Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.
http://www.jstor.org
TheWorldwideExpansionof Higher
Educationin the TwentiethCentury
Evan Schofer John W. Meyer
Universityof Minnesota StanfordUniversity
The authors analyze the rapid worldwide expansion of higher educational enrollments
over the twentieth century using pooled panel regressions. Expansion is higher in
economically developed countries (in some but not all analyses) as classic theories
would have it. Growth is greater where secondary enrollments are high and where state
control over education is low, consistent with conflict and competition theories.
Institutional theories get strong support. growth patterns are similar in all types of
countries, are especially high in countries more linked to world society, and sharply
accelerate in virtually all countries after 1960. The authors theorize and operationalize
the institutional processes involved, which include scientization, democratization and the
expansion of human rights, the rise of development planning, and the structuration of the
world polity. Withthese changes, a new model of society became institutionalized
globally-one in which schooled knowledge and personnel were seen as appropriatefor
a wide variety of social positions, and in which many more young people were seen as
appropriate candidates for higher education. An older vision of education as
contributing to a more closed society and occupational system-with associated fears of
"over-education "-was replaced by an open-system picture of education as useful
"humancapital "for unlimited progress. The global trends are so strong that developing
countries now have higher enrollment rates than European countries did only afew
decades ago, and currently about one-fifth of the world cohort is now enrolled in higher
education.
AMERICANSOCIOLOGICAL
REVIEW,2005, VOL. 70 (December-898-92o)
EXPANSION
WORLDWIDE OF HIGHEREDUCATION 899
Prior studies have noted growing tertiary ly affectedit. Meyeret al. (1977), in theircross-
enrollments in particular societies, and have nationalstudyof tertiaryenrollmentratiosfrom
typically invokedexplanatoryframeworksthat 1950 to 1970, found very rapid increases in
emphasized the local context. In the United enrollments in all types of countries. Tertiary
States, it is conventionalto call attentionto the expansion, in fact, considerably outpaced the
"GI Bill." In fact, the GI Bill produces only a rapid expansion of mass education during the
minor "blip" that is dwarfed by subsequent
period.Similarly,Windolf's(1997) longitudinal
expansion (see section A of the ASR Online study of enrollments in several industrialized
Supplement: http://www2.asanet.org/ countriesfrom 1850 to 1990 finds rapidgrowth,
journals/asr/2005/tocO48.html). InThirdWorld
especially afterWorldWarII.Yetthese studies,
countries,scholarscommonly point to decolo-
too, were unableto find countrycharacteristics
nization as generating expansion. In Europe, that strongly affect the rate of expansion-not
case researchersrefer to Americaninfluences,
even standardvariablessuchas economic devel-
the weakening of standards associated with
democratization,or the local political context opment.1
The character and extent of the overall
(Wittrock 1993). These idiosyncraticexplana-
tions may help make sense of some nationalor explanatoryproblem is indicated in Figure 1,
which presentsglobalhighereducationalenroll-
period-specific variation,but they have diffi-
culty accounting for what turns out to be a ments per 10,000 capitaover the twentiethcen-
ratheruniversalglobal trend. tury.The figure shows the extremegrowththat
Comparativeresearchershave made the crit- is concentratedin the latterpartof the century.
ical observation that the twentieth-century
expansion of higher education is a worldwide
phenomenon.Riddle's(1990, 1993) studyof the 1 Windolf
doesfindsomemodesteffects,suchas
historical formationof universities found that a positiveeffectof unemployment,whichgenerates
expansion is very general in the modern peri- a short-termincreasein enrollments.Nevertheless,
od, andwas unableto isolate factorsthatstrong- thefundamental causesof expansionremainunclear.
