# Can AI finance a universal dividend?

**Status: executed illustrative scenario accounting, 27 October 2026.**

## The question

If AI increases economic output, how much of that increase could reach every resident as a publicly funded payment after tax collection and implementation costs? The model tests financing arithmetic, not whether AI will deliver the assumed gains or whether a payment would meet basic needs.

## What the numbers mean

The economy produces 100 units in a baseline year. Every monetary number below is a percentage point of that baseline annual gross domestic product. The modeled period is one year after the assumed deployment, not 2027 or a cumulative forecast. Prices and population are held fixed. There is no currency conversion. All residents, including children, share a transfer equally. Consequently a transfer pool equal to 1% of GDP pays each resident 1% of baseline GDP per person per year. GDP per person is **not** a household income or poverty threshold.

All numeric inputs are author-selected teaching assumptions. They are **not** estimates from the International Monetary Fund, Jamaica, or another source. External sources support the channels examined, not these parameter values. The package is intentionally not calibrated to Jamaica: it has no verified local adoption, displaced wages, taxable-profit retention or delivery-cost series. It should not be marketed as a Jamaican forecast.

## Results, rounded for publication

| Annual change, % of baseline GDP | Inclusive gains | Concentrated returns | Weak gains and disruption | Promise before funding |
|---|---:|---:|---:|---:|
| Additional productivity value before disruption | 7.20 | 7.20 | 1.00 | 7.20 |
| Output lost during disruption | 0.50 | 0.50 | 1.50 | 0.50 |
| Net output change | 6.70 | 6.70 | -0.50 | 6.70 |
| Wage-income change, before tax | 3.02 | -2.00 | -2.20 | 3.02 |
| Capital-income change, before tax | 3.68 | 8.70 | 1.70 | 3.68 |
| Additional public revenue, net of lost wage-tax receipts | 1.52 | 0.04 | -0.19 | 1.52 |
| New public implementation and transition spending | 0.40 | 0.60 | 0.60 | 0.40 |
| Fully funded payment reaching residents | **1.07** | **0.00** | **0.00** | **1.07** |
| Other funding required before any new dividend | 0.00 | 0.56 | 0.79 | 0.00 |
| Promised payment, where specified | Not specified | Not specified | Not specified | 4.00 |
| Funding gap if that promise is paid, including delivery cost | Not specified | Not specified | Not specified | **3.09** |

Rounding can cause columns to differ by 0.01. Charts must use the unrounded JSON. The 1.07 is a **universal dividend in this example, with adequacy untested**, not an adequate universal basic income or cash available now. Zero means no dividend financed by these incremental revenues, not that existing social protection vanishes or all financing options are exhausted. In the fourth case 1.07 is affordable and 4.00 is the proposed alternative; **do not add them**.

## Five findings in plain language

1. More output does not automatically fund a payment. In the inclusive illustration, a 6.7% output increase leaves 1.07% of baseline GDP for residents after tax collection, public transition spending and delivery costs.
2. Who receives the gains matters. The concentrated-return case has the same output growth, but falling wage receipts and low collection from capital leave no new dividend after transition costs. This is a bundled scenario comparison, not proof of one variable's causal effect.
3. Workers can lose income while output rises. Aggregate wage income falls by 2.00% of baseline GDP in the concentrated illustration. This is a wage-income change, not a prediction of job losses, inequality or any individual's outcome.
4. A payment promise needs a funding source. A 4% GDP transfer in the inclusive economy leaves a 3.09% GDP gap once delivery costs are included. Debt, other taxes or spending cuts would change the policy and are not silently assumed.
5. Better collection changes the result even with unchanged output. Holding every other inclusive assumption fixed, effective capital collection of 5%, 15% and 25% yields funded dividends of approximately 0.37%, 0.72% and 1.07% of baseline GDP. These are sensitivity settings, not recommended tax rates.

## Inputs and their purposes

| Input | Inclusive | Concentrated | Weak | Promise | Purpose and provenance |
|---|---:|---:|---:|---:|---|
| Share of economic activity adopting | 60% | 60% | 25% | 60% | Teaching setting; an output-weighted adoption share, not percentage of jobs or firms |
| Productivity improvement in adopting activity | 12% | 12% | 4% | 12% | Teaching setting; improvement after recurring private AI expenses; excludes separately listed disruption |
| Output disruption, % baseline GDP | 0.5 | 0.5 | 1.5 | 0.5 | Teaching setting; lost real production, separate from income redistribution |
| Worker share of new net value | 60% | 15% | 40% | 60% | Teaching setting; splits the net value between wages and capital |
| Existing wage income shifted to profit, % baseline GDP | 1 | 3 | 2 | 1 | Teaching setting; replacement of wage payments by capital income, not an additional loss of GDP |
| Effective labor-income tax yield | 20% | 20% | 20% | 20% | Teaching setting; yield on wage-income change, including negative receipts; not a statutory rate |
| Effective capital-income tax yield | 25% | 5% | 15% | 25% | Teaching setting; receipts after exemptions, leakage and collection constraints; not a statutory rate |
| New public implementation/transition spending, % GDP | 0.4 | 0.6 | 0.6 | 0.4 | Teaching setting; retraining, adjustment support and public implementation, not private costs already deducted |
| Delivery administration as share of gross dividend budget | 5% | 5% | 5% | 5% | Teaching setting; incurred only if a transfer is delivered; 95% reaches residents |
| Eligible share of population | 100% | 100% | 100% | 100% | Definition of the universal illustration, including children |
| Promised net transfer, % GDP | None | None | None | 4 | Teaching choice demonstrating a promise exceeding new funding; not an adequacy threshold |

The income split is deliberately simplified. This is a two-factor teaching economy, not a full national-accounts decomposition: depreciation, net product taxes and cross-border factor-income flows are not separately modeled. Income shifted from wages to capital can reflect labor replacement, but it does not identify job counts. The model does not observe who owns that capital, domestic versus foreign ownership or household-level incidence. The effective collection setting abstracts from those issues; a country study must model them explicitly.

## Accounting and safeguards

Productivity value minus disruption equals new net output. New wage income plus new capital income equals that output exactly. Wage income shifted to profit is counted once, not as extra output. Tax is a transfer to government, not a second economic gain. A universal payment is redistribution, not another addition to GDP. Its administration and public transition costs consume the available fiscal resources.

The funded payment uses only nonnegative incremental revenue remaining after public costs. A funding shortfall is reported separately instead of creating a negative transfer. The promise scenario keeps the inclusive economic assumptions unchanged and calculates the additional financing needed to deliver the promise. Existing services and budgets are otherwise unchanged; this is not a proposal to replace targeted benefits.

## What the model cannot say

- No probability forecast, Monte Carlo confidence interval or empirical causal estimate is produced. Animation should reveal the accounting stages, not look like a forecast trajectory.
- There is no inflation, price elasticity, import leakage, exchange-rate, debt-interest or supply-response equation. An unfunded promise can create financing and demand pressures, but a 3.09% funding gap is **not a 3.09% inflation prediction**. Financing method and available supply matter.
- No poverty or welfare conclusion follows from GDP or the dividend alone. Household composition, who loses wages, existing benefits, disability support and access to essential services matter.
- The same funds could support targeted transfers, wage insurance, public services, training or a mixed package. Their relative benefits require distributional evidence, not this model.
- Adoption and productivity are not estimated independently, and not every percentage setting is feasible. There is no behavior feedback from taxation, demand or investment. The grid is a stress test of the arithmetic, not a policy optimizer.

## Jamaica reproduction checklist

Use a common dated price base and official annual output and population, then replace every teaching setting with documented local estimates or explicit scenario ranges. Consult STATIN for national accounts, population, employment and wages; the Ministry of Finance for receipts and delivery budgets; the Bank of Jamaica for macroeconomic context. Record imported AI-service costs and foreign ownership/tax leakage. Test households losing wages and existing targeted benefits separately. Do not splice nominal GDP, purchasing-power-parity output, different-year population or monthly transfers without conversion. A source link is not evidence that a missing parameter has been measured.

## Sources read, checked 27 October 2026

1. Brollo, F., Dabla-Norris, E., de Mooij, R., Garcia-Macia, D., Hanappi, T., Liu, L., and Nguyen, A. D. M. (2024). *Broadening the Gains from Generative AI: The Role of Fiscal Policies*. IMF Staff Discussion Note 2024/002. https://www.imf.org/en/Publications/Staff-Discussion-Notes/Issues/2024/06/11/Broadening-the-Gains-from-Generative-AI-The-Role-of-Fiscal-Policies-549639 . Landing-page abstract and policy overview read, not full paper. Supports considering productivity, labor disruption, inequality and fiscal distribution under uncertainty. **Does not supply model inputs.**
2. Dabla-Norris, E., and de Mooij, R. (17 June 2024). *Fiscal Policy Can Help Broaden the Gains of AI to Humanity*. IMF Blog. https://www.imf.org/en/Blogs/Articles/2024/06/17/fiscal-policy-can-help-broaden-the-gains-of-ai-to-humanity . Full retrieved page read, sections on social protection, capital-income tax, emerging-market exposure and public services. Supports the financing and protection channels. **Does not forecast this dividend or recommend these illustrative rates.**

A Jamaica 2026 IMF report search surfaced a 2024 population/GDP-per-capita baseline, but the report page returned 403. It was not used to calibrate this model. This avoids treating a search excerpt as a verified numerical baseline.

## Files and reproduction

Run `python3 model.py`. It creates `scenarios.json`, `scenarios.csv`, `sensitivity.json`, `test-results.txt` and SHA-256 manifest. The code uses only the Python standard library and runs deterministically. `formulas.tex` preserves mathematical source. Public charts should include a static text/table fallback, exact unit labels, negative values and a reduced-motion setting. Do not animate hypothetical adoption as observed history.

## Formula key

G is added productivity value before disruption; a is the output-weighted adoption share; p is improvement within adopting activity; L is output disruption; V is net new output. W and K denote wage and capital income, with Δ meaning change; s is the worker share of new value; D is existing wage income shifted to capital. The two t parameters are effective tax yields. R is incremental public revenue; C is new public implementation and transition spending; B is the available gross dividend budget; h is administration's share of that budget; U is transfer reaching residents; A is administration cost; F is other funding needed. P is a proposed larger resident payment and F with subscript P is its financing gap. For currency scaling, Y is baseline annual GDP, N is population, e is the eligible population fraction, and lowercase u is annual payment per eligible person.