Economics of Education Review 29 (2010) 200–213

Contents lists available at ScienceDirect

Economics of Education Review

journal homepage: www.elsevier.com/locate/econedurev

Cultural capital and its effects on education outcomes夽

Lucia Tramonte ∗ , J. Douglas Willms
University of New Brunswick, Canada

a r t i c l e i n f o a b s t r a c t

Article history: In this study we distinguished between two forms of cultural capital, one that is static,
Received 12 June 2009 representing the highbrow activities and practices of parents, and one that is relational,
Accepted 12 June 2009
representing cultural interactions and communication between children and their parents.
We used data for 28 countries from the 2000 Programme for International Student Assess-
Keywords: ment to examine whether these two types of cultural capital were associated with students’
Cultural capital
reading literacy, sense of belonging at school, and occupational aspirations, after control-
Human capital
ling for traditional measures of socioeconomic status. We examined whether one type of
Educational achievement
School segregation cultural capital had stronger effects than the other and whether their effects differed across
outcomes and across countries. The results provide compelling evidence that dynamic cul-
tural capital has strong effects on students’ schooling outcomes, while static cultural capital
has more modest effects.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction Economists have stressed the mediating role that

schools play on the path between family background and
Education is a key factor for predicting social mobil- student outcomes. The dominant approach has been to
ity in most industrialized societies (Blau & Duncan, 1967; search for aspects of classroom and school inputs that
Sewell & Hauser, 1975). Most studies of social mobility affect student outcomes after controlling for family inputs,
have found that academic achievement and occupational and to estimate their effects using education ‘produc-
attainment are largely determined by people’s family ori- tion functions’ (Hanushek, 1989; Levin, 1974). Henry Levin
gin and educational experiences (Bielby, 1981; Kerckhoff, contributed to this work in two important ways. First,
1996; Sewell, Hauser, & Featherman, 1976). These studies he brought cost-benefit analysis to the production func-
have focused primarily on the roles of socioeconomic sta- tion enterprise, relating the costs of inputs to measured
tus (SES), family structure, and family resources, including gains in learning (Levin, 1989). Second, he questioned the
economic, cultural and social capital (Lareau & Weininger, conventional wisdom that reallocating inputs based on pro-
2003; Sirin, 2005). Dominant status groups hold economic, duction functions would increase the efficiency of schools;
political, and symbolic power (Collins, 1971), and their suc- rather he argued that dramatic and sustainable change
cess depends on the use of their social and cultural capital required dramatic changes in the structural and organisa-
in strategic ways. tional features of schools that would bring disadvantaged
students into the mainstream, including incentives for
teachers, efficient access to information, the use of produc-
夽 Note: The authors are appreciative of funding from the Social Sci- tive technology, and better assessment strategies (Levin,
ences and Humanities Research Council for its support of the collaborative 1997).
research program, Raising and Levelling the Bar in Children’s Cognitive, Sociological explanations of the association between
Behavioural and Health Outcomes (Grant Number 512-2003-1016). children’s educational outcomes and family background
∗ Corresponding author at: Canadian Research Institute for Social Policy
have often referred to economic, social, and cultural capital
Suite 300 Keirstead Hall, Fredericton, New Brunswick, Canada E3B 5A3.
Tel.: +1 506 458 7012. as the main components of parental resources (De Graaf, De
E-mail address: [email protected] (L. Tramonte). Graaf, & Kraaykamp, 2000). Current research is concerned

0272-7757/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.econedurev.2009.06.003
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 201

with three main aspects of this process: differences among This study uses data from the 2000 Programme for Inter-
families in the magnitude of their investment in their chil- national Student Assessment (PISA), a large comparative
dren’s human capital; social reproduction via differential study conducted in 28 Organisation for Economic Cooper-
family access to the cultural capital required for the child to ation and Development (OECD) countries, and in a number
succeed in school and for parents to mediate and intercede of non-OECD countries, under the aegis of the OECD.1 We
on their child’s behalf; and social reproduction associ- attempt to address two sets of questions relevant to the
ated with families’ access to the social capital that enables relationship between schooling outcomes and cultural cap-
children to succeed in school (Farkas, 2003). The human ital:
capital investment paradigm (Becker, 1964; Mincer, 1958,
1974; Schultz, 1960, 1981) has provided durable support
(1) Do the two forms of cultural capital affect reading lit-
that education and training are useful assets in produc-
eracy, sense of belonging at school, and occupational
tion (Farkas, 2003). Economists have recently included a
aspirations, after controlling for other family-related
new focus on the effects of schools and intervention pro-
determinants of these social outcomes? If so, does the
grams on student values, habits, and behaviours (Bowles
strength of the relationships, or their pattern, differ
& Gintis, 2000, 2002; Dunifon, Duncan, & Brooks-Gunn,
across outcomes?
2001; Heckman & Lochner, 2000; Heckman & Rubinstein,
(2) Do static and relational cultural capital operate dif-
2001). Yet this paradigm, using the notion of families’
ferently in school systems that are highly selective
calibrated cost/benefit investments in their children’s edu-
compared with those that are less selective? If this is
cation, leaves unexplained why low income families are
the case, we expect that the two types of capital differ
generally unable to help children achieve school success
also in their effects on student performance within and
(Farkas, 2003).
between schools.
The research on cultural capital helps address this ques-
tion. It suggests that individuals possess different amounts
of cultural capital which explains why some students 1.1. Social mobility in schooling systems
meet school standards, are accepted at college, and finally
achieve higher levels of education, and why other students Two theoretical frameworks have been used in this
do not (Swidler, 1986; Lareau, 1989). Schools promote par- work to explain how social inequalities are embedded in
ticular linguistic structures, authority patterns, and types schooling systems: the distinction between sponsored and
of curricula. Children from higher SES families are already contest mobility by Turner (1960), and Bourdieu’s (1977,
familiar with these social arrangements when they enter 1984, 1986) concepts of distinction and social reproduction.
school, and therefore they do not perceive school as an Turner (1960) argued that societies emphasize social
intimidating place. Their ongoing experiences at home help mobility to varying degrees, and that there are con-
them adapt to school and maintain the pursuit of academic stant strains shaping the educational system to conform
achievement (Lamont & Lareau, 1988). Elements of family to the established norm.2 However, students’ selection
life, especially cultural resources, are invested as capital to into secondary school is a powerful mechanism shaping
align students’ expectations with school norms and help educational systems, and a key element defining contest
solve problems concerning social acceptance (Lamont & and sponsored mobility (Turner, 1960; Kerckhoff, 1975;
Lareau, 1988; Lareau, 1989). The argument is that low Kinloch, 1969). Early selection is therefore a crucial element
income parents fail to support their children in succeed- of our analysis.
ing in school not because they see too low a payoff to such
action, but because they lack the skills, habits, and knowl-
edge needed to effectively assist them (Swidler, 1986; 1
The Programme for International Student Assessment (PISA) is an
Farkas, 1996). The human capital approach and the research internationally standardised assessment developed jointly by OECD
on cultural capital share a common core: the application of member countries through the OECD’s Directorate for Education. It is
administered to 15-year olds in schools. The survey was implemented
resources to building skills and habits in children by parents
in 43 countries in the 1st assessment in 2000 and 2002, in 41 countries
and other adults in the extended family and neighbourhood in the 2nd assessment in 2003, in 57 countries in the 3rd assessment in
(Farkas, 1996, 2003). 2006, and in 62 countries in the 4th assessment in 2009. Tests are typically
Our work attempts to describe and explain variabil- administered to between 4500 and 10,000 students in each country.
2
ity in national patterns of the influence of family cultural Turner’s main assumption is that within a formally open class sys-
tem, the organizing folk norm defining the accepted mode of upward
resources on school achievement. Our operational defini- mobility shapes the school system, and may be even more crucial than the
tions of cultural resources distinguish between two types of extent of upward mobility.” (Turner, 1960, p. 856) The folk norms to which
cultural capital: one that is static, which includes ‘highbrow’ Turner refers to are contest mobility and sponsored mobility. In contest
activities and practices; another that is relational, which is mobility systems, elite status is the prize in an open contest, which is won
through students’ own efforts. The prize of successful upward mobility is
concerned with the cultural interactions and communica-
not awarded by the established elite. In sponsored mobility systems, the
tion between children and their parents. Conceptually, we established elite chooses ‘recruits’ it wishes to be admitted to its ranks, and
consider the unit of analysis to differ for these two forms of carefully inducts them into the norms of elite status. The early selection
capital: static cultural capital is an expression of the family’s of students plays a dominant role in this process, but sponsorship systems
socioeconomic advantage, while relational cultural capital also differ from contest systems in their curricula, systems of examina-
tions, and the processes governing access to further education. Although
embodies the resources and experiences of children that the English school system has often been described as a sponsored system
they can use in society to interact strategically and suc- and the US as a contest system, all systems have structural features that
cessfully in achieving their goals. can be considered sponsored or contest.
202 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

In contrast, Bourdieu’s concept of social reproduction very close to that of human capital in economics (Robbins,
focuses on the differential socialization of individuals 1991; Throsby, 1999). Sometimes definitions of human cap-
coming from different social classes. This socialization pro- ital within economics explicitly include culture as one of
vides children with a sense of what is comfortable or its components (Throsby, 1999; Costanza & Daly, 1992;
natural—what Bourdieu calls ‘habitus’. The selection mech- Chiswick, 1983); however, Woodbury (1993) has pointed
anisms governing sponsored mobility are similar to those out how building culture into the human capital frame-
supporting social reproduction. In both cases cultural and work empties the theory of empirical content because no
social resources are the necessary ‘passwords’ to succeed independent assessment of ‘culture’ is possible.
in the selection process for elite status. The essence of cul- Bourdieu’s conceptual arguments do not provide a clear
tural capital is that its effects are institutionalized (Lamont direction for proceeding with empirical tests, and conse-
& Lareau, 1988; De Graaf et al., 2000): schools are places quently have been only weakly linked to specific empirical
where codes from higher socio-economic status groups are analyses.3 The research evidence concerning the definition
recognized and where the possession of cultural capital is and operationalization of cultural resources is mixed. A
rewarded. High income parents may have highbrow skills glaring problem is that some studies assess the possession
and habits that are of little productive value but provide of cultural attributes and resources, particularly ‘highbrow’
their children with signals of high cultural capital for their preferences, tastes, and attitudes, while others do not dis-
teachers (Farkas, 2003). tinguish cultural capital from the common measures of SES.
Our main thesis is that static cultural capital, as a marker Also, most studies do not distinguish between the cultural
of social status, is more important in sponsored school sys- resources of the parents from those of their children, treat-
tems, where established elites control the enrolment of ing cultural capital as a family characteristic.
their children in high-status schools. Their children’s ability To explain educational reproduction, we need to con-
and effort play a lesser role in school success. In contrast, sider two important mechanisms of exchange: one is the
we expect relational cultural capital to be more important in effect of students’ cultural resources on their educational
contest systems, as it is the cultural interactions and strate- attainment; the second is the transmission of cultural
gic communication between children and their parents that resources from parents to children. The latter process is
gives some children an upper hand in the different types of somewhat complicated: while one may easily agree that
contests they encounter as they proceed through school. elements of cultural lifestyle are transmitted from one
Generally, selective school systems tend to have greater generation to the next, there is an important difference
variance between schools in their academic outcomes than between the possession and the activation of cultural
within them, while in contest systems much of the vari- resources (Lareau & Horvat, 1999). It is possible to trans-
ance in academic outcomes is within schools. If our thesis mit cultural resources, but the beneficiaries have to activate
holds, static cultural capital will be more important in them.4
sponsored systems than in contest systems as it helps chil-
dren achieve success in early recruitment, but thereafter,
the skills that are transmitted through relational cultural
3
Most of the fixed characteristics associated with the construct of
capital will contribute to their success within schools. In
cultural capital come from two of Bourdieu’s works, “Reproduction” and
contest systems, however, relational cultural capital will “Distinction” (Bourdieu, 1977, 1984, 1986). The interpretation of Bourdieu’s
play a dominant role in both school selection and in chil- concept and the attempts to operationalize it have a long history. DiMaggio
dren’s success within schools. Our analysis includes three (1982) considered cultural capital as a factor capable of completely filling
different types of schooling outcomes, but we do not have out models of the ‘status achievement process’. He observed an association
between cultural capital and students’ grades; cultural capital acted as a
a strong thesis on whether static versus relational cultural resource that supplemented ability enabling students to succeed in school.
capital has relatively stronger effects on a particular type The argument is that teachers communicate more easily with students
of outcome. who participate in elite status cultures, and perceive them as more gifted
Our analysis is not at odds with Bourdieu’s concept of than students who lack the distinctive tastes, traits, and styles. Draw-
ing from DiMaggio’s (1982) work, most of the subsequent Anglophone
cultural capital. Rather, we feel that a distinction between
research maintained two intrinsically related conceptions of cultural cap-
the parents’ possession of ‘highbrow’ static cultural capital ital: it comprises prestigious, highbrow activities and it is distinguished
and the activation of cultural capital through their relation- conceptually and causally from the effect of ability (Lareau & Weininger,
ships with their children can further our understanding of 2003). DiMaggio’s (1982) definition of cultural capital is rather restrictive,
how cultural capital operates in different types of school and therefore researchers who adopted his suggestions found it difficult
to explain the mechanism of educational reproduction.
systems. This observation can potentially direct debates 4
Recent studies have considered parental and students’ cultural inter-
about educational inequality towards discussions of equity. ests, including engagement in reading practices, as forms of cultural
capital. For example, De Graaf et al. (2000) considered parents’ cultural
1.2. Cultural capital and human capital resources, but distinguished between participation in the ‘beaux-arts’
and engagement in reading. Their research in the Netherlands found
that parents’ mastery of highbrow cultural codes brought less advan-
Researchers in the fields of education and sociology tage to students’ educational careers than parents’ reading behaviour. The
have attempted to define the concept of cultural capital in same distinction between beaux-arts participation and parents’ verbal
ways that are consistent with some of the definitions pro- and reading ability was used by Crook (1997) in a study of the effects
posed by Bourdieu. Cultural capital exists in three forms: of parental cultural practices in Australia. He also found parents’ reading
behaviours to have stronger effects on student outcomes than partici-
in embodied state, in objectified state, and in an institu- pation in the beaux-arts. Their operationalization of cultural capital is
tionalized state (Bourdieu, 1986). As a concept developed consistent with DiMaggio (1982) in that participation in elite activities
and conceived in an individualistic form, cultural capital is is separate from ability. Sullivan (2001) proposed a much broader defi-
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 203

1.3. Static versus relational cultural capital is related to several schooling outcomes, including aca-
demic achievement (Finn & Rock, 1997) and the successful
Our analysis examines the effects of two types of transition from middle to secondary school (Newman,
cultural capital. Static cultural capital includes both the Lohman, Newman, Myers, & Smith, 2000). Willms (2003a)
possession of high culture goods, such as art works, musical argued that sense of belonging at school and engagement
instruments, and classical music; and highbrow activities, should be considered important outcomes in their own
such as going to museums or the ballet or theatre. Parents right, as they represent a disposition towards learning,
hold and manage ‘static’ cultural resources, which they working with others, and functioning in a social institution.
share in the household with their children. The ‘static’ We consider student aspirations important also, as stu-
attribution indicates that it is relatively constant, perhaps dents’ interpersonal relationships affect their educational
even more so than income or level of education. The and occupational aspirations.
cultural goods and activities are available to children, Early studies on status attainment focused on the
but they are not necessarily an expression of individual importance of ‘significant others’ – peers, parents, and
choices, tastes, and preferences.5 teachers – in mediating the effects of socioeconomic
Relational cultural capital includes cultural resources and background and ability on aspirations and educational
activities that are expressed in the relationships between and occupational attainment (Duncan, Haller, & Portes,
parents and children. It includes discussions between chil- 1968; Haller & Butterworth, 1960; Sewell, Haller, & Portes,
dren and parents on cultural, political, and social matters, 1969; Sewell & Shah, 1968). Although peers and parents
on school activities, and on books that children have read. undoubtedly play a strong role in shaping educational and
This form of cultural capital is relational because it requires occupational aspirations, we maintain that schools also
an investment of parents and children in an on-going rela- play a prominent role.
tionship that transcends the economic capital available to
them. The ideas of activation, engagement, and promo- 2. Materials and methods
tion of cultural interests are embedded in this construct.
2.1. Data and sample
We consider the distinction between static and relational
cultural capital important because static cultural capital
The data for this study are a subset of the 2000 PISA,
may only reflect parents’ own choices and lifestyles, while
a reading literacy skills assessment conducted with
relational cultural capital reflects how that capital is trans-
224,058 students nested in 8364 schools from 28 OECD
mitted and employed.
countries (Organisation for Economic Cooperation and
The PISA 2000 study is the first large-scale international
Development, 2001, 2003). We limited our study to OECD
study to explicitly attempt to include indicators of cultural
countries.6 The target population for the PISA survey was
capital, and the student questionnaire includes questions
all 15-year-old students in each country. The data were
that can be roughly categorized into static and relational
derived from a comprehensive test of reading literacy
cultural capital. However, the study was not designed with
and a student background questionnaire covering family
our schema in mind and therefore the classification of items
structure; education and occupation of the mother and
as static and relational is far from perfect.
father; various measures of the attitudes, habits and
Our analysis examines the effect of these two forms of
expectations of students; and the relationships of students
cultural capital on three social outcomes: reading literacy,
with their peers, parents, and teachers. The student data
sense of belonging at school, and occupational aspirations.
set provides measures of our three outcome measures.
Each of these outcomes is influenced by practices at home,
We constructed measures of static and relational capital,
but also by the school environment and teaching practices.
family structure, and SES. Our analysis does not include
Willms’s (2003a) analysis of reading literacy and sense
school-level measures available from a questionnaire
of belonging at school found that these outcomes vary
administered to school administrators, but it does take
considerably among countries and among schools within
into account the hierarchical structure of the data set, with
countries. Moreover, their relationships with a traditional
pupils nested within schools, and schools within countries.
measure of SES varied considerably among countries.
The variables used in the analysis are described below.
Sense of belonging at school is considered a key compo-
nent of the broader construct, student engagement, which
2.2. Measures

2.2.1. Reading literacy

nition of cultural capital. It distinguished parents’ cultural capital from
children’s cultural capital, and the conception of parents’ cultural cap-
PISA defines reading literacy as the ability to under-
ital comprised both reading behaviours and highbrow possessions and stand, use, and reflect on written texts in order to
activities. Similarly, her indicators of children’s cultural capital included participate effectively in everyday life. The measure of read-
information about pupils’ activities, cultural knowledge, and vocabulary. ing literacy is based on students’ performance on a wide
She found that cultural capital had a large, direct effect on children’s edu-
range of literacy tasks with varying difficulty. The test items
cational attainment in the UK, but only partially explained social class
inequalities in attainment.
5
For example, going to a ballet performance could be considered rela-
6
tional in that parents are making an effort to transmit their cultural capital. The non-OECD countries tend to have lower levels of economic devel-
It is a question of whether this is more of an outward expression of parents’ opment and their schooling systems tend to be highly segregated because
possession of cultural capital versus an attempt to inculcate an apprecia- of the differential allocation of low and high SES students into rural public,
tion of the arts in their children. urban public, and private schools (Willms, 2006).
204 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

were scaled using techniques from item response theory, within countries and among countries. For a discussion on
and standardized to have a mean of 500, and a standard the importance of considering reliability in a multilevel
deviation of 100 for all OECD countries. framework, see Rowan, Raudenbush, and Kang (1991).

