1: \begin{abstract}

2: This paper considers the nonparametric regression model with an additive error

3: that is dependent on the explanatory variables. As is common in empirical studies in

4: epidemiology and economics, it also supposes that valid instrumental variables are

5: observed. A classical example in microeconomics considers the consumer demand function

6: as a function of the price of goods and the income, both variables often considered as endogenous.

7: In this framework, the economic theory also imposes shape restrictions on the demand function,

8: like integrability conditions. Motivated by this illustration in microeconomics,

9: we study an estimator of a nonparametric constrained regression

10: function using instrumental variables by means of Tikhonov regularization.

11: We derive rates of convergence for the regularized model both in a deterministic

12: and stochastic setting under the assumption that the true regression function

13: satisfies a projected source condition including, because of the

14: non-convexity of the imposed constraints, an additional smallness condition.

15: \end{abstract}