추정오차를 감안한 팩터 기반 최적 조합과 자산 기반 최적 조합 우위 비교

Abstract

Purpose - This paper conducts both simulation and empirical study to compare the factor-based asset allocation and asset-class-based asset allocation. We focus on how close the investment opportunity sets from these two methods get to the true investment opportunity set. Design/methodology/approach - First, we conduct a simulation study which allows us to observe the true minimum variance frontier (MVF). From the simulated asset market, we compare the MVFs created from three factors and three portfolios. We also adopt the confidence region in Jobson (1991) to compare these two MVFs. Second, we use U.S. asset market (stock, bond, cash) data to do the same comparison. Findings - Both studies consistently show that the confidence region for the MVF created from the factors is much narrower, implying that it is more precisely estimated. This concludes that the factor-based asset allocation suffers much less from the estimation errors. Research implications or Originality - Conventional research in this line of literature focuses on the expansion of the investment opportunity set in comparing the two asset allocation methods. However, we focus on comparing the estimation errors in the MVFs from these asset allocation methods.