human capital development before age five$ - princeton university
TRANSCRIPT
CHAPTER1515Human Capital Development beforeAge FiveI
Douglas Almond, Janet CurrieColumbia University
Contents1. Introduction 13162. Conceptual Framework 1322
2.1. Complementarity 13232.2. Fixed investments 1324
2.2.1. Remediation 13252.3. Responsive investments 1326
3. Methods 13283.1. Power 13333.2. Data constraints 1336
3.2.1. Leveraging existing datasets 13363.2.2. Improvements in the production of administrative data 13383.2.3. Additional issues 1339
4. Empirical Literature: Evidence of Long Term Consequences 13404.1. Prenatal environment 1341
4.1.1. Maternal health 13414.1.2. Economic shocks 13674.1.3. Air pollution 1367
4.2. Early childhood environment 13704.2.1. Infections 13714.2.2. Health status 13714.2.3. Home environment 13944.2.4. Toxic exposures 13964.2.5. Summary re: long term effects of fetal and early childhood environment 1396
5. Empirical Literature: Policy Responses 13965.1. Income enhancement 13975.2. Near-cash programs 14055.3. Early intervention programs 1419
5.3.1. Home visiting 14255.3.2. US supplemental feeding program for women, infants, and children (WIC) 14265.3.3. Child care 14335.3.4. Health insurance 1453
I We thank Maya Rossin and David Munroe for excellent research assistance, participants in the Berkeley Handbookof Labor Economics Conference in November 2009 for helpful comments, and Christine Pal and Hongyan Zhao forproofreading the equations.
E-mail addresses: [email protected] (Douglas Almond), [email protected] (Janet Currie).
Handbook of Labor Economics, Volume 4b ISSN 0169-7218, DOI 10.1016/S0169-7218(11)02413-0c© 2010 Elsevier Ltd. All rights reserved. 1315
1316 Douglas Almond and Janet Currie
6. Discussion and Conclusions 1467Appendix A 1468Appendix B 1471Appendix C 1472Appendix D 1475References 1476
AbstractThis chapter seeks to set out what economists have learned about the effects of early childhoodinfluences on later life outcomes, and about ameliorating the effects of negative influences. We beginwith a brief overview of the theory which illustrates that evidence of a causal relationship between ashock in early childhood and a future outcome says little about whether the relationship in question isbiological or immutable. We then survey recent work which shows that events before five years old canhave large long term impacts on adult outcomes. Child and family characteristics measured at schoolentry do as much to explain future outcomes as factors that labor economists have more traditionallyfocused on, such as years of education. Yet while children can be permanently damaged at this age,an important message is that the damage can often be remediated. We provide a brief overview ofevidence regarding the effectiveness of different types of policies to provide remediation.We concludewith a list of some of the many outstanding questions for future research.
JEL classification: I12; I21; J13; J24; Q53
Keywords: Human capital; Early childhood; Health; Fetal origins
1. INTRODUCTION
The last decade has seen a blossoming of research on the long term effects of earlychildhood conditions across a range of disciplines. In economics, the focus is on howhuman capital accumulation responds to the early childhood environment. In 2000,there were no articles on this topic in the Journal of Political Economy, Quarterly Journalof Economics, or the American Economic Review (excluding the Papers and Proceedings),but there have been five or six per year in these journals since 2005. This work hasbeen spurred by a growing realization that early life conditions can have persistentand profound impacts on later life. Table 1 summarizes several longitudinal studieswhich suggest that characteristics that are measured as of age 7 can explain a greatdeal of the variation in educational attainment, earnings as of the early 30s, and theprobability of employment. For example, McLeod and Kaiser (2004) use data fromthe National Longitudinal Surveys and find that children’s test scores and backgroundvariables measured as of ages 6 to 8 predict about 12% of the variation in the probabilityof high school completion and about 11% of the variation in the probability of collegecompletion. Currie and Thomas (1999b) use data from the 1958 British Birth Cohortstudy and find that 4% to 5% of the variation in employment at age 33 can be predicted,and as much as 20% of the variation in wages. Cunha and Heckman (2008) and Cunhaet al. (2010) estimate structural models in which initial endowments and investments feed
Human Capital Development before Age Five 1317
Table1
How
muchof
thediffe
rences
inlatero
utcomes
(testscores,ed
ucationa
lattainm
ent,earnings)can
beexplaine
dby
early
child
hood
factors?
Stud
yan
dda
taInpu
ts/in
term
ediate
variab
les
Outco
mes
Results
Stud
iesus
ingNationa
lLon
gitudina
lSurve
yof
Youth-ch
ildSa
mpleData
Chi
ldho
odem
otio
nala
ndbe
havi
oral
prob
lem
sand
educ
atio
nal
atta
inm
ent
(McL
eod
and
Kai
ser,
2004
)N
LSY-
child
sam
ple
data
onch
ildre
n6-
8in
1986
.Fi
vew
aves
thro
ugh
2000
(whe
nch
ildre
n20
-22)
n=
424.
Em
otio
nala
ndbe
havi
oral
prob
lem
sata
ge6-
8m
easu
red
byB
PI,m
othe
rem
otio
nal
prob
lem
sand
delin
quen
cy,
pove
rty
stat
us,m
othe
r’sA
FQT
scor
e,m
othe
r’sed
ucat
ion,
mot
her’s
mar
itals
tatu
sand
age,
child
’sag
e,se
x,an
dra
ce,
dum
my
for
LBW
.
Dum
my
for
grad
uatin
ghi
ghsc
hool
by20
00,
dum
my
for
enro
lling
inco
llege
by20
00.
R-s
quar
edfo
rpre
dic
ting
HS
gra
duat
ion
:O
nly
child
emot
iona
land
beha
vior
alpr
oble
ms:
0.04
6A
ddch
ildan
dm
othe
rde
mog
raph
ics:
0.11
1A
ddm
othe
r’sem
otio
nalp
robl
emsa
ndde
linqu
ency
:0.
124.
R-s
quar
edfo
rpre
dic
ting
colle
geen
rollm
ent:
Onl
ych
ildem
otio
nala
ndbe
havi
oral
prob
lem
s:0.
017
Add
child
and
mot
her
dem
ogra
phic
s:0.
093
Add
mot
her’s
emot
iona
lpro
blem
sand
delin
quen
cy:0.
112.
(con
tinue
don
next
page
)
1318 Douglas Almond and Janet Currie
Table1(con
tinue
d)Stud
yan
dda
taInpu
ts/in
term
ediate
variab
les
Outco
mes
Results
Form
ulat
ing,
iden
tifyi
ng,an
des
timat
ing
the
tech
nolo
gyof
cogn
itive
and
nonc
ogni
tive
skill
form
atio
n(C
unha
and
Hec
kman
,20
08)N
LSY-
child
sam
ple
data
onw
hite
mal
es.
n=
1053
.
Hom
een
viro
nmen
tmea
sure
dby
HO
ME
scor
e.M
easu
reso
fpa
rent
alin
vest
men
ts:nu
mbe
rof
child
’sbo
oks,
whe
ther
the
child
hasa
mus
ical
inst
rum
ent,
whe
ther
the
fam
ilyre
ceiv
esa
daily
new
spap
er,w
heth
erth
ech
ildre
ceiv
essp
ecia
lles
sons
,ho
wof
ten
the
child
goes
tom
useu
ms,
and
how
ofte
nth
ech
ildgo
esto
the
thea
ter.
Mea
sure
sofc
hild
cogn
itive
skill
s:PI
AT
test
scor
esat
vari
ous
ages
.M
easu
reso
fchi
ld’s
non-
cogn
itive
skill
s:B
PIat
vari
ousa
ges.
Log
earn
ings
and
likel
ihoo
dof
grad
uatio
nfr
omhi
ghsc
hool
.E
stim
ated
ady
nam
icfa
ctor
mod
elth
atex
ploi
tscr
oss-
equa
tion
rest
rict
ions
for
iden
tifica
tion.
Est
imat
edeff
ects
ofco
gniti
vean
dno
n-co
gniti
vesk
illsa
sw
ella
spar
enta
lin
vest
men
tsat
differ
ent
stag
esof
the
child
’slif
ecyc
leon
log
earn
ings
and
likel
ihoo
dof
grad
uatio
nfr
omhi
ghsc
hool
.
A10
%in
crea
sein
pare
ntal
inve
stm
ents
atag
es6-
7in
crea
sese
arni
ngsb
y24
.9%
(12.
5%th
roug
hco
gniti
vesk
ills,
12.4
%th
roug
hno
n-co
gniti
vesk
ills)
,an
din
crea
sesl
ikel
ihoo
dof
grad
uatin
ghi
ghsc
hool
by64
.4%
(54.
8%th
roug
hco
gniti
vesk
ills,
9.6%
thro
ugh
non-
cogn
itive
skill
s).
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1319
Table1(con
tinue
d)Stud
yan
dda
taInpu
ts/in
term
ediate
variab
les
Outco
mes
Results
Est
imat
ing
the
tech
nolo
gyof
cogn
itive
and
non-
cogn
itive
skill
form
atio
n(C
unha
etal
.,20
10)
NLS
Y-ch
ildsa
mpl
efirs
t-bo
rnw
hite
child
ren.
N=
2207
.
Exa
mpl
esof
mea
sure
sofc
hild
cogn
itive
skill
s:M
otor
-soc
ial
deve
lopm
enta
tbir
th;PI
AT
read
ing-
com
preh
ensio
nat
ages
5-6;
PIA
Tm
ath
atag
es13
-14.
Exa
mpl
esof
mea
sure
sofc
hild
non-
cogn
itive
skill
s:B
PIat
vari
ousa
ges;
“fri
endl
ines
sat
birt
h”.E
xam
ples
ofm
easu
reso
fpa
rent
alin
vest
men
ts:“H
owof
ten
child
goes
onou
tings
duri
ngye
arof
birt
h”;“H
owof
ten
mom
read
sto
child
duri
ngye
arof
birt
h”;“C
hild
hasa
CD
play
er,ag
es3-
4”;H
owof
ten
child
ispr
aise
d,ag
es13
-14”
.
Mai
nou
tcom
eva
riab
le:
com
plet
edye
arso
fed
ucat
ion
byag
e19
.M
ulti-
stag
epr
oduc
tion
func
tions
estim
ated
inw
hich
the
prod
uctiv
ityof
late
rin
vest
men
tsde
pend
son
earl
yin
vest
men
tsan
don
the
stoc
kof
cogn
itive
and
non-
cogn
itive
skill
sand
inve
stm
ents
are
endo
geno
us.
Iden
tifica
tion
base
don
nonl
inea
rfa
ctor
mod
els
with
endo
geno
usin
puts
.In
pref
erre
dsp
ecifi
catio
n,us
edm
axim
um-l
ikel
ihoo
dm
etho
dsto
estim
ate
the
prod
uctio
nte
chno
logy
.
34%
ofva
riat
ion
ined
ucat
iona
latt
ainm
enti
sex
plai
ned
bym
easu
reso
fcog
nitiv
ean
dno
n-co
gniti
veca
pabi
litie
s.16
%is
due
toad
oles
cent
cogn
itive
capa
bilit
ies;
12%
isdu
eto
adol
esce
ntno
n-co
gniti
veca
pabi
litie
s.Pa
rent
alen
dow
men
ts/i
nves
tmen
tsac
coun
tfor
15%
ofva
riat
ion
ined
ucat
iona
latt
ainm
ent.
(con
tinue
don
next
page
)
1320 Douglas Almond and Janet Currie
Table1(con
tinue
d)Stud
yan
dda
taInpu
ts/in
term
ediate
variab
les
Outco
mes
Results
Stud
iesus
ingNationa
lChild
Dev
elop
men
tStudy
(195
8BritishBirthCo
hort)D
ata
Abi
lity,
fam
ily,
educ
atio
n,an
dea
rnin
gsin
Bri
tain
(Dea
rden
,19
98)
Focu
son
indi
vidu
alsw
hopa
rtic
ipat
edin
wav
es4
and
5of
the
surv
eyin
1981
and
1991
,w
how
ere
empl
oyee
sin
1991
.n=
2597
mal
es,
2362
fem
ales
.
Mat
han
dve
rbal
abili
ty(a
ge7)
,ty
peof
scho
ol,fa
mily
char
acte
rist
ics—
teac
her’s
asse
ssm
ento
fint
eres
tsho
wn
bypa
rent
sin
child
’sed
ucat
ion
at7;
type
ofsc
hool
atte
nded
at16
fam
ily’s
fina
ncia
lsta
tusa
t11
and
16;re
gion
dum
mie
s;fa
ther
’sSE
S;pa
rent
s’ed
ucat
ion
leve
ls.
Year
soff
ull-
time
educ
atio
n;E
arni
ngsa
tag
e33
.
R-s
quar
edfo
red
uca
tion
asoutc
om
e:In
clud
ing
alle
xpla
nato
ryva
riab
les:
0.33
-0.3
4.R
-squar
edfo
rea
rnin
gsas
outc
om
e:B
asel
ine
earn
ings
equa
tion
incl
udin
gon
lyye
ars
educ
atio
n:0.
15(m
ales
),0.
25(f
emal
es)
Add
read
ing
and
mat
hsc
ores
atag
e7,
scho
olty
pean
dre
gion
aldu
mm
ies:
0.26
(mal
es),
0.31
(fem
ales
)In
clud
ing
alle
xpla
nato
ryva
riab
les:.
029
(mal
es),
0.41
for
fem
ales
.
Ear
lyte
stsc
ores
,so
cioe
cono
mic
stat
us,an
dfu
ture
outc
omes
(Cur
rie
and
Tho
mas
,19
99b)
Rea
ding
and
mat
hte
stsc
ores
atag
e7,
mot
her
and
fath
er’s
SES
and
educ
atio
n,bi
rth
wei
ght,
othe
rch
ildba
ckgr
ound
vari
able
sat
age
7.
Num
ber
ofO
-lev
elpa
sses
ofex
amsb
yag
e16
;em
ploy
edat
age
23,
33;lo
gw
age
atag
e23
,33
.
R-s
quar
edfo
rpre
dic
ting
age
16ex
ampas
ses:
Rea
ding
and
mat
hsc
ores
only
:0.
21-0
.22
Add
othe
rba
ckgr
ound
vari
able
s:0.
31-0
.32.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1321
Table1(con
tinue
d)Stud
yan
dda
taInpu
ts/in
term
ediate
variab
les
Outco
mes
Results
Full
sam
ple
size
(bas
edon
resp
onse
sat
7):n=
14,02
2.
R-s
quar
edfo
rpre
dic
ting
emplo
ymen
tat
33:
Rea
ding
and
mat
hte
stsc
ores
only
:0.
01A
ddot
her
back
grou
ndva
riab
les:
0.04
-0.0
5.R
-squar
edfo
rpre
dic
ting
log
wag
eat
33:
Rea
ding
and
mat
hte
stsc
ores
only
:0.
08-0
.09
Add
othe
rba
ckgr
ound
vari
able
s:0.
18-0
.20.
The
last
ing
impa
ctof
child
hood
heal
than
dci
rcum
stan
ce(C
ase
etal
.,20
05)
n=
14,3
25(7
016
men
,70
39w
omen
)
Mot
her’s
and
fath
er’s
educ
atio
nan
dSE
S,LB
W,in
dica
tors
for
mod
erat
e,he
avy,
and
vari
edm
ater
nals
mok
ing
duri
ngpr
egna
ncy,
num
ber
ofch
roni
cco
nditi
onsa
tage
7an
d16
.
Num
ber
ofO
-lev
elpa
sses
ofex
amsb
yag
e16
;ad
ulth
ealth
stat
usat
age
42;pa
rt-t
ime
orfu
llem
ploy
men
tata
ge42
.
R-s
quar
edfo
rpre
dic
ting
age
16ex
ampas
ses/
adult
hea
lth/e
mplo
ymen
tat
42.
Mot
her’s
educ
atio
nan
dSE
S:0.
062/
0.08
2/0.
076
Fath
er’s
educ
atio
nan
dSE
S:0.
241/
0.18
9/0.
173
LBW
and
mat
erna
lsm
okin
g:0.
024/
0.08
6/0.
052.
Exp
lain
ing
inte
rgen
erat
iona
lin
com
epe
rsist
ence
:no
n-co
gniti
vesk
ills,
abili
ty,an
ded
ucat
ion
(Bla
nden
etal
.,20
06)
A.N
CD
S:19
58co
hort
Focu
son
2163
mal
es.
B.B
ritis
hC
ohor
tSt
udy:
1970
coho
rtFo
cuso
n33
40m
ales
.
A.Fa
mily
inco
me
atag
e16
read
ing
and
mat
hte
stsc
ores
atag
e11
,sc
ores
for
“beh
avio
ral
synd
rom
es”
atag
e11
,O
-lev
elex
amsc
ores
atag
e16
.B
.Ye
arso
fpre
scho
oled
ucat
ion,
birt
hw
eigh
t,he
ight
at5
and
10,
emot
iona
l/be
havi
oral
scor
esat
ages
5,10
,&
16,fa
mily
inco
me
atag
es10
and
16,re
adin
gan
dm
ath
test
scor
esat
age
10,IQ
atag
e10
,du
mm
yfo
rH
Sde
gree
,ex
amsc
ores
atag
e16
.
A.E
arni
ngsa
tage
33.
B.E
arni
ngsa
tage
30.
R-s
quar
edM
odel
A.B
irth
wei
ght,
child
hood
heal
th,an
dag
e11
test
scor
eson
ly:0.
116
Incl
udin
g“b
ehav
iora
lsyn
drom
es”
atag
e11
:0.
151
All
vari
able
s:0.
263.
R-s
quar
edM
odel
B.B
irth
wei
ght,
child
hood
heal
th,an
dag
e10
test
scor
eson
ly:0.
075.
Incl
udin
gem
otio
nal/
beha
vior
alch
arac
teri
stic
sat
age
10:0.
087.
All
vari
able
s:0.
222.
1322 Douglas Almond and Janet Currie
through to later outcomes; they arrive at estimates that are of a similar order of magnitudefor education and wages. To put these results in context, labor economists generally feelthat they are doing well if they can explain 30% of the variation in wages in a humancapital earnings function.
This chapter seeks to set out what economists have learned about the importanceof early childhood influences on later life outcomes, and about ameliorating the effectsof negative influences. We begin with a brief overview of the theory which illustratesthat evidence of a causal relationship between a shock in early childhood and a futureoutcome says little about whether the relationship in question is biological or immutable.Parental and social responses are likely to be extremely important in either magnifying ormitigating the effects of a shock. Given that this is the case, it can sometimes be difficult tointerpret the wealth of empirical evidence that is accumulating in terms of an underlyingstructural framework.
The theoretical framework is laid out in Section 2 and followed by a brief discussionof methods in Section 3. We do not attempt to cover issues such as identification andinstrumental variables methods, which are covered in some depth elsewhere (cf Angristand Pischke (2009)). Instead, we focus on several issues that come up frequently in theearly influences literature, including estimation using small samples and the potentiallyhigh return to better data.
Section 4 discusses the evidence for long-term effects of early life influences in greaterdetail, while Section 5 focuses on the evidence regarding remediation programs. Thediscussion of early life influences is divided into two subsections corresponding to in uteroinfluences and after birth influences. The discussion of remediation programs starts fromthe most general sort of program, income transfers, and goes on to discuss interventionsthat are increasingly targeted at specific domains. In surveying the evidence we haveattempted to focus on recent papers, and especially those that propose a plausible strategyfor identifying causal effects. We have focused on papers that emphasize early childhood,but in instances in which only evidence regarding effects on older children is available,we have sometimes strayed from this rule. A summary of most of the papers discussed inthese sections is presented in Tables 4 through 13. A list of acronyms used in the tablesappears in Appendix A. We conclude with a summary and a discussion of outstandingquestions for future research in Section 6.
2. CONCEPTUAL FRAMEWORKGrossman (1972) models health as a stock variable that varies over time in response toinvestments and depreciation. Because some positive portion of the previous period’shealth stock vanishes in each period (e.g., age in years), the effect of the health stock andhealth investments further removed in time from the current period tends to fade out. Asindividuals age, the early childhood health stock and the prior health investments that itembodies become progressively less important.
Human Capital Development before Age Five 1323
In contrast, the “early influences” literature asks whether health and investments inearly childhood have sustained effects on adult outcomes. The magnitude of these effectsmay persist or even increase as individuals age because childhood development occurs indistinct stages that are more or less influential of adult outcomes.
Defining h as health or human capital at the completion of childhood, we can retainthe linearity of h in investments and the prior health stock as in Grossman (1972), butleave open whether there is indeed “fade out” (i.e. depreciation). For simplicity, we willconsider a simple two-period childhood.1 We can consider production of h:
h = A[γ I1 + (1− γ )I2], (1)
where: {I1 ∼= investments during childhood through age 5I2 ∼= investments during childhood after age 5.
For a given level of total investments I1 + I2, the allocation of investments betweenperiod 1 and 2 will affect h for γ 6= 0.5. If γ > 0.5, then health at the end of period1 is more important to h than investments in the second period, and if γ A > 1, h mayrespond more than one-for-one with I1. Thus, (1) admits the possibility that certainchildhood periods may exert a disproportionate effect on adult outcomes that does notnecessarily decline monotonically with age. This functional form says more than “earlylife” matters; it suggests that early-childhood events may be more influential than laterchildhood events.
2.1. ComplementarityThe assumption that inputs at different stages of childhood have linear effects is commonin economics. While it opens the door to “early origins,” perfect substitutability betweenfirst and second period investments in (1) is a strong assumption. The absence ofcomplementarity implies that all investments should be concentrated in one period(up to a discount factor) and no investments should be made during the low-returnperiod. In addition, with basic preference assumptions, perfect substitutability “hard-wires” the optimal investment response to early-life shocks to be compensatory, as seenin Section 2.3.
As suggested by Heckman (2007), a more flexible “developmental” technology is theconstant elasticity of substitution (CES) function:
h = A[γ Iφ1 + (1− γ )I
φ
2
]1/φ, (2)
1 See Zweifel et al. (2009) for a two period version of the Grossman (1972) model.
1324 Douglas Almond and Janet Currie
For a given total investment level I1+ I2, how the allocation between period 1 and 2 willalso affect h depends on the elasticity of substitution, 1/(1−φ),and the share parameter,γ . For φ = 1 (perfect substitutability of investments), (2) reduces to (1).
Heckman (2007) highlights two features of “capacity formation” beyond thosecaptured in (2). First, there may be “dynamic complementarities” which imply thatinvestments in period t are more productive when there is a high level of capability inperiod t − 1. For example, if the factor productivity term A in (2) were an increasingfunction of h0, the health endowment immediately prior to period 1, this would raisethe return to investments during childhood. Second, there may be “self-productivity”which implies that higher levels of capacity in one period create higher levels of capacityin future periods. This feature is especially noteworthy when h is multidimensional, asit would imply that “cross-effects” are positive, e.g. health in period 1 leads to highercognitive ability in period 2. “Self-productivity” is more trivial in the unidimensionalcase like Grossman (1972)—even though the effect of earlier health stocks tends to fadeout as the time passes, there is still memory as long as depreciation in each period is lessthan total (i.e., when δ < 1 in Zweifel et al. (2009)).
Here, we will use the basic framework in (2) to consider the effect of exogenousshocks µg to health investments that occur during the first childhood period.2 Webegin with the simplest case, where investments do not respond to µg (and denote theseinvestments I1 and I2). Net investments in the first period are:
I1 + µg.
We assume thatµg is independent of I1. Whileµg can be positive or negative, we assumeI1 + µg > 0. We will then relax the assumption of fixed investments, and considerendogenous responses to investments in the second period, i.e. δ I ∗2 /δµg, and how thisinvestment response may mediate the observed effect on h.
2.2. Fixed investmentsConceptually, we can trace out the effect of µg while holding other inputs fixed, i.e., weassume no investment response to this shock in either period. Albeit implicitly, mostbiomedical and epidemiological studies in the “early origins” literature aim to inform usabout this ceteris paribus, “biological” relationship.
In this two-period CES production function adopted from Heckman (2007), theimpact of an early-life shock on adult outcomes is:
δh
δµg= γ A
[γ ( I1 + µg)
φ+ (1− γ ) Iφ2
](1−φ)/φ( I1 + µg)
φ−1. (3)
2 We include the subscript here because environmental influences at some aggregated geographic level g may provideexogenous variation in early childhood investments.
Human Capital Development before Age Five 1325
The simplest production technology is the perfect substitutability case where φ = 1.In this case:
δh
δµg= γ A.
Damage to adult human capital is proportional to the share parameter on period 1investments, and is unrelated to the investment level I1.
For less than perfect substitutability between periods, there is diminishing marginalproductivity of the investment inputs. Thus, shocks experienced at different baselineinvestment levels have heterogenous effects on h. Other things equal, those with higherbaseline levels of investment will experience more muted effects in h than those wherebaseline investment is low. A recurring empirical finding is that long-term damage dueto shocks is more likely among poorer families (Currie and Hyson, 1999). This is in partdue to the fact that children in poorer families are subject to more or larger early-lifeshocks (Case et al., 2002; Currie and Stabile, 2003). However, it is also possible that thesame shock will have a greater impact among children in poorer families if these childrenhave lower period t investment levels to begin with. This occurs because they are on asteeper portion of the production function. Ceteris paribus, this would tend to accentuatethe effect of an equivalent-sized µg shock on h among poor families.3,4
2.2.1. RemediationIs it possible to alter “bad” early trajectories? In other words, what is the effect of a shockµ′g > 0 experienced during the second period on h? Remediation is of interest to theextent that (3) is substantially less than zero. However, large damage to h from µg per sesays little about the potential effectiveness of remediation in the second period, as bothinitial damage and remediation are distinct functions of the three parameters A, γ , and φ.
The effectiveness of remediation relative to initial damage is:
δh/δµ′gδh/δµg
=1− γγ
(I1 + µg
I2 + µ′
g
)1−φ
. (4)
Thus, for I1 > I2 and a given value of γ , a unit of remediation will be more effectiveat low elasticities of substitution—the lack of I2 was the more critical shortfall priorto the shock. If I1 < I2 high elasticities of substitution increase the effectiveness ofremediation—adding to the existing abundance of I2 remains effective.
3 I.e. δ2h/δµgδ I1 < 0. On the other hand, δ2h/δµgδ I2 > 0 so lower period two investments would tend to reducedamage to h from µg . The ratio of the former effect to the latter is proportional to γ /(1 − γ ) (Chiang, 1984). Thus,damage from a period 1 shock is more likely to be concentrated among poor families when the period-1 share parameter(γ ) is high.
4 The cross-effect δ2h/δµgδ I2 is similar to dynamic complementarity, but Heckman (2007) reserves this term for thecross-partial between the stock and flow, i.e. δ2h/δh0δ It for t = 1, 2 in the example of Section 2.1.
1326 Douglas Almond and Janet Currie
Fortunately, it is not necessary to observe investments and estimate all threeparameters in order to assess the scope for remediation. In some cases, we merely need toobserve how a shock in the second period, µ′g, affects h. Furthermore, this does notnecessarily require a distinct shock in addition to µg. In an overlapping generationsframework, the same shock, µg = µ′g could affect one cohort in the first childhoodperiod (but not the second) and an older cohort in the second period (but not the first).For a small, “double-barreled” shock, we would have reduced form estimates of both thedamage in (3) and the potential to alter trajectories in (4).5 For example, in addition toobserving how income during the prenatal period affects newborn health (Kehrer andWolin, 1979), we might also be able to see how parental income affects the health of pre-school age children to gain a sense of what opportunities there are to remediate negativeincome shocks experienced during pregnancy.
2.3. Responsive investmentsMost analyses of “early origins” focus on estimating the reduced form effect, δh/δµg.Whether this empirical relationship represents a purely biological effect or also includesthe effect of responsive investments is an open question. In general, to the extent that“early origins” are important, so too will be any response of childhood investments toµg.For expositional purposes, we will consider µg < 0 and responses that either magnify orattenuate initial damage.
Unless the investment response is costless, damage estimates that monetize δh/δµg
alone will tend to understate total damage. In the extreme, investment responses couldfully offset the effect of early-life shocks on h, but this would not mean that such shockswere costless (Deschnes and Greenstone, 2007). More generally, the damage from early-life shocks will be understated if we focus only on long-term effects and there arecompensatory investments (i.e. investments that are negatively correlated with the early-life shock (δ I ∗2 /δµg < 0)). The cost of investments that help remediate damage shouldbe included. But even when the response is reinforcing (δ I ∗2 /δµg > 0), total costs canstill be understated by focussing on the reduced form damage to h alone (see below).
To consider correlated investment responses more formally, we assume parentsobserve µg at the end of the first period. The direction of the investment response—whether reinforcing or compensatory—will be shaped by how substitutable period 2investments are for those in period 1. If substitutability is high, the optimal response willtend to be compensatory, and thereby help offset damage to h.
A compensatory response is readily seen in the case of perfect substitutability. Cunhaand Heckman (2007) observed that economic models commonly assume that productionat different stages of childhood are perfect substitutes. When φ = 1, (2) reduces to:
h = A[γ (I1 + µg)+ (1− γ )I2
]. (5)
5 How parameters of the production function might be recovered is discussed in Appendix D.
Human Capital Development before Age Five 1327
This linear production technology is akin to that used in a previous Handbookchapter on intergenerational mobility (Solon, 1999), which likewise considered parentalinvestments in children’s human capital. Further, Solon (1999) assumed parent’s utilitytrades off own consumption against the child’s human capital:
Up = U (C, h), (6)
where p denotes parents and C their consumption. The budget constraint is:
Yp = C + I1 + I2/(1+ r). (7)
With standard preferences, changes to h through µg will “unbalance” the marginalutilities in h versus C .6 If µg is negative, the marginal utility of h becomes too highrelative to that in consumption. The technology in (5) permits parents to convert someconsumption into h at a constant rate. This will cause I ∗2 to increase, which attenuates theeffect of the µg damage. This attenuation comes at the cost of reduced parental utility.Similarly, if µg is positive parents will “spend the bounty” (at least in part), reduce I ∗2 andincrease consumption. Again, this will temper effects on h, leading to an understatementof biological effects in analyses that ignore investments (or parental utility). In either case,perfect substitutability hard-wires the response to be compensatory.
The polar opposite technology is perfect complementarity between childhood stages,i.e., a Leontieff production function. Here, a compensatory strategy would be completelyineffective in mitigating changes to h. As h is determined by the minimum of period1 and period 2 investments, optimal period 2 investments should reinforce µg. Ifµg is negative, parents would seek to reduce I2 and consume more. Despite higherconsumption, parents’ utility is reduced on net due to the shock (or this bundle of lowerh and higher C would have been selected absent µg). Again, the full-cost of a negativeµg shock is understated when parental utility is ignored.
The crossover between reinforcing and compensating responses of I ∗2 will occurat an intermediate parameter value of substitutability. (The fixed investments case ofSection 2.2 can be seen to reflect an optimized response at this point of balance betweenreinforcing and compensating responses). The value of φ at this point of balance willdepend on the functional form of parental preferences in (6), as shown for CES utility inAppendix B.
To take a familiar example, assume a Cobb-Douglas utility function of the form:
Up = (1− α)logC + αlog h. (8)
6 Obviously, the marginal utilities themselves will not be the same but equal subject to discount factor, preferenceparameters, and prices of C versus I , which have been ignored.
1328 Douglas Almond and Janet Currie
If the production technology is also Cobb-Douglas (φ = 0), then no change to I ∗2 iswarranted. If instead substitution between period 1 and period 2 is relatively easy (φ > 0),compensating for the shock is optimal. If substitution is relatively difficult (φ < 0), thenparents should “go with the flow” and reinforce. For this reason, whether conventionalreduced form analyses under or over-state “biological” effects (effects with I2 held fixed)depends on how easy it is to substitute the timing of investments across childhood. Ifthe elasticity of substitution across periods is low, then it may be optimal for parents toreinforce the effect of a shock.
Tension between preferences and the production technology may also be relevantfor within-family investment decisions. For example, Behrman et al. (1982) consideredparental preferences that parameterize varying degrees of “inequality aversion” among(multiple) children. Depending on the strength of parents’ inequality aversionrelative to the production technology (as reflected by φ), parents may reinforceor compensate exogenous within-family differences in early-life health and humancapital. If substitutability between periods of childhood is sufficiently difficult (lowφ), reinforcement of sibling differences will be optimal. This reinforcement may beoptimal even when the parents place a higher weight in their utility function on theaccumulation of human capital by the less able sibling (see Appendix C). Thus, empiricalevidence that some parents reinforce early-life shocks could reveal less about “humannature” than it would reveal about the developmental nature of the childhood productiontechnology.
3. METHODSAs discussed above, we confine our discussion to methodological issues that seemparticularly germane to the early influences literature. One of these is the question ofwhen sibling fixed effects (or maternal fixed effects) estimation is appropriate. Fixedeffects can be a powerful way to eliminate confounding from shared family backgroundcharacteristics, even when these are not fully observed. This approach is particularlyeffective when the direction of unobserved sibling-specific confounders can be signed.For example, Currie and Thomas (1995) find that in the cross-section, children whowere in Head Start do worse than children who were not. However, compared to theirown siblings, Head Start children do better. Since there is little evidence that Head Startchildren are “favored” by parents or otherwise (on the contrary, in families where onechild attends and the other does not, children who attended were more likely to havespent their preschool years in poverty), these contrasting results suggest that unobservedfamily characteristics are correlated both with Head Start attendance and poor childoutcomes. When the effect of such characteristics is accounted for, the positive effectsof Head Start are apparent.
However, fixed effects can not control for sibling-specific factors. The theorydiscussed above suggests that it may be optimal for parents to either reinforce or
Human Capital Development before Age Five 1329
compensate for the effects of early shocks by altering their own investment behaviors.Whether parents do or do not reinforce/compensate obviously has implications for theinterpretation of models estimated using family fixed effects. If on average, familiescompensate, then fixed effects estimates will understate the total effect of the shock(when the compensation behavior is unobserved or otherwise not accounted for). Insome circumstances, such a bias might be benign in the sense that any significantcoefficient could then be interpreted as a lower bound on the total effect. It is likelyto be more problematic if parents systematically reinforce shocks, because then any effectthat is observed results from a combination of underlying effects and parental reactionsrather than the shock itself. In the extreme, if parents seized on a characteristic that wasunrelated to ability and systematically favored children who had that characteristic, thenresearchers might wrongly conclude that the characteristic was in fact linked to successeven in the absence of parental responses.
The issue of how parents allocate resources between siblings has received a good dealof attention in economics, starting with Becker and Tomes (1976) and Behrman et al.(1982). Some empirical studies from developing countries find evidence of reinforcingbehavior (see Rosenzweig and Paul Schultz (1982), Rosenzweig and Wolpin (1988) andPitt et al. (1990)). Empirical tests of these theories in developed countries such as theUnited States and Britain generally use adult outcomes such as completed education as aproxy for parental investments (see for example, Griliches (1979), Behrman et al. (1994)and Ashenfelter and Rouse (1998)).
Several recent studies have used birth weight as a measure of the child’s endowmentand asked whether explicit measures of parental investments during early childhood arerelated to birth weight. For example, Datar et al. (2010) use data from the NationalLongitudinal Survey of Youth-Child and show that low birth weight children are lesslikely to be breastfed, have fewer well-baby visits, are less likely to be immunized, andare less likely to attend preschool than normal birth weight siblings. However, all of thesedifferences could be due to poorer health among the low birth weight children. Forexample, if a child is receiving many visits for sick care, they may receive fewer visits forwell care and this will not say anything about parental investment behaviors. Hence, Dataret al. (2010) also look at how the presence of low birth weight siblings in the householdaffects the investments received by normal birth weight children. They find no effect ofhaving a low birth weight sibling on breastfeeding, immunizations, or preschool. Theonly statistically significant interaction is for well-baby care. This could however, be dueto transactions costs. It may be the case that if the low birth weight sibling is getting a lotof medical care, it is less costly to bring the normal birth weight child in for care as well,for example.
Del Bono et al. (2008) estimate a model that allows endowments of other children toaffect parental investments in the index child. They find, however, that the results fromthis dynamic model are remarkably similar to those of mother fixed effects models in most
1330 Douglas Almond and Janet Currie
cases. Moreover, although they find a positive effect of birth weight on breastfeeding, theeffect is very small in magnitude.
We conducted our own investigation of this issue using data on twins from theEarly Childhood Longitudinal Study-Birth Cohort (ECLS-B), using twin differencesto control for potential confounders. At the same time, twins routinely have largedifferences in endowment in the form of birth weight. Table 2 presents estimates forall twins (with and without controls for gender), same sex twins, and identical twins.Overall, there are very few significant differences in the treatment of these twins: Parentsseemed to be more concerned about whether the low birth weight twin was ready forschool, and to delay introducing solid food (but this is only significant in the identicaltwin pairs). We see no evidence that parents are more likely to praise, caress, spank orotherwise treat children differently, and despite their worries about school readiness,parents have similar expectations regarding college for both twins. This table largelyreplicates the basic finding of Royer (2009), who also considered parental investmentsand birth weight differences in the ECLS-B data. In particular, Royer (2009) focussedon investments soon after birth, finding that breastfeeding, NICU admission, and othermeasures of neonatal medical care did not vary with within twin pair birth weightdifferences.
The parental investment response has also been explored in the context of naturalexperiments. Kelly (2009) asked whether observed parental investments (e.g., time spentreading to child) were related to flu-induced damages to test scores in the 1958 Britishbirth cohort study but did not detect an investment response.
In an interesting contribution to this literature, Hsin (2009) looks at the relationshipbetween children’s endowments, measured using birth weight, and maternal time useusing data from the Child Supplement of the Panel Study of Income Dynamics. Shefinds that overall, there is little relationship between low birth weight and maternaltime investments. However, she argues that this masks important differences by maternalsocioeconomic status. In particular, she finds that in models with maternal fixed effects,less educated women spend less time with their low birth weight children, while moreeducated women spend more time. This finding is based on only 65 sibling pairs whohad differences in the incidence of low birth weight, and so requires some corroboration.
Still, one interpretation of this result in the context of the Section 2 framework is that theelasticity of substitution between C and h varies by socioeconomic status. In particular,if ϕpoor > ϕrich, low income parents tend to view their consumption and children’s has relatively good substitutes. This would lead low-income parents to be more likely toreinforce a negative shock than high-income parents (assuming that the developmentaltechnology, captured by γ and φ, does not vary by socioeconomic status). A secondpossible interpretation of the finding is that parents’ responses may reflect their budgetconstraint more than their preferences. If parents would like to invest in both children,
but have only enough resources to invest adequately in one, then they may be forced
Human Capital Development before Age Five 1331
Table 2 Estimated effect of birth weight on parental investments within twin pairs, estimates fromthe early childhood longitudinal study.Outcome All twins Same sex
twinsIdenticaltwins
9month survey
1 if child was ever breastfed 0.0183 0.0187 0.0031[0.0238] [0.0277] [0.0355]1550 1000 350
1 if child is now being breastfed 0.0038 −0.0039 −0.0007[0.0126] [0.0152] [0.001]1550 1000 350
How long child was breastfed in months, −0.0753 −0.2165 −0.343given breastfed [0.1752] [0.204] [0.3182]
800 500 150Age solid food was introduced in months, −0.1802 −0.2478 −0.6660*given introduced [0.1523] [0.1906] [0.2914]
1550 1000 350Number of well-baby visits 0.283 0.3803 0.5797
[0.1883] [0.2414] [0.5253]1550 1000 350
Number of well-baby visits 0.1956 0.2329 0.2668only children in excellent or very good health [0.1624] [0.1944] [0.3799]
1500 950 3001 if caregiver praises child −0.0015 −0.051 0.096
[0.0941] [0.1189] [0.2089]1250 800 250
1 if caregiver avoids negative comments −0.0051 −0.0077 0[0.0055] [0.0084] [.]1250 800 250
1 if somewhat difficult or difficult to raise −0.0181 −0.0772 −0.0946(caregiver report) [0.0583] [0.0712] [0.1395]
1550 1000 3501 if not at all difficult or not very difficult to raise 0.1065 0.153 0.2237(caregiver report) [0.0707] [0.0812] [0.1195]
1550 1000 350
2-year survey
1 if caress/kiss/hug child 0.0228 0.0055 0.0021[0.0266] [0.0254] [0.0049]1350 850 300
(continued on next page)
1332 Douglas Almond and Janet Currie
Table 2 (continued)
Outcome All twins Same sextwins
Identicaltwins
1 if spank/slap child −0.0195 −0.0095 −0.0048[0.0249] [0.0192] [0.0316]1350 850 300
1 if time spent calming child > 1 hr usually 0.0317 −0.024 0.0719[0.0646] [0.0759] [0.093]1450 950 300
1 if somewhat difficult or difficult to raise −0.0432 −0.0901 −0.1412(caregiver report) [0.0555] [0.0621] [0.086]
1450 950 3001 if not at all difficult or not very difficult to raise(caregiver report)
−0.0031[0.0757]
0.068[0.0869]
0.0527[0.1258]
1450 950 300Age when stopped feeding formula in months −0.1903 −0.4504 −0.5903
[0.255] [0.3204] [0.7844]1150 750 250
Age when stopped breastfeeding in months −0.1492 −0.0267 −0.0422[0.5981] [0.044] [0.069]100 50 50
Preschool survey
1 if parent expects child to enter kindergarten early −0.0082[0.012]
−0.0071[0.0102]
0[0]
1300 800 2501 if parent concerned about −0.1435** −0.1299* −0.1099child’s kindergarten readiness [0.0554] [0.0636] [0.1253]
1300 850 2501 if expect child to get ≥ 4 yrs of college −0.0073 0.0069 0.0228
[0.0272] [0.0327] [0.0264]1350 850 300
Number of servings of milk in the past 7 days −0.0598 −0.0577 0.0819[0.2074] [0.2278] [0.2489]1350 850 300
Number of servings of vegetables past 7 days 0.0632 0.2131 0.0871[0.2634] [0.3027] [0.4091]1350 850 300
Standard errors clustered on the mother are shown in brackets with sample sizes below. Twin pairs in which a child hada congenital anomaly are omitted. Birth weight measured in kilograms. Each entry is from a separate regression of thedependent variable on birth weight and a mother fixed effect. Models in column 1 also control for child gender. Samplesizes are rounded to the nearest multiple of 50.Significance levels: *p < 0.10, **p < 0.05, ***p < 0.001.
Human Capital Development before Age Five 1333
to choose the more well endowed child.7 Interventions that relaxed resource constraintswould have quite different effects in this case than in the case in which parents preferredto maximize the welfare of a favored child. More empirical work on this question seemswarranted. For example, the PSID-CDS in 1997 and 2002 has time diary data for severalthousand sibling pairs which have not been analyzed for this purpose.
Parent’s choices are determined in part by the technologies they face, and thesetechnologies may change over time, with implications for the potential biases in fixedeffects estimates.8 For example, Currie and Hyson (1999) asked whether the long termeffects of low birth weight differed by various measures of parental socioeconomic statusin the 1958 British birth cohort. They found little evidence that they did (except thatlow birth weight women from higher SES backgrounds were less likely to suffer frompoor health as adults). But it is possible that this is because there were few effectiveinterventions for low birth weight infants in 1958. In contrast, Currie and Moretti (2007)looked at Californian mothers born in the late 1960s and 70s and find that women bornin low income zip codes were less educated and more likely to live in a low incomezipcode than sisters born in better circumstances. Moreover, women who were low birthweight were more likely to transmit low birth weight to their own children if they wereborn in low income zip codes, suggesting that early disadvantage compounded the initialeffects of low birth weight.
To the extent that behavioral responses to early-life shocks are important empirically,they will affect estimates of long-term effects whether family fixed effects are employedor not. Our conclusion is that users of fixed effects designs should consider any evidencethat may be available about individual child-level characteristics and whether parentsare reinforcing or compensating for the particular early childhood event at issue. Thisinformation will inform the appropriate interpretation of the estimates. There is littleevidence at present that parents in developed countries systematically reinforce orcompensate for early childhood events, but more research is needed on this question.
3.1. PowerGiven that there are relatively few data sets with information about early childhoodinfluences and future outcomes, economists may be tempted to make use of relativelysmall data sets that happen to have the requisite variables. Power calculations can behelpful in determining ex ante whether analysis of a particular data set is likely to yieldany interesting findings. Table 3 provides two sample calculations. The first half of thetable considers the relationship between birth weight and future educational attainment
7 In the siblings model of Appendix C, this can be seen in the case of a Leontief production technology, where the secondperiod investments generate increases in h only up to the level of first period investments. If parental income Y fallsbelow the cost of maintaining the initial investment level 2 I +µg in the second period, it may be optimal to invest fullyin child b (i.e. I∗2b = I ), but not in child a (i.e., I∗2a < I + µg ), who experiences the negative first-period shock.
8 For example, the effectiveness of remedial investments would change over time if γ varied with the birth cohort.Remediation would be more effective for later cohorts if γt > γt+1 in Eq. (4).
1334 Douglas Almond and Janet Currie
Table3
Samplepo
wer
calculations.
Given
atrue
popu
lation
effects
ize,wha
tisthepo
wer
ofasize
alph
a=0.05
test
agains
tthe
nullhy
pothesisthat
thereisno
effect
ford
ifferen
tsam
plesizes?
Basisstud
yAssum
ptions
Samplesize
Power
Bla
cket
al.(2
007)
Key
resu
lt:a
1%in
crea
sein
birt
hw
eigh
tin
crea
sest
hepr
obab
ility
ofhi
ghsc
hool
com
plet
ion
by0.
09pe
rcen
tage
poin
ts.
Bir
thw
eigh
tsam
ple
sum
mar
yst
ats(
twin
s):
mea
n=
2598
g,SD=
612
g.Pr
obab
ility
ofH
Sgr
adsa
mpl
esu
mm
ary
stat
s:m
ean=
0.73
,SD=
0.44
.
Tru
em
odel
:Pr
ob(H
SGR
AD
)=0.
7+
0.1∗
ln(b
irth
wei
ght)+
erro
rC
alcu
latio
nof
erro
rva
rian
cean
dSD
:Le
ty
=Pr
ob(H
SGR
AD
),x
=ln
(bir
thw
eigh
t),
e=
erro
rVa
r(y)
=0.
44ˆ2
=0.
19Va
r(x)
=0.
26ˆ2
=0.
07(w
here
SD(x)=
0.26
,ac
cord
ing
toth
edi
stri
butio
nof
ln(b
irth
wei
ght)
).If
y=
0.7+
0.1x+
e,an
dx
and
ear
ein
depe
nden
t,Va
r(e)
=Va
r(y)−
(0.1
ˆ2)∗
Var(
x)=
0.19−
(0.1
ˆ2)∗
(0.0
7)=
0.19
.So
,SD(e)=
sqrt(V
ar(e))=
0.44
The
refo
re,as
sum
e:bi
rthw
eigh
t∼N
(259
8,61
2),
and
take
the
natu
rall
ogof
birt
hwei
ght.
erro
r∼
N(0,0.
44).
100
300
500
600
700
800
900
1000
1250
1500
1620
2000
2200
2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
0.09
70.
167
0.26
30.
298
0.35
10.
376
0.40
90.
446
0.53
10.
617
0.66
00.
744
0.75
00.
825
0.89
20.
928
0.96
20.
975
0.98
20.
993
0.99
40.
996
0.99
9
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1335
Table3(con
tinue
d)
Given
asamplesize,h
owlargewou
ldthetrue
effects
izeha
veto
bein
orde
rtobe
able
tode
tect
itwithrelia
blepo
wer
usingatest
ofsize
alph
a=0.05
?
Basisstud
yAssum
ptions
True
B1Po
wer
Con
ley
etal
.(2
007)
Sibl
ing
sam
ple
from
PSID
(n=
1360
)M
odel
:y=
B0+
B1∗
x+
erro
rA
ssum
e:z∼
N(2
598,
612)
,x=
ln(z)
erro
r∼
N(0,0.
44)
sam
ple
size=
1500
0.00
50.
010.
020.
030.
040.
050.
060.
070.
080.
090.
10.
120.
150.
170.
2
0.04
60.
047
0.07
70.
092
0.14
60.
198
0.27
40.
354
0.46
10.
525
0.63
10.
769
0.92
60.
975
0.99
Pow
erca
lcul
atio
nsar
eba
sed
onM
onte
Car
losim
ulat
ions
with
1000
repl
icat
ions
.
1336 Douglas Almond and Janet Currie
as in Black et al. (2007). Their key result was that a 1% increase in birth weight increasedhigh school completion by 0.09 percentage points. The example shows that underreasonable assumptions about the distribution of birth weight and schooling attainment,it requires a sample of about 4000 children to be able to detect this effect in an OLSregression. We can also turn the question around and ask, given a sample of a certain size,
how large would an effect have to be before we could be reasonably certain of findingit in our data? The second half of the table shows that if we were looking for an effectof birth weight on a particular outcome in a sample of 1300 children, the coefficient on(the log of) birth weight would have to be at least 0.15 before we could detect it withreasonable confidence. If we have reason to believe that the effect is smaller, then it is notlikely to be useful to estimate the model without more data.
3.2. Data constraintsThe lack of large-scale longitudinal data (i.e. data that follows the same persons over time)has been a frequent obstacle to evaluating the long-term impacts of early life influences.Nevertheless, the answer may not always be to undertake collection of new longitudinaldata. Drawbacks include the high costs of data collection; the fact that long termoutcomes cannot be assessed for some time; and the fact that limiting sample attritionis particularly costly. Unchecked, attrition in longitudinal data can pose challenges forinference.
3.2.1. Leveraging existing datasetsIn many cases, existing cross-sectional microdata can serve as a platform for constructinglongitudinal datasets. First, it may be possible to add retrospective questions to ongoingdata collections. Second, it may be possible to merge new group-level informationto existing data sets. Third, it may be possible to merge administrative data sets byindividual in order to address previously unanswerable questions. The primary obstacleto implementing each of these data strategies is frequently data security. Depending onthe approach adopted, there are different demands on data security, as described below.
Smith (2009) and Garces et al. (2002) are examples of adding retrospective questionsto existing data collections. Smith had retrospective questions about health in childhoodadded to the Panel Study of Income Dynamics (PSID). The PSID began in the 1960swith a representative national sample, and has followed the original respondents and theirfamily members every since. Using these data, Smith (2009) is able to show that adultrespondents who were in poor health during childhood have lower earnings than theirown siblings who were not in poor health. Such comparisons are possible because thePSID has data on large numbers of sibling pairs. Garces et al. (2002) added retrospectivequestions about Head Start participation to the PSID, and were able to show that youngadults who had attended Head Start had higher educational attainment, and were lesslikely to have been booked or charged with a crime than siblings who had not attended.
Human Capital Development before Age Five 1337
While these approaches may enable analyses of long-term impacts even in the absenceof suitable “off the shelf ” longitudinal data, they have their drawbacks. First, retrospectivedata may be reported with error, although it may be possible to assess the extent ofreporting error using data from other sources. Second, only outcomes that are already inthe data can be assessed, so the need for serendipity remains. Still, the method is promisingenough to suggest that on-going, government funded data collections should build-inmechanisms whereby researchers can propose the addition of questions to subsequentwaves of the survey.
