A Curation Approach to Identity Management: The Costs of Combining Identity Expression and Suppression

Identity Expression and Suppression

Identity expression is defined as intentionally integrating one’s identity into the workplace by openly embracing, revealing, displaying, or bringing attention to it (Cha et al., 2019; Madera et al., 2012; Roberts & Nkomo, 2025). Identity management research has illuminated a range of expression behaviors that an employee can engage in at work. These can include positive distinctiveness behaviors that both portray one’s identity positively and inform others, such as intentionally challenging stereotypes (Cha & Roberts, 2019), incorporating desirable identity-based attributes into the conversation (Bell & Nkomo, 2001; Roberts, 2005), and educating others about one’s group (Creed & Scully, 2000). Expression can also include disclosures about one’s identity, such as sharing activities one engages in or experiences with discrimination (Arnett, 2023; Phillips et al., 2009). Finally, through their attire, music, or food, individuals can intentionally shape which features related to their identity others observe (Madera et al., 2012).

Identity suppression is defined as intentionally downplaying, concealing, or minimizing one’s identity at work (Roberts & Nkomo, 2025). More than simply a lack of expression, suppression involves going out of one’s way to hide aspects of an identity. Suppression can emerge through various behaviors, such as altering one’s name to resemble the majority culture (Kang et al., 2016), code-switching to mirror workplace norms by changing one’s accent, speech, or hair (McCluney et al., 2021), avoiding conversations about threatening identity-relevant topics (McCluney et al., 2017), dodging identity-based questions (Button, 2001), and pretending to adhere to values that are inconsistent with one’s identity (Hewlin, 2003). Given that race and ethnicity are typically visible identities, suppression in the form of fully passing as a member of the dominant group is less common (Clair et al., 2005; Ragins, 2008; cf. Leary, 1999). However, employees of color can nonetheless engage in social recategorization, wherein they not only downplay their membership in a devalued minority group but also emphasize their affiliation with a more highly regarded group, such as a professional identity (Roberts, 2005).

Consequences of Expression and Suppression

When it comes to understanding the consequences of expression and suppression, research typically focuses on one of two interrelated directions. The first direction focuses on external outcomes, often examining how specific behaviors and tactics influence one’s image, relationships, and career success. This research has historically positioned marginalized identities as liabilities at work because drawing attention to such identities bears the risk of highlighting stereotypes and eliciting negative evaluations from others, making suppression a prudent choice (Kang et al., 2016; McCluney et al., 2021). Recently, however, scholars have shown increased interest in the ways in which marginalized identities such as race and ethnicity are not only liabilities but also resources, meaning sources of value (Cha & Roberts, 2019; Creary et al., 2015). Scholars taking this more optimistic view have highlighted ways in which expression can be professionally beneficial. For example, drawing on meaningful details about a marginalized identity can facilitate deeper connections with coworkers, disrupt negative assumptions about one’s group, or enable more insightful work deliverables (Arnett, 2023; Cha & Roberts, 2019; Roberts, 2005; Thomas, 1993). Yet, these same scholars acknowledge that risks remain, such as when an attempt to draw on one’s identity in meaningful ways inadvertently triggers anxiety in others, alienates valued colleagues, or narrows career opportunities. Thus, marginalized identities can be both resources and liabilities, and expressing these identities can improve or undermine one’s image, relationships, and success (Cha & Roberts, 2019).

The second research direction focuses on internal outcomes, such as psychological experiences. This research typically moves beyond specific behaviors to instead consider the cumulative frequency with which a person expresses or suppresses in their workplace (i.e., whether they do so rarely or a great deal), which can be shaped both by dispositional (e.g., self-monitoring) and contextual factors relevant to one’s workplace (e.g., diversity climate, industry norms; Clair et al., 2005; Roberts, 2005). This body of work examines the frequency with which a person engages in either expression or suppression and how this relates to person-level psychological outcomes such as employees’ overall well-being and organizational commitment. The consensus in this line of work is that expression is beneficial for well-being, which, in turn, fuels organizational commitment and retention, whereas suppression undermines these outcomes (Day & Schoenrade, 1997; Griffith & Hebl, 2002; Jones & King, 2014; Madera et al., 2012; Martinez et al., 2017; Ragins, 2008; Ragins et al., 2007; Roberts et al., 2008). For example, the more open that LGBTQ employees were about their sexual orientation (a continuum ranging from full suppression to open expression), the lower their experience of conflict and the higher their job satisfaction and commitment were (Day & Schoenrade, 1997). Among Black doctors, engaging in frequent positive distinctiveness (a form of expression) reduced depression and increased career commitment due to the behavior’s potential for “enhancing one’s sense of authentic self-expression and providing a sense of psychological empowerment” (Roberts et al., 2008, p. 285). The opposite was predicted for social recategorization (a form of suppression), as the authors theorized that “the need to suppress one’s social identity in the professional context may engender feelings of identity conflict and poor fit into the profession” (Roberts et al., 2008, p. 285).

Taken together, these two directions of literature highlight the double-edged nature of both expression and suppression: Suppression can be professionally prudent but is psychologically taxing, whereas expression is psychologically uplifting but professionally risky. In light of these trade-offs, scholars have astutely pointed out that some employees may opt to engage in both expression and suppression at work. Although this possibility has received only limited attention in the literature, to the extent that scholars have discussed the consequences of combining expression and suppression, they have focused on the potential external consequences (Clair et al., 2005; Roberts et al., 2008; Smith et al., 2019). For instance, although Roberts and colleagues’ (2008) theorizing primarily focused on internal consequences of identity management, they included a valuable point about how some individuals may both express and suppress, and highlighted how this combination may be advantageous professionally and interpersonally.

Building on this idea, we note that a person concerned about the potential interpersonal risks of expression may choose to occasionally suppress, perhaps in the hope that they can achieve better career prospects while still attaining the psychological benefits of expression. In a similar vein, a person who is psychologically taxed from suppression may decide to occasionally express, hoping that doing so can relieve their distress without jeopardizing their career. Thus, people who both express and suppress may be seeking an optimal solution wherein they can achieve both professional success and psychological well-being. In light of these possibilities, it is important to gain a holistic understanding of the implications of engaging in both expression and suppression. Building on the foundation laid by researchers highlighting the potential interpersonal and professional benefits of this combination, we seek to address an open question regarding its intrapersonal ramifications.

The Curation Approach Heightens Ambivalence

Ambivalence is “simultaneously positive and negative orientations toward an object” (Ashforth et al., 2014, p. 1454). More than just the mere existence of opposing orientations (often referred to as “objective ambivalence”), ambivalence is most potent when it involves an experience of being pulled between opposing sides (also referred to as “subjective ambivalence”; Priester & Petty, 1996, p. 432), such as being torn between the advantages and disadvantages of a new policy (van Harreveld et al., 2009a) or being conflicted between the desire to appear smart vs. to be authentic (Emmons & King, 1988). Multifaceted objects that have both positive and negative features are a common source of ambivalence (Ashforth et al., 2014; Smith & Berg, 1987), and minority identities are ripe for such tensions given that their multifaceted nature can make them a resource or a liability (Cha & Roberts, 2019). We therefore focus on ambivalence that captures the extent to which employees are subjectively torn between positive thoughts regarding their identity (as a resource) and negative thoughts regarding their identity (as a liability). Although ambivalence can include both cognitive components (e.g., torn between conflicting thoughts) and emotional components (e.g., torn between conflicting emotions) (Rothman et al., 2017), we focus on the cognitive aspect of ambivalence given our focus on conflicting views of one’s identity as a resource versus liability. Notably, one need not personally endorse the liability side to experience ambivalence; to experience ambivalence, it is sufficient to perceive the identity as a resource yet also perceive that others hold the identity in low regard or associate it with stereotypes (Pratt, 2000; Priester & Petty, 2001).

The ambivalence literature reveals that when grappling with a multifaceted object like a minority identity, individuals recognize the potential for ambivalence and thus engage in different strategies to manage and potentially attenuate subsequent experiences of ambivalence (Ashforth et al., 2014). These strategies focus, to differing degrees, on the opposing orientations underlying the multifaceted object, in this case on perceptions of the identity as either a resource or a liability; consequently, the strategies vary in how effectively they reduce ambivalence (Ashforth et al., 2014). Drawing on this foundation, we consider how different identity management approaches map onto specific strategies from the ambivalence literature and the potential consequences for downstream experiences of ambivalence.

We contend that a curation approach is associated with the most potent experience of ambivalence. Individuals invoking a curation approach of high levels of expression and suppression use what scholars refer to as a compromise response for managing ambivalence, which occurs when a person enacts behaviors consistent with both sides of opposing orientations (Ashforth et al., 2014). Engaging in both expression and suppression keeps the opposing orientations of one’s identity top of mind, thereby making these orientations chronically accessible. When opposing sides of multifaceted objects, like minority identities, remain salient, a person is most likely to experience ambivalence in the form of conflict between contradictory facets of the object (Newby-Clark et al., 2002). For instance, employees of color who express to leverage their identity as a resource may experience tension because of increased awareness of how they have made themselves vulnerable to liabilities (Cha & Roberts, 2019). We argue that curators are especially likely to perceive expression through this ambivalent lens because liabilities are already salient to them due to curators’ frequent suppression. Similarly, individuals who suppress to protect themselves from liabilities can experience conflict over concealing a valued aspect of the self (Roberts, 2005; Roberts et al., 2008). We propose that curators are especially likely to experience suppression as an acute source of ambivalence because the positive side of their identity—the valued part of themselves they feel they must hide—is top of mind for these individuals given that they often express.

In contrast, people who either primarily express or primarily suppress minimize ambivalence by invoking a domination response that prioritizes one orientation and discounts the other (Ashforth et al., 2014). Indeed, a fundamental tenet of navigating conflicting cognitions is that focusing on one orientation reduces experiences of tension (Festinger, 1957; Nordgren et al., 2006). Employees who primarily express (pairing high expression with low suppression) use a domination strategy by focusing on their identity as a valued resource and downplaying its liabilities (Crocker & Major, 1989; Tajfel & Turner, 1986). These individuals may still perceive risks or liabilities, including in the wake of expression; however, they overshadow or discredit liabilities in a way that counteracts ambivalence (Ashforth et al., 2014). For instance, Mei, a Chinese American woman, could challenge a negative stereotype about Asian people, encounter a negative response from a coworker, and still keep ambivalence in check by discrediting the coworker and reminding herself of the value she gets from staying true to herself and speaking up on behalf of her group (Ellemers et al., 2002; Roberts et al., 2008). Thus, individuals who primarily express are able to mitigate ambivalence, allowing perceived resources to eclipse liabilities. Similarly, employees who primarily suppress (combining high suppression with low expression) also invoke a domination response focused on a single orientation, liabilities rather than resources, which protects them from bearing the full brunt of ambivalence (Ashforth et al., 2014). Individuals following this approach avoid liabilities at all costs and downplay their identity’s relevance as a resource. Although suppression can be seen as sacrificing a valued aspect of their identity (Roberts et al., 2008), individuals following this identity management approach ease ambivalence by rationalizing that it is sensible to blend in to protect their career or that they are deferring bringing attention to their identity until they make it to the top (Meyerson & Scully, 1995; Thomas, 1993).

Finally, people who pair low suppression with low expression invoke what the ambivalence literature refers to as an avoidance response of minimizing both the positive and negative sides of a multifaceted object (Ashforth et al., 2014). They ignore concerns that their identity could be a liability (and thus they may engage in low suppression by answering honestly if asked a question about their identity), and they do not engage deeply with thoughts about their identity as a resource (hence why they do not express). The tendency to chronically disengage from both positive and negative thoughts associated with one’s identity facilitates relatively lower ambivalence (Ashforth et al., 2014). The culmination of these patterns is that we expect ambivalence to be strongest when expression is combined with suppression, consistent with a curation approach. We therefore predict an interaction effect:

Ambivalence Leads to Psychological Strain

Researchers prioritize understanding the consequences of identity management for well-being (Clair et al., 2005; Day & Schoenrade, 1997; Griffith & Hebl, 2002; Hewlin, 2009; Mohr et al., 2019), such as anxiety (subjective feelings of apprehension and nervousness, often caused by uncertainty; Raghunathan & Pham, 1999; Spielberger, 1985), stress (a perceived inability to cope with environmental stressors; Hobfoll, 1989), and exhaustion (a depletion of emotional energy; Moore, 2000). Although distinct, these outcomes are components of a broader construct of psychological strain: adverse emotional and emotion-related responses to stressors (Beehr et al., 2000; Chang et al., 2007; Côté, 2005; Karasek, 1979).

Ambivalence is important to evaluate in the context of psychological strain because it is aversive; that is, the experience of having opposing orientations toward an object is discomforting (Ashforth et al., 2014). Having ambivalent attitudes about an object, such as genetically modified food, can trigger negative feelings, including anxiety (Nordgren et al., 2006), as can experiencing conflict about the advantages and disadvantages of personal goals, such as trying “to be all things to all people” (Emmons & King, 1988, p. 1042). This occurs because ambivalence violates expectations of self-consistency (Festinger, 1957). While awareness of inconsistency and conflicting thoughts can sometimes be reconciled (e.g., having to both care for animals and euthanize animals is reconciled by viewing euthanasia as a form of care) in ways that can mitigate the negative ramifications of ambivalence (Ashforth et al., 2024), in many situations no clear resolution exists. Under these conditions, ambivalence is most likely to trigger psychological strain (Ashforth et al., 2014; Emmons & King, 1998). For a curation approach, conflicting thoughts related to one’s race or ethnicity violate a general desire to see oneself as consistent, and such thoughts are especially likely to disrupt well-being because they stem from something core to the self: one’s identity. Moreover, opposing perceptions of one’s identity as a resource and liability do not lend themselves to easy reconciliation. People can struggle to see liabilities as helpful or as synergistic with a view of their identity as a resource, which causes these contradictory thoughts to intensify psychological strain.

