Many empirical studies challenged the parametric approach to estimate nonlinear house price models and to use instead nonparametric or semiparametric model specifications. In the area of environmental analysis using house price models, Stock (1991) and McMillen and Thorsen (2000) applied semi-parametric models to estimate the impact of hazardous wastes on house prices. Furthermore, estimating nonparametricaly house price models is gaining now additional momentum because Ekeland et al. (2002) provided a theoretical justification to use such an empirical approach to estimate hedonic price models. The objective of this paper is to assess the impact of agricultural pollution in a livestock intensive region of France using alternative non-parametric and semi-parametric hedonic price models. We propose a two stage estimation procedure were the first step is based on the partially linear model introduced by Robinson (1988). In the second step, once the parametric part of the model has been estimated on the house characteristics, we propose four specifications for the remaining functional form on the environmental variables. A parametric, a fully nonparametric, an additive, and a single index specifications are proposed and estimated using appropriate econometric procedures similar to those developed by Horowitz and Lee (2001), Iwata et al. (2000) and Robinson (1988). The performances of these models are examined, compared with respect to the benchmark nonparametric model and specification tests are proposed.