Engendering Racial Perceptions

Abstract

Intersectionality emphasizes that race, class, and gender distinctions are inextricably intertwined, but fully interrogating the co-constitution of these axes of stratification has proven difficult to implement in large-scale quantitative analyses. We address this gap by exploring gender differences in how social status shapes race in the United States. Building on previous research showing that changes in the racial classifications of others are influenced by social status, we use longitudinal data to examine how differences in social class position might affect racial classification differently for women and men. In doing so, we provide further support for the claim that race, class, and gender are not independent axes of stratification; rather they intersect, creating dynamic feedback loops that maintain the complex structure of social inequality in the United States.

Keywords

gender race stratification incarceration unemployment poverty welfare

For many gender scholars there is an inherent tension between survey-based, quantitative methods and the study of intersectionality. As McCall (2005) notes, there is considerable scorn for the mathematical language (“additive” vs. “multiplicative”) often used to describe the insights of intersectionality and frustration at the implication that the entire paradigm can be reduced to a series of “interaction effects” in a multivariate model. At the same time, research from a traditional stratification perspective is often dismissive of insights gleaned from case studies and personal narratives, arguing that claims about how systems of inequality operate cannot be validated without more generalizable data. As a result, conversations across these methodological, epistemological, and disciplinary divides remain limited, despite shared interest in understanding the mechanisms that perpetuate social difference and division (McCall 2005).

We aim to further the study of intersectionality by demonstrating that statistical modeling is not necessarily antithetical to examining the intersections of race, class, and gender. In doing so, we take seriously recent critiques that stress the importance of asking dynamic, process-oriented, and relational questions about how race, gender, and social class divisions overlap in people’s lives (Choo and Ferree 2010). Drawing on nearly 20 years of longitudinal data from the National Longitudinal Survey of Youth (NLSY, 1979 cohort), we examine whether changes in how people are perceived racially by others are influenced by social class position in ways that differ by gender. Specifically, we ask: If Americans are more likely to be seen as Black and less likely to be seen as white after experiencing decreases in status, such as becoming incarcerated or impoverished (Saperstein and Penner 2012), are these effects stronger for women or men? Or do different sets of class-relevant factors matter depending on the gender of the person being racially classified?

Our results highlight that an individual’s position in the “matrix of domination” (Collins 1990) is not necessarily fixed. Consistent with the idea that race is a changeable marker of status encoded onto the body (Omi and Winant 1994), we find that class mobility can trigger racial mobility, both upward and downward, and the specific trajectories often vary for women and men. For example, while some status cues, such as living in the suburbs, make both women and men more likely to be seen as white, others have differential effects either in magnitude, significance, or both, as when receiving welfare makes women (but not men) more likely to subsequently be seen as Black and incarceration makes men (but not women) less likely to subsequently be seen as white. These findings not only illustrate how intersectional processes affect individuals in their everyday lives, they also provide further evidence, nationally representative and longitudinal, of one of the central claims of the field: that race and class are gendered, and gender and class are racialized.

Race, Class, Gender, and Inequality

The term intersectionality was coined by Crenshaw (1989) to describe the unique situation facing Black women pursuing redress for their antidiscrimination claims in U.S. courts. Was the nature of their complaint racial, or based on their gender? Rather than argue both, their lawyers had to decide which aspect of discrimination made the more compelling case. Crenshaw (1989) insisted that the two experiences could not be separated so easily.

This simple, but profound, insight—dating at least to Smith and Stewart (1983) and echoed in the work of Collins (1990)—launched a new area of inquiry across the social sciences. More than 20 years later, it has been credited with bringing marginalized groups to the center of academic scholarship, and the original trinity of intersectionality (race–class–gender) has expanded to include studies of how sexual orientation, disability, nationality, and age are experienced differently by different groups (Choo and Ferree 2010).

Implementing an Intersectional Analysis

Over time, both the “what” and the “how” of studying intersectionality have been in dispute. Some suggest the term is more of a buzzword than a coherent theoretical or methodological paradigm (Davis 2008). Others identify a spectrum of intersectional approaches ranging from postmodernist “anticategorical” to comparative “intercategorical” analysis (McCall 2005). A growing quantitative literature directly engages intersectionality and evaluates its empirical claims (e.g., Harnois 2010; Steinbugler, Press, and Dias 2006). Yet, it is often implied, if not explicitly stated, that qualitative research methods are the most appropriate means of studying the intersectional complexity of social difference, in part because they tend to be better suited to identifying mechanisms and process rather than broad patterns (Choo and Ferree 2010).

A common critique of quantitative studies of inequality is that, even when they consider the effects of race, class, and gender (and/or other categories of difference), they often assess each effect independently and ask which axis of stratification is most prominent in a given situation. This approach goes against the claim that the various dimensions are inextricably intertwined and always operating in conjunction. We adopt an intermediate position, in line with others (e.g., Browne and Misra 2003; Timberlake and Estes 2007), that asserts the extent of the co-constitution of race, class, and gender inequality is an empirical question and may vary across contexts. Following Glenn (2002), we use standard racial and class categories as “anchors,” without assuming that they are static, and explore how they change relative to one another over an individual’s life course. That is, we are both questioning the categories’ essentialism and using their purported existence for analytical leverage to demonstrate how boundaries are drawn between them.

Our particular analytical focus is whether gender shapes the relationship between social status and race. Ideally, we would examine not only how the racial categorization of individuals changes over time in response to changes in social position, but also how sex categorization changes as a function of whether respondents conform to gendered racial and class norms (cf. Freeman and Ambady 2011). The NLSY does include data on how interviewers classified the respondents’ sex each year from 1979 to 1998, but there is relatively little change over time, rendering statistical analysis less useful.¹ It would be interesting to examine the causes and consequences of changes in sex categorization, as well as whether gender fluidity and racial fluidity are related, but we are unable to do so in this study, and so focus instead on whether the respondent’s self-identified sex influences the effect of social position on racial categorization.²

Race, Gender, and the Content of Our Stereotypes

In the United States, Black people are stereotyped as musical and athletic, but also lazy, unintelligent, and unethical (Blair, Judd, and Fallman 2004; Devine and Elliot 1995). Simply mentioning the phrase “inner city” cues racial stereotypes and affects whites’ attitudes toward crime policy (Hurwitz and Peffley 2005). Even biracial Blacks (Harris and Khanna 2010) and new immigrants of African descent (Waters 1999) accept these negative stereotypes and use them to explain why native (and monoracial) Blacks are overrepresented at the bottom of the social hierarchy.

However, American racial stereotypes, from Ronald Reagan’s “Welfare Queen” to Willie Horton on a prison furlough, have never been only about race. These ostensibly racial cues have both social class and gender dimensions. Like racial stereotypes, gender stereotypes and perceptions of innate difference reinforce hierarchies between women and men. Women are expected to be warm, sensitive, and interested in children, but not stubborn, controlling, or promiscuous, while men are supposed to be self-reliant and assertive, but not weak or gullible (Prentice and Carranza 2002). These widespread, “hegemonic” ideals facilitate gendered status distinctions, such that those who are unable to meet stereotypic expectations are relegated to subordinate or “pariah” positions in the social order (Schippers 2007).

A growing body of research highlights the importance of understanding how race and gender stereotypes and their associated value judgments intersect. For example, while men are typically perceived to be more competent than women (Correll and Ridgeway 2003), these evaluations do not necessarily hold for Black men (Wingfield 2009). Pervasive stereotypes of Black men in the United States portray them as shiftless (Kirschenman and Neckerman 1991), irresponsible fathers (Glauber 2008) and prone to violence and criminal behavior (Eberhardt et al. 2004). Black women, on the other hand, are widely perceived as single mothers (Kennelly 1999), loud and confrontational (Weitz and Gordon 1993), and undeserving of government benefits (Gilens 1999).

Even where men and women of the same race face similar stereotypes, there are often differences in the degree to which the stereotypes are thought to apply. For example, although Blacks are stereotyped as less intelligent than whites, Black women are seen as more intelligent than Black men (Steinbugler, Press, and Dias 2006). These persistent associations lead Browne and Misra (2003) to conclude that women and men of color face different stereotypes that lead to discrimination.

Experimental research in social psychology provides support for this perspective as well. Studies show that when race and gender are cued simultaneously, people combine the characteristics to generate categorizations (Freeman and Ambady 2011) and behavioral expectations (Correll and Ridgeway 2003), rather than rely on one piece of information or the other. This research also demonstrates that evaluations related to a person’s organizational roles “become nested within the prior, automatic categorization of that person as male or female and take on slightly different meanings as a result” (Ridgeway 1997).

At the same time, although category attributions of race and gender are largely automatic and occur within the first few seconds of perceiving someone, research shows that introducing other seemingly contradictory characteristics or status cues can alter these perceptions (Freeman et al. 2011; Purdie-Vaughns and Eibach 2008). For example, racially categorizing admired Blacks and disliked whites takes longer and prompts more “mistakes” than categorizing admired whites and disliked Blacks (Richeson and Trawalter 2005). In this study, we examine whether similar processes occur during nonexperimental interactions, as Americans sort through a host of information to help them determine how to racially categorize someone. In particular, we are interested in whether the effects of social status on racial perceptions vary significantly by the target’s gender.

Data and Methods

Our data come from the 1979 cohort of the NLSY, a representative sample of 12,686 U.S. men and women who were 14 to 22 years of age when first surveyed in 1979. Respondents were eligible to be interviewed every year thereafter, until 1994, when interviews began occurring biennially. We utilize data from 1979 to 1998, the most recent year in which the interviewers recorded their racial classification of the respondents. To define our study populations, we use the respondent’s self-reported sex from 1979.

Racial Classification

Our dependent variable is the interviewer’s classification of the respondent’s race, which was collected in all but one survey year from 1979 to 1998. Interviewers were instructed to classify the respondent’s race at the end of the interview. Thus, we do not have their first impression of the respondent’s race; we have a classification colored by the respondent’s answers during the survey interview. This is ideal for assessing the effects of social status on racial classification, because the interviewer heard a range of information about the respondent, from their income and education to their employment and marital history, prior to recording the respondent’s race.

Interviewers were not given any special instructions as to how to classify the respondents by race (NLS 2006); the categories available to them were “Black,” “White,” and “Other.” Descriptive statistics on how respondents were racially classified by the interviewer are reported in Table 1, along with our other key variables. Of the observations where respondents have racial classifications in consecutive survey years, 6 percent are described by a different race than in the previous person-year, and 20 percent of the individuals in the sample experienced at least one change in how they were racially classified between 1979 and 1998.³

Table 1:

Descriptive Statistics

	Women	Men
Racial classification
Percentage classified as white by interviewer	67.8	67.2
Percentage classified as Black by interviewer	26.6	27.1
Percentage classified as other by interviewer	5.6	5.7
Respondent characteristics
Percentage unemployed over 4 months	8.9	10.1
Percentage below poverty line	22.9	17.1
Percentage incarcerated	0.2	2.6
Percentage receiving welfare	17.5	6.8
Percentage married with children	31.5	24.9
Percentage married without children	11.2	10.0
Percentage unmarried with children	19.9	3.8
Percentage unmarried without children	37.4	61.3
Percentage with 4 or more children in household	3.2	1.5
Percentage living in an inner city	18.2	19.0
Percentage living in a suburb	30.4	30.0
Mean years of education	12.4	12.2
Percentage self-identified as Hispanic in 1979	17.3	16.5
Percentage self-identified as multiple races in 1979	11.8	9.1
Percentage born outside the United States	6.8	7.0
Percentage living in the South	39.5	37.4
Mean age	25.9	25.7
Interviewer characteristics
Percentage female	95.8	88.7
Percentage self-identified as white	84.2	84.6
Percentage self-identified as Black	11.4	10.9
Percentage self-identified as other	4.4	4.5
Mean age	49.4	49.6
Percentage less than a high school degree	1.3	1.5
Percentage high school graduate	22.1	20.3
Percentage some college/vocational degree	37.9	38.7
Percentage bachelor’s degree or more	38.7	39.4

SOURCE: 1979 National Longitudinal Survey of Youth.

NOTE: Statistics for person-years (N=182,224; representing as many as 12,686 unique persons in a given year).

