identity and schooling: some lessons for the economics · pdf fileidentity and schooling: some...

Journal of Economic Literature Vol. XL (December 2002) pp. 1167–1201

Identity and Schooling:Some Lessons for the Economics

of Education

GEORGE A. AKERLOF and RACHEL E. KRANTON1

1167

1. Introduction

T HIS REVIEW culls noneconomic litera-ture on education—by sociologists, an-

thropologists, reformers, and practitioners—to present a new economic theory ofstudents and schools. This theory integratesa sociological view of education with eco-nomic analysis. In a classic economic model,students choose effort (or time) in school tobalance its discounted return with its oppor-tunity cost, and resources determine schoolquality.2 Economists now often append so-cial interactions to this model but rarely ex-amine the nature of these interactions. Byreviewing literature outside of economics,we develop a theory where a student’s pri-mary motivation is his or her identity and the

quality of a school depends on how studentsfit in a school’s social setting.3 This theoryoutlines fresh directions for research in theeconomics of education.

With the focus on identity and schools,this review gives a new perspective on the al-location of resources in education. We oftensee vast differences in educational outcomeswith little difference in quantity of re-sources. These differences remain a puzzleto economists; as one of our referees argued,“economists do not have good models for ex-plaining why school resources do or do not(as often found) affect the returns to school-ing.” The framework we develop might fillthis gap. We translate key sociological con-cepts into an economic model of studentsand schools. At the forefront are a school’ssocial setting and students’ backgrounds. Weshow how sociological variables may affecteducational outcomes, and how resourcesmay change the returns to different policies.

The literature we review represents a ma-jor (in our view, the predominant) line in thesociology of education. The concepts weadopt—ideal type, identity, and social cate-gory—are as basic to sociology as supply and

1 Akerlof: University of California, Berkeley;Kranton: University of Maryland. We thank CydFremmer, Nisha Malhotra, and Eric Verhoogen for in-valuable research assistance, Robert Akerlof, PaulBeaudry, Pierre Fortin, Abdeslam Maghraoui, CraigRiddell, and Janet Yellen for help and comments, andJohn McMillan and three anonymous referees for theircritical review which helped us greatly improve the pa-per. Akerlof is grateful to the Canadian Institute forAdvanced Research, the MacArthur Foundation, andthe Brookings Institution for financial support.Kranton thanks the Institute for Advanced Study,where she spent 2001–2002 as a Deutsche BankMember of the School of Social Science, for its hospi-tality and financial support.

2 We review the economic literature below.

3 Thus we follow Judith Harris (1998), who writes“children want to be successful children,” and JamesColeman (1961), who argues adolescents want to fit in.

demand to economics. In schools of educa-tion, the syllabi of core courses are filledwith works that apply these concepts toschooling.4 Of course, there is now much so-ciology that is indistinguishable from eco-nomics. Economics, however, has notadopted core sociological concepts. Thisomission may be for good reason: we haveheard many economists say these conceptsare too fuzzy to inform either theory or em-pirical work. Nonetheless, it might beworthwhile to consider how central socio-logical concepts may explain outcomes thatcannot be explained by a standard economicmodel. Without a model that mirrors this so-ciology, economic analysis produces onlypartial answers to key questions. For exam-ple, what is the impact of resources on aca-demic attainment? What are the importantelements of school reform? Economists cananswer whether additional resources en-hance schooling, but not when these re-sources will be effective, or why. The mod-els below give tentative answers anddirections for empirical investigation. Theeffectiveness of resources will depend uponinteraction between the resources used andstudent formation of academic identity.

Our interpretation of the noneconomicliterature relies on previous work, Akerlofand Kranton (2000), which advances a no-tion of utility where an individual’s identity,or sense of self, is salient. An individual gainsutility when her actions and those of othersenhance her self-image. Furthermore, self-image, or identity, is associated with the so-cial environment: People think of them-selves and others in terms of different socialcategories. Examples of social categories in-clude racial and ethnic designations, and inthe school context include, for example,“jock” and “nerd.” Prescriptions give theideal, or stereotypical physical attributes and

behavior, of people in each category.Individuals then gain or lose utility insofar asthey belong to social categories with high orlow social status and their attributes and be-havior match the ideal of their category. Weuse this basic model of utility to understandstudent behavior in schools.

In this introduction we first describe ourreadings in the noneconomic education lit-erature and our interpretation of it. We thendiscuss what may be missing from the eco-nomic literature and why the inclusion ofthese sociological variables could make a difference to economic analysis.

1.1 Outline and Interpretation of Noneconomic Literature

We begin with James Coleman’s (1961)study of ten Illinois high schools in the late1950s, Adolescent Society. Coleman pio-neered the research on students’ socialarrangements. From student questionnaires,Coleman found that students literally dividethemselves into social categories—categoriesthat should be familiar to anyone who has at-tended a United States high school: “nerds,”“jocks,” “leading crowd,” “burnouts,” andsuch. Associated with each category is anideal in terms of physical attributes and be-havior. Coleman found that these categoriesinfluenced academic performance. Follow-ing Adolescent Society, numerous scholars,especially Penelope Eckert (1989), RichardEverhart (1983), and Douglas Foley (1990),have described student motivation and socialdivision in fine ethnographic detail.5

We depict this work in a model where stu-dents maximize utility by making twochoices. They choose their social category,and they choose effort in school.6 Whenchoosing categories, students try to fit in:They consider the match between their own

1168 Journal of Economic Literature, Vol. XL (December 2002)

4 See, for example, Harvard University’s ProfessorSara Lawrence-Lightfoot’s syllabus on the sociology ofeducation: http://icommons.harvard.edu/ ~gse-a107/syl-labus/Syllabus.html.

5 For antecedents, see August Hollingshead (1949).6 We use the word “choice” advisedly. As in classic

utility theory, we do not presume that individuals areconscious of the reasons for their choices. See Akerlofand Kranton (2000) for discussion.

characteristics and the ideal characteristicsof jocks, burnouts, and nerds. When choos-ing effort, students also try to fit in. Theyconsider the match between their own ac-tions and the ideal behavior of their chosencategories. Ethnographies of high schoolsreveal that social divisions and students’identities can be a dominant influence onachievement.

We then consider the view—held by histo-rians, contemporary sociologists, anthropolo-gists, and educators—of schools as institu-tions, with social goals.7 In this view, schoolsnot only impart skills. Schools also impart animage of ideal students, in terms of charac-teristics and behavior.8 School rituals—peprallies, home-room announcements, assem-blies—and day-to-day interactions in class-rooms, hallways, and gymnasiums reveal thenature of this ideal. Teachers, administrators,and coaches praise and reward some stu-dents, while they disapprove of and punishothers. These features and occasions definewhat we call a school’s social category and itsideal student. Students with backgroundssimilar to this category readily identify withthe school. Others, however, do not fit in soeasily. Sociologists consider these interac-tions important, but economic research hasyet to examine this setting.

We first consider assimilation in Americanschools of the early twentieth century, as de-scribed by historians. In the model, schoolspromote a single social category, and stu-dents have two choices. They choosewhether or not to adopt the school’s socialcategory, and they choose their effort. Thehistorians see these choices as particularlypoignant for students whose backgrounds

conflict with the school’s ideal. The studentsare likely to resent the implication, inherentin their schooling, that there is somethingwrong with their backgrounds, and by exten-sion, that there is something wrong withthemselves. To avoid a loss in self-image, astudent rejects the school and consequentlyexerts low levels of effort.9 Studies as re-cently as the 1990’s reveal similar problemsamong today’s African-American students,Hispanic students, and other minorities.10

We use the same model to capture educa-tors’ views of contemporary United Statespublic schools. This view, best exposited byArthur Powell, Eleanor Farrar, and DavidCohen (1985), is that the typical U.S. highschool today is “Shopping Mall High.” In thelanguage of our framework, such schools failto promote a particular social category.11

Instead, principals and teachers preach tol-erance, and students are allowed widechoice in classes and curricula. Educatorssay that such laxity produces academicmediocrity. In our interpretation, school ad-ministrators face a trade-off between pro-moting a single ideal or offering students achoice. With choice, more students in a di-verse population find ways to identify withthe school. The outcome is more “demo-cratic” in that more students are engaged inthe school, but some students acquire lowerskills.

Akerlof and Kranton: Identity and Schooling 1169

7 Thus Allan Bloom (1987, p. 26) writes that “Everyeducational system . . . wants to produce certain typesof human beings.”

8 Anthony Bryk, Valerie Lee, and Peter Holland(1993, pp. 146–47), for example, describe theStatement of Philosophy of St. Ignatius School, whichwe discuss below. Allen Peshkin (1985) provides thedescription of the ideal student at a fundamentalistChristian school.

9 Pediatricians Andrea Bonny et al. (2000) found that“school connectedness,” the extent to which studentsidentify with their schools, is an indicator of safer be-haviors and better health outcomes. School connected-ness was a better predictor than “family connected-ness.” In this paper, we explore the various aspects ofschooling that can affect school connectedness andthereby academic achievement.

10 A recent article in the Harvard Education Letter,for example, tells how researchers now consider the“connectedness” or relationship between students andtheir schools as central to the gap in test scores be-tween whites and other students (Michael Sadowski2001).

11 Powell, Farrar, and Cohen (1985) describe thetypical U.S. public high school as Shopping Mall Highbecause of the great deal of choice given to studentsand the failure of the schools to impose educational values.

We then consider school reform programsand the difference between public and pri-vate schools. Educators argue that legal in-novations and changes in social attitudes inthe 1960s and 1970s have made it difficultfor public schools to promote standards forstudents’ behavior.12 In contrast, privateschool administrators and teachers spendconsiderable resources to delineate pre-scriptions for student behavior and ensurethat students identify with the school and itsideals. Successful experiments in publicschool reform, such as the Central Park EastElementary and Secondary Schools inHarlem and the Comer Schools in NewHaven, have similar strategies. We see theseschools, in our economic terminology, as in-vesting in students’ self-images and relation-ships with the school. The schools reducethe initial social differences among the stu-dents and create a community, with an idealof academic excellence.13

1.2 The Economics of Education: A BriefSummary and Two Missing Pieces

Here we review the economic literature,with special attention to the influence of sociology on economic research and an

assessment of what is still missing from economic analysis.

Since Theodore Schultz (1960) and GaryBecker (1964) introduced the concept of hu-man capital, economists have been con-cerned with resources devoted to educationand the return to education. A large body ofempirical work has examined the impact ofresources on education outcomes. Promi-nent recent examples include David Cardand Alan Krueger (1992a,b), Julian Betts(1995), Ronald Ferguson (1998b), EricHanushek (1996), Caroline Hoxby (2000),Krueger and Diane Whitmore (1999). Thesestudies have variously viewed resources interms of school expenditures, teacher-stu-dent ratios, and teacher quality, and the re-turns to education in terms of earnings, aswell as other intermediate measures such asenhanced test scores, continuation rates,and rates of college application. As we willdiscuss further in the conclusion, the socio-logical literature, in contrast, is remarkablysilent on the return to resources devoted toeducation. The framework we develop sug-gests a synthesis of the sociological and eco-nomic viewpoints, providing insights intohow resources can be effectively deployed.

The delivery of education services andmarket structure is a second classic questionin the economics of education. In Free toChoose, Milton Friedman and Rose Fried-man (1980) propose privatization and schoolvouchers as an antidote to the agency prob-lems inherent in governmental monopolyschools. Economists have since consideredboth theoretically and empirically the impli-cations of market structure on educationalattainment. Empirical studies have mea-sured the relative effectiveness of public andprivate schools (e.g. William Evans andRobert Schwab 1995) and asked whethercompetition increases school quality (e.g.Hoxby 1996). On the theoretical side,Charles Manski (1992) and Dennis Eppleand Richard Romano (1998) build models ofprivate and public schools and ask howvoucher policies affect the set of students


12 We discuss the legal innovations below. For dis-cussion of the social climate, see Gerald Grant (1988).

13 In their textbook for educators, Edward Wynneand Kevin Ryan (1997) emphasize the role of commu-nity in a school. They claim that “proper student con-duct is enhanced by a code of conduct that not onlyprohibits wrongdoing, but also encourages students todo things that immediately help others.” Their check-list for school principals of methods to promote charac-ter shows concretely some of the ways in which schoolscan enhance a sense of community. These include suchitems as academic team competitions in or amongschools (e.g., math or spelling bees), band or choir,cheerleading, classroom or building clean-up, classmonitors, messengers, hall guards, and office assistants,crossing guards and patrol duty, community service,dramatic presentations, fund-raising, clubs, sports,school newspaper, peer tutoring, well-organized aca-demic group projects, library aides, athletic or sports-manship awards, certificates, mention in school news-paper, mention in newsletter to parents, mention overpublic address system, mention on report card, notehome to parents, pep rally, posting name or photo, goldstar or sticker, and special jackets or garments (pp.xxiii–xxiv).

that attend each type of school. The latterpaper considers peer group effects in educa-tional attainment; private schools offer fel-lowships to high-quality students becausethey enhance the learning of other students.Of course, peer group effects are a promi-nent topic in sociology and are now a well-researched area within economics.14 How-ever, as we have already noted and willdiscuss extensively below, the sociologicalliterature indicates that the difference be-tween private and public schools goes be-yond peer group effects and the selection(or self-selection) of students. Privateschools pursue different social goals andhave greater freedom to invest in the iden-tity of their students.

Following Charles Tiebout (1956), schoolfunding is a third topic that has occupiedeconomic researchers. Roland Benabou(1993) and Raquel Fernandez and RichardRogerson (1996), for example, offer theo-retical studies of community formation andits impact on school funding. Recently,economists have considered a key sociologi-cal variable: ethnic homogeneity. ClaudiaGoldin and Lawrence Katz (1997) find thatethnically homogeneous communities ledthe “high-school movement” in the earlytwentieth-century United States. More ho-mogeneous communities had higher schoolfunding and greater high-school atten-dance. Alberto Alesina, Reza Baqir, andWilliam Easterly (1999) show the continuedpositive impact of homogeneity on schoolfunding in the United States, and Edward

Miguel (1999) has shown that tribally di-verse school districts in Kenya have lowerlevels of funding.15 The sociological litera-ture we interpret gives a window on basicmicroeconomic effects of ethnicity on thesupply and demand for education. The mo-tivation of students themselves may derivefrom their ethnic background and thematch between themselves and schools.Schools as institutions may themselves pro-mote a particular social category and idealstudent; they may also adjust their policiesto accommodate an ethnically diverse stu-dent population.

Other papers in economics have consid-ered less traditional questions in education,such as ideology imparted by schooling.Samuel Bowles and Herbert Gintis (1976)argue that the U.S. educational system wasdesigned to produce compliant workers.Michael Kremer and Andrei Sarychev(2000) build a political economy modelwhere schools inculcate their students withan ideology. In a democracy, the populacemay vote for a public school system (andagainst school choice), in order to promoteideological homogeneity.

