The problems with conducting a gender analysis at the household level

My main task this summer is to conduct a gender analysis of the food security situation in Rwanda. The primary data I have is from a 2015 household survey focusing on the food security and vulnerability conditions throughout different regions of Rwanda. As the survey was conducted at the household level, there is limited limited information about individual members of the household. Thus, the main way to analyze the data from a gender perspective is to compare female-headed to male-headed households, which is how these analyses are often done.

This comparison between female-headed households (FHH) and male-headed households (MHH) had at first seemed straight forward, but the more I look at the data, the more I find this level of analysis to be misleading in terms of gender comparisons. I've found five main problems to this type of analysis:

1. The FHH and MHH categories are, by definition, not equivalent.

When conducting the CFSVA household survey, enumerators categorize households as female-headed only in the case that there is no male who is part of the household and capable of making decisions. This means that the average FHH is likely to have older, widowed heads and fewer members because of the lack of a spouse. This in turn increases the likelihood of having a disabled member (because of the older age of the household head) and a higher percentage of dependents (simply because of not having a spouse). These factors all increase the likelihood of food insecurity while only being partially related to gender.

On the other hand, a woman will be considered part of an MHH if she is married to someone with the mental facilities to make decisions, regardless of how much money she earns compared to her husband or how much she participates in decision-making.

The categorization inherently makes FHH more vulnerable than MHH for multiple reasons besides gender.

FHH, as I found, are still much more vulnerable than MHH in terms of food insecurity, and I am therefore convinced that an analysis at the household level is useful for understanding and targeting the vulnerable subset of FHH. However, to understand this as a "gender analysis" misses some of the key reasons that FHH are more vulnerable than MHH and may incorrectly attribute the discrepancy to gender differences.

This also makes a comparison between FHH and MHH less meaningful than a gender comparison at the individual level because the households tend to differ in many areas. A household with a young married heterosexual couple and their kids would automatically be labeled as an MHH, whereas the FHH that it would be compared to would almost certainly have an older widow.

2. The CFSVA categorization of FHH hides the households led by females while males are not present.

The CFSVA labels households where husbands are not present for long periods of time (due to detention, migration, work, etc.) as MHH. Despite not having a man physically in the house, the enumerators consider these households to be male headed because there is a male who could make some decisions remotely. This limits our ability to analyze a subsection of households where women are in charge of most decisions and often finances. It also distorts the two categories by labeling these households as MHH, which could in fact make MHH on average seem poorer or more vulnerable than if these families had not been included (if it turns out that these households are in fact poorer).

Other studies tackle this problem by disaggregating MHH, de jure FHH, and de facto FHH. They recognize that households with a man who is not present but might still send remittances or contribute to decision-making are neither the same as de jure FHH nor MHH and separate them.

One study from the Rwandan government also conducted in 2015 used this category and found that 6.4% of households were temporarily headed by females whose husbands had been away for a long time. This shows that this is a significant enough number of households that their categorization could in fact affect overall averages.

3. An analysis at the household level distorts gender distinctions in agricultural practices, coping mechanisms, and food preparation.

Using the monolithic unit of the household suggests the conditions of everyone within the household are the same. Given that women perform different roles than men in terms of food security, we cannot simply assume their food security and vulnerability conditions would be the same within the household. For instance, in some cultures, women serve meals to their families first and eat last. Food shortages might then impact them more than other household members, if there is less left for them to eat. If we look at the household level, we miss this difference.

Additionally, as women tend to be responsible for coping for food shortages in both MHH and FHH, a comparison of the two types of households may not show the same differences as a comparison of households where women are present and where women are not present. It would also not indicate which household members WFP and other agencies should target to make a difference in food or nutrition practices because it does not allow us to specify who plays what role.

4. Finally, categorizing homes as having 1 head of household reinforces patriarchal ideas of the family.

When enumerators label households as headed by either a man or a woman, they support the notion that only one person can be the primary decision-maker in a home. Moreover, by assigning the man to be the household head in all but exceptional cases, they perpetuate ideas of male dominance when households might not otherwise think of themselves in those terms.

It is rarely the case that only one person makes decisions at home. In terms of food preparation and nutrition, it is in fact often the woman making decisions, even if her husband might make larger decisions on buying and selling land, etc. The household head notion erases these nuances, making it more difficult to conduct an analysis and fortifying ideas of hierarchy within the family.

For the time being, I will continue to conduct the analysis at the household level because this is the primary data available and because, as mentioned, FHH do in fact tend to be some of the most vulnerable, even if only somewhat because of gender differences. However, I want to make sure to highlight these problems because they should at least inform the way we interpret my results. Ideally, they could also change the way data is collected in the future.