Scholarly Culture and Academic Performance in 42 Nations

Abstract

Exposure to books and high culture provides important academic advantages. But the reasons for this are hotly disputed. Elite
closure theory posits that culture merely signals children's elite status to gatekeepers who then grant them unjust advantages.
But other theories suggest that scholarly culture provides cognitive skills that improve academic performance, which schools
justly reward. We attempt to adjudicate between these theories using data on academic performance from 42 national samples
with 200,144 cases from OECD's PISA. We find that a key aspect of scholarly culture, the number of books in the family home,
exerts a strong influence on academic performance in ways consistent with the cognitive skill hypothesis, regardless of the
nation's ideology, political history, or level of development.

Full text

© The Author 2014. Published by Oxford University Press on behalf of the
University of North Carolina at Chapel Hill. All rights reserved. For permissions,
please e-mail: journals.permissions@oup.com.
Social Forces 00(0) 1–34, Month 2014
doi: 10.1093/sf/sou030
Address correspondence to M. D. R. Evans; Email: MariahEv2@gmail.com. This paper was supported
in part by a grant to Evans from the Nevada Agricultural Experiment Station in connection with
CSREES W2001 (later W3001).
Scholarly Culture Test Scores
Scholarly Culture and Academic Performance
in 42 Nations
M. D. R. Evans, University of Nevada
Jonathan Kelley, International Survey Center and University of Nevada
Joanna Sikora, Australian National University
Exposure to books and high culture provides important academic advantages. But
the reasons for this are hotly disputed. Elite closure theory posits that culture
merely signals children’s elite status to gatekeepers who then grant them unjust
advantages. But other theories suggest that scholarly culture provides cognitive skills
that improve academic performance, which schools justly reward. We attempt to adju-
dicate between these theories using data on academic performance from 42 national
samples with 200,144 cases from OECD’s PISA. We find that a key aspect of scholarly
culture, the number of books in the family home, exerts a strong influence on academic
performance in ways consistent with the cognitive skill hypothesis, regardless of the
nation’s ideology, political history, or level of development.
Throughout the world, life chances are shaped by education. Hence, both policy
concerns about improving opportunities for the disadvantaged and theoretical
questions about the lifelong consequences of childhood socialization have spurred
scholars in the Blau-Duncan paradigm to explore how family background affects
education. Research has revealed that parents’ education and occupational status
have substantial effects on their offspring’s education in all countries yet stud-
ied and has estimated the magnitude of these effects with considerable preci-
sion (Blau and Duncan 1967; Marks, Cresswell, and Ainley 2006; Reisel 2011;
Treiman and Yip 1989; Warren and Hauser 1997). These effects are net of many
schools’ efforts to educate children regardless of their family situations, indicat-
ing that it is very difficult fully to compensate for low parental education and
occupational status, characteristics that are difficult to change. In search of barri-
ers that may be easier to breach, research has proceeded to explore the influences
of other features of the cognitive environment of the home.
Notably, research has demonstrated that scholarly culture—homes that
abound in books and in which the way of life involves esteeming, reading, and
enjoying books—is associated with higher educational attainment in several rich
Scholarly Culture Test Scores 1
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Western countries, net of social class and many other factors (de Graaf 1986; de
Graaf, de Graaf, and Kraaykamp 2000; Evans and Kelley 2000). Intriguingly,
researchers curious about the generality of these results also discovered them in
Hungary (Ganzeboom, de Graaf, and Robert 1990), then in a wider set of Eastern
European countries (Kraaykamp and Nieuwbeerta 2000), and more recently in
a broad range of countries throughout the world: market-oriented, Communist,
post-Communist, advanced, and developing alike (Evans et al. 2010).
How does scholarly culture affect educational attainment? One important possi-
bility is that academic performance is the mechanism, the key intermediate variable.
Moreover, tests with different dependent/response variables are highly valuable,
since the more domains in which hypotheses derived from a theory stand (or fall),
the stronger the case for or against the theory (Stinchcombe 1968). Spaeth’s schol-
arly culture thesis holds that the way of life embedded in the scholarly culture
develops cognitive skills, capacities, and tastes that enhance educational perfor-
mance and thereby both encourage and enable children to go further in school.
By contrast, the highly influential elite closure/social reproduction/cultural repro-
duction theory (much of it inspired by Bourdieu, his followers, and those claim-
ing—perhaps not always correctly—to follow in his path) holds that exposure to
books endows children with a style that signals their elite status. Gatekeepers such
as teachers (and later employers) recognize those signals and then provide elite chil-
dren with access to unmerited advantages that are hoarded from outsiders.
The evidence to date supports the scholarly culture model for educational
attainment (both for years of education and in terms of stages/transitions), rather
than the elite closure model (Crook 1997a; de Graaf, de Graaf, and Kraaykamp
2000; Evans and Kelley 2000; Evans et al. 2010; Goldthorpe 2007; Kingston
2001; Tzanakis 2011; but see Rivera 2012). Nonetheless, the elite closure model
retains a tenacious hold on the sociological imagination, so further tests are
needed, and extending the hypotheses to include predictions about the potential
mechanism—academic performance—can deepen our understanding.
The OECD’s PISA (Program for International Student Assessment) surveys
provide excellent material for this analysis, because the two theories have very
different implications for the effects of scholarly culture on children’s perfor-
mance on a test that is anonymously graded and designed to minimize class and
ethnic biases. Specifically, this paper examines reading achievement at age 15,
based on tests conducted on over 200,000 students in 42 nations—a very broad,
albeit still incomplete, array of countries.
Prior Research1 and This Paper’s Hypotheses
As noted above, two broad themes dominate research on the linkage between
scholarly culture in the home and educational outcomes. One focuses on schol-
arly culture as a “toolkit” of competencies, skills, and funds of knowledge
(Kelley and Evans 2002; Miller, Kohn, and Schooler 1985; Pearlin and Kohn
1966; Spaeth 1976, 1979, 1989; Swidler 1986). The other is a widely endorsed,
but diversely named, “elite closure” or “cultural capital” or “signaling” or
“secret handshake” model focusing on elites’ hoarding advantages by using
2 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

cultural signals to recognize fellow members and to exclude others (Goblot
1925 [1973]; Bourdieu 1984; Kingston 2000; van de Werfhorst 2010).2 The
theoretical rationales for both are well summarized and assessed in Crook
(1997a) and Kingston (2001).
Theory: Scholarly Culture
The scholarly culture hypothesis holds that reading provides cognitive skills that
enhance educational performance. A home with books as an integral part of the
way of life encourages children to read for pleasure and encourages discussion
among family members about what they read, thereby providing children with
information, vocabulary, imaginative richness, wide horizons, and skills for dis-
covery and play (Bus and Ijzendoorn 1995; Dronkers 1992; Persson 2012; Price
2012). Stemming from work by Spaeth and P. de Graaf (de Graaf 1986; Spaeth
1976), this approach suggests a substantive connection linking scholarly resources
to cognitive skill and complexity (Bidwell 1989; Farkas 1996; Kohn, Naoi, et al.
1990; Kohn, Slomczynski, et al. 1997; Krashen 2004; Teachman, Paasch, and
Carver 1997). That generates a substantial effect of home library size on edu-
cational performance. In this view, books are a concrete resource. Moreover,
the size of the home library indicates the family’s commitment to investing in
knowledge (Crook 1997b; Dronkers 1992) and suggests that conversations of
parents with their children will include references to books and imaginative play
growing out of them. For example, research in the Netherlands has shown that
both role modeling via parental reading customs and explicit parental mentor-
ing about reading influence their offspring’s educational attainment (Notten and
Kraaykamp 2010). Moreover, an emerging body of experimental research sug-
gests that supplying books to children for summer reading can eliminate the sum-
mer setback (Heyns 1978) anticipated for them and enhance their performance
on standardized reading tests (Allington et al. 2010; Kim 2006; Kim and White
2008). Thus, a substantial foundation of research encourages proponents of the
scholarly culture theory to hypothesize that books enhance cognitive capacities in
ways that are directly useful in school, improving academic performance.3 They
are both a resource in themselves and an indicator of other resources.
Theory: Elite Closure
By contrast, the elite closure hypothesis posits that, with the transition to very
large societies and the waning of distant kinship bonds, elites have developed
a new method of identification and status maintenance, a “secret handshake”
equivalent, to enable them to recognize one another and control entry into the
elite. These arguments have been put forward by many writers, some inspired
by Bourdieu (1984, 1986); there are particularly clear expositions in Kingston
(2000), Lamont and Lareau (1988), and van de Werfhorst (2010). We focus on
the stratification-related elite-closure theory—a theory that has played the major
role in a lively research stream of research in the status attainment tradition
(Kingston 2001; Tzanakis 2011); this has been dubbed the “domesticated” ver-
sion of the argument (Goldthorpe 2007; see also Swartz 1997).
Scholarly Culture Test Scores 3
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

According to the elite closure thesis, arts spectatorship—especially attendance
at opera, ballet, classical music, and theater performances—are often-suggested
signals (this assumes that these are distinctively elite activities). Once a new-
comer is recognized as elite, existing elite members offer help and assistance.
So, for example, children of elite families attend performances of opera, ballet,
classical music, and plays; because of that, they are recognized as elite by their
schoolteachers, who then reward them beyond their just deserts.4 This is class
favoritism; the logic is the same as for ethnic, racial, or religious favoritism.
The elite closure hypothesis can be seen as a specific application of a more
general social psychological theory emphasizing the role of arbitrary signals in
hierarchy and boundary maintenance, for which there is strong evidence in the
experimental literature (Ridgeway and Erickson 2000; Ridgeway et al. 2009).
The logic of this highly influential argument, stemming in part from the classical
concept of status groups as bounded entities (Weber 1968), is certainly seduc-
tive.5 The question in the stratification-related work is not whether arbitrary sig-
nals can enable and reinforce hierarchy and boundary maintenance, but whether
scholarly culture is an instance of an arbitrary signal of this kind.
The evidence to date is not supportive of elite closure. Empirically, prior
research has not found the unified high-culture lifestyle it predicts. Instead, ever
since P. de Graaf (1986) first raised the empirical issue, every study with mul-
tiple indicators has found that reading practices are distinct from arts specta-
torship, that is, attendance at plays, opera, classical music concerts, and art
galleries (e.g., Crook 1997b; de Graaf and de Graaf 2001; de Graaf, de Graaf,
and Kraaykamp 2000; Evans and Kelley 2002a). To date, these analyses cover
only two societies (Australia and the Netherlands), both highly developed but
not otherwise distinctive so far as we can tell (appendix table 1, panel 2). Hence,
more research is needed, but the available evidence is against any unified high-
culture lifestyle. Moreover, the only studies that find “cultural capital effects”
are those that ignore this clearly established distinction and shoehorn reading
indicators into their cultural capital indices. Furthermore, prior evidence sug-
gests channel-specific inheritance with scholarly culture strongly linked across
the generations, with parents’ arts spectatorship strongly linked to offspring’s
arts spectatorship, but with links between scholarly culture and arts-related cul-
ture being created anew in each generation (Evans et al. 2010, figure 6).
Most of the research to date uses educational attainment (years of schooling,
completion of educational stages) as the dependent variable. Nonetheless, the
theories have clear implications for school performance (reading ability, math
skills). Exploring effects on this new dependent variable provides a valuable
extension of the domain of things explained. Existing research on it mixes schol-
arly culture with arts spectatorship, but at least suggests that it is reasonable to
expect an effect of home library size on standardized reading-test scores (Byun,
Schofer, and Kim 2012; Park 2008).
Differences between the Theories: Functional Form (H1)
The theories differ on whether home library size matters only for elite entry
or whether it is also important for the “ordinary success” that enables people
4 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