The connection between ambivalence and psychological strain is particularly pronounced when one takes action, because doing so increases uncertainty about the ramifications of one’s choice (Rothman et al., 2017). For example, having ambivalent attitudes toward a policy and then writing an endorsement or rejection of it led to anxiety and regret, specifically due to individuals’ concerns over the outcomes of their actions (van Harreveld et al., 2009a). Because the potential for losses looms larger than the potential for gains, ambivalence may make individuals especially likely to fixate on the undesirable outcomes of their choices, engage in counterfactual thinking, and thus experience negative emotions (van Harreveld et al., 2009b). Ambivalence can also lead to distress about whether actions fall short of one’s goals. For instance, seeking support from an ambivalent friend can cause the seeker to “question the intent, sincerity, or possible subsequent consequences of support,” which in turn triggers both anxiety and physiological signs of stress about the inability to attain support (Holt-Lunstad et al., 2007, p. 279). For those using a curation approach, expressing or suppressing involves frequently taking action in a way that not only highlights the conflict between the resources and liabilities of their identity but also leads to distress about the consequences of their decisions and whether they are getting the support needed for career advancement. As ambivalence makes identity-related trade-offs readily accessible, Daniel, the aforementioned Black physician, may feel uncertain and anxious about whether broaching the topic of race with a colleague helps or hurts his image as an expert as well as his relationships (Cha & Roberts, 2019; Dumas et al., 2008). He may also feel stressed about whether suppressing beloved activities relating to his race unnecessarily estranges him from a valued identity (Bell, 1990). The culmination of experiences like this may contribute to an exacerbated sense of psychological strain. We therefore predict the following:

Psychological Strain Increases Turnover Intentions

Scholars have long recognized that employees who are unhappy at work cope with their dissatisfaction by looking for better circumstances (Maslach & Jackson, 1981). Psychological strain has the ability to trigger psychosomatic issues, drain psychological resources, and undermine one’s sense of accomplishment and self-esteem at work (Cheng & McCarthy, 2018; Lee & Ashforth, 1993; Melamed et al., 1999). Because of these negative ripple effects, psychological strain is considered a fundamental influence on behavior and the extent to which people determine the need for change (Chang et al., 2007). Indeed, psychological strain serves as a cue that a new organizational environment may be needed, and therefore leads to withdrawal and turnover intentions in one’s current organizational environment (Cheng & McCarthy, 2018; Lee & Ashforth, 1993). Identity management approaches can be guided by individual preferences but are heavily influenced by context (Goffman, 1963; Roberts, 2005). Employees who experience psychological strain may conclude that a new organizational environment can afford them the opportunity to enact a new, less psychologically taxing identity management approach. We thus expect that for people who draw on a curation approach involving high levels of both expression and suppression, psychological strain will translate into greater turnover intentions.

We test our theoretical framework across two studies and a supplemental study focused on employees of color. Study 1 is a preregistered organizational study testing the effects of expression and suppression on psychological strain (H2) and turnover intentions (H3) in a small sample of employees of color. Study 1 also includes exploratory tests of our hypotheses with a broader sample that includes women. Study 2 is a preregistered longitudinal survey in which we test our full model (H1–H3), including mediation via ambivalence. The supplemental study tests H2–H3 and is reported in Appendix E of the supplementary online materials (SOM) because it was not preregistered and was a single-time-point study with online participants. We conclude with an internal meta-analysis (across Study 1, Study 2, and the supplemental study) and p-curve analysis to summarize our key findings. Our preregistrations, supplementary online materials (SOM), and data/analysis files are available on ResearchBox (https://researchbox.org/6050). ¹ Following Willroth and Atherton (2024), we provide a full table of deviations from pre-registration in SOM Appendix A.

Sample and Procedure

We collected data for Study 1 in April 2022. Participants were employees in a public-sector organization. Two C-suite officers sent an email inviting all employees to complete a survey about workplace climate. We did not have a predetermined sample size and aimed to recruit as many employees as possible. The participants were asked to focus on one of their non-work identities and to report their selection in a text box prior to responding to the key measures. Of the 558 employees who completed the survey, we included racial or ethnic minority participants, who completed the expression and suppression scales about a minority racial or ethnic identity or a closely related identity (e.g., nationality, social class, religion, each of which is a component of cultural identity; Ely & Thomas, 2001). This yielded a sample of 41 participants (51.2 percent men, 46.3 percent women, 2.4 percent non-binary; modal age: 40–49; 75.6 percent Hispanic/Latinx, 14.6 percent Black, 12.2 percent Asian American, 7.3 percent Native American, 2.4 percent Middle Eastern, 29.3 percent White given some participants identified as multiracial or as White Hispanic/Latinx). On average, the participants had worked at the organization for 8.88 years (SD = 7.46). The participants completed the study measures below and measures unrelated to the present research.

Measures

Expression and suppression

Although there are existing scales that measure or relate to expression and suppression (Madera et al., 2012; Roberts et al., 2008, 2014; Wessel et al., 2020), we sought to create new scales, for two key reasons. First, some existing measures focus on specific instantiations of these constructs (e.g., Roberts et al., 2008 measured positive distinctiveness and social recategorization), rather than measuring expression and suppression broadly. Second, even broader scales (e.g., Madera et al., 2012) do not address employees discussing or avoiding certain risky, identity-related topics likely to be relevant to marginalized groups broadly speaking (including people of color), such as societal issues (e.g., mega-threats and social movements; DeCelles et al., 2020; Leigh & Melwani, 2019) or identity-related experiences at work (e.g., micro-aggressions; King et al., 2023; Sue et al., 2007). We therefore developed new measures to capture a broader range of identity management behaviors. SOM Appendix B describes in detail the stages of our scale development process, following recommendations from Hinkin (1995, 1998). This process included deductive and inductive item development; content validation; confirmatory factor analyses (CFA), which showed that a two-factor model fit better than a one-factor model; and validation, which demonstrated that our scale was related to but distinct from existing measures. See Table 1 for the final items.

OPEN IN VIEWER

Table 1.

Expression and Suppression Scale Items and Confirmatory Factor Analysis Factor Loadings*

Item	Factor
	1	2
Scale 1: Expression
1. At work, I discuss things related to my identity with others.	.82
2. At work, I discuss issues or events in society that relate to this identity.	.81
3. At work, I educate others about my identity and background.	.80
4. I open up to my coworkers about how my identity shapes my experiences within our organization.	.80
5. At work, I share details about activities I engage in relating to this identity.	.78
6. At work, I open up about how my experiences in life are shaped by my identity.	.77
7. I share details about meaningful dates, holidays, and traditions relating to this identity with coworkers.	.74
8. I express aspects of my identity at work.	.70
9. I display signs of this identity in my workspace (e.g., pictures, objects).	.56
10. I wear clothes, emblems, or accessories that reflect my identity at work.	.50
Scale 2: Suppression
1. At work, I hide details about activities I engage in relating to this identity.		.84
2. At work, I hide thoughts and feelings about how my identity shapes the way I am treated in my organization.		.83
3. I try to hide certain aspects of my identity when interacting with coworkers.		.83
4. I suppress aspects of my identity at work.		.81
5. I try to suppress behaviors or mannerisms that are associated with this identity at work.		.80
6. At work, I intentionally withhold thoughts and feelings about how my identity influences my day-to-day experiences in life.		.77
7. During conversations with coworkers, I purposely steer the conversation away from certain topics that could bring attention to my identity.		.76
8. At work, I hide aspects of myself that could bring attention to stereotypes relating to my identity.		.75
9. I conceal or camouflage signs of this identity in my workspace (e.g., pictures, objects).		.70
10. There are certain things that I do as a member of my identity group that I purposely hide from my coworkers (e.g., foods I eat, music I listen to).		.67

*

Factor loadings come from the confirmatory factor analysis conducted as part of the scale development process, described in the Supplementary Online Materials, Appendix B. Underlined items are the eight-item, shortened version used in Study 1.

