Both theory and evidence suggest an ambiguous relationship between business tax compliance and geographic proximity to tax offices. We study this issue using a large-scale natural field experiment with Indonesia’s tax authority involving 12,000 micro, small, and medium enterprises (MSMEs). Businesses were randomly assigned to receive deterrence, information, or public goods letters, or no message. All letters improved compliance, with deterrence messages producing the largest gains - substantially increasing filing rates and raising monthly tax payments. Each dollar spent on deterrence letters generated about US$30 in additional revenue over the course of a year. We observe high compliance among non-treated MSMEs near metropolitan tax offices and find that enforcement messages successfully raise compliance in non-metropolitan regions to comparable levels. However, targeting already compliant MSMEs near metropolitan tax offices backfires, underscoring the need for geographically tailored tax administration strategies. These results provide novel experimental evidence on the relation between geographic proximity and the effectiveness of tax enforcement, helping to reconcile mixed findings in the tax compliance literature.