You are on page 1of 12

In this chapter I discuss evidence, tirst, on graduation rates and, next, on

grade-point averages (GPAs, tor under-represented ethnic group (LPLG,


and non-LPLG students in LS universities. I then go on to reviev and
analyze studies ot an issue vith important implications tor claims made by
critics ot attirmative action (AA, in admissions to LS higher educational insti-
tutions: the ettect ot collegeuniversity selectivity on student academic
pertormance.
URLG and non-URLG graduation rates
1here is much more evidence available on graduation rates at the undergrad-
uate level than at the graduate level, and the evidence is more plentitul tor
Atrican Americans than tor members ot other under-represented groups.
Bunzel (1988, vas one ot the tirst to drav attention to the relatively lov
graduation rates ot LPLG students at major LS universities. Bunzel cited
tigures tor the Lniversity ot Calitornia at Berkeley (LC-Berkeley, in the
1980s, according to vhich 66 per cent ot White students could be expected
to graduate vithin tive years, vhereas the corresponding tigure tor Hispanic
Americans is 41 per cent and tor Blacks 27 per cent.
1hernstrom and 1hernstrom (1997: 391, 1able 2, have provided more
comprehensive data on time trends ot college attendance as vell as graduation
rates by Blacks and Whites in the LS. During the three decades trom 1963 to
1993 years in vhich preterential admissions via AA to Black students became
videspread the percentage ot Blacks attending colleges rose trom 13 per cent
to 43 per cent, as the percentage ot Whites increased trom 26 per cent to 33 per
cent. 1he percentage ot Black students actually completing tour years ot college
rose only trom 7 per cent to 13 per cent, as compared to an increase trom 13
per cent to 26 per cent tor Whites. 1his implies that the traction ot undergrad-
uate students actually completing tour years ot college tell trom 43 per cent to
34 per cent tor Blacks, vhile remaining close to 30 per cent tor Whites.
1
Another body ot evidence highlighted by 1hernstrom and 1hernstrom
addresses ditterential pertormance by Blacks and Whites on the Scholastic
Aptitude (or Assessment, 1est (SA1,, the standardized test videly utilized by
11 Affirnative action and
acadenic perfornance
in US universities
2004 Thomas E.Wei sskopf
admissions ottices as an important indicator ot the academic qualitications ot
college applicants.
2
1hat Blacks (and Hispanics,, on average, score less highly
on the SA1 than Whites (or Asians, has been a vell-knovn and persistent
tact since such data vere tirst collected.
3
1here is also considerable evidence
that graduation rates among students at any given college or university are
positively correlated vith the students` SA1 scores especially at relatively
competitive universities.
4
1hernstrom and 1hernstrom (1997: 4069, esp.
1able 9, tocus attention on the relationship betveen the mean SA1-score gap
and the graduation-rate gap betveen Blacks and Whites accepted into thir-
teen tour-year colleges at elite universities vhere AA is practiced. 1he mean
SA1-score gaps vary trom 93 at Harvard to 288 at LC-Berkeley, the corre-
sponding graduation-rate gaps vary trom 2 per cent at Harvard to 26 per cent
at LC-Berkeley.
3
As 1hernstrom and 1hernstrom emphasize, there is a clear
positive correlation betveen the size ot the mean SA1 gap and the size ot the
graduation-rate gap betveen Black and White students.
1he mean SA1 score ot entering tirst-year undergraduates serves as a rough
proxy tor the degree ot selectivity as vell as the degree ot competitiveness ot
LS universities. 1hernstrom and 1hernstrom`s data reveal also a signiticant posi-
tive correlation betveen the degree ot selectivity ot an elite university and the
average rate ot graduation tor both Black and White students. 1his suggests that
the greater success ot more elite universities in graduating Black students may
have something to do vith their quality as vell as vith the smaller mean SA1
gaps ot their entering classes a point that I vill pursue belov (see pp. 1637,.
lurther evidence on the relationship betveen SA1 scores, college selectivity,
and academic pertormance has been compiled and analyzed by Datcher Ioury
and Garman (1993,. 1hey utilized a rich database ot college-goers trom a repre-
sentative nationvide sample ot male high-school graduates ot 1972, vhose
college choices span a much broader range than the elite ones highlighted by
1hernstrom and 1hernstrom. 1able 11.1 summarizes Datcher Ioury and
Garman`s tindings tor graduation rates as a tunction ot student SA1 scores and
college selectivity, controlling tor tamily background as vell as gender.
1able 11.1 Graduation rates related to ovn SA1 scores and college nedian
SA1 scores
9 1, College meian 8AT 9 1,

