Every time the net neutrality debate flares up, the ISP industry and its anti-net neutrality allies come up with some reason why leaving unfettered gatekeeper power in the hands of the people who invented the cable video bundle is awesome rather than something that needs oversight to prevent rip offs and anticompetitive behavior. It used to be “net neutrality/Title II will kill investment.” This claim has been repeatedly disproven (you can see some Free Press explanation for why this is nonsense here, here and here). Furthermore, Covid showing the truly massive dimensions of the persistent digital divide has largely discredited “deregulation will spur investment — really!” to all but the most diehard true believers.
With Title II back on the table again, we are seeing the repetition of yet another talking point that sounds plausible but turns out to be totally wrong when you actually dig into the evidence. ISPs and their defenders are repeatedly claiming that the U.S. did better than other net neutrality countries (specifically, the EU27) when it came to handling the crush of Covid-19 induced traffic. Unsurprisingly, they credit the lack of regulation for this amazing response. Once again, this claim does not hold up to real scrutiny.
As with the investment nonsense, this is a highly complicated area and therefore subject to a lot of spin and heated arguments over what the data actually show and how to explain it. It is made even more difficult by the complete lack of any official statistics (or, as the recent BITAG report put it more politely: “Data sources vary from independent measurement systems to self-reported internal company sources.” (P 7 n.1) So I will just give a few headlines up top and dig into the details below.
Contrary to industry boosterism, everything was not awesome for networks during Covid. As one industry observer put it: “By ‘handling’ the volumes they mean that their networks are not crashing and shutting down. But I think there is a whole lot more to these headlines than what they are telling the public.” For reports from the actual time about U.S. problems, see here, here, and here.
The U.S. Performed Worse Than Some Countries With Net Neutrality Laws. Studies vary, but one important one looked at not simply the EU and U.S., but also the European Free Trade Association (EFTA) and Canada. EFTA member states have the same net neutrality mandates as the EU (sometimes referred to as the EU27, referring to the full member 27 as distinct from the EFTA). Canada has treated broadband as a telecom service for something like 2 decades now, and has similar net neutrality laws to the U.S. 2016 rules. As this study found the U.S. internet traffic as a whole suffered a 4.9% increase in congestion as compared to 7.25% for the entire EU27, but this was significantly higher than for EFTA (3.3%) or Canada (2.4%). Additionally, when surveyed a week later, EFTA and Canada had made significantly greater progress on reducing congestion than the U.S. Furthermore, the U.S. numbers were for the largest cities with the strongest networks. If you start taking out members of the EU27 who aren’t considered our economic peers, the numbers for Europe improve to be comparable with those of the U.S. So sure, there were some differences but they had nothing to do with net neutrality regulations.
There isn’t a lot of evidence to support the “U.S. did better than the EU” claim. While you can find some studies that support the thesis that the U.S. did “better” by some set of metrics, there are a lot of other studies that show that from a consumer perspective, E.U. and U.S. subscribers had similar experiences. See here, here, here, and here.
The Netflix Red Herring. The “EU asked YouTube and Netflix to downgrade traffic” factoid beloved of ISPs and their supporters is a red herring. Yes, EU regulators approached Netflix, YouTube when lockdowns began to reduce the quality of their video from high-def to standard. But this was a prophylactic precaution to head off a potential concern, not a response to congestion. Only in the U.S. — and only among industry and Libertarians — would the idea of government and all industry sectors coordinating and accepting “a joint responsibility to take steps to ensure the smooth functioning of the internet” be regarded as a sign of weakness or regulatory overreach rather than a simple statement of reasonable prudence and preparedness.
More below . . .
Why Is Measuring This So Damn Hard?
Measuring the technical performance of internet networks and the impact of events like the Covid lockdown is incredibly, insanely hard. First, we have absolutely no standards to describe what we are talking about. Are we talking about how an average user experiences their internet experience? OK, are we worried about download speed? Upload speed? Both? Access to specific popular sites (like Amazon) as opposed to applications like Zoom? What about latency? Aggregate performance v. specific time of day?
You get the idea. We have many potential metrics we can use to describe whether the internet was or wasn’t impacted by the surge in use from the Covid-19 traffic. Even if we agree on what we want to measure, say the performance of latency sensitive applications like video conferencing, we don’t have agreed on metrics for how to measure impact. Jitter? Freeze? Dropped calls?
