Impact evaluation of a community nutrition and livelihood program on child nutrition in rural Bangladesh

Abstract Given the high prevalence of child undernutrition in Bangladesh, multi‐sectoral approaches involving livelihood promotion have potential to mitigate the burden of undernutrition. This study examined the impact of an economic development (ED) program providing poultry assets, gardening skills and saving training added to the Positive Deviant (PD)/Hearth program (PDH/ED), compared to PD/Hearth only (PDH). A total of 1029 children who attended PD/Hearth sessions in September–November 2018 at 6–13 months of age were enrolled in the cohort study in July–August 2019. The cohort, comprised of 532 children in the PDH/ED group and 593 children in the PDH group, was reassessed in November 2020. The program impact on child nutrition, food security, crop production, dietary quality and household income was estimated using a difference‐in‐differences approach accounting for the sociodemographic differences between PDH/ED and PDH groups. Compared to the PDH group, the PDH/ED group showed increases in child dietary diversity score (DDS) (+0.32), child minimum dietary diversity (13.7 percentage points [pp]), and maternal DDS (+0.28) (all p < 0.05). From 2019 to 2020, the PDH/ED households improved food security by 12.6 pp and diversified crop production (bananas (9.7 pp), papaya (11.1 pp), carrots (3.8 pp) and lemons (5.9 pp)), and increased the proportion of annual income ≥60,000 Taka by 12.4 pp and last month income ≥5000 Taka by 7.8 pp, compared to PDH group (all p < 0.05). However, there was no impact on child nutritional status, morbidity, livestock ownership and total annual/last income. Incorporating an ED program into nutrition programming could benefit food security and dietary diversity in rural Bangladesh.


| INTRODUCTION
Globally, an estimated 149 million children under 5 years old are stunted, while an estimated 47.0 million are wasted, as of 2020 (United Nations Children's Fund World Health Organization & World Bank Group, 2021). South Asia alone is home to about one-third of stunted children (more than 54.3 million) and more than half (25.0 million) of all wasted children.
Various nutrition-specific interventions, such as supplementation with micronutrients and energy-balanced foods and behavior change efforts to promote optimal complementary feeding, are effective strategies for addressing child stunting (Bhutta et al., 2013). However, implementing a combination of effective nutrition-specific interventions at a 90% scale in 34 low-and middle-income countries would address only 20% of the burden of child stunting (Bhutta et al., 2013). Nutrition-sensitive approaches to enhance food security, child care, and hygienic environmental and health services have shown the potential to mitigate the burden of child stunting (Ruel et al., 2018). Economic status can play a key role in improving nutritional quality in micronutrient intake and daily energy consumption by increasing purchasing power, particularly benefiting those most in poverty (Ramos et al., 2021). Previous literature analyzing the prevalence of child stunting and gross domestic production (GDP) per person in the country reported that a 10% increase in GDP per person predicts an almost 6% reduction in stunting among the ultrapoor population. (McGovern et al., 2017;Ruel & Alderman, 2013). In addition, nutrition-sensitive programs have led to positive evidence on food security and dietary diversity. A study examining household-level food security in Kenya, Uganda, and Tanzania found that wider crop variety was associated with food security (Silvestri et al., 2015). Other studies implementing asset transfer in Ethiopia (Irenso & Atomsa, 2018) and Mozambique (Johnson et al., 2015) showed enhanced household food consumption and dietary diversity.
Multifaceted interventions including nutrition-sensitive approaches, however, are delivered through multiple channels and comprehensive modalities. Despite programs having apparent and outcome-oriented logic models, the innate complexity of such programming makes it challenging to understand which activities contributed to the expected outcomes.

World Vision Bangladesh (WVB) and World Vision Korea codesigned and implemented the Bangladesh Rajshahi Division
Maternal and Child Nutrition (BRDMCN) project in 2018-2020 . The BRDMCN attempted to improve child nutrition through social behavior change communication (SBCC) activities and economic development (ED) program, transferring assets for income generation. The ED was based on an ultrapoor graduation approach, which is a multifaceted livelihood program that has demonstrated its multidimensional effects on improving food security, food consumption, and asset growth in six randomized controlled trials, including in Bangladesh (Banerjee et al., 2015). However, the impact of a graduation approach on child nutrition is less reported. A recent study in Burkina Faso found a 22% reduction in food insecurity after 2 years of implementing a graduation approach nutrition program with cash transfer and productive asset transfer (Bouguen & Dillon, 2021).
The present study attempted to explore the potential of the ED program with a graduation approach to strengthen the impact of the SBCC-oriented Positive Deviant (PD)/Hearth program, on child nutrition. Specifically, the study examined if the ED program added to the PD/Hearth program enhanced the impact on reducing child underweight and morbidity, improving child and maternal dietary diversity, ensuring household food security, increasing crop production activities, and improving income generation, compared to exposure to the PD/Hearth program only.

