LONDON, Feb 27 (Reuters) - An AI productivity boom, if it materialises, may help buy major economies more time to clean up their strained public finances, economists say, although it won't do the heavy lifting.
The stakes couldn't be higher. Debt is above 100% of output across most rich economies and set to rise given ageing population costs, interest bills and pressure to spend more on defence and climate change.
U.S. policymakers are already optimistic on AI-driven growth and economists say the technology
has the potential to shake the world out of a post-2008 productivity slump by boosting workers' efficiency and freeing them to focus on more productive tasks.
Higher economic growth would make government spending and debt loads potentially more manageable and help fend off scrutiny from bond vigilantes.
To sketch the impact on public finances if AI does boost labour productivity over the long term, the OECD and three high- profile economists shared early estimates with Reuters.
Filiz Unsal, the OECD's deputy director of economic policy and research, said an AI productivity surge, were it to increase employment, would lower debt across OECD countries, from the U.S. to Germany and Japan, by 10 percentage points from the roughly 150% of output the organisation expects in 2036.
That would still be a sharp rise from 110% currently.
Much will depend on whether job creation eventually outweighs any job losses from automation, as well as whether firms pass on higher profits by raising wages and how governments manage their overall spending.
In the United States, two of the other economists projected debt rising more slowly to roughly 120% over the next decade from around 100% of output now in their best-case scenarios. One saw little change.
"Productivity is like magic... It helps the fiscal dynamics dramatically," said Idanna Appio, one of the economists, who was previously at the New York Federal Reserve and is now a fund manager at First Eagle Investment Management.
"But our fiscal problems are well beyond what productivity can fix," Appio said.
DEMOGRAPHICS TO LIMIT AI IMPACT
For now, ratings agency S&P assumes no major public finance impact by the end of the decade.
"The one (path) that the (U.S.) administration is hoping for would be you get saved by the bell," said Mark Patrick, head of macro and country risk at Teachers Insurance and Annuity Association of America, but added: that isn't "something we can set our clocks by."
The economists did not provide estimates for other countries. However, AI could boost productivity in Britain in line with the U.S., but by half as much in Italy and Japan due to lower adoption rates and smaller sectors that could benefit from AI, OECD research has found.
Ultimately, fiscal dynamics will determine how much AI-driven productivity can counterbalance rising debt. Demographics is the biggest challenge.
"The root of the debt issue is with ageing demographics and the entitlements that are tied to that," said Kevin Khang, who heads global economic research at Vanguard, the world's second- largest asset manager.
Addressing it "requires getting the fiscal house in order and (AI is) just buying us the time," he said.
Khang sees a scenario where AI boosts U.S. growth to average 3% through 2040 as most likely. The Fed sees potential growth around 2%.
He estimates higher growth and tax revenues would slow U.S. debt growth to around 120% of output by the late 2030s. That's much lower than the 180% he foresees - higher than others - if AI disappoints, growth slows and market pressure raises borrowing costs.
Bond investors have been quick to punish governments for fiscal largesse ever since bond yields rose sharply after the pandemic across rich economies.
Appio said declining U.S. immigration added to the demographic challenge.
"The labour shock offsets any (AI) productivity growth," she said, but added she would be much more worried without AI.
UNCERTAINTY AROUND TAX, SPENDING
Economy-wide productivity gains should raise revenues. But if AI reduces employment or competition, and profits and capital - often taxed less than labour - see most benefit, revenues could disappoint.
On the spending side, public sector efficiency gains could help costs, but the risk is spending rising alongside growth.
That's why Kent Smetters, director of the University of Pennsylvania's Penn Wharton Budget Model analysis group, expects a tiny impact on U.S. debt in a decade.
Even if growth is higher than he currently expects, that would still have little impact on curbing social security, which makes up a fifth of federal spending, because claims are indexed to average wages, Smetters said. Other labour costs the government covers would also rise if productivity raises private sector wages, he added.
"It's very important to see whether wages are going to increase," the OECD's Unsal said, adding that wage growth was more likely if AI doesn't increase employment.
There are of course debt costs, which will depend on whether productivity raises real interest rates, a debate already emerging at the Fed, and how long growth outpaces any rise, economists said.
Clearly, they don't have crystal balls. A shock could quickly upend this debate.
A recession could mean "the AI boom may not come quick enough before the market gets nervous about the fiscal trajectory," said Barclays' global head of economics research Christian Keller.
(Reporting by Yoruk Bahceli, additional reporting by Rodrigo Campos in New York; Editing by Dhara Ranasinghe and Susan Fenton)
