Psych Term Paper

Please see the attached files to see the Term Paper Outline, and the 3 scholarly articles attached that will needed to be refrenced to in the essay.

The social norm I wish to break is “Traffic light regulations” (will be highlighted in yellow in Term Paper Outline Doc.)

In the outline is asks you to predict the outcome of breaking this social norm, which then I give you the ability to make up your own prediction and what could happen in society if this norm was broken.

Please read outline carefully and follow APA formatting.

Thank you

Psyc 241 Term Paper

Could you break a social norm?

Purpose: To examine the power of social situations to influence individual behaviour

Instructions: For this assignment, you will identify a social norm, break that social norm, and

then observe and report other people’s reactions to this violation. In addition to observing the

reactions of others, you should also pay close attention to your own thoughts and feelings. Some

examples of social norms that a person might break: stand facing backwards in an elevator or, if

you are male, wearing barrettes in your hair.

There are 4 stages to this assignment.

1. First, I would like to pick a social norm. I would like you to do some research on social

norm. That is to say, look at the existing research data to see if you can find any research

related to why social norm exist – and, if possible, if there is any research on the specific

social norm that you have chosen to violate. Please include at least 3 peer reviewed

articles.

2. Once you have done your research, I would like you to predict what is going to happen

when you break the social norm.

3. Next, you will break the social norm. How and for how long you do this is up to you, I

will caution you, however, that if the time period is too short, you will not have much to

write about. As you are breaking the social norm, pay close attention to your reactions

and the reactions of others. If you wanted to have one of your friends help you observe

others’ reactions, this would be ok too.

4. Finally, you will write about your experience in 6-8 pages (double spaced, following

APA 6 formatting). In addition to the information discussed already (background

information on social norms, and predictions about what will happen when you break

your social norm), I would also like you to provide information on what you did (describe

your research).

Relevant Resource for APA: http://owl.english.purdue.edu/owl/resource/560/01/

Reading Articles in Social Psychology:

http://arts.uwaterloo.ca/~sspencer/psych253/readart.htm

**The social norm I wish to break is: “Traffic lights, stopping at red light and moving at green”

Pacific Sociological Association

Measuring Social Norms
Author(s): Sanford Labovitz and Robert Hagedorn
Reviewed work(s):
Source: The Pacific Sociological Review, Vol. 16, No. 3 (Jul., 1973), pp. 283-303
Published by: University of California Press
Stable URL: http://www.jstor.org/stable/1388488

Accessed: 22/11/2012 21:20

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
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 support@jstor.org.

University of California Press and Pacific Sociological Association are collaborating with JSTOR to digitize,
preserve and extend access to The Pacific Sociological Review.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=ucal

http://www.jstor.org/stable/1388488?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

MEASURING SOCIAL NORMS

SANFORD LABOVITZ
University of Calgary

ROBERT HAGEDORN
University of Victoria

Although social norms have been considered to be an
important sociological concept (Gibbs, 1965; Morris, 1956;
Jackson, 1966; Sumner, 1906; Homans, 1950; Blake and Davis,
1964), they are seldom measured in a systematic way. In the
relatively few inquiries in which they are operationalized, there
is little concensus on the most reliable and valid techniques. The
purposes of this article are to explore the nature of social norm
measurement and to demonstrate the utility of using more than
one measurement technique. The basic idea is that different
techniques are likely to be biased in different ways.’ These
biases may “cancel each other out” so that if two or more
techniques produce similar conclusions, we can have more
confidence in the techniques (for each acts as a validity check
on the others) and more confidence in the conclusions. Further,
given the theoretical vagueness of social norms, a variety of
approaches designed to tap their various meanings seems to be a
sound research strategy.

SOCIAL NORMS DEFINED

Most definitions of social norms are based on social evalu-
ations or social sanctions or both (see the summary by Gibbs,
Pacific Sociological Rev., Vol. 16 No. 3, July 1973 01973 Pacific Sociological Assn.

[2831

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[284] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

1965). Evaluations refer to statements of should or ought, or
should not or ought not. If evaluations are stressed, norms are
defined as standards of conduct that should or should not be
followed. Sanctions refer to any kind of reactions to behavior,
positive or negative, that attempt to alter the behavior, or
increase or decrease its frequency. If sanctions are stressed,
norms are defined as standards of conduct that are rewarded or
punished. Stated otherwise, people receive physical or psycho-
logical rewards and punishments for their behavior, which
encourages or discourages them to conduct themselves in a
similar way in the future.

THE IMPUTED IMPORTANCE OF SOCIAL NORMS

If most people behave in accordance with at least some
norms-that is, if they behave as they should according to
certain standards of society-then those norms may account, in
part, for the behavioral patterns in a society; and different
norms among societies may account, in part, for behavioral
variation. Social norms, consequently, may be important links
to order and predictability, which would make them critical
factors in social control.

Furthermore, if social norms are learned, they may be
modified over time; and if they are truly a causal link to
behavior and interaction patterns, then normative modifications
may account for certain social changes. Along this line,
normative conflict and normative ambiguity also may lead to
social changes such as a breakdown in the social order.

Besides their imputed link to social change and their unifying
and controlling nature, conformity and deviance are usually
defined in terms of following or not following social norms. To
establish whether conformity or deviance has occurred, con-
sequently, it is first necessary to establish the norms in
question.

These conceptual treatments of the importance of social
norms are largely speculative. The degree of predictive and
explanatory power must be tested before their importance can

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [285]

be established, and any test depends on reliable and valid
measurement.

MEASUREMENT OF SOCIAL NORMS

Techniques for measurement may be subsumed under four
types: (1) questioning people, (2) inferring norms from be-
havior, (3) postulated effects-that is, measuring the implica-
tions of hypothetical norms, and (4) written documents. Each
technique appears to have particular kinds of biases that may
not characterize the others.

NORMS MEASURED BY DIRECT QUESTIONS

Questioning respondents may involve the direct approach
(“Should people go to church?”) or the use of a hypothetical or
actual situation (“What should the teacher do to the student
who was caught cheating on the final examination?”). Ques-
tions, furthermore, may stress the evaluative aspects of social
norms-that is, should or should not statements-or they may
stress sanctions-that is, rewards and punishments. Studies may
use one or a combination of such questions. The important
point is, however, that responses to the questions determine the
content and dimensions of social norms.

Questionnaire items are always subject to potential biases.
For example, some respondents may lie or have faulty
memories; some simply are not qualified to provide certain
kinds of data; and some responses are invalid because respond-
ents may give answers they feel the researcher wants or that
represent their community rather than themselves.

An illustration of measuring social norms by questioning is a
study by Turk (1965) concerning student participation in the
classroom situation. In response to an article by Wrong (1961)
stressing the “over-socialized conception of man,” Turk feels
that the author has overstated his case. In contrast to Wrong’s
thesis that norms conflict with individual motives, he hypothe-

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[286] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

sizes that agreement on social norms applied to others will be
greater than agreement when self-applied.

Turk tested his agreement thesis on two introductory
sociology classes at Duke University in the early 1960s. His
problem was to specify the social norm on student participation
in the classroom; specifically, on the amount of classroom time
a student should use to express a lengthy opinion. To test the
thesis, he obtained from each student the amount of time they
felt they should use and the amount of time they felt others in
the class should use. Each class was divided into two groups and
each group answered only one of the following questions:

If you were to express a lengthy opinion in a regular one-period class
at Duke, what is the longest you should take so that class time will
not be monopolized? minutes and seconds.
If a classmate were to express a lengthy opinion in a regular
one-period class at Duke, what is the longest he should take so that
class time will not be monopolized? minutes and seconds.

The results of the study support the thesis. Agreement, as
measured by the variance, was greater with regard to others
than with regard to oneself. Students, on the average, allotted
themselves more time to express an opinion than they allotted
to others.

Another study measuring social norms by direct questioning
was carried out by the authors. The study measures the range of
acceptable behavior of student involvement in determining the
content of courses and is based on a conceptualization by
Jackson (1960, 1964, 1966; Glick and Jackson, 1970). Jackson
maintains that social norms are measured on two dimensions.
First, there is a behavioral dimension specifying the amount or
extent that is acceptable-for example, the degree to which a
foreman supervised his workers, the number of times a person
speaks in group therapy, or the degree to which citizens are
involved in community affairs. The second dimension is an
evaluation of the degree of approval or disapproval of the
extent or range of behavior-for example, most people may
approve of a person speaking from two to four times in group

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [287]

therapy, but disapprove of those who remain silent or repeated-
ly speak out and, thereby, dominate the discussion. Note that
Jackson uses approval rather than should in his specification of
social norms.

In two classes-a senior sociology class at the University of
Southern California and an introductory sociology class at the
University of Victoria-students were asked the following:
“According to an evaluation scale ranging from +3 (highly
approve) to -3 (highly disapprove) and containing the values,
+3, +2, 0, -1, -2, -3, what percentage of student suggestions
on course content should be accepted by a professor: 0, 10, 20,
30, 40, 50, 60, 70, 80, 90, 100?” The mean evaluational
response (approval to disapproval) was then calculated for each
percentage and plotted as shown in Figure 1.

Not only does the figure show the acceptable and unaccept-
able ranges of behavior for both classes, but it indicates the
intensity of feeling associated with each behavioral distinction
by the heights and depths of the curves. Perhaps the most
striking difference between the two classes is the substantially
greater intensity of feeling by the USC students. Their curve
reaches a greater height (over +2 average), as well as a lower
depth (almost the maximum negative evaluation of a -3 at both
extremes). The second major distinction is the higher acceptable
range of professional behavior (in accepting student suggestions)
for the class at USC. Although their acceptable ranges overlap
to some extent, the class at Victoria positively evaluates an
acceptance of only about 15% of the student suggestions; for
the USC class, the lowest positive evaluation is just above 30%.
At the other extreme of acceptable behavior, UV positively
evaluates up to about 55%, while USC approves, on the average,
about 65%. It seems clear that not only do students in the USC
class want a larger influence over what is taught than those in
the class at UV, but they feel more strongly about it.

Besides measuring social norms by direct questions, a
researcher may ask respondents to select the acceptable or
unacceptable behavior in a hypothetical story. Rather than
asking directly, “What is the norm governing a specific

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[288] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

Figure 1. MEAN EVALUATION SCORES FOR TWO CLASSESa ON THE PER-
CENTAGE OF STUDENT SUGGESTIONS ON COURSE CONTENT
THAT ARE ACCEPTED BY THE PROFESSOR

(highly
approve) +3

Evaluation 4-USC Class
Scale +1

–UV Class

(indifferent) 0
10 20 30 40 50 60 70 80d 90 100

-1

-2
Percentage of Student Suggestions -2

(highly
disapprove) -3

a. Senior sociology class at the University of Southern California in 1970, and an
introductory sociology class at the University of Victoria in 1970.

behavior?” we ask, “What would the norm be if the following
situation occurred?”

A good example of the use of a hypothetical story is a study
carried out by Angell (1962). Based on a sample of 800 (out of
approximately 3,000,000) in Detroit and contiguous areas in
1955-1956, he studied the “proper” standard of conduct in
areas of religion, civil liberties, family privacy, and race
relations. In the attempt to determine the social norms
governing a given problem situation, respondents were asked to
choose one of three alternatives after hearing a brief hypothet-

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [2891

ical story. The hypothetical story in the area of civil liberties is
given below.

The story concerns the problem of freedom of speech- for
atheists as opposed to the religious orientation of the society:

At a college, a student group plans to ask someone who does not
believe in God (an atheist) to give a speech on his views toward
religion. Some people want the president of the college to do
something about this speech. What should the president do?
(a) He should allow the speech to be made.
(b) He should ask students to call off the speech, but not forbid it.
(c) He should not allow the speech to be made.

Fifty percent of the respondents selected alternative a, 25%
alternative b, and 21% alternative c. The results, consequently,
support the freedom of speech for atheists, but they are not
conclusive. Combining the responses to alternatives b and c
shows that 46% of the sample are opposed to this type of
freedom of speech for atheists-that is, they do not want the
speech to be made, and they are willing to do something about
it. It may be concluded tentatively that, although there is a
tendency to support it, substantial dissensus characterizes the
sample on the following social norm: “Freedom of speech
should be afforded to atheists to the extent of allowing them to
speak on campus.”

Because norms seldom if ever apply universally, Samuel A.
Stouffer (1949) tried to specify the situations in which some
are applicable and also specify the situations characterized by
normative conflict (for a more recent application of Stouffer’s
design, see Stoodley, 1959). The nature of the conflicting
norms and the specific situation, according to the author, are
important factors because they at least partially determine a
person’s choices among a set of behavioral alternatives.

The study is based on 196 students at Harvard and Radcliffe
who were mostly undergraduates. The major concerns of the
study are over the issues of cheating on examinations and
whether the social norms about cheating change with the

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[2901 PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

situation. One of his major propositions is that persons strain to
conform to social norms, even in conflicting situations.

Based on anonymous questionnaires, each student was told
to imagine that he was proctoring an examination. During the
examination, the proctor catches a student cheating (by
copying answers from written notes). The students were asked
what they would do in this situation under a given set of
different circumstances. They were asked what they would do,
for example, if they did not know the student, and then they
were asked what they would do if the student was a roommate
or a close friend. Under such situations, the alternative actions
listed for the proctor were (1) “Take away his notes and exam
book, dismiss him, and report him for cheating”; (2) “Take
away his notes, let him finish the exam, but report him for
cheating”; (3) “If he can be led to withdraw from the exam on
some excuse, do not report him for cheating; otherwise report
him”; (4) “Take away his notes, but let him finish the exam,
and not report him for cheating.”

Stouffer found that those actions approved by the students
were not entirely consistent with what the students thought
would be approved by university authorities. Students ap-
proved, to illustrate, alternatives 3, 4, and 5 much more than
their imputed approval to the university authorities. He found,
further, that being a roommate or friend dramatically altered
the approved actions. These results indicate that student norms
and norms imputed to university authorities are not in complete
agreement on cheating on examinations. The student, when
acting as proctor, therefore, may be under conflicting norms-
that is, the norms of authorities and students do not overlap to
some extent. The extent to which they do not overlap results in
conflict for the student in making a decision. In such conflict
situations, many students selected a compromise alternative,
and, if the student perceived an overlap between his preference
and the preference of the authorities, he was quite likely to
select the alternative in which both approved.

Despite these studies, the question remains, do norms predict
behavior? Stated otherwise, to what extent do people conform

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [291]

to norms? Partially taking their cue from Stouffer’s study, the
authors have carried out a small inquiry on norms and behavior
in cheating on examinations in college. Questionnaires were
handed out in several classrooms, from 1968-1970, at the
University of Southern California and California State College
at Fullerton (now California State University, Fullerton). The
number of students completing the anonymous questionnaires
was 296, which represents almost a 100% return. Social norms
were measured by two questions: “How often should college
students cheat on exams?” (always, usually, sometimes, seldom,
never); and “In your opinion, how often do other students
think college students should cheat on exams?” (always,
usually, sometimes, seldom, never). The first question is
designed to tap the cheating norm of each student as he feels it
applies; the second is designed to tap the imputed norm of
other students. To ascertain the predictive nature of the
cheating norm, two perceived behavioral questions were asked:
“How often have you cheated on exams in college?” (very
frequently, frequently, seldom, never); and “How often have
you seen other college students cheating on exams?” (very
frequently, frequently, seldom, never). About 76% of the
respondents say that college students should never cheat on
exams. This measure suggests that the social norm governing the
examination situation is that college students should not cheat.
About 50% of the students, however, report that they have
cheated in their college career; and 82% report that they have
seen other students cheating. These results strongly suggest that
students are violating the norm to a large extent. Given,
however, that besides the social norm there may be many
factors that may influence a student to cheat (importance of
the course to the student, ability of the student, fairness of the
instructor, and the like), is the cheating norm at least partially
constraining on the individual? That is, can behavior be
predicted by the norm to some extent? Examination of the data
indicates that the self-designated norm on cheating is related to
a student’s reported frequency of cheating (,y = .60). Fifty-seven
percent of those students who feel you should never cheat

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[292] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

report that they, in fact, never do cheat; furthermore, 75% of
the students who say you should cheat to some extent, report
that they have, in fact, cheated to some extent.

NORMS INFERRED FROM BEHAVIOR

Perhaps the most prevalent way of designating social norms is
by inferring from behavior. Social scientists observe some
behavior pattern (a frequently repeated behavior) and impute a
norm from it.2 In England, to illustrate, almost every marriage
is monogamous, so a norm is imputed that marriages should be
monogamous. In the United States, most marriages occur in
June, so perhaps people should marry in June. Since small girls
do play house, small girls should play house. Finally, since
people do cheat on their income tax returns, they should cheat
on them.

The social scientist inferring norms from behavior must be
quite cautious of three factors. First, the “facts” may be
incorrect. Do most people in fact cheat on their income tax
returns? What percentage of the small girls in a society actually
play house? And is this percentage substantially larger than that
for small boys? Inferring norms from behavior is a difficult
enough process without having misperceived the behavior in
question. Second, even if the facts are correct, they may not be
reflecting a social norm; other factors beside social norms may
account for the behavior. Most people may be monogamous, for
example, because this marital relationship is the most economi-
cally feasible in certain types of societies; marriage rates may go
up in June because this is the first month college students are
out from school; and many people may cheat on their income
tax returns because they are very likely to get away with it. We
are not implying that these are the actual or only reasons for
the behavior, but only that they are reasonable alternatives to
the imputation of social norms. A third factor becomes
extremely important in inferring norms from behavior when the
researcher is trying to explain the behavior in question. If the
explanation is in terms of the norm (norms constrain people to

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [293]

behave in certain ways), then the reasoning is tautological or
circular in nature.

A survey by Converse and Campbell (1968) illustrates a
rather sophisticated attempt to infer voting norms from the
voting behavior of Jews, Catholics, labor union members, and
blacks. Their study of the 1956 presidential election is based on
interviews of a random sample of 1,772 adults in the United
States living in private households. The authors concentrated on
establishing the existence of group voting norms such as persons
should vote as their group specifies or as the group leaders
specify. In this case, the four groups of Jews, Catholics, labor
union members, and blacks dominantly push a Democratic
voting norm-that is, members should vote for the Democratic
presidential nominee and for the slate of Democratic candidates
in general.

If members of a group vote alike or are similar on any type of
behavior, argue Converse and Campbell, they may either be
influenced by a group norm which constrains them to behave
that way, or they may be responding to the same life
experiences. Responding to the same life experiences by
behaving alike may merely reflect the notion that similar
experiences lead to similar responses. If so, similar voting
behavior is not due to group voting norms. Their task,
consequently, is to demonstrate that members’ voting behavior
is not a response to similar life experiences.

To demonstrate that similar life experiences are not the sole
cause of a group’s homogeneous voting behavior, Converse and
Campbell compared the “distinctiveness” of voting by group
members when paired with nonmembers of similar character-
istics. Catholics, to illustrate, are more likely to vote Demo-
cratic than non-Catholics of equivalent ethnic background, sex,
age, income, education, occupation, residence location, and
region. Since these background characteristics are indicative of
similar life experiences and are equated for Catholics and
non-Catholics, the only major difference between the two
groups is religious membership. It should be noted that the
comparison did not use all possible characteristics to equate the

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[294] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

groups, and that religious membership also is a life experience.
Although these two factors do not negate the interpretation
that group norms are one of the causes of the distinctive voting
behavior of Catholics, it should caution us in accepting the
conclusion too readily. All four groups proved to be distinctive
in voting to some extent-that is, when compared with
nonmembers of similar characteristics, members of each group
voted more homogeneously. These results are consistent with
the group influence or group norm thesis.

Although the inferred norm is a possible interpretation for
the relatively homogeneous voting behavior in each group,
another possible interpretation is similar life experiences,
because they could not control on all background character-
istics. They may, consequently, have inferred norms where none
actually exist, they have inferred the wrong norm, or, perhaps,
they have inferred the correct norm, but it is too weak or
unknown to have much influence on behavior. Either way, their
conclusion that group voting norms influence members’ voting
behavior may be in error.

Another illustration of inferring norms from behavior is
taken from several “quasi-experiments” reported by Harold
Garfinkel (1964). Based on the disruption of everyday occur-
rences, Garfinkel inferred the existence of what may be called
common understandings or social expectancies between per-
sons. These understandings or expectancies can be viewed as
social norms, because they regulate behavior by designating
what one should or should not do in a given situation. The task
for sociology, according to Gardinkel, is to establish these
understandings and expectancies-that is, social norms-that
underlie everyday life. He attempts to do this by disrupting
everyday activities and observing the consequences. The unmis-
takable disruptive consequences of his quasi-experiments is
consistent with the notion that interaction is sustained by the
unnoticed rules of proper behavior (norms).

To illustrate Garfinkel’s techniques consider the game of
tick-tack-toe. A quasi-experiment may be set up by letting the
naive subject put his X down first; the experimenter, instead of

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [2951

placing his O in an appropriate place, erases the subject’s X and
places it elsewhere. Now the task of the experimenter is merely
to observe the behavior or reaction of the subject. A common
reaction is to state that the experimenter is unfair or perhaps
even to get angry; either way, the negative reaction leads to the
assumption that a rule of the game has been violated-that is, a
social norm has not been followed.

Other examples of his quasi-experiments are bargaining for
standard priced merchandise; in your own household, assume
that you are no longer a son or daughter, but a boarder, and
address your parents formally; and, in ordinnary conversation,
bring your face close to the other person’s face. The inferred
norm in the last example is that there is a proper distance that
should be maintained when conversing.

MEASURING THE IMPLICATIONS OF POSTULATED NORMS

If social norms are treated hypothetically, their implications
may be measured or tested. Only the imputed behavioral
consequences are measured; if the consequences of the unmeas-
ured standard are predicted, then they support the idea that
there is, in fact, a social norm as specified.

Compared to the other three measuring techniques of
questioning, inferring from behavior and written documents the
use of postulated effects is a more indirect method for
determining social norms. The measured factors are the possible
consequences of the postulated norm. Because these conse-
quences could be caused by conditions other than the norm in
question,this technique seems more problematic than the three
more direct measures. Even if the imputed consequences are
observed, rather than validating the postulated norms, one can
only say that the data are consistent with it.

Suppose social norms make human behavior predictable and
establish order in society. What would be some of the
behavioral consequences of a normless situation? The ability to
anticipate or predict the behavior of others seems crucial to
individual stability. If human behavior is haphazard or “ran-

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[296] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

dom,” individuals are likely to experience extreme stress.
Normlessness, furthermore, means that the established order is
gone-society and its subgroups lack cohesiveness. Two likely
results of normlessness, consequently, are low cohesiveness and
personal stress. Without actually measuring the degree to which
either a situation is normless or there is a normative breakdown,
group cohesiveness and individual stress can be determined.
Observations of low cohesiveness and high personal stress in
areas ravished by flood or hurricane support the assumed
normlessness of the situations.

A theory linking status integration to suicide by Gibbs and
Martin (1964) provides a test of the implication of hypothetical
norms. Their theory is based on five interrelated propositions
from which one testable hypothesis is deduced: suicide rates
vary inversely with status integration. Status integration refers
to the occupancy of different positions in society in which the
social norms and roles do not conflict. Being a male and an
airline pilot, to illustrate, does not lead to much conflict, but
being a mother and a businesswoman might. The first two
statuses are highly integrated (male pilots), while the last two
are not (businesswomen who are mothers). A high suicide rate
would be predicted for this type of woman, because the
expectations or social norms for businesswomen (devote a large
amount of time to business) often conflict with those of being a
mother (devote a large amount of time to the children).

The theory’s five propositions and derived testable hypoth-
esis are:

Proposition 1: The suicide rate of a population varies inversely with
the stability and durability of social relationships within that
population.
Proposition 2: The stability and durability of social relationships
within a population vary directly with the extent to which
individuals in that population conform to the patterned and socially
sanctioned demands and expectations-that is, social norms-placed
upon them by others.
Proposition 3: The extent to which individuals in a population
conform to patterned and socially sanctioned demands and expecta-

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [297]

tions placed upon them by others varies inversely with the extent to
which individuals in the population are confronted with role
conflicts.
Proposition 4: The extent to which individuals in a population are
confronted with role conflicts varies directly with the extent to
which individuals occupy incompatible statuses in that population.
Proposition 5: The extent to which individuals occupy incompatible
statuses in a population varies inversely with the degree of status
integration in that population.
Hypothesis: The suicide rate of a population varies inversely with
the degree of status integration in that population.

Observations consistent with the deduced hypothesis lend
support to the untestable propositions; support, however, does
not mean the propositions are proved, but only that they are
consistent with the results and could be correct. Gibbs and
Martin’s theory has proved to be predictive to some extent. The
implied norms in the propositions, consequently, are partially
supported by tests of the derived hypothesis.

NORMS FROM WRITTEN DOCUMENTS

The official documents of formal organizations, incorporated
cities, counties, states, provinces, communes, countries, and
international treaties and associations (like the United Nations)
specify certain laws and rules that govern the people in
question. These formally stated laws and rules usually have
negative sanctions attached to them which indicate the nature
of punishment for violators. Fines or imprisonment may be
imposed for violating national laws; firing or laying off may
accompany disregard for organizational rules; ostracism may be
the fate of violators in certain small, relatively isolated societies;
and breaking off diplomatic relations or even war may be the
result of treaty violation. No matter what the sanctions or
evaluations, sociologists may use statements in official docu-
ments to indicate the existence of social norms. Societal laws
against killing and stealing, organizational rules on authority

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[298] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

and communication, and international treaties on trade rela-
tions may all be treated as social norms.

Actually, any written material can be used to ascertain the
existence of social norms. Books, newspapers, magazines, and
even the notes for speeches may be relevant for the study of
norms. Such material can be scrutinized (e.g., by content
analysis) for evaluational statements of should and should not
and for statements on positive and negative sanctions.

A limitation in using written documents is that they may be
outmoded or may not apply to the group in question. There is
always a time lapse between any written document and the
present, which may result in laws that are no longer enforced.
These laws may still be “on the books,” but they are not
enforced and may not be known (by police or citizen), and
therefore they do not govern the behavior of individuals.

A small inquiry using written documents on societal laws was
carried out by Thorsten Sellin (1967) on the relation between
homicide and capital punishment. Sellin claims that the effects
of the death penalty (and especially its possible deterrence
effect) cannot be ascertained by comparing states that have
abolished it with states that have retained it. Those that have
abolished the death peanalty do, in fact, have lower homicide
rates, but the comparison ignores the vast economic, social, and
political differences between the retentionist and abolitionist
states. These factors, argues Sellin, may account for differences
in the homicide rates rather than the abolition or retention of
the death penalty.

To partially control (hold constant) economic, social, and
political differences, Sellin compared homicide rates for con-
tiguous retentionist and abolitionist states. Being contiguous,
the states are more likely to have similar characteristics than are
noncontiguous states. The underlying rationale is that, if the
states that can be compared are different only on whether or
not they have abolished the death penalty, then the effects of
this factor can be directly determined without being con-
founded by the presence of other factors. The use of contiguous
states, of course, only partially controls for other factors,
because even these states are at least somewhat distinctive.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [299]

Based on several contiguous state comparisons, homicide
death rates (per 100,000 population) were determined for the
period, 1920-1963. The comparisons indicate that the death
penalty does not affect homicide death rates. Abolitionist and
retentionist states cannot be distinguished on the basis of
comparing their respective rates of homicide.

DEGREE OF NORMNESS

Is a social norm a quality or a quantity? Is it best to orient
social norm measures to statements that are differentiated in
kind (“one should be married by the age of 30” or “one should
believe in God”), or in degree (“75% of the population feels
that one should be married sometime between the ages of 19 to
38” of “63% of the population feels that one should have a
moderate to extensive belief in God”)?

In general, sociological statements on social norms have
stressed their qualitative nature. Either a norm exists for a
specified group or it does not. This orientation seems unrealistic
in characterizing social groups. Most groups are heterogeneous
to some extent, and large groups tend to be characterized by
subgroups, which may develop their own social norms. Further-
more, as the Jackson model indicates, norms may refer
differentially to a range of behavior. People may fluctuate in
their intensity of feeling depending on the degree or severity of
the behavior in question. These factors suggest that it may be
valuable to measure social norms quantitatively to tap the
narrow-to-wide variations in their dimensions.

Viewed quantitatively, in terms of the degree of “normness,”
the following dimensions may be considered: (1) the range of
behavior specified, (2) percentage of the group supporting or
not supporting a “should” statement, (3) the intensity of
feeling associated with a “should” statement, (4) the intensity
of feeling associated with different degrees of behavior, (5) the
severity of the sanctions attached to the conformity or violation
of a behavioral pattern, (6) the degree of conformity to a
“should” statement, (7) the extent to which a “should”

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[300] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

statement is known among members of a social group, (8) the
degree to which sanctions are uniformly applied, and (9) the
degree to which a “should” statement is internalized. Although
measuring the degree of “normness” may prove to be a sound
strategy, other than the model proposed by Jackson, very little
in the way of theorizing or inquiry has been done in this regard.

SUMMARY AND IMPLICATIONS

Each of the four techniques for measuring social norms
(questions, inferring from behavior, postulation, and written
documents) has obvious biases. Questions may produce invalid
answers, because respondents may not know what should or
should not be or they may not give truthful answers. Questions,
furthermore, may not have the same meaning to each respond-
ent. Measurement by inferring from behavior may be biased,
because any specific behavior simply may not indicate a social
norm. Marriage rates, for example, may increase in June in
certain areas, not because one should get married, but because
in this month college students begin their vacations. Besides the
possibility that behavior may not reflect a social norm, the
attempt to explain a behavior by a norm leads to circular
reasoning if the norm is first measured by inferring from the
behavior in question. A major bias from measurement by
postulation is that the implications of hypothetical norms may
be in error. Finally, the use of written documents may not
adequately specify social norms, because they may be outdated
or they may not apply to the group in question.

If two or more of these techniques are used to establish the
nature or extent of social norms, then the effects of these biases
tend to be cancelled. For example, suppose we want to know
whether people in a given community should be church
members. We could ask respondents directly (“Should people in
this community be a member of a church?”), but a positive
answer could be based on what the respondent thinks others in
the community would say rather than what he believes. So we

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [301]

may attempt to measure the norm by noting whether, in fact,
most people in the community do belong to a church (infer
from behavior). If almost all do belong, this could indicate that
one should belong. It could also indicate, however, a prevalent
personal feeling that each wants to belong without indicating
that others should belong. Some may be church members,
furthermore, to establish business contacts or to take part in
social activities rather than belonging because it is right and
proper.

Suppose now we attempt to measure the nature of this
potential norm by postulation. If people should be church
members, then they are likely to send their children to Sunday
school, and they are likely to have a comparatively low divorce
rate. We could then observe the implications of such a norm by
finding out the frequency of Sunday school attendance and
comparing the divorce rate of the community to other
community divorce rates where the dominant feeling is not to
be a church member. Of course, even if supported, these
implications may reflect something other than the social norm
in question. Finally, we may check the community newspaper
(a written document) and try to discern its sentiment regarding
church membership. The newspaper, however, may express the
sentiments of only a few in the community. If all four
measurements, or any two or three, consistently show the same
results (for example, that people should be church members),
then confidence can be placed in the conclusion and the validity
of the different measures is supported. At this modest stage in
the development of sociological knowledge and given the vague
conceptualization and treatment of social norms, the use of
several different measurement techniques may prove to be a
sound research strategy.

NOTES

1. For explications of this view, see Webb et al. (1966), Blalock (1971), Denzin
(1970: chs. 3, 5), and Denzin (1971). The essential notion, if supported by sound

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

[302] PACIFIC SOCIOLOGICAL REVIEW / JULY 1973

rationales, is that a theoretical concept requires multiple operations or measurements.
A single measure reflects several dimensions or parameters; seldom, if ever, will it
reflect just one. Therefore, any particular measure is a function of more factors than
just the concept it represents. With regard to the desired dimension, these other
factors are biases. Because different measures tend to have different biases, the use of
several measures may cancel out their effects. This process of “triangulation,”
consequently, is designed to reduce or eliminate rival interpretations to the empirical
findings.

