LOG IN창 닫기

  • 회원님의 아이디와 패스워드를 입력해 주세요.
  • 회원이 아니시면 아래 [회원가입]을 눌러 회원가입을 해주시기 바랍니다.

아이디 저장

   

아이디 중복검사창 닫기

HONGGIDONG
사용 가능한 회원 아이디 입니다.

E-mail 중복확인창 닫기

honggildong@naver.com
사용 가능한 E-mail 주소 입니다.

우편번호 검색창 닫기

검색

SEARCH창 닫기

비밀번호 찾기

아이디

성명

E-mail

학술자료 검색

고빈도 데이터(HFD: High Frequency Data)를 활용한 페어 트레이딩(Pairs Trading) 전략의 성과 특성에 관한 연구

  • 윤주영 미래에셋맵스자산운용 이사
  • 김강휘 하버드대학교 공학ㆍ응용과학대학원 석사과정
본 연구는 대표적인 시장중립 매매 전략 중 하나인 페어 트레이딩 전략을 고빈도 데이 터를 활용할 수 있는 형태로 변형하여 국내 시장에 적용한 연구이다. 기존 연구와의 차별성은, 일별 종가가 아닌 고빈도 데이터를 활용하여 성과의 특성을 연구하였다는 점, 그리고 이를 위해 매매 신호 추출 시 칼만필터(Kalman Filter)를 이용하여 시변 (Time-adaptive) 회귀상수를 계산한 뒤 종목 간 스프레드를 계산하였다는 점이다. 전략 성과 측정을 위한 대상종목은 거래 특성을 고려하여 유동성이 높고 차입이 원활 한 대형주 유니버스로 제한하였고, 성과 분석 결과 보수적인 거래비용을 고려했음에 도 시장국면에 무관한 유의적인 양의 수익을 얻을 수 있었다. 전략의 주요 성과 특징 으로는, 시장 하락기에 성과가 상대적으로 더 좋게 나타났고, 업종 간 성과의 차별성 을 확인할 수 있었으며, 장 시작 직후 및 장 마감 근처에 진입한 매매에서 상대적으로 높은 승률과 수익률이 실현되었다. 한편 표본 평균회귀강도(Mean-reversion Strength), 표본 정보비율(Information Ratio), 표본ADF(Augmented Dickey-Fuller) 테 스트 t-통계량을 추가적으로 고려하여 합산 통계량 순위 상위 페어를 선정한 뒤 선택 적 매매를 할 경우 성과가 더욱 향상되는 것을 확인할 수 있었다.
시장중립,페어 트레이딩,통계적 차익거래,칼만 필터,고빈도 거래

Performance Analysis of Pairs Trading Strategy Utilizing High Frequency Data : Evidence from the Korean Stock Market

  • Jooyoung Yun
  • Kangwhee Kim
The pairs trading is a strategy often adopted to identify arbitrage opportunity based on historical equilibrium in spread between the share prices. Basically, an investor evaluates the current position of the spread based on its historical fluctuations and seizes the moment when the current spread deviates from its historical mean level by a pre-determined threshold. In this study, the well-known basic pairs trading strategy, one of typical market neutral strategies, is modified so as to utilize high frequency data, and it is also applied to the constituent shares of the KOSPI (The Korea Composite Stock Price Index) 100 index. We also introduce the use of the high frequency equity data in strategy modeling, although the industry practice for market neutral hedge funds is to use daily sampling frequency of equity data in designing a trading model. In this study, intraday stock prices data sampled at a 30-minute interval is used for the strategy, and the performance is analyzed in high frequency domain. The data set covers the horizon from the 1st of October 2008 to the 31st of July 2010, which includes bullish, bearish, and flat market periods within the horizon. We highlight how perfor- mance varies depending on market condition, industry group, and timing of the market entry. This study is distinguished from the most previous works on the traditional pairs trading strategy in that we introduce the use of high frequency data in strategy modeling instead of daily closing prices, which allows us to analyze the performance of the strategy in high frequency domain. More specifically, we extract the trading signal, which is based on the spread between stocks of comprising a pair, by estimating time adaptive regression coefficient using the Kalman filter scheme. This study is the first practice in the realm of the high frequency market neutral trading strategy that extracts trading signal by estimating time adaptive regression coefficient using the Kalman filter scheme. Moreover, our loss-cut strategy clears position if holding duration exceeds the pre-determined maximum trading duration. As for the underlying universe for the strategy, we confine ourselves by considering only the most liquid top 100 stocks in terms of larger trading amounts and higher liquidity as a basket for our experiment. This is to get rid of other external variables that may add undesirable noises to the overall performance which would make it difficult to analyze pure performance of the strategy itself. We analyze the results of out-of-sample performance test from various angles. Major findings include that arbitrage profitability is, in fact, present without being subject to market condition even when conservative transaction costs are taken into account. In particular, our strategy outperforms better in the bear market condition, showing 2.55% of average rate of return per trade in bearish period, which is higher than 0.80% in the bullish period and 0.39% in the flat period. The performance of the pair trading strategy varies depending on the industry group. Those industry groups dominantly influenced by domestic demand i.e. non-cyclical in Korea such as Distribution, Household & Personal Products, and Automobiles and Components show relatively higher winning ratios and average rates of return per trade whereas the industry groups involving fast-paced technological development and variable international demand such as Technology Hardware and Software & Services give out relatively low statistics. The results also demonstrate that the performance of the strategy is dependent upon the time when trading position is set up during daily trading hours; the performance of trades entered around at opening and closing of the daily market appears to be relatively superior to that of trades executed in the rest of daily trading hours in terms of the average rate of return per trade, the winning ratio, and the information ratio. Recalling that intraday volatility pattern is generally formed in U-shape, the strategy seems to achieve higher performance in intraday time zone with higher volatility, which also corresponds to our previous finding that our strategy returns outstanding figures in volatile market trend. Besides, the strategy seems to take advantage of inefficiency derived from where stock price reflects the undisclosed market information. Furthermore, we introduce an enhanced version of the pair trading strategy and compare the performances with the basic strategy. One difference between the basic and the enhanced model is that it selects high-ranking pairs to trade for the next time period based on a set of in-sample statistics, which includes in-sample ADF (Augmented Dickey-Fuller) test t-statistics, in-sample information ratio, and in-sample mean reversion strength. In our study, moving window with the size of 2 weeks is considered as an in-sample period. It is verified that the performance of the enhanced strategy had better profitability and reliability compared with our basic strategy.
Market Neutral,Pairs Trading,Statistical Arbitrage,Kalman Filter,High Frequency Trading