The Australian Economy: Q4 2025

Introduction

I run several Bayesian macroeconomic models of the Australian economy. They fit the data to the model structure and estimate how much each variable is contributing to the key unobservables – the natural rate of unemployment and the potential growth rate. What follows is what those models are currently telling me.

But first, let's quickly recap on the RBA's policy rate, the current state of Australian inflation, and the current unemployment rate.

The Speed Limit

The model puts the speed limit – the rate at which the economy can grow without generating above-target inflation – at somewhere between 1.8 and 2.2 per cent through the year. GDP grew at 2.6 per cent in Q4 2025. We are above it. Above-target inflation is the predictable result.

The major institutions sit in a similar range, though with some spread. Treasury's December 2025 MYEFO implies around 2.5 per cent, making it the most optimistic. The RBA sits closer to 2 per cent. ANZ and the OECD are in the middle at roughly 2.25–2.3 per cent. The IMF, Deloitte, Oxford Economics and KPMG cluster around 2–2.1 per cent. For reasons I will return to below, the optimistic end of these estimates deserves scepticism.

The Output Gap

How far is the economy from potential? The answer depends on whether you allow the Phillips curve slope to shift across regimes.                   

This chart compares two model variants. The simple model (orange line) uses a single Phillips curve slope across the full sample. The complex model (blue line) allows the slope to shift across three regimes: pre-GFC, post-GFC, and post-COVID. There are other differences in these models, but we will focus on the regime-switching here.

The difference is striking. The complex model estimates the output gap at around 0.3 per cent of potential GDP – the economy is running hot. The simple model puts it near zero.

Why the divergence? The complex model's regime-switching Phillips curve steepens post-COVID, meaning even a modest positive output gap generates meaningful inflationary pressure. The simple model, forced to average across regimes, estimates a flatter curve. A flatter curve needs a larger gap to explain the same inflation – so when inflation moderates, it infers the gap has closed.

There is a deeper point here. The flatter Phillips curve estimated through the 2010s allowed the economy to run near capacity on significantly lower interest rates without triggering inflation. That was the post-GFC world of low rates and missing inflation. The regime shift following the pandemic means lower rates are now more likely to produce an output gap and inflation. The neutral rate implications follow directly. The post-GFC era of low interest rates is over.

What Is Driving Inflation

In Q3 and Q4 2025, trimmed mean inflation ran at around 4 per cent annualised – well above the 2–3 per cent target band. The model's decomposition of that excess is uncomfortable reading.

Supply-side factors explain a negligible share. Demand-side factors explain a larger share. But roughly half of the above-target inflation falls into the regression residual. The model cannot account for it.

That residual could be housing wealth effects, a fiscal impulse from government spending, immigration-driven demand that shows up in prices before it registers in the labour market, or population growth running ahead of supply capacity. Any or all of the above, or something else entirely. The honest answer is: we don't know.

This matters because the RBA's instrument – the cash rate – only works on demand-driven inflation. If half the inflation is coming from somewhere the model cannot identify, the RBA cannot confidently claim its policy settings are calibrated correctly. That is a reason for caution, not comfort.

The NAIRU

The model puts the non-accelerating inflation rate of unemployment – the lowest unemployment rate consistent with inflation at target – at most likely between 4.6 and 4.8 per cent.

The institutional estimates generally sit below this range. Treasury has the NAIRU at 4.25 per cent, a figure the Treasury Secretary stated publicly in late 2024. The IMF lands at around 4.2 per cent, though it acknowledges a wide modelling range of 3.9 to 4.5 per cent. CBA estimates it is around 4.4 per cent. The RBA's published working assumption has been around 4.5 per cent, and Zac Gross's 2025 peer-reviewed work defends that benchmark. The partial exception is the RBA's internal model estimates released under FOI, which was around 4.75 per cent – closer to this analysis than the RBA's own published figure.

The gap matters. If the true NAIRU is 4.7 per cent and Treasury is working with 4.25 per cent, they are misreading how much slack the labour market actually has. 

The Neutral Interest Rate

In standard macroeconomic theory, the long-run neutral nominal rate is roughly equal to potential growth plus the inflation target – around 4.5 per cent in Australia's case. Since the global financial crisis, central banks held rates well below this for over a decade without generating persistent inflation. Global quantitative easing (QE) distorted the picture (even in Australia, which was a late comer to QE).

