The Reserve Bank of Australia (RBA, the Bank) has had an inflation target roughly since 1993. Thirty years is long enough to judge it. This post does that with six charts, five of them built from my own models, and it reaches a verdict that is kinder to the bank before the GFC than after it. The bank has had two failures since then. They were different in kind, and only one of them was a choice.
Where the target came from
The target was not born into calm conditions. For nearly twenty years to 1990, Australian inflation averaged close to 10 per cent. It ran above 5 per cent in almost every quarter, peaked near 18 per cent in 1975 and was still above 10 per cent as late as 1983. The recession of 1990 to 1991 broke it, at great cost. Australia had paid that cost once already, in the recession of the early 1980s, and inflation had come back within three years. The target was the attempt to make sure it did not have to be paid a third time.
The chart shows underlying inflation, not headline, to reduce the quarter-to-quarter noise. Trimmed mean and weighted median tell much the same story, and this post uses underlying inflation throughout.
The chart starts in 1983 and shows the end of that era, the fall through the early 1990s, and what followed. Underlying inflation spent most of the next thirty years inside or close to the 2 to 3 per cent band, with two exceptions: the resources boom of 2008 and the breakout of 2022. Every judgment in this post should be read against the twenty years before the chart begins. The bank's first job was to make people believe that inflation would stay low. Its second job was not to give that belief away.
The cost was paid in jobs
Inflation fell fast. Unemployment did not.
The recession that broke inflation pushed unemployment to about 11 per cent in late 1992. It was still above 8 per cent five years later, did not get below 6 per cent until about 2003, and did not reach 4 per cent until 2008. Inflation was inside the band by 1993. The labour market took fifteen years.
But the chart shows something the raw rate does not. On my estimates the natural rate of unemployment, u* or NAIRU, was itself falling over those years, from around 10.7 per cent to about 5.3 per cent, and the unemployment rate tracked it closely all the way down. Most of the long recovery was structural. The economy was not running with slack for fifteen years. It was running close to a natural rate that was itself coming down.
The model identifies the fall. It does not say why it happened, and the why is complicated. The fall coincided with two decades of structural reform, and I think that is most of the story, but the chart cannot prove it and this post does not try. What the chart can say is that the unemployment rate consistent with stable inflation came down from near 11 per cent to about 5 per cent, and that shift, more than anything the Reserve Bank did, is what let unemployment fall for fifteen years while inflation stayed inside the band.
The line then stops falling. From 2008 the natural rate flattens at around 5 per cent and has drifted only a little since. What follows should be read with that in mind: the level of unemployment the economy can hold without generating inflation has not moved much in fifteen years, and redefining full employment does not move it either.
The model cannot say much about the cost of the disinflation itself. The first three years of the sample are shaded because u* is poorly identified there, and that is precisely when the cost was paid. The raw numbers are the only witness: 11 per cent unemployment, two years above 10 per cent, five years above 8 per cent. Whatever the natural rate was doing, that was a lot of people out of work for a long time, and it is the memory that shaped the next thirty years of policy.
Two later episodes matter more for the argument. From 2013 to 2019 unemployment sat above u* for six years. That is the slack the bank could not close, and it is why the demand component of inflation ran negative for most of that decade. And from 2022 the unemployment rate has been below u* and still was in Q2 2026, 4.35 per cent against 4.74 per cent. The labour market is tight, not loose. Keep that in mind when we get to the mandate.
How to judge a central bank
The method is simple. Inflation has causes, and the cash rate is only one of them. If you want to know whether policy was well set, you need to know three things: what was pushing inflation, where the neutral rate was, and how far the bank moved the cash rate relative to neutral. The next three charts answer those questions in turn.
What inflation is made of
The third chart decomposes quarterly inflation into five parts: the target itself, expectations sitting above target, demand pressure from the labour market, supply shocks from import prices and global supply chains, and noise. The model is a joint estimate of the natural rate of unemployment and potential output, so the demand component is measured against a moving benchmark rather than a fixed one.
Read left to right, the chart divides into four eras. In the mid 1990s expectations still sat above target, the purple band, and demand was modestly positive. From 1997 to about 2020 the purple band disappears. That is the target doing its job: people stopped expecting inflation above 2.5 per cent, and the bank got the credit for it. Demand, the orange band, ran positive from 2004 to 2008 and then negative for almost all of 2012 to 2021. Supply, the dark blue, spiked in 2020 and again in 2022. And in 2022 the purple band came back.
The chart makes one point that matters for everything that follows. The 2008 breakout was predominantly demand. The 2022 breakout was supply and demand together, with the demand part about the same size as 2008. The bank's response to those two demand shocks was very different.