It is possible that these changes in respondents’ racial classification from one survey year to the next were the result of mistakes made by the interviewers if, for example, they were in a hurry to complete their remarks and meant to check “White” but mistakenly checked “Black.” To the degree that the mistakes were randomly distributed, this explanation would decrease the signal-to-noise ratio in our analyses, making it difficult to find evidence of a relationship between gender, social position, and racial classification. For additional discussion of why changes in the racial perceptions of the same individual should not be interpreted primarily as errors, see Saperstein and Penner (2012).

However, we do not assume that the interviewer’s classification is “true” or correct in any objective sense. We also do not assume that the interviewer’s racial classification necessarily aligns with how the respondent would self-identify; rather we view self-identification and classification by others as capturing different information about an individual’s race and racial experiences (Saperstein 2008). We regard the interviewer’s racial classification as a measure of how the respondent is likely to be perceived, and presumably treated, by others. Indeed, some scholars have argued that racial classification is actually a more appropriate measure than self-identification for understanding phenomena such as discrimination (Telles and Lim 1998). For this reason, understanding who gets racially classified as what, and under what circumstances, is an important question for sociological analysis.

Social Status

Our primary independent variables are repeated measures of the respondent’s social position, such as long-term unemployment, poverty, incarceration, and welfare receipt. Because of widespread racialized stereotypes regarding who goes to prison or collects welfare benefits (Timberlake and Estes 2007), we expect that people who experience a loss of status on any of these dimensions will be more likely to be classified as Black, and less likely to be classified as white. Further, given the gendered nature of these stereotypes, we expect some changes in social position to have different effects on the racial classifications of women compared to men.

Our measure of unemployment indicates whether the respondents ever reported being unemployed for more than 16 weeks in the previous calendar year, and welfare receipt indicates whether the respondents, or their partners, ever reported receiving Aid to Families with Dependent Children (AFDC and later TANF), food stamps, Supplemental Security Income (SSI), or any other public assistance during the previous calendar year. To capture the experience of poverty, we examine whether the NLSY ever indicated that the respondent’s total family income was below the poverty level in the previous calendar year. To measure incarceration, we use data on whether respondents were ever interviewed in jail or prison. These factors are not particularly common in any given year (e.g., in all person-years, just 1 percent of the sample is incarcerated on average, and 12 percent receive welfare benefits), so we focus on differences between respondents who have and have not ever experienced these changes in their social position. Given evidence of the lasting effect of incarceration on racial classification (Saperstein and Penner 2010), we suspect these status cues not only operate directly—as when the interview takes place in prison, or information is given explicitly in response to a question—but also shape perceptions of respondents indirectly and might accumulate over time.

We also examine other class-relevant factors linked to racial stereotypes, such as marital and parity status (indicators for married with children, married without children, and unmarried with children, with unmarried without children as the omitted category), family size (an indicator for having four or more children living in the household), neighborhood type (indicators for living in the inner city and living in the suburbs, with all else as the omitted category), and years of completed education. Changes in these characteristics are more common, so we allow these measures of status to vary from one year to the next. These factors are also likely to be immediate status cues during the survey interview, as the vast majority of interviews take place in the respondent’s home.⁴

Other Respondent Controls

The likelihood of being racially classified in a particular way and experiencing the changes in social status detailed above are both related to other characteristics of the respondent. Thus, we include controls for the respondent’s age (in years), country of birth (an indicator for born outside the United States), whether the respondent gave multiple racial/ethnic origin responses or reported a Hispanic origin in 1979, and whether the respondent currently lives in the South.⁵

Interviewer Characteristics

We control for demographic characteristics of the interviewer, in part because of research noting differences in interviewers’ perception of respondents’ skin tone (Hill 2002). These controls include age, race (self-identification as white, Black, or other), sex (an indicator for identifying as female), and education (indicators for some college and college graduates compared to everyone else). We also include year-fixed effects (i.e., indicators for the year of interview) in each model. These account for trends in classification caused by changes in the social meaning of race (e.g., due to increased immigration or the multiracial movement), changes in the salience of racial stereotypes due to media coverage of current events (e.g., the OJ Simpson trial), as well as year-to-year changes in survey design, question wording, or interviewer training.

Rather than use only cases with full information on all of our independent variables, we generate indicators for whether data are missing or not, recode missing data to zero, and include both the standard controls and indicators for missing data in our models. Interviewer characteristics account for most of the missing cases, and supplementary analyses (available upon request) confirm that models using listwise deletion produce similar results.

Models

To examine the effects of social position on racial classification, we estimate logistic regression models predicting whether respondents are classified as white (vs. non-white) and Black (vs. non-Black) in a given year. We use dichotomous outcome variables to explore which characteristics move individuals either toward or away from the traditional poles of racial privilege and stigma in the United States: whiteness and Blackness.⁶ The models examining whiteness control for whether the respondent was racially classified as white (vs. non-white) in the previous year, while the models examining Blackness control for whether the respondent was racially classified as Black (vs. non-Black) in the previous year. This allows us to isolate, for example, which factors are related to the likelihood of being classified as white in the current year, net of how the respondent was racially classified previously.

We estimate models separately for men and women to examine whether there are differences in which characteristics are related to the interviewer’s racial classification. We also use fully interacted models to test whether the effects of social position on racial classification are statistically significantly different between women and men. These results are based on models that include all of our status variables together, which indicate, for example, whether there is an effect of unemployment on racial classification net of the other factors. With the exception of education (discussed below), the results are similar when we examine models that include only one status variable at a time.⁷

Results

We begin by presenting descriptive evidence that racial classification changes over time with similar frequency for women and men. We then estimate regression models to determine whether changes in perceived race are related to differences in social status, net of characteristics of both the survey interviewer and respondent—and whether or not the patterns differ between women and men.

