo,World Bank.
Kim, Woochan, and Shang-Jin Wei, 2000, “Foreign Portfolio Investors Before and During a Crisis,” Journal of international Economics 56, pp. 77–96.
Morck, Randall, and Fan Yang, 2001, “The Mysterious Growing Value of S&P Membership,” NBER Working Paper 8654.
Richards, Anthony J., 2002, “Momentum Investing and Price Impacts of Foreign Investors in Emerging Equity Markets: Evidence from Five Asian Economies,” mimeo, International Monetary Fund.
Schinasi, Gary J., and R. Todd Smith, 2000, “Portfolio Diversification, Leverage, and Financial Contagion,” IMF Staff Papers, Vol. 47, Number 2 (Washington, D.C.: International Monetary Fund).
Sias, Richard W., Laura T. Starks, and Sheridan Titman, 2001, “The Price Impact of Institutional Trading,” mimeo.
Tesar, Linda L., and Ingrid M. Warner, Jeffrey A. Frenkel, 1994, “International Securities Transactions and Portfolio Choice,” The Internationalization of Equity Markets, (Chicago, IL: University of Chicago Press) pp. 185–216.
Warmer, Vincent A., 1995, “Aggregate Mutual Fund Rows and Security Returns,” Journal of Financial Economics 39, pp. 209–235.
Zheng, Lu, 1999, “Who Moves the Market? A Study of Stock Prices and Investment Cashflows,” (Ann Arbor, MI: University of Michigan Business School).
This paper was prepared by Anthony Richards (APD).
See, for example, Hahm and Mishkin (2000) for an analysis of Korean asset prices during the crisis period.
The Korea Stock Exchange’s Fact Book provides further details on the KSE.
There are a few exceptions, for public corporations and in certain industries (e.g., telecommunications and airlines) where there are limits on aggregate foreign holdings, typically in the range of 33–49 percent.
About 35 Korean companies have some form of ADR or GDR arrangement on U.S. or European markets, though only a small number of these have good liquidity. There are also three closed-end funds traded on U.S. exchanges.
See Salomon Smith Barney, Emerging Markets Equity Allocator, January 2002, which estimates the total trading cost for establishing a $2 million position in Korean stocks at 0.22 percent, versus a median of 0.92 percent for 25 other emerging markets.
See Chopra et al. (2001) for an outline of corporate and financial sector reforms.
For a foreign investor’s view on Korean corporate governance, see the presentation by John Lee on “Scudder’s View on Corporate Governance” at www.aicg.org.
Each trading day on the KSE and KOSDAQ markets begins and ends with a batch auction. All orders placed in the automated system between 8 a.m. and 9 a.m. are held and processed in an opening auction that determines opening prices. Similarly, all orders placed between 2:50 p.m. and 3 p.m. are held and processed (along with earlier orders that were placed at “limit or market at close”) in a closing auction at 3 p.m. that determines closing prices. Overnight returns are defined as the change between the closing value of the KSE on one day and the opening value on the next day, with intraday returns defined correspondingly.
The Philadelphia Semiconductor index is an index of the stock prices of 16 U.S. semiconductor-related companies and is closely watched in Korea given the importance of semiconductors and computer-related stocks in Korea.
The adjusted R-squared of similar regressions using monthly returns is somewhat higher, at about 0.55.
The variables that were tested for inclusion in the equation were as follows: U.S. market returns—the previous day’s return on the Dow Jones Industrial index, the Nasdaq Composite index, the S&P500 index, and the Philadelphia Semiconductor stock index, and the change in the yield spread on Korean sovereign bonds (as measured by the JP Morgan EMBI spread); Asian timezone returns—the same day daily return in the Topix index, Hang Seng index, the Straits Times index, the Taiwan Stock Exchange index, and the All Ordinaries index; domestic variables—the change in the three-year government bond yield, the change in the overnight call rate, and the percentage change in the value of the won in Seoul trading. For each of the previous variables, one lagged value was also included. Other control variables include day of the week dummies, and two lags of the returns on the KOSPI and KOSDAQ index.
Given that many of the possible explanatory variables are highly correlated, a stepwise procedure seemed most appropriate to identify those which appear to be most correlated. Such procedures run the risk of “data mining”, but in the current case—with sample sizes typically over 600 observations—the risk of spurious correlations seems fairly low.
Here, and for the remainder of the paper, net inflows are measured as a percent of total market capitalization.