180
160
.-140
S120
100
800
60
40
20
00
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Year
Figure1. WorldHigher-Education
Studentsper 10,000Capita,1900-2000.
goo AMERICAN REVIEW
SOCIOLOGICAL
We bring to this problemnew dataand argu- previously reviewed all suggest the following
ments. Our data spans a longer period of time hypothesis:
and more countries than do prior studies. We
Hypothesis1: Economicdevelopmentproduces
conduct a more sophisticatedset of analyses of the expansion of higher education.
enrollmentexpansion, and we provide a more
systematic and comprehensiveexaminationof In fact, however, empirical researchers do
competing arguments.Moreover,we develop a not find very strong effects of national indus-
new global explanationto accountfor the broad- trialization or economic development on the
based postwar expansion of higher education. expansion of education at any level (Meyer et
We argue that a new model of society became al. 1977, 1992b;Windolf 1997). Moreover,the
institutionalized in that period, reflected in rapidexpansionof highereducationin the 1960s
trends toward increasing democratization, does not coincide with especially large histor-
human rights, scientization, and development ical changesin occupationalstructures,job skill
planning.This global institutionaland cultural requirements,2or labor market demands that
change paved the way for hyper-expansionof would create a need for massive expansion of
highereducation.We also contributeto the neo- higher education(Wyatt2005; see also Morris
institutionalliterature,in which ourargumentis and Western 1999). In response to the theoret-
rooted,by providinggreaterprecision-in argu- ical and empirical problems of functionalist
mentandoperationalization-regardingthe his- arguments, two broad lines of thought have
torical driversof institutionalchange. evolved. Institutionaltheory treats functional
We begin with a discussion of nationalvari- ideas aboutsocioeconomic developmentas cul-
ation in higher education. Later, we develop tural or ideological myths and models rather
our global institutionalargumentand turnto a than as reflections of functional realities.
statistical analysis of higher-educationenroll- Conflictandorganizationaltheoriestreatsocioe-
ments from 1900 to 2000. conomic development as providing resources
and grounds for competition rather than for
OF NATIONAL
EXPLANATIONS functional efficiency. We discuss these theo-
VARIATION ries in turn.
FUNcTIONALISM OFCONFLICT,
THEORIES AND
COMPETITION,
Traditionalsocioeconomicfunctionalism,often ORGANIZATION
employedas a baselineor strawman in the field A very distinctiveidea arises from strongcon-
(Collins 1971), argues that national economic flict and competition theories: as education
development, rationalization,and differentia- becomes importantin the attainmentof social
tion createcollectiveandindividualdemandsfor
both specialized trainingand loyalty to a com- status, groups and individuals compete more
mon culture. On the supply side, development intensivelyfor success in education,producing
also creates the resources to meet the needs. inflationary credential expansion far beyond
any originalfunctionalrequirements(Bourdieu
Higher education can also be seen as a con- and Passeron 1977; Collins 1971, 1979).
sumptiongood thatmight especially be desired A principaltheme in conflict theories is that
by individuals in societies where basic needs elite groups use education to perpetuate the
have been met. The core propositionis thatthe
dominance of their status-group culture, not
expansionof highereducationis stronglyaffect-
ed by national economic development. simply to ensure their children's success
Traditionalfunctionalism also has variants (Bourdieu and Passeron 1977; Collins 1971,
that shift the main argumentfrom the national 1979). This theme has implicationsfor the con-
to the global level. Here the expansionof high-
er education in core countries aids in coordi-
nating or controllingrelationshipsin the world 2AsMorrisandWestern (1999:34)pointout,"The
(see Clark 1992; Dale and Robertson 2002). evidencefor a generalincreasein the demandfor
Again, the predictionschangevery little despite skills is weak... [andthe] evidencefor the more
variations in political tone (e.g., Wallerstein specifichypothesisof a technology-drivenincrease
1974 and elsewhere).Thus,the lines of thought in demandforskillsis equallyweak."
OF HIGHEREDUCATION gox
EXPANSION
WORLDWIDE
350
300
?
250
.
S200
/.
o
S150
E 10
z
E 50
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Year
-x- Democratization --o-- HumanRights -- Scientization
--- Development
Planning - WorldPolityStructuration
global ones. Most important,nations are not social distinctionsand obtainadvantagesin the
equally likely to conform to models that are stratificationsystem. Thus,
taken for granted in the world polity. World
Hypothesis 7: Higher educationexpands more
influences may strongly affect Europeansoci-
rapidly when secondary education enroll-
eties, which are enmeshed in the European ments are high.
Union and countless international organiza-
tions, and less strongly affect peripheralsoci-
eties such as North Korea or Bhutan. Nations
ANALYSIS:DATAAND
QUANTITATIVE
more densely linked to the worldpolity should
METHODS
conform the most (see Schofer and Hironaka
2005). We turnnow to quantitativeanalyses of growth
in higher-educationalenrollments from 1900
Hypothesis 5: Higher educationexpands most
to 2000. Our primary dataset comprises ten
rapidlyin countrieslinked in organization
and identity to world models. decadal cross-national panels that are pooled
together.Panels include the lagged dependent
An implicationof strongversions of institu- variable and other covariates measured 10
tional theories is that countriesrespond to the years prior to the dependent variable.
pressures of world models more than to their Compared with cross-sectional designs, this
own histories.Indeed,if theirown historiespro- approachreducesconcernsregardingthe direc-
duce globally unacceptable effects, they are tion of causality. We employ a randomeffects
likely to correct themselves. This produces a generalized least squares (GLS) regression
prediction dramaticallyopposed to the ideas model with robust standard errors, because
aboutinertiafound in much organizationalthe- pooling violates the independenceassumption
ory, and suggests a kind of negative inertia: of ordinary least squares (OLS) regression.
Hypothesis6: Countrieswith low ratesof high- Thus, we include an additionalerrorterm (Ui)
er-educationalexpansion in given periods to model the correlated errorthat occurs with
tend to have "corrective"higher rates in the same country appearingmultiple times in
our dataset:
subsequentperiods.
= + Ui + ei (1)
Yi(t2)iYl a + bXi(tl) + cYi(tl)
~I
THEEXPANSION
OFSECONDARY
EDUCATION Variablesused in the analyses are described
in the following sections. Descriptive statistics
Arguments from all the traditions discussed can be found in Section B of the ASR Online
earlierlink expansionof secondaryeducationto Supplement (http://www2.asanet.org/
growth of tertiaryenrollments.Most common journals/asr/2005/toc048.html).
is an obvious functional demographic argu-
ment: secondary graduates are a requisite of
DEPENDENT VARIABLE
tertiaryexpansion, and havingmore of them is
likely to increase tertiaryenrollments. HIGHER-EDUCATION ENROLLMENTS PER10,000
Froman institutionalpointof view,the expan- CAPITA. UNESCOhas collected enrollmentdata
sion of secondaryeducationis deeplyenmeshed based on country reports since the 1950s
with tertiaryexpansion,though more as a spu- (UNESCO2000, 2004). The criticalvariablefor
rious correlatethan a direct causal effect. The our purposes is the numberof tertiarystudents
global models and discourses that supportter- within a nation which, accordingto UNESCO,
tiary expansion also affect secondary educa- approximatesthe students falling into ISCED
tion (Chabbott2002; Meyeret al. 1977, 1992b). (International Standard Classification of
Finally, competition and conflict theorists Education)categories of 5 and 6 (i.e., exclud-
point out that the growth of secondary educa- ing post-secondary vocational/technical pro-
tion reflects increased status competition and grams). Banks (2001), working from similar
leads to credential inflation. Once secondary sorts of nationaldata,collected enrollmentdata
educationis greatlyexpanded,it becomes nec- fromthe latenineteenthcenturyto around1980.
essary to obtain a tertiarydegree to maintain The Banks codebook states that "every effort
906 AMERICAN SOCIOLOGICALREVIEW
Eastern-bloc socialist societies from 1970 to reporting,and so the curves start at different
1990 to addresshistoricallyspecific arguments points in time. The strikingfeatureof the results
that those countries limited the expansion of is how similar the growth curves are between
higher education (see earlierdiscussion). these dramaticallydifferentsorts of countries.
Enrollmentgrowthacceleratesin every type of
X POST-1990 DUMMY.
