SEIFA – a geek’s analysis of the statistical analysis

SEIFA – a geek’s analysis of the statistical analysis

As we delve into the new SEIFA data released this week by the ABS, Glenn explains how the different socio-economic variables and indicators are weighted to give us a picture of advantage or disadvantage in Australia.

The SEIFA technical paper is a statistician’s dream bedtime read!

It’s always interesting to see which variables come out as being correlated with advantage or disadvantage in the SEIFA indexes each Census. This can tell us quite a bit about our society in itself – the fact that these variables and the weightings of them are different each time is one of main reasons SEIFA indexes are not considered comparable over time.

In 2016, a total of 16 measures were used to construct the Index of Disadvantage, and 25 were used to construct the Advantage/Disadvantage index. The full list can be found in the SEIFA Technical Paper.

Each indicator is given a weighting score, which indicates how much it contributes to the total index, and how correlated it is with other characteristics indicating advantage or disadvantage. These fall out of the main Principal Component Analysis method.

It’s worth remembering that, while the original “candidate list” of characteristics is hand-picked, which ones end up being used in the index and which ones are dropped is based on a set of rules about the level of correlation they have with each other – so the result is a real measure of our changing society.

In both the Index of Advantage/Disadvantage, and the Index of Disadvantage, Household Income appears as the strongest measure of disadvantage (low income) and advantage (high income). In 2016 these were defined as under $26,000 per annum and over $78,000 per annum respectively. These got a weighting of around 0.8 in the positive or the negative depending on the income level (the highest possible weighting is 1).

Other measures of interest:

  • No motor vehicles in the household – a weak indicator of disadvantage (0.33) – but a large number of cars is not highly correlated with advantage (0.26) and was dropped. This has always been a difficult one – many cars could mean you are well off, or just that you can’t access public transport very easily. The weighting on this variable as an indicator of disadvantage is still in there but has reduced significantly since 2011.
  • The proportion of dwellings owned outright without a mortgage is not considered a strong measure of advantage, whereas it once was. It was dropped from the classification with a weight of only 0.22. This is really significant because you would think that full home ownership at times of increasing house prices is a really strong indicator of wealth. But it may be due to the fact that it is now largely elderly pensioners who fully own their home and they may be asset rich but cash poor, and areas with a lot of these households tend to score highly for other measures of disadvantage.
  • Poor English proficiency has been retained as a weak indicator of disadvantage (-0.3), but dropped from the advantage/disadvantage index because its weighting was too low (-0.22). This probably indicates the rapid multicultural growth of Australian society – in many areas, other languages now have a critical mass and having low English proficiency doesn’t necessarily exclude people from society as much as it used to. Particularly in areas where that language is widely spoken.
  • After having a high income, having a high mortgage repayment (defined as $646 or more per week) had the next highest loading for Advantage (at 0.72). Paying high rent ($470/week) had a much lower correlation at 0.47.
  • A lack of Internet access is now a very highly weighted indicator of disadvantage (-0.78). This indicates that access to information is now a basic resource needed to function in society.
  • After this, one of the more highly correlated indicators (-0.76) was households with children who had neither parent with a job. This has often been mentioned as a strong indicator of entrenched, perhaps generational disadvantage.

It’s the currency of these indicators which makes SEIFA such a strong indicator of a range of socio-economic issues beyond what is included in the Census itself, and so it’s always worth a look to see what actually makes it up.

We are currently working to update our online tools with the new SEIFA data – bookmark our updates page to keep up to date with the latest releases.

.id is a team of population experts who combine online tools and consulting services to help local governments and organisations decide where and when to locate their facilities and services, to meet the needs of changing populations. Access our local government area information tools here.

Glenn - The Census Expert

Glenn is our resident Census expert. After ten years working at the ABS, Glenn's deep knowledge of the Census has been a crucial input in the development of our community profiles. These tools help everyday people uncover the rich and important stories about our communities that are often hidden deep in the Census data. Glenn is also our most prolific blogger - if you're reading this, you've just finished reading one of his blogs. Take a quick look at the front page of our blog and you'll no doubt find more of Glenn's latest work. As a client manager, Glenn travels the country giving sought-after briefings to councils and communities (these are also great opportunities for Glenn to tend to his rankings in Geolocation games such as Munzee and Geocaching).

Leave a Reply