How Many People Actually Have Broadband?
Broadband service is critical to success and opportunity in the current economy — even before COVID. Why then is it so hard to find out what the actual situation is when it comes to who has it and who doesn’t? And what can be done about it?
According to the FCC, 14.5 million, or only 4.3% of Americans, don’t have access to fixed terrestrial broadband. But according to the Census’s ACS survey, 30% of American households don’t have fixed terrestrial high-speed broadband in their home. The drastic difference in these numbers is: 1) over-counting by the FCC, 2) households that don’t sign up for broadband, likely due to lack of resources and high costs, which is in turn due to lack of competition.
If you’re sold on the problem, maybe skip to a discussion of solutions below.
It helps to look at the big picture. According to the FCC Form 477 data, 99% of urban households have access to broadband. There are a few reasons this number is so high: 1) there is a well known flaw in the FCC data where once a provider covers one house in a census block, they’re considered to serve the whole block. (More on this below). 2) While I’ve taken out satellite service, fixed wireless is included here. I’m not aware of many apartment buildings, or even houses, where you could put an antenna on the roof with line-of-sight to a central point for service. Still, there is a lot of coax in cities, so 99% isn’t completely impossible.
What this chart brings into stark relief is we have a #DigitalDivide problem in urban and rural areas. Urban areas aren’t accessing the service, and rural areas don’t have service to access. Despite 99% access, only 65% of urban households have broadband service. This gap is almost certainly due to lack of competition, high prices, and the general systemic racism that you see everywhere when you start looking. Unless they are forced, there just isn’t an incentive for broadband providers to bring additional services to poor neighborhoods to lower prices.
Rural areas tell a different story. Only 84.7% of rural households have access to broadband according to the inflated FCC data. And only 56.6% have broadband service. This suggests that an increase in access will increase the number of households with service — though it is a large gap to close.
Let’s zoom in for some examples. Here’s a block where I lived in Brooklyn for a long time. The FCC data says I had access to 4 broadband services. Two are fixed wireless providers, one of whom claims gigabit symmetrical and the other of whom claims 200 Mbps symmetrical (hmm…). I remember calling one of them when I thought about sharing internet throughout the building (thankfully I thought better of it). The answer was they’d have to come out to roof and see if there was line of sight to their radio. With the amount of new construction, it wasn’t likely. Certainly not an option for an individual apartment. Then there’s regular cable internet from Charter. For a long time that was it — no options, in one of the densest areas of the country, with fiber everywhere.
I believe my building was the first on the block to get Fios. Watching Fios go in, you get an appreciation for how expensive it must be for Verizon to do this, and why they reneged on their agreement with NYC to provide everyone with service. They have to run the fiber underground from the main street down the side street, drill up the street into the building (two pipes — a primary and backup), install a huge switching panel, then drill through stairwells and run wires in hallways to get to an apartment door. All of that is done with contractors. A quick search on their website shows many of the other addresses on the block still don’t have access — but the FCC thinks they do.
Though it is minuscule in the aggregate, the FCC data can undercount broadband access as well. Let’s look at the first example FreePress found:
The FCC data says the 7 people who live there (and 5 unserved “locations”) don’t have access to broadband, despite most of the neighboring blocks having access through Cox, and in some cases Frontier. Whether none of these people have asked for service, or whether it is simply a data error on the part of the providers, it is agreeable that this community is probably not hurting for access.
A strident defender of these auction would probably point out that there is a reserve price for each block group. Theoretically, a reverse auction, especially when there is a provider that blankets coverage from space at almost 0 marginal cost, would lower the price in this easy-to-connect block. Indeed the winner in this block group, GeoLinks, went into the final round against SpaceX and two other satellite providers who bid lower but couldn’t compete against a Gigabit offering given the weighting for Gigabit and against slower speeds. Hopefully the FCC review process, or the provider’s inability to deliver, negates their winning. Regardless, even a potential $10,764 subsidy over 10 years is simply too much, and it should not have been included in the auction.
If we zoom out just slightly, we see how well this area is doing: 98.7% of the 943 households in the block group have access to broadband according to the FCC. More importantly, according to the ACS, 94% of households actually have broadband service. We might find a few unconnected people in block group 3 in this tract — it’s the Germany-shaped piece that looks like it has a ravine through it — 85.6% of households have broadband service but the FCC reports 98.7% access.
Data is imperfect — we knew that. While I’ve argued that we need to keep moving forward even with our imperfect data, we should also do something about it.
For years, FCC Commissioner Jessica Rosenworcel has pointed out that maps have problems. In her Pew speech, she recalls this Senate Commerce Committee hearing where Senators from both parties lambast the FCC’s maps.
Commissioner Rosenworcel outlines what she calls the “three c’s” of getting the broadband maps right: coordination between federal agencies, correct maps, and creative ideas. Under creative ideas, she tosses out a couple interesting ones including validating a provider’s offering before they’re funded, using postal trucks or other federal infrastructure to take real world samples (in this case wireless), or crowdsourcing to produce real world data.
Before we get into more ideas, it’s important to outline goals. There is so much inequity built into the current system — in the maps, but also in the incentives for providers — that it has to be an explicit goal of the new maps to break down those walls and build up new more fair ones. For example, there is no cost data on the Form 477 data, or in the ACS data. Nor is there latency data which is crucial to the actual internet experience, especially for learning and working online.
One thought: can modeling be used to smooth out some of the rough edges in the data? Given how well-connected the surrounding areas are, and the income of the residents, a model could almost certainly de-prioritize that Los Angeles golf club census block’s need for additional connectivity help. For example, exclude block groups that have > 95% probability that 90% or more or households have broadband service. No one is served by models obscuring what’s happening, so publish — as the FCC does well — the process, and post the eligible block groups before they’re finalized, as they did in the RDOF.
I can tell you for sure the available data will be just as good at predicting where the FCC data is overestimating broadband service. (Spoiler: it will be highly correlated with income and race ).
Another thought: I can use fast.com or speedtest.net to test my latency (ping times) against their servers. Why not reverse the process? Many routers will accept pings against them. Set up fairly distributed servers, and ping…ping…ping all day and all night, recording the times from these servers to end user routers. My router won’t respond to a ping, but if I choose my IP-neighbor by adding 1 to my IP, turn off my Wifi on my phone, I get 53ms avg to my neighbor.
It shouldn’t be a surprise that FCC broadband coverage data provided by 3,000 different ISPs with their own interests and data practices generates messy data. Let’s not be paralyzed by it.