Nonparametric modelling of biodiversity: Determinants of threatened species

This study uses a sample of 71 countries and nonparametric quantile and partial regressions to model a number of threatened species (reptiles, mammals, fish, birds, trees, plants) in relation to various economic and environmental variables (GDPc, CO₂ emissions, agricultural production, energy intensity, protected areas, population and income inequality). From the analysis and due to high asymmetric distribution of the dependent variables it seems that a linear regression is not adequate and cannot capture properly the dimension of the threatened species. We find that using OLS instead of non-parametric techniques over- or under-estimates the parameters which may have serious policy implications.