In this thesis, the efficient portfolios using stochastic dominance rules are obtained, and compared with the efficient portfolios obtained by the E-V criteria and E-SV criteria. In E-V model and E-SV model, it is assumed that the distribution of returns is normal, or that the investor's utility function is quadratic. But this assumption is not rational in real world.
Test shows the unreality of the above assumption, so a new criteria, namely, Stochastic Dominance rule, for obtaining efficient sets, which uses whole distribution as decision variables, is introduced. The particular attraction of Stochastic Dominance is that its results are consistent with expected utility hypothesis without depending on a particular mathematical form of utility function or one specific type distribution of returns.
E-V efficient sets are obtained by L.P., E-SV efficient sets by Q.P. and SD efficient sets by Fortran Program.
Major results are as follows;
(1) K-S test shows that the distribution of returns is not normal. More than 50 percent of the 59 stocks have not normal distribution,
(2) the differences between the E-V and SD efficient sets are great, as might be expected,
(3) in the upper range of mean and variance, the three sets are virtually identical, but most of the portfolios that are E-V efficient but not SSD or TSD occur in the lower range of the mean and variance,
(4) the E-SV efficient sets are subsets of the TSD efficient set and the size of the two sets are almost same, so E-SV criteria can be said better than E-V criteria.
The above results support that, in real world, E-V criteria is not appropriate to obtain efficient portfolios, and that E-SV criteria is relatively good, and that SD criteria is nearest to real world.