Broadband Cherry-Picking

According to Akamai’s State of the Internet reports, America’s average Internet connection speed ranking improved from 22nd place in 2009 to 13th place at the end of 2011, and other sources show that a majority of the world’s LTE users reside in the United States. These trends reflect positively on U. S. Internet policy, but our friends at New America Foundation are not impressed.

Since 2008, the NAF’s Open Technology Institute has published papers arguing that American broadband service is inferior to services in the rest of the world and getting worse. They blame this (counter-factual) state of affairs on our reliance on facilities-based competition (cable vs. DSL and FTTH vs. wireless) instead of mandated wholesale access to our various networks at prices set by regulatory fiat. They also want more facilities-based competition as long as it’s from government-owned or funded networks, using TV White Spaces, FTTH, and various other technologies.

Their position stems in part from a particularly quirky reading of a provision of the 1996 Telecom Act that was meant to create a market for competitive telephone services over the lines owned by the phone companies that were created by the break-up of Ma Bell. They also have tremendous confidence in the ability of small towns to operate their own broadband networks.

The latest installment of NAF’s ongoing broadband critique, ”The Cost of Connectivity”, adds some new wrinkles to the story that are particularly misleading. It seeks to compare the costs of triple-play bundles of TV, phone, and Internet service around the world which go considerably beyond simple connectivity. It also purports to compare worldwide broadband services at the $35/month level, an especially intriguing price point. Finally, it tries to identify the fastest Internet services available in selected cities worldwide.

The Cost of Content

On the first issue, the most important basic fact is that the pricing of triple play packages is dominated by the costs that service providers have to pay for content retransmission rights. Even the programming that is broadcast free-to-air must be licensed by the local service provider when it’s transmitted by wire or satellite, and providers pay additional fees for ESPN, Disney, MTV, AMC, HBO, and the like. So triple play price comparisons between the United States and other countries do not measure the “cost of connectivity” as much as they measure the “cost of content.”

While it’s an interesting exercise to compare content prices worldwide, any exercise of that kind isn’t meaningful unless we include the costs of content creation, which NAF doesn’t touch. If you live in Los Angeles, your cable bill pays most of Los Angeles Dodgers star Matt Kemp’s $160 million contract and Angels first baseman Albert Pujols’ $240 million deal. These teams have this kind of money to spend because they have cable contracts worth $3 - 4 billion over 20 years, and that’s just for baseball. NFL football is even more costly, and then there’s basketball, hockey, the Olympics and the rest.

America’s love for sports is one reason that cable bills are as high as they are, and our brilliantly produced TV shows and movies are others. It’s not surprising that the nation whose TV programming is the most costly and popular in the world (with the possible exception of Bollywood) should have the highest prices for content, is it?

Cable TV is also not as mature in other countries as it is in the United States. France doesn’t have a nation-wide cable network, so it relies on over-the-air, satellite and IPTV to deliver video to consumer households. Some companies, notably Free.fr –  the darling of the “U.S. broadband is too expensive” crowd  – offer very aggressive pricing for triple- and quad- play bundles in order to acquire market share, but the quality of their service offerings is so weak that customer retention is low even at rock-bottom prices. France loves its soccer, but the TV rights for its entire Ligue 1 (professional league) are only worth $150 million/year and there’s not much of interest in the Free.fr channel guide that’s not available free-to-air. So raw price comparisons for TV/broadband bundles in such different countries are little more than public policy trivia without a deep analysis of cultural context that NAF fails to provide.

Telephone Taxes

Another component of the triple play bundle is telephone service, and I’m more willing to agree with NAF that Americans pay too much for plain old telephone service. But the low-cost overseas options are Internet-based VoIP rather than classic circuit switched telephone service, so they’re naturally cheaper to provide.  Moreover, U.S. phone providers have to pay an array of fees and taxes (which inflate the NAF price comparisons) like the Spanish-American War Tax and a Universal Service Fund contribution from people in the cities to cover service in the countryside at the same prices that city dwellers pay.

