A good mix against ultra‐poverty? Evidence from a Randomized Controlled Trial (RCT) in Bangladesh

• Published date: 01 November 2021
• Author: Atiya Rahman,Anindita Bhattacharjee,Narayan Das
• Date: 01 November 2021
• DOI: http://doi.org/10.1111/rode.12809

2052

|

wileyonlinelibrary.com/journal/rode Rev Dev Econ. 2021;25:2052–2083.

© 2021 John Wiley & Sons Ltd

1

|

INTRODUCTION

In conversations on the heterogeneity of poverty, it is essential to keep in mind that the poorest are not

like “[t]he poor but a little bit poorer. They may benefit from policies to help the poor, but need other

policies as well” (Sen et al., 2004). Microcredit and transfer (as grants) programs are among the most

notable initiatives for poverty reduction. However, the literature on microcredit documents mixed evi-

dence on the effects on income and consumption (Banerjee, Esther, Glennerster, et al., 2015; Banerjee,

Karlan, et al., 2015; Pitt & Khandker,1998). Furthermore, microcredit programs face challenges in

reaching the poorest (or the ultra- poor). Both demand- and supply- side constraints are responsible for

Received: 18 August 2020

|

Revised: 18 June 2021

|

Accepted: 21 June 2021

DOI: 10.1111/rode.12809

REGULAR ARTICLE

A good mix against ultra- poverty? Evidence from a

Randomized Controlled Trial (RCT) in Bangladesh

AtiyaRahman1

|

AninditaBhattacharjee1,2*

|

NarayanDas1

*BRAC Institute of Governance and Development (BIGD) is a former affiliation for Anindita Bhattacharjee.

1Economics Cluster, BRAC Institute of

Governance and Development (BIGD),

BRAC University, Dhaka, Bangladesh

2Evidence & Learning, Programme

Development and Quality, Save the

Children International, Dhaka, Bangladesh

Correspondence

Atiya Rahman, BRAC Institute of

Governance and Development (BIGD),

BRAC University, Dhaka 1212,

Bangladesh.

Email: atiya.rahman@bracu.ac.bd

Funding information

UK’s Foreign Commonwealth &

Development Office (FCDO), Department

of Foreign Affairs and Trade (DFAT) of

the Australian Government, and BRAC—

the donors of the Ultra- Poor Graduation

Program— through the FCDO- DFAT-

BRAC Strategic Partnership Agreement

(SPA)

Abstract

Existing evidence shows that programs that provide grants

to productive assets along with training to very poor women

increase labor supply, earnings, and consumption. In con-

trast, evidence on the effect of microcredit on these out-

comes is mixed. In this paper, we examine the effect of a

hybrid of the two approaches— credit and grant— on the

livelihoods of the ultra- poor in Bangladesh. A randomized

evaluation of the hybrid intervention shows that it increases

labor supply of working- age women, household income,

productive assets, savings, and consumption expenditures.

The benefit– cost ratio of the intervention is estimated to be

8.47.

KEYWORDS

hybrid, livelihood, randomized controlled trial, ultra- poor

JEL CLASSIFICATION

C9; I3

|

2053

RAHMAN et Al.

these challenges (Hashemi,1997; Hulme & Mosley,1996; Morduch,1998; Rahman & Razzaque,2000).

Even if the ultra- poor par ticipate in microfinance, they are often unable to benefit from it (Ahmed

etal.,2009; Morduch,1998). The literature on transfer programs, on the contrary, shows that they are

highly effective in reducing ultra- poverty (Bandiera et al.,2017; Banerjee, Esther, Goldberg, et al.,

2015; Banerjee et al., 2018; Blattman etal.,2016). Typically, transfer programs provide the ultra- poor

with productive assets, training, consumption allowances, health support, and follow- up advice and

supervision. The transfer program (hereafter, grants- only intervention or model) studied by Banerjee,

Esther, Goldberg et al. (2015), Banerjee et al. (2018), and Bandiera etal.(2017) was originally designed

and implemented by BRAC, the largest nongovernmental organization (NGO) in the world.

BRAC’s grants- only intervention is designed for the ultra- poor who are demographically vulnerable.

A large proportion of households targeted by this intervention are female headed (52%) and have no phys-

ically active male member (41%).1 As discussed in Section2, because of using a set of strict targeting

criteria, a segment of ultra- poor households is not covered by the grants- only intervention. Therefore,

a policy question is how the ultra- poor households that are not covered by the grants- only intervention

could be sustainably lifted out of poverty. One solution could be expanding the grants- only model to

include this segment. However, it may be that these ultra- poor households do not need the costly grants-

only intervention as they are better endowed, especially in terms of demographic resources and the pri-

mary source of livelihood, compared to those who are generally targeted by the grants- only intervention.

In this paper, we study a program that combines microcredit and grants (hereafter referred to as

hybrid intervention or model) to address ultra- poverty in Bangladesh. Both the grants- only and hybrid

models are implemented under the umbrella of the Ultra- Poor Graduation (UPG) program of BRAC.

The graduation program, developed in 2002, implemented just the grants- only intervention until 2006.

The hybrid intervention was introduced in 2007. It includes enterprise development training, soft

loans, necessary input supports, consumption allowance, and health support. Both the grants- only

and hybrid interventions of the UPG program target the ultra- poor. But the target group of the latter is

comparatively well- off in terms of socioeconomic characteristics. The aim of implementing different

approaches under the graduation program is to address heterogeneity (in terms of livelihood opportu-

nities, demographic characteristics, etc.) among the ultra- poor.

