Mobile device location data offer a low-cost alternative for measuring visitation to outdoor recreation sites and are known to correlate with official visitation counts. Less is known about whether these data can recover recreation demand and consumer surplus comparable to surveybased methods. We compare travel cost models estimated using mobile device and survey data for 17 U.S. National Park Service sites. Results are mixed. We examine the roles of aggregation and sampling bias and use LASSO regression to assess whether sample and site characteristics explain discrepancies.