| Study settings
The BRDMCN program was implemented in three upazilas (the second-lowest tier of regional administration), Joypurhat, Panchbibi, and Dhamoirhat in the Rajshahi division from March 2018-December 2020 . This study area features a typically poor and agricultural community, has been hit by natural disasters and presents higher child undernutrition than Dhaka division (World Vision Bangladesh, 2017). Based on the BRDMCN, an inclusive impact and process evaluation study has been conducted to understand the implementation process of major intervention components. Due to the spread of coronavirus disease 2019 (COVID-19), the Bangladesh government has reinforced lockdowns several times since March 2020 (Bangladesh, 2020). The BRDMCN downscaled the planned activities and modified the intervention modality in 2020 to reduce interpersonal contacts and avoid large-group gatherings, while adaptive procedures, such as meetings with fewer participants, were also implemented .

Key points/highlights
• Incorporating asset provision and saving training into nutrition programming would help households safeguard the dietary needs of women and young children.
• Future nutrition-sensitive approaches aimed at improving nutritional status should identify the point along the impact pathway where the program impact is likely to be weakened and incorporate appropriate complementary measures accordingly.
• Program strengthening is needed so that mothers can use food resources and household income to increase the consumption of animal protein and micronutrient-rich foods for child growth.

| Intervention
The details of BRDMCN have been described elsewhere . The intervention delivery and monitoring activities of the BRDMCN were conducted by 285 community facilitators, 29 community supervisors, and nine program officers. Community facilitators were recruited from the project area and trained to lead and facilitate the PD/Hearth and ED program, including screening ultrapoor households, facilitating Hearth sessions, conducting follow-up visits and child weight monitoring, and managing monthly group savings group meetings . Each community facilitator was in charge of providing guides on nutrition care and asset management to 300 households.
Community supervisors oversaw the activities of community facilitators and helped manage monthly group savings meetings.

| PD/Hearth program
The SBCC component in the BRDMCN included PD/Hearth, a monthly child growth monitoring promotion activity, nutrition education at the community level, and community water, sanitation, and hygiene campaigns. As a major intervention component of BRDMCN, the PD/Hearth program aimed to rehabilitate underweight children and promote appropriate nutrition and hygiene behaviors among their caregivers (World Vision, 2014). The PD/Hearth programs showed positive impacts on child nutritional status in other program settings (Kang et al., 2017;Roche et al., 2017;Young et al., 2021).
Community facilitators screened children aged 6-59 months with weight-for-age z-score (WAZ) less than − 1.0 in the community.
The screened mothers and children with WAZ score < −1 attended a 2-week, community facilitator-led Hearth session. Through these Hearth sessions, the community facilitators transferred knowledge of healthy behaviors based on Positive Deviant Inquiry findings, which included: breastfeeding, complementary feeding, hygiene practices, caring, health-seeking, and family planning. The participating mothers prepared and fed their children complimentary meals using local and low-cost ingredients that are dense in micronutrients (i.e., vegetables and eggs). Following the Hearth sessions, community facilitators conducted 2-3 home visits over 2 weeks to support mothers to continue practicing what they learned from Hearth sessions at home. Community facilitators weighed the children who attended Hearth sessions at enrollment, 12 and 30 days, and 3 and 6 months after the Hearth session. A total of 11,984 children enrolled in PD/Hearth sessions in BRDMCN. A nested research study based on weight monitoring data of approximately 6500 PD/Hearth children indicated that PD/Hearth participation held a greater benefit for children enrolled at a younger age (Kim, Biswas, et al., 2021).