A second way to address long-term questions is to merge new information at thegroup level to existing data sets. The merge generally requires the use of geocoded data.For some purposes, such as exploring variations in policies across states, only a stateidentifier is required. For other purposes, such as examining the effects of traffic patternson asthma, ideally the researcher would have access to exact latitude and longitude.There are many examples in which this approach has been successfully employed. Forexample, Ludwig and Miller (2007) study the long term effects of Head Start, whichexploits the fact that the Office of Economic Opportunity initially offered the 300poorest counties in the country assistance in applying for Head Start. They show, usingdata from the National Educational Longitudinal Surveys, that children who were incounties just poor enough to be eligible for assistance were much more likely to haveattended Head Start than children in counties that were just ineligible. They go on toshow that child mortality rates in the relevant age ranges were lower in counties whoseHead Start enrollments were higher due to the OEO assistance. Using Census datathey find that education is higher for people living in areas with higher former HeadStart enrollment rates. Unfortunately, however, neither the decennial Census nor theAmerican Community Survey collect county of birth, so they cannot identify peoplewho were born in these counties (substantial measurement error is obviously introducedby using county of residence or county where someone went to school as a proxy forcounty of birth). An exciting crop of new research would be enabled by the additionof Census survey questions on county of birth, as well as county of residence at keydevelopmental ages (e.g., ages 5 and 14).9
In addition to the observational approaches described above, an intriguing possibilityis that participants in a completed randomized trial could be followed up. Forinstance, Rush et al. (1980) conducted a randomized intervention of a prenatal nutritionprogram in Harlem during the early 1970s. Following these children over time would
9 In another example, Currie and Gruber (1996a) were able to examine the effects of the Medicaid expansions on theutilization of care among children by merging state-level information on Medicaid policy to data from the NationalHealth Interview Survey (NHIS). At the time, this was only possible because one of the authors had access to the NHISstate codes through his work at the Treasury Department. It has since become easier to access geocoded health dataeither by traveling to Washington to work with the data, or by using it in one of the secure data centers that Censusand the National Center for Health Statistics (NCHS) support. However, it remains a source of frustration to healthresearchers that NCHS does not make state codes and/or codes for large counties available on its public use data sets.
1338 Douglas Almond and Janet Currie
allow researchers to evaluate cognitive outcomes in secondary school, and it might bepossible to collect retrospective data on parental investments during childhood, and toevaluate whether parental investments were affected by the randomization.
A third approach to leveraging existing data merges administrative records frommultiple sources at the individual level, which obviously requires personal identifiers suchas names and birth dates or social security numbers. Access to such identifiers is especiallysensitive. Nevertheless, it constitutes a powerful way to address many questions of interest.Several important studies have successfully exploited this approach outside of the US. Forexample, Black et al. (2005) and Black et al. (2007) use Norwegian data on all twins bornover 30 years to look at long-term effects of birth weight, birth order, and family sizeon educational attainment. Currie et al. (2010) use Canadian data on siblings to examinethe effects of health shocks in childhood on future educational attainment and welfareuse. Almond et al. (2009) use Swedish data to look at the long term effects of low-levelradiation exposure from the Chernobyl disaster on children’s educational attainment.
In the US, Doyle (2008) uses administrative data from child protective services andthe criminal justice system in Illinois to examine the effects of foster care. He shows firstthat there is considerable variation between foster care case workers in whether or nota child will be sent to foster care. Moreover, whether a child is assigned to a particularworker is random, depending on who is on duty at the time a call is received. Using thisvariation, Doyle shows that the marginal child assigned to foster care is significantly morelikely to be incarcerated in future. These examples exploit large sample sizes, objectiveindicators of outcomes, sibling or cohort comparisons, as well as a long follow up period.
Some limitations of using existing data include the fact that administrative data sets oftencontain relatively little background information, and that outcomes are limited to thosethat are collected in the data bases. Finally, the application process to obtain individual-matched data is often protracted.
Looking forward, the major challenge to research that involves either merging newinformation to existing data sets, or merging administrative data sets to each other, is thatprivacy concerns are making it increasingly difficult to obtain data just as it is becomingmore feasible to link them. In some cases, access to public use data has deteriorated.
For example, for many years, individual level Vital Statistics Natality data from birthcertificates included the state of birth, and the county (for counties with over 100,000population). Since 2005, however, these data elements have been suppressed and it isnow necessary to get special permission to obtain US Vital Statistics data with geocodes.
3.2.2. Improvements in the production of administrative dataThere are several “first best” potential solutions to these problems. First, creators oflarge data sets need to be sensitive to the fact that their data may well be useful foraddressing questions that they have not envisaged. In order to preserve the ability to usedata to answer future questions, it is essential to retain information that can be used for
Human Capital Development before Age Five 1339
linkage. At a minimum, this should include geographic identifiers at the smallest levelof disaggregation that is feasible (for example a Census tract). Ideally, personal identifierswould also be preserved.
Second, more effort needs to be expended in order to make sensitive data available toresearchers. A range of mechanisms exist that protect privacy while enabling research:
1. Suppress small cells or merge small cells in public use data files. For example, NCHSdata sets such as NHIS could be released with state identifiers for large states, and withidentifiers for groups of smaller states.
2. Add small amounts of “noise” to public use data sets, or do data swapping in orderto prevent identification of outliers. For example, Cornell University is coordinatingthe NSF-Census Bureau Synthetic Data Project which seeks to develop public-use“analytically valid synthetic data” from micro datasets customarily accessed at secureCensus Research Data Centers.
3. Create model servers. In this approach, users login to estimate models using the truedata, but get back output that does not allow individuals to be identified.
4. Data use agreements. The National Longitudinal Survey of Youth and the NationalEducational Longitudinal Survey have successfully employed data use agreementswith qualified users for many years, and without any documented instances of datadisclosure.
5. Creation of de-identified merged files. For example, Currie et al. (2009) asked thestate of New Jersey to merge birth records with information about the location ofpollution sources, and create a de-identified file. This allows them to study the effectof air pollution on infant health.
6. Secure data facilities. The Census Research Data Centers have facilitated access tomuch confidential data, although researchers who are not located close to the facilitiesmay still face large costs of accessing them.
These approaches to data dissemination have been explored in the statistics literaturefor more than 20 years (see Dalenius and Reiss (1982)), and have been much discussed atCensus (see for example, Reznek (2007)).
3.2.3. Additional issuesWe conclude with two new and relatively unexplored data issues. First, how caneconomists make effective use of the burgeoning literature on biomarkers? Thesemeasures have recently been added to existing health surveys, such as the NationalLongitudinal Study of Adolescent Health data. Biomarkers include not only informationabout genetic variations but also hormones such as cortisol (which is often interpretedas a measure of stress). It is tempting to think of these markers as potential instrumentalvariables (Fletcher and Lehrer, 2009). For instance, if it was known that a particular genewas linked to alcoholism, then one might think of using the gene as an instrument for
1340 Douglas Almond and Janet Currie
alcoholism. The potential pitfall in this approach is clear if we consider using somethinglike skin color as an instrument in a human capital earnings function–clearly, skin colormay predict educational attainment, but it may also have a direct effect on earnings. Justbecause a variable is “biological” does not mean that it satisfies the criteria for a validinstrument.
A second issue is the evolving nature of what constitutes a “birth cohort.”Improvements in neonatal medicine have meant that stillbirths and fetal deaths thatwould previously have been excluded from the Census of live births may be increasinglyimportant, e.g. MacDorman et al. (2005). Such a compositional effect on live births mayhave first-order implications for program evaluation and the long-term effects literature.Indeed, both the right and left tails of the birth weight distribution have elongated overtime—in 1970 there were many fewer live births with birthweight either less than 1500g or over 4000 g. To date there has been little research exploring the implications of thesecompositional changes.
In summary, there are many secrets currently locked in existing data that researchersdo not have access to. Economists have been skillful in navigating the many datachallenges inherent in the analysis of long-term (and sometimes latent) effects.Nevertheless, we need to explore ways to make more of these data available, and to moreresearchers. In many cases, this will be a more cost effective and timely way to answerimportant questions than carrying out new data collections.
4. EMPIRICAL LITERATURE: EVIDENCE OF LONG TERM CONSEQUENCES
What is of importance is the year of birth of the generation or group of individuals underconsideration. Each generation after the age of 5 years seems to carry along with it thesame relative mortality throughout adult life, and even into extreme old age.
Kermack et al. (1934) in The Lancet (emphasis added).
In this section, we summarize recent empirical research findings that experiencesbefore five have persistent effects, shaping human capital in particular. A hallmark of thiswork is the attention paid to identification strategies that seek to isolate causal effectsof the early childhood environment. An intriguing sub-current is the possibility thatsome of these effects may remain latent during childhood (at least from the researcher’sperspective) until manifested in either adolescence or adulthood. Recently, economistshave begun to ask how parents or other investors in human capital (e.g. school districts)respond to early-life shocks, as suggested by the conceptual framework in Section 2.3.
As the excerpt from Kermack et al. (1934) indicates, the idea that early childhoodexperiences may have important, persistent effects did not originate recently, nor didit first appear in economics. An extensive epidemiological literature has focussed onthe early childhood environment, nutrition in particular, and its relationship to health
Human Capital Development before Age Five 1341
outcomes in adulthood. For a recent survey, see Gluckman and Hanson (2006). Thisliterature has been criticized within epidemiology for credulous empirical comparisons(see, e.g. Rasmussen (2001) or editorial in The Lancet [2001]). Absent clearly-articulatedidentification strategies, health determinants that are difficult to observe and are thereforeomitted from the analysis (e.g., parental concern) are presumably correlated with thetreatment and can thereby generate the semblance of “fetal origins” linkages, even whensuch effects do not exist.
4.1. Prenatal environmentIn the 1990s, David J. Barker popularized and developed the argument that disruptionsto the prenatal environment presage chronic health conditions in adulthood, includingheart disease and diabetes (Barker, 1992). Growth is most rapid prenatally and in earlychildhood. When growth is rapid, disruptions to development caused by the adverseenvironmental conditions may exert life-long health effects. Barker’s “fetal origins”perspective contrasted with the view that pregnant mothers functioned as an effectivebuffer for the fetus against environmental insults.10
In Table 4, we categorize prenatal environmental exposures into three groups.Specifically, we differentiate among factors affecting maternal and thereby fetalhealth (e.g. nutrition and infection), economic shocks (e.g. recessions), and pollution(e.g. ambient lead).
4.1.1. Maternal healthCurrie and Hyson (1999) broke ground in economics by exploring whether “fetalorigins” (FO) effects were confined to chronic health conditions in adulthood, or mightextend to human capital measures. Using the British National Child DevelopmentSurvey, low birth weight children were more than 25% less likely to pass English andmath O-level tests, and were also less likely to be employed. The finding that testscores were substantially affected was surprising, as epidemiologists routinely posited fetal“brain sparing” mechanisms, whereby adverse in utero conditions were parried through aplacental triage that prioritized neural development over the body, see, e.g., Scherjonet al. (1996). Furthermore, Stein et al. (1975)’s influential study found no effect ofprenatal exposure to the Dutch Hunger Winter on IQ.
Currie and Hyson (1999) were followed by a series of papers that exploited differencesin birthweight among siblings and explored their relationship to sibling differences incompleted schooling. In relatively small samples (approximately 800 families), Conleyand Bennett (2001) found negative but imprecise effects of low birth weight oneducational attainment. Statistically significant effects of low birth weight on educationalattainment were found when birth weight was interacted with being poor, but in general
10 For example, it has been argued that nausea and vomiting in early pregnancy (morning sickness) is an adaptive responseto prevent maternal ingestion of foods that might be noxious to the fetus.
1342 Douglas Almond and Janet Currie
Table4
Pren
ataleff
ectson
laterchild
andad
ulto
utcomes.
Stud
yan
dda
taStud
yde
sign
Results
Effects
ofmaterna
lhea
lth
Isth
eim
pact
ofhe
alth
shoc
kscu
shio
ned
byec
onom
icst
atus
?T
heca
seof
low
birt
hw
eigh
t.C
urri
ean
dH
yson
(199
9)N
CD
S19
58co
hort
.N=
11,6
09at
age
20,10
,267
atag
e23
,94
02at
age
33.
Mul
tivar
iate
regr
essio
nw
ithnu
mer
ousb
ackg
roun
dan
dde
mog
raph
icco
ntro
ls(in
clud
ing
mat
erna
lgra
ndfa
ther
’sSE
S,bi
rth
orde
r,an
dm
ater
nals
mok
ing
duri
ngpr
egna
ncy)
.K
eyex
plan
ator
yva
riab
lesa
rein
dica
tors
for
LBW
,SE
S,an
dth
ein
tera
ctio
ns.
Out
com
esm
easu
red
are
the
num
ber
ofO
-lev
elpa
sses
atag
e16
(tra
nscr
ipts
colle
cted
atag
e20
),em
ploy
men
t,w
ages
and
heal
thst
atus
atag
es23
and
33.SE
Sas
signe
dus
ing
fath
er’s
soci
alcl
assi
n19
58(o
rm
othe
r’sSE
Sif
fath
eris
miss
ing)
.
LBW
child
ren
are
38-4
4%le
sslik
ely
topa
ssM
ath
O-l
evel
.LB
Wfe
mal
esar
e25
%le
sslik
ely
topa
ssE
nglis
hO
-lev
elte
sts.
LBW
fem
ales
are
16%
less
likel
yto
beem
ploy
edfu
ll-tim
eat
age
23,LB
Wm
ales
are
9%le
sslik
ely
tobe
empl
oyed
full-
time
atag
e33
.LB
Wfe
mal
esar
e54
%m
ore
likel
yto
have
fair
/poo
rhe
alth
atag
e23
,LB
Wm
ales
are
43%
mor
elik
ely
toha
vefa
ir/p
oor
heal
that
age
33.
Few
signi
fica
ntdi
ffer
ence
sby
SES.
Ret
urns
tobi
rthw
eigh
t(B
ehrm
anan
dR
osen
zwei
g,20
04).
Mon
ozyg
otic
fem
ale
twin
sbor
n19
36-1
955
from
the
Min
neso
taTw
insR
egist
ryfo
r19
94an
dth
ebi
rth
wei
ghts
ofth
eir
child
ren.
N=
804
twin
s,an
d60
8tw
in-m
othe
rpa
irs.
Twin
fixe
deff
ects
estim
ates
com
pare
dto
OLS
.To
expl
ore
gene
raliz
abili
tyof
find
ings
from
twin
ssam
ple,
wei
ghte
dth
esa
mpl
eus
ing
the
US
singl
eton
dist
ribu
tion
offe
talg
row
thra
tes.
Res
ultsfr
om
twin
ssa
mple
:1
oz.pe
rw
eek
ofpr
egna
ncy
incr
ease
infe
talg
row
thle
adst
oin
crea
seso
f5%
,in
scho
olin
gat
tain
men
t;2%
inhe
ight
;8%
hour
lyw
ages
.R
esultsfr
om
twin
ssa
mple
wei
ghte
dusi
ng
singl
eton
dis
trib
ution
:1
oz.pe
rw
eek
ofpr
egna
ncy
incr
ease
infe
talg
row
thle
adst
oin
crea
seof
5%in
scho
olin
gat
tain
men
t;1.
7%in
heig
ht;no
signi
fica
nteff
ecto
nw
ages
.O
LSun
dere
stim
ates
effec
tsof
birt
hw
eigh
tby
50%
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1343
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
cost
sofl
owbi
rth
wei
ght
(Alm
ond
etal
.,20
05).
Link
edbi
rth
and
infa
ntde
ath
file
sfor
US
for
1983
-85,
1989
-91
and
1995
-97.
Hos
pita
lcos
tsfr
omhe
alth
care
cost
and
utili
zatio
npr
ojec
tsta
tein
patie
ntda
taba
sefo
r19
95-2
000
inN
ewYo
rkan
dN
ewJe
rsey
.N
CH
SN=
189,
036
twin
s,49
7,13
9sin
glet
ons.
HC
UP
N=
44,4
10.
Twin
fixe
deff
ects
toes
timat
eeff
ect
oflo
wbi
rth
wei
ght(
LBW
)on
hosp
italc
osts
,he
alth
atbi
rth,
and
infa
ntm
orta
lity.
Also
estim
ated
impa
ctof
mat
erna
lsm
okin
gdu
ring
preg
nanc
yon
heal
tham
ong
singl
eton
birt
hs,co
ntro
lling
for
num
erou
sbac
kgro
und
and
dem
ogra
phic
char
acte
rist
icsu
sing
OLS
and
prop
ensit
ysc
ore
mat
chin
g.
Res
ultsusi
ng
twin
fixe
deff
ects
:1
SDin
crea
sein
birt
hw
eigh
tlea
dsto
:0.
08SD
decr
ease
inho
spita
lcos
tsfo
rde
liver
yan
din
itial
care
0.03
SDde
crea
sein
infa
ntm
orta
lity
rate
s0.
03SD
incr
ease
inA
pgar
scor
es0.
01SD
decr
ease
inus
eof
assis
ted
vent
ilato
raf
ter
birt
h(O
LSes
timat
esw
/out
twin
fixe
deff
ects
are
0.51
SD,0.
41SD
,0.
51SD
,0.
25SD
).R
esultsfr
om
OLS
on
effec
tsofm
ater
nal
smoki
ng:
Mat
erna
lsm
okin
gre
duce
sbir
thw
eigh
tby
200
g(6
%);
incr
ease
slik
elih
ood
that
infa
ntis
LBW
(<25
00g)
bym
ore
than
100%
(mea
n=
0.06
1).
No
stat
istic
ally
signi
fica
nteff
ects
onA
pgar
scor
e,in
fant
mor
talit
yra
tes,
orus
eof
assis
ted
vent
ilato
rat
birt
h.
The
1918
influe
nza
pand
emic
and
subs
eque
nthe
alth
outc
omes
(Alm
ond
and
Maz
umde
r,20
05).
Dat
afr
omSI
PPfo
r19
84-1
996.
N=
25,1
69.
Com
pare
coho
rtsi
nut
ero
befo
re,
duri
ngan
daf
ter
Oct
.19
18flu
pand
emic
.R
egre
ssio
nses
timat
eco
hort
effec
tsin
clud
ing
surv
eyye
ardu
mm
iesa
ndqu
adra
ticin
age
inte
ract
edw
ithsu
rvey
year
.
Indi
vidu
alsb
orn
in19
19ar
e10
%m
ore
likel
yto
bein
fair
orpo
orhe
alth
,al
soin
crea
seso
f19%
,35
%,13
%,17
%in
trou
ble
hear
ing,
spea
king
,lif
ting
and
wal
king
.
Est
imat
ing
the
impa
ctof
larg
eci
gare
tte
tax
hike
s:th
eca
seof
mat
erna
lsm
okin
gan
din
fant
birt
hw
eigh
t(Li
enan
dE
vans
,20
05).
1990
-199
7U
SN
atal
ityfile
s.D
ata
onci
gare
tte
taxe
sfro
mth
eta
xbu
rden
onto
bacc
o,va
riou
syea
rs.
IVus
ing
tax
hike
sin
four
stat
esas
inst
rum
ents
for
mat
erna
lsm
okin
gdu
ring
preg
nanc
y.C
ontr
olle
dfo
rst
ate
and
mon
thof
conc
eptio
nan
dba
ckgr
ound
char
acte
rist
ics.
Mat
erna
lsm
okin
gdu
ring
preg
nanc
yre
duce
sbir
thw
eigh
tby
5.4%
and
incr
ease
slik
elih
ood
oflo
wbi
rth
wei
ghtb
y∼
100%
.
(con
tinue
don
next
page
)
1344 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Long
term
effec
tsof
inut
ero
expo
sure
toth
e19
18in
flue
nza
pand
emic
inth
epo
st-1
940
US
popu
latio
n(A
lmon
d,20
06).
1960
-198
0U
SC
ensu
sdat
a.19
17-1
919
Vita
lSta
tistic
sdat
aon
mor
talit
y.Fo
r19
60n=
114,
031,
for
1970
n=
308,
785,
for
1980
n=
471,
803.
Est
imat
edeff
ects
ofin
uter
oin
flue
nza
expo
sure
byco
mpa
ring
coho
rtsb
orn
imm
edia
tely
befo
re,
duri
ng,an
daf
ter
the
1918
pand
emic
and
byem
ploy
ing
the
idio
sync
ratic
geog
raph
icva
riat
ion
inin
tens
ityof
expo
sure
toco
nduc
tw
ithin
-coh
orta
naly
sis.E
xpos
edco
hort=
thos
ebo
rnin
1919
.Su
rrou
ndin
gco
hort
s=th
ose
born
in19
18an
d19
20.U
sed
mul
tivar
iate
regr
essio
nw
ithdu
mm
yfo
rbi
rth
coho
rt=
1919
,an
da
quad
ratic
coho
rttr
end
tom
easu
rede
part
ures
inth
e19
19bi
rth
coho
rtou
tcom
esfr
omth
etr
end.
For
geog
raph
icco
mpa
riso
n,us
edda
taon
viru
sstr
engt
hby
wee
kas
wel
lasd
ata
onep
idem
ictim
ing
byC
ensu
sdiv
ision
toyi
eld
am
easu
reof
aver
age
pand
emic
viru
lenc
eby
divi
sion.
The
nes
timat
edm
ultiv
aria
tere
gres
sion
incl
udin
gth
evi
rule
nce
mea
sure
,st
ate
and
year
ofbi
rth
fixe
deff
ects
,th
ein
fant
mor
talit
yra
tein
stat
ean
dye
arof
birt
h,an
dth
eat
triti
onof
birt
hco
hort
inth
eC
ensu
sdat
a.
Est
imat
ion
resu
ltsco
mpar
ing
bir
thco
hort
s:T
he19
19bi
rth
coho
rt,co
mpa
red
with
the
coho
rttr
end:
was
4%-5
%(1
3%-1
5%am
ong
trea
ted)
less
likel
yto
com
plet
ehi
ghsc
hool
rece
ived
0.6%
-1.6
%fe
wer
year
sofe
duca
tion
had
1%-3
%le
ssto
tali
ncom
e(f
orm
ales
only
,20
05do
llars
)ha
d1%
-2%
low
erso
cioe
cono
mic
stat
usin
dex
(Dun
can
inde
x)w
as1%
-2%
mor
elik
ely
toha
vea
disa
bilit
yth
atlim
itsw
ork
(for
mal
eson
ly)
had
12%
high
erav
erag
ew
elfa
repa
ymen
t(fo
rw
omen
).E
stim
ates
are
sligh
tlyla
rger
for
nonw
hite
subg
roup
.Est
imat
ion
resu
ltsusi
ng
geogra
phic
vari
atio
nan
dst
ate
fixe
deff
ects
:(F
orth
e19
19bi
rth
coho
rt,us
edav
erag
em
ater
nali
nfec
tion
rate=
1/3)
Mat
erna
linf
ectio
n:re
duce
ssch
oolin
gby
2.2%
redu
cesp
roba
bilit
yof
high
scho
olgr
adua
tion
by0.
05%
decr
ease
sann
uali
ncom
eby
6%re
duce
ssoc
ioec
onom
icst
atus
inde
xby
2%-3
%.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1345
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Exp
lain
ing
siblin
gdi
ffer
ence
sin
achi
evem
enta
ndbe
havi
oral
outc
omes
:th
eim
port
ance
ofw
ithin
-an
dbe
twee
n-fa
mily
fact
ors(
Con
ley
etal
.,20
07)
Dat
afr
omch
ildde
velo
pmen
tsu
pple
men
tofP
SID
.N=
1360
.
Sibl
ing
fixe
deff
ects
toex
amin
eeff
ects
ofbi
rth
wei
ght,
birt
hor
der,
and
gend
eron
late
rou
tcom
es.
Cog
nitiv
eou
tcom
esm
easu
red
byth
eW
oodc
ock-
John
son
revi
sed
test
sofa
chie
vem
ent.
Beh
avio
ral
outc
omes
mea
sure
dby
the
beha
vior
alpr
oble
msi
ndex
(BPI
).C
ontr
olfo
rfa
mily
-an
dch
ild-s
peci
fic
char
acte
rist
ics.
No
stat
istic
ally
signi
fica
nteff
ects
ofbi
rth
wei
ghto
nB
PIor
onco
gniti
veas
sess
men
tsfo
rw
hole
sam
ple.
For
blac
ks,po
sitiv
eeff
ecto
fbir
thw
eigh
ton
cogn
itive
asse
ssm
ents
.
Twin
differ
ence
sin
birt
hw
eigh
t:th
eeff
ects
ofge
noty
pean
dpr
enat
alen
viro
nmen
ton
neon
atal
and
post
nata
lmor
talit
y(C
onle
yet
al.,
2006
)D
ata
ontw
inbi
rths
from
the
1995
-97
mat
ched
mul
tiple
birt
hda
taba
se.
N=
258,
823.
Twin
fixe
deff
ects
mod
elso
feffec
tsof
birt
hw
eigh
ton
mor
talit
yfo
rsa
me-
sex
and
mix
ed-s
expa
irs.
1lb
incr
ease
inbi
rth
wei
ghtl
eads
to:
9%(1
0%)r
educ
tion
inin
fant
mor
talit
yfo
rm
ixed
-sex
(sam
e-se
x)7%
(8%
)red
uctio
nin
neon
atal
mor
talit
yfo
rm
ixed
-sex
(sam
e-se
x)2%
redu
ctio
nin
post
-neo
nata
lmor
talit
yfo
rm
ixed
-sex
and
sam
e-se
x.Fo
rfu
ll-te
rmtw
ins,
mix
ed-s
exeff
ects
muc
hla
rger
than
sam
e-se
xeff
ects
.
(con
tinue
don
next
page
)
1346 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Bio
logy
asde
stin
y?Sh
ort-
and
long
-run
dete
rmin
ants
ofin
terg
ener
atio
nalt
rans
miss
ion
ofbi
rth
wei
ght(
Cur
rie
and
Mor
etti,
2007
).C
alifo
rnia
nata
lity
data
for
child
ren
born
betw
een
1989
and
2001
and
thei
rm
othe
rs(if
born
inC
A)
born
betw
een
1970
and
1974
.M
othe
rsw
hoar
esis
ters
are
mat
ched
usin
ggr
andm
othe
r’sna
me.
n=
638,
497
birt
hs.
Exa
min
eeff
ecto
fmot
her’s
birt
hw
eigh
ton
child
’sbi
rth
wei
ghti
nm
odel
swith
gran
dmot
her
fixe
deff
ects
.E
xam
ine
inte
ract
ions
ofm
othe
r’sbi
rth
wei
ghta
ndgr
andm
othe
r’sSE
Sat
time
ofm
othe
r’sbi
rth
aspr
oxie
dby
inco
me
inzi
pco
deof
mot
her’s
birt
h.E
xam
ine
effec
tofm
ater
nall
owbi
rth
wei
ghto
nm
othe
r’sSE
Sat
time
she
give
sbir
th.
Mot
her’s
low
birt
hw
eigh
tinc
reas
eslik
elih
ood
that
child
islo
wbi
rth
wei
ghtb
yab
out5
0%.T
hein
cide
nce
ofch
ildlo
wbi
rth
wei
ghti
s7%
high
erif
mot
her
was
born
into
high
pove
rty
zip
code
than
into
low
pove
rty
zip
code
.C
hild
ren
born
into
poor
hous
ehol
dsar
e0.
7%m
ore
likel
yto
belo
wbi
rth
wei
ghti
fthe
irm
othe
rsw
ere
low
birt
hw
eigh
t;ch
ildre
nbo
rnin
tono
n-po
orho
useh
olds
are
0.4%
mor
elik
ely
tobe
low
birt
hw
eigh
tift
heir
mot
hers
wer
elo
wbi
rth
wei
ght(
so,po
vert
yra
isest
hepr
obab
ility
oftr
ansm
issio
nof
low
birt
hw
eigh
tby
88%
).B
eing
low
birt
hw
eigh
tisa
ssoc
iate
dw
itha
loss
of$1
10in
futu
rein
com
e,on
aver
age,
ona
base
line
inco
me
of$1
0,09
6(in
1970
dolla
rs).
Bei
nglo
wbi
rth
wei
ghti
ncre
ases
the
prob
abili
tyof
livin
gin
ahi
gh-p
over
tyne
ighb
orho
odby
3%re
lativ
eto
the
base
line.
Bei
nglo
wbi
rth
wei
ghtr
educ
esfu
ture
educ
atio
nala
ttai
nmen
tby
0.1
year
s.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1347
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
From
the
crad
leto
the
labo
rm
arke
t:th
eeff
ecto
fbir
thw
eigh
ton
adul
tout
com
es(B
lack
etal
.,20
07).
Bir
thre
cord
sfro
mN
orw
ayfo
r19
67-1
997.
Dro
pped
cong
enita
ldef
ects
.M
atch
edto
regi
stry
data
oned
ucat
ion
and
labo
rm
arke
tou
tcom
esan
dto
mili
tary
reco
rds.
n=
33,3
66tw
inpa
irs.
Twin
fixe
deff
ects
,co
ntro
lling
for
mot
her
and
birt
h-sp
ecifi
cva
riab
les.
Log(
birt
hw
eigh
t)is
prim
ary
inde
pend
entv
aria
ble.
10%
incr
ease
inbi
rth
wei
ght:
redu
ces1
-yea
rm
orta
lity
by13
%in
crea
ses5
min
APG
AR
scor
eby
0.3%
incr
ease
spro
babi
lity
ofhi
ghsc
hool
com
plet
ion
by1.
2%in
crea
sesf
ull-
time
earn
ings
by1%
.M
ale
outc
om
esat
age
18-2
0:10
%in
crea
sein
birt
hw
eigh
t:in
crea
sesh
eigh
tby
0.3%
incr
ease
sBM
Iby
0.5%
incr
ease
sIQ
by1.
1%(s
cale
of1-
9).
Effec
tsofm
oth
er’s
bir
thw
eigh
ton
child’s
bir
thw
eight:
10%
incr
ease
inm
othe
r’sbi
rth
wei
ght:
incr
ease
schi
ld’s
birt
hw
eigh
tby
1.5%
.
(con
tinue
don
next
page
)
1348 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
influe
nce
ofea
rly-
life
even
tson
hum
anca
pita
l,he
alth
stat
usan
dla
bor
mar
keto
utco
mes
over
the
life
cour
se(J
ohns
onan
dSc
hoen
i,20
07)P
SID
.A
dult
sam
ple
born
betw
een
1951
and
1975
.N=
5160
peop
lean
d16
55fa
mili
es.C
hild
sam
ple
0-12
year
sold
in19
97.
N=
1127
mot
hers
and
2239
child
ren.
Sibl
ing
fixe
deff
ects
toes
timat
eim
pact
ofm
ater
nals
mok
ing
duri
ngpr
egna
ncy
onbi
rth
outc
omes
.A
lsoes
timat
edm
odel
usin
gO
LS,
cont
rolli
ngfo
rgr
andp
aren
tsm
okin
gin
adol
esce
nce,
mat
erna
lbi
rth
wei
ght,
and
pate
rnal
smok
ing,
amon
got
her
back
grou
ndch
arac
teri
stic
s.
Not
e:m
eans
are
notr
epor
ted,
sore
lativ
eeff
ects
cann
otbe
calc
ulat
ed.
Effec
tsofin
com
ean
dm
oth
er’s
bir
thw
eigh
ton
child’s
bir
thw
eigh
t:A
nin
crea
sein
inco
me
of$1
0,00
0ra
isesc
hild
’sbi
rth
wei
ght
by0.
12lb
sifm
othe
rw
aslo
wbi
rth
wei
ght(
by0.
02lb
sif
mot
her
was
notl
owbi
rth
wei
ght)
for
afa
mily
with
inco
me
of$7
500
(199
7do
llars
).H
avin
gno
heal
thin
sura
nce
incr
ease
sthe
prob
abili
tyof
low
birt
hw
eigh
tby
10pp
.E
ffec
tson
child
hea
lth
outc
om
es:
(hea
lthin
dex
1-10
0)Lo
wbi
rth
wei
ghts
iblin
gsha
vea
1.67
poin
tlow
erhe
alth
inde
x.Pr
ivat
ehe
alth
insu
ranc
ein
crea
sesh
ealth
inde
xby
1.02
poin
ts.
A$1
0,00
0in
crea
sein
inco
me
for
fam
ilies
with
$15,
000-
50,0
00in
com
ein
crea
sesh
ealth
inde
xby
0.53
pp.
Effec
tson
adult
outc
om
es:
Low
birt
hw
eigh
tsib
lings
:4.
7pp
mor
elik
ely
tobe
high
scho
oldr
opou
ts,ha
vea
3.7
poin
tlow
erhe
alth
inde
x(1
-100
scal
e).
Low
birt
hw
eigh
tbro
ther
s(sa
mpl
eof
mal
eson
ly):
are
4.3
ppm
ore
likel
yto
have
nopo
sitiv
eea
rnin
gs(s
igat
10%
leve
l)ha
ve$2
966
less
annu
alea
rnin
gs(s
igat
10%
leve
l).
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1349
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Mat
erna
lsm
okin
gdu
ring
preg
nanc
yan
dea
rly
child
outc
omes
(Tom
iney
,20
07)
Chi
ldre
nof
NC
DS
mot
hers
born
betw
een
1973
and
2000
.n=
2799
mot
hers
and
6291
siblin
gch
ildre
n.
Sibl
ing
fixe
deff
ects
toes
timat
eim
pact
ofm
ater
nals
mok
ing
duri
ngpr
egna
ncy
onbi
rth
outc
omes
.A
lsoes
timat
edm
odel
usin
gO
LS,
cont
rolli
ngfo
rgr
andp
aren
tsm
okin
gin
adol
esce
nce,
mat
erna
lbi
rth
wei
ght,
and
pate
rnal
smok
ing,
amon
got
her
back
grou
ndch
arac
teri
stic
s.
Not
e:on
lyre
port
ing
resu
ltsfr
omsib
ling
fixe
deff
ects
regr
essio
nhe
re.
Mat
erna
lsm
okin
gdu
ring
preg
nanc
yre
duce
sbir
thw
eigh
tby
1.7%
.N
ost
atist
ical
lysig
nifica
nteff
ecto
fmat
erna
lsm
okin
gdu
ring
preg
nanc
yon
prob
abili
tyof
havi
nga
low
birt
hw
eigh
tchi
ld,
pre-
term
gest
atio
n,or
wee
ksof
gest
atio
n.N
ost
atist
ical
lysig
nifica
nteff
ects
ofm
ater
nals
mok
ing
amon
gm
othe
rsw
hoqu
itby
mon
th5
ofpr
egna
ncy.
Larg
ereff
ects
ofm
ater
nal
smok
ing
onbi
rth
wei
ghta
mon
glo
wed
ucat
edw
omen
.
Can
api
ntpe
rda
yaff
ecty
our
child
’spa
y?T
heeff
ecto
fpre
nata
lal
coho
lexp
osur
eon
adul
tou
tcom
es(N
ilsso
n,20
08).
Dat
afr
omSw
edish
LOU
ISE
data
base
onfirs
t-bo
rnin
divi
dual
sbor
nbe
twee
n19
64an
d19
72.
n=
353,
742.
Swed
ishna
tura
lexp
erim
enti
nw
hich
alco
hola
vaila
bilit
yin
2tr
eatm
entr
egio
nsin
crea
sed
shar
ply
asre
gula
rgr
ocer
yst
ores
wer
eal
low
edto
mar
kets
tron
gbe
erfo
r6
mon
thsd
urin
g19
67w
ithth
em
inim
umag
efo
rpu
rcha
sebe
ing
16(in
stea
dof
21).
Diff
eren
ce-i
n-di
ffer
ence
-in-
differ
ence
com
pari
ngun
der-
21m
othe
rsw
ithol
der
mot
hers
intr
eatm
enta
ndco
ntro
lre
gion
spre
-,du
ring
,an
dpo
st-
expe
rim
ent.
Bas
elin
ees
timat
ions
focu
sed
onch
ildre
nco
ncei
ved
prio
rto
the
expe
rim
ent,
but
expo
sed
inut
ero
toth
eex
peri
men
t.C
ontr
olle
dfo
rqu
arte
ran
dco
unty
ofbi
rth
fixe
deff
ects
.
DD
Dre
sults:
Year
sofs
choo
ling:
decr
ease
dby
0.27
(2.1
%)y
ears
for
who
lesa
mpl
e;by
0.47
year
sfor
mal
es;no
stat
istic
ally
signi
fica
nteff
ectf
orfe
mal
es.
HS
grad
uatio
n:de
crea
sed
by0.
4%fo
rw
hole
sam
ple;
by10
%fo
rm
ales
;no
stat
.sig
nifica
nteff
ectf
orfe
mal
es.
Gra
duat
ion
from
high
ered
ucat
ion:
decr
ease
dby
16%
for
who
lesa
mpl
e;by
35%
for
mal
es;no
stat
.sig
nifica
nteff
ectf
orfe
mal
es.
Ear
ning
sata
ge32
:de
crea
sed
by24
.1%
for
who
lesa
mpl
e;by
22.8
%fo
rm
ales
;by
17.7
%fo
rfe
mal
es.
Prob
abili
tyno
inco
me
atag
e32
:in
crea
sed
by74
%w
hole
sam
ple.
Prop
ortio
non
wel
fare
:in
crea
sed
by90
%fo
rw
hole
sam
ple;
by5.
1pp
for
mal
es.
(con
tinue
don
next
page
)
1350 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Shor
t-,m
ediu
m-
and
long
-ter
mco
nseq
uenc
esof
poor
infa
nthe
alth
(Ore
opou
lose
tal.,
2008
).D
ata
from
the
Man
itoba
Cen
ter
for
Hea
lthPo
licy
mat
chin
gpr
ovin
cial
heal
thin
sura
nce
clai
msw
ithbi
rth
reco
rds,
educ
atio
nalr
ecor
ds,an
dso
cial
assis
tanc
ere
cord
s.In
clud
esal
lch
ildre
nbo
rnin
Man
itoba
from
1979
to19
85.n=
54,1
23sib
lings
and
1742
twin
s.
Sibl
ing
and
twin
fixe
deff
ects
.U
sed
thre
em
easu
reso
finf
anth
ealth
:bi
rth
wei
ght,
5-m
inA
pgar
scor
e,an
dge
stat
iona
llen
gth
inw
eeks
;us
eddu
mm
iesf
ordi
ffer
ent
cate
gori
esof
the
vari
able
sto
estim
ate
nonl
inea
reff
ects
.A
lsoes
timat
edm
odel
susin
gO
LSfo
rth
ew
hole
sam
ple,
for
the
siblin
gssa
mpl
e,an
dfo
rth
etw
inss
ampl
ew
ithou
tfam
ilyfixe
deff
ects
.
Not
e:on
lyre
port
ing
resu
ltsfr
omre
gres
sions
that
incl
uded
fam
ilyfixe
deff
ects
here
.B
W=
birt
hw
eigh
t.R
elat
ive
toA
pgar
scor
e=
10,
BW>
3500
g,ge
stat
ion
40-4
1w
eeks
low
erva
lues
.E
ffec
tson
infa
ntm
ort
ality:
incr
ease
infa
ntm
oral
ity.
E.g
.in
siblin
gsa
mpl
e(in
fant
mor
talit
y=
0.01
1).
Apg
arsc
ore<
6in
crea
sesp
roba
bilit
yof
infa
ntm
orta
lity
by31
.9pp
BW<
1000
gin
crea
sesp
roba
bilit
yof
infa
ntm
orta
lity
by87
.2pp
Ges
tatio
n<
36w
eeks
incr
ease
spro
babi
lity
ofin
fant
mor
talit
yby
11.9
pp.
Effec
ton
langu
age
arts
score
(tak
enin
gra
de
12):
Apg
arsc
ore<
6de
crea
sest
ests
core
by0.
1of
SD(s
ig.at
10%
leve
l)B
W25
01-3
000
gde
crea
sest
ests
core
by0.
04of
SDE
ffec
ton
pro
bab
ility
ofre
achin
ggra
de
12by
age
17:
Apg
arsc
ore<
6de
crea
sesg
rade
12pr
obab
ility
by4.
1pp
(sig
.at
10%
)B
W10
01-1
500
gde
crea
sesg
rade
12pr
obab
ility
by14
.1pp
Ges
tatio
n<
36w
eeks
decr
ease
sgra
de12
prob
abili
tyby
4.0
pp.
Effec
ton
soci
alas
sist
ance
take
-up
duri
ng
ages
18-2
1.25
:B
W<
1000
decr
ease
spro
babi
lity
ofta
ke-u
pby
21.5
pp.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1351
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Bir
thco
hort
and
the
blac
k-w
hite
achi
evem
entg
ap:th
ero
leof
heal
thso
onaf
ter
birt
h(C
hay
etal
.,20
09).
Dat
afr
omth
ena
tiona
lass
essm
ento
fed
ucat
iona
lpro
gres
slon
g-te
rmtr
ends
for
1971
-200
4.A
FQT
data
from
US
mili
tary
for
1976
-200
1fo
rm
ale
appl
ican
ts17
-20.
n=
2,64
9,57
3w
hite
mal
esan
dn=
1,10
3,74
8bl
ack
mal
es.H
ospi
tald
ischa
rge
rate
sfr
omN
atio
nalH
ealth
Inte
rvie
wSu
rvey
.
Reg
ress
ion
ofte
stsc
ores
onye
ar,
age,
and
subj
ectfi
xed
effec
ts(s
epar
atel
yfo
rbl
acks
and
whi
tes)
usin
gN
AE
P-LT
Tan
dA
FQT
test
scor
eda
ta.In
AFQ
Tte
stsc
ore
data
,co
rrec
tfor
sele
ctio
nbi
asus
ing
inve
rse
prob
abili
tyw
eigh
ting
from
Nat
ality
and
Cen
susd
ata.
Est
imat
edre
gres
sions
sepa
rate
lyfo
rbl
acks
and
whi
tesa
ndfo
rth
eN
orth
,th
eSo
uth,
the
Rus
tbel
t,th
ede
epSo
uth,
and
indi
vidu
alst
ates
with
inth
eSo
uth
and
Nor
th.A
lsoes
timat
eddi
ffer
ence
-in-
differ
ence
-in-
differ
ence
(DD
D)
mod
els,
com
pari
ngbl
ack
and
whi
tete
stsc
ores
betw
een
coho
rtsb
orn
in19
60-6
2an
d19
70-7
2in
the
Sout
hre
lativ
eto
the
Rus
tbel
t,co
ntro
lling
for
regi
on-s
peci
fic,
race
-spe
cific
age-
by-t
ime
effec
tsan
dra
ce-b
y-re
gion
-by-
time
effec
ts.
Use
dpo
stne
onat
alm
orta
lity
rate
(PN
MR
)asa
prox
yfo
rin
fant
heal
then
viro
nmen
tto
asse
ssim
pact
ofin
fant
heal
thon
the
test
scor
ega
p.
NA
EP
-LT
Tdat
a:T
hebl
ack-
whi
tete
stsc
ore
gap
decl
ined
from
1SD
to0.
6of
SDbe
twee
n19
71an
d20
04.T
heco
nver
genc
ew
aspr
imar
ilydu
eto
larg
ein
crea
sesi
nbl
ack
test
scor
esin
the
1980
s.R
egre
ssio
nre
sults
indi
cate
that
the
conv
erge
nce
was
due
toco
hort
effec
ts,ra
ther
than
time
effec
ts.
AFQ
Tdat
a:T
hebl
ack-
whi
tega
pin
AFQ
Tte
stsc
ores
decl
ined
byab
out
19%
betw
een
the
1962
-63
and
1972
birt
hco
hort
s.T
hede
clin
ein
the
gap
isab
out0
.3SD
sgre
ater
inth
eSo
uth
rela
tive
toth
eR
ustb
eltb
etw
een
1960
-62
and
1970
-72
coho
rts.
Con
verg
ence
inPN
MR
expl
ains
52%
ofth
eva
riat
ion
acro
ssst
ates
inA
FQT
conv
erge
nce.
Effec
tsofac
cess
tohosp
ital
s:A
30pp
incr
ease
inbl
ack
hosp
italb
irth
rate
sfro
m19
62-6
4to
1968
-70
incr
ease
scoh
ortA
FQT
scor
esby
7.5
perc
entil
epo
ints
.A
blac
kch
ildw
hoga
ined
adm
issio
nto
aho
spita
lbef
ore
age
4ha
da
0.7-
1SD
gain
inA
FQT
scor
eat
age
17-1
8re
lativ
eto
abl
ack
child
who
did
not.
(con
tinue
don
next
page
)
1352 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Sepa
rate
dat
birt
h:U
Stw
ines
timat
esof
the
effec
tsof
birt
hw
eigh
t(R
oyer
,20
09)
Cal
iforn
iabi
rth
reco
rds.
n=
3028
sam
e-se
xfe
mal
etw
inbi
rths
1960
-198
2.D
ata
from
earl
ych
ildho
odlo
ngitu
dina
lst
udy-
birt
hco
hort
.n=
1496
twin
birt
hs.
Twin
fixe
deff
ects
,al
low
ing
non-
linea
reff
ects
ofbi
rth
wei
ght.
Tw
infixe
d-e
ffec
tre
sults:
250
gin
crea
sein
birt
hw
eigh
tlea
dsto
:8.
3%de
crea
sein
prob
abili
tyof
infa
ntm
orta
lity
0.82
day
decr
ease
inst
ayin
hosp
italp
ost-
birt
h0.
02-0
.04
ofSD
incr
ease
inm
enta
l/m
otor
test
scor
e0.
2%in
crea
sein
educ
.at
tain
men
tofm
othe
rat
child
birt
h0.
5%in
crea
sein
child
’sbi
rth
wei
ght
11%
decr
ease
inpr
egna
ncy
com
plic
atio
ns.
No
stat
istic
ally
signi
fica
nteff
ects
ofbi
rth
wei
ghto
nad
ult
heal
thou
tcom
essu
chas
hype
rten
sion,
anem
ia,an
ddi
abet
es.
No
stat
istic
ally
signi
fica
nteff
ects
ofbi
rth
wei
ghto
nin
com
e-re
late
dm
easu
res.
200
gin
crea
sein
birt
hw
eigh
tlea
dsto
a0.
5pp
incr
ease
inpr
obab
ility
ofbe
ing
obse
rved
inad
ulth
ood—
smal
lsel
ectio
nbi
as.Fo
und
nost
atist
ical
lysig
nifica
ntev
iden
ceof
com
pens
atin
gor
rein
forc
ing
pare
ntal
inve
stm
ents
whe
nco
nsid
erin
gea
rly
med
ical
care
and
brea
stfe
edin
g.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1353
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
long
-ter
mec
onom
icim
pact
ofin
uter
oan
dpo
stna
tale
xpos
ure
tom
alar
ia(B
arre
ca,20
10).
Mal
aria
mor
talit
yfr
omU
Sm
orta
lity
stat
istic
s190
0-19
36.
n=
1147
stat
e-ye
arob
serv
atio
ns.C
limat
eda
tan=
1813
obs.
Adu
ltou
tcom
esfr
om19
60C
ensu
s.A
llda
tam
erge
dat
stat
e/ye
arof
birt
hle
vel.
IVus
ing
frac
tion
ofda
ysin
“mal
aria
-ide
al”
tem
pera
ture
rang
eas
IVfo
rm
alar
iade
aths
inst
ate
and
year
.
Not
e:O
nly
resu
ltsfr
omth
eIV
regr
essio
nar
ere
port
edhe
re.
Exp
osur
eto
10ad
ditio
nalm
alar
iade
aths
per
100,
000
inha
bita
ntsc
ause
s3.4
%le
ssye
arso
fsch
oolin
gE
xpos
ure
tom
alar
iaca
nac
coun
tfor
appr
oxim
atel
y25
%of
the
differ
ence
inye
arso
fsch
oolin
gbe
twee
nco
hort
sbor
nin
high
and
low
mal
aria
stat
es.
Long
-run
long
evity
effec
tsof
anu
triti
onsh
ock
inea
rly
life:
The
Dut
chFa
min
eof
1846
-47
(Van
Den
Ber
get
al.,
inpr
ess)
.H
istor
ical
sam
ple
ofth
eN
ethe
rlan
ds.E
xpos
edco
hort
born
9/1/
1846
-6/1/
1848
.N
on-e
xpos
edbo
rn9/
1/18
48-9/1/
1855
and
9/1/
1837
-9/1/
1944
.
Key
inde
pend
entv
aria
ble
isex
posu
reto
fam
ine
atbi
rth.
Inst
rum
enta
cces
sto
food
with
vari
atio
nsin
year
lyav
erag
ere
alm
arke
tpri
ceso
frye
and
pota
toes
for
thre
edi
ffer
entr
egio
ns.
Con
trol
led
for
mac
roec
onom
icco
nditi
ons,
infa
ntm
orta
lity
rate
s,an
din
divi
dual
dem
ogra
phic
and
soci
o-ec
onom
icch
arac
teri
stic
s.
Res
ultsfr
om
nonpar
amet
ric
regre
ssio
n:
Res
idua
llife
expe
ctan
cyat
age
50is
3.1
year
ssho
rter
for
expo
sed
men
than
for
men
born
afte
rth
efa
min
e;1.
4ye
ars
shor
ter
than
for
men
born
befo
refa
min
e.K
olm
ogor
ov-S
mir
nov
test
sugg
ests
that
the
surv
ival
curv
esaf
ter
age
50di
ffer
signi
fica
ntly
for
expo
sed
and
cont
rolm
en.
Max
differ
ence
indi
stri
butio
n=
0.15
atag
e56
.R
esultsfr
om
par
amet
ric
surv
ival
model
s:E
xpos
ure
tofa
min
eat
birt
hfo
rm
enre
duce
sres
idua
llife
expe
ctan
cyat
age
50by
4.2
year
s.N
ost
atist
ical
lysig
nifica
ntre
sults
for
wom
en.
(con
tinue
don
next
page
)
1354 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Mat
erna
lstr
essa
ndch
ildw
ell-
bein
g:ev
iden
cefr
omsib
lings
(Aiz
eret
al.,
2009
).N
atio
nal
Col
labo
rativ
ePe
rina
talP
roje
ct.
Bir
ths1
959-
65in
Prov
iden
cean
dB
osto
n.N=
1103
birt
hsto
915
wom
en.16
3sib
lings
insa
mpl
e.
Sibl
ing
fixe
deff
ects
.E
stim
ated
effec
tofd
evia
tions
from
mea
nco
rtiso
llev
els(
cort
isolm
easu
res
mat
erna
lstr
ess)
duri
ngpr
egna
ncy
onou
tcom
es.
No
signi
fica
nteff
ects
onbi
rth
wei
ghto
rm
ater
nalp
ostn
atal
inve
stm
ents
.E
xpos
ure
toto
pqu
artil
eof
cort
isoll
evel
dist
ribu
tion
(rel
ativ
eto
the
mid
dle)
lead
sto
47%
ofSD
decr
ease
inve
rbal
IQat
age
7(s
ign.
at10
%le
vel).
1SD
devi
atio
nin
cort
isoll
evel
slea
dsto
26%
ofSD
decr
ease
ined
ucat
iona
latt
ainm
ent.
Exp
osur
eto
top
quar
tile
ofco
rtiso
llev
eldi
stri
butio
n(r
elat
ive
toth
em
iddl
e)le
adst
o51
%of
SDde
crea
sein
educ
atio
nal
atta
inm
ent.
The
scou
rge
ofA
sian
flu:
the
phys
ical
and
cogn
itive
deve
lopm
ento
faco
hort
ofB
ritis
hch
ildre
nin
uter
odu
ring
the
Asia
nIn
flue
nza
Pand
emic
of19
57,fr
ombi
rth
until
age
11(K
elly
,20
09).
NC
DS
1958
coho
rt.
N=
16,7
65at
birt
h,14
,358
at7,
14,0
69at
11.
Effec
toffl
uon
each
coho
rtm
embe
rid
entifi
edus
ing
vari
atio
nin
inci
denc
eof
epid
emic
bylo
cal
auth
ority
ofbi
rth.
Epi
dem
icpe
aked
whe
nco
hort
mem
bers
wer
e17
-23
wee
ksge
stat
ion.
1/3
ofw
omen
ofch
ild-b
eari
ngag
ew
ere
infe
cted
,he
nce
true
trea
tmen
teff
ectc
anbe
estim
ated
bym
ultip
lyin
gre
sults
by3.
