How we do it


We used a combination of secondary and primary data analysis, using mixed methods. We focused on one case study country, Tanzania. Five research methods were used.

1. Review of existing household surveys (post-1960) and censuses (post-1950) in sub-Saharan Africa involving textual analysis of description of differences in standard household definitions by country, and survey type, including change over time. Review included household survey documents (e.g.: enumerators’ manuals, methodology reports). It covered all censuses and national surveys for which documentation could be found either on the internet or in the libraries of SOAS and LSE. Censuses for: Botswana, Ethiopia, Gambia, Ghana, Kenya, Malawi, Nigeria, South Africa Tanzania, Zambia and Zimbabwe. Surveys examined for a subset of countries: Gambia, Ghana, Kenya, Malawi, Tanzania and Zambia and included: Demographic and Health Survey (DHS), World Health Survey (WHS), World Fertility Survey (WFS), Multiple Indicator Cluster Survey (MICS), Living Standards Measurement Survey (LSMS), , Integrated household surveys, Core Welfare Indicator Questionnaires (CWIQ), and Household Budget Surveys. Not all surveys were undertaken in all countries.

2. In-depth semi-structured interviews (n=41 with 54 individuals) with household survey producers and consumers, national and international, to identify how the concept of the household is produced, understood and used at different stages of data production and use. Interviews focused on:

  • What they understood by the term ‘household’
  • Understandings of the concept of household in survey data collection
  • Awareness of which types of individuals are ex/included in survey ‘households’ and why
  • Perceptions of the divergences between survey ‘households’ and their own experience of the primary social units in Tanzania where applicable
  • Their reflections on particular populations who might be poorly served / represented in household surveys

Recorded interviews were transcribed verbatim and analysed for both content and language used. Content was analysed to ascertain respondents’ knowledge and understanding of the conventions dictating household definitions in that country and whether there are steps in the chain of data production and analysis where unfounded assumptions are made about how ‘household’ has been used.

3. In-depth qualitative study of local concepts of the household for selected national sub-populations, identified on the basis of interviews with data producers and consumers. Sub-populations included:

  • Urban communities with high proportion of temporary migrants (two suburbs of Dar es Salaam, n=24)
  • pastoral and polygamous groups (Maasai ethnic group, n=8, but 218 people because these locally defined households were very large)
  • a rural area with high levels of temporary migration (Rufiji area, n=20).

4. Scenario modelling of the impact of the different household concepts to estimate the size and direction of impacts of different household concepts on a range of socio-demographic outcomes.

5. Literature mapping of articles (n=2,393) drawn from across the social sciences, published between 2003 and 2009. A systematic mapping of the literature in major social science databases based on keywords and phrases. This established the scale of awareness of the nuances of household in the published literature.

Ethical clearance was obtained from the LSE’s Research Committee. No ethical issues arose during the course of the work.