An Age–Period–Cohort Analysis of Political Tolerance in the United States

Abstract

We employ hierarchical age–period–cohort models and the 1974–2010 General Social Survey data to examine changes in the political tolerance of gays and lesbians, communists, racists, and anti-religionists. Results show period-based growth in political tolerance, cohort-based growth in tolerance of anti-religionists, baby boomers are particularly tolerant, and political tolerance is associated with changes in college education. The findings suggest that liberalizing trends in political tolerance are largely motivated by changes among Americans as a whole, not cohort replacement, that baby boomers are unique in their social and political perspectives, and that aggregate changes in higher education are correlated with changes in political tolerance.

ACKNOWLEDGMENTS

The authors wish to thank Jacob Cheadle, Kristen Olson, the editors of The Sociological Quarterly, and the anonymous reviewers for their helpful comments and suggestions.

NOTES

Notes

¹ There are, of course, many factors that may influence an individual's support for the civil rights of stigmatized groups, which we attempt to account for in our analyses (see Sullivan et al. 1982; Gibson 2006). Here, however, we are focusing on societal change in views of stigmatized groups.

² Respondents were asked: “What is your religious preference? Is it Protestant, Catholic, Jewish, some other religion, or no religion?” The religious nonaffiliation variable compares those who say they have no religion (= 1) with all other respondents.

³ Political scientists argue that individual-level variation in political tolerance is largely attributed to psychological factors (predominantly dogmatism), perceptions of group threat, and political perspectives (Sullivan et al. 1982; Gibson 2006). Unfortunately, there are no relevant measures of dogmatism or perceived threat of the outgroups in question across multiple years of the GSS. In regard to political perspectives, while we cannot assess support for the norms of democracy (Gibson 2006), we can measure respondents' political orientations. Specifically, we include a measure of political orientation that ranges from extremely liberal (= 1) to extremely conservative (= 7).

⁴ A dummy variable for female respondents controls for sex. Race is assessed with dummy variables for African-American and “other” race respondents, with whites as the reference category. Family income is measured with a continuous measure of income in 1986 dollars, logged to adjust for the skewed distribution. Dummy variables for currently married respondents and those with children under the age of 18 living in their homes control for family formation. Religious service attendance is an ordinal variable ranging from never (= 0) to more than once a week (= 8). Urbanity is based on dummy variables for respondents in the 100 largest standard metropolitan statistical areas (SMSAs) (“urban”), suburbs of the 100 largest SMSAs (“suburban”), and rural areas, with other urban areas as the reference category. A dummy variable for respondents in South Census Region controls for regional variation. Descriptive statistics for all independent variables are available upon request from the lead author.

⁵ For instance, the predicted probability for birth cohort j = exp(β₀ + u_j) / (1 + exp(β₀ + u_j)). These can be seen as “conditional” predicted probabilities, as opposed to “population-averaged” or “marginal” predicted probabilities (Skrondal and Rabe-Hesketh 2009). The difference between conditional and marginal predicted probabilities are quite small when the intra-class correlation (ICC) is below .5 (Skrondal and Rabe-Hesketh 2009), and all ICCs are well below .5 in the models presented in this article.