John Bishop’s (1998) paper on students’social norms comes closest to the sociologyand modeling we present below. His work,like ours, focuses on individual student moti-vation in the school’s social context. He askswhy social norms favor athletic over aca-demic performance. He argues that aca-demic success rewards only the individual,while athletic success rewards students in theschool as a whole. Hence, students would co-operate in learning, rather than harass thosewho achieve, if academic rewards dependedon outside testing and competition. The soci-ology we discuss below, and our modeling ofit, paints a more microeconomic picture ofstudent motivation, as derived from students’identities. We examine how social categories,


14 A body of work in economics considers schoolsand peer group effects on educational outcomes: e.g.,Hanushek (1971), Anita Summers and Barbara Wolfe(1977), Evans, Wallace Oates and Schwab (1992),Alejandro Gaviria and Steven Rafael (1997), DavidLevine and Gary Painter (2000), and Bruce Sacerdote(2000). These authors and others attempt to disentan-gle the impact of peer behavior on achievement fromother student and school characteristics, where thepeer group is often defined at the school level. Themodeling below emphasizes the social characteristicsthat lead students to choose different peer groupswithin a school, and how school policies affect thischoice.

15 Miguel also provides a theoretical model in whichheterogeneity reduces school quality, and thereforelowers the return to funding.

and the prescriptions for those categories, af-fect academic achievement, as well as howschools can influence students’ choices.

While economists have begun to examinemany substantive areas of the sociology ofeducation, our reading of the literature indi-cates that two related themes have thus fareluded economic analysis. First is a sociolog-ical view of the student as the primary deci-sion maker. Second is the conception of theschool as a social institution. In sections 2and 3 below, respectively, we elaborate onthese themes. Each section has (i) a discus-sion of the noneconomic literature, (ii) a ba-sic model that gives a guideline of how eco-nomic theory could and should capture therelevant sociology, and (iii) a discussion ofexisting evidence that supports the model-ing. In section 4 we integrate these sociolog-ical themes into an economic analysis of re-source allocation within schools. Section 5considers black-white differences in educa-tion. Section 6 reviews the implications ofthe theory in this paper for empirical re-search, and section 7 concludes.

2. Identity and Utility

We begin by modeling the sociology ofhigh-school life presented by Coleman and hisfollowers. We posit a model of student prefer-ences where self-image, or identity, is salient,based on the general model of utility inAkerlof and Kranton (2000). Much of the re-search in social psychology that is the basis forthis model was conducted with school-agechildren as subjects, thus making it appropri-ate to adapt this model to the field of educa-tion. We describe our model along with a briefsummary of this social psychology research.

2.1 A Student Utility Function

We construct a model of a student’s utilitywhere identity, or self-image, is salient. As ina standard model of education, a student’s

utility will depend on her effort in schooland the pecuniary returns to this effort. Letei be individual i’s effort in school, and letk(ei) be i’s skills, or human capital.16

Typically, the pecuniary benefits of ei are thereturns to skills in the labor market.Pecuniary costs include i’s foregone leisureor cost of effort.17 A standard utility functionwould posit i’s utility is a function of incomeand effort: Ui = Ui(w . k(ei), ei), where w isthe wage rate per unit of skill.

We now add identity to this utility func-tion. As in Akerlof and Kranton (2000), webegin with a set of social categories C.Again, these categories may include racialand ethnic designations, gender, and in theschool context include, for example, “jocks”and “nerds.” Prescriptions P give the idealcharacteristics and behavior for each cate-gory. An individual i is assigned to a cate-gory, and we denote this assignment ci.

18 i’sself-image, Ii, depends on the match be-tween her behavior and characteristics withthe ideals for her category. Student i’s util-ity is then Ui = Ui(wk(ei), ei, Ii), whereIi =�Ii(ei, ci; ei, P) and ei are i’s characteristics.We use the word identity to describe both astudent’s assigned category and the payoffsassociated with self-image. Thus, we speakof gains or losses in identity, which are thegains or losses in utility derived from theterm Ii.

This model describes behavior revealed bysocial psychology experiments. These experi-ments demonstrate how assigned social cate-gories influence behavior. The classic RobbersCave experiment is one illustration. In this ex-periment, two groups of randomly selectedeleven-year-old boys were taken to a park in


16 Our concept of “effort” summarizes the student’sinput into education. It is thus a composite of both thelength of time spent in school and the student’s effortwhile there.

17 Other costs include foregone wages and direct ex-penditures on education such as books and tuition.

18 In a more general framework, i could also assignothers to social categories.

Oklahoma, where they were kept apart for aweek. During this week, researchers foundthe boys developed a sense of belonging totheir group. When they met for a tournamentin the second week, the eleven-year-oldequivalent of war broke out, complete withname-calling and stereotyping.19 Minimalgroup experiments show that competition isnot necessary for category assignments to af-fect behavior. Experiments, with both chil-dren and adults, show that subjects are morelikely to give rewards to those assigned thesame label than to others, even when the as-signment is random, recipients are anony-mous, and there is no impact on own pay-offs.20

Outside of experiments, the behavior ofschool-age children also reveals an apprecia-tion of social categories. Children are awareof gender categories, and beyond the ages offive or six, self-segregate into groups of boysand girls.21 Even very young children mayhave some understanding of the social cate-gories and prescriptions for “blacks” and“whites.” Lisa Delpit (1995, p. 48) tells of thenew black first-grader who asked her blackteacher why she was speaking in a differentvoice, a voice blacks use when interactingwith whites: “ ‘Teacher, how come you talkin’like a white person? You talkin’ just like mymomma talk when she get on the phone.’ ”22

2.2 High School: Jocks, Nerds, and Burnouts

We adapt the preceding model to capturethe behavior of high-school students de-scribed in Coleman (1961) and later studies;e.g., Everhart (1983), Eckert (1989), Foley(1990). This model is a prototype, and we useit throughout the paper. The sociologicalstudies all find that high-school students inthe United States divide into social groups;they try to “fit in” to various categories, suchas “jock,” “burnout,” and “nerd.” Coleman’s(1961) Adolescent Society established the re-lationship between such categories andachievement. His work followed AugustHollingshead’s Elmtown’s Youth (1949),which found a strong correlation betweenstudents’ class background and their behav-ior in school. Coleman derived students’ so-cial categories from survey questions such as“Would you say that you are a part of theleading crowd?” and prescriptions fromquestions such as “Among the crowd you goaround with, which of the things below areimportant . . . : be a good dancer, have sharpclothes, have a good reputation, stir up a lit-tle excitement, have money, smoking, beingup on cars, know what’s going on in the worldof popular singers and movie stars.”23,24

These categories and prescriptions forthe “leading crowd” were associated withstudents’ self-images and academic per-formance. Students in the leading crowdappeared to have an enhanced self-image;they were only half as likely “to want to besomeone else” as other students.25 Their


19 For description and discussion of this experiment,see Roger Brown (1986).

20 For description and discussion of minimal groupexperiments and children subjects, see Brown (1986)and Henri Tajfel and John Turner (1979, pp. 13–15).

21 For an excellent survey of the research on groupdistinctions and social categories, see MargaretWetherell (1996, pp. 219–27).

22 Students also understand ethnic social categories,and may behave accordingly. In a recent survey of highschool students in Miami and San Diego, RubénRumbaut (2000) finds that grade point average variedaccording to students’ ethnic self-identification(Hispanic, Vietnamese, etc.) controlling for parents’ socioeconomic status. He also found that the more students identified as being American, the lower wasacademic achievement.

23 See Coleman (1961, Boys’ Fall Questionnaire, p.3). Another example: “Now rank the following fouritems in terms of their importance to you: groups andactivities outside school, activities associated withschool, having a good time, a good reputation.” (p. 2).

24 Coleman’s categories were derived from answersto questionnaires; thus there remains some question asto the extent to which his social categories were im-posed by the researcher; but later ethnographic studiesof high schools show clearly Coleman’s categories correspond to students’ own classifications.

25 Coleman (1961, p. 225).

self-categorization seems to have affectedtheir behavior. In schools where academicperformance was not a criterion for being inthe leading crowd, high-IQ students re-duced their performance: The best studentswere less likely to be those with the highestIQ.26

Many ethnographies have followedAdolescent Society in describing how schoolsocial settings affect behavior. Eckert’s(1989) study of the “jocks” and “burnouts” ina high school on the outskirts of Detroit, forexample, shows that students see themselvesin terms of these social categories and be-have accordingly. She reports jocks andburnouts as behaving differently: in “cloth-ing, territory, substance use, language, de-meanor, academic behavior, and activi-ties.”27 Each of these behaviors signifiesdifference and opposition between the twogroups. Jocks wear pastels, burnouts weardark colors. Burnouts smoke, jocks abstain.Jocks hang out around the lockers and avoidthe courtyard, burnouts hang out in thecourtyard and avoid the lockers.28 Moregenerally, jocks accept the school’s authority;burnouts reject it.

To capture this behavior, we construct amodel where students both choose effort inschool and divide themselves into three so-cial categories: leading crowd, nerds, andburnouts; the addition of nerds correspondsto the groupings found by Jere Cohen(1979) in a factor analysis of student ques-tionnaire responses.

Consider a population of students normal-ized to size one. Each student has two exoge-nously given characteristics. The first is ability,ni, that is, a complement to effort in the pro-duction of skills. The second is physical ap-pearance, or “looks,” li. Looks and ability areboth independently and uniformly distributedon [0,1]. A student’s marketable skills dependson effort ei and ability ni, ki = ki(ni,ei) = ni ei.The pecuniary cost of effort is 1=2 (ei)2.

As for identity payoffs, prescriptions givethe ideal characteristics for each social cate-gory L, leading crowd; N, nerds; and B,burnouts. The ideal L has looks l = 1; theideal N has ability n = 1; and the burnoutsdo not have an ideal.29 Prescriptions alsodictate ideal effort levels, with e(N) > e(L)> e(B). A student’s self-image depends onher own category, ci, and the extent towhich her own attributes and behavior cor-respond to her category’s ideals. A studentfor whom ci = L earns identity payoffs IL t(1 �li), where t is a positive parameter scal-ing the identity loss from i’s distance fromher ideal. A student who sees herself as anN earns IN � t(1 �ni). The ethnographiessuggest that IL >�IN >�IB so that students inthe leading crowd have a more rewardingself-image than nerds or burnouts. For con-venience, we normalize the initial self-im-age of burnouts to zero, IB = 0. A studentwill lose utility 1/2 (ei �e(ci))

2 for deviationsfrom the prescribed effort for her respec-tive category, ci.

The parameter t is critical to this and sub-sequent models. It measures how difficult itis for students with different ascriptive char-acteristics to fit in a group. That is, t is pro-portional to the translation of ascriptivecharacteristics into social differences.

Together we have a utility function for astudent i who is, respectively, L, N, or B, as:


26 A similar finding obtained when comparing boysand girls. Girls, for whom outstanding achievement wasless socially valued than for boys, also had lower varia-tion in grades. This finding suggests that, like a girl weknow who cried after winning a math prize, the girls inColeman’s survey reduced their academic achieve-ment. Nor does the composition of socioeconomic sta-tus for the different schools offer an explanation—sincethe importance of scholarship for membership in theleading crowd does not vary systematically with schoolsocioeconomic status.

27 Eckert (1989, p. 69).28 See Eckert (1989, p. 50, p. 51 ff, p. 58 and p. 53).

29 It is straightforward to model endogenous ideals.E.g., suppose the ideal is the average of members of thegroup. The equilibrium ideal then emerges from stu-dents’ decisions to join different groups. Similar resultsconcerning effort in school obtain.

Ui(L) = p

"w ki

1

2e

2i

#

+ (1 p)

"IL t(1 li)

1

2(ei e(L))2

#

Ui(N) = p

"w ki

1

2e

2i

#(1)

+ (1 p)

"IN t(1 ni)

1

2(ei e(N ))2

#

Ui(B) = p

"w ki

1

2e

2i

#

+ (1 p)

"1

2(ei e(B))2

#

where 0 £ p £ 1 denotes the weight on thepecuniary benefits and costs of effort.

To maximize utility, a student chooses a cat-egory and an effort level. The special casewhere p = 0 describes behavior when stu-dents care only about their current social situ-ation. Ethnographies suggest such a low valueof p. In the West Texas High School describedby Foley (1990, p. 101): “Most students cameto school for extracurricular activities and thesocial scene.” For all but a few “brains,” aca-demics was a side show. In the model, a stu-dent chooses her category to balance socialstatus with “fitting in.” A student with suffi-ciently low l will reject the leading crowd. Sheloses IL but no longer suffers the loss t(1 �li).Students whose n is less than (1 �IN/t) andwhose l is also less than (1 �IL/t) choose to beburnouts—they are unable to obtain positiveutility either as a nerd or as a member of theleading crowd. For p = 0, each student ichooses effort to match the ideal for her chosen category c*

i ; that is, e*i = e(c*

i).This analysis shows the contrast between

the outcomes of a purely economic modeland a purely sociological model. In thepurely economic model, when p = 1, aver-age effort is w/2 and skill aquisition is w/3.

Here, when p = 0, identity is the only de-terminant of the input of student effort,which is, on average30:

(1 IN=t)(1 IL=t)e(B) (2)

+ [(IN=t)(1 IL=t + 1=2(IN =t)]e(N )

+ [IL=t 1=2(IN =t)2]e(L):

Neither effort nor skill acquisition is deter-mined by the wage, but by the social param-eters: IN, IL, e(B), e(N), e(L), and t.31 For IL> IN, there is a group of students with highn but also high l who are in the leadingcrowd despite high levels of academic abil-ity. These high-ability students, just as inColeman’s schools, reduce their perfor-mance below their high-ability peers.

In (2), skill acquisition is not responsive tothe wage, but is responsive to the degree ofsocial difference, t, in a school. When t ishigh, it is harder for students with low l andn to fit in as part of the leading crowd or as anerd. Skill acquisition is low, since more stu-dents become burnouts and exert low effort.On the other hand, when t =�0, all studentsare in the leading crowd, average effort ise(L), and skill acquisition is e(L)/2.

