Crawford Misuses the Survey Data

This is the second installment of a two part series of posts on Susan Crawford’s 8300 word post, Responding to Distorted Op-Eds Published by the New York Times. In the first installment, we looked at primarily at the technical claims that Crawford got wrong and her general framing of the issues. In this one, we’ll examine factual claims Crawford makes on the basis of her foray into actual research. As noted previously, this blog post is Crawford’s first attempt to buttress her analysis of U. S. broadband policy on research data; her book, Captive Audience, relied on blog posts and news articles for its references, not exactly reliable sources.

Common Research Errors

Just as Crawford makes technical errors about the differences between fiber, cable, and DSL, she makes research errors with respect to broadband deployment (AKA “buildout”) and adoption (AKA “uptake”.) For example:

Crawford says: “The FCC recently reported that only 0.3% of DSL connections in America provide speeds of 25 Mbps or higher.” There is no source for this claim, but the most recent data from NTIA & FCC is cited in the White House’s June, 2013 report Four Years of Broadband Growth. They say (Figure 3) that 7.21% of American homes have access to a DSL service that runs at 25 Mbps or more.  Crawford is off by 2400 percent. Good show.

Crawford says nobody wants fast DSL: “In general, takeup of U-Verse has been slow; it has taken U-Verse six years to reach the penetration that FiOS reached right off the bat.” Yet telecompetitor says U-verse is currently adding subscribers faster than cable. Is the history more important than the present trend?

Crawford claims: “Cisco estimates that wireless data accounts for only 2% or so of all data traffic, and that wireless will still represent just 9% of traffic in four years.” But this is what Cisco actually says (page 2): “Traffic from wireless and mobile devices will exceed traffic from wired devices by 2016. By 2017, wired devices will account for 45 percent of IP traffic, while Wi-Fi and mobile devices will account for 55 percent of IP traffic.” Crawford is off by 500 percent.

Crawford claims an FCC report says less that 11 percent of U. S. homes are served by mobile at speeds above 10 Mbps, but there is no such claim in the report. In fact, the FCC says (Table 11) none are served at that speed, but the report is based on pre-LTE data. With four networks building LTE – capable of speeds well above 10 Mbps – the current figure would be in excess of 60 percent and on its way to more than 90 percent.

Crawford claims a European Commission report says: “the U.S. is behind South Korea, the UAE, Hong Kong, Japan, Taiwan, Latvia, Lithuania, Norway, Sweden, Slovakia, Bulgaria, Portugal, Iceland, Denmark, Estonia, Finland, and Norway,” yet the report makes no such claim and doesn’t even mention most of the non-European countries on Crawford’s list. It actually says this (page 49): “Fixed broadband penetration in the EU was slightly higher than in Japan and just below that of the US as of July 2012.” Europe’s fiber buildout is well behind that of the US. According to OECD data (page 23), FTTH buildout in the US was ranked sixth in 2010, behind only Japan, Korea, Slovakia, Finland, and Denmark. It’s likely higher now because the US has installed more fiber miles per year that the rest of the world except China since then.

Changing the Subject

Crawford tries to refute a claim I made about broadband speed by changing the subject to broadband subscription rates. I said:

Much of the disparity between perception and reality has to do with timing. Before the recession, American Internet service was on a very different path, not keeping pace with large sections of Western Europe and East Asia.

But that began to change as the economy turned around. Private investment and advances in technology, brought about by a competition policy that encouraged cable and phone companies to improve their networks, have propelled America’s networks forward.

It should be clear that I’m talking about “America’s networks,” but Crawford’s counter claim that “American growth in high-speed Internet access has plateaued” is based on largely on subscription rates: “But growth in adoption slowed around 2007 and hasn’t picked back up. In 2008, over 100 million Americans, about one-third of the population, did not have an Internet connection at home.” Subscription growth is an interesting subject, but it’s not germane to my point.

Crawford mixes the adoption data – which are actually lower in the EU than in the US – with buildout claims such as this one: “According to the EU, 14.8% of European fixed connections provide speeds of at least 30 Mbps (up from 9% a year ago), mainly thanks to the expansion of cable DOCSIS 3.0 lines. Penetration has gone from 2.5% to 4.2%.” Yet the FCC’s data says that 82 percent of American homes are passed by a cable service offering speeds in excess of 50 Mbps and 17.8% are passed by an FTTH service offering even higher speeds. There is easily twice as much cable in the US as in the EU, as even Crawford admits: “39.4% of the households in Europe are reached by a cable DOCSIS 3.0 connection that is certainly capable of 100 Mbps downloads.“

In fact, the US has more fiber and more cable per household than Europe does. Crawford confuses buildout with uptake again in claiming an FCC report says: “half a percent of fixed connections in the US meet the 100 Mbps mark.” 100 Mbps connections are offered to roughly half of Americans. (See analysis in The Whole Picture, page 21.) She knows the difference when it comes to criticizing a claim by McAdam that seemed to confuse buildout with takeup as well (“It is not true that only 2% of households in the EU have access to 100Mbps service.”)

