Married men commonly earn more than unmarried. Explanations emphasize either productivity gains from marriage or positive selection into marriage. Using longitudinal data on men in Sweden 1947–2022, combining historical population records for a regional sample with full-count modern administrative registers, we analyze information on earnings and occupation for both men and their spouses. We estimate linear regression, two-way fixed-effects, event-study models, and residualized quantile regression to capture aspects of the marriage premium. We find a stable marriage premium over 70 years. While OLS estimates are substantially attenuated in fixed-effects models, indicating an important role for selection, a positive premium remains. There were large changes over time in the underlying mechanisms. Until the 1980s, the premium was highest for men with lower earnings, while this pattern reversed after 1990 when the premium was larger at higher earnings. Before 1970 there is some support for a causal marriage premium, while this changed in later periods, when there was strong earnings growth in the years leading up to marriage. The persisting marriage premium suggests that marriage is still crucial for men’ labor market outcomes.