2.2.2. Sense of belonging at school 2.2.8. Relational cultural capital

Sense of belonging at school is a composite index gen- A standardised index, with a mean of zero and a standard
erated by a set of questions related pupils’ perceptions and deviation of one for all OECD countries, was constructed
feelings about their presence at school (Willms, 2003a). from responses to questions related to cultural exchanges,
This measure was also standardized to have a mean of 500 communication, and elaboration of cultural experiences.8
and a standard deviation of 100 for all OECD countries. The estimated reliability coefficients are 0.68, 0.65,
and 0.99 at the individual, school and country levels
2.2.3. Occupational aspirations respectively. These results suggest that the measure can
Students were asked, “What kind of job do you expect powerfully distinguish among countries in their average
to have when you are about 30 years old?” Responses were scores, but is less reliable in distinguishing among schools
coded and scaled on the international occupational index within countries or among individuals. We conclude that
developed by Ganzeboom, De Graaf, and Trieman (1992). this measure reliably distinguishes among countries, which
As with reading literacy, the measure was scaled to have a is pertinent to one of the primary aims of our analyses.
mean of 500, and a standard deviation of 100 for all OECD However, the estimates of the slopes within countries are
countries. likely to be underestimated, given the lower reliability at
With the three outcome measures scaled in this way, the individual and school levels.
one can interpret estimates of the magnitude of observed
differences among countries and regression parameters as 2.2.9. Variation between-schools in literacy skills
‘effect sizes,’ as 100 points on each scale is equivalent to For each country we estimated the proportion of vari-
one standard deviation. ance in reading literacy that was between schools. This
is the ratio of the between-school variance to the total
2.2.4. Mothers’ and fathers’ level of education variance, called Rho, which was estimated with a ‘null’ hier-
Parents’ education was measured by students’ reports of archical linear model (Raudenbush & Bryk, 2002). We treat
the levels of education achieved by the father and mother, this as proxy for the cumulative effects of the selection
based on the 1997 International Standard Classification of of students into different types of schools. In many OECD
Education (ISCED). These were recoded into years of edu- countries students are selected as early as age 10 or 11,
cation. while in others they are selected during the transition from
elementary to secondary school (Willms, Smith, Zhang,
2.2.5. Parental occupation & Tramonte, 2006). In many school systems there is also
Parental occupational prestige was measured by the
selection into a private sector, which generally increases
International Socio-Economic Index of Occupational Sta-
between-school variance.
tus (ISEI), which gauges the attributes of occupation that
We include Rho in our final model as a proxy for
convert a person’s education into income.
the selectivity of the school system. Ideally one would
use external indicators of this construct. The principal
2.2.6. Sex questionnaire of the PISA study includes some measures
A dichotomous variable “female” was created with relevant to school selection, but these pertain mainly to the
females coded one and males zero. selection of students into schools at the secondary level
rather than the cumulative selection of students into dif-
2.2.7. Static cultural capital ferent types of schools during their school careers. Also,
A standardised index, with a mean of zero and a standard one does not usually like to include a covariate that is
deviation of one for all OECD countries, was constructed derived from the outcome measure. However, in this case
from students’ responses to questions related to cultural the measure is at the country level, far removed from
consumptions and possessions.7 the student-level outcomes. Moreover, measures of Rho
The reliability of this measure at the individual level, derived from either mathematics or science performance
using Cronbach’s alpha, is 0.68. At the school and country are highly correlated with the measure based on reading
levels, the reliability is 0.84 and 0.95 respectively. The latter performance.
estimates were obtained based on a three-level ‘null’ hier-
archical linear model, with students nested within schools, 2.3. Statistical methods
and schools nested within countries. The results reveal
that the measure adequately distinguishes among schools The complex structure of the PISA data set requires
consideration of two important features. The PISA study

7
The following items comprise the construct of Static Cultural Capital:
8
how many books there are in your home? How often have you visited The following items comprise a measure of Relational Cultural Capital:
museums or art galleries? How often have you attended opera, ballet, and do your parents discuss political or social issues with you? Do your parents
classic symphony? How often have you watched live theatre? How often discuss books, films, or television programs with you? Do your parents
do your parents listen to classical music with you? Do you have musical discuss how well you are doing at school? Do your parents spend time
instruments at home? Do you have classic literature at home? Do you have just talking with you? Do you like talking about books with other people?
books of poetry at home? Do you have works of art at home? Do you enjoy going to a bookstore or a library?
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 205

employed a matrix sampling design whereby each stu- ˇ00k = 000 + u00k
dent was given only a sample of the items from the full ˇ10k = 100 + u10k
reading test. Each student was assigned five ‘plausible val- ˇ20k = 200 + u20k
(2c)
ues,’ estimates of what they might have achieved had they ˇ30k = 300 + u30k
completed the full test. Also, because the sampling design ˇ40k = 400 + u40k
was a two-stage clustered design, the data base includes a ˇ50k = 500 + u50k
set of sampling weights derived from a technique called
Balanced Repeated Replicates. Correct estimates of the At level 1 (Eq. (2a)) the reading score for the ith stu-
standard errors of statistical estimates (e.g., means, stan- dent in the jth school and the kth country is modelled as a
dard deviations, regression coefficients) can be obtained by function of student-level covariates and a random student-
using the set of plausible values in concert with the repli- level error term. This is identical to the base model (Eq. (1))
cate weights. The hierarchical structure of the data can also except that the model is conceived to have different param-
be taken into account by fitting a hierarchical linear model eters for each school and country. At level 2 (Eq. (2b)) the
to the data (Raudenbush & Bryk, 2002), using each of the intercepts and regression coefficients at level 1 are mod-
five plausible values (see Willms & Smith, 2005). elled as random effects; that is, they are allowed to vary
For each of the three outcome measures, we fit a among schools and among countries. The level 3 model
basic regression model separately for each country, using (Eq. (2c)) allows the parameters for each school to vary
ordinary least-squares (OLS) estimation and the PV-BRR among countries. These are described as a grand mean plus
technique. The base model is: a random term that varies around the grand mean.
In essence, the model fits a separate regression model for
every school within every country. In our exploratory work
Readingi = ˇ0 + ˇ1 (Relational Capital)i we fit separate two-level models, one for each country,
+ ˇ2 (Static Capital)i which yielded virtually identical results. However, we pre-
fer the three-level model as HLM provides an explicit test as
+ ˇ3 (Parental Occupation)i to whether the variation in the coefficients at levels 2 and
3 are statistically significant. For example, we can ask, “Do
+ ˇ4 (Parental Education)i
the effects of relational and static capital on reading (after
+ ˇ5 (Female)i + εi (1) controlling for other variables in the model) vary among
schools within countries?” The analysis provides estimates
of the average effects of relational and static capital (ˇ10k
The model also includes two dummy variables indicat-
and ˇ10k , respectively) which we expect to be positive and
ing missing data for parental occupation and education. The
greater than the effects associated with parental occupa-
same model was fit for sense of belonging at school and
tion and education. The analysis also provides estimates of
occupational aspirations. The first question is addressed by
the variance in r1jk and r2jk and chi-square tests are used to
examining the strength and pattern of the coefficients, ˇ1
discern whether each variance is greater than zero. We also
and ˇ2 , across countries.
ask, “Do the effects of relational capital and static capital
To test the hypotheses concerning variation among
vary among countries?” The analysis yields estimates of the
schools and countries in these relationships we fit a three-
variance of u10k and u20k which are also tested with a chi-
level hierarchical linear model (HLM), with students nested
square test. For each of the three social outcomes, a general
within schools, and schools nested within countries. The
model was fitted and examined, and then the parameters
three-level hierarchical model is described with Eq. (2)
that did not vary among countries were fixed.
below using the notation set out by Raudenbush and Bryk
The third level of the model can be extended to include
(2002, see p. 231).
country-level variables. In our case, we added Rho, the
proportion of the variance in reading achievement that is
Readingijk = 0jk + 1jk (Relational Capital)jk between schools, to the first three equations in (2c):

