What are we supposed to get from that chart? It looks like a subset of the data performs differently from the set? That's not useful information. Via regression

What are we supposed to get from that chart? It looks like a subset of the data performs differently from the set? That's not useful information. Via regression to the mean the larger data set will always be more "normal" either lower or higher than the smaller data set. If you wanted a good comparison you'd have to be comparing an equal sized subset of the DJI that only included boards with low diversity. Junk conclusions from junk data. When it was actually studied with rigor, e.g. the study you linked to (https://jfin-swufe.springeropen.com/articles/10.1186/s40854-021-00265-x#Sec6 https://jfin-swufe.springeropen.com/articles/10.1186/s40854-...) "We found evidence for a positive effect of women on corporate boards on both measures of company performance—ROA and PER"