Return to the town Agency Laboratorymodel 0.4.0
Reproducible Monte Carlo experiment

One town tells a story.
1,000 towns reveal the distribution.

Start with a fair control world: equal effort, starting wealth, actions, and opportunity. Then switch on transportation, health, access, and education constraints one at a time—or all together—and compare the resulting distributions.

1,000 independent worlds120,000 simulated lives364 daily transitions100% effort, locked
02 / Results

The result does not exist until you run it.

Choose a reproducible seed and parameter set, then generate the full distribution. No headline value is prefilled.

03 / Operational definitions

Every number must say what it measures.

01

Stable at year end

Food, housing, and safety are all above the failure threshold on the final day.

Essay reference: roughly 90%
02

Any basic need failed

Food, housing, or safety crosses its failure threshold on at least one day during the year.

Broader path measure derived from the essay's three needs
03

Ever in compound failure

Two or more basic needs fail simultaneously on at least one day during the year.

Essay reference: nearly half
04

Trapped most of the year

Compound failure persists for more than half of the simulated year.

Essay reference: about 1 in 24
05

Ever displaced

Housing security crosses the displacement threshold at any point in the year.

Mechanism output—not an essay target
04 / Scientific boundary

This estimates the model—not the country.

Monte Carlo repetition measures uncertainty produced by these rules. It does not make the assumptions empirically true.

  • Residents are synthetic agents, not survey respondents.
  • The 95% intervals describe uncertainty in the simulated world mean.
  • The control is calibrated only to the essay's three published aggregates; it is not fitted to population data.
  • Constraint layers are explicit model assumptions, not empirical causal estimates.
  • Every parameter and seed is exported with the raw outcome table.