UBI: The Policy No One Believes In — Until It Pays Their Bills

People often ask me about the future and Universal Basic Income. I can’t predict the future. But I’ve had a short answer for a long time: it won’t be a choice. Here’s why.

“UBI won’t be championed by idealists. It’ll be pushed through by the same people who swore it was impossible — right up until it started paying their bills.”

For decades, UBI was the ultimate political mirage — the moonshot economists debated, futurists blogged about, and politicians politely waved away. The problem wasn’t the math. It was the politics.

That shifts the moment the professional-managerial class — the donors, the opinion shapers, the “knowledge workers” — start feeling the squeeze they once observed from a safe distance.

Picture the moment a $200k/year strategy consultant realizes GPT-7 does their job better, faster, and without the lunch breaks. That’s when “handouts” quietly rebrand as “transitional support.”

And when that happens, the political calculus flips overnight.

UBI won’t start as UBI. It’ll be rolled out as a “temporary emergency measure,” a short-term patch to keep the middle class afloat. But political gravity will swell it. Institutional inertia will cement it. My prediction is that by 2032, we’ll have UBI in everything but name — even as politicians insist it’s just a stopgap, while campaigning on raising the monthly amount.

The irony? The loudest defenders of this “impossible” policy will be the same voices who spent thirty years insisting it could never happen. Not because they’ve embraced the theory — but because they’ve embraced the deposits.

UBI won’t arrive with a bang. It’ll arrive like a mountain: slow, silent, and immovable.

And once that mountain forms? Good luck moving it.

Update: August 2025 → March 2026

In August 2025, I wrote that universal basic income would not arrive as an ideological triumph. It would be pushed through as a "temporary emergency measure" — by the same people who spent thirty years insisting it was impossible, once they found themselves facing the pressure firsthand.

Six months later, you don't need AI to see this process already underway.

A new disability support system took effect on September 1, 2025. Disability assessment normally lasts until age 67; reassessment can be requested if circumstances change. Partial disability recipients receive 82% of full benefits with a 350,000 ISK monthly earnings exemption. The Icelandic Medical Association has argued that, for people entitled to projected pension-fund disability benefits, combined payments can reach prior earnings. Actual payments depend on individual rights and income. The share of people aged 18–66 receiving disability or rehabilitation pensions rose from 6.3% on January 1, 2000 to 9.4% on January 1, 2024.

This is not a critique of those who depend on the system. It is an observation about the mechanics.

The 2025 disability reform is not universal basic income, either in name or in substance. Eligibility remains conditional, and benefits remain income-tested. It would be misleading to treat the reform as confirmation of my prediction.

What it does illustrate is something narrower but still important: Iceland is already reconsidering how income security should relate to a person's capacity to participate in the labour market.

That distinction matters. Income security is not the same as UBI, and the existence of one does not prove the inevitability of the other.

A note on the earnings exemption. The 350,000 ISK a month I cited in March matches the special earnings exemption (250,000) plus the general one (100,000) as the system launched in September 2025. From January 1, 2026 the special exemption is 263,000 ISK a month.

The political question remains open.

Clarification added November 3, 2026.

Update: November 3, 2026 — The Human Dividend

What if the future of work isn't less work, but different work?

When I wrote this article in August 2025, my argument was political: universal basic income would not arrive because politicians were suddenly persuaded by the theory. It would arrive when economic insecurity reached people who had believed themselves immune to it.

I still think that is a plausible trajectory. But over the past year a more interesting question has emerged, and it is not simply whether artificial intelligence will eliminate jobs.

What happens when we can produce more without needing proportionally more human labour?

In September 2026, registered unemployment in Iceland was 4.0%, compared with 3.5% a year earlier. In October, collective redundancy notices covered 281 employees, 227 of them in public administration. These figures do not show that AI caused the job losses. The economic picture is more complicated than that. (The Statistics Iceland labour force survey measures something different and put seasonally adjusted unemployment at 5.7% in September.)

The underlying technological question remains. An organisation might double its output without doubling its workforce. It might stop hiring rather than start firing. Productivity might rise while opportunities for new workers become harder to find.

A revised Stanford study offers an early signal. Using US payroll data through June 2026, it finds that workers aged 22 to 25 in occupations highly exposed to AI have fallen behind comparable groups, mostly through reduced hiring rather than increased dismissals. The authors describe the findings as early and descriptive. This is not evidence of an economy-wide employment collapse. It is a reason to look beyond unemployment rates.

