Kenya's long-term basic income study: cash works, and how it is delivered matters

In 2017 the charity GiveDirectly began the largest and longest randomised study of a universal basic income yet attempted: in 195 villages in two rural Kenyan counties, every adult was enrolled in one of three cash programmes, while 100 further villages served as controls. One group was promised monthly payments for twelve years, another received the same monthly payments for two years, and a third received roughly the same total as a single lump sum. The first full results, in a working paper by Abhijit Banerjee, Michael Faye, Alan Krueger, Paul Niehaus and Tavneet Suri dated 15 September 2023, cover the first two years, when all three groups had received about the same amount of money. They show local economic expansion, no fall in total work, a shift from wage labour to self-employment, and lower depression — and they show that how cash is delivered matters as much as how much is delivered. This study places those findings alongside the earlier GiveDirectly evidence by Johannes Haushofer and Jeremy Shapiro and by Dennis Egger and colleagues, and sets out what the Kenyan evidence can and cannot tell societies debating basic income in very different economies.

Key findings

• Design. Of 295 study villages in Bomet and Siaya counties, 100 were controls; in 44 villages every adult received about US$0.75 a day ($1.88 at purchasing-power parity), paid monthly, for twelve years; in 80 villages adults received the same payments for two years; and in 71 villages adults received a one-off payment of about $500 ($1,250 PPP), delivered in two instalments two months apart. About 23,000 adults were treated, each individually rather than as a household.
• Local economies grew. In long-term villages, the net revenues of local enterprises rose by slightly more than 50%; almost all of the expansion came from non-agricultural businesses, mainly retail, whose number rose by 25%. Consumer prices in nearby markets showed no statistically significant change.
• Total work did not fall. Households reduced hours of wage work in agriculture and increased hours of non-agricultural self-employment by slightly more, leaving total household labour supply unchanged. The authors find "no evidence of UBI promoting 'laziness'".
• The lump sum matched or beat the long-term stream on economic outcomes; the two-year stream had noticeably smaller effects on enterprises, revenues and costs, despite delivering the same money to date.
• Wellbeing improved, but differently by arm. All three arms reduced depression; the reduction in the lump-sum arm was about half that in the long-term arm. Self-reported happiness rose by similar amounts in all arms.
• Earlier GiveDirectly trials point the same way. Haushofer and Shapiro (2016) found household consumption rising from $158 to $193 PPP a month nine months after transfers began, and large gains in psychological wellbeing. Egger and colleagues (2022) found a local transfer multiplier of 2.5 and minimal price inflation after one-off transfers of about $1,000 to more than 10,500 households in 653 villages.
• The evidence is early. The 2023 results cover two years of a twelve-year programme, come from a working paper, and describe rural Kenya with near-universal mobile money. They are strong evidence about cash in poor rural economies and weak evidence about basic income in rich ones.

Research question and scope

This study asks what the Kenyan UBI experiment shows, two years in, about four questions that dominate basic income debates. Does unconditional cash make people work less? Does it generate local economic activity or just consumption? Does a long-term guarantee change behaviour compared with a temporary programme or a lump sum of equal value? And does cash improve wellbeing?

The core source is the September 2023 working paper by Banerjee, Faye, Krueger, Niehaus and Suri. We also use GiveDirectly's published descriptions of the programme, two peer-reviewed studies of earlier GiveDirectly transfers in the same region (Haushofer and Shapiro, 2016; Egger, Haushofer, Miguel, Niehaus and Walker, 2022), and the review of UBI in developing countries by Banerjee, Niehaus and Suri (2019). We do not evaluate GiveDirectly as an organisation, and we do not attempt to project the twelve-year results.

Background

Why test basic income in Kenya?

Banerjee, Niehaus and Suri argue that the case for universal transfers in low-income countries rests on a dilemma. Governments often lack the data and capacity to target transfers well — identifying the poor imposes time and psychological costs on applicants and produces errors that exclude some of the most needy. Yet they also face tight budgets, which makes universal transfers expensive. If cash raises recipients' earnings, through investment, risk-taking or local demand, some of that tension eases. The Kenyan experiment was designed to test whether it does.

