Colorblind and multicultural diversity strategies create identity management pressure

Colorblind and multicultural diversity strategies may create identity management pressure , leading minorities to assert or distance from their racial identity. In two experiments ( N = 307, 279), Asian and Asian American participants in the United States completed racial identification measures, contemplated employment at a company expressing a multicultural, colorblind, or control strategy, and completed measures assessing ingroup similarity and comfort in the company. In the colorblind condition, participants who were strongly identified with their racial ingroup downplayed similarity to the ingroup and expressed less comfort relative to multicultural and control conditions. Participants who were weakly identified reported more similarity (but inconsistently) and more comfort in the colorblind relative to multicultural and control conditions. Thus, diversity strategies convey different meanings to strongly and weakly identified Asian individuals, with the former responding to colorblindness with identity distancing and the latter with identity assertion. Multiculturalism does not alter the typical pattern expected, with strongly identified asserting their identity more than weakly identified.

White people often fit into their workplaces relatively seamlessly, but people of color face mixed messages about how to present themselves and navigate their minority identity (Compton, 2016).
For example, people of color sometimes report pressure to distance from their racial or ethnic ingroup in professional contexts, perhaps because it is not valued (Debose, 1992;Derks, van Laar, Ellemers, & Raghoe, 2015;Garner & Rubin, 1986;Myers-Scotton, 1993;Roberts, 2005;Thomas, 1993), but also report the opposite pressure to assert their ingroup identity and confirm identity-based expectations (Kanter, 1977;Sinclair, Huntsinger, Skorinko, & Hardin, 2005). While most people navigate workplace impression management concerns (DuBrin, 2011;Wieland, 2010), people of color or other underrepresented groups may more frequently experience identity management pressure, or pressure to assert or distance from their ingroup identity to fit contextual norms.
Identity management pressure can come from a range of environmental cues. For example, scholars have examined how discrimination cues (e.g., Derks, Van Laar, & Ellemers, 2016) or interpersonal demands (Sinclair et al., 2005) can create pressure to assert or distance from one's ingroup identity. Yet, no research has examined the impact of workplace diversity messages on identity management decisions (but see Lyons, Wessel, Ghumman, Ryan, & Kim, 2014).
Diversity strategies are one way by which organizations send messages about identity management. Diversity strategies represent sets of ideas and practices about how people from different backgrounds should interact, relate to, and accommodate each other, with the goal of promoting and managing diversity. Despite increasing interest in diversity strategies (e.g., Apfelbaum, Stephens, & Reagans, 2016;Gündemir, Dovidio, Homan, & Dreu, 2016;Jansen, Vos, Otten, Podsiadlowski, & van der Zee, 2016;Plaut, Thomas, & Goren, 2009;Purdie-Vaughns, Steele, Davies, Ditlmann, & Crosby, 2008), most research has focused on affective reactions or workplace performance expectations, rather than how they affect cognitive representations of the self.
In the present research, we examine novel questions about whether organizational diversity strategies serve as important contextual cues that lead people of color to engage in identity management, such as self-group distancing. A one-size-fits-all approach to understanding these processes would be suboptimal because of individual differences in the importance people of color attribute to their racial identity. Thus, we take a more nuanced approach that considers the role of racial identification in tailoring diversity strategies to individuals. A nuanced approach to preventing identity management pressure is important because identity distancing and assertion processes may be taxing (Harter, 2002;Schmader & Sedikides, 2017), and optimizing diversity strategies is crucial for optimizing workplace outcomes (Apfelbaum et al., 2016;Jansen et al., 2016;Plaut et al., 2009;Purdie-Vaughns et al., 2008).