Changes in Race over Time

Previous research has documented the existence of fluidity in racial classification in the United States for the same individuals at different points in their life course (Brown, Hitlin, and Elder Jr. 2007; Saperstein and Penner 2012). However, whether the level or patterns of change differ by gender has yet to be empirically explored. Table 2 presents year-to-year changes in racial classification in percentage terms. Each row is restricted to respondents who were classified as white, Black, or other, respectively, in the previous year and reports how they were classified by the interviewer in the current year. These categories are mutually exclusive as measured in the NLSY; thus, if racial classifications did not change over time, all of the observations would be in the diagonal cells. Observations in the off-diagonal cells indicate, for example, that on average 1.3 percent of women who were classified as Black in the previous year are classified as white in the current year. Similarly, on average 3.7 percent of women classified as white in the previous year are currently classified as other. Comparing the results for women (Panel A) and men (Panel B) reveals that the overall level of change in racial classification is almost identical. It is not the case that only women, or only men, experience changes in racial classification. Rather, racial fluidity characterizes the lived experience of race equally for women and men.⁸

Table 2:

Changes in Interviewers’ Racial Classification of Respondents Between Two Interviews

	Current year’s race
Previous year’s race	White	Black	Other	Total
Panel A: Women
White (%)	95.8	0.5	3.7	100
(N)	52,656	294	2,040	54,990
Black	1.3	98.2	0.5	100
	290	21,365	109	21,764
Other	44.9	2.6	52.5	100
	2,016	117	2,353	4,486
Total	67.7	26.8	5.5	100
	54,962	21,776	4,502	81,240

Panel B: Men

White (%)	95.8	0.5	3.7	100
(N)	51,065	260	1,996	53,321
Black	1.2	98.4	0.4	100
	263	21,185	90	21,538
Other	45.3	2.0	52.7	100
	2,025	88	2,355	4,468
Total	67.3	27.1	5.6	100
	53,353	21,533	4,441	79,327

SOURCE: 1979 National Longitudinal Survey of Youth.

NOTE: Observations represent person-years. Percentages sum across rows.

When interpreting the results in Table 2, it is important to remember that we are pooling data for individuals over time, so our unit of analysis is a person-year. Thus, the percentages describe, on average, how an individual classified as a particular race in the previous year will be seen in the current year, and not what proportion of people who have ever been classified as a particular race will ever make the transition in question (e.g., change their racial classification from Black to white between two years). These percentages are higher: Among respondents who have ever been classified as Black, 14 percent of individuals experience a change in classification from Black to white in consecutive interviews, and 4 percent experience a change from Black to other. Among respondents ever classified as white, 5 percent experience a change from white to Black in consecutive interviews, while 19 percent experience a change from white to other. Finally, among respondents ever classified as other, 88 percent experience at least one classification shift from other to white, and 8 percent experience a shift from other to Black in consecutive interviews. Like the fluidity observed in the year-to-year changes, the likelihood of ever experiencing a particular racial transition does not vary by gender.

The Effects of Social Status on Race

Having established that both women and men experience racial fluidity, we now turn to modeling which social factors influence racial classification. Though changes in race are equally likely for women and men overall, it does not necessarily follow that the changes for men and women are driven by the same factors, as we demonstrate below.

Table 3 presents odds ratios from two sets of logistic regression models estimating how interviewers racially classified women and men, alongside p values from fully interacted models testing whether the coefficients for women and men differ. The first two models examine the odds of being classified as white (vs. all else) separately for women and men. The third column in the table presents the p value from a model in which all independent variables were interacted with gender to test whether the differences in odds ratios for women and men are statistically significant (full model results available upon request). The other three columns in the table present analogous models examining the odds of being classified as Black (vs. all else).

Table 3:

Odds Ratios From Logistic Regression Models Predicting Racial Classification By Interviewer

	Classified as White			Classified as Black
	Women	Men	p value of Women-Men	Women	Men	p value of Women-Men
Ever long-term unemployed	0.71***	0.77***	.062	1.52***	1.33***	.129
	(−9.59)	(−7.33)		(6.38)	(3.66)
Ever below poverty line	0.74***	0.65***	.011	1.34***	1.527***	.207
	(−7.03)	(−10.07)		(4.09)	(5.33)
Ever incarcerated	1.00	0.64***	.090	0.88	1.57**	.278
	(−0.01)	(−5.93)		(−0.26)	(2.95)
Ever received welfare	0.75***	0.93	.000	1.47***	1.09	.019
	(−5.97)	(−1.45)		(4.61)	(0.87)
Married with children	1.261***	1.37***	.213	0.77**	0.70***	.466
	(4.68)	(7.27)		(−3.10)	(−4.19)
Married without children	1.42***	1.53***	.356	0.81	0.58***	.078
	(5.90)	(7.61)		(−1.82)	(−3.92)
Unmarried with children	0.65***	0.85	.009	1.72***	1.58*	.739
	(−7.78)	(−1.95)		(4.97)	(2.30)
Four or more children	0.67***	0.80	.306	1.41	1.07	.371
	(−3.95)	(−1.45)		(1.93)	(0.24)
Lives in an inner city	0.62***	0.65***	.311	1.88***	1.49***	.053
	(−5.10)	(−5.14)		(5.44)	(3.39)
Lives in a suburb	1.17**	1.22**	.182	0.88	0.87	.662
	(2.59)	(3.20)		(−1.43)	(−1.47)
Education (years)	0.97**	1.02*	.000	1.06**	0.98	.008
	(−2.84)	(2.48)		(2.71)	(−0.95)
Same race previously	160.50***	173.19***	.324	4189.2***	5913.6***	.011
	(58.00)	(52.99)		(74.55)	(79.88)
Hispanic in 1979	1.10	1.11	.961	0.19***	0.21***	.760
	(0.54)	(0.55)		(−5.69)	(−4.14)
Multiple races in 1979	1.72***	1.70***	.870	0.69**	0.68**	.947
	(7.87)	(6.82)		(−2.81)	(−2.96)
Born outside the US	0.88	1.05	.290	1.17	0.60	.093
	(−0.98)	(0.30)		(0.65)	(−1.55)
Lives in the South	0.63***	0.63***	.787	1.87***	1.79***	.523
	(−5.54)	(−5.77)		(7.49)	(6.47)
Age (years)	1.11**	1.01	.052	0.86*	1.04	.050
	(2.61)	(0.22)		(−2.00)	(0.57)
Age-squared	1.00*	1.00	.145	1.00	1.00	.085
	(−2.28)	(−0.52)		(1.82)	(−0.43)
Female interviewer	1.30	1.33	.968	0.57	0.67	.394
	(1.05)	(1.66)		(−1.78)	(−1.44)
Interviewer age (years)	1.01	1.01	.379	1.01	1.01	.540
	(1.83)	(1.72)		(1.94)	(1.71)
Interviewer some college	0.94	0.95	.946	1.00	1.07	.581
	(−0.45)	(−0.40)		(−0.04)	(0.49)
Interviewer college grad	1.09	0.98	.233	0.81	1.13	.022
	(0.68)	(−0.15)		(−1.89)	(0.83)
Black interviewer	0.34***	0.31***	.489	3.14***	3.31***	.801
	(−7.44)	(−9.58)		(6.18)	(5.47)
Other interviewer	0.56	0.64	.595	0.97	0.59	.292
(non-black, non-white)	(−1.44)	(−1.13)		(−0.09)	(−1.68)
Observations (person-years)	81,237	79,297		81,233	79,297