| ED program
Graduation approaches target ultrapoor who live on <$1.9 per day, and are multidimensionally vulnerable (e.g., health, nutrition, and food security) (World Vision Bangladesh & BRAC, 2018). Graduation programs varied by context but shared some components: (1) productive asset transfer; (2) consumption support; (3) technical skills to manage assets; (4) home visits, (5) saving practices, and (6) health education and services (Abdul Latif Jameel Poverty Action Lab J-PAL, 2015). The ED program in the BRDMCN adopted the components of graduation programs such as asset transfer for income generation, asset management skills and saving practices. The ED program used to work with poor and ultrapoor beneficiaries, those who do not have adequate land or capital for large-scale cultivation ( Table S1). The objectives of the ED program were to enhance the production of homestead vegetables, fruit tree plantation and small-scale livestock (duck and chicken) rearing such that the intended beneficiaries had a greater capacity to cultivate sources of nutrition for their family and to consume their products as well.
A secondary aim was to provide beneficiaries with supplemental income through the selling of surplus vegetables, eggs, chicken, and ducklings after satisfying their family requirements. They were also encouraged to purchase any necessary nutritious foods that they were not producing on their own farms, utilizing surplus income .
The ED program provided a total of 2960 ultrapoor participants with assets based on their living environment or previous microbusiness experience for the purpose of income generation or food sourcing. A total of 2089 households each received 11-17 ducks; 870 received 11 chickens; 2075 received vegetable seeds and gardening training; and 1511 received 1-4 nursery fruit trees in September-December 2018 (Volpenhein et al., 2022). Quarterly saving training was given to the participants, in addition to technical support from the Department of Agriculture Extension and Livestock.
Participants also attended monthly group meetings led by community facilitators.
Regular household visits were made by community facilitators to see whether the participants needed additional asset management support or guidance. Using an excel-based monitoring tool developed by the implementation team, community facilitators recorded the number/amount of assets received, assets lost, production, purchases, sales, consumption, and derived income every 3 months . Using these monitoring records, a nested study found ED beneficiaries sustained asset holding and income generation over 18 months in 2019-2020 (Volpenhein et al., 2022). difference in the prevalence of underweight, with 5% type 1 error and 80% power, between the PDH children and those who participated in the PD/Hearth alongside the ED program at the end of the project period.

| Outcomes
The primary outcomes of interest were the difference in WAZ and the prevalence of underweight (WAZ < −2) among children in the PDH/ED group compared with the PDH group following 2 years of program implementation (2018-2020). Secondary outcomes were the difference between the PDH/ED group and the PDH group in the proportion of child and maternal dietary diversity score (DDS)/ minimum dietary diversity (MDD), household food security, household crop production/livestock rearing activities, annual and monthly income (Bangladesh Taka) and child morbidity from 2019 to 2020.

| Data collection and field procedure
The research team initially developed paper-based questionnaires, which were translated into Bangla and transferred to an online F I G U R E 1 † Unmet: The subject moved out of the area or was not at home at the time of the survey visit. ‡ Refused: The subject decided not to participate at the time of the survey. survey tool. The questions were reviewed by researchers and WVB's nutrition and agriculture experts for content and appropriateness. The project team hired a survey firm in Dhaka to conduct the surveys and the research team trained the survey firm interviewers and partnered with the implementation team to supervise data collection in the field. A KOBO tool (www. kobotoolbox.org) was used to administer a tablet-based survey.
Interviewers were recruited by the survey firm in 2019 and recruited by World Vision in 2020. Researchers conducted a 3day training, and pilot tests were administered before the household survey. In 2019, the interviewers collected sociodemographic data, anthropometrics, diet outcomes, food security, crop production, livestock possession and child morbidity data. Child height and mid-upper arm circumference were not measured in 2020 due to the reinforced social distancing.
DDS and MDD were assessed using a standard methodology to calculate Infant and young child feeding indicators (World Health Organization, 2010). Children's DDS ranged from 0 to 7 using the following food groups: (1) grains, roots and tubers; (2) legumes and nuts; (3) dairy products (i.e., milk, yogurt, and cheese); (4) flesh foods (i.e., meat, fish, poultry and liver/organ meats); (5) eggs; (6) vitamin A-rich fruits and vegetables; and (7) other fruits and vegetables. MDD was defined as the proportion of children who consumed four or more foods out of seven food groups in the past 24 h. Maternal DDS (range: 0-10) was based on 10 food groups: (1) starchy staples; (2) beans and peas; (3) nuts and seeds; (4) dairy; (5) flesh foods (meat, fish); (6) eggs; (7) vitamin A rich dark green vegetables; (8)  A nine-item Household Food Insecurity Access Scale (HFIAS) module was used to assess household food insecurity (anxiety, quantity and quality) (Food and Nutrition Technical Assistance Project, 2007). Responses for each food insecurity item were followed with frequency questions (rarely, sometimes, and often).
Following the standard categorization of HFIAS, the responses were categorized as food secure versus food insecure (mildly, moderately, and severely).
Mothers were asked if the household produced any type of crops in the past year and if so, were asked about the amount of crop produced for the following items: rice, banana, guava, mango, papaya, carrot, dark green leafy vegetables (DGLV) and lemon. Rice and beans are the type of crops commonly harvested in the study area. The items of mango, papaya, lemon, and DGLV were based on the assets provided through the ED program. We assumed that the households who received assets would expand their farming/gardening activities into other crop items. Mothers were asked if they owned any livestock and how many cows, goats, lambs, chickens and/or ducks the household currently owned. These livestock species were commonly found in the project area.
Mothers in the study were asked about total income in the past year and in the past month. These income variables respectively were categorized into binary variables using the following cut-offs: ≥60,000 Taka ($720) versus <60,000 Taka and ≥5000 Taka ($60) versus <5000 Taka. The cut-off of 5000 Taka for monthly income and 60,000 Taka

