Cost per averted DALY, shown all the way down

What changed on 24 September 2026

This annex was published in August on the yields and burden inputs then in use. Four corrections have since been made, two of which move figures against the programs. They are listed here so the August version and this one can be compared line by line.

Rwanda no longer pays for itself on these figures. Contracted carbon revenue covers 68% of the program budget and 55% of the fully loaded cost, against the 106% this annex published in August, and the break-even price is $22.80 per tonne against $15.42 implied by the contracts. The water volume contradiction that the August version carried, 40% school-day intake in the mortality chain against 30% in the delivery chain, is still open and one value will be chosen for the next version. A second reconciliation is open: in Asili and Helvetas the credits now carried in the portfolio model divide by their yields to a smaller served population than the health chains use, 2.45M person-years against 4.43M in Asili and 0.93M against 1.27M in Helvetas. The credit ramp and the service population are being reconciled program by program, and until they are, those two health chains rest on the larger population and their cost per DALY is the more favorable of the two readings.

Four numbers per program, not one

Every program in the portfolio is scored on the same four metrics. Together they answer the two questions a funder should insist on: what does a dollar buy, and would the claim survive a referee.

Each model also reports self-sustainability: the ratio of carbon revenue to delivery cost, in the base case and at the contracted downside price floor. Moral-weight sensitivity is run under three published frameworks (GiveWell default, egalitarian, wellbeing-anchored); the GiveWell default is the headline.

Amazi Meza Rwanda, restated on the all-cause basis

The flagship school safe-water program: 2,000 schools and 1.98 million students at scale, Gold Standard registered, Article 6 credits contracted and issuing since 2024. The headline below is computed on the all-cause mortality basis GiveWell uses for water; the strict morbidity-only frame is retained as a sensitivity floor.

The restatement matters because bases were previously mixed: the strict headline ($6,021) counts school-age illness only, with zero mortality, a frame roughly twenty times stricter than the all-cause convention GiveWell's own water models adopted after the Kremer et al. meta-analysis. This annex now reports every program's headline on that one basis, with the strict chains retained in the workbooks as sensitivity floors. The fund's primary metric, marginal donor cost, is what carbon revenue drives toward zero regardless of frame; the restated gross figures show the programs are competitive even before revenue.

The same program, in GiveWell’s own dialect

This chain rebuilds the Rwanda model with the structure and conventions of GiveWell’s published water-quality cost-effectiveness analyses: GBD baseline mortality, their pooled all-cause effect, their adjustment vocabulary, their 2020 moral weights, and their units-per-dollar-versus-cash bottom line. Every parameter is stated; the reconciled cost basis allocates portfolio overhead rather than footnoting it.

Reconciled cost basis

The chain, parameter by parameter

Results: units of value per dollar, versus cash

Mechanics, per 100,000 covered (best guess): 2.61 under-5 deaths averted (×113.16) + 1.92 over-5 deaths (×134) + 28.6 YLDs (×2.3) + 24.8% development effects + 25% other effects = 964 units of value against $90,480 of donor cost at the no-carbon rung, 0.0107 units per dollar, 3.2× the GiveDirectly benchmark. On the fully loaded $0.90 cost basis, contracted carbon covers 55% of cost, and 68% on the program-budget basis; both coverage figures are stated so neither does silent work. GiveWell lowered its funding bar from 8× to 6× cash in 2026. On the corrected school-age mortality rate the school layer clears that bar only at contracted carbon prices (7.1×); the $10 floor (5.0×) and the no-carbon rung (3.2×) sit below it, and at those rungs the school layer’s case rests on the take-home extension below.

Program+: the same sheet, extended to the household

The school chain above prices water that never leaves the school: its setting factors assume a school filter delivers 33–66% of a household filter’s benefit, and its under-five sibling row is zero. Program+, the take-home extension now entering a funded, Emory-led three-arm evaluation, is the program that fills those cells. Structurally, three things change: the sibling row activates (roughly 0.30 under-fives reached per student, study-measured), the setting factors rise toward household parity, and household parity itself is well-anchored, because the take-home hardware is the same device class the Rwanda trials randomized. The cost side changes too: household hardware is roughly ten times the school program’s cost per student-year, so this is the part of the portfolio where carbon leverage is not a bonus but the load-bearing wall.

