How Diversity Hurts Performance (Revisited)

It occurred to me that the previous post I had on Diversity and Performance relied on a 1998 literature review.

I decided to look for something more recent, and exploring the issue of diversity more broadly, and it had broadly similar results:

The Role Of Context In Work Team Diversity Research: A Meta-Analytic Review (2009)

"Our review indicated that approximately 60 percent of the direct effects reported in past research were nonsignificant for various diversity attributes. Among the remainder, 20 percent of the effects reported were significantly positive, and 20 percent were significantly negative...

We first examined the correlations between all types of diversity and performance and obtained a near-zero, nonsignificant result (r .01, k 117, 95% CI .02 to .00). This initial result corroborated past meta-analytic findings (e.g., Webber & Donahue, 2001). We then conducted separate analyses for relations- and task-oriented diversity and found a different pattern of results for each type of diversity. For relations-oriented diversity [Ed: Gender/Race/age], we found a very weak negative but significant relationship with performance (r .03, k 69, 95% CI .05 to .02). The relationship between task-oriented diversity [Ed: Function/Education/Tenure] and performance was also very weak but positive and significant (r .04, k 48, 95% CI .02 to .06)... functional background diversity was most positively related to performance (r .13, k 20, 95% CI .09 to .17) and that age diversity showed the most negative performance effect...

Gender diversity had a significant, negative effect on team performance in majority male occupational settings (r .09, k 12, 95% CI .12 to .05). The effect of gender diversity was significantly positive in relatively gender balanced settings (r .11, k 7, 95% CI .06 to .15)... The average correlation was significantly negative in majority white occupations (r .07, k 10, 95% CI .10 to .04) and positive in relatively balanced occupations (r .11, k 6, 95% CI .07 to .14)...

Task-oriented diversity showed more positive performance effects in majority male and white settings...

As predicted in Hypothesis 2a, relations-oriented diversity had a positive effect on performance in service industries (r .07, k 21, 95% CI .05 to .09). Inconsistently with Hypothesis 2a, however, in the manufacturing industry setting, the effect of relations-oriented diversity was negative (r .04, k 16, 95% CI .07 to .01) and interestingly, relations-oriented diversity displayed the strongest negative performance effect in high-technology industry settings (r .18, k 21, 95% CI .20 to .15)...

Among teams with low interdependence relations-oriented diversity was positively related to performance (r .08), and among teams with moderate and high interdependence, relations-oriented diversity was negatively related to performance...

Hypothesis 4 proposed that the negative effects of relations-oriented diversity would be strengthened in long-term teams. We found strong support for relations-oriented diversity: the categorical moderator model was highly significant (QB[1] 222.91, p .01). The performance effect of relations-oriented diversity was positive in relatively short-term teams (r .09, k 23, 95% CI .07 to .12) but became negative in more stable or long-term teams (r .14; k 43, 95% CI .16 to .12)...

Our findings indicated that the industry settings in which teams were embedded also had interesting implications for team-level diversity-based outcomes. We found that relations-oriented diversity had positive effects in service industry settings and slightly negative effects in manufacturing settings. In addition to the market competence perspective (Richard et al., 2007) discussed in developing our hypotheses, some additional considerations should also perhaps be taken into account. Service settings (for example, retail establishments and restaurants) involve front-line customer contact, and the costs of interactions based on negative categorizations are high in this context. Hence, firms embedded in these industries may engage in proactive diversity management efforts to address gender-, ethnicity-, and age-based issues in the workplace"


Unsurprisingly, on the whole diversity based on shallow demographic attributes (gender/race/age, which I shall henceforth call "diversity") was shown to be bad whereas real diversity in an attribute actually indicative of a different background, i.e. functional background, was shown to be good.

Some might interpret "diversity" sometimes having a U shaped relationship with performance (where "diversity" is low, increasing it reduces performance before it increases it) as evidence that we need affirmative action, but this assumes the gains from "diversity" will outweigh the harm of affirmative action (i.e. bringing in unqualified people to make the numbers is going to worse than negate the positive effects of "diversity"). Plus, the sample of studies all used real world teams, so low initial "diversity" is probably correlated with the task at hand, i.e. increasing gender diversity in a predominantly male engineering team may lead to worse performance because of the nature of engineering, whereas increasing gender diversity in a relatively balanced marketing team may lead to better performance because of the nature of marketing.

And given the spiel about "diversity" in tech it is interesting that this form of "diversity" was the most negative in high technology settings (they say this may be because white men have more training/mentoring/coaching/better performance reviews, i.e. visibility, but the results are what they are, and we don't know if they get more visibility because of discrimination or because of performance - which is actually what we would expect in the presence of affirmative action).

Also it is interesting that if you need to work with other teams, "diversity" is bad, and that the longer you work with your team, the more negative "diversity" is.