Given organizational constraints, participants completed a shortened, four-item version of expression and suppression (see the underlined items in Table 1) on a scale of 1 to 5 (“not at all” to “very much”). For each measure, we selected one broad item (“I express/suppress aspects of my identity at work”) and then the three highest-loading items in our CFA, with one exception: We chose item 4 over item 2 for expression due to its greater organizational relevance. Participants chose the identity for which they would complete the actions described in each item and indicated this choice in a free-response question. Specifically, they were told that they would answer questions about navigating a non-work identity at work, were given example identities (e.g., race, ethnicity, nationality, religion, social class, gender, etc.), and were asked to “please write one non-work identity that you would like to focus on for the questions on the following page.” The next page showed the expression and suppression items, and the identity participants had written earlier was inserted as piped text in the instructions (i.e., “Below are statements about the identity you listed on the prior page: [piped text]. For the purposes of the statements below, the word ‘identity’ refers to this identity ([piped text]).”

Preregistered Tests of Hypotheses

Table 2 shows descriptives and correlations of our key measures for this sample. For our analyses, expression and suppression were mean-centered. To conduct simple moderation analyses, we used Hayes’s (2022) Process macro (Model 1), with percentile bootstrapping (10,000 samples, 95 percent confidence intervals). Following Yzerbyt and colleagues (2018), we examined moderated mediation by first applying causal steps (MacKinnon et al., 2000); then we tested the index of moderated mediation by using Hayes’s (2022) Process macro (Model 8) with percentile bootstrapping (10,000 samples, 95 percent confidence intervals). In the analyses, we used the 16th, 50th, and 84th percentiles as the low, average, and high levels of the moderator, respectively.

OPEN IN VIEWER

Table 2.

Means, Standard Deviations, and Correlations in Study 1 (Organizational Study)*

Variable	M(SD)	1	2	3	4
1. Expression	2.51(0.97)	.90
2. Suppression	2.15(0.90)	–.25	.90
3. Psychological strain (emotional exhaustion)	3.64(1.34)	.05	.30 +	.76
4. Turnover intentions	3.21(1.58)	.01	.36 •	.66 •••	.83

+

p < .10; ^•p < .05; ^••p < .01; ^•••p < .001

*

In diagonals, we report scale reliabilities.

Psychological strain (emotional exhaustion)

We found a positive relationship between suppression and emotional exhaustion (b = 0.51, SE = 0.22, p = .03, 95 percent CI [0.07, 0.96]). Although the relationship between expression and emotional exhaustion was nonsignificant (b = 0.39, SE = 0.22, p = .08, 95 percent CI [–0.06, 0.84]), it was qualified by a significant interaction (b = 0.64, SE = 0.26, p = .02, 95 percent CI [0.12, 1.16]), supporting Hypothesis 2a. Specifically, expression was positively related to emotional exhaustion when suppression was high (b = 1.09, SE = 0.42, p = .01, 95 percent CI [0.24, 1.95]) but not when suppression was low (b = −0.34, SE = 0.29, p = .26, 95 percent CI [–0.93, 0.26]). To further probe the interaction, we also examined the relationship between suppression and emotional exhaustion at low and high levels of expression. Suppression had a positive relationship with emotional exhaustion when expression was high (b = 1.19, SE = 0.36, p < .01, 95 percent CI [0.47, 1.91]) but had no relationship when expression was low (b = −0.13, SE = 0.34, p = .70, 95 percent CI [–0.81, 0.55]). These patterns indicate that individuals enacting high levels of both expression and suppression—a curation approach—experienced the most emotional exhaustion, supporting Hypothesis 2a. Figure 2 shows this result.

OPEN IN VIEWER

Figure 2.

Study 1: Interactive Relationship of Expression and Suppression with Psychological Strain (Emotional Exhaustion)

Turnover intentions via psychological strain (emotional exhaustion)

Suppression had a positive relationship with turnover intentions (b = 0.69, SE = 0.27, p = .01, 95 percent CI [0.14, 1.23]), and there was no relationship between expression and turnover intentions (b = 0.33, SE = 0.27, p = .23, 95 percent CI [–0.22, 0.87]). The interaction was directionally (but not significantly) consistent with Hypothesis 3a (b = 0.48, SE = 0.31, p = .13, 95 percent CI [–0.15, 1.12]). Specifically, expression showed a directionally but not significantly positive relationship with turnover intentions when suppression was high (b = 0.86, SE = 0.51, p = .10, 95 percent CI [–0.18, 1.90]) but not when suppression was low (b = −0.23, SE = 0.36, p = .53, 95 percent CI [–0.95, 0.49]). We then examined suppression’s relationship with turnover intentions at low and high levels of expression. We found that suppression had a significantly positive relationship with turnover intentions when expression was high (b = 1.20, SE = 0.43, p = .01, 95 percent CI [0.32, 2.08]) but had no relationship when expression was low (b = 0.20, SE = 0.41, p = .63, 95 percent CI [–0.63, 1.02]). The overall pattern of results was consistent with Hypothesis 3a, such that a curation approach was highest in turnover intentions. Figure 3 shows this result.

OPEN IN VIEWER

Figure 3.

Study 1: Interactive Relationship of Expression and Suppression with Turnover Intentions

We next examined expression’s indirect relationship with turnover intentions via emotional exhaustion, moderated by suppression. Supporting Hypothesis 3b, the results showed a significant interaction between expression and suppression on emotional exhaustion (p = .02; see above), a significant positive association between emotional exhaustion and turnover intentions (b = 0.71, SE = 0.17, p < .001, 95 percent CI [0.37, 1.04]), and a significant index of moderated mediation (b = 0.45, SE = 0.23, 95 percent CI [0.02, 0.95]). Specifically, expression had a positive indirect relationship with turnover intentions via emotional exhaustion when suppression was high (b = 0.77, SE = 0.34, 95 percent CI [0.19, 1.52]) but not when suppression was low (b = −0.24, SE = 0.28, 95 percent CI [–0.79, 0.35]). We then also examined suppression’s indirect relationship to turnover intentions via emotional exhaustion, moderated by expression. Suppression had a positive indirect relationship with turnover intentions via emotional exhaustion when expression was high (b = 0.84, SE = 0.33, 95 percent CI [0.23, 1.51]) but had no indirect relationship when expression was low (b = −0.09, SE = 0.31, 95 percent CI [–0.74, 0.51]).

Additional Analyses Examining Curation Among Other Marginalized Groups

As described in our summary of deviations from preregistration (SOM Appendix A, #6), we moved our preregistered analyses with employees who have any marginalized identity (e.g., based on gender, sexual orientation, disability, etc.) to the supplementary online materials. Our hypotheses were not supported in this broader sample. However, in additional exploratory analyses, we focused on the two most common visible marginalized identities in our dataset: employees of color who answered our expression and suppression questions about identities related to race or ethnicity (n = 41 from our sample above) and women employees who answered about a gender, family, or caregiver identity (n = 130). When conducting analyses with these two groups combined, we found a significant interaction between expression and suppression on both emotional exhaustion and turnover intentions, ps < .05. Furthermore, we found significant moderated mediation. These results suggest that our findings regarding the detrimental effects of curation for employees of color likely extend to women employees. See SOM Appendix C for detailed results.

