Wednesday, November 19, 2014

SALM explanation

In a couple of recent posts on the Small Areas Labour Market data I raised some queries on correlating the previous geographical areas with the current SA2 outcomes: Becalmed by Salm; More on SALM. The following response has been received from the Department of Employment:

It is likely that the changes in Central Cairns between the old Small Area Labour Markets (SALM) series, based on the ABS’ 2006 Australian Standard Geographical Classification, and the new SALM series, based on the ABS’ 2011  Australian Statistical Geography Standard, are due to changes in the way the Newstart (NSA) and Youth Allowance (other) (YAo) data (used as weights to apportion SA4 unemployment data between the SA2s within an SA4 – in this case the SA4 of Cairns) are sourced.  Under the old series, NSA and YAo data were sourced by postcode, and then concorded to Statistical Local Areas (SLAs).  Under the new series, the NSA and YAo data have been sourced directly at the SA2 level (this wasn’t possible when the old system was developed, but is possible now).  All of Central Cairns, but also much of the surrounding area (including significant parts of the old SLAs of Cairns (C) – Mt Whitfield, Cairns (C) – City and Cairns (C) – Western Suburbs), was covered by the postcode 4870 under the old SALM system, and differences in the allocation of NSA and YAo data between the concordance (published by the ABS, based on overall population) and the new data available directly at the SA2 level is the main reason for the differences between the new and old regions (it is worth noting that the difference between the old Labour Force Region of Far North, and the new SA4 of Cairns, has also contributed to the changes, although by not as large an amount).
It is not clear as to why the unemployment rates for the SA2s in the old Central Suburbs SLA are higher in SALM than they were at the 2011 Census.  The accuracy of the Census figures can be affected by the nature of the not stated populations in each area and the fact that the method of collecting the Census labour force figures is different from that used in the ABS Labour Force Survey. On the other hand, the SALM estimates are estimates, rather than a Census count or a direct survey (both of which are prohibitively costly to do at a small area level on a quarterly basis), and in a small number of SA2s the methodology used can lead to figures that can over or under-estimate the actual unemployment rate at the SA2 level.
In terms of the participation rate data from the Census, the system used in SALM takes the same approach as the ABS, and excludes the ‘labour force status not stated’ population from the calculation of the participation rate.  These participation rates are then applied to the latest available ABS population data at the SA2 level to calculate weights with which to apportion the SA4 labour force estimate between the SA2s within the SA4.  This provides labour force levels at the SA2 level that are consistent with data published by the ABS by SA4, while also reflecting changes in population growth since the Census.
Sourcing the Centrelink data input directly at the more detailed SA2 level rather than postcode would certainly make a difference in the  unemployment range across the Cairns suburbs. As frequently noted the 4870 postcode is among the largest populations in Queensland with some significant demographic variations. I still suspect census data related to ATSI population and employment may have some influence on the variation between the two series.

1 comment: