A Mathematics of Immigration

In December 2021 the Australian Treasury published a modelling paper that almost nobody read properly. It used a tool called FIONA — the Fiscal Impact of New Australians model — to estimate what a year’s intake of permanent migrants would contribute to, or take from, government budgets over the rest of their lives.

The findings got quoted widely. Skilled migrants: plus $198,000 each in net present value. Family stream: minus $126,000. Humanitarian stream: minus $400,000. Those figures have been used ever since to argue that Australia should tilt harder toward skilled migration, and they are real numbers from a serious model.

The figure almost nobody quotes is the one Treasury put alongside them as a benchmark. The average Australian-born citizen: minus $85,000.

Not because Australians are a drain on anything. Because that is simply what a full human lifetime costs a government when you count childhood schooling, healthcare, aged care and pensions against a working life of tax. Over a whole life, most people in a developed welfare state run a fiscal deficit. The taxes are paid by a minority of very high earners and by the fact that the economy grows.

Once you know that, every number in this debate changes meaning. A family-stream migrant at minus $126,000 is not a drain compared to a citizen — they are about $41,000 worse than one. A skilled migrant at plus $198,000 is not $198,000 better than the alternative; relative to the native-born benchmark they are roughly $283,000 better. And the humanitarian stream, at minus $400,000, is genuinely expensive on this measure, which is worth saying plainly rather than omitting.

Almost every public claim about the fiscal impact of immigration quietly compares migrants against zero. The correct comparison is against a citizen, and citizens do not score zero.

This article is about what the immigration economics literature actually shows, where it is contested, and what the fiscal frame leaves out. It takes seriously that reasonable people disagree here — and that on several of the most-quoted numbers, the experts are not disagreeing politely but publicly accusing each other of error.

Why the accounting method decides the answer

Almost all of the disagreement in this field comes from methodology rather than data, so it is worth understanding three choices that drive the results.

Snapshot versus lifetime. A point-in-time budget snapshot counts what a group pays and receives this year. It will always make recent arrivals look expensive, because settlement support and schooling for their children land early while peak tax payments arrive fifteen years later. Generational accounting instead discounts a whole life of taxes and benefits to present value. It is the better method, but it requires forecasting wages, employment and policy decades ahead, and small changes in those assumptions move the answer enormously.

Which government. Australian federal and state finances are modelled together in FIONA. Most American work is federal-only. That matters because the costs of immigration — schools, emergency healthcare, local roads — fall disproportionately on states and municipalities, while the income tax flows to the centre. A federal-only study will therefore look more positive than a consolidated one. This is a real structural mismatch, not an accounting artefact.

How you treat public goods and descendants. Do you charge each migrant a share of defence spending and debt interest, which they do not increase by arriving? Do you count their children, who in most datasets are strong net contributors? Including the second generation makes almost any immigration programme look good. Excluding it makes most look worse. Neither choice is obviously correct, and studies that reach opposite conclusions usually differ here rather than on the underlying data.

The Australian evidence

Treasury’s FIONA model remains the most careful public estimate available for any country, precisely because it consolidates Commonwealth and state budgets. Its headline result is that the 2018–19 permanent migrant cohort was fiscally positive at both levels, and about $127,000 per person better than the population overall.

Australian Treasury, FIONA model, 2018–19 permanent migrant cohort, published December 2021. The right-hand column is the comparison that matters and is usually omitted.

Three cautions belong with this table. It models a single pre-pandemic cohort, and both the migration programme and the economy have changed since. It measures fiscal flows to government, not economic contribution, which is a different and larger thing. And it says nothing about whether a humanitarian intake is worth its cost — that is a question about legal obligations and moral commitments, not one a fiscal model can answer. A government that admits refugees is not making a failed investment; it is doing something it did not undertake for fiscal reasons in the first place.

The American numbers, and the fight about them

The US evidence is where care is most needed, because the two most-cited recent estimates are in open conflict and are routinely quoted together as though they agreed.

