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

This paper proposes a flexible modelling approach for either stated or revealed preference nonmarket valuation studies based on Generalized Additive Models (GAMs). GAMs encompass non-linear relationships via smooth functions, thereby avoiding imposing restrictive parametric specifications, which are rarely supported by economic theory. The smoothing parameters, which regulate the overall flexibility and the degree of non-linearity of the model, are directly estimated from the data using the penalized likelihood approach introduced by Wood (2000). Theoretically consistent WTP measures are derived using relatively simple modification of the standard algorithms proposed for parametric models. Two case studies, a contingent valuation and a travel cost application, illustrate the practical advantages of GAMs with respect to the conventional modelling techniques which dominate the literature