White racial attitudes: Does it matter how we ask?

Abstract

Much of our systematic knowledge of the nature and correlates of white racial attitudes is based on sample surveys administered by interviewers. Questions are often raised about the validity of such data due to social desirability pressures to appear racially liberal. Using a quasi-experimental design, this study investigates the nature and extent of this bias and its relevance to understanding white racial attitudes. Specifically, responses are compared in three conditions of administration: a standard condition where interviewers asked all the questions; a modified face-to-face condition where respondents answered a subset of questions in a self-administered form; and a completely non-interviewer condition where questionnaires were mailed to respondents. The primary hypothesis of this study was that white respondents will express less liberal racial attitudes as conditions increase in confidentiality, and that this should be especially true of those with more education. The underlying assumption is that liberal racial attitudes have become increasingly expected in the United States, and that the presence of an interviewer increases pressures on respondents to say what is "socially desirable." More educated respondents would be more aware of such norms and feel pressure with greater force. These hypotheses were believed to operate differently for different types of racial questions, so questions that varied in their likely sensitivity to social desirability were used. The original sample consisted of essentially a cross-section of the white population of the metropolitan Detroit area, divided randomly into the three conditions. The three samples were comparable in response rates and on most demographic variables, with the key exception being a higher educational level for the mail survey, which was then controlled through weighting and multivariate analysis. Although there was initial support for the main hypotheses, some puzzling findings led to a second hypothesis: less educated respondents are subject to another form of social pressure--a tendency to agree with cliched statements. This tendency, known in survey methods literature as "acquiescence," might be expected to be heightened in a survey condition where an interviewer reads a plausible sounding statement. When the racial questions were reanalyzed in terms of both social desirability and acquiescence, the results provided support for both theoretical assumptions. This quantitative analysis was supplemented by depth interviews with a small sub-sample of mail survey respondents. These additional qualitative data were used to enrich the interpretation of the survey results, especially to show the complexity of the racial attitudes of white Americans. This complexity might otherwise be missed when we speak too simply of "social desirability.".