Should the Average Tax Rate Be Marginalized?

Economic theory assumes that taxpayers use their true marginal tax rate (MTR) to guide their economic decisions. However, complexity of the personal income tax system implies that taxpayers may incorrectly perceive true marginal prices and incentives. We first develop an updating model that formalizes this proposition. A prediction of this model is that an unexpected innovation in the previous year's average tax rate (ATR) influences the perception of the MTR in the current year, even though the MTR is not in fact changing between the two years. This model generalizes the \schmeduling" hypothesis of Liebman and Zeckhauser (2004), who suggest that taxpayers use the ATR in place of the MTR in making their decisions. Then, assuming that taxpayers react to their perceived after-tax price as economic theory would suggest, we test this prediction empirically by examining whether household labor income responds to predictable (but not necessarily predicted) variation in the previous year's ATR due to eligibility for the Child Tax Credit, which depends on the exact timing of a child's 17th birthday. We find that household labor income decreases in response to losing eligibility for the Child Tax Credit. This finding is inconsistent with the rational taxpayer hypothesis, but consistent with the schmeduling hypothesis. Our robustness tests do not provide any consistent evidence that this result is entirely driven by an omitted variable bias due to a direct timing of birth effect. We also discuss the welfare consequences of schmeduling.