When Numbers Start Making Decisions · Season Three, “The People the Average Does Not See” · Article 1
1. Does moving house make the same student a different person?
Imagine two university students whose family incomes, parents’ education, rent, transport costs and study pressures are almost identical. One lives in a statistical area classified as low socioeconomic status. The other lives across an administrative boundary in an area ranked just above the cut-off. If the system can observe only their addresses, the first may be counted in a university’s low-SES student population while the second is not.
Australia’s demand-driven Needs-based Funding, introduced from 2026, has to make classifications of this kind. Its guidance uses the Australian Bureau of Statistics’ Socio-Economic Indexes for Areas, or SEIFA, and specifically the Index of Education and Occupation. A student’s first reported residential address is coded to Statistical Area Level 1. SA1s covering the lowest 25 per cent of Australia’s population aged 15 to 64 are treated as low SES. Where an address cannot be coded accurately to an SA1, the first reported postcode may be used instead. Department of Education: Needs-based Funding Guidance
The classification contributes to the funding received by the university. It does not declare that an individual student is poor, and it does not automatically give that student a fixed payment. Universities can use the additional funding for measures such as scholarships, emergency assistance, mentoring and academic support. Department of Education: Demand-driven Needs-based Funding
The important question is therefore not whether SEIFA is useful. It is when a number describing a place may stand in for a person who lives there and then help distribute resources.
2. SEIFA describes the composition of an area, not the life of an individual
SEIFA is a family of four indexes. The Index of Relative Socio-economic Disadvantage focuses on disadvantage; the Index of Relative Socio-economic Advantage and Disadvantage covers both ends; the Index of Economic Resources concentrates on income and housing resources; and the Index of Education and Occupation measures the educational and occupational composition of an area.
The indexes combine census variables at SA1 level and standardise the resulting scores. A national mean of 1,000 and a standard deviation of 100 make areas comparable, after which they can be arranged into ranks, percentiles and deciles. The score is a relative position, not a quantity of disadvantage. An area with a score of 1,000 is not twice as advantaged as one with a score of 500. ABS: SEIFA 2021 Methodology
Most importantly, the ABS assigns SEIFA to areas, not people. A low-ranked area contains some highly educated and well-paid residents; a high-ranked area contains unemployed people, low-income tenants and families experiencing sudden hardship. Inferring the circumstances of every resident from an area score is the ecological fallacy in administrative form.
Area indexes nevertheless have genuine value. Government cannot repeatedly investigate every household’s assets, debts, caring responsibilities and recent shocks for every funding calculation. Address data are relatively stable and inexpensive to administer. They can also identify concentrations of limited opportunity that should be answered through place-based outreach. SEIFA measures a structural environment: the opportunities and resources likely to surround a person, not the entirety of that person’s life.
3. The bottom quarter is an administrative choice, not a natural boundary
SEIFA produces continuous scores and ranks. It does not itself announce that a person at the 24.9th percentile deserves support while one at the 25.1st does not. Needs-based Funding turns the lowest quarter into a category because a national program needs a defined population, a calculable budget and a reproducible rule.
Two compressions occur. Census variables about education and occupation become one area index; the continuous index then becomes a binary low-SES or non-low-SES classification. Each compression improves administrative usability and loses information.
SA1s immediately either side of the line may be nearly indistinguishable in transport, rent and educational opportunity. Yet the students attached to them can contribute differently to a university’s funding calculation. This does not prove that the 25 per cent threshold is irrational. Limited programs require scope. It does mean that the boundary must be justified by the purpose.
If funding is intended for outreach into underrepresented places, an area measure may be exactly the right level. If the decision concerns whether one student receives emergency financial relief, address alone is plainly insufficient. A measure suited to allocating resources among institutions should not automatically acquire authority to refuse an individual.
4. An area average cannot see internal variation
Even a small SA1 contains people with different incomes, tenures, family structures and needs. Mixed-income suburbs, student districts, new developments and rapidly gentrifying neighbourhoods make the mismatch especially visible. At larger geographies, population-weighted aggregation conceals still more variation.
