Overlapping or not overlapping in dual frame surveys?

The last wave (2014) of the CATI / CAMI Health Barometer survey is considered, conducted among the French population among the 15 and the 75 years.

The survey deals with health issues (health status, alcohol consumption, drugs, etc.) and lasts on average 32 minutes; It was possible to interview 15.635 respondents, with a response rate of 47%, the 51% concluded the interview via mobile, the remaining through fixed. The waste has not been replaced. The survey also detects if the respondent has another landline / mobile phone.

The objective of the survey is to understand if the sample interviewed with an approach Overlapping it is more or less representative of the sample interviewed with approach screening (used in waves prior to 2014).

The approach screening provides for the interview by mobile phone only of those who, during the presentation of the survey, declare that they own only the mobile phone. Those who own both mobile and fixed phones are interviewed by fixed and excluded from the sample of mobile numbers. The difficulties that arise in contacting those who have both a mobile and a landline number but use mobile almost exclusively have led to the search for a new method.

The solution adopted in the 2014 is the approach Overlapping which provides for the establishment of two samples, one of fixed numbers and one of mobile numbers as per the approach screening; in this case, however, the interview is conducted regardless of whether or not there is another fixed number. To try to get the interview are made up to 40 attempts.

The comparisons considered are:

  • CTO (cellular-telephone-only) vs Population (surveyed by the 2012 Labor Force Survey): they include more people under 45 years of age, more residents in cities greater than 2000 inhabitants.
  • dual users interviewed in the sample of mobile numbers vs dual users respondents in the sample of fixed numbers: among the first there are more men, more people under 55 years of age, more residents in large municipalities (> 2000 inhab.), higher educational qualifications, a greater number of employed at high and more people who do not live alone. Furthermore, there are significant differences in terms of attitudes to smoking, alcoholism and drug use; these differences disappear after weighing the data.

Finally, it is noted that the difference between cellular contact costs and fixed contact costs is much higher in the approach screening (mobile phones in this case only include CTOs that cost 180% more than landlines on average, while in theOverlapping the difference is equal to 10%). In absolute terms the two approaches have similar costs and through theOverlapping it is easier to find the answers of the dual users; however, the response rate is lower in the approach Overlapping, due to the greater number of mobile phones present.

LTOs (landline-telephone-only) have peculiar characteristics (older, live in rural areas, live alone and have a low educational qualification) which does not justify the transition to a sampling that considers mobile phones alone.

Taken from:
Effects of using an overlapping dual-frame design on estimates of health behaviors: a French general population telephone survey

In the Journal of Survey Statistics and Methodology, volume 5, number 2, June 2017 Richard, Andler, Gautier, Guignard, Leon and Beck analyze the conduct of interviews by dual-frame overlapping compared to the approach screening.

Leave a comment