Accounts of why elites overthrow democracies usually turn on interest: how much a given elite loses under democratic rule and gains under a dictatorship. Naidu, Robinson and Young add a second variable. Even among people with identical interests, they argue, some are better placed than others to make a coup succeed, and position here means position in a social network.

Their model treats coups as generating rents and predicts that the more central a family sits in the elite network, the likelier it is to join. To test this they built a dataset of Haitian elite social ties and matched it against firm-level records for import businesses.

Central families, they find, were disproportionately involved in the 1991 overthrow of the democratic Aristide government. Afterward, retail prices rose faster for the staple goods those families imported, and children born in the cities during the years without democracy show worse health outcomes.