As we will see further below, the precedingmodel allows an assessment of school policiesthat affect social categories and prescriptions.Policies that change any of the social parame-ters will almost always have an effect on edu-cational outcomes as long as p is not equal toone. For example, let us modify the modelslightly to demonstrate the effect of schoolathletic programs, perhaps the most pervasiveAmerican school policy that affects students’social arrangements. Foley’s (1990) WestTexas school, for example, accorded specialimportance to football, and football playerswere among the most popular students in


30 Here we discuss only the interior (nonboundary)solutions.

31 For intermediate values of p, of course, skill acquisition will depend on both the pecuniary returnsto education and the social structure of the school.

school, as Coleman found, statistically, thatbeing on the football team dramatically af-fected a boy’s probability of being in the “lead-ing crowd.”32 Let each student have an ath-letic ability ai that is independently anduniformly distributed on [0,1].33 The ideal foran athlete is a = 1, and athletes are membersof the leading crowd (earning identity payoffsIL �t(1 �ai), and suffering 1/2 (ei �e(L))2 fordeviations from the ideal effort e(L)). With theadditional choice of “athlete,” some studentswith high academic ability, n, but with high a,join the leading crowd. Students with low land n but with high a also choose to be ath-letes. That is, athletics allows an entree intothe leading crowd. Athletics turns some nerdsinto “jocks” and some burnouts into jocks. Theeffect on average skills may be either positiveor negative, but unambiguously, there arefewer students in the two extreme categories.In this way, school athletic programs “democ-ratize” skill acquisition. This outcome corre-sponds to the description of Foley’s school,where working-class Mexican-Americansfound their way into the leading crowd by ex-celling at football. The Mexican-Americanplayers were mostly from “low white collar”(54 percent) or “blue collar” families (20 per-cent), but they were much more likely thantheir peers to pursue a college education andto end up in middle-class jobs.34

2.3 Empirical Evidence and High Schooland Beyond

Here, as elsewhere in the paper, we shalldiscuss some of the empirical implications ofour model that differentiate it from more tra-ditional economic analysis of education. Themodel implies: first, there are differentgroups (or social categories); second, mem-bers of these different social categories have

different prescriptions; third, members of thedifferent social categories will have differentbehaviors, predictable from the prescriptions;and fourth, school policies can change the divisions into social categories and prescrip-tions and thereby affect behavior.

The testing of each implication is chal-lenged by a problem of identification. Howcan we know that being in one social cate-gory or another reflects anything more thanindividual tastes and endowments? For ex-ample, there may be no such thing as a groupof nerds: those who are called nerds may justbe those who are smarter and more academ-ically inclined. The empirical task is to estab-lish that membership in a social category, independent of tastes, affects behavior.

The model suggests in principle how totest for the existence of social categories andtheir influence. In the model a student juston the margin of being in one category oranother, for example, of being a nerd or amember of the leading crowd, will have dis-continuous effort, depending upon whethershe chooses to be a nerd or in the leadingcrowd. With data on “looks” and “ability” asindependent variables, and “effort” as thedependent variable, and with assumptionsabout the distributions of error terms, itwould be possible to estimate the parame-ters of the model. Such an estimated modelis identified by the nonlinearity in the be-havior of effort, which will behave discontin-uously at the boundary between social cate-gories. But the potential presence ofunobservables poses serious problems in in-terpreting such nonlinear estimations inpractice. Consider the simplest possiblemodel, where effort depends only on ability,which has observed and unobserved compo-nents that are not perfectly correlated. Nowsuppose that the researcher tries to test theeffect of group membership. She will use ef-fort as a dependent variable to estimate theslope of effort on observed ability, and shewill also estimate a parameter that gives thelevel of ability that divides the nerds fromothers. She will falsely accept the hypothesis


32 See Coleman (1961, figure 4.10, p. 131).33 Hollingshead (1975) found that athletic participa-

tion was positively correlated with social class. See alsoColeman (1961).

34 See Foley (1990, p. 138 and tables 2 and 4, pp.232–33).

of categorization because the observed ef-fort will be correlated with unobserved abil-ity. It will appear falsely to her that there aretwo groups, whereas in truth there is onlyone group, with effort only dependent onability.

Despite these problems, such groupingsmust sometimes be identifiable. The mostprominent division of social category is uni-versally known and is commonly used byeconometricians. This is the division of thepopulation by gender. A common economet-ric exercise is to check whether men andwomen behave differently. Tests of differencein behavior by gender avoid the problem ofidentification because the econometriciancan take as given the categorization of maleand female. In addition, the observation ofmembership in these two groups is madewith almost no error. We propose that re-searchers use the same sort of informationthat accurately divides people into male andfemale to make other group identifications.The differences between male and female areunambiguous to almost all observers becausemen and women wish to signify their gender.Boys typically do not want to be classified asgirls, nor girls classified as boys. They there-fore signify their gender in a variety of waysthat are obvious to each other and withoutambiguity to any ethnographer. For groupsthat are less universal, econometricians needto use ethnographic information that pro-vides the signifiers, for example, on dress andadornment whereby members of the differ-ent social groups indicate their membership,to themselves and to others. The descriptionof the differences between jocks and nerdsdescribed by Eckert suggests the existence ofsuch markers. Like males and females, theydress differently, they adorn themselves dif-ferently, they have different manners andmannerisms. Since these signs are supposedto be read by others, there should be rela-tively little error, as in gender identification,as to how the groups are divided.

It is useful here to give a taste ofColeman’s empirical strategy since it illus-

trates the combination of being part ethnog-rapher, part statistician. His analysis showsthat different school groups within schoolsand across schools have different prescrip-tions. He first identifies members of theschool elites by asking students who they“want to be like,” or, alternatively, who is in“the leading crowd.”35 He then teases outthe prescriptions by asking about the quali-ties of those in the elites and also how theydiffer from the rest of the student body. Forexample, he shows that members of thefootball team are much more likely to be inthe elite;36 they also have better grades.37

Among girls, the elites are much more likelyto value participation in school activities,and somewhat more likely to want to bepopular.38 He describes in some detail howthe prescriptions of elites differ from thoseof the rest of the student body, how pre-scriptions differ by sex, and how they differschool-to-school by, for example, comparingthe role of car ownership, the role of popularsongs and movies, the role of reputation, andthe factors making for popularity.39 Heshows that there is consistency of opinion inmany different questions, as would be ex-pected if schools have a common under-stood culture. For example, there is a highcorrelation—not among freshman, who havejust arrived, but among seniors—in rankingthe role of high grades as a determinant ofpopularity and also of membership in theleading crowd.40 He shows that in twoschools that place special emphasis on intel-lectual activities, boys and girls both placeespecially high ranking on grades as a deter-minant of popularity.41 He views this finding


35 Coleman (1961, pp. 6, 12). More precisely, theleading crowd was defined by answers to the followingquestion: “If a fellow (girl) came here to your schooland wanted to get in with the leading crowd, what fel-lows (girls) should he (she) get to be friends with?”

36 Coleman (1961, p. 131).37 Coleman (1961, p. 82).38 Coleman (1961, p. 137).39 Coleman (1961, ch. IV, pp. 97–142).40 Coleman (1961, p. 74).41 Coleman (1961, p. 78).

as the result of a small natural experiment(of course with undetermined problems ofselection bias) that those who are placedinto different school environments will endup with different prescriptions.

Although a researcher could use cross sec-tion data in the fashion of Coleman, paneldata is necessary to test the model against thepossibility that the observed correlations aremerely due to the correlation between groupmembership and unobserved tastes. The twocases have very different policy implications.If the model describes reality, the shift of aperson from one group to another will resultin a change in their prescriptions and theirbehavior. If, on the other hand, the correla-tion between prescriptions and group behav-ior are only due to unobserved tastes, suchshifts will have no effect on behavior. Indeed,to test for one of these possibilities relative toanother, one needs panel data where the

shift from one category to another was ran-domly distributed, as in either a controlled ora natural experiment. We defer the discus-sion of such data until later in the paper, afterwe have seen some effects of different edu-cational experiments. We will find that theremay be natural and controlled experimentsthat could be used to test some of the predic-tions of the model.

We ourselves engaged in a crude attempt tosee whether behavior is associated with socialcategory. Adolescent Society was not only fol-lowed by ethnographies, but also by the con-struction of a large data set, High School andBeyond. The questionnaires of AdolescentSociety and High School and Beyond, which isthe source for our tables, are strikingly similar,including questions about self-esteem andmembership in the leading crowd. The prob-lems with these tables serve as a useful back-drop for discussion of the needs of empirical


TABLE 1ATTITUDES OF STUDENTS BY SOCIAL CATEGORY: ODDS RATIO FROM LOGIT REGRESSIONS

Positive Attitude Think I Am School Spirit View School Toward Selfe Depressedf No Good At Allg Excellenth Discipline as Fairi

Nerdsa 1.32** .74** .80** .99 1.33**

N = 19018 (.03) (.04) (.03) (.02) (.05)

Athletesb 1.37** .65** .95 1.33** 1.21**

N =�25528 (.03) (.03) (.03) (.03) (.04)

Leading Crowdc 2.09** .46** .50** 1.33** 1.19**

N =�42230 (.05) (.02) (.02) (.03) (.05)

Burnoutsd .93** 1.75** 1.88** .90** .76**

N =�9279 (.02) (.08) (.07) (.02) (.04)

Independent Variable

DependentVariable: Social Category

Source: High School and Beyond. aNerd: took advanced English or advanced math course; bAthlete: participated invarsity sports if senior, in varsity or other athletic teams, if junior; cMember of Leading Crowd: Regards self as pop-ular; dBurnout: had disciplinary problems in school in last year; ePositive attitude toward self: agrees strongly withstatement “I take a positive attitude of myself”; fDepressed: Respondents (sophomores only) felt depressed or un-happy a lot in past few weeks; gThink I am no good at all: Strong agreement with statement “At times I think I amno good at all”; hRespondent rates school spirit as excellent; iRespondent rates fairness of discipline as excellent;jRespondent answers true to: Do you dread English class? kRespondent answers true to: Do you dread math class?;

work that tries to “test” our model and showsthe need for data at an ethnographic or near-ethnographic level.

Table 1 categorizes students who reportedtaking advanced English or math classes as“nerds;” students on sports teams as “ath-letes;” those who responded “yes” to “areyou a member of the leading crowd in yourschool?” as in the “leading crowd,” and thosewho reported disciplinary problems as“burnouts.” The nerds, athletes, and mem-bers of the leading crowd appear to havebetter opinions of themselves than studentsoutside these respective categories. Theywere more likely to agree strongly with thestatement “I take a positive attitude of my-self;” less likely to be “depressed or unhappya lot” in the past few weeks; and less likely toagree with the statement “I think I am nogood at all.” Burnouts show the oppositepattern. Similarly, as in the model, athletes,

nerds, and members of the leading crowdappear to show more positive attitudes to-ward school than students outside their re-spective category. Table 1 shows that theywere considerably less likely to view schooldiscipline as poor; less likely to have beenbored a lot; more likely to be disappointed ifthey do not graduate from college; and morelikely to like working hard in school.Burnout attitudes had the opposite pattern.These observations are consistent with thepremise of our model that athletes, nerds,and the leading crowd all identify withschool; burnouts reject it. In table 2 we seethat burnouts are less likely to be from thetop quintile of socioeconomic status (SES)and considerably more likely to be from thebottom quintile than non-burnouts, and theopposite pattern describes the SES of ath-letes, nerds, and members of the leadingcrowd.


TABLE 1 (cont.)

Dread English Dread Math Take ACT or Bored a Like Working Disappointed if Do Classj Classk SATl Lotm Hard at Schooln Not Go to Collegeo

.64** .68** 2.24** .84** 1.82** 2.34**

(.02) (.02) (.06) (.03) (.03) (.05)

.90** .86** 2.79** .77** 1.16** 1.58**

(.02) (.02) (.07) (.02) (.02) (.03)

.71** .92** 1.28** .58** 1.41** 1.50**

(.02) (.03) (.03) (.02) (.03) (.03)

1.75** 1.50** .86** 1.47** .49** .54**

(.06) (.05) (.02) (.05) (.01) (.01)

lACT or SAT: seniors only; mBored a lot: Respondent felt bored a lot in last few weeks; nWorking hard at school: re-spondent likes working hard at school; oDisappointed if do not go to college: agreement with statement “I will bedisappointed if I do not go to college.” a, b, c, d are the dependent variables; each of them is regressed on the inde-pendent variables (e, f, g, h, i, j, k, l, m, n, o) one-by-one; standard errors are in parentheses; ** indicates significantat 1% significance level. Example: Odds Ratio (OR) = Odds for depressed(Nerd)/Odds for depressed (non-Nerd) =.74; where Odds for depressed = probability of depressed/probability of not depressed; logit (p) = a + bX; OR = eb.

While tables 1 and 2 appear to substanti-ate our model’s assumptions, other interpre-tations are equally consistent with the data.The leading problem of interpretation is inthe classification of students by group. Con-sider our classification of “nerds.” We haveclassified nerds by their behavior, as thosewho take an honors course. Irrespective ofwhether nerds form a social category thatwill affect their attitudes toward academicwork, we might also expect those who havetastes for taking such courses to have all ofthe attitudes attributed to them in the table,such as “positive attitude toward self,” un-likely to be “depressed,” unlikely “to think Iam no good at all,” etc. The same can also besaid of burnouts and athletes. Our definitionof “leading crowd” yields another problem:this is a self-classification; at best it reflectsthe perception of the interviewee. She may

be overly optimistic about her position orshe may be knowingly stretching the truth.We would expect that both over-optimistsand truth-stretchers give much the same an-swers as those who are true members of theleading crowd. The problems with the inter-pretation of the groups in tables 1 and 2 con-firm the importance of systematic meansoutside the questionnaire to classify groups,as we suggested could be done by the mar-riage of ethnographic observation and statis-tical use of questionnaires.

3. The School as an Institution

To many scholars outside economics,schools are social institutions with socialgoals. Schools not only impart skills; they im-part the characteristics and behavior of idealstudents. Here we discuss this second aspect


TABLE 2SOCIOECONOMIC STATUS OF PARENTS AND SOCIAL CATEGORY: ODDS RATIO FROM LOGIT REGRESSIONS

Nerdsa Athletesb Leading Crowdc Burnoutsd

N =�19018 N =��25528 N =�42230 N =�9279

Top 2.07** 1.71** 1.56** .66**

(.05)* (.04) (.04) (.02)Fourth 1.17** 1.16** 1.29** .85**

(.03) (.02) (.04) (.03)*

Third .93 1.03 1.06* .97(.02) (.02) (.03) (.03)

Second .71** .84** .92** 1.11**

(.02) (.02) (.02) (.03)*

Bottom .57** .57** .56** 1.54**

(.01) (.01) (.01) (.04)

Source: High School and Beyond. aNerd: took advanced English or advanced math course. bAthlete: participatedin varsity sports if senior, in varsity or other athletic teams if junior; cMember of Leading Crowd: Regards self aspopular; dBurnout: had disciplinary problems in school in last year. eDuncan index of parents’ socioeconomic sta-tus from parents’ survey.Standard errors are reported in parentheses; *Significant at 5% level; **significant at 1% level; example: Odds Ratio(OR) = Odds for Nerd (bottom quintile)/Odds for Nerd (not from bottom quintile) = .57, where Odds for Nerd =probability of Nerd/probability of not Nerd; logit(p) = a + bX; OR = eb.

