Universal Basic Income: Still a Utopian Response to AI

Rising unemployment among young people and recent graduates in North America has renewed fears over how artificial intelligence could disrupt the job market. And from there, it’s not that much of a leap to start talking about universal basic income (UBI). The assumption is that artificial intelligence is different from other revolutionary technologies. Unlike the steam engine or computers, it automates knowledge work, not just repetitive tasks.

The idea—notably espoused by the Big Tech leaders in the US—is that AI will be a revolutionary disrupter that eliminates jobs several times faster than it can create them. If that happens, any attempt to retrain millions of laid‑off workers would be futile, as replacement jobs just won’t be created. To survive, a market‑based system would need to offer some kind of guaranteed income, regardless of one’s employment status.

Of all the possible outcomes, this is obviously one of the worst. And even then, the conditions under which such a system would be economically sustainable still need to be set out. The most commonly proposed system, which would pay for that income with the productivity gains from AI, would take a massive effort to implement.

The Latest Incarnation of Solow’s Productivity Paradox

The worst‑case scenario is based on the idea that productivity gains are both substantial and broad‑based. But the data doesn’t back that up just yet. A recent National Bureau of Economic Research survey External link. of nearly 6,000 executives in the US, UK, Germany and Australia found that 89% of managers haven’t seen a measurable change in productivity over the past three years, despite rapid AI adoption. This echoes the paradox first expressed by Robert Solow back in 1987: “You can see the computer age everywhere but in the productivity statistics.”

Microeconomic studies do show efficiency gains ranging from 10% to 50% for specific tasks. These are some of the promising factors fuelling the stock market’s AI boom. But history has shown that converting localized gains into broader macroeconomic growth requires a drastic restructuring of production processes, the timing of which remains uncertain. One unusual aspect of the current situation is that the public sector—under heavy pressure to improve efficiency—may be among the first to reap the benefits. Yet given the challenges of measuring productivity in government settings, it’s little solace.

Four Conditions That Are Hard to Meet

But suppose those productivity gains finally do materialize. To pay for UBI, four conditions need to be met simultaneously.

The first is that productivity gains must be both large and lasting. It’s already clear that we can’t just take that as a given. Right now, designing a permanent framework to redistribute income based on a hypothetical fiscal windfall would be like building on sand.

The second condition is that the government must be able to tax those gains. But AI‑related productivity gains may materialize in ways that fall outside the traditional tax base. A company that has replaced its workforce with AI could domicile its profits in a favourable jurisdiction. Unlike factories or workers, digital capital is inherently mobile. It’s precisely because human labour is the least mobile factor of production that it has, by default, borne most of the tax burden in Western democracies.

The third condition is that the prisoner’s dilemma posed by international tax competition must be resolved. The OECD attempted to address this with a 15% global minimum tax on multinationals, but that’s still well below what a UBI program would require. This is especially true given the many exceptions to that tax, which just goes to show how difficult it is to take concerted action. Most importantly, international tax coordination efforts move at a glacial pace, while technological disruption tends to blaze ahead at lightning speed.

Here we see the same asymmetry that hampers climate policy: The problem is global, but the policy tools used to address it are fragmented and constrained by national borders. Finland introduced the first carbon tax in 1990. Thirty‑six years later, only a quarter of the world’s emissions are covered by carbon pricing, with rates that vary from one jurisdiction to the next. How long will it take to agree on how to properly tax intellectual property that can move freely across jurisdictions?

And then there’s the fourth condition, which is perhaps the most overlooked: The productivity gains must not be entirely absorbed by the disruption itself. Let’s say AI does generate a net economic surplus that can be redistributed. If that surplus is mostly used to compensate the affected workers, the challenge currently facing Western societies—namely the sustainability of the welfare state—will remain unresolved.

In fact, it could get even worse. For example, even if all four conditions are met and a UBI can be introduced, the circumstances could subsequently change: Mobility, optimization or other factors could shrink the tax base just as spending pressures start to intensify. This would be a nightmare scenario for governments that are already on shaky fiscal ground—weakened by pandemic debt, aging populations and the sudden need to ramp up defence spending, while remaining caught between rising public expectations and diminishing fiscal capacity.

Finally, we’re left with the question that even the biggest supporters of UBI have yet to answer: Can governments design a tax system robust enough to pay for such a program without obscuring the market economy signals that make technological innovation—and its long‑term dividends—possible in the first place? Doing so would involve some creativity, to say the least. To put it in Derrida’s terms, what’s needed is to “think about the possibility of the impossible.”