A statistically optimal estimator of semivariance

This paper develops a statistical estimator for the semivariance. The estimator is shown to be asymptotically unbiased, consistent, and computationally efficient. The estimator is superior to mere approximations to semivariance that have appeared in the literature recently. The suggested use of semivariance by Harry Markowitz in 1959 in cases where distributions are not symmetric should now be easily implementable by workers in portfolio analysis and other areas where risk is measured.