Delivery infrastructure

The study would have been far harder elsewhere. Kenya has one of the world's most developed mobile money systems: as early as 2014 an estimated 96% of Kenyan households had a mobile money account, mostly with M-PESA. Suri and Jack (2016) estimated that access to M-PESA itself lifted 194,000 households, about 2% of Kenyan households, out of poverty. GiveDirectly paid all transfers by mobile money, which made monthly payments to tens of thousands of individuals cheap and traceable; every household in the study area had access to at least one mobile money agent.

The setting

Bomet and Siaya are among Kenya's poorer counties. At baseline, 85% of households reported experiencing hunger at least once in the previous year, maize consumption — typically about 40% of total consumption — was $0.60 PPP per person per day, 73% of households had a farm enterprise and 21% a non-farm enterprise, and households owned 1.7 acres of land on average. 86% had a phone but only 13% a bank account. Villages averaged 52 households and 118 eligible adults, and the average village had three markets within 5 km.

Method and sources

Sampling and randomisation

The researchers listed all 390 villages in four sub-counties, kept 325 villages of 30–100 households (30–73 in Bomet) to improve statistical power for a fixed budget, and randomly selected 295 for the study. A census in April–June 2017 counted 15,189 households and 73,511 people, of whom 23,082 were adults eligible for transfers. Randomisation was stratified by location, a geographic unit containing about 14 villages.

The long-term transfer was calculated as enough to cover the most basic needs, such as staple food and perhaps some health and education spending. Teenagers aged 15–17 in long-term villages were told they would start receiving payments when they turned 18. The lump sum was designed to equal, in present-value terms, the short-term stream discounted at 8% a year. GiveDirectly's public description gives the monthly amount as $22.50 and the lump sum as about $500; it describes the project as costing about $30 million.

Two additional "nudges" — encouragement to plan spending and a reminder about an interest-bearing mobile savings account — were cross-randomised at household level but are not analysed in the paper. Payments started between January and June 2018. People who moved into a treatment village after enrolment were not eligible.

Data

The researchers surveyed 30 households per village. Baseline work in autumn 2017 included in-person surveys of 8,753 household heads and phone surveys of 4,383 spouses on sensitive topics such as domestic violence. The endline, between October 2019 and January 2020, included 8,292 household heads, 407 heads of migrant households and phone surveys of 1,034 individual migrants; 6,969 spouses; 295 village elders; price surveys in 104 markets with up to three observations each for up to 175 goods; a census of 9,850 non-agricultural enterprises; and detailed surveys of 3,418 enterprises. Household response rates reached 97% in all arms. A short phone survey in autumn 2022 added data on land values, alcohol use and rotating savings and credit associations (ROSCAs). Ten families of outcomes were pre-registered.

Timing and comparability

At the endline, the lump-sum, long-term and short-term arms had received on average 92%, 87% and 92% of the two-year amount. The comparison therefore holds the cumulative money roughly constant: the arms differ in timing (all at once versus monthly) and in expectations (only the long-term arm expected more to come).

How the atlas weighs the evidence

The design is strong: large, cluster-randomised, pre-registered, with high response rates and village-level measurement of spillovers. The main caveats are that the paper is a working paper rather than a peer-reviewed article in the version we cite (a Crossref search in October 2026 found no journal version; the paper was publicly evaluated by The Unjournal in 2024), that the comparisons between treatment arms were not powered for two-sided tests of equality, and that the authors include GiveDirectly's co-founder and a board member (Faye and Niehaus serve without compensation as GiveDirectly directors, a fact the paper discloses). We read the comparisons with the control group as well established and the comparisons between arms as suggestive.

