GiveDirectly (Kenya)

Unconditional cash transfers of about $1,000 per household, sent by mobile money to poor households in East Africa. A faithful re-implementation of GiveWell's November 2024 model, Kenya column. GiveWell uses this program as the benchmark ("1× cash") for everything else it funds.

Organisation: GiveDirectly
Based on: GiveWell, GiveDirectly's Cash for Poverty Relief Program (November 2024) (spreadsheet)
Published: November 2024, retrieved October 2026
Licence: CC BY-NC-SA 3.0 US
Variants: GiveDirectly (Malawi)
Model file: givedirectly.json · change history

Our model: GiveWell's analysis plus your assumptions and our accounting adjustments (salary sacrifice, credit for shared outcomes).

1.0× as much good per dollar as GiveDirectly · rank 3 of 3

Joint low 0.30×, joint high 3.3×. In its own unit: 8.60 × 10⁻³ GiveWell units of value per dollar. Change your assumptions

70% of the uncertainty comes from judgment calls and shared assumptions.
17 inputs on the result's path have no range, so they add no uncertainty here.
Source analysis is 23 months old.

The evaluation

Summary

GiveDirectly sends about $1,000 by mobile money to each poor household it enrols in East Africa, with no conditions. GiveWell estimates the Kenya program at 0.0087 units of value per dollar. That is 2.6 times GiveWell’s older estimate for cash, the “× cash” figure it still benchmarks against. We reproduce that model exactly and use it as the site’s yardstick: every other charity is shown as a multiple of this one.

With our default assumptions the result barely moves (0.0086 units per dollar). Most of GiveDirectly’s cost is the transfers themselves, so our salary-sacrifice adjustment has little to act on. Its range runs from 0.3× to 3.3× itself. The width comes almost entirely from GiveWell’s own uncertainty about spillovers and long-run returns.

What the charity does

GiveDirectly identifies villages with high poverty rates, enrols eligible households, and sends one lump-sum transfer, now about $865 in Kenya. About 84 cents of each donated dollar reaches recipients; the rest pays for targeting, enrolment, delivery and overhead. Recipients decide how to spend it. In trials, households consume a little over half in the first year and invest the rest, mostly in livestock, roofs and small businesses.

How the impact happens

Cash raises the household’s consumption in the first year, by about 43% over a baseline of $1.79 a day.
The invested share pays returns over the next nine years.
The money spent locally raises the incomes of neighbours who didn’t receive a transfer (spillovers).
Preliminary evidence suggests the transfers reduce deaths of young children.

GiveWell values consumption on a log scale, so a dollar counts for more the poorer the person, and converts averted deaths with its moral weights.

How the model works

The model follows GiveWell’s November 2024 spreadsheet row by row. It works through program reach, then recipient consumption, spillovers and mortality. It finishes with supplementary adjustments (morbidity, development, fraud and fungibility, together +8%). Country data such as purchasing-power conversion and child mortality come from our shared Kenya facts, so other Kenya models will use the same numbers.

What drives the result

The three biggest levers are:

How much GiveWell values consumption against deaths (a moral weight). It moves the result by about ±22%.
How far to trust the spillover estimate. It rests on one study (Egger et al. 2022), significant only at the 10% level, so GiveWell discounts it heavily. Its internal and external validity adjustments each move the result by about ±20%.
How long the investment returns last. This sits within GiveWell’s ±40% range.

No single input changes the result by more than about a quarter. The joint range is wider because it sets them all at once.

Where we differ from GiveWell

Only in the adjustments we add on top. These are judgments of ours and are labelled as such.

Salary sacrifice: we treat the gap between staff pay and what staff could earn elsewhere as an in-kind donation. For GiveDirectly it is small: staff pay is about 8% of cost and the gap about 15%, so the result falls by about 1.4%.
Shared credit: GiveDirectly delivers transfers itself, so full credit and shared credit give the same answer.

Switch to GiveWell’s model above to see their figures exactly.

Biggest uncertainties

The mortality effect comes from preliminary, unpublished results. GiveWell halves it, and a different read of the final paper could move the result by more than our range shows.
Spillovers are estimated in Western Kenya in 2014–2017; today’s programs run in different markets.
Our staff cost and pay-gap figures are guesses until we extract GiveDirectly’s audited financials.

