Government heat map ‘wake up call’ to stop…

Climate Media Centre Advocacy groups have welcomed the release of the Federal Government’s…

Compulsory income management doing more harm than good:…

Charles Darwin University Media Release Compulsory income management (CIM) in the Northern Territory…

Flicker of Hope: Biden’s Throwaway Lines on Assange

Walking stiffly, largely distracted, and struggling to focus on the bare essentials,…

Seizing a Future Made in Australia

Climate Council Media Release THE CLIMATE COUNCIL celebrates today's announcement that the Future…

The Meanjin essay: The Voice and Australia's democracy…

With Stephen Charles AO KC The dire state of truth in Australia’s civic…

Haunted by waters

By James Moore We were young when we lived near the Rio Grande…

The price of victimhood: The Higgins/Lehrmann gravy train

By Bert Hetebry I’m not much good at sums, but I can imagine…

An Open Letter: Save Toondah - it’s the…

By Callen Sorensen Karklis Dear Readers, Seventeen years ago I was inspired by…


What’s happening with the Labour Force?

Before you believe all the spin that economists and politicians put on the latest Labour Force figures, you would have to think that there’s something odd going on with November’s labour market, just as there was with the figures for October 2015.

The most glaring oddity for the November figures is the number of jobs created, 71,400 in total, of which 29,700 were part time.

If we take this figure as reported that means over the 22 working days of the month, 3245 new jobs were created, every day. Put another way, that’s over 400 new jobs every working hour of the month. That’s a tough one to swallow.

And before the government gets too excited, one needs to point out that while 71,400 jobs were ‘created’ the number of people unemployed decreased by only 2800. How come? Because, according to the ABS the participation rate increased.

That is, unemployed people who had either lost their job or who had been unemployed and had previously told the survey caller they were not looking for work, were now saying in the new survey that they were looking for work, so they were added to the new count as unemployed.

But these people were still unemployed whether they were counted or not, which casts a shadow over the classification of the participation rate. At some earlier stage, these people had a job and either left it, or were sacked.

Therefore, at some point those 68,600 people in the November survey classified as being part of an increased participation rate, lost their jobs.

That resembles a revolving door with almost as many going out as are coming in. This does not translate to an improving economy. Considering that these are seasonally adjusted figures, their volatility suggests there is something wrong with the formula.

marThe following is an explanation of the formula used by the ABS. This description is taken directly from Peter Martin’s article in the Sydney Morning Herald one month ago, when he reported on the October figures. The heading was, ABS labour force figures can’t be believed – employment didn’t jump by 58,600 jobs in October”

I’m happy to let Peter’s explanation stand because reading the explanation from the ABS is a lot more complicated. If however you wish to forge you way through it, you will find it here.

Peter wrote,

“Every month the Bureau surveys around 26,000 households and asks who in them worked in the past two weeks and who did not. It surveys the same households month after month for eight months, then on the ninth month it abandons one eighth of its survey and replaces it with a new eighth, chosen at random. The new eighth stays in for eight months, and so on.

The idea is to get both continuity and change, so that it is not always surveying the same 26,000 households.

Every so often there’s something unusual about the one-eighth that is ejected from the survey (such as being highly unemployed) or something unusual about the one-eighth that is brought in (such as being highly employed). Or something unusual about both.

When that happens the employment total can jump (or dive) even if the employment experience of those who remained in the survey didn’t change, or didn’t change much.

That seems to be what happened in October. Yes, there was something of a “Turnbull Effect”, or perhaps it was a “Daniel Andrews Effect” given that the jump in employment was concentrated in Victoria. But it wasn’t as big as is claimed.

It is possible to work this out by examining what the Bureau calls the “gross flows”. These are the movements into and out of employment between the October and November surveys among only those households who were in both surveys. The gross flows leave out both the households that left the survey in October and the households that joined in November.

The gross flows show a genuine jump in employment in October, but one only half as big as the one published.