Discussion

Study 1 provides some initial evidence for our thesis that racial and ethnic minority employees using a curation approach (high expression and high suppression) experience the greatest psychological strain and, therefore, higher levels of turnover intentions in organizational settings. Moreover, Study 1 provides suggestive evidence that our hypotheses regarding curation apply not only to employees of color but also to women. However, Study 1 did not include a measure of ambivalence and is limited in its ability to address common method bias (Podsakoff et al., 2003). In addition, the small sample size rendered the study underpowered (post-hoc power = 31 percent for emotional exhaustion and 19 percent for turnover intentions) and prevented robustness checks using controls, due to a low variable-to-observation ratio (Tanaka, 1987). Study 2 addresses these limitations. Additionally, while the results of Study 1 support our predictions, there is also a somewhat surprising result: When expression is low, suppression did not relate to psychological strain or turnover intentions, despite long-standing research suggesting that suppression should be positively related to these outcomes (Hewlin, 2009; Madera et al., 2012; Ragins et al., 2007; Roberts et al., 2014). Study 2 examines whether this result persists with a larger sample size.

Participants and Procedure

Given that we were interested in recruiting a narrow segment of the population (racial/ethnic minority employees in white collar jobs) to complete an overtime study that required a high sample size and was subject to attrition across time points, our goal was to first attract as many eligible participants from the Prolific pool as possible, with a goal of achieving a minimum final sample of 200 racial/ethnic minority participants who met all of our eligibility criteria and passed all of our pre-registered exclusion criteria. ² Of the 2,855 participants who completed a pre-screening survey (January 2023), 1,276 met our eligibility criteria (racial/ethnic minority, employed in a U.S. white-collar job) and passed preregistered attention checks. We invited them to a four-week study (throughout April 2023) involving a weekly survey. The participants (N = 503) who completed Time 1 were invited to complete the remaining waves. To encourage participation, compensation increased at each wave ($3.00, $3.25, $3.50, and $3.75), plus a $2 bonus for completing all four. Of the 396 who completed all four waves, 22 failed at least one preregistered attention check, yielding a final sample of 374 (57 percent men, 42 percent women, 1 percent non-binary; 80.5 percent heterosexual, 19.5 percent other sexuality; 40.6 percent Asian, 31.8 percent Black, 25.9 percent Hispanic/Latinx, 17.1 percent White, 2.4 percent Indigenous, 1.1 percent Arab, and 1.1 percent Prefer to self-describe; M_age = 36.21 years; M_{organizational tenure} = 6.37 years).

Measures

The four time points were spaced apart by one week. We collected key measures at each time point, except for control variables, which we assessed only at Time 1. In addition to the measures described below, the participants completed exploratory outcomes, contextual moderators, individual difference moderators, and free response questions. See SOM Appendix D for details.

Preregistered Tests of Hypotheses

Means, standard deviations, and correlations of key measures are included in Table 3. While each of our key measures was collected at each time point, our preregistered plan specified analyzing each at a single time point (e.g., expression and suppression at T2). We adopted this between-subjects design given that we conceptualized identity management as an individual-level approach. ³ Although we used a between-subjects design, the longitudinal structure, spanning different time points for different model stages, helped to reduce concerns about common method bias (Podsakoff et al., 2003).

OPEN IN VIEWER

Table 3.

Means, Standard Deviations, and Correlations in Study 2 (Longitudinal Study)*

Variable	M(SD)	1	2	3	4	5
1. Expression	2.16(0.94)	.95
2. Suppression	1.87(0.98)	–.11 •	.96
3. Ambivalence	2.40(1.09)	–.09 +	.32 •••	.87
4. Psychological strain	0.00(0.89)	–.11 •	.47 •••	.69 •••	.87
5. Turnover intentions	3.38(2.03)	–.29 •••	.31 •••	.45 •••	.56 •••	.94

+

p < .10; ^•p < .05; ^••p < .01; ^•••p < .001

*

In diagonals, we report scale reliabilities.

To test our preregistered hypotheses, we examined the interactive relationship between T2 expression and T2 suppression with T3 ambivalence, T3 psychological strain, and T4 turnover intentions. When included in the model, controls were from T1. Expression and suppression were mean-centered in all analyses. We used the Process macro (Hayes, 2022) for simple moderation (model 1), moderated mediation (model 8), and sequential moderated mediation (model 86), with 10,000 bootstrapped samples to estimate 95 percent confidence intervals. As in Study 1, we examined mediation by using causal steps and an index-based test of moderated mediation (Yzerbyt et al., 2018). We again used the 16th, 50th, and 84th percentiles to represent low, average, and high levels of the moderator, respectively. We reported results without controls; however, results with preregistered controls (gender, sexual orientation, age, job tenure, and self-monitoring) were consistent. See SOM Appendix D for the full results with controls.

Ambivalence

To test Hypothesis 1, we conducted a simple moderation analysis. Suppression had a positive relationship with ambivalence (b = 0.40, SE = 0.06, p < .001, 95 percent CI [0.29, 0.51]), while expression had no relationship (b = −0.05, SE = 0.06, p = .38, 95 percent CI [–0.16, 0.06]). Supporting Hypothesis 1, their interaction was significant (b = 0.21, SE = 0.06, p < .001, 95 percent CI [0.10, 0.33]). The relationship between expression and ambivalence was positive when suppression was high (b = 0.19, SE = 0.09, p = .03, 95 percent CI [0.01, 0.37]) but negative when suppression was low (b = −0.24, SE = 0.07, p = .001, 95 percent CI [–0.38, –0.09]). We then assessed the relationship between suppression and ambivalence moderated by expression: Suppression was positively associated with ambivalence when expression was low (b = 0.20, SE = 0.07, p < .01, 95 percent CI [0.07, 0.33]) and even more so when expression was high (b = 0.63, SE = 0.09, p < .001, 95 percent CI [0.44, 0.81]). The highest ambivalence was therefore observed under high expression and suppression (i.e., the curation approach); see Figure 4. These findings support Hypothesis 1.

OPEN IN VIEWER

Figure 4.

Study 2: Interactive Relationship of Expression and Suppression with Ambivalence

Psychological strain via ambivalence

To test Hypothesis 2a, we first examined the interaction of expression and suppression on psychological strain. Suppression had a positive relationship with psychological strain (b = 0.46, SE = 0.04, p < .001, 95 percent CI [0.38, 0.55]), while expression had no relationship (b = −0.04, SE = 0.04, p = .33, 95 percent CI [–0.13, 0.04]). As predicted by H2a, their interaction was significant (b = 0.14, SE = 0.04, p = .001, 95 percent CI [0.06, 0.23]). Consistent with our expectations, when suppression was high, expression had a directionally although not significantly positive relationship with psychological strain (b = 0.12, SE = 0.07, p = .08, 95 percent CI [–0.02, 0.26]); when suppression was low, the relationship was significantly negative (b = −0.17, SE = 0.06, p < .01, 95 percent CI [–0.28, –0.06]). We then assessed the relationship between suppression and psychological strain moderated by expression. Suppression was positively related to psychological strain when expression was low (b = 0.32, SE = 0.05, p < .001, 95 percent CI [0.22, 0.43]) and even more so when expression was high (b = 0.61, SE = 0.07, p < .001, 95 percent CI [0.47, 0.75]). Accordingly, the curation approach (high expression and suppression) was associated with the greatest psychological strain, consistent with Hypothesis 2a; see Figure 5.

