I remember one of the old hands in the Queensland Government Statistician’s Office telling me back around 2000 that they once queried the ABS about the reliability of labour force data for Queensland, and were told the main point of the Labour Force Survey is to produce a reliable national unemployment rate estimate, which is essential for the RBA in setting monetary policy. In other words, you should be grateful you’re getting any State data at all. So there have always been doubts about the reliability of labour force data at the State level, and this appears to be the reason Queensland Treasury has consistently used trend estimates in its briefings, even though the seasonally adjusted data are much more likely to identify turning points in the labour market.
Given the concerns about the State-level labour force data, I’ve always been super skeptical about the reliability of data at the regional level, and recent batches of regional data have confirmed that skepticism. Loose Change in a post earlier today (Employment breakdowns) points out the wacky labour force data for North Queensland we’ve seen recently. These numbers are so wacky that I’m starting to worry about the quality of the administration of the Labour Force Survey in North Queensland. I would encourage the ABS to undertake an urgent review of these data to establish whether the extreme volatility is due to (a) sampling error associated with the small sample size in the region or (b) failure to consistently obtain representative samples of households in this region.
Might get a better estimate of movement from centrelink customer data (if that’s available?) for unemployment. Census should be pretty good for August 2011 as a base too. ABS budget cuts/efficiency dividend will likely mean that Nth QLD employment figures wont be high up their list of priorities. I haven’t looked at the ABS figures for Nth QLD but one would hope the rotation policy for the survey would keep that volatility to a minimum (7/8th of the sample is the same each month).
Thanks for the comment, Ben. There are really large movements over a 12 month period in the NQ data (e.g. a drop in the participation rate of six percentage points based on 12-month moving averages). I’ll have a look at the Centrelink data. I know DEEWR base their small area labour market estimates on these data but they benchmark to the ABS regional data so I expect the DEEWR data will have the same problems. Hence looking at the raw Centrelink data may be the best way forward.
Totally agree. I have relied on Roy Morgan Research for unemployment figures for some time. Frankly, I gave up on ABS for reliable and meaningful statistics some years ago. I was advised by Roy Morgan in October 2012 that they are aware the RBA “look closely at our figures and one would hope they factor them in when making decisions about the level of interest rates in the country”.
I believe the Roy Morgan figures capture hidden unemployment that is not apparent in the standard ABS figures.
Thanks for the comment, Chris. I’ll check out the Roy Morgan data.
the ABS also publishes ‘underemployment’. People just need to dig a little bit deeper than the headlines. Generally speaking, the ABS figures are based on a vastly more robust/larger sample than Roy Morgan. I know everybody loves to hate the ABS cos they think the Government is trying to have a lend of ’em but the reality is the ABS is doing a pretty good job! The ABS so much better resourced. there is just no comparison. They are about as independent an organisation as any in Australia.
Thanks for that Ben. My interest is “labour force underutilisation rate”. Of course the ABS is better resourced than Roy Morgan and that is the problem. I don’t love to hate the ABS I just prefer to take my sources from others, with reference to the ABS, rather than blindly relying on just the one source. One can data mine and cube ABS figures as much as one likes but there will always be a discrepancy. Between the two providers, with respect to your views, I believe Roy Morgan publishes more realistic figures.
I do find it hard to believe that unemployment has shot up over the past couple of years from ~7% to 10.9% using the RM. That happens when there is a deep recession. We’re not in a deep recession. On my numbers (which are now a couple of months out) Centrelink NSA numbers have only increased from about 640K to 680K over the past 3 years (To Dec 2012). RM would talk about ‘discouraged’ job seekers but we’ve also had some growth in the lab force ’employed’ category + relatively strong AWE numbers + trend GDP + the numbers on family payments would have increased and or the number of $ payments on family payments would have increased. As these people become ‘unemployed’ their centrelink payments (particularly FTB A/B) would naturally increase – simply hasn’t happened. Another concern with any non-ABS survey is the rather awful response rates these private sector surveys get. I’m really not sure what’s going with the RM numbers atm. It’s very odd. They did track pretty well for years but the recent couple of years are weird!
The North-NW numbers are those seriously questionable. The regional data is obviously always questionable and volatile but when adjusted for trend, or a 12 month average, the data for Far North does correlate or make sense with what is perceived to be happening. However, North-NW is almost non-sensible?
It cant be the monthly volatility we already know about and sample size when we get such massive divergence in 12 month averages. Additionally the anomaly in North-NW employment is entirely female employment.
This has been previously reported in the Townsville Bulletin with retail getting the most blame? Really? Clearly something has gone badly wrong in a not insignificant region!
Re comments here on the Roy Morgan survey I think Gene hit the nail on the head here that what mattered was a reliable consistent national standard related to RBA policy. This seems to mostly work given our economic record?
The purpose of the survey is not an accurate count of unemployed but rather a statistically consistent series to correlate with policy. It does this ok at a national level! The Roy Morgan survey has some methodological problems in this regard and I particularly note the previous comment that the RM survey is realistic which entirely misses that point.
There was a heap of commentary on RM early this year when the data diverged from ABS. Mostly uncomplimentary to the RM methodology?