Dependent Variable: Social

Category

Independent Variable: SES by Quintilee

of the sociological approach to educationthat is missing from current economic mod-eling. In this approach, students’ identitiesdelineate whether a student accepts or re-jects the school itself. We have already seenEckert’s description that a primary divisionbetween jocks and burnouts is that jocks ac-cept the school and participate in its activi-ties, while burnouts reject it.

Why would students reject their school? Alarge part of the sociology of schoolingpoints to systematic social differences be-tween students and schools as a reason.Histories of urban education, for example,emphasize the clash between theAmericanizing schools of the early twentiethcentury and their immigrant students.Students were not just taught reading, writ-ing, and arithmetic. In school, they were alsocorrected in details of comportment, includ-ing what to wear and how to speak.42 Theteachers believed they were teaching thestudents what they needed to be economi-cally successful. The skills were socially neu-tral—just logic and reason.43 But that is nothow it seemed to all students. One protago-nist, quoted prominently by Robert Hampel(1986), relates that his loving Englishteacher disapproved of his clothing andmanners and saw him as a “filthy little slum

child.” This episode shows the culture clashemphasized by education historians, thateven the most caring teachers can unknow-ingly offend their students and convey thatthey are inferior.

In our modeling below, schools face atradeoff that captures the historical andethnographic accounts of schooling in theUnited States. These accounts, and ourmodel, see the school as having a choice be-tween promoting a student ideal closer to eco-nomically useful cultural norms and skills andan ideal closer to the students’ social back-grounds.44 It may seem strange to an econo-mist that such a trade-off exists. But sociolo-gists’ accounts reveal that skills, and certainlynorms of dress and comportment, are not so-cially neutral. Even such skills as proficiencyin English and mathematics are associatedwith particular social categories, and a cur-riculum can privilege certain social groups. Inthis case, the promotion of certain skills andprescriptions for behavior can alienate stu-dents from different social backgrounds.

Paul Willis’ (1977) classic day-to-day ac-count of working-class British adolescentsillustrates. To Willis, the school itself, as an


42 According to David Tyack’s (1974, p. 234) history,turn-of-the-century public schools demanded “total assimilation . . . [and] creat[ed] a sense of shame atbeing ‘foreign.’ ” See also Hampel (1986).

43 Bloom (1987, p. 27) characterizes the social cate-gory whence immigrant students were being rescued asthat of the “traditional communities where myth andpassion as well as severe discipline, authority, and theextended family produced an instinctive unqualified,even fanatic patriotism.” He contrasts this with theideal of the schools: “the rational and industrious man,who was honest, respected the laws, and was dedicatedto the family . . . Above all he was to know the rightsdoctrine; the Constitution, which embodied it; andAmerican history, which presented and celebrated thefounding of a nation ‘conceived in liberty and dedi-cated in the proposition that all men are createdequal.’ ” For the purposes of this paper it does not mat-ter whether Bloom has given the correct interpretationof the contrasts; all that matters is that the protagonistsbelieved in these differences.

44 Bowles, Gintis, and Melissa Osborne (2001) dis-cuss the extent to which schooling increases earningsby affecting personality. Curiously, the inclusion ofmeasures of cognitive ability in earnings equations re-duces the coefficient on schooling relatively little—tak-ing an average over different studies, by only 18 per-cent. (See p. 1149). Personality seems to matter more.For example, in an earnings equation with PSID datathe introduction of a battery of motivational and behav-ioral traits reduces the coefficient of years of schoolingon earnings by 37 percent (see p. 1164). These results,and others, are suggestive (although certainly not con-clusive) that behavioral traits developed in school areimportant components of the school-based skills thatare of economic advantage. Exactly consistent withthese results and the argument in this section, Bowles,Gintis, and Osborne hypothesize that much of the eco-nomic benefit of schooling results from “subjecting stu-dents to types of social interactions and systems of re-ward that replicate the social interactions and rewardsystems of the workplace, providing positive reinforce-ment for some behaviors and personalities and sanc-tions for others” (p. 1167).

institution, embodies a set of prescriptions,derived from teachers who value order anddiscipline, as well as the administration thatsponsors programs to “reform” working-class youth. Willis’ account shows students’powerful reactions to such school messages.He follows a group who call themselves “thelads,” who resist schooling and end up, liketheir parents, in working-class jobs. Thelads’ opposition to the school can be foundin almost everything they do. They breakschool rules by drinking, smoking, disrupt-ing class, and, especially, by generating a“laff” (practical joke). Their clothes suggestsexual maturity, just as smoking and drinkingindicate their rejection of school rules meantfor children. Just prior to graduation, thelads got drunk at a pub at lunch time and re-turned to school, where the school authori-ties punished them by refusing to let themgraduate. Teachers were surprised at thelads’ spree because they could have waiteduntil the evening to drink. The teachersmissed the symbolism of the lads’ last “laff.”

In other ethnographies, we see that thelads behave like “burnouts” everywhere.Foley (1990) relates a similar West Texas ex-ample of resistant behavior by Mexican-American male high-school students.45

Foley relates hitching a ride with some“vatos” to an away football game. As theboys light up joints on the way, they plan tochase the local girls and provoke a fight withthe local males, thereby using a “respectableevent to be disrespectful, rebellious andcool.”46 Just as the lads use their school’srules against alcohol and smoking as thecontext for their displays, the vatos use theirschool’s preoccupation with football as thebackdrop for their resistance.

Willis and Foley see it as no coincidencethat the lads and the vatos come from work-ing-class families. To them, the schools are

insulting. The only vato who made it into themiddle class47 reflected, ten years later, onhis high-school experience. In his commentswe see the key features of the models thatfollow—the anger at teachers and schooland the corresponding social divisions, withsome students accepting the school and oth-ers rejecting it:

We were really angry about the way the teacherstreated us. They looked down on us and neverreally tried to help us. A lot of us were real smartkids, but we never figured that the school wasgoing to do anything for us . . . We were the vi-olent macho types, I guess. They’d [the teach-ers] manipulate the nerds into school and books.There was a real separation between us and thenerds and the jocks.48

A sequence of three models will capturethe historical development of U.S. schools inthe twentieth century. The first model cap-tures historians’ depiction of the initial pe-riod, in the early twentieth century: schoolshave a single ideal, which students may ac-cept or reject. The second model conformsto the sociological picture of contemporarypublic education: schools adjust to their di-verse student bodies by allowing students tochoose among different ideals within thesame school. The third model captures thenature of reform programs: schools reducethe social difference between students andtheir schools; students then identify withthese schools, which promote high academicachievement.

In our review, rules and disciplinaryprocesses will appear prominently as evidenceof the ideals and prescriptions of schools.Sociologists view such observations as espe-cially revealing. In his study of the Puritans,Kai Erikson (1966) observed that rules anddisciplinary proceedings delineate the bound-aries of a community. Punishment is an occa-sion to draw the line between acceptable andunacceptable behavior. Accordingly, thosewho willingly disobeyed the rules are not fullmembers of the community. In the school


45 For Mexican Americans, see also Angela Valen-zuela (1999). Signithia Fordham (1996) provides astudy of African-American students.

46 Foley (1990, p. 58).47 Foley (1990, p. 140).48 Foley (1990, p. 139).

context, we have already seen how the ladsand the vatos used disobedience of schoolrules to demonstrate their rejection of theschool and its ideals.

3.1 A Model of a School with a Single Ideal

A model with a single ideal reflects his-torians’ view of late nineteenth- and earlytwentieth-century U.S. public schooling.Associated with the school is a social cate-gory, which we label S. The school’s prescrip-tions give the ideal characteristic and behav-ior of a student in this category. Supposestudents’ characteristics ei, such as class, eth-nicity, or other social attributes, are uni-formly distributed on [m �s/2, m + s/2],where m is the mean and s represents the di-versity of the student population.49 Let s Î[m �s/2, m + s/2] denote the ideal charac-teristics of S, and let e(S) > 0 denote theideal effort level. We call s the school’s ideal.

As in the model of the previous section, astudent’s self-image depends on her cate-gory and the extent to which her own attri-butes and behavior match the ideals for hercategory. A student i who identifies with theschool category, S, earns IS �t(s �ei) for ei £s and earns IS for ei > s. The parameter tagain measures the extent to which studentswith different characteristics do not fit theschool’s category. A student suffers an iden-tity loss 1/2(ei �e(S))2 for diverging from the ideal effort level e(S). A student i whodoes not identify with the school, a“(B)urnout,” earns identity payoffs diIBwhere 0 £ di £ 1 is the level of student i’s

disruption in the school. This disruption expresses an alternative identity, as seen inthe behavior of the lads and the vatos. A Bwill also lose utility of 1=2 (ei �e(B))2 for deviating from the prescribed effort levele(B), which we normalize to zero.

We assume that the school ideal, s, affectsmarketable skills. In particular, the greateris s, the more marketable skills a student ob-tains for a given effort level. This assump-tion reflects associations between certainsocial categories and the curriculum, behav-ioral prescriptions, and other aspects of eco-nomic success. Let the production functionfor skills take the form ki = sei (1 � di),where skills are increasing in s and ei butdecreasing in disruption di. With these as-sumptions, an S and a B have utility, respec-tively,

Ui(S) = p

"s ei

1

2e

2i

#

+ (1 p)

"IS

1

2(ei e(S))2

#for °i > s

= p

"s ei

1

2e

2i

#(3)

+ (1 p)

"IS t(s °i)

1

2(ei e(S))2

#

for °i < s

Ui(B) = p

"s ei(1 di)

1

2e

2i

#

+ (1 p)

"di IB

1

2(ei e(B))2

#:

As previously, the special case p = 0 is thepure sociological model. In this case, any stu-dent who does not identify with the schoolwill choose di = 1. The number of burnoutsis determined by the position of the school, s,relative to the distribution of the students,and the relative gain of identifying with


49 Students’ characteristics e, in general, can includea variety of traits. Some will be correlated with a family’ssocioeconomic status, race, and ethnicity. Others will bepurely idiosyncratic. Most economic models consideridiosyncratic traits as tastes, randomly distributed in thepopulation. In our setting, these traits reflect a student’ssocial background. We consider how these socially de-fined traits affect effort in school. Idiosyncratic compo-nents would explain the differences in outcomes withinsocial groups. For instance, there are “burnouts” even atall-white middle- and upper-class schools. A poignantfictional example, J.D. Salinger’s Holden Caulfield,speaks in the voice of a burnout.

school I º�IS �IB. Let b denote the fractionof burnouts. We have

(4)≠ =

1

2+

s m I=t

º

and mean skill acquisition, denoted as K, of

(5)K = s e(S)

Ã1

2

s m I=t

º

!

For s sufficiently above the median student,K declines in s even though higher s con-tributes directly to skill acquisition. As s in-creases, more students reject the school, re-ducing the total skill level. The parameter tis critical. The greater is t, the greater the so-cial differences between the students andthe school. For a given s, as t increases, morestudents reject the school, and skills de-cline.50

3.2 Hamilton High to Shopping Mall High

The modeling above captures events at“Hamilton High,” events that sociologistsbelieve reflect general trends in U.S. educa-tion. The World We Created at HamiltonHigh by Gerald Grant (1988) gives a de-tailed account of an upstate New York highschool from the 1950s to the 1980s. This his-tory relates a great deal of discord—even theclosure of the school for some period oftime. It also relates how the school cameback together as a place of learning. To aneconomist Hamilton High seems to be only acurious story: it does not fit into the typicalmodel. There is massive change in educa-tional outcomes, but no change in resources.Shifts in parameters of the model here, how-ever, easily capture Grant’s history of theschool.

In the 1950s and early 1960s, HamiltonHigh’s student population was homogeneous(s was small) and its ideal, s, was in the cen-ter of the population of white middle-class

students. The school’s prescriptions wereclear. The principal had few doubts that hisjob was to “enforce middle-class standardsof courtesy and respect, emphasize a collegepreparatory curriculum and put winningteams on the Hamilton field.”51 Students ac-tively participated in the school activities, inthe newspaper, Greek letter fraternities andsororities, the girls’ club, and the a capellachorus, and such.

Grant then relates what happened after asignificant number of poor black studentsentered the school in the late 1960s underorders of forced integration. In terms of ourmodel, the diversity, s, of the students rose;most of the new students’ characteristicswere considerably below the school’s ideal s.The model predicts exactly what happened:the number of burnouts and the disruptionin the school increased. The new studentsreacted almost immediately to rejection bythe white students and the faculty.52 Day-to-day there were clashes, arising from whatblack students and their parents saw asracism and unfair application of school rules.The following interaction was a typical class-room exchange:

Teacher (to black student): Please sit down andstop talking.Student: I was only seeing if I could borrow apencil and a piece of paper for that quiz youwere talking about.Teacher: You know you’re supposed to be inyour seat.Student: But you will give me a zero if I don’thave a quiz paper.Teacher (slightly exasperated): Sit down. You’resupposed to bring those things to class or bor-row them before class.Student (voice rising): Why you picking on me?You don’t pick on white kids who borrow a pieceof paper.53

Anger rose to the point that riots closed theschool. The riots began after a fraternityparty behind the school where white stu-dents told a group of blacks to “get out of


50 Here, as before, the text describes the nature ofinterior solutions.

51 Grant (1988, p. 241).52 For changes in s holding constant the upper

bound m + s/2 , the number of burnouts, b, is increas-ing in s.

53 Grant (1988, p. 36).

our school.” The next Monday, some of thenew students tore up the school cafeteria. Achemistry teacher summarized students’feelings:

The black students were responding to the waythey were being treated. You know, it’s like thesewhite teachers don’t really care anything aboutme . . . He isn’t teaching me anything. Youknow, it’s a handout sheet every day or it’s a filmeveryday. The teacher may be making racist re-marks overt or subtle . . . 54 .

The principal of the school looking back atthe period said: “[the school] gave them themessage that ‘some people did notbelong.’ ”55

As the story of Hamilton High continues,we see the school adjusting to its diverse stu-dent body. Gradually, the school of Greek-letter clubs and a capella choruses fadedaway. A new school emerged, whose definingfeatures were tolerance and students’ rightsto choice.56 A slightly revised model belowcaptures this new Hamilton High, which issimilar to what has been elsewhere describedas the typical contemporary U.S. high school,Shopping Mall High (Powell et al. 1985).

3.3 A Model of Shopping Mall High

In Hamilton High, as in other schoolsthroughout the country, students in the1970s were granted greater rights. Gonewere the days of unquestioned in loco paren-tis and a consensus on the school’s ideals.57

At Hamilton High, new rules essentiallyeliminated teachers’ authority to enforce aca-demic and other behavioral standards.