The first job AI eliminates may be the job that never gets created.

The wrong question

For decades, we have organised economic policy around employment. We want businesses to grow because growing businesses create jobs. We want communities to thrive because thriving communities provide work. We measure success through participation, productivity, wages and growth.

But suppose the relationship between production and human labour begins to change. A business might serve twice as many customers with the same workforce. Public administration might become significantly less expensive. AI systems might take over a growing share of repetitive analytical and administrative tasks.

There may still be enormous demand for human work. Just not necessarily the same work. We will still need people to care for the elderly, teach children, support families, rehabilitate patients, build relationships, create art and hold communities together. Some of these activities may become more valuable precisely because so much else can be automated.

The economic problem then becomes a distribution problem.

Who receives the benefits of productivity when those benefits are no longer distributed primarily through wages?

Iceland as a laboratory

On October 7, 2026 the Icelandic Regional Development Institute announced additional funding of ISK 80 million annually for five years for Brothættar byggðir, its programme for vulnerable rural communities, and a new track called Sterkari byggðir from 2027.

Regional development has traditionally been tied to employment opportunities. I want to explore another possibility: what if predictable income itself could become part of regional infrastructure? Not instead of housing, education, healthcare or jobs. Alongside them.

Consider someone who wants to stay in a small community, care for an ageing parent and develop a small business. The opportunity may exist, but the financial risk may make staying impossible. A predictable income floor could change that calculation. It might create room for entrepreneurship, part-time work, education, caregiving and civic participation.

Or it might not. That is what a properly designed experiment should establish.

The Human Dividend

I am planning an exploratory economic simulation for Iceland through 2045. The question is whether productivity growth from AI and automation could, under identifiable assumptions, finance both an income floor and an expansion of human-centred services.

In the model, automation does not mean everyone stops working. It means that some productive activities require fewer human hours, while society deliberately invests more resources in care, education, healthcare, culture and community development.

The intended scenarios will be mathematical experiments, not forecasts. They do not establish that the required productivity growth, taxation or redistribution can actually be achieved, and the financing challenge is substantial. For illustration, a payment of ISK 150,000 per month to about 309,000 adults (Statistics Iceland, January 1, 2026) would cost roughly ISK 557 billion a year, before taxes, offsets or changes to existing benefits.

The question is not merely whether Iceland could generate enough output. It is how much of that output would be available for redistribution, which public services must still be financed, and how much real purchasing power the payments would provide.

Recent research also reminds us that unconditional cash transfers can reduce paid working hours. In an experiment in two US states published in the Quarterly Journal of Economics in August 2026, participants who received USD 1,000 a month for three years worked one to two hours less per week. That must be part of any serious model, not dismissed.

But I believe there is a larger question worth investigating. If automation makes part of our existing production significantly more efficient, could we deliberately convert some of that gain into greater economic security and more time for human needs? I call that possibility the Human Dividend.

Beyond universal basic income

I began this article by arguing that UBI would eventually become politically unavoidable. I am not ready to claim that prediction has been confirmed. Nor do I think UBI alone answers the most important questions.

Income security matters. So does access to technology. And so does ownership of the systems that create the wealth. A society could become dramatically more productive while the benefits accumulate among a small number of asset owners. Or it could develop institutions that distribute those benefits more widely, through wages, lower prices, public services, shared ownership or income guarantees. That is a choice about economic institutions, not an automatic consequence of technological progress.

And perhaps this is where the original argument becomes more interesting.

What if the real promise of artificial intelligence is not that humans become economically unnecessary, but that we finally have the productive capacity to invest more of our human time in one another?

We should not assume that this future will happen. We should model it, test it, and measure what it would actually require.

The mountain may still be forming. But the more important question is no longer simply whether it will arrive. It is what we choose to build on it.

Sources

Directorate of Labour: registered unemployment, September 2026

Directorate of Labour: collective redundancies, October 2026

Statistics Iceland: labour market, September 2026

Statistics Iceland: population by age, January 1, 2026

Icelandic Regional Development Institute: expansion of Brothættar byggðir, October 7, 2026

Brynjolfsson, Chandar and Chen: Canaries in the Coal Mine? (Stanford Digital Economy Lab, revised August 12, 2026)

Vivalt et al. (2026): The Employment Effects of a Guaranteed Income, Quarterly Journal of Economics 141(4)

Regulation no. 1490/2025 on the earnings exemption for disability benefits