The third demountable classroom arrived in the winter school holidays, craned over the fence and bolted onto the square of asphalt where the Year 5s had played handball the term before. The family four streets away — she collects pathology samples, he drives a delivery van, their two children started at the school in 2022 and 2024 — did not object to it. Demountables are ordinary. Every parent in an outer-suburban growth corridor knows what they look like on a hot February afternoon. What they noticed was quieter: the library had been partitioned to make a small-group teaching room, the canteen queue now ran in two shifts, and the permanent building block that once appeared in a departmental planning document was still, three years later, a line on a page rather than a slab on the ground.
Nothing about that school is a scandal. Relocatable classrooms are a legitimate and often excellent tool: fast, cheap relative to permanent construction, and designed precisely so a school can absorb a surge without turning children away at the gate. The signal is not that a demountable arrives. The signal is how long it stays. A buffer used for one year is working as designed. A buffer used for five, on a site where the permanent build keeps slipping down the capital program, is a capital works pipeline that has fallen behind the population it serves.
This article is about that gap, and it is deliberately not about housing. The collision between the post-pandemic migration surge and the rental market is examined in full elsewhere on this site. What follows is everything else — the schools, the emergency departments, the roads and buses and trains, the water storages and the electricity connections that were planned, funded and sequenced on population projections written before the surge, and are still being delivered on those older assumptions.
Start with the correction: migration is falling, not rising
Net overseas migration is not at record levels. It is falling, and it has been falling for some time. Writing otherwise in the present tense would be wrong, and the argument that follows is stronger without it.
Net overseas migration — NOM — is the difference between the number of people arriving in Australia intending to stay twelve months or more and the number departing for twelve months or more. It is not the same thing as the permanent Migration Program, which is a separate annual ceiling for permanent visas set in the Budget. NOM is dominated by temporary visa holders, chiefly international students, temporary graduates, working-holiday makers and temporary skilled workers, and it is that temporary flow that makes the total swing.
The peak of the post-pandemic surge was 2022-23 (Centre for Population). In calendar 2025, net overseas migration was approximately 301,000 (ABS, Overseas Migration; as reported). The Centre for Population's 2025 Population Statement forecasts NOM of 260,000 for 2025-26, and population growth moderating to around 1.3 per cent — close to Australia's long-run average rather than an outlier. The direction of travel has been unambiguous for more than two years.
Why it is falling matters as much as the fact that it is. The Centre for Population attributes the decline to three things operating together. The first is the exhaustion of post-pandemic pent-up demand: the students, workers and family members whose plans were frozen between March 2020 and early 2022 have largely moved through the system, and a rebound from an artificially suppressed base cannot repeat itself. The second was always coming — the large cohort of temporary visa holders who arrived in 2022 and 2023 is now reaching the maximum stay their visas allow, which raises departures and lowers the net figure without any new arrival being refused. The third is deliberate policy: the Genuine Student test, higher English-language thresholds, and a broader set of student-visa integrity measures aimed at low-quality providers.
That third factor deserves to be stated plainly and without grudging qualification: it is the government's own action, and it worked in the intended direction. The Commonwealth changed the settings it controls, and the number it targeted came down. A critique that refuses to say so is not analysis; it is advocacy. So why write this article at all? Because the flow and the stock are different things, and the entire policy problem lives in the difference between them.
A school place not built in 2023 is still not built in 2026. The flow slowing does not build it.
Migration is a flow: people per year, a number ministerial direction, visa criteria and processing priority can turn up or down within a quarter. Infrastructure is a stock: classrooms, beds, lane kilometres, gigalitres, transformer capacity — things that exist or do not, and take years to bring into being. When the flow ran ahead of the stock for three years, the stock did not catch up once the flow slowed. It stayed where it was, serving a population permanently larger than the one it was sized for. Australia's population is not going back to its 2021 level. The deficit accumulated in 2022, 2023 and 2024 is an opening balance, not a temporary spike.
What the Commonwealth controls, and what it does not
The constitutional division has to be set out, because almost every sloppy version of this argument fails at exactly this point. The Commonwealth Parliament has power over immigration and naturalisation under section 51 of the Constitution. Visa classes, eligibility criteria, English-language thresholds, post-study work rights, processing priorities and the size of the permanent Migration Program are all federal. So is the Budget forecast of net overseas migration — the number every other planner in the country works from.