The Fiber Story

Crawford disputes my claim that the story she tells about American networks – that they’re second rate and falling behind Europe and Asia is false – by further misusing the EC data.  She says “The fiber story in the U.S. is meager” and “Other countries are moving ahead” and tries to support these charges with EC data from the study previously cited. In particular, she cites EC data to argue for U. S. mediocrity: “In Japan, 42% of all fixed high-speed Internet access lines are already FTTH. In South Korea, 58% of all lines are FTTH.” The context of the EC’s FTTH data is important, so let’s look at it in context. The EC data is on page 52 of their study, and it reads as follows:

In terms of NGA technologies, cable is by far the market leader having 57.4% of high-speed lines. Over 90% of European cable networks have been upgraded to Docsis 3.0 capable of download speeds well above 30 Mbps. Furthermore, cable operators have also migrated the majority of their customer base to NGA. VDSL has been progressing much more slowly, as only  a  fraction  of  xDSL  lines  have  been  upgraded  to  VDSL.  FTTH  and  FTTB  have  a combined share of 25.8% within NGA lines, and only 5.1% of all fixed broadband lines as opposed  to  42%  in  Japan,  58%  in  South  Korea  and  9%  in  the  US⁶⁴. NGA lines in total account for 20.3% of EU fixed broadband lines as opposed to 12.2% a year ago.

According to the EC analysis Crawford cites, the US has nearly twice as many active fiber connections per household as the EU (and twice as many ultra-high-speed cable connections, as noted previously.)  It’s worth noting the terminology the EC study uses: “NGA” means “Next Generation Access”, or any broadband network capable of 30 Mbps or more. EC doesn’t discriminate on the basis of wire type: any technology that can exceed 30 Mbps is fine by this analysis as it opens up the most demanding applications of today and the near future. DOCSIS, VDSL, LTE, and FTTx can all qualify as NGA technologies from the EC perspective.

Reversing the Data

But the remarkable use of data here is Crawford’s attempt to take data from the EC that shows the U. S. far ahead of Europe and re-purpose it to show the U. S. falling behind. Of course, there are a few European states with higher FTTH buildout than the U. S. currently has in terms of percentage of households. But look at who they are: Slovakia, Finland, and Denmark, small nations where most people live in highly dense cities. Finland has a lot of land area, of course, but most of it is populated only by reindeer and is not covered with fiber. Despite low mean population density, Finland has very high median population density (in the cities.)

The data cited do not support Crawford’s claim that the U. S. is falling behind Europe, but what about Asia? As we know, Korea and Japan have very extensive fiber builds, Korea’s concentrated on Seoul and Japan’s more widespread. They also have DSL, cable, and LTE. The fiber builds took place in these countries a decade ago. The current trend in these countries is something Crawford doesn’t want to consider: people, especially young people, are cutting the cord on FTTH and going all-in for LTE. Carriers are slashing prices for fiber to try and get them to stay:

In the wake of slowing subscriber growth NTT East and NTT West have cut prices for their fibre-to-the-home (FTTH) service by 34%, from JPY5,460 (USD67) to JPY3,600. The move is widely seen as a measure to stem the flow of customers leaving fixed broadband services in favour of mobile broadband platforms such as Long Term Evolution (LTE). According to TeleGeography’s GlobalComms Database, NTT’s net addition of FTTH subscribers fell from 2.046 million during the twelve months ending June 2010 to 1.756 million in the twelve months ending June 2011, and fell further to 1.277 million over the twelve months to June 2012. Australian technology news site Delimiter cites sources at NTT East and NTT West as being convinced that the main reason for the slowing FTTH take-up is due to many younger users preferring not to pay for a household-based FTTH service when they are already paying for their own smartphone LTE data plan. Unlike its smaller rival KDDI, NTT is prohibited from offering FTTH and LTE from its subsidiary NTT DoCoMo in a single bundled offering.

Oops. There’s even more irony, as the defection from FTTH to LTE is coming at a time when all-you-can-eat wireless data plans have been replaced with metered ones:

The irony of slowing FTTH subscriber growth in Japan and the rise of LTE is that it comes at a time when DoCoMo has actually significantly rationalised its mobile broadband pricing, sweeping away the old all-you-can-eat era with a cheapest price of ¥4,935 (US$60) per month for just 3GB of data – albeit at potential super-fast LTE download speeds of 75Mbps.

