How It's Constructed
The firm SafeGraph has provided several datasets to researchers to help society respond to COVID-19. Their Social Distancing Metric database contains aggregated, anonymized, privacy-safe data on a range of spatial behaviors of mobile devices. No single indicator in the SafeGraph dataset adequately captures all aspects of mobility and engagement, and each is noisy and subject to idiosyncrasies.
Our index therefore summarizes the information in seven different variables, each measured daily at the county level and relative to its weekday-specific average over January–February. The variables are:
- Fraction of devices leaving home in a day.
- Fraction of devices away from home for three to six hours at a fixed location.
- Fraction of devices away from home longer than six hours at a fixed location.
- An adjusted average of daytime hours spent at home.
- Fraction of devices taking trips longer than 16 kilometers (10 miles).
- Fraction of devices taking trips less than 2 kilometers (1.2 miles).
- Average time spent at locations far from home.
These variables are combined via principal component analysis, which extracts a weighted average of the seven-variable series that best explains their variation. The resulting combination is our county-level MEI. We then aggregate the county-level MEIs to the metropolitan statistical area (MSA) and state and national levels. All indices are scaled so that the national index averages zero over January-February and is -100 for the week ended April 11. Regions with values less than -100 indicate their mobility fell more than the national average.
The MEI was developed by Tyler Atkinson, business economist; Jim Dolmas, senior research economist; Christoffer Koch, senior research economist; Evan Koenig, senior vice president; Karel Mertens, senior economic policy advisor; Anthony Murphy, senior economic policy advisor; and Kei-Mu Yi, senior vice president at the Federal Reserve Bank of Dallas.