HOW BEST TO FLIP-FLOP IF YOU MUST - INTEGER DYNAMIC-STOCHASTIC PROGRAMMING FOR EITHER-OR

Standard portfolio analysis presumes one can blend different securitie s continuously. When one must choose all of one portfolio or all of an other, we are in stochastic digital programming: either-or, zero-or-on e choice. The algorithm for doing this optimally is shown to be simple r than in real variable maximizing, a switch from the usual extra comp lexities of digital programming. The Bellman multi-period dynamic prog ramming is shown, paradoxically, to make it possible for a risk-averse investor to want sometimes to embrace an unfair gamble. The superiori ty of within-time diversification over across-time diversification car ries over to this flip-flop case.