Understanding White Identity Management in a Changing America

Download data and study materials from OSF

Principal investigator:

Eric Shuman

New York University

Email: eshuman1650@gmail.com

Homepage: https://wp.nyu.edu/knowleslab/people/


Sample size: 1631

Field period: 04/26/2023-09/22/2023

Abstract
The United States is experiencing significant social change, compelling White Americans to grapple more directly with their white identity. This paper examines how White Americans manage their identity amidst societal shifts using a new measure of advantaged identity management, representative data (N = 2648), and latent profile analysis. The findings reveal five subgroups of White Americans, each managing their identity differently. Four profiles correspond to the main advantaged identity management strategies (defend, deny, distance, dismantle), with a fifth using strategies flexibly. Of 15 predictions regarding how valuing hierarchy, meritocracy, and egalitarianism predict profile membership, 13 were supported. These profiles show contrasting attitudes toward social change, with defender-deniers opposing, denier-distancers moderately opposing, distancers remaining neutral, and dismantlers supporting change. These findings provide some of the first empirical evidence for a theorized model of white identity management and suggest that how White Americans manage their identity has important implications for social change.
Hypotheses

These analyses have three main goals: 1) to identify different classes of White Americans based on how they manage their white identity 2) to test the theoretical aspects of advantaged identity management developed in Knowles et al. (2014) and Shuman et al. (2022) using these classes, and 3) to test the effects of framing inequality either as advantage or disadvantage on support for equality-advancing policies aimed among each of the classes.

Research Goal 1 – Identify Classes
We will use tidyLPA with Mplus Integration (Rosenberg, et al., 2018) for R (R Core Team, 2021) to test LPA solutions for our five indicator variables (Defend, Deny, Distance from Inequality, Distance from Identity, Dismantle). We are expecting to find a similar 5-class solution to that which we found in our prior research (for more detail see previous pre-registration): Defender-Deniers, Denier-Distancers, Flexible-Ambivalents, Distancers, and Dismantlers. However, we are open to the fact that now that we are using a representative sample we may uncover new classes, which were not present in large enough numbers in our earlier samples to be detected and that the exact patterns of means within each class may change somewhat from those observed in our earlier research.

Research Goal 2 – Test Theory of Identity Management
Thus, we make the following predictions regarding how the values will be linked to membership in each of the classes (summarized in Table 2 in online preregistration see https://osf.io/xgh9c/?view_only=51d63155c8444762ae88032bea6ce3c6). Membership in the Defender-Denier class should be strongly predicted by SDO, strongly negatively predicted by universalism, and moderately related to the Preference for Merit Principle. Membership in the Denier-Distancer class should be significantly predicted by SDO and the Preference for Merit Principle, but negatively predicted by universalism. Membership in the Flexible/Ambivalent class should be weakly predicted by SDO and the Preference for Merit Principle but negatively related to universalism. Membership in the Distancer class should be predicted by the Preference for Merit Principle and universalism, and negatively predicted by SDO. Membership in the Dismantler class should be strongly predicted by universalism, and negatively predicted by the Preference for Merit Principle and strongly negatively predicted by SDO.

Research Goal 3 – Consequences of Class Membership
Finally, the effects of framing inequality either as advantage or disadvantage on support for policies aimed at advancing equality among each of the classes. Given that the strategies reflect the ways in which White Americans understand racial advantage, their choice of strategies should shape reactions to inequality frames that do (or do not) highlight that advantage (see Figure 3 for hypotheses). In this way, testing how inequality frames affect White Americans who choose different identity-management strategies also serves as a test of the predictive validity of this theoretical approach, in addition to potentially resolving a debate in the literature. Namely, we expect that Dismantlers, who are highly aware of their privilege and motivated by the moral-image threat it poses (Knowles et al., 2014), will be more supportive of redistributive policies when inequality is framed as ingroup advantage (Lowery et al., 2012), because this framing triggers their main motivation. On the other hand, Distancers, who recognize the existence of inequality but attempt to distance themselves from privilege, should react more positively to framings of inequality as disadvantages faced by Black Americans—as this does not directly connect them to the problem. Further, we predict that Denier-Distancers, who do not believe inequality is a major problem, will be less supportive of policies regardless of framing. Their denial is primarily driven by the psychological threats posed by recognizing inequality, and thus advantage framing which increases threat (Dover, 2022) is likely to only deepen their opposition. Defender-Deniers, who are committed to defending the racial hierarchy, will likely be strongly opposed to any policies advancing equality regardless of framing. Since they are committed to defending the hierarchy it is unlikely that any framing of inequality can motivate them to support equality. Finally, because Flexible-Ambivalents employ many strategies, it is difficult to predict how they will react to the two framings, but observing which framing is more effective will still be informative as it will give us insight in how to best increase support for equality among members of this group.

Experimental Manipulations
We examined the effects of framing inequality either as advantage or disadvantage on support for policies aimed at advancing equality among each of the profiles. Given that the strategies reflect the ways in which White Americans understand racial advantage, their choice of strategies should shape reactions to inequality frames that do (or do not) highlight that advantage. After completing all other measures except for the measure of support for policies that would reduce inequality, participants were randomly assigned to view one of two infographics. These infographics presented the same factual information, but in the advantage framing condition this information was presented as advantages held by White Americans (e.g. School districts serving predominately White neighborhoods receive roughly $23 billion more in funding than their Black counterparts), whereas in the disadvantage framing condition, the same information was presented as disadvantages faced by Black Americans (e.g. School districts serving predominately Black neighborhoods receive roughly $23 billion more in funding than their White counterparts). This manipulation was taken almost exactly from Dietze and Craig (2020), which found that framing economic inequality as disadvantages increased support for action. The formatting was exactly the same, and we presented information about inequalities in the three out five domains Dietze and Craig (2020) in manipulations (i.e. education, employment, and health). While Dietze and Craig (2020) also presented information about inequalities in food access and home ownership, we chose instead to present information about inequality in the criminal justice system as this is one of the most frequently discussed ongoing racial inequalities. The specific facts presented also differed from Dietze and Craig’s (2020) manipulations, as ours were about racial rather than economic inequality. Otherwise, there were no differences so this could be seen as a conceptual replication of Dietze and Craig (2020) in the domain of racial rather than economic inequality.
Outcomes

Support for Policies to Advance Equality

Rate the extent to which you agree with the following statements: 1 strongly disagree to 7 strongly agree
1. American society has a responsibility to compensate Black people with jobs and education through programs such as affirmative action.
2. The government should ensure that schools receive equal funding regardless of the racial make-up of the students.
3. The government should invest more money in hospitals and clinics that primarily serve Black communities.
4. There should be large scale criminal justice reform to address racial inequalities in the justice system.

Summary of Results

Analytic Strategy
Our primary analyses were conducted in three steps. First, using latent profile analysis (LPA), we determined the number and nature of the subgroups in each sample using the five identity-management strategies as profile indicators. We report the full details of these analyses including how we selected the final number of profiles in the supplemental materials for the interested reader, but in the main text focus on presenting the results of the final model. Second, using multinominal regression, we tested theoretically derived predictions about which values predict membership in the profiles. Third, we examine the social and political outcomes of membership in each of the profiles by regressing our outcome variables on the posterior probabilities of membership in each profile (while setting the model’s intercept term to 0). This approach yields predicted values of the outcomes when the probability of membership in a given subgroup is 1 ( Pastor et al., 2007).

Step 1: Latent Profile Analysis
As a form of mixture modeling, LPA is premised on the notion that variables’ observed distributions may reflect unobserved subgroups (“profiles”) of individuals (Oberski, 2016). However, unlike cluster analysis, LPA has the advantage of being model-based, with criteria for helping to select an optimal solution. Based on these criteria (AIC, BIC, VLMR tests, entropy, etc., see overview in Table 1) and theoretical interpretability, we selected the 5-profile solution as the best solution in both studies (for more details on model selection see supplemental materials). Additionally, a multigroup LPA indicated that these two solutions did not significantly vary between the two studies further supporting their validity (see supplemental materials). Below we describe each of the five profiles both in terms of their scores on the indicators used in the LPA, i.e. the different identity management strategies (for an overview see Figure 2), and in terms of demographics . We examined demographic characteristics for each profile by regressing the posterior probabilities onto all the demographics. We give a brief summary of selected demographics here (based on Study 2 data as this was the representative sample), and present full demographic information on each profile for both studies in the supplementary materials.