Almost none of the infrastructure discussed here is federal. Public schools are built and maintained by state education departments. Public hospitals are owned and run by state health services; the Commonwealth co-funds activity through the National Health Reform Agreement but does not build wards. Urban roads, buses, trams and suburban rail are state. Water utilities are state-owned or state-regulated. Land-use planning — the approvals deciding when a growth corridor is released and what services it must be given — is state and local. The Commonwealth funds land transport through the Infrastructure Investment Program and makes grants to the states, most importantly the GST pool. But it does not hold a shovel. On the question this article asks — did the classrooms, beds and connections get built in time — the delivering government is almost always a state government, of whichever party.
That is not a defence of the Commonwealth. It is the precise shape of the problem. One level of government sets the demand signal. A different level owns the capital response. Nothing binds them together.
The strongest case for the policy
A rebound from an artificial zero is not a new normal — and migration builds capacity as well as demanding it
Put the government's case at its strongest and it is genuinely formidable. For close to two years the Australian border was effectively closed by federal direction, and in 2020-21 net overseas migration turned negative — more people left than arrived, for the first time since the Second World War (ABS Overseas Migration, approx.). Australia entered 2022 with a population several hundred thousand smaller than the pre-pandemic trend implied. Measuring the rebound against the closed-border years, rather than against the trend line the closure interrupted, systematically overstates it. A surge measured from an artificial zero is arithmetic, not policy. The Centre for Population's own explanation of the current decline — pent-up demand exhausted, visa cohorts reaching maximum stay — confirms that reading: the surge behaved exactly as a catch-up should, and then it stopped.
Second, migration is not only a source of demand for infrastructure. It is a source of the capacity that builds and staffs it. Construction, engineering and healthcare have been among the most persistent shortage occupations in the country for years (Jobs and Skills Australia, Skills Priority List, various). Every additional civil engineer, electrician, project manager, registered nurse and radiographer admitted through the skilled streams is a unit of delivery capacity, not just a unit of demand. A program cut hard enough to relieve pressure on services would also thin the workforce that expands them, and that second effect operates on a longer lag than the first.
Third, migration expands the tax base that pays for capital works. Migrants on skilled and student pathways are overwhelmingly of working age, arrive with education already paid for by someone else, and pay income tax, GST and, indirectly, state payroll tax for decades. Treasury's Intergenerational Report 2023 is explicit that sustained migration improves the Commonwealth's fiscal position over the long run by slowing population ageing and lifting the ratio of workers to retirees. A closed-border counterfactual is fiscally worse, not better — fewer taxpayers supporting the same age-related spending means less money for schools and hospitals, not more.
Fourth, and most awkwardly for critics, the delivery failure sits with a different level of government from the one setting the migration dial. The Commonwealth does not choose when a growth corridor is rezoned, when a school site is acquired, or how a state sequences its capital program against its credit rating. States had their own projections, their own planning departments and their own choices. Blaming a federal number for a state delivery outcome collapses a genuine federation coordination problem into a partisan one.
And finally: the government did act. It tightened student-visa integrity, introduced the Genuine Student test and raised English-language requirements, and net overseas migration fell. On its own published forecast it will fall further. A critique that treats the migration number as an ongoing, unaddressed failure is arguing with a version of the data that no longer exists.
The evidence: two clocks, running at different speeds
How fast each clock actually runs
The analytical core of this topic is a mismatch between two clocks: the migration clock, which the Commonwealth winds, and the capital works clock, which the states wind and which cannot be sped up much by wanting it faster. A migration setting can change almost immediately. A ministerial direction reordering visa-processing priority takes effect the day it is issued. A change to the Genuine Student test, an English-language threshold or a financial-capacity requirement is made by legislative instrument or policy guidance and flows through to grant volumes within months. The permanent Migration Program ceiling is reset every Budget. In practice the Commonwealth can move the annual population growth signal by a hundred thousand people or more inside a single fiscal year, and in 2023-24 and 2024-25 it did.
