The question of blame
Last week the RBA's Deputy Governor, Andrew Hauser, sounded a distinctly hawkish note. On X, Cameron Kusher asked whether the RBA had reflected on its narrow path and realised it had made things worse for longer. I said you can't blame the bank for following the direction the government gave it. Zac Gross replied that the whole point of an independent central bank is that you blame it only for interest rate decisions.
That deserves a proper answer. Independence covers the instrument, not the objective. The government sets the objective. And on this occasion, the government got the objective wrong twice, then pushed hard on a bad estimate.
Where the narrow path began
The bank was too tight in 2016 to 2019. In September 2022, Gross and Andrew Leigh published a paper using the RBA's MARTIN model to compare actual policy with an optimal simple rule. They found the cash rate was too high in 2016 to 2019, holding inflation below the band and costing jobs. Leigh was by then a minister in the government that had just commissioned the RBA Review.
Gross and Leigh put it down to financial stability: the bank held rates higher than inflation and unemployment warranted because it was worried about house prices and household debt. They make a good case. Australia had the highest policy rate among its peers, so they argue the zero lower bound was not the binding constraint. Even so, I suspect the bank was also nervous about the lower bound.
The paper supplied the diagnosis. The 2023 Review and the new Statement on the Conduct of Monetary Policy supplied the prescription: full employment as an objective in its own right, with the speed of return to target to be balanced against it.
Full employment has been in the Reserve Bank Act since 1959. What changed is its weight. Under inflation targeting, price stability came first, as the way to deliver jobs over time. Since 2023, employment has been an objective in its own right, to be balanced against inflation in the short run. Same words in the Act, very different instructions to the bank.
The 2 to 3 per cent target has not changed. What changed is how hard the RBA pushes toward it. The Statement fixes the destination: the bank sets policy so inflation is expected to return to the midpoint of the target. It leaves the speed open. The timeframe depends on economic circumstances and should, where necessary, balance the price stability and full employment objectives.
A trade-off written into the brief
Having two goals is not the problem. Full employment has always been one of them. The problem is telling the bank to balance jobs against inflation in the short run. That builds in a conflict whenever inflation is above target. Then every step toward the target costs jobs, and the jobs leg tells the bank to ease off. When inflation is below target, the two goals point the same way and there is nothing to balance. When they conflict, the bank splits the difference. That is the narrow path. The destination stayed the same. The slow speed of travel was set by the mandate.
The numbers show it. The RBA peaked at 4.35 per cent in November 2023. The Fed reached 5.25 to 5.5 per cent, the Bank of England 5.25 per cent and the RBNZ 5.5 per cent, with broadly similar inflation peaks. The RBA stopped short, eased in 2025, and inflation never settled at the midpoint.
The trade-off between jobs and inflation only exists in the short run. The reason is expectations. When demand pushes unemployment below u*, firms struggle to hire and workers gain bargaining power, so wages and prices rise faster than people expected. For a while, that surprise lets unemployment sit low. But people learn. Wage claims and price setting start to build in the higher inflation. To hold unemployment below u*, the bank would have to keep surprising them, with inflation rising ever faster. That cannot last. Eventually unemployment returns to u*, which is set by structural factors such as wage setting, skills and how well workers match to jobs, not by monetary policy. What remains is the higher inflation.
This is the Friedman and Phelps insight from the late 1960s, and the 1970s proved it. Monetary policy can move unemployment for a time. It cannot hold it below u* for good. Sitting on above-target inflation does not buy lasting jobs. It raises inflation expectations, and if those become entrenched, breaking them takes a recession.
That is what happened from the mid-1970s to the early 1990s. Each recession left more people in long-term unemployment, and the jobless rate ratcheted higher. Inflation can fall fast, but the labour market takes years to heal. Unemployment peaked near 11 per cent after the early-90s recession and took most of the decade to recover. Once the RBA began targeting inflation, it moved early and decisively, as in 1994, and the jobs recovery held.
And the inflation-first approach worked for jobs. Once the bank got on top of inflation in the 1990s, u* fell substantially, settling around 5 to 6 per cent by the late 2000s. Low, stable inflation is what let unemployment fall and stay down. Balancing jobs against inflation puts that at risk to protect jobs in the short run. Even the Fed accepts monetary policy can't set long-run employment. In practice, getting inflation right is how it serves the jobs side too.
The lesson is simple. Small moves early avoid big ones later. The jobs you protect at the start are the jobs you lose at the end.
This is not an argument for strict inflation targeting. The board should diagnose what is driving inflation and respond accordingly: look through a supply shock when inflation starts inside the band, lean against demand when it doesn't. Full employment, properly defined as the NAIRU, is the goal. Price stability is how you achieve and sustain it. What I object to is treating jobs as a short-run trade-off against inflation. That is what the current balancing clause invites.
By that rule, looking through the first wave of the post-pandemic inflation was defensible. In 2021 inflation started inside the band, and the shock was mostly supply: shipping, energy and goods. But the mix shifted. By the peak it was roughly half supply and half demand, and by 2024 it was almost entirely a domestic story, led by services. The error was not looking through the supply shock at the start. It was staying gentle after the shock had stopped being about supply, and after inflation had spent years outside the band. The same is true now. The rise in trimmed mean through 2026 is domestic, not the Iran war’s oil premium.
The loss function behind it
The Statement says "balance" without saying how. The Gross and Leigh paper supplies the answer. To find the optimal path, they minimised a quadratic loss function: the squared gap between inflation and 2.5 per cent, plus the squared gap between unemployment and the NAIRU, plus a penalty on changing the cash rate. Inflation and unemployment each got a weight of one. Smoothing got a half. They justified the equal weights by citing Janet Yellen in 2012 on how the Fed reads its dual mandate. It is the natural way to read the brief.
That function describes the narrow path. Trade a bit more inflation for a bit less unemployment, every quarter, as if each trade were temporary. The bank did not need to adopt the formula. It only needed to behave as the formula prescribes, and a balancing mandate invites exactly that.
In the model, each trade really is temporary. MARTIN treats inflation expectations as exogenous. They enter the wage and trimmed mean equations, but they sit where the modeller puts them and do not respond to policy. In long-run simulations they converge to 2.5 per cent. The NAIRU is exogenous too. So in MARTIN, a period of above-target inflation can never shift expectations, and it can never raise u*. The trade-off can be made again and again at no lasting cost. That is the logic error built into the lens. A short-run trade-off is treated as one you can keep making. In the real economy, the trades add up. Expectations drift, and the short-run trade-off becomes a long-run cost.
Quoting Yellen in 2012 is also a stretch. Each Fed chair and RBA governor has to be read against the era and the problems they faced. Yellen spoke at the zero lower bound, with unemployment near 8 per cent and inflation below target. Both goals pointed the same way, so equal weights cost nothing. The same was true in Australia in 2016 to 2019. A mandate has to work when the goals conflict.
Right diagnosis, wrong prescription
The Gross and Leigh diagnosis implied a narrow fix. If the error was leaning against house prices, the remedy was to stop doing that. Leave household debt to APRA. Make the target symmetric in practice, so running below the band is treated as seriously as running above it. Neither weakens the bank's grip on inflation when inflation is high. Nothing in the diagnosis required changing the weights. A dual-mandate loss function that began as a modelling lens became the way to read the brief.
A mandate is not a patch. It applies in every condition, at the zero lower bound and in an inflation breakout alike. Good rules are built for the hardest case, not the last one. The 2016 to 2019 problem belonged to a regime of low inflation and rates near zero. The cure applied everywhere, and the first real test it faced was the inflation surge of 2022. It failed that test.
Two definitions of full employment
The Statement then muddied the target itself. It carries two definitions of full employment, side by side. The government's objective is aspirational: everyone who wants a job can find one without searching for too long. The bank's is the NAIRU: the maximum level of employment consistent with low and stable inflation.
A framework document should not define its central term twice. The ambiguity invites the looser reading, and Peter Tulip has argued that is what happened: the aspirational definition leaked into the bank's target and pulled it below the sustainable rate. I agree. The NAIRU is the right benchmark, and it should be the only one.
A wishful view of the NAIRU
Treasury moved its NAIRU assumption to 4.25 per cent just as inflation was peaking, and the government's employment agenda leaned on the idea that a lower jobless rate was sustainable.
This is not hindsight. Even at the time, estimates in the low 4s looked wishful to a number of observers, Tulip and me included. A year ago I had u* between 4.5 and 5 per cent.
My models put u* between 4.6 and 4.9 per cent across four specifications. Unemployment fell below every one of them in 2021 and has stayed there since. At the 2022 low it sat more than a point under. That is a labour market running hot for four years, under a mandate that told the bank to protect it. Through 2025, with u* between 4.6 and 4.9 and unemployment around 4.2, there was no case for the cuts the bank delivered. Gross's own NAIRU tracker puts the NAIRU near 4.9 per cent currently.
The same chart shows the other side. In 2016 to 2019, unemployment sat above u*. The bank was too tight then. On that much, Gross and Leigh were right.
The chart above runs to the June quarter. Monthly unemployment has since risen to around 4.7 per cent.
