Three questions for the advocates
Stephen Kirchner wrote last week about the Greenspan rule and the RBA, building on a post by Lars Christensen. I could only read the teaser from Kirchner, but the rule is built on nominal GDP targeting: the target is the inflation target plus an assumed trend growth rate. The exchange that followed on X sharpened my earlier view.
nGDP targeting comes in two flavours. Some argue the level of nominal GDP is what should be targeted: set a path, and return to it after any miss. Others argue for the growth rate, and let past misses go. Most academic advocates prefer the level. The Greenspan rule is a growth rule.
Advocates promise that an nGDP target protects jobs better than an inflation target, and that it leaves the central bank with less discretion. They also present it as the simpler regime. So there are three questions. Does it protect jobs? Does it reduce discretion? Is it simpler? I am not convinced.
How the two regimes work
Inflation targeting works like this. The Bank aims for inflation of 2 to 3 per cent, with 2.5 as the midpoint. When inflation looks set to run above that, it raises the cash rate to slow demand. When inflation looks set to run below, it cuts. To judge where inflation is heading, it looks at how much slack the economy has: unemployment against u*, and output against potential. The target is a fixed number. Real growth is whatever the economy delivers.
nGDP targeting works like this. Nominal GDP is the dollar value of everything the economy produces, so its growth is real growth plus inflation. The bank picks a target for that sum, say 5 per cent, built from 2.5 per cent inflation and 2.5 per cent assumed trend growth. When nominal growth runs above the target, it raises rates. When nominal growth runs below, it cuts. The target does not depend on how the total divides. Real growth of 2.5 with inflation of 2.5 is on target. So is real growth of 1 with inflation of 4.
Under a growth target, each year starts afresh. Under a level target, the bank must return nominal GDP to its path, so a weak year is followed by catch-up and a strong year by restraint. Either can be set for total nGDP or for nGDP per capita.
Mostly the same regime
The two regimes are more alike than the debate suggests. Typically, when demand runs hot, inflation and nominal income both rise, and both regimes call for tightening. When a supply shock hits and is expected to pass, nGDP targeting ignores it. So does flexible inflation targeting. The RBA looks through supply shocks. The textbook position for New Keynesian policy is to look through a supply shock if expectations are well anchored, and the supply shock is reasonably expected to be transitory. There is no doubling down on output loss.
The regimes part in one case: a lasting shift in productivity. Under inflation targeting, what is kept constant is the rate of inflation, and real growth takes the variation. Under nGDP targeting, nominal growth is kept constant and inflation takes the variation.
Jobs that never arrive
The main employment promise rests on that one case. If productivity growth falls and the bank holds nominal spending to its old path, the argument goes, demand is protected and so are jobs.
In the long run this cannot work. A lasting fall in productivity is a fall in potential. Holding nominal demand to the old path recovers no output. It changes the split between real growth and prices. The long-run Phillips curve is vertical, so unemployment returns to u* either way. What remains is the higher inflation.
The better claim is about the adjustment. Wages are sticky downwards, so if real wages have to fall, it is easier to let inflation do it than to cut pay. That matters after a sudden shock, and flexible inflation targeting already allows for it by looking through the first round of price rises. It matters much less for a slow decline in productivity growth. With inflation at 2.5 per cent, wages can keep rising in dollar terms even when productivity is flat. Nobody's pay has to be cut. The insurance nGDP targeting offers is one that inflation targeting already carries, without the higher inflation afterwards.
Potential growth is growth in labour supply plus growth in productivity. Australian potential growth has slid from about 3.3 per cent in 2004 to about 2 per cent now, and most of that fall is productivity. A nominal growth target set in 2004 and left alone would by now mean inflation more than a point higher, with no extra jobs to show for it.
My models currently disagree about potential growth by about 0.2 percentage points. The fall since 2004 is about 1.3 points.