The Profiles

Defender-Deniers. Defender-Deniers made up 8% of the Study 1 sample and 10% of the Study 2 sample. These White Americans are characterized by high levels of the defend and deny strategies—and very low levels of the dismantle strategy. In fact, all members of this profile are at or above the midpoint on the defend and deny and at or below the midpoint on dismantle. While Defender-Deniers tended to score above the midpoint on the two forms of distancing, we see high dispersion around these means and notice that the subgroup’s distancing scores are similar to those in other profiles—thus indicating that distancing is not a distinguishing feature of this profile. Members of this subgroup of White Americans believe that any inequalities between White and Black Americans are legitimate, while also minimizing the extent of those inequalities. A person most characteristic of the Defender-Denier profile was most likely to be male (67%), aged 54, very conservative (4.01 on a 1-5 scale), and a Republican (76%). In terms of religious identity, they were most likely to identify as Protestant/Just Christian (51%) or Catholic (18%). They were not likely to have earned a college degree (25%) and were estimated to earn around $53,000 per year.

Denier-Distancers. Denier-Distancers comprised 21% of the Study 1 sample and 19% of the Study 2 sample. These White Americans are characterized by high levels of the deny and distancing from identity strategies: All members of this profile are at or above the midpoint on deny and tend to score very high on distancing from identity, although a few cases dip below the midpoint. Although Denier-Distancers are almost as low on dismantling as are Defender-Deniers, Denier-Distancers tend not to defend racial inequality. This subgroup of White Americans asserts that racial inequality is minimal or nonexistent and prefers to avoid thinking about themselves in racial terms. In terms of demographics, a person most characteristic of the Denier-Distancer profile was most likely to be male (54%), aged 54, conservative (3.88 on a 1-5 scale), and a Republican (76%). In terms of religious identity, they were most likely to identify as Protestant/Just Christian (67%) but had a higher likelihood of being “None/Other” (20%) than the Defender-Deniers. They were not likely to have earned a college degree (37%) and were estimated to earn around $57,000 per year.

Distancers. Distancers made up 25% of the Study 1 sample and 26% of the Study 2 sample. These White Americans are characterized by high levels of the two distancing strategies. Indeed, these White people distance from identity and inequality more than all other strategies. Distancers score low on defend and deny but display middling levels of and high variance on the dismantling strategy. These White Americans acknowledge that there is racial inequality and that it is illegitimate—but they do not regard it as highly self-relevant. Demographically speaking, A person most characteristic of the Distancer profile was most likely to be female (52%) aged 52, moderately liberal (2.78 on a 1-5 scale), and a Democrat (63%). In terms of religious identity, this was one of the most religiously diverse groups, with a moderate likelihood of being Protestant/Just Christian (38%), and a relatively equal likelihood of being Atheist/Agnostic (18%) or Catholic (17%) or None/Other (19%). They were relatively equally likely to have earned a college degree (47%) versus not (53%) and were estimated to earn around $54,000 per year.

Dismantlers. Dismantlers comprise 28% of Study 1 sample and 20% of Study 2 sample. These White Americans strongly endorse dismantling and display very low levels of the defend and deny strategies. Dismantlers also distance from identity to a moderate extent; however, they do so less than all other profiles and display high variance on this indicator, suggesting that distancing from identity is not a defining feature of this subgroup. In general, these White Americans acknowledge that racial inequality exists, is illegitimate, and should be confronted—while also regarding it as self-relevant. A person most characteristic of the Dismantler profile was mostly to be female (58%), aged 47, liberal (1.75 on a 1-5 scale), and a Democrat (96%). In terms of religious identity, this was one of the least religious groups, with moderate likelihood of being Atheist/Agnostic (41%) and an equal likelihood of being None/Other (21%) or Protestant/Just Christian (21%). They were very likely to have earned a college degree (70%) and were estimated to earn around $73,000 per year.