Arbitration guidelines, adopted in 1972, al-lowed students to initiate grievance proce-dures “when the behavior of any staff mem-ber willfully imposes upon a student theethical, social or political values of the staffmember.”58 Teachers only rarely correctedstudent behavior, inside or outside the class-room. A teacher who asked a repeatedlytardy student to bring a note from homewhen late was told by the parent: “stop wor-rying my child just because you have a mid-dle-class hang-up about time.”59 Teacherswho tried to punish students for cheating of-ten had to defend their allegations to theprincipal, to the student’s parents, and, whencalled upon, to the student’s lawyers.

Students also gained rights to choose theircurriculum. The school instituted moreelective courses; only ten of eighteen creditsrequired for graduation were specified.Even though formal tracking was elimi-nated, students now tracked themselves.Those interested in academics sought outthe best teachers, and those not interestedchose less challenging courses. As the schoolturned to laissez faire, the troubles dieddown. Learning took root again, but the pre-vious standards no longer applied.

Grant’s description of the new HamiltonHigh conforms to educators’ description oftypical U.S. high schools across the country(Hampel 1986). Powell et al. (1985) callthese high schools “shopping malls,” wherestudents are treated as customers. Theschools give them what they want, makinglittle or no attempt to change their values.The Shopping Mall High School appears notto be just the result of changes in the 1960sas reported by Grant and Hampel, butrather to be the much longer evolutionaryoutcome of continued democratization ofU.S. schools. This democratization was evenmore dramatic earlier in the century (seeKatz and Goldin 1997), especially as “life


54 Grant (1988, p. 35).55 Grant (1988, p. 38).56 The problems at Hamilton High are an extreme

version of the day-to-day struggle to maintain schoolorder. Everhart, along with Willis (1977), Foley (1990),Lois Weiss (1990), and others, paint a remarkably simi-lar picture of the many small and large ways studentsassert themselves against teachers’ authority, disruptingclass and school operations. When students do notidentify with the school and accept its authority, learn-ing does not occur.

57 We discuss in detail below the legal innovationsthat advanced students’ rights.

58 These guidelines were distributed to all studentsin a handbook outlining their rights (Grant 1988, p. 53).

59 Grant (1988, p. 54).

skills” curricula were introduced for thenon-college-bound who were said to be“preparing for life.”60

We now capture the essence of aShopping Mall High School in a modelwhere a school may adjust to its studentpopulation. The school has two options. Itcan select a single ideal. Alternatively, theschool can give the students a choice be-tween two ideals. The school makes this se-lection to maximize the total skills, K.61 Theanalysis shows each of the school’s options inturn, then determines which of the two op-tions is optimal.

Single-s optimum. Recall that K takes intoaccount students’ choices whether or not toidentify with the school. As we have alreadyseen, K = s . e(S) . [1 �b], where b� is thefraction of burnouts. The choice of s in-volves a trade-off: Increasing s increasesskills directly, but reduces [1 �b], the num-ber of students who identify with the school.The optimal s balances these effects. From(4) and (5) above, the optimal s that maxi-mizes K is:

(6)s

¤ = min

"m +

º

2;

1

2

Ãm +

º

2+

I

t

!#

Notice again the impact of the social param-eter t; the optimal ideal s* is decreasing in t.The more students view themselves as dif-ferent from the school, the more the schoolmust reduce its ideal to engage students inthe school and increase skills.62

Double-s Optimum. Consider now aschool that can give students a choice.Instead of one ideal, the school can set twoideals sH and sM, where sH > sM and sH is as-sociated with higher academic achievement.There are now three social categories:(H)onors students, with ideals sH and e(H);(M)iddle students with ideals sM and e(M),and (B)urnouts, with no ideal and ideal effortlevel e(B), where e(H) > e(M) > e(B) º�0.

Middle students are a prominent categoryin Theodore Sizer’s (1984) description ofUnited States’ public schools as well asPowell et al.’s Shopping Mall High School.These students are largely uninspired—tolearn or to misbehave. They make their wayto the end of their schooling, without beingprompted to investigate, learn, or question.In Horace’s Hope Sizer relates shadowingone such student, Martha, for the course ofher school day. Passing from class to class, inwhich she is taught “pap [sic],” Marthamaintains a mask of passivity.

As before, a student’s identity depends onthe match between own characteristics andbehavior and the ideal of her category. A stu-dent i choosing category ci = H,M earnsidentity payoffs IS �t(sc �ei) 1/2 (ei �e(ci))

2

for ei £�sc and earns IS �1/2 (ei �e(ci))2 for ei

>�sc. Burnouts earn the identity payoffs of theprevious model. We again consider students’choices of effort and category for p =�0.63

There will be (1/s)(m + s/2 �sH) H stu-dents; (1/s)(sH �sM + I/t) M students; and(1/s)(sM �I/t �m + s/2) B’s. Mean skills as afunction of sH and sM are:


60 This debate is well-documented in Edward Krug(1964) and (1972). Diane Ravitch (1983, p. 46) excori-ates progressive educators, who championed nontradi-tional high school curricula, emphasizing life skills relat-ing to health, vocation, and family and community life.

61 Certainly, a school administration could haveother goals in choosing its category and prescriptions,as would be the case in religious schools. We discussthis possibility below.

62 This level of s is invariant to changes in s and mwhen m + s/2 is held constant. The optimal school af-ter integration at Hamilton High would be the same asthe optimal school before integration. However, if weincluded the many externalities of disruption, theschool would lower s to account for the new students.

63 Once again, the text concentrates on nonboundarysolutions. ei is again assumed uniformly distributed onthe range [m s/2, m + s/2]. An appendix availableupon request gives the full specification and analysis ofthe model. In the analysis we assume that all studentslocated above sH choose H. For sH=sM, we assume thatstudents located above sH=sM choose H (and exerte(H)) and students located below sH=sM that choose toidentify with a school choose M and exert e(M). Thisassumption prevents a discontinuity in the numbers ofH and M students at sH = sM. In what is written below,we assume sM > m s/2 + I/t. This assumption iswithout loss of generality because it is never optimal forthe school to set sM at a lower level.

K = sH

"1

2+

m sH

º

#e(H) (7)

+ SM

"sH sM

º

+I

º t

#e(M)

When setting sH and sM, the school faces twotrade-offs. First, as with one category, raisingideals sH and sM directly increases skill at-tainment, but increases the number ofburnouts. Second, there is competition be-tween the categories. Increasing sH increasesthe skills of the honor students, but more stu-dents will opt for the middle category, withlower educational attainment. An appendixprovides the optimal s*

Hand s *M as functions of

I, m, s, and t: s*H and s *

M are increasing in mand s, and decreasing in t. The distance (s*

Hs *M) increases with t and s.

Selection Between Single-s or Double-sSchool. When is it optimal for a school to givestudents a choice rather than to promote asingle social ideal? The results match the his-torical accounts. Comparison of the totalskills for one category, s*, and of two, s*

H ands *M, shows that when the social distinctions are

large (t is large or s is large), the schoolachieves higher skills by providing two cate-gories. Two categories increases the numberof students who identify with the school.When the social distinctions are small, (t issmall or s is small), students are more likelyto identify with the school, and a school willmaximize skills by eliminating choice andproviding a single standard. The events atHamilton High are thus consistent with themodel. In the short run, the single initialvalue of s is fixed and there was massive dis-ruption. This disruption eased in the long runas the school moved from a single-s ideal to adouble-s ideal with choice and tolerance.

4. Resource Use: Investments in Identityand School Reform

In this section, we introduce resources intoour modeling of schools. This exercise inte-grates the sociological view of schools with a

classic economic view, and, we believe, cap-tures the essence of contemporary school re-form programs. In the previous generation,educators promoted large schools to take ad-vantage of economies of scale; schools shouldbe consolidated so as to offer a variety of ad-vanced courses (James Conant 1959).Current school reform promotes smallschools, with unified programs. We argue thatschool reform on both the left and the rightcan be characterized in this same way: re-formed schools use resources to create acommunity, represented in the model as a re-duction in the parameter t.64 Students withlow values of t do not differentiate themselvesfrom each other according to their homebackground. This leaves schools free tochoose an s leading to high skill acquisition.

4.1 Central Park East Elementary andSecondary Schools

Central Park East Elementary and Sec-ondary Schools is a poster child for school re-form and illustrates the value of creating acommunity and inducing students to identifywith the school. There is no doubt aboutCPESS’ success. In East Harlem, a neighbor-hood where students are more likely to go tojail than to college, this school has almost nodropouts; it sends 90 percent of its graduateson to college of whom 90 percent graduate.65

Accounts of CPESS suggest that its successlies in students’ and teachers’ identificationwith the school and its academic ideals. Thisidentification is no accident. From the very be-ginning, school administrators set out to createa new type of school, with a strong sense ofcommunity. The school takes students out oftheir troubled community and isolates them ina different world. The importance of creating anew, different social category is apparent toDeborah Meier, the founder-director. In her


64 We could also model investments that change thedistribution of students’ characteristics. Similar resultsobtain. Reforms can also involve prescriptions for ef-fort in school. We discuss disciplinary reforms below.

65 Indeed the school has been dubbed the “miraclein East Harlem” (Seymour Fliegel 1993).

words: “We committed ourselves openly andloudly to being different.”66 She further em-phasizes: “We wanted no ‘we’ versus ‘they’ inthis community.”67

The unusual pedagogy and arrangementsat CPESS underlies the group identity of theschool. The curriculum is generated by theideas of students themselves, and examiningthese ideas systematically using “The FiveHabits of Mind.” These five habits requireasking and answering: “How do we knowwhat we know?” “Who’s speaking?” “Whatcauses what?” “How might things have beendifferent?” and “Who cares?“68 The school’sother practices help students identify witheach other and the school. The school hassmall class size, small overall school size,multiple-year student-teacher assignments,and, in higher grades, relatively few period-to-period shifts between teachers—all mea-sures to encourage familiarity among stu-dents and teachers. The emphasis onstudent ideas and student presentations andprojects as well as lengthy open-endedteacher-parent conferencing, are all de-signed to make each student feel as if shebelongs, as if to a family.

4.2 A Resource-Use Model of School Reform

We interpret the creation of a school com-munity as a choice about the allocation of re-sources within a school. The model showswhen resources devoted to such a reformprogram will enhance educational attain-ment. Suppose a school has resources r = 1which it can divide into two uses, those di-rectly devoted to the teaching of skills, inproportion a, with the remainder, 1 – a, de-voted to creating community and reducingt, the social differences between the stu-dents and the school.69 Skills increase

directly in a: ki = asei (1 �di), at the ex-pense of social difference: we have t(a),which is also increasing in a.

The school’s objective function (for p = 0)is now in the two-s and one-s case respec-tively:

K = ¬sH

"1

2+

m sH

º

#e(H)

+ ¬sM

"sH sM

º

+I

º t(¬)

#e(M)

2 s Case

K = ¬ s e(S) (8)Ã

s m I=t(¬)

¬

!

1 s Case

In the two-s case, the school chooses sH, sMand a; in the one-s case it chooses s and a.Solving first for the optimal ideals—s* in the1 – s case and s*

H , s *M in the two-s case—we

see that optimal ideals are increasing in (1 a), the fraction of resources devoted to re-ducing social differences. As students iden-tify more with the school, the school canpromote an ideal that is more amenable tomarketable skills. With an internal solution,a satisfies the following condition, whichshow the costs and benefits of such an in-vestment:

dK(¬)

d¬

=K

¬

¬ t

0(¬)I

º t(¬)

"sM e(M)

#

= 0 2 s Case

dK(¬)

d¬

=K

¬

¬ t

0(¬)I

º t(¬)

"s

¤e(S)

#(9)

= 0 1 s Case


66 Meier (1995, p. 30).67 Meier (1995, p. 20).68 Meier (1995, p. 50).69 It is possible that resources devoted to reducing t

are also directly complementary to skill acquisition—asmany would argue that CPESS’ special curriculum enhances learning.

º

¤

1

2

where s* and s *M are the optimal choices for a

given a. In both cases, investments in reduc-ing t result in a direct loss in skills due to di-version of resources, reflected in the firstterm K/a. The benefit is an increase in thenumber of students who identify with theschool. The second terms show the changein skills of these students.

There is an interior solution for a whent(1) is sufficiently high.70 When students ini-tially think of themselves as quite differentfrom the school, a small investment to re-duce t can have a large impact on skills. Inthis case, it is optimal to spend resources tochange this social parameter. Furthermore,with sufficient investment in t, it is optimalfor the school to promote a single ideal.

The model suggests that there is muchgreater similarity between the “liberal” and“conservative” reform agendas than is appar-ent in the often shrill debate. Both empha-size special curricula that create school com-munities. That is, both invest to reduce t inorder to raise the ideal s. Both are then starkcontrasts to the Shopping Mall High whichmakes no attempt to change social parame-ters. We have already seen the distinct peda-gogy and student ideal in CPESS. We seesimilar elements in Core KnowledgeCharter Schools, a prime example of conser-vative reform. In these schools, studentsshould learn what Hirsch (1996) identifies as“Core Knowledge.” The central idea of theschool—its s in the language of the model—is the role of this curriculum in “promot[ing]a community of learners—for adults andchildren.”71 The aim of the curriculum andother aspects of the reform is to trump theeffect of students’ backgrounds on schoolparticipation. The rules of a school in Parker,Colorado illustrate. Students must wear uni-forms and obey a dress code so strict it

details socks, which “must be worn in a coor-dinated color with the school uniform andworn in a matching pair of the same color.”Furthermore, students and parents mustsign an agreement to abide by the school’scode of discipline.72 These rules and codesof discipline, as we shall discuss next, delin-eate the boundaries of a school communityand play important roles in reform pro-grams.

4.3 Discipline in Schools

Establishing and enforcing a school’s codeof discipline involves a substantial use of re-sources and defines the school community.In the noneconomic accounts of schools andof educational reform, rules and disciplineare a common theme. This emphasis echoesKai Erikson’s (1966) observations that byobeying the rules one becomes a part of thecommunity, by breaking them, one becomesan outcast.

The reforms by James Comer of two NewHaven schools exemplify the use of re-sources to establish and enforce the discipli-nary code.73 Comer describes the situationwhen he arrived at Baldwin ElementarySchool as “shocking.” Desperate teacherswere unable to establish order. Childrenmilled around, yelling and screaming, call-ing the teacher and each other names.74

Five years later order reigned, as depictedon the cover of his book, School Power: aclassroom with all students neatly dressed,smiling at their desks, eagerly raising theirhands.75


70 At a = 0, the condition is always positive, hencewe must have a > 0. At a = 1, the condition is nega-tive for t¢(a) sufficiently large (recall that t¢(a) is posi-tive).

71 See: http://www.coreknowledge. org/CKproto2/about/index.htm#BEN.

72 See: http://www.ckcs.net/.73 CPESS is another example of how a school’s disci-

plinary procedures delineates the community. Studentssent to the Director’s office for misconduct are ledthrough the Five Habits of Mind to sort out the prob-lem (Meier (1988, p. 50)). Such a method would be in-effectual if the students did not already identify withthe school and its particular precepts.

