DSL Prime: The Data Is Wrong

Look closer at broadband statistics around the world, and you'll find discrepancies that can be fixed, says a researcher. Here are her results.

by Dave Burstein of DSL Prime and Future of TV and the Web Video Summit [June 7, 2007]

Shara Evans: Numbers Need More Precision
Countries may be closer than they appear in the rankings
Key Takeaway: Most countries are much closer together than the rankings suggest
Shara Evans of Market Clarity produced a remarkable labor of love, a 60 page report that looks deeply at how the broadband statistics are derived internationally. The handful of us who care about the details of the numbers will spend hours pouring over the analysis. Most interesting to everyone else is her discussion of "bands"—a set of nations essentially so close that they are identical even if one is ranked 4th and another 8th. There's no simple test for the "margin of error" in data compiled from many sources, but clearly differences of a few percentage points could be data errors rather than actual results.

The OECD figures are carefully collected and sensibly analyzed. In general, they are highly accurate. In particular, I have looked several times over the years at individual numbers in the OECD tables I had reason to doubt, and confirmed them. However, Evans found reputable data sources in many of the countries with substantially different figures than the OECD, both higher and lower. Few countries have totally accurate official record collection, with definitions, dates and accuracy varying between the different sources. Some countries effectively separate business from residential figures, but other make arbitrary or no differentiation.

An extreme example of bad data from reputable sources is the Pew Study. This respected non-profit contacted 3,000 U.S. adults and claimed "the margin of error on the overall sample is +/- 2 percent." They however had a massive error, claiming "As of March 2006, DSL connections constitute half (50 percent) of all home broadband connections and cable modems have a 41 percent share." In fact, the cable share was then about 5 points higher than DSL, based on the audited financial statements of the companies filed with the SEC. Errors that large are rare, but available figures should be presumed precise to 1 percent, and errors of 5 points are certainly possible. (I have no clue on why Pew was so far off, only some of which is a high count of wireless. I hope they go back and check.)

I've long suspected that Beijing figures did not include many apartment complexes that were wired for 10 to 100 megabits by the builder. I've recently heard from Russia that similar "building provided" broadband is a very large factor not in the government count.

This study (and others less reputable) is already being abused to pretend that real policy failures are simply misinterpreted data. Korea, Singapore, Hong Kong and most of Scandinavia are well ahead. Canada, Taiwan and Switzerland are doing well. The U.S., UK, and Australia are significantly behind. Spain, Germany, and Italy have even lower penetration.

In Shara's spirit, here's a banded ordering based on households rather than population, using Point-Topic's Global Broadband statistics. Point-Topic (like OECD) offers an highly accurate data set. P-T gives household as well as population rankings, which the OECD has avoided because the count of "households" is notoriously hard to pin down. Rankings by populations are generally similar, however.

Banded Country Ranking, by households

Band 1: Proving what's possible

• South Korea 89 percent
• Hong Kong 84 percent
• Iceland 76 percent

Band 2: Doing well

• Netherlands 71 percent
• Denmark 70 percent
• Singapore 70 percent
• Israel 69 percent
• Switzerland 67 percent

Band 3: Not embarrassing

• Canada 62 percent
• Taiwan 61 percent
• Norway 60 percent
• Finland 59 percent
• France 56 percent
• Japan 54 percent

Band 4: Muddling

• UK 52 percent
• Belgium 52 percent
• Sweden 52 percent
• Estonia 50 percent
• USA 50 percent
• Australia 50 percent

Band 5: Disappointing

• Spain 46 percent
• Ireland 43 percent
• Portugal 42 percent
• Austria 42 percent
• Slovenia 40 percent
• Italy 39 percent
• Germany 38 percent
• New Zealand 34 percent

I have an item I held over explaining some of these numbers. Martin's comment that rural areas are lower was suggestive, because rural areas are also poor. The U.S., with a Gini coefficient of 0.45, has more relative poverty than any other developed country, which is a thoroughly unattractive explanation of why we are behind. In affluent countries like this, the key variable explaining take rate is price, which explains 50 to 80 percent of the differences. Price, in turn, is largely a function of competition—more than four (Korea, Japan, France) produces leaders, duopolies like the U.S. generally fall behind, and countries close to monopoly (Spain, Italy, and Germany until recently) have the worst results.

Copyright 2007 Dave Burstein.
The DSL Prime Newsletter is reprinted with permission.

"The power of the printing press belongs solely to those who own the presses"
—A.J. Leibling

The Internet is the cheapest printing press ever invented.