SOURCE: 1979 National Longitudinal Survey of Youth.

NOTE: All models include survey year fixed effects and indicators for missing data. z statistics account for clustering on interviewers. “Same race previously” indicates whether respondents were classified as white in the previous interview in the models predicting classification as white, and whether respondents were classified as Black in the previous interview in the models predicting classification as Black. The columns labeled “Women” and “Men” report odds ratios from separate models for women and men, and the column labeled “p value of Women-Men” reports the p value from a fully interacted model testing whether the effect of each variable was the same for men and women.

p < .05, ** p < .01, *** p < .001.

For both women and men, racial classification is influenced by a wide variety of status characteristics. Looking first at column 1, the odds ratio for having ever experienced long-term unemployment (.705) indicates that, net of all other measured factors, including whether she was classified as white in the previous year, having been unemployed reduces the odds of a woman being classified as white in the current year by roughly 30 percent. In fact, for women, all but one of our status variables is a significant predictor of being classified by the interviewer as white. The second model presents the analogous results for men, and 8 of the 11 status variables are statistically significant predictors of being classified as white among men. Results from the other pair of models can be interpreted similarly, with odds ratios greater than one indicating that a given factor increases the odds of being classified as Black, and odds ratios less than one indicating that factor decreases the odds of being classified as Black. Overall, 30 of the 44 possible associations between social position and racial classification are statistically significant and in the expected direction, demonstrating the inextricable link between race and class in the United States.

Focusing specifically on gender differences in the relationship between social position and racial classification, we examine p values comparing the odds ratios for women and men (see columns 3 and 6). A statistically significant p value (p < .05) for the gender interaction of a given characteristic indicates that there is a measurable difference in the effect of that characteristic on the racial classifications of women and men. We find six instances where the odds ratios are significantly different at the .05 level and four additional factors with gender differences that are approaching significance (p < .10). Poverty, for example, decreases both women’s and men’s odds of being classified as white, but the p value indicates that having ever been in poverty has a stronger effect (indicated by the odds ratio furthest from 1) on the racial classification of men compared to women. Finding that poverty has a stronger effect for men is consistent with traditional gender roles that emphasize men’s responsibility as breadwinners (Reskin and Padavic 1994). Likewise, consistent with gendered stereotypes around welfare (Timberlake and Estes 2007), we see that women who ever received welfare benefits have lower odds of being seen as white and higher odds of being seen as Black, whereas men’s racial classifications are not affected by reporting a history of welfare receipt (e.g., food stamps, SSI and other public assistance).

However, there are also characteristics that significantly predict racial classification for both women and men where the estimated effects do not differ significantly by gender. Nonsignificant p values in columns 3 and 6 suggest that some status cues affect racial classification equally for everyone. For example, both women and men are significantly more likely to be seen as white if they are married (with or without children), or live in the suburbs, and significantly less likely to be seen as white if they live in the inner city.

It is also worth noting that the factors that affect whether people are seen as white are not necessarily the same factors that affect whether they are seen as Black. For example, living in a suburb increases both women’s and men’s odds of being classified as white, but does not affect their odds of being classified as Black. We also find that the odds ratio for having been classified as Black in the previous year is significantly larger for men. This suggests that “Blackness” in general is stickier for men: net of the status variables examined here, a man classified as Black this year has higher odds of being classified as Black again next year than a woman who is classified as Black this year.

Perhaps the most puzzling findings are for education. The results suggest education has a significant effect on the odds of both women’s and men’s racial classifications, but it affects these perceptions in different ways. For men, more educational attainment significantly increases the odds of being classified as white, as one might expect given the controversial but oft-repeated association between academic success and “acting white” (see Tyson, Darity Jr, and Castellino 2005). However, for women, being more educated significantly decreases their odds of being classified as white, and increases their odds of being classified as Black. In interpreting these results, it is important to remember that the odds ratios represent the effects of education net of all other factors in the model. Models including only education—along with year-fixed effects, the previous year’s racial classification, and self-identification as Hispanic or multiracial—indicate that more education is positively associated with whiteness and negatively associated with Blackness for both women and men (results available on request).

The difference between the education only and full models suggests that the effect of education on women’s racial classification operates through other demographic characteristics or aspects of social status; once factors such as welfare, poverty, unmarried parenthood, and living in the South are taken into consideration, the positive association between education and whiteness becomes negative. In contrast, for men, education remains positively associated with whiteness even when other considerations are held constant. Our data are unable to speak to the specific mechanisms behind these differing results. However, finding that greater educational attainment is less incongruent with Blackness for women is in keeping with previous research that highlights gender differences in perceived intelligence (Steinbugler, Press, and Dias 2006).

Discussion

The results from our analysis of gender differences in the relationship between social position and racial classification can be summarized by three broad patterns, illustrated in Figure 1. First, as depicted in the pair of bars on the left, there are some status distinctions, such as living in the inner city, that are gender neutral; they affect racial classification similarly for women and men, both in terms of their magnitude and statistical significance. Second, some factors have statistically significant effects for both women and men, but the magnitude of the effect differs. For example, as depicted in the middle pair of bars, having ever lived in poverty reduces the odds of being classified as white for both women and men, but the effect is significantly stronger for men. Third, some factors differ in both magnitude and statistical significance, as seen in the pair of bars on the right, where a history of welfare receipt affects the odds of women being classified as white, but has no significant net effect for men.

