What is the relationship between the United States' uniquely democratic nominating procedure — the direct primary — and the incumbency advantage? I argue that the adoption of primary elections incentivizes legislators to cultivate and voters to use non-party reputations, leading to an increase in the incumbency advantage. To test this argument, I combine a regression discontinuity approach for estimating the incumbency advantage with a panel design to identify effects of direct primary adoption. The estimates from this differences-in-discontinuities design suggest that the adoption of the direct primary increased the incumbency advantage by about two percentage points. These results offer one possible institutional basis for the incumbency advantage in the US House of Representatives and provide new evidence for the central importance of primary elections in American politics.