Counting same sex couples in France from a new survey on families

Elisabeth Morand, Institut National d'Études Démographiques (INED)
Nicolas Razafindratsima, Institut National d'Études Démographiques (INED)
Laurent Toulemon, Institut National d'Études Démographiques (INED)

Enumerating same-sex couples is a challenge for official statistics, as well as studying the demographics of these couples is a challenge for research. Despite these couples gained in many European countries the right to an official recognition through a specific legal status or the right to marriage, their identification in censuses and surveys is very difficult. Even if the description is restricted to coresident couples enumerated in the same household, census data based on cross-tabulation of individual answers to questions on couple situation and sex do not suffice, because a very limited number of errors in the sex of one or the other member of different-sex couples leads to a strong relative proportion of “false same-sex couples”. In order to avoid such errors, it is necessary to include specific categories in the “couple” question and/or to explicitly ask a question about the sex of the partner. This was done in France for the first time in a one-percent survey, the Family and dwellings survey, which took place within the 2011 census, allowing a precise enumeration of same-sex couples. As the sampling unit was the enumerator zone, the sample suffers from large cluster effects. Taking these effects into account is necessary to estimate the variance of the counts. Furthermore, the sample can be used as a training sample for identifying same-sex couples among households included in the census, based on the information present in the census. Using data mining methods, it is then possible to build rules from the sample and to use them to infer counts of same-sex couples in the whole population. We will compare the estimates based on the survey itself, the 2011 yearly census survey, and complete census data based on five yearly waves. Note: the paper could also fit in the “Data and method” Theme.