AbstractWe attempt to reconcile two puzzles in the non‐profit (NP) literature: (i) studies often find that NPs grossly underspend on fundraising; (ii) crowd‐out effect estimates of government grants on private donations vary widely, suggesting both crowd‐out and crowd‐in. We examine these impacts via a semi‐parametric instrumental variables strategy. We find heterogeneous effects across NPs, fundraising efforts, and government grant intensity. Smaller NPs appear to drive the presence of crowd‐out; estimates for the marginal productivity of fundraising are consistent with leading theories of NP optimizing behaviour.