| Statistical analysis
An exploratory data analysis calculated mean (SD) and percentages for all variables, including outcomes. Comparability of sociodemographic characteristics was tested between PDH group and PDH/ED group with data collected in 2019 using χ 2 tests for categorical variables and student t-tests for continuous variables.
A difference-in-differences (DID) analysis was used to estimate the program impact for 1 year from 2018 to 2020 on primary outcomes (child underweight and WAZ) and from 2019 to The impact estimation in the DID regressions was adjusted for child age, gender, maternal age, father's occupation and education, and household building materials that differed in PDH and PDH/ED groups in 2019. All DID models accounted for clustering within individuals using the "cluster" option to obtain robust standard errors. All data analyses were done using STATA version 17 SE (StataCorp, 2021). A p-value of 0.05 was considered statistically significant.

| Intervention implementation
Regarding PD/Hearth participation, there was no difference in the days of session attendance between the PDH and PDH/ED groups (11.5 days vs. 11.5 days) but the number of follow-up visits by community facilitators was slightly higher in the PDH/ ED versus PDH (10.5 times vs. 9.2 times) (

| Sociodemographic characteristics in 2019
In both groups, the majority of fathers were between 25 and 34

| Program impact
The program impact on WAZ (0.026 z-score; p = 0.56) and ED group in 2020 with significant increases in producing banana (9.7 pp; p < 0.001), papaya (11.7 pp; p < 0.001), carrot (3.8 pp; p < 0.001) and lemon (5.9 pp; p = 0.02), compared to the PDH group. There was no significant difference in the amount of crops produced between two groups (Table S4) (Tables 5 and S5).
Total income in the past year (p = 0.62) and in the past month income (p = 0.64) also did not show significant improvement in the PDH/ED group, compared to the PDH group (Table 6). However, after stratification of income, the result showed a significant increase in the proportion of households with total income in the past year ≥60,000 Taka by 12.4 pp (0.003) and the past month income of ≥5000 Taka by 7.8 pp (p = 0.008) in the PDH/ED group, compared to the PDH group. The PDH/ED program showed no significant differences in the impact on child morbidity, compared to the PDH group (Table S6).

