Interaction Effects in Spady and Tinto's Conceptual Models of College Attrition

Author(s): Ernest T. Pascarella and Patrick T. Terenzini

Source: Sociology of Education , Oct., 1979, Vol. 52, No. 4 (Oct., 1979), pp. 197-210
Published by: American Sociological Association

Stable URL: https://www.jstor.org/stable/2112401

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend
access to Sociology of Education

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN SPADY'S AND TINTO'S CONCEPTUAL
MODELS OF COLLEGE DROPOUT*

ERNEST T. PASCARELLA PATRICK T. TERENZINI

University of Illinois, Chicago Circle State University of New York at Albany

Sociology of Education 1979, Vol. 52 (October):197-210

This study investigated the main and interaction effects of student characteristics and measures
of social and academic integration on voluntary freshman withdrawal decisions. After
controlling for the influence of twelve student entering characteristics, thirteen measures of
social and academic integration contributed significant R,2 increases in the explanation of
voluntary withdrawalfrom college. The particular influence of specific measures of social and
academic integration for both sexes, however, was dependent upon levels of student entering
characteristics and levels of social and academic integration in other areas. The most consis-
tent pattern of interaction effects concerned the influence of the frequency and quality of
student-faculty relationships. Such relationships were most important in positively influencing
the persistence of freshmen with entering characteristics and levels of academic and social
integration predictive of withdrawal.

In their explanatory models of the col- collegiate experiences on persistence, but

lege dropout process both Spady (1970, also for investigating potentially important
1971) and Tinto (1975) have emphasized interrelationships between and within
the processes of social and academic inte- these- sets of variables. While the former
gration as critical influences on student focuses on questions of "main effects"
persistence, Each model views (e.g., What are the relative contributions
persistence/withdrawal decisions largely of measures of social and academic inte-
as the result of a longitudinal process of gration when student entering char-
associations between the student and the acteristics are taken into account?), the
academic and social systems of the in- latter is concerned more with questions of
stitution. The student comes to a particu- statistical interactions (e.g., What are the
lar institution with background char- relative influences of different types of so-
acteristics which; in turn, partially deter- cial and academic integration for different
mine how the student will relate to an kinds of students?).
institution's social and academic systems. Based on the extensive lists of studies
The nature and quality of these associa- reviewed by Spady (1970), Tinto (1975)
tions "lead to varying levels of normative and, more recently, Pantages and Creedon
and structural integration in those col- (1978), it would appear that the investiga-
legiate systems" (Tinto, 1975, p. 103). As- tion of interaction effects has received
suming that external influences are held relatively little attention. As a result, a
constant, the higher the levels of integra- number of important sociological and
tion into the social and academic systems educational issues vis-a-vis college stu-
of an institution the less likely the student dent attrition have yet to be addressed.
is to withdraw voluntarily. For example, are certain aspects of social
The complexity of these two models is and academic integration most important
perhaps a strength in that they provide a in positively influencing the persistence of
conceptual framework not only for students with initially high levels of pre-
assessing the relative influence of student college characteristics such as degree as-
entering characteristics and subsequent piration, commitment to college comple-
tion or academic aptitude? Such interac-
* An earlier version of this paper was presented at tion effects would suggest that the major
the annual meeting of the Amprican Educational Re-
benefits of certain college experiences and
search Association, San Francisco: April 8, 1979.
involvements tend to accrue to those stu-
Address correspondence to Dr. Ernest T. Pascarella,
College of Education, University of Illinois, Chicago dents whose entering levels of educational
Circle, Box 4348, Chicago, Illinois 60680. aspiration and aptitude make their persis-

197

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 198 PASCARELLA AND TERENZINI

tence in college quite likely to begin with. particularly important influence on both
For the purpose of this study such in- academic and social integration. Indeed,
teraction effects will be referred to as ac- evidence from a number of studies
centuating. I (Gekoski and Schwartz, 1961; Pascarella
Conversely, do the influences of other and Terenzini, 1976, 1977; Spady, 1971)
experiences within the institution's aca- suggests that such aspects of student-
demic and social systems tend to compen- faculty relationships as the frequency of
sate for initially low levels of such enter- student-faculty informal contact beyond
ing characteristics as degree aspiration, the classroom are in fact positively asso-
commitment to college completion, aca- ciated with college persistence.
demic ability or parental educational at- Moreover, Tinto (1975) suggests the
tainment? Such interaction effects (re- possibility of an interaction between ex-
ferred to as compensatory) would suggest tent of social contact with faculty and the
that the benefits of certain college experi- sex of the student, social contact being
ences may be sufficient to override enter- somewhat more important for women
ing traits which often typify the dropout- than men. Thus, within the framework of
prone student. identifying accentuating and compensa-
A conceptually related question tory influences in the Spady-Tinto mod-
suggested by Tinto (1975) is whether or els, a particular focus was given in this
not high levels of social integration can study to measures of student-faculty re-
compensate for low levels of academic lationships.
integration in influencing persistence, and Specifically, the present study had three
vice versa. Similarly, Spady (1970) has basic purposes: 1) to determine the inter-
suggested the concept of "sufficiency": active influence of measures of social and
once a threshold level of integration has academic integration with various mea-
been achieved, the student's continuation sures of student entering characteristics in
in school is more likely. Spady applies the prediction of voluntary persistence/
this view to both social and academic withdrawal decisions, 2) to identify in-
integration simultaneously; however, it teractions between measures of social
may well be that a "sufficient" level of integration and measures of academic
integration attained in one domain (either integration, and 3) to identify the interac-
academic or social) may counterbalance tions between student-faculty relation-
or even outweigh a lack of integration in ships and various student entering char-
the other. The present study sought to acteristics and other forms of social and
identify such compensatory or accen- academic integration.
tuating interactions within the conceptual
framework of the Spady-Tinto models.
It is clear from both models that differ- METHOD
ent types of collegiate involvements and
Design and Sample
experiences are conceptually associated
with different types of integration (e.g., The study was longitudinal and was
relationships with peers tend to influence conducted at a large, independent, resi-
social integration while grade perfor- dential university in New York State with
mance essentially influences structural aca- a total undergraduate enrollment of ap-
demic integration). One aspect of the stu- proximately 10,000 students. In July 1976,
dent's experience, however-his or her a simple random sample of 1905 persons
relationships with faculty-is seen as a was drawn by computer from the total
population of incoming freshmen. Sample
1 An analogous, though not totally comparable, members were sent a detailed question-
type of accentuating influence has been suggested by naire designed to assess their expectations
Feldman and Newcomb (1969), Newcomb, Brown, of a variety of aspects of the college expe-
Kulik, Reimer, and Revelle (1970) and Wilson, Gaff,
rience, as well as to collect selected back-
Dienst, Wood, and Bavry (1975) in considering the
effects of different college involvements and inter-
ground information. Usable responses
personal experiences on various educational out- were received from 1457 students (76.5
comes other than persistence. percent who subsequently enrolled. Dur-