But QE is now being unwound through quantitative tightening (QT). Global rates are back well above the zero lower bound. We are closer to a conventional interest rate environment than at any point in nearly twenty years. That reopens the question of whether the current policy rate of 3.85 per cent is actually restrictive – or whether it is closer to neutral, or even accommodative. My intuition is that policy is still inappropriately accommodative. 

With hindsight, the rate cuts in the first half of 2025 look ill-judged. Inflation moved back up. Unemployment moved below the NAIRU. The data are consistent with policy easing into an already-tight economy. The RBA is now reversing cuts it should not have made.

The Taylor Rule, an instrument that fell out of favour in the low interest rate world after the GFC suggests policy rates should be set at five and a quarter per cent to arrest the current inflation outbreak.

The Real Problem: Multi-Factor Productivity

At the bottom of the garden, the persistent failure of multi-factor productivity growth is doing quiet but serious damage to everything discussed above. Australia's MFP level is roughly where it was twenty-five years ago; stagnant – flat over two decades.

For the speed limit: potential GDP growth is the sum of labour force growth, capital accumulation, and MFP. If MFP contributes nothing – or drags – the speed limit depends entirely on workforce growth and investment. Workforce growth is slowing as immigration normalises and participation plateaus. Capital deepening has diminishing returns. An economy in this position has a structurally lower speed limit than one where MFP is pulling its weight. The 2–2.5 per cent estimates from the major institutions may themselves be optimistic if MFP remains negative.

For the NAIRU: if productivity gains are coming from capital deepening rather than genuine efficiency improvements, the cost of those gains is embedded in the capital stock. Firms pay for productivity through investment rather than getting it through better organisation or technology. That means the economy can sustain less real wages growth before it becomes inflationary. The NAIRU is higher than it would otherwise be. This is structural, not cyclical.

There is also a fiscal dimension. Budget projections that assume nominal GDP will compound at a reasonable rate over the forward estimates are implicitly assuming productivity will do some of the heavy lifting. If MFP is not delivering, both the speed limit and the fiscal arithmetic are quietly overstated. Treasury's 2.5 per cent potential growth assumption embeds a productivity contribution the data do not support.

Technical notes

Model specification

The complex model is a Bayesian state-space system estimated on Australian quarterly data. The key equations in the model are as follows:

Student-t random walk for the NAIRU
  NAIRU_t = NAIRU_{t-1} + ε_t,  ε ~ StudentT(ν=4, σ)

Potential output – Cobb-Douglas production function
  Y*_t = Y*_{t-1} + α·ΔK + (1-α)·ΔL + ΔMFP + ε

Okun's Law (gap-to-gap with persistence)
  U_gap = τ₂·U_gap_{-1} + τ₁·Y_gap + ε

Phillips Curve for price inflation (regime-switching)
  π = quarterly(π_exp) + γ_regime·u_gap + ρ·Δ4ρm + ξ·GSCPI² + ε

Wage Phillips curve (regime-switching)
  Δulc = α + π_exp + γ_regime·u_gap + λ·ΔU/U + ε

IS curve
  y_gap = ρ·y_gap_{-1} − β_is·r_gap_{-2} + γ_fi·fiscal_{-1} + ε

Hourly compensation Phillips curve (regime-switching)
  Δhcoe = α + π_exp + γ_regime·u_gap + λ·ΔU/U + ψ·MFP + ε

Participation rate
  Δpr = β_pr·(U_{-1} − NAIRU_{-1}) + ε

Employment – labour demand
  Δemp = α + β_ygap·output_gap + β_wage·real_wage_gap + ε

Exchange rate
  Δe = ρ·Δe_{-1} + β_r·r_gap_{-1} + ε

Import price pass-through
  Δ4ρm = β_pt·Δ4twi_{-1} + β_oil·Δ4oil_{-1} + ρ·Δ4ρm_{-1} + ε

Net exports
  Δ(NX/Y) = β₁·output_gap + β₂·Δtwi + ε

The NAIRU uses a Student-t innovation distribution (ν=4) rather than Gaussian to accommodate occasional larger structural shifts without contaminating the smooth underlying trend. The Phillips curve and wage equations use regime-switching on the slope coefficient to allow the inflation-unemployment trade-off to vary across different inflation regimes.