In this paper, we consider the portfolio selection problem. For 60 selected Hong Kong stocks, we form portfolios based on 10 years of observations. The classical Markowitz mean-variance model is compared with a downside risk optimization model for the Hong Kong stock market, fn particular, the downside risk model provided better downside risk protection in the sense of offering lower risk and higher return for the chosen data. The results also indicate that longer holding periods tend to result in higher portfolio returns. Lastly, the downside risk optimization approach allows investors to specify both required return levels and target return rates in order to reflect their risk attitudes.