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 199

ing the spring semester of the following need to take such pre-enrollment or en-
year (1977), a second questionnaire was tering characteristics into account. The
mailed to these 1457 students seeking in- following measures of entering char-
formation on the reality of their college acteristics, suggested as potentially im-
experience. After a mail follow-up, usable portant correlates of persistence by var-
responses were received from 773 ious critical reviews of college attrition
freshmen (53.1 percent response). Chi- research (e.g., Cope and Hannah, 1975;
square goodness-of-fit tests (used to de- Pantages and Creedon, 1978; Spady, 1970;
termine the similarity of sample and Tinto, 1975), were statistically controlled
population characteristics) indicated that in the study:
the 773 freshmen were representative of
Sex
the freshman population from which they Racial/Ethnic Origin (non-minority or
were drawn with respect to sex, racial/ minority)
ethnic origin, college of enrollment, aca- Initial Program of Enrollment (liberal arts or
demic aptitude (SAT scores) and professional, decided on prior to registra-
freshman year cumulative grade-point av- tion)
erage. Academic Aptitude (combined Scholastic
Aptitude Test (SAT) scores)
A review of each student's records in
High School Achievement (percentile rank
September 1977, indicated that 10 of the
in high school class)
773 had been dismissed or advised to
Number of High School Extracurricular Ac-
withdraw because of unsatisfactory aca- tivities (of 2 hours or more per week on the
demic performance, and 90 (50 men and 40 average)
women) had voluntarily withdrawn from Expected Number of Informal Contacts with
the university at the end of their freshman Faculty (per month of ten minutes or more
year. Since the study focuses on voluntary outside of class)
withdrawal, the 10 academic dismissals Parents' Combined Annual Income (in
were excluded from all further analyses. thousands of dollars)
Mother's and Father's Combined Formal
(The voluntary withdrawal rate of 11.8
Education (seven categories from "some
percent in the sample is comparable to
grammar school" to "graduate degree")
previous known freshman year voluntary
Student's Highest Expected Academic De-
withdrawal rates at this institution.) gree (Bachelors to Ph.D., Ed.D., M.D.,
While it might be argued that both the J.D.)
Tinto and Spady models are intended to Importance of Graduating from College
explain attrition during the second, third ("Extremely important" to "not at all im-
or fourth years of college as well as in the portant")
first year, evidence from Iffert (1958), Rank of this University as a College Choice
Eckland (1964), Marsh (1966), and (1st Choice to 4th or lower choice)
Pre-enrollment Confidence that Choosing to
Rootman (1972) strongly suggests that
Attend this University was the Right Deci-
attrition is heaviest at the end of the
sion ("extremely confident" to "not at all
freshman year. Based on this evidence, it
confident").
was judged that analyses using a
freshman year sample would provide a With the exception of academic
reasonable estimate of those aspects of achievement, which was freshman year
the models being investigated. grade point average obtained from official
university records, measures used to
assess the concepts of social and aca-
Variables and Instruments
demic integration were obtained from stu-
According to the Spady-Tinto models dent responses on the follow-up question-
student background characteristics, en- naire. The follow-up instrument asked
tering aspirations and pre-college experi- students to: 1) indicate their extent of in-
ences influence both persistence/ volvement in extracurricular activities
withdrawal decisions and the quality of during the freshman year (of two hours or
the students' experiences with an institu- more per week on the average), 2) respond
tion's social and academic systems. Thus, to a series of 34 Likert items designed to
attempts to operationalize the models measure various dimensions of social and

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
200 PASCARELLA AND TERENZINI

Table 1. Alpha Reliabilities and the Two High;est Loading Items on Five Factorially Derived Measures of
Social and Academic Integration

Factor Alpha
Scale/Items Loading Reliability

Peer-Group Relations (7 items) .84

1. Since coming to this university I have developed close personal relationships
with other students. .82
2. The student friendships I have developed at this university have been personally
satisfying. .82
Informal Relations With Faculty (5 items) .83
I . My non-classroom interactions with faculty have had a positive influence on my
personal growth, values and attitudes. .86
2. My non-classroom interactions with faculty have had a positive influence on my
career goals and aspirations. .83
Faculty Concern for Teaching and Student Development (5 items) .82
I . Few of the faculty members I have had contact with are generally interested in
students.a - .77
2. Few of the faculty members I have had contact with are generally outstanding or
superior teachers.a -.72
Academic and Intellectual Development (7 items) .74
I . I am satisfied with the extent of my intellectual development since enrolling in
this university. .68
2. My academic experience has had a positive influence on my intellectual growth
and interest in ideas. .67
InstitutionallGoal Commitment (6 items) .71
1. It is important for me to graduate from college. .69
2. I am confident that I made the right decision in choosing to attend this univer-
sity. .63