Now the other clock. A new government primary school in a growth corridor requires a site reserved through a precinct structure plan or equivalent, then acquired, designed, funded in a state capital program, tendered and built. Where the land is already in public hands and funding is allocated promptly, three to four years from decision to first students is a good outcome; where land must be acquired or a planning dispute intervenes, five or more is common (state education department capital programs and state budget papers, various). A major hospital redevelopment — a new emergency department, expanded theatres, additional inpatient beds — is a five-to-ten-year proposition from business case to commissioning, assuming the workforce to staff it is recruited in parallel. Urban heavy rail is slowest: ten to fifteen years is normal for a new line, and the largest projects in Sydney, Melbourne and Brisbane have run longer.
Electricity is the exception that proves the rule, and it is instructive. Connecting an individual new house is quick. Augmenting the distribution network that serves a whole new estate — zone substations, feeders, transformer capacity — is not: it requires easements, planning approvals, long-lead equipment with global supply queues, and, critically, it must be funded inside a five-year regulatory control period determined by the Australian Energy Regulator. A distributor that did not forecast the load in its last determination cannot simply spend the money now.
The path of the number, drawn honestly
Schools: the demountable as an index of lag
School enrolment is the cleanest place to see the two clocks diverge, because the demand signal arrives with a predictable delay and the supply response does not. A person arriving on a student visa does not generate a primary school place; a family arriving with a four-year-old generates one the following year and then for thirteen consecutive years. That lag is a gift to planners: it is a demand signal you can see coming. The ABS publishes national enrolment counts annually in Schools, Australia, and state education departments run their own projections at the individual school level. Nobody had to guess.
Where the pressure landed was equally predictable. Australia's fastest-growing local government areas are consistently outer-suburban growth councils — Melbourne's west and south-east, Sydney's north-west and south-west, Brisbane's western and southern corridors, Perth's north (ABS, Regional Population, recent releases). These are where new estates are released, where new arrivals and existing residents alike move for cheaper land, and where the school portfolio is thinnest and least able to absorb a surge inside an existing building.
The management tool is the relocatable classroom, and every state uses it. The policy question is the ratio of buffer to permanent stock and the duration of its use. State budget papers and Auditor-General reports across several jurisdictions have examined school capital programs during this period and consistently found pressure in growth corridors, with permanent builds sequenced behind demand rather than ahead of it. Where a build slips, the consequences fall unevenly inside a school. Specialist spaces go first: the library becomes a classroom, the art room becomes a classroom, the hall gets a timetable. None of that is visible in an enrolment statistic, and all of it is visible to a child.
The federal dimension here is narrow. The Commonwealth is the majority public funder of non-government schools and a minority funder of government schools, on recurrent terms set under the Australian Education Act 2013. Recurrent funding pays teachers and running costs; it does not build classrooms. Public school capital is a state responsibility funded from state budgets.
Hospitals: a denominator problem
Hospital capacity is measured in beds, theatres, treatment spaces and staff. Demand is measured per head of population. When the denominator grows faster than the numerator, performance degrades even if nothing inside the hospital gets worse. The relevant published series are well established and independent. The Australian Institute of Health and Welfare publishes emergency department presentations and waiting times, and elective surgery waiting times, annually; the Productivity Commission's Report on Government Services compiles the same measures on a comparable state-by-state basis each January. Across the period since 2019 those series show rising presentation volumes, and elective surgery waiting lists that lengthened sharply when non-urgent surgery was suspended by public health direction during the pandemic and have been worked down unevenly and incompletely since (AIHW; Productivity Commission, RoGS, editions through the 2020s).
Attribution has to be careful here: coincided with is not caused by. Emergency department and elective surgery pressure has causes with nothing to do with population growth — the deferred-care backlog from suspended surgery, an ageing population with more complex chronic conditions, workforce attrition among nurses and junior doctors, and difficulty accessing bulk-billed general practice pushing lower-acuity presentations into EDs. Migration is one term in that equation and not the largest one.
But it is a term, and it is the one that sets the denominator. A hospital's footprint and staffed bed count are fixed by decisions made years earlier against a catchment population projection. Where a growth-corridor catchment grew faster than the projection the hospital was sized against, the mismatch is structural: it cannot be resolved by working harder, and it will not be resolved by the flow slowing, because the catchment does not shrink.