No
stat
istic
ally
signi
fica
ntev
iden
ceof
rein
forc
ing
orco
mpe
nsat
ing
pare
ntal
inve
stm
ents
.1
SDin
crea
sein
epid
emic
inte
nsity
decr
ease
sbir
thw
eigh
tby
0.03
-0.3
63SD
;de
crea
sest
ests
core
sby
0.06
7of
SDat
age
7,by
0.04
3of
SDat
age
11;
incr
ease
sdet
rim
enta
leffec
tofm
othe
rpr
eecl
amps
iafr
om−
0.07
5to−
0.11
SDon
birt
hw
eigh
t;in
crea
sesd
etri
men
tale
ffec
tofm
othe
rsm
okin
gby
0.03
ofSD
onbi
rth
wei
ght;
incr
ease
sdet
rim
enta
leffec
tofm
othe
run
der
8st
one
wei
ght
from−
0.54
to−
0.61
ofSD
onbi
rth
wei
ght
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1355
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Do
low
erbi
rth
wei
ghtb
abie
sha
velo
wer
grad
es?
Twin
fixe
deff
ecta
ndin
stru
men
talv
aria
bles
evid
ence
from
Tai
wan
(Lin
and
Liu,
2009
).B
irth
Cer
tifica
teda
tafo
rT
aiw
an.H
igh
scho
olen
tran
ceex
amre
sults
from
Com
mitt
eeof
Bas
icC
ompe
tenc
eTe
st.
N=
118,
658.
Twin
sam
ple
n=
7772
.
Twin
fixe
deff
ects
and
IVus
ing
the
publ
iche
alth
budg
etan
dnu
mbe
rof
doct
orsi
nco
unty
whe
rech
ildw
asbo
rnas
inst
rum
ents
for
child
’sbi
rth
wei
ght.
Tw
inFE/I
Vfo
rm
oth
ersw
/<9
yrsed
&<
25yr
sold
Incr
ease
inbi
rth
wei
ghto
f100
gle
adst
o:In
crea
seC
hine
sesc
ore:
0.5%
/8.2
%In
crea
seE
nglis
hsc
ore:
0.3%
/8.1
%*
Incr
ease
Mat
hsc
ore:
0.7%
/12.
6%In
crea
seN
atur
alSc
ienc
esc
ore:
0.8%
/11.
8%In
crea
seSo
cial
Scie
nce
scor
e:0.
4%/4
.0%
**
mea
nssig
nifica
ntat
10%
leve
lIV
resu
ltsfo
rot
her
subg
roup
ssho
wsm
alle
r/no
effec
ts.
Poor
,hu
ngry
and
stup
id:
num
erac
yan
dth
eim
pact
ofhi
ghfo
odpr
ices
inin
dust
rial
izin
gB
rita
in,17
50-1
850
(Bat
enet
al.,
2007
).D
ata
onag
ehe
apin
gfr
om19
51to
1881
Bri
tish
Cen
sus.
Info
rmat
ion
onpo
orre
lieff
rom
Boy
er(1
990)
.
Age
heap
ing
isro
undi
ngag
eto
near
est5
or10
.W
hipp
lein
dex
(WI)=
num
ber
ofag
esth
atar
em
ultip
lesr
elat
ive
toex
pect
ednu
mbe
rgi
ven
unifo
rmag
edi
stri
butio
n.R
egre
ssio
nsus
eW
Ias
depe
nden
tvar
iabl
e.W
heat
pric
esan
dpo
orre
liefm
easu
resa
rein
depe
nden
tvar
iabl
es.
Dur
ing
the
Rev
olut
iona
ryan
dN
apol
eoni
cW
ars,
the
pric
eof
whe
atal
mos
tdou
bled
—an
dth
enu
mbe
rof
erro
neou
slyre
port
edag
esat
mul
tiple
sof5
doub
led
from
4%to
8%.M
enan
dw
omen
born
inde
cade
swith
high
erW
Iso
rted
into
jobs
that
had
low
erin
telli
genc
ere
quir
emen
ts.1
SDin
crea
sein
the
Whi
pple
inde
xas
soci
ated
with
a2.
8%(r
elat
ive
tom
edia
nea
rnin
gs)d
ecre
ase
inea
rnin
gs.
(con
tinue
don
next
page
)
1356 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Impa
ctsof
econ
omicsh
ocks
Bir
thw
eigh
tand
inco
me:
inte
ract
ions
acro
ssge
nera
tions
(Con
ley
and
Ben
nett
,20
01)
Dat
afr
omPS
ID.Tw
osa
mpl
es:
(1)c
hild
ren
born
1986
-199
2,n=
1654
;(2
)rea
ched
19by
1992
,n=
1388
indi
vidu
alsi
n76
6fa
mili
es.
Sibl
ing
fixe
deff
ects
.Lo
wbi
rth
wei
ghtc
hild
who
spen
t6ye
arsa
tpov
erty
line
isle
sslik
ely
togr
adua
tehi
ghsc
hool
than
norm
albi
rth
wei
ght
child
who
spen
t6ye
arsa
tpov
erty
line.
Low
birt
hw
eigh
tch
ildw
hosp
ent6
year
swith
inco
me
5tim
esth
epo
vert
ylin
eis
aslik
ely
togr
adua
tehi
ghsc
hool
asno
rmal
birt
hw
eigh
tch
ildof
sam
ein
com
e(s
igni
fica
ntat
10%
leve
l).
Eco
nom
icco
nditi
onse
arly
inlif
ean
din
divi
dual
mor
talit
y(V
anD
enB
erg
etal
.,in
pres
s).D
ata
from
hist
oric
alsa
mpl
eof
the
Net
herl
ands
onin
divi
dual
sbor
n18
12-1
903
with
date
ofde
ath
obse
rved
by20
00.n=
9276
.M
erge
dw
ithhi
stor
ical
mac
rotim
ese
ries
data
.
Com
pare
din
divi
dual
sbor
nin
boom
swith
thos
ebo
rnin
the
subs
eque
ntre
cess
ions
.C
ontr
olle
dfo
rw
arsa
ndep
idem
ics.
Com
pari
ngth
ose
born
inbo
omof
1872
-187
6w
ithth
ose
born
inre
cess
ion
of18
77-1
881:
Boo
mT|T>
2=
66.0
year
s,R
eces
sion
T|T>
2=
62.5
year
sB
oom
T|T>
5=
70.8
year
s,R
eces
sion
T|T>
5=
67.5
year
sR
egre
ssio
nre
sults:
Boo
mT|T>
2is
1.58
year
sgre
ater
than
Rec
essio
nT|T>
2N
otat
ion:
T|T>
2=
aver
age
lifet
ime
give
nsu
rviv
alpa
stag
e2
T|T>
5=
aver
age
lifet
ime
give
nsu
rviv
alpa
stag
e5
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1357
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Evi
denc
eon
earl
y-lif
ein
com
ean
dla
te-l
ifehe
alth
from
Am
eric
a’sD
ustb
owle
ra(C
utle
ret
al.,
2007
).T
heH
ealth
and
Ret
irem
entS
tudy
.N=
8739
peop
lebo
rnbe
twee
n19
29an
d19
41.A
gric
ultu
rald
ata
from
Nat
iona
lAgr
icul
tura
lSta
tistic
sSe
rvic
e.In
com
eda
tafr
omB
urea
uof
Eco
nom
icA
naly
sis.
Mea
sure
dec
onom
icco
nditi
onsi
nut
ero
usin
gin
com
ean
dyi
eld
from
the
sam
eca
lend
arye
arfo
rth
ose
born
in3r
dor
4th
quar
ter
ofth
eye
aran
dfr
omth
epr
evio
usca
lend
arye
arfo
rth
ose
born
in1s
tor
2nd
quar
ter.
Reg
ress
ions
incl
ude
the
inut
ero
econ
omic
cond
ition
mea
sure
(log
inco
me,
log
yiel
d),du
mm
yfo
rw
heth
erre
spon
dent
’sfa
ther
was
afa
rmer
and
inte
ract
ion
ofth
edu
mm
yw
ithth
eec
onom
icco
nditi
ons.
Con
trol
led
for
regi
onan
dye
arof
birt
hfixe
deff
ects
,re
gion
-spe
cific
linea
rtim
etr
ends
,an
dre
gion
-yea
rin
fant
deat
han
dbi
rth
rate
s,an
dot
her
dem
ogra
phic
char
acte
rist
ics.
No
stat
istic
ally
signi
fica
ntre
latio
nshi
pbe
twee
npo
orea
rly-
life
econ
omic
cond
ition
sdur
ing
the
Dus
tbow
land
late
-life
heal
thou
tcom
essu
chas
hear
tcon
ditio
ns,st
roke
,di
abet
es,hy
pert
ensio
n,ar
thri
tis,ps
ychi
atri
cco
nditi
ons,
etc.
(con
tinue
don
next
page
)
1358 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Long
-run
heal
thim
pact
sof
inco
me
shoc
ks:w
ine
and
Phyl
loxe
rain
19th
cent
ury
Fran
ce(B
aner
jee
etal
.,fo
rthc
omin
g).Se
epa
per
refe
renc
esfo
rso
urce
ofda
taon
win
epr
oduc
tion,
num
ber
ofbi
rths
,an
din
fant
mor
talit
y.D
epar
tmen
tlev
elda
taon
heig
hts
from
mili
tary
reco
rds1
872-
1912
.n=
3485
year
-dep
artm
ents
.
Reg
iona
lvar
iatio
nin
ala
rge
nega
tive
inco
me
shoc
kca
used
byPh
yllo
xera
atta
ckso
nFr
ench
vine
yard
sbet
wee
n18
63an
d18
90.
Shoc
kdu
mm
yeq
ualt
o0
pre
Phyl
loxe
raan
daf
ter
1890
(whe
ngr
aftin
gso
lutio
nfo
und)
.D
umm
yeq
ual1
whe
nw
ine
prod
uctio
n<
80%
ofpr
e-le
vel.
Diff
eren
ce-i
n-di
ffer
ence
com
pari
ngch
ildre
nbo
rnin
affec
ted
and
unaff
ecte
dar
easb
efor
ean
daf
ter.
Dec
reas
ein
win
epr
oduc
tion
was
notc
ompe
nsat
edby
anin
crea
sein
othe
rag
ricu
ltura
lpro
duct
ion
(e.g
.w
heat
),su
gges
ting
that
this
was
trul
ya
larg
ene
gativ
ein
com
esh
ock.
Mai
noutc
om
es:
3-5%
decl
ine
inhe
ight
atag
e20
for
thos
ebo
rnin
win
e-gr
owin
gfa
mili
esdu
ring
the
year
that
thei
rre
gion
was
affec
ted
byPh
yllo
xera
.T
hose
born
inPh
yllo
xera
-affec
ted
year
are
0.35
-0.3
8pp
mor
elik
ely
tobe
shor
ter
than
1.56
cm.
No
stat
istic
ally
signi
fica
nteff
ects
onot
her
mea
sure
sofh
ealth
orlif
eex
pect
ancy
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1359
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Impa
ctsof
environm
entalsho
cks
Life
span
depe
ndso
nth
em
onth
ofbi
rth
(Dob
lham
mer
and
Vaup
el,20
01)
Dat
aon
popu
latio
nsof
Den
mar
k,A
ustr
ia,an
dA
ustr
alia
.D
enm
ark:
Long
itudi
nald
ata
base
don
popu
latio
nre
gist
ryfo
r19
68-2
000
n=
1,37
1,00
3.A
ust
ria:
Dat
afr
omde
ath
cert
ifica
tesf
or19
88-1
996
n=
681,
677
Aust
ralia:
Dat
afr
omde
ath
cert
ifica
tesf
rom
1993
-199
7n=
219,
820
Also
used
data
onin
divi
dual
sbo
rnin
Bri
tain
who
died
inA
ustr
alia
:n=
43,0
74D
ata
onin
fant
mor
talit
yfo
rD
enm
ark
Use
dtt
ests
tope
rfor
mpa
irw
iseco
mpa
riso
nsbe
twee
nm
ean
age
atde
ath
byqu
arte
rof
birt
h.To
test
whe
ther
the
seas
onal
differ
ence
inth
eri
skof
deat
hac
coun
tsfo
rdi
ffer
ence
sin
adul
tlife
span
bym
onth
ofbi
rth,
calc
ulat
edm
onth
lyde
viat
ions
from
annu
alde
ath
rate
san
dus
edw
eigh
ted
leas
tsqu
ares
regr
essio
nw
ithdu
mm
iesf
orm
onth
ofbi
rth,
curr
entm
onth
,ag
esin
cela
stbi
rthd
ayin
mon
ths,
sex,
and
birt
hco
hort
.To
test
whe
ther
sele
ctiv
esu
rviv
alor
debi
litat
ion
duri
ngth
e1s
tyea
rof
life
expl
ains
differ
ence
sin
life
expe
ctan
cyat
age
50,ca
lcul
ated
mon
thly
deat
hra
tes
duri
ng1s
tyea
rof
life
and
mon
thly
devi
atio
nsfr
oman
nual
deat
hra
tes
duri
ng1s
tyea
rof
life;
then
used
am
ultiv
aria
tere
gres
sion
with
out
cont
rols
for
sex
and
birt
hco
hort
.
Not
e:A
utum
n-Sp
ring=
aver
age
differ
ence
inag
eat
deat
hbe
twee
npe
ople
born
inA
utum
n(O
ct-D
ec)a
ndin
the
Spri
ng(A
pr-J
un).
Den
mar
k:M
ean
rem
aini
nglif
eex
pect
ancy
atag
e50=
27.5
2ye
ars
0.19
year
ssho
rter
lifes
pans
for
thos
ebo
rnin
2nd
quar
ter
0.12
year
slon
ger
lifes
pans
for
thos
ebo
rnin
4th
quar
ter
Cor
rela
tion
betw
een
infa
ntm
orta
lity
attim
eof
birt
han
dad
ultm
orta
lity
afte
rag
e50=
0.87
.A
ust
ria:
Ave
rage
age
atde
ath=
77.7
0ye
ars
0.28
year
ssho
rter
lifes
pans
for
thos
ebo
rnin
2nd
quar
ter
Aust
ria:
aver
age
age
atde
ath=
77.7
0ye
ars
0.28
year
ssho
rter
lifes
pans
for
thos
ebo
rnin
2nd
quar
ter
0.32
year
slon
ger
lifes
pans
for
thos
ebo
rnin
4th
quar
ter
Add
ition
alre
sults
show
nfo
rsp
ecifi
cca
uses
ofde
ath.
Aust
ralia:
Mea
nag
eof
deat
h=
78.0
0ye
arsf
orth
ose
born
in2n
dqu
arte
r;m
ean
age
ofde
ath=
77.6
5ye
arsf
orth
ose
born
in4t
hqu
arte
rB
ritis
him
mig
rant
sbor
nN
ov.-
Jan.
have
age
ofde
ath
0.36
year
shig
her
than
nativ
es.T
hose
imm
igra
ntsb
orn
Mar
-May
have
age
ofde
ath
0.26
year
slow
erth
anA
ustr
alia
nna
tives
.
(con
tinue
don
next
page
)
1360 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Air
qual
ity,in
fant
mor
talit
y,an
dth
eC
lean
Air
Act
of19
70(C
hay
and
Gre
enst
one,
2003
a,b)
Cou
nty-
leve
lmor
talit
yan
dna
talit
yda
ta19
69-1
974.
Ann
ual
mon
itor
leve
ldat
aon
tota
lsu
spen
ded
part
icle
s(T
SP)f
rom
EPA
.N=
501
coun
ty-y
ears
.
Cle
anA
irA
ctim
pose
dre
gula
tions
onpo
llute
rsin
coun
tiesw
ithT
SPco
ncen
trat
ions
exce
edin
gfe
dera
lce
iling
s.U
sed
nona
ttai
nmen
tsta
tus
asan
inst
rum
entf
orch
ange
sin
TSP
.A
lsous
edre
gres
sion
disc
ontin
uity
met
hods
toex
amin
eeff
ecto
fTSP
onin
fant
mor
talit
y.
1%de
clin
ein
TSP
pollu
tion
resu
ltsin
0.5%
decl
ine
inin
fant
mor
talit
yra
te.M
oste
ffec
tsdr
iven
byre
duct
ion
ofde
aths
occu
rrin
gw
ithin
1m
onth
ofbi
rth.
Air
pollu
tion
and
infa
nthe
alth
:w
hatc
anw
ele
arn
from
Cal
iforn
ia’s
rece
ntex
peri
ence
?(C
urri
ean
dN
eide
ll,20
05)
Dat
aon
pollu
tion
from
Cal
iforn
iaE
PA.In
divi
dual
-lev
elin
fant
mor
talit
yda
tafr
omC
alifo
rnia
vita
lsta
tistic
s19
89-2
000.
n=
206,
353.
Est
imat
edlin
ear
mod
elst
hat
appr
oxim
ate
haza
rdm
odel
s,w
here
the
risk
ofde
ath
isde
fine
dov
erw
eeks
oflif
e,an
dle
ngth
oflif
eis
cont
rolle
dfo
rw
itha
flex
ible
nonp
aram
etri
csp
line.
Con
trol
led
for
pren
atal
and
post
nata
lpol
lutio
nex
posu
re,w
eath
er,ch
ild’s
age,
and
num
erou
soth
erch
ildan
dfa
mily
char
acte
rist
ics,
asw
ella
smon
th,
year
,an
dzi
pco
defixe
deff
ects
.E
xam
ined
effec
tsof
ozon
e(O
3),
carb
onm
onox
ide
(CO
),an
dpa
rtic
ulat
em
atte
r(P
M10
)on
infa
ntm
orta
lity.
1.1
unit
redu
ctio
nin
post
nata
lexp
osur
eto
CO
(the
actu
alre
duct
ion
that
occu
rred
inC
Ain
the
1990
s)sa
ved
991
infa
ntliv
es,−
4.6%
decr
ease
inin
fant
mor
talit
yra
te.N
ost
atist
ical
lysig
nifica
nteff
ects
ofpr
enat
alex
posu
reto
any
ofth
epo
lluta
nts.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1361
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Pren
atal
expo
sure
tora
dioa
ctiv
efa
llout
(fro
mC
hern
obyl
)and
scho
olou
tcom
esin
Swed
en(A
lmon
det
al.,
2009
).A
llSw
edes
born
1983
-198
8,n=
562,
637.
Dat
aon
radi
atio
nfr
omSw
edish
Geo
logi
calS
urve
yat
pari
shle
vel.
Thr
eeem
piri
cals
trat
egie
s:(1
)C
ohor
tcom
pari
sons
—C
ompa
red
coho
rtsi
nut
ero
befo
re,du
ring
,an
daf
ter
Che
rnob
ylw
ithpa
rtic
ular
focu
son
coho
rt8-
25w
eeks
post
conc
eptio
n.(2
)With
in-c
ohor
tcom
pari
sons
-U
sege
ogra
phic
vari
atio
nin
leve
lsof
expo
sure
.D
efine
4re
gion
s,R
0(le
aste
xpos
ure)
,R
1,R
2,R
3.(3
)Diff
-in-
diff
siblin
gco
mpa
riso
n—C
ompa
red
thos
eex
pose
dto
radi
atio
nin
uter
o8-
25to
siblin
gsw
how
ere
not.
Res
ultsfr
om
(1):
Prob
abili
tyof
qual
ifyin
gfo
rhi
ghsc
hool
redu
ced
by0.
2%fo
rco
hort
inut
ero
duri
ngra
diat
ion;
by0.
6%fo
rco
hort
inut
ero
8-25
.G
rade
sred
uced
by0.
4%fo
rco
hort
inut
ero
duri
ngra
diat
ion;
by0.
6%fo
rco
hort
inut
ero
8-25
.R
esultsfr
om
(2):
(R3
rela
tive
toR
0)Pr
obab
ility
ofqu
alify
ing
for
high
scho
olre
duce
dby
3.6%
for
coho
rtin
uter
odu
ring
radi
atio
nin
R3;
by4%
for
coho
rtin
uter
o8-
25in
R3.
Gra
desr
educ
edby
3%fo
rco
hort
inut
ero
duri
ngra
diat
ion
inR
3;by
5.2%
for
coho
rtin
uter
o8-
25in
R3.
Res
ultsfr
om
(3):
(R3
rela
tive
toR
0)D
iffer
ence
inpr
obab
ility
ofqu
alify
ing
for
high
scho
olbe
twee
nsib
lings
incr
ease
dby
6%in
R3.
Diff
eren
cein
grad
esbe
twee
nsib
lings
incr
ease
dby
8%in
R3.
(con
tinue
don
next
page
)
1362 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Air
pollu
tion
and
infa
nthe
alth
:le
sson
sfro
mN
ewJe
rsey
(Cur
rie
etal
.,20
09).
Pollu
tion
data
from
EPA
.In
divi
dual
-lev
elda
taon
infa
ntbi
rths
and
deat
hsfr
omth
eN
ewJe
rsey
Dep
artm
ento
fH
ealth
1989
-200
3.n=
283,
393
for
mot
her
fixe
deff
ects
mod
els.
Mot
her
fixe
deff
ects
toes
timat
eim
pact
ofex
posu
reto
pollu
tion
(dur
ing
and
afte
rbi
rth)
onin
fant
heal
thou
tcom
es.A
irpo
llutio
nm
easu
red
from
air
qual
itym
onito
rsan
das
signe
dto
each
child
base
don
hom
ead
dres
sofm
othe
r.A
lsoin
clud
edin
tera
ctio
nsb/
nva
riab
lefo
rpo
llutio
nex
posu
rean
dm
ater
nals
mok
ing
asw
ella
soth
erm
ater
nalc
hara
cter
istic
s.C
ontr
olle
dfo
rw
eath
er,po
llutio
nm
onito
rlo
catio
ns,tim
etr
ends
,se
ason
aleff
ects
,an
dot
her
back
grou
ndch
arac
teri
stic
s.
1un
itch
ange
inm
ean
CO
expo
sure
duri
ngla
sttr
imes
ter
ofpr
egna
ncy
decr
ease
save
rage
birt
hw
eigh
tby
0.5%
,in
crea
ses
likel
ihoo
dof
low
birt
hw
eigh
tby
8%,an
dde
crea
sesg
esta
tion
by0.
2%.
1un
itch
ange
inm
ean
CO
expo
sure
infirs
t2w
eeks
afte
rbi
rth
incr
ease
slik
elih
ood
ofin
fant
mor
talit
yby
2.5%
.E
stim
ated
that
a1
unit
decr
ease
inm
ean
CO
expo
sure
infirs
t2
wee
ksaf
ter
birt
hw
ould
save
17.6
per
100,
000
lives
.E
ffec
tsof
CO
expo
sure
onin
fant
heal
that
birt
har
e2-
6tim
esla
rger
for
smok
ersa
ndm
othe
rsw
hoar
eov
erag
e35
.E
ffec
tsof
PM10
(par
ticul
ate
mat
ter)
and
ozon
ear
eno
tcon
siste
ntly
signi
fica
ntac
ross
the
spec
ifica
tions
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1363
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Feta
lexp
osur
esto
toxi
cre
leas
esan
din
fant
heal
th(C
urri
eet
al.,
2009
)To
xic
Rel
ease
Inve
ntor
yda
taat
coun
ty-y
ear
leve
lmat
ched
toco
unty
-yea
rle
veln
atal
ityan
din
fant
mor
talit
yda
ta.
N=
5279
.
Mul
tivar
iate
regr
essio
nof
infa
nthe
alth
outc
omes
onam
ount
ofto
xic
rele
ases
inea
chco
unty
and
year
,co
ntro
lling
for
dem
ogra
phic
and
soci
o-ec
onom
icch
arac
teri
stic
s,m
othe
rdr
inki
ngor
smok
ing
duri
ngpr
egna
ncy,
coun
tyem
ploy
men
t,an
dco
unty
and
year
fixe
deff
ects
.C
ompa
red
effec
tsof
deve
lopm
enta
ltox
inst
oot
her
toxi
ns,an
d“f
ugiti
ve”
air
rele
ases
to“s
tack
”ai
rre
leas
es(s
ince
emiss
ions
that
goup
asm
oke
stac
kar
em
ore
likel
yto
betr
eate
din
som
ew
ay,
and
henc
ew
illaff
ectt
hose
inth
evi
cini
tyle
ss).
An
addi
tiona
ltho
usan
dpo
unds
per
squa
rem
ileof
allt
oxic
rele
ases
lead
sto:
0.02
%de
crea
sein
leng
thof
gest
atio
n0.
04%
decr
ease
inbi
rth
wei
ght
0.1%
incr
ease
inpr
obab
ility
oflo
wbi
rth
wei
ght
0.8%
incr
ease
inpr
obab
ility
ofve
rylo
wbi
rth
wei
ght
1.3%
incr
ease
inin
fant
mor
talit
yLa
rger
effec
tsfo
rde
velo
pmen
talt
oxin
sand
for
fugi
tive
air
rele
ases
than
for
othe
rto
xins
orfo
rst
ack
air
rele
ases
.A
2-SD
incr
ease
inle
adre
leas
esde
crea
sesg
esta
tion
by0.
02%
and
decr
ease
sbir
thw
eigh
tby
0.05
%.A
2-SD
incr
ease
inca
dmiu
mre
leas
esde
crea
sesg
esta
tion
by0.
03%
,de
crea
ses
birt
hw
eigh
tby
0.07
%,in
crea
sesp
roba
bilit
yof
low
birt
hw
eigh
tby
1.2%
,in
crea
sesp
roba
bilit
yof
very
low
birt
hw
eigh
tby
1.4%
,an
din
crea
sesi
nfan
tmor
talit
yby
5%.Si
mila
rre
sults
for
tolu
ene,
epic
hlor
ohyd
rin.
Red
uctio
nsin
rele
ases
over
1988
-199
9ca
nac
coun
tfor
3.9%
ofth
ere
duct
ion
inin
fant
mor
talit
yov
erth
esa
me
time
peri
od.
(con
tinue
don
next
page
)
1364 Douglas Almond and Janet Currie
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Tra
ffic
cong
estio
nan
din
fant
heal
th:ev
iden
cefr
omE
-ZPa
ss(C
urri
ean
dW
alke
r,20
09)
Bir
thre
cord
sfor
Penn
sylv
ania
and
New
Jers
ey,19
94-2
003.
Dat
aon
hous
ing
pric
esfo
rN
ewJe
rsey
.N=
727,
954
for
diff
-in-
diff
.N=
232,
399
for
mot
her
fixe
deff
ects
.
Iden
tifica
tion
due
toin
trod
uctio
nof
elec
tron
icto
llco
llect
ion
(E-Z
Pass
),w
hich
redu
ced
traffi
cco
nges
tion
and
mot
orve
hicl
eem
issio
nsin
the
vici
nity
ofhi
ghw
ayto
llpl
azas
.D
iffer
ence
-in-
differ
ence
,co
mpa
ring
infa
ntsb
orn
tom
othe
rsliv
ing
near
toll
plaz
asto
infa
nts
born
tom
othe
rsliv
ing
near
busy
road
way
s(bu
taw
ayfr
omto
llpl
azas
),be
fore
and
afte
rin
trod
uctio
nof
E-Z
Pass
.A
lsoes
timat
edim
pact
sofe
xpos
ure
toE
-ZPa
sson
infa
nthe
alth
usin
gm
othe
rfixe
deff
ects
met
hods
.C
ontr
olle
dfo
rva
riou
sbac
kgro
und
char
acte
rist
ics,
year
and
mon
thfixe
deff
ects
,an
dpl
aza-
spec
ific
time
tren
ds.A
lsoes
timat
edim
pact
ofE
-ZPa
ssin
trod
uctio
non
hous
ing
pric
esne
arth
eto
llpl
aza.
Res
ultsfr
om
Diff
eren
ce-i
n-D
iffer
ence
Met
hod
:R
educ
tions
intr
affic
cong
estio
nge
nera
ted
byin
trod
uctio
nof
E-Z
Pass
redu
ced
inci
denc
eof
prem
atur
ebi
rth
by10
.8%
and
low
birt
hw
eigh
tby
11.8
%fo
rch
ildre
nof
mot
hers
livin
gw
/in
2km
ofa
toll
plaz
a.Fo
rth
ose
livin
gw
/in
3km
ofa
toll
plaz
a,eff
ects
are
7.3%
for
prem
atur
ebi
rths
and
8.4%
for
low
birt
hw
eigh
t.Si
mila
rre
sults
usin
gm
othe
rfixe
deff
ects
.N
oeff
ects
ofE
-ZPa
ssin
trod
uctio
non
hous
ing
pric
esor
dem
ogra
phic
com
posit
ion
ofm
othe
rsliv
ing
near
toll
plaz
as.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1365
Table4(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Cau
tion,
Dri
vers
!Chi
ldre
nPr
esen
t.T
raffi
c,Po
llutio
n,an
dIn
fant
Hea
lth(K
nitt
elet
al.,
2009
).T
raffi
cda
tafr
omth
efr
eew
aype
rfor
man
cem
easu
rem
ents
yste
m.Po
llutio
nda
tafr
omE
PA.B
irth
data
from
Cal
iforn
iaD
ept.
ofPu
blic
Hea
lth20
02-2
006.
N=
373,
800.
IVus
ing
traffi
csh
ocks
(due
toac
cide
ntso
rro
adcl
osur
es)t
oin
stru
men
tfor
air
pollu
tion.
Pref
erre
dsp
ecifi
catio
nsin
clud
etr
affic
flow
,de
lays
,an
din
tera
ctio
nsbe
twee
ntr
affic
and
wea
ther
.
Not
e:on
lyre
sults
from
the
IVre
gres
sion
are
repo
rted
here
.1
unit
decr
ease
inex
posu
reto
part
icul
ate
mat
ter
(PM
10)l
eads
to5%
decr
ease
inin
fant
mor
talit
yra
te(s
aves
14liv
espe
r10
0,00
0bi
rths
).
Unl
esso
ther
wise
note
d,on
lyre
sults
signi
fica
ntat
5%le
vela
rere
port
ed.
1366 Douglas Almond and Janet Currie
sample size prevented detection of all but the largest effects (see Section 3.1). Usinga comparable sample size, Behrman and Rosenzweig (2004) found the schooling ofidentical female twins was nearly one-third of a year longer for a pound increase in birthweight (454 grams), with relatively imprecise effects on adult BMI or wages.
In light of the above power concerns, Currie and Moretti (2007) matched mothersto their sisters in half a million birth records from California. Here, low birth weightwas found to have statistically significant negative impacts on educational attainmentand the likelihood of living in a wealthy neighborhood. However, the estimatedmagnitudes of the main effects were more modest: low birth weight increased thelikelihood of living in a poor neighborhood by 3% and reduced educational attainmentapproximately one month on average. Like Conley and Bennett (2001), the relationshipwas substantially stronger for the interaction between low birth weight and being bornin poor neighborhoods.
In a sample of Norwegian twins, Black et al. (2007) also found long-term effects ofbirth weight, but did not detect any heterogeneity in the strength of this relationshipby parental socioeconomic status.11 Oreopoulos et al. (2008) find similar results forCanada and Lin and Liu (2009) find positive long term effects of birth weight inTaiwan. Royer (2009) found long-term health and educational effects within Californiatwin pairs, with a weaker effect of birth weight than several other studies, esp. Blacket al. (2007). Responsive investments could account for this discrepancy if they differedbetween California and elsewhere (within twin pairs). Alternatively, there may be morehomogeneity with respect to socioeconomic status in Scandinavia than in California. Asdescribed in Section 3, Royer (2009) analyzed investment measures directly with theECLS-B data, concluding that there was no evidence of compensatory or reinforcinginvestments (see Section 2.2).
Following a literature in demography on seasonal health effects, Doblhammer andVaupel (2001) and Costa and Lahey (2005) focused on the potential long-term healtheffects of birth season. A common finding is that in the northern hemisphere, peopleborn in the last quarter of the year have longer life expectancies than those born inthe second quarter. Both the availability of nutrients can vary seasonally (particularlyhistorically), as does the likelihood of common infections (e.g., pneumonia). Therefore,either nutrition or infection could drive this observed pattern. Almond (2006) focused onprenatal exposure to the 1918 Influenza Pandemic, estimating that children of infectedmothers were 15% less likely to graduate high school and wages were between 5 and9% lower. Kelly (2009) found negative effects of prenatal exposure to 1957 “Asianflu” in Britain on test scores, though the estimated magnitudes were relatively modest.Interestingly, while birth weight was reduced by flu exposure, this effect appears to beindependent of the test score effect. Finally, Field et al. (2009) found that prenatal iodine
11 Royer (2009) notes that Black et al. (2007) find a “negligible effect of birth weight on high school completion for the1967-1976 birth cohort, but for individuals born between 1977 and 1986, the estimate is nearly six times as large”.
Human Capital Development before Age Five 1367
supplementation raised educational attainment in Tanzania by half a year of schooling,with larger impacts for girls.
4.1.2. Economic shocksA second set of papers considers economic shocks around the time of birth. Here, healthin adulthood tends to be the focus (not human capital), and findings are perhaps lessconsistent than in the studies of nutrition and infection described above. Van Den Berget al. (2006)’s basic result is that adult survival in the Netherlands is reduced for thoseborn during economic downturns. In contrast, Cutler et al. (2007) detected no long termmorbidity effects in the Health and Retirement Survey data for cohorts born during theDustbowl era of 1930s. Banerjee et al. (forthcoming) found that shocks to the productivecapacity of French vineyards did not have detectable effects on life expectancy or healthoutcomes, but did reduce height in adulthood. Baten et al. (2007) related variations ingrain prices in the decade of birth to numeracy using an ingenious measure based on“age heaping” in the British Censuses between 1851 and 1881. Persons who are morenumerate are less likely to round their ages to multiples of 5 or 10. They find that childrenborn in decades with high grain prices were less numerate by this index.
4.1.3. Air pollutionThe third strand of the literature examines the effect of pollution on fetal health.Epidemiological studies have demonstrated links between very severe pollution episodesand mortality: one of the most famous focused on a “killer fog” in London, Englandand found dramatic increases in cardiopulmonary mortality (Logan and Glasg, 1953).Previous epidemiological research on the effects of moderate pollution levels on prenatalhealth suggests negative effects but have produced inconsistent results. Cross-sectionaldifferences in ambient pollution are usually correlated with other determinants of fetalhealth, perhaps more systematically than with nutritional or disease exposures consideredabove. Many of the pollution studies have minimal (if any) controls for these potentialconfounders. Banzhaf and Walsh (2008) found that high-income families move out ofpolluted areas, while poor people in-migrate. These two groups are also likely to providediffering levels of (non-pollution) investments in their children, so that fetuses and infantsexposed to lower levels of pollution may tend to receive, e.g., better quality prenatalcare. If these factors are unaccounted for, this would lead to an upward bias in estimates.Alternatively, certain pollution emissions tend to be concentrated in urban areas, andindividuals in urban areas may be more educated and have better access to health care,factors that may improve health. Omitting these factors would lead to a downward bias,suggesting the overall direction of bias from confounding is unclear.
Two studies by Chay and Greenstone (2003a,b) address the problem of omittedconfounders by focusing on “natural experiments” provided by the implementation ofthe Clean Air Act of 1970 and the recession of the early 1980s. Both the Clean Air Actand the recession induced sharper reductions in particulates in some counties than in
1368 Douglas Almond and Janet Currie
others, and they use this exogenous variation in levels of pollution at the county-yearlevel to identify its effects. They estimate that a one unit decline in particulates caused bythe implementation of the Clean Air Act (recession) led to between five and eight (fourand seven) fewer infant deaths per 100,000 live births. They also find some evidence thatthe decline in Total Suspended Particles (TSPs) led to reductions in the incidence of lowbirth weight. However, only TSPs were measured at that time, so that they could notstudy the effects of other pollutants. And the levels of particulates studied by Chay andGreenstone are much higher than those prevalent today; for example, PM10 (particulatematter of 10 microns or less) levels have fallen by nearly 50% from 1980 to 2000.
Several recent studies consider natural experiments at more recently-encounteredpollution levels. For example, Currie et al. (2009) use data from birth certificates inNew Jersey in which they know the exact location of the mothers residence, and birthsto the same mother can be linked. They focus on a sample of mothers who live nearpollution monitors and show that variations in pollution from carbon monoxide (whichcomes largely from vehicle exhaust) reduces birth weight and gestation. Currie andWalker (2009) exploit a natural experiment having to do with introduction of electronictoll collection devices (E-ZPass) in New Jersey and Pennsylvania. Since much of thepollution produced by automobiles occurs when idling or accelerating back to highwayspeed, electronic toll collection greatly reduces auto emissions in the vicinity of a tollplaza. Currie and Walker (2009) compare mothers near toll plazas to those who live nearbusy roadways but further from toll plazas and find that E-ZPass increased birth weightand gestation. They show that they obtain similar estimates following mothers over timeand estimating mother fixed effects models. These papers are notable in part because ithas proven more difficult to demonstrate effects of pollution on fetal health than on infanthealth, as discussed further below. Hence, it appears that being in utero may be protectiveagainst at some forms of toxic exposure (such as particulates) but not others.
This literature on the effects of air pollution is closely related to that on smoking.
Smoking is, after all, the most important source of indoor air pollution. Medicalresearch has shown that nicotine constricts the oxygen supply to the fetus, so thereis an obvious mechanism for smoking to affect infant health. Indeed, there is nearunanimity in the medical literature that smoking is the most important preventable causeof low birth weight. Economists have focused on ways to address heterogeneity in otherdeterminants of birth outcomes that are likely associated with smoking. Tominey (2007)found that relative to a conventional multivariate control specification, roughly one-
third of the harm from smoking to birth weight is explained by unobservable traits ofthe mother. Moreover, the reduction in birth weight from smoking was substantiallylarger for low-SES mothers. In a much larger sample, Currie et al. (2009) showedthat smoking significantly reduced birth weight, even when comparisons are restrictedto within-sibling differences. Moreover, Currie et al. (2009) document a significantinteraction effect between exposure to carbon monoxide exposure and infant health
Human Capital Development before Age Five 1369
in the production of low birth weight, which may help explain the heterogeneity inbirth weight effects reported by Tominey (2007). Aizer and Stroud (2009) note thatimpacts of smoking on birth weight are generally much smaller in sibling comparisonsthan in OLS and matching-based estimates. Positing that attenuation bias is accentuatedin the sibling comparisons, Aizer and Stroud (2009) use serum cotinine levels as aninstrument for measurement error in smoking and find that sibling comparisons yieldsimilar birth weight impacts (around 150 g). Lien and Evans (2005) use increases instate excise taxes as an instrument for smoking and find large effects of smoking onbirth weight (182 g) as a result. Using propensity score matching, Almond et al. (2005)document a large decrease in birth weight from prenatal smoking (203 g), but argue thatthis weight decrease is weakly associated with alternative measures of infant health, suchas prematurity, APGAR score, ventilator use, and infant mortality.
Some recently-released data will enable new research on smoking’s short and long-term effects. In 2005, twelve states began using the new US Standard Certificate ofLive Birth (2003 revision). Along with other new data elements (e.g., on surfactantreplacement therapy), smoking behavior is reported by trimester. It will be useful toconsider whether smoking’s impact on birth weight varies by trimester, and also whethersmoking is more closely tied to other measures of newborn health if it occurs early versuslate in pregnancy. Second, there is relatively little research by economists on the long-term effects of prenatal exposure to smoking. Between 1990 and 2003, there were 113increases in state excise taxes on cigarettes (Lien and Evans, 2005).12 Since 2005, theAmerican Community Survey records both state and quarter of birth, permitting linkageof these data to the changes in state excise taxes during pregnancy.
Almond et al. (2009) examine the effect of pollution from the Chernobyl disasteron the Swedish cohort that was in utero at the time of the disaster. Since the pathof the radiation was very well measured, they can compare affected children to thosewho were not affected as well as to those born in the affected areas just prior to thedisaster. They find that in the affected cohort those who suffered the greatest radiationexposure were 3% less likely to qualify for high school, and had 6% lower math grades(the measure closest to IQ). The estimated effects were much larger within families.A possible interpretation is that cognitive damage from Chernobyl was reinforced byparents. An alternative hypothesis is that there were fixed characteristics of families thatwere correlated both with poor outcomes and with exposure. It might be the case, forexample, that pregnant mothers of higher socioeconomic status were more likely to stayindoors or move temporarily to reduce radiation exposure.13
To summarize, the recent “fetal origins” literature in economics finds substantialeffects of prenatal health on subsequent human capital and health. As we discuss in12 Some states enacted earlier excise taxes: the “average state tax rate increased from 5.7 cents in 1964 to 15.5 cents in
1984” (Farrelly et al., 2003); high 1970s inflation can be an additional potential source of identification as excise taxeswere set nominally.
13 That said, the radiation dosages estimated in Sweden were not believed harmful at the time of the Chernobyl accident.
1370 Douglas Almond and Janet Currie
Section 5, this suggests a positive role for policies that improve human capital by affectingthe birth endowment. That is, despite being congenital (i.e. present from birth), thisresearch indicates that the birth endowment is malleable in ways that shape human capital.This finding has potentially radical implications for public policy since it suggests that oneof the more effective ways to improve children’s long term outcomes might be to targetwomen of child bearing age in addition to focusing on children after birth.
4.2. Early childhood environmentIt would be surprising to find that a very severe shock in early childhood (e.g., a headinjury, or emotional trauma) had no effect on an individual. Therefore, a more interestingquestion from the point of view of research is how developmental linkages operatingat the individual level affect human capital formation in the aggregate. To answer thisquestion, we need to know how many children are affected by negative early childhoodexperiences that could plausibly exert persistent effects? How big and long-lasting arethe effects of less severe early childhood shocks relative to more severe shocks? Takentogether, how much of the differences in adult attainments might be accounted for bythings that happen to children between birth and age five? Furthermore, how are theselinkages between shocks and outcomes mediated or moderated by third factors? Forexample, is the effect of childhood lead exposure on subsequent test scores stronger forfamilies of lower socioeconomic status (i.e. is the interaction with SES an important one)and if so why? Alternatively, is the effect of injury mediated by health status, or is thecausal pathway a direct one to cognition?
We might also wish to know how parents respond to early childhood shocks. Todate, there has been less focus on this question in the early childhood period than inthe prenatal period, perhaps because it seems less plausible to hope to uncover a “pure”biological effect of a childhood shock given that children are embedded in families and insociety. However, this embeddedness opens the possibility that a richer set of behavioralresponses—of the kind considered by economists—might be at play. Furthermore, earlychildhood admits a wider set of environmental influences than the prenatal period. Forexample, abuse in early childhood can be distinguished from malnutrition, a distinctionmore difficult for the in utero period, and these may have quite different effects.
We define early childhood as starting at birth and ending at age five. From anempirical standpoint, early childhood so defined offers advantages and disadvantagesover analyses that focus on the prenatal period. Mortality is substantially lower duringearly childhood than in utero, which reduces the scope for selective attrition caused byenvironmental shocks to affect the composition of survivors. On the other hand, it isunlikely that environmental sensitivity during early childhood tapers discontinuously atany precise age (including age five). From a refutability perspective, we cannot makesharp temporal comparisons of a cohort “just exposed” to a shock during early childhoodto a neighboring cohort “just unexposed” by virtue of its being too old to be sensitive.
Human Capital Development before Age Five 1371
Moreover, it will often be difficult to know a priori whether prenatal or postnatal exposureis more influential.14 Thus, studies of early childhood exposures tend to emphasize cross-sectional sources of variation, including that at the geographic and individual level. Thestudies reviewed in this section focus on tracing out the relationships between events inearly childhood and future outcomes, and are summarized in Table 5.
4.2.1. InfectionsInsofar as specific health shocks are considered, infections are the most commonlystudied. In epidemiology, long-term health effects of infections—and the inflammationresponse they trigger—has been explored extensively, e.g. Crimmins and Finch (2006).Outcomes analyzed by economists include height, health status, educational attainment,test scores, and labor market outcomes. The estimated impacts tend to be large. Usinggeographic differences in hookworm infection rates across the US South, Bleakley (2007)found that eradication after 1910 increased literacy rates but did not increase the amountof completed schooling, except for Black children. The literacy improvement was muchlarger among Blacks than Whites, and stronger among women then men. The returnto education increased substantially, and Bleakley (2007) estimated that hookworminfection throughout childhood reduced wages in adulthood by as much as 40%. Caseand Paxson (2009) focussed on reductions in US childhood mortality from typhoid,malaria, measles, influenza, and diarrhea during the first half of the 20th Century. Theyfound that improvements in the disease environment in one’s state of birth were mirroredby improved cognitive performance at older ages, but like Bleakley (2007), this effectdid not seem to operate through increased years of schooling. However, the estimatedcognitive impacts in Case and Paxson (2009) were not robust to the inclusion of state-specific time trends in their models.
Chay et al. (2009) found that reduced exposure to pneumonia and diarrhea in earlychildhood among Blacks during the late 1960s raised subsequent AFQT and NAEPscores towards those of Whites. Changes in postneonatal mortality rates (dominated byinfections) explained between 50% and 80% of the (large) reduction in the Black-WhiteAFQT gap. Finally, Bozzoli et al. (forthcoming) highlight that in developing countries,high average mortality rates cause the selection effect of early childhood mortality tooverwhelm the “scarring” effect. Thus, the positive relationship between early childhoodhealth and subsequent human capital may be absent in analyses that do not account forselective attrition in high mortality settings.
4.2.2. Health statusMany of the studies reviewed in Table 5 investigate the link between health in childhoodand future cognitive or labor market outcomes. These studies can be viewed as a subset of
14 For example, early postnatal exposure to Pandemic influenza apparently had a larger impact on hearing than didprenatal flu exposure (Heider, 1934).
1372 Douglas Almond and Janet Currie
Table5
Impa
ctsof
early
child
hood
shocks
onlatero
utcomes.
Stud
yan
dda
taStud
yde
sign
Results
Child
hood
physical
andmen
talh
ealth
Men
talh
ealth
inch
ildho
odan
dhu
man
capi
tal(
Cur
rie
and
Stab
ile,
2006
).D
ata
from
NLS
CY
and
NLS
Y.C
anad
ian
NLS
CY
:da
taon
child
ren
aged
4-11
in19
94.
Men
talh
ealth
scre
enin
gin
1994
;ou
tcom
esm
easu
red
in20
02.
n=
5604
.U
SN
LSY
:da
taon
child
ren
aged
4-11
in19
94.
Men
talh
ealth
scor
eav
erag
edov
er19
90-1
994.
n=
3758
.
Sibl
ing
fixe
deff
ects
.H
yper
activ
ityan
dag
gres
sion
scor
esba
sed
on“s
cree
ner”
ques
tions
aske
dof
all
child
ren.
Est
imat
edeff
ecto
fhy
pera
ctiv
ityon
grad
ere
petit
ion,
read
ing
and
mat
hsc
ores
,sp
ecia
led
ucat
ion,
and
delin
quen
cy.
Con
trol
led
for
indi
vidu
alba
ckgr
ound
char
acte
rist
ics.
Est
imat
edsa
me
mod
elom
ittin
gch
ildre
nw
ithot
her
lear
ning
disa
bilit
iesb
esid
esth
ose
inth
em
ain
expl
anat
ory
vari
able
.
1unit
chan
gein
hyper
activi
tysc
ore
:in
crea
sesp
roba
bilit
yof
grad
ere
tent
ion
by10
-12%
inbo
thU
San
dC
anad
ade
crea
sesm
ath
scor
esby
0.04
-0.0
7SD
inbo
thU
San
dC
anad
ain
crea
sesp
roba
bilit
yof
bein
gin
spec
iale
dby
11%
inU
Sde
crea
sesr
eadi
ngsc
ores
by0.
05SD
inU
S1
unit
chan
gein
conduct
dis
ord
ersc
ore
(in
the
US
only
):in
crea
sesp
roba
bilit
yof
grad
ere
tent
ion
by10
%de
crea
sesm
ath
scor
esby
0.02
SDde
crea
sesr
eadi
ngsc
ores
by0.
03SD
1unit
chan
gein
aggre
ssio
nsc
ore
(in
Can
ada
only
):de
crea
sesp
roba
bilit
yth
ata
yout
hag
ed16
-19
isin
scho
olby
4% high
depr
essio
nsc
ores
incr
ease
prob
abili
tyof
grad
ere
tent
ion
by10
%in
both
US
and
Can
ada
nosig
nifica
nteff
ects
ofin
tera
ctio
nbe
twee
nm
enta
lhea
lthsc
ores
and
inco
me
orm
ater
nale
duca
tion. (con
tinue
don
next
page
)
Human Capital Development before Age Five 1373
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Dise
ase
and
deve
lopm
ent:
Evi
denc
efr
omho
okw
orm
erad
icat
ion
inth
eA
mer
ican
Sout
h(B
leak
ley,
2007
).D
ata
onho
okw
orm
infe
ctio
nra
tesf
rom
the
Roc
kefe
ller
Sani
tary
Com
miss
ion
surv
eysf
or19
10-1
914.
Cen
susd
ata
from
IPU
MS
for
1880
-199
0.n=
115
stat
e-ye
ars.
Iden
tifica
tion
due
todi
ffer
ent
pre-
erad
icat
ion
hook
wor
min
fect
ion
rate
sin
differ
ents
tate
s.C
ompa
rein
divi
dual
sbor
nbe
fore
and
afte
rer
adic
atio
nca
mpa
igns
fund
edby
the
Roc
kefe
ller
Sani
tary
Com
miss
ion
(RSC
).A
ctiv
eye
arso
fRSC
wer
e19
10-1
915.
Con
sider
edco
ntem
pora
neou
seffec
tson
child
ren
asw
ella
slon
g-te
rmeff
ects
onad
ultw
ages
and
educ
atio
nala
ttai
nmen
t.C
ontr
olle
dfo
rge
ogra
phic
and
year
fixe
deff
ects
,as
wel
lass
ome
indi
vidu
alch
arac
teri
stic
s.In
regr
essio
nsfo
rco
ntem
pora
neou
seff
ects
,ge
ogra
phic
units
are
stat
eec
onom
icar
eas.
Inre
gres
sions
for
long
-ter
meff
ects
,ge
ogra
phic
units
are
stat
esof
birt
h.
Conte
mpora
neo
useff
ects
ofin
fect
ionson
childre
n:
1SD
incr
ease
inla
gged
hook
wor
min
fect
ion
asso
ciat
edw
ith:
0.18
-0.2
5SD
decr
ease
insc
hool
enro
llmen
t0.
21-0
.28
SDde
crea
sein
full-
time
scho
olen
rollm
ent
0.1
SDde
crea
sein
liter
acy
Res
ults
robu
stto
incl
usio
nof
stat
e-ye
arfixe
deff
ects
,co
ntro
lling
for
mea
n-re
vers
ion
insc
hool
ing,
and
usin
gst
ate-
leve
linf
ectio
nra
tes.
Larg
ereff
ects
for
blac
ksth
anfo
rw
hite
s.N
oco
ntem
pora
neou
seffec
tson
adul
ts.
Long
-ter
meff
ects
ofin
fect
ionsin
childhood
on
adults:
Bei
ngin
fect
edw
ithho
okw
orm
inon
e’sc
hild
hood
lead
sto
are
duct
ion
inw
ages
of43
%an
da
decr
ease
inre
turn
sto
scho
olin
gby
5%.80
%re
duct
ion
inw
ages
due
toho
okw
orm
infe
ctio
nsex
plai
ned
byre
duce
dre
turn
sto
scho
olin
g.B
eing
infe
cted
with
hook
wor
min
one’
schi
ldho
odle
adst
oa
redu
ctio
nin
occu
patio
nali
ncom
esc
ore
by23
%an
da
decr
ease
inD
unca
n’sS
ocio
-Eco
nom
icIn
dex
by42
%.