| Norms conveyed by diversity strategies
The two most prominent diversity strategies represent opposing approaches to navigating diversity and difference. Whereas some organizations advocate a multicultural strategy that celebrates diversity and group differences, others instead advocate a colorblind strategy (sometimes termed a value-in-homogeneity approach) that downplays difference and focuses on the similarity of their workforce (but see Gündemir et al., 2016;Gündemir, Martin, & Homan, 2019 for a discussion of a form of colorblindness that celebrates individual differences).
Multiculturalism has advantages over colorblindness in the workplace, particularly for workplace well-being among people of color (see Gündemir et al., 2019;Plaut, 2002;Rattan & Ambady, 2013 for reviews). However, it is premature to conclude that multiculturalism is universally beneficial for people of color.
Multiculturalism has notable drawbacks, such as increasing stereotyping of racial minorities (Gutiérrez & Unzueta, 2010;Wolsko, Park, Judd, & Wittenbrink, 2000). When primed with multicultural as opposed to colorblind strategies, Whites stereotype African American and Latinx individuals more on both positive and negative attributes (Wolsko et al., 2000). Additionally, White Americans like stereotypical minorities more when primed with multiculturalism but like counter-stereotypical minorities more when primed with colorblindness (Gutiérrez & Unzueta, 2010). Taken together, this research suggests that multiculturalism creates an expectation for people of color to stay within the bounds of their racial identity, or even an expectation to assert their identity. Colorblindness, however, creates an expectation for people of color to distance from their identity.
Accordingly, people of color may feel pressure to behave in line with the norms of an organization's diversity strategy (i.e., experience identity management pressure). Indeed, workplace norms and cultures send messages about desired behavior (e.g., Gelfand, Leslie, Keller, & de Dreu, 2012), and people are particularly attuned to impression management in workplace contexts (DuBrin, 2011). The social identity model of de-individuation effects (SIDE model; Reicher, Spears, & Postmes, 1995) contends that there is a strategic component of social identity-this component monitors environmental cues to determine whether identity assertion or distancing is appropriate. For example, when women are motivated to affiliate a man (e.g., because he holds an authority role) and learn that he holds stereotypical views of women, women strategically present themselves as consistent those stereotypes (Pack & Zanna, 1975).
Similarly, African Americans who interact with a team leader holding stereotype-consistent views present themselves as more stereotypical (i.e., less academically invested; Sinclair et al., 2005). However, they do not present as more stereotypical when interacting with someone holding no power. These findings highlight the strategic component of identity assertion.
These strategies apply to identity distancing as well. For example, in an effort to increase their likelihood of obtaining a job, African Americans sometimes "whiten", or remove stereotypic information from their resumes (Kang, DeCelles, Tilcsik, & Jun, 2016). When companies express a multicultural strategy, however, participants whiten their resumes less, suggesting that they are more willing to assert their racial identity. Conversely, colorblind strategies may create pressure to disconfirm stereotypes or distance from one's identity (see Gutiérrez & Unzueta, 2010).

| The role of racial identification in identity assertion and distancing
Although diversity strategies may send a message about workplace norms, individual characteristics may lead to different responses. In particular, strength of group identification shapes how group members respond to identity management pressure (Ellemers, Spears, & Doosje, 2002). People who are strongly identified with their group are motivated to express or assert their identity even if it is not valued in a particular context (Ellemers et al.,, 2002). Weakly identified group members, on the other hand, are more likely to strategically assert or distance from their identity in response to environmental norms or other strategic concerns (e.g., Ellemers, Barreto, & Spears, 1999;Pickett, Bonner, & Coleman, 2002;Spears, Doosje, & Ellemers, 1997), as they are primarily concerned with individual mobility or improving their own social status. For example, when psychology students learn that their group is perceived negatively in comparison to business students, those weakly identified with psychology distance from the psychology identity by reducing their levels of self-stereotyping (i.e., presenting oneself as similar to prototypical ingroup members; Spears et al., 1997). Students who are strongly identified with psychology report similar levels of self-stereotyping irrespective of how their group is perceived.
Weakly identified group members also assert their identity when it is contextually beneficial. For example, when weakly identified group members have the opportunity to increase either their group's status or their own individual status, they only prioritize the group when they know that other members of their group will be able to see their decision (Barreto & Ellemers, 2000). When their decisions are anonymous, they prioritize individual over group status, again demonstrating a propensity to engage in identity assertion for strategic purposes.People who are strongly identified with the group are more likely to prioritize group advancement regardless of whether their decisions are known or anonymous.

| Hypotheses
In the present research, we also argue that weakly and strongly identified group members may respond differently to diversity strategies. We discuss our expectations for different identification levels first (Hypothesis 1), then expectations for relationships within diversity strategies (Hypothesis 2; see Figure 1 for a prediction graph). For hypotheses by identification level, we expect little responsiveness to diversity strategies among people of color who are strongly identified with their racial ingroup, such that there will be no difference in racial identity assertion/distancing across conditions (Hypothesis 1a). In contrast, we expect people of color who are weakly identified to comply more strongly with contextual cues, such that those in the multicultural condition assert their ingroup identity more (i.e., distance less) than those in the colorblind and control conditions (Hypothesis 1b).
With respect to hypotheses by condition, in the control and colorblind conditions, we expect stronger identification with the racial ingroup to be associated with more identity assertion (i.e., lower distancing; Hypothesis 2a and 2b, respectively). Under typical circumstances (i.e., in the control condition), the stronger a person identifies with their racial group, the more likely they are to assert their racial identity (Ellemers et al., 2002;Hogg & Turner, 1987).
Because colorblindness has historically been the default model in American company contexts (Plaut, 2002), participants' racial identity assertion in the colorblind condition may not differ from how they present themselves by default. In the multicultural condition, identity assertion will be high irrespective of the level of identification (Hypothesis 2c), leading to no relationship between identification and identity assertion and distancing.

