Chapter Two
Students Educating Students:
The Emerging Role of Peer Effects in Higher
Education
George Goethals, Gordon Winston, and David Zimmerman
G
oethals, Winston, and Zimmerman investigate the implicit assumption that students learn better in the company of stronger students than with weaker ones. If this widely accepted
assumption is valid, then students themselves are the only suppliers of an important input to
educational production and, likewise, competition among colleges and universities for top students may
be rational and justifiable. The authors review the literature on peer effects, and describe their methodological approach to investigating the existence and strength of peer effects on learning.
“The real intellectual life of a body of undergraduates, if there be any, manifests itself, not in the
classroom, but in what they do and talk of and set before themselves as their favorite objects between
classes and lectures. You will see the true life of a college … where youths get together and let themselves go upon their favorite themes—in the effect their studies have upon them when no compulsion
of any kind is on them, and they are not thinking to be called to a reckoning of what they know.”
Woodrow Wilson
I. Introduction
1
of able students than with weak ones. This suggests
The proposition is simple: the quality of the edu-
a primary reason colleges and universities, and
cation a student gets at a college or university
potential students, care so about the quality of an
depends both on the institution's resources—faculty,
institution's students—why selectivity looms so
facilities, libraries—and importantly on the quality of
large in quality rankings like U.S. News and World
his or her fellow students. The student simply learns
Report's and why student quality is such a deadly
more—better, faster, more deeply—in the company
serious business to colleges and universities.
25
Put that way, the proposition seems reason-
legally turn students away do create enclave hon-
able, persuasive, and appealing—we can usually
ors colleges or quality differentiated campuses
get by simply by asserting it. But as we've looked
where they can). Peer effects also help explain why
more closely at those peer effects, we have encoun-
competition for student quality is driving an increas-
tered an increasingly complicated, subtle, and often
ingly fierce competition among the most selective
slippery set of issues: at base, not much is known
colleges and universities.
about peer effects in higher education, despite their
The long debate in K-12 policy circles on the
potential importance. The purpose of this paper is, in
merits of sorting students by their abilities often rests
a sense, to describe the structure of our ignorance—
on implicit assumptions about the symmetry of peer
what it looks like, why it matters, and how it might be
effects—whether weak students pull strong students
overcome—a research agenda. So this is very much
down more than strong students lift weak students
a work in progress, but we have come to feel that it's
up. Proponents of sorting assume that mainstream-
essential to frame the questions clearly at the outset.
ing will do harm to able students that can't be offset
To that end, the next section asks why peer
by gains to the less able. Opponents assume that
effects matter. Section III provides a brief review of
gains to the less able will outweigh any losses to the
relevant prior research on peer effects. Section IV
more able. If hard evidence could be found to show
specifies the key questions researchers must con-
that one or the other of these is most usual, even
front, along with a discussion of possible mecha-
from the college level, it would help to focus that dis-
nisms for the transmission of peer effects. Section V
cussion.
lays out our proposed research strategy. Section VI
If peer effects work, and if they start to work in
offers some conclusions.
the early grades, the strong arguments for early
intervention with disadvantaged children are made
II. Why Do Peer Effects Matter?
stronger since early improvements will not only
Peer effects appear to be central to the way
raise the performance of a child directly but,
educational services are produced and, through
through the effect of that improvement on his
that, to the structure of the “firms” and “markets”
peers, have a multiplier effect over time. Dollars
that make up higher education. Most specifically,
spent early will have a larger payoff than dollars
peer effects help explain why selectivity among
spent late, other things being equal.
applicants is considered essential to educational
In the long debate about the role of increased
quality and why the firms in this industry sacrifice
resources in improving education, it has always
significant revenues by massively turning away
been difficult to separate neighborhood (peer)
customers, hence revenues, to maintain student
effects from the effects of resources in K-12
quality (and why the public universities that can't
schools—neighborhoods well endowed with able
26
peers are usually well endowed with resources, too.2
on educational outcomes. A different role for peer
Colleges and universities, in contrast, bring together
effects has emerged from the debate on public affir-
students from varied backgrounds to share the
mative action policy where the argument has been
same resource levels, so they should allow that sep-
made that since it is socially/morally/politically unac-
aration between peers and resources to be made
ceptable to let minority representation fall significant-
more effectively. Public educational policy will be
ly in our best public colleges and universities—and
better informed with a clearer identification of those
no proxy has appeared to replace the explicit con-
two forces.
sideration of race in admissions—it will be neces-
Much is being made of the threat to conven-
sary to set lower standards for all students in order
tional higher education from high-tech (and often for-
to achieve acceptable levels of minority representa-
profit) competition—Peter Drucker has famously
tion. That, if peer effects are important, will under-
predicted the imminent end of the university as we
mine the quality of public education in a way that
have known it. Yet if peer effects are both important
affirmative action never did.4
and difficult to generate through electronic media,
The most basic puzzle in all of this may be not
there will be severe limits on the kind and quality of
about peer effects, per se, but more fundamentally
education those new information technologies can
about why everyone cares so about an institution's
replace. The distinction between training and educa-
student quality—college administrations and boards,
tion may become increasingly central.
alumni, students, parents, U.S. News, etc. That con-
Affirmative action appears to be sensitive to
cern supports an economic anomaly of major pro-
peer effects in two ways. It's not a big step from the
portions: colleges restrict their supply (limit
destructive stereotype anxiety that psychologist
enrollments) in the face of strong and persistent
Claude Steele of Stanford University identified as
demand by admitting only those applicants of high-
reducing the academic confidence and competence
est quality, a policy that clearly costs the institution
of black students to the role of peer expectations and
tuition revenues and (on a more idealistic plane) the
values in triggering or suppressing that anxiety. 3 The
opportunity to serve more students. Take Williams
importance of a comfort zone in which minority stu-
College as a convenient example. If we accepted all
dents encounter positive peer effects is evident in
of the 4,500 students who apply for our 500 seats in
the Andrew W. Mellon Foundation's College and
the freshman class each year, our current yield rate
Beyond Data in Historically Black Colleges and
of around 50 percent would mean that 2,250 of them
Universities. Parallels with the more general effects
would come. After four years of doing this at an aver-
of institutional ethos, of course, suggests that these
age net tuition of $23,350, we'd have over $40 mil-
peer effects, while they may be supportive of the stu-
lion in additional revenues each year. Ignoring the
dent, can have either a positive or negative impact
fact that we'd also have a lot more in cost, what's
27
going on? And, while this is anomaly is clearest at
meet and get to know fellow students who will some-
the best institutions, very few colleges or universities
times be helpful in later life—and good students are
have truly open admissions, and that number is
more likely to achieve greater success and hence be
shrinking.
