Julia recently posted this useful article, which indicates the latest unemployment figures may well be way out of whack: https://www.aussiestockforums.com/forums/showpost.php?p=397107&postcount=11
Of course the major headline in the papers, was that unemployment was not quite as bad as expected. But that doesn't quite stack up with the large number of job layoffs announced recently.
Anyhow, I've had a bit more of a look into how the ABS calculates the figures (as published on the ABS website in various notes), and it's quite interesting, and more importantly there are several areas that seem prone to problems, *especially* in January data!
* ABS takes a survey based method to calculate the employment figures, based on latest population and labour estimates of total workers. Therefore they interview in person, and then over the phone in subsequent interviews people for 8 months, with 1/8 being replaced each month (hence getting a new sample set in that period).
* It *does not* go on Centrelink figures, which is possibly contrary to what many might think. The survey only captures 0.24% of the population. The sample size was last year slashed by the Rudd government!
* Unemployed persons are those that listed they were actively looking for work in the 4 weeks proceeding the survey week. Of course, if one was a recipient of a multi-month redundancy, one would be likely to take it easy for several weeks, before considering looking for work (most especially over the xmas period).
* The January survey is done 2 weeks following the 7th and 13th of the month. Now I can see a problem here for those not employed end of the year, start of January, where a lot of employers aren't even around, aren't conducting interviews. Most potential employees aren't going to be actively seeking work, hence probably won't match the criteria of being "unemployed".
* The December figures were revised upwards by some +0.3% in the latest unemployment figures! If similarly the January figures were to be revised upwards, the seasonally adjusted unemployment figure could well turn out to be 5.1% - into the 5's.
* At the bottom of each release, it's worth looking at the 95% confidence errors. In January, unemployment was between -0.1% and +0.7% of the figures. Hence there is a 95% chance the seasonal unemployment figures lie anywhere within the range 4.4% and 5.5% !. A pretty large range, and what could be the difference between "hey they weren't that bad", and "oh damn, we're screwed!"
* The trend series is the main item people tend to look at, this is derived with a "surrogate of a Henderson moving average" of the 6 most recent seasonal figures. So hence if the seasonal figures are later revised UP, so will the trend series.
* I can see a problem with the survey based method in isolated country towns, small mining towns, especially likely to have been hit with recent mining layoffs. Is the ABS really going to send a surveyor 1000km into the middle of nowhere to take employment surveys?? Doesn't seem likely to me. Could be a very large gap in the survey.
* Survey method seem partially adequate, but is not going to capture people who are frequent movers, those that frankly have better things to do, as these surveys can take 30 minutes or so to fill out, even over the phone. I can see large numbers, many different groups of people simply not the type that could be bothered participating when asked. This could introduce bias in the figures, especially with some demographics.
Please correct any errors above if you have any better information.
Now I'm not necessarily having a go at the ABS, they are just using whatever figures they have available to them, but with recent cuts to sample size, and the above, it's worth keeping in mind to be a little skeptical.
Of course the major headline in the papers, was that unemployment was not quite as bad as expected. But that doesn't quite stack up with the large number of job layoffs announced recently.
Anyhow, I've had a bit more of a look into how the ABS calculates the figures (as published on the ABS website in various notes), and it's quite interesting, and more importantly there are several areas that seem prone to problems, *especially* in January data!
* ABS takes a survey based method to calculate the employment figures, based on latest population and labour estimates of total workers. Therefore they interview in person, and then over the phone in subsequent interviews people for 8 months, with 1/8 being replaced each month (hence getting a new sample set in that period).
* It *does not* go on Centrelink figures, which is possibly contrary to what many might think. The survey only captures 0.24% of the population. The sample size was last year slashed by the Rudd government!
* Unemployed persons are those that listed they were actively looking for work in the 4 weeks proceeding the survey week. Of course, if one was a recipient of a multi-month redundancy, one would be likely to take it easy for several weeks, before considering looking for work (most especially over the xmas period).
* The January survey is done 2 weeks following the 7th and 13th of the month. Now I can see a problem here for those not employed end of the year, start of January, where a lot of employers aren't even around, aren't conducting interviews. Most potential employees aren't going to be actively seeking work, hence probably won't match the criteria of being "unemployed".
* The December figures were revised upwards by some +0.3% in the latest unemployment figures! If similarly the January figures were to be revised upwards, the seasonally adjusted unemployment figure could well turn out to be 5.1% - into the 5's.
* At the bottom of each release, it's worth looking at the 95% confidence errors. In January, unemployment was between -0.1% and +0.7% of the figures. Hence there is a 95% chance the seasonal unemployment figures lie anywhere within the range 4.4% and 5.5% !. A pretty large range, and what could be the difference between "hey they weren't that bad", and "oh damn, we're screwed!"
* The trend series is the main item people tend to look at, this is derived with a "surrogate of a Henderson moving average" of the 6 most recent seasonal figures. So hence if the seasonal figures are later revised UP, so will the trend series.
* I can see a problem with the survey based method in isolated country towns, small mining towns, especially likely to have been hit with recent mining layoffs. Is the ABS really going to send a surveyor 1000km into the middle of nowhere to take employment surveys?? Doesn't seem likely to me. Could be a very large gap in the survey.
* Survey method seem partially adequate, but is not going to capture people who are frequent movers, those that frankly have better things to do, as these surveys can take 30 minutes or so to fill out, even over the phone. I can see large numbers, many different groups of people simply not the type that could be bothered participating when asked. This could introduce bias in the figures, especially with some demographics.
Please correct any errors above if you have any better information.
Now I'm not necessarily having a go at the ABS, they are just using whatever figures they have available to them, but with recent cuts to sample size, and the above, it's worth keeping in mind to be a little skeptical.
