Assessing the heterogeneous causal effects of endogenous variables is of a strong interest in econometrics. In this context, the varying-coefficient models have a natural potential to model the heterogeneous effects as non-parametric functions. On the one hand, we can introduce the use of instruments in the varying-coefficient model framework with a control function approach. On the other hand, the usual conditional independence condition of instruments assumed in the fully non-parametric case can be relaxed to express the coefficient function with respect to the instruments. We illustrate our approach by estimating the effect of the American public health insurance Medicaid on the household savings by using an adaptation of an available and effective R package implementation.