Three errors pointing one way
Put a balancing mandate together with a loose definition of full employment and an optimistic NAIRU, and every signal says go gently. If full employment is around 4 per cent, unemployment in the high 3s looks like balance, not overheating. All three errors pointed the same way, and inflation got the benefit of the doubt for too long.
The government also pushed hard for the jobs-protecting reading. In September 2024 the Treasurer said rate rises had smashed the economy, and framed the shared objective as getting on top of inflation without smashing an economy that was already weak. Bullock held the line that year, warning that entrenched inflation would eventually require disinflation with lasting costs through higher unemployment. The cuts came in 2025, under the new Monetary Policy Board and the new Statement.
The bill comes due
The latest monthly figure puts unemployment around 4.7 per cent, inside the range of my u* estimates. So the labour market is roughly back in balance, yet trimmed mean inflation is still around 3.6 per cent. That is what entrenched expectations look like: inflation that persists without a tight labour market to drive it.
Seven likely objections
A dual mandate is best practice. The error was the definition. This is Tulip's view, and he is right on both counts. Two goals are fine, provided the bank is not expected to balance one against the other in the short run. The double definition of full employment was a real error, and both Tulip and I called NAIRU estimates in the low 4s wishful at the time. But fixing the definition does not fix the balancing clause. Get u* right and the bank still faces a choice whenever inflation is above target. That is now: unemployment near u*, trimmed mean at 3.6 per cent. Getting back to 2.5 per cent inflation means unemployment above u* for a while. A brief that treats jobs as a short-run trade-off against inflation tells the bank to go easy on exactly that. Jobs are the goal. Price stability is how you get there.
The narrow path was a communication tool. Zac's reply on X was that the phrase did not change the target. Agreed. But a communication tool for what? The narrow path told us the bank would bring inflation down slowly, and only eventually, to protect jobs. That is the equal-weight loss function in action: trade the inflation gap against the unemployment gap, quarter by quarter, as if each trade were free. The phrase did not change the target. It told us how the bank was weighing the trade-off.
The trade-off is short run, so the mandate is harmless. If price stability and full employment don't conflict over time, the argument goes, a balancing clause costs nothing in the long run. But the short run is where the mandate does its damage. It invites the bank to compromise: go slow, protect jobs, let inflation sit a little high for a little longer. Then expectations drift, and the long run you were promised arrives with a higher u* and a recession attached. Letting inflation fester is how a short-run choice becomes a long-run cost.
The Fed balances its goals, so why shouldn’t the RBA? The Fed's balanced approach is largely window dressing. It has long held that monetary policy cannot set the long-run level of employment, so in practice it targets inflation and calls the result maximum employment. And the one time it leaned into the jobs side, with its 2020 framework and its focus on employment shortfalls, it fell behind on inflation and then had to hike harder than anyone. That rather makes my point. In August 2025 the Fed abandoned both flexible average inflation targeting and the shortfalls language, returning to something close to its 2012 framework. The 2020 changes were a fix for the zero lower bound, the same regime-specific problem that shaped the Australian mandate. The Fed recognised its fix had misfired and reversed it. Australia has not.
The bank was too tight in 2016 to 2019. Agreed, but the cause matters. A bank worried about house prices needs to leave household debt to APRA and treat the target symmetrically. It does not need a permanent change to its objective. The diagnosis was right. The prescription was wrong.
Inflation is coming down anyway. Trimmed mean is still around 3.6 per cent, years after it first left the band. That is not a slow path working. It is a slow path running out of road. And the RBA burned credibility on the narrow path. It now has to talk an even tougher game to be taken seriously, or expectations will drift. That is why it is talking tough.
A correctly specified loss function would have tightened harder. With the NAIRU near 4.9, as Gross’s own tracker has it, both gaps said tighten in 2022 to 2024: inflation well above target, unemployment well below u*. Agreed. A loss function with the right NAIRU would have done better than the bank did. But that is why the three errors matter together. Plug in a NAIRU in the low 4s and the jobs gap shrinks, leaving the inflation gap to be traded off against a labour market that looked merely balanced. The balancing frame was only as good as the estimate fed into it, and the estimate was wishful.
Blame the brief
Blame the bank for its calls, by all means. But the government wrote the brief, and it got the objective wrong twice. It built a permanent short-run trade-off into the bank’s objectives to fix a no-longer-relevant, regime-specific problem. It muddied the definition of full employment, setting an aspirational goal beside the NAIRU. Then it pushed hard on a bad estimate, an optimistic view of where the NAIRU sat, and handed the package to the bank just as inflation broke out.
No comments:
Post a Comment