Matching father to children in the British Household Panel Survey

[Cheng Mun Kwang]

Hello,

I am working on a research in applied economics project. My research theme is intergenerational mobility in the UK, in particular how does a father's income affect his children's income.

Currently, I'm having problems matching the father to his children in the British Household Panel Survey. My understanding is that I have to merge information from the wEGOALT and the wINDRESP files using the variable pid or hid and pno but I'm not too sure how to go about doing this in STATA or if I am on the right track.

I would appreciate it if anyone can shed some light on this matter. Thank you

[Andrew Musau]

help merge

should guide you on how to do this. Having opened the wEGOALT file in Stata, you probably want to

merge m:1 pid pno using wINDRESP

[Stephen Jenkins]

If you wish to link children and parents using BHPS data, then: wINDALL is the file containing all enumerated individuals (including children); wINDRESP a file containing respondent adults; wEGOALT contains information about the relationship between each "ego" and "alter" pair; wHHRESP contains household income variables of various kinds (you need to be careful about which one you want), whereas each adult's earnings, if in work, are on wINDRESP. If you want a "disposable household income" variable rather than "gross household income" variable, you need to register with the UK Data Service and get the "BHPS net income variables" created by me and colleagues. Note that variable "relationship to household reference person" (wHGR2R on wINDALL) may be one way to proceed, though not good enough if some of your fathers are not the household reference person. So, the matching and merging for your project is actually quite complicated and is likely to involve 3 file types for each BHPS wave. I strongly recommend that you consult the on-line BHPS documentation and, even better, take one of the ISER training courses.

[Cheng Mun Kwang]

Thank you for your reply. I am thinking of using the wFIYRL (annual labour income) variable in WINDRESP as my measure of income. What is the difference between using wREL (relationship of alter to ego) in wEGOALT and using wHGR2R in wINDALL as suggested by you?

[Stephen Jenkins]

I mentioned wHGR2R as a possibility; I didn't say it was better, and I mentioned a potential source of difference already. To repeat: I strongly advise that you read the BHPS online documentation -- volumes A and B -- before proceeding further. (I.e. don't rely on me.) This is not straightforward stuff. At the same time, also check out the different measures of earnings and income that are available, and decide which is best for your research question and why. (NB "Labour income" = earnings, which is not the same as "income" as understood by most people, and what you mentioned in your first post.)

[Cheng Mun Kwang]

Thank you for your advice. I am using annual labour income and annual household labour income data from the first 10 waves of the BHPS. This is not really Stata-related but I am wondering if I have to deflate the labour income data.