| Present studies
In the present studies, we will focus specifically on Asian American and Asian participants, who are one of the fastest growing racial minority groups in US society (Colby & Ortman, 2014), but who are underrepresented in professions such as management (Johnson & Sy, 2016). Furthermore, they have been neglected in research on diversity strategies thus far and face unique identity management issues in the workplace. Whereas some groups largely face negative stereotypes in professional domains (e.g., Black individuals stereotyped as low in intelligence; Devine & Elliot, 1995), some of the workplace-relevant stereotypes Asian individuals face are considered positive or are more mixed (e.g., intelligent, math-oriented, obedient, and loyal; Ho & Jackson, 2001;Lin, Kwan, Cheung, & Fiske, 2005). Thus, they may be especially willing to strategically assert their racial identity in response to workplace norms because some stereotypes of their group are relatively compatible with the workplace. On the other hand, they might nonetheless be inclined to distance from their identity because even positive stereotypes are perceptually linked with negative ones (Czopp, Kay, & Cheryan, 2015;Siy & Cheryan, 2016) and because the negative stereotypes that exist (e.g., cold, unlikable, and unassertive; Lin et al., 2005) are not considered conducive to success in all domains. Indeed, some negative stereotypes of Asians are proffered as an explanation for their underrepresentation in management (Sy et al., 2010).
To examine these ideas, Asian American and Asian participants learned about an organization that advanced either a multicultural or colorblind diversity strategy or no diversity strategy (control condition) and then considered employment at that organization.
They imagined engaging in the recruitment process and completed a range of measures assessing their racial identity assertion or distancing (e.g., reports of their similarity to the racial ingroup; Studies 1 and 2) and their comfort in the organizational context (Study 2).
We defined racial identity assertion and distancing relative to the control group, such that reporting more ingroup similarity relative to the control group was classified as identity assertion. Reporting Weak idenƟficaƟon Strong idenƟficaƟon IdenƟty asserƟon

MulƟcultural
Colorblind Control less ingroup similarity relative to the control group was classified as identity distancing.

| S TUDY 1
In Study 1, Asian American and Asian participants learned about a multicultural, colorblind, or control company context and then responded to measures assessing racial identity assertion, including self-reported racial ingroup similarity, stereotypical self-descriptions, and interest in stereotypical activities.

| Participants
We recruited 325 undergraduate students of an Asian ethnic background from a large public university in the United States (55% female; M age = 19.48, SD = 1.54) during the 2013-2014 academic year. Participants completed the online experiment in exchange for partial psychology course credit. Eighteen participants were omitted from analyses due to missing data or because they did not identify as Asian or Asian American. Of the remaining 307 participants, the most common ethnicities were Chinese (32%), Korean (14%), Taiwanese (8%), and Vietnamese (8%).
We collected data until the end of an academic term. We ran a sensitivity analysis using GPower 3.1.5 (Faul, Erdfelder, Buchner, & Lang, 2009) to determine the effect size needed to detect an interaction between racial identification and diversity condition (i.e., the linear bivariate regression test for the difference between slopes across two groups, the colorblind vs. multicultural comparison). Using α = .05 with n = 107 and n = 96 (residual σ = 1.23, σ_x1 = 1.40, σ_x2 = 1.03), we needed a change in slope of b = 0.42 across conditions to achieve 80% power. Similar research (Derks et al., 2015) examining contextual effects on ingroup distancing showed a change in slope of b = 1.89, which suggested that we had sufficient statistical power.

| Procedure
The procedures in all experiments received ethical approval from the University of Washington Human Subjects Division. After consenting to participate in the research, participants read a recruitment brochure from a fictitious company, CCG Business Consulting, and were asked to imagine that they were considering working at the company. The brochures contained neutral images of an office space, a list of the company's services (e.g., advice, planning, and consulting), and their diversity strategy contained in a section entitled "Our Staff". For the manipulation of diversity strategy, participants were randomly assigned to read that the company advocated a multicultural or colorblind strategy to diversity management (adapted from Purdie-Vaughns et al., 2008) or were given no information about the company's diversity strategy within the brochure. In the multicultural and colorblind conditions, the statements were identical other than their focus on celebrating differences (e.g., "we train our ethnically and culturally diverse workforce to embrace their differences") as opposed to similarities (e.g., "we train our ethnically and culturally diverse workforce to embrace their similarities"). The design of the control condition was also identical to the other conditions, but it had no statement about their staff (see Appendix S1). After reading the manipulation, participants summarized the brochure in their own words and then completed a range of self-stereotyping measures. 1

