The downside of Affirmative Action and what can be done about it

Avoiding academic mismatch increases graduation rates dramatically. [Link]

There is now increasing evidence that students who receive large preferences of any kind—whether based on race, athletic ability, alumni connections or other considerations—experience some clear negative effects: Students end up with poor grades (usually in the bottom fifth of their class), lower graduation rates, extremely high attrition rates from science and engineering majors, substantial self-segregation on campus, lower self-esteem and far greater difficulty passing licensing tests (such as bar exams for lawyers).
The most encouraging part of this research is the parallel finding that these same students have dramatically better outcomes if they go to schools where their level of academic preparation is much closer to that of the median student. That is, black and Hispanic students—as well as the smaller numbers of preferentially admitted athletes and children of donors—excel when they avoid the problem of what has come to be called "mismatch." Jareau Hall's experience is representative of scenes that play out every fall at selective schools across the country. Black and Hispanic high-school seniors are actually more likely than similar whites to aspire to careers in science and engineering (which, along with technology and math, make up the so-called STEM fields), as first demonstrated by Dartmouth psychologist Rogers Elliott in 1996, and since confirmed in other studies. Tens of thousands of minority students receive preferences to attend schools where they feel overwhelmed, especially in STEM classes. As a result, the studies have found, they switch from science courses and migrate to other fields that, if not actually easier, are at least graded less harshly and are less sequential in their teaching. The end result: Whites are seven times more likely than blacks to go on to get doctorates in STEM fields.
Mr. Elliott's research was ignored in the widely heralded 1998 book "The Shape of the River," in which former Princeton President William Bowen and former Harvard President Derek Bok argued that the effects of affirmative action were unambiguously good. But a few years later, University of Virginia psychologists Fred Smyth and John McArdle got access to the data set created by Messrs. Bowen and Bok, which included data on academic aspirations, admissions and performance for tens of thousands of students at 28 major universities. Using the same data, they showed that students receiving large preferences were nearly 80% more likely to complete STEM degrees if they avoided mismatch by going to a less-elite school.
Research on law schools by one of us (Richard Sander)—hotly disputed by some scholars when it was published in 2005 by the Stanford Law Review and now confirmed by economist Doug Williams—found that mismatch essentially doubled the rate at which blacks and Hispanics failed bar exams. Under existing preference policies, only one in three blacks entering law school graduates and passes the bar on his or her first attempt (compared with two in three whites). Simply by reducing mismatch, we could get this ratio up to one in two.
This same dynamic turned up when several leading educators wanted to find out why so few black students went on to become professors. Funded by the Council of Ivy League Presidents, sociologists Stephen Cole and the late Elinor Barber surveyed thousands of young African-American students entering a broad cross-section of selective schools. The 2003 Cole-Barber book, "Increasing Faculty Diversity," concluded that large racial preferences and the ensuing mismatch led directly to lower grades and diminished intellectual self-confidence. They found that promising young black students who wanted to become professors abandoned their academic aspirations in droves, while similar black students who weren't mismatched were far more likely to stay the course.
The work ties into a second major finding, that of social mismatch. The central legal justification for using race in university admissions is the need to produce a healthy learning environment by fostering diverse classroom viewpoints and cross-racial friendships. But it turns out that these effects are also heavily influenced by the presence of large preferences. Economics professor Peter Arcidiacono and his colleagues at Duke University found in a 2011 study that students were much more likely to become friends with classmates they saw as academically similar to themselves. Students with large preferences were more likely to self-segregate and find themselves socially isolated. The reason wasn't racism. At Duke University, for example, large numbers of whites and blacks formed friendships at the outset of college. But for those with large academic gaps, the friendships atrophied. Using their multischool results, Mr. Arcidiacono and his colleagues concluded that smaller preferences at the most selective schools would tend to increase both the likelihood and the number of cross-racial friendships at elite schools in general, despite declines in the numbers of black and Hispanic students at the most elite schools.