Where the neutral rate was
The fourth chart estimates the nominal neutral cash rate, r*, two ways: from the bond market, as the world real rate plus an Australian wedge, and from the RBA's own reaction function. Both add long run inflation expectations to convert a real rate to a nominal one.
The two models agree on the shape. Neutral fell from about 6 per cent in the mid 1990s to around 4 per cent through the 2000s, then slid to about 2.2 per cent by 2020. With expectations near 2.5 per cent, that puts the real neutral rate at or below zero. It has since recovered to somewhere between 3.3 and 3.9 per cent, depending on the model.
This chart is why the cash rate on its own tells you nothing. A cash rate of 1.5 per cent in 2016 was barely easy because neutral had fallen to meet it. A cash rate of 4.35 per cent in 2023 was barely tight for the same reason.
How tight policy actually was
The fifth chart subtracts each estimate of neutral from the cash rate. Positive means tight, negative means easy. Against it I have plotted the inflation gap: trimmed mean inflation less the midpoint of the target, with the supply component from the third chart netted out. The green band marks a gap within 0.5pp of target, which I treat as close enough.
This is the scorecard. A well run cycle is one where the RBA's cash rate policy stance moves and the gap stays inside the band. A badly run cycle is one where the gap moves first and the stance chases it.
The pre-emptive strike of 1994
The bank's first test under the target was its cleanest pass. Through 1994 the RBA's policy stance went from about 1pp easy to 2pp tight. The inflation gap peaked at 0.6pp and came straight back. The bank moved before the gap did, and inflation never got going. When the gap turned negative in 1996 the bank eased, and the cycle closed with expectations lower than they had started. This is what the target was meant to look like.
The long grind to 2008
The resources boom gave the bank its longest test. From 2002 the RBA's policy stance rose almost without interruption for six years, from near zero to 3.8pp tight on the bond market measure and 3.1pp on the reaction function measure. Through most of that period the inflation gap sat between 0.2pp and 0.7pp. High side of target, but inside the band.
That flat gap is consistent with the tightening doing its job. Demand was building the whole time, the orange band in the third chart shows it, and underlying inflation stayed contained until the final stage of the boom.
The cycle failed only at the very end. In 2007 to 2008 the gap jumped from 0.4pp to 2.4pp in about a year while the stance was already at its highest level in the sample. Whether the bank ran out of room or the terms of trade shock was simply too big to offset, the result was the same: a breakout after six years of holding the line. Then the GFC arrived and the question became moot. The stance fell from nearly 4pp tight to 1pp easy inside twelve months, the sharpest reversal on the chart, and it happened while the gap was still above 2pp. That was the right call. It is also the last unambiguous one.
A decade with no room to move
The years from 2011 to 2021 look like a failure on the fifth chart. The stance eased steadily, the gap drifted below the band from 2016 and stayed there, and the bank never closed it. Demand ran negative for nearly a decade and unemployment sat above the natural rate for six years of it.
But the fourth chart explains why. Neutral was falling almost as fast as the cash rate. The bank kept cutting, but neutral kept falling underneath it, so much of the apparent easing never translated into easier policy relative to neutral. The stance never got more than about 1.5pp easy before 2020. From 2016 the cash rate was 1.5 per cent and the floor was close. The bank rationed its last few cuts because it could see the end of them.
This was a failure of the framework, not of judgment. A 2 to 3 per cent target with the cash rate as the main tool is very hard to hit when the real neutral rate is near or below zero. Whether the bank should have gone to unconventional policy earlier is a fair question. But the constraint was global r*, and the RBA did not set it.
When everything went wrong: the pandemic
The pandemic period is different. Here the bank had room, and the errors were its own.
The chain ran like this. In November 2020, the bank forecast a long stretch of weak inflation, a reasonable prior after a decade of undershooting, but a bad call on the recovery. On that forecast it said, repeatedly, that it did not expect the conditions for a rate rise to be met before 2024. The bank called that a forecast. The market and the public heard a commitment, and the bank did nothing to correct them until it was too late. To make the message credible it put a target on the April 2024 bond. When the data turned in late 2021 the bond target became indefensible, the market broke it in a matter of days, and the bank abandoned it in November 2021 without the orderly exit it had promised. That left forward guidance nobody believed and a market pricing the hikes the bank had said would not come.
The fifth chart shows the cost. The gap went from nearly 2pp negative to more than 2pp positive before the RBA's stance had crossed zero. The bank was roughly a year behind, and when it moved in May 2022 it was following the market rather than leading it.
The third chart shows the deeper cost. The purple band came back. Expectations rose above target for the first time since the 1990s and, on my estimates, they have not fully returned. The model shows the return and its timing, not its cause, and global inflation alone would have moved expectations somewhat. But the component that had been absent for two decades came back in the year the bank's guidance collapsed, and I think that is not a coincidence. Whatever the cause, it is the credibility the 1990s bought, and it is being spent.