FORMERCOMMUNIST country after about 1960. Sub-SaharanAfrica
The measure is coded 1 for former socialist lags substantially throughout the period, but
societies in the period after 1990. With the even therewe see thatgrowthratesin the region
demise of the Soviet Union, we expect former are high in proportionalterms. We also find
constraints on higher education to disappear, extremely similar patterns when nations are
yielding rapidgrowth in post-Soviet societies. broken out by other criteria, such as level of
development, and in plots of individual coun-
RESULTS tries.
Figure 4 presents coefficients of variation
THETRENDS for higher education enrollments by region
We begin with a descriptivelook at enrollment between 1900 and 2000. The case base grows
growthoverthe twentiethcentury,to explorethe and becomes more diverse over time as new
extentto which enrollmentexpansionreflects a countriesbecome independent,andso in factthe
common "global" process, or a process that extent of convergence is understated.
involves merely certain subsets of nations. Nevertheless,we see thatvariationamongcoun-
Figure3 presentsenrollmentsper capitabroken triesdecreasesovertime withineach region,and
out by worldregion.We includeonly cases with in the world as a whole. Whereas one might
data at most points throughoutthe period of expect expansion to be associated with diver-
350
300
250
200
150
100
50
01 --------
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Year
-- IndustrializedWest --- EasternEurope - - Central/SouthAmerica
-- Asia MiddleEast/NorthAfrica --o--- Sub-SaharanAfrica
.
2.5-
0 1.5
0.5
0
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Year
-- World West
--- Industrialized -o- EasternEurope
America ------ Asia
-+- Central/South -- MiddleEast/North
Africa
-o-- Sub-Saharan
Africa
gence and inequality,we insteadsee increasing These figures paint a picture of universal
isomorphismover time. expansionand increasingglobal isomorphism.
We also explored gender differences in par- The evidence is broadly consistent with our
ticipation by looking at disaggregatedtertiary global institutionalargument,and casts doubt
enrollment ratios, which are available in the upon perspectives that predict divergence or
postwarera. Figure 5 presentsthe global aver- sharpregional differences.
age of enrollmentratiosfor independentnations
from 1950 to 2000. At the start of the period, REGRESSION
MODELS
higher education enrollmentratios were quite
low: just over 2 percent for men and less than Tables 1 and 2 presentresultsof randomeffects
1 percent for women in the average country. GLS regression models with robust standard
The initial expansion disproportionately errors predicting national higher education
involved men, increasingthe gender gap from enrollments per capita. Table 1 examines our
about 1.5 points to over 3 points. In raw terms, main global-level argument, along with key
however,the gap begins to shrinkstartingafter national-levelcontrol variables.Models 1 to 4
1970, and parity is achieved around 1990. By introduce our four indicators of the changed
2000 the average female enrollment ratio is model of society in which higher education
higherthanthe male averageby 6 points, a dif- could be viewed as universallyvaluablefor indi-
ference of almost 25 percent. It is noteworthy viduals and societies (Hypothesis4). In Model
that the "new gender gap,"which has recently 1, we see that global democratization has a
been observedin studiesof industrializedcoun- strong positive and significant effect on high-
tries, is also evident in our global averages. er-education enrollment. Education expands
The overalltrend,however,is similarfor men faster in time periods when democracyis more
and women. Enrollmentratios grow by more prevalentin the world. Model 2 shows a posi-
than an order of magnitude over the period, tive and significant effect of nationaldevelop-
dwarfingthe between-genderdifferences. ment planning. As nations mobilize around
91o AMERICAN
SOCIOLOGICAL
REVIEW
30
25
20
ES15
10
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Year
I- Male-*- Female
Figure 5. Higher-Education
EnrollmentRatioby Gender,GlobalAverageof IndependentNations, 1950-2000.
modern conceptions of national development, Table I also includes a variable for INGO
enrollments grow faster. In Model 3, we see a membership(logged), the standardmeasureof
positive, significant effect of the global expan- national linkage to global organizations and
sion of science. Finally,the intensifiedstructure discourses. INGO membershiphas a positive
of the world polity, measuredby the prolifera- and significant effect on tertiaryenrollments.
tion of internationalorganizations,has a posi- Consistent with the prior neo-institutionallit-
tive and significant effect on tertiary
erature,societies most deeply embeddedin the
enrollments.
Model 5 includes an index thatcombines the organizationalstructuresof world society are
historicalvariablesfromthe priorfourmodels: influenced the most. Hypothesis 5 is support-
ed.
democratization,scientization,the rise of devel-
In standardized terms, the two main neo-
opment planning, and the increasedstructura-
tion of the world polity itself. Like each institutional variables are the largest statisti-
component,the overall index has a strongpos- cally significant coefficients in Model 5.l1 Our
itive and significant effect on higher-education world-level index (of democratization,scienti-
enrollments, consistent with our central argu- zation, etc) and the INGOlinkagevariablehave
ment (Hypothesis 4). As these global trends a greater impact than conventionalpredictors
emergeanda new worldviewbecomestakenfor
granted-a processthatacceleratesin the 1950s,
1960s, and 1970s-nations and individualsini- such as a linear"time"variable(see section C of the
tiate a massive shift towardgreaterenrollments ASROnlineSupplement:http://www2.asanet.org/
in tertiaryeducation.9 journals/asr/2005/tocO48.html).
10STATA8.0 cannotproducestandardized
coef-
ficientsin regressionmodelswith robuststandard
errors.Commentsaboutstandardized coefficients
9 The "worldindex" variableremains significant werethereforebasedon modelswithordinarystan-
when comparedagainst other baseline hypotheses, darderrors.
EXPANSION
WORLDWIDE OFHIGHEREDUCATION 91i
(1.46)
Structurationof the world polity - 9.47***
(1.79)
Indexof world-levelvariables 3.25***
(.61)
National-LevelVariables
INGOmemberships(log) 5.37*** 4.34** 4.10** 3.92** 3.76**
(1.47) (1.48) (1.52) (1.52) (1.50)
Secondaryenrollmentper 10,000capita .06*** .07*** .07*** .07*** .06***
(.01) (.01) (.01) (.01) (.01)
Economicdevelopment 3.14*** 3.00*** 3.03*** 3.08*** 3.21***
(.94) (.93) (.93) (.93) (.93)
Ethno-linguisticfractionalization -. 15* -. 15* -. 14* -. 15* -. 15*
(.06) (.06) (.06) (.06) (.06)
Constant -5.17 3.54 -14.83 -57.47** 22.89*
(8.16) (8.96) (8.29) (11.85) (10.26)
Observations 687 687 687 687 687
AdjustedR2 .82 .81 .81 .81 .81
Note: Robuststandarderrorsappearin parentheses.GLS = generalizedleast squares;INGO= international
nongovernmental organization.
* p < .05; ** p < .01; *** p < .001 (one-tailfordirectionalhypothesesnotedabove;otherwisetwo-tailtest).
such as secondary education and economic We also see positive and significanteffects of
development. economic developmentrunningacrossthe mod-
Models in Table 1 also include a measureof els in Table 1 (butnote contraryresultsin Table
secondary education expansion, which has a 3). Findingsin TableI areconsistentwith a vari-
positiveandhighly significanteffect on tertiary ety of functionalarguments(discussed earlier),
enrollments.Hypothesis7 is supported.As dis- rangingfrom classical modernizationtheory to
cussed earlier, several argumentspredict this narroweconomic argumentsaboutcosts. We do
effect, includingdemographicarguments,com- not have leverage to adjudicateamong them.
petition/credentialinflationarguments,andthe Evidencefromlongitudinalstudiesof core coun-
neo-institutionalpredictionthatsecondaryedu- tries shows little effect of societal industrializa-
cationwill expandalongwith tertiaryexpansion tion or other structural economic changes
due to global culturalmodels that stress edu-
cation. We cannot directly adjudicate among
these claims with the availabledata.11 secondary education is excluded from the model.
Secondary enrollments may reflect national sub-
scriptionto world pro-educationalmodels, and thus
be a direct consequence of the "worldindex."If so,
11 In exploratoryanalyseswe noted thatthe effect then the positive secondary-enrollmenteffect partly
of the "world index" is substantially larger when reflects a neo-institutionaldynamic.
912 AMERICAN
SOCIOLOGICAL
REVIEW
into the sample, which are the most ethnical- andothers.The only limitationof the 1970-2000
ly diverse countries in the world. In the early analyses is that we cannot effectively examine
part of the century, where the sample is dom- the "worldindex"and relatedglobal trends, as
inatedby Europeand LatinAmerica,the effect they are severely truncated.Most of the varia-
is generally positive and occasionally borders tion in those variablesoccurspriorto 1970, and
on significance. thus we do not include them in the post-1970
The effect of economic development is analyses.
weaker in the early period, remainingpositive Table 3 presents results from the contem-
but falling far short of statistical significance. porary period with improved measures. Our
This may be due to a numberof wealthy coun- basic findings regarding INGO linkage and
tries (e.g., Germany)maintainingstronglimits secondary enrollmentsremain stable through-
on tertiary expansion. out Table 3. Ethno-linguistic fractionalization
maintains a negative sign that is typically sig-
REGRESSION MODEL
DIAGNOSTICS, nificant (or nearly so). The notable difference
ANDALTERNATE
SPECIFICATION, MEASURES is a weakening of the economic development
effect, which is not significant in any mod-
Furtheranalyses examined sources of error, els.16 We do find an economic development
bias, and potential alternative explanations. effect when secondary enrollment is exclud-
Regression diagnostics were generally unre- ed from the model, suggesting the lattermedi-
markable,except as noted. We exploreda wide ates the former.There is no effect of GDP per
range of model specifications and additional capita, however, over and above the effect of
control variables, including: primary enroll-
secondary enrollment.
ment ratios, population, population growth, Table 3 explores a varietyof additionalvari-
political autocracy,colonial history and dom- ables. We first examine domestic economic
inant colonial power, national religion and
inequality, measured by Gini coefficients.
religious composition, civil war, the World Inequality has a negative but nonsignificant
Wars, the Great Depression, international effect on subsequenttertiary-enrollment expan-
trade, world system position, OECD sion. We also look at two measures of global-
(Organizationfor Economic Cooperationand ization: trade openness and foreign direct
Development) membership, and others. The investment (FDI). Both have small positive
additional analyses are not included in Tables effects, the latterof which is statisticallysignifi-
I and 2 either because effects were not statis- cant. It may be the case that FDI brings labor
tically significant or because they signifi- force opportunitiesthat provide incentives for
cantly reduced the sample size due to missing increasedparticipationin highereducation.
data. In any case, none of these variables
Finally, we examine two measures of tech-
altered our main findings.
nological innovation and the scientific labor
As a furthercheck on our results, we con- force: patents per million capita (log), and the
ductedanalyses from 1970 to 2000, when high- numberof scientists in R&D per million capi-
er quality data was available.15For instance, ta (log). Neither variable is associated with
we were able to use gross tertiaryenrollment
higher education expansion, casting doubt on
ratiosratherthanper capitameasures,and GDP fashionable argumentsabout the "knowledge
as ourmeasureof economic developmentrather
society." This finding makes sense given that
than iron and steel production.In these analy- education expands very rapidly everywhere,
ses, we were able to examine a wider range of even in very poor nations that possess little in
independentvariables,includingforeign direct the way of high technology, innovation, or a
investment,Gini income inequalitycoefficients, knowledge-based economy.
gender enrollmentratios, measures of techno-
logical innovation(e.g., patents),R&D (research
& development) personnel and expenditure,
16Thedisappearanceof theGDPeffectis notdue
to therestrictedtimeperiodof the analysis.Rather,
it occursas we improveourmeasures-specifically,
'5Wepooledfive-yearpanelsto maintaina large whenwe useenrollment ratiosas thedependentvari-
samplesize. able.
EXPANSION
WORLDWIDE OFHIGHEREDUCATION 915