Findings: the village economy

Enterprises and revenues

The most striking result is at the village level. In long-term villages, enterprise revenues and costs both rose significantly, and net revenues — revenues minus costs, excluding the cost of household labour — rose by slightly more than 50%. Almost all of the change came from non-agricultural enterprises, whose net revenues nearly doubled and whose number rose by 25%. Within that sector, nearly the entire effect came from retail. The number of items sold within 2 km of the village doubled, an effect that narrowly missed statistical significance (p = 0.114). Households spent a larger share of their budgets inside their own village. The authors describe the picture as "an expanding distribution system", financed in part by the transfers and meeting the extra local demand they created.

Agricultural enterprises grew less in proportional terms and not significantly, although the value of agricultural assets rose by 35%, partly through higher land prices and partly through more livestock. The total value of transfers in a treated village amounted to about 15% of total spending in control villages — a large local shock.

Prices, wages and migration

The authors could not reject the hypothesis that consumer prices in nearby markets were unchanged, for both agricultural and non-agricultural goods, although their estimates were not precise enough to rule out meaningful changes. In the same region, Egger and colleagues had precisely estimated consumer-goods inflation of 0.1% after transfers equal to about 15% of local GDP. Wages paid to labour hired from outside the village for agricultural and construction work rose in all arms, mostly significantly, and reservation wages for public-works jobs rose by smaller amounts, suggesting that higher labour demand rather than reduced supply drove the change. Treated villages also attracted new residents: lump-sum villages had 13% more households.

Labour supply and occupations

Total household labour supply did not change. The long-term arm saw a significant reduction in hours of wage work, all in agriculture, and a slightly larger increase in non-agricultural self-employment. Equilibrium wage employment did not change, which the authors interpret as workers from outside the village taking up the slack. In the authors' words, "there is no evidence of UBI promoting 'laziness,' but evidence of substantial effects on occupational choice."

Findings: households

Consumption and food

Total consumption effects were moderate: between 6% and 11% of the control median, and between 19% and 33% of the average amount transferred. The long-term arm had the smallest effect (about 5%) and the only one not statistically significant. Food consumption rose by similar, significant amounts in all arms, with gains in protein consumption and food variety; in the long-term arm protein consumption rose by close to 25% at the bottom of the distribution and closer to 15% near the top. Respondents in treated villages reported seeing fewer neighbours drinking daily and were less likely to see drinking as a problem.

Assets, savings and land

Productive assets for agriculture and for non-agricultural uses increased substantially; financial assets rose modestly, with significant increases in liquid savings and no significant change in debt. Combining these with consumer durables, households accumulated between $580 and $1,150 of non-land assets, depending on the arm, against roughly $3,000 of transfers. Self-assessed land values rose by even more than the transfers received, entirely through appreciation of existing land rather than purchases. The authors treat this cautiously: part may reflect investment in structures, but part may simply reflect a belief that treated villages had become more desirable places to live.

Participation in ROSCAs — informal groups whose members pool regular contributions and take turns receiving the pot — was "enormously higher" in the long-term arm than in the others. This is consistent with the authors' interpretation that long-term recipients saved up for larger investments.

Education, health and child outcomes

Education spending rose in all arms, significantly in the short-term and lump-sum arms, but years of schooling did not change. The share of children scoring more than 250 in the national primary-school examination (KCPE) rose significantly, by about 14%, only in the lump-sum arm. Health spending fell slightly in two arms (not significantly) and rose significantly in the lump-sum arm. A pre-specified child anthropometric index improved only in the short-term arm, which also had the largest effect on food consumption.

Mental health and wellbeing

All three treatments significantly reduced standard measures of depression (the CES-D scale), but the reduction in the lump-sum arm was about half that in the long-term arm, a significant difference. All three raised self-reported happiness on Cantril's 0–10 ladder by similar amounts, about 0.15 of a step. Life satisfaction improved most clearly in the short-term arm. Domestic violence fell in all arms, significantly only in the lump-sum arm. One result is puzzling: the long-term transfer had the most negative effect on a locus-of-control measure, appearing to make households feel less in control of their lives even as it made them happier.

Findings: does the delivery design matter?

