<p>So this is the key section (in my judgment of Atkinson’s remarks:</p>
<p>"In brief, the study shows that the SAT II is a far better predictor of college grades than the SAT I. The combination of high school grades and the three SAT IIs account for 22.2 percent of the variance in first-year college grades. When the SAT I is added to the combination of high school grades and the SAT IIs, the explained variance increases from 22.2 percent to 22.3 percent, a trivial increment. </p>
<p>The data indicate that the predictive validity of the SAT II is much less affected by differences in socioeconomic background than is the SAT I. After controlling for family income and parents’ education, the predictive power of the SAT II is undiminished, whereas the relationship between SAT I scores and UC grades virtually disappears. The SAT II is not only a better predictor, but also a fairer test insofar as it is demonstrably less sensitive than the SAT I to differences in family income and parents’ education."</p>
<p>Finally, what they found was that the writing SAT II was a better predictor of first-year college performance than the SAT I. In other words, they could throw out the SAT I altogether, keep the graded writing sample, and do a better job of predicting college performance than if they threw out the writing sample and kept the SAT I.</p>
<p>The new writing sample is virtually useless - because (if it is like the writing SAT II), it works.</p>