The bill comes later. Neither regime abolishes the business cycle. Shocks still arrive and forecasts are still wrong, so at some point nominal growth runs hot and has to be pushed back. Under a level target the overshoot has to be reversed as well. Pushing back on nGDP is the same thing as pushing back on inflation: rates held above neutral and unemployment above u*. It is harder work if inflation has been left to roam with productivity, because expectations have nothing fixed to return to. Advocates say expectations would anchor on nominal income instead. That is untested, and wage and price setters here respond to inflation. Inflation can fall fast, but the labour market takes years to heal.
The regime also assumes the public will accept whatever inflation the arithmetic delivers. Voters around the world punished governments for the post-pandemic inflation. The day a government decides 4 per cent is intolerable, the disinflation starts from expectations that were never anchored to an inflation rate.
Insurance against the wrong risk
The better version of the jobs argument is about the zero lower bound. With a fixed nGDP target, lower productivity growth means higher inflation, so nominal interest rates sit higher on average and the bank has more room to cut before it reaches zero. A level target adds a promise: any shortfall in nominal income will be made up, so policy stays loose after a slump. Neither gives the bank a new tool at zero. The first is a higher inflation target by another route. The second works through expectations, and only if the promise to overshoot later is believed. It is also separate from nGDP. A bank can promise catch-up on prices alone, as the Fed did with average inflation targeting from 2020 to 2025. The zero lower bound is a challenge for any regime. It is Keynes's liquidity trap.
Michael Woodford, who shared this year's Nobel prize, proposed a nominal GDP level path in 2012 as a way to commit to catch-up at the zero lower bound. He also said the path should be adjusted for changes in potential.
There is a point here. Zac Gross and Andrew Leigh showed that policy was too tight in 2016 to 2019. The Bank did itself no favours in the second half of the 2010s. The cash rate was 1.5 per cent and inflation was below the band. The inflation target already said cut. The Bank did not follow its own target. Advocates need to explain why it would have followed a different one more faithfully.
And nGDP targeting ties the extra inflation to productivity, not to the neutral rate. In theory the two move together. In practice they have come apart. In the 2010s the neutral rate was at its lowest, while trend productivity growth was still around 1 per cent. The rule would have added little inflation when the lower bound was closest. Today productivity growth is near 0.3 per cent on the RBA's numbers, the Governor said in September that there is a general consensus the neutral rate has risen, and the cash rate has been 4.85 per cent since 3 November. The rule would deliver the extra inflation now, when it is least needed. We are not fighting yesterday's war.
What nominal GDP was signalling
Take a growth target of 5 per cent, or 3.5 per cent per head. It would not have shown the 2016 to 2019 problem. Export prices are counted in nominal GDP but not in the CPI, so a rise in iron ore and coal prices lifts nominal income without lifting consumer inflation. That is what happened. Nominal growth ran between about 3.5 and 8 per cent in those years while inflation sat below the band. On the observed growth rates, the signal was to hold.
The post-pandemic years are worse. In 2023 and 2024 nominal GDP per capita grew by between 1 and 2 per cent a year, because real output per head was falling and the terms of trade were coming off. Underlying inflation was between 4 and 6 per cent. On those figures, a per capita nGDP rule pointed to rate cuts. Today nominal growth is 5 per cent and 3.5 per cent per head. The rule says hold, with trimmed mean inflation at 3.7 per cent.
To be fair, a growth rule gets one right. Nominal growth was above 10 per cent from mid-2021, and it would have tightened well before May 2022. But the signal came from the rebound off 2020 and from iron ore above US$200 a tonne, while Sydney and Melbourne were in lockdown and underlying inflation was below the band. It is hard to see any board raising rates into a lockdown on the strength of iron ore prices. The Board still needs to judge the world it is in.
My suspicion is that an nGDP rule would not ease into its policy settings. It would jump between the brake and the accelerator. Since 2014 the nominal numbers have said ease, then tighten in 2017, then brake hard in 2021, then accelerate in 2023 with inflation above the band. Stop-go policy is no recipe for steady employment.
Set and forget
We can set an inflation target and forget it. We cannot set an nGDP target and forget it.
Australia's 2 to 3 per cent target has stood since 1993. Economists argue about whether that is the best number. But over that time trend productivity growth has moved by more than two points, and nobody had to reopen the target.