8iuenis' own 8ATs hiie males Crauaiion raies (7) Black males

700 or less 42 39 38 26
701830 47 60 36 39
8311,000 34 39
Over 1,000 34 37
Over 830 77 32

ource: Datcher Ioury and Garman (1993: 304, 1able 4,.
2004 Thomas E.Wei sskopf
It one uses the median SA1 score again as a proxy tor college selectivity
and competitiveness, 1able 11.1 suggests that Black male students in each
SA1 bracket have a signiticantly lover probability ot graduating at a more
competitive college (vith a median SA1 ot 1,000 rather than 900,. By
contrast, the probability ot graduation tor White male students in each SA1
bracket is actually higher at a more competitive college. (I vill return to this
anomaly belov, see pp. 1637., 1hernstrom and 1hernstrom (1997: 410, note
that the average White student in this sample had an SA1 score ot 1,011 and
vent to a college vith a median SA1 ot 1,019, vhile the average Black
student had an SA1 score ot 768 and vent to a college vith a median SA1 ot
936.
6
1hey take this as evidence that White students are typically vell
matched to their college, but that Black students are litted by AA into exces-
sively competitive colleges trom vhich they graduate at lover rates than
they vould have it properly matched. 1his interence is also dravn by Datcher
Ioury and Garman, vho conclude that 1he principal cost ot attending more
selective schools tor mismatched` Blacks is the lover probability ot graduation
and the subsequently lover earnings tor those vho tail to complete college
(Datcher Ioury and Garman 1993: 3067,. 1his conclusion, hovever, is not
easy to reconcile vith the evidence in the above table that the extent ot the
reduction in graduation rates tor Blacks at more selective colleges rises vith
their SA1 bracket, i.e. the graduation reduction is greater tor Black students
vho are less mismatched.
An alternative perspective is provided by Boven and Bok`s tindings on
graduation rates ot students vho entered selective colleges in 1976 and 1989.
1hese tindings, tor various ethnic groups ot students and tor the three
ditterent selectivity groups ot colleges in their sample ot tventy-eight, are
presented in 1able 11.3. 1he tigures in the table shov the percentage ot
students vho graduated vithin six years trom the college they entered.
7
Graduation rates trom each selectivity group ot colleges in the Boven and
Bok sample increased signiticantly trom the 1976 to the 1989 cohort.
1able 11.2 GPAs related to ovn SA1 scores and college nedian SA1 scores

9 1, College meian 8AT 9 1,

8iuenis' own 8ATs hiie males CPAs Black males

700 or less 2.66 2.66 2.38 2.24
701830 2.67 2.73 3.03 2.73
8311,000 2.73 2.81
Over 1,000 3.07 3.10
Over 830 2.87 2.63

ource: Datcher Ioury and Garman (1993: 304, 1able 4,.
2004 Thomas E.Wei sskopf
Ditterences among ethnic groups are evident tor both the 1976 and 1989
cohorts. Among students in the latter cohort, LPLG members (Blacks,
Hispanics, and Native Americans, graduated trom the Boven and Bok
colleges at a somevhat lover rate than Whites, but in both cases these rates
vere much higher than the corresponding graduation rates trom a representa-
tive sample ot all LS colleges.
1able 11.3 indicates also that in Boven and Bok`s sample, graduation rates
tor both Whites and LPLGs rose vith the selectivity ot the college. Boven
and Bok`s tindings about the positive relationship betveen graduation rates
and the degree ot selectivity ot a college remain robust even vhen they
undertake multivariate regression analysis to control tor a variety ot entry
qualitications including scores on the nationvide standardized college entry
test (the SA1,, secondary-school grades, gender, and socio-economic status.
8
1hese tindings stand in sharp contrast to those ot 1hernstrom and
1hernstrom (1997, and Datcher Ioury and Garman (1993,, vhich suggest a
negative relationship betveen college selectivity and graduation rates, I vill
return to this controversy belov (see pp. 1637,.
Lvidence on graduation rates trom graduate and protessional schools,
separately tor ditterent ethnic groups, is rather scarce. Once again, lav schools
have been the object ot the most systematic studies ot ditterences among
students trom ditterent ethnic groups. In her comprehensive study ot
1able 11.3 Graduation rates, by student ethnic group and college selectivity,
1976 and 1989
Bowen an Bok's
seleciiciiy group
Year All (7) hiie
(7)
Black
(7)
Hispanic-
American
(7)
Naiice-
American
(7)

1 1976 86 87 81 77 70
2 1976 78 80 67 67 64
3 1976 72 74 36 33 46

1 1989 94 93 83 92 79
2 1989 83 86 76 79 89
3 1989 81 82 68 71 72

All tour-year
colleges*
1989 38 39 40

ource: Boven and Bok (1998: 37, 3769, 1ables D.3.1 and D.3.2,.
^ote:
* A representative sample ot tour-year colleges.

2004 Thomas E.Wei sskopf
199091 applicants to American Bar Association-approved lav schools,
Wightman (1997, tound that Black and Hispanic-American students vho
enrolled in these schools had somevhat lover graduation rates than White
students. 1his is hardly surprising, the reasons as to vhy students do not
complete lav school are as likely to be economic as academic, and LPLG
students are typically much more heavily burdened by tinancial problems.
What is more relevant to the issue ot the impact ot AA on ditterential ethnic
graduation rates is that Wightman`s analysis shoved that there vas no signiti-
cant ditterence betveen the graduation rates ot those LPLG students vho
vould have been accepted to lav schools in the absence ot AA and the grad-
uation rates ot those vho oved their admission to AA.
In their study ot the consequences ot AA at the elite Lniversity ot
Michigan (LM, Iav School, Iempert et al. (2000a, tound that under-
represented minority students, although admitted vith quantitative entry
credentials that (on average, vere considerably lover than those ot White
students, did just about as vell in their lav-school studies. 1hey report that
96 per cent ot the minority students and 98.3 per cent ot the vhite
students vho entered the LM Iav School betveen 1983
9
and 1992 gradu-
ated trom Michigan vith the .D. [Doctor ot Iavs| degree (Iempert et al.
2000a: 422,. Ot those White and minority students vho entered but did
not graduate, most lett in academic good-standing to attend other schools
or to pursue other careers.
URLG and non-URLG grade perfornance
As in the case ot graduation rates, much ot the available data on ditterential
grade pertormance by LPLG and non-LPLG students in the LS has been
compiled tor Blacks and Whites. Moreover, the tirst reports on such ditteren-
tials came trom critics ot attirmative action. 1hus Bunzel (1988, provided
intormation on group-vise GPAs at LC-Berkeley in 1986, according to
vhich 93 per cent ot White and Asian-American students had GPAs ot 2.0 or
better, vhereas the corresponding tigure tor Chicanos vas 76 per cent and tor
Blacks 72 per cent.
1he tirst systematic and statistically sophisticated investigation ot ditter-
ences in GPAs betveen Blacks and Whites vas carried out by Datcher Ioury
and Garman (1993,, utilizing their database ot college-goers trom a represen-
tative nationvide sample ot male high-school graduates ot 1972. Among
these college-goers the mean cumulative GPA vas 2.91 tor Whites and 2.69
tor Blacks (on a scale running trom 0 to 4,.
10
Datcher Ioury and Garman
analyzed Black and White student GPAs as a tunction ot student SA1 scores
and college selectivity, controlling tor tamily background as vell as gender, in
the same vay as they analyzed graduation rates. 1heir tindings are summa-
rized in 1able 11.2.
1he evidence in 1able 11.2 indicates that Black male students in each
SA1 bracket achieve on average a signiticantly lover GPA vhen they attend
2004 Thomas E.Wei sskopf
a more competitive college (vith a median SA1 ot 1,000 rather than 900,,
vhereas tor White male students, college median SA1 makes little ditterence.
1hese tindings suggest that Black students vould do better attending less
selective colleges. In this case, the tigures on comparative GPA reductions tor
Black males attending more selective colleges indicate that the extent ot GPA
reduction is positively correlated vith the extent ot the mismatch betveen
the student`s SA1 score and the median college SA1 (though the GPAs tor
Black males in the middle SA1 bracket seem anomalously high,.
Data compiled by Boven and Bok (1998: 728, tor the cohort ot students
entering their College and Beyond (C8B, sample ot selective colleges in
1989 shov that there is a signiticant ditterence betveen Blacks and Whites in
grade pertormance: the average GPA is 2.61 tor Blacks and 3.13 tor Whites.
1he lover average GPA tor Black students is not explainable simply in terms
ot lover entry test scores. Boven and Bok tind that the average Black GPA ot
2.61 is in tact less than vould be predicted on the basis ot their average SA1
score, this retlects vhat has been characterized as underpertormance.
11
In
Boven and Bok`s sample the average Black graduate ranks in the tventy-third
percentile (trom the bottom, ot the graduating class, vhereas the average
White ranks in the titty-third percentile. Lven controlling tor SA1 scores and
other available college-entry characteristics, Boven and Bok tind that Blacks
graduate on average vith a signiticantly lover class rank than Whites. As one
vould expect, Boven and Bok also tind (1998: 383, 1able D.3.6, that,
controlling tor college-entry characteristics, both Blacks and Whites are likely
to graduate at a lover class rank trom a more selective college. 1his implies
that they are likely to graduate trom a more selective college vith a lover
GPA as vell unless the overall distribution ot grades in that college is signit-
icantly higher than in the alternative, less selective school.
1hat under-represented minority students compile lover cumulative GPAs
than Whites is a tinding that applies to the graduate as vell as the undergraduate
level ot post-secondary education. In the case ot lav schools, many studies have
shovn: (1, that Black, Hispanic-, and Native-American students enter vith
considerably lover average cumulative undergraduate GPAs and considerably
lover average scores on the nationvide standardized Iav School Admissions
1est (the ISA1, than do White and Asian-American students, (2, that an index
based on these tvo indicators is a statistically signiticant predictor ot grades
earned by students in their tirst year ot study, and (3, that LPLG students on
average earn lover grades and graduate at lover class ranks than do non-LPLG
students.
12
Iempert et al. (2000a: 43968,, in the context ot their analysis ot
attirmative action at the LM Iav School trom the late 1960s to the mid-1990s,
contirmed these tindings and tound also that an index combining the tvo pre-
lav-school quantitative indicators has a signiticant ettect on the cumulative as
vell as the tirst-year lav-school GPAs ot LM Iav School students.
Boven and Bok assert that the phenomenon ot underpertormance charac-
terizes Black students (on average, at the graduate as vell as at the
undergraduate level.
13
1heir ovn tindings at the undergraduate level are
2004 Thomas E.Wei sskopf
corroborated and extended by Vars and Boven (1998,, vho made use ot data
on students vho entered in 1989 a subset ot eleven colleges in the same C8B
database. 1hey tound that student GPAs are correlated vith student SA1s
(although less so tor Blacks than tor Whites,, and that in every SA1 bracket,
Blacks earn lover grades than Whites, even atter controlling tor other vari-
ables like high-school grades and tamily socio-economic status. Interestingly,
the pertormance gap is greatest tor those Blacks vith the highest SA1s. As
they note, tor reasons not vell understood, many gitted Blacks are not
reaching their tull academic potential at selective institutions (Vars and
Boven 1998: 438,.
1he tact that the average level ot academic pertormance ot Black students
remains lover than that ot White students, even controlling tor standardized
test scores (and otten other entry characteristics as vell,, has been a source ot
much concern and discussion. 1he evidence tor this kind ot underpertor-
mance on the part ot Atrican Americans is especially strong at selective LS
colleges and universities. What could account tor such tindings
1here are tvo possible reasons tor tindings ot underpertormance, each ot
vhich encompasses a variety ot possible explanations. On the one hand, the
studies may tail to control tor certain characteristics ot entering students that
are both ditticult to measure and on the vhole less tavorable tor Blacks than
tor Whites. lor example, Blacks entering college typically come trom home,
community, and school environments that are less conducive to good study
habits and learning practices in vays that are not necessarily retlected in test
scores or other readily measured entry characteristics.
14
On the other hand, it
may be the case that higher educational institutions tend not to be as
successtul in educating Black students as they are in educating othervise
similar White students. lor this there are several ditterent possible explana-
tions, each vith very ditterent implications.
13
One such explanation tor this ditterence in educational ettectiveness is that
predominantly White taculty members may not have the cultural avareness to
teach Black students as vell as they do White students. Another possibility is
that Black students may tend to tind their in-college experience more ditticult
than do White students, tacing, tor example, greater problems adjusting to a
nev (and White-dominated, college campus environment, greater vulnera-
bility to race-related social tensions, and possibly subtle (it not overt, torms ot
racial discrimination. Yet another possibility is that many Black students are
characterized by a torm ot stereotype bias that inhibits them trom
pertorming vell on examinations and in other competitive academic
activities.
16
linally, it is possible that a considerable number ot Black students
in a predominantly White environment may develop an alternative subculture
that disparages academic achievement as White, peer pressure along these
lines could inhibit even highly capable students trom doing as vell as they
othervise might.
17
Over the last thirty years, selective LS colleges and universities have made
increasingly vell-intormed ettorts to assist Black students in overcoming some
2004 Thomas E.Wei sskopf
