In a world of uncertainty, there’s one thing markets can rely upon. When it comes to Australia’s jobs report, it’s to expect the unexpected.
The notoriously volatile seasonally-adjusted figures, regularly making a mockery of those brave enough to forecast them, certainly has its critics. Only last year the former head of the ABS stated that recent tweaks to seasonal adjustments within the survey meant that the value of the figures weren’t worth the paper they were printed on.
In an excellent note released yesterday, Tapas Strickland, Skye Masters and Ray Attrill, members of the NAB’s economics team, look at the mechanics of the survey in an attempt to explain the wild monthly swings.
Here’s a snippet from their December jobs report explaining the reason for the ongoing data volatility.
The employment numbers are based on a survey of around 26,000 households a month (or 0.32% of the population). The survey sample rotates every month so that 1/8 of the sample is new each month. Combined with a 94% response rate on average, it means that around 20% of the sample is new each month.
The new parts of the sample can induce considerable volatility in the headline employment growth numbers if those incoming are more active in the labour market than the existing sample – indicated by a higher than average participation rate. This occurred in February, October and November – all months of very strong employment prints of 40k+.
Essentially, when new respondents are more active in the labour market than those they replace, it tends to lead to significant jobs growth being reported for the month. Given the survey is based on just 0.32% of the Australian population, the effects are amplified when applied to the broader Australian populous.
The chart below from the NAB reveals just how large these amplifications can be on the monthly job figures.
After analysing the seasonally adjusted job figures for October and November — reported by the ABS as gains of 56,100 and 71,400 respectively — Strickland, Masters and Attrill came up with an astounding discovery.
Almost all of the reported job growth was entirely due to the survey’s sample rotation.
They explain. Our emphasis in bold.
Sample rotation issues have overstated underlying jobs growth over the past two months. Almost all of November and half of October’s growth were due to sample rotation effects.
The chart below demonstrates the impact that the survey rotation, along with existing survey respondents who failed to respond in prior months, influenced the jobs figures in October and November.
Their contribution to the monthly jobs growth is shown in red.
In what could almost be adopted by the ABS as a disclaimer to accompany its seasonally adjusted data, the NAB suggest that “sample rotation effects should be taken into account when interpreting implausibly large headline employment changes”.
As for December’s jobs report, released on Thursday this week at 11.30AM AEDT, the NAB expect the sample rotation effects of the past two months to diminish, predicting that employment will decline by 25,000 for the month.
In a survey of 23 economists polled by Bloomberg, the median estimate is for employment to decline by 10,000.
Estimates range from a decline of 45,000 to an increase of 15,000, demonstrating the large degree of uncertainty that accompanies the report each and every month.