«n.5,

3%:
.21.: .

.3 c. n
55.4..
. viz: .
x $5.... ,
21...: l. .4:
nuurvzha. ... 2
its. 13:. ,

L
u, x 3.3.9...“

oil-t.
131."
.fv'llvll. .
$5.3 ... A}
. til-.69
f u. a

”5 F vi.

[1.f‘l‘z.

‘9.»

.31.; .15;
Q .t I«‘

suffix
l a. i}?

i .r: ..n...

IF ‘1 .

u..h..4.. 2...:
‘ A. .fv. . :l
.1 .1..~.
‘, v 2'51... v
1.... {Wadi l.
y! 1 .. . 3.0
y. 3.... J . 34;.va
:u . . fun. i 5.51:... . . I
E ’1 .. , z 4
x.‘ \ .. 22:4". .V f

.4. .1.
, 1., .

 

..fln;p.n..,....rflafl§§§.5. ., . ‘

.. .f. .. . . ... .3
. q . . .f 9...: ii .5... 1‘: Hi: ,

|\\\\\\\\l\\\\|\\\\\\ W

3 1293 00887

\ \\\\\\\\\\\i\\\\€\5\\‘\‘\4\i\\7\\\m

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

CHANGES IN OUTCOMES OF THE
COLLEGE ENTRANCE EXAMINATION BOARD
SCHOLASTIC APTITUDE TEST PROGRAM
FROM 1988 TO 1991

presented by

Mary Mayumi Kino

has been accepted towards fulfillment
of the requirements for

' P h D degree in

Educational Psychology,
& Special Education

\MQAK

Major professor
Date WW 2: If? 3

 

MS U i: an Affirmative Action/Equal Opportunity Institution 0- 12771

 

 

 

 

 

 

 

 

PLACE iN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ll

MSU is An Affirmative ActiorVEquol Opportunity institution
czuMprns-ot

 

CHANGES IN OUTCOMES OF THE COLLEGE ENTRANCE EXAMINATION BOARD
SCHOLASTIC APTITUDE TEST PROGRAM FROM 1988 TO 1991

BY

Mary Mayumi Kino

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Counseling, Educational Psychology, and
Special Education

1993

ABSTRACT

CHANGES IN OUTCOMES OF THE COLLEGE ENTRANCE EXAMINATION BOARD
SCHOLASTIC APTITUDE TEST PROGRAM FROM 1988 TO 1991

BY

Mary Mayumi Kino

Changes in outcomes of the College Entrance Examination

Board’s Scholastic_Antitude_Iest (CEEB 5A1) are examined for

college-bound seniors. 'fhe outcomes include college—bound
seniors': (a) gender, ethnicity, and parent educational
attainment compositions; (b) academic preparation; (c)

academic aspirations; and (d) mean SAI-Total scores. Changes
in these outcomes are described with frequency summaries,
analyses of proportions, analyses of observed group means,
and the analyses of adjusted means.

Compositional changes in parent education and ethnicity
were significant factors in the score decline of both
observed and adjusted means. Observed mean scores of
examinees from high parent education groups increased, but
the proportions of respondents from these groups declined.
The proportions of examinees from low parent education groups
increased significantly, and. their observed :mean. score
declines dominated patterns of observed mean score increases
found among other groups. Emaminees from ethnic minority
groups, traditionally associated with low mean SALE-Total
scores, became more numerous, while the proportions of the

White examinee group, traditionally associated with high mean

agirTotal scores, decreased over time. Between 1988 and
1991, the population of college-bound examinees who took the
5A1 became increasingly disadvantaged on both ethnicity and
parent educational attainment variables.

Patterns of net mean score change among traditionally
high-scoring groups revealed a ”ceiling effect," likely
explained by the fact these groups were already obtaining
relatively high mean scores. Net mean score declines among
traditionally low-scoring groups and their increasing
proportions in the samples implied a change in the self—
selection ratio on the SAT.

The trend in predicted mean fiAI-Total scores, adjusted
at the demographic composition of the 1988 cohort, revealed
that more than half of the nine—point score decline in
unadjusted means was explained by compositional change
between 1988 and 1991. The findings of this study are
consistent with the hypothesis that change in observed mean
$51 scores between 1988 and 1991 was attributable, in part,

to change in the constituency of the examinee population.

 

ACKNOWLEDGMENTS

Towards the completion of this study, I found myself
dreaming about actually being in a position to write
acknowledgments. It is with great pleasure that I now thank
the many individuals who contributed to this undertaking.

I am grateful to the members of my committee, Betsy Jane
Becker, John E. Hunter, William A. Mehrens, Stephen W.
Raudenbush, and Herbert C. Rudman, for their support, advice,
and helpful comments with the dissertation from the proposal
stage to the final version.

William H. Schmidt, Director of the United States
National Research Center for the Third International
Mathematics and Science Study, made it possible for me to
complete the dissertation during my final year of graduate
studies as a research assistant on his research projects.

I owe an enormous debt to my adviser, Herb Rudman. I
deeply appreciate his patience, understanding, and
willingness to discuss at any time the dissertation at its
various stages of development. Many more times than he would
probably like to recollect, I prevailed upon his wisdom from
37 years of distinguished service to the profession of

education. What could have become a hazardous project, in

iv

the end, became a rewarding one under his guidance and
advice, both editorial and otherwise.

Finally, I thank my parents for their unwavering belief
and support throughout my graduate studies and the writing of
the dissertation. Their confidence in my abilities and
capabilities were enduring and crucial during many phases of
this endeavor. Without their help and encouragement, I am

certain that I would not now be writing these words.

TABLE OF CONTENTS

LIST OF TABLES ............................................ X
CHAPTER 1: AN OVERVIEW ................................... 1
Purpose ................................................. 1
Need .................................................... 2
Theory .................................................. 6
Overview ................................................ 8
CHAPTER 2: REVIEW OF THE LITERATURE ...................... 9
What Does the SAT Really Measure? ....................... 9
The Use of SAT in College Admissions .................... 10
Changes in Mean SAT Scores Over Time .................... 11
Sources of Change in Mean SAT Scores .................... 13
Test Scale Drift ..................................... 13
Constituency of Examinees ............................ 15
Gender composition .............................. 17
Ethnic group composition ........................ l7
Socioeconomic status ............................ 18

High school rank in class and course-taking

patterns ........................................ 19

College plans ................................... 20

Educational degree goals ........................ 21

Underlying Ability Distributions ..................... 21

Summary ................................................. 23
vi

vii

CHAPTER 3: ELEMENTS OF THE STUDY ......................... 26
Research Questions ...................................... 26
Structure of the Data Set ............................... 26
Outcomes of the CEEB SAT ................................ 28

The Student Descriptive Questionnaire ................ 28
Traditional demographic characteristics ......... 29
Academic preparation ............................ 30
Academic aspirations ............................ 31

SAT-Total Score ...................................... 32

Hypotheses and Methodology .............................. 33

Question 1 ........................................... 33

Question 2 ........................................... 35

Question 3 ........................................... 36

Question 4 ........................................... 37

Question 5 ........................................... 38

Question 6 ........................................... 39

Analyses ................................................ 39

Question 1: Changes in Demographic

Characteristics ...................................... 41

Question 2: Changes in Academic Preparation ......... 41

Question 3: Changes in Academic Aspirations ......... 42

Question 4: Changes in SAT-Total Score
Distribution ......................................... 42

Question 5: Changes in Mean SAT—Total Score by
Subgroups ............................................ 42

Question 6: Mean Score Decline Explained by Change
in Demographic Composition ........................... 43

 

viii

CHAPTER 4: RESULTS ....................................... 45
Section 1: Analyses of Proportions ..................... 45
Demographic Composition .............................. 45
Gender .......................................... 45
Ethnic group membership ......................... 46
Highest level of parent’s educational J
attainment ...................................... 48
Academic Preparation ................................. 52
Academic Aspirations ................................. 58
Intended college major .......................... 58
Highest degree aspirations ...................... 60
Section 2: Analyses of Means ........................... 63
SAT—Total Score Distributions ........................ 63
Coding of Variables .................................. 65
Preliminary Analyses ................................. 66
Single—Factor Categories ............................. 68
Gender .......................................... 68
Ethnic group membership ......................... 71
Parent educational attainment ................... 75
High school class rank .......................... 78
Two—Factor Categories ................................ 81

Parent education and ethnic group membership.... 81
Ethnic group membership and class rank .......... 85
Parent education and class rank ................. 89

Three-Factor Categories: Parent Education, Class
Rank, and Ethnic Group Membership .................... 91

Mean Score Decline Explained by Compositional
Change ............................................... 101

Synopsis ............................................. 106

 

ix

CHAPTER 5: INTERPRETATION AND DISCUSSION ................. 107
Demographic Compositions ............................. 107
Academic Preparation ................................. 109
Academic Aspirations ................................. 110
Mean SAT-Total Scores ................................ 111
Mean SAT-Total Scores by Specific Subgroups .......... 112

Gender .......................................... 112
Ethnic groups ................................... 113
Parent education ................................ 113
High school class rank .......................... 114
Parent education and ethnicity .................. 114
Class rank and ethnicity ........................ 115
Parent education and class rank ................. 115
Parent education, class rank, and ethnicity ..... 116
Adjusted Mean SAT-Total Scores .................. 119
Limitations ............................................. 120
Conclusions and Recommendations ......................... 122

APPENDIX A: SUMMARIES OF 1980-1987 CEEB OUTCOMES ......... 127

APPENDIX B: MEAN SAT-TOTAL OUTCOMES FOR 1988-1991 ........ 135

APPENDIX C: ADJUSTED MEAN SAT—TOTALS FOR 1988-1991 ....... 161

APPENDIX D: SENSITIVITY ANALYSES. ADJUSTED MEAN SAT-
TOTALS .................................................... 163

LIST OF REFERENCES ........................................ 170

 

b;

 

LI ST OF TABLES

Table 1: High School Graduates & Percent College-Bound
Seniors ................................................... 4

Table 2: Mean SAT—Total & Subtest Scores ................. 12

Table 3: Percent 17-year Old High School Graduates &

Drop-Outs ................................................. 16
Table 4: Percent of Examinees Who Are College—Bound

Seniors ................................................... 27
Table 5: Effective Sample Sizes Used for Analyses ........ 28

Table 6: Proportions (Frequencies) of Sample by Gender ...45

Table 7: Proportions (Frequencies) of Ethnic Minority
Groups .................................................... 47

Table 8: Proportions (Frequencies) of Parent's Highest
Education ................................................. 50

Table 9: Proportions (Frequencies) of Course-Taking

Patterns .................................................. 54
Table 10: Proportions (Frequencies) of High School

Class Rank ................................................ 57
Table 11: Proportions (Frequencies) of Intended College
Majors .................................................... 59
Table 12: Proportions (Frequencies) of Highest Degree

Goals ..................................................... 61
Table 13: SAT—Total Score Distribution Summary ........... 63
Table 14: SAT-Total Score, Measures of Central

Tendency .................................................. 65
Table 15: Variables Used in Analyses of Means ............ 66
Table 16: Analysis of Variance for Year .................. 67
Table 17: Analysis of Variance for Gender & Year ......... 69

Table 18: Raw Mean SAT-Total (Std Dev), Sample Size n
by Gender Groups .......................................... 70

xi

Table 19: Analysis of Variance for Ethnic Groups &
Year ...................................................... 72

Table 20: Raw Mean SAT-Total (Std Dev), Sample Size n
by Ethnic Groups .......................................... 73

Table 21: Analysis of Variance for Parent Education
Groups & Year ............................................. 75

Table 22: Raw Mean SAT-Total (Std Dev), Sample Size n
by Parent Education Groups ................................ 76

Table 23: Analysis of Variance for Class Rank Groups &
Year ...................................................... 78

Table 24: Raw Mean SAT—Total (Std Dev), Sample Size n
by High School Class Rank Groups .......................... 80

Table 25: Analysis of Variance for Parent Education,
Ethnic Groups & Year ...................................... 82

Table 26: Net Mean SAT-Total Score Change, Parent Education
& Ethnic Groups ........................................... 83

Table 27: Analysis of Variance for Class Rank, Ethnic
Groups & Year ............................................. 86

Table 28: Net Mean SAT-Total Score Change, High School
Class Rank & Ethnic Groups ................................ 87

Table 29: Range of Net Mean SAT—Total Score Change for
1988-1991, Ethnic & Class Rank Groups ..................... 89

Table 30: Analysis of Variance for Parent Education,
Class Rank Groups & Year .................................. 89

Table 31: Net Mean SAT—Total Score Change, Parent
Education & Class Rank Groups ............................. 91

Table 32: Analysis of Variance for Parent Education,
Class Rank, Ethnic Groups & Year .......................... 92

Table 33: Ethnic Groups With Significant Linear Trend
in Mean SAT Scores by Parent Education 8. Class Rank
Groups .................................................... 93

Table 34: Ethnic Groups With Highest Mean SAT Scores,
by Parent Education & Class Rank Groups ................... 95

Table 35: Ethnic Groups With Lowest Mean SAT Scores, by
Parent Education & Class Rank Groups ...................... 96

Table 36: Net Mean SAT-Total Score Change, Parent
Education, Ethnicity, & High School Class Rank Groups ..... 97

 

 

xii

Table 37: Adjusted Mean SAT-Total Scores ................. 102
Table 38: Net Mean Change, Adjusted Scores, by Ethnic

Subgroups & Models ........................................ 104
Table 39: Mean SAT—Total for Selected & Excluded Cases ...121

Table A-1: 1980 — 1987, Percent (Frequencies) of Gender
Composition ............................................... 127

Table A—2: 1980 — 1987, Proportions (Frequencies) of
Ethnic Groups ............................................. 128

Table A-3: 1980 - 1985, Proportions (Frequencies) of
Parent Income Distributions ............................... 129

Table A—4: 1980 - 1987, Proportions (Frequencies) of
Class Rank Distributions .................................. 130

Table A-5: 1980 - 1987, Proportions (Frequencies) of
Academic Preparation ...................................... 131

Table A-6: 1980 - 1987, Proportions (Frequencies) of
Intended College Majors ................................... 132

Table A-7: 1980 - 1987, Proportions (Frequencies) of
Highest Degree Goals ...................................... 133

Table A-8: 1980 - 1991, Number of College—Bound
Seniors ................................................... 134

Table B-1: Raw Mean SAT-Total (Std Dev), Sample Size
n, Parent Education & Ethnic Groups ....................... 135

Table B-2: Raw Mean SAT-Total (Std Dev), Sample Size
n, Class Rank & Ethnic Groups ............................. 139

Table B-3: Raw Mean SAT—Total (Std Dev), Sample Size
n, Parent Education & Class Rank Groups ................... 143

Table B-4: Raw Mean SAT-Total (Std Dev), Sample Size
n, Parent Education, Class Rank, & Ethnic Groups .......... 145

Table C-l: 1988 Predictor Means, Separate Year
Regression Estimates for Intact Annual Cohorts ............ 161

Table C-2: 1988 Predictor Means, Separate Year
Regression Estimates by Ethnic Subgroups .................. 162

Table D-1: Parent Education, Ethnicity, Gender, & Class
Rank Main Effects Model ................................... 165

Table D-2: Parent Education, Ethnicity, Gender, Class
Rank, & 'Parent Education-by-Class Rank" Model ............ 167

 

 

CHAPTER 1

AN OVERVIEW

Purpose
The purpose of this dissertation is to investigate the
ways in which outcomes of the College Entrance Examination

Board’s ScholasLiQ_Aptitude_IesL (CEEB SAT) changed between

1988 and 1991. These outcomes include the following:

(a) traditional demographic characteristics of college-bound
seniors who took the SAT. These variables include self-
reported ethnicity, gender, and highest level of
parents' educational attainment;

(b) indices of academic preparation” 'Phese 'variables
include high school rank in graduating class, and years
of study in English, foreign language, mathematics,
natural science, and social sciences;

(C) academic aspirations. These variables include intended
college major and highest educational degree goals; and

(d) SAT-Total score distributions of college-bound high
school seniors.

Mean SAT-Total scores for particular subgroups of the sample,

their net mean SAT-Total score changes from 1988 to 1991, and

patterns of score change among these groups are examined.

The study concludes with an assessment of the observed mean

score decline explained by change in demographic composition

between 1988 and 1991.

i

 

 

Need

The College Entrance Examination Board (CEEB) Sghglgfitig
Aptitndg_Tg§; (SAT) has been and continues to be an important
standardized testing' prograni in. the college admissions
process. A large proportion of higher educational
institutions uses information from the CEEB SAT program and
other supplementary sources to make admissions decisions, and
to assist and to guide admitted students. (See, for example,
American Association of Collegiate Registrars & Admissions
Officers [AACRAO] & CEEB, 1980.) Changes in the world
economy and increasing competition for employment place an
increasing demand on access to higher education.
Consequently, outcomes of the CEEB SAT are important to many
different audiences.

Nearly three decades ago, concerns about international
competition were prominent among the general public.
Outcomes on large-scale testing programs like the CEEB SAT
which involved large samples of students at or near the end
of their secondary school education were used as one index of
the academic progress of a national group. When, upon close
examination, it was found that the students of that period
were performing significantly less well than students of a
decade earlier on the SAT, considerable attention was focused
on the possible causes and implications of the observed SAT
score decline (CEEB, 1977; National Commission on Excellence

in Education, 1983).

 

From 1977 to 1980, mean SAT scores continued to decline.
From 1980 to 1987, mean SAT scores stabilized and then
increased slightly. Despite this reversal in score decline,
the slight increase in mean SAT scores was insufficient to
recover the accumulated decline in mean SAT scores since
detection of the initial score decline in 1967. During the
most recent SAT administrations from 1988 to 1991, mean SAT
scores again declined with each successive year of
administration (CEEB, 1991). In 1991, mean SAT scores were
below the mean SAT scores of 1967. Nearly two decades have
elapsed since the initial study of the 1967 — 1977 SAT score
decline. No single factor could account for the early SAT
score decline, and it is unlikely that any single factor can
account for the current decline in mean SAT scores.

In addition to the observed decline in mean scores from
1988 to 1991, the variance of SAT—Total scores consistently
increased over time (CEEB, 1980, 1981, 1982, 1983, 1984.
1985, 1986a, 1987, 1991). The increasing spread of SAT score
distributions suggests the presence of changing
characteristics either in the SAT or in the population of
examinees who took the SAT. In what ways did the
constituency of SAT—takers change from 1988 to 1991? Are
there different patterns of mean score change on the SAT
between 1988 and 1991 for different demographic groups?

It is inappropriate to extrapolate the outcomes from the
CEEB SAT to all high school students in the United States.

The population of examinees who take the SAJL is

 

 

self—selected, and these examinees are not necessarily
representative of all high school students, of all students
seeking admission to higher education institutions, or of all
students who ultimately enroll in college. College-bound
high school seniors who take the SAT do not necessarily
comprise a representative sample of the nation’s college—
bound high school seniors. They do, however, comprise a
majority of the .SAT:taking population and form a
representative sample of the SAT-taking population.
Furthermore, college-bound seniors who take the SAT currently
comprise about three-fourths of all high school graduates who
enroll in college each year as shown in Table 1 (CEEB, 1991;
United States Department of Education [USDE] & Office of

Educational Research & Improvement [OERI], 1992).

Table 1
. l E 1 J E 1 : J] _ i E .

 

High School College-Bound Percent College—Bound

 

Year Graduates Seniors Seniors
1988 1,575,000 1,134,364 72
1989 1,463,000 1,088,223 75
1990 1,410,000 1,025,523 73
1991 1,420,000 1,032,685 73

 

The CEEB SAT outcomes are a rich source of information
about a substantial number of the nation's graduating high

school seniors. Outcomes include but are not limited to

information about SAT scores, high school achievement,

 

 

 

academic aspirations, and traditional demographic
information. An investigation and interpretation of these
outcomes in the appropriate context can yield important
findings. Annual summaries of SAT scores, for example, have
shown fluctuations in mean SAT scores and consistently lower
mean SAT scores since 1967 (CEEB, 1991).

Not only have mean SAT scores changed since 1967,
demographic characteristics of examinees who took the SAT
also changed over time. The constituency of SAT-takers who
are members of various ethnic groups fluctuates as world
politics affect the number of U. S. immigrants, who also seek
access to higher education in the U. S. The ethnic
composition of the examinee population continues to change
and the size of the examinee population continues to increase
(CEEB, 1991).

The rising costs associated with higher education, the
changing U. S. and global economies and their impact on
available financial resources from federal agencies, and the
ability of public and private educational institutions to
award financial aid and scholarships to individuals -- all
affect who can and who cannot, and who does and who does not
pursue higher education following the completion of secondary
school. In what ways did these changing demographics affect

the CEEB SAT outcomes between 1988 and 1991?

 

Theory

While the decline in mean SAT scores has been documented
since 1964 (CEEB, 1977, 1980, 1981, 1982, 1983, 1984, 1985.
1986a, 1987, 1991), the implications of the observed decline
are less explicit. There are three potential sources of the
mean score decline: (a) changes in the characteristics of
the test; (b) changes in the demographics of the examinee
population: that is to say, a different group of examinees
takes the test than that for which the test was originally
intended; and (c) changes in the inherent characteristics --
for example, ”ability” or ”aptitude” —— of the general
population, independent of demographic characteristics (CEEB,
1977).

To investigate the first ”source” of change —- changes
in SAT characteristics -- is beyond the scope of this study
and its discussion is limited to the literature review at
this time. Since 1967, the concepts tested and the test
content of the SAT have changed to reflect changes in the
academic curriculum (Donlon, 1984). Changes in item formats
on the SAT have occurred infrequently, and only after
considerable and comprehensive research efforts by College
Board. and Educational Testing Service (ETS) staff and
researchers. The last major change in item format occurred
in 1974 when the Quantitative Comparison (QC) item type was
introduced and included with the Regular Multiple—Choice (RM)

item type (Braswell, 1991).

Item development and test development at ETS are
conducted under rigorous quality control procedures
throughout each phase of the process, including objectives
development, item specifications, reliability and validity
studies, differential item functioning research, test
blueprints and test specifications, and test form assembly
(P. Kenney & M. Zieky, personal communications, July, 1989).

Item equating places test items on a common scale and
controls for the effect of different examinee groups who
respond to different test items. Test score equating
similarly places test scores on a common metric, controlling
for the effect due to particular test forms. Successive test
forms are systematically equated to predecessor forms, taking
into account item difficulty, indices for Emmential item
bias, and item discrimination. (See Angoff, 1984; or Wilks,
1961, for a more complete discussion of item and test score
equating.) The content and item formats of the SAT have
changed considerably since its first administration in 1926.
For the purposes of this study, however, any changes in test
content and item format administered to the 1988 — 1991
college-bound seniors are negligible.

The latter two potential ”sources” of change -- changes
in demographic characteristics of the examinee population,
and changes in the inherent characteristics of the general
population —- are investigated in the dissertation through
empirical data analyses. Compositional changes in gender,

ethnicity, and parent's educational attainment are examined

through frequency summaries and tests of trend in proportions
between 1988 and 1991. Academic preparation and academic
aspirations are similarly summarized as indicators of
”interest characteristics.” To the extent that the SAT is a
measure of general readiness for college-level work, changes
in this underlying characteristic for the college—bound
population are examined. The SAT—Total score distribution
from 1988 to 1991 is described empirically. Mean SAT—Total
scores, significance tests for linear trend in mean SAT-Total
scores, and net mean SAT-Total score change are computed for
specified subgroups of the samples included in the analyses.
Different patterns of net mean score change and their
implications are discussed in the context of a changing

constituency of college-bound high school SAT examinees.

Overview

Chapter 2 contains a review of the relevant literature.
The structure of the data set, outcomes of the CEEB SAT, and
the design and strategy used to investigate the research
questions are discussed in Chapter 3. Chapter 4 addresses
the results of the statistical analyses and consists of two
sections: Section. 1 includes frequency summaries and
compositional analyses; Section 2 pertains to the analyses of
mean SAT scores. Chapter 5 presents an interpretation and a
discussion of the research findings. Limitations of the
study and recommendations for future research conclude the

dissertation.

CHAPTER 2

REVIEW OF THE LITERATURE

What Does the SAT Really Measure?

Perceptions of the CEEB and ETS on the SAT have changed
over time from a measure of academic achievement in specific
academic subjects to a measure of general readiness for
college—level work.enui developed aptitude (Angoff, 1986;
Boldt, 1986; Jackson, 1980).

According to Crouse (1985) and Hunter (1986), tests like
the SAT that cover a wide range of academic abilities predict
educational and economic success better than tests that cover
a single ability. This finding is consistent with research
on the dominance of general abilities (9) over specific
abilities in the prediction of performance (Thorndike, 1986).
Crouse (1985) analyzed the results from the National
Longitudinal Study (NLS) of the high school class of 1972 and
found that the SAT, as a measure of accumulated school
experience, tended to be predictive of first-year academic
performance in college. Hunter (1980, 1986) and Hunter and
Hunter (1984) provided extensive reviews and re-analyses from
three comprehensive validation data bases and found that
cognitive ability tests have higher predictive validity than
tests that measure discrete abilities. Although Crouse
questioned the practical utility and significance of the SAT
over alternative measures of academic achievement to predict

college success, both Crouse and Hunter provided persuasive

 

10

evidence that aptitude and cognitive ability tests are both
highly geloaded and that they have considerable predictive

validity for later success.

The Use of SAT in College Admissions

Many higher educational institutions have used SAT
scores as one of the selection criteria in the college
admissions. process. Some literature (Schaeffner, 1985;
Smith, 1980; Trusheim, 1984; Willingham, 1986; Willingham &
Breland, 1982) suggests that an increasing number of post—
secondary educational institutions are ”de-emphasizing” the
use of SAT scores in the admissions process and giving
greater consideration to an applicant’s high school record,
biographical information, personal statement, and areas of
interest.

Increasing numbers of colleges and universities may be
relying on information from a variety of sources, but
standardized test scores like the SAT score retain an
important role in college admissions. A 1979 survey among
four-year colleges conducted by the College Board and the
American Association of Collegiate Registrars and Admissions
Officers (AACRAO) showed that among 726 of its 981
respondents, the SAT and other standardized test scores were
considered in college admissions decisions (AACRAO & CEEB,
1980). Despite the criticisms of its utility in college
admissions (Gottfredson & Crouse, 1986; Slack & Porter, 1980;

Trusheim, 1984; Trusheim & Crouse, 1982, 1984), the SAT

 

 

11

provides a standardized measure which can be used to compare
different individuals who come from diverse backgrounds
(Jackson, 1980; Jacobson, 1986).

An additional consideration which enhances the
relevance, validity, and role of the SAT in the college
admissions process is grade inflation. Bejar and Blew
(1981), for example, found that the mean high school grade
point average consistently increased from 1964 to 1978,
during the same period when mean SAT scores steadily
declined. If grade inflation is still a viable construct,
then the SAT, in the presence of declining validity of high
school grade point average as an index of high school
achievement and as a predictor of college success, is

instrumental in predicting academic success in college.

Changes in Mean SAT Scores Over Time

ETS and. other independent researchers periodically
examine SAT scores. The present SAT scale-score is based
upon the scale for the first group of SAT-takers of 1942, to
which subsequent SAT forms have been successively equated.
The scale has a mean of 500 and a standard deviation of 100
(CEEB, 1986b). The mean of the SAT has changed over time.
As shown in Table 2, mean SAT scores declined by almost one-
half of a standard deviation on the SAT-verbal (SATrV) and
nearly one-third of a standard deviation on the SAT—

mathematics (SAT-M) subtests from 1967 to 1981.