Now let’s make it even more complicated. Where in the network are the issues and how do the issues interact with each other. For example, one of the things that congested early in some places were the traffic exchange points between specific countries, notably in Italy and in Spain. As commentors noted, this was in part because of the release on the first weekend of lockdown of new versions of the video games Call of Duty and Fortnite. Additionally, these international exchange points are not subject to any kind of net neutrality or Title II regulation, and never were. So congestion in this part of the network has nothing to do with the regulatory environment.
So the first problem, as I pointed out over a year ago when this started, is identifying where in the chain we find degredation and congestion from the increase in Covid traffic. The first step closest to the user is the Wi-Fi router. If that gets congested, the individual user is going to suffer a slowdown. Then there is the last mile/first mile, the broadband provider to which the end user subscribers to get access to the internet. Then you have various pieces of the “cloud,” the largely unseen companies that route traffic globally through local, national and international exchange points. Finally, you have the actual site on the internet you are trying to reach. If Amazon is getting a ton of traffic, then traffic to and from Amazon will move slowly (which it did in the first few weeks) until Amazon increases its own capacity.
Of these many segments, only one — the “last mile” provider, the broadband internet access service (BIAS) provider — is impacted by net neutrality/ Title II classification. And even within the last mile, the medium makes a huge difference. copper-based DSL works differently than cable DOCSIS which works differently from fixed wireless, fiber, etc. So even trying to measure the impact of Covid lockdown on “last mile” speed accurately is very difficult.
Absolutely Zero Consistent Measurements Even Within the Same Segments of the Network.
Our problem in getting accurate measurements don’t end there. As any serious expert acknowledges, we don’t have a a good source of standardized measurements that is (a) open to all investigators; (b) consistent across all networks; and (c) agreed to by all experts as definitive. So, for example, you can get a lot of good data from places like M-Labs, but that is not going to give you the same data as the ISPs’ own data which they use to manage their internal networks. Nor are ISPs consistent between each other over how they measure or what they weight. While of course there is a great deal of similarity, because ISPs are dealing with the basic problem of managing networks in the face of a massive spike in use, ISPs will take different approaches and therefore measurements will vary enough to make precise comparisons for research purposes difficult.
Of course, the ISPs control access to their own performance data. They decide how they want to share it, if they want to share it at all. For a variety of reasons, other segments of the network are more transparent. So, like the old joke of the drunk looking for his lost key under the street lamp because ‘the light’s better here,’ we tend to overweight what measurements we can get. Finally, by homogenizing to get national averages, we blur the problems that arise from differences in things like regional connectivity or differences in technology. If I say speed on a national basis slowed down by 4%, that doesn’t tell me about the millions of people who have low quality broadband whose speeds already hit rock bottom before lockdown, the regions where we didn’t see as much increase in demand because those states did not take lockdown seriously, v. the largest cities where slowdowns surged.
Like a broken record, I keep pushing for the FCC to establish these standards and collect data. This, of course, requires firm statutory authority, like Title II. But I digress. The key point I want to make here is that even before we get to what the data actually show, we need to acknowledge our terrible data situation and just how hard it is to use that data to draw policy conclusions — especially really broad policy conclusions like “who needs Title II because our broadband did awesome during lockdown! (Except for the 40 million or so who didn’t have sufficient broadband in the first place.”
So, Was Everything In Fact Awesome During Lockdown? Or at least better than countries with net neutrality rules?
Now we get to the actual thing industry and anti-Title II folks keep saying. Everything worked out awesomely during Covid lockdown, especially compared to countries that impose “heavy handed, investment killing, anti-innovation, puppy strangling, kitten drowning nasty bad bad net neutrality regulation.” Are either of these claims true?
Not really.
One of the thing corporate America has done very well over the last generation or two is to severely lower our expectations on things. If you have to call customer service for anything, you consider it a major victory if you get off the phone in less than an hour and actually get the problem solved — never mind that you still got overcharged or otherwise ripped off in the first place. “Excellent customer service” now means “at least they didn’t spit on me or punch me in the back of the head.” This is particularly true for the cable industry, which is now the broadband industry (as documented in this delightful South Park from a decade ago).