a Scores on

academic integration suggested by the number of times during each semester of

Spady-Tinto models, and 3) indicate the their freshman year they had met infor-
frequency of their informal contact with mally, outside of class, with a faculty
faculty for six different purposes. member for the following purposes: "to
A principal components factor analysis get basic information and advice about my
of the Likert-item instrument yielded five academic program," "to discuss matters
factors termed: "peer group relations," related to my future career," "to help re-
"academic and intellectual develop- solve a disturbing personal problem," "to
ment," "informal relations with faculty," discuss intellectual or course-related
"faculty concern for teaching and student matters," "to discuss a campus issue or
development" and ",institutional/goal problem" and "to socialize informally."3
commitment."2 The alpha (internal con-
I A preliminary examination of the distributions of
sistency) reliabilities and the two highest
the frequency of contact in each category indicated a
loading items on each scale are shown in substantial positive skew (i.e., a large proportion of
Table 1. (The complete item sets and fac- the students had few or no contacts while a compar-
tor matrix are available on request.) atively small number had many contacts). A similar
examination of the distributions of all other variables
In order to measure frequency of stu-
in the investigation also indicated a substantial posi-
dents' informal contact with faculty, re- tive skew in four variables: number of high school
spondents were asked to indicate the extracurricular activities, expected informal contact
with faculty, parents' combined annual income and
2 While it was initially expected that items freshman year extracurricular involvement. In cor-
measuring primarily commitment to the goal of relational analyses such skewed distributions can
graduation and those measuring primarily institu- lead to the inordinate impact of outlier cases and,
tional commitment might form two separate factors, quite possibly, to specious results (Walberg and
the rotation of different numbers of factors as well as Rasher, 1976). To reduce the impact of such outlier
the use of different selection techniques for rotation cases and to reduce skewness in the distributions,
(e.g. eigenvalues = 1.00 or greater versus the scree each of the above variables was transformed to a
test) failed to support this expectation. It was there- natural logarithm prior to any statistical analysis
fore decided to proceed on the evidence of the factor (Walberg and Rasher, 1976). To avoid the indeter-
analysis even though it differed slightly from Spady minancies of taking a natural log of 0.0 a constant of
and Tinto's theoretical expectations. 1.0 was added to each measure.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 201

Only contacts of 10 minutes or more were While Tinto's (1975) graphic display of
to be counted. This operational measure his model appears to place institutional
of student-faculty contact was drawn from and goal commitment at the beginning and
an instrument employed by Wilson, end of the student's college career, it
Wood, and Gaff (1974). The six types of seems reasonably clear from his discus-
contact generally appeared to fall into the sion that it is an ongoing process which
two broad categories of academic integra- both influences and, in turn, is influenced
tion (academic advising, intellectual dis- by students' experiences with the social
cussions, career concerns) and social and academic systems of the institution. It
integration (campus issues, socializing was therefore decided to let actual volun-
informally and resolving personal prob- tary freshman year persistence/
lems).4 On an earlier sample the same six withdrawal decisions represent outcome
types of contact were found to discrimi- levels of institutional/goal commitment,
nate significantly between voluntary and to treat the measure of institutional/
freshman year withdrawals and persisters goal commitment obtained early during
with the influence of six academic apti- the spring semester as an additional indi-
tude and personality dimensions taken cator of social and academic integration.
into account (Pascarella and Terenzini, The dependent variable, freshman year
1977). voluntary persistence/withdrawal behav-
The variables were grouped in concep- ior was dummy coded I = persisters and 0
tual categories of social and academic = voluntary withdrawals. Data on this
integration as follows: social inte- variable were obtained from official uni-
gration-involvement in extracurricular versity records.
activities, the peer-group relations and
informal relations with faculty scales, and
Statistical Analysis
frequency of informal contacts with fac-
ulty to discuss campus issues, to socialize Prior analysis of these data to determine
informally and to resolve a personal the predictive validity of the five facto-
problem; academic integration- rially derived Likert-item scales indicated
freshman year grade point average, the significant interactions between sex of
academic and intellectual development student and the peer-group relations and
and faculty concern for teaching and stu- institutional/goal commitment scales
dent development scales, and informal (Pascarella and Terenzini, in press).
contacts with faculty to obtain advice and Peer-group relations had a stronger posi-
information about academic programs, to tive influence on persistence for women
discuss intellectual matters and to discuss while institutional/goal commitment was
career concerns. The institutional/goal more important for men. A further pre-
commitment scale, which appeared to liminary analysis in the present investiga-
combine items measuring institutional tion, indicated an additional sex x con-
commitment with those measuring goal tacts to resolve personal problems in-
commitment, was treated separately as a teraction. Frequency of faculty contacts
measure representing both academic and to discuss a personal problem was posi-
social integration.'

4Not surprisingly, frequency of informal contact general categories of social and academic integra-
with faculty was positively related to the quality and tion. The present assignment was based primarily on
impact of student-faculty relationships, as measured ajudgment of what the Spady-Tinto models mean by
by the informal relations with faculty and faculty academic and social integration and what the vari-
concern for teaching and student development ables in the study measure. For example, the infor-
scales. For men the median correlation between the mal relations with faculty scale was assigned to so-
six contact categories and the informal relations with cial integration because it also contained items con-
faculty scale was .21, and .14 between the six contact cerning the development of student-faculty
categories and the faculty concern for teaching and friendships and student satisfaction with their infor-
student development scale. For women the corre- mal relationships with faculty. On the other hand,
sponding median correlations were .18 and .10 re- the faculty concern for teaching and student devel-
spectively. opment scale focused more on teaching and
I Clearly an argument might be made for a some- classroom relationships and was thus assigned to
what different assignment of specific variables to the academic integration.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 202 PASCARELLA AND TERENZINI

tively (though non-significantly) related to controlled, in a series of preliminary dis-

persistence for women but negatively re- criminant analyses. Significant interac-
lated to persistence for men. Such evi- tions in any set were included in the final
dence suggests that the patterns of vari- discriminant analyses only if the entire set
ables influencing persistence/withdrawal of terms in the preliminary analyses was
behavior may differ by sex. Thus, data in associated with a significant increase in
the present study were analyzed sepa- the explained variance (canonical R2 or
rately by sex. RC2) in persistence/withdrawal decisions
Two-group discriminant function (Huberty, 1975).
analysis, which is essentially multiple re- Prior to any statistical analysis, the
gression analysis with a categorical de- overall sample of 763 subjects was ran-
pendent variable (Huberty, 1975), was the domly divided into two samples of ap-
major statistical tool employed. The vari- proximately 2/3 (n = 497) and 1/3 (n =
ables were entered in a setwise manner 266). The larger of these two was used as a
with pre-enrollment characteristics en- calibration sample for all statistical
tered first, academic integration variables analyses, while the smaller was held out
entered second, social integration vari- as a validation sample. The discriminant
ables entered third and the institutional/ function derived on the calibration sample
goal commitment scale entered last. was used to classify subsequent persisters
In order to test for the presence of ac- and withdrawals in the calibration group
centuating or compensatory interactions and was then applied to the raw data of the
among variables, two preliminary sets of validation sample in a similar, but inde-
cross-product, interaction terms were pendent, classification analysis (Huberty,
formed. The first set crossed the thirteen 1975). To achieve a conservative estimate
measures of social and academic integra- of correct classification, prior probabilities
tion (including institutional/goal commit- for all classification analyses were set at
ment as a measure of both) with six en- .50 for each group. Such a strategy would
tering characteristics: race, initial college tend to depress overall correct classifica-
of enrollment, combined SAT scores, tion when group sizes are quite different
highest degree expected, importance of (e.g., 11.8 percent and 88.2 percent in the
graduating from college and parents' present study), by eliminating much of the
combined formal education. The second correct classification due to extremely
set crossed each measure of social inte- high prior probabilities in one group.
gration with each measure of academic
integration. (Institutional/goal commit-
ment as a measure of both social and aca- RESULTS

demic integration was crossed with all

other social and academic integration Table 2 presents, by sex, the results of
variables.) the setwise discriminant analysis. As the
To investigate the particular influencetable indicates, the set of entering char-
of student-faculty relationship variables,acteristics failed to make a significant
the second set of cross-product terms alsocontribution to RC2 for either sex. Both the
included all other measures of social and academic and social integration sets made
academic integration crossed with each of significant contributions for each sex al-
the eight measures of student-faculty re- though the relative sizes of the unique
lationships (i.e., the six categories of contributions were reversed: academic
non-class contacts as measures of the integration (controlling for pre-enrollment
amount of student-faculty informal con- and social integration variables) = .101 for
tact, and the informal relations with fac-men and .049 for women; social integra-
ulty and faculty concern for teaching and tion (controlling for pre-enrollment and
student development scales as measures academic integration variables) = .060 for
men and .115 for women. Moreover, the
of the quality and impact of student-faculty
relationships). institutional/goal commitment scale made
Each set of interaction terms was then a significant RC2 contribution for men but
entered separately, with all main effects not women.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 203

Table 2. Discriminant Analysis Summary Based on Calibration Sample (N = 497)

Men (N = 270) Women (N = 227)

Change Change
in in
Variables Re2 Re2 df F Re2 Re2 df F
Main Effects
Pre-enrollment/Entering
Characteristics .036 .036 12/257 0.80 .056 .056 12/214 1.06
Academic Integration .244 .208 6/251 11.51** .183 .127 6/208 5.39**
Social Integration .304 .060 6/245 3.52* .298 .115 6/202 5.52**
Institutional/Goal Commitment .370 .066 1/244 '25.56** .308 .010 1/201 2.90
Interactions
(5 for men, 7 for women) .476 .106 5/239 9.67** .553 .245 7/194 15.19**

Note: Variables were entered in pre-determined order. Row 1 shows the effect of Pre-enrollment/Entering
Characteristics only; each subsequent row shows the effect of the indicated variable set after controlling for
the previous variable sets.
When Social Integration was entered before Academic Integration, the Academic Integration set made a
contribution to RC2 of .101 for men (p < .001) and .049 for women (p < .05).
Only those interaction effects which made a significant contribution in the preliminary analysis were included
here.
p < .05.
p < .001.

Table 3 presents the standardized dis- resents voluntary freshman withdrawal,

criminant weights for each main effects the negative weight may to some extent
variable. These discriminant weights are reflect the fact that, with all other factors
analogous to beta weights in multiple re- in the main effects model considered,
gression. The relative size and sign of the male students performing particularly well
weight indicates the magnitude and direc- academically are likely to transfer to other
tion of the contribution of a particular institutions generally regarded as more
variable to persistence/withdrawal deci- prestigious and academically demanding.
sions with the influence of all other vari- This may be a phenomenon particularly
ables taken into account. Generally the evident during the freshman year when
pattern of significant discriminant weights institutional or academic program com-
suggests that institutional/goal commit- mitments are less firmly embedded than in
ment and, in particular, range of relation- subsequent years. Such an explanation
ships with faculty (e.g. faculty concern for would seem consistent with both Feldman
teaching and student development, and and Newcomb's (1969) and Rootman's
informal contacts to discuss intellectual (1972) concept of lack of "person-role" fit
matters and campus issues) have a some- between the student and the normative
what stronger positive influence on per- climate of the institution as a determinant
sistence for men than for women.6 The of withdrawal. In this particular instance
quality of peer relationships, however, the academically successful student may
appears more important for women than regard the normative climate of his
for men. present institution as insufficiently chal-
At first glance the significant, negative lenging vis-a-vis his level of performance
discriminant weight for male grade-point and educational goals and, at the same
average (-.308) is somewhat surprising, time, perceive that academic environ-
given the general theoretical expectation. ments more congruent with his own dem-
However, as the dependent variable rep- onstrated capabilities and level of aca-
demic motivation are available
elsewhere.7
6 With ali pre-enrollment and all other measures of
social and academic integration controlled, the eight 7 Further indirect evidence that voluntary male
student-faculty relationship variables were associ- leavers were not "pushed out" by low academic
ated with an R,2 increase of .164 for men (p<.001),
marks is suggested by their significantly (p<.Ol)
which was over 50 percent greater then the RC2 in- higher freshman gradepoint average than non-
crease of .103 for women (p<.01). voluntary leavers.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
204 PASCARELLA AND TERENZINI

Table 3. Standardized Discriminant Weights and F Ratios for Unique Contributions to RC2 of All Main
Effects Variables (Based on Calibration Sample: N = 497)

Men (N = 270) Women (N = 227)

Standardized Standardized
Discriminant Discriminant
Variables Weighta Fb Weighta Fb

Pre-Enroliment/Entering Characteristics
Racial/ethnic origin - .079 0.48 -.137 0.86
Initial program of enrollment .091 0.62 .008 0.01
High school rank in class -.081 0.44 .126 0.68
Number of high school extracurricular
activitiesc .015 0.02 -.107 0.59
Expected informal contact with facultyc .034 0.11 .167 1.55
Parents' combined annual income .076 0.49 -.202 2.16
Parents' combined level of formal education .094 0.71 -.008 0.01
Academic aptitude .238 3.65 .249 2.23
Highest expected academic degree -.047 0.14 .043 0.08
Importance of graduating from colleged -.119 0.97 - .012 0.07
Rank of this university as a college choice -.027 0.05 .007 0.01
Confidence that choosing to attend
this university was the right choiced 130 1.21 .033 0.05
Academic Integration
Freshman grade point average -.308 5.22* -.053 0.09
Academic and intellectual development .150 1.42 -.198 1.58
Faculty concern for teaching and
student development .329 7.87** .266 3.49
Informal contacts with faculty to:C
1. Obtain information about courses and
academic programs -.107 0.77 -.314 5.16*
2. Discuss career concerns .067 0.28 .129 0.54
3. Discuss intellectual matters .408 10.05** .205 1.27
Social Integration
Extracurricular activitiese - .031 0.08 .205 2.08
Peer-group relationse - .209 2.87. .482 11.62**
Informal relations with faculty .304 6.45* .397 7.34**
Informal contacts with faculty to:C
1. Help resolve a personal probleme -.334 7.24** .102 0.55
2. Discuss a campus issue .282 4.43* .078 0.31
3. Socialize informally .093 0.48 .162 1.44
Institutional/Goal Commitmente .574 25.40** .254 3.03

a Persistence/withdrawal decisions scored 1 =persistenc

analysis, the size and sign of the discriminant weights represent the influemce of a particular variable with all
other variables in the model held constant.
b All other main effects variables held constant; degrees of freedom= 1/244 for men and 1/201 for women.
c As natural logarithms.
d Scored in reverse; l=highest or most positive, 4=lowest or least positive.
e Significant difference between men and women in the influence of this variable on persistence/withdrawal
decisions.
* < .05 **p < .01.

Similarly, the significant, negative dis- negative influence on male persistence.

criminant weights for contacts with fac- Indecisions about academic programs and
ulty to resolve a personal problem (-.334 majors may have a similar negative influ-
for men), and to obtain information about ence for women, although the meaning of
courses and academic programs (-.314 the negative sign in this instance would
for women) are somewhat surprising. On seem less clear.
further consideration, however, such Results of the preliminary discriminant
findings may have little to do with faculty analyses, conducted to determine if the
relationships. Rather, for men they may two sets of interaction terms made signifi-
merely reflect the presence of difficult cant contributions to the increase in R,2
personal problems and the fact that such with all main effects held constant,
problems are a particularly important yielded significant results in all cases. The

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 205

Rc2 increase for the total set of interac- .001 (recall that prior probabilities for the
tions between entering characteristics and classification analyses were set at .50 for
measures of social and academic integra- each group). Such results suggest that the
tion was .255 (from .370 to .625) for men functions based on the variables in the
and .328 (from .308 to .636) for women, analyses have both substantial dis-
both significant at p < .05. Similarly, the criminating power and reasonably high
corresponding R02 increase for the set of stability across samples (Huberty, 1975).
interactions between social and academic Table 4 provides the regression equa-
integration variables, and between social tions for plotting each of the twelve sig-
and academic integration variables and all nificant interactions. As Table 4 indicates,
student-faculty relationship variables was there were three significant interactions
.209 (from .370 to .579) for men and .348 between entering characteristics and
(from .308 to .656) for women, both sig-student-faculty relationship variables (in-
nificant at p < .05. Results of these teractions 1 for men, 6 and 7 for women),
analyses justified further investigation of The interactions involved the entering
individual interactions. To obtain the most characteristics: level of parents' formal
parsimonious model, only those interac- education (for both sexes), importance of
tions making significant unique contribu- graduating from college (for women) and
tions to the R 2 increase in each prclimi- the student-faculty relationship variables:
nary analysis (i.e., controlling for all main frequency of contacts to discuss career
effects and all other interactions) were concerns (men) and faculty concern for
selected for inclusion in the final discri- teaching and student development
minant analyses. (women). (Both faculty relationship vari-
Table 2 also shows the increases in R 2 ables were considered measures of aca-
associated with the addition of the demic integration.)
selected interaction terms to the main ef- As indicated by the signs of the unstan-
fects model. For men the final discrimi- dardized regression weights in Table 4,
nant model with all main effects and five there was a consistent compensatory pat-
interactions (which made significant tern for all three interactions.8 Spe-
unique contributions in the preliminary cifically, contacts with faculty to discuss
analyses) had an R 2 of .476 (df = 30/239, p career concerns had their most important
< .001) and an R 2 increase of .106 (df - positive influence on persistence for
5/239, p < .001) associated with the in- freshman men whose parents had rela-
teractions. For women the discriminant tively low levels of formal edcuation.
model with all main effects and seven in- Similarly, perceptions of faculty mem-
teraction had an R 2 of .553 (df = 32/194, pbers' concern for teaching and students
< .001) and an R 2 increase of .245 (df = were most important in positively in-
7/194, p < .001) associated with the in- fluencing the persistence of women who
teractions. initially attached relatively low levels of
Results of the classification analyses on importance to college graduation or who
the calibration sample indicated that 90.37 came from families where parents them-
percent of the men and 94.69 percent of selves had completed relatively little for-
the women who subsequently either with- mal education. (Note that pre-enrollment
drew voluntarily at the end of their importance of graduating was scored 1 =
freshman year or re-registered for their extremely important to 4 = not at all im-
sophomore year were correctly identified. portant.) As levels of parental education
When the function was applied to the raw and (for women) pre-enrollment impor-
data of the validation sample, the drop tancein of graduating increased, relation-
correct identification of subsequent per- ships with faculty had a decreasing influ-
sisters and withdrawers was 5.37 percent ence on students' persistence.
for men (correct classification - 85.0 per-
cent) and 6.63 percent for women (correct
8 The general nature of each interaction may be
classification = 88.06 percent). All four verified by substituting arbitrary high and low values
correct classification percentages were (e.g. 10 = high and 5 = low) for each variable and
significantly better than chance at p < plotting the four predicted points that result.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
206 PASCARELLA AND TERENZINI

Table 4. Regression Equations for Significant Interactions

Interaction Fa Equationb
Men
1. Level of parents' education (PE)
x Contacts to discuss issues re-
lated to future career (CFC) 4.13* yC= .016 (PE) + .150 (CFC) - .014 (PE x CFC)
2. Highest expected academic de-
gree (HD) xFreshman grade
point average (GPA) 5.39* y = -.114 (HD - .190 (GPA) + .042 (HD x GPA)
3. Initial program of enrollment
(IPE) x Peer-group relations
(PG) 13.57** y = .647 (IPE) + .004 (PG) - .025 (IPE x PG)
4. Institutional/goal commitment
(IGC) x Contacts to discuss in-
tellectual matters (CIM) 16.59** y = .039 (IGC) + .441 (CIM) - .015 (IGC x CIM)
5. Academic/Intellectual develop-
ment (AID) x Faculty concern
for teaching and student devel-
opment (FCTSD) 11.73** y = .057 (AID) + .091 (FCTSD) - .003 (AID x FCTSD)
Women
6. Level of parents education (PE)
x Faculty concern for teaching
and student development
(FCTSD) 9.34** y = .078 (PE) + .051 (FCTSD) - .005 (PE x FCTSD)
7. Importance of graduating from
college (IG)d x Faculty concern
for teaching and student devel-
opment (FCTSD) 9.35** y = .085 (IG)d + .051 (FCTSD) + .025 (IG x FCTSD)
8. Importance of graduating from
college (IG)d x Peer-group rela-
tions (PG) 12.28** y = .085 (IG)d + .128 (PG) - .018 (IG x PE)
9. Peer-group relations (PG) x
Contacts to discuss intellectual
matters (CIM) 7.87** y = .128 (PG) + .256 (CIM) - .009 (PG x CIM)
10. Institutional/Goal commitment
(IGC) x Peer-group relations
(PG) 20.13** y = .107 (IG) + .128 (PG) - .004 (IGC x PG)
11. Informal relations with faculty
(IF) x Faculty concern for
teaching and student develop-
ment (FCTSD) 9.28** y = .038 (IF) + .051 (FCTSD) - .002 (IF x FCTSD)
12. Informal relations with faculty
(IF) x Contacts to obtain infor-
mation about courses and aca-
demic programs (CCAP) 9.77** y = .038 (IF) - .218 (CCAP) + .015 (IF x CCAP)
a All main effects and all other interactions held constant; degrees of freedom = 1/239 for men
women.
h Weights for each equation are with all main effects and all other interactions statistically controlled;
constant = -1.79 for men and -3.75 for women.
I Predicted persistence/voluntary withdrawal decisions.
d Importance of graduating from college scored 1 = extremely important to 4 = not at all important.
* p < .05 ** p < .01.

A pattern somewhat different from the general, positively associated with vol-
compensatory relationships discussed untary withdrawal (as discussed above),
above emerges for the interaction (shown the association is most pronounced for
as 2 for men in Table 4) between highest those men with the lowest levels of aca-
expected academic degree at college en- demic degree aspiration at the time of
trance and freshman year academic per- college entrance. For freshman men with
formance (a measure of academic integra- relatively high pre-college degree aspira-
tion). The interaction effect suggests that, tions, level of academic performance
while male academic achievement is, in would appear to be somewhat less impor-

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 207

tant in leading to voluntary withdrawal. In considered measures of academic integra-