Both build on the National Academies of Sciences 2016 study, which remains the reference work. The NAS itself did not produce one number; it produced a range of scenarios, and which scenario you pick decides your conclusion. Read as the net future value of immigrants already present under a worst-case fiscal path, it gives a negative figure. Read as the lifetime impact of immigrants from arrival onward, its own tables give net benefits of roughly $92,000 to $186,000 per person. Advocates on both sides quote the NAS, honestly, and reach opposite conclusions.

Daniel Di Martino at the Manhattan Institute extended the NAS framework with narrower age bands, a federal-only scope, and the exclusion of immigrants’ US-born descendants. He finds the average new immigrant modestly positive — around $10,000 in present value — with enormous variation underneath: an immigrant arriving before 35 with a graduate degree reduces the federal deficit by more than $1 million, while immigrants without college degrees are net burdens of up to about $400,000. Notably, his own framework puts the average native-born American at roughly minus $250,000, the same structural point the Australian data makes.

David Bier at the Cato Institute argues the Manhattan model contains outright errors. His central technical claim is that it conflates income tax liability with refundable tax credits, growing both in a way that overstates the cost of low-income immigrants. Correcting that alone, he calculates, flips whole categories: the average immigrant without a high school diploma arriving between 18 and 24 moves from about minus $315,000 to plus $45,000. That is not a nuance. It reverses the sign on the most politically contested group in the entire debate.

The original version of this argument cited Cato and the Manhattan Institute in the same sentence as joint evidence. They are not allies here. They are having a public fight about whether one of them has made an arithmetic error.

A reader is entitled to know that this is unresolved. The honest summary is narrower than either camp’s: young, highly educated immigrants are strongly fiscally positive under every model anyone has published; older arrivals are negative under every model; and for working-age immigrants without tertiary education, the answer depends on technical choices that specialists are currently disputing in public. Anyone who tells you that group is definitively a burden, or definitively a benefit, is selecting a study rather than reporting a finding.

It is also worth noting the priors in the room. The Manhattan Institute and the Cato Institute are both free-market think tanks, but with different commitments — Cato is strongly pro-immigration on libertarian grounds, Manhattan more restrictionist on fiscal ones. Neither is a neutral referee, and the NAS, for all its ambiguity, is the closer thing to one.

What the fiscal frame leaves out

There is a deeper issue that the numbers above cannot reach, and it deserves stating before the policy discussion rather than after.

A lifetime fiscal impact model answers one question: what does this person do to the government’s budget? That is a legitimate question, and taxpayers are entitled to ask it. But it is not the same as asking what a person contributes to a society, and treating it as the objective function has three consequences worth being explicit about.

●        It excludes the migrant’s own welfare entirely. A framework that maximises net fiscal return to the destination country treats the person as an input. That is a defensible way to design a migration programme — states do act in the interests of existing citizens — but it is a political choice, not a neutral technique, and it should be argued for rather than assumed.

●        It scores care work, parenting and community contribution at zero or negative. On this measure a grandparent who enables two parents to work full-time is a cost. So is a nurse, relative to a derivatives trader. The model is not wrong; it is measuring one thing.

●        It is a measure of tax systems as much as of people. A country with a flatter tax structure and thinner benefits will make every migrant look more fiscally positive without anyone’s contribution changing at all.

None of this is an argument against fiscal analysis. It is an argument for labelling it — and for noticing that the same analysis applied to citizens produces minus $85,000, which nobody proposes to act on.

Two systems: selection versus the market

Developed economies sit on a spectrum between state-run points selection and employer-driven market allocation, and the comparison is genuinely instructive.

Australia and Canada screen for age, qualifications, language and occupation, and Australia allocates roughly 70% of permanent places to the skill stream. This produces an unusually educated migrant population: close to six in ten Australian migrants hold a university degree or higher, against roughly four in ten of the native-born — the largest such gap in the OECD.

The United States screens far less. Employment-based preferences account for a minority of green cards, with most allocated through family ties and humanitarian pathways, and high-skilled work visas are capped well below employer demand. The system absorbs a wider range of skill profiles, produces exceptional outcomes at the top — immigrant-founded firms, patents, research output — and, because legal economic channels are capped below demand, has historically generated large unauthorised inflows into low-wage sectors, with costs concentrated at state and local level.

