FCC En Banc Hearing
on Broadband and the Digital Future
July 21, 2008
The FCC held a special En Banc Hearing on Broadband and the Digital Future at CMU in Pittsburgh, on July 21, 2008.
Good evening. I'm delighted to be here to talk about issues of broadband and the digital divide, an area of research specialization. I just left a place where connectivity is not the primary concern -- the hospital. They've recently started free WiFi service for patients and their families, and I observed that even though it isn't used by everyone, when it is used, it's very useful. My wife and I just had twins, and so if I'm sleepier or even less coherent than usual, I have a good excuse.
work is global in its scope, but I've also to focused on the digital divide in
I won't go into the challenges of measuring broadband access, the reasons for any disparities, or even contentious discussions on what is broadband? I think we will all agree that broadband increasingly important, and it is used for a number of societal functions including education, seeking a job, online commerce and banking, voter information, etc. It’s also a great equalizer, being used by those who are geographically near, or those with disabilities. Pew and others have documented how those with broadband have different online habits than those using dial-up or shared access at a library, school, etc.
Intuitively, we can get a sense that as networks grow in size and value, those who are not in the network become further and further behind. However, conventional wisdom and traditional analysis indicates we are overall better off societally. But what about the individuals who are excluded? In some work just recently presented, done with Ernest Wilson of USC, I attempted to formalize the study of network exclusion. I will briefly summarize some high-level findings, I believe they are quite disturbing.
SLIDE 2 – Network Effects Matter
Metcalfe's Law is one of the most famous network laws, which points out that the value of a network doesn't just depend on you, it depends on how many others are part of the network. If you have a phone, you need other people with a phone as well. For such a system, Metcalfe's Law can be approximated as saying that the value of a network is proportional to the square of the number of users.
SLIDE 3 – How do we Measure Disparity?
There are some corrections or modified Laws based on the type of network (such as linear, for broadcast), social networks, etc. However, Metcalfe's Law and all the other Network Laws not only show monotonically increasing value with size, they are all based on network inclusion. Any disparity, which I will argue actually becomes a cost, is simply calculated as the difference between those in the network and those outside, who are assigned a zero value.
Say we have a network with 19 members. Metcalfe's Law would imply a value proportional to 361, or, per person, 361/19 = 19. But surely the disparity for those who are excluded is different whether just one person is excluded, for an applicable population of 20, or whether 81 people are outside the network, out of a population of 100!
SLIDE 4 – An Exclusion-Based Framing
We've proposed an alternative exclusion-based framing that takes into account both the growing included network, which may be an inherently superior network such as broadband (or immunized, insured, etc.) as well as the number of people who are excluded.
The included (growing) network matters not only because it gives signals to dynamic processes, (including supply and demand) but complementary networks (like content, software, etc.) will shift towards such a network. Even if you have dial-up, and nothing else changes in terms of number of dial-up subscribers, you’ll find content will become increasingly larger, flashier, etc. When we examined a number of leading websites over time, their size has increased between five- and ten-fold in the last few years.
SLIDE 5 – Exclusion Really Matters as Networks Grow
In a traditional framing, the individual disparity between included and excluded increases, but is modest (e.g., linear in Metcalfe’s Law). When we apply our exclusion based framing, we find that the cost of exclusion begins to look exponential as fewer and fewer people remain excluded. To our surprise, we found this is true for any underlying network structure (network law)! You’ll see that as a new technology emerges, inclusion confers a competitive advantage to those few who have it. As it becomes widespread, not having it is what hurts.
SLIDE 6 – Exclusion has Societal Costs as well
Those who are excluded from the "included" network will often resort to alternative or parallel networks. These can be less efficient, more expensive, less environmentally green, etc. If I cannot, say, download a government form of the Web, I might have to write, call, go there in person etc. Going may involve long lines, taking a bus, driving etc. This has implications on time, energy, environment, etc.
One closing but major point based on this is that worrying about exclusion is more important than just for the individuals who are excluded. Because of a number of factors, including alternative and parallel networks that societally we demand be operational, exclusion imposes costs on everyone, included those who are so-called included. The classic example is health insurance, where the uninsured disproportionately end up in emergency rooms, sometimes having skipped less expensive preventative care. Not only are the costs often borne by taxpayers, such cases also delay services for other critical patients.