What isn’t modelled

Room for more funding, organisational strength and track record matter for a giving decision; they aren’t scored here yet. Effects on recipients’ wellbeing beyond consumption and mortality enter only through GiveWell’s supplementary adjustments.

Sources

GiveWell, GiveDirectly’s Cash for Poverty Relief Program (November 2024), and its cost-effectiveness spreadsheet. CC BY-NC-SA 3.0 US.
GiveWell, Re-evaluating the impact of unconditional cash transfers (blog, November 2024).
Egger, Haushofer, Miguel, Niehaus and Walker (2022), General Equilibrium Effects of Cash Transfers, Econometrica.

What drives this result

Each bar moves one input from the low to the high end of its range, holding the rest at their central values. Log scale, as a multiple of GiveDirectly. Dashed line: central estimate.

0.50×1.0×2.0×

Value of raising ln(consumption) by 1 for one person for one year

0.78×1.2×

External validity adjustment (spillovers)

0.79×1.2×

Internal validity adjustment (spillovers)

0.81×1.2×

Value of year 2–10 gains, relative to year 1

0.81×1.2×

Baseline annual consumption per person (PPP)

0.90×1.1×

Distributional adjustment (richer non-recipients)

0.91×1.1×

Internal validity adjustment (mortality)

0.94×1.1×

Transfers as a share of total cost

0.94×1.1×

observed estimate judgment or shared assumption

Shows the 8 inputs that move the result most. Ranges come from the source where it gives them; explainers say when a range is ours. Source: charityevals model of GiveWell, GiveDirectly's Cash for Poverty Relief Program (November 2024).

The model, row by row

Click any row for its explanation, source and formula. Formulas link to the rows they use. Values update with your assumptions.

