What SEIFA measures in Social Atlas, and why it can't be compared over time

SEIFA is the measure councils reach for most often when they need to show relative disadvantage, and it's the one most often quoted in a way it can't support. Two things go wrong repeatedly: comparing a suburb's percentile against an LGA's, and comparing this Census against the last. This article explains what the index measures, why neither comparison works and how to phrase a disadvantage claim so it holds up.

What SEIFA is

SEIFA stands for the Socio-Economic Indexes for Areas. The ABS builds the indexes from a set of Census variables and combines them into a single score for each area. The ABS calculates SEIFA; .id doesn't. There are four SEIFA indexes, but only two are widely used: the Index of Relative Socio-Economic Disadvantage and the Index of Relative Socio-Economic Advantage and Disadvantage.

Community Profile carries the two SEIFA indexes at the LGA and suburb level - ranking your area against every other LGA in the state, and small areas within your LGA against the LGA and benchmarks. Social Atlas maps socio-economic disadvantage across your LGA at the SA1 level.

SEIFA indexes are standardised to have a mean of 1,000 and standard deviation of 100, but are not exactly a normal distribution, with the Index of Disadvantage having a longer tail at the lower end (more disadvantaged). Higher numbers always represent lower disadvantage or higher advantage.

A SEIFA score describes a place, not the people in it

This is the distinction that matters most for how you write about it.

A score describes an area. People live in a disadvantaged area; they are not themselves necessarily disadvantaged. An area with a higher score can still contain households with characteristics of disadvantage, and an area with a lower score can contain households across the whole range.

So 'residents of X are disadvantaged' is a claim SEIFA doesn't support. 'X ranks among the most disadvantaged areas in the state' is a claim it does. The wording matters for how a council phrases disadvantage claims in strategies and grant applications, where a reviewer who knows the measure will notice. Using the correct terminology also avoids stigmatising individuals by the area in which they live.

Why a suburb percentile and an LGA percentile aren't comparable

A SEIFA score on its own doesn't mean much. It's a dimensionless number on a non-ratio scale, so a difference of, say, 10 points at one point on the scale doesn't equal the same difference higher or lower. Its meaning comes from where it sits in a ranked list, and the list is made up of areas of the same type.

An LGA score places the LGA in a percentile of a ranked list of LGAs. A suburb score places the suburb in a percentile of a ranked list of suburbs. However, in Community Profile the small areas list shows suburb percentiles for all small areas, whether they are suburbs, LGAs or other combinations of areas. These percentiles are intended to give an idea of where an area would sit in a ranked list of Australia's approximately 15,000 suburbs if it were a suburb. It's not an exact measure where the areas are not standard geographies, but it can still be very useful.

What to do about it: when stating the percentile for a small area, say that this is approximately where the area would sit on a ranked list of suburbs. If the area is close to a suburb or a combination of two or three, that's a fair comparison. If it's an LGA or wider region, avoid making that comparison and use an LGA percentile instead.

Averaging scores doesn't work either. The next thing many officers try is combining small areas by averaging their SEIFA scores. A raw mean isn't valid - the scores are relative rankings, not quantities you can add up. A combined score for a wider area can be derived using a population-weighted average, which is the approach .id uses where a combined figure is needed.

Why only one Census year is shown

The ABS discourages comparing SEIFA across Censuses, and Social Atlas follows that advice rather than presenting a comparison that can't be relied on.

The components of the index change between Censuses. That means a movement in score may reflect a change in the index rather than a change in the area, and there's no way to separate the two from the published scores. Showing a single Census year is a deliberate choice not to invite an invalid comparison.

This derives from the fact that the technique of Principal Component Analysis uses a set of correlations which are redefined each Census, so the weighting of the input variables changes each Census, and some variables are added while others drop out entirely. For instance, in 2016 the second highest-weighted variable in the Index of Disadvantage was '% of households with no internet access'. In 2021 that was no longer a part of the Census and couldn't be included.

There's a second thing to keep in mind: SEIFA is relative. An area can fall in the rankings while its circumstances improve, because other areas improved faster - and, as above, because the variables behind the index may have changed too. A lower percentile than last Census doesn't necessarily mean an area has gone backwards.

What to do about it: to look at change over time, compare the underlying Census characteristics - incomes, education levels, occupations and the like - which are measured consistently across Censuses and can be compared directly. You can also compare an area's ranking across Censuses to a limited extent, with the caveat that the variables contributing to the index may have changed between the two years.

Where to find SEIFA in Social Atlas and Community Profile

Social Atlas maps socio-economic disadvantage across the LGA at SA1 level, so you can see how it's distributed within suburbs rather than only between them. This is the view that shows pockets of disadvantage inside an otherwise higher socio-economic area, which an LGA or suburb figure averages away.

Community Profile reports SEIFA for the LGA and for small areas, with benchmark comparisons, which is the form that suits a written report.

SA1 is the lowest level SEIFA is available for.

Frequently asked questions

Why is my suburb's percentile different from my council's?

They're ranked against different lists. Your suburb sits in a ranked list of suburbs, your LGA in a ranked list of LGAs. The same score can fall in quite different percentiles on the two lists, and neither is wrong. Quote the one that matches the comparison you're making, and name the comparison group. Note that in Community Profile, on the small areas page, percentiles are shown from the suburb list, even for LGAs and regions, so this may differ from the percentile on an LGA list sourced directly from the ABS.

Why isn't there a 2016 SEIFA on my site?

Because the components of the index change between Censuses, so a comparison between the two years wouldn't mean what it appears to mean. The ABS discourages it, and showing one Census year is deliberate. To track change over time, compare the underlying Census characteristics instead - they are measured consistently across Censuses.

Does a low SEIFA score mean the people who live there are disadvantaged?

No. SEIFA describes areas. People live in a disadvantaged area; they aren't necessarily disadvantaged themselves. Write about areas, not about residents.

Can I average SEIFA scores across several areas?

Not with a simple average. SEIFA scores are relative rankings, so a raw mean of several areas' scores isn't a valid figure. A combined score can be derived using a population-weighted average, which is the approach .id uses where a combined figure is needed.

How is SEIFA calculated?

By principal component analysis across a set of Census variables. The ABS calculates it and .id doesn't adjust it. A range of variables are included in each index, but the major ones are incomes (percentages of high and low), education levels (percentage with university or trade qualifications), housing size (number of bedrooms and motor vehicles) and occupations (managerial and professional versus machinery operators and labourers). The full list of SEIFA variables and weightings for each index is in the 2021 SEIFA Technical Paper on the ABS website.

How detailed does SEIFA go?

SA1 is the lowest level, which is what Social Atlas maps. There is nothing below it. SEIFA scores at the most detailed geographic level vary the most, while the larger an area is the closer to the average (1,000) it will be. SEIFA scores for large areas like states and capital cities are not very useful as they are close to the mean.