Around half of that figure of 58,600 newly-employed workers appears to have been a statistical artifact, caused by a difference in the type of households that left the survey and the type that joined.”

Remember, these are the seasonally adjusted figures we are referring to. If you would rather accept the trend figures instead of the seasonally adjusted figures then you will be even more confused. If anything, they seem to paint a more believable picture.

One has to think though, that there must be a better formula to accurately gauge the state of the labour market.

Suggestions are more than welcome.

Like what we do at The AIMN?

You’ll like it even more knowing that your donation will help us to keep up the good fight.

Chuck in a few bucks and see just how far it goes!

Your contribution to help with the running costs of this site will be gratefully accepted.

You can donate through PayPal or credit card via the button below, or donate via bank transfer: BSB: 062500; A/c no: 10495969

Donate Button


Login here Register here
  1. John Lord

    That figures John.

  2. keerti

    Add all that to the slavery for the dole scheme and f….knows where we are.

  3. Chris

    And labour hire company manipulation at the behest of governments….Here a one day shift that we are offering to our client employers at a discount rate….

  4. Wally

    Does the sample surveyed also provide the data used to determine the number of people unemployed? I can understand the reasoning for using survey data to determine changes in employment numbers, it would be impossible to survey every employer but if the unemployment figure is determined from the same source it has little if any usefulness. I always assumed unemployment figures were determined by the number of people registered with Centrelink as unemployed.

  5. John Kelly

    Wally, using the numbers registered for employment would seem the obvious starting point.

  6. Ella

    John, thank for the interesting read.
    A member of our family has been doing an office management course, all the while looking for work and attending one of those employment services agencies.. She is employable. In the last 2 years this company has provided her with one job interview. She sees them every 2 weeks I wonder how much this employment service charges the gov. It is an absolute rip off. The fact is that that jobs are as scarce as hen’s teeth. Regardless of the spin the gov puts on it.

  7. Alison White

    I would have thought Centrelink database figures would be the absolutely obvious place to source the figures. Hell, they could even identify under employment too. Why on earth is unemployment calculated that way!?

  8. Kaye Lee

    Sooooo….they want us to believe that full time employment went up by 41,600 people and part-time jobs went up by 29,700 people……but “Seasonally adjusted monthly hours worked in all jobs decreased 12.7 million hours (0.8%) in November 2015 to 1,645.9 million hours.”

  9. Geoff Andrews

    1. I would have thought that surveying families would be the most inefficient and inaccurate method
    (“Ah, shit! Here’s that bloody survey thing again. What’ll we tell ’em this month?”)
    2. I would have thought that a similar survey of businesses would be better and the way to do that for ALL businesses is to add a couple of questions in their monthly (or three monthly) BAS returns. “How many part time/full time employees on your payroll?” This would give a measure of the monthly and three monthly trends, which should agree (?)
    3. The most accurate measure must be every 5 years on census night so whatever methods that are being used now can be calibrated against the census.
    4. It could well be that sales at McDonalds or consumption of beer might give a more accurate measure of unemployment. 🙂

  10. Bacchus

    The problem with relying on Centrelink figures is that they don’t count those who are unemployed (or underemployed) but not receiving any Centrelink benefits. As an example, if one of a working husband and wife become unemployed, often the remaining working partner earns too much for the unemployed partner to qualify for Newstart, yet the unemployed person is still looking for, and available to start work in the coming week (part of the definition used to determine if someone is unemployed).

  11. Bacchus

    Of course ripping $78 million out of the ABS budget and sacking or not replacing 350 staff (including the chief statistician) might have something to do with it as well 😉

  12. Michael Taylor

    The problem with relying on Centrelink figures is that they don’t count those who are unemployed (or underemployed) but not receiving any Centrelink benefits.

    Bacchus, my former Branch in my former department used the Centrelink database to release the official figures of the number of Newstart recipients etc. Under Howard, the recipients whose payments were suspended (for say 8 weeks for a violation) were not counted as receiving Newstart. It was a sure way of making it look as though the recipient population had fallen. It gave the government the chance to say “Look at us. Our Welfare to Work policies are working”. Sneaky.