Addresses change too. A student may move for study, alternate between a parent’s home and a rental, or experience unstable accommodation. Using the first reported address gives the formula a fixed reference point, but it can preserve a situation that is no longer current. Falling back to a postcode increases the area and therefore the heterogeneity.
The IEO can see only the education and occupation variables available in census data. It cannot directly see a bank balance, debt, disability-related costs, family violence, caring duties, housing insecurity or a recent redundancy. Two areas can arrive at the same score through very different structures.
Data accuracy and measurement validity must therefore be separated. A student’s address may be recorded perfectly and geocoded without error, while the inference about how much support that student needs remains wrong. The failure lies not in the input but in treating an environmental proxy as a personal attribute.
5. Review has to exist at three levels
Authority is distributed along a chain. The ABS chooses variables and calculates area rankings. The Department of Education chooses the IEO, the population and the bottom-quarter rule. Universities report addresses and receive funding, then decide how support reaches students. No participant in the chain sees the whole picture.
A student can generally seek correction of inaccurate personal information, but correcting an address is not the same as contesting an SA1’s SEIFA ranking. The latter is a population result generated by a national method. Proving personal hardship may not change how the university is counted; living in a low-SES area does not guarantee receipt of a particular service.
Review should therefore be divided into three layers. Data review asks whether the address, identity and enrolment are correct. Classification review checks the SEIFA version, geographic coding and quartile. Needs review allows a person excluded by the proxy to establish hardship through individual evidence such as changed income, unstable housing or caring burdens.
The first two layers make the formula accurate. The third prevents an accurate proxy from leaving a real person outside support. Regulators must also ask whether institutional funding actually benefits the intended students. A correct allocation is not proof of a delivered outcome.
6. Why institutions need a number that is not the person
Socioeconomic position is produced through income, occupation, housing, services, family relationships and change over time. It contains no naturally occurring line. To calculate, compare and audit national funding, institutions impose an area, an index and a quartile. This is a forced structure: a necessary simplification that should remain visibly incomplete.
Its legitimacy depends on consequences. A rough geographic proxy is more defensible when it opens additional institutional resources than when it closes a person’s access. It is more defensible when universities can use other evidence to help students than when the classification becomes conclusive. It is more defensible when versions, coding failures and expenditure are auditable than when “the data” appears as an unexplained verdict.
The asymmetry matters. False inclusion in an outreach pool may direct some support to a student whose area overstates disadvantage. False exclusion can leave an equally needy student invisible. When individual support is at stake, the second error deserves a route around the proxy.
7. A practical test for area-based decisions
Before an area index is used to influence an individual or institutional consequence, ask:
- Does the decision concern a place, an institution or a person?
- Which SEIFA index and version are being used, and why is that index relevant?
- How much variation exists inside the selected geography?
- What happens near the percentile boundary?
- What important forms of hardship are absent from the variables?
- Can inaccurate source data and geographic coding be corrected?
- Can individual evidence override the area proxy where the decision is personal?
- Is the funding’s real distribution audited after the formula is applied?
The ABS itself advises users to interpret SEIFA at area level and to avoid assigning area characteristics to individuals. ABS: Using and Interpreting SEIFA That warning should not be treated as a footnote. It is a limit on the authority of the result.
Conclusion: a map may guide resources without defining a person
SEIFA allows Australia to see patterns that individual anecdotes cannot reveal. It can show where educational and occupational disadvantage is concentrated and make national support more systematic. Those are substantial achievements.
But an area ranking does not turn its residents into an average. My judgement is that SEIFA is legitimate for distributing institutional and place-based resources only when its level is made explicit, its errors are auditable and personal support remains open to personal evidence. The address may be evidence about a student’s environment. It is not a complete biography.
The first lesson of Season Three is simple: the authority of a group measure should weaken as a decision moves from the group towards the individual.
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.