SERIES

IZA DP No. 6099

PAPER

The Intergenerational Transmission of Human Capital:

Exploring the Role of Skills and Health
Using Data on Adoptees and Twins
DISCUSSION

Petter Lundborg
Martin Nordin
Dan Olof Rooth

October 2011

Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
 The Intergenerational Transmission of
Human Capital:
Exploring the Role of Skills and Health
Using Data on Adoptees and Twins

Petter Lundborg
Lund University, VU University Amsterdam,
Tinbergen Institute, HEP, CED and IZA

Martin Nordin
Lund University and HEP

Dan Olof Rooth

Linnaeus University, Lund University,
CED, CReAM and IZA

Discussion Paper No. 6099

October 2011

IZA

P.O. Box 7240

53072 Bonn
Germany

Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: [email protected]

Any opinions expressed here are those of the author(s) and not those of IZA. Research published in
this series may include views on policy, but the institute itself takes no institutional policy positions.

The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
organization supported by Deutsche Post Foundation. The center is associated with the University of
Bonn and offers a stimulating research environment through its international network, workshops and
conferences, data service, project support, research visits and doctoral program. IZA engages in (i)
original and internationally competitive research in all fields of labor economics, (ii) development of
policy concepts, and (iii) dissemination of research results and concepts to the interested public.

IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 6099
October 2011

ABSTRACT

The Intergenerational Transmission of Human Capital:

Exploring the Role of Skills and Health
Using Data on Adoptees and Twins*

In this paper, we focus on possible causal mechanisms behind the intergenerational

transmission of human capital. For this purpose, we use both an adoption and a twin design
and study the effect of parents’ education on their children’s cognitive skills, non-cognitive
skills, and health. Our results show that greater parental education increases children’s
cognitive and non-cognitive skills, as well as their health. These results suggest that the
effect of parents’ education on children’s education may work partly through the positive
effect that parental education has on children’s skills and health.

JEL Classification: I12, I11, J14, J12, C41

Keywords: intergenerational transmission, human capital, education, health,

cognitive skills, non-cognitive skills, adoptees, twins

Corresponding author:

Petter Lundborg
Lund University
Department of Economics
P.O. Box 7082
220 07 Lund
Sweden
E-mail: [email protected]

*
The authors thank Anders Björklund, Paul Devereux, Peter Fredriksson, Lars Kirkebøen, Mårten
Palme, Erik Plug, Marte Rønning, Jonas Vlachos and seminar participants at UCD Dublin, Lund
University, Stockholm University, Statistics Norway, and the EALE 2011 conference for useful
comments and suggestions. A research grant from the Centre for Economic Demography (CED) at
Lund University is gratefully acknowledged.
I Introduction
There is a strong and well known correlation between parents’human capital
and that of their children. A recent literature has established that at least
part of this transmission of human capital seems to re‡ect a ’nurturing’e¤ect
of parents’education on their children’s education (for recent overviews of the
literature, see Black and Deveraux 2010, Björklund and Salvanes 2010, and
Holmlund et al. 2010). This literature has mostly been based on studies of
adopted children and children of twin parents (e.g. Behrman and Rosenzweig
2002; Sacerdote, 2002; Plug, 2004; Björklund et al. 2006; Holmlund et al.,
2010).1
Much less is known about the causal mechanisms involved in the in-
tergenerational transmission of human capital. Do the estimates re‡ect, for
instance, that more educated parents get healthier children, who, in turn, go
on to obtain more schooling? A growing literature points to the importance
of early "health capital" for education and earnings (see Currie 2009 and Al-
mond and Currie 2010 for recent overviews). Or do the estimates re‡ect that
more educated parents get smarter children or children with greater social
skills? Studies by James Heckman and others emphasise the importance of
skill acquisition early in life for later life outcomes (Heckman et al. 2006).
In this paper, we focus on possible mechanisms behind the intergen-
erational transmission of human capital. For this purpose, we use both an
adoption and a twin design and study the e¤ect of parents’ education on
their children’s cognitive skills, non-cognitive skills, and health. This allow us
to shed some light on what is re‡ected in previous twin and adoption-based
estimates on the transmission of education across generations. If parental
schooling a¤ects child outcomes such as skills and health, this also provides
a possible explanation to how the intergenerational transmission of education
arises.
We believe that improving the understanding of the underlying mech-
anisms that explains the transmission of education across generations is a nat-
ural next step in the literature.2 First of all, opening the black box of possible
causal mechanisms would increase our understanding of what it is that well-
educated parents transmit to their children. This could be useful information
for policy-makers. For instance, policies that increase the level of education in
society could be considered as more valuable if they also improve the education
of people’s o¤spring through increasing the o¤spring’s skills and the health,
rather than only through improving their study results in a narrower sense.
Second, policies that encourage people to invest in education may be more
desirable to society if education promotes intergenerational mobility through
1
There is also a small number of IV-studies, where changes in compulsory schooling laws
have been used for identi…cation (see for instance Black et al. 2005; Holmlund et al. 2010).
2
The importance of understanding the mechanisms is also argued in a recent overview
by Black and Deveraux (2010), where the authors write: "The focus of current research
is on establishing a link between parent and child education; understanding the underlying
mechanisms is a clear direction for future research." A similar view is expressed by Holmlund
et al. (2010), who conclude that: "From our perspective, the roadmap for future research
lies in a better understanding of the mechanisms that explain how parental schooling is
passed on to the next generation."

2
improving the skills and health of children, rather than through for instance
only promoting access to higher education or through nepotism.3
Third, knowledge about the underlying mechanisms would also allow
policy-makers to directly target those mechanisms in order to reach policy goals
such as increased intergenerational mobility.4 Finally, increased knowledge
about the mechanisms is of interest from a child development perspective.
Improving children’s skills and health is often regarded as an important goal
in itself for society and understanding the role that parental education plays in
the "child quality production function" should therefore be useful for policy-
makers.
Our results show that greater parental education increases children’s
cognitive and non-cognitive skills, as well as their health. Moreover, the es-
timates are rather similar using both the adoption and twin design. In addi-
tion, we obtain some pronounced gender di¤erences, where fathers’education
matters more for the formation of cognitive and non-cognitive skills, whereas
mothers’education matter more for the child’s health.
We also show that our estimates appear very reasonable, given the results
obtained in previous studies on the intergenerational transmission of human
capital and on the returns to early life characteristics. Focusing only on cog-
nitive and non-cognitive skills, we show that the e¤ect of parental education
on such child skills could explain at least half of the causal e¤ect of parental
schooling on children’s schooling obtained in previous studies. Our results
thus suggest that skills and health early in life may be important mechanisms
through which the intergenerational transmission of education arises.
The paper proceeds as follows. The next section reviews some of the
relevant literature on the intergenerational transmission of human capital. In
section 3, we describe the data used in the study. In Section 4, we then describe
and discuss the two empirical strategies used in this paper; the adoption design
and the twin design. Section 5 presents the results and contrasts the results
obtained from the two alternative identi…cation strategies. In Section 6, we
summarize our …ndings and draw some conclusions.

II Previous literature
We start this section by …rst reviewing some of the recent studies that use an
adoption or twin design to study the intergenerational transmission of educa-
tion. We then comment on the few studies that have explored possible causal
mechanisms, using an adoption or twin design. Finally, we provide a brief
3
This also relates to the discussion on the socially optimal level of intergenerational
mobility (Black and Devereaux 2010). In order to determine a socially optimal level of
mobility, it is crucial to understand the underlying causes of the intergenerational transmis-
sion of education. A society may be less willing to accept low mobility if it mainly results
from nepotism rather than from skills that children of well-educated parents get from their
parents.
4
For instance, if parents’education primarily a¤ects children’s education by improving
child health outcomes, health interventions targeted toward low-educated families could
weaken the link between parents’ and their children’s education. The same could be said
for programs aimed at improving skills early in life, such as the Perry pre-school program
(see Heckman et al. 2006).

3
review of some of the IV-literature that may yield some insights into possible
mechanisms. Note that when we report estimates for the intergenerational
transmission of human capital it is, unless otherwise stated, from speci…ca-
tions including fathers’and mothers’education separately, i.e. not accounting
for assortative mating.
In the earlier adoption studies by Bruce Sacerdote (2000, 2002), data
on Korean adoptees in the U.S was used to estimate the relationship between
parents’years of schooling and children’s schooling. He found that one addi-
tional year of schooling of the mother and of the father increased the children’s
years of schooling by .22 and .16, respectively.5 Larger estimates were obtained
by Plug (2004), using a sample of U.S adoptees. When controlling for adop-
tive fathers’education, however, the e¤ect of mothers’education was no longer
signi…cant, whereas the e¤ect of fathers’education remained unaltered when
controlling for the spouse’s education.
Later adoption studies usually obtain smaller ’nurturing’e¤ects. Sac-
erdote (2007) obtained an estimate for mothers’education of .09, using data
on Korean adoptees who were randomly assigned to families.6 Smaller e¤ects,
0.07 for mothers and 0.11 for fathers, were also reported in the novel study by
Björklund et al. (2006), where data on both the biological and adoptive par-
ents of adopted children were used. For adoptive mothers, the e¤ect decreased
dramatically when controlling for spouse’s education, thus con…rming earlier
results by Plug (2004).
A number of recent adoption studies, using data from Scandinavian coun-
tries, in general agree with the estimates obtained in Sacerdote (2007) and
Björklund et al. (2006). In Holmlund et al. (2010), using data on both in-
ternationally adopted children and children adopted within Sweden, similar
e¤ects were obtained for both mothers’ and fathers’ education with no es-
timate being larger than 0.11 for children adopted in Sweden and 0.04 for
international adoptees.7
In addition to the adoption studies, there has been a small number of
twin studies on the intergenerational transmission of education. Using data
on monozygotic twin parents, Behrman and Rosenzweig (2002) showed that
no signi…cant association between mothers’schooling and children’s schooling
remained when genes and common environment were accounted for. The e¤ect
of fathers’education was still signi…cant, however, with an estimate of 0.36.
Holmlund et al. (2010) also found that fathers’education mattered more,
using a sample of pooled MZ and DZ twins, but with a smaller point esti-
mate of 0.12. Using Norwegian data on pooled MZ and DZ twins, Pronzato
(2010) found larger e¤ects of mothers’education, being around 0.10, compared
to previous twin studies but still obtained the result that fathers’ education
mattered more (0.16).8
5
There is also an even earlier study by Dearden et al. (1997) that uses data on 41
adopted children from the British NCDS data. The results suggest even larger estimates of
the intergenerational transmission of education compared to Sacerdote (2000, 2002).
6
In a footnote, Sacerdote (2007) reports that similar results were obtained for fathers’
education.
7
Using school performance as an outcome, similar estimates were also obtained in a
Norwegian context, using data on Korean adoptees (Haegeland et al. 2010).
8
Haegeland et al (2010), studying school performance, found that fathers’ education

4
 Contrasting evidence was obtained in Amin et al. (2011a), where the
results showed that only mothers’education mattered, when using a sample
consisting of MZ twins only. Moreover, this e¤ect was driven completely by the
e¤ect of mothers’schooling on their daughter’s schooling, where one additional
year of mother’s schooling increased daughter’s schooling by 0.10. Moreover,
in Bingley et al. (2009), where data on Danish MZ twins were used, the
importance of mothers’education became larger in more recent cohorts.
There are also a few IV-studies on the topic, such as Black et al.
(2005). Using a reform of the Norwegian schooling system as an instrument,
they …nd no causal relationship between the father’s education and the child’s
education level. For mother’s they report a small e¤ect on the son’s education
(but not on the daughter’s). Similar IV-results were obtained by Holmlund et
al. (2010), where mothers’education were found to be slightly more important.
For an overview of IV-studies using other educational outcomes of the children,
such as grade repetition etc, see Holmlund et al. (2010).
Our overall conclusion from the studies cited above is that there seems
to exist a nurturing e¤ect of parents’education, although more recent adop-
tion and twins studies have estimated this transmission to be rather small,
although not negligible. We next review the small literature using a twin or
adoption design that have provided estimates on the causal e¤ect of parents’
education on children’s outcomes, such as health and skills. These studies
provide results that may be informative on the causal mechanisms involved in
the intergenerational transmission of education.
In Sacerdote (2000), the relation between parents’years of schooling
and children’s cognitive test scores was estimated for a small sample of Korean
adoptees in the U.S.9 The results suggest a signi…cant nurturing e¤ect of both
mothers and fathers’ education. In both cases, the estimates are about half
the size compared to the estimates obtained in the control group of own-birth
children. In Sacerdote (2007), the relation between mothers’ education and
children’s BMI and drinking behaviour was estimated. For both outcomes, the
e¤ect of adoptive mothers’education is small and insigni…cant.
Using Chinese data, Chen and Li (2009) estimate the e¤ect of mothers’
education on children’s height-for-age and …nd the e¤ect to be of similar size for
both adoptive mothers and own-birth mothers. The similarity of the results
for adopted children and own-birth children suggests that the relationship
between parental education and child height is primarily caused by parental
nurturing and not by genetics (the estimates suggest that about 80% of the
relationship could be attributed to nurture).10
mattered more for the child’s exam scores.
9
Sacerdote (2000) also uses a sample of adoptees from the British NCDS. For the analysis
of the e¤ect of fathers’ years of schooling, the sample is very small, 81 children, and the
e¤ect on various child test scores at age 7 is imprecisely estimated and always insigni…cant.
In addition, data on 117 adoptees from the Colorado adoption project is used to estimate
the e¤ect of adoptive mothers’education on child test scores at ages 6 and 7. The estimates
are insigni…cant and a lot smaller than the corresponding estimates for the e¤ect of the
biological mother’s education.
10
The large nurturing component obtained by Chen & Li (2009) contrasts greatly with
the results discussed above using adopted children to study the intergenerational correlation
in schooling. The relatively greater role of nurture versus nature is also somewhat surpris-

5
 The only twin study on the relationship between parental education
and child health outcomes that we are aware of is a working paper by Bingley et
al. (2009). They use Danish data and estimate the e¤ect of parental schooling
on the birth weight of the o¤spring. The results show that mothers’schooling
increase the birth weight of their o¤spring but that fathers’schooling have no
e¤ect.
Besides the adoption and twin studies on the relation between parents’
education and child outcomes other than education, there is also a small IV-
literature on the topic.11 The …rst IV study on the topic appears to be Currie
and Moretti (2003), who used the expansion in the number of colleges in the
US between 1940 and 1996 as an instrument for mothers’ schooling. They
found that the increase in college attendance of women improved children’s
health, in terms of birth weight and gestational age and led to less smoking
among mothers.
In a study using UK data, Lindeboom et al. (2009) used the com-
pulsory schooling reform in 1947 in Britain to estimate the causal e¤ect of
schooling on child health. They found no evidence of a causal e¤ect of parental
schooling on child health outcomes at birth and at ages 7, 11, and 16. Chevalier
and Sullivan (2007), however, found that the reform had heterogeneous e¤ects
and that the most impacted groups experienced larger changes in infant birth
weight.
Carneiro et al. (2007) used U.S data and instrumented mothers’edu-
cation with the presence of colleges at age 17 in the state of residence and with
local labor market conditions. They found a strong e¤ect of maternal educa-
tion on child cognitive outcomes, but also on measures of behavior problems.
Finally, McCrary and Royer (2011) combined data on exact birth dates
and age-at-school entry policies in the spirit of Angrist and Kreuger’s quarter
of birth instrument. Using data on all births in California during the years
1989-2002, and using birth date as an instrument for schooling, they …nd no
evidence of a causal e¤ect of schooling on infant health outcomes.
The literature review above suggests that there is rather mixed evi-
dence regarding the causal e¤ect of parental schooling on child outcomes such
as health and cognitive skills in developed countries.12 In a comprehensive sur-
ing, given that the average age of the adoptive children at follow-up is only 27.5 month,
which means that the “nurturing period” is relatively short. The results may thus suggest
that nurturing factors are more important for children’s health development in developing
countries, but less important for educational attainment in developed countries.
11
Note, however, that the results from this literature may be hard to generalize to the
twin and adoption literature, since the IV-estimates re‡ect Local Average Treatment E¤ects,
where the instruments used normally a¤ect people at the lower end of the educational
distribution, while results using adopted children e¤ectively exploits variation in education
in a group of well-educated parents. The twin design exploit di¤erences in education between
twin parents, where the di¤erences are often found to be quite evenly distributed across the
education distribution.
12
There also exist some studies in a developing country context, where the e¤ect of
parental education is ususally found to be signi…cant. Breierova and Du‡o (2004), for
instance, exploited the primary school construction program in Indonesia during the 1970s.
During this period, a large number of new schools were constructed but the intensity of the
construction program varied across the districts in Indonesia. Using program intensity as an
instrument for schooling, they …nd that parents’schooling have a negative e¤ect on infant

6
vey on the literature on children’s health and parental socioeconomic status,
Currie (2009) concludes that it is still di¢ cult to prove that the relationship
between parental background and children’s health is a causal relationship
and that more research is needed. It could be noted, though, that there is
some agreement between the IV-studies and the studies using an adoption
and twin design. Both Bingley et al. (2009), Currie and Moretti (2003), and
Chevalier and Sullivan (2007) …nd evidence that maternal education a¤ects
birth weight. Moreover, both Carneiro et al. (2007) and Sacerdote (2000) …nd
e¤ects on cognitive outcomes.13

III Data
For the purpose of this study, we merged a number of di¤erent registers. This
includes the Register of the Total Population (RTB, with information on birth
date, date of immigration, sex, country of birth, and parents’country of birth),
the Multi Generation Register (Flergenerationsregistret, with identi…cation of
siblings on the mothers side through a family ID), the Swedish National Service
Administration, the Swedish Register of Education (UREG), the National Tax
Records, and the Swedish Twin Registry (with information on twin zygosity).
We identify adopted children through an adoption indicator that is
available in the RTB-register from Statistics Sweden. From this register, we
select all male foreign-born adoptees born between 1965 and 1978. In our sam-
ple, most adopted children came from South Korea (24%), India (14%), Chile
(10%), Thailand (9%), Colombia (7%) and Sri Lanka (6%). These countries
constitute 70% of the countries in our sample.
In order to avoid the risk that some foreign-born adoptees are adopted
by relatives, which would create a genetic link, we only include adoptees
adopted by Swedish-born parents. By further restricting the sample to those
who have enlisted to the military, and who have non-missing data on parental
education and on our outcome measures, our sample of adoptees varies between
3,375 and 3,741, depending on the outcome studied.
Our twin sample consists of all children to twin parents that were born
between 1950-1978 and who have enlisted.14 In order to determine zygosity
of the twins, we have merged our data with zygosity information from the
mortality.
13
There are also a number of studies that estimates the relationship between various out-
comes across generations, where the goal is not to establish causality, but where the results
may still point to potentially important mechanisms for the intergenerational transmission
of education. These includes, for instance, Black et al. (2009), Anger and Heineck (2009),
Grönqvist et al. (2010), and Björklund et al. (2010), who all study the transmission of
IQ across generations, and Lindqvist & Hjalmarsson (2010) and Lindqvist & Hjalmarsson
(2011a), who study the intergenerational transmission of crime and drunk driving. In a
recent paper, however, Lindqvist & Hjalmarsson (2011b) use an adoption design, including
both information on the adopted and biological parents, to study the intergenerational trans-
mission of crime. Moreover, Amin et al. (2011b), studies the intergenerational transmission
of income, using a twin design.
14
Note that for adopted children, the sample only includes those born 1965 and onwards.
The reason is that there were almost no international adoptions taking place in Sweden
before 1965.

7
Swedish twin registry. In this manner, we are able to determine zygosity for
82 % of the parents in our dataset.15 Our dataset includes between 2,983
and 3,216 children of monozygotic twin parents, depending on the outcome
studied.
Our third sample consists of the population of "own-birth" sons, born
during the periods corresponding to the periods used in the adoptee and twin
samples. With this sample, we are unable to provide "baseline" estimates
for a sample of sons with biological parents. We will then contrast these
results with the ones obtained when breaking the genetic link through the
adoption and twin design. Restricting the sample of own-birth sons to those
with non-missing observations on parental education gives us between 851,128
and 903,891 observations, again, depending on the outcome studied.
In our empirical analyses, we focus on three set of outcomes; cognitive
skills, non-cognitive skills, and health. These are measured at age 18, during
the military enlistment tests. It should be noted that during our study period,
in principle every male Swedish citizen enlisted for the military when turning
18. Enlistment was mandatory and there is very little attrition in the data.
Refusal to enlist leads to a …ne, and eventually to imprisonment. Individuals
are only exempted from enlistment if they are imprisoned, if they have ever
been convicted for heavy crimes (which mostly concerns violence-related and
abuse-related crimes), or if they are in care institutions and are deemed to be
unable to function in a war situation. During our study period, the annual
cohort size of men turning 18 was about 50,000. Per cohort, around 1,250 (i.e.,
0.25%) were exempted from enlistment.16
Cognitive skills are measured using a test similar in style to the AFQT in
the US. The test is called the Enlistment Battery 80 and includes four separate
tests; Instructions, Synonyms, Metal Folding and Technical Comprehension.
The separate scores of these tests are aggregated into a standard composite
measure calculated by the military enlistment service. In our analyses, we use
a normalized version of this score.
Non-cognitive skills are measured through interviews carried out by cer-
ti…ed psychologists employed by the Swedish army. The ultimate purpose of
the interview is to evaluate the conscript’s ability to perform military service
and to function in a war situation. This is achieved through an assessment of
the enlistee’s psychological stability and endurance, capability of taking ini-
tiatives, responsibility, and social competence. The assessment results in a
composite enlistment score of non-cognitive skills, ranging from 1 to 9, which
we then normalize in our analyses.
Though the original purpose of the non-cognitive skill measure used here is
to evaluate peoples’suitability to serve in a war situation, the character traits
15
The twin registry determines zygosity based on survey questions regarding co-twin sim-
ilarity. The method used has been found to classify twins with an accuracy of 95 % or more
(see Lichtenstein, 2002).
16
Note also that the incentives to deliberately underperform at the enlistment tests are
rather limited. The reason is that, during our study period, the results of the tests had
no impact on the probability of doing military service or not, since almost all people that
enlisted during our study period also completed military service. Instead, the test results
merely in‡uenced the individual’s placement within the army, meaning that poorer results
typically led to a less quali…ed and meriting placement.

8
valued by the military psychologists are similar to the traits usually included
in measures of non-cognitive skills. The measure has also been found to be a
strong predictor of adult earnings, independently of cognitive skills (Lindqvist
and Vestman 2010).
We consider two health outcomes in our empirical analyses; a measure
of overall health and height.17 The former is an unidimensional global health
measure, which is based on the severity of the individual’s health conditions
(both physical and mental) and which is used by the military to determine the
enlistees suitability with respect to type of military service. In our analyses,
we transform their ordered categories into a 13-step scale, ranging from worst
possible health to perfect health. The measure of height is taken by the sta¤
at the enlistment o¢ ces and we thus have none of the problems encountered
when using self-reports on height.
Table 1 reports descriptive statistics for our three samples; own-birth
sons, adopted sons and own-birth sons with a monozygotic twin-parent. Whereas
mean health is similar in the group of own-birth children and the children of
twin parents, adoptees have substantially worse health. Moreover, the height
of adopted sons is, on average, considerably shorter than the height of own-
birth sons (with or without a twin parent). Finally, adoptive parents are
signi…cantly more educated than the parents of own-birth sons and adopted
sons have lower cognitive and non-cognitive test scores compared to own-birth
sons.

IV Empirical model
The typical reduced form equation used to estimate the intergenerational
transmission of human capital from a parent p to a child c can be written:

Yc = + S p + gp + f p + c
: (1)
In this paper, Y c represents child outcomes, such as cognitive and non-
cognitive skills and health, whereas S p represents the schooling of the parent.
Moreover, g p represents any genetic endowments that are passed on from the
biological parent to the child, whereas f p represents any non-genetically trans-
mitted characteristic of the parent other than schooling, such as child-rearing
skills. Unobserved child-speci…c factors are denoted c . The coe¢ cient mea-
sures the causal e¤ect of parental education, S p ; on child outcomes. The
and coe¢ cients measure the extent to which genetic endowments and non-
genetic endowments are transmitted to the child. As shown by Holmlund et
al. (2010), the least-square estimator of can be written as:

b cov(S p ; g p ) cov(S p ; f p )
p lim = + + (2)
var(S p ) var(S p )
p p p p
Thus, in order to identify , we need cov(S ;g )
var(S p )
= 0 and cov(S ;f )
var(S p )
= 0 or
= 0 and = 0. This is unlikely to hold in a sample of own-birth children
17
An adult individual’s height has been denoted “probably the best single indicator of his
or her dietary and infectious disease history during childhood” (Elo and Preston, 1992).

9
and parents. For instance, it seems likely that the education of a parent is
related to certain genes of the parent. At the same time, these genes may be
important predictors of the child’s skills and health. Similarly, non-genetic
factors, such as child-rearing talents of the parent may be related to both the
education of the parent and to child outcomes. The adoption and the twin
design o¤er two alternative ways of adressing these concerns.

The adoption approach

For adoptees, we can write the equation relating parents’education to child
outcomes as:

Y ac = AD + AD S
ap
+ AD g
bp
+ AD f
ap
+ ac
(3)
where ac refers to the adopted child, ap to the adoptive parent, and bp to
the biological parent. In this speci…cation, only the education of one of the
adoptive parents is included but in our empirical analysis, we will also employ
speci…cations were we include the spouse’s education.
In order to identify AD , the adoption literature employs two standard as-
sumptions. First, adopted children are assumed to be randomly allocated to
their adoptive parents. If the adoption agency instead would match the adop-
tive child with the adoptive parents, or if the education of the adoptive parents
is related to characteristics of the biological parents, an arti…cial genetic link
may be constructed and cov(S ap ; g bp ) 6= 0. This would also be the case if adop-
tive parents were able to choose who to adopt and more well-educated parents
would choose more healthy children or children with greater innate skills.
Holmlund et al. (2010) used data on international adoptees in Sweden and
found some weak evidence of selection of adoptees. For instance, more highly
educated mothers seemed to prefer older children and boys. The estimated
e¤ects were very small, however, and when controlling for these observable
factors, the estimated intergenerational associations hardly changed at all.
Besides controlling for observable characteristics of the adoptees, we will con-
duct a sensitivity analysis in order to examine whether child-parent matching
is a problem in our study.
It should be noted that adoptive parents in Sweden are not able to choose
a particular child for adoption. Instead, a strict procedure is followed, where
the parents seeking to adopt …rst have to seek permission from the social
authorities, who will undertake a thorough investigation about the person’s
or couple’s ability to take care of an adopted child (SOU 2003). If granted
permission, the couple can contact a government-accredited adoption organi-
zation and …le an application to adopt. By choosing a particular organization,
who often specialises in children from particular countries, the parents are able
to choose what country to adopt from. They are also able to state a prefer-
ence regarding the gender and age of the child, although there is no guarantee
that this preference will be ful…lled. Thus, the parents may in‡uence certain
characteristics of the adoptive child, but these characteristics are observable
to us.18
18
Another worry is that the parents or the adoption centers in the home countries of the
adoptees will choose adoptive parents based on certain characteristics. It is very common,

10
 The second assumption required in the adoption design is that any non-
genetic characteristics of the parents, that may also directly a¤ect the out-
comes of the adopted child, are unrelated to parental schooling, i.e. that
cov(S ap ; f ap ) 6= 0. There is nothing in the adoption design that guarantees
this and one can only speculate on the likely direction of the bias that arises
from violating this assumption. For child-rearing skills, for instance, it is a
priori not clear if such skills are negatively or positively related to parental
education (Björklund et al. 2006). If there is a trade-o¤ between investing
in child-rearing skills and schooling they may be negatively related. On the
other hand, education is often assumed to be related to an increased e¢ ciency
of household production, which may include child-rearing. In any case, if the
second assumption needed in the adoption design is violated, the estimated
"nurturing" e¤ect would include both the e¤ect of parental education and any
other non-genetic factors that are related to both parental schooling and child
outcomes.
Even if internal validity would be ful…lled in the adoption design, there are
several threats to the external validity of the results. As shown in Table 1, there
are large di¤erences in the distribution of characteristics between adopted sons
and own-birth sons and between own-birth parents and adoptive parents. If
these di¤erences also mean that the e¤ect of parental education is di¤erent for
adopted sons compared to own-birth sons, this would make a comparison of
estimates obtained using the di¤erent designs complicated. We will therefore
devote parts of the results section to investigate if there is any evidence of such
di¤erential e¤ects. In particular, we will investigate if there is any evidence
that parents treat own-birth and adopted sons di¤erently, if adoptive families
are more nurturing, and if nurturing is less e¤ective for adopted sons, due to
exposure to adverse early life conditions.

The twin approach

In the twin approach, data on monozygotic twins, who share genes and
early family environment, is used. Here, the in‡uence of such shared factors is
removed by relating di¤erences in schooling between monozygotic twin parents
to di¤erences in their children’s skills and health. We may thus write the
intergenerational transmission equation as:

Yc = TW Sp + TW gp + TW fp + c
(4)
however, that the children given up for adoption are children that have been abandoned
by their mother, in which case the mother is unknown (SOU 2003). Even if the mother is
known, it seems that she has little in‡uence over the selection of adoptive parents, once she
has given a written approval to the organization to adopt away her child. The adoption
organization may, however, require certain conditions to be ful…lled in order to allow a couple
to adopt and these requirements vary somewhat by country. In South Korea, for instance,
the adoption organization only allows married couples below the age of 44 to adopt. It is
also required that the marriage has lasted for at least 3 years and that the spouses have
at least 11 years of schooling. In the other large adoption countries, the most important
criteria seem to concern the spouses age and their marital status. For a thorough description
of the various conditions demanded in the di¤erent countries, see SOU (2003).

11
Since MZ twins are genetically identical, we have that g p = 0.19 In order
for the twin approach to provide an estimate of the causal e¤ect, a number
of additional assumptions needs to be made. In our case, the most important
assumption is that schooling di¤erences between identical twin parents are
unrelated to di¤erences in non-genetic characteristics that may also a¤ect the
outcomes of the child. To achieve this, the twin literature also assumes that
f p = 0 within twin pairs. Thus, by di¤erencing out g p and f p , one may
obtain an estimate of the causal e¤ect T W . Another requirement is of course
that some twin parents di¤er in terms of schooling, so that S p is not equal
to zero for all twin pairs.
The assumption that schooling di¤erences within twin pairs are unre-
lated to di¤erences in unobserved non-genetic characteristics has been ques-
tioned by, for instance, Bound and Solon (1999). It would fail, for instance, if
schooling di¤erences within twin pairs are associated with di¤erences in abil-
ity, early life health, or parenting skills. While there is mixed evidence on the
topic, there are some evidence suggesting that birth weight di¤erences between
twins may be associated with schooling di¤erences. If birth weight di¤erences
would also have an independent e¤ect on child outcomes, twin estimates would
be biased (for an overview, see Holmlund et al 2010). Moreover, in a recent
paper by Lundborg et al. (2011), health di¤erences between MZ twins, using a
comprehensive set of health measures taken at age 18, did not predict school-
ing di¤erences between the twins. Note, however, that even if birth weight
di¤erences would predict schooling di¤erences among our twin parents, this
is only a problem if the same birth weight di¤erences are directly associated
with the children’s outcomes.20
Related to the question of unobserved di¤erences within twin pairs is
the issue of assortative mating. In the equation above, we allow the e¤ect of
parental education to also include e¤ects that run through assortative mating.
This means that we still get an unbiased estimate of the schooling of the twin
parent but that the e¤ect also includes the endowments and schooling of the
spouse (Antonovics and Goldberger 2005; Holmlund et al. 2010). We will
contrast the results obtained using this speci…cation to those obtained when
controlling for the education of the spouse. The latter results would then give
the e¤ect of education that does not run through assortative mating. However,
since it is not possible to di¤erence out the endowments that vary between the
spouses, these results will most likely be upward biased. The reason is that
higher educated parents are also more likely to marry spouses with favourable
endowments, generating a positive correlation between the education of the
twin parent and the error term.
Another assumption commonly made in the twin design is that mea-
surement errors do not play an important role. As shown by Griliches (1979),
19
This also requires that we exclude certain types of gene-environment interactions.
20
There is very little evidence suggesting that parental birth weight has a causal e¤ect on
children’s outcomes. Currie & Moretti (2007) estimate a signi…cant relationship between
mothers’birth weight and the birth weight of their children, however, using data on siblings.
For ability, we are aware of no studies that have established a causal e¤ect of parental IQ
on child IQ. In a recent working paper by Sandewall et al. (2010), however, they show that
di¤erences in the cognitive test score among MZ twins at age 18 predict di¤erences in later
schooling.

12
the downward bias induced by normally distributed measurement errors will
be exaggerated when using sibling or twin-…xed-e¤ects models. Any reduction
in the estimated coe¢ cients when imposing sibling-…xed e¤ects may therefore
be caused by the more severe downward bias resulting from measurement er-
rors, rather than from removal of twin-pair-speci…c unobserved heterogeneity.
Since we rely on register data on education, we believe that the measurement
error issue is less of a problem in this study.21

V Results
Adopted sons
We start our analyses by estimating the relationship between parental ed-
ucation and child outcomes for own-birth sons and adopted sons. Table 2
shows the results, where separate regressions for mother’s and father’s school-
ing are estimated. Table 3 shows the corresponding results when controlling
for spouse’s education. All regressions on adoptees control for the adoptees
age at immigration and country of origin.
Focusing …rst on the results for cognitive and non-cognitive outcomes,
the signi…cant estimates for own-birth sons suggest that one additional year
of fathers’ schooling increases cognitive and non-cognitive skills by .099 and
.057, respectively. The corresponding estimates for mothers’ education are
almost identical, being .096 and .056, respectively. Moving on to the adoptees,
however, the estimates are substantially reduced in magnitude. Moreover,
the relative importance of fathers’versus mothers’years of schooling changes
in a non-trivial way. For cognitive skills the estimate for fathers’ education
now decreases to .027, while the corresponding estimate of mothers education
decreases to .014, both being statistically signi…cant at the 5 percent level.
Similar pattern as for cognitive skills are obtained for the e¤ect of
adoptive fathers’education on non-cognitive skills, where the point estimate
is reduced to .017, while the corresponding estimate for mothers’education is
close to zero and statistically insigni…cant. When we control for the spouse’s
education in the regressions on cognitive and non-cognitive skills, as shown
in Table 3, the e¤ect of adoption mothers’education in both cases skills ap-
proaches zero, whereas the e¤ect of adoption father’s education on cognitive
and non-cognitive skills remains almost unchanged.
Our results for adoptees thus suggest that only the education of the
father has a nurturing impact on the child’s cognitive and non-cognitive test
scores. For mothers, we interpret the association between mothers’education
and children’s skills to mainly re‡ect inherited abilities and assortative mat-
ing. The latter result is interesting, given the common …nding in the adoption
literature that mothers’education matter less than fathers education for chil-
dren’s years of schooling. One interpretation of our …nding is thus that the
21
Holmlund et al. (2008) addressed the measurement error problem in the Swedish data
using the approach suggested by Ashenfelter and Kreuger (1994). Instrumenting the edu-
cation measure with another, independent, measure of the same education did not a¤ect
the results. Moreover, Holmlund et al. (2010) found the reliability ratio in the education
measure to be relatively high.

13
larger positive e¤ect of fathers’education on their children’s education partly
arises because well-educated fathers, but not well-educated mothers, improve
the skills of their sons.
At this point, we can only speculate on the mechanism behind the
…nding that only fathers’education matters. A possible non-causal explana-
tion would be that there exists a negative correlation between child-rearing
skills and education among mothers but not among fathers.22 Another expla-
nation could be that the result re‡ects specialization within the household,
where fathers take a greater responsibility for helping with tasks that improve
children’s cognitive and non-cognitive skills, whereas mothers take a greater
responsibility for other household work. Yet another explanation could be
that there is a stronger correlation between education and income for fathers
than for mothers, so that the e¤ect of fathers’ education also picks up the
e¤ect of income to a greater extent.23 We can test this by including controls
for income in the regressions. This only marginally a¤ected the estimates for
parents’ education and the decline in the size of the coe¢ cients was almost
identical for mothers and fathers.24
The relative importance of fathers’ versus mothers’ education is not
as straightforward when considering our health outcomes. Although fathers’
education seems to matter more than mothers’education for the overall health
of their for own-birth sons, this result does not hold up in the sample of
adoptees. Here, mothers’ education is associated with a greater increase in
health and the e¤ect is signi…cant at the 10 percent level, although it should
be noted that the di¤erence in point estimates between mothers and fathers
is rather small. The gender di¤erence becomes more pronounced, however,
when accounting for spouses education, where the point estimate of fathers’
education is halved in magnitude. These results cast some doubts on the
hypothesis that there is a negative correlation between child-rearing skills and
mothers’ education. Rather, the result suggests that fathers and mothers
specialize in di¤erent aspects of child-rearing and that education is generally
bene…cial for the various child outcomes considered.
For our other measure of early life health, height, the results for own-
22
A downward bias would also arise if there is a negative correlation between education
and certain favorable unobserved characteristics in the sample of adoptive parents (Black
et al. 2010). The reason is that adoptive parents who are allowed to adopt despite poor
observable characteristics are likely to have better on average unobserved characteristics that
compensate for their poor observable characteristics. This would imply that the estimated
e¤ect for the sample of adoptive parents is downward biased. If the negative correlation
between education and favourable unobserved characteristics is stronger for mothers than
fathers, one reason being that the screening procedure prior to adoption may put more
emphasis on mothers’characteristics, this could explain why the estimated e¤ects are smaller
for mothers. There is nothing in the formal guidelines for adoption, however, which supports
such a di¤erential treatment of males and females.
23
Yet another explanation for the result that fathers’ education matters more would be
that measurement errors are more severe for the measure of mothers’ years of schooling.
Since our measures are based on register data, we see no reason why this would be the case
in our study. Moreover, Holmlund et al. (2008) report that the reliability ratio for education
in these register is 0.95 for both mothers and fathers.
24
For cognitive skills, the e¤ects of fathers and mothers education declined to .022 and
.009 when controlling for income. The corresponding estimates for non-cognitive skills were
.013 and .000.

14
birth sons suggest that mothers’education are of somewhat more importance.
In the sample of adopted sons, however, the education of the father matters
more, with an estimate that is signi…cant at the 10 percent level. This result
is strengthened when accounting for assortative mating, where the estimate of
mothers’education approaches zero. In sum, the results for health outcomes
are more mixed when it comes to the relative importance of mothers’ and
fathers’education, although the results suggest small but signi…cant nurturing
e¤ects.
As discussed in the methods section, a crucial assumption in the adop-
tion design is that adopted children are randomly allocated to their adoptive
parents. Before proceeding with the results for twins, we shed some light on
the credibility of this assumption by regressing parental education on various
observable characteristics of the adopted children. The results, reported in
Table 4, suggest that some matching appears to be take place, since 4 out of
12 coe¢ cients from the regressions come out as signi…cant. The magnitudes
of the coe¢ cients are quite small, however.
The results suggest, for instance, that parental education is positively and
signi…cantly related to the probability of adopting a child from Columbia. On
the other hand, there is a negative correlation between parents’education and
the probability of being adopted from South Korea. Since children from South
Korea outperforms even Swedish-born children in some of the enlistment tests,
this rather suggests that, if anything, some negative selection is present.
Further evidence that the magnitude of the matching is limited is
revealed by the fact that controlling for country of birth in the regressions in
Table 2 does not change the results to any important extent. In fact, some of
the resulting estimates for the adopted children becomes somewhat stronger
in magnitude, which is what one would expected if there is negative relation
between parental education and the probability of adopting a child from "high-
quality" countries.
For age at adoption, we do not …nd any signi…cant relationship with
parental education. In sum, we believe that the results presented in Table 4 do
not suggest any problematic matching that we cannot deal with. We also note
that our results on potential non-random allocation of adoptees are similar to
the results found in Holmlund et al. (2010), where there was no evidence of
any important parent-child matching.

Sons with a monozygotic twin parent

We continue to explore the intergenerational transmission of human capital
using the twin design. The results are shown in Table 2, where we also include
the OLS results for the twin sample in order to get some idea about the
representativeness of the twin sample. Ideally, the OLS estimates should not
di¤er by large compared to the result for the main sample of own-birth children,
which is also what most of the results in Table 2 suggest. For cognitive skills,
for instance, the own-birth estimates of fathers’ and mothers’ education are
.099 and .096, which is very similar to the OLS estimates obtained for the twin
sample of .097 and .100, respectively.
Moving on to the twin …xed-e¤ect results, and starting with cogni-

15
tive skills, the results suggest a similar ‘nurturing’ e¤ect of both mothers’
and fathers’ education of .050 and .044, respectively. For fathers, this esti-
mate is rather similar to the corresponding estimate obtained in the sample
of adoptees. For mothers, the di¤erences are somewhat larger, although one
should note that the point estimates are quite small in both cases and one
should thus not exaggerate the di¤erences. The results for cognitive skills do
not change much when controlling for spouse’s education, as shown in Table
5. Note that this result di¤ers from the result obtained for adoptive mothers,
where the e¤ect approached zero when accounting for the spouse’s education.
The results for non-cognitive skills are much in line with the results
obtained for the adoptees. While the OLS estimates on the twin sample sug-
gest an equal e¤ect of mothers’ and fathers’ education on their children’s
non-cognitive skills, the …xed e¤ects estimates again suggest that the father’s
education matters more. An additional year of schooling increases the non-
cognitive test results by .041, whereas the corresponding estimate for mothers
is .019 and insigni…cant. Accounting for spouse’s education does, again, not
change this pattern.
For our overall health measure, the twin-based estimates again suggest
that mothers’ education matters more. While the OLS estimates suggests
an equal e¤ect of twin mothers’and twin fathers’schooling, the …xed e¤ects
estimates are only signi…cant for twin mothers, in line with the results obtained
for adoptees. Moreover, the estimated coe¢ cient di¤ers substantially from
that of fathers, which is small and has the wrong sign. The estimate, 0.168,
suggests that one additional year of mother’s schooling increases the child’s
health with about 5 percent of a standard deviation. For height, the estimates
are roughly halved in magnitude when imposing …xed e¤ects and similar, but
imprecisely measured, for twin fathers and twin mothers. None of these results
change to any dramatic extent when accounting for spouse’s education.

Comparing the adoption and twin estimates

In sum, the results obtained using both the twin sample and adoptee sample
suggest that cognitive skills, non-cognitive skills, and health may be an impor-
tant mechanism through which the intergenerational transmission of schooling
arises. For most outcomes studied, the results are relatively similar across
the research designs, despite the fact that the samples are very di¤erent and
that very di¤erent assumptions are used. A general picture that emerges is
that fathers’education seems to matter somewhat more for the development
of their son’s skills, whereas mothers’education matters more for the devel-
opment of their health. Since it is rather well established that early life skills
and health a¤ect later educational attainment, our results thus provide some
clues to what the underlying mechanisms behind the transmission of human
capital are.
Even though the results came out remarkably similar across research
designs, there is a tendency towards smaller magnitudes in the sample of
adoptees. Moreover, an often expressed concern is that the external validity
of adoption estimates may be low. Adoptive parents may, for various reasons,
be more or less e¤ective in bringing skills to their adopted children compared

16
to biological parents. There could also be something that makes adopted chil-
dren gain less skills from having well-educated children compared to children
of twin parents. We will next address some hypotheses that may a¤ect the
external validity of the adoption results.
First, we examine di¤erences in the distribution of characteristics be-
tween the group of adoptees, sons of twin parents, and own-birth children.
Clearly, both adoptive parents and adoptees di¤er from the other groups in
many ways, as revealed by the descriptive statistics in Table 1. First of all,
adoptive parents are on average more highly educated. Adoptive fathers and
mothers have on average 12.7 and 12.8 years of schooling, compared to 10.6
years for twin parents. At the same time, adoptees on average score substan-
tially worse on the skills tests during the enlistment compared to sons of twin
parents.
Given the rather substantial di¤erences in the distribution of charac-
teristics between the adoption and twin sample, it becomes even more surpris-
ing to …nd the e¤ects of education on skills and health to be rather similar
across the designs. This suggests that the e¤ects of parental schooling are
not radically di¤erent across the di¤erent groups considered. We will further
check, however, the external validity of the adoption result, which will also
shed some light on possible reasons for the di¤erence in the magnitude of the
estimates obtained in the adoption and the twin design
In order to investigate if adoptive parents are more or less nurturing
compared to other parents, we can compare the e¤ect of education for own-
birth sons with an adopted sibling with the e¤ect for own-birth sons without
an adopted sibling (see e.g. Plug 2004; Holmlund et al. 2010). If adoptive
parents are more nurturing, we would expect a greater e¤ect for own-birth
children with an adopted sibling compared to the e¤ect on own-birth children
without an adopted sibling.
For fathers, the e¤ect of parents’education on the biological child’s cogni-
tive skills, is more or less the same, whether or not the child has an adopted
sibling (Table A1). For mothers, however, the estimate is about double in size
if the child has an adopted sibling. A similar increase in the magnitude of the
e¤ect is obtained for health. These results are consistent with the idea that
that adoptive mothers are more nurturing than biological mothers.25 On the
other hand, the e¤ect on height is now halved in magnitude and we therefore
refrain from any de…nitive statements about the whether adoptive mothers
and fathers are more or less nurturing.
An alternative interpretation of the results above is that mothers who
adopt a child invest less in their biological child and more in the adopted child.
We further test for any evidence that parents treat adopted children di¤erently
from own-birth children by comparing the e¤ect of parental education for
adopted sons with an own-birth sibling to the corresponding e¤ect for those
without a own-birth sibling (Table A2). If the latter e¤ect is greater, this would
support the idea that parents treat adopted children worse. We obtain no
strong evidence for this hypothesis, since the resulting point estimates for most
25
Note, however, that we now rely on a di¤erent sample, where we cannot rule out that
the e¤ects for families with only own-birth children would also have been di¤erent.

17
outcomes are rather similar in both types of families and since the estimates
are imprecisely estimated in the mixed families. These …ndings support those
obtained in previous studies, such as Björklund et al. (2006) and Holmlund et
al. (2010).
Finally, we test for any signs that nurturing is less e¤ective for adopted
sons compared to own-birth sons. The idea here is that those adopted early
are less exposed to detrimental conditions in their home country, which means
that nurturing may be more e¤ective.26 By comparing the e¤ects to those
who were adopted at later ages, we can get some indications if nurturing is
less e¤ective for those adopted late. For cognitive skills, we do not obtain any
large di¤erences in the results for those adopted before and after the age of 1,
neither for mothers, nor for mothers (Table A3).
For non-cognitive skills, the e¤ect of fathers’education is smaller for
those adopted later, although one should note that the absolute di¤erence
in point estimates is not great. The greatest di¤erence is obtained for overall
health, where the e¤ect of both mothers’and fathers’education is substantially
greater if the child was adopted before the age of 1. This supports the idea
that parental education has less of an e¤ect on health if the child had been
exposed to adverse living conditions early in life. For those adopted before the
age of 1, both mothers and fathers’schooling show a positive impact on the
child’s health, although the e¤ect of mothers’ schooling is still greater. For
our height measure, our estimates are imprecise, although the point estimates
are actually slightly larger for those arriving later.
To sum up, we have shown that although the characteristics of the
adoptive parents and adoptees di¤er greatly from those of the other samples
used, we obtain rather few indications that the e¤ect of parents’ schooling
di¤er greatly between adoptive parents and twin parents. The clearest result
was obtained for health, were the sensitivity analysis supported the hypothesis
that parents’schooling has less of an e¤ect if the adopted child was adopted
late. This may also explain some of the smaller e¤ect on health obtained for
adoptees compared to children of twin parents. We leave this to future research
to further explore.

Are earlier grades and tracking a¤ecting child skills and health at
18?
Our results so far are consistent with the idea that parental education matters
for children’s education partly because parental education improves children’s
skills. One could not rule out the possibility, however, that child skills, as
measured at age 18, are a¤ected by earlier study results or choice of track in
school, so that parental education a¤ects children’s skills by improving their
study outcomes. We analyzed this by analyzing the link between parental
education and children’s grades for the subsample on which we have grade
information. The results showed small and insigni…cant e¤ect for grades in
26
Evidence from Romanian adoptees adopted to UK, for instance, suggests that if the
adoption took place during the …rst 6 months of life, catch-up in terms of height and IQ
was almost complete (Rutter et al. 1998). For those adopted later, catch-up was less, but
still substantial.

18
language and mathematics at age 15, however.

How large are the e¤ects and how do they relate to previous …nd-
ings?
The estimates we obtain in the adoption and twin design may appear rather
small, but so are also the previous estimates on the e¤ect of parents’educa-
tion on children’s education that have been obtained in Scandinavian countries.
But are our results reasonable, given the previous estimates on the intergen-
erational transmission of human capital? We can shed some light on this, by
examining to what extent our estimates, together with previous estimates on
the returns to skills, could generate the observed transmission of schooling.
Starting with cognitive skills and twins, our twin-based estimates for
fathers and mothers schooling converge around 0.05 in most speci…cations.
This means that one additional year of fathers’schooling is associated with a
5 percent of a standard deviation increase in the child’s cognitive skills. Sande-
wall et al. (2010) provide twin-based estimates of the returns to cognitive skills
and Lundborg et al. (2011) provides twin-based estimates on the returns to
both non-cognitive and cognitive skills, where in both studies a one standard
deviation increase in cognitive skills is associated with about 0.5 additional
years of schooling. This means that a 5 percent of a standard deviation in-
crease in cognitive skills, which resulted from one additional year of parental
schooling, is associated with 0.025 additional years of schooling by the child.
If we assume a similar return to non-cognitive skills, as indicated by
the results of Lundborg et al. (2011), and take our estimated e¤ect where one
additional year of fathers’schooling is associated with a 4 percent of a standard
deviation increase in non-cognitive skills, this would imply a .020 increase in
years of schooling of the child. Since previous twin-based estimates of the
e¤ect of fathers’ schooling on children’s schooling obtained in Scandinavian
countries are often found to be in the ballpark of .06 - .10, our results for
cognitive and non-cognitive skills would thus be able to explain a large part
of this causal transmission of schooling across generations. This is reassuring
and makes our estimates appear very reasonable.
Finally, we may use our estimates to say something about the policy
relevance of our results. If we assume that one additional year of parental
schooling increases the child’s skills in a way that makes him obtain about
0.05 additional years of schooling, this also means that 3 additional year of
schooling, equivalent to a bachelor degree, would increase child schooling by
about 0.15 years, if we assume linear returns. Since previous studies, such
that Lundborg et al. (2009), have established that the wage returns to years
of schooling in Sweden is about 6 percent, this would increase yearly income
increase by almost 1 percent. One a life-time basis, this is not trivial. Since
we show, however, that skills are most likely an important mechanism through
which the transmission of human capital arises, the bene…cial e¤ects of parents’
education are most likely not restricted to income but translates into other
areas as well. Previous studies on the social returns to early life skills, such
as Heckman et al. (2006), have shown that increases in non-cognitive skills
leads to decrease in crime among other things. This means that although

19
the e¤ect of one additional year of parents’schooling on one particular child
outcome may be small, the total e¤ect added up over di¤erent outcomes may
be substantial.

VI Summary and conclusion

A recent literature, using an adoption and twin design, has established that
the intergenerational transmission of education partly re‡ects a causal e¤ect
of parents’ education. In this paper, we provide some clues regarding what
it is that well-educated parents bring to their children, so that their children
obtain more education. We focused on factors such as early life skills and
health that could plausibly be important mechanisms for the intergenerational
transmission of education and employed the most common methods used in
the recent literature; the adoption and the twin design.
Our results suggest that cognitive and non-cognitive skills, as well as
health, may be important factors in understanding the intergenerational trans-
mission of human capital. In both the adoption and the twin design, parents’
schooling had a positive impact on these child outcomes and the estimates were
rather similar across the twin and adoption designs. Our results also suggest
that fathers’education was of greater importance for the development of the
child’s cognitive and non-cognitive skills, whereas mothers’education matter
more for the child’s health. Since it is well established that early life skills
and health a¤ects later educational attainment, our results thus suggest that
part of the causal e¤ect of parental schooling on children’s schooling re‡ects
an e¤ect of parental schooling on children’s skills and health.
We believe that our estimates appear very reasonable, given the pre-
vious estimates using an adoption and a twin design that estimate the inter-
generational transmission of schooling in Scandinavian countries. Using our
estimates of the e¤ect of fathers’ schooling on skills, and in addition using
previous …ndings on the returns to skills, we showed that our estimates nicely
…tted in with those obtained for the transmission of education in previous
adoption and twin studies in Scandinavian countries. Moreover, we showed
that although small, the estimated e¤ects are far for trivial, especially when
one considers that an increase in various types of skills, as well as health, most
likely has positive e¤ects on a far wider range of outcomes than just earnings.
Returning to some of the policy issues discussed in the introduction,
our results suggest that interventions that increase the level of schooling in
society may have bene…ts that not only transcends across generations but also
across a range of child outcomes. This is a useful …nding since education
is something that can be manipulated by policy-makers. We believe that a
fruitful path for future research will be to continue to try to understand the
mechanisms behind the intergenerational transmission of human capital.

20
References
Almond, D & Currie, J. 2010. "Human Capital Development Before Age
Five," NBER Working Papers 15827, National Bureau of Economic Research.
Amin, V., Lundborg, P., Rooth, DO (2011a). Mothers’do matter: New ev-
idence on the e¤ect of parents’schooling and children’s schooling using Swedish
twin data. IZA Discussion Paper No. 5946.
Amin, V., Lundborg, P., Rooth, DO (2011b). Following in your father’s
footsteps: A note on the intergenerational transmission of income between
twin fathers and their sons. IZA Discussion Paper No. 5990.
Anger, Silke and Guido Heineck (2009), “Do Smart Parents raise smart
children? The intergenerational transmission of cognitive abilities”, Working
paper no. 156 (SOEP, DIW, Berlin).
Antonovics, Kate and Arthur S. Goldberger. 2005. “Does Increasing
Women’s Schooling Raise the Schooling of the Next Generation? Comment.”
American Economic Review, 95(5): 1738-1744.
Ashenfelter, Orley and Alan B. Krueger. 1994. “Estimates of the Returns
to Schooling from a New Sample of Twins.”American Economic Review, 84(5):
1157-1173.
Behrman, Jere R. and Mark R. Rosenzweig (2002), “Does increasing women’s
schooling raise the schooling of the next generation?”American Economic Re-
view 92:323-334.
Bingley, P., Christensen, K., & Jensen, V. M. (2009). Parental Schooling
and Child Development: Learning from Twin Parents. SFI working paper.
Björklund, A., Jäntti, M. and Solon, G., 2005, “In‡uences of Nature and
Nurture on Earnings Variation: A Report on a Study of Various Sibling Types
in Sweden”. In Unequal Chances: Family Background and Economic Success,
edited by Bowles, S., Gintis, H. and Osborne Groves, M., Princeton: Princeton
University Press, pp. 145-164.
Björklund, A., Lindahl, M. and Plug, E., 2006, “The Origins of Inter-
generational Associations: Lessons from Swedish Adoption Data”, Quarterly
Journal of Economics 121, pp. 999-1028.
Björklund, A. and Salvanes, K. (2010). Education and Family Background:
Mechanisms and Policies. In: E.A. Hanushek, S. Machin, and L. Woessmann
(eds.), Handbook in Economics of Education, Vol. 3, North Holland: pp.
201-247.
Björklund Anders, Karin Hederos Eriksson, and Markus Jäntti (2010), “IQ
and family background: Are associations strong or weak?” The B.E. Journal
of Economic Analysis and Policy (Contributions) 10(1).
Black, S., Devereux, P. and Salvanes, K., 2005, ”Why the Apple Doesn’t
Fall Far: Understanding Intergenerational Transmission of Human Capital”,
American Economic Review 95(1), pp. 437-449.
Black, Sandra, Paul Devereux, and Kjell Salvanes (2009), “Like father, like
son? A note on the intergenerational transmission of IQ scores”, Economics
Letters 105: 138-140.
Black, SE, Devereux, P. (2010). Recent Developments in Intergenerational
Mobility. Forthcoming in: O. Ashenfelter and D. Card (eds.), Handbook of
Labor Economics, Amsterdam: Elsevier.

21
 Bound, J. and Solon, G., 1999, “Double trouble: on the value of twins-
based estimation of the return to schooling”, Economics of Education Review,
18(2), pp. 169-192.
Breierova, L. & Esther Du‡o, 2004. "The Impact of Education on Fertility
and Child Mortality: Do Fathers Really Matter Less Than Mothers?," NBER
Working Papers 10513, National Bureau of Economic Research,
Carneiro, Pedro, Costas Meghir, and Matthias Parey. 2007. “Maternal
Education, Home Environments. and the Development of Children and Ado-
lescents.”Institute of Fiscal Studies Working Paper W07/15.
Chen, Y, Li, H. (2009). Mother’s education and child health: Is there a
nurturing e¤ect? Journal of Health Economics: 28(2): 413-426.
Chevalier, Arnaud, and Vincent O’Sullivan. 2007.“Mother’s Education and
Birth Weight.”IZA Discussion Paper 2640.
Chevalier, Arnaud. 2004. “Parental Education and Child’s Education: A
Natural Experiment”, IZA discussion paper No. 1153.
Chou, Shin-Yi, Jin-Tan Liu, Michael Grossman, Theodore Joyce. 2007.
"Parental Education and Child Health: Evidence from a Natural Experiment
in Taiwan." NBER Working Paper No. 13466.
Currie, J (2009). "Healthy, Wealthy, and Wise: Socioeconomic Status,
Poor Health in Childhood, and Human Capital Development," Journal of Eco-
nomic Literature, vol. 47(1), pages 87-122
Currie, Janet. and Enrico Moretti 2003. "Mother’s Education and the In-
tergenerational Transmission of Human Capital: Evidence from College Open-
ings," Quarterly Journal of Economics VCXVIII: 1495-1532.
Currie, Janet. and Enrico Moretti (2007). "Biology As Destiny? Short and
Long-Run Determinants of Intergenerational Transmission of Birth Weight"
Journal of Labor Economics, 25(2): 231-264.
Dearden, Lorraine S., Stephen Machin and Howard Reed. (1997). “Inter-
generational Mobility in Britain.”Economic Journal, 110(440): 47-64.
Elo, I.T. and S.H. Preston. 1992. "E¤ects of early-life conditions on adult
mortality: a review." Population index, 58, 186-212.
Griliches, Zvi (1979), "Sibling Models and Data in Economics: Beginnings
of a Survey," Journal of Political Economy, University of Chicago Press, vol.
87(5), pages S37-64.
Grönqvist, Erik, Björn Öckert, and Jonas Vlachos (2009), “The Intergen-
erational transmission of cognitive and non-cognitive abilities.”CEPR Discus-
sion Paper 7908.
Hægeland, Torbjørn, Lars Kirkebøen, Oddbjørn Raaum and Kjell Salvanes.
2010. “Why Children of College Graduates Outperform their Schoolmates: A
study of Cousins and Adoptees.”Mimeo, University of Bergen.
Heckman, J., Stixrud, J. and S. Urzua. 2006. “The E¤ects of Cognitive
and Noncognitive Abilities on Labor Market Outcomes and Social Behavior.”
Journal of Labor Economics 24(3):411-482.
Holmlund, Helena, Mikael Lindahl and Erik Plug. 2008. ”The Causal
E¤ect of Parent’s Schooling on Children’s Schooling: A Comparison of Esti-
mation Methods.”IZA Discusson Paper No. 3630.
Holmlund H, Lindahl, M, Plug, E. (2010), The Causal E¤ect of Parents’
Schooling on Children’s Schooling: A Comparison of Estimation Methods.

22
forthcoming Journal of Economic Literature.
Lichtenstein, Paul, Ulf de Faire, Birgitta Floderus, Magnus Svartengren,
Pia Svedberg and Nancy L. Pedersen (2002), ”The Swedish Twin Registry: A
unique resource for clinical, epidemiological and genetic studies”, Journal of
Internal Medicine, 252: 184-205.
Lindeboom, M., van der Klaauw, B., Nozal, AL. (2009). Parental educa-
tion and child health: evidence from a schooling reform. Journal of Health
Economics 28, 109-131.
Lindqvist, M., and Hjalmarsson, R. (2010). Driving Under the In‡uence
of Our Fathers. The B.E. Journal of Economic Analysis & Policy 10(1): 1-15.
Lindqvist, M., and Hjalmarsson, R. (2011a). Like Godfather, Like Son:
Explaining the Intergenerational Nature of Crime. Forthcoming in Journal of
Human Resources.
Lindqvist, M., and Hjalmarsson, R. (2011b). The Origins of Intergenera-
tional Associations in Crime: Lessons from Swedish Adoption Data" CEPR
Discussion Paper No. 8318, April 2011.
Lindqvist, E., Vestman, R. (2010). The Labor Market Returns to Cognitive
and Noncognitive Ability: Evidence from the Swedish Enlistment" forthcom-
ing in American Economic Journal: Applied Economics.
Lundborg, P., Nystedt, P., Rooth, DO. (2009). The Height Premium in
Earnings: The Role of Cognitive Ability and Physical Strength. IZA Working
Paper 4266.
Lundborg, P., Nilsson, A., Rooth, D. (2011). Does Health Predict School-
ing Di¤erences Within Monozygotic Twin Pairs? IZA Discussion Paper No.
503.
McCrary, J., Royer, H. (2011). "The E¤ect of Maternal Education on
Fertility and Infant Health: Evidence from School Entry Policies Using Exact
Date of Birth" American Economic Review, forthcoming.
Plug, E., 2004, “Estimating the E¤ect of Mother’s Schooling on Children’s
Schooling Using a Sample of Adoptees”, American Economic Review 94(1),
pp. 358-368.
Plug, E. and Vijverberg, W., 2003, Schooling, Family Background, and
Adoption: Is It Nature or Is It Nurture?”, Journal of Political Economy,
111(3), pp. 611–641.
Plug, E. and Vijverberg, W., 2005, “Does Family Income Matter for School-
ing Outcomes? Using Adoptees as a Natural Experiment”, Economic Journal
115(506), pp. 879-906.
Pronzato, Chiara. (2010) “An Examination of Paternal and Maternal Inter-
generational Transmission of Schooling." Forthcoming in Journal of Population
Economics.
Rutter, M. et al. (1998), Developmental catch-up, and de…cit, following
adoption after severe global early privation, Journal of Child Psychology and
Psychiatry, 39, 465-476.
Sacerdote, Bruce. 2000. “The Nature and Nurture of Economic Out-
comes.”NBER Working Paper 7949.
Sacerdote, B., 2002, “The Nature and Nurture of Economic Outcomes,”
American Economic Review Papers and Proceedings, 92(2), 344–348.
Sacerdote, Bruce (2007), “How large are the e¤ects from changes in family

23
environment? A study of Korean American adoptees”, Quarterly Journal of
Economics 122: 119-157.
Sandewall, Ö., Cesarini, D., Johannesson, M., 2009. The Co-Twin Method-
ology and Returns to Schooling –Testing a Critical Assumption. IFN Working
Paper no. 806, IFN, Stockholm.
SOU, 2003, "Adoptionsutredningens betänkande Adoption - till vilket pris?",
SOU 2003:49.

24
 Table 1: Descriptive statistics. Data from the Swedish twin registry.

Own-birth sons Adopted sons Son of twins

Mean St. dev Mean St. dev Mean St. dev
Cognitive ability 5.20 1.90 4.22 1.91 5.24 1.83
Non-cognitive ability 5.19 1.66 4.74 1.71 5.33 1.59
Health 2.69 4.23 4.20 4.8 2.22 3.69
Height 179.6 6.48 170.2 6.59 179.68 6.19
Age at adoption 1.69 1.92
Father’s education 1.69 1.92 12.69 2.73 10.56 3.12
Mother’s education 11.49 2.30 12.80 2.61 10.63 2.71
Notes: This table reports descriptive statistics.

25
 Table 2: Estimates of the relationship between parental education level and son’s outcomes.

Cognitive skills Non-cognitive skills

Own-birth Adopted Twins Own-birth Adopted Twins
OLS FE OLS FE
Father’s 0.099 0.027 0.097 0.044 0.057 0.017 0.061 1,165
education (0.000)*** (0.006)*** (0.008)*** (0.019)** (0.000)*** (0.006)** (0.009)*** (0.021)*
N (pair, di¤) 748,598 3,692 1,245 1,245 721,017 3,375 1,165 1,165
(475, 239) (449, 226)
Mother’s 0.096 0.014 0.100 0.050 0.056 0.003 0.055 0.019
education (0.000)*** (0.006)** (0.008)*** (0.017)*** (0.000)*** (0.007) (0.008)*** (0.019)
N (pair) 880,791 3,692 1,928 1,928 851,128 3,375 1,818 1,818

26
Health Height
Own-birth Adopted Twins Own-birth Adopted Twins
OLS FE OLS FE
Father’s 0.046 .040 .012 -.057 0.177 0.070 0.250 0.114
education (.002)*** (.030) (.035) (.084) (0.002)*** (0.037)* (0.056)*** (0.118)
N (pair, di¤) 769,031 3,631 1,303 1,303 745,841 3,741 1,228 1,228
(492, 248) (468, 238)
Mother’s 0.018 .054 .013 .168 0.216 0.032 0.207 0.112
education (.002)*** (.031)* (.031) (.076)** (0.002)*** (0.039) (0.055)*** (0.299)
N (pair) 903,891 3,631 1,982 1,982 877,845 3,741 1,913 1,913
(716, 310) (694, 296)
Notes: Each coe¢ cient comes from separate estimation. The dependent variables are the son’s cognitive
skills, non-cognitive skills, health, and height at age 18. For adopted children age at adoption and country of
birth dummies are included in the model. For the …xed e¤ect model for cousins with monozygotic twin
parents the number of pairs is reported together with the number of observations. In the twin regressions, the
sample size refers to the number of children of complete pairs of twin parents. The numbers within
parentheses refer to the number of complete pairs of twin parents and the number of twin parents where
there exists a di¤erence in education. Standard errors are shown below the coe¢ cients.
Table 3: Estimates of the relationship between parental education level and
son’s outcomes when accounting for assortative mating. Adopted sons.

Cognitive Non-cognitive Health Height

Father’s education 0.027 0.021 .019 0.073
(0.006)*** (0.007)** (.034) (0.043)*
Mother’s education -0.001 -0.008 .044 -0.007
(0.007) (0.008) (.036) (0.045)
N 3,692 3,375 3,631 3,741
Each coe¢ cient comes from separate estimation. The dependent variables are
the son’s cognitive skills, non-cognitive skills, health, and height at age 18.
from linear OLS and FE regressions. For adopted children age at adoption and
country of birth dummies are included in the model. Standard errors are shown
below the coe¢ cients.

27
Table 4: Estimating the relationship between parental education level and
country of birth and age at adoption.

Chile Columbia India South Korea Sri Lanka Thailand Age adopted
Father’s education .002 .003 .001 -.007 -.005 -.002 -.016
(.002) (.002)* (.002) (.003)** (.002)*** (.002) (.013)
Mother’s education -.003 .005 .001 -.012 .003 .000 .004
(.002) (.002)** (.003) (.003)*** (.002)* (.002) (.014)
Notes: The dependent variable is shown in the top of the column. A linear probability model
is estimated. Standard errors in parenthesis.

28
Table 5: Estimates of the relationship between parental education level and
son’s outcomes when accounting for assortative mating. Twin fathers and
mothers.

Cognitive Non-cognitive Health Height

Twin fathers
Twin father’s education 0.033 0.041 -0.055 0.115
(0.019)* (0.022)* (.061) (0.119)
Mother’s education 0.056 0.001 -0.009 0.000
(0.016)*** (0.019) (0.073) (0.102)
N 1,245 (475) 1,165 (449) 1,303 (492) 1,228 (468)

Twin mothers
Twin father’s education 0.052 0.038 0.067 0.139
(0.013)*** (0.016)** (.061) (0.081)*
Mother’s education 0.043 0.013 0.159 0.092
(0.017)*** (0.019) (.076)** (0.100)
N 1,928 (698) 1,818 (661) 1,982 (716) 1,913 (694)

Notes: The dependent variables are the son’s cognitive skills, non-cognitive
skills, health, and height at age 18. Standard errors are shown below the
coe¢ cients.

29
Table 6: The e¤ect of parental education for own-birth sons with an adopted
sibling.

Cognitive Non-cognitive Health Height

Father’s education 0.058 0.015 0.033 0.111
(0.009)*** (0.009) (0.042) (0.070)
Mother’s education 0.097 0.018 0.049 0.105
(0.009)*** (0.010)* (0.047) (0.068)
N 1,444 1,389 1,278 1,450
Notes: In all columns age at adoption and country of birth dummies are.
included in the modelStandard errors are shown below the coe¢ cients.

30
Table 7: The e¤ect of parental education for adopted sons with a biological
sibling.

Cognitive Non-cognitive Health Height

Father’s education 0.016 0.020 0.002 0.097
(0.011) (0.013) (0.063) (0.077)
Mother’s education 0.019 0.012 0.037 -0.049
(0.012) (0.014) (0.069) (0.085)
N 819 754 797 823
Notes: In all columns age at adoption and country of birth dummies are.
included in the modelStandard errors are shown below the coe¢ cients.

31
 Table 8: Estimating the relationship between parental education level and country of birth and age at adoption.

Age at migration 1 Age at migration 1

Cognitive Non-cognitive Health Height Cognitive Non-cognitive Health Height
Father’s .020 .008 .003 .079 .034 .030 .097 .047

32
education (.007)*** (0.008) (.036) (.045)* (.010)*** (.012)** (.053)* (.067)
Mother’s .014 -.003 -.006 .052 .014 .017 .163 -.017
education (.017)** (.008) (.038) (.047) .014 (.012) (.055)*** (.070)
N 2,555 2,341 2,494 2,587 1,137 1,034 1,137 1,154
Notes: The dependent variable is shown in the top of the column. A linear probability model
is estimated. Standard errors in parenthesis.