In my last post I argued the Aussie dollar is being held up by the carry rather than by the fundamentals. The biggest of those rotten fundamentals is productivity growth. Australian labour productivity sits below where it was in 2019. Capital deepening has collapsed to zero. Multifactor productivity peaked in 2004 and has gone nowhere since. The slump is deep and it has been persistent.
This matters because we want real wage growth, cost-of-living relief and inflation back to target, all at once. We cannot have all three. Real wage growth without productivity growth is either inflation (the wage rise gets eaten by prices) or profit compression (the wage rise eats margins until firms cut investment and employment). The arithmetic doesn't bend. With capital deepening at zero and labour productivity going nowhere, there is no fairy dust that delivers the desired outcomes. The supply side is the binding constraint and nobody wants to talk about it.
I covered the long sweep of this story in Australia's Productivity Problem: A Lost Decade and Why AI Won't Save Us in January and the measurement framework in Multi-factor Productivity: The Solow Residual in December. This post picks up where those left off. The diagnosis hasn't changed. The question I want to push on here is why the slump has been so deep and so persistent, and what the answer means for real wage growth, for inflation, and for what monetary policy can and can't do. Along the way we will make a number of comparisons with the United States, which is currently experiencing a labour productivity boom.
The headline picture
Take the chart above. Labour productivity climbed steadily from 1978 until around 2022, then stalled and gave some back. If we ignore the pandemic, it has been flat since 2015. Multifactor productivity did most of its work between 1992 and 2004 and has gone nowhere since. As I argued in the lost decade post, we have had one good twelve-year window of genuine productivity growth in the last forty-five years. Bookended by stagnation. The labour productivity line keeps rising for another decade beyond the MFP peak only because firms kept piling capital onto each worker. When capital deepening stalled around 2015, labour productivity stalled with it.
Compared with the US: Australian labour productivity sits below where it was in 2019; American productivity over the same period is up about twelve per cent. Even on the market-sector measure that strips out the non-market composition effect, Australia is 11 points behind the US since 2019, and both economies are below their pre-pandemic trend, but the US is back on trend while Australia is well below it.
The identity
The cleanest way to see this is through the productivity identity. Output per hour worked equals capital per hour times output per unit of capital. Y/L = (K/L) × (Y/K).
Capital deepening (orange) did almost all of the heavy lifting from 1978 to about 2015, more than doubling over the period. Capital deepening just means capital per hour worked: the stock of machines, equipment, software, buildings and other productive assets that each worker has to work with. When workers have more and better capital, they produce more per hour. That's how labour productivity grows.
Capital productivity (green), the output you get per unit of capital, went sideways then fell. Output per unit of capital is 30 per cent lower today than it was in 1978. The whole labour productivity story is a capital deepening story. When capital per hour worked stopped growing around 2015, labour productivity (blue; output per hour worked) stopped growing too. The two lines flatten together on the chart because the identity says they must.
The growth rate makes the slowdown sharper. Capital deepening averaged around 4 per cent a year through the mining super-cycle decade of 2006 to 2015. It has averaged essentially zero since. Capital per worker grew faster in that ten-year window than at any other time in the series. It has not grown at all in the ten years since. The engine that drove four decades of labour productivity growth stopped when mining investment stopped.
Why so deep and so persistent
The slump has multiple drivers and they compound. Each is structural, each has been building for two decades, and each resists the kind of incremental fix that politics can deliver. Together they explain why a decade of low rates, two decades of strong terms of trade, the largest migration program per capita in the OECD, and a banking system flush with credit have not been enough to lift the productivity number. The structural drivers are covered below.
Housing investment eats business investment
The biggest structural drag is the way Australia allocates capital, and the root cause is planning. Restrictive zoning, slow approvals, height limits, heritage overlays, minimum lot sizes and the political economy of established suburbs combine to keep land supply tight in the places people want to live. Tight supply meets growing demand and the adjustment happens through price. Land prices rise. Existing dwellings become more valuable. Capital flows in to capture the price appreciation. Every dollar that goes into bidding up a Sydney house is a dollar that didn't go into a robot, a software stack, a warehouse automation system or a factory upgrade.