Row	Value	Kind
Program reach How many people a fixed donation reaches. The $1M donation is arbitrary; it cancels out in the per-dollar result.
	$1.00M	Judgment
Arbitrary scale for the calculation. Any amount gives the same result per dollar. Source: cell `New CEA!B4` in the original sheet
	$865	Observed
Average one-time transfer in Kenya, from GiveDirectly's program data. Source: GiveDirectly program data via GiveWell, cell `New CEA!B5` in the original sheet
	84%79% – 89%	Observed
The rest pays for targeting, enrolment, delivery and overhead. Range is GiveWell's own ±6% Monte Carlo bound. Source: GiveDirectly financials via GiveWell, cell `New CEA!B6; Inputs!D5:F5` in the original sheet
	971	Calculated
Formula`Donation * Transfers as a share of total cost / Transfer per household (nominal USD)` The source reports 971; this model reproduces it.
shared fact	2.37	Calculated
Formula`Kenya: PPP conversion factor (2017 international $ per US $)` From the shared Kenya facts, so every model working in Kenya uses the same conversion. A dollar buys about 2.4 times as much in rural Kenya as in the US. Shared fact from Kenya: every model that needs it reads the same value. Source: World Bank, via GiveWell's GiveDirectly CEA. The source reports 2.37; this model reproduces it.
	$2,049	Calculated
Formula`PPP adjustment (2017 prices) * Transfer per household (nominal USD)` The source reports $2,049; this model reproduces it.
	4.2	Observed
Everyone in the household is assumed to share the transfer equally. Source: GiveDirectly data via GiveWell, cell `Baseline!B4` in the original sheet
	$488	Calculated
Formula`Transfer per household (PPP) / Average household size` The source reports $488; this model reproduces it.
	4,079	Calculated
Formula`Average household size * Households receiving a transfer` The source reports 4,079; this model reproduces it.
Consumption gains to recipients Recipients spend part of the transfer in year one and invest the rest, which pays off over the next nine years. Gains are valued on a log scale: a dollar is worth more to someone poorer.
	$652$554 – $750	Observed
About $1.79 a day. Poorer recipients gain more from the same transfer, which is a large part of why GiveWell's 2024 update raised its estimate. Range is GiveWell's ±15%. Source: GiveWell baseline consumption estimate, cell `Baseline!B36; Inputs!D11:F11` in the original sheet
	43%	Estimate
Spent on assets such as livestock, roofs and businesses rather than consumed in year one. Returns on it show up in years 2–10. Source: GiveWell, from GiveDirectly RCTs, cell `New CEA!B15` in the original sheet
	57%	Calculated
Formula`1 - Share of the transfer invested` The source reports 57%; this model reproduces it.
	$278	Calculated
Formula`Share consumed in year 1 * Transfer per person (PPP)` The source reports $278; this model reproduces it.
	42.7%	Calculated
Formula`Extra consumption per person in year 1 (PPP) / Baseline annual consumption per person (PPP)` The source reports 42.7%; this model reproduces it.
	0.355	Calculated
Formula`ln(1 + Increase over baseline in year 1)` The source reports 0.355; this model reproduces it.
	0.513	Calculated
Formula`Increase in ln(consumption) in year 1 * Value of raising ln(consumption) by 1 for one person for one year` The source reports 0.513; this model reproduces it.
	2,090	Calculated
Formula`Units of value per person, year 1 * People reached` The source reports 2,090; this model reproduces it.
	123%73.8% – 172%	Estimate
Returns on the invested 43%, decaying over nine years and discounted at 4% a year. GiveWell's best guess has gains fading faster than the Egger et al. long-run data suggest; with Egger persistence the bottom line rises by about 40%. Range is GiveWell's ±40%, its widest on any recipient input. Source: GiveWell Consumption tab, cell `Consumption!B23; Inputs!D23:F23` in the original sheet
	4,663	Calculated
Formula`Units of value from year 1 consumption * (1 + Value of year 2–10 gains, relative to year 1)` The source reports 4,663; this model reproduces it.
	90%	Judgment
Gains are uneven: some go to recipients above the average baseline, who value them less on a log scale. GiveWell knocks 10% off. Source: cell `New CEA!B25` in the original sheet
	4,196	Calculated
Formula`Units of value from years 1–10 * Distributional adjustment (richer recipients)` The source reports 4,196; this model reproduces it.
Spillovers to non-recipients Cash spent locally raises neighbours' incomes too. Based on Egger et al. (2022), heavily discounted for doubts about how well it replicates.
	2.5×	Observed
Each $1 transferred in Western Kenya produced about $2.50 of local economic activity. Only significant at the 10% level. Source: Egger et al. (2022), Econometrica
	70.3%	Estimate
Part of the 2.5× is the recipients' own consumption, already counted above. This keeps only the share accruing to others. Source: GiveWell GE tab, cell `GE!B7` in the original sheet
	60%27% – 93%	Judgment
Regresses a single, marginally significant study towards GiveWell's prior that spillovers are smaller. Range is GiveWell's own, and very wide. Source: cell `New CEA!B33; Inputs!D33:F33` in the original sheet
	80%32% – 128%	Judgment
How well the 2014–2017 Western Kenya result carries over to today's program. Above 100% means it could be stronger now. Source: cell `New CEA!B34; Inputs!D34:F34` in the original sheet
	80%60% – 100%	Judgment
Spillovers partly reach better-off traders and business owners, who value a dollar less. Source: cell `New CEA!B35; Inputs!D35:F35` in the original sheet
	67.5%	Calculated
Formula`Egger et al. economic multiplier * Distributional adjustment (richer non-recipients) * Share of the multiplier that isn't recipients' own gain * Internal validity adjustment (spillovers) * External validity adjustment (spillovers)` The source reports 67.5%; this model reproduces it.
	2,833	Calculated
Formula`Recipient consumption benefits * Spillover benefits as a share of recipient benefits` The source reports 2,833; this model reproduces it.
