Matching children and young people to parents no longer in the same household
I am using BHPS together with UKHLS and I am trying to match children and young people to their parents in the dataset, even if they are no longer in the household.
Using BHPS, this seems straightforward as fnspid and mnspid are constant throughout the dataset, so you can merge by creating a separate mother/father file and then merging pid to mnspid/fnspid from indresp.
However, in UKHLS, merging on fnspid and mnspid does not provide as many matches. Are these pidps? Is there a better way of matching which does not use household indicators?
I had a look at the previous query using w_indall in UKHLS but this seems to lead to duplicate cases in mnspid which cannot then be merged?
Updated by Gundi Knies over 7 years ago
- Category set to Data analysis
- Status changed from New to Feedback
- Assignee set to Megan Scott
- Target version set to X M
- % Done changed from 0 to 90
the variables mnpid and fnpid refer to the natural mother/father’s PIDP in the UKHLS. That PIDP may change over time as a result of new information coming through in a particular wave, for example, as the parent realises we make a distinction between natural and adoptive parents and they are not in fact a natural parent. Another route is that the parent’s gender may be incorrectly recorded in the study data base in a particular wave (resulting in mother and father switch) or there may be same-sex parents but only one can be recorded in the gender-specific pointer. Another route for inconsistencies across studies is that the information is asked differently: in the BHPS, household relationships are inferred from the relationship to the head of household; in the UKHLS relationships of each person are asked directly.
There will not be many cases with cross-wave inconsistencies. But (like all data errors/inconsistencies) these should be investigated; if you have a better idea about why the inconsistencies occur you can decide for yourself how to deal with these cases in your specific analysis.
Note that the UKHLS provides both the PID and the PIDP for respondents who already participated in BHPS W1-18. However, BHPS sample members who did not already appear in BHPS W1-18 will have their PID in the UKHLS set to -8 "inapplicable" (like all members of the new UKHLS samples). Therefore, for those pairs who occur in the BHPS W1-18 you could rely solely on the PID but for all others you have to use PIDP (or generate a new unique ID that works across studies). Once you have created an ID that works across studies and cleaned the parent-pointers, linkage is straightforward (as you describe).
Hope this helps, best wishes,