In recent years, global major exchanges such as the NYSE, LSE, and TSE have faced an unprecedented amount of a specific type of trading known as high frequency trading (HFT). Although researchers and practitioners have not agreed on the precise definition of HFT, the Securities and Exchange Commission (SEC) loosely defines it as “a trading that uses high-speed computer systems to monitor market data and submit a large number of orders to the market.” HFT began in the late 1990s, and its proportion of total trading volume has rapidly increased during the past decade. In 2009, a rough estimation (by TABB group) indicated that HFT accounted for 60% and 30% of the total trading volumes in the US and European equity markets, respectively. To accommodate and process surging HFT-related orders, exchange-level projects to expand order processing capacity and reduce trade execution time are under way among a number of global exchanges, including the Korea Stock Exchange. On March 23, 2009, the Korea Exchange migrated to the EXTURE (‘EXchange + fuTURE’), an upgraded trade settlement system. The launch of this technically advanced trading system might enable investors to implement HFT in the Korean stock market. However, as far as the authors know, little is known about HFT in the Korean stock market. In this study, we find a special type of HFT has emerged as a conspicuous market phenomenon characterized by the extremely fast repetition of order submissions and cancellations. It closely resembles the ‘strategic runs’ observed on the NASDAQ (Hasbrouck and Saar (2013)). The strategic runs discovered in the Korean stock market consist of a chain of normal orders and its cancellation within a certain period of no more than one second between each consecutive order. We further examine the effect that strategic runs have on market quality in terms of short-term liquidity and volatility, measured in intraday ten-second intervals. To measure liquidity, we use proportional spread and the number of shares at the best bid, or ask prices divided by total number of shares outstanding. The highest price minus the lowest price divided by the bid-ask midpoint is used to calculate volatility. During the 10-second intervals when strategic runs are most concentrated, they attract liquidity by submitting higher proportions of marketable limit orders. This leads to increased spreads, decreased depth, and higher volatility. Furthermore, 10-second interval intraday vector auto regression (VAR) results indicate that strategic runs deteriorate market quality by decreasing market depth and raising volatility. We do not, however, find a statistically significant relationship between strategic runs and spreads. Overall, strategic runs appear to have negative effects on market quality. Our results must be interpreted with caution. Because the sample period of VAR analysis is as short as three months, one cannot make sure whether our main results still hold for the post sample period. Another problem is that strategic run is only a special type of high frequency trading. It implies that we cannot generalize our results based on strategic runs for other types of high frequency trading that we do not know yet. We hope these issues will be examined in the future research. While major exchanges are making considerable capital investments into IT infrastructure and network system recently to accommodate and attract high frequency trading, it is unclear what the socioeconomic benefit of the investment is and who makes profits (or suffers losses) in high frequency trading. As high frequency trading is increasingly dominant in the stock market, individual investors who cannot access stock market as fast as high frequency traders will leave the market, and this may lead to reduction in liquidity and eventually market failure. This paper contributes to the finance literature and the government’s policymaking as follows. First, this study is the first to not only verify that strategic runs exist in the Korean stock market, but also to examine their characteristics and effect on market quality. Second, this study finds strategic runs without incurring errors thanks to the use of account-level trades and quotes data. In contrast, the data used by Hasbrouck and Saar (2013) are subject to inherent limitation, as they do not include account level information. For example, they cannot identify whether two consecutive orders are submitted by the same trader, which makes it difficult to extract strategic runs from quote data. Our study is free from such problems because the data used enable us to identify which traders submit which orders. Finally, given that HFT is in the early stages in the Korean stock market, our results might provide useful information for financial authorities regarding whether they will need to regulate HFT in the Korean stock market.