For further discussion of different possible explanations of correlations between flows and returns, see e.g., Engel and Lehnert (2000), Clark and Berko (1997), and Sias, Starks and Titman (2001).
For the base broadening effect to explain positive correlation, the price increases that are generally thought to accompany equity market liberalization would have to occur at least partly through the process of increased ownership, rather than occurring immediately at the time of liberalization.
Alternatively, the model of Brennan and Cao (1997) would suggest that under certain circumstances positive correlation between flows and returns could result from foreigners being less informed than domestic investors.
See Cho, Kho and Stulz (1999, 2001) and Kim and Wei (2001) for other empirical work using the KSE data at the individual stock level.
See, for example, Kaminsky et al. (2000) and Borensztein and Gelos (2000) for studies of mutual fund portfolio behavior in emerging markets.
These differences are made more stark when viewed in annualized terms, of about 300 percent on net inflow days and -85 percent on net outflow days.
By contrast, much of the previous work analyzing the linkages between investor flows and equity prices—most notably the U.S. literature on institutional investor purchases and flows into mutual funds—has used monthly or quarterly data, which does not adequately allow one to assess if returns cause flows or vice versa.
Positive net purchases following increases in foreign stock prices could, however, also be consistent with a simple portfolio model where the increase in the value of one asset changes portfolio weights and induces purchases of the other asset (see Schinasi and Smith (2001).
Given that flows have positive autocorrelation, it follows that a boost to inflows on one day is associated with further inflows on subsequent days. However, to the extent that flows are somewhat predictable, it should only be the surprise or unexpected component of flows that impacts upon prices, with the expected component having no impact (see Warmer (1995)) To test this, a series for “expected” foreign flows on day t was constructed based on a regression similar to that in Table IV.3, but using only variables predetermined at the end of Korean trading on day t-1. Unexpected flows were then derived as actual flows less expected flows. When net foreign inflows are decomposed in this way and added to the equation in paragraph 13, the coefficient on unexpected flows is 0.11 and highly significant (t-statistic of 4.6) while the coefficient on expected flows is approximately zero and insignificant. Thus, as expected the coefficient on surprise inflows is larger than the coefficient on total flows, corresponding to a price impact of 11 percent for a surprise inflow equivalent to 1 percent of market capitalization.
Grinblatt and Keloharju (2000, p. 66) suggest that the contrarian trading behavior of unsophisticated Finnish investors (especially households) may be due to their being “overly eager to cash out on winning stocks or to buy losing stocks.”
This would be consistent with data for Korea’s beta with respect to the world market return. For example, Salomon Smith Barney’s Emerging Markets Equity Allocator, January 2002, estimates Korea’s beta at 2.2, the second highest in their sample of 26 emerging markets.
About 8–10 percent of daily trading occurs in the opening auction. Some feedback trading cannot be ruled out in this auction as the total buy and sell orders are known to investors in the lead-up to the opening auction.
The Froot et. al (2001) results are also far higher than the estimates in Dahlquist and Robertson (2001) which would appear to imply a price impact of 3.4 percent for inflows into Sweden.
The one exception to the finding that foreign inflows are associated with price increases appears to be the KOSDAQ market.
See Richards (2002) for more details of the analysis for these three countries, and the relationship between price impacts estimated using data of different frequency.
The candidate variables include domestic variables (changes in the three-year bond yield and overnight call rate), domestic and foreign stock market movements, net foreign inflows into the KSE, changes in the value of the yen and euro against the dollar, and day of the week dummies.
Elasticities of 0.20–0.45 are lower than the unit elasticity that market participants have sometimes spoken of. The estimated elasticities seem more plausible as measures of the longer-run average correlation. Indeed, it is not uncommon to see temporarily high correlations between relatively different countries, which market participants rely on as rules-of-thumb until the correlations break down entirely.
The spreads for the EMBI components (and the U.S. corporate debt variables) are defined as the yield on the securities less the yield on U.S treasury securities of comparable maturities.
The variables included as potential regressors included: changes in the average sovereign spread for other BBB rated countries (an average of EMBI global spreads for Malaysia, Poland and South Africa) and in the overall EMBI+ spread; changes in U.S. corporate spreads on BBB and BB rated bonds; the change in the three-year government bond yield in Korea and in the overnight call rate; the percentage change in the won exchange rate; net purchases by foreigners on the KSE over the five day period; and equity returns in Korean, regional and U.S. stock markets.
T-statistics are based on Newey-West standard errors to take account of the moving average error term from using overlapping five-day return observations.