The OECD’s recent Australian work is the strongest evidence for the selection model, and it should be quoted accurately, because it is regularly overstated. The findings are that a region with a 10% larger migrant share has on average a 1.3% larger regional wage differential, which the OECD reads as indicating higher labour productivity; that a one percentage point rise in the employment share of higher-educated migrants raises regional patent applications by 4.8% over five years; and that a one percentage point rise in migrant inflow raises native-born employment by 0.53%.

Two qualifications matter. The productivity result is a comparison across regions, not a measured change over time, so the causal reading is an inference. And the OECD’s own finding on wages is that migration boosts native employment and does not affect native wages — which is good news, but not the same as the claim, often made, that migration raises native wages.

The same body of work contains a finding rarely quoted by advocates: despite their higher education, migrants in Australia have lower employment rates than the native-born, driven mainly by low labour force participation among migrant women. A selection system that admits highly qualified people and then fails to employ them is buying human capital and leaving it on the shelf. That, rather than intake volume, is arguably the largest efficiency loss in the Australian programme.

The human capital transfer

When a developed economy admits a foreign-trained engineer or physician at 25, it acquires two decades of education that someone else paid for. The origin country funded the schooling; the destination captures the tax.

This is a real transfer and a large one. Estimates of the cost of raising and educating a person to tertiary qualification vary enormously by country and by what is counted — public spending alone, or public plus household — so a single global figure is not available and should not be invented. What can be said is that the recipient country avoids roughly twenty years of net public outlay per skilled arrival, and that this is a substantial part of why skilled migration scores so well fiscally. It is not productivity magic. It is a subsidy, paid by a poorer country.

The brain drain literature has moved on from treating this as straightforwardly harmful, and the offsets are well documented: remittances that frequently exceed aid flows, diaspora effects on trade and investment, return migration carrying skills back, and the incentive effect whereby the prospect of emigration raises domestic investment in education, leaving more educated people behind than would otherwise exist.

But the offsets are not uniform, and the strongest counter-case deserves its due. For small states and for specific professions — physicians and nurses from sub-Saharan Africa and the Pacific being the clearest cases — the losses are concentrated and severe in ways remittances do not repair. The WHO maintains a code of practice on international health-worker recruitment for exactly this reason. A wealthy country solving its aged-care staffing problem by recruiting nurses from a country with a worse nurse-to-population ratio is doing something that a purely national fiscal calculation will score as a triumph.

Housing, infrastructure and the timing problem

The most politically potent objection to immigration in Australia and Canada is housing, and it needs handling carefully because the evidence is weaker than the rhetoric on both sides.

The mechanism is not in dispute. New arrivals add housing demand immediately. Supply responds slowly, because construction runs through zoning approvals, finance, materials and labour. In the short run that pushes rents up and advantages existing owners over renters and new arrivals alike.

The magnitude is very much in dispute. The Reserve Bank of Australia’s own estimate is that a population increase of 50,000 might raise private rents by around 0.5% relative to baseline, and it concluded that the surge in international students accounted for only a small share of post-pandemic rent increases. The RBA has also noted that the supply and demand effects of stronger-than-expected population growth appeared roughly to offset in aggregate inflation.

Meanwhile the supply-side evidence is strong. Work by Kendall and Tulip at the RBA attributes a large share of Australian housing costs to zoning and planning restrictions that hold prices well above the physical cost of supply. The National Housing Supply and Affordability Council projects completions of roughly 938,000 homes over five years against a 1.2 million target — a shortfall that exists independently of migration settings.

And there is a feedback loop that demand-side arguments tend to miss: migrants build houses. Cutting skilled migration reduces the construction workforce, which reduces completions, which worsens the shortage the cut was meant to fix. Several analysts argue that sharp migration cuts could raise Australian house prices for this reason. The honest position is that migration contributes to short-run rental pressure, that supply constraints dominate the long run, and that the two interact in ways that make simple volume cuts an unreliable instrument.

Can this be modelled?

