Read on Twitter
I've been teaching my students that you can never rule out statistical discrimination, so you cannot prove racism (taste-based discrimination). Now I think statistical discrimination is a cop out for inaction & the toleration of racial inequality. #ShutDownSTEM #BlackLivesMatter 
This thread is also related to @marreguant’ excellent thread on H0, your null hypothesis: if your null is that there is no racism, and you have statistical discrimination at your disposal as an explanation, you can never confidently rule out the null. https://twitter.com/MarReguant/status/1267944963129651200?s=20
For example, in @mariannebertra2 & @m_sendhil’s famous experiment, they found that callback rates for job interviews are 50% lower for fictitious candidates with Black-sounding names https://www.aeaweb.org/articles?id=10.1257/0002828042002561. Is this racism (aka taste-based discrimination)?
Even if the resume is identical but for the name, Black-sounding names may convey something outside the resume (we don’t know employers' full information set). https://scholar.harvard.edu/fryer/publications/causes-and-consequences-distinctively-black-names) argues exactly so much in that uniquely Black names are correlated with worse socioeconomics.
Therefore, if your null is that there is no racism (taste-based discrimination) but only statistical discrimination, it is almost impossible to disprove. #BlackLivesMatter
Instead of dedicating ourselves to the Sisyphian task of disproving statistical discrimination, we should evaluate policies & interventions that increase racial equality. #BlackLivesMatter
The framework of statistical discrimination may not be entirely useless. But, instead of being a cop out for inaction, it should be used e.g. as a way to understand factors involved in discrimination in order to come up with better policies. Thoughts @kerwinkcharles @NeumarkDN?
