Causal Recreation Demand Estimation with Cell Phone Mobility Data

Abstract

Estimating welfare impacts of environmental disruptions in recreation demand requires credible causal inference, yet zero market shares limit applications across regions and time. We integrate empirical Bayes posterior mean estimation into a two-step random coefficient logit model to incorporate sites with low or zero visitation without biasing demand estimates. We apply this framework to the 2021 Huntington Beach oil spill using high frequency cell phone mobility data and combine synthetic difference-in-differences with Bayes-adjusted shares to estimate causal effects on recreation demand and welfare. Results indicate welfare losses at Huntington Beach totaling $1.0 million, while nearby beaches recovered.