of greater help. So colleges and universities sell an
Peer effects offer one explanation for such
investment not just in “what you know,” but also in
strange behavior in seeing a college's student-cus-
“who-you’ll-know-later,” and that investment may
tomers as the only potential suppliers of an impor-
sensibly be recognized by both students and institu-
tant input to educational production and one that
tions as important. (In days past, marriage contacts
varies markedly among individuals. If peer effects
would have been a significant submarket in this kind
exist, institutions care about their student quality
of investment.)
because of their peculiar customer-input production
But what seems most often to be presented as
technology. 5 And if peer effects exist, students care
an alternative to peer effects on learning is some sort
about their fellow students for the same reason—
of status association—that colleges provide a stu-
because they are both an indicator of and contribu-
dent with utility-enhancing association with the rich
tor to high quality education.
and famous (or bright and nerdy). Higher education
But other explanations are often presented as
institutions are clubs formed to establish restrictive
alternatives to peer effects—though frequently by
association or, alternatively, they are restaurants,
way of citing education as an example of something
part of whose appeal comes from association with
like status associations. Still, it is useful to look more
fellow diners. There is no investment component in
closely at these alternatives.
this, just the pleasure of cozying up to others who
themselves have social status. More selective col-
Some are easy. Alumni care about the quality of
leges provide more impressive fellow-customers,
students at the Old Alma Mater (more generally, the
hence more of that kind of satisfaction.
quality of the institution) because its reputation
redounds to them—by implication, the better the cur-
A complementary but different source of utility
rent students, the better the alumni must have been
attaches simply to winning. Selective colleges are,
in their day to pass through those same portals. Ex
by definition, hard to get into, so being accepted is a
post distinction. Institutions, on the other hand, care
source of achievement—like Edmond Hillary's
about their student quality for straightforward mar-
explanation for wanting to climb Mount Everest,
keting reasons—good students provide a visible
“Because it's there.”
endorsement effect that implies institutional quality:
At the other end of a continuum of social nobili-
“Those who have choices choose us.”
ty would appear to be a college's idealistic objective
More seriously, the students of an institution
of offering their excellent education to those who'll
may provide significant networking effects. They will
use it to best contribute to society. Colleges and uni-
28
versities see themselves as educating future leaders,
• that it's probably useful, in service of clarity, to
so they admit those applicants whose prospects of
stick to common usage and let peer effects
becoming influential leaders are the greatest. This is
describe only students’ effects on other stu-
6
often explicit in the wealthiest institutions.
dents' learning
Economically, these alternatives have to do with
• that in light of the very high costs colleges and
(a) increasing demand (marketing), (b) providing util-
universities incur—in foregone “sales volume”
ity directly (status or winning), (c) producing a differ-
and tuition revenues, if nothing else—through
ent, non-learning, product (networking), or (d)
service of student quality, it remains unclear
serving idealistic social objectives. These alterna-
why most of these things would be of suffi-
tives are very different from one another and from
cient importance to justify those costs to the
the peer effects that simply increase the quality of
institution, absent peer effects on learning
educational services and learning that a college or
• that the customer-input technology of higher
university produces. And it's clear that the four alter-
education does not appear to be very com-
natives are not mutually exclusive. A car, for a famil-
mon in other production processes—certainly
iar example, can provide both transportation and
not as common as the status and association
status, and the status component will be much influ-
effects that have received recent attention.
enced by who else owns that kind of car. But your
So we're back to “Do peer effects really exist?”
Mercedes isn't any safer, nor will it stop shorter or
It's important whether other satisfactions accrue to
hold the road better, if other Mercedes owners are
alumni or students or parents from the selectivity of
rich or famous or klutzes or Grand Prix drivers.
colleges and universities. It is also important whether
The implications of all of this for our examina-
there's an underlying educational rationale—affecting
tion of peer effects would appear to be:
learning—for that selectivity. We have to concede
• that these alternatives are very different, even
that we hope the answer is yes. Like most educators,
if all help us understand why people and
we'd like to find that peer effects on learning—a col-
administrations may care about an institu-
lege's primary mission—are real rather than face the
tion's student quality
idea that all that attention to student quality is driven
• that all of them are probably at work in higher
by less noble motives. In any event, it seems impor-
education, especially at the top of the hierar-
tant to know whether the magnitude of any educa-
chy—at Harvard and Swarthmore and Duke
tional benefits associated with peer effects warrant
the resources expended to achieve them.
• that because the peer effects alternatives are
III. What Is the Evidence So Far?
not mutually exclusive, persuasive evidence
We're not starting completely from scratch, so
that one exists doesn't suggest the absence
it's useful to briefly sketch out what others have
of another
29
found out about peer effects—in higher education,
duction function provides a useful way of organizing
K-12, in psychology labs, and in other social sci-
a diverse literature. Formally, the education produc-
ences. Indeed, one of the interesting aspects of
tion function might be specified as follows:
studying peer effects is that it has attracted the atten-
Educational outputs = f(inputs)
tion of researchers from such a wide variety of disciplinary
perspectives—usually
with
indicating that educational outputs (grades, atti-
different
tudes, achievement, retention, etc.) are a function of
motivating questions, but all linked by the important
the educational inputs. A useful way of distinguishing
role ascribed to peer effects.
between the array of inputs is to separate them into
Certainly the most influential piece of social sci-
those associated with student's prior performance,
ence research incorporating peer effects is the
family background characteristics, peer (or neigh-
famous study, Equality of Educational Opportunity,
borhood) characteristics, and other institutional char-
completed more than 30 years ago.7 This study,
acteristics. These four inputs, combined with
known more commonly as the “Coleman Report,”
different outputs, can be used to characterize much
popularized the use of education production func-
of the social science literature on peer effects. It is
tions—equations relating schooling inputs (e.g., fam-
worth noting that the inclusion of a measure of prior
ily background, peers, student-teacher ratios,
performance allows for the use of a value added
spending per student) to schooling outputs (e.g.,
measure of output. That is, we can see whether a
grades, retention). Including more than a half million
change in some input is associated with a change in
students from approximately 3,000 elementary and
some output.
