A noted test expert, Mr. Rosner has discussed his research on the process by which SAT items are pretested. In brief, Mr. Rosner has discovered that pretest questions answered correctly more by whites than students of color are the ones then chosen for inclusion on the SAT exam.

Short Bio: Jay Rosner
Jay Rosner, the Executive Director of The Princeton Review Foundation, is a lawyer and an admission test expert, with a specialization in assisting historically excluded students on tests such as the SAT, GRE, MCAT and LSAT. In 2001 he testified as an expert witness on testing in the University of Michigan Law School affirmative action trial in Detroit. He is a consultant to KIPP schools and to Bob Moses' Algebra Project. He is based in the San Francisco bay area, but his work is national in scope. The Foundation he heads currently runs programs jointly with such national organizations as the Thurgood Marshall Scholarship Fund, the NAACP, and the Tavis Smiley Foundation and local organizations like the Asian Pacific Fund and Aspira of New York City. Jay holds a BA from the University of Pennsylvania, and a JD from Widener University. Prior to attending law school, Jay was a public high school math teacher for two years. He is the father of two daughters: one a recent college graduate, the other a college senior.
An excerpt from one of his latest articles is pasted below.
Preferences? On the SAT, They Happen to be White
_____________________________
___________________
I've been a critic of the SAT for over 20 years, particularly regarding its impact on African American, Latino and Native American students. Recently, the president of a well-known northeastern university e-mailed me this question: "What do you say to people when they ask about the quantitative section of the SAT I? Surely there is no racial preference when it comes to this subject." In response, I just show folks two SAT questions and a bit of additional data. I don't need to do anything overly technical. The only concepts I use here are percentages and subtraction.
Educational Testing Service (ETS) is the company that puts together the SAT (and the GRE, GMAT, etc.). Below are two SAT math questions that ETS pretested. ETS pretests every potential SAT question by putting it into an unscored section of the SAT so that ETS can try it out. In other words, ETS wants to see ahead of time how tens of thousands of students do on it before the question is considered for later use on a scored SAT section. Thus, ETS finds out in pretesting the overall percentage of students answering the question correctly, along with the percentage of whites, blacks, Latinos, men, women, etc., answering correctly.
---------------------------------------------------------
Question #1:
If v2x is an integer, which of the following must be an integer?
a) vx
b) x
c) 4x
d) x
e) 2x
Is there a racial preference in this question? It happens that 7% more BLACKS that whites answered this question correctly. I'm an SAT expert, and I don't know why.
----------------------------------------------------------
Question #2:
If the area of a square is 4x, what is the length of a side?
a) x
b) 2x
c) 4x
d) x
e) 2x
Is there a racial preference in this question? It happens that 11% more WHITES than blacks answered this question correctly. Again, I don't know why.
------------------------------------------------------------
If you had the above statistics from pretesting (which ETS had) and needed to choose one of these questions to use on the SAT, which onewould you choose? ETS selected and used the second question, and rejected the first question. My testimony in the University of Michigan Law School case documented how ETS systematically rejects for use virtually every pretested question that favors blacks, and selects for use many pretested SAT questions that favor whites by 20% or more. Question #2 above, with an 11% difference, is slightly less than the average white/black difference of 15% on the SAT math questions that I researched. And, note that 4 of the 5 answer choices are identical. The white/black test score gap is a cumulative result of the individual questions that are chosen. I have data on 240 math questions used on 4 real SATs - exactly 1 question "favored" blacks (by 3%), and all the other 239 "favored" whites (by an average of 15%). Does this sound fair to you? Someone saying, "Surely, there is no racial preference (on the math section)," is certainly expressing what is commonly accepted wisdom. It's just wrong. There is no obvious bias, no bias that I can show you in the content of these or virtually any other SAT question. But there is a massive, unconscionable white preference created through question selection. I believe that this is not done by ETS with a racist motive, but it is a consistent, reliable, predictable and foreseeable pattern on the SAT.
------------------------------------------------------------------------
The answer to Question #1 is c. To get the correct answer, you must substitute 1/2 for x. The answer to Question #2 is b. Squaring a side of length 2x produces the correct answer.
------------------------------------------------------------------------