By Richard Arnold and Jeremy Beer
Anyone and everyone living in Austin is by now quite familiar with the
controversy that raged for approximately two weeks on and off the UT campus
following remarks concerning race and academic achievement made by UT Law
Professor Lino Graglia. Legislators called for Graglia's head, shame-faced
administrators apologized profusely to anyone who would listen, angry community
organizations issued forth thundering denouncements, leftist campus groups
emulated their '60s heroes with grievance lists, faculty signed joint statements
condemning Graglia's retrograde views, Student Government president Marlen
Whitley joined with two other black students in filing a racial harassment
complaint against Graglia, and on and on, ad nauseaum if not ad infinitum.
The psychology of that response is in itself interesting, and could bear,
we believe, a sustained and detailed analysis. For there were two questions
that the reasonable listener immediately asked: (1) What exactly did Graglia
mean?, and (2) Is it true? With few exceptions, neither question was posed
by the media, and neither was addressed by Graglia's detractors. So here
we address these questions, fully aware that in discussions of race and
ethnicity, honesty is the best way to make enemies.
Graglia's statement that "Blacks and Mexican-Americans are not academically
competitive with whites in selective institutions" can be taken in
at least three ways: (1) Without racial preferences in admissions, proportionately
fewer blacks and Hispanics will be admitted to selective institutions than
will whites; (2) given the same objective qualifications, blacks and Hispanics
will do worse than whites at selective institutions; (3) Among students
admitted to selective institutions under current affirmative action practices,
blacks and Hispanics fare significantly worse than do whites.
Only No. 2 can in any way be considered controversial, for as we will show,
Nos. 1 and 3 are uncontested empirical facts -- in fact, they refer to the
very same empirical facts. So did Graglia's statement have meaning No. 2?
Not at all, as later clarifications by Graglia made clear. In fact, Graglia's
detractors evidently didn't think he meant No. 2 either -- at least no one
made it clear that they thought No. 2 was what Graglia meant; in any case
No. 2, as we will show, is false.
So at least part of the controversy must have had something to do with Nos.
1 an 3, which, as we noted above, essentially refer to the same phenomenon.
Is it true that blacks and Hispanics are less academically qualified - on
average, by conventional, objective standards - than are whites to attend
top schools? Without doubt. The mean Verbal and Mathematical SAT scores
for white college-bound seniors in 1995 were 448 and 498, respectively.
The comparable scores for blacks were 356 and 388. It adds up to an average
SAT score gap between blacks and whites of 202 points. This is down from
258 in 1976, but is still very substantial.
The Hispanic-white difference is typically between one-half and two-thirds
the magnitude of the black-white difference. Comparisons of Hispanic and
white students who took the LSAT in 1992 and the GRE in 1987-88 offer a
typical illustration. On the LSAT, the average score of Latinos was one
"standard deviation" below the average score for whites, which
means that the average Latino student had a score equal to the 16th percentile
for whites. On the three sections of the GRE, the difference between whites
and Hispanics was less but still substantial, anywhere from one-half a standard
deviation to three-quarters a standard deviation.
Of course, selective schools sample from the upper end of the distribution
of SAT scores. What does the racial gap look like there? It is magnified.
103, 872 black students took the SAT in 1995. 107 of them had Math scores
of 750 above, and 184 scored at least 700 on the Verbal section. Compare
those numbers to those of Asian-American students. Of the 81, 514 Asians
that took the SAT in 1995, 3,827 scored at least 750 on the Math section
and 1,476 scored at least 700 on the verbal.
We can look at the numbers in terms of percentages: 1.7% of blacks had scores
of 600 or more on the Verbal SAT. For whites, the percentage was 9.6%, and
for Asians 10%. The differences are magnified on the Math SAT. 2% of blacks
had scores of 650 or better, compared to 13.4% of whites and a whopping
25.8% of Asians.
Even if we restrict ourselves to those students actually admitted to selective
institutions, the SAT gap between blacks and whites is large. Among elite
institutions the gap is smallest at Harvard, which generally gets its pick
of the top scoring black students in the country. The gap between black
and white SAT scores is 95 points. At Princeton it is 150 points. At Stanford
it is 171, Rice 271, and Berkeley 288 (all of these data are from the freshmen
classes of 1992). The median gap between Hispanic and white students at
26 selective schools was 129 points.
When lucid, most everyone acknowledges racial and ethnic differences
in academic achievement. But, the critics say, these differences are not
as real as they may seem. The so-called "objective" tests are
biased against minorities. They are the result of poverty. And discrimination.