The Commonwealth's role is real but indirect. Under the National Health Reform Agreement it contributes a share of the efficient growth in the cost of public hospital services — around 45 per cent of growth, subject to a cap on the annual increase in the Commonwealth contribution in the order of 6.5 per cent (approx.). That is activity funding: it pays for services delivered, not capital. If a state needs a new emergency department, the state builds it, and where the capped Commonwealth contribution grows more slowly than demand, the residual falls on the state budget.
The Commonwealth can change the population forecast in a budget. The state has to change the hospital in a decade.
Transport: the fast lever and the slow one
Transport is where the asymmetry is easiest to feel and hardest to fix, because the useful capacity is exactly the kind that takes longest to build.
Growth corridors are, by construction, car-dependent. They are released at the urban fringe, typically ahead of fixed public transport, and the initial travel pattern is a long car commute toward employment closer to the centre. Arterial road capacity is finite, and congestion is non-linear: throughput degrades slowly as volume rises and then collapses past a threshold. A corridor absorbing 5 per cent more trips a year for three years does not experience 15 per cent more delay. Past the threshold it experiences considerably more.
Two levers are available and they run on different clocks. Bus service kilometres can be added inside a budget year: vehicles, drivers and a timetable, and while driver recruitment is a real constraint, it is a months-to-a-year problem. Rail is a decade problem. Extending a line, adding a station, or building the turnback and stabling capacity that lets a timetable run more services requires planning approval, land, procurement and construction on a horizon that outlasts several parliaments.
Infrastructure Australia, the Commonwealth's independent advisory body, has documented this pattern repeatedly in its Australian Infrastructure Audit work and in the Infrastructure Priority List. Outer-suburban access to jobs, services and public transport has been a recurring entry, identified as a national problem well before the post-pandemic surge (Infrastructure Australia, various editions). The point is not that nobody noticed. Everybody noticed. The point is that noticing produced advice, and advice does not produce a rail line.
In 2023 an independent strategic review of the Commonwealth's Infrastructure Investment Program found the pipeline overcommitted relative to both the funding envelope and the construction sector's capacity to deliver it, and the government subsequently discontinued around fifty projects and reprofiled others (approx.). That is defensible on its own terms — an overcommitted pipeline inflates costs and delivers nothing faster. It is also a concrete instance of the mismatch: at the moment the population signal was at its highest, the deliverable pipeline was being trimmed to what the sector could actually build.
Water: the longest lead time of all
Urban water security has the longest planning horizon of any service considered here, and the least political attention until the moment it fails. Supply capacity is fixed in the medium term; demand is the product of population and per-capita consumption. Australian cities achieved large and durable reductions in per-capita use through and after the Millennium Drought — restrictions, pricing, efficient appliances, dual-flush and low-flow standards — and those reductions have masked a great deal of population growth. That is a genuine success and it is not infinitely repeatable. Once the easy efficiency is banked, population growth translates into demand growth close to one-for-one.
Augmentation is slow. The desalination plants built for Sydney, Melbourne, Perth, Adelaide and the Gold Coast between roughly 2006 and 2012 followed the Millennium Drought, and each took several years to build after several more years of argument about whether to build it at all. Recycling and indirect potable reuse face longer timelines again. No water augmentation responds to a demand surprise inside an electoral cycle. And urban water is entirely a state responsibility — state-owned utilities, state price regulators, state approvals.
Electricity and gas connections: the regulatory clock
The last piece is the least visible and, in growth corridors, one of the most consequential: getting a new estate connected to the grid at adequate capacity. Distribution network businesses have their allowable revenue and capital expenditure determined by the Australian Energy Regulator in five-year regulatory control periods, built on forecasts of connections and load. A distributor that underforecast growth in a corridor cannot simply spend more in response to actual demand: it is spending inside an envelope set years earlier, and adjusting it requires a regulatory process. That is not a flaw in the framework — the framework exists to stop network businesses gold-plating assets and charging consumers for them, a problem Australia learned about expensively in the 2010s. But it is a five-year clock bolted onto an annual demand signal.