We evaluate the 2016 cohort of the hybrid intervention using a randomized controlled trial (RCT). In

2016, BRAC selected 11 districts to implement both its grants- only and hybrid models. Out of the eight

branch offices selected by BRAC from each of these districts, we randomly selected two branch offices

and assigned them to control.2 The remaining six branch offices from each district served as treatment

areas. We conducted a baseline survey in April– July 2016, covering 8,973 eligible ultra- poor house-

holds for the hybrid intervention (1,931 from control areas and 7,042 from treatment areas). Following

the baseline survey, BRAC offered the intervention to eligible ultra- poor households from treatment

areas. Those in the control areas did not receive any support from BRAC. The intervention continued

until December 2017. During the follow- up survey conducted in December 2018, we intended to cover

all eligible households from control areas and two- thirds of the eligible households for the hybrid in-

tervention (i.e., ~4,700 households) due to fund constraints. The survey, however, could successfully

visit 1,739 households from the control group and 4,212 households from the treatment group. Baseline

data reveal that the randomization is balanced across observable characteristics. Among eligible house-

holds for the hybrid intervention, 64% earned per- capita income of less than USD 1.90 per day at the

2016 purchasing power parity (PPP) exchange rate, indicating that a substantial proportion of the tar-

geted households were income- extreme- poor at baseline.3 Due to imperfect compliance, we estimate

the intention- to- treat (ITT) effect and the local average treatment effect (LATE).

Our results show that the hybrid intervention increases the labor supply of working- age women.

The increase in time devoted to livestock and poultry rearing is the main driver of this increased labor

2054

|

RAHMAN et Al.

supply. Further results show that working- age men also increase their time devoted to livestock and

poultry rearing. Results (LATE) further indicate that the intervention increases per- capita household

income by 19% and the value of productive assets by 135%. The intervention has positive effects on

outstanding loans (i.e., the amount of loan yet to be paid back) and savings almost 1year after the

completion of the program cycle. The estimated effect on consumption expenditure is positive and

statistically significant, suggesting that the program increases the welfare of ultra- poor households.

The benefit– cost ratio (BCR) of the hybrid model is estimated to be 8.47. As mentioned earlier, the

ultra- poor households selected by BRAC for receiving the hybrid intervention are relatively well- off

compared to those selected for its grants- only intervention.4 Therefore, our results imply that the

hybrid approach is effective for the relatively well- off ultra- poor. Whether this model can be effec-

tive for all types of ultra- poor households, however, requires further investigation. Nonetheless, our

results suggest that the model can be scaled up to favorably assist a significant segment of the ultra-

poor instead of trying to support all of them through the more expensive grants- only intervention.

This study extends the existing set of literature on microcredit and transfer programs for poverty re-

duction (Bandiera etal.,2017; Banerjee, Esther, Glennerster, et al., 2015, Banerjee, Esther, Goldberg,

et al., 2015, Banerjee, Karlan, et al., 2015; Blattman etal., 2016; Pitt & Khandker,1998). Most of

these studies evaluate microfinance and transfer programs separately. To the best of our knowledge,

ours is the first study to analyze the effect of a hybrid of the two approaches on the ultra- poor using

RCT.5 Microfinance generally fails to have an impact on the ultra- poor due to supply- and demand-

side constraints (Hashemi, 1997), while the grants- only intervention of the UPG program targets

a segment of the ultra- poor (Bandiera etal., 2017).6 Therefore, it is a policy question whether the

extreme- poor who are not reached by microfinance or the grants- only intervention— a program that

requires a large amount of investment— could be supported by a hybrid model to lift them out of pov-

erty. The findings of our paper are thus useful for development organizations, including NGOs, that

have been replicating BRAC’s UPG program.

2

|

OVERVIEW OF BRAC’S UPG PROGRAM

In 2002, BRAC started a pilot program titled “Challenging the Frontiers of Poverty Reduction:

Targeting the Ultra- Poor” for addressing ultra- poverty in Bangladesh, later renamed as the Ultra-

Poor Graduation program.7 The program selects ultra- poor households through a rigorous, multistage

process that includes community wealth ranking and household survey. Initially, BRAC selects sub-

districts based on poverty mapping by the World Food Programme (WFP). Then, it identifies com-

munities with a high concentration of poverty in the chosen subdistricts, based on program staff's

knowledge and discussion with field staff from other BRAC programs.8

In the selected villages, BRAC staff conduct a participatory wealth- ranking exercise, where the

households of a community are ranked into several wealth groups (e.g., very poor, poor, middle class,

and nonpoor). Then, the staff conduct a survey on households from the bottom two or three wealth

ranks to check against the predetermined eligibility criteria (discussed later). Based on the survey and

further verification by senior- level staff, a decisive list of selected ultra- poor households is prepared.

Until 2006, the UPG program offered a grants- only model in which the ultra- poor received assets

(mostly livestock and poultry) as grants along with other support (consumption allowance, training on

income- generating activity, etc.). In 2007, it started implementing two different models— the previous

grants- only model and a newly initiated hybrid model that combined credit with grant— to address

heterogeneity among the ultra- poor. The participants of the grants- only model are poorer than those

of the hybrid model. This study focuses on the hybrid model.