    The business rules were also changed as to how many hours a person needed to be working in order to be classified as ’employed’. It was reduced to some piddly amount (such as a couple of hours a week), again, so false figures could give Howard and Hockey (my then Minister) something the brag about. Also sneaky.

  13. Chris

    Michael Taylor Yes there is definitely lots of ‘sneaky’….. and unfortunately it all involves screwing around vulnerable people.

  14. Bacchus

    Yep Michael – Newstart figures are very susceptible to political “manipulation”, but as John points out, the Labour Force Survey is also not providing us with what we need to know either. Since having huge budget cuts made, the ABS has changed the method of collecting the monthly data, moving to more online surveys and less face-to-face or phone interviews.

    Basically there are eight groups interviewed every month for eight months to the Labour Force data. Each month, one group drops out (after their eight months is complete) and a new group joins the survey. The make-up of the group dropping out and the new group can lead to anomalies in the data (a particular group may consist of more or less than the average number of unemployed, skewing the data).

    Our old friend Bill Mitchell had a go at explaining this last year: ABS revises labour force methodology – things are much worse than we thought

  15. Kyran

    The International Labour Organisation defined unemployment as those working for pay for less than one hour a week, around about 1982. My recollection (haven’t been able to source articles yet) is that little johnnie adopted that definition as our definition (and instructed ABS accordingly) around about 1998. The reference to Centrelink figures is interesting, as a methodology. Pretty sure little johnnie got rid of the CES, which would have been a far more accurate source of facts than a neutered ABS. As a matter of interest, one of the article’s predates this post and say’s the same thing, in 2012.
    Thank you Mr Kelly. Take care

  16. Wally

    Michael Taylor

    I appreciate the point you make about those who are unemployed but not receiving payments but surely Centrelink can include those they know of who fall into that category in the figures. If you wanted to know how many people were on newstart you wouldn’t do a survey you would ask Centrelink how many people were paid newstart in the last fortnight. Same principle applies they just need to include people in the appropriate categories from their database, the most basic data mining algorithm would spit out the correct figure.

  17. Bacchus

    Except Wally, Centrelink DON’T know “about those who are unemployed but not receiving payments”. How can they? They’re not registered with the Department of Human Services’ systems. And they are actually a significant proportion of the un (or under) employed. The unemployment figures aren’t a measure of those on Newstart – they’re a measure of those who are seeking work no matter their economic circumstance…

  18. Wally


    I am not suggesting that Centrelink can report on people they are not aware of but there are a lot of people registered with Cntrelink that are unemployed but for various reasons do not receive benefits. In some cases they work casual and earn to much to receive benefits in a given period or their partner is employed, others have not been unemployed long enough to qualify for payment.

    There would be very few unemployed people who are not on Centrelinks books in some form and the criteria (working less than 12 hours a week I think?) to be considered unemployed would exclude most who are under employed.

    At the end of the day the public are interested in the most accurate figure of unemployment and using Centrelink numbers must be beneficial in estimating the number.

  19. Bacchus

    I appreciate where you’re coming from Wally – unfortunately the reality is vastly different to your imagination… 😉

    “There would be very few unemployed people who are not on Centrelinks books in some form” – No, there is a very large number of unemployed (and more so underemployed) who have never crossed paths with Centrelink or whose dealings with that organisation are no longer current.

    Centrelink data has its uses (as Michael pointed out earlier), but presenting a reasonable guestimate of unemployment is definitely not one of them – there are far too many outside of its tentacles (as well as many receiving benefits but not technically meeting the ILO definition of being unemployed) for it to be very useful.

    To be classified as unemployed a person needs to meet the following three criteria:
    – not working more than one hour in the reference week;
    – actively looking for work in previous four weeks; and
    – be available to start work in the reference week.

