Australia might need the NBN, but the OECD data is meaningless
International broadband measures run wide but not deep
No, please no, not another argument about whose position the OECD broadband data supports. This happens every time there's an OECD data release, or even when there's no release, but merely a statement about the last data.
It starts, as always, with a political interpretation of OECD broadband data. Since this is practically the only use to which the data is put (at least in Australia), it has taken on a form of political trolling: whoever might be the relevant minister at any given moment (whose position on a variety of topics is already known) will use the OECD data to illustrate his point.
Journalists, who are trained never to let a ministerial press release pass, then take one of three possible approaches to the ministerial statement: write it into a story without further research or revision; call the minister’s opposite number for a formal refutation; or call some analysts for interpretation of the OECD data.
Alternatively, you could try and parse the OECD data for yourself, which is what I’m about to do.
How many carriers?
The first problem with using this data for a serious policy debate is that it's a mile wide, and about an inch-and-a-half deep. Let's look at the data gathering for broadband price comparisons:
A set of three operators per country has been chosen for comparison: the incumbent telecommunications operator, the largest cable provider (if there is cable coverage) and one alternative provider, if available, over DSL, cable or fibre.
For Australia, this yields either two or three carriers as the sample: Telstra, as the incumbent operator; either Telstra or Optus for cable prices (Telstra is the largest provider, but Optus is the largest non-Telstra cable provider), and one other.
Whether or not the OECD data fairly represents Australia’s price range depends entirely on the identity of the unnamed third provider. It may be someone who offers seriously competitive pricing, or it may not. We don’t know.
The second problem is that "price per megabit per second" is a meaningless measure for any country in which download caps are common. Speed does not determine plan price in Australia: download limits do. A 24 Mbps ADSL2+ service here might go out the door at $19.95, but with only a skinny download allowance (Telstra’s best price right now is $9.95 per month – but you only get 2 Gbytes per month).
In the middle of 2010, the price of a gigabyte’s worth of downloads in Australia ranged upwards from 35 cents.
Australia's maximum price "per megabit per second" will have been for a premium plan – one with "unlimited" downloads (although the regulator, the ACCC, doesn’t like that expression) or with 500 GB or 1 TB allowance. The top price, in other words, doesn’t reflect speed, but what you can do with the connection.
The point is that the OECD is seeking to “normalize” a complex set of service prices to a simple benchmark. This process inevitably demands compromises and, ultimately, inaccurate representations of data.
An average of what?
The third problem is that the bulk of the OECD’s tables – everything except the table reporting price range – talks about “average monthly subscription price”.
The “average price” of broadband plans, as advertised by a maximum of three providers, doesn’t tell us anything useful. As I noted in my second point, the most expensive plans on a per-Mbps basis are those with the highest allowances; they also have the smallest number of users, because they’re so expensive – but those plans drag the average upwards.
Averaging the advertised price tells us nothing about what users actually pay for broadband. Broadband prices advertised by Telstra, Optus and iiNet (apologies to others who might feel left out) today range from $10 per month to $120 per month.
But nobody’s paying $120 per month – well, hardly anybody – and I know this for a fact, because the providers’ ARPU (average revenue per user, as reported in their financials) sits around $40 or $50 per month. More than $50 ARPU gets champagne corks popping.
With so many shortcomings in the OECD data, you would think it were a sitting duck for an opposition communications spokesman wanting to make a media splash.
Distribution of access
“It’s all very well for the minister to justify his expensive and unnecessary National Broadband Network with reference to the OECD data, but we’re all aware that this data is not a sound basis for serious policy-making,” Malcolm Turnbull might have said – but didn’t.
Instead, he responded with a bit of statistical sleight-of-hand of his own, about the way broadband access is tied to income.
His statement was that "94 per cent of households earning more than $120,000 a year have access to the internet at home compared to only 43 per cent of households earning less than $40,000 have access to the internet."
Like all good mislead-with-statistics statements, it's absolutely true. It's also incomplete.
There is, as Turnbull said, a close correlation between income and access to broadband. But there’s also a close correlation between income and location. Households with an annual income less than $40,000 are also more likely to live far from the cities where broadband is most available.
Income is correlated not to one “broadband metric”, but two. People need to have the income to buy a broadband service – but even if someone on a low income wanted broadband, they’re most likely to live outside the footprint of broadband services.
For them, the question of price "per megabit per second" is irrelevant. ®