it not all ot the above obstacles to academic success. 1he more prestigious
and vell-endoved institutions have been able to do more along these lines.
lev can credibly claim, hovever, that they have tully equalized the learning
environment tor Black and White students.
Analysis of the relationship betveen college selectivity
and student acadenic perfornance
In the previous tvo sections I have presented evidence mainly on the
comparative academic pertormance ot LPLG and non-LPLG students
attending the same higher educational institutions. In discussing such
evidence, hovever, I noted hov several studies sought to estimate the vay in
vhich the academic pertormance ot a given student vould likely vary across
institutions ot dittering selectivity and quality. As noted in Chapter 8, the rela-
tionship betveen the selectivity ot a higher educational institution and the
academic pertormance ot given students is critical to the assessment ot several
ot the claims made by proponents or opponents ot AA in higher education
especially claims related to the argument that AA results in a mismatch
betveen AA beneticiaries and their universities. In this chapter, theretore, I
vill discuss in more detail the nature and implications ot studies that have
addressed this relationship. (1hus tar such studies have been done only in the
LS.,
Some ot the studies on ditterential academic pertormance ot Whites and
Blacks, vhich I discussed in the preceding sections, also addressed the claim
that Black students vho enter selective colleges by virtue ot AA tend to
pertorm poorly or drop out ot those colleges because they are academically
mismatched. 1his is the implication ot 1hernstrom and 1hernstom`s (1997,
comparison ot the Black-White SA1 gap vith the corresponding graduation
gap at elite colleges, and this is also the implication ot Datcher Ioury and
Garman`s (1993, tindings on the relationship betveen selectivity and both
graduation rates and GPAs. Yet the Boven and Bok (1998, study points to
the opposite conclusion vith respect to graduation rates. And there remains
an unresolved anomaly in Datcher Ioury and Garman`s tinding that the prob-
ability ot graduation tor Whites increases vith college selectivity vhereas tor
Blacks it decreases. In order to reconcile the apparent inconsistencies in the
above tindings vith respect to the ettect ot selectivity, it vill be helptul to
consider more generally hov to analyze the determinants ot a given student`s
academic pertormance.
What are the tactors that one might expect to intluence a student`s
academic pertormance at college beyond the natural-born ability and readi-
ness to apply ettort that the student brings to college 1he tolloving tactors
are surely relevant:
Iamil socioeconomic |ackrouno 1his is important tor at least tvo distinct
reasons. lirst, a better-ott tamily can provide more economic security
2004 Thomas E.Wei sskopf
making it less likely that a student vill be troubled in school by
economic concerns, or burdened at school by having to take on a paying
job, or obliged to quit school to vork tull-time at a paying job. Secondly,
a better-ott tamily can provide a student vith a home and community
environment more conducive to the development ot good studying and
learning habits.
Prior acaoemic achie:ement A student vho enters college vith better
academic qualitications as measured by indicators such as hisher high-
school SA1 score, class rank, or GPA, and the competitiveness ot the high
school attended vill tend to do better in hisher college studies.
ualit of the collee Higher-quality colleges (as measured, e.g., by the
median SA1 ot students attending, have higher-quality students and
taculties, and tend to have more resources, they are likely to do a better
job ot educating students because students vill have better-prepared class-
mates, better teachers, more andor better support programs and tacilities,
and probably also more tavorable social-educational norms.
Match to the acaoemic en:ironment 1o the extent that a student`s ovn
academic qualitications match those ot hisher peers, it is more likely that
the student vill thrive rather than sutter in the college`s educational envi-
ronment.
18
Match to the social en:ironment 1he more that a student`s ovn
socialcultural background matches that ot hisher college campus envi-
ronment, the more likely the student is to adjust to and prosper
(academically as vell as socially, at the college.
AA policies lead to changes in the ethnic composition ot college student
bodies in such a vay as to increase the salience ot the last three ot the above
tactors. By redistributing LPLG students upvard and non-LPLG students
dovnvard in the selectivityquality spectrum ot colleges, they put the typical
LPLG student into a higher-quality college. It the upvard redistribution ot
LPLG students has the ettect ot putting the typical LPLG student into a
college vhere hisher typical non-LPLG peer has stronger academic qualiti-
cations vhich is amply contirmed by the evidence revieved in Chapter 9,