    Someone receiving Newstart who is currently under a medical review does not meet these criteria, but is still receiving the benefit.

    Someone receiving Newstart who is awaiting the outcome of a Disability Support Pension application does not meet these criteria, but is still receiving the benefit.

    Someone undertaking “training” provided by a Job Network provider does not meet these criteria, but is still receiving the benefit.

    Someone undertaking Unpaid Work Experience does not meet these criteria, but is still receiving the benefit.

    Centrelink data is really only useful to tell you how many people are “on the dole”, not how many people are actively seeking employment and available to start work next week…

  20. trishcorry

    I like to study the Job Vacancy data. Mainly for the reason’s Michael has pointed out here.

  21. trishcorry

    and good article John. You were quick off the mark.

  22. Bacchus

    Perhaps I should have just linked to the ABS’ own explanation:

    Comparisons are often made between the official unemployment estimates produced by the ABS and counts of recipients of government job seeker income support produced by the Department of Social Services. While both measures inform policy makers and analysts about people seeking employment, the two measures differ in many ways.

    This article explores the definition, scope and concepts behind the two measures and also examines the key differences between them.


  23. Wally


    It is not what I imagine it is what I know can be achieved from data mining.

    Someone receiving Newstart who is currently under a medical review does not meet these criteria, but is still receiving the benefit.

    Someone receiving Newstart who is awaiting the outcome of a Disability Support Pension application does not meet these criteria, but is still receiving the benefit.

    Someone undertaking “training” provided by a Job Network provider does not meet these criteria, but is still receiving the benefit.

    Someone undertaking Unpaid Work Experience does not meet these criteria, but is still receiving the benefit.

    All of this information is in the Centrelink database so all you need to do is create a query that matches your criteria and click a button for the result to be spat out, it really is that simple.

    Centrelink data is really only useful to tell you how many people are “on the dole”, not how many people are actively seeking employment and available to start work next week

    I can understand why you can believe the above to be true but please accept from someone who has extensive experience designing and developing relational databases that it is incorrect.

  24. Stove-pipe

    Here’s an idea: ask those surveyed “Do you have a full time job?”
    Full time job meaning 38 hours a week, for the same employer, doing the same job.

  25. Bacchus


    You can’t get information out of a database that isn’t in there to begin with. Did you read the link to the ABS?

    Figure 1 shows that while there is a group of individuals who were unemployed and received a government job seeker income support payment in 2011-12, the majority of the unemployed were not receiving NSA or YAO. It also shows that not all people receiving NSA or YAO were classified as unemployed according to the ABS definition. In 2011-12, just over one-third (36%) of these recipients were defined as unemployed, with the remaining recipients classified as employed (26%) or not in the labour force (38%). Furthermore, the SIH estimates that only 30% of all unemployed people were receiving NSA or YAO in 2011-12.

    [emphasis added]

    With up to 70% of unemployed NOT having an entry in the Centrelink database, I don’t care how good you are at designing a RDBMS to store and extract information, if it is not in the database, you can’t extract it with any sort of magical query 😉

  26. Bacchus

    They do Stove-pipe…

    The LFS also collects a range of other information about the population. For employed people, information is collected on hours worked and whether they work full-time or part-time, whether they want and are available to work more hours, and their industry, occupation and status in employment. For people who are currently unemployed, the survey collects information about whether they are looking for a full-time or part-time job, how long they have been looking, and the characteristics of their last job (industry, occupation and reasons for leaving). Estimates of the labour force characteristics of families are also produced from the LFS1.

    The Household form is also a component of the MPS. The Household form collects personal characteristics of each member of the household. These characteristics include sex, age, marital status, relationship in household, participation in school, country of birth, and, where applicable, year of arrival in Australia. Please note the Household form is not included in the attached LFS questionnaire.

