1. 1 A MULTIFACETED PROGRAM CAUSES LASTING PROGRESS FOR THE VERY POOR:EVIDENCE FROM SIX COUNTRIES BY : ABHIJIT BANERJEE, ESTHER DUFLO, NATHANAEL GOLDBERG, DEAN KARLAN,* ROBERT OSEI, WILLIAM PARIENTÉ, JEREMY SHAPIRO, BRAM THUYSBAERT, CHRISTOPHER UDRY 30 th May 2015 This work was carried out with the aid of a grant from the International Development Research Centre, Canada.

2. 2 About this Study Funded by Ford Foundation 3ie U.S. Agency for International Development (USAID) Study received approval from Yale University Human Subjects Committee MIT Human Subjects Committee, IRB Innovations for Poverty Action Human Subjects Committee A multifaceted program causes lasting progress for the very poor:Evidence from six countries Sites : Ethiopia, Ghana, Honduras, India, Pakistan, Peru

3. 3 Originates from, BRAC, a big Bangladeshi NGO that originally came up with this approach to tackle abject poverty, calls it a “graduation programme”. Various NGOs, including Heifer International, Oxfam and World Vision, give cows, goats or chickens to poor people in developing countries This is to enable them to earn an income selling milk or eggs. But what if the recipients are so hungry that they end up eating their grants. BRAC’s idea was to give those in the graduation programme not just chickens but also training on how to keep them, temporary income support to help them to resist the inevitable temptation to eat them, and repeated visits from programme workers to reinforce the training and bolster participants’ confidence.

4. 4 Overview of the Study Problem Investigate whether a multifaceted graduation program can help the extreme poor establish sustainable self-employment activities and generate lasting improvements in their well-being. How they did? The program targets the poorest members in a village and provides Productive asset grant, Training and support, Life skills coaching, Temporary cash consumption support, Access to savings accounts Health information or services

5. 5 What is the rationale for doing this? This multipronged approach is relatively expensive, but the theory of change is that the combination of these activities is necessary and sufficient to obtain a persistent impact. We do not test whether each of the program dimensions is individually necessary. But examined the “sufficiency” claim: A year after the conclusion of the program, and 3 years after the asset transfer, are program participants earning more income and achieving stable improvements in their well-being?

6. 6 Where they did this ? Ethiopia, Ghana, Honduras, India, Pakistan, and Peru – Representing 3 continent Implementing partners selected eligible villages based on being in geographies associated with extreme poverty. Then identified the poorest of the poor in these villages through a participatory wealth-ranking process Participatory Wealth Ranking is a method whereby communities themselves define who the poorest or the better-off are About half the eligible participants were assigned to treatment, and half to control. We conducted a baseline survey on all eligible participants, as well as an endline at the end of the intervention (typically 24 months after the start of the intervention) and a second endline 1 year after the first endline

7. 7 What is the outcome of the study? Results At the end of the intervention, we found statistically significant impacts on all 10 key outcomes or indices. One year after the end of the intervention, 36 months after the productive asset transfer, 8 out of 10 indices still showed statistically significant gains. monthly consumption of food had risen by around 5%, Household income had also risen, he value of participants’ assets had increased by 15%, person in the programme spent an average of 17.5 more minutes a day working Conclusion Primary goal, to substantially increase consumption of the very poor, is achieved by the conclusion of the program and maintained 1 year later. The estimated benefits are higher than the costs in five out of six sites. We establish that a multifaceted approach to increasing income and well- being for the ultrapoor is sustainable and cost-effective

8. 8 Study Timeliness How they implemented this study? About half the eligible participants were assigned to treatment, and half to control. In three of the sites, to measure within village spillovers, we also randomized half of villages to treatment and half to control.

9. 9 How they Implemented this Study?

10. 10 Household Eligibility Criteria

11. 11 Intervention - Saving & Health

12. 12 Intervention - Asset Transfer & Consumption Support

13. 13 Interpreting Values ( Standardized Mean, Q-value, F test for equality) Standardized Mean We construct indices first by defining each outcome Y ijl (outcome k, for observation i in family j, within country l) so that higher values correspond to better outcomes. Then we standardize each outcome into a z-score, by subtracting the country control group mean at the corresponding survey round and dividing by the country l’s control group standard deviation (SD) at the corresponding survey round We then average all the z-scores, and again standardize to the control group within each country and round Q- Value Corrected for the potential issue of simultaneous inference using multiple inference testing Calculated q-values using the Benjamini-Hochberg step-up method to control for the false discovery rate. Our q-value is the smallest a at which the null hypothesis is rejected F – Test F-test of equality of coefficients across sites, with q-values, An F statistic is a value you get when you run an ANOVA test or a regression analysis to find out if the means between two populations are significantly different. Tests for the hypothesis that the results are the same for all countries for each outcome variable. The hypothesis is rejected for almost all pooled outcomes (Table 3), which suggests that there is significant site-by-site variation

14. 14 Interpretation on Results

15. 15

16. 16 Pooled average intent-to- treat effects, endline 2 at a glance. This figure summarizes treatment effects presented in tables Treatment effects on continuous variables are presented in SD units. Each entry shows the OLS estimate and 95% confidence interval for that outcome. ( ♦ ) Statistically significant, 5% level; ( ⋄ ) not statistically significant, 5% level.

17. 17 Cost benefit Analysis The ultimate goal of the program is to durably increase consumption, not merely to increase asset holding. Using total consumption as the measure for benefits, the total benefit-cost ratios presented in row 11 indicate that with the exception of Honduras, the programs all have benefits greater than their costs (ranging from 133% in Ghana to 433% in India).

18. 18 Evaluation Effective implementation of randomized control trail Differences in implementing in across region/ continents Cultural issues Level of living standard (level of PWR can be compared across countries? Standard asset transfer across countries ( Impact of Goat and Hens in Ethiopia may be significantly difference in India Level of training and education is subjective, depend on the quality of enumerator, too many variable to control Within control groups Across countries Large number of outcome variables are reported. Therefore, we expect some of the variables to show significant results due to chance. Differences in sample size across countries, but in analysis it not weighted  but SD mean rectify the issue? More can be learned about how to optimize the design and implementation of the program The costs of the schemes, which varied from $414 per participant in India to $3,122 in Peru. Reason why?

19. 19 Evaluation Which intervention driving the change, too many drivers, would have numerically studied, which variable is successful in overall impact. Will the impact diminish over time – diminishing marginal returns Study covers long term period of time, creates a risk of seasonal patterns or unexpected events ( droughts, financial crisis etc.)

20. 20 Any Questions ?