한국 주식수익률의 장기 예측가능성과 분산분해분석에 관한 실증 연구 (Empirical Study on the Long-run Predictability and the Variance Decomposition of Korean Stock Returns)

서지정보

· 발행기관 : 한국금융학회
· 수록지 정보 : 금융연구 / 26권 / 4호 / 61 ~ 91페이지
· 저자명 : 정동준

초록

이 논문에서는 한국 주식시장 통계 자료를 월별로 1981년 1월부터 2011년 12월까지 사용하였고, 이 분야에서 표준적인 모형이라고 판단되는 Campbell(1991)과 Hodrick(1992)의 모형을 사용하여 추정하였다. 이 논문의 주요 추정 결과를 간단히 요약하면 다음과 같다. 첫째, 현재의 주식수익률을 예측하는데 있어서, 과거의 주식수익률, 배당수익률, 주가수익비율은 장단기에 있어서 예측력을 갖는다. 그러나, Fama and French(1988)가 주장하였던, 현재의 주가수익률을 과거의 배당수익률들로 예측할 경우, 과거 시차가 장기로 갈수록 계수 값이 커진다는 현상은 한국 주식시장에서는 발견되지 않았다. 둘째, 현재의 주식수익률을 예측하는데 있어서, 과거의 실질이자율들은 영향을 안준다는 의미에서 중립적인 역할을 하였다. 셋째, 미국 주식수익률에서 발견되는 평균 회귀 현상은 한국 주식수익률에서는 발견되지 않았다. 마지막으로, 주식의 초과수익률의 예상치 못한 부분의 분산의 거의 대부분인 73%를 미래의 실질 배당성장률에 대한 분산이 설명하였다. 그리고, 이러한 분산 분해 결과는 앞에서 논의한 한국 주식수익률의 장기 예측가능성에 대한 추정 결과와 일치하는 측면이 있었다.

영어초록

This paper empirically investigates whether the long-run time-series characteristics of U.S. stock returns exist in Korean stock returns. If there are differences between the two countries’ stock returns, the reason may be that different time-lengths and composition of the two total stock returns with individual dividends.
In this study, the standard model as defined by Campbell (1991) and Hodrick (1992) was used for estimations of Korean stock market data for the sample period (January 1981 to December 2011).
To summarize, the major estimation results of this paper are as follows:First, to predict current stock returns, past stock returns, past dividend yields, and past price-earnings ratios for forecasting power in both the short- and long-run, we used Fama and French (1988) assertion that long-horizon stock returns were more predictable than short-horizon stock returns. However, the power of the dividend yield to forecast stock returns increased with the length of stock returns horizon was not found in Korean stock market until recently. The reason is thought to be the different compositions of the two countries’ total stock returns with individual dividends, in the sense that the dividend portion in Korean stock returns is especially smaller than that of U.S. stock returns.
Second, forecasting power was not indicated because past real interest rates had a neutral effect on current stock returns. It is thought that Korean stock returns tend to respond more sensitively to business-cycle variables such as dividends and earnings rather than real interest rates.
Third, the mean reversion phenomenon that had been found in American stock returns was not found in Korean stock returns until recently.
Fourth, 73% of the variance was unexpected for excess stock returns, as explained by the variance of future real dividend growth rates. Furthermore, the estimation results for variance decomposition were nearly consistent with the long-run predictability of Korean stock returns.
Finally, the different properties of the long-run time-series Korean stock returns from those of the U.S. stock returns are thought as the following three reasons. The first reason is that the different time-lengths of the two countries’ stock returns. The second is that the different compositions of the two countries’ total stock returns with individual dividends. The third is that they were attributed to the different degrees of development of each countries’ short-term financial and bond markets; i.e., where the interest rate variables were determined.

참고자료

· 없음