Markowitz revisited: mean-variance models in financial portfolio analysis (Q2706425)

This paper gives a detailed treatment of mean-variance approaches to finding optimal portfolios. It discusses first the single-period case with general returns and then the multiperiod case for scenario trees where the returns are random variables taking only a finite number of values. The main emphasis is on a thorough theoretical analysis of several variants of the basic problem and the relations between the corresponding solutions. The situations under consideration are 'risky assets only', 'risky assets and riskless cash', 'risky assets, cash, and guaranteed total loss', and in addition 'downside risk' in the single-period case. The idea of an asset with guaranteed total loss formalizes the possibility of extracting surplus money to avoid penalization from an overperformance; this is related to (approximate) downside risk minimization.