상관계수 정보를 이용한 투자전략의 유효성에 관한 연구 - 한국 주식시장에서의 실증분석

Abstract

This study conducted empirical tests on the effectiveness of the portfolio strategy based on correlation coefficient information. The objective of this study is to demonstrate that a part of the factor structure in stock returns reflects variations in correlation coefficient information. This study collected data from the KSRI, KIS-FAS, and ECOS to construct portfolios of stocks sorted by various characteristics of returns. Sample firms were selected from common stocks listed on the Korean Stock Exchange (KSE) for the period from January 1, 1981 to December 31, 2002. Initially, 511 companies were chosen for the empirical analysis. Ang et al. (2001) found that stocks with higher correlations with the market have higher expected returns in the declining U.S. stock market. To compare the U.S. stock market, this study tries to test the effectiveness of correlation information in the Korean Stock Market by portfolios based on the correlation information. Main results are summarized as follows. First, the expected return on a portfolio of stocks with the lowest correlation has the highest expected returns among 10 portfolios in the declining Korean Stock market. This study also focuses on the observation period to calculate the correlation, because previous works found that the length of the observation period determines the characteristics of time-varying correlations. The finding of this study is that the 6-month observation period gives more effective information than other observation periods. Second, to compare the performance of the investment strategy based on the correlation information, this study investigates the performance of the portfolios sorted by betas and non-systematic risks. In the declining Korean Stock market, an investment strategy based on the correlation information shows better performance than others. Third, controlling for the market beta, the size effect, the book-to-market effect, and the momentum effect, the correlation coefficient information has some explanatory power for the returns of stocks. Finally, this study constructs a correlation premium factor that goes long stocks with low correlations and goes short stocks with high correlations. The correlation premium results in a statistically significant average return of 1.38% per month.