First look: Summary of the New FCC Broadband Maps

The FCC released the new maps on Friday of last week, and I’m ready to share the first version of an aggregated analysis. I’ll provide the top lines here and more detailed methodology and caveats are below. Overall, there are 112 million Broadband Serviceable Locations (BSLs) in the country (excluding territories). 6.9% of the BSLs are unserved, which is 7.7 million. 5.3% of the BSLs are underserved, or 6.0 million nationally.

The Unserved and Underserved numbers provided are how I expect the calculation to be done for the BEAD program: it excludes LEO satellite service, and also excludes service provided over unlicensed fixed wireless.

Here’s a chart for how this data might look as part of the BEAD allocation.

Here’s a spreadsheet with both the BDC numbers and the potential allocations.

It’s important to remember that the denominator in these calculations is BSLs, not housing units. BSLs include small businesses. Also, a single BSL could be an apartment building with many housing units in it, or a building of condos or townhouses that are 1 BSL for the purposes of the Fabric but have have multiple housing units. 112 million BSLs seems like a reasonable number to expect given that there 140 million housing units in the 2020 Census.

13.7 million unserved and underserved BSLs surprises on the low side, if anything. Consensus estimates were unserved and underserved at 21-23 million housing units (remembering again the different denominator between BSLs and housing units).

UPDATE 11/22: Only UT licensed FWA remains missing. (There are 5 files that appear to be missing some data. I haven’t included those files at all, so if those files are corrected I can rerun these numbers. In Colorado the file of cable ISP service is missing so I’ve used Missouri’s percent of unserved and underserved as a placeholder. Three of the missing files are licensed FWA (OK, UT, VT). Oregon is missing copper ISP locations served.)

As far I can tell, there isn’t a file that shows whether a location is residential, commercial, or both. So comparisons to Census data won’t be able to differentiate between commercial and residential BSLs at this time.

This is a tremendously large dataset. I’ve imported 632 million records of service at a location for this analysis (so far). It’s possible these numbers change as more files become available, or our understanding of what’s provided in the data improves. I’ll update this post or issue a new one with significant updates.

I’ve only begun to think about the types of analysis that should be done against this data. If you have ideas for analysis that you think would be interesting, drop a comment below.