The Trouble with Broadband Deployment Statistics
It seems hardly a week goes by without somebody sounding the alarm bell on the “crisis” in US broadband deployment. While we all share the common goal of bringing affordable broadband service to all Americans, it’s unfortunate that the most frequently cited source of broadband deployment – the semi-annual Organization for Economic Cooperation and Development (OECD) numbers – contains a variety of inaccuracies.
So, to help set the record straight, this week we thought we would take a look at the OECD broadband study and the real state of broadband today. In our first two installments, we’ll examine flaws in several units of measurement utilized in the OECD study. We’ll then move on to fact check the “miracle” of Japanese Broadband and finish by analyzing why consumers aren’t connecting to services that are already available.
It depends on what your definition of a subscription is.
The most significant flaw in OECD’s methodology is their measurement unit of subscriptions per 100 inhabitants. Average household size plays havoc with the “inhabitants” calculation, creating some serious unintended consequences. We’ll cover that in tomorrow’s post.
Equally problematic, however, is what they do and do not consider to be a subscription. In OECD’s definition of what constitutes a “broadband subscription,” there is no distinction drawn between business DSL or cable lines and residential DSL or cable lines, but there is a specific exclusion of direct fiber and T1 lines for businesses. As a result, some businesses are counted and others are not.
OECD’s data fails to capture the tens of millions of U.S. workers that access the Internet via these special access connections.
The OECD measure also fails to count the approximately 16 million college students in the U.S., most of whom have access to both wired and wireless High-Speed Internet (HSI) service. Also uncounted are the HSI users that access WiFi connections, and the growing number of mobile wireless and “Hot Spot” customers.
Undercounting these populations negatively impacts the US ranking, but counting them would be problematic as well. Because so many people have broadband access at home, at work, via their mobile device, at college, or through some other connection, the risk of double or triple counting becomes fairly great.
Some have suggested that a better metric would be to simply measure the number of residential households that are subscribed. The distinction is really very stark. For instance, in the US, roughly 57 million households subscribe to cable, DSL, fiber, satellite, or fixed wireless service. Using a measurement of how many “residences” have access would more accurately reflect the real state of residential broadband consumption, and would vault the U.S. ahead of 9 European countries which were ranked higher in terms of household penetration in the OECD rankings of December 2006. (Note: Household data is not available for Korea, Canada, and Australia, so it’s unclear where they would rank).
::
Send to a friend
February 6th, 2008 at 3:52 pm
This is a great idea to talk about Michael.
A quick run through of your post had the following section catch my eye.
“Equally problematic, however, is what they do and do not consider to be a subscription. In OECD’s definition of what constitutes a “broadband subscription,” there is no distinction drawn between business DSL or cable lines and residential DSL or cable lines, but there is a specific exclusion of direct fiber and T1 lines for businesses. As a result, some businesses are counted and others are not.
OECD’s data fails to capture the tens of millions of U.S. workers that access the Internet via these special access connections.”
I am not completely sure I understand what you are saying, could you ellaborate please.
Also, if the Internet connections are not being counted in the U.S., they are also not counted in other countries. I am not that accessing the Internet in special access connections is so much greater in other countries than in the U.S. It is still worth attempting to fix the missing data that could provide a better, clearer picture.
I think the ‘crisis’ rhetoric concerning broadband in the U.S. is only highlighted by the OECD data. Prices, download speeds, and so forth are strikingly different in the U.S. than in other OECD countries. ( I am sure you will talk about this later). In some aspects the U.S. is lagging behind and in other ways the U.S. is leading. However I prefer to think that the rhetoric is aimed at providing fast, effecient, affordable broadband connectivity at a person’s home, not a special access connection location.
Cheers.
February 6th, 2008 at 4:09 pm
Unfortunately, the study assumes a DSL connection is a DSL connection regardless of whether it’s a home or business. A T1, however, is unlikely to be a home connection.