An nGDP target has trend growth inside it. When the trend moves, either the target moves or inflation does. Kirchner's reply to me was that the rule can include a time-varying g*, as any output gap measure would, and that it is hard to get around taking a view on trend growth. I agree with both. But a time-varying target adds discretion. Someone has to decide the new trend, and decide when to change it.
Who cuts the target
Productivity growth zigzags by several points a year around a trend that moves by about a tenth of a point. It takes the best part of a decade to tell a lower trend from a bad run. A ten-year average is five years stale when it arrives. In the meantime every lobby group will say give it another quarter to come good. Raising a nominal target would be easy. Lowering one would not.
The problem is visible today. The RBA assumes trend productivity growth of 0.7 per cent and potential growth a little above 2 per cent. Treasury assumes 1.2 per cent and 2½ per cent. In the May Budget, Treasury kept its 1.2 per cent and pushed out the date for reaching it from two years to five. Most commentators, me included, think g* is about 2 per cent. Over the past ten years productivity growth has averaged about 0.1 per cent.
If an nGDP target were set this year, Treasury's number gives 5 per cent and the Bank's gives 4.5. The target would be agreed between the Treasurer and the Bank, as the inflation target is now, and Treasury's number would be hard to keep out of it. If the truth is 2, the 5 per cent target delivers 3 per cent inflation and reports it as success.
Discretion moves to the target
So the discretion does not go away. It is reduced on following the target and increased on setting the target. That is the worse place for it. A rate decision is made in public, eight times a year, and can be reversed at the next meeting. A target is set rarely, under lobbying, and a stale one runs for years.
The error also lands differently. Under inflation targeting, a wrong view of g* means the Bank misjudges the stance and inflation misses. That miss is a failure against the target, and the Bank has to respond. Under an nGDP target, g* sits inside the target. A stale estimate arrives as inflation while nominal growth is on target. Everyone can see the inflation, but the bank has met its objective and nothing requires it to act.
More moving parts
Productivity is not the only choice hidden in the target. Should it be total nGDP or nGDP per capita? A total target has an assumed rate of population growth inside it, so its meaning shifts with the migration intake. A per capita target avoids that in principle. But migration is complicated, because it is a time-varying mix of students and workers. A worker adds full-time hours. A student works limited hours and counts the same in the denominator. In 2023 and 2024 the gap between the two measures widened from about 1.5 points to about 2.5. The choice of denominator would have changed the reading from mildly weak to deeply weak.
A level target also needs a path, and everything depends on it. The chart shows two, one growing at 4.2 per cent a year and one at 5.2. Against the first, nominal GDP was on target in 2016 to 2019 and is now about 12 per cent too high. Against the second, it was too low in 2016 to 2019 and is on target now. One point on the growth rate, about the size of the fall in potential growth, is the difference between easing and tightening. The pandemic makes the choice harder still. Keep the old path and the bank must run nominal growth below trend for years. Rebase and the whole price jump is ratified.
The data are a problem too. Inflation and unemployment are monthly. Nominal GDP is quarterly, about two months late, and revised for years. So the bank would steer by its own forecast of nGDP, as it steers by a forecast of inflation now. That forecast would be built from the monthly inflation and labour data. The simple rule ends up needing the same models and the same judgement.
Back to the three questions
Does nGDP targeting protect jobs? No. For demand shocks and passing supply shocks it does what flexible inflation targeting already does. When productivity falls for good, it delivers higher inflation and the same unemployment.
Does it reduce discretion? No. It moves discretion from setting the rate to setting the target, where decisions are rarer, more political and slower to correct.
Is it simpler? No. It needs a view on trend growth, a choice between level and growth, a choice between total and per capita, a starting point for the path, and a forecast of a number that arrives late and gets revised.
There is a useful idea in nGDP targeting. Nominal demand is worth watching as a cross-check. But as a regime it is inflation targeting plus a bet that trend productivity holds still. In Australia that bet has been losing for twenty years.




No comments:
Post a Comment