An analysis of the SA4 level of Australia's new statistical geography

This paper analyses the new SA4 regionalisation, which is an aggregation of SA2s and was first implemented by the Australian Bureau of Statistics at the 2011 Census. The paper uses two different algorithms to produce alternative aggregations of SA2s using commuting flows reported in the Census. We then employ fuzzy set principles to fine-tune all three aggregations, firstly based on the proportion of flows that are self-contained within regions, and secondly based on the proportion of interaction that occurs throughout the regions. Finally we test the various resulting aggregations against each other using a number of summary statistics, some of which are based on fuzzy set principles. It is found that the alternative regionalisations perform slightly better than the SA4 regionalisation and that fine-tuning does indeed slightly improve all groupings.