We propose a new explanation for the persistence of gravity in international investment flows based on new facts about large cross-border commercial real estate transactions. Buyers in these transactions preferentially match with counterparties from own or proximate countries; such affinity-based matching helps alleviate financial investment frictions. We set up and structurally estimate a model of capital allocation in a decentralized market with an investment friction, which delivers the price, volume, and counterparty matching patterns in the data. The model shows that if clusters of high-affinity counterparties lie along historical routes, as in the data, preferential matching can perpetuate gravity relationships.