Generalized Additive Models for Nonmarket Valuation via Revealed or Stated Preference Methods

Single-site recreation demand and dichotomous choice contingent valuation analyses are typically conducted by implementing models containing strong parametric assumptions, which are rarely underpinned by theoretical arguments. This work illustrates how these assumptions can be relaxed and the estimation conducted semiparametrically by using generalized additive models (GAMs). This approach directly estimates the degree of the variables’ nonlinearities from the data, thereby avoiding subjective choices on the smoothing parameters and offering many advantages when compared to the conventional modeling techniques that dominate the environmental economics literature. Additionally, this paper illustrates how GAMs can be specified to construct theoretically consistent willingness-to-pay measures.