Using ACS's broadband adoption numbers for a real picture of the Digital Divide

The FCC data vastly overstates progress in closing the digital divide. And a close look at Census data shows it.

As Ajit Pai’s term as Chair of the FCC winds down, his supporters are trying to shape his legacy and accomplishments. There is bipartisan agreement that there’s one metric above all others: In a world where affordable broadband is critical to learning and success in the new economy, we have to quickly close the digital divide. That’s why FCC press releases from Pai’s office tout that Americans lacking broadband access “fell to 14.5 million — a 46% decrease from the end of 2016” which is only 4.3% of Americans lacking access. But this supposed progress in closing the digital divide is grossly overblown. The Census’s American Community Survey shows it’s really 35.9 million American households (~93.7 million people) without broadband service, 29.2% of all households, and a decrease of a mere 7.6% since 2016.

The ACS is a yearly detailed survey about all aspects of American life. The Census goes to great lengths to provide statistical rigor to this survey — and they accomplish their goal. But a close look provides more color to the broadband portion of the survey. If you look on the ACS’s site at the broadband section, they say 82.7% of American households have a broadband subscription, Pew surveys say 73%, and I’m claiming ACS says only 70.8% — what’s the difference. Well, if you look closely, the Census is counting a “cellular data plan” as a type of broadband.

The real 2019 number is “broadband such as cable, fiber optic, or DSL” and is nested under “broadband of any type”. A cellular plan with data caps, overages, and slow speeds is just not a permanent internet solution. I don’t know why the Census includes cellular (and satellite) in its broadband category in summary reports. It’s comforting to know that the actual ACS questionnaire uses common terms to identify types and levels of service. The FCC’s benchmark for high-speed broadband is “access to terrestrial (not satellite) fixed (not cellular/mobile) broadband with speeds of at least 25/3 Mbps” (parenthesis mine), and it’s clear that the matching category in the ACS data is this question 11.b.

So how are we doing at closing this actual broadband service divide? Not so great.

In 2016, there were 38.8 million American households without a real broadband subscription. In 2019, there were 35.9 million households — a drop of a mere 7.6%. The best performers aren’t the rural areas Pai leads you to believe: DC has 23% less households without broadband; suburban-heavy NJ has 13% less; and California and Arkansas have 11% less. Alaska, as they are happy to tell you, actually gained 13% in households without broadband. Montana, Vermont, Wyoming, and Delaware also gained small numbers of households without service (though likely within the margin of error).

Link: Clickable version of the above

A quick note on why Pai’s numbers are so different from the Census’s sampled numbers: I, and many others, have written about how the FCC systematically overestimates broadband coverage by counting a census block as having “access” if a single location in the block is covered. The word “access” there is important too. The ability to sign up for broadband doesn’t do any good if not affordable, and — especially in urban areas — there is a big gap between households who have access, and those that can afford to have service. The Census is asking about broadband service. If the digital divide is closed when all households in America have quality, affordable broadband, that’s a measure of service and the ACS is a more accurate metric.

The single biggest — maybe only — metric to measure FCC’s success in the Trump/Pai era is broadband service. Pai and his allies are using inflated numbers to put a positive spin on the impact of their corporate-friendly policies. The truth is we’re barely better than where he started.