William Hakes- Decision Sciences Article Summary
1) Portfolio Selection Using Stochastic Dominance Criteria”. Decision Sciences; Atlanta; Fall 1998; John R. McNamara. Volume 29 Issue 4, pg 785-801.
2) John R. McNamara: Iacocca Professor in Business and Economics at Lehigh University, Bethlehem, PA 18015-3117, email@example.com.
3) Is Stochastic Dominance a better method for portfolio selection than other more traditional measures?
4) Applied Statistical: This paper uses a sample of 60 NYSE stocks from Compustat and data from the S&P 500 Index from 1981-1996. A linear programming second-degree stochastic dominance (LPSSD) model is compared to a traditional quadratic programming (QP) model.
5) The central issue taken with traditional LP or QP portfolio models is that they use the first two moments of a series as the basis for optimization. This ignores the 3rd moment (skewness), which if known, would provide insight into the derivation of the standard deviation and possible asymmetries that would be important to the portfolio selection process. Stochastic dominance criteria appears to generate higher expected returns in most cases versus the traditional QP. In the few cases where the QP generated higher returns, the variance in the model most understated the actual risk, in the sense that negative skewness or very low returns were not given proper consideration. Stochastic criteria have several advantages over mean-variance models, including sensitivity to skewness and the capacity to use smaller sets of data.
6) The article would appeal to both academics and practitioners interested in portfolio selection, not just in stocks, but perhaps in project selection as well. In terms of applied vs. theoretical interests, undoubtedly an applied researcher would gain the most by reading this article. However, the theory behind the use of stochastic dominance vs. traditional methods might be of interest to the average person interested in portfolio selection.
7) I hope to publish an article along the lines of this paper, particularly in the area of stochastic modeling. The literature in finance, investing, and in many ways management sciences has to do with more traditional methods of analysis, like regression, the Capital Asset Pricing model, etc. All of these methods fail to capture the upper moments of return distributions, and in fact the results are often under or overstated in terms of the financial risk that is modeled in each respective model. Stochastic modeling takes the study of investments to a more applied level, and thus I hope to apply it in my own article in the future.