secondary schools, Coleman and his associates
Much of the literature contributed by econo-
sought to measure the features of school environ-
mists focuses on the impact school spending has on
ment that led to differences in student attainment. A
either grades or wages. 9 To disentangle the effects
key finding of this study was that “…a pupil's
spending might have on student performance, it is
achievement is strongly related to the educational
necessary to control for other variables—such as
backgrounds and aspirations of the other students in
the quality of the peer environment—that are likely
the school.” Indeed, peer characteristics were found
to be correlated with spending. Peer effects are, for
to be notably more important than teacher charac-
this task, simply a nuisance that must be statistical-
teristics or non-social aspects of the school. This
ly controlled to enable the researcher to accomplish
report spawned a virtual cottage industry of
their chosen objective of measuring the benefits of
researchers attempting to pin down the parameters
additional spending. Typically, a measure of a
of the education production function.8 While there
school's average student quality—usually average
continues to be great controversy over the determi-
SAT scores—is included in wage or grade equa-
nants of student achievement, the educational pro-
tions and usually has a significant and positive coef-
30
ficient. 10 Indeed, it often looks as though
their own abilities, students were best off if they were
resources—spending per student—may work pri-
in the top group of a school that sorted by ability and
marily through its effect on attracting student quality
worst off in the bottom group of such a school.
and allowing greater selectivity. In all, the evidence
The economics literature has also considered
of peer effects is positive, but not strong. It should
an important methodological issue that is perva-
be noted that the use of a peer effect argument in
sive in all research on peer effects; people often
most modern educational production functions
select with whom they associate. This contrasts
shows, at least, economists prior belief in the poten-
sharply with an experimental situation in which we
tial importance of peers effects.
might randomly assign people to differing peer
Where economists have used peer effect evi-
environments and then measure their effect on
11
dence explicitly, they have drawn on a small K-12
educational attainment. If the peers with whom
literature, especially a study of approximately 7,000
one associates are linked with attributes of the
Montreal students between the first and third grades.
person that also affects their attainment (and
12
This study found compelling evidence that peer
which are unobservable to the researcher) then we
effects were both important and nonlinear. That is,
might falsely attribute a peer effect where one
student performance rose with the average class-
does not exist. For example, suppose people with
room IQ score. The increase, however, slowed as
low ability friends tend to do worse in school.
the mean IQ rose. It should be noted that their model
Perhaps they would have done poorly even if they
did not allow the measured peer effect to vary with
didn't associate with such people. That is, what
the students initial ability level.
might at first look like a peer effect might really be
A recent K-12 study used the impressive NCDS
a case of birds of a feather flocking together. At
British data that follows the entire cohort of children
least two studies by economists have looked at the
born in a particular week in Britain in 1958. This
issue of such selection bias in the context of peer
study related the children's standardized math and
effects. Evans et al. 14 studied peer effects in the
reading scores—taken at the ages of 7 and 11—to
context of teen pregnancy and school dropout
measures of their parents and schooling inputs.13
behavior. They use the econometric technique of
Peer effects were captured both by the varying
instrumental variables to attempt to control for
socioeconomic background of the student's peers,
such bias, finding that peer effects disappear once
along with the “streaming” of students by ability with-
such concerns are incorporated into the empirical
in schools. They found clear evidence that peer
model. Steven Rivkin1 5 questions these results and
effects were positive and their data suggested, too,
shows that the results are sensitive to the type of
that they were nonlinear—that poor students were
instrumental variable used. These papers suggest
helped more than strong students were hurt. Given
the importance of taking the selection issue seri-
31
ously. They also suggest the value of a good
the ecological approach20—have also studied the
experimental or quasi-experimental approach to
impact peers have on grades. Almost all of these
the measurement of peer effects—something we
studies focus on the elementary and secondary
pursue below.
school level.2 1 A relevant strand of this issue stud-
While it is not possible in this paper to survey
ies the factors that influence a person's vulnerabil-
the study of peer effects from all disciplinary per-
ity to influence. In addition, there is an interesting
spectives, it is probably worthwhile to summarize
debate in the area of child and adolescent devel-
briefly the various strands from the diverse litera-
opment over the relative importance of the family
tures that have considered such effects.
versus peers in a person’s development. Indeed,
in her recent book, The Nurture Assumption,
Sociologists have spent considerable time
Judith Harris argues that peers are much more
studying neighborhood effects—particularly in the
important than parents in human development.2 2 In
context of urban poverty. The central issue in this lit-
sum, the psychological literature affirms the exis-
erature is whether and how proximity to concentrat-
tence and importance of peer effects for elemen-
ed poverty increases the odds that an individual will
tary and high school students. Again, selection
be or remain poor. Jencks and Meyer 16 present a
issues seem problematic in the nonexperimental
useful taxonomy of models including contagion the-
empirical studies. At the college level, Pascarella
ories, theories of collective socialization, competition
and Terenzini 23 summarize the effect of peers on a
theories, and relative deprivation theories (see also
variety of outcomes including attitudes, values,
Crane, 1991; Rosenbaum, 1993).17 Writings by
and educational attainment.
William Julius Wilson combines elements of these
Finally, it is worth noting that educational
approaches in his influential book The Truly
researchers from various disciplines have con-
Disadvantaged.1 8
sidered the benefits of “peer assisted learning.”
