This study contributes to the literature that highlights the penalties of education-occupation mismatch in terms of earnings across different employment statuses. Most existing literature analyzing the education-occupation mismatch has focused on paid employees, overlooking self-employed individuals, and has not controlled for sample selection bias and unobserved heterogeneity bias simultaneously. Therefore, the objective of this study is to analyze the impact of education mismatch on earnings across different employment statuses after correcting for both sample selection bias and unobserved heterogeneity bias. To achieve this objective, we applied the methodology of Duncan and Hoffman (1981) to the Pakistan Social and Living Standards Measurement (PSLM), 2019-20. Our results show that after controlling for unobserved heterogeneity bias and sample selection bias, overeducation has no positive value for both paid employees and the self-employed. The returns from overeducation based on the OLS model might be overestimated if overeducated workers possess lower average ability levels, whereas the returns of adequately educated individuals increase after correcting for the bias and are significantly higher for self-employed individuals compared to paid employees.