From the snappy folks at Westpac:
In regard to sample rotation, this month the ABS notes that most of the change in employment can be explained by the match sample – it is not related to sample roll. But a word of caution, this analysis is done in original terms and Jan is very seasonal reporting a –265.2k decline in total employment in original terms.
Also worth noting that last month the ABS said;
“… looking ahead to the January 2016 estimates, it is important to note that the outgoing rotation group in December 2015, which will be replaced by a new incoming rotation group in January 2016, has a relatively low employment to population ratio (59.7% in December). If the new incoming rotation group has an employment-to-population ratio that is more similar to the average (61.9% for all rotation groups in December 2015), or is relatively higher than the average, there is likely to be some degree of contribution to employment growth from this rotation group change.”
That is, the sample rolling out has a lower than average attachment to the labour force than the population. If the new sample rolling in is more like average, or above average, then the sample roll will make a positive contribution to the survey outcome. It does not appear that this ended up being the case in January.
For what it is worth, this is what the ABS is saying for February;
“… looking ahead to the February 2016 estimates, the outgoing rotation group in January 2016, which will be replaced by a new incoming rotation group in February 2016, had a similar employment to population ratio (60.1% in January) to other rotation groups. In original terms if the incoming group in February 2016 has an employment to population ratio that is equally similar to the average (60.5% for all rotation groups in January 2016), this rotation group change is unlikely to contribute significantly to employment growth.
This will be summarised in the February 2016 issue. The outgoing rotation group in January 2016 also had a relatively high unemployment rate (7.2% in January 2016) compared to the other rotation groups (6.5% for all rotation groups in January 2016). If the incoming group in February 2016 has an unemployment rate that is more similar to the average, or is relatively lower than the average, there may be some degree of contribution to the unemployment rate from this rotation group change.”
In summary, if the incoming sample is more like the average of the total sample then the new sample is unlikely to make a contribution any change in employment but it could lead to fall in the unemployment when rounded to one decimal place. We shall wait and see.
And from a reinvigorated Bloxo:
Recent employment data appear to confirm what we had suspected; the extraordinary strength in jobs growth seen in October and November was most likely overstated. In those two months, employment rose by a seasonally-adjusted total of 133k. Then, over December and January employment dipped by a combined -9k.
Rather than reflecting a stop-start economy, the volatility in the numbers likely reflects statistical noise thrown up by sampling changes and difficulty in determining the seasonal pattern. This has made assessing the true underlying labour market trends very difficult, as we wrote about recently in ‘Is Australia’s jobs growth really that strong?’ (11 February).
We therefore need to take into account a broader range of labour market indicators. Most of these other indicators, which we have combined into a labour market index, tell a similar story; the labour market has been improving over the past year or two, but not to the degree shown by the official figures. Employment growth of 2.6% still looks somewhat overstated, so some further softness might be expected over coming months.
The only surety is Numberwang!