(con
tinue
don
next
page
)
1374 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Adu
lthe
alth
and
child
hood
dise
ase
(Boz
zoli
etal
.,fo
rthc
omin
g)D
ata
onhe
ight
from
the
Eur
opea
nC
omm
unity
Hou
seho
ldPa
nel,
the
Hea
lthSu
rvey
ofE
ngla
nd,an
dth
eN
atio
nalH
ealth
Inte
rvie
wSu
rvey
inth
eU
Sfo
rin
divi
dual
sbo
rnin
1950
-198
0.D
ata
onpo
st-n
eona
talm
orta
lity
from
the
Wor
ldH
ealth
Org
aniz
atio
n.n=
316
coun
try-
year
s.A
lsous
edda
taon
wom
en’s
heig
hts
from
the
inte
rnat
iona
lsys
tem
ofD
emog
raph
ican
dH
ealth
Surv
eys
onw
omen
aged
15-4
9in
mor
eth
an40
coun
trie
sin
the
late
1990
s-20
00s.
Dat
aon
infa
ntm
orta
lity
from
the
Uni
ted
Nat
ions
popu
latio
ndi
visio
n.n=
1514
coun
try-
year
s.
Ana
lyze
dre
latio
nshi
pb/
npo
st-n
eona
talm
orta
lity
(PN
M,
deat
haf
ter
28da
ysan
dbe
fore
firs
tbi
rthd
ay)a
ndad
ulth
eigh
t.PN
Mis
am
easu
reof
child
hood
heal
then
viro
nmen
t.O
LSre
gres
sion
with
adul
thei
ghta
sout
com
eva
riab
le,co
ntro
lling
for
coun
try
and
year
fixe
deff
ects
and
atim
etr
end.
Also
cont
rolle
dfo
rne
onat
alm
orta
lity
rate
sand
GD
Pin
year
and
coun
try
ofbi
rth.
Con
sider
edm
orta
lity
from
pneu
mon
ia,
inte
stin
aldi
seas
e,co
ngen
ital
anom
alie
s,an
dot
her
caus
esse
para
tely
.
Surv
ivor
sare
expe
cted
tobe
posit
ivel
yse
lect
edre
lativ
eto
thos
ew
hodi
ed,bu
tmay
still
best
unte
dby
illne
ss.In
poor
coun
trie
s,th
ese
lect
ion
effec
tdom
inat
es,w
here
asin
rich
coun
trie
s(w
ithlo
wm
orta
lity
rate
s)th
est
untin
geff
ect
dom
inat
es.O
vera
llco
rrel
atio
nb/
nPN
Man
dav
erag
ead
ult
heig
htof
the
sam
eco
hort
ina
give
nco
untr
y=−
0.79
.In
the
US,
PNM
was
3xla
rger
than
inSw
eden
in19
70.T
hisd
iffac
coun
tsfo
r20
-30%
ofth
e2-
cmdi
ffin
aver
age
heig
htb/
n30
-yr-
old
Am
eric
ansa
ndSw
edes
in20
00.A
fter
cont
rolli
ngfo
rPN
M,no
rela
tions
hip
b/n
adul
thei
ghta
ndG
DP
inth
eye
aran
dco
untr
yof
birt
h.B
igge
stde
term
inan
tofd
iffer
ence
sin
PNM
rate
sacr
ossc
ount
ries
ism
orta
lity
from
inte
stin
aldi
seas
e,fo
llow
edby
mor
talit
yfr
ompn
eum
onia
.O
utof
the
4de
term
inan
tsof
PNM
,on
lym
orta
lity
from
pneu
mon
iaha
sasig
nifica
ntne
gativ
eeff
ecto
nad
ulth
eigh
t. (con
tinue
don
next
page
)
Human Capital Development before Age Five 1375
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Stat
ure
and
stat
us:he
ight
,ab
ility
and
labo
rm
arke
tout
com
es(C
ase
and
Paxs
on,20
08a)
.D
ata
from
NC
DS
(on
coho
rtof
indi
vidu
als
born
inw
eek
of3/
3/19
58in
Bri
tain
),B
CS,
NLS
Y(o
nsib
lings
inth
eU
S),an
dFr
agile
Fam
ilies
and
Chi
ldW
ell-
Bei
ngSt
udy
(on
youn
gch
ildre
nin
the
US)
.N
CD
S:n=
9155
;B
CS:
n=
9003
;N
LSY
:n=
13,8
84(t
otal
,no
tsib
lings
);Fr
agile
Fam
ilies
:n=
2150
.
Ana
lyze
dre
latio
nshi
pb/
nad
ult
heig
htan
dco
gniti
veab
ility
asa
pote
ntia
lexp
lana
tion
for
the
signi
fica
ntan
dpo
sitiv
ere
latio
nshi
pb/
nad
ulth
eigh
tand
earn
ings
that
isob
serv
ed.A
dult
heig
htis
apr
oxy
for
earl
ych
ildho
odhe
alth
cond
ition
s.O
LSre
gres
sion
ofco
gniti
vete
stsc
ores
onhe
ight
-for
-age
(HFA
)z-
scor
es,c
ontr
ollin
gfo
rba
ckgr
ound
and
dem
ogra
phic
vari
able
s(us
ing
NC
DS,
BC
San
dfr
agile
fam
ilies
data
).Sa
me
regr
essio
nw
ithm
othe
rfixe
deff
ects
usin
gN
LSY
79da
ta.O
LSre
gres
sion
oflo
gho
urly
earn
ings
onad
ulth
eigh
t,co
ntro
lling
for
cogn
itive
test
scor
esat
ages
7,10
,an
d11
,an
dot
her
back
grou
ndan
dde
mog
raph
icva
riab
les.
Child’s
hei
ghtan
dco
gnitiv
ete
stsc
ore
sat
age
3:1
SDin
crea
sein
HFA
z-sc
ore
linke
dto
a0.
05-0
.1SD
incr
ease
inPP
VT
scor
e.C
hild’s
hei
ghtan
dco
gnitiv
ete
stsc
ore
sat
ages
5-10
:(r
epor
ting
only
mot
her
fixe
deff
ects
resu
lts)
1SD
incr
ease
inH
FAz-
scor
elin
ked
toin
crea
sed
PIA
Tm
ath
scor
e,PI
AT
read
ing
reco
gniti
onsc
ore,
and
PIA
Tre
adin
gco
mpr
ehen
sion
scor
eby
0.03
SD.H
eigh
tdoe
snot
expl
ain
differ
ence
sin
test
scor
esac
ross
raci
algr
oups
.A
dult
hei
ght,
earn
ings
,an
dco
gnitiv
ete
stsc
ore
s:In
clus
ion
ofco
gniti
vete
stsc
ores
atag
es7,
10,an
d11
mak
esth
eco
effici
ento
nad
ulth
eigh
tins
igni
fica
ntfo
rpr
edic
ting
hour
lyw
ages
.H
eigh
tdiff
eren
ceb/
nm
enan
dw
omen
does
nota
ccou
ntfo
rth
edi
ffer
ence
inea
rnin
gs.
(con
tinue
don
next
page
)
1376 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Hei
ght,
heal
than
dco
gniti
vefu
nctio
nat
olde
rag
es(C
ase
and
Paxs
on,20
08b)
.D
ata
from
HR
Son
men
and
wom
enag
ed50
and
olde
rfr
om19
96to
2004
.n=
72,2
58.
OLS
regr
essio
nof
heal
than
dco
gniti
onm
easu
reso
nad
ult
heig
ht,co
ntro
lling
for
age,
surv
eyw
ave,
race
and
sex.
Adu
lthe
ight
isa
prox
yfo
rea
rly
child
hood
heal
than
dnu
triti
on.A
lsoes
timat
edm
odel
sinc
ludi
ngco
ntro
lsfo
rch
ildho
odhe
alth
(sel
f-re
port
ed),
com
plet
eded
ucat
ion,
and
adu
mm
yfo
rem
ploy
men
tin
aw
hite
-col
lar
occu
patio
n.
1in
incr
ease
inhe
ight
asso
ciat
edw
ith:
0.9%
incr
ease
inde
laye
dw
ord
reca
llsc
ore
0.3%
incr
ease
inpr
obab
ility
ofbe
ing
able
toco
untb
ackw
ards
0.3%
incr
ease
inpr
obab
ility
ofkn
owin
gth
eda
te1.
4%de
crea
sein
depr
essio
nsc
ore
0.4%
decr
ease
inhe
alth
stat
ussc
ale
(1=
exce
llent
,5=
poor
)C
hild
hood
heal
th,co
mpl
eted
educ
atio
n,an
dem
ploy
men
tin
aw
hite
-col
lar
occu
patio
nal
lpos
itive
lyre
late
dto
adul
tco
gniti
vefu
nctio
nan
dhe
alth
.In
clus
ion
ofth
ese
cont
rols
mak
esth
eco
effici
ents
onad
ulth
eigh
tins
igni
fica
ntex
cept
for
the
case
sofd
elay
edw
ord
reca
llan
dde
pres
sion.
The
role
ofch
ildho
odhe
alth
for
the
inte
rgen
erat
iona
ltra
nsm
issio
nof
hum
anca
pita
l:ev
iden
cefr
omad
min
istra
tive
data
(Sal
man
dSc
hunk
,20
08).
Adm
inist
rativ
eda
tafr
omth
ede
part
men
tof
heal
thse
rvic
esin
the
city
ofO
snab
ruec
k,G
erm
any,
colle
cted
duri
ngoffi
cial
scho
olen
tran
cem
edic
alex
amin
atio
nson
child
ren
aged
6be
twee
n20
02-2
005.
Sibl
ing
sam
ple:
n=
947.
321
child
ren
had
atle
asto
nepa
rent
w/
colle
gede
gree
.
Sibl
ing
fixe
deff
ects
toes
timat
eim
pact
ofhe
alth
cond
ition
son
cogn
itive
and
verb
alab
ility
atsc
hool
entr
ance
.C
ontr
olle
dfo
rin
divi
dual
back
grou
ndch
arac
teri
stic
s.U
sed
the
Oax
aca
(197
3)de
com
posit
ion
and
ano
npar
amet
ric
deco
mpo
sitio
nto
estim
ate
how
muc
hof
the
gap
b/n
child
ren
ofco
llege
-edu
cate
dpa
rent
sand
child
ren
ofle
ssed
ucat
edpa
rent
scan
beex
plai
ned
bych
roni
cch
ildho
odhe
alth
cond
ition
s.
Men
talh
ealth
cond
ition
sred
uce
cogn
itive
abili
tyby
10%
for
the
who
lesa
mpl
e,by
9%fo
rco
llege
-edu
cate
dsa
mpl
e,by
11%
for
less
-edu
cate
dsa
mpl
e.A
sthm
are
duce
scog
nitiv
eab
ility
by8%
for
less
-edu
cate
dsa
mpl
eon
ly.M
enta
lhea
lthco
nditi
onsr
educ
eve
rbal
abili
tyby
11%
for
the
who
lesa
mpl
e,an
dby
13%
for
the
less
-edu
cate
dsa
mpl
e.H
ealth
cond
ition
sex
plai
n18
%of
the
gap
inco
gniti
veab
ility
and
65%
ofth
ega
pin
lang
uage
abili
tyb/
nch
ildre
nof
colle
ge-e
duca
ted
and
less
-edu
cate
dpa
rent
s.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1377
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Long
-ter
mec
onom
icco
stso
fps
ycho
logi
calp
robl
emsd
urin
gch
ildho
od(S
mith
and
Smith
,20
08)
Dat
afr
omPS
IDon
siblin
gs(w
itha
spec
ials
uppl
emen
tdes
igne
dby
the
auth
orsi
nth
e20
07w
ave,
aski
ngre
tros
pect
ive
ques
tions
abou
tchi
ldho
odhe
alth
).Sa
mpl
eco
nsist
sofs
iblin
gch
ildre
nof
the
orig
inal
part
icip
ants
who
wer
eat
leas
t16
in19
68.Si
blin
gch
ildre
nw
ere
atle
ast2
5in
2005
.
Sibl
ing
fixe
deff
ects
toes
timat
eth
eim
pact
ofre
port
ing
havi
ngha
dch
ildho
odps
ycho
logi
calp
robl
ems
(mea
sure
dby
depr
essio
n,dr
ugor
alco
hola
buse
orot
her
psyc
holo
gica
lpro
blem
sbef
ore
age
17)o
nla
ter
life
soci
o-ec
onom
icou
tcom
es.C
ontr
olle
dfo
rin
divi
dual
child
hood
phys
ical
illne
sses
(ast
hma,
diab
etes
,al
lerg
icco
nditi
ons,
and
man
yot
hers
)as
wel
lasf
amily
back
grou
ndch
arac
teri
stic
s.
Hav
ing
had
psyc
holo
gica
lpro
blem
sdur
ing
child
hood
lead
sto
:20
%re
duct
ion
inad
ulte
arni
ngs(
$10,
400
less
per
year
;$1
7,53
4le
ssfa
mily
asse
ts)
redu
ctio
nin
num
bero
fwee
ksw
orke
dby
5.76
wee
kspe
ryea
r11
ppre
duct
ion
inlik
elih
ood
ofge
ttin
gm
arri
ed33
.5pp
redu
ctio
nin
educ
atio
nala
ttai
nmen
t(m
eans
not
repo
rted
,so
can’
tcal
cula
tere
lativ
eeff
ects
izes
)C
ost
sofch
ildhood
psy
cholo
gic
alpro
ble
ms:
Life
time
cost
:$3
00,0
00lo
ssin
fam
ilyin
com
e;$3
.2tr
illio
nco
stfo
ral
ltho
seaff
ecte
d
Ear
lylif
ehe
alth
and
cogn
itive
func
tion
inol
dag
e(C
ase
and
Paxs
on,20
09)
Reg
ion-
leve
lhist
oric
alda
taon
mor
talit
yfr
omin
fect
ious
dise
ases
,as
wel
last
otal
infa
ntm
orta
lity.
Dat
afo
r19
00-1
936
from
Gra
ntM
iller
’sda
taar
chiv
eon
NB
ER
web
site.
Dat
afo
r19
37-1
950
from
vita
lsta
tistic
sdoc
umen
ts.D
ata
onla
ter
life
outc
omes
from
Hea
lthan
dR
etir
emen
tStu
dyfo
r19
96-2
004
onm
enan
dw
omen
aged
50-9
0.n=
60,0
00.
Iden
tifica
tion
due
tova
riat
ion
inm
orta
lity
rate
sacr
osst
ime
and
regi
ons.
OLS
regr
essio
nof
late
rlif
eou
tcom
eson
log
mor
talit
yra
tesf
rom
vari
ousi
nfec
tious
dise
ases
inye
arof
birt
han
din
2nd
year
oflif
ein
regi
onof
birt
h.C
ontr
olle
dfo
rag
e,se
x,ra
ce,an
dcu
rren
tCen
susr
egio
nof
resid
ence
.
Dec
reas
ein
infa
ntm
orta
lity
byha
lfdu
ring
2nd
year
oflif
eas
soci
ated
w/
anin
crea
sein
dela
yed
wor
dre
call
scor
eby
0.1
SD.Si
gnifi
cant
nega
tive
impa
ctso
ftyp
hoid
,in
flue
nza,
and
diar
rhea
mor
talit
yin
2nd
year
oflif
eon
dela
yed
wor
dre
call
scor
e.(M
eans
notr
epor
ted
soca
n’tc
alcu
late
rela
tive
effec
tsiz
es).
Wea
ker
asso
ciat
ions
b/n
dise
ase
mor
talit
yan
dab
ility
toco
untb
ackw
ards
.N
osig
nifica
ntim
pact
sofd
iseas
em
orta
lity
and
over
allm
orta
lity
duri
ngye
arof
birt
h,on
cem
orta
lity
in2n
dye
arof
life
isin
clud
ed,on
eith
erof
the
late
rlif
eou
tcom
es.R
esul
tsno
trob
ustt
oad
ding
Cen
sus
regi
on-s
peci
fic
time
tren
ds.
(con
tinue
don
next
page
)
1378 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Chi
ldhe
alth
and
youn
gad
ult
outc
omes
(Cur
rie
etal
.,20
10)
Adm
inist
rativ
eda
taon
publ
iche
alth
insu
ranc
ere
cord
sfro
mth
eC
anad
ian
prov
ince
ofM
anito
baon
child
ren
born
b/n
1979
and
1987
,fo
llow
edun
til20
06.
n=
50,0
00.
Sibl
ing
fixe
deff
ects
ofth
elo
ngte
rmeff
ects
ofhe
alth
prob
lem
sat
vari
ousc
hild
ages
cont
rolli
ngfo
rhe
alth
atbi
rth
(with
birt
hw
eigh
tan
dco
ngen
itala
nom
alie
s).K
eyex
plan
ator
yva
riab
lese
xam
ined
are:
asth
ma,
maj
orin
juri
es,
AD
HD
,co
nduc
tdiso
rder
s,an
dot
her
maj
orhe
alth
prob
lem
sat
ages
0-3,
4-8,
9-13
,an
d14
-18.
Key
outc
ome
vari
able
s:ac
hiev
emen
ton
stan
dard
ized
test
onla
ngua
gear
tsin
grad
e12
,w
heth
erch
ildto
okco
llege
-pre
para
tory
mat
hco
urse
sin
high
scho
ol,w
heth
erch
ildis
ingr
ade
12by
age
17,an
dw
elfa
repa
rtic
ipat
ion
afte
rbe
com
ing
elig
ible
at18
.
Res
ultsre
port
edbel
owar
eonly
those
that
contr
olf
or
hea
lth
pro
ble
msat
alla
gegro
ups:
An
addi
tiona
lmaj
orhe
alth
cond
ition
atag
es0-
3/4-
8/14
-18
incr
ease
sthe
prob
abili
tyof
bein
gon
wel
fare
by10
%/9
%/3
1%.
Effec
tsof
maj
orhe
alth
cond
ition
saty
oung
erag
eson
educ
atio
nalo
utco
mes
nots
igni
fica
ntw
hen
cont
rolli
ngfo
rhe
alth
atag
es14
-18.
AD
HD
/conduct
dis
ord
ersdia
gnosi
s:at
ages
0-3/
4-8/
9-13/14
-18
decr
ease
spro
babi
lity
ofbe
ing
ingr
ade
12by
age
17by
4%/1
0%/1
7%/1
9%;at
ages
4-8/
9-13/14
-18
incr
ease
spro
babi
lity
ofbe
ing
onw
elfa
reby
38%
/44%
/109
%;
atag
es4-
8/9-
13/14
-18
decr
ease
spro
babi
lity
ofta
king
aco
llege
prep
mat
hcl
assb
y11
%/2
5%/3
5%;at
ages
4-8/
9-13/14
-18
decr
ease
slite
racy
scor
eby
0.15
SD/0
.23
SD/0
.27
SD.
Effec
tsof
asth
ma
atyo
unge
rag
esno
tsta
tistic
ally
signi
fica
nton
cehe
alth
atag
es14
-18
isco
ntro
lled
for.
Maj
orin
jury
:at
ages
0-3/
14-1
8in
crea
sesp
roba
bilit
yof
bein
gon
wel
fare
by7%
/9%
;at
ages
9-13/14
-18
decr
ease
spr
obab
ility
ofbe
ing
ingr
ade
12by
age
17by
2%/2
%;at
ages
9-13/14
-18
decr
ease
spro
babi
lity
ofta
king
aco
llege
prep
mat
hcl
assb
y6%
/8%
;at
ages
9-13/14
-18
decr
ease
slite
racy
scor
eby
0.03
SD/0
.03
SD.C
hild
ren
who
have
am
ajor
phys
ical
heal
thco
nditi
onan
dth
enre
cove
rdo
noth
ave
signi
fica
ntad
vers
eou
tcom
es.C
hild
ren
with
men
talh
ealth
cond
ition
s,ch
ildre
nw
ithm
ajor
heal
thco
nditi
onsa
tage
s14
-18,
and
thos
ew
ithco
nditi
onst
hatp
ersis
tfor
mul
tiple
age
peri
odss
uffer
wor
seou
tcom
es.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1379
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
impa
ctof
child
hood
heal
thon
adul
tlab
orm
arke
tout
com
es(S
mith
,20
09)
Dat
aon
siblin
gsfr
omPS
ID.
Chi
ldho
odhe
alth
mea
sure
das
ase
lf-re
port
edre
tros
pect
ive
heal
thin
dex
rega
rdin
ghe
alth
atag
esyo
unge
rth
an17
.A
dult
outc
omes
mea
sure
din
1999
.n=
2248
.
Sibl
ing
fixe
deff
ects
toes
timat
eim
pact
ofse
lf-re
port
edre
tros
pect
ive
child
hood
heal
thst
atus
(on
a5-
poin
tsca
le)b
efor
eag
e16
onad
ulte
arni
ngs,
empl
oym
ent,
educ
atio
n,m
arita
lst
atus
,et
c.,co
ntro
lling
for
dem
ogra
phic
sand
fam
ilyba
ckgr
ound
.
Bet
ter
heal
thdu
ring
child
hood
incr
ease
sadu
ltfa
mily
inco
me
by24
%,in
crea
sesa
dult
fam
ilyw
ealth
by20
0%(r
elat
ive
tom
ean=
$200
0),an
din
crea
sesa
dult
earn
ings
by25
%.
Bet
ter
heal
thdu
ring
child
hood
incr
ease
spro
babi
lity
ofha
ving
wor
ked
the
year
befo
reby
5.4
pp(m
ean
notr
epor
ted)
.A
bout
2/3
ofov
eral
lim
pact
ofpo
orch
ildho
odhe
alth
onad
ultf
amily
inco
me
ispr
esen
tata
ge25
;th
ere
mai
ning
1/3
isdu
eto
aslo
wer
grow
thpa
thaf
ter
age
25du
eto
poor
child
hood
heal
th.A
bout
1/2
ofov
eral
lim
pact
ofpo
orch
ildho
odhe
alth
onin
divi
dual
earn
ings
ispr
esen
tata
ge25
;th
ere
mai
ning
1/2
isdu
eto
aslo
wer
grow
thpa
thaf
ter
age
25du
eto
poor
child
hood
heal
th.N
ost
atist
ical
lysig
nifica
ntim
pact
sofc
hild
hood
heal
thon
educ
atio
nala
ttai
nmen
t.
The
effec
tofc
hild
hood
cond
uct
diso
rder
onhu
man
capi
tal(
Vuj
icet
al.,
2008
)D
ata
from
the
Aus
tral
ian
Twin
Reg
ister
ontw
insb
orn
betw
een
1964
and
1971
.D
ata
colle
cted
in19
89-1
990
and
1996
-200
0n=
5322
twin
s;22
50id
entic
altw
ins.
Twin
fixe
deff
ects
toes
timat
eim
pact
ofch
ildho
odco
nduc
tdi
sord
er(m
easu
red
byva
riou
sin
dica
tors
)bas
edon
diag
nost
iccr
iteri
afr
omps
ychi
atry
oned
ucat
iona
latt
ainm
enta
ndcr
imin
albe
havi
orin
adul
thoo
d.C
ontr
olle
dfo
rbi
rth
wei
ght,
timin
gof
the
onse
tofc
ondu
ctdi
sord
ers,
and
othe
rfa
mily
and
indi
vidu
alba
ckgr
ound
char
acte
rist
ics.
Con
duct
edse
para
tean
alys
esfo
ral
ltw
insa
ndid
entic
altw
ins.
Chi
ldho
odco
nduc
tdiso
rder
slea
dto
:5-
16%
decr
ease
inlik
elih
ood
ofhi
ghsc
hool
grad
uatio
n10
0-22
8%in
crea
sein
likel
ihoo
dof
bein
gar
rest
ed(m
ean=
0.07
)6-
68%
incr
ease
inlik
elih
ood
ofgr
ade
rete
ntio
n20
-60%
incr
ease
inlik
elih
ood
ofha
ving
3+jo
bqu
its(n
otsig
nifica
ntin
iden
tical
twin
sam
ple)
50-3
25%
incr
ease
inlik
elih
ood
ofte
lling
lies(
mea
n=
0.04
)37
-526
%in
crea
sein
likel
ihoo
dof
goin
gto
jail
(mea
n=
0.01
9)E
arlie
roc
curr
ence
ofco
nduc
tdiso
rder
hasl
arge
rne
gativ
eeff
ects
than
late
roc
curr
ence
.
(con
tinue
don
next
page
)
1380 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Cau
sesa
ndco
nseq
uenc
esof
earl
ylif
ehe
alth
(Cas
ean
dPa
xson
,20
10a)
Dat
afr
omN
CD
S,B
CS,
PSID
,W
hite
hall
IISt
udy
(long
itudi
nal
stud
yof
Bri
tish
civi
lser
vant
sb/n
ages
34an
d71
,co
llect
edfr
om19
85to
2001
),H
RS,
and
NLS
Y79
.Se
eC
ase
and
Paxs
on(2
008a
,b)f
orm
ore
info
onth
eda
tase
ts.n=
11,6
48(N
CD
S),
11,1
81(B
CS)
,63
,995
(PSI
D),
29,7
74(W
hite
hall
II),
66,2
69(H
RS)
,n=
3200
-46,
000
(NLS
Y79
).
Mul
tivar
iate
regr
essio
nof
educ
atio
nala
ttai
nmen
t,em
ploy
men
t,lo
gea
rnin
gs,an
dse
lf-re
port
edhe
alth
stat
usan
dco
gniti
vefu
nctio
non
adul
thei
ght
inin
ches
(pro
xyfo
rea
rly
child
hood
heal
th)u
sing
5lo
ngitu
dina
ldat
ase
ts.C
ontr
olle
dfo
rag
e,et
hnic
ity,se
x,an
dsu
rvey
wav
e.U
sed
siblin
gfixe
deff
ects
inN
LSY
79to
unde
rsta
ndw
hat
aspe
ctso
fear
lylif
ehe
alth
adul
the
ight
capt
ures
—re
gres
sed
child
ren’
stes
tsco
res,
grad
ele
vel,
self-
perc
eptio
nin
scho
ol,an
dch
ildho
odhe
alth
outc
omes
onch
ildre
n’sH
AZ
.
Res
ults
from
NC
DS,
BC
S,PS
ID,W
hite
hall
II,an
dH
RS:
One
inch
ofhe
ight
isas
soci
ated
with
:0.
05-0
.16
mor
eye
ars
ofsc
hool
ing;
0.2-
0.6
ppin
crea
sein
likel
ihoo
dof
empl
oym
ent;
0.01
2-0.
028
incr
ease
inav
erag
eho
urly
earn
ings
for
men
;0.
007-
0.02
7in
crea
sein
aver
age
hour
lyea
rnin
gsfo
rw
omen
.A
4-in
incr
ease
inhe
ight
lead
sto:
8%de
crea
sein
prob
abili
tyof
long
-sta
ndin
gill
ness
;40
%de
crea
sein
prob
abili
tyof
disa
bilit
y;4%
ofSD
decr
ease
inde
pres
sion
scor
e.R
esul
tsfr
omN
LSY
79(r
epor
ting
siblin
gfixe
deff
ects
only
)in
dica
teth
at1
pt.in
crea
sein
HA
Zle
adst
o:0.
3%in
crea
sein
PIA
Tm
ath
scor
e;0.
1%in
crea
sein
PIA
Tre
adin
gre
cogn
ition
scor
e;0.
2%in
crea
sein
PIA
Tre
adin
gco
mpr
ehen
sion
scor
e;0.
9%in
crea
sein
digi
tals
pan
test
scor
e;0.
4%in
crea
sein
PPV
Tsc
ore.
2%de
crea
sein
likel
ihoo
dth
atch
ildis
inap
prop
riat
egr
ade
leve
lfor
age
(age
s6-1
4).1%
incr
ease
inch
ilddo
ing
scho
olw
ork
quic
kly;
1%in
crea
sein
child
rem
embe
ring
thin
gsea
sily;
0.9%
incr
ease
into
tals
chol
astic
com
pete
nce
scor
e.8%
decr
ease
inlik
elih
ood
that
child
has
limiti
ngem
otio
nal/
neur
olog
ical
cond
ition
;0.
105
incr
ease
inbi
rth
wei
ghtz
-sco
re(m
ean
notr
epor
ted)
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1381
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
long
reac
hof
child
hood
heal
than
dci
rcum
stan
ce:ev
iden
cefr
omth
eW
hite
hall
IISt
udy
(Cas
ean
dPa
xson
,20
10b)
Dat
afr
omth
eW
hite
hall
IISt
udy
(see
abov
efo
rde
scri
ptio
n).
n=
10,3
08.
Sinc
em
ostW
hite
hall
IIst
udy
part
icip
ants
belo
ngto
the
high
est
occu
patio
nalc
lass
es,es
timat
edth
ese
lect
ion
effec
tusin
gda
tafr
omN
CD
San
dB
CS
for
com
pari
son.
OLS
regr
essio
nsof
initi
alpl
acem
enta
ndpr
omot
ion
inW
hite
hall
onhe
ight
,fa
mily
back
grou
ndch
arac
teri
stic
s,ed
ucat
iona
latt
ainm
ent,
and
othe
rin
divi
dual
char
acte
rist
ics.
Also
estim
ated
indi
vidu
alfixe
deff
ects
and
firs
t-di
ffer
ence
dre
gres
sions
offu
ture
prom
otio
nson
self-
repo
rted
heal
th,an
dof
futu
rehe
alth
ongr
ade
inW
hite
hall.
Sele
ctiv
ityof
Whi
teha
llII
stud
yat
tenu
ates
the
impa
ctof
fath
er’s
soci
alcl
asso
nva
riou
sout
com
esre
lativ
ees
timat
esth
atw
ould
beob
tain
edin
the
full
popu
latio
n.O
LSre
gres
sions
indi
cate
signi
fica
ntco
rrel
atio
nsbe
twee
nfa
mily
back
grou
nd(f
athe
r’sso
cial
clas
s),in
itial
plac
emen
tand
likel
ihoo
dof
prom
otio
n.Si
gnifi
cant
rela
tions
hip
betw
een
initi
alpl
acem
ent
and
prom
otio
n,he
ight
,an
dse
lf-re
port
edhe
alth
.A
lthou
ghso
me
ofth
eeff
ects
ofhe
alth
onpl
acem
enta
ndpr
omot
ion
are
med
iate
dby
educ
atio
nala
ttai
nmen
t,th
ere
isan
inde
pend
ent
effec
tofh
ealth
.In
divi
dual
fixe
deff
ects
estim
ates
indi
cate
that
coho
rtm
embe
rsw
hore
port
“exc
elle
nt”
or“v
ery
good
”he
alth
are
13%
mor
elik
ely
tobe
prom
oted
(rel
ativ
eto
low
est
grad
e).N
oeff
ecto
fgra
dein
Whi
teha
llon
futu
rehe
alth
.
(con
tinue
don
next
page
)
1382 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Hom
een
vironm
ent
The
impa
ctof
mat
erna
lalc
ohol
and
illic
itdr
ugus
eon
child
ren’
sbe
havi
orpr
oble
ms:
evid
ence
from
the
child
ren
ofth
eN
atio
nal
Long
itudi
nalS
urve
yof
Yout
h(C
hatt
erji
and
Mar
kow
itz,20
00)
Dat
afr
omth
e19
88,19
92,an
d19
94w
aves
ofN
LSY-
Chi
ldsu
rvey
son
child
ren
who
wer
e4-
14yr
sold
duri
ngth
esu
rvey
year
s.n=
6194
child
ren
into
tal
sam
ple;
n=
2498
mot
hers
who
have
siblin
gsin
sam
ple;
n=
7546
obsf
orch
ildre
nw
hoha
ve2
or3
obse
rvat
ions
insa
mpl
e.
Est
imat
edim
pact
ofm
ater
nal
subs
tanc
eab
use
onch
ildre
n’s
beha
vior
prob
lem
susin
gO
LS,IV
,ch
ild-s
peci
fic
&m
ater
nalf
amily
fixe
deff
ects
mod
els.
InIV
,us
edal
coho
land
illic
itdr
ugpr
ices
asin
stru
men
tsfo
rm
ater
nal
subs
tanc
eab
use.
Con
trol
led
for
num
erou
sdem
ogra
phic
and
back
grou
ndva
riab
les.
Beh
avio
rpr
oble
msm
easu
red
byth
eB
ehav
ior
Prob
lem
sInd
ex(h
ighe
rB
PIm
eans
mor
epr
oble
ms)
.
No
stat
istic
ally
signi
fica
ntre
latio
nshi
pbe
twee
nm
ater
nal
subs
tanc
eab
use
and
child
beha
vior
inIV
regr
essio
ns,bu
t1st
stag
ere
latio
nshi
pis
wea
k.Fi
xed
effec
tssu
gges
ttha
t1m
ore
drin
kco
nsum
edby
mot
her
inpa
stm
onth
incr
ease
schi
ldB
PIby
1%.
Mat
erna
lmar
ijuan
aus
ein
past
mon
thin
crea
sesc
hild
BPI
by7-
8%.
Mat
erna
lcoc
aine
use
inpa
stm
onth
incr
ease
schi
ldB
PIby
19%
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1383
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Pare
ntal
empl
oym
enta
ndch
ildco
gniti
vede
velo
pmen
t(R
uhm
,20
04).
Dat
afr
omth
eC
hild
ren
ofN
LSY
79fo
r19
86-1
996
(mot
hers
aged
29-3
8at
the
end
of19
95).
n=
3042
.
OLS
toes
timat
eth
eim
pact
ofm
ater
nalw
ork
inth
eye
arpr
ior
toch
ild’s
birt
h,an
din
the
firs
t4
year
sofc
hild
’slif
eon
child
cogn
itive
test
scor
esat
ages
3-4
and
5-6.
Use
da
rich
seto
fco
ntro
lsfo
rm
ater
nalb
ackg
roun
d,de
mog
raph
ic,an
dso
cio-
econ
omic
char
acte
rist
ics,
asw
ella
sass
essm
enty
ear
fixe
deff
ects
.M
ater
nalw
ork
mea
sure
dbo
thby
hour
sand
wee
ksw
orke
ddu
ring
the
year
.C
ontr
olle
dfo
rfa
mily
inco
me.
20m
ore
hrs
.w
ork
edea
chw
eek
bym
oth
erduri
ng
child’s
1stye
ar:
Dec
reas
ePP
VT
scor
eby
0.06
-0.1
0SD
(age
s3-4
)20
more
hrs
.w
ork
edea
chw
eek
bym
oth
erduri
ng
child’s
age
2-3:
Dec
reas
ePI
AT-
Rea
ding
scor
eby
0.06
-0.0
8SD
(age
s5-6
)D
ecre
ase
PIA
T-M
ath
scor
eby
0.05
-0.0
6SD
(age
s5-6
)R
esul
tsro
bust
toal
tern
ativ
esp
ecifi
catio
ns(m
easu
ring
empl
oym
entb
yw
eeks
wor
ked
orpa
rt-t
ime
vs.no
empl
oym
ent)
.M
ore
nega
tive
effec
tson
PPV
Tan
dPI
AT-
Mat
hsc
ores
for
boys
than
girl
s.M
ore
nega
tive
effec
tson
PIA
T-R
eadi
ngsc
ores
for
girl
stha
nbo
ys.M
ore
nega
tive
effec
tson
PPV
Tan
PIA
T-R
eadi
ngsc
ores
for
whi
test
han
blac
ks.M
ore
nega
tive
effec
tson
PIA
T-M
ath
scor
esfo
rbl
acks
than
whi
tes.
Mat
erni
tyle
ave,
earl
ym
ater
nal
empl
oym
enta
ndch
ildhe
alth
and
deve
lopm
enti
nth
eU
S(B
erge
ret
al.,
2005
)D
ata
from
the
Chi
ldre
nof
NLS
Y79
onch
ildre
nbo
rnb/
n19
88an
d19
96.n=
1678
.
OLS
and
prop
ensit
ysc
ore
mat
chin
gto
estim
ate
impa
ctof
mat
erna
lret
urn
tow
ork
with
in12
wee
ksof
birt
hon
child
heal
than
dde
velo
pmen
talo
utco
mes
.M
odel
sco
ntro
lfor
num
erou
sbac
kgro
und
char
acte
rist
icsa
swel
lass
tate
and
year
ofbi
rth
fixe
deff
ects
.Sa
mpl
elim
ited
tom
othe
rsw
how
orke
dpr
e-bi
rth.
Moth
erre
turn
ing
tow
ork
within
12w
eeks
ofgiv
ing
bir
thle
adsto
:2.
5%de
crea
sein
likel
ihoo
dof
wel
l-ba
byvi
sit12
.8%
decr
ease
inlik
elih
ood
ofan
ybr
east
feed
ing
40.4
%de
crea
sein
num
ber
ofw
eeks
brea
stfe
d4.
3%de
crea
sein
child
gett
ing
allD
PT/P
olio
imm
uniz
atio
nsR
esul
tsfr
omO
LSco
nsist
entw
ithre
sults
from
prop
ensit
ysc
ore
mat
chin
gm
etho
ds.La
rger
nega
tive
effec
tsof
retu
rnin
gto
wor
kfu
ll-tim
ew
ithin
12w
eeks
ofgi
ving
birt
hco
mpa
red
toan
yw
ork
atal
l.
(con
tinue
don
next
page
)
1384 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Evi
denc
efr
omm
ater
nity
leav
eex
pans
ions
ofth
eim
pact
ofm
ater
nalc
are
onea
rly
child
deve
lopm
ent(
Bak
eran
dM
illig
an,
2010
)D
ata
from
the
NLS
CY
inC
anad
aon
child
ren
upto
age
29m
onth
san
dth
eir
mot
hers
in19
98-2
003.
Iden
tifica
tion
due
tola
rge
expa
nsio
nsin
pare
ntal
leav
epo
licie
sin
Can
ada
onD
ecem
ber
31,20
00(f
rom
25to
50w
eeks
ofle
ave)
.O
LSre
gres
sion.
Key
expl
anat
ory
vari
able
isa
dum
my
for
birt
haf
ter
Dec
embe
r31
,200
0.O
utco
mes
incl
ude
child
deve
lopm
enti
ndic
ator
s,m
ater
nal
empl
oym
ent,
and
use
ofch
ildca
re.C
ontr
olle
dfo
rva
riou
sin
divi
dual
and
fam
ilyba
ckgr
ound
char
acte
rist
icsa
swel
lasl
abor
mar
ketv
aria
bles
.O
bser
vatio
nsfr
omQ
uebe
com
itted
due
toch
ange
sin
child
care
polic
ieso
ver
the
sam
etim
epe
riod
.O
bser
vatio
nsfr
omsin
gle-
pare
ntfa
mili
esom
itted
b/c
ofex
pans
ions
inbe
nefits
toth
ose
fam
ilies
.Su
bsam
ple
anal
ysis
for
wom
enw
hore
turn
edto
wor
kw
ithin
1ye
arof
child
’sbi
rth.
Exp
ansio
nsin
pare
ntal
leav
epo
licie
sled
toa
48-5
8%in
crea
sein
mon
thso
fmat
erna
lcar
edu
ring
the
1sty
ear
oflif
ean
da
25-2
9pp
decr
ease
inno
n-pa
rent
alca
re.N
ost
atist
ical
lysig
nifica
nteff
ects
ofpa
rent
alle
ave
expa
nsio
nson
child
deve
lopm
enti
ndic
ator
s(m
otor
/soc
iald
evel
opm
ent,
beha
vior
,ph
ysic
alab
ility
,co
gniti
vede
velo
pmen
t).
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1385
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
effec
tofe
xpan
sions
inm
ater
nity
leav
eco
vera
geon
child
ren’
slon
g-te
rmou
tcom
es(D
ustm
ann
and
Scho
nber
g,20
09)
Adm
inist
rativ
eda
taon
stud
ents
inG
erm
any
atte
ndin
gpu
blic
scho
ols
inH
esse
,B
avar
ia,an
dSc
hles
wig
-Hol
stei
n(3
stat
esin
Ger
man
y)fo
r20
02-0
3to
2005
-06.
n=
101,
257.
Adm
inist
rativ
eda
taon
soci
alse
curi
tyre
cord
sfor
Ger
man
indi
vidu
alsb
orn
b/n
July
1977
and
June
1980
.n=
140,
387
Iden
tifica
tion
due
to3
polic
yre
form
sin
Ger
man
yth
atex
pand
edun
paid
and
paid
mat
erni
tyle
ave
toes
timat
eim
pact
onad
ultw
ages
,em
ploy
men
t,an
ded
ucat
iona
lout
com
es.Fi
rst
refo
rmin
1979
incr
ease
dpa
idle
ave
from
2to
6m
onth
s.Se
cond
refo
rmin
1986
incr
ease
dpa
idle
ave
from
6to
10m
onth
s.Thi
rdre
form
in19
92in
crea
sed
unpa
idle
ave
from
18to
36m
onth
s.O
LSre
gres
sion,
com
pari
ngch
ildre
nbo
rnon
em
onth
befo
rean
daf
ter
each
refo
rm,c
ontr
ollin
gfo
rba
ckgr
ound
vari
able
sand
stat
efixe
deff
ects
.A
lsodi
ffer
ence
-in-
differ
ence
,co
mpa
ring
child
ren
born
befo
rean
daf
ter
leav
eex
pans
ions
with
child
ren
born
inth
esa
me
mon
ths
the
year
befo
re(h
ence
nota
ffec
ted
byex
pans
ions
).
Des
pite
signi
fica
ntde
lays
inm
ater
nalr
etur
nto
wor
kdu
eto
polic
yre
form
s,th
ere
are
nost
atist
ical
lysig
nifica
ntim
pact
sof
expa
nsio
nsin
mat
erni
tyle
ave
polic
ies(
paid
orun
paid
)on
any
long
-run
outc
omes
(wag
es,em
ploy
men
t,se
lect
ive
high
scho
olat
tend
ance
,gr
ade
rete
ntio
nan
dgr
ade
atte
ndan
ce).
(con
tinue
don
next
page
)
1386 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Chi
ldpr
otec
tion
and
adul
tcri
me:
usin
gin
vest
igat
oras
signm
entt
oes
timat
eca
usal
effec
tsof
fost
erca
re(D
oyle
,20
08)
Dat
afr
omco
mpu
teri
zed
crim
inal
hist
ory
syst
emfr
omth
eIl
linoi
sSt
ate
Polic
eon
arre
stsa
ndim
priso
nmen
tup
toag
e31
for
2000
-200
5,lin
ked
with
child
abus
ein
vest
igat
ion
data
onin
divi
dual
swho
are
18-3
5in
2005
.n=
23,2
54.
IVm
odel
soft
heeff
ecto
ffos
ter
care
plac
emen
ton
mea
sure
sof
crim
inal
activ
ity.Id
entifi
catio
ndu
eto
the
fact
that
child
prot
ectio
nca
sesa
rera
ndom
ized
toin
vest
igat
ors,
and
inve
stig
ator
sin
flue
nce
whe
ther
ach
ildis
plac
edin
fost
erca
reor
rem
ains
atho
me.
Mod
elal
low
sfor
trea
tmen
teffec
the
tero
gene
ity—
used
rand
omco
effici
entm
odel
.IV
whe
reth
ein
stru
men
tist
hein
vest
igat
or’s
prob
abili
tyof
usin
gfo
ster
plac
emen
trel
ativ
eto
the
othe
rin
vest
igat
ors.
Chi
ldre
non
the
mar
gin
for
plac
emen
tint
ofo
ster
care
are
1.5
times
mor
elik
ely
tobe
arre
sted
,2.
68tim
esm
ore
likel
yto
have
ase
nten
ceof
guilt
y/w
ithhe
ld,3.
41tim
esm
ore
likel
yto
bese
nten
ced
topr
ison
ifth
eyar
epl
aced
into
fost
erca
re.
Chi
ldre
non
the
mar
gin
are
likel
yto
incl
ude
Afr
ican
-Am
eric
ans,
girl
s,an
dyo
ung
adol
esce
nts.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1387
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Long
-ter
mco
nseq
uenc
esof
child
abus
ean
dne
glec
ton
adul
tec
onom
icw
ell-
bein
g(C
urri
ean
dW
idom
,20
09).
Sam
ple
ofab
used
/neg
lect
edch
ildre
nba
sed
onco
urt-
subs
tant
iate
dca
seso
fch
ildho
odph
ysic
alan
dse
xual
abus
ein
1967
-197
1in
one
Mid
wes
tern
met
ropo
litan
coun
ty.
Mal
trea
ted
child
ren
mat
ched
toco
ntro
lson
the
basis
ofse
x,ra
ce,
and
elem
enta
rysc
hool
clas
s/ho
spita
lofb
irth
.C
ases
rest
rict
edto
child
ren
11yr
sor
youn
ger.
Adu
ltou
tcom
esm
easu
red
atm
ean
age
41.U
sed
info
colle
cted
in19
89-9
5an
d20
03-0
4in
terv
iew
s.n=
1195
in19
89-9
5;n=
807
in20
03-0
4.M
atch
edsa
mpl
e(b
oth
mem
bers
ofm
atch
edpa
irin
terv
iew
eddu
ring
the
2w
aves
):n=
358.
Mul
tivar
iate
regr
essio
nw
ithke
yex
plan
ator
yva
riab
lebe
ing
adu
mm
yfo
rha
ving
been
mal
trea
ted.
Con
trol
led
for
dem
ogra
phic
and
fam
ilyba
ckgr
ound
char
acte
rist
ics,
asw
ell
asqu
arte
rof
year
attim
eof
inte
rvie
w.A
lsoes
timat
edm
odel
sse
para
tely
for
mal
esan
dfe
mal
esan
dfo
rth
esu
bsam
ple
ofpa
rtic
ipan
tsw
hose
fam
ilies
rece
ived
food
stam
psor
wel
fare
whe
nth
eyw
ere
child
ren.
Indiv
idual
sw
ho
wer
em
altr
eate
das
childre
n:
com
plet
e4.
3%le
ssye
arso
fsch
oolin
g(1
989-
95)
scor
e5.
3%lo
wer
onIQ
test
(198
9-95
)ha
ve24
%lo
wer
impu
ted
earn
ings
(200
3-04
)ar
e0.
52tim
esas
likel
yto
have
ask
illed
job
(198
9-95
)ar
e0.
46tim
esas
likel
yto
beem
ploy
ed(2
003-
04)
are
0.58
times
aslik
ely
toow
na
vehi
cle
(200
3-04
)Fo
rm
ales
,th
eon
lysig
nifica
nteff
ects
ofm
altr
eatm
enta
refo
rye
arso
fsch
oolin
gan
dha
ving
ask
illed
job.
For
fem
ales
,sig
nifica
ntne
gativ
eeff
ects
ofm
altr
eatm
ento
nye
arso
fsc
hool
ing,
IQte
stsc
ores
,im
pute
dea
rnin
gs,be
ing
empl
oyed
,ow
ning
aba
nkac
coun
t,ow
ning
ast
ock,
owni
nga
vehi
cle,
and
owni
nga
hom
e.E
ffec
tsla
rger
for
fem
ales
than
for
mal
es.
(con
tinue
don
next
page
)
1388 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Chi
ldm
altr
eatm
enta
ndcr
ime
(Cur
rie
and
Teki
n,20
06)
Dat
afr
omth
eN
atio
nal
Long
itudi
nalS
tudy
ofA
dole
scen
tH
ealth
(Add
Hea
lth).
Firs
twav
e:19
94-9
5;la
stw
ave:
2001
-200
2.n=
13,5
09(f
ulls
ampl
e),n=
928
(tw
inss
ampl
e).
OLS
,pr
open
sity
scor
em
atch
ing,
and
twin
fixe
deff
ects
met
hods
.E
stim
ate
the
impa
ctof
mal
trea
tmen
ton
crim
inal
activ
ity.
Con
trol
led
for
num
erou
sde
mog
raph
ican
dba
ckgr
ound
char
acte
rist
icsa
swel
lass
tate
fixe
deff
ects
.A
lsoco
ntro
lled
for
pare
ntal
repo
rtso
fchi
ldbe
ing
bad-
tem
pere
dor
havi
nga
lear
ning
disa
bilit
y.
Onl
yre
sults
that
are
signi
fica
ntin
all3
mai
nsp
ecifi
catio
ns(O
LS,pr
open
sity
scor
es,an
dtw
infixe
deff
ects
)are
repo
rted
.C
hildre
nw
ho
exper
ience
dan
ym
altr
eatm
entar
e:99
-134
%m
ore
likel
yto
com
mit
any
non-
drug
rela
ted
crim
e(m
ean=
0.10
9)28
8-48
9%m
ore
likel
yto
com
mit
abu
rgla
ry(m
ean=
0.00
9)11
3-18
1%m
ore
likel
yto
dam
age
prop
erty
(mea
n=
0.05
2)10
6-13
1%m
ore
likel
yto
com
mit
anas
saul
t(m
ean=
0.04
9)18
3-22
2%m
ore
likel
yto
com
mit
ath
eft>
$50
(mea
n=
0.01
8)76
-101
%m
ore
likel
yto
com
mit
any
hard
-dru
gre
late
dcr
ime
(mea
n=
0.08
5)96
-103
%m
ore
likel
yto
bea
crim
evi
ctim
(mea
n=
0.07
7)Pr
obab
ility
ofcr
ime
incr
ease
sifa
pers
onsu
ffer
smul
tiple
form
sofm
altr
eatm
ent.
Bei
nga
vict
imof
mal
trea
tmen
tdo
uble
sthe
prob
abili
tyth
atan
indi
vidu
alis
conv
icte
das
aju
veni
le.
Cost
-Ben
efit:
Est
imat
edco
stso
fcri
me
indu
ced
byab
use=
$8.8
-68.
6bi
llion
/yea
rE
stim
ated
cost
sofh
ome
visit
ing
prog
ram
s(th
atha
vebe
ensh
own
tore
duce
case
sofm
altr
eatm
entb
y50
%)=
$14
billi
on/y
ear.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1389
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
effec
tofm
ater
nald
epre
ssio
nan
dsu
bsta
nce
abus
eon
child
hum
anca
pita
ldev
elop
men
t(F
rank
and
Mea
ra,20
09)
Dat
afr
omch
ildre
nof
NLS
Y79
onch
ildre
nag
ed1-
5in
1987
.n=
1587
.
OLS
regr
essio
nof
the
impa
ctof
mat
erna
ldep
ress
ion
and
subs
tanc
eab
use
onm
ater
nali
nput
sand
child
ren’
sout
com
esin
grad
es1-
5an
d6-
9.C
ontr
olle
dfo
ra
rich
set
ofde
mog
raph
ican
dba
ckgr
ound
vari
able
sinc
ludi
ngfa
mily
mea
sure
sofp
aren
tand
siblin
gbe
havi
oran
dhe
alth
.Fo
rro
bust
ness
,es
timat
edm
odel
susin
gm
othe
rfixe
deff
ects
and
prop
ensit
ysc
ores
.
Effec
tson
mat
ernal
inputs
:M
ater
nald
epre
ssio
nle
adst
oa
0.2
SDde
crea
sein
emot
iona
lst
imul
atio
nsu
b-co
mpo
nent
ofth
eH
OM
Esc
ore
(age
s7-1
0,11
-14)
.M
ater
nals
ubst
ance
abus
ele
adst
oa
0.23
SDde
crea
sein
emot
iona
lstim
ulat
ion
sub-
com
pone
ntof
the
HO
ME
scor
e(a
ges1
1-14
).E
ffec
tson
child
outc
om
es:
Mat
erna
ldep
ress
ion
lead
sto
a0.
46SD
incr
ease
inbe
havi
oral
prob
lem
sind
ex(a
ges7
-10,
11-1
4).
Mat
erna
lalc
ohol
abus
ele
adst
oa
0.31
SDde
crea
sein
child
’sPI
AT
mat
hsc
ore
(age
s11-
14);
0.29
SDin
crea
sein
beha
vior
alpr
oble
msi
ndex
(age
s7-1
0);37
7%de
crea
sein
likel
ihoo
dof
ever
bein
gsu
spen
ded
orex
pelle
dat
any
age
(mea
n=
0.22
).