 

12

Table 2

WW5

 

Year SAT-V ASAT-V SAT-M ASAT-M SAT-Total ASAT-Total

 

 

 

1967 466 492 958

1968 466 0 492 O 958 O
1969 463 -3 493 +1 956 -2
1970 460 -3 488 -5 948 -8
1971 455 -5 488 O 943 -5
1972 453 -2 484 -4 937 -6
1973 445 -8 481 —3 926 -11
1974 444 -1 480 -1 924 —2
1975 434 ~10 472 -8 906 -18
1976 431 -3 472 O 903 -3
1977 429 —2 470 -2 899 -4
1978 429 O 468 -2 897 —2
1979 427 -2 467 -1 894 -3
1980 424 -3 466 —1 890 —4
1981 424 O 466 O 890 O
1982 426 +2 467 +1 893 +3
1983 425 -1 468 +1 893 O
1984 426 +1 471 +3 897 +4
1985 431 +5 475 +4 906 +9
1986 431 O 475 O 906 O
1987 430 -1 476 +1 906 O

SAT-V = mean SAT-verbal score

ASAT-V mean change in mean SAT—verbal score
SAT-M = mean SAT—mathematics score

ASAT—M mean change in mean SAT-mathematics score
SAT-Total mean SAT-Total score

ASAT-Total = mean change in mean SAT—Total score

 

13

Beginning in 1982, mean SAT—M scores began to increase
slightly, as did mean SAT-V scores, but these slight
increases were insufficient to recover the observed decline
in mean SAT scores from 1967 - 1977. From 1980 to 1987, mean
SAT-M scores increased a net ten scale—score points. During
the same period, mean SAT—V scores increased a net six scale-
score points. Both mean SAT—V and mean SAT-M scores declined
a net 36 scale-score points and 16 scale—score points
respectively from 1967 to 1987. On the basis of mean SAT
scores alone, the SAT-takers of 1987 performed less well than

SAT—takers from 1967.

Sources of Change in Mean SAT Scores
Changes in mean SAT scores can be attributed to three
factors: (a) changes attributed.tx> test scale drift; (tn
changes in the constituency of SAT—takers; and (c) changes in
the underlying ”ability” distribution after accounting for

changes in the constituency of the examinee population.

W

Test scale drift is the extent to which a scale is
invariant over time (Angoff, 1971). The raw scores of Test A
are transformed to scaled scores on a specifically defined
common scale at time A. Test B is statistically equated to
Test A and also placed on the same scale at a later time B.
In the absence of scale drift, examinees should obtain

approximately the same scores regardless which test they

14

take. An upward scale drift is detected when the obtained
scale—score on Test B is greater than the obtained scale-
score on Test A for the same group of examinees.

Subsequent to the 1967-1977 decline in mean SAT scores,
an expert panel was appointed by the CEEB to seek
explanations for the decline in mean SAT scores. An integral
part of the investigation was the examination of psychometric
and statistical characteristics of the SAT (CEEB, 1977).

SAT forms have been successively equated since 1942, and
test scale drift has been a continuing concern. Various
methods of SAT form equating were researched to minimize the
overall error in equating successive test forms over time.
The result of this early research was the development of the
equating technique to ”braid” test forms. (See, for example,
McGee, 1961; or Wilks, 1961.) In a study of SAT scale drift
among test forms from 1948 to 1953, Stewart (1966) found that
the SAT-V scale had drifted upwards 20 scale-score points,
and that the SAT-V scale then remained stable from 1953 to
1963.

Modu and Stern (1975) found that the mean SAT-V scale
drift was about 14 points upward from 1963 to 1973, and that
the SAT-M mean scale drift was upward 17 points from 1963 to
1973. In a follow-up study, Modu and Stern (1977) found that
the mean SAT scale drift was approximately five scale-score
Points upward on the SAT-V from 1963 to 1973, and that during
the same period, the drift in SAT—M scale was upward 14

scale-score points. Although these scale-score drifts alone

15

were considered statistically non-significant (Beaton,
Hilton, & Schrader, 1977), these scale drifts worsened the
observed decline in mean SAT scores and, in effect, widened
the gap in mean SAT scores over time.

Different methods of equating tests can yield different
results. Could the upward scale drifts of the SAT have been
attributed to the choice of equating method? To examine the
accuracy of various methods used to equate tests, Marco,
Petersen, and Stewart (1983) compared the results of
different equating methods subjected to varying conditions.
In their study, non-linear and linear equating methods were
applied to tests with varying difficulty and respondents’
ability levels. They concluded that the most consistently
accurate results were obtained when item response theoretic
(IRT) methods were used.

In another study, Petersen, Cook, and Stocking (1983)
compared the effects of several equating methods on the
stability of SAT—V and SAT—M scales, and concluded that IRT
methods produced less scale drift for the SAT+V scale and
about the same scale drift for SAT-M scales as if
conventional linear equating procedures had been used. The
threat of scale drift has become less severe with the use of

more precise IRT methods to equate SAT forms (Donlon, 1984).

i . E .
According to the most recent statistics compiled by the

U. S. Department of Education and the Office of Educational

 

16

Research and Improvement (USDE & OERI, 1992), the enrollment
of pupils in all U. S. high schools at grades 9 - 12
decreased slightly from 12,893,000 students 1J1 1988, to
12,472,000 in 1990, increased again to 12,610,000 in 1991,
and is expected to increase through fall of the year 2002
with a projected enrollment of 15,206,000 students. Table 3
shows the numbers and proportions of U. S. high school
graduates and the percentage of high school drop-outs

relative to the population of 17—year olds from 1967 to 1991.

 

 

Table 3
7- ' —
Total Percent of 17-year‘ Percent of 17-year
Year Graduates old Graduates old Drop-Outs
1967 2,695,000 76.3 17.0
1979 3,043,000 71.4 14.6
1988 2,727,000 72.5 12.9
1991 2,485,000 75.6 12.5

 

Between 1988 and 1991, the proportion of first-time
college enrollees in all U. S. institutions of higher
education increased steadily from 13,055,000 in 1988, to
14,157,000 in 1991, and is expected to increase to 16,030,000
students in the year 2002 (USDE & OERI, 1992). To the extent
that these figures include college—bound seniors who took the
SAT from 1988 to 1991 and examinees who are expected to take
the SAT by the year 2002, the population of examinees has

grown in number and the number of college-bound seniors

17

projected to take the SAT will be substantial. The high
school drop-out rate has been declining, the high school
graduation rate has been increasing, and the numbers of
first-time degree enrollees in college have increased
(Beaton, Hilton, & Schrader, 1977; USDE & OERI, 1992).
Characteristics of those who take the SAT have changed
since 1967. Prior to 1967, most college-bound seniors who
took the SAT applied to traditionally selective and
prestigious institutions. Since 1967, and among students who
took the SAT, the diversity in the types of institutions to
which students have applied for admission has increased
(Beaton, Hilton, & Schrader, 1977; USDE & OERI, 1992).
W. In what ways did demographic
characteristics of college-bound seniors who took the SAT
between 1988 and 1991 change? Although mean SAT scores
appeared to stabilize in the early 19803, the characteristics
of examinees who took the SAT continued to change. The
percentage of women who took the SAT increased from 42.7% in
1967, 48.7% in 1972, 50% in 1976, to 51.7% in 1980. Table A—
l in Appendix A shows the numbers and proportions of men and

women who took the SAT between 1980 and 1987.

Eghnig_grgnn_ggmpgsitign. Table A-2 in Appendix A shows
the changes in ethnic composition of college-bound seniors
who took the SAT between 1980 and 1987. The proportions of
all college-bound seniors who took the SAT and who belonged
to African-American, American-Indian, and Puerto Rican ethnic

groups remained relatively stable between 1980 and 1987

 

 

 

18

(CEEB, 1980, 1981, 1982, 1983, 1984, 1985, 1987). Both the
proportions and numbers of Hispanic-Americans and Asian-
Americans who took the SAT steadily increased during this
period, while the proportions of White majority SATrtakers
declined moderately.

Sagigeggngmig_status. Table A—3 in Appendix A shows the
numbers and proportions of all college-bound seniors who took
the SAT and their reported annual parent income as an index
of socioeconomic status (SES) from 1980 to 1985. Until 1985,
the percentage of reported annual parent income from lower
income categories ($29,999 or less) decreased over time,
while the percentage of reported annual parent income from
higher income categories ($30,000 or more) increased over
time.

The accuracy of these self-reports as an index of SES
is, however, questionable. Although. increasing costs
associated with higher education and the decreasing
availability of financial assistance might have limited
college access to those students whose parents earned higher
annual incomes, this did not necessarily mean that greater
proportions of college—bound seniors who took the SAT came
from higher SES groups between 1980 and 1985. The median
reported annual parent income may have increased from year to
year, but the same, fixed categories of annual income are
typically used over several years with no adjustments for
economic change. For example, a $40,000 - $50,000 annual

income category represented the second highest income

 

l9

category, i.e., upper SES, between 1980 and 1985 on the
W (SEQ). The same category
of annual income represented only the fourth highest annual
income category, i.e., middle SES, beginning in 1987 on the
SEQ.

The accuracy of SES distributions as indicated by
reported annual income from 1985 to 1987 on the SILQ
overshadows the appropriateness of this information as an
index of SES. The accuracy of annual parent income
categories as an index of SES is therefore questionable.

Level of education and income tend to be positively
correlated with SES (USDE & OERI, 1991; 1992). Information
about parents’ highest level of educational attainment on the
SD_Q was first collected in 1987. Of the college-bound
seniors of 1987, 4% or 44,110 respondents indicated neither
parent held a high school diploma; 38% or 373,428 respondents
indicated their parents' highest educational attainment was a
high school diploma; 7% or 65,674 respondents indicated their
parents' highest educational attainment was an associate’s or
a two-year degree; 27% or 270,899 respondents reported their
parents' highest educational attainment at a: bachelor’s
degree; and 24% or 240,975 respondents indicated their
parents' highest level of educational attainment as a
graduate or professional degree (CEEB, 1987).

As shown in Table A-4 in Appendix A, consistently more than

20% of college-bound seniors from 1980 — 1987 reported a rank

 

 

20

of highest-tenth in their graduating class, and an increasing
percentage of more than 25% reported a class rank in the
second quintile. The percentage of college-bound seniors who
reported a class rank in the third quintile declined during
this period, but consistently captured more than 20% of each
annual cohort. College—bound seniors who reported a class
rank in the fourth quintile or lower comprised less than 5%
of all respondents between 1980 and 1987.

Table A-5 in Appendix A shows the numbers and
proportions of respondents from 1980 tx> 1987 who studied a
typical college-bound curriculum, defined by minimal years of
study in certain academic subjects. Both the proportions of
respondents who studied at least four years of English, and
the proportions of respondents who studied at least three
years of social science/history increased steadily from 1980
to 1987. The proportions of respondents who studied at least
three years of mathematics, and the proportions of
respondents who studied at least two years of a foreign or
classical language declined in the early 19805, but began to
increase during the nudrl9805. Data about course-taking
patterns in the natural sciences were unavailable for this
period.

Wag. Table A—6 in Appendix A shows the
proportions of planned college majors among all college-bound
seniors who took the SAT between 1980 and 1987. Intended
majors in Business comprised a large and increasing

proportion of college—bound seniors who took the SAT between

 

21

1980 and 1987. The proportion of respondents who planned to
major in Biological Science declined during the same period.
The proportions of college-bound seniors who intended to
major in the Physical Sciences increased, then began to
decline between 1980 and 1987, while an increasing proportion
of respondents intended to major in the Social Sciences. The
proportions of respondents with unspecified plans fluctuated
and increased to nearly 8% in 1987.

Educatignal_degree_ggals. Data in Table A-7 of Appendix
A show that the majority of college—bound seniors who took
the SAT aspired to a bachelor's degree between 1980 and 1987.
Approximately 25%, and an increasing proportion of
respondents, indicated their highest degree goals of a
master's degree. Less than 20% indicated highest degree
goals at the doctorate or professional degree level.
Consistently less than 6% of respondents held highest degree
goals of a two-year or associate's degree. Nearly 20% of

respondents were undecided about their highest degree goals.

3 J . 1.]. . .1 .

Mean SAT-V, SAT—M, and SAT-Total scores declined from
1967 to 1981. The observed mean score decline during this
period was 68 scale-score points on the combined SAT-V and
SAT-M subtests among college-bound seniors. According to
Modu and Stern (1975), the observed decline in SAT scores
combined with an upward scale drift resulted in a more

pronounced SAT score decline than is evident on the basis of

22

the observed declines alone. Upward scale drifts of 17
points on the verbal and 14 points on the mathematics
subtests were detected between 1967 and 1977. These upward
scale drifts meant that mean SAT scores actually declined 54
scale-score points on the SAT-V and 36 scale-score points on
the SAT-M between 1967 and 1977. During this time, mean SAT-
Total scores dropped 90 scale-score points among college—
bound seniors who took the SAT. The mean score decline was
greater than one-half of one standard deviation on the SAT-V
and nearly one—third of one standard deviation on the SAT-M
subtests for that period. Barely one-third of the 1977
college-bound seniors performed as well as half of the 1963
college—bound seniors (Beaton, Hilton, & Schrader, 1977;
CEEB, 1977).

Additional evidence from a composite study of Project
TALENT and the National Longitudinal Study of 1972 showed
that these changes in ability distributions were not unique
to the SAT. Similar declines in mean test scores on the Iowa
Tests of Educational Development, the Minnesota Scholastic
Aptitude Test, and the American College Testing Program from
the 19605 through the 1970s corroborated the 1967 - 1977 mean
SAT score decline (American College Testing Program, 1977;
American Council on Education, 1967, 1972; National Center
for Educational Statistics, 1982; Pepin, Knepper, Bales,
Bartell, Shulman, & Williams, 1982). The common theme of
mean score declines among samples of students drawn from the

same population during the same time period suggested an

23

underlying commonalty which could not be explained by
differences in testing programs alone.

Was the earlier SAT score decline of 1966 - 1976 simply
a reflection of change in the demographic composition of the
examinee population, or an indication of ”real” change in the
ability or aptitude of the general population? If greater
numbers of examinees who took the SAT came from disadvantaged
backgrounds, then declining means could be attributed to
changes in demographic compositions rather than to an erosion
in the ability or aptitude of the general population.

Did the decline in mean SAT scores, in fact, continue
throughout the 19805? In reference to the slight reversal in
the SAT score decline during 1980 - 1984, Turnbull noted
that, ”. . . only hindsight after a few more years will tell
us for sure whether or not the long slide [of SAT scores]

ended and reversed course in 1980. . . ” (1985, p. 9).

Summary

The summary statistics and proportions of all college—
bound high school seniors who took the SAT through 1987
indicated changes in demographic characteristics, academic
preparation, and academic aspirations. One limitation of
these descriptions is that they are based on the number of
students who chose to respond to a particular question on the
SDQ, in: infer changes across different characteristics on
the basis of these figures alone, e.g., changes in intended

college major by ethnic group membership, is potentially

 

24

misleading because there is no guarantee that the group of
respondents to a particular SQQ question is identical from
one question to another. An additional limitation of these
earlier aggregate descriptions is that interaction effects
and contextual interpretations are not documented explicitly.
How much of the earlier score decline could be attributed to
compositional change, and how much of the earlier score
decline left unexplained by demographic change have not been
ascertained.

A substantial number of college-bound seniors take the
SAT each year, and many higher educational institutions use
outcomes from the CEEB SAT as one criterion for college
selection, admission, placement, or guidance of prospective
students. The changes in observed mean SAT scores and in the
constituency and characteristics of SAT-takers from 1980 to
1987 have been documented. The observed decline in mean SAT
scores between 1988 and 1991 requires further examination of
the possible sources of this recent score decline.

Elements of the study, data structure, research
questions, and methodology are described in the next chapter.
Descriptive summaries and frequencies of characteristics are
examined for each annual cohort of college-bound high school
seniors who took the SAT between 1988 and 1991. These
characteristics include demographic information, descriptions
about academic preparation, and characteristics pertaining to
academic aspirations. Proportions and significance tests for

trend in proportions over time are computed. Mean SAT-Total

25

scores, significance tests for linear trend in the mean SAT—
Total scores, and net mean SATrTotal score change between
1988 anui 1991 are computed for specific subgroups of the
samples included in this study. The portion of the observed
score decline explained by changes in demographic
characteristics of examinees who took the SAT between 1988

and 1991 is assessed.

(l)

(2)

(3)

(4)

(5)

(6)

CHAPTER 3

ELEMENTS OF THE STUDY

Research Questions
In what ways did the demographic characteristics of
college-bound high school seniors who took the SAT
change between 1988 and 1991?
In what ways did the academic preparation of college-
bound high school seniors who took the SAT change
between 1988 and 1991?
In what ways did the academic aspirations of college-
bound high school seniors who took the SAT change
between 1988 and 1991?
In what ways did mean SAT scores of college—bound
seniors change between 1988 and 1991?
In what ways did mean SAT-Total scores for different
gender, ethnicity, parent education, and high school
class rank groups in the sample change between 1988 and
1991?
How much of the observed mean SAT-Total score decline
between 1988 and 1991 is attributable to changes in

demographic composition?

Structure of the Data Set

The total number of college-bound high school seniors

who took the SAT remained relatively stable during the most

recent decade. There are currently about 1,000,000 total

26

 

27

college—bound seniors who take the SAT each year. (See Table
A—8 in Appendix A.) Among the SAT—taking population are men
and women who are resuming post-secondary education, men and
women who are returning to college to pursue another field of
study, men and women who are returning to college to receive
additional training towards a new career venture, and men and
women who had postponed their post-secondary education until
a later time in their lives. Despite increasing numbers of
SAT-takers who are not high school seniors, the majority of

SAT-takers are college-bound seniors, as shown in Table 4.

Table 4
E . 1 i J] _ 1 i .

 

 

Year Percent
1987 75.9
1988 75.2
1989 75.9
1990 74.9
1991 73.9

 

The data for this study consist of a 10% simple random
sample of college-bound seniors drawn by ETS. The sample
includes only college-bound seniors who took both SAT-V and
SAT-M subtests of the SAT in academic year 1988, 1989, 1990,
or 1991.

Available cases with complete information on all

relevant variables were included in the analyses.

28

Statistical analyses were performed on the IBM 3090-200J
mainframe system with the SAS System version 6.07 Statistical
Package. The first phase of the analyses includes frequency
summaries and tests for trend in proportions over time. The
second phase of the analyses includes descriptions of the
SAT-Total score distribution, an examinathmn of mean SAT:
Total scores for specific combinations of gender, ethnicity,
parent education, and high school class rank groups, and an
analysis of adjusted mean SAT-Total scores for the four—year
sample. Table 5 shows the effective sample sizes used in the

analyses.

Table 5
EE . i J . 1 E J

 

 

Year Unrestricted Cases Selected Cases
1988 102,495 68,344
1989 116,057 77,368
1990 109,275 72,836
1991 109,654 73,150

 

Outcomes of the CEEB SAT

1 i i . . . .
The SLudenLDesctintiALLQuestionnaire (SEQ) is a self-
report survey of prior course work, current academic

standing, vocational and extracurricular interests, and a

29

variety of student background variables (Donlon, 1984). The
use of self—reported information for prediction is often
criticized for its questionable validity and reliability.

Baird (1976), however, states:

There is considerable evidence that self-reported grades
can be as useful as school-reported grades as predictors
of college grades. . . and. . . [these] measures of
accomplishment could be used for the early
identification of students with the potential for high
level accomplishment (p. 34).
One reason to use information from the SEQ is that a subset
of questions has consistently appeared since its first
introduction in 1972, and the same form of the SEQ has been
used from 1988 to 1991. Nearly 70% of the questions that
appear in present forms of the SEQ have appeared in similar
or identical forms in earlier versions of the SEQ (Donlon,
1984).

WWW Included in
the outcomes of the SEQ are the traditional demographic
variables of gender, ethnicity, parents’ highest level of
educational attainment, anxi annual parent income. Ethnic
groups consist of the following categories: .African—
American, American—Indian, Asian-American, Mexican—American,
Puerto Rican, other Hispanic-American, ”Other” unspecified
ethnic minority groups, and White.

Although an index of annual parent income is reported in
the CEEB SAT outcomes, this information is not used as an
index of SES in this study. The categories of annual parent

income on the SEQ are unadjusted for economic factors, and as

 

30

a measure of SES, are not necessarily comparable from year to
year or consistent with the changing distributions of annual
parent incomes. The relevance of self-reported annual income
on the SEQ as an index of SES, therefore, is questionable.
Data compiled by the USDE and the OERI (1992) indicated that
parents' highest level of educational attainment was
positively correlated with median annual income. Parents’
highest level of educational attainment tends to be a acme
stable and consistent index of SES than the CEEB's categories
of annual parent income. Therefore, only parents' highest
level of educational attainment is used as an index of SES in
this study. The categories of parents’ highest educational
attainment are: ”Elementary School,” ”Some High School,”
”High School Diploma,” ”Two-year/Associate,” ”Some College,”
”College ‘Degree,’ ”Somer Graduate School,” and
”Graduate/Professional Degree.”

Agademig_pxepazagign. Self-reported information about
students’ course—taking patterns, number of years of study in
English, foreign language, mathematics, natural science, and
social science or history in high school, overall high school
grade point average, and rank in graduating class are
examined.

Considering the potential effect of high school grade
inflation. on reported overall high school grade {point
average, rank in graduating class is used in the analyses as
an index of high school academic achievement instead of

overall high school grade point average. If high school

 

31

grade inflation is a legitimate concern, then the variability
of high school academic achievement on the basis of overall
high school grades will tend to be reduced: Most college—
bound high school seniors will tend to receive inflated
grades in high school. Although overall high school grade
point average is positively correlated with rank in high
school class in the samples of this study, the distribution
of overall high school grade point average is more negatively
skewed than the distribution of high school class rank.

Agademig_jmmugfigigna. The SEQ includes questionnaire
items about students' highest degree goals and intended
college major.

Intended college major consists of the following
categories: ”Biological Sciences and related areas”; ”Arts
and Humanities”; ”Business, Commerce, and Communications”;
”Physical Sciences and related areas”; ”Social Science and
related areas”; ”Other areas, ” which include
general/interdisciplinary studies, technical, and trade; and
”Undecided.”

Information about college-bound seniors” academic
aspirations is only tested for trends in proportions over
time. Aspirations and career interests are nebulous and
subject to change over time, particularly during one’s post-
secondary education. Boldt (1986) suggested that one's
performance on the SAT can affect his or her actualized
academic and vocational plans, and that intended and actual

goals are not necessarily identical. Information about

32

college-bound seniors' future aspirations can be informative,
but these data are only used in the analyses of proportions,
and they are excluded from the analyses of mean SAT-Total

scores between 1988 and 1991.

W

Predictive validity studies of SAT performance combined
with high school record for first—year college grade point
average have been conducted (Breland, 1979; Weitzman, 1982;
Wilson, 1983). In those predictive validity studies, SAT-M
and SAT-V components were treated separately or were
arithmetically combined to form one predictor, the SAT-Total

score.

SAT-verbal and SAT-mathematical scores have been used as
two distinct predictors. The two scores can also be
simply summed (or averaged) so that the SAT sum can be
treated as a single predictor (Donlon, 1984, p. 142).

The SAT—V and SAT—M subtests are positively correlated
(Pearson product-moment correlation, pXy = 0.65 ~ 0.70). A
high, positive correlation suggests that higher SAT-V scores
tend to be associated with higher .SAJywi scores.
Concurrently, lower SAT+V scores tend to be associated with
lower SAT—M scores. A combined administration of SAT-V and
SAT-M subtests yields a test twice the length of either
subtest alone, thus increasing reliability to 0.92 (Boldt,
1986; Coffman, 1966; Pruzek & Coffman, 1966; Rock & werts,
1979). The arithmetic sum of SATrV and SAT-M scores is used

in all analyses of mean SAT-Total scores.

33
Hypotheses and Methodology

0 - '00 ° r .rq ,. 0'0 0‘ Q‘fl'! «09' r. , -
E 1] _] I 1. 1 1 J . 1 1 1 E 1
WW?

There are several hypotheses: (a) the proportions of
women who took the SAT did not change between 1988 and 1991;
(b) the proportions of African-American, Asian-American,
Mexican—American, Puerto Rican, and other Hispanic-American
ethnic minority examinees increased between 1988 and 1991;
and (c) the proportions of examinees from parent education
groups of less than high school increased between 1988 and
1991. This set of questions is addressed through frequency
summaries and analyses of proportions among each of the
gender, ethnicity, and parent educational attainment groups
between 1988 anui 1991. Planned orthogonal contrasts for
linear and quadratic trends over time in the proportions of
gender, ethnicity, and parent's educational attainment are
tested for statistical significance. The associated

hypotheses are:

H (18‘): ‘P at 0, and

0 lin

‘P #0,

quad

with the alternatives,

Ha). -
H1 . Thin — 0, and

\Pquad : 0 ’

where 93in is a linear contrast to test for presence of trend

in the proportions of women between 1988 and 1991. and 92mm.

is an orthogonal contrast to test for presence of a

34

curvilinear trend in gender composition between 1988 and
1991. This set of hypotheses tests whether the proportions

of college-bound women who took the SAT changed between 1988
and 1991.
The associated hypotheses for change in the proportions

of ethnic minority groups are:

H ”b’: lem s o, and

0

W’ Z O,

3 quad

with the alternatives,
H Mb).

1 ‘I’jlin > O, and

‘Pj d<0,

mm
where qfiihlis a contrast to test for positive linear trend in
the proportions of each ethnic minority group j —— African—
Americans, Asian—Americans, Mexican-Americans, Puerto Ricans,
and other Hispanic minority groups —- between 1988 and 1991,
and “Gama is an orthogonal contrast to test for a declining
curvilinear trend in ethnic minority group j proportions
between 1988 and 1991. Separate tests are performed for each
ethnic group j.

The associated hypotheses for change in the proportions
of examinees from parent education groups of less than high

school are:

ac)
Ho : ‘l’lin S 0, and

q' 2 O,

quad

with the alternatives,

H MC): 9'

1 lin > O, and

‘I’quad < O,

35

where ‘I’ is a linear contrast and ‘I‘ is a quadratic

lin quad
contrast to test for change in the proportions of examinees
in each of the two lowest parent education groups —-
”elementary school” and ”some high school” -- between 1988

and 1991.

0 ‘°"'.00,.0. Q 0! 0.0 ‘0 0 .00 00. 9‘ 4 O-z!"
W?

This question is investigated through frequency
summaries and analyses of pmoportions among college-bound
seniors who studied a typical college—bound curriculum
minimally comprised of:

(a) four years of English;

(b) two years of foreign language;

(c) three years of mathematics;

(d) three years of natural science; and

(e) three years of social science and history.
Performance on the SAT tends to be positively correlated with
academic preparation. Academic rank in high school class is
an index of cumulative academic achievement, and the
distribution of rank in graduating class among college-bound
seniors in the sample is also examined between 1988 and 1991.
Declining proportions of college-bound seniors who studied a
college-bound curriculum, declining proportions of examinees
with high academic class rank, or increasing proportions of

examinees with low academic class rank could provide

36

explanations for the observed mean SAT—Total score decline

between 1988 and 1991. The associated hypotheses are:
H (2): ‘1' == 0, and

with the alternatives,
(2),
H1 . ‘Plin at 0, and

q’ t O,

quad

where ‘I’1i is a linear contrast and ‘P is an orthogonal

n quad
quadratic contrast to test for presence of trends in the
proportions of examinees who studied a college-bound high
school curriculum. Similar hypotheses apply to test for
presence of a linear trend in the proportions of examinees
with. high. academic class rank, and. separate tests are

performed for the high academic class rank group.

