Research summary · working paper, draft 3.2, November 2026
If automation pushed unemployment to 20%, could lower prices and a tax-financed benefit preserve the consumption of the workers who lost their jobs? Under what conditions?
Short answer: yes in most cases, but only if automation raises output enough and the government can tax profits when public debt reaches its cap. In the central specification, a benefit that tops the displaced up to their pre-automation income, financed by a profit levy of up to 15% once the debt cap binds, keeps them at 90% or more of their consumption in 277 of 432 cases (64%), against 211 (49%) when the government can only cut the benefit. Making the levy costly shrinks this advantage and, in some specifications, trades it for more fiscal crises. And even when their consumption is preserved, the displaced fall behind those who keep their jobs unless part of the benefit is indexed to wages.
This is a stylised accounting model with a partial calibration to Israel, plus a descriptive look at the data. It describes conditions and mechanisms. It is not a forecast.
Each point is the median productivity gain g* at which both displaced groups keep 90% of their consumption, for a given ceiling on the profit levy. To the right of the line the median combination of assumptions is preserved; to the left it is not. The bars show the quartiles across 144 combinations of the other assumptions.
| Rule when debt reaches the cap | Costless levy | Levy with step-301 costs | Erosion elasticity 0.9 |
|---|---|---|---|
| No levy (benefit cut, equity at present value) | 0.463 (0.344–0.625) | 0.338 (0.175–0.500) | 0.463 (0.344–0.625) |
| Levy up to 15% (central) | 0.325 (0.200–0.450) | 0.250 (0.125–0.325) | 0.375 (0.275–0.506) |
| Levy up to 30% | 0.200 (0.125–0.300) | 0.150 (0.100–0.200) | 0.350 (0.275–0.425) |
| Benefits only | 0.500 (0.350–0.675) | 0.350 (0.175–0.525) | 0.500 (0.350–0.675) |
The step-301 costs are tax-base erosion with elasticity 0.5, owners who consume 0.4 of their income instead of 0.9, and equity priced after the levy. With these costs every threshold falls, even without a levy, so the comparison that matters is between the levy rows and the benefits-only row. An erosion elasticity of 0.9 is above published estimates (about 0.2 for the United States and 0.5–0.6 for Germany); there the levy lowers the threshold less, but at a 15% ceiling it still preserves 246 of 432 cases against 214 with no levy, with 37 fiscal crises instead of 110.
g is not total factor productivity. It is the efficiency gain in a scenario that also displaces 20% of workers: at g ≈ 0.38, output in year 20 equals output in the economy without automation. Data: frontier_web.json · step 305 note.
277 / 432
With a levy of up to 15% of profits before the benefit is cut, the displaced keep 90% of their consumption in 64% of cases, with no severe fiscal crisis; with benefit cuts alone, 49%. A 30% ceiling raises this to 78%. The result assumes that firms keep investing: if investment falls with unemployment, preservation ranges down to 35%, and to 11% in a stress case calibrated to OECD investment data (30% with an interest-rate rule bounded at zero).
g* = 0.33
Under benefit cuts alone, the median case needs automation that leaves output about 21% above the economy without automation (g* = 0.50). With a levy of up to 15% and equity priced at present value, output about 9% below it is enough. This comparison changes both the levy and equity pricing. Giving the levy costs narrows its advantage, which disappears only at a tax-base erosion elasticity of about 2.4, roughly four times the upper end of published estimates.
+29% vs +3%
In the median case the displaced end close to where they would have been without automation (+3% to +4% of consumption), while retained workers gain about 29%. Both displaced groups end below the no-automation control in 47% of cases. Indexing half of the benefit to the wages of the employed, with a floor, closes that gap, at a benefit cost of 5.6% of output instead of 3.9%.
On the empirical side, firm AI adoption in 13 European countries (848 observations) shows no robust association with employment in exposed occupations (β = −0.55, p = 0.39), and in 50 European industries about 0.6 of a fall in unit labour cost reaches relative prices within a year and about 1 within five years, an estimate that may be biased upwards. Neither is a causal estimate.
The full study (in Hebrew, with translated pages) → · Research log →
Parnas, I. (2026). Cheaper Goods, Thinner Budgets: Can Lower Prices and Tax-Financed Benefits Protect Workers Displaced by Automation? (Version 3.1). Zenodo. https://doi.org/10.5281/zenodo.24134846
MacroVision is independent research by Ido Parnas, with no institutional affiliation and no supervisor. The text was revised after three rounds of informal review based on referee-style reports produced with AI tools. That review is not peer review, and the paper has not yet been peer reviewed.
Comments, corrections and press enquiries: idopar@gmail.com. Nothing on this site is investment or financial advice.