Estimating Women Multidimensional Poverty from Different Datasets

[Kehinde Atoloye]

I want to estimate women multidimensional poverty using MICS dataset. The problem I have is that, the variables I need are not in same dataset. I have WASH household variables in the household dataset, I have education variables in the household member dataset and then the family type variable in the women dataset. How do I go about combining these dataset with women as the unit of analysis. Thanks.

[Andrew Musau]

I am not familiar with these datasets but there should be an individual identifier in the household level dataset or a household identifier in the individual level datasets. You normally would use these identifiers possibly in combination with a year or wave variable to combine the datasets. Look into the codebooks accompanying these datasets for this information.

[Maarten Buis]

In addition to Andrew Musau s comments: the command in Stata that you would use for that is merge.

[George Ford]

use householddata, clear
joinby hhid using familydata , unmatched(master) _merge(_merge_family)
tab _merge_family  // tells you what didn't merge

I'd check the familydata to make sure there's aren't many family members. If so, you need to figure out how to limit familydata to the person you're interested in from householdata.

[Kehinde Atoloye]

Thank you all. I was able to do the merging. I have another issue which has been an argument between myself and my colleagues. The topic of what I am working on is Household Typology and Women Poverty. Like I said earlier, I am measuring the women's poverty by capturing their situations through the household data on education and standard of living. I merged the household member file with the women's file on a 1:1 basis and then in turn merged it with the household file on a m:1 basis. The poverty indicators I used which is based on

Sabina Alkire, Usha Kanagaratnam and Nicolai Suppa (2011). MPI Methodological Note 05 (OPHI Briefing 05)

approach. Now, I concluded that the unit of analysis is the women but my superior argues it is the household. I used variables from household to compute the poverty indicators and used the characteristics of the household and household head as independent variables against the MPI indicator in a random-effect logistic model. Should the women be the unit of analysis or the household. From my own understanding, I believe the women is the unit of analysis but my superior argues it is the household simply because the characteristics of the women are household characteristics capturing their situation. Merging was m:1, i.e., many women to one household. Please I need your insights on this. Thank you.