The Evolution of Security Designs

This paper embeds security design in a model of evolutionary learning. We consider a competitive and perfect financial market where agents, as in Allen and Gale (1988), have heterogeneous valuations for cash flows. Our point of departure is that, instead of assuming that agents are endowed with rational expectations, we model their behavior as the product of adaptive learning. Our results demonstrate that adaptive learning profoundly affects security design. Securities are mispriced even in the long run and optional designs trade off underpricing against intrinsic value maximization. The evolutionary dominant security design calls for issuing securities that engender large losses with a small but positive probability, and otherwise produce stable payoffs. These designs are almost the exact opposite of the pure state claims which are optimal in the rational expectations framework but are roughly consistent with what one would expect given the decision making heuristics documented in the behavioural economics literature.