The past decade has witnessed significant advances in the microeconometric analysis of recreation data. In this review, we focus on two areas in which these innovations have been especially prolific: accounting for unobserved preference heterogeneity and controlling for unobserved and possibly endogenous site characteristics, such as congestion. Failure to appropriately address these issues with the nonlinear models typically used in recreation demand analysis can severely bias parameter and welfare estimates. We consider these issues of widespread importance within and beyond recreation demand applications. We also expect these estimation challenges to become more ubiquitous as the field gradually moves toward region-wide, multisite applications in reaction to large-scale environmental changes.