Recently several developing and transitional countries changed their personal income tax from fairly progressive to áat in an e§ort to improve e¢ ciency. But how do taxes a§ect incentives when people can sometimes tax evade and pay bribes? In this paper, I address this question by focusing on the e§ects of personal income tax progressivity on the decision to become self-employed. I develop a theoretical model of tax evading self-employed individuals who pay bribes to tax authorities. The model predicts that progressivity a§ects the decision to become self-employed even if people tax evade. I then test this prediction empirically using three sources of data. First, I use Russian longitudinal data and estimate the e§ects of progressivity on the individual decision to become self-employed. Second, I construct a data set of personal income tax schedules for 95 countries over 20 years and estimate the e§ects of progressivity on number of micro enterprises at the aggregate level. Third, I use Living Standards Measurement Surveys from 8 developing countries to estimate how much people are evading and the e§ect of progressivity on the amount that is not evaded. I Önd that increases in progressivity decrease the probability of choosing self-employment and decrease the number of micro enterprises. I also Önd that in countries with high tax evasion and frequent bribes, self-employment is less responsive to taxes than in the U.S.