The load is also changing shape faster than the forecasts that sized the network. Victoria has required most new homes needing a planning permit to be all-electric since 1 January 2024, and the ACT has moved similarly. Electric vehicle uptake, rooftop solar and household batteries all change the size and daily shape of the load a new estate places on a local network. AEMO's biennial Integrated System Plan models the transmission and generation implications nationally, but the pinch point for a family moving into a new house is local: the zone substation, the feeder, the transformer. Long-lead electrical equipment sits in global supply queues that lengthened after 2021, adding another year or two no Australian government controls.
The critique: nobody owns the question
Set the evidence out end to end and the failure is not that migration was too high. It is that the two decisions were never joined. The Commonwealth sets a migration number and publishes a NOM forecast in Budget Paper No. 1 every year. The Centre for Population turns that into national and state population projections, which state treasuries and planning agencies then use. The information pipeline exists and it works. What does not exist is any mechanism by which the infrastructure consequence feeds back into the migration decision, or by which the migration decision binds anyone to deliver the infrastructure.
There is no infrastructure capacity test attached to the setting of migration levels. No Commonwealth process requires the minister, before changing student-visa settings or the permanent program ceiling, to obtain and publish an assessment of whether schools, hospitals, water and network capacity in the receiving regions can absorb the resulting population. No state government has a veto, a right of consultation with legal force, or an automatic funding entitlement triggered by a change in the federal number. Infrastructure Australia advises on projects; it has no role in migration settings.
The reverse loop is equally absent. When the Commonwealth's own NOM forecasts were repeatedly exceeded through 2022 and 2023 — a matter of public record in successive Budget updates — nothing in the federal financial architecture automatically increased state capital funding in response. GST distribution does adjust for population, but through Commonwealth Grants Commission assessments that use averaged historical data and therefore respond with a lag of years, and it is untied revenue a state may spend on anything. It is not a capital response mechanism.
Compare the one area where a coordination mechanism was actually built. Housing supply got the National Housing Accord, agreed at National Cabinet in August 2023, with a national target, an incentive payment structure and a reporting cadence. Reasonable people disagree sharply about whether it is working. But it exists: a forum, a target, and a number ministers are asked about. There is no equivalent for school places in growth corridors, emergency department capacity, water security, or network augmentation. The same demand signal produced one coordination mechanism and none for everything else.
The result is an accountability vacuum with a familiar shape. Ask why a growth-corridor school is in its fifth year of demountables and the state department points to a capital program constrained by a budget constrained in turn by the state's debt position. Ask why the population grew faster than projected and the answer runs to federal visa settings. Ask the Commonwealth and the answer is that it does not build schools. Every one of those answers is true. That is the problem: a system in which every participant can truthfully explain why the outcome is not their responsibility will produce that outcome again.
Separate the categories cleanly, because this site's credibility depends on it. The border closure that created the pent-up demand was imposed by the previous government, in circumstances that were defensible at the time; the current government inherited the rebound, not caused it. The state capital programs that fell behind were set by state governments, of both parties, answering to their own budgets. What the current Commonwealth government owns is narrower: it set the visa settings that governed how fast the catch-up was allowed to run and how long it ran for; it published NOM forecasts that were exceeded repeatedly without a corresponding capacity assessment; and it did not build, and has not built, an institutional mechanism to align the demand signal it controls with the capital delivery it does not. That last omission is the critique, and it is a critique about institutional design rather than about any number of people.
What they got right, and the honest caveats
Several things belong on the credit side of the ledger, and they should be said without hedging. The migration number is down: net overseas migration peaked in 2022-23 and has fallen substantially since, to approximately 301,000 in calendar 2025, with 260,000 forecast for 2025-26 (ABS Overseas Migration; Centre for Population, 2025 Population Statement). Population growth is moderating toward around 1.3 per cent in 2025-26, close to Australia's long-run average. Whatever else is true, the flow is no longer accelerating.
Government policy contributed to that. The Genuine Student test, higher English-language thresholds and the broader student-visa integrity agenda were deliberate measures aimed at a specific part of the flow, and that part came down. It is possible to argue the measures came later than they should have. It is not honest to argue they did nothing.