Both psychologists and sociologists (and
These studies show the benefits of group ver-
economists) have studied the impact peers have
sus solitary learning. In as much as they are
on adolescent substance abuse and teen preg-
interested in designing “optimal learning envi-
nancy. Here the literature commonly involves the
ronments,” the effects of peers is important to
definition of the relevant network of friends and
their analyses.2 4
linking the behavior of the friends to the individual
under study. 1 9 Deleterious effects of substance
IV. Do Peer Effects Exist in
Higher Education?
abusing friends is often found, but the problem of
selection bias seems particularly troublesome in
Given the discussion above, it seems that there
this area.
are several recurring questions that are central to
Developmental psychologists—often following
32
understanding peer effects:
have to be similar in some meaningful sense
• Do peer effects exist in higher education?
in order to influence each other? Would a stu-
They might be the product of wishful thinking
dent who is significantly below the peer aver-
at elite and selective institutions trying to con-
age do better or worse than expected? What
vert their snob appeal and their prestige
other factors will influence the magnitude of
race—with the promise of association with the
the peer effect?
rich and famous—into something of genuine
• Who are the peers? It's necessary to limit
educational and social value. But they could
these questions to a sub-domain of social
be very real. Right now, peer effects seem to
psychology—to exclude parents and profes-
be a litmus for optimists—or romantics—
sors and other employees of the college or
since reactions to the plausibility of a genuine
university. Where might peer influences be
effect on learning is either taken as obvious or
most easily detected? On the team? In the
dismissed with haste and often derision as
dining hall? In the dorm room or the debate
hopelessly idealistic. So, examining their exis-
team or…? Which of the possible sources are
tence has to be the first order of business,
most important?
and controlling for potentially confounding
• Are peers individuals or groups, or is a broad-
variables—particularly given the self selection
er institutional ethos more influential? And
issues discussed above—is the central
how much can the institution intentionally
empirical challenge.
shape, or change, that ethos?
• Are peer effects nonlinear? That is, do the
• Do peer effects work with equal force for bet-
benefits of improving the peer environment
ter (good behavior and academic perform-
diminish at some point? The answer to this
ance) or for worse (binge drinking, drugs, and
question is at the heart of a utilitarian justifica-
the lost opportunities of a bad crowd)?
tion of mixing students of different ability lev-
• Do peer effects need physical proximity or
els. Would the aggregate of learning in higher
can they function just as effectively through
education be increased if students were ran-
cyberspace?
domly assigned to different schools? Would
• Do peer effects endure or do they dissipate
the loss in learning a high ability student suf-
rapidly over time?
fers by being moved to a lower average peer
setting be offset by the gains a weaker student garners in a better peer setting? In what
A great deal of evidence from studies in social
way are peer effects dependent upon a stu-
psychology addresses these issues and suggests
dent's ability level?
that peers do in fact exert tremendous influence on
college students in numerous ways. We’ll review this
• Is there an intimidation effect? Do students
33
literature briefly and suggest its relevance to policy
when schemas do change, or undergo accommo-
questions in higher education. An economic frame-
dation, during development. They become more
work for understanding peer effects is found in the
complex or they are replaced by new or highly
appendix.
revised schemas. Then the person views the world
differently. This process takes place very rapidly during cognitive development in childhood, but it contin-
Talk and Its Cognitive Consequences
One of the things that people do is talk.
ues throughout the life span. It is this process that
Certainly the talking and discussing that college stu-
professors strive to affect. Get students to think in
different ways. Get them to pay attention to different
dents do has important effects. Some of these
effects are directly relevant to the institution’s goal of
things. Peers who can provide new information or
graduating well-educated students, students who
ideas that challenge existing schemas, or who can
directly model different ways of thinking or problem
are knowledgeable, savvy, thoughtful, and wise. Talk
among peers can help students develop these qual-
solving can also have the effect of changing our con-
ities in a number of ways. First, students can trans-
ceptions or mental models of the world and its workings.
mit information. They can explain things to one
another—facts, concepts, perspectives, techniques,
In the first instance then, peers provide us
strategies for learning. Second, peers can affect the
with new information to expand our data base, or
ways students think and therefore the way they
the way we organize and deploy that data base to
acquire and process new information. Jean Piaget
further understand the world. Psychologists make
discussed the development of cognitive schemas, or
the important distinction between crystallized and
knowledge structures, in childhood. Schemas can
fluid intelligence; simply put, what we know from
be thought of as our general knowledge about the
absorbing and processing information vs. the men-
physical world, events, people, groups, activities,
tal horsepower we use to make sense of new data
etc. Often new information goes through a process
and plan adaptive behavior. Talk is one of the
of assimilation to schemas. That is, the information
activities that can enhance both. While there are
is interpreted in terms of existing schemas, stored in
most likely limits on the levels of both crystallized
familiar categories, and simply fit into comfortable
and fluid intelligence that people reach, input from
and familiar ways of looking at the world.
books, mass media, life experiences, professors,
Sometimes, however, the data don’t fit the existing
and peers can increase those levels.
25
schemas, and a schema actually changes to better
Talk is a two-way street. It has more effect than
fit the data. This is the process of accommodation.
merely providing input to increase knowledge or
Academics often joke that if the data don’t fit the the-
change modes of thinking. When people talk rather
ory, so much the worse for the data. The schemas
than listen, they have to switch their cognitive tun-
are kept intact. However, exciting things happen
34
ing.26 They move from the reception mode, where
tionship level communication concerns the people
they simply take information in, to the transmission
talking—what each one thinks of himself or herself,
mode, where they have to organize their thoughts
what he or she thinks of the other person, and what
into a coherent story that makes sense to others.
he or she thinks of their relationship. Most often com-
Doing so requires lots of cognitive work, sometimes
munication at the content level is done in words.
at the expense of remembering and transmitting
Communication at the relationship level is done
everything in detail. However, this teaching function
nonverbally, through gesture, tone of voice, facial
generates a great deal of learning and understand-
expression, etc. In his book on presidential charac-
ing—in the teacher! So, there is benefit from being
ter, James David Barber 2 8 notes that any presiden-
on either side of the talking table. It helps to hear
tial decision is the story of a rational man calculating
what others have to say, and to absorb or to
and an emotional man feeling. Conversation simi-
process and organize what one hears, and it helps
larly has its rational and emotional elements. The
to speak to others, and organize our own knowl-
relational and emotional aspects make talk much
edge and ideas in ways that communicate effec-
more complicated and introduce a range of consid-
tively.
erations that affect how the process of peer educa-
Can all of this be done by letter, telephone, or e-
tion unfolds. These relational issues can greatly
mail? Perhaps some of it can. But the richness of
enhance, but also reduce, the educational impact of
human communication lies in more than the words
peers.