Mortality Preliminary evidence that cash reduces child deaths, halved for doubts about the study.
	12.9%	Observed
From GiveDirectly's household rosters. Source: GiveWell Equivalence tab, cell `Equivalence!B20/B41` in the original sheet
	525	Calculated
Formula`Share of people reached who are under five * People reached` The source reports 525; this model reproduces it.
shared fact	4.11%	Calculated
Formula`Kenya: Under-five mortality rate (2023)` National average, from the shared Kenya facts. Shared fact from Kenya: every model that needs it reads the same value. Source: UN IGME via GiveWell. The source reports 4.11%; this model reproduces it.
	5.7%	Observed
57 per 1,000. Compared with the national rate for the same years to estimate how much riskier GiveDirectly's target population is. Source: cell `New CEA!B44` in the original sheet
shared fact	4.63%	Calculated
Formula`Kenya: Under-five mortality rate (2016)` National rate in the Egger et al. study period, from the shared Kenya facts. Shared fact from Kenya: every model that needs it reads the same value. Source: UNICEF via GiveWell. The source reports 4.63%; this model reproduces it.
	1.23×	Calculated
Formula`Under-five mortality in the Egger et al. sample (2015–2017) / Under-five mortality, national (2016)` The source reports 1.23×; this model reproduces it.
	5.06%	Calculated
Formula`Under-five mortality rate, Kenya (2023) * Extra risk in GiveDirectly's target population` The source reports 5.06%; this model reproduces it.
	45%	Observed
A 45% drop in all-cause child mortality. Preliminary and unpublished at the time. Source: Preliminary results from the Egger et al. Kenya trial, cell `New CEA!B48` in the original sheet
	50%25% – 75%	Judgment
Halves the effect because one preliminary result this large is likely an overestimate. Range is GiveWell's own. Source: cell `New CEA!B49; Inputs!D49:F49` in the original sheet
	22.5%	Calculated
Formula`Reduction in under-five mortality from the transfer (unadjusted) * Internal validity adjustment (mortality)` The source reports 22.5%; this model reproduces it.
	5.99	Calculated
Formula`Children under five reached * Under-five mortality in the target population * Reduction in under-five mortality (adjusted)` The source reports 5.99; this model reproduces it.
	694	Calculated
Formula`Under-five deaths averted * Value of averting the death of a child under five` The source reports 694; this model reproduces it.
	50%	Judgment
Cash plausibly reduces adult deaths too, but there is little data. GiveWell adds half the child benefit. Source: cell `New CEA!B54` in the original sheet
	1,042	Calculated
Formula`Units of value from child deaths averted * (1 + Adult mortality benefits, as a share of child benefits)` The source reports 1,042; this model reproduces it.
Supplementary adjustments Effects GiveWell doesn't model explicitly, each as a % of the modelled benefits.
	10%	Judgment
Fewer illnesses short of death. Source: cell `New CEA!B59` in the original sheet
	5%	Judgment
Long-run gains for children, e.g. from better nutrition. Source: cell `New CEA!B60` in the original sheet
	2%	Judgment
Net effect of the donation drawing in, or displacing, other funders. Source: cell `New CEA!B61` in the original sheet
	-2%	Judgment
Transfers that are diverted or lost. Source: cell `New CEA!B62` in the original sheet
	-5%	Judgment
Money given for one country may free up GiveDirectly money for less cost-effective work. Source: cell `New CEA!B63` in the original sheet
	-2%	Judgment
Non-recipients may feel worse off by comparison. Source: cell `New CEA!B64` in the original sheet
	108%	Calculated
Formula`1 + Reduced morbidity + Developmental effects + Leverage and funging + Risk of wastage or fraud + Within-organisation fungibility + Psychological spillovers` The source reports 108%; this model reproduces it.
Bottom line
	8,071	Calculated
Formula`Recipient consumption benefits + Spillover benefits + Mortality benefits` The source reports 8,071; this model reproduces it.
	8,717	Calculated
Formula`Modelled benefits (consumption, spillovers, mortality) * Total adjustment factor` The source reports 8,717; this model reproduces it.
	8.72 × 10⁻³	Calculated
Formula`Total benefits / Donation` The source reports 8.72 × 10⁻³; this model reproduces it.
	3.35 × 10⁻³	Estimate
GiveWell still expresses everything as a multiple of this older figure ("× cash"). Kept so this model reproduces GiveWell's headline numbers. Source: cell `New CEA!B28` in the original sheet Context only: the bottom line doesn't use this row.
	2.6×	Calculated
Formula`Units of value per dollar / GiveWell's previous estimate for cash (units of value / $)` The source reports 2.6×; this model reproduces it. Context only: the bottom line doesn't use this row.
Your dollar's share charityevals adjustments on top of GiveWell's model. With the default assumptions they change little for GiveDirectly, because most of its cost is the transfers themselves.
charityevals	8%4% – 12%	Judgment
About 16% of GiveDirectly's costs are not transfers (see the cost share above). We guess half of that is staff pay. GiveDirectly publishes audited financials that could pin this down; we haven't extracted them yet. Range is ours.
charityevals	15%0% – 35%	Judgment
How much less staff earn than they could elsewhere, as a share of what they'd earn. Field staff in Kenya are paid at or near local market rates; headquarters staff in tech, finance and operations likely take a larger cut. Range is ours.
	$14,118	Calculated
Formula`Donation * Staff pay as a share of total cost * Pay gap versus comparable for-profit jobs / (1 - Pay gap versus comparable for-profit jobs)`
	98.6%	Calculated
Formula`Donation / (Donation + if(Count staff salary sacrifice as a donation, Salary forgone by staff (in-kind donation), 0))`
	8.60 × 10⁻³	Calculated
Formula`Units of value per dollar * Share of the full cost paid by cash donations`