Scaled by nominal GDP the story is sharper. Business capital sits at 2.5 times GDP, almost exactly where it was in the late 1980s. The productive base of the economy has not deepened relative to the economy it has to support. Housing has gone from around 2.0 times GDP in the early 1990s to 4.2 times GDP today. Houses have more than doubled relative to the economy. Productive capital has gone nowhere.
The share chart is the cleanest version of the story. In 1989 business capital was 55 per cent of the combined stock and housing was 45 per cent. Today housing is 62 per cent and business capital is 38 per cent. The crossover happened around 1997 and the gap has widened almost continuously since. This is the inversion of an economy. We have spent thirty years turning ourselves from a country whose capital was mostly productive into a country whose capital is mostly residential.
Households have made the same choice the aggregate numbers show. Housing has gone from 53 per cent of household net worth in 1988 to 63 per cent today, and from 47 to 53 per cent of total assets. The shift happened in the early 2000s and has held there for two decades. This is not the residue of a one-off price move. It is a sustained allocation decision by Australian households to hold their wealth in housing rather than in business equity, productive assets, or anything else. When sellers receive proceeds they recycle them back into the same asset class. The system ratchets deeper into housing with each turnover.
Tax settings amplify the problem but they don't originate it. Negative gearing, the capital gains tax discount and the family home exemption all make housing more attractive as an asset. Without supply constraints, those concessions would mostly produce more houses rather than higher prices. With supply constraints, they get capitalised into land values and pull more capital into the bidding war. Planning is the binding constraint. Tax is the accelerant.
The banks have noticed. Housing now accounts for the overwhelming share of bank lending. Business credit has been a falling share for two decades. Bank capital rules that favour mortgage lending against existing dwellings make the tilt worse, but they too are operating on a price signal that planning created. A banking system that lends mostly against existing houses is not a banking system that funds productivity growth. I covered the household debt side of this picture back in 2012 in Debt. The structural pattern hasn't shifted; it has hardened.
Migration as accelerant
Cheap and abundant labour substitutes for capital. Why buy a machine when you can hire another worker on a student visa to do the job for less than the machine costs to finance and maintain? The decision is rational for each individual firm. The cumulative effect across the economy is a capital stock that no longer grows per worker.
The composition matters as much as the level. Permanent skilled migration has run at 30 to 50 thousand a year for two decades. Steady. The volatility in the headline NOM number comes almost entirely from the temporary pathways: students, working holiday makers, visitors and other temporary visa categories. The 2022-23 spike to nearly 560 thousand was overwhelmingly temporary, not permanent skilled. The 2024-25 number of around 320 thousand is still well above the pre-pandemic norm and still dominated by temporary visas.
The stock measure is sharper than the flow. There are now around 1.83 million temporary visa holders working in Australia, up from 1.15 million in early 2021. The biggest single category is New Zealand SCV holders at around 500 thousand, but these have full work rights and behave like permanent residents in the labour market, so they sit outside the productivity argument that follows. Stripping them out leaves around 1.33 million genuinely temporary working visa holders. Students (390 thousand), bridging visa holders (290 thousand), temporary skilled (216 thousand), working holiday makers (203 thousand) and graduate visa holders (175 thousand) make up the bulk of that. The student, bridging and graduate categories have grown fastest.
These are estimates built from Home Affairs quarterly visa stock data multiplied by ABS ACTEID 2021 employment rates by visa class. The employment rates are held constant at their 2021 Census-linked values, so if recent cohorts (especially students after the uncapped-hours rule changes) are working more intensively, the true working stock is higher than 1.83 million. The figure is conservative to moderate.
The policy mechanism that turned this from a skilled migration program into a low-wage labour pipeline was the Temporary Skilled Migration Income Threshold. The TSMIT is the wage floor for the temporary skilled visa program. It was frozen at \$53,900 from 2013 to 2023. Over that decade median wages rose substantially and the threshold quietly stopped filtering for skill. Jobs that paid above median in 2013 paid below median by 2020, and the "skilled" program progressively captured lower-skilled work. The threshold was lifted and indexed in 2023 (it now sits around \$73,150 and has been renamed the Core Skills Income Threshold) but the decade of degraded selection left a large stock of workers admitted under criteria that no longer apply. The Migration Review under Martin Parkinson identified this clearly. The fix going forward is the indexed threshold. The unresolved question is what to do with the existing stock.