Figure 1:

Gendered patterns in how social status shapes race

Overall, the results are consistent with a priori expectations of how common racial stereotypes are gendered. For status distinctions that tend to affect men and women equally, such as being married, we find no gender differences in their effects on racial classification. However, where status distinctions and their associated stereotypes are strongly gendered, as with welfare or incarceration, we find significant status effects on racial classification only for the expected group. Between these extremes are factors like poverty and parenthood, with stereotypes that can be applied to both men and women but are often deployed unequally (or asymmetrically), as when out-of-wedlock births are seen as a problem only for women of color (Furstenberg 2003).

We can also read the content of our racial stereotypes and their gendered dimensions backward, inferring them from the significant effects in Table 3. From this perspective, our results suggest that Blackness is associated with unemployment, poverty, unmarried parenthood, and inner city residence, for both men and women. Whiteness, on the other hand, is associated with living in the suburbs and being married (regardless of parental status). By focusing on the p value testing for significant differences in the effects of a given factor, one can also infer the degree to which racial stereotypes are gendered. The primary example here is welfare, which has significant gender differences in its effects on being classified as white and Black, with significant effects for women but not men.

Our results imply that the stereotypes shaping racial classifications vary by gender, and thus suggest that factors triggering racial discrimination are also likely to vary. This echoes social psychological research on racial prototypicality showing that the more characteristics someone has that are associated with Blackness the more likely they are to be discriminated against (Eberhardt et al. 2006). However, our work also supports previous research finding that the prototypical Black woman and Black man (or white woman and white man) are not identical (cf. Timberlake and Estes 2007). Taken together, our study calls attention to the fact that before racial discrimination can occur, potential discriminators must first racially classify their targets, and the factors that inform these classifications vary by gender. Revealing a history of welfare receipt may put a woman at greater risk for racial discrimination, because it increases the likelihood that she will be perceived as Black, but the same is not true for men. From a policy perspective, it is important to understand how the antecedents of discrimination might vary, especially if other axes of difference (e.g. gender, sexuality, class, disability) are subject to similar classification processes.⁹

Destabilizing Categories or Reinforcing Hierarchy?

The micro-level processes discussed above do not simply reflect macro-level structures of inequality, they actively “(re)constitute broader relations of inequality” (Pyke 1996, 546). As such, finding that changes in race are tied to social mobility in gendered ways has important implications for how we think about macrostructural aspects of intersectionality. In particular, our results highlight that changes in race, gender, or class status do not necessarily imply a significant weakening of boundaries or an undermining of categories, as has often been predicted in previous research (e.g., see Alba 2005; Wilkins 2004). Instead, our findings reveal a deeply intertwined system of inequality in which individual fluidity can perversely reinforce existing structural inequalities.

To illustrate, we take Collins’s (1990) metaphor of the “matrix of domination” literally, invoking a multidimensional array of cells that encode a hierarchy of social status (though using “geometrical” images also has its limitations; see West and Fenstermaker 1995). Typically, the boundaries separating categories, or cells, along the dimension of race (or gender) are seen as static and rigid, restricting mobility and causing individuals’ experiences to be shaped or constrained by their social location. So, when people experience class mobility, moving into the cells above or below their current structural position, it is assumed their trajectory is restricted to the same racialized column of cells. Our results reveal the potential for diagonal movement between cells, as social mobility can involve multiple dimensions of stratification simultaneously. For example, when a man’s income falls below the poverty line, our results show that this not only affects his class position but also lowers his odds of being classified as white and raises his odds of being classified as Black.

However, highlighting the permeability of racial boundaries is not necessarily the same as saying they become blurred or unstable. We argue instead that these unexpected patterns of mobility are serving, intentionally or not, to maintain perceived differences between groups. When confronted with counter-stereotypical status shifts, the larger system of inequality can be stabilized by redefining individuals—moving them between cells of the existing matrix—instead of redefining the categories or their associated stereotypes and lived experiences. Rather than revise tropes about Black women and welfare, the matrix is reinforced as the racial categorizations of women who are (and are not) on welfare get reinterpreted.

Indeed, we suggest the strength of these social boundaries is in their selectivity, not their impermeability per se. If the movement of individuals across cell (or category) boundaries is selective and based on status characteristics, the social meaning of the categories will remain in equilibrium, and their associated stereotypes will continue to be reinforced over time. To the extent that individual boundary crossing increases homogeneity within a given category or cell, this process can also increase social distance, as selective fluidity slowly redefines shared experiences. Thus, our findings suggest that the micro-level mobility we observe not only maintains the content of a particular social category, but also positions cells vis-à-vis other cells in the larger structure of inequality.¹⁰

By highlighting the relationship between selective racial fluidity and the maintenance of hierarchy, our study provides an important counterpoint to the claim, common in literary criticism and cultural studies, as well as among members of the multiracial movement, that the existence of multiplicity and flexibility has the potential to undermine categorical racial distinctions. In a similar vein, Wilkins asserts that revealing the racial performances of individuals “destabilizes racial categories, adding elasticity to their boundaries and molding their content” (Wilkins 2004, 119). As performances imply an audience, studies of racial classification and identity performance effectively examine opposite sides of the same coin. While we agree that acts of performing race, class, and gender offer the potential to destabilize categories through individual agency at any given point in time, our results suggest that over a longer period, identity performances (or at least the racial classifications that follow from them) are not transformative of the larger system. We echo Butler (1993) in noting that, theoretically, the fact that norms are reproduced through individual actions introduces the possibility of revolutionary change, but most of the time people give normatively “correct” performances, which help to maintain stability and affirm the existing hierarchy.

For example, the “wannabes” in Wilkins’s study perform Puerto Rican-ness (rather than whiteness) to gain entrée into hip-hop culture and access to relationships with Black and Puerto Rican men. Although their behavior both calls into question the putative biological and cultural foundations of “race” and exposes the norms and stereotypes that circumscribe authentic membership, it leaves the overall hierarchy that frowns upon interracial relationships intact. In the long run, to the extent that their performances are successful (i.e., they are accepted and categorized as Puerto Rican by others), any normative transgression in their presentations of self will fade away into the appearance of compliant racial homogamy. This micro-level fluidity helps to reinforce inequality by concealing challenges to the hegemonic order; or, put another way, our expectations literally shape what we see and do not see, rendering potential points of incongruity effectively invisible.

