Our current knowledge of the causes of party unity rests heavily on the analysis of average unity scores of party groups from different countries. This study design invites two related problems: By aggregating unity scores we miss valuable variance at the level of disaggregated votes, and by comparing these aggregate scores across time and countries we might confound institutional effects with an unobserved case-specific selection bias of roll-call votes. In taking advantage of the laboratory-like conditions of the 16 sub-national parliaments of Germany and shifting the level of analysis to party unity in every single vote this article addresses both problems. Analysing 8607 unity scores, it is shown that the voting context is an important moderator of institutional effects on party unity. Specifically, it is shown that government status boosts party unity particularly within legislative important votes. Furthermore, the unity-boosting effect of slim majorities is only present for government parties and particularly strong when legislative consequential decisions are taken. Beyond that I also show that roll-call vote request, increasing ideological distances and norms of party loyalty increase party unity.