The impacts of money in US politics have long been debated. Building on principal-agent models, we test whether and to what degree companies’ political donations lead to their favoured treatment in federal procurement. We expect the impact of donations on favouritism to vary by the strength of control by political principals over their bureaucratic agents. We compile a comprehensive dataset of published federal contracts and registered campaign contributions for 2004-2015. We develop risk indices capturing tendering practices and outcomes likely characterised by favouritism. Using fixed effects regressions, matching, and regression discontinuity analyses, we find confirming evidence for our theory. A large increase in donations from 10,000 USD to 5 million USD increases favouritism risks by about 1/4th standard deviation. These effects are largely partisan, with firms donating to the party that holds the presidency showing higher risk. Donations influence favouritism risks most in less independent agencies: the same donation increases the risk of favouritism by an additional 1/3rd standard deviation in agencies least insulated from politics. Exploiting sign-off thresholds, we demonstrate that donating contractors are subject to less scrutiny by political appointees.