http://www.iser.essex.ac.uk/publications/working-papers/iser/2009-09

Drawing on sociolinguistic research on conversations and its recent application to survey interviewing in particular, we examined a ‘remind-still’ dependent interviewing design used in a set of BHPS questions to ascertain current employment characteristics. Our theoretical approach applies the principle of cooperation and the tendency for agreement in conversations to the use of yes-no interrogatives in this set of questions. These principles imply that when respondents need to indicate a change in circumstances their tendency is to provide additional information than required by the question by way of explanation. This extra talk adds to the common ground of known information obtained during the interview. We argue that this leads sort of thing leads to subsequent departures from standard interviewing procedures as interviewers are then faced with explicit or inferred knowledge of answers to subsequent survey questions. Under conditions of no change, we assert that problems of respondent cognition and this conversational tendency are reduced, thereby minimising the likelihood of interviewer departures from standard procedures for these reasons. Controlling for respondent age, sex and education, we find that dependent interviewing questions are no different from independent questions in the occurrence of cognition problems. Dependent interviewing did seem to reduce the amount of behaviour indicative of such problems, however. Also we found a weak but significant association between interviewer departures from standard interviewing practice and the occurrence of respondent cognition problems regardless of question type. Dependent interviewing did, however, impact the survey interaction itself. We found that dependent interviewing questions were nearly 6.5 times more likely to be interrupted by respondents than regular survey questions. We also found that respondents were 2.7 times more likely to engage in answer elaboration and other forms of conversation at dependent interviewing questions than routine survey questions. Such verbal action seems to occur when respondent circumstances have changed; respondents were nearly 6 times more likely to engage in elaborations at dependent questions under conditions of change in their circumstances than no change in circumstances. We found that interviewers were nearly 3 times more likely to depart from standard interviewing procedures at follow-ups to negated dependent interviewing questions, and they were almost 2.5 times more likely to depart from standard practice at subsequent points in the interview under conditions of no change in respondent circumstances. These results suggest that under conditions of no change, dependent interviewing is consistent with conversational principles when there is no change in circumstances. We believe that this identifies a mechanism by which the observation of reduced error in the amount of change between waves of a panel survey. However under conditions of change in respondent circumstances, dependent interviewing works against conversational principles. Under conditions of change, departures from standard interviewing practice are common because respondents are more likely to engage in extraneous talk. The violations of standard interviewing procedure we observe seem to result from the tension between audience designed survey questions and the emergent common ground in the survey interview enriched by the use of dependent interviewing.

We examine how dependent interviewing affects verbal interaction between interviewers and respondents in questions obtaining current employment details in the British Household Panel Study. Respondents experience few cognition problems when answering DI questions, but interruption and elaboration are likely at PDI questions. These behaviours occur when respondent circumstances have changed. Departures from standardised interviewing are also likely when circumstances change. DI seems to reduce the accuracy of detail about such change since we observe interviewer behaviour that others find to produce inaccurate data. Nevertheless, these results may explain why DI reduces the odds of spurious change between waves of panels. Working Paper 2009-09 2009-03-27 S.C. Noah Uhrig Institute for Social and Economic Research scnuhrig@essex.ac.uk Emanuela Sala Institute for Social and Economic Research esala@essex.ac.uk