The Intergenerational Transmission of Adiposity across Countries
Peter Dolton
(University of Sussex)
Centre for Economic Performance
London School of Economics)
and
Mimi Xiao
(University of Sussex)
Abstract
There is a worldwide epidemic of obesity. We are only slowly beginning to understand its
causes and consequences. This paper addresses one important component of the crisis –
namely the extent to which obesity – or more generally - adiposity is passed down from one
generation to the next. Using the BMI as a measure of adiposity, we find that the
intergenerational transmission of adiposity is relatively high and very comparable across time
and countries – even if these countries are at a very different stage of economic development.
Our second key finding is that this intergenerational transmission mechanism is very different
across the distribution of children‘s BMI. Most specifically it is up to double for the fattest
children what it is for the thinnest children. This has enormous consequences for the health
of the world‘s children.
Address for Correspondence:
Prof Peter Dolton
Department of Economic
University of Sussex,
Brighton
BN1 9SL
1
Introduction
There is a worldwide epidemic of obesity and obesity has become one of the foremost major
public health problem in most countries. In 2013, the US spent 190 billion dollars on obesityrelated health expenses.
We are only slowly beginning to understand the causes and
consequences of childhood obesity. This paper addresses one important component of the
crisis – namely the extent to which obesity – or more generally - adiposity is passed down
from one generation to the next.
Hence, our underlying question is: what is the driving force behind rising childhood obesity?
Adiposity is a result of both the genetic process of inheritance and a result of decisions made
in families. To understand the process of obesity it is crucial to understand the
intergenerational transmission mechanism behind this. Evidence ( Maes, Neale, & Eaves,
1997 ) suggests that adiposity is affected by both environmental and genetic factors. Clearly,
the intergenerational transmission mechanism we are studying here operates through both
these two channels. So it is transmitted through family environmental factors, which take the
form of intra-household mechanism (how the resources are allocated within the family), and
it is also affected by genetic factors through a direct channel. Therefore, through exploring
the elasticity of adiposity ( measured by the BMI score) across generations in different
countries, we attempt to reveal the underlying intergenerational relationship in
anthropometric characteristics.
In order to give you a sense of what the underlying relationship between parents and child‘s
BMI looks like – we first of all present some basic non-parametric graphs of the data. First
we plot a non-parametric scatterplot based on the raw data. Figure 1 below is the local
weighted scatter smoothing of the log of father‘ BMI variable against the log of their
children‘s BMI variable, the slopes capture the magnitude of the intergenerational elasticity.
They suggest that the slopes are relatively flat and nearly parallel across countries, this
implies that the intergenerational elasticity are relatively high and approximately constant
across countries.
2
Figure 1a Lowess log (father’s BMI variable ) and log (child’s BMI variable)
2.8
3
3.2
3.4
3.6
Lowess
2
2.5
3
log bmi_father
China 1989-2009
British 1970-1996
US 1988-1994
3.5
4
Indonesia 1993-2007
England 1995-2010
Figure 1b Lowess log (mother’s BMI variable ) and log (child’s BMI variable)
2.9
3
3.1
3.2
3.3
3.4
Lowess
2
2.5
3
log bmi_mother
China 1989-2009
British 1970-1996
US 1988-1994
3.5
Indonesia 1993-2007
England 1995-2010
3
4
2. Evidence on the Intergenerational Transmission Mechanism.
Intergenerational studies originate with Francis Galton (1877). By running a regression of the
offspring‘s height on their parents‘ height, he argued that an individual‘s characteristics are
correlated with those of their parents and at the same time ―regress to mediocrity‖. More
specifically, the individual characteristics (such as height and weight) are closer to the
population mean than those of their parents (Galton, 1892). This finding was the basis of
Becker-Tomas model (1979, 1986) on intergenerational human capital transmission
(Goldberger, 1989; Han & Mulligan; Mulligan, 1999).
Most of these intergenerational transmission studies are about the transmission of income or
educational achievement outcomes. Since this transmission relates to the equality (or
otherwise) of individual opportunity over time, which exerts profound influences on the
social mobility. The strength of this transmission is usually measured by the elasticity of
children‘s income with respect to their parents‘ income (or the intergenerational elasticity of
income, hereafter called IIE). The larger is the IIE the more it means that the children‘s
position on the ―income ladder‖ is determined by their parents‘ income position. We would
naturally be concerned if this elasticity of the transmission mechanism was (too) large as it
would imply that both equity and efficiency of the society would be undermined. Specifically,
it means that the smaller is this elasticity the smaller the role played by one‘s parents in the
determination of the child‘s outcome in school or the labour market.
In terms of the discrepancy in IIE across countries, partly due to the restriction of data which
covers multiple generations, most of these studies are conducted in the US or European
countries. In the US, the consensus estimated IIE is ― 0.4 or a bit higher‖ ( for instance, 0.473
using PSID by Grawe (2004), 0.542 using NLSY sample born 1957-64 by Bratsberg et al.
(2007), this is higher than Canada (0.2 using register data (Corak et al. ,1999), 0.152 using
IID Canadian Intergenerational Income Data and 0.381 using PSID Panel Study of Income
Dynamics data (Grawe, 2004) and most of the European countries except for Britain ( 0.45
using NCDS 1958 cohort Bratsberg,et al., 2007) and Italy (0.48 using data Italy Survey on
Household Income and Wealth (SHIW), Piraino, 2007). The IIE estimates in Nordic
countries
or Scandinavian societies may be the lowest, ranging from 0.2 to 0.3 ( see
Pekkarinen, et al., 2009; Björklund,et al., 2012 ). In contrast, the IIE in China is perhaps at
the top of the list---0.63: a Chinese father‘s income 10 percent above the paternal cohort
4
mean will be associated with his son having an income 6.3 percent above the filial cohort
mean ( Gong, 2012 ). Using the Urban Household Education and Employment Survey
(UHEES) and the Urban Household Income and Expenditure Survey in1987-2004 ( UHIFS) ,
Gong et al. (2012) show the IIE in China is 0.63 for father-son, 0.97 for father-daughter, 0.36
for mother-son, and 0.64 for mother-daughter, and education is one of the most crucial
channels through which earnings ability is transmitted across generations. However, other
factors such as genes and health are also potentially important pathways of intergenerational
transmission in China.
Regarding the intergenerational transmission of education achievement, the intergenerational
mechanism can be written in the following formula, in which case the child‘s education
achievement is measured by the human capital: Hc= F (Yc, Hp, Ac), thus the mechanism
usually uses three main channels. The first channel is through the effects of parental income,
higher educated parents tend to have more income, so they have more resources to invest in
child‘s education (Yc) ; the second is through the effects of parental education (Hp), since
higher-educated parents may invest in child‘s education in a more efficient way. In addition
to these two indirect channels, parental education may also affect child‘s education through a
direct channel, which is usually proxied by the genetic inheritance of ability (Ac).
Empirically, the first channel can be decomposed into the effects of current parental income
and the effects of permanent parental income, of which the latter normally plays the dominant
role and might be measured by family fixed effects (Carneiro and Heckman, 2003). The
second channel also includes the motivation effects since higher educated parents may
encourage their children to go for a higher level of education (Boudon, 1974 ). The third
channel is normally conducted by comparing children of twin pairs ( Behrman and
Rosenzweig, 2002) or between biological and adopted children with variation in education
(Björklund et al., 2006), the general conclusion is that the intergenerational correlation in
education cannot be fully attributed to the genetic factors. The intergenerational correlations
of education is estimated at 0.3 ~ 0.4 (Alburg, 1998), but a more popular measure of this
intergenerational educational relationship is intergenerational education elasticity (hereafter
called IEE) , which varies from 0.14~ 0.45 in the USA (Mulligan, 1999) to 0.25 ~0.4 in the
UK ( Dearden et al., 1997). It is worth mentioning that a large body of the literature looks at
the intergenerational elasticity of IQ (hereafter called IQE), which is used to measure the
5
intergenerational relationship in the third channel, the magnitude of IQE ranges from 0.3 to
0.5 ( Solon, 2004; Anger and Heineck , 2010; van Leeuwen et al.,2008 ).