Most of these 1.83 million temporary workers are counted in Australia's Estimated Resident Population (ERP; anyone here for 12 of the last 16 months qualifies) and therefore in the standard Labour Force Survey (LFS). The labour force in the Labour Account includes workers not in the LFS: the military (~60k), non-resident workers, workers aged < 15 years and diplomatic personnel. The Labour Account vs Labour Force Survey gap suggests only around 250-300 thousand non-resident workers sit outside the LFS frame, mostly short-stay visitors and working holiday makers in their first year. The chart below shows the gap has been broadly stable at this level for thirty years. The other 1.5 to 1.6 million temporary visa workers are in the denominator of labour productivity, in employment counts, in hours-worked totals. Their effect on the productivity number isn't hidden in some separate statistical bucket. It's built into the headline numbers we've been looking at.
A skilled migration program lifts productivity. A temporary migration program that brings in workers for sectors that would otherwise automate (hospitality, retail, aged care, logistics, food delivery) suppresses it. The counterfactual is hard to prove cleanly, but the mechanism is sound: when labour is cheap and available, the threshold at which firms substitute capital for labour rises. Australia has been running the cheap-labour program at scale. Students, working holiday makers, bridging visa holders and graduate visa holders dominate the genuinely temporary working cohort, and they concentrate in exactly the sectors where capital substitution would otherwise be advancing. The labour is cheap, available and willing. The capital investment doesn't happen. The productivity number doesn't move.
This is the accelerant on top of the housing capital misallocation problem, not the underlying cause. But it's a powerful accelerant. Around 1.33 million genuinely temporary working visa holders (close to 9 per cent of the workforce) are concentrated in low-wage, automation-vulnerable sectors. That's 1.33 million reasons firms don't need to invest in capital deepening to get the work done.
This is not to say there are no genuine skill shortages in Australia. Regional healthcare faces real constraints. So do some engineering specialties, parts of construction, and advanced digital roles. The point is that the temporary visa program operates at a scale far larger than those genuine shortages, and it concentrates in sectors where the binding constraint isn't skill scarcity but the willingness of firms to pay more or invest in capital to do without the marginal worker.
The fiscal case for temporary migration is real and worth acknowledging. Temporary visa holders work, pay income tax and GST, get little or no access to Medicare or welfare, and international students pay full-fee tuition that cross-subsidises university research. Abul Rizvi, who knows the system better than most, makes this case regularly. He's right on the fiscal arithmetic.
But fiscal contribution and productivity effect are different ledgers. A worker who is fiscally net positive can still be substituting for capital that would otherwise have been deployed. The capital-deepening incentive depends on the relative price of labour and capital at the margin, not on the fiscal position of the marginal worker. Net positive fiscally, drag on productivity. Both can be true. The harder question is how to weigh the fiscal upside against the productivity drag.
Energy
Australian industrial electricity prices are up 18 per cent in real terms since 2010. US industrial electricity prices are down 8 per cent. That's a 26-point relative divergence over fifteen years. The chart shows indices not levels, so it doesn't tell you which country had cheaper power in 2010. It tells you Australia made industrial electricity more expensive while the US made it cheaper.
The US went the other way on shale and LNG. The US is now the world's largest LNG exporter, but the shale supply response was large and fast enough to keep domestic gas prices low even as exports ramped up. Domestic users and exporters both got what they wanted because supply expanded to meet both. Australia exports gas at similar scale but supply hasn't expanded the same way, partly because of geology and partly because of state-level moratoriums on onshore gas development. Domestic users now pay export-parity prices because there isn't enough domestic supply to insulate them. The US got cheap industrial power as a dividend of the shale revolution. Australia got expensive industrial power, in part as a consequence of exporting the resource without expanding the supply.
The household side of the same story. Residential electricity prices have risen 82 per cent faster than the rest of CPI since 1990, with most of the gap opening up after 2008. Industrial prices have come back from the 2018-19 peak partly because the political pressure to lower them was real. Household prices haven't, because households can't relocate or substitute the way industrial users can. The costs that came out of industrial bills got shunted onto household bills. It's a hidden cross-subsidy from households to industry, and the political economy of doing anything about it is correspondingly difficult: any fix that lowers industrial prices further now has to confront fifteen years of accumulated household bill increases.
