Effects of interviewers on response to income and wealth items

TL;DR

This study investigates how interviewer expectations influence respondents' willingness to disclose income and wealth data in surveys, highlighting the importance of interviewer effects on survey accuracy.

Contribution

It demonstrates that interviewer expectations are associated with response rates to financial questions and evaluates methods for handling missing data in this context.

Findings

Higher interviewer expected response rates correlate with increased respondent disclosure.

Model averaging approaches did not outperform simpler missing data methods.

Interviewer expectations provide valuable insights for improving survey response modeling.

Abstract

Item nonresponse to financial questions is a persistent source of survey error, especially in interviewer-administered surveys. We examine whether interviewers' expectations about respondents' willingness to report income are associated with actual item responses to income and asset questions in Wave 6 of the Survey of Health, Ageing and Retirement in Europe (SHARE). Using data from 41,934 respondents in 12 countries, linked to interviewer survey and roster information, we analyze responses to four financial items with substantial nonresponse. We compare three approaches to handling missing covariates: complete-case analysis, multiple imputation (fill-in methods), and a generalized missing-indicator framework with information-criterion-based model averaging. Across most specifications, respondents interviewed by interviewers with higher expected income response rates are more likely to provide financial information. However, model averaging does not yield clear gains over simpler approaches. The results suggest that interviewer expectations contain useful information for understanding and modeling item nonresponse to sensitive financial items, with potential implications for interviewer training and survey fieldwork design.