[TOTM: The following is part of a blog series by TOTM guests and authors on the law, economics, and policy of the ongoing COVID-19 pandemic. The entire series of posts is available here.
This post is authored by Justin “Gus” Hurwitz, (Associate Professor of Law & Co-director, Space, Cyber, and Telecom Law Program, University of Nebraska; Director of Law & Economics Programs, ICLE).]
I’m a big fan of APM Marketplace, including Molly Wood’s tech coverage. But they tend to slip into advocacy mode—I think without realizing it—when it comes to telecom issues. This was on full display earlier this week in a story on widespread decisions by ISPs to lift data caps during the ongoing COVID-19 crisis (available here, the segment runs from 4:30-7:30).
As background, all major ISPs have lifted data caps on their Internet service offerings. This is in recognition of the fact that most Americans are spending more time at home right now. During this time, many of us are teleworking, so making more intensive use of our Internet connections during the day; many have children at home during the day who are using the Internet for both education and entertainment; and we are going out less in the evening so making more use of services like streaming video for evening entertainment. All of these activities require bandwidth—and, like many businesses around the country, ISPs are taking steps (such as eliminating data caps) that will prevent undue consumer harm as we work to cope with COVID-19.
The Marketplace take on data caps
After introducing the segment, Wood and Marketplace host Kai Ryssdal turn to a misinformation and insinuation-laden discussion of telecommunications policy. Wood asserts that one of the ISPs’ “big arguments against net neutrality regulation” was that they “need [data] caps to prevent congestion on networks.” Ryssdal responds by asking, coyly, “so were they just fibbing? I mean … ya know …”
Wood responds that “there have been times when these arguments were very legitimate,” citing the early days of 4G networks. She then asserts that the United States has “some of the most expensive Internet speeds in the developed world” before jumping to the assertion that advocates will now have the “data to say that [data] caps are unnecessary.” She then goes on to argue—and here she loses any pretense of reporter neutrality—that “we are seeing that the Internet really is a utility” and that “frankly, there’s no, uhm, ongoing economic argument for [data caps].” She even notes that we can “hear [her] trying to be professional” in the discussion.
Unpacking that mess
It’s hard to know where to start with Wood & Ryssdal discussion, such a muddled mess it is. Needless to say, it is unfortunate to see tech reporters doing what tech reporters seem to do best: confusing poor and thinly veiled policy arguments for news.
Let’s start with Wood’s first claim, that ISPs (and, for that matter, others) have long argued that data caps are required to manage congestion and that this has been one of their chief arguments against net neutrality regulations. This is simply not true.
Consider the 2015 Open Internet Order (OIO)—the net neutrality regulations adopted by the FCC under President Obama. The OIO discusses data caps (“usage allowances”) in paragraphs 151-153. It explains:
The record also reflects differing views over some broadband providers’ practices with respect to usage allowances (also called “data caps”). … Usage allowances may benefit consumers by offering them more choices over a greater range of service options, and, for mobile broadband networks, such plans are the industry norm today, in part reflecting the different capacity issues on mobile networks. Conversely, some commenters have expressed concern that such practices can potentially be used by broadband providers to disadvantage competing over-the-top providers. Given the unresolved debate concerning the benefits and drawbacks of data allowances and usage-based pricing plans,[FN373] we decline to make blanket findings about these practices and will address concerns under the no-unreasonable interference/disadvantage on a case-by-case basis.
[FN373] Regarding usage-based pricing plans, there is similar disagreement over whether these practices are beneficial or harmful for promoting an open Internet. Compare Bright House Comments at 20 (“Variable pricing can serve as a useful technique for reducing prices for low usage (as Time Warner Cable has done) as well as for fairly apportioning greater costs to the highest users.”) with Public Knowledge Comments at 58 (“Pricing connectivity according to data consumption is like a return to the use of time. Once again, it requires consumers keep meticulous track of what they are doing online. With every new web page, new video, or new app a consumer must consider how close they are to their monthly cap. . . . Inevitably, this type of meter-watching freezes innovation.”), and ICLE & TechFreedom Policy Comments at 32 (“The fact of the matter is that, depending on background conditions, either usage-based pricing or flat-rate pricing could be discriminatory.”).
The 2017 Restoring Internet Freedom Order (RIFO), which rescinded much of the OIO, offers little discussion of data caps—its approach to them follows that of the OIO, requiring that ISPs are free to adopt but must disclose data cap policies. It does, however, note that small ISPs expressed concern, and provided evidence, that fear of lawsuits had forced small ISPs to abandon policies like data caps, “which would have benefited its customers by lowering its cost of Internet transport.” (See paragraphs 104 and 249.) The 2010 OIO makes no reference to data caps or usage allowances.
