Thursday, April 11, 2019

Half of All U.S. Households are Not Using 25 Mbps Speeds? Impossible

Many complain that Federal Communications Commission data on broadband speeds is incomplete, misleading or wrong. Fair enough. The data will likely never be as good as gathered by Speedtest and other organizations that test actual user sessions, in the concrete. On the other hand, those tests are not “scientific” in the sense of using controlled, weighted samples.

But just a bit of logic suggests many of the complaints about U.S. broadband speeds cannot be correct, either.

Roughly 10 percent of U.S. households are in rural areas, the places where it is most expensive to install fast fixed network internet access facilities, and where the greatest speed gaps--compared to urban areas--almost certainly continue to exist.

In its own work with TV white spaces, Microsoft has targeted perhaps  two million people, or roughly a million households, that have no fixed network internet access. That assumes there are two people living in a typical household, which is below the U.S. average of roughly 2.3 to 2.5 per household.

Recall that the definition of broadband is 25 Mbps downstream. Microsoft has argued that 20 million people (about 10 million homes) or perhaps eight percent of the population (perhaps four percent of homes) cannot get such speeds from any fixed network service provider.

Ignoring for the moment access at such speeds by satellite, fixed wireless or mobile, that is about the dimensions of the potential rural broadband problem. Perhaps nobody would dispute a potential speed gap for those 10 million or so homes.
But Microsoft claims about half of U.S. residents cannot get 25 Mbps service. That is hard to believe. There simply are not enough rural households to create a gap that large when urban and suburban areas, where 92 percent of people live, have access to speeds far higher than 25 Mbps.

In 2018, U.S. fixed internet service provider speeds averaged 96 Mbps downstream, according to Speedtest.


We can freely admit that the FCC data is based on sampling and inferences from that reported data. But we should at least be skeptical, and apply some tests of sanity, when claims of the magnitude of the speed gaps are made. But Ookla’s data is likely quite representative of internet users.

Consider the incentive to test a connection speed. What sort of user is most likely to do so, and when? In my own experience, people test when they suspect a problem, not because they simply want to enjoy how fast their connections are. If anything, the Ookla tests should overemphasize tests by people who suspect they have a problem (slow router, for example).

Personally, I only test when I think I have a problem.

It simply is not mathematically possible for half of U.S. homes to be using less than 25 Mbps when the number of homes in that range are assumed to be rural households, which represent less than 10 percent of total homes.

Even if 100 percent of rural homes could not buy 25 Mbps service, such locations only represent eight percent or so of all locations.

When average speeds are above 100 Mbps nationwide, it is not reasonable to conclude that those eight percent of locations--even if all were using connections slower than 25 Mbps--could represent half of all households.

Microsoft’s claims seem impossible to believe, even if we all agree the FCC data is incomplete or even wrong.

No comments:

Many Winners and Losers from Generative AI

Perhaps there is no contradiction between low historical total factor annual productivity gains and high expected generative artificial inte...