Conferences

The quality of survey data depends on the way in which information is collected. To limit bias, the interviewers are instructed to ensure that the interviews take place in private without anyone else present. However, they cannot impose the interview conditions and in many cases one or more outsiders are present during all or part of the interview. To evaluate bias on respondents’ answers, the interviewer indicates at the end of the interview if a third person was present, if they intervened and if, in the interviewer’s opinion, this presence influenced the respondent’s answers. Though it is essential to know if a third person was present, it is more problematic to expect the interviewer to assess the resulting bias. Is he or she really capable of doing so?

These reservations, combined with the constraints of survey design, led to the adoption of a different strategy for the first wave of the Generations and Gender Survey conducted in France by INED and INSEE in 2005. Rather than obtaining retrospective information (at the end of the survey), factual indicators of the presence of third persons were included at different points in the questionnaire. The interviewer was asked to answer the following question: “Interviewer: Are there, at this point in the interview, any other persons present in addition to the respondent? Yes / No” and, if yes “Please indicate who is now present: Spouse / Children / Brothers and Sisters / Friends / Other”. A total of five indicators are included in the course of the questionnaire.

This approach gives a more precise description of interview conditions than do other surveys, and provides a means to detect comings and goings (persons present at the beginning, but not at the end, or the arrival of someone during the interview). Although one quarter of interviews were affected by the presence of a third person at the beginning or the end, the proportion rose to one-third when these comings and goings were taken into account. The third person was usually the respondent’s spouse. It is important to have a precise description of the interview conditions because of the reporting biases that can result. The example of the breakdown of parental tasks within the couple tends to indicate that the presence of the spouse exercises a certain “control” over the respondent’s answers, while his or her absence encourages the respondent to “idealize” by presenting a more even breakdown, more in keeping with a certain egalitarian standard. The presence of the spouse also seems to encourage recall when the respondent is trying to date certain events in their conjugal life. However, although these results seem to indicate that, contrary to the instructions given to interviewers, it is preferable for interviews to take place in company of the spouse, the presence of a third party – particularly that of the spouse – means that past conjugal histories are reported less fully, especially on the male side.