2. A classic sociological study in which informal work norms were inferred from
behavior by the method of participant observation was carried out by Roethlisberger
and Dickson (1939).

REFERENCES

Angell, Robert C.
1962 “Preferences for moral norms in three problem areas.” Amer. J. of

Sociology 67 (May): 650-660.
Blake, Judith and Kingsley Davis

1964 “Norms, values, and sanctions.” Pp. 456-484 in R.E.L. Faris (ed.)
Handbook of Modern Sociology. Chicago: Rand McNally.

Blalock, Hubert M.
1971 “Aggregation and measurement error.” Social Forces 50 (December):

151-165.
Converse, Philip and Angus Campbell

1968 “Political standards in secondary groups.” Pp. 199-211 in D. Cartwright
and A. Zander (eds.) Group Dynamics, New York: Harper & Row.

Denzin, Norman K.
1970 The Research Act. Chicago: Aldine-Atherton.
1971 “The logic of naturalistic inquiry.” Social Forces 50 (December): 166-182.

Durkheim, Emile
1968 The Rules of Sociological Method. New York: Free Press.

Garfinkel, Harold
1964 “Studies of the routine grounds of everyday activities.” Social Problems 11

(Winter): 225-250.
Gibbs, Jack P.

1965 “Norms: the problem of definition and classification.” Amer. J. of
Sociology 70 (March): 586-594.

— and Walter T. Martin
1964 Status Integration and Suicide. Eugene: Univ. of Oregon Press.

Glick, Oren W. and Jay Jackson
1970 “Effects of normative similarity on group formation among college

freshmen.” Pacific Soc. Rev. 13 (Fall): 236-269.
Homans, George C.

1950 The Human Group. New York: Harcourt, Brace.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Labovitz, Hagedorn / MEASURING SOCIAL NORMS [3031

Jackson, Jay
1960 “Structural characteristics of norms.” Pp. 131-163 in N. B. Henry (ed.)

Dynamics of Instructional Groups. Chicago: Univ. of Chicago Press.
1964 “The normative regulation of authoritative bahavior.” Pp. 213-241 in W. J.

Gore and J. W. Dyson (eds.) The Making of Decisions. New York: Free
Press.

1966 “A conceptual and measurement model for norms and roles.” Pacific Soc.
Rev. 9 (Spring): 35-47.

Morris, Richart T.
1956 “A typology of norms.” Amer. Soc. Rev. 21 (October): 610-613.

Roethlisberger, F. J. and W. J. Dickson
1939 Management and the Worker. Cambridge, Mass.: Harvard Univ. Press.

Sellin, Thorsten
1967 “Homicides in retentionist and abolitionist states.” Pp. 135-138 in T.

Sellin (ed.) Capital Punishment. New York: Harper & Row.
Stoodley, Bartlett H.

1959 “A cross-cultural study of structure and conflict in social norms.” Amer. J.
of Sociology 65 (July): 39-48.

Stouffer, Samuel A.
1949 “An analysis of conflicting social norms.” Amer. Soc. Rev. 14 (December):

707-717.
Sumner, William Graham

1906 Folkways. Boston: Ginn.
Turk, Herman

1965 “An inquiry into the undersocialized conception of man.” Social Forces
43 (May): 518-521.

Webb, Eugene J., Donald T. Campbell, Richard D. Schwartz, and Lee Sechrest
1966 Unobtrusive Measures: Nonreactive Research in the Social Sciences.

Chicago: Rand McNally.
Wrong, Dennis

1961 “The oversocialized conceptions of man in modern sociology.” Amer. Soc.
Rev. 26 (April): 183-193.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:20:52 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Article Contents

p. 283
p. 284
p. 285
p. 286
p. 287
p. 288
p. 289
p. 290
p. 291
p. 292
p. 293
p. 294
p. 295
p. 296
p. 297
p. 298
p. 299
p. 300
p. 301
p. 302
p. 303

Issue Table of Contents

The Pacific Sociological Review, Vol. 16, No. 3 (Jul., 1973), pp. 283-405
Front Matter [pp. 304-400]
Measuring Social Norms [pp. 283-303]
Feminist Attitudes as Related to Sex of the Interviewer [pp. 305-314]
Structural Supports for the Development of Professionalism among Police Administrators [pp. 315-343]
Leisure Sports as “Ephemeral Roles”: An Exploratory Study [pp. 345-356]
Values, Authoritarianism, and Antagonism toward Ethnic Minorities: A Swiss Replication [pp. 357-376]
Toward a Theory of Interpersonal Trust [pp. 377-399]
A Comment on the Clemente (MFD) and Gibbs-Martin (MID) Measures of the Division of Labor: Their Relation to Amemiya’s Index of Economic Differentiation (IED) [pp. 401-405]
Back Matter

The Review of Economic Studies, Ltd.

Social Norms and Community Enforcement
Author(s): Michihiro Kandori
Reviewed work(s):
Source: The Review of Economic Studies, Vol. 59, No. 1 (Jan., 1992), pp. 63-80
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/2297925

Accessed: 22/11/2012 21:28

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

Oxford University Press and The Review of Economic Studies, Ltd. are collaborating with JSTOR to digitize,
preserve and extend access to The Review of Economic Studies.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=oup

http://www.jstor.org/stable/2297925?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

Review of Economic Studies (1992) 59, 63-80 0034-6527/92/00030063$02.00

?) 1992 The Review of Economic Studies Limited

Social Norms and Community

En orcement
MICHIHIRO KANDORI

Princeton University

First version received November 1989; final version accepted August 1991 (Eds.)

The present paper extends the theory of self-enforcing agreements in a long-term relationship
(the Folk Theorem in repeated games) to the situation where agents change their partners over
time. Cooperation is sustained because defection against one agent causes sanction by others,
and the paper shows how such a “social norm” is sustained by self-interested agents u-der various
degrees of observability. Two main results are presented. The first one is an example where a
community can sustain cooperation even when each agent knows nothing more than his personal
experience. The second shows a Folk Theorem that the community can realize any mutually
beneficial outcomes when each agent carries a label such as reputation, membership, or licence,
which are revised in a systematic way.

1. INTRODUCTION

It is widely recognized that in many economic transactions, informal means are employed
to execute mutually beneficial agreements. As S. Macaulay (1963) points out, “social
pressure” and “reputation” are perhaps more widely used than formal contracts

and

filing suits. In many cases, people behave honestly because honesty is rewarded and/or
defection is punished in future transactions.

Such informal enforcement mechanisms can be classified into two categories. One
is personal enforcement, in which cheating triggers retaliation by the victim. These
mechanisms are effective only if quick and substantial retaliations are available-that is,
they work best in frequent and long-term relationships. The Folk Theorem in the repeated
game literature (Rubinstein (1979) and Fudenberg and Maskin (1986)) provides a formal
model of personal enforcement, showing that any mutually beneficial outcome can be
sustained as a subgame-perfect equilibrium if the same set of agents frequently play the
same stage game ad infinitum. However, many important transactions are infrequent i

nature. As economic historians argue, the division of labour and specialization are
important driving forces of economic progress. Potential gains are larger in diverse
transactions with different specialists than with fixed partners. Therefore, the control of
incentives in such an infrequent trade is of vital importance to understand the organization
of economic transactions. This observation leads to the second category of informal
enforcement mechanisms, community enforcement, where agents change their partners
over time and dishonest behaviour against one partner causes sanctions by other members
in the society. The present paper is devoted to the study of such mechanisms.

The specification of desirable behaviour together with sanction rules in a community
may be regarded as a social norm, and we analyze how such social norms work to support
efficient outcomes in infrequent transactions. Our approach to this problem assumes the
standard axiom of economics that agents only care about their own utility; that is, we
do not assume that people follow a social norm for its own sake, but we investigate how
such a rule is sustained by self-interested community members. For a social norm to be

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

64 REVIEW OF ECONOMIC STUDIES

sustainable, it must provide proper incentives to the members in every respect. Thus, not
only are deviators from the desired behaviour punished, but a person who fails to punish
is in turn punished. In other words, we will investigate the (subgame-perfect or sequential)
equilibria of the game played by community members.

In small communities where members can observe each other’s behaviour, community
enforcement works in much the same way as personal enforcement. We can easily modify
the usual Folk Theorem to show that any efficient and individually rational outcome can
be sustained when there are frequent interactions among the community members as a
whole, and this is true even if the transactions among any given pair of agents are
infrequent. This observation shows that changing partners itself is unimportant and the
crux of the matter is information transmission among the community members. Given
this, we propose the following research programme: What is the minimal information
transmission necessary to sustain efficient outcomes by community enforcement? As a
first step to answer this question, the present paper shows what a community can achieve
under various degrees of observability.

When the observability is not perfect, each agent typically possesses a piece of
information about the history of trades in the community which may not be known to
others. The presence of such private information characterizes the main theoretical
difference between our models and the models of standard repeated games, where all
relevant information is assumed to be common knowledge among all agents in the game.’
We encounter new problems and the analysis of such games turns out to be a non-trivial
extension of the usual repeated games. The source of difficulty is illustrated by the
following observation. Suppose the community can somehow “mark” deviators, say, by
putting dark spots on their foreheads, and suppose that the community norm requires
that an agent should cooperate if and only if the partner is unspotted. Now consider the
agents’ incentives to follow the norm. Clearly, no one wishes to deviate from the
equilibrium path if the punishment is severe enough. To show that this social norm is
an equilibrium, however, we must also show that everybody has incentives to follow the
norm after any history. Consider two unspotted players matched to each other, and
suppose that one of them is likely to encounter many spotted partners in the future. If
the punishment is costly to carry out, this may destroy the incentives for them to cooperate,
because one of them doesn’t have enough stake in the future. And just as in any games
with private information, their incentives depend on how this information is known to
them, which is determined by their personal experiences (for example, how many deviators
each of them has seen and when they were observed). Checking incentives and specifying
equilibrium behaviour in those cases can potentially be very complicated, because (i)
private information (what the players have observed) and its distribution will be increas-
ingly complicated over time, and (ii) the players’ private information do not come from
a common prior distribution after deviations occur. Those points show that the game
does not have the usual recursive structure which is possessed by the standard models
of repeated games with public information. The analysis of models with public informa-
tion is greatly simplified because after any history the continuation strategies correspond
to an equilibrium in the original game (see Abreu, Pearce and Stacchetti (1990)). In
contrast, in our model continuation strategies correspond to a part of (partially) correlated

1. See the papers on the Folk Theorems under perfect information cited above, as well as the literature
on repeated games with imperfect monitoring which assumes that the signal in the game is publicly observed
by all players (Abreu, Pearce and Stacchetti (1990), Green and Porter (1984), Fudenberg, Levine and Maskin
(1989), Kandori (1989)). A recent paper by Fudenberg and Levine (1991) deals with the case of privately
observed signals, but they relax the notion of equilibrium to cope with the difficulty.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 65

equilibrium where the correlation device is the private observation of the past history.
Point (i) above says that the distribution of the correlation device is non-stationary and
point (ii) says that the players do not share the same prior of the correlation device.
Hence it is not obvious if the seemingly straightforward social norm described above is
in fact an equilibrium; players’ behaviour off the equilibrium may need to be modified,
and the modifications can be highly complicated.

The paper presents two main results. The first one is an example where a community
can sustain cooperation even when each agent only observes the results of the trades he
engaged in. An agent knows what his partners did to him, but he does not observe their
identity nor what they have done to other agents. We show that even in that case,
cooperative behaviour can sometimes be sustained. The social norm supporting cooper-
ation in such a situation uses a contagious process of defection, and the scope and
limitations of such a social norm will be discussed. The second results gives the Folk
Theorem under mild assumptions, assuming the existence of a mechanism or institution
which processes a certain class of information honestly. Being interested in the minimal
information dissemination, we put the following restrictions on the form of information
transmission.

1. A label is attached to each agent.
2. Before executing trade each agent observes his and his partner’s label.
3. A player and his partner’s actions and labels today determine their labels tomorrow.

Requirement 2 says that each agent has only local knowledge of the current state of the
community, and requirement 3 indicates that the revision of labels does not require
knowledge of the entire society, and allows the information processing to be potentially
decentralized. Reputation, membership, citizenship, social status, and credit cards can
be regarded as examples of this class of information transmission mechanisms.2 We show
how such an information processing device, which does not have any enforcement power
of its own, facilitates community enforcement of efficient trade. These mechanisms are
studied in the pioneering work by Okuno-Fujiwara and Postlewaite (1989) and they show
that the Folk Theorem holds in a random matching game when such information trans-
mission is available. However, as we have seen, the presence of private information
complicates the analysis, and to overcome this difficulty they either assume infinite
population and uniform random matching (where no one expects to meet a defector after
any finite number of deviations), or employ an equilibrium concept which is weaker than
sequential equilibrium. Under mild assumptions on the payoff functions, however, the
present paper proposes a simple strategy profile which defines an exact (i.e. sequential)
equilibrium for any population size and matching rule, but still is able to support any
feasible and individually rational payoffs. The equilibrium reveals the importance of
such notions as repentance and forgiveness in social norms.

The role of information in matching games is also analyzed by Rubinstein and
Wolinsky (1990) in a different context. In their model, buyers and sellers are randomly
matched, and once a seller and a buyer agree to trade, they leave the market. So the
structure of their model is different from repeated games, but they derive results which
are similar in flavour to ours. Their analysis shows that if each player has enough

2. The following may be helpful to motivate the reader who is familiar with the dynamic programming
decomposition of repeated games (See Abreu, Pearce and Stacchetti (1990)). If we treat the whole matching
game as a huge repeated game with all community members, continuation payoffs serve as labels summarizing
all relevant information about the past history, if each player can observe all players’ labels and the revision
of each label can depend on the actions and labels of all players. The question is whether we can find something
analogous which works in a decentralized fashion.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

66 REVIEW OF ECONOMIC STUDIES

information about other players’ history, there exist many equilibria which are not
sustained when such information is unavailable.

The paper is organized as follows. Section 2 introduces the formal model of repeated
matching games. Section 3 shows that community enforcement works in much the same
way as personal enforcement under public observability. Section 4 provides an example
where a community can sustain an efficient outcome even when each agent observes the
history of his transactions only. Section 5 deals with the Folk Theorem when labels are
attached to agents. The final section discusses the related literature and possible extensions
of our model.

2. REPEATED MATCHING GAMES

The basic structure of the repeated matching game is described as follows. The set of
players N = {1, .. , 2n} is partitioned into two sets of equal size, N1 = {1,…, n} and
N2 = {n .1… ., 2n}, where Nk is the set of type-k players (k = 1, 2). In each stage, each
type-1 player is matched with a type-2 player according to some matching rule, and they
play a two-player stage game. This procedure is repeated infinitely and each player’s
total payoff is the expected sum of his stage payoffs discounted by 8 E (0, 1). Most of
the results in this paper do not depend on the way players are matched. For example,
they can be endogenous and history-dependent. Let ,(i, t) be player i’s match at time t.
For some cases, we assume uniform random matching in which

Prob {,u(i, t) =j} = 1/n for all i E N1 and j e N2 and for all t

and the matching in each stage is independent. The stage game is described by a payoff
function g: A – lR2, where A = A1 x A2 and Ak is the finite set of actions for type-k players
(k = 1, 2). The minimax point M1 E A for type-1 players is defined by

Me E arg mina2cA2 (maxaleAl g1(a1, a2))

Ml E arg maxa1eAl g1(al, M’)

and M2 is defined similarly. For simplicity, the “mutual minimaxing point” (M2, Ml)
is denoted by m = (ml, M2). We normalize the payoffs in such a way that the minimax
payoffs are equal to zero (g1(M’) = g2(M2) = 0). Finally, the set of feasible and
individually rational payoffs in the stage game is defined by

V = {v e cog (A) I v >> O},

where cog (A) is the convex hull of the set g(A).

3. FOLK THEOREMS UNDER PUBLIC OBSERVABILITY

We first point out that, if all agents’ past actions are publicly observable, every point
v E V is sustained by a perfect equilibrium, even when agents change their partners over
time. Perhaps the simplest way to prove the assertion is to utilize the same strategies as
the Folk Theorem for two-player repeated game.

Proposition 1. If v E V is supported by an equilibrium in the two-player repeated game
for some 8, then it is also supported by an equilibrium in the matching game for the same
8 with arbitrary population size and matching rule.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 67

The point of this assertion is that the same 8 works for any population and matching
rule; under public observability, each agent has as strong an incentive to cooperate as if
he faced the same partner in each period. This is true even when the chance of meeting
the same partner in the future is very small, or even zero.

Proof. The players in the matching game start by playing the equilibrium path of
the two-player game. If the type-i player deviates (1= 1, 2), all type-I’s are punished by
all of the other type, using the same punishment strategies as in the two-player game.
The same principle applies to any further deviations, and simultaneous deviations are
ignored. Then, it is clear that the incentives of each player are identical with those in
the two-player game because each player encounters the same sequence of action profiles
as in the two-player repeated game, only the opponents being changed over time. Hence
the prescribed strategies are in fact a perfect equilibrium. 11

In the above proposition, an equilibrium is constructed where a large number of
innocent players are held responsible for the deviation of a defector. Since punishment
never occurs on the equilibrium path, this does not entail actual welfare loss, but such
an equilibrium is intuitively unappealing. If only the deviator is to be punished and
innocent pairs are to play the originally prescribed actions, however, more conditions
are necessary to achieve subgame perfection. As was discussed in the introduction,
difficulties may arise in providing incentives for innocent players after some players’
defections. To maintain the faith of innocent agents, we have to avoid the situation

where

too much burden of punishment is imposed on them. Such a situation arises when the
community is highly populated with “guilty” agents, or the matching is highly non-
uniform. Introducing some “forgiveness” in the social norm, we can avoid the first
problem.

Proposition 2. Under uniform random matching, any payoff point v E V is sustained
by an equilibrium where only defectors are punished, if 8 E (8*, 1), where 8* can be chosen
independently of the population size.

Remark. Without the independence of population, the statement would be trivial
because personal retaliations may be effective when players are very patient.

Proof. Let a* E A be the action profile achieving the payoff v. On the equilibrium
path, action profile a* is played.3 If a player deviates from the equilibrium path, he and
his opponents will play mutual minimax m for T periods and then revert to the original
actions a*. Meanwhile, all the other players keep on playing a*. When the deviator
defects while he is punished, the T-period punishment is restarted. If another player
deviates when a player is punished, the former deviator is forgiven and only the latest
deviator is punished. Simultaneous deviations are ignored.

Let x = g(m) and define the average punishment payoffs

V=(1 8 )x+8 v.

Choose 8T E (0, 1) so that V >> 0. Then, no player deviates when he is punished, because
he can earn at most 0+ 8Vk (k = 1, 2) by a deviation, which is less than the original payoff
Vk.

3. Or use a correlated strategy if necessary, and assume that the corelation devices are observable. The
same remark will apply to all propositions below.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

68 REVIEW OF ECONOMIC STUDIES

When a type-k player (k = 1, 2) is innocent, his average payoff is at least

fl = (1 – 8)Xk + (S- _T)[(1 – 1/n)vk + (1/n)xk] + 8TVk,

while by a deviation he receives at most

I = (1 -8)Vk + 8Vk,

where v* = maxa A gk(a). As 8 1 I with ST held constant,

Hj-H_-I (1-_ T)(1 – 1/n)(vk -Xk) > 0,

and therefore the deviation is unprofitable. Since the difference HI-IH’ is smallest for
n = 2, the lower bound on 8 for n = 2 works for all larger n. 11

Those assertions show that under public observability it is unimportant who punishes
the defector; community enforcement works just as well as personal retaliation in the
usual repeated games. In this sense, observability in the community is a substitute for
having a long-term frequent relationship with a fixed partners.

4. COOPERATION WITHOUT INFORMATION PROCESSING

This section analyzes an example where very limited observability is svailable, and
illustrates what kind of difficulty a community confronts in such a case. The example
shows the ability and limitations of social norms with very little information, which helps
to explain why a community needs a certain amount of information to sustain its norms.
In this section, the stage game is assumed to be the prisoner’s dilemma game described
by Table 1.

C d

C 1,1 -i,1+g

d i+g,-i 0,0

TABLE 1

The numbers g (the gain from defection) and I (the loss when cheated) in the above
table are positive, and action c stands for cooperation and d stands for defection. The
first number in each entry indicates the row player’s payoff and the second is the column
player’s. Since the game is symmetric, assume n pairs are formed randomly out of 2n
players without any distinction between type-I and type-2. For notational convenience,
let M = 2n denote the population size.

We consider the case in which each player observes only the history of action profiles
in the stage games which he has played. In this situation, a player knows nothing about
the identity of players or what has gone on in the rest of the community. Direct
communication among the players is assumed to be non-existent. We call this case no
information processing. In this situation, there is no way to implement the equilibria in
the Folk Theorems under perfect information described in Section 3, and personal

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 69

retaliation is not available either. One may further conjecture that the only equilibrium
is the repetition of the Nash equilibrium in the stage game. However, we will show that
the community may sometimes still sustain cooperation by what may be called the
‘contagious equilibrium”.

The idea of the “contagious equilibrium” is that trust is attached to the community
as a whole, not to each individual. Therefore, a single defection by a member means the
end of the whole community trust, and a player who sees dishonest behaviour starts
cheating all of his opponents. As a result, defection spreads like an epidemic and
cooperation in the whole community breaks down. We are going to show that this
somewhat peculiar social norm is indeed justified as a sequential equilibrium of the game.4

To describe the equilibrium conditions, let us introduce the “types” of players. A
player is of type c if nobody has deviated in the past history of his stage games, and
otherwise he becomes type d. This means that, once a player cheats or is cheated, he
becomes type d for good. The equilibrium is described by a simple rule; each player
chooses the action which is equal to his type.

Theorem 1. Under uniformly random matching, the contagious strategy described
above constitutes a sequential equilibrium strategy for any given g and M if 8 and 1 are
sufficiently large.

Remark. If a player can believe that a large number of players have already been
cheated whenever he seees a defection, then it is in his best interest to defect forever, as
is prescribed by the contagious strategy. This system of beliefs, however, is not consistent
in the sense of Kreps-Wilson (1982)5so that it is not compatible with sequential equili-
brium. For example, if a player sees a defection in the first period, he must believe that
everybody else is cooperating, because the definition of consistent beliefs requires that
defections by different players are statistically independent. By the same token, when a
player defects when he has seen no defection, he must believe that he is the first person
to defect in the society. Therefore, checking the incentive to follow contagious process
of defection is a non-trivial problem.

The intuition of Theorem 1 is straightforward. When a player defects, his future
payoffs are destroyed by the contagious process of defection. The loss of future payoffs
outweighs the instantaneous gain if the player does not discount the future payoffs much.
Hence a player never deviates from the equilibrium path if 8 is close to unity. On the
other hand, once the contagious process has started, a player faces the following trade-off.
If a player chooses c rather than d, he can slow down the contagious process to enjoy
high payoffs in the future. However, playing c is costly in terms of instantaneous gain.
In particular, there is always a possibility of meeting a d-type, in which case playing c
induces loss l. Therefore, if 1 is large enough, it is in a player’s best interest to follow
the contagious process once it has started.

To derive the formal equilibrium conditions, let us introduce some notation. Let X,
be the number of type d players at time t, and define the M x M transition probability

4. Harrington (1989) independently discovered a similar equilibrium in the case of Poisson matching in
continuous time.

5. A system of beliefs is consistent if it is derived by the following procedure. First, perturb each player’s
strategy so that all actions are taken with positive probabilities. The perturbation of strategies is required to
be independent across players and over time. Then we can unambiguously define the beliefs by Bayes’ rule,
because all information sets are reached with positive probabilities. Finally, take the limit of the beliefs as the
perturbation tends to zero.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

70 REVIEW OF ECONOMIC STUDIES

matrix A=(aij) where aij=Prob(Xt?1=jjX,=i). Also define B=(bij) by bij=
Prob (X,+1 = j I X, = i and one of the d-types deviates to play c at time t), and let H = B – A.
Matrix H indicates how the diffusion of d-types is delayed by unilateral deviation of
one of d-types.6 The probability that a d-type player encounters a c-type is summarized
by a column vector,

pM (M-1,M-2,…,1,0)T

where the ith element of p is the conditional probability that a d-type player meets a
c-type when there are i d-types. Finally, ei stands for 1 x M vector whose ith element is
1 with zeros everywhere else. With this preparation, the formal equilibrium conditions
are given as follows.

Lemma. The contagious strategy constitutes a sequential equilibrium if

1′ (1- )e,(I-BA)-lp (1)
1+g

and

M k g + – I

\ M 11+ / ?ekH(I-8A)-Yp for k=2,3,…,M. (2)

The proofs of Theorem 1 and the Lemma are given in the Appendix.

Note that the contagious equilibrium is more effective at deterring defection than
personal retaliation, where a defector is punished only by the person he cheated. This
is because under the contagious equilibrium a defector not only loses cooperation with
the victim, but also some of the future payoffs with other players are destroyed by the
contagious process. More precisely, if (g, 6) is such that a player is just indifferent
between cooperating and cheating under personal retaliation, he strictly prefers to cooper-
ate under the contagious equilibrium. Since the incentive to start the contagious process
(the inequality (2)) can be maintained when being cheated is costly (when 1 is large),
there are situations in which cooperation is sustained by the contagious equilibrium, but
not by personal retaliation.

There are two things going on in this equilibrium; information transmission and
punishment. A player who was cheated starts cheating, which has the effect of telling
the other members that someone has deviated, and of punishing the cheater who cannot
be identified. The players have positive incentives to follow such behaviour because the
community trust will completely break down sooner or later irrespective of his action,
so that there is no point in sticking to honest behaviour (see condition (2)). As a result,
quick breakdown of community trust occurs after a deviation in a self-enforcing way.

Three points should be noted about the limitations of social norms under no informa-
tion processing:

(1) The above social norm does not have as much cheating deterrence power as the
community enforcement under perfect information, because a cheater is not
punished immediately. And the conditions for sustaining cooperation under

6. For the closed forms of A and H, see the original version of this paper (CARESS working paper No.
89-14, University of Pennsylvania, 1989).

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

K-ANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 71

perfect information are much weaker, so that there exists a case in which
cooperation is sustained under perfect information but not under no information
processing.

(2) In a large population, it is difficult to sustain cooperation under no information
processing, because it takes a long time for a defector to meet players who have
met someone who have met … the player who was cheated by the defector.
Given this observation, it is straightforward to show the following impossibility
result.

Proposition 3. Consider a general matching game defined in Section 2 with uniform
random matching and no information processing. Suppose v E V is not the value of a Nash
equilibrium of the stage game. Then, for any 8, v cannot be sustained by a sequential
equilibrium if the population size is large enough.

The proof is contained in the Appendix.

(3) Even when cooperation is sustained, the norm under no information transmission
has the unfortunate feature that innocent players will necessarily be punished.
Also, it is fragile in the sense that a little bit of noise (“trembling hands”) causes
complete breakdown of cooperation in the community. This may be the main
reason why we do not observe such a norm very often. Given this point, we feel
that a norm should be evaluated not only by its equilibrium payoffs, but also by
its (suitably defined) “robustness”.

Given those limitations of social norms under no information processing, we will
now turn to the case where information is transmitted in a systematic way. Unlike the
case of no information transmission, we will show that we can construct equilibria which
work for any population size and are robust in a suitably defined sense.

5. LOCAL INFORMATION PROCESSING

This section is devoted to the study of social norms when there is a mechanism or
institution which systematically processes some information among community members.
The processing of information is treated as exogenous, and it is assumed that information
is transmitted honestly. Assuming the existence of such a mechanism our main concern
here is to identify a class of simple and “robust” equilibria which sustain any feasible
and individually rational outcomes by community enforcement with as little information
as possible. To this end, we require a series of desirable properties and show that there
exists a class of equilibria which satisfy all of them.

The first requirement is informational decentralization. We require that both com-
munity members’ decision making and the update of information can be made without
the knowledge of the entire society. Although computer networks in advanced societies
may facilitate instant dissemination of information about all the community members,
the presumption here is that such a central coordination device is costly. We restrict our
attention to the situations where each agent carries a label and the necessary information
is transmitted by agents’ labels. Reputation, membership, citizenship, social status, and
credit cards can be regarded as examples of the labels. When two players are matched,
they observe each other’s label first and then take some actions. After that, their labels
are updated depending only on their original labels and actions by a given rule. Thus,
the choice of actions and the revision of labels in this setting are based only on the local

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

72 REVIEW OF ECONOMIC STUDIES

information, and this information structure is referred to as local information processing.
Such a mechanism is first introduced by Okuno-Fujiwara and Postlewaite (1989). It is
formally defined as follows.

Definition 1. A matching game with local information processing has the following
information structure.

1. A state zi(t) E Zk is assigned to player i E Nk (k = 1, 2) at time t.
2. When player i and j meet at time t and take actions (ai(t), aj(t)), their next states

are determined by

(zi (t + 1), zj (t + 1)) = T(zi (t), zj (t), ai (t), aj (t)).

3. At time t, player i can observe at least (zi(t), Z.(i, t)(t)) before choosing his action.

Although this information structure potentially can convey large amounts of information
when the community can utilize a large number of labels, we are interested in finding as
simple a process as possible in this class. The minimal number of labels will be discussed
later in this section.

The second requirement is the simplicity of decision making and the robustness with
respect to the information structure. In potential applications of the present model, it is
highly likely that the players may be able to observe something other than the labels of
their partners, so it is desirable that the equilibrium does not depend on the fine detail
of the information structure. In other words, the labels should be sufficient statistics and
it should summarize all relevant information for agents’ decision making. An equilibrium
in which each labels are sufficient statistics is called straightforward.

Definition 2. A sequential equilibrium in a matching game with local information
processing is straightforward if given that all other players’ choice of actions depends
only on their and their partners’ labels, a player’s best response also depends only on
his and his partner’s labels, even if he had more information than those;

aj(t) = Jk(Zi(t), , 0(t)) for i E Nk, k = 1, 2.

In a straightforward equilibrium, each agent does not need any information other than
his and his partner’s labels, and this, together with the local information processing,
represents a strong form of informational decentralization.

Third, we also require that the equilibrium should be independent of such fine details
of the game as the matching rule and the size of population. Since matching in the real
world is determined in a rather complicated way, the applicability of the model would
be limited if it depended too much on the specification of the matching mechanism. Note
that this requirement drastically reduces the information an agent needs to know. To
check if the norm is actually an equilibrium, one does not have to know the number of
people in the community, or how people are matched.

The next requirement is the stability of the equilibrium. As the example in the
previous section shows, a community may be able to sustain an efficient outcome under
limited observability by using harsh punishments, but the equilibrium may be fragile in
the sense that a small amount of noise (deviation) causes the breakdown of the honest
behaviour in the whole community. On the contrary, if the equilibrium always goes back
to the original payoff point, it is robust to the mistakes of players and it also allows
players to test various actions in order to learn the social norm. A strong notion of
dynamic stability is defined as follows.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 73

Definition 3. An equilibrium sustaining payoffs v E V is globally stable if for any
given finite history of actions h,

lim,,c, E(vi(t) I h) = Vk for all i E N.

where vi (t) is player i’s continuation payoffs at t and E ( * I h) is the conditional expectation.

Finally, equilibria should be simple. For a social norm to be effective, the rule must
be easy to understand. Especially when there are newcomers to the community, simplicity
is crucial for them to learn and follow the community norm. An important measure of
the simplicity of an equilibrium is the number of different actions prescribed to the players
on and off the equilibrium path. Another measure would be the number of the states
(#Zk, k = 1, 2). Those numbers should be small for an equilibrium to be appealing.

In the rest of this section, it will be shown that those requirements can be achieved
for some important classes of stage games. The first class of games is those in which
only one type of player has some incentive problems. A canonical example is a credit
market, where the lender has a choice between lending money or not, and the borrower
can either pay back or not. An efficient choice of action is to lend and to pay back, where
only the borrower can gain by unilateral deviation. In such cases, it is well known that
a straightforward and globally stable equilibrium can be constructed to sustain the efficient
point. For completeness, let us define this class of games and briefly state the result.

