Saturday, October 12, 2013

An insignificant figurer

The ABS labour force survey statistics came out this week with analysis from Conus and Grog's Gamut:

Conus: Surprise fall in Aus unemployment rate
Grog's Gamut: Australia's Unemployment Rate steady at 5.7%

The headlines belie underlying agreement in their analysis. The difference of course is between a fall in the widely reported seasonally adjusted number and a flat trend number. Both Conus and Grog also refer to the impact of rounding on the seasonally adjusted data as explained at Grog:
And anyway if we get down to the nitty gritty, the seasonally adjusted figure in August was 5.7647% so it only just got rounded up to 5.8%, and September’s number was 5.6479%, so it only just got round down to 5.6%. Thus while it looks like a 0.2 percentage point drop it’s really only a 0.12 percentage point drop.
 It was curious to then see this tweet come through from Queensland Treasurer Tim Nicholls:
Employment growth 0.2% in Qld in Sept and trend participation rates up every month in Sept qtr.
So lets look at the September quarter trend participation rates for Queensland:

June   65.8%
July   65.9%       
Aug   65.9%
Sept   65.9%

Looks pretty flat to me, eh? You see the ABS publish this data to one decimal point however the ABS spreadsheet downloads are to seven decimal points which becomes for the September quarter:

June   65.8364146%
July   65.8605645%       
Aug   65.8971742%
Sept   65.9365453%

So was the participation rate up every month in the quarter or not? Perhaps here we need to consider significant figures and for simplicity I will refer on this to Wikipedia:
The significant figures of a number are those digits that carry meaning contributing to its precision. This includes all digits except:
Spurious digits introduced, for example, by calculations carried out to greater precision than that of the original data, or measurements reported to a greater precision than the equipment supports.
Given the known issues around the ABS labour force survey data it could be suggested that the significance of any number beyond the first decimal published by ABS may be approximately zero? Note that while Conus / Grog have drawn attention to the rounded numbers as cause for caution Treasurer Tim has instead attributed significance to 'precise' numbers where there is none!

Even more curious is that this was one of the key significant (sic) highlights he could find in the Queensland employment data? Since the state election in March 2012 the Queensland trend participation rate has now declined by 0.9pcpts while the national rate has declined by only 0.3pcpts.


Example of an insignificant figurer

No comments:

Post a Comment