Certain Silence

This GAO Report (pdf) got a lot of play last summer in the context of the Net Neutrality debate, because it called into question the FCC’s overly optimistic broadband numbers and broadband competition was a key talking point for both sides. I read it this weekend because I’m trying to get a sense of the problem, to understand the extent and shape of the “broadband digital divide.” It’s a fascinating read, for a couple of reasons that I’ll spread out over a couple of posts.

The first thing to note is that the report is not actually intended to evaluate the FCC’s broadband stats. As part of the Internet Tax Nondiscrimination Act  the GAO had to evaluate the impact of Internet taxation on broadband infrastructure deployment and broadband adoption, and they stumbled onto the FCC because they needed reliable baseline data on who had access to what. They quickly found that the FCC’s data was essentially useless.

The GAO’s methodology is a bit strange, it basically consists of a survey the bought from Knowledge Networks of a random sample of 3,000 Americans, with a 48% response rate (giving a charitable 7% margin of error on the resulting data), supplemented by a series of interviews with “stakeholders,” everyone from industry reps to state public utility commissions. The report thus slips back and forth between empirical data and industry conventional wisdom in a way that’s at best confusing and at times misleading. There’s also a bizarre process wherein, after the GAO finished its initial draft, it submitted the report to the FCC and the stakeholders for their feedback, “and incorporated it where appropriate.” I’m not sure what to make of that.

What’s especially crazy is that to evaluate the effect of Internet taxation, they had to use a binary variable (Internet tax? Yes/No) in their econometric modeling, because they didn’t have any information on broadband pricing. None. So they couldn’t account for differences in the rate of taxation in their analysis, and they couldn’t consider whether price influenced broadband adoption. The report is pretty much useless as an evaluative tool (your tax dollars at work!), but it illustrates problem number one in analyzing the “broadband digital divide.” No good data.