As for the elaboration, US businesses often rely on T1 lines because that was the only option available for some time. By discounting T1 lines, but counting DSL lines, the US fares worse than countries where DSL is more prevalent for businesses. Their business connections are counted while ours are not.
So your assumption that they’re equally skewed in the US and other nations isn’t necessarily accurate. T1 lines would be discounted everywhere, but if one country was more likely to employ them, it would have a negative impact on that country, despite it having – and specifically because it has – a lot of them.
That’s one area of contention where some have made the suggestion that residential connections be the focus of such studies. If you focus only on residential connections, the US is actually well ahead of many countries the current methodology inflates.
February 7th, 2008 at 3:55 pm
Michael,
I believe a T1 line is a fiber optic line. Since all fiber lines of
256 kbit/s are counted I don’t see an issue there.
Let me sum up what I think you are saying just to be sure. Because a T1 line is subscribed to by one person only, each T1 connection is counted once. Even though in the U.S. one T1 line may connect many people to the Internet. Whereas in other countries, DSL was used, and since DSL is subscribed to one person and not used by many, the numbers are inflated for the DSL countries because there is a lesser ratio of users per connection subscriber? Where a T1 in the U.S. might allow 100 people access to the Internet, a DSL connection might allow fewer people access, thus more DSL connections are needed.
If I have understood you correctly, then we would need to adjust the OECD to include all persons that connect to the Internet per subscription. Logistically, I do not see how that is feasible. Have you any ideas?
Cheers,
Wyatt
February 7th, 2008 at 4:35 pm
A T1 line is not a fiber optic line. A T1 line is made of 24 8-bit channels (essentially a normal twisted-pair phone line). When you look at a T1 block in your phone closet, it actually looks like a normal punch block and is largely indistinguishable from your phones. A T1 line operates at 1.544 m-bits. Fiber optic lines are different. Fiber optic lines are much faster than a T1.
The heart of your question isn’t dependent on that, so we’ll move on. A number of years ago, I worked in an office in Phoenix that had about 20 people connected via one 768k DSL line. They would be counted. When I first moved to DC, I worked in an office with about 50 people connected via a T1 (again, at 1.544 m-bits). They were not counted.
The issue that creates for the OECD study is it a) reduces the number of connections counted, b) reduces the average speed of those connections, and c) discounts 50 people who are covered.
Now both of those were businesses. At the time, I had a DSL line at home as well (I now have cable
. If you count both businesses and home connections, I was double counted in Phoenix and not in DC. Trying to measure broadband deployment when what you count and what you don’t is arbitrary becomes problematic.
As I mentioned in the post, some have suggested that you eliminate the counting of businesses at all and simply measure residential connections whether they are DSl, cable, Wi-Fi, T1, or anything else.
February 7th, 2008 at 4:47 pm
[...] « The Trouble with Broadband Deployment Statistics [...]
February 8th, 2008 at 4:16 pm
Do we need to count connections or subscriptions? I think it would be difficult to count the number of people connected per subscription.
Cheers
February 8th, 2008 at 4:27 pm
That actually gets directly to the heart of the problem I addressed in the second installment of this series. Counting the number of subscriptions as a function of “per 100 inhabitants” becomes problematic for exactly that reason.
Due to fluctuations in household size, 100 subscriptions in the US reaches about 20% more poeple than the same 100 subscriptions in Denmark, for instance. It adversely impacts our ranking depsite having more people covered per connection.
February 8th, 2008 at 4:29 pm
Sorry Michael, had not gotten that far yet. Been a busy day.
February 8th, 2008 at 4:30 pm
No worries. Just thought I’d point you that way.
May 5th, 2008 at 7:45 pm
[...] National Cable & Telecommunications Association blog did a series of posts back in February about the OECD study. There seems to be three basic criticisms. First, [...]
May 19th, 2008 at 6:11 pm
[...] The Trouble with Broadband Deployment Statistics [...]