The identification of the average causal effects of a treatment in observational studies is typically based either on the unconfoundedness assumption or the availability of an instrument. When available, instruments may also be used to test for the unconfoundedness assumption (exogeneity of the treatment). In this project, we propose a method to non-parametrically test for the unconfoundedness assumption. We give conditions under which the test has power and we perform a simulation study and apply the results to an observational study of the effect of job practice on employment.