(con
tinue
don
next
page
)
1390 Douglas Almond and Janet Currie
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Environm
entalsho
cks
Env
ironm
enta
lpol
icy
asso
cial
polic
y?T
heim
pact
ofch
ildho
odle
adex
posu
reon
crim
e(R
eyes
,20
07).
Dat
aon
lead
inga
solin
efo
r19
50-1
990
ona
stat
e-by
-sta
tele
velf
rom
Year
lyR
epor
tof
Gas
olin
eSa
lesb
ySt
ate,
Petr
oleu
mM
arke
ting
Ann
ual,
and
Petr
oleu
mPr
oduc
tsSu
rvey
.D
ata
onai
rle
adex
posu
refr
omth
eE
PA’s
Aer
omet
ric
Info
rmat
ion
Ret
riev
alSy
stem
.D
ata
onpe
rca
pita
crim
era
tesf
rom
the
Uni
form
Cri
me
Rep
orts
com
pile
dby
the
Fede
ral
Bur
eau
ofIn
vest
igat
ion
for
1985
-200
2.D
ata
onin
divi
dual
bloo
dle
adle
vels
from
NH
AN
ES
for
1976
-198
0on
ly.
Use
dst
ate-
by-y
ear
vari
atio
nin
the
decl
ine
ofle
adex
posu
refr
omga
solin
ebe
twee
n19
75an
d19
85du
eto
enfo
rcem
ento
fthe
Cle
anA
irA
ctto
iden
tify
the
impa
ctso
fch
ildho
odle
adex
posu
reon
viol
entc
rim
ein
youn
gad
ulth
ood.
Also
estim
ated
the
impa
ctof
lead
expo
sure
onle
vels
ofle
adin
bloo
ddu
ring
child
hood
for
robu
stne
ss.
Prim
ary
mea
sure
ofle
adex
posu
reis
the
gram
sofl
ead
per
gallo
nof
gaso
line
inau
tom
obile
sour
ces.
Cal
cula
ted
air
lead
expo
sure
(mea
sure
ofav
erag
eco
ncen
trat
ion
ofle
adin
the
air
inea
chst
ate
and
year
).C
alcu
late
deff
ectiv
ele
adex
posu
rere
lativ
eto
each
crim
ein
each
stat
ean
dye
aras
the
wei
ghte
dav
erag
eof
lead
expo
sure
b/n
ages
0an
d3
for
allc
ohor
tsof
arre
stee
s.R
egre
ssio
nof
per
capi
tacr
ime
rate
oneff
ectiv
ele
adex
posu
re,
cont
rolli
ngfo
rst
ate
and
year
fixe
deff
ects
and
othe
rst
ate-
year
char
acte
rist
icst
hatc
ould
affec
tcr
ime
rate
s.
Ela
stic
ityof
viol
entc
rim
ew
ithre
spec
tto
lead
expo
sure=
0.8.
Cha
nges
inch
ildho
odle
adex
posu
rein
the
1970
sare
resp
onsib
lefo
ra
56%
decr
ease
invi
olen
tcri
me
inth
e19
90s(
whi
leab
ortio
nis
resp
onsib
lefo
ra
29%
drop
invi
olen
tcri
me)
.W
eak
evid
ence
for
link
b/n
lead
expo
sure
and
mur
der
orpr
oper
tycr
imes
.C
ost
-Ben
efitA
nal
ysis
:C
umul
ativ
eso
cial
cost
ofsw
itchi
ngto
unle
aded
gaso
line:
$15-
65bi
llion
.C
umul
ativ
eso
cial
bene
fito
fred
uced
crim
efr
omre
duce
dle
adex
posu
re:$1
.2tr
illio
n.N
ote,
due
toda
talim
itatio
ns,it
isno
tpos
sible
tom
ake
adi
rect
conn
ectio
nbe
twee
nbl
ood
lead
leve
ls,le
adin
gaso
line,
and
crim
inal
activ
ityat
the
indi
vidu
alle
vel.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1391
Table5(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
long
-ter
meff
ects
ofea
rly
child
hood
lead
expo
sure
:evi
denc
efr
omsh
arp
chan
gesi
nlo
cala
irle
adle
vels
indu
ced
byth
eph
ase-
outo
fle
aded
gaso
line
(Nils
son,
2009
)D
ata
onai
rle
adex
posu
refr
omSw
edish
Env
ironm
enta
lPr
otec
tion
Age
ncy
that
uses
ana
tionw
ide
grid
ofm
osss
ampl
esco
llect
edin
1000
loca
tions
even
lysp
read
acro
ssSw
eden
for
aver
age
lead
expo
sure
leve
lsfo
r19
72-7
4,19
77-7
9,an
d19
82-8
4.In
divi
dual
outc
ome
data
from
the
Edu
catio
nald
atab
ase
atth
eIn
stitu
tefo
rLa
bor
Mar
ketP
olic
yE
valu
atio
nin
Upp
sala
,Sw
eden
onth
ose
born
in19
72-7
4,19
77-7
9,an
d19
82-8
4.n=
797,
889
Iden
tifica
tion
due
toge
ogra
phic
alva
riat
ion
inch
ildho
odle
adex
posu
redu
eto
apo
licy
inSw
eden
that
indu
ced
aph
ase-
out
ofle
aded
gaso
line
betw
een
1973
and
1981
.D
iffer
ence
-in-
differ
ence
com
pari
ngch
ildre
nbo
rnin
mun
icip
aliti
esth
atex
peri
ence
dla
rge
redu
ctio
nsin
lead
expo
sure
toth
ose
born
inm
unic
ipal
ities
with
rela
tivel
yno
chan
ge,be
fore
and
afte
rpo
licy
wen
tint
oeff
ect,
toes
timat
eim
pact
ofch
ildho
odle
adex
posu
reon
late
rlif
eou
tcom
es.A
lsous
edm
othe
rfixe
deff
ectm
etho
ds,co
mpa
ring
siblin
gsw
hoex
peri
ence
ddi
ffer
ent
leve
lsof
lead
expo
sure
inch
ildho
od.C
ontr
olle
dfo
rva
riou
sba
ckgr
ound
char
acte
rist
icsa
swel
las
year
ofbi
rth
and
mun
icip
ality
ofbi
rth
fixe
deff
ects
.
Ade
crea
sein
bloo
dle
adle
vels
by30
mg/
kgle
adst
o:3%
decr
ease
inlik
elih
ood
ofbe
ing
inth
elo
wer
end
ofth
eG
PAdi
stri
butio
n0.
6pp
incr
ease
inav
erag
eIQ
for
mal
es0.
05yr
incr
ease
insc
hool
ing
atta
inm
ent
0.7
ppin
crea
sein
likel
ihoo
dof
grad
uatin
ghi
ghsc
hool
0.5
ppde
crea
sein
likel
ihoo
dof
rece
ivin
gw
elfa
rein
adul
thoo
dE
ffec
tsof
lead
are
stro
nger
atlo
wer
end
ofab
ility
/ski
llsdi
stri
butio
nan
dfo
rch
ildre
nfr
ompo
orer
hous
ehol
ds.E
ffec
tsof
lead
are
stro
nger
whe
nex
posu
reis
atag
es0-
2th
anw
hen
expo
sure
isat
ages
5-7.
No
stat
istic
ally
signi
fica
ntdi
ffer
ence
ineff
ects
byge
nder
.N
onlin
ear
effec
ts—
signi
fica
ntan
dne
gativ
eeff
ects
ofle
adex
posu
reab
ove
75th
perc
entil
e(>
50m
g/kg
).N
ost
atist
ical
lysig
nifica
nteff
ects
ofex
posu
reto
lead
belo
w75
thpe
rcen
tile
leve
ls.E
stim
ated
that
ther
eex
istsa
bloo
dle
adth
resh
old
of60
mg/
Lbe
low
whi
chre
duct
ions
inle
adex
posu
reha
veno
effec
ton
outc
omes
.Si
mila
rre
sults
usin
gm
othe
rfixe
deff
ects
.N
ote:
mea
nsno
trep
orte
d,so
rela
tive
effec
tsiz
esca
n’tb
eca
lcul
ated
.
Unl
esso
ther
wise
note
d,on
lyre
sults
signi
fica
ntat
the
5%le
vela
rere
port
ed.Pe
rcen
tcha
nges
repo
rted
rela
tive
toth
em
ean.
1392 Douglas Almond and Janet Currie
a broader literature asking whether income affects health, and how health affects income?For example, using cross-sectional US data, Case et al. (2002) find a striking relationshipbetween family income and a child’s reported health status, which becomes stronger aschildren age. Their motivation for looking at children is that the child’s health is unlikelyto have a large direct effect on family income, so that the direction of causality is relativelyclear. Currie and Stabile (2003) investigate this relationship using Canadian panel dataand argue that one reason the relationship between income and child health increasesover time is that poorer children are subject to many more negative health shocks. Infact, in Canada, this is the dominant mechanism driving the relationship (which is notsurprising given that all Canadian children have public health insurance so that gaps intreatment rates are small).15
The question we focus on here is how much poor health in childhood, in turn, affectsfuture outcomes. One of the chief ambiguities in answering this question is what wemean by health in childhood. While it has become conventional to measure fetal healthusing birth weight (though there may be better measures, see Almond, Chay and Lee,2005) there are a wide variety of different possible measures of child health, ranging frommaternal reports about the child’s general health status through questions about diagnosesof specific chronic conditions, to the occurrence of “adverse events”.
Case and Paxson (2008a,b, 2010a,b) do not have a direct measure of child health, butargue that adult height is a good proxy for early child health. This is a useful observationgiven that most surveys of adults have no direct information on child health. Heightat age 5 is affected by a range of early health shocks including fetal conditions, poornutrition, and illness. In turn, height at age 5 is strongly predictive of adult height. Andlike birthweight, it is predictive of many shorter and longer range outcomes.
A second problem is that it is often unclear whether the ill health dates from aparticular period (e.g. an injury) or whether it might reflect a continuing, perhaps acongenital, condition. For example, Smith (2009) uses data from the Panel Study ofIncome Dynamics which asked young adults a retrospective question about their healthstatus before age 16. In models with sibling fixed effects, he finds that the sib withthe worse health had significantly lower earnings, although educational attainment wasnot significantly affected. He also finds using data from the Health and RetirementSurvey that reports of general poor health in childhood do tend to be correlated in theexpected way with the presence of specific health conditions. However, it is not possibleto ascertain that the negative effects are due to poor health at any particular “critical”window. Salm and Schunk (2008) attempt to deal with this problem using detailed health
15 Condliffe and Link (2008) argue that in the US, differential access to care also plays a role in the steepening of therelationship between income and child health with age. A number of studies have investigated this relationship, dubbed“the gradient”, in other countries (cf Currie et al. (2007), Case et al. (2008), Doyle et al. (2005), Khanam et al. (2009)).Propper et al. (2007) and Khanam et al. (2009) are particularly interesting because they find that when maternal mentalhealth is controlled, the relationship between family income and child health disappears, suggesting that it is mediatedlargely by factors that affect maternal mental health.
Human Capital Development before Age Five 1393
information from a medical examination of young German children entering school. Inmodels with sibling fixed effects, they find a significant relationship between poor mentalhealth and asthma on the one hand, and measures of cognitive functioning on the other.They control for the child’s birth weight in an effort to distinguish between the effectsof health at birth and health after birth (though to the extent that birth weight is animperfect measure of health at birth, it is possible that the other health measures partlycapture congenital conditions).
Currie et al. (2010) use administrative data from the Canadian province of Manitoba’spublic health insurance system to follow children from birth through young adulthood.
Using information about all contacts with medical providers, they construct measuresof whether children suffered injuries, asthma, mental health problems or other healthproblems at ages 0 to 3, 4 to 8, 9 to 13, and 14 to 18. It is interesting that even ina large sample, there were relatively few children with specific health problems otherthan injuries, asthma, or mental health problems, so that it was necessary to groupthe remaining problems together. They then look at the relationship between healthat various ages, educational attainment, and use of social assistance as a young adult insibling fixed effects models that also control flexibly for birth weight and for the presenceof congenital anomalies. The results are perhaps surprising in view of the conceptualframework developed in Section 3. When entered by themselves, early childhood healthconditions (at age 0-3 and at age 4-8) are predictive of future outcomes, conditionalon health at birth. However, when early physical health conditions are entered alongwith later ones, generally only the later ones matter. This result suggests that physicalhealth in early childhood affects future outcomes largely because it affects future health(i.e., subsequent health mediates the relationship), and not because there is a direct linkbetween early physical health status and cognition. In contrast, mental health conditionsat early ages seem to have significant negative effects on future outcomes even if there areno intermediate report of a mental health condition. This result suggests that commonmental health problems such as Attention Deficit Hyperactivity Disorder (ADHD, alsocalled Attention Deficit Disorder or ADD) or Conduct Disorders (i.e. disorders usuallyinvolving abnormal aggression and anti-social behavior) may impair the process of humancapital accumulation even if they do not lead to diagnoses of mental health disorders inadulthood.
Several recent papers focus specifically on measures of mental health conditions.Currie and Stabile (2006) use questions similar to those on mental health “screeners”which were administered to large samples of children in the US and Canada in twonational surveys. They find that children whose scores indicated mental health problemsin 1994 had worse outcomes as of 2002-4 than siblings without such problems. Theycontrolled for birth weight (among other variables) and estimated models with andwithout including children with diagnosed learning disabilities. In all specifications, theyfound negative effects of high ADHD scores on test scores on schooling attainment.
1394 Douglas Almond and Janet Currie
Smith and Smith (2008) report similar results using data from the PSID which includesretrospective questions about mental health problems before age 16. Like Smith (2009)and Currie et al. (2010) they estimate models with sibling fixed effects, and findsignificant long term effects of mental health conditions which are much larger thanthose of physical health conditions. Vujic et al. (2008) focus on conduct disorders usinga panel of Australian twins and find that conduct disorder before age 18 has strongnegative effects on the probability of high school graduation as well as positive effectson the probability of criminal activity. None of these three papers focus specifically onmeasures of mental health conditions before age 5 but “externalizing” mental healthconditions such as ADHD and Conduct Disorder typically manifest themselves at earlyages. Finally, although they are conceptually distinct, many survey measures of mentalhealth resemble measures of “non-cognitive skills”. Hence, one might interpret evidencethat non-cognitive skills in childhood are important determinants of future outcomes asfurther evidence of the importance of early mental health conditions (Blanden et al.,2006; Heckman and Rubinstein, 2001).
4.2.3. Home environmentThe home is one of the most important environments affecting a young child and thereis a vast literature in related disciplines investigating the relationship between differentaspects of the home environment and child outcomes. We do not attempt to summarizethis literature here, but pick three aspects that may be most salient: Maternal mental healthand/or substance abuse, maternal employment, and child abuse/foster care (which maybe considered to be an extreme result of bad parenting). Given the importance of childmental health and non-cognitive skills, it is interesting to ask how maternal mental healthaffects child outcomes? Frank and Meara (2009) examine this question using data fromthe National Longitudinal Survey of Youth. They include a rich set of control variables(mother’s cognitive test score, grandparent’s substance abuse, permanent income) andestimate models with mother fixed effects and models with propensity scores. Theirestimates suggest large effects (relative to the effects of income) of contemporaneousmaternal depression on the quality of the home environment and on children’s behavioralproblems, but little effect on math and reading scores. Estimates of the effects of maternalsubstance abuse are mixed, which echoes the findings of Chatterji and Markowitz (2000)using the same data. Unfortunately, the authors are not able to look at the long termeffects of maternal depression experienced by children aged 0 to 5 because the depressionquestions in the NLSY have been added only recently. As these panel data are extendedin time, further investigation of this issue is warranted.
There is also a large literature, including some papers by economists, examining theeffect of maternal employment at early ages on child outcomes. Much of this literaturesuffers from the lack of an appropriate conceptual framework. If we think of childoutcomes being produced via some combination of inputs, then the important questionis how maternal employment affects the inputs chosen? This will evidently depend on
Human Capital Development before Age Five 1395
how much her employment income relaxes the household budget constraint, and theprice and quality of the child care alternatives and other inputs that are available. Some ofthe literature on maternal employment seems to implicitly assume that the mother’s timeis such an important and unique input that no purchased input can adequately replace it.This may possibly be the case but is a strong assumption. If the mother’s time is replaceableat some price, then one might expect maternal employment to have quite different effectson women with different levels of household income (moreover, mother’s time may notall be of equal quality, so that it is easier to replace some mother’s time than others withthe market). This argument suggests that it is extremely important to consider explicitlythe quality of the mother’s time inputs and the availability of potential substitute inputsin models of maternal employment, something that is difficult to do in most availabledata sets. Studies that rely on regression methods and propensity score matching (seeBerger et al. (2005) and Ruhm (2004) often find small negative effects of maternalemployment (especially in the first year) on children’s cognitive development. However,two recent studies using variation in maternity leave provisions find that while moregenerous maternity leave policies are associated with increased maternal employment,there is little effect on children’s outcomes (Baker and Milligan, 2010; Dustmann andSchonberg, 2009). Dustmann and Schonberg (2009) have data that permit cohortsaffected by expansions in German maternity leave laws to be followed for many years.They see no effect of maternal employment on educational attainment or wages.
Finally, there are a few papers examining the effects of child abuse/foster care on childoutcomes. This is a difficult area to investigate because it is hard to imagine that abuse (orneglect) can be divorced from other characteristics of the household. Currie and Widom(2009) use data from a prospective longitudinal study in which abused children (thetreatments) were matched to controls. After following these children until their mid 40s,they found that the abused children were less likely to be employed, had lower earnings,and fewer assets, and that these patterns were particularly pronounced among women. Itis possible that these results are driven by unobserved differences between the treatmentsand controls, although focusing on various subsets of the data (e.g. children whosemothers were on welfare; children of single mothers) produced similar results. Currieand Tekin (2006) use data from the National Longitudinal Study of Adolescent Healthto examine the effect of having been abused before age 7 on the propensity to commitcrime. They find strong effects which are quite similar in OLS, sibling, and twin fixedeffects models. It is possible that these results reflect a characteristic of an individual child(such as difficult temperament) which makes it both more likely that they will be abusedand more likely that they will commit crime. However, controlling directly for measuresof temperament and genetic endowments does not alter the results. The Doyle (2008)study of the effects of foster care on the marginal child is also summarized in Table 5.
1396 Douglas Almond and Janet Currie
4.2.4. Toxic exposuresEpidemiological studies of postnatal pollution exposure and infant mortality have yieldedmixed results and many are likely to suffer from omitted variables bias. Currie and Neidell(2005) examine the effect of more recent (lower) levels of pollution on infant health,along with the role of specific pollutants in addition to particulates (only TSPs weremeasured during the time periods analyzed by Chay and Greenstone (2003a,b)). Usingwithin-zip code variation in pollution levels, they find that a one unit reduction in carbonmonoxide over the 1990s in California saved 18 infant lives per 100,000 live births.However, unlike Currie et al. (2009) they were unable to find any consistent evidence ofpollution effects on health at birth, probably because of the crudeness of their measure ofmaternal location.
Reyes (2007) found large effects of banning leaded gasoline on crime in the US, butresults were not robust to state-specific time trends despite a relatively long panel of state-level lead measurements. Nilsson (2009) considered reductions in ambient lead levelsin Sweden following the banning of lead in gasoline and measures possible exposuresusing the concentrations of lead in 1000 moss (bryophyte) collection sites that havebeen maintained by the Swedish environmental protection agency since the early 1970s.Nilsson (2009) found that early childhood exposure reduced human capital, as reflectedby both grades and graduation rates. These effects persisted when comparisons wererestricted within siblings, and were substantially larger for low-income families.
4.2.5. Summary re: long term effects of fetal and early childhood environmentThe last 10 years have seen an upsurge of empirical work on the long-term effects ofearly childhood. As a result, much has been learned. We can state fairly definitivelythat at least some things that happen before age five have long-term consequences forhealth and human capital. Moreover, these effects are sufficiently large and general toshape outcomes at the population level. On balance, effects of fetal exposure tend to besomewhat larger than postnatal effects, but there are important exceptions. Mental healthis a prime example. Mental health conditions and non-cognitive skills seem to have large,persistent effects independent of those captured by measures of child health at birth.
5. EMPIRICAL LITERATURE: POLICY RESPONSES
The evidence discussed above indicates that prenatal and early childhood often have acritical influence on later life outcomes. However, by itself this evidence says little aboutthe effectiveness of remediation. Hence, this section discusses evidence about whetherremediation in the zero to five period can be effective in shaping future outcomes. Inso doing, we take a step away from explicit consideration of an early-childhood shockug as in Section 2. Instead, we focus on the specific public policies that may be able toalter developmental trajectories, often in disadvantaged sub-populations. We begin withprograms that raise income, and then move on to programs that target specific domains.
Human Capital Development before Age Five 1397
5.1. Income enhancementIn the model sketched above, there are many ways for poverty to affect child outcomes.Even with identical preferences, poorer parents will make different investment choicesthan richer ones. In particular, poor families will optimize at lower investment (andconsumption) levels and thereby have children with lower health and human capital,other things equal. Further, poor parents may face different input prices for certain goods,or have access to different production technologies. Providing cash transfers addresses thebudgetary problems without necessarily changing the production technology. Hence, it isof interest to see whether cash transfers, in and of themselves, can improve outcomes. It ishowever, remarkably difficult to find examples of policies that increase incomes withoutpotentially having a direct effect on outcomes. For example, many studies of cash welfareprograms have demonstrated that children who are or have been on welfare remain worseoff on average than other children. This does not necessarily mean, however, that welfarehas failed them. Without welfare, their situation might have been even worse. Bergeret al. (2009) explore the relationship between family income, home environments, childmental health outcomes, and cognitive test scores using data from the Fragile Familiesand Child Well-being Study which follows a cohort of five thousand children born inseveral large US cities between 1998 and 2000. They show that all of the measures of thehome environment they examine (which include measures of parenting skills as well asphysical aspects of the home) are highly related to income and that controlling for thesemeasures reduces the effects of income on outcomes considerably.
Levine and Zimmerman (2000) showed that children who spent time on welfarescored lower than other children on a range of tests, but that this difference disappearedwhen the test scores of their mothers were controlled for, suggesting that welfare hadlittle effect either positive or negative. Similarly, Levine and Zimmerman (2000) arguethat children of welfare mothers were more likely to grow up to be welfare mothers,mainly because of other characteristics of the household they grew up in.
Currie and Cole (1993) compare siblings in families where the mother receivedwelfare while one child was in utero, but not while the other child was in utero, and find nodifference in the birth weight of the siblings. Given that research has shown little evidenceof positive effects of cash welfare on children, it is not surprising that the literatureevaluating welfare reform in the United States has produced similarly null findings. TheNational Research Council (Smolensky and Gootman, 2003) concluded that “no strongtrends have emerged, either negative or positive, in indicators of parent well-being orchild development across the years just preceding and following the implementationof [welfare reform]”. However, US welfare reform was a complex intervention thatchanged many parameters of daily life by, for example, imposing work requirements onrecipients.
Conditional tax credits represent an alternative approach to providing income to poorfamilies, and hence to poor children. The early years of the Clinton administration in the
1398 Douglas Almond and Janet Currie
United States saw a huge expansion of the Earned Income Tax Credit (EITC), whilein the UK, the Working Families Tax Credit approximately doubled in 1999. Theseare tax credits available to poor working families. Their essential feature is that theyare “refundable”—in other words, a family whose credit exceeds its taxes receives thedifference in cash. The tax credits are like welfare in that they give cash payments topoor families. But like welfare reform, the tax credits are a complex intervention inthat recipients need to work and file tax returns in order to be eligible, and a great dealof research has shown that such tax credits affect maternal labor supply and marriagepatterns (Eissa and Liebman, 1996; Meyer and Rosenbaum, 2001; Blundell, 2006). Thisis because the size of the payment increases with earnings up to a maximum level beforebeing phased out, so that it creates an incentive to work among the poorest householdsbut a work disincentive for households in the phase-out range. In the US, the numberof recipients grew from 12.5 million families in 1990 to 19.8 million in 2003, and themaximum credit grew from $953 to $4204. The rapid expansion of this formerly obscureprogram run through the tax system has resulted in cash transfers to low-income familiesthat were much larger than those that were available under welfare. Gundersen and Ziliak(2004) estimate that the EITC accounted for half of the reduction in after-tax poverty thatoccurred over the 1990s (the other half being mainly accounted for by strong economicgrowth).
Table 6 provides an overview of some of the research on the effects of income onchildren. Dahl and Lochner (2005) use variation in the amount of the EITC householdsare eligible for over time and household type to identify the effects of household incomeand find that each $1000 of income improves childrens’ test scores by 2% to 4% of astandard deviation. An attractive feature of the changes in the EITC is that householdsmay well have regarded them as permanent, so this experiment may approximate theeffects of changes in permanent rather than transitory income. Their result implies,though, that it would take on the order of a $10,000 transfer to having an educationallymeaningful effect on test scores.
Milligan and Stabile (2008) take advantage of a natural experiment resulting fromchanges in Canadian child benefits. These benefits vary across provinces and werereformed at different times. An advantage of their research is that the changes in incomewere not tied to other changes in family behavior, in contrast to programs like the EITC.
They find that an extra $1000 of child benefits leads to an increase of about 0.07 ofa standard deviation in the math scores and in the Peabody Picture Vocabulary Test, astandardized test of language ability for four to six year old children. If we think of achange of a third or a half a standard deviation in test scores as a meaningful educationaleffect, then these results suggest that an increase of as little as $5000 in family incomehas a meaningful effect. Milligan and Stabile (2008) go beyond Dahl and Lochnerby examining effects on other indicators. They find that higher child benefits loweraggression in children and decrease depression scores for mothers. They do not find much
Human Capital Development before Age Five 1399
Table6
Effectsof
incomeon
birthan
dearly
child
hood
outcom
es:evide
ncefrom
theUSan
darou
ndtheworld.
Stud
yan
dda
taStud
yde
sign
Results
The
link
betw
een
AFD
Cpa
rtic
ipat
ion
and
birt
hwei
ght(
Cur
rie
and
Col
e,19
93).
Dat
afr
omN
LSY
mer
ged
with
stat
ean
dco
unty
leve
lda
ta.D
ata
onch
ildre
nbo
rnbe
twee
n19
79-1
988(n=
5000).
IV:us
ing
para
met
erso
fAFD
C,FS
Pan
dM
edic
aid
asin
stru
men
ts,co
ntro
lling
for
child
char
acte
rist
icsa
ndre
gion
fixe
deff
ects
.Se
para
tere
gres
sions
for
poor
blac
kan
dpo
orw
hite
wom
en.Si
blin
gFE
:co
mpa
ring
sibsw
here
mot
her
onA
FDC
duri
ngpr
egna
ncy
toot
hers
.
OLS
resu
ltssu
gges
ttha
tAFD
Cha
sneg
ativ
eeff
ects
.Si
blin
gfixe
deff
ects
indi
cate
nosig
nifica
nteff
ects
.IV
resu
ltssu
gges
ttha
tAFD
Cdu
ring
preg
nanc
yca
uses
larg
ein
crea
sesi
nm
ean
birt
hw
eigh
tam
ong
poor
whi
teso
nly.
Hen
ce,
evid
ence
sugg
ests
that
AFD
Cha
sneu
tral
orpo
sitiv
era
ther
than
nega
tive
effec
ts.N
egat
ive
estim
ates
driv
enby
sele
ctio
n.
Doe
smon
eyre
ally
mat
ter?
estim
atin
gim
pact
soff
amily
inco
me
onch
ildre
n’sa
chie
vem
entw
ithda
tafr
omra
ndom
-ass
ignm
ente
xper
imen
ts(M
orri
seta
l.,20
04).
Dat
afr
omfo
urst
udie
stha
teva
luat
ed8
wel
fare
and
anti-
pove
rty
prog
ram
swith
rand
omiz
edde
signs
:C
onne
ctic
ut’s
Jobs
Firs
t,th
eLo
sAng
eles
Jobs
Firs
tG
AIN
,th
eN
ewB
runs
wic
kan
dB
ritis
hC
olum
bia
sites
ofth
eC
anad
ian
Self-
Suffi
cien
cyPr
ojec
t,an
dth
eA
tlant
a,G
A,G
rand
Rap
ids,
MI,
and
Riv
ersid
e,C
Asit
esof
the
Nat
iona
lEva
luat
ion
ofW
elfa
reto
Wor
kSt
rate
gies
.n=
18,47
1ch
ildre
nag
ed2-
15at
the
time
ofra
ndom
assig
nmen
t.
IVm
odel
suse
rand
omas
signm
enta
sin
stru
men
tsfo
rin
com
e,w
elfa
rere
ceip
t,an
dem
ploy
men
tto
estim
ate
impa
ctof
inco
me
onch
ildre
n’sc
ogni
tive
achi
evem
ent.
Incl
uded
dum
mie
sfor
sites
inbo
thst
ages
ofth
ere
gres
sion
anal
ysis.
Cog
nitiv
eac
hiev
emen
tmea
sure
dw
ithte
stsc
ores
and
pare
nt/t
each
erre
port
s.C
ontr
olle
dfo
rva
riou
sbas
elin
eba
ckgr
ound
fam
ilych
arac
teri
stic
s.
A$1
000
incr
ease
inan
nual
inco
me
raise
sco
gniti
veac
hiev
emen
tby
0.06
ofSD
for
child
ren
aged
2-5.
No
stat
istic
ally
signi
fica
nteff
ects
ofin
com
eon
child
ren
aged
6-9
or10
-15.
A3-
SDin
crea
sein
the
prop
ortio
nof
quar
ters
that
wel
fare
isre
ceiv
edle
adst
oa
1.5
SDde
crea
sein
cogn
itive
achi
evem
enta
mon
g10
-15
year
-old
s.
(con
tinue
don
next
page
)
1400 Douglas Almond and Janet Currie
Table6(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Who
bene
fits
from
child
bene
fit?
(Blo
wet
al.,
2005
).D
ata
from
UK
Fam
ilyE
xpen
ditu
reSu
rvey
1980
-200
0.n=
9811
two-
pare
ntho
useh
olds
;n=
2920
one-
pare
ntho
useh
olds
.
Iden
tifica
tion
from
vari
atio
nin
real
valu
eof
Chi
ldB
enefi
t(C
B)1
980-
2000
due
toin
flat
ion
and
polic
ych
ange
s.C
alcu
late
dan
ticip
ated
and
unan
ticip
ated
chan
gesi
nC
Bpa
ymen
ts.E
xam
ine
the
prop
ensit
yto
spen
dC
Bon
child
good
s(ch
ildre
n’s
clot
hing
)and
adul
tgoo
ds(a
lcoh
ol,
toba
cco
and
adul
tclo
thin
g).
Spen
din
goutofunan
tici
pat
edvs
.an
tici
pat
edch
ange
inC
B:
15x
grea
ter
for
alco
holi
n2-
pare
ntho
useh
olds
;10
xgr
eate
rfo
rad
ultc
loth
ing
inlo
ne-p
aren
tho
useh
olds
.R
esul
tssu
gges
ttha
tpar
ents
fully
insu
rech
ildre
nag
ains
tinc
ome
shoc
ksso
that
unan
ticip
ated
chan
geso
nly
affec
tspe
ndin
gon
adul
tgoo
ds.
Effec
tsof
EIT
Con
birt
han
din
fant
heal
thou
tcom
es.(B
aker
,20
08).
US
Vita
lSta
tistic
sNat
ality
data
file
1989
-96.
(n=
781,
535)
.E
xclu
deob
serv
atio
nsfr
om19
94.U
sed
data
from
1997
Mar
chC
PSto
com
pute
prop
ortio
nof
trea
tmen
tgro
upth
atis
elig
ible
for
EIT
C.
Diff
eren
ce-i
n-D
iffer
ence
-in-
Diff
eren
ces
(DD
D).
Exp
loit
the
larg
eex
pans
ion
ofE
ITC
in19
93(p
hase
din
1994
-96)
that
incr
ease
dbe
nefits
tofa
mili
esw
ith2
orm
ore
child
ren.
“Tre
atm
ent”
grou
p=
mot
hers
givi
ngbi
rth
to3r
dor
high
erch
ild.C
ontr
ol1
=m
othe
rsgi
ving
birt
hto
1sto
r2n
dch
ild.C
ontr
ol2
=m
othe
rsgi
ving
birt
hto
1stc
hild
.U
sed
mot
hers
with
less
than
Hig
hsc
hool
asa
prox
yfo
rE
ITC
elig
ibili
ty.E
ffec
tof
inte
rest
isin
tera
ctio
nbe
twee
n<
HS*
trea
tmen
t*af
ter
EIT
Cex
pans
ion.
Bir
thw
eigh
t:In
crea
sed
by0.
4%fo
ral
lwom
enIn
cide
nce
oflo
wbi
rth
wei
ght:
Dec
reas
edby
3.7%
for
all
No
stat
istic
ally
signi
fica
nteff
ects
on#
pren
atal
visit
sor
mat
erna
lsm
okin
gdu
ring
preg
nanc
y.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1401
Table6(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
impa
ctof
fam
ilyin
com
eon
child
achi
evem
ent:
evid
ence
from
the
earn
edin
com
eta
xcr
edit
(Dah
land
Loch
ner,
2008
).D
ata
from
NLS
Yon
child
ren
and
thei
rm
othe
rsfo
r19
88-2
000.
n=
4720
child
ren
(252
7m
othe
rs).
Iden
tifica
tion
base
don
EIT
Cex
pans
ions
in19
87-1
993
and
1993
-199
7th
atin
crea
sed
bene
fits
for
low
-an
dm
iddl
e-in
com
e.U
sed
simul
ated
inst
rum
enta
lva
riab
les(
IV):
IVis
pred
icte
dch
ange
inE
ITC
inco
me.
Con
trol
led
for
fam
ilyba
ckgr
ound
and
dem
ogra
phic
vari
able
san
dfo
rtim
e-va
ryin
gst
ate-
spec
ific
polic
ies
that
mig
htaff
ectc
hild
outc
omes
.E
xam
ine
cont
empo
rane
ousa
ndla
gged
inco
me.
Also
estim
ate
OLS
and
child
fixe
deff
ects
mod
els.
A$1
000
incr
ease
inin
com
era
isesc
ombi
ned
mat
han
dre
adin
gsc
ores
by0.
06SD
.La
rger
gain
sfro
mco
ntem
pora
neou
sinc
ome
for
child
ren
aged
5to
10th
anfo
rth
ose
aged
11-1
5.La
rger
gain
sfor
child
ren
from
disa
dvan
tage
dfa
mili
es.
Do
child
tax
bene
fits
affec
tthe
wel
l-be
ing
ofch
ildre
n?E
vide
nce
from
Can
adia
nch
ildbe
nefit
expa
nsio
ns(M
illig
anan
dSt
abile
,20
08).
Dat
afr
omsu
rvey
ofla
bor
and
inco
me
dyna
mic
son
fam
ilies
with
child
ren
for
1999
-200
4fo
rbe
nefit
simul
atio
n.D
ata
from
Can
adia
nN
atio
nalL
ongi
tudi
nalS
tudy
ofC
hild
ren
&Yo
uth
for
1994
-200
5fo
rch
ildre
n10
and
unde
r(n=
56,0
00).
Inst
rum
enta
lvar
iabl
es.U
sed
vari
atio
nin
child
bene
fits
acro
sstim
e,pr
ovin
ces,
and
num
ber
ofch
ildre
npe
rfa
mily
tode
velo
pa
mea
sure
ofbe
nefiti
ncom
eas
IVfo
rch
ildbe
nefits
inre
gres
sions
ofth
eeff
ecto
fchi
ldbe
nefits
onch
ildou
tcom
es.C
ontr
olle
dfo
rpr
ovin
ce-y
ear
fixe
deff
ects
.A
lsore
duce
dfo
rmsf
orsim
ulat
edbe
nefits
’effec
ton
child
outc
omes
.
An
incr
ease
in$
1000
insim
ulat
edbe
nefits
lead
sto
:1.
5%re
duct
ion
inlik
elih
ood
ofle
arni
ngdi
sabi
lity
(ifm
om<
HS)
;3.
6%de
clin
eco
nduc
tdiso
rder
/agg
ress
ion
scor
e4-
10;4.
3%of
SDde
clin
em
ater
nald
epre
ssio
n;11
.6%
SDde
clin
ein
mat
erna
ldep
ress
ion
(ifm
om<
HS)
;1.
1%de
clin
ein
child
ever
expe
rien
cing
hung
er(if
mom
<H
S)La
rger
posit
ive
impa
ctso
ned
ucat
ion
and
phys
ical
heal
thfo
rbo
ysth
angi
rls;
larg
erpo
sitiv
eim
pact
son
men
talh
ealth
for
girl
s.E
ffec
tson
mat
han
dvo
cabu
lary
scor
es,be
havi
oral
outc
omes
,m
ater
nald
epre
ssio
nan
dlik
elih
ood
ofev
erex
peri
enci
nghu
nger
pers
istat
leas
t4ye
ars.
(con
tinue
don
next
page
)
1402 Douglas Almond and Janet Currie
Table6(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Sout
hA
fric
anC
hild
Supp
ortG
rant
(CSG
):U
ncon
ditio
nalc
ash
tran
sfer
prog
ram
with
paym
ents
mad
eto
child
’spr
imar
yca
regi
ver.
Inte
nded
toco
ver
the
coun
try’
spoo
rest
30%
ofch
ildre
n.St
udy
ofeff
ects
onch
ildnu
triti
on(A
guer
oet
al.,
2009
).D
ata
from
Kw
aZul
u-N
atal
Inco
me
Dyn
amic
sStu
dy.M
ain
outc
ome
mea
sure
:H
eigh
t-fo
r-A
gez-
scor
e(H
AZ
).T=
245,
C=
886.
Exp
loit
vary
ing
leng
thso
fexp
osur
eto
CSG
amon
gch
ildre
nag
ed0-
3du
eto
the
timin
gof
the
rollo
utof
CSG
.Fi
taqu
adra
ticO
LSm
odel
toap
plic
atio
nde
lay
(num
ber
ofda
ysbe
twee
npr
ogra
mcr
eatio
nan
dch
ild’s
enro
llmen
t)us
ing
data
for
child
ren
born≤
2ye
arsp
rior
tosu
rvey
date
.C
alcu
late
dan
expe
cted
appl
icat
ion
dela
yva
riab
lefo
rea
chch
ildco
nditi
onal
onbi
rth
date
,lo
catio
nan
dfa
mily
char
acte
rist
ics,
define
das
%de
viat
ion
inap
plic
atio
nde
lay
from
expe
cted
dela
y.C
ondi
tiona
lon
fam
ilych
arac
teri
stic
s,de
viat
ion
inde
lay,
and
obse
rvab
les,
the
exte
ntof
CSG
trea
tmen
tsho
uld
bera
ndom
.Use
gene
raliz
edpr
open
sity
scor
es,
MLE
.
No
gain
sin
HA
Zfo
rtr
eatm
ents
cove
ring
≤20
%of
the
0-3
win
dow
.H
AZ
0.20
high
erfo
rtr
eatm
entc
over
ing
2/3
ofw
indo
wth
anfo
ra
child
rece
ivin
gtr
eatm
entc
over
ing
1%of
win
dow
.B
enefi
t-co
stra
tio
ofC
SG
:B
etw
een
1.06
and
1.48
(est
imat
ing
lifet
ime
earn
ings
gain
sfro
mga
insi
nH
AZ
,us
ing
annu
al5%
disc
ount
rate
,an
das
sum
ing
unem
ploy
ed33
%of
time)
.R
esul
tsro
bust
toch
ecki
ngfo
rag
eco
hort
effec
tsan
dlo
catio
n/sp
acia
leffec
ts.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1403
Table6(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Cond
itiona
lCashTran
sfer
Prog
rams(CCT
)
Mex
ican
“Pro
gres
a”pr
ogra
m(n
ow“O
port
unid
ades
”):C
ondi
tiona
lcas
htr
ansf
erso
f20-
30%
ofho
useh
old
inco
me
ever
y2
mon
ths.
Fam
ilies
mus
ttak
eyo
ung
child
ren
tohe
alth
clin
ics,
imm
uniz
atio
ns,ge
tade
quat
epr
enat
alca
re,&
rece
ive
nutr
ition
supp
lem
ent.
(Fam
ilies
also
toke
epol
der
child
ren
insc
hool
)(G
ertle
r,20
04).
Dat
afr
omsu
rvey
colle
cted
from
expe
rim
enta
lvill
ages
.Sa
mpl
esiz
es:77
03ch
ildre
nun
der
age
3at
base
line;
1501
new
born
s;H
eigh
tan
alys
is:T=
1049
,C=
503;
Ane
mia
anal
ysis:
T=
1404
,C=
608.
Prog
ram
phas
edin
to32
0tr
eatm
ent
villa
gesa
nd18
5co
ntro
lvill
ages
rand
omly
sele
cted
.C
ontr
olvi
llage
srec
eive
dbe
nefits
2ye
arsa
fter
star
tofp
rogr
am.T
hey
did
not
know
this
wou
ldbe
the
case
.In
vest
igat
e0/
1tr
eatm
entd
umm
yas
wel
lasd
iffer
ent
leng
thso
fpro
gram
expo
sure
.
Tre
atm
enteff
ects
:N
ewbo
rns2
5.3%
less
likel
yto
beill
inpa
stm
onth
.C
hild
ren
0-3
22.3
%le
sslik
ely
tobe
illin
past
mon
th.
Hei
ght(
12-3
6m
os.)
1.2%
high
er.
Prob
abili
tyof
bein
gan
emic
(12-
48m
os.)
25.5
%le
ss.
(con
tinue
don
next
page
)
1404 Douglas Almond and Janet Currie
Table6(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
“Ate
ncio
na
Cri
sis”
inru
ral
Nic
arag
ua.C
ondi
tiona
lcas
htr
ansf
er∼
15%
aver
age
per
capi
taex
pend
iture
s.W
omen
rece
ive
paym
ents
ever
y2
mon
thsi
fpre
scho
olch
ildre
nta
ken
for
regu
lar
visit
sto
heal
thcl
inic
s.(O
lder
child
ren
mus
tal
sobe
enro
lled
insc
hool
)(M
acou
rset
al.,
2008
).Sa
mpl
esiz
es:T=
3002
,C=
1019
.A
dditi
onal
data
from
the
2001
Nic
arag
uaLi
ving
Stan
dard
sM
easu
rem
entS
tudy
.
Ran
dom
ized
expe
rim
enta
ldes
ign.
4gr
oups
:C
CT,
CC
T+
voca
tiona
ltra
inin
g,C
CT+
prod
uctiv
ein
vest
men
tgra
nt,
cont
rol.
Subg
roup
anal
ysis
byge
nder
and
age.
Ana
lyze
dva
riou
stra
nsm
issio
nm
echa
nism
sby
estim
atin
gex
pend
iture
son
differ
entf
oods
(to
mea
sure
nutr
ient
inta
ke),
and
differ
ence
sin
indi
cato
rsfo
rea
rly
child
hood
deve
lopm
ent.
Tre
atm
enteff
ects
:D
evel
opm
enta
lscr
eene
r(D
DST
soci
al-p
erso
nal):
incr
ease
by0.
13SD
s.D
DST
lang
uage
:in
crea
seby
0.17
SDs.
Rec
eptiv
evo
cabu
lary
(TV
IP):
incr
ease
by0.
22SD
s.(C
hild
ren
mad
eup
appr
ox.1.
5m
onth
sof
dela
yon
TV
IP.)
Effec
tsby
age:
DD
STla
ngua
ge:up
0.06
SDs
0-35
mos
.,up
0.20
SDs6
0-83
mos
.T
VIP
:up
0.05
SDs3
6-59
mos
.,up
0.36
SDs
60-8
3m
os.(O
ldes
tchi
ldre
nm
ade
up2.
4m
onth
sofd
elay
onT
VIP
)N
ost
atist
ical
lysig
nifica
ntdi
ffer
ence
sby
gend
er.
Tra
nsm
issio
nm
echa
nism
s:B
ette
rdi
et;m
ore
likel
yto
have
book
s,pa
per,
penc
ils;m
ore
likel
yto
bere
adto
;m
ore
likel
yto
have
chec
kup,
vita
min
s,de
-wor
min
gdr
ugs;
mor
elik
ely
tobe
enro
lled;
mor
elik
ely
tose
edo
ctor
whe
nill
.
Unl
esso
ther
wise
note
d,on
lyre
sults
signi
fica
ntat
5%le
vela
rere
port
ed.
Perc
entc
hang
es(d
enot
edby
%in
stea
dof
“pp”
)are
repo
rted
rela
tive
toth
em
ean,
ifm
eans
wer
ere
port
edin
the
pape
r.“T
”=
Tre
atm
entg
roup
,“C
”=
Con
trol
grou
p.O
utco
mes
wri
tten
as“T>
C”
mea
nth
attr
eatm
ento
utco
me
grea
ter
than
cont
rolo
utco
me
at5%
signi
fica
nce
leve
l.O
utco
mes
wri
tten
as“T=
C”
mea
nth
atth
ere
isno
stat
istic
ally
signi
fica
ntdi
ffer
ence
inou
tcom
esbe
twee
n2
grou
psat
5%le
vel.
Human Capital Development before Age Five 1405
impact on physical health measures, though they do find a decrease in families reportingthat their children went hungry. There is some evidence of gender differences, with girlsshowing greater responsiveness to income on the mental health and behavioral scoreswhile boys show greater responsiveness on test scores.
These findings are extremely intriguing, but raise several questions. First, do theeffects of income vary depending on the child’s age? Morris et al. (2004) argue thatincome is more important at younger ages, though persistent poverty is worst of all.Second, are there really gender effects in the impact of income, and if so why? Third,the effects that Milligan and Stabile find are roughly twice those found by Dahl andLochner. Is this because the former study a pure income transfer while the latter studya tied transfer? Fourth, will the effects last, or will they be subject to “fade out” as thechildren grow older?
Table 6 also includes examples from a growing literature analyzing “conditional cashtransfer programs” (CCTs). These are programs that tie transfers to specific behavior onthe part of the family. For example, the parents may be required to make sure that thechildren attend school or get medical care in return for the transfer. These programs havebecome increasingly popular in developing countries, and have also been implementedto a limited extent in rich countries (for example, there is a program in New York Citywhich is being evaluated by Manpower Development Research Corporation). By theirnature, CCTs are complex programs that cannot tell us about the pure impacts of income.Still, these programs have attracted attention because randomized controlled trials haveshown at least short-term results. It is difficult however to compare across programs, giventhat they all tend to focus on different outcomes.
Given this positive evidence about the effects of income, it is a puzzle why so much aidto poor families is transferred in kind. Currie and Gahvari (2008) survey the many reasonsfor this phenomena that have been offered in the economics literature and concludethat the most likely reasons aid is offered in kind are agency problems, paternalism,and politics. In a nutshell, policy makers and the voters they represent may be moreconcerned with ensuring that children have medical care than with maximizing theirparent’s utility, even if the parent’s utility is assumed to be affected by the children’s accessto health care. Politics come in because coherent lobby groups (such as doctors, teachers,or farmers) may have incentives to advocate for various types of in kind programs. Inany case, in kind programs are an important feature of aid policies in all Organizationfor Economic Cooperation and Development states, accounting for over 10% of GDP ifhealth care and educational programs are included. In what follows, we first discuss “nearcash programs” and then programs whose benefits are less fungible with cash.
5.2. Near-cash programsPrograms such as the US Food Stamp Program (FSP, now renamed the SupplementalNutrition Assistance Program, or SNAP) and housing assistance are often referred to
1406 Douglas Almond and Janet Currie
as “near cash” programs because they typically offer households benefits that are worthless than what the household would have spent on food or housing in any case. Hence,
canonical microeconomic theory suggests that households should think of them asequivalent to cash and that they should have the same impact as the equivalent cashtransfer would have. In the case of food stamps, it has proven difficult to test thisprediction because the program parameters are set largely at the national level, so thatthere is only time series variation.
Currie (2003) provides an overview of the program, and the research on its effectsthat had been conducted up to that point. Schanzenbach (forthcoming) uses data from afood stamp cash out experiment to examine the effect on food spending. She finds that aminority of households actually received more in food stamps than they would otherwisespend on food. In these constrained households, families did spend more on food thanthey would have otherwise, while in other households, food stamps had the same effectas cash. Unfortunately, there is little evidence that constrained households bought foodsthat were likely to have beneficial effects; they seem, for example, to have spent some ofthe “extra” food money on products such as soda.
Hoynes and Schanzenbach (2009) use variation from the introduction of the FSP toidentify its effects on food spending. The FSP began as a small pilot program in 1961, andgradually expanded over the next 13 years: In 1971, national eligibility standards wereestablished, and all states were required to inform eligible households about the program.
In 1974, states were required to extend the program statewide if any areas of the stateparticipated. Using data from the PSID, the introduction of the FSP was associated withan 18% increase in food expenditures in the full sample, with somewhat larger effects inthe most disadvantaged households. They find that the marginal propensity to consume(MPC) food out of food stamp income was 0.16 compared to 0.09 for cash income. Thus,it does seem that many households were constrained to spend more on food than theyotherwise would have (or alternatively, that the person receiving the food stamps had astronger preference for food than the person controlling cash income in the household).From a policy maker’s point of view, this means that the FSP has a bigger impact on foodspending than an equivalent cash transfer. Still, it is a leaky bucket if only 16 cents ofevery dollar transferred goes to food.
Bingley and Walker (2007) conduct an investigation of the Welfare Milk Program inthe UK. They identify the effect of the program on household milk expenditures usinga large change in eligibility for the program that had differential effects by householdtype. They find that about 80% of a transfer of free milk is crowded out by reductionsin milk purchases by the household. This estimate is quite similar to that of Hoynes andSchanzenbach, though it still suggests that the in kind transfer is having some effect onthe composition of spending. Details of these two studies are shown in Table 7.
Given that these programs appear to have some effect on food expenditures, it isreasonable to ask what effect they have on child outcomes. There is a substantial older
Human Capital Development before Age Five 1407
Table7
Impa
ctsof
Food
Stam
pson
birthan
dearly
child
hood
outcom
es.
Stud
yStud
yde
sign
Results
Con
sum
ptio
nre
spon
sest
oin
-kin
dtr
ansf
ers:
evid
ence
from
the
intr
oduc
tion
ofth
eFo
odSt
amp
prog
ram
(Hoy
nesa
ndSc
hanz
enba
ch,20
09).
Dat
afr
omPS
IDfo
r19
68-7
8an
dfr
omth
e19
60,19
70,an
d19
80de
cenn
ial
cens
uses
.n=
39,6
23pe
rson
-yea
rob
s.
Diff
eren
ce-i
n-di
ffer
ence
usin
gva
riat
ion
inco
unty
-lev
elim
plem
enta
tion
ofFS
Pto
estim
ate
impa
ctof
FSP
onfo
odco
nsum
ptio
nan
dla
bor
supp
ly.C
ontr
olle
dfo
rco
unty
and
year
fixe
deff
ects
asw
ella
sst
ate
linea
rtim
etr
ends
.In
clud
edtr
ends
inte
ract
edw
/pr
e-tr
eatm
entc
hara
cter
istic
san
dth
ree
mea
sure
sofa
nnua
lper
capi
taco
unty
tran
sfer
paym
ents
.
Intr
oduc
tion
ofFS
Pis
asso
ciat
edw
ith:18
%in
crea
sein
tota
lfoo
dex
pend
iture
s(w
hole
sam
ple)
;26
-28%
incr
ease
into
talf
ood
expe
nditu
resf
orfe
mal
e-he
aded
HH
s;6-
13%
incr
ease
into
talf
ood
exp.
for
non-
whi
tefe
mal
e-he
aded
HH
s.N
osig
nifica
nteff
ecto
nm
eals
outa
ndca
shex
pend
iture
son
food
atho
me.
Ela
stic
ityof
food
spen
ding
with
resp
ectt
oin
com
e=
0.30
.M
PCfo
rfo
odou
tofc
ash
inco
me=
0.09
(for
who
lesa
mpl
e);
MPC
for
food
outo
fcas
hin
com
e=
0.11
1(<
$25,
000
inco
me)
;M
PCfo
rfo
odou
tofF
SPin
com
e=
0.16
(for
who
lesa
mpl
e);
MPC
for
food
outo
fFSP
inco
me=
0.23
8(<
$25,
000
inco
me)
.D
ecre
ase
inw
heth
erth
eH
Hhe
adre
port
sany
wor
kby
21%
.
The
re’s
nosu
chth
ing
asa
free
lunc
h:A
ltrui
stic
pare
ntsa
ndth
ere
spon
seof
hous
ehol
dfo
odex
pend
iture
sto
nutr
ition
prog
ram
refo
rms(
Bin
gley
and
Wal
ker,
2007
).D
ata
from
UK
Fam
ilyE
xpen
ditu
reSu
rvey
sfor
1981
-199
2.n=
29,22
2.
Ana
lyze
d3
nutr
ition
prog
ram
sin
the
UK
:fr
eesc
hool
lunc
hfo
rch
ildre
nfr
ompo
orH
Hs,
free
milk
topo
orH
Hsw
/pr
e-sc
hool
child
ren,
and
free
milk
atda
yca
refo
rpr
e-sc
hool
ersi
nat
tend
ance
rega
rdle
ssof
inco
me.
For
iden
tifica
tion,
expl
oite
d19
88re
form
that
ende
del
igib
ility
for
poor
HH
sw
ithw
orki
ngpa
rent
s.D
iffer
ence
-in-
differ
ence
(DD
).A
lsodi
dD
Dus
ing
the
fact
that
free
scho
ollu
nche
sav
aila
ble
only
duri
ngte
rmtim
e,an
dsu
mm
erho
liday
sbeg
inea
rlie
rin
Scot
land
.
Free
scho
ollu
nch
redu
cesf
ood
expe
nditu
reby
15%
ofth
epu
rcha
sepr
ice
ofth
elu
nch.
Free
pint
ofm
ilkre
duce
smilk
expe
nditu
reby
80%
.
(con
tinue
don
next
page
)
1408 Douglas Almond and Janet Currie
Table7(con
tinue
d)Stud
yStud
yde
sign
Results
Impa
ctof
Food
Stam
pPr
ogra
m(F
SP)o
nbi
rthw
eigh
t,ne
onat
alm
orta
lity,
and
fert
ility
(Alm
ond
etal
.,20
09).
Bir
than
dde
ath
mic
roda
tafr
omth
eN
atio
nal
Cen
ter
for
Hea
lthSt
atist
ics
mer
ged
with
coun
ty-l
evel
data
for
1968
-77.
FSP
data
from
USD
A.
Cou
nty
char
acte
rist
icsf
rom
1960
City
and
Cou
nty
Dat
aB
ook.
Dat
aon
gove
rnm
entt
rans
fers
and
per-
capi
tain
com
efr
omR
EIS
.Pa
rtic
ipat
ion
rate
scal
cula
ted
usin
gC
PS.n=
97,78
5w
hite
s;27
,274
blac
ks.
Diff
eren
ce-i
n-di
ffer
ence
,us
ing
the
fact
that
FSP
was
intr
oduc
edto
differ
entc
ount
iesa
tdi
ffer
entt
imes
due
toav
aila
ble
fund
ing
and
polic
ych
ange
s.K
eypo
licy/
trea
tmen
tva
riab
leis
the
mon
than
dye
arth
atea
chco
unty
impl
emen
ted
FSP.
Est
imat
edth
eim
pact
ofFS
Pon
coun
ty-l
evel
birt
hou
tcom
es,us
ing
coun
tyan
dtim
efixe
deff
ects
.M
ain
outc
omes
conc
erne
dw
ithav
aila
bilit
yof
FSP
duri
ng3r
dtr
imes
ter
ofpr
egna
ncy.
Intr
oduc
tion
ofFS
Pin
3rd
trim
este
rle
dto
:0.
06-0
.08%
(0.1
-0.2
%)i
ncre
ase
inbi
rth
wei
ghtf
orw
hite
s(bl
acks
);a
1%(0
.7-1
.5%
)dec
reas
ein
frac
tion
oflo
wbi
rth
wei
ghtf
orw
hite
s(bl
acks
).In
signi
fica
ntim
pact
sofe
xpos
ure
toFS
Pdu
ring
earl
ier
trim
este
rs.
Res
ults
robu
stto
addi
ngco
unty
&tim
efixe
d-eff
ecta
ndot
her
cont
rols,
asw
ella
svar
ious
time
tren
dsto
the
anal
ysis.
Res
ults
robu
stto
cond
uctin
gan
even
tstu
dyan
alys
is.
Effec
tsof
FSP
bene
fits
onw
eigh
tga
ined
byex
pect
antm
othe
rs(B
aum
,20
08).
Dat
afr
omth
eN
LSY.
Lim
ited
tolo
w-i
ncom
ebl
ack
and
Hisp
anic
wom
enw
/pr
egna
ncy
info
rmat
ion
inth
esu
rvey
s.n=
1477
preg
nanc
y-le
velo
bs.
Ran
dom
effec
tsm
odel
susin
gH
eckm
anan
dSi
nger
met
hod
tom
odel
unob
serv
edhe
tero
gene
ity.D
epen
dent
vari
able
isw
heth
erw
omen
gain
corr
ecta
mou
ntof
wei
ghtd
urin
gpr
egna
ncy
base
don
pre-
preg
nanc
yB
MI.
Ass
ume
stat
eva
riat
ion
inFS
Pel
igib
ility
rule
s,an
dpr
ogra
mad
min
istra
tion
affec
tsFS
Pta
keup
butn
otw
eigh
tgai
n.C
ontr
olfo
rge
stat
ion,
pre-
preg
nanc
yFS
P,W
IC.
Incr
easin
gav
erag
em
onth
lyFS
Pbe
nefits
from
$0to
$100
decr
ease
spro
babi
lity
ofga
inin
gto
olit
tlew
eigh
tby
11.8
-13.
7%.N
oeff
ecto
npr
obab
ility
ofga
inin
gto
om
uch
wei
ght.
No
stat
istic
ally
signi
fica
ntdi
ffer
ence
ineff
ects
ofFS
Pon
wei
ght
gain
betw
een
firs
t-tim
ean
dno
n-firs
t-tim
em
othe
rs.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1409
Table7(con
tinue
d)Stud
yStud
yde
sign
Results
Impa
ctof
FSP
onbi
rth
outc
omes
inC
alifo
rnia
(Cur
rie
and
Mor
etti,
2008
).D
ata
onFS
Pfr
omst
ate
reco
rdsa
ndR
EIS
.D
ata
from
birt
hre
cord
sin
CA
for
1960
-74.
Agg
rega
ted
data
into
cells
define
dus
ing
coun
ty,ra
ce,ye
arof
birt
h,m
ater
nala
gegr
oup,
pari
ty,a
ndth
eth
ird
ofth
eye
ar.n=
38,4
75ce
lls.
Diff
eren
ce-i
n-di
ffer
ence
usin
gco
unty
-lev
elva
riat
ion
intim
ing
ofFS
Pin
trod
uctio
n.FS
Pm
easu
red
usin
gdu
mm
y(=
1if
FSP
intr
oduc
ed),
log
expe
nditu
res,
orlo
gpa
rtic
ipat
ion.
FSP
dum
my
refe
rsto
9m
onth
spri
orto
birt
h.C
ount
yfixe
deff
ects
and
coun
ty-s
peci
fic
time
tren
dsin
clud
ed.
Exa
min
edte
enag
em
othe
rsan
dLA
coun
tyse
para
tely
.
FSP
intr
oduc
tion
led
toa
10%
incr
ease
innu
mbe
rof
firs
tbir
thst
ow
hite
teen
mot
hers
(onl
yin
Los
Ang
eles
);a
24%
incr
ease
innu
mbe
rof
firs
tbir
ths
tobl
ack
teen
mot
hers
;a
12%
incr
ease
innu
mbe
rof
firs
tbir
thst
oal
lbla
cks;
a0.
1%in
crea
sein
prob
abili
tyin
fant
1500
-200
0g
surv
ives
for
whi
tes;
a4%
decr
ease
inpr
obab
ility
infa
nt<
3000
gsu
rviv
esfo
rbl
acks
;a
4%in
crea
sein
prob
abili
tyof
low
birt
hw
eigh
tam
ong
whi
tete
ens.
Unl
esso
ther
wise
note
d,al
lrep
orte
dre
sults
are
stat
istic
ally
signi
fica
ntat
5%le
vel.
Perc
entc
hang
es(d
enot
edby
%in
stea
dof
“pp”
)are
repo
rted
rela
tive
toth
em
ean.
1410 Douglas Almond and Janet Currie
literature examining this question (see Currie (2003) for a summary). The modal studycompares eligible participants to eligible non-participants using a multiple regressionmodel. The main problem with drawing inferences about the efficacy of the FSPfrom this exercise is that participants are likely to differ from eligible non-participantsin ways that are not observed by the researcher. Thus, for example, Basiotis et al.(1998) and Butler and Raymond (1996) both find that participation in the FSP reducesconsumption of some important nutrients. Since it is hard to imagine how giving peoplefood coupons could do this, one suspects that these results are driven by negative selectioninto the FSP program.
Several recent papers examining the effects of the FSP on young children aresummarized in Table 7. Currie and Moretti (2008) were the first to try to use variationin the timing of the introduction of the Food Stamp program to look at effects onbirth outcomes. Using Vital Statistics Natality data from California, they find that theintroduction of the FSP increased the number of births, particularly in Los AngelesCounty. They also find some evidence that the FSP increased the probability of fetalsurvival among the lightest white infants, but the effect is very small, and only detectablein Los Angeles. Notably, the FSP increased (rather than decreased) the probability of lowbirth weight but the estimated effect is small, and concentrated among teenagers givingbirth for the first time. Thus, it appears that in California, the FSP increased fertility andinfant survival (in some groups) with overall zero or negative effects on the distributionof birth weight.
Almond et al. (forthcoming) examine the same question using national data, and focuson receipt of the FSP during the third trimester, when the fetus typically puts on most ofthe weight the baby will have at birth. In contrast to Currie and Moretti, they find thatthe introduction of the FSP increased birth weights for whites and had even larger effectson blacks. The percentage reductions in the incidence of low birth weight were greaterthan the percentage increases in mean birth weight, suggesting that the FSP had its largesteffects at the bottom of the birth weight distribution. Almond et al. find no effect of FoodStamp receipt in the first trimester of pregnancy and much weaker evidence for effects ofreceipt in the second. This suggests that one reason for the contrast between their resultsand those of Currie and Moretti is that the latter did not focus narrowly enough on therelevant part of pregnancy. Moreover, Almond et al. find larger effects in the South thanin other regions, raising the possibility that overall effects were smaller in California thanin other regions. Finally, it is possible that the effects in California are obscured by thesubstantial in-migration that the state experienced over this period.
Baum (2008) examines the effects of the FSP on weight gain among pregnant women,
with particular attention to whether women gained either less than the recommendedamount or greater than the recommended amount given their pre-pregnancy bodymass index. He estimates a simultaneous equations model in which weight gain andFSP participation are jointly determined. FSP participation is assumed to be affected by
Human Capital Development before Age Five 1411
various state-level rules about eligibility, outreach and so on. One difficulty is that theserules may be affected by other characteristics of states (such as overall generosity of socialprograms) which have direct effects on weight gain (e.g. through superior access to healthcare during pregnancy). Baum finds that FSP participation reduces the probability thatwomen experienced inadequate weight gain during pregnancy, but has no effect on theprobability that they gained too much weight. Since inadequate maternal weight gain isan important risk factor for low birth weight, it is likely that FSP had a positive effect onbirth weights among affected mothers.
As discussed above, the other large category of “near cash” offer subsidized housing.
Many OECD countries have large housing assistance programs, but their effects onfamilies are understudied. In fact, we were able to find only one paper that examinedthe effects of housing programs on the outcomes of children less than five, and onlya handful that examined effects on children at all. These studies are summarized inTable 8.
Since by design, families receiving housing assistance are among the poorest of thepoor, it is clearly important to address the endogeneity of program receipt. Currie andYelowitz (2000) look at the effects of living in a public housing project in families withtwo children. They combine information from the Census and from the Survey ofIncome and Program Participation in a two-sample instrumental variables frameworkwhere the instrument for receipt of housing assistance is the sex composition of thesiblings (families with a boy and a girl are entitled to larger apartments, and so are morelikely to take up housing benefits). They find that families living in projects are less likelyto be subject to overcrowding and that the children are much less likely to have been heldback in school. The latter effect is three times bigger for boys (who are more likely to beheld back in any case) than for girls. Since most “holding back” occurs at younger ages(Kindergarten and grade 1), this suggests that this type of assistance is in fact beneficialfor young children.
Goux and Maurin (2005) focus on the effect of overcrowding in France using asimilar instrumental variables strategy: They argue that children in families in which thetwo eldest children are the same sex are more likely to live in crowded conditions inchildhood. They also propose an alternative strategy in which crowding is instrumentedwith whether or not the parent was born in an urban area—parents who are from urbanareas are more likely to live in crowded conditions. They find evidence consistent withCurrie and Yelowitz in that crowding has a large and significant effect on the probabilitythat a child falls behind in school and eventually drops out.
Fertig and Reingold (2007) examine the effect of receipt of public housing assistanceusing data from the Fragile Families Study and three instruments: the gender compositionof children in the household, the supply of public housing in each location, and thelength of waiting lists in each location. They find mixed estimates of effects on maternalhealth and little evidence of an effect on child health, though their samples are quite
1412 Douglas Almond and Janet Currie
Table8
Effectsof
housingan
dne
ighb
orho
odson
child
outcom
es.
Stud
yan
dda
taStud
yde
sign
Results
Hou
sing
andch
ildou
tcom
es
Are
publ
icho
usin
gpr
ojec
tsgo
odfo
rki
ds?
(Cur
rie
and
Yelo
witz
,20
00).
Dat
afr
omSI
PPfo
r19
92-1
993,
Mar
chC
PSfo
r19
90-1
995,
and
US
Cen
susf
or19
90on
fam
ilies
w/
2ch
ildre
n(u
nder
18)a
ndin
com
e<
$50,
000.
n=
279,
129.
Two-
sam
ple
inst
rum
enta
lvar
iabl
e(T
SIV
)oft
heeff
ects
ofliv
ing
inpu
blic
hous
ing
proj
ects
onch
ild’s
educ
atio
n,ho
usin
gco
nditi
ons.
Out
com
eva
riab
lesa
rein
SIPP
and
Cen
sus.
End
ogen
ous
regr
esso
ris
inC
PS.In
stru
men
tisa
dum
my
for
siblin
gsbe
ing
ofdi
ffer
ents
exes
since
fam
ilies
with
differ
ents
exch
ildre
nge
tlar
ger
apar
tmen
tsin
publ
icho
usin
gth
anfa
mili
esw
ithsa
me
sex
child
ren.
Con
trol
led
for
per-
capi
taav
aila
bilit
yof
proj
ects
,vo
uche
rs,Se
ctio
n8
subs
idie
s,as
wel
las
othe
rne
ighb
orho
odan
dfa
mily
back
grou
ndch
arac
teri
stic
s.
Firs
tsta
gere
sults
:H
avin
gsib
lings
ofdi
ffer
ent
sex
incr
ease
slik
elih
ood
offa
mily
livin
gin
proj
ectb
y24
%.
Seco
ndst
age
resu
lts:Fa
mili
esw
holiv
ein
proj
ects
are
16pp
(mea
n=
0.04
)les
slik
ely
tobe
over
crow
ded;
and
12pp
(mea
n=
0.02
)les
slik
ely
toliv
ein
high
-den
sity
hous
ing.
Chi
ldre
nw
holiv
ein
proj
ects
are
111%
less
likel
yto
have
been
held
back
insc
hool
(larg
ereff
ects
for
boys
than
for
girl
s).B
lack
child
ren
who
live
inpr
ojec
tsar
e19
%le
sslik
ely
toha
vebe
enhe
ldba
ckin
scho
ol.N
ost
atist
ical
lysig
nifica
nteff
ectf
orw
hite
child
ren.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1413
Table8(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
long
-ter
meff
ects
ofpu
blic
hous
ing
onse
lf-su
ffici
ency
(New
man
and
Har
knes
s,20
02)
Dat
afr
omPS
ID—
Ass
isted
hous
ing
data
base
onco
hort
sbor
nb/
n19
57an
d19
67.Sa
mpl
elim
ited
toin
divi
dual
swho
sefa
mili
esw
ere
elig
ible
for
publ
icho
usin
gb/
nag
es10
and
16.
n=
1183
.A
dult
outc
omes
mea
sure
dat
ages
20-2
7.
Am
emiy
age
nera
lized
leas
tsqu
ares
regr
essio
nw
here
the
inst
rum
entw
asth
eve
ctor
ofre
sidua
lsfr
oma
regr
essio
nof
the
num
ber
ofpu
blic
hous
ing
units
per
elig
ible
fam
ilyin
the
area
onde
mog
raph
icch
arac
teri
stic
soft
hear
ea.
Inst
rum
ente
dfo
rw
heth
erch
ildliv
edin
publ
icho
usin
gto
estim
ate
impa
cton
vari
ousa
dult
outc
omes
.C
ontr
olle
dfo
rnu
mer
ousb
ackg
roun
dch
arac
teri
stic
sand
stat
ean
dye
arfixe
deff
ects
.
Livi
ngin
publ
icho
usin
gas
ach
ildle
adst
oan
incr
ease
inth
epr
obab
ility
ofan
yem
ploy
men
tb/
nag
es25
and
27of
7.8%
;an
incr
ease
inan
nual
earn
ings
b/n
ages
20an
d27
of14
.3%
;an
incr
ease
inth
enu
mbe
rof
year
snot
onw
elfa
reb/
nag
es20
and
27of
11.3
%.N
ost
atist
ical
lysig
nifica
nteff
ecto
nho
useh
old
earn
ings
rela
tive
toth
epo
vert
ylin
e.N
ote:
abov
ere
sults
signi
fica
ntat
6%le
vel.
Publ
icH
ousin
g,H
ousin
gVo
uche
rs,an
dSt
uden
tA
chie
vem
ent:
Evi
denc
efr
omPu
blic
Hou
sing
Dem
oliti
onsi
nC
hica
go(J
acob
,20
04)
Adm
inist
rativ
eda
tafr
omth
eC
hica
goH
ousin
gA
utho
rity
and
Chi
cago
Publ
icSc
hool
sfor
1991
-200
2on
stud
ents
who
lived
inhi
gh-r
isepu
blic
hous
ing
for
atle
asto
nese
mes
ter.
n=
10,5
56.
Inst
rum
enta
lvar
iabl
esre
gres
sion
ofth
eeff
ecto
fliv
ing
ina
publ
icho
usin
gpr
ojec
ton
stud
ent
outc
omes
.Id
entifi
catio
nfr
omva
riat
ion
intim
ing
ofho
usin
gde
mol
ition
sin
Chi
cago
inth
e19
90s.
Inst
rum
ente
dfo
rliv
ing
inpu
blic
hous
ing
with
adu
mm
yfo
rw
heth
erth
est
uden
tliv
edin
aun
itsc
hedu
led
for
dem
oliti
onat
the
time
ofth
ecl
osur
ean
noun
cem
ent.
Con
trol
sfor
back
grou
ndch
arac
teri
stic
sand
proj
ecta
ndye
arfixe
deff
ects
.
No
stat
istic
ally
signi
fica
nteff
ecto
fliv
ing
inhi
gh-r
isepr
ojec
tson
stud
ento
utco
mes
.Li
ving
ina
build
ing
that
was
dem
olish
edle
ads
toan
8.2%
incr
ease
inpr
obab
ility
that
stud
ents>
14yr
sdro
pou
tofs
choo
lwith
in3
year
srel
ativ
eto
thos
eth
atliv
edin
build
ings
sche
dule
dfo
rde
mol
ition
that
had
noty
etbe
ende
mol
ished
(12%
mor
elik
ely
for
girl
s;4%
mor
elik
ely
for
boys
). (con
tinue
don
next
page
)
1414 Douglas Almond and Janet Currie
Table8(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Publ
icho
usin
g,he
alth
and
heal
thbe
havi
ors:
isth
ere
aco
nnec
tion?
(Fer
tigan
dR
eing
old,
2007
).D
ata
from
the
frag
ilefa
mili
esan
dch
ildw
ell-
bein
gst
udy
onch
ildre
nbo
rnin
20U
Sci
tiesb
etw
een
1998
and
2000
.Sa
mpl
e1:
T=
422,
C=
2,05
5;Sa
mpl
e2:
T=
323,
C=
1,99
9;Sa
mpl
e3:
T=
150,
C=
919.
Inst
rum
enta
lvar
iabl
esre
gres
sion.
Thr
eein
stru
men
ts:ge
nder
com
posit
ion
ofho
useh
old,
the
supp
lyof
publ
icho
usin
gin
each
city
,an
dle
ngth
ofw
aitin
glis
t.T
hree
subs
ampl
es.
“Con
trol
”gr
oup
ism
othe
rsw
hoha
vebe
low
80%
ofm
edia
nin
com
ein
thei
rar
eabu
tare
noti
npu
blic
hous
ing.
“Tre
atm
ent1
”is
mot
hers
who
live
inpu
blic
hous
ing
attim
eof
firs
tint
ervi
ewim
med
iate
lyaf
ter
child
birt
h.“T
reat
men
t2”
ism
othe
rsw
hoha
vem
oved
into
publ
icho
usin
gbe
twee
nch
ildbi
rth
and
seco
ndin
terv
iew
atch
ild’s
age
1(“
mov
e-in
”su
bsam
ple)
.“T
reat
men
t3”
isth
e“‘
mov
e-in
”su
bsam
ple
limite
dto
mot
hers
with
two
orth
ree
child
ren
atse
cond
inte
rvie
w.C
ontr
olle
dfo
rfa
mily
,ba
ckgr
ound
,ne
ighb
orho
od,an
dot
her
dem
ogra
phic
and
soci
o-ec
onom
icch
arac
teri
stic
s.
Not
e:on
lyre
port
ing
IVes
timat
esw
ithco
mpl
ete
com
bina
tion
ofin
stru
men
tshe
re.
Effec
tsofm
ovin
gin
topublic
housi
ng
bet
wee
nch
ildbir
than
done-
year
inte
rvie
w:
1.63
poin
tinc
reas
ein
mot
her’s
heal
thin
dex
(5po
ints
cale
,m
ean
notr
epor
ted,
can’
tca
lcul
ate
effec
tsiz
e)55
ppde
crea
sein
likel
ihoo
dth
atm
othe
rha
slim
iting
heal
thco
nditi
on(m
ean=
0.08
)41
ppde
crea
sein
likel
ihoo
dth
atm
othe
ris
unde
rwei
ght(
mea
n=
0.07
)18
%in
crea
sein
mot
her’s
BM
ID
ecre
ase
indo
mes
ticvi
olen
ce(im
prec
iseco
effici
ente
stim
ate)
No
stat
istic
ally
signi
fica
nteff
ects
onch
ild’s
birt
hw
eigh
tor
PPV
Tsc
ores
.
Neigh
borhoo
dch
aracteristicsan
dch
ildou
tcom
es
The
long
-run
cons
eque
nces
ofliv
ing
ina
poor
neig
hbor
hood
(Ore
opou
los,
2003
)D
ata
from
the
inte
rgen
erat
iona
lin
com
eda
taba
sefo
r19
78-1
999
(tak
enfr
omin
com
eta
xfile
s)on
indi
vidu
alsl
ivin
gin
Toro
nto,
born
b/n
1963
and
1970
.n=
4060
.
Am
ong
fam
ilies
who
appl
yfo
rho
usin
gpr
ojec
ts,
assig
nmen
tto
apa
rtic
ular
proj
ecti
sapp
roxi
mat
ely
rand
om—
base
don
1sta
vaila
ble
unit.
Com
pare
dho
usin
g-pr
ojec
tmea
nsof
vari
ousa
dult
outc
omes
acro
ssne
ighb
orho
odqu
ality
(mea
sure
dby
dens
ityof
hous
ing,
tota
lsiz
eof
proj
ect,
prop
ortio
nof
the
Cen
sust
ract
belo
wth
elo
w-i
ncom
ecu
toff
,an
dw
heth
erth
epr
ojec
tisa
llhi
gh-r
ises)
.
The
qual
ityof
hous
ing
proj
ecto
rne
ighb
orho
odha
sno
stat
istic
ally
signi
fica
ntim
pact
onto
tali
ncom
e,an
nual
earn
ings
,or
num
ber
ofye
arso
nw
elfa
rein
adul
thoo
d.Fa
mily
char
acte
rist
icsa
ccou
ntfo
rup
to30
%of
tota
lvar
iatio
nin
adul
tout
com
es.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1415
Table8(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Nei
ghbo
rhoo
dE
ffec
tson
Cri
me
for
Fem
ale
and
Mal
eYo
uth:
Evi
denc
efr
oma
Ran
dom
ized
Hou
sing
Vouc
her
Exp
erim
ent
(Klin
get
al.,
2005
)D
ata
from
Mov
ing
toO
ppor
tuni
ties(
MT
O)
expe
rim
ent.
Fam
ilies
livin
gin
publ
icho
usin
gpr
ojec
tsin
Bos
ton,
LosA
ngel
es,N
ewYo
rkC
ity,
Bal
timor
e,an
dC
hica
gow
ere
rand
omly
chos
ento
stay
inth
eir
curr
enth
ome,
tore
ceiv
ea
Sect
ion
8ho
usin
gvo
uche
r,or
tore
ceiv
ea
Sect
ion
8ho
usin
gvo
uche
rre
stri
cted
tous
ein
neig
hbor
hood
sw
/po
vert
yra
tele
ssth
an10
%.
Surv
eysc
ondu
cted
in20
02.D
ata
onyo
uths
aged
15-2
5.D
ata
onar
rest
reco
rdsf
orM
TO
stat
es.
n=
4475
.
OLS
regr
essio
nof
crim
eou
tcom
eson
dum
my
for
whe
ther
fam
ilyus
eda
vouc
her
(inst
rum
ente
dby
whe
ther
fam
ilyw
asas
signe
dto
trea
tmen
tgro
up)
oron
vouc
her
type
,an
don
anag
e-tr
eatm
ent
inte
ract
ion
vari
able
.In
stru
men
tnec
essa
rysin
ceno
tall
fam
ilies
who
wer
eas
signe
da
vouc
her
chos
eto
use
it-
som
ere
mai
ned
inth
eir
initi
alho
usin
g.
Ass
ignm
entt
oa
rest
rict
edvo
uche
rle
adst
oa
31%
decr
ease
invi
olen
tcri
me
arre
stsf
orfe
mal
eyo
uths
.U
seof
rest
rict
edvo
uche
rle
ads
toa
76%
redu
ctio
nin
viol
entc
rim
ear
rest
sam
ong
fem
ale
yout
hs.N
ost
atist
ical
lysig
nifica
nteff
ects
onm
ales
for
viol
entc
rim
ear
rest
s.A
ssig
nmen
tto
are
stri
cted
vouc
her
lead
sto
a33
%de
crea
sein
prop
erty
crim
ear
rest
sfor
fem
ale
yout
hs.U
seof
rest
rict
edvo
uche
rle
adst
oa
85%
decr
ease
inpr
oper
tycr
ime
arre
stsf
orfe
mal
eyo
uths
.Ass
ignm
entt
oa
rest
rict
edvo
uche
rle
adst
oa
32%
incr
ease
inpr
oper
tycr
ime
arre
stsf
orm
ale
yout
hs.U
seof
rest
rict
edvo
uche
rle
adst
oa
77%
incr
ease
inpr
oper
tycr
ime
arre
stsf
orm
ale
yout
hs.
(con
tinue
don
next
page
)
1416 Douglas Almond and Janet Currie
Table8(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Nei
ghbo
rhoo
dsan
dA
cade
mic
Ach
ieve
men
t(Sa
nbon
mat
suet
al.,
2006
)D
ata
from
MT
Oex
peri
men
tsu
rvey
sfor
2002
(see
abov
e)on
yout
hsag
ed6-
20.n=
5074
.
Sam
ere
gres
sion
asK
ling
etal
.(2
005)
for
educ
atio
nalo
utco
mes
.C
hild
ren
infa
mili
esas
signe
dto
are
stri
cted
vouc
her
atte
ndbe
tter
qual
itysc
hool
s:th
em
ean
rank
ofsc
hool
son
stat
eex
amsi
s30%
grea
ter
than
the
cont
rolg
roup
mea
n,an
dth
em
ean
prop
ortio
nof
scho
ol-l
unch
-elig
ible
child
ren
is8%
low
er.N
ost
atist
ical
lysig
nifica
nteff
ects
ofei
ther
vouc
her
onch
ilded
ucat
iona
lout
com
es(s
choo
latt
enda
nce,
whe
ther
child
does
hom
ewor
k,ch
ild’s
beha
vior
incl
ass,
whe
ther
child
atte
nded
clas
sfo
rgi
fted
stud
ents
orcl
assf
orsp
ecia
lhel
p,an
dW
oodc
ock-
John
son
Rev
ised
read
ing
and
mat
hte
stsc
ores
).
Clo
seN
eigh
borh
oods
Mat
ter:
Nei
ghbo
rhoo
dE
ffec
tson
Ear
lyPe
rfor
man
ceat
Scho
ol(G
oux
and
Mau
rin,
2005
).D
ata
from
the
Fren
chLa
bor
Forc
eSu
rvey
for
1991
-200
2on
16-y
ear-
old
yout
hs.n=
1311
6.
Aut
hors
assu
me
that
ach
ild’s
birt
hdat
ew
illim
pact
his/
her
own
educ
atio
nalo
utco
mes
,bu
tnot
the
outc
omes
ofhi
s/he
rne
ighb
ors.
(Pro
port
ion
of15
-yr-
olds
held
back
agr
ade
is15
pphi
gher
amon
gth
ose
born
atth
een
dof
the
year
rela
tive
toth
ose
born
atth
ebe
ginn
ing
ofth
eye
ar).
IVre
gres
sion,
whe
refirs
tsta
gesa
re:re
gres
sion
ofbe
ing
held
back
atag
e15
ontim
ing
ofbi
rthd
ayan
dre
gres
sion
ofpr
opor
tion
ofne
ighb
orho
odyo
uths
held
back
atag
e15
onth
eir
birt
hday
s.Se
cond
stag
eis
regr
essio
nfo
rbe
ing
held
back
atag
e16
onha
ving
been
held
back
atag
e15
and
prop
ortio
nof
neig
hbor
hood
yout
hshe
ldba
ckat
age
15.C
ontr
olle
dfo
rye
arfixe
deff
ects
,ge
nder
,an
dna
tiona
lity.
1SD
incr
ease
inth
epr
opor
tion
ofne
ighb
orin
gyo
uth
that
wer
ehe
ldba
ckin
scho
olat
age
15in
crea
sest
hepr
obab
ility
ofbe
ing
held
back
atag
e16
by0.
2SD
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1417
Table8(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Exp
erim
enta
lana
lysis
ofne
ighb
orho
odeff
ects
(Klin
get
al.,
2007
).D
ata
from
MT
Oex
peri
men
tsu
rvey
sfor
2002
(see
abov
e)on
yout
hsag
ed15
-20.
n=
1807
.
Sam
ere
gres
sion
as(K
ling
etal
.,20
05)f
orhe
alth
and
beha
vior
outc
omes
.A
ssig
nmen
tto
anun
rest
rict
edvo
uche
rle
ads
toa
decr
ease
inlik
elih
ood
ofex
peri
enci
ngan
xiet
ysy
mpt
omsb
y62
%/9
1%am
ong
fem
ales
/mal
es;a
decr
ease
inm
ariju
ana
use
by44
%am
ong
fem
ales
;an
incr
ease
inth
elik
elih
ood
ofno
n-sp
orts
inju
ryby
130%
amon
gm
ales
;an
incr
ease
inin
cide
nce
ofsm
okin
gby
121%
amon
gm
ales
.A
ssig
nmen
tto
rest
rict
edvo
uche
rle
adst
oa
decr
ease
inlik
elih
ood
ofex
peri
enci
ngps
ycho
logi
cal
dist
ress
by10
0%am
ong
fem
ales
;a
decr
ease
inlik
elih
ood
ofex
peri
enci
ngan
xiet
yby
57%
amon
gfe
mal
es;a
decr
ease
inm
ariju
ana
use
by50
%am
ong
fem
ales
;an
incr
ease
inlik
elih
ood
ofno
n-sp
orts
inju
ryby
140%
amon
gm
ales
;an
incr
ease
inin
cide
nce
ofsm
okin
gby
82%
amon
gm
ales
.
Unl
esso
ther
wise
note
d,on
lyre
sults
signi
fica
ntat
5%le
vela
rere
port
edhe
re.
1418 Douglas Almond and Janet Currie
small. Newman and Harkness (2002) use data from the PSID to examine the effectof living in public housing as a child on future earnings and employment. Living inpublic housing is instrumented using the residual from a regression of local housingsupply on the demographic characteristics of the area. They find that public housingis associated with increases in the probability of any employment (from 88% to about95%) and increases in annual earnings (by $1861 from a mean of $11,210). While allof these instrumental variables strategies are subject to caveats (is gender compositionreally uncorrelated with sibling’s outcomes? Are characteristics of local housing marketsassociated with unobserved factors such as the quality of schools that might also affectchild outcomes?) they certainly all point in a similar direction.
An important question is whether public housing assistance benefits children morethan the equivalent cash transfer. It is difficult to answer this question given the availabledata. However, it is possible to eliminate some possible channels through which publichousing programs might have different effects. One is that public housing programsmay constrain the recipient’s choice of neighborhoods, with either positive or negativeeffects. Jacob (2004) studies students displaced by demolitions of the most notoriousChicago high-rise projects. The US Congress passed a law in 1996 that required localhousing authorities to destroy units if the cost of renovating and maintaining them wasgreater than the cost of providing a voucher for 20 years. Jacobs argues that the orderin which doomed buildings were destroyed was approximately random. For example, inJanuary 1999, the pipes froze in some buildings in the Robert Taylor Homes, whichmeant that those buildings were demolished before others in the same complex. Bycomparing children who stayed in buildings scheduled to be demolished to otherswho had already been displaced by demolitions, he obtains a measure of the effect ofliving in high-rise public housing. Despite the fact that the high rises in Jacob’s studywere among the most notorious public housing projects in the country, he finds verylittle effect of relocation on children’s educational outcomes. However, this may bebecause for the most part, children stayed in the same neighborhoods and in the sameschools.
The most exhaustive examination of the effects of giving vouchers to project residentsis an ongoing experiment called “Moving to Opportunity” (MTO). MTO was inspiredby the Gautreaux program in Chicago, which resulted from a consent decree designedto desegregate Chicago’s public housing by relocating some black inner-city residentsto white suburbs. MTO is a large-scale social experiment that is being conducted inChicago, New York, Los Angeles, Boston and Baltimore (see Orr et al. (2003), Klinget al. (2005) and Sanbonmatsu et al. (2006)). Between 1994 and 1998, volunteers frompublic housing projects were assigned by lottery to one of three groups. The first groupreceived a voucher that could only be used to rent housing in a low-poverty area (aCensus tract with a poverty rate less than 10%). This group also received help locating asuitable apartment (referred to here as the “MTO group”). The second group received
Human Capital Development before Age Five 1419
a voucher which they could use to rent an apartment in any neighborhood. The thirdgroup was the control and received no vouchers or assistance although they were eligibleto remain in their project apartment. Families in the first group did move to lower povertyneighborhoods and the new neighborhoods of the MTO group were considerably safer.The move to new neighborhoods had positive effects on the mental health and schoolingattainment of girls, and negative effects on the probability that they were ever arrested.But surprisingly, MTO either had no effect, or negative effects, on boys. Boys in theexperimental group were 13% more likely than controls to have ever been arrested. Thisincrease was due largely to increases in property crimes. These boys also report morerisky behaviors such as drug and alcohol use. And boys in the MTO and voucher groupswere more likely to suffer injuries. These differences between boys and girls are apparenteven within families (Orr et al., 2003).
It remains to be seen how the long-term outcomes of the MTO children willdiffer from controls. Oreopoulos (2003) uses data from Canadian income tax recordsto examine the earnings of adults who lived in public housing projects in Toronto aschildren. There are large differences between projects in Toronto, both in terms of thedensity of the projects, and in terms of the poverty of the neighborhoods. Oreopoulosargues that the type of project a family lives in is approximately randomly assigned becausethe family is offered whatever happens to be available when they get to the top of thewaiting list. Oreopoulos finds that once the characteristics of the family are controlled,the neighborhood has no effect on future earnings or on the likelihood that someoneworks.
The findings on near cash programs can be summarized as follows. There is credibleevidence that the FSP may improve birth weight. More work remains to be done todetermine whether it has positive effects on the nutrition of children after birth, whethersimilar programs in other countries have positive effects, and whether this particular typeof in kind program has effects that are different than cash subsidies to poor households.The evidence regarding housing programs also suggests that they can be beneficial tofamilies, but offers little guidance about the important question of whether housingprograms matter primarily because they subsidize family incomes or operate throughsome other mechanism. It seems doubtful, given the available evidence, that housingprograms benefit child outcomes primarily by improving their neighborhoods (especiallysince many housing projects are located in less desirable neighborhoods).
5.3. Early intervention programsMany programs specifically seek to intervene in the lives of poor children in order toimprove their outcomes. Three interventions that have been shown to be effective arenurse home visiting programs, nutritional supplementation for pregnant women, andquality early childhood education programs. Table 9 summarizes some recent evidenceabout home visiting programs.
1420 Douglas Almond and Janet Currie
Table9
Rand
omized
trialsof
homevisitin
gprog
rams.
Stud
y/prog
ram
name
Data,prog
ram
description,
andstud
yde
sign
Results
The
Com
preh
ensiv
eC
hild
Dev
elop
men
tPr
ogra
m(C
CD
P)(S
t.Pi
erre
and
Layz
er,19
99)
Biw
eekl
yvi
sitss
tart
ing
0-1,
endi
ngat
5ye
ars.
Popu
latio
nse
rved
:43
%A
fric
an-A
mer
ican
,26
%H
ispan
ic;al
lbel
owpo
vert
y.B
ackg
roun
dof
Hom
eV
isito
rs:pa
rapr
ofes
siona
lsSa
mpl
esiz
es:T=
2,21
3,C=
2,19
7E
valu
atio
nSi
tes:
21sit
esth
roug
hout
the
US.
Age
ofch
ildre
nat
last
follo
w-u
p:5
year
sold
Dev
elop
men
talC
heck
list:
T=
57.9
3,C=
57.5
1Fo
und
nosig
nifica
nteff
ects
onw
ide
rang
eof
outc
omes
incl
udin
g:D
evel
opm
enta
ndB
ehav
ior
scor
es,m
edic
alca
re,
mor
talit
y,H
OM
Esc
ores
,m
ater
nald
epre
ssio
n,w
elfa
reus
e,m
ater
nali
ncom
e,ed
ucat
ion
orem
ploy
men
t,m
ater
nal
subs
tanc
eus
e.To
talc
ostp
erpa
rtic
ipan
t:$
37,4
88.To
tal
bene
fitp
erpa
rtic
ipan
t:$9
1.N
etpr
esen
tval
ue=−
$37,
397.
Hea
lthy
star
t(D
ugga
net
al.,
1999
,20
04;H
ardi
nget
al.,
2007
;D
uMon
teta
l.,20
06)
Wee
kly
visit
s,fa
ding
toqu
arte
rly
age
ofpa
rtic
ipat
ion:
birt
hto
5ye
ars.
popu
latio
nse
rved
:lo
w-i
ncom
e,at
-risk
fam
ilies
ofne
wbo
rnsr
ecru
ited
thro
ugh
anH
SPsc
reen
ing
and
asse
ssm
entp
roto
col.
All
Eng
lish-
spea
king
.B
ackg
roun
dof
hom
evi
sitor
s:pa
rapr
ofes
siona
ls.Sa
mpl
esiz
es:A
lask
a:T=
179,
C=
185;
Haw
aii:
T=
373,
C=
270,
6sit
es;
Vir
gini
a:T=
422,
C=
197,
2sit
es.
19ad
ditio
nals
itesd
iscus
sed
in(H
ardi
nget
al.,
2007
).A
geof
child
ren
atla
stfo
llow
-up:
2ye
ars
old
(3in
San
Die
go)
Som
epo
sitiv
eeff
ects
onpa
rent
ing
prac
tices
and
nega
tive
effec
tson
dom
estic
viol
ence
inso
me
sites
.E
.g.H
awai
ipar
tner
viol
ence
:T=
16%
,C=
24%
.Le
ssco
rpor
al/v
erba
lpun
ishm
entT
<C
(odd
srat
io0.
59).
Hea
ltheff
ects
inso
me
sites
butn
otot
hers
,e.
g.V
A:bi
rth
com
plic
atio
nsT=
0.2,
C=
0.48
;N
ewYo
rk:lo
wbi
rth
wei
ghtT=
3.3%
,C=
8.3%
;m
ater
nald
epre
ssio
nT=
23%
,C=
38%
.In
crea
sesi
nch
ildB
ayle
ySc
ale
for
Infa
ntD
evel
opm
enti
nSa
nD
iego
,A
rkan
sas.
Incr
ease
sin
mat
erna
ledu
catio
nan
dde
crea
sesi
nse
riou
schi
ldab
use
only
inN
ewYo
rk(T
12.5
%le
ssth
anC
).A
llsit
este
sted
for
aw
ide
rang
eof
poss
ible
effec
ts,w
ithge
nera
llyin
signi
fica
nteff
ects
onot
her
mea
sure
sofc
hild
wel
lbei
ngan
dch
ildab
use.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1421
Table9(con
tinue
d)Stud
y/prog
ram
name
Data,prog
ram
description,
andstud
yde
sign
Results
The
nurs
e-fa
mily
part
ners
hip
prog
ram
(Old
seta
l.,19
99)
Wee
kly
visit
s,fa
ding
tom
onth
ly,pr
enat
alto
2yr
s.Po
pula
tion
serv
ed:di
sadv
anta
ged
firs
t-tim
em
othe
rsle
ssth
an30
wee
kspr
egna
nt(6
2%un
mar
ried
,47
%te
enag
e,23
%po
or,
unm
arri
edan
dte
enag
e).
Bac
kgro
und
ofho
me
visit
ors:
Nur
ses.
Sam
ple
sizes
:C
1=
90,C
2=
94,T
3=
100
T4=
116
(see
belo
wfo
rde
scri
ptio
nof
grou
ps).
Eva
luat
ion
site:
Elm
ira,
New
York
.C
1=
scre
enin
g;C
2=
scre
enin
g&
tran
spor
tatio
n;T
3=
scre
enin
g,tr
ansp
orta
tion,
&pr
enat
alho
me
visit
s;T
4=
scre
enin
g,tr
ansp
orta
tion,
pren
atal
and
post
nata
lhom
evi
sits.
Pren
atal
anal
ysis:
T=
T3+
T4,
C=
C1+
C2;
Post
nata
lana
lysis
:T=
T4,
C=
C1+
C2.
Age
ofch
ildre
nat
last
follo
w-u
p:15
year
sold
.
Pre-
term
birt
hsfo
rw
omen
who
smok
edm
ore
than
4ci
gare
ttes
per
day:
T=
2.08
%,C=
9.81
%(m
othe
rsal
sole
sslik
ely
tosm
oke
duri
ngpr
egna
ncy,
bett
ernu
triti
ondu
ring
preg
nanc
y,m
ore
likel
yto
use
WIC
).Fo
rch
ildre
n:fe
wer
emer
genc
yro
omvi
sitsa
t0-1
2,12
-24
mon
ths.
AT
15-y
rfo
llow
-up
Mot
her’s
num
ber
ofm
onth
srec
eivi
ngA
FDC
:T
4=
60.4
,C=
90.3
;M
othe
r’ssu
bsta
nce
use
impa
irm
ents
:T
4=
0.41
,C=
0.73
;M
othe
r’sar
rest
s:T
4=
0.18
,C=
0.58
;C
onvi
ctio
ns:
T4=
0.06
,C=
0.28
;Su
bsta
ntia
ted
repo
rtso
fchi
ldab
use
and
negl
ect,
0to
15yr
s:T
4=
0.29
,C=
0.54
;C
hild
’sin
cide
nce
ofar
rest
s:T
4=
0.20
,C=
0.45
;C
hild
’sco
nvic
tions
and
prob
atio
nvi
olat
ions:
T4=
0.09
,C=
0.47
;C
hild
’snu
mbe
rof
sex
part
ners
:T
4=
0.92
,C=
2.48
;C
hild
’snu
mbe
rof
days
dran
kal
coho
l:T
4=
1.09
,C=
2.49
.C
ost
-ben
efitan
alys
is:
Tota
lcos
tper
child
:$1
0,30
0;To
talb
enefi
tper
child
:$
30,0
00.N
etpr
esen
tval
ue=+
$19,
700.
Mos
tben
efits
due
tore
duce
dcr
ime
onpa
rtof
the
child
and
redu
ctio
nsin
child
abus
eon
part
ofpa
rent
.
(con
tinue
don
next
page
)
1422 Douglas Almond and Janet Currie
Table9(con
tinue
d)Stud
y/prog
ram
name
Data,prog
ram
description,
andstud
yde
sign
Results
The
nurs
e-fa
mily
part
ners
hip
prog
ram
(Old
seta
l.,20
07)
Wee
kly
visit
s,fa
ding
tom
onth
ly,pr
enat
alto
2yr
s.Po
pula
tion
serv
ed:
firs
t-tim
em
othe
rsle
ssth
an29
wee
kspr
egna
ntan
dat
leas
t2of
the
follo
win
g:un
mar
ried
,le
ssth
an12
yrso
fed
ucat
ion,
orun
empl
oyed
;92
%A
fric
an-A
mer
ican
,98
%un
mar
ried
,64
%18
yrso
fage
oryo
unge
r,85
%at
orbe
low
pove
rty
leve
l.B
ackg
roun
dof
hom
evi
sitor
s:nu
rses
.Sa
mpl
esiz
es:C
1=
166,
C2=
515,
T3=
230,
T4=
228
(see
belo
wfo
rde
scri
ptio
nof
grou
ps).
Eva
luat
ion
sites
:M
emph
is,Te
nnes
see.
C1=
tran
spor
tatio
n;C
2=
scre
enin
g&
tran
spor
tatio
n;T
3=
scre
enin
g,tr
ansp
orta
tion,
pren
atal
hom
evi
sits,
one
visit
post
part
umin
hosp
ital,
one
post
part
umvi
sitat
hom
e;T
4=
T3
plus
hom
evi
sitst
hrou
ghch
ild’s
2nd
birt
hday
.Pr
enat
alan
alys
is:T=
T3+
T4,
C=
C1+
C2;
Post
nata
lana
lysis
:T=
T4,
C=
C2.
Age
ofch
ildre
nat
last
follo
w-u
p:9
year
sold
.
Dur
ing
firs
t2ye
arso
fchi
ld’s
life:
Num
ber
ofhe
alth
enco
unte
rsfo
rin
juri
es/i
nges
tions
:T=
0.43
,C=
0.56
Num
ber
ofou
tpat
ient
visit
sfor
inju
ries
/ing
estio
ns:
T=
0.11
,C=
0.20
Num
ber
ofda
ysho
spita
lized
for
inju
ries
/ing
estio
ns:
T=
0.04
,C=
0.18
Mot
her
atte
mpt
edbr
east
-fee
ding
:T=
26%
,C=
16%
Subs
eque
ntliv
ebi
rths
:T=
22%
,C=
31%
AT
9-yr
follo
w-u
pC
hild
GPA
(low
-res
ourc
eon
ly):
T=
2.68
,C=
2.44
Rea
ding
and
mat
hac
hiev
emen
t(lo
w-r
esou
rce
only
):T=
44.8
9,C=
35.7
2M
othe
r’s#
mon
thsw
ithcu
rren
tpar
tner
:T=
51.8
9,C=
44.4
8N
umbe
rof
mon
thso
nA
FDC
/TA
NF
per
year
:T=
5.21,C=
5.92
Num
ber
ofm
onth
son
food
stam
pspe
rye
ar:T=
6.98
,C=
7.80
Mat
erna
lmas
tery
:T=
101.
03,C=
99.5
0N
o.of
mon
thsw
ithem
ploy
edpa
rtne
r:T=
46.0
4,C=
48.4
3N
osig
nifica
nteff
ects
onE
Rvi
sitsf
orin
juri
esin
firs
t2
year
s,he
alth
atbi
rth,
use
ofm
edic
alca
re,m
ater
nalh
ealth
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1423
Table9(con
tinue
d)Stud
y/prog
ram
name
Data,prog
ram
description,
andstud
yde
sign
Results
Ear
lyst
art(
Ferg
usso
net
al.,
2006
)W
eekl
yvi
sitsf
or1s
tmon
th,th
enva
ryin
gag
eof
part
icip
atio
n:pr
enat
alto
3ye
ars.
Popu
latio
nSe
rved
:fa
mili
esre
crui
ted
thro
ugh
the
sam
esc
reen
ing
proc
essa
sin
Haw
aiiH
ealth
ySt
art.
Bac
kgro
und
ofho
me
visit
ors:
“fam
ilysu
ppor
tw
orke
rs”
with
nurs
ing
orso
cial
wor
kqu
alifi
catio
ns.
Sam
ple
sizes
:T=
220,
C=
223.
1sit
ein
New
Zea
land
.A
geof
child
ren
atla
stfo
llow
-up:
3ye
arso
ld.
Ave
rage
num
ber
ofdo
ctor
’svi
sits0
-36
mo:
T>
Cby
0.24
SD.
Perc
enta
geof
up-t
o-da
tew
ell-
child
chec
ks0-
36m
o:T>
Cby
0.25
SD.
Perc
enta
geen
rolle
dw
/de
ntist
0-36
mo:
T>
Cby
0.20
SD.
Perc
enta
geat
tend
edho
spita
lfor
acci
dent
/inj
ury
orac
cide
ntal
poiso
ning
0-36
mo:
T<
Cby
0.22
SD.
Mea
ndu
ratio
nof
earl
ych
ildho
oded
ucat
ion:
T>
Cby
0.22
SD.
Mea
nnu
mbe
rof
com
mun
ityse
rvic
eco
ntac
ts:T>
Cby
0.31
SD.
Posit
ive
pare
ntin
gat
titud
essc
ore:
T>
Cby
0.26
SD.
Non
-pun
itive
pare
ntin
gat
titud
essc
ore:
T>
Cby
0.22
SD.
Ove
rall
pare
ntin
gsc
ore:
T>
Cby
0.27
SD.
Perc
enta
geof
pare
ntal
repo
rtof
seve
reph
ysic
alas
saul
t:T<
Cby
0.26
SD.
Chi
ldin
tern
aliz
ing
(neg
ativ
e)be
havi
orsc
ore*
:T<
Cby
0.26
SD.
Chi
ldto
taln
egat
ive
beha
vior
scor
e*:T<
Cby
0.24
SD.
(con
tinue
don
next
page
)
1424 Douglas Almond and Janet Currie
Table9(con
tinue
d)Stud
y/prog
ram
name
Data,prog
ram
description,
andstud
yde
sign
Results
Effec
tiven
esso
fhom
evi
sitat
ion
bypu
blic
heal
thnu
rses
inpr
even
tion
ofth
ere
curr
ence
ofch
ildph
ysic
alab
use
and
negl
ect:
ara
ndom
ized
cont
rolle
dtr
ial
(Mac
Mill
anet
al.,
2005
)
Wee
kly
visit
sfor
firs
t6m
onth
s,th
enbi
wee
kly
Age
ofpa
rtic
ipat
ion:
Ent
ry:0-
13yr
sold
,pr
ogra
mla
sted
3ye
ars.
