This paper examines to what extent non-random sorting of spouses affects earnings inequality while explicitly disentangling effects from increasing assortativeness in couple formation from changing patterns of couples' labor supply behavior. Using German micro data, earnings distributions of observed and randomly matched couples are compared to each other. Earnings of hypothetical couples are adjusted for changes in hours worked given the differences in the household context using predictions based on a structural model of labor supply. The main finding is that the impact of marital sorting on earnings inequality has been underestimated in previous approaches. Predicting hours worked for hypothetical couples reveals a strong disequalizing impact of nonrandom sorting on inequality which is stable since the 1980s. Taking labor supply choices as given would suggest a smaller effect. This suggests that increasing earnings correlation among couples is to a considerable extent driven by changing patterns of labor market behavior rather than changes in the assortativeness in couple formation.