The earlier engine (v4.22), kept for exploration

Why five measures. They answer different questions. The stock measures ask whether people have a cushion; the flow measures ask whether this year's income covers this year's needs; extreme poverty asks whether people are homeless. A policy can move one and not another, so each is reported against where people started (year 0) as well as against the Baseline.

Stock: what people own at the end of the run. Wealth poverty counts net wealth below POVERTY_LINE ($25,000). BLEI poverty counts people whose savings, income and BU cover fewer than 30 days of basic living costs. BLEI counts BU and cost relief from the day someone enrols, so its year-0 figure is shown under Baseline rules, with the figure under your settings' rules beneath it.

Flow: what people earn in the final year. Cash income is wage income plus CCO conversion proceeds; PTH appreciation is left out, as a capital gain. Relative income poverty counts cash income below 60% of that scenario's own median, so when most incomes rise together it can stay flat or even rise. Basket poverty compares income with the cost of a living-wage basket (LIVING_WAGE_ANNUAL, adjusted for inflation): "net" uses the prices each person actually pays after CCO, PTF and PTH cost relief; "gross" uses full prices. At year 0 the median person's wage income ($39,945) is below the basket ($49,370), so about two-thirds of people start in basket poverty. That starting gap is what drives the Baseline's decline (Known Limitations; CONTRIBUTING.md v4.17).

Extreme: homeless, with necessities from charity if at all. The engine models no housing, mental illness or vocation, and at the US rate (about 22 per 10,000 people, HUD's 2025 point-in-time count) this is about one person in a 500-agent run. So it is an estimated share built from three pathways, and these rows read per 10,000 people. Economic (73% at year 0) moves with the share of people in housing distress, whose income plus savings cannot cover this year's basket; CCO income and PTF and PTH cost relief lower it. Serious mental illness (25%) does not respond to income, so CCO alone leaves it unchanged; wellness zones, which need both PTH and SZH, lower it by up to half (Housing First trial evidence), scaled by zone coherence. Voluntary (2%) covers ascetics, itinerant monastics and renunciants; it is the same in every scenario, and the share will be updated as data sources become available. This definition differs from the World Bank's, which counts income below $3.00 a day. See Known Limitations and CONTRIBUTING.md v4.18.

Measure. Housing distress is the v4.18 measure: this year's cash income plus savings cannot cover this year's basket. For each seed the study runs your settings with recessions off and with recessions on under each rule, on the same agents, recession path and random draws. The excess is the difference in recession years, averaged over every recession year of every seed, for CCO participants and non-participants separately. Excess extreme poverty converts the all-agent excess through the v4.18 overlay's economic pathway. Extra BU is the added allocation as a share of the base allocation over the run.

Shock-neutral. The multiplier at which participants' excess distress in recession years is zero, found by false position from ×1 and ×3 (widened to ×4 if needed). At the reference settings it is ×1.36 over 300 seeds (node harness.js stabilizer), shown as ×1.35 on the slider; the scaled rule's equivalent is about +2.8% BU per 1% income loss. These are the sliders' defaults. The Research Hub's ×1.20 offsets about half of the rise. The neutral point is set by a roughly fixed amount of extra relief, about 7% of the basket, so it is the same in every environment at the same BU and 0% inflation, and higher where inflation erodes an unindexed BU: ×1.40 in Adverse Environment and ×1.55 in Stress Test (v4.21).

Why the default is a target, not the peak of a curve. Returns to a larger increase are close to linear, because the engine has no price response to BU issuance and no budget constraint. A larger increase never looks worse, and the model cannot test whether it would be inflationary. The default is the smallest increase that fully offsets a recession for the people it reaches.

How the BU amount reaches a household (v4.19). Through v4.18, CCO's cost relief was a flat 20% of the basket whatever the BU amount, so a larger BU helped only through conversion proceeds, which runYear() credits once a year. The framework describes BU as redeemable for essential goods, so from v4.19 the relief scales with the BU actually issued: 20% at the $1,200 reference, in proportion above and below it, up to a 50% ceiling (a placeholder). Every $1,200 preset is unchanged. Under the v4.18 rule the shock-neutral multiplier was ×2.35.

Reach. A BU increase reaches only CCO participants. Emergency enrollment extends the CCO cost relief to non-participants in triggered years; its 50% take-up default is a placeholder with no source.

The same lever can target places as well as times. A region that wants to attract residents, such as a cold-climate area with little winter sunlight, could add a BU premium for residents who meet a residency requirement, much as Alaska's Permanent Fund Dividend pays each qualifying resident after a full calendar year of residence. Whether such payments move people is not well established: a 2019 University of Alaska review lists the dividend's effect on in-migration among the questions not yet studied. The closest measured case, Tulsa Remote, pays remote workers $10,000 to relocate; an Upjohn Institute evaluation (Bartik, 2025) estimates that 58–70% of participants would not have moved without it. Remote workers are a different population from households in poverty, and the same evaluation notes that added housing supply is needed to realise the benefits. A payment tied to a place can also be absorbed by higher local rents; Public Trust Housing, whose costs are set outside the private rental market, could limit that. This simulation has no geography, so it reports no result for this use.