Popu
latio
nse
rved
:pa
rent
swho
have
are
port
edep
isode
ofph
ysic
alab
use
orne
glec
tin
the
3m
onth
spri
orto
join
ing
prog
ram
.B
ackg
roun
dof
hom
evi
sitor
s:pu
blic
heal
thnu
rses
.Sa
mpl
esiz
es:T=
73,C=
66.1
site
inH
amilt
on,C
anad
a.N
ote:
Con
trol
grou
pre
ceiv
edst
anda
rdse
rvic
esfr
omth
ech
ildpr
otec
tion
agen
cyC
PA).
Tre
atm
entg
roup
also
rece
ived
the
stan
dard
serv
ices
inad
ditio
nto
the
hom
evi
sitin
g.A
geof
child
ren
atla
stfo
llow
-up:
Vari
ed(3
-yea
rfo
llow
-up)
.
Rec
urre
nce
ofph
ysic
alab
use
orne
glec
tbas
edon
hosp
ital
reco
rds:
T=
24%
,C=
11%
Also
test
ed,bu
tfou
ndno
stat
istic
ally
signi
fica
nteff
ects
on:
Rec
urre
nce
ofab
use
orne
glec
tbas
edon
CPA
reco
rds;
HO
ME
scor
e;ch
ild’s
beha
vior
al,an
xiet
y,at
tent
ion
prob
lem
s,ag
gres
sion
orco
nduc
tdiso
rder
scor
es;pa
rent
ing
beha
vior
scor
esor
CA
Psc
ore.
Eco
nom
icev
alua
tion
ofan
inte
nsiv
eho
me
visit
ing
prog
ram
me
for
vuln
erab
lefa
mili
es:A
cost
-eff
ectiv
enes
sana
lysis
ofa
publ
iche
alth
inte
rven
tion
(McI
ntos
het
al.,
2009
)
Wee
kly
visit
sA
geof
Part
icip
atio
n:pr
enat
alto
18m
onth
sold
.Po
pula
tion
Serv
ed:Pr
egna
ntw
omen
iden
tified
byco
mm
unity
mid
wiv
esas
bein
gat
risk
for
child
abus
eor
negl
ect.
Bac
kgro
und
ofho
me
visit
ors:
para
prof
essio
nals
who
rece
ived
prog
ram
trai
ning
.Sa
mpl
eSi
zes:
T=
67,C=
64.
1sit
ein
the
UK
.A
geof
Chi
ldre
nat
Last
Follo
w-u
p:1
year
old
Mat
erna
lsen
sitiv
ity(C
AR
EIn
dex)
:T>
Cby
13%
.In
fant
coop
erat
iven
ess(
CA
RE
Inde
x):T>
Cby
18%
.Li
kelih
ood
infa
ntpl
aced
onch
ildpr
otec
tion
regi
ster
:T>
C(1
.35
times
mor
elik
ely)
.Pr
opor
tion
ofch
ildre
nre
mov
edfr
omho
me:
T=
6%,
C=
0%.
Som
eco
st-b
enefi
tana
lysis
from
Aos
etal
.(2
004)
,Te
chni
calA
ppen
dix.
“T”
refe
rsto
trea
tmen
tgro
up,“C
”;re
fers
toco
ntro
lgro
up.“T=
C”
mea
nsno
disc
erna
ble
differ
ence
betw
een
grou
psat
5%sig
nifica
nce
leve
l.
Human Capital Development before Age Five 1425
5.3.1. Home visitingUnlike many social programs, home visiting has been subject to numerous evaluationsusing randomized control trials. A recent survey appears in Howard and Brooks-Gunn(2009). David Olds and collaborators have developed a particular model for home visitingand conducted randomized controlled trials in a number of settings (Olds et al., 1999,
2007) to evaluate it. Olds’ programs focus on families that are at risk because the motheris young, poor, uneducated and/or unmarried, and involve home visits by trained publichealth nurses from the prenatal period up to two years postpartum. The evaluationshave shown many positive effects on maternal behavior, and on child outcomes. As oftwo years of age, children in the Elmira New York were much less likely to have beenseen in a hospital emergency room for unintentional injuries or ingestion of poisonoussubstances, although this finding was not replicated at other study sites. As of age 15,
children of visited mothers were less likely to have been arrested or to have run away fromhome, had fewer sexual partners, and smoked and drank less. The children were also lesslikely to have been involved in verified incidents of child maltreatment. This finding isimportant given the high incidence of maltreatment among US children (and especiallyamong poor children), and the negative outcomes of maltreated children discussed above.
There was little evidence of effects on cognition at four years of age (except amongchildren of initially heavy smokers), though one might expect the documented reductionin delinquent behavior among the teens to be associated with improvements in eventualschooling attainment.
In Olds’ model, using nurses as home visitors is viewed as key to getting goodresults. This may be because nurse home visitors are more acceptable to parents thansocial workers or community workers because families may want medical services. Arandomized trial of nurses versus trained paraprofessionals (Olds et al., 2002) suggests thatthe effects that can be obtained by paraprofessionals are smaller. Also, the Olds programsare strongly targeted at families considered to be at risk and so they do not shed lighton the cost-effectiveness of the universal home visiting programs for pregnant womenand/or newborns that exist in many countries.
Olds’ positive results do not imply that all home visiting programs are equallyeffective. In fact, Table 9 suggests that the average home visiting program has relativelysmall effects. They often improve parenting in subtle ways and may result in someimprovements in specific health outcomes. However, these may not be sufficient tojustify the cost of a large scale program (Aos et al. (2004) offers a cost benefit analysisof several programs). Home visiting programs can be viewed as a type of parentingprogram—presumably the reason why Olds’ home visitors improved outcomes is becausethey taught mothers to be better parents. Since parents are so important to children,
programs that seek to improve parenting practices are perennially popular. Yet studies ofthese programs suggest that it is remarkably difficult to change parents’ behavior and thatmany attempted interventions are unsuccessful. The most successful parenting programs
1426 Douglas Almond and Janet Currie
are those that combine parent education with some other intervention that parents want,such as visits by nurses (as in Olds case) or child care (Brooks-Gunn and Markham, 2005).
5.3.2. US supplemental feeding program for women, infants, and children (WIC)A second type of early intervention program that has been extensively studied is theUS Supplemental Feeding Program for Women, Infants, and Children (WIC). As itsname implies, WIC is a program targeted at pregnant and lactating women, infants, andchildren up to age 5. Participants receive vouchers that can be redeemed for particulartypes of food at participating retailers. Participants must generally go to the WIC officeto receive the vouchers, and generally receive nutrition education services at that time.Many WIC offices are run out of clinics and may also facilitate access to medicalcare. Dozens of studies (many of them reviewed in Currie (2003)) have shown thatparticipation in WIC during pregnancy is associated with longer gestations, higher birthweights, and generally healthier infants, and that the effects tend to be largest for childrenborn to the most disadvantaged mothers. Economists have critiqued these studies, onthe grounds that there may be unobservable variables that are correlated with WICparticipation among eligibles and also with better birth outcomes. Moreover, it may beimplausible to expect WIC to have an effect on pre-term birth. A recent Institute ofMedicine report on the subject reviewed the evidence and concluded that randomizedtrials of many different interventions with women at risk of pre-term birth had failedto find effects (Behrman and Butler, 2007). So it might be surprising to find an effectfor WIC, when more specific and intensive interventions aimed at preventing pre-termbirth have generally failed.
A number of new studies have attempted to deal with various aspects of this critique,as shown in Table 10. Bitler and Currie (2005) look at data from the Pregnancy RiskMonitoring System, which contains very detailed data from new mothers obtainedby combining data from birth records and survey data taken from women before andafter pregnancy. They directly address the question of selection bias by examining thepopulation of mothers eligible for Medicaid (all of whom are adjunctively eligiblefor WIC) and asking how participants differ from non-participants along observabledimensions. They find that the WIC women are more disadvantaged than the non-participants along all observables. This finding does not prove that WIC women are alsonegatively selected in terms of unobservable variables, but it does mean that women whowere very negatively selected in terms of education, health, family relationships and soon would have to have other attributes that were systematically correlated with positiveoutcomes. Like previous studies, Bitler and Currie also find that WIC participation isassociated with higher maternal weight gain, longer gestation, and higher birth weight,particularly among women on public assistance, high school dropouts, teen mothers, andsingle mothers.
Joyce et al. (2004) adopt a similar strategy with regard to selection, and focus on asample of first births to women who initiated prenatal care in the first four months of
Human Capital Development before Age Five 1427
Table10
Selected
stud
iesof
specialsup
plem
entalfee
ding
prog
ram
forw
omen
,infan
ts,and
child
ren(W
IC).
Stud
yStud
yde
sign
Results
Impa
cton
birthou
tcom
es
Bitl
eran
dC
urri
e(2
005)
.D
ata
from
Preg
nanc
yR
iskA
sses
smen
tM
onito
ring
Syst
em(P
RA
MS)
,(n=
60,73
1).
Com
pare
dW
ICpa
rtic
ipan
tsan
dno
n-pa
rtic
ipan
tsin
the
sam
ple
ofw
omen
who
sede
liver
iesw
ere
paid
byM
edic
aid.
Add
ress
edse
lect
ion
bias
byco
mpa
ring
abr
oad
rang
eof
obse
rvab
lech
arac
teri
stic
sbet
wee
nel
igib
leW
ICpa
rtic
ipan
tsan
dno
n-pa
rtic
ipan
ts.
WIC
mot
hers
are
nega
tivel
yse
lect
edin
toth
epr
ogra
mre
lativ
eto
allM
edic
aid
reci
pien
ts.W
ICpa
rtic
ipan
tsar
e1.
4-1.
5tim
esm
ore
likel
yto
have
had
pren
atal
care
in1s
ttri
mes
ter;
0.7
times
aslik
ely
togi
vebi
rth
tolo
wbi
rth
wei
ghti
nfan
t;0.
9tim
esas
likel
yto
give
birt
hto
infa
nts
who
are
belo
wth
e25
thpe
rcen
tile
ofbi
rth
wei
ghtg
iven
gest
atio
nala
ge;0.
9tim
esas
likel
yto
have
thei
rin
fant
adm
itted
toth
eIn
tens
ive
Car
eU
nit.
WIC
asso
ciat
edw
ithin
crea
sesi
nm
ater
nalw
eigh
tgai
n,ge
stat
ion,
and
birt
hw
eigh
t.La
rger
impa
ctfo
rm
ore
disa
dvan
tage
dw
omen
(suc
has
thos
ew
hore
ceiv
edpu
blic
assis
tanc
e,hi
ghsc
hool
drop
-out
s,te
enm
othe
rs,sin
gle
mot
hers
).
(con
tinue
don
next
page
)
1428 Douglas Almond and Janet Currie
Table10
(con
tinue
d)Stud
yStud
yde
sign
Results
Figl
ioet
al.(2
009)
.M
atch
edda
taon
birt
hsin
Flor
ida
duri
ng19
97-2
001
with
scho
olre
cord
sofo
lder
siblin
gsof
the
infa
ntst
oid
entif
yw
heth
erth
eol
der
child
rece
ives
free
lunc
hor
redu
ced-
pric
elu
nch
thou
ghth
eN
SLP
(tho
seon
NSL
Par
eW
ICel
igib
le).
Mar
gina
llyel
igib
le:
n=
2530
;M
argi
nally
inel
igib
le:
n=
1744
.
Inst
rum
enta
lvar
iabl
esco
mpa
ring
mar
gina
llyel
igib
lean
din
elig
ible
WIC
wom
en.M
argi
nally
inel
igib
lefo
rW
IC=
fam
ilies
notp
artic
ipat
ing
inN
LSP
duri
ngpr
egna
ncy,
butp
artic
ipat
ing
duri
ngei
ther
the
year
befo
reor
afte
r.M
argi
nally
elig
ible=
fam
ilies
that
rece
ived
NSL
Pdu
ring
the
preg
nanc
y,bu
tdid
notr
ecei
veth
elu
nche
satl
east
one
adja
cent
year
.A
fede
ralp
olic
ych
ange
that
incr
ease
din
com
ere
port
ing
requ
irem
ents
for
WIC
elig
ibili
tyin
Sept
embe
r19
99,m
ade
itm
ore
diffi
cult
for
elig
ible
sto
obta
inW
IC.T
hein
stru
men
tfor
WIC
part
icip
atio
nis
anin
tera
ctio
nbe
twee
nan
indi
cato
rfo
rth
epo
licy
chan
gean
del
igib
ility
.
WIC
part
icip
atio
nre
duce
sthe
likel
ihoo
dof
low
birt
hw
eigh
tby
12.9
pp(im
prec
isees
timat
e).N
ost
atist
ical
lysig
nifica
nteff
ecto
fWIC
onge
stat
iona
lage
orlik
elih
ood
ofpr
emat
urity
.
Gue
orgu
ieva
etal
.(2
009)
.D
ata
onm
othe
r-in
fant
pair
sfro
mbi
rth
file
son
alls
ingl
eton
birt
hsin
Flor
ida
hosp
itals
betw
een
1996
and
2004
.M
erge
dw
ithM
edic
aid
elig
ibili
tyan
dW
ICpa
rtic
ipat
ion
data
from
the
Flor
ida
Dep
artm
ento
fHea
lth.
(n=
369,
535)
.
Adj
uste
dfo
rse
lect
ion
usin
gpr
open
sity
scor
es.
“Tre
atm
ent”
ispe
rcen
tday
son
WIC
duri
ngpr
egna
ncy.
Mai
nou
tcom
eis
SGA
(“sm
allf
orge
stat
iona
lage
”).S
epar
ate
anal
yses
for
full-
term
,lat
epr
e-te
rm,ve
rypr
e-te
rm,an
dex
trem
ely
pre-
term
birt
hs.
A10
%in
crea
sein
perc
ento
ftim
edu
ring
preg
nanc
yin
WIC
asso
ciat
edw
itha
2.5%
decr
ease
inpr
obab
ility
ofa
full-
term
and
SGA
infa
nt;2.
0%de
crea
sein
prob
abili
tyof
ala
tepr
e-te
rman
dSG
Ain
fant
;3.
7%de
crea
sein
prob
abili
tyof
ave
rypr
e-te
rman
dSG
Ain
fant
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1429
Table10
(con
tinue
d)Stud
yStud
yde
sign
Results
(Joy
ceet
al.,
2004
).D
ata
from
birt
hce
rtifi
cate
file
sin
New
York
City
betw
een
1988
and
2001
onw
omen
who
wer
eon
Med
icai
dan
d/or
WIC
duri
ngpr
egna
ncy.
(n=
35,4
15in
1988
-199
0;n=
50,6
59in
1994
-199
6;n=
52,6
08in
1999
-200
1).
Mul
tivar
iate
regr
essio
nw
ithdu
mm
iesf
orW
ICpa
rtic
ipat
ion,
inte
ract
edw
ithye
arof
birt
h.To
addr
esss
elec
tion
bias
,lim
ited
anal
ysis
tow
omen
with
firs
tbir
thsw
hoin
itiat
edpr
enat
alca
rein
firs
t4m
onth
sofp
regn
ancy
.E
stim
ated
sepa
rate
mod
elsb
yra
ce,et
hnic
ity,na
tivity
,an
dpa
rity
.E
stim
ated
sam
em
odel
com
pari
ngtw
inbi
rths
.
Am
ong
US-
born
blac
ks,W
ICpa
rtic
ipan
tsar
e2.
4pp
less
likel
yto
expe
rien
cea
low
birt
hw
eigh
tbir
thth
anno
n-W
ICpa
rtic
ipan
tsin
1988
-199
1.N
ost
atist
ical
lysig
nifica
nteff
ects
ofW
ICpa
rtic
ipat
ion
for
fore
ign-
born
Hisp
anic
wom
en.In
twin
anal
ysis,
for
US-
born
blac
ks,W
ICpa
rtic
ipat
ion
isas
soci
ated
with
a55
gin
crea
sein
birt
hw
eigh
tad
just
edfo
rge
stat
ion,
and
a3.
9pp
decr
ease
inSG
Ara
tes.
Effec
tsbi
gges
tfor
US-
born
blac
kw
omen
unde
rag
e25
.(N
ote,
mea
nsfo
rsu
bgro
upsn
otre
port
ed,so
effec
tsiz
esca
n’tb
eca
lcul
ated
).
Joyc
eet
al.(2
007)
.Pr
egna
ncy
Nut
ritio
nSu
rvei
llanc
eSy
stem
(PN
SS)d
ata
(n=
2,87
0,03
1).
Incl
uded
allw
omen
who
wer
een
rolle
din
WIC
duri
ngpr
egna
ncy
and
re-e
nrol
led
post
part
um.
Com
pari
son
grou
pis
wom
enw
hoen
rolle
din
WIC
afte
rde
liver
ybu
tw
ere
note
xpos
edto
WIC
duri
ngpr
egna
ncy.
Mul
tivar
iate
regr
essio
nw
ithdu
mm
iesf
orW
ICin
each
trim
este
r.To
deal
with
sele
ctio
nbi
as,es
timat
edse
para
tem
odel
sby
race
/eth
nici
ty.A
lsoan
alyz
edsu
bgro
upsw
hose
pre-
preg
nanc
ych
arac
teri
stic
sput
wom
enat
high
risk
for
anem
ia,lo
ww
eigh
tgai
n,an
din
trau
teri
negr
owth
reta
rdat
ion.
Fina
lly,
anal
yzed
subg
roup
offirs
t-bi
rths
.
Con
ditio
nalo
nge
stat
iona
lage
,m
ean
birt
hw
eigh
tis4
0g
grea
ter
amon
gpr
enat
alen
rolle
esth
anpo
stpa
rtum
enro
llees
.R
ates
ofSG
Aar
e1.
7pp
(14%
)le
ss,an
dra
teso
fter
mlo
wbi
rth
wei
ght
are
0.7
pp(3
0%)l
ess.
Diff
eren
cebe
twee
n1s
tand
3rd
trim
este
ren
rolle
esin
mea
nbi
rth
wei
ghti
s13.
5g.
(con
tinue
don
next
page
)
1430 Douglas Almond and Janet Currie
Table10
(con
tinue
d)Stud
yStud
yde
sign
Results
Kow
ales
ki-J
ones
and
Dun
can
(200
2).
NLS
YM
othe
r-C
hild
data
.20
00ch
ildre
n19
90-9
6.10
4sib
ling
pair
s,71
pair
sin
whi
chon
ech
ildpa
rtic
ipat
edan
don
edi
dn’t.
Sibl
ing
fixe
deff
ects
.In
crea
seof
7ou
nces
inm
ean
birt
hwei
ght.
Posit
ive
effec
ton
tem
pera
men
tsco
re.N
oeff
ecto
nso
cial
orm
otor
skill
stes
tsco
res.
(Hoy
nese
tal.,
2009
).D
ata
onco
unty
-lev
elW
ICav
aila
bilit
yin
1971
-197
5an
d19
78-1
982
from
lists
oflo
cala
genc
iest
hatp
rovi
ded
WIC
serv
ices
.D
ata
onbi
rth
wei
ghtf
rom
vita
lsta
tistic
srec
ords
.D
ata
onpo
pula
tion
ofw
omen
byco
unty
-yea
rfr
omth
eC
AN
CE
R-S
EE
Rpo
pula
tion
data
.D
ata
onva
riou
scon
trol
vari
able
sfr
om19
70IP
UM
S.(n=
18,5
17co
unty
-yea
rce
lls)
Diff
eren
ce-i
n-di
ffer
ence
,us
ing
vari
atio
nin
the
timin
gof
WIC
impl
emen
tatio
nac
ross
differ
ent
coun
tiesi
n19
72-1
979,
toes
timat
eim
pact
ofW
ICav
aila
bilit
yon
birt
hw
eigh
t.C
ontr
olle
dfo
rth
ree
mea
sure
sofp
erca
pita
gove
rnm
entt
rans
fers
,an
indi
cato
rfo
rFo
odSt
amp
prog
ram
avai
labi
lity,
othe
rde
mog
raph
ics,
and
coun
tyan
dye
arfixe
deff
ects
,an
dst
ate-
year
fixe
deff
ects
.Sc
aled
estim
ates
byW
ICpa
rtic
ipat
ion
rate
s.A
lsoco
nduc
ted
sub-
grou
pan
alys
isin
coun
ty-y
ear-
mat
erna
ledu
catio
nle
vel
cells
.
IfW
ICis
avai
labl
ein
aco
unty
byth
eth
ird
trim
este
r,av
erag
ebi
rth
wei
ght
incr
ease
sby
0.1%
(est
imat
eno
tsca
led
bypa
rtic
ipat
ion
rate
).A
mon
gw
omen
with
low
leve
lsof
educ
atio
n,W
ICin
crea
ses
aver
age
birt
hw
eigh
tby
10%
(est
imat
esc
aled
bypa
rtic
ipat
ion
rate
)and
redu
ces
the
frac
tion
ofbi
rths
clas
sified
aslo
wbi
rth
wei
ghtb
y11
%(e
stim
ate
scal
edby
part
icip
atio
nra
te).
No
effec
tsof
WIC
onfe
rtili
ty—
sore
sults
notd
rive
nby
sele
ctio
nin
tobi
rth.
Impa
cton
brea
stfeed
ingan
dinfant
feed
ingpractices
Cha
tter
jiet
al.(2
002)
.N
LSY
Mot
her-
Chi
ldfile
.12
82ch
ildre
nbo
rn19
91-9
5.97
0sib
lings
born
1989
-95.
IVw
ithW
ICst
ate
prog
ram
char
acte
rist
icsa
sin
stru
men
ts.Si
blin
gfixe
deff
ects
.O
LSan
dIV
indi
cate
WIC
redu
ces
brea
stfe
edin
gin
itiat
ion,
butn
oeff
ecto
ndu
ratio
n.Fi
xed
effec
tsug
gest
sre
duct
ions
inle
ngth
brea
stfe
edin
g.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1431
Table10
(con
tinue
d)Stud
yStud
yde
sign
Results
Impa
cton
nutritionan
dhe
alth
outcom
esof
child
ren
Bla
cket
al.(2
004)
Dat
afr
omsu
rvey
sad
min
ister
edby
The
Chi
ldre
n’s
Sent
inel
Nut
ritio
nalA
sses
smen
tPr
ogra
min
am
ulti-
site
stud
yat
urba
nm
edic
alce
nter
sin
5st
ates
and
Was
hing
ton,
DC
in19
98-2
001
(n=
5923
:53
95re
ceiv
edW
ICas
sista
nce,
528
did
not)
.
Com
pare
dW
IC-e
ligib
lefa
mili
esw
hopa
rtic
ipat
edin
WIC
with
thos
ew
hodi
dno
tdue
tose
lf-re
port
edac
cess
prob
lem
s.M
ultiv
aria
tere
gres
sion
incl
udin
ga
part
icip
atio
ndu
mm
y,co
ntro
lling
for
rele
vant
back
grou
ndch
arac
teri
stic
s.
Com
pare
dto
infa
ntsw
hore
ceiv
edW
ICas
sista
nce,
thos
ew
hodi
dno
trec
eive
WIC
assis
tanc
ew
ere
mor
elik
ely
tobe
unde
rwei
ght(
wei
ght-
for-
age
z-sc
ore=−
0.23
vs.0.
009)
,sh
ort
(leng
th-f
or-a
gez-
scor
e=−
0.23
vs.
−0.
02),
and
perc
eive
das
havi
ngfa
iror
poor
heal
th(a
djus
ted
odds
ratio=
1.92
).N
ost
atist
ical
lysig
nifica
ntdi
ffer
ence
sin
rate
soff
ood
inse
curi
ty.
Lee
and
Mac
key-
Bila
ver
(200
7).
Dat
afr
omIL
Inte
grat
edD
atab
ase
onC
hild
ren’
sSer
vice
s.In
clud
esin
foon
Med
icai
d,FS
P,W
ICen
rollm
ent.
All
ILch
ildre
nbo
rnbe
twee
n19
90an
d19
96w
hoen
tere
dM
edic
aid
with
infirs
tmon
th.T
rack
edto
2001
.To
tal
sam
ple=
252,
246.
Sibl
ing
FEsa
mpl
e=
36,2
77.
Mul
tivar
iate
regr
essio
nw
ithsib
ling
fixe
d-eff
ects
,us
ing
part
icip
atio
ndu
mm
iesf
orFS
P,W
IC,an
dFS
P-W
ICjo
intly
.(N
ote,
siblin
gFE
only
mak
ese
nse
for
WIC
,eff
ects
ofFS
Pes
timat
edus
ing
OLS
.A
lso,
mos
tfam
ilies
onW
ICal
sore
ceiv
eFS
P).
Effec
tofW
IC:A
buse
orne
glec
trat
esde
crea
seby
84%
(mea
n=
0.10
).In
cide
nce
ofan
emia
decr
ease
by41
%(m
ean=
0.19
5).Fa
ilure
toth
rive
decr
ease
by78
%(m
ean=
0.12
8).
Nut
ritio
nald
efici
ency
decr
ease
by11
5%(m
ean=
0.03
8).
Unl
esso
ther
wise
note
d,on
lyre
sults
signi
fica
ntat
5%le
vela
rere
port
edhe
re.
Perc
entc
hang
es(d
enot
edby
%in
stea
dof
“pp”
)are
repo
rted
rela
tive
toth
em
ean.
1432 Douglas Almond and Janet Currie
pregnancy in order to ensure that participants and non-participants were more likely tobe similar in terms of unobservables. In their sample of women giving birth in NewYork City, they find positive effects of WIC among US born black women, but notin other groups. Joyce et al. (2007) use a national sample of women, compare womenwho enrolled in WIC pre and post delivery, and focus on whether infants are smallfor gestational age (SGA). If one does not believe that WIC can affect gestation, thenfocusing on SGA is appropriate because it is not affected by gestational age. They findthat the incidence of SGA is lower for the prenatal enrollees than for the postpartumenrollees. Gueorguieva et al. (2009) use a large sample of births from Florida and tryto deal with potential selection using propensity score matching. They side step theissue of whether WIC affects gestation by presenting separate analyses for pregnanciesof different length, and focusing on SGA. They find that longer participation in WIC isassociated with reductions in the incidence of SGA. Kowaleski-Jones and Duncan (2002)examine sibling pairs from the NLSY and find that WIC participation is associated withan increase of seven ounces in birth weight. However, the number of pairs in which onechild participated and one did not is quite small, so it would be useful to try to replicatethis finding in a larger sample of siblings.
Figlio et al. (2009) present an innovative instrumental variables strategy using alarge sample of births from Florida that have been merged to school records of oldersiblings. While the characteristics of WIC programs vary across states, they do notshow a lot of variation over time, and previous analyses have demonstrated that thesecharacteristics are weak instruments (Bitler and Currie, 2005). Figlio et al. (2009) firsttry to select participant and non-participant groups who are very similar. They do this bydefining “marginally ineligible” families as those who participated in the National SchoolLunch Program (NLSP) in the year before or after the birth, but did not participatein the birth year. Thus, the study focuses on families whose incomes hover around theeligibility threshold for NSLP, which is the same as the eligibility threshold for WIC. Theinstrument is a change in income reporting requirements for WIC in Sept. 1999 whichmade it more difficult for eligible families to receive benefits. Figlio et al. (2009) find thatWIC participation reduces the probability of low birth weight, but find no significanteffect on gestational age or prematurity.
There has been much less study of the effects of WIC on other outcomes, orother groups of participants. A couple of studies that make some attempt to deal withthe selection issue are summarized in Table 10. One problem with WIC is that itsubsidizes baby formula, which is likely to discourage breast-feeding. Chatterji andBrooks-Gunn (2004) use the NLSY Mother-Child file and estimate both sibling fixedeffects models and instrumental variables models using characteristics of state programsas instruments. They find that WIC reduces breast feeding initiation and the length ofbreastfeeding. However, these results are subject to the caveats above (i.e. small samplesand possibly weak instruments). Turning to the effects of WIC on older children, Black
Human Capital Development before Age Five 1433
et al. (2004) compare WIC eligible participants and those who did not participate dueto “access problems”. These problems were assessed based on the families own reportsabout why they were not participating. They found that infants who received WIC wereless likely to be underweight, short, or perceived by their parents to be in fair or poorhealth. Lee and Mackey-Bilaver (2007) use a large data base from Illinois that integratesadministrative data from several sources. Using sibling fixed effects models, they find thatsiblings who received WIC were less likely to be anemic, to have exhibited failure-to-thrive, or other nutritional deficiencies, and that the infants were less likely to be abusedor neglected. As discussed above, one issue in the interpretation of these findings is whyone infant would receive WIC while the other did not?
In one of the most interesting recent studies, Hoynes et al. (2009) use the initialroll-out of the WIC program in the 1970s to identify its effects. They find that theimplementation of WIC increased average birth weight by 10% and decreased thefraction of low birth weight births. They did not find any evidence of changes in fertility.
In summary, the latest group of studies of WIC during pregnancy largely support thefindings of earlier studies which consistently found beneficial effects on infant health. Thefinding is remarkable because WIC benefits are relatively modest (often amounting toabout $40 per month) and Americans are generally well fed (if not overfed at least in termsof total calories). Research that attempted to peer into the “black box” and shed light onwhy the program is effective would be extremely interesting. Another question that criesout for future research is whether WIC benefits infants and children (i.e. children whoparticipate after birth)? While a few studies suggest that it does, the effects of WIC in thispopulation has been subject to much less scrutiny than the effects on newborns.
5.3.3. Child careThere have been many evaluations of early intervention programs delivered throughthe provision of child care. One reason for focusing on early intervention through theprovision of quality child care is that the majority of young children are likely to beplaced in some form of care. In 2008, 64% of women with children under 3 workedfor pay (US Bureau of Labor Statistics, 2009). While the US may be an outlier in thisrespect, labor force participation among women with children is high and rising inmany other economies. Blau and Currie (2006) provide an overview of the literature onearly intervention through child care. Many studies concern experimental evaluationsof model programs that serve relatively small numbers of children and involve intensiveservices delivered by well-trained and well-supervised staff. These studies generallyfind that early intervention has long-lasting effects on schooling attainment and otheroutcomes such as teen pregnancy and crime, even if it does not result in any lastingincrease in cognitive test scores. These results point to the tremendous importanceof “non-cognitive skills” (cf Heckman and Rubinstein (2001)) or alternatively, tothe importance of mental as well as physical health in the production of good childoutcomes (Currie and Stabile, 2006).
1434 Douglas Almond and Janet Currie
A few of the most notable model programs are summarized in Table 11. Two studies of“model” early intervention child care programs stand out because they randomly assignedchildren to treatment and control group, had low dropout rates, and followed childrenover many years. They are the Carolina Abecedarian Project and the Perry PreschoolProject. Both found positive effects on schooling. A recent cost-benefit analysis of theAbcedarian data through age 21 found that each dollar spent on Abecedarian saved taxpayers four dollars. And by focusing only on cost savings, this calculation does not eveninclude the value of higher achievement to the individual children and society (Masse andBarnett, 2002). Each dollar spent on Perry Preschool has been estimated to have saved upto seven dollars in social costs (Karoly et al., 1998), although this high benefit-cost ratiois driven largely by the effect of the intervention on crime, which in turn depends on ahandful of individuals.
Anderson (2008) conducts a re-analysis of the Perry Preschool and Abcedarian data(and a third intervention called the Early Training Project) and finds that like theMTO public housing experiment, the significant effects of the intervention were largelyconcentrated among girls. In addition to analyzing the data by gender, Anderson payscareful attention to the idea that there may be a reporting bias in the published studies ofearly intervention experiments; that is, researchers who found largely null effects of theexperiment might still be able to publish results focusing on one or two positive outcomesout of many outcomes investigated. Conversely, if all effects tended in the same direction,
but there was insufficient power to detect significant effects on each outcome, it mightbe possible to detect a significant effect on an index of the outcomes. Anderson findspositive effects (for girls) on a summary index of effects, and the effects are quite large atabout a half a standard deviation. This study highlights an interesting question, which iswhether it is generally easier to intervene with girls than with boys, and why that mightbe the case?
The fact that special interventions like Perry Preschool or Abcedarian had an effecton at least some target children does not prove that the types of programs typicallyavailable to poor inner-city children will do so. Head Start is a preschool program fordisadvantaged 3, 4, and 5 year olds which currently serves about 800,000 children eachyear. It is funded as a federal-local matching grant program and over time, federal fundinghas increased from $96 million when the program began in 1965 to about $7 billion in2009 (plus additional “stimulus” funds). Head Start is not of the same quality as the modelinterventions, and the quality varies from center to center. But Head Start centers havehistorically been of higher average quality than other preschool programs available to lowincome people. This is because, in contrast to the private child care market, there are fewvery low-quality Head Start programs (see Blau and Currie (2006) for an overview ofpreschool quality issues).
An experimental evaluation of Head Start has recently been conducted (Puma et al.,2010). The evaluation compares Head Start children to peers who may or may not be in
Human Capital Development before Age Five 1435
Table11
Selected
recent
evalua
tions
ofearly
child
hood
prog
ramswith
rand
omized
design
s.Stud
y/prog
ram
namea
Data,prog
ram
description,
andstud
yde
sign
Results
Car
olin
aA
bece
dari
anfo
llow
-up
and
cost
-ben
efit
anal
ysis
at21
year
sofa
ge(B
arne
ttan
dM
asse
,20
07)
Pres
choo
lers
:fu
ll-da
ych
ildca
reSc
hool
age:
pare
ntpr
ogra
mSa
mpl
esiz
es:
Initi
al:T=
57,C=
54A
ge8:
T=
48,C=
42A
ge15
:T=
48,C=
44A
ge21
:T=
53,C=
51A
geof
part
icip
atio
nin
prog
ram
:E
ntry
:6
wee
ksto
3m
onth
sold
Exi
t:5
to8
year
s
Follo
w-u
pat
21ye
arsofag
e:G
rade
rete
ntio
n:T=
34%
,C=
65%
,ag
e21
Spec
iale
duca
tion:
T=
31%
,C=
49%
,ag
e21
Hig
hsc
hool
drop
out:
T=
33%
,C=
49%
,ag
e21
Col
lege
atte
ndan
ce:T=
36%
,C=
13%
,ag
e21
Cri
me
rate
:T=
C,ag
e21
Em
ploy
men
tsta
tus:
T=
Cat
age
21A
vera
geag
efirs
tchi
ldbo
rn:T>
Cat
age
21C
ost
-Ben
efitA
nal
ysis
:(u
sing
5%di
scou
ntra
te,$2
002)
Net
cost
per
child=
$34,
599
Net
bene
fito
fpro
gram=
$72,
591
per
part
icip
ant
Infa
ntH
ealth
and
Dev
elop
men
tPro
ject
Follo
w-u
pat
18ye
arso
fag
e(M
cCor
mic
ket
al.,
2006
)
Hom
evi
sits,
full-
day
child
care
Sam
ple
sizes
:In
itial
:T=
377,
C=
608
Follo
wup
atag
e8:
T=
336,
C=
538
Follo
wup
atag
e18
:T=
254,
C=
381
(div
ided
in2
grou
ps:lig
hter
low
birt
hw
eigh
t(LL
BW
)and
heav
ier
low
birt
hw
eigh
t(H
LBW
))A
geof
part
icip
atio
nin
prog
ram
:E
ntry
:bi
rth
(hom
evi
sits)
,1
year
(car
e).
Exi
t:3
year
s
Mat
hac
hiev
emen
t:T>
Cby
6.8%
,ag
e18
HLB
WR
eadi
ngac
hiev
emen
t:T>
Cby
5.6%
,ag
e18
HLB
WR
isky
beha
vior
s:T>
Cby
23.3
%,ag
e18
HLB
WIQ
:T=
C,ag
e18
HLB
WN
ote:
For
allo
utco
mes
:T=
C,ag
e18
LLB
W
(con
tinue
don
next
page
)
1436 Douglas Almond and Janet Currie
Table11
(con
tinue
d)Stud
y/prog
ram
namea
Data,prog
ram
description,
andstud
yde
sign
Results
Are
eval
uatio
nof
earl
ych
ildho
odin
terv
entio
n—A
bece
dari
an,Pe
rry
Pres
choo
land
Ear
lyT
rain
ing
Proj
ect—
with
emph
asis
onge
nder
differ
ence
sand
mul
tiple
infe
renc
e(A
nder
son,
2008
)
Abe
ceda
rian
:T=
57,C=
54Pe
rry:
T=
58,C=
65E
TP:
T=
44,C=
21A
geso
fent
ry:
Abe
ceda
rian
/Per
ry/E
TP:
4.4
mo.
/3yr
s./3-
4yr
s
Out
com
esin
clud
e:IQ
,gr
ade
repe
titio
n,sp
ecia
led.
,hi
ghsc
hool
,co
llege
atte
ndan
ce,em
ploy
men
t,ea
rnin
gs,re
ceip
ttr
ansf
ers,
arre
sts,
conv
ictio
ns,dr
ugus
e,te
enpr
egna
ncy,
mar
riag
e.Su
mm
ary
inde
xpo
olsm
ultip
leou
tcom
esfo
ra
singl
ete
st.Se
para
tete
stsb
yge
nder
.E
ffec
tson
sum
mar
yin
dex
for
girl
s5-1
2:A
BC
/Per
ry:in
crea
seby
0.45/0.
54SD
s.E
ffec
tson
sum
mar
yin
dex
for
girl
s13-
19:
AB
C/P
erry
/ET
P:in
crea
seby
0.42
/0.6
1/0.
46SD
s.E
ffec
tson
sum
mar
yin
dex
for
wom
enov
er21
-40:
AB
C/P
erry
:in
crea
seby
0.45/0.
36SD
s.N
ost
atist
ical
lysig
nifica
nteff
ects
onm
ales
ofan
yag
e.
Hig
h/Sc
ope
Perr
yPr
esch
oolp
roje
ctfo
llow
-up
and
cost
-ben
efit
anal
ysis
at40
year
sofa
ge(B
arne
ttet
al.,
2006
)
Hom
evi
sits,
Pres
choo
lpro
gram
Sam
ple
sizes
:In
itial
:T=
58,C=
65A
ge40
:T=
56,C=
63A
geof
part
icip
atio
nin
prog
ram
:E
ntry
:3
to4
year
s,E
xit:
5ye
ars
Arr
ests
:T=
32%
,C=
48%
;In
jail:
T=
6%,C=
17%
;R
epor
tofs
topp
ing
wor
kfo
rhe
alth
reas
ons:
T=
43%
,C=
55%
;H
ard
drug
use:
T=
22%
,C=
29%
;A
bort
ions
:T=
17%
,C=
32%
Cost
-ben
efitan
alys
is:
Mai
nre
sult:
Ben
efito
f$12
.90
for
each
$1co
stM
ostb
enefi
tsdu
eto
redu
ced
crim
era
tesf
orm
ales
Cos
t:$1
5,82
7($
2000
)per
stud
ent
Tota
lNet
Priv
ate
Ben
efit=
$17,
730
per
part
icip
ant
Tota
lNet
Publ
icB
enefi
t=$1
80,4
55pe
rpa
rtic
ipan
t
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1437
Table11
(con
tinue
d)Stud
y/prog
ram
namea
Data,prog
ram
description,
andstud
yde
sign
Results
Nat
iona
leva
luat
ion
ofE
arly
Hea
dSt
art
(Adm
inist
ratio
non
Chi
ldre
n,Yo
uth
and
Fam
ilies
,20
02&
Love
etal
.,20
05)C
ost-
Ben
efit
(Aos
etal
.,20
04).
17E
arly
Hea
dSt
arts
itess
elec
ted
tore
flec
tE
HS
prog
ram
sfun
ded
in19
95-9
6.R
ando
mas
signm
entw
ithin
site
(pos
sible
give
nw
aitl
ists)
.Sa
mpl
e:T
=15
13,C
=14
88.
Age
ofpa
rtic
ipat
ion
inpr
ogra
m:E
ntry
at0-
1ye
ar,ex
it3
year
s
Men
talD
evel
opm
entI
ndex
(MD
I):T>
Cby
0.12
SDPP
VT-
III
Voca
bula
rysc
ore:
T>
Cby
0.13
SDPe
rcen
tage
with
PPV
Tsc
ore<
85pt
s:T<
Cby
0.12
SDA
ggre
ssiv
ebe
havi
or:T<
Cby
0.11
SDSu
ppor
tiven
essd
urin
gpa
rent
-chi
ldpl
ay:T>
Cby
0.15
SDH
OM
Esc
ore:
T>
Cby
0.11
SDIn
dex
ofse
veri
tyof
disc
iplin
e:T<
Cby
0.11
SDN
ost
atist
ical
lysig
nifica
nteff
ects
onpa
rent
alm
enta
lor
phys
ical
heal
thor
onm
easu
reso
ffam
ilyfu
nctio
ning
.C
ost
-ben
efitan
alys
is:
Tota
lCos
tper
child
:$2
0,97
2To
talB
enefi
tper
child
:$4
768,
NPV
:−
$16,
203
Hea
dSt
arti
mpa
ctst
udy
(US
Dep
artm
ento
fH
ealth
and
Hum
anSe
rvic
es,20
10)
Con
gres
siona
lly-m
anda
ted
stud
yof
Hea
dSt
art.
Chi
ldre
nfr
omw
aitl
istsr
ando
mly
assig
ned
toon
eof
383
rand
omly
sele
cted
Hea
dSt
artc
ente
rsac
ross
23di
ffer
ents
tate
s.B
asel
ine
data
colle
cted
infa
ll20
02;an
nual
spri
ngfo
llow
-ups
thro
ugh
spri
ng20
06.
Sam
ple
Size
s:T=
2783
,C=
1884
.E
ntry
:3-
4ye
arso
ld;E
xit:
4-5
year
sold
Sum
mar
yof
effec
tsfo
r4-
year
-old
entr
yco
hort
aten
dof
1stg
rade
:PP
VT
:T>
Cby
0.09
SD;W
ithdr
awn
beha
vior
:T<
Cby
0.13
SD;Sh
y/so
cial
lyre
ticen
t:T>
Cby
0.19
SD;Pr
oble
msw
ithte
ache
rin
tera
ctio
n:T>
Cby
0.13
SD.
No
stat
istic
ally
signi
fica
ntim
pact
sata
ge4,
kind
erga
rten
,or
1stg
rade
on:m
ath
scor
es,Sp
anish
lang
uage
test
s,or
alco
mpr
ehen
sion,
and
seve
ralp
aren
t-an
dte
ache
r-re
port
edm
easu
reso
fem
otio
nala
ndbe
havi
oral
outc
omes
.N
ost
atist
ical
lysig
nifica
ntim
pact
satk
inde
rgar
ten
or1s
tgra
deon
:sc
hool
acco
mpl
ishm
ents
,pr
omot
ion,
lang
uage
and
liter
acy
abili
ty,m
ath
abili
ty,an
dso
cial
stud
iesa
ndsc
ienc
eab
ility
.Su
mm
ary
ofeff
ects
for
3-ye
ar-o
lden
try
coho
rtas
of1s
tgra
de(s
elec
ted
resu
lts):
Ora
lcom
preh
ensio
n:T>
Cby
0.08
SD.N
osig
nifica
nteff
ecto
not
her
outc
omes
.
(con
tinue
don
next
page
)
1438 Douglas Almond and Janet Currie
Table11
(con
tinue
d)Stud
y/prog
ram
namea
Data,prog
ram
description,
andstud
yde
sign
Results
The
rate
sofr
etur
nto
the
Hig
h/Sc
ope
Perr
yPr
esch
oolP
rogr
am(H
eckm
anet
al.,
2009
)
See
Bar
nett
etal
.(2
006)
entr
yfo
rsa
mpl
esiz
esan
din
form
atio
nab
outt
hePe
rry
Pres
choo
lPro
gram
.R
ando
miz
atio
nw
asco
mpr
omise
dbe
caus
eof
reas
signm
enta
fter
initi
alra
ndom
izat
ion.
Stan
dard
erro
rson
the
rate
ofre
turn
estim
ates
adju
sted
for
failu
reof
rand
omiz
atio
nus
ing
boot
stra
ppin
gan
dM
onte
Car
lom
etho
ds.
Also
,adj
uste
dfo
rthe
dead
wei
ghtl
ossd
ueto
taxa
tion
(ass
umin
g0%
,50
%,an
d10
0%de
adw
eigh
tlos
ses)
.U
sed
aw
ide
vari
ety
ofm
etho
dsfo
rw
ithin
-sam
ple
impu
tatio
nof
miss
ing
earn
ings
.U
sed
loca
ldat
aon
cost
sof
educ
atio
n,w
elfa
repa
rtic
ipat
ion,
and
crim
ein
stea
dof
natio
nald
ata,
whe
reve
rpo
ssib
le.
Use
dse
vera
lmet
hods
toex
trap
olat
ebe
nefits
and
cost
sbey
ond
age
40(a
fter
last
follo
w-u
p).U
sed
seve
ralm
easu
reso
fthe
stat
istic
alco
stof
life
toes
timat
eco
stso
fm
urde
r.
Est
imat
edso
cial
rate
sofr
etur
nto
the
Perr
yPr
esch
ool
Prog
ram
are
7%-1
0%.E
stim
ated
bene
fit-
cost
ratio
=2.
2to
31.5
(dep
ends
ondi
scou
ntra
teus
ed,an
dth
em
easu
reof
cost
ofm
urde
r).
Thr
ough
outt
heta
ble,
“T”
refe
rsto
trea
tmen
tgro
upan
d“C
”re
fers
toco
ntro
lor
com
pari
son
grou
p.O
utco
mes
liste
das
T>
Cor
C>
Tw
ere
stat
istic
ally
signi
fica
ntat
the
5%le
vel.
aPr
ogra
msa
regr
oupe
dsu
chth
atth
ose
enro
lling
child
ren
youn
ger
than
thre
eye
arso
ldap
pear
firs
t,fo
llow
edby
thos
een
rolli
ngch
ildre
naf
ter
age
thre
e.
Human Capital Development before Age Five 1439
some other form of preschool (including state-funded preschools modeled in Head Start).In fact, the majority of children who did not attend Head Start did end up attending someother preschool program. Even relative to this baseline, initial results show that Head Startchildren make gains, particularly in terms of language ability. But children are followedonly into the first grade, and so this evaluation did not address the important issue ofwhether Head Start has longer term effects. This example illustrates one of the limitationsof experiments for the study of longer-term effects, which is that one may have to waita long time for evidence to accumulate. There has also been a federal evaluation of EarlyHead Start (EHS), a version of the program geared to infants and toddlers under threeyears old. As Table 11 shows, EHS has small positive effects on cognitive test scores andsome measures of behavior though Aos et al. (2004) conclude that it does not pass acost-benefit test.
Table 12 summarizes notable non-experimental evaluations of Head Start and otherpublic preschool programs. In a series of studies Currie and Thomas use nationalpublicly-available survey data to try to measure the effect of Head Start. In most ofthese studies, they compare the outcomes of children who attended Head Start tosiblings who did not attend. As discussed above, sibling fixed effects control for manyshared characteristics of children, but are not a panacea. However, careful examination ofdifferences between participant and non-participant children within families suggestedthat the Head Start sibling typically attends when the family is relatively disadvantaged.
For example, a young single mother might have her first child attend Head Start.If she then marries, her next child will enjoy higher income and be ineligible forHead Start. Currie and Thomas found no within-family differences in birth weight orother individual characteristics of the children. They also investigated spillover effects,which as discussed above, can bias the estimated effect of Head Start. They foundsome evidence (Garces et al., 2002) that having an older sibling attend Head Start hadpositive effects on younger siblings. In all, it seems likely that sibling fixed effects modelsunderstate the true effect of Head Start.
Nevertheless, they found significant positive effects of Head Start on educationalattainments among white youths, and reductions in the probabilities of being bookedor charged with a crime among black youths (Garces et al., 2002). Test score gainsfor blacks and whites were initially the same, but these gains tended to fade out morequickly for black than white students, perhaps because black former Head Start studentstypically attend worse schools than other students (Currie and Thomas, 1995). Effectswere especially large for Hispanic children (Garces et al., 2002).
More recently, Deming (2009) replicates the results of Currie and Thomas (1995)using the same cohort of NLSY children observed at older ages. Like Anderson, hefocuses on an index of outcomes (although he also reports results for separate outcomes)and finds that Head Start results in an increase of 0.23 standard deviations, which isequivalent to about 1/3 of the gap between Head Start and other children. He notes
1440 Douglas Almond and Janet Currie
Table12
Selected
stud
iesof
large-scalepu
blicearly
child
hood
prog
rams.
Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Evalua
tion
sof
Hea
dStart
Doe
sHea
dSt
artm
ake
adi
ffer
ence
?D
oesH
ead
Star
the
lpH
ispan
icch
ildre
n?(C
urri
ean
dT
hom
as,19
95,
1999
a).N
LSC
M.
Sam
ple
size:
T=
896,
C=
911
Hisp
anic
stud
y:T=
182,
C=
568
Ent
ry:3
to5
year
s;E
xit:
5to
6ye
ars.
Est
imat
esib
ling
fixe
deff
ects
mod
els
ofeff
ects
ofH
ead
Star
tand
othe
rpr
esch
oola
tten
danc
eon
vari
ous
outc
omes
.E
xam
ine
differ
ence
sbe
twee
nsib
lings
that
mig
htpo
tent
ially
expl
ain
differ
entia
lat
tend
ance
bysib
lings
.
Ach
ieve
men
tte
sts:
T>
C(1/3
SDw
hite
sand
Hisp
anic
sonl
y)G
rade
rete
ntio
n:T<
C(∼
50%
whi
tesa
ndH
ispan
icso
nly)
Imm
uniz
atio
nra
tes:
T>
C(8
%-1
1%)
Chi
ldhe
ight
-for
-age
:T=
C
Long
term
effec
tsof
Hea
dSt
art(
Gar
cese
tal.,
2002
).PS
ID,Sa
mpl
esiz
e:T=
583,
C=
3502
.E
ntry
:3
to4
year
s;E
xit5
to6
year
s.
Com
pare
dH
ead
Star
tpar
ticip
ants
toth
eir
own
siblin
gsw
hodi
dno
tpa
rtic
ipat
e.O
utco
mes
mea
sure
dbe
twee
nag
es18
and
31.
Ret
rosp
ectiv
ere
port
son
Hea
dSt
art
part
icip
atio
n.
Hig
hsc
hool
grad
uatio
n:T>
C(∼
25%
for
whi
teso
nly)
Arr
ests
T<
C(∼
50%
for
Afr
ican
-Am
eric
anso
nly)
Col
lege
T>
C(∼
25%
for
Whi
tes)
Teen
preg
nanc
yT=
CW
elfa
reT=
C
Effec
tofH
ead
Star
ton
heal
than
dsc
hool
ing
(Lud
wig
and
Mill
er,20
05).
Vita
lSta
tistic
sCom
pres
sed
Mor
talit
yFi
les1
973-
83;
Indi
vidu
alda
tafr
omN
ELS
,w
here
T=
649,
C=
674.
Reg
ress
ion
disc
ontin
uity
arou
ndcu
toff
atw
hich
coun
tiesw
ere
elig
ible
for
assis
tanc
ein
appl
ying
for
Hea
dSt
arti
n19
65.T=
300
poor
est
coun
tiesi
n19
65,C=
301-
600t
hpo
ores
tcou
ntie
s.80
%of
trea
tmen
tco
untie
srec
eive
dfu
ndin
gvs
.43
%of
allc
ount
iesn
atio
nwid
e.