And don't forget disparities in school funding. It is rarely acknowledged,
but the data exist with which to test these criticisms. They all fail miserably
to account for the racial/ethnic gap in academic achievement.
The discrimination hypothesis is inherently unfalsifiable; discrimination
is held to be at once subtle and pervasive. It affects every measure we
use; therefore it is itself hopelessly immeasurable. We can only say that
given the radical decline of white racism in recent years (at least as suggested
by survey data), we do not think that discrimination is a plausible cause
of more than a small part of the entire racial/ethnic gap in academic achievement.
A quite different argument holds that, sure, there are real differences
in academic achievement between ethnic and racial groups, but this only
reflects differences in the level of poverty among these groups. Substitute
"school funding" for "poverty" in that sentence and
you have another argument. But does poverty play a causal role in academic
achievement? Does school funding?
Let's take school funding first. Sociologist James Coleman, as lead researcher
for a U.S. government study of this and related matters, was the first to
look at this idea in a systematic, empirical manner. The Coleman Report,
as it came to be called, found that differences in school funding played
only a minuscule role in determining differences in academic achievement.
Subsequent studies - even studies by committed Marxists like Christopher
Jencks - have found the same thing. There is no causal relationship between
the amount of money a school spends per pupil and those pupils' academic
achievement.
The effects of poverty are a little trickier to study. There is no doubt
some threshold level of poverty exists, below which academic potential is
seriously thwarted. Kids growing up in the most wretched socioeconomic conditions
are at an overwhelming disadvantage. But when we talk of poverty in America
we are, except in rare cases, talking about an absolute standard of living
that is quite good compared to what was considered "poverty" just
fifty years ago. Wretched poverty just is not common in America.
With that caveat, may we safely conclude that differences in socioeconomic
status can not account for more than a tiny fraction of race differences
in academic achievement? It appears so. Many studies of this question have
been conducted, and the results strongly support this conclusion, but just
a few statistics are needed to drive the point home. Using 1995 numbers,
the average Verbal and Math SAT scores of blacks who came from families
with incomes of $70,000 or more were 407 and 442. The average Verbal and
Math SAT scores of whites who came from families with incomes of $10,000
or less were 409 and 460. Thus, poor whites had total SAT scores averaging
20 points higher than those of privileged blacks.
Now for the "test bias" charge, so often heard. The critic
employing this argument generally means something like this: academic achievement
tests like the SAT, GRE, LSAT, etc., simply don't measure what is important
for minorities (i.e., blacks and Hispanics) with regard to school achievement.
That is, the tests don't predict future performance as well for these minorities
as they do for whites. This is the crux of the test bias criticism: differential
validity.
Dozens of studies have investigated this issue. There is remarkable consensus
among them: there is no evidence that tests like the SAT are worse at predicting
minority academic achievement than academic achievement for whites. (The
interested reader should consult Arthur Jensen's Bias in Mental Testing
for the most thorough treatment of the subject in the literature.) This
fact has an important consequence, as we shall see.
Around the time of the Hopwood decision, former UT President Robert Berdahl
published an open letter in The Daily Texan which exemplified the profound
ignorance (or perhaps disingenuousness) concerning the validity of scholastic
aptitude testing displayed by many of the proponents of race-equity admissions
policies. Berdahl informed the public that UT has two diversity-related
goals: increased minority enrollment, and increased minority retention.
What Berdahl failed to realize (or admit) is that, absent a radical change
in the UT applicant pool, these goals are mutually exclusive.
When a school uses different minimum criteria for admission for members
of different identifiable groups, group differences in academic performance
are guaranteed to follow. The simple reason for this is that the traditional
criteria for college admission - high school performance (GPA or class rank)
and SAT scores - are valid predictors of college success. "Validity"
simply means that those who perform well on the predictors succeed in college
at a higher rate than those who perform poorly on the predictors; this much
should be obvious. Perhaps less obvious is the fact already mentioned: these
predictors are equally valid for members of different groups.
Thus, for example, black students and white students who enter UT with high
school GPAs of 3.9 and SAT combined scores of 1500 earn similarly high GPAs
and graduate at the same (high) rate from UT. For the same reason, blacks
and whites with equally low high school GPAs and SAT scores perform equally
poorly in college. Unfortunately, minorities and non-minorities differ,
on average, on these indicators. Though this causes policy-makers great
distress, the objective observer must conclude that realization of Berdahl's
dual goals is an actuarial impossibility. Use uniform admissions criteria,
and racial/ethnic diversity suffers; use group-differentiated admissions
criteria, and group differences in scholastic performance result.
