Cohabitation Relationships Derived from Register Data

Carel Harmsen, Statistics Netherlands
Henrico Witvliet, Statistics Netherlands

Many couples start living together or decide to break up without an official marriage or divorce. A database containing information derived from registers on start and end dates of cohabitation relationships was not readily available yet. However, building such a database appeared to be possible by using longitudinal register data on marriages and births, partnerships for income taxes and social security benefits, moving to an address together and on where people live, in combination with an imputation strategy applied to ‘ambiguous couples’, being two persons living together at one address but of whom we did not have any other officially registered information about their relationship. The imputation strategy is also used to reduce the bias that would arise in the database when only using the longitudinal register data. For couples that recently started living together at the same address, less register information about their common future is available. As a result, less of them could be directly designated as a cohabiting couple. Having created this file, we can study relationships over time at the individual level. Furthermore, since the file can be linked to other registers, an abundance of additional research opportunities into the phenomenon of cohabitation has become available.

Many couples start living together or decide to break up without an official marriage or divorce. A database containing information derived from registers on start and end dates of cohabitation relationships was not readily available yet. Now, we have built such a database using longitudinal register data on marriages and births, data on partnerships for income taxes and social security benefits, data on moving to an address together and on where people live. We will discuss the methodology behind the construction of this file. How do we deal with ‘ambiguous couples’, i.e., two persons living together at one address but of whom we do not have any other officially registered information about their relationship? And how do we reduce the bias that is introduced into the database by using longitudinal register data?

In the Netherlands, marriages and divorces are registered, but starting and ending an intimate cohabitation relationship is not, while the majority of new cohabiting couples are unmarried. Hence, by using data about marriages and divorces we underestimated the amount of starting and ending cohabitation relationships. There is also a household database available, but the definition of a couple in that database is not completely applicable to the situation of intimate cohabitation. Improved data about cohabitation relationships was desirable and we have searched for a better way of reaching this. Not only to improve official statistics about cohabitation but also to create a sound basis to study this phenomenon over time at the individual level. Because we have the possibility of linking this file to other registers, many studies are waiting to be conducted.

Our aim was to determine as best we could whether two persons at an address are an intimate cohabiting couple and if so, what the start and (if applicable) end date of that relationship are, based on register information. We defined a cohabitation relationship as a marriage-like relationship between two people who live together at the same address but who do not necessarily have to be married. As a proxy for the ‘real’ start and end date of the cohabitation relationship we used the date that these two persons start living together at the same address and the date that they do not live together at the same address anymore.

Our strategy is, in short, as follows. First, we select all couples of whom we know from a register source that they have been, still are or will become married, parents of a common child or partners for income taxes or social security benefits. By combining these data with data on address occupation, we decide whether two persons living at an address are a cohabiting couple. For example: two persons live at the same address at a certain moment in time and marry afterwards. These persons can be classified as an intimate cohabiting couple from the moment they started living together because their marriage was registered. One of the problems we encountered was that this strategy causes a bias in the database. For people that recently started living together at the same address, less information about their common future is available. Related to this issue is the question what to do when we do not know whether two individuals living at an address are a cohabiting couple or not, because they are not seen as partners in any register. We have tried to solve these issues by applying an imputation strategy.

Presented in Session 1234: Posters