초록 close

Markowitz(1953)의 평균-분산 모형 이후 최적 포트폴리오 구성에 관한 많은 연구가 진행되어 왔으나 투자자의 선호체계에 대한 논란이 지속되고 있다. 본 논문은 분산 외에 VaR, LPM, CVaR 등 3개의 위험지표를 사용한 자산배분을 실증적으로 비교분석하였다. 분석결과, 기대수익률 선택에 따라 적정위험지표가 다소 차이는 있지만 CVaR가 위험조정수익률 측면에서 유리한 것으로 나타났으며 상승 및 하락국면으로 구분할 경우에는 적정위험지표가 각각 상이하였다. 한편 기대수익률에 역사적 평균 적용시 LPM, No-view 및 전문가 전망 적용시 CVaR의 누적수익률이 높았다. 따라서, 투자자는 위험조정수익률과 누적수익률 제고를 목적으로 할 경우 위험지표로 CVaR를 고려할 수 있고, 정확성 측면에서 기대수익이 리스크보다 포트폴리오에 큰 영향을 미치므로 기대수익률 추정능력의 제고가 필요하다.


The optimal portfolio composition has been of special interest and of great importance in the finance literature since the introduction of the mean-variance model by Markowitz (1952), although the proper preferences of investors have still been consistently debated. In this paper, we compare the performance of asset allocations with four different risk indicators, such as the VaR, the LPM, the CVaR, and the variance. Our empirical results indicate that each risk indicator performs slightly differently, depending on which measure of the expected rate of return is selected (such as the historical mean, LPM, and No-view). The appropriate risk indicator changes somewhat depending upon the chosen expected return. If the historical average return to the expected return is used, LPM offers the highest return and the standard deviation. When “No-view” to the expected return is applied, CVaR shows the highest return with the lowest risk. Also, if the expert forecasts for the expected return are considered, the returns for Variance and CVaR are highest. In addition, when stochastic dominances are compared, CVaR has the second degree stochastic dominance over VaR when historical return is applied to expected return. CVaR stochastically dominates in the second degree over VaR and LPM when “No-view” was applied, and CVAR has the second degree stochastic dominance over VaR and LPM when the expert forecasts are applied. In general, the CVaR has an advantage in improving a risk adjusted return regardless of different measures of expected returns, but when the time period is divided into higher return and lower return periods, the CVaR does not necessary perform better. That is, the appropriate risk indicators are different over various market periods. During an upward trending market, the Sharpe ratio is the highest for LPM in the case of historical return, VaR in the case of “No-view” and the expert forecasts, respectively. In contrast, during a down trending stock market with downside risks, the Sharpe Ratios of Variance and CVAR in the case of historical return, LPM in the case of “No-view”, and the variance in the case of expert forecasts are the highest, respectively. Furthermore, our results suggest that the expected rates of return have greater significant effect than the risk measures upon portfolio performance. When differences in asset weights for various risk indicators are compared, with a few exceptions significant discrepancies are found when either the historical return or “No-view” is used for expected return.