1: \begin{abstract}

2: We extend nonparametric regression smoothing splines to a context where there is endogeneity and instrumental variables are available.  Unlike popular existing estimators, the resulting estimator is one-step and  relies on a unique regularization parameter. We derive uniform rates of the convergence for the estimator and its first derivative. 

3: We also address the issue of imposing monotonicity in estimation.

4: Simulations confirm the good performances of our estimator compared to two-step procedures. Our method yields economically sensible results when used to estimate Engel curves.

5: \end{abstract}

6: