This paper analyzes the self-employment decision among Swedish-born male employees. The main objective of the paper is to investigate the impact of the relation between the actual and the predicted income on the probability to become self-employed. The predicted income is calculated from a standard income regression with controls for age, education, family status, family background and place of residence. By construction of a ratio between the actual and the predicted income we identify three groups of employees: (1) employees who have an actual income lower than the predicted income (underpaid), (2) employees with an actual income close to the predicted one, and (3) employees with an actual income higher than the predicted one (overpaid). The first question is if individuals who are "overpaid" or "underpaid" are more likely to become self-employed than those who are paid as we can expect. Our main finding is that employees who receive an income that differs from the one predicted by the income regression are more likely to become self-employed. We also analyse the effect of the ratio on four different measures of success as self-employed: income from self-employment, number of employees, turnover of the firm, and the probability to have a firm registered as a limited liability company. The general conclusion is that those who performed well as employees are also more successful as self-employed.