• EndNote
• RefWorks
• Scholar's Aid
• BibTeX

Export Citation Cancel

The term structure of interest rates is closely related to the macro economy, in that it contains information about the current state of the economy and reflects the bond market participants’ expectations regarding the economy’s future path. It can be effectively used to predict major economic variables, including inflation, and therefore has potential as an indicator variable in the establishment of monetary policy. Due to its economic importance, many researchers in the field of finance and macroeconomics have studied the modeling of the term structure of interest rates. In finance, the focus of such research has been the prediction of interest rates and the pricing of interest rate-related derivatives. In macroeconomics, researchers have mainly attempted to understand the changes in the term structure of interest rates in the context of monetary policy implementation. Since the pioneering work of Ang and Piazzesi (2003), however, macro-finance term structure models combining the advantages of the two traditional approaches have provided a useful means of interpreting the movements of the term structure in the economic context while maintaining the consistency between short- and long-term interest rates with the no-arbitrage restrictions. In this study, we apply the unspanned macro-finance term structure model (unspanned MTSM) to analyze the empirical relationship between the term structure of interest rates and the macroeconomic variables using Korean data. An ‘unspanned’ model is free from the unrealistic assumption that the macroeconomic variables can be entirely generated by the first Nth-principal components of the term structure (macro-spanning restriction). It is also more flexible in the sense that the model permits feedback between the term structure and macroeconomic variables. The unspanned MTSM incorporated in this paper was suggested in Joslin, Priebsch, and Singleton (2012) and extended to the multi-lag form in Joslin, Le, and Singleton (2013). In the empirical section, we first conduct the macro-spanning test to determine the validity of the empirical analysis using the unspanned MTSM. The test results show that the term structure factors only explain some of the variations in the real economy (R2 = 13.43%) and the price variables (R2 = 15.40%), confirming the validity of modeling ‘unspanned’ macro variables. Empirically, we identify the macroeconomic variables with the highest explanatory power over the term structure of the interest rate in Korea. We estimate the model with the three term structure factors (level, slope, and curvature) and two combinations of the 82 key macroeconomic variables (i.e. 82C2 = 3,321 models are estimated). The performance of each combination of macroeconomic variables is measured by the various information criteria (AIC, SIC, and HQIC). In our analysis, the combination of the ‘nonfarm payrolls’ in the real economy variable group and ‘consumer price index excluding agricultural products and oils’ in the inflation variable group provides the best performance in explaining the Korean term structure of interest rates. The robustness of these variables is maintained even when we add more macroeconomic variables (i.e. combining the three macroeconomic variables) to the macro-finance term structure model. In the three macroeconomic variable cases, the third additional variable with a high explanatory power is ‘credit spreads’, the spread between AAA-rated and BBB+-rated corporate bond spread. We believe that the ‘nonfarm payrolls’ has the highest explanatory power, as noted by Wu and Zhang (2008) in the US case, because this variable carries information about not only the real economy growth, but also inflation. Then, using the identified three macroeconomic variables, we estimate the unspanned MTSM and analyze the effect that macroeconomic factors have on the market price of risk. Due to the ambiguity of the economic meaning of ‘nonfarm payrolls’, we use the ‘gross domestic product’—the second-best performing real economy variable—in the estimation. The results of this analysis show that the inflation variables have a pro-cyclical effect on the market price of the slope factor while the real economy variables have a pro-cyclical (counter-cyclical) effect on the market price of the level (curvature) factor. However, the third macroeconomic variable—credit spreads—has a negligible effect on the bond risk premium. Finally, we conduct the impulse-response and variance decomposition analysis to show the empirical relationship between the three term structure factors and the three identified macroeconomic variables. We observe that a positive shock on the real economy variables increases the level factor, and as the credit spread widens, the level factor falls. A positive shock on the inflation variable also increases the slope factor, which widens the spread between short- and long-term yields. Finally, through variance decomposition, we confirm that the three macroeconomic variables explain 91% of the variation in the level factor, and 91.42% and 95.39% of the slope and curvature factors in the 36-month forecasting period.