The 2026-27 Federal Budget began addressing this through an east coast gas reservation scheme, announced on 8 May 2026 and starting 1 July 2027. Major LNG exporters will be required to reserve 20 per cent of production for the domestic market, applied to spot gas and new contracts rather than existing export commitments. Compliance is tied to export approvals. A proposed 25 per cent windfall profits tax was ruled out in favour of the supply-side intervention. Whether the design delivers the modest domestic oversupply that Bowen has flagged as the objective depends on details still being worked through. The structural problem is real, the policy response is real, and the test will be whether industrial electricity prices come down enough to restore the productivity dividend that cheap power delivers in the US.
Manufacturing base
Manufacturing has fallen as a share of GDP in both countries, but the Australian decline is sharper and started from a lower base. Australian manufacturing has gone from 10.7 per cent of GDP in 2003 to 5.1 per cent today. US manufacturing has gone from 13 per cent to 9.4 per cent over the same period. The Australian share is now around half the US share. The sectors that used to drive productivity gains through process improvement and capital deepening have shrunk to the point where they no longer move the national average. There isn't much left to lose.
Business dynamism: bad looks good
Productivity growth happens through churn. New firms with better methods displace older firms with worse ones.
The headline ABS picture says churn is rising. Entry rates are at 19 per cent of the opening business stock, up from 15.5 per cent a decade ago. Exit rates are at 17 per cent, up from 15. Net entry is positive. By the standard reading, Australian business dynamism is in good shape.
It isn't. The ABS series counts ABN/GST register entries, which captures every sole trader, contractor and gig worker who registers an ABN to do delivery work, freelance, drive for a rideshare app or sell their labour through a personal services entity. The surge in entries since 2020 is mostly this. It is the labour market restructuring into piece work, not new firms forming with employees, processes and growth potential.
The composition matters because the productivity-relevant kind of dynamism is the opposite of what's happening. The kind of firm that drives productivity growth hires staff, invests in capital, develops processes and competes with incumbents on something other than wage cost. The kind of entry the chart is picking up replaces an employee with a contractor and books it as a new business. Each substitution shows up as a churn data point. None of it lifts productivity.
This is the same mechanism as the migration story in a different guise. Cheap, available, flexible labour substitutes for capital and for permanent employment. Migration delivers it through new arrivals taking low-wage roles. ABN proliferation delivers it through existing workers being reclassified as their own business. The headline dynamism number rises. The productivity number doesn't.
The cleaner evidence is in the BLADE-based research. The RBA Bulletin's September 2023 piece on recent productivity trends shows the picture clearly. Employing business entries (the dark blue line in the top panel) have been flat at 10 to 12 per cent for the entire period. Non-employing entries (yellow, mostly sole traders) have been volatile and surging. The headline ABN entry rate is rising because of the non-employing line, not because employing firms with growth potential are entering at a faster rate. Job mobility, the rate at which workers move between firms, has been flat at around 2.5 to 3 per cent throughout, which means labour reallocation isn't accelerating either. The Hambur and Andrews 2023 RDP "Doing Less, with Less" extends this with firm-level data and shows the longer-run decline more starkly. The kind of dynamism that drives productivity growth has been flat for over a decade. The headline ABN entry rate is picking up something different.
R&D and intangibles
The composition shift in exports
The composition of Australia's goods exports shows the structural story directly. Manufactures (the blue band) made up around 22 per cent of goods exports in the late 1980s. They are around 12 per cent today. Crude materials (largely iron ore) and mineral fuels (coal, gas) have grown from a combined 35 per cent to around 60 per cent. The export base has been progressively concentrated into bulk commodities.
The levels chart shows where the money is. Crude materials run at around \$60 billion per quarter, mineral fuels at around \$30 billion, manufactures at around \$20 billion. The post-2008 surge in total export value is almost entirely commodities. Manufactures have grown in absolute terms but lost share dramatically.
This matters for productivity because complex manufacturing is where process innovation, capital deepening and intangible investment compound. Commodity exports generate revenue but limited productivity uplift beyond the initial extraction infrastructure. An economy whose export base is iron ore, coal and gas is an economy with fewer firms doing the sophisticated things that drive long-run productivity growth. The complex firms that do exist are too small to lift the average.
