For my thesis, I'm working with a panel dataset that tracks individuals within households. To allow for merging, the dataset provides the household ID in the previous round for each individual. What I'm wondering is how to decide based on this information that a household is "the same" in two different rounds of data collection.

As I'm looking at the socio-economic outcomes of various farming practices, my concern is that there are unobservable household characteristics that influence the farmer's returns to farming but then also correlate with various household outcomes, such as education, assets, etc. To control for this, I'd like to use household fixed effects.

However, I'm wondering which features of a household would justify me saying "this is the same household and these effects are fixed over time". For instance, I don't want to use the household head because if the head changes and every other aspect of the household is identical, the impacts on farming outcomes shouldn't be too large. My question hence is whether there are commonly accepted characteristics for identifying households over time?

For reference, the data I'm using is the Living Standards Measurement Survey - Integrated Surveys on Agriculture, by the World Bank.


1 Answer 1


I don't get it, you say you have a household ID. If it is the same across waves, then it is the same household over waves. Most likely the people who collected the data will be able to answer the question of whether this is the same household better than you.

The exact definition of who constitutes a household is necessarily arbitrary and varies to some degree between countries. However, the large surveys tend to gravitate around a definition that includes everybody who has lived in the same "building" (whatever that may be) and at least one meal together for more than 6 months in the last year. Then they tend to narrow this down e.g. excluding workers being paid while living there etc. You can check out a questionnaire from a large survey, for instance Living Standard Measurement Survey by the World Bank, at the beginning of the household roster they usually have a section describing who is part of the household.

Households will be given a unique id the first time they are interviewed and followed using this id over the subsequent waves (there is no kind of probability-based matching going on). To re-identify the same households, surveys will use address / GPS, name, and contact details.

In sum: There is really not much point worrying about this. Just group based on hhid and spend your time on other problems in the data (which likely there will be enough of ;))

  • $\begingroup$ Thanks for your answer. The issue is that I want to track "the same" household over multiple waves of the survey - but each household is assigned a new identifier for each year. For each individual, I know which household ID they had in previous waves. Obviously, household composition changes over time, so that doesn't help me decide which of the individuals belongs to the "old" household, and which has split off. $\endgroup$
    – Alison
    Feb 19, 2021 at 18:08
  • $\begingroup$ Ah, now I get it. And can you follow individuals over time, do you have unique person id? $\endgroup$
    – Papayapap
    Feb 20, 2021 at 13:52
  • $\begingroup$ I haven't used LSMS, but if it is a panel survey then there should be a merge identifier and it should usually be laid out in the data documentation. No way the survey collectors would distribute the data without a merger, unless it is different households in each wave. $\endgroup$
    – Papayapap
    Feb 20, 2021 at 13:56
  • $\begingroup$ The merge identifier is the individual ID! Hence the difficulty in deciding what constitutes a household. $\endgroup$
    – Alison
    Feb 20, 2021 at 18:11
  • $\begingroup$ Okay, in case you have Info on household Head (?) then you can use the previous wave hhid from this person and all people with the same current hhid? $\endgroup$
    – Papayapap
    Feb 20, 2021 at 19:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.