There is also a growing literature on the intergenerational correlation in various health
outcomes, such as weight, height , BMI, self-rated measures (Coneus and Spieß, 2008),
depression ( Akbulut and Kugler ,2007), and smoking behaviours (Loureiro et al. ,2006).
Using data from the German Socio-Economic Panel (SOEP), Coneus and Spieß (2008)
estimate the intergenerational relationship of both father and mother and children. Regarding
the fixed effects estimates, using actual BMI and obesity as the anthropometric measures ,
they find that father‘s BMI has a significantly positive effect on child‘s BMI ( with a
coefficient of 0.57, but the effects of mother‘s is not significant), whereas mother‘s obesity is
strongly associated with child‘s obesity with a coefficient of 0.26. They claim this as a
―transmission‖ rather than merely ―link‖. However, in their data, child‘s health outcomes are
provided by mother rather than professionals, additionally, father and mother‘s health are
self-reported, this might lead to biased estimates from measurement error. Classen ( 2010 )
estimates the intergenerational elasticity of BMI when both generations are between the ages
of 16 and 24. Applying a regression which only controls for mother and child‘s BMI, he finds
an elasticity of 0.35 between mother and child‘s BMI, but typically this sort of long data is
not available. As Black and Devereux (2010) review, few papers have claimed a causal link,
since the family environmental factors may affect the health outcome of both parents and
children. Some studies try to address this by differencing out fixed family characteristics
through comparing ―sibling mothers‖ or looking within ―twin pairs of mothers‖, assuming
twins share the same environment and genetics (Currie and Moretti ,2007; Black, Devereux,
and Salvanes, 2007; Royer ,2009)
In the case of income and education, the intergenerational elasticity varies largely across
countries, and the environmental channels are exploited more often than the genetic channel.
Whether this transmission mechanism is applicable to the case of anthropometric outcome
( measured by BMI variable in this paper) generates the main motivation of our study. The
difference of IBE from IIE and IEE hinges on the relative role and the interaction of
environmental and genetic forces in the intergenerational transmission, our assumption is that
in the transmission of BMI variable, a smaller fraction of the operation forces are open to
manipulation (such as the diet change with the household---environmental change ), and a
6
larger fraction of the forces are driven by the ―natural process‖. In other words, in the case of
health outcome such as the BMI variable, they are more likely to be inherited genetically
regardless of the change in environment. If this assumption is true, our estimation for IBE
may provide a lower bound of the intergenerational correlation in any characteristics
including the income and education.
In other words, assuming the intergenerational
transmission of anthropometric outcome is entirely determined by the genetic traits, if our
IBE is closer to the IIE in Scandinavian societies where the IIE is the lowest---0.2, it may
imply that the relationship between parents and child cannot be lower than this threshold, in
spite of the change in either family environment ( such as the shift of nutrition pattern ) or
socioeconomic environment ( such as the innovation or marketing campaign in food industry).
In addition to ―regression to the mean‖ in the inheritability of BMI variable, the degree of this
inheritability (IBE) may vary across child‘s BMI distribution and this variation usually relates
to the family‘s socioeconomic status in the society. The general conclusion in the literature is:
in either developed countries or developing countries, the intergenerational correlation in
health measure tends to be stronger at lower SES levels ( see, for example, Currie and Moretti,
2007;
Bhalotra and Rawlings , 2009). In developing countries, this strong correlation
emerges at the lower levels of BMI, whereas in developed countries such as the US, this also
occurs at higher levels of BMI (Classen, 2010, Laitinen et al., 2001; Scholder et al., 2012),
one explanation is that in these countries, fast food industry is more developed, these
―unhealthy‖ food are generally cheaper than ―healthy‖ food , thus lower income families tend
to consume these ―unhealthy‖ food which is viewed as a cause of obesity.
Thus far, we can see these intergenerational studies are essentially derived from the model of
―Regression to the mean‖ (Becker or even dates back to Galton). Based on this framework,
the intergenerational transmission of BMI variable in this paper can be modelled in the
following way:
(
Where
)
(
)
(1)
denotes child‘s BMI variable,
or mother‘s or both), and
represents parents‘ BMI variable ( father‘s
vc denotes the random determinants of
7
. Thus,
, the intergenerational elasticity of BMI variable ( hereafter called IBE) measures the
degree of inheritability of BMI variable if we assume BMI is determined genetically as the
innate ability in Becker‘s innate ability model.
3. An Empirical Model of Intergenerational BMI Transmission
In this section, we outline an empirical model on intergenerational transmission of BMI. This
model is directly analogous to Becker‘s model on intergenerational transmission of income.
In Becker‘s model, parents allocate their income between the child‘s income, and his own
consumption, to optimize their utility. In our model, the outcome of interest is a child‘s BMI
health which can be invested by parents sacrificing their own consumption (literally- maybe
even their own food) – to promote the improved health-as measured by BMI of the child.
Hence here,
denotes the child‘s health as the intergenerational outcome we are interested
in. The child‘s health is a function of parent‘s income, , and a genetic endowment , which
is determined exogenously. Since children cannot choose their parents and the genetic traits
they inherit from them, then this endowment factor is taken as given (by the exogenous
―natural selection process‖).
(1)
Where
is decomposed into genetic factors,
, and environmental factors,
.
(2)
From this point on, we will use lower case letters to denote observable variables which we
obtain data on or can proxy for. Let subscript
index the parent and
index the child,
substituting (2) into equation (1), we obtain
(3)
8
As for the correlation between parents‘ genes and the child‘s genes, we consider the child‘s
as a production function of its father‘s genes,
genes,
and its mother‘s genes
, which
is assumed to take the form of Cobb-Douglas function1.
(4)
denotes the child‘s multiplicative scaled genetic transmission and
Where
denotes the
stochastic term. Equation (4) can be thought of as the ‗translog biological production
function‘, which maps parents‘ health outcomes into children‘s outcomes.
( )
( )
(
(
)
)
( )
(5)
Now consider substituting equation (5) into equation (3) where we assume that mothers and
fathers‘ BMI measures are sufficient statistics for their health and the environmental factors
are individual specific and captured by the term . We now wish to estimate the following
equation (6) in a cross-section framework.7
( )
(
(
)
)
(6)
where indexes individual child observations and
captures the transmitted stochastic error
term Equation (6) shows that child‘s health outcome
health outcome,
is a function of child ‘s father‘s
and the child‘s mother‘s health outcome,
,
denotes a vector of
parental income factors (or family environmental factors), such as the categories of
household income per capita, father and mother‘s SES factors. As both child‘s health and
parents‘ health are affected by time-invariant unobserved individual heterogeneity,
, such
as eating habits, health behavior and genetic components. we estimate equation (6) in an
individual fixed effect framework.
(
1
)
(
)
(
)
(7)
The justification of this functional form is purely for analytical tractability, and is not strict necessary. One similar
specification is in Rosenzweig, (1983), where birth weight is a function of health behavior such as smoking behaviour of the
parent.
9
Where denote observations referenced to a specific time period (or wave of the data). This
equation takes into account an individual fixed effect . Based on this equation and using
BMI as the measure of health, we estimate the intergenerational elasticity of health outcome,
conditional on time-variant family environmental factors which are captured by
.
However, it is also reasonable to assume that unobserved family environmental factors
remain constant. For instance, the family routine, such as eating and sleeping time shared
among household members, normally do not change over time; More importantly, the pattern
of food allocation among household members normally remain constant, these patterns,
together with, who is in control of the family income (Thomas, 1990), who takes a higher
level of energy-intensive activities (Pitt et al., 1990), whether the parents have a preference
for sons (Qian, 2008) or lower-birth-order children (Dasgupta, 1995), can affect both parental
and children‘s BMI outcome. Thus, household fixed effects are applied to estimate equation
(6), ie, fixed effects model is estimated using the following equation (8).
(
)
(
)
(
)
(8)
Where indexes household observations. In household , child ‘s health is a function of
father‘s health,
and mother‘s health,
,
denotes the household fixed effects. This
equation can only be identified when we have data on siblings for which the
distinct and the
effects are
are the same. We have a subset of our data for which we can estimate this
model- namely when we have data as more than one child in each household.
In the regression tables which follow we estimate single parent version of equation (7) and (8)
by variously restricting
, and
, then examining just
to examine the
pure effect of both parents but no conditioning factors ( simplified ‗ both parents equation‘).
In order to investigate the simplistic dynamic pattern of child‘s BMI measure, equation (9) is
estimated by including the previous value of child‘s BMI measure
the individual unobserved heterogeneity when
, Here we can net out
. Next, the age of the child is
included, in the regression equation (10), to allow for the possible of bias associated with the
use of the WHO software for China where children have comparatively lower than average
10
BMI, if this varies by age. The inclusion of the age of the child as a regressor also acts as a
proxy for the length of time the child has been exposed to the treatment of the family
environmental effect.
(
)
(
)
(
)
(
)
(9)
(
)
(
)
(
)
(
)
(10)
The empirical estimation will be conducted in several stages. First, we estimate the IBE at the
central level. Single parent version of equation (7) and (8) (father-child IBE and mother-child
(
IBE) are estimated using all the individual-wave observations,
) is also accounted
for to investigate the dynamic change of child‘s BMI variable, then equation (7), (8) are
estimated in a fixed effects framework. Second, we estimate single parent version of equation
(7) and (8) in terms of different levels of family social economic status, which includes
different levels of family income, mother‘s education, father‘s occupation and the time
duration when the family was in poverty, respectively. Third, applying single parent version
of equation (7) and (8), we estimate the IBE across different quantiles of child‘s BMI variable,
and compare the results with those using simplified both parents equation and equation (9).
Fourth, applying single parent version of equation (7) and (8), we estimate the IBE by age
group. Finally, we repeat our estimation of the IBE on samples aged 16~18 years old to
examine the probability that this relationship is structurally different when the children have
become adults.
4. Data and Measurement Issues.
The most widely used measure of body fat, or adiposity, is the Body Mass Index (BMI)
which is calculated using the following formula
[
(
)
(
)
]
, since we have
problems that when dealing with the measurement of Body Mass of children of different age,
we will not use the BMI per se, but the z-score of the BMI as adjusted by age and gender of
the child. This conversion can be made using the 2006 WHO Growth Standards for preschool
11
children and the 2007 WHO Growth Reference for school age children and adolescents. In
Stata, this conversion is implemented using a program from the WHO website2.
As mentioned in the literature review, the majority of intergenerational studies use elasticity
(eg. IIE and IEE) as a measure of the intergenerational relationship. To facilitate the
comparison of our results on anthropometric with other intergenerational results, we also
adopt elasticity as the measure and thus transform BMI z-score (ranging from -6 to 6) to BMI
variable (which ranges from 1th to 100th) to obtain the logarithm form in the calculation of
elasticity. In this way, this percentile indicates the relative position of individual BMI z-score
within the CHNS sample, since BMI z-score is based on the comparison with the WHO
population (the z-score normally distributed population constructed by the international
reference population using ‗Anthro‘), the BMI z-score percentile (hereafter called BMI
variable) is a relative anthropometric indicator based on the comparison with the WHO
population and then within the CHNS. Simply speaking, since the BMI z-score ranges from 6 to 6, and we can‘t take the log of a negative number, we transform this into a percentile of
BMI z-score to facilitate the taking of logs and the computation of an intergenerational BMI
elasticity.
An important problem we face is exactly how we correlate a child‘s BMI on their parents
BMI. Clearly a child‘s BMI is a function of their age and gender –so a simply correlation of
child‘s BMI against parents BMI would not allow for this factor. One way to examine the
intergenerational transmission is to wait until the child is an adult and then correlate the two
BMIs. This is what Classen did. There are two problems with this – firstly the logistic one is
that there is very little data when one has the child‘s height and weight observed when they
are an adult – as well as having their parents height and weight at the same time. The other
big problem with this is that we are mainly concerned with childhood obesity and so waiting
until they are adult does not help us.
One way around these problems is to take the child‘s BMI when they are young – use the
WHO‘s program to compute the child‘s BMI z score which explicitly allows for both the
2
They can be downloaded from http://www.who.int/childgrowth/software/en/ for Child growth standards
(0~5 years old) and http://www.who.int/growthref/tools/en/ for Growth reference (5~19 years old). See
Appendix 2 for a detailed description and discussion of BMI z-score.
12
child‘s age and gender. Once we have this z score we can then ask what would their BMI be
with such a z score when they are adult. The assumption that we have to make here is that
we are assuming that the child would remain in the same position in the distribution when
they are adult as when they are a child. Note that by doing this we are not, de facto,
assuming that this is what will happen to that child when they are an adult – but rather simply
getting an estimate of what the adult BMI is consistent with a given z score for the child.
Although this is a strong assumption there is only one other way to proceed.
This would be to simply use the child‘s BMI (as it is- even if they are very young) as the
dependent variable in a regression on the parents BMI on the assumption that if we control
for the child‘s age, gender, age squared and an interaction of the child with that of their
gender then we would have conditioned out for the non-linear effect of age on gender3. We
use both these methods as a robustness check on our findings. Fortunately they do not differ
much in their findings – with the former method giving lower variance in the tails than the
latter variable. We will therefore use the first method in each of our country datasets. We
report the second method in a n Appendix available on request for those interested.
In the course on doing this research we had considered if there was any alternative way of
retrieving the IBE. We contemplated using the WHO to generate z scores or percentiles and
using these logged metrics. It turns out that the estimation of the BMI elasticity is sensitive
to any possible transformation of its scale. – ie. to z scores or percentiles. So keeping the
analysis simple has many virtues. It turns out that estimating the model in the log of BMI or
the BMi itself does not make much difference – the elasticity is slightly smaller when
estimated without logging. But since taking logs allows albeit crudely – for general nonlinearity in the data and has the nice property that is preserves the constant elasticity across
the values of the BMI then we adopt it here. This means also that it forces the elasticity to be
a constant – which has the virtue that its first derivative (and hence the elasticity) is constant
across the whole range of the BMI score.
3
The weakness of this method is that we have to assume that we can net out for the whole non-linear process
of the child’s BMI rising as they age.
13
5 Empirical Evidence of intergenerational transmission
5.1 OLS
Applying equation (1), we estimate the IBE on CHNS data, Indonesian Family Life Survey
(IFLS) data, British 1970 cohort data, Health Survey for England (HSE) data and National
Health and Nutrition Examination Survey (NHNAES), respectively.
Table 1a reports the results on IBE when equation (1) controls for father‘s BMI variable
alone. It suggests that the father-child IBE estimates range from 0.198 in Indonesian sample
to 0.235 in BCS sample, and they do not vary substantially across countries ( this is different
from the previous studies on IIE and IEE). For the UK, The IBE estimate on BCS sample
(0.235) is close to that from HSE sample (0.202) 4 .These results suggest that the
responsiveness of child‘s BMI variable to parents‘ BMI variable is around 0.20 and the extent
of this ― inheritability‖ is quite constant across countries. In other words, if the father‘ s BMI
variable is 50% above the mean of their generation, on average his child‘s BMI variable
would be around 10% ( 50% * 0.20) above the mean of the children‘s generation, and this
seems to be regardless of the ethnicity and economic growth level. In a similar way, Table 1b
presents mother-child IBE estimates from these samples, and we see a similar pattern as
given by father-child IBE estimates. In addition, comparing Table 1a and 1b, we can see that
in general, the mother-child IBE estimates are greater than father-child IBEs, the only
exception is found in CHNS sample. Next, we incorporate both father and mother‘s BMI
variables (
(
) and
(
)) into equation (1), and the results are reported in
Table 1c. As we expect, once we control for both father and mother‘s BMI variables, the
sizes of paternal and maternal BMI effects shrink significantly compared with Table 1a and
Table 1b, with the dominance of father‘s BMI effects .
4
Notice the HSE was collected from 1995 to 2010, and the BCS 1970 survey tracks the cohorts born in 1970
until they reached 26 years (1996). This seems to tell us that the IBE in Britain was declining from 1970 to 2010,
if HSE and BCS 1970 can talk to each other ( by which I mean they have similar design and structure).
14
Table 1a father-child
China
Indonesia
BCS
HSE
logbmi_father 0.225*** 0.198*** 0.235*** 0.202***
(0.0154) (0.0106) (0.0104) (0.00948)
Constant
2.356*** 2.407*** 2.348*** 2.547***
(0.0476) (0.0324) (0.0331) (0.0312)
Observations
R-squared
13,990
18,570
21,505
26,316
0.024
0.024
0.035
0.022
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
NHANES3
0.204***
(0.0171)
2.603***
(0.0557)
6,515
0.023
Table 1b mother-child
China
Indonesia
BCS
HSE
logbmi_mother 0.186*** 0.165*** 0.201*** 0.188***
(0.0144) (0.00877) (0.00832) (0.00789)
Constant
2.477*** 2.500*** 2.468*** 2.600***
(0.0444) (0.0274) (0.0260) (0.0256)
Observations
R-squared
13,990
18,570
22,650
26,316
0.018
0.022
0.040
0.031
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
NHANES3
0.166***
(0.0130)
2.733***
(0.0418)
6,515
0.027
Table 1c father-mother-child
China
logbmi_father
0.193***
(0.0155)
logbmi_mother 0.148***
(0.0143)
Constant
2.000***
(0.0594)
Observations
R-squared
Indonesia
BCS
HSE
NHANES3
0.160***
(0.0108)
0.129***
(0.00891)
2.119***
(0.0377)
0.200***
(0.0103)
0.176***
(0.00850)
1.908***
(0.0390)
0.166***
(0.00937)
0.165***
(0.00787)
2.127***
(0.0369)
0.164***
(0.0173)
0.139***
(0.0131)
2.284***
(0.0624)
13,990
18,570
21,246
26,316
0.035
0.037
0.065
0.045
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
15
6,515
0.041
5.2 Quantile
Thus far, the estimates for IBE are at the conditional mean of child‘s BMI variable. In order
to explore the variation of IBE across different quantiles of child‘s BMI variable (or whether
the association of mother or father‘s BMI variable is constant across multiple quantiles of the
child‘s BMI variable distribution), in this section, we estimate the quantile elasticities of BMI
variable between father and child at different points in the distribution of the child‘s BMI zscore.
The results are displayed in Figure 2. They suggest that the degree of BMI transmission
shows a roughly increasing trend throughout child‘s BMI distribution in all the samples,
father-child IBE tend to be stronger at the higher level of child‗s BMI variable. In other
words, the effects of shared environmental and genetic factors between father and child tend
to be larger for the fatter children.
16
Figure 2: Quantile estimates of IBE relative to OLS elasticity
Elasticity father-child China
0.30
0.20
Elasticity
0.20
0.00
0.10
0.10
0.15
Elasticity
0.25
0.40
0.30
Elasticity father-child Indonesia
.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95
Quantile (conditional log bmi_adchild)
.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95
Quantile (conditional log bmi_adchild)
Elasticity father-child BCS1970
0.20
0.10
0.10
0.15
0.15
Elasticity
Elasticity
0.20
0.25
0.30
0.25
Elasticity father-child HSE
.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95
Quantile (conditional log bmi_adchild)
.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95
Quantile (conditional log bmi_adultchild)
17
0.20
0.15
0.10
0.05
Elasticity
0.25
0.30
Elasticity father-child NHANES3
.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95
Quantile (conditional log bmi_adultchild)
5.3 Transition Matrices and Markov Chain
5.3.1 Transition Matrices
In this section, instead of using continuous variable of BMI, we use categories of BMI status
as the anthropometric measure to investigate the probabilities of transition between BMI
status across generations. The individuals with average BMI z-score under -1.64 are
classified as underweight, -1.64~1.04 as normal weight, 1.04~1.64 as overweight, and above
1.64 as obese.
18
Table 2 China
Child’s BMI status
Full sample
Mother’s
(N= 14011)
distribution
BMI z-score
<-1.64
-1.64-1.04
1.04-1.64
>1.64
Category
Underweight
Normal
Overweight
Obese
Mother’s
<-1.64
Underweight
25.45%
65.45%
2.4%
6.7%
1.2%
BMI status
-1.64-1.04
Normal
11.3%
74.5%
6.57%
7.6%
82.94%
1.04-1.64
Overweight
6.25 %
75%
9.4%
9.3%
11.65%
>1.64
Obese
6.6%
70.7%
10.86%
11.76%
4.7%
10.68%
74.84%
7.09%
8.05%
Child‘s distribution
19
Table 3: Indonesia
Child’s BMI status
Full sample
Mother’s
(N= 18755)
distribution
BMI z-
<-1.64
-1.64-1.04
1.04-1.64
>1.64
score
Category
Underweight
Normal
Overweight
Obese
Mother’s
<-1.64
Underweight
30.1%
62.54%
2.2%
5.18%
3.2%
BMI status
-1.64-1.04
Normal
18.11 %
71.61 %
4.04%
6.24%
70.36%
1.04-1.64
Overweight
12.47 %
70.38%
5.32%
7.83 %
14.24%
>1.64
Obese
9.73 %
72.85 %
7.3%
10.08%
12.22%
16.67 %
71.86%
4.6%
6.9%
Child‘s distribution
20
Table 4 UK (BCS 1970 cohorts)
Child’s BMI status
Full sample
Mother’s
distribution
BMI z-score
<-1.64
-1.64-1.04
1.04-1.64
>1.64
Category
Underweight
Normal
Overweight
Obese
Mother’s
<-1.64
Underweight
1.42 %
45.39 %
24.11 %
29.08 %
2.14 %
BMI status
-1.64-1.04
Normal
1.50 %
44.03 %
18.15 %
36.32 %
53.59 %
1.04-1.64
Overweight
1.01 %
39.35 %
18.16 %
41.48 %
13.55%
>1.64
Obese
0.54 %
35.26 %
16.91 %
47.28 %
30.72 %
1.14 %
40.73 %
17.90 %
40.23 %
Child‘s distribution
Table 5: UK ( HSE )
Child’s BMI status
Full sample
Mother’s
distribution
BMI z-score
<-1.64
-1.64-1.04
1.04-1.64
>1.64
Category
Underweight
Normal
Overweight
Obese
Mother’s
<-1.64
Underweight
11.39 %
62.03 %
8.86 %
17.72 %
0.30 %
BMI status
-1.64-1.04
Normal
2.15 %
59.37 %
14.09 %
24.38 %
48.37 %
1.04-1.64
Overweight
1.30 %
55.67 %
14.68 %
28.34 %
20.01 %
>1.64
Obese
1.09 %
46.63 %
16.45 %
35.83%
31.31 %
1.68 %
54.65 %
14.93 %
28.74 %
Child‘s distribution
21
Table 6 US (NHANES 3)
Child’s BMI status
Full sample
Mother’s
distribution
BMI z-score
<-1.64 ( 891)
-1.64-1.04
1.04-1.64
>1.64
Category
Underweight
Normal
Overweight
Obese
Mother’s
<-1.64
Underweight
10.22%
76.7%
10.22%
2.84%
0.78%
BMI status
-1.64-1.04
Normal
3.90%
80.30%
9.49%
6.32%
75.4 %
1.04-1.64
Overweight
2.54%
71.58%
14.54%
11.34%
13.23 %
>1.64
Obese
2.17%
66.49%
15.88%
15.46%
10.59 %
3.56 %
77.33%
10.90%
8.20%
Child‘s distribution
22
Table 2~6 describe the distribution of child‘s BMI status conditional on mother‘s BMI status
in China, Indonesia, UK and US, respectively. This is similar to transition matrices across the
discretized bivariate distribution, the classes are defined underlying the transition matrix in
terms of quantile groups of the marginal distribution. The interaction terms between mother
and child of different BMI status provides the matrix of intergenerational transition
probabilities. For instance, in Table 2, the numbers in the first row of this matrix indicate that
of the total numbers of children whose mothers were ―underweight‖, 25.45 percent were
―underweight‖, 65.45% were ―normal‖, 2.4% were ―overweight, and 6.7% were ―obese‖.
Similarly for the second and third row. It suggests that there is a high degree of persistence in
the same BMI status across generations.
Thus, in China, the intergenerational ―underweight‖ transmission is higher than that of other
categories ; The intergenerational ―underweight‖ transmission is even higher in Indonesia,
but Indonesian mothers tend to be more overweight; The intergenerational transmission of
higher BMI status ( overweight and obesity) is higher than other categories in the UK, and
the BMI status tends to move up to a higher level in the offspring; These are even greater in
the US.
Overall, the distribution of BMI status transition reveals a wide disparity in the strength of
intergenerational transmission across different BMI status and across different countries. In
underdeveloped economies, such as China and Indonesia, the intergenerational transition
tends to be stronger at the lower levels of BMI status such as ―underweight‖; In contrast, in
developed economies, such as the UK and the US, the transition tends to shift towards
―normal‖ and higher levels of BMI status, overweight and obesity.
5.3.2 Markov Chain
As we see, Table 2~6 display the matrices of BMI status transition probabilities across
generations in different countries. Based on these matrices, we build up a Markov chain to
predict the future changes in the distribution of BMI status.
Using the current child‘s BMI distribution as the initial distribution and based on these
probabilities transition matrices, the BMI distribution in three generations is obtained by
multiplying them in the following way:
Initial distribution : a=
0.1070 0.7480 0.0710 0.0810
23
Transition matrix: b=
0.2550 0.6550 0.0240 0.0670
0.1130 0.7450 0.0660 0.0760
0.0630 0.7500 0.0940 0.0930
0.0660 0.7070 0.1090 0.1180
a*b*b*b = 0.1243 0.7365 0.0666 0.0799
When we multiply the initial distribution by multiple powers of transition matrices, we find
that ―the rows even of the third power of the basic matrix of transition probabilities are
already getting close to one another‖. When we raise the matrix to higher powers, we find
that the Markov chain process of this transition matrix inevitably moves towards an
equilibrium situation, where the constant probability vector would be
a*b*b*b *b*b= 0.1244 0.7367 0.0666 0.0799 ( the 5th higher power)
This vector shows what the proportions in different BMI status tend to be---and to remain--when the process has reached equilibrium. We find this vector ( the future distribution ) does
not go further from the current distribution ( the child‘s BMI distribution in the CHNS
sample), this indicates that the transition process may have already been following these
matrices for a while, so that the current distribution has already reached somewhere near the
equilibrium. In other words, according to the Markov Chain, the BMI distribution in China
may remain as the current state in five generations if there is not a ―shock‖ ( by which I mean
a dramatic environmental change, such as the food innovation) which might fundamentally
change the transition matrices, or if there are no unobserved status-specific effects.
Likewise, we predict the future distribution in five generations for other countries. The results
are displayed in Table 7, along with the initial distribution ( child‘s BMI distribution ) as a
comparison5. We can see that as in the CHNS sample, in other countries the distribution in
five generations is also quite close to the initial distribution. However, the process of iteration
varies across countries. The distribution in China and UK converges after three generations,
whereas those in Indonesia and the US does not show a clear tendency of convergence.
5
The reason why we choose five iteration is that through the iteration of matrix, it seems that five is the optimal
number of iterations through which we can see whether the resultant vector move towards convergence or not.
24
Overall, based on the Markov Chain, the future distribution does not vary much from the
initial distribution. If this proximation of future distribution to the current distribution can be
viewed as equilibrium, this may indicate that the current distribution has reached equilibrium
in almost all the countries. However, whether this process converges and how long it takes to
converge varies across countries. And more importantly, our results are based on the strong
assumptions of Markov Chain.
Table 7: The initial distribution ( child‘s BMI distribution ) and final distribution after five
generations
Initial
Final
Initial
Final
Initial
Underweight
0.107
0.124
0.167
0.196
0.036
Normal
0.748
0.737
0.719
0.693
0.773
Overweight
0.071
0.067
0.046
0.039
0.011
Obese
0.081
0.080
0.069
0.063
0.082
HSE (pooled)
Final
Initial
Final
Initial
Final
Initial
0.035
0.040
0.038
0.011
0.010
0.017
0.706
0.801
0.803
0.407
0.396
0.550
0.095
0.100
0.105
0.179
0.177
0.150
0.067
0.060
0.060
0.402
0.417
0.290
HSE (1995)
Final
Initial
0.019
0.015
0.557
0.550
0.148
0.140
0.282
0.290
HSE (2010)
Final
Initial
0.014
0.020
0.548
0.530
0.140
0.140
0.291
0.300
Final
0.040
0.529
0.141
0.281
China
Indonesia
UK (BCS
cohorts)
Spain
US
Table 8: HSE 1995 after one generation, compared with HSE (2010)
HSE (1995)
Initial
After one
generation
Compare with
HSE (2010)
Underweight
0.015
Normal
0.550
Overweight
0.140
Obese
0.290
0.014
0.550
0.140
0.291
0.020
0.530
0.140
0.300
25
6. Conclusions and Policy Implications
This paper has examined the intergenerational transmission of BMI or adiposity across
generations in four countries across the world. Using the BMI as a measure of adiposity, we
find that the intergenerational transmission of adiposity is relatively high and very
comparable across time and countries – even if these countries are at a very different stage of
economic development. . This suggests that the intergenerational transmission mechanism
which naturally occurs in biology via the transmission of both parental genetic inheritance
and also the shared environment of the family when the child is growing up determine a
significant fraction of the child‘s likely BMI makeup as an adult. At the mean of the
distribution we find that the father and mother each separately account for around 20% of the
child‘s BMI. Since this effect is linear and additively separable for these two parents then we
have found that the joint effect of the family and its associated genetic makeup accounts for
around 40% of the child‘s likely BMI.
Our second key finding is that this intergenerational transmission mechanism is very different
across the distribution of children‘s BMI. Most specifically it is up to double for the fattest
children what it is for the thinnest children. This has enormous consequences for the health
of the world‘s children. Specifically we find that over 30% of the fattest child‘s BMI is
determined by the mother and 30% by the father. Hence jointly they account for up to 60%
of the child‘s likely BMI. In contrast this corresponding fraction is only around 30% for the
thinnest child.
To sum up, our evidence from different countries‘ data suggests that there is a strong
consistency in the IBE estimates across countries and ethnicities, this consistency is different
from what the previous studies find on IIE and IEE. The literature on the transmission of
intergenerational elasticity has found that there is a substantial disparity in the IIE and IEE
estimates across different countries and different datasets. Ranging from as little as .1 to as
much as .6 when they consider only the relationship of income of the child with the income
of a single parent.
This first important implication of our research is that it puts the emphasis firmly on the
family in terms of understand the huge fraction of adiposity determination. Specifically we
need to look no further than the simply biological process of genetic inheritance from parents
to child and what happens to the child when they are very young to explain a huge fraction of
26
what they become as fat or thin adults. We have no way (with the data available to us) of
splitting up the IBE into that which is due to genetic inheritance and that which is due to the
family environment – but what we do know is that jointly these two influences determine a
sizeable faction of what can happen to children. One way of thinking about this process is to
suggest that – in the extreme – the thinnest child in the data – who is clearly not fat at all still
inherits 35% of their BMI from their parents – so that this is the lowest bound on how much
may be due to the biological process of inheritance. Some fraction of the difference between
their inheritance and that of the fat child with a .6 elasticity may still be due to biology but it
seems likely that this could be more to do with what goes on inside the family – namely how
much exercise is taken, what the family diet is like; whether they use a car for transport, how
much TV is watched and generally how active they are.
The corollary of our findings is that logically – certainly for fat children much of the damage
is done a the beginning of their lives. Most children are doomed to have the BMI which is
pre-programmed into them by the genes they inherit and the lifestyle their family lead when
they are a child. This means that the is a strictly limited amount any public intervention can
do to promote health later in life. Much of the damage will have already been done.
27
Appendix (Data description)
CHNS data
The Chinese data here uses the longitudinal data from eight waves (1989, 1991, 1993, 1997,
2000, 2004, 2006, and 2009) of the China Health and Nutrition Survey ( CHNS ). Based on
the definition of response rates that those who participated in previous survey rounds
remaining in the current survey ( Popkin, 2010 ), the response rates of this data were 88% at
individual level and 90 % at household level. This data contains detailed information on
health outcomes, demographic, anthropometric measures of all members of the sampled
households, including height and weight. It is noteworthy that these anthropometric
measures are medically measured rather than self-reported which are mostly used in the
literature. In addition, it includes information on economic and non-economic indicators such
as education, household income and labor market outcomes.
Our sample is restricted to children under 18 years old with information ( especially
anthropometric information) on both the biological father and mother. We choose 18 as the
threshold since age 18 is used to distinguish between adult and child in the CHNS physic
exam dataset where the anthropometric information is included. Additionally, children within
this age range normally live with their parents and rely on their parents for nutritional intake
and health care. As a result, this sample includes 14, 082 person-wave observations made up
by 6,045 children with 3,975 fathers and 3,974 mothers. In other words, our sample includes
6,045 sets of father, mother and children.
Indonesian Family Life Survey (IFLS)
The Indonesian Family Life Survey (IFLS) is an on-going longitudinal survey data which started in
1993. The sample used here is drawn from 1993, 2000 and 2007 waves of the survey, it is
representative of 83% of the Indonesian population and contains over 30,000 individuals living in
13 of the 27 provinces in Indonesia. This survey includes a range of health measures for both
parents and children. It is noteworthy that as in CHNS data, the anthropometric outcome in
IFLS survey was also measured by trained nurses rather than self-reported. Additionally, the
IFLS data also includes information on socioeconomic factors such as education and income.
Thus, the IFLS data is similar to CHNS data in terms of the survey design and measure
methods, this similarity improves the comparability of results based on these two datasets.
28
The sample is restricted to those aged from 0 to 14 years old in each wave and have both
parents and household‘s information. It is noteworthy that this is different from the CHNS
data, where the children sample comprises those aged between 0 and 18 years old.
In addition, in the Indonesian Family Life Survey (IFLS), we also consider step/adopted
children as the sample. The adopted or step children account for around 1% of the whole
sample in each wave, for these children, the information on their parents use the step parents‘
‘
rather than biological parents 6.
British 1970 cohort study
The 1970 British Cohort Study is an ongoing follow up study of 17,200 babies born in
England, Scotland, Wales and Northern Ireland between 5 and 11 April 1970 who are still
living in Britain (excluding Northern Ireland). The survey was conducted when the cohorts at
birth, aged 5 ( in 1975 ) , 10 ( in 1980), 16 ( in 1986) , 26 ( in 1996), 30 ( in 1999-2000) , 34
( in 2004-2005) and 38 ( in 2008-2009). The samples at the age 5 and 10 were augmented
since immigrants born in the same week were added in. In this paper we use the cohorts in
the first five waves (sweeps).
At the birth, the questionnaires were completed by midwife and the supplementary
information was collected from clinical records. As the cohorts got older, the approach of
survey changed, parents were interviewed by the health stuff and questionnaires were
completed by teachers. In terms of the anthropometric information, the height and weight
were measured at the age of 10 and self-reported at the age of 26 ( Shaheen, et al., 1999)
Health Survey for England
The Health Survey for England is designated to be nationally representative of people of
different age, gender, geographic region and socio-demographic circumstances. 7 It was
started in 1991 and has been conducted annually since then. The survey combines
6
Comparing step/adopted children BMI with the BMI of both their biological and step parents is beyond the
scope of this dissertation. Several studies using adopted children have been conducted in order to investigate
whether genetic influence or environmental factors that has stronger impact on child‘s obesity (Sørensen 1992;
Maes 1997; and Silventoinen, 2010). The results from these studies support for the importance of genetic factors,
where genetic factors had a strong effect on the variation of body mass index (BMI), whereas environment
indicators only show a weak influence.
7
―The 1991 and 1992 surveys had a limited population sample of about 3,000 and 4,000 adults respectively. For
1993 to 1996 adult sample was boosted to about 16,000 to enable analysis by socio-economic characteristics and
health regions. In 1995 for the first time a sample of about 4,000 children was also introduced. In the 1997
Health Survey the sample was about 7,000 children and 9,000 adults. In 1998 the sample was again about
16,000 adults and 4,000 children. ‖
29
questionnaire-based answers with physical measurements and the analysis of blood sample.
Each year‘s survey has a particular focus on a disease or condition or population group, but
height, weight and general health are covered each year. An interview with household
members is followed by a nurse visit. Thus, there are both self-reported and medicallymeasured height and weight in this data. In the computation of BMI z-score, we use ―htval‖
and ―wtval‖ in the survey which are referred to as the ―valid‖ height and weight.
NHANES
The National Health and Nutrition Examination Survey (NHANES) is a program of studies
designed to assess the health and nutritional status of adults and children in the United States.
Four surveys of this type have been conducted since 1970:
1. 1971-75—National Health and Nutrition Examination Survey I (NHANES I);
2. 1976-80—National Health and Nutrition Examination Survey II (NHANES II);
3. 1982-84—Hispanic Health and Nutrition Examination Survey (HHANES); and
4. 1988-94—National Health and Nutrition Examination Survey (NHANES III) and
5. 1999-present--National Health and Nutrition Examination Survey (Continuous NHANES)
Note in NHANES data, there is only personal identification variable (seqn), there is no
household id on the public release file, the relationship of a participant to the household
reference person is not publicly released8. Thus, we cannot track down the participants‘
parents via father and mother‘s id ( as in CHNS and Indonesian data), or identify the potential
parents via the household id ( as in England data). In other words, there is no way to identify
the parents by ID. However, in one of these surveys---NHANES III9, there is a family
background section in the youth file, where limited characteristics of the parents were
collected, including mother and father‘s height and weight. Even though their age is not
available, the availability of their height and weight still provides an opportunity to compute
their BMI z-score, where their age is assumed 19 years old. This assumption on age is the
same as we compute parents‘ z-score in the previous datasets ( the adults‘ age above 19 are
8
with the exception of dietary data, the relationship of the sample participant to the proxy is not publicly
released, either.
9
Note in NHANES2, there is also mother and father‘s height and weight in the youth file of on the Web of
Tutorial, but most of them are the same, I suspect that they are wrong, so not use them
30
assumed 19 when they run through the WHO macro). Therefore, NHANES III allows us the
estimate the IBE in the United States between 1988 and 1994.
NHANES III, conducted between 1988 and 1994, included about 40,000 people selected
from households in 81 counties across the United States. In NHANES III, black Americans
and Mexican Americans were selected in large proportions , each of these groups comprised
separately 30 percent of the sample. It was the first survey to include infants as young as 2
months of age and to include adults with no upper age limit. Our sample is obtained by
merging the youth data which includes child‘s age and parents‘ height weight with
examination data which includes child‘s final (medically measured) height weight. Any BMI
zscores outside the range of -6 and 6 are viewed as implausible and are dropped from the
sample (Li et al., 2009) . Our final sample includes 6,582 pairs of father, mother and child.
31
References
Abrevaya, J., & Tang, H. (2011). Body mass index in families: spousal correlation,
endogeneity, and intergenerational transmission. Empirical Economics, 41(3), 841864.
Ahlburg, D. (1998). Intergenerational transmission of health. American Economic Review,
88(2), 265-270.
Anderson, P. M., Butcher, K. F., and Levine, P. B. (2003). Maternal employment and
overweight children. Journal of Health Economics, 22(3), 477-504.
Anderson, P. M., Butcher, K. F., and Schanzenbach, D. W. (2007). Childhood disadvantage
and obesity: Is nurture trumping nature? : National Bureau of Economic Research.
Anderson, P. M. (2012). Parental employment, family routines and childhood
obesity. Economics and Human Biology, 10 (4) (2012), pp. 340–351
Anger, S., and Heineck, G. (2010). Do smart parents raise smart children? The
intergenerational
transmission
of
cognitive
abilities. Journal
of
Population
Economics, 23(3), 1105-1132.
Becker, G. S. (1981). A Treatise on the Family. Cambridge, MA: Harvard University Press.
Becker, G. S., and Barro, R. J. (1986). A reformulation of the economic theory of fertility.
National Bureau of Economic Research.
Bell, A. C., Ge, K., and Popkin, B. M. (2012). The road to obesity or the path to prevention:
motorized transportation and obesity in China. Obesity Research,10(4), 277-283.
Behrman, J. R. (1988). Intrahousehold allocation of nutrients in rural India: Are boys favored?
Do parents exhibit inequality aversion? Oxford Economic Papers, 32-54.
Behrman, J. R., Rosenzweig, M. R., and Taubman, P. (1994). Endowments and the allocation
of schooling in the family and in the marriage market: the twins experiment. Journal
of Political Economy, 1131-1174.
Behrman, J. R., and Rosenzweig, M. R. (2002). Does increasing women's schooling raise the
schooling of the next generation?. American Economic Review, 323-334.
Björklund, A., Lindahl, M., and Plug, E. (2006). The origins of intergenerational associations:
Lessons from Swedish adoption data. The Quarterly Journal of Economics, 121(3),
999-1028.
Björklund, A., Roine, J., and Waldenström, D. (2012). Intergenerational top income mobility
in Sweden: Capitalist dynasties in the land of equal opportunity?. Journal of Public
Economics, 96(5), 474-484.
32
Black, S. E., Devereux, P. J., and Salvanes, K. G. (2003). Why the apple doesn't fall far:
Understanding intergenerational transmission of human capital (No. w10066).
National Bureau of Economic Research.
Blanden, J. (2008). IntergenerationalIncome Mobility in a Comparative Perspective.
Education and Inequality Across Europe, Edward Edgar, forthcoming.
Bratsberg, B., K. Røed, O. Raaum, R. Naylor, M. Jäntti, T. Eriksson and E. Österbacka
(2007), Nonlinearities in intergenerational earnings mobility: Consequences
for crosscountry comparisons, Economic Journal, 117(519), C72–C92.
Boudon, R. (1974). Education, Opportunity, and Social Inequality: Changing Prospects in
Western Society.
Björklund, A., and Jäntti, M. (2009). Intergenerational income mobility and the role of family
background. Oxford Handbook of Economic Inequality, Oxford University Press,
Oxford.
Carneiro, P., and Heckman, J. J. (2002). The Evidence on Credit Constraints in PostSecondary Schooling. The Economic Journal, 112(482), 705-734.
Carneiro, P., and Heckman, J. J. (2003). Human capital policy, in Heckman, J. J., and
Krueger, A. B. (2005). Inequality in America: What Role for Human Capital
Policies?, MIT Press Books, 1.
Case, A., Fertig, A., and Paxson, C. (2005). The lasting impact of childhood health and
circumstance. Journal of Health Economics, 24(2), 365-389.
Chevalier, A. (2004). Parental education and child's education: a natural experiment, IZA
Working Paper No. 1153.
Classen, T. J. (2010). Measures of the intergenerational transmission of body mass index
between mothers and their children in the United States, 1981–2004. Economics and
Human Biology, 8(1), 30-43.
Cole, T. J., Faith, M. S., Pietrobelli, A., and Heo, M. (2005). What is the best measure of
adiposity change in growing children: BMI, BMI%, BMI z-score or BMI
centile?. European Journal of Clinical Nutrition, 59(3), 419-425.
Coneus, K., & Spiess, C. K. (2012). The intergenerational transmission of health in early
childhood—Evidence from the German Socio-Economic Panel Study. Economics &
Human Biology, 10(1), 89-97.
Conley, D., & Bennett, N. G. (2000). Is biology destiny? Birth weight and life
chances. American Sociological Review, 458-467.
Corak, M. and A. Heisz (1999), The Intergenerational Earnings and Income Mobility of
33
Canadian Men: Evidence from Longitudinal Income Tax Data, Journal of Human
Resources, 34/3: 504-33.
Cremer, H., and Pestieau, P. (1991). Bequests, filial attention and fertility. Economica, 359375.
Currie, J., and Moretti, E. (2003). Mother's education and the intergenerational transmission
of human capital: Evidence from college openings. Quarterly Journal of Economics,
118(4), 1495-1532.
Currie, J., and Moretti, E. (2005). Biology as destiny? Short and long-run determinants of
intergenerational transmission of birth weight: National Bureau of Economic
Research Cambridge, Mass., USA.
Dasgupta, Partha. (1993). An Inquiry into Well-Being and Destitution. New York: Oxford
University Press Inc.
Dearden, L., Machin, S., and Reed, H. (1997). Intergenerational mobility in Britain. The
Economic Journal, 47-66.
Desai, S., and Alva, S. (1998). Maternal education and child health: Is there a strong causal
relationship?. Demography, 35(1), 71-81.
Du, S., Lu, B., Zhai, F., and Popkin, B. M. (2002). A new stage of the nutrition transition
in China. Public Health Nutrition, 5(1A), 169-174
Emanuel, I., Filakti, H., Alberman, E., & Evans, S. J. (1992). Intergenerational studies
human birthweight from the 1958 birth cohort. 1. Evidence for a multigenerational
effect. BJOG: An International Journal of Obstetrics & Gynaecology, 99(1), 67-74.
Fang, P., Dong, S., Xiao, J., Liu, C., Feng, X., and Wang, Y. (2010). Regional inequality in
health and its determinants: evidence from China. Health Policy,94(1), 14.
Galton, F.(1869). Hereditary Genius: An Inquiry into Its Laws and Consequences.London:
Macmillan.
Galton, F.(1877). Typical Laws of Heredity. Proc. Royal Inst. Great Britain 8 (February
1877): 282–301.
Goldberger, A. S. (1989). Economic and Mechanical Models of Intergenerational
Transmission. American Economic Review, 79(3), 504-513.
Gong, H., Leigh, A., and Meng, X. (2012). Intergenerational income mobility in urban
China. Review of Income and Wealth, 58 (3), 481-503
Goode, A., Mavromaras, K., and Smith, M. (2008). Intergenerational transmission of healthy
eating behaviour and the role of household income. Available at SSRN 1145914.
34
Gorber, S. C., Tremblay, M., Moher, D., and Gorber, B. (2007). A comparison of direct vs.
self-report measures for assessing height, weight and body mass index: a systematic
review. Obesity reviews, 8(4), 307-326.
Grawe, N. D. (2004). Intergenerational mobility for whom? The experience of high-and
low-earning sons in international perspective. Generational income mobility in
North America and Europe, 58-89.
Gruber, J. (2009). The Problems of Disadvantaged Youth: An Economic Perspective:
University of Chicago Press.
Guo, X., Mroz, T. A., Popkin, B. M., and Zhai, F. (2000). Structural change in the impact of
income on food consumption in China, 1989–1993. Economic Development and
Cultural Change, 48(4), 737-760.
Han, S., and Mulligan, C. B. (2001). Human capital, heterogeneity and estimated degrees of
intergenerational mobility. The Economic Journal, 111(470), 207-243
Kalmijn, M. (1994). Assortative mating by cultural and economic occupational
status. American Journal of Sociology, 422-452.
Karlin, S. (1983). Association arrays in assessing forms of dependencies between bivariate
random variables. Proceedings of the National Academy of Sciences, 80(2), 647-651.
Kanbur, R., and Zhang, X. (2005). Fifty years of regional inequality in China: a journey
through
central
planning,
reform,
and
openness. Review
of
Development
Economics, 9(1), 87-106.
Kanbur, R., and Zhang, X. (1999). Which regional inequality? The evolution of rural–urban
and inland–coastal inequality in China from 1983 to 1995. Journal of Comparative
Economics, 27(4), 686-701.
Klein, L. R., and Ozmucur, S. (2003). The estimation of China's economic growth
rate. Journal of Economic and Social Measurement, 28(4), 187-202.
Laitinen, J., Power, C., and Järvelin, M. R. (2001). Family social class, maternal body mass
index, childhood body mass index, and age at menarche as predictors of adult obesity.
American Journal of Clinical Nutrition, 74(3), 287-294.
Lee, R. D., and Miller, T. (1991). Population growth, externalities to childbearing, and
fertility policy in developing countries. Proceedings of the World Bank Annual
Conference on Development Economics 1990. 275-304.
Lefgren, L., Sims, D., and Lindquist, M. J. (2012). Rich Dad, Smart Dad: Decomposing the
Intergenerational Transmission of Income. Journal of Political Economy, 120(2), 268303.
35
Lehmann, E. L. (1966). Some concepts of dependence. The Annals of Mathematical
Statistics, 1137-1153.
Li, L., Law, C., Lo Conte, R., and Power, C. (2009). Intergenerational influences on
childhood body mass index: the effect of parental body mass index trajectories.
American Journal of Clinical Nutrition, 89(2), 551-557.
Linabery, A. M., Nahhas, R. W., Johnson, W., Choh, A. C., Towne, B., Odegaard, A. O., ...
and Demerath, E. W. (2012). Stronger influence of maternal than paternal obesity on
infant and early childhood body mass index: the Fels Longitudinal Study. Pediatric
Obesity.
Martin, M. A. (2008). The intergenerational correlation in weight: how genetic resemblance
reveals the social role of families. American Journal of Sociology, 114(Suppl), S67.
Mare, R. D. (1991). Five decades of educational assortative mating. American Sociological
Review, 15-32.
Manser, M., and Brown, M. (1980). Marriage and household decision-making: A bargaining
analysis. International Economic Review, 31-44.
Monteiro, C. A., Moura, E. C., Conde, W. L., and Popkin, B. M. (2004). Socioeconomic
status and obesity in adult populations of developing countries: a review. Bulletin of
the World Health Organization, 82(12), 940-946.
Morgan, S. L. (2000). Richer and taller: stature and living standards in China, 19791995. The China Journal, (44), 1-39.
Mulligan, C. B. (1999). Galton versus the human capital approach to inheritance. Journal of
Political Economy, 107(6 PART 2), S184-S224.
Onis, M. D., Onyango, A. W., Borghi, E., Siyam, A., Nishida, C., and Siekmann, J. (2007).
Development of a WHO growth reference for school-aged children and adolescents.
Bulletin of the World Health Organization, 85(9), 660-667.
Oreopoulos, P., Page, M. E., and Stevens, A. H. (2006). The intergenerational effects of
compulsory schooling. Journal of Labor Economics, 24(4), 729-760.
Patterson, S. (04/08/2011). Obesity in China: Waistlines are Expanding Twice as Fast as
GDP Retrieved 28/11, 2012, from
http://uschina.usc.edu/w_usci/showarticle.aspx?articleID=16595andAspxAutoDetectCookieS
upport=1
Pekkarinen, T., Uusitalo, R., and Kerr, S. (2009). School tracking and intergenerational
income mobility: Evidence from the Finnish comprehensive school reform. Journal of
Public Economics, 93(7), 965-973.
36
Piraino, P. (2007). Comparable estimates of intergenerational income mobility in Italy. The
BE Journal of Economic Analysis and Policy, 7(2).
Pitt, M. M., Rosenzweig, M. R., and Hassan, M. N. (1990). Productivity, health, and
inequality in the intrahousehold distribution of food in low-income countries. The
American Economic Review, 1139-1156.
Pollak, R. A. (2005). Bargaining power in marriage: Earnings, wage rates and household
production (No. w11239). National Bureau of Economic Research.
Plomin, R., DeFries, J. C., and Loehlin, J. C. (1977). Genotype-environment interaction and
correlation in the analysis of human behavior. Psychological Bulletin, 84(2), 309.
Popkin, B. M. (2001). The nutrition transition and obesity in the developing world. The
Journal of Nutrition, 131(3), 871S-873S.
Popkin, B. M. (2008). The nutrition transition: an overview of world patterns of
change. Nutrition Reviews, 62(s2), S140-S143.
Popkin, B. M., Du, S., Zhai, F., and Zhang, B. (2010). Cohort Profile: The China Health and
Nutrition Survey—monitoring and understanding socio-economic and health change
in China, 1989–2011. International Journal of Epidemiology, 39(6), 1435-1440.
Qian, N. (2008). Missing women and the price of tea in China: The effect of sex-specific
earnings on sex imbalance. The Quarterly Journal of Economics,123(3), 1251-1285.
The National Bureau of Statistics of China (2001). Statistics Yearbook 2000. Beijing: China
Statistics Press
Rosenzweig, M. R., and Schultz, T. P. (1983). Estimating a household production function:
Heterogeneity, the demand for health inputs, and their effects on birth weight. The
Journal of Political Economy, 723-746.
Royer, H. (2009). Separated at birth: US twin estimates of the effects of birth
weight. American Economic Journal: Applied Economics, 49-85.
Sen, A. (1990). More than 100 million women are missing. The New York Review of
Books, 37.
Scarr, S., and McCartney, K. (1983). How people make their own environments: A theory of
genotype→ environment effects. Child Development, 424-435.
Scholder, S. V. H. K., Smith, G. D., Lawlor, D. A., Propper, C., and Windmeijer, F. (2012).
The effect of fat mass on educational attainment: Examining the sensitivity to
different identification strategies. Economics and Human Biology, 10 (4),405-418.
37
Sherry, B., Jefferds, M. E., and Grummer-Strawn, L. M. (2007). Accuracy of adolescent selfreport of height and weight in assessing overweight status: a literature
review. Archives of Pediatrics and Adolescent Medicine, 161(12), 1154.
Solon, G. 1999. Intergenerational Mobility in the Labor Market, in Handbook of Labor
Economics, Volume 3A. Orley Ashenfelter and David Card, eds. Amsterdam:
North-Holland, pp. 1761-800.
Solon G. (2004). A model of intergenerational mobility variation over time and space. In
Generational Income Mobility in North America and Europe, ed. M Corak, pp. 38-47.
Cambridge, UK: Cambridge Univ. Press
Spencer, E. A., Appleby, P. N., Davey, G. K., and Key, T. J. (2002). Validity of self-reported
height and weight in 4808 EPIC–Oxford participants. Public Health Nutrition, 5(04),
561-565.
Stiglitz, J. E. (1969). Distribution of income and wealth among individuals.Econometrica:
Journal of the Econometric Society, 382-397.
Trannoy, A., Tubeuf, S., Jusot, F., & Devaux, M. (2010). Inequality of opportunities in health
in France: a first pass. Health Economics, 19(8), 921-938.
Thomas, D. (1990). Intra-household resource allocation: An inferential approach. Journal
of Human Resources, 635-664.
Thomas, D., Strauss, J., and Henriques, M. H. (1991). How does mother's education
affect child height?. Journal of Human Resources, 183-211.
Thompson, O. The Intergenerational Transmission of Health Status: Estimates
and Mechanisms.
Van Leeuwen, M., Van Den Berg, S. M., and Boomsma, D. I. (2008). A twin-family study of
general IQ. Learning and Individual Differences, 18(1), 76-88.
Wang, Y., and Chen, H. J. (2012). Use of Percentiles and Z-Scores in Anthropometry.
Handbook of Anthropometry: Physical Measures of Human Form in Health and
Disease, 29.
World Health Organization. (2000). The Asia-Pacific perspective: redefining obesity and its
treatment. Geneva, Switzerland: World Health Organization.
WHO, E. C. (2004). Appropriate body-mass index for Asian populations and its implications
for policy and intervention strategies. Lancet, 363(9403), 157.
Wu, Y. (2006). Overweight and obesity in China. British Medical Journal, 333(7564), 362363
38
Yang, D. T. (1999). Urban-biased policies and rising income inequality in China. The
American Economic Review, 89(2), 306-310.
Zhang, X., and Kanbur, R. (2005). Spatial inequality in education and health care in
China. China Economic Review, 16(2), 189-204.
39