Research has demonstrated that couples have similar partisan preferences, a finding associated with political polarization. However, it remains debated to what extent different mechanisms contribute to this homogamy. Analyzing dyadic panel data from the German Socio-Economic Panel 1984–2020, we distinguish analytically between (1) direct political matching (i.e., partner selection on matching party preferences); (2) indirect political matching (i.e., social structural homogamy with political homogamy as a by-product); and (3) couples’ political alignment over time, to explain party preference similarity. First, we study matching among recently formed couples using an innovative method that compares real-world couples with three types of counterfactuals: couples that are matched (1) randomly, (2) by multidimensional social structural characteristics, and (3) by maximizing similarity in party preference. Second, we study couples’ political alignment over the course of relationships, tracking real-world couples over time and controlling for macro-level changes in the party-political landscape. Results indicate substantial political homogamy among recently formed couples, which is best explained by political matching (i.e., direct selection based on partisan preferences). Effects of social structural homogamy appear weak in comparison and rather stable across cohorts. Couples further align in their partisan preferences over time, but this effect is countered by an increasing heterogeneity of the German political landscape.