short, a freshman male may be more likely tion. As indicated by the regression
to withdraw voluntarily from this particu- weights, there was a consistent compens-
lar institution if he enters with relatively atory pattern across all three interactions.
low degree aspirations but performs par- Specifically, perceptions of faculty con-
ticularly well academically. Since demon- cern for teaching and students had their
strated academic success may foster strongest positive influence on persistence
higher levels of educational aspiration, for men with the lowest scores on the
such evidence of success may lead the academic/intellectual development scale.
student to believe that in enrolling in this Moreover, non-class contacts with faculty
particular institution he has significantly which focused on discussion of intellec-
underestimated his academic capabilities. tual matters were most important in posi-
To the extent that such a belief leads to tively influencing the persistence of men
the realization of incongruence with the with relatively low levels of institutional/
normative academic climate of the col- goal commitment and women with rela-
lege, the probability of withdrawal seems tively low ratings of the quality and impact
likely to increase. of their relations with peers. A conceptu-
The peer-group relations scale also ally similar, compensatory interaction was
showed significant interactions with en- also found for women (interaction 10)
tering characteristics. Interaction 3 be- between institutional/goal commitment
tween the peer-group relations scale and and scores on the peer-group relations
initial program of enrollment suggested scale.
that, with all other variables controlled, The compensatory nature of the above
peer-group involvement had a somewhat four interactions, and in particular peer-
stronger positive association with with- group relations (social integration) x
drawal for men enrolled in the liberal arts contacts with faculty to discuss intellec-
than for men in pre-professional pro- tual matters (academic integration), af-
grams.9 For women, the nature of the in- fords at least modest support for Tinto's
teraction between the peer-group relations (1975) hypothesis that high levels of aca-
scale and pre-enrollment importance of demic integration will tend to compensate
graduating from college (interaction 8) for low levels of social integration, and
was generally accentuating. That is, the vice versa. A generally similar pattern of
quality and impact of relations with peers compensatory interactions (11 and 12 in
were most important in positively in- Table 4) was found between student-
fluencing the persistence of women who, faculty relationship variables considered
at entrance, attached a relatively high as measures of social integration (i.e., in-
level of importance to college graduation. formal relations with faculty), and as mea-
As further shown in Table 4, sures of academic integration (i.e., faculty
there were three significant interactions concern for teaching and student devel-
between student-faculty relationship mea- opment, and non-class contacts for aca-
sures and other measures of social and demic advising). In terms of positive in-
academic integration (interactions 4 and 5 fluence on persistence, the quality of
for men and 9 for women). Both student- specific informal, non-class contacts with
faculty relationship variables involved faculty was most important for women
(i.e., contacts with faculty to discuss in- who perceived faculty in general as having
tellectual matters and faculty concern for little concern for effective classroom
teaching and student development) were teaching or students, and vice versa. The
quality of informal relations with faculty
9 Given theoretical expectations, this is a rather was also most important for women hav-
difficult interaction to explain. One possible, though ing frequent contacts with faculty for
highly tentative explanation, however, is that liberal course or program advice. However, as
arts men may be less clearly goal and career oriented
the latter variable had a significant, nega-
than men in pre-professional majors. Thus, for the
former, excessive peer-group involvement may have tive association with persistence, dis-
a greater potential for interfering with successful cussed above, this interaction might also
integration in other areas. be considered compensatory.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 208 PASCARELLA AND TERENZINI

IMPLICATIONS the student brings to college, a finding

generally consistent with earlier research
It is particularly important to stress the on voluntary withdrawals (Terenzini and
tentativeness of these findings, not only Pascarella, 1978).10 Thus, there may be
because of the limited nature of the sam- important determinants of freshman year
ple (i.e., single year, single institution) but persistence which are not merely the re-
also because of potential weaknesses in sult of the kinds of students enrolled, but
the measures employed to assess the di- rather are subject to the influence of in-
mensions of social and academic integra- stitutional policies and programs which
tion. This is a particularly relevant caution affect the student after he or she arrives
in the investigation of interaction effects on campus. This may be particularly true
in non-experimental data. Even under the if such programs and policies can posi-
relatively tighter controls of experiments tively influence the quality of relation-
or quasi-experiments, interactions have ships with faculty for men, and both fac-
been difficult to replicate (Snow, 1976). ulty relationships and peer relationships in
Thus, a replication of this study, or similar the case of women.
investigations on samples of students from Clearly such a finding can in no way be
different institutional settings would lend construed to mean that such factors as
additional credibility to the results. race, prior achievement, academic apti-
Given these limitations, however, the tude or educational aspirations do not sig-
findings would nevertheless suggest cer- nificantly influence success or persistence
tain implications for the theoretical mod- in college. A substantial body of research
els investigated and for the notion of per- exists to suggest otherwise (e.g. Cope and
sistence in college. First, it would seem Hannah, 1975; Pantages and Creedon,
that the findings firmly underscore 1978; Marsh, 1966; Spady, 1970; Tinto,
Spady's (1970) and Tinto's (1975) concep- 1975). Rather, the modest influence of
tion of the sociological complexity of the these variables in the present study may
influences on student persistence/ in large measure be attributable to the fact
withdrawal decisions. While measures of that the dependent variable was persis-
student entering characteristics and tence versus voluntary withdrawal. Had
freshman year social and academic inte- the dependent variable combined volun-
gration explained .370 and .308 of the tary and non-voluntary (academic) with-
variance in freshman male and female per- drawal, or focused only on academic fail-
sistence respectively, the addition of mul- ures, it seems likely that variables associ-
tiplicative terms, assessing the influence ated with educational attainment and aca-
of different dimensions of social and aca- demic achievement such as aspirations,
demic integration for different kinds of aptitude, racial/ethnic origin and prior
students, increased the explained variance academic achievement may have had a
.106 for men and .245 for women. Such substantially greater influence in the
evidence suggests that future investiga- model. (It also seems less likely that a
tions designed to increase understanding negative discriminant weight would have
of the proce-ss of student persistence/ been found for academic achievements.)
withdrawal decisions (rather than simply It is also important to note that the ex-
predicting who stays and who leaves) may tent of influence of various measures of
well need to take into account such in- social and academic integration was not
teractions among variables in order to independent of the particular background
capture what may be a particularly com- characteristics which students bring to
plex pattern of social-psychological re-
lationships.
"I As shown in Table 2, measures of social and
Second, the findings suggest that, in academic integration explained .334 of the variance
terms of main effects influence on persis-for men and .252 for women, versus .036 and .056
tence, what happens during the freshman respectively for entering characteristics. Further-
more, none of the twelve entering characteristics had
year may be more important than the par- significant discriminant weights for either men or
ticular commitments, background char- women when considered within the framework of the
acteristics, aspirations or aptitudes which main effects model.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 INTERACTION EFFECTS IN COLLEGE DROPOUT MODELS 209