Definition 4. A stage game g has a one-sided incentive problem for the type-1 player
at a* E A if a* is a best response to a* and there is a Nash equilibrium in ao in the stage
game such that g1(a?) < gi(a*). A symmetric definition applies to the type-2 player.

Then, the following assertion is rather obvious; cooperation is sustained because if a
player with the incentive problem cheats, he will simply be punished by his partners for
a finite time by the one-shot Nash Equilibrium.

Proposition 4. If g has a one-sided incentive problem at a* E A, then, under local
information processing, there exists a straightforward and globally stable equilibrium support-
ing a*, if 6 E (8, 1) for some 5*, and the statement is independent of the matching rule and
the population size.

Proof See Appendix. II

The credit bureau is an example of the possible application of the above proposition.
Credit bureaus are private institutions which keep track of consumers’ credit history and
sell the information to potential lenders, who decide whether to give credit to a consumer
depending on the record. A recent paper by D. Klein (1989) analyzes a model of credit
bureaus. Another example is the reputation of a firm which can potentially cheat its
consumers by supplying low-quality goods. This was analyzed by B. Klein and K. B.
Leffler (1981). Avner Greif’s (1989) remarkable analysis of medieval trade reveals the
potential applicability of the idea of community enforcement to the problems of economic
history before the creation of law enforcement agencies. He shows some evidence that
medieval Jewish traders controlled their agents’ behaviour through their reputation, and
again the structure of the model is characterized by the one-sidedness of the incentive
problem (only the agents have the opportunity to cheat). The point is that the simple
form of community enforcement in those examples is due to the one-sidedness of the

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

74 REVIEW OF ECONOMIC STUDIES

incentive problem, and is not readily extended to the general case of two-sided incentive
problems unless we impose some assumptions.

If both types of players have incentive problems, a more sophisticated treatment is
necessary. The source of the difficulty is to check the incentives of “honest” players
matched together when there are many “guilty” players in the society.7 The second class
of the games we consider cover those cases and are characterized by the following property.
(Recall that m represents the mutual minimax point.)

Assumption (Al). There exists r c A such that

g1(ml, r2) > gl(m) ? gl(r, iM2)

g2(r, m2) > g2(m)?’-g2(m1, r2).

This property has the following implications. We will introduce an equilibrium in which
two different punishments are used. If a guilty player meets another guilty player, they
mutually minimax each other. If a guilty player and an innocent player are matched, the
innocent minimaxes the guilty but the latter is not supposed to minimax the former;
instead, he “repents” by choosing an action r, which is less harmful for the opponent
(the strict inequalities in (Al)) but more costly for himself (the weak inequalities). By
those dual punishments, honest behaviour is enforced for innocent players even though
they expect to encounter guilty players for a long time in the near future. The property
(Al) is a natural assumption and satisfied by many important games. Examples in this
class include the prisoner’s dilemma game, where m is defection and r is cooperation,
and bilateral trades where the minimax point is no trade and r is interpreted as paying
a fine to innocent players. With those two punishments, we can construct a “robust”
equilibrium which sustains any mutually beneficial outcomes.

Theorem 2. Under the Assumption (Al), every point v E V is sustained by a straight-
forward and globally stable equilibrium with local information processing, if 8 E (8*, 1) for
some 8*, which is independent of the matching rule and the population size. Furthermore,
only three actions are prescribed to each player.

Proof: The state spaces are finite sets Z1 = Z2= Z = {O, 1,…, T}. State 0 indicates
that the player is innocent and otherwise he is guilty. The T states for guilty players
count the number of punishments which last for T periods. Let a*e A be the action
profile to achieve the designated payoffs v. The equilibrium utilizes only three actions,
a* , m, and r. If two innocent players are matched, they choose the designated action
a*. If two guilty players meet, they mutually minimax each other. If an innocent player
encounters a guilty player, the former minimaxes the latter but the latter chooses the
“repenting” action r defined in (Al). That is, for z E Z x Z,

a* ifz=(0,0)

z (MI,r2) ifz1=O,z2#O
(r1, M2) if z1$ 0, Z2 =0
m if z1, Z2O.-

7. Recall the discussion in the introduction about the difficulty of the social norm with “spotted” deviators.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 75

The state transition obeys a simple rule; any deviation starts a T-period punishment.
For zeZxZ,

0 if z1 = 0 and a1 = u1(z)

r(z, a) = z, + 1 (mod T+ 1) if z, 0 0 and a, = ul(z)
I if a1 a (z)$

and the symmetric formula applies to type-2 players.
Then the following conditions ensure that type-1 players cannot profitably deviate

from the prescribed strategies. Symmetric conditions apply to type-2 players.

T)gI~~~~~~~~~~~~~3 (- g(r,, M2) + 8 TVI 0 (3)

(1-_ T) mm {g1(m1, r2), v1}+S Tv1 ? (1- 8)v* + 6[(l – T)g1(m)+ Tv1], (4)

where v* = maxaeA g1(a). Inequality (3) guarantees that a type-1 player follows the
designated action when he is guilty. This is seen as follows. If a player is guilty, his
payoff is at least

x(1 ) + Ax(2) + * * * + S T- 1x(T) +( T + T+1. _ .)VI, (5)

where x(t) is either g1(m) or g1(r,, M2) depending on the opponent. Note that if all
other players conform to the equilibrium strategies, every player will be “forgiven” and
will become innocent T periods ahead, so that the player will surely be matched with
honest players and will receive v1 after that. If he deviates, on the other hand, he can
earn at most

0 + 8x(2) + 5 Tx(T) + 8 Tgl(rl, M2) + (S T+I + 5 T+2 … )VI. (6)

Note that at T + 1 periods ahead he will get g1(rl, M2) because his partner will be innocent.
Thus the difference between (5) and (6) is

X (1) -S gl (r, ,m2) + 16 TVI (- )(rl, M2) + 5 VI

which is non-negative by (3).
Inequality (4) assures innocent player’s actions to be in their best interest. The

left-hand side of that inequality is the lower bound of an innocent player’s average payoff
when he conforms to the equilibrium strategy, and the right-hand side is the upper bound
when he deviates.

Next, we will show that (3) and (4) can be satisfied for some 8 and T. Equation (3)
is satisfied if 8 T is close to one, because v >> 0. If we increase 8 holding 8 T constant, the
right-hand side of equation (4) approaches

(1 8T)g1(m)+ 8TV1,

which is strictly less than the left-hand side of (4), because v, > 0 g1(m) and g1(m1, r2) >

gl(m) (by (Al)).
Finally, the equilibrium is globally stable because given any history, all players will

be forgiven and go back to the original payoff point v after at most T periods. 11

Finally, let us examine the minimum number of states (#Zk, k = 1, 2) to obtain the
Folk Theorem. When random state transitions are allowed, the informational requirement
may be further weakened because in that case counting the number of punishments may
be unnecessary. Suppose each guilty player has an independent chance of being forgiven
in each period. Note that varying the probability of such chances is similar to varying

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

76 REVIEW OF ECONOMIC STUDIES

the number of deterministic punishments. In the standard repeated game, equilibrium
can be constructed in either way. In the matching game, however, stochastic punishment
requires more conditions. This is because unlike in the deterministic case, a player may
encounter a guilty player when he is forgiven, which may destroy the incentives of guilty
players. Therefore, an innocent player’s payoff when he is matched with a guilty player
should not be too small. If this is guaranteed, we can reduce the number of the states
to just two. In a sense, this is the minimal information necessary to sustain any efficient
outcomes. Formally, we need the following assumption.

Assumption (A2) There exists r E A such that

gl(ml, r2)gl(m) -g,(r, m2) O

g2(r1, m2)>g2(m)-g2(mI, r2)?O-

For example, this assumption is satisfied by bilateral trades in which exchange of
commodities is mutually beneficial; in that case, m stands for retaining the endowments
and r for exchange.

Theorem 3. When assumption (A2) is satisfied and stochastic state transition is allowed,
any payoff v E V is sustained by a straightforward and globally stable equilibrium under local
information processing with two states for each player, if 8 E (8*, 1), and 8* can be chosen
independently of the matching rule and the population size.

Proof. See Appendix. ||

Remark. If there is a random event which is observable to all the members in the
community, then forgiving guilty players can be coordinated. That is, all guilty players
can be forgiven at once based on the outcome of the publicly observable event. When
this is possible, a guilty player always expects to be matched with innocent players
whenever he is forgiven, and therefore assumption (Al) is sufficient to establish the same
result as Theorem 2.

6. CONCLUDING REMARKS

In this paper, we have shown how informal sanctions by community members can induce
desirable behaviour in infrequent trades. Specifically, it is shown that a simple action
rule and local information transmission are sufficient to sustain any mutually beneficial
outcomes under weak conditions.

Technically, our model is an extension of the theory of repeated games to the case
of matching games. Early literature on the study of repeated matching games includes
R. Rosenthal (1979), and R. Rosenthal and H. Landau (1979). The former deals with a
model with adverse selection and proves the existence of Markov equilibrium when
players observe their partners’ last actions. The latter analyzes examples of equilibria in
a bargaining game with a specific local information transmission mechanism, which they
call reputation. The attempt to generalize the Folk Theorem of repeated games to the
case of matching games was initiated by P. Milgrom, D. North and B. Weingast (1990),
and M. Okuno-Fujiwara and A. Postlewaite (1989). The former analyzed concrete
examples of information transmission mechanisms and the latter introduced the notion

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 77

of local information processing. Both of them, however, mainly deal with the infinite
population case to avoid potentially complicated problems of incentives on off-equilibrium
paths. Our paper shows that such problems can be resolved in a simple way if the stage
game satisfies a certain weak condition. Equilibria constructed in our paper work for
any population size and any matching rule, and are robust to changes in information
structures.

Perhaps the most important question which is unanswered by the present paper
concerns the way in which the information transmission postulated in our model is
implemented. Three important problems arise in that respect.

1. What is the cost of information transmission?
2. Who pays the cost?
3. How are proper incentives maintained in the information transmission

mechanism?

Papers by Milgrom et al. (1990) and Klein (1988) analyse some of those problems for
such institutions as “law merchants” during the early middle ages and credit bureaus in
modern times, but general theoretical examination of those problems is yet to be done.

APPENDIX

Proof of Theorem 1 and Lemma. By the Principle of Dynamic Programming, we have only to check that
one-shot deviations from the strategy are unprofitable after any history (see Abreu (1988)). The condition that
a one-shot deviation from the equilibrium path is unprofitable is

?, 708=o Ve1A’p(I +g). (*)

The left-hand side is the payoff from cooperating forever, and the right-hand side is the payoff when the player
defects forever. The expression e1A’p is the probability of meeting a c-type at time t given that the player was
the first person to deviate at time 0. Since the equilibrium requires that a player play defection after he has
defected, he receives payoff (1+ g) if he is matched with a c-type, and gets zero payoff otherwise. Summing
up the geometric sequence shows that condition (*) is equivalent to inequality (1).

To show the unprofitability of deviation from an off-equilibrium path (i.e. playing c after observing any
defection including his own), we need in principle to specify a player’s belief on X, (the number of type-d
players), because we are interested in a sequential equilibrium. Rather than following this procedure, we
identify a sufficient condition for a one-shot deviation from an off-equilibrium path to be unprofitable under
any consistent beliefs. This not only avoids complication but also guarantees that the equilibrium is robust to
refinements of sequential equilibria. The condition we are going to use is that a one-shot deviation from playing
d forever is unprofitable for a d-type player given any number of d-types (i.e. X, = k for all k = 2, 3, M);

Z:obekA’p (I+g) ?()(- I+bZ obek eBA p (1I+g).
(M – I M ( MI)

The left-hand side is the payoff from playing d forever when there are k d-type players including the player
himself, and right-hand side is what the player receives when he plays c today and then plays d forever;
(M – k)/(M – 1) and (k – 1)/(M – 1) are the probabilities of meeting type-c and type-d players respectively,
and ekB is the distribution of the number of type d players tomorrow given that there are k d-types and one
of them (the player) deviates to play c today. This can be manipulated to get inequality (2).

Given M and g, the equilibrium conditions (1) and (2) can be satisfied for large 8 and I because lim8_1
(I -M8A)-lp

(I – 8A)-lp =0 b’A’p = I’ 0 b/Ap = (I – 5A)-lp,

where A is a matrix obtained by replacing the last column of A by zeros. Such replacement is justified because
X, = M is the absorbing state and the Mth element of p is zero. Given this, we have only to show the existence

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

78 REVIEW OF ECONOMIC STUDIES

of (I – A)’. Since the number of d-types never declines, A is upper-triangular and so is (I – A). The determinant
of an upper-triangular matrix is the products of its diagonal elements, which are all strictly positive for (I – A). 11

Proof of Proposition 3. Without loss of generality, suppose the type-I player has an incentive to deviate
from v, and suppose the gain from defecting is d, > 0. Let A, = maxa,A g1(a) -minaeA g,(a). Given a strategy,
suppose V and V’ be the continuation payoffs for a type-I player when he conforms to a strategy and deviates
from it respectively (a symmetric argument applies to type-2 players). Then, for sufficiently large n, V – V’ < d, and therefore the only equilibrium is the repetition of a one-shot Nash equilibrium. This is because

?’I “t
, 00 3 ̂*tmin 12′, l,n} asn .

The inequality is explained as follows. Let I, be the set of players whose behaviour at T+ t is affected by the
defection at time T. If player i is the defector at T, I, = {i, s(i, T)} and for t> 1, I, = {Il e I,-, orj = ,u(k, t-i1)
for some k E It-1}. The number of such players is maximal when all members of I, are matched with players
in N\I,, so that an upper bound of #I, is min {2′, 2n}. Hence min {2′, n} is an upper bound of the number
of type-2 players in I, so that min {2′, n}/n is an upper bound of the probability for the type-I defector to
meet a type-2 player in I, at T+ t. 11

Proof of Proposition 4. Suppose the type-i players have an incentive problem, and let Z1, = {0, 1,…,T}
be the state space for type-I and let type-2 players’ state be identically equal to zero. The state transition rule
is the following.

~0a) ~(0, 0) if a,= a*,
(1, 0) otherwise

and r(k, 0) = k+ 1 (mod T+ 1) for k = 1, . . ., T and for all a E A. The action choice is given by

ak, a* ifk=0* I
ao otherwise

Thus, after a deviation, a type-1 player is punished for T periods by one-shot Nash equilibrium. Type-2 players
have no incentive to deviate because their future payoffs are independent of their actions and they are always
taking one-shot best responses. If T and 8 are sufficiently large, the loss caused by the T periods punishment
becomes greater than the gain from a one-shot deviation, and therefore the type-2 conforms to the prescribed
strategy. 11

Proof of Theorem 3. The state spaces are Z, = = Z = {0, 1}, where 0 indicates innocence and 1 means
that the player is guilty. Let a* E A be the action profile which achieves v. The action rule is, for z E Z,

(a* if z=(0,0)

(m,, r2) if zI =?, Z2=

(rl,m2) if zl=i,z2=0
m if z= Z2 = I

The state transition is determined as follows. If z, (t) = 0 and player i does not deviate, then z, (t + 1) = 0, and
if he deviates, z, (t + 1) = 1. When zi (t) = 1, z, (t + 1) = I if he deviates, and if he conforms,

I with probability p
z(t+ 1) 0 with probability I-p.

We will check the incentives of type-I players. Symmetric arguments apply to type-2 players. When a type-1
player is guilty, his total payoff is

rI = X()+k=2 (8p)k (x(k) – u(k)) + 28k-1u(k),

where

u(k) = 0(k)g,(m,, r2) + (1 – 0(k))v,

x(k) = 0(k)gl(m) + (1 – 0(k))g,(rj, M,

and 0(k) is the probability of meeting a guilty opponent after k periods. Note that when the player is forgiven,
his payoff changes from x(k) to u(k). If he deviates, he earns at most

11=O+ Ax(2)+ 82(px(3) + (I -p)u(3)) +.*

k=k2 (8P)2 (x(k) – u(k)) +Ek=2 k1u(k)

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

KANDORI SOCIAL NORMS AND COMMUNITY ENFORCEMENT 79

The difference is

-H = xMl + (pSa- ) Y,k =2 (8p ((k) – u(k)).

Note that x(k) – u(k) = 6(k)(gl(m)) – gl(ml, r2)) + (1 – 0(k))(gl(r, iM2) – Vl) 0 and
Assumption (A2). Therefore,

HII-I>x()+ lP X g1(r1, M2) ->x(l) -g1(r, m2) 0

as St 1 with Sp E (0, 1) held constant. Thus we have shown that any deviation is unprofitable for guilty players.
When a player is innocent, his total payoff is

U = E-
k-l u(k),

and one-shot deviation earns at most U’= maxaeA g1(a)+I1′. Therefore,

U – U’ = (u(1) -maxaEA gl(a) + 8’k=2 (8p) k2(u(k) – x(k)).

Since u(k)-x(k)>O, this is positive if 8 and 8p are close to 1. 11

Acknowledgement. This paper is based on Chapter 4 of my thesis (Kandori (1989)). I would like to
thank my advisors, Kenneth Arrow, Darrell Duffie, and especially Paul Milgrom for their guidance. Comments
and discussion by Peter DeMarzo, Joe Harrington, Masahiro Okuno-Fujiwara, Andy Postlewaite, Ariel Rubin-
stein and Barry Weingast much improved the paper. All remaining errors are mine. I am also very grateful to
the editors of the Review of Economic Studies for giving me an opportunity to present this paper at the London
School of Economics, Universitat Autonoma de Barcelona, and Tel Aviv University in the RES European
Meetings, 1989.

REFERENCES
ABREU, D. (1988), “On the Theory of Infinitely Repeated Games with Discounting”, Econometrica, 56, 383-96.
ABREU, D., PEARCE, D. and STACCHETTI, E. (1990), “Toward a Theory of Discounted Repeated Games

with Imperfect Monitoring”, Econometrica, 58, 1041-1063.
BEN-PORATH, Y. (1980), “The F-Connection: Families, Friends, and Firms and the Organization of

Exchange”, Population and Development Review, 6, 1-30.
FUDENBERG, D. and LEVINE, D. (1991), “An Approximate Folk Theorem with Imperfect Private Informa-

tion”, Journal of Economic Theory, 54, 26-47.
FUDENBERG, D., LEVINE, D. and MASKIN, E. (1989), “The Folk Theorem with Imperfect Public Informa-

tion” (mimeo).
FUDENBERG, D. and MASKIN, E. (1986), “The Folk Theorem in Repeated Games with Discounting or

with Incomplete Information”, Econometrica, 50, 533-554.
GREIF, A. (1989), “Reputation and Coalitions in Medieval Trade”, Journal of Economic History, 49, 857-882.
HARRINGTON, J. (1989), “Cooperation in Social Settings” (mimeo, Johns Hopkins University).
KANDORI, M. (1989) Information and Coordination in Strategic Interaction over Time (Ph. D. thesis, Stanford

University).
KANDORI, M. (1989), “Repeated Games Played by Overlapping Generations of Players”, Review of Economic

Studies (forthcoming).
KANDORI, M. (1989), “The Use of Information in Repeated Games with Imperfect Monitoring”, Review of

Economic Studies (forthcoming).
KLEIN, B. and LEFFLER, K. B. (1981), “The Role of Market Forces in Assuring Contractual Performance”,

Journal of Political Economy, 89, 615-641.
KLEIN, D. (1989), “Cooperation through Collective Enforcement: A Game-Theoretic Model of Credit Bureaus”

(mimeo, U. C. Irvine).
KREPS, D. M. and WILSON, R. (1982), “Sequential Equilibria”, Econometrica, 50, 863-894.
MACAULAY, S. (1963), “Non-Contractual Relations in Business: A Preliminary Study”, American Sociological

Review, 28, 55-67.
MILGROM, P., NORTH, D. and WEINGAST, B. (1990), “The Role of Institutions in the Revival of Trade:

The Law Merchant, Private Judges, and the Champagne Fairs”, Economics and Politics, 2, 1-23.
NORTH, D. (1987), “Institutions, Transaction Costs and Economic Growth”, Economic Inquiry, 25, 419-428.
OKUNO-FUJIWARA, M. and POSTLEWAITE, A. (1989), “Social Norms in Random Matching Game”

(mimeo., University of Tokyo and University of Pennsylvania).
ROSENTHAL, R. W. (1979), “Sequences of Games with Varying Opponents”, Econometrica, 47, 1353-1366.

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

80 REVIEW OF ECONOMIC STUDIES

ROSENTHAL, R. W. and LANDAU, H. J. (1979), “A Game-Theoretic Analysis of Bargaining with Reputa-
tions”, Journal of Mathematical Psychology, 20, 233-255.

RUBINSTEIN, A. (1979), “Equilibrium in Supergames with the Overtaking Criterion”, Journal of Economic
Theory, 21, 1-9.

RUBINSTEIN, A. and WOLINSKY, A. (1990), “Decentralized Trading, Strategic Behavior and the Walrasian
Outcome”, Review of Economic Studies, 57, 63-78.

WILLIAMSON, 0. (1985) Economic Institution of Capitalism (New York: Free Press).

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:28:25 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Article Contents

p.63
p.64
p.65
p.66
p.67
p.68
p.69
p.70
p.71
p.72
p.73
p.74
p.75
p.76
p.77
p.78
p.79
p.80

Issue Table of Contents

The Review of Economic Studies, Vol. 59, No. 1 (Jan., 1992), pp. 1-225
Volume Information
Front Matter
On the Behaviour of Commodity Prices [pp.1-23]
The Demand for M1 in the U.S.A., 1960-1988 [pp.25-61]
Social Norms and Community Enforcement [pp.63-80]
Repeated Games Played by Overlapping Generations of Players [pp.81-92]
The Dynamics of Pretrial Negotiation [pp.93-108]
Convergence to Rational Expectations in a Stationary Linear Game [pp.109-123]
Competitive Profits in the Long Run [pp.125-142]
Price Leadership [pp.143-162]
Iterated Elimination of Dominated Strategies in a Bertrand-Edgeworth Model [pp.163-176]
Disagreement in Markets with Matching and Bargaining [pp.177-185]
A Bargaining Model with Incomplete Information [pp.187-203]
Strategic Delay in Bargaining with Two-Sided Uncertainty [pp.205-225]
Errata: Variable Returns to Scale, Non-Uniqueness of Equilibrium and the Gains from International Trade [p.227]
Errata: Volume Information [p.227]
Back Matter

Social Norms and Social Roles
Author(s): Cass R. Sunstein
Reviewed work(s):
Source: Columbia Law Review, Vol. 96, No. 4 (May, 1996), pp. 903-968
Published by: Columbia Law Review Association, Inc.
Stable URL: http://www.jstor.org/stable/1123430

Accessed: 22/11/2012 21:25

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

Columbia Law Review Association, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
Columbia Law Review.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=clra

http://www.jstor.org/stable/1123430?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

Cass R Sunstein*

I. Tales of Rationality and Choice ……………………… 904
A. Ultimatums and Fairness ……………………….. 904
B. Littering ………………………………………. 905
C. Smoking, Rationality, and Race ……….. …………. 905
D. Recycling in the Hamptons ………………………. 906
E. John Jones …………………………………….. 906
F. The Point of this Article ………………………… 907
G. An Insufficiently Charted Domain . …… ………… 910

II. Definitions and Concepts …………………………… 914
A. Social Norm s …………………………………… 914

1. In G eneral ………………………………….. 914
2. Intrinsic Value, Reputational Value, and Self-

conception …………………………………. 916
3. Norms and Freedom …………………………. 917
4. Exit and Entry ………………………………. 919
5. Norms and Nonfungible Cash …………………. 920

B. Social Roles …………………………………. .. 921
1. In General ………………………………….. 921
2. Roles and Freedom ………………………….. 922
3. Roles and Law ………………………………. 923
4. Citizens and Consumers ………………………. 923

C. Social M eanings ………………………………… 925
D. Social Norms, Social Roles, and Social Meanings……. 928
E. Divisions in the Self and Norm Bandwagons ……….. 929
F. Beliefs About Facts ……………………………… 930

III. Choices and Preferences ……………………………. 931
A. Preferences as Choices ………………………….. 932
B. Preferences Behind Choices ……………………… 935
C. Against the Idea of Preferences …………………… 938

IV. Choices, Norms, Roles, and Meanings ………………… 939

* Karl N. Llewellyn Distinguished Service Professor, Law School and Department of
Political Science, University of Chicago. This is a written version of the Coase Lecture,
delivered at the University of Chicago on November 28, 1995. I am grateful to Bruce
Ackerman, Richard Craswell, Gertrud Fremling, Elizabeth Garrett, Daniel Kahan, William
Landes, Larry Lessig, John Lott, Martha Nussbaum, Susan Moller Okin, Eric Posner,
Richard Posner, Joseph Raz, David Strauss, and Richard Thaler for valuable discussions
and comments. I am also grateful to participants in very helpful workshops at the
University of Chicago and at New York University.

This article is dedicated to the memory of Jean Hampton, an extraordinary friend,
colleague, and philosopher who was also one of the nicest people in the world. Jean
warmly encouraged my efforts to grapple with these problems and in particular with the
expressive function of law, a topic that she did much to illuminate. Her premature death
is an unfathomable loss for many people; her brilliant and humane work will provide
enduring illumination for lawyers as well as political theorists and philosophers.

903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

904 COLUMBIA LAW REVIEW [Vol. 96:903

A. In General …………………………………….. 939
1. Choices and Norms ………………………….. 939
2. Choices and Roles …………………………… 940
3. Choices and Meanings ……………………….. 940
4. Preferences …………………………………. 941

B. Willingness to Pay vs. Willingness to Accept: The Place
of Shame ……………………………………… 941

C. Anomalies, Shame, Altruism, and Free-Riding ……… 944
V. Government Action: On Autonomy and Tools ………… 947

A. Norms and Paternalism ………………………….. 947
B. Tools …………………………………………. 948
C. Levels and Institutions ……… …………………… 952

VI. Government Action: Five Grounds …………………… 953
A. Some Unusual Collective Action Problems . ………. 955

1. Standard Accounts …………………………… 955
2. Puzzles …………………………………….. 955
3. Legal Responses …………………………….. 958

B. Citizens and Consumers …………………………. 959
C. Risk, Autonomy, and Well-Being . …………………. 961
D . C aste …………………………………………. 962
E. The Expressive Function of Law . …………………. 964

VII. Blocked Grounds …………………………………. 965
Conclusion …………………………………………….. 967

I. TALES OF RATIONALITY AND CHOICE

A. Ultimatums and Fairness’

Economists have invented a game: the ultimatum game. The people
who run the game give some money, on a provisional basis, to the first of
two players. The first player is told to offer some part of the money to the
second player. If the second player accepts that amount, he can keep
what is offered, and the first player gets to keep the rest. But if the sec-
ond player rejects the offer, neither player gets anything. Both players
are informed that these are the rules. No bargaining is allowed. Using
standard assumptions about rationality, self-interest, and choice, econo-
mists predict that the first player should offer a penny and the second
player should accept.

This is not what happens. Offers usually average between 30% and
40% of the total. Offers of less than 20% are often rejected. Often there

1. See Alvin E. Roth, Bargaining Experiments, in The Handbook of Experimental
Economics 253, 270-74, 282-88, 298-302 (John H. Kagel & Alvin E. Roth eds., 1995)
(exploring concept of fairness in ultimatum games); Colin Camerer & Richard H. Thaler,
Ultimatums, Dictators and Manners, J. Econ. Persp., Spring 1995, at 209, 216-18
(discussing impact of fairness on ultimatum games).

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

is a 50-50 division. These results cut across the level of the stakes and
across diverse cultures.2

B. Littering

Why do people litter? Why don’t they throw things out instead? So-
cial psychologist Robert Cialdini tried to find out.3 He placed flyers
under the windshield wipers of cars and waited to see what drivers would
do with them. Cialdini made arrangements so that before reaching their
cars, some people would see someone (a Cialdini associate) walk past
them, pick up from the street a bag from a fast-food restaurant, and
throw it in the trash can. Of the group who both saw the responsible
behavior and noticed the flyers, almost none threw them on the street.
In the control experiment, with no one showing responsible behavior,
over 1/3 of the drivers threw the flyers on the street.4

C. Smoking, Rationality, and Race

About 400,000 Americans die each year from smoking-related
causes.5 Government has tried to reduce smoking through educational

campaigns designed to inform people of the risks. Indeed the govern-
ment has now initiated a large-scale program to reduce smoking, espe-
cially among teenagers.6 Despite this fact, about one million Americans

begin smoking each year, many of them teenagers,7 and people worry
that educational campaigns will succeed, if at all, only with well-educated

people.
But consider this. Nationally, 22.9% of white teenagers smoked in

1993, a number that basically has been unchanged in the last decade.
But in the same year, only about 4.4% of African-American teenagers
smoked, a number that is four times smaller than the number a decade
before.8 What accounts for this difference? Part of the explanation ap-

2. See Camerer & Thaler, supra note 1, at 210-11. In December 1995, the ultimatum
game was played in a law school classroom (consisting of about 70 students) in an
experiment run by Richard Thaler and me. The results-consistent with the usual ones-
will be discussed in Christine M.Jolls & Richard H. Thaler, Behavioral Law and Economics
(1996) (unpublished manuscript).

3. See Robert B. Cialdini et al., 24 Advances Experimental Soc. Psychol. 201 (1991).
4. See id. at 221-23.
5. SeeJ. Michael McGinnis & William H. Foege, Actual Causes of Death in the United

States, JAMA, Nov. 10, 1993, at 2207, 2208.
6. See Excerpts From Clinton News Conference on His Tobacco Order, N.Y. Times,

Aug. 11, 1995, at A18.
7. See Michelle Ingrassia & Karen Springen, Waiting to Exhale, Newsweek, May 1,

1995, at 76, 76.
8. See id. It is notable that the reduction in smoking among African-American

teenagers was 50% between 1984 and 1992, whereas the reduction among African-
American adults was 25%, a difference that suggests a substantial effect from social norms
among teenagers. See U.S. Department of Commerce, Statistical Abstract of the United
States 1995, at 144 (1995).

1996] 905

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

pears to lie in differing understandings of what is fashionable. And part
of that difference may lie in a private antismoking campaign in the
African-American community, symbolized most dramatically by posters in
Harlem subways showing a skeleton resembling the Marlboro man and

lighting a cigarette for a black child. The caption reads: “They used to
make us pick it. Now they want us to smoke it.”9

D. Recycling in the Hamptons

In East Hampton, New York-part of the famous and wealthy
“Hamptons”-what used to be called the East Hampton Dump is now the
East Hampton Recycling and Disposal Center. At the East Hampton
Recycling and Disposal Center, there are separate bins for green glass,
clear glass, newspapers, tin cans, paper other than newspaper, and more.

Almost every day in August (the most popular period in the

Hamptons), residents can be found patiently separating their garbage for

placement in the relevant bins. Sometimes this takes a long time. The
people at the Center tend to own expensive cars-Mercedes Benzes,
BMWs-that are parked near the bins. As they separate their garbage,
they look happy.

E. John Jones

John Jones lives in California. Here is a description of some aspects
of his behavior.

1. He buys smoke alarms and installs them in three rooms in his
house.

2. He loves chocolate and ice cream, and eats a lot of both. He also
eats a fair amount of frozen foods; he makes sure that they are “lean”
whenever he has a choice. According to his doctor, he is slightly over his
ideal weight.

3. On warm days, he likes to ride his bicycle to and from work, and
he enjoys riding his bicycle on busy city streets, even though he has heard
about a number of collisions there.

4. He is happily married. He tries to share the work around the
house, but he doesn’t much like domestic labor. He does less than his
share. He acknowledges that this is both true and unfair, and he sup-
ports many policies that are conventionally described as “feminist.”

