We propose a Bayesian nonparametric instrumental variable approach that allows us to correct for endogeneity bias in regression models where the covariate effects enter with unknown functional form. Bias correction relies ...