The prices we pay for telephone service in the United States reflect our historic commitment to uniform pricing of communications services across the nation despite the varying costs of providing service according to population density and distance. So even a cursory examination finds that most broadband services provided by cable companies or traditional phone companies are priced the same in hard to serve areas of low population density and cheap to serve high-rise apartments in the big cities. Broadband service doesn’t have to be priced this way, of course, but our carriers have always done it that way with a few exceptions in areas where they’re especially eager to attract new customers or to retain old ones.

In contrast, many of the low priced foreign and domestic competitors NAF features serve almost exclusively high-density urban areas. It’s much less expensive to serve urban customers, so firms that specialize on the best parts of the urban market can naturally price their services lower than large American ISPs that must serve urban, suburban and rural customers can. The effect of this study’s focus on city-to-city comparisons is to distort the playing field by comparing the prices charged in low-cost and high-cost scenarios as if they were the same.

Of course, NAF and their allies respond by saying that U.S. ISPs also serve high density, low cost urban areas and could charge lower prices there.  But if they did, they would have to charge higher prices in poor urban areas with low subscribership and in low density suburban and rural communities, and NAF would then be complaining bitterly about unfair discrimination against the urban and rural poor. In most cases, telephone companies and cable franchise holders are required to provide universal service, so cherry-picking the most affluent neighborhoods and ignoring the others simply isn’t an option for them.

Cherry Picking

NAF cherry picks data within the United States as well.  The easiest way to make a distorted comparison about broadband service would be to contrast one with uniform pricing across a large market and one that serves a small and selected market consisting of customers who are especially easy to serve, in part because of high rates of computer ownership and a strong interest in broadband. NAF commits exactly this offense, comparing Webpass  of San Francisco and Sonic.net of Santa Rosa, CA with AT&T and Comcast. Webpass describes itself as a "Building Specific ISP [offering] service to over 200 buildings in the Bay Area."  To qualify for service, the building has to have been built in the last 15 years, have more than 20 units, and be located in San Francisco, Emeryville, Oakland, or Berkeley.

Sonic has a similar business model, providing gigabit fiber service to just 700 homes in Sebastopol, CA in addition to a DSL product provided over AT&T’s lines (and offered at AT&T’s prices) to a larger swath of the Bay Area. I was once a Sonic customer, and my experience was that the product failed to live up to expectations; it was supposed to include web hosting, but Sonic’s engineers couldn’t quite figure out how to implement NFS byte range locking in a way that agreed with Movable Type’s requirements. I had a blog that could not be viewed and updated at the same time, a major inconvenience. But my travails with Sonic were long ago, and I’m sure they have many happy customers today.

The following map shows the deployment of the Sonic fiber network in Sebastopol. (The active portion is in green.)

Now compare this with the map of Comcast’s Sebastopol deployment according to data from the FCC’s National Broadband Map.

First, let’s look at the plain map of Sebastopol.

Now let’s overlay the Comcast DOCSIS 3.0 service area in blue:

Comcast serves all the homes Sonic does, but it also serves many more, including ones in much less densely populated parts of Sonoma County.  Can you imagine the reaction if Comcast announced it was only going to serve 700 homes in the city core of Sebastopol and stop serving all the rest?

What actually happens to people who live in hard-to-serve areas when companies like Webpass and Sonic pick all the low-hanging fruit from the broadband tree? They can only suffer in the long run when the high margin, low cost areas are cherry picked by the premium network that doesn’t bother to extend infrastructure into the countryside.  Because broadband has high fixed costs (the costs of stringing cable, setting up home connections, installing neighborhood boxes, etc.), when Sonic or other cherry pickers set up in places like downtown Sebastopol they cut revenues for the incumbents much more than they cut their costs.  As a result, the costs to serve other broadband users go up.  In the overall national scheme of things, the cherry-pickers are nothing to celebrate, so the city comparison is inherently the wrong way to judge national policies.

Arbitrary Decisions

The second question the report asks is what level of broadband service people can buy in various cities for $35. In the United States, broadband prices tend to start at $20/month for a low-end, introductory service and then rise according to capacity. For example, Time Warner Cable offers the following one year deals in Brooklyn:

3 Mbps               $20 10 Mbps             $30 20 Mbps             $40 30 Mbps             $50 50 Mbps             $80

Picking $35 as a cut-off point excludes the best value services from Time Warner.

