Covid Alliance: County Mobility Flows
Enabling app developers, government teams, and policy analysts to work together to power insights on COVID-19
The COVID Alliance’s first released dataset in partnership with the Snowflake Data Exchange shows daily movement between counties in the US. This data gives state and local decision-makers a clear understanding of the ongoing interconnectedness of their municipalities with those around them.
The data gives an approximation of movement into each county from every neighboring county over the course of each day. It uses location data from unique mobile devices representing >10% of the population and county-level population data to closely approximate the total number of people moving across county borders. Counties where coverage of the initial population is too low to be representative have been excluded.
The main table here is county_flow, a compendium of 2M county-to-county flows across the country over the last 2 weeks (see sample attached).
For example, from the sample attached, we see that 3.7M people who have been in Orange County have also been in Los Angeles County sometime before their entrance into Orange County.
By contrast, looking deeper in the table, we see that there were approximately only 54 people in the US who traveled from Eureka County, Nevada to Marion County, Oregon.
- County_from — a unique county identifier representing where travel originated from maintained by the Alliance
- County_to — a unique county identifier representing the target of travel for the counts in the remaining columns
- County_from_name, state_from — The county and state where the travel originated from
- County_to_name, state_to — The county and state towards where travel was directed
- County_to_device_count — How many devices (phones, bluetooth, etc.) were observed in the county where travel originated in the last 2 weeks in the Covid Alliance’s database. This can be useful for normalization and scaling.
- County_to_inferred_pop — Census-rebalanced county population estimate derived from device
- Flow_into_county_observed — How many people traveled from county_from to county_to in the last two weeks as measured by all devices (phones, bluetooth, etc.) in the Covid Alliance’s database.
- Flow_into_county_inferred — How many people traveled from county_from to county_to in the last two weeks, rebalanced from Flow_into_county_observed according to census distribution to approximate the true population flow into the county.
- Flow_into_county_pct — How many people traveled from county_from to county_to in the last two weeks, expressed as a relative percentage with the population of county_to as the denominator.
- Note this number can be >1 for counties that are sparsely populated but where many people travel through, such as highway corridors.
- For example, the table indicates that 1716 people who have been in Kenedy County, Texas have previously been in Williamson, Texas, despite the census records showing a population of only 404 for Kenedy County.
Table: exchange.metadata.us_counties — Information about each US county catalogued by the Alliance.
- lat_min, lon_min, lat_max, lon_max — The bounding box of the county. Note the Alliance uses more detailed polygons for computing the county_flow table, but this offers a rough placement of where the county exists on the map.
Table: exchange.metadata.demographics — Copied from the Starschema, except it is annotated with Alliance’s internal county_id.
Please note that we have excised county-to-county flow for any counties where Flow_into_county_observed < 25 or Flow_into_county_inferred < 25 in compliance with our privacy principles.