Wednesday, 13 July 2016

Lies, damned lies, and Irish statistics

In March, Ireland's statistical agency, the Central Statistics Office (CSO), estimated that Ireland's GDP had grown in real terms in 2015 by 7.8%.

Yesterday the CSO came out with a revised estimate. It now says that Ireland's GDP in 2015 grew by - wait for it - 26.3%.

This is both absurd, and yet technically correct. Absurd, because as the Irish Times commentary headline put it, 'Crazy growth figures bear scant relationship to reality'. Yet technically correct, because the CSO says it follows the methodology of the "European version of the current UN mandated international standards for national accounts statistics, the System of National Accounts (SNA) 2008". And there's nothing wrong with doing that: our own Statistics NZ uses the same UN approach (details here if you're ever looking for them).

What's actually happened is that, for assorted tax reasons, over a short period of time in 2015, a number of international companies shifted the domicile of patents they own, or aircraft they lease, or their own corporate domicile, to Ireland. And, apparently, if you apply the standard national accounts methodology to those transactions, you get 26.3% GDP growth. I say "apparently" because the logic of some of the accounting escapes me, but let's take it at face value that the Irish statisticians cranked the right handles and out came the "right" UN-consistent answer.

There is now, as you can imagine, a big barney going on in Ireland about the reliability of the GDP statistics and how can people tell how the economy is actually behaving, but I was struck by two other thoughts.

One was the complete absence of any helpful explanation from the CSO. Here is the complete text of their statistical release.

Is there any attempt to reconcile their earlier 7.8% stab at it with the new 23.6%? No. Is there anything helpful at all about what actually drove the new results? No. Nothing. Zip. Nada.

Or in Irish, neamhní, faic, dada, rud ar bith*.

The relevance to us in New Zealand is that there's been a bit of a debate, here and overseas, about how far statistical agencies ought to go in providing analysis or commentary on the statistics they produce. Statistics New Zealand, I'm pleased to say, is down the right end of this debate, and goes some way to help users understand what's going on.  As an example, the commentary on the latest GDP release, for the March quarter, told us that "The anticipated El Niño weather pattern was not as severe as expected. The normal seasonal fall in milk production was less pronounced than usual, resulting in seasonally adjusted volumes of milk produced increasing slightly", which is helpful when you're trying to make sense of the agricultural production component of GDP.

We don't want Stats to veer off the reservation into opinion or editorial, but we most certainly do want them, at a minimum, to keep up the level of explanation they currently provide. As for the CSO, it badly needs to develop some customer focus and join the 21st century**.

The other thought I had was the silliness of some media and financial market reactions to small changes in GDP from what they had expected. As the Irish example has inadvertently reminded us, GDP is an estimate, a more or less rough stab at the aggregate level of economic activity. It comes with various kinds of measurement and survey error, and has complex and debatable inbuilt assumptions, and not just the Irish ones around intellectual property and official domicile. The measurement of the output of the financial sector, for example, is a contentious issue.

Our latest official stab at GDP growth for the full year to March is 2.4% (or 2.8% just comparing March '16 with March '15). The reality is that "low to mid 2's" is probably just as good a description.

*Pronounced navnee, fack, dodduh, rud er bih, though the 'd's are more like the 'th' in the English 'the'. I particularly like faic, as in "the statistics make faic-all sense".

** (Update July 14) This is too harsh. While I'm still of the view that the statistical release was inadequate, the CSO did supplement it with a separate press release (see comments below). Yesterday the CSO also announced that, while it will continue to estimate GDP/GNP according to the international rules (as it is obliged to), it is also convening a new consultative group to look at "how best to provide insight and understanding of all aspects of the Irish economy", including "whether new presentations of existing information would improve understanding". That's a good move. In that context I hope they have a look at moving on from bare bones presentation of the data.


  1. See

  2. Thanks for that. I've probably been too harsh on the CSO - I hadn't gone to the press release area of the website, I went to the statistical release itself, and there's no clarification or explanation there about what's happened (and, I might add, no link on the release page to the explanatory press release), and what text there is completely ignores the elephant in the room.
    But that said, you're quite right, the CSO did put out a press release explaining the guts of what happened, and I didn't see it.
    The press release could have been improved by giving (if confidentiality allowed) some indication of the quantum of these corporate asset effects: I'd guess most if not all of the change from 7.8% to 26.3% was down to these corporate actions, but it would have been useful to know if the 'underlying' 7.8%, if I can call it that, was still in play or had been revised.