Emil O W Kirkegaard

similar methodological issues as audit studies that claim to have found racial discrimination.) More directly you have the replication of Bertrand and Mullainathan I mention in my post about white supremacy, which strongly suggests the effect in that study was really class-base

discrimination. To be clear, my claim is not that this evidence for the role played by statistical discrimination and other kinds of taste-based discrimination beside racial is *conclusive*, only that there is plenty of studies that found effect sizes for this stuff that could

explain a large share of the effects found by audit studies that claim what they found is racial discrimination when really their design doesn't allow to rule out statistical discrimination and class-based discrimination.

On the main problem identified by Heck though, I think there is a guy who devised a method which he claims solves it (was it Neumark?), but I haven't read his paper yet.

I am confused. Certainly evid for "stat discrim" exists. I do not see how evidence for said *existence* constitutes any explanation for audit studies, which personal info constant. I hope next tweet conveys why I don't follow your argument

Me: discrim in audit studies replicate* You: But much is explained by stat discrim Me: Got evidence? You: Lots of evid. of stat discrim Me: But your claim was not "there is evid of stat discrim." Your claim was "stat discrim explains audit studies." I still see no supporting evid

* To get to the point that we agree that that orig statement was actually true, we had to dance through @pnin1957 's tweets seeming to contest that, but which we now agree was simply a different point that did nothing to contest "discrim in audit studies replicates."

Okay here is my reasoning. You have plenty of audit studies that find discrimination, but their design usually doesn't allow you to tell whether it's racial discrimination, statistical discrimination or class-based discrimination. Now, we also have plenty of evidence that there