NAF seems to simply play fast and loose with the facts with respect to standard pricing vs. introductory pricing. American broadband providers customarily offer a low rate for the first year and a negotiated rate in following years. The NAF report mysteriously omits the Comcast service available in many American cities (including San Francisco) that provides 30 Mbps for only $30 per month, even though its one year term falls within their stated guidelines, and it includes shorter term introductory rates in Germany even though they fall outside the guidelines.

Hidden Costs

NAF’s method produces some distortion because it only measures the bill sent to the consumer by the broadband company and ignores the effects of tax incentives and subsidies. In many countries, tax incentives are used to encourage investments in broadband infrastructure. This is the case in Korea, where NAF reports consumers paying a modest $23/month for 100 Mbps broadband service. It’s probably reasonable to double that rate to take tax incentives into account, and to double it again based on the cost of providing service. As ITIF explained in “Explaining International Broadband Leadership“, Korea has extremely high urbanicity, which means that most people live in high-rise apartment buildings in the large cities. This is the ideal scenario for broadband deployment.

In Paris, Free.fr is able to pull fiber through the sewer system for free, and the people of Chattanooga who enjoy a gigabit fiber network benefit from a federal grant for the smart grid system that runs over the fiber network as well.

Another hidden cost is levied by nations in which the government builds network infrastructure. This is the case in Sweden, Singapore, Australia, and the Netherlands. Taxpayers fund infrastructure build-outs through tax payments that lower the cost of providing services and hence allow providers to charge lower prices. The honest way to consider pricing in such countries would include the tax-funded expenses of the infrastructure company, but NAF doesn’t do that.

Falling Upward

The NAF report goes to some lengths to criticize the average connection speed in the U. S., claiming that we have “fallen to 13th internationally in average measured connection speed.” Two points need to be made about this.

First, average Internet connection speeds are falling worldwide according to Akamai:

The global average connection speed saw an unusual, and fairly significant, decline in the fourth quarter of 2011, dropping to 2.3 Mbps, as shown in Figure 8. It was reflected in declines in eight of the top 10 countries, as well as the United States. Globally, 93 countries/regions that qualified for inclusion saw average connection speeds decline, ranging from a loss of just 0.3% in Kyrgyzstan to a 31% drop in Kuwait.

Average connection speeds in declining because more and more users are connecting to the Internet from smartphones and tablets that aren’t wired to the Internet directly. Since the average connection speed of wireless devices is slower than that of wired devices, the increased number of wireless devices depresses the overall average, even when wired speeds are improving. Outside the U. S., the vast majority of these connections are made over 3G wireless technologies, as the majority of the world’s LTE adoption has taken place in the U. S.

Despite America’s leadership in LTE, or perhaps because of it, our average connection speed ranking is actually improving, despite NAF’s claim to the contrary. The evidence that our standing is improving comes by comparing our current speed ranking, 13^th, to our ranking in 2009, reported by one source as “between 15^th and 21^st in the world in terms of broadband access, adoption, speeds and prices.” The source is a report titled “Building a 21st Century Broadband Superhighway” by NAF’s Benjamin Lennett, the author of ”The Cost of Connectivity”.

NAF’s 2009 ranking of the United States was based on OECD rankings using a very different methodology than Akamai. If we compare Akamai’s ranking of the United States  in the fourth quarter of 2009 to our ranking in the fourth quarter of 2011, we see a rather dramatic rise from 22^nd in 2009 to 13^th in 2011. Apparently, “falling” means something different to the NAF researchers than it does to the rest of us.

The connection speeds in rural America are certainly low compared to those in any urban setting, because traditional DSL becomes slower as the distance of the user from the telephone switching office increases. (See “Explaining International Broadband Leadership” for a fuller discussion.) This effect becomes minimized as DOCSIS 3.0 cable and LTE become technologies of choice in rural America, options that are not available to most of the world.