What does this tell us about Wood’s characterization of policy debates about data caps? The only discussion of congestion as a basis for data caps comes in the context of mobile networks. Wood gets this right: data caps have been, and continue to be, important for managing data use on mobile networks. But most people would be hard pressed to argue that these concerns are not still valid: the only people who have not experienced congestion on their mobile devices are those who do not use mobile networks.
But the discussion of data caps on broadband networks has nothing to do with congestion management. The argument against data caps is that they can be used anticompetitively. Cable companies, for instance, could use data caps to harm unaffiliated streaming video providers (that is, Netflix) in order to protect their own video services from competition; or they could exclude preferred services from data caps in order to protect them from competitors.
The argument for data caps, on the other hand, is about the cost of Internet service. Data caps are a way of offering lower priced service to lower-need users. Or, conversely, they are a way of apportioning the cost of those networks in proportion to the intensity of a given user’s usage. Higher-intensity users are more likely to be Internet enthusiasts; lower-intensity users are more likely to use it for basic tasks, perhaps no more than e-mail or light web browsing. What’s more, if all users faced the same prices regardless of their usage, there would be no marginal cost to incremental usage: users (and content providers) would have no incentive not to use more bandwidth. This does not mean that users would face congestion without data caps—ISPs may, instead, be forced to invest in higher capacity interconnection agreements. (Importantly, interconnection agreements are often priced in terms of aggregate data transfered, not the speeds of those data transfers—that is, they are written in terms of data caps!—so it is entirely possible that an ISP would need to pay for greater interconnection capacity despite not experiencing any congestion on its network!)
In other words, the economic argument for data caps, recognized by the FCC under both the Obama and Trump administrations, is that they allow more people to connect to the Internet by allowing a lower-priced access tier, and that they keep average prices lower by creating incentives not to consume bandwidth merely because you can. In more technical economic terms, they allow potentially beneficial price discrimination and eliminate a potential moral hazard. Contrary to Wood’s snarky, unprofessional, response to Ryssdal’s question, there is emphatically not “no ongoing economic argument” for data caps.
Why lifting data caps during this crisis ain’t no thing
Even if the purpose of data caps were to manage congestion, Wood’s discussion again misses the mark. She argues that the ability to lift caps during the current crisis demonstrates that they are not needed during non-crisis periods. But the usage patterns that we are concerned about facilitating during this period are not normal, and cannot meaningfully be used to make policy decisions relevant to normal periods.
The reason for this is captured in the below image from a recent Cloudflare discussion of how Internet usage patterns are changing during the crisis:
This image shows US Internet usage as measured by Cloudflare. The red line is the usage on March 13 (the peak is President Trump’s announcement of a state of emergency). The grey lines are the preceding several days of traffic. (The x-axis is UTC time; ET is UCT-4.) Although this image was designed to show the measurable spike in traffic corresponding to the President’s speech, it also shows typical weekday usage patterns. The large “hump” on the left side shows evening hours in the United States. The right side of the graph shows usage throughout the day. (This chart shows nation-wide usage trends, which span multiple time zones. If it were to focus on a single time zone, there would be a clear dip between daytime “business” and evening “home” hours, as can be seen here.)
More important, what this chart demonstrates is that the “peak” in usage occurs in the evening, when everyone is at home watching their Netflix. It does not occur during the daytime hours—the hours during which telecommuters are likely to be video conferencing or VPN’ing in to their work networks, or during which students are likely to be doing homework or conferencing into their meetings. And, to the extent that there will be an increase in daytime usage, it will be somewhat offset by (likely significantly) decreased usage due to coming economic lethargy. (For Kai Ryssdal, lethargy is synonymous with recession; for Aaron Sorkin fans, it is synonymous with bagel).
This illustrates one of the fundamental challenges with pricing access to networks. Networks are designed to carry their peak load capacity. When they are operating below capacity, the marginal cost of additional usage is extremely low; once they exceed that capacity, the marginal cost of additional usage is extremely high. If you price network access based upon the average usage, you are going to get significant usage during peak hours; if you price access based upon the peak-hour marginal cost, you are going to get significant deadweight loss (under-use) during non-peak hours).
Data caps are one way to deal with this issue. Since most users making the most intensive use of the network are all doing so at the same time (at peak hour), this incremental cost either discourages this use or provides the revenue necessary to expand capacity to accommodate their use. But data caps do not make sense during non-peak hours, when marginal cost is nearly zero. Indeed, imposing increased costs on users during non-peak hours is regressive. It creates deadweight losses during those hours (and, in principle, also during peak hours: ideally, we would price non-peak-hour usage less than peak-hour usage in order to “shave the peak” (a synonym, I kid you not, for “flatten the curve”)).
