Keywords: interviewer effects, social distance, CAPI, India

A rich literature over the past several decades has examined interviewer effects arising due to interviewers’ demographics, attitudes, interviewing behaviors, etc. What is less studied is how survey estimates are impacted due to the social distance between interviewers and respondents. Since a survey interview is essentially a structured social interaction, the respondent and the interviewer are both likely to be influenced by the perceived social differences between them. For example, an interviewer may mentally develop a set of expected responses based on a respondent’s background and the respondent may align her responses to conform to social expectations, leading to biased estimates.

Social distance effects have been studied in the context of the race of the interviewer and respondent in the United States, but little literature has explored social distance effects in the developing world. This is ironical, since many such societies have complex social structures warranting deeper study. Moreover, while standardized interviewing is the preferred paradigm in the Western world, such a paradigm is difficult to adopt in other parts of the world where the interviewer would be treated as a guest and strictly standardized interviewing would be considered artificial and socially awkward. Thus, the interviewing in such geographic contexts proceeds more conversationally, potentially subjecting estimates to more social distance effects.

Our study focuses on India, a society known for its pronounced socio-linguistic diversity and the presence of historically entrenched social norms and hierarchies (caste). We use data from the third wave of the India Human Development Survey (IHDS), a large-scale pan-India panel survey of 47,000 households, fieldwork of which was conducted between April 2022 and June 2024. Our primary measure of social distance is caste and secondarily, education, age, and marital status.

Our focus is on sensitive questions in the Main Household instrument (where the respondent and the interviewer are both typically males) and the Eligible Woman (EW) instrument (where the respondent and the interviewer are both female). Preliminary analysis shows, for example, that lower caste female respondents report a much lower agreement with domestic violence situations (e.g., “Is it okay for a husband to beat a woman when under the influence of alcohol”) when interviewed by higher caste interviewers than when interviewed by persons from their own caste, all other factors being adjusted for. We study the impact of the interview location (in the bedroom, outside the dwelling, etc.) and privacy in moderating these social distance effects.
Unlike race, caste can be difficult to predict from skin color. It is therefore intriguing to find the presence of such effects. To further understand the mechanisms producing social distance effects, we: (a) Listen to a sample of recorded interviews and analyze interviewer-respondent interactions (b) Use data from in-depth interviews conducted on interviewers.