In telephony networks, our phone system still allows pulse dialing in parallel to touchtone, ostensibly for the very few people who might need it. It may be cheaper to just give them touchtone instruments (?). Similarly, if we consider older operating systems like Windows 98 on computers, they are no longer supported and are unlikely to be patched from a security perspective. It is precisely these machines, which are still online, that are responsible for a very large fraction of the Internet zombies, Trojans, attacks etc.
Will asymmetries and inequalities disappear in society? No, and I'm not advocating uniformity or homogenization of networks. However, broadband might be something so important to society, like roads and electricity, that we need to actively address the 4 A's of the digital divide: Awareness, Availability, Accessibility, and Affordability.
Comments Presented (including Q&A)
Slides used (for Testimony and Comments, including Q&A)
These slides (.pdf) are based on work on network exclusion, available in my publications. The Annenberg Seminar I gave has more slides, and our TPRC paper is the first cut at this analysis (updated since then).
give below my personal comments/reactions to the Hearing. I
couldn't stay for all the Public Comments, and provide only reactions
to the second Panel, on broadband. [Boy I wish I had a blog up
During the opening statements by the Commissioners, an analogy was made to roads. While the analogies to roads may be many, and ones I myself have made on a number of occasions, I think the differences are also quite important. Roads were always viewed as a public good, with rare cases of private ownership. Telecoms in the US were a regulated utility, mostly in the hands of private entities. Interestingly, if we consider rural electricity, a substantial fraction of the US gets its service from rural cooperative providers, who are mostly nonprofit.
Dr. Wallsten pointed out the folly of relying too much on international rankings and comparisons. I agree. Per capita versus household numbers may certainly shift the rankings a little. However, I strongly disagree that the trends are therefore meaningless. Regardless of whether we examine households were per capita, as long as the methodology is consistent, a trend showing declining US broadband rank is relevant and disturbing. His suggestion that we need better numbers is also very accurate -- the limitations of things like FCC form 477 (which are at a ZIP code level) or where state regulators have data on DSL but not cable are well known. However, if one can make a claim that things may not be so bad because we just don't have true data, one can also then make the claim that things could be much worse because we don't have true data! Waiting for better data and scientific consensus before taking action to me is reminiscent of those who wish to deny or at least defer action on global warming.
After my comments comparing healthcare and broadband, Dr. Wallsten observed that something like broadband represents an enormous opportunity cost when we have other challenges facing US society. While I certainly would put healthcare as being more important than broadband, I emphatically reject the premise that these are mutually exclusive efforts. (This is without even thinking about the Mother of all Opportunity Costs...)
As a society, we just need to make decisions on what exclusion we do or don't find acceptable. After that, we can figure out the best means of achieving the desired goals once we are willing to accept the problem and limitations of current policies. My own work and discussions with a number of leaders in the field indicates that it is not that expensive to provide broadband, even in rural areas.
Based on the discussions, where Dr. Wallsten pointed out that underserved are not rural per se but poorer communities. I think there may be some overlap, but if we find there are two types of underserved regions, rural and poorer non-rural areas, then it would be important to understand the differences and their reasons.
In some ways, rural areas are intellectually easier to understand. Because of lower densities, much more expensive backhaul, etc. it costs more than in other regions. It costs what the techno-economics state it will cost. At a fair and reasonable price, I do not believe ability or willingness to pay is a major barrier, at least not more so than the bulk of the United States.
Uplinking is a MAJOR cost and bottleneck. One carrier mentioned at a meeting "we don't price discriminate against rural users - all are the same." Yes, but their PoP is nowhere nearby. Local loop charges will be killing then.
In the rural US, fiber is cheaper (on poles) than people think. The challenge is the 3 utilities who have rights to or ownership of the poles (and share amongst themselves): power, cable, and phone. They claim there is no more space [x feet separation] for a new entrant, even though fiber will not interfere at all. A few years ago (I spent some time helping with broadband in Rural PA), as far as I can remember, the rental for space in rural PA was $5/pole/yr. In a mile, we're talking $100/yr. Assuming we can get 3 homes in a linear mile, which is feasible for the majority of rural areas, this comes to $33/year/home, or just $3/month for FIBER. Capex and installation are medium, one-time costs that are also not nearly as high for existing poles compared to new ducts underground.
