Notes and caveats on the data used in the analysis. These aren’t written up nicely for sharing or necessarily complete (rather they’re for my own possible reference later) but are provided here in the interest of transparency.
ABS for 2016 Census. Two relevant questions asked are:
- MSTP: all population assigned to either married, widowed, divorced, separated, never married. Didn’t use for main answer as: it doesn’t show if de facto, although some might have been included in married;
- MDCP: registered marriage/de facto/not married. ~11% not assigned (much higher % for elderly). Not assigned aka not classifiable means visitors, incomplete response, not in private dwelling (e.g. hotel, nursing home, prison, hospital), though 11% seems too much for these types. Excludes married but separate houses. Chose this dataset as it excludes smaller amount than MSTP excluding de facto. ”Not assigned’ managed by ignoring i.e. in effect, people in this category proportionally allocated to other categories.
Neither category includes people in a relationship other than de facto or married. For this I used the study done in https://aifs.gov.au/sites/default/files/institute/pubs/fm2011/fm87/fm87.pdf of the 2005 HILDA survey. While 9% of adults are in non-Census relationships, this varies by age, so I didn’t just want to use an average. I didn’t have access to the actual data so I’ve had to estimate the age distribution largely by eye-balling the graphs, though for a big chunk of the age range I was able to calculate the figures based off combinations of the tables in the article and the Census data.
For gender, the information that I can access from HILDA only shows that men are somewhat more likely to be in a non-residential relationship. The information doesn’t split gender variance by age group, but 18-25 63% Male; 45+ 65% Male, and in between 39% male. I’ve worked that in with the census data to arrive at the following estimates. Not perfect but the importance of this type of relationship is strongly overtaken by census type relationships from mid-20s onwards.
The Census data is from 2016, but the data from HILDA is from 2005. That mismatch isn’t ideal, however this article’s intention is more to look at circumstances within a fixed age group, rather than following changes of cohorts as they move through age groups.
For suburbs, I used 2011 Census data as 2016 MDCP by suburb didn’t seem to be available at the time. I don’t have access to HILDA data on a suburb-by-suburb basis, and considering it’s a Australian-wide sample rather than a census it probably wouldn’t be appropriate to use it at such a granular level. Instead I have used the HILDA average of 24% of people not identified by the Census as coupled are in fact in a relationship. Individuals that the Census didn’t place into a relationship category have been excluded from the analysis, and commentary excludes any suburbs with a very small sample size of reported relationships.
For all the data, conclusions will less reliable at the extremes (late teens and from around 75 onward) due to small sample sizes and mismatches in datasets. I’ve excluded some later age ranges from the charts where the results are less robust, and to enhance clarity of the more common age groups.