On the gross basis Program+ sits below the cash benchmark, because household hardware is expensive per student and this dialect’s validity stack roughly halves effect sizes relative to the direct Kremer-basis chain in the donor-cost section above, which is why that section’s band reads lower. The case for Program+ stands on two legs this table makes explicit: carbon revenue on the household extension, which carries the optimistic case to about 14× cash, and the funded study, which exists to collapse the two widest bands (usage and cost-per-child) and to measure the sibling reach that separates the best-guess and optimistic columns. Each pre-study cell maps to a study output: under-five reach and spillover from Phase 3 household visits, usage from continuous Lume sensors, the setting-factor increment from the three-arm Program+ versus Program contrast, and cost-per-child from implementation tracking. When those cells fill, this table stops being a forecast.

The entire model, cell by cell

Everything below is the complete v1 workbook: every input with its provenance, every calculation in the order it runs, every scenario. Nothing is summarized away.

All rows are proforma inputs from the base scenario of the portfolio financial model; net credits carry the ~5% Article 6 host-country retirement already deducted, on the yield now in use, 0.0328 credits per person-year, adopted 27 August 2026 from the independent due diligence reconstruction under Gold Standard SDWS V2.0. Credits issued to date: 16,153 tCO₂e (Sep 2024) plus 17,758 pending.

Every figure on this page divides by this one cost. The build fails programmatically if any downstream number uses a different denominator. Deliberately excluded from this cost: an allocated share of portfolio-level management, verification, and sensor operations (an optimistic omission, priced in scenario S7), and post-2033 operating years within the 15-year crediting period (a conservative omission: the steady-state years are the cheapest). The organization's published "∼$12M investment" reconciles with this basis as a cross-check; it is not the denominator.

Recomputed robustness statement, not asserted: no single input moves the health frame anywhere near the cash-transfer bar on full cost, and none reverses the climate finding. The self-financing conclusion is the fragile one, and its reversal condition is stated in section F and S8.

The donor-cost ladder

The same program, the same DALYs, three cost bases. The metric that matters to a donor is the third.

Context for these magnitudes: the classical WHO benchmark rates an intervention “very cost-effective” below 1× GDP per capita per DALY averted (about $1,000 in Rwanda), and later opportunity-cost work puts practical thresholds nearer half of GDP per capita. Those benchmarks assume mortality-inclusive, all-population models; the strict frame here counts morbidity only, among students only. On the fund's primary metric, marginal donor cost, the school layer sits at $139 to $259 per averted DALY at contracted prices, within the WHO benchmark and above the GiveWell bar.

Where the marginal donor dollar goes: from schools into households

The $0 row is the school layer, and it is deliberately not the end of the story. School filtration reaches children only during school hours: evenings, weekends, and school breaks are unfiltered, and under-five siblings, the group at peak diarrheal mortality, never touch the school filter. The health-maximizing move is Program+: extending filtration into households through gravity household filters (~$50), personal filter bottles (~$20), and transport containers (~$5) with structured behavior change. Household filtration is itself creditable (the Tubeho Neza and LifeStraw programs are carbon-financed household filtration), so the marginal donor cost is the all-in cost net of the extension's own carbon revenue, and it falls toward zero as revenue coverage rises. A funded, Emory-led GiveWell design-and-evaluation study ($790K, August 2026 to July 2027; 15 schools in three arms with a pure control) is measuring the two numbers that decide the answer: cost-per-child and sensor-verified usage. Pre-study, the model bounds the marginal donor cost per averted DALY on the same all-cause mortality basis GiveWell itself adopted for water (Kremer et al. 2023 meta-analysis: 28–30% all-cause under-five mortality reduction; GiveWell's stricter chlorination-only pooling: 14%):