    You can see a sample and more detailed information about the LFS here:

    6232.0 – Information Paper: Questionnaires Used in the Labour Force Survey, July 2014

  27. Kaye Lee

    If you look at the confidence intervals for the seasonally adjusted figures from October to November it says that they are 95% sure that

    Total Employment went up somewhere between 13 000 and 129 800

    The change in Total Unemployment was somewhere between a decrease of 39 800 up to an increase of 34 200$File/62020_nov%202015.pdf

    In the three years to December 2014, Australia’s working age population grew by just over a million people which means we need about 28,000 new jobs a month just to keep up with growth let alone reduce the number of unemployed.

  28. paul walter

    That is read that is to know the meaning of the word despair.

  29. Keith Woolsey

    stats, damn stats, a pollie’s best friend to weave as they see fit.

  30. Pingback: What’s happening with the Labour Force? | THE VIEW FROM MY GARDEN

  31. Wally


    You make the assumption that 70% of unemployed people are not on the Centrelink database because of the quote “Furthermore, the SIH estimates that only 30% of all unemployed people were receiving NSA or YAO in 2011-12” and I do not think that assumption is correct. I believe that a much greater percentage of unemployed are included in the database regardless of receiving benefits or not.

    Obviously using Centrelink data will not be 100% accurate but neither is a survey that only includes 0.32% of the population, as Kaye Lee points out the survey aims to be 95% accurate.

    Both methods have margins of error the obvious solution would be to use both and then compare results and quote the mean and median from the combined figures.

  32. Wally


    “I don’t care how good you are at designing a RDBMS to store and extract information,”

    You do not build RDBMS (relational database management systems) for data mining purposes, typically you would execute an SQL statement at the command prompt in a text console and the result would be displayed within the console. The following is a basic example (would bear no resemblance to the Centrelink database fields) of an SQL query that would return the count or sum of records received for a particular customer. The conditions to match with the “WHERE” clause in an SQL query can be extremely complex, multiple or a range of matches possible and multiple conditions can be included on multiple fields using “AND” “OR” clauses.

    SELECT COUNT(CustomerID) AS OrdersFromCustomerID FROM Orders WHERE CustomerID=7

  33. Bacchus

    SELECT * FROM Centrelink_Customers
    WHERE Status=’unemployed’
    AND RegisteredWithCentrelink=’0′;

    Your query has returned 0 results.

    “You do not build RDBMS (relational database management systems) for data mining purposes,”

    Never heard of CRM Wally? 😉

  34. Wally


    I have played with CRM software but why would a software engineer waste money buying expensive software to do what they should be capable of doing for themselves? And using CRM is not building a RDBMS is it.

    The fact that so many companies are marketing CRM software is testament to the benefits of data mining, CRM software provides a lot of capability to less tech savvy end users.

    All bullshit and bravado aside our difference of opinion comes down to if it is reasonable to assume that 70% of unemployed people are not registered with Centrelink. I have not found anything to confirm it but I don’t think you can access full services of the Job Network without being registered with Centrelink, you can use the facilities such as Internet access but would not be assigned to a coordinator.

  35. Bacchus

    Alright Wally – I concede. You obviously know everything about anything worth knowing 😀

    You know that DBAs and systems analysts are never used in the most critical phase of setting up a CRM system – the database design.
    You know that RDBMS are never designed specifically to facilitate analysis of failures in complex systems.
    You know how to write a magical query that will return reasonably accurate data on the employment numbers from the Centrelink database even though up to 70% of the unemployed (we don’t know exactly) may not be captured in that database, and NO information is captured about the working population, especially those employed but who’d like more hours.

    Bravo Wally – you’re obviously a brilliant person 😯

  36. Chris

    Bacchus….a little ‘arrogant’, yes…..

Leave a Reply

Your email address will not be published. Required fields are marked *

The maximum upload file size: 2 MB. You can upload: image, audio, video, document, spreadsheet, interactive, text, archive, code, other. Links to YouTube, Facebook, Twitter and other services inserted in the comment text will be automatically embedded. Drop file here

Return to home page