Economics Working Paper 15114
The federal tax deduction for tuition potentially increases investments in postsecondary education at minimal administrative cost. We assess whether it actually does this using regression discontinuity methods on the income cutoffs that govern eligibility for the deduction. Although many eligible households take nearly the maximum deduction allowed, we find no evidence that it affects attending college (at all), attending full- versus part-time, attending fourversus two-year college, the resources experienced in college, the amount paid for college, or student loans. Our analysis suggests that the deduction's inefficacy may be due to issues of salience, timing, and the method of receipt. We argue that the deduction might increase collegegoing if it were modified in simple ways that would not increase costs but would make it more likely to relax liquidity constraints and be perceived as a price change (which they is) as opposed to an income change. We outline how such modifications could be tested. This study has independent applied econometrics interest because households who would be just above a cut-off manage their incomes so that they fall slightly below it. This income management generates bias due to reverse causality, and we explore how to choose "doughnut-holes" that avoid bias without undue loss of statistical power.