As a coherent risk measure, CVaR or expected shortfall (ES) is limited in terms of applying equal weights to the extreme loss beyond Value-at-Risk regardless of investors' risk aversion. Acerbi (2002, 2004) introduced spectral risk measures (SRMs) that reflect investors' subjective risk aversion. In this study, portfolios are composed of three different Korean financial assets: the KOSPI index, won-dollar exchange rates, and the government bond. The asset allocations derived from ES and SRMs with various risk aversion coefficients are compared. The SRMs model converges to the ES model by imposing equal weights to the loss beyond VaR. The results show that when investors are more risk averse, the weights for high-risk stocks decrease and the weights for low-risk government bonds increase. The efficient frontiers of ES and SRMs show that the risk taken depends on the degree of risk aversion, and that investors select the lower risk portfolio when they are more risk averse. The efficient frontier of ES is one of the various efficient frontiers of SRMs, which implies that asset allocation based solely on ES is not appropriate for very risk-averse investors or conservatively managed portfolios.