Enhancing adaptive capacities of coastal communities, especially women, to cope with climate change induced salinity

The project is implemented by UNDP Bangladesh in the districts of Khulna and Satkhira. This impact study focusses on two interventions of the project: the "adaptive livelihoods" component (LH), and the "drinking water" component (DW).

The DW component consists mainly of constructing rainwater harvesting systems (RWHS), ensuring water supply during the dry season. The use is exclusively directed at drinking water, not at water resources for agriculture, cooking or personal hygiene. The RWHS can be installed in appropriate sizes at the household level, community level or institutional level. The plan is to install 13,308 RWHS at the household level, for which a small co-financing is required (this could also be contributed in-kind by helping to build the RWHS); 228 RWHS at the community level, each covering approximately 25–50 households (e.g. at mosques, temples or other community buildings); 19 RWHS at the institutional level, each covering approximately 75–100 households (e.g. at schools or other government institutions); and 41 pond embankments and filtration systems. At least 20 per cent of target households in each ward should receive a household-level RWHS. Water user groups and water management committees will be formed to ensure sustainable planning and maintenance of the water solutions.

The LH component targets exclusively women. They will be organized into Women livelihood groups (WLGs) of approximately 25 women each, for a total of 1,017 WLGs. Each WLG will jointly select three out of eight adaptive livelihood options, according to their preferences, for which they will be trained as a group. After completion of the training, they will be asked to select two out of the three LHs they were trained on, for which they will receive the necessary input. The eight LH options were selected with the goal of being appropriate for women’s engagement and empowerment as well as suitable for local market conditions. There are three production cycles: for the first cycle, in-kind support for the necessary inputs will be received; for the second cycle, a cash transfer from the Government of Bangladesh (around BDT 20,000 [USD 235] per beneficiary) will be received; for the third cycle, loans can be taken up by a microfinance institution to slowly phase in financial independence.

2022-02-01

2022-11-01

Household Food Consumption Score
Household Food Insecurity Access Scale score
Household total income
Household total expenditure
Household food expenditure
Household income diversification
Household income stability
Household savings
Household preparedness for shocks (self-reported)

Household Food Consumption Score: following the methodology developed by the World Food Program
Household Food Insecurity Access Scale score: following the methodology developed by the INDDEX Project
Household income stability: proxied by a cross-sectional estimate of income volatility following the three-step Feasible Generalized Least Squares procedure proposed in Chaudhuri, Jalan and Suryahadi (2002). "Assessing household vulnerability to poverty from cross-sectional data: a methodology and estimates from Indonesia". Department of Economics, Columbia University. Discussion paper No. 0102-52
Household income diversification: proxied by the inverse Hirschman–Herfindahl Index calculated using income shares from various sources (following Idowu, Awoyemi, Omonona, and Falusi, 2011. Non-farm income diversification and poverty among rural farm households in southwest Nigeria. European Journal of Social Sciences, vol. 21. pp. 163-177). Income sources are categorized following the RIGA Project.

Time allocation of women
Time allocated by women to fetching water
Women's decision-making power in the household
Women’s participation in social life
Adoption of adaptive livelihoods
Household profits from adaptive livelihoods
Income generated by women in the household

The study design consists of a clustered phase-in RCT.
For the purpose of the trial, the "adaptive livelihood" component of the project is rolled out in two phases, i.e. one group of beneficiaries started project activities in February 2022 - the "early group - while the other group will start project activities in October 2022 - the "late" group.
The project covers 101 wards in 39 Union Parishads (UPs) across 5 upazilas (sub-districts). Randomization at the household (or individual) level and even ward level was not feasible. Therefore, UPs were chosen as the randomization units, i.e. UPs are the clusters in our trial.
Out of the 39 project UPs, 25 were randomly allocated to the "early" group; and 14 to the "late" group. Randomization was stratified by upazila, allowing to control for upazila-level heterogeneity, including potential implementer bias (UNDP partners with three NGOs for project implementation, and each NGO catchment area is defined based on upazilas).
Note that baseline sampling was done amongst those households identified as eligible to both the "adaptive livelihood" and the "drinking water" components of the project. Baseline data were collected in both groups in October 2021, and endline data will be collected from the same households in October 2022, before the second phase of implementation starts.
It is important to note that staggered implementation is only done for the "adaptive livelihood" component - the "drinking water" component will start in all project areas at the same time, during the "early" phase.
The proposed design will allow us to estimate the causal impacts of the "adaptive livelihood" component of the project, by comparing endline outcomes of the "early" group to those of the "late" group - the latter effectively serving as the control group in our study. Baseline data are used to check balance after randomization, and to control (potentially) for baseline values in regression analysis.
As households in the study sample also benefit from the "drinking water" component, we will be able to explore whether the key mechanism intended by the project is indeed at play (i.e. that "drinking water" solutions allow women to re-allocate time from fetching water towards income-generating adaptive livelihoods) - this part of the analysis may not yield causal claims, as the "drinking water" component was not controlled.

Randomization was done in office, by a computer, using Stata.

The unit of randomization in our study is the Union Parishad (UP).

Yes

39 Union Parishads (UPs)

Endline target = 2,652 households Baseline target = 3,120 households, assuming up to 15% attrition

25 clusters in the "early" group - Target = 1,700 households at endline (2,000 at baseline)
14 clusters in the "late" group (effectively the control group in our analysis) - 952 households at endline (1,120 at baseline)

Main outcome = Household yearly income Baseline mean outcome = 90,530 Bangladeshi Taka (BDT) Baseline standard deviation = 55,233 UP-level Intra-Cluster Correlation = 0.054 Endline mean outcome (calculated) = 103,991 BDT Standardized Minimum Detectable Effect = +14.9% Note that, although we present results in the case of an increase in household yearly income, our study is powered to also capture a decrease of the same magnitude.