Family scholarly culture and educational success: Books and schooling in 27 nations

Abstract

Children growing up in homes with many books get 3 years more schooling than children from bookless homes, independent of their parents’ education, occupation, and class. This is as great an advantage as having university educated rather than unschooled parents, and twice the advantage of having a professional rather than an unskilled father. It holds equally in rich nations and in poor; in the past and in the present; under Communism, capitalism, and Apartheid; and most strongly in China. Data are from representative national samples in 27 nations, with over 70,000 cases, analyzed using multi-level linear and probit models with multiple imputation of missing data.

Full text

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
Research in Social Stratification and Mobility xxx (2010) xxx–xxx
A
vailable online at www.sciencedirect.com
Family scholarly culture and educational success: Books and
schooling in 27 nations
M.D.R. Evans a,∗, Jonathan Kelley b, Joanna Sikorac, Donald J. Treimand
aSociology Department, University of Nevada, 1664 N. Virginia St., MSS 300, Reno, NV 89557-0042, United States
bInternational Survey Center and University of Nevada, Reno, United States
cAustralian National University, United States
dUniversity of California at Los Angeles, United States
Received 1 December 2007; received in revised form 4 August 2009; accepted 12 January 2010
Abstract
Children growing up in homes with many books get 3 years more schooling than children from bookless homes, independent of
their parents’ education, occupation, and class. This is as great an advantage as having university educated rather than unschooled
parents, and twice the advantage of having a professional rather than an unskilled father. It holds equally in rich nations and in
poor; in the past and in the present; under Communism, capitalism, and Apartheid; and most strongly in China. Data are from
representative national samples in 27 nations, with over 70,000 cases, analyzed using multi-level linear and probit models with
multiple imputation of missing data.
© 2010 Published by Elsevier Ltd on behalf of International Sociological Association Research Committee 28 on Social Stratification
and Mobility.
Keywords: Social stratification; Education; Books; Scholarly culture; Elite closure; Cultural capital; Home literacy environment; Culture; Schooling;
Cross-national
Throughout the world, education is the key to good
jobs and high incomes, and hence a central concern in
sociology. Research in the Blau-Duncan tradition long
ago delineated, with considerable precision, how par-
ents’ education, occupational status, and class shape
their offspring’s educational careers and how this does
(or does not) vary between nations, over time, and in
response to government policy. Researchers have now
begun to explore a wide range of other parental resources.
Promising among them is scholarly culture – the way
of life in homes where books are numerous, esteemed,
read, and enjoyed. Recent research on several West-
∗Corresponding author. Tel.: +1 775 784 6333.
E-mail address: mariah@international-survey.org
(M.D.R. Evans).
ern nations, as well as Hungary, suggests that parents’
scholarly culture enhances children’s educational attain-
ment, net of other things (Crook, 1997a; de Graaf, 1986;
de Graaf, de Graaf, & Kraaykamp, 2000; DiMaggio,
1982; Evans & Kelley, 2002; Ganzeboom, de Graaf, &
Robert, 1990). The effect is usually “weak but signifi-
cant” (Aschaffenburg & Maas, 1997), sometimes absent,
and varies greatly depending on which aspect of culture
is measured (Kingston, 2001). It most likely comes about
because children from cultured homes perform better in
school (Park, 2008).
We pursue this promising beginning, measuring par-
ents’ scholarly culture in a consistent manner by the
number of books in the home and estimating its effect on
children’s education in 27 nations, net of a comprehen-
sive, consistently measured set of control variables. We
0276-5624/$ – see front matter © 2010 Published by Elsevier Ltd on behalf of International Sociological Association Research Committee 28 on Social Stratification and Mobility.
doi:10.1016/j.rssm.2010.01.002

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
2M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
seek to establish whether it has an impact on children’s
education only in a handful of rich Western nations at
the end of the 20th century, or whether it is important
in all rich nations, or in all market economies, or under
Communism, or only in recent decades rather than in
past generations.
It is not yet clear why scholarly culture has an impact.
If it is an elite conspiracy using essentially arbitrary cul-
tural signals to recognize fellow members and to exclude
others (Bourdieu, 1984), it is likely to be important only
in a few nations under particular historical circumstances
and vulnerable to changes in government policy. But
if scholarly culture provides skills and knowledge that
are central to literacy and numeracy, and hence valu-
able in schools everywhere, it is likely to be important
throughout the world and little affected by historical
circumstance or government policy.
To address these issues, we analyze a broad array of
countries at different levels of economic development, in
different historical periods, with varying cultural back-
grounds, following diverse social and political policies.
Multiple, diverse tests make for a stronger theory test-
ing program (Stinchcombe, 1968). Data are from the
World Inequality Study, a large database compiled from
high quality, representative national samples (27 nations
for this analysis; N= 73,349), analyzed with multi-level
linear and probit models with multiple imputation of
missing values.
Plan of the paper. We first describe our theory. The
following sections detail our evidentiary base, define
variables, and describe our estimation procedures. A
descriptive section then gives basic information for each
country. The main analytic section estimates the effect of
parents’ scholarly culture in the pooled sample, assesses
interactions, evaluates historical differences, and gives
country-specific details. Sensitivity analyses investigate
the robustness of our results. A summary and discussion
conclude.
1. Theory
Research on the role of culture in status attainment
has largely focused on its influence on formal educa-
tion. Within this research tradition, two broad themes
have emerged. On the one hand, a scholarly culture
model (e.g. Crook, 1997a; de Graaf, 1986; Evans &
Kelley, 2002; Teachman, 1987) views the effects of
books and reading on education as a consequence of
culture being a “toolkit” of competencies, skills, and
funds of knowledge, with complexity an important theme
(Kohn, Naoi, Schoenbach, Schooler, & Slomczynski,
1990; Spaeth, 1976; Swidler, 1986). The key proposition
is that scholarly culture enhances learning and perfor-
mance in school. On the other hand, an elite closure
model (incisively summarized and critiqued in Kingston,
2001) views the effects of scholarly culture as a conse-
quence of the elite using essentially arbitrary cultural
signals to recognize fellow members and to exclude
others (Bourdieu, 1984; Goblot, 1925 [1973]). The key
proposition here is that elites arbitrarily choose high cul-
ture as a signal of elite membership but that it has no
substantive impact on task performance.
Nomenclature in this area is varied and confusing.
What we call the “elite closure” model focusing on
elite discrimination via cultural markers or “secret hand-
shakes” is far removed from the natural interpretation of
its alternative name of “cultural capital” which suggests
skills and knowledge analogous to “human capital”. We
describe skills and information useful in school work
as “scholarly culture” rather than as “human capital”
because the economists’ “human capital” usage sug-
gests that that acquiring skills and information is a
“cost” to the individual (notably in income forgone),
whereas our approach posits it as a normal part of their
lifestyle. Reading for pleasure, acting out stories based
on favorite books, playing charades or word games on
a winter’s night – all of these are intrinsically reward-
ing, even if they also enhance skills that are rewarded in
school.
The scholarly culture and elite closure perspectives
make largely similar predictions about the effects of
scholarly culture in rich Western democracies with stable
elites. But they have different implications for poor peo-
ple, past times, and Communist governments. If bookish
origins provide cognitive toolkits that enhance task per-
formance, as the scholarly culture theory claims, they
are likely to be valued under diverse regimes and val-
ued by modest families, not just the elite. For example,
there is no reason why Communist educational sys-
tems should value competence less than educational
systems in market societies. But if the scholarly culture
merely provides arbitrary signals, “secret handshakes”
that allow elites to monopolize advantages, then it seems
extremely unlikely that societies differing in regime ori-
entation – from Communist to fascist, from those intent
on exploding privilege to those aggressively entrenching
it – and that differ in culture, in wealth, and in many other
ways should just coincidentally settle on the same sig-
nal. Thus, the broad array of countries and time periods
analyzed in this paper provides new insights into these
theoretical issues.
Because, as will be seen, the results support the schol-
arly culture model, we focus on that here and return to the
implications for alternative theories in the Discussion.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 3
2. Scholarly culture
2.1. Cultural toolkit
Stemming from work by Spaeth (1976) and de Graaf
(1986), the scholarly culture hypothesis holds that read-
ing provides cognitive skills that enhance educational
attainment, a cultural toolkit. A home in which books are
an integral part of the way of life will encourage chil-
dren to read for pleasure, thereby providing them with
information, vocabulary, imaginative richness, and wide
horizons (Dronkers, 1992).
The usage of the term “culture” here refers not to
overarching values, but rather to culture as an everyday,
routine set of practices and preferences that are engaged
with material objects (books, in this case) and with activ-
ities (reading, talking about books, using knowledge).
This has also been called “mini-culture” (Evans & Lukic,
1998) or “habitus” (Bourdieu, 1984).1The argument is
that it generates generalized cognitive skills [or “com-
plexity” (Kohn et al., 1997), or a “toolkit” (Swidler,
1986)] which are useful in problem solving in formal
education, even on topics distant from the books’ subject
matter.2
This approach suggests that a substantive connection
between scholarly resources and performance in school
accounts for much of culture’s effect on educational
attainment (Bidwell, 1989; Teachman, 1987). Books and
reading are a concrete resource and indicate a cognitively
complex way of life that enhances intellectual capacities
in ways directly useful in school, improving academic
1An alternative term, used particularly in the analysis of early child-
hood cognitive performance, is “home literacy environment” (e.g.
Farkas & Hibel, 2008). This is also a clear term, but we prefer the
culture terminology because it is used in this sense in the literature on
which we draw most closely and because it makes explicit the links to
larger theoretical issues which are important here.
2It is logically possible that books could be separated from such a
culture – perhaps being inherited from a maiden aunt – and that they
are so intrinsically attractive that children devour them voraciously
in the absence of any support or encouragement from parents. But it
is not likely: In Australia (where we have information both on home
library size and on reading, unlike the international data), of parents
with 100 or more books, only 2% never read, and on average, their
children noticed them reading more than once a week. The correlation
between owning books and reading is fully 0.64. Consider also the dust
that gathers on school library shelves unless children are assigned to
work there: Availability is not enough. So interpreting a home library
as an indicator of participation in a scholarly culture/subculture is rea-
sonable. Indeed, close study of reading, books, and related matters in
the homes of young children shows that the number of books in the
home is “...one of the more accurate measures of parental interest in
providing instruction to the child” (Farkas & Hibel, 2008).
performance. For example, the larger the home library,
the better children perform on standardized reading tests,
net of parents’ education, across a broad range of coun-
tries (Park, 2008).
Because it generates skills and knowledge central to
schooling, scholarly culture should enhance educational
achievement in all societies, rich and poor alike; in all
political systems, Communist and capitalist alike; and in
the past as well as the present.
2.2. Diminishing marginal returns
The scholarly culture approach suggests diminishing
marginal returns to cognitive resources. The reason for
this is that culture is not analogous to material capital
in one crucial respect. Material capital is exclusive: only
one person can own it at a time, it is a zero sum resource.
In contrast, culture is shared: many people can possess
it at the same time. As a consequence, the gain from
a gift of culture depends on what culture the recipient
had to begin with. For example, suppose that my uncle
the archaeologist gives me a chance to work in his dig
one summer. If I were previously ignorant of archaeol-
ogy, I would learn a vast amount: new words, new skills,
and new knowledge. But if I had been there the summer
before, or if I already had a PhD in archaeology, I would
learn much less, since I would already know much of
what is on offer. So a gift of culture has diminishing
marginal returns.
Parents’ home library, and the vocabulary, skills, and
knowledge that come from it, are shared resources. More
formally, let Book1be the set of all words, ideas, infor-
mation, and skills to be acquired from the 1st book in
the home library; Book2those for the 2nd book; and so
on to BookN. Then the set of all words, ideas, etc. in the
parents’ home library, pBooks, is the union of Book1to
BookN:
pBooks =Book1∪Book2∪···∪BookN(1)
Because working vocabularies are somewhat limited,
because the information base of any particular book is
limited, and because key interpretive strategies are mas-
tered relatively early (with more sophisticated ones being
variations on established themes), the existing “stocks”
of information, vocabulary, etc., will in general grow
with library size. But each new book contributes pro-
gressively less new material, adds less to the “stocks”
than did the books acquired earlier, since much is redun-
dant. For example, the first reading of Green Eggs and
Ham, opens up a whole new world; and, after that, most
of the words in Fox in Socks would be still be new;
but by the time one has mastered Hamlet, the vocab-

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
4M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
ulary and interpretive strategies required for Macbeth
are only a slight advance. Thus the relationship between
the number of books in the home and the scholarly cul-
ture/cultural toolkit gained from them will be non-linear,
showing diminishing marginal increments for each addi-
tional book.
2.3. Interactions
Now consider what gains to their cognitive toolkits
children get out of their parents’ books (pBooks), their
parents’ education (pEduc), and so on. If Vtis the set
of vocabulary, ideas, information, etc., that a child pos-
sesses at time t(for example, when they are 8 years old),
then their stock of scholarly culture at the end of the fol-
lowing year is the union of (1) what they already had and
(2) what they are exposed to that year from their parents’
home library (for example, 3% of the total), less what of
that they already knew (pBooks ∩Vt); and (3) plus what
they are exposed to because of their parents’ education,
less what of that they already knew (pEduc ∩Vt), and so
on:
Vt+1=Vt∪b1(pBooks −(pBooks ∩Vt))
∪b2(pEduc −(pEduc ∩Vt))
∪...etc.
(2)
where bi( ) is a function, analogous to a regression coeffi-
cient, which chooses a certain proportion of the members
of a set. The same logic applies whether it is the respon-
dent or any other family members reading the books:
they will bring their enhanced scholarly culture into their
normal interactions with the rest of the family.
The model of Eq. (2) implies a negative interaction
between number of books in the parents’ home library,
pBooks, and parents’ education, pEduc – the more one
already has from one source and so the greater Vt,
the less any increment from the other can contribute.
Thus children from poorly educated families will bene-
fit proportionately more from scholarly culture than will
children from well-educated families.
2.4. Cultural preferences
In addition to providing skills and knowledge, a large
home library is a manifestation of the family’s prefer-
ences: an indication that they enjoy and value scholarly
culture, that they find ideas congenial, reading agreeable,
complex and intellectually demanding work attractive. It
shows a commitment to investing in knowledge, and per-
haps in schooling. It suggests that conversations between
parents and their children will include references to
books and imaginative ideas growing out of them. In
short, a large library reveals a preference for the scholarly
culture.
Moreover, a home library is a relatively pure mea-
sure of preference. It is little contaminated by financial
constraints – books are not expensive, unlike educa-
tional credentials which cost orders of magnitude more
in income forgone and tuition paid. Nor are there finan-
cial attractions; one reads a book at home for pleasure,
but one goes to college in part for the income and secu-
rity it ensures afterward in the labor market. Nor are
there many practical constraints: love, war, and wander-
lust may upset well-laid educational plans, but books are
to be had anywhere and anytime.
Our argument thus suggests:
H1. Scholarly culture confers educational advantage:
parents’ participation in scholarly culture will enhance
children’s educational attainment, net of the parents’ for-
mal education and social class, and do so in all nations,
throughout history, and regardless of government policy.
H2. Biggest gains at the bottom: an increase in schol-
arly culture has the greatest impact on children from
families with little scholarly culture.3
H3. Interaction with formal education: scholarly cul-
ture has a greater effect on children’s education if their
parents are poorly educated.4
Other approaches offer counter-arguments to all these
hypotheses; we take them up in the Discussion, where
we reflect on them in light of the evidence.
3. Data
The World Inequality Study (Kelley, Evans, & Sikora,
2007) pools data from several major projects, all based on
representative national samples incorporating detailed
information on father’s and respondent’s occupation (4-
digit ISCO or equivalent) and detailed educational data.
They are: the International Social Survey Program ISSP
(Zentralarchiv fuer Empirische Sozialforschung, 2002);
the Social Stratification in Eastern Europe SSEE, sur-
veys (Szelenyi & Treiman, 1994); the Life Histories and
Social Change in Contemporary China Survey (Treiman
3For example, going from 10 books to 20 will matter more than
going from 40 to 50.
4For example, going from 10 books to 20 in a family where both
parents only finished primary school will help more than going from
10 to 20 books in a family where both parents finished high school.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 5
et al., 1996); the International Survey of Economic Atti-
tudes (Kelley et al., 1998); and the Social Change in
South Africa survey (Treiman & Lewin, 1993). We ana-
lyze only surveys which include a question on home
library size.5With this restriction, there are 73,249
cases with valid answers in 27 nations ranging from
poor countries (Philippines, China) to the rich coun-
tries of Northwestern Europe and its overseas extensions.
Because education typically occurs early in the life cycle,
and because these surveys have retrospective data about
family background, they provide temporal depth reach-
ing back to World War II for older cohorts. Into these
individual level data we have merged contextual level
data on economic development (GDP when respondent
was 15 and beginning to make school/career decisions)
and political-institutional context (Communist vs. non-
Communist).
In the country-by-country analyses, we provide race-
specific estimates for South Africa, because the pattern of
effects is particularly telling for our theory and because
educational segregation was extreme. We also provide
analyses for China as a whole and for the urban and
rural sectors separately, because residential controls and
differentiation of privilege between the cities and the
countryside (Unger, 1982) could well lead to different
links between scholarly culture and education. Internal
differentiation in other nations is less institutionalized.
East Germany is treated as a separate nation as most
respondents were educated under the former Commu-
nist regime, in very different circumstances from West
Germany.
3.1. Imputation of missing data
In order to use all the information that respondents
provide, we did multiple imputation of missing data sep-
arately for each society, following the general approach
of King and colleagues (Honaker, Joseph, King, Scheve,
& Singh, 2003; King, Honaker, Joseph, & Scheve, 2001),
in practice a regression-based technique augmented by
a random component to the imputed value. We did the
5Australia: International Social Science Survey/Australia,
1984–2003 Kelley et al; Bulgaria 1993 SSEE; Canada 1999 ISSP;
Chile 1999 ISSP; China 1996 Treiman et al; Cyprus 1999 ISSP; Czech
Republic 1993 SSEE and 1999 ISSP; France 1999 ISSP; Germany
East 1999 ISSP; Germany West 1999 ISSP; Hungary 1993 SSEE and
1999 ISSP; Israel 1999 ISSP; Japan 1999 ISSP; Latvia 1999 ISSP;
Netherlands 1999 ISSP; New Zealand 1999 ISSP; Norway 1999 ISSP;
Philippines 1999 ISSP; Poland 1993/4 SSEE and 1999 ISSP; Portugal
1999 ISSP; Russia 1993 SSEE and 1999 ISSP; Slovakia 1993 SSEE
and 1999 ISSP; Slovenia 1999 ISSP; South Africa 1992 Treiman et
al.; Spain 1999 ISSP; Sweden 1999 ISSP; USA 1999 GSS/ISSP.
imputations, under the assumption that the data are miss-
ing at random (MAR), separately for each country to
take advantage of country-specific patterns. These and
related procedures have desirable properties when data
are “missing completely at random” (MCAR) or, as is
reasonable to assume here, “missing at random” (MAR);
they also perform well in simulations (Allison, 2000;
Schafer, 1997). Given our many 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, & Van Hoewyk,
2004),6choosing options to estimate models similar to
King’s.
4. Measurement
4.1. Scholarly culture: number of books in the
parents’ home
To measure parents’ scholarly culture, the surveys
asked the number of books in the parents’ home when the
respondent was young.7The question usually followed
items on parents’ education and occupation:
About how many books were there around your family’s house
when you were 14 years old?
None
1or2
Around 10
Around 20
Around 50
Around 100
Around 200
Around 500
1000 or more
This item has good measurement properties. First,
it is reported reliably. Test–retest reliability over a 5-
year period is r= .76 in Australia, compared to .75 for
parents’ education, .81 for father’s occupational status,
.60 for father’s ownership, and .58 for father supervisor
(N= 1150; calculations by the authors).
Second, the item also appears to be a valid indicator
of scholarly culture in the home. Following pioneering
work in the Netherlands (de Graaf, 1986), research in
several nations shows that parents’ books (home library
size) is correlated with other aspects of scholarly cul-
ture, including how often parents read “serious novels or
poetry”; read science, mathematics or technology; read
6On the advantages of this method see Downey, Hippel, and Broh
(2004).
7The time reference was usually age 14; in some nations it was age
15 or 16 according to local conventions.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
6M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Fig. 1. Description: books in parents’ home and children’s education. Means, unadjusted. Pooled data from 31 nations; N= 72,787.
Source:Table A.2.
“other serious books like history or biography”; and went
to the library; and, moreover, that it is clearly distinct
from arts spectatorship involving attendance at drama,
art museums, classical, music and dance performances
(Crook, 1997b; de Graaf et al., 2000; Evans & Kelley,
2002; Sullivan, 2001; Zimdars, Sullivan, & Heath 2009).
Moreover, analysis of many different aspects of the home
environment finds that home library size has strong pre-
dictive validity as an indicator of parents’ attraction to
the teaching role vis a vis their children (Farkas & Hibel,
2008).
Thus, this single measure is conceptually central, well
correlated with other aspects of scholarly culture, reli-
ably reported by respondents decades after they left their
parents’ home, and applies to a wide variety of cultures,
not just rich Western societies.
We analyze the natural log of this, because both the-
ory and our results suggest diminishing marginal effects
(see Fig. 1, below). Characteristically, when there is a
non-linear relationship between two variables such that a
one-unit change at low values of the independent variable
produces relatively large gains in the dependent variable,
but these gains shrink at higher values of the independent
variable, a natural log transformation of the independent
variable linearizes the relationship for estimation pur-
poses (Allen, 1997; Allison, 1999). A familiar example
would be the effect of income on happiness – big gains
in happiness with each extra dollar when incomes are
small, smaller gains in happiness with each extra dol-
lar as incomes get large. Many dose–response models
are also linear in the log of the focal independent vari-
able. We also assessed three alternatives: (1) a linear plus
quadratic form was inferior to the log specification, (2) a
linear plus square root form was inferior to the log spec-
ification, and (3) using a set of 7 dummy variables added
nothing of consequence. Accordingly, we used the more
powerful and parsimonious log specification.
4.2. Education
Education was measured by questions appropriate to
each nation and then recoded by the original investigators
into equivalent years of formal education (Jagodzinski
& Uher, 2001; Kelley et al., 1998; Treiman, 1994;
Treiman et al., 1993, 1996). Additional, country-specific
information on degrees and other qualifications is avail-
able for some nations; using this we have sometimes
readjusted the original investigators’ scores to be more
internationally comparable. Country-by-country details
are available on www.international-survey.org.
We allow for the possibility that the effects of schol-
arly culture and family background may differ by level
of education by analyzing several key stages: year 9, sec-
ondary school, and university. We estimate total effects,
contrasting those who completed year 9 or higher with
everyone else; contrasting those who finished secondary
school or higher with everyone else; and contrasting uni-
versity graduates with everyone else. This allows for
discontinuities without entailing selectivity bias prob-
lems that arise from analyzing transitions sequentially
(Cameron & Heckman, 1998).
4.3. Measurement of class, national context, and
other variables
Since the number of books is likely to be corre-
lated with other important determinants of education, we
introduce them as controls. These include parents’ edu-
cation and class, and an indicator for Eastern Europe,
which has unusually high levels of education for its
modest level of economic development.8
8Expert opinions are divided over the estimation of GDP in Eastern
Europe during Communist times (Lancieri, 1993). This issue is beyond

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 7
4.3.1. Economic development when respondent was
young
We also include a measure of the nation’s GDP per
capita when respondent was age 15. Much research
on children’s performance on achievement tests has
included the level of economic development as a pre-
dictor variable (e.g. Heyneman & Loxley, 1982; Park,
2008), but, to our knowledge, this is the first analy-
sis to include it in a model of the effects of scholarly
culture and parental education on educational attain-
ment. For that we need economic development at the
time the schooling took place – for example around
1935 for respondents born in 1920, or around 1995 for
respondents born in 1980. Historical estimates of this
sort are not readily available. We have developed rough
but plausible estimates for them. They are available
for download on our website www.international-survey.
org.
Table A.1 details the measurement of these variables.
5. Methods
5.1. Model
We assume that children’s education is a function
of parents’ scholarly culture, family background, and
national characteristics:
Education
=b1+b2ln ParentsBooks +b3ParentsEducation
+b4FathersOccupation +b5FatherOwner
+b6FatherPetitBourgeois +b7FatherSupervisor
+b8Male +b9LnParentBooks
×ParentsEducation +b10EasternEurope
+b11 ln GDP15 +e(3)
For country-by-country analyses the national level
variables (Eastern Europe and ln GDP15 drop out). We
make no assumption about causal order among variables
on the right-hand side.
Our estimates give the direct effects of the right-hand
side variables, regardless of how they came about. The
intervening mechanisms are interesting but beyond the
scope of this analysis.
the scope of this paper, so we have simply used what we believe to be
the best available estimates.
5.2. Estimation
We use multi-level linear and probit regression to esti-
mate our core models. For the country-specific analyses,
we use OLS (with robust Huber-White standard errors)
and probit models.
5.3. Presentation of results: first differences
The crucial issue is how much schooling is implied
for students from families with different numbers of
books in the home. For models that are fully linear this
is a well-known matter of applying appropriate means
to the regression parameters and taking differences in
predicted values (Jones & Kelley, 1984). But it is not
so simple for our curvilinear dependent variable, the
natural log of books, or for the intrinsically non-linear
probit estimates. We therefore present first differences
in predicted values, which are closely related to partial
derivatives (King, Murray, Salomon, & Tandon, 2004).
We estimate them from a whole population standardiza-
tion (Kelley & Evans, 1995) which makes comparisons
using a common reference population, here the pooled
set of respondents.9The results depend both on the
equation and on the population chosen as a baseline for
comparison. Specifically, the predicted education of chil-
dren from bookless homes is obtained by changing every
case in the sample to have no books while leaving all
other variables unchanged; computing predicted values
for every case (using the coefficients in Table A.3); and
then averaging. The answer is that they would expect to
get 9.4 years of education. Children from otherwise iden-
tical families with 500 books would expect to get 12.6
years of education. The difference between these two,
an advantage of 12.6 −9.4 = 3.2 years, is the key result.
Confidence intervals for these differences are computed
by bootstrap methods. The effects we discuss throughout
the paper are obtained in this way.
We have chosen comparison points reflecting the full
range of values, from high (but not extraordinary) to low
(but still reasonable). Specifically, for books we compare
families with 500 books to those with none (roughly the
top 20% vs. the bottom 10% worldwide).10 For parents’
education, we compare college graduates (15 or 16 years
of education) to people who left school after 3rd Grade.
9Our reference population greatly over-represents Australia because
of its large sample. But this makes little practical difference since Aus-
tralia is unexceptional (see Table 1). For example, the effect of growing
up in a family with no books versus one with 500 books is 3.2 years
including Australia and 3.3 excluding it entirely.
10 Scoring the true zeroes as 0.5, so the log is defined.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
8M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Table 1
Description: percentages and means for 31 societies, circa 1999. N= 73,349a.
Society (sorted by
mean years of
education)
A: Books in parents’ home B: Family background C: Education Casesa
% None %
Around
10
%
Around
25
%
Around
75
%
Around
500+
Total= 100%Mean Mean, parents’
education
Mean, father’s
occupation
% Father
owner
% Father
petty
bourg
% Father
supervise
Mean, GDP at
age 15 (Index.
1990 USA = 1)
% Grade
9 or more
% Secondary+ % University Mean,
years
United States 3 25 19 34 18 100 112 10.9 40 10 12 37 .60 95 85 27 13.4 1095
Canada 1 17 17 39 25 100 142 11.8 45 18 10 37 .58 97 84 32 13.2 813
Netherlands 3 16 20 47 14 100 134 9.8 44 11 8 29 .46 92 61 23 12.9 1422
New Zealand 2 14 13 43 28 100 164 10.0 41 24 12 36 .46 91 56 29 12.9 957
Israel 4 9 11 41 35 100 224 9.8 42 16 18 30 .33 88 75 23 12.8 942
France 3 18 16 40 23 100 143 8.1 44 17 16 33 .39 91 60 32 12.6 1774
Latviab2 5 11 37 45 100 265 9.7 37 3 9 24 .21 91 61 17 12.5 950
Norway 1 10 16 44 28 100 158 8.7 44 4 8 32 .45 91 57 20 12.4 1058
South Africa: White 2 20 14 39 26 100 152 10.6 45 27 4 45 .20 95 68 17 12.3 2110
Japan 6 22 18 39 15 100 94 7.7 39 20 38 31 .34 91 74 18 12.3 1102
Czech Republic 3 11 12 41 33 100 187 9.6 35 3 6 17 .21 91 63 15 12.3 6848
Russia 16 18 13 30 23 100 146 7.1 34 1 3 21 .17 81 57 26 12.0 5554
Slovakia 6 24 19 36 15 100 88 8.7 32 2 9 16 .19 86 61 13 12.0 5275
Sweden 1 13 15 42 29 100 166 8.1 39 20 10 33 .47 80 49 14 11.4 953
Australia 3 18 17 42 21 100 129 8.9 40 16 11 38 .42 82 40 20 11.1 14,843
Cyprus 9 34 21 31 6 100 46 7.1 31 12 29 12 .22 78 67 20 11.0 874
Slovenia 11 27 19 33 10 100 69 7.5 28 4 21 20 .25 70 44 11 10.9 891
Germany-East 8 16 20 38 19 100 111 7.1 34 7 5 27 .24 65 17 12 10.9 455
Poland 18 26 18 28 10 100 68 7.5 27 4 35 13 .16 72 39 10 10.8 4088
Philippines 15 54 16 12 2 100 22 7.3 25 10 48 19 .09 62 31 17 10.8 966
Hungary 16 23 15 27 20 100 117 7.0 28 4 20 11 .15 63 38 12 10.6 5418
Bulgaria 30 25 12 21 12 100 81 5.8 26 2 17 8 .13 66 35 12 10.5 4221
Germany-West 7 22 20 35 16 100 96 7.0 37 13 9 27 .38 46 16 9 10.1 822
Spain 12 31 20 27 10 100 61 5.5 27 5 28 11 .30 57 33 15 9.8 1013
South Africa: Asian 4 46 22 23 6 100 42 4.2 32 17 14 23 .21 59 25 5 9.3 658
Chile 14 48 15 19 4 100 30 5.5 26 9 26 17 .15 54 39 10 9.3 1190
China: Urban born 9 31 25 32 3 100 66 3.7 32 7 6 18 .03 76 26 4 9.0 1814
South Africa: Colored 16 39 15 21 8 100 53 5.6 25 7 5 17 .21 48 17 3 8.3 707
South Africa: Black 31 40 10 13 7 100 48 3.7 23 6 6 12 .22 36 14 1 6.8 3517
Portugal 17 47 12 17 8 100 52 3.3 27 10 28 15 .20 26 15 6 6.4 1019
China: Rural born 25 44 19 12 0 100 25 1.6 9 3 12 4 .03 41 10 1 5.9 3738
(Total) 10 23 16 33 18 100 112 7.5 33 8.5 13 22 .25 75 45 15 10.8 73,349
aNumber of cases is for respondent’s education. Ns for other variables vary slightly due to missing data. GDP figures for China and South Africa refer to the whole nation, not the subgroups
shown.
bThe question on books in the home was asked in a slightly non-standard way, which probably leads to an upward bias.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 9
For father’s occupation, we compare free professionals at
the top of the occupational hierarchy (ISCO major group
1, prestige 58 and over) to farm laborers at the bottom.
For GDP, we compare US levels, at the top, with Chi-
nese levels, near the bottom with GDP per capita around
10% of the US. Other variables are dichotomous, so we
compare those with the characteristic to those lacking
it.
6. Description
Pooling all the countries, around 10% of respondents
grew up with no books at home (Table 1, Panel A, bottom
row). 23% had around 10 books and 16% had about 25;
some 33% had around 75 books, and 18% grew up with
hundreds of books. The average is 112.
6.1. Who has books?
Books are more numerous in rich nations, but mainly
because they have more well-educated, high status fam-
ilies. Hardly any Americans grew up without books
and 18% had over 500; the average was 112, the same
as the worldwide average. But this is around 80 less
that would be expected given America’s high GNP
and well-educated families, so America’s reputation for
philistinism may not be entirely undeserved. Book own-
ership in Western Europe varies greatly. People growing
up in Eastern Europe under Communism had neither
more or fewer books than expected given their GNP
and family background. (Details on these analyses are
available at www.international-survey.org.)
6.2. Parents’ books and children’s education
Pooling people from all the countries together, chil-
dren who grew up without books completed around 7
years of education on average (Fig. 1, Panel A, and
Table A.2). Those growing up with a couple of dozen
books completed 11 years, and offspring of the most
bookish parents completed 14 years of education, about
the level of an American junior college degree. Thus, on
average, 7 years of education separate those who grew up
without books in the home from those who grew up with
500 or more, a huge difference. Each additional book is
associated with greater gains in educational attainment
in families with few books than in families where there
are already many books.
The same pattern holds for each educational transition
(Fig. 1, Panel B and Table A.2). We will next see whether
these bivariate associations persist after adjustment for
our control variables.
7. Analysis
7.1. Years of education
Home library size has a very substantial effect on
educational attainment, even adjusting for parents’ edu-
cation, father’s occupational status, and other family
background characteristics (Table 2). Growing up in a
home with 500 books would propel a child 3.2 years fur-
ther in education, on average, than would growing up in
an otherwise similar home with few or no books (t= 65,
p< .001).11
This is a large effect both absolutely and in compar-
ison with other influences on education (Fig. 2, Panel
A and Table 2). (1) The difference between a bookless
home and one with a 500-book library is as great as the
difference between having parents who are barely liter-
ate (3 years of education) and having university educated
parents (15 or 16 years of education).12 Thus, a home
library is as important as parents’ education, the most
important variable in the standard educational attainment
model. (2) Moreover a home library is twice as impor-
tant as father’s occupation: only 1.6 years of education
separates children of farm laborers at the bottom of the
hierarchy from professionals’ children at the top, all else
equal. This is just half the 3.2-year home library gap. (3)
Even though rich countries have for generations made
huge investments in education, the difference between
being born into a society as poor as China and a society
as rich as the United States amounts to just 2 years of edu-
cation, net of family background. This is less than two-
thirds the gap that separates children reared in bookless
homes those born into 500-book homes, all else equal.
7.2. Educational transitions
The major effect of books in the home is evident
throughout the educational career: in finishing Year
9, finishing secondary school, and going to university
(Table 2). (1) A child growing up in a family with 500
books is 33 percentage points more likely to finish Year
9than an otherwise identical child from a home with no
books. (2) A child from a 500-book family is 36 per-
11 This is the first difference in predicted years of education at differ-
ent levels of home library size, with other variables held constant by
whole population standardization, as described in Section 5.
12 It is reasonable that there should be effects of parents’ education net
of home library size because going further in the educational system
contributes both substantive knowledgeand also knowledge about how
the system works which parents can use to guide and advise their
children (Lucas, 2001).

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
10 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Table 2
Estimated gain in education from various sources. First differencesafrom multi-levellinear and probit regression models with bootstrapped confidence
intervals; multiple imputation of missing data separately for each nation. Pooled data from 31 societies. N= 77,758a.
Variable (and comparison) Years of education
(mean)
Complete year 9+
(percent)
Complete secondary+
(percent)
Complete university
(percent)
Gain 95% C.I. Gain 95% C.I. Gain 95% C.I. Gain 95% C.I.
Parents’ books (log) (500
books vs. 1 book)
3.2 3.1–.3 33% 32–34 36% 35–37 19% 18–20
Parents’ education (15
years vs. 3 years)
3.2 3.1–3.3 26% 25–27 36% 35–38 16% 15–17
Father’s occupation (higher
professional vs. farm
laborer)
1.6 1.5–1.7 10% 9–11 15% 14–17 12% 11–14
Father owner (yes vs. no) 0.1 0.0–0.2 ns −3to−1 1% 0–3 1% 0–2
Father petit bourgeois (yes
vs. no)
ns −0.1 to 0.0 ns −5to−3ns−1 to 0 2% 1–3
Father supervisor (yes vs.
no)
0.4 0.3–0.4 3% 2–4 3% 2–4 3% 2–4
GDP when young (USA vs.
China)
2.0 1.9–2.1 16% 15–17 14% 13–15 4% 3–5
Eastern Europe (yes vs. no) 1.4 1.4–1.5 6% 6–7 11% 11–12 1% 0–2
Male (yes vs. no) 0.6 0.5–0.6 5% 4–5 4% 3–5 4% 3–4
Source: Calculated from Table A.3. Bootstrap standard errors based on 200 iterations.
aFor example in row 1, someone born into a family that had only 1 book but was otherwise average in parents’ education, father’s occupation,
GDP, and the rest, would expect to get 9.4 years of education themselves. Another person from an otherwise identical family with 500 books would
expect to get 12.6 years of education. The difference between these two, an advantage of 3.2 years due just to having more books in the family, is
reported in column 1. The standard error of this estimate, given in column 2, implies that we can expect, with 95% confidence that the true advantage
is between 3.1 and 3.3 years. The calculation for percent completing grade 9 or more in school, reported in column 3, is similar: someone from the
first family has a 59% chance of finishing year 9, someone from the second a 91% chance, a difference of 33 percentage points, rounded. ns: not
significantly different from zero at p< .001, two-tailed.
centage points more likely to graduate from high school
than an otherwise similar child without a home library.
(3) Finally, a child from a family rich in books is 19 per-
centage points more likely to complete university than
a comparable child growing up without a home library.
Books are more important than parents’ education or any
other influence in the usual model (Fig. 2, Panel B).
Thus the largest gains from a home library are below
university level, at year 9 and year 12.13
7.3. Parents’ books and parents’ education:
interactions?
Our theory implies that scholarly culture will have a
strong effect at all educational levels, even far below the
13 This difference is statistically significant, as can be seen from the
confidence intervals. Note that this refers to percentage point differ-
ences, for example 62% from homes rich in books finishing secondary
school versus 26% from bookless homes, a difference of 36 percent-
age points. This is compared to 24% versus 5% finishing university,
a difference of “only” 19 percentage points. But proportionally, com-
ing from a home rich in books increases secondary school completion
by 62/26 = 2.4 times while it increases university completion more,
24/5 = 4.8 times.
elite, with the greatest impact on children whose par-
ents are poorly educated (Hypotheses 2 and 3) – hence
a (negative) interaction between parents’ books and par-
ents’ education. Fig. 3 gives the evidence (details in
Table A.4).
7.3.1. Unschooled parents
Take as a baseline the 7.6 years of education predicted
for a child whose parents had no books and little or no
schooling (Fig. 3, Panel A, lower left corner).14 Parents
like this are common in Portugal or China today and
were common in past times in many other nations. We
assume the family was average in other respects. Had this
family owned 25 books instead of none at all, their child
could have expected over 2 more years of schooling, and
a further 2 years had they owned 500 books (Panel A,
lower right corner). The gain for such families amounts in
all to 4.3 years of education, comparing bookless homes
to book-rich homes, a huge difference.
14 This is higher than their actual education, 5.0 years (Appendix
DAppendix Table D), because the predicted value standardizes for
father’s occupation, GNP, and other background variables) which tend
to have very low values for such a child. Thus, part of their low
educational attainment reflects influences controlled in our model.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 11
Fig. 2. Estimated gain in education from various sources: first differences from multivariate models in Table 3.N= 77,758.
Fig. 3. Estimated education by parents’ home library and parents’ education, controlling for other aspects of family background and national
characteristics. Multi-level linear and probit estimates evaluated by whole population standardization. Pooled data from 31 societies with multiple
imputation of missing data; N= 77,758.
Source:Table A.4.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
12 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
7.3.2. Parents with primary schooling
A child whose parents had no books but had been to
primary school and were otherwise average could expect
8.8 years of schooling (Panel A, second row). Had the
family instead owned 25 books, the child could have
expected almost 2 more years of schooling, and almost a
further 2 years with 500 books. The gain amounts in all
to 3.5 years. Thus a home library is a great advantage for
children whose parents went only to primary school, but
not quite as great as for children of uneducated parents.
7.3.3. Parents with incomplete secondary schooling
A child whose parents had no books and left school
after 8th or 9th Grade (Year 8 or 9) – still common in
Western Europe and even more common in the past –
could expect 10 years of schooling (Panel A, third row).
Had the family instead been rich in books, with 500 in
their home library, the child would expect to finish high
school and perhaps go a bit beyond, a gain of 2.8 years.
Thus a home library is still a big advantage, but not as
large as for children from less educated homes.
7.3.4. Parents with high school education
A child whose parents had no books, although high
school graduates and otherwise average, would not
expect to finish high school themselves. Had the par-
ents instead been rich in books, their child could expect
to finish high school comfortably and go a year or so
beyond. In all, a large home library confers a 2.1-year
advantage: large, but not as large as for less educated
parents.
7.3.5. University educated parents
Finally, a child whose parents (very unusually) had no
books despite being university educated, could expect to
get 1.1 years less education than a child from a similar
home rich in books. This is only a modest advantage, far
less than in illiterate homes.
Thus for years of schooling completed, the advan-
tage conferred by a home library is largest for less
educated parents and smallest for university educated
parents (t=−26, p< .001; Table A.3). This is as pre-
dicted (Hypothesis 3) but contrary to Bourdieu who
expected the largest gain for the offspring of the elite.
7.4. Parents’ books and educational transitions:
interactions?
7.4.1. Year 9
A home library is a striking advantage in getting
through Year 9 in school other things being equal (Fig. 3,
Panel B). Only 40% of children from bookless homes
with unschooled parents can be expected to finish Year
9, compared to 88% of children with unschooled but
book-rich parents, a huge 48 percentage point advantage.
The home library advantage is 39 percentage points for
children with primary educated parents, 29 percentage
points for parents with incomplete secondary, 19 per-
centage points for high school educated parents, and just
9 percentage points for university educated parents. Thus
books matter most when parents have little education
(t=−9, p< .001, Table A.3).
7.4.2. High school
A home library is a big advantage in getting children
through high school, for illiterate and university edu-
cated parents alike (Panel C). For unschooled parents,
the advantage of a large home library is 33 percentage
points, about the same as the 37 point advantage for
primary educated parents, 40 for incomplete secondary
parents, 41 for parents with high school education, and
38 for university educated parents. Thus for high school
completion, unlike Year 9, books matter more or less
equally for all families (t= 1, ns).
7.4.3. University
For children whose parents are almost illiterate, a
home library confers a small but real advantage (Panel
D). Only 3% in bookless homes would go to university
compared to 13% where there is a large home library, an
advantage of 10 percentage points. This advantage rises
to 15 percentage points for primary school parents, 21
for parents with incomplete secondary schooling, 28 for
high school educated parents, and fully 37 percentage
points for university educated parents.
Thus in getting children through university, the advan-
tage conferred by a home library is largest for university
educated parents and least for poorly educated parents.
This is in agreement with the elite closure hypothesis,
but contrary to our Hypothesis 3 (t=6,p< .001).
7.5. History and political policy
Does the great advantage of growing up in cultured
homes with many books hold mainly in modern times,
among technologically advanced, prosperous, market
economies? Or did it also hold in past times and under
different political systems? The world of the 1940s
was very different; technology was simpler, nations far
poorer, and economies ravaged by war. The old world
was then transformed by the great post-war boom and
the rise of the Welfare State in the West, by Apartheid
in South Africa, and by Communism in the East. The
period covered by our data saw huge educational upgrad-

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 13
Fig. 4. History: predicted years of education for respondents coming of age in different time periods and in different types of nation. Regression
estimates separately for each time period and type of nation. Predicted values control other variables by whole population standardization, using the
population of that time period and type of nation as the reference population. 95% confidence intervals. 31 nations; N= 75,648.
Source:Table A.5.
ing throughout the world (Ganzeboom & Treiman, 1993;
Treiman & Yip, 1989). How did scholarly culture fare
in these very different conditions? Fig. 4 shows that the
advantage of books in the home is found throughout the
whole range of institutional arrangements covered here,
as predicted by Hypothesis 1 (details are in Table A.5).
Scholarly culture’s advantage goes back for genera-
tions, as far back as the memory of survey respondents
can take us, and in all political systems. (1) In the West,
the advantage was large before World War II (Panel A).
It remained large despite the post-war boom, economic
stagnation in the 1970s, and globalization at the end
of the century. (2) In Eastern Europe before Commu-
nism, scholarly culture was as important as in the West
(Panel B). The imposition of Communism was a “nat-
ural experiment” changing elites and their orientations
and presumably their membership signals, but scholarly
culture remained important: the educational result of a
generation of Communism was little different than a gen-
eration of Western free markets.15 (3) Being born into a
15 Many other studies have found this, for example (Ganzeboom &
Nieuwbeerta, 1999), although some suggest scholarly culture became
family with many books was a huge advantage in pre-
revolutionary China (Panel C). Then the Communists
under Mao seized power and later the Cultural Revolu-
tion followed – one of history’s most extreme, sustained,
and bloody attempts to undermine the advantages held by
elite families. But the advantage of coming from a family
rich in books remained no less than in pre-revolutionary
times, indeed far larger than in the West or in Communist
Eastern Europe. Only in the most recent period, when
market reforms began to spread in China, was there any
change and even then the effect remained large.16 (4)
Culture has been a valuable resource for non-white South
Africans17 as far back as the 1940s, with the advantage at
least as large as in Western Europe or in Eastern Europe
under Communism.
In short, scholarly culture is a great resource for the
oppressed, whatever their color, whomever their oppres-
sor, and whatever the historical circumstances.
somewhat less important toward the end of the Communist period.
16 For a detailed treatment of trends in literacy and their determinants
in China, see Treiman (2006).
17 Because the analysis is done separately by time period, we need to
combine the several non-white groups in order to have sufficient cases.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
14 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Table 3
Influences on respondent’s education in each nation separately: estimated gain in education (first differences) from multiple regression controlling
family background (Eq. (3)), with multiple imputation of missing data and bootstrapped standard errors. 31 societies, circa 1999. N= 77,758.
Nation Parents’ have 500
books vs. 1 book
95% C.I. Parents have 15 years
education vs. 3 years
95% C.I.
China (all) 6.6 6.2–7.0 3.1 2.6–3.6
China: Rural born 6.6 6.2–7.0 2.8 2.2–3.5
China: urban born 5.5 4.7–6.2 2.5 1.9–3.1
Chile 5.3 4.3–6.3 4.2 3.3–5.0
Spain 4.7 3.6–5.8 6.5 5.4–7.7
Norway 4.3 3.4–5.2 1.7 0.8–2.5
Portugal 4.1 3.3–5.0 4.2 3.2–5.2
Philippines 4.0 3.0–5.1 3.3 2.5–4.1
South Africa: Black 4.0 3.6–4.4 4.5 4.1–4.9
Israel 3.9 3.0–4.8 2.5 1.7–3.4
South Africa: Asian 3.8 3.0–4.7 3.1 2.3–3.9
Sweden 3.6 2.5–4.7 2.1 0.7–3.4
Hungary 3.5 3.2–3.8 3.0 2.6–3.4
Cyprus 3.5 2.3–4.7 3.7 2.8–4.5
South Africa (All) 3.5 3.2–3.8 4.8 4.5–5.0
France 3.4 2.6–4.2 2.1 1.4–2.8
Poland 3.3 3.0–3.7 2.6 2.1–3.0
Latviaa3.3 2.4–4.2 1.1 0.3–1.9
Slovakia 3.2 2.9–3.5 2.9 2.4–3.3
Germany-West 2.8 2.1–3.6 2.7 1.2–4.2
New Zealand 2.8 1.9–3.6 1.9 1.1–2.7
South Africa: Colored 2.8 2.0–3.5 3.9 3.2–4.7
Czech Republic 2.7 2.5–3.0 2.7 2.2–3.1
Australia 2.7 2.5–2.9 1.9 1.7–2.1
Bulgaria 2.7 2.4–3.0 3.8 2.5–4.2
Slovenia 2.7 1.8–3.5 3.0 2.0–4.1
Germany-East 2.6 1.5–3.7 −0.1 −1.7 to 1.5
Russia 2.6 2.3–2.9 2.3 2.0–2.6
United States 2.4 1.7–3.1 2.3 1.6–2.9
Japan 2.2 1.6–2.8 2.1 1.6–2.6
Netherlands 2.1 1.5–2.7 2.8 1.9–3.8
South Africa: White 1.8 1.4–2.1 2.9 2.3–3.4
Canada 1.6 0.7–2.5 1.6 0.7–2.5
All (mean) 3.7 3.0
Source: Calculated from Table A.6.
aFor clarity, the interaction between books and parents’ education is omitted; it is small in all societies, although statistically significant in many.
7.6. Country-by-country results
Growing up in homes with numerous books greatly
boosts educational success in every country in our sam-
ple (Table 3). For example in the US – where the
advantage is relatively modest – a child whose parents
have 500 books can expect to get about 2 or 3 years more
education than a comparable child from a bookless home.
The advantage in Australia and West Germany is similar;
not far different in Bourdieu’s France; larger in Norway
and Spain (4 or 5 years); and largest in China (6 or 7
years).
There is no evidence of a trade-off between books
and formal education (Fig. 5). On the contrary, par-
ents’ books tend to give a large advantage precisely
in countries where parents’ education also gives a
large advantage (r= .36, t= 2.08, p< .05, calculated with
nations as the unit of analysis). Spain, Chile, China, and
Portugal stand out as societies where both books and for-
mal education matter a lot, as they do for South African
blacks. On average, parents’ books matter a bit more than
parents’ education.
7.7. Extensions and sensitivity analyses
7.7.1. Measurement of occupational status
Status attainment models can be sensitive to the way
occupational status is measured so we re-estimated our
models using the well-known ISEI scores (Ganzeboom
& Treiman, 1996, 2003) instead of the Worldwide Status

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 15
Fig. 5. Effect of books and parents’ education. Impact on child’s edu-
cation: 1st differences from Table 3.
scores. But in fact, this makes no discernible difference
(Table A.7).
7.7.2. Functional form
Our model (Eq. (3)) is highly parametric, with strong
assumptions (albeit ones well justified by theory and pre-
liminary analyses) about measurement, functional form,
and interactions. A useful contrast is provided by “near-
est neighbor” methods developed in economics to assess
treatment effects with minimal parametric assumptions
(Abadie, Drukker, Herr, & Imbens, 2004); these are con-
ceptually similar to “hot deck” methods for imputing
missing data. They require dichotomizing the indepen-
dent variable, a major loss of information for continuous
variables like ours. Dichotomizing parents’ books near
the median, our parametric model implies a difference of
1.14 years of education for those growing up in families
above and below the median (with a standard error of
.024), other things being equal. Matching on the same
variables used in our model, Imbens’ nearest neighbor
estimate is 1.21 (with a standard error of .034). This is
comfortingly close.
Random measurement error is serious problem, lead-
ing to a variety of biases, as has long been known
(Blalock, 1965; Kelley, 1973). For the variables at hand,
the only comprehensive set of reliabilities available
with which to correct for attenuation are for Australia
(Table A.8, column 1). Using standard structural equa-
tion methods (Arbuckle & Wothke, 1999), we corrected
our key analysis on the assumption that the Australian
reliabilities would apply, at least approximately, to the
other nations (Table A.3, Panel B). This is not an
unproblematic assumption. However, ignoring measure-
ment error is equivalent to assuming that all variables
are measured without error, an even more problematic
assumption.
Both books and parents’ education are reliably mea-
sured, with a test–retest correlation of r= .76 over a five
year period. This compares to .81 for father’s occupa-
tional status and around .60 for other aspects of class.
Corrections for attenuation slightly increase the effect
of books (by 7%) and parents’ education (by 20%).
The (considerable) effect of father’s status is reduced
to zero. In uncorrected analyses, it seems to have been
picking up variance properly attributable to parents’ edu-
cation and perhaps books. The effect of GDP goes down
by almost half, probably for the same reason. Other
effects, all small to begin with, change modestly in both
directions.
Applying the Australian reliabilities to the Australian
data (N= 14,982), a less problematic matter, leads to
similar conclusions. Corrections for attenuation slightly
increase the effect of parents’ books (by 17%) and
parents’ education (by 2%). The effect of father’s occu-
pational status also increases slightly, unlike in the
international data. Other effects are all small to begin
with.
Overall, these analyses suggest that both parents’
books and parents’ education have strong, independent,
and about equally large effects on children’s education
even after correcting for attenuation due to random mea-
surement error. The effects are much larger than those
of any other class or background variable we have mea-
sured.
7.7.3. Omitted variables
The Australian data allow us to introduce addi-
tional controls not available in the international data
(Table A.8); they are based on 14,718 cases from a series
of representative national samples conducted between
1984 and 2003. For the matters at hand, Australia is unex-
ceptional, so there is a reasonable chance that the results
will generalize to other countries (see Tables 1 and 3,
Fig. 5, and Table A.3). In particular, the impact of par-
ents’ books on children’s education is much the same in
Australia (.69, corrected for attenuation) as in the pooled
sample (.61). Corrected for attenuation due to random
measurement error, our usual model implies that parents’
books have a standardized effect of .27 on children’s edu-
cation, net of other things (Table A.8, column 2). We take
this as the baseline and assess the effect of introducing
further controls.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
16 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
7.7.4. Academic ability and father’s scholarly
habitus
One might have legitimate concerns that home library
size was at least in part a proxy for respondent’s aca-
demic ability – on the argument that intelligent parents
buy more books than usual and have smarter children
than average. We measure academic ability by a stan-
dard multiple-item test adapted for survey use (reliability
r= .53 over a 5-year period). Alternatively, books might
be a proxy for father’s scholarly habitus, with fathers in
occupations where books are common acquiring schol-
arly skills and preferences that lead them both to buy
books and to further their children’s educational careers.
We measure habitus by the average home library size for
people in father’s 4-digit occupation.
Including these controls reduced the standardized
effect of home library size (.27 in the base model) by
about a quarter (to .20; Table A.8, column 3).18 Thus
the effects we have estimated in the main analysis would
probably remain at least three-quarters of their present
size, were able to control academic ability. So even
with this possible reduction, the effects remain large and
important.
7.7.5. Parents’ income and wealth
Controlling for parents’ income is not possible in the
usual models based on retrospective reports of parents’
characteristics because respondents’ reports of their par-
ents’ income – unlike their reports of parents’ education
and occupation – are not reliable. An alternative to the
retrospective approach is to take respondents as the par-
ents and get child data from their proxy reports of their
grown children’s education, confining the analysis to
children age 25 or older (column 4). We measure family
income by respondents’ reports of their current income,
which are reasonably reliable (r= .713), taking that as
a proxy for their income in the past when their children
were growing up. This suggests that parents’ income has
a modest effect on children’s education (.06). But even
with the income control, the effect of home library size
is significant and substantial (.15).
Further evidence comes from the OECD’s PISA study
of academic performance (reading and vocabulary tests)
among high school students in over 40 nations, with over
200,000 cases (OECD, 2002). It has a plausible retro-
spective measure of parents’ income and wealth based
on a series of questions about possessions in the home.
18 This is probably an over-correction since it seems likely that there
is some reciprocal influence between home library size and ability,
which is beyond the scope of the present analysis.
Fig. 6. Influences on home library size. Structural equation estimates
correcting for random measurement error; Australia, 1984–2003; N=.
In these data (Table A.8, column 5), parents’ income
and wealth has a substantial effect on children’s per-
formance, .13 in standardized terms (Kelley, Evans, &
Sikora, 2005). But even controlling for that as well as
parents’ education and occupation, home library size
remains important, .25; indeed it has by far the largest
effect.19
7.7.6. Respondent’s marks/grades in school
One of the ways in which a home library might
influence children’s achievement is by improving their
performance in school. Controlling for the usual vari-
ables plus academic ability and father’s scholarly
habitus, we find that books’ effect is indeed strong and
highly significant in Australia, as in the OECD data (col-
umn 6). In fact, it is second only to academic ability. This
lends further strength to our causal interpretation of the
association between scholarly culture and educational
attainment.
7.7.7. The origin of home libraries
Where do libraries come from – who acquires a large
library? The answer is unequivocal: a taste for books is
largely inherited (Fig. 6 and Table A.8, column 7). Par-
ents’ library size is by far the dominant influence on one’s
own home library size, .62 in standardized terms (see
also Crook, 1997a). Importantly, once other factors are
19 None of our data sets has information on personality, which might
be relevant. However, a study in the Netherlands using standard
multiple-item measures of the “big five” personality traits together
with standard stratification-related measures, finds family background
effects do not change when the personality variables are included (van
Eijck & de Graaf, 1995).

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 17
taken into account, education, occupational status, and
class are irrelevant. Prosperous people buy more books,
although not a lot more, .13. Academically able peo-
ple buy many more books, .18. Working in a scholarly
habitus also leads people to buy a few more books. Men
buy fewer than women. Net of other things, young peo-
ple growing up in recent years, when Australia’s GNP
was larger, themselves buy fewer books than comparable
people in the past.
Thus it seems that scholarly culture, and the taste for
books that it brings, flows from generation to generation
largely of its own accord, little affected by education,
occupational status, or other aspects of class. Academ-
ically able people, and those who work in a habitus
infused with scholarly culture, sometimes join in. But
the main story is continuity within the family. Parents
give their infants toy books to play with in the bath;
read stories to little children at bed-time; give books as
presents to older children; talk, explain, imagine, fanta-
size, and play with words unceasingly. Their children get
a taste for all this, learn the words, master the skills, buy
the books. And that pays off handsomely in school.
8. Summary
In sum, we find that parents’ commitment to schol-
arly culture, manifest by a large home library, greatly
enhances their children’s educational attainment. It
does so not only in the rich, long-democratic, market-
oriented nations of Western Europe and its overseas
extensions, but also in Eastern Europe, in Asia, in
South America, and in South Africa. The effect remains
strong after controlling for well-known sources of
educational advantage: parents’ education, father’s occu-
pation, father’s class and ownership situation, gender,
GDP when growing up, and nation. Moreover, the effect
is strong across the whole political spectrum – in every
one of our 27 countries and as far back in history as
our survey data can take us. It was strong in societies
whose educational systems were redesigned explicitly
to eliminate class privilege (Eastern Europe under Com-
munism, China during the Cultural Revolution). It was
strong among the underclass in a society designed to
maintain group privilege (South African blacks under
Apartheid).
Regardless of how many books the family already
has, each addition to a home library helps the children
get a little farther in school. But the gains are not equally
great across the entire range; instead they are larger at
the bottom, far below elite level, in getting children from
modest families a little further along in the first few
years of school. Moreover, having books in the home
has a greater impact on children from the least edu-
cated families, not on children of the university educated
elite.
Sensitivity analyses in one country where richer data
are available (Australia) suggest that the strong impact of
a home library persists even after controlling for random
measurement error, academic ability, father’s habitus,
and parents’ income. Moreover it suggests that schol-
arly culture, and the taste for books that it brings, persists
from generation to generation within families largely of
its own accord, independent of education and class.
9. Theoretical implications
Our preferred scholarly culture theory leads to three
predictions, almost all supported by the evidence pre-
sented here. First, because scholarly culture provides
skills and knowledge that are central to literacy and
numeracy, and hence valuable in schools everywhere,
it implies that parents’ participation in scholarly cul-
ture will enhance children’s educational attainment in
all societies, net of the parents’ formal education and
social class (Hypothesis 1). As we have seen, the evi-
dence strongly supports this hypothesis. Moreover it
also suggests that social and economic policies have
little effect on the advantages conferred by scholarly cul-
ture; instead, the advantage is large in all nations, at all
times, under all political regimes. This does not mean
that regimes do not try to use education to produce ide-
ological conformity, but rather it suggests that scholarly
culture confers skill sets that are valued regardless of
the dominant ideology, something diaspora peoples have
long known.
The results also support our prediction that an
increase in scholarly culture has the greatest impact on
children from families with little scholarly culture to
begin with (Hypothesis 2). It is at the bottom, where
books are rare, that each additional book matters most,
not among the literate elite: each additional book yields
more “bang for your book” among the book-poor than
among the book-rich.
Further, scholarly culture matters more if parents are
poorly educated, but matters less if parents are well-
educated (Hypothesis 3). This is true for education as
a whole (measured in years) and in the early stages of a
child’s educational career, around year 9. But, contrary
to our hypothesis, it is not true for completing secondary
school and quite the reverse is true for university.
All this suggests that scholarly culture provides skills
and competencies that are useful in school, or that it
reflects a preference for and enjoyment of books and
reading that makes schooling congenial, or enjoyable.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
18 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
This may be one contributing reason why school reform
is such a challenge. Ever since the original Coleman
report, school effects on academic achievement are often
found to be weak (Pong & Hao, 2007; Schneider &
Keesler, 2007), dashing, or at least sobering, hopes that
newer statistical methods would reveal substantial but
previously undetected effects, and hence identify points
of leverage. Note that our results do not in any way imply
that formal schooling cannot compensate for the absence
of scholarly culture in the home; but they do highlight
the fact that children from homes lacking in scholarly
culture may require special attention.
All this suggests that scholarly culture provides skills
and competencies that are useful in school, or that it
reflects a preference for and enjoyment of books and
reading that makes schooling congenial, or both.
9.1. Cultural mobility
The cultural mobility thesis proposes that cultural
skills and knowledge are not a monopoly of the elite nor a
means of preserving privilege from one generation of the
elite to the next (as Bourdieu would have it) but instead
are widely available and perhaps especially valuable for
children from modest families, contributing to their edu-
cational success and upward mobility (Aschaffenburg &
Maas, 1997; Blaskó, 2003; DiMaggio, 1982; Kingston,
2001). This is a descriptive claim, without consensus
among its adherents as to why culture is important, in
what nations, or in which historical circumstances.20
Our data clearly fit this description: most of culture’s
effects are independent of family background and espe-
cially beneficial to children from modest families. This
has important policy implications, suggesting that cul-
ture should be encouraged in schools, not scorned as
irrelevant or as merely a tool for preserving elite privi-
lege.
However, that is not the whole story: culture simulta-
neously helps to transmit advantage from one generation
to the next. Further analysis (not shown) suggests that
25–35% of the advantage of growing up in a well-
educated, high status family comes about indirectly
because such families provide a richer cultural environ-
ment for their children, which in turn gives the children
an enduring advantage in school. So scholarly cul-
ture produces both (much) cultural mobility and (some)
inherited privilege.
20 It is consistent with estimates that about 25% of the variance in
school marks in the US is attributable to family factors other than aca-
demic ability and the usual stratification measures (Teachman, 1996).
10. Rejected alternative theories
10.1. Elite closure/cultural capital in the West
What does the evidence say about applying Bour-
dieu’s “elite closure”/“cultural capital” theory about
elites hoarding advantages by using essentially arbi-
trary cultural signals to recognize fellow members and to
exclude others (Bourdieu, 1984; Goblot, 1925 [1973]) to
the particular case of scholarly culture and educational
attainment? Bourdieu envisions an elite, as arguably in
France, with enough power to get its way in schools and
universities; with a preference for furthering the inter-
ests of its own members; and living in a society where
elite membership is recognizable by involvement in the
scholarly culture. This is a plausible description of many
developed Western nations, and so Bourdieu’s thesis –
like ours – implies that in these nations children from
families with many books will get more education. And
this they do.
Bourdieu’s theory also implies that the payoffs go
to children of the elite. But we have seen that children
from poorly educated homes where books are scarce in
fact benefit greatly, indeed more so than children from
elite homes. For example, a child whose parents had no
books and only a few years of education – about as far
from the elite as you can get in the modern world –
could expect to get 7.6 years of education, other things
being equal. But if their parents had 25 books, that
would rise to 9.8 years, more than 2 years higher. Thus,
just a hint of scholarly culture conveys a large advan-
tage to children from modest families. It is hard to see
how this has anything to do with elite gate keeping:
the parents might number among the respectable poor,
but not among the elite. And the evidence in general is
that scholarly culture’s gains are manifest for children
from families on the lower rungs of the social hierarchy,
indeed generally greater for them than for children from
higher ranking families. Elite closure cannot account for
that.
It remains possible that elite closure operates in addi-
tion to other (more powerful) factors at university level.
A small but statistically significant interaction points to
scholarly culture mattering more for children from well-
educated, possibly elite families. This is consistent with
elite closure, but also with other hypotheses, for example
the argument that educated parents are expert guides to
the educational system, focusing their children’s skills
effectively (Lucas, 2001). In general, if completing uni-
versity requires a combination of traits – say, ability,
motivation, and financial resources – then there will be
a positive interaction. Ability will matter most where

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 19
there is also sufficient motivation and money; motivation
will matter most when it is matched with ability; and so
forth.
10.2. Elite closure/cultural capital under
Communism
If Bourdieu and other elite theorists from Mosca to
Mills are right in thinking that ruling elites have power
over schools and universities in (mostly pluralist and
democratic) Western nations, then ruling elites in (cen-
tralized and authoritarian) Communist societies must
have at least as much power, probably more. But when
Communists were in power they did not favor children of
the old regime, instead Communist elites favored work-
ers, peasants, children from Communist families, and
other politically approved groups. If home library size
and the natural expressions of a way of life steeped in
scholarly culture were simply signals, arbitrary markers
of membership in the old elite, then they would certainly
not have been rewarded under Communism (Connelly,
2000; Unger, 1982).
Thus, our finding that books in the home was just
as much an advantage in different historical periods in
Eastern Europe as in the modern West, and even more of
an advantage in China (in both the urban and rural sec-
tors), is especially revealing. First, it suggests that there
is an intrinsic advantage in growing up around books, an
advantage so great that it prevailseven in the face of indif-
ference or outright hostility from ruling elites, as during
the Cultural Revolution in China. That the advantage is
present for non-whites in the face of the white elite’s
Apartheid policies in past decades in South Africa is
another striking example.
Second, since the advantage shows up in all Com-
munist nations throughout their varied and checkered
history, as well as in South Africa and the West, it is
unlikely to be something mainly determined by govern-
ment policies, even dramatic ones. Instead, it must reflect
some fundamental cause common to a wide range of
societies and diverse historical periods. We propose that
it is an indicator of a cognitive toolkit that even the most
ideologically driven societies cannot afford to ignore.
Third, since the advantage shows up just as clearly
under Communism, where it cannot be attributed to the
elite, as it does in the West, Occam’s Razor suggests there
is no need to follow Bourdieu and attribute the advantage
in the West to elite machination.
We suggest that it is time to doubt Bourdieu’s elite clo-
sure/cultural capital explanation of why scholarly culture
matters to status attainment in the West. No serious anal-
ysis has ever supported the stronger forms of this thesis
(Kingston, 2001)21. But the idea is so attractive that it has
been tempting to dismiss contrary results as reflecting
the peculiarities of the society investigated (most stud-
ies deal with a single country) or imperfections of data
or method. But now that the theory’s inadequacy has
been demonstrated for dozens of societies, with diverse
elites and varied institutions in diverse historical periods,
the accumulation of contrary evidence should persuade
its adherents to consider the alternatives. This does not
rule out the possibility that elites use other “secret hand-
shakes”, but it is no longer plausible to interpret scholarly
culture’s impact on education as one of them.
11. Conclusion
We suggest that the standard model of educational
attainment should be extended to include parents’ schol-
arly culture, as measured by the number of books in the
parents’ home. Scholarly culture has a powerful impact
on children’s education throughout the world, in rich
nations and in poor, under Communism and under capi-
talism, under good governments and under bad, in the
present generation and as far back in history as now
living memory can take us. It helps children from all lev-
els of the social hierarchy, but especially those from the
bottom. A book-oriented home environment, we argue,
endows children with tools that are directly useful in
learning at school: vocabulary, information, comprehen-
sion skills, imagination, broad horizons of history and
geography, familiarity with good writing, understanding
of the importance of evidence in argument, and many
others. In short, families matter not just for the material
21 Another weak variant of the thesis runs thus: Elites choose the
content of “sorting” examinations in the educational system to suit
themselves and therefore the content of the books that will be relevant
to educational success (for example, Communists choose Marx, and
religious elites choose St. Thomas Aquinas). We would not deny that
they try! What our findings imply is that the cognitive gains from grow-
ing up in the scholarly culture are large enough that they dominate such
attempts. The fact that the results reveal a substantial scholarly culture
effect for black South Africans growing up under Apartheid (whose
home libraries were surely not full of pro-Apartheid books) is strongly
contrary to a theory implying that only books consistent with the dom-
inant culture produce educational success. Similarly, the strong effect
in China shortly after the Communist revolution (when few of these
books would have been Communist classics) also supports the view
that scholarly culture, regardless of its ideological orientation, confers
skills that lead to educational success, regardless of the regime’s ide-
ological orientation. All this suggests that scholarly culture provides
skills and competencies that are useful in school, or that it reflects a
preference for and enjoyment of books and reading that makes school-
ing congenial, or both. This is not to say that the choice of reading
matter does not vary by class, as it demonstrably does, for example in
newspaper readership (Chan & Goldthorpe, 2007).

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
20 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
resources they provide, not just because of parents’ for-
mal educational skills, but also – often more importantly
– because of the scholarly culture they embody.
Acknowledgements
Development of the World Inequality Study database
was supported by the Australian Research Council’s
Research Infrastructure and Equipment Facility (REIF)
grant # R00002808 to the Melbourne Institute of Applied
Economic and Social Research, University of Mel-
bourne. This paper was supported in part by a grant to
Evans from the Nevada Agricultural Experiment Sta-
tion in connection with CSREES W2001. We thank Greg
Elliot, Calvin Goldscheider, Max Haller, Patrick Heller,
Dennis Hogan, John Modell, Mark Suchman, Michael
White, and Krzysztof Zagorski for their comments.
Appendix A.
See Tables A.1–A.8
.
Table A.1
Measurement.
Books in the parent’s home is measured as described in the text. We analyze the natural log (counting “no books” as 1).
Education is measured in years, with adjustments described in the text. Year 9,Secondary, and University are dummy variables scored 1 for those
who completed that stage or higher and 0 for everyone else. Year 9 is around the end of compulsory education in many European nations and,
even in recent decades, many end their schooling then; in poor countries and past generations, many did not even go that far in school.
Secondary schooling finishes at 12 years in most nations. We define 16 years of education or more as university completion except for
British-model educational systems (Australia, Cyprus, Great Britain, India, Ireland, New Zealand, Northern Ireland, and South Africa) where an
ordinary university degree takes 1 year less. Russia has a compressed educational system with university following 10 years of primary and
secondary schooling, rather than the 12 years usual in most countries, so we treat it like the British systems.
Parents’years of education were measured by questions appropriate to each nation and then recoded in the same way as respondent’s years of
education.
Interaction: Our main model uses an interaction = (ln number of books)×(parents’ education)
Father’s occupational status is coded from the surveys’ original 4 digit International Standard Classification of Occupations (ISCO) or similarly
detailed country-specific scores into Treiman’s 14 category Standard International Classification of Occupations, and thence into Worldwide
Status Scores (Kelley, 1990; Treiman, 1977). These scores order the 14 Treiman categories from 0 (farm laborers) through 1.0 (higher
professionals). These scores are often used in international research (e.g. Kelley & Evans, 1995; Sikora, 2005) and in the US are essentially
interchangeable with Duncan SEI scores.
The self-employed are all those working in their own business, farm, or professional practice, regardless of whether they are incorporated or not
and regardless of whether the self-employed person receives their compensation in the form of salary, profit, or dividend. We define the father
petit bourgeois as self-employed without employees and father owner as self-employed with employees that they supervise (Robinson & Kelley,
1979); these are mostly small entrepreneurs. These are treated as two separate dummy variables.
Father supervisor is scored 1 for fathers who supervise others at work and zero for fathers who do not, based on a direct question.
Gender is a dummy variable scored 1 for men and 0 for women.
GDP per capita when respondent was age 15 is measured at parity-purchasing-power and scored as an index with USA in 1995= 1. GDP data for
recent years are from the Penn World Table Version 6.1 (Heston, Summers, & Aten, 2002). The projection back into the past is based on Banks’
cross-national time series (Banks, 1976), calibrated to Penn World Table definitions using the years that overlap, together with further
assumptions by the authors. Some of our assumptions are heroic, but we believe our estimates are the best yet available. Technical details and a
downloadable version of the database are available on our website, www.international-survey.org.
Eastern Europe. Residence in an Eastern European country is a dummy variable, scored 1 for Eastern Europeans and 0 otherwise. Preliminary
analysis suggested that no other country groups require special treatment.
Table A.2
Books in parents’ home and child’s education: descriptive means and percentages, not adjusting for other differences. 31 societies, circa 1999.
Books in parents’ home: Difference: 500 −none
None Around 10 Around 25 Around 75 Around 500+
Panel A: all nations
Education (mean years) 6.8 9.0 10.8 12.2 13.7 7.0
Grade 9 or more 31% 57% 79% 90% 96% 65%
Secondary or more 12% 24% 40% 58% 78% 66%
University 3% 5% 9% 19% 36% 34%
(Cases) 7,309 16,937 11,655 23,816 13,070

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 21
Table A.3
Influences on education: multi-level linear and probit regression models (Panel A) and structural equation estimates correcting for random measurement error using Australian reliabilities (Panel
B). Multiple imputation of missing data done separately for each nation. Pooled data from 31 societies. N= 77,758.a.
Variable A. Multi-level linear and probit estimates (metric coefficients and t-statistics) B. Metric structural equation estimates (years of education)
Years of education Complete Year 9+ Complete secondary+ Complete university World Change Australia N= 14,982 Change
btbtbt bt Raw Cor. Raw Cor.
Parents’ books (log) .78 65 .24 36 .16 25 .12 13 .55 .61 .07 .52 .69 .17
Parents’ education (years) .38 60 .13 38 .08 25 .04 9.27 .47 .20 .19 .21 .02
Father’s occupational status 1.41 25 .48 14 .44 17 .52 19 1.31 .09 −1.22 1.64 1.68 .05
Father owner (yes vs. no) .22 5ns 1.08 4ns3.22 .62 .40 .03 .02 −.01
Father petty bourg (yes vs. no) ns 1ns −3ns 1ns 3nsns–−.13 −.39 −.26
Father supervisor (yes vs. no) .40 10 .20 8.17 9.10 5.27 −.52 −.79 .30 −.05 −.35
GDP when R young .90 28 .35 22 .19 11 ns 3.74 .36 −.38 – – –
Eastern Europe (yes vs. no) 1.11 20 .31 11 ns 3 ns 01.43 1.06 −.38 – – –
Male (yes vs. no) .59 28 .22 20 .14 14 .19 16 .58 .57 −.01 .41 .46 .05
(Interaction: row 1 ×row 2) −.036 −26 −.008 −9ns 1 .006 6––––– –
(constant) 6.69 89 −.46 −12 −1.28 −33 −2.33 −49 6.91 5.25 −1.66 6.46 5.76 −.69
(std error of ui) (1.12) – (.58) – (.48) – (32) – – – – – – –
(rho)b(.14) – (.25) – (.19) – (.09) – – – – – – –
(R2or pseudo R2)c(.44) – (.30) – (.25) – (.16) – (.44) (.49) (.24) (.27)
ns, not significantly different from zero at p< .001, two-tailed.
aFive multiple imputations were used for missing data in Panel A. Reliabilities are from Table A.8.
bProportion of total variance contributed by second level variation.
cCorrelation between the predicted value and the dependent variable, squared.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
22 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Table A.4
Influence of books in the home and parents’ education on respondent’s education, raw and adjusted by the models of Table A.3 using whole population standardization. 31 societies, circa 1999.
Panel A. Actual (mean or proportion) Difference
500 −0
books
Panel B. Predicted (mean or proportion) Difference
500 −0
books
None Around
10
Around
25
Around
75
Around
500
None Around
10
Around
25
Around
75
Around
500
1. Years of education
Parents: university 12.6 13.4 14.0 14.6 15.2 2.6 12.7 13.1 13.3 13.5 13.8 1.1
Parents: secondary 11.3 11.8 12.3 13.2 14.0 2.7 11.1 11.9 12.2 12.6 13.2 2.1
Parents: 7–9 years 9.6 10.4 11.3 12.1 12.9 3.3 10.0 11.0 11.4 11.9 12.8 2.8
Parents: primary 8.5 9.3 10.4 11.2 12.2 3.8 8.8 10.1 10.6 11.2 12.3 3.5
Parents: up to 3 years 5.0 7.0 8.9 10.0 10.8 5.8 7.6 9.2 9.8 10.6 11.9 4.3
2. Complete year 9
Parents: university .84 .94 .97 .99 .99 .15 .89 .94 .95 .97 .98 .09
Parents: secondary .82 .91 .95 .97 .99 .17 .78 .88 .91 .93 .97 .19
Parents: 7–9 years .61 .76 .86 .92 .95 .34 .66 .81 .85 .89 .95 .29
Parents: primary .41 .55 .71 .81 .86 .46 .53 .72 .78 .84 .92 .39
Parents: up to 3 years .17 .37 .61 .74 .72 .55 .40 .61 .69 .77 .88 .48
3. Secondary school
Parents: university .74 .82 .84 .88 .92 .18 .49 .65 .71 .78 .87 .38
Parents: secondary .49 .61 .64 .75 .84 .35 .36 .52 .59 .66 .77 .41
Parents: 7–9 years .23 .33 .43 .56 .68 .45 .28 .42 .48 .56 .68 .40
Parents: primary .14 .22 .34 .44 .57 .43 .20 .33 .38 .45 .58 .37
Parents: up to 3 years .07 .13 .23 .35 .46 .39 .14 .24 .29 .35 .47 .33
4. University
Parents: university .21 .32 .38 .48 .58 .37 .08 .18 .23 .30 .45 .37
Parents: secondary .11 .13 .16 .26 .37 .26 .06 .13 .17 .22 .34 .28
Parents: 7–9 years .04 .07 .10 .17 .26 .22 .05 .10 .13 .17 .26 .21
Parents: primary .03 .05 .08 .12 .22 .19 .04 .07 .09 .12 .19 .15
Parents: up to 3 years .02 .02 .04 .09 .15 .13 .03 .05 .07 .09 .13 .10
Number of cases (Panel A) or 95% C.I. for predicted years of education (Panel B)
Parents: university 19 108 149 1,216 2,967 12.6–12.8 13.0–13.2 13.2–13.3 13.4–13.5 13.7–13.9
Parents: secondary 92 746 1,176 5,204 4,406 11.0–11.2 11.8–12.0 12.2–12.3 12.5–12.6 13.2–13.3
Parents: 7–9 years 1,027 4,918 4,847 10,958 4,246 9.9–10.0 11.0–11.03 11.4–11.4 11.9–11.9 12.7–12.8
Parents: p×rimary 2,064 5,156 3,286 4,478 1,020 8.7–8.8 10.1–10.12 10.6–10.6 11.2–11.3 12.3–12.4
Parents: up to 3 years 3,876 5,463 1,877 1,366 205 7.5–7.7 9.1–9.2 9.8–9.9 10.5–10.6 11.8–12.0

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 23
Table A.5
Predicted years of education for respondents coming of age in different historical periods and different types of nation: regression estimates with multiple imputation of missing data. 31 societies;
N= 75,648.
Type of society A. Predicted years of education
other things being equalaand
95% confidence intervals
B. OLS Regression equation, with multiple
imputation of missing valuesb
One
book
95% CI 500
books
95% CI Difference,
500
books −one
book
95% CI Parents’
books
(ln)
t Parents’
educa-
tion
(years)
t Father’
occupa-
tion
(0–1)
Father
owner
Father
petty
bour-
geoisie
Father
super-
visor
GDP when R
young (ln
index, USA
1990 = 1)
Male (Constant) R-squared (Cases)
A. Market economies
Up to 1949 8.0 7.8–8.2 11.1 10.9–11.23.1 2.8–3.4 .50 20.4 .30 20.5 1.9 ns ns 0.5 1.2 0.9 6.2 .37 6,937
1950–1964 9.2 9.0–9.3 12.3 12.2–12.53.1 2.9–3.4 .53 22.9 .31 24.6 1.7 ns ns 0.2 0.5 0.6 6.2 .34 8,788
1965–1979 10.5 10.4–10.713.3 13.3–13.42.8 2.6–3.04 .47 23.9 .24 25.1 1.3 0.4 ns ns ns 0.3 7.5 .26 11,136
1980 on 11.7 11.4–12.014.2 14.0–14.32.5 2.1–2.8 .41 11.9 .28 19.8 0.9 ns ns ns ns ns 8.4 .28 4,234
B. Eastern Europe
Up to 1949c8.4 8.3–8.5 11.5 11.3–11.73.1 2.8–3.3 .49 23.9 .30 20.1 1.5 ns ns ns ns 1.5 5.0 .34 7,632
1950–1964 9.9 9.8–10.012.9 12.8–13.03.0 2.8–3.2 .49 28.0 .20 17.7 1.2 ns -0.4 0.7 ns 0.6 8.5 .29 10,254
1965–1979 10.8 10.7–11.013.5 13.4–13.62.7 2.5–2.9 .44 28.9 .22 22.6 0.9 ns ns 0.4 -0.3 ns 8.0 .27 12,540
1980 on 10.8 10.5–11.013.1 13.0–13.32.3 2.0–2.7 .38 14.0 .20 12.0 0.8 ns ns 0.5 ns -0.3 8.2 .27 3,689
C. China
Up to 1949 2.6 2.3–3.0 8.8 7.6–10.06.2 4.9–7.4 1.00 11.8 .39 11.8ns ns ns ns –d2.4 .6 .39 598
1950–1964 3.5 3.3–3.8 10.6 10.1–11.17.1 6.5–7.7 1.15 22.5 .29 6.4 1.5 ns ns ns – 1.7 2.0 .44 1,361
1965–1979 4.9 4.6–5.2 11.9 11.6–12.37.0 6.4–7.6 1.18 27.8 .15 5.7 1.8 ns ns ns – 1.5 3.2 .39 2,490
1980 on 6.2 5.7–6.6 10.7 10.3–11.14.5 3.8–5.3 .80 13.7 .20 7.1 1.4 ns ns ns – 1.0 4.2 .38 1,107
D. South Africa: Non-white
Up to 1949 3.8 3.6–4.1 8.3 7.5–9.1 4.5 3.5–5.3 .65 8.4 .44 11.1 2.4nsns ns–
d0.8 2.0 .39 885
1950–1964 5.4 5.2–5.7 10.2 9.7–10.64.8 4.2–5.3 .75 13.7 .36 11.1 2.0 ns ns ns – 0.8 3.1 .39 1,245
1965–1979 7.4 7.2–7.7 10.9 10.5–11.3.5 3.0–4.0 .54 14.7 .32 14.8 2.1 ns 0.8 ns – 0.6 4.7 .36 2,310
1980 on 7.9 7.3–8.6 11.7 11.1–12.3.8 2.7–4.9 .62 8.1 .25 5.5ns ns ns ns – ns 5.9 .30 442
[4] White South Africans are omitted.
ns, italicized coefficients are not significantly different from zero at p<.01, two-tailed (or p< .001 for samples over 5000).
aOther variables evaluated at the mean, separately for each type of society. For example, father’s occupation in Panel A is evaluated at the mean for market economies.
bThe predicted values in Panel A are from equations including an interaction between parents’ books and parents’ education. For clarity, this interaction is omitted from the equations in Panel
B; hence the coefficient for parents’ books reflects its unique effects, and similarly the coefficient for parents’ education reflects its unique effects.
cExcept Russia, which was already Communist
dOmitted in single nation analyses, as too few groups for reliable estimates.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
24 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
Table A.6
Influences on respondent’s education in each nation separately: OLS regression estimates with multiple imputation of missing data. 31 societies, circa 1999. N=77,758.
Nation Parents’
books
(ln)
tParent’s
educa-
tion
tFather’s
occupa-
tion
Father
owner
Father
petit
bour-
geois
Father
supervi-
sor
GNP when 15 Male (Constant) (R-squared) Cases
China: all 1.13 (41.91) .17 (9.36) 1.53 ns −1.09 ns 1.08 1.50 6.69 (.46) 5,556
China: rural born 1.09 (33.6) .17 (6.2) 1.27 ns ns ns 1.65 1.90 ns (.42) 3,740
China: urban born .98 (20.0) .15 (6.4) ns −1.28 −1.68 ns ns .76 ns (.40) 1,816
Chile .93 (11.8) .32 (9.6) 1.64 ns ns ns 3.57 ns 12.0 (.53) 1,192
Spain .82 (10.1) .50 (11.7) ns ns ns ns 2.77 .88 ns (.59) 1,016
Cyprus .72 (8.0) .29 (7.4) ns ns ns .95 2.85 .51 ns (.61) 874
Philippines .67 (7.5) .26 (7.4) 2.38 ns ns ns 2.25 ns 12.8 (.32) 966
Norway .66 (8.7) .14 (3.8) 1.76 ns ns ns ns ns ns (.26) 1,100
Israel .66 (9.9) .21 (5.4) ns ns ns ns ns ns 8.3 (.30) 952
Portugal .65 (10.2) .39 (10.0) 2.16 ns ns 1.20 2.30 .98 6.5 (.62) 1,019
South Africa: Asian .63 (8.5) .23 (6.6 ns ns .68 ns 2.56 1.53 9.1 (.46) 658
South Africa: Black .60 (20.3) .36 (19.9) 1.67 ns .29 ns 1.95 .47 7.0 (.44) 3,517
Sweden .58 (6.7) .19 (3.8) ns ns ns ns 1.95 ns 8.9 (.25) 964
Hungary .57 (24.0) .26 (15.0) .84 −.48 −.38 .52 .53 .79 7.6 (.42) 5,461
South Africa: all .56 (27.03) .40 (36.23) 1.36 .65 .83 1.00 1.14 .70 5.81 (.57) 6,992
France .56 (8.8) .17 (6.3) 1.40 ns ns .13 1.30 ns ns (.26) 1,783
Latviaa.53 (7.8) .09 (2.9) ns ns ns ns ns ns 9.3 (.18) 951
Poland .52 (20.0) .21 (12.0) 1.46 ns ns ns 1.39 .22 10.1 (.40) 4,103
Slovakia .52 (21.4) .23 (11.7) .92 ns ns ns .55 .83 8.6 (.31) 5,318
Slovenia .45 (7.2) .26 (5.3) ns ns ns ns .91 .54 8.3 (.34) 893
Australia .44 (29.1) .16 (17.3) 1.70 ns ns .31 2.60 .50 9.2 (.31) 14,982
Russia .44 (15.1) .18 (11.2) .85 ns ns .47 ns ns 9.9 (.28) 5,622
South Africa: Colored .44 (5.8) .35 (10.2) ns ns ns ns .96 ns 6.0 (.47) 707
Germany-West .43 (7.0) .33 (6.9) 1.48 ns ns .16 .80 .13 6.5 (.35) 829
New Zealand .42 (7.2) .16 (5.3) ns ns ns ns 2.59 ns 11.2 (.28) 966
Germany-East .42 (4.7) ns ns ns ns ns 1.03 .72 .62 ns (.17) 456
Bulgaria .42 (17.4) .32 (21.4) ns ns ns ns .29 .30 8.1 (.41) 4,450
Czech Republic .42 (19.6) .23 (13.7) 1.69 ns ns .31 .20 1.10 7.5 (.29) 6,861
United States .40 (7.1) .19 (6.7) 1.66 ns ns ns ns ns 9.3 (.28) 1,095
Japan .35 (7.7) .17 (7.3) .98 ns ns ns 1.03 .74 10.5 (.43) 1,103
Netherlands .35 (6.3) .23 (6.0) 1.98 ns ns ns 2.29 .74 ns (.20) 1,434
South Africa: White .27 (8.2) .24 (10.9) 1.13 .34 ns ns .50 .62 8.4 (.27) 2,110
Canada .25 (3.5) .14 (3.6) 1.99 ns ns ns 1.28 ns ns (.19) 820
ns, italicized coefficients are not significantly different from zero at p<.01, two-tailed (or p< .001 for samples over 5000).
aFor clarity, the interaction between books and parents’ education is omitted; it is small in all societies, although statistically significant in many. Thus the coefficient for parents’ books reflects
its unique effects, and similarly the coefficient for parents’ education reflects its unique effects.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 25
Table A.7
Alternative measures of father’s occupational status: ISEI scores and Worldwidescores. Estimated gain in education from various sources. Multi-level
linear and probit regression models. Pooled data from 31 societies; N= 49,294 cases with complete information on all variables.a.
Variable (and comparison) Treiman-Ganzeboom ISEI Kelley Worldwide scoresb
bt bt
Parents’ books (ln; 500 books vs. 1 book) .88 62 .88 62
Parents’ education (15 years vs. 3 years) .44 56 .43 55
Father’s occupation (higher professional vs. farm laborer) .03 27 .02 31
GDP when R young (USA vs. China) .73 16 .73 16
Eastern Europe (yes vs. no) .87 11 .93 12
Male (yes vs. no) .40 16 .40 16
(Interaction: row 1 ×row 2) −.05 −29 −.05 −30
(constant) 5.74 57 6.36 66
(std error of ui) (1.14) – (1.14) –
(rho)c(.16) – (.16) –
ns, not significantly different from zero at p< .001, two-tailed.
aProbit estimates for completion of year 9, high school completion, and university completion also show that the Treiman-Ganzeboom and Kelley
status scores are virtually interchangeable. Details are available on request.
bPredicted values using the Worldwide scores are correlated r= .995 with predicted values using ISEI scores.
cProportion of total variance contributed by second level variation.
Table A.8
Extensions and sensitivity analyses: structural equation models correcting for attenuation due to random measurement error, estimated by maximum
likelihood; standardized coefficients. Persons 25 and older. Australia, 1984–2003 (column 5 is OECD PISA data for 43 nations).
Independent variable Test–retest
reliabilitya
R’s
education
R’s
education
Son or daughter’s
education
PISA: marks
in school
Australia: Marks
in school
R’s books
nowb
(1) (2) (3) (4) (5) (6) (7)
Parents’ books (ln, metric) .756 .27 .20 .15 .25 .13 .62
Parents’ education .752 .14 .13 .28 .07 ns ns
Father’s occupation .809 .16 .24 .10 .17 ns ns
Father owner .600 ns .08 – – ns ns
Father petit bourgeois .485 ns ns – – ns ns
Father supervisor .582 ns ns – – ns ns
Male .988 .09 .06 –.04 – −.07 −.11
GDP when young –c.29 .36 ns – .09 −.18
Academic ability (IQ) .527 – .32 ns – .27 .18
Father’s habitusd–c– ns – – ns .09
ln incomee.713 – – .06 .13 – .13
(R-squared) – (40%) (50%) (20%) (22%) (77%) (61%)
(Cases) – 14,718 11,248 3,402 228,170 11,248 8,260
ns, coefficients in italics are not significantly different from zero at p<.001 (for large samples) or p< .05 (for samples under 5,000).
aTest–retest reliabilities are based on approximately 1,150 cases,varying slightly due to missing data, over a 5 year interval. The income reliability
is over 2.7 years, based on 6,474 cases. For intelligence, the reliability is Cronbach’s alpha, based on 16,090 cases. The reliability for respondent’s
education is .868.
bBooks in respondent’s home (ln). For this column, all variables except Row 1refer to respondent’s characteristics, not to their parents’
characteristics. Thus, row 2 is respondent’s education, row 3 respondent’s occupation, and so on. See Fig. 6 for details.
cReliability not known; we have conservatively assumed .900.
dAverage number of books owned for those in father’s (4-digit ISCO) occupation.
eNot available for the usual analysis based on respondents’ reports of their own education and their parents’ SES. Available only for alternative
analyses based on respondents’ reports of their adult (age 25+) sons’ and daughters’ education and respondents’ own SES in Col. 9. The reliability
shown is therefore for respondents’ own family income, not respondents’ reports of their parents’ income. For the OECD analysis, parents’ income
is proxied by a multiple-item scale of possessions.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
26 M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx
References
Abadie, A., Drukker, D., Herr, J. L., & Imbens, G. W. (2004). Imple-
menting matching estimators for average treatment effects in stata.
Stata Journal,4, 290–311.
Allison, P. D. (2000). Multiple imputation for missing data: A caution-
ary tale. Sociological Methods and Research,28, 301–309.
Arbuckle, J. L., & Wothke, W. (1999). Amos 4.0 user’s guide. Chicago:
SmallWaters Corp.
Aschaffenburg, K., & Maas, I. (1997). Cultural and educational
careers: The dynamics of social reproduction. American Socio-
logical Review,62, 573–587.
Banks, A. S. (1976). Cross-national time series, 1815–1973. [machine
readable datafile]. Ann Arbor, Michigan: Inter-university Consor-
tium for Political and Social Research.
Bidwell, C. E. (1989). The meaning of educational attainment.
Research in Sociology of Education and Socialization,8, 117–138.
Blalock, H. M. (1965). Some implications of random measurement
error for causal inferences. American Journal of Sociology,71,
37–47.
Blaskó, Z. (2003). Cultural reproduction or cultural mobility? Review
of Sociology of the Hungarian Sociological Association,9, 5–26.
Bourdieu, P. (1984). Distinction: A social critique of the judgment of
taste. Cambridge, MA: Harvard University Press.
Cameron, S. V., & Heckman, J. J. (1998). Life cycle schooling and
dynamic selection bias: Models and evidence for five cohorts of
American males. Journal of Political Economy,106, 262–333.
Chan, T. W., & Goldthorpe, J. H. (2007). Social status and newspaper
readership. American Journal of Sociology,112, 1095–1134.
Connelly, J. (2000). Captive university: The sovietization of East Ger-
man, Czech, and Polish higher education, 1945–1956. Chapel Hill:
University of North Carolina Press.
Crook, C. J. (1997a). Cultural practices and socioeconomic attain-
ment. Westport, CT: Greenwood Press.
Crook, C. J. (1997b). The dimensionality of stratification-related cul-
tural practices in Australia. Australian and New Zealand Journal
of Sociology (now the Journal of Sociology),33, 226–238.
de Graaf, N. D., de Graaf, P. M., & Kraaykamp, G. (2000). Parental
cultural capital and educational attainment in the Netherlands: A
refinement of the cultural capital perspective. Sociology of Educa-
tion,73, 92–111.
de Graaf, P. M. (1986). The impact of financial and cultural resources on
educational attainment in the Netherlands. Sociology of Education,
59, 237–246.
DiMaggio, P. (1982). Cultural capital and school success: The impact
of status culture participation on the grades of U.S. High School
Students. American Sociological Review,47, 189–201.
Downey, D. B., von Hippel, P. T., & Broh, B. A. (2004). Are
schools the great equalizer? Cognitive inequality during the sum-
mer months and the school year.American Sociological Review,69,
613–643.
Dronkers, J. (1992). Parents, love, and money: The relations between
parental class, cognitive skill, educational attainment, occupation,
marriage, and family income. International Perspectives on Edu-
cation and Society,2, 277–293.
Evans, M. D. R., & Kelley, J. (2002). Cultural resources and educational
success: The Beaux arts vs. scholarly culture. In M. Evans, & J.
Kelley (Eds.), Australian economy and society 2001: Education,
work, and welfare (pp. 54–65). Sydney: Federation Press.
Evans, M. D. R., & Lukic, T. (1998). Immigrant women’s labor force
participation in Australia: The impact of resources and family-level
cultural practices. Gender Issues,16, 52–83.
Farkas, G., & Hibel, J. (2008). Being unready for school: Factors affect-
ing risk and resilience. In A. C. Crouter (Ed.), Disparities in school
readiness how families contributeto transitions to school. Mahwah,
NJ: Lawrence Erlbaum.
Ganzeboom, H. B. G., de Graaf, P. M., & Robert, P. (1990). Cul-
tural reproduction theory on socialist ground. Research in Social
Stratification and Mobility,9, 79–104.
Ganzeboom, H. B. G., & Nieuwbeerta, P. (1999). Access to education
in six eastern European countries between 1940 and 1985. Results
of a cross-national survey. Communist and Post-Communist Stud-
ies,32, 339–357.
Ganzeboom, H. B. G., & Treiman, D. J. (1993). Preliminary results on
educational expansion and education achievement in comparative
perspective. In P. L. J. Hermkens (Ed.), Solidarity of generations:
Demographic, economic, and social change, and its consequences
(pp. 457–506). Amsterdam: Thesis Publishers.
Ganzeboom, H. B. G., & Treiman, D. J. (2003). Three internationally
standardized measures for comparative research on occupational
status. In C. Wolf (Ed.), Advances in cross-national comparison.
A European working book for demographic and socio-economic
variables (pp. 159–193). New York: Kluwer Academic Press.
Ganzeboom, H. B. G., & Treiman, D. J. (1996). Internationally com-
parable measures of occupational status for the 1988 International
Standard Classification of Occupations. Social Science Research,
25, 201–239.
Goblot, E. (1925 [1973]). Cultural education as a middle class enclave.
In T. Burns (Ed.), The sociology of literature and drama (pp.
433–444). Harmondsworth: Penguin.
Heston, A., Summers, R., & Aten, B. (2002). Penn world table ver-
sion 6.1. Center for International Comparisons at the University of
Pennsylvania (CICUP). (electronic data file)
Heyneman, S. P., & Loxley, W. A. (1982). Influences on academic
achievement across high and low income countries: A re-analysis
of IEA data. Sociology of Education,55, 13–21.
Honaker, J., Joseph, A., King, G., Scheve, K., & Singh, N. (2003).
Amelia: A program for missing data. Department of Government,
Harvard University., available at:. http://gking.harvard.edu
Jagodzinski, K., & Uher, R. (2001). International social surveyprogram
1998: Ideology of inequality—Pooled international file, ZA. In
Codebook and MRDF. Koeln: Zentralarchiv.
Jones, F. L., & Kelley, J. (1984). Decomposing differences between
groups: A cautionary note on measuring discrimination. Sociolog-
ical Methods and Research,12, 323–343.
Kelley, J. (1973). Causal chain models for the socioeconomic career.
American Sociological Review,38, 481–493.
Kelley, J., & Evans, M. D. R. (1995). Class and class conflict in six
western nations. American Sociological Review,60, 157–178.
Kelley, J., Evans, M. D. R., & Sikora, J. (2005). Family background,
books, and academic performance in 43 OECD nations. Melbourne
Institute of Applied Economic and Social Research, University of
Melbourne., available at:. http://www.international-survey.org
Kelley, J., Evans, M. D. R., & Sikora, J. (2007). World inequality study:
Codebook and machine readable data file, 1987–2004. Interna-
tional Survey Center. http://www.international-survey.org
Kelley, J., Zagorski, K., Evans, M. D. R., Ervasti, H., Ganzeboom, H.
B. G., Gijsberts, M., et al. (1998). ISEA, rounds 1 and 2 (Inter-
national Survey of Economic Attitudes). Codebook and MRDF.
International Survey Center. http://www.international-survey.org
King, G., Honaker, J., Joseph, A., & Scheve, K. (2001). Analyz-
ing incomplete political science data: An alternative algorithm
for multiple imputation. American Political Science Review,85,
49–69.

Please cite this article in press as: Evans, M. D. R., et al. Family scholarly culture and educational success: Books and schooling
in 27 nations. Research in Social Stratification and Mobility (2010), doi:10.1016/j.rssm.2010.01.002
ARTICLE IN PRESS
+Model
RSSM-98; No. of Pages 27
M.D.R. Evans et al. / Research in Social Stratification and Mobility xxx (2010) xxx–xxx 27
King, G., Murray, C. J. L., Salomon, J. A., & Tandon, A. (2004).
Enhancing the validity and cross-cultural comparability of mea-
surement in survey research. American Political Science Review,
98, 191–207.
Kingston, P. W. (2001). The unfulfilled promise of cultural capital
theory. Sociology of Education, 88–99.
Kohn, M. L., Naoi, A., Schoenbach, C., Schooler, C., & Slomczynski,
K. M. (1990). Position in the class-structure and psychological
functioning in the United-States, Japan, and Poland. American
Journal of Sociology,95, 964–1008.
Kohn, M. L., Slomczynski, K. M., Janicka, K., Khmelko, V., Mach,
B. W., Paniotto, V., et al. (1997). Social structure and personality
under conditions of radical social change: A comparative anal-
ysis of Poland and Ukraine. American Sociological Review,62,
614–638.
Lancieri, Elio. (1993). Dollar GNP estimates for Central and East-
ern Europe 1970–90: A survey and a comparison with western
countries. World Development,21, 161–175.
Lucas, S. R. (2001). Effectively maintained inequality: Education tran-
sitions, track mobility, and social background effects. American
Journal of Sociology,106, 1642–1690.
OECD. (2002). Pisa 2000 Technical Report. Paris: Organization for
Economic Co-operation and Development.
Park, H. (2008). Home literacy environments and children’s reading
performance: A comparative study of 25 countries. Educational
Research and Evaluation,14, 489–505.
Pong, S.-l., & Hao, L. (2007). Neighborhood and school factors in the
school performance of immigrants’ children. International Migra-
tion Review,41, 206–241.
Raghunathan, T. E., Solenberger, P. W., & Van Hoewyk, J. (2004).
IVEware: Imputation and variance estimation software. Univer-
sity of Michigan: Survey Methodology Program, Survey Research
Center, Institute for Social Research.
Robinson, R. V., & Kelley, J. (1979). Class as conceived by Marx
and Dahrendorf: Effects on income inequality and politics in the
United States and Great Britain. American Sociological Review,
44, 38–58.
Schafer, J. L. (1997). Analysis of incomplete multivariate data. Lon-
don: Chapman and Hall.
Schneider, B. L., & Keesler S V.A. (2007). School reform 2007: Trans-
forming education into a scientific enterprise. Annual Review of
Sociology,33, 197–217.
Spaeth, J. L. (1976). Cognitive complexity: A dimension underlying
the socioeconomic achievement process. In D. Featherman (Ed.),
Schooling and achievement in American Society (pp. 103–131).
New York: Academic Press.
Stinchcombe, A. L. (1968). The Logic of Scientific Inference. In Con-
structing social theories. Chicago: University of Chicago Press.,
pp. 15–29.
Sullivan, A. (2001). Cultural capital and educational attainment. Soci-
ology,35, 893–912.
Swidler, A. (1986). Culture in action—Symbols and strategies. Amer-
ican Sociological Review,51, 273–286.
Szelenyi, I., & Treiman, D. J. (1994). Social stratification in East-
ern Europe after 1989: Codebook and MRDF. Los Angeles:
UCLA Institute for Social Science Research, Social Science Data
Archive.
Teachman, Jay D. (1996). Intellectual skill and academic performance:
Do families bias the relationship? Sociology of Education,69,
35–48.
Teachman, Jay D. (1987). Family background. Educational resources
and educational attainment. American Sociological Review,52,
548–557.
Treiman, D. J. (1994). Social stratification in Eastern Europe after
1989: Provisional codebook general population survey [MRDF].
Los Angeles: UCLA Institute for Social Science Research, Social
Science Data Archive [distributor].
Treiman, D. J., et al. (1993). Social stratification in South Africa
[MRDF]. Los Angeles: UCLA Institute for Social Science
Research, Social Science Data Archive [distributor].
Treiman, D. J., & Lewin, A. (1993). Survey of socioeconomic oppor-
tunity and achievement in South Africa. Codebook and MRDF.
Los Angeles: UCLA Institute for Social Science Research. Unpub-
lished.
Treiman, D. J., Walder, A., Zheng, H., Li, Q., Pan, S., Shi, X.,
et al. (1996). Life histories and social change in contemporary
China (LHSCCC) survey. Codebook and MRDF. Los Angeles:
UCLA, Institute for Social Science Research, Social Science Data
Archive.
Treiman, D. J., & Kam-Bor, Y. (1989). Educational and occupational
attainment in 21 countries. In M. L. Kohn (Ed.), Cross-
National research in sociology (pp. 373–394). Newbury Park:
Sage.
Unger, J. (1982). Education under Mao: Class and competition in
Canton schools, 1960–1980. New York: Columbia University
Press.
van Eijck, K., & de Graaf, P. M. (1995). The effects of family structure
on the educational attainment of siblings in Hungary. European
Sociological Review,11, 273–292.
Zentralarchiv fuer Empirische Sozialforschung. (2002). Interna-
tional Social Survey Programme 1999/2000: Social Inequality-3,
international pooled file. Codebook and machine-readable data
file. Cologne, Germany: Zentralarchiv fuer Empirische Sozial-
forschung.
Zimdars, A., Sullivan, A., & Heath, A. (2009). Elite higher education
admissions in the arts and sciences: Is cultural capital the key?
Sociology,43, 648–666.