Source: Banerjee, Faye, Krueger, Niehaus and Suri (2023). Comparisons between arms were not powered for two-sided tests; differences described as "larger" or "smaller" are not all statistically significant.

The authors interpret the pattern through a simple model of savings- and credit-constrained households, some of which have access to a business opportunity requiring a lumpy up-front investment with high returns afterwards. Lump-sum recipients can make the investment immediately. Long-term recipients have a reason to save towards it, because they expect a long stream of payments to come; ROSCAs are one way of doing so. Short-term recipients, who receive the same money but know it will stop soon, face neither the immediate capital nor the long horizon, and so spend more on consumption, food and household items. The authors stress that this is not the only possible explanation, but argue that savings and credit constraints, and some degree of local increasing returns, must play a role in any explanation.

GiveDirectly's own summary draws a policy conclusion: short-term monthly payments — the design used by most UBI pilots and much cash assistance — performed least well on nearly all economic measures, though they still had positive effects. This is a fair reading of the point estimates, but the atlas notes that the short-term arm had the clearest gains in food consumption, child anthropometrics and life satisfaction. Which design is "best" depends on whether the objective is investment and enterprise or immediate welfare.

Related evidence from Kenya

Haushofer and Shapiro (2016)

The first major randomised evaluation of GiveDirectly, published in the Quarterly Journal of Economics, studied transfers in rural Kenya that were unconditional, large and concentrated in time. The researchers randomised at village and household levels and, within the treatment group, randomised the recipient's gender (wife or husband), the timing (lump sum or monthly instalments) and the amount ($404 PPP or $1,525 PPP). Household monthly consumption rose from $158 PPP to $193 PPP nine months after the transfers began, and recipients experienced large increases in psychological wellbeing. There was no overall effect on the stress hormone cortisol. Monthly transfers were more likely to improve food security, while lump sums were more likely to be spent on durable goods — an early sign of the savings and credit constraints that the UBI study later examined at scale.

Egger, Haushofer, Miguel, Niehaus and Walker (2022)

This study, published in Econometrica, provided one-time transfers of about $1,000 to more than 10,500 poor households across 653 randomised villages, a fiscal shock of more than 15% of local GDP. It found large effects on recipients' consumption and assets, large positive spillovers to non-recipient households and firms, minimal price inflation and an estimated local transfer multiplier of 2.5. It is the strongest evidence that cash in such settings stimulates local economies rather than simply bidding up prices.

What the evidence does not show

• It does not yet show long-term effects. The 2023 results describe the first two years of a twelve-year programme. Whether long-term villages pull ahead, converge with lump-sum villages or fall back as novelty fades is unknown. A separate working paper examined the COVID-19 period, but the main long-run results are still to come.
• It does not show what a national UBI would do to public finances. The transfers were funded by private donors, not by taxes. A government would have to finance them, with consequences that a donor-funded experiment cannot capture, and Kenya's own social spending is far below the amounts transferred in the study. GiveDirectly notes that Kenya's spending on all social services amounted to $93 per person in the year before its 2023 summary.
• It does not transfer directly to rich economies. The mechanisms that drive the Kenyan results — credit constraints, lumpy business investments, idle local capacity and a large informal sector — are far weaker in countries with developed financial systems and existing safety nets. As Verho and colleagues note in their study of Finland, findings from developing countries are not directly applicable to advanced welfare states.
• It does not separate universality from cash. Because every adult in treated villages was enrolled, the village-level results combine individual responses with spillovers. That is a strength for studying UBI, but it means the results cannot simply be scaled from individuals.
• It does not resolve the land-value puzzle. Reported land appreciation exceeded the transfers. Whether this is real wealth creation, a local price effect or optimism is not settled.
• It does not settle which delivery design is best. The arm comparisons were not powered for two-sided tests, and different arms performed best on different outcomes.

