I'm looking for open data on household social networks at the micro level in Africa. Preferably I'd like to get piece together a dataset that has networks of households within villages dispersed all across Africa. Am I a dreamer to even think such a dataset is possible to piece together right now? Any village network datasets in Africa?

Data on online African social networks with some sort of geographical or ethnic identification would also work.

  • What type of analysis do you plan to carry out with this data? Cross-section/village/component comparisons? Dyadic regressions? Answering these might help find suitable data.
    – sheß
    Aug 13, 2015 at 14:32
  • Essentially, I want to compare average degree and similar network summary measures across different regions or ethnic groups. I think online data could suffice, but physical data would be preferable.
    – LJB
    Aug 13, 2015 at 19:08
  • Maybe, but that really depends heavily on the context you're working on, you could use household- or family size as a proxy for average degree or something.
    – sheß
    Aug 13, 2015 at 19:25

1 Answer 1


This doesn't exist as of now. There are very limited datasets on social networks at the micro-level. Especially if you're interested in full village census, you are restricted to very few data sets which you will only get by asking the authors of the respective papers themselves. We are currenly working on a networks data set covering over fifty villages in the Gambia, but this won't be available anytime soon and is also not comparable to data collected in other countries.

Most often used in existing research is the network data by De Weerd and Dercon (http://edi-europe.com/docs/JDE_deweerdtdercon.pdf) which is from a single village in TZ.

Data that has comparable networks survey data on the micro level from several different country does not exists.

Maybe if you describe somewhat more detailed what exactly you need, someone may suggest other options. Non-network survey data could proxy for some aspects of network characteristics. Or worldwide online/trade/misc data could be used, even though this would also not give you household-level networks.

For example -but that really depends heavily on the context- you could use household- or family size as a proxy for network characteristics, such as average degree. This would be available from various sources.

Maybe your best shot is to look at large scale surveys and figure out which questions of those are suitable in your context. I think the WVS doesn’t have a lot in terms of social ties. But from the DHS roster you could extract how many non-household members stayed overnight in the last night, which could already be a decent proxy for some form of social network, provided that you find variation in that. Maybe you could also look at how many people without blood-ties are currently staying in each household. A bid odd, but potentially useful is the number of people one had sexual intercourse with, which is also asked in some DHS questionnaires, but I guess that’s too far out of your theoretical framework.

Your Answer

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

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