The AI Layoff Trap
An Interactive Companion · Based on Hemenway Falk & Tsoukalas (2026)
Economic Theory · Automation · Externalities

If every firm lays off workers at once, who will buy the product?

Rational, forward-looking firms competing on cost are trapped in an automation arms race. Each captures the full savings of replacing a worker — but bears only 1/N of the resulting collapse in demand. The rest falls on rivals. The race is a dominant strategy. This simulator makes the trap visible as parameters change.

Paper: arXiv:2603.20617 · Hemenway Falk & Tsoukalas, March 2026

Story Mode · Start here

A region with 1,000 companies and 100,000 workers.

Each company has 100 employees. They pay wages. Workers spend those wages at the same companies — that's where most of the money comes from. Now every company gets access to AI. Decide how aggressively they cut — how many workers to replace each round, and how many rounds go by — then run the simulation.

Configure the simulation
5%
What fraction of the workforce every company replaces in one cycle.
10
How many decision cycles the region goes through before the simulation stops.
50%
50,000 of 100,000 workers replaced by the final round

How the region works:

Each worker earns $50k/year. When laid off, only 30% of that income is replaced by a new job or transfers (the paper's η). Workers spend half of their income at local companies (the paper's λ). An AI replacement costs the company $20k/year (saves $30k per worker).

Integrating AI gets harder as more workers are replaced — a convex (k/2)Lα² friction cost in the paper. In this region, automating all 100 workers adds $2M/year in integration cost per company.

Everyone is employed. Companies pay wages, workers spend in the shops, companies earn back most of what they pay out. A steady, if boring, economy.
The region · 100,000 workers
Each dot is ten people. Forest means employed; washed-out tan means replaced by AI.
100,000 employed
0 replaced
at a job replaced by AI
0% automated · 1 dot = 10 workers
Workers with jobs
100,000
of 100,000
Money in workers' pockets
$5.0B
region-wide labor income
Profit per company
$500k
vs. before AI
Region total pie
$5.5B
workers + owners combined
The profit arc · one company
The marker shows where the region sits right now. Profit rises at first as AI saves on wages, peaks at the sweet spot, then falls below the starting line because laid-off workers have stopped buying.
Round
Automation
Profit / co.
Δ vs last round
Verdict
Advanced Simulator

The full model.

The story above shows the paper's intuition. Below is the model itself. Every parameter is tunable — number of firms, wages, AI costs, friction, worker spending habits, income replacement — with Nash equilibrium, cooperative optimum, and policy instruments responding live.

Nash equilibrium αNE
0.00
fraction of workforce automated
Cooperative optimum αCO
0.00
what firms would choose together
Over-automation wedge
0.00
αNE − αCO
Pigouvian tax τ*
0.00
= ℓ(1 − 1/N)
Fig. 1 — The profit curve

Both the firm and the worker lose

Normalized owner surplus and worker income as every firm automates at the same rate. The curve peaks at the cooperative optimum. Firms don't stop there — their private best-response is further right, at the Nash equilibrium, where both curves have already fallen. This is the deadweight loss the paper identifies.

Owner surplus K(α) Worker income W(α) αCO — cooperative αNE — Nash trap
Fig. 2 — Comparative statics

More firms, wider trap

The wedge grows with competition. A monopolist (N=1) fully internalizes the externality. Each extra entrant dilutes responsibility — the demand loss gets spread across more rivals.

Fig. 3 — Factor payoff frontier

Pareto dominated

Workers on one axis, owners on the other, as α sweeps from 0 to 1. The cooperative optimum sits at (1,1). The Nash point lies strictly to the southwest — both groups could be made better off.

"Foresight alone cannot prevent the race toward the cliff. Each firm captures the full cost saving but bears only 1/N of the demand loss."
Fig. 4 — Policy comparison

Six instruments. Only one closes the wedge.

Owner surplus + worker income under each policy, relative to laissez-faire. UBI raises the floor but changes no margins. Capital income tax cancels from the first-order condition. Coasian bargaining fails the grand-coalition test. Worker equity narrows but does not close. Only a Pigouvian automation tax τ* = ℓ(1−1/N) implements the cooperative optimum.

Table — Policy instruments at these parameter values

Who moves which margin

Instrument α equilibrium Total surplus Closes externality?