With the huge volatility in State unemployment rate estimates, demonstrated by the ABS seasonally adjusted estimate of Queensland’s unemployment rate falling from 6.4% in January to 5.6% in February, and the SA rate increasing from 6.8% to 7.7%, the ABS really needs to invest greater resources into generating better State-level estimates. I did my best to explain what the latest labour force data mean for Queensland to the Courier-Mail yesterday (see Qld February 2016 unemployment rate adjusted to 5.6%):
Economist Gene Tunny, from Adept Economics, said southeast Queensland was cracking ahead while regional areas were struggling.
He said while the volatile figures were hard to read, there definitely was a continuing trend in jobs growth.
“Gold Coast is going well with the Commonwealth Games and Brisbane with construction but other areas have the loss of the resources industry,” he said. “Overall, I think the state economy is really a bit lacklustre.”
On yesterday’s figures, I would highly recommend Paul Syvret’s excellent column in today’s Courier-Mail, in which he too expresses concern about the reliability of the data (Qld unemployment data for February 2016 explained):
The problem we have right now is that the labour force data is wildly erratic, making it hard for policy makers to get a handle on what is actually happening in the real economy.
The volatility in the State-level estimates should add to concerns about the reliability of the national seasonally adjusted unemployment rate, which is a key indicator monitored by the Reserve Bank in advising on monetary policy. The ABS may promote its trend estimates, but they are poor substitutes for reliable seasonally adjusted data. As I have noted before, the trend estimates are simply smoothed, filtered versions of the underlying seasonally adjusted data, and the trend estimates will tend to disguise turning points that may be occurring, because they are smoothing the time series. Eventually the trend data will be revised, and turning points will be revealed, but analysts will miss the turning points when they really need to see them if they rely on the trend data alone.
Having being part of the survey sample the deficiencies are obvious. To only take one week of each month over a 6 month period would provide fluctuations that don’t reflect whether some is employed or not as we would expect it would. It only provides a snapshot of a families employment status of that week, how many hours they worked, how many hours that would normally work, if they didn’t work why not, and if they didn’t work that week are they actively looking for work. ( define active)
Further to this is the screening process as to whether they will accept you as someone to take part in the sample survey. Regional Qld in particular would be open to large fluctuations in unemployment due to the large numbers of people working in mining and seasonal industries such as agriculture and tourism as well.
A place like Townsville for example the unemployment rate here is distorted by the large defence presence in the city. Nearly 7000 defence personnel are not allowed to be included in the survey, so none of the members of that persons household are included either. Even allowing only one other person in each household that is employed and not included that would leave 14,000 employed persons that are excluded from the figures. That is a very large amount out of a sample of 250,000 people and virtually leaves the Townsville figure irrelevant, particularly from the total employed persons column.
For what its worth I just look at Petes trend at Conus, he seems to iron out the bumps.
Very good point about the defence population in Townsville, Glen. And, yes, Pete’s trend series is worth looking at. Thanks for the comment.
Pingback: Last Week in Queensland – 21 March 2016 | Ethical Consulting Services