FCC Update on New Broadband Maps

and lingering concerns about changes to methodology and public access

Fifteen months ago I wrote a piece for Slate about the FCC’s new broadband maps, their importance, the magnitude of the changes, and some cautions about how hard it would be. Though it’s hard to believe, the stakes are even higher now. The FCC has been making admirable progress, and last week they opened up their system for ISPs to input coverage data against the new Broadband Serviceable Location Fabric. Two concerns I had at the time persist: 1) for all the focus on where ISP’s provide broadband service (the numerator), how many people and locations we are trying serve (the denominator) may be a more influential number; and 2) public access to this data in its raw form is important when big changes like this are being made quickly.

Currently, we measure the Homework Gap with this simple formula — how many Americans aren’t able to get broadband service at home, divided by the total number of Americans. In November 2020, the Pai FCC released numbers that said only 14.5 million Americans, 4.3% of the total, lacked access to broadband at the end of 2019. When the Rosenworcel FCC released the numbers as December 2020 with much less fanfare, the unserved numbers has shrunk by more than half, to 7.3 million Americans unserved, just over 2%. (We know these number underestimate the Gap; it likely about double the Form 477 estimates.)

The formula above matches with the policy objective: make sure Americans have broadband available at their homes so they aren’t in Taco Bell’s parking lot. (Hopefully they can afford the service, too).

Going forward, we’ll have a new congressionally-mandated formula:

Broadband serviceable locations (BSL) are, according to the Broadband DATA Act:

The Commission shall create a common dataset of all locations in the United States where fixed broadband internet access service can be installed, as determined by the Commission.

It’s becoming clear that these two denominators could be quite different. Any differences are material. Most of the BEAD funding is dispersed based on how many unserved BSLs each state has compared to the total nationally. As I showed in a previous post and spreadsheet, these allocations are very sensitive to changes in the number of relative unserved locations by state. We expect the allocations to be different than Form 477 data suggests because the new maps will be more granular; changes moving from population and housing units to BSLs may be a new dimension that catches people by surprise.

One potential difference is that a residential broadband serviceable location is the not the same as a housing unit. For example, an apartment building is a single broadband serviceable location. While the FCC will endeavor to include the number of housing units for each location, they’re only able to do so “to the extent such data are readily available”. If there isn’t a consistent relationship between BSLs and housing units (or population), we lose the ability to report on the Homework Gap as a percentage of residential population.

As if to highlight the complexity of adding non-residential locations, the FCC has created a distinction between locations that would generally purchase “mass-market” broadband Internet vs locations that “commonly subscribe to enterprise/business data broadband services.” While the intention is obvious - subsidies for deployment shouldn’t go to large office buildings served by business Internet plans - the implementation seems messy. And let’s remember that the BEAD subsidies are based on state level unserved location totals.

The amount of change, and these complications, only make public access to data more important. Congress authorized the FCC to sign a contract with a private company, and presumably the FCC’s contract with CostQuest specifies that they can’t release the raw location data publicly. While the FCC may make a map available to the public with individual locations, they should consider doing more. For example, the Fabric aggregated against 2020 Census blocks likely provides enough granularity for analysis without disclosing proprietary data.

With a dataset of the Fabric aggregated by Census block, I’d be looking to compare how the new Fabric compares to existing Census estimates of population and housing units, and how our current understanding of the unserved changes when Form 477 data is applied to new Fabric.

These are big changes in methodology. Let’s ensure we’re having a public conversation about what’s changing and how.