5. He buckles his seat belt whenever he is in a car. His own car is a
Volvo, and he bought it partly because it is said to be an especially safe
car.

6. He is not worried about the risk of an earthquake in California.
On some days, he says that he doesn’t think that an earthquake is very
likely; on other days, he claims to be “fatalistic about earthquakes.”

9. Ingrassia & Springen, supra note 7, at 76.

906 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

7. He does not recycle. He considers recycling a personal “irrita-
tion.” He is mildly embarrassed about this, but he has not changed his
behavior.

8. He considers himself an environmentalist; his votes reflect his en-
thusiasm for environmentalism. He supports aggressive regulation
designed to encourage conservation and to protect people from risks to
their life and health. In fact he is in favor of mandatory recycling,
notwithstanding his own failure to recycle.

9. In his own mind, his resources fall in various mental “compart-
ments.” Some money is reserved for retirement. Some money is saved
for charitable donations. Some money is kept for vacation. Some money
is for monthly bills. His forms of mental accounting are very diverse. He
is fully aware of this.

Is Jones inconsistent or irrational? Is Jones risk-averse or risk-in-
clined? What is Jones’s dollar valuation of a human life, or of his own
life?

F. The Point of this Article

My goal in this article is to challenge some widely held understand-
ings of rationality, choice, and freedom, and to use that challenge to de-
velop some conclusions about human behavior and the appropriate uses
and domain of law. I particularly seek to understand and defend the
place of law in “norm management.”

I urge that behavior is pervasively a function of norms; that norms
account for many apparent oddities or anomalies’° in human behavior;
that changes in norms might be the best way to improve social well-being;
and that government deserves to have, and in any case inevitably does
have, a large role in norm management.1′ As I will suggest, norm man-
agement is an important strategy for accomplishing the objectives of law,
whatever those objectives may be. One of my goals is to show how this is
so.

10. I understand anomalies by reference to economic understandings of rationality,
see infra Parts IV.B-C.

11. Objections to norm management from the standpoint of liberty would overlap a
great deal with objections to the use of “external preferences” as a basis for law. See
Ronald Dworkin, Taking Rights Seriously 240-55 (1977). I believe that there is no general
objection to the use of external preferences-the use of people’s beliefs about what other
people’s beliefs should be-and also that there is no objection to government’s effort to
change norms. When government attempts to inculcate an anti-littering norm, or to
inculcate a norm in favor of paying taxes, its behavior is not objectionable merely because
norms are being managed. Any objection, to be persuasive, depends on an argument that
government is invading rights. A robust set of rights should be in place to limit norm
management, see infra Part VII; but so long as it is, there should be no objection to norm
management itself. Cf. Joseph Raz, Liberalism, Scepticism, and Democracy, in Ethics in
the Public Domain: Essays in the Morality of Law and Politics 82, 94-98 (1994) (arguing
that “external preferences” should not be automatically excluded from the political
domain).

1996] 907

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

Part of my motivation is therefore practical. Consider the following
table:12

Deaths from Preventable Risks in the United States

Percent of
Risk Total Deaths (Range) Total Deaths/yr

Tobacco 19 14-19 400,000
Diet/activity 14 14-27 300,000
Alcohol 5 3-10 100,000
Microbial 4 90,000
Toxic agents 3 3-6 60,000
Firearms 2 35,000
Sexual behavior 1 30,000
Motor vehicles 1 25,000
Illicit drugs <1 - 20,000

Existing social norms encourage much risk-taking behavior; and al-
most all of these risks of death could be much reduced with different
norms. Consider smoking, diet/activity, alcohol, firearms, sexual behav-
ior, motor vehicles, and illicit drugs as causes of death. In all these cases,
new norms could save lives.13 A regulatory policy that targets social
norms may well be the cheapest and most effective strategy available to a
government seeking to discourage risky behavior. It may complement or
work more efficiently than existing regulatory approaches.

Social norms are also part and parcel of systems of race and sex

equality. If norms changed, existing inequalities would be greatly re-
duced.’4 It is thus transparently important to see whether shifts in social
norms, brought about through law, might operate to save lives and other-
wise improve human well-being.’5

But part of my motivation is theoretical. It involves a conceptual puz-
zle. In the last decade there has been an intense debate about whether
and to what extent law should try to change people’s “preferences.”‘6
But the term “preferences” is highly ambiguous, and it is not clear what

12. Adapted from McGinnis & Foege, supra note 5, at 2208.
13. See Marilyn Chase, Besides Saving Lives, Wearing Helmet When Cycling Is Cool,

Wall St.J., Sept. 18, 1995, at B1 (noting dramatically shifting social norms with respect to
bicycle helmets); see also Richard E. Nisbett & Dov Cohen, Culture of Honor: The
Psychology of Violence in the South (1996) (discussing role of social norms in producing
violence as a result of judgments about “honor”).

14. See United Nations Development Programme, Human Development Report
1995, at 11-28, 99-124 (1995); Susan M. Okin, A Clash of Basic Rights? Women’s Human
Rights, Identity Formation and Cultural Difference (1995) (unpublished manuscript).

15. I give this idea more content below, see infra Parts V.A, VI.C. For the moment I
use it as a placeholder for the reader’s preferred account.

16. See, e.g., Richard A. Epstein, Simple Rules for a Complex World 308-12 (1995);
Robin West, Authority, Autonomy, and Choice: The Role of Consent in the Moral and
Political Visions of Franz Kafka and Richard Posner, 99 Harv. L. Rev. 385 (1985); Richard

908 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

the participants in this debate are actually disputing when they say that
“preferences” should or should not be respected by law. I attempt to clar-

ify possible meanings of the term. I also suggest that when the idea of a
“preference” is unpacked, it becomes plain that the term is often too ab-
stract and coarse-grained to be a reliable foundation for either normative
or positive work. We will thus find reason to doubt the elaborate edifice
of social science based on the notion of “preference.” The ultimate task
is to separate positive, descriptive, and normative inquiries more sharply
and, in the process, to try to untangle different motivational states and
their influences on choices.

More particularly, I aim to make a set of conceptual or descriptive
points:

1. Existing social conditions are often more fragile than might be

supposed, because they depend on social norms to which-and this is the

key point-people may not have much allegiance. What I will call norm

entrepreneurs-people interested in changing social norms-can exploit
this fact. If successful, they produce what I will call norm bandwagons and
norm cascades.17 Norm bandwagons occur when small shifts lead to large
ones, as people join the “bandwagon”; norm cascades occur when there
are rapid shifts in norms. Successful law and policy try to take advantage
of learning about norms and norm change.

2. Sometimes people do not behave as economists predict. Many
important and well-known anomalies in human behavior are best ex-

plained by reference to social norms and to the fact that people feel
shame when they violate those norms.18 Thus when people deviate from
economic predictions-when they appear not to maximize their “ex-

pected utility”-it is often because of norms.
3. There is no simple contrast between “rationality” or “rational self-

interest” and social norms.19 Individual rationality is a function of social
norms. The costs and benefits of action, from the standpoint of individ-
ual agents, include the consequences of acting inconsistently with social
norms. Many efforts to drive a wedge between rationality and social
norms rest on obscure “state of nature” thinking, that is, on efforts to
discern what people would like or prefer if social norms did not exist.
Those efforts are doomed to failure.20

A. Posner, The Ethical Significance of Free Choice: A Reply to Professor West, 99 Harv. L.
Rev. 1431 (1986).

17. A valuable discussion, one that has much influenced the presentation here, is
Timur Kuran, Private Truths, Public Lies 71-73 (1995). Kuran is, however, principally
interested in the concealment or falsification of preferences; I am interested in norms that
alter behavior as well as talk and hence in a somewhat more general phenomenon.

18. See infra Part IV.B.
19. See infra Part IV.C.
20. A qualification is necessary if the definition of rationality is normative and

defended as such. In that case it would be possible to say that a certain norm is irrational
because (for example) it makes lives worse.

1996] 909

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

4. For many purposes, it would be best to dispense with the idea of
“preferences,”21 despite the pervasiveness of that idea in positive social
science and in arguments about the appropriate domains of law and the
state. In normative work, the idea of “preferences” elides morally impor-
tant distinctions among the motivations and mental states of human
agents. In positive work, the idea is too coarse-grained, in the sense that
it disregards contextual factors that produce diverse choices in diverse
settings. People’s choices are a function of norms, which operate as
“taxes” or “subsidies”; and the content of norms depends on the context.
Instead of speaking of “preferences,” we might assess choice in terms of
(1) intrinsic value, (2) reputational effects, and (3) effects on self-
conception.

I also aim to make two normative claims involving the appropriate
domain of law. These claims have a great deal to do with law’s expressive
function, by which I mean the function of law in expressing social values
with the particular goal of shifting social norms.

1. There can be a serious obstacle to freedom in the fact that indi-
vidual choices are a function of social norms, social meanings, and social
roles, which individual agents may deplore, and over which individual
agents have little or no control. Norms can tax or subsidize choice. Col-
lective action-in the form of information campaigns, persuasion, eco-
nomic incentives, or legal coercion-may be necessary to enable people
to change norms that they do not like.

2. Some norms are obstacles to human well-being and autonomy. It
is appropriate for law to alter norms if they diminish well-being by, for
example, encouraging people to shorten their lives by driving very fast,
using firearms, or taking dangerous drugs. It is appropriate for law to
alter norms if they diminish autonomy by, for example, discouraging peo-
ple from becoming educated or exposed to diverse conceptions of the
good.

G. An Insufficiently Charted Domain

Libertarians, some economic analysts of law, and many liberals22 give
inadequate attention to the pervasive functions of social norms, social
meanings, and social roles. Often it is said that in a free society, govern-
ments should respect both choices and preferences. But the case for re-

21. See infra Part III. I think that Gary Becker moves in the direction of dispensing
with the idea in Gary S. Becker, Accounting For Taste (forthcoming 1996), because he
disaggregates the term. I am not sure, however, that Becker would accept my
characterization.

22. The liberal tradition is very complex on this count, and I will not try to sort out its
various strands here. I believe that all of the arguments made here fit well within central
strands of that tradition. For relevant discussion, see Stephen Holmes, Passions and
Constraint: On the Theory of Liberal Democracy 13-41 (1995). On autonomy, see, e.g.,
Thomas Nagel, The View From Nowhere 113-20 (1986); Joseph Raz, The Morality of
Freedom 369-99 (1986).

[Vol. 96:903 910

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

specting these things depends partly on their consequences and genesis,
and as I have indicated, the determinants of choices (indeed the very
meaning of the term “preference”) remain obscure.23 We should agree
that social norms play a part in determining choices; that people’s
choices are a function of their particular social role; and that the social or

expressive meaning of acts is an ingredient in choice.24 Of course norms

vary a great deal across cultures, and sometimes even within cultures. We
should try to see when social norms, social roles, and social meanings are
obstacles to human well-being, and whether something might be done to

change them, even if people are making “choices,” even if there is
neither force nor fraud, and whether or not there is “harm to others.”

One of my central points here is that individual agents have little
control over social norms, social meanings, and social roles, even when

they wish these to be very different from what they are.25 This is not an

argument against norms, meanings, and roles. Human beings can live,
and human liberty can exist, only within a system of norms, meanings,
and roles; but in any particular form, these things can impose severe re-
strictions on well-being and autonomy.

As I have suggested, agents who seek to make changes in norms face
a collective action problem. For example, it is impossible for an individ-
ual to alter norms determining whether the act of smoking seems daring,
or the act of recycling seems exotic, or the act of opposing sexual harass-
ment seems humorless. This is so even though the relevant norms greatly
influence behavior. If, for example, smokers seem like pitiful dupes
rather than exciting daredevils, the incidence of smoking will go down. If

people who fail to recycle are seen as oddballs, more people will recycle.
If the role of secretary is not associated with susceptibility to unwanted
sexual attention, there will be less unwanted sexual attention. The point
bears very much on current public disputes. If single parenthood is stig-
matized, social practices will change accordingly; if homosexual mar-

riages are consistent with social norms, social practices will be much al-
tered. In all of these cases, individual actors need to act together in order
to produce the relevant shifts.

More particularly, I hope to draw attention to the fact that people’s
conception of appropriate action and even of their “interest” is very

23. Illuminating discussions include Elizabeth Anderson, Value in Ethics and
Economics 1-90 (1993); Amartya Sen, Behavior and the Concept of Preference, in Choice,
Welfare and Measurement 54 (1982); Jean Hampton, The Failure of Expected-Utility
Theory as a Theory of Reason, 10 Econ. & Phil. 195 (1994); Amartya Sen, Internal
Consistency of Choice, 61 Econometrica 495 (1993) [hereinafter Sen, Internal
Consistency].

24. For especially instructive discussion, see Lawrence Lessig, The Regulation of
Social Meaning, 62 U. Chi. L. Rev. 943 (1995). Though I have referred to this paper at
various points, my presentation here owes a general debt to Lessig’s argument and in
particular to his emphasis on the collective action problem presented by social meanings.
See id. at 991-1016.

25. See id. at 1000-07.

1996] 911

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

much a function of the particular social role in which they find them-
selves. This is true of (for example) judges, lawyers, doctors, parents,
children, waiters, wives, husbands, colleagues, friends, and law school
deans. Attention to the place of social role shows that for many purposes,
the contrast between “rationality” and social norms26 is unhelpful. What
is rational for an agent is a function of, and mediated by, social roles and
associated norms.27 And when social norms appear not to be present, it
is only because they are so taken for granted that they seem invisible.

At the same time, norms and roles-operating as taxes on or subsi-
dies to action-can create a division between the judgments and desires
that are displayed publicly and the judgments and desires that would be
displayed without current norms and roles.28 People’s private judgments
and desires diverge greatly from public appearances. For this reason cur-
rent social states can be far more fragile than is generally thought-small
shocks to publicly endorsed norms and roles decrease the cost of display-
ing deviant norms and rapidly bring about large-scale changes in publicly
displayed judgments and desires. Hence societies experience norm band-
wagons and norm cascades. Norm bandwagons occur when the lowered
cost of expressing new norms encourages an ever-increasing number of
people to reject previously popular norms, to a “tipping point” where it is
adherence to the old norms that produces social disapproval.29 Norm
cascades occur when societies experience rapid shifts toward new
norms.30 Something of this kind happened with the attack on apartheid
in South Africa, the fall of Communism, the election of Ronald Reagan,
the use of the term “liberal” as one of opprobrium, the rise of the femi-
nist movement, and the current assault on affirmative action.

26. See Jon Elster, The Cement of Society 1-16 (1989) [hereinafter Elster, The
Cement of Society]. It might be possible to define rationality in a way that abstracts from
social norms, seeJon Elster, Norms of Revenge, 100 Ethics 862, 872-76 (1990) [hereinafter
Elster, Norms of Revenge], but it would be hard to make robust predictions on the basis of

any such definition. The problem lies in a norm-free specification of the “ends” that
rational actors pursue. If we understand rationality in purely instrumental terms, we will
be unable to make any predictions at all; if we understand it in purely economic terms, we
will find much social irrationality and hence make bad predictions, and in any case the

pursuit of wealth will inevitably have some relation to social norms. (Elster sees revenge
behavior as inconsistent with rationality, but individuals who seek revenge get hedonic
benefit from getting revenge. Hence Elster’s conception of rationality is normatively
invested in a way that seems plausible but needs defense, and that in any case is designed
for normative rather than positive purposes.) See also infra Part IV.C.

27. See Viviana A. Zelizer, Morals and Markets: The Development of Life Insurance
in the United States 41-89 (1983) (discussing conflicts between social norms and the
purchase of life insurance).

28. See Kuran, supra note 17, at 3-5, 22-44.
29. See id. at 71-73.
30. Cf. Viviana A. Zelizer, Pricing the Priceless Child: The Changing Social Value of

Children (1985) (describing changing economic and sentimental valuation of children);
Sushil Bikhchandani et al., A Theory of Fads, Fashion, Custom, and Cultural Change as
Informational Cascades, 100 J. Pol. Econ. 992 (1992) (explaining how informational
cascades can contribute to the rapid spread of new behaviors).

912 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

To spell out the most general point emerging from the discussion:
The notion of a “preference” can be deeply confusing, and in many of its
uses, it impairs both positive and normative analysis of law. In its stan-
dard form, a preference is supposed to be something that lies behind
choices and that is more abstract and general than choices are.31 But
what lies behind choices is not a thing but an unruly amalgam of

things32-aspirations, tastes, physical states, responses to existing roles
and norms, values, judgments, emotions, drives, beliefs, whims. The in-
teraction of these forces will produce outcomes of a particular sort in
accordance with the particular context. Hence we might say that prefer-
ences are constructed, rather than elicited, by social situations,33 in the sense
that they are very much a function of the setting and the prevailing
norms.

This point bears on the role of government, which cannot avoid af-

fecting social norms. A market economy will, for example, have predict-
able effects on norms,34 and historically it has been justified on just this

ground, as a way of softening social divisions by allowing people to inter-
act with one another on a mutually beneficial basis.35 A good deal of

governmental action is self-consciously designed to change norms, mean-

ings, or roles, and in that way to increase the individual benefits or de-
crease the individual costs associated with certain acts. Thus government
might try to inculcate or to remove shame, fear of which can be a powerful
deterrent to behavior. The inculcation of shame operates as a kind of
tax; the removal of shame might be seen as the elimination of a tax or
even as a kind of subsidy.

There is a thin line between education and provision of information
on the one hand and attempted norm-change on the other. In fact we
will see that in the process of norm management, government has a

31. This is the idea behind much of Gary Becker’s work. See, e.g., Gary S. Becker, A
Treatise on the Family (enl. ed. 1991). For Becker’s most recent statement-which is, I
believe, in a somewhat different spirit-see Becker, supra note 21.

32. See Gary S. Becker, Nobel Lecture: The Economic Way of Looking at Behavior,
in The Essence of Becker 633, 633 (Ram6n Febrero & Pedro S. Schwartz eds., 1995) (“An
important step in extending the traditional theory of individual rational choice … is to
incorporate into the theory a much richer class of attitudes, preferences, and
calculations.”).

33. Cf. Paul Slovic, The Construction of Preference, 50 Am. Psychol. 364 (1995)
(arguing that preferences are shaped by situation-specific cognitive processes). I mean to
use the idea of construction somewhat more broadly than does Slovic. Note in this
connection the striking study by Samuels and Ross, showing that people cooperate when a
certain game is denominated “Community” but not when the same game is denominated
“Wall Street.” See S.M. Samuels & L. Ross, Reputations Versus Labels: The Power of
Situational Effects in the Prisoner’s Dilemma Game (1993) (unpublished manuscript); see
also Lee Ross & Andrew Ward, Naive Realism: Implications for Social Conflicts and
Misunderstandings in Values and Knowledge (Terrance Brown et al. eds., forthcoming
1996).

34. See generally Albert 0. Hirschman, The Passions and The Interests (1977).
35. See id. at 69-93.

1996] 913

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

number of tools. In a democratic society, it ought to be willing to use
them.

This Article comes in seven parts. Part II discusses the basic con-

cepts: social norms, social roles, and social meanings. Part III explores
the relationship between preferences and choices; it explains the limited
usefulness of the idea of “preferences.” Part IV places special emphasis
on the role of pride and shame in human behavior. Part V offers some

general comments on the role of government in altering norms. Part VI

explores more particular bases for governmental or legal action. It shows
how government might overcome collective action problems, diminish
risk, and enhance autonomy through managing norms. Part VII briefly
describes when and why government action may be illegitimate or ill-
advised.

II. DEFINITIONS AND CONCEPTS

In this section I discuss three ideas that will play an important part
throughout this Article: social norms, social roles, and social meanings.
Social norms are the most central of these, since they provide the founda-
tion for both social roles and social meanings. I also connect these ideas
to problems of collective action, movements for legal and social change,
and some issues about the goals and functions of law.

A. Social Norms

1. In general. – The term “social norms” might be understood in

many different ways.36 For present purposes the differences among the

possible definitions are not very important, and we can rely on conven-
tional understandings. If a definition is thought necessary, we might, very
roughly, understand “norms” to be social attitudes of approval and disap-
proval, specifying what ought to be done and what ought not to be done.
Some norms set good manners, for example, about how to hold one’s
fork; others reflect morally abhorrent views, as in the taboo on interracial
relations; others reflect hard-won moral commitments, as in the norm

against racial epithets. In fact there are social norms about nearly every
aspect of human behavior. There are norms about littering, dating,
smoking, singing, when to stand, when to sit, when to show anger, when,
how, and with whom to express affection, when to talk, when to listen,
when to discuss personal matters, when to use contractions, when (and
with respect to what) to purchase insurance.

36. I am understanding norms in an unusually broad sense. Narrower definitions are
possible and many useful distinctions might be drawn. See, e.g., Elster, Norms of Revenge,
supra note 26, at 863-66; Philip Pettit, Virtus Normativa: Rational Choice Perspectives, 100
Ethics 725, 728-32 (1990); see also H.L.A. Hart, The Concept of Law 82-91 (2d ed. 1994)
(discussing rules of obligation, a subclass of norms); David K. Lewis, Convention 99-100
(1969) (dealing with the relation between conventions and norms).

914 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

“It isn’t done” is a frequent reaction to certain disapproved con-
duct37-even though the relevant “it” is indeed done. The governing at-
titudes span an exceptionally wide range. They may or may not begin or
maintain themselves as a result of reflective judgments about, for exam-

ple, fairness and utility.38 They may be based on simple, intuitive judg-
ments about what is healthy behavior; these judgments may eventually
turn into moral commitments unmoored from health judgments. Social
norms may or may not promote liberty and well-being; they may or may
not be easily malleable or go very deep into people’s understandings.
Sometimes norms are codified in law.39 This is true, for example, with
norms governing littering, respecting private property, and banning dis-
crimination on the basis of race or sex.

Whether or not law is also at work, social norms are enforced

through social sanctions that are, to say the least, pervasive. “Political
correctness” is no isolated phenomenon. It is ubiquitous.40 It occurs
whenever reputational incentives impose high costs on deviant behavior.
The relevant sanctions create a range of unpleasant (but sometimes

pleasant) emotional states in people who have violated norms. If some-
one behaves in a way inconsistent with social norms, public disapproval
may produce embarrassment or perhaps shame and a desire to hide.
Sometimes the unpleasant feelings brought about by violations of social
norms are intense, and the social consequences of these feelings, and
(perhaps even more) of anticipating them, can be substantial.4″

37. See Elster, The Cement of Society, supra note 26, at 132-33.
38. I do not offer an account here of the emergence of norms. For relevant

discussion, see Elster, The Cement of Society, supra note 26; Allan Gibbard, Wise Choices,
Apt Feelings: A Theory of Normative Judgment (1990); Cristina Bicchieri, Norms of
Cooperation, 100 Ethics 838 (1990). Nor do I discuss the relationship between social
norms and moral judgments about, for example, fairness. Social norms are sometimes
supported by such judgments, but sometimes not. Social norms are sometimes a product
of such judgments, but sometimes not. On the relation between behavior and judgments
about fairness-more precisely, in my view, social norms rooted in judgments of fairness-
see Daniel Kahneman et al., Fairness and the Assumptions of Economics, in Richard H.
Thaler, Quasi Rational Economics 220 (1991).

39. For discussions of law, norms, and insurance, see Zelizer, supra note 30, at
113-37; Zelizer, supra note 27, at 27-39.

40. In fact the term itself is an exercise in norm management. It brands people with
certain views as weak followers of convention, rather than people who are following moral
convictions of their own. If one believes that one’s own views are “politically correct,” one
is likely to feel a bit embarrassed by them; if one thinks that one is avoiding “political
correctness,” one is likely to be proud of one’s independence and fortitude. Consider the
suggestion that “[a] lot of young women don’t want to be called feminists because, hey,
listen to Rush Limbaugh, and you’ve heard it all. It’s equated with being lesbian, fat, ugly.”
Karen De Witt, Feminists Gather to Affirm Relevancy of Their Movement, N.Y. Times, Feb.
3, 1996, at 9.

41. The persistent urge to conform to social norms has been demonstrated in a good
deal of work in social psychology. The classic study is S.E. Asch, Effects of Group Pressure
upon the Modification and Distortion ofJudgments, in Groups, Leadership, and Men 177
(Harold Guetzkow ed., 1951).

915 1996]

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

Much socially desirable behavior is attributable to social norms. For
example, lack of education produces a good deal of shame in certain
communities; so too with the decision to be promiscuous, or to use alco-
hol, cigarettes, or unlawful drugs. The result can be salutary incentives.
But social norms can produce undesirable incentives as well. Thus, in
some communities, norms encourage a lack of education, promiscuity, or
abuse of alcohol or unlawful drugs.

2. Intrinsic value, reputational value, and self-conception. – From these

points we might conclude that choice among options is a function not

only of (a) the intrinsic value of the option-a book, a job, a drink-but
also of (b) the reputational benefit or cost of the choice and of (c) the
effects of the choice on one’s self-conception.42 The intrinsic value refers
to whether, apart from reputational effects or effects on the agent’s self-

conception, the option is fun, illuminating, pleasant, interesting, and so
forth. You may watch a television show on public broadcasting not only
because it is fun or illuminating, but also because there are reputational
advantages from doing so and advantages as well from the standpoint of

enhancing your self-conception. Someone may buy a certain book (say, a
book about Shakespeare) and not another book (say, a book by Stephen
King) largely because she wants to think of herself as the sort of person
who reads about Shakespeare. Someone may buy and wear an expensive
piece of clothing not only or not mostly because he thinks it looks nice,
but also or mostly because he wants other people to see him wearing that

piece of clothing. Social norms are a key determinant in reputational
benefit or cost. They can much affect (though they need not determine)
self-conception as well.

Changes in social norms can influence choices if intrinsic value is
held constant, by altering the effects of reputational incentives and conse-

quences for self-conception.43 If littering produces shame or disap-
proval, behavior will shift; if race discrimination produces reputational
cost rather than benefit, fewer people will discriminate on the basis of
race; if parenthood out of wedlock is stigmatized, there should be less

parenthood out of wedlock.44 Indeed, obedience of law is built in large

42. Cf. Kuran, supra note 17, at 24-38, which has much influenced my presentation
here. Kuran discusses reputational and intrinsic utility; he also refers to what he calls
“expressive utility,” by which he means the expressive value of the act. See id. at 30-35.
Some people may, for example, want to deviate from existing norms because of the
expressive value of deviation. But I think it is better to speak of effects on the agent’s self-
conception; this is a more general idea that includes, but is not limited to, what Kuran calls
expressive value.

43. See id. at 35-38.
44. See Elijah Anderson, Streetwise: Race, Class, and Change in an Urban

Community 112-37 (1990) (describing impact of social norms on patterns of childbirth
and parenting).

916 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

1996] SOCIAL NORMS AND SOCIAL ROLES 917

part on the perceived reputational consequences of law violation.45 Of
course those consequences might be favorable rather than unfavorable.

The three-part division is a bit crude, since perceptions of intrinsic
value will be often a function of social norms. Those perceptions do not
exist in a vacuum. Perceived intrinsic value is emphatically a result of
social forces-consider responses to opera, rock music, rap-and reputa-
tional pressures can interact in complex ways with beliefs about intrinsic
value. People’s self-conceptions are very divergent, and each of our self-

conceptions has many dimensions; for example, many of us may want not
to be conformists, but also want not to diverge too much from what other

people do and think. I offer the three-part division as a place to start.

3. Norms and freedom. – In a way social norms reduce freedom, un-
derstood very broadly as the power to do whatever one would like to do.46
Certainly norms stop people from doing things that (if the norms were
different) they would like to do, and people sometimes would like the
norms to change. In particular, norms can drive a wedge between peo-
ple’s public actions or statements and their private judgments or
desires.47 Some people very much want norms to be something other
than what they are, and they regret the fact (if they recognize it) that they
have no power to change them. Thus a teenager might lament the exist-
ence of social norms in favor of (say) carrying a firearm or using drugs;
but while the norms are in place, he might well carry a firearm and use
drugs, and in a sense feel compelled to do so.

It would, however, be quite ludicrous to deplore social norms, to see
them only as constraints on freedom, or to wish for them to disappear. In
fact norms make freedom possible. Social life is not feasible-not even
imaginable-without them.48 In the absence of social norms, we would
be unable to understand one another.49 Norms establish conventions
about the meanings of actions. Social norms are thus facilitative as well as

constraining.50 If everyone knows the norms concerning a raised voice or
wearing bluejeans, then people can raise their voices or wear bluejeans
without having to decide what these actions mean.

45. See Tom R. Tyler, Why People Obey the Law 45 (1990) (showing a significant
correlation between fear of peer disapproval and compliance).

46. I do not mean to endorse this odd conception of freedom. Many unobjectionable
things-like speed limit laws or high prices-reduce freedom, thus understood. I seek
only to draw attention to the fact that norms can constrain behavior and choice even
though some or many people would like them to be otherwise.

47. See Kuran, supra note 17, at 3-12.
48. This is a theme of Elster, The Cement of Society, supra note 26.
49. Hence cross-cultural understandings are sometimes made difficult by the fact that

social norms are different in different cultures, so that meanings have to be translated, and
people may be unaware of that fact. Consider the example of whistling at sporting events:
In America, whistling connotes approval; in Europe, it is a form of “booing.”

50. Cf. Holmes, supra note 22 (arguing that constitutional constraints improve
governance).

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

There is a further and, for present purposes, an especially important
point. Good social norms solve collective action problems by encourag-
ing people to do useful things that they would not do without the rele-
vant norms.5′ Consider voting, littering, behaving courteously, keeping
promises, cleaning up after one’s dog, writing tenure letters, and doing
one’s share of administrative work. Without social norms, coercion or
economic incentives-perhaps with large financial investments-would
be required to solve collective action problems. Norms can be an espe-
cially cheap way of ensuring against the unfortunate consequences of
prisoner’s dilemmas.

On the other hand, social norms may be ineffective. This is so partly
because some people like to incur the disapproval that follows norm-viola-
tion, and hence some people like to “flout convention” by rejecting pre-
vailing norms by, for example, smoking, playing loud music in public, or

wearing unusual clothes. Of course people who violate generally held
social norms might be behaving consistently with particular norms in a
relevant subculture. This is true when people in a certain group wear

clothing of a certain kind, or when people who smoke cigarettes receive

peer group approval simply by virtue of the fact that they are violating
more broadly held norms. (Hence those who reject generally held norms

may be the most committed of conformists; they are following the norms
of a subcommunity, as when teenagers choose a form of dress that vio-
lates generally held norms but imposes a rigid orthodoxy on the

subgroup.)
There are many possible reasons for rejecting prevailing norms.

Some people depart from the prevailing norm because of their reflective
judgments. Such people think, on reflection, that the norm is too silly or
too unworthy to affect behavior, or that relevant roles diminish autonomy
or well-being. Marrying someone of a different race may reflect thisjudg-

51. See Edna Ullmann-Margalit, The Emergence of Norms 18-133 (1977). For an
important discussion of how norms produce social order, and solve collective action
problems, in the absence of legal constraints, see Robert C. Ellickson, Order Without Law
(1991). But norm changes need not produce Pareto improvements; there can be losers as
well as winners even in the face of a solution to a collective action problem, as when, for
example, some members of a small group reject a ban on littering. Moreover, there is a
crucial question about which norms are taken as given, and which are put up for grabs, in
the sort of analysis that celebrates certain norms as solving collective action problems. See
infra Part VI.A. For an illuminating discussion of why norms might be inefficient, see Eric
A. Posner, Law, Economics, and Inefficient Norms, 144 U. Pa. L. Rev. (forthcoming 1996).

A growing literature discusses the place of norms in ordering behavior entirely outside
of law. See, e.g., Lisa Bernstein, Opting Out of the Legal System: Extralegal Contractual
Relations in the Diamond Industry, 21 J. Legal Stud. 115 (1992); Robert D. Cooter,
Structural Adjudication and the New Law Merchant: A Model of Decentralized Law, 14
Int’l Rev. L. & Econ. 215 (1994). This literature raises interesting questions about the
extralegal development of norms and the extent to which law should build on those
norms.

