Paul A. Samuelson, Limited Liability, Short Selling, Bounded Utility, and Infinite-Variance Stable Distributions, The Journal of Financial and Quantitative Analysis, Vol. 11, No. 3 (Sep., 1976), pp. 485-503

The Markowitz analysis of efficient portfolio selection, which can be interpreted as solving the quadratic-programming problem of minimizing the variance of a normal variate subject to each prescribed mean value, easily can be generalized (in the special case of independently distributed investments) to the concave-programming problem of minimizing the “dispersion” of a stable Pareto-Lévy variate subject to each prescribed mean value. Some further generalizations involving interdependent distributions will also be presented here.

“Don't put all your eggs in one basket, ” is a familiar adage. Economists, such as Marschak, Markowitz, and Tobin, who work only with mean income and its variance, can give specific content to this rule—namely, putting a fixed total of wealth equally into independently, identically distributed investments will leave the mean gain unchanged and will minimize the variance.