High-frequency smartphone location data lets me observe millions of real visitation decisions, to recreation sites, grocery stores, clinics, and neighborhoods, at a scale and granularity survey data can't match. I combine these mobility panels with structural demand models, difference-in-differences, and other causal designs to recover willingness-to-pay for environmental quality, food access, and local amenities, and to separate genuine behavioral change from the redistribution of activity across nearby locations.
The throughline across my projects is not a single model but a shared way of using mobility data as a revealed-preference instrument, paired with whatever identification strategy the setting calls for, applied across the areas below.
I use mobility pings as a revealed-preference instrument for recreation demand — recovering willingness-to-pay for water quality, quantifying the cost of site closures and disruptions, and measuring how disasters and environmental hazards reshape where people spend their time.
I study how grocery access, dollar-store entry, and food retail shocks affect household nutrition and welfare, pairing mobility-based shopping trips with scanner-level purchase data and structural demand models to separate demand-side from supply-side drivers of food access gaps.
Behind each project is a proprietary smartphone mobility panel paired with public administrative or geographic data, cleaned and matched to a choice set of destinations. The methods vary with the question, but the data pipeline is largely shared, and figuring out its promises and pitfalls is itself part of the research agenda. If you're curious about the broader range of economic applications this kind of data supports, our survey paper below is a good place to start.
Polygon-based POI panel with weekly/monthly visit counts. Used to define choice sets and measure visitation across recreation, retail, and restaurant contexts.
Data ↗Timestamp-level device trajectories with customizable POI definitions, enabling fine-grained visit-episode identification within a secure compute environment.
Data ↗Raw device-ping data accessible for local computation, enabling large-scale individual mobility modeling and custom trip construction.
Data ↗I'm always glad to hear from researchers working on food access, environmental amenities, or urban and infrastructure questions where mobility data could help. If that's you, feel free to reach out.