| Nutritional status
The ED program improved food security, dietary diversity, crop production and household income earning above the poverty line, but had no impact on child underweight. This result implies that the impact of ED program operated along the anticipated impact pathway  but was insufficient in influencing nutritional status. These findings serve as a reminder of the challenges of nutrition-sensitive programs, which need careful midcourse corrections in the coordination of multifaceted activities to improve child nutritional status, rather than addressing the immediate causes of undernutrition (UNICEF, 2021) The purpose of typical graduation models did not consider addressing undernutrition but rather focused to raise the socioeconomic status of the ultrapoor (Banerjee et al., 2015). For this reason, few studies included nutritional status in their program outcomes. A graduation program in Bangladesh called targeting the ultrapoor (TUP) demonstrated the program impact on wasting and weight-for-height z-score in a randomized controlled trial, and the effect was also positive in poor nonparticipants (Raza et al., 2018). The TUP program had a 4-year follow-up that occurred later in childhood, but the ED program had a relatively short 2-year follow-up for underweight outcomes with an unprecedented COVID-19 pandemic.
A pre-post evaluation of income-generating activities targeting extremely poor families in Bangladesh showed a reduction of the prevalence of stunting by 13%, but no reduction in the prevalence of wasting and underweight (Goto et al., 2019). Even though food security and net income improved along the impact pathway, the absence of control or comparison groups in the study design constrains the interpretation of program impact (Goto et al., 2019).
A qualitative study, which was nested within the parental evaluation based on the BRDMCN, was conducted in 2019. When asked about the question of ED program purpose, "improving child nutrition" was less frequently mentioned among ED mothers, compared to replies indicating a desire to raise socioeconomic status . The fact that child nutrition is not prioritized as highly in asset utilization suggests that the impact pathway of the ED program would not be dense enough to result in better nutritional status.
The cohort children who were mostly underweight at the time of PD/Hearth enrollment experienced a dramatic return in nutritional status between 2019 and 2020. This unexpected T A B L E 5 Program impact on food security, crop production and livestock activities* Although there was an overall improvement in nutritional status from 2018 to 2019, instability in weight changes strengthens the notion of the sensitive nature of child nutrition status to environmental influence. This situation reflected the negative impact of the COVID-19 pandemic on child nutrition due to downscaled program activities in 2020 . This finding mirrors previous studies that reported children of poor families are disproportionately affected by COVID-19 (McConnell et al., 2021). Although our study findings imply that assets provided by the ED program could provide a buffer against economic hardship during COVID-19, we are unable to rule out the possibility that the pandemic could lessen the impact of the ED program, which would have otherwise improved nutritional status.

| Diet
An increase in the consumption of more diverse foods was more apparent in the PDH/ED group than in the PDH group. Legumes  (Tables S2 and S3). This result suggests that the PDH/ED group was afforded the ability to buy legumes using additional income. The observed lack of impact on egg consumption among children in 2020 compared to 2019 might be due to an already increased consumption of eggs among the PDH/ED group before the assessment in 2019, as poultry assets were distributed in 2018. Our study found that the ED program TUP played an essential role in helping households safeguard their dietary needs. However, the consumption of animal protein and micronutrient-rich foods, which are essential for child growth, did not significantly increase. Also, the present study did not assess the amount and frequency of foods the study children truly consumed.
Taken together, the enhanced dietary diversity did not address undernutrition in the study children. | 11 of 15 empowerment and self-esteem in making food consumption-related decisions for their households . In addition, saving practices in 2020 were higher in the PDH/ED (70.8%) than in the PDH group (53.3%) (p < 0.05; data not shown).

| Food security and income generation
The proportion of households that earned annual income (≥60,000 Taka) and monthly income (≥5000 Taka) above the extreme poverty line increased in the PDH/ED group compared to the PDH group. This finding implies that the ED program could give greater financial benefits to the extremely poor than the PDH program alone.
Additional evidence supporting the results is that annual income from selling ducks and eggs during 2019 was USD 29 (IQR: 12-56) among 1200 duckling beneficiaries (Volpenhein et al., 2022).
The limited improvement in total income might be due to soaring food prices and income reduction in 2020 which further restricted the population's purchasing power (IPC-IG & ROSA, 2020). Studies reported an increase in food insecurity at almost twice the rate of the previous year in Bangladesh (Egger et al., 2021) and a high rate of losing or experiencing income reduction in the country, with 35% of the workforce depending on a daily wage Mottaleb et al., 2020). This financial risk might also push ED participants to consume foods available through household assets rather than expanding businesses to earn income. Interestingly enough, a graduation approach showed a potential to protect the poor during economic shock: the program participants who passed the poverty threshold secured their job (Rahman & Bandiera, 2021).