pp. 1403, and above, pp. 1603 then AA policies simultaneously increase
the extent ot the academic mismatch betveen the typical LPLG student and
hisher college. AA policies also increase the extent ot the social mismatch tor
the typical LPLG student, insotar as they lead some students avay trom
predominantly LPLG colleges (like the historically Black colleges (HBCs,,
and into predominantly non-LPLG colleges (like the selective colleges that
practice AA in admissions,.
Critics ot AA claim that the negative ettect on academic pertormance trom
increasing the academic mismatch taced by the typical LPLG student vill
outveigh any positive ettect derived trom the correspondingly increased
quality ot the college attended. Proponents ot AA claim just the opposite. 1he
tact that these claims are simultaneously and directly opposed to one another
2004 Thomas E.Wei sskopf
should tacilitate empirical testing ot their validity by means ot an analysis ot
the ettect ot increased college selectivity on the academic pertormance ot AA
beneticiaries. lor such an analysis to be statistically legitimate and tor its tind-
ings to be compelling, hovever, it must control as ettectively as possible tor
intluences on student academic pertormance other than selectivity itselt. In
particular, it must control tor the other three tactors in the list above tamily
socio-economic background, prior academic achievement, and match to the
social environment so that ditterences in academic pertormance due to vari-
ation in these tactors are not contused vith ditterences ot college quality and
academic mismatch due to variance in college selectivity. With this in mind,
let us examine the available empirical evidence.
1hernstrom and 1hernstrom`s (1997, implicit correlation ot the size ot the
Black-White SA1 gap vith the size ot the Black-White graduation gap across
highly selective colleges (vhere a large percentage ot the Black students can
be assumed to be AA beneticiaries, does not prove that an AA-induced SA1
mismatch betveen Black students and their college peers is the cause ot their
lover graduation rates. 1his is because the correlation tails to control tor
ditterences in the socio-economic background and the prior academic
achievement ot Black students at ditterent colleges. 1hose Black students in
the 1hernstrom and 1hernstrom sample vho vere more SA1-mismatched
and more likely to drop out ot college might have dropped out tor reasons
that vould have applied vherever they had attended college e.g. because ot
economic insecurity associated vith a lov socio-economic status tamily
background,
19
or because ot inadequate prior educational preparation.
Datcher Ioury and Garman (1993,, utilizing a much richer database on
students going to a much broader variety ot colleges, address the major limita-
tions ot the 1hernstrom and 1hernstrom tindings by introducing into their
analysis ot student academic pertormance (as a tunction ot college selectivity,
variables tor student tamily background and prior academic achievement.
1hese variables control reasonably vell, though tar trom pertectly, tor ditter-
ences in relevant characteristics ot students attending colleges ot varying
selectivity. But Datcher Ioury and Garman included HBCs in their sample,
vithout controlling tor the resultant ditterences in the matching ot Black
students and White students to the social environments ot HBCs as distinct
trom White-dominated colleges.
Boven and Bok (1998, did control tor social match or mismatch, by the
simple expedient ot contining their sample to elite (highly selective, colleges
and thus excluding all HBCs. 1hey tabulated many ot their tindings on
ditterential Black and White student academic pertormance as averages, tor
their tull sample ot colleges and tor three sub-groups distinguished according
to level ot selectivity. lrom such tabulations they tound a positive relation-
ship betveen graduation rates and college selectivity, but those tindings are
subject to the same criticism levied above against the 1hernstrom and
1hernstrom tindings. Boven and Bok vent on, hovever, to undertake multi-
variate regression analyses ot various measures ot student pertormance, in
2004 Thomas E.Wei sskopf
vhich they vere able to control to a considerable extent tor tamily back-
ground and prior academic achievement, and they tound that the positive
relationship betveen graduation rates and college selectivity, vhile somevhat
veakened, remained robust.
In a sophisticated ettort to address these issues, Kane (1998: 4408, made
use ot a longitudinal database tor a nationvide representative sample ot high-
school graduates ot 1982 similar to Datcher Ioury and Garman`s database,
but trom observations ten years later. Kane vas able to distinguish observa-
tions on Hispanic Americans trom observations on (non-Hispanic, Blacks and
non-Hispanic others, in the research discussed belov, he combined Hispanic-
American and Black students into a single under-represented minority