NAF touts speeds and prices in places such as Hong Kong that provide extremely fast connections within Hong Kong’s territory but not to the rest of the world. Hong Kong’s broadband ISPs claim to provide 1 Gbps and 500 Mbps downstream services (and much slower upstream service) but these speeds are for local transfers only. Tests indicate that Hong Kong actually connects to the rest of the world at very low rates: “These tests further showed international speeds ranging between 0.5Mbps and 7.6Mbps – not as impressive as many people would have imagined.”

So is it really correct to claim, as NAF does, that “Broadband connectivity in San Francisco lags behind global cities like Seoul and Hong Kong”? San Francisco lags behind Seoul in many cases, but it’s nowhere near lagging behind Hong Kong, where FTTH connections slow to old ADSL rates at the border. The LUS Fiber service in Lafayette, Louisiana has the same limitation:  Users connect within the LUS service area at 100 Mbps, but to the Internet as a whole at 10, 30, or 50 Mbps.

It’s dishonest to try to pass off a high local rate as a high Internet rate, and it’s a fair guess that FTTH users in Riga, Bucharest, Prague, and Chattanooga don’t connect to the global Internet at rates anywhere near their “local area network” speeds. By NAF’s method of rating “Internet speed” the gigabit Ethernet connection from which I'm writing this post provides me with a gigabit Internet connection, which it doesn’t, of course.

What Does it All Mean?

Following the NAF reports on American broadband from year to year, the one constant is a steady drumbeat of criticism for the “facilities-based competition” model of broadband deployment we have in the United States. NAF would abandon this system for a wholesale/retail ISP model similar to the one we used when we accessed the Internet from dial-up modems and had dozens of ISPs to choose from. In order to make a case for switching to a system of intra-modal competition, it’s first necessary to make the case that the American system of inter-modal competition is deficient, so that’s the task NAF sets for itself.

However, to claim that Americans pay high prices for inferior service, it’s necessary to ignore much of the data on prices, service quality, content, differences in geography and density, government incentives, and connection speed trends.  Taking the straightforward view, wired broadband service in the U. S. has improved relative to the rest of the world since 2009: While we once ranked 22^nd, we now rank 13^th, a major movement in the right direction. This has come about primarily because most Americans connect to the Internet over cable and fiber rather than traditional DSL, but DSL speeds are improving in the U. S. as well, because AT&T and Century Link have replaced a great deal of legacy ADSL with much faster VDSL+ (up to 50 Mbps unshared) in many areas. Meanwhile, the nations that have adopted the wholesale/retail model are stuck in an ADSL copper wire ghetto from which they can only escape by pouring tax money into fiber projects.

While our wireline connection speeds have improved, so have our mobile connections, as most the LTE users in the world reside in the U. S. The United States is clearly bucking the predictions that NAF has made about our infrastructure on a regular basis since 2008.

Will NAF change its tune in the face of the evidence? It’s unlikely that future NAF broadband reports will improve on the shoddy methodology in the current and previous ones, because they’re accomplishing an efficient propaganda purpose. A considerable group of bloggers and journalists take these reports at face value and trumpet their findings as gospel.

One particularly striking example of this credulous acceptance took place at Forbes magazine, where managing editor Bruce Upbin wrote a blog post declaring “Complacent Telcos Deliver Americans Third Rate Broadband Service At High Prices.” In comments following the post, Upbin declared “I would kill for such inferior service [as consumers in Hong Kong can buy]. I pay $45/mo in Brooklyn for a measly 1.5 Mbps on a good day from Time Warner.”

This appears to be a case of Stockholm Syndrome, because Time Warner service in Upbin’s neighborhood provides 20 Mbps service for $40/month or 30 Mbps for $50, much better effective speeds to the Internet as a whole than the people of Hong Kong can buy for any price.

Upbin is in good company, however: “The Cost of Connectivity” was endorsed by The Verge, WebPro News, Ars Technica, USA Today and a host of others despite its evident flaws. As long as journalists and bloggers are willing to accept NAF’s outlandish claims, we can expect such reports to continue, unfortunately, and for Internet policy to focus on the wrong problems.