Home » Topics » Wealth Index » Construction of wealth quintiles
| Re: Construction of wealth quintiles [message #17967 is a reply to message #17966] |
Thu, 01 August 2019 13:00   |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
The wealth quintiles can be constructed from the PR file, using hv271 (the continuous wealth index), hv005 (the household weight), and hv102 (de jure residence). The goal is to have an equal number of weighted de-jure individual in the PR file in each quintile. Note that the residence criterion is de jure rather than de facto. For most purposes, DHS uses the de facto criterion. For this variable we use de jure, partly to be consistent (for this index) with MICS, which uses de jure for most purposes.
In Stata, this would mean opening the PR file and running this command:
xtile hv270_test=hv271 [pweight=hv005] if hv102==1, n(5)
If you do it the way I just described, you will split households on the boundaries of the quintiles, and that must be avoided. Everyone in the same household should be in the same quintile.
Instead, it is best to work from the HR file and then transfer the values to the PR and other files.
Open the HR file and then run these lines:
gen wt=hv005*hv012
xtile test=hv271 [pweight=wt], n(5)
I just tested this with the latest Angola HR file and I see that four households are misclassified with these lines (out of more than 16,000 households). This may have to do with how ties are handled in Stata, versus how they are handled with SPSS, which I believe was used to calculate the quintiles for this survey. I would not consider this a problem.
To calculate deciles you would just replace n(5) with n(10) in the xtile command. Then you could merge with other files OR you could identify the numerical values of the five additional cut points and just apply them to the other files.
|
|
|
|
Current Time: Tue Nov 26 14:55:42 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
