Escaping Poverty: Comparing Livelihood Approaches for the Ultra-Poor in Ghana

How to cite

Karlan, Dean; Ofori-Atta, Angela; Barker, Nathan; Bryan, Gharad; Udry, Christopher. 2020. "Escaping Poverty: Comparing Livelihood Approaches for the Ultra-Poor in Ghana." AEA RCT Registry. AEARCTR-0005453. https://doi.org/10.1257/rct.5453-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Dean Karlan* , Angela Ofori-Atta , Nathan Barker , Gharad Bryan , Christopher Udry

* Primary investigator

Registry ID
AEARCTR-0005453
DOI
https://doi.org/10.1257/rct.5453-1.0
First registered
Last updated
Status
On going
Trial dates
2016-01-01 – 2021-12-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/5453

Abstract

The Escaping Poverty (EP) project was developed in response to two major challenges facing developing nations: low incomes and poor mental health. The two issues are undoubtedly linked, yet programs to increase earnings or reduce depression often operate in separate siloes. The EP study aims to bridge this gap. EP randomized whether or not households received cognitive behavioral therapy (CBT) prior to receiving a multi-faceted economic program (often referred to as a “graduation” model), to see the extent to which improving an individual’s mental health better enables them to take advantage of an economic opportunity. Moreover, by randomizing which components of the graduation program participants receive, the study seeks to investigate which aspects are most crucial to its success.

Intervention

The EP study has two main interventions: a mental health intervention (CBT) as well as an economic intervention (graduation program).

Individuals randomized to receive CBT participated in a twelve-week program that focused on productive thinking, dealing with negative thought distortions, setting goals and time management, and other related topics. The participants met once a week for approximately 90 minutes at a time in a group of roughly 10 people from the same community, all of the same gender. They were led by one CBT counselor and one assistant. The counselor in each community was someone with a bachelor’s degree or higher, in about half of the cases with a degree in psychology. These counselors received two weeks of classroom training and one week of practice with individuals who agreed to attend a part of the curriculum.

The graduation program combined several different components. First, participants were given a productive asset (for example, four goats) to be used to generate income, and they received training from technical experts on how to manage the asset. Individuals were also offered health insurance and access to a savings account. In the immediate months after receiving the asset, a random subset of households were additionally given regular small cash transfers to minimize risk that they would liquidate the assets to finance immediate consumption needs. Finally, all households receiving the asset were randomized to receive individual coaching visits, group coaching visits, or no coaching for a two year period following asset delivery. Coaching was led by field agents who was role was to provide support, answer questions about how to best manage the assets, and ensure that they continue to work productively.

Lastly, the experimental design included a lump sum cash transfer arm (as well as control group), in order to compare the multi-faceted approach against a one-time cash transfer.

Experimental design

As part of the EP study, participants were randomized at the community level into one of four experimental conditions:

- Pure control
- Treatment - CBT only
- Treatment - GUP model (coaching was at the household level)
- Treatment - Heifer model (coaching was at the group level)

The second stage of randomization assigned households in the three treatment groups to various sub-treatment categories. First, households were randomized to receive or not receive CBT. To mitigate concerns surrounding mixed-gender CBT groups, we randomized all treatment communities to offer either female-only or male-only CBT programs. Subsequently, households with an individual (either the household head or his/her spouse) of the CBT-assigned gender were randomly assigned to receive CBT. Households without a member of the assigned gender were excluded from the CBT treatment e.g., a single mother living in a community assigned to male-only CBT.

After CBT was randomized, households in treatment communities were then assigned to the remaining possible treatment categories. For example, a household assigned to CBT in a GUP community would be randomized to: Pure CBT; Assets, CBT, and CS; or Assets, CBT, and no CS. This second tier of the randomization did not consider gender.

Design details

Randomization method
We performed a computerized randomization.
Randomization unit
Since we used a two-stage randomization design, pure control communities are randomized at the village level and all other groups are randomized at the household level.
Clusters
97 community-level treatment assignments, 5,773 individual-level treatment assignments (from the second stage of randomization)
Observations
7,330 households
Arms
1,557 households - Control, 380 households - CBT only, 3,173 households - GUP, 2,220 households - Heifer
Primary outcomes
Per Capita Consumption, Total Asset Wealth, Total Income, Mental Health Index, Women's Empowerment/Intra-Household Violence
Secondary outcomes
Secondary/Downstream Outcomes: Index of Financial Inclusion, Index of Physical Health, Index of Political Involvement; Mechanism Variables: Total Hours Worked (Last 24 Hours), Index of Grit and Self-Efficacy

Cited by

No citation data available for this registration yet.

Keywords