What this all means
During the current crisis, we are seeing a significant increase in usage during non-peak hours. This imposes nearly zero incremental cost on ISPs. Indeed, it is arguably to their benefit to encourage use during this time, to “flatten the curve” of usage in the evening, when networks are, in fact, likely to experience congestion.
But there is a flipside, which we have seen develop over the past few days: how do we manage peak-hour traffic? On Thursday, the EU asked Netflix to reduce the quality of its streaming video in order to avoid congestion. Netflix is the single greatest driver of consumer-focused Internet traffic. And while being able to watch the Great British Bake Off in ultra-high definition 3D HDR 4K may be totally awesome, its value pales in comparison to keeping the American economy functioning.
Wood suggests that ISPs’ decision to lift data caps is of relevance to the network neutrality debate. It isn’t. But the impact of Netflix traffic on competing applications may be. The net neutrality debate created unmitigated hysteria about prioritizing traffic on the Internet. Many ISPs have said outright that they won’t even consider investing in prioritization technologies because of the uncertainty around the regulatory treatment of such technologies. But such technologies clearly have uses today. Video conferencing and Voice over IP protocols should be prioritized over streaming video. Packets to and from government, healthcare, university, and other educational institutions should be prioritized over Netflix traffic. It is hard to take anyone who would disagree with this proposition seriously. Yet the net neutrality debate almost entirely foreclosed development of these technologies. While they may exist, they are not in widespread deployment, and are not familiar to consumers or consumer-facing network engineers.
To the very limited extent that data caps are relevant to net neutrality policy, it is about ensuring that millions of people binge watching Bojack Horseman (seriously, don’t do it!) don’t interfere with children Skyping with their grandparents, a professor giving a lecture to her class, or a sales manager coordinating with his team to try to keep the supply chain moving.
Gus: This is fabulous and I wish I had the thought and data to do it. Just two quick points that may help. One is that to the extent peak-period congestion is the issue, the best response in theory is peak-period pricing, not data caps (see electricity). I don’t know why time-of-use pricing isn’t feasible or used; you or others may know. A second is that I don’t think most or any data caps have been hard, but rather that one can go beyond a data cap but at a higher price. If I’m right, a cap is kind of the opposite of a volume discount, but still consistent with efficient and probably equitable price discrimination.
Thanks Tim! And I agree completely. Data caps are a just an ok-ish proxy for time-of-day or peak-hour pricing. Given that re’s a strong correlation between high data use generally and high data use during peak hours, it’s a decent proxy — and one that is reasonably intuitive to consumers. Consumer intuition is an important point, because we want to avoid complex usage arithmetic for consumers.
In the long run, and no matter whether Internet access is treated as a utility or not, I am confident that we will eventually transition to a more electricity-like model. Everyone will have more capacity than could ever be needed by a typical consumer household (and, really, 1gbps symmetric probably satisfies that) and users will be billed on a usage basis. But we’re not there yet, given the number of unconnected and underconnected households.
On the consumer perception point, see also this post by Josh Gans, https://digitopoly.org/2015/10/20/broadband-usage-caps-are-probably-inefficient/ , and my comment in response, part of which was:
“I don’t think it requires much optimism – just an appreciation for the regulatory overlay. The FCC very nearly prohibited usage caps outright in the Open Internet Order, and reserved for itself the right to monitor them closely, and limit or prohibit them if they so choose. Three years ago, when Comcast first started testing usage tiers, Senator Wyden floated a bill that could have nipped them in the bud. Mobile network operators are under constant attack from regulatory advocates for their use of data caps. And so on, and so on. Now, how long do you think Comcast would have been able to offer data tiers that kicked in at a level that would have included, say, 50% of all broadband users? Not very long. By introducing metering at the 300GB level, Comcast keeps access “open” for 98% of users, and can claim (rightly) that instead of casual users subsidizing access by the 2%, it’s the other way around.
IOW, the possible *regulatory* response to pricing mechanisms that might be misunderstood by consumers might also help to explain the current system versus more granular usage-based pricing.
Great points, Tim. On the latter, indeed this is why Berin and I in a related previous post years ago tried to home in on the notion of “tiers” versus “caps”.
On the first point, from that post:
“Pricing structures for broadband are still evolving. Just this year, Comcast moved from its original 250 GB cap—which it never enforced—to today’s 300 GB basic tier, and other broadband providers will likely follow suit. Those plans will probably continue to evolve towards pricing structures that minimize network congestion—like offering periods of unmetered use in the middle of the night, when network use plummets. That would go a long way to allaying concerns about the effect of tiered plans on competition, since Netflix could send your favorite shows and the next movies in your queue to the device of your choice while you sleep. But pricing structures also have to allow sensible, fair recovery of costs—which the Wyden bill would simply ban.”