Evaluating Precision, Privacy, and Representation with Cell Phone Data

Abstract

While mobility data has emerged as a promising alternative for assessing the economic value of recreation, their reliability depends critically on how data are processed and protected. This paper systematically evaluates how key data-handling practices affect the accuracy of recreation demand estimation. Using a random utility model to analyze recreation visits at Cape Cod beaches from 2019-2022, we evaluate how methodological choices influence the marginal willingness to pay (MWTP) for avoiding fecal bacteria contamination. We find an average MWTP of $8.92 per visit when using the proposed practices, such as refined visit definitions, sampling weights, and long-term choice sets. Deviations from these practices can introduce significant biases: relaxing the minimum dwell time and applying differential privacy reduce MWTP by 57% and 65%, respectively, while short-term choice set definition inflates it by 10%. By demonstrating the sensitivity of welfare estimates to data-processing decisions, this study highlights the importance of transparent and judicious mobility data practices for credible environmental valuation and evidence-based policymaking.