Ingroup similarity
The primary measure assessed participants' reported similarity to their racial ingroup, which has been used as an effective measure of identity assertion and distancing in other research (Spears et al., 1997).

Stereotypical self-descriptions
Participants also responded to a less direct measure of self-stereotyping that assessed how stereotypically they described themselves on specific stereotype dimensions. This approach has also served as an effective measure of identity assertion and distancing in other research (Derks, van Laar, Ellemers, & de Groot, 2011;Derks et al., 2015). On a 1 (Not at all descriptive of me) to 7 (Very descriptive of me) scale, they responded to the statement "In this workplace, I would feel..." for 32 traits. The questions comprised several traits rated as stereotypical of Asians in research by Ho and Jackson (2001) and Katz and Braly (1933), along with filler traits. Our independent piloting showed that 11 of these traits were judged by Asian individuals themselves to encompass cultural stereotypes of Asians (see Appendix S1 for details about the pilot study). We averaged these 11 traits (loyal, intelligent, hard-working, educated, mathematical, self-disciplined, traditional, competitive, quiet, obedient, nerdy) to form a measure where higher values indicated more stereotypical self-descriptions (α = .80).

Interest in stereotypical activities
Next, in line with Derks et al. (2015), participants answered a series of questions assessing their interest in stereotypical activities (adapted from Steele & Aronson, 1995 scale, they responded to 18 items, but only two were judged as encompassing cultural stereotypes of Asians in our independent piloting (see Appendix S1 for details): "I enjoy Math oriented courses" and "I enjoy Science oriented courses". We averaged these two items to form a measure where higher values indicate more interest in stereotypical activities, but the reliability of the measure was poor (ρ = .63).

Racial identification
Next, participants responded to the identity subscale of Luhtanen Although racial identification was measured after the manipulation, a subset of participants had also participated in a large pre-screening questionnaire at the beginning of the school year (n = 135). The questionnaire contained the same measure of racial identification and was related to racial identification in this study at r(133) = .55, p < .001.

Manipulation check and demographics
Participants next responded to a question assessing their under-  Table 1.

| Primary analyses
A potential caveat of our research methodology was that we measured our moderator variable, racial identification, after the manipulation of diversity strategy and the dependent variables. We addressed this concern in two ways. Before proceeding with racial identification as a moderator in analyses, we checked and confirmed that racial identification was not different across the multicultural 005. Indeed, this is consistent with theorizing and findings that the identity centrality dimension of racial identification tends to be relatively stable across situations (Sellers, Rowley, Chavous, Shelton, & Smith, 1997).
Secondly, as noted in the Method, a subset of participants completed racial identification before the start of the study in a pre-screening session. All results reported below were the same when using the pre-screening measure of racial identification or the measure at the end of the study. In order to boost statistical power, we report results below including identification measured at the end of the study, but we report exact statistics for the pre-screening measure (Tables S3 and S4) in Appendix S1.

Analytic strategy
We first dummy coded diversity condition, designating the colorblind condition as the reference group. Thus, one variable compared the multicultural condition to the colorblind condition, and the other compared the control condition to the colorblind condition. We also mean-centered racial identification before entering it into the model and calculating interaction terms.
We ran a hierarchical moderated regression with centered racial identification and the two dummy coded diversity condition variables in the first step. Two-way interactions between multiculturalism and racial identification and between colorblindness and racial identification were entered into the second step. If the two interaction terms together in Step 2 accounted for a significant change (α = .05) in the variance accounted for (ΔR 2 ), then we examined the individual interaction terms further. For statistically significant interactions, we ran simple slope and simple effects analyses using the PROCESS macro (Hayes, 2013). In simple effects analyses, we broke down the effect of condition for those high (1 SD above the mean) TA B L E 1 Means, standard deviations, and correlations between primary Experiment 1 variables and low (1 SD below the mean) on racial identification, consistent with recommendations by Aiken and West (1991). We also conducted the Johnson-Neyman technique (Johnson & Neyman, 1936;Preacher, Curran, & Bauer, 2006), reported in Table 3, which supplements the simple effects analysis by testing for the precise value of racial identification at which the differences by condition become statistically significant.

