“Page 42:
Whatever the combination of sources of test bias might be, genuine bias against a minority will show up in a way that leaves no room for doubt: It will under-predict the test taker’s performance in the classroom or on the job.
Whether predictive validity is the same for different groups can be subjected to rigorous statistical scrutiny, and it has been, repeatedly. The results are unambiguous, whether the thing being predicted is grades in school or performance on the job. The major tests do not under-predict the performance of Africans or Latins.”