Maximizing expected utility is widely accepted in theoretical work, but extension to empirical models is more controversial. We state two theorems which outline conditions when mean-variance procedures such as quadratic programming can be used to maximize expected utility. These conditions are often assumed or satisfied in empirical studies.
This paper was published in Johnson, Donald A., and Michael D. Boehlje. "Conditions for Transforming an Expected Utility Problem into a Mean-Variance Analysis." Unpublished paper, Iowa State University (1979).
Johnson, Donald and Boehlje, Michael, "Conditions For Transforming An Expected Utility Problem Into A Mean-Variance Analysis" (1979). Economic Staff Paper Series. 75.