Improving food and nutrition security by enhancing women’s empowerment in Bangladesh

The intervention will consist of two parts, (1) a leaflet intervention and a (2) women's empowerment training.

Part 1
The leaflet intervention consists of a leaflet on diarrhoea prevention behaviours and an application exercise that allows participants to work with, internalise, and digest the information provided on the leaflet. The application exercise will probably be shaped as a group discussion. The exercise will not explicitly include aspects related to women’s empowerment. Women who have not followed the women's empowerment training should not learn lessons related to empowerment during this exercise. However, at the same time, the exercise should allow for women who have followed the training modules to engage with the lessons related to empowerment that they have learnt in the training.

Part 2
The women's empowerment training specifically focuses on empowerment in household decision making. There will be no mention of child diarrhoea. The training is designed around active learning strategies. It is based on the assumption that necessary knowledge and insights reside within the local community, and thus emphasises learning from oneself and one’s peers. At its core lies the psychological principle of WISE interventions, which outlines that behavioural change results from altering maladaptive perceptions. Targeting these perceptions, rather than behaviours itself, can create sustainable and lasting behavioural change.

The training consists of four modules, of which three include only women, and the last one includes both women and their husbands. Each module comprises three theory-based components aiming to change psychological perceptions: introducing new perceptions, interdependence, and commitment. The first module focuses on perceptions related to roles, the second module on self-efficacy, the third module on leveraging social networks, and the fourth module on husband and wife teamwork.

2024-03-15

2024-07-15

1. Psychological well-being
2. Personal Empowerment
3. Household decision-making
4. Household food insecurity access scale (HFIAS)
5. Dietary Diversity
6. Health
7. Waste disposal
8. Handwashing

Full details of the variables can be found in the attached pre-analysis plan.
Psychological well-being is a composite index that measures the overall psychological satisfaction and connection in the respondent and her husband’s life. We collect data on these measures from both women and men It is comprosed of the following sub-indicies:
a. Meaning-making
b. Social Connectedness
c. Social Support
d. Teamwork
e. Controlling behavior
The final index is then calculated as (a) + (b) + (c) + (d) – (e)
• psychology_woman = meaning_making_woman + Connect_woman + support_woman + team_woman - control_woman
• psychology_man = meaning_making_man + connect_man + support_man + team_man - control_man

Personal empowerment is the measure of agency and self-efficacy reported by the respondent and her husband. It is the sum of two indicies:
a. Agency index
b. Self-Efficacy
This index will be created for both women and their husbands separately.
• empowerment_woman = agency_woman + self_efficacy_woman
• empowerment_man = agency_man + self-efficacy_man

Household decision-making measures the household decision making initiative the wife has. The variable uses information from both woman and man survey to generate decision making for woman. It captures whether the woman is able to initiate discussions on important household decisions with their partner. For each decision where the woman has initiatiated or participated in a discussion with their partner they get a score of 1 and for those that they haven’t they get a score of 0. Additionally for the same statements if the man also reports that the wife bought up the particular topic for discussion the statement gets additional score of 1. If the woman reports discussion but the partner does not then no additional score is added. Lastly, if the woman reports not bringing up a topic for discussion and the partner disagrees then no additional score is added as well.

Household food insecurity access scale (HFIAS): The HFIAS module yields information on food insecurity (access) at the household level. The variable is calculated for each household by assigning a code for the food insecurity (access) category in which it falls. The categories are coded as 0 = Food Secure, 1=Mildly Food Insecure Access, 2=Moderately Food Insecure Access, 3=Severely Food Insecure Access. The computation of this variable will follow the guidelines developed by USAID.

Dietary Diversity: Dietary diversity scores are calculated by summing the number of food groups consumed by the individual respondent and the reference child (under 5 years of age) over the 24-hour recall period. Consumption of food based on major food groups as proposed by FAO will be created.

Variable health captures the number of household members that were sick in the past two weeks preceeding the survey. It is computed as two separate variables to differentiate between sickness in general family member and child health.
a. number of members in the household who were sick in the last two weeks.(sick_mem = D1R1)
b. number of children under the age of 5 who were sick in the last two weeks.(sick_child = D1R2)
Further, health is also measured through anthopometric measures of children (AN1 to AN11.2). Using the height and weight of children we will compute the z-scores for height for age and weight for age, as well ahd BMI of both child and mother. In doing so we will follow the guidelines published by WHO.

Waste disposal observation is an observational measure of the availability of proper waste disposal facilities for waste from latrine, child diapers, and garbage in a household.

Handwashing behavior self-report is a self-reported measure of handwashing behavior. Specifically, we define this as the standardized mean of the following six indicators for handwashing practices of the respondent in last 24 hours:
a. Handwash after defecation: frequency
b. Handwash after defecation: quality
c. Handwash before preparing food: frequency
d. Handwash before preparing food: quality:
e. Handwash before eating: frequency:
f. Handwash before eating: quality:

We randomize 152 VOs in our sample such that both the treatment and the control groups have 76 VOs each. We further randomize the 76 VOs into two groups of 38 VOs each such that one group recieves the leaflet intervention and the other does not. Our study will therefore have the following four groups:

Group 1: Training only group - All female Brac members from this group will get the psychology training.

Group 2: Training + Leaflet group - All female Brac members in this group will get the psychology training and the leaflet intervention.

Group 3: Leaflet only group: All femal Brac members in this group will only get the leaflet intervention.

Gropu 4: Control group: All female Brac members in this group do not receive any training or leaflet intervetnion.

Not available

Randomization done in office using STATA

Cluster randomization at village organization level

Yes

152 clusters

Total observations are 1034 men and 1034 women (so 1034 households)

We randomize 152 VOs in our sample such that both the treatment and the control groups have 76 VOs each. We further randomize the 76 VOs into two groups of 38 VOs each such that one group recieves the leaflet intervention and the other does not.

Since we have a full factorial design with 2 interventions (i.e. training and leaflet), we assume that we have 1034/2=517 individuals for each intervention and similarly, J is 152/2= 76 groups which implies we have on average (n) 7 women per group. Further, we assume the following regarding rest of the parameters: Significance level: 0.05; power: 80% and P =0.5. The most critical variable is \rho. Therefore, we do calculations for a range of values for \rho We assume: 0.01<\rho<0.20. \rho\rho0.01<\rho<0.20 Simple calculations for MDES than suggest that the MDES of the training we will be able to detect varies between 0.23 and 0.34 for the one sided test and 0.26 and 0.39 for the two sided test. Second, we use the above formula for analysis that combines training only and training+ leaflet groups together as one training group and leaflet only and control group as one control group. In this case we assume that we have 1034 individuals for each intervention and similarly, J is 152 groups which implies we have on average (n) 7 women per group. Further, we assume the following regarding rest of the parameters: Significance level: 0.05; power: 80% and P =0.5. Again, we do calculations for a range of values for \rho We assume: 0.01<\rho<0.20. Simple calculations for MDES than suggest that the MDES of the training we will be able to detect varies between 0.16 and 0.24 for the one sided test and 0.18 and 0.27 for the two sided test.