+ 2jk (Static Capital)jk ˇ00k = 000 + 001 (Rho)k + u00k

ˇ10k = 100 + 101 (Rho)k + u10k (3)
+ 3jk (Parental Occupation)jk
ˇ20k = 200 + 201 (Rho)k + u20k
+ 4jk (Parental Education)jk
One of advantages of hierarchical modelling is that the
+ 5jk (Female)jk + eijk (2a) model can be manipulated to yield separate estimates of
the within- and between-school slopes associated with
relational and static cultural capital. The within-school
slopes can be estimated by centering the covariates at level
1 (Eq. (2a)) around their school means, rather than around
0jk = ˇ00k + r0jk the grand mean. After fitting this model, we examined
1jk = ˇ10k + r1jk graphically the bivariate relationship between the within-
2jk = ˇ20k + r2jk school slopes and Rho, and then extended the model to
(2b)
3jk = ˇ30k + r3jk include Rho at level 3, as per Eq. (3).
4jk = ˇ40k + r4jk The between-school slopes were estimated by dropping
5jk = ˇ50k + r5jk the covariates for relational and static cultural capital at
206 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

level 1 and adding the mean levels of these two variables examining these results in a multilevel model, it is clear
in the level 2 equation for the intercepts. This model asks that there are large and statistically significant differences
whether the average levels of school performance, adjusted among countries in reading performance, sense of belong-
for sex, parental education and parental occupation, are ing at school, and occupational aspirations. The mean
related to mean levels of relational and static cultural cap- scores for reading performance have been reported earlier
ital within the school. These school-level relationships are in the international report (OECD, 2001), and mean scores
allowed to vary among countries. Therefore, as with the for sense of belonging at school were reported by Willms
within-school slopes, one can examine graphically their (2003a).
relationship with Rho, and provide an explicit test of the
statistical significance of the observed relationships. This (1) Do the two forms of cultural capital affect reading lit-
model is described with Eq. (4): eracy, sense of belonging at school, and occupational
aspirations, after controlling for other family-related
Readingijk = 0jk + 3jk (Parental Occupation)jk
determinants of these social outcomes? If so, does the
+ 4jk (Parental Education)jk strength of the relationships, or their pattern, differ
across outcomes?
+ 5jk (Female)jk + eijk (4)
Table 2 displays the effects of static and relational cul-
tural capital on reading literacy, sense of belonging at
0jk = ˇ00k + ˇ01k (Dynamic Capital)·k school, and occupational aspirations, while controlling for
+ ˇ02k (Relational Capital)·k + r0jk gender, parental education, and parental occupation. Recall
3jk = ˇ30k + r3jk that the measures of relational and static cultural capital
4jk = ˇ40k + r4jk were standardized to OECD norms. Therefore the effects
5jk = ˇ50k + r5jk shown in Table 2 for these variables indicate the changes in
the outcome measures associated with one standard devi-
ˇ00k = 000 + 001 (Rho)k + u00k ation increases in either static or relational cultural capital.
ˇ01k = 010 + 011 (Rho)k + u01k Considering all OECD countries together, there is a sta-
ˇ02k = 020 + 021 (Rho)k + u02k tistically significant effect of both relational and static
ˇ30k = 300 + u30k capital for all three outcomes. The effect sizes are par-
ˇ40k = 400 + u40k ticularly large for relational and static cultural capitals
ˇ50k = 500 + u50k on reading, and for relational cultural capital on sense of
belonging at school. The other effects are of modest size,
We did not include measures of family structure or ranging from 1.4 to 5.6 points. These effects cannot be
immigrant status in our statistical models as we wanted compared directly with those of parental education and
estimates of the effects of cultural capital, overall, and occupation, as these effects indicate the change in the
within and between schools, unadjusted for family struc- outcome measure associated with one additional year of
ture or immigrant status. Thus, our goal was not in trying education and a one-point increase on the ISEI occupation
to maximize explained variance, but to understand how scale. The effect of parental education for reading is rel-
the relationships between student outcomes and the two atively small, but statistically significant. The effect is not
forms of cultural capital differ within and between schools statistically significant for sense of belonging at school and
and among countries. aspirations. This is noteworthy as it suggests that additional
controls using traditional measures of SES would unlikely
3. Results have a mediating effect on our measures of cultural capital.
The effect of parental occupation, net of other covariates in
Table 1 provides descriptive statistics of all variables the model, is very strong for reading literacy (18.7 points),
used in the analyses. As noted earlier, the three outcome but weaker for sense of belonging at school (3.7 point) and
measures were scaled to have a mean of 500 and a stan- occupational aspirations (9.7 points). The effects of parents’
dard deviation of 100 for all OECD countries.9 Even before occupation are statistically significant for all three outcome
measures. Females scored considerably higher than males
on the reading test—the difference is nearly 25 points. How-
9
One of the dilemmas in scaling data from international studies is ever, sex effects were not statistically significant for sense
whether to determine the international mean based of data weighted of belonging at school, and for occupational aspirations the
at the pupil-level, which is called ‘house weighting’, or weighted at the average score for females was about 5 points higher than
country level, with each country contributing equally, which is called
that of males.
‘senate weighting.’ For PISA, an international mean score based on house
weighting would be heavily dominated by the US and Mexico, which have If we again consider the average across all OECD coun-
large populations of 15-year-old students, while countries like Iceland and tries, the results indicate that the effect on reading literacy
Luxemburg would contribute relatively little to the weighted mean. In of relational cultural capital is similar to that of static
contrast, the senate weighted approach gives equal weight to small and cultural capital: 14.2 points compared with 13.9 points.
large countries alike, and so it too has its shortcomings. The PISA Technical
Advisory Group opted for the second approach, senate weighting, which
we have also used. Also, Netherlands was not used in calculating the OECD
mean as its sample did not meet PISA standards in 2000. We also excluded aspirations, the means are not exactly 500 because of missing data. These
Netherlands in our scaling of the data. In the case of sense of belonging and issues have very little impact on any of the international comparisons.
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 207

Table 1
Descriptive analysis of the variables included in the model.

Country Sample Size Reading literacy Sense of belonging Occupational Relational Static cultural Rho
aspirations cultural capital capital
Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.

Australia 5,176 528 102 495 97 489 95 −0.198 1.037 −0.055 0.987 0.183
Austria 4,745 507 93 526 109 482 98 −0.312 0.999 0.076 1.004 0.598
Belgium 6,670 507 107 479 90 494 102 −0.294 1.011 −0.382 1.043 0.597
Canada 29,687 534 95 512 110 503 88 −0.034 0.975 −0.018 0.989 0.176
Czech Rep 5,365 492 96 471 77 474 106 0.155 0.876 0.326 0.942 0.532
Denmark 4,235 497 98 513 104 564 106 0.175 0.937 −0.026 0.992 0.190
Finland 4,864 546 89 502 96 488 105 0.065 0.847 0.039 0.939 0.125
France 4,673 505 92 486 94 506 108 0.243 0.965 −0.345 0.983 0.502
Germany 5,073 484 111 518 107 488 110 −0.297 0.999 −0.003 1.009 0.592
Greece 4,672 474 97 498 95 493 86 0.330 0.883 0.114 0.859 0.509
Hungary 4,887 480 94 514 97 476 101 0.421 0.802 0.415 0.912 0.668
Iceland 3,372 507 92 514 109 506 102 −0.160 0.988 0.589 0.777 0.082
Ireland 3,854 527 94 508 101 483 93 −0.080 0.954 −0.056 0.966 0.177
Italy 4,984 487 91 500 92 489 89 0.538 0.871 0.294 0.896 0.548
Japan 5,256 522 86 465 89 529 108 0.011 1.148 −0.333 0.913 0.461
Korea 4,982 525 70 461 81 502 85 −0.464 1.127 0.013 0.866 0.375
Luxemburg 3,528 441 100 505 110 510 108 −0.332 1.019 −0.133 1.055 0.308
Mexico 4,600 422 86 509 98 523 75 0.018 0.927 −0.666 1.002 0.534
Netherlands 2,503 532 89 499 84 473 96 −0.177 0.955 −0.430 0.944 0.505
New Zealand 3,667 529 108 498 98 496 98 0.012 0.991 −0.129 0.982 0.162
Norway 4,147 505 104 512 104 492 101 −0.258 0.964 0.127 0.987 0.103
Poland 3,654 479 100 461 85 507 100 0.049 0.980 0.135 0.944 0.624
Portugal 4,585 470 97 501 88 501 86 0.278 0.882 −0.169 0.988 0.369
Spain 6,214 493 85 499 91 498 95 0.179 0.919 0.165 0.955 0.204
Sweden 4,416 516 92 527 103 489 99 −0.184 0.945 0.090 0.970 0.093
Switzerland 6,100 494 102 520 105 497 111 −0.179 1.028 0.015 1.012 0.432
United Kingdom 9,340 523 100 513 101 493 97 0.067 0.915 −0.053 1.064 0.217
United States 3,846 504 105 494 111 524 90 0.247 1.083 −0.067 1.055 0.294

Total 159,095 500 100 501.9 101.1 499.2 98.2 −0.007 0.99 −0.016 1.00 0.363

However, the effect of relational cultural capital is greater education. The superscripts, S and C, indicate that this
than that of static cultural capital for sense of belonging effect varies significantly among schools within countries
at school – 10.5 compared with 4.5 – and for occupational and among countries. Similarly, the effect attributable to
aspirations – 5.6 compared with 1.4. The results for all OECD parental occupation, 17.04, is the increase in reading pro-
countries presented in Table 2 indicate that the effects ficiency associated with a one standard deviation increase
of relational and static cultural capital are considerably in parental occupation scores. These effects also varies at
stronger for reading literacy than for sense of belonging both the school and country levels. Finally, the achievement
at school and occupational aspirations. When these effects gap between males and females is 27.6 points, with females
are estimated in a multilevel model, the pattern is the scoring higher than males. The size of this achievement gap
same; although the effects of static cultural capital on occu- also varies among schools and countries.
pational aspirations are slightly larger. These results are The second model introduces the variables for relational
shown in Table 3. and static cultural capital. There are three key findings in
these results: first, the average effect of relational capi-
(2) Do static and relational cultural capital operate dif- tal is greater than that of static cultural capital; second,
ferently in school systems that are highly selective the effects for both factors vary among schools within
compared with those that are less selective? If this is countries and among countries; third, the two measures
the case, we expect that the two types of capital differ of cultural capital exert effects on reading performance
also in their effects on student performance within and that are independent of the effects of parental education
between schools. and occupation. Thus, they explain some but not all of the
effects associated with the traditional measures of socioe-
Table 3 shows the results for three separate three-level conomic status.
hierarchical models. The first model includes only the vari- The third model introduces Rho, the measure of
ables denoting sex of the student and their parents’ level between-school segregation. It has a strong negative rela-
of education and occupational status. The intercept for the tionship with average school performance as well as the
analyses indicates the average mean score for schools after slopes associated with relational and static cultural capital.
adjusting for the covariates in the model. The superscripts, S The results associated with sense of belonging at school
and C, indicate that the adjusted means vary among schools and occupational aspirations follow the same pattern; how-
within countries and among countries. The average effect ever, the effects attributable to parental education and
of parental education is 3.7, which is the increase in the occupation and the two forms of cultural capital are not
outcome associated with a one-year increase in parental as strong.
 208
Table 2
Regression coefficients estimating the effects of relational and static cultural capital on student reading performance, sense of belonging, and occupational aspirations, controlling for parents’ education, parents’
occupation, and sex.
Reading Sense of belonging Occupational aspirations
Constant Relational Static Parents’ Parents’ Female Constant Relational Static Parents’ Parents’ Female Constant RelationalStatic Parents’ Parents’ Female

