Household data in the Norwegian Labour Force Survey

This is a technical document describing quality analysis of survey data at the household level. The Norwegian Labour Force Survey uses registered nuclear family as the sampling unit, while actual household is the required unit in data transmitted to EUROSTAT. The purpose of this project was to evaluate how the sampled unit serves as a proxy for the actual household, and what improvements can be made. Register-based data sources have been linked to survey data at the micro level, in order to compare household identification and classification in detail. Some macro level comparison has been made for the labour market indicator “workless households”.

Some main results and conclusions:

• The response survey sample is biased with respect to household size, mainly due to higher nonresponse rate among one-person-households.

• The main divergence between various data sources for household information is the classification of 1- and 2-person-units.

• Comprehensive register-based household data differs considerably from the sampling unit in the LFS, which is ‘registered nuclear family’. On the other hand, the more comprehensive data require longer production time and thus not considered as an ideal sampling frame.

• The survey-based household identification module consists of a set of questions asked to a subsample. This considerably improves the quality of household size distribution. Specifically, many registered ‘one-person-households’ are identified as defacto couples.

Statistics Norway is in the process of developing new IT-solution for interviewing the defacto household members who are not in the sampling frame, only identified by the survey. Currently, they are identified but not interviewed about their labour market situation.