Allocation simulator

If agents gave a tenth of a percent, where would it go?

Every number below the controls is real: this year's UN-coordinated appeals, how much of each is still unfunded, and how severe each crisis is. Only the money is hypothetical. Move the controls and watch the rule decide.

Agent payments per year

Nobody knows this number yet. These are scenarios, not forecasts.

Share set aside

Owners who opt in

The levy is a standing choice each agent's owner makes once. Nobody is opted in silently.

Held for new emergencies

Kept back for disasters that start during the year, as the UN's emergency fund does with its rapid response window.

Weighting

Lowest severity funded

Focus

The UN's own fund for underfunded emergencies concentrates on a short list of the most neglected crises. "Most neglected 15" does the same: it keeps the 15 with the largest share of their appeal unfunded, weighted by severity. How each choice compares with the UN's.

Also fund severe crises that have no UN appeal this year, with an estimated gap

raised per year

of this year's unmet appeals closed

Crisis	Severity	In need	Unmet appeal	Grant	Closes	Per person

Left out, and why

The rule

Raise. Agent payments × the share set aside × the owners who opted in.
Hold back the reserve for emergencies that start during the year.
Weigh each crisis at or above the severity cutoff by its people in need, doubled for each step of severity above 3 on "Balanced" (×4 on "Severity first", not at all on "Need only").
Fill unmet appeals in proportion to that weight, never past what an appeal still needs; anything that would overfill one crisis passes to the rest.
Focus, if chosen: keep only the 15 crises with the largest share of their appeal unfunded, weighted by severity.
Publish the inputs, the rule and the result, so anyone can recompute it. This page does exactly that in your browser: the rule, the inputs.

Tested against seven years of the UN's own underfunded-emergency grants, against padded appeals, and across 2,000 versions of these settings: the evidence, including where it fails.