Oh dear, maybe Putnam's effect isn't real at all. Silly that he spent years hiding it then.

Edit: Not a good meta-analysis. Not only is it not quantitative in meta-analyzing effect sizes, they write:

In addition, high levels of ethnic heterogeneity often go hand in hand with economic inequality and low levels of economic prosperity (Letki 2008, Phan 2008). As economic deprivation and inequality may mediate the observed relationships between ethnic heterogeneity and social cohesion, studies in which economic characteristics are included in the explanatory model can make a firmer claim that they are not reporting spurious effects of
heterogeneity.

Thus, they prefer studies that have a high likelihood of over-correcting.

The worst paper of this sort that I have seen in the one finding that diversity has a positive impact on group performance. For their pure groups, they used groups consisting only of Hispanics. For their diverse groups, they used mixed groups of Hispanics, Whites and/or Asians. There were no pure White or Asian groups. Their effect would seem to be explainable from the group differences in cognitive ability + a simple compositional model.

Unfortunately, I cannot remember what the paper is called.

When genomic data becomes more widespread, one can directly measure diversity in the genes for a group of persons and use that as a measure of racial diversity. Of course, it is possible to get a culturally homogeneous group from a racially diverse group, but this rarely happens in practice. Perhaps e.g. in Silicon Valley with Europeans and Asians, and at universities that do not use affirmative action (MIT?).