from disadvantaged backgrounds to climb a little higher on the socioeconomic
ladder. Empirically, this implies relationships with very different functional
forms.
The scholarly culture thesis suggests diminishing returns: For a child from a
bookless family, each additional exposure to scholarly culture is likely to make a
real contribution to their funds of skill and information. But if a child is already
adept in the scholarly culture, the likelihood is that the new resources they
encounter will contain few unfamiliar elements, because their existing “stocks”
of skill and information are already large. Thus, going from 0 books to 50 (for
example) will greatly increase educational attainment, while going from 400 to
450 will increase it only slightly. So, the relationship between home library size
and school performance should be approximately linear in the log of the number
of books.
By contrast, the elite closure thesis holds that increasing home library size
from 0 books to 50 (to continue the example) should have little if any effect:
Such children are not serious candidates for elite entry in the first place. Whether
they have a few more books, or a few less, is irrelevant. On this argument,
books begin to matter only near the boundary that separates the elite from the
masses: For someone near the borderline, the difference between 400 and 450
might just tip the scales. Thus, there should be little difference between 0 books
and a small library, but more of a difference between a middle-sized library and
a large one.6
Prior research on educational attainment finds that the largest gains come
from increased numbers of books at the bottom (as scholarly culture theory
predicts), not at the top (as elite closure theories predict): Additional books help
children from book-deprived families go a little further in school. This is true
worldwide (Evans et al. 2010). It seems reasonable to anticipate that the same
will hold true for academic performance on standardized reading tests.7 If so,
that too is a matter of ordinary success, not elite entry. So:
H1. (Scholarly culture hypothesis: Biggest gains at the bottom) The edu-
cational payoff for participation in the scholarly culture will be greater
where the baseline level is low.
Empirically, the functional form will be that the natural log of the num-
ber of books in the home will be linearly related to reading achievement.
By contrast, the elite closure model predicts a Matthew effect, with advantages
going more to those who are already well off: Educational returns to the number
of books increase rather than diminish. So having, for example, 15 books rather
than 10 does not matter, but additional books at the high end, say 315 versus
310, could be the passport to elite entry. Thus, the relationship should be rather
flat across the smaller home library sizes and then kick up sharply at larger home
library sizes. A positive quadratic term or a spline function would be likely to
capture this functional form nicely.8
Thus, the elite closure approach suggests the opposite of diminishing mar-
ginal returns: Educational returns should be an increasing function of books in
the home, because it is an issue of inclusion into an elite:
Scholarly Culture Test Scores 5
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

H1-Alt. (Elite closure alternative: Biggest gains at the top) The educa-
tional payoff for participation in the scholarly culture will be greater
where the baseline level is high.
Empirically, the number of books in the home will have a curvilinear
relationship with reading achievement, flat at first and then increasingly
steep.
Differences between the Theories: Generality in Advanced
Societies (H2)
Existing research first found that scholarly culture in the home is associated with
more schooling for children in a handful of rich Western countries (de Graaf
1986; de Graaf, de Graaf, and Kraaykamp 2000; Evans and Kelley 2000). This
is net of parents’ education, father’s occupational status, urban/rural residence,
number of siblings, and other control variables. This was extended to Hungary
(Ganzeboom, de Graaf, and Robert 1990) and then throughout much of the
world (Evans et al. 2010).
By extension, scholarly culture theory implies that culture’s effects on edu-
cational performance (reading skills and the like) will be found in all nations
because of the intrinsic connection between scholarly culture and cognitive
skills. Thus:
H2. (Scholarly culture: Gains in all advanced societies) Scholarly cul-
ture will enhance children’s educational performance in all advanced
nations.
By contrast, the elite closure/cultural reproduction hypothesis sees the connec-
tion between scholarly culture and academic success as essentially arbitrary:
Any unique signal that reliably distinguishes an existing elite (accent, vocabu-
lary, manners, appearance, food or leisure preferences, etc.) would work. So,
there should be a great deal of variation among countries. Moreover, it will
often matter who the elite is, and what sort of people they want to favor:
H2-Alt. (Elite closure alternative: Gains depend on the elite’s choice of
signals) Book ownership will enhance children’s educational attainment
only in some nations.
Differences between the Theories: Developing Societies (H3)
In prior research, the positive effect of books on educational attainment has
been found throughout a broad range of countries across the world, so scholarly
culture does not appear to be a signal of membership in a particular elite, such
as the elite of France or the United States. This is consistent with the scholarly
culture theory. It is contrary to the expectation from elite closure theory that a
scholarly culture effect should be absent at low levels of development, emerging
only with socioeconomic development as the strength of kin ties wanes and as
the increasing scale of societies (and hence of elites) demands new signals of elite
membership.
6 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Research to date has focused largely on scholarly culture’s influence on years
of education in economically advanced societies. But it also has found sub-
stantial effects on test performance for fourth graders in 25 nations in a wide
range of societies (Park 2008), consistent with scholarly culture’s implication of
universality. Hence:
H3. (Scholarly culture: Gains in developing societies, too) Scholarly
culture will enhance children’s educational performance in developing
societies as well as in advanced nations.
By contrast, elite closure theory posits that the link of home library size to edu-
cational success is arbitrary, so it should matter who the elite is, and what sort
of people they want to favor. A conservative, French elite might want to favor
the offspring of a well-read bourgeois family. But a devout religious elite in Latin
America, or an ethnic elite in the Balkans, or a political elite in Russia or China
would more likely choose to favor the devout, or co-ethnics, or the politically
faithful:
H3-Alt. (Elite closure alternative: Few gains in developing societies)
Different signals of elite membership will hold at different levels of
socioeconomic development, with less developed societies relying on
particularistic ties, especially religion, kinship, ethnicity, or politics,
rather than on signals related to high culture.
Additional Explorations: Compensation and Cumulative Advantage
In addition to the questions addressed above, it is also reasonable to ask whether
home libraries are especially important for those from low-status origins.
Evidence of greater educational gains for children whose parents have little edu-
cation has been found for the Netherlands (de Graaf, de Graaf, and Kraaykamp
2000) and subsequently in many other countries (Evans et al. 2010, figure 3).9
We dub this important hypothesis the GGK compensation hypothesis:
H4. (GGK compensation hypothesis) Scholarly culture will especially
enhance the educational performance of children whose parents are low
in the social hierarchy.
Specifically, there will be interaction effects such that the log number of
books will have especially large positive effects for children whose par-
ents have little education and/or low occupational status.
Cumulative advantage (the Matthew effect) is an alternative: that home libraries
interact synergistically with high-status origins:
H4-Alt. (Cumulative advantage alternative) Scholarly culture will espe-
cially enhance the educational performance of children whose parents
are high in the social hierarchy.
Specifically, there will be interaction effects such that the log number of
books will have especially large positive effects for children whose par-
ents are highly educated and/or of high occupational status.
Scholarly Culture Test Scores 7
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Data and Measurement
Data
Data are from the year 2000 round of the OECD’s Program for International
Student Assessment (PISA), conducted in 43 nations (OECD 2002; see also
Institute of Education Sciences 2012a,b). The survey instrument is highly regarded,
and its administration appears to be acceptable in most countries (Marks 2005).
Following Marks, we have omitted Japan because they did not include parents’
education. The samples are intended to be representative of 15-year-old students
in each country, although there are some uncertainties and differences in retention
to the relevant age so that the samples are biased toward higher-SES backgrounds
(Sikora and Saha 2010).10 A more troubling possibility is that the lust for national
prestige may have led one or two nations to select especially high-achieving stu-
dents from low socioeconomic backgrounds, which would produce downwardly
biased estimates for parents’ SES together with upwardly biased means.
We used multiple imputation of missing data separately for each society, fol-
lowing the general approach of King et al., in practice a regression-based tech-
nique augmented by a random component to the imputed value (King et al.
2001). These procedures have desirable properties when data are missing at ran-
dom (MAR), as is reasonable to assume here, and perform well in simulations
(Allison 2000; Schafer 1997). Given our large samples, King et al’s attractive
software was impractical, so we used IVEware from the Survey Research Center
at the University of Michigan (Raghunathan, Solenberger, and van Hoewyk
2004). Alternative imputations using MICE (Royston 2004), choosing options
to make it estimate models similar to King’s, lead to essentially identical results.
Omitted Variables
Unfortunately, PISA has no measure of academic ability, although it could have
been included (Baumert et al. 2009). Australian data suggest that including abil-
ity would reduce home library’s effect on years of education and on school marks
by about one-quarter (Evans et al. 2010). Since Australia is in no way atypical
(appendix table 1, panel 2), we tentatively assume, pending future research, that
the bias would be similar in other nations and that the same magnitude of bias
applies to performance.
PISA also lacks data on personality. However, a study in the Netherlands
including standard multiple-item measures of the “big five” personality traits
finds that adding them increases variance explained but leaves family back-
ground effects unchanged (van Eijck and de Graaf 2001). Since the Netherlands
is not atypical (appendix table 1, panel 2), we assume that the same holds for
other countries.
Measurement
We measure academic performance by PISA’s combined reading scale, which
includes subscales focused on retrieving information, interpreting texts, and
8 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

reflection and evaluation. This is the most reliable of PISA’s achievement
measures, .93, and is highly correlated with science and mathematics achieve-
ment (Marks 2005). The PISA achievement tests are well suited for testing our
hypotheses because they are oriented toward learning and cognitive skills rather
than toward specific curricula that form the focus of some alternative interna-
tional data sets (Institute of Education Sciences 2012a).
For clarity, we express performance as grade equivalents. PISA reports per-
formance in an arbitrary metric, scaled to a mean of 500 and standard deviation
of 100. Without loss of generality, we rescale these arbitrary scores. PISA stu-
dents, all 15 years old, are almost all in either the ninth or tenth grade. In PISA’s
metric, the ninth graders average 460 points (rounded) while the tenth graders,
having completed an extra year of school, average 503. Thus, an extra year of
schooling increases performance by 43 points on average. We take that 43 as our
yardstick. For example, if one student scores 20 points higher than another, we
report that as being about half a grade equivalent higher (viz. 20/43 = .47 grade
equivalents). Nothing is lost in this way: Multiplying our figures by 43 gives
PISA style points. But there is a gain in clarity by putting things into the familiar
metric of a year’s worth of schooling.11
We measure parents’ scholarly culture using the number of books in the student’s
family home (home library size). This facilitates comparison with prior research on
educational attainment. The number of books is conceptually central and is well
correlated with other aspects of scholarly culture (Evans et al. 2010). Following
pioneering work in the Netherlands (de Graaf 1986), research in Australia and
Britain also found that home library size is highly correlated with other aspects
of scholarly culture, including how often parents read “serious novels or poetry”;
read science, mathematics, or technology; read “other serious books like history or
biography”; and how often they went to the library (Evans and Kelley 2002a). It is
clearly distinct from arts spectatorship involving attendance at drama, art muse-
ums, classical music, and dance performances (Crook 1997b; de Graaf, de Graaf,
and Kraaykamp 2000; Sullivan 2001). In the PISA data, parents’ home library size
is correlated r = .38 with having computers, calculators, and educational software;
r = .19 with recreational reading and library use; r = .20 with having a diction-
ary; and r = .22 with a measure of family interest in political and social issues.
Australian panel data show that respondents report home library size reliably: Test-
retest reliability over a five-year period is high, r = .76 (Evans et al. 2010).
Information on the kinds of books in the home library is not available in
PISA: PISA’s great strengths are the range of countries included and the fact that
it allows us to extend the analysis to effects on performance as well as on attain-
ment. Thus, the effects we observe are an average of the effects of different kinds
of books. It seems likely that books of higher quality would have greater effects,
but that remains a matter for future research.
In the analysis, we mainly use the natural log of the number of books rather
than a simple count. The reasons for this are set out in detail in the hypotheses
above, with empirical substantiation in the results section.
Measurement of individual-level control variables largely follows prior research.
Averaging mother’s and father’s years of education is the single best representation
Scholarly Culture Test Scores 9
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

of parents’ education for stratification-related outcomes (Korupp, Ganzeboom,
and van der Lippe 2002; Treiman, Ganzeboom, and Rijken 2003), so we use it
here. Similarly, parents’ occupational status averages mother’s and father’s occu-
pational status. The four-digit ISCO 1988 code was recoded into Treiman’s (1977)
14 nominal categories and thence into Worldwide Status Scores (Kelley 1990).
A plausible alternative, ISEI scoring (Ganzeboom and Treiman 1996), produces
equivalent results (Evans et al. 2010). Parents’ wealth is the sum of six z-scored
possessions: dishwasher, separate bedroom, number of telephones, number of
TVs, number of cars, and number of bathrooms. Average inter-item correlation
in this index is r = .28, and alpha reliability = .68. These explanatory variables are
based on student reports, so it is noteworthy that a careful study shows that youth
as young as 11 years old report parents’ occupations accurately (West, Sweeting,
and Speed 2001). Developing nations are those with GDP per capita less than 50
percent of the United States in the year 2000, based on World Bank estimates.
Methods
First, prior research finds that, in a wide range of countries, the relationship
between parental home library size and the number of years of education the
respondent has completed is linear when library size is transformed by the natu-
ral log (i.e., the natural log of the number of books has a linear relationship with
educational attainment; see Evans et al. [2010]). However, as described above,
the elite closure theory has a very different prediction about the functional form
of the relationship between home library size and reading test score, anticipat-
ing that the relationship will be relatively flat across the small home library sizes
and then will kick up sharply, as elite entrance becomes a serious possibility.
Accordingly, to assess the functional form of the relationship, we use confidence
bands around a fractional polynomial fitted line—very much like a piecewise
regression (Royston, Ambler, and Sauerbrei 1999). This is a bivariate analysis
with no controls, and is shown in figure 1.
For hypotheses 2 and 3, we use OLS regression, which provides the best linear
unbiased estimates for models of this sort. We focus on first differences based on
the metric coefficients to show the magnitude of the effect and allow comparisons
across countries. Because the functional form of the relationship between each of
our independent variables and reading scores has not been firmly established by
prior research, we include quadratic terms for each of the independent variables,
thereby allowing a flexible degree of curvature. About half of these quadratic
terms are statistically significant at p < .05, and a few are large (see appendix
table 2). The first differences in predicted values (from the estimates in appendix
table 2) in table 1 and figure 1 summarize the effects of the linear and quadratic
terms in a single number; they are closely akin to partial derivatives (King et al.
2001). We estimate them from a whole population standardization (Kelley and
Evans 1995);12 this makes comparisons using a common reference population,
here the pooled international sample.
In order to obtain a summary estimate across many nations of the association
between the number of books in the student’s home and the student’s reading
10 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

performance, and to account for differences due to socioeconomic develop-
ment, we also conducted a multilevel analysis (a random effects GLS regression
with country as the group variable) including the variables just described plus
national GDP, with results shown in appendix table 1 and figure 2. The goal of
the multilevel analysis is to develop a single best characterization of the world-
wide pattern and to assess whether the association between home library size
and score on the reading test is net of socioeconomic development, as well as net
of parents’ education, parents’ occupation, parents’ wealth, and gender.
We also present standardized effects from the multilevel model and from
sheaf coefficients (which give the standardized coefficients representing the com-
bined effects of groups of variables [Whitt 1986]). There are two sets of sheaf
coefficients given for the multilevel model. The first set shows the effects of all
the family background effects taken together. The second set “unpacks” family
background into its components (home library size, parents’ education, parents’
occupation, parents’ wealth).
Results
Functional Form (H1): Biggest Gains at the Top or at the Bottom?
The scholarly culture theory predicts that as home library size expands from
nothing to just a few, and then a few more, reading test performance will increase
very steeply. But as we reach large library sizes, further gains are harder to come
by, since each additional book provides fewer new skills and less new informa-
tion, so the relationship flattens out. Thus, the relationship is approximately lin-
ear in the log of library size (H1). By contrast, elite closure theory suggests that
the relationship should be fairly flat at the beginning, where there is no realistic
possibility of elite entry. But then it should get much steeper, once elite entry is a
possibility (H1-Alt). Figure 1 shows the evidence.
Worldwide, the evidence is strongly in favor of the scholarly culture theory:
The first few books give by far the greatest benefit (figure 1). There is no sign
of the flat beginning followed by a steep rise that is predicted by elite closure
theory.
Generality in Advanced Societies (H2)
Scholarly culture arguments predict that books in the home will enhance chil-
dren’s educational performance in all advanced nations, regardless of elite
orientation (H2). This prediction is supported by the finding that the effect is
statistically significant in all 42 countries, with their diverse array of elite ori-
entations (table 1). This holds in traditional strongholds of bourgeois culture
such as France and Germany. It holds throughout the freewheeling Anglo-Celtic
world. It holds even in nations with strong egalitarian/redistributive institutions
like Scandinavia and post-Communist Eastern and East-Central Europe.
The evidence is based on regressions controlling for parents’ education, par-
ents’ occupational status, family wealth, and gender. It focuses on first differ-
ences, which are more readily interpretable than metric regression coefficients
Scholarly Culture Test Scores 11
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

because of the quadratic terms that give our model flexibility (the metric coef-
ficients are in appendix tables 1 and 2).
Home library size has a statistically significant effect (at the .01 level or better)
on reading performance in all advanced societies (table 1; details in appendix
table 1). In most cases, the influence is large, as shown by the substantial stan-
dardized regression coefficients (column 1). Further, the magnitude estimates
(column 2) show that the gain from a 500-book home library rather than from
a one-book library is equivalent to the gain of somewhere between one or two
additional years in school (for Finland, the Netherlands, Canada, etc.) and three
additional years (for Hungary, Germany, the UK, etc.). In the United States, the
gain is about two-and-a-half years. These results are consistent with scholarly
culture theory.
These results are, however, contrary to elite closure theory’s prediction that
home library size matters in some nations but not in others (H2-Alt). For exam-
ple, elite closure arguments would predict no effect of home library size in
the more egalitarian advanced societies, anticipating that these elites would
prefer memberships signals not associated with 19th-century bourgeois cul-
ture. However, books have a statistically significant effect in all the famously
egalitarian Scandinavian countries (table 1). At the top, the effects in Sweden,
Norway, and Iceland are very similar to the United States (about 2.5 grade
equivalents; standardized coefficients of .21 to .24). The effect is rather smaller
in Denmark (1.9 grade equivalents; .19 standardized) and in Finland (1.6 grade
equivalents; .15 standardized). These are widely scattered within the range of
Figure 1. Panel A. Theories: Scholarly culture and elite closure. Predicted values
assuming elite entry is around 500 books. Mean reading score. Panel B. Results: Even a
few books improve reading skill. 95-percent confidence intervals. Mean reading score, no
controls
7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0
Reading performance (grade level)
0 50 100 250 500 750
Number of books in the family home
Elite closure theory
Scholarly culture
Predicted effect of books in the home on reading performance
12 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

rich Western societies with less redistributive political systems. Similarly, home
library size has significant effects throughout the post-Communist countries,
from those that retain strongly dirigiste and redistributive systems to those that
have transited to more market-oriented economies. The effects are similar in
nations with left-wing governments and those with right-wing governments.
In all, the effects differ in size but none is negligible. All this is consistent with
scholarly culture’s predictions (H2) and contrary to elite closure theory’s pre-
dictions (H2-Alt).
Effects in Developing Societies (H3)
Scholarly culture arguments imply that home libraries enhance children’s educa-
tional performance in developing societies as well as in advanced nations (H3).
This is because of scholarly culture’s intrinsic connection with cognitive com-
plexity and skills, important in schools throughout the world. In contrast, elite
closure arguments suggest that different signals of elite membership—especially
particularistic ties of kinship and ethnicity—would apply in developing societies
(H3-Alt).
The data show that developing countries (in italics in table 1) are in no way
unique. The largest effects are found in them (Hungary, the Czech Republic),
but so are the smallest (Indonesia, Brazil, Macedonia). Even in nations with the
smallest effects, books matter enough to be policy relevant, with the gain from
a 500-book home library equivalent to an additional three-quarters of a year of
schooling.
These results are consistent with scholarly culture theory. They are clearly
contrary to elite closure theory (H3-Alt).
Figure 1. continued
7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11 .0
Reading performance (grade level)
0 50100 250 500 750
Number of books in the family home
PISA data for 42 nations, year 2000. N= 214,561
Scholarly Culture Test Scores 13
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Table 1. Family Influences on School Achievement Tests in Each Nation Separately. First Differences Estimated by OLS from the Model of Appendix
Table 2. PISA data for 42 nations, year 2000a
Nation
(developing
nations in italics)
Books in home:
Standardized
effectc
Predicted gain in achievement: Grade equivalents (1st differences)b
R-squared Cases
Books in
home: 500 vs
none
Parents’
education:
12 years vs. 4
Parents’ status:
Professional vs.
farm
Parents’ wealth:
top 10% vs
bottom 10%
Hungary .30 3.4 1.9 1.6 –0.6 .28 4,599
Czech Republic .26 3.3 2.0 2.1 –0.5 .25 5,177
Austria .29 3.0 1.0 1.3 0.5 .20 4,240
Germany .25 2.9 1.4 2.0 0.8 .26 4,434
New Zealand .24 2.9 0.8 1.6 0.5 .18 3,116
United Kingdom .25 2.7 0.7 2.2 ns .19 8,290
France .28 2.6 0.5 1.8 0.5 .22 4,043
Spain .25 2.6 0.8 1.1 ns .19 5,555
Latvia .20 2.6 0.7 1.6 –1.3 .14 3,624
Luxembourg .26 2.5 0.7 2.2 0.6 .25 2,569
United States .25 2.5 ns 1.4 1.0 .21 2,963
Switzerland .24 2.5 1.3 2.0 0.4 .23 5,562
Sweden .23 2.5 0.8 2.1 0.5 .17 4,148
Iceland .21 2.5 1.4 0.9 –0.4 .15 3,138
Ireland .24 2.4 0.8 1.9 ns .16 3,679
Norway .22 2.4 1.4 1.7 0.8 .16 3,764
Portugal .24 2.4 ns 2.2 0.5 .21 4,308
Russia .22 2.4 1.2 1.3 –1.2 .17 6,274
14 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Bulgaria .23 2.2 1.0 2.1 –0.5 .26 3,967
Korea (South) .26 2.2 0.7 ns ns .13 4,626
Argentina .22 2.2 0.5 1.4 0.6 .22 3,395
Australia .19 2.1 0.7 2.0 0.7 .18 4,786
Belgium .22 2.1 1.5 2.4 0.3 .21 5,960
Albania .22 2.1 1.3 1.2 –0.8 .20 4,159
Peru .22 2.1 1.0 1.7 ns .23 3,906
Greece .19 2.1 0.6 1.1 ns .15 4,307
Italy .20 2.0 0.9 1.2 ns .15 4,742
Canada .18 2.0 0.7 1.7 0.5 .16 28,180
Liechtenstein .18 2.0 ns 1.4 ns .17 282
Denmark .19 1.9 1.7 1.0 ns .19 3,779
Netherlands .23 1.9 1.1 1.8 .11 .19 2,320
Poland .20 1.8 0.5 1.7 –1.1 .18 3,108
Hong Kong .21 1.8 0.6 ns –0.4 .10 4,177
Chile .19 1.7 0.9 1.8 0.5 .22 4,584
Mexico .18 1.6 0.8 0.9 0.4 .21 4,123
Finland .15 1.6 0.5 1.1 0.6 .15 4,539
Thailand .17 1.4 0.4 0.7 0.7 .16 4,680
Israel .13 1.4 0.8 1.4 1.0 .15 3,032
Romania .10 1.0 0.6 0.9 –0.7 .05 4,405
(Continued)
Scholarly Culture Test Scores 15
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Table 1. continued
Nation
(developing
nations in italics)
Books in home:
Standardized
effectc
Predicted gain in achievement: Grade equivalents (1st differences)b
R-squared Cases
Books in
home: 500 vs
none
Parents’
education:
12 years vs. 4
Parents’ status:
Professional vs.
farm
Parents’ wealth:
top 10% vs
bottom 10%
Indonesia .10 0.8 0.4 0.7 ns .08 5,318
Brazil .08 0.8 0.6 1.1 1.6 .18 4,306
Macedonia .07 0.7 1.9 1.6 –0.6 .21 3,980
Pooled: Multilevel
model (rho = .05;
42 nations)d
.21 2.2 0.7 1.6 0.2 .31 200,144
ns: Not significantly different from zero at p < .01, two-tailed. (.05 for Liechtenstein, which has a very small sample.)
aSource: OECD Program for International Student Assessment (PISA). This is the combined reading scale, combining subscales focused on retrieving
information, interpreting texts, and reflection & evaluation. It is the most reliable of PISA’s scales, .93, and is highly correlated with scales measuring
science achievement and mathematics achievement.
bExample. In row 1, if two Hungarian students come from similar families in parents’ education, parents’ occupation, and wealth, differing only in how
many books there were in the home, the student from a home with 500 books would typically perform at the 10th-grade level while the student whose
parents had no books would perform at only 6th- or 7th-grade level, a difference of 3.4 grade levels (row 1, column 2). Analogously, if students differ only
in parents’ education, a student whose parents have 12 years of education would typically perform at the 9th-grade level, while a student whose parents
had only 4 years of education would preformed at about the 7th-grade level, a difference of 1.9 grade levels (row 1, column 3).
cThe standardized effect of books is from the corresponding OLS model, and is therefore approximate.
dIn addition to the variables shown, GDP and its square are included in the multilevel model. Other things being equal, a student from a country as rich as
the United States would be expected to perform 2.0 grade levels better than a student from a very poor nation with a GPD per capita at 10 percent of the
US level.
16 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

The Worldwide Pattern
Scholarly culture arguments imply that home libraries improve children’s aca-
demic performance and should therefore increase success in any academically
demanding, meritocratic school system. They would not necessarily help in reli-
gious school systems that focus on memorization or personal beliefs; nor in
art, music, drama, or athletic institutions, or the like. But most schools in most
nations are academically demanding and mostly meritocratic. So, it is logical to
make an estimate of the usual worldwide pattern linking books in the home to
performance and to compare it with the other “usual suspects” in academic per-
formance. These include gender (girls generally do better than boys), the nation’s
level of development (students from rich countries generally do better than those
from poor nations), and other aspects of family background (parents’ education,
occupational status, and wealth, all of which are advantageous). The results are
in figure 2 and appendix table 1. To compare the impact of home library size
to other well-known influences, we summarize their effects using standardized
sheaf coefficients (as discussed in the methods section).13
First consider the big picture, combining all aspects of family background
(books, parents’ education, occupation, and wealth) to see their joint impact
on academic performance. Family matters hugely, as is well known. The fam-
ily background sheaf coefficient is a very substantial .36, far larger than gender
with its sheaf of .13 and slightly larger than the sheaf of .32 for the nation’s per-
capita GDP (figure 2, top panel).
Second, consider the relative importance of the different family background
influences (figure 2, bottom panel). Home library size is clearly the most impor-
tant of them, with a sheaf coefficient of .21. Parents’ occupational status follows
Figure 2. Relative importance of different influences on reading performance in 42 nations,
year 2000. Multi-level analysis (from appendix table 1)
.32
.13
.36
.32
.13
.01
.16
.08
.21
.00 .10 .20.30 .40
Country GNP
Female
Parents' wealth
Parents's occupational
status
Parents' education
Books in the family (ln)
IN DETAIL:
Country GNP
Female
Family background
OVERALL:
Standardized effect (sheaf coefficient)
Detail Overall
Scholarly Culture Test Scores 17
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

(.16), then parents’ education (.08), and finally, far below in last place, parents’
wealth (a mere .01). At this level of detail, a nation’s per-capita GDP is the larg-
est single influence.
Note, however, that per-capita GDP differences are not over the whole
range of nations, but mainly between poor nations (like Indonesia or Brazil)
and nations at Western European levels of development. Once a nation reaches
European levels of development, further increases—for example to the US level
or above—are no benefit to academic performance (figure 3). Indeed, they may
slightly harm performance. To put things in perspective, going from a family
with no books to one with a 500-book home library has about the same impact
on academic performance as migrating from Indonesia to the UK, or from Brazil
to France.
Thus, the impact of home library size is substantial, either in terms of the mag-
nitude of the gains, in terms of its relative importance compared to other long-
documented influences, or in comparison to gains from international migration.
H4 (Scholarly Culture Compensation: GGK Hypothesis)
Children from high-status families (university-educated parents with profes-
sional or administrative jobs) do indeed achieve higher academic test scores
when there are many books in the home: Those with 500 books perform half a
grade better than those with just 100 (appendix table 3). This modest advantage
is consistent with both elite closure and scholarly culture predictions (hypoth-
eses H4 and H4-Alt).
Figure 3. More developed nations do better, up to a point. PISA data for 42 nations, year 2000.
N = 200,144. 95-percent confidence intervals. Predicted values from multilevel regression
(from appendix table 1)
6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0
Reading (grade level)
0 Indonesia Brazil, World Korea France, UK USA
Development: GDP per capita (Index:USA=1)
18 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

In contrast, books matter hugely for children from low-status homes (par-
ents not high school graduates, working at semi-skilled or unskilled jobs, with
fewer than 100 books in the home). For those modest families, 75 books enable
children to perform one-and-a-half grades better than their peers from book-
less families, a huge difference. That is as predicted by the GGK hypothesis of
scholarly culture synergistic compensation (hypothesis H4). But it is entirely
inconsistent with elite closure arguments (hypothesis H4-Alt): No gatekeeper
is going to think that a child from a blue-collar family with parents who never
even finished high school comes from an elite family just because they have 75
books at home.
Summary
In sum, we find that books in the home have a positive “payoff” in improved
test scores throughout the world. Regardless of how many books the family
already has, each addition to the home library helps children do better on a
reading test that is carefully designed, comprehensive, structured to minimize
class and ethnic biases, and anonymously graded. Growing up in the scholarly
culture provides important academic skills.
But the gains are not equally great across the entire cultural hierarchy. Instead,
they are larger at the bottom, far below elite level: Each additional book has a
greater impact on the performance of someone who had only a small home
library than it does on the performance of someone from a home overflowing
with books. Thus, the second book and the third book have much larger impacts
than the 102nd or 103rd. This, too, is true not only for the sample as a whole,
but in each of the countries as well. This constitutes strong support for schol-
arly culture theory’s hypothesis H1 (biggest gains at the bottom) and is strongly
contrary to elite closure theory’s hypothesis H1-Alt (biggest gains at the top).
The relationship is strong, clear, and statistically significant in every one of the
42 countries. This is true even after controlling for well-known, potent family
influences on educational performance—parents’ education, parents’ occupa-
tion, and family wealth. Importantly, the effect is found regardless of the elite’s
ideology: in the egalitarian societies of Scandinavia and egalitarian-heritage
post-Communist societies, in rich market-oriented advanced societies, in the
new world as well as the old, in Asia as well as Europe and the Americas. This
evidence is consistent with scholarly culture theory’s hypothesis H2 (gains in
all advanced societies), and contrary to elite closure theory’s hypothesis H2-Alt
(gains depend on the elite’s choice of signals).
Moreover, books have a strong impact across the whole broad spectrum of
socioeconomic development, from poor nations to rich nations. The effect is
large and statistically significant in all the developing societies in the study; it
is very large in some of them. This is consistent with scholarly culture theory’s
hypothesis H3 (gains in developing societies, too) and contrary to elite closure
theory’s hypothesis H3-Alt (few gains in developing societies).
Importantly, books especially benefit children from disadvantaged families.
They enhance the academic performance of children from families at all educa-
Scholarly Culture Test Scores 19
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

tional and occupational levels, but the enhancement is greater for families with
little education and low-status occupations. This strong compensatory effect
of scholarly culture is as predicted by hypothesis H4 (the GGK compensation
hypothesis) and is contrary to expectations based on the cumulative advantage
hypothesis H4-Alt (the Matthew effect).
Implications
Scholarly Culture Theory versus Elite Closure Theory
All three tests in this paper (hypotheses H1, H2, and H3) support scholarly cul-
ture theory and undermine elite closure theory. Moreover, the multilevel analysis
provides a working hypothesis for future research on other countries or in other
times: On average, the difference in reading test scores between children from
bookless homes and homes with 500 books is as large as that to be expected
for an additional 2.2 years in school. These gains have to do with gains in aca-
demic performance on objective, internationally normed tests: Elite closure can-
not account for them. But they are perfectly consistent with the hypothesis that
scholarly culture endows children with education-related skills.
This estimate may be upwardly biased by PISA’s omission of an academic
ability/intelligence measure, as noted earlier. If the bias is of the same degree
as for educational attainment—which we can estimate in other data sets—it
would reduce books’ effect by about one quarter. That would leave it equiva-
lent to something over one-and-a-half years of schooling (2.2 × 0.75 = 1.65).
That is still a very substantial effect. Thus, the evidence remains strongly in
favor of the scholarly culture theory and strongly contrary to the elite closure
theory.
Note that these results on home library size in the educational domain do not
speak to the issue of elite closure in society at large, nor even to elite closure in
the educational domain linked to other arbitrary class signals (such as accent or
name), or to religion, ethnicity, politics, or the like. But for understanding the
links between culture and education, we should orient theoretical work toward
understanding how scholarly culture works, not toward elite exclusion: It would
be wiser to spend our energies on the horses that are still running rather than on
beating dead ones.
Cultural Mobility
Consider finally Dimaggio’s cultural mobility hypothesis that the elite pretend to
openness by allowing a small number of exceptional youth from the lower rungs
of society to enter their ranks. To the contrary, we have shown that the impact
of scholarly culture is mostly about ordinary success, not about elites. Homes
that read a little more nudge disadvantaged children a little further ahead, but
not into a different sphere. Home library size makes a clear, robust impact large
enough to be well worthy of policy attention but, in magnitude, it usually leads
to a small boost in status, not to elite entry.
20 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Evidence Issues
As with all evidence, the evidence we use here is imperfect. (1) The scope is
not yet universal, because only one or two dominantly Muslim countries are
included and because the only sub-Saharan nation included is South Africa, with
its atypical wealth and even more atypical social history. (2) Ultimately, research
needs to replicate the analysis using other performance measures to ensure that
the findings are not specific to the PISA tests (other rounds of PISA already
cover science and mathematics). However, the risk that our findings are specific
to PISA seems low, given how strongly they echo prior findings for educational
attainment. (3) Repeating the analysis on a data set with an academic ability
measure (inexplicably omitted from PISA) would be valuable. (4) Information
on the quality of books available in the home library should be included in
future studies. The information to hand does not show whether any book will
do, or whether high-quality books enhance performance more strongly. That
seems likely, but evidence is needed.
Future Research
It would be valuable to replicate these analyses on additional countries, particu-
larly Muslim and sub-Saharan nations; Egypt and Nigeria would be particularly
valuable. There are also new questions that can be addressed in existing data.
For example, how much do the children of cultured families cluster in particu-
lar schools and how much does that matter? And what about social climate—
does the scholarly culture of the time and place where children grew up matter?
In particular, does growing up in bookish times and places boost the achieve-
ment of children from bookless homes? It would also be useful to consider other
educational outcomes. For example, it may be that children from families with
small home libraries are more likely to suffer summer setback (Vale et al. 2012).
Answers to these questions will also help us understand how the scholarly cul-
ture works.
Policy Implications
While the association between home library size and academic performance
is clear, there are still questions about the mechanisms involved. In particular,
to what degree is the scholarly culture of the family the driver, with the num-
ber of books merely representing the scholarly culture way of life (in which
case, books are not an especially effective policy lever) and to what degree is it
the books themselves? Some experimental research placing books in children’s
hands suggests that books themselves do matter, raising reading scores especially
for children from disadvantaged homes (Allington et al. 2010; Kim 2006). If
so, a strong policy recommendation in favor of book drives is justified. More
speculatively, it could also be that providing children’s books to young mothers
would help start scholarly culture growing in the family. In all, it is clear that
scholarly culture is vastly important to educational performance. So, anything
Scholarly Culture Test Scores 21
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

that helps scholarly culture flourish is likely to help children, especially those
from disadvantaged families, do better in school.
Notes
1. Unless otherwise noted, the findings described in this section are from multivariate
models controlling for family background factors, particularly parents’ education
and occupation. Parents’ income is rarely included because it is less reliably mea-
sured and generally has modest effects, if any, on educational outcomes (e.g., Fejgin
1995; Ganzach 2000; Mayer 1997; Marks 2005). Hence, these findings are probably
not distorted by omitted-variable bias.
2. Nomenclature in this area is varied and confusing. What we call “elite closure” (but
is often called “cultural capital”) focusing on elite discrimination is far removed
from the natural interpretation of “cultural capital” as skills and knowledge analo-
gous to “human capital.” The term “cultural capital” has been operationalized in
both elite closure mode and much more broadly to include virtually any cultural
features that might enhance education (Andersen and Hansen 2012).
3. We describe skills and information useful in schoolwork as “scholarly culture” rather
than as “human capital” because economists’ usage of “capital” has the implication
that it is a “cost” to the individual (usually in income forgone), whereas we posit it
as a normal part of the scholarly culture lifestyle. Reading for pleasure, acting out
stories from favorite books, shared humorous literary references, playing charades
on a winter’s night—all are intrinsically rewarding, even if they also enhance reading
skills, vocabulary, knowledge, and other skills that are rewarded in school. Thus, the
scholarly culture hypothesis suggests a much less instrumental parenting approach
than “human capital” or “concerted cultivation” (Lareau 2011).
4. Empirical research unequivocally exonerates teachers from this gatekeeper role,
revealing instead persistent attempts to compensate for deficits in students’ home
cognitive environment (Portes, Fernandez-Kelly, and Haller 2009).
5. A committed social-closure theorist could claim that educational systems themselves
are in part signaling mechanisms, but the evidence is against it (Evans and Kelley
2002b).
6. This is a slightly different approach to hypothesis 2a tested by de Graaf, de Graaf,
and Kraaykamp (2000) for educational attainment in the Netherlands.
7. By way of illustration, an experimental book distribution program found that the
enhancements in cognitive skill were significantly greater for children from book-
scarce homes than for their peers from homes rich in books (Kim 2006). The eight
books provided by the program to fourth graders—including some from the Captain
Underpants series—were surely not providing signals of elite membership.
8. There is an implicit assumption here that insofar as elite membership is signaled by
reading, large home libraries are the likely source—the larger the home library, the
more likely that the complex, demanding books that are plausible signals of elite
status are found there. Until the home library is well past the basic level, there is not
even a question of signals of elite membership.
9. De Graaf et al. generously associated this important finding with Dimaggio’s “cultural
mobility hypothesis,” (Dimaggio and Mohr 1985) but as that hypothesis concerns
elite entry and their finding concerns mobility from extremely low-status positions to
middle-status positions, it will be clearer simply to dub it the GGK hypothesis.
10. Careful comparisons with benchmark data in several countries suggest that some
PISA samples over-represent students from high-SES families (Ferreira and Gignoux
22 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

2010). This limits the generalizability of descriptive statistics, but multivariate results
should be representative, albeit with caveats, since SES variables are included in our
models.
11. The computation is: Grade_equivalent = (PISA_score – 502.5) / (502.5 – 459.8),
where 502.5 is the mean PISA score of tenth graders and 459.8 that for ninth grad-
ers. The subtraction in the numerator just centers the score at the performance of
the average tenth grader. Thus, a score of 0 would be performance at the level of an
average tenth grader; a score of –1 would be performance at the ninth-grade level,
and –2 at the eighth-grade level; analogously, a score of +1 would be performance at
the 11th-grade level and +2 at the 12th-grade level. The gains above the tenth-grade
level and the losses below ninth grade are approximate, assuming that progress from
grade to grade is always at the 43 PISA points per year we observe between ninth
and tenth grades.
12. For nonlinear equations, the standard predicted value for an artificial case with aver-
age characteristics may be far from the average and the slopes can vary greatly from
person to person. So, there is no simple summary. What we have done, a “whole
population standardization,” is to calculate predicted values for every person in the
sample and average those figures. The results depend both on the equation and on
the population chosen as a baseline for comparison, and are analogous to econo-
mists’ “average derivative.”
13. These range in theory from –1.0 (indicating a hugely important negative effect) to
0 (indicating no effect) to 1.0 (a hugely important positive effect). In practice, for
individual-level data where the variables are not related by definition, the actual
range seems to be roughly –0.5 to 0.5.
Scholarly Culture Test Scores 23
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Appendix Table 1. Influences on reading performance (measured in grade equivalents).
Multi-level regressions. PISA 2000
Coefficient (1)
Standard
error (2) z (3)
Panel 1: Main model
Number of books in parents’ home (ln) 0.31 0.003 94.3
Parents’ education (years) 0.07 0.002 40.1
Parents’ education, squared 0.00 0.000 −8.5
Parents’ occupational status (index, 0 to 1) 1.56 0.023 67.3
Parents’ status, squared −0.49 0.073 −6.7
Wealth (possessions index, 0 to 1) 0.24 0.019 12.5
Wealth index, squared 0.06 0.050 1.2
GDP per capita at PPP (index; USA =1) 2.73 0.268 10.2
GDP, squared −5.23 0.870 −6.0
Female (dummy variable, female = 1) 0.63 0.009 71.3
Intercept 5.17 0.161 32.2
(N of nations) 42 – –
(N of individuals) 200,144 – –
(R-squared) 0.31 – –
Panel 2: Are Australia or the Netherlands distinctive?a
(All variables of Panel 1 are included but
not shown)
– – –
Australia (dummy variable, Australia =1) 0.14 0.470 0.3
Netherlands (dummy, Netherlands =1) 0.79 0.471 1.7
(R-squared) 0.31 – –
All coefficients in Panel 1 are statistically significant at p < .001 two-tailed, except the ones
shown in italics. In Panel 2 neither of the country dummy variables is significant even at
p < .05 .
aN = 5176 for Australia and 2503 for the Netherlands.
24 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Appendix Table 2. Family influences on school achievement tests (reading grade level) in each nation separately. Metric regression coefficients and
robust standard errors. These are used to calculate the results in Table 1. PISA data for 42 nations, year 2000
Country (and UN
country code)
Books
in home
(ln)
Parents’
education
Education
squared
Parents’
status
Status
squared
Wealth
(index)
Wealth
squared Female Intercept R-squared
N of
cases
8. Albania 0.3045 0.1114 −0.0142 1.2460 −1.4208 −0.7650 0.4980 0.8440 4.1180 0.20 4,159
0.0221 0.0143 0.0035 0.1639 0.4628 0.3195 0.6540 0.0616 0.2079
32. Argentina 0.3115 0.0585 −0.0010 1.3725 −1.8056 0.6321 0.4384 0.8634 5.6656 0.22 3,395
0.0283 0.0129 0.0025 0.2144 0.5886 0.1625 0.4308 0.0724 0.1399
36. Australia 0.3086 0.1111 0.0049 2.0214 −0.7430 0.6763 −0.7030 0.7122 6.0349 0.18 4,786
0.0249 0.0184 0.0030 0.1749 0.5384 0.1550 0.4156 0.0602 0.2104
40. Austria 0.4372 0.0702 −0.0129 1.2814 1.0211 0.5116 −1.4012 0.6039 6.2930 0.20 4,240
0.0253 0.0140 0.0026 0.1771 0.5694 0.1458 0.4258 0.0594 0.1507
56. Belgium 0.3075 0.0754 −0.0281 2.4249 −1.2807 0.2857 −1.4420 0.5540 7.1768 0.21 5,960
0.0193 0.0112 0.0020 0.1489 0.4614 0.1127 0.3552 0.0529 0.1343
76. Brazil 0.1107 0.0696 −0.0018 1.0517 −0.8154 1.6177 1.1201 0.3925 5.6667 0.18 4,306
0.0215 0.0138 0.0019 0.1867 0.6046 0.1581 0.3803 0.0580 0.1409
100. Bulgaria 0.3248 0.1758 0.0131 2.0995 0.0886 −0.5139 −1.2627 0.8433 3.5444 0.26 3,967
0.0214 0.0246 0.0042 0.1865 0.6235 0.2049 0.4966 0.0634 0.2875
124. Canada 0.2957 0.0732 −0.0024 1.6989 −0.2471 0.5193 −0.0595 0.7401 6.6210 0.16 28,180
0.0099 0.0062 0.0010 0.0617 0.1975 0.0626 0.1670 0.0244 0.0755
152. Chile 0.2473 0.0993 −0.0043 1.8293 −1.1987 0.4897 0.8985 0.4744 5.1396 0.22 4,584
0.0204 0.0129 0.0022 0.1735 0.4510 0.1450 0.3555 0.0532 0.1451
203. Czech
Republic
0.4769 0.2328 −0.0058 2.1005 −1.6467 −0.5238 −0.8796 0.5933 3.6045 0.25 5,177
0.0272 0.0253 0.0036 0.1820 0.5429 0.1626 0.4215 0.0535 0.2958
(Continued)
Scholarly Culture Test Scores 25
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Appendix Table 2. continued
Country (and UN
country code)
Books
in home
(ln)
Parents’
education
Education
squared
Parents’
status
Status
squared
Wealth
(index)
Wealth
squared Female Intercept R-squared
N of
cases
208. Denmark 0.2750 0.2070 −0.0018 1.0344 1.0319 0.0663 −0.5344 0.5941 5.2519 0.19 3,779
0.0248 0.0178 0.0027 0.1884 0.6619 0.1620 0.4481 0.0647 0.2098
246. Finland 0.2319 0.0458 −0.0031 1.1043 0.5448 0.574 −1.0644 1.0886 8.0937 0.15 4,539
0.0248 0.0122 0.0034 0.1497 0.4776 0.1277 0.3977 0.0569 0.1480
250. France 0.3818 0.0338 −0.0069 1.7989 −1.1943 0.5419 −0.8376 0.4961 6.9933 0.22 4,043
0.0212 0.0132 0.0023 0.1618 0.4935 0.1244 0.3934 0.0586 0.1465
276. Germany 0.4227 0.1627 −0.0032 1.9677 −1.8017 0.8339 −1.7473 0.6893 4.4939 0.26 4,434
0.0262 0.0151 0.0024 0.1751 0.5420 0.1614 0.4338 0.0601 0.1844
300. Greece 0.2978 0.0792 0.0011 1.1199 0.1667 −0.2061 −1.1370 0.7973 6.5461 0.15 4,307
0.0248 0.0129 0.0029 0.1514 0.4363 0.1327 0.4086 0.0623 0.1464
344. Hong Kong 0.2571 0.0930 0.0032 0.4286 −0.4017 −0.4253 −1.6602 0.2414 8.9106 0.10 4,177
0.0205 0.0138 0.0022 0.1918 0.6688 0.1928 0.5056 0.0565 0.1362
348. Hungary 0.4879 0.2139 −0.0072 1.6116 0.0924 −0.5517 0.1377 0.5918 3.8666 0.28 4,599
0.0269 0.0294 0.0066 0.1882 0.5244 0.1401 0.3667 0.0526 0.2858
352. Iceland 0.3590 0.1299 −0.0117 0.8685 0.7395 −0.3785 −0.9148 0.7374 6.4333 0.15 3,138
0.0320 0.0172 0.0029 0.1815 0.5839 0.2118 0.5405 0.0703 0.2649
360. Indonesia 0.1147 0.0595 0.0026 0.6775 −1.4937 −0.2067 −0.5268 0.3062 6.1271 0.08 5,318
0.0165 0.0112 0.0015 0.1194 0.3040 0.3611 0.7087 0.0460 0.2174 0.00 0
372. Ireland 0.3515 0.0211 −0.0185 1.9462 0.3972 0.1334 0.3618 0.5905 7.6661 0.16 3,679
0.0250 0.0145 0.0039 0.1772 0.5987 0.1293 0.3912 0.0652 0.1534
26 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

376. Israel 0.1960 0.1275 0.0078 1.3570 −1.3497 0.9710 −0.5467 0.2830 5.5963 0.15 3,032
0.0285 0.0183 0.0036 0.2523 0.6650 0.1252 0.4124 0.0759 0.2260
380. Italy 0.2959 0.0707 −0.0113 1.2388 −0.0244 0.2048 −0.7461 0.6915 6.7973 0.15 4,742
0.0225 0.0124 0.0027 0.1795 0.5306 0.1320 0.3946 0.0565 0.1194
410. Korea,
Republic of
0.3134 0.0801 −0.0010 0.2350 0.0671 −0.2303 −0.4871 0.3936 7.9967 0.13 4,626
0.0196 0.0100 0.0021 0.1199 0.4048 0.1235 0.3646 0.0462 0.1273
428. Latvia 0.3699 0.0810 −0.0026 1.5525 −1.8798 −1.3346 −2.1111 0.9060 5.6981 0.14 3,624
0.0314 0.0330 0.0052 0.1787 0.5982 0.3756 0.7613 0.0716 0.4365
438.
Liechtenstein
0.2823 0.1397 0.0011 1.4131 −2.5703 1.0927 −1.6988 0.2106 5.7728 0.17 282
0.1069 0.0681 0.0119 0.6194 2.1309 0.6590 1.7571 0.2294 0.6867
442.
Luxembourg
0.3649 0.0544 −0.0079 2.1615 −1.1832 0.6360 −1.6768 0.5916 5.5435 0.25 2,569
0.0298 0.0124 0.0020 0.2207 0.7023 0.2093 0.5349 0.0787 0.1645
484. Mexico 0.2351 0.1039 0.0005 0.9143 −1.2773 0.4456 −0.0002 0.4002 6.1609 0.21 4,123
0.0214 0.0145 0.0017 0.1837 0.4890 0.1466 0.3597 0.0570 0.1334
528. Netherlands 0.2679 0.0778 −0.0141 1.7525 −0.4271 0.1153 −1.0104 0.5548 7.9520 0.19 2,320
0.0252 0.0188 0.0033 0.2254 0.7497 0.1837 0.5503 0.0758 0.1860
554. New
Zealand
0.4213 0.0776 −0.0056 1.6161 0.4297 0.4834 −0.0600 0.9394 6.2205 0.18 3,116
0.0330 0.0155 0.0024 0.1968 0.6340 0.1625 0.4934 0.0772 0.2118
578. Norway 0.3488 0.1165 −0.0159 1.6688 0.7814 0.7579 −2.2722 0.7606 5.5512 0.16 3,764
0.0286 0.0195 0.0030 0.1891 0.5889 0.2190 0.5414 0.0703 0.2496
604. Peru 0.3039 0.1378 0.0038 1.6773 −2.0320 0.5535 0.9074 0.1740 3.3346 0.23 3,906
0.0230 0.0132 0.0028 0.1822 0.5349 0.2171 0.4948 0.0631 0.1697
616. Poland 0.2665 0.1167 0.0142 1.7172 −0.8820 −1.0824 −1.4099 0.6740 6.2898 0.18 3,108
0.0243 0.0526 0.0099 0.2408 0.6988 0.2130 0.5255 0.0736 0.5878
(Continued)
Scholarly Culture Test Scores 27
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Appendix Table 2. continued
Country (and UN
country code)
Books
in home
(ln)
Parents’
education
Education
squared
Parents’
status
Status
squared
Wealth
(index)
Wealth
squared Female Intercept R-squared
N of
cases
620. Portugal 0.3478 −0.0036 −0.0038 2.2020 −1.1363 0.5036 −1.2010 0.5115 6.9342 0.21 4,308
0.0249 0.0120 0.0023 0.1780 0.5348 0.1241 0.3530 0.0594 0.1097
642. Romania 0.1504 0.0657 −0.0030 0.8814 −1.8467 −0.7302 −1.5764 0.2609 7.5104 0.05 4,405
0.0264 0.0178 0.0045 0.1853 0.5138 0.2221 0.5201 0.0652 0.2218
643. Russian
Federation
0.3403 0.0810 −0.0167 1.9936 −0.1131 −1.1736 −1.9243 0.7185 5.8710 0.17 6,274
0.0201 0.0230 0.0044 0.1282 0.3906 0.2492 0.5035 0.0495 0.2925
724. Spain 0.3717 0.0711 −0.0077 1.0445 −0.6208 −0.0374 −0.1947 0.4916 6.9361 0.19 5,555
0.0221 0.0112 0.0020 0.1450 0.4231 0.1019 0.3005 0.0478 0.1166
752. Sweden 0.3642 0.0326 −0.0163 2.1250 0.3879 0.5010 −0.7206 0.7032 6.6292 0.17 4,148
0.0263 0.0183 0.0039 0.1779 0.5981 0.2013 0.4682 0.0602 0.2175
756. Switzerland 0.3634 0.1195 −0.0113 2.0354 −1.6008 0.4184 −1.4788 0.5258 5.7523 0.23 5,562
0.0223 0.0138 0.0028 0.1462 0.4448 0.1175 0.3457 0.0535 0.1411
764. Thailand 0.2084 0.0695 0.0047 0.6614 −1.3846 0.6836 0.2370 0.7700 6.6576 0.16 4,680
0.0193 0.0141 0.0023 0.1758 0.4801 0.1543 0.3756 0.0523 0.1436
807. Macedonia 0.0981 0.2080 −0.0076 1.6096 −0.0717 −0.6278 −0.2998 0.9150 3.7753 0.21 3,980
0.0237 0.0134 0.0031 0.1709 0.5152 0.1487 0.4007 0.0586 0.1634
826. United
Kingdom
0.3917 0.0575 −0.0076 2.2192 −0.2480 0.2826 −0.3914 0.5331 6.7772 0.19 8,290
0.0175 0.0112 0.0019 0.1264 0.4320 0.1113 0.3081 0.0448 0.1311
840. United
States
0.3638 0.0359 −0.0005 1.4257 0.1122 1.0352 1.2138 0.5773 6.1197 0.21 2,963
0.0275 0.0190 0.0032 0.2123 0.6522 0.2642 0.6356 0.0746 0.2304
Source: OECD Program for International Student Assessment (PISA). This is the combined reading scale, combining sub-scales focused on retrieving
information, interpreting texts, and reflection & evaluation. It is the most reliable of PISA’s scales, .93, and is highly correlated with scales measuring
science achievement and mathematics achievement.
28 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Appendix Table 3. Interactions with class or status: Influences on reading performance
(measured in grade equivalents). Multi-level regression. PISA 2000
Parents are high ranking:
Both university educated,
in administrative or
professional jobs, with
over 100 books in the
home
Parents are low ranking:
Did not graduate from
high school, work in
semi-skilled or unskilled
jobs, have fewer than 100
books
Coefficient
(1)
Standard
error (2)
Coefficient
(3)
Standard
error (4)
Panel 1. Multi-level regression
Number of books in parents’
home (ln)
0.30 0.032 0.29 0.009
GDP per capita at PPP (index;
USA =1)
1.82 0.550 2.23 0.397
Female (dummy variable,
female = 1)
0.65 0.037 0.52 0.023
Intercept 7.57 0.399 5.95 0.257
(N of nations) 42 – 42 –
(N of individuals) 12,148 – 29,159 –
(R-squared) 0.09 – 0.20 –
(rho) 0.20 – 0.14 –
Panel 2. Predicted valuesa
Grade
equivalent
Standard
error
Grade
equivalent
Standard
error
No books in parents’ home – – 7.05 0.173
25 books in parents’ home – – 8.19 0.170
50 books in parents’ home – – 8.39 0.170
75 books in parents’ home – – 8.51 0.171
100 books in parents’ home 10.41 0.231 – –
250 books in parents’ home 10.69 0.228 – –
500 books in parents’ home 10.90 0.228 – –
750 books in parents’ home 11.02 0.228 – –
All coefficients are statistically significant at p < .001 two-tailed.
aWhole population standardization, using the sample analyzed as the reference group.
Scholarly Culture Test Scores 29
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

About the Authors
M. D. R. Evans, Professor of Sociology at the University of Nevada, coordinates
the Applied Statistics Program and mentors students in the Social Psychology
PhD program. As a research duo, Evans and Jonathan Kelley study status
attainment (“Family Scholarly Culture. . .,” RSSM 2010) and bioethics (Nature
Biotechnology 2011). They teamed with Clayton Peoples on inequality attitudes
(Social Science Quarterly 2010). Evans and Kelley also conduct research with
additional collaborators. Evans studies environmental and natural resource
sociology with Kimberly Rollins.
Jonathan Kelley directs the World Inequality Study (43 nations; 134 sur-
veys; N = 800,000; www.international-survey.org) and is Adjunct Professor of
Sociology, University of Nevada. He does duo work with Evans; with Krzysztof
Zagorski, Nate Breznau, C. G. E. Kelley, and S. M. C. Kelley, they study wom-
en’s employment (IJPOR 2009), religion (JSSR 2011), attitudes toward gays,
and subjective well-being (Social Indicators Research 2013). Kelley recently
began a major project on the social experience of climate change.
Joanna Sikora is Senior Lecturer in Sociology at Australian National University.
Her research interests involve educational inequalities, social stratification, and
mobility in comparative perspective, including adolescents’ occupational expec-
tations. She has authored recent articles with A. Pokropek on gender segregation
of science career plans (Science Education 2012) and gendered inheritance of sci-
ence career preferences (International Journal of Science Education 2012); and
with L. Saha on talent loss (International Journal of Contemporary Sociology
and Educational Practice and Theory).
References
Allington, Richard L., Anne McGill-Franzen, Gregory Camilli, Lunetta Williams, Jennifer Graff, Jacqueline
Zeig, Courtney Zmach, and Rhonda Nowak. 2010. “Addressing Summer Reading Setback among
Economically Disadvantaged Elementary Students.” Reading Psychology 31:411–27.
Allison, Paul D. 2000. “Multiple Imputation for Missing Data.” Sociological Methods and Research
28:301–9.
Andersen, Patrick L., and Marianne N. Hansen. 2012. “Class and Cultural Capital.” European Sociological
Review 28:607–21.
Baumert, Juergen, Oliver Luedtkea, Ulrich Trautweina, and Martin Brunner. 2009. “Large-Scale Student
Assessment Studies Measure the Results of Processes of Knowledge Acquisition.” Educational
Research Review 4:165–76.
Bidwell, Charles E. 1989. “The Meaning of Educational Attainment.” Research in Sociology of Education
and Socialization 8:117–38.
Blau, Peter M., and Otis Dudley Duncan. 1967. The American Occupational Structure. New York: Free Press.
Bourdieu, Pierre. 1984. Distinction. Cambridge, MA: Harvard University Press.
______. 1986. The Forms of Capital. New York: Greenwood Press.
Bus, Adriana G., and Marinus H. van Ijzendoorn. 1995. “Joint Book Reading Makes for Success in Learning
to Read.” Review of Educational Research 65:1–21.
Byun, Soo-yong, Evan Schofer, and Kyung-keun Kim. 2012. “Revisiting the Role of Cultural Capital in East
Asian Educational Systems.” Sociology of Education 85:219–39.
30 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Crook, Christopher J. 1997a. Cultural Practices and Socioeconomic Attainment. Westport, CT: Greenwood
Press.
______. 1997b. “The Dimensionality of Stratification-Related Cultural Practices in Australia.” Australian
and New Zealand Journal of Sociology 33:226–38.
de Graaf, Nan Dirk, and Paul M. de Graaf. 2001. “Formal and Popular Dimensions of Cultural Capital.”
Netherlands Journal of Social Sciences 38:167–86.
de Graaf, Nan Dirk, Paul M. de Graaf, and Gerbert Kraaykamp. 2000. “Parental Cultural Capital and
Educational Attainment in the Netherlands.” Sociology of Education 73:92–111.
de Graaf, Paul M. 1986. “The Impact of Financial and Cultural Resources on Educational Attainment in the
Netherlands.” Sociology of Education 59:237–46.
Dimaggio, Paul, and J. Mohr. 1985. “Cultural Capital, Educational Attainment, and Marital Selection.”
American Journal of Sociology 90:1231–61.
Dronkers, Jaap. 1992. “Parents, Love, and Money.” International Perspectives on Education and Society
2:277–93.
Evans, M. D. R., and Jonathan Kelley. 2000. “The Beaux Arts versus Scholarly Culture.” Australian Social
Monitor 3(2):41–48.
______, eds. 2002a. “Cultural Resources and Educational Success.” In Australian Economy and Society
2001: Education, Work, and Welfare, 54–65. Sydney: Federation Press.
______, eds. 2002b. “Why Is Education Rewarded?” In Australian Economy and Society 2001: Education,
Work, and Welfare, chapter 2. Sydney: Federation Press.
Evans, M. D. R., Jonathan Kelley, Joanna Sikora, and Donald J. Treiman. 2010. “Family Scholarly Culture
and Educational Success.” Research in Social Stratification and Mobility 28:171–97.
Farkas, George. 1996. Human Capital or Cultural Capital? New York: De Gruyter.
Fejgin, Naomi. 1995. “Factors Contributing to the Academic Excellence of American Jewish and Asian
Students.” Sociology of Education 68:18–30.
Ferreira, Francisco H. G., and Jeremie Gignoux. 2010. “Educational Inequality and Its Intergenerational
Persistence.” In 31st IARIW Conference, edited by IARIW. St. Gallen, Switzerland.
Ganzach, Yoav. 2000. “Parents’ Education, Cognitive Ability, Educational Expectations, and Educational
Attainment.” British Journal of Educational Psychology 70:419–41.
Ganzeboom, Harry B. G., Paul M. de Graaf, and Peter Robert. 1990. “Cultural Reproduction Theory on
Socialist Ground.” Research in Social Stratification and Mobility 9:79–104.
Ganzeboom, Harry B. G., and Donald J. Treiman. 1996. “Internationally Comparable Measures of
Occupational Status.” Social Science Research 25:201–39.
Goblot, Edmond. 1925 [1973]. “Cultural Education as a Middle-Class Enclave.” In The Sociology of
Literature and Drama, edited by Tom Burns, 433–44. Harmondsworth, UK: Penguin.
Goldthorpe, John H. 2007. “Cultural Capital.” Acta Sociologica 50:211–29.
Heyns, Barbara. 1978. Summer Learning and the Effects of Schooling. New York: Academic Press.
Institute of Education Sciences. 2012a. “How Does PISA Differ from Other International Assessments?”
http://nces.ed.gov/surveys/pisa/faq.asp#6. In PISA USA. Washington, DC: NCES.
______. 2012b. “How Does PISA Select a Representative Sample of Students?” http://nces.ed.gov/
su rveys/pisa/faq.asp#4. In PISA USA. Washington, DC: NCES.
Kelley, Jonathan. 1990. “The Failure of a Paradigm.” In John Goldthorpe, edited by C. Modgil, 319–46.
London: Falmer Press.
Kelley, Jonathan, and M. D. R. Evans. 1995. “Class and Class Conflict in Six Western Nations.” American
Sociological Review 60:157–78.
______. 2002. “Does Education Add Value? Ideal Rewards to Education in Three Nations.” Australian
Social Monitor 5(4):100–112.
Scholarly Culture Test Scores 31
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Kim, James S. 2006. “Effects of a Voluntary Summer Reading Intervention on Reading Achievement.”
Educational Evaluation and Policy Analysis 28:335–55.
Kim, James S., and Thomas G. White. 2008. “Scaffolding Voluntary Summer Reading for Children in
Grades 3 to 5.” Scientific Studies of Reading 12:1–23.
King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. “Analyzing Incomplete Political
Science Data.” American Political Science Review 85:49–69.
Kingston, Paul W. 2000. “Class Culture.” In The Classless Society, 119–48. Stanford, CA: Stanford
University Press.
______. 2001. “The Unfulfilled Promise of Cultural Capital Theory.” Sociology of Education 74:88–99.
Kohn, Melvin L., A. Naoi, Carrie Schoenbach, Carmi Schooler, and Kazimiersz M. Slomczynski. 1990.
“Position in the Class-Structure and Psychological Functioning in the United-States, Japan, and
Poland.” American Journal of Sociology 95:964–1008.
Kohn, Melvin L., Kazimiersz M. Slomczynski, Krystina Janicka, V. Khmelko, Bogdan W. Mach, V. Paniotto,
W. Zaborowski, R. Gutierrez, and C. Heyman. 1997. “Social Structure and Personality under Conditions
of Radical Social Change.” American Sociological Review 62:614–38.
Korupp, Sylvia E., Harry B. G. Ganzeboom, and Tanja van der Lippe. 2002. “Do Mothers Matter?” Quality &
Quantity 36:17–42.
Kraaykamp, Gerbert, and Paul Nieuwbeerta. 2000. “Parental Background and Lifestyle Differentiation in
Eastern Europe.” Social Science Research 29:92–122.
Krashen, Stephen. 2004. The Power of Reading. Portsmouth, NH: Heinemann.
Lamont, Michelle, and Annette Lareau. 1988. “Cultural Capital.” Sociological Theory 6:153–68.
Lareau, Annette. 2011. Unequal Childhoods. 2nd ed. Berkeley: University of California Press.
Marks, Gary N. 2005. “Cross-National Differences and Accounting for Social Class Inequalities in
Education.” International Sociology 20:483–505.
Marks, Gary N., John Cresswell, and John Ainley. 2006. “Explaining Socioeconomic Inequalities in Student
Achievement.” Educational Research and Evaluation 12:105–28.
Mayer, Susan E. 1997. What Money Can’t Buy: Family Income and Children’s Life Chances. Cambridge,
MA: Harvard University Press.
Miller, Karen A., Melvin L. Kohn, and Carmi Schooler. 1985. “Educational Self-Direction and the Cognitive
Functioning of Students.” Social Forces 63:923–44.
Notten, Natascha, and Gerbert Kraaykamp. 2010. “Parental Media Socialization and Educational
Attainment.” Research in Social Stratification and Mobility 28:453–64.
OECD. 2002. PISA 2000 Technical Report. Paris: Organization for Economic Co-Operation and Development.
Park, Hyunjoon. 2008. “Home Literacy Environments and Children’s Reading Performance.” Educational
Research and Evaluation 14:489–505.
Pearlin, Leonard I., and Melvin L. Kohn. 1966. “Social Class, Occupation, and Parental Values.” American
Sociological Review 31:466–79.
Persson, Mikael. 2012. “Does Type of Education Affect Political Participation?” Scandinavian Political
Studies.
Portes, Alejandro, Patricia Fernandez-Kelly, and William Haller. 2009. “The Adaptation of the Immigrant
Second Generation in America.” Journal of Ethnic and Migration Studies 35:1077–104.
Price, Leah. 2012. How to Do Things with Books in Victorian Britain. Princeton, NJ: Princeton University
Press.
Raghunathan, T. E., Peter W. Solenberger, and John van Hoewyk. 2004. IVEware: Imputation and Variance
Estimation Software. University of Michigan: Survey Methodology Program, SRC, ISR.
Reisel, Liza. 2011. “Two Paths to Inequality in Educational Outcomes.” Sociology of Education 84:261–80.
Ridgeway, Cecilia L., and Kristan G. Erickson. 2000. “Creating and Spreading Status Beliefs.” American
Journal of Sociology 106:579–615.
32 Social Forces
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from

Ridgeway, Cecilia L., Yan E. Li, Kristan Erickson, Kristen Backor, and Justine Tinkler. 2009. “How Easily
Does a Social Difference Become a Status Distinction?” American Sociological Review 74:44–62.
Rivera, Lauren A. 2012. “Hiring as Cultural Matching.” American Sociological Review 77:999–1022.
Royston, Patrick. 2004. “Multiple Imputation of Missing Values.” Stata Journal 4:227–41.
Royston, Patrick, Gareth Ambler, and Willi Sauerbrei. 1999. “The Use of Fractional Polynomials to Model
Continuous Risk Variables in Epidemiology.” International Journal of Epidemiology 28:964–74.
Schafer, Joseph L. 1997. Analysis of Incomplete Multivariate Data. London: Chapman and Hall.
Sikora, Joanna, and Lawrence Saha. 2010. “New Directions in National Education Policymaking: Student
Career Plans in International Achievement Studies.” In The Impact of International Achievement
Studies on National Education Policymaking. edited by Alex Wiseman, 83–115. Bingley, UK: Emerald
Group Publishing.
Spaeth, Joe L. 1976. “Cognitive Complexity.” In Schooling and Achievement in American Society, edited by
David Featherman, 103–31. New York: Academic Press.
______. 1979. “Vertical Differentiation among Occupations.” American Sociological Review 44:746–62.
______. 1989. “Occupational Status, Resource Control, and Earnings.” Research in Social Stratification
and Mobility 8:303–319.
Stinchcombe, Arthur L. 1968. “The Logic of Scientific Inference.” In Constructing Social Theories, 15–37.
Chicago: University of Chicago Press.
Sullivan, Ann. 2001. “Cultural Capital and Educational Attainment.” Sociology 35:893–912.
Swartz, David. 1997. Culture and Power: The Sociology of Pierre Bourdieu. Chicago: University of Chicago
Press.
Swidler, A. 1986. “Culture in Action.” American Sociological Review 51:273–86.
Teachman, Jay D., Kathleen Paasch, and Karen Carver. 1997. “Social Capital and the Generation of Human
Capital.” Social Forces 75:1343–59.
Treiman, D. J., H. B. G. Ganzeboom, and Susanne Rijken. 2003. “Educational Expansion and Educational
Achievement in Comparative Perspective.” Working Paper, California Center for Population Research.
Treiman, Donald J. 1977. Occupational Prestige in Comparative Perspective. New York: Academic Press.
Treiman, Donald J., and Kam-Bor Yip. 1989. “Educational and Occupational Attainment in 21 Countries.”
In Cross-National Research in Sociology, edited by Melvin L. Kohn, 373–94. Newbury Park: SAGE.
Tzanakis, Michael. 2011. “Bourdieu’s Social Reproduction Thesis and the Role of Cultural Capital in
Educational Attainment.” Educate 11:76–90.
Vale, Colleen, Mary Weaven, Anne Davies, Neil Hooley, Kristy Davidson, and Daniel Loton. 2012.
“Growth in Literacy and Numeracy Achievement.” Australian Educational Researcher DOI: 10.1007/
s13384-012-0065-9.
van de Werfhorst, Herman G. 2010. “Cultural Capital.” British Journal of Sociology of Education 31:157–69.
van Eijck, Koen, and Paul M. de Graaf. 2001. “The Impact of Personality Traits on Educational Attainment.”
Mens-en-Maatschappij 76:285–302.
Warren, John R., and Robert M. Hauser. 1997. “Social Stratification across Three Generations.” American
Sociological Review 62:561–72.
Weber, Max. 1968. Economy and Society. New York: Bedminster.
West, Patrick, Helen Sweeting, and Ewen Speed. 2001. “We Really Do Know What You Do.” Sociology
35:539–60.
Whitt, Hugh P. 1986. “The Sheaf Coefficient.” Social Science Research 15:174–89.
Scholarly Culture Test Scores 33
by guest on January 11, 2016http://sf.oxfordjournals.org/Downloaded from