The tradable sector has been concentrated into bulk commodities. The non-tradable sector has absorbed most of the employment growth, which the next section covers.
The non-market sector grows
The non-market sector has absorbed a steadily rising share of Australian employment for three decades. Public administration and safety, education and training, and health care and social assistance together employed about 21 per cent of workers in the mid-1980s. They now employ almost 32 per cent. Over the same period, the market sector share has fallen from roughly 79 per cent to 68 per cent.
This isn't a recent surge driven by one government or one election cycle. It's a continuous three-decade trend, with the pace picking up slightly in the last five years. The shift is also broadly based: public administration and safety has only risen moderately. Most of the expansion is health, social assistance and education, which makes sense given an ageing population, the NDIS, and a growing economy needing more education.
These are sectors where output measurement is intrinsically difficult because many services are provided outside markets and lack observable prices. In the national accounts, output is therefore often inferred from inputs or costs, which tends to mechanically suppress measured productivity growth relative to the market sector. As the non-market share rises, the headline productivity number is dragged down by composition even if the underlying activity is genuinely valuable.
This does not imply these sectors lack social value. Many represent normal features of a richer and older society. The issue is that they generate weaker measured productivity growth and therefore drag on aggregate productivity statistics as their employment share rises. The Australian shift is sharper than demographics alone would suggest, partly because the migration program has steered new workers into these sectors and partly because the public sector has expanded faster than the population it serves.
The market-sector productivity series strips out much of this compositional effect, and Australia still underperforms the United States, though by less than the headline figures suggest. Both effects matter: market-sector productivity has genuinely weakened, and the long-run shift toward non-market employment compounds the slowdown in aggregate productivity growth.
Will AI save us
The current optimistic narrative is that AI will deliver a productivity surge that fixes all of this. I'm sceptical for general reasons I covered in the lost decade post: AI requires extraordinary capital investment in hardware that depreciates rapidly, inference is compute-intensive at the point of use, and a meaningful share of the gross productivity potential may be absorbed by capital costs rather than showing up in aggregate productivity statistics. If AI was billed at what it currently costs to provide, it would not be competitive as staff augmentation or replacement. Current pricing is being subsidised by venture capital that is funding land-grab market share rather than sustainable margins. The more difficult question is whether AI will be competitive in time, as model efficiency improves and compute costs fall, and whether the productivity case survives once venture capital subsidies fall off.
The Australia-specific story is sharper. AI amplifies existing institutional advantages. Countries with strong capital formation, abundant cheap energy, deep intangible investment and sophisticated firms will capture disproportionate gains. Australia is short on most of those complements. The productivity gains from AI accrue at two ends: the supply end (building the chips, training the models, running the data centres, generating the inference) and the demand end (firms deploying AI tools to lift output per worker). Australia is unlikely to capture much of the first and may struggle to capture as much of the second as the current optimism assumes.
The supply-side challenge is straightforward. AI compute is electricity-intensive and hyperscalers are concentrating investment in jurisdictions with abundant, reliable and relatively cheap industrial power. Australian industrial electricity prices are up 18 per cent in real terms while US prices are down. We will not build frontier AI infrastructure at scale here. Whatever AI Australian firms use will be imported as a service from US and Chinese providers, with most of the supply-side rents accruing offshore.
The demand side is where Australia could still capture gains, but the same structural drags that have stalled capital deepening for more than a decade will constrain AI adoption too. Cheap, abundant labour substitutes for AI just as readily as it substitutes for conventional capital. The hospitality manager who might otherwise deploy an AI booking and inventory system can hire another student visa holder to do it manually. The aged care provider considering AI scheduling and records management can hire another bridging visa holder. The technology changes. The relative input prices don't.
This is the central Australian problem for AI adoption. In large parts of the market sector, firms face weak incentives to automate because the temporary visa program keeps labour cheap and available. AI hits the same substitution barrier that has held back capital deepening more broadly.
Mining is the genuine exception. Mining doesn't run on temporary visa labour and the capital intensity is already high. Rio, BHP and Fortescue are running autonomous haulage, AI exploration models and predictive maintenance at scale. The mining sector will capture meaningful AI productivity gains. Professional services will get software-layer gains. But these are sector-specific bright spots in an economy where the broader market sector has been stalled, and they won't lift the national average enough to close the gap with the US.