The transparency infrastructure is good and it has been maintained. The Centre for Population was established within Treasury in 2019, under the previous government, and the current government has continued and funded its annual Population Statement. That series — published methodology, state-level detail, and a forecast that can later be marked against the outcome — is the reason an article like this can be written with real numbers rather than assertion. A government that publishes the number its critics will use against it deserves to be told so. The same applies to the ABS overseas migration and regional population series, and to the AIHW and Productivity Commission health series relied on above.
Infrastructure Australia's independence was strengthened rather than weakened during this period through the 2023 reforms to its governance and remit, and the 2023 review of the Infrastructure Investment Program was a genuine attempt to match a pipeline to deliverable capacity rather than to announcement value. Trimming an overcommitted pipeline is unglamorous and politically costly. On the evidence, it was the right call.
Now the caveats, stated with equal clarity. A forecast is a forecast: the 260,000 figure for 2025-26 is a projection, not an outcome, and the recent history of NOM forecasting in Australia is a history of forecasts being wrong in both directions by large margins. And the deepest caveat is the one the steel-man raised: an infrastructure constraint that is partly a workforce constraint cannot be solved by admitting fewer of the people who staff the workforce. Any mechanism built to align migration with infrastructure must be capable of concluding that the answer is more skilled migration in specific occupations, not less, or it is not an analytical mechanism at all.
The risk to the future
A capital stock deficit that outlives the flow that created it
The forward risk is not a population risk. It is a delivery-lag risk with a specific shape. Australia's population is now permanently larger than the population the current stock of schools, hospitals, arterial roads, water supply and distribution networks was sized against, and that stock turns over slowly: a school built today serves for fifty years, a hospital redevelopment for thirty, a rail line for a century. Catching up therefore does not mean building at the rate of population growth. It means building above that rate for long enough to close a gap opened over three years — while also replacing ageing assets and delivering the energy transition, from the same finite pool of engineers, project managers and trades.
If that does not happen, the consequence is not dramatic and it will not produce a crisis anyone can date. It is a slow, compounding decline in the quality of ordinary public services in exactly the places where the population grew fastest: larger classes and fewer specialist spaces in growth-corridor schools, longer emergency department waits and elective surgery queues in growth-corridor hospitals, longer commutes for people who already have the longest ones, tighter restrictions in a dry water sequence, and network constraints in new estates. It degrades service quality for everyone in those catchments — people who have lived there thirty years and people who arrived last year, identically, because a queue does not check anyone's visa history.
The second-order risk is political. When public services degrade in a growing area, the explanation easiest to reach for is the visible change rather than the invisible one — a capital program sequenced years ago against a projection that turned out wrong. Keeping the infrastructure question and the migration question separable is not a rhetorical nicety. It is the difference between a debate about capital works sequencing, which is solvable, and a debate about who belongs here, which is not a policy debate at all.
The forward-looking test is therefore narrow, checkable, and independent of any view about the right level of migration. Is there any institution in the Australian federation whose job it is to answer the question "can the receiving regions absorb this population, and if not, what has to be built and by when?" — and to be held accountable for the answer? As at 2026, on the public record, there is not. Until there is, the same asymmetry will produce the same result the next time the demand signal moves, in whichever direction it moves.
What would actually close the gap
A critique that names no remedy is only half an argument, so the shape of one is worth setting out — while being honest that it is a design problem before it is a spending problem, and that this site is not in a position to price it. The minimum credible version is an infrastructure capacity statement published alongside the migration program settings each year: a public, regionally disaggregated assessment of what the projected population implies for school places, hospital capacity, water supply and network capacity in the receiving regions, and what is already funded against it. It would not need to bind anyone. It would only need to exist, appear on a fixed cadence, and be checkable against outcomes later — exactly the discipline the Population Statement already imposes on the population forecast itself. The mechanism is proven. It has simply never been extended from the demand side to the supply side.
The harder version involves money and federation politics: some form of population-triggered Commonwealth capital contribution to the states for social infrastructure in high-growth regions. That is a substantial reform with substantial risks, and designing it is beyond this article. But the absence of any such mechanism, in a federation where one government controls the population signal and another controls the buildings, is the fact this whole topic rests on. The family with the third demountable will not read a Population Statement. What they will notice is whether the permanent block appears before their younger child finishes primary school — and the answer no longer has anything to do with how many people arrive next year.