918 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

ment; sharing domestic labor on an equal basis almost certainly does.52
In other cases, as discussed above, the departure simply expresses defiance,
and the real desire is to flout convention, whatever the norm is.53 Many
apparently odd practices involving dress and manners are rooted in this
phenomenon; some people find defiance an intrinsic good, and what
they are defying is more or less incidental.54 In still other cases, the de-
parture is the expression of an individual desire or taste, which the person
would pursue whether or not it is inconsistent with social roles and ac-
companying norms. Consider the view that Coca-Cola actually is better
than all other drinks, a view that might be reflected in unconventional
drink selections in many imaginable places.

The fact that some people like to reject social norms is highly rele-
vant to law. For example, a serious problem with legal efforts to inculcate
social norms is that the source of the effort may be disqualifying. Such
efforts may be futile or even counterproductive. If Nancy Reagan tells
teenagers to “just say no” to drugs, many teenagers may think that it is
very good to say “yes.”55 It is said that propaganda efforts in the former
Soviet Union failed simply because the source of the propaganda was not
trusted; hence the government’s effort to inculcate norms of its choosing
fell on deaf ears.56 These points bear on the regulation of social risks-a
principal concern of this article-particularly in the areas of teenage
smoking and potentially dangerous sexual activity. Efforts by private or
public authorities to stigmatize certain acts may have the opposite effect.

4. Exit and entry. – The fact that norms are contested within a heter-

ogeneous society can lead to the creation of many diverse norm communi-

52. See Susan M. Okin, Justice, Gender, and the Family 149-55 (1989) (discussing
gender inequality in domestic labor).

53. Consider the scene in The Wild One in which Marlon Brando, teenage rebel, is
asked, “Johnny, what are you rebelling against?” and responds, “Whaddaya got?” described
by David Hinckley, The “W” Word, Chi. Trib., Sept. 6, 1990, at 17C.

54. See James Miller, “Democracy is in the Streets”: From Port Huron to the Siege of
Chicago 59 (1987) (reporting revealing statements made by Tom Hayden describing some
1960s civil rights leaders as seeking to make a personal statement rather than to protect
civil rights for their own sake).

55. Compare President Clinton’s remarks on teenage smoking, which are very much
in the spirit of what I am suggesting here:

[I]t’s simply not true that cultural changes and legal bars, together, cannot work
to reduce consumption …. If you make it clearly illegal, more inaccessible, you
reduce the lure of advertising, and then you have an affirmative campaign, a
positive campaign, so that you don’t say, “Just Say No.” You give young people
information and you make it the smart, the cool, the hip thing to do to take care
of yourself and keep yourself healthy and alive.

Excerpts from Clinton News Conference on His Tobacco Order, supra note 6, at A18.
56. Hence a great deal of attention has been paid in recent years to the role of trust in

regulatory policy. See Paul Slovic, Perception of Risk: Reflections on the Psychometric
Paradigm, in Social Theories of Risk 117, 151-52 (Sheldon Krimsky & Dominic Golding
eds., 1992); Richard H. Pildes & Cass R. Sunstein, Reinventing the Regulatory State, 62 U.
Chi. L. Rev. 1, 40-43, 107-12 (1995); Chauncey Starr, Risk Management, Assessment, and
Acceptability, 5 Risk Analysis 97 (1985).

1996] 919

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

ties. People who are dissatisfied with prevailing norms can vote with their
feet, using the power of “exit” to leave norm communities they dislike
and to enter into groups built on more congenial norms. They may even
create, or seek to enter, self-consciously deviant subcommunities. Many
American high schools reflect this phenomenon, as students find groups
that are defined in a relatively crisp way, and as groups intermingle only
on occasion. The most conspicuous norm communities are private; but
in a system with state and local governments and free mobility, there are
public norm communities too, as people sort themselves by choosing lo-
calities with congenial norms. The norms of such communities may be
reflected in supportive state or local laws in which, for example, homo-
sexual relations are permitted, or traditional morality is enforced. And
once sorting has begun, processes of “voice” can increase the intensity
and uniformity of governing norms. Homogenous local communities
may insist on their own norms. National rights are often a safeguard
against the possibly injurious character of this process.

To the extent that people can freely enter and exit norm communi-
ties, such communities offer important protection against oppressive
norms. On the other hand, it can be very costly to exit from the norm
community in which one finds oneself, and the fact that one has been
raised in that community may make other options seem unthinkable,57
even though they might be much better.58

5. Norms and nonfungible cash. – Because of prevailing social norms
within relevant communities, money itself is not fungible.59 The point
deserves separate discussion, for it has particular importance for law and
legal policy, and it has not received the attention it deserves.

Consider, for example, the fact that money is often compartmental-
ized.60 Some money is specially reserved for the support of children.
Some money is for gifts. Some is for one’s own special fun. Some money
is for summer vacation. Some money is for a rainy day. Thus a study of
practices in Orange County, California, says that residents keep

a variety of domestic “cash stashes”-“generally one in the bill-
fold of each adult, children’s allowances and piggy banks, a
‘petty cash’ fund in a teapot-equivalent, a dish of change for
parking meters or laundry”-or “banked stashes of money,” in-
cluding Christmas club savings and accounts designated for spe-

57. See Kuran, supra note 17, at 176-95 (describing “the unthinkable and the
unthought”).

58. Apparently dissident communities may be based on norms that are highly reactive
to and even defined by reference to more generally held norms; it might be better if the
community as a whole could do something about those norms. These points bear on the
legitimate role of government.

59. See generally Viviana A. Zelizer, The Social Meaning of Money (1994) (proposing
differentiated model of money that recognizes certain monies as nonfungible,
nonportable, subjective, and heterogeneous).

60. See Richard H. Thaler, Mental Accounting Matters (Dec. 1993) (unpublished
manuscript, on file with the Columbia Law Review).

920 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

cial expenditures such as property or other taxes, vacations, or
home and car insurance payments.61

Social theorists have often feared that the use of money would “flatten”
social life, by erasing qualitative distinctions among goods. Ironically,
however, social life, pervaded as it is by social norms, has “unflattened”
money by insisting on qualitative distinctions.62 “There is no single, uni-
form, generalized money, but multiple monies: people earmark different
currencies for many or perhaps all types of social interactions …. And

people will in fact respond with anger, shock, or ridicule to the ‘misuse’
of monies for the wrong circumstances or social relations … .”63

The existence of norms involving money suggests that it may not be
possible to infer global judgments from particular consumption choices
that are dependent on context-specific norms. Of course a wide range of
norms ban the use of dollars as a reason for action, and sometimes law
fortifies such norms; consider prohibitions on prostitution, vote-trading,
and surrogate parenthood.64 Here, however, I am concerned with norms
that “subsidize” or “tax” certain uses of money. The point is important
because economists often claim that particular choices demonstrate or
reveal general valuations.65 But this may be a mistake, because particular
choices are a function of social norms that may be limited to the particu-
lar context. In a different context, the governing norms may be quite
different. Thus, for example, the refusal to insure certain goods may be a
product of social norms that are limited to the particular context of insur-
ance. If so, it may be wrong to draw from insurance choices general con-
clusions about (say) the appropriate domain of tort law.66

B. Social Roles

1. In general. – Many norms are intensely role-specific, and law is
often self-consciously concerned with social roles. Because the subject of
roles raises distinctive issues, I now discuss some of the relationships
among roles, norms, and law.

Consider the following social roles: doctor, employee, waiter, law
school dean, wife, friend, pet-owner, colleague, student. Each of these
roles is accompanied by a remarkably complex network of appropriate
norms. The network is not easily reduced to rules, but people know,

61. Zelizer, supra note 59, at 5 (citing Jean Lave, Cognition in Practice 132-33
(1988)).

62. For an impressive argument to this effect, see Thaler, supra note 60.
63. Zelizer, supra note 59, at 18-19.
64. See Cass R. Sunstein, On the Expressive Function of Law, 144 U. Pa. L. Rev.

(forthcoming 1996).
65. See, e.g., W. Kip Viscusi, Fatal Tradeoffs: Public and Private Responsibilities for

Risk 34-74 (1992).
66. See Steven P. Croley & Jon D. Hanson, The Nonpecuniary Costs of Accidents:

Pain-and-Suffering Damages in Tort Law, 108 Harv. L. Rev. 1787, 1851-53 (1995). For a
more detailed discussion, see Sunstein, supra note 64.

1996] 921

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

often very well, what they are.67 If you are a waiter, and treat your restau-
rant’s patrons the way you treat your friends, you probably will not be a
waiter for very long (except perhaps in California). If you are a student,
and treat a teacher as if he were your employee at the local factory, you
will be perceived as misbehaving very badly. If you treat friends the way
doctors treat patients, or lawyers treat clients, you probably won’t have

many friends. People rapidly internalize social norms about what their
roles entail. Violations of role-specific norms can seem jarring and pro-
duce prompt social punishment (or reward).

Roles are accompanied by a wide range of included and excluded
reasons for action.68 These may or may not be a product of law; they are

certainly a product of social norms. In your capacity as lawyer, you can
act only on the basis of certain reasons. For example, you may reveal

something told to you in confidence only to prevent a crime; you cannot
breach a confidence on the ground that it would be economically profit-
able to do so. In your capacity as judge, you may look only at a restricted
set of considerations, a set far more restricted than those you may ex-
amine if you are a legislator. In your capacity as friend, you are not per-
mitted to violate a confidence on the ground that it would be fun to

gossip about what you have been told. Confusion of roles-is X speaking
as a friend or as a colleague? is the judge a closet legislator? what exactly
is my relationship to my employer?-can cause uncertainty, awkwardness,
or much worse.

2. Roles and freedom. – Are social roles an obstacle to freedom?69 In
a way the answer is yes, since people often would like to do things that
their role forbids, and since people often would like to change the nature
of their roles. But this would be a far too simple conclusion. Without
roles, life would be very hard to negotiate. Like social norms, social roles
are facilitating as well as constraining. The existence of a clear role
makes communication much easier. By sharply constraining the domain
of permissible actions, roles simplify things and, in that sense, increase

personal freedom.
Of course some of the norms associated with certain social roles are

silly or even oppressive, and some people deplore them for this reason.
What can they do? Large-scale changes in social definitions of roles nor-

mally require collective action, whether private or public-a point with
considerable importance for those interested in the appropriate domain
of law. But sometimes individual people act in ways inconsistent with
their roles precisely in order to draw attention to their silly or oppressive
character. Thus a slave in the pre-civil War South might decide not to act

67. Cf. John R Searle, The Construction of Social Reality 127-47 (1995) (describing
the “background” knowledge which is necessary to constitute a role as I understand it
here).

68. Cf. Joseph Raz, Practical Reason and Norms 35-48 (1975) (discussing
exclusionary reasons).

69. The same qualification is necessary here as in supra note 46.

922 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

deferentially;70 a student might raise his voice against an abusive teacher;
a woman in an unequal society might insist that domestic labor be shared;
a homosexual man might “flaunt it”; a teacher in business school might
wear bluejeans. In fact attacks on norms associated with particular roles
often have a great deal to do with perceptions of injustice.

Many roles are ascriptive and not chosen, even in post-feudal socie-
ties. We cannot fully control the roles in which we find ourselves. To be
sure, people have power to assume or not to assume some roles. You can
decide whether to be a spouse, a parent, a teacher, a dean, and so forth.
And within limits, you can decide what it means to be any of these things;
people certainly can alter the roles associated with parent, wife, and hus-
band, even if they cannot do a great deal about the meanings associated
with their choices. But many roles are assigned rather than voluntarily
assumed-child, man, African-American, old person, short person, and
more. A role that is assigned might be described as a status, a distinctive
kind of role that, if surrounded by objectionable norms, raises special
problems.71 And many roles cannot easily be rejected in most societies-
driver, employee, student, citizen, family member.

3. Roles and law. – Prevailing roles and norms can be fortified by
legal requirements; they may even owe their existence to law. There are
many specific legal provisions for people occupying different roles-par-
ents, spouses, employers, employees, home-owners, nuclear power plant
operators, animal owners, doctors, stock brokers, landlords, automobile
sellers, and others. Law can help constitute roles. Much of the law relat-

ing to families, employer-employee relations, and professional obliga-
tions (lawyers, doctors, architects, and others) has this feature. In fact
many roles seem “natural” even though they owe their origin to social
and even legal conventions.72

Often law tries to redefine roles. In recent years, this has happened
with respect to the roles of employee, husband, father, disabled person,
and judge. Thus, for example, the law has said that husbands may not

rape their wives; that absent fathers owe duties of support to their chil-
dren; that disabled people have certain rights of access to the workplace.
All of these measures can be seen as attempts to create new or better
norms to define the relevant roles.

4. Citizens and consumers. – Of course each of us occupies many dif-
ferent roles, and there is much to be said about the constraints imposed

70. Cf. infra note 226.
71. See infra Part VI.D. An illuminating discussion in the context of gender is Mary

Anne C. Case, Disaggregating Gender from Sex and Sexual Orientation: The Effeminate
Man in The Law and Feminist Jurisprudence, 105 Yale LJ. 1 (1995).

72. See Lessig, supra note 24, at 949-51 (discussing constructed character of social
meanings); cf. Elster, Norms of Revenge, supra note 26, at 871-72 (The urge for
vengeance “is not a spontaneous, presocial anger directed at another person (or, for that
matter, a material object) that has harmed or hurt us. The passion for revenge is
embedded in a way of life that revolves around the notion of honor, and in which ‘boiling
blood’ is a socially recognized category.”).

1996] 923

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

by these diverse roles. But for present purposes, an especially important
and pervasive difference involves the relationship between citizen and
consumer.

Return to John Jones in the fifth tale.73 The example shows that in

your capacity as a citizen, you might urge a result-with respect to (say)
the duties of polluters or commercial broadcasters-that is quite differ-
ent from what you seek through your market behavior in your capacity as
a consumer.74 Acting as citizens, many people try to change social prac-
tices, and they often try to do this by changing social norms associated
with a particular role. Sometimes these efforts are a function of the role
of citizen and associated norms. Sometimes these efforts are undertaken
via law. In their private capacity-as consumers, employers, or family
members-people may do something that they believe, on balance, to be

unjust, and as citizens, they may support measures that better reflect their
convictions.75 Sometimes efforts to change norms and roles reflect an

understanding that human beings are selfish or have weakness of will and
that measures should be taken to ensure behavior that, on reflection, we
would like to follow.76

In addition, citizens do or say things just because of existing social
norms, which impose sanctions on publicly expressed dissident behavior
or judgments; in their private capacity, people may be freer to do or say
as they (in a sense, and subject to the prevailing norms) wish. In all cases
the difference is connected to the fact that a citizen is helping to make a

judgment not simply for himself but for a collectivity.77 In this sense
there are important contextual differences between market behavior and

voting behavior. The former does not affect the collectivity in the same

way, and hence those concerned, for example, with protecting the envi-
ronment may believe that their own behavior is largely irrelevant, whereas

73. See supra Part I.E.
74. See Mark Sagoff, The Economy of the Earth 7-28 (1988) (describing the

distinction between citizen and consumer).
75. This is the central topic of a poem by Vachel Lindsay, “Why I Voted the Socialist

Ticket.” Unfortunately, the poem is truly horrible:
I am unjust, but I can strive for justice.
My life’s unkind, but I can vote for kindness.
I, the unloving, say life should be lovely.
I, that am blind, cry out against my blindness.

Vachel Lindsay, Why I Voted the Socialist Ticket, in 1 The Poetry of Vachel Lindsay 64
(Dennis Camp ed., 1984). Of course it is possible, too, that in their private behavior
people will be unselfish, whereas people might be motivated by greed in their public
arena. I am discussing possibilities only, not making general suggestions about the realms
in which people are most likely to be motivated by norms of fairness.

76. SeeJon Elster, Ulysses and the Sirens 36-111 (1979); Holmes, supra note 22, at
134-77.

77. People appear to behave accordingly even though a single vote probably matters
no more than a single market decision.

924 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

1996] SOCIAL NORMS AND SOCIAL ROLES 925

laws can make a great deal of difference.78 Largely for this reason, the
role of citizen is accompanied by norms that can discourage selfishness
and encourage attention to the public good.79

In fact many efforts to change law are at least partly an outgrowth of
the difference between citizens and consumers.80 Consider laws outlaw-

ing sexual harassment, providing incentives to share domestic labor,81 or

granting workers a right to unionize. It should be clear that in such cases,
there is no simple relationship among choices, preferences, norms, and
roles. There may be conflict or tension between two or more of these.

C. Social Meanings

Social norms help people assign “social meaning” to human behav-
ior. With this term I refer to the expressive dimension of conduct (not
excluding speech) in the relevant community.82 Social meaning is a

product of social norms.83
The expressive dimension of conduct is, very simply, the attitudes and

commitments that the conduct signals. A complex body of First Amendment
doctrine deals with the problem of “expressive conduct,” that is, acts that

carry an expressive purpose and effect, such as flag-burning and draft

card-burning.84 But most conduct (including words) has an expressive
function-not in the sense that the actor necessarily intends to communi-
cate a message, but in the sense that people will take the conduct to be

expressing certain attitudes and commitments.85 Advertisers are, of
course, well aware of this point, and much advertising is an effort to affect
the social meaning of a product or of a purchasing decision.86 Thus ad-

78. The very idea that I treat the prevention of environmental damage just like
buying a private good is itself quite absurd. The amount I am ready to pay for my
toothpaste is typically not affected by the amount you pay for yours. But it would
be amazing if the payment I am ready to make to save nature is totally
independent of what others are ready to pay for it, since it is specifically a social
concern. The ‘lone ranger’ model of environmental evaluation confounds the
nature of the problem at hand.

Amartya Sen, Environmental Evaluation and Social Choice: Contingent Valuation and the
Market Analogy, 46Japanese Econ. Rev. 23, 29 (1995).

79. I am describing a possibility, not a certainty. Political behavior is very often selfish.
80. See Sen, supra note 78, at 26-31, 34 (distinguishing consumer willingness to pay

from citizen judgments).
81. See Okin, supra note 52, at 149-55.
82. See Lessig, supra note 24, at 949-58.
83. Cf. Thomas C. Schelling, Micromotives and Macrobehavior 213-14 (1978)

(explaining how hockey players’s reluctance to wear helmets is a product of social norms);
Lessig, supra note 24, at 1014-16 (discussing factors that control or manage social
meaning).

84. See generally Geoffrey R. Stone et al., Constitutional Law 1101-1403 (2d ed.
1991).

85. See Lessig, supra note 24, at 951-55.
86. Consider an advertisement in the Atlanta airport, “Remember those nerds you

hated in high school? Now you can hire them.” The advertisement is for an accounting
firm. Its goal is to change the social meaning of a hire, so that people will not feel that they

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

vertisements often focus not on the intrinsic value of the product, but
instead on its reputational value and (equally important) its conse-
quences for self-conception-making the purchaser seem smart, sophisti-
cated, and in control of the situation. Political advertisements can have
similar functions. They often attempt to alter the social meaning of a
vote by making the opponent’s supporters seem like whiners or dupes
and voters for the advertised candidate seem decent and good.

The meaning of acts is very much a function of context and culture.
Consider some examples. If I light up a cigarette, I will, in certain parts
of the United States, be signalling something relatively precise and very
bad about myself, my self-conception, and my concern for others. In
other parts of the United States, the signals are very different. In France,
a smoker gives still different signals. If you fail to attend church-or if
you do attend church, and tell everyone about it87-your act will have
particular meanings, and these have everything to do with the community
in which you find yourself. If I decide not to get married, or not to have
children, my act will convey a restricted range of possible meanings, and I
will not have much control over those meanings. (If I were a woman, my
decisions to this effect would have a quite different set of meanings. The
meaning of a woman’s not marrying or having children is quite different
from a man’s.) If you buy insurance on your children’s lives, you will be
signalling something about your conception of your child, and in some
times and places, the signal is very bad.88

Language also has social meanings, extending far beyond the words
themselves and reflected in the attitudes and commitments signalled by
how people talk. Context determines those meanings. The words, ‘You
look great today,” can have many different possible social meanings.
Consider their use from a mother to a fifteen-year-old daughter, from a
male employer to a female employee, from a doctor to a convalescent
patient, from a homosexual male student to a male classmate. If you re-
fer to women as “ladies,” you are also making (whatever your intentions)
a certain set of statements about yourself and about your views on gender
issues. A description of certain Americans as “blacks” will have a different
meaning in 1996 (after the rise of the term “African-American”) from
what it was in (say) 1976.

As with social norms and social roles, the social meanings of acts are
something about which individuals can do relatively little (most of the
time).89 If a lawyer drives a Harley-Davidson motorcycle to Wall Street,
his own attitude toward his act will have little relation to what other peo-

are confused and need a superior mind, but instead feel superior in their capacity to hire a
servant.

87. See Stephen L. Carter, The Culture of Disbelief: How American Law and Politics
Trivialize Religious Devotion (1993), which can be understood as an attack on, and an
effort to alter, the social meaning of being religious in certain parts of the United States.

88. See Zelizer, supra note 30, at 121-24.
89. See Lessig, supra note 24, at 993-1007 (exploring the collective action problem).

926 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

pie take his act to mean. If a European visitor to the United States ad-
dresses women as “Miss” or “Mrs.,” the social meaning of his choice will
have little to do with his intentions (unless, perhaps, people know that he
is a visitor). If a nonsmoker asks someone not to smoke, the social mean-
ing of the act will be quite different in New York in 1996 from what it was
in the same city in 1966, and different as well from what it is in Germany
in 1996. This is a pervasive characteristic of social meanings. If you
buckle your seat belt in Boston, you will communicate no insult to the
driver; but things would have been very different twenty years ago, and
within certain subgroups the buckling of a belt still connotes cowardice
or accusation.90 In fact in California, it is a passenger’s failure to buckle
that is an act of disrespect to the driver, who, under California law, may
be held liable if you have not buckled.

To take a science fiction-ish example: If you lived in a society of
vegetarians, the act of eating meat-at, let us suppose, specially desig-
nated animal flesh restaurants-would be very different from what it is in
a society of meat-eaters.91 And if you lived in a society of vegetarians, you
might well choose not to eat meat, because social meaning would impose
a decisive cost on meat-eating. The meanings of actions are set by forces
that are emphatically human but that are largely outside of the control of
the individual agent.

The expressive dimension of action has everything to do with the
actor’s particular social role-with the way in which acts conform to, or
violate, expectations associated with the role. Because of the social mean-
ing of action, a lawyer acting as a lawyer may not make jokes that would
be perfectly acceptable with her family or friends. If a law school dean
wears shorts to teach contracts, or calls students by their first name, he
will be signalling something important and a bit radical, certainly at the
University of Chicago Law School; at a small college, the signals would be
altogether different. If a President endorses atheism, or if an American
judge speaks critically about the drafters of the American Constitution,92
people will be outraged, very much because of the role in which presi-
dents and judges find themselves. What can be said and done-in terms
of social meanings and social norms-is a function of roles.

On the other hand, there are contexts in which a person or a small
group of people may make inroads on social meanings. In a household, a
woman may be able to alter, a little or a lot, the social meaning to her
family of her refusal to do dishes or to make dinner, or her decision to go
out with colleagues at night. In a company, a single person or a small
group may be able to alter the social meaning of discourteous or aggres-
sive behavior formerly taken as natural or as good, or even as definitive of

90. See id. at 952.
91. I am grateful to Thomas Nagel for this example.
92. See Thurgood Marshall, Commentary, Reflections on the Bicentennial of the

United States Constitution, 101 Harv. L. Rev. 1 (1987); William B. Reynolds, Another View:
Our Magnificent Constitution, 40 Vand. L. Rev. 1343 (1987).

1996] 927

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

membership in a certain group. With communication over time, signifi-
cant changes may occur. But the most entrenched social meanings are-
by definition-not movable without concerted action on the part of
many people. Hence private groups often attempt to bring about

changes in meanings and the norms that produce them.93 Religious or-
ganizations, feminist groups, animal rights activists, and groups challeng-
ing “political correctness” are prominent recent examples. Often they
have been highly successful; sometimes they produce norm cascades.

D. Social Norms, Social Roles, and Social Meanings

What is the relationship among social norms, social roles, and social

meanings? As I have said, social norms determine the social meaning of
action. Social roles are similarly a product of social norms. But social
norms can also be an artifact of social meaning. Suppose that the social

meaning of condom use is a confession, an accusation,94 or a statement,
“I am not a spontaneous person.” If so, there will be a social norm dis-

couraging condom use.95 There are many public efforts to change this
norm.96 Social norms can also be produced by social roles. A university
teacher has a certain role, and the social norms governing acts by such a
teacher-use of a first name, informal clothing, references to popular
music-are an outgrowth of that role. Of course roles have no content

apart from social norms. Suppose that the social meaning of a male teen-

ager’s refusal to fight when insulted is, “I am cowardly.” The norm in
favor of fighting, in such circumstances, is a product of the role. There is
now a public effort to change this very norm.97

93. See Zelizer, supra note 27, at 41-89 (describing changing social meaning of life
insurance). Life insurance companies engaged in self-conscious efforts at norm
management, from the view that life should not be commodified or an object of gamble, to
the view that good fathers and husbands, at least, protect their families in this fashion:
“‘The necessity that exists for every head of family to make proper provision for the
sustenance of those dear to him after his death, is freely acknowledged and there is no
contingency whereby a man can stand excused from making such a provision.'” Id. at 56
(quoting Manhattan Life Insurance Co., Life Insurance 19 (1852)).

94. See Jennifer Steinhauer, At a Clinic, Young Men Talk of Sex, N.Y. Times, Sept. 6,
1995, at C7 (describing men’s failure to wear condoms as a product of a fear that such use
will be taken as evidence of infidelity).

95. See Elliot Aronson, The Social Animal 89-90 (6th ed. 1992); Lessig, supra note
24, at 1019-25.

96. See, e.g., Steinhauer, supra note 94 (describing efforts to encourage condom
use).

97. Thus there is an effort to publicize a new symbol: a fist covered by a hand. The
meaning of this symbol is: Though I am offended, I will not fight. The publicity campaign
is designed to change the meaning of a refusal to fight back.

928 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

E. Divisions in the Self98 and Norm Bandwagons

I have noted that social norms can make people act and talk publicly
in ways that are different from how they actually think, or from how they
act and talk privately. It is time to explore this point.

People often act in accordance with norms that they wish were other-
wise or even despise. Under the apartheid regime in South Africa, public
criticism of apartheid-at least within South Africa-much understated
private opposition to apartheid. The same was true for Communist re-
gimes.99 In a social group that punishes atheists or agnostics, few people
may confess their uncertainty about whether God exists; in a group of
atheists, few people may talk about their religious faith. Even in democ-
racies, the deterrent effect of social norms on acts and beliefs creates a

sharp disjunction between public acts (including speech) and private
thought.100 Hence a state of affairs may persist even though there is

widespread opposition to it. And eventually the norms may affect private
thought itself.’10 When, for example, social norms discourage pollution,
smoking, or acts of sexual harassment, people may come in their private
thoughts to see such things as unacceptable. Of course this might not

happen. But over the long run, behavior that is inconsistent with existing
social norms might not be defended publicly, and for this reason people
may come not to accept that behavior even in their most private
moments. 102

Political actors might be able to exploit private dissatisfaction with

existing norms in order to bring about large-scale social change. In fact

many political participants can be described as norm entrepreneurs; con-
sider Martin Luther King, Jr., William Bennett, Louis Farrakhan, Catha-
rine MacKinnon, Ronald Reagan, and Jerry Falwell. As we have seen, in-
dividuals who favor changes in norms face a free rider problem; norm

entrepreneurs can alert people to the existence of a shared complaint
and can suggest a collective solution. Thus political actors, whether pub-
lic or private, can exploit widespread dissatisfaction with existing norms

by (a) signalling their own commitment to change, (b) creating coali-
tions, (c) making defiance of the norms seem or be less costly, and (d)
making compliance with new norms seem or be more beneficial. We

might say that the intrinsic value of some option may be held constant,
but the reputational and self-conception values may shift dramatically.

98. Cf. Kuran, supra note 17, at 43-44 (discussing the “divided self”).
99. See id. at 118-27.
100. Thus the idea of a “silent majority” is used to allow people to voice their private

views by assuring them that there are many more people who share their views than
appears; this is an effort to reduce the perceived reputational costs associated with stating
seemingly unpopular views.

101. This is because of the desire to reduce cognitive dissonance, seeJon Elster, Sour
Grapes 109-40 (1983), or because the privately held thoughts disappear through lack of
public use, see Kuran, supra note 17, at 176-95.

102. For an intriguing discussion, see Kuran, supra note 17, at 176-95.

1996] 929

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

Thus there can be a “tipping point” when norms start to push in new
directions.

When the free rider problem begins to be solved, through reducing
the cost of acting inconsistently with prevailing norms, private thoughts
will be stated publicly, and things can shift very quickly. Something of
this sort happened in both South Africa and Eastern Europe, producing
more rapid and more peaceful changes than anyone anticipated.103 Part
of the reason is that hostility to the regimes was widespread and intense-
but inconsistent with existing social norms and hence much underesti-
mated. When the norms began to collapse, the regimes collapsed too.

The point bears on norm bandwagons. People may support an ex-
isting norm publicly not because they are genuinely committed to it, but
because they fear social sanctions. As I have said, there is a bandwagon
effect when those sanctions diminish or disappear, as many people join
the group opposing the existing norm and urging a new one. The result
can be astonishingly rapid change.104 An effect of this kind occurred
with two opposing and recent movements-the feminist movement and
the recent opposition to “political correctness” in the university.

F. Beliefs About Facts

Choices, meanings, roles, and norms are commonly based on beliefs
about relevant facts. Someone may believe, for example, that cigarette
smoking is not dangerous, and he may smoke partly for that reason. If he
really believed that smoking was dangerous, perhaps he would not
smoke.105 Choices are pervasively a function of beliefs. The same is true
for social norms. Consider the dramatic recent shifts in social norms gov-
erning smoking. Such norms have a great deal to do with prevailing be-
liefs about whether smoking causes harm to nonsmokers. When the be-
lief shifts, the norm shifts as well.106

Norms about behavior are interpenetrated with beliefs about harm
and risk. Thus many religiously-grounded norms about personal cleanli-
ness and hygiene owe their origins to beliefs about what is healthy; but
the norms often outstrip the beliefs and receive a kind of moral ground-
ing that is not simply reducible to an instrumental judgment about likely
risks. When someone violates a norm relating to hygiene, people’s reac-
tion is different-more stern and more deeply moralized-than it would
be if the reaction were based solely on the incremental increase in risk.

103. See id. at 261-88 (discussing fall of Communism).
104. See id. at 288 (“A specific law, regulation, policy, norm, or custom can be

abruptly abandoned when people who have helped sustain it suddenly discover a common
desire for change.”).

105. See Slovic, supra note 56, at 139 (discussing teenagers’s misperception of risks of
smoking).

106. Cf. Mary Douglas & Aaron Wildavsky, Risk and Culture (1982) (exploring
perceptions of and reactions to risk).

930 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

There are complex interactions between understandings of facts and
social roles. Certainly beliefs about facts help generate roles. Thus be-
liefs about natural differences between men and women, or blacks and
whites, affect social understandings about the appropriate roles of men
and women or blacks and whites. When people see that apparent differ-
ences between social groups are not grounded in fact, the roles associ-
ated with group members may shift accordingly. Thus attacks on claimed
natural differences have affected perceptions of appropriate roles. But
the process of change is often delayed because of the collective action

problems faced by people in changing norms that have outstripped the
beliefs on which they were originally based.

The converse is also true: Understandings of facts may be a function
of roles and accompanying norms. There are complex scientific litera-
tures on the differences between men and women; much of the relevant
work, even in its most scientistic forms, rests palpably on conceptions of
roles and of surrounding norms.’07 When the social role of a certain

group is to be laborers, or wives, it may be hard for most people to believe
that all people are (in some relevant sense) equal.