74 Comer (1980, p. 76).75 The children also achieved higher levels of aca-

demic achievement. Prior to the reforms, the schoolhad performance ratings in math and reading at thebottom of New Haven’s public schools.

How did Comer achieve this transforma-tion? Comer says that the success of his pro-gram comes from students’ internalizationof school values. This process involved useof resources. Teachers were trained how torespond to children who acted up. A centralfacility was established where children couldtake hold of their feelings, relax and play, be-fore being returned to the classroom, readyto obey. Comer demonstrates the nature ofthe disciplinary process with the case of anangry fifth grader who had attacked asmaller child with his belt. Rather than sim-ply punishing the student who misbehaved,the teacher, who had been trained to lookfor causes for misbehavior, wormed out ofthe boy that he was upset because his fatherhad been denied a pass from jail forChristmas. She helped him write a letter tohis father, but, at the same time, she alsomade him understand he could not take outhis feelings on other children.76 That is, chil-dren were taught not only how to read, butalso how to obey the rules.

Just as sociologists view schools’ rules anddisciplinary procedures as a window onschools’ ideals, they also view lack of rulesand lack of discipline as indicative of the fail-ure of a school to promote academic andother values. Grant’s description of thenewly tolerant Hamilton High of the 1970sillustrates. The old consensus on appropriatebehavior and rules had broken down withnothing to replace it. The principal ex-plained:

We assumed through the fifties and the sixtiesthat we had a set of rules that applied to every-one when in fact they didn’t. The black commu-nity was basically compliant with the expecta-tion of the white community. Then things beganto change. Then you get into a situation [where]you say what are the rules? When I startedteaching at Hamilton the kids were sent home ifthey wore dungarees. I laugh [to think we everhad such a rule]. But [in 1970] there were norules.77

A teacher who caught a boy cheating wastold that she needed a witness.78 When thesame conscientious teacher gave a studentan F for a ghosted paper, she was forced tohand over all her documents to the student’slawyer, who was seeking evidence of preju-dicial treatment.79 A student group, whichcalled itself the Protesters, opposed all in-fringements on students’ rights, includingthe school prerogative of searching studentlockers.80,81

4.4 Public versus Private Schools

The modeling above also describes a dif-ference between private and public schools.Economists and noneconomists have bothurged public support for private schooling.While economists focus on agency issues,John Chubb and Terry Moe (1988), for ex-ample, advance a different argument for pri-vatization. They see a benefit in the freedomprivate school administrators have to setgoals and mobilize resources to achieve


76 Comer (1980, p. 118).77 Grant (1988, p. 40)

78 Grant (1988, p. 53).79 Grant (1988, p. 53).80 Grant (1988, pp. 60–61).81 Judicial interpretations of the law governing

schools’ authority over students reveals changes in atti-tudes regarding school authority that coincided in timewith the changes in the disciplinary system at HamiltonHigh. In the opinions of the landmark cases, as in ourmodel, public schools have only limited authority to af-fect students’ identity (see Anne Dupre 1996). TheTinker v. Des Moines (1969) and Goss v. Lopez (1975)decisions demarcate the break from the traditionalview of schools as agents of parents (under the doctrineof in loco parentis) with the authority to enforce codesof discipline and mold students’ morals and values.These decisions established that students are “persons”under the Constitution and do not shed their rights “atthe schoolyard gate.” Dupre (1996, p. 54) Tinker v. DesMoines concerns three students who were suspendedafter refusing to remove armbands in protest of theVietnam War. The Supreme Court ruled that theschool could not punish the students, since “Studentsin school as well as out of school are ‘persons’ underour Constitution. They possess fundamental rightswhich the state must respect . . . In the absence of aspecific showing of constitutionally valid reasons to reg-ulate their speech, students are entitled to freedom ofexpression of their views.” Goss v. Lopez gave an analo-gous interpretation of students’ rights to due process.See Dupre (1996, p. 57 ff).

them. In terms of the model, setting goalscan be seen as establishing an ideal s; themobilization of resources can be seen as theuse of resources to establish a communitythat will accept these goals. Private school-ing is thus seen as relaxing constraints on investments to establish identity.

Chubb and Moe (1988, tables 4 to 8) sup-port their view with data from the Teachersand Administrators Survey supplement toHigh School and Beyond. Batteries of re-sponses show that private school principalshave less interference from school boardsand other administrators over curriculum,instruction, and discipline. They face lessconstraints on school policy and proceduresconcerning personnel, and hiring and firing.They have more support from parents in set-ting their goals and set greater goal clarityfor their schools. In addition, Chubb andMoe show that the private school principalsare viewed more affirmatively by theirteaching staff, which is just one sign of aschool community.

Peshkin’s (1985) account of a Christianschool in Illinois provides an extreme exam-ple of a private school that reaches into allaspects of a student’s life. Parents and chil-dren sign detailed contracts, where parentsnot only grant the school the right to enforcea firm Christian code of discipline, theypromise to enforce the school’s code athome.82 The model characterized reductionin t as narrowing the gap between schooland home: this unusual measure brings thehome closer to the school ideal.

4.5 Empirical Evidence

What evidence do we have that the estab-lishment of community affects educationalattainment? Are Comer’s schools and CPESSjust anecdotes? We argue the contrary. Theseprograms changed the outcomes for popula-tions of students. In epidemiology, a popula-tion that shows particular resistance orpredilection for a disease is a research

opportunity. Researchers ask what makesthis population different than the rest of us.Similarly, such reform programs can providevaluable information concerning the inputsto education.

Bryk and his colleagues (1993) conductsuch a study. Catholic schools are known fortheir higher educational outcomes.83 Bryk etal. (1993) describe in detail how Catholicschools establish such a school communitywhere students accept the school’s goals andideals. These schools make their ideals clearin their statements of philosophy. For exam-ple, one school defined its ideal student, who“should be marked by a number of charac-teristics: . . . intellectually competent, . . .loving, . . . a person of faith, . . . [and]committed to doing justice.”84 Each of thedesirable characteristics is described in de-tail. Bryk et al. describe the arrangementsthat foster students’ acceptance of theseideals (in our model investments in t). All theteachers assume responsibility for shapingstudent character and are supposed to be in-volved in many aspects of students’ lives.85

Lack of specialization makes this feasible, asthe English teacher of the morning is likelyto be the counselor at lunch time and possi-bly the soccer coach of the afternoon.86


82 See Peshkin (1985, p. 90).

83 Coleman, Hoffer, and Kilgore (1982) show thatCatholic school students have higher test scores withlower standard deviations than public school students.See table 6–1, p. 127. The question remains to what ex-tent these differences can be attributed to selectionbias or school quality. Joseph Altonji, Todd Elder andChristopher Taber (2000), who also review the exten-sive literature on this subject, correct for selection biasby using the bias in the observables to correct for thebias in unobservables. They concentrate their analysison a sample of students who were in Catholic schools ineighth grade, some of whom continued to Catholichigh schools, and some of whom switched to publicschools. They find that those who continued in theCatholic schools had considerably greater chance ofgraduating and attending college.

84 See Bryk et al. (1993, p. 146).85 Bryk et al. (1993, p. 141): “Teachers convey an in-

trusive interest in students’ lives that extends beyondthe classroom door into virtually every facet of schoollife. In some cases it extends even to students’ homesand families.”

86 Bryk et al. (1993, p. 141).

Wide participation in school activities, in-cluding a greater fraction of students on ath-letic teams and shared religious activities,foster the school community. This commu-nity blurs distinctions between school andhome: in the terms of our model, it reducesthe social differences t.

These differences in community betweenthe public schools and the Catholic schoolsare supported by statistical evidence. FromHigh School and Beyond and the associatedAdministrator and Teacher Survey, Bryk etal. (p. 284) construct an aggregate index ofschool community from 23 sub-components,including such items as the likelihood that ateacher will know a given student, an indexof students’ perceptions of teacher interest,and an index of agreement on goals betweenteachers and administrators. The value ofthe aggregate index of community forCatholic schools is 2.35 standard deviationsfrom the public school mean, as, in fact, theCatholic schools were more communal thanthe public schools for each and every one ofthe subcomponents.87

5. Black-White Differences in SchoolAttitudes and Achievement

Our emphasis on social categories andidentity captures arguments by leadingscholars of black education such as JamesBanks, Comer, Delpit, Ferguson, Fordham,

John Ogbu, and Claude Steele.88 Clashes be-tween black students and their school, as inHamilton High, appear frequently in studiesof black-white differences in educational at-tainment. Anthropological studies show howschool routines and curricula can convey toblack students that there is something“wrong” with them and their background.Delpit’s classroom accounts show, for exam-ple, how teachers’ presumption of superior-ity of standard English can (unknowingly) in-sult speakers of black dialect. In a readinglesson, a girl renders the text: “Yesterday Iwashed my brother’s clothes,” as “Yesterday Iwash my bruvver close.” The teacher cor-rects her. But the student has done some-thing far more sophisticated than read: shehas translated the passage into her own di-alect. Instead of being praised, the girl is toldthat she has made a mistake.89 In microcosm,this incident captures some of the tensionbetween students and schools with sociallydifferent ideals. This tension, Ogbu (1997)argues, leads to an “oppositional culture” onthe part of black students.90

5.1 Empirical Interpretation and Need for Future Work

The empirical studies of black/white differ-ences in education reveal a paradox, whichmight be resolved by the modeling in this pa-per. Tests of the theory of oppositional culture


88 See for example: Banks and Dresden Grambs(1971), Comer (1980), Delpit (1995), Ferguson(1998a,b,c, 2001), Fordham (1996), Ogbu (1971), andSteele and Aronson (1998).

89 Delpit (1995, pp. 58–59). Delpit believes teachingstandard English is especially important in poor Blackneighborhoods. She suggests, for example, that Englishbe taught as drama. Delpit (1995, p. 51) also relates thefollowing interchange between a teacher and a four-year-old Black boy: Teacher: Good morning, Tony, how are you? Tony: I be’s fine. Teacher: Tony, I said, How are you? Tony: (with raised voice) I be’s fine. Teacher: No, Tony, I said how are you? Tony: (angrily) I done told you I be’s fine and I ain’ttelling you no more.

90 See also Fordham (1996).

87 The subcomponent indices measure: teacheragreement on student goals, reported teacher consen-sus on beliefs and values, teacher beliefs that studentscan learn, teacher and administration agreement thatstudents can learn, teacher and administration agree-ment on standards of discipline, student consensusabout teacher role, track and course-taking commonal-ity of students, teacher knowledge of students (inclass), percent students involved in extracurricular ac-tivities, percent students in leadership roles, percentteachers who obtain help from colleagues, teacher co-operation with colleagues, teacher time planning withother teachers, staff commitment to evaluation, partici-pation in faculty social events, teacher time in extendedroles, percent of teachers involved in extracurricularactivities, teacher knowledge of students (beyondclass), teacher contact with students outside of class,and student perception of student interest.

suppose that if black students are angered byschool culture, they should have less favorableattitudes toward school than white students.The existing evidence suggests the contrary.Philip Cook and Jens Ludwig (1997) showthat, compared to non-Hispanic whites, blackshave about the same expectations for educa-tional attainment (high school, college, etc.)and about the same school attendance andlevels of effort.91 Table 3, with slightly differ-ent tabulations from High School and Beyond,shows blacks have better school attitudes thanwhites. On average compared to whites,blacks are less likely to “dread English class,”less likely to “dread math class,” more likely toperceive “school spirit as excellent,” morelikely to report a “positive attitude toward self”and much more likely to “like working hard inschool.” These findings seem paradoxical inview of the large and persistent gap betweenblack and white test scores (ChristopherJencks and Meredith Phillips (1998)).

The possibility that schools adjust theircurriculum and ideal to their student bodyyields a possible explanation why blacks ap-pear to have better attitudes toward schoolthan whites, but they have lower test scores.In the model of section 3 schools adjusted sso that more students will identify with theschool. In this case, it is likely that schoolswith a majority white (black) student popu-lation emphasize an ideal that conforms totheir white (black) students. As a conse-quence, it is likely that black students, forexample, are much more likely to “dread”English class and math class in white schoolsthan in black schools, with the reverse forwhite students. Table 4 shows that when wedivide the High School and Beyond sampleaccordingly, blacks in almost totally whiteschools are, indeed, significantly more likelyto dread English and math than their coun-terparts in all-black schools. Symmetrically,white students in almost totally black schoolsare more likely to dread English; similarly,they also have a positive (although statisti-

cally insignificant) coefficient on the likeli-hood of dreading math relative to their peersat all-white schools.92

Using data collected by Bishop at ShakerHeights High School, Ferguson (2001) hasalso examined the relation between theblack/white grade gap and racial differencesin attitude toward school. Like Cook andLudwig, and like ourselves, Ferguson findslittle difference between black and white at-titudinal scores. Thus, while within race dif-ferences in attitudes are well-correlated withGPA at Shaker Heights, differences betweenraces explain little of the gap.93 This findingsuggests that subjects’ interpretation of theattitudinal questions differs by race.Ferguson’s ethnographic findings of differ-ences between black and white high-schoolstudents—some from schools other thanShaker Heights—indeed suggest that blacksand whites segregate themselves into differ-ent groups. He emphasizes that black stu-dents do not want to appear disloyal to theirrace.94 The modeling in this paper suggeststhat with such differences there should be apresumption that prescriptions, and there-fore ideal levels of effort, will differ by group.We thus might expect that blacks andwhites95 will give similar responses to ques-tions that subjects either consciously or un-consciously interpret as concerning the de-sirability of effort relative to the ideal of theirrespective group. With such interpretation,similarity of response does not imply similar-ity of ideal effort by group. Consider, for ex-ample, the meaning of student scalings of thefollowing three statements: “It’s not cool tobe competitive about grades.” “It’s not coolto frequently volunteer answers or com-ments in class.” “It’s annoying when students


91 See also Roslyn Mickelson (1990).

92 There could, of course, be other theories, such asselection bias that could explain these facts.

93 See Ferguson (2001, p. 348).94 See Ferguson (2001, pp. 377–78).95 We would expect similar lack of difference be-

tween honors students and middle students atShopping Mall High, except insofar as these two groupsmix.

try to get teachers off track in class.”96 Thebehavior that students would define as “be-ing competitive,” “volunteering answers,” or“getting teachers off track” is likely to differby group. For example, while nerds andburnouts both disapprove of “being competi-tive about grades,” the objective behaviorenvisaged by the two groups will be very different. Ferguson’s accompaniment ofethnographic observation with his statisticalanalysis puts such responses in context. Itsuggests the need for further work that basesattitudinal questions on ethnographic obser-vation. Those questions should solicit stu-dents’ attitudes toward objectively definedbehaviors, with no ambiguity that all stu-

dents, regardless of their social group, willhave the same objective interpretation of thequestions. Such study, in our opinion, is thenext natural step in resolution of the Cook-Ludwig paradox.