Our examination of how micro-level fluidity reinforces the macro-level structures of inequality also stands in contrast to standard research on inequality, which focuses on how being located in a particular gender and/or race category constrains upward mobility and other forms of status attainment. Placing racial (and gender) fluidity alongside class mobility both recasts previous research and raises new research questions. From this perspective, it becomes important not only to examine how interlocking axes of oppression shape the experiences and opportunities of individuals in a particular social location, but also to ask who moves between which categories, how mobility is experienced, and what are the barriers to (or the avenues that facilitate) this movement (cf. Bettie 2002). Learning that some social boundaries are easier to cross while others are less permeable, or that movement along one dimension of status is likely to result in movement on another, helps us to better understand how various cells (or different axes of stratification) are organized in relation to each other and illuminates otherwise invisible aspects of the structure of the matrix.

Conclusion

This study demonstrates how longitudinal survey data, with repeated measures of key axes of stratification, can help further efforts to understand the dynamics of intersectional inequalities in the United States. Choo and Ferree argue that, by examining changes over time, “macrostructures of inequality” can be linked to “microstructures of social construction of meaning,” revealing feedback loops among the various axes of stratification (Choo and Ferree 2010, 146). Our results illuminate one such loop by identifying gendered patterns in how class relevant factors shape racial classification in the United States. In some cases, the gendered effects match a priori expectations based on widespread stereotypes about criminal propensities and “welfare queens.” In others, they point to unexplored variation in how women and men experience both social and racial mobility. For example, previous research has asked what happens when upward mobility through educational attainment interacts with racial identities that are implicitly “classed” (Bettie 2002; Cole and Omari 2003), but little has been done empirically to examine whether this process differs by gender.

Of course, though our findings speak to the presence of gendered differences in how social and racial status are intertwined, they cannot speak to how mobility along these multiple dimensions is experienced. Thus, we hope our results will inspire additional qualitative work on the antecedents and consequences of crossing what are typically viewed as rigid category boundaries, and what this movement means in other aspects of people’s lives (see, e.g., Schilt 2006). Our study also highlights the need for further macro- and meso-level explorations of how race-making processes (Wacquant 2002) interact with gendered organizations and social systems (Acker 2006; Ridgeway 1997) to produce different sets of racialized expectations for men and women. To fully understand how race, class, and gender interact, in “a process by which each relationship of inequality works on and through the others” (Choo and Ferree 2010, 36), it is crucial for future research on these issues to be developed iteratively and from a range of methodological perspectives.

Footnotes

Authors’ Note:

The authors contributed equally to this paper. We are grateful for financial support from the University of California Center for New Racial Studies and the Clayman Institute for Gender Research, research assistance from Alison Perlberg, and to Catherine Bolzendahl, Shelley Correll, Catherine Harnois, Jocelyn Hollander, Cecilia Ridgeway, Laurel Westbrook, and the editors for their helpful comments.

Notes

Andrew Penner is an assistant professor of sociology at UC Irvine. His research focuses on gender and race inequality in the labor market and educational system. He is currently involved in projects examining gender inequality in transition economies, racial fluidity in the United States, and international gender differences in education.

Aliya Saperstein is an assistant professor of sociology at Stanford University. Her work explores the construction of social difference and how the operationalization of concepts like race and ethnicity affects studies of social inequality. Her current research examines trends in the measurement of sex and gender, and race and ethnicity, in national surveys.

References

Acker

Joan

. 2006. Inequality regimes: Gender, class, and race in organizations. Gender & Society 20:441-64.

Alba

Richard

. 2005. Bright vs. blurred boundaries: Second-generation assimilation and exclusion in France, Germany, and the United States. Ethnic and Racial Studies 28:20-49.

Bettie

Julie

. 2002. Exceptions to the rule: Upwardly mobile white and Mexican American high school girls. Gender & Society 16:403-22.

Blair

I. V.

Judd

C. M.

Fallman

J. L.

2004. The automaticity of race and Afrocentric facial features in social judgments. Journal of Personality and Social Psychology 87:763-78.

Brown

J. Scott

Hitlin

Steven

Elder

Glen H.

Jr.

2007. The importance of being other: A natural experiment about lived race over time. Social Science Research 36:159-74.

Browne

Irene

Misra

Joya

. 2003. The intersection of gender and race in labor markets. Annual Review of Sociology 29:487-513.

Butler

Judith

. 1993. Bodies that matter. New York, NY: Routledge.

Choo

Hae Yeon

Ferree

Myra Marx

. 2010. Practicing intersectionality in sociological research. Sociological Theory 28:129-49.

Cole

Elizabeth R.

Omari

Safiya R.

2003. Race, class and the dilemmas of upward mobility for African Americans. Journal of Social Issues 59:785-802.

10.

Collins

Patricia Hill

. 1990. Black feminist thought. London: HarperCollins.

11.

Correll

Shelley J.

Ridgeway

Cecilia L.

2003. Expectation states theory. In Handbook of social psychology, edited by Delamater

John

. New York, NY: Kluwer Academic/Plenum.

12.

Crenshaw

Kimberle

. 1989. Demarginalizing the intersection of race and sex. University of Chicago Legal Forum 1989:139-67.

13.

Davis

Kathy

. 2008. Intersectionality as buzzword: A sociology of science perspective on what makes a feminist theory successful. Feminist Theory 9:67-85.

14.

Devine

P. G.

Elliot

A. J.

1995. Are racial stereotypes really fading? The Princeton trilogy revisited. Personality and Social Psychology Bulletin 11:1139-50.

15.

Eberhardt

Jennifer L.

Davies

Paul G.

Purdie-Vaughns

Valerie J

Johnson

Sheri Lynn

. 2006. Looking deathworthy: Perceived stereotypicality of Black defendants predicts capital-sentencing outcomes. Psychological Science 17:383-86.

16.

Eberhardt

Jennifer L.

Goff

Phillip Atiba

Purdie

Valerie J.

Davies

Paul G.