exchanged. Face-to-face interaction conveys a
Interaction
and
social
comparison.
wealth of nonverbal information that qualifies and
Psychologists have noted that a great deal of self-
amplifies the spoken word. It offers a relational and
evaluation takes place through social comparison.29
emotional context for talk that provides the potential
We evaluate our opinions and abilities by comparing
for greatly enhanced learning. This learning won’t
them with those of other people. This process is not
always be fun and easy. But much can be gained
always objective. We want to find out whether our
from it.
own opinions are correct and our abilities are good,
but we also want to find out whether the opinions of
our peers are indeed correct and good. When stu-
Social Interaction and Its Personal
Consequences
dents discuss an important issue with their peers,
Students of human communication argue that
there are complex comparison and influence
communication always takes place on two levels
processes that push toward consensus within the
simultaneously: the content level and the relation-
group, either through opinion change or the rejection
ship level. At the content level communication con-
from the group of people with deviant views. There
cerns the world at large—ideas, events, people,
are also likely to be implicit ability evaluations. Am I
objects, tasks, problems, and values.27 At the rela-
smarter or morally superior to my peers? These abil-
35
ity evaluation processes sometimes produce com-
processes that take place in conversation.
petition, the formation of coalitions, or attempts to
Group polarization effects. When people
undermine other people’s performance, but general-
engage in the social comparison of values, the
ly serve to improve one’s own performance. But
expression of a group’s values often becomes more
there may also be rejection of people with discrepant
extreme. Colleges and universities need to be aware
ability levels, just as there is rejection of those with
of the values that various campus groups adopt and
deviant opinions.
accentuate.
One specific and highly important consequence
When people compare opinions and judgments
of the social comparison of abilities is the setting of
that are related to important values, the comparison
levels of aspiration. In general, people are strongly
process resembles ability comparison more than
motivated to perform as well as possible and to
ordinary opinion comparison. When important val-
develop abilities as fully as possible. But when peo-
ues are involved, people compare how well they
ple cease comparing with others, they make impor-
exemplify or support those values, and doing that
tant adjustments to their levels of aspiration. When
well may be as important as having a high level of an
they cease comparing with others who are less able,
important ability. Thus, rather than reaching some
their level of aspiration rises. When they cease com-
kind of middling consensus or compromise, as often
paring with others who are more able, their level of
happens with ordinary opinion comparison, people
aspiration drops. Levels of aspiration have important
compete to support the value at least as much, or a
impacts on academic engagement and perform-
little bit more, than their peers, and thus the group
ance. We will return to this issue below.
polarizes, and expresses the value more extremely.
What do these processes imply for peer educa-
One common example is a group making judgments
tion? Interacting people will engage in frequent dis-
about the appropriate level of risk. Often the group
cussions in sorting out their views on important
will make a risky shift and take action that is riskier
issues. They will try to clarify and influence one
than the course of action proposed by the average
another's values in an effort to reach consensus.
individual. In this way important values get highlight-
These influence processes may serve the goals of
ed or exaggerated in the group.
either self-evaluation or self-validation, but one way
In the college setting, if the value that is exag-
or another they will provide a spur to discussion, the
gerated in this way is one that a college or universi-
clarifying of position, the exchange of information,
ty wants to support and emphasize, a happy result
the setting of priorities, etc. These overtly content
has been achieved for the institution. But if the polar-
level communications are energized by concerns
ized value is counter to the overarching educational
with self and relationship. Cognitive development,
values and goals of the institution, there’s trouble.
the enhancement of both crystallized and fluid intel-
Again, colleges and universities need to be alive to
ligence, can be motivated by the comparison
36
the groups that form, the values that guide them, and
models who seem attainable and admirable, and in
how those values can become polarized, for good or
fact behave in ways that enhance intellectual devel-
bad.
opment. Sometimes students do not choose the role
Self-definition and identification. In our society,
models college administrators would choose for
college is often a time for identity seeking and self-
them. There are a variety of reasons for this. One is
definition. People attempt to construct an identity in
that their estimation of their chances of successfully
numerous ways. They try out a variety of behaviors
identifying is low. Another is that peer group values
and roles until they find a self-defining set that fits
do not make emulating them very attractive. We will
both their deep sense of self and other people’s per-
return to this problem below.
ceptions and expectations. One important way of
Observational learning. Closely related to the
adopting roles is through the process of identifica-
notion of identification is the idea that people will
tion. If we perceive others who are admirable in
adopt behaviors that they observe others perform if
important ways, we will imitate them and try to be as
the behavior leads to some kind of reward or rein-
much like them as possible. They serve as important
forcement. While this process may have relatively lit-
role models. To the extent that a college values intel-
tle to do with emotion or relationship, it is true that
lectual achievement and academic success, per-
students learn not only from listening to their peers
haps because of recognition by the institution’s
but also from watching them. We can learn from oth-
faculty or administration, students may try to emulate
ers about effective ways to manage time, resolve
those who are successful. In identifying with those
conflicts, solve problems, and succeed academical-
who succeed, they will identify with the activities that
ly. For example, learning how to take notes, plan
make them successful. They will develop interests
exam answers, and compile references can be
and adopt behaviors that make them similar and can
accomplished through watching peers. Again, this
lead to similar successes.
kind of learning doesn’t depend on admiration for or
Our capacity to perceive that we are or can be
emulation of a peer, just the perception that imitating
similar to very talented others is extraordinary. It can
his or her behavior is likely to lead to a good result.
be self-deceptive, in that we may never be or come
It obviously makes an important difference how stu-
to be as able as many of those we emulate. Still, the
dents are influenced to judge what is and is not a
effort may bring success and development that
good result.
would not happen otherwise. Sometimes however,
Group Formation and Peer Effects
in College
we may decide that we have very little chance of
being like others who are highly successful and
seem highly admirable. We may cease comparing
There seems little doubt that students will influ-
with them and cease identifying with them.
ence each other. They will have direct effects on
Individuals can be influenced most positively by role
cognitive capacity through talk, both as speakers
37
and as audience. They become sources of informa-
the norms and values of the institution. They will
tion for evaluating opinions and abilities, and there-
have a crucial effect on the extent to which students
by affect judgment, belief, value, level of aspiration,
identify with the educational mission of the college.
