We show how tax kinks can be used to estimate the marginal propensity to consume (MPC). Tax kinks create discrete changes in the relationship between taxable income and disposable income, which – under a set of testable assumptions – enables causal identification of the spending response to changes in disposable income. We apply our new approach using administrative data from Denmark. Using a regression kink design (RKD), we estimate a MPC of 0.6 (s.e. 0.1) for taxpayers at the top tax kink. We show theoretically the conditions under which tax kinks provide variation in transitory or permanent income and deduce that the RKD exploits predominately transitory variation in income. Our implementation addresses threats to the research design from manipulation of taxable income through labor supply and other behavioral responses. Our findings inform the targeting of fiscal stimulus and social insurance programs to high income earners.