We exploit a recent Swedish tax reform, implying higher marginal tax rates for the top 5% of the earnings distribution, to learn about earnings responses to higher taxes. Using a simple and graphical cross-sectional method, we estimate an earnings elasticity of 0.16, evaluated four years after the reform. Our analysis indicates that the response is driven by adjustments in wage rates, and not by changes to hours, job switches or income shifting. We interpret the magnitude and preciseness of the response using a realistically calibrated simulation model in which people face uncertain marginal tax rates due to earnings dynamics.