So when we ask “how did our broadband network do in the Covid Lockdown,” we need to take that into account. Basically, you could have spent the first 2 months with your kids’ virtual classes freezing every 5 minutes and getting only every 3rd word on you office Zoom conference and people would still say “it worked fine” because at least the damned thing worked some of the time and every now and then I got through an entire episode of Bridgerton without crashing during one of the steamy parts. Basically, unless your broadband service collapsed for a solid week, the general response on “how well did your broadband work during Covid?” is going to be “about the same as always.”
Even if we wanted to rely on the overall customer impression of how service worked, we don’t have a good way to aggregate it. And, as I said above, even if consumers experienced significant difficulties, we would have no idea where whether that was the result of a last mile problem or something else.
All of which is to say, the constant “yay, the internet didn’t collapse so it must have been awesome!” that we keep hearing from Republicans in particular is not the measurement we should be using. If everything is a success unless there is a prolonged catastrophe, then eventually we are going to have catastrophes. And, in any event, the point here is “Yay, the Internet didn’t collapse” does not prove diddly about the impact of net neutrality regulation, which is what the industry wants to claim.
So when we look at actual data, we discover two things. As far as we can make decent cliche-to-cliche comparisons, the U.S. and the EU did about equally well in terms of internet performance degredation and slow downs. See here, here, here, here, and here. Additionally, some broadband providers have been particularly challenged on upload speed rather than download speed, but because we measure internet performance by download speed they get to report that everything is awesome.
This is not to say that you can’t find studies showing that the U.S., aggregated and on average, had a statistically smaller degredation of traffic due to congestion than the EU27, aggregated and on average. You can. But the difference is negligible from a consumer perspective (about 2%). More importantly, in the same study, other countries that classify broadband as telecommunications and impose net neutrality rules did about 2% better than the U.S. Perhaps net neutrality just improves efficiency in Canada and Iceland but not in Greece? More likely, as serious reports have suggested, it has a lot more to do with things like how much content is cached near the local network than with the regulatory regime.
But the EU Asked Netflix and YouTube to Downgrade Streaming!
All of which brings us to the fact that kicked off this talking point in the first place. In the opening days of the lockdown, EU Internet Market Commissioner Thierry Breton went to the largest providers of video downloads — Netflix and video game providers — and asked them to limit the size of downloads for 30 days to prevent any issues. If the United States had a responsible government, we would have engaged in a similar exercise in prudence. Not because we were experiencing problems, but because in a crisis you want to ask people to conserve resources and behave responsibily.
As explained more fully in this article here, these actions were merely part of an overall coordinated effort by the EU regulators (notably the Body of European Regulators of Electronic Communications, aka BEREC) to keep tabs on the emerging situation and prevent any crisis from emerging. Again, it is something of a testament to how out of whack life in Washington is in the Post-Trump era that functional oversight to ensure that critical infrastructure avoids a crisis is regarded as some sort of confession of weakness when real manly macho governments rely on hopes and prayers. But that is what this basically boils down to, “hee hee, look at those silly Europeans acting like grown ups and taking precautions! Don’t they know the right thing to do is have the Chair of the FCC frantically call up ISPs and beg them to sign a voluntary pledge to behave?”
Conclusion
There are actually a bunch of important lessons to learn from the EU Covid-19 response wrt net neutrality. Notably, the BERC issued an interpretive document in the first week of the lockdown explaining that net neutrality rules did not prevent ISPs from managing their networks in the face of the crisis, but they were not allowed to use the situation as an excuse for paid prioritization. Hopefully I will get a chance to blog about the lessons of Covid for the future of internet policy another time.
But to sum up, the argument that the United States did better than Europe because of the repeal of net neutrality and Title II is not supported by any set of reasonable facts — although you can cherrypick a few things for hand-waiving purposes. It is difficult to show whether or not last-mile broadband networks in the U.S. actually performed better than in the EU at all. And the evidence that does support that conclusion shows that the U.S. performed worse than other developed countries with net neutrality.
Not that I expect the above 2500+ words to matter to those who wish to believe it is true. But for those who actually care about policy, and for the upcoming court cases and FCC proceedings, I expect having all the links and facts in one place to be useful.
Stay tuned . . .
Typo! “hand-waiving”