Helvetas Madagascar: the Ranovola worksheet

Gravity-fed and solar-pumped chlorinated water in Diana, Menabe, and Amoron'i Mania, targeting 350,000 people by 2030.

This is the earliest-stage model in the portfolio: the program launched in 2025, no credits have been issued yet, and it is the first project on an Article 6 pathway in a country with no prior host-country authorization. The model says so, scores itself accordingly, and publishes every input anyway.

PRELIMINARY · v1 · AUGUST 2026 · EARLIEST-STAGE · NOT YET THROUGH ADVERSARIAL REVIEW

COST BASIS: CARBON FINANCING ONLY · NOT FULL DELIVERY COST · NOT GIVEWELL-COMPARABLE
01The Frame

What this model can and cannot claim

Helvetas and its donors build the water systems; the carbon channel finances verification and service. This model prices only the carbon channel, and says so on every figure.

What the cost includes

Virridy's carbon-side outlay only: the $210,000 guaranteed first purchase (the first 7,000 credits at $30) plus every later credit purchased on delivery at the contracted buy price, $2.75M in total over 2025–2033. This is the money the carbon market puts into the program.

What it deliberately does not

Partner construction capex and donor funding for the RANO WASH delivery model sit outside the portfolio financial model, so a full-delivery-cost $/DALY cannot be computed yet. Until it can, the dollars-per-DALY figures here describe what the carbon financing buys, never what the program costs, and must not be quoted against GiveWell benchmarks.

$10.31/t
carbon financing per credited tonne, 17.9× below the $185 social cost of carbon.
$20.21/t
per real tonne after the integrity adjustment, still 9.2× below the SCC.
0.51
credit integrity ratio, the portfolio's lowest: the 0.21 yield predates the 2026 methodology reconstruction, and the baseline-practice residual is hard-flagged at 0.60.
$2.16
of carbon financing per person-year of chlorinated water service; 1.27M person-years over the window, derived from credits at the booked yield.
$572
per DALY averted on the carbon-financing basis (raw; $2,083 evidence-adjusted at the portfolio's lowest composite, 0.274). A financing-efficiency figure, not program cost-effectiveness.
4,816
DALYs averted (raw central): 95% from under-five mortality at a deliberately conservative 12% all-cause reduction; the Kremer meta-analytic 25% runs as a scenario, not the central case.

Partner-published context, not a model input: Helvetas reports 51% of children under five in its program regions affected by diarrheal disease. The model's burden inputs are flagged judgment values pending country data, which is the main reason this is a preliminary.

02The Worksheet

The entire model, cell by cell

Same protocol, same transparency as the flagship worksheet: every input with its provenance, every calculation in order.

Provenance key: Proforma input Evidence-anchored Judgment · flagged Formula
↓ Download the workbook (.xlsx)
A · Program series, 2025–2033 (portfolio financial model, base scenario)
Series202520262027202820292030203120322033Total
Projected credits (tCO₂e)0011,75118,05128,55139,05149,55160,05160,051267,057
Buy price to partner ($/cr)9.209.209.309.509.609.709.809.909.90
Population served (derived)0055,95785,957135,957185,957235,957285,957285,9571,271,700 person-yrs
Virridy carbon outlay ($)00254,184171,485274,090378,795485,600594,505594,5052,753,163

Credits and buy prices are proforma inputs (contracted offtake: 51,000 credits on the voluntary market, 2028–2030; first issuance 2027, already rolled forward one year). Population is a formula, credits ÷ 0.21 yield, and is flagged twice: the yield is booked but pending the 2026 methodology reconstruction, and the derived 2030 population (186K) sits below the program's 350K target because crediting lags service by roughly two years. The 2027 outlay row includes the $210,000 guaranteed first purchase (first 7,000 credits at $30).

B · Cost basis: one denominator, honestly labeled
TOTAL CARBON-FINANCING COST, 2025–2033$2,753,163Formula first batch + on-delivery purchases; no advances for this program
Person-years of service financed1,271,700Formula sum of the derived population row
Financing per person-year$2.16Formula cost ÷ person-years

Every figure on this page divides by this one cost, and the build fails programmatically if any figure uses a different denominator. What this basis is not: it contains no partner construction capex and no donor funding, so it is not the program's cost-effectiveness. When partner cost data is available, a v2 adds the full-delivery-cost frame alongside this one.