6. Further Implications for FutureEmpirical Work

Here we sketch some further implicationsof the model for empirical work. Thus far,the empirical work that we have discussedhas, mostly, used non-experimental data—as, for example, in the studies of Cook andLudwig and of Ferguson. But we have men-tioned both controlled and natural experi-ments. For example, the Comer schoolswere a controlled experiment, Hamilton


96 See Ferguson (2001, pp. 355–57) for these ques-tions and others.

TABLE 3ATTITUDES OF STUDENTS BLACK/WHITE: ODDS RATIO FROM LOGIT REGRESSIONS

Positive Attitude Think I Am School Spirit View School Toward Selfa Depressedb No Good At Allc Excellentd Discipline as Faire

Male

Black 2.27** 1.51** 1.12 1.13** 1.03(.09) (.15) (.09) (.05) (.07)

White .44** .66** .89 .88** .97(.02) (.07) (.07) (.04) (.07)

Female

Black 3.65** 1.52** .84* 1.15** 1.28**

(.13) (.11) (.06) (.04) (.09)

White .27** .66** 1.19* .87** .78**

(.01) (.05) (.08) (.03) (.05)

Independent Variable

DependentVariable

Source: High School and Beyond. aPositive attitude toward self: agrees strongly with statement “I take a positiveattitude of myself”; bDepressed: Respondents (sophomores only) felt depressed or unhappy a lot in past fewweeks; cThink I am no good at all: Strong agreement with statement “At times I think I am no good at all”;dRespondent rates school spirit as excellent; eRespondent rates fairness of discipline as excellent; fRespondent an-swers true to: Do you dread English class? gRespondent answers true to: Do you dread math class?; hACT or SAT:seniors only; iBored a lot: Respondent felt bored a lot in last few weeks; jWorking hard at school: respondent likes

High a natural experiment. Indeed, Hamil-ton High gives a dramatic instance of aschool’s response to an exogenous (in thiscase, court-ordered) change in the diversityof its student body. For this reason it is agood test of our model. Hampel (1986) re-lates similar disturbances in other schoolswith forced integration in the same time pe-riod. The work of Grant and Hampel shouldbe carried further. Whether changes in testscores relate to court-ordered busing can beexamined statistically. The model predictsthat, as in Hamilton High, the impulse re-sponse to court-ordered busing will be a re-duction in test scores, followed by a less thancomplete recovery.

Another example where our theory sug-gests a broader scope of empirical work inthe economics of education comes from

Krueger and Whitmore (1999). They exam-ined the consequences of Tennessee’s Starexperiment, in which students in grades K to3 were randomly assigned to small classes.For the duration of the experiment, blackand free-lunch subjects had large and statis-tically significant gains in the StanfordAchievement Test. After returning to nor-mal-sized classes, these gains disappeared.Despite this disappearance, eight years afterthey had left the experiment, blacks who hadbeen in the small classes were twenty-fivepercent more likely to take college entranceexams (ACT or SAT) than those in controlgroups.97 Our perspective suggests that suchan astounding finding should be followed up,


97 40.2 percent compared to 31.7. Krueger andWhitmore (1999, p. 11).

TABLE 3 (cont.)

Dread English Dread Math Taken ACT Bored a Like Working Disappointed if Do Classf Classg or SATh Loti Hard at Schoolj Not Go to Collegek

Male

.55** .86* .90* 1.14 2.23** 1.78**

(.04) (.05) (.04) (.08) (.09) (.08)

1.83** 1.16** 1.11** .88 .45** .56**

(.12) (.07) (.05) (.06) (.02) (.02)

Female

.67** .73** .86** 1.39** 2.04** 2.43**

(.04) (.04) (.04) (.08) (.08) (.10)

1.49** 1.37** 1.16** .72** .49** .41**

(.09) (.07) (.05) (.04) (.02) (.02)

working hard at school; kDisappointed if do not go to college: agreement with statement “I will be disappointed ifI do not go to college.”*Significant at 5% level; **significant at 1% level. Example: Odds Ratio (OR) = Odds for Depressed (BlackMale)/Odds for Depressed (non-Black Male) =�1.51, where Odds for depressed = probability of depressed/proba-bility of not depressed; logit (p) = a + bX; OR = eb.

since the economics of education should notjust tabulate the effects of experiments.Research should also ascertain how and whythe experimental subjects were subsequentlydifferent.

Our theory suggests a possible testable ex-planation: the familiar environments af-forded by small class sizes trumped thesestudents’ tendency to view school as a placefor others, rather than for them. Otherwise,it is hard to explain why almost a decade laterthey were more likely to see themselves assuitable candidates for colleges requiring en-trance exams. Empirically we would expectto see this difference in attitude towardschool reflected in greater participation inschool activities.

Using currently available data, it is possi-ble to use measures of school participationto test the hypothesis that successful pastexperiments are associated with a greatersense of belonging to the school commu-nity. High-school yearbooks provide an in-dicator of school participation. Typicallythe captions of photographs of sports teamsand clubs list the names of participants.Also, appearance for the graduation picturein the yearbook itself is another (minimal)form of school participation. With the Starexperiment, even now, it would be possibleto use such data to test for differences inhigh-school participation by those in theexperimental and the control groups. Itwould also be possible to see whether such


TABLE 4.COEFFICIENTS FROM LOGIT REGRESSIONS FOR DREAD ENGLISH/DREAD MATH BY BLACK/WHITE

AND RACIAL COMPOSITION OF SCHOOL

Dread Englishd Dread Mathe Dread English Dread MathN =�2995 N =�3011 N =�19727 N =�19678

Blackf Whitef

All White in Schoola .76* .72* — —(.34) (.32)

Few Black in School .24 .11 .10** .08*

(.15) (.14) (.04) (.04)

Half Black in School .18 .06 .05 .01(.14) (.13) (.05) (.05)

Mostly Black in School .12 .02 .07 .10(.15) (.13) (.09) (.09)

All Black in Schoolb — — 1.15** .31(.31) (.31)

SESc .48 .50 3.52** 1.52**

(.62) (.58) (.23) (.22)

Constant 1.47** 1.09** .77** .77**

(.12) (.11) (.03) (.03)

Source: High School and Beyond. aNo black students in respondents’ ninth grade; ball black students in respon-dents’ ninth grade; cDuncan index of parents’ socioeconomic status from parents’ survey (measure of SES is theDuncan index times 10,000); cno black students in respondents’ ninth grade; dAffirmative response: Do you dreadEnglish class?; eAffirmative response: Do you dread math class?; fInclusive of Hispanics by race.Standard errors are reported in parentheses. **Significant at 1% significance level; *significant at 5% significancelevel. Regressions for Black: All black in school was omitted dummy. Regressions for White: All white in schoolwas omitted dummy. Regression: logit(p) = a + bX.

difference in participation was correlatedwith taking the ACT.

Such follow-up is also possible for theComer experiment. Test scores at theComer schools rose markedly in the first fiveyears of the experiment. The yearbookrecords of these students in high schoolshould show whether the effects of the ex-periment continued, in the form of greaterparticipation in school life. This evidenceshould be particularly powerful since thehigh-school record of increased participa-tion should mirror the increasing success ofthe experiment in its first five years.98 Suchstudies would expand the boundaries of em-pirical work in the economics of education,as their goal is not only to measure the con-sequences of changes in resources, but also,more deeply, to resolve, the reasons forthese consequences.

While a considerable econometric litera-ture has reexamined the assertions of supe-rior performance of Catholic and other pri-vate schools by Coleman and others, there isstill the need to reexamine the more ambi-tious assertions in this literature regardingthe reasons for such superiority. Bryk et al.and Chubb and Moe have established goodcorrelations, respectively, between indicesof community and of school independenceand student performance. But causality runsin two directions—especially since affluentschool districts and schools with high pro-portions of middle class students are likelyto find it easier both to establish communityand to give school administrators large scope

for discretion.99 Researchers should reexam-ine the extent to which school communityand administrator independence causally affect student performance.

7. Conclusion

This review considers a sociological per-spective on students and schools and inte-grates this perspective into an economicmodel. This view focuses on variables—identity, social categories, and schoolideals—that have been beyond the purviewof the economic theory of education. Theframework we suggest captures phenomenathat could not be explained by standard eco-nomic modeling. It easily explains the se-quence of events at Hamilton High, the im-plications of school discipline and changes inthe legal status of students, the success ofnotable school reform experiments such asCPESS and the Comer schools, the trade-offs in the Shopping Mall High, and argu-ments of non-economists in favor of schoolvouchers and charter schools.

The last model incorporates all the previ-ous modeling in the paper to consider the useof school resources to change student identi-ties. The model combines the fundamentalconcern of economists—the allocation of re-sources—with a basic concern of educationscholars—the dependence of students’


98 Establishing the role of school participation from such experiments would test the hunches ofHollingshead in Elmtown’s Youth, where he found astrong relation between participation in school activi-ties and social class. (See Hollingshead 1975, pp.143–50). Participation was especially low by the stu-dents of the lowest class, who lived near the canal andhad a high rate of drop out. A girl of one of the lowerclasses explains to Hollingshead the reasons why shedoes not go to football games: “Well, why go? We’remade to feel out of place and that is the way it is.”(1975, p. 150). This paper has explored the conse-quences of such feelings much more generally.

99 For example, Chubb and Moe (1990) define lowand high performance schools as those respectively inthe bottom and top quartiles of sophomore-to-seniorimprovement on a test administered to the HighSchool and Beyond panel. In the high performanceschools principals had more motivation to control theschools. In schools in the top quartile of improvement,61.9 percent had higher than average motivation tocontrol, as measured by their index, in contrast to only18.9 percent in the bottom quartile. (See Chubb andMoe 1990, p. 85). These tabulations and others show adifference in performance between successful and un-successful schools in the exercise of leadership of prin-cipals. While Chubb and Moe are well aware thatcausality runs in both directions, their attempts to es-tablish causality of school independence are not con-vincing. Quite possibly with the type of panel dataavailable to them from High School and Beyond suchcausality cannot be identified.

achievement on the gap between their ownself-images and the person the school intendsthem to be. In our reading of the noneco-nomic literature, we were struck by the ab-sence of comment on the role of resources inschooling. Consider three examples. InPowell et al.’s account of Shopping Mall High,a defining moment occurs when a teacherfails to reprimand a student for complainingabout an assignment. The authors considerthis response as a failure to set standards andchange the student’s own expectations ofwhat she could achieve: in the language ofour models she failed to promote an aca-demic ideal. But straightforward economicsoffers another interpretation: the teacher isincompetent or lazy. If incompetent, theschool lacks resources; if lazy, she lacks theright incentives. Descriptions of CPESS andthe Comer Schools similarly fail to considerthe exceptional resources available to theseschools, not in terms of budget, perhaps, butin terms of exceptional teaching and adminis-tration.

The perspective offered by our models isthus consistent with the view of Hanushek(1986) and Ferguson (1998b) that resources,especially in the form of high teacher quality,are critical ingredients for school reform. Butthe sociological perspective says more. It saysthat the resource use is likely to be ineffectivewhen students’ backgrounds are antithetical tothe academic values that schools should pro-mote and the schools are constrained againstinvesting in identity.100,101 This perspective, aswe have seen, offers a possible explanationwhy the experimental seniors in the Star ex-periment had such a high rate of taking theACT or SAT exam. Sizer’s (1984) claims aboutteachers and student self-image can similarlybe interpreted as a reason for a relation

between resources and academic achieve-ment. Sizer sees the typical U.S. high school as unable to affect student character since thetypical teacher meets with 120 students every day,102 making close student-teacher relations all but impossible.

This synthesis of the sociology and the eco-nomics shows why new research in econom-ics should delve into the black box of theschool. Our task is not only to determinewhether an experiment or resources increaseeducational achievement. Our task is also toexplain why. The sociological approach out-lines possible answers, and directions for newresearch within the economics of education.

In closing we should mention two impor-tant omissions from this review and from itstheoretical framework. We have focused onthe identity and behavior of students, butthe motivations of teachers and administra-tors are also keys to the success of differentschools. As we saw briefly in the accounts ofCPESS, the teachers and principal identi-fied with the goals of the school. Indeedthey created them. The committed CatholicSchool teachers and administrators, de-scribed by Bryk et al., contrasts with the dis-engaged public school teacher, described bySizer.103 Thus the extent to which teachersidentify with their school’s mission may be asimportant as differences in student motiva-tion in explaining the gap in performancebetween the two systems.

Second, reflecting our own bias as econo-mists, our models assume that schools maxi-mize students’ marketable skills. But skills areonly one of the goals of some schools and ofthe parents who choose them. Religiousschools, for example, often eschew economicgoals in favor of religious goals. In somecases, they view their primary mission as theseparation of the saved from the damned, as


100 On average teachers with greater academic skillswill be better motivated themselves, and will thus morenaturally motivate their students.

101 Jeffrey Mirel (1999) has documented, in the caseof Detroit, the decline in the resources available for anever more difficult task in maintaining a pro-education,pro-academic morale for the period 1930–80.

102 Sizer (1984, p. 97).103 See Bryk et al. (1993) and Sizer (1984). Parents’

and administrators’ identities also play a role in the op-eration of the school. Comer (1980) and Ogbu (1997)emphasize the role of parents in their children’s identi-fication with or against school.

suggested by the fears of the principal inPeshkin’s ethnography: “The devil’s crowd[the unconverted] is after our kids.”104 A simi-lar desire for separation lay behind thevoucher-supported private academies estab-lished (unconstitutionally) in the wake ofBrown v. Board of Education of Topeka,Kansas (Jeffrey Henig 1993).105 The goals andcurricula of public schools are the product ofelected school boards; the nature of theseschools, their ideals, may then derive from the political economy of a community.106

If schools’ goals include promoting certainideologies, school choice may be neither skill-increasing nor ideologically neutral.107

REFERENCES

Akerlof, George A. and Rachel E. Kranton. 2000.“Economics and Identity,” Quart. J. Econ. 105:3, pp.715–53.

Alesina, Alberto; Reza Baqir and William Easterly.1999. “Public Goods and Ethnic Divisions,” Quart.J. Econ. 114:4, pp. 1243–84.

Altonji, Joseph G.; Todd R. Elder and Christopher R.Taber. 2000. “Selection on Observed and Unob-served Variables: Assessing the Effectiveness ofCatholic Schools,” NBER work. paper 7831.

Banks, James A. and Dresden Grambs, eds. 1972.Black Self-Concept: Implications for Education andSocial Science. NY: McGraw-Hill.

Becker, Gary. 1964. Human Capital: A Theoretical andEmpirical Analysis with Special Reference to Educa-tion. NY: Columbia U. Press.

Benabou, Roland. 1993. “Workings of a City: Location,Education, and Production,” Quart. J. Econ. 108:3,pp. 619–52.

