Model Uncertainty and Liquidity

Extreme market outcomes are often followed by a lack of liquidity and a lack of trade. This market collapse seems particularly acute for derivative markets where traders rely heavily on a specific empirical model. Asset pricing and trading, in these cases, are intrinsically model dependent. Moreover, observed behavior of traders and institutions suggests that attitudes toward ``model uncertainty'' may be qualitatively different than Savage rationality would suggest. For example, a large emphasis is placed on ``worst-case scenarios'' through the pervasive use of ``stress testing'' and ``value-at-risk'' calculations. In this paper we use Knightian uncertainty to describe model uncertainty, and use Choquet-expected-utility preferences to characterize investors aversion to this uncertainty. We show that an increase in model uncertainty can lead to a reduction in liquidity as measured by the bid-ask spread set by a monopoly market maker. In addition, the non-standard nature of hedging model uncertainty can lead to broader portfolio adjustment effects like ``flight to quality'' and ``contagion.''