The Automation Dividend
How a city could pay the drivers that driverless taxis replace, and what it would cost.
Driverless taxis should spread as fast as they are proven safe. The drivers they replace should be paid. A city can do both: charge driverless operators a levy on fares, pay it into a fund for drivers, and borrow against the fund so the money arrives when the jobs go.
For London, a 15% levy would pay 40,000 established drivers £1,250 a month for three years, tapering over five more. The borrowing is repaid by year 13.
Why not slow the cars down
In 2024, 1,602 people died on Britain’s roads.1Department for Transport, Reported road casualties Great Britain, annual report: 2024. Over 56.7 million miles, Waymo’s cars had fewer injury crashes than human drivers in most categories and more in none.2Kusano et al., Comparison of Waymo rider-only crash rates by crash type to human benchmarks at 56.7 million miles, Traffic Injury Prevention, 2025. Waymo keeps a running version on its safety impact page. Independent reviews of its worst crashes mostly blame the other driver.3Timothy B. Lee, Very few of Waymo’s most serious crashes were Waymo’s fault, Understanding AI, September 2025. Of 45 serious crashes over six months, he judged 37 mostly or entirely the other driver’s fault. If that holds, every year of delay costs lives.
So delay is the wrong answer to job losses, but it is the one politicians reach for. California’s legislature voted to require a human in every autonomous truck; the governor vetoed it.4California AB 316 (2023). Governor’s veto message. Bills like that will keep coming while drivers are offered nothing.
Who loses
The gains are spread thin: cheaper rides and safer roads for millions. The losses are concentrated. London has over 110,000 licensed private hire drivers, plus the black-cab trade.5TfL licensing figures for July 2025, reported by Taxi-Point: 110,668 licensed private hire drivers. Workers displaced like this mostly end up in lower-paid work,6David Autor, Work of the Past, Work of the Future, NBER w25588, 2019. and places with more automation see lower employment and wages.7Daron Acemoglu and Pascual Restrepo, Robots and Jobs: Evidence from US Labor Markets, Journal of Political Economy, 2020.
A technology like this can leave everyone better off, but only if some of the gain reaches the people who lose.8Anton Korinek and Joseph Stiglitz, Artificial Intelligence and Its Implications for Income Distribution and Unemployment, NBER w24174, 2017. Markets don’t do that on their own.
Pay now or pay later
Losers organise; winners don’t. New York’s taxi medallions fell from about $1m in 2014 to almost nothing once Uber arrived, leaving owners with loans they couldn’t repay.9Jacobs, Distressed Drivers, Columbia Human Rights Law Review, 2022. Relief came in 2023, after a hunger strike and most of a decade.10NYC Mayor’s Office, announcement of medallion debt relief, January 2023: $356m forgiven across 1,789 owners. On the hunger strike, THE CITY, November 2021. Some owners were still stuck in 2024.11New York Focus, September 2024.
Cities pay for displacement either way. Paying late costs more and poisons what comes next. London drivers have already declared a “state of emergency” over Waymo’s pilot.12Taxi-Point on the App Drivers and Couriers Union. San Francisco fought the state’s approval of driverless rides, and Cruise lost its permit within months.13CPUC Resolution TL-19144, August 2023; KQED on the vote. The city’s challenge was decided in 2025. California’s DMV suspended Cruise’s driverless permit in October 2023. Operators survive bad press. They don’t survive a council that turns against them.
The mechanism
The levy is a percentage of gross fares on driverless trips that start in the city. It grows with the market, and it is one line in the operator’s accounts, so it is easy to audit.
The credit is cash, paid to drivers who were working before the scheme was announced, so nobody can qualify by getting a licence later. Full-time drivers get the full amount; part-timers get a share. There is no retraining condition. America’s main retraining programme for displaced workers left participants with less income than people who stayed out of it.14Schochet et al., Estimated Impacts for Participants in the Trade Adjustment Assistance Program, Mathematica for the US Department of Labor, 2012. Cash does reduce work slightly: $1,000 a month cut hours by about 1.3 a week in the largest trial.15Vivalt et al., The Employment Effects of a Guaranteed Income, NBER w32784, 2024. Labour force participation among recipients fell by two percentage points. A credit worth half of past earnings, which then tapers, keeps that small.
Payments start when driverless cars reach 10% of trips, which is when earnings start to fall. After that they follow a fixed schedule, so drivers can plan.
Drivers lose income early, when fleets are small and the levy raises little. The levy only becomes large a decade later. So the city borrows against future levy receipts and repays from the mature levy. Cities already fund roads and regeneration this way.16Federal Highway Administration, Tax increment financing.
London, roughly
The biggest assumption is market growth. Driverless rides will be cheaper and available at any hour, so they will create trips that don’t happen today and replace many that now happen in private cars. Bullish forecasts have fleets carrying almost all passenger miles within a decade.17RethinkX, Rethinking Transportation 2020–2030, 2017, which predicted autonomous fleets would carry 95% of US passenger miles within ten years of approval. Fares fall as volumes rise, and the levy is charged on fares, so what matters is spending, not trips. I assume spending on rides triples. If robotaxis replace most private cars it could quadruple; if they only replace taxis it might barely grow.
On that basis each driver gets £1,250 a month from year 5 for three years, then less each year for five more: £82,500 in all, and £3.3bn across every driver. Borrowing peaks at £1.6bn in year 9 and is repaid by year 13.
The fund, year by year
Solvent. Borrowing peaks at £1.6bn in year 9 and is repaid by year 13.