The rest of Asia is pretty much devoid of FTTH. China has very little FTTH in relative terms and most of Asia doesn’t even dream of it: India, Indonesia, Pakistan, Viet Nam, or anywhere else other than Korea and Japan have no FTTH at all. China has an aggressive goal: it wants to connect ten percent of its 400 million households to fiber by 2015, so they’re well behind the U. S. and simply making the rational move for a country with no cable.

No Laggard in the Global Context

So the U. S. is not a laggard in the global context, even if a few small countries without a significant cable footprint have decided, rationally, to skip a generation and go straight from ADSL to fiber. That’s certainly a sensible choice for them. In the overall scheme of things, fiber is most valuable as a backhaul or aggregation technology that only needs to come to the home when there are no bottlenecks anywhere else and the copper wire in the neighborhood is past its sell-by date. The U. S. fiber deployment plan follows that template: we install it where and when we need it, not sooner and not later. Cable uses fiber for backhaul, as does DSL and cellular. The U. S. installs 19 million miles of the stuff a year, more than all of Europe and more than all of Asia if we exclude China. That’s not what “falling behind” looks like.

Crawford’s blog post is most remarkable where it addresses the issue of observed speeds. In Captive Audience, she cited an Akamai State of the Internet report from Q4, 2009 to the effect that the U. S. was in 22^nd place overall and falling in average (TCP) connection speed. My op-ed pointed out the most recent Akamai SOTI report had the U. S. in 8^th and rising on this very measure. I’ve asked Crawford repeatedly why she reached all the way back to 2009 to denounce U. S. broadband in a book published in 2013, and have not gotten an answer. The blog post doesn’t address this question either, it simply questions the validity of the Akamai data and goes to other sources that don’t allow us to track standings over time and concluding that “Tussling over contestable rankings is not a good use of our time” before going through surveys of offered speeds and promised speeds to muddy the waters.

This is simply dishonest: Crawford was perfectly happy with Akamai’s data in January and there’s no doubt that actual, measured speeds are more important than offers that don’t live up to their claims. This is epidemic in Europe, where the EC has found that actual speeds lag offered ones by an enormous margin, as we reported on this blog in June. Overall, the speed of Europe’s broadband connections is only 74% of promised speeds, while in the U. S., ISPs deliver 96% of advertised speeds. As the U. S. has twice as many cable and fiber connections per household as Europe, our overall average connection speed is higher than Europe’s, where over 70% of broadband connections use DSL and there is virtually no LTE.

Who Chooses The Speed?

It’s particularly noteworthy that Crawford blames the supposed speed shortfall in the U. S. on carriers and is silent about the discrepancy between advertised and actual speeds in Europe:

The FCC sets a benchmark at 25Mbps, and tells us that only 9-14% of fixed connections meet it as of June 2012.  This is an interesting fact: DOCSIS 3.0 and fiber lines are certainly capable of providing 25 Mbps speeds, but the companies that operate these networks generally choose not to implement this capability.

In fact, all of the major DOCSIS carriers in the US offer speeds faster than 25 Mbps: Comcast goes up to 100 Mbps in most places and even higher in some, and Time Warner Cable offers 50 Mbps in most of its footprint. Crawford is making the classic error here, confusing consumer choice for lower-cost, lower speed plans with the capability of the network itself. Crawford claims DOCSIS 3 speeds are “not routinely available to Americans, and if they are they're very expensive” when 82% of Americans can purchase speeds of 50 Mbps and higher for less than the price of the most popular cable TV plan.

Crawford concludes with her typical claim that Verizon and AT&T have an effective duopoly on LTE (why does she care since it’s an inferior technology with no real value?) This is an increasingly out-of-touch observation in light of the dynamics of the mobile industry. Just last week, analyst reports say something very different is happening:

UBS telecom analyst John Hodulik has identified several trends that show T-Mobile and Sprint are gaining the upper hand on AT&T and Verizon. It’s still early, but we could be witnessing a competitive renaissance.

Oops again.

Conclusion

So what we have here is a host of analytical errors on common survey data that attempt to force conclusions that the data itself do not support. On the whole, the U. S. is ahead of Europe and Asia in broadband deployment, speed, adoption and price, with the exception of a few small countries that have either deliberately sought bragging rights (Japan and Korea) or who have skipped cable and gone straight from DSL to fiber (the former Soviet bloc.)

The policies that the U. S. is pursuing for broadband are rational with respect to our circumstances and the actual needs of broadband users. Our Internet economy is not being held back by our broadband networks and will not be held back for years to come under the gloomiest forecast.

Network quality is less important in America than the challenge of getting more people connected to the Internet by either wired or mobile broadband. This is the 21^st century, after all.