2004. Seeing Black: Race, crime and visual processing. Journal of Personality and Social Psychology 87:876-93.

17.

Freeman

Jon B.

Ambady

Nalini

. 2011. A dynamic interactive theory of person construal. Psychological Review 118:247-79.

18.

Freeman

Jon

Penner

Andrew M.

Saperstein

Aliya

Scheutz

Matthias

Ambady

Nalini

. 2011. Looking the part: Social status cues shape racial perception. PLoS One 6:e25107.

19.

Furstenberg

Frank E.

2003. Teenage childbearing as a public issue and private concern. Annual Review of Sociology 29:23-39.

20.

Gilens

Martin

. 1999. Why Americans hate welfare. Chicago, IL: University of Chicago Press.

21.

Giles

Howard

Bourhis

Richard Y.

1976. Voice and racial categorization in Britain. Communication Monographs 43:108-14.

22.

Glauber

Rebecca

. 2008. Race and gender in families and at work: The fatherhood wage premium. Gender & Society 22:8-30.

23.

Glenn

Evelyn Nakano

. 2002. Unequal freedom. Cambridge, MA: Harvard University Press.

24.

Harnois

Catherine E.

2010. Race, gender and the Black women’s standpoint. Sociological Forum 25:68-85.

25.

Harris

Cherise A.

Khanna

Nikki

. 2010. Black is, Black ain’t: Biracials, middle-class Blacks, and the social construction of Blackness. Sociological Spectrum 30:639-70.

26.

Hill

Mark

. 2002. Race of the interviewer and perception of skin color. American Sociological Review 67:99-108.

27.

Hurwitz

Peffley

2005. Playing the race card in the post-Willie Horton era: The impact of racialized code words on support for punitive crime policy. Public Opinion Quarterly 69:99-112.

28.

Kennelly

Ivy

. 1999. That single-mother element: How white employers typify Black women. Gender & Society 13:168-92.

29.

Kirschenman

Neckerman

1991. “We’d love to hire them but…”: The meaning of race for employers. In The urban underclass, edited by Jencks

Peterson

P. E.

Washington, DC: Brookings Institution.

30.

McCall

Leslie

. 2005. The complexity of intersectionality. Signs 30:1771-1800.

31.

Omi

Michael

Winant

Howard

. 1994. Racial formation in the United States. New York, NY: Routledge.

32.

Prentice

Deborah A.

Carranza

Erica

. 2002. What women and men should be, shouldn’t be, are allowed to be, and don’t have to be: The contents of prescriptive gender stereotypes. Psychology of Women Quarterly 26:269-81.

33.

Purdie-Vaughns

Valerie

Eibach

Richard P.

2008. Intersectional invisibility: The ideological sources and social consequences of non-prototypicality. Sex Roles 59:377-91.

34.

Pyke

Karen D.

1996. Class-based masculinities: The interdependence of gender, class, and interpersonal power. Gender & Society 10:527-49.

35.

Reskin

Barbara F.

Padavic

Irene

. 1994. Women and men at work. Thousand Oaks, CA: Pine Forge Press.

36.

Richeson

Jennifer A.

Trawalter

Sophie

. 2005. On the categorization of admired and disliked exemplars of admired and disliked racial groups. Journal of Personality and Social Psychology 89:517-30.

37.

Ridgeway

Cecilia L.

1997. Interaction and the conservation of gender inequality: Considering employment. American Sociological Review 62:218-35.

38.

Ridgeway

Cecilia L.

Kricheli-Katz

Tamar

. 2013. Intersecting cultural beliefs in social relations: Gender, race, and class binds and freedoms. Gender & Society.

39.

Saperstein

Aliya

. 2008. (Re)Modeling race: Moving from intrinsic characteristic to multidimensional marker of status. In Racism in post-race America: New theories, new directions, edited by Gallagher

Charles

. Chapel Hill, NC: Social Forces.

40.

Saperstein

Aliya

Penner

Andrew M.

2012. Racial fluidity and inequality in the United States. American Journal of Sociology 118:676-727.

41.

Saperstein

Aliya

Penner

Andrew M.

2010. The race of a criminal record: How incarceration colors racial perception. Social Problems 57:92-113.

42.

Schilt

Kristen

. 2006. Just one of the guys? How transmen make gender visible at work. Gender & Society 20:465-90.

43.

Schippers

Mimi

. 2007. Recovering the feminine other: Masculinity, femininity, and gender hegemony. Theory and Society 36:85-102.

44.

Smith

Stewart

A. J.

1983. Approaches to studying racism and sexism in Black women’s lives. Journal of Social Issues 39:1-15.

45.

Steinbugler

Amy C.

Press

Julie E.

Dias

Janice Johnson

. 2006. Gender, race, and affirmative action: Operationalizing intersectionality in survey research. Gender & Society 20:805-25.

46.

Telles

Edward

Lim

Nelson

. 1998. Does it matter who answers the race question? Racial classification and income inequality in Brazil. Demography 35:465-74.

47.

Timberlake

Jeffrey M.

Estes

Sarah

. 2007. Do racial and ethnic stereotypes depend on the sex of target group members? Sociological Quarterly 48:399-433.

48.

Tyson

Karolyn

Darity

William

Jr. Castellino

Domini R.

2005. It’s not “a Black thing”: Understanding the burden of acting white and other dilemmas of high achievement. American Sociological Review 70:582-605.

49.

Wacquant

Loïc

. 2002. From slavery to mass incarceration: Rethinking the “race question” in the U.S. New Left Review 13:41-60.

50.

Waters

Mary

. 1999. Black identities. Cambridge, MA: Harvard University Press; New York: Russell Sage Foundation.

51.

Weitz

Gordon

1993. Images of Black women among Anglo college students. Sex Roles 28:19-34.

52.

West

Candace

Fenstermaker

Sarah

. 1995. Doing difference. Gender & Society 9:8-37.

53.

Wilkins

Amy C.

2004. Puerto Rican wannabes: Sexual spectacle and the marking of race, class, and gender boundaries. Gender & Society 18:103-21.

54.

Wingfield

Adia Harvey

. 2009. Racializing the glass escalator: Reconsidering men’s experiences with women’s work. Gender & Society 23:5-26.