The education bill signed by George W. Bush in early January includes the
provision that federal education dollars may not be available to states that fail to close
the achievement gaps between “minority and nonminority students, and between disadvantaged
children and their more advantaged peers.”
Just how big are these gaps? Consider the grade 4 math results from the 2000 National
Assessment of Educational Progress: Fewer than half of African American and Hispanic
students nationwide attained the “basic” level, compared with 80 percent of White students.
And only 46 percent of fourth graders who were eligible for the federal free or reducedprice
lunch program, compared with 79 percent of their ineligible (wealthier) counterparts,
reached the “basic” level.
These achievement gaps and the inequities in opportunity in which they are rooted have
repercussions far beyond elementary school; they are also at the core of a key dilemma in
college admissions. In the short term, an admissions policy that dictates the selection of the
most academically prepared students is likely to yield an entering class with few African
American, Latino, Native American or low-income students.
So how can colleges reward applicants who are well-prepared, while simultaneously assembling
a group of students who come from a variety of ethnic and socioeconomic backgrounds?
The admissions decision process is further complicated by the fact that academic
excellence is an attribute of individuals, whereas diversity can only be evaluated by considering
the entering class as a whole. How can the qualifications of individuals and the characteristics
of an entire class be evaluated simultaneously?
An interesting new idea for balancing the academic and nonacademic goals of admissions
policy has emerged from the testing arena. Researchers Peter J. Pashley and Andrea
E. Thornton from the Law School Admission Council (which makes the Law School
Admission Test, or LSAT) developed a computerized method of assembling an entering
class with certain prescribed characteristics.Their approach, outlined in a 1999 report called
“Crafting an Incoming Law School Class” is based on established optimization techniques
from the field of operations research.
In theory, an admissions officer could say,“I want to admit 500 students, of which at least
20 percent should be ethnic minorities, at least 25 percent should be low-income, and at least
35 percent should be from within the state. Given those conditions, I want the students with
the highest possible combination of test scores and grades.”
In a conventional selection technique, this goal might be attempted by ranking individuals
using index scores based on test performance and grades, with points added for the other
characteristics deemed desirable (for example,minority, low-income or state-resident status).
But a ranking process is not well-suited to the task at hand.There’s no easy way to assign the
right number of points to each attribute of interest, nor is there any assurance that the top 500
on the resulting list will have the desired characteristics when considered as a group.
Instead of ranking individuals, the new method evaluates sets of 500 candidates. The
computer then produces a list of the best group of 500 admits, based on the established
criteria.
Pashley and Thornton provided an example of their optimization technique, in which
they attempted to mimic the results obtained through the actual admissions process at a
particular (unidentified) law school. One of their goals was to produce a group of admits as
ethnically diverse as those actually chosen, but without using race as an explicit criterion.
Toward this end, they considered such factors as the percentage of minorities within the
applicant’s area of residency, the percentage of minorities at the applicant’s undergraduate
school, and the nature of the undergraduate campus environment—urban, suburban or
rural. (The researchers used applicant data files, along with information from the College
Board and the U.S. Census.) Given that various conditions involving these and other demographic
factors were met, an index combining LSAT scores and undergraduate grades was
to be maximized.
The results showed that there was an overlap of about 75 percent between the students
“admitted” using the computerized optimization approach and those who had been
admitted under ordinary procedures.The ethnic composition of the two overlapping groups
of students was quite similar.
To some, a selection method like this one is anathema—another instance of mindless
computerization of a complex process. But if this approach were to be adopted, human
involvement would be no less essential than in many existing admissions procedures.
Admissions personnel would still need to determine what factors were to be considered in
the decision process and how to measure them. Along with academic and demographic
characteristics, ratings of personal qualities, as reflected in letters of recommendation and
candidates’ statements, could be taken into account, as they are in more conventional
selection methods. Admissions policy makers would also need to establish target values for
each admissions criterion: the minimum acceptable percentage of students from low-income
backgrounds; the desired percentage of students from within the state, and so on. Finally, it
would, of course, be essential to provide for a review of the computerized listings by
admissions personnel, to allow for possible modification.
The researchers’ use of data on the degree of minority representation in applicants’
communities and schools may be of particular interest in California, where the quest for
student diversity is impeded by Proposition 209, which forbids the explicit consideration of
race or ethnicity in admissions to public universities.
And overall, the proposed approach
seems to be well-suited to the admissions objectives
of the University of California.
At a November conference at UC Santa
Barbara, UC President Richard Atkinson reaffirmed
the recommendation he made one year
ago—that achievement tests, and not the SAT I,
be used in UC admissions decisions. “Our goal
in setting admissions requirements,” he said,
“should be to reward excellence in all its forms
and to minimize…the barriers students face in
realizing their potential. In other words, to honor
both the ideal of merit and the ideal of broad
educational opportunity.”
Atkinson’s words serve to highlight the official
goal of the UC admissions process as expressed in a 1998 resolution by the UC Regents
—to enroll a student body that both “demonstrates high academic achievement or
exceptional personal talent, and…encompasses the broad diversity of cultural, racial, geographic
and socioeconomic backgrounds characteristic of California.”
The approach to assembling an incoming class suggested by Pashley and Thornton
embodies a recognition that evaluating each application separately may not be the best way
to assemble an ideal group of students. One argument for using such a procedure—at least
as a starting point—is that computers are simply more efficient at assembling groups with
specified characteristics than are human decision makers.The computerized approach also
makes it easier to evaluate the effects of shifts in admissions criteria that may be under consideration.
But perhaps the key advantage of the proposed selection process is that it forces policymakers
to translate both the academic and nonacademic admissions criteria into an explicit
set of rules, and, in doing so, to spell out the context in which test scores are to be judged.
In his speech at UC Santa Barbara, President Atkinson lamented the fact that “we will
never devise the perfect test—a test that accurately assesses students irrespective of parental
education and income, the quality of local schools, and the kind of community students live
in.” But it is not the “job” of tests to further the social policy goal of broadening educational
opportunity, and neither the SAT I nor any current or future test can be expected to do so.
The role of admissions tests is a much more limited one—to assess applicants’ academic
preparedness for college. It is admissions policymakers who must determine how best to
maximize academic excellence while increasing the diversity of the student body.
