TVSEP junior researcher Mark Brooks presented his paper 'PAPI is gone, but errors remain - Non-sampling errors in household surveys' in the panel session 'Survey/Method' on the 14th day of the triennial International Conference of Agricultural Economists, which took place on the 17th-31st of August. Data collection in household surveys based in developing countries has transitioned from Paper and Pencil Interviews (PAPI) to Computer Assisted Personal Interviews (CAPI) as a result of technological advances. CAPI is found to essentially eliminate item missing data, whereas measurement error, while being significantly reduced, remains.
The paper analyses the determinants of non-sampling errors arising during the interview. A comprehensive negative binomial regression model is established, which accounts for characteristics of the interviewer and respondent as well as the interview and survey environment. The main drivers of non-sampling error are identified as stemming from quantitative and qualitative interviewer and respondent characteristics. Notably, congruency of ethnicity is found to be a key predictor of data quality in countries with diverse cultures.