How we compare programs
The core idea
Two charities can both be honest, well run and low on overhead, and still differ a hundredfold in what a dollar buys. That gap, not overhead ratios, is what we care about. We compare programs on the cost of a concrete outcome: a death averted, a year of healthy life, a child in school, a ton of carbon dioxide kept out of the air.
Outcome measures we use
| Measure | What it means | Where it comes from |
|---|---|---|
| Cost per death averted | Dollars spent per person who would otherwise have died | Randomized trials plus program cost data (GiveWell, program evaluations) |
| Cost per DALY averted | Dollars per year of healthy life gained; combines deaths and illness | WHO and IHME burden estimates applied to trial results |
| Cost per unit of income | Dollars per dollar of income increase for recipients, often stated as a multiple of cash transfers | GiveWell and Open Philanthropy models, GiveDirectly evaluations |
| Cost per ton of CO2e | Dollars per ton of greenhouse gas kept out of or removed from the atmosphere | Founders Pledge and Giving Green analyses |
| Cost per animal-year | Dollars per animal moved from a worse to a better living condition, per year | Animal Charity Evaluators, Rethink Priorities |
Who we rely on
We do not run our own trials. We summarize evaluators that publish their models and their reasoning: GiveWell (global health), Open Philanthropy, Founders Pledge and Giving Green (climate), Animal Charity Evaluators (animal welfare), the Happier Lives Institute (wellbeing), and the academic literature behind them. When evaluators disagree, we say so and explain why.
What a cost-per-outcome number cannot tell you
- It is an estimate. Most rest on a handful of trials plus assumptions about how well results carry over to new places. Treat a 2x difference as noise; treat a 10x difference as real.
- It changes. Funding gaps close, prices move, new evidence lands. Every figure we print carries the date and the source.
- It hides values. Comparing a child's life against a ton of carbon or a chicken's living space requires moral judgments. We make ours explicit and show the numbers so you can apply your own.
- Marginal, not average. The question is what your next dollar buys, not what the charity's average dollar bought last year. A fully funded program has a marginal value near zero even if its average is superb.
Our own rules
- Every number links to a source and states the date we read it.
- No charity pays for coverage. We accept no referral fees.
- We publish corrections at the top of the article, dated.
- We write in plain language. If a term needs explaining, we explain it the first time it appears.