Ingroup similarity
As shown in Table 2 and Figure 2, an interaction emerged between racial identification and diversity condition, but the pattern of results was unexpected. Inconsistent with Hypothesis 1a, participants who were strongly identified with their racial group reported lower ingroup similarity in the colorblind compared to multicultural and control conditions. This suggests that strongly identified Asians complied with the norm expressed in a colorblind context that individuals should downplay group identities. In contrast to Hypothesis 1b, weakly identified participants self-reported higher ingroup similarity in the colorblind compared to multicultural and control conditions. Both of these patterns are precisely the opposite of our predictions. Table 3 shows simple effects analyses for the effect of condition on ingroup similarity separated by weak and strong racial identification, as well as simple slope analyses.
Simple slope analyses also deviated from expectations, with the exception of Hypothesis 2a, where stronger identification with the racial ingroup was associated with reports of more ingroup similarity in the control condition. Inconsistent with Hypothesis 2b, however, there was no relationship between identification and ingroup similarity in the colorblind condition. Also inconsistent with Hypothesis 2c, in the multicultural condition, stronger identification with the racial ingroup was associated with more ingroup similarity. Overall, the multicultural condition showed the positive relationship expected of the colorblind condition (and was similar to the control condition), and the colorblind condition showed the null relationship expected of the multicultural condition.

Stereotypical self-descriptions
As shown in Table 2, an interaction emerged between racial identification and diversity condition and showed a similar pattern to the ingroup similarity measure. Inconsistent with Hypothesis 1a, participants who were strongly identified with their racial group de- Step 1

Racial iden fica on
to multicultural and control conditions. This suggests a similar compliance as for the ingroup similarity measure, where participants conformed to the colorblind norm of downplaying their group identity. In contrast to Hypothesis 1b, weakly identified participants described themselves more stereotypically in the colorblind compared to multicultural condition (but not the control condition). Table 3 shows simple effects analyses for the effect of condition on stereotypical self-descriptions separated by weak and strong racial identification, as well as simple slope analyses.
Simple slope analyses also deviated from expectations. Inconsistent with Hypothesis 2a (and the ingroup similarity measure) and 2b, there was no relationship between identification and stereotypical self-descriptions in the control or colorblind conditions. Also inconsistent with Hypothesis 2c, in the multicultural condition, stronger identification with the racial ingroup was associated with less stereotypical self-descriptions. Again, the multicultural condition showed the positive relationship expected of the colorblind condition, and the colorblind condition showed the null relationship expected of the multicultural condition.

Interest in stereotypical activities
Inconsistent with Hypotheses 1 and 2, the interactions between diversity condition and racial identification did not contribute to a significant change in the variance accounted for (ΔR 2 < .01, ps > .114; see Table 2), so we did not examine the interactions further. 2

| Discussion
We expected that a multicultural diversity strategy would encourage identity assertion, whereas colorblindness would encourage identity distancing. Additionally, we expected that weakly identified Asian and Asian American individuals would be most susceptible to pressure to conform to the norms expressed by diversity strategies.
These hypotheses were not confirmed-we instead observed the reverse pattern. Strongly identified participants unexpectedly complied with the norms expressed by diversity strategies, distancing from their identity in the colorblind compared to the multicultural and control conditions. Weakly identified participants were more likely to assert their identity in the colorblind (relative to the control condition), rather than the multicultural condition, which shows a rejection of colorblind norms. These findings were largely consistent across reports of ingroup similarity and stereotypical selfdescriptions, with some minor inconsistencies across measures for the control condition.
Given the unexpected findings in Study 1, we sought to replicate and further understand the effect. One possible explanation for these results is that colorblindness sends a safety cue to weakly identified Asian and Asian American individuals, which makes them more comfortable expressing similarity to their racial ingroup.
Identity safety cues signal that one's stigmatized identity will not negatively impact them in a particular setting (Chaney, Sanchez, & Remedios, 2016;Davies, Spencer, & Steele, 2005). Although identity safety cues typically acknowledge and affirm group identities (e.g., Chaney et al., 2016;Purdie-Vaughns et al., 2008), people who are weakly identified with a stigmatized ingroup may instead view a colorblind strategy as a signal of safety. Among people who are 2 We also included a similar measure that is described in Appendix S1. Participants responded to an open-ended question where they described their activities and interests. Independent raters assessed their responses for sterotypicality, but we did not find an effect of diversity condition with this activity measure either. Activities and interests are less relevant to the workplace so may not be an ideal way to capture changes in stereotypical presentation, as also demonstrated through the self-report activity measure in Study 1.

