This little snippet at the foot of today’s building consents release from Statistics NZ shows how kind I was to readers and to statisticians by declining to use seasonally adjusted figures many years ago.
In a note to today’s release, Statistics NZ said: “We have improved the way we calculate the seasonally adjusted number of new homes consented. We now include an adjustment for the timing of Easter. As a result, the seasonally adjusted increase in the number of new homes consented in August 2017 has been revised down from 10% to 5.9%.”
For its latest seasonal adjustment calculations, Statistics NZ said: “The seasonally adjusted number of new dwellings consented fell 2.3%, following a 5.9% rise in August. For houses only, the seasonally adjusted number fell 1.7%, following a 3.1% fall in August.
“The trend for the number of new dwellings consented increased, and is at its highest level since early 2004.”
I haven’t gone back to Statistics NZ to ask what Easter has to do with August. Such a large revision – what looks like an admission of a 41% miscalculation, and lacking a real, credible explanation – will keep me wary of these adjustments for some time yet.
In these columns, you’ll continue to get comparisons from year to year, one month against the same month. Or, as I noted in May 2008, my solution when I quoted then-Government Statistician Geoff Bascand, showing the difficulty Statistics NZ had with seasonal adjustments: “The earlier occurrence of the Easter holidays in March, rather than April, may have contributed to this increase, although the exact effect is difficult to measure.”
My solution was to lump the 2 months together, March + April, when comparing hotel occupancy, for example.
But the changes are refreshing
Government Statistician Liz MacPherson warned of this month’s change in the September release on building consents, under the heading Upcoming changes to seasonally adjusted & trend series. I’ve repeated her message below:
“We are improving the way we calculate the seasonally adjusted & trend series in building consents issued. These changes will be introduced in the September 2017 release (published on 31 October 2017).
“All seasonally adjusted series will now include an adjustment for the timing of Easter. This will account for when Easter moves between March & April. This change will affect the entire time series.
“We are also updating the way we treat outliers in the trend for the value of non-residential building consents. Currently, we exclude consents with a value of $50 million or more from the calculation of the trend. This threshold will be increased to $100 million, backdated to 2006. Currently, these outliers are only excluded from the monthly trend. For consistency, we will now also exclude these outliers from the quarterly trend.”
Despite my scepticism about some calculations, I’m enjoying the changes emanating under new leadership at Statistics NZ. They’re aimed at giving more people better information that they can use – a worthy cause.
Statistics NZ: Building consents issued seasonal adjustment and trend changes in September 2017
Related story today: New home consents jammed in 1000/year range
2 October 2017: A new understanding of seasonal adjustment
13 May 2008: Campers lift March accommodation use, but hotel & motel occupancy down
1 February 2008: Statistics, lies & don’t knows
2 September 2006: Pick an apple, an orange and you can concoct statistical fruitcake
13 May 2006: Late Easter takes March occupancy down
8 May 2006: Don’t believe everything you read…
Attribution: Statistics NZ release.