- Credit per driver, full rate
- £1,250 a month, from year 5
- Paid to each driver in total
- £82,500 over 8 years
- Levy raised over 30 years
- £24bn
- Credits paid, all drivers
- £3.3bn
- Interest paid on the bonds
- £376m
- Balance after 30 years
- £20bn in hand
Year-by-year figures
| Year | Driverless share | Levy | Credits | Balance |
|---|---|---|---|---|
| 1 | 2% | £9m | – | £9m |
| 2 | 3% | £15m | – | £25m |
| 3 | 5% | £26m | – | £51m |
| 4 | 9% | £44m | – | £94m |
| 5 | 13% | £74m | £600m | −£432m |
| 6 | 20% | £124m | £600m | −£928m |
| 7 | 29% | £203m | £600m | −£1.4bn |
| 8 | 39% | £316m | £500m | −£1.6bn |
| 9 | 51% | £459m | £400m | −£1.6bn |
| 10 | 61% | £616m | £300m | −£1.4bn |
| 11 | 70% | £764m | £200m | −£880m |
| 12 | 77% | £886m | £100m | −£134m |
| 13 | 81% | £978m | – | £838m |
| 14 | 85% | £1bn | – | £1.9bn |
| 15 | 87% | £1.1bn | – | £3bn |
| 16 | 88% | £1.1bn | – | £4.1bn |
| 17 | 89% | £1.1bn | – | £5.2bn |
| 18 | 89% | £1.1bn | – | £6.3bn |
| 19 | 90% | £1.1bn | – | £7.5bn |
| 20 | 90% | £1.1bn | – | £8.6bn |
| 21 | 90% | £1.2bn | – | £9.8bn |
| 22 | 90% | £1.2bn | – | £11bn |
| 23 | 90% | £1.2bn | – | £12bn |
| 24 | 90% | £1.2bn | – | £13bn |
| 25 | 90% | £1.2bn | – | £14bn |
| 26 | 90% | £1.2bn | – | £16bn |
| 27 | 90% | £1.2bn | – | £17bn |
| 28 | 90% | £1.2bn | – | £18bn |
| 29 | 90% | £1.2bn | – | £19bn |
| 30 | 90% | £1.2bn | – | £20bn |
How it works. Each year, driverless fare revenue is the ride market times the driverless share of trips. The levy is a percentage of that, less 5% for running the fund. Credits are drivers times past earnings times the credit rate, scaled by where the year falls in the schedule. Deficits pay interest at the bond rate; surpluses earn nothing. Nothing else is modelled.
Fixed assumptions. London’s taxi and private hire market is £3bn a year in fares, a round number consistent with driver numbers and earnings. A full-time driver earns £30,000 a year after costs. Eligible drivers are about a third of licence holders, since many licences are barely used. Driverless trips pass 10% of the market in year 5, half in year 9, and level off at 90%.
It holds up. With no market growth at all, the debt still clears by year 20. At 6.5% interest, by year 13. There is room to be more generous: 65% of past earnings for five full years still clears by year 16. What breaks it is a low rate. At 5%, with the market only doubling, the fund is still £101m in debt after thirty years.
Once the debt is repaid, the levy raises about £1.2bn a year. The law should say now what happens to it: cut the rate, or fund the next group displaced.
Objections
Operators will leave. They can’t take the trips with them: a London ride happens in London. Smaller cities have less leverage, which argues for cities acting together.
Cities may not have the power.13See note 13. In Britain, driverless services will run under national permits.18Automated Vehicles Act 2024, which creates automated passenger services permits. So the levy belongs in the permit, with receipts passed to local funds. This is the objection most likely to stop it.
Riders will pay some of it. Yes: passed on in full, a 15% levy turns a £20 ride into £23. Riders are also the biggest winners from cheaper, safer taxis.
The fund will be raided. The US tobacco settlement was meant for health; about a third went on it, and states sold future payments to plug budgets.19GAO, Tobacco Settlement: States’ Allocations of Payments, 2007; ProPublica, How Wall Street tobacco deals left states with billions in toxic debt. So make each credit a legal entitlement, and put the fund under an independent trustee, as Alaska does.20Alaska Permanent Fund Corporation, history.
The bonds are risky. Two or three operators stand behind them. A city guarantee on first losses lowers the rate and makes the risk visible.
“Driver” is hard to define. A fixed cut-off date and proportional credits settle most cases. Appeals settle the rest.
Robot taxes always fail.21The Delvaux report (A8-0005/2017) floated a tax on robots; Parliament left it out of the resolution it adopted. Gates made his case in an interview with Quartz. This isn’t one. It is a charge on a licence the city grants, for one activity, for a fixed period. It is closer to a congestion charge.
It doesn’t help lorry drivers. True. Freight needs a national answer.
Why operators should agree
The levy buys permits that stay granted. Drivers paid from it have a stake in the service running. In 1960, West Coast dockers let employers mechanise the ports in exchange for a $29m fund for the existing workforce.22Mechanization and Modernization Agreement, 1960. The employers’ $29m, paid over five and a half years, protected the jobs and earnings of the registered workforce. Containers followed. This is the same deal, with the city as the other party. Operators who agree early set the terms; New York shows the alternative.
After drivers
Drivers will be the first automation job losses everyone can see: cars on the street with nobody in them. Up to eight million UK jobs could follow.23IPPR, Transformed by AI, 2024. Most have no licence to attach a levy to. But the first case sets the expectation. Get it right and the next argument is about the rate. Get it wrong and every later technology inherits the fight.
Waymo plans driverless rides in London in 2026.24Waymo, Hello, London, October 2025. Agree the terms before then.