| Crop production and livestock management
The diversified profile of crop production among the PDH/ED group is associated with a higher DDS and a higher percentage of households with income above the poverty line than in the PDH group. It was likely that the PDH/ED group introduced other types of crops using gardening skills and income depending on their interest. Such an adaptation process was observed gradually with the choice of poultry rearing. In another study that analyzed the monitoring data of all 2960 ED beneficiaries, we found that beneficiaries who did not receive duck as assets began to adopt chicken over time (0%-18%), while those who were not provided with chicken adopted chicken rearing gradually (0%-46.6%) (Volpenhein et al., 2022). Meanwhile, these fluid adoptions remind us that we are not able to completely rule out the potential of a spillover effect from the ED program on the non-ED group concerning poultry rearing or crop production.
No lack of observed difference between groups in the number of poultry currently owned can be explained by a few reasons. The rearing scope of ducks varied seasonally and could be reduced in early winter due to scarcity of pond water and feeding resources and depending on the market price of duck (Khan et al., 2018).
Additionally, a potential duck plague that also seasonally occurs in winter may lead to the precautionary selling of ducks before their livestock is hurt. Another study in the BRDMCN that monitored 2095 ED beneficiaries during 2019 indicated that the scale of duck assets during April-June was the peak, while the highest number of death and sales occurred during the prewinter season; the lowest asset number was reported from October through December (Volpenhein et al., 2022). The COVID-19 pandemic also might compel study participants to sell their poultry to make cash urgently, resulting in a reduction of possession of poultry at the time of assessment.

| Implication for future programs and research
Our research reveals the high potential of the ED program to improve child nutrition along the impact pathway of nutrition-sensitive agricultural programming (SPRING, 2014 Further research should be planned to answer these questions. The current study did not quantify women's empowerment or social capital at the community level. These factors should be measured in future studies as one the key drivers of program impact. The sustained impact of SBCC approaches including PD/Hearth is likely to be weakened under the COVID-19 lockdown. However, programs ensuring food security such as the ED have the potential to complement the nutritional impact of SBCC activities. Many low-income countries, such as Malawi, adopted food distribution and transfer payments to mitigate the direct impact early in the pandemic and to ensure that people had adequate food access (Jiang et al., 2021).
However, social assistances are rarely of a substantial amount and generally provides only short-term solutions, insufficient in averting the long-withstanding effect brought about by the pandemic. An evaluation of the food assistance program provided by the government and private sectors in the Philippines showed that the food packages lacked nutritional value and could only address short-term hunger while potentially increasing the risk of acute and chronic diseases among those already in poverty (Ong et al., 2020). Even the social assistance distributed in high-income countries, such as the United States, was found to be insufficient in alleviating financial hardships and food insecurity from already low-income families (Kim, 2021). We need to consider how ED program participants can continue to strengthen their household's food security while keeping livestock as an asset even during a crisis like the COVID-19 pandemic. Selling surplus livestock and eggs may be a sensible short-term solution as the effect of the pandemic persists. However, program managers and community facilitators should also encourage other forms of income-generating activities for a successful program impact.

| Strength and limitations
Our large sample size, the presence of a comparison group, and the parallel approach in the evaluation process of an ongoing program contributed to the strength of our study. However, our study has its limitations.
First, except for weight measurement at PD/Hearth sessions, other key data began to be collected in 2019, 1 year after program implementation, when the cohort was generated. The present study, therefore, did not have a "true baseline" as our first measured outcomes were most likely to have shown improvements since program implementation. This limitation influenced our ability to compare to the endline.
Second, height and other anthropometric measurements that require physical contact were not collected due to social distancing restrictions due to COVID-19. Weight-for-age measures were collected according to PD/Hearth guidelines; hence, we could not measure the long-term program impact on stunting by 2020. Weight is more sensitive to environmental influence.
Third, the study participants were not randomized into the intervention groups and were not comparable in socioeconomic status. The households in the PDH/ED group had poorer socioeconomic status as the asset transfer targeted the poor or ultrapoor.
This limitation was adjusted for in estimating the program impact by adjusting for socioeconomic and demographic factors. However, it cannot totally exclude the possibility that the results may have been affected by socioeconomic status. Fourth, there is a probability of recall bias in the information collected on the amount of crop produced and monthly or annual income.

| CONCLUSION
Our study found that adding an ED program to the PD/Hearth program benefits food security, dietary diversity, crop diversification and income among the poor with children at risk of undernutrition, compared to the PDH program only. Our study provides supporting evidence that asset transfer programs based on a graduation approach may also potentially offer some protection against economic shock but need to be strengthened to make an impact on nutrition outcomes.