OPEN IN VIEWER

Figure 5.

Study 2: Interactive Relationship of Expression and Suppression with Psychological Strain

We next assessed the indirect relationship between expression and strain via ambivalence moderated by suppression (H2b). In the first causal step, as reported above, we found a significant interaction between expression and suppression on ambivalence (p < .001). In the next causal step, we found that ambivalence was significantly related to psychological strain (b = 0.49, SE = 0.03, p < .001, 95 percent CI [0.43, 0.55]). Finally, the index of moderated mediation was significant (b = 0.10, SE = 0.03, 95 percent CI [0.05, 0.16]), supporting Hypothesis 2b. When suppression was high, expression had a positive indirect relationship with psychological strain via ambivalence (b = 0.09, SE = 0.05, 95 percent CI [0.01, 0.18]): Greater expression was associated with greater ambivalence, which, in turn, was associated with greater strain. The opposite pattern emerged when suppression was low: There was a negative indirect relationship (b = −0.12, SE = 0.03, 95 percent CI [–0.18, –0.05]), such that expression was associated with less ambivalence and, in turn, less psychological strain. We also assessed the indirect relationship between suppression and strain via ambivalence moderated by expression. Suppression had a consistently positive indirect relationship with psychological strain via ambivalence, but this indirect relationship was stronger when expression was high (b = 0.31, SE = 0.04, 95 percent CI [0.23, 0.39]) compared to low (b = 0.10, SE = 0.04, 95 percent CI [0.02, 0.18]).

Turnover intentions via ambivalence and psychological strain

To test Hypothesis 3a, we first examined the interaction of expression and suppression on turnover intentions. We found, consistent with past research, a positive relationship between suppression and turnover intentions (b = 0.58, SE = 0.10, p < .001, 95 percent CI [0.37, 0.78]). Expression was negatively related to turnover intentions (b = −0.56, SE = 0.10, p < .001, 95 percent CI [–0.77, –0.36]). Contrary to Hypothesis 3a, there was no interaction between expression and suppression (b = 0.007, SE = 0.11, p = .94, 95 percent CI [–0.20, 0.22]). Thus, expression was associated with lower, and suppression was associated with higher, turnover intentions regardless of the other one’s levels; see Figure 6.

OPEN IN VIEWER

Figure 6.

Study 2: Interactive Relationship of Expression and Suppression with Turnover Intentions

Despite the lack of an interaction, conditional indirect effects are still possible (Hayes, 2022), whereby high suppression may amplify expression’s indirect relationship with turnover intentions via ambivalence and strain (H3b). We therefore tested sequential moderated mediation via ambivalence followed by strain. As reported above, there was a significant interaction of expression and suppression on ambivalence (p < .001). In the next two causal steps, we found that ambivalence had a significant positive relationship with psychological strain (b = 0.57, SE = 0.03, p < .001, 95 percent CI [0.51, 0.63]), and psychological strain had a significant positive relationship with turnover intentions (b = 1.01, SE = 0.14, p < .001, 95 percent CI [0.73, 1.28]). Finally, the index of moderated mediation was significant (b = 0.12, SE = 0.04, 95 percent CI [0.05, 0.20]). When suppression was high, expression had a significant positive indirect relationship with turnover intentions (b = 0.11, SE = 0.06, 95 percent CI [0.01, 0.23]), such that higher levels of expression were associated with greater ambivalence, which, in turn, was associated with greater psychological strain; this strain, in turn, was associated with higher turnover intentions. When suppression was low, the opposite pattern emerged: Expression had a significant negative indirect relationship with turnover intentions (b = −0.13, SE = 0.05, 95 percent CI [–0.23, –0.05]), such that higher levels of expression were associated with lower ambivalence, which, in turn, was associated with lower psychological strain; this lower strain, in turn, was associated with lower turnover intentions. We also assessed the indirect relationship between suppression and turnover intentions via ambivalence and strain moderated by expression. Suppression had a positive indirect relationship with turnover intentions, though this relationship was more potent at high expression (b = 0.31, SE = 0.07, 95 percent CI [0.19, 0.45]) than at low expression (b = 0.10, SE = 0.04, 95 percent CI [0.02, 0.19]). SOM Appendix D (Table D1) shows a table summarizing all mediation results.

Additional Analyses: Robustness Checks

We conducted additional analyses (reported in SOM Appendix D) suggesting that the proposed model fits the data better than alternative causal models do. First, we examined a reverse causality model (e.g., perhaps psychological strain causes people to engage in certain identity management approaches). Specifically, we examined a model in which turnover intentions (T2) led to psychological strain (T3), which, in turn, led to ambivalence (T3), and this ambivalence, in turn, led to both expression (T4) and suppression (T4). Second, we also tested a simultaneous model in which ambivalence and psychological strain are parallel mediators, contrasting it with our hypothesized sequential model. In both cases, our predicted model exhibited better fit.

Discussion

Study 2 provides evidence that a curation approach (high in both expression and suppression) is linked to significant psychological strain among employees of color. This pattern emerges because although expression is associated with diminished psychological strain when suppression is low (consistent with past theorizing), this relationship is reversed when suppression is high. Moreover, consistent with existing theory, suppression was positively associated with psychological strain at all levels of expression; however, we reveal that this relationship is strongest among people of color who engage in high levels of expression. The resulting greater psychological strain experienced by curators is associated with greater turnover intentions. Importantly, Study 2 reveals that ambivalence—feeling torn between opposing thoughts—is a driving factor underlying these negative intrapersonal outcomes.

Although we found that the psychological strain associated with the curation approach heightens turnover intentions, Study 2 did not find that employees of color who use a curation approach are the most inclined to leave their organizations overall. The reason for this is unclear, as we found no evidence in our exploratory analyses that curation has advantages, such as generating better insights into complex organizational dilemmas, fueling optimism about one’s leadership potential, or enabling greater job satisfaction (see SOM Appendix D). Of note, in a supplemental study, we found further evidence for our prediction that the curation approach leads to higher turnover intentions (see SOM Appendix E), indicating the possibility that moderators may strengthen or weaken this relationship.