Betts, Julian R. 1995. “Does School Quality Matter?Evidence from the National Longitudinal Survey ofYouth,” Rev. Econ. Statist. 77:2, pp. 231–50.

Bishop, John H. 1998. “Nerd Harassment, Incentives,School Priority and Learning,” mimeo, Cornell U.

Bloom, Allan. 1987. The Closing of the AmericanMind. NY: Simon and Schuster.

Bonny, Andrea; Maria T. Britto, Brenda K. Kloster-mann, Richard W. Hornung and Gail S. Slap. 2000.“Social Disconnectedness: Identifying Adolescentsat Risk,” Pediatrics 106:5, pp. 1017–21.

Bowles, Samuel and Herbert Gintis. 1976. Schooling inCapitalist America: Educational Reform and theContradictions of Economic Life. NY: Basic Books.

Bowles, Samuel; Herbert Gintis and Melissa Osborne.2001. “The Determinants of Earnings: A BehavioralApproach,” J. Econ. Lit. 39:4, pp. 1137–76.

Brown, Roger. 1986. Social Psychology: The SecondEdition. NY: Free Press.

Bryk, Anthony S.; Valerie E. Lee and Peter B. Holland.1993. Catholic Schools and the Common Good.Cambridge, MA: Harvard U. Press.

Card, David and Alan B. Krueger. 1992a. “Does SchoolQuality Matter? Returns to Education and the Char-acteristics of Public Schools in the United States,” J.Polit. Econ. 100:1, pp. 1–40.

———. 1992b. “School Quality and Black-White Rela-tive Earnings: A Direct Assessment,” Quart. J. Econ.107:1, pp. 151–200.

Chubb, John E. and Terry Moe. 1988. “Politics, Mar-kets, and the Organization of Schools,” Amer. Polit.Sci. Rev. 82:4, pp. 1065–87.

———. 1990. Politics, Markets and America’s Schools.Washington, DC: Brookings Institution Press.

Cohen, Jere. 1979. “High School Subcultures and theAdult World,” Adolescence 14:55, pp. 491–502.

Coleman, James S. 1961. The Adolescent Society: TheSocial Life of the Teenager and Its Impact on Educa-tion. NY: The Free Press.

Coleman, James S.; Thomas Hoffer and Sally Kilgore.1982. High School Achievement: Public, Catholicand Private Schools Compared. NY: Basic Books.


104 See Peshkin (1985, p. 289).105 Some African-American charter schools also have

separatist teachings (Lynell George 1992).106 See Foley (1990) for school board battles and the

history of Anglo and Mexican-American relations in theWest Texas town he studied. Many current educationpolicies, including vouchers, sex education in publicschools, bilingual instruction, etc. are being deter-mined at the ballot box.

107 The political economy model of Kremer andSarychev (1998) deals nicely with the issue of ideologi-cal choice in public and private schools. In their modelparents choose among schools that promote differentideologies. After their education, people with differentideologies have less gains from trade. Parents wanttheir children to be like them ideologically, but theyalso want their children to be able to communicatewith others for economic gain. A school’s ideal in ourdescriptive model is similar to school ideology inKremer and Sarychev, and our model produces a re-lated trade-off. Schools interested in students’ skillshave a reason to choose the ideology of the dominantculture because that is the idiom of commerce, butthey have reason to mirror their students’ preferences.Our description of school reform which empowers in-vestments in identity and relaxes the public school con-straint in these investments yields different analysis ofthe potential benefits and costs of school privatization.Consider the seeming paradox in the Netherlands ofideological homogeneity and private schooling, whichKremer and Sarychev explain as the lucky draw frommultiple equilibria. In our descriptive model, privateschooling will result in more homogeneity than publicschooling if parents care sufficiently about children’sskills. Private schools will devote resources, a, to reduceany conflict between students’ identity and the domi-nant culture, and then choose an ideal, s, that conformsto the dominant culture as well. In contrast, con-strained public schools must mimic the heterogeneityalready existing in society. Friedman (1955) discussesthe issue of school ideological concerns.

Comer, James P. 1980. School Power: Implications ofan Intervention Project. NY: The Free Press.

———. 1988. Maggie’s American Dream: The Life andTimes of a Black Family. NY: New American Li-brary.

Conant, James B. 1959. The American High School To-day: A First Report to Interested Citizens. NY: McGraw-Hill.

Cook, Philip J. and Jens Ludwig. 1997. “Weighing theBurden of ‘Acting White’: Are There Race Differ-ences in Attitudes Toward Education?” J. Polit.Anal. Manage. 16:2, pp. 256–78.

Delpit, Lisa. 1995. Other People’s Children: CulturalConflict in the Classroom. NY: The New Press.

Dupre, Anne Proffitt. 1996. “Should Students HaveConstitutional Rights? Keeping Order in the PublicSchools,” George Washington Law Rev. 65:1, pp.49–105.

Eckert, Penelope. 1989. Jocks and Burnouts: SocialCategories and Identity in the High School. NY:Teachers College Press.

Epple, Dennis and Richard E. Romano. 1998. “Com-petition Between Private and Public Schools, Vouch-ers, and Peer Group Effects,” Amer. Econ. Rev. 88:1,pp. 33–62.

Erikson, Kai. 1966. The Wayward Puritans: A Study inthe Sociology of Deviance. NY: Wiley.

Evans, William N. and Robert M. Schwab. 1995. “Fin-ishing High School and Starting College: DoCatholic Schools Make a Difference?” Quart. J.Econ. 110:4, pp. 941–74.

Evans, William N.; Wallace E. Oates and Robert M.Schwab. 1992. “Measuring Peer Group Effects: AStudy of Teenage Behavior,” J. Polit. Econ. 100:5,pp. 966–91.

Everhart, Richard. 1983. Reading, Writing, and Resis-tance: Adolescence and Labor in a Junior HighSchool. London: Routledge & Kegan Paul.

Ferguson, Ronald F. 1998a. “Teachers’ Expectationsand the Black-White Test Score Gap,” in The Black-White Test Score Gap. Christopher Jencks andMeredith Phillips, eds. Washington, DC: BrookingsInstitution Press, pp. 273–317.

———. 1998b. “Can Schools Narrow the Test ScoreGap?” in. Jencks and Phillips, op. cit., pp. 318–74.

———. 1998c. “Comment” on Philip J. Cook and JensLudwig, “Weighing the ‘Burden of ‘Acting White’:Are There Race Differences in Attitudes TowardEducation?” in Jencks and Phillips, op. cit., pp.394–97.

———. 2001. “A Diagnostic Analysis of Black-WhiteGPA Disparities in Shaker Heights, Ohio,” Brook-ings. Pap. Educ. Pol. 2001, pp. 347–414.

Fernandez, Raquel and Richard Rogerson. 1996. “In-come Distribution, Communities, and the Quality ofPublic Education,” Quart. J. Econ. 111:1, pp.135–64.

Fliegel, Seymour. 1993. Miracle in East Harlem: TheFight for Choice in Public Education. NY: RandomHouse.

Foley, Douglas E. 1990. Learning Capitalist Culture:Deep in the Heart of Texas. Philadelphia: U. Penn-sylvania Press.

Fordham, Signithia. 1996. Blacked Out: Dilemmas ofRace, Identity and Success at Capital High. Chicago:U. Chicago Press.

Friedman, Milton. 1955. “The Role of Government inEducation,” in Economics and the Public Interest.Robert A. Solo, ed. New Brunswick, NJ: Rutgers U.Press, pp. 123–44.

Friedman, Milton and Rose Friedman. 1980. Free toChoose: A Personal Statement. NY: Harcourt Brace.

Gaviria, Alejandro and Steven Raphael. 1997. “School-Based Peer Effects and Juvenile Behavior,” U. Cal.San Diego discuss. paper 97–21.

George, Lynell. 1992. No Crystal Stair: African-Amer-icans in the City of Angels. London: Verso.

Goldin, Claudia and Lawrence F. Katz. 1997. “Why theUnited States Led in Education: Lessons from Sec-ondary School Expansion, 1910–1940,” NBER work.paper 6144.

Grant, Gerald. 1988. The World We Created at Hamil-ton High. Cambridge, MA: Harvard U. Press.

Hampel, Robert L. 1986. The Last Little Citadel:American High Schools Since 1940 . Boston:Houghton Mifflin.

Hanushek, Eric A. 1971. “Teacher Characteristics andGains in Student Achievement,” Amer. Econ. Rev.Pap. Proceed. 61:2, pp. 280–88.

———. 1986. “The Economics of Schooling: Produc-tion and Efficiency of Public Schools,” J. Econ. Lit.24:3, pp. 1141–77.

Hanushek, Eric A.; Steven G. Rivkin and Lori L. Tay-lor. 1996. “Aggregation and Estimated Effects ofSchool Resources,” Rev. Econ. Statist. 78:4, pp.611–27.

Harris, Judith R. 1998. The Nurture Assumption: WhyChildren Turn Out the Way They Do. NY: Simonand Schuster.

Henig, Jeffrey R. 1993. Rethinking School Choice:Limits of the Market Metaphor. Princeton: PrincetonU. Press.

Hollingshead, August B. 1949. Elmtown’s Youth: TheImpact of Social Class on Adolescents. NY: John Wi-ley.

———. 1975. Elmtown’s Youth and Elmtown Revis-ited. NY: John Wiley.

Hoxby, Caroline M. 1996. “The Effects of PrivateSchool Vouchers on Schools and Students,” inHolding Schools Accountable: Performance-Based Reform in Education. H. Ladd, ed. Wash-ington, DC: Brookings Institution Press, pp.177–208.

———. 2000. “The Effects of Class Size on StudentAchievement: New Evidence from Population Varia-tion,” Quart. J. Econ. 115:4, pp. 1239–85.

Jencks, Christopher and Meredith Phillips. 1998. TheBlack White Test Score Gap. Washington, DC:Brookings Institution Press.

Kremer, Michael and Andrei Sarychev. 1998. “Why DoGovernments Operate Schools?” mimeo, M.I.T.

Krueger, Alan B. and Diane M. Whitmore. 1999. “TheEffect of Attending a Small Class in the EarlyGrades on College-Test Taking and Middle SchoolTest Results: Evidence from Project STAR,” mimeo,Ind. Rel., Princeton U.


Krug, Edward A. 1964. The Shaping of the AmericanHigh School: 1880–1920. Madison, WI: U. Wiscon-sin Press.

———. 1972. The Shaping of the American HighSchool: 1920–1941. Madison, WI: U. WisconsinPress.

Levine, David I. and Gary Painter. 2000. “Are Mea-sured School Effects Just Sorting?” mimeo, U. Cal.Berkeley.

Manski, Charles F. 1992. “Educational Choice (Vouch-ers) and Social Mobility,” Econ. Ed. Rev. 11:4, pp.351–69.

Meier, Deborah. 1995. The Power of Their Ideas.Boston: Beacon Press.

Mickelson, Roslyn A. 1990. “The Attitude-Achieve-ment Paradox Among Black Americans,” SociologyEd. 63:1, pp. 44–61.

Miguel, Edward. 1999. “Ethnic Diversity and SchoolFunding in Kenya,” mimeo, Harvard U.

Mirel, Jeffrey. 1999. The Rise and Fall of an UrbanSchool System: Detroit, 1907–1981, 2nd ed. Ann Ar-bor: U. Michigan Press.

Ogbu, John U. 1974. The Next Generation: An Ethnog-raphy of Education in an Urban Neighborhood. NY:Academic Press.

———. 1997. “Beyond Language: Ebonics, ProperEnglish and Identity in a Black American SpeechCommunity,” mimeo, U. Cal. Berkeley, Dept. An-thropology.

Powell, Arthur G.; Eleanor Farrar, and David K. Co-hen. 1985. The Shopping Mall High School: Winnersand Losers in the Educational Marketplace. Boston:Houghton Mifflin.

Peshkin, Allen. 1986. God’s Choice: The Total World ofa Fundamentalist Christian School. Chicago: U.Chicago Press.

Ravitch, Diane. 1983. The Troubled Crusade: Ameri-can Education 1945–1980. NY: Basic Books.

Rumbaut, Rubén G. 2000. “Children of Immigrantsand Their Achievement: The Role of Family, Accul-turation, Social Class, Gender, Ethnicity, and SchoolContexts,” mimeo, Michigan State U.

Sacerdote, Bruce. 2000. “Peer Effects with RandomAssignment: Results for Dartmouth Roommates,”NBER working paper 7469.

Sadowski, Michael. 2001. “Closing the Gap OneSchool at a Time,” Harvard Education Letter,May/June. http://www.edletter.org/past/issues/2001–mj/gap.shtml.

Schultz, Theodore W. 1960. “Capital Formation by Ed-ucation,” J. Polit. Econ. 68, pp. 571–83.

Sizer, Theodore R. 1984. Horace’s Compromise: WhatWorks for the American High School. Boston:Houghton Mifflin.

———. 1996. Horace’s Hope: What Works for theAmerican High School. Boston: Houghton Mifflin.

Steele, Claude M. and Joshua Aronson. 1998. “Stereo-type Threat and Test Performance of AcademicallySuccessful Americans,” in Jencks and Phillips, op.cit., pp. 401–27.

Summers, Anita A. and Barbara L. Wolfe. 1977:2. “DoSchools Make A Difference?” Amer. Econ. Rev. Pap.Proceed. 67, pp. 639–52.

Tajfel, Henri and John Turner. 1979. “An IntegrativeTheory of Intergroup Conflict,” in The Social Psy-chology of Intergroup Relations, William G. Austinand Stephen Worchel, eds. Monterey, CA:Wadsworth, pp. 33–47.

Tiebout, Charles M. 1956. “A Pure Theory of LocalExpenditures,” J. Polit. Econ. 64:5, pp. 416–24.

Tyack, David. 1974. The One Best System. Cambridge,MA: Harvard U. Press.

Valenzuela, Angela. 1999. Subtractive Schooling: U.S.-Mexican Youth and the Politics of Caring. Albany:State U. New York Press.

Weiss, Lois. 1990. Working Class Without Work: HighSchool Students in a De-industrializing Economy.NY: Routledge.

Wetherell, Margaret. 1996. “Group Conflict and theSocial Psychology of Racism,” in Identities, Groupsand Social Issues. Margaret Wetherell, ed. Thou-sand Oaks, CA: Sage Pub., pp. 219–27.

Willis, Paul R. 1977. Learning to Labour: How Work-ing Class Kids Get Working Class Jobs. Westmead,Farnborough, Hants, England: Saxon House.

Wynne, Edward A. and Kevin Ryan. 1997. ReclaimingOur Schools: Teaching Character, Academics andDiscipline. 2nd ed. Upper Saddle River, NJ: Pren-tice-Hall.


identity and schooling: some lessons for the economics · pdf fileidentity and schooling: some...

Documents