Judgments about fact are similarly entangled with social norms.
When most people smoke, it is hard for most people to believe that smok-

ing is dangerous; in particular, smokers may not want to believe that

smoking is dangerous. The norm affects the belief, just as the belief af-
fects the norm. In fact norms and judgments about risk are hard to sepa-
rate. When people are asked to judge risks in terms of their seriousness,
they respond not simply in terms of aggregate expected death rates but
also with reference to moral judgments, connected with prevailing social
norms, about the voluntariness of the risk, its distribution, its dreaded-
ness, its potentially catastrophic character, and so forth.108 And norms
and roles pervasively influence judgments about facts.109

III. CHOICES AND PREFERENCES

If we attend to the functions of norms, meanings, and roles, how will
we understand the relationship between choices and preferences? How
will we understand the notion of “preference” itself? My proposition here
is that attention to the functions of norms, meanings, and roles chal-

lenges our understanding of notions like “choice” and “preference” in
much of modern social science.

An initial and important problem is that the term “preference” is

quite ambiguous. Suppose someone takes a job as a welder, recycles
newspapers, or buys aspirin rather than chocolate bars. When we say that

107. See Helen E. Longino, Science as Social Knowledge: Values and Objectivity in
Scientific Inquiry 103-32 (1990).

108. See Slovic, supra note 56, at 120; Pildes & Sunstein, supra note 56, at 57-58.
109. See Kuran, supra note 17, at 176-95.

1996] 931

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

someone “prefers” to do as he chose, what exactly do we mean?110 There
are two major possibilities. First, preferences may be choices. Second,
preferences may be the mental states that stand behind choices. Atten-
tion to the place of norms, meanings, and roles complicates both of these
ideas; it suggests that for many purposes, it would be best to dispense with
the term “preference” altogether, and to work with more fine-grained
ideas. It follows that much of legal and political debate-about whether
the state should respect “preferences”-is based on a confusion stem-

ming from inadequate specification of what “preferences” are. The point
very much bears on the continuing puzzle of paternalistic government.

A. Preferences As Choices

The idea of a “preference” might be understood as simply a choice,
as in the idea, influential within economics from the work of Paul
Samuelson, of the “revealed preference.”111 On this view, preferences are
choices. This approach seems promising, because it makes it unneces-

sary to inquire into the mental states that accompany choices. Perhaps
we can work from behavior alone, and make predictions on the basis of
behavior alone. Perhaps behavior itself will show valuations that can be
used not only for positive but also for normative purposes.112 And there
is no doubt that it is illuminating to catalogue choices.

If, however, this is what we are doing, it is unnecessary and perhaps
misleading to use the notion of a “preference,” which seems to be in-
tended to explain or to back something called choices. If we are really
talking about choices, we can dispense with the idea of preferences en-

tirely. We will have a list of choices and should speak only in terms of that
list.

Perhaps this is merely a semantic quibble. Perhaps social theorists
can work with the list for positive or normative purposes. But if they are

really working just with choices, they will encounter many problems.
Choices are inarticulate, and hence imperfect predictors of behavior
without an account of what lies behind them.113 From the bare fact of

(particular) choices, it is not always possible to make robust claims about
future choices. This is because particular choices depend on the context
and on the norms, meanings, and roles prevailing in that context. When

110. The question is not adequately sorted out in the legal literature on both sides of
current debates about government respecting “preferences.” Thus-to take a not entirely
random example-Cass R. Sunstein, Legal Interference With Private Preferences, 53 U.
Chi. L. Rev. 1129 (1986), takes the term “preferences” as if it is clear; the analysis is
weakened by the failure to discuss the ambiguities in the term.

111. Paul A. Samuelson, Consumption Theory in Terms of Revealed Preference, 15
Economica 243, 243 (1948). I do not deal here with formal models of the sort described
and criticized in Hampton, supra note 23.

112. See, e.g., Viscusi, supra note 65.
113. See Aronson, supra note 95, at 187-89; Sen, Internal Consistency, supra note 23,

at 496-98; Amos Tversky & Itamar Simonson, Context-Dependent Preferences, 39 Mgmt.
Science 1179, 1187 (1993).

932 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

changes occur in the context, new norms may come into play. And be-
cause of norms, even the weakest axioms of revealed preference theory
can fail.114 John Jones, in the fifth tale,115 presents an illustration; his
particular choices do not allow observers to offer general predictions.
But for the present let us take a simpler example. IfJones prefers X over
Y, we might think that the mere introduction of an undesired third alter-
native, Z, ought not to change this preference. After all, Jones prefers X
to Y, and he would have to be an odd person to prefer Y to X simply
because of the introduction of Z.

But we can readily imagine cases in which the new alternative Z has
precisely this effect. Jones might, for example, always select the second
largest piece of cake, or he might want to be a person of relative modera-
tion. If Z is an especially large piece of cake, his preference for X (once
the second largest, now the third) over Y (once the largest, now the sec-
ond), will shift. Empirical work has encountered an effect called “ex-
tremeness aversion,” in which people make choices that avoid the ex-
tremes.116 Extremeness aversion is at least partly a product of social
norms. People are generally taught to avoid extremes, and people who
make extreme choices seem like malcontents, oddballs, or (never a word
of praise) extremists.117 There are many examples. A voter might, for
example, choose a Republican candidate over a Democratic candidate;
but the introduction of some third candidate (say, Ross Perot) may lead
him to choose the Democrat, because it makes some new characteristic
salient to voters,118 or because it shifts the outcome produced by the deci-
sion, making it moderate when it would otherwise be extreme. Barry
Goldwater’s famous slogan-“extremism in the defense of liberty is no
vice … and moderation in the pursuit of justice is no virtue”‘l9-was an
(unsuccessful) effort to alter the social meaning of a vote for Goldwater.

114. See Tversky & Simonson, supra note 113, at 1187-88. Perhaps the weakest of
these is the idea that if A prefers X to Y, the introduction of a third alternative, Z, ought
not to make A prefer Y to X.

115. See supra Part I.E.
116. See Tversky & Simonson, supra note 113, at 1183.
117. Cf. R.W. Apple Jr., Judge is Witness: His Burden of Proof, N.Y. Times, Sept. 24,

1987, at A26 (describing confirmation strategy of avoiding extremist-sounding testimony);
In re Bork: Ex-ChiefJustice, Law Professor and Carter Aide Present Views, N.Y. Times,
Sept. 24, 1987, at A20 (excerpt of Bork hearings addressing question of whether Bork
deserved the term “extremist”).

118. Cf. Jonathan W. Leland, Generalized Similarity Judgments: An Alternative
Explanation for Choice Anomalies, 9 J. Risk & Uncertainty 151, 169 (1994) (presenting
model of “generalized similarity judgments” that “predicts violations of tenets of
rationality”); Itamar Simonson & Amos Tversky, Choice in Context: Tradeoff Contrast and
Extremeness Aversion, 29 J. Marketing Res. 281, 289-92 (1992) (arguing that the
importance to consumers of certain product characteristics varies with the number of
product choices); Slovic, supra note 33, at 364 (discussing dimensions of choice).

119. William Safire, What’s an Extremist, N.Y. Times,Jan. 14, 1996, § 6 (Magazine), at
14.

1996] 933

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

In fact extremeness aversion is itself a contextual “taste,” one that
varies with both setting and governing norms. In a society or subculture
in which extremeness is prized, there may be no such aversion (assuming
the idea would be intelligible). All of us have seen subcultures in which
extremeness is a virtue (and does not quite count as extremeness). In
fact many people who are generally averse to extremeness are, in one
setting or another, quite willing or even delighted to be extreme; a per-
son who dresses in ordinary ways may be delighted to wear loud ties, or to
make odd dessert choices. This is partly because the social meaning of
extremeness in that context is different from what it usually is. It may, for
example, signal boldness and confidence rather than self-destructiveness
and peculiarity.

More broadly, the social meaning of action can encourage people to
make a certain choice, by taxing or subsidizing it; the meaning, and
hence the choice, very much depends on the context and the norms that
accompany it. If you are in a certain social group, you may well choose a
drink of brandy or wine over Coca-Cola simply because of local practices.
The choice of Coca-Cola may signal excessive informality, an unwilling-
ness to unwind and enjoy oneself, or even disrespect. But in a different
group, your choice may be different (and all this regardless of what you
would choose if you were in your house alone120). You may purchase an
American car, or display the flag on July 4, or engage in risky behavior
because of existing norms in your community. Perhaps your purchase of
a non-American car would signal a lack of patriotism; perhaps your fail-
ure to display the flag would be taken as a political protest whether or not
you meant it that way. If you run a local television station, your decision
whether to allow violent programming is very much a function of prevail-
ing norms, even if such programming would attract a large audience.

Role is, of course, an especially important determinant of choice. A
teacher may refuse an alcoholic drink at a party with students, just be-
cause he is a teacher; students may choose a certain kind of music just
because they are students and their teachers are there. A teacher’s will-
ingness to have a drink with students signals a host of characteristics with
which the teacher may not want to be associated. A student group’s deci-
sion to play rap music at a party with teachers may signal disrespect or
(what may be the same thing) excessive friendliness. Because of the im-
portance of context and surrounding norms, a choice of one good over
another may tell us little about further choices unless we know a lot about
the motivations and context of the choice.

These points suggest that to explain or predict behavior, it is impor-
tant not only to know about choices but also to have some account of
what underlies choices, or of what choices are for, and in this way to intro-

120. I do not suggest that that choice would be norm-free or that it should be taken as
definitive of what preferences “really” are.

934 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

duce an account of motivation.121 Here social norms, meanings, and
roles will be crucial. And if this is right, it is impossible to explain behav-
ior by reference to choices, without using the very apparatus that the “re-
vealed preference” idea was intended to eliminate.122

Normative arguments on the basis of choices alone123 will also run
into serious trouble. After finding that a person or group makes certain
choices, we may think we can infer what or how much that person or
group values, and from this make policy recommendations about, for ex-
ample, workplace safety or environmental protection. But this is hazard-
ous. Choices are a function of norms, which may be limited to certain
settings. Choices need not suggest acontextual valuation of social goods,
and thus even if we want to respect people’s valuations, we will have to
look not at, but behind, choices.

B. Preferences Behind Choices

Let us turn, then, to another, more promising conception of a “pref-
erence.” The term is often meant to refer not to choices themselves, but
to something that lies behind, and accounts for, choices. Gary Becker’s
work is in this tradition.124 The idea has obvious advantages. It seeks to
provide the motivational story behind choices, and if the motivational
story is uncovered (and sufficiently simple), positive work should be pos-
sible. And if we can identify what lies behind choices, perhaps we can get
a sense of people’s own conception of what promotes their well-being.
Knowledge of this conception is surely relevant for purposes of both eth-
ics and politics.125 Hence an emphasis on social norms should not be
seen as an attack on rational choice approaches to social and political
problems. From the standpoint of an individual agent, norms provide a
part of the background against which costs and benefits are assessed;
more specifically, they help identify some of the costs and benefits of ac-
tion. From the standpoint of the individual agent, this is hardly irra-
tional, and it is hardly inconsistent with self-interest. (Whether certain
norms are rational for society as a whole is a different question. Un-
doubtedly some of them are not.)

121. See Sen, Internal Consistency, supra note 23, at 496-98.
122. See Samuelson, supra note 111, at 243.
123. See Viscusi, supra note 65.
124. See, e.g., GeorgeJ. Stigler & Gary S. Becker, De Gustibus Non Est Disputandum,

67 Am. Econ. Rev. 76 (1977).
125. There are, however, problems with so using preferences. First, some preferences

are adaptive to an unjust status quo; when an unjust status quo is responsible for
preferences, the status quo cannot be justified by reference to the preferences. See Elster,
supra note 101, at 133-40. Second, the preference may well take the form it does for
highly contextual reasons, and should not be thought to be the agent’s preference in some
abstract or global way. In another context, the preference may take a different form. See
supra Part I.E. Third, the preference may be an artifact of social pressures directed against
public statements or acts of a certain kind. See Kuran, supra note 17, at 22-44.

1996] 935

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

I think that positive and normative work can and should operate in
this way. But the idea introduces difficulties of its own-indeed, the diffi-
culties that the “revealed preference” notion was intended to overcome.
These difficulties make the idea of “preference,” as the agent behind
choice, too crude. RecallJohnJones, the protagonist of the fifth tale.126
Can we provide an account of Jones’s motivation or “preferences”? No

simple answer would make sense. Several possibilities do present them-
selves. From the fact thatJones pays a certain premium for automobile

safety, we might judge that he is risk-averse-someone who prefers to
avoid danger-and we might even attempt to generate numbers captur-
ing his own conception of the value of his life.127 But when it comes to

bicycle-riding, Jones is somewhat reckless. And in his capacity as voter,
Jones’s valuations appear still more complex. To get a full account of his
motivation, we need to know many details-something like a personality
profile.

This point raises some larger issues. If we think of a preference as

something that lies behind choice, what is it exactly? Plainly it is a dispo-
sition or a mental state of some kind.128 And plainly people do have dis-

positions of various sorts. But internal mental states can be extraordi-

narily complex. People’s decisions are based on a complex of whims,
responses to norms, second-order preferences, aspirations, judgments,
emotions, drives of various kinds, conceptions of role, and more, with all
these producing particular results depending on the context. What lies
behind a choice in one setting may be quite different from what lies be-
hind a choice in a different time and place.

In this light, it can be hard to make predictions about individuals or

groups without knowing a great deal. Of course the value of positive
work lies in what the evidence shows. If predictions can be made with

simple accounts of preference, so much the better. Of course there are

regularities in people’s behavior, and these regularities can be connected
to people’s dispositions. But general dispositions of various kinds-to
avoid extremes, to comply with (or violate) norms, to fit with expecta-
tions associated with role, to drink beer rather than wine-manifest
themselves in particular choices only in accordance with context. No sim-

ple thing called a “preference” accounts for choice. Preferences are not
the building-blocks for a theory of decision; if positive work is to be done

very well, whatever we call a “preference” needs to be unpacked further.
Of course an excessively detailed account of the ingredients of “pref-

erence” may make predictions impossible. A standard that relies on too

many factors will not be a basis for social science. We might, then, begin
by distinguishing among intrinsic value, reputational effects, and effects
on self-conception, and by seeing whether changes in, for example,

126. See supra Part I.E.
127. This is the basic project in Viscusi, supra note 65, at 34-74.
128. Of course mental states have relations to physical entities. For a controversial

account, see Paul M. Churchland, The Engine of Reason, The Seat of the Soul (1995).

936 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

reputational effects induce changes in choice.129 This might well be a

productive direction in which social science will eventually move.

Shifting from positive to normative, we can see that the complexity
of mental states also makes it hard for governments to know how to re-

spond to people’s choices. Take an important and contested issue in de-

velopment policy. To what extent should a national government or inter-
national institutions take people’s “preferences” as given? Perhaps more

important: What does this idea even mean? For example, a poor woman
in India may in some sense not wish to be literate, especially if she faces
severe sanctions from trying to do so.’30 Because of those sanctions, she

may “prefer” not to go to school, or fail to “choose” to go to school, and
she may resist certain efforts to teach her to read. We might call these
her “conventional” preferences, and they are certainly real.

But in what sense is it helpful to say, in such a case, that she “prefers”
the status quo? She may also have certain aspirations-to have self-re-

spect, to be capable of doing a range of things, not to be wholly depen-
dent on others. We may call these her “fundamental” preferences (in the
sense that she probably will consider them deeper than her particular
choices), and they are real too. In these circumstances, her decision not
to learn to read is a function of social norms directly connected with her
role, and with the social meaning of her education for people in her

group. Norms and roles impose costs on choice, and as the situation has
been described, those costs are decisive for her. But on reflection she

may wish both roles and meanings to be different. Her general aspira-
tions may conflict with her more particular choices.

She may well see all of this already, or at least under appropriate
conditions. If she does, and if she is taught to read and delighted by that
fact, should we say that her preference has been changed? The answer

depends on how her preference is specified. In some ways the process of

learning is consistent with her deeper or more fundamental “prefer-
ences.” (Of course there are complex issues of interpretation in the de-

scription of some preferences as fundamental and others as conventional.

Any such description will have large evaluative dimensions.)
The example is merely illustrative. The same questions might be

asked about smoking, drug use, single parenthood, risky sexual activity,
recycling, and much more. Collective efforts to discourage damaging or

risky behavior, or to encourage norms that promote well-being or solve
collective action problems, might well be consistent with people’s under-

lying aspirations and judgments.
It may be tempting to think that in cases of this kind, there is a con-

flict between preferences and metapreferences, or between preferences

129. See Kuran, supra note 17, at 24-38.
130. See the accounts in Amartya Sen, Inequality Reexamined 117-28 (1994);

Amartya Sen, Commodities and Capabilities 73-83 (1985) [hereinafter Sen, Commodities
and Capabilities]; Martha Nussbaum, Aristotelian Social Democracy, in Liberalism and the
Good 203, 216 (R. Bruce Douglass et al. eds., 1990).

1996] 937

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

and preferences about preferences.’13 This is indeed an important phe-
nomenon, one that also complicates certain antipaternalistic arguments.
But it is different from what I am now describing. People have prefer-
ences about their preferences, but they also have preferences at different
levels of precision and generality.132 Sometimes people are indeed
forced to do things that (it makes sense to say) they do not wish to do.
But it is far from simple to identify a “preference” as something that
stands behind choices and explains them.

This is not to say that the idea is entirely unhelpful. For some pur-
poses it may aid analysis. We may be able to posit that a legislator “pre-
fers” to be reelected and we may use that preference for positive pur-
poses. Social scientists may be able to posit that most people prefer to
obey social norms most of the time, and make predictions accordingly.
Of course some people regularly like some foods better than others

merely on grounds of taste, and they choose accordingly. But these are
the simplest cases, and even here there are complexities. And in other
cases, things are far more difficult.

C. Against the Idea of Preferences

From all this we might conclude that for many purposes, the whole
idea of a “preference” is confused and misleading because it is ambiguous
between choices and underlying psychological forces, and because the
mental operations that produce choices are a function of a great many
factors, prominently including social norms and social roles. For many
purposes, it might well be best to dispense altogether with the idea of

preference and to work instead with choices on the one hand and with

complex and somewhat unruly mental states on the other-or to relate
choices to more concrete sources of subjective value (intrinsic value,
reputational effects, and effects on self-conception) or to a wide set of
influences (norms, roles, price changes, increases in leisure time). I be-
lieve that this point gives us reason to doubt the elaborate edifice of social

131. See Gary S. Becker, Habits, Addictions, and Traditions, in The Essence of
Becker, supra note 32, at 218, 232 (referring to metapreferences); cf. Harry G. Frankfurt,
Freedom of the Will and the Concept of a Person, in The Importance of What We Care
About 11, 11-25 (1988) (distinguishing between first order and second order desires).

132. Human motivation includes many ingredients: bodily appetites, emotions,
reasoned judgments, and more. Social norms do not play the same role in all of these.
Hunger and thirst certainly are affected by social norms; one’s attitude toward these things
is hardly unmediated by learning. But hunger and thirst are connected with physical
states, and in extreme forms they are likely to be unpleasant no matter what the prevailing
norms may be-even if the unpleasantness can be counteracted by attitudes of certain
sorts. The case of racial attitudes should be analyzed quite differently. The emotions
present a special case: Early childhood experiences affect fear and jealousy in a way that
they may not some other attitudes. A serious problem with the idea of “preference” is that
it runs all these together. As we will see, the conflation makes it harder to understand how
governmental practices might shape attitudes in different contexts.

938 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

science based on notions of “preference” or “metapreference”133
(though much of the edifice can remain if reconstructed on different
foundations).134 The eventual task is to separate positive, descriptive,
and normative inquiries more sharply, and in the process to try to untan-
gle relevant mental states and their influences on choices.’35

IV. CHOICES, NORMS, ROLES, AND MEANINGS

In this section I discuss in more detail how social norms, roles, and
meanings affect choice. I also deal with the place of shame and pride in
explaining certain anomalies in human behavior. In the process I offer
some notations on how this point bears on law and legal policy, especially
by showing how law can affect choice by affecting the sources of shame
and pride and associated norms.

A. In General

What is the relationship between choices and social norms, between
choices and social roles, and between choices and social meanings?

1. Choices and norms. – Choices are, of course, affected by social
norms. As I have said, we can understand a norm-with respect to
choices-as a subsidy or a tax. If you would otherwise like to dress color-
fully, and if the relevant norm fortifies that choice, the norm is a subsidy;
but norms governing dress can operate as a tax too, as they do in many
professional (and nonprofessional) settings.

The mere fact that a norm is as it is need not be decisive with respect
to choices. Any particular social norm is only a relevant factor. The ex-
tent of its effect will depend on five factors: (1) the intensity of the norm
(exactly how much opprobrium attaches to a violation?); (2) the nature
of the norm (what kind of attitude is signalled by a violation? what kind of

133. This is the approach defended by Gary Becker. See Becker, supra note 131. In
Becker’s view, preferences change, but metapreferences stay constant. There are many
puzzles here. Metapreferences may change too. Someone may want to be the sort of

person who has many friends and goes to many parties (whether or not one is that sort of

person); but this metapreference may shift. Someone may want to be the sort of person
who likes classical music (even if one prefers rock and roll); but this may shift. It may be

possible to posit a human motivation of some sort-to have a good life, or to have a lot of

money-and see how far we can get with the assumption. Perhaps some of the motivations
that it is helpful to posit can be described as metapreferences. But certainly one’s

preferences about one’s preferences shift over time, and it may even be possible to model
those shifts. Note that metapreferences, like preferences, are a product of social norms.

Becker has, however, recently recast his approach to place emphasis on “social capital”
rather than metapreferences, and this recasting fits well with much of what I am suggesting
here. See Becker, supra note 21.

134. Cf. Gary S. Becker, Irrational Behavior and Economic Theory, in The Essence of
Becker, supra note 32, at 18 (discussing dispensability of the economic assumption of
rational behavior by individuals).

135. Becker, supra note 32, and Kuran, supra note 17, at 326-48, make some
movement in that direction.

1996] 939

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

attitude is provoked by violators?); (3) the agent’s attitude toward the
norm and the opprobrium occasioned by its violation (does the agent
like to be seen as a defiant person? how does the agent react to social
opprobrium?); (4) the possibility of social approval or forgiveness among
relevant subgroups (will the agent’s peer group support the norm-defy-
ing act?); and (5) the nature and weight of the other ingredients in
choice, including competing norms, intrinsic value, and effects on self-
conception (what must a norm-complier sacrifice?).

Some people like to make choices that violate social norms, or do
not much mind doing so. Some people are willing to ask whether some,
many, or all norms really make sense. Some people are willing to choose
in such a way so as to violate norms if the “price” of compliance is high, as
it is, perhaps, for some people who are asked to refrain from smoking, or
for others who are asked to clean up after their dogs. If norm violators
are viewed in relevant places as somewhat courageous-or as freedom
fighters engaged in a form of civil disobedience-norm violation will not
be terribly uncommon. This is true for smokers in some places, and also
for people engaged in homosexual relations. With respect to both,
whether to be “in the closet” is partly a function of social norms.

2. Choices and roles. – Choices are pervasively a function of social
role. A teacher might dress a certain way, take a certain salary, or refuse
to talk about the current President in class, because of what is (under-
stood to be) entailed by the social role of teacher. A teacher might do
some of this even though she would greatly wish that the social role of
“teacher” had different implications. A patient might act a certain way
toward her doctor just because of the nature of norms associated with the
doctor-patient relationship, even if the patient would like that relation-
ship to be different.

In some cultures, a social role entails a great deal of gift-giving to
hierarchical superiors.136 Social inferiors who do not give gifts are signal-
ling something bad or dangerous about themselves, to the effect that they
are disrespectful, rebellious, or angry at their superiors. The “choice” to
offer gifts is a function of role. This is pervasively true of human choices,
and if it is not visible, it is only because the particular role is so deeply
internalized. The practice of giving gifts to hierarchical superiors may
seem exotic to most Westerners; but it has parallels in a wide range of
social gestures signalling an understanding of social hierarchies and
roles.

3. Choices and meanings. – Like a social norm, the social meaning of
an act can operate as a subsidy to choices or as a tax on them. Suppose
that the social meaning of listening carefully to others, and not talking a
great deal of the time, is to reveal yourself as a good person who cares

136. Consider, for example, contemporary China.

940 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

about others and about what other people have to say. If so, the act137 of

listening carefully to others is “subsidized” by its social meaning, in the
sense that the benefits of listening carefully are increased, by the social
meaning, over what they would otherwise be. If, by contrast, the relevant
social norm is that talking a lot, interrupting, and claiming the spotlight
reveals strength of character, confidence, and an understanding of one’s
general importance-whereas listening reveals weakness, low status, and
confusion-the social meaning of listening carefully is a “tax” on the
choice of doing so.138 Thus courtesy can be made more or less costly by
social meaning. There are many examples in social life, with the amount
of the subsidy or the tax varying with context.

4. Preferences. – What is the relationship between social norms, so-
cial meanings, and social roles on the one hand, and “preferences” on
the other? If preferences are what lie behind choices, then norms, mean-
ings, and roles are part of the complex and unruly set of factors that help
produce choices. People usually do not want to feel shame, and for this
reason they may act consistently with a social norm, even if they believe
the norm is wrong and wish the norm were otherwise. Some people may
not want to face the hostility of others and may obey the norm for that
reason. Some people may do things to counter the social norm, perhaps
to show a reflective judgment that the norm is wrong, perhaps for the
sake of defiance. We might say, then, that any preference for an action is

partly a function of social norms and the agent’s attitude toward those
norms.

B. Willingness To Pay vs. Willingness To Accept: The Place of Shame

Once we incorporate social norms into our account of choice, we
can explain some anomalies in common theories of rationality. As the
five tales at the beginning of the Article show,139 seemingly anomalous
behavior can result from anticipated shame or pride. My goal here is to

explain how social norms help account for anomalies in ways that bear on

positive social science in general. Many such anomalies qualify as such at
least partly because of social norms. In particular, behavior is a product
of shame and pride, or anticipated shame and pride, and changes in be-
havior often reflect anticipations of shame and pride (as seen in all five
tales in Part I).

137. Not the preference, which is a confusing concept in this setting, for reasons
discussed above.

138. See generally Diego Gambetta, “Claro!”: How to Get Deliberative Democracy
Wrong (1995) (unpublished manuscript, on file with the Columbia Law Review).
Gambetta describes the culture of “claro” in Italy and South American countries, in
accordance with which any statement of any kind is greeted with the word “claro!”-
signalling, “of course; I knew that”-and in which any statement of ignorance is treated as
an admission of weakness. See id. at 1, 6-13. In such a culture, the social meaning of
surprise and learning is different from what it is elsewhere.

139. See supra Parts I.A-E.

1996] 941

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

Recent empirical work suggests that many claims in economics140
rest on an intriguingly false assumption, one that indicates that it some-
times may be impossible for government to take preferences “as they
are.”’41 The basic finding is this: The initial grant of an entitlement of
some good X to some person A can make A value X far more than he
would if X had been initially allocated to B. (It also makes B value it less
than he otherwise would.) The initial allocation-the legal rule saying
who owns what before people begin to contract with one another-serves
to create, to legitimate, and to reinforce social understandings about pre-
sumptive rights of ownership. The effect of the initial allocation of a
commodity or an entitlement is commonly described as the “endowment
effect.”142

This point has received considerable empirical confirmation, often
in the context of environmental amenities. One study found that people
would demand about five times as much to allow destruction of trees in a
park as they would pay to prevent the destruction of those same trees.143
When hunters were questioned about the potential loss of their hunting
licenses, they said that they would be willing to pay an average of $247 to
prevent the loss-but would demand no less than $1044 to accept it.’44
In another study, participants required payments to accept degradation
of visibility ranging from five to more than sixteen times higher than
their valuations based on how much they were willing to pay to prevent
the same degradation.145 A related experiment tried to ascertain the
“existence value” of a tree-like houseplant. The subjects were told that
any trees not sold or kept would be killed at the end of the experiment.
The mean willingness to pay (WTP) to avoid the “kill” option was $7.81.
The mean willingness to accept (WTA) payment to allow a tree to be
killed was $18.43.146

140. Including the Coase theorem.
141. See Richard H. Thaler, Toward a Positive Theory of Consumer Choice, in Quasi

Rational Economics, supra note 38, at 3, 7-10. A good overview is Elizabeth Hoffman &
Matthew L. Spitzer, Willingness to Pay vs. Willingness to Accept: Legal and Economic
Implications, 71 Wash. U. L.Q. 59 (1993). See also W.R. Dubourg et al., Imprecise
Preferences and the WTP-WTA Disparity, 9J. Risk & Uncertainty 115, 116 (1994) (arguing
that imprecision of people’s preferences is a factor explaining WTP-WTA disparities).

142. It was first so-called in Richard Thaler, Toward a Positive Theory of Consumer
Choice, 1 J. Econ. Behav. & Org. 39 (1980). This essay, along with others of similar
interest, can be found in Quasi Rational Economics, supra note 38.

143. See David S. Brookshire & Don L. Coursey, Measuring the Value of a Public
Good: An Empirical Comparison of Elicitation Procedures, 77 Am. Econ. Rev. 554, 562-63
(1987).

144. SeeJudd Hammack & Gardner M. Brown,Jr., Waterfowl and Wetlands: Towards
Bioeconomic Analysis 26-27 (1974).

145. See Robert D. Rowe et al., An Experiment on the Economic Value of Visibility, 7
J. Envtl. Econ. & Mgmt. 1, 10 (1980).

146. See Rebecca R. Boyce et al., An Experimental Examination of Intrinsic Values as
a Source of the WTA-WTP Disparity, 82 Am. Econ. Rev. 1366, 1370 (1992).

942 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AAD SOCIAL ROLES

In general, the range of the disparity appears to vary from slight dis-
parities to a ratio of more than four to one, with WTA usually doubling
WTP. Environmental goods tend to reflect a disparity of factors from two
to over ten.147 In some environmental experiments involving trees, the
WTA/WTP ratio is extraordinarily high, averaging around 75/1.148

What explains this phenomenon? There are many possibilities, and
none is likely to be exhaustive.149 My suggestion is that some of the dif-
ference between WTP and WTA has a great deal to do with social norms
and social meanings. If someone says that she is willing to accept $X to
allow the extinction of a species, the meaning of her action is altogether
different from what it is if she says that she is willing to pay $X (and no
more) to prevent the extinction. Under prevailing social norms, one
ought not to accept even a great deal of money to allow destruction of an
environmental amenity-partly because the good at issue is collectively
owned, partly because its loss may be irreversible, and partly because it is
not thought to be commensurable with its cash equivalent (in the sense
that it is not valued in the same way or along a single metric).

In these circumstances, people who announce their willingness to
accept cash for the loss of a pond or a species feel shame. They believe
that they are assuming responsibility for the destruction of something in-
trinsically valuable, not replaceable, and owned by many people. Because
of the risk of shame, people will demand a great deal, and they may even
refuse any amount that is offered.150 By contrast, those who refuse to pay
an enormous or infinite amount to save an environmental amenity do not
feel the same degree of shame (if they feel shame at all).151 They are
confronted with a different set of social norms.

Take an analogy. If someone is asked how much she would be will-
ing to accept to allow her dog’s life to be shortened by six months-or
how much she would be willing to accept to allow her dog to suffer severe

147. See id. at 1366.
148. See Brookshire & Coursey, supra note 143, at 561-62.
149. See Cass R. Sunstein, Endogenous Preferences, Environmental Law, 22 J. Legal

Stud. 217, 227-30, 242-53 (1993).
150. Thus in surveys nearly 50% of people sometimes refuse to name any amount.

See Rowe et al., supra note 145, at 9.
151. There is a background issue here about the relation between shame and guilt.

