And also separating out Indian, Chinese, and Other Asian. The majority of many tech workforces are broadly Asian, along with over half of the global population. It doesn't make sense to lump them all in together. It may be acceptable in other industries in the US where Asian representation is in line with the US standard of ~6%, but certainly not in tech. If a single bucket accounts for more than half of your entire sample, and that bucket can trivially be broken down further into sub-buckets, then you should absolutely do so.

This is something that tech in particular just keeps getting wrong over and over. E.g. look at the 2019 Google diversity report [1] -- on page 52 it shows that 55.5% of new tech hires in the most recent year were Asian, vs 0.7% for the smallest category (Native American). Break out that 55.5%, please!

[1] https://kstatic.googleusercontent.com/files/25badfc6b6d1b33f...

They do in their model. Being Asian carries with it the worst relative odds of being hired, though it is not responsible for the overall negative non-white male coefficient (Black and Hispanic applicants also have negative coefficients in the logit)

Seems to make sense relative to each other. D&I initiatives often times target Black and Hispanic candidates, but not Asian since they are not underrepresented.