Suchergebnisse

Although there have been several investigations into race-of-interviewer effects on black R's, little is known about the same type of effects on white R's. The 1971 Detroit Area Study at the U of Michigan included a systematic variation of race of interviewer for both black & white R's. 106 out of 1,485 whites were included in this experimental variation. The results showed whites to be as susceptible to race-of-interviewer effects as black R's, & thus called into question earlier interpretations of such effects as being due to asymmetrical social deference. Black interviewers received less antiblack responses for 4 racial items than did their white counterparts. The literature on such effects on black R's shows that white interviewers receive less antiwhite or militant responses from black R's than black interviewers. The underlying process for both races seems to be one of avoiding responses that might offend the interviewer of the 'opposing' race & being frank (or franker) with members of one's own race. Those affected by race-of-interviewer, however, tend to be different for the 2 races. Education seems to be a key factor. Lower educated blacks & higher educated whites tend to be more susceptible to race-of-interviewer effects. 2 Tables. AA.

Although `racial matching' of interviewer and subject may often be appropriate, as a political strategy it risks marginalisation of black researchers and, as a methodological approach, its assumption of a single `truth' is open to challenge. These issues are explored through an examination of public and private accounts and their relationship to race-of-interviewer effects and a discussion of the author's experience of conducting interviews with prospective black foster parents. The final section explores the role of the white researcher.

This paper reports the results of research examining the issue of non‐response bias to mailed surveys. The hypothesis for this experitnent was that lower response rates of black citizens to mail questionnaires could be overcome, in part, by using black sponsoring institutions. A city was selected which contained both a traditionally black and a traditionally white state supported university. Two samples of randomly chosen registered voters from selected precincts received the same mailed survey. The samples consisted of equal numbers of black and white registered voters. Some precincts were racially segregated (all black or all white) and others integrated and were from middle and working class neighborhoods. If a total design method is used (in contrast to a single shot survey design) institutional afiliation does not affect response rates of black or white citizens. The research helps to elucidate empirically the complexity of the response rate hypothesis as it applies to race.

Research was conducted to ascertain what effects the race of the interviewer has on survey responses, & to determine whether racial or interpersonal deference provides the most satisfactory explanation. The extent to which SS amplifies the race-of-interviewer effect (RE) was also tested. The data were gathered in a 1974 study of 944 Atlanta, Ga, area high school seniors. Each R was interviewed for 1 hour, face-to-face, by 1 of 12 professional, Mc F interviewers, 8 of whom were white & 4 black. Twelve items (4 racially neutral & 8 racially valenced) were presented as part of a longer instrument. In addition, following the interview each R filled out a paper & pencil form without reporting the answers to the interviewer. It was hypothesized that these latter items would be freer from the RE than those collected in the conventional way. All items were closed-ended questions. Bias arising from the RE was found to occur among both black & white Rs. However, such bias was limited to items mentioning the race of the interviewer, in an interview with a member of the opposite race. The direction of the bias was always toward deference to the interviewer's race. Low SS Rs showed a sporadic tendency toward increased deference. RE persists in spite of the increased privacy of the written responses. In all cases, the magnitude of the RE rarely exceeded 3% of variance explained. It seems unlikely in most instances that the RE will produce serious misinterpretations in the substantive realm. 3 Tables, Appendix. Modified AA.