Unemployment gulf revisited

By Leith van Onselen

Earlier this month, Australia’s unofficial provider of labour force data, Roy Morgan Research, released its employment figures for March, whereby it estimated that 9.3% of Australians were unemployed, down -0.4% from February 2012.

As explained previously, Roy Morgan Research measures employment differently from the Australian Bureau of Statistics (ABS), which is Australia’s official provider of labour force data:

According to the ABS definition, a person who has worked for one hour or more for payment or someone who has worked without pay in a family business, is considered employed regardless of whether they consider themselves employed or not.

The ABS definition also details that if a respondent is not actively looking for work (ie: applying for work, answering job advertisements, being registered with Centre-link or tendering for work), they are not considered to be unemployed.

The Roy Morgan survey, in contrast, defines any respondent who is not employed full or part-time and who is looking for paid employment as being unemployed…

Since Roy Morgan uses a broader definition of unemployment than the ABS, it necessarily reports a higher unemployment figure. In addition, Roy Morgan’s measure tends to be far more volatile, owing to the fact that it draws on a smaller sample than the ABS and is not seasonally adjusted.

Following the release of the ABS labour force data for January, which registered a surprise fall in Australia’s unemployment rate from 5.2% to 5.1%, the head of Roy Morgan Research, Gary Morgan, claimed that the ABS unemployment estimate “defy belief and common sense” and suggested that “the ABS figures are diverging from reality” in light of the “constant stream of companies announcing retrenchments in the early stages of 2012″.

Gary Morgan also argued that the ABS measure is a lagging indicator as it “only classifies an unemployed person as part of the labour force if, when surveyed, they have been actively looking for work in the four weeks up to the end of the reference week“, whereas the Roy Morgan estimate “provides a more accurate picture of what is happening on the ground” since its “estimates are taken at that point in time“.

In light of the release of the ABS March labour force data last week, which produced no change in the official unemployment rate (5.2% seasonally adjusted) despite the creation of 44,000 jobs, I considered it worthwhile to once again compare and contrast the ABS unemployment estimate with Roy Morgan’s March release.

The below chart plots the ABS seasonally adjusted (SA) and non-seasonally adjusted (NSA) unemployment rate series against Roy Morgan’s NSA series:

Focusing on the non-seasonally adjusted (NSA) series only, you can see that there was some convergence in the Roy Morgan and ABS measures in March, although they remain far apart, historically so. In January, the Roy Morgan estimate was 5.2% higher than the official ABS measure – the highest divergence in the series’ 11-year history. However, in March, the difference fell to 4.1%, which is still way above the average divergence of 2.2% over the life of the series (see below chart).

When both series are charted on a 3-month moving average (3MMA) basis (in order to reduce volatility), you can see that they tend to be highly correlated – it’s just that something weird has happened recently that has pushed the two series wide apart:

I noted previously that it remains to be seen whether Roy Morgan’s estimate is right or wrong with respect to the direction of the labour market. That is, either Roy Morgan’s unemployment rate will retrace sharply in coming months (beyond usual seasonal movements), the ABS figure will deteriorate toward Roy Morgan’s, or some combination of the two.

Given the strong ABS employment figures for March, Roy Morgan’s survey at this stage looks to be the outlier. However, with the ongoing weakness in the retail sector, the number of dwelling approvals contracting, and the rate of credit issuance remaining highly subdued, it is still possible to see the two indexes as intact. The different methodology of the Roy Morgan index is more sensitive to underemployment and may be more strongly registering the churn we are seeing in the labour market between sectors. I still expect to see the two converge at a higher official rate in the coming months.

[email protected]


Unconventional Economist


  1. “The Roy Morgan survey, in contrast, defines any respondent who is not employed full or part-time and who is looking for paid employment as being unemployed…”

    I think that is the key statement.

    As more people become under-employed they are counted in the RM survey but not in the ABS figures. If it was just the four week lag in being ‘unemployed’ according to the ABS definition we would have seen more of a convergence by now.

  2. What’s the margin of error and confidence interval of the Roy Morgan estimate?

    For the ABS, there is a 95% chance that the true value of the estimate lies within +/- 0.2%

    • dumb_non_economist

      OMG, I face having a million and one people jump on me for this, but what if the explanation for the divergence between Roy Morgan and the ABS lies in the other 5% and therefor the true value is outside +/- 0.2%? I realise the odds are against that, but you can hardly say these are normal times!

      • Yes, the ABS has a slim probability of being wrong by, let’s say, 0.4% to the down side. But it’s several months now that they are measuring unemployment at around 5.2%, so that probability decreases.

        The ABS participation rate has been falling recently and fast (not last month) that could explain the unusual large divergence between the two measures.

        The ABS always uses the same methodology so you can compare their data to previous years.

        The difference between the 2 sets is now 4.1% which is higher than average, but seems to be coming back to being withing the usual range.

        I think it would help if it was known what the margin of error and confidence of the Roy Morgan measure is, being a survey on a sample.

  3. Under-employment is the major issue here. You can work one hour a week and still be counted as employed; however common sense would tell you that you can’t afford your rent or electricity on one hour per week. Given the extreme casualisation of the workforce that’s occurred in the last ten years I’d say there are a lot of people in that “can’t make ends meet” category.

    • In retail, I’m not allowed to employ someone for one hour. The minimum I am allowed to employ someone is THREE hours in one shift, except for a school-age child, whom I’m allowed to employ for 1.5 hours minimum after school ……….. or so I thought, please correct me if I am wrong.

  4. Would it be possible or practical to either count, the number of CentreLink benefits titled JobStart or NewStart claimed, or the number of people registered or resigerting with a JobService Provider in order to obtain these benefits, to get somekind of idea of what unemployment may be?

    Hawke-Keating in their early days of Government utilised (I think it was called back then the CES) CES data to see what unemployment was. They then adopted this “Labor Force Survey” for their measure. I guess when they claim their policies reduced unemployment they weren’t kidding. I jest and digress.

    Why just look at the ABS, Roy Morgan, ABS vs Roy Morgan, ABS this, Roy Morgan that, analysis for unemployment? Why doesn’t some one count CentreLink claims as a basis for their statistical analysis?

    I don’t see ABS and Roy Morgan unemployment analysis, it is more like Roy Morgan is a scalar multiple of the ABS, that’s what those graphs show.

  5. Think we can confidently take a stab in between, i.e. unemployment and underemployment are significant. There is also another useful index, i.e. the Clarius Skills Index which is more general but measures demand for employees in particular occupational areas each quarter, generally quite interesting read and shows how out of touch pollies and media are…….

  6. According to the ABS definition

    Which is actually the ILO definition, adopted by international agreement in 1982.

    The ABS Labour Force Survey also looks at underemployment – current trend underemployment is 7.6%, see Table 22 or 23 of the Labour Force stats.

  7. Government policies that ‘warehouse’ young people in school ie the 17 year of age minimum school leaving age (federal policy) and an increase in the number of Uni and TAFE places available. I believe this was around 50K this year alone.

    Zoologists and personal trainers abound.

    Kick that ‘unemployment’ can … one more time.

    • Next up lots of training courses catering for the older unemployed as well? Have seen that done (feeling old and ancient).
      It’s great if people can retrain and seek new careers but those sort of issues can explain in part why unemployment figures don’t always seem to reflect the reality.