L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

cultural cultural education occupation cultural cultural education occupation cultural cultural education occupation
capital capital capital capital capital capital
Australia 501.3 (0.5) 14.2 (0.3) 13.9 (0.5) 5.2 (0.1) 18.7 (0.4) 24.7 (0.8) 492.0 (2.6) 4.2 (2.4) 4.3 (2.0) 1.5 (0.8) 0.4 (2.1) −1.5 (3.2) 489.5 (2.5) 11.3 (2.0) 5.1 (1.9) 0.9 (0.8) 10.9 (2.3) −0.7 (3.0)
Austria 510.2 (2.1) 19.0 (1.7) 11.2 (2.0) 3.4 (0.6) 19.0 (2.0) 16.6 (4.0) 526.2 (2.4) 6.8 (2.1) −0.2 (2.2) 0.5 (0.9) 0.3 (2.5) −9.4 (4.2) 484.5 (2.0) 2.8 (1.7) 9.0 (1.7) 0.9 (0.6) 16.5 (2.0) 15.7 (5.1)
Belgium 512.8 (2.7) 6.5 (1.3) 20.2 (1.8) 3.4 (0.6) 23.2 (2.5) 25.9 (4.4) 481.7 (1.3) 9.5 (1.2) 3.1 (1.3) 0.8 (0.4) 0.2 (1.4) 6.9 (2.3) 499.3 (2.8) 6.5 (1.5) 6.7 (1.3) 0.0 (0.4) 19.5 (1.9) 7.3 (3.0)
Canada 524.0 (1.4) 19.9 (0.9) 11.6 (1.0) 4.1 (0.4) 16.2 (1.0) 23.8 (1.5) 509.9 (1.3) 6.1 (1.1) 6.0 (1.1) 0.7 (0.5) 1.0 (1.3) 2.1 (1.9) 503.8 (0.8) 5.0 (0.9) 4.3 (0.9) 1.7 (0.4) 4.8 (0.9) 13.9 (1.4)
Czech 469.1 (3.2) 14.6 (1.8) 18.3 (3.7) 9.1 (2.3) 21.8 (3.2) 25.2 (4.2) 465.0 (1.5) 13.4 (1.4) 5.1 (1.4) 0.5 (0.7) 1.6 (1.7) −6.1 (2.9) 472.8 (4.2) 11.5 (2.1) 6.7 (3.5) 2.5 (2.9) 16.9 (4.3) −1.7 (4.5)
Republic
Denmark 489.0 (1.8) 26.3 (1.7) 9.9 (1.7) 8.9 (0.7) 10.1 (1.8) 14.8 (2.8) 510.5 (2.1) 3.3 (1.9) 3.3 (1.8) 3.5 (0.9) −2.4 (2.1) −12.1 (3.8) 563.3 (1.8) 8.2 (1.8) 6.8 (1.9) 0.6 (0.9) 8.3 (2.1) 12.2 (3.1)
Finland 542.7 (2.5) 24.4 (2.4) 9.6 (1.8) 3.1 (0.5) 11.3 (1.8) 39.4 (2.7) 500.4 (1.6) 8.0 (1.8) 2.1 (1.7) 0.7 (0.5) −2.5 (1.7) −19.9 (2.9) 491.0 (2.0) 7.3 (2.0) 1.3 (2.3) 3.4 (0.7) 13.3 (1.4) 2.3 (3.0)
France 509.8 (2.0) 10.8 (1.9) 22.5 (1.7) 2.9 (0.5) 17.6 (1.7) 20.0 (2.7) 484.7 (1.8) 10.5 (2.0) 4.5 (1.6) 0.2 (0.5) −2.0 (1.8) −8.0 (2.7) 508.6 (2.5) 7.8 (2.3) 8.2 (2.3) −0.9 (0.8) 14.7 (1.8) 1.0 (3.6)
Germany 482.3 (1.9) 9.6 (2.2) 19.7 (4.5) 5.4 (0.9) 26.3 (3.2) 27.8 (3.9) 518.5 (1.7) 5.8 (1.6) 6.1 (2.4) 0.6 (0.8) 1.3 (2.3) −8.3 (3.5) 490.8 (1.8) 5.0 (2.2) 7.0 (3.1) 0.0 (1.1) 9.0 (3.0) −1.0 (4.0)
Greece 466.6 (4.2) 15.9 (1.9) 16.6 (2.5) 2.8 (0.6) 17.6 (2.0) 29.4 (3.7) 487.4 (2.2) 17.1 (2.1) 4.2 (2.5) −0.8 (0.5) −1.8 (1.5) −3.4 (3.7) 496.9 (2.7) 4.8 (2.1) 4.7 (2.5) 1.2 (0.6) 11.0 (1.4) 4.4 (3.3)
Hungary 455.5 (2.8) 12.8 (1.9) 26.7 (2.1) 11.0 (1.1) 16.1 (2.1) 22.8 (4.0) 499.8 (2.3) 14.3 (2.6) 12.3 (3.0) 1.8 (1.5) 0.6 (2.2) −4.7 (3.0) 470.5 (3.6) 8.1 (2.1) 7.9 (2.6) 7.4 (1.4) 13.8 (2.4) 17.0 (4.0)
Iceland 491.2 (2.2) 20.5 (1.8) 13.1 (2.6) 2.6 (0.6) 8.5 (1.7) 32.0 (3.1) 508.2 (2.6) 5.8 (2.1) 6.5 (2.9) 1.3 (0.7) 3.1 (2.4) −9.0 (4.4) 504.2 (2.4) 3.9 (2.0) 3.9 (2.6) 0.5 (0.7) 11.7 (1.9) −1.9 (3.2)
Ireland 523.8 (2.6) 16.3 (1.9) 13.3 (2.1) 2.3 (0.8) 21.8 (1.9) 18.9 (3.9) 507.5 (2.2) 7.4 (2.1) 1.4 (2.1) 0.1 (0.7) −1.0 (1.9) −5.2 (4.3) 487.3 (1.7) 7.5 (1.8) 5.8 (2.0) 0.8 (0.7) 12.1 (2.2) 3.5 (3.6)
Italy 473.8 (3.1) 16.7 (1.9) 12.1 (2.3) 2.9 (0.7) 13.5 (2.3) 28.7 (5.8) 488.8 (1.9) 11.2 (1.7) 6.6 (1.7) −0.3 (0.8) −5.1 (2.0) −3.1 (3.3) 488.9 (2.3) 6.8 (1.6) 5.9 (1.8) 1.8 (0.5) 11.8 (1.9) 16.3 (3.9)
Japan 538.6 (5.2) 19.0 (1.7) 15.1 (1.8) 1.2 (1.3) 2.5 (2.3) 15.9 (5.1) 467.8 (4.8) 8.0 (1.4) 3.8 (1.7) 1.1 (1.6) 0.7 (1.7) −1.8 (3.3) 526.1 (5.6) −4.3 (1.7) 1.7 (1.9) 0.1 (1.7) 5.2 (2.4) 21.8 (4.7)
Korea 535.8 (2.4) 16.9 (1.2) 7.8 (1.6) 2.9 (0.3) 4.0 (1.7) 8.3 (4.8) 466.3 (1.9) 11.8 (1.2) 11.7 (1.5) −0.4 (0.5) 3.6 (1.7) −20.4 (2.6) 511.3 (1.6) −1.3 (1.3) 0.6 (1.9) 0.7 (0.5) 7.4 (1.7) −10.3 (3.0)
Luxemburg 458.0 (1.6) 2.0 (1.9) 27.2 (1.8) 3.1 (0.5) 21.9 (1.9) 22.8 (3.2) 511.8 (2.1) 4.4 (2.1) 9.3 (2.2) 1.9 (0.6) 6.2 (2.5) −2.7 (4.1) 510.3 (1.5) 3.2 (1.8) 0.2 (1.9) −0.6 (0.6) 8.7 (2.5) −23.0 (4.7)
Mexico 449.1 (3.1) 13.9 (1.6) 14.7 (1.8) 3.9 (0.5) 13.0 (1.7) 17.2 (2.8) 513.1 (3.1) 20.6 (2.1) 3.9 (2.1) 1.1 (0.6) 3.0 (2.1) 4.1 (3.3) 526.8 (1.7) 1.6 (1.4) 0.8 (1.4) −0.4 (0.4) 5.6 (1.6) −9.6 (2.4)
Netherlands 534.7 (2.6) 19.3 (2.0) 11.0 (1.8) 2.7 (0.8) 17.1 (2.6) 18.4 (4.7) 496.5 (2.6) 7.4 (2.7) −3.7 (2.1) 0.5 (0.6) 4.1 (1.7) 2.8 (4.1) 475.3 (2.4) 9.5 (2.5) 1.2 (2.2) 1.4 (0.8) 11.4 (2.7) −10.3 (4.6)
New Zealand 522.0 (2.1) 13.1 (1.8) 12.9 (1.8) 1.9 (0.7) 23.7 (2.0) 37.9 (4.8) 496.5 (2.6) 7.4 (2.7) −3.7 (2.1) 0.5 (0.6) 4.1 (1.7) 2.8 (4.1) 492.3 (1.9) 7.3 (1.5) 3.9 (2.1) 1.7 (0.8) 11.8 (1.7) 4.6 (3.4)
Norway 496.6 (2.7) 21.3 (1.9) 19.3 (2.2) 1.6 (0.8) 15.9 (2.0) 34.0 (3.6) 506.2 (3.0) 7.1 (2.2) 4.8 (2.4) 2.1 (0.6) 2.7 (2.1) −10.7 (3.9) 486.2 (2.3) 3.7 (2.2) 5.2 (2.0) 2.9 (0.9) 14.8 (2.1) 12.6 (3.2)
Poland 478.1 (3.6) 11.6 (2.0) 15.0 (2.5) 5.6 (1.2) 19.6 (2.7) 26.7 (5.7) 458.9 (2.0) 10.1 (2.1) 14.2 (2.5) −1.1 (1.2) 4.1 (2.5) −1.3 (4.3) 507.8 (2.8) 0.8 (2.0) 7.7 (1.9) 2.1 (1.0) 13.0 (2.3) 24.7 (5.3)
Portugal 470.2 (3.5) 27.0 (2.0) 15.5 (1.8) 0.7 (0.6) 23.9 (2.1) 14.3 (3.2) 499.9 (1.9) 17.1 (1.8) 10.3 (1.5) 0.4 (0.4) 3.6 (1.6) −8.3 (2.9) 506.3 (2.1) 5.0 (1.9) 6.9 (2.0) 0.0 (0.5) 10.5 (1.6) −0.6 (3.0)
Spain 493.5 (2.0) 20.5 (1.4) 14.6 (1.6) 3.1 (0.3) 11.3 (1.4) 15.3 (2.4) 494.3 (1.8) 11.4 (1.8) 3.7 (1.9) −0.3 (0.5) −1.0 (1.7) 3.1 (3.1) 504.8 (2.2) 4.7 (1.6) 5.5 (2.2) 2.0 (0.4) 7.4 (1.6) 1.2 (2.9)
Sweden 511.4 (2.3) 16.2 (1.7) 16.4 (1.7) 1.2 (1.0) 17.4 (1.7) 30.7 (2.3) 526.2 (2.1) 7.3 (1.9) 1.4 (2.0) 1.8 (0.9) −3.4 (2.0)−27.2 (3.5) 490.9 (2.0) 7.3 (1.9) 9.6 (1.7) 0.0 (0.9) 11.9 (1.8) 16.0 (3.1)
Switzerland 497.0 (3.1) 17.9 (1.5) 13.6 (1.8) 4.6 (0.7) 23.3 (1.9) 17.1 (3.4) 519.4 (1.9) 1.3 (2.0) 6.0 (2.3) −0.4 (0.7) 3.5 (2.0) −11.8 (3.6) 499.6 (1.9) 0.8 (1.9) 6.1 (1.9) 1.7 (0.6) 13.6 (2.0) 8.1 (4.1)
United 521.0 (1.8) 15.6 (1.6) 16.2 (1.9) 5.2 (0.7) 24.5 (1.7) 18.3 (3.1) 511.7 (1.5) 7.2 (2.0) 2.6 (1.5) 0.5 (0.7) 4.5 (1.5) −6.5 (2.5) 492.9 (2.1) 8.1 (2.0) 4.8 (1.5) 0.7 (0.7) 12.4 (1.7) −3.9 (3.0)
Kingdom
United 501.3 (4.1) 9.6 (2.2) 23.7 (2.8) 5.3 (1.3) 18.8 (2.5) 21.8 (3.4) 489.5 (2.6) 12.5 (2.4) 3.9 (2.2) 1.6 (1.0) 2.1 (2.4) 5.7 (4.0) 516.6 (2.5) 7.9 (1.3) −0.5 (1.8) −0.6 (1.0) 4.5 (1.8) 4.8 (2.7)
States
OECD 501.3 (0.5) 14.2 (0.3) 13.9 (0.5) 5.2 (0.1) 18.7 (0.4) 24.7 (0.8) 493.9 (0.8) 10.5 (0.8) 4.5 (0.7) −0.6 (0.3) 3.7 (0.9) −1.4 (1.3) 505.9 (0.7) 5.6 (0.6) 1.4 (0.7) −0.4 (0.3) 9.7 (0.7) 4.8 (1.2)
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 209

The key findings emanating from the three-level analy-

4.3SC (0.7)

3.4SC (0.8)

1.0SC (0.3)
497.0SC (3.1)

9.6SC (1.0)
4.7 (10.1)

−9.9 (2.5)

−4.6 (2.8)

5.0 (2.6)
ses can be best understood with graphical displays. Fig. 1

Model 3
shows the relationship between the adjusted mean reading
literacy scores and the proportion of variation in reading
scores that is between schools. Recall that the latter mea-
sure is treated as a proxy for student selection, and although

496.8SC (3.0)

5.3SC (0.7)

5.0SC (2.6)
1.0SC (0.3)
3.9SC (1.0)

9.6SC (1.0)
direct measures of the nature and timing of student selec-
tion would be preferable, this measure serves to highlight
Model 2

Note: Effects that vary significantly among schools within countries are denoted with a superscript S; effects that vary significantly among countries are denoted with a superscript C.
the important relationships. For example, countries like
Occupational aspirations

Austria, Germany, and Czech Republic have high scores on

this index. These countries allocate students early in their
school career to academically oriented schools versus voca-
498.9SC (3.4)

7.4SC (2.6)
1.5SC (0.3)
11.0SC (1.2)
tionally oriented schools (e.g., see Straková, Tomášek, &
Model 1

Willms, 2006). Mexico also has a high score on this index,

owing mainly to the ‘learning divides’ associated with rural
schools serving very poor populations, and selection into
private schools in urban areas (Willms, 2006). In contrast,
the Scandinavian countries, which are known for their inte-
−15.47 (16.0)

grated school systems, have relatively low scores on this

4.6SC (0.6)

0.5SC (0.2)
503.8SC (5.1)

−5.9SC (1.4)
9.0SC (1.0)

0.7C (0.5)
1.9 (3.3)

6.3 (3.5)

index.
Model 3

A strong negative effect associated with increasing seg-

regation is readily apparent in Fig. 1. The multilevel analysis
reveals that it is statistically significant, and that an increase
in Rho of 0.1 is associated with a decrease in adjusted read-
505.3SC (4.7)

−7.1SC (3.8)
8.8SC (0.8)

0.5SC (0.3)
4.0SC (0.5)

ing scores of about 6 points (−56.7 divided by 10). There

0.9C (1.1)

are three countries that are notable exceptions: Japan and

Model 2

Korea have relatively high scores, despite high levels of

The relationship between reading performance, sense of belonging, and occupational aspiration and contextual factors.

segregation. It may be that the scores in these countries

Sense of belonging

are bolstered by participation in tutoring programs. Bray

(1999, 2003) estimated that between one-quarter to three-
1.2SC (0.3)

−3.2SC (4.0)
505.9SC (4.1)

3.0SC (1.2)

quarters of students in Asian countries receive tutoring on

Model 1

a regular basis throughout their primary and secondary

school years. The third exception is Netherlands, which we
noted earlier may not have had an adequately representa-
tive sample in PISA 2000.
When we examined the country-by-country results in
497.4SC (4.4)

13.2SC (0.9)

11.4SC (0.9)

2.1SC (0.4)

20.4SC (2.1)
13.0SC (1.5)
−56.7 (16.6)

Table 2, we observed a pattern suggesting that countries

−23.7 (2.9)

−10.2 (2.9)
Model 3

with high average levels of adjusted reading literacy scores

tended to have stronger effects associated with relational
cultural capital, while those with lower scores tended to
have stronger effects associated with static cultural capital.
The multilevel modelling of these relationships allowed us
502.7SC (4.7)

12.3SC (0.9)

2.1SC (0.4)

20.3SC (2.1)
12.9SC (1.5)
15.4SC (1.0)

to discern how these relationships vary within and among

Model 2

countries, and determine whether this variation is also

associated with the segregation of students within and
between schools. The findings are presented in Table 3 and
Fig. 2.
Reading literacy

17.04SC (1.8)

An understanding of these relationships is not straight-

501.9SC (4.2)

3.7SC (0.4)

27.6SC (2.1)

forward, as the slopes associated with SES, or in this case

Model 1

static and relational cultural capital, can differ within and

among schools. For example, in a highly segregated system,
such as Germany, the slope attributable to SES is typically
quite large, while within schools the slope is gradual as the
majority of students in each school come from similar fam-
Relational cultural capital
Student-level variables

ily backgrounds (see Willms, 2010). Fig. 2 shows the among-

Static cultural capital

and within-school relationships between static and rela-

Parent occupation
Parent education

tional cultural capital and the proportion of reading literacy

Adjusted mean

skills that is among schools.

The top two scatter-plots in Fig. 2 show the relation-
Table 3

Female

ships among countries. The first of these plots shows the

Rho

Rho

Rho

relationship between the relational cultural capital among-

210 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

Fig. 1. The association between adjusted mean reading literacy and the variation among schools in reading performance.

school slopes and the index of segregation. The magnitude tance in sponsored systems. In contrast, static cultural
of the among-school slopes are all positive, ranging from capital is important in both types of school systems.
about 5 to 60. The plot also indicates a positive but weak
relationship between the among-school slopes and the 4. Discussion
index of segregation, with steeper slopes in more segre-
gated systems. The effect is 43.1 ( 011 in Eq. (4)), which is In this study we argue that there are two types of cul-
not statistically significant (p < 0.05; standard error is 21.6). tural capital, one that is static, representing the highbrow
The second of the top two plots shows the relationship activities and practices of parents, and one that is rela-
between the static cultural capital among-school slopes tional, representing children’s activation of communication
and the index of segregation. The magnitude of the among- and cultural interactions. We used data for 28 countries
school slopes are all positive, ranging from about 15 to 90.
The plot indicates a strong, positive relationship between
the among-school slopes and the index of segregation, with
much steeper slopes in more segregated systems. The effect
is 68.9 ( 021 in Eq. (4)), which is statistically significant
(p < 0.05; standard error is 18.6).
The next two scatter-plots, at the bottom of Fig. 2, dis-
play the relationships for the within-school slopes. The first
of the two plots shows the relationship between the rela-
tional cultural capital slopes and the index of segregation.
The slopes range in size from about 8 to 25, and there is
a strong negative relationship between the within-school
slopes and the index of segregation. The effect is −22.6
( 101 in Eq. (4)), which is statistically significant (p < 0.05;
standard error is 3.4). The second of the two plots shows
the relationship between the static cultural capital slopes
and the index of segregation. The slopes range in size from
about 3 to 20, and there is also a strong negative relation-
ship between the within-school slopes and the index of
segregation. The effect is −26.2 ( 201 in Eq. (4)), which is
statistically significant (p < 0.05; standard error is 3.0).
Estimates of the combined among- and within-school
effects of relational and cultural capital are provided in
Table 3. Relational cultural capital is an important asset in
contest schooling systems, but is of relatively little impor-
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 211

Fig. 2. The association between dynamic and static cultural capital slopes, among and within schools, versus the variation among schools in reading
performance.

special programs within the public sector that are more while the within-school relationships were weaker than
attractive to families with greater economic resources. In among-school relationships in sponsored systems. Static
contrast, vertical segregation mainly stems from the selec- cultural capital plays a relatively minor role within schools
tion of students into particular schools based on their in both contest and sponsored systems.
ability at certain points in their school career. Our approach to cultural capital and our empirical find-
A partitioning of the effects of the two types of cap- ings are not at odds with Lareau and Weininger’s (2003)
ital revealed that the effects vary in their among-school definition of cultural capital, or their call for detailed
and within-school relationships among countries, and this micro-analyses of the formal and informal ways students
depends on the type of school system. Our interpretation of are evaluated. However, the PISA data lack information
these results is that students’ static cultural capital renders on many of these important mechanisms. For example,
them eligible for recruitment in sponsored systems, while it would be useful to have cross-cultural data on par-
in contest systems it does not afford them as much advan- ents’ and children’s reading habits, use of television, and
tage. Relational cultural capital plays a lesser role than static engagement in other cultural and recreational activities.
cultural capital in determining what type of school a child We also want to understand the nature of parents’ inter-
attends. In contrast, once students are ‘in the door,’ rela- actions with their children, and the extent to which they
tional cultural capital has a stronger relationship than static make an effort to transmit cultural capital to their chil-
cultural capital, particularly in contest systems. Therefore, dren. This focus would be close to the work on parenting
in contest systems we observed stronger relationships for styles (e.g., Beyer, 1995; Lamborn, Dornbusch, & Steinberg,
relational cultural capital within schools than among them, 1996; McLaughlin & Vacha, 1992; Steinberg, Lamborn,
212 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213

Dornbusch, & Darling, 1992; Taylor, Hinton, & Wilson, 1995) ing inequalities. It may be, therefore, that children’s school
and somewhat related to research on parental involvement experience in these countries is not that much different
(e.g., Astone & McLanahan, 1991; Epstein, 1987; Fehrmann, from that of other OECD countries, but the wider culture
Keith, & Reimers, 1987; Lareau, 1987; Stevenson & Baker, supports the translation of static to relational cultural cap-
1987), especially as it pertains to direct parental involve- ital in positive ways that support the efforts of schools. We
ment at home (e.g., see Ho & Willms, 1996). believe that all three mechanisms operate simultaneously.
Our findings suggest that part of the answer to why Levin’s work has been influential here as well. He set
some schools and schooling systems perform better than out a model for educational reform based on the belief that
others lies in a complex interplay between schools’ formal disadvantaged students could succeed if they were afforded
and informal evaluation criteria, the level of segregation some of the same approaches to learning as gifted students,
of the educational system, parents’ skills and strategies including opportunities to construct knowledge through
to advance their child’s schooling, and the ways students hands-on projects and research (Levin, 1989). He devel-
respond and activate their cultural resources. oped the concept of “accelerated schools” (Levin, 1996,
Two important pieces of the puzzle that are missing 1998) which embodied a three-part philosophy: (1) unity
are how parents’ static cultural capital and children’s rela- of purpose, (2) empowerment coupled with responsibility,
tional capital relate to students’ acquisition of literacy skills and (3) building on students’ strengths through the use of
in contest and sponsored schooling systems during the strategies commonly used for gifted learners. The slogan
primary school years, and the role that the two forms of was “accelerate, don’t remediate”, which is consistent with
cultural capital plays thereafter, particularly as students the need for students to translate static to relational cultural
with differing levels of ability encounter various selection capital in a contest system.
points that channel them into particular schools or school
programs. Most children make a successful transition from
learning-to-read to reading-to-learn by the end of grade 3 References
or 4, and thereafter are able to take advantage of the learn-
Astone, N. M., & McLanahan, S. S. (1991). Family structure, parental prac-
ing opportunities that lie ahead. But a large percentage of tices and high school completion. American Sociological Review, 56,
children do not make this transition successfully (Beswick 309–320.
& Sloat, 2006); estimates for the US for example are about Becker, G. (1964). Human capital. New York: Natl. Bur. Econ. Res.
Beswick, J. F., & Sloat, E. A. (2006). Early literacy success: A matter of social
40%, based on the National Assessment of Reading Profi- justice. Education Canada, 46, 23–26.
ciency (Fletcher & Lyon, 1998). A disproportionate number Beyer, S. (1995). Maternal employment and children’s academic achieve-
of these children are from low SES families (Willms, 2006). ment: Parenting styles as mediating variable. Developmental Review,
15, 212–253.
It is a reasonable hypothesis that in contest systems the cru- Bielby, W. T. (1981). Models of status attainment. In D. J. Treiman, & R. V.
cial ‘contest’ is the acquisition of strong literacy skills during Robinson (Eds.), Research in social stratification and mobility. Green-
the primary years, while in sponsored systems there may wich: JAI Press Inc.
Blau, P., & Duncan, O. (1967). The American occupational structure. New
be more options, including tutoring, special private schools, York: Free Press.
and alternative programs. Bourdieu, P. (1977). Cultural reproduction and social reproduction. In J.
Karabel, & A. H. Halsey (Eds.), Power and ideology in education. New
York: Oxford University Press.
5. Conclusions
Bourdieu, P. (1984). Distinction: A social critique of the judgement of taste.
London: Routledge.
The differences among countries in the patterns of Bourdieu, P. (1986). The forms of capital. In J. C. Richardson (Ed.), Hand-
book of theory and research for the sociology of education. New York:
effects associated with relational and static cultural capi-
Greenwood Press.
tal, and in the effects of parental education and occupation Bowles, S., & Gintis, H. (2000). Does schooling raise earnings by making
also warrant discussion. Willms’ (2003b, 2003c, 2006) people smarter? In K. Arrow, S. Bowles, & S. Durlauf (Eds.), Meritocracy
hypothesis of converging gradients holds that socioeco- and economic inequality (pp. 118–136). Princeton, NJ: Princeton Univ.
Press.
nomic gradients in literacy skills tend to converge at higher Bowles, S., & Gintis, H. (2002). Schooling in capitalist America revised.
levels of socioeconomic status. This suggests that youth Sociology of Education, 75, 1–18.
from high SES backgrounds tend to do well in any juris- Bray, M. (1999). The shadow education system: Private tutoring and its
implication for planners. Paris: UNESCO International Institute for Edu-
diction, while those with low SES backgrounds tend to cational Planning.
vary considerably in their proficiency among jurisdictions. Bray, M. (2003). Financing education in Asian and Pacific countries. In J.
Our findings are consistent with this hypothesis, but they P. Keeves, & R. Watanabe (Eds.), International handbook of educational
research in the Asia-Pacific region. Dordrecht/Boston/London: Kluwer
add a new dimension. Willms attributed the convergence Academic Publishers.
of socioeconomic gradients to two plausible mechanisms. Chiswick, B. R. (1983). The earnings and human capital of American Jews.
One is that successful jurisdictions achieved high and grad- Journal of Human Resources, 18, 313–336.
Collins, R. (1971). Functional and conflict theories of educational stratifi-
ual gradients by successful school policies and practices
cation. American Sociological Review, 36, 1002–1019.
that bolstered the performance of low SES students. The Costanza, R., & Daly, H. E. (1992). Natural capital and sustainable develop-
other is that in jurisdictions with low quality schooling, ment. Conservation Biology, 6, 37–46.
Crook, C. J. (1997). Cultural Practices and Socio-economic Attainment: the
high SES families provide additional educational support
Australian Experience. Westport, CT: Greenwood Press.
for their child with extra help at home, or with extra tutor- De Graaf, N. D., De Graaf, P. M., & Kraaykamp, G. (2000). Parental cultural
ing. The findings in this study suggest that there is a third capital and educational attainment in the Netherlands: A refinement
plausible mechanism: jurisdictions with high and gradual of the cultural capital perspective. Sociology of Education, 73, 92–111.
DiMaggio, P. (1982). Cultural capital and social success: The impact of
gradients achieve this by cultural mechanisms that support status culture participation on the grades of high school students.
all students, thereby increasing performance and reduc- American Sociological Review, 47, 189–201.
 L. Tramonte, J.D. Willms / Economics of Education Review 29 (2010) 200–213 213

Duncan, O. D., Haller, A. O., & Portes, A. (1968). Peer influences on aspira- Organisation for Economic Cooperation and Development (OECD). 2001.
tions: A reinterpretation. American Journal of Sociology, 74, 119–137. Knowledge and skills for life: First results from the OECD Pro-
Dunifon, R., Duncan, G. J., & Brooks-Gunn, J. (2001). As ye sweep, so shall gramme for International Student Assessment (PISA) 2000. Paris:
ye reap. American Economic Review, 91, 150–154. OECD: author.
Epstein, J. L. (1987). Parent involvement: What research says to adminis- OECD, and UNESCO Institute for Statistics. 2003. Literacy skills for the
trators. Education and Urban Society, 19, 119–136. world of tomorrow: Further results from PISA 2000. Paris/Montreal:
Farkas, G. (1996). In A. de Gruyter (Ed.), Human capital or cultural capital? authors.
Ethnicity and poverty groups in an urban school district. New York. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Appli-
Farkas, G. (2003). Cognitive skills and noncognitive traits and behaviors in cations and data analysis methods (2nd ed.). Thousand Oaks: Sage
stratification processes. Annual Review of Sociology, 29, 541–562. Publications.
Fehrmann, P. G., Keith, T. Z., & Reimers, T. M. (1987). Home influence on Robbins, D. (1991). The work of Pierre Bourdieu: Recognizing society. Milton
school learning: Direct and indirect effects of parental involvement on Keynes: Open University Press.
high school grades. Journal of Educational Research, 80, 330–337. Rowan, B., Raudenbush, S. W., & Kang, S. J. (1991). School climate in sec-
Finn, J. D., & Rock, D. A. (1997). Academic success among students at risk ondary schools. In S. W. Raudenbush, & J. D. Willms (Eds.), Schools,
for school failure. Journal of Applied Psychology, 82, 221–234. classrooms, and pupils: International studies of schooling from a multi-
Fletcher, J., & Lyon, R. (1998). Reading: A research-based approach. In W. level perspective. San Diego: Academic Press.
Evers (Ed.), What’s gone wrong in America’s classrooms. Palo Alto, CA: Schultz, T. W. (1960). Capital formation by education. Journal of Political
Hoover Institution Press, Stanford University. Economy, 68, 571–583.
Ganzeboom, H. B. C., De Graaf, P. M., & Trieman, D. J. (1992). A standard Schultz, T. W. (1981). Investing in people. Berkeley: University of California
international index of occupational status. Social Science Research, 21, Press.
1–56. Sewell, W. H., & Hauser, R. M. (1975). Education, occupation, and earnings:
Haller, A. O., & Butterworth, C. E. (1960). Peer influences on levels of occu- Achievement in the early career. New York: Academic Press.
pational and educational aspirations. Social Forces, 38, 289–295. Sewell, W. H., Hauser, R. M., & Featherman, D. (Eds.). (1976). Schooling and
Hanushek, E. A. (1989). The economics of schooling: Production and effi- achievement in American society. New York: Academic Press.
ciency in public schools. Journal of Economic Literature, 24, 1141–1177. Sewell, W. H., Haller, A. O., & Portes, A. (1969). The educational and early
Heckman, J., & L. Lochner. 2000. Distinguishing between learning-by- occupational attainment process. American Sociological Review, 34,
doing and on-the-job training. Working Paper. 82–92.
Heckman, J. J., & Rubinstein, Y. (2001). The importance of noncognitive Sewell, W. H., & Shah, V. P. (1968). Parents’ education and children’s edu-
skills: Lessons from the GED testing program. American Economic cational aspirations and achievements. American Sociological Review,
Review, 91, 145–149. 33, 191–209.
Ho, E., & Willms, J. D. (1996). The effects of parental involvement on eighth Sirin, S. R. (2005). Socioeconomic status and academic achievement: A
grade achievement. Sociology of Education, 69, 126–141. meta-analytic review of research. Review of Educational Research, 75,
Kerckhoff, A. C. (1975). Stratification processes and outcomes in England 417–453.
and the United States. American Sociological Review, 39, 789–801. Steinberg, L., Lamborn, L. S. D., Dornbusch, S. M., & Darling, N. (1992).
Kerckhoff, A. C. (Ed.). (1996). Generating social stratification. Boulder: West- Impact of parenting practices on adolescent achievement: Authori-
view Press. tative parenting, school involvement, and encouragement to succeed.
Kinloch, G. C. (1969). Sponsored and contest mobility among college grad- Child Development, 63, 1266–1281.
uates: Measurement of the relative openness of a social structure. Stevenson, D. L., & Baker, D. P. (1987). The family-school relation and the
Sociology of Education, 42, 350–367. child’s schools performance. Child Development, 58, 1348–1357.
Lamborn, S. D., Dornbusch, S. M., & Steinberg, L. (1996). Ethnicity and com- Straková, J., Tomášek, V., & Willms, J. D. (2006). Education inequalities in
munity context as moderators of the relations between family decision the Czech Republic. Prospects, XXXVI, 517–538.
making and adolescent adjustment. Child Development, 67, 283–301. Sullivan, A. (2001). Cultural capital and education attainment. Sociology,
Lamont, M., & Lareau, A. M. (1988). Cultural capital: Allusions, gaps and 35, 893–912.
glissandos in recent theoretical developments. Sociological Theory, 6, Swidler, A. (1986). Culture in action: Symbols and strategies. American
153–168. Sociological Review, 51, 273–286.
Lareau, A. M. (1987). Social class differences in family-school relationships: Taylor, L. C., Hinton, I. D., & Wilson, M. N. (1995). Parental influences on
The importance of cultural capital. Sociology of Education, 60, 73–85. academic performance in African-American students. Journal of Child
Lareau, A. M. (1989). Home advantage. London: Falmer Press. and Family Studies, 4, 293–302.
Lareau, A. M., & Horvat, E. (1999). Moments of social inclusion and exclu- Throsby, D. (1999). Cultural capital. Journal of Cultural Economics, 23, 3–
sion: Race, class, and cultural capital in family-school relationships. 12.
Sociology of Education, 72, 37–53. Turner, R. (1960). Sponsored and contest mobility and the school system.
Lareau, A. M., & Weininger, E. B. (2003). Cultural capital in educational American Sociological Review, 25, 855–867.
research: A critical assessment. Theory and Society, 32, 567–606. Willms, J. D. (2003a). Student engagement at school: A sense of belonging
Levin, H. M. (1974). Measuring efficiency in educational production. Public and participation. Paris: OECD.
Finance Quarterly, 2, 3–24. Willms, J. D. (2003b). Literacy proficiency of youth: Evidence of converging
Levin, H. M. (1989). Economics of investment in educationally disadvan- socioeconomic gradients. International Journal of Educational Research,
taged students. American Economic Review, 79(2), 52–56. 39, 247–252.
Levin, H. M. (1996). Accelerated schools after eight years. In L. Schauble, & Willms, J. D. (2003c). Ten hypotheses about socioeconomic gradients and
R. Glaser (Eds.), Innovations in learning: New environments for education community differences in children’s developmental outcomes. Ottawa:
(pp. 329–352). Mahwah, NJ: Lawrence Erlbaum. Applied Research Branch of Human Resources Development Canada.
Levin, H. M. (1997). Raising school productivity: An X-efficiency approach. Willms, J. D. (2006). Learning divides: Ten policy questions about the perfor-
Economic of Education Review, 16(3), 303–313. mance and equity of schools and schooling systems. Montreal: UNESCO
Levin, H. M. (1998). Accelerated schools: A decade of evolution. In A. Har- Institute for Statistics.
greaves, A. Lieberman, M. Fullan, & D. Hopkins (Eds.), International Willms, J. D. (2010). School Composition and Contextual Effects on Student
Handbook of Educational Change (pp. 807–830). Dordrecht: Kluwer Outcomes. Teachers College Record, 112, 3–4.
Academic Publishers. Willms, J. D., & T. M. Smith. 2005. A manual for conducting analyses with
McLaughlin, T. F., & Vacha, E. F. (1992). The at-risk student: A proposal for data from TIMSS and PISA. Report prepared for the UNESCO Institute
action. Journal of Instructional Psychology, 19, 66–68. for Statistics.
Mincer, J. (1958). Investment in human capital and personal income dis- Willms, J. D., Smith, T. M., Zhang, Y., & Tramonte, L. (2006). Raising and
tribution. Journal of Political Economy, 66, 281–302. levelling the learning bar in Central and Eastern Europe. Prospects,
Mincer, J. (1974). Schooling, Experience, and Earnings. New York: Columbia XXXVI, 411–418.
University Press. Woodbury, S. A. (1993). Culture and human capital: Theory and evidence
Newman, B. M., Lohman, B. J., Newman, P. R., Myers, M. C., & Smith, V. C. or theory versus evidence. In W. Darity Jr. (Ed.), Labour economics:
(2000). Experiences of urban youth navigating the transition to ninth Problems in analyzing labor (pp. 239–267). Boston: Markets. Kluwer
grade. Journal of Youth and Society, 31, 387–416. Academic Publishers.