Against the bar: GiveWell targets under roughly $150 per averted DALY (top charities model at $3,000–$5,500 per life saved, about $135 per DALY at this chain's 33-YLL convention), and Renaissance Philanthropy's stated threshold is $100. The upside scenario clears both (~$58 per DALY, ~$1,900 per life saved); the central scenario (~$290, ~$210 at contracted prices) sits within roughly 1.5–2× of the bar before any further revenue growth; and every scenario falls toward $0 as third-party revenue coverage rises, which is the fund's mechanism. The strictest frame, applying the effect only to the diarrheal-attributed 9% of deaths, multiplies these figures roughly tenfold and is retained in the workbook as a sensitivity floor; GiveWell's own water models do not use it. Figures are pre-study; the study's measured cost-per-child and sensor-verified usage replace the two widest bands.

GiveWell's machinery, plus rules GiveWell doesn't need

The protocol adopts the standard machinery of professional cost-effectiveness analysis, then adds discipline specific to carbon-financed delivery.

Adversarial review before any figure is final

Every model faces a three-seat referee panel before its numbers are cited: an econometrician (recomputes every chain from raw inputs), an uncertainty analyst (attacks the sensitivity and scenario structure), and a skeptical grantmaker briefed to argue against funding, who for this portfolio also takes the carbon-market critic's chair: suppressed-demand baselines, self-reported usage, and additionality are attacked before an outside critic does. The worked example above is v1 and enters that review now; the independent, GiveWell-funded evaluation now under way (Emory, University of Arkansas, University of Michigan, Resilience Research Solutions; Matthew Freeman, Lead PI) supersedes all of it.

Delivery inputs draw on the portfolio's continuous verification layer: 1,273 monitored water sites, with pooled water quality moving from 24% of samples testing safe at baseline (n=4,943) to 90% under monitored delivery (n=1,685). See the evidence base on the main page. Sensor credibility is applied only where the sensor actually measures: water safety at the point of delivery, not household behavior, and the models keep that boundary explicit.

All seven programs, each on its own basis

Every program is modeled with the same protocol; each row links to that program's full cell-by-cell worksheet and downloadable workbook. The cost bases differ by design, so the rows are labeled rather than averaged: a single portfolio $/DALY would mix denominators, and this annex does not do that.

How to read the table. Each program divides its own cost basis by the credits it carries in the portfolio model on the yields in use, adopted 20 August 2026 from independent due diligence and, for Rwanda, 27 August. Those yields replace the ones this annex published in August: Burundi 0.163 (was 0.170), Asili 0.414 (was 0.243), Water Mission 0.204 (was 0.200), Rwanda 0.0328 (was 0.053), MWA 0.087 (was 0.103), Helvetas 0.141 (was 0.210) and LifeStraw 0.020 (was 0.049). Across the seven the portfolio moves from 5.53M to 5.25M credits for 2025–2033, because Asili rises while the rest fall. Asili’s increase comes from the wood-to-charcoal conversion factor in V2.0 applied to a 92% charcoal baseline, and it is the number here that most deserves scrutiny. Every program’s integrity-adjusted cost per tonne still sits between 3.3× and 74× below the $185 social cost of carbon on its own stated basis.

Burden inputs were corrected on 24 September against UN IGME 2024 and World Bank 2025 population shares, which the August annex carried as flagged judgment values. Burundi’s under-five death rate falls from 0.011 to 0.00965 and its under-five share from 17.5% to 15.3%, so its DALYs fall 18% and its cost per DALY rises. DR Congo’s rises from 0.014 to 0.0188 and Madagascar’s from 0.009 to 0.0129, so both programs were conservative and their figures improve. Rwanda’s school-age rate falls from 0.001 to 0.00043, which is the largest single correction in this annex. The LifeStraw Kenya health figure is withdrawn until Kenya-specific burden and incidence inputs replace the Rwanda values the chain was built from.

What does not roll up: a single portfolio $/DALY, because the denominators differ by design, and the lowest per-DALY rows (Asili, Helvetas) sit on partial cost bases, which is why the basis column exists. Only Rwanda and Burundi carry full delivery cost. All models are v1 preliminary and in adversarial review; the two flagship models are restated on the unified all-cause basis above, and the partner-program rows remain on their originally stated chains with corrected burden inputs.