The deeper issue is that AI adoption depends heavily on complementary intangible investment: process redesign, organisational capital, data infrastructure and workforce training. These are precisely the categories Australia has persistently under-invested in. Firms that aren't investing in the complements required for AI deployment won't realise the largest productivity gains from AI itself.
The best case for Australia is that we import AI-enhanced products and services from the countries that build them and capture genuine but second-order gains through adoption in mining, professional services and the sectors where the labour-substitution barrier doesn't bind. The worst case is that the global productivity frontier moves further away while Australia stays where it is. AI alone won't reverse fifteen years of structural drag. The same problems that have delivered flat capital deepening will deliver patchy AI uptake.
Policy paralysis
The last serious productivity-enhancing reform agenda was Hawke-Keating through early Howard. The GST in 2000 was the last big structural reform. Tax, competition, federation, industrial relations: all politically too hard to touch at scale. The institutional settings that drove the 1990s productivity surge have not been refreshed in twenty-five years. The 2026-27 Federal Budget had incremental measures (the gas reservation scheme is real, the housing measures are real) but no serious observer would call it a productivity-building reform agenda in the Hawke-Keating sense. The housing measures in particular don't address the binding constraint: without planning reform, tax adjustments on the demand side get capitalised into land prices. You can see the result in the MFP line in the headline chart. Multifactor productivity did most of its work in the decade after those reforms and very little since. I made roughly this point back in 2012 in Productivity Growth: A Mixed Story. The mix has only got worse since.
Why this is the inflation problem
This is where the productivity story connects back to the inflation story. Wages don't have to be high in absolute terms to be inflationary. They just have to outrun productivity. With productivity going backwards, even modest wage growth becomes a cost-push impulse.
Unit labour costs in Australia are up nearly thirty percent from a 2019 base. In the US, twenty. The ten-point gap is the productivity gap with a wage multiplier on top. That cost-push impulse keeps feeding into services prices. Monetary policy can compress demand. It cannot fix a supply side that has broken.
The RBA can hold the cash rate at 4.35 percent for as long as it likes. It will slow consumption. It will not make a hospitality worker more productive, build a factory, restore a manufacturing base, reform planning, lower industrial electricity prices, or rebuild firm dynamism.
What it would take
The honest answer is a reform agenda of Hawke-Keating scale. Planning reform that lets land be used at the densities and on the timelines a growing economy needs. Tax reform that stops capitalising scarcity into land values. Competition reform that lowers barriers to entry across protected sectors. Energy policy that delivers cheap industrial power rather than expensive virtue. Migration policy with a properly indexed skilled-visa wage floor that filters for genuine skill premium rather than degrading into a low-wage rubber stamp. An R&D regime that actually moves the dial on business investment in intangibles.
None of that is on the table. The political economy of each piece is brutal. Each constituency that benefits from the status quo will fight, and the diffuse beneficiaries of reform will not organise. So the productivity line will probably keep going sideways or backwards. Unit labour costs will keep diverging from the US. Inflation will keep getting a cost-push contribution that the RBA cannot fix. And the AUD will keep being held up by the carry until it isn't.
The bottom line: aspirations without foundations
The productivity slump is the most important macro fact about the Australian economy right now. It is the supply-side problem behind the inflation problem behind the rate differential behind the currency. It is deep, it has been persistent for two decades, and it will not be fixed by the next budget, by the next rate decision, or by waiting it out.
The political class wants real wage growth, cost-of-living relief and inflation back to target. The arithmetic doesn't deliver all three without productivity growth. Real wage growth without productivity growth is either inflation or profit compression. Cost-of-living relief without productivity growth is fiscal transfers that have to be funded by someone. Inflation back to target without productivity growth needs demand suppression that costs jobs and investment. None of these promises is coherent without a supply-side foundation, and the supply-side foundation has been crumbling for twenty years.
The market sector has stopped getting better at producing things per worker. Capital deepening is at zero. R&D is half the OECD average. Business dynamism is structurally broken. Manufacturing is five per cent of GDP and falling. Until those things change, the inflation problem is partly outside monetary policy's reach, the cost-of-living crisis is partly outside fiscal policy's reach, and the wage growth aspirations are aspirations without foundations.
No comments:
Post a Comment