As regulators sift through heaps of data to try to develop a national broadband plan, one key measurement they are trying to nail down is the average speed of the Internet services available to households across the country. A central question: Should network operators be held accoutable for the speeds they say they can provide or the actual speeds they deliver?
It is widely believed that Internet service providers like Verizon, AT&T and Comcast typically advertise that they offer much faster connections than the speeds actually experienced by their customers. Last week, Blair Levin, who leads the FCC's broadband task force, said while speaking on a panel I moderated that ISPs typically provide only half their advertised speeds. In August, the Commerce Department came under some fire for changing the broadband data reporting requirements for large carriers in the broadband stimulus grant process. Instead of having to report the actual speeds delivered to customers in a given market, carriers were allowed to only report the speeds they advertised in those markets. Public interest groups Public Knowledge and Free Press criticized the change, saying the resulting data would be innacurate.
As I was reading the Berkman Center's Broadband Study recently submitted to the FCC, I noticed some data that, at first glance, seems to contradict those conclusions. The study compared the advertised Internet speeds as reported by OECD to the actual Internet speeds reported by consumers using Speedtest.net, a Web site that lets users test their connection speeds. The chart below, which is on page 51 of the 232-page report, shows that the actual download speeds experienced by Internet subscribers were in fact faster--significantly faster, in some countries-- than the speeds advertised by carriers.
The study points out the data is far from reliable and gives a few reasons why. For example, users who know enough to measure their connection speeds likely have above-average Internet skills and, therefore, likely subscribe to faster services. And Speedtest.net users may be self-selecting because they have high speeds they want to test, so the results may be upwardly biased. Still, the report's authors say "these data are potentially useful for, at a minimum, offering an additional source of insight on actual performance of networks."
This is clearly only a sliver of the massive amount of information the FCC is trying to make sense of as it tries to find some concrete answers. But it shows the potential dichotomy between the real Internet landscape and widely held beliefs in the policy world. It also underscores the difficult task ahead of the FCC.