Critics of traditional admissions criteria often point to examples of individuals
who, though they had only mediocre scores on the admissions criteria, nevertheless
have distinguished themselves academically. They also cite the many examples
of brilliant failures. That these examples abound cannot reasonably be disputed.
However, the purveyors of these anecdotes fail to see the implications of
the fact that these cases are equally common among minorities and non-minorities,
and furthermore that these cases cannot be reliably predicted.
We are left, then, with a very straightforward situation; though some low-scorers
succeed, and some high-scorers fail, these people are relatively uncommon,
occur in all groups, and are difficult to identify. If we lower the standards
for admission for one group, we may catch a few low-scoring individuals
who are ultimately successful, but their numbers will be overwhelmed by
those who are well predicted; and for low-scorers the predictions are ones
of failure. The result of lowering the standards for a group would be that,
due to the high failure rates of those who do not meet the standard admissions
criteria, but do meet the lower preferred-group standard, the average level
of success of that entire group is diminished, which is exactly what we
have witnessed for the past 30 years.
This is not idle speculation. Schools in which racial/ethnic groups differ
most on the SAT have the largest discrepancy in drop-out rates between those
groups. At Harvard, where the SAT gap between black and white students in
1992 was 95 points, 5% of blacks and 3% of whites dropped out of school.
At Columbia, where the gap between average scores for blacks and whites
on the SAT was 182 points, 25% of blacks dropped out vs. only 12% of whites.
And at Cal-Berkeley, where the gap is an astonishing 288 points, fully 42%
of blacks dropped out of school vs. only 16% of whites. And, you should
not think that these drop-outs are drawn randomly from their respective
groups. The drop-outs, black and white, disproportionately consist of those
whose high school grades and SAT scores were among the lowest for admitted
students. The correlation is strong and robust.
Whether people like it or not, the SAT and performance in high school are
by far the best predictors of academic success in college. That these predictors
are equally valid for all groups, and that groups differ, on average, on
these predictors may cause people some discomfort. However, if issues regarding
race, scholastic admissions and scholastic performance are to be debated,
the debate must be grounded in reality. These facts do not provide the answers
to our problems, but without them progress toward a reasonable solution
is impossible.
Although we believe a careful reading of our arguments should obviate
the need for this section , experience tells us otherwise. For starters,
the facts concerning average group differences tell us nothing about individual
merit. There exist many black and Hispanic applicants whose grades and SAT
scores surpass those of the majority of white applicants, while there similarly
exist many white applicants who are not competitive with the majority of
black or Hispanic applicants. Thus, group differences are ones of average
differences, not absolute differences.
It is ironic that the people often labeled as racist are those who merely
point out these facts, while their accusers often make statements with far
worse racist implications, though they undoubtedly are unaware of their
sins. For example, in the wake (as people like to call it) of the Hopwood
decision discouraged opponents asserted that the decision would produce
a "lily-white" student body. In our opinion this assertion is
more racist (and false) than anything the supporters of the Hopwood decision
said, for it implies that no minority applicants would be accepted on the
basis of merit. This is not only false, but outrageous, as this year's enrollment
statistics indicate. The fact that nobody at the time seemed to appreciate,
or at least didn't want to discuss, the implications of the "lily-white"
assertion is the best evidence that a great many people just aren't aware
of the sorts of issues we've discussed in this article.
One last point on this - it does not follow from the fact that group differences
in academic achievement exist at present that they will continue to exist
in the future. It does however seem to us that if we don't honestly confront
the existence and respect the practical importance of these differences,
it is unlikely that we will ever do anything to attenuate them, whatever
their source.
Was Graglia right? In implying that, today, blacks and Hispanics would
not be proportionately represented at elite academic institutions in the
absence of racial preferences, yes he was. Undeniably. In asserting that
this is probably the result of deep and imbedded cultural differences between
groups, he may have been right. There is some evidence to support that view,
but at present we just can't say.
However, we can say, with complete confidence, that racial/ethnic differences
in academic achievement, especially those between blacks and whites, cannot
be attributed to poverty or disparities in school funding. It is also clear
that traditional, objective admissions criteria are not "biased"
against minorities; they predict equally well for blacks, whites, and Hispanics.
We can not rule out the effects of a subtle and pervasive "atmosphere"
of racial discrimination, but neither can we (or anyone else) affirm its
existence and/or effect on academic achievement. Finally, race/ethnic differences
on predictors that are equally valid for those racial or ethnic groups will
inevitably mean differences in minority retention, unless, of course, race-blind
admissions policies are used, in which case retention rates will be the
same but races/ethnic groups will not be proportionately represented.
We submit that any reasonable discussion must take these facts under consideration.
Richard Arnold and Jeremy Beer are graduate students in psychology.