1Sinergia Animal157×
2GiveDirectly (Malawi)1.5×
3GiveDirectly (Kenya)1.0×

Your assumptions

Moral weights and exchange rates are judgment calls. Drag them and every result on the site recomputes.

How impact is counted

Count staff salary sacrifice as a donation

What this means

Many nonprofit staff earn less than they could in comparable for-profit jobs. That gap is a donation in kind: the staff pay part of the cost of the impact. When this is on, the cost of the work is cash spending plus that forgone salary, so your dollar is credited with only its share of the impact.

When it's off, as in GiveWell's and ACE's models, all of the impact is credited to cash donations. Each model estimates its own staff costs and pay gap, and says how.

Give full credit for outcomes shared with other organisations

What this means

Most outcomes need several organisations: a corporate pledge follows campaigns by several groups. Off (the default, and ACE's approach) credits each charity with its estimated share. On credits each with the whole outcome, on the view that removing any necessary link would have stopped it.

Shares never add up across charities under full credit, so comparisons between charities that share outcomes get harder. Use it to see how much a ranking depends on the choice.

Exchange rates between causes

Units of value per human DALY averted2

Why this value

Bridge from health to GiveWell's units. A rough derivation: GiveWell values an under-five death at 116 units, and a child death in sub-Saharan Africa costs roughly 55–65 years of life, so one year of healthy life is worth about 2 units.

This ignores GiveWell's age weighting and treats years lost to disability like years lost to death. Ours, not GiveWell's.

Human DALYs per suffering-adjusted day1.10 × 10⁻³

2.70 × 10⁻⁴2.70 × 10⁻³

Why this value

Suffering-adjusted days (SADs), from Ambitious Impact, count days of animal suffering weighted by pain intensity and by a species' welfare range relative to humans. ACE uses them.

We read one SAD as one human-equivalent day of disabling-level pain and give that day a disability weight of 0.4 (0.1–1.0), so one SAD is 0.4 / 365 DALY. AIM doesn't publish an official SAD-to-DALY rate, so this is the least grounded number on the site and often decides whether animal charities rank above or below global health.

Extra weight on animal suffering, beyond welfare ranges1×

0.05×1×

Why this value

SADs already scale suffering by each species' welfare range. At 1 you take those welfare ranges at face value. Below 1 you discount animal suffering further, for example because you're unsure animals are sentient or because you give more weight to humans for other reasons.

Moral weights

Value of raising ln(consumption) by 1 for one person for one year1.44

1.081.8

Why this value

GiveWell defines one unit of value as doubling one person's consumption for one year. Doubling raises ln(consumption) by ln 2, so one unit of ln(consumption) is worth 1/ln 2 ≈ 1.44 units.

The low and high ends are the ±25% uniform range GiveWell itself uses in its Monte Carlo inputs. Moving this moves the value of consumption gains relative to deaths averted.

Value of averting the death of a child under five116

70160

Why this value

GiveWell's moral weight: averting one under-five death is worth as much as doubling consumption for 116 person-years. It comes from a mix of donor surveys, a survey of people in Ghana and Kenya, and staff judgment.

GiveWell doesn't vary this in its own uncertainty analysis. The 70–160 range is ours, to show how much the ranking leans on it.

Economics

Annual discount rate4%

0%8%

Why this value

How much less a benefit a year from now counts than one today, compounding. GiveWell and ACE both use 4%.

Used where a model discounts explicitly (ACE's cage-free campaigns). GiveWell's GiveDirectly model has the 4% built into its year 2–10 consumption ratio, so changing this doesn't affect it yet.

1.0× as much good per dollar as GiveDirectly · rank 3 of 3

The evaluation

Summary

What the charity does

How the impact happens

How the model works

What drives the result

Where we differ from GiveWell

Biggest uncertainties

What isn’t modelled

Sources

What drives this result

The model, row by row

Program reach

Consumption gains to recipients

Spillovers to non-recipients

Mortality

Supplementary adjustments

Bottom line

Your dollar's share