9W] _] 1 1. 1 1 J . 1 1 1 l
W?

Did the distributions of intended college majors among
college—bound seniors change between 1988 and 1991? Did

highest degree goals among college-bound students who took

the SAT between 1988 and 1991 change?

The associated hypotheses are:

Ba),

H ‘Pklin = 0, and

o
T):

quad = 0’

with the alternatives,

Ba).
H1 . ‘I’klin at 0, and

‘I’kquad at 0 ,

37

where ‘I’klin is a linear contrast and ‘quuad is an orthogonal

contrast to test for presence of trend in the proportions of
respondents in each of the k degree goal categories --
”Certificate,” ”Two-year/Associate,” ”Bachelor's,”
”Master’s,” ”Doctor/Professional,” or ”Undecided” —— between
1988 and 1991. Similar hypotheses to test for trends in the

proportions of intended college majors are:
(313)

HO : ‘I’mlin = O, and
‘l‘mquad = 0,
with the alternatives,
Him”: ‘I’mlin ¢ 0, and
‘quuad ¢ 0,
where ‘l‘mlin is a linear contrast and ‘l’mquad is an orthogonal

contrast to test for presence of trend in the proportions of
respondents in each of the m college majors -- Biological
Science and related areas; Business/Commerce/Communications;
Humanities and the Arts; Physical Sciences and related areas;
Social Sciences and related areas; Other miscellaneous areas,
which includes general and interdisciplinary studies, trade,
skilled crafts, and technical or trade; and Undecided.
O ‘ 09 ° r .r-. .- 0.0. u‘q! ‘ 0.‘ o o ‘0‘-
W1?

It is hypothesized that mean SAT-Total scores declined
between 1988 and 1991. This question is addressed by the

following set of planned orthogonal contrasts:

(4)
H0 : \Plin 2 O, and

9‘ S O,

quad

38

with the alternatives,

4
H ()z

1 \an < 0, and

9’ > O,

quad
where “Him is a linear contrast to test for a negative trend

in mean SAT-Total scores, and Tgmm is an orthogonal contrast
to test for a curvilinear trend in mean SATeTotal scores
between 1988 and 1991. This set of hypotheses tests whether
the mean SAT-Total scores of college-bound seniors who took
the SAT declined between 1988 and 1991. Mean SAT—Total
scores, variances, and indices of excess and skewness of the
SAT—Total score distribution from 1988 to 1991 are examined
for changes in the shape of the Sign-Total score

distributions.

000 . «e. 0 o o r 9‘ a“. - quO- 9‘ .--o '33

1313145121?

This problem is addressed through the analyses of mean
SAT-Total scores for various subgroups of the sample.
Significance tests for linear trend in the mean SAT-Total
scores and net mean SAT-Total score changes between 1988 and
1991 are computed by subgroups. The subgroups are defined by
several factors:

(a) single-factor categories of ”gender,” ”ethnicity,”

”parent education,” and ”high school class rank”;

(b) two-factor, crossed categories of ”ethnicity/parent
education,” ”ethnicity/class rank,” and ”parent

education/class rank”; and

39

(c) three-factor, crossed categories of ”parent education/
ethnicity/class rank” groups.

Specifically, is the net mean score change between 1988 and

1991 for the Asian—American group similar to the net mean

score changes for other ethnic groups? Do different patterns

of net mean score change emerge for different class rank and

parent education groups?

This problem is addressed through the analysis of
adjusted mean SAT-Total scores. Separate year regression
coefficients are estimated for a common prediction model. To
compute the adjusted means, the 1988 mean of each predictor
in the model is held constant, while regression estimates are
allowed to vary by year. Competing regression models are
analyzed as sensitivity tests, and trends in these adjusted

mean SAT—Total scores are examined.

Analyses
The data set is analyzed in two ways: (a) contingency
tables are used to summarize frequencies, compute

proportions, and test for presence of trends in the
proportions over time; and (b) mean SAT-Total scores and
standard deviations are computed by specific subgroups in the

four-year sample to detect significant trends in mean scores,

40

to summarize patterns of net mean score change, and to
examine trends in adjusted mean scores between 1988 and 1991.

The outcomes (Hi the CEEB’s .SEQ_ consist of many
categorical 'variables. Research. questions (1) - (3),
delineated earlier in this chapter, are addressed through
analyses of proportions and significance tests for trend in
the proportions between 1988 and 1991.

Trend tests in the proportions and in mean SAT—Total
scores are limited to two orthogonal contrasts for linear and
quadratic trends. Although four equally-spaced time
intervals permit a comparison of at most three orthogonal
contrasts, the interpretation of a third orthogonal contrast
for cubic trend is less meaningful, and therefore is not
pursued in the present study. Mathematically, a minimum of
two unique points defines a line, and a minimum of three
unique points defines a quadratic function. The presence of
a linear trend over four time intervals is substantively more
meaningful than the presence of a quadratic trend over only
four time intervals. For the analyses of mean SAT-Total
scores, only significance tests for linear trends in the
means between 1988 and 1991 are computed.

The effective sample sizes of this study are
sufficiently large and exceed 65,000 for each annual cohort.
Significance tests among proportions and means will tend to

show statistical significance for small differences. Except

where noted, conservative type I error rates (a-levels) are

41

used throughout the study in order to detect larger, ”real”

differences among these large samples.

Frequency summaries and proportions are examined for
each of the gender, ethnicity, and parent’s highest
educational attainment variables. The proportions of:

(a) college-bound women who took the SAT;

(b) college-bound seniors of African—American, Asian-
American, Mexican—American, Puerto Rican, or other
Hispanic-American ethnic minority groups; and

(c) college-bound seniors whose parent’s highest level of
educational attainment was less than a high school
diploma,

are tested for linear and quadratic trends over time.

i . 2. i] . i . .

Course-taking patterns of college-bound senior SAT
examinees are examined between 1988 and 1991. The
proportions of college-bound seniors who studied a typical
college-bound curriculum are tested for linear and quadratic
trends over time. The distribution of rank in graduating
class between 1988 and 1991 among college-bound seniors in
the sample is examined. The proportions of examinees’ class
rank (highest-tenth, second-tenth, second quintile, third
quintile, fourth quintile, and fifth quintile) are tested for

linear and quadratic trends over time.

42

The proportions of intended college majors in each area
(Biological Sciences and related areas; Arts and Humanities;
Business, Commerce, and Communications; Physical Sciences and
related areas; Social Sciences and related areas; Other
areas; and Undecided) are tested for trends over time. The
proportions of cmfllege-bound seniors’ highest degree goal
(Certificate, Two-year/Associate, Bachelors, Masters,
Professional/Doctorate, and Undecided) are tested for trends

over time.

. I' 1 . _ J . .1 .

Mean SAT-Total scores of college—bound seniors who took
the SAT between 1988 and 1991 are tested for negative linear
trend. Large sample sizes increase the power of the
significance tests, small differences are :more readily
detected, and statistical significance is expected. Indices
of excess and skewness of the SAT—Total score distribution
between 1988 and 1991 are examined for changes in the shape

of the SAT-Total score distributions.

: . 5. :1 . i _ 1 E 1 E 1
Mean SAT-Total scores, standard deviations, and subgroup
sample sizes are computed for each of the following groups
separately by annual cohort:
(a) gender groups,

(b) ethnic groups,

43

(c) parent education groups,

(d) high school class rank groups,

(e) parent education and high school class rank groups,

(f) high school class rank and ethnic groups,

(g) parent education and ethnic groups,

(h) parent education, ethnicity, and high school class
rank groups.

College—bound seniors who take the SAT are a self-
selected population, and range restriction in the responses
is evident. For example, less than five percent of college-
bound seniors reported a high school class rank in the lowest
40% of their class, and more than 40% reported a class rank
in the highest 20% of their class between 1988 and 1991.
Some categories of parent education and of high school class
rank variables are aggregated to provide sufficient subsample
sizes and to facilitate the interpretation of results.
Significance tests for linear trend in the means and net mean
SAT-Total score change from 1988 to 1991 are computed for
each subgroup.

. . . . .
Qn2SIJIukTflI___lkEMLTSQQI3_J2a1LIna_JEQZUALQQQ_JEL_Changfi_ELn

An adjusted mean SAT-Total score, statistically
controlled at the demographic composition of the 1988 cohort,
is predicted for each annual cohort in the study. The
presence or absence of a trend in these predicted values is
an indication of the extent to which the observed score

decline between 1988 and 1991 can be explained by change in

44

demographic compositions. Trends in separate year regression
coefficients among specific demographic groups indicates the
presence of time-by-demographic interactions. The trend in
adjusted means for such subgroups differs from the trend in

adjusted mean scores for the overall group.

CHAPTER 4

RESULTS

Section 1: Analyses of Proportions

1. i . .

Sam. The percentage of female SAT examinees was
consistently greater than 52% between 1988 and 1991. The
frequencies and proportions of males and females in the
sample are shown in Table 6 for each annual cohort included

in the analyses.

Table 6
. I . I E i J 1 i 1

 

 

Year Male Female
1988 0.478 0.522
(32,671) (35,673)
1989 0.476 0.524
(36,817) (40,551)
1990 0.475 0.525
(34,612) (38,224)
1991 0.475 0.525
(34,741) (38,409)

 

A two—tailed test for trend in proportions of female SAT

examinees indicated no statistical significance at a combined

a-level of 0.001 for two contrasts between 1988 and 1991

[(TS)lin = 1.193; (TS)quad = —0.51], and the hypotheses,
(1a)
H0 : ‘I’lin at O, and
‘i’quad at O,

45

46

were rejected in favor of the alternatives,
Ha) _
H1 . ‘Plin _ 0,

1‘

and
mad = 0.
Neither linear nor quadratic trend in the proportions of
females was statistically significant. The gender
composition remained constant between 1988 and 1991, and more
than half of the examinee population consisted of female
examinees, who on average obtained lower SAT—Total scores
than males.

SW. The proportions of self-
described ethnic minority group members in the sample changed
between 1988 and 1991. The proportions of White majority
examinees were consistently larger than the proportions of
all ethnic minority groups of college-bound seniors. Of non-
White ethnic minority groups, African-Americans comprised the
largest group, followed by Asian-Americans, Mexican—
Americans, and other Hispanic-Americans (neither Mexican-
American nor Puerto Rican). This pattern persisted from 1988
to 1991, and tests for trend in proportions revealed
statistically significant linear trends (combined a-level of
0.001 for two contrasts) for all ethnic minority groups with
the exception of ”Other” unspecified ethnic groups.
According to these results, the population of examinees
became increasingly disadvantaged over time on the ethnicity
variable. Proportional data, frequencies, and trend tests

appear in Table 7.

 

47

Saamofiummuoom

.Hoo.o no Hm>ma-o confinEOU m om u:EUAOAamflm

 

 

.oEHu Ho>o odouu Omumuomso no 585: no monomoua you.“ and» ooawouuo}?

omH.mm mmm.mm mom.mm mam.mm mmmcflemxm Hmuoe
Immo.+mc Immm.+mc Imom.mmc Iomm.mmv

.oo.v .m+.om- mmm.o mmm.o mom.o mmm.o muses
INHH.HV Iomo.Hc Imomc Immmc

va.m- «0.0 mHo.o vHo.o mHo.o mHo.o >ufloflcaum .nmauo.
Iomm.HV Imom.flc loam.ac ImmH.Hc

mm.o- .mo.oa emo.o mmo.o omo.o mHo.o daemomflm nmauo
lemme Imomc Immmc Immmc

+0.0- .mm.m HHo.o HHo.o oao.o moo.o amoam ounmsm
lfimfi.mc Immm.fic Imam.flv Imam.ac

om.o .mo.HH omo.o mmo.o «No.0 Hmo.o consumea-emofixmz
Immm.mc .mov.mv Imma.mv Imeo.mc

mm.m- .mm.mH mmo.o mmo.o mmo.o mmo.o cmoaume<-amflm<
Imomc Immmc Immm.Hc lemma

.mH.oH- .NH.OH- moo.o oHo.o mHo.o HHo.o cmaocH-cmuauwe<
Ifimm.mc .oma.mc Iamm.mc Imam.mc

mo.H .mv.a mmo.o mmo.o mmo.o mmo.o cmuHume<-cmufluua

omso sea

Imec Imev HmmH ommH mmmfl mmmfi

 

 

b wands

48

The hypothesis,

ub),

H ‘lein S 0

0
was rejected for each of the j ethnic minority groups tested
(African-Americans, Asian-Americans, Mexican-Americans,
Puerto Ricans, and other Hispanic-Americans) in favor of the
alternative,

1b
H ( ):
1

‘I’jlin > O.

No statistically significant quadratic trends were detected
among the five target ethnic minority group proportions
tested.

The increase in proportions of ethnic minorities was
most pronounced. for the Asian-American. group, and the
proportions of White majority respondents showed a
significant negative linear trend between 1988 and 1991.
Significant quadratic trends in proportions were detected for
the American-Indian group and White majority group,
indicating that the rate of the linear decline in proportions
for both of these groups decreased over time.

The proportions of ethnic minority groups traditionally
associated with lower SAT-Total scores increased, while the
proportions of White examinees and other ethnic groups
traditionally associated. with higher ,SATrTotal scores
decreased. The compositional changes in SAT-takers’
ethnicity implied that overall, the population of examinees
became increasingly disadvantaged on the ethnicity variable.

. 1 J 1 E , 1 . J . .

Parent’s highest level of education was used as an index of

49

SES instead of reported annual parent income categories on
the SEQ. Among respondents who reported different levels of
highest educational attainment between parents, the highest
level of education, regardless of parent’s gender, was used.
The proportions of respondents with highest parent’s
educational attainment of ”two-year degree” or less increased
between 1988 and 1991. Respondents whose parent’s highest
educational attainment was a bachelor’s degree or higher
decreased between 1988 and 1991. The largest proportions of
highest parent education were found among the
graduate/professional degree category, consistently more than
23% for each annual cohort. Significance tests for trend in
proportions showed positive linear trends among low parent
educational attainment groups (”some high school” or less),
and negative linear trends among high parent educational
attainment groups (”some graduate school” or higher). With
regard to parent education, the population of examinees
became increasingly disadvantaged over time. These data

appear in Table 8.

 

50

.Hoo.o no Hm>mH-o vacanEOU a on namoauficmflm

 

 

 

hzmoflumflumfim .oEflu Ho>o Ucouu ofiumupmso Ho Romain mo mocmmoua no“ ammo pmflmmDuoga

omH.mm mmm.mm mmm.bm vem.mm moodfismxm Hmuoe
Ammm.bac Aamm.bav Aomb.mav Ambm.mav

mn.o Amv.mu mmm.o bmm.o mvm.o vvm.o Hmcofimmououm\oumzoouo
Imam.ev xbmm.vv Imam.vv Abvm.vv

0H.H Aes.mn mmo.o woo.o mmo.o mmo.o Hoocom coho oEom
Amom.mav Ammv.mav Iamm.mav Ammm.vfiv

oa.o mm.Hu mam.o mam.o vam.o vam.o omoaaoo
AmmH.HHV Abmm.aav Aamm.aae Ammm.omv

Hm.Hn oo.H mmH.o mma.o mma.o mmm.o omoaaou oEom
xmvm.mv Hovm.mv Ammw.mc Ambv.mv

mm.ou mo.m mmH.o mmH.o mmH.o vma.o common Hmohum
Ammm.aav Amm>.HHV Amflv.mav Ammm.oav

me.o ms.a mma.o mmH.o oma.o mmH.o Hoocom seem
Amma.mc Ammm.av Amao.mv Amsm.av

mm.a Amm.m mmo.o mmo.o mmo.o mmo.o Hoonom coax oEom
AmmH.Hv Ammo.ac Immo.av xwmmv

mm.o «ma.m mao.o mHo.o mHo.o mao.o Hoonom Shoucofioam

DU

a. Ammo enamel HmmH ommH mmmH mmmH Hm>mq coaumosom

 

 

.1

51

The hypothesis,

H(1w: ‘V

0 so,

Mn
was rejected in favor of the alternative,

HIHC): ‘I’nn > 0,
at an a-level of 0.001 for two orthogonal contrasts for the
two lowest parent educational attainment groups. No
statistically significant quadratic trends were detected,
suggesting no change in the direction of significant linear
trends among the two lowest and the two highest parent
educational attainment groups. These data showed that the
proportions of parent education groups in the 1988 - 1991
cohorts changed among the two highest and the two lowest
parent educational attainment categories reported on the SEQ.
The proportions of lower parent education levels of
”elementary school” and ”some high school” both increased,
while proportions of higher parent education levels of ”some
graduate school” or ”graduate/professional degree” both
decreased between 1988 and 1991.

On the basis of parent educational attainment, the
population of examinees became increasingly disadvantaged
between 1988 and 1991. The proportions of examinees from low
parent education groups traditionally associated with low
.SATrTotal scores increased, while the proportions of
examinees from the highest parent education groups,

traditionally associated with high .SATrTotal scores,

decreased between 1988 and 1991.

52

W
In addition to information about college-bound seniors’

traditional demographic characteristics, information about
their course-taking patterns was examined” The most
frequently reported high school curriculum consisted of the
following:

(a) four or more years of English;

(b) two or more years of a foreign or classical language;

(c) three or more years of mathematics;

(d) three or more years of natural sciences; and

(e) three or more years of social sciences and history.

Proportions of college—bound seniors in the sample who

reported studying at least these levels of course preparation
in high school were computed for each annual cohort.
Proportions of college-bound seniors in the sample whose
academic preparation was less than the typical college-bound
curriculum in any of the five areas described above, (a)-(e),
were also tested for the presence of trends over time. An
increase in the proportions of respondents who studied less
than the typical college-bound curriculunl may be one
explanation for the decline in mean SAT-Total scores. Tests
for non-zero linear trend in proportions showed statistical
significance across all categories of academic study. The
proportions of respondents in the sample who studied less
than four years of English increased, and the proportions of
respondents who studied less than the typical college-bound

curriculum decreased between 1988 and 1991. Proportions of

53

respondents who studied a typical college-bound curriculum or
more increased.
The hypothesis,

(2)
0 : ‘P = O,

H lin
was rejected for the course-taking patterns of English,
foreign language, mathematics, science, social science, and
college-bound study in favor of the alternative,

m)

t 0.

Statistically significant negative linear trends in the
proportions were detected for respondents who reported
studying less than two years of foreign language, three years
of mathematics, three years of science, and three years of
social science. The proportions of respondents who studied a
college-bound curriculum increased between 1988 and 1991. A
statistically significant positive linear trend in
proportions of respondents who studied less than four years
of English was detected. Significant quadratic trends in
proportions were found among respondents who reported
studying less than four years of Emglish, three years of
science, or a college-bound curriculum. Table 9 shows the

proportions and frequencies of the sample and their self-

reported high school academic preparation.

.aoo.o wo Ho>oaud pocmnsoo m on unmoflumcofim

54

 

 

Saamoflumflumom .oEflo uo>o osouu ofiumuomso no Moon: mo oocomouq How umou ooamounozaA
omH.mb mmw.mb mmm.>m vvm.mm moocfieoxm Hoooe
Amma.mmv Ammo.bmv Am0b.mmc Aaom.amv
moo.vau som.ma Hem.o Hnm.o Hem.o mam.o ESHDOAHHSU ocsomuomoHHou
Aavm.oav Ammv.oav AmmN.HHV Hmmm.oav
bb.H Amm.bu ova.o vva.o mva.o va.o oocoflom Hmfioom mumox m v
Ammb.mac Imam.mav Amam.mac Amvv.mav
«oo.w «Hm.nan mma.o HmH.o mmH.o mmm.o oodoflom mumom m v
Ammv.me Homv.mv xvab.mv Ammm.mv
Hm.m Avm.mn bvo.o mvo.o mvo.o mmo.o mumumeocumz munch m v
Ava.mv Ammo.mv Amvb.mv Ammm.mv
HH.m «mm.mu mma.o vma.o mmH.o mmH.o mama cofiouom memo» N v
Ammm.mv Amwm.wv Ammm.mv Imam.bv
Aom.m 4mm.va mma.o mma.o HHH.o oaa.o nomaocm mumox v v
capo ass
Amev Amev Hmma omma mmma mama >ozum no mumo»

 

 

m OHQMB

55

These findings suggest that positive trends in the
proportions of fewer years of English study, of more years of
science study, and of study in a college-bound curriculum
were not strictly linear between 1988 and 1991. Adthough
increasing proportions of respondents reported studying a
college-bound curriculum, the rate of this increase
diminished between 1988 and 1991. One interpretation of this
finding is that more examinees studied fewer years of
English, despite the increase in proportions that reported
studying a typical college-bound curriculum, of which four
years of English study is one component.

An alternative interpretation is that the increase in
proportions of fewer years of English study reflected the
course-taking patterns of increasing numbers of non-English-
speaking high school seniors whose high school academic work
excluded English study. This interpretation is consistent
with the observed increase in proportions of college—bound
academic preparation in non-English academic subject areas
and the increase in proportions of ethnic minority examinees
between 1988 and 1991.

High school class rank, as an index of cumulative
academic achievement, was examined. More than 20% reported
an academic standing in each of the highest tenth and second
tenth of their high school class. More than 27% of
respondents in each annual cohort reported a class rank in
the second quintile. The third quintile captured more than

23% of respondents in each annual cohort between 1988 and

56

1991. The fourth and fifth quintiles, together, captured

less than 5% of respondents in each annual cohort. At a

combined a—level of 0.001 for two orthogonal contrasts, none

of the proportions of class rank revealed significant linear
trends between 1988 and 1991. Only the proportions of
respondents who reported a class rank in the third quintile
showed a statistically significant quadratic trend from 1988
to 1991.

The hypotheses for insignificant trends in the

proportions of respondents’ class ranks from 1988 to 1991,

H a): 9’ = 0, and

0 lin

3‘ = 0,

quad

were consistent with the data, with the exception of trend in

proportions among the third quintile, for which the
alternative,

Him): ‘I’quad at O,
was favored. Table 10 shows proportions, frequencies, and

trend tests for the class rank distributions between 1988 and

1991.

57

.Hoo.o uo Ho>oHn5 pmdflQEoo o no unnumuflcomm

 

 

SHHmoHummumum .oeflu uo>o Ucouu ofioouomso Ho umocfia Mo monomond HON ammo poHflmunotses
omH.mb mmm.mb mmm.bb vem.mm momdflemxm Hmooe
Aommv Reeve Amway Amovv
om.H ma.m boo.o moo.o moo.o woo.o oaflucmdo cuumm
Aabm.mv Ammn.mv Ammm.mv Aamw.mv
mm.H mm.o mmo.o nmo.o hmo.o mmo.o oafiucmso condom
Ammm.mav Hmow.bav Ammm.mH. Ammv.mav
Anv.m mH.m mvm.o mmm.o smm.o Hem.o oaflucwso ouflca
Amov.omv Ammv.omv Ammm.amv Ammv.mav
bH.o mm.mu mnm.o omm.o vmm.o emm.o oamocfiso ocooom
Avom.mfiv Anmm.mav AvHH.>HV Abvm.vav
mo.mu mo.o ham.o omm.o Hmm.o hHN.o coca ocooom
Ammv.mav AHHm.mHV Avmm.mav Avmm.vav
mm.mn vm.on Ham.o bam.o va.o mam.o nuoa unocmmm
no
a. $5 5:2; $2 ommfl $3. $2 scam 32o

 

OH magma

58

i . . .

MW. Between 1988 and 1991, the
leading choice of intended college major among college-bound
.SAT_ examinees was Business or Social Science, each
consistently chosen by more than 20% of each annual cohort.
Although Business comprised one of the largest groups of
planned college majors, a significant negative linear trend
in proportions was detected (combined a—level of 0.001),
indicating a decline in interest in this area over time.
Tests for trend in proportions of intended majors in Physical
Science, Humanities or the Arts, and proportions of Undecided
majors showed no significant trends. Significance tests for
trend in proportions of intended Biological Science or Social
Science majors were positive and linear (combined a-level of
0.001), indicating an increase in interest over time in these
areas of college study. The proportions of Other Area majors
showed.a: statistically significant negative linear trend.

The data appear in Table 11.

59

.aoo.o no am>ma-o omaaQEOU m um unmeaaacmam

 

 

haemomomfluoom .oEau uo>o ocoau Oaumaooso Ho Hmocma mo monomoua How umou Omaamuuozes
oma.mn mmm.mb wmm.bb vvm.mw moocfleoxm amuoe
Hamb.mv amab.mv aamw.mv amvm.mv
om.o vm.m Hmo.o Hmo.o mvo.o mvo.o coowooocs
avmo.av amoa.av amam.av ammv.av
mm.mu .ma.oau vac.o mao.o bao.o Hmo.o mou< Hocuo
amwm.bav amvo.bav aahm.mav ammm.vav
vo.mu .mo.w mmm.o vmm.o Hmm.o mam.o cosmaom Hmfloom
awvo.mv ammm.hv ammm.mv Amvv.bv
mo.a mn.o oaa.o moa.o moa.o moa.o muH<\mofiufldmsdm
avvo.bav aamn.mac AHNH.HNV amma.mav
«ow.mu +mm.amu mmm.o mmm.o mnm.o omm.o monoEEou\mmodmmsm
ammv.aav amom.aav ahva.mav ammo.aav
mv.a mv.mn bma.o mma.o bma.o mma.o oozoflom acummmnm
abmm.vav awvh.mav amwb.mav ammm.oav
Asb.m «mv.ma mma.o mba.o mma.o oma.o cocoaom advamoHOAm
DO
a. last asamev amma omma mmma mmma some: mmmaaoo

 

aHqfluaIuduAdduIuuuauqaHIHqIAuaHuuuquuuufiluqaqudadHN

aa magma

6O

Highe§L_degzee_aapiraLiQnS. The majority of respondents
in the sample aspired to a bachelor’s degree or higher.
Tests for presence of trend in proportions for a combined a-
level of 0.001 for two orthogonal contrasts showed
statistically significant negative linear trends among
respondents who aspired to a certificate, a bachelor’s
degree, or were undecided about their degree goals. No
significant change 'was detected. in the jproportions of
respondents whose degree goal was an associate or two-year
degree. .A statistically significant positive linear trend
(combined a-level of 0.001 for two orthogonal contrasts) was
detected among college-bound seniors aspiring to masters
degrees and doctor/professional degrees.

From 1988, consistently more than 40%, and by 1991, more
than 50% of respondents aspired to advanced graduate degrees.
Table 12 shows the proportions of respondents’ highest degree

goals.

61

.HO0.0 MO H¢>¢HI5 UwGMQEOU m um UCMUMMMgfim

 

 

xgmofiumflumbm .mEflu um>o ccmuu ofiumucmso .Ho ummcfla mo wocmmmua “Om umwu pwHwMuuozeg

oma.mb wmm.mb mmm.bb wvm.mm mmmcflemxm Houoe
Aomv.mav Ambv.mav Ammm.vav AomH.mHV

ma.o «mm.mu vma.o mmH.o mma.o mmH.o pmpfiomoco
“mom.mav Abvm.vav Amvm.vav Achm.mav

mm.o «mm.va cam.o oom.o me.o HmH.o HmcoflmmwMOHm\Houooo
Amhm.amv ANNH.HNV Ammm.amv Ammm.mav

mm.oa «mm.m mmm.o omm.o mmm.o wbm.o wumummz
”www.mav Aomm.omv Abvb.mmv Abbw.omv

mm.ou iv>.vflu mmm.o Nmm.o «mm.o vom.o mHonnomm
Ammv.av Abmv.av Abvm.av Abmm.av

mm.o mm.o omo.o omo.o omo.o omo.o mmummo ummxum\mumfioomm<
Amom.av Ambm.av Ambb.av Abbn.av

m>.Hn amm.vu mmo.o mmo.o mmo.o mmo.o mumofiufiuumu

nflaAmev HzfiAva Hmma omma mama mmma Hmow mwummo

 

NH magma

62

On the basis of either gender or high school class rank
compositions alone, the SAT, examinee population remained
unchanged between 1988 and 1991. More than half of each
annual cohort in the study was comprised of females and of
respondents with class rank in the top 40% of their class.
On average, males tended to obtain higher SAT-Total scores
than females, and on average, examinees with higher class
ranks tended to obtain higher SAT-Total scores than examinees
with lower class ranks. These differences maintained, and no
significant variation in these patterns was detected between
1988 and 1991.

Significant changes in composition were detected for
both ethnicity and parent education variables. The number of
respondents from ethnic minority groups and from low parent
education groups, traditionally associated with low SAT-Total
scores, increased over time. In general, the population of
examinees became increasingly disadvantaged on ethnicity and

parent education variables.

63

Section 2: Analyses of Means
_ J . .1 .
Means and variances of SAIL-Total scores, indices of
excess and skewness of the fiAI-Total score distribution for

1988 to 1991 are shown in Table 13.

Table 13

_ J . .1 . i

 

 

 

*An alternative form of kurtosis is excess =

Yeax' Mean SAT—Total Variance Skewness Excess*

1988 923 43,010 0.1563 —0.4609

1989 923 43,616 0.1745 -0.4544

1990 920 44,706 0.1721 —0.4700

1991 914 44,610 0.2008 -0.4329
6<x-u)‘

4

0'

Planned orthogonal contrasts for negative linear and
quadratic trends in mean EAT-Total scores over time for a

combined a-level of 0.001 for two comparisons,

no”): Tun 2 o, and
‘I’quad S 0,

with the alternatives,

H1”): Wm. < o, and
unad > 0,

resulted in the rejection of H0 in favor of the alternative

[(TS)1h1== —8.254; (TS)mmd = 3.351]. Linear and quadratic

trends in mean SAT—Total scores were both statistically

64

significant, showed a linear decline in mean SAT-Total scores

over the four-year period, and indicated a steeper rate of

decline in mean scores over time.

The indices of skewness and excess for the fiAIrTotal
score distributions between 1988 and 1991 suggested the
following:

(a) as the indices of excess fluctuated and tended toward
zero, the score distribution became less broad;

(b) as the indices of skewness fluctuated and increased
positively, the distance from the mean to the mode of
the distribution in standard deviation units tended to
increase. A higher positive index of skewness indicated
a more positively skewed distribution of scores.

For the four annual cohorts included in the analyses, the

SAT-Total score distribution became increasingly positively

skewed with negative excess, relative to the normal

distribution. As the SAT-Total score distribution became
more positively skewed, the mean became the largest of the
values of central tendency. Table 14 lists SAT-Total score
medians and modes by annual cohort as alternative measures of

central tendency to the mean.

 

65

Table 14
WW

 

 

Year Mean Median Mode
1988 923 920 860
1989 923 920 860
1990 920 910 880
1991 914 910 900

 

The medians of the fiAI—Total score distributions from
1988 to 1991 showed a net score decline of ten scale-score
points, comparable to the observed decline in the means. The
modes of the SAT-Total score distributions for 1988 to 1991,
however, revealed a 40-point increase in the most frequently

attained EAT-Total score.

W3

Table 15 lists the variables which were included in the
analyses of means. Some categories of parent education and
high school class rank were aggregated to provide sufficient
subsample sizes for crossed factor cells, e.g., American
Indian examinees in the lowest 60% of their class with parent
educational attainment of high school diploma or less.
Collapsing some categories within a variable reduced the
number of distinct subgroups, but provided a more coherent

and substantively meaningful interpretation of the results.

Table 15

66

. l] i . J E

 

Variable

Definition and values

 

Year

Ethnicity

Gender

Parent
Education

Class Rank

= Sample year - 88 [YEAR = 0, 1, 2, 3]

oAfrican-American [BLCK = 1]
oAmerican-Indian [AMIN = 1]
oAsian—American [ASIA = l]
oMexican—American [MEX = 1]
oPuerto Rican [PRIC = 1]
oother Hispanic—American (neither
Mexican nor Puerto Rican) [HISP = 1]
oWhite [WHIT; all AMIN ~ OTHR = O]
-"Other' unspecified ethnic group
[OTHR = l]

omale [RSEX = 1]
ofemale [RSEX = 0]

oHigh School or Less: [PEd = 12]
Elementary School
Some High School
High School
~2—year/Associate Degree: [PEd = 14]
2-year or Associate Degree
Some College
Business or Technical School
oCollege [PEd = 16]
oGraduate/Professional Degree: [PEd = 20]
Some Graduate School
Graduate/Professional Degree

oHighest Tenth [RANK = 95]
oSecond Tenth [RANK = 85]
oSecond Quintile [RANK = 70]
oLowest 60%: [RANK = 30]

Third Quintile

Fourth Quintile

Fifth Quintile

 

1' . 1

Results from. frequency summaries. and. compositional

analyses were used to determine the specific categories of

67

parent education and high school class rank to be aggregated
for subsequent analyses. Each of the four resulting parent
education categories captured approximately 25% of each
annual cohort after aggregating some of the nine originally
reported educational attainment levels. High school class
rank categories were similarly treated and produced four
categories to capture approximately equal proportions of
respondents from each annual cohort.

The results from analysis of variance by regression of
the SAT-Total scores in the pooled four-year sample, using

”YEAR" as the only predictor, are shown in Table 16.

Table 16

J . E . E

 

 

 

Source df Sum of Squares Mean Square P
Value
Model(YEAR) 1 3076395 3076395 70
Error 291696 12833751801 43997
Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R? = 0.02%

Less than one percent of the variance in SAT-Total
scores for the pooled sample was explained using 'YEAR' as a
predictor 'variable. A. statistically’ significant. F-test
indicated the parameter estimate for ”YEAR” was non-zero in
this model. The observed net mean decline of nine scale—

score points for the pooled sample over a four-year period

68

indicated the presence of a substantive decline. The ”YEAR"
variable alone, however, was an insufficient predictor of
SAT-Total scores.

Additional models were tested and their partitioned
sources of variance by regression are described. Each of the
"models” outlined below includes main effects by time, by
single- or multiple—factors, and their interaction effects
with time.

The analyses of means are limited to three crossed
factors and their interactions with time to udnimize data
management problems encountered in early analyses, and to
facilitate interpretation of the findings. Preliminary
analyses for single factors of gender, ethnicity, parent
education, and high school class rank revealed that class
rank (26.56%), parent education (10.71%), and ethnicity
(8.34%) were the three factors with the largest proportion of

explained variance in SAT—Total scores for the pooled sample.

5' J _ : .
Gender; Table 17 shows the analysis of variance by
regression results for the model with gender, year, and year—

by—gender interaction.

 

69

Table 17

J . E . E i

 

 

 

Source df Sum of Squares Mean Square P
Value
Model 3 248459598 82819866 1919
YEAR 1 3076395 3076395 71
RSEX 1 245053848 245053848 5678
YEAR/RSEX 1 329355 329355 7
Error 291694 12588368598 43156
Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 1.94%

Results from the analysis of variance indicated that the
time-by—gender interaction effect was statistically non-
significant, but that main effects for gender and for year
were non-trivial. Although there appeared to be significant
differences between males' and females' mean m-Total
scores, the mean score differential was constant between 1988
and 1991. The FHtest for main effect year indicated that
annual mean SAIL-Total scores for the pooled sample were
statistically different across time.

Table 18 shows raw mean SAIL-Total scores, standard
deviations, and sample sizes by gender for separate years.
The computed z-tests for trend in the means showed
significant negative linear trends (p<:.0125) for both males
and females, with a more pronounced negative trend for males

than for females.

H

70

F

 

 

Hvu.vm Nam.vm uwm.mm Humrmm
m.oa: ammm.>u Amh.mamv Awm.mamv Aoo.mamv Amm.oamv mam:
mm.mvm mo.omm mm.mmm om.mmm
mov.mm www.mm Hmmtgv mum.mm
v.mn «oom.vu Amm.momv Ao>.momv Amm.oomv AH¢.oomv mHmEmm
mo.mmw 5v.mmm oo.mmw wv.mmm
wocmcu cfia
:82 662 5.5 33 omma mmma mmma xmm

 

q 0:.

ma manna

71

Preliminary analyses of mean SAT-Total scores for gender
groups, crossed with other factors, indicated no systematic
pattern of gender differences. The results from both Tables
17 and 18 show that, on average:

(a) males consistently scored more than 50 scale—score
points higher than females;

(b) mean SAIL—Total scores for males and for females
declined; but

(c) males' and females’ mean SAT-Total score declines were

parallel and did not differ significantly between 1988

and 1991.

No compositional changes were detected between 1988 and 1991
by gender groups. The absence of a significant time-by-
gender interaction in mean SAT-Total scores and the apparent
stability of the gender composition implied that gender was
not a significant source of explanation for the observed mean
SAT—Total score decline between 1988 and 1991. Gender was,
therefore, excluded as a: factor from further analyses of
observed group means over time.

Ethnig_£nxnuL_membership. Significant compositional
changes in ethnicity were detected between 1988 and 1991,
particularly for Asian—Americans, Mexican-Americans, non-
Mexican/non-Puerto Rican Hispanic-Americans, and for African-
Americans.

Table 19 shows the analysis of variance by regression

results. The nmdel includes main effects for year, eight

72

ethnic groups represented by seven indicator variables, and

year-by-ethnicity group interactions.

Table 19
J . E . E 1 . i

 

 

 

Source df Sum of Squares Mean Square P Value
Model 15 1070882053 71392137 1770
YEAR 1 3076395 3076395 76
Ethnicity 7 1066847376 152406768 3778
YEAR/Ethnic 7 958281 136897 3
Error 291682 11765946144 40338
Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 8.34%

The F—tests for interaction effects and main effects
using sequential sums of squares indicated a statistically
non-significant time-by-ethnic group interaction effect, and
significant main effects for ethnic groups and for year.
Less than 10% of the variance in SAIL—Total scores was
explained by year and ethnic group main effects, and year-by-
ethnic group interactions. Significant main effects for
ethnic group and for year suggested stable differences across
time between different ethnic groups.

Table 20 lists the raw mean SAT-Total scores, standard
deviations, and sample sizes by year for each of the eight

ethnic groups.

73

 

 

New.” mmm mmm mmm
Amm.mmmv Amm.mmmc Amm.mm~c Lom.mmmc

m.mH- Hmm.H- mm.mmm mo.mmm mm.mmm mm.wom mmeo
mum.vm mmk.wm mow.mm mmk.mm
ivo.Homc 1mm.Homv Amo.oomv imfl.mmflc

m.m- .Hvo.m- Hm.omm mm.vmm mm.mmm Hm.mmm eHmz
mmu.~ mmm.~ mmm.~ mm~.~
Amm.momc Aflm.momc Amm.momc AmH.momc

m.mH- .mmm.m- Ho.mmm mo.amm mm.mmm mm.amm mmHm
vmm ARR mmk mwm
lmo.mmflc 1mm.mmac 1mm.omfiv Amm.mmfic

m.m mmm. mm.omm mm.mmm mm.mmm mm.amm onm
~m~.m mmm.~ mmm.~ mvv.~
Amm.mmfiv imm.mmfic iom.mmac 1H~.mmfic

N.H- mmm.- mm.mmm Hm.mmm mm.mmm No.0mm xmz
Hmm.m Nev.w 6mm.m mvm.m
Amm.mmflc Amm.mmac Aom.HmHV Lom.mmflc

v.0H- .mmm.m- mm.mvm oo.mmm mm.mmm Hm.mmm xoqm
mmk.m Gmw.m mmm.m mwm.v
Amm.mmmc Aom.Hmmc Amm.mmmc Amm.ommc

m.m mmo.m mm.mmm mm.mmm mm.mmm mm.mvm «Hm4
mmm men mmm.~ VGA
Amm.mmfic 1mm.~mflc Amm.mmflc 1mm.omfic

N.m omm. mm.mmm mm.omm «H.0mm NH.mmm sz<

wmcmnu c
cggzuwz Hilmev Hmma ommH mmmH mmmH meHonmem

 

om magma

74

The computed z-tests for trend in the means showed
statistically significant (p;: .0125) negative linear trends
for African-Americans, Hispanic-Americans (neither Mexican-
American nor Puerto Rican), and Whites. The most pronounced
negative linear trend was detected for Whites, although the
net mean SAT—Total score change between 1988 and 1991 was
only 5 scale-score points.

For each annual cohort in the study, the Asian—American
mean SAT-Total score was consistently higher than the mean
scores of all other ethnic groups. The African-American mean
was consistently the lowest mean SAT-Total score in each
annual cohort. American-Indians', Puerto Ricans’, and Asian-
Americans' mean SAT-Total scores showed a net mean increase
of more than seven scale-score points between 1988 and 1991,
a net mean change opposite in direction from the White
majority and other ethnic minority groups.

The mean SAT—Total score for 'Other' unspecified ethnic
group examinees showed the largest net mean decline of 17
scale—score points between 1988 and 1991. Hispanic-
Americans' and African-Americans’ mean SAT-Total scores also
showed net mean declines of more than 10 scale-score points
between 1988 and 1991.

Although the ethnic group composition of college-bound
high school seniors who took the SAT changed between 1988 and
1991, White examinees still comprised more than 70% of each
annual cohort. The net mean SAT-Total score decline for

White examinees, combined with the net mean SAT-Total score

75

decline among other ethnic minority groups, attenuated the
net mean SAT—Total score increases found among American-
Indians, Asian—Americans, and Puerto Ricans between 1988 and
1991.

W Table 21 shows the

partitioned sources of variance by regression analyses.

Table 21
1 . E . E i .

 

 

 

Source df Sum of Squares Mean Square F
Value
Model 3 1374556414 458185471 11660
YEAR 1 3076395 3076395 78
PEd 1 1369414643 1369414643 34849
YEAR/PEd 1 2065375 2065375 53
Error 291694 11462271782 39295
Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 10.71%

The analysis of variance results indicated the presence
of a significant year—by-parent education interaction effect
in the pooled four-year sample. The change in mean SAT-Total
scores between 1988 and 1991 varied between different parent
education groups. Table 22 lists the raw means, standard
deviations, sample sizes, net mean SAT-Total score change
between 1988 and 1991, and computed z-tests for trend in the
means for each of the four parent education groups, by

separate year.

76

 

 

 

mmm.~m mom.mm mmm.mm kam.~m
Amv.ofimv AHm.momc 1mm.momv Aom.momv Hmcommmmuoum
v.m mom.H Hm.moo.H mm.moo.a mo.moo.a mm.moo.a \mumsomuo
mom.m~ mmv.m~ mmm.GH mwe.v~
Aom.mmfic 1mm.mmac Amm.mmflc Amfi.mmfiv
0.6- .ofiv.m- mo.oem om.mmm oo.mmm oa.mmm mmmflaoo
www.mm kmm.mm www.mm www.mm
AHm.HmHV 1mm.mmac Amo.mmac me.Hch mumfloommm
m.mH- .mfim.m- mm.amm om.mmm mm.omm ma.vmm \ummm-m
mam.m~ mmm.v~ www.mm mmv.m~
Afim.mmfiv 1mm.Hch Amm.mmfiv Amo.mmfiv mmmq no
m.NH- .mom.m- mm.mfim OH.mHm mm.mmm mv.mmm Hoonum comm
wmcmnu : ZOHHEUDQM
:mglsmz Hzlmev HmmH ommH mmmH mmma azummm
o o .0 o: .o c 0.... ..- o ..c) ..u

NN magma

77

Computed z-tests for presence of linear trend in the
means showed significant (p;: .0125) negative linear trends
for parent education groups at the college level or lower.
No significant linear trend was detected for the highest
parent education group of graduate/professional degree, but
this group showed a net mean SAT-Total score increase of two
scale-score points between 1988 and 1991, a change opposite
in direction from the other three parent education groups.

Mean SAT-Total scores of parent education groups of
”college degree" consistently declined over the four—year
period, with a net mean score decline of six scale-score
points between 1988 and 1991. For parent education groups of
less than college, the net mean score decline was 12 scale-
score points. More broadly:

(a) The higher the level of parent educational attainment,
the higher the mean SAT-Total score. This pattern
maintained for each annual sample in the study.

(b) The parent education gap in mean SAT-Total scores
between ”high school or less” and
”graduate/professional" parent education groups was
nearly 200 scale-score points between 1988 and 1991.
This gap appeared to increase over the four-year period.

(c) The lowest parent educational attainment group and the
highest educational attainment group showed the largest
net score decline and the largest net score increase,

respectively, over the four—year period.

78

(d) The compositional changes of parent education groups
between 1988 and 1991 indicated that increasing
proportions of examinees from lower parent education
backgrounds obtained lower mean SAIL-Total scores.
Examinees from the highest parent education backgrounds
obtained the highest mean SAIL—Total scores, but the
proportions of these groups declined over time.

High_sghggl_gla§s_rank. Table 23 shows the partitioned

sources of variance by regression analyses.

Table 23

J . E . E J 1 E

 

 

 

Source df Sum of Squares Mean Square F Value
Model 3 3409545040 1136515013 35165
YEAR 1 3076395 3076395 95
RANK 1 3406307984 3406307984 105396
YEAR/RANK 1 160661 160661 5
Error 291694 9427283156 32319
Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 26.56%

This model of SAT—Total scores with high school class
rank, year, and year-by—rank effects, accounted for the
largest proportion of explained variance in SAT-Total scores
relative to other models with single—factor effects. The
year-by-rank interaction effect was not significant, although

significant main effects for year and for rank were detected.

79

Table 24 lists the raw mean SAT-Total scores, standard
deviations, and sample sizes for each of the four class rank
groups, separately by year. Net mean SAT-Total score change
between 1988 and 1991 and significance tests for presence of
linear trend in the means were computed for class rank
groups.

Significant (p< .0125) negative linear trends in mean
scores were detected for each of the four class rank
categories. The most pronounced negative linear trend was
detected for the lowest 60% class rank group. This same
group also showed the largest net mean SAT-Total score
decline between 1988 to 1991, relative to other class rank

groups .

80

 

 

www.mw www.mw www.mw vmm.vm
Amw.mwav Amv.mmav Ama.mmav AHH.m>HV woa
«.m- .mmm.m- mm.mmo.H om.moa.a mm.¢oa.a mm.moa.a ummnmmm
vom.mm mmm.m~ www.mm mmm.w~
Amm.mmav Avm.mmav Amm.omav Abm.mmav wOH
H.6u momm.mu >v.mmm mm.mmm mm.amm mm.o>m pcoomm
mcv.om mmv.om Nmm.am www.mm
AHm.v>Hv Amw.mmav Aom.HmHV Aoa.mmav maflucflso
H.5I «mmm.vu ma.mmm mm.omm Hv.¢mm mm.mmm pcoomm
mmm.mm mom.om mom.~N mmm.m~
ANH.NmHV Aom.mwav Amo.mwav Aom.mmav woo
N.HH- mamm.bu mn.mbm mo.omm mm.vmm mm.mmn umm3oq
wmcmcu cma
cmgsowz Amev Hmma omma mmma mmma xz<m

 

vm Gamma

81

Patterns of change in the mean SAT—Total scores for
class rank groups indicated the following:

(a) The higher (or better) the class rank, the higher the
mean SAT-Total score. This pattern maintained for each
of the four annual cohorts in the study.

(b) Among the three lowest class rank groups (”lowest 60%,”
”second quintile,” and ”second 10%”) the higher the
class rank, the smaller the net mean fiAIrTotal score
decline. The pattern of net mean score decline was
interrupted by the class rank group of ‘highest 10%.”
This highest class rank group also showed the largest
standard deviations compared to time three lower class
rank groups.

(c) The class rank gap between ”lowest 60%” and ”highest
10%” class rank groups revealed a difference of more
than 300 scale—score points in mean SAT-Total scores for

each annual cohort.

W
W Table 25
shows the partitioned sources of variance by regression

analyses.

82

 

 

 

Table 25

;oo. ' o . ..o ‘ o 'o, ‘0 0. . u n o o

X23:

Source df Sum of Squares Mean Square P Value

Model 31 2154508120 69500262 1898
YEAR 1 3076395 3076395 84
PEd 1 1369414643 1369414643 37390
Ethnicity 7 766097960 109442566 2988
PEd/Ethnic 7 13180112 1882873 51
YEAR/PEd 1 1116955 1116955 30
YEAR/Ethnic 7 1014275 144896 4
YEAR/PEd/ 7 607789 86827 2
Ethnic

Error 291666 10682320076 36625

Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 16.78%

Analysis of variance outcomes showed that neither ”year-
by-parent. education-by-ethnicity” run: ”year-by-ethnicity'
interaction. effects ‘were statistically significant. A.
significant ”year-by-parent education” effect was, however,
detected, consistent with the earlier analysis of variance
outcomes with the single-factor model of parent education.
”Parent educationrby-ethnicity” group interaction effects
were statistically significant, suggesting stable differences
between ethnic groups matched on parent education.

Table B-1 in Appendix B lists the raw mean SAIrTotal
scores, standard deviations, sample sizes, and computed 2-
tests for linear trend in mean SAT-Total scores for ethnic

groups, matched by parent education levels for each annual

83

cohort 1J1 the study. Statistically significant negative
linear trends at the p<:.0125 level in mean SAT-Total scores
were detected among White examinees with ”high school or
less” and '2-year/associate” parent education backgrounds,
and ”Other” unspecified. ethnic: minority' examinees ‘with
”graduate/professional” parent education backgrounds.

Table 26 shows the net mean SAT-Total score change
between 1988 and 1991 for ethnic groups matched by parent
education levels. Negative quantities represent a net mean
score decline between 1988 and 1991; positive quantities

represent a net mean score increase during the same period.

 

 

Table 26
\‘ V‘o. , - 0 v 0 ‘ iol" '- ‘0 0 o .0 9!
Dreams
High 2—year/ College Graduate/
School Associate PrOfeSSional
AMIN -ll.1 +16.9 +0.1 +15.7
ASIA +10.4 +8.1 +14.l +6.3
BLCK -6.0 -7.8 -8.1 ~10.9
MEX +5.3 -12.9 +0.1 +24.0
PRIC +25.9 +19.7 -O.5 -36.0
HISP -11.4 -16.8 —16.1 -17.8
WHIT -12.1* -9.3* —4.9 +4.3
OTHR -l3.1 +1.5 -13.3 -41.3*

 

 

*These groups showed statistically significant linear
trends in means at p < .0125.

84

Within parent education groups, there were three
distinct patterns:

(a) Asian-Americans consistently showed a net mean SAT-Total
score increase;
(b) African-Americans consistently showed a net mean SAI-

Total score decline; and
(c) Hispanic-Americans (neither Mexican-American nor Puerto

Rican) consistently showed a net mean score decline

between 1988 and 1991.

The largest net mean score decline was detected among
”Other” unspecified ethnic examinees from
”graduate/professional” parent education backgrounds. The
largest net mean score increase was detected among Puerto
Rican examinees from ”high school or less” parent education
backgrounds.

For the White examinee group, net mean SAT-Total score
change between 1988 and 1991 was consistent with the overall
pattern found among parent education groups: The higher the
level of parent education, the smaller the net mean score
decline. Despite the increases in ethnic minority
compositions and their differential patterns of net mean SAI-
Total score change, the net mean score pattern of the
predominantly White examinee population dominated these other
patterns of net mean score change.

Several findings, different from the outcomes of White
parent education groups' net mean SAT-Total score changes,

are notable. For the African-American group, the net mean

 

85

.SAIrTotal score decline became more pronounced with
increasing levels of parent education. A similar pattern was
found among the Puerto Rican group, although the effect was
more dramatrc: The two lowest parent education groups --
”high school or less,” and ”2-year/associate” -- revealed net
mean score increases of nearly 20 scale-score points, while
the highest parent education group, ”graduate/professional,”
showed a net mean score decline of more than 30 scale—score
points. These patterns were opposite in direction from the
patterns observed for White examinees across increasing
levels of parent education.

African—American, Puerto Rican, Hispanic—American, and
”Other” unspecified ethnic groups all showed the largest net
mean SAT-Total score declines between 1988 and 1991 at the
highest parent education group, ”graduate/professional.”
American-Indian, Asian-American, Mexican-American, and White
examinee groups with ”graduate/professional” parent education
backgrounds all showed net mean score increases between 1988
and 1991.

Wren}; Table 27 shows

the partitioned sources of variance by regression analyses.

‘ .mo‘gfilm ..I
‘

A

86

Table 27

J . E . E :1 1 1 . E

 

 

 

Source df Sum of Squares Mean Square F Value

Model 31 4198648822 135440284 4573
YEAR 1 3076395 3076395 104
RANK 1 3406307984 3406307984 115013
Ethnicity 7 730452988 104350427 3523
RANK/Ethnic 7 57376417 8196631 277
YEAR/RANK 1 108658 108658 4
YEAR/Ethnic 7 907944 129706 4
YEAR/RANK/ 7 325059 46437 2
Ethnic

Error 291666 8638179374 29617

Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 32.71%

Results from the analysis of variance by regression
showed that none of the interaction effects with time were
significant. The interaction effect of class rank-by—
ethnicity group was, however, statistically significant.

Table B—2 in Appendix B lists the raw mean SAIrTotal
scores, standard deviations, sample sizes, and computed 2-
tests for presence of linear trend in mean SAT-Total scores
among each of the ethnic groups, matched on class rank.
Statistically significant negative linear trends were
detected among White examinees in the lowest 60% and the
second quintile class ranks, and among African-Americans in
the highest 10% class rank. Statistically significant
positive linear trends were detected among Asian-Americans in

the second 10% and the highest 10% class ranks. Even after

87

matching CH1 class ranks, significant differences between
ethnic groups' mean SAT-Total scores remained.

Table 28 shows the net mean SAT-Total score change
between 1988 and 1991 for ethnic groups matched on high
school class rank. Negative quantities represent a net mean
score decline from 1988 to 1991; positive quantities

represent a net mean score increase during the same period.

 

 

Table 28
\‘ v‘qo ; - o - o ‘ orqoo- ,o. goo .. {-0.
We
Lowest Second Second Highest
60% Quintile 10% 10%
AMIN -4.5 +9.9 +25.4 -22.3
ASIA +4.6 +13.2 +18.8* +14.9*
BLCK -7.1 —8.1 -3.8 -26.4*
MEX +6.6 —8.6 +7.4 -25.0
PRIC +9.6 —1.7 +9.3 +47.9
HISP -7.8 +3.2 -6.3 -22.7
WHIT -9.1* -4.5* -1.8 -4.5
OTHR +4.8 +18.9 -25.2 -58.8

 

 

*These groups showed statistically significant linear
trends in means at p < .0125.

Among African—American and among White groups, there was
a consistently negative net mean SAT-Total score change
between 1988 and 1991 across all four class rank categories.
For the White group, the most pronounced negative net score

decline was found among examinees in the lowest 60% class

:3

m

88

rank. For African-American examinees, the most pronounced
net mean score decline was found among the highest 10% class
rank, with more than a 26 scale-score point drop over the
four-year period.

The Asian-American group showed consistently positive
net mean SAT-Total score changes across all four class rank
categories between 1988 and 1991. The most pronounced net
mean score increase was found in the second 10% class rank
group, revealed by a net mean increase of 18 scale-score
points. The net mean score increase was slightly lower, at
15 scale-score points, for the highest 10% class rank of the
Asian—American group.

Although no other systematic patterns were detected in
net mean SAI-Total score changes between different ethnic
groups matched by class rank, the range of net mean SAT-Total
score change differed in significant ways between ethnic
groups and across class rank categories. For example, the
range of net mean SAleTotal score change within the lowest
60% class rank for 1988 to 1991 was 18.7 scale-score points
(maximum net decline of 9.6 points, maximum net increase of
9.1 points). Table 29 shows the range of net mean SAT-Total

scores between ethnic groups within the four class ranks.

FL'““1“""§)

89

Table 29

Bangs of Me: Mean sem_mgta] Score Change :9: 1288-122]
WW

 

 

 

Lowest Second Second Highest
60% Quintile 10% 10%
Max net decline -9.1 -8.6 -25.2 -58.8
Max net increase +9.6 +18.9 +25.4 +47.9
Range 18.7 27.5 50.6 106.7

 

 

EarenL_education_and_glass_rank. Table 30 shows the

partitioned sources of variance by regression analyses.

 

 

 

Table 30

so. ' o , ',, - to 'q ‘0 o . it! . :29. 0.0

&_Year

Source df Sum of Squares Mean Square F Value

Model 7 4316566430 616652347 21111
YEAR 1 3076395 3076395 105
PEd 1 1369414643 1369414643 46882
RANK 1 2910757873 2910757873 9965
PEd/RANK 1 32204307 32204307 1102
YEAR/PEd 1 1020844 1020844 35
YEAR/RANK 1 25210 25210 1
YEAR/PEd/ 1 67157 67157 2
RANK

Error 291690 8520261766 29210

Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 33.63%

90

Results from the analysis of variance indicated the only
significant interaction effect with year was the ”parent
education-by—year” effect. A significant interaction effect
of parent education—by-class rank was detected, suggesting
time invariant but stable differences between ”parent
education-by-class rank” groups.

Table B—3 in Appendix B lists the raw mean SAT-Total
scores, standard deviations, sample sizes, and computed 2—
tests for presence of significant linear trends in mean SAI-
Total scores among each of the ”parent education-by-class
rank” groups, by separate year.

Significant negative linear trends were detected among
the lowest 60% class rank and among the highest 10% class
rank with ”high school or less” parent education, all four
class rank groups with ”2—year/associate” parent education
backgrounds, and the lowest 60% and the second quintile class
rank categories with ”college” parent education backgrounds.

A significant positive linear trend was detected for the
second 10% class rank group with ”graduate/professional”
parent education background.

Table 31 shows the net mean SAIL-Total score change
between 1988 and 1991 for parent education and class rank
categories. Negative quantities represent net mean score
declines, and positive quantities represent net mean score

increases.

T4. 7. -.‘n..'.'3‘7';“?1

91

Table 31

W
W

 

 

High 2-year/ College Grmhmte/
School Associate Professional
30 -13.1* -8.7* —9.7* -1.3
70 -4.6 —8.5* -9.8* —3.8
85 —8.1 -10.5* -4.8 +8.1*
95 —15.5* —l3.1* -6.2 +0.3

 

 

*These groups showed statistically significant linear
trends in means at p < .0125.

For both class rank categories of lowest 60% and second
quintile, the net mean .SAI:Tptal score change was
consistently negative among all parent education groups. For
class rank groups of second 10% and highest 10%, the highest
parent education group of ”graduate/professional” level
showed the only non-negative net mean score change between
1988 and 1991. For the class rank group of highest 10%,
there was an ordered progression in net mean score change
with increasing levels of parent education: The most
pronounced net score decline within this class rank category

was among ”high school” or less parent education groups.

Table 32 shows the partitioned sources of variance by

regression analyses.

4"
an!

T.- A in“ I
l A u
n

92

 

 

 

Table 32

$0. ' o - '-o ‘ o '- ‘0 o q '09 , {-0. 90'

W

Source df Sum of Squares Mean Square F Value

Model 63
YEAR 1 3076395 3076395 126
PEd 1 1367546742 1367546742 56273
RANK 1 2905802372 2905802372 119570
Ethnicity 7 729599875 104228554 4289
YEAR/PEd 1 1019321 1019321 42
YEAR/RANK 1 23777 23777 1
YEAR/Ethnic 7 1025486 146498 6
PEd/RANK 1 33438655 33438655 1376
RANK/Ethnic 7 42821799 6117400 252
PEd/Ethnic 7 9352378 1336054 55
YEAR/PEd/ 1 29684 29684 1
RANK
YEAR/RANK/ 7 314459 44923 2
Ethnic
YEAR/PEd/ 7 279684 39955 2
Ethnic
PEd/RANK/ 7 654763575 93537653 3849
Ethnic
YEAR/PEd/ 7 521144 74449 3
RANK/Ethnic

Error 291634 7087212844 24302

Total 291697 12836828197

Percent variance SAT-Total scores explained by the
model, R2 = 44.79%

The partitioned sources of variance in the pooled four-
year sample explained by main effects and interaction effects
showed a non—significant ”parent education-by—rank—by-
ethnicity—by-year” interaction effect, but a significant
”parent education-by-rank-by—ethnicity' interaction effect.

Table B-4 in Appendix B lists raw mean SAT-Total scores,

standard deviations, sample sizes, and computed z—tests for

*1 .1 Aer-Ff“!!!
|
A.

93

presence of linear trends in mean SAT-Total scores for each
”parent education-by-class rank-by-ethnicity” group,
separately by annual cohort.

Statistically’ significant linear trends. for (ethnic
groups matched on parent education and on class rank
categories are identified in Table 33. Empty cells indicate
no statistically significant linear trends were detected for

any ethnic group.

Table 33

1.; “15"E' . 1’ 5
WW

 

 

 

 

 

High School 2-year/ College Graduate/
or less Associate Professional

Lowest White (-) White (-)
60%
Second White (-) ASIA (+)
Quintile
Second White (-) ASIA (+)
10%
Highest White (-) White (-) ”Other” (-)
10% PRIC (+)

 

 

 

 

 

Statistically significant negative linear trends in mean
scores among White examinees from lower parent education
groups and across all class rank categories were detected.
Puerto Rican examinees in the highest 10% of their class from
lOW' parent education backgrounds showed. a significant
positive linear trend in mean scores between 1988 and 1991.

Asian—American examinees with class ranks in the 60th to 90th

‘1
...In

.K in. i-V‘L‘ 4'
l o

94

percentiles from high parent education backgrounds showed
significant positive linear trends in mean SAT-Total scores
between 1988 and 1991.

Table 34 identifies ethnic groups with the highest mean
SAT-Total scores between 1988 and 1991, matched by parent
education and class rank groups.

The White examinee group obtained the highest mean SAI-
Total scores between 1988 and 1991 when matched on parent
education and on class rank groups. For the upper extremes
of class rank and high parent education distributions,
”Other” unspecified ethnic groups and Asian-Americans
obtained the highest mean SAT-Total scores between 1988 and
1991. The summary in Table 34 shows the different patterns
of highest mean SAT-Total scores by ethnic groups matched on
parent education and class rank.

Table 35 lists ethnic groups with the lowest mean SAI-
Total scores between 1988 and 1991, matched on parent
education and on class rank. Within parent education and
class rank groups, the African—American group obtained the
lowest mean SAT-Total scores. For higher parent education
groups, Puerto Ricans obtained the lowest mean score for some
years of the study. No other distinct patterns of lowest

mean score attainment were apparent.

if“ ..Atnfi.“38 .‘fl

95

 

.mdsoum xcmn mmmfio Ucm coflumospm bemoan m.HHmo wnu co cocoons .wuoom
Hmboeanw some ummzmfi: wnb Um: QSObm oflcnuw cm :ofin3 CH umm> mnu ucwmwbdwu mononucwnMQ 2H muwnesz H

 

 

 

 

 

 

 

 

 

 

 

 

:3 cm?
Aomv moans
Aamlmmv Gomm< Ammv numnuo. Aamv Gmfim< Aomv Gmfim<
Awmv nhwnuou Ammv cmfiUGH GMOmHmE< Aomnmmv muflSS Aad .mm .mwv mufi£3 mba umwnmmm
Aam .omv Gmfim<
Ammv wuޣ3
Ammv uuwnuoa Aamummv muw£3_ Aamnwmv mufinz Aamummv mum£3 woa Ucoowm
Aam .omv cmflmd mamucflno
Amm .mmv mufi£3 Aamlmmv muwn3 Aamnmwv mumgz Aflmnmwv wuflnz Ucoumm
Aam .mmv .umnuo. Ammv .nmnuo.
Aom .mmv muflnz Aam .om .mmv wuwnz Aamnwmv muwnz Aamummv mums3 wow umoSOA
Hmcofimmmmonm mumfloOmw< mmma Ho
\mumnpmuo mmmaaoo \Hmmhum Hoonom nmfim
H o o .9.“ r . .0. u o .c s. 4. a o a ..c) «...p o .< o o ..-

wm magma

96

.mdsonm xcmu mmmHo pcm cofiumozpw bamboo m.HHwo wan co pwcoume .wpoom
HmuoelHdw cmwe umwon wnu can agonm oflczow cm noflnz CH umm> ocu ucwmwudwu mononucwHMQ CH muwbesz m

 

 

 

 

 

 

 

 

 

 

 

Ammv coofim ouumsm Aamummv xomHm woa

Aam .om .mmv xomHm” Aamummv :moflm chumsm Aamummv xomHm Ammv cmofim Ouumzm ummnmwm

Aomv xomHm Aamv xomHm Aomv cmowm condom «OH

13.3.2: 582 3.85 Smummv 582 335m :mummv xva 3m .mm .mm: xomHm ccouwm

Ammv xomam Aomv cmoflm ouumsm maflucfioo

lam.om.mwv cmoflm Ouuwom AHm .mm .mmv xomam Aamnmmv xomHm Aamummv xomHm pgoomm

Aomv cmomm Ouumsm Aamv cmofim ouumom wow

Aam .mm .mmv xomam Aomummv xomHm Aamnmmv xomHm AHm:mmv xomam ummzoq

Hmcoflmmwuoum wumfloomm< mmwa no
\mumncmuo mmmaaou \ummmum Hoosom zmwm
m 0 0 .0.“ v . .0. . .0 0.. u. c 0 u 0.-) «.0 . .. 0 .00

mm wanna

97

Table 36 shows the net mean SAT-Total score change
between 1988 and 1991 for ethnic groups matched on parent
educational attainment and class rank. Negative quantities
represent a net mean score decline; positive quantities

reflect a net mean score increase.

Table 36

 

 

 

   

 

 

 

 

   

Ifighsdmbl Lowest Second Second Highest
Parent Ed 60% Quintile 10% 10%

AMIN —34.7 +28.0 +63.8 -148.3
ASIA +2.8 +12.8 +21.l +22.6
BLCK +5.9 —8.4 -13.6 -20.8
MEX +ll.7 -9.0 +24.9 -24.9
PRIC +18.4 +35.1 -33.9 +140.7*
HISP -26.1 -2.2 +13.0 +28.4
WHIT . -1.5 -5.3
OTHR +29.5 -32.8 -28.9

2-Yr/Associate Lowest Second Second Highest

Parent Ed 60% Quintile 10% 10%

AMIN +39.6 -34.6 +17.0 -5.1
ASIA +5.5 +3.1 +5.9 +25.6
BLCK +0.4 —13.6 +6.0 —37.2
MEX +1.3 -9.7 ~22.3 ~45 8
PRIC +27.8 -28.8 +87.5 —6.0
HISP -1.6 —l9.2 -31.6 S
WHIT "" i -2.4 -8.9*
OTHR -18.7 +53.7 +3o.1 +21.7

 

 

98

Table 36 (Cont’d)

 

 

   

 

 

 

 

   

College Parent Lowest Second Second Highest
Ed 60% Quintile 10% 10%

AMIN -11.4 +73.8 -69.1 -41.9
ASIA +15.9 —l.O +13.9 +19.9
BLCK —22.9 -9.1 —1.5 +17.2
MEX -27.1 -26.7 +26.0 +48.7
PRIC -31.1 -28.1 +55.9 +57.3
HISP .4 +1.7 —59.4 -35.o
WHIT fl. :5 —8.4* -3.6
OTHR w -23 4 .. +0.1 -19.2

Graduate/ Lowest Second Second Highest

Fjgfiifggl 60% Quintile 10% 10%
AMIN —26.7 -2.9 +80.7 +0.1
ASIA -3.o +26.1* +36.4* +7.5
BLCK .6 +2.3 +10.0 -28.6
MEX .6 +37.1 -5.9 -1.8
PRIC .9 -41 3 -15.4 +5.7
HISP .8 -4 o +6.2 .8
WHIT - 4 . 5 +6 . 2 I;
OTHR fl +24.9 Si +16.6 -45.4 + -75.6*

 

 

*These groups showed statistically significant linear
trends in means at p < .0125.

White examinees from ”college” level or less parent
education backgrounds consistently showed net mean score
declines between 1988 and 1991, even after matching White
parent education groups on the class rank variable. Aside
from an overall net mean score decline, no additional

patterns were detected for White examinees among these parent

99

education groups. The most pronounced net mean score change
for White examinees was a 15 scale—score point decline among
the lowest 60% class rank matched on lowest parent education
background.

Asian—American examinees from ”2—year/associate” or less
parent education backgrounds consistently showed net mean
score increases between 1988 and 1991 across all four class
rank categories. The higher the class rank, the larger the
net mean score increase for Asian—American examinees in the
lowest parent education group.

African-American examinees from ”high school” or less
parent education backgrounds showed larger net mean score
declines across increasing class rank categories. A net mean
score increase was detected between 1988 and 1991 among
examinees in this group matched on lowest 60% class rank.

African-American examinees with ”college” parent
education backgrounds showed diminishing net mean score
declines across increasing class rank categories. African-
American examinees in the highest class rank category from
the ”college” parent education group showed a net mean score
increase of 17 scale-score points. This pattern of net mean
score change relative to increasing class rank was opposite
in direction from the pattern of observed net mean score
decline among the lowest parent education groups.

The net mean score change between 1988 and 1991 among
Hispanic-Americans (neither‘ Mexican-American. nor Puerto

Rican) in the lowest parent education group, and for

.?f**’*1

100

Mexican-Americans and Puerto Ricans in the ”college” parent
education groups showed patterns of net mean score increases
across increasing class rank categories.

Mexican-Americans with ”2—year/associate” parent
education backgrounds, and ”Other” unspecified ethnic
minority groups with the highest parent educational levels,
both showed patterns of progressively larger net mean score
declines across increasing class ranks.

For American-Indian and Asian-American examinees from
”graduate/professional” parent education groups matched on
class rank, the higher the class rank, the smaller the net
mean score decline (or the larger the net mean score
increase). This pattern was interrupted by smaller net mean
score changes between the second 10% and the highest 10%
class rank groups. This finding was consistent with a
ceiling effect, in which the highest ”ability” examinees were
already obtaining high scores.

The pattern of net mean SAIrTotal score change among
White examinees from the lowest and the highest class rank
categories was consistent across increasing levels of parent
education. The higher the level of parent education, the
smaller the net mean SAT-Total score decline. For White
examinees in the lowest 60% or the highest 10% of their class
from ”graduate/professional” parent education backgrounds, a
net mean score increase in mean scores was found between 1988
and 1991. This pattern, highlighted by the shaded cells in

Table 36, ‘was consistent with. the overall pattern of

.‘t‘

101

diminishing net mean score decline across increasing levels

of parent education.

Mean_Score_Degline_Exnlained_b1_Comnositional_Change

Demographic variables, as defined in this study,
included ethnicity, parent’s highest educational attainment,
and gender.

To assess how much of the nine-point decline in
unadjusted mean SAT-Total scores was explained by change in
demographic composition, mean SAT-Total scores, adjusted at
the 1988 demographic composition, were predicted. Separate
year regression estimates were computed for each annual

cohort. The basic regression equation is:

Y1 = Bo + B1(Parent Ed): + Bz(EthniC)i +

(Equation 1)
BKRSEXM + EL
Mean values of the predictors -- parent education,
ethnic group indicators, and gender -- were computed for the

1988 cohort. These mean values were held constant in each of
the separate year regressions, while the regression
coefficients were allowed to vary by cohort year. The 1988
predictor means and separate year regression estimates for
annual cohorts are presented in Table C—1 of Appendix C.
Statistically significant trends were detected in the
year-by-year regression coefficients for Puerto Rican and for
Asian-American ethnic groups, while the regression

coefficients among other ethnic groups remained relatively

 

102

stable over time. The between-year differences in the
regression coefficients for Puerto Rican and for Asian—
American groups implied the presence of a time-by-ethnic
group interaction. Subsequently, adjusted means were
computed separately for Puerto Rican, and for Asian—American
groups. The regression estimates for these two ethnic
subgroups appear in Table C-2 of Appendix C.

The predicted mean SAT—Total scores, adjusted at the

1988 demographic composition, are shown in Table 37.

Table 37

WW

 

 

Net Mean

GROUP 1988 1989 1990 1991 (flange,
Adjusted

Overall Group 923.16 925.28 923.58 919.51 -3.65
Puerto Rican 803.84 808.67 805.34 814.78 +10.94

Asian-American 943.98 947.84 954.61 953.73 +9.75

 

For the overall group, changes in demographic
characteristics accounted for more than half of the
unadjusted score decline between 1988 and 1991.
Statistically significant differences in separate ‘year
regression slopes for the Puerto Rican and for the Asian-
American ethnic subgroups indicated that the trend. in
adjusted means for these two ethnic groups were opposite in
direction from the detected trend in adjusted means for the

overall group.

103

Two additional prediction equations were analyzed as
tests of sensitivity to the prediction model using only
demographic predictors. Adjusted mean SAT-Total scores for
the overall group, for Puerto Rican, and for Asian-American
ethnic subgroups were computed for both tests of sensitivity.
Both tests included main effect class rank as a factor in
addition to demographic predictors. The second test of
sensitivity also included a two-way interaction effect,
”parent education-by-class rank.” Appendix D presents the
regression estimates and predicted adjusted mean SAT-Total
scores for both tests of sensitivity.

The trend in adjusted mean SAT-Total scores for Asian-
American and for Puerto Rican ethnic subgroups was opposite
in direction to the trend in adjusted means for the overall
group. Adding main effect class rank to the prediction model
with demographic predictors enhanced the positive trend in
adjusted mean scores for the Asian-American and the Puerto
Rican subgroups. For the Asian—American subgroup, the
positive trend in mean SAT-Total scores was enhanced by
approximately three scale-score points; for the Puerto Rican
subgroup, the positive trend in mean SAIrTotal scores was
enhanced by nearly seven scale-score points.

Inclusbmn of the two-way interaction effect, ”parent
education—by-class rank,” provided no additional information
over class rank and demographic main effects for the Asian—
American ethnic subgroup. For the Puerto Rican ethnic

subgroup, adding the two-way interaction effect to main

 

104

effects class rank and demographic predictors enhanced the
positive trend in adjusted mean SAIL—Total scores by an
additional three scale-score points.

For the overall group, both tests of sensitivity yielded
virtually identical results to the model which only used
demographic predictors. The inclusion of main effects class
rank or the interaction effect, ”parent education—by-class
rank,” contributed no additional information in the
explanation of the mean score decline between 1988 and 1991
for the overall group. The net mean change in adjusted
scores, by' prediction. model and ethnic subgroups, are

summarized in Table 38.

Table 38

1 i' i 1 1 . 1
Models

 

 

Model Overall Puerto Rican. Asian-American
Demographic only -—3.65 +10.94 +9.75
Demographic + -2.78 -E17.59 ‘+12.94
Rank
Demographic + -3.19 +21.11 +12.90

Rank + PEd/Rank

 

For the overall group, compositional changes in parent
education, ethnicity, and gender explained more than half of
the nine-point decline in observed mean SAT-total scores
between 1988 and 1991. For Asian-American and for Puerto
Rican ethnic subgroups, the trend in adjusted means was

upward, approximately ten scale-score points.

105

Class rank was a significant factor in the prediction of
adjusted means both for Asian-Americans and for Puerto Rican
ethnic subgroups. Compositionally, the Puerto Rican examinee
subpopulation became increasingly disadvantaged on class rank
and parent education variables. The Asian—American examinee
subpopulation became increasingly disadvantaged on the parent
education variable as the proportions of respondents from the
lowest parent education groups increased over time.

The trend in adjusted mean SAT-Total scores suggested
that more than half of the nine-point score decline was
attributed to compositional change between 1988 and 1991, but
the remainder of the net score decline in unadjusted means
remained unexplained by either class rank or ”parent
education-by—class rank.”

For Asian-American and for Puerto Rican ethnic
subgroups, the trend in adjusted mean scores was opposite in
direction from the trend for the overall group. Although the
Asian—American group became increasingly disadvantaged on the
parent education variable and the Puerto Rican group became
increasingly disadvantaged on parent education and class rank
variables, both of these ethnic subgroups obtained higher
mean SAT-Total scores than might have been expected.
Statistically controlling for class rank enhanced the
positive trend in adjusted mean scores for these two ethnic

subgroups.

106

Synopsis

The analyses of proportions of examinee characteristics
and the analyses of mean SAT-Total scores for the pooled
sample by specific subgroups revealed some significant
sources of the observed score decline between 1988 and 1991.
The constituency of ,SAIxtakers changed over time,
particularly in terms of ethnicity and parent education
compositions. Different patterns of net mean SAT-Total score
change between 1988 and 1991 were detected among specific
subgroups of the pooled sample.

For non Asian-American/non—Puerto Rican ethnic groups,
the trend in adjusted mean SAT—Total scores revealed that
nearly half of the mean score decline between 1988 and 1991
was attributable to compositional change in demographic
characteristics. The remainder of the mean score decline
remains unexplained.

For Puerto Rican and Asian-American ethnic subgroups,
positive trends in adjusted mean scores suggested that, as a
group, their performance on the SAT was better than expected
after controlling for change in demographic composition. An
interpretation of these findings and their implications are

presented in the next chapter.

 

CHAPTER 5

INTERPRETATION AND DISCUSSION

1. . .

No significant trends over time were detected in the
proportions of female examinees who took the SAT. The gender
composition of college-bound seniors remained unchanged
between 1988 and 1991.

The composition of ethnicity and parent’s education
groups in the annual cohorts revealed significant differences
over time. The proportions of African—American, Asian-
American, Mexican-American, Puerto Rican, and other Hispanic-
American respondents increased between 1988 and 1991, with

the most pronounced increase in proportions of Asian-American

respondents . Proportions of ”Other” unspecified ethnic
groups remained unchanged, The proportions of White
examinees decreased over time. The examinee population

became increasingly disadvantaged on ethnicity as increasing
proportions of traditionally low—scoring ethnic groups, and
decreasing proportions of White examinees, who traditionally
obtain higher SAT scores, took the SAT between 1988 and 1991.
Significant changes were detected among the two lowest
and two highest parent education group categories. The
proportions of respondents from low parent education groups
—- less than a high school diploma -- increased, while the

proportions of respondents from higher parent education

107

108

groups -- some graduate school or a graduate/professional
degree —- decreased between 1988 and 1991.

Overall, the population of examinees became more
disadvantaged on ethnicity and parent education variables.
Males retained their mean SAT-Total score advantage over
females, and female examinees consistently outnumbered male
examinees.

If highest level of parent’s educational attainment is
an accurate reflection of SES, and if performance on a
scholastic aptitude measure like the £511 is positively
correlated with SES, then changes in the composition of
respondents from different parent education groups is
expected to manifest change in mean SAT-Total scores between
1988 and 1991: More examinees from low SES backgrounds, who
traditionally obtain lower mean SAT-Total scores, took the
SAT between 1988 and 1991. Concurrently, fewer examinees
from high SES backgrounds, who traditionally obtain higher
mean SAT-Total scores, took the SAT between 1988 and 1991.
Trends in mean SAT—Total scores, adjusted for ethnicity,
gender, and parent education groups, were consistent with
this hypothesis. Approximately half of the score decline in
unadjusted means was explained by compositional change in
demographic characteristics in the overall group.

The trend in adjusted means for Puerto Rican and for
Asian-American ethnic subgroups differed from the trend for
the overall group. The adjusted mean SAT-Total scores for

these two subgroups revealed a positive trend in adjusted

.. ‘hf-WA P~lfi- .'

109

means, opposite in direction from the trend in adjusted means
for the overall group. Although the Puerto Rican mean SAI—
Total score was lower than the mean for the overall group,
their predicted mean scores, adjusted for parent education
and gender, showed a net increase of nearly eleven scale-
score points, a trend opposite in direction from other non
Asian-American ethnic groups. A similar trend was detected

in Asian-Americans' adjusted mean SAT-Total scores.

3 . .

Information about the course-taking patterns of college-
bound seniors in the sample showed some initially anomalous
patterns. Despite the decline in mean SAT-Total scores,
self-reported course-taking patterns of respondents indicated
that increasing proportions of college-bound seniors studied
some or all of the elements of a typical college-bound
curriculum, defined by minimal number of years of study in
certain academic subjects. The proportions of respondents
who studied all of the elements of a college-bound curriculum
increased between 1988 and 1991. These results were opposite
in direction from the expected findings if fewer years of
study in college-bound curricula was a tenable explanation of
the decline in mean SAT—Total scores between 1988 and 1991.
The increase in proportions of respondents who reported fewer
years of English study was consistent with the expectation of

increasing' numbers of less ‘well-prepared college-bound

 

110

SAT-takers. The increase in proportions of respondents who
were less-well prepared in English was contrary to the
finding that increasing proportions of respondents reported
studying a college-bound curriculum, which includes four or
more years of English study. This finding, together with the
increases in ethnic minority proportions between 1988 and
1991, suggests one possible explanation for the increase in
fewer years of English study: An increasing proportion of
linguistic minorities and immigrants to the U.S. took the SAT
between 1988 and 1991, and their high school academic
preparation included all elements of a college-bound high
school curriculum except English.

The proportions and distributions of high school class
rank categories were stable between 1988 and 1991. Except
for a quadratic trend in the proportions of respondents who
reported a class rank in the third quintile, no statistically
significant trends in the proportions of high school class
rank categories were detected between 1988 and 1991. Thus,
class rank alone was not an explicit source of the change in
mean SAT scores. Despite compositional changes in ethnicity,
parent's educational attainment, and academic course-taking
patterns, the distribution of high school class rank remained

unchanged between 1988 and 1991.

i . . .
Academic aspirations of college-bound seniors in the

sample changed significantly in highest degree goals and in

111

areas of intended college major between 1988 and 1991.
Specifically, the proportions of respondents whose highest
degree goal was less than college remained constant or
declined between 1988 and 1991, while the proportions of
respondents whose highest degree goal was a
”graduate/professional” degree increased.

The proportions of intended Biology majors and Social
Science majors increased significantly between 1988 and 1991.
Significant declines in the proportions of planned Business
majors and planned majors in ”other” unspecified areas were
detected over the same period. The decrease in proportions
of intended majors in other unspecified areas (which included
technical, trade, skilled labor, and general or
interdisciplinary studies), suggested that college-bound
seniors in the sample had better defined their intended
college majors by the time they took the SAT. The increase
in proportions of intended Biology majors and Social Science
majors, together with the increase in proportions of advanced
graduate degree goals, suggested that larger numbers of
college-bound examinees aspired to careers in the medical/
health care, legal, and academic professions, which

traditionally require high levels of educational attainment.

W
A net mean SAT-score decline of nine scale-score points

and an increasing variance in SAT-Total scores were detected

between 1988 and 1991. The descriptive statistics of the

112

SAT—Total score distributions showed that scores became
increasingly positively skewed between 1988 and 1991, and
that the shape of the score distributions became less flat
over time.

Although the means and the medians of the SAT-Total
score distribution declined between 1988 and 1991, the modes
of the SAT-Total score distribution increased by 40 scale-
score points. As the mean, median, and mode of the score
distribution approached a common score, these shifts in
measures of central tendency reflected a tendency toward the
normal distribution. These observations, together with the
results from compositional analyses, suggested an increasing
self-selection ratio on the SAT over time: Increasing
numbers and a growing proportion of high school seniors
aspired to higher education and took the SAT to meet those

goals.

_ J i 1 i 'E' 5]

Gender. Mean SAT-Total scores for both males and
females declined between 1988 and 1991. As a group, males
consistently scored 50 scale-score points higher than females
between 1988 and 1991 on the SAT. This difference was
statistically and substantively significant, but the effect
was constant and time invariant. The composition of males
and females in each annual cohort remained constant. Gender,
as a factor, was not a significant contributor to the

observed mean score decline between 1988 and 1991.

113

W. Mean SAT-Total scores between ethnic
groups differed by as much as 200 scale-score points, a
significant difference of one full standard deviation. As a
group, and on the basis of ethnicity alone, African-Americans
consistently obtained the lowest mean SAT—Total score while
Asian-Americans consistently obtained the highest mean score
between 1988 and 1991. The differences in mean scores
between ethnic groups were stable and time invariant, but the
proportions of ethnic groups changed. The changes in ethnic
composition and the differences between ethnic group means
between 1988 and 1991 implied that, in general, the examinee
population became increasingly disadvantaged over time.

Barent_edngatign. Significant differences in mean SAT-
Total scores between different parent education groups, and
their interaction effects over time were detected. As
expected, parent educational attainment appeared to correlate
significantly with mean score: The higher the parent
education background, the higher the mean score.
Furthermore, mean scores of examinees from the lowest parent
education groups declined over time, while mean scores of
examinees from the highest parent education groups increased
over time. The compositional change in parent educational
attainment groups, between-group differences in mean SAT-
Total scores by parent education categories, and
predominantly declining trends in the means among lower

parent education groups, all suggested that the examinee

114

population became increasingly disadvantaged on the parent
education variable between 1988 and 1991.
fiigh_§chgol_gla§a_rank. The SAT is a test of general
readiness for college-level work, and reflects some measure
of cumulative learning and experience. It is expected, then,

that of the four SDQ variables considered singly in the

analyses of means —- gender, ethnicity, parent education, and
high school class rank -- class rank accounts for the largest
proportion of explained variance in SATrTotal scores. The

difference in mean SAT-Total scores between ”highest 10%” and
”lowest 60%” class rank groups was more than 300 scale-score
points, a significant difference of one and one-half standard
deviations. The higher the class rank, the higher the mean
SAT-Total score. Across all four class rank groups,
statistically significant negative linear trends in the mean
scores were detected from 1988 to 1991. (mm the basis of
class rank alone, mean SAT-Total scores uniformly declined
over time, across all class ranks. A statistically
significant interaction effect with year was therefore not
detected.

2arent_educatign_and_ethnigity, Significant differences
were detected between ”parent education-by-ethnicity” groups,
although the interaction effect with year was non—
significant. The only significant year interaction effect
was detected for ”parent education-by-year.” Mean SAT-Total
score differences between ethnic groups remained, and

different patterns of net mean score change were found for

115

ethnic groups after matching on parent education background.
The Asian-American net mean score change between 1988 and
1991 was consistently positive while the net mean score
change for the .African-American. group» was consistently
negative across all parent education groups. For African—
American and Puerto—Rican groups, the higher the parent
education background, the greater the net mean score decline
between 1988 and 1991. For White examinees, the effect was
opposite in direction: The higher the parent education
background, the smaller the net mean score decline.

Clasa_;ank_and_ethnigity. Significant differences were
found between ”class rank—by-ethnicity” groups, but these
differences were not susceptible to interaction effects with
year. Net mean SAT-Total score change between ethnic groups
showed different patterns of change after matching on class
rank. Net mean score changes for Asian-Americans were
consistently positive between 1988 and 1991 across all four
class rank categories, but these patterns were attenuated by
patterns of net mean score decline across all four class
ranks among White examinees, who comprised the majority of
the sample, as well as for the largest ethnic minority group
of African-Americans.

W. Of the three models
that included two SEQ variables to seek explanations for the
mean score decline, parent education, class rank, parent

education-by-class rank, and their interaction effects with

39:25

116

year accounted for more than 33% of the variance in SAT—Total
scores in the pooled four—year sample.

Time invariant but significant differences in mean
scores between ”parent education-by-class rank” groups were
detected. For the ”highest 10%” class rank, the higher the
level of parent education, the smaller the net mean score
decline. The only net mean score increase was detected among
the highest 20% class ranks from ”graduate/professional”
parent education groups. No distinct patterns of net mean
score change among parent education groups matched on class
ranks were detected.

W A
significant three-way interaction effect between parent
education, ethnicity, and class rank revealed differences
between ethnic groups after matching for parent education and
class rank. variables. Within. ethnic groups, distinct
patterns emerged across parent education categories after
matching (n1 class rank, and across class rank categories
after matching on parent education.

Mean score differences between ethnic groups remained
after matching for class rank and parent education, and
several distinct patterns of net mean score change were found
for all ethnic groups included in the study.

The analyses of mean SAT-Total scores grouped by parent
education/class rank/ethnicity categories suggested. the

following general themes.

 

 

117

WWW” Within
ethnic groups matched on parent education categories, the
higher the class rank, the smaller the net mean score
decline. This pattern pertained tn) African—Americans,
Mexican-Americans, and Puerto Ricans from ”college” parent
education groups; and Hispanic—Americans with ”high school”
or less parent education backgrounds.

This pattern also pertained to American-Indians and
Asian-Americans from ”graduate/professional” parent
education backgrounds, with what appeared to be a ”ceiling
effect” among the highest class rank category. These
findings were consistent with the hypothesis of a changing
self-selection ratio on the SAT among lower class rank and
lower parent education groups.

WM As a group. the number and
proportions of Asian-Americans from lower parent education
backgrounds steadily increased and, across all class ranks,
showed consistent net mean score increases between 1988 and
1991. Although no distinction was made between Asian
nationals, Asian immigrants, and subsequent generations of
Asian-Americans who took the SAT in this study, this
pattern among ”Asian-Americans” from lower parent education
groups was consistent with the hypothesis that larger
numbers and growing proportions of Asian immigrants took
the SAT between 1988 and 1991.

W The group of White

examinees from college level or less parent education

*Lxgyfic 1. E711
_' :-

118

backgrounds showed negative net mean score changes across
all class rank categories. Among White examinees in the
”highest 10%” and ”lowest 60%” class ranks, the higher the
level of parent education, the smaller the net mean score
decline: Net mean score increases were found among the
highest parent education categories for these examinees.
Although significant changes in the ethnic composition
of SAT—takers were detected between 1988 and 1991, White
examinees still comprised the vast majority of each annual
cohort. Net mean score declines which were detected across
all class rank categories of the White examinee group
dominated much of the observed mean score decline in SAT—
Total scores between 1988 and 1991. Distinct patterns of
net mean score increase were found, but these patterns were
attenuated by the patterns of net mean score decline found
among the ethnic majority group.
fl2EmELJLL__QQle§x_fflQLQLS. Within ethnic groups matched on
parent education, the higher the class rank, the larger the
net mean score decline. This pattern contrasts the pattern
described in Theme 1, and describes the pattern of mean
SAT—Total scores between 1988 and 1991 for African-
Americans with ”high school” or less parent education,
Mexican-Americans with ”2-year/associate” parent education,
and ”Other” unspecified ethnic groups from
”graduate/professional” parent education backgrounds.
There are several inconclusive interpretations of this

pattern. Some self-reported class ranks may have been

119

inflated in order to enhance respondents' chances for
opportunities in higher education. Larger net mean score
declines among higher class rank. categories may be
indicative of within-ethnic group change, explained by a
growing proportion of foreign nationals, recent immigrants,
and linguistic minorities with lower competence in English
language skills.

MW. Mean SAT. scores --
adjusted for ethnicity, parent education, and gender
compositions -- were predicted for 1988, 1989, 1990, and 1991
cohorts with the 1988 cohort as the referent group. Although
changes in demographic characteristics did not account
completely for the nine-point decline in unadjusted mean SAT-
Total scores between 1988 and 1991, compositional change in
parent education, ethnicity, and gender did account for
nearly one-half of the observed nine-point decline. The
implication is that the observed nine—point decline was less
severe than initially appeared —- more than half of the score
decline reflected compositional change in the examinee
population. However, the remainder of the score decline is
left unexplained by the present study.

For Asian-American and Puerto Rican examinee subgroups,
performance on the SAT actually improved over time despite an
increased disadvantage on the parent education variable. For
these two ethnic subgroups, compositional change in class
rank was a significant factor in the prediction of adjusted

mean SAT-Total scores. Based on these findings, there

120

appeared to be significant within-ethnic group differences
which contributed to the patterns of net mean score change

for Puerto Rican and for Asian-American ethnic subgroups.

The generalizability of the findings from this study are
limited to college—bound seniors who took the SAT, and to
college-bound seniors in the annual cohorts of this study.
College-bound seniors who take the SAT comprise a self-
selected sample, and these results suggest that the extent of
self—selection has changed from the early 19605 as increasing
proportions of examinees from certain demographic groups who
might not have aspired to higher education in previous years
are now taking the SAT as a first step in the college
application process.

The selection of cases from the data sets for analyses
further restricted within-sample variability. The mean SAT-
Total scores for selected cases, for example, were
consistently higher than the mean SAT-Total scores of
excluded cases for each year in the study. Table 39 shows
the mean SAT-Total scores for the excluded and for the

selected cases.

121

Table 39
WW

 

 

Year Mean SAT Mean SAT Difference Percent Cases
of Excluded of Selected in Means Included in
Cases Cases Analyses
1988 857 923 66 66.68
1989 856 923 67 66.66
1990 853 920 67 66.65
1991 848 914 66 66.71

 

The mean SAT-Total scores for the group of excluded
cases were consistently lower than the mean for the group of
selected cases. Respondents on the SEQ may differ from non-
respondents who take the SAT in ways other than in obtained
mean SAT—Total scores, but such differences could not be
ascertained without a common measure as a basis of
comparison. SAT-Total scores were the only common measure
between selected and excluded cases for the present study.
Selection of cases may threaten the generalizability of
results. However, the patterns of:

(a) a consistent difference in mean SAT-Total scores between
excluded and selected groups from year to year;
(b) identical trend of nine-point decline in unadjusted mean

SAT-Total scores for excluded and selected cases; and
(c) a consistent percentage of cases selected for analyses

from. each. annual cohort using the same inclusion

criteria from year to year,

reduce the threat of generalizability of results. The

122

patterns and conclusions of this study are relevant to the

population of college-bound seniors who took the SAT between

1988 and 1991.

Conclusions and Recommendations

The purposes of this study were to examine changes in
the CEEB's SAT outcomes between 1988 and 1991, and to seek
explanations for the observed score decline. Compositional
change in the examinee population was hypothesized as one
explanation for the mean SAT-Total score decline between 1988
and 1991.

Results from compositional analyses and analyses of
means revealed significant differences and substantive
changes in the composition of the examinee population and
their obtained mean SAT-Total scores. Results from neither
analyses alone were sufficient to formulate the conclusions
of this study. The findings from both analyses together
provided a clearer, but still incomplete, perspective of the
changes among college-bound high school seniors who took the
SAT.

Gender and class rank variables alone provided weak
explanations of the observed score decline in the present
study.

Compositional change in ethnicity and parent education
groups, and differences between ethnic and parent education
groups were significant contributors to the explanation of

observed mean score decline between 1988 and 1991. The

 

§~

 

123

increasing proportions of examinees from ethnic minority
groups and low parent education groups, and declining trends
in their obtained mean SAT-Total scores affected the observed
trend in mean scores in the overall sample.

In general, the SAT examinee population became
increasingly disadvantaged on ethnicity and parent education
variables. Examinees from demographic groups traditionally
associated with low mean SAT-Total scores became more
numerous over time.

The results of this study are consistent with the
hypothesis that compositional change among examinees was a
significant source of explanation in the observed mean score
decline between 1988 and 1991. Substantive differences were
detected between ethnic groups -- in the proportions of
ethnic group constituencies, and in the mean SAT—Total scores
they obtained —- even after matching demographic groups on
parent education and class rank variables. Differences
within ethnic groups require further investigation.

Whether the self—selection ratio on the SAT among
traditionally low—scoring demographic groups increased, or
whether the proportions of immigrants to the U. S. and
linguistic minorities increased, the constituency of the
examinee population changed in significant ways.

Changes in characteristics of the examinee population
and the effects of these changes have several implications
for the developers of large-scale standardized testing

programs. If the changes in ethnic composition and the

.‘P arms-334!

 

124

differences between ethnic groups, for example, also pertain
to examinee populations at earlier stages of education, these
changes and. differences :necessitate the (development of
relevant and appropriate norms for standardized testing
programs. Item and test development may similarly be
affected by the need for studies of cultural bias, content
validity, and differential item functioning more relevant to
a changing examinee population.

Interpretations of test results and other outcomes must
be within the appropriate context. Test results and other
outcomes from large-scale testing programs that presuppose an
American educational experience may lose their relevance and
substantive meaning when the test is administered to a group
of individuals who receive their education outside of the
American educational system.

As shown in this study, the interpretation of trends in
observed test performance can be tricky business. Trends in
unadjusted mean test scores, for example, could reflect
compositional changes, rather than. a ”real” change in
underlying aptitude or ability. The social science
laboratory is rife with many complex factors, and raw
observations are often masked by subtle interactions which
are, at best, difficult to decipher and variables that are
often impossible to isolate. Statistically controlling for
some of these complex factors provides social scientists with

a viable alternative to investigate some of these effects.

 

 

125

The purpose of this study was to seek explanations for
the observed score decline between 1988 and 1991. How much
of the net decline could be attributed to compositional
factors was assessed through the analysis of trends in
adjusted means. According to the methods used in this study,
half of the score decline in unadjusted means was explained
by demographic change in the examinee population. Half of
the unadjusted mean score decline remained unexplained by
compositional change as defined in this study.

Two possible explanations exist for the unexplained
portion of the score decline: (1) additional variables that
could more completely describe compositional change are
needed; or (2) changes in other constructs independent of
composition, like ability or aptitude, account for the
remainder of the score decline. If increases in the
population of immigrants and linguistic minorities are viable
considerations in the explanation of the score decline, then
these factors must be measured. Citizenship, residency
status, first language learned, and predominant language used
are examples of variables which could capture this type of
information.

Improved measures of family structure, parent
educational attainment, or other variables which better
capture a respondent’s SES, merit attention. In the present
study, SES is indexed only by one parent's highest
educational attainment, whether that parent is ani active

agent in the respondent's intellectual and personal growth

 

126

and development, or whether that parent predominantly resides
in the examinee's household. Parent's highest educational
attainment in the present study presupposes that it is an
accurate reflection of the respondent’s SES. Change in
family' structure is EH1 important consideration :hi the
determination of SES. The assessment of a respondent's SES
is complex, and is likely influenced by the relationships
between many variables.

Additional years of data, preceding 1988 and following
1991, should be included to test whether patterns of mean
score change and compositional changes in the examinee
population maintain over longer periods of time.

Corroborative evidence may be found by comparing the outcomes

of the present study -- compositionally and by test
performance -- with outcomes from other large-scale testing
programs . The impact of the immigrant and linguistic

minority populations on observed outcomes of testing
programs, and changes in the self-selection ratio on tests
like the SAT can provide substantively important implications

for both developers and users of tests and their outcomes.

- .1!

V5;
.
I
.

 

APPENDIX A
SUMMARIES OF 1980-1987 CEEB OUTCOMES

Table A-1

APPENDIX A
SUMMARIES OF 1980-1987 CEEB OUTCOMES

 

 

Year Males Females Total
1980 48.3 51.7

(478,443) (513,071) 991,514
1981 48.1 51.9

(478,625) (515,708) 994,333
1982 48.2 51.8

(476,556) (512,124) 988,680
1983 48.3 51.7

(465,092) (498,117) 963,209
1984 48.2 51.8

(464,910) (499,829) 964,739
1985 48.3 51.7

(471,995) (505,370) 977,365
1986 48.1 51.9

(481,477) (519,271) 1,000,748
1987 48.2 51.8

(520,326) (560,100) 1,080,426

 

127

128

 

pwflufiucopm Doc mascum omcnum emcuo Ham mopsHoch nonuo

cmofluwsdncmflm<
cmxmmH< mmCDHUCmV cmflch-cmoHHwE<

“Amuflcmdmmm cmon obnmnm::o: uwcuo w

“wumn3
.mocmoenu

"cmomm Ouumzm
.mcmufiuoe<ucmofixwz mopsfiocfiv
“oewnom mcHBOHHOH on» mBOHHOH m manna ca poms mofiuomoumo mazmuwnEmE ozone oficnum one

“AmuwpcmHmH oawfiomm a
Umcmammm

.Ao>onm

.mamucwmuo mopsao:mv
“cmoHHmE<ucmoHuw< “50>fiumz

.uonuo “oufinz “annex Ouwmsm “Haucmmwo

 

 

 

“cmoflnwecucmofixwz “xumHm “cmecanmoHuw54 "pom: wuw3 mwfluommumo QfinmnmnEmE anoum oecnum mcfl3oHHOu mnu .wmma Cu anu@ v
.cofiumuumficHEpm Edm wmmfi ocu HON wHQMAHm>m mmz coflumEMOMGH omnamnmoemp oz m
Amoa.mav Ammm.amv Ammm.omv Amme.mav 5mmm.mav xvmm.omv Amom.omv
mao.o «No.0 mmo.o mmo.o Nmo.o mmo.o mmo.o .Hmnuo.
imam.mmmv 5mm>.mamv Ammm.mamv Aoao.oamv x¢m¢.mmmv Imam.bvmv 5mmm.mvmv
mmm.o oom.o mom.o Haw.o bam.o mam.o Hmm.o mumnz
xvom.oav Ammo.HHv Amom.aav AmHm.oav Ammv.oav AMmm.oav 5mmm.mv
oao.o mHo.o mao.o mHo.o «Ho.o HHo.o HHo.o amoflm ouwmbm
Imam.mmv Ammm.mvv xomm.mmv Aamm.mmv xvmo.emv Ammm.amv Ammm.mmv
mmo.o mvo.o mvo.o mvo.o mmo.o vmo.o mmo.o cmomwwemncmflm<
Amow.mmv Ammm.mav Amma.wav xvmm.mav Ammm.mav 5mmm.mav Amme.mav
mmo.o mmo.o omo.o mHo.o wao.o mao.o bao.o cmomuoE¢uomcmomfim
Ammo.mmv Ammm.mmv Ammo.owv Amma.mmv 5mmm.mmv Amma.mmv AHNm.mwv
mmo.o mwo.o Hmo.o mmo.o mmo.o omo.o Hmo.o :mofiwmemucmomwmm
Aboa.oav imam.vv 5mmm.ev xmom.vv xnmm.vv Ammo.mv xmvm.vv
oao.o moo.o moo.o moo.o moo.o moo.o moo.o cmflchncmoflme<
bmma mwma vwma mwma Nmma Hmma omma asouw

 

 

mud magma

129

1') rim-ado! H Kim

wEOUCa ucwpma amzccm pmumzflpmc: was EOuu pwmcmno wwa can man so mEOoCa ucmuma amsccm no mmauommumo wnu

.m>onm Umumaa moauomwumo

.emma ca mcacaammm m

 

 

 

 

Avmv.maav Ammm.mmaa Amom.vmav onm.maav Avmm.mmv Avam.mmv

amm.o mma.o mma.o Nva.o maa.o amo.o woos Ho oooomm

Amma.aoav norm.ooav Amva.omv AmNm.aav Ammv.vav Amoo.amv

vaa.o ama.o aaa.o ooa.o mmo.o wao.o mmmme n oooovm

Amma.omav Aaov.mvav Amao.avav Ammv.mvav Aava.mmav Ammm.maav

aaa.o maa.o maa.o vaa.o «ma.o ava.o mmmmm n oooomm

xavv.vmv aama.mmv Aamm.moav Amma.maav Amam.omav Amam.maav

maa.o mma.o maa.o ava.o ava.o ava.o mmmmm I ooovmm

Amaa.mmv Amam.voav Ammm.aaav ammm.omav Ammm.mvav Amam.mmaa

ama.o ama.o mea.o aaa.o aaa.o ama.o mmmmm u oooaam

Ammm.mav xaao.mav aama.amv Amov.aoav Awmo.vmav Aamv.mmaa

Noa.o moa.o baa.o ama.o Nma.o mma.o mmmma u ooomam

Aaam.mav Amam.aav 5amm.amv Ammm.aav Amom.amv Aaom.moav

mmo.o aao.o amo.o ooa.o aaa.o mma.o mmmaa : oooam

Ammo.mma Aoaa.amv Arum.amv Ammo.omv Aoam.mmv Avmo.mma

amo.o mmo.o mmo.o mmo.o vvo.o mvo.o ooomm Hops:
mama vama mama mama aama oama wEooGH ucmumm

00 0 «:0 0 .c o 0 c 0c-0c -00 000 o I ..

mud manna

130

 

 

 

 

 

Aaoo.av Amvm.amv aama.vmmv Amom.mamv Avom.aomv Amom.amav
ao.o v0.0 vm.o mN.o NN.o aN.o mama
Amam.av Aeam.amv xaov.mmmv Avom.avmv Amom.maav 5mmw.maaa
ao.o v0.0 mm.o am.o am.o am.o aama
Ammm.mv 50am.amv Amaa.mvmv Ammm.ammv Aomm.oaav Amoa.maav
ao.o v0.0 am.o am.o am.o aN.o mama
Aaaa.mv Aomm.omv Amma.vmmv Aamm.ammv Ammm.aaav Ammm.vaaa
ao.o v0.0 mm.o am.o am.o am.o aama
Ammm.vv Avvm.mmv ooam.mmmv Amam.vmmv Amvm.vaav 5mmm.maav
ao.o mo.o am.o am.o mm.o mm.o mama
Abov.vv Aamm.mmv Aomb.ammv xamo.bmmv xmvm.amav Aaav.maav
ao.o mo.o am.o mm.o mm.o mm.o mama
Amvv.vv Aomm.bmv aama.mmmv Amvm.ammv Amam.vmav Aomm.omav
ao.o mo.o am.o mm.o mm.o am.o aama
Avvv.vv Amaa.amv xaom.ommv Aaom.ammv Aamm.mmaa xaam.amav
ao.o mo.o am.o mm.o mm.o mm.o oama
:uuam auaam zuuam nuuam enema shame Ham»
numwzoa unuusom upwane ncaooom nocoowm ummnmam
00 0 0 .06" 6 c 0c—0u 00 000 u

vu< magma

131

 

 

Table A-5
’30 - ’37 '7000 00 -0|-0 ‘ 0 , .0-u ' -0. .
3+ Years
Year 4+ Years 3+ Years 2+ Years Social
English Math Language Science
1980 0.80 0.29 0.36 0.40
(745,318) (264,985) (329,062) (370,206)
1981 0.82 0.27 0.35 0.40
(758,580) (255,239) (326,938) (372,726)
1982 0.82 0.26 0.35 0.41
(757,536) (240,880) (323,424) (371,794)
1983 0.83 0.25 0.35 0.41
(736,880) (222,160) (310,943) (363,958)
1984 0.83 0.24 0.35 0.41
(752,318) (218,743) (309,734) (371,852)
1985 0.84 0.24 0.35 0.42
(766,346) (218,318) (318,671) (376,974)
1987 0.88 0.29 0.36 0.44
(868,907) (287,702) (358,510) (433,248)

 

132

a woman Umpsaocav
“mmaUSDm vacnum

“20aum>uomcoo\>uumoH0u
w mndomoaaza “oamse

msowcmaawomaz
“COaumuspw meDaUCav

“mwocoaom amoamanan

“mommamCMa cmaquH “wusumuwuaa\nmaamcm

1A>00aozo>ma

 

 

.Ammwum unapwowua may Ca meSaUCa Doc mpsum uo mmauommumo “Uwpaompc: “amc0a0m00>

“oocmaom >umDaaaE
mmwum pwumawu a mwocoaom awaoom
meum>m\mocoaom woundeoo UmUDaUCaV mmmum pmumawu a mwocoaom amoam>nm
“manuasoaumm cmoDaUCav mmwum Conwaou w mwocmaom amoaUOaOam

“wocmaom monunaa
“ammocmaom amoam>na
“macaumoaGSEEou a

“moasocoow ween
“moaDMEwnums

. OUHOEEOU . mmwcfimfim

Amwusuaso w mucuman “magnumowm
umcaummcamco “mammamcm
“AaMUame a nuamwn
“amuum Hmumwnu “GOamaamu

“bum “cmammp amucmscoua>co\wu:uomuanoum owUSaUCav mmauacmesm w muu<

"momma wmwaaoo Umpcmuca HON dd” map so pom: mumB mmauomwumo amumcmm meaBOaaow mnu

.mmma 6» Hoaum m

.cwuuau3mu mm: acaumwso manu ow aaaQaMDuwQ Eoua

 

 

 

 

Adflwv wuaMCCOameso w>audauomwo uchSDw mnu cwnB .COaumuumaCaEUm 94m aama och no“ mammaam>m ma ceaumEHOuCa canmmumoswp oz a
Ammm.aav Ammo.>mav Amam.maav Amom.mmmv aama.mmav Avvm.aaav mama
omo.o oom.o oma.o oam.o oaa.o oma.o
5amm.mmv “mam.amav Ammm.omav 5amm.aamv Amao.maav Avmm.omv mama
mao.o oaa.o aam.o avm.o aaa.o aaa.o
Amaa.mmv Ammm.omav Aoam.aamv amam.aomv Amam.mmav Amma.mav aama
aao.o aaa.o avm.o mmm.o mma.o amo.o
Ammo.amv Aamm.mvav Aoaa.ommv Ammm.amav xowm.aaav Awam.aav mama
mmo.o aaa.o mmm.o amm.o mma.o moa.o
Ammm.nmv xovm.amav Aaaa.aomv Amma.oomv amma.vmav Ammm.ama mama
wac.o mma.o mmm.o vmm.o mma.o oaa.o
Ammm.mav Aava.maav Aomm.maav Aaao.momv Aamo.maav Ammo.moav aama
mao.o aaa.o mom.o mmm.o mom.o maa.o
Amma.aav Aaom.v>av Avam.aaav Ammo.momv Aamm.maav Aomv.aaav oama
aao.o mma.o aaa.o mmm.o bom.o mma.o
wmnuo mocmaom mocmaom mmmCamsm mocwaom mmauacmEdm Haw»
awaoom amoammnm amoaUOaOam w muH¢
a. .a .6) :0: 0-00: 0 «. 0¢a0c 00. 000 o I 0n.

and OHQMB

133

 

 

 

 

Ammo.aamv xmv0.aaav Ammo.ammv xaaa.mmmv Aomm.vvv
mm.o ma.o am.o om.o v0.0 mama
Amom.amav Aomm.maav Roma.avmv Aomo.mmma Ammm.ovv
ma.o ma.o mm.o aa.o mo.o mama
Amam.mmav Amam.eaav Aamm.ammv Aaam.mmmv Amma.wvv
aa.o aa.o am.o aa.o mo.o aama
xamm.amav Aavm.mmav Rema.mmmv Aoao.mamv Ammm.mev
aa.o aa.o am.o aa.o mo.o mama
Amva.wmav Aaaa.amav Avaa.mmmv Ammm.vmmv xaam.avv
ma.o aa.o am.o mm.o mo.o mama
Aamm.maav Aaom.oaav 5amm.ammv Amom.mmmv xamm.omv
om.o ma.o mm.o mm.o mo.o aama
Ammm.aaav Avaa.amav Aomo.mmmv aama.ammv Amam.mmv
om.o aa.o mm.o mm.o mo.o oama
pwcaompco amGOammmmoum muwummz mHOamnomm onwaoommd ummw
\uoDooo \Hmmmum
.0 0..- -.0 0 . 01.0- 00 000 .

mt< manna

134

Table A-8

155: _ 1551 1 E J] _ j .

 

 

 

Year Total

1980 991,514
1981 994,333
1982 988,680
1983 963,209
1984 964,739
1985 977,365
1986 1,000,748
1987 1,080,426
1988 1,134,364
1989 1,088,223
1990 1,025,523
1991 1,032,685

 

 

 

APPENDIX B
MEAN SAT-TOTAL OUTCOMES FOR 1988-1991

 

Table B-1

WW

MEAN SAT-TOTAL OUTCOMES FOR 1988-1991

APPENDIX B

 

 

3 . 1 . E
PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
High School or Less
775.77 787.50 775.87 764.71 -0.773
AMIN (171.91) (158.40) (177.19) (164.64)
194 356 179 140
838.48 848.86 850.19 848.93 1.152
ASIA (203.92) (205.18) (209.54) (211.70)
881 1,117 1,240 1,312
693.81 701.91 697.28 687.82 -1.390
BLCK (159.99) (157.58) (157.25) (152.78)
1,690 1,959 1,913 2,051
764.08 778.17 769.38 769.39 0.265
MEX (171.52) (168.10) (177.18) (165.83)
640 841 892 1,022
713.47 734.73 718.06 739.38 1.340
PRIC (156.33) (172.24) (184.65) (173.61)
239 279 314 288
760.16 763.79 746.83 748.75 -1.407
HISP (174.44) (178.21) (176.78) (175.97)
380 494 556 577
859.96 862.88 853.50 847.87 —5.460*
WHIT (181.35) (181.76) (183.72) (181.24)
9,279 10,188 9,535 9,468
779.22 804.31 776.59 766.11 —1.151
OTHR (192.50) (204.35) (197.72) (182.09)
180 232 223 252

 

135

 

 

136
Table B-1 (Cont'd)
PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
2-year/Associate
823.51 822.17 822.16 840.37 0.928
AMIN (183.55) (161.50) (193.88) (177.04)
242 378 255 188
891.66 895.01 898.94 899.80 1.293
ASIA (202.59) (200.12) (209.06) (205.78)
788 1,003 1,067 1,080
744.02 746.66 736.72 736.24 -2.060
BLCK (172.25) (170.62) (166.65) (166.41)
1,942 2,389 2,351 2,506
843.31 834.63 835.98 830.43 -1.072
MEX (185.56) (168.26) (189.54) (176.93)
456 574 604 668
793.27 810.10 807.90 813.01 0.946
PRIC (186.75) (190.69) (173.66) (184.44)
156 205 186 239
847.76 848.74 847.49 831.00 —1.175
HISP (186.40) (192.12) (187.10) (187.30)
294 406 422 460
906.63 904.35 901.18 897.30 —4.619*
WHIT (185.01) (187.64) (185.39) (184.97)
14,812 16,396 15,436 15,147
841.04 820.14 841.52 842.50 0.413
OTHR (208.66) (196.09) (202.46) (207.18)
192 215 256 256

 

 

 

137

 

 

Table B-1 (Cont’d)
PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)11n
ETHNIC ITY
College
878.20 851.89 890.88 878.27 0.558
AMIN (189.25) (162.44) (189.24) (180.33)
167 264 148 110
951.95 959.95 970.15 966.02 1.872
ASIA (210.00) (207.74) (208.50) (219.90)
987 1,240 1,390 1,378
788.55 785.54 793.59 780.48 —0.641
BLCK (171.38) (184.05) (179.21) (177.90)
874 1,024 998 1,066
891.24 877.65 904.65 891.33 0.451
MEX (179.28) (191.25) (186.58) (196.44)
153 204 228 241
814.23 796.18 794.35 813.78 -0.041
PRIC (187.25) (190.61) (168.38) (203.79)
111 123 124 148
891.34 894.10 915.39 875.24 -0.433
HISP (211.95) (185.17) (184.50) (183.49)
172 197 217 250
961.42 963.64 961.89 956.52 -2.134
WHIT (189.36) (189.86) (192.61) (190.13)
12,021 13,349 12,221 12,114
898.42 917.88 919.31 885.10 —0.560
OTHR (203.56) (224.55) (218.43) (213.74)
177 160 173 202

 

 

 

138
Table B-1 (Cont’d)
PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Graduate/Professional
912.61 871.40 896.94 928.32 1.009
AMIN (192.13) (165.80) (184.61) (202.61)
161 301 160 131
1,051.62 1,048.39 1,059.76 1,057.93 1.190
ASIA (230.97) (231.66) (231.18) (231.51)
1,389 1,823 1,779 1,988
842.03 833.90 841.42 831.08 -0.987
BLCK (197.88) (201.71) (195.68) (194.04)
1,136 1,222 1,180 1,198
896.90 927.14 923.55 920.85 1.171
MEX (197.53) (201.29) (194.93) (194.70)
197 259 256 260
862.48 842.46 860.65 826.48 —1.170
PRIC (198.54) (198.21) (226.11) (220.32)
137 175 153 159
936.11 927.49 916.79 918.28 —1.375
HISP (207.70) (209.63) (208.82) (214.84)
339 423 443 499
1,014.00 1,019.22 1,020.76 1,018.25 2.123
WHIT (198.02) (199.66) (201.25) (201.02)
17,678 19,270 17,595 17,350
1,038.29 1,016.59 1,011.67 997.04 —2.319*
OTHR (227.23) (237.02) (237.22) (218.39)
280 302 342 402

 

*t—ratio for linear contrast,

a-level of 0.0125.

statistically significant at an

 

 

Table B-2

139

Euna_nEa3n_jU1I;IIua1l_1jEui_Lr2LLl_iEannlsLJSizeriit_933uas_lbznhtii

 

 

Eudniis_12royuxa
CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Lowest 60%
758.13 763.23 764.86 753.61 —0.278
AMIN (158.23) (137.18) (162.57) (150.52)
310 536 296 205
776.16 783.55 779.70 780.76 0.389
ASIA (176.71) (176.23) (173.46) (177.55)
790 1,013 1,063 1,191
684.88 683.04 679.65 677.81 -1.906
BLCK (147.54) (145.65) (143.97) (140.57)
2,253 2,601 2,567 2,804
721.32 731.72 715.68 727.97 0.135
MEX (150.73) (142.59) (143.92) (152.41)
438 530 606 700
712.31 704.33 698.97 721.87 0.532
PRIC (155.11) (154.78) (147.16) (158.07)
212 261 273 310
739.50 742.75 737.28 731.70 -0.872
HISP (153.13) (155.55) (151.31) (156.10)
337 488 508 583
809.09 806.41 803.23 799.97 -5.353*
WHIT (156.64) (158.69) (159.07) (156.80)
14,954 16,020 14,988 15,243
772.64 774.25 774.11 777.40 0.288
OTHR (180.30) (193.95) (175.32) (185.99)
231 259 304 350

 

:m' V
l

r“

1

 

 

 

140
Table B-2 (Cont'd)
CLASS RANK/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
Second Quintile
823.81 822.10 826.60 833.71 0.645
AMIN (159.99) (141.32) (182.09) (167.11)
197 390 209 175
867.98 874.62 880.64 881.21 1.886
ASIA (186.19) (187.07) (190.24) (193.14)
1,035 1,257 1,410 1,443
753.09 748.27 748.91 744.96 -1.376
BLCK (167.43) (162.68) (162.82) (162.38)
1,679 1,891 1,866 1,985
793.97 791.57 807.51 785.35 +0.318
MEX (161.80) (161.18) (164.54) (157.08)
453 586 586 609
770.72 781.57 776.18 768.98 -0.195
PRIC (179.45) (171.73) (174.81) (173.77)
194 216 220 226
822.08 817.34 812.73 825.24 0.123
HISP (180.48) (174.50) (176.46) (190.45)
351 429 477 540
907.76 908.47 904.11 903.27 —3.031*
WHIT (162.53) (161.85) (163.94) (162.81)
15,312 16,938 15,415 15,119
854.63 851.63 843.33 873.54 0.899
OTHR (189.55) (198.40) (193.41) (200.03)
227 245 240 311

 

 

 

 

 

141
Table B-2 (Cont’d)
CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Second 10%
893.21 907.47 894.73 918.62 0.974
AMIN (168.79) (162.82) (165.66) (167.42)
156 249 148 109
958.44 964.82 975.88 977.20 2.608*
ASIA (198.79) (194.75) (202.29) (199.98)
982 1,338 1,387 1,459
‘g
806.61 811.83 805.80 802.81 -0.728 J
BLCK (180.39) (181.26) (174.68) (183.12)
1,045 1,292 1,199 1,244 5
|
E
863.96 867.04 870.12 871.35 0.666 ‘\
MEX (168.19) (169.15) (181.57) (164.68)
318 412 433 495
819.20 813.63 811.38 828.52 0.380
PRIC (189.52) (190.25) (198.47) (185.64)
137 182 152 176
874.31 922.96 897.91 867.99 —0.914
HISP (188.85) (194.58) (202.55) (191.07)
269 365 383 359
994.43 996.71 993.09 992.68 —1.293
WHIT (165.06) (166.93) (168.45) (168.69)
11,723 13,070 12,070 11,825
948.89 933.20 924.84 923.67 —1.281
OTHR (215.29) (214.97) (230.04) (219.18)
207 206 225 237

 

 

 

 

142
Table B-2 (Cont’d)
CLASS RANK/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
Highest 10%
1,056.73 989.11 1,037.19 1,034.38 -0.221
AMIN (177.42) (175.22) (190.65) (186.42)
101 124 89 80
1,121.87 1,119.59 1,128.20 1,136.76 2.273*
ASIA (196.34) (201.16) (199.09) (201.35)
1,238 1,575 1,616 1,665
921.47 915.79 916.70 895.09 —2.452*
BLCK (192.97) (194.03) (189.53) (192.50)
665 810 810 788
996.88 982.29 986.62 971.83 -1.493
MEX (182.49) (174.58) (197.48) (184.05)
237 350 355 387
900.50 940.57 916.74 948.44 1.347
PRIC (194.88) (190.20) (230.11) (227.68)
100 123 132 122
1,034.21 1,010.21 1,005.74 1,011.51 —1.355
HISP (196.10) (202.83) (201.18) (186.46)
228 238 270 304
1,121.50 1,122.54 1,120.84 1,117.05 -2.173
WHIT (166.75) (169.35) (171.12) (171.27)
11,801 13,175 12,314 11,892
1,113.29 1,082.31 1,101.42 1,054.53 -2.186
OTHR (221.29) (226.53) (213.01) (219.20)
164 199 225 214

 

*t-ratio for linear contrast,

a-level of 0.0125.

statistically significant at an

 

 

Table B-3

W
W

143

 

 

PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)11n
CLASS RANK
High School or less
729.64 729.09 720.08 716.57 —5.097*
Lowest 60% (152.66) (149.89) (147.09) (146.24)
4,580 5,197 5,219 5,460
807.69 811.21 802.22 803.07 —2.013
2nd (162.21) (158.67) (165.01) (164.88)
Quintile 3,982 4,577 4,238 4,257
875.11 882.82 877.50 866.99 —2.052
Second 10% (174.51) (174.59) (178.88) (176.94)
2,818 3,211 3,033 3,065
1,001.37 1,000.08 993.19 985.85 -3.087*
Highest 10% (181.46) (184.69) (188.71) (183.45)
2,103 2,481 2,362 2,328
2—yr/Associate
768.96 764.44 763.08 760.31 -3.228*
Lowest 60% (151.11) (152.39) (151.80) (150.38)
5,932 6,788 6,591 6,777
860.57 858.35 857.07 852.06 -2.827*
2nd (162.14) (161.99) (162.16) (162.68)
Quintile 5,706 6,416 6,057 5,953
940.36 937.36 929.77 929.90 -3.306*
Second 10% (167.47) (169.89) (174.62) (168.50)
3,855 4,575 4,302 4,255
1,061.61 1,056.63 1,052.99 1,048.49 -3.263*
Highest 10% (171.47) (174.85) (173.36) (177.00)
3,389 3,787 3,627 3,559

 

«—

 

 

 

144
Table B—3 (Cont’d)
PARENT
EDUCATION/ 1988 1989 1990 1991 (TS)11n
CLASS RANK
College
815.68 812.55 813.14 805.97 —2.619*
Lowest 60% (156.06) (157.57) (161.22) (158.16)
4,106 4,537 3,997 4,219
908.19 909.91 902.46 898.39 —3.356*
2nd (161.29) (162.99) (163.69) (161.58)
Quintile 4,199 4,680 4,429 4,424 ran
993.99 993.17 991.77 989.22 -1.208 ‘
Second 10% (166.49) (168.15) (165.71) (169.66) '
3,169 3,703 3,546 3,452 1»
1,116.42 1,115.27 1,113.14 1,110.20 —1.599
Highest 10% (164.08) (165.12) (172.00) (171.18)
3,188 3,641 3,527 3,414
Graduate/Professional
842.18 842.27 841.06 840.86 -0.482
Lowest 60% (166.62) (170.51) (170.38) (169.21)
4,907 5,186 4,798 4,930
948.81 945.41 946.61 945.03 -0.995
2nd (169.62) (170.50) (171.64) (174.42)
Quintile 5,561 6,279 5,699 5,774
1,032.76 1,035.79 1,038.43 1,040.89 2.515*
Second 10% (169.05) (172.76) (173.33) (173.61)
4,995 5,625 5,116 5,132
1,165.89 1,165.17 1,169.10 1,166.14 0.481
Highest 10% (165.64) (171.35) (169.68) (172.31)
5,854 6,685 6,295 6,151

 

*t-ratio for linear contrast,

a-level of 0.0125.

statistically significant at an

 

Table B-4

145

Parent Education Group = High School or Less

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
Lowest 60%
728.24 738.44 702.16 693.50 —1.738
AMIN (165.90) (138.08) (127.07) (152.09)
91 154 74 60
711.37 716.52 712.22 714.21 0.094
ASIA (159.61) (155.88) (148.30) (163.16)
205 253 293 335
629.81 645.89 637.92 635.75 0.493
BLCK (126.72) (130.45) (127.29) (121.66)
691 806 840 892
678.97 696.65 684.14 690.66 0.557
MEX (146.25) (129.28) (133.63) (138.07)
195 239 285 331
670.56 665.86 661.28 688.91 0.828
PRIC (135.69) (152.59) (145.41) (139.74)
90 99 133 119
701.59 708.49 699.76 675.46 -1.547
HISP (161.59) (155.95) (154.74) (135.07)
107 179 210 227
759.10 755.10 748.73 744.15 ~4.537*
WHIT (146.51) (146.22) (143.21) (142.29)
3,142 3,393 3,315 3,406
704.07 705.54 720.00 733.56 1.222
OTHR (153.43) (173.10) (148.86) (166.52)
59 74 69 90

 

Table B-4

(Cont'd)

146

Parent Education Group = High School or Less

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Second Quintile
755.00 779.25 794.48 782.98 1.047
AMIN (148.08) (140.07) (186.62) (142.69)
46 107 58 47
798.99 810.40 795.39 811.75 0.487
ASIA (177.50) (175.79) (179.12) (189.23)
247 298 343 343
693.21 695.99 694.11 684.78 —0.960
BLCK (146.56) (146.17) (145.58) (138.73)
489 529 525 534
747.50 743.96 751.37 738.46 -0.474
MEX (146.73) (143.12) (150.35) (142.00)
204 283 262 293
714.32 737.74 745.36 749.47 1.415
PRIC (146.69) (156.70) (178.09) (167.66)
81 84 84 76
744.85 725.55 725.99 742.65 —0.105
HISP (154.03) (148.21) (153.52) (164.46)
130 146 162 162
840.72 845.31 836.66 839.20 -1.013
WHIT (152.44) (146.97) (154.48) (153.73)
2,739 3,071 2,737 2,733
766.30 770.85 738.06 733.48 —1.331
OTHR (161.62) (173.42) (178.97) (166.80)
46 59 67 69

 

2-

 

Table B—4

(Cont'd)

147

Parent Education Group = High School or Less

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)1in
ETHNICITY
Second 10%
808.46 858.78 826.36 872.22 1.311
AMIN (150.32) (159.59) (150.95) (159.16)
39 74 33 27
855.50 861.28 875.37 876.58 1.588
ASIA (176.36) (182.91) (186.38) (190.90)
229 297 315 354
744.29 744.92 764.93 730.72 -0.534
BLCK (161.86) (158.26) (158.51) (163.35)
317 388 337 389
799.54 823.64 810.36 824.45 1.222
MEX (147.71) (161.95) (165.82) (153.79)
151 176 192 227
772.14 760.76 711.23 738.20 —1.324
PRIC (185.27) (174.36) (175.91) (178.70)
42 66 57 61
777.39 840.75 793.74 790.36 0.105
HISP (171.14) (186.29) (195.34) (172.60)
92 106 123 112
915.51 925.06 918.87 910.25 -1.330
WHIT (161.29) (158.35) (164.12) (159.28)
1,900 2,051 1,923 1,848
802.71 845.09 753.21 773.83 -1.539
OTHR (183.51) (180.82) (183.86) (175.42)
48 53 53 47

 

 

Table B-4

(Cont'd)

148

Parent Education Group = High School or Less

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Highest 10%
998.33 938.10 969.29 850.00 -1.818
AMIN (114.39) (202.60) (226.32) (162.60)
18 21 14 6
998.05 1,002.23 1,027.65 1,020.68 1.670
ASIA (196.67) (198.26) (188.10) (186.14)
200 269 289 280
841.55 835.76 833.41 820.76 —1.249
BLCK (169.86) (165.75) (165.01) (169.40)
193 236 211 236
926.56 926.15 907.58 901.70 -1.290
MEX (182.72) (167.39) (204.37) (163.44)
90 143 153 171
764.62 896.33 859.25 905.31 2.613*
PRIC (163.79) (150.02) (238.82) (190.91)
26 30 40 32
890.98 880.00 869.67 919.34 0.708
HISP (186.60) (190.90) (189.89) (179.62)
51 63 61 76
1,036.21 1,037.53 1,025.06 1,024.56 —2.503*
WHIT (162.02) (167.60) (170.79) (165.10)
1,498 1,673 1,560 1,481
923.70 959.13 1,003.82 870.87 -0.652
OTHR (246.41) (212.67) (190.30) (203.90)
27 46 34 46

 

m—

 

?" ‘

 

Table B-4

Parent Education Group = 2-year/Associate

(Cont’d)

149

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Lowest 60%
731.98 750.40 750.61 771.55 1.662
AMIN (136.76) (129.22) (163.41) (156.64)
101 151 114 71
767.29 769.58 774.73 772.75 0.428
ASIA (168.77) (155.49) (174.62) (164.07)
188 236 237 258
679.85 682.87 675.64 680.29 0.291
BLCK (136.77) (142.46) (139.57) (139.12)
811 950 961 1,070
741.33 745.25 717.14 742.67 0.501
MEX (150.26) (131.30) (133.07) (139.02)
150 179 206 225
736.72 718.00 736.67 764.49 1.185
PRIC (166.09) (151.30) (134.03) (157.12)
58 70 60 89
745.43 725.33 744.77 743.84 -0.244
HISP (141.79) (134.67) (134.09) (154.41)
94 137 128 151
788.02 783.11 783.45 779.00 —2.783*
WHIT (146.63) (150.36) (147.15) (145.90)
4,470 4,989 4,775 4,809
757.67 728.55 752.00 738.94 —0.384
OTHR (172.43) (149.15) (156.54) (163.27)
60 76 110 104

 

 

 

Table B—4

Parent Education Group = 2-year/Associate

(Cont'd)

150

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
Second Quintile
833.68 821.13 821.85 799.04 —1.051
AMIN (174.45) (140.46) (193.04) (153.59)
57 115 65 52
840.18 840.21 847.90 843.24 0.353
ASIA (167.82) (178.08) (179.26) (177.11)
228 290 324 293
743.60 735.62 735.04 730.05 -1.528
BLCK (159.02) (148.91) (147.82) (151.07)
584 710 677 748
816.85 816.65 831.44 807.19 -0.261
MEX (152.73) (150.35) (157.61) (157.00)
143 161 167 185
800.00 819.27 792.81 771.21 —1.055
PRIC (192.66) (191.15) (152.89) (167.37)
51 55 57 66
840.69 848.06 816.06 821.48 —1.184
HISP (177.98) (167.01) (167.62) (176.67)
87 108 127 142
880.34 880.58 879.56 877.92 -0.808
WHIT (153.96) (154.07) (153.60) (152.66)
4,494 4,918 4,578 4,412
807.74 813.22 826.45 861.45 1.711
OTHR (189.75) (186.85) (163.13) (159.08)
62 59 62 55

 

271—3077“

If ‘ inn:

.-

 

151

Table B—4 (Cont'd)

Parent Education Group = 2-year/Associate

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Second 10%
884.81 879.71 887.95 901.82 0.512
AMIN (170.29) (164.65) (177.59) (165.48)
54 70 44 33
921.99 936.91 937.64 927.86 0.345
ASIA (177.77) (166.82) (201.76) (172.17)
176 246 258 285
795.28 802.64 778.28 801.30 0.164
BLCK (169.66) (171.96) (164.10) (170.28)
345 450 448 431
905.96 852.14 885.82 883.71 -0.486
MEX (163.66) (148.96) (172.63) (160.02)
89 126 134 151
797.31 850.59 865.38 884.82 2.219
PRIC (160.41) (181.71) (201.68) (176.85)
26 51 39 56
883.38 912.43 907.75 851.78 —1.125
HISP (174.71) (175.45) (188.43) (171.51)
65 103 102 90
963.16 961.93 954.44 954.24 -2.727*
WHIT (156.22) (159.80) (161.26) (157.34)
3,066 3,483 3,232 3,148
847.94 871.09 898.67 878.03 0.895
OTHR (186.10) (183.03) (173.19) (214.51)
34 46 45 61

 

Table B-4

Parent Education Group = 2-year/Associate

(Cont’d)

152

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Highest 10%
1,002.00 987.14 987.19 996.88 -0.112
AMIN (185.41) (160.42) (192.01) (153.53)
30 42 32 32
1,043.62 1,047.36 1,044.07 1,069.26 1.292
ASIA (187.44) (187.51) (184.89) (190.95)
196 231 248 244
915.30 901.65 892.26 878.09 -2.160
BLCK (194.72) (181.86) (184.46) (185.33)
202 279 265 257
1,025.81 989.17 1,027.32 980.00 —1.190
MEX (171.97) (157.77) (181.09) (182.22)
74 108 97 107
928.10 943.79 904.33 922.14 —0.303
PRIC (195.46) (190.35) (179.97) (230.58)
21 29 30 28
1,012.71 1,028.45 1,016.62 995.19 -0.663
HISP (163.47) (186.37) (159.21) (171.40)
48 58 65 77
1,077.35 1,077.76 1,072.72 1,068.37 -2.321*
WHIT (161.72) (163.97) (160.81) (164.02)
2,782 3,006 2,851 2,778
1,030.83 967.94 1,052.05 1,052.50 1.003
OTHR (202.73) (215.02) (233.23) (190.70)
36 34 39 36

 

 

 

153

Table B-4 (Cont'd)

Parent Education Group = College

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Lowest 60%
795.78 779.02 817.31 784.39 —0.045
AMIN (166.45) (134.49) (155.68) (132.65)
64 112 52 41
790.76 804.90 810.15 806.68 1.028
ASIA (177.29) (177.06) (172.40) (171.32)
197 245 267 268
731.78 708.77 711.00 708.84 -2.040
BLCK (146.48) (141.06) (143.28) (139.52)
354 414 360 422
791.40 749.62 789.64 764.29 -0.450
MEX (136.15) (150.03) (153.26) (169.00)
43 52 55 70
735.36 719.09 726.86 704.29 —0.684
PRIC (168.78) (160.95) (134.51) (170.88)
28 44 35 49
759.06 778.62 782.60 779.44 0.725
HISP (161.89) (145.67) (122.82) (153.74)
53 65 50 71
828.92 828.62 827.41 822.71 —1.661
WHIT (151.50) (152.56) (158.47) (153.52)
3,309 3,558 3,124 3,234
787.93 834.26 787.78 764.53 —1.105
OTHR (187.93) (215.44) (170.17) (171.60)
58 47 54 64

 

 

 

Table B-4

Parent Education Group = College

(Cont’d)

154

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Second Quintile
825.85 850.70 852.14 899.67 1.962
AMIN (139.55) (132.73) (183.74) (169.19)
53 86 42 30
881.46 901.64 906.03 880.51 -0.034
ASIA (177.97) (174.53) (169.87) (179.41)
268 292 368 370
785.84 774.47 786.51 776.74 -0.364
BLCK (161.14) (160.76) (159.01) (162.59)
269 295 315 328
870.21 812.17 875.06 843.54 -0.180
MEX (164.37) (161.66) (160.66) (152.93)
48 60 81 65
820.00 809.71 779.43 791.89 —0.817
PRIC (209.92) (175.32) (157.35) (147.61)
31 35 35 37
857.80 867.26 865.59 859.51 0.035
HISP (182.23) (138.54) (143.03) (168.13)
50 62 68 81
923.52 926.64 915.75 915.13 -3.080*
WHIT (154.12) (156.60) (158.50) (152.89)
3,427 3,804 3,477 3,449
888.49 819.13 906.28 888.59 0.839
OTHR (160.56) (156.69) (188.00) (191.12)
53 46 43 64

 

 

Table B-4

(Cont’d)

155

Parent Education Group = College

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Second 10%
962.14 957.80 932.00 893.08 —1.822
AMIN (119.52) (167.01) (163.99) (171.58)
28 41 30 26
983.59 968.35 995.48 997.49 1.383
ASIA (184.71) (172.30) (177.78) (193.18)
237 340 361 334
828.94 854.53 841.26 827.47 —0.307
BLCK (173.21) (186.25) (175.16) (166.44)
160 201 182 194
915.48 941.57 930.42 941.50 0.603
MEX (158.53) (142.62) (190.97) (160.71)
31 51 48 60
817.69 776.67 803.33 873.57 1.342
PRIC (176.87) (194.39) (122.81) (150.83)
26 18 21 28
957.19 1,000.81 971.25 897.82 —1.708
HISP (177.39) (152.45) (153.83) (171.40)
32 37 48 55
1,009.09 1,008.59 1,004.40 1,005.48 -1.102
WHIT (157.19) (160.50) (156.48) (158.46)
2,611 2,986’ 2,818 2,715
960.68 922.41 995.00 941.50 -0.105
OTHR (187.04) (206.75) (226.28) (216.71)
44 29 38 40

 

 

)

i’:" 5.131 ' _.

 

Table B-4

(Cont’d)

156

Parent Education Group = College

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)1n1
ETHNICITY
Highest 10%
1,137.27 1,008.80 1,066.67 1,095.38 —0.396
AMIN (161.49) (164.43) (177.83) (153.06)
22 25 24 13
1,103.33 1,103.64 1,115.25 1,123.23 1.669
ASIA (163.54) (185.77) (186.25) (189.68)
285 363 394 406
946.37 971.32 958.72 963.61 0.487
BLCK (172.15) (200.30) (177.85) (199.76)
91 114 141 122
1,038.06 1,056.34 1,074.77 1,086.74 1.346
MEX (177.66) (165.16) (132.16) (155.24)
31 41 44 46
888.85 921.92 876.06 946.18 0.754
PRIC (162.88) (194.42) (204.60) (246.38)
26 26 33 34
1,069.19 1,053.64 1,059.41 1,034.19 —0.763
HISP (198.86) (192.19) (193.52) (161.48)
37 33 51 43
1,127.43 1,125.91 1,126.30 1,119.45 —1.720
WHIT (157.79) (156.08) (162.42) (161.18)
2,674 3,001 2,802 2,716
1,089.09 1,137.37 1,045.26 1,039.12 -1.471
OTHR (189.16) (162.09) (203.39) (200.72)
22 38 38 34

 

Table B-4

(Cont’d)

157

Parent Education Group = Graduate/Professional

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)1111
ETHNICITY
Lowest 60%
812.78 796.72 828.04 786.06 -0.513
AMIN (153.06) (141.31) (174.96) (127.82)
54 119 56 33
836.55 837.42 827.89 833.52 -0.365
ASIA (178.01) (188.45) (176.15) (185.06)
200 279 266 330
749.19 728.14 747.66 729.64 —1.104
BLCK (164.23) (165.09) (155.64) (156.01)
397 431 406 420
766.20 815.50 792.67 815.81 1.371
MEX (136.69) (174.53) (165.69) (184.06)
50 60 60 74
759.44 750.21 738.44 740.57 —0.632
PRIC (152.13) (144.64) (154.67) (170.85)
36 48 45 53
769.16 800.56 776.08 787.99 0.483
HISP (140.90) (166.47) (158.62) (162.23)
83 107 120 134
855.11 858.19 856.11 857.28 0.379
WHIT (163.48) (164.97) (167.12) (162.72)
4,033 4,080 3,774 3,794
847.78 866.77 850.56 872.72 0.563
OTHR (180.93) (202.71) (204.23) (205.19)
54 62 71 92

 

 

Table B—4

(Cont'd)

158

Parent Education Group = Graduate/Professional

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)lin
ETHNICITY
Second Quintile
884.63 849.39 851.59 881.74 -0.058
AMIN (153.15) (141.21) (154.47) (180.68)
41 82 44 46
935.68 930.93 961.97 961.76 2.418*
ASIA (190.01) (191.33) (189.93) (189.14)
292 377 375 437
830.27 829.22 824.30 832.59 0.049
BLCK (178.89) (178.33) (181.73) (171.88)
337 357 349 375
837.93 891.59 876.45 875.00 0.986
MEX (186.07) (181.70) (167.26) (156.86)
58 82 76 66
820.65 796.43 810.91 779.36 —0.809
PRIC (169.51) (156.76) (202.33) (209.92)
31 42 44 47
901.07 879.20 896.33 897.10 -0.065
HISP (176.83) (183.53) (180.27) (205.95)
84 113 120 155
962.11 959.41 959.61 957.64 —1.222
WHIT (162.41) (163.02) (163.63) (166.13)
4,652 5,145 4,623 4,525
933.03 956.91 922.65 949.67 0.165
OTHR (193.98) (203.48) (187.68) (197.43)
66 81 68 123

 

 

 

Table B—4

(Cont’d)

159

Parent Education Group = Graduate/Professional

 

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11X1
ETHNICITY
Second 10%
945.43 961.88 929.76 1,026.09 1.630
AMIN (181.68) (138.37) (151.73) (135.87)
35 64 41 23
1,029.12 1,044.86 1,051.92 1,065.49 2.695*
ASIA (200.17) (197.07) (197.62) (183.84)
340 455 453 486
896.73 896.88 890.52 906.78 0.419
BLCK (188.08) (184.05) (182.87) (197.01)
223 253 232 230
957.45 963.90 979.83 951.58 —0.016
MEX (168.52) (193.96) (171.65) (163.86)
47 59 59 57
879.30 861.91 919.14 863.87 -0.076
PRIC (206.53) (202.89) (193.61) (183.44)
43 47 35 31
945.25 981.09 973.27 951.47 0.123
HISP (176.90) (201.84) (193.73) (201.47)
80 119 110 102
1,044.49 1,047.84 1,050.62 1,050.68 1.896
WHIT (159.07) (162.35) (162.64) (165.06)
4,146 4,550 4,097 4,114
1,071.48 1,033.72 1,010.34 1,026.07 -1.717
OTHR (184.46) (218.86) (224.93) (190.71)
81 78 89 89

 

 

Table B-4

(Cont'd)

160

Parent Education Group = Graduate/Professional

 

 

CLASS RANK/ 1988 1989 1990 1991 (TS)11n
ETHNICITY
Highest 10%
1,086.45 1,007.50 1,134.21 1,086.55 0.792
AMIN (187.91) (183.20) (132.97) (209.19)
31 36 19 29
1,203.36 1,195.51 1,208.53 1,210.86 1.109
ASIA (177.31) (181.69) (181.71) (185.90)
557 712 685 735
1,001.96 1,006.96 1,010.62 973.41 —1.304
BLCK (190.24) (195.08) (181.99) (183.45)
179 181 193 173
1,066.19 1,055.52 1,056.56 1,064.44 —0.040
MEX (157.79) (183.18) (173.20) (176.62)
42 58 61 63
1,021.11 985.79 1,055.17 1,026.79 0.496
PRIC (172.03) (211.51) (245.38) (231.97)
27 38 29 28
1,110.76 1,078.21 1,057.96 1,078.98 -1.434
HISP (170.84) (183.38) (200.32) (183.60)
92 84 93 108
1,169.94 1,171.08 1,174.02 1,171.10 0.627
WHIT (158.50) (162.95) (160.91) (162.37)
4,847 5,495 5,101 4,917
1,222.41 1,174.44 1,166.14 1,146.84 -2.793*
OTHR (161.91) (214.52) (196.31) (186.27)
79 81 114 98

 

*t-ratio for linear contrast,

a-level of 0.0125.

statistically significant at an

 

APPENDIX C
ADJUSTED MEAN SAT-TOTALS FOR 1988-1991

 

 

APPENDIX C
ADJUSTED MEAN SAT-TOTALS FOR 1988-1991

 

 

 

 

Table C—1
.0: ’ ‘0 0 9". -0._ . e e. .‘0 e 00 u. e 0
WW5
1988
Parameter Predictor 1983 1989 1990 1991
Mean
Intercept 614.317 614.794 592.487 585.640
PED 15.906 19.139 19.124 20.551 20.824
BLCK 0.083 —164.989 -165.162 -163.838 -166.371
ASIA 0.059 1.216 3.027 12.279 15.592
MEX 0.021 -80.117 -83.921 -79.930 -78.288
HISP 0.017 -88.830 -88.366 -85.796 -90.614
OTHR 0.012 -48.336 -44.907 -45.442 -52.653
AMIN 0.011 -87.368 -89.017 -87.294 -74.441
PRIC 0.009 -138.928 -136.143 -136.987 -123.359
RSEX 0.478 50.265 54.045 48.026 44.506
Adjusted Means 923.16 925.28 923.58 919.51

 

 

161

 

 

 

 

 

 

 

 

 

162
Table C-2
°:: - -. . 0-.. -.. . - -. :-. - .. u. - .
Wm
Puerto Ricangsubgroup
Parameter Pre1d9i8c8tor 1988 1989 1990 1991
Mean
Intercept 614.317 614.794 592.487 585.640
PED 15.906 19.139 19.124 20.551 20.824
PRIC 1.000 —138.928 -136.143 -136.987 -123.359
RSEX 0.478 50.265 54.045 48.026 44.506
Adjusted Means 803.84 808.67 805.34 814.78 “
Asian-American subgggup
Parameter Pre1d918c8tor 1988 1989 1990 1991
Mean
Intercept 614.317 614.794 592.487 585.640
PED 15.906 19.139 19.124 20.551 20.824
ASIA 1.000 1.216 3.027 12.279 15.592
RSEX 0.478 50.265 54.045 48.026 44.506

 

Adjusted Means 943.98 947.84 954.61 953.73

 

 

APPENDIX D
SENSITIVITY ANALYSES, ADJUSTED MEAN SAT-TOTALS

 

APPENDIX D
SENSITIVITY ANALYSES, ADJUSTED MEAN SAT-TOTALS

Two additional regression models were used to compute
adjusted mean SAT-Total scores as tests of sensitivity to the
prediction model which only used demographic characteristics.
The first of these models included the ”class rank” main
effect in addition to the main effects of demographic
characteristics as defined in this study. The second of
these models included a two-way interaction effect of ”parent
education-by—class rank.”

Although the composition of examinees in the samples did
not vary between 1988 and 1991 on the basis of class rank
alone, class rank was strongly correlated to performance on
the SAT, and was therefore included as a factor in addition
to demographic predictors in a separate prediction model.

The basic prediction equation is:

Y1 = Bo + BMParent Ed)1-+ BAEthnich +

E t' 2
03(RSEX)1 + B4(RANK)1- , e1. ( qua 10n )

Parent education, as an index of SES, is traditionally
highly correlated with high school class rank: The higher
the level of parent's educational attainment, the better the
index of academic performance in high school. A deviation
from this relationship was detected for immigrant populations
and linguistic minorities from low parent education
backgrounds but whose parents nonetheless held high

expectations of academic performance and scholastic

163

 

 

164

achievement. Hence, the two-way interaction effect of
”parent education-by-class rank” was included as a factor in
the second prediction equation of the sensitivity analyses.

The basic prediction equation is:

Y1 = B0 + B1(Parent Ed): + fl2(EthniC)1 +

E t' 3
BMRSEXM + BdRANKh + BHPEDRANKM + 8L ( qua ion )

The 1988 means of each predictor in both predictor
equations were computed, and separate year regression
coefficients were estimated. Adjusted mean SAT-Total scores
were predicted using the same 1988 predictor means,
concurrently allowing the set of regression coefficients to
vary by cohort year. These results appear in Tables D-1 and

D-2.

 

 

165

Table D-l

Intact Annual Cohorts

 

 

 

Parameter Pre1d9i8c8tor 1988 1989 1990 1991
Mean
Intercept 404.896 408.054 389.711 387.408
PED 15.906 15.559 15.282 16.322 16.645
BLCK 0.083 -137.950 -138.980 ~137.116 -l38.893
ASIA 0.059 -21.665 -18.655 -9.383 -4.625
MEX 0.021 -85.786 -85.761 —78.344 -76.996
HISP 0.017 -80.213 -79.089 -78.233 -80.787
OTHR 0.012 -41.157 -43.500 -41.272 -44.261
AMIN 0.011 -61.286 ~61.643 -59.505 -55.909
PRIC 0.009 -128.327 -124.673 -123.453 -106.636
RSEX 0.478 56.309 58.560 54.967 51.125
RANK 67.145 3.903 3.914 3.931 3.882

 

Adjusted Means 923.14 923.82 922.38 920.36

 

 

 

 

 

 

 

 

 

 

166
Table D-l (Cont’d)
Puerto Rican subgroup
1988
Parameter Predictor 1988 1989 1990 1991
Mean
Intercept 404.896 408.054 389.711 387.408
PED 15.906 15.559 15.282 16.322 16.645
PRIC 1.000 -128.327 —124.673 —123.453 -106.636
RSEX 0.478 56.309 58.560 54.967 51.125
RANK 67.145 3.903 3.914 3.931 3.882
Adjusted Means 813.03 817.25 816.10 830.62
Asian-American subgroup
1988
Parameter Predictor 1988 1989 1990 1991
Mean
Intercept 404.896 408.054 389.711 387.408
PED 15.906 15.559 15.282 16.322 16.645
ASIA 1.000 -21.665 -18.655 -9.383 -4.625
RSEX 0.478 56.309 58.560 54.967 51.125
RANK 67.145 3.903 3.914 3.931 3.882

 

Adjusted Means 919.69 923.27 930.17 932.63

 

 

Intact Annual Cohorts

167

 

 

Earmmmer Prdifiior 1988 1989 1990 1991
Mean

Intercept 558.294 558.077 552.316 548.619
PED 15.906 5.763 5.644 5.842 6.244
BLCK 0.083 ~138.336 -139.106 -137.526 -139.139
ASIA 0.059 -22.170 -18.892 -9.537 -4.751
MEX 0.021 ~80.753 -80.356 -78.272 -76.612
HISP 0.017 -80.045 -79.253 -78.294 -80.760
OTHR 0.012 -44.654 -43.623 -42.017 -44.223
AMIN 0.011 -61.872 -59.341 -59.443 -56.326
PRIC 0.009 -129.270 -124.827 -124.067 -103.637
RSEX 0.478 56.567 58.868 55.106 51.296
RANK 67.145 1.631 1.704 1.534 1.487
PED/RANK 1077.149 0.144 0.140 0.153 0.153
Adjusted Means 923.40 923.21 922.16 920.21

 

 

 

168

Table D-2 (Cont’d)

Puerto Rican supgroup

 

 

 

Parameter PreldgchBtor 1933 1939 1990 1991
Mean

Intercept 558.294 558.077 552.316 548.619
PED 15.906 5.763 5.644 5.842 6.244
PRIC 1.000 -129.270 —124.827 -124.067 -103.637
RSEX 0.478 56.567 58.868 55.106 51.296
RANK 67.145 1.631 1.704 1.534 1.487
PED/RANK 1077.149 0.144 0.140 0.153 0.153
Adjusted Means 812.35 816.38 815.82 833.47

 

Asian-American subgroup

 

 

Pammeter Préfifiior 1988 1989 1990 1991
Mean
Intercept 558.294 558.077 552.316 548.619
PED 15.906 5.763 5.644 5.842 6.244
ASIA 1.000 —22.170 -18.892 -9.537 -4.751
RSEX 0.478 56.567 58.868 55.106 51.296
RANK 67.145 1.631 1.704 1.534 1.487
PED/RANK 1077.149 0.144 0.140 0.153 0.153
Adjusted Means 919.45 922.31 929.85 932.35

 

For the overall group, both prediction models as tests
of sensitivity yielded virtually identical adjusted mean SAT-
Total scores. The inclusion of a two-way interaction effect
of ”parent education-by-class rank” yielded no discernible
difference in the trend of adjusted mean scores over ”class
rank” and demographic main effects. Compared to the model
using demographic predictors alone, the addition of class

rank as 61 factor accounted for one additional scale-score

 

 

 

169

point of the unadjusted nine-point decline. Whereas changes
in demographic characteristics alone accounted for five
points of the nine point decline, additionally adjusting for
changes in class rank composition between 1988 and 1991
accounted for six points of the nine point decline.

For the Puerto Rican subgroup, adding main effect ”class
rank” to demographic predictors resulted in a net mean 17-
point increase between 1988 and 1991. For the Asian-American
subgroup, adding class rank to demographic characteristics in
the prediction equation resulted in a net mean 13-point
increase between 1988 and 1991. For Puerto Rican and Asian-
American ethnic subgroups, adding the ”class rank” main
effect enhanced the trend in adjusted mean SATsTotal scores
over demographic predictors alone by six scale-score points,
and three scale-score points, respectively.

The two-way interaction effect, ”parent education-by-
class rank,” enhanced the positive trend in adjusted mean
SAT-Total scores, but only for the Puerto Rican ethnic
subgroup, an effect of three additional scale-score points
over ”class rank” and demographic predictors. For both
Puerto Rican and .Asian-American subgroups, class rank
enhanced the positive trend in adjusted mean SAT-Total scores

between 1988 and 1991.

 

LI ST OF REFERENCES

 

LI ST OF REFERENCES

American Association of Collegiate Registrars and Admissions
Officers, & the College Board. (1980). undergraduate

1' . _ 1 1' . E . . 1 1' .
W. New York: College Entrance

Examination Board.

American College Testing Program. (1977). The High Seheel
W. Iowa City, IA: American

College Testing Program.

American Council on Education. (1967). National norms for
entering college freshmen: Fall 1966. AQE_Beeee;eh

Reports, 10.1.52, (1).

American Council on Education. (1972). The American
freshman: National norms for Fall 1972. ,AflE_Reeeeren

Benorts. 18211.1, (5).

Angoff, W. H. (1971). The Admissions Testing Program. In
w. H. Angoff (Ed.), Waring
WW

. . . 1 . 1 1 J . . i
AenieyemenL_TeeLe. Princeton, NJ: College Entrance

Examination Board.

Angoff, w. H. (1984). W.

Princeton, NJ: Educational Testing Service.

Angoff, W. H. (1986). Some contributions of the College
Board SAT to psychometric theory and practice.

W741

Baird. L. L. (1976). MW
nerfermenee. [College Board Research Memorandum, No.
76-7.] New York: College Entrance Examination Board.

Beaton, A. E., Hilton, T. L., & Schrader, W. B. (1977).
Changes in the verbal abilities of high school seniors,
college entrants, and SAT candidates between 1960 and

1972. Appendix to W
LWW

deeline. New York: College Entrance Examination Board.

." 2LR-n-T V

 

171

Bejar, I. I., & Blew, E. O. (1981). Grade inflation and the
validity of the Scholastic Aptitude Test. Ameriean

WW. 18(2) , 143-156.

Boldt, R. F. (1986). W
eelleges. [College Entrance Examination Board RR 86-3 &

Educational Testing Service RR 86-24.] New York:
College Entrance Examination Board.

Braswell, J. (1991). deryiem Qf_shanges_in_the_SAI
W Paper presented at the annual

meeting of the National Council on Measurement in '
Education, Chicago, IL.

Breland, H. M. (1979). WWW
En;ranee_Measnres. [College Board Research Monograph,
No. 79—08.] New York: College Entrance Examination
Board.

Coffman, w. E. (1966). W
W. [College
Board Research and Development Report 65-6, No. 17.]
Princeton, NJ: Educational Testing Service.

 

College Entrance Examination Board. (1977). Qn_£nrrher
E . . . E 1 3 . J 1
W New York: College

Entrance Examination Board.

College Entrance Examination Board. (1980). gellegeehennd
i . . 155: E'J E i 3 1. _
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1981). gellegeehennd
. . 1581 E'] E SEI i E 1. I !_
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1982). gellegeehennd
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1983). gellege;bennd
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1984). Cellegeehennd
Takers. Princeton, NJ: Educational Testing Service.

 

172

College Entrance Examination Board. (1985). Cellegeehennd
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1986a). gellegeehennd
i . . 1585 E'] E i i 1. _
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1986b). Measures_in
the_College_Admissions_£rocess, New York: College

Entrance Examination Board.

College Entrance Examination Board. (1987). Cellegeebennd
i . . JEEZ E'] E i i 1. _
Takers. Princeton, NJ: Educational Testing Service.

College Entrance Examination Board. (1991). gellege;bennd
i . . 195] E'] E i i 1. _
Takers. Princeton, NJ: Educational Testing Service.

Crouse, J. (1985). Does the SAT help colleges make better
selection decisions? Haryard_Educational_Bexiew 55.
195— 219.

Donlon, T. F. (Ed.). (1984). Ihe_College_Board_Iechnical

3] 1 E 1 E 1 J . . 3 j
Aehieyemenr_Tesrs, New York: College Entrance

Examination Board.

Gottfredson, L. S. & Crouse, J. (1986). Validity versus
utility of mental tests. Example of the SAT. Senrnal

of_Yocational_Behayior. 29(1), 363-78.

Hunter, J. E. (1980). The_dimensienalitx_of_the_seneral
. .
AntitudeETest_5atterx_lQAIE%Tan%_the_dominapce_of:r .

ef_ieh_nerfermanee. Washington, D.C.: United States
Employment Service.

Hunter, J. E. (1986). Cognitive ability, cognitive
aptitudes, job knowledge, and job performance. Jenrnal

of_Yocational_Beharior 22(3), 340-362.

Hunter, J. E. , & Hunter, R. F. (1984). Validity and utility
of alternate predictors of job performance.

Psychological_5nlletin 25, 72- 98.

Jackson, R. (1980). The Scholastic Aptitude Test: A
response to Slack and Porter' 5 ”Critical Appraisal.

Hafrard_Edncational_Eeyiew 50(3), 382- 391.

 

 

173

Jacobson, R. L. (1986). Selective colleges’ use of SAT is
unshaken by controversies. Chrgn1cle_efi_fi1gher
Educat19n. 52(18). 1. 8.

Marco, G. L., Petersen, N. S., & Stewart, E. E. (1983). A
_ J J . E . 3 'J'
Egnar1ng_Mede1s. [Educational Testing Service Research
Memorandum, No. 83—2.] Princeton, NJ: Educational
Testing Service.

McGee, V. E. (1961). Towards a maximally efficient system
of braiding for SAT equating. In S. S. Wilks (Ed.),
5cal1ng_and_EQnat1ns_£911ege_59ard_1ests. pp 86 - 96.
Princeton, NJ: Educational Testing Service.

Modu. c. C.. & Stern. J. (1975). Ihe.&tahilitx_gf_the_5AT
Seere_§ea1e. [College Board Research and Development
Report 74-75, No. 3.] Princeton, NJ: Educational
Testing Service.

Modu, C. C., & Stern, J. (1977). The stability of the SAT-
verbal score scale. Appendix to Qn_£urrher_flxam1nar1enr
RenQzL_Qf_the_AdM1sorx_2anel_Qn_the_5;holastic_Ant1tude
Tesr_Seere_Deel1ne. New York: College Entrance
Examination Board.

National Commission on Excellence in Education. (1983). A
Nar1en_ar_k1sk. Washington, DC. U. S. Department of
Education.

Pepin, A., Knepper, P., Bales, S. N., Bartell, E., Shulman,
c. H., & Williams, M. c. (1982). Trends_and_£atterns.

Washington, D.C.: American Council on Education.

Pruzek, R. M., & Coffman, W. E. (1966). .A_£aerer_Analys1s
E 1 1 . J i . E 1 ] J . . 3
Tesr. [College Board Research and Development Report
65-6, No. 10.] Princeton, NJ: Educational Testing
Service.

Rock, D. A., & Werts, C. E. (1979). Censrrner_ya11d1ry_er
1 E J . _ . . J : E'
srudy. [College Board Research and Development Report

78-79, No. 5]. Princeton, NJ: Educational Testing
Service.

Schaeffner, P. E. (1985L Competitive admission practices

when the SAT is optional Q9urnal_gf_fligher_fidusatign.
55(1). 55- 72.

v:

 

 

174

Slack, W. V., & Porter, D. (1980). The Scholastic Aptitude
Test: A critical appraisal. Haryard_£dnear1ena1
Rex1ew. 59(2). 154—175.

Smith. c. L. (1980). Se1ected_pred1ctors_pf_academ1c
ach1eyement_of_cQllege_freshmen_u11h_lomer_range
admiss1pns_test_scpres. Unpublished doctoral

dissertation thesis, University of Alabama.

Stewart. E. E. (1966). Ihe_5tab111tx_pf_the_5AI;Merbal
Seare_Sea1e. [Research Bulletin RB-66-37, College
Entrance Examination Board RDR-66-7.] Princeton, NJ:
Educational Testing Service.

Thorndike, R. L. (1986). The role of general ability in
prediction. Jonrnal_pf_YQcat1onal_Behay1or. 22(1). 332-
39.

Trusheim, D. w. (1984). Effects_pf_collegesl_use_of_the_531
i . . j . . .
Unpublished doctoral dissertation thesis, University of
Delaware.

Trusheim, D. W., & Crouse, J. H. (1982). The ETS admissions
formula: Does the SAT add useful information? £h1

WI 5.4(l)! 59‘61-

Trusheim, D. W., & Crouse. J. H. (1984). The SAT and
traditional predictive validity: A critical assessment.

J9nrnal_pf_Cpllege_Adm1ss1ons. 194. 8—12.

Turnbull. w. w. (1985). 5Ludent_Change1_£rpgram_9hange1

Whx_1he_SAT_Scores_Kept_£all1ng. [College Board Report
No. 85-2 & Educational Testing Service RR No. 85-28.]
New York: College Entrance Examination Board.

U. S. Department of Education, U. S. Office of Educational
Research & Improvement, Center for Education Statistics,
National Center for Education Statistics. (1992).
D1gest_of_Edncatipn_51a11sticsi_1222. Washington. D.C.:
Author.

Weitzman, R. A. (1982). The prediction of college
achievement by the Scholastic Aptitude Test and the high
school record. aonrnal_of_Edncat1pnal_Measnrement.
12(1). 179-91.

wilks. s. 3. (Ed.). (1961). 5cal1ns_and_Egnat1ns_Cpllege
Beard_Tesrs. Princeton, NJ: Educational Testing
Service.

175

Willingham, w. w. (1986). WWW
mmmaumummmmnm New York:

College Entrance Examination Board.

Willingham, W. W., & Breland, H. (1982). Persena1_Qna11r1es
and_Cgllege_Adm1ss1ens. New York: College Entrance

Examination Board.

Wilson, K. M. (1983). WW
3' . E i E E 1 1
Year. [College Board Report No. 83-2 and Educational

Testing Service Research Report No. 83-2.] New York:
College Entrance Examination Board.

176
General_References

Astin, A. W., King-Hemond, M., & Richardson, G. T. (1982).
1 . 1 . . J E J] 1582.
Los Angeles: American Council on Education.

Bishop, Y., Finberg, S., & Holland, P. (1975). D1serere
Mult1yar1ate_Analxs1s1__Ihegrx_and_Eractice. Cambridge.
MA: MIT Press.

Burton, N. W., Lewis, C., & Robertson, N. (1988). Sex
D1fferences_1n_SAI_Scpres. [College Entrance
Examination Board Research Report RR 88-9 & Educational
Testing Service Research Report RR 88-58.] New York:
College Entrance Examination Board.

Burton, N. W., Morgan, R., Lewis, C., & Robertson, N. J.
(1989). Ihe_pred1ct1xe_xal1d112.91.5A1_and_ISWE_1tem
txpes_1pr_ethnic_and_gender_groups. Paper presented at
the annual meeting of the American Educational Research
Association, San Francisco, CA.

Kelly, F. J., Beggs, D. L., McNeil, K. A., Eichelberger, T.,
& Lyon. J. (1969). Mnlt1ple_Regress1cn_Apprpach.
Carbondale, IL: Southern Illinois University Press.

Mosteller, F., & Tukey, J. W. (1977). Dara_Ana1ys1s_and
Regress1en. Reading, MA: Addison-Wesley.

National Center for Education Statistics. (1982). The
Cond1t1pn_of_3ducat1gn. Washington. D c.: National
Center for Education Statistics.

Neter, J., & Wasserman, W. (1974). Appl1ea_L1near
Statist1gal_Mede1s. Homewood, IL: Richard D. Irwin,
Inc.

Weisberg. S. (1985). Appl1ed_L1near_Begress1on. New York:
John Wiley & Sons, Inc.

 

 

 

11111111111):