We document three facts: (i) Higher educated workers are more likely to moonlight; (ii) conditional on education, workers with higher wages are less likely to moonlight; and (iii) the prevalence of moonlighting is declining over time for all education groups. We develop an equilibrium model of the labor market to explain these patterns. A dominating income effect explains the negative correlation of moonlighting with productivity in the cross section and the downward trend over time. A higher part-to-full time pay differential for skilled workers (a comparative advantage) explains the positive correlation with education. We provide empirical evidence of the comparative advantage using CPS data. We calibrate the model to 1994 cross-sectional data and assess its ability to reproduce the 2017 data. The driving forces are productivity variables and the proportion of skilled workers. The model accounts for 56% of the moonlighting trend for skilled workers, and 67% for unskilled workers.