Implications for institution design

1. Duration and credibility are design variables. The same money produced different behaviour depending on whether people expected it to continue. A basic income that is credibly permanent may encourage saving and investment that a temporary pilot does not. Pilots of a few years therefore systematically understate — or at least mismeasure — the behavioural effects of a permanent scheme.
2. Lump sums deserve a place in the toolbox. Where households face lumpy investment opportunities and weak credit markets, a single larger payment may achieve more than an equivalent stream. Some programmes could let recipients choose; an earlier GiveDirectly study cited by the authors found that recipients offered the choice overwhelmingly chose lump sums.
3. Measure the village, not only the household. The largest effects in Kenya were in local enterprise, wages and migration. Evaluations that study only recipients miss spillovers to non-recipients, positive or negative.
4. Infrastructure comes first. Cheap, secure, individual-level delivery through mobile money made the experiment possible. States considering universal transfers need payment rails, identification and grievance mechanisms before they need a benefit level.
5. Individual entitlement changes household dynamics. Paying every adult individually meant that a woman leaving her partner could take her remaining transfers with her. Whether transfers are paid to individuals or households is a substantive design choice, not an administrative detail.
6. Do not overstate portability. The Kenyan results are a strong argument for cash transfers in poor rural economies. They are a weak argument, on their own, for or against basic income in high-income countries facing automation.

Open questions

• How will long-term villages compare with lump-sum and short-term villages after the short-term payments end and as the twelve-year programme continues?
• What happens when the long-term payments stop, and do households anticipate the end years in advance?
• Do the enterprise gains persist, or were they a one-time expansion of retail to meet a temporary demand shock?
• How would results change if transfers were financed by local or national taxation rather than external donors?
• What explains the negative effect of long-term transfers on locus of control?

Sources

1. Banerjee, A., Faye, M., Krueger, A., Niehaus, P. and Suri, T. (2023). Universal Basic Income: Short-Term Results from a Long-Term Experiment in Kenya. Working paper, 15 September 2023. PDF
2. GiveDirectly. "Basic Income" — programme and study description. givedirectly.org
3. GiveDirectly (2023). "Early findings from the world's largest UBI study." 6 December 2023. givedirectly.org
4. Haushofer, J. and Shapiro, J. (2016). "The Short-term Impact of Unconditional Cash Transfers to the Poor: Experimental Evidence from Kenya." Quarterly Journal of Economics 131(4): 1973–2042. doi:10.1093/qje/qjw025
5. Egger, D., Haushofer, J., Miguel, E., Niehaus, P. and Walker, M. (2022). "General Equilibrium Effects of Cash Transfers: Experimental Evidence From Kenya." Econometrica 90(6): 2603–2643. doi:10.3982/ECTA17945
6. Banerjee, A., Niehaus, P. and Suri, T. (2019). "Universal Basic Income in the Developing World." Annual Review of Economics 11: 959–983. doi:10.1146/annurev-economics-080218-030229
7. Suri, T. and Jack, W. (2016). "The long-run poverty and gender impacts of mobile money." Science 354(6317): 1288–1292. doi:10.1126/science.aah5309
8. Gentilini, U., Grosh, M., Rigolini, J. and Yemtsov, R. (eds) (2020). Exploring Universal Basic Income: A Guide to Navigating Concepts, Evidence, and Practices. World Bank. PDF
9. Verho, J., Hämäläinen, K. and Kanninen, O. (2022). "Removing Welfare Traps: Employment Responses in the Finnish Basic Income Experiment." American Economic Journal: Economic Policy 14(1): 501–522. doi:10.1257/pol.20200143
10. Jones, D. and Marinescu, I. (2022). "The Labor Market Impacts of Universal and Permanent Cash Transfers: Evidence from the Alaska Permanent Fund." American Economic Journal: Economic Policy 14(2): 315–340. doi:10.1257/pol.20190299
11. Vivalt, E., Rhodes, E., Bartik, A. W., Broockman, D., Krause, P. and Miller, S. (2026). "The Employment Effects of a Guaranteed Income: Experimental Evidence from Two U.S. States." Quarterly Journal of Economics 141(4): 2675–2734. doi:10.1093/qje/qjag042