We use detailed micro-level data on consumption and income for a 1998-2004 panel of Russian households to study the effects of the flat income tax reform in 2001. We show that the gap between household expenditures and reported income is a meaningful measure of tax evasion. We use the difference-indifference and regression discontinuity approaches to assess the response of tax evasion and worker productivity to the flat tax. We find that the tax evasion response (10-12%) is larger than the productivity response (0-4%), and thus increased tax revenues and reported earnings are largely driven by improved tax compliance.