Sometimes shame is described as a result of violations of a widely shared standard of
proper behavior; by contrast, guilt is described as a transgression of other people’s rights.
See John Rawls, A Theory of Justice 483-84 (1971). Sometimes guilt is described as a
feeling of wrongdoing experienced by individuals consulting their own conscience in an
individualistic culture, whereas shame reflects violation of a community’s widely shared
norms. See Stuart Schneiderman, Saving Face: America and the Politics of Shame 5-6
(1995). On this view, America is sometimes described as a “guilt culture” andJapan as a
“shame culture.” See id. at 5. But these descriptions are controversial.

There is also an underlying question about the extent to which people actually believe
that the norm is right. Of course people may comply with norms (because of the
reputational cost of violating them) that they reject in principle; their compliance may be
more likely if they approve of the norm for articulable or firmly held reasons.

1996] 943

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

pain for (say) one week-she might well say: “No amount is sufficient.”
The question is very different if a veterinarian is asking someone whether
unusual and expensive medical procedures should be used to prolong a
dog’s life or to reduce its pain. Here the answer need not be: “No
amount is too high.”

Some intriguing work suggests that the disparity between WTA and
WTP is connected with the assignment of moral responsibility for the de-
struction of environmental assets, which are perceived as intrinsic goods.
The WTA measure assigns responsibility to the individual. The WTP
measure does so more ambiguously.152 These findings are consistent
with the norm-based explanation I am offering here. People want to
avoid or minimize the feeling that they have been morally culpable for
producing the loss of an environmental amenity.153 Feelings of moral
culpability are tightly connected with prevailing social norms.

C. Anomalies, Shame, Altruism, and Free-Riding

The general point-that choices are a function of norms and the
sentiments they produce-relates to the first, second, and fourth tales in
Part I;154 we are now in a position to explain these apparent anomalies as
well. When two people are to divide an amount given to them under the
stated conditions, the offeror in the ultimatum game feels shame given
prevailing norms155-that he is demonstrating that he is a greedy and
even horrible person-if he offers a penny or a dollar from a sum of (say)
$200. If a sum is given to two people under the conditions of the game,
good people share; they do not try to keep almost all of the money for
themselves. For his part, the offeree feels mistreated-treated in a con-
temptuous way-if a small or token amount is suggested.156 The social
meaning of the statement, “How about five cents for you?” is contempt;
the social meaning of responding, “Great!” is a willingness to be
dishonored. 157

152. This is an apparently pervasive social judgment about responsibility, but it must
of course be defended; it is not self-evident. Consider the distinction between acts and
omissions, of which this is an example; the line between the two has a great deal to do with
ascriptions of moral responsibility. I am trying to describe the norm, not to defend it.

153. See Boyce et al., supra note 146, at 1371.
154. See supra Parts I.A-B, D.
155. He may well feel guilt as well; he may accept the prevailing norm in principle.
156. See Roth, supra note 1, at 270-74, 298-302.
157. Cf. Richard H. Thaler, The Psychology of Choice and the Assumptions of

Economics, in Quasi Rational Economics, supra note 38, at 137 (describing fallibility of
rational choice theory); Daniel Kahneman et al., Fairness as a Constraint on Profit
Seeking: Entitlements in the Market, in Quasi Rational Economics, supra note 38, at 199
(discussing role of fairness in commercial settings). Of course feelings of dishonor may be
based on bad norms, and it may be appropriate for people to try to change them. Cf.
Elster, The Cement of Society, supra note 26, at 116-21 (discussing feuding); supra note
97 (describing current efforts to stop people from fighting when insulted).

944 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

So, too, even affluent people may recycle if prevailing norms make it
shameful to refuse to do so, or if prevailing norms make it seem wonder-
ful to engage in menial labor. Those who separate their garbage may
enjoy the reputational benefits of doing so. Or they may enjoy the effects
on their self-conception of contributing to what is perceived as a social
good. Hence it should come as no surprise that people in East Hampton
are willing to recycle, and even to separate, their garbage in a public
space. 158

Experimental work shows that people contribute to a shared good,
and refuse to free ride, far more often than economists predict.159 It also
shows that agents are willing to cooperate, and hence to solve collective
action problems without coercion, if most people are seen as cooperators;
in such circumstances the social meaning of noncooperation is greed or
selfishness. When a number of people free ride, and are seen to free
ride, cooperation breaks down; in such circumstances the social meaning
of cooperation is a willingness to be a “dupe” or a “sucker.” The desire to
contribute to a collective good is palpably a function of social norms. If
social norms do not lead most people to contribute, contributions de-
crease steadily and dramatically.160 The second tale161-involving litter-
ing and cooperation-should therefore be taken as a metaphor for many
social outcomes.

My suggestion, then, is that apparent puzzles of rationality are often
a product of social norms and moral judgments that are intertwined with
those norms. When people appear to behave irrationally, in the sense
that they violate predictions based on economic assumptions, it is often
because social norms are at work. In such circumstances, nothing need
be irrational in the normative sense about the underlying behavior; such
people are not confused or inconsistent; they see existing norms (and
associated meanings) as imposing costs on choice. Of course a full expla-
nation of these matters would have to include an account of norms as
well-why are they as they are?-and there is a risk that a reference to
social norms will become a conclusory response to any apparently anoma-
lous results. But once we identify prevailing norms, we may be able to
make robust predictions about behavior and also to test those predictions

158. See supra Part I.D.
159. See Charles A. Holt, Industrial Organization: A Survey of Laboratory Research,

in The Handbook of Experimental Economics, supra note 1, at 349, 409-11; John O.
Ledyard, Public Goods: A Survey of Experimental Research, in The Handbook of
Experimental Economics, supra note 1, at 111, 141-69; John M. Orbell et al., Explaining
Discussion-Induced Cooperation, 54J. Personality & Soc. Psychol. 811, 811 (1988). Note
also that cooperation increases when people can talk with one another; discussion
significantly raises contribution rates, perhaps because it increases empathy and the shame
associated with noncooperation. See id. at 811-12, 818. Of course discussion may increase
cooperation because it can introduce more assurance of compliance; but the greater
assurance itself comes from norms.

160. See Kuran, supra note 17, at 48-49.
161. See supra Part I.B.

1996] 945

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

by asking, for example, whether the anomalous behavior continues when
it is not observed publicly. We also may know something about the pros-
pects for social change through law. If we know how laws affect behavior,
we can more accurately generate laws that target norms. And if law gen-
erates a change in norms, behavior may change too. (Consider the sec-
ond, third and fourth tales.162)

These points suggest a more complex and adventurous one: For
many purposes, it is not sensible to oppose “rationality” and social norms.
Of course we could define “rationality” in many ways, and economists
have offered restricted definitions that are useful for many purposes.163
Such definitions, whether or not abstracted from norms, may be valuable
for positive purposes. The test of a positive theory lies in its usefulness in
making predictions. My central point here is that norms will create many
of the costs and benefits that enter into an agent’s decision. It follows
that from the standpoint of the individual agent, rational decision is very
much a function of social norms. Norms help show what it makes sense
to do; norms even help identify costs and benefits as such.

It might seem natural to suppose that it is in people’s interest not to
pick up their garbage (see the second tale164) and that social norms
against littering add a new or artificial factor to the individual calculus.
But if we make this supposition, we are saying something about the indi-
vidual’s calculus without the anti-littering norm. What is the basis for any
particular conception of how the calculus will come out? No such con-
ception will be free of an array of ends, seen as such partly because of
social influences, including social norms. Why, for example, is picking
up garbage a cost rather than a benefit? Why, for that matter, is voting,
cooperating, seeking revenge, putting one’s fork on the right-hand side
of the plate, covering up one’s nose as one sneezes, or for that matter
violating rather than complying with norms, best treated as a cost rather
than a benefit?

An implicit (but undefended and obscure) state of nature theory
seems to lie at the heart of many distinctions between social norms and
rationality, or between social norms and rational self-interest. That is, the
distinction seems to make sense only on the basis of some supposition
about what people would do in some natural state lacking social norms.
But we are not likely to be able to make much progress in that way. As I
have suggested, a positive theory of human rationality will likely do best if

162. See supra Parts I.B-D.
163. See, e.g., Douglas G. Baird et al., Game Theory and the Law 6-46 (1994);

Graham Loomes & Robert Sugden, Regret Theory: An Alternative Theory of Rational
Choice Under Uncertainty, 92 Econ. J. 805 (1982); Amos Tversky & Daniel Kahneman,
Advances in Prospect Theory: Cumulative Representation of Uncertainty, 5 J. Risk &
Uncertainty 297 (1992).

164. See supra Part I.B.

946 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

it sees norms as taxes on or subsidies to choice, and hence as part of the

array of considerations that people face in making decisions.165

V. GOVERNMENT ACTION: ON AUTONOMY AND TOOLS

Sometimes it seems desirable for individuals or societies to change
choices. Government might attempt to change, or help people to

change, choices by changing social norms, social meanings, and social
roles. In fact changes in norms may be the cheapest and most effective

way to make things better, whatever are our criteria for assessing that mat-
ter. The relation between behavior and norms has yet to receive sus-
tained attention; when we attend to that relation, we see that government
has a policy instrument of great potential value. In this section I deal with
some general points. I turn to more specific grounds, and more specific
objections, in Parts VI and VII.

To be sure, private power to create norm communities may make
government action less necessary or less desirable. Often the best step is
to allow those communities to be formed and to see how they work out.
But sometimes it is too costly for individuals to create or join those com-
munities, and sometimes the generally held norm is too damaging to
human well-being. These issues cannot be solved in the abstract; the

judgment depends on the details. But it is clear that norms can create

problems of various sorts and that collective action may be required in
some cases.

A. Norms and Paternalism

Common objections to “paternalism” or “meddling” are not easy to
sustain in such contexts. Recall that people usually do not choose norms,
meanings, and roles; all of these are (within limits) imposed. As I have
said, it would be ludicrous to deplore norms, meanings, and roles. They
make life possible and they much facilitate social engagement. They pro-
vide the context within which free interaction is possible. Nonetheless,
some of them operate as severe limits on autonomy or well-being, and

certainly they should not be treated as fixed or given regardless of their
content or consequences.

Private groups can test or even change norms. Indeed, the testing of
current norms, meanings, and roles is a crucial function of groups that
are intermediate between citizens and the state. Religious groups are in
this sense norm entrepreneurs; the same is true for environmental and
civil rights organizations. But sometimes private groups are unable to
produce desirable change on their own. This is a point missed by the
idea that the sole basis for government action is to avoid force, fraud, and

165. See Kuran, supra note 17; Becker, supra note 32.

1996] 947

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

“harm to others.”’66 Obstacles to autonomy and to good lives can also
come from bad roles, norms, and meanings.

Changes in norms and meanings can promote human well-being.
Often all or most people would, on reflection, like to see a change in a
particular norm;167 and yet they cannot bring about the change on their
own, because in his individual capacity, each person has limited power to
alter meanings, norms, or roles. The case of mandatory helmets for
hockey players is a familiar example. Hockey players may prefer not to
wear helmets if the meaning of helmet-wearing is cowardice; but their
preferred solution, available only through a league mandate, is a system
in which all are required to wear helmets, and hence players wear hel-
mets without signalling cowardice.168 Of course shifts in norms, mean-
ings, and roles are pervasive. Consider, for example, changing norms
with respect to smoking, littering, drug use, polluting, racial discrimina-
tion, sexual relations outside of marriage, the roles of women and men,
and interracial relationships.

In fact it is often hard to know what people would “like” or prefer,
because their judgments and desires are entangled with norms, mean-
ings, and roles, and because once one or more of these is changed, they
may be better off, either objectively or subjectively. If government
changes the social meaning of smoking (see the third tale169), has it ac-
ted illegitimately? What if most people, or most smokers, would, on re-
flection, want smoking to have a different meaning? Or suppose that gov-
ernment tries to change an aspect of a certain social role, like that of
unwed fathers, high school teachers, homosexuals, or workers. Surely the
consequences of the change matter; surely it matters if the change is sup-
ported by (most or all) unwed fathers, high school teachers, homosexu-
als, or workers, and if members of each group face a collective action
problem. The justification for government action is firm if all or almost
all people support it as a means of solving a collective action problem.

B. Tools

Suppose that government wants to change norms, meanings, or
roles. It has many different tools for doing so; some of these are mildly
intrusive while others may foreclose choice. Government may restrict it-
self to education, understood as simple statements of fact designed to en-

166. See Epstein, supra note 16, at 37-49; Richard A. Posner, Sex and Reason 2-4
(1992). I think that both Epstein and Posner give inadequate attention to the liberty-
limiting effects of norms, roles, and meanings. The classic source of the “harm to others”
idea, Mill, offers subtle arguments on the point. See John S. Mill, The Subjection of
Women (1869); John S. Mill, On Liberty (1859) (developing a complex discussion of
pressures to conform).

167. Their desire to this effect is likely to be a product of norms that are being held
constant. See infra Part VIA. 2.

168. See Schelling, supra note 83, at 213-14; Lessig, supra note 24, at 967-68.
169. See supra Part I.C.

948 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

sure accurate beliefs. We have seen that norms, meanings, and roles can
be a function of beliefs and that beliefs are mutable.170 Perhaps prevail-
ing beliefs are false and warrant correction. People may think that AIDS
is a disease limited to homosexuals, that smoking does not hurt non-
smokers, or that there is no relation between cholesterol and heart dis-
ease or between diet and cancer. Changing norms with respect to smok-
ing are almost certainly a result-at least in part-of information from
the government about health risks.171

Of course government may try to ensure accurate beliefs in order to
persuade people to do something new or different. It may seek to in-
crease or decrease shame; or it may not be thinking much about norms.
But so long as government restricts itself to the provision of accurate in-
formation, there should be no objection to governmental efforts to cor-
rect false beliefs, even if the correction affects norms, meanings, and
roles. In fact the change along this dimension may be the most impor-
tant consequence of education, which may, for example, remove certain
kinds of shame on the part of people who want to buckle their seat belts
or abstain from premarital sex.

Government may also attempt to engage in persuasion, understood as
a self-conscious effort to alter attitudes and choices rather than simply to
offer information.172 Consider the third tale;173 assume that some such
advertisement had been issued by state officials. Perhaps it would have
been effective (though its social meaning would have been altogether dif-
ferent if issued by officials rather than by members of the private African-
American community). The ‘just say no” policy for drugs falls in the cate-
gory of attempted persuasion; so, too, with efforts to control AIDS by
strongly encouraging the use of condoms or abstinence from sex. Here
government does not restrict itself to provision of information, but in-
stead uses rhetoric and vivid images to change norms, meanings, or roles,
and in this way attempts to persuade people to choose a certain course.174

170. Beliefs, however, are not easily mutable because of cognitive dissonance. See
George A. Akerlof & William T. Dickens, The Economic Consequences of Cognitive
Dissonance, in George A. Akerlof, An Economic Theorist’s Book of Tales 123 (1984);
Aronson, supra note 95, at 171-239.

171. See W. Kip Viscusi, Smoking: Making the Risky Decision 47-60 (1992) (relating
smoking trends and public information).

172. Cf. id. at 144-46 (criticizing persuasion as “propaganda” in the context of
smoking, and arguing for information instead). Viscusi treats norms and meaning as
exogenous givens and as some ineradicable part of the guiding normative criterion,
“choice.” If these are not fixed and not chosen, the attack on persuasion is much weaker
because the background contains limits on liberty that government and citizens may seek
to overcome. See, e.g., Excerpts from Clinton News Conference on His Tobacco Order,
supra note 6, at A18.

173. See supra Part I.C.
174. This kind of approach is much more effective than the mere provision of facts.

See Aronson, supra note 95, at 89-90. Part of the reason is status quo bias. See William
Samuelson & Richard Zeckhauser, Status Quo Bias in Decision Making, 1 J. Risk &
Uncertainty 7, 8 (1988) (describing how people sometimes retain their particular choices

1996] 949

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

If government is interested in persuasion, it will understand that “ref-
erence states” matter a great deal, and it will use this understanding to
lead people in the right direction. People are generally loss averse, in the
sense that they are made more unhappy by a loss of a certain amount
than they are made happy by an equivalent gain.175 Hence reactions to
information turn a good deal on how the information is framed. A claim
that breast cancer examinations will increase the number of lives saved
(over what it would otherwise be) is far less effective than a claim that the
failure to examine one’s breast will decrease the number of lives saved
(over what it would otherwise be).176 Similarly, energy conservation pro-
grams are far more effective if they point to the dollars lost through fail-
ure to conserve than if they point to the dollars saved through conserva-
tion.177 The difference has a great deal to do with social norms and
social meanings and in particular with shame. If government is inter-
ested in changing behavior, it will take advantage of the phenomenon of
loss aversion.

Some people think that although the provision of information can
be justified, government should rarely or never attempt to persuade.178
But if norms, roles, and meanings are beyond individual control, and
sometimes bad, this thought is hard to sustain, at least if government is

subject to democratic controls.
Consider in this connection the problem of smoking and the lessons

of the third tale.179 Among African-Americans between 18 and 24, the

smoking rate has fallen from 37.1% in 1965, to 31.8% in 1979, to 20.4%
in 1987, to 10.3% in 1992.180 Among whites in the same age group, the
rate fell from 38.4% in 1965 to 27.8% in 1987-but it has remained more
or less constant since that time.181 The change among African-American
teenagers is universally described as a ” ‘public health success story,’ ” but
one that government officials cannot explain.182 Though no one has a
full account of this phenomenon, changing social norms appear to be

instead of changing to a preferred alternative); see also Dubourg et al., supra note 141, at
128 (discussing overwhelming importance of the “reference state” in determining
reactions to changes in risks).

175. See Daniel Kahneman et al., Experimental Tests of the Endowment Effect and
the Coase Theorem, in Quasi Rational Economics, supra note 38, at 167, 182-86.

176. See Beth E. Meyerowitz & Shelly Chaiken, The Effect of Message Framing on
Breast Self-Examination Attitudes, Intentions, and Behavior, 52 J. Personality & Soc.
Psychol. 500, 506-09 (1987).

177. See Marti H. Gonzales et al., Using Social Cognition and Persuasion to Promote
Energy Conservation: A Quasi-Experiment, 18J. Applied Soc. Psychol. 1049, 1062 (1988).

178. See Viscusi, supra note 65, at 149-59.
179. See supra Part I.C.
180. See U.S. Department of Commerce, supra note 8, at 144.
181. See id.
182. Donald Bradley, Few Black Teens Smoke, But Why?, Kan. City Star, Aug. 12,

1995, atAl (quoting Michael Eriksen, Director of the Office on Smoking and Health at the
Centers for Disease Control and Prevention).

950 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

playing a substantial role.’83 Smoking does not have the same cachet in
the African-American community that it has among whites. If govern-
ment could bring about a general change in social norms-through, for
example, attempts at persuasion-it is hardly clear that there would be a
good objection to its behavior. Consider related findings about the close
relation between norms and alcohol abuse.184

Government also might use economic instruments to tax or subsidize
choices. Of course education is assisted publicly, as are day care, muse-
ums, and public broadcasting (at least as of this writing). Alcoholic
drinks, tobacco products, generation of waste, and some polluting activi-
ties are met with taxes (although some of these are subsidized too). We
can understand some economic incentives as efforts in part to counteract
social meanings, social norms, or social roles with financial benefits or
penalties designed to produce a good “equilibrium.” A social meaning
that is perceived to be bad might be “matched” with a financial disincen-
tive. Some such incentives may amount to efforts to change social norms
or social meanings. If a good becomes more expensive, the social mean-
ing of choosing it may change. Cigarette smoking, for example, may be-
come less popular if most people cannot afford to smoke. Of course it is
hard to predict the effects of changes in price on norms.

Government also might impose time, place, and manner restrictions. It

might ban smoking in public places. It might say that television shows
containing violence may be shown only in certain time slots. It might
require government itself to choose low-polluting motor vehicles. It
might ban affirmative action in the public sector but allow it in the pri-
vate sector. Strategies of this kind might affect the social meaning of the

183. See id. at A19. It might be speculated that there has been a shift from cigarettes
to some other harmful substance, such as illicit drugs. But it is notable that African-
Americans have lower lifetime rates of reported use of illicit drugs (especially cocaine)
than whites. The 1990 National Household Survey on Drug Abuse finds that 40 to 50
percent fewer African-Americans than whites under age 35 report any lifetime use of
cocaine. (About 20% more African-Americans than whites over age 35 report lifetime
cocaine experience.) For use in the last year, African-Americans between 12 and 17 report
the lowest rates of use of cocaine-1.7% compared to 2.3% for whites. African-Americans
in the 12 to 17 age group consistently posted lower prevalence rates than whites for
lifetime, past year, and past month use. (Surveys in 1985, 1988 and 1990.) All of these
data suggest that young African-Americans use illicit drugs at a lower rate than do young
white people. These data do not address directly the substitution of cocaine for cigarettes.
But it appears that while African-Americans between 12 and 17 use cocaine at lower rates
than whites (according to 1985, 1988, and 1990 surveys), rates for both groups have
decreased from 1985 to 1990. The white rate has decreased from 4.2% to 2.3% and the
African-American rate has gone from 2.4% to 1.7%. Thus it seems that the public health
success story regarding African-American teenagers and cigarette smoking also extends to
cocaine use. And there may be another public health success story to be told with respect
to cocaine use among white teenagers. All data is from Denise B. Kandel, The Social
Demography of Drug Use, 69 Milbank Q. 365, 395-99 (1991). I am grateful to Tracy
Meares for help on this point.

184. See William J. Sonenstuhl, Working Sober: The Transformation of an
Occupational Drinking Culture (forthcoming 1996).

1996] 951

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

relevant activity very generally. But they do not foreclose choice entirely;
they channel it instead. Of course in cases of this sort, government is

directly affecting behavior. What I am suggesting is that it may be seeking
as well to affect social norms and social meanings by affecting the well-

springs of shame and pride. The social meaning of smoking is, for exam-

ple, affected by virtue of the existence of time, place, and manner restric-
tions, even in places lacking those restrictions.

The most intrusive kind of government action is of course straightfor-
ward coercion. Thus government might prohibit the use of certain drugs;
require everyone to recycle or buckle their seat belts; or make education
mandatory for people under a certain age.

C. Levels and Institutions

A polity concerned to change norms could proceed through differ-
ent levels and with different institutional arrangements. Should it en-

courage private behavior through creative public-private partnerships?
Should it, for example, fund private groups that are engaged in discour-

aging teenage pregnancy, drug use, and promiscuity? There is also a

question about whether national action or local action is better. Norm

management could occur in towns; it could happen at the state level; it
could be a self-conscious effort of the national government.

The question about appropriate levels of government is hard to an-
swer in the abstract. Much depends on the details. But two notations

may be appropriate. First, public-private partnerships are often the most
fruitful of all possible approaches. Purely governmental efforts at norm

management may fail for lack of trust; imagine, for example, if the na-
tional government had engaged in the anti-smoking campaign that was

apparently so successful among African-American teenagers. Often a gov-
ernment concerned with bad norms does best by working with charitable
and religious organizations which can have a high level of competence,
experience, and trust.

Second, it is probably best to have a presumption in favor of the
lowest possible level of government.185 The lowest level is closest to the

people, and in that sense most responsive to it, and most likely to be
trusted by it. The risk of futility or positive harm is accordingly dimin-
ished. Of course there are familiar counterarguments, and sometimes
national efforts at norm management are well-justified. But a nation that
is concerned about existing norms should exploit the possibilities that
exist in a system committed to federalism.

185. See Michael W. McConnell, Federalism: Evaluating the Founders’ Design, 54 U.
Chi. L. Rev. 1484, 1491-1511 (1987) (reviewing Raoul Berger, Federalism: The Founders’
Design (1987)); Andrzej Rapaczynski, From Sovereignty to Process: TheJurisprudence of
Federalism after Garcia, 1985 Sup. Ct. Rev. 341, 400, 402-03, 407-08; Charles M. Tiebout,
A Pure Theory of Local Expenditures, 64J. Pol. Econ. 416 (1956).

952 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

VI. GOVERNMENT ACTION: FIVE GROUNDS

In this section I discuss several grounds for governmental efforts to
change norms, meanings, and roles. The unifying theme is the expressive
function of law-a term that I use to identify the function of law in expres-
sing social values and in encouraging social norms to move in particular
directions.186

My discussion will not by any means exhaust law’s expressive uses,’87
and I do not discuss any of these grounds in much detail. I offer instead
a brief sketch of some possibilities. An account of legitimate grounds for
changes in norms will overlap a great deal with-and may even be reduc-
ible to-an account of the legitimate bases for governmental action. My
purpose here is hardly to offer such an account, but instead to see how
some fairly standard ideas might be brought to bear on the particular
subjects under discussion. In none of these cases do I urge that social
norms should be free from scrutiny on the merits by individual citizens.
Government ought not to inculcate norms that cannot be supported and
evaluated publicly, and on the basis of reasons. Like rules, norms will
typically have a degree of crudeness and rigidity, and it is entirely appro-
priate for citizens to conclude that there are contexts in which even good
norms make no sense if they are mandated.188

Nor do I urge that the mere fact that behavior is a function of norms
provides some warrant for governmental action. To defend governmen-
tal action, some claim has to be offered about how such action will make
things better. A reference to norms helps undermine the view that gov-
ernment should restrict itself to the satisfaction of “preferences.” We
have seen that this term is highly ambiguous and that norms can often
undermine human well-being. But to defend a change, the idea of well-
being has to be specified, and it has to be shown that the change would
improve matters under the relevant criteria.

To introduce the analysis, we might make a few initial distinctions.
1. In some cases, all or almost all people will support an effort to

change a norm, a meaning, or a role. In the case of hockey helmets, for

186. For more discussion, see Sunstein, supra note 64.
187. In particular there is an intriguing set of laws designed to fortify norms

regulating the use of money as a reason for action. See id.
188. There is a currently prominent and controversial example: governmental efforts

to stigmatize unwed parenthood. At first glance, it seems legitimate for public and private
institutions to try to inculcate norms that make people feel ashamed if they produce
children who will have horrible life prospects. For this reason, efforts to inculcate norms

against unwed parenthood-in cases in which the children who result will have miserable
lives-are fully supportable. But there are several questions that might be raised about this
conclusion: (a) In some cases, unwed parents will not have done anything wrong, and
hence they are entitled not to feel ashamed; (b) The fact that the children will have
miserable lives may be a result of a background of social injustice; (c) Shame is heavy
medicine, and some people might think that government should feel free to inculcate
norms but not the particular norm of shame in this setting. These are merely questions
about the project of shame inculcation in this context; they do not purport to be answers.

1996] 953

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

example, there may be near-unanimous agreement that things would be
better if the meaning of helmet-wearing were not cowardice. In the case
of cleaning up after one’s dog, almost everyone may agree that things
would be better if there was a norm in favor of cleaning up. Of course
any such agreement may depend on norms that should themselves be

brought into question.189 We might say that the agreement can be “im-

peached” by showing that people do not, on reflection, endorse the
norms that produce agreement, or that there are problems with those
norms.

2. In some cases, all or almost all people will agree that behavior is a
function not of intrinsic value, and not of effects on people’s (well-con-
sidered) judgments about their self-conception, but instead a function of
the reputational consequences of choice; and they will agree too that it
would be better if the reputational consequences were different. All or
almost all people might agree, for example, that use of drugs in a certain

community stems from the reputational benefits of using drugs and the

reputational costs of refusing to do so. They also might agree that things
would be better if using drugs produced reputational harm rather than
benefit; they might act accordingly via law (with education, attempted
persuasion, economic incentives, or coercion). In this kind of case, there
is also a possibility of “impeachment” of the agreement along lines sug-
gested above.

3. In some cases, existing norms may be part and parcel of a caste

system. They may turn a morally irrelevant characteristic-race and gen-
der are the most obvious examples-into a signalling device with respect
to social role and associated norms. If a caste system is unjust, it is appro-
priate to alter norms, roles, and meanings that perpetuate it.

4. In some cases, existing norms undermine people’s well-being or

autonomy, by discouraging them from being exposed to diverse concep-
tions of the good and from giving critical scrutiny to their own concep-
tions, in such a way as to make it impossible for them to be, in any sense,
masters of the narratives of their own lives.190

These are brief notations on some highly controversial subjects.
Needless to say, I cannot attempt here to defend an account of the appro-
priate role of the state; my purpose is more modestly to connect the pro-
ject of norm management with some familiar ideas about what the state

legitimately does. Notably, there are cases in which it is unnecessary to
choose among two or more of the general grounds for government ac-
tion; people with varying theoretical commitments might believe that a
particular action makes sense. Hence political participants might achieve

189. See infra Part VI.A.
190. On this conception of autonomy, see Raz, supra note 22, at 400-29. This idea is

very controversial as a foundational norm. See, e.g., John Rawls, Political Liberalism
134-35 (1994). But political liberals might well favor an effort to ensure against norms
that prevent people from being or acquiring the capacities to be free and equal citizens.
See id. at 174-76.

954 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

an incompletely theorized agreement on a particular outcome-an agreement
on what steps make best sense, unaccompanied by a shared understand-
ing of why, exactly, they make the best sense.191 Those interested in pos-
sible changes in norms would do well to take advantage of such
agreements.

A. Some Unusual Collective Action Problems

1. Standard accounts. – In a now-standard account, offered by many
economists, lawyers, and political scientists,192 social practices would be
inefficient were it not for certain social norms; such norms solve collec-
tive action problems. They do the work of law.193 They may provide con-
ventions on which everyone voluntarily settles; table manners are exam-
ples. Or they may solve prisoner’s dilemmas through social sanctions
imposed on deviants; this is true of the idea that people should clean up
after their dogs. And the absence of such norms makes some existing
practices highly inefficient.

Let us begin here with the simple and conventional case of littering,
captured in the second tale.194 Under ordinary assumptions, each per-
son may well litter-if the costs of throwing things in the garbage are
wholly internalized, whereas the benefits of doing so are spread across a
wide range of people. In the standard account, “rational” individuals,
acting in their “self-interest,” will produce a great deal of litter, perhaps
so much that legal regulation ultimately is required. This idea helps ex-
plain legal responses to environmental degradation, as in the cases of
mandatory recycling, taxes on or fees for polluting activity, and command
and control regulation.

2. Puzzles. – It is illuminating to see things this way, but many ques-
tions might be raised by the standard account. The relevant changes do
not bring about Pareto improvements, which occur only when everyone
gains or at least some gain and no one loses. But in the cases at hand,
some people are losers; in fact many people may be losers. With a
mandatory seat belt law, for example, there will be people who dislike
wearing seat belts and resent the coercion, no matter the number of peo-
ple who wear them. Their objection has nothing to do with the social
meaning of wearing belts; it involves the intrinsic inconvenience of doing
so.

We should, in this light, distinguish the very simplest cases-in which
all or nearly all people favor a change in norms-from cases in which

191. See Cass R. Sunstein, Incompletely Theorized Agreements, 108 Harv. L. Rev.
1733, 1735-36 (1995); Cass R. Sunstein, Legal Reasoning and Political Conflict ch. 2
(1996) [hereinafter Sunstein, Legal Reasoning].

192. See, e.g., Ellickson, supra note 51, at 123-36; Ullmann-Margalit, supra note 51, at
18-133.

193. See Ellickson, supra note 51, at 123; Elster, The Cement of Society, supra note
26, at 125-40; Ullmann-Margalit, supra note 51, at 89-93.

194. See supra Part I.B.

1996] 955

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

there are bare majorities in favor of such a change. If everyone would
favor a situation in which people pick up after their dogs, but this result
cannot be brought about without government action, the case is easy:
Government action should be initiated. But if 65% of people would favor
the change, and 35% like the status quo, we have a harder case. Perhaps
the change would be favored under the Kaldor-Hicks criterion; but that
criterion is highly controversial.195 To decide whether government ac-
tion is appropriate, it is thus necessary to take a stand on a large question
in political theory; the fact that norm-change is involved means that the
setting is unusual, but the basic analysis is not.196

Even more fundamentally, the words “rational” and “self-interest” ob-
scure a great deal, since they take so much for granted. At first glance it
appears that there is no sharp dichotomy between rationality and social
norms or between self-interest and social norms; taking the words in their
ordinary sense, what is rational and what is in an agent’s self-interest are
functions of social norms.197 Return to our second story’98 and suppose
that there is a social norm that everyone should pick up litter.199 If the
norm is in place, people who act in their rational self-interest will not
litter. In the second tale, were the non-litterers or members of the con-
trol group acting in their rational self-interest? What is rational, and what
promotes self-interest, depends on many exogenous factors, including ex-
isting norms.200

There is a further point,201 and it raises serious difficulties for the
traditional account of the relationship between norms and collective ac-

195. See Richard A. Posner, The Economics of Justice 91-92 (1981); cf. Ronald
Dworkin, Is Wealth a Value?, in A Matter of Principle 237 (1985) (criticizing using “social
wealth” maximization as a basis for social action).

196. To know whether government action is appropriate, we also need to ask whether
government is pursuing legitimate ends and also whether the action is likely to be effective.
See infra Part VII.

197. See the qualification in supra pp. 946-47. Of course nothing I have said is
inconsistent with the view that we should adopt a tractable definition of rationality for
positive purposes and see how that definition helps in predicting behavior. What might be
adapted to this kind of approach is an understanding that the incentives created by norms
will play a role in an individual’s judgment about what is the rational thing to do.

198. See supra Part I.B.
199. Cf. Carol M. Rose, Introduction: Approaching Property, in Property and

Persuasion: Essays on the History, Theory, and Rhetoric of Ownership 1, 2-3 (1994)
(discussing role of norms in providing environmental protection).

200. For an interesting qualification, see Holmes, supra note 22, at 42-68. Self-
interest might be self-consciously defined in narrowly economic terms, in which case it is
not formal but, on the contrary, has a point-the defeat of sometimes corrosive social
norms such as honor, which can in some circumstances cause people to fight. Hence
Holmes argues that the early liberal enthusiasm for self-interest was designed not to
disparage altruism but instead to call attention to the fact that if people are pursuing
economic goals in a self-interested way, they may not do the sorts of mischief associated
with fanatics and dogmatists of various stripes. See id. at 53-62.

201. I am grateful to Ronald Dworkin for raising this problem.

956 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

tion problems.202 The difficulty is that any collective action problem can
be characterized as such only because of a wide range of norms, and not

only because of the particular norm that is said to be producing the prob
lem. The traditional account focuses on one norm, but the problem is a
product of a wide range of them.

Suppose, for example, that there is no norm against littering; that
people think that there is too much litter; and that they would like to
create a new, anti-littering norm. Would it be right to say that this is a
case involving a collective action problem that would be served best with
the aid of a new social norm, that is, a norm against littering? The state-
ment would not be false, but it would be misleading and incomplete.
What gives rise to the collective action problem is an array of individual
judgments and desires that are themselves (in all likelihood) a function
of social norms. There are, for example, norms against clutter, norms

involving certain conceptions of aesthetics, norms about public spaces. If

people “want” a new norm-the norm against littering-their desire
probably stems from many other norms, such as norms favoring clean
rather than dirty parks, norms in favor of shared rather than maldis-
tributed burdens, norms in favor of solutions through norms rather than
coercion or fines.

When a situation is supposed to create a prisoner’s dilemma that
would be satisfied by some norm Z, the situation presupposes a range of
norms A through Y which are being held constant and not being put in
contention. Then the question becomes: Why is it that norm Z (say, the
norm with respect to littering) is put into question, rather than some
other norm (say, the norm favoring clean parks)? Why should a norm be
established in favor of cleaning up after one’s dog, instead of changing
the norm producing unwillingness to be exposed to the relevant mess?
This question has yet to be addressed in existing work on collective action
and social norms. And the question has a powerful Coasian dimen-
sion.203 The traditional account takes a set of norms as given, without

seeing that they themselves might be altered, rather than altering the par-
ticular norm that has been put into question.204

Several answers might be offered. On the Coasian view, we might ask
which norms can be changed most cheaply. Just as one person in a legal
controversy might be the cheapest cost avoider, so one norm in a collec-
tive action problem might be the cheapest target of norm management.

202. See Ellickson, supra note 51, at 149.
203. See R.H. Coase, The Problem of Social Cost, 3J.L. & Econ. 1, 42-43 (1960).
204. Hence the traditional view is, roughly speaking, Pigouvian. In describing

pollution as a problem of externalities, Pigou did not see that the pollution problem is a
function notjust of the polluter’s most recent act, but also of something that is being taken
for granted, that is the numerous acts and omissions of both polluters and pollutees. See
id. at 34. In seeing a certain norm Z as creating or solving collective action problems, the
traditional view does not see that norms A through Y are also important sources of the
problems. Perhaps norm A, B or C might be altered more cheaply than norm Z could be.

1996] 957

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

It is plausible, for example, to think that it is much more efficient to
create a norm in favor of cleaning up than to create a norm making peo-
ple approve of clutter and mess. On this view, we would not inquire into
the merits or basis of existing norms, but ask more simply which norms
can be altered at the lowest cost. I speculate that an implicit judgment of
this kind lies behind the traditional approach to this problem.

Alternatively, and departing from economic criteria, we might put at
issue those norms that are not part of the relevant agents’s own deepest
convictions and self-understandings. Typically the norms thought to
solve collective action problems seem to be a form of “tinkering,” encour-

aging conduct that preserves what people believe most deeply, have

thought through, or most take for granted. Finally, we might not look to

agents’s convictions but venture instead an objective account of human
needs and human interests. On this view, we should create a norm

against littering, rather than a norm in favor of clutter and mess, because
lives are objectively better with the first norm than with the second. On
such an approach, a collective action problem exists because if reason-
able agents could agree on the norm in question, things would be better
rather than worse. It is not clear, however, that this way of seeing things
can coexist with ordinary understandings of collective action problems,
which are rooted in subjective desires. A possible conclusion of what I
have said thus far is that in the context of norms, the ordinary under-

standings face a serious conceptual problem.
3. Legal responses. – A well-functioning society needs many norms

that make it rational for people, acting in their self-interest, to avoid col-
lective action problems. When such problems exist, it is because of the
social norms that make rational self-interest take a certain form-in
favor, for example, of refusing to clean up or otherwise failing to contrib-
ute to shared projects.205 If we draw back from the puzzles just discussed,
we can make a simple point for simple cases: A large task for society and
for law is to try to inculcate the relevant norms. Effective responses pro-
mote efficiency206 and simultaneously enhance a form of freedom by pro-
ducing outcomes that citizens reflectivelyjudge best but cannot obtain on
their own.

Much legal regulation has this goal. Such regulation might even
consist of direct coercion, designed to generate good norms or to pick up
the slack in their absence. There are laws designed to ensure that every-
one picks up after their dog; that people do not litter; that people do not
smoke in certain places. These laws are rarely if ever enforced through
criminal prosecutions. But they have an effect in shaping social norms
and social meanings. They do this in large part because there is a general

205. Of course it may be possible to create property rights as a way of diminishing or
eliminating the collective action problem. See Terry L. Anderson & Donald R. Leal, Free
Market Environmentalism 3 (1991).

206. The complications from the immediately preceding discussion should be kept in
mind.

958 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

norm in favor of obeying the law. The relevant laws help to inculcate
both shame and pride; they help define the appropriate sources of these
things. (Of course people have to be brought up to have a sense of
shame, and to have a certain understanding of what sorts of things call
for shame, notably including certain paradigm cases from which analo-
gies appear.207) They readjust the personal calculation, making what is
rational, and what is in one’s self-interest, different from what they were
before.

The key point is that such a change may be supported by the reflec-
tive judgments of all or most people. When it is, there should be no
objection in principle.208 The point very much bears on the phenome-
non of norm bandwagons. People may actually reject existing norms but
fail to state their opposition publicly, and once public opposition be-
comes less costly, new norms may rapidly come into place.

B. Citizens and Consumers

John Jones, the protagonist of the fifth tale,209 is in one way quite
usual: There is an evident and pervasive difference between people’s
choices as consumers and their choices as citizens. This is because peo-
ple are choosing quite different things.210 In their private capacity, peo-
ple may watch silly situation comedies; but they may also support, as citi-
zens, the use of government resources to assist public broadcasting.
Some people seek stringent laws protecting the environment or endan-
gered species even though they do not use the public parks or derive
material benefits from protection of endangered species-and even
though in their private behavior, they are unwilling to do much to pro-
tect environmental amenities. The mere existence of certain environ-
mental goods seems to be highly valued by political participants, even if
they are not willing to back up the valuation with dollars in private mar-
kets. Of course many people give to organizations that support environ-
mental protection. But what people favor as political participants can be
different from what they favor as consumers. It is in part for this reason
that democratic outcomes are distinct from those that emerge from
markets.211

In fact a good deal of empirical work shows that people’s judgments
about politics are not a product of their self-interest, narrowly under-

207. See Churchland, supra note 128, at 272-78 (describing role of paradigm cases in
mental functioning); Charles E. Larmore, Patterns of Moral Complexity 1-5 (1987)
(discussing moral examples). Creativity in politics and law usually consists of assimilating
new cases to the paradigm ones. See Sunstein, Legal Reasoning, supra note 191, at ch. 3.

208. Some cases can be imagined in which the response would be invasive of rights
and therefore, by hypothesis, unacceptable. See infra Part VII.

209. See supra Part I.E.
210. See Sen, supra note 78, at 26-31, 34.
211. Of course there are other reasons, including the free-rider problem, since

citizens may support laws for whose costs other people are responsible.

1996] 959

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

stood. People without health care are not more likely to support laws
creating a right to health care; people recently victimized by crime are
not more likely to support aggressive policies against crime. Norms and
values are instead the principal determinant of political judgment.212

The disjunction between political and consumption choices presents
a puzzle. Would it make sense to say that consumer behavior is a better
or more realistic reflection of “actual” preferences than is political behav-
ior?213 In light of the fact that choices depend on context, and do not
exist in the abstract, the very notion of a “better reflection” of “actual”

preferences is a confusing one; there is no such thing as an “actual” (in
the sense of unitary or acontextual) preference in these settings. The
difference might be explained by the fact that political behavior reflects

judgments made for a collectivity.214 For this reason it reflects a variety of
social norms that are distinctive to the context of politics.

Because of the governing norms, citizens may seek to implement in-
dividual and collective aspirations in political behavior but not in private
consumption.215 As citizens, people may seek the aid of the law to bring
about a social state that they consider to be higher than what emerges
from market ordering. People may, in their capacity as political actors,
attempt to promote altruistic or other-regarding goals, which diverge
from the self-interested preferences sometimes characteristic of markets.
Political decisions might also vindicate metapreferences or second-order

preferences. People have wishes about their wishes, and sometimes they
try to vindicate those second-order wishes, including considered judg-
ments about what is best, through law. And norms with respect to public
discussion may impose “taxes” on public statements of various sorts-per-
haps requiring them to be “laundered,”216 perhaps inducing conformity
by punishing certain dissident views that might be reflected in other

spheres.
In all of these ways, the norms at work in democratic arenas can pro-

duce different choices from those produced by markets. It would be

wrong to say that the market choices are more “real” or “true.” The ques-
tion of which choices should govern for purposes of law and policy de-

pends on a range of contextual issues that cannot be resolved by refer-

212. See Dennis Chong, Values, Norms, and Interests in the Explanation of Social
Conflict, 144 U. Pa. L. Rev. (forthcoming 1996); David O. Sears & Carolyn L. Funk, Self-
Interest in Americans’ Political Opinions, in Beyond Self-Interest 147, 150-64, 168-70
(JaneJ. Mansbridge ed., 1990).

213. As suggested in Viscusi, supra note 65, at 4-5.
214. See Sen, supra note 78, at 29.
215. This is a possibility and only a contingent one. We could imagine a society in

which people were self-interested in political behavior but aspirational in markets or at
least civil society. What I am discussing here is a function of prevailing norms.

216. See Robert E. Goodin, Laundering Preferences, in Foundations of Social Choice
Theory 75, 86-91 Jon Elster & Aanund Hylland eds., 1986). By “laundering” Goodin
means a form of cleaning, whereby preferences, expressed in public, are stated in terms
that avoid what speakers consider illegitimate or invidious. See id. at 75-76.

960 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

ence to notions of “choice” and “preference” alone. Of course some
collective judgments might be wrong, productive of inefficiency, invasive
of rights, or confused. I take up this point below.

C. Risk, Autonomy, and Well-Being
Thus far I have tried to avoid the most controversial questions and to

build from common understandings. Let us venture now into more com-

plicated territory, connected with the ideas of well-being and autonomy.
Outside of the domain of autonomy, we might think that there are

obstacles to well-being in norms that encourage people to take high risks

simply because the failure to do so signals-in light of those norms-
cowardice or worse. Suppose that people engage in certain behavior not
because of its intrinsic value but because of the reputational cost of fail-

ing to do so. We are dealing, then, with classes of cases in which the

danger accompanying choice means that intrinsic utility is not high but

risk-taking behavior persists because of social norms. Suppose too that

people dislike the norms that generate that reputational cost, or that on
reflection, people would dislike those norms. If so, norms that en-

courage people to carry guns, use dangerous drugs, drive well over the

speed limit, engage in unsafe sex, and so forth may properly be an object
of governmental attack, because of their potentially pernicious effects on

people’s lives. Information and persuasion are probably the best tools for

government to use. But in some cases economic incentives or coercion

may be justified.
Of course a liberal society might also want to ensure that all of its

citizens are autonomous.217 For the moment let us understand that no-
tion in a way that leaves open many questions but that will be helpful for
our limited purposes here. A citizen can be understood as autonomous
insofar as she is able to choose among a set of reasonably good options
and to be reflective and deliberative about her choice.218 A society can
be understood as self-governing, and as politically autonomous, to the
extent that its citizens face a range of reasonably good options and exer-
cise capacities of reflection and deliberation about their choice.

It should be clear that social norms, meanings, and roles may under-
mine individual autonomy. Above all this is because norms can compro-
mise autonomy itself, by stigmatizing it. People may believe, on reflec-
tion, that the act of being well-educated should not be a source of shame;
but in some communities, a good deal of education may be inconsistent
with prevailing social norms.219 Or exposure to diverse options, and re-

217. A form of liberal perfectionism is defended in Raz, supra note 22, at 369-429.
218. Cf. Anderson, supra note 23, at 141-43 (discussing autonomy and market

choice).
219. This is very much true with respect to employment for women. See, e.g., Martha

Chen, A Matter of Survival: Women’s Right to Employment in India and Bangladesh, in
Women, Culture, and Development: A Study of Human Capabilities 37, 37-50 (Martha C.
Nussbaum & Jonathan Glover eds., 1995).

1996] 961

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

flection about which is best, may seem inconsistent with existing norms.
In such cases autonomy cannot easily220 exist without collective assist-
ance; people are able to produce the norms, meanings, and roles that

they reflectively endorse only with governmental involvement. Some-

thing must be done collectively if the situation is to be changed.
To promote autonomy, a society might seek to ensure that everyone

has a minimal degree of education, a certain level of exposure to diverse

conceptions of the good, and what might be considered the material ba-
ses of autonomy: food, shelter, and freedom from criminal violence.221
In modest forms this project is fully compatible with political liberalism;
perfectionist liberals might insist on a good deal in this vein in order to
allow people to be (more or less) masters of the narratives of their own
lives.222 In either case, social norms can undermine the liberal project,
and government might try to alter them in order to promote autonomy.
People may have adopted prevailing norms rationally in view of limits in

existing opportunities, as when norms discourage schooling for people
whose prospects are poor;223 but such norms are nonetheless an obstacle
to autonomy.

A government that seeks to promote autonomy might well work

against efforts by subcommunities to require conformity to a single defin-

ing creed. In fact conflicts between antidiscrimination principles and

religious liberty have everything to do with perceived limits on govern-
mental ability to change norms, meanings, and roles in subgroups that

deny autonomy. These conflicts are generally resolved in favor of the
latter, especially in the area of sex equality. But if we attend to the auton-

omy-denying effects of norms and meanings, it might well make sense to
resolve the conflicts against subgroups, even religious ones.

D. Caste

Social norms, social meanings, and social roles bear very much on

problems of discrimination. For someone who is African-American, fe-
male, or handicapped, existing meanings are possible obstacles to auton-

omy or well-being. The social meaning of being female, for example,
may bear on a range of choices at home and in the workforce. So, too,
for the social meaning of particular choices by women, which may di-

verge sharply from similar choices from men: not to have children, to

220. Of course some people may be able to defy existing norms against autonomy.
221. See Raz, supra note 22, at 407-08.
222. See id. at 408.
223. Cf. Elster, supra note 101, at 109-40 (describing adaptive preferences); Sen,

Commodities and Capabilities, supra note 130, at 81-83 (in India, the “perception of relative
needs of different members of the family may be closely related to social influences”). Just
as preferences can be adaptive to an unjust status quo, so can norms and meanings be a
predictable outgrowth of limited autonomy or heteronomy.

962 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

work as an airline pilot, to hate cooking, to dress in a certain way.224 We
have seen that a status is a particular kind of role; statuses are imposed
rather than chosen. Thus statuses are an occasional target of law. In par-
ticular, law might attempt to overcome a collective action problem faced
by victims of discrimination in changing prevailing norms, meanings, and
roles.

In American law, problems of discrimination typically raise issues of
caste. We might say that we have a system with caste-like features when a
highly visible and morally irrelevant factor is turned, by social and legal
practices, into a systematic source of societal disadvantage.225 An impor-
tant and disastrous feature of this situation is the signalling effect of the
characteristic that is shared by lower caste members. That characteristic
promotes a certain social role for caste members, since it is associated
with a range of undesirable or otherwise stigmatizing traits. Often an at-
tack on a caste system amounts to an attack on that social role and its
associated social norms-especially as a result of behavioral norms
shared, or thought to be shared, by members of the lower caste.226 For
lower caste members, the problem is that the shared characteristic carries
with it a meaning-stupidity, passivity, venality-that cannot be con-
trolled by individual agents.

Laws forbidding discrimination might well have, as one of their pur-
poses and effects, a change in the signalling effect of the relevant charac-
teristic. They might be designed to encourage people to be self-con-
scious about registering the preexisting signal.

Suppose, for example, that it was thought important to alter social
norms about gender relations. The social role of “being a woman” is asso-
ciated with a wide range of social norms and social meanings. There are
many examples from the present and recent past. Thus it may be that
women do the majority of domestic labor. In these circumstances, a man
who does most of the domestic labor might seem odd, or in some way
woman-ish, and a woman who asks for something like equality in domes-
tic labor might seem odd, selfish, or in some way man-ish.227 Or a woman
in her fifties might be seen as fundamentally different from a man in his
fifties, because of social norms associated with gender. Or a single wo-
man might be stigmatized, or inquired about, in ways fundamentally dif-
ferent from what happens to a similarly situated man. Or a woman who

224. For what might be seen as an argument in favor of changing the social meaning
of biological characteristics, see Case, supra note 71, at 9-36.

225. See Cass R. Sunstein, The Partial Constitution 338-46 (1993); Cass R. Sunstein,
The Anticaste Principle, 92 Mich. L. Rev. 2410, 2428-33 (1994).

226. Note in this regard Richard Wright’s suggestion that unlike most blacks in the
south, he would “act straight and human.” See Richard Wright, Black Boy: A Record of
Childhood and Youth 162 (1937). Wright says that most blacks, even those who felt great
resentment, pretended to accept their inferior position. See id. at 160-62.

227. Cf. Zelizer, supra note 59, at 36-70 (discussing “dole” and “allowance” in context
of use of domestic money by women).

1996] 963

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

complains about apparently mild forms of sexual harassment might seem
to be a radical, a troublemaker, or someone without a sense of humor.

For lower caste members, a wide range of “choices” might emerge
from the underlying social norms. These choices might reflect adapta-
tion by lower caste members to existing injustice; they might be a product
of the social opprobrium attached to violation of social norms by lower
caste members.228 The choices might even be called “preferences”;229
certainly desires can be affected. But many women believe on reflection
that the social meaning of being a woman is bad for them and that it
should be changed. These women face a collective action problem that
may be solved best via law.230 In any case a caste system tends to deny
autonomy to lower caste members. Thus we have a particular reason for
legal intervention: to overcome the obstacles to autonomy produced by a
system of caste, obstacles that cannot be overcome by individual agents.

This is simply a stylized discussion of the problems faced by people
who live within a caste system, and who might seek to enlist the law to
make things better.

E. The Expressive Function of Law

Many laws have an expressive function. They “make a statement”
about how much, and how, a good or bad should be valued. They are an
effort to constitute and to affect social meanings, social norms, and social
roles. Most simply, they are designed to change existing norms and to
influence behavior in that fashion.

Of course human goods are valued in different ways; people have a
wide variety of evaluative stances toward relationships and goods.231 Laws
with expressive functions are often designed to promote a certain way of
valuing certain goods. Many such laws are intended to say that specified
goods should be valued in a way that deters thinking of them as mere objects for
use. Laws forbidding the purchase and sale of certain goods can be so
understood. A ban on the sale of children is designed (among other
things) to say that children should be valued in a way that forbids the
acceptance of cash as a reason for taking them out of parental care. A
ban on vote-selling can be viewed similarly. We might understand such a
law as an effort to make a certain statement about the pricelessness-not

228. See Kuran, supra note 17, at 196-204 (discussing preference adaptation of lower
caste members in India).

229. See Becker, supra note 31 (adopting an economic approach to family
relationships and choices); cf. Amartya K. Sen, Gender and Cooperative Conflicts, in
Persistent Inequalities 123, 126-28 (Irene Tinker ed., 1990) (describing social factors that
lead women to adapt to, and accept, injustice).

230. See Kuran, supra note 17, at 134 (“Because society will generally ostracize anyone
who abandons the caste system, the potential member of an anticaste colony is likely to
withhold his participation until it appears likely to succeed. With other potential members
reasoning likewise, the colony will remain unformed.”).

231. See Anderson, supra note 23, at 1-16.

964 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

the infinite value-of the right to vote. In the environmental area, de-
bates over market valuation are partly debates over this question.232
Some people think that market valuation should be viewed with suspi-
cion; it threatens to affect social norms and social meanings in undesir-
able ways.

Laws with expressive justifications are most plausibly defended on
the ground that they will in fact affect social norms and move them in

appropriate directions.233 But at this stage there are empirical questions:
Do laws really affect social norms and social meanings? Under what con-
ditions? Much work remains to be done on these important questions.

VII. BLOCKED GROUNDS

What I have suggested here should unsettle some common under-

standings about government “paternalism” and “meddling.” If private
choices are a function of roles, norms, and meanings over which private
people have no sovereignty, many imaginable initiatives are consistent
with individual autonomy, rightly conceived. But this conclusion ought
not to suggest that government should be licensed to do whatever it
wishes. There should be firm limits on government’s efforts to engage in
norm management.

Of course government action should often be rejected on simple
pragmatic grounds-because, for example, it is likely to be futile or

counterproductive.234 The “just say no” campaign with respect to drug
use probably falls in this category. Perhaps efforts to stigmatize teenage
smoking will backfire and make smoking seem bold or glamorous. Per-

haps government has mistakenly concluded that there is a collective ac-
tion problem calling for governmental response. Perhaps most people
are happy that littering is not stigmatized.

It is also true that government interference with norms, roles, or

meanings might compound a collective action problem, aid well-organ-
ized private groups promoting unjust goals, or aggravate a caste-like situa-
tion. Imagine an effort to promote the use of cigarettes, alcohol, or drugs,
or to discourage the buckling of seat belts, or to increase the opprobrium
associated with the role of being a homosexual. Nothing I have said sug-

232. See Steven Kelman, What Price Incentives? 52-83 (1981).
233. See Zelizer, supra note 27, at 16-17 (discussing prohibitions on life insurance).

In making these claims, I do not endorse the use of law to “make statements”
independently of consequences. Though I cannot defend the point here, official use of
law merely to “make statements” risks fanaticism and pointlessness. See Sunstein, supra
note 64. Indeed, emphasis on consequences-and on the empirical questions raised by
arguments over consequences-has the salutary function of reducing the volume and the
heat raised by policy disputes. It does so by shifting debate from basic values to issues that
are, at least in principle, subject to empirical testing. I believe that this is an important
social function of approaches to law and policy that are based on consequences; it is also
part of the appeal of utilitarianism.

234. Cf. Raz, supra note 22, at 428-29 (discussing pragmatic objections to pursuit of
perfectionist policies).

1996] 965

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

gests that government is not properly criticized when it engages in activi-
ties of this sort. But any such criticism should be on the merits, not on
the ground that government may not interfere with private preferences
or choices.

There is a final and especially important point. A liberal society lim-
its the permissible bases for governmental action.235 Ordinarily people
describe these limits as “rights.” Rights operate as constraints on norm
management as on everything else that government does. A full account
of these limits would be far too ambitious for an article of this sort; but a
few notes will be helpful.

Some government action designed to change norms, meanings, and
roles might be based on religious grounds; these should be banned as
reasons for public action. At least in the American constitutional system,
for example, it is unacceptable for government to attempt to legislate on
the ground that the divinity of Jesus Christ requires a certain state of
affairs.

It would be equally unacceptable to base government action on
grounds that deny the basic equality of human beings-as in efforts to
encourage norms that treat members of racial minorities as second-class
citizens. Many efforts to change norms would impair autonomy rather
than promote it. If government attempted to promote norms ensuring
that women occupy traditional roles, its attempt would be illegitimate-
not because it is an attempt at norm management, but because the at-
tempted management invades rights of sex equality.

In any case some human interests are properly denominated rights,
and efforts to change norms, meanings, and roles should not be allowed
to invade rights.236 Many imaginable efforts ought to be rejected because
of this risk. Consider, for example, a suggestion that the meaning of re-
fusing to allow government officials into your home is now “personal
courage and independence”-accompanied by the not implausible
thought that things would be better if the meaning were “unpatriotic un-
willingness to cooperate with the crime-fighting effort,” culminating in a
proposal that all people should be required to open their homes to the
government. There is a collective action problem here. But if it is be-
lieved that people should have a right to keep government officials from
their homes, this proposal should be rejected.

Political liberals go further and urge rejection of any ground for ac-
tion that is based on a “comprehensive view.”237 Of course there are

235. For discussion, see, e.g., Dworkin, supra note 11, at 266-78; Larmore, supra note
207, at 42-68; Raz, supra note 22, at 400-29.

236. I do not mean to say anything here about the status or basis of rights.
237. See Larmore, supra note 207, at 40-90 (arguing that liberal government must

remain neutral by not selecting among divergent “conceptions of the good life”); Rawls,
supra note 190, at 58-66, 133-72 (describing political principles on which those with
diverse “comprehensive views” can rely).

966 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

SOCIAL NORMS AND SOCIAL ROLES

many complexities in this claim.238 What is important for present pur-
poses is that on any sound view of liberalism, there is no general basis, in

principle, for objection to proposals of the sort I have suggested. Political
liberals ought to acknowledge, for example, that social roles and social

meanings may undermine the equality and liberty of citizens and that

changes require collective action. The constraints imposed by political
liberalism impose no bans on those changes.

CONCLUSION

Many claims about the appropriate limits of law are insufficiently ap-
preciative of the pervasive effects of social norms, social meanings, and
social roles. In fact these effects have yet to receive much attention.239
But-to say the least-the impact of law on human behavior is an impor-
tant matter for lawmakers to understand, and that impact has everything
to do with social norms. An understanding of norms will therefore help
illuminate effective regulatory policy. Many of the most dramatic gains in
health and safety policy are a product of changes in norms, meanings,
and roles. Many of the most severe problems in current societies are a

product of unfortunate norms, meanings, and roles.
Norms relate to some broader issues as well. Often it is said that the

common law, and a liberal regime dedicated to freedom, take “prefer-
ences” as they are and do not seek to change them. But the term “prefer-
ences” is highly ambiguous. If the term is meant to refer to “choices,” it
should be understood that choices are very much a function of context,
including governing norms, meanings, and roles. Certainly the particular
choices made by people in markets-in their capacity as consumers or
laborers-do not suggest global or acontextual valuations of relevant

goods. If the term “preferences” is meant to refer not to choices but to
the motivational and mental states behind choices, it is important to rec-

ognize that those mental states include assessments of social norms, the

expressive meaning of acts, and the expectations associated with a daz-

zling variety of social roles. Norms and roles affect both public action
and public talk, in ways that can much disguise how people think pri-
vately. This point has large implications. In many settings, it would be
best to dispense with the idea of “preferences,” and to shift instead to
more concrete ideas, including intrinsic value, reputational effects, and
effects on self-conception.

238. See Rawls, supra note 190, at 133-72.
239. The principal contributions are Dan M. Kahan, What Do Alternative Sanctions

Mean?, 63 U. Chi. L. Rev. (forthcoming 1996) (discussing how norms affect the public
assessment of criminal punishment); Lessig, supra note 24 (arguing that law should
account for, and regulate, social meaning); and from a different direction, Ellickson, supra
note 51 (discussing norm of neighborliness in land-use bargaining); see also Kuran, supra
note 17 (describing role of “preference falsification” in reflecting and shaping social
norms).

1996] 967

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

COLUMBIA LAW REVIEW

We have also seen that norms can be far more fragile than they ap-
pear. Hence “norm entrepreneurs” can help solve collective action

problems, and hence “norm bandwagons” and cascades are common.
Sometimes law interacts with the efforts of norm entrepreneurs, facilitat-

ing or blunting their efforts, and sometimes law ratifies or accelerates-
or halts-norm bandwagons and cascades.

While social life would be impossible without norms, meanings, and
roles, individual people have little control over these things. The result
can be severe limits on human well-being and autonomy. Certainly there
is a problem with existing norms when all or almost all people would seek
a change. There may well be a problem when reputational incentives
lead people to do what they would otherwise refuse to do, at least if the
relevant norms deny people the preconditions for autonomy or otherwise
undermine well-being. In fact lives are shortened and unjustified in-

equalities are perpetuated by the existence of many current norms.

People need collective help if they want to change norms, meanings,
or roles. Collective help may be futile or counterproductive; it may be

illegitimately motivated. But these matters require an inquiry into the

particular context. The issue should not be foreclosed by resort to con-

fusing claims about the need to respect private choice.

968 [Vol. 96:903

This content downloaded by the authorized user from 192.168.72.232 on Thu, 22 Nov 2012 21:25:38 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Article Contents

p. 903
p. 904
p. 905
p. 906
p. 907
p. 908
p. 909
p. 910
p. 911
p. 912
p. 913
p. 914
p. 915
p. 916
p. 917
p. 918
p. 919
p. 920
p. 921
p. 922
p. 923
p. 924
p. 925
p. 926
p. 927
p. 928
p. 929
p. 930
p. 931
p. 932
p. 933
p. 934
p. 935
p. 936
p. 937
p. 938
p. 939
p. 940
p. 941
p. 942
p. 943
p. 944
p. 945
p. 946
p. 947
p. 948
p. 949
p. 950
p. 951
p. 952
p. 953
p. 954
p. 955
p. 956
p. 957
p. 958
p. 959
p. 960
p. 961
p. 962
p. 963
p. 964
p. 965
p. 966
p. 967
p. 968

Issue Table of Contents

Columbia Law Review, Vol. 96, No. 4 (May, 1996), pp. 809-1092
Front Matter
The Consent Paradigm: Tribal Sovereignty at the Millennium [pp. 809-902]
Social Norms and Social Roles [pp. 903-968]
Notes
Putting Precedent in Its Place: Stare Decisis and Federal Predictions of State Law [pp. 969-1018]
Villainous Verdicts? Rethinking the Nineteenth-Century French Jury [pp. 1019-1061]
Review Essay
Review: A New Class of Lawyers: The Therapeutic as Rights Talk [pp. 1062-1092]
Back Matter

Turn in your highest-quality paper
Get a qualified writer to help you with

“ Psych Term Paper ”

Get high-quality paper

Guarantee! All work is written by expert writers!

Still stressed from student homework?

Get quality assistance from academic writers!

Order now