TA B L E 3
Simple effects and simple slope analysis on ingroup similarity and stereotypical self-description in Study 1 Note: The first three rows show simple slopes (i.e., the effect of racial identification on the dependent measures separated by condition) for statistically significant interactions. The second set of effects shows the simple effects (i.e., the effect of diversity condition on the dependent measures separated by weak and strong racial identification). J-N refers to the Johnson-Neyman technique, which supplements the simple effects analysis by testing for the precise value of racial identification at which the differences by condition become statistically significant at p < .05. Cells with hyphens indicate that any difference across condition is not statistically significant within the observed range of racial identification.
strongly identified, on the other hand, colorblindness may create discomfort, leading them to downplay their similarity to their ingroup.

| S TUDY 2
To understand whether colorblindness sends a safety cue to weakly identified participants, but the reverse for those strongly identified, we examined participants' trust and comfort in the company setting, in addition to ingroup similarity. We focused specifically on ingroup similarity (and dropped stereotypical self-descriptions and interest in stereotypical activities) to keep the study a reasonable length and because we considered it the most direct proxy for identity assertion and distancing. We also changed the organizational context from a consulting company to an engineering company to explore the generalizability of the phenomenon. Finally, we did not include the control group in this study because it did not differ from the multicultural condition in Study 1.

| Participants
We Participants completed the online experiment in exchange for partial psychology course credit. Three participants were omitted from analyses due to missing data. Of the remaining 279 participants, the most common ethnicities were Chinese (50%), Korean (13%), and Vietnamese (13%).
We collected data until the end of an academic term. We ran a sensitivity analysis using GPower 3.1.5 to determine the effect size needed to detect an interaction between racial identification and diversity condition (i.e., the linear bivariate regression test for the difference between slopes across two groups, the colorblind versus multicultural comparison). Using α = .05 with n = 148 and n = 130 (residual σ = 1.07, σ_x1 = 1.24, σ_x2 = 1.23), we needed a change in slope of b = 0.29 across conditions to achieve 80% power. Similar research (Derks et al., 2015) examining contextual effects on ingroup distancing showed a change in slope of b = 1.89, suggesting that we had sufficient statistical power.

| Procedure
After consenting to participate in the research, participants followed the same procedure as in Study 1, but were randomly assigned either to the colorblind or multicultural condition. The brochure manipulations discussed an engineering firm and were also bolstered with employee testimonials that supported the diversity strategy expressed in the brochure. Both employees stated "It seems like we're in a unique position of having many different cultural and ethnic groups working at our company", but the employee in the multicultural condition added, "and I think the company considers this a great asset because different cultural groups bring different perspectives, providing a richness in styles of interaction and problem solving strategies". The employee in the colorblind condition instead said, "and, importantly, I think the company sees us all first and foremost as human beings, encouraging us to treat each other the same". Appendix S1 contains the full testimonials and brochures.
Participants once again completed a measure of ingroup similarity (α = .73) and racial identification (α = .79), but did not complete measures of stereotypical self-descriptions or interest in stereotypical activities. In addition, they responded to a measure of trust and comfort (e.g., "I think I would trust the management to treat me fairly"; "I think I could 'be myself' at a company like CCG"; Purdie-Vaughns et al., 2008) toward the company setting on a 1 (Completely disagree) to 7 (Completely agree) scale (α = .95). All items were averaged to form a measure where stronger values indicate more trust and comfort.

| Results
Correlations between all measures are reported in Table 4.

| Primary analyses
Before proceeding with racial identification as a moderator in analyses, we checked and confirmed that racial identification did not

Analytic strategy
We used the same analytic strategy as in Study 1, but with only one dummy coded variable for diversity condition, with the colorblind condition as the reference group.

Ingroup similarity
As shown in Table 5 and Figure 3, an interaction emerged between racial identification and diversity condition, and all patterns matched those of Study 1. Inconsistent with Hypothesis 1a, participants who were strongly identified with their racial group reported lower ingroup similarity in the colorblind compared to multicultural condition. The statistical test for those weakly identified was not significant, but in contrast to Hypothesis 1b, it showed the same pattern as in Study 1, that participants selfreported higher ingroup similarity in the colorblind compared to multicultural condition. Table 6 shows simple effects analyses for the effect of condition on all dependent measures separated by weak and strong racial identification, as well as simple slope analyses.
Simple slope analyses also deviated from hypotheses but matched Study 1. Inconsistent with Hypothesis 2b, there was no relationship between identification and ingroup similarity in the colorblind condition. Also inconsistent with Hypothesis 2c, in the multicultural condition, stronger identification with the racial ingroup was associated with more ingroup similarity. Again, the multicultural condition showed the positive relationship expected of the colorblind condition, and the colorblind condition showed the null relationship expected of the multicultural condition.

