Abstract The Torrens (or registration) and the recording title systems apply different principles to resolve conflicting claims to land title. This paper develops a theoretical model of how expected title risk and transactions costs affect land value across the two systems and ultimately concludes that the Torrens system leads to higher property values, ceteris paribus. It also suggests that empirical studies of the title system–land value nexus need to control for self‐selection effects in the data. We use the simultaneous existence of two alternative title systems in Cook County, Illinois, as a natural experiment for comparing land values under each system. The estimates indicate that the Torrens system increases land value relative to the recording system when controlling for self‐selection effects.