and, ultimately, ability. They can serve as models for
How should colleges act to influence their students’
identification and observational learning. Given the
definitions of appropriate groupings? Will a better
enormous effects peers can have, it is crucial to ask
result come from trying to define groups in larger,
the question, What peers? Within the larger group of
diverse units or by encouraging smaller more
undergraduates, students form smaller peer groups.
homogenous groupings? The groupings that result
These groupings will be largely based on students’
will have profound effects on the extent to which stu-
choices (self-selection), and those choices are large-
dents identify with the academic mission, how they
ly based on similarity. However, they are also based
set their level of aspiration, what images they build
on proximity. Proximity is something over which insti-
into their self-concepts, how well they perform in the
tutions can exert some control: examples include
classroom, and their broader social values. It proba-
how colleges and universities arrange housing and
bly makes sense for colleges and universities to
dining; control or influence course choices, or the
think about how they can affect these groupings.
amount of time students spend interacting in curric-
V. How Do We Propose to Find Out?
ular groups such as lab sessions and discussions
groups, or in extracurricular groups, such as athletic
The framing that has proven useful so far looks
teams or ethnically-based social groups. All can
at (a) an individual, (b) his or her peers, (c) peer
have an important impact on peer groupings and
characteristics, and (d) the individual's behavior.
therefore peer influences.
Agent-environment-input-output. Clearly, there's lots
A number of important issues come into play
of room for expansion on each of these. Which indi-
here. First, there is almost certainly going to be some
viduals? What peers? Which peer characteristics?
degree of hostility between groups. This need not be
What individual behaviors? In what setting?
open warfare, but there are probably inescapable
Briefly, we're taking two broad approaches:
tendencies toward ingroup favoritism and outgroup
A. Experiments in the psychology lab
derogation. Therefore, the groups that form have
One approach is to study live groups interacting in
enormous impact. If left to their own devices, stu-
the laboratory. We have begun a series of studies
dents will join groups that are comfortable and famil-
where we look at the impact of two peers on one fel-
iar, based on the similarity of group members. What
low student. All three students read and discuss arti-
actions should colleges take to foster or impose
cles from the “Week in Review” section of The New
uncomfortable interactions and groupings? Second,
York Times. We hope to explore the benefits of inter-
each group will develop its own set of norms and val-
action itself. What do students get from discussion
ues. Those values will vary in their consistency with
and interaction compared to a solo environment
38
where interaction with peers is limited or nonexist-
ural experiments whenever possible.
ent? What are the effects of substitutes for face-to-
One plausible place to look for such a natural
face interaction? Does e-mail have the same
experiment is in student housing assignments.
beneficial effects? What about conference calls, or
Suppose we identify situations where students are
teleconferencing? Our hypothesis is that face-to-
randomly assigned a room during their first (or sub-
face interaction with live peers will produce measur-
sequent) years of study. They will, in effect, inherit a
able increases in learning and in motivation for
roommate by virtue of the housing lottery. Such an
learning. We will attempt to find out by observing stu-
assignment will also (randomly) place them in a par-
dents in a series of experiments where they interact
ticular entry—part of a particular residence. In short,
to varying degrees with their peers.
they will have been assigned a particular (room-
Some preliminary results indicate that students
mate/entry/house) peer environment. Students with
benefit from three different kinds of discussion.
different racial or socioeconomic or academic back-
Compared to issues that they have simply read
grounds are roomed together. This random assign-
about but not discussed, students report more inter-
ment eliminates the selection problems that plague
est in pursuing issues that they have 1) both read
many peer studies. We might ask, for example,
about and discussed, 2) issues that they have not
whether students tend to over or under perform con-
read about but have discussed with peers who have
ditional upon the SAT scores of their roommate. That
read about them, and 3) issues that they have read
is, would knowing the SAT scores of a person's
about and explained to peers who have not read
roommate help us to forecast their GPA—once
about them. These findings are extremely prelimi-
we've controlled for their own SAT scores? We can
nary. But they suggest the promise of studying peer
ask whether the effect is nonlinear. We can also see
effects in the laboratory, and learning how peers can
if the over or underperformance depends on either
enhance the educational experience.
the student's own SAT score or on the gap in SAT
B. Observation and econometric analysis
scores between the roommates. We can measure
of behavior
the presence or absence of peer environment
Once we move from an experimental setting, we
effects at the house, entry, and roommate level of
must be very careful to avoid the issues generated
proximity. A possible extension of this approach
by the selection problem discussed above. We have
would involve designing freshman room assignment
two possible approaches. Either we must find a situ-
strategies to improve the generation of meaningful
ation that simulates an experiment vis-`a-vis the ran-
peer evidence. We might also merge to this data
dom assignment of different peer environments (i.e.,
information on any junior advisors associated with
a natural or quasi-experiment) or we must control for
the various entries in the student houses.
other factors that are linked with both achievement
The College and Beyond database—created by
and peer associations. Our preference is to use nat-
the Andrew W. Mellon Foundation—provides an
39
extraordinarily rich data set for the study of such
We hope we have provided a convincing case
peer effects. It supplies detailed information on the
that peer effects are worthy of systematic study.
college experiences of a total of approximately
The notion that students educate students is cen-
90,000 undergraduate students from 34 selective
tral to a diverse set of issues including selective
colleges and universities in cohorts entering in
admissions, affirmative action, distance learning,
1951, 1976, and 1989. Matching housing data with
and the economic anomaly of college tuition being
this database allows the application of the empiri-
well below cost. We have attempted to frame the
cal approach described above. It also allows the
questions researchers must confront in contem-
exciting possibility of using a broad range of edu-
plating peer effects. We have also suggested an
cational outputs and peer characteristics in the
empirical strategy—employing both experimental
analyses. The large sample sizes are also likely to
and observational methods—for measuring such
allow a precise analysis of students at the top and
effects. Our intention, in taking this first step, is to
bottom of the SAT distribution. It also allows an
frame the issues in a way that will be productive,
investigation of factors that facilitate or discourage
to us and to others, in thinking about peer effects.
peer influences.
