Published on Development Impact

Yes we can! Unbundling poverty graduation programs for cost-effective impact

This page in:

There has been a fair amount of discussion in this blog of multifaceted, poverty graduation programs (now often called ‘economic inclusion’ programs). Berk covered results from a big multicountry study, as well as some related qualitative work and Markus has talked about longer run impacts from Bandiera and co.

These can be complex, expensive programs. There are serious questions about whether they can be scaled by governments and whether some components might be more cost effective than others. Today we bring you some neat new results from a paper that we wrote with B Karimou, D Karlan, H Kazianga, W Pariente, C Thomas, C Udry and K Wright, and came out in Nature a week ago (you can also access the longer but older working paper version here).

The setting is Niger, one of the poorer countries on the planet. The women who are to be the main beneficiaries face a range of challenges, for example: 7 percent of them are literate, they live on average over an hour from the nearest market, and three quarters of them live in households who reported changing consumption in response to a shock in the last year.  In the face of this extreme poverty, we take a look at three permutations of a graduation program added to the government regular cash transfer program (which provides about $16 per month). Before getting to the variants, there is a base set of interventions common to all that includes savings groups, coaching and entrepreneurship training. In addition, one group (randomly selected) gets a lump-sum cash grant (about US$127) (capital package). Another group gets a psychosocial set of interventions that includes socioemotional skills training plus community sensitization on aspirations and social norms (psychosocial package). And the third group gets the full package. We compare the impacts of these different variants to a control group of folks getting only regular cash transfers.

This program was designed to be much cheaper than graduation programs so it could be scaled through government systems.  It reached around 15,000 beneficiary households in the phase that we look at. And implementation was led by the government – building on the cash transfer program infrastructure. Much of the actual activities were run by government field agents, with the exception of some of the training where NGOs were hired to bring on trainers on shorter-term contracts.

We use three rounds of data to estimate impacts, with one follow up 6 months after the interventions and another at 18 months.  This figure from the paper summarizes a lot of the main outcomes:

figure showing impacts on a range of outcomes


There were significant and meaningful impacts on consumption and food security. For the capital and full arms, these happened by the first follow up and were sustained in the second follow up. The psychosocial package started with lower impacts but grew between the first and second follow-up.


One of the main mechanisms for these consumption impacts was new businesses and business revenue overall. The beneficiary herself was significantly more likely to start a business, but so did others in the household. By the second follow up annual household business revenue was up by $318 in the capital arm, $334 in the psychosocial arm, and $541 for the full package. In a country where the GDP per capita in 2019 was $554, these are meaningful figures (Kathleen blogged about these businesses last week). And, in case you were wondering, business profits also show significant impact.  Livestock revenue, while contributing less to total household revenue than businesses, was also boosted by all three arms, with the full (+$73) and capital (+$70) packages outpacing the psychosocial arm (+$35). Farm revenues tell a different story with the psychosocial (+$91) and full (+$80) arms outpacing the capital arm (+$60) at the second follow up.

Other dimensions of well-being also show meaningful improvements. As the figure indicates, we see significant improvements in mental health and self efficacy for all arms. Focusing on mental health, which measured things like life satisfaction, inner peace and depression, we see effects ranging from 0.15 s.d. in the capital arm, 0.21 s.d. in the psychosocial arm and 0.26 in the full arm at the second follow up.  Looking at an index of social cohesion and community closeness, which included things like trust in other community members and reported tensions, we see all three arms having significant impacts at the second follow up. However, when we dig a little deeper into women’s empowerment, we see a hint of two different stories emerge. For the psychosocial arm, there are significantly larger impacts on social cohesion and community closeness at the second follow up than for either the capital arm or the full arm. This effect seems to be driven by things like a reduced likelihood of having enemies and caring more about the village. And women also seem to report better perceptions of within-household relations. The capital and the full arms, on the other hand, show higher impacts (significantly different at 10 percent) on women’s control over earnings than the psychosocial arm, driven in part by increased livestock assets and control over these revenues.  Finally, in no arm do we see an increase in women’s control over household decision making power, perhaps because their increase in revenue isn’t enough to get them close to 50 percent of total household revenue. 

Overall, these three low-cost packages show remarkable impacts, and those impacts are sustained over time. But which is the most cost effective? Focusing on consumption impacts and leaving aside asset accumulation and other things(where there were some differences across arms), we compute a range of rates of return. Taking a conservative assumption that benefits dissipate by 50% per year after the second follow up, the internal rate of return for the psychosocial arm comes out at 66 percent, while the full comes in at 44 percent and the capital is at 15 percent. Both packages with psychosocial interventions are more cost-effective than the capital package. Taken together these returns give us markedly higher cost-benefit ratios than some other graduation evaluations (such as the discussed by Berk or Bedoya and co. in Afghanistan)

This work is the first in a series of four countries in which this type of un-bundling economic inclusion programs delivered through government systems. Stay tuned for more results. 


Thomas Bossuroy

World Bank Sr. Economist

Markus Goldstein

Lead Economist, Africa Gender Innovation Lab and Chief Economists Office

Julia Vaillant

Senior Economist, World Bank’s Africa Gender Innovation Lab

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000