• EndNote
• RefWorks
• Scholar's Aid
• BibTeX

Export Citation Cancel

Based on the relationship between fund performance and the subsequent net flows, this study examines the net flow persistence. Earlier works have focused largely on the positive relationship between fund performance and the subsequent net flows. The most valuable information for fund investment is the past performance of each fund, yet one fund investor may consider the previous 6-month performance most important while another focuses on the previous 36-month performance. Investors’ interpretations of the concept of ‘past performance’ vary, and fund performance affects fund investors’ investment horizons differently. The economic motivation for this study is how fund investors use past fund performance to inform decision making, and how long their responses persist. In this study, we answer the following questions. First, what is the period for measuring past fund performance that has the most significant effect on net flows? Although earlier studies have fixed the past performance period, we use several past periods for measuring fund performance such as 1, 3, 6, 12, and 36 months. Some investors may chase the prior 1-year fund performance while others note the prior 2-year performance. Although different fund investors may make different investment decisions according to the prior fund performance, there could be a period that has the most influential effect on fund investment decisions. Second, how long does the fund performance affect the net flows? To the best of our knowledge, most studies have examined the relationship between the prior fund performance and the subsequent 1- or 3-month net flows. We take a different approach. The past performance of equity funds may affect future flows, but as time passes, the effect disappears. The net flows may follow fund performance for 1 month, 3 months, or longer. We define this duration effect of past performance as ‘flow persistence’ and it reflects the response duration of fund investors. The existence of flow persistence can be explained as follows. The fund performance information may be realized by fund investors after a time delay, or the fund investors may react to the information slowly. In this study, we test the flow persistence by examining the subsequent 1, 3, 6, 12, 24, and 36 months’ net flows following past performance. This concept of flow persistence is significantly different from the performance persistence of equity funds. Our goal is to provide an overall perspective on how fund investors react to past performance information. We measure the equity funds’ performance in multiple ways while analyzing the effects of past period performance and flow persistence; that is, past performance is measured by cumulative return, 1-factor alpha, 3-factor alpha, and 4-factor alpha. Each alpha is a risk-adjusted return. During the sample period, which ran from January 2004 to June 2012, the Korean fund markets were severely hit by the global financial crisis that originated in the U.S. Before the crisis, the fund markets were growing rapidly, but during the crisis, the global and Korean stock markets were sluggish, which led to the poor performance of equity funds. Numerous fund investors were surprised by plummeting fund returns. As a consequence, equity fund redemptions increased in the post-crisis period. In this study, we divide the entire period into two sub-periods, pre- and post-crisis, to determine the effect that the global financial crisis had on the relationship between fund performance and net flows. The empirical findings are as follows. First, the 1-year performance has the most significant influence on the net flows of equity funds, which implies that fund investors make their own investment decisions largely based on the prior 1-year performance rather than other performance periods. Our results support earlier studies that use prior 1-year performance as information to predict subsequent net flows. Second, the flow persistence lasts for 6 to 12 months. This reveals that prior fund performance affects the subsequent future net flows for 6 to 12 months. Third, we also test the effects of the past performance period and the flow persistence by measuring fund performance in many different ways. Additional tests prove the robustness of our results. Finally, we find that both the flow-performance relationship and the effect of flow persistence weaken during the post-crisis period. This study tests the effects of past fund performance as information for predicting subsequent net flow. We believe that our findings could provide new perspectives on how fund investors respond to past performance as information.