Another
promising
application
of
the
College and Beyond data involves the use of
information provided by Alexander Astin and his
colleagues on students’ pre-collegiate aspirations during 1976 and 1989. This data allows several interesting questions to be studied. First,
student aspirations—aggregated to the level of
the school—might be used to provide a measure
of the institution's ethos. Along these lines, we
might ask whether a student's stated pre-collegiate aspirations are more or less likely to be
borne out under different institutional environments. To what extent, for example, are students'
choice of major affected by the aspirations of
their classmates (after controlling for the students prior aspirations).
VI. Conclusion
40
Appendix
ways an individual student's behavior (his choice of
activities) can be influenced by others (peers).
An analytical framework for studying peer
effects can be developed using a straightforward
In the Becker household production framework,
microeconomic analysis—but it's worth briefly
31
repeating for economists and describing for others.
chased inputs (like groceries or educational servic-
an individual customer is seen to combine pur-
es) with his own time and effort to produce the things
The firm (college) and customer (student) are
he really wants (nutrition, learning…). Since all these
seen as separate entities.
activities are set within a 24-hour day, he's got to
The firm uses various inputs in making its prod-
make choices on what to do (hence what not to do)
uct, including materials, labor, and capital services—
and how to do it, choices that determine what he'll
heating oil, faculty time and effort, and labs and
buy, how much he'll work, what he'll earn, and how
buildings. The possibilities for combining inputs to
much satisfaction he'll get out of life. Like the firm, his
make the product are limited and specific to a prod-
household production choices are constrained by his
uct—those possibilities describe the firm's produc-
understanding of appropriate production technolo-
tion technology (a production function). All firms in an
gies—how to cook, how to study, etc.—and shaped
industry are assumed to have pretty much the same
by his particular, individual preferences. Some peo-
technological choices.
ple will do things (produce household activities)
Peer effects on learning are described as a cus-
more efficiently than others, simply because they
tomer-input technology in the important sense that
use a better production technology (they're smarter,
one of the inputs to the firm's production can be
more disciplined, more practiced, stronger, younger,
bought only from the same customers who buy its
etc.). Some people will do different things simply
product. That's seen as a fact built into the technolo-
because they prefer them over other activities.
gy of making the product (educational services). So
So the college or university is seen to produce
with peer effects, a student both buys educational
educational services that it sells to student-cus-
services from the institution and, simultaneously,
tomers at the same time that it gets peer quality from
provides his own services (in the form of an educa-
those same students (in different amounts according
tional interaction with other students) to the college.
Rothschild and White (1995) and Winston (1996)
to the individual student's abilities). The student, in
30
turn, is seen to take those educational services from
have developed some of the implications of this
the institution and combine them (efficiently or not)
technology and the resulting markets and prices.
with his own time and energy to produce, finally,
For the customer, its formal representation as
learning.
consumer/household is even more useful in thinking
Peer effects, then, are the influences that
about peer effects, since it generates a list of the
41
other students will have on his learning (for better
10 – How much she earns or expects from
or for worse).
other activities—ditto.
11 – When the satisfactions of studying accrue
The formal representation of all this as an optimization problem for the student generates a sur-
to her—immediately or with a delay.
prisingly rich and useful list of ways those peer
12 – When the satisfactions of other activities
accrue—ditto.
influences might work. Most basic is the fact that,
within a 24-hour day, doing more of one thing means
13 – How she feels about the present and the
doing less of something else, so all activity choice is
future—her discount rate and/or impa-
based on the relative attractiveness of activities—
tience or willingness to wait for delayed
something can be very satisfying but still not be cho-
gratification, and any myopia that will
sen to do simply because it's trumped by something
induce problems of self-discipline.
that's even more satisfying to do (with some further
14 – Whether the satisfactions of studying are
complications like costs that needn't detain us here).
intrinsic—the pleasure of learning, per se,
(the life of the mind)—or extrinsic and
So a student's learning can be affected by her
instrumental.
peers by their changing:
15 – Whether learning is seen as an investment
1 – How much she likes learning—her utility
activity that affects the production of other
function.
activities (like making it easier to read a
2 – How much she likes doing other things—
magazine or enjoy an opera).
ditto.
16 – Whether the satisfactions of the activity
3 – How efficiently she learns—her household
are those of doing it (process utility) or of
production function.
having done it (goal utility) and, if those are
4 – How efficiently she does other things—ditto.
in conflict (as in struggling with an addic-
5 – What she has to pay for inputs used in
tion), her ability at self-control.
learning—prices.
17 – Her total (exogenous) endowments of
6 – What she has to pay for inputs used in other
energy and unearned income (high-ener-
activities—ditto.
gy people, wealthy people, sick people,
7 – How hard studying is for her—its use of
and financial aid people).
her limited energy (a specific component
Finally, it seems useful that in generating a
of No. 3).
description of the generic “activity a,” the model
8 – How hard other activities are for her—ditto.
forces attention to which activities we expect to see
9 – How much she earns or expects to earn
influenced by an academic peer effect. Some are
from studying—its wage rate.
obvious (studying) while some aren't (late night bull
42
sessions). But most basic would seem to be whether
Endnotes
we are thinking of outcomes as activities, like study-
1
ing, or outcomes as results, like grades or the accu-
Martin Trow, “The Organizational Context,” in T.
Newcomb and E. Wilson (eds.) College Peer
mulation of human capital and skills for future use. If
Groups (Chicago: Aldine Publishing Company,
it's the latter, we need a bit more complicated model
1966).
that includes an investment function that describes
how the individual turns the learning activity into
2
human capital and how that human capital is to be
Eric Hanushek, “The Economics of Schooling:
Production and Efficiency in Public Schools,”
measured or observed. Models of this sort have
Journal of Economic Literature, 24(3) (1986):
been used to analyze addiction and self-control with
1141-1177.
that investment/outcomes feature.
3
Claude Steele and Joshua Aronson, “Stereotype
References
Threat and the Intellectual Test Performance of
D. G. Myers and H. Lamm, “The Group Polarization
African Americans,” Journal of Personality and
Phenomenon,” Psychological Bulletin 83 (1976):
Social Psychology 69 (1995): 797-811.
602-627.
H. Tajel, and J. C. Tuner, “The Social Identity Theory
4 Jeffrey
of Intergroup Behavior,” in S. Worchel and W. G.
(Feb. 23-March 2,1998).
