Diversity Thresholds: How social norms, visibility, and scrutiny relate to group composition

Introduction

As institutional Diversity, Equity, Inclusion and Justice (DEIJ) efforts have become increasingly prevalent, public scrutiny (critical attention paid to particular behaviors) has increasingly been recognized as an effective tool to encourage such efforts. The #OscarSoWhite critique of the Academy of Motion Picture Arts and Sciences is one such example. Although previous scholarship has shown that “diversity” is interpreted differently across individuals and contexts, this study aims to understand how groups concerned with diversity-related reputational threat identify thresholds for “sufficient” diversity and respond to external scrutiny.

The authors theorize that when forming “diverse” groups, individuals aim to match the average levels of diversity among peer groups to avoid associated scrutiny – particularly for highly visible groups. Additionally, the authors highlight findings that individuals find lofty goals highly motivating but pause their efforts once salient goal thresholds are reached, suggesting that peer groups will coalesce around a similar threshold. Thus, intuitively, additional underrepresented members within the group will be added at lower rates once the peer group diversity threshold is met. Further, groups will not act on their diversity goals unless there is risk of scrutiny and visibility. If these hypotheses are accurate, there should be an overabundance of highly visible peer groups with identical levels of diversity. To test these hypotheses, the authors conduct six analyses. This study demonstrates the significance of external factors such as peer group social norms and external scrutiny in setting – and expanding – expectations for diversity in hiring and group composition. In a time of increasing DEIJ investment, this study offers critical insight into implicit barriers and potential accelerators to this effort.

Dr. Edward H. Chang, PhD, is an Assistant Professor of Business Administration in the Negotiation, Organizations, and Markets Unit at Harvard Business School. Dr. Katherine L. Milkman, PhD, is the James G. Dinan Professor of Operations, Information, and Decisions at the Wharton School and Professor of Medical Ethics and Health Policy at the Perelman School of Medicine of the University of Pennsylvania. Dr. Dolly Chugh, PhD, is an Associate Professor of Management and Organization at New York University Stern School of Business. Dr. Modupe Akinola, PhD, is an Associate Professor of Management and Faculty Director of the Sanford C. Bernstein & Co. Center for Leadership and Ethics at Columbia Business School.

Methods and Findings

The authors first test the hypothesis that group diversity levels will coalesce around the peer group average (the descriptive social norm), and that this effect will increase for high visibility companies, by examining gender diversity among companies boards within the S&P 1500 and higher profile S&P 500 companies in 2013. This analysis found that there were significantly more (+12%) boards with exactly two women and significantly fewer (-8%) boards with no women than would be expected if gender diversity was randomly distributed across all boards. This pattern of statistically significant cluster of boards around the contemporary descriptive social norm held true across 12 years of historical data. Among higher visibility S&P 500 companies this effect was exaggerated, with 45% more boards with two women and 45% fewer boards with zero women than would be expected in a random distribution. More specifically, an ordinary least squares regression showed that the level of media attention received among S&P 1500 companies in 2012 significantly predicted the overabundance of exactly two female board members in 2013. Notably, the authors did not see any clustering effect for race – a diversity trait not publicly scrutinized during this time period.

Second, the authors test the hypothesis that additional diverse members would be added to groups at a slower rate once the group has reached the peer group diversity threshold, and whether this varied based on the visibility of the group, by analyzing additions to company boards among the S&P 1500 from 2004 to 2013. The authors found that company boards that already included at least 2 women were less likely to add an additional female member – and this trait was the primary predictor for that behavior. This pattern was exaggerated among the more visible S&P 500 companies and those that received greater media attention in the previous year. Notably, the authors did not find a clustering effect for racial diversity – a diversity trait that was not scrutinized – in this time period. The authors also created an online experiment (N=479) to replicate the conditions of this study in a controlled setting, where they could vary the number of women already on a board and control for the quality of potential new candidates. They found that individuals were less likely to add additional female candidates to boards with two existing female members, regardless of the quality of the candidate.

Next, the authors created an experiment among 556 business students to test the salience of social norm thresholds and threat of scrutiny in a controlled setting. They asked students to select an additional speaker for a hypothetical conference and experimentally varied the gender diversity threshold and level of scrutiny. When gender was scrutinized, participants significantly preferred female candidates if they believed they had not met the social norm for gender diversity. However, when gender was not scrutinized, participants displayed no preference for female candidates, regardless of whether they had achieved the social norm for gender diversity. The authors also replicated this experiment in an online setting (N=200) and added monetary stakes for participants. They found that the online results mirrored the in-person study: when gender was scrutinized, participants significantly preferred female candidates if they were below the diversity threshold, even though the female candidate was more expensive to recruit. When gender was not scrutinized, participants displayed no preference for female candidates, regardless of the social norm.

Finally, the authors created an online experiment (N=603) to test whether diversity thresholds also applied to race and group visibility had a similar effect as scrutiny. Participants were significantly more likely to prefer Black candidates if they believed they were below the diversity threshold, even though they received lesser financial compensation for selecting the Black candidate. Further, participants were significantly more likely to prefer Black candidates if they believed their decision was highly visible.

Conclusions

This study demonstrates the relevance of peer-defined diversity thresholds,  visibility, and anticipated scrutiny to realize group diversity. These findings explain the clustering of group diversity around diversity thresholds (a phenomenon they term “twokenism”) and variance in the degree of clustering by group visibility and threat of scrutiny. Practically, this research identifies conditions in which minoritized individuals are more or less likely to be selected for groups and points toward potential strategies to encourage continued diversification. 

These strategies could include employing public scrutiny or idealized social norms to motivate group diversification as an alternative to emphasizing individual behaviors like hiring bias. However, the authors also note several important caveats to their proposed strategies, such as the greater effectiveness of positive attention for high performing groups rather than punitive reinforcements for poor performers. To overcome the tendency to stall diversification once the peer-defined threshold is met, scrutiny should be maintained until more ambitious goals are reached. Relevant social norms should also be shifted from peer group averages (descriptive norms) to ideals (injunctive norms) in order to establish more ambitious diversity targets. 

The authors point to two limitations of this study. First, they only examined the behavior of individuals, and it’s reasonable to expect that group decision-making is more complex. Secondly, it is possible that their findings do not translate to all external settings. The authors also identify several potential areas for further research to extend the current study, including: investigations about ways in which diversity-related norms (including processes for shifting descriptive norms)can hurt rather than help minoritized groups; relevance of these findings for other organizational contexts; the specific psychological mechanisms underlying the effect of descriptive norms and scrutiny; and interventions that decrease the salience of descriptive norms in favor of injunctive norms.