• EndNote
• RefWorks
• Scholar's Aid
• BibTeX

Export Citation Cancel

We provide evidence supporting the presence of investors who prefer lottery-like stocks in the Korean stock market. Our empirical findings are as follows. First, we document that higher MAX stocks earn lower average returns, with MAX defined as the maximum daily return over the past month, as in Bali, Cakici, and Whitelaw (2011. Maxing out: Stocks as lotteries and the cross-section of expected returns. Journal of Financial Economics 99, 427-446). Specifically, we find that average risk-adjusted return differences between stocks in the lowest and highest MAX quintiles is 1.39 percent per month over our sample period. This finding suggests that investors have a preference for lottery-like assets, i.e., assets that have a relatively small chance of a large payoff. Given this preference, investors may be willing to pay more for stocks with lottery-like payoffs, prompting such stocks to have lower returns in the future. In this study, we measure the propensity for a stock to deliver lottery-like payoffs on the basis of MAX, defined as extreme positive returns over the past month, and find a negative relationship between MAX and expected returns in the cross-section. In other words, we demonstrate that lottery-like stocks, which presumably exhibit high MAX, have low expected returns in the Korean stock market, confirming previous findings for the U.S. market. Our finding of a negative relationship between MAX and expected returns appears to be robust to various cross-sectional effects such as size; book-to-market; and momentum, liquidity, and short-term return reversals. Both portfolio sorts and cross-sectional regressions reveal that the MAX-return relationship continues to be significant after controlling for other effects. Further, we find little evidence that the idiosyncratic volatility puzzle, i.e., the negative relationship between the idiosyncratic volatility and average returns in the cross-section documented by Ang, Hodrick, Xing, and Zhang (2006. The cross-section of volatility and expected returns. Journal of Finance 61, 259-299), account for the relationship between MAX and returns. Considering that MAX is, on average, positively related to idiosyncratic volatility, one can argue that the negative MAX-return relationship is a different appearance of the idiosyncratic volatility puzzle. However, our empirical results reveal that the MAX effect is robust to controls for the effect of idiosyncratic volatilities. Second, we find that stocks with high MAX tend to be small and low-priced, and have higher idiosyncratic volatility. Moreover, we observe that high MAX stocks are, on average, heavily traded by retail investors. MAX exhibits a monotonically increasing pattern in retail trading proportions (RTP) in the cross-section, where we define a stock’s RTP as the monthly buyer- and seller-initiated retail trading volume divided by the total trading volume of the stock in that month. This finding is consistent with the literature, which regards small, low-priced stocks exhibiting high idiosyncratic volatilities and skewness to have lottery-like features. Last, and most importantly, we find that the negative relationship between MAX and average returns is more prominent among stocks with higher retail trading proportions, using a unique dataset that enables us to identify retail investors’ trades. In particular, we find that the MAX-return relationship is more significant and negative among stocks with higher RTP. The strategy of selling high MAX stocks and buying low MAX stocks earns, on average, 1.87 (0.17) percent per month among stocks in the highest (lowest) RTP quintiles. This evidence is consistent with previous studies arguing that retail investors are more likely to have a greater gambling propensity. Collectively, we contribute to the literature on investors’ preferences by presenting evidence of investors’ preference for lottery-like stocks in the Korean stock market. We also contribute to the literature on retail investors’ trading behavior. We use a unique dataset from the Korean stock market that enables the identification of retail investors’ trades while providing empirical evidence that retail investors are more associated with the preference for lottery-like assets.

• EndNote
• RefWorks
• Scholar's Aid
• BibTeX

Export Citation Cancel

In this paper, we aim to survey studies on financial intermediation in Korean financial markets with a view to understanding and evaluating the current status of financial intermediation services in Korea. Specific issues discussed include the relationship between economic growth and financial development, financial intermediation services such as liquidity provision, economizing transaction costs, production and transmission of information, monitoring, and the effects of relationship banking in Korea. We also discuss corporate restructuring driven by Korean banks as a special form of financial intermediation at the final stage of client firms’ lives, policy finance, and shadow banking which is regarded as an important factor influencing the future of financial intermediation in Korea. This survey reveals that while banks’ intermediation services are known to be special and value-additive in the US, for example, Korean studies have not shown the same. That is, research results show that banks’ intermediation services in Korea have been either insignificant or sometimes less than value-additive as benefits from intermediation services have been outweighed by associated costs. This result is partly because banks have had superior negotiation power to borrower firms in Korea, and partly because of the government’s dominating influence on bank behavior. It may also be because of sampling bias in terms of data period, as most studies use data near the 1997 financial crisis.

• EndNote
• RefWorks
• Scholar's Aid
• BibTeX

Export Citation Cancel