Trust and comfort
As shown in Table 5 and Figure 4, an interaction emerged between racial identification and diversity condition, and the pattern matched that of the ingroup similarity measure. Participants who were strongly identified with their racial group trusted the colorblind company less than the multicultural company. However, those weakly identified trusted the colorblind company more than the multicultural company. Table 6 shows simple effects analyses for the effect of condition separated by weak and strong racial identification, as well as simple slope analyses.
Simple slope analyses also matched the ingroup similarity measure, showing no relationship between identification and trust in the colorblind company. In the multicultural company, stronger identification with the racial ingroup was associated with more trust and comfort.

Mediated moderation analysis
Using the PROCESS macro (Hayes, 2013)

| Discussion
Study 2 partially replicated findings from Study 1 and also demonstrated parallel findings on a measure of trust and comfort. However, trust and comfort did not mediate the interaction between diversity condition and racial identification on ingroup similarity. Taken together, these exploratory findings suggest that colorblindness sends a safety cue to weakly identified Asian individuals. It may also make them more comfortable asserting ingroup similarity, but this pattern was only statistically significant in Study 1 and should be interpreted cautiously. However, colorblindness creates discomfort for those strongly identified and also leads them to downplay similarity to their ingroup.

| G ENER AL D ISCUSS I ON
This research examined whether diversity strategies convey messages about how to express group identity in organizational settings, placing pressure on people of color to assert or to distance from their racial identity, depending on their levels of racial identification.
We predicted that strongly identified Asian and Asian American individuals would be resolute in their identity management irrespective of the presence of various diversity initiatives. However, that weakly

TA B L E 5 Hierarchical regression on ingroup similarity and trust and comfort in Study 2
Predictor Ingroup similarity Trust and comfort Step identified Asian and Asian American individuals would be more likely to assert (vs. downplay) their racial identity when exposed to a multicultural compared to colorblind organizational norm. Our findings did not confirm these predictions.
Two studies instead demonstrated that weakly identified Asian and Asian American individuals were more likely to go against the norm and assert similarity to their racial ingroup in a colorblind compared to a multicultural (and control) setting. However, the evidence Note: The first three rows show simple slopes (i.e., the effect of racial identification on the dependent measures separated by condition) for statistically significant interactions. The second set of effects shows the simple effects (i.e., the effect of diversity condition on the dependent measures separated by weak and strong racial identification). J-N refers to the Johnson-Neyman technique, which supplements the simple effects analysis by testing for the precise value of racial identification at which the differences by condition become statistically significant at p < .05. Cells with hyphens indicate that any difference across condition is not statistically significant within the observed range of racial identification.

Racial iden fica n
Colorblind Mul cultural Colorblind Mul cultural for this finding was somewhat inconsistent across studies. They also felt more trust and comfort in the colorblind setting. In contrast, strongly identified Asian individuals unexpectedly complied with the norms expressed by the diversity strategy, downplaying similarity to their ingroup in the colorblind relative to the multicultural (and control) setting. Further, they felt less trust and comfort in the colorblind setting. Thus, the colorblind context disrupted the standard identity assertion and distancing process for weakly and strongly identified Asian Americans and showed parallel patterns for trust and comfort in the settings. Trust and comfort did not mediate the effects on identity assertion and distancing, however, leaving open the question of why participants engaged in identity distancing and assertion.
The findings, while reliable across both studies for strongly identified participants, were unexpected. Why would strongly identified Asians distance from their identity as a function of colorblind (weaker identity focus) and multicultural (stronger identity focus) contexts? The strong interdependence characteristic of Asian cultures (Markus & Kitayama, 1991), which can lead to a focus on fitting in and conforming to situational expectations, may help explain this pattern. Those who are most strongly identified with their Asian ingroup may be the most attuned to cultural practices, and therefore the most likely to engage in identity-based conformity.
The findings for weakly identified Asians were less consistent. While the pattern of identity assertion was the same across studies, weakly identified Asians did not show a significant level of identity assertion in the colorblind relative to multicultural and control conditions in Study 2. Thus, it is unclear whether this pattern of findings is reliable. Nonetheless, one potential framework for understanding the findings is in research exploring the "queen bee" phenomenon. When women and people of color are weakly identified with their gender or racial group and expect discrimination in that workplace, they show reactance and distance from their identity rather than assert it (Derks et al., 2011(Derks et al., , 2015.
Because the diversity strategies were both framed as a strategy for achieving inclusion, we did not expect them to be sufficiently threatening to elicit reactance; however, future research should explore this possibility.
A final unexpected finding was that participants showed similar reactions in the multicultural and control conditions, despite some scholars suggesting that colorblindness is the default diversity strategy in the US context (Plaut, 2002). One possibility is that organizational norms in the United States are changing, with an increased focus on celebrating differences as part of their initiatives.
Indeed, there is an increasing demand for diversity training programs (Bezrukova, Jehn, & Spell, 2012), and harnessing the benefits of employee differences may be one component of these.