category, and the remaining White category includes Asian Americans. He
sought to predict student academic pertormance both probability ot gradu-
ation and cumulative GPA as a tunction ot tamily background (tamily
income and parental education,, prior academic achievement (measured by
individual SA1 and high-school GPA,, and qualityselectivity ot college
(college median SA1,. Holding the tamily background and prior academic
achievement indicators constant, Kane tound that greater college selectivity is
associated vith a higher probability ot graduating and vith a slightly lover
GPA.
20
His tindings support Boven and Bok`s rejection ot the academic
mismatch claim tor graduation rates, and they are consistent also vith Boven
and Bok`s contirmation ot the mismatch claim vith respect to class rank.
Kane`s initial results vere also tairly consistent vith those ot Datcher Ioury
and Garman`s, in that greater college selectivity vas associated vith a higher
probability ot graduation tor Whites but much less so tor Blacks and Hispanic
Americans (Kane did not tind the negative relationship tor Blacks highlighted
by Datcher Ioury and Garman,. When he controlled tor HBCs, hovever, he
tound that this ditterence disappeared greater selectivity is associated vith
higher probability ot graduation tor both under-represented minority and
White students. Lvidently the ditterence tound by Datcher Ioury and
Garman vas due to the tact that Black students attending HBCs have rela-
tively lov SA1 scores but relatively high graduation rates. Kane`s tindings on
college selectivity lend strong support to those vho reject the academic
mismatch claim, and they also explain the seemingly anomalous Datcher
Ioury and Garman result about the tavorable ettect ot college selectivity on
White graduation rates. Kane is hesitant to claim a positive ettect ot college
selectivity on graduation rates because ot the possibility that the omission ot
variables on student academic competence (other than SA1 scores and high-
school grades, may have biased upvards his estimates ot the selectivity ettect,
but he sees no reason that an upvard bias vould ditter by race, and he
unequivocally rejects a negative selectivity ettect.
21
1he evidence I have revieved here, in its totality, appears to reject very
convincingly the academic mismatch hypothesis as applied to the probability
ot graduation trom college ot LPLG students. 1he evidence indicates that
students ot any ethnic group are no less likely to graduate trom a more selec-
2004 Thomas E.Wei sskopf
tive college, though they vill graduate vith a lover class rank trom a more
selective college. Since more selective colleges can be expected on the vhole
to impart higher-quality educational services to their students, one can drav
the conclusion that it is generally in the academic interest ot any given
student to attend the most selective college to vhich heshe can gain access.
1his implies, ot course, that AA action policies do serve the best academic
interests ot the LPLG students vhom they attect, by the same token, they
harm the academic interests ot the non-LPLG students vhom they attect.
1here is one respect, hovever, in vhich the evidence is not sutticient to
support the contention that LPLG students are vell served academically by
AA policies. It is not clear that Black students generally have just as high a
probability ot graduating trom highly selective predominantly White colleges
as they do trom less selective HBCs. Boven and Bok`s tindings do not speak
to this issue at all, since they excluded HBCs trom their sample.
22
When
Datcher Ioury and Garman and Kane include observations trom HBCs in
their analysis, their tindings are at best ambiguous vith respect to the
contention that increased college selectivity does not have a negative ettect on
the probability ot graduation ot Black students.
1hus it may be true that Black students vho seize an opportunity (ot the
kind attorded by AA, to go to a more selective, predominantly White college
pay an academic price, in terms ot a probability ot graduation that is lover
than the one they vould have had at an HBC. 1his vould constitute a cost
attributable to mismatch. It vould have to be attributed, hovever, to a social
rather than an academic mismatch, tor the evidence suggests that academic
mismatch alone holding the social environment constant does not result
in a lover probability ot student graduation. Moreover, even it there is an
expected cost in lover probability ot graduation to be paid by a Black
student shitting trom an HBC to a more selective, predominantly White
college, that cost may be more than ottset by a higher expected economic
return to graduation trom an elite college. 1o determine vhether or not it is
in the economic interest ot Black students to make such a shitt requires
evidence on the vay in vhich college selectivity attects post-college career
prospects an issue to vhich I vill turn in Chapter 13.
2004 Thomas E.Wei sskopf

You might also like