Individual assumptions behind the RBA’s forecasts may be reasonable, but their combined effect is a systematically less optimistic assessment of Australia’s future supply capacity than Treasury’s longer-term outlook.
- Every forecast depends on assumptions. Each may be defensible on its own, but when they all move in the same direction, the overall picture becomes less credible.
- Treasury and the RBA reach different conclusions from different assumptions about trend productivity growth, labour-force participation and the stable-inflation rate of unemployment. Treasury’s long-term projections are consistently less pessimistic than the RBA’s shorter-term estimates.
- The risks around the RBA’s assumptions appear asymmetric. They implicitly assume that recent productivity headwinds will persist despite stronger global investment, domestic policy changes and the potential contribution of AI. Extending that weakness beyond the forecast period may also conflict with evidence that global interest rates are likely to remain above their pre-pandemic levels.
Forecasts become harder to substantiate the further into the future they extend. Uncertainty about AI ranges from expectations of extraordinary productivity gains to severe economic disruption, making a continuation of recent trends an appealing fallback. The key question is how much of the past should be extrapolated.
This is the context for the latest Intergenerational Review. Treasury assumes productivity growth will average 1.2% a year over the next 40 years, consistent with the 30-year average. It also expects labour-force participation to continue its long-term upward trend until 2040.
The RBA takes a weaker view, assuming productivity growth of 0.7% a year, in line with the past 20 years and the period following the global financial crisis and pandemic. It also expects participation to be broadly flat or lower in the near term.
Both approaches have limitations. Averaging over round-number periods can capture a downturn without the subsequent recovery, or the reverse. Peak-to-peak or trough-to-trough comparisons are preferable, although they still cannot avoid the weakness evident in recent years.
In fairness, the RBA’s productivity assumptions concern the near term and are not directly comparable with Treasury’s 40-year projections, as the Governor has emphasised. The divergence over participation rates is nevertheless genuine and was first identified by Ryan Wells in May.
The RBA’s more pessimistic assumptions may unintentionally reinforce a narrative that Australia’s growth prospects are weaker than they are. Some commentators attribute productivity growth almost entirely to domestic policy reform, pointing to the deregulation associated with the Hawke-Keating era and the absence of comparable reforms since. That perspective overlooks the global nature of the productivity slowdown.
It also assumes that the factors restraining productivity will continue for decades. Conditions that reduced risk appetite, investment and capital accumulation after the global financial crisis may not recur to the same extent absent another major crisis. The current AI investment boom and renewed appetite for innovation make it difficult to predict a return to the earlier period’s lack of productivity-enhancing technology.
For Australia to achieve average productivity growth of 0.7% over the next 20 years, the economy would need to experience another compositional drag similar to the recent expansion of the care economy and decline in measured mining production. Decarbonisation is likely to reduce capital-intensive, high-productivity coal output and weigh further on mining productivity. However, assuming that critical and transition minerals provide no offset would be difficult to justify.
AI may take time to lift measured productivity, but some benefit could emerge before the end of the RBA’s forecast horizon in 2028. Recent UK GDP data already suggest an AI-related boost to service-sector activity. Claiming that the rest of the world will benefit while Australia will not requires a clear explanation for the divergence.
The Productivity Commission estimated in December that AI could increase multifactor productivity by 2.3 percentage points over the next decade, averaging roughly 0.25 percentage points of additional annual growth. Labour-productivity growth could rise somewhat more as data-centre investment expands the capital stock.
That estimate may not fully reflect the improvement in leading AI models released over the summer of 2025–26, particularly coding tools and agentic software. Limited AI adoption in the Australian public sector may also cause policymakers to underestimate both technological capability and the breadth of private-sector uptake.
A simple illustration shows how optimistic assumptions are needed to extend the RBA’s near-term productivity projection. If AI made only 10% of the economy 10% more productive over four years, the economy-wide productivity gain would be about 1%. That would still add roughly 0.25 percentage point to annual productivity growth for four years, lifting the 0.7% assumption to almost 1%.
Beyond the RBA’s short-term horizon, persistently low productivity growth also sits uneasily with the prospect of a higher global average interest rate.
RBA Deputy Governor Andrew Hauser has cited nearly 700 years of bond-yield data to argue that the post-financial-crisis period of low rates was exceptional. Global interest rates balance worldwide saving and investment, and investment associated with AI and other projects is already evident. It is difficult to assume that this investment will raise global yields while delivering no productivity gains. Even increased defence spending can generate productivity benefits, as Ukraine’s expanding drone industry illustrates.
When inflation is above target, a central bank is right to favour caution over technological optimism. Policies that are slightly too restrictive can bring inflation back to target more quickly, which is consistent with an inflation-targeting mandate.
The concern is that understandable assumptions can become systematically one-sided. Arguments for productivity growth below the 0.7% 20-year average have yet to be demonstrated, while credible reasons exist for it to be somewhat higher.
The RBA’s participation-rate assumptions are also low relative to near-term demographics and Treasury’s revised projections. The latest August labour-market data show participation rising again, to just 12½ basis points below the all-time high recorded in January 2025. That supports the view that the RBA’s participation outlook is too pessimistic.
The Intergenerational Review’s estimate of the NAIRU, or stable-inflation rate of unemployment, is also roughly 0.25 percentage point below RBA estimates revealed through freedom-of-information requests and the Governor’s recent comments.
Together, these three RBA assumptions imply weaker supply-capacity growth and therefore greater inflationary pressure for any given level of demand. Each assumption may appear reasonable in isolation, but their combined downward bias suggests a pattern. The tilt is unlikely to be deliberate, yet it points to the likely direction of medium-term surprises.
