<p>
If you are looking for undeniable evidence that purely race-based AA occurs, I agree that there is none to be found. However, most people are looking for some sort of model to bring reason to the process. A race-based AA model fits the available data extremely well. Perhaps a better model will be developed, and five years down the road we will all reconvene and laugh about how silly it was to believe that race-based AA existed. But until a better alternative is presented, I think most people here will choose this theory because quite frankly it describes reality pretty well.
</p>
<p>Agreed. Although I think that an adcom bias favoring uncommon applicants (such as URM who are not disadvantaged due to other factors) is not a terrible theory.</p>