| Caveats and future directions
One important limitation of the present studies is that Asian and Asian American individuals are a heterogeneous group in terms of both nationality and ethnicity. There are some commonalities in the nature of the stereotypes and prejudice that Asian individuals face in a US context (see Cheryan & Tsai, 2007;Zou & Cheryan, 2017), but there is also a wide range of experiences across different ethnic groups that might impact their perceptions of organizational diversity strategies. Given the dearth of research on Asian and Asian American individuals' responses to diversity strategies, we considered this research an important step toward understanding their perceptions; however, future research would benefit from collecting larger samples of specific ethnic groups to understand potential divergences in perceptions.
Relatedly, the present findings may not extend to other groups that face different stereotypes or have a different history of discrimination. Indeed, some research has shown less clear evidence of strategic identity management among African Americans in response to social expectations (Sinclair, Hardin, & Lowery, 2006).
In spite of potential benefits of colorblindness for weakly identified Asian individuals in a US context, the broader research literature suggests that multiculturalism may be a more promising strategy overall (Gündemir et al., 2016;Purdie-Vaughns et al., 2008; but see Jansen, Otten, & van der Zee, 2015;Plaut, Garnett, Buffardi, & Sanchez-Burks, 2011;Purdie-Vaughns & Walton, 2011). Perhaps the most promising strategies will combine some components of both multiculturalism and colorblindness, ensuring that people's group identities are valued, while also still valuing individuality and acknowledging variation within groups.
Indeed, one approach that combines multicultural with equal opportunity rhetoric, known as multicultural meritocracy, retains the benefits of multiculturalism, but reduces its negative consequences on outcomes such as stereotyping (Gündemir, Homan, Usova, & Galinsky, 2017). Similarly, a type of colorblindness that celebrates interindividual differences, rather than subgroup differences, has shown similar benefits as multiculturalism (Gündemir et al., 2016).
Practically speaking, merging these approaches could involve celebrating a wide range of types of diversity-for example, having a cultural day for a specific ethnic group might make those who are weakly identified feel uncomfortable and put them in the spotlight.
However, having an optional diversity day that celebrates a range of groups, including individual-level characteristics such as personality, might prove more palatable. Nonetheless, it is important that attempts to broaden diversity do not make strongly identified racial group members feel that the initiatives have been diluted and are no longer genuine efforts (see Smith, Morgan, King, Hebl, & Peddie, 2012).
Finally, it will be important for future research to understand the consequences of strategic responses to diversity approaches. In particular, if strongly identified group members feel pressure to distance from their identity in a colorblind context, this could have implications for their feelings of authenticity in the workplace. People prefer to be seen in ways that are consistent with their own self-concept (Swann & Read, 1981), so downplaying an important identity could have negative affective consequences as well as hurt feelings of workplace fit (Harter, 2002;Schmader & Sedikides, 2017).

| Conclusion
Colorblindness and multiculturalism can have quite different consequences for Asian and Asian Americans of different identification levels, serving as a safety cue for some, but creating discomfort for others. Organizations intending to promote an inclusive environment for their workforce should be aware that one size may not fit all and that multiple strategies may be needed in order to create safe environments for a range of groups and individuals.

CO N FLI C T O F I NTE R E S T
The authors have no affiliations or financial interests that present a conflict of interest in the process of conducting and reporting this research. The manuscript adheres to the ethical guidelines specified in the APA and BPS Codes of Conduct. We obtained ethical approval through the University of Washington Human Subjects Division.

TR A N S PA R EN C Y S TATEM ENT
All materials and data files for both studies are available at https:// osf.io/53v8f/ .