Lots of commentary about the terrible jobs data for April in Townsville released by the ABS on Thursday (see here). The “original”, unadjusted data from the ABS (and the only series for the regional data that they produce) had the unemployment rate at 13.9%. Our own Conus Trend came in at 10.5%. The 12 month moving average (simply the last 12 months of original data added together and divided by 12) was 8.9%. So which one’s right?
The Queensland Government, in the shape of Employment Minister Grace Grace, have come out and said that “The most accurate measure of Townsville’s unemployment rate is the 12 month rolling average, and this is currently 8.9 per cent.”
Unfortunately for Ms Grace this statement alone shows a basic lack of understanding of the data. It’s not a question of which is the “most accurate”. The original ABS data is not “accurate” in the sense that it tells us precisely what the unemployment rate is. It is based on a (relatively small) survey and is therefore simply a best-guess, or estimate, based on the data they have collected. This original data is the basis, and only basis, for the calculation of the 12 month moving average. We also use this original estimate data to calculate our Conus Trend. To suggest one is more “accurate” than the others doesn’t make sense. Two are simply mathematical constructs of the other.
The more useful question we should be asking is “which measure is more likely to provide us with meaningful information?” To answer that question we can simply turn to the experts; the ABS themselves. The ABS say..”Trend estimates are usually preferred to compare data at different points in time as potentially misleading seasonal patterns, residual noise and irregular influences have been removed.” Indeed they go so far as to tag their own original data with a “Do Not Use” label for making comparisons. We would agree and suggest that if we want to get a meaningful picture of what is happening with this data set we need to consider a Trend series.
Unfortunately when it comes to regional data the ABS simply do not have the resources to provide Trend estimates for all data sets. We have therefore stepped in to fill that gap. Using a highly sophisticated statistical model we have created the Conus Trend series for all the Queensland Regions. We would argue that this, rather than the original or moving average series, are what we should be looking at. Note that the ABS don’t rely on a simple moving average of original data for any of their own series but rather construct seasonally adjusted and Trend series; and for good reason. A simple moving average tells us next to nothing about what trends are underlying the data and gives equal weighting to both data from a year ago and a month ago.
Is the unemployment rate in Townsville precisely 10.5% as our Conus Trend suggests? Almost certainly not, but we would argue that this is the closest we can get to a real idea of what the actual rate might be.