It is tempting to write the whole problem as an optimisation — choose the annual intake that maximises welfare, subject to housing and infrastructure constraints. Something like:

maxₖ  W = Σₜ δᵗ [ Y(Mₜ, Hₜ) + F(Mₜ, Hₜ) − C_housing(Mₜ, ΔSₜ) − C_infra(Mₜ, Kₜ) − C_integ(Mₜ, σₜ) ]

where M is intake, H the human capital of entrants, ΔS housing completions, K infrastructure capacity, and σ an index of integration support. Differentiate with respect to M and you get the intuitive result: admit more when human capital and productivity spillovers are high, admit less when housing supply stalls or integration capacity is underfunded.

It is worth being candid about what this buys. It is a way of stating the trade-off clearly. It is not a model anybody can estimate: the functional forms are unspecified, the parameters have not been measured, and the choice of what enters W — particularly the exclusion of migrant welfare — is the political argument rather than a technical input to it. Writing an equation around a disagreement does not resolve it, and false precision is a real hazard in this field, where confident numbers already do more work than they can bear.

The genuinely useful output of this framing is not an optimal number. It is the observation that intake and absorptive capacity are separate policy levers that are usually discussed as one. Most of the political heat is about M. Most of the available improvement is in ΔS, K and σ.

What follows

Several conclusions survive the uncertainty above, and they are worth separating from the ones that do not.

●        Age at arrival is the single strongest predictor of fiscal outcome, and this is robust across every model in both countries. It is also the least controversial lever available.

●        Linking intake to absorptive capacity is sound regardless of where you stand on volume. If housing completions and infrastructure investment set the constraint, then planning migration against those measures is better than setting a headline number and arguing about it annually.

●        The federal–local fiscal mismatch is real and fixable. Where the centre collects the income tax and municipalities carry the service costs, transfers to local government are a straightforward correction, and one that would reduce local opposition on its own terms.

●        Credential recognition is the highest-return underused policy. A country that screens for qualifications and then does not recognise them has paid for human capital it refuses to use. The employment-rate gap in the OECD data is the evidence.

●        Fiscal impact should not be the only criterion, and cannot be for humanitarian intake. Those obligations are chosen on other grounds, and the fiscal cost is the price of the commitment, not an argument against it.

Five questions for any immigration number

1.  Compared to what? If the figure is not benchmarked against a native-born citizen, it is close to meaningless. The Australian benchmark is minus $85,000; the American one, on one prominent model, is around minus $250,000.

2.  Snapshot or lifetime? A single-year budget figure will always make recent arrivals look expensive. Check whether the number discounts a whole life.

3.  Which government’s books? Federal-only studies look more positive than consolidated ones, because the costs sit with states and councils. Both are honest; they are not comparable.

4.  Are the descendants counted? Including the second generation improves almost any result. Excluding it worsens almost any result. Studies that disagree usually differ here.

5.  Who is disputing it? On the most contested category — working-age immigrants without degrees — two think tanks are currently accusing each other of technical error. Any source that presents that question as settled has picked a side.

Sources and notes on this revision

Corrections made to the earlier draft of this piece:

●        The Cato Institute and the Manhattan Institute were cited together as joint evidence for a positive average fiscal impact. They are in open methodological dispute; Cato alleges a specific error in the Manhattan model that reverses the sign for low-skilled immigrants. This is now the centre of the American section.

●        The claim that “across all education tiers, the average immigrant pays more in federal taxes than they consume” contradicted the draft’s own table, which showed non-college immigrants at minus $100,000 to minus $400,000. Removed.

●        The skilled-migrant figure of $198,000 was described as being “compared to incumbent native-born baselines.” It is an absolute net present value. Relative to the native-born benchmark the figure is roughly $283,000. The native-born baseline of minus $85,000 has been moved from a table footnote to the centre of the argument.

●        The humanitarian stream (minus $400,000 in the same Treasury model) was omitted from the draft’s table. It has been restored, along with the point that humanitarian intake is not undertaken for fiscal reasons.

●        The OECD result was described as showing a wage uplift for native-born workers. The OECD finding is that migration boosts native employment and does not affect native wages. The productivity result is a cross-regional comparison, not a measured change over time.

