Scott Wallsten

The next administration should launch a concerted broadband data-collection and analysis effort to support smart, timely, and informed decision-making by the Federal Communications Commission (FCC) and other agencies that work on broadband, such as the Rural Utilities Service. Specifically, the FCC should collect (or work with others to collect) comprehensive data on the following eight indicators:

The key missing component of nearly every proposal to solve the connectivity problem is evidence — evidence suggesting the ideas are likely to work and ways to use evidence in the future to evaluate whether they did work. Otherwise, we are likely throwing money away. Understanding what works and what doesn’t requires data collection and research now and in the future. It doesn’t have to be this way. The pandemic did not only lay bare the implications of the digital divide, it also created a laboratory for studying how best to bridge the divide.

We evaluate a program by a private Internet Service Provider intended to encourage low-income households to subscribe to broadband internet service.

New, more accurate and detailed broadband maps are on their way. The telecom policy crowd fervently hopes the data upgrade will help us better address digital divides and other issues. But maps and data alone won’t solve anything. Skill, expertise and time will all be required to study and use the new maps, and the resources required grow as the datasets become larger and more complex.

Senator Ed Markey (D-MA) introduced a bill with real potential to mitigate the digital divide. Most proposals simply call for more money for existing programs or for new programs without evidence they will help. Real-world experience, however, has demonstrated how little we truly understand about why many low-income people do not subscribe. The Markey bill tackles this underlying issue.

The seemingly interminable wait for the Court’s decision in Mozilla v. FCC is finally at an end. In its 186-page decision, the Court described how it considered economic concepts, arguments, and expert reports. It upheld the Federal Communications Commission’s 2018 Order but rejected its claimed preemption authority. As we have learned, the net neutrality debates will never die, but they may now change venue.

A long-standing public policy goal has been ensuring that almost all citizens are connected to some minimum level of communications services. This paper evaluates Comcast’s “voluntary commitment” to introduce a low-income broadband program that Comcast has branded “Internet Essentials (IE).” We use data from the US Census Current Population Survey (CPS) and the National Broadband Map and a differences-in-differences approach to evaluate the program’s effects on subscription rates for eligible households.

[Analysis] Criticisms of the 2017 Restoring Internet Freedom process apply also to the 2014 Open Internet process. A key difference between the comment process in 2017 compared to 2014 is that form letters became more sophisticated and more difficult to identify. In the future, bad actors are likely to continue improving their ability to make form letters appear unique, hide their origins, or simply make the comment process unmanageable.

[Commentary] Releasing the network neutrality draft order early had some unintended consequences. It created a flurry of activity when everybody with an opinion felt they had to re-litigate their arguments. Far too many chose it as an opportunity to hurl invectives at those with differing opinions, contributing to the downfall of productive debate. Despite the increasing vitriol during the weeks before the vote, releasing the draft order prior to the Federal Communications Commission’s vote is one key to making the FCC regulatory process more transparent.

By the time the Federal Communications Commission’s ferociously controversial net neutrality draft Order was released on November 22, 2017, more than 22 million comments were submitted to the FCC through its new application programming interface (API). This avalanche of public input is impossible to navigate and interpret using human labor alone. Machine learning tools are uniquely suited to navigating and interpreting such a large amount of information.