The mandate and the shallow peak
The RBA Review reported in 2023 and recommended a clearer dual mandate, putting full employment alongside price stability. The government broadly adopted that framework. Whatever the merits of that as a long run principle, it arrived at an awkward moment: a bank that had just lost credibility, facing an inflation gap of more than 2pp, was being asked to put greater emphasis on the other side of the ledger. The Review was built to fix the 2010s. It was delivered into the 2020s.
The mandate change came with a message. Unemployment had fallen to about 3.5 per cent in 2022, its lowest in fifty years, and the government made a great deal of it. The Employment White Paper of 2023 then defined full employment as a state in which "everyone who wants a job should be able to find one without having to search for too long", and said that the NAIRU "should not be confused with, nor constrain, longer-term policy objectives". It put no number on the definition. My reading of the message is that the rate the economy had reached was being treated as closer to full employment than the RBA's estimate of 4.5 per cent, and that policy should not throw it away lightly.
The White Paper's distinction is defensible as far as it goes. The NAIRU is the short run constraint, and structural policy should try to lower it over time. But aspiration is not capacity. Until something actually moves u*, monetary policy has to work with the economy it has, not the one the government would like it to be. On my estimates the natural rate was about 4.7 per cent then and still is. The rate the government was celebrating was more than 1pp below it, and the third chart shows the result: the demand band stayed positive through 2023 and 2024 while unemployment sat below u*.
What the bank did next is on the fifth chart. The stance peaked at 1.7pp tight on the bond market measure and 0.8pp on the reaction function measure. Set that against 2008, when a demand shock of similar size was met with a stance of 3.1pp to 3.8pp. The response was less than half as large, and it was held for less time. Then the bank started cutting in 2025 with underlying inflation still above the band. Go back to the first chart: both measures bottomed around 2.7 to 2.8 per cent in mid 2025, at the top of the band, and both have since turned back up. Inflation never got back to the midpoint before the easing began.
The charts cannot see the RBA Board's motives, and there are other explanations for a shallow cycle: the sensitivity of mortgaged households, fear of recession, the initial diagnosis that the shock was mostly supply. But a bank asked to weigh employment more heavily, in a year when the government was celebrating 3.5 per cent unemployment, did exactly what that instruction would predict, and it did so with unemployment below the natural rate the whole time. It was not trading inflation for jobs. It was being cautious with both sides of the ledger in its favour. I think the mandate was a contributing cause. Readers can weigh the alternatives.
Where that leaves us
On my decomposition, supply has faded to almost nothing. What remains above target is demand and expectations, and neither is something the bank can look through. The inflation gap on my measure is 1.08pp and rising. Unemployment in Q2 2026 was 4.35 per cent against a natural rate of 4.74 per cent. The RBA's policy stance is between 0.46pp and 1.02pp on the tight side, depending on which neutral rate you believe.
The bank's own reading has been catching up with mine. It raised the cash rate three times in the first half of 2026, to 4.35 per cent in May, and has held since. The August Statement put trimmed mean inflation at 3.6 per cent, attributed it to "ongoing economy-wide capacity pressures" as well as some pass-through of Middle East costs, judged the labour market "a little tighter than full employment", and described policy as "somewhat restrictive". Strip out the institutional language and the diagnosis is close to mine: domestic capacity pressure, a tight labour market, and a stance the bank thought was probably enough.
Its forecast showed what "probably enough" bought. Trimmed mean inflation left the band in 2022. On the August path, conditioned on market pricing that had less than one more rise in it, inflation stays above 3 per cent until early-to-mid 2027 and reaches the midpoint in 2028, six years after it left.
That lasted about a month. On 8 September the deputy governor said inflation is too high and asked whether the Board had done enough. The market now prices at least one more rise this year, possibly two, and a cash rate near 5 per cent by mid 2027. The first move may come at the late September meeting.
So where does that leave the verdict? A demand gap above 1pp, unemployment below the natural rate and a stance under 1pp: under the framework that ran from 1993 to 2021, that was a tightening signal. Under the current one it took until September to read it that way. The bank has had two failures since the GFC. The first was forced on it by a world where neutral fell below the point at which the target could be hit. The second was not. Neutral has recovered to nearly 4 per cent, the bank has room, and the excuse that covered the 2010s no longer applies.
The record says the RBA does well against demand shocks it believes in. It believed in 1994 and 2008. It did not believe in 2022, and it has been softly softly ever since. The 1990s taught Australia that credibility is expensive to buy. The 2020s have shown how quickly it can be put at risk. If demand pressure is still holding underlying inflation above target a year from now, it will be much harder to blame the constraint that excused the 2010s.
No comments:
Post a Comment