VALIDITY OF WELFARE MOTHERS' INTERVIEW RESPONSES.

The article focuses on the response accuracy of a sample of poor African American welfare mothers in New York. They were New York City residents receiving public assistance in 1966 who had been interviewed by the National Opinion Research Center for a study on the use of health services. The African American welfare mothers in the study voted much less often than other groups studied. Only 29 percent voted in the 1964 election. Nevertheless, they acknowledge their nonvoting with as much candor as other groups. Two sets of characteristics distinguish the accurate from the biased respondent on the voting questions. The biased respondent is high on community integration. The fact that the same characteristics that are associated with voting are associated with response bias distorts not only the absolute figures but also the analysis of relationships among variables. It is usually assumed that social distance between respondent and interviewer will adversely affect response accuracy. Disparities in social status were limited to socioeconomic status, education and age. It was expected that welfare mothers would give the interviewers most unlike themselves a more biased, idealized story.