●        The OECD’s finding that migrants in Australia have lower employment rates than the native-born despite higher education was absent. It has been added, as it is the strongest evidence for the draft’s own credential-recognition recommendation.

●        The housing section asserted near-zero supply elasticity and strong rent effects without qualification. RBA estimates of magnitude, the zoning literature, and the construction-labour feedback loop have been added.

●        A sentence on the cost of educating a tertiary-qualified worker terminated mid-clause around a missing figure. Rather than supply a number, the passage now explains why a single global estimate is not available.

●        Brain drain offsets were listed without the counter-case. The concentrated losses in health workforces of small and low-income states, and the WHO code of practice, have been added.

●        Sources including Wikipedia, a translation company’s blog, a relocation company’s blog and a migration law firm’s newsletter were carrying substantive policy claims. They have been replaced with the primary documents. Ten of the twenty-seven references in the draft were never cited in the text, and several were duplicates of each other.

On framing: the draft described admitting foreign-trained workers as “human capital arbitrage” in which destination countries “capture” peak productive years. The economics of that observation is sound and is retained. The vocabulary treats people as assets, which will read badly to a general audience and obscures that the fiscal-maximisation objective is a political choice rather than a neutral one. That choice is now made explicit rather than assumed.

References

1. Varela, P. et al. (2021). The Lifetime Fiscal Impact of the Australian Permanent Migration Program. Australian Treasury (FIONA model), December 2021. Source of the +$198,000, −$126,000, −$400,000 and −$85,000 figures and the $127,000 differential. https://treasury.gov.au/sites/default/files/2021-12/p2021-220773_1.pdf

2. National Academies of Sciences, Engineering, and Medicine (2016). The Economic and Fiscal Consequences of Immigration. https://www.nationalacademies.org/projects/DBASSE-CNSTAT-13-03/publication/23550

3. Di Martino, D. The Lifetime Fiscal Impact of Immigrants and The Fiscal Impact of Immigration (2025 Update). Manhattan Institute. https://manhattan.institute/article/the-fiscal-impact-of-immigration-2025-update

4. Bier, D. The Fiscal Impact of Immigration in the United States and the working-paper critique of the Manhattan Institute model. Cato Institute. https://www.cato.org/sites/cato.org/files/2024-11/working-paper-82-update.pdf

5. OECD (2023). Migration and Regional Productivity: Evidence from Individual Wages in Australia; and OECD (2024), Migration and Regional Innovation in Australia, Regional Development Papers No. 66. https://www.oecd.org/en/publications/migration-and-regional-innovation-in-australia_6d6ff472-en.html

6. Centre for Population. OECD: Findings on the effects of migration on Australia’s economy — including the 0.53% native employment result and the finding of no effect on native wages. https://population.gov.au/publications/research/oecd-findings-effects-migration-australias-economy

7. OECD. Migration in Australia — education profile of migrants and the employment-rate gap. https://www.oecd.org/en/topics/sub-issues/migration-regional-and-local-development/migration-in-australia.html

8. Reserve Bank of Australia (July 2025). International Students and the Australian Economy, RBA Bulletin. https://www.rba.gov.au/publications/bulletin/2025/jul/international-students-and-the-australian-economy.html

9. Kendall, R. & Tulip, P. The Effect of Zoning on Housing Prices, RBA Research Discussion Paper; and National Housing Supply and Affordability Council, State of the Housing System 2025.

10. OECD (2024). Society at a Glance — foreign-born population shares across OECD countries. https://www.oecd.org/en/publications/society-at-a-glance-2024_918d8db3-en/full-report/migration_faa3ffd8.html

11. OECD (2017). How Immigrants Contribute to Developing Countries’ Economies; and World Health Organization, Global Code of Practice on the International Recruitment of Health Personnel (2010). https://www.oecd.org/en/publications/how-immigrants-contribute-to-developing-countries-economies_9789264288737-en.html

12. Migration Policy Institute. Competing Approaches to Selecting Economic Immigrants: Points-Based vs. Demand-Driven Systems. https://www.migrationpolicy.org/publication/competing-approaches-selecting-economic-immigrants-points-based-vs-demand-driven