Link Search Menu Expand Document

COVIDcast Geographic Coding

The geo_value field returned by the API specifies the geographic location whose estimate is being reported. Estimates are available for several possible geo_types:

  • county: County-level estimates are reported by the county’s five-digit FIPS code. All FIPS codes are reported using pre-2015 FIPS code assignments, except for FIPS codes used by the jhu-csse and usa-facts sources. These are reported exactly as the sources report their data; see below. FIPS codes ending in 000 are not valid counties, and instead represent “megacounties” we construct; see below.
  • hrr: Hospital Referral Region, units designed to represent regional health care markets. There are roughly 300 HRRs in the United States. A map is available here. We report HRRs by their number (non-consecutive, between 1 and 457).
  • msa: Metropolitan Statistical Area, as defined by the Office of Management and Budget. The Census Bureau provides detailed definitions of these regions. We report MSAs by their CBSA ID number.
  • dma: Designated Market Areas represent geographic regions with their own media markets, as defined by Nielsen.
  • state: The 50 states, identified by their two-digit postal abbreviation (in lower case). Estimates for Puerto Rico are available as state pr; Washington, D.C. is available as state dc.

Some signals are not available for all geo_types, since they may be reported by their original sources with different levels of aggregation.

Table of contents

  1. Small Sample Sizes and “Megacounties”
  2. Coding Exceptions

Small Sample Sizes and “Megacounties”

Most sources do not report the same amount of data for every county; for example, the survey sources rely on survey responses each day, and many counties may have comparatively few survey responses. We do not report individual county estimates when small sample sizes would make estimates unreliable or would allow identification of respondents, violating privacy and confidentiality agreements. Additional considerations for specific signals are discussed in the source and signal documentation.

On each day, in each state, we collect together the data from all counties with insufficient data to be individually reported. These counties are combined into a single “megacounty”. For example, if only five counties in a state have sufficient data to be reported, the remaining counties will form one megacounty representing the rest of that state. Megacounty estimates are reported with a FIPS code ending with 000, which is never a FIPS code for a real county. For example, megacounty estimates for the state of New York are reported with FIPS code 36000, since 36 is the FIPS code prefix for New York.

These megacounty estimates are used on our COVIDcast map and in the county maps produced by our API clients, to color in the background of states and graphically represent the “rest of” states whose counties are not all individually reported.

Warning: As sample sizes vary from day to day, the counties composing the megacounty can vary daily; the geographic area covered by the megacounty is simply the state minus the counties reported for that day. The megacounty construction also depends on the specific source and signal, so on one day, megacounty 36000 can cover a different geographic area for the doctor-visits source than it does for the fb-survey source. Do not try to compare megacounty estimates across time or between signals.

Coding Exceptions

  1. The cases and deaths data from JHU CSSE has some geographic exceptions in its coding and reporting; see its documentation for more details.
  2. The cases and deaths data from USAFacts also has geographic exceptions; see its documentation for details.