Support #459

derived variable for total number of children

Added by Alexey Bessudnov about 8 years ago. Updated about 8 years ago.

Derived variables
Start date:
% Done:




I'd like to create variables for a total number of biological children for individuals in each wave. My understanding is that nnatch gives me the number of biological children currently residing in the household, and lnprnt is for the number of children outside hh in wave 1 and only applies to new entrants in later waves. What I want is a series of variables for the total number of children in each wave (no matter dead or alive, residing in the hh or living separately), i.e. something along these lines:

pidp a_nch b_nch c_nch d_nch e_nch
1 0 0 0 1 1 (a person has a new child between waves 3 and 4)
2 1 1 1 1 1 (a person enters the study with one child and doesn't have any more children)

Next I'd like to do the same for the total number of adoptive children.

Any advice on this would be much appreciated.

Thank you.


Updated by Gundi Knies about 8 years ago

  • Status changed from New to In Progress
  • Assignee changed from Gundi Knies to Alexey Bessudnov
  • Target version set to X M
  • % Done changed from 0 to 80

Hi Alexey,

first, to confirm that there is no derived variable which provides this information. Hence, you are right that you will have to construct your own.

Second, you are right that the information stored in _nnatch, _allch, and _adoptch relate to any children below age 16 currently ennumerated in the household. If you are just interested in children in that age range, using this info will be an option. Note, however, that this information is derived from the household grid and we found that there were many reporting or recording errors with this information. We therefore also provide a corrected version of the relationships (see _relationship_dv on data file _egoalt) and then provide pointers to children's parents (e.g. _mnpid is the pidp of the respondent's biological mother on files such as _indall). You may prefer to use this edited version. Also note that some children change from biological to half-siblings to step-children etc.

Third, you say you are interested in "children dead or alive". For this, in addition to _lnprnt, which is asked off men only, you may want to consider _preg, which asks women about any pregnancies they may have had. These questions always refer back to the last interview and some respondents miss a wave.

We do not ask about any non-resident adoptive children a respondent may have been "a parent" to in the past (or even between waves if they happened to have lived in a different household context for some period of time).

If you are interested in this information for the BHPS sample, you may want to use the fertility history file and read the user guide to that file. Even if you decide to not consider the BHPS sample, the user guide may offer insights into the likely data issues you will face, and how they may be dealt with. The link to that file is as follows:

Hope this helps,


Updated by Alexey Bessudnov about 8 years ago

Hi Gundi,

Many thanks for your detailed and timely reply. Just a couple of things to clarify. You said that lnprnt is asked of men only. Is there any way to find out if a woman had biological children before she entered the study if they don't live in the same household?

Essentially what I want to do is a) to limit the sample to people who didn't have children at the point of entry to the study, 2) to look at the fertility history for the duration of the study for this subset. What I'm going to do then is to look at a combination of nnatch (or the corrected version) and lnprnt (and their analogues for adopted children) at wave 1 for everyone and in waves 2 to 5 for new entrants only and filter out people with values > 0 on any of these variables. For the remaining subset I use the combination of nnatch and lnprnt to code the total number of biological children in and outside the hh in waves 2 to 5. Does this sound about right?

Thank you,



Updated by Gundi Knies about 8 years ago

Hi Alexey,
apologies, it turns out the _lnprnt is asked of men and women who appear not to have any natural children living in the current household (based on _nnatch). I copy directly from the W2 questionnaire:

"Have you ever fathered/had any biological children?"
if ff_ivlolw = 2|3|MIS & ff_everint <> 1 //new entrant never interviewed, excluding rising 16 year olds
and if GRIDVARIABLES.NNatCh = 0 //number of biological children in household is 0

That seems to be a pretty straightforward indicator to select your sample! You may want to do some sensitivity analyses and check whether everybody who should have been asked the question was asked (and vice versa) given _nnatch may have been incorrect (using _relationship_dv).

I don't think you can do the same thing for adoptive children but I may be wrong. i can only advise to check the questionnaires carefully for anything that may work in your specific context. What works in one study may not work in another and often times it is a good idea to use different indicators! From the top of my head, the information on the number of living relatives (see the W3/W5 family networks module) may offer some additional insights for you.



Updated by Alexey Bessudnov about 8 years ago

Thank you. This answers my questions.


Updated by Gundi Knies about 8 years ago

  • Status changed from In Progress to Resolved
  • % Done changed from 80 to 100



Updated by Redmine Admin about 8 years ago

  • Status changed from Resolved to Closed

Also available in: Atom PDF