News
In a blog entry earlier this month at the site of the always-engaging Washington Post Monkey Cage, Spencer Piston of Syracuse University suggested that “lighter-skinned Latinos are more likely than darker-skinned Latinos to identify as Republican.” Some days later, Karthick Ramakrishnan of UC Riverside responded that even if skin complexion and partisanship are correlated, Latinos are on the whole Democrats. As he puts it, “the Democratic Party has a sizable net advantage in party identification, even among lighter-skinned Latinos. This is a point that can be easily overlooked when we focus on the direction of the relationship between skin tone and partisanship, without paying attention to absolute levels of partisanship among Latinos.”
In this brief remark, I wish to circle back to the possible correlation between skin tone and partisan identification among Latinos. Is there in fact such a relationship? The 2012 American National Election Study offers scant evidence of this. Approximately 450 self-identified Latinos took part in the face-to-face household portion of this study. At the end of the survey, interviewers noted the skin tone of each of these respondents based on a ten-point scale (1=very fair complexion, 10=very dark complexion). The correlation between skin tone and the standard seven-point measure of party identification is -.065, which implies that Latino citizens with fairer skin lean slightly more towards the Republicans—or, as Karthick Ramakrishnan would have it, are slightly less committed to the Democrats. But this correlation does not rise to the level of statistical significance using standard benchmarks (p=.165). If sampling weights are applied to the data, which the ANES strongly recommends, then the correlation drops to -.038 (p=.585).
Of course, summary correlation coefficients can sometimes mask noteworthy nonlinear relationships between two variables. There is not any appreciable masking in this instance, however. Below is a contour density plot that illustrates the distribution of Latino respondents along the “skin color” and “party identification” dimensions. A lowess smoothing plot is provided to explore potential curves or dips. All in all, no interesting patterns emerge. Pigment tells us nothing about partisanship, at least with respect to the Latinos in the 2012 ANES.
James McCann is a 2014-2015 Visiting Scholar and Professor of Political Science at Purdue University.