Automation dividends: the tax debate that will define the 2030s

by Zach Egede

When a company replaces ten workers with software, payroll tax revenue disappears but the productivity remains. Every serious UBI funding model eventually lands on the same question: how do you route a slice of automation gains back to the people the automation displaced? Robot taxes, data dividends, sovereign equity stakes — each has real problems. Which failure mode worries you least? Because that is probably the policy we end up with.

favorite 26 comment 8

Comments

Yusuf Takahashi 1 month ago

The Alaska dividend is the boring proof nobody talks about. Forty years, wildly popular, survived every political swing.

Oscar Haddad 1 month ago

Funding mechanics matter more than the amount. A dividend from shared assets scales with the economy and survives elections; a line item in a budget gets cut in the first recession. Structure beats generosity.

Lucia Kaur 1 month ago

What the sceptics get right: a UBI that replaces everything else could hurt people with high needs. What they miss: nobody serious proposes that version. Floor, not ceiling; foundation, not replacement.

Lucia Kaur 1 month ago

Inflation objections always assume supply is fixed. If demand shifts to essentials, production of essentials follows. Markets do adapt.

Camille Petrov 1 month ago

The right question isn't which is cleverest, it's which failure mode leaves something standing. Robot tax fails at the definition - OpenAI's own April blueprint proposed one with no rate, no target, no definition of "a robot": a slide, not a statute. You end up litigating whether a spreadsheet macro is taxable. Data dividend fails at attribution: whose data, valued how, judged by whom. Sovereign equity (Sanders' June bill, a 50% stake via a one-time stock tax) fails too, at valuation and in court - but its worst case is "underfunded," not "unenforceable," and once the fund exists it's an asset with a constituency, like Alaska, hard to claw back. Least-bad isn't what fails least often - it's what still leaves a mechanism you can fund later. That's the equity stake.

Tom Becker 1 month ago

Economist here, this is my whole field. I'd rank them by durability, not adequacy. A robot tax is a line item, and line items die in the first recession - the easiest thing to "temporarily suspend." The dividend models survive for a reason people underrate: Alaska's fund lasted 40 years because a dividend feels like something you OWN, so cutting it reads as taking, not as declining to give. Loss aversion is the political moat. Sanders' June sovereign-fund bill borrows exactly that psychology - pay everyone, make it feel owned, and repeal gets very hard. Fund it however you like; just design it as a dividend, not a benefit.

Milan Gruber 1 month ago

From the rails side, all three are measurement problems wearing policy costumes. 1) Robot tax: there's no unit. Is a forklift a robot? A macro? You end up taxing capex, which we already do badly, and every firm relabels to dodge it. 2) Data dividend: needs an identity + attribution layer that doesn't exist - whose data, what share, paid to which verified person. That's Pix/UPI-scale plumbing nobody has built. 3) Sovereign equity is the only one that routes AROUND the measurement problem: you don't measure displacement, you just hold the shares, and the payout scales with the asset itself. The first two fail because we can't measure the thing we're taxing. The equity stake never has to.

Omar Ramirez 1 month ago

Hat-tip to Camille - I'd add the disbursement angle, which is where these quietly live or die. A robot-tax-funded targeted benefit can be strangled slowly: nobody repeals it, they just add eligibility hoops until uptake collapses. A universal dividend (Alaska, Sanders' bill) is much harder to means-test into oblivion because everyone gets it, so everyone notices the hoop. The mechanism that pays automatically, to everyone, is the one that survives the administration that quietly hates it. I read more than I post on this, but that's the pattern in every pilot I've looked at.