Assumptions underlying RBA forecasts might be individually reasonable but in combination they suggest a systematically downbeat view of supply capacity growth relative to Treasury’s longer-term view.
- Any forecast rests on assumptions. Each might be individually reasonable, but if they all tilt in the same direction, they look less plausible in combination.
- Different assumptions can also drive different conclusions, as can be seen in Treasury and RBA views on trend growth in supply capacity. Treasury’s long-term view on both productivity growth and labour force participation trends is systematically less downbeat than the RBA’s (shorter-term) view. Treasury’s estimate of the stable-inflation rate of unemployment is also lower than the RBA’s estimate.
- We consider the risks around the RBA’s view to be asymmetric. The RBA’s assumptions implicitly require a repeat of the factors that dragged recent productivity outcomes down, despite strong global investment, policy changes domestically and the possible benefits of AI. Some observers appear to extrapolate this weakness beyond the RBA’s forecast horizon. But this would be hard to square with the reasonable belief that the global interest rate structure will be higher in future than it was pre-pandemic. A systematic tilt, even if unintentional and understandable, may point to the likely direction of future surprises.
Any forecast rests on assumptions, and the further out you forecast, the harder it is to ground those assumptions in something real. Add the uncertainty about AI, where predictions range from “unbounded prosperity without needing to work” to “we will all be dead in a couple of years”, and you can see why assuming the future looks like the past is an attractive fall-back position.
How much of the past, though? This is the context for the debate about the Intergenerational Review (IGR) released this week. In one corner, Treasury assumes that future productivity growth will average 1.2%yr over the next 40 years, in line with the 30-year average, and that labour force participation will continue to trend up, as it has for the past five decades, until 2040. In the other corner, the RBA assumes that productivity growth will average 0.7%yr, in line with the past 20 years – essentially the post-GFC and pandemic period – and that participation rates will be flat to down in the near term.
Both productivity assumptions have limitations: round-number-of-year averages might be distorted by capturing a downturn without the matching upturn, or vice versa. Peak-to-peak or trough-to-trough is always preferable. Even that calculation cannot escape the weakness of recent years, though.
To be fair, the RBA’s productivity assumptions relate only to the short term and are not strictly comparable to the IGR, as the Governor has emphasised this week. By contrast, the near-term divergence in forecasts for the participation rate, which my colleague Ryan Wells first highlighted in May, is a genuine divergence.
The RBA’s more downbeat set of assumptions could be unwittingly contributing to a self-fulfilling talking-down of Australia’s prospects that clearly appeals to some commentators outside the public sector. Many in this group view productivity growth as stemming wholly or mostly from policy “reforms”, harking back to the halcyon Hawke–Keating era of deregulation. Proponents point to the lack of reform recently to justify this downbeat view of growth capacity. We think this misses the global nature of that slowdown.
Part of the trouble with the downbeat view of future productivity is that it implicitly assumes that the factors weighing on productivity growth in the past will continue. But there are good reasons not to expect the same post-GFC weakness in risk appetite, and so investment and capital accumulation, decades later and absent another crisis. The AI boom and associated resurgence of risk appetite make it hard to argue that we will see the same lack of new productivity-enhancing technologies and general malaise as that period.
For Australia specifically, assuming the same low average productivity growth as the past 20 years would require assuming a compositional drag similar to a repeat of the run-up of the care economy and decline in measured mining production over the past 15 years. To be fair, decarbonisation does imply further declines in (capital-intensive, high-productivity) coal production, dragging mining productivity down further. It would be a stretch, though, to assume expansion in critical and transition minerals provides no offset.
We can grant that AI’s impact on productivity will take a while to come through and still expect some boost in coming years – at least by the end of the RBA’s forecast horizon in 2028. Indeed, some observers already see an AI-induced boost to service-sector activity in the recent upside surprise in UK GDP. Some productivity pessimists might want to argue that the rest of the world will benefit from the AI boom, but somehow not Australia, or not until later. Such an extraordinary claim requires a mechanism explaining that divergence, though.
The Productivity Commission estimated back in December that AI could boost multifactor productivity by 2.3%pt over the course of the next decade, or roughly a quarter-point per year extra growth on average – a big difference in the scheme of usual growth rates in output and productivity. Labour productivity growth would be boosted a little more; the data centre boom is expanding the capital stock. But an estimate published in December is unlikely to adequately capture the step change in capability of the main models, especially the coding and agentic harnesses, that emerged over the summer of 2025/26. Given what we understand to be limited adoption of AI in the Australian public sector, policymakers might also be systematically underestimating both the technology’s capabilities and the breadth of uptake in the private sector.
As an illustration of how downbeat you need to be to extrapolate the RBA’s near-term assumption further out, suppose that all AI does is make about 10% of the economy 10% more productive over four years. This extra 1% of economy-wide productivity is well below most estimates of the impact. It would nonetheless add a quarter-point on annual productivity growth for four years, bringing a downbeat 0.7% assumption up to almost 1%.
A deeper challenge is that, beyond the RBA’s near-term horizon, low productivity growth is hard to square with a higher future average for interest rates globally.
We note, for example, recent comments by RBA Deputy Governor Hauser, pointing to 700 years of bond yield data to suggest that the post-GFC period of low interest rates was an anomaly, something we noted two years ago. The global structure of interest rates emerges to balance global saving and investment, and stronger investment for AI and other purposes is clearly in evidence. We find it implausible to assume that all that investment happens, pushing up global yields, and none of it delivers higher productivity. Even increased defence spending has historically had productivity payoffs, as Ukraine’s drone manufacturing ecosystem has exemplified recently.
It is understandable that when inflation is above target, a central bank would rather err on the side of caution than techno-optimism. They know that being downbeat, and running policy a little too tight, means a faster return of inflation to target, which is no bad thing for an inflation-targeting central bank.
It matters, though, when those understandable assumptions are systematically one-sided. If there are arguments that productivity growth over the next few years will be even lower than the 0.7%yr 20-year average, we are yet to see them. In contrast, there are reasonable arguments for it to be a bit higher.
Similarly, the RBA’s participation rate assumptions are low given near-term demographic trends, and out of step with Treasury’s revised projections. Today’s labour market data for August shows the participation rate rising again, to just 12½ basis points below the January 2025 all-time high. This supports the idea that RBA’s view on participation trends is too downbeat. The IGR’s NAIRU (stable-inflation rate of unemployment) assumption is also a quarter-point or so below RBA estimates revealed in various FOI requests and the Governor’s comments this week.
Together, the three RBA assumptions imply weaker growth in supply capacity and thus more inflationary pressure for any given level of demand. If every one of your assumptions seems individually plausible, but all skew in the one direction, it starts to look like a pattern. The thumb on the scale is unlikely to be intentional, but it does suggest the likely direction of future surprises in the medium term.