The Perils of Identity Management: Curation, Expression, and Suppression

Our research contributes to a long-standing conversation in the identity management literature regarding the costs and benefits of identity expression and suppression in the workplace. Both expression and suppression are associated with trade-offs: Expression is often positioned as beneficial for well-being but professionally risky, and suppression is viewed as psychologically taxing but potentially necessary for success (Cha et al., 2019; Clair et al., 2005; Hewlin, 2003; Phillips et al., 2009). In an effort to resolve the dilemma of how to best navigate a minority identity, the literature is replete with examples of scholars looking for ways that employees of color can engage in identity expressions that are less professionally risky, such as sharing positive or meaningful aspects of one’s identity at work (Arnett, 2023; Phillips et al., 2009; Roberts, 2005; Thomas, 1993). Yet, these solutions often have hidden risks of their own, such as the potential to inadvertently cause anxiety, elicit stereotypes, or violate others’ perceptions of propriety. Perhaps because of these persistent trade-offs, scholars have proposed that it could benefit employees of color to both express and suppress (Roberts et al., 2008), for example by engaging in meaningful conversations that bring attention to positive aspects of one’s group (Arnett, 2023) but purposely disengaging from topics that could cause anxiety or friction such as systemic discrimination (McCluney et al., 2017). While we see the potential professional value of this approach, our research demonstrates that curation may not simply surmount professional risks but, rather, shift those risks internally, such that a person of color who wishes to embrace curation may shoulder a greater psychological burden. Moreover, while the intent of enacting curation could be to enable professional advancement, the psychological strain associated with this approach may prompt employees of color to exit their organizations before achieving their objective of getting ahead. This dilemma suggests a previously undiscussed double bind for those who express: They can either be an open book about their minority identity and risk being pushed out of their organizations due to negative perceptions of their identity, or they can curate, contorting themselves to the point where they feel compelled to leave. To the extent that our findings extend to women, they highlight yet another double bind they must navigate with regard to their gender identity in the workplace (Rudman & Phelan, 2008). Overall, our research illuminates not only the hidden psychological costs of curation but also the elusiveness of viable solutions that satisfy a dual desire for career prosperity and internal vitality.

Our findings reveal a critical boundary condition to the long-standing finding that expression, as well as authenticity more broadly, is beneficial for the well-being and commitment of employees of color (Cha et al., 2019; Madera et al., 2012; Ragins et al., 2007; Roberts et al., 2008). Indeed, these notions about the intrapersonal benefits of authenticity are so deeply ingrained in the literature that scholars developing new authenticity scales use relationships with well-being (higher vitality, lower negative affect and stress) and commitment (higher dedication) as evidence of construct validity (van den Bosch & Taris, 2014). Notably, our findings show that expression’s ability to improve psychological and retention outcomes is not simply eliminated among people who suppress but is actually reversed: Expression psychologically harms employees of color who engage in suppression and makes them actively consider alternative employment options. As such, our findings suggest that the well-established well-being benefits of expression, and authenticity in general, may be limited to individuals who refrain from suppression at work.

While our research uncovers a critical boundary condition to expression, we do not intend our interactive findings to cast aspersion upon expression but, rather, to reveal the pernicious nature of suppression. We advance the literature by revealing that suppression is problematic not only due to its negative main effect on well-being (as suggested by prior works; Bell, 1990; Clair et al., 2005; Hewlin, 2003) but also because it undermines or even reverses the benefits of expression. In this way, suppression essentially harms employees of color twice over: It not only robs them of the ability to be authentic when they conceal, but it also takes away their ability to benefit from authenticity when they engage in expression.

Broadening the Lens of Identity Management

Perhaps an even broader revelation that stems from our work is that, to gain a holistic understanding of the consequences of expression or suppression, researchers must move beyond understanding them in isolation and, instead, consider them jointly. Existing research on intrapersonal consequences has laid valuable groundwork for understanding how expression and suppression separately influence individual-level outcomes. Building on this foundation, we contend that to truly understand the complexities of the whole person, researchers should account for the broader portfolio of identity management, including the extent to which an individual expresses and suppresses. Using this portfolio perspective, we propose that past and future research focused on the consequences of expression and suppression would benefit from considering how the predictions for one variable are shaped by the other. For instance, past research focused on the psychological benefits of high expression may be implicitly theorizing about high expression combined with low suppression (e.g., Madera et al., 2012; Roberts et al., 2008), whereas research considering ambivalence associated with expression may be implicitly theorizing about high expression combined with high suppression (e.g., Cha & Roberts, 2019). Additionally, considering the psychological costs of suppression may benefit from considering the broad portfolio of identity management—including how much a person intentionally draws attention to their identity—when they attempt to understand the full landscape of challenges that employees of color grapple with at work (e.g., Hewlin, 2009; McCluney et al., 2019).

Through this research, we make explicit something that has been implicit in the literature: the notion that expression and suppression are two distinct dimensions. Past research has acknowledged that the same person can enact both expression and suppression (Roberts et al., 2008). However, the literature still reflects the idea that expression and suppression could be two ends of the same continuum and, thus, need not or perhaps even cannot be considered in conjunction. This assumption emerges when, for example, hypotheses about their effects are considered as opposites (e.g., if suppression is expected to decrease well-being and retention, then expression is predicted to increase these outcomes). Our explicit theorizing about the two-dimensional nature of expression and suppression pushes the literature toward considering outcomes that may have been overlooked, such as interactive effects as investigated in this research, and other possibilities, such as distinct effects on different outcomes.

Through this research, we contribute to a broader conversation within the diversity and identity management domains by illustrating high levels of nuance as requisite for addressing the complexities inherent in identity-related challenges. For instance, intersectionality research has shown that studies of the separate effects of race and gender are not sufficient for understanding the compounding impact of intersectional identities like being a Black woman (Dupree, 2024; Hall et al., 2019; Smith et al., 2019). In a similar vein, our research shows the importance of investigating complex identity management approaches whose outcomes cannot be clearly derived from existing knowledge about the distinct effects of expression and suppression.

The Importance of Integrating the Ambivalence Literature into Identity Management

Our research adds more theoretical and empirical depth to an ongoing conversation about the complex relationship between identity management and ambivalence. Cha and Roberts (2019) primarily focused on ambivalence as an antecedent to identity management, but they also recognized that expression can be a trigger for ambivalence, particularly due to the tension from viewing one’s identity as both a resource and a liability. Taking an intrapersonal outcomes perspective, our research builds on the foundation set by existing research to demonstrate that expression is not necessarily associated with greater ambivalence. Indeed, when expression was combined with low suppression, ambivalence was actually lower. However, among individuals who combine expression with the seemingly oppositional behavior of high suppression (consistent with a curation approach), ambivalence was amplified. While shifting between oppositional behaviors can be associated with ambivalence, scholars also acknowledge the need for deeper insight into the psychological ramifications of such an approach (Pratt & Doucet, 2000). We therefore engage in deeper integration of the ambivalence literature to understand why curation is detrimental. We theorize that curation involves a compromise strategy for managing ambivalence, which is especially likely to keep opposing thoughts (i.e., one’s identity as a resource and liability) salient in a way that can lead to negative psychological experiences, particularly when the two sides cannot be easily reconciled (Ashforth et al., 2014).