Even when there is fiber, and there's much more fiber even in Africa (power companies) or India (incumbent, who goes into every sub-district, ~precinct, with fiber) than people realize, clarity on who can use it, at what price, etc. is what prevents it being used well. Fiber is such a special medium that we only need one for all the ostensible traffic out of many homes. I think relying on physical layer competition (and economists say we need at least 3) will be a red herring. We don't build 3 highways to a town in the name of competition. Congestion, maybe. The competition is Ford vs. GM vs. Toyota. DHL vs. Fedex vs UPS. Contractor/toll operator company A vs. B, etc. [A paper on such Open Access for Africa is here: FiberAfrica]
If you examine my commentary on broadband and healthcare both showing bimodal distributions of pretty good versus pretty bad, can we think of any major infrastructure where private players (the market) built out near ubiquitous presence solely through competitive forces without regulatory or other incentives? Markets are great, but I think we need to examine what they leave behind.
Coming to urban areas, where in theory DSL is available, and perhaps even cable broadband, is it simply a matter of cost? Many carriers in the DSL space offer introductory plans for about $15 or $20 per month. We can likely add a couple of dollars due to taxes and fees. I think to see widespread penetration total costs need to be under $10. And I also think, based calculations and discussions, at $10 per month it would still be slightly profitable for the carriers, especially at the margin. In India, parts of China, and a few other places, introductory broadband (admittedly with usage caps) is only about $5-6 per month. The problem is no carrier in the United States wants to offer eight or $10 DSL in the same market where "cheap" broadband retails for $30. Consider an elasticity based argument. For argument's sake, lacking any better information, let's assume an elasticity of 1. So if I cut down my DSL price by a factor of three, subscriptions would increase by a factor of three. My revenues would be approximately neutral. But my profit would fall dramatically, much more so than the equivalency in revenues. One can play around with the numbers a little, but I'm proposing this hypothesis as being one factor why we don’t have cheaper DSL for some areas where the service already exists.
During Mr. Quinn's Statement, where he talked about AT&T's capital expenditure woes, I think there may be a little bit of "true, but so?" at hand. Hasn't investment always been somewhat cyclic, so that there would be periods of massive investment followed by lower investment while supply and demand try to reach an equilibrium? He complained that backbone capacity was going to run out in three years, prompting large investment again. I don't know the details of how upgrades are planned, but optical fibers operating under DWDM are capable of additional wavelengths with medium instead of enormous investments; this excludes the fact that the optical fibers themselves are in place.
If there will be growth, growth in the backbone may be much easier to manage than growth in the last mile, forget for new last mile connections. Using the rule of thumb that a 4x growth in capacity costs only 2x or 2.5x (plus or minus), given the enormous growth in paid volume, the average cost may remain stable or even decline. Consider long-haul point-to-point connections, where the costs have continuously fallen over time. I don't think anyone can claim that more usage hurts consumers by raising average costs.
Regarding my own presentation, most of the press and blogs were very positive, though one blog did claim that I botched Metcalfe's Law (which the author states shows geometric growth). I'm happy to respond to any direct queries, but will mention that I explicitly stated that one could approximate Metcalfe's Law by showing the value being proportional to the square of the number of users, which is correct at an order of magnitude level.
The intellectual challenges in dealing with exclusion are enormous, and my ongoing work in this field is only one of multiple efforts dealing with network exclusion and the digital divide. How does one measure excluded when, by definition, they are excluded and not measured?
I'm sure most people are familiar with the phrase "Six degrees of separation" based on Stanley Milgram's work in the 1950s on getting letters from one end of the US to the other only through direct contacts. What is much less well known is that the average hop count that is now part of lay lexicon was only for the letters that made it to their destination. In his published study, only about a quarter of the letters made it, and in earlier trials, scholars have discovered that far fewer actually made it. Thus, my work on an exclusion-based framing to the digital divide presents a new framework.
In closing, kudos to the FCC for going around the country for hearings. I only wish there were a better manner to have more non-linear discussions, instead of hearing everything, in order, which necessarily includes speeches or statements, limiting time for discussion and broader participation.