Why Don't Taxpayers Bunch at Kink Points?

We introduce a new theory for optimizing frictions that prevents agents from choosing the threshold where the slope (kink) or intercept (notch) of a constraint changes. By making these frictions a function of observables, we can estimate why agents do not report exactly at predicted thresholds. Our new approach also estimates a tax elasticity that controls for the optimizing friction explicitly. Using U.S. tax returns data, we find that a) observables account for a substantial share of optimizing frictions, b) self-employment, capital gains, pension income, and tax credits are the most important optimizing frictions, and c) elasticity estimates that explicitly control for frictions are higher than estimates that do not