Yet, an even deeper dive into the ambivalence literature provides greater insight into our research and reveals the value of better integrating ambivalence theory into the identity management literature. Ambivalence theory asserts that the mere presence of ambivalence need not be detrimental; rather, various factors, including how a person navigates their ambivalence, can influence outcomes (Ashforth et al., 2014; Rothman et al., 2017). Thus, although our theory and findings suggest that curation invokes a compromise strategy that keeps opposing thoughts alive, the ambivalence literature reveals that acute awareness of opposing sides need not always be as psychologically detrimental as our research indicates. When the two opposing sides are synergistic or reconcilable, employees can move beyond a compromise strategy to a more adaptive way of managing ambivalence, referred to as holism (Ashforth et al., 2024). A holistic response is susceptible to similar challenges with regard to keeping both orientations alive but may have redeeming qualities in that it involves finding ways to channel the opposing orientations toward some productive end or overall positive perspective (Ashforth et al., 2014; Pratt & Pradies, 2011). To illustrate, consider research on tempered radicals, employees who are radical about social change but also feel pressure to temper this passion to succeed at work (Meyerson, 2001). Tempered radicals experience ambivalence because they are balancing two antithetical viewpoints: the view of oneself as an insider to an organization, who is seeking to get ahead and often conforms to norms to do so, and the view of oneself as an outsider, often based on demographic differences or different belief systems, who is seeking to shift the status quo by enacting social change within the organization (Meyerson & Scully, 1995). Yet, these opposing thoughts can be understood in a synergistic way, such that being an outsider enables a person to see the need for social change, and being an insider enables them to enact it (Meyerson & Scully, 1995). In this way, oppositional thoughts, although challenging, are appraised as adding value to the broader cause of social change, leading ambivalence scholars to characterize tempered radicalism as an example of holism rather than compromise (Ashforth et al., 2014).

This counterfactual regarding holism raises the question of why curation is not similarly a form of holism and, instead, seems to be a compromise that accentuates ambivalence and yields negative intrapersonal outcomes. In theory, it seems as though curation, like tempered radicalism, could have synergies. Indeed, Roberts (2005) characterized certain forms of expression (particularly positive distinctiveness, which focuses on shaping a positive narrative related to one’s identity) and certain forms of suppression (specifically social recategorization, which combines downplaying one’s marginalized identity with affiliating with positively regarded identities) as both enabling career success. To the extent that expression and suppression are perceived as enabling the same goal (i.e., career success), they can be viewed as synergistic. Yet, empirical research did not find evidence of this possibility, instead finding that suppression is associated with lower, not greater, perceived career success (Roberts et al., 2014). These findings are consistent with our premise that suppression largely evokes negative thoughts that are incompatible with viewing one’s identity as a resource, rather than positive thoughts that could be viewed as synergistic with a resource view. This supports our theorizing that curation involves constantly grappling with irreconcilably incompatible thoughts, and this is the basis through which negative downstream intrapersonal consequences emerge.

Overall, it is clear that both ambivalence and identity management theory are highly nuanced; yet, the identity management literature to date has only scratched the surface of integrating ambivalence theory. We hope our research will spark future identity management studies and theoretical frameworks that engage more deeply in ambivalence theory, as the answers that scholars are seeking regarding how employees from marginalized groups can effectively harness resources and mitigate liabilities may lie within the ambivalence literature.

Practical Implications

Organizations may be inadvertently losing diverse talent because employees of color are straining themselves to curate the perfect workplace image. One of the most likely causes of this predicament is when organizations send mixed messages, encouraging employees to be themselves without accounting for the risks that may compel employees to suppress. Organizational leaders often feel ambivalent about diversity (Creary, 2025), and employees frequently detect leaders’ ambivalence through mixed messages (Ashforth et al., 2014). These mixed messages can come out in various ways. For instance, organizations can say they value diversity yet also pigeonhole employees of color (Ely & Thomas, 2001) or reject notions that discrimination exists within their walls (Kaiser et al., 2013). Following the cue that diversity is valued, employees may seek opportunities to express. Yet, an undercurrent of risk—that their race or experiences related to it could be a liability—may cause them to suppress.

Employees who perceive fewer mixed messages may be less likely to curate. Take, for example, the concern that trying to leverage one’s identity as a resource by speaking up about negative assumptions about one’s group may lead to being seen as too disruptive (Ashford et al., 1998; Cha & Roberts, 2019; Rattan et al., 2022), which can, in turn, cause marginalized employees to suppress (Shelton & Stewart, 2004). A tech organization can make it clear that although speaking up about concerns, including concerns with discrimination, is viewed as an unwanted disruption in some places, the organization values and actively seeks disruption as a way to fulfill its values of innovation and continuous improvement. In this way, the organization is taking something traditionally viewed as a liability associated with a minority identity and reframing it as valued in a way that can enhance self-esteem and self-efficacy (Tajfel & Turner, 1986). By reframing a potential liability, such as a concern about being viewed as disruptive, as part of the reason that a minority identity is an asset, organizations can send a consistent message that employees’ identities are valued. Such messages may make employees less inclined to curate.

Limitations and Future Directions

Theoretical and empirical work extending beyond the scope of this research will be useful for understanding how, when, and why people curate, as well as the associated outcomes. First, building on the above practical implications, we see promising opportunities to build on our research by examining the antecedents of curation. Doing so may be an important step toward creating inclusive and equitable environments for employees of color. Future research can also examine whether employees of color who invoke a curation approach are doomed to worse well-being or whether there are avenues for overcoming the psychological strain uncovered in our study. Curation is an alluring solution, but it also functions as a potential trap for employees who may unwittingly strain themselves by trying to craft the perfect image at work. Our findings suggest that this trap occurs because racial and ethnic minority employees view expression and suppression as separate, non-synergistic identity management approaches, fostering a compromise response to ambivalence. We encourage future research to build on the ambivalence theoretical framework by investigating whether it is possible to promote a holistic response to ambivalence whereby expression and suppression are viewed as synergistic. Such research could help to bridge our findings with studies suggesting that curation could yield positive career-related outcomes (e.g., Roberts et al., 2008).

Future studies can also further investigate when and why curation leads to greater turnover intentions. In Study 2, we found that while curation is associated with the highest levels of psychological strain, employees drawing on this approach did not exhibit the highest turnover intentions, despite our finding that psychological strain as a mechanism is associated with greater turnover intentions. These Study 2 results suggest there is a mechanism that, at least in some circumstances, counteracts the negative relationship between psychological strain and turnover intentions. However, we found no evidence in our exploratory analyses that curation has other advantages (see SOM Appendix D). One possible explanation that we did not explore in the present research is that, although curation is psychologically straining, it nonetheless successfully shields employees from discrimination and, thus, enhances their commitment to the organization. Indeed, employees sometimes resolve their ambivalence by accepting the bad parts of their organization along with the good in a way that strengthens their bond to the organization (Brickman, 1987; Rothman et al., 2017). We see fruitful opportunities to explore this possibility in future studies.

Future research can further examine the relationships and causal chain underlying the effects shown in our studies. We used longitudinal data to lower the chances of common method bias, which could inflate relationships among our variables. However, we acknowledge that our primary evidence is based on self-reported, between-person analyses that are correlational in nature. Our study is therefore limited in its ability to definitively establish causality. Although this limits our ability to infer causality, we encourage future studies to use methodologies that build on the foundation we developed in this research.

Our framework establishes new footing for research on other marginalized identity groups in the workplace. Our initial foray into whether the negative consequences of curation extend to other marginalized groups suggests that the challenges of curation most likely apply to people with visible, historically marginalized social identities, such as gender, as well as race and ethnicity. That said, although we examined a range of potentially invisible social identities (e.g., sexual orientation, disability), each one was extremely underrepresented within our Study 1 dataset, which means we do not know whether our findings also apply to specific invisible social identities. We encourage future research examining how expression and suppression may interactively influence intrapersonal outcomes for both visible and invisible identities and whether the visibility of a marginalized identity may serve as a boundary condition to our findings. We hope that our work sparks continued inquiry into the complex and nuanced dynamics of identity management, particularly among employees with traditionally marginalized social identities.