Online Program

Statistical surveys among homeless people, improving methods for a better coverage

*Maryse Marpsat, INSEE (National Institute for Statistics and Economics Studies) 
Martine Quaglia, INED 

Keywords: Surveys on homelessness - Improving methods - improving coverage

Constructing a representative sample of the homeless population brings out many methodological and ethical questions. In the 1980’s, US researchers and statisticians have been the first to use methods like Time Location Sampling (a form of indirect sampling), in surveys where homeless people were sampled through the services they used.

In France the National Institute of Demography (Ined), in 1995 and 1998, and the National Institute of Statistics (Insee) for its 2001 national survey (SD2001), have built samples of users of services such as shelters and soup kitchens mainly meant for the homeless. Among them, the homeless sleeping in places not meant for habitation were reached only if they used soup kitchens. Weightings were calculated with the weight sharing method (Lavallée, Deville).

The issue of the unsheltered homeless is rather complicated. Field observations made during the 2002 Ined survey of people met by outreach services seemed to show that those who refused to go to shelters in the evening or at night could accept, after an unsheltered night, to go and spend part of the morning in a place where breakfast is served.

In order to reach those who, sleeping in places not meant for habitation, don’t use lunch or dinner services, can we enlarge the sample frame and include other services such as drop-in centres, places providing breakfast, and outreach services ?

The EMSA2009 survey conducted in Toulouse in January/February 2009 was designed to answer this question and thus prepare the 2012 Insee/Ined national survey. We will present the EMSA2009 survey and the context in which it was conducted, give results on the use of services by the people interviewed, and information on the decisions taken for the 2012 national survey.