Determining validity and identifying nonresponse bias in a survey requesting income data

Abstract

Two procedures for validating incomes self-reported in ranges by University of Kentucky doctoral graduates, one comparing grouped data to income data collected by the National Academy of Sciences (1974) for the same year, 1973, the other comparing individual self-report data in a double-blind process to state income tax records, indicating that the self-report data were generally accurate. A check of respondents versus nonrespondents in various categories (sex, age, graduation year, Holland type, Biglan type) showed no difference in the proportion of response in any category. Comparison of nonrespondent income with respondent income in state income tax records revealed that nonrespondents averaged almost $3,500.00 less income during that year than did respondents. This finding held true across all categories but two.