Flexible-Ambivalents. This subgroup makes up 17% of the Study 1 sample and 24% of Study 2 sample. These White Americans appear ambivalent in their stances regarding inequality and endorse most of the identity-management strategies to some extent. That said, Flexible-Ambivalents deploy the distancing significantly more than the other strategies, which may comport with their general ambivalence. They dismantle significantly less than all other strategies, suggesting that they lack a strong commitment to equality. We suspect that these White Americans lack strong ideological commitments and thus may deploy any strategy that is most effective in a given situation. A person most characteristic of this profile was most likely to be to be male (54%), aged 51, moderately conservative (3.40 on a 1-5 scale) and a Republican (54%). This group was highly likely to be Christian: Protestant/Just Christian (49%) or Catholic (24%). They were very unlikely to have earned a college degree (30%) and were estimated to earn around $48,000 per year.

Step 2: Testing Links Between Subgroup Membership and Underlying Values
Having uncovered a consistent set of five White identity management profiles in two independent samples, we proceeded to test our predictions concerning links between subgroup membership and White Americans’ underlying social values (see Knowles, et al, 2014; Shuman et al., 2024). Specifically, we examined how hierarchy, meritocracy, and egalitarianism values relate to the probability of membership in each of the five subgroups. In three multinomial logistic regression analyses, we regressed profile membership (coded as a nominal variable) on a value while weighting the observations by the corresponding profile frequency—thus accounting for the subgroups’ differing base rates. We then calculated the predicted change in profile membership probability when moving from low levels (i.e., the 16th percentile) to high levels (i.e., the 84th percentile) of the value, testing the significance of this change relative to 0.

For a summary of our predictions versus the observed effects in Studies 1 and 2, see Figure 4. We did not begin this project with predictions concerning the Flexible-Ambivalent profile. Nonetheless, in keeping with this subgroup’s roughly equal endorsement of every strategy, we developed the expectation that membership in the Flexible-Ambivalent profile might vary somewhat as a function of all three values. Overall, 13 of our 15 effects were in the predicted direction and 11 were of the predicted size—providing generally strong support for the 4-D model of advantaged identity management. We next describe in detail the results of our multinomial logistic models.

Valuing hierarchy was strongly linked to profile membership in both Study 1 (2(3) = 1342.21, p < .001, McFadden’s Pseudo-r2 = .20) and Study 2 (2(4) = 2948.2, p < .001, McFadden’s Pseudo-r2 = .11). Participants high (vs. low) in SDO were significantly more likely to be Defender-Deniers and significantly less likely to be Dismantlers. Membership in the Distancer was most likely when SDO was relatively low, and membership in the Denier-Distancer and Flexible-Ambivalent profiles was most likely when SDO was moderate (see Figure 5). Meritocratic values were strongly linked to profile membership in Study 1 (2(4) = 2479.3, p < .001, McFadden’s Pseudo-r2 = .22) and Study 2 (2(4) = 4267.5, p < .001, McFadden’s Pseudo-r2 = .16). Participants high (vs. low) in support for meritocracy were significantly more likely to be Defender-Deniers and Denier-Distancers, and significantly less likely to be Dismantlers. Membership in the Distancer and Flexible-Ambivalent profiles was most likely when support for meritocracy was moderate to high (see Figure 6).Egalitarian values were strongly linked to profile membership in Study 1 2(4) = 1671.8, p < .001, McFadden’s Pseudo-r2 = .15) and Study 2 (2(4) = 1967.1, p < .001, McFadden’s Pseudo-r2 = .08). Participants high (vs. low) in universalism were significantly more likely to be Dismantlers and significantly less likely to be Defender-Deniers. Membership in the Flexible-Ambivalent profile was most likely when universalism was moderate. Membership in the Distancer profile also tended to increase with universalism, though not to the same extent as did membership in the Dismantler profile.

Step 3: Outcomes of Profile Membership
In Study 2, we also explored whether a type of framing manipulation known to increase support for action to combat economic inequality (Dietze & Craig, 2020) might also increase support for action to reduce racial equality—at least among some groups of White Americans. However, we observed no overall effect of the framing manipulation, nor any effect within any specific identity-management profile, indicating that framing manipulations of this sort may not generalize to racial issues. Despite this, we were able to use the outcome measure of support for inequality-advancing policies to again test our hypotheses regarding different profiles’ stances toward social change. The results generally supported our hypotheses, with Defender-Deniers displaying the strongest opposition to these policies, followed by Denier-Distancers, Flexible-Ambivalents, Distancers, and finally Dismantlers (who were highly supportive of such policies; see Figure 9 and detailed analyses in supplemental materials).