The southern coast of Bangladesh is highly susceptible to flooding, with climate change exacerbating the frequency and intensity of these events. Households affected by floods often resort to “maladaptive” coping strategies such as reducing expenditures, selling assets, borrowing, and drawing down savings—with possible gendered effects. Globally, social assistance programs such as cash or in-kind transfers have shown promise in supporting resource-poor rural households to better cope with weather shocks associated with climate change. But evidence is limited on how social assistance affects households specifically in the context of flooding in Bangladesh, and even fewer studies differentiate effects by gender.
In a forthcoming brief supported by the CGIAR Initiative on Gender Equality (HER+) and the CGIAR Initiative on Fragility, Conflict, and Migration (FCM), researchers assess how a social assistance program implemented in southern Bangladesh between 2012 and 2014 influenced the effects of flooding. The analysis examines how flooding impacted consumption and diets among resource-poor rural households that received the program, compared to households that did not, with a particular focus on gendered differences.
Program design and survey data
The social assistance program studied—the Transfer Modality Research Initiative (TMRI)—was implemented by the World Food Programme from May 2012 to April 2014 and was designed as a randomized control trial. TMRI provided monthly cash or food transfers, with or without nutrition behavior change communication (BCC), to mothers of young children in resource-poor rural households. While TMRI was implemented in both the northwest and in the coastal south of Bangladesh, this analysis focuses on the latter (referred to as the “South”; Figure 1).
Across the South, fifty villages were randomly assigned to each of the five intervention arms, and 10 eligible households were chosen from each village. Thus, 500 households were assigned to each of the following:
- “Control”: no intervention
- “Cash”: unconditional monthly cash transfers of 1,500 taka (approximately US$19, or 25% of households’ pre-program monthly consumption)
- “Food”: unconditional monthly food rations (rice, lentils, oil)
- “½ Cash, ½ Food”: half the monthly cash transfers and half the monthly food rations
- “Food+BCC”: monthly food rations along with nutrition BCC
The BCC focused on promoting adoption of recommended practices for young children’s nutrition and health, through a combination of group sessions, home visits, and community meetings. Some BCC sessions discussed producing micronutrient-rich foods in order to access healthy diverse diets.
Three rounds of longitudinal survey data were collected on TMRI treatment and control households immediately before and during the intervention: a baseline survey from March–April 2012, a midline survey in June 2013, and an endline survey in April 2014.
Flood measures
Since TMRI survey rounds occurred between March and June in 2012–2014, the study considered flooding in the prior monsoon seasons, in other words the 3rd quarter of 2011–2013. Remote sensing data were used to construct a standardized “flooding extent z-score”, to capture flooding anomalies relative to each union’s own historical distribution.
As expected, due to randomization, the distribution of flooding extent z-scores between 2011–2013 did not meaningfully differ between the TMRI control and treatment groups (Figure 2). The distributions show substantial variation in flooding over these years and locations, while also reflecting these events were not among the most extreme in recent history (for example, z-scores exceeded 3 at other time points).
Estimating how TMRI influenced the effects of flooding on consumption and diets
Key outcomes constructed from the TMRI survey data include: (1) household consumption expenditures and (2) the individual-level Global Diet Quality Score (GDQS), as well as “high-risk GDQS” which measures whether an individual’s score falls in a range associated with high risk of nutrient inadequacy or non-communicable disease.
The analysis estimates impacts of a one-standard-deviation increase in prior flooding among households which received no TMRI treatment, as well as the impacts of prior flooding among households which received TMRI. The core presentation of results shows 95% confidence intervals for the estimated effects of
- “Flood with No Treatment”: for households in the control group, the impact of an increase by 1 in prior flood extent z-score, relative to the mean flood extent z-score in this sample; and
- “Flood with Treatment”: for households that received cash transfers, food transfers, half cash and half food transfers, or food transfers with BCC, the impact of an increase by 1 in prior flood extent z-score, relative to the mean flood extent z-score in this sample.
Findings
The study finds that, among poor households in southern Bangladesh which did not receive TMRI, an increase in flooding extent z-score during the prior monsoon season did not meaningfully affect household consumption several months later but did lead to significantly reduced household savings (Figures 3 and 4). Among households that did receive TMRI, an increase in lagged flooding did not reduce household savings. These results suggest that households in the control group (although poor) kept precautionary savings to cope with possible flooding in this flood-prone setting, but TMRI allowed households to avoid needing to draw down these savings, presumably due to transfer resources.
The study also finds that, among poor households which did not receive TMRI, an increase in prior flooding led to small but significant reductions in diet quality for both men and women aged 15–49 years. Among TMRI treatment households, an increase in flooding led to a small improvement in diet quality among both men and women, corresponding to a 13% decline in the probability of both men and women having high-risk GDQS (3-percentage point reduction; Figure 5).
Further analysis suggests that, for both men and women, without TMRI, prior flooding substantially decreased individual intake of legumes and of fish. With TMRI, the effect of prior flooding was not meaningfully different for fish but led to an increase in legume intake (driven by transfers) as well as for fruits and vegetables (driven by transfers+BCC, possibly related to changes in crop production). Improvements appear stronger for TMRI households that received both transfers and BCC.
Study implications
These patterns suggest that, in the absence of TMRI, poor households attempted to cope with prior flooding and smooth consumption by drawing down precautionary savings and adjusting household members’ diets. Receiving TMRI allowed households to cope without resorting to using savings and without adverse impacts on men’s and women’s diet quality. These findings are consistent with recent qualitative evidence, as well as early quantitative evidence on coping responses to flooding in Bangladesh. With a few exceptions, the analysis shows limited differential impacts by gender, suggesting that households’ coping responses to prior flooding—with or without TMRI—generally did not appear to worsen gender equality in terms of diets.
Findings indicate that, among poor households in southern Bangladesh, social assistance allows smoothing consumption several months after moderate flooding without reducing savings or worsening diets, with benefits for both men and women. Results suggest possibly stronger improvements when transfers are combined with nutrition BCC, implying that this nutrition-sensitive design may simultaneously promote improved child diets, household poverty reduction, women’s empowerment, and adaptation to flooding.
Written by: Shalini Roy, Senior Research Fellow, Poverty, Gender, and Inclusion Unit, IFPRI
Top photo: M. Aminul Islam Khandaker/IFPRI-Bangladesh