Gross flows in the Australian labour market

The prior two months I have written about gross flows in relation to the state of the Aussie labour market, but have been too busy to provide supporting detail.

In this post, I aim to give a bit of that additional detail. There is a lot of ‘non-standard’ information that’s available in the ABS’s Labour Force survey (Catalogue 6202), and I only just skim the surface.

First of all, some background detail. The labour force survey is a survey of households. The survey object is a physical address; and once an address is added to the sample, it remains in the sample for the following seven months (so the address is surveyed once per month for eight months). If folks move in or move out of an address that’s in the survey frame, they drop into or out of the survey.

The repeated sampling of the same address means that we can ‘match’ a responding person from month to month in many cases. The fact that one eighth of the sample is dropped each month, and a new eighth is added, means that as a maximum about 7/8^th (or 87.5%) of the sample can be matched. In practice, that number is ~80% – as folks move house, go on holiday, go to hospital etc.

This data set is useful by itself, as it allows one to track the transition of units between labour market states; it is also useful to compare the matched estimates and the unmatched estimates.

The latter is the simpler case (and also makes clear a potential issue with the gross flows data) so I’ll deal with that case first of all, and then look at the gross flows data. The main thing to be aware of is that the unmatched sample has different properties than the sample (that is, folks are not ‘unmatchable’ at random), and doing things in this order properly highlights that fact first up.

When you think about it, it makes sense that the ‘matchable’ and ‘unmatchable’ are different. My friends with secure jobs tend to own their own home and to stay put from month to month. My friends with a more marginal attachment to the labour force tend to both move about a lot and also tend to more frequent unemployment. Sometimes they move from job to job, and other times they move from work to holiday / unemployment.

Thus, the matched sample unemployment rate is persistently lower than the full sample unemployment rate. You can see this in the above chart (note, I seasonally adjusted the full sample unemployment rate myself, rather than use the ABS estimate, as I didn’t want the different adjustment method to contaminate the spread estimate, which I show below).

The job prospects of the unmatchable are pro-cyclical, so the spread between the matched sample and full sample unemployment rate also tends to be pro-cyclical. You can see in the above chart that the recent increase in the unemployment rate coincides with a widening of the difference between the matched and full sample unemployment rates.

It makes intuitive sense that increases in the unemployment rate tend to coincide with a souring of the employment prospects of the marginally attached. Microeconomic theory supports this conclusion. Marginally attached workers ought to be the cheapest workers to shed — often simply by not renewing a contract.

Also, while the data isn’t sufficient to be sure of it, it seems probable that highly cyclical sectors would prefer these employment arrangements — which means that the sectors that employ such workers may also be the ones that are most vulnerable to a downturn.

The gross flows data support the contention that the supply-demand balance in the labour market has moved from ‘better bid’ to ‘better offered’. However, it’s a subtle shift so far, and not a melt-down.

The above chart shows the hazard rate – the probability that an employed worker will become unemployed. It appears to have bottomed out some time at the start of this year at just below 0.8% (that is 0.8% of employed workers lose their job each month), and is now probably trending up modestly (in August, it fell ~0.04% to 0.845%).

The finder-rate, shown above, is the probability that a previously unemployed person will find a job in the next period. This measure had showed particular weakness in June and July, but improved markedly in August (+3.5pts to 23.8%). The recent volatility is a worry, but if the current level holds, I would say that things are sound.

It is this measure that collapsed most markedly in the post LEH period, and it’s worth noting that it fell with a lag – so we are not out of the woods just yet. A part of that lag was due to the maths of an increase in the number of unemployed, which pushed down the finder rate, for a given (person) outflow from unemployment to work.

Transitions from part time to full time work also suggest weakness in the labour market. The probability of a matched unit moving from part time to full time employment declined from ~13% earlier this year to a low of just under 11% in July. There was improvement in August (to ~12%), and consequently the trend has turned up again – however it’s a volatile estimate, and there was a similar dive-and-bounce in late 2008, so I reserve judgement at this point.

On the basis of the above, I think the Australian labour market has been softening for a few months. The unmatched (or marginally attached) appear to have been losing their jobs at a faster pace, the hazard rate appears to have troughed and has risen a little, and the ease with which one might transition from unemployment to employment has declined somewhat. Also, it’s getting harder for part timers to graduate to full time work.

At this point, these developments merely confirm that things have softened. If these trends are sustained, the unemployment rate will continue to drift up, and at some point the RBA will reconsider their 4.75% policy rate.

The market is priced for this to occur in October; however that seems too soon to me – unless there is a (bigger) financial crisis to come out of Europe. The Bank’s problem in the Q3 SOMP was that they forecast too strong growth, and too high inflation.

It will take a few quarters of evidence to change this – I’m guessing a least two quarters. It simply takes time to turn a 3.25% inflation forecast into a 2.5% inflation forecast (which is what’s required to get the bank to ease monetary policy).

So, is the RBA’s next move a cut? I don’t know … what I do know is that the market has basically priced in the worst (a crash / crisis), so there’s probably money in the Aussie short end.