Impact Evaluation of the DFID Programme to Accelerate Improved Nutrition of the Extreme Poor in Bangladesh

Undernutrition is widespread in Bangladesh. In 2011, according to the Bangladesh Demographic and Health Survey, 41.3 per cent of children under age five were stunted, 36.4 per cent were underweight, 15.6 per cent were wasted, and more than 50 per cent were anaemic. Prevalences were even higher among extremely poor households. Meanwhile, evidence from South Asia shows that high rates of economic growth and reduction in poverty have not led to similarly large reductions in undernutrition (see 1, 2, 3). These findings have suggested that improvements in income alone may not be sufficient to improve nutritional status. Extensive research has also shown that the critical window for nutritional interventions is during the "first thousand days" of life (see 4, 5), from the time when a child is in utero until about two years of age. Based on this accumulated evidence, growing attention has come to introducing nutrition interventions that target children's "first thousand days" alongside household poverty reduction programmes. In particular, there has been growing emphasis on nutrition interventions that aim to improve infant and young child feeding practices-through increasing nutritional knowledge of women who are pregnant, lactating, or likely to be pregnant in the future-as well as to improve the nutritional status of these women themselves.

Although there exists considerable evidence assessing the effectiveness of various livelihoods interventions and other social protection programmes, as well as some evidence on the effectiveness of various direct nutritional interventions, little research directly assesses how an integrated livelihoods and nutrition programme might compare with livelihoods support alone. There are several reasons why the combination of nutrition and livelihoods support may have nutritional benefits over and above livelihoods support only. First, a key constraint to improved nutritional status may be insufficient knowledge of appropriate infant and young child feeding practices (for example, the appropriate duration of exclusive breastfeeding, the appropriate frequency and diversity of child feeding thereafter, etc.). If this is the case, then improving income alone will not necessarily lead to improved feeding practices. Second, there may be synergies between the two types of support. For example, even if a mother's knowledge of infant and young child feeding practices improves, she may still need access to sufficient resources for undertaking those practices (such as income to purchase the recommended types of food), which can be facilitated through a livelihoods intervention. Third, there may be other dynamics shifted through the direct nutrition intervention that mediate how the livelihoods intervention affects nutritional status. For example, if a direct nutrition intervention targeting women improves women's bargaining power within the household, and if women tend to prefer devoting more resources to young children's nutrition (e.g., Quisumbing and Maluccio 2003), the result may also be larger impacts on nutritional status than livelihoods support alone.

The DFID Programme to Accelerate Improved Nutrition for the Extreme Poor in Bangladesh aims to improve nutrition outcomes for young children, pregnant and lactating mothers, and adolescent girls. Its approach is to integrate direct nutrition interventions into the livelihood support currently provided to extremely poor households in Bangladesh through three existing programmes: the Chars Livelihoods Programme (CLP), the Shiree Economic Empowerment of the Poorest Programme (Shiree or EEP, within which the investigators focus on the Concern subproject), and the Urban Partnership for Poverty Reduction Programme (UPPR).

In order to rigorously and independently assess the impacts of these integrated nutrition and livelihoods programmes, DFID has collaborated with research partners and implementation partners to undertake a mixed methods impact evaluation, entitled "Impact Evaluation of the DFID Programme to Accelerate Improved Nutrition for the Extreme Poor in Bangladesh." The evaluation team includes IDS (the lead organisation), IFPRI, ITAD, CNRS, and BRAC University. The evaluation uses mixed quantitative and qualitative methods within a strong theory-based design to assess the impacts of the integrated programmes on nutritional status.

The quantitative impact component involves a baseline survey (conducted in September-November 2013) and an endline survey (to be conducted in November-December 2015). The exploratory/explanatory component includes a qualitative subcomponent (for which the first phase of fieldwork has been ongoing since February 2014), as well as a process evaluation subcomponent (ongoing since July 2014, final results not yet available). The cost effectiveness component began in August 2014 and will be completed in early 2016 following the quantitative endline survey completion.

The three key research questions regarding programme impact that will be addressed are:

• What is the impact on nutrition outcomes of receiving a combination of livelihoods and direct nutrition interventions (denoting this scenario (L+N)), relative to receiving a livelihoods intervention only (denoting this scenario (L))?
• What is the impact on nutrition outcomes of receiving a combination of livelihoods and direct nutrition interventions (L+N), relative to receiving no intervention (denoting this scenario (C) for control)?
• What is the impact on nutrition outcomes of receiving a livelihoods intervention only (L), relative to receiving no intervention (C)? This will pertain only to the urban group served by UPPR.

In order to construct a proxy for the (L+N) households in the counterfactual (L) scenario, randomisation is used. Among the households that already receive the livelihoods intervention at baseline, half are randomly assigned to additionally receive the nutrition intervention after the baseline (denoted the (L+N) group). The remaining half continue to receive only the livelihoods intervention (denoting the (L) group). Randomisation is conducted at the level of primary sampling units (PSUs) that cover an entire locality, rather than at the level of individual households. The randomisation makes it very likely that characteristics of the (L) and (L+N) groups will on average be similar at baseline. (L) is then a valid proxy for (L+N), and average differences between the groups at endline can be interpreted as impacts caused only by the addition of the nutrition component rather than pre-existing differences.

In order to construct a proxy for the (L+N) households in the counterfactual (C) scenario of no intervention, non-randomised approaches are used. Since none of the original livelihoods interventions was rolled out following a randomised control trial design, there is no obvious set of comparable non-beneficiaries to serve as the counterfactual. Because a control group is nonetheless required to assess the absolute benefits of either (L) or (L + N) interventions, attempts were made to construct the best possible control group out of non-randomly selected non-beneficiaries. It is important to emphasise that a non-random control group is not expected to be on average identical to beneficiary households. In the baseline survey, the objective was simply to sample a group of non-beneficiaries as similar as possible to the beneficiaries except for receipt of the intervention.

Major topics areas covered by the qualitative data collection tools include the following:

• Social, economic, institutional and political context of the community
• Local practices, resources, customs in regards to health, hygiene, nutrition and care of children, pregnant and lactation mothers, adolescent girls
• For (L) and (L+N) sites: Perceived impact of the livelihood intervention
• For (L+N) sites: Perceived synergies and disconnects between the nutrition and livelihood interventions in the communities
• For (L+N) sites: Micro-dynamics of the nutrition intervention at the community level and how beneficiaries perceive/experience the intervention

Inclusion criteria

• Children (boys and girls) 0-36 months
• Pregnant and breastfeeding women
• Adolescent girls (aged 13-19)
• Households must meet eligibility criteria of each of the three programmes (CLP, Shiree, UPPR) except for the control group in UPPR.

Exclusion criteria

• None