Governments default on their creditors in a variety of ways: Some cease payment on foreign bondholders; others default on domestic lenders, repay only part of their debt, or inflate it away. Existing models by economists assume that default is a strategy to appropriate foreign capital, ignore domestic politics, and thus fail to explain almost a third of all cases of non-payment of sovereign debt since 1983. In this paper, we present a formal model explaining the choice of a government on which creditors to default and whom to spare. We find that the probability of default on different creditor types varies with the ability of governments to tax different domestic voter groups. As our model predicts a nonlinear relationship, we use double machine learning to test our predictions. We illustrate the underlying mechanisms with two short case studies comparing Ghana (which defaulted on domestic and foreign creditors) and Zambia (which spared domestic lenders).