Excerpts From the Klitgaard Report

The following are excerpts from the preliminary report on admissions by Robert E. Klitgaard '68, special assistant to President Bok.

116. The tests are biased at the right tail, in the predictive sense discussed above: a 700 is not a 700 is not a 700. But the direction most people anticipate. Test scores at the right tail often overpredict the later academic performance of women and minorities compared to other students. In fact, at the right tail women perform slightly worse than their scores would predict; blacks also perform worse. Blacks may obtain grades up to half a standard deviation below what their scores would have led one to expect. Examples from the College, the Law School, and the Kennedy School support this point.

117. In other words, an admissions policy at Harvard that took race into account in "adjusting" the scores of blacks to reflect their later performance would adjust the scores downward.

118. Equally startling is the underprediction of the academic performance of Jewish students. Using test scores without taking accout of ethnicity ends up discriminating in favor of blacks and against Jews--again, in terms of their later academic performance. These facts are not widely recognized in discussions of testing or admissions policies.

119. These biases at the right tall can be given a relatively straightforward statistical explanation in terms of "regression toward the mean." Since Jews have higher scores than blacks, a very high score is closer to the Jewish mean than to the white or black means. Statistically one would therefore expect on average that the regression effect would be greatest for blacks and least for Jews, so that the use of a common prediction equation for all three groups would overestimate the subsequent performance of blacks and underestimate that of Jews. Qualitatively, this explanation fits the observed facts. Quantitatively, its adequacy deserves further study, since other explanations are also plausible.

120. Affirmative action programs go beyond the purview of the present report, but a few observations may be relevant. In fields where high test scores seem to be a prerequisite for adequate academic performance, "affirmative action" may imply remedial education rather than lower standards or greater recruitment efforts in very limited pools.

122. Across universities and fields, black students tend to be about 1 to 1 1/2 standard deviations below whites on test scores and significantly lower on grades as well. Both blacks and whites may notice this pervasive disparity. In theory, this may be an undesirable result of affirmative action at universities like Harvard. If elite universities did not compete so heavily for blacks these students might attend slightly lesser institutions where they might compete as intellectual equals. Conceptually one might even imagine a ripple effect, where blacks would end up academically equal to whites at all but the very bottom institutions, as a result of not being admitted at institutions where they were not equally capable. The consequence might be a greater overall perceived feeling of equality by both blacks and whites. Of course, other considerations are also germane. Such conceptual models help one appreciate the effects of affirmative action across universities, not just from one university's viewpoint.

150. The optimal mix of career interests, personality characteristics, backgrounds, and so forth has apparently received little empirical or theoretical study. The idea of a "critical mass" of minority students or women, for example, seems to be supported by few empirical studies. How the average student benefits from studying at college with different numbers or proportions of blacks or Californians or athletes or socialists is not clear.