Rosen, “Damage Control,” The New Yorker,
Austin (eds.) The Psychology of Intergroup
5
Relations (Chicago: Nelson-Hall, 1986).
Michael Rothschild and Lawrence White, “The
Analytics of the Pricing of Higher Education and
Other Services in Which the Customers Are Inputs,”
Journal of Political Economy, 103(3) (1995): 573586. See also Gordon C. Winston, “The Economic
Structure of Higher Education: Subsidies, Customerinputs, and Hierarchy,” Williams Project on the
Economics of Higher Education, Discussion Paper #
40 (1996).
6
Robert Klitgaard, Choosing Elites (Basic Books,
1985). See also Henry Rosovsky, The University :
An Owner’s Manual (New York : W.W. Norton,
1990).
43
7
James Samuel Coleman et al., Equality of
Functions,” Journal of Public Economics, 10(1)
Educational Opportunity (Washington, D.C.: U.S.
(1978): 97-196.
Dept. of Health, Education, and Welfare, Office of
13
Education, 1966).
Matter? Peer Group versus Schooling Effects on
D. Robertson and J. Symons, “Do Peer Groups
Academic
8
A. Summers and B. Wolfe, “Do Schools Make a
Attainment,”
London
School
of
Economics Centre for Economic Performance.
Difference?” The American Economic Review, 67(4)
Discussion Paper (1996): 311.
(1977): 639-652. See also Hanushek, 1986.
14
9 Gary
W. Evans, W. Oats, and R. Schwab, “Measuring
Burtless, ed, Does Money Matter? The Effect
Peer Group Effects: A Study of Teenage Behavior”
of School Resources on Student Achievement and
Journal of Political Economy, 100(5) (1992): 966-991.
Adult Success (Washington, D.C.: Brookings
15
Institution Press, 1996).
Steven Rivkin, “The Estimation of Peer Group
Effects,” Amherst College mimeo (1997).
10
R. Ehrenberg and D. Brewer, “Does It Pay to
Attend Elite Private Colleges? Evidence from the
16
Senior High School Class of 1980,” Research in
Social Consequences of Growing up in Poor
Labor Economics. 15 (1996): 239-271. See also
Neighborhood: A Review,” (Evanston, Ill.: Center for
Jere R. Behrman, Jill Constantine, Lori Kletzer,
Urban Affairs and Policy, 1989).
Christopher Jencks and Susan Meyer, “The
Michael McPherson, and Morton Owen Schapiro,
“Impact of College Quality Choices on Wages: Are
17
There Differences Among Demographic Groups?”
Ghettos and Neighborhood Effects on Dropping Out
Williams Project on the Economics of Higher
and Teenage Childbearing,” American Journal of
Education, Discussion Paper # 38 (1996), and
Sociology (1991). See also J. Rosenbaum, "Black
Sarah Turner, A Note on Changes in Returns to
Pioneers—Do Their Moves to the Suburbs Increase
College Quality, mimeo (1996).
Economic Opportunity for Mothers and Children?”
Jonathan Crane, “The Epidemic Theory of
Housing Policy Debate 2(4) (1991): 1179-1213.
11
Michael McPherson and Morton Schapiro,
“Selective Admission and the Public Interest” (New
18
York: The College Entrance Examination Board,
(Chicago: University of Chicago Press, 1987).
William J. Wilson, The Truly Disadvantaged
1990).
19
N. Mounts and L. Steinberg, “An Ecological
12 V. Henderson, P. Mieszkowski, and Y. Sauvageau,
Analysis of Peer Influence on Adolescent Grade
“Peer Group Effects and Educational Production
Point Average and Drug Use,” Developmental
44
Psychology 31(6) (1995) : 915-922.
(New York: Humanities Press,1951), 1st edition,
1926.
20
Urie Bronfenbrenner, The Ecology of Human
Development : Experiments by Nature and Design
26 R.
(Cambridge, Mass.: Harvard University Press, 1979).
Communication,” Journal of Abnormal and Social
B. Zajonc, “The Process of Cognitive Tuning in
Psychology 61 (1960):159-167.
21
Concha Delgado-Gaitan, “Adolescent Peer Influ-
ence and Differential School Performance,” Journal
27
of Adolescent Research 1(4) (1986): 449-462. See
Pragmatics of Human Communication (New York:
also J. Ide et al., “Peer Group Influence on
Norton, 1967).
P. Watzlawick, J. H. Beavin, and D. D. Jackson,
Educational Outcomes: A Quantitative Synthesis,”
Journal of Educational Psychology 73(4) (1981):
28
472-484.
Performance in the White House (Englewood Cliffs,
James Barber, Presidential Character: Predicting
N.J.: Prentice Hall 1992): 4.
22
Judith Harris, The Nurture Assumption: Why
Children Turn Out the Way They Do (The Free
29
Press, 1998).
Processes,” Human Relations 7 (1954):117-140.
23
30
E. Pascarella and P. Terenzini, How College
L. Festinger, “A Theory of Social Comparison
Rothschild and White, “The Analytics of the
Affects Students (San Francisco: Jossey-Bass
Pricing of Higher Education and Other Services in
Publishers, 1991).
Which the Customers Are Inputs,” Journal of Political
Economy, 103(3) (1995): 573-586. See also
24
L. Alexander, R. Gur, and L. Patterson, “Peer
Winston, “The Economic Structure of Higher
Assisted Learning,” Improving Human Performance
Education:
Subsidies,
Customer-inputs,
and
Quarterly 3(4) (1974): 175-186. See also S. Fraser
Hierarchy,” Williams Project on the Economics of
et al. “Two, Three, or Four Heads Are Better than
Higher Education, Discussion Paper # 40 (1996).
One: Modification of College Performance by Peer
Monitoring,” Journal of Educational Psychology
31
69(2) (1977): 101-108, and R. Joiner, D. Messer, P.
Approach to Economic Behavior,” Journal of
Light, and K. Littleton “Peer Interaction and Peer
Economic Behavior and Organization 8(4) (1987):
Presence in Computer Based Problem Solving: A
567-585.
Research Note,” Cognition and Instruction 13(4)
(1995): 583-584.
25
J. Piaget, The Child’s Conception of the World
45
Gordon C. Winston, “Activity Choice: A New