college or of other social and academic have substantial levels of formal post-
experiences during the freshman year. secondary education.
Thus, even if particularly effective in- Similarly, high levels of academic inte-
stitutional programs or policies designed gration such as frequent informal contacts
to reduce student withdrawal could be with faculty focusing on intellectual mat-
established (e.g., increasing opportunities ters or perceptions of faculty as particu-
for student-faculty informal contact), it larly concerned about teaching and stu-
may well be that the positive benefits of dents appeared to compensate for low
those policies or programs will accrue levels of social and academic integration
differentially rather than generally. That in other areas. 11 For men such aspects of
is, rather than benefiting all students their relationships with faculty tended to
equally, the impacts of such programs on compensate for low levels of
student retention are likely to be mediated institutional/goal commitment and aca-
by the characteristics of the students in- demic and intellectual development. For
volved and the levels of social and aca- women, however, frequent contacts with
demic integration of those students in faculty focusing on intellectual issues
other areas. tended to compensate for low levels of
A third implication of the findings fo- satisfaction with the quality and impact of
cuses on this very notion of the varying peer relationships. This latter finding, in
influence of different dimensions of social particular provides reasonably clear sup-
and academic integration for different port for Tinto's hypothesis of a potentially
kinds of students. Particularly noteworthy compensatory association between social
in this regard was the differential pattern and academic integration.
of influences associated with measures of
student-faculty relationships. Various
REFERENCES
measures of the frequency and quality of
student-faculty relationships made signifi-
Cope, R. and W. Hannah
cant contributions to the prediction of
1975 "Revolving college doors: The causes and
male and female persistence with the in- consequences of dropping out, stopping out
fluence of twelve entering characteristics and transferring." New York: Wiley.
and five other measures of social and aca- Eckland, B.
1964 "College dropouts who come back." Har-
demic integration held constant. (This
vard Educational Review 34:402-20.
finding is generally consistent with those Gekoski, N. and S. Schwartz
of Pascarella and Terenzini, 1977 and 1961 "Student mortality and related factors."
Terenzini and Pascarella, 1978). Extent of Journal of Educational Research 54:192-
194.
influence in any one dimension of
Feldman, K. and T. Newcomb
student-faculty relationships, however,
1969 "The impact of college on students." San
generally depended upon student back- Francisco: Jossey-Bass.
ground characteristics, commitment to Huberty, C.
graduation and level of integration in other 1975 "Discriminant Analysis." Review of Edu-
cational Research 45:543-98.
areas.
Iffert, R.
The nature of these interaction effects 1958 "Retention and withdrawal of college stu-
suggests that such measures of student- dents." Washington, D.C.: U.S. Depart-
faculty relationships as frequency of in- ment of Health, Education and Walfare,
formal contacts to discuss career concerns Bulletin No. 1.

or perceptions of degree of faculty con-

cern for teaching and students function in
" This paper has stressed the compensatory side
a compensatory manner in terms of their
of the interactions between measures of student-
influence on freshman persistence. That faculty relationships and other student char-
is, they may provide interpersonal links acteristics. It may be important, however, to ac-
with important adults in the institution knowledge the reverse interpretation. That is, as
which tend to compensate for the influ- levels of parental education, entering commitment to
graduation and extent of social and academic inte-
ence of an initially low commitment to the
gration in other areas increased, the influence of a
goal of graduation or the relative absence student's relationships with faculty on his or her
of parental role models who themselves decisions to. persist or withdraw diminished.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms
 210 PASCARELLA AND TERENZINI

Marsh, L. ress report." Pp. 50-105 in L. Shulman,

1966 "College dropouts-A Review." Personnel (ed.) Review of Research in Education, 4.
and Guidance Journal 44:475-81. Itasca, Ill.: Peacock..
Newcomb, T., D. Brown, J. Kulik, D. Reimer, and Spady, W.
W. Revelle 1970 "Dropouts from higher education: An
1970 "Self-selection and change." Pp. 137-60 in interdisciplinary review and synthesis."
J. Gaff (ed.), The Cluster College. San Interchange 1: 109-21.
Francisco: Jossey-Bass. Spady, W.
1971 "Dropouts from higher education: Toward
Pantages, T. and C. Creedon
an empirical model." Interchange 2:38-62.
1978 "Studies of college attrition: 1950-1975."
Terenzini, P. and E. Pascarella
Review of Educational Research 48:49-
1978 "The relation of students' pre-college char-
101.
acteristics and freshman year experience to
Pascarella, E. and P. Terenzini voluntary attrition." Research in Higher
1976 "Informal interaction with faculty and Education 9:347-66.
freshman ratings of the academic and non- Tinto, V.
academic experience of college." Journal of 1975 "Dropout from higher education: A
Educational Research 70:35-41. theoretical synthesis of recent research."
1977 "Patterns of student-faculty informal in- Review of Educational Research 45:89-
teraction beyond the classroom and volun- 125.
tary freshman attrition." Journal of Higher Walberg, H. and S. Rasher
Education 48:540-52. 1976 "Improving regression models." Journal of
1979 "Predicting freshman persistence and vol- Educational Statistics 1:253-77.
untary dropout decisions from a theoretical Wilson, R., J. Gaff, E. Dienst, L. Wood, and J.
model." Journal of Higher Education, in Bavry
press.
1975 "College professors and their impact on
Rootman, I. students." New York: Wiley.
1972 "Voluntary withdrawal from a total adult Wilson, R., L. Wood, and J. Gaff
socializing organization: A model." Sociol- 1974 "Social-psychological accessibility and
ogy of Education 45:258-70. faculty-student interaction beyond the
Snow, R. classroom." Sociology of Education 47:
1976 "Research on aptitude for learning: A prog- 74-92.

MANUSCRIPTS FOR THE

ASA ROSE SOCIOLOGY SERIES

Manuscripts ( 100 to 300 typed pages) are solicited for publication in

the ASA Arnold and Caroline Rose Monograph Series. The Series
welcomes a variety of types of sociological work-qualitative or quan-
titative empirical studies, and theoretical or methodological treatises.
An author should submit three copies of a manuscript for consideration
to the Series Editor, Professor Robin M. Williams, Jr., Department of
Sociology, Cornell University, Ithaca, New York 14853.

This content downloaded from

200.130.19.162 on Sat, 02 Sep 2023 18:33:27 +00:00
All use subject to https://about.jstor.org/terms