Interviewers and Data Quality in a Less Developed Setting

The problem of non-sampling error in the application of survey methods to developing countries has been raised often. This paper examines interviewer characteristics as one possible source of non- sampling error. Intensive field supervision and ethnographic cross-checks were used in Nepal to generate data on the technical quality of data from interviews. Interviewers were assigned to interactions with respondents randomly and the study incurred zero non-response. Variations in the amount of technical errors, "don't know" responses and false information gathered during the interview are analyzed as indicators of data quality. The paper examines three hypotheses. First, that interviewers are more careful in irregular interactions. Second, that respondents provide better information when interviewed by someone with similar characteristics. Third, that respondents provide better information when interviewed by females. Same-gender, cross- gender, same-ethnicity and cross-ethnicity interviews are examined controlling for a variety of interviewer characteristics to test these hypotheses. The evidence provides some support for the conclusion that interviewers are more careful in irregular interactions, but no support for the idea that matching interviewers and respondents by characteristics improves data quality. When the sex of the interviewer has an influence, female interviewers produce higher quality data than male interviewers.