This study improves upon the standard "dummy variable" approach to modeling fish consumption advisories by jointly estimating a "perceived hazard" model and a site choice model. The perceived hazard model overcomes the shortcomings of the dummy variable model, namely that all anglers respond equally to advisories and that all anglers know of and believe the advisories. We find that anglers' perceived hazards associated with consumption advisories do affect product (recreational site) choice. Anglers' perceptions also affect welfare measures, where the benefits of contaminant removal follow a more reasonable pattern than that of the dummy variable approach. The joint perceived hazard/product choice model is applicable to a wide variety of risky choices with which consumers are faced.