Ambiguity and Reality

Model builders face ambiguity about the true data generating process. Consequently, they need to deal with ambiguity attitudes (inside uncertainty) and ambiguous financial reality (outside uncertainty) when developing and estimating financial models. We introduce a novel approach for systematically dealing with outside uncertainty in addition to inside uncertainty in a tractable way. By bounding the effects of ambiguous data features, we avoid the adverse consequences of outside uncertainty, such as strongly biased equity premiums and investment policies. In a real data application, we show that asset managers can be more reliably evaluated using our bounded-influence approach.