How do politicians enforce vote-buying agreements when ballots are secret? This asymmetric information problem lies at the heart of the vote-buying literature, yet existing research has largely described and examined individual enforcement strategies in isolation, rather than their relative effectiveness. We address this gap and compare the effectiveness of five common enforcement strategies in a randomized vignette experiment in Angola: turnout buying, questioning ballot secrecy, machine politics, reciprocity, and a novel signaling mechanism. We find that only tangible enforcement strategies—monitoring turnout and casting doubt on ballot secrecy—significantly increase voting intentions. By contrast, clientelistic enforcement strategies based on promises of future benefits have no effect and even backfire among politically interested voters. Our results suggest that the effectiveness of vote buying depends on the type of enforcement employed and that strengthening ballot secrecy and voter information may reduce the effectiveness of vote-buying. A complementary list experiment confirms the prevalence and provides robustness for our measurement of vote-buying.