C · Health chain inputs (financing basis)
InputValueProvenanceBasis
Under-five share of served population17%Judgment · flaggedMadagascar age structure; DHS/GBD lookup pending.
Diarrheal incidence, under-five1.0 episodes/child-yrJudgment · flaggedGBD lookup pending; high-burden setting.
Diarrheal incidence, over-five0.35 episodes/person-yrJudgment · flaggedGBD lookup pending.
Diarrhea reduction, piped treated supply30%Evidence-anchoredWolf et al. 2022 (Lancet), piped water of higher quality.
Enrollment (share of population connected)85%Judgment · flaggedNo enrollment data; program launched 2025.
Usage (share of drinking water from system)70%Judgment · flaggedNo usage survey; continuous monitoring not yet deployed at scale here, so unlike the flagship this cell is not sensor-measured.
YLD per episode0.002215 DALYFormula0.188 disability weight × 4.3 days ÷ 365.
Under-five death rate9 / 1,000 child-yrsJudgment · flaggedMadagascar order of magnitude; GBD lookup pending.
Diarrheal share of under-five deaths9%Judgment · flaggedCross-check input only; not multiplied into the central estimate (see D).
All-cause under-five mortality reduction12%Evidence-anchoredGiveWell-style conservative estimate for water treatment; the Kremer et al. 2023 meta-analytic ~25% runs as scenario S4, not the central case.
YLL per under-five death averted33Evidence-anchoredStandard life-table basis.
D · Calculation chain
StepValueComputation
Under-five person-years216,1891,271,700 × 17%
Diarrheal episodes at baseline585,618216,189 × 1.0 + 1,055,511 × 0.35
Episodes averted104,533585,618 × 30% × (85% × 70% effective coverage)
YLD DALYs averted232104,533 × 0.002215
Baseline under-five deaths in served population1,946216,189 × 9/1,000
Deaths averted (central, conservative)138.91,946 × 12% all-cause × 59.5% effective coverage
Cross-check: diarrhea-attributable pathway alone31.31,946 × 9% share × 30% × 59.5%. The all-cause estimate implies ~4.4× the diarrhea-only prediction, the pattern the Kremer meta-analysis documents. Stated, not hidden.
YLL DALYs averted4,584138.9 × 33
TOTAL DALYs averted (raw)4,816YLD + YLL; 95% mortality-driven
$ / DALY, financing basis, raw$572$2,753,163 ÷ 4,816
Evidence adjustment (E)× 0.2744,816 → 1,322 DALYs adjusted
$ / DALY, financing basis, adjusted$2,083$2,753,163 ÷ 1,322
xCash, GiveWell / egalitarian / wellbeing weights1.82× / 1.34× / 1.90×mortality weight 117 / 87 / 120 per death, morbidity 2.3 / 1.0 / 3.6 per DALY; financing basis, so not comparable to full-cost models
xCash, moral-uncertainty weighted1.72×0.45 / 0.25 / 0.30 prior
E · Evidence-adjustment pipeline
FactorScoreBasis
Internal validity0.70Meta-anchored effect sizes, but every exposure input is judgment.
External validity0.70Trial contexts transferred to rural Madagascar piped schemes.
Publication bias0.80Systematic-review anchors.
Implementation quality0.70Launched 2025, no issuance record, first Article 6 pathway with no prior host-country authorization.
Composite0.274Formula the lowest composite in the portfolio, as it should be for the earliest-stage program.
F · Credit integrity and climate metrics
Residual baseline-practice factor0.60Judgment · hard flag The 0.21 yield predates the 2026 methodology reconstruction that cut comparable reconstructed yields by ~45%; this is the portfolio's most exposed baseline. A measured baseline survey and the reconstruction replace it.
Delivery effectiveness0.85Judgment · flagged Assumed pending sensor deployment; unlike the flagship, not yet sensor-measured.
Leakage / rebound1.00Negligible for water treatment.
Credit integrity ratio0.51Formula 0.60 × 0.85 × 1.00, the portfolio's lowest.
$ / tonne, credited basis$10.31$2,753,163 ÷ 267,057 t · 17.9× below the $185 SCC
$ / tonne, integrity-adjusted$20.21$10.31 ÷ 0.51 · 9.2× below the SCC. Fails the SCC test only below integrity 0.056, implausible even under full reconstruction.
03Scenarios

One-way scenarios; no Monte Carlo on a preliminary

ScenarioVaried inputCostDALYs$ / DALYxCash GWReading
Base case$2,753,1634,816$5721.82×central, conservative mortality
S1 · first issuance slips to 2028ramp −1 yr$2,158,6583,710$5821.79×the already-realized risk repeats; economics barely move, the window truncates
S2 · contracted offtake only51,000 cr$636,510920$6921.50×downside book; $12.39 per credited tonne
S3a · integrity 0.400.40$2,753,1634,816$5721.82×$25.78 per real tonne, 7.2× below SCC
S3b · integrity 0.800.80$2,753,1634,816$5721.82×$12.89 per real tonne, 14.4× below SCC
S4 · Kremer 25% all-cause mortality0.25$2,753,1639,782$2813.66×meta-analytic upper anchor, scenario only

Recomputed robustness statement: the climate finding holds across the whole integrity band, and no scenario moves financing-basis health figures by more than ~2.4×. The binding risk for this program is not economics but issuance: registration is pending, the country pathway is new, and the ramp has already slipped once. That is a timing risk the fund's first-loss capital is designed to carry.

04Critique

The model's critique of itself

DimensionRatingAssessment
Effect-size qualityWEAKEvery exposure input is a flagged judgment pending GBD/DHS data; nothing program-specific has been measured yet. The 0.274 composite is the portfolio's lowest, deliberately.
Counterfactual robustnessWEAKUnmodeled in v1. Other WASH funders are active in Madagascar; a named-funder funging pass is required for v2. Carbon-side additionality is an assumption about the future until first issuance.
Moral-weight transparencyMODERATEAll frameworks reported. Mortality carries 95% of the DALYs, so under-five weighting drives everything; revisit when burden data arrives.
Attribution and integrityFLAGThe cost basis is Virridy's financing only, so the $/DALY is not program cost-effectiveness and is labeled as such everywhere it appears. Integrity 0.51 is the portfolio's lowest and anticipates the yield reconstruction.
Excluded factorsFLAGPartner capex (excluded by construction and labeled), time savings from piped proximity (likely material, conservative omission), water-quantity and hygiene effects, post-2033 operation.
Ranking stabilityMODERATEClimate case stable, 7.2–14.4× below SCC across the integrity band. Health figures span ~2.4× across scenarios. The decision-relevant risk is issuance timing, not cost-effectiveness.