OECD vs. SpeedTest

by on May 5, 2008 · 8 comments

Nate Anderson points to a new report on broadband around the world that I’m looking forward to reading. I have to say I’m skeptical of this sort of thing, though:

Critics of the current US approach to spurring broadband deployment and adoption point out that the country has been falling on most broadband metrics throughout the decade. One of the most reliable, that issued by the OECD, shows the US falling from 4th place in 2001 to 15th place in 2007. While this ranking in particular has come under criticism from staunchly pro-market groups, the ITIF’s analysis shows that these numbers are the most accurate we have. According to an ITIF analysis of various OECD surveys, the US is in 15th place worldwide and it lags numerous other countries in price, speed, and availability—a trifecta of lost opportunities.

With an average broadband speed of 4.9Mbps, the US is being Chariots of Fire-d by South Korea (49.5Mbps), Japan (63.6Mbps), Finland (21.7Mbps), Sweden (16.8Mbps), and France (17.6Mbps), among others. Not only that, but the price paid per megabyte in the US ($2.83) is substantially higher than those countries, all of which come in at less than $0.50 per megabyte.

Now, this site is a tool for measuring the speed of your broadband connection, and it purports to have data from around the world. I have no idea how reliable their methodology is generally, or how good their testing equipment is around the world, but I’ve used it in several different places in the US and it at least seems reliable around here. According to their measurements, the US has an average broadband speed of 5.3 mbps, roughly what the OECD study said. But the numbers for the other countries cited are wildly different: Japan is 13 mbps, Sweden is 8.7 mbps, South Korea is 6.1 mbps, and France is 5.5 mbps. If these numbers are right, te US is behind Sweden and Japan, and slightly behind South Korea and France, but we’re not nearly as far behind the curve as the OECD reports would suggest.

And then there’s this:

The ITIF warns against simply implementing the policies that have worked for other countries, however, and it notes that a good percentage of the difference can be chalked up to non-policy factors like density. For instance, more than half of all South Koreans lived in apartment buildings that are much easier to wire with fiber connections than are the sprawling American suburbs.

Now, I haven’t examined SpeedTest’s methodology, so they might have all sorts of problems that make their results suspect. But it’s at least one data point suggesting that the OECD data might be flawed. And I think the very fact that there seems to be only one widely cited ranking out there ought to make us somewhat suspicious of its findings. Scott Wallsten had bad things to say about the OECD numbers on our podcast. Is there other work out there analyzing the quality of the OECD rankings?

  • http://ditzler.blogspot.com Wyatt Ditzler

    Tim, some random discussions I and some people at CableTechTalk had a bit ago.
    (http://www.cabletechtalk.com/news-items/2008/02/06/the-trouble-with-broadband-deployment-statistics/)

    Also this report came out of the same discussion:
    (http://www.phoenix-center.org/pcpp/PCPP29Final.pdf)

    With the apparent flaws of the OECD numbers, I still think that they are the best measurement we have at this time. That is to say there is a lot of room for improvement.

  • http://ditzler.blogspot.com Wyatt Ditzler

    Tim, some random discussions I and some people at CableTechTalk had a bit ago.
    (http://www.cabletechtalk.com/news-items/2008/02…)

    Also this report came out of the same discussion:
    (http://www.phoenix-center.org/pcpp/PCPP29Final.pdf)

    With the apparent flaws of the OECD numbers, I still think that they are the best measurement we have at this time. That is to say there is a lot of room for improvement.

  • http://tieguy.org/ Luis

    SpeedTest’s methodology is ‘we test whoever visits the site.’ It is about as unscientific and invalid a polling methodology as you can get.

  • http://tieguy.org/ Luis

    SpeedTest’s methodology is ‘we test whoever visits the site.’ It is about as unscientific and invalid a polling methodology as you can get.

  • http://internetthought.blogspot.com Rudolf

    The OECD’s methodology is that they publish the numbers as they are provided by the member states and/or as they are available from public sources.

    What the OECD publishes is either the broadband penetration numbers they receive from the nations themselves (FCC, department of commerce etc provide those numbers for the USA) and they look at the offers from the largest operators in each nation to be able to do a price broadband comparisson. For a good explanation of how this works see the explanation Taylor Reynolds of the OECD provides. http://www.itwire.com/content/view/12224/1154/

    What you are doing is you measure the throughput to a US site (Speedtest) from Korea and then compare this with the numbers as given. In doing this you forget that trans-pacific traffic is still a bit expensive and therefore often rationed. So while Koreans may get 50mbit/s on the network of their provider, of net in Korea this already may be lower, but towards the US this is definitely lower. Furthermore and on top of this, Speedtest relies on people going to their website and testing the speed. It just might very well be that Koreans are not too interested in the speedtest website, seriously affecting its usability.

  • http://internetthought.blogspot.com Rudolf

    The OECD’s methodology is that they publish the numbers as they are provided by the member states and/or as they are available from public sources.

    What the OECD publishes is either the broadband penetration numbers they receive from the nations themselves (FCC, department of commerce etc provide those numbers for the USA) and they look at the offers from the largest operators in each nation to be able to do a price broadband comparisson. For a good explanation of how this works see the explanation Taylor Reynolds of the OECD provides. http://www.itwire.com/content/view/12224/1154/

    What you are doing is you measure the throughput to a US site (Speedtest) from Korea and then compare this with the numbers as given. In doing this you forget that trans-pacific traffic is still a bit expensive and therefore often rationed. So while Koreans may get 50mbit/s on the network of their provider, of net in Korea this already may be lower, but towards the US this is definitely lower. Furthermore and on top of this, Speedtest relies on people going to their website and testing the speed. It just might very well be that Koreans are not too interested in the speedtest website, seriously affecting its usability.

  • Pingback: 1300 number

  • Pingback: fantasy football asia

Previous post:

Next post: