ArticlePDF Available

Unequal before death: The effect of paternal education on children's old-age mortality in the United States

March 2024
Population Studies 78(2):1-27

DOI:10.1080/00324728.2023.2284766

Authors:

Hamid Noghanibehambari

Austin Peay State University

Content uploaded by Hamid Noghanibehambari

Content may be subject to copyright.

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rpst20

Population Studies

A Journal of Demography

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/rpst20

Unequal before death: The effect of paternal

education on children’s old-age mortality in the

United States

Hamid Noghanibehambari & Jason Fletcher

To cite this article: Hamid Noghanibehambari & Jason Fletcher (06 Mar 2024): Unequal before

death: The effect of paternal education on children’s old-age mortality in the United States,

Population Studies, DOI: 10.1080/00324728.2023.2284766

To link to this article: https://doi.org/10.1080/00324728.2023.2284766

View supplementary material

Published online: 06 Mar 2024.

Submit your article to this journal

Article views: 82

View related articles

View Crossmark data

Reasearch Article

Unequal before death: The effect of paternal education

on children’s old-age mortality in the United States

Hamid Noghanibehambari and Jason Fletcher

University of Wisconsin–Madison

A growing body of research documents the relevance of parental education as a marker of family socio-

economic status for children’s later-life health outcomes. A strand of this literature evaluates how the

early-life environment shapes mortality outcomes during infancy and childhood. However, the evidence

on mortality during the life course and old age is limited. This paper contributes to the literature by

analysing the association between paternal education and children’s old-age mortality. We use data from

Social Security Administration death records over the years 1988–2005 linked to the United States 1940

Census. Applying a family(cousin)- fixed-effects model to account for shared environment, childhood

exposures, and common endowments that may confound the long-term links, we find that having a father

with a college or high-school education, compared with elementary/no education, is associated with a 4.6-

or 2.6-month-higher age at death, respectively, for the child, conditional on them surviving to age 47.

Supplementary material for this article is available at: http://dx.doi.org/10.1080/00324728.2023.2284766

Keywords: education; family fixed effects; mortality; life expectancy; historical data; social benefits;

externality

[Submitted April 2022; Final version accepted June 2023]

Introduction

Health disparities in old age partially reflect socio-

economic conditions in early life (Hayward and

Gorman 2004; Engelman et al. 2010; Elo et al.

2014; Lazuka 2019). Parental education, as a

major component and determinant of socio-

economic conditions, may play an important part

in the later-life health and mortality outcomes of

their children. Several studies suggest that parental

education and family socio-economic status (SES)

shape early childhood mortality (Hatt and Waters

2006; Gakidou et al. 2010; Gage et al. 2013; Balaj

et al. 2021). However, the evidence for mortality

through the life course is less established, despite

the potential importance and policy relevance of

a long-term association. In a recent study,

Huebener (2019) investigates the effects of par-

ental education on children’s life expectancy in

Germany. The results from survival analysis show

that higher parental education is correlated with

higher survival rates for children’s mortality, con-

ditional on survival up to age 65. Also, the

effects of father’s education are smaller than

those of mother’s education and are insignificant

in most cases. Lee and Ryff (2019) use Midlife in

the United States data and explore how a range

of early-life adversities affect later-life mortality.

They find evidence that early-life SES is associated

with adult mortality risk. In a similar study, Montez

and Hayward (2011) use the Health and Retire-

ment Study to explore the effects of early-life con-

ditions on old-age mortality. They find that non-

Hispanic white people and those whose fathers

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Population Studies, 2024

https://doi.org/10.1080/00324728.2023.2284766

are less than high-school educated display higher

risks of mortality.

The current study extends this line of research by

establishing an association between paternal edu-

cation and children’s longevity in the United States

(US). We construct a longitudinal sample based on

the full-count 1940 Census linked with Social Secur-

ity Administration (SSA) death records for the years

1988–2005. The results of family-fixed-effects

models, which compare the longevity of cousins

within a family tree, suggest that higher paternal

education significantly increases children’s age at

death and that the cross-sectional ordinary least

squares (OLS) estimates do not generate a notice-

able bias in the results. The preferred specification,

which includes a wide array of family controls and

fixed effects, in addition to father’s family fixed

effects, implies that fathers with a college education

have children with 4.6 months of additional lifespan

compared with low-educated fathers. Similarly, the

ages at death of children whose fathers are high-

school or middle-school educated are 2.6 or 1.8

months higher, respectively, than for those with a

father with elementary/no education.

In addition, we find substantial heterogeneity in

the effects. The effects are more pronounced

among Black people, males, those in families with

lower SES, and those residing in the South census

region. Moreover, we find complementarity

between father’s and mother’s schooling: the

effects of father’s education are larger when the

mother is more educated. Further analyses show

that father’s education affects the material resources

of a family. Highly educated fathers are more likely

to be in the labour force, have higher income, have

more children, own a house, and work in occupations

with higher income scores. Also, families engage in

assortative matching in the marriage market and

more highly educated fathers are also more likely

to be matched with more highly educated mothers.

While the data offer these channels as the mediating

paths between father’s education and children’s

longevity, we also discuss alternative pathways

suggested by the literature.

This study contributes to the growing literature on

education and mortality in three ways. First, we assess

endogeneity issues by implementing a family-fixed-

effects strategy, thus extending the literature. While

some studies take advantage of compulsory schooling

as the shock to parental education in order to explore

children’s life-cycle outcomes, hardly any studies

apply a family-fixed-effects model to explore this

long-term link specifically in the US context (see

e.g. Currie and Moretti 2003; Oreopoulos et al.

2006; Lindeboom et al. 2009; Chou et al. 2010;

Chalfin and Deza 2018,2019; Hamad et al. 2018;

Cui et al. 2019; Noghanibehambari et al. 2022).

Second, while previous studies exploit data with

small sample sizes and low power to explore parental

education on children’s health and longevity, the con-

structed data set in the current study offers unprece-

dented and unparalleled sample sizes to explore this

long-term link using US data (Montez and Hayward

2011; Huebener 2019,2020,2021; Lee and Ryff

2019). Third, this paper contributes to the growing lit-

erature on non-monetary returns to education and

intergenerational effects of human capital (Currie

and Moretti 2003; Silles 2011,2017;Suhonen and Kar-

hunen 2019). This aspect of the current study has

important policy implications. It adds to understand-

ing of the usually unobserved and long-term benefits

of intervention policies that aim to improve edu-

cational outcomes.

The rest of the paper is organized as follows. We

first provide a brief review of the literature and then

discuss the process of constructing the final sample

and the sample selection criteria. Next we introduce

the econometric framework and underlying assump-

tions in the models. The next two sections go over

the results and suggest several mechanism channels.

We finish with some concluding remarks.

Literature review

There are several ways through which paternal edu-

cation may affect children’s old-age mortality. We

categorize these mediating channels into three

general pathways: (1) improvements in children’s

own education; (2) improvements in available

material resources during childhood; (3) the associ-

ation between father’s and mother’s education

through assortative matching. In this section, we

discuss each channel briefly.

First, individuals with higher education also tend

to have more highly educated parents. Several

studies establish a causal link between parental

schooling and children’s educational attainments

(Behrman and Rosenzweig 2002; Daouli et al.

2010; Pronzato 2012; Zeng and Xie 2014; Dickson

et al. 2016; Zhou and Dasgupta 2017; Lundborg

et al. 2018; Agüero and Ramachandran 2020). Edu-

cation, in turn, may affect health and mortality in

several ways; for example by providing better

health-related knowledge, change in social peers,

improved resources through changes in lifetime

earnings, and safer occupations with better health

insurance (Acemoglu and Angrist 2000; Card 2001;

2Hamid Noghanibehambari and Jason Fletcher

Grimard and Parent 2007; Conti et al. 2010; Goldin

and Katz 2010; Cutler et al. 2015; Fletcher 2015;

Fletcher et al. 2021; Lleras-Muney 2022; Lleras-

Muney et al. 2022; Noghanibehambari 2022a).

Higher lifetime earnings can translate into a health-

ier environment and better health-related resources

to improve health status and increase life expectancy

(Lindahl 2005; Snyder and Evans 2006; Gonzalez

and Quast 2015; Fitzpatrick and Moore 2018;

Lefèbvre et al. 2019).

In addition, some studies also establish a link

between education and health-related behaviour,

such as drinking, smoking, and obesity-related

habits (Fletcher and Frisvold 2009; Tenn et al.

2010; Fletcher and Frisvold 2011; Cawley et al.

2013; Cohen et al. 2013; Maralani 2013; Koning

et al. 2015;Kim2016; Barcellos et al. 2018). Drink-

ing, smoking, and obesity are also linked to mortality

outcomes (Pampel 2005; Leon et al. 2007; Cecchini

et al. 2010; Ahima and Lazar 2013; Carter et al.

2015). Several studies explore the causal effect of

education on health, longevity, and mortality regard-

less of the pathway. While some studies find a pro-

tective effect of education against mortality

(Lleras-Muney 2005; van Kippersluis et al. 2011;

Fischer et al. 2013; Gathmann et al. 2015; Buckles

et al. 2016; Davies et al. 2016; Galama et al. 2018;

Halpern-Manners et al. 2020), others reach a null

result and fail to detect a significant impact (Mazum-

der 2008; Lindeboom et al. 2009; Cutler and Lleras-

Muney 2012;Clark and Royer 2013; Leuven et al.

2016; Meghir et al. 2018; Barcellos et al. 2019; Albar-

rán et al. 2020). For instance, Halpern-Manners et al.

(2020) use linked US 1940 Census and SSA death

records and implement a twin fixed-effects strategy.

They show that education significantly increases

longevity; however, their results suggest that

accounting for common endowments through a

twin fixed-effects strategy causes the coefficients to

drop slightly. They find that an additional year of

schooling is associated with a 4.2-month-higher age

at death. Lleras-Muney (2005) explores the effect

of own education on mortality using changes in com-

pulsory schooling laws and child labour laws during

the years 1915–39 as the instrument for education.

She constructs synthetic cohorts using US decennial

censuses to compute 10-year mortality rates. She

finds that an additional year of education reduces

the 10-year mortality rate by approximately 3.6 per-

centage points. However, other studies that

implement changes in compulsory schooling as the

shock to education find different results (Mazumder

2008; Fletcher 2015). For instance, Black et al. (2015)

refine Lleras-Muney’s(2005) estimations by

providing more precise measures of mortality using

vital statistics data and find that including cohort

fixed effects absorbs all the effects on mortality.

Fletcher (2015) implements the same strategy on a

relatively large sample of individual data and finds

significant evidence for the impacts of own edu-

cation on a wide range of health conditions and

self-reported health outcomes. However, the effect

of education on mortality is not precisely estimated,

although the magnitude of the effect is large.

Second, parental education can change the avail-

able material resources that matter for prenatal or

postnatal development, leaving an initial health

endowment by improving birth outcomes and nur-

turing child growth during early life (Almond and

Currie 2011; Almond et al. 2018). Education-

induced increases in parental income can affect

child health through increases in consumption of

health-related inputs (e.g. better prenatal care

during pregnancy, better health insurance, and

better healthcare utilization during childhood) and/

or other non-health-related consumption that

affects child development (e.g. better food, cleaner

neighbourhood of residence, and better health

environment). These improvements influence

health at birth and at postnatal ages, which may

change the trajectory of life-cycle outcomes

(Almond et al. 2011; Hoynes et al. 2011; Hoynes

et al. 2015; Hoynes et al. 2016; Rosales-Rueda

2018; Noghanibehambari and Salari 2020). For

example, Almond et al. (2011) show that the intro-

duction of the Food Stamp Program (an anti-

poverty programme that provided resources for the

poor in the US) was associated with improved

birth outcomes. Hoynes et al. (2016) complement

this analysis, showing that those individuals who

benefited from the programme during their child-

hood displayed improved health outcomes during

adulthood.

Third, in an assortative matching marriage market,

paternal education is also reflected in maternal edu-

cation, and the latter, in turn, affects child health out-

comes as a complement to the effects of paternal

education (Currie and Moretti 2003; Chen and Li

2009; Ali and Elsayed 2018; Chang 2018;Keats

2018; Shen 2018; Noghanibehambari et al. 2022).

Maternal education can also influence the initial

health endowment of offspring through health be-

haviour channels such as smoking and drinking,

both of which are documented to be associated

with negative infant health outcomes (Yan 2014;

Barreca and Page 2015). Several studies establish

the long-term link between health at birth and in

infancy and childhood on health and mortality at

Paternal education and children’s old-age mortality 3

older ages (Behrman and Rosenzweig 2004; Black

et al. 2007; Royer 2009; Wherry et al. 2018; Mar-

uyama and Heinesen 2020; Goodman-Bacon 2021).

Several studies examine the impacts of parental

education on the life-cycle outcomes of their chil-

dren. For instance, Carneiro et al. (2013) use

matched data from female respondents in the

National Longitudinal Survey of Youth (1979

cohort) to explore the effect of maternal education

on children’s outcomes. They find considerable

returns to mother’s education in terms of test

scores and Behaviour Problems Index. Nevertheless,

they do not find an effect on children’s overweight

measures, although the coefficients are negative.

Chou et al. (2010) exploit the sharp changes in com-

pulsory schooling laws in Taiwan accompanied by

the construction of a series of new high schools to

assess the effect of parental education on infant

health and mortality. They find significant reductions

in low-birthweight infants and infant mortality. The

effects are quite similar for both maternal and

paternal education. Of particular relevance to the

current study, Huebener (2019) investigates the

effects of parental education on children’s life

expectancy in Germany. The survival analysis

results show that higher parental education is associ-

ated with lower mortality for children conditional on

survival up to age 65. Also, the effects of father’s

education are smaller than those of mother’s edu-

cation and are insignificant in most cases. Lundborg

et al. (2014) explore the education–health associ-

ation using a compulsory schooling reform in

Sweden and find that increasing mother’s education

has a positive effect on sons’height, health, physical

capacity, and cognitive and non-cognitive abilities.

The results of other similar studies are inconclusive

(Caldwell and McDonald 1982; Thomas et al. 1990,

1991; Breierova and Duflo 2004; Alderman et al.

2006; Lindeboom et al. 2009; Gakidou et al. 2010;

McCrary and Royer 2011). For instance, Lindeboom

et al. (2009) show that the UK schooling reform,

which raised the minimum school leaving age by

one year in 1947, did not have a significant impact

on the health of children of affected cohorts.

Data and sample construction

The primary source of data is the ‘Numident’

(Numerical Identification System) files of the SSA

death records (1988–2005) linked to the full-count

1940 Census in the US. The Numident data are pro-

vided by the CenSoc project (Goldstein et al. 2021).

The linkage technique employs the ‘ABE fully

automated approach’, which is based on first name,

last name, and age, to identify individuals (Abra-

mitzky et al. 2012,2014,2019). We should note that

CenSoc uses exact name to match and later evaluates

the robustness of this method compared with three

other methods (using raw names, New York State

Identification and Intelligence System (NYSIIS)

standardization, and the Jaro–Winkler distance

method) to link individuals across data sets (Ferrie

1996; Abramitzky et al. 2012,2019; Abramitzky

et al. 2021; Breen and Osborne 2022).

Data from the 1940 Census and other historical

full-count censuses between 1900 and 1930 are

extracted from Ruggles et al. (2020). Since women

often change their last name after marriage, we are

unable to carry out the merging for mothers when

we explore historical censuses to search for the

family tree. Therefore, our primary variable of inter-

est is father’s education, and we impose sample

restrictions based partly on father’s characteristics.

We use the information on dates of birth and

death to construct our measure of longevity: age at

death. We restrict the sample to cohorts born

1923–40, as children usually leave the household by

age 18 and would thus not be recorded in the par-

ental household in the 1940 Census. We also limit

the sample to those whose fathers are aged 25–55

in 1940. The Numident-census-linked data contain

a wide array of parental information, including par-

ental education, which can be used to estimate

long-run effects. However, we are concerned that

even after controlling for a full battery of character-

istics and fixed effects, such correlations may fail to

capture the full effects of family environment, neigh-

bourhood influences, and local health environment,

as well as children’s initial abilities and genetic

endowments. For instance, the local availability of

schools and colleges may influence paternal edu-

cation and also children’s health. Moreover, these

institutes are also located in places with different

characteristics compared with places with fewer

schooling options; these local factors may result in

better health accumulation during childhood and

also affect old-age longevity. In addition, more able

fathers may acquire more schooling and also have

more able children, who live longer due to factors

that operate through genetic mechanisms.

To control for a set of these potential confounders,

we rely on variation in educational attainment

between fathers who are siblings in assessing the out-

comes of their children (who are cousins). We argue

that siblings are more likely to experience similar

exposures to internal and external shocks during

childhood than two unrelated individuals in pooled

4Hamid Noghanibehambari and Jason Fletcher

samples. They are plausibly similarly exposed to local

economic shocks, school availability, nurturing

environments, parental culture regarding education,

and genetic endowment compared with two unre-

lated individuals with similar characteristics. We

should note that implementing family-fixed-effects

models and exploiting within-sibling variation does

not account for the full set of shared environments

and genetic endowments. Moreover, unlike studies

that implement a twin-fixed-effects strategy, this so-

called cousin-fixed-effects method does not account

for non-shared genetic traits. We also note that we

cannot use twin- or sibling-fixed-effects models to

analyse children’s outcomes, as there is no variation

in paternal education. A remaining strategy would

be to use fathers who are identical twins but are

discordant for educational attainment; in this case,

the children (cousins) would share the same amount

of genetic similarity as traditional biological siblings

from the same father but would be exposed to

larger unshared non-genetic variation than would

biological siblings raised in the same household.

During the first decades of the twentieth century,

there were many early-life adversities that could

have affected accumulation of human capital and

health endowment with long-lasting legacies for sub-

sequent generations: these include the Spanish flu,

the Great Depression, two world wars, and many

social and political movements (Almond 2006; Myrs-

kylä et al. 2013; Evans et al. 2016; Kose et al. 2021;

Noghanibehambari and Engelman 2022; Schmitz

and Duque 2022; Noghanibehambari and Fletcher

2023). The potential differential exposure to these

adversities may confound cross-sibling comparisons.

However, there are two benefits of such a method

(see Boardman and Fletcher 2015). First, comparing

two siblings still accounts for a portion of childhood

environment and shared genetic traits and hence

provides more reliable estimates than comparing

two arbitrary individuals with similar characteristics.

Second, in the Main results subsection, we show the

surprising stability in coefficients across OLS and

family-fixed-effects models. The fact that taking

into account a portion of shared childhood

exposures and common innate endowments does

not change the effects suggests that these factors

do not confound the estimates. Hence, it is likely

that other unobservables do not confound the

results either. Therefore, we expect within-sibling

variation in education to provide a good framework

for reducing the impact of environmental factors

that may influence education and also have an

impact on the health of the family’s subsequent

generations.

To detect cousins within our Numident-census-

linked sample, we search for the father’s family tree

in the decennial full-count censuses between 1900

and 1930 and then identify cousins in the 1940

Census. To identify grandparents and build a family

tree, we start with historical identification of the

father and use historical crosswalk data files provided

by the Census Linking Project (Abramitzky et al.

2020) to search for their family tree in historical full-

count censuses between 1900 and 1930. In Appendix

A (supplementary material), we replicate the results

using IPUMS Multigenerational Longitudinal

Project (MLP) data, which provide similar crosswalks

across historical censuses. We find very similar effects

when we use MLP as the linking source.

As Abramitzky et al. (2019) explain, various auto-

mated linking techniques give a less than 5 per cent

false positive rate. To assure that our results are

not driven by false positive rates (negative match

wrongly categorized as a positive match), we con-

sider a match between two historical identification

values if for all techniques we have a positive

match. We consider two persons to be siblings if

they share at least one parent when they appear in

historical censuses as children. As an example,

assume that in 1940 there are two persons named

A and B with their fathers’unique identifiers being

FA and FB. From the 1930 Census, we deduce that

FA’s and FB’s birth years are 1887 and 1891, respect-

ively. The reason for using the 1930 Census is that the

1940 Census does not report birth year, whereas pre-

vious censuses do. Also, using father’s age to deduce

birth year is problematic due to potential measure-

ment errors (e.g. age heaping). We start with the

1900 Census (the earliest year that FA and FB

could appear in the census data) and search for

their fathers’and mothers’identifiers. If both identi-

fiers point to the same father or mother, we assume

that FA and FB are siblings and hence A and B are

cousins. In the Additional analyses subsection, we

show that the main results are robust and quite

similar if we focus on those with the same father

and the same mother. Such selection reduces our

current sample by only 4.65 per cent.

We are able to identify 132,810 fathers for whom

parental information is available in previous cen-

suses and whose siblings also have children in the

1940 pooled sample. Hereafter, we refer to these

observations as the ‘sibling-fathers sample’. This

sample covers individuals in cohorts born between

1923 and 1940 and who die in the years 1988–2005.

Therefore, age at death in our sample varies

between 47 and 82. Figure 1 shows the geographic

distribution of places in which fathers’families are

Paternal education and children’s old-age mortality 5

identified. The Northeast region shows higher con-

centrations of identified family trees than the other

regions (South, West, and Midwest).

Figure 2 provides a visual depiction of the process

of final sample construction. The population of indi-

viduals born in 1923–40, who we observe in the 1940

Census and who satisfy our sample selection criteria,

is roughly 32 million. About 7 per cent of these indi-

viduals die in 1988–2005 and are present in the

Numident data. Further, we need to drop from the

sample those fathers whose brothers cannot be

traced in the historical censuses. We are able to

locate about 36 per cent of fathers’households in his-

torical censuses from 1900 to 1930. Finally, we need

to drop those sibling fathers whose children are not

present in Numident death records. These restric-

tions allow roughly 18 per cent of observations that

remain from previous steps to be retained in the

final sample. Overall, the final sample covers 4.2

per cent of the observations of the original sample.

Table 1 reports summary statistics for the original

population, the Numident-census-linked sample, and

the sibling-fathers sample, respectively. Our measure

of mortality, age at death, is similar in the pooled and

sibling-fathers samples. Compared with the original

population, father’s schooling is 0.15 years longer

in the pooled sample and 0.07 years shorter in the

sibling-fathers sample. For the sibling-fathers

sample, we also report the percentage discordant

(percentage of within-family fathers with different

values) for the primary variable of interest (father’s

years of schooling) and the outcome (age at death).

Well over 99 per cent of children in the sibling-

fathers sample die at different ages. Among fathers

within a family, roughly 75 per cent have different

years of schooling (measured in whole years).

The final samples display slightly different demo-

graphic characteristics from the original population.

To account for these differences, we treat the data as

a longitudinal panel with consecutive attrition issues

and reweight the data to adjust for these discrepan-

cies. We follow Halpern-Manners et al. (2020) and

weight the regressions by the inverse of the prob-

ability of linkage between the 1940 Census, Numi-

dent, and historical censuses using probit models

conditioning on individual and family covariates. In

Appendix B (supplementary material), we discuss

the construction of weights.

Econometric method

We start our analysis with a standard OLS estimation

that attempts to account for unobserved factors,

applying a series of fixed effects and family covari-

ates as follows:

Dibsj =

1Sij +

2Xi+

3Zj+

bs +1ibsj, (1)

where the outcome is age at death for individual i

who is from birth cohort band birthplace (state of

birth) sand whose family is indexed by j. The par-

ameter Srepresents a series of dummies for

Figure 1 Distribution of identified households in the final sample (from historical censuses (1900–30) linked to

the 1940 Census and Numident (1988–2005))

Source: Authors’analysis of Numident files linked to 1940 Census data.

6Hamid Noghanibehambari and Jason Fletcher

father’s schooling. Specifically, we include three

binary indicators to capture college education,

high-school education, and middle-school edu-

cation; the reference group is elementary/no edu-

cation. College education is a dummy that equals

one if the father has any college education and

zero otherwise; high-school education is a dummy

that equals one if the maximum years of father’s

schooling is 9–12 and zero otherwise; and middle-

school education is a dummy that equals one if the

maximum years of father’s schooling is 6–8and

zero otherwise.

Birth-state-by-birth-year fixed effects are included

. We use birth-state fixed effects as studies show

that place of birth is more important for lifetime out-

comes than place of residence (Xu, Engelman et al.

2020; Xu, Wu et al. 2020; Xu et al. 2021). However,

in the robustness checks, we include a battery of

fixed effects for current state and county of residence

in 1940, and we find very similar coefficients.

The matrix Xincludes individual controls, such as

sex, race, and origin. Family-level controls, rep-

resented in Z, include dummies for mother’s edu-

cation (with an indicator for missing values),

Figure 2 Steps in selection of the final sample from the original population

Source: As for Figure 1.

Paternal education and children’s old-age mortality 7

father’s number of children, father’s marital status,

father’s labour force status, father’s homeownership

status, and dummies for father’s occupation (blue-

collar or farming sector, with white-collar as refer-

ence group). We cluster standard errors at the

family level. Later we show that the level of cluster-

ing does not change the standard errors, and cluster-

ing at other dimensions reveals the same set of

results (Robustness checks subsection).

A primary concern with these regressions is their

failure to account for unobservables either at the

local-area level or at the father’s family level. For

instance, the availability of schools and colleges

may influence father’s education. Moreover, these

institutes are also located in places with different

characteristics compared with places with fewer

schooling options; these local factors may result in

better health accumulation during childhood and

affect old-age longevity. In addition, more able

fathers may acquire more schooling and also have

more able children who live longer through channels

that operate through inherent endowments. To

account for these confounders, we refine equation

(1) by including (father’s) family fixed effects:.

Dibsj =

1Sij +

2Xi+

bs +1ibsj, (2)

where

represents (father’s) family fixed effects for

the sibling-fathers sample. All other parameters are

similar to those in equation (1). We cluster the stand-

ard errors at the family level.

This strategy compares the mortality outcomes of

cousins whose fathers share the same family and

childhood environment, and, on average, 50 per

cent of their genetic endowments (Hoekstra et al.

2008). However, in addition to the genetic

Table 1 Summary statistics for the original population, the Numident-census-linked sample, and the sibling-fathers sample

Original

population (1940

Census)

Pooled sample

(linked with

Numident death

records)

Sibling-fathers

sample (linked with

grandparents and

Numident)

Mean SD Mean SD Mean SD

Age at death (months) ––829.421 74.037 829.701 73.249

Within-family percentage discordant –– – – 0.998 0.046

Father’s years of schooling 7.914 3.610 8.067 3.383 7.840 3.022

Within-family percentage discordant –– – – 0.750 0.433

Father’s education

None/elementary 0.168 0.374 0.137 0.344 0.124 0.329

Middle school 0.499 0.500 0.531 0.499 0.600 0.490

High school 0.250 0.433 0.254 0.435 0.221 0.415

College 0.082 0.274 0.078 0.267 0.055 0.229

Female 0.490 0.500 0.416 0.495 0.410 0.492

Race

White 0.907 0.291 0.933 0.250 0.954 0.210

Black 0.089 0.284 0.064 0.245 0.044 0.205

Other 0.005 0.068 0.003 0.055 0.002 0.049

Origin

Hispanic 0.022 0.145 0.013 0.111 0.008 0.089

Father’s labour force status 0.965 0.184 0.966 0.181 0.968 0.176

Father’s occupation

Blue-collar 0.043 0.202 0.040 0.197 0.030 0.172

Farm 0.193 0.394 0.181 0.385 0.227 0.419

Father is married 0.983 0.130 0.983 0.129 0.985 0.123

Number of children in the household 3.569 2.069 3.658 2.062 3.914 2.098

Father homeowner 0.397 0.489 0.421 0.494 0.431 0.495

Mother’s education

Zero 0.110 0.313 0.086 0.280 0.071 0.257

Less than High School 0.469 0.499 0.493 0.500 0.538 0.499

High School 0.324 0.468 0.330 0.470 0.315 0.464

Some college 0.068 0.251 0.065 0.246 0.052 0.223

Missing 0.028 0.167 0.026 0.161 0.023 0.152

Observations 32,145,395 2,126,236 132,810

Note: SD is the standard deviation.

Source: Authors’analysis of Numident files linked to 1940 Census data.

8Hamid Noghanibehambari and Jason Fletcher

differences, two potential factors could still con-

found the estimations from equation (2), and we

should be aware of these when interpreting the

results. First, some studies in sociology, psychology,

and psychopathology document differences among

siblings in social and behavioural outcomes that

can be partly attributable to non-shared environ-

ments and non-shared exposures (Dunn and

Plomin 1991; Anderson et al. 1994; Conley et al.

2007; Jensen and McHale 2015). Second, the

shared environment is not equally experienced by

all children, and parents may engage in differential

treatment of their children. This differential treat-

ment could be related to child characteristics (e.g.

sex, health) or parents’characteristics (e.g. edu-

cation) and could reveal compensatory or reinfor-

cing behaviour (Almond and Mazumder 2013;

Frijters et al. 2013; Grätz and Torche 2016; Restrepo

2016; Fletcher et al. 2020). To the extent that non-

shared experiences and differential treatments influ-

ence paternal education and may also appear in

investments in their children’s health, this generates

a bias in equation (2). In the Selection on unobserva-

bles subsection, we show that these confounders

would need to have a large degree of influence to

show considerable effects on the coefficients.

Results

Main results

We start our analysis by discussing the results for spe-

cifications that exclude family fixed effects. These raw

OLS resultsare reported in panel A1, Table 2.Weadd

more covariates across columns, and the effects are

quite robust across models. In the full specification

in column (4), which adds a wide array of father’scon-

trols, maternaleducation, and fixed effects, we can see

that children with college-educated and high-school-

educated fathers live 4.4 and 2.8 months longer,

respectively (compared with those whose fathers

have elementary/no education). In panel A2, we

add father’s family fixed effects to all columns. The

marginal effects are surprisingly stable and compar-

able to the OLS effects in panel A1. Another interest-

ing aspect of both sets of results is the relatively large

and very stable R-squared values. In Appendix C

(supplementary material), we examine this feature

further. We find that the highest contributor to the

R-squared is family fixed effects and the second is

birth-year fixed effects. The contributions of all

other covariates and fixed effects are less than 1 per

cent.

The evidence suggests that after controlling for

fixed characteristics of families in which fathers are

raised and taking into account (although only par-

tially) genetic endowments across siblings, children

whose fathers have a college, high-school, or

middle-school education live 4.6, 2.6, or 1.8 months

longer, respectively, than those with low-educated

fathers. These effects are equivalent to 74, 42, and

29 per cent of the gap between females and males

in the outcome after controlling for other factors

(the coefficient for the ‘female’dummy in the full

specification of column (4) is 6.17).

In panel B, we replace the measure of father’s edu-

cation with a continuous measure of father’s years of

schooling. We observe a similarly robust and consist-

ent pattern across specifications and when compar-

ing OLS models with those that add family fixed

effects (panels B1 and B2). For instance, a one-

standard-deviation change in father’s years of

schooling (roughly three years) is associated with

1.1-month-higher longevity in both models (column

(4)).

These results are in line with the findings of

Montez and Hayward (2011), who examine the

association between measures of childhood family

SES and later-life mortality. They find that children

whose fathers have low levels of education (less

than eight years of schooling) are roughly 13–20

per cent more likely to die at each given age con-

ditional on survival up to age 50. Huebener (2019)

uses data from Germany to examine the association

between parental education and children’s life

expectancy. He finds that, conditional on survival

to age 65, those with more highly educated

mothers live roughly two years longer than those

with low-educated mothers. However, he does not

find significant effects of father’s education on chil-

dren’s longevity. Furthermore, we can also

compare our estimated effects with studies that

examine the association between own education

and mortality. For instance, Halpern-Manners et al.

(2020) implement a twin strategy to explore the

effect of education on mortality among white males

born between 1910 and 1920 and find that an

additional year of schooling raises age at death by

4.2 months. One aspect of their findings is of most

interest and is in line with our results: they show

the effects across unpaired and paired samples of

twins, siblings, and people living in the same neigh-

bourhood. The marginal effects are very similar

across different samples, either paired or unpaired,

and whether including twin fixed effects or not; this

is consistent with limited endogeneity issues in the

long-term links between education and mortality.

Paternal education and children’s old-age mortality 9

Our estimated effects are also comparable to those

in the study by Fletcher and Noghanibehambari

(2021), who explore the impact of county-level

college opening during adolescent years on old-age

longevity. They find that a four-year college (offering

bachelor’s degrees and above) opening in the county

of residence when individuals are 17 years old is

associated with 0.13-months-higher longevity. Their

treatment-on-treated calculation suggests that having

any college education raises age at death by about

12 months. The estimated effects in Tabl e 2 suggest

that having a father with any college education

raises the child’s age at death by 4.6 months, about

one-third of the effect of own college education as

reported by Fletcher and Noghanibehambari (2021).

Cohort selection and truncation

Due to left and right truncation of the data, average

age at death differs across cohorts. For instance, the

average age at death of cohorts born in 1923–30

Table 2 Effects of father’s education on children’s old-age mortality

Outcome: Age at

death (months) (1) (2) (3) (4)

Panel A. Education dummies

Panel A1. OLS models

Father’s education

Middle school 1.83*** 1.95*** 1.92*** 1.95***

(0.57) (0.57) (0.57) (0.58)

High school 2.61*** 2.78*** 2.76*** 2.84***

(0.64) (0.64) (0.65) (0.68)

College 4.73*** 4.77*** 4.43*** 4.35***

(0.88) (0.88) (0.95) (1.03)

Observations 132,810 132,810 132,810 132,810

R-squared 0.48 0.48 0.48 0.48

Panel A2. Family-fixed-effects models

Father’s education

Middle school 1.84** 1.87** 1.79** 1.82**

(0.86) (0.86) (0.86) (0.86)

High school 2.58** 2.64*** 2.51** 2.58**

(1.02) (1.02) (1.03) (1.04)

College 4.54*** 4.75*** 4.46*** 4.59***

(1.43) (1.43) (1.50) (1.55)

Observations 132,810 132,810 132,810 132,810

R-squared 0.71 0.71 0.71 0.71

Panel B. Years of schooling

Panel B1. OLS models

Father’s years of schooling 0.35*** 0.36*** 0.35*** 0.36***

(0.06) (0.06) (0.07) (0.08)

Observations 132,810 132,810 132,810 132,810

R-squared 0.48 0.48 0.48 0.48

Panel B2. Family-fixed-effects models

Father’s years of schooling 0.37*** 0.37*** 0.35*** 0.37***

(0.10) (0.10) (0.11) (0.11)

Observations 132,810 132,810 132,810 132,810

R-squared 0.71 0.71 0.71 0.71

Individual controls ✓✓✓✓

Birth-state FE ✓✓✓✓

Birth-year FE ✓✓✓✓

Birth-state-by-birth-year FE ✓✓✓

Father’s controls ✓✓

Mother’s controls ✓

***p< 0.01, **p< 0.05, *p< 0.10.

Notes: Standard errors, clustered at the family level, are in parentheses. The regressions are weighted using the inverse of the probability of

linkage between the 1940 Census and Numident using probit models conditioning on covariates. Individual controls include a dummy for

sex. Father’s controls include father’s occupation dummies, father’s marital status dummies, father’s number of children in the household in

1940, and father’s age. Mother’s controls include dummies for educational attainment. FE refers to fixed effects.

Source: As for Table 1.

10 Hamid Noghanibehambari and Jason Fletcher

varies between 57 and 82 years and that of the 1931–

40 cohorts varies between 47 and 74 years. There-

fore, a portion of the variation in the final sample

comes from comparing individuals in earlier

cohorts (who die at relatively older ages) with

those in later cohorts (who die at younger ages due

to truncation). Since earlier cohorts contain more

individuals with lower paternal education, part of

the results could reflect this selection due to the trun-

cated nature of the final sample. To test for this

concern empirically, we attempt to compare

cohorts that die at similar ages. In so doing, we

extract two subsamples from the final sample: a sub-

sample that includes those in the 1923–30 cohorts

who die at ages 65–75 (death ages that overlap

between 1923–30 and 1931–40 cohorts due to trunca-

tion) and a subsample of those in the 1931–40

cohorts who die at ages 57–65 (thus excluding the

common death ages across earlier and later

cohorts). We replicate the main results of family-

fixed-effects models for these two groups in Table

3. The full models in column (4) across panels

suggest very similar effects. For instance, the effects

of father’s years of schooling are 0.38 and 0.34 for

earlier and later cohorts, respectively (panels B1

and B2). Moreover, these estimates are fairly

similar to our main results, which further supports

the conclusion that truncation and selection issues

are not the driver of the main results.

Robustness checks

In Table 4, we evaluate the robustness of the results

of the family-fixed-effects models to alternative spe-

cifications, sample selections, and functional forms.

Column (1) reports the fully parametrized family-

fixed-effects results of column (4) of panel A2 of

Table 2 as the benchmark marginal effects. Column

(2) shows the results for an unweighted regression.

The magnitudes and standard errors remain quite

similar to the benchmark estimates.

One concern is the dominance of outliers in

influencing the results. In column (3) we use the

sample from column (1) and drop those families

in which the standard deviation of schooling

across sibling fathers is more than 4.0 (dropping

7 per cent of observations). The resulting coeffi-

cients are slightly larger than the benchmark esti-

mate for the college education dummy and

smaller than the benchmark effects for high-

school and middle-school education dummies.

Further, we use the sample from column (1) and

drop observations that are greater than or less

than two standard deviations from the mean of

father’s schooling and age at death. The results,

reported in columns (4) and (5), are comparable

to the benchmark results.

The main analysis sample is based on individuals

aged 18 at most as the threshold for being observed

within a household so as to locate their father’s

characteristics. In column (6) we observe a slightly

smaller coefficient for middle-school and high-

school education dummies when we restrict the

sample to those aged 12 at most. However, the

effect of college education is larger and remains stat-

istically significant.

Next, we examine the sensitivity of standard errors

to alternative clustering levels. In columns (7) and

(8), we replicate the regression in column (1) but

cluster standard errors at the birth-state and the

1940-county-of-residence levels. In column (9), we

use two-way clustering at the county-of-residence

and birth-state levels. The statistical significance of

the results does not appear to be sensitive to a

specific clustering level.

We continue by exploring the robustness of the

results to additional controls and alternative specifi-

cations. The marginal effects are practically

unchanged when we add father’s occupational

income score as an additional covariate as well as

fixed effects for father’s birth state and father’s

birth year (column (10)). Furthermore, the effects

are comparable to the benchmark estimate when

we add a county-of-residence fixed effect (column

(11)), a county-of-residence-by-birth-cohort fixed

effect (column (12)), and birth-state-by-race and

birth-year-by-race fixed effects (column (13)).

As a final check, we explore the functional form

sensitivity of the results. We replace the outcome

variable with the logarithm of age at death

(column (14)). Although it is not intuitive to inter-

pret the semi-elasticity coefficients, they are statistic-

ally significant, suggesting that the linear

measurement of variables is not a problem in our

analyses.

In Table 5, we replicate these results without

family fixed effects. The effects across columns are

very similar to the benchmark estimates in column

(1) as well as to the results of Table 4.

Subsamples

The results reveal substantial heterogeneity across

subpopulations. We use father’s years of schooling

as the main independent variable in this section to

ease the two-by-two comparisons. In Figure 3 we

Paternal education and children’s old-age mortality 11

show the effects across a variety of subsamples, from

implementing family-fixed-effects and OLS strat-

egies, respectively. The figure shows the estimated

coefficients and their 90 per cent confidence

intervals (horizontal axis) for each subsample (verti-

cal axis) and for each model (dashed line for OLS

and solid line for family-fixed-effects). In these

regressions, we include the full specification as per

Table 3 Effects of father’s education on children’s old-age mortality: restricting the sample to specific cohorts and ages at

death

Outcome: Age at

death (months)

Fixed-effects models

(1) (2) (3) (4)

Panel A. Education dummies

Panel A1. 1923–30 cohorts, age at death 65–75

Father’s education

Middle school 1.76 1.29 1.27 1.33

(1.31) (1.31) (1.31) (1.31)

High school 2.25 1.78 1.77 2.08

(1.53) (1.53) (1.54) (1.57)

College 6.56*** 5.74*** 4.68** 5.05**

(2.15) (2.16) (2.24) (2.32)

Observations 30,926 30,916 30,916 30,916

R-squared 0.48 0.49 0.49 0.49

Mean DV 837.8 837.8 837.8 837.8

Panel A2. 1931–40 cohorts, age at death 57–65

Father’s education:

Middle school 1.56 1.24 1.36 1.38

(1.80) (1.81) (1.81) (1.81)

High school 1.66 1.21 1.51 1.57

(2.11) (2.12) (2.14) (2.17)

College 3.90 4.04 4.13 4.15

(3.11) (3.15) (3.30) (3.41)

Observations 30,128 30,124 30,124 30,124

R-squared 0.62 0.63 0.63 0.63

Mean DV 961.5 961.5 961.5 961.5

Panel B. Years of schooling

Panel B1. 1923–30 cohorts, age at death 65–75

Father’s years of schooling 0.46*** 0.40** 0.34** 0.38**

(0.16) (0.16) (0.16) (0.17)

Observations 30,926 30,916 30,916 30,916

R-squared 0.48 0.49 0.49 0.49

Mean DV 837.8 837.8 837.8 837.8

Panel B2. 1931–40 cohorts, age at death 57–65

Father’s years of schooling 0.34 0.30 0.33 0.34

(0.22) (0.22) (0.23) (0.24)

Observations 30,128 30,124 30,124 30,124

R-squared 0.62 0.63 0.63 0.63

Mean DV 961.5 961.5 961.5 961.5

Family FE ✓✓✓✓

Individual controls ✓✓✓✓

Birth-state FE ✓✓✓✓

Birth-year FE ✓✓✓✓

Birth-state-by-birth-year FE ✓✓✓

Father’s controls ✓✓

Mother’s controls ✓

***p< 0.01, **p< 0.05, *p< 0.10.

Notes: Standard errors, clustered at the family level, are in parentheses. The regressions are weighted using the inverse of the probability of

linkage between the 1940 Census and Numident using probit models conditioning on covariates. Individual controls include a dummy for

sex. Father’s controls include father’s occupation dummies, father’s marital status dummies, father’s number of children in the household in

1940, and father’s age. Mother’s controls include dummies for educational attainment. FE refers to fixed effects.

Source: As for Table 1.

12 Hamid Noghanibehambari and Jason Fletcher

Table 4 Effects of father’s education on children’s old-age mortality: robustness checks of family-fixed-effects strategy

Column (4),

panel A2,

Table 2 Unweighted regression

Drop if within-family SD in

schooling>4

Drop if schooling

.Mean +2SD |

,Mean −2SD

Drop if death age

.Mean +2SD |

,Mean −2SD

Drop if age in 1940

>12

Clustering SE

at birth state

(1) (2) (3) (4) (5) (6) (7)

Father’s education

Middle

school

1.82** 1.97** 1.21 1.41* 1.46 1.08 1.82

(0.86) (0.82) (0.96) (0.81) (0.94) (1.21) (1.09)

High

school

2.58** 2.78*** 1.86 2.18** 2.23** 2.22 2.58*

(1.04) (1.00) (1.19) (0.99) (1.12) (1.46) (1.34)

College 4.59*** 4.62*** 5.88*** 4.88*** 5.66** 5.11** 4.59**

(1.55) (1.49) (2.00) (1.46) (2.67) (2.21) (1.77)

Observations 132,810 132,810 123,676 125,387 119,608 67,729 132,810

R-squared 0.71 0.66 0.71 0.67 0.71 0.67 0.71

Clustering at

county of

residence

Two-way clustering at

county of residence

and birth state

Adding father’s birth-year

FE, birth-state FE, and

occupational income score

Adding county-of-

residence FE

Adding county-of-

residence-by-cohort

Adding birth-state-

by-race FE and

birth-year-by-race

Log of

outcome

(8) (9) (10) (11) (12) (13) (14)

Father’s education

Middle

school

1.82* 1.82** 1.86** 1.61* 1.80 1.58* 0.00**

(0.97) (0.77) (0.89) (0.87) (1.16) (0.87) (0.00)

High

school

2.58** 2.58*** 2.68** 2.19** 3.02** 2.28** 0.00**

(1.16) (0.96) (1.08) (1.06) (1.38) (1.05) (0.00)

College 4.59*** 4.59*** 4.44*** 4.23*** 6.24*** 4.24*** 0.01***

(1.75) (1.48) (1.59) (1.56) (1.98) (1.55) (0.00)

Observations 132,810 132,810 128,257 132,723 110,265 132,779 132,810

R-squared 0.71 0.71 0.71 0.72 0.80 0.71 0.71

***p< 0.01, **p< 0.05, *p< 0.10.

Note: Standard errors (SE), clustered at the father’s family level (except for columns (7)–(9)), are in parentheses. The regressions are weighted using the inverse of the probability of linkage between the

1940 Census, historical censuses, and Numident using probit models conditioning on individuals’sex, race, and origin, father’s occupation dummies, mother’s education dummies, father’s marital status

dummies, father’s number of children in the household at 1940, father’s age, and individual birth-state-by-birth-year fixed effects (FE). All regressions include individual race, sex, and origin dummies,

fixed effects for birth state by birth year, and controls for family covariates including mother’s education, father’s labour force status, father’s marital status, father’s occupation type (blue-collar, farmer),

father’s total number of children, and a dummy for father owning the dwelling. SD refers to the standard deviation.

Source: As for Table 1.

Paternal education and children’s old-age mortality 13

Table 5 Effects of father’s education on children’s old-age mortality: robustness checks of OLS regressions

Column (4),

panel A1,

Table 2

Unweighted

regression

Drop if within-family SD

in schooling>4

Drop if schooling

.Mean +2×SD |

,Mean −2×SD

Drop if death age

.Mean +2×SD |

,Mean −2×SD

Drop if age in

1940>12

Clustering SE

at birth state

(1) (2) (3) (4) (5) (6) (7)

Father’s education

Middle

school

1.95*** 2.23*** 2.12*** 1.86*** 1.84*** 1.94*** 1.95***

(0.58) (0.51) (0.61) (0.51) (0.62) (0.70) (0.54)

High

school

2.84*** 3.30*** 2.81*** 2.90*** 2.75*** 2.65*** 2.84***

(0.68) (0.61) (0.72) (0.60) (0.72) (0.82) (0.63)

College 4.35*** 4.92*** 5.27*** 4.53*** 4.30** 4.47*** 4.35***

(1.03) (0.93) (1.15) (0.91) (1.79) (1.24) (0.90)

Observations 132,810 132,810 123,678 127,970 123,906 85,687 132,810

R-squared 0.48 0.43 0.48 0.44 0.48 0.39 0.48

Clustering at

county of

residence

Two-way clustering at

county-of-residence

and birth-state

Adding father’s birth-year

FE, birth-state FE, and

occupational income score

Adding county-of-

residence FE

Adding county-of-

residence-by-cohort FE

Adding birth-state-

by-race FE and

birth-year-by-race

Log of

outcome

(8) (9) (10) (11) (12) (13) (14)

Father’s education

Middle

school

1.95*** 1.95*** 2.17*** 1.73*** 2.32*** 1.91*** 0.00***

(0.58) (0.46) (0.59) (0.58) (0.68) (0.58) (0.00)

High

school

2.84*** 2.84*** 3.22*** 2.67*** 3.20*** 2.84*** 0.00***

(0.69) (0.58) (0.70) (0.68) (0.80) (0.68) (0.00)

College 4.35*** 4.35*** 4.73*** 4.30*** 4.56*** 4.29*** 0.01***

(1.05) (0.91) (1.05) (1.04) (1.20) (1.02) (0.00)

Observations 132,810 132,810 129,900 132,762 119,607 132,793 132,810

R-squared 0.48 0.48 0.49 0.50 0.58 0.49 0.48

***p< 0.01, **p< 0.05, *p< 0.10.

Notes: Standard errors (SE), clustered at the father’s family level (except for columns (7)–(9)), are in parentheses. The regressions are weighted using the inverse of the probability of linkage between the

father’s total number of children, and a dummy for father owning the dwelling. SD refers to the standard deviation.

Source: As for Table 1.

14 Hamid Noghanibehambari and Jason Fletcher

column (4) of Table 2, including birth-state-by-birth-

year fixed effects and a full set of family controls. For

the family-fixed-effects models, we also add family

fixed effects in addition to all other fixed effects

and covariates included in the OLS regressions.

To provide a benchmark for comparison, we show

the marginal effect (and its 90 per cent confidence

interval) of paternal education on age at death for

the full sample in the first line. The following lines

report the results for subsamples based on race

(white, Black), origin (Hispanic), sex (males,

females), census division region of residence (North-

east, Midwest, South, and West), mother’s education

(less than high school, at least high school), socio-

economic index (SEI: low, high), and ownership of

dwelling (owner, renter). While we show the OLS

results for comparison purposes, our primary refer-

ence analysis is the family-fixed-effects model.

We observe larger effects for Black and Hispanic

people compared with white people. However, in

both cases, the effects are statistically insignificant,

which limits additional comments. This is due

mainly to the small numbers of observations in

these groups reducing statistical power. For

instance, there are only 808 Hispanic and 4,419

Black people in our sample. The main reason for

these small sample sizes is the difficulty in linking

these subpopulations across several data sources.

For instance, the probability of merging between

Numident and the 1940 Census is lower for Black

individuals (see Ta b l e 1). Nonetheless, there is evi-

dence that linked observations of Black individuals

Figure 3 Heterogeneity of the main results across subsamples implementing father’s family-fixed-effects strat-

egy and OLS strategy

Notes: The 90 per cent confidence intervals are based on standard errors that are clustered at the father’s family level. The

regressions are weighted using the inverse of the probability of linkage between the 1940 Census, historical censuses, and

Numident using probit models conditioning on covariates. All regressions include individual race, sex, and origin

dummies, fixed effects for birth state by birth year, and controls for family covariates including mother’s education,

father’s labour force status, father’s marital status, father’s occupation type (blue-collar, farmer), father’s total number of

children, and a dummy for father’s being ownership of the dwelling.

Source: As for Figure 1.

Paternal education and children’s old-age mortality 15

are representative of the Black population nation-

ally (Breen and Osborne 2022). However, the

difference between the share of Black (and Hispa-

nic) individuals in our final sample and the original

population is much larger than the difference

between the share of Black (and Hispanic) individ-

uals in Numident-census-linked data and the orig-

inal population. Therefore, we should exercise

caution in interpreting the heterogeneity analyses

by race, as there could be selection based on unob-

servables that make the final Black and Hispanic

samples less representative of their corresponding

populations. One concern that may arise is that

the estimates are biased due to incorrect links and

that the incorrect links might be more prevalent in

harder-to-link subpopulations such as Black and

Hispanic groups. We should note that the Numi-

dent–census linking is based primarily on name

commonalities, although some harder-to-link sub-

populations, such as Black and immigrant groups,

aremorelikelytobeenumeratedwitherror,

hence lower linking rates and at the same time

differential education–longevity associations. We

discuss the endogenous merging concern in Appen-

dix D (supplementary material). Specifically, we

explore whether more educated fathers are more

or less likely to be linked from the original 1940

Census population to the Numident data. We find

very small education-linking coefficients that,

even in the worst scenario, will induce an ignorable

bias into our estimations. Moreover, we observe a

very similar coefficient when we look at whites, a

subpopulation with more accurate links, suggesting

that the effects are unlikely to be confounded by the

linking rules of harder-to-link subpopulations. Fur-

thermore, incorrect links will lead to unrelated

people being assigned into a family unit. Therefore,

we expect to observe underestimated coefficients

and attenuated effects, as the effects of education

on longevity will be smaller in comparisons of

non–family members.

Studies suggest that the health–education gradient,

more specifically the mortality–education gradient, is

a function of childhood and adulthood local-area

characteristics, and hence the gradients vary by geo-

graphic regions (Sheehan et al. 2018; Kemp and

Montez 2020). Therefore, we also expect to observe

heterogeneous effects of paternal education on child

mortality across different regions. In the subsample

analyses based on different census regions, we

observe larger impacts among people residing in the

South compared with other regions (Figure 3).

The evidence also suggests complementarity

between mother’s and father’s educational

attainment. The effect of father’s education is con-

siderably smaller (and statistically insignificant)

when mother’s education is low (0.16 vs 0.37) and

slightly larger (and statistically significant) when

the mother is educated at least to high-school level

(0.61 vs 0.37).

Finally, there is evidence that the education–

health gradient depends on the socio-demographic

status of families (Barrow and Rouse 2005; Kimbro

et al. 2008; Seeman et al. 2008). To examine this

source of heterogeneity, we replicate the results

across families who are above and below median

SEI. The results in Figure 3 suggest larger effects

among families who are below the median socio-

economic score. The association is statistically insig-

nificant for the above-median group. We also

examine heterogeneity by ownership of dwelling,

but we do not find differences in the effects across

owners vs renters.

Additional analyses

For the main results, we consider two fathers to be

siblings if, in historical censuses, they share the

same family and at least one of their parents is the

same. We now impose a stricter condition for two

persons to be siblings: both parents should be alive

(and living within the same family) in historical cen-

suses to be identified. Therefore, we assume two

fathers are siblings if they have the same parents

and both of their parents are present in the house-

hold. The resulting sample size is only 4.65 per cent

smaller than the final sample for the main results.

We replicate the main results for this analytic

sample and report them in Table 6. The OLS

results are very similar to the main results.

However, the family-fixed-effects estimates are

slightly larger than those reported in Table 2.

Selection on unobservables

Equation (2) differences out all shared genetic

characteristics, shared family features, and shared

childhood environment exposures across siblings in

the father’s family. However, it does not account

for non-shared environment, non-shared genetic

characteristics, or any discriminatory behaviour of

parents. The fact that the OLS estimates are very

close to the family-fixed-effects estimates suggests

that fathers’childhood experiences are not con-

founding the estimates. However, we may be con-

cerned that while unobservable characteristics of

16 Hamid Noghanibehambari and Jason Fletcher

families are not introducing bias in our estimates,

their differential impacts on members of a family

could be confounding them. To gauge the degree to

which these unobservables could be biasing the esti-

mates, we apply a series of artificial unobservable

shocks with a pre-specified relationship with the

outcome and explanatory variables and calculate

the marginal effects in each scenario.

To begin, we regress the outcome (age at death)

on a series of observables, namely mother’s school-

ing, father’s labour force status, father’s income,

family’s SEI, father’s occupational income score,

father’s marital status, father’s number of children,

and father’s ownership of dwelling. We use the pre-

dicted value of this regression as a combination of

observables. The correlation of this variable with

father’s education is 0.03 and with the outcome is

0.10. We posit that the differential effect of

unobservable characteristics could be as strong as

the effect of a combination of all observables,

although only in extreme cases. Therefore, we gener-

ate an artificial random variable (z) that is correlated

with a value between −0.1 and 0.1 with both the

outcome and explanatory variables. We then

implement a full specification of the family-fixed-

effects model (column (4), panel A2, Table 2). To

reach more precise correlations, we implement a

semi–Monte Carlo simulation, running these

regressions 1,000 times for each combination of cor-

relations and computing the average of estimated

coefficients. Here, in order to ease interpretation of

the findings and construct a correlation matrix, we

use years of schooling instead of the three education

dummies. The calculated point estimates are

reported in Figure 4.

Each point estimate in this figure is associated

with a regression that adds an artificial confounder

(z) with a pre-defined correlation with father’s

years of schooling (corr(z,x) as shown on the x-

axis) and a pre-defined correlation with the

outcome (corr(z,y)). For illustration purposes, we

group and connect all point estimates with a specific

corr(z,y) and reveal their specified correlation in the

figure’s legend. As we would expect, when the corre-

lation of artificial variable (z) with father’s education

(x) is zero, there is no bias in the coefficients and

they converge to the benchmark effect of 0.37.

Also, when zis not correlated with the outcome

(y), it should not change the marginal effects, as

shown by the green-triangle dotted line in Figure 4.

The artificial bias generated by zchanges the mar-

ginal effects, but they vary between 0.22 and 0.48

(still far from zero) and are statistically significant

at conventional levels in all cases. There are two

extreme cases where the coefficients drop below

0.3. The first is where corr(z,y)=0.1 and

Table 6 Effects of father’s education and children’s old-age mortality: fathers with both parents present in historical

censuses

Outcome: Age at

death (months)

OLS Family-fixed-effects strategy

(1) (2) (3) (4) (5) (6)

Father’s education

Middle school 2.01*** 1.96*** 1.99*** 2.32*** 2.24** 2.26**

(0.59) (0.59) (0.59) (0.89) (0.89) (0.89)

High school 2.97*** 2.94*** 3.00*** 3.18*** 3.06*** 3.15***

(0.66) (0.67) (0.70) (1.05) (1.06) (1.07)

College 4.74*** 4.44*** 4.34*** 5.17*** 5.12*** 5.22***

(0.90) (0.98) (1.06) (1.47) (1.53) (1.58)

Observations 126,632 126,632 126,632 126,632 126,632 126,632

R-squared 0.49 0.49 0.49 0.71 0.71 0.71

Father’s family FE ✓✓✓

Birth-state and birth-year FE ✓✓✓✓✓✓

Individual controls ✓✓✓✓✓✓

Birth-state-by-birth-year FE ✓✓ ✓✓

Father’s characteristics ✓✓

Mother’s education control ✓✓

***p< 0.01, **p< 0.05, *p< 0.10.

Notes: Standard errors, clustered at the family level, are in parentheses. The regressions are weighted using the inverse of the probability of

linkage between the 1940 Census, historical censuses, and Numident using probit models conditioning on individual sex, race, and origin.

Father’s controls include father’s labour force status, father’s marital status, father’s occupation type (blue-collar, farmer), father’s total

number of children, and a dummy for father owning the dwelling. Individual controls include race, sex, and origin dummies. FE refers

to fixed effects. Columns (1)–(3) include the same set of covariates as columns (1), (2) and (4) in Table 2.

Source: As for Table 1.

Paternal education and children’s old-age mortality 17

corr(z,x)=−0.1, which is an unobservable that

inhibits father’s education but appears positive and

strong for children’s old-age health, a hard-to-find

unobservable. The second is where both yand x

show a relatively strong and positive correlation

(0.1) with the artificial unobservable z. For instance,

if there is a personality trait that leads to higher edu-

cation in a father and not his brother, that personal-

ity trait is as important for his child’s longevity as all

other socio-economic and demographic character-

istics of his family, including the child’s mother’s edu-

cation, a genetic trait, or differential treatment by

parents that outweighs all other observables. In this

case, we expect the real marginal effects to decrease

to 0.2, and in more extreme cases to zero.

Potential mechanisms

In this section, we explore potential mechanism

channels between father’s education and old-age

longevity. An important channel is children’s edu-

cation, through the intergenerational transmission

process. Since we do not have data on the com-

pleted education of children (they are still too

young to have completed their education in

1940), we turn to alternative measures based on

variables available in the 1940 Census, where the

data report the highest grade attained. We

compute the median grade level for each age. To

avoid measurement errors in reported age, we cal-

culate age based on exact date of birth available in

Figure 4 Monte Carlo simulation results from adding an additional regressor (z) with a pre-specified corre-

lation with father’s years of schooling (x) and child’s age at death (y) in a family-fixed-effects strategy

Notes: The regressions are weighted using the inverse of the probability of linkage between the 1940 Census, historical cen-

suses, and Numident using probit models conditioning on individuals’sex, race, and origin, father’s occupation dummies,

mother’s education dummies, father’s marital status dummies, father’s number of children in the household at 1940,

father’s age, and individual’s birth-state-by-birth-year fixed effects. All regressions include sex, race, and origin dummies,

fixed effects for birth year by birth state, and a series of family controls including mother’s years of schooling, father’s

labour force status, father’s marital status, father’s occupation type (blue-collar, farmer), father’s total number of children,

and a dummy for father’s ownership of the dwelling.

Source: As for Figure 1.

18 Hamid Noghanibehambari and Jason Fletcher

Table 7 Effects of father’s education on families’socio-economic characteristics: OLS and family-fixed-effects strategies

Child’s old-

for-grade

status

Child’s school

attendance

status

Mother’s

years of

schooling

Employed more

than 48 weeks

last year

Father in

the labour

force

Log

father’s

income

Father’s Socio-

Economic

Index

Father’s

occupational

income score

Father’s

number of

children

Father

owns

house

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Panel A. OLS models

Father’s education

Middle

school

−0.13*** 0.07*** 2.22*** 0.08*** 0.02*** 0.26*** 4.02*** 2.05*** −0.61*** 0.05***

(0.00) (0.00) (0.03) (0.01) (0.00) (0.04) (0.17) (0.09) (0.03) (0.01)

High

school

−0.18*** 0.10*** 4.05*** 0.18*** 0.03*** 0.50*** 14.75*** 5.85*** −1.37*** 0.07***

(0.00) (0.00) (0.04) (0.01) (0.00) (0.04) (0.23) (0.11) (0.03) (0.01)

College −0.21*** 0.11*** 6.46*** 0.27*** 0.03*** 1.00*** 37.96*** 13.61*** −1.72*** 0.16***

(0.00) (0.01) (0.06) (0.01) (0.00) (0.04) (0.42) (0.24) (0.04) (0.01)

Observations 110,063 110,063 129,629 132,810 132,810 97,523 128,038 129,903 132,810 132,810

R-squared 0.22 0.30 0.40 0.06 0.01 0.05 0.25 0.20 0.11 0.06

Mean DV 0.12 0.86 8.06 0.62 0.96 7.18 23.99 23.06 3.94 0.41

Panel B. Family-fixed-effects models

Father’s education

Middle

school

−0.09*** 0.04*** 1.50*** 0.05*** 0.01*** 0.16*** 3.02*** 1.55*** −0.32*** 0.03***

(0.01) (0.01) (0.04) (0.01) (0.00) (0.05) (0.25) (0.13) (0.04) (0.01)

High

school

−0.12*** 0.06*** 2.74*** 0.11*** 0.01*** 0.32*** 10.04*** 3.92*** −0.77*** 0.05***

(0.01) (0.01) (0.05) (0.01) (0.00) (0.05) (0.33) (0.16) (0.04) (0.01)

College −0.14*** 0.07*** 4.73*** 0.16*** 0.02*** 0.66*** 28.77*** 10.15*** −1.06*** 0.09***

(0.01) (0.01) (0.08) (0.01) (0.01) (0.07) (0.57) (0.30) (0.05) (0.01)

Observations 100,171 100,171 127,649 132,810 132,810 84,562 125,307 128,263 132,810 132,810

R-squared 0.57 0.64 0.78 0.63 0.60 0.69 0.72 0.71 0.68 0.64

Mean DV 0.12 0.87 8.06 0.62 0.96 7.17 23.96 23.06 3.94 0.41

***p< 0.01, **p< 0.05, *p< 0.10.

Notes: Standard errors, clustered at the family level, are in parentheses. The regressions are weighted using the inverse of the probability of linkage between the 1940 Census, historical censuses, and

Numident using probit models conditioning on individual sex, race, and origin, father’s occupation dummies, mother’s education dummies, father’s marital status dummies, father’s number of

children in the household at 1940, father’s age, and individual birth-state-by-birth-year fixed effects. All regressions include individual race, sex, and origin dummies, fixed effects for birth state by

birth year. The outcome in column (1), child’s old-for-grade status, is a dummy indicating the age of child being at least two years less than the median age of the grade they have attained. DV

stands for dependent variable.

Source: As for Table 1.

Paternal education and children’s old-age mortality 19

the SSA Numident files and designate an old-for-

grade status to those who do attend school but

whose grade level lies at least two years below

the age-specific median grade level in the data.

Silles (2017) uses a similar method to determine

the grade retention status of children. However,

to obtain a precise grade retention status, we

need information on exact state-level enrolment-

age policies as well as the degree to which individ-

uals comply with those policies. We call this vari-

able ‘old for grade’, as some individuals start

school later than others and not all people who

are old for grade experience a grade retention.

We also use data on school attendance to gen-

erate a dummy indicating whether or not a

person is currently attending school. We exclude

individuals under six years old. We explore the

effect of father’s schooling on these outcomes

using the sibling-fathers sample. The results are

reported in Table 7 , columns (1) and (2), for the

OLS (panel A) and family-fixed-effects (panel

B) strategies. All regressions include the fixed

effects and individual covariates used in the

main results in Tab le 2 . The family-fixed-effects

models suggest that having a college-educated

or high-school-educated father is associated with

a 13.5- or 12.1-percentage-point lower likelihood

of old-for-grade status and a 6.6- or 6.2-percen-

tage-point higher likelihood of attending school,

respectively.

We continue to explore possible mechanisms by

exploiting the available data on paternal socio-

economic measures. These results are reported in

columns (3)–(10) of Table 7. There are strong associ-

ations between father’s education and mother’s

years of schooling and between father’s education

and a range of other paternal characteristics, includ-

ing longer duration of employment, being in the

labour force, having a higher income or SEI,

having fewer children, and owning the dwelling of

residence. All these effects can be translated into

improvements in well-being, better access to

material resources, better healthcare access, better

access to high-quality insurance, and a generally

healthier environment during childhood, all of

which in turn add to children’s health capital over

their childhood and can be detected in their old-

age longevity and mortality improvements (Van

Den Berg et al. 2006,2009,2011,2015; Strand and

Kunst 2006a,2006b; Fletcher et al. 2010; Black

et al. 2015; Scholte et al. 2015; Wherry and Meyer

2016; Sohn 2017; Fletcher 2018; Wherry et al. 2018;

Goodman-Bacon 2021; Noghanibehambari 2022b,

2022c).

Conclusion

In an early book, Mangold (1920)suggeststhat

parental SES, as well as parental education and

knowledge, can directly or through other channels

affect child health outcomes, such as disability and

mortality. Several case studies also suggest that the

parental non-genetic package (education, income,

housing, wealth, health knowledge, etc.) influences

child development and health outcomes and that

this correlation is not mere coincidence (Levy

1919;WileandDavis1939; Gough 1946). Recent

research establishes this beyond-coincidence

relationship and documents a direct effect of the

parental package and specifically parental edu-

cation on children’s health outcomes (Currie

2009;Chouetal.2010; Ross and Mirowsky 2011;

Reinhold and Jürges 2012; Currie and Goodman

2020). The current study adds to this long-standing

literature by evaluating the effect of father’sedu-

cation on an important and precise measure of

their children’s long-run health: old-age mortality.

To account for shared characteristics of the

environment in which fathers grow up and to

control for the shared family experiences and child-

hood exposures that affect the father’s educational

decisions and also confound the children’s old-age

mortality equation, we implement a family-fixed-

effects model. This strategy compares the within-

family old-age mortality outcomes of cousins

whose fathers experience and share the same

family but obtain different years of schooling. Com-

paring the family-fixed-effects estimates with the

OLS results suggests that common exposures and

shared characteristics of siblings generate very little

bias in the coefficients. The findings imply very

robust and consistent evidence that father’s edu-

cation is positively associated with children’s longev-

ity. On average, children of college-educated fathers

live 4.6 months more than low-educated fathers, con-

ditional on survival up to age 47. This is roughly 74

per cent of the gap between females and males in

the outcome (age at death).

Notes and acknowledgements

1 Hamid Noghanibehambari is based in the Center for

Demography of Health and Aging, University of Wis-

consin–Madison. Jason Fletcher is based in La Follette

School of Public Affairs, University of Wisconsin–

Madison.

2 Please direct all correspondence to Hamid Noghanibe-

hambari, Center for Demography of Health and

20 Hamid Noghanibehambari and Jason Fletcher

Aging, University of Wisconsin–Madison, 180 Observa-

tory Drive, Unit 4401 Madison Wisconsin 53706, USA;

or by Email: noghanibehambarih@apsu.edu.

3 The authors would like to acknowledge financial

support from the National Institute on Aging (NIA),

grant number R01AG060109, and the Center for Demo-

graphy of Health and Aging at the University of Wis-

consin–Madison under NIA core grant number P30

AG17266.

Disclosure statement

No potential conflict of interest was reported by the

authors.

References

Abramitzky, R., L. Boustan, and K. Eriksson. 2012.

Europe’s tired, poor, huddled masses: Self-selection

and economic outcomes in the age of mass migration,

American Economic Review 102(5): 1832–1856.

https://doi.org/10.1257/AER.102.5.1832

Abramitzky, R., L. Boustan, and K. Eriksson. 2014.A

nation of immigrants: Assimilation and economic out-

comes in the age of mass migration, Journal of

Political Economy 122(3): 467–506. https://doi.org/10.

1086675805

Abramitzky, R., L. Boustan, and K. Eriksson. 2019. To the

New World and back again: Return migrants in the age

of mass migration, ILR Review 72(2): 300–322. https://

doi.org/10.1177/0019793917726981

Abramitzky, R., L. Boustan, K. Eriksson, J. Feigenbaum,

and S. Pérez. 2021. Automated linking of historical

data, Journal of Economic Literature 59(3): 865–918.

https://doi.org/10.1257/JEL.20201599

Abramitzky, R., L. Boustan, and M. Rashid. 2020. Census

Linking Project: Version 1.0 [dataset]. Available at:

https://censuslinkingproject.org

Acemoglu, D. and J. Angrist. 2000. How large are human-

capital externalities? Evidence from compulsory

schooling laws, NBER Macroeconomics Annual 15: 9–

59. https://doi.org/10.1086/654403

Agüero, J. M. and M. Ramachandran. 2020. The interge-

nerational transmission of schooling among the edu-

cation-rationed, Journal of Human Resources 55(2):

504–538. https://doi.org/10.3368/jhr.55.2.0816.8143R

Ahima, R. S. and M. A. Lazar. 2013. The health risk of

obesity—Better metrics imperative, Science 341(6148):

856–858. https://doi.org/10.1126/science.1241244

Albarrán, P., M. Hidalgo-Hidalgo, and I. Iturbe-Ormaetxe.

2020. Education and adult health: Is there a causal

effect?, Social Science & Medicine 249: 112830. https://

doi.org/10.1016/j.socscimed.2020.112830

Alderman, H., J. Hoddinott, and B. Kinsey. 2006. Long

term consequences of early childhood malnutrition,

Oxford Economic Papers 58(3): 450–474. https://doi.

org/10.1093/oep/gpl008

Ali, F. R. M. and M. A. A. Elsayed. 2018.The effect of par-

ental education on child health: Quasi-experimental

evidence from a reduction in the length of primary

schooling in Egypt, Health Economics 27(4): 649–662.

https://doi.org/10.1002/hec.3622

Almond, D. 2006. Is the 1918 influenza pandemic over?

Long-term effects of in utero influenza exposure in the

post-1940 US population, Journal of Political Economy

114(4): 672–712. https://doi.org/10.1086/507154

Almond, D. and J. Currie. 2011. Killing me softly: The fetal

origins hypothesis, Journal of Economic Perspectives 25

(3): 153–172. https://doi.org/10.1257/jep.25.3.153

Almond, D., J. Currie, and V. Duque. 2018. Childhood cir-

cumstances and adult outcomes: Act II, Journal of

Economic Literature 56(4): 1360–1446. https://doi.org/

10.1257/jel.20171164

Almond, D., H. W. Hoynes, and D. W. Schanzenbach. 2011.

Inside the war on poverty: The impact of food stamps

on birth outcomes, Review of Economics and Statistics

93(2): 387–403. https://doi.org/10.1162/REST_a_00089

Almond, D. and B. Mazumder. 2013. Fetal origins and

parental responses, Annual Review of Economics 5:

37–56. https://doi.org/10.1146/annurev-economics-

082912-110145

Anderson, E. R., E. M. Hetherington, D. Reiss, and G.

Howe. 1994. Parents’nonshared treatment of siblings

and the development of social competence during ado-

lescence, Journal of Family Psychology 8(3): 303–320.

https://doi.org/10.1037/0893-3200.8.3.303

Balaj, M., H. W. York, K. Sripada, E. Besnier, H. D.

Vonen, A. Aravkin, J. Friedman, et al. 2021. Parental

education and inequalities in child mortality: A global

systematic review and meta-analysis, The Lancet 398

(10300): 608–620. https://doi.org/10.1016/S0140-6736

(21)00534-1

Barcellos, S. H., L. S. Carvalho, and P. Turley. 2018.

Education can reduce health differences related to

genetic risk of obesity, Proceedings of the National

Academy of Sciences 115(42). https://doi.org/10.1073/

pnas.1802909115

Barcellos, S. H., L. S. Carvalho, and P. Turley. 2019.

Distributional effects of education on health, NBER

Working Paper. Available at: https://doi.org/10.3386/

W25898

Barreca, A. and M. Page. 2015. A pint for a pound?

Minimum drinking age laws and birth outcomes,

Health Economics 24(4): 400–418. https://doi.org/10.

1002/hec.3026

Barrow, L. and C. E. Rouse. 2005. Do returns to schooling

differ by race and ethnicity?, American Economic

Paternal education and children’s old-age mortality 21

Review 95(2): 83–87. https://doi.org/10.1257/

000282805774670130

Behrman, J. R. and M. R. Rosenzweig. 2002. Does increas-

ing women’s schooling raise the schooling of the next

generation?, American Economic Review 92(1): 323–

334. https://doi.org/10.1257/000282802760015757

Behrman, J. R. and M. R. Rosenzweig. 2004. Returns

to birthweight, Review of Economics and Statistics 86(2):

586–601. https://doi.org/10.1162/003465304323031139

Black, S. E., P. J. Devereux, and K. G. Salvanes. 2007. From

the cradle to the labor market? The effect of birth

weight on adult outcomes, The Quarterly Journal of

Economics 122(1): 409–439. https://doi.org/10.1162/

qjec.122.1.409

Black, D. A., Y. C. Hsu, and L. J. Taylor. 2015. The effect of

early-life education on later-life mortality, Journal of

Health Economics 44: 1–9. https://doi.org/10.1016/j.

jhealeco.2015.07.007

Boardman, J. D. and J. M. Fletcher. 2015. To cause or not

to cause? That is the question, but identical twins might

not have all of the answers, Social Science & Medicine

127: 198–200. https://doi.org/10.1016/j.socscimed.2014.

10.013

Breen, C. F. and M. Osborne. 2022.An Assessment of

CenSoc Match Quality. Available at: https://doi.org/10.

31235/OSF.IO/BJ5MD

Breierova, L. and E. Duflo. 2004. The impact of education

on fertility and child mortality: Do fathers really matter

less than mothers? NBER Working Paper. Available at:

https://doi.org/10.3386/W10513

Buckles, K., A. Hagemann, O. Malamud, M. Morrill, and

A. Wozniak. 2016. The effect of college education on

mortality, Journal of Health Economics 50: 99–114.

https://doi.org/10.1016/j.jhealeco.2016.08.002

Caldwell, J. and P. McDonald. 1982. Influence of maternal

education on infant and child mortality: Levels and

causes, Health Policy and Education 2(3-4): 251–267.

https://doi.org/10.1016/0165-2281(82)90012-1

Card, D. 2001. Estimating the return to schooling: Progress

on some persistent econometric problems,

Econometrica 69(5): 1127–1160. https://doi.org/10.

1111/1468-0262.00237

Carneiro, P., C. Meghir, and M. Parey. 2013. Maternal edu-

cation, home environments, and the development of

children and adolescents, Journal of the European

Economic Association 11: 123–160. https://doi.org/10.

1111/j.1542-4774.2012.01096.x

Carter, B. D., C. Abnet, D. Feskanich, N. Freedman, P.

Hartge, C. Lewis, J. Ockene, et al. 2015. Smoking and

mortality —Beyond established causes, New England

Journal of Medicine 372(7): 631–640. https://doi.org/

10.1056/NEJMsa1407211

Cawley, J., D. Frisvold, and C. Meyerhoefer. 2013.The

impact of physical education on obesity among

elementary school children, Journal of Health

Economics 32(4): 743–755. https://doi.org/10.1016/j.

jhealeco.2013.04.006

Cecchini, M., F. Sassi, J. A. Lauer, Y. Y. Lee, V. Guajardo-

Barron, and D. Chisholm. 2010. Tackling of unhealthy

diets, physical inactivity, and obesity: Health effects

and cost-effectiveness, The Lancet 376(9754): 1775–

1784. https://doi.org/10.1016/S0140-6736(10)61514-0

Chalfin, A. and M. Deza. 2018. The effect of parental edu-

cation on children’s drug and alcohol use, AEA Papers

and Proceedings 108: 373–378. https://doi.org/10.1257/

pandp.20181105

Chalfin, A. and M. Deza. 2019. The intergenerational

effects of education on delinquency, Journal of

Economic Behavior & Organization 159: 553–571.

https://doi.org/10.1016/j.jebo.2017.07.034

Chang, L. V. 2018. Information, education, and health

behaviors: Evidence from the MMR vaccine autism

controversy, Health Economics 27(7): 1043–1062.

https://doi.org/10.1002/hec.3645

Chen, Y. and H. Li. 2009. Mother’s education and child

health: Is there a nurturing effect?, Journal of Health

Economics 28(2): 413–426. https://doi.org/10.1016/j.

jhealeco.2008.10.005

Chou, S. Y., J. T. Liu, M. Grossman, and T. Joyce. 2010.

Parental education and child health: Evidence from a

natural experiment in Taiwan, American Economic

Journal: Applied Economics 2(1): 33–61. https://doi.

org/10.1257/app.2.1.33

Clark, D. and H. Royer. 2013. The effect of education on

adult mortality and health: Evidence from Britain,

American Economic Review 103(6): 2087–2120.

https://doi.org/10.1257/aer.103.6.2087

Cohen, A. K., D. H. Rehkopf, J. Deardorff, and B. Abrams.

2013. Education and obesity at age 40 among American

adults, Social Science & Medicine 78(1): 34–41. https://

doi.org/10.1016/j.socscimed.2012.11.025

Conley, D., K. M. Pfeiffer, and M. Velez. 2007. Explaining

sibling differences in achievement and behavioral out-

comes: The importance of within- and between-family

factors, Social Science Research 36(3): 1087–1104.

https://doi.org/10.1016/j.ssresearch.2006.09.002

Conti, G., J. Heckman, and S. Urzua. 2010. The education-

health gradient, American Economic Review 100(2):

234–238. https://doi.org/10.1257/aer.100.2.234

Cui, Y., H. Liu, and L. Zhao. 2019. Mother’s education and

child development: Evidence from the compulsory

school reform in China, Journal of Comparative

Economics 47(3): 669–692. https://doi.org/10.1016/j.jce.

2019.04.001

Currie, Janet. 2009. Healthy, wealthy, and wise:

Socioeconomic status, poor health in childhood, and

human capital development, Journal of Economic

Literature 47(1): 87–122. https://doi.org/10.1257/jel.47.1.87

22 Hamid Noghanibehambari and Jason Fletcher

Currie, J. and J. Goodman. 2020. Parental socioeconomic

status, child health, and human capital, The

Economics of Education: A Comprehensive Overview:

239–248. https://doi.org/10.1016/B978-0-12-815391-8.

00018-5

Currie, J. and E. Moretti. 2003. Mother’s education and the

intergenerational transmission of human capital:

Evidence from college openings, The Quarterly

Journal of Economics 118(4): 1495–1532. https://doi.

org/10.1162/003355303322552856

Cutler, D. M., W. Huang, and A. Lleras-Muney. 2015.

When does education matter? The protective effect of

education for cohorts graduating in bad times, Social

Science & Medicine 127: 63–73. https://doi.org/10.1016/

j.socscimed.2014.07.056

Cutler, D. M. and A. Lleras-Muney. 2012.Education and

Health: Insights from International Comparisons.

Available at: https://doi.org/10.3386/W17738

Daouli, J., M. Demoussis, and N. Giannakopoulos. 2010.

Mothers, fathers and daughters: Intergenerational

transmission of education in Greece, Economics of

Education Review 29(1): 83–93. https://doi.org/10.

1016/j.econedurev.2009.02.006

Davies, N. M., M. Dickson, G. D. Smith, G. J. Van Den

Berg, F. Windmeijer, G. van den Berg, and F.

Windmeijer. 2016. The causal effects of education on

Health, Mortality, Cognition, Well-being, and Income

in the UK Biobank, BioRxiv. Available at: https://doi.

org/10.1101/074815

Dickson, M., P. Gregg, and H. Robinson. 2016. Early, late

or never? When does parental education impact child

outcomes?, The Economic Journal 126(596): F184–

F231. https://doi.org/10.1111/ecoj.12356

Dunn, J. and R. Plomin. 1991. Why are siblings so differ-

ent? The significance of differences in sibling experi-

ences within the family, Family Process 30(3): 271–

283. https://doi.org/10.1111/j.1545-5300.1991.00271.x

Elo, I. T., P. Martikainen, and M. Myrskylä. 2014.

Socioeconomic status across the life course and all-

cause and cause-specific mortality in Finland, Social

Science & Medicine 119: 198–206. https://doi.org/10.

1016/j.socscimed.2013.11.037

Engelman, M., V. Canudas-Romo, and E. M. Agree. 2010.

The implications of increased survivorship for mortality

variation in aging populations, Population and

Development Review 36(3): 511–539. https://doi.org/

10.1111/J.1728-4457.2010.00344.X

Evans, M. F., E. Helland, J. Klick, and A. Patel. 2016.The

developmental effect of state alcohol prohibitions at

the turn of the twentieth century, Economic Inquiry

54(2): 762–777. https://doi.org/10.1111/ecin.12303

Ferrie, J. P. 1996. A new sample of males linked from the

public use microdata sample of the 1850 U.S. federal

census of population to the 1860 U.S. federal census

manuscript schedules, Historical Methods: A Journal of

Quantitative and Interdisciplinary History 29(4): 141–

156. https://doi.org/10.1080/01615440.1996.10112735

Fischer, M., M. Karlsson, and T. Nilsson. 2013. Effects of

compulsory schooling on mortality: Evidence from

Sweden, International Journal of Environmental

Research and Public Health 10(8): 3596–3618. https://

doi.org/10.3390/IJERPH10083596

Fitzpatrick, M. D. and T. J. Moore. 2018. The mortality

effects of retirement: Evidence from social security eli-

gibility at age 62, Journal of Public Economics 157: 121–

137. https://doi.org/10.1016/j.jpubeco.2017.12.001

Fletcher, J. M. 2015. New evidence of the effects of edu-

cation on health in the US: Compulsory schooling

laws revisited, Social Science & Medicine 127: 101–

107. https://doi.org/10.1080/19485565.2018.1501267

Fletcher, J. M. 2018. New evidence on the impacts of early

exposure to the 1918 influenza pandemic on old-age

mortality, Biodemography and Social Biology 64(2):

123–126. https://doi.org/10.1080/19485565.2018.1501267

Fletcher, J. M. and D. E. Frisvold. 2009. Higher education

and health investments: Does more schooling affect

preventive health care use?, Journal of Human

Capital 3(2): 144–176. https://doi.org/10.1086/645090

Fletcher, J. M. and D. E. Frisvold. 2011. College selectivity

and young adult health behaviors, Economics of

Education Review 30(5): 826–837. https://doi.org/10.

1016/J.ECONEDUREV.2011.04.005

Fletcher, J. M., J. C. Green, and M. J. Neidell. 2010. Long

term effects of childhood asthma on adult health,

Journal of Health Economics 29(3): 377–387. https://

doi.org/10.1016/J.JHEALECO.2010.03.007

Fletcher, J. M. and H. Noghanibehambari. 2021.The

Effects of Education on Mortality: Evidence Using

College Expansions. Available at: https://doi.org/10.

3386/W29423

Fletcher, J. M., Y. Wu, Z. Zhao, and Q. Lu. 2020. The pro-

duction of within-family inequality: Insights and impli-

cations of integrating genetic data, BioRxiv. Available

at: https://doi.org/10.1101/2020.06.06.137778

Fletcher, J., M. Topping, F. Zheng, and Q. Lu. 2021.The

effects of education on cognition in older age:

Evidence from genotyped siblings, Social Science &

Medicine 280: 114044. https://doi.org/10.1016/j.

socscimed.2021.114044

Frijters, P., D. W. Johnston, M. Shah, and M. A. Shields.

2013. Intrahousehold resource allocation: Do parents

reduce or reinforce child ability gaps?, Demography

50(6): 2187–2208. https://doi.org/10.1007/s13524-013-

0224-2

Gage, T. B., F. Fang, E. O’Neill, and G. DiRienzo. 2013.

Maternal education, birth weight, and infant mortality

in the United States, Demography 50(2): 615–635.

https://doi.org/10.1007/s13524-012-0148-2

Paternal education and children’s old-age mortality 23

Gakidou, E., K. Cowling, R. Lozano, and C. J. Murray.

2010. Increased educational attainment and its effect

on child mortality in 175 countries between 1970 and

2009: A systematic analysis, The Lancet 376(9745):

959–974. https://doi.org/10.1016/S0140-6736(10)61257-3

Galama, T. J., A. Lleras-Muney, and H. van Kippersluis.

2018. The effect of education on health and mortality:

A review of experimental and quasi-experimental evi-

dence, in Oxford Research Encyclopedia of

Economics and Finance.https://doi.org/10.1093/

acrefore/9780190625979.013.7

Gathmann, C., H. Jürges, and S. Reinhold. 2015.

Compulsory schooling reforms, education and mor-

tality in twentieth century Europe, Social Science &

Medicine 127: 74–82. https://doi.org/10.1016/J.

SOCSCIMED.2014.01.037

Goldin, C. and L. F. Katz. 2010.The Race Between

Education and Technology. Harvard University Press.

https://doi.org/10.2307/j.ctvjf9x5x

Goldstein, J. R., M. Alexander, C. Breen, A. M. González,

F. Menares, M. Osborne, M. Snyder, and U. Yildirim.

2021. Censoc Project. CenSoc Mortality File: Version

2.0. Berkeley: University of California. Available at:

https://censoc.berkeley.edu/data/

Gonzalez, F. and T. Quast. 2015. Non-linearities in the

relationship between aggregate income and mortality

rates, Eastern Economic Journal 41(1): 51–69. https://

doi.org/10.1057/EEJ.2013.36

Goodman-Bacon, A. 2021. The long-run effects of child-

hood insurance coverage: Medicaid implementation,

adult health, and labor market outcomes, American

Economic Review 111(8): 2550–2593. https://doi.org/

10.1257/AER.20171671

Gough, H. G. 1946. The relationship of socioeconomic

status to personality inventory and achievement test

scores, Journal of Educational Psychology 37(9): 527–

540. https://doi.org/10.1037/H0055013

Grätz, M. and F. Torche. 2016. Compensation or reinforce-

ment? The stratification of parental responses to chil-

dren’s early ability, Demography 53(6): 1883–1904.

https://doi.org/10.1007/s13524-016-0527-1

Grimard, F. and D. Parent. 2007. Education and smoking:

Were Vietnam war draft avoiders also more likely to

avoid smoking?, Journal of Health Economics 26(5):

896–926. https://doi.org/10.1016/J.JHEALECO.2007.

03.004

Halpern-Manners, A., J. Helgertz, J. R. Warren, and E.

Roberts. 2020.The effects of education on mortality:

Evidence from linked U.S. census and administrative

mortality data, Demography 57(4): 1513–1541. https://

doi.org/10.1007/S13524-020-00892-6

Hamad, R., H. Elser, D. C. Tran, D. H. Rehkopf, and S. N.

Goodman. 2018. How and why studies disagree about

the effects of education on health: A systematic

review and meta-analysis of studies of compulsory

schooling laws, Social Science & Medicine 212: 168–

178. https://doi.org/10.1016/j.socscimed.2018.07.016

Hatt, L. E. and H. R. Waters. 2006. Determinants of child

morbidity in Latin America: A pooled analysis of inter-

actions between parental education and economic

status, Social Science & Medicine 62(2): 375–386.

https://doi.org/10.1016/j.socscimed.2005.06.007

Hayward, M. D. and B. K. Gorman. 2004. The long arm of

childhood: The influence of early-life social conditions

on men’s mortality, Demography 41(1): 87–107.

https://doi.org/10.1353/DEM.2004.0005

Hoekstra, C., Z. Z. Zhao, C. B. Lambalk, G. Willemsen, N.

G. Martin, D. I. Boomsma, and G. W. Montgomery. 2008.

Dizygotic twinning, Human Reproduction Update 14(1):

37–47. https://doi.org/10.1093/HUMUPD/DMM036

Hoynes, H., D. Miller, and D. Simon. 2015. Income, the

earned income tax credit, and infant health, American

Economic Journal: Economic Policy 7(1): 172–211.

https://doi.org/10.1257/pol.20120179

Hoynes, H., M. Page, and A. H. Stevens. 2011. Can tar-

geted transfers improve birth outcomes?. Evidence

from the introduction of the WIC program, Journal of

Public Economics 95(7-8): 813–827. https://doi.org/10.

1016/j.jpubeco.2010.12.006

Hoynes, H., D. W. Schanzenbach, and D. Almond. 2016.

Long-run impacts of childhood access to the safety

net, American Economic Review 106(4): 903–934.

https://doi.org/10.1257/aer.20130375

Huebener, M. 2019. Life expectancy and parental edu-

cation, Social Science & Medicine 232: 351–365.

https://doi.org/10.1016/J.SOCSCIMED.2019.04.034

Huebener, M. 2020. Parental education and children’s

health throughout life, The Economics of Education:

AComprehensive Overview:91–102. https://doi.org/10.

1016/B978-0-12-815391-8.00007-0

Huebener, M. 2021. The effects of education on health: An

intergenerational perspective, SSRN Electronic

Journal.https://doi.org/10.2139/SSRN.3249909

Jensen, A. C. and S. M. McHale. 2015. What makes siblings

different? The development of sibling differences in

academic achievement and interests, Journal of

Family Psychology 29(3): 469–478. https://doi.org/10.

1037/FAM0000090

Keats, A. 2018. Women’s schooling, fertility, and child

health outcomes: Evidence from Uganda’s free

primary education program, Journal of Development

Economics 135: 142–159. https://doi.org/10.1016/J.

JDEVECO.2018.07.002

Kemp, B. and J. K. Montez. 2020. Why does the impor-

tance of education for health differ across the United

States?, Socius: Sociological Research for a Dynamic

World 6: 237802311989954. https://doi.org/10.1177/

2378023119899545

24 Hamid Noghanibehambari and Jason Fletcher

Kim, Y. J. 2016. The long-run effect of education on obesity

in the US, Economics & Human Biology 21: 100–109.

https://doi.org/10.1016/J.EHB.2015.12.003

Kimbro, R. T., S. Bzostek, N. Goldman, and G. Rodríguez.

2008. Race, ethnicity, and the education gradient in

health, Health Affairs 27(2): 361–372. https://doi.org/

10.1377/HLTHAFF.27.2.361

Koning, P., D. Webbink, and N. G. Martin. 2015. The effect

of education on smoking behavior: New evidence from

smoking durations of a sample of twins, Empirical

Economics 48(4): 1479–1497. https://doi.org/10.1007/

S00181-014-0842-6

Kose, E., E. Kuka, and N. Shenhav. 2021. Women’s suf-

frage and children’s education, American Economic

Journal: Economic Policy 13(3): 374–405. https://doi.

org/10.1257/POL.20180677

Lazuka, V. 2019. Early-life assets in oldest-old age:

Evidence from primary care reform in early twentieth

century Sweden, Demography 56(2): 679–706. https://

doi.org/10.1007/s13524-018-0758-4

Lee, C. and C. D. Ryff. 2019. Pathways linking combi-

nations of early-life adversities to adult mortality:

Tales that vary by gender, Social Science & Medicine

240: 112566. https://doi.org/10.1016/J.SOCSCIMED.

2019.112566

Lefèbvre, M., P. Pestieau, and G. Ponthiere. 2019.

Premature mortality and poverty measurement in an

OLG economy, Journal of Population Economics 32

(2): 621–664. https://doi.org/10.1007/S00148-018-0688-X

Leon, D. A., L. Saburova, S. Tomkins, E. Andreev, N.

Kiryanov, M. McKee, and V. M. Shkolnikov. 2007.

Hazardous alcohol drinking and premature mortality

in Russia: A population based case-control study, The

Lancet 369(9578): 2001–2009. https://doi.org/10.1016/

S0140-6736(07)60941-6

Leuven, E., E. Plug, and M. Rønning. 2016. Education and

cancer risk, Labour Economics 43: 106–121. https://doi.

org/10.1016/J.LABECO.2016.06.006

Levy, J. 1919. Reduction of infant mortality by economic

adjustment and by health education, American

Journal of Public Health 9(9): 676–681. https://doi.org/

10.2105/AJPH.9.9.676

Lindahl, M. 2005. Estimating the effect of income on

health and mortality using lottery prizes as an exogen-

ous source of variation in income, Journal of Human

Resources XL(1): 144–168. https://doi.org/10.3368/

JHR.XL.1.144

Lindeboom, M., A. Llena-Nozal, and B. van der Klaauw.

2009. Parental education and child health: Evidence

from a schooling reform, Journal of Health Economics

28(1): 109–131. https://doi.org/10.1016/J.JHEALECO.

2008.08.003

Lleras-Muney, A. 2005. The relationship between edu-

cation and adult mortality in the United States,

Review of Economic Studies 72(1): 189–221. https://

doi.org/10.1111/0034-6527.00329

Lleras-Muney, A. 2022. Education and income gradients

in longevity: The role of policy, Canadian Journal of

Economics/Revue Canadienne d’économique 55(1): 5–

37. https://doi.org/10.1111/CAJE.12582

Lleras-Muney, A., J. Price, and D. Yue. 2022. The associ-

ation between educational attainment and longevity

using individual-level data from the 1940 census,

Journal of Health Economics 84(7): 102649. https://

doi.org/10.1016/j.jhealeco.2022.102649

Lundborg, P., A. Nilsson, and D.-O. Rooth. 2014. Parental

education and offspring outcomes: Evidence from the

Swedish compulsory school reform, American

Economic Journal: Applied Economics 6(1): 253–278.

https://doi.org/10.1257/APP.6.1.253

Lundborg, P., M. Nordin, and D. O. Rooth. 2018.Theinter-

generational transmission of human capital: The role of

skills and health, Journal of Population Economics 31

(4): 1035–1065. https://doi.org/10.1007/s00148-018-0702-3

Mangold, G. B. 1920.Problems of Child Welfare. Norwood,

MA: Macmillan Press.

Maralani, V. 2013. Educational inequalities in smoking:

The role of initiation versus quitting, Social Science &

Medicine 84: 129–137. https://doi.org/10.1016/J.

SOCSCIMED.2013.01.007

Maruyama, S. and E. Heinesen. 2020.Another look at

returns to birthweight, Jour nal of H e alt h E con omic s

70: 102269. https://doi.org/10.1016/j.jhealeco.2019.102269

Mazumder, B. 2008. Does education improve health? A

reexamination of the evidence from compulsory

schooling laws, Economic Perspectives 32(Q II): 2–16.

https://ssrn.com/abstract=1134064

McCrary, J. and H. Royer. 2011. The effect of female edu-

cation on fertility and infant health: Evidence from

school entry policies using exact date of birth,

American Economic Review 101(1): 158–195. https://

doi.org/10.1257/aer.101.1.158

Meghir, C., M. Palme, and E. Simeonova. 2018. Education

and mortality: Evidence from a social experiment,

American Economic Journal: Applied Economics 10

(2): 234–256. https://doi.org/10.1257/APP.20150365

Montez, J. and M. D. Hayward. 2011. Early life conditions

and later life mortality, in International Handbook of

Adult Mortality Vol. 2, pp. 187–206. https://doi.org/10.

1007/978-90-481-9996-9_9

Myrskylä, M., N. K. Mehta, and V. W. Chang. 2013. Early

life exposure to the 1918 influenza pandemic and old-

age mortality by cause of death, American Journal of

Public Health 103(7): e83–e90. https://doi.org/10.2105/

AJPH.2012.301060

Noghanibehambari, H. 2022a. In utero exposure to natural

disasters and later-life mortality: Evidence from earth-

quakes in the early twentieth century, Social Science

Paternal education and children’s old-age mortality 25

& Medicine 307: 115189. https://doi.org/10.1016/J.

SOCSCIMED.2022.115189

Noghanibehambari, H. 2022b. Intergenerational health

effects of Medicaid, Economics & Human Biology 45:

101114. https://doi.org/10.1016/J.EHB.2022.101114

Noghanibehambari, H. 2022c. School finance reforms and

juvenile crime, American Law and Economics Review

24(1): 1–86. https://doi.org/10.1093/ALER/AHAC001

Noghanibehambari, H. and M. Engelman. 2022. Social

insurance programs and later-life mortality: Evidence

from new deal relief spending, Journal of Health

Economics 86: 102690. https://doi.org/10.1016/J.

JHEALECO.2022.102690

Noghanibehambari, H. and J. Fletcher. 2023. Childhood

exposure to birth registration laws and old-age mor-

tality, Health Economics 32(3): 735–743. https://doi.

org/10.1002/HEC.4643

Noghanibehambari, H. and M. Salari. 2020. Health

benefits of social insurance, Health Economics 29(12):

1813–1822. https://doi.org/10.1002/hec.4170

Noghanibehambari, H., M. Salari, and N. Tavassoli. 2022.

Maternal human capital and infants’health outcomes:

Evidence from minimum dropout age policies in the

US, SSM - Population Health 19: 101163. https://doi.

org/10.1016/J.SSMPH.2022.101163

Oreopoulos, P., M. E. Page, and A. H. Stevens. 2006.The

intergenerational effects of compulsory schooling,

Journal of Labor Economics 24(4): 729–760. https://

doi.org/10.1086/506484/ASSET/IMAGES/LARGE/

FG2.JPEG

Pampel, F. 2005. Forecasting sex differences in mortality in

high income nations: The contribution of smoking,

Demographic Research 13(18): 455–484. https://doi.

org/10.4054/DEMRES.2005.13.18

Pronzato, C. 2012. An examination of paternal and

maternal intergenerational transmission of schooling,

Journal of Population Economics 25(2): 591–608.

https://doi.org/10.1007/S00148-010-0311-2

Reinhold, S. and H. Jürges. 2012. Parental income and

child health in Germany, Health Economics 21(5):

562–579. https://doi.org/10.1002/HEC.1732

Restrepo, B. J. 2016. Parental investment responses to a

low birth weight outcome: Who compensates and who

reinforces?, Journal of Population Economics 29(4):

969–989. https://doi.org/10.1007/s00148-016-0590-3

Rosales-Rueda, M. 2018. The impact of early life shocks on

human capital formation: Evidence from El Niño

floods in Ecuador, Journal of Health Economics 62:

13–44. https://doi.org/10.1016/J.JHEALECO.2018.07.

003

Ross, C. E. and J. Mirowsky. 2011. The interaction of per-

sonal and parental education on health, Social Science

& Medicine 72(4): 591–599. https://doi.org/10.1016/J.

SOCSCIMED.2010.11.028

Royer, H. 2009. Separated at girth: US twin estimates of

the effects of birth weight, American Economic

Journal: Applied Economics 1(1): 49–85. https://doi.

org/10.1257/app.1.1.49

Ruggles, S., S. Flood, R. Goeken, J. Grover, and E. Meyer.

2020. IPUMS USA: Version 10.0 [dataset].,

Minneapolis, MN: IPUMS. Available at: https://doi.

org/10.18128/D010.V10.0

Schmitz, L. L. and V. Duque. 2022. In utero exposure to

the Great Depression is reflected in late-life epigenetic

aging signatures, Proceedings of the National Academy

of Sciences 119(46). https://doi.org/10.1073/pnas.

2208530119

Scholte, R. S., G. J. Van Den Berg, and M. Lindeboom.

2015. Long-run effects of gestation during the Dutch

Hunger Winter famine on labor market and hospitaliz-

ation outcomes, Journal of Health Economics 39: 17–30.

https://doi.org/10.1016/J.JHEALECO.2014.10.002

Seeman, T., S. S. Merkin, E. Crimmins, B. Koretz, S.

Charette, and A. Karlamangla. 2008. Education,

income and ethnic differences in cumulative biological

risk profiles in a national sample of US adults:

NHANES III (1988–1994), Social Science & Medicine

66(1): 72–87. https://doi.org/10.1016/J.SOCSCIMED.

2007.08.027

Sheehan, C., J. K. Montez, and I. Sasson. 2018. Does the

functional form of the association between education

and mortality differ by U.S. region?, Biodemography

and Social Biology 64(1): 63–81. https://doi.org/10.

1080/19485565.2018.1468239

Shen, M. 2018. The effects of school desegregation on

infant health, Economics & Human Biology 30:

104–118. https://doi.org/10.1016/j.ehb.2018.06.002

Silles, M. A. 2011. The intergenerational effects of parental

schooling on the cognitive and non-cognitive develop-

ment of children, Economics of Education Review 30

(2): 258–268. https://doi.org/10.1016/J.ECONEDUREV.

2010.09.002

Silles, M. A. 2017. The intergenerational transmission of

education: New evidence from adoptions in the USA,

Economica 84(336): 748–778. https://doi.org/10.1111/

ECCA.12240

Snyder, S. E. and W. N. Evans. 2006. The effect of income

on mortality: Evidence from the social security Notch,

Review of Economics and Statistics 88(3): 482–495.

https://doi.org/10.1162/rest.88.3.482

Sohn, H. 2017. Medicaid’s lasting impressions: Population

health and insurance at birth, Social Science & Medicine

177: 205–212. https://doi.org/10.1016/j.socscimed.2017.

01.043

Strand, B. H. and A. Kunst. 2006a. Childhood socioeco-

nomic position and cause-specific mortality in early

adulthood, American Journal of Epidemiology 165(1):

85–93. https://doi.org/10.1093/aje/kwj352

26 Hamid Noghanibehambari and Jason Fletcher

Strand, B. H. and A. Kunst. 2006b. Childhood socioeco-

nomic status and suicide mortality in early adulthood

among Norwegian men and women. A prospective

study of Norwegians born between 1955 and 1965 fol-

lowed for suicide from 1990 to 2001, Social Science &

Medicine 63(11): 2825–2834. https://doi.org/10.1016/J.

SOCSCIMED.2006.07.020

Suhonen, T. and H. Karhunen. 2019. The intergenerational

effects of parental higher education: Evidence from

changes in university accessibility, Journal of Public

Economics 176: 195–217. https://doi.org/10.1016/J.

JPUBECO.2019.07.001

Tenn, S., D. A. Herman, and B. Wendling. 2010. The role of

education in the production of health: An empirical

analysis of smoking behavior, Journal of Health

Economics 29(3): 404–417. https://doi.org/10.1016/J.

JHEALECO.2010.03.001

Thomas, D., J. Strauss, and M. H. Henriques. 1990. Child

survival, height for age and household characteristics

in Brazil, Journal of Development Economics 33(2):

197–234. https://doi.org/10.1016/0304-3878(90)90022-4

Thomas, D., J. Strauss, and M.-H. Henriques. 1991. How

does mother’s education affect child height?, The

Journal of Human Resources 26(2): 183. https://doi.

org/10.2307/145920

Van Den Berg, G. J., G. Doblhammer, and K. Christensen.

2009. Exogenous determinants of early-life conditions,

and mortality later in life, Social Science & Medicine

68(9): 1591–1598. https://doi.org/10.1016/j.socscimed.

2009.02.007

Van Den Berg, G. J., G. Doblhammer-Reiter, and K.

Christensen. 2011. Being born under adverse economic

conditions leads to a higher cardiovascular mortality

rate later in life: Evidence based on individuals born at

different stages of the business cycle, Demography 48

(2): 507–530. https://doi.org/10.1007/s13524-011-0021-8

Van Den Berg, G. J., and S. Gupta. 2015. The role of mar-

riage in the causal pathway from economic conditions

early in life to mortality, Journal of Health Economics

40: 141–158. https://doi.org/10.1016/j.jhealeco.2014.02.

004

Van Den Berg, G. J., M. Lindeboom, and F. Portrait. 2006.

Economic conditions early in life and individual mor-

tality, American Economic Review 96(1): 290–302.

https://doi.org/10.1257/000282806776157740

van Kippersluis, H., O. O’Donnell, and E. van Doorslaer.

2011. Long-run returns to education, Journal of

Human Resources 46(4): 695–721. https://doi.org/10.

3368/jhr.46.4.695

Wherry, L. R. and B. D. Meyer. 2016. Saving teens: Using a

policy discontinuity to estimate the effects of medicaid

eligibility, Journal of Human Resources 51(3): 556–588.

https://doi.org/10.3368/jhr.51.3.0913-5918R1

Wherry, L. R., S. Miller, R. Kaestner, and B. D. Meyer.

2018. Childhood medicaid coverage and later-life

health care utilization, The Review of Economics and

Statistics 100(2): 287–302. https://doi.org/10.1162/

REST_A_00677

Wile, I. S. and R. M. Davis. 1939. Behavior differentials of

children with IQ’s 120 and above and IQ’s79and

below, with some reference to socio-economic status,

American Journal of Orthopsychiatry 9(3): 529–539.

https://doi.org/10.1111/J.1939-0025.1939.TB05622.X

Xu, W., M. Engelman, A. Palloni, and J. Fletcher. 2020.

Where and when: Sharpening the lens on geographic

disparities in mortality, SSM - Population Health 12:

100680. https://doi.org/10.1016/J.SSMPH.2020.100680

Xu, W., M. Topping, and J. Fletcher. 2021. State of birth

and cardiovascular disease mortality: Multilevel ana-

lyses of the National Longitudinal Mortality Study,

SSM - Population Health 15: 100875. https://doi.org/

10.1016/J.SSMPH.2021.100875

Xu, W., C. Wu, and J. Fletcher. 2020. Assessment of changes in

place of death of older adults who died from dementia in

the United States, 2000–2014: A time-series cross-sec-

tional analysis, BMC Public Health 20(1): 1–13. https://

doi.org/10.1186/S12889-020-08894-0

Yan, J. 2014. The effects of a minimum cigarette purchase

age of 21 on prenatal smoking and infant health,

Eastern Economic Journal 40(3): 289–308. https://doi.

org/10.1057/eej.2013.42

Zeng, Z. and Y. Xie. 2014. The effects of grandparents on

children’s schooling: Evidence from rural China,

Demography 51(2): 599–617. https://doi.org/10.1007/

S13524-013-0275-4

Zhou, D. and A. Dasgupta. 2017. Understanding the inter-

generational transmission of human capital: Evidence

from a quasi-natural experiment in China, Review of

Development Economics 21(2): 321–352. https://doi.

org/10.1111/RODE.12266

Paternal education and children’s old-age mortality 27

ResearchGate has not been able to resolve any citations for this publication.

Childhood exposure to birth registration laws and old‐age mortality

Article

Full-text available

Dec 2022
HEALTH ECON

This paper studies the effects of the enactment of birth registration laws, as the official universal and uniform method of recording births, across US states in the first decades of the 20th century on old‐age longevity for children affected by these laws. We show that establishing birth registration laws has long‐term benefits for old‐age health. The benefits are primarily driven by states with an effective child labor policy, suggesting that registering births helps the enforcement of child labor laws which in turn operate as the mechanism channel to improve old‐age longevity. A treatment‐on‐treated calculation suggests an increase of 0.6 years of longevity from not working during childhood due to the birth registration law.

In utero exposure to the Great Depression is reflected in late-life epigenetic aging signatures

Article

Full-text available

Nov 2022
P NATL ACAD SCI USA

Research on maternal-fetal epigenetic programming argues that adverse exposures to the intrauterine environment can have long-term effects on adult morbidity and mortality. However, causal research on epigenetic programming in humans at a population level is rare and is often unable to separate intrauterine effects from conditions in the postnatal period that may continue to impact child development. In this study, we used a quasi-natural experiment that leverages state-year variation in economic shocks during the Great Depression to examine the causal effect of environmental exposures in early life on late-life accelerated epigenetic aging for 832 participants in the US Health and Retirement Study (HRS). HRS is the first population-representative study to collect epigenome-wide DNA methylation data that has the sample size and geographic variation necessary to exploit quasi-random variation in state environments, which expands possibilities for causal research in epigenetics. Our findings suggest that exposure to changing economic conditions in the 1930s had lasting impacts on next-generation epigenetic aging signatures that were developed to predict mortality risk (GrimAge) and physiological decline (DunedinPoAm). We show that these effects are localized to the in utero period specifically as opposed to the preconception, postnatal, childhood, or early adolescent periods. After evaluating endogenous shifts in mortality and fertility related to Depression-era birth cohorts, we conclude that these effects likely represent lower bound estimates of the true impacts of the economic shock on long-term epigenetic aging.

Social Insurance Programs and Later-Life Mortality: Evidence from New Deal Relief Spending*

Article

Full-text available

Oct 2022
J HEALTH ECON

A growing body of research explores the long-run effects of social programs and welfare spending. However, evidence linking welfare support in early life with longevity is limited. We add to this literature by evaluating the effect of in-utero and early-life exposure to the largest increases in welfare spending in the US history under the New Deal programs. Using Social Security Administration death records linked with the 1940-census and spending data for 115 major cities, we show that the spending is correlated with improvements in old-age longevity. A treatment-on-treated calculation focused on a period when spending rose by approximately 1900 percent finds that a 100 percent rise in municipal spending in the year of birth is associated with roughly 3.5 months higher longevity. We show that these effects are not driven by endogenous selection of births, selective fertility, endogenous migration, and sample selection caused by endogenous data linking. Additional analysis suggests that rises in education and socioeconomic status are likely channels of impact.

Maternal human capital and infants’ health outcomes: Evidence from minimum dropout age policies in the US

Article

Full-text available

Sep 2022

The purpose of this cross-sectional study is to examine the causal relationship of maternal education and infants' health outcomes. Using birth certificate data over the years 1970–2004 and exploiting the space-time variation in Minimum Dropout Age laws to solve the endogeneity of education, we find a sizeable effect of mothers' education on their birth outcomes. An additional year of maternal education is associated with a reduction in incidences of low birth weight and preterm birth by 15.2 and 12.7 percent, respectively. The estimates are robust across various specifications and even when allowing mothers’ cohort-of-birth to vary across regions. The results suggest that the candidate mechanisms of impact include improvements in timing, quantity, and quality of prenatal care, lower negative health behavior during pregnancy such as smoking and drinking, and higher spousal education.

In utero exposure to natural disasters and later-life mortality: Evidence from earthquakes in the early twentieth century

Article

Full-text available

Jul 2022

Hamid Noghanibehambari

A growing body of research explores the effects of prenatal insults caused by natural disasters on life-cycle outcomes. This paper joins the literature by exploring the long-run effects of prenatal exposure to earthquakes on adulthood and old-age mortality. Using Social Security Administration death records (1975–2005) linked with the full-count 1940 US census and implementing a difference-in-difference methodology, I show that exposure during the first trimester is associated with 1.8 months lower longevity, conditional on survival up to age 34. This effect is equivalent to 22 percent of the white-nonwhite gap in the outcome. The results implicate the relevance of the in-utero period and specifically the onset of pregnancy for later-life health outcomes.

School Finance Reforms and Juvenile Crime

Article

Full-text available

May 2022

Hamid Noghanibehambari

Several states initiated school finance reforms during the post-1990s, commonly named the "adequacy" era, with the primary purpose of providing adequate funding for low-income school districts. This paper uses the space-time variation in court-ordered reforms in this period as shocks to school spending and investigates its effects on juvenile arrest rates and risky behaviors. Using a 2SLS-DDD approach and a wide range of datasets, I find that exposure to reform reduces the juvenile arrest rates, increases the likelihood of high school graduation, increases the time spent on educational activities, and reduces risky behaviors at schools. A 10% rise in real per-pupil spending is associated with 7.4 fewer arrests per 1,000 in the population aged 15-19. This rise is equivalent to a reduction of roughly 90,806 arrests annually. It also implies a minimum of 20% return in school spending due to the avoided costs of deterred crimes.

An Assessment of CenSoc Match Quality

Preprint

Jun 2022

The CenSoc datasets link individual-level 1940 Census records to Social Security death records using deterministic record linkage algorithms. In this technical report, we describe our record linkage methodology and assess the accuracy and representativeness of the CenSoc Version 2.1 matches.

The association between educational attainment and longevity using individual-level data from the 1940 census

Article

Jun 2022

We combine individual data from the 1940 full-count census with death records and other information available on the Family Tree at familysearch.org to create the largest individual dataset to date (17 million) to study the association between years of schooling and age at death. Conditional on surviving to age 35, one additional year of education is associated with roughly 0.4 more years of life for both men and women for cohorts born 1906-1915 and smaller for earlier cohorts. Focusing on the 1906-1915 cohort we find that this association is identical when we use sibling or twin fixed effects. This association varies substantially by place of birth. For men, the association is stronger in places with greater incomes, higher quality of school, and larger investments in public health. Women also exhibit great heterogeneity in the association, but our measures of the childhood environment do not explain it.

Education and income gradients in longevity: The role of policy

Article

Mar 2022

Adriana Lleras-Muney

Education and income are strong predictors of health and longevity. In the last 20 years, many efforts have been made to understand if these relationships are causal and what the possible role of policy should be as a result. The evidence from various studies is ambiguous: the effects of education and income policies on health are heterogeneous and vary over time and across places and populations. I discuss explanations for these disparate results and suggest directions for future research. Gradients d'éducation et de revenu dans la longévité : le rôle de la politique. L'éducation et le revenu sont de bons prédicteurs de la santé et de la longévité. Au cours des 20 dernières années, de nombreux efforts ont été déployés pour comprendre si ces relations sont causales et quel devrait être le rôle éventuel de la politique en conséquence. Les preuves issues de diverses études sont ambiguës. Les effets des politiques d'éducation et de revenu sur la santé sont hétérogènes et varient dans le temps, selon les lieux et les populations. J'examine diverses hypothèses susceptibles d'expliquer ces résultats hétérogènes et suggère de nouvelles avenues de recherche.

Intergenerational Health Effects of Medicaid

Article

Jan 2022
ECON HUM BIOL

Hamid Noghanibehambari

This paper investigates the effects of the introduction of Medicaid during the 1960s on next generations’ birth outcomes. A federal mandate that all states must widen the coverage to all cash welfare recipients generated cross-state var iations in Medicaid eligibility, specifically among nonwhites who largely overrepresented the target population. I implement a reduced-form difference-in-differences strategy that compares the birth outcomes of mothers born in states with higher cash welfare recipiency versus low welfare recipiency and different years relative to the Medicaid implementation year. Using Natality data (1970-2004), I find that Medicaid significantly improves birth outcomes. The effects are considerably larger among nonwhites, specifically blacks. The effects do not appear to be driven by preexisting trends in birth outcomes, preexisting trends in households’ socioeconomic characteristics, changes in other welfare expenditures, and selective fertility. A back-of-an-envelope calculation points to a minimum of 3.9 percent social externality of Medicaid through income rises due to next generations’ improvements in birth outcomes.

Unequal before death: The effect of paternal education on children's old-age mortality in the United States

Recommended publications

Long‐run intergenerational health benefits of women empowerment: Evidence from suffrage movements in...

Water is Life, Clean Water Means Health: The Effect of Early-Life Exposure to the City-Wide Water Fi...

Long-Term Health Benefits of Occupational Licensing: Evidence from Midwifery Laws

The Siren Song of Cicadas: Early-Life Pesticide Exposure and Later-Life Male Mortality