Effec
tsof
part
icip
atio
nin
Hea
dSt
art:
Mor
talit
y,ag
e5-
9:T<
Cby
35%
-79%
Hig
hsc
hool
com
plet
ion
rate
s:T>
Cby
5.2%
-8.5
%So
me
colle
ge+
:16
.2%
-22.
4%fo
rol
dest
coho
rton
ly
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1441
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Hea
dSt
artP
artic
ipat
ion
and
Chi
ldho
odO
besit
y(F
risv
old,
2006
).PS
IDC
hild
Dev
elop
men
tSup
plem
ent.
Sam
ple
size=
1332
.
Est
imat
edth
eeff
ecto
fHea
dSt
arto
nlik
elih
ood
ofa
child
bein
gov
erw
eigh
tor
obes
e.A
ssum
eth
at#
ofsp
aces
avai
labl
ein
aco
mm
unity
isa
valid
inst
rum
entf
orH
ead
Star
tpa
rtic
ipat
ion.
Hea
dSt
artr
educ
espr
obab
ility
ofob
esity
atag
es5-
10am
ong
blac
ks.N
oeff
ecti
nfu
llsa
mpl
eof
child
ren
orin
child
ren
over
10.E
stim
ates
are
larg
ere
lativ
eto
sam
ple
mea
nsim
plyi
ng∼
100%
redu
ctio
nsin
over
wei
ght/
obes
ity.
Evi
denc
efr
omH
ead
Star
tO
nLi
fecy
cle
Skill
Dev
elop
men
t(D
emin
g,20
09).
Dat
afr
omN
LSY
Chi
ldre
nfo
rco
hort
enro
lled
inH
ead
Star
tbet
wee
n19
84an
d19
90.C
hild
ren
inst
udy
atle
ast5
year
sold
in19
90.
Sam
ple
size:
3415
tota
l.
Sibl
ing
fixe
deff
ects
estim
ates
ofbe
nefits
ofH
ead
Star
t.Te
stsc
ores
:T>
Cby
0.14
5SD
ages
5-6,
by0.
133
SDag
es7-
10,by
0.05
5SD
ages
11-1
4.N
onco
gniti
vesc
hool
-age
outc
omes
inde
x:T>
Cby
0.26
5SD
.Lo
ng-t
erm
effec
ton
inde
x*of
youn
gad
ulto
utco
mes
:T>
Cby
0.22
8SD
.La
rge
fade
-out
inte
stsc
ores
ofA
fric
anA
mer
ican
s,no
nefo
rw
hite
sor
Hisp
anic
s.N
oeff
ects
oncr
imin
alac
tivity
.Sum
mar
y:H
ead
Star
tin
crea
sesi
ndex
oflo
ngte
rmou
tcom
esby
0.23
SD(∼
1/3
ofga
pat
tend
eesa
ndot
hers
).Pr
ojec
ting
wag
esim
plie
stha
tben
efits
(∼$1
500
ingr
eate
rea
rnin
gspe
rye
ar)e
xcee
dpr
ogra
mco
stso
f∼
$600
0.*
inde
xin
clud
es:gr
adua
tehi
ghsc
hool
,co
mpl
ete
1yr
colle
geid
le(n
ojo
b,no
tin
scho
ol),
poor
heal
th.
(con
tinue
don
next
page
)
1442 Douglas Almond and Janet Currie
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Prev
entin
gbe
havi
orpr
oble
msi
nch
ildho
odan
dad
oles
cenc
e:E
vide
nce
from
Hea
dSt
art(
Car
neiro
and
Gin
ja,20
08).
NLS
YC
hild
ren
data
.Sa
mpl
esiz
e=
1786
mal
es.
Beh
avio
rpr
oble
ms,
grad
ere
p.an
dob
esity
at12
-13.
Dep
ress
ion,
crim
e,an
dob
esity
at16
-17.
Old
esti
nda
ta—
born
in19
74;yo
unge
stin
data
—bo
rnin
1992
.
Hea
dSt
arte
ligib
ility
rule
scre
ate
disc
ontin
uitie
sin
inco
me
elig
ibili
ty.
Com
pare
fam
ilies
abov
ean
dbe
low
the
cuto
ff.Id
entifi
catio
nst
rate
gyre
quir
esth
atfa
mili
esar
eno
tabl
eto
stra
tegi
cally
loca
teab
ove
orbe
low
cuto
ff.
USIN
GR
ED
UC
ED
FO
RM
EST
IMAT
ION
Hea
dSta
rtpar
tici
pat
ion
impac
tson
12-1
3ye
ar-o
ldm
ales
:be
havi
oral
prob
lem
sind
exde
crea
sesb
y38
%pr
obab
ility
ofgr
ade
rete
ntio
nde
crea
sesb
y33
.3%
prob
abili
tyof
obes
ityde
crea
sesb
y∼
100%
for
blac
kson
lyH
ead
Sta
rtpar
tici
pat
ion
impac
tson
16-1
7ye
ar-o
ldm
ales
:D
epre
ssio
n(C
ESD
)dec
reas
esby
23.4
%pr
obab
ility
ofob
esity
decr
ease
sby
57.9
%pr
obab
ility
ofbe
ing
sent
ence
dde
crea
sesb
y>
100%
for
blac
kson
lyU
SIN
GST
RU
CT
UR
AL
EQ
UAT
ION
SH
ead
Sta
rtpar
tici
pat
ion
impac
tson
12-1
3ye
ar-o
ldm
ales
:pr
obab
ility
ofgr
ade
rete
ntio
nde
crea
sesb
y>
100%
Hea
dSta
rtpar
tici
pat
ion
impac
tson
16-1
7ye
ar-o
ldm
ales
:pr
obab
ility
ofbe
ing
sent
ence
dde
crea
sesb
y>
100%
prob
abili
tyof
obes
ityde
crea
sesb
y>
100%
prob
abili
tyof
bein
gse
nten
ced
decr
ease
sby>
100%
for
blac
kson
lypr
obab
ility
ofob
esity
decr
ease
sby>
100%
for
blac
kson
lyN
ote:
base
line
mea
nsar
efo
rsa
mpl
eof
child
ren
with
inco
mes
betw
een
5%an
d19
5%of
Hea
dSt
arte
ligib
ility
cut-
off.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1443
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Inve
stin
gin
heal
th:th
elo
ng-t
erm
impa
ctof
Hea
dSt
arto
nsm
okin
g(A
nder
son
etal
.,20
09).
Dat
afr
omth
ePS
ID.U
sed
smok
ing
data
from
1999
and
2003
onin
divi
dual
sage
d21
-36
in19
99.n=
922
in19
99;
n=
1005
in20
03.
Com
pare
dsm
okin
gof
siblin
gsw
hodi
dan
ddi
dno
tatt
end
Hea
dSt
arto
ran
ypr
esch
oolu
sing
siblin
gfixe
deff
ects
.C
ontr
olle
dfo
rfa
mily
back
grou
ndch
arac
teri
stic
sspe
cific
toth
eag
ech
ildre
nw
ere
elig
ible
for
Hea
dSt
art.
Exa
min
edsib
ling
differ
ence
stha
tmig
htbe
pred
ictiv
eof
Hea
dSt
arta
tten
danc
e.E
xam
ined
spill
over
effec
tsby
incl
udin
gin
tera
ctio
nsbe
twee
nH
ead
Star
tand
birt
hor
der.
Res
ultsfr
om
1999
dat
a:H
ead
Star
tpar
ticip
ants
are
58%
less
likel
yto
smok
eth
ansib
lings
.R
esultsfr
om
2003
dat
a:H
ead
Star
tpar
ticip
ants
are
65%
less
likel
yto
smok
eth
ansib
lings
.In
clud
ing
cont
rolf
ored
ucat
iona
latt
ainm
entm
akes
resu
ltsst
atist
ical
lyin
signi
fica
nt.
Cost
-Ben
efit:
PVre
duct
ion
insm
okin
gis
$996
7pe
rpa
rtic
ipan
t(us
ing
3%di
scou
ntra
te,ac
coun
ting
for
med
ical
expe
nses
and
prod
uctiv
itylo
sses
)Ave
rage
cost
per
Hea
dSt
art
part
icip
anti
n20
03is
$709
2.D
epen
ding
ondi
scou
ntra
teus
ed,
the
valu
eof
redu
ctio
nin
smok
ing
isas
soci
ated
with
36-1
41%
ofpr
ogra
mco
sts.
Exp
andi
ngex
posu
re:ca
nin
crea
sing
the
daily
expo
sure
toH
ead
Star
tred
uce
child
hood
obes
ity?
(Fri
svol
dan
dLu
men
g,20
09)
Adm
inist
rativ
eda
tafr
oma
Mic
higa
nH
ead
Star
tfor
2002
-200
6.n=
1833
obs.
(fro
m15
32ch
ildre
n,sin
ceso
me
atte
ndfo
rm
ultip
leye
ars)
Full-
day
clas
ssa
mpl
e=
424
obs.
Hal
f-da
ycl
asss
ampl
e=
1409
obs.
Est
imat
edth
eeff
ecto
fful
l-da
yvs
.ha
lf-da
yH
ead
Star
ton
obes
ityat
end
ofsc
hool
year
.Id
entifi
catio
nvi
ael
imin
atio
nof
agr
antw
hich
led
toa
decr
ease
inth
e#
offu
ll-da
ycl
asse
sfr
om16
clas
sroo
msi
n20
02to
4cl
assr
oom
sin
2003
.(I
V=
%fu
ll-da
yfu
nded
slots
).C
ontr
olsf
orob
serv
able
fam
ilych
arac
teri
stic
s.
Fir
stSta
geR
esults:
10pp
incr
ease
inpe
rcen
tage
offu
ll-da
yslo
tsin
crea
sesl
ikel
ihoo
dof
full-
day
atte
ndan
ceby
85%
(rel
ativ
eto
base
line=
11%
enro
llmen
tin
full-
day
slots
in20
03).
Sec
ond
Sta
geR
esults:
Full-
day
enro
llmen
tin
Hea
dSt
art
decr
ease
slik
elih
ood
ofob
esity
by14
3%.T
hisi
mpl
iest
hat
child
ren
who
atte
nded
full-
day
clas
sesw
ould
have
been
alm
ost3
lbsh
eavi
erha
dth
eyat
tend
edha
lf-da
ycl
asse
s.Si
mul
atio
nre
sults
sugg
estt
hatt
he14
3%ch
ange
inth
elik
elih
ood
ofob
esity
can
beex
plai
ned
bya
chan
gein
calo
ric
inta
keof
75ca
lori
espe
rda
yw
ithno
chan
gein
phys
ical
activ
ity.
(con
tinue
don
next
page
)
1444 Douglas Almond and Janet Currie
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Stud
iesof
publicpre-K/Kan
dch
ildcare
prog
rams
Impa
ctof
earl
ych
ildho
odca
rean
ded
ucat
ion
onch
ildre
n’sp
resc
hool
cogn
itive
deve
lopm
ent:
Can
adia
nre
sults
from
ala
rge
quas
i-ex
peri
men
t(Le
febv
reet
al.,
2006
)D
ata
from
Can
ada’s
NLS
CY
on4-
and
5-yr
old
child
ren
from
5co
nsec
utiv
ecy
cles
.n=
15,54
6
Iden
tifica
tion
from
chan
gesi
nQ
uebe
c’sc
hild
care
subs
idie
s.O
nSe
p.1,
1997
,ch
ildca
refa
cilit
ies
offer
ed$5
-per
-ful
l-da
yse
rvic
esto
child
ren
who
wer
e4
yrso
ldby
Sep.
30th
.In
the
follo
win
gye
ars,
age
cuto
ffsd
ecre
ased
and
the
num
ber
ofsp
aces
incr
ease
d.N
osim
ilar
polic
ies
inot
her
prov
ince
sofC
anad
ain
1994
-200
3.D
iffer
ence
-in-
differ
ence
(DD
)and
Diff
eren
ce-i
n-di
ffer
ence
-in
-diff
eren
ce(D
DD
)des
igns
,co
mpa
ring
Que
bec’
spre
scho
olch
ildre
nw
ithch
ildre
nof
simila
rag
esfr
omot
her
prov
ince
susin
gth
efa
ctth
atdi
ffer
entc
ohor
tsof
child
ren
wer
eex
pose
dto
differ
entn
umbe
rsof
trea
tmen
tyea
rs.
Subs
idie
sinc
reas
eth
enu
mbe
rof
hour
sin
child
care
by5.
5-7.
4ho
ursp
erw
eek
for
child
ren
aged
4-5.
Effec
tlar
ger
for
mot
hers
inhi
ghes
tedu
catio
nalg
roup
.N
oeff
ects
on4-
year
-old
child
ren’
sPPV
Tsc
ores
.Dec
reas
ein
PPV
Tsc
ores
of0.
33SD
for
5-ye
ar-o
ldch
ildre
n.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1445
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Prom
otin
gsc
hool
read
ines
sin
Okl
ahom
a:A
nev
alua
tion
ofTu
lsa’s
Pre-
KPr
ogra
m(G
orm
ley
and
Gay
er,20
06)a .
Dat
afr
omTu
lsaPu
blic
Scho
ols(
TPS
)on
test
scor
esof
Pre-
Kan
dK
inde
rgar
ten
child
ren
from
test
adm
inist
ered
inA
ug.20
01.
T=
1112
,C=
1284
(T=
child
ren
who
just
com
plet
edPr
e-K
,C=
child
ren
who
are
abou
tto
begi
nPr
e-K
).E
ntry
:4
year
sold
;E
xit:
5ye
arso
ld.
Reg
ress
ion
disc
ontin
uity
desig
nar
ising
from
cuto
ffof
Sept
.1
toen
roll
inPr
e-K
ina
give
nye
ar.
Com
pare
kind
erga
rten
child
ren
who
just
com
plet
edPr
e-K
with
sligh
tlyyo
unge
rchi
ldre
nw
how
ere
inel
igib
leto
atte
nd.U
sed
quad
ratic
poly
nom
ial
tofitu
nder
lyin
gag
e/te
stsc
ore
rela
tions
hip.
Cog
nitiv
e/kn
owle
dge
scor
e:T>
Cby
0.39
SDM
otor
skill
ssco
re:T>
Cby
0.24
SDLa
ngua
gesc
ore:
T>
Cby
0.38
SDLa
rges
tim
pact
sfor
Hisp
anic
s,fo
llow
edby
blac
ks,lit
tleim
pact
for
whi
tes.
Chi
ldre
nw
hoqu
alify
for
free
scho
ollu
nch
have
larg
erim
pact
stha
not
her
child
ren.
Doe
sPre
kind
erga
rten
Impr
ove
Scho
olPr
epar
atio
nan
dPe
rfor
man
ce?
(Mag
nuso
net
al.,
2007
)D
ata
from
EC
LS-K
.n=
10,22
4.
Prim
ary
met
hod
isa
mul
tivar
iate
regr
essio
nto
estim
ate
the
impa
ctof
Pre-
Kat
tend
ance
onva
riou
sou
tcom
es.R
obus
tnes
sche
cksu
sing
teac
her
fixe
deff
ects
,pr
open
sity
scor
em
atch
ing,
and
inst
rum
enta
lvar
iabl
es(I
V).
IVis
differ
entm
easu
reso
fac
cess
topr
e-K
ina
give
nst
ate.
Dep
ende
ntva
riab
lesa
rem
easu
red
inth
efa
llof
kind
erga
rten
and
inth
esp
ring
offirs
tgra
deto
asse
ssan
yla
stin
gim
pact
sofP
re-K
.
Pre-
Kat
tend
ance
:in
crea
sesr
eadi
ngsc
ores
atsc
hool
entr
yby
0.86
SD(I
V);
incr
ease
sagg
ress
ion
atsc
hool
entr
yby
0.69
SD(I
V).
No
effec
ton
mat
hsc
ores
orse
lfco
ntro
lin
IV.E
ffec
tsiz
esfo
ral
lout
com
esar
ela
rger
for
Pre-
Kth
anfo
rot
her
form
sof
child
care
,bu
tPre
-Kch
ildre
nha
vedi
ffer
entc
hara
cter
istic
stha
not
her
child
ren.
Am
ong
child
ren
atte
ndin
gPr
e-K
inth
esa
me
publ
icsc
hool
asth
eir
kind
erga
rten
,hi
gher
read
ing
scor
esar
eno
tacc
ompa
nied
byin
crea
sed
beha
vior
alpr
oble
ms.
For
disa
dvan
tage
dch
ildre
n,co
gniti
vega
insa
rem
ore
last
ing
than
inth
ew
hole
sam
ple.
Effec
tsiz
esfo
rco
gniti
veou
tcom
esm
uch
low
erin
spri
ngof
1stg
rade
than
atsc
hool
entr
y.E
ffec
tsiz
esfo
rbe
havi
oral
outc
omes
are
the
sam
e.
(con
tinue
don
next
page
)
1446 Douglas Almond and Janet Currie
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
The
Effec
tsof
Okl
ahom
a’sPr
e-K
Prog
ram
onH
ispan
icC
hild
ren
(Gor
mle
y,20
08)
Test
sadm
inist
ered
byTu
lsapu
blic
scho
olsi
nA
ug.20
06.
T=
194,
C=
295.
(T=
child
ren
who
just
com
plet
edPr
e-K
,C=
child
ren
who
are
abou
tto
begi
nPr
e-K
).E
ntry
:4
year
sold
;E
xit:
5ye
arso
ld.
See
Gor
mle
yan
dG
ayer
(200
6)en
try.
Lett
er-W
ord
Iden
tifica
tion
Test
scor
e:T>
Cby
0.84
6SD
Spel
ling
scor
e:T>
Cby
0.52
SDA
pplie
dPr
oble
mss
core
:T>
Cby
0.38
SDSi
gnifi
cant
effec
tson
lyfo
rH
ispan
icst
uden
tsw
hose
prim
ary
lang
uage
atho
me
isSp
anish
.
Uni
vers
alC
hild
Car
e,M
ater
nalL
abor
Supp
ly,an
dFa
mily
Wel
l-B
eing
(Bak
eret
al.,
2008
).C
anad
ian
NLS
CY
(199
4-20
03)
incl
udes
only
mar
ried
wom
enan
dth
eir
child
ren.
Ave
rage
of20
00ch
ildre
nat
each
age
per
yr.Pr
imar
ysa
mpl
eag
es0-
4,ro
bust
ness
chec
ksag
es8-
11.
Com
pare
outc
omes
inQ
uebe
c,w
hich
bega
n$5
per
day
dayc
are
for
4ye
arol
dsin
1997
,ex
tend
edpr
ogra
mto
3ye
arol
dsin
1998
,2
year
olds
in19
99,an
dal
lchi
ldre
n<
2in
2000
,to
the
rest
ofC
anad
a.D
iffer
ence
indi
ffer
ence
s.
Incr
ease
inus
eof
any
child
care
/ins
titut
iona
lcar
e/m
othe
rsw
orki
ngby
35%
/>10
0%/1
4.5%
.C
row
ding
outo
foth
erin
form
alch
ildca
re.
Incr
ease
inem
otio
nald
isord
eran
xiet
ysc
ore
(phy
sical
aggr
essio
nan
dop
posit
ion)
by12
%(9
%)f
or2-
3yr
olds
.D
ecre
ase
inst
anda
rdiz
edm
otor
and
soci
alde
velo
pmen
tsco
reby
1.7%
.In
crea
sein
mot
her
depr
essio
nsc
ore
by10
%.
40%
ofth
eco
stof
the
child
care
subs
idy
isoff
setb
yin
crea
sed
taxe
son
extr
ala
bor
supp
ly.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1447
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Impa
ctso
fNew
Mex
ico
Pre-
Kon
Chi
ldre
n’sS
choo
lR
eadi
ness
atK
inde
rgar
ten
Ent
ry:R
esul
tsfr
omth
eSe
cond
Year
ofa
Gro
win
gIn
itiat
ive
(Hus
tedt
etal
.,20
08).
Dat
aon
child
ren
who
part
icip
ated
inth
e2n
dye
arof
the
Pre-
Kpr
ogra
mdu
ring
2006
-200
7an
den
tere
dki
nder
gart
enin
Fall
2007
.T=
405,
C=
519
(T=
child
ren
who
just
com
plet
edPr
e-K
,C=
child
ren
who
are
abou
tto
begi
nPr
e-K
).E
ntry
:4
year
sold
;E
xit:
5ye
arso
ld.
Reg
ress
ion
disc
ontin
uity
desig
ndu
eto
abi
rthd
ayel
igib
ility
cut-
offof
Aug
.31
stto
enro
llin
Pre-
Kin
agi
ven
year
.C
ompa
red
“you
ng”
kind
erga
rten
child
ren
who
just
com
plet
edPr
e-K
with
sligh
tlyyo
unge
rch
ildre
nw
hoar
eab
outt
obe
ing
Pre-
K.U
sed
linea
rm
odel
for
voca
bula
rysc
ore
asde
pend
ent
vari
able
,qu
adra
ticm
odel
for
earl
ylit
erac
ysc
ore
asde
pend
entv
aria
ble,
cubi
cm
odel
for
mat
hsc
ore
asde
pend
entv
aria
ble.
Voca
bula
ry(P
PVT
)sco
re:T>
Cby
0.25
SDM
ath
scor
e:T>
Cby
0.50
SDE
arly
liter
acy
scor
e:T>
Cby
0.59
SDN
ost
atist
ical
lysig
nifica
ntdi
ffer
ence
ineff
ects
betw
een
Pre-
Kpr
ogra
msf
unde
dby
the
Publ
icE
duca
tion
Dep
artm
enta
ndth
ose
fund
edby
the
Chi
ldre
n,Yo
uth
and
Fam
ilies
Dep
artm
ent.
(con
tinue
don
next
page
)
1448 Douglas Almond and Janet Currie
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
An
Effec
tiven
ess-
Bas
edE
valu
atio
nof
Five
Stat
ePr
e-K
inde
rgar
ten
Prog
ram
s(W
ong
etal
.,20
08).
Dat
aon
test
scor
esfr
omfa
ll20
04fr
omM
ichi
gan,
New
Jers
ey,
Okl
ahom
a,So
uth
Car
olin
aan
dW
estV
irgi
nia.
Sam
ple
sizes
:T=
485,
C=
386
(MI)
;T=
1177
,C=
895
(NJ)
;T=
431,
C=
407
(OK
);T=
353,
C=
424
(SC
);T=
379,
C=
341
(WV
).(T=
child
ren
who
just
com
plet
edPr
e-K
,C=
child
ren
who
are
abou
tto
begi
nPr
e-K
).E
ntry
:4
year
sold
;E
xit:
5ye
arso
ld.
Reg
ress
ion
disc
ontin
uity
desig
ndu
eto
ast
rict
birt
hday
elig
ibili
tycu
t-off
.Lo
oked
ateff
ects
ize
differ
ence
sdue
topr
ogra
mm
atic
vari
atio
nbe
twee
nst
ates
.U
sed
apo
lyno
mia
lap
prox
imat
ion
toth
eco
ntin
uous
func
tion
onth
eas
signm
entv
aria
ble
inth
ere
gres
sions
.
Inte
nt-
to-T
reat
Res
ults:
MI:
T>
Cby
0.53
SDfo
rm
ath;
1.09
SDfo
rPr
intA
war
enes
s.N
J:T>
Cby
0.36
SDfo
rPP
VT,
0.23
SDfo
rm
ath,
0.32
SDfo
rPr
intA
war
enes
s.O
K:T>
Cby
0.28
SDfo
rPPV
T,0.
78SD
forP
rint
Aw
aren
ess.
WV
:T>
Cby
0.92
SDfo
rPr
intA
war
enes
s.Tre
atm
ent-
on
-Tre
ated
Res
ults:
MI:
T>
Cby
0.47
SDfo
rm
ath,
0.96
SDfo
rPr
intA
war
enes
s.N
J:T>
Cby
0.36
SDfo
rPP
VT,
0.23
SDfo
rm
ath,
0.50
SDfo
rPr
intA
war
enes
s.O
K:T>
Cby
0.29
SDfo
rPP
VT.
SC:T>
Cby
0.79
SDfo
rPr
intA
war
enes
s.N
ost
atist
ical
lysig
nifica
ntre
sults
for
WV.
No
clea
rre
latio
nshi
pbe
twee
nst
ate
fund
ing
for
Pre-
Kpr
ogra
msa
ndeff
ects
izes
.St
ate
Pre-
Kpr
ogra
msh
ave
larg
ereff
ects
izes
than
Hea
dSt
art.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1449
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
Do
Inve
stm
ents
inU
nive
rsal
Ear
lyE
duca
tion
Pay
Off?
Long
-ter
mE
ffec
tsof
Intr
oduc
ing
Kin
derg
arte
nsin
toPu
blic
Scho
ols(
Cas
cio,
2009
).D
ata
from
4D
ecen
nial
cens
uses
for
1970
,19
80,an
d20
00fr
omth
ePu
blic
Use
Mic
roda
taSa
mpl
es.n=
840
whi
tes,
425
blac
ks.D
ata
from
PSID
onH
ead
Star
tenr
ollm
ent:
n=
174
whi
tes,
126
blac
ks.
Ana
lyze
deff
ecto
fexp
ansio
nof
publ
icki
nder
gart
enon
long
-ter
mou
tcom
es.Id
entifi
catio
nfr
omth
eva
riat
ion
inth
etim
ing
offu
ndin
gin
itiat
ives
amon
gtr
eate
dst
ates
.E
vent
stud
ym
odel
,co
mpa
ring
indi
vidu
als
aged
5be
fore
and
afte
rth
ein
itiat
ives
wer
eim
plem
ente
d.In
clud
eddu
mm
iesf
orco
hort
sint
erac
ted
with
dum
mie
sfor
3di
ffer
entg
roup
sof
trea
ted
stat
esde
fine
don
the
basis
ofav
erag
eed
ucat
ion
expe
nditu
repe
rpu
pili
nth
eea
rly
1960
s.A
lsoco
ntro
lled
for
coho
rt-b
y-re
gion
-of-
birt
hfixe
deff
ects
and
stat
efixe
deff
ects
.U
nits
ofob
serv
atio
nar
eco
hort
-sta
tece
lls.
Whi
tech
ildre
nag
ed5
afte
rth
ety
pica
lsta
tere
form
are
2.5%
less
likel
yto
behi
ghsc
hool
drop
-out
sand
22%
less
likel
yto
bein
stitu
tiona
lized
asad
ults
.N
osig
nifica
nteff
ects
ongr
ade
rete
ntio
n,co
llege
atte
ndan
ce,em
ploy
men
t,or
earn
ings
.N
osig
nifica
nteff
ects
for
blac
ks,de
spite
com
para
ble
incr
ease
sin
enro
llmen
tin
publ
icki
nder
gart
ensp
ostr
efor
m.Po
tent
ial
expl
anat
ion
isth
atst
ate
fund
ing
for
publ
icki
nder
gart
ens
redu
ced
the
likel
ihoo
dth
ata
blac
k5-
year
-old
atte
nded
Hea
dSt
artb
y∼
100%
.R
educ
tion
inH
ead
Star
tatt
enda
nce
may
acco
untf
or16
%of
the
1.13
ppin
crea
sein
the
blac
k-w
hite
gap
inhi
ghsc
hool
drop
-out
rate
s.D
iffer
ence
ineff
ects
oned
ucat
iona
latt
ainm
entb
etw
een
whi
tesa
ndbl
acks
are
driv
enby
fem
ales
.
(con
tinue
don
next
page
)
1450 Douglas Almond and Janet Currie
Table12
(con
tinue
d)Stud
y/prog
ram
namean
dda
taStud
yde
sign
Results
No
Chi
ldLe
ftB
ehin
d:U
nive
rsal
Chi
ldC
are
and
Chi
ldre
n’sL
ong-
Run
Out
com
es(H
avne
sand
Mog
stad
,20
09).
Dat
afr
omSt
atist
icsN
orw
ayon
indi
vidu
alsf
rom
1967
to20
06.H
ouse
hold
info
rmat
ion
from
the
Cen
tral
Popu
latio
nR
egist
er.
Adm
inist
rativ
eda
taon
child
care
inst
itutio
nsan
dth
eir
loca
tions
for
1972
-199
6.R
estr
icte
dsa
mpl
eto
indi
vidu
alsw
hose
mot
hers
wer
em
arri
edat
the
end
of19
75.n=
499,
026
child
ren;
318,
367
fam
ilies
.A
dult
outc
omes
mea
sure
dbe
twee
nag
es30
and
33.
Diff
eren
ce-i
n-di
ffer
ence
s.E
xplo
ited
ach
ildca
rere
form
in19
75in
Nor
way
,w
hich
assig
ned
resp
onsib
ility
for
child
care
tolo
cal
gove
rnm
ents
,an
dth
usre
sulte
din
grea
tvar
iatio
nin
child
care
cove
rage
for
child
ren
aged
3-6
both
betw
een
coho
rtsa
ndac
ross
mun
icip
aliti
es.
T=
mun
icip
aliti
esw
here
child
care
expa
nded
alo
t;C=
mun
icip
aliti
esw
here
child
care
did
notc
hang
em
uch.
Com
pare
dch
ange
sin
outc
omes
for
trea
tmen
tand
cont
rol
adul
tsw
how
ere
3-6
year
sold
befo
rean
daf
ter
the
refo
rm.A
lsoin
vest
igat
edhe
tero
gene
ityof
effec
ts.
Con
trol
led
for
vari
ousc
hild
and
fam
ily-s
peci
fic
char
acte
rist
ics,
asw
ell
asm
unic
ipal
ityfixe
deff
ects
.R
obus
tnes
sche
cks:
incl
uded
atim
etr
end,
chec
ked
for
plac
ebo
effec
tby
com
pari
ngth
etw
ogr
oups
befo
reth
ere
form
,an
dus
edsib
ling
fixe
deff
ects
com
pari
ngsib
lings
expo
sed
toth
ere
form
toth
ose
who
wer
eno
t.
TT
Effec
ts(I
TT
effec
tsin
par
enth
eses
)O
nem
ore
child
care
plac
e:in
crea
sese
duca
tiona
latt
ainm
entb
y2.
7%(0
.4%
)de
crea
sesp
roba
bilit
yof
drop
ping
outo
fHS
by22
.8%
(3.8
%)
incr
ease
spro
babi
lity
ofco
llege
enro
llmen
tby
17%
(2.5
%)
decr
ease
spro
babi
lity
ofha
ving
little
orno
earn
ings
by23
.3%
(3.9
%)
incr
ease
spro
babi
lity
ofha
ving
aver
age
earn
ings
by7.
4%(1
.3%
)de
crea
sesp
roba
bilit
yof
bein
gon
wel
fare
by31
.9%
(5.6
%)
decr
ease
spro
babi
lity
ofpa
rent
hood
by10
.4%
(1.8
%)
incr
ease
spro
babi
lity
ofbe
ing
singl
ew
ithno
child
ren
by23
.3%
(4%
)A
lmos
tall
ofth
ere
duct
ion
inpr
obab
ility
ofbe
ing
alo
wea
rner
isdr
iven
byfe
mal
es.W
omen
mor
elik
ely
tode
lay
child
bear
ing
and
fam
ilyfo
rmat
ion
than
men
asa
resu
ltof
incr
ease
dch
ildca
re.M
ostb
enefi
tsof
univ
ersa
lchi
ldca
rear
efo
rch
ildre
nof
low
-edu
cate
dm
othe
rs.
Subs
idiz
edfo
rmal
child
care
crow
dsou
tinf
orm
alch
ildca
rew
ithal
mos
tno
neti
ncre
ase
into
talc
hild
care
use
orm
ater
nal
empl
oym
ent.
No
impa
ctof
child
care
refo
rmon
mat
erna
led
ucat
ion.
Cost
-Ben
efitA
nal
ysis
:C
ost:
Ann
ualb
udge
tary
cost
per
child
care
plac
e=
$540
0B
enefi
t:In
crea
sein
0.35
yrso
fedu
catio
nim
plie
san
incr
ease
in$2
7,00
0in
lifet
ime
earn
ings
.
See
Kar
oly
etal
.(2
005)
for
mor
ein
form
atio
nab
outs
ome
ofth
est
udie
sdes
crib
edin
this
tabl
e.U
nles
soth
erw
iseno
ted,
none
ofth
ese
eval
uatio
nsw
asra
ndom
ized
.“T
”re
fers
toth
etr
eatm
ent,
and
“C”
refe
rsto
the
cont
rolo
rco
mpa
riso
ngr
oup.
T>
Cm
eans
that
the
differ
ence
was
signi
fica
ntat
the
5%le
vel.
aA
very
simila
rst
udy
byG
orm
ley
etal
.(2
005)
eval
uate
sthe
effec
tsof
Okl
ahom
a’sU
nive
rsal
Pre-
Kpr
ogra
mon
scho
olre
adin
essu
sing
the
sam
ere
gres
sion
disc
ontin
uity
desig
n,bu
tmea
suri
ngou
tcom
esw
ithth
eW
oodc
ock-
John
son
Ach
ieve
men
tTes
t.T
hey
find
a0.
79SD
incr
ease
inth
eLe
tter
-Wor
dId
entifi
catio
nsc
ore,
a0.
64SD
incr
ease
inth
eSp
ellin
gsc
ore,
and
a0.
38SD
incr
ease
inth
eA
pplie
dPr
oble
mss
core
.
Human Capital Development before Age Five 1451
that projected gains in earnings are enough to offset the cost of the program, so that thereis a positive cost/benefit ratio. Carneiro and Ginja (2008) use the same data but a differentidentification strategy: they focus on families around the cutoff for income eligibility forthe program and compare families who are just below (and therefore eligible) to thosewho are just above (and therefore ineligible). A potential problem with this strategy isthat it implicitly assumes that families cannot game the system by reducing their incomesin order to become eligible for the program. Consistent with other studies, they findpositive effects of Head Start attendance on adolescents including reductions in behaviorproblems, grade repetition, depression, and obesity.
Since its inception, Head Start has aimed to improve a broad range of child outcomes(not only test scores). When the program was launched in 1965, the Office of EconomicOpportunity assisted the 300 poorest counties in applying for Head Start funds, andthese counties were significantly more likely than other counties to receive funds. Usinga regression discontinuity design, Ludwig and Miller (2007) show that mortality fromcauses likely to be affected by Head Start fell among children 5 to 9 in the assisted countiesrelative to the others. Mortality did not fall in slightly older cohorts who would not havebeen affected by the introduction of the program.
Frisvold (2006) and Frisvold and Lumeng (2009) also focus on health effects byexamining the effect of Head Start on obesity. The former instruments Head Startattendance using the number of Head Start places available in the community, whilethe later takes advantage of a cut in a Michigan Head Start program which resulted in theconversion of a number of full-day Head Start places to half day places. Both studies findlarge and significant effects of Head Start on the incidence of obesity. In defense of theirestimates, which some might find implausibly large, Frisvold and Lumeng point out thata reduction of only 75 calories per day (i.e. less than a slice of bread or an apple) wouldbe sufficient to yield their results. In small children, even small changes in diet may havelarge cumulative effects. Anderson et al. (2009) follow Garces et al. and use sibling fixedeffects and data from the PSID to estimate the effect of Head Start on smoking as anadult. Again, they find large effects.
Head Start has served as a model for state preschools targeted to low-income childrenin states such as California, and also for new (non-compulsory) universal preschoolprograms in Georgia, and Oklahoma. The best available evaluations of universalpreschool programs highlight the importance of providing a high quality program thatis utilized by the neediest children. Baker et al. (2008) examine the introduction of auniversal, $5 per day (later $7), preschool program in the Canadian province of Quebec.
The authors find a strong response to the subsidy in terms of maternal labor supply andthe likelihood of using care, but they find negative effects on children for a range ofoutcomes. Lefebvre et al. (2006) focus on the same natural experiment and examine theeffects on children’s vocabulary scores, which have been shown to be a good predictorof schooling attainment in early grades. They find strong evidence of negative effects.
1452 Douglas Almond and Janet Currie
In interpreting this study, it is important to consider who was affected by the program.
Because poor children were already eligible for child care subsidies, the marginal childaffected by this program was a child who probably would have stayed home with his orher middle-class, married, mother, and instead was put into child care. Moreover, themarginal child care slot made available by the program was of low quality—the suddeninflux of children into care caused the province to place more emphasis on makingslots available than on regulating their quality. Hence, the study should be viewed as theconsequence of moving middle class children from home care to relatively poor qualitycare. It is not possible to draw any conclusion from this study about the effect of drawingpoor children into care of good quality, which is what model preschool programs andHead Start aim to do.
Gormley and Gayer (2006) examine the effects of Oklahoma’s universal pre-Kprogram, which is run through the public schools and is thought to be of high quality.They take advantage of strict age cutoffs for the program and compare children who hadjust attended for a year to similar children who were ineligible to attend because theywere slightly younger. They find a 52% gain in pre-reading skills, a 27% gain in pre-
writing skills, and a 21% gain in pre-math skills. These results suggest that a high qualityuniversal pre-K program might well have positive effects, though one would have to trackchildren longer to determine whether these initial gains translate into longer term gainsin schooling attainment. Several other recent studies use a similar regression discontinuitydesign, including Hustedt et al. (2008) and Wong et al. (2008) who examine state pre-Kprograms in five states. These studies find uniformly positive effects. It has been arguedin fact, that the effects of quality state preschool programs are larger than those of HeadStart. However, it is difficult to control for pre-existing differences between the HeadStart children and children who attend other preschools. For example in Magnusonet al. (2007), the preschool children had systematically higher incomes than those whoattended Head Start.
A handful of studies examine the long-term effects of public pre-school orkindergarten programs. Cascio (2009) uses data from four decennial censuses to analyzethe impact of introducing kindergarten into public schools in the US, where kindergartenwas phased in on a state-by-state basis. Using a cohort-based design, she finds that whitechildren born in adopting states after the reform were less likely to dropout of high schooland less likely to be institutionalized as adults. However, she finds no significant effect forblacks, which may be due to significant crowd out of blacks from other programs, suchas Head Start. Like Anderson, she finds that the effects were larger for girls. Havnes andMogstad (2009) study a 1975 policy change in Norway which increased the availability ofregulated child care in some areas but not in others. They find that children “exposed” tomore child care received more education and were more likely to have earnings as adults.Once again, much of the benefit was concentrated among females, and children of lesseducated mothers were particularly likely to benefit. In terms of mechanisms, they find
Human Capital Development before Age Five 1453
that the increase in formal care largely displaced informal care, without much net effecton the mother’s labor force participation.
Finally, it is worth mentioning the “Sure Start” program in England and Wales. Thisprogram aimed to provide early intervention services in disadvantaged neighborhoodsbut allowed a wide variety of program models, which obviously complicates anassessment of the program. An evaluation was conducted by comparing communities thatwere early adopters to those that adopted later. A second evaluation compared Sure Startchildren to children from similar neighborhoods who were drawn from the MillenniumCohort study. This second study used propensity scores to balance the samples. Thefirst evaluation found that the most disadvantaged households were actually doing morepoorly in intervention areas than in other areas (NESS, 2005), while the second foundsome evidence of positive effects (NESS, 2008). Following the first evaluation, there hasbeen a move to standardize the intervention and most communities are now offering SureStart Children’s Centers. This latest incarnation of the program remains to be evaluated.
This discussion shows the value of using a framework for the production of childquality as a lens for the interpretation of the program evaluation literature. As discussedabove, child human capital is produced using inputs that may come from either the familyor from other sources. A program that augments the resources available to the child islikely to have positive effects (subject of course to diminishing returns), while a programthat reduces the resources available to the child is likely to have negative effects. Hence, aprogram that causes poor quality group time to be substituted for relatively high qualitymaternal time can have a negative effect and vice versa. The important point is that it ispossible to intervene effectively and to improve the trajectories of young children.
5.3.4. Health insuranceHealth insurance is not an intervention program in the sense of the programs describedabove. Yet, there is a good deal of evidence that access to health insurance improveschildren’s health at birth and afterwards. Much of the evidence comes from studies ofthe introduction, or expansion, of health insurance benefits. Some of this literature issummarized in Table 13. For example, Hanratty (1996) examined the introduction ofpublic health insurance in Canada, which was phased in on a province-by-provincebasis. Using county-level panel data, she finds that the introduction of health insurancewas associated with a decline of four percent in the infant mortality rate, and that theincidence of low birth weight also decreased by 1.3% for all parents and by 8.9% forsingle parents. Currie and Gruber (1996) conduct a similar exercise for the US, focusingon an expansion of public health insurance to pregnant women and infants. They findthat the effects vary depending on whether the expansion covered the poorest women, orwomen somewhat higher in the income distribution. Narrowly targeted expansions thatincreased the fraction of the poorest women eligible by 30%, reduced low birth weightby 7.8%, and reduced infant mortality by 11.5%. Expansions of eligibility of a similarmagnitude to women of higher incomes had very small effects on the incidence of low
1454 Douglas Almond and Janet Currie
Table13
Effectsof
Med
icaidan
dothe
rpub
liche
alth
insuranceon
birthan
dearly
child
hood
outcom
es.
Stud
yan
dda
taStud
yde
sign
Results
Effects
onbirthweigh
tand
health
atbirth
The
effica
cyan
dco
stof
chan
gesi
nth
eM
edic
aid
elig
ibili
tyof
preg
nant
wom
en(C
urri
ean
dG
rube
r,19
96).
Not
e:A
utho
rsal
soco
nduc
tan
anal
ysis
ofM
edic
aid
take
-up,
whi
chis
noti
nclu
ded
inth
ere
sults
here
.D
ata
from
CPS
and
Vita
lSt
atist
ics.
Dat
aon
Med
icai
dex
pend
iture
sfr
omth
eH
ealth
Car
eFi
nanc
ing
Adm
inist
ratio
n.D
ata
onth
eus
eof
med
ical
serv
ices
bypr
egna
ntw
omen
from
the
NLS
Y.Si
mul
ated
mod
elfo
rta
rget
edch
ange
sest
imat
edfo
r19
79-1
992.
Sim
ulat
edm
odel
for
broa
dch
ange
ses
timat
edfo
r19
87-1
992.
(n=
600)
.
Exp
loite
dva
riat
ion
betw
een
stat
esin
the
timin
gof
expa
nsio
nsof
Med
icai
del
igib
ility
.U
sea
fixe
dsa
mpl
eto
simul
ate
the
frac
tion
elig
ible
unde
rdi
ffer
ents
tate
rule
s.D
istin
guish
“tar
gete
d”ch
ange
saff
ectin
gve
rylo
win
com
ew
omen
from
“bro
ad”
chan
gest
ow
omen
furt
her
upth
ein
com
edi
stri
butio
n.In
stru
men
tthe
actu
alfr
actio
nof
wom
enel
igib
lein
each
stat
ean
dye
arw
ithth
esim
ulat
edel
igib
ility
mea
sure
.C
ontr
olle
dfo
rst
ate
fixe
deff
ects
and
time
vary
ing
stat
ech
arac
teri
stic
s.
The
perc
enta
geof
15-4
4yr
old
wom
enel
igib
lefo
rM
edic
aid
(had
they
beco
me
preg
nant
)ros
efr
om12
.4%
to43
.3%
b/n
1979
and
1991
.A
30%
incr
ease
inel
igib
ility
lead
sto
a1.
9%de
crea
sein
inci
denc
eof
low
birt
hw
eigh
t(sig
.at
10%
leve
l)an
da
8.5%
decr
ease
inin
fant
mor
talit
yra
te.Fo
rta
rget
edpr
ogra
mch
ange
s,a
30%
incr
ease
inel
igib
ility
decr
ease
slow
birt
hw
eigh
t(in
fant
mor
talit
y)by
7.8%
(11.
5%).
For
broa
dpr
ogra
mch
ange
sa30
%in
crea
sein
elig
ibili
tyde
crea
sesl
owbi
rth
wei
ght(
infa
ntm
orta
lity)
by0.
2%(2
.9%
).
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1455
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Can
adia
nN
atio
nalH
ealth
Insu
ranc
ean
dIn
fant
Hea
lth(H
anra
tty,
1996
).C
ount
y-le
velp
anel
data
onin
fant
mor
talit
yfr
om10
prov
ince
sin
1960
-197
5fr
omth
eC
ensu
sofC
anad
aan
dfr
omC
anad
a’sD
ivisi
onof
Vita
lSta
tistic
s.(n=
204
coun
ties)
.D
ata
onbi
rth
wei
ghtf
rom
asa
mpl
eof
all
birt
hre
cord
sin
Can
ada
from
1960
to19
74.
Use
dva
riat
ion
intim
ing
ofim
plem
enta
tion
ofna
tiona
lhea
lthin
sura
nce
acro
sspr
ovin
cesi
nC
anad
aov
er19
62-7
2.Lo
gito
fout
com
eson
adu
mm
yfo
rha
ving
natio
nalh
ealth
insu
ranc
ein
apa
rtic
ular
coun
ty-y
ear,
cont
rolli
ngfo
rde
mog
raph
ican
dso
cio-
econ
omic
fact
ors,
atim
etr
end,
and
year
fixe
deff
ects
.
Intr
oduc
tion
ofna
tiona
lhea
lthin
sura
nce
lead
sto
decl
ines
of4%
inin
fant
mor
talit
yra
tes;
1.3%
inlo
wbi
rth
wei
ght
(who
lesa
mpl
e);8.
9%in
low
birt
hw
eigh
t(sin
gle
pare
nts)
.N
oim
pact
onbi
rth
wei
ghta
mon
gm
arri
edw
omen
.
Cha
nges
inpr
enat
alca
retim
ing
and
low
birt
hw
eigh
tby
race
and
soci
oeco
nom
icst
atus
:Im
plic
atio
nsfo
rth
eM
edic
aid
expa
nsio
nsfo
rpr
egna
ntw
omen
(Dub
ayet
al.,
2001
).D
ata
onbi
rths
from
the
1980
,19
86,an
d19
93N
atal
ityFi
les.
n=
8,10
0,00
0bi
rths
.
Diff
eren
ce-i
n-di
ffer
ence
,su
btra
ctin
gdi
ffer
ence
inob
stet
rica
lout
com
es(r
ates
ofla
tein
itiat
ion
ofpr
enat
alca
rean
dra
teso
flo
wbi
rth
wei
ght)
b/n
1980
and
1986
from
differ
ence
inou
tcom
esb/
n19
86an
d19
90,w
ithin
soci
oeco
nom
ic(S
ES)
grou
ps.
Also
com
pare
dch
ange
sin
obst
etri
cal
outc
omes
in19
86-9
3b/
nw
omen
oflo
wan
dhi
ghSE
S(s
ince
high
SES
wom
enw
ere
nota
ffec
ted
byM
edic
aid
expa
nsio
ns).
SES
define
dby
mar
itals
tatu
sand
year
sof
scho
olin
g.M
edic
aid
expa
nsio
nsoc
curr
edin
1986
-93.
Con
trol
led
for
year
,ag
eof
mot
her,
pari
ty,an
dag
e-pa
rity
inte
ract
ions
.
Res
ultsfr
om
diff
-in
-diff
within
SES
:M
edic
aid
expa
nsio
nsas
soci
ated
with
decr
ease
sof:
12%
-21%
pren
atal
care
initi
atio
naf
ter
1stt
rim
este
rfo
rw
hite
wom
en;3%
-5%
inlo
wbi
rth
wei
ghta
mon
gw
hite
wom
enw
ith<
12ye
arse
duca
tion;
10%
-13%
inpr
enat
alca
rein
itiat
ion
afte
r1s
ttri
mes
ter
for
blac
kw
omen
with<
12ye
arse
duca
tion;
13%
-27%
inpr
enat
alca
rein
itiat
ion
afte
r1s
ttri
mes
ter
for
blac
kw
omen
with
12-1
5ye
arso
fedu
catio
n;12
%-3
5%in
pren
atal
care
initi
atio
naf
ter
1stt
rim
este
rfo
rbl
ack
wom
enw
ith>
15ye
arso
fedu
catio
n.A
ssoc
iatio
nw
itha
3%in
crea
sein
likel
ihoo
dof
low
birt
hw
eigh
tfor
unm
arri
edbl
ack
wom
enw
ith<
12ye
ars
educ
atio
n.Si
mila
rre
sults
usin
gdi
ff-i
n-di
ffac
ross
SES
for
1986
-93.
(con
tinue
don
next
page
)
1456 Douglas Almond and Janet Currie
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Effec
tsof
Med
icai
dex
pans
ions
and
wel
fare
cont
ract
ions
onpr
enat
alca
rean
din
fant
heal
th(C
urri
ean
dG
rogg
er,
2002
).D
ata
onbi
rth
outc
omes
from
VSD
Nfile
s19
90-1
996.
Dat
aon
feta
lde
aths
from
vita
lsta
tistic
sfe
tald
eath
sdet
aile
dre
cord
s.D
ata
onM
edic
aid
adm
inist
rativ
ere
form
sfr
omN
atio
nalG
over
nor’s
Ass
ocia
tion
Mat
erna
land
Chi
ldH
ealth
new
slett
ers.
Dat
aon
wel
fare
case
load
sfr
omth
eU
SD
epar
tmen
tof
Hea
lthan
dH
uman
Serv
ices
.n=
3,98
5,96
8w
hite
s;4,
014,
935
blac
ks.
Logi
treg
ress
ion
that
incl
udes
dum
mie
sfor
stat
e-le
velM
edic
aid
adm
inist
rativ
ere
form
s;in
com
eel
igib
ility
cuto
ffs;
rate
sof
wel
fare
part
icip
atio
n;un
empl
oym
entr
ates
;an
dm
ater
nalo
bser
vabl
ech
arac
teri
stic
s.A
lsoes
timat
edau
xilia
ryre
gres
sions
that
exam
ine
the
effec
tofp
olic
yva
riab
leso
nag
greg
ate
Med
icai
dca
selo
ads.
Med
icai
dca
selo
adin
crea
sesb
y0.
233
fore
ach
1%in
crea
sein
wel
fare
rate
.M
edic
aid
case
load
incr
ease
sby
0.66
4fo
rea
ch10
0%in
crea
sein
inco
me
cut-
off(f
orth
ose
notr
ecei
ving
cash
bene
fits
).In
crea
sein
inco
me
cuto
fffr
om10
0%to
200%
ofpo
vert
ylin
ein
crea
sesp
roba
bilit
yof
adeq
uate
pren
atal
care
by0.
4%fo
rw
hite
s.2
ppin
crea
sein
wel
fare
rate
asso
ciat
edw
/1.
1%in
crea
sein
prob
abili
tyth
atpr
enat
alca
rew
asin
itiat
edin
1stt
rim
este
r;0.
7%in
crea
sein
prob
abili
tyof
adeq
uate
care
for
whi
tes;
2%in
crea
sefo
rbo
thfo
rbl
acks
.In
crea
sein
inco
me
cuto
fffr
om10
0%to
200%
ofpo
vert
ylin
eas
soci
ated
w/
ade
crea
seof
1720
feta
ldea
thsp
erye
aram
ong
blac
ks.2%
incr
ease
inw
elfa
reas
soci
ated
w/
10%
decr
ease
infe
tald
eath
sper
year
amon
gbl
acks
.M
ost
adm
inist
rativ
ere
form
shav
eno
effec
t.B
utus
ing
mai
l-in
form
s(in
stea
dof
in-p
erso
nin
terv
iew
s)in
crea
sesp
roba
bilit
yth
atpr
enat
alca
rew
asin
itiat
edin
1stt
rim
este
rby
0.7%
for
blac
ksan
dsh
orte
rfo
rmsi
ncre
ase
prob
abili
tyth
atpr
enat
alca
rew
asin
itiat
edin
1stt
rim
este
rby
3%fo
rw
hite
s.U
sing
mai
l-in
form
sdec
reas
espr
obab
ility
oflo
wbi
rth
wei
ghtb
y26
%;of
very
low
birt
hw
eigh
tby
38%
for
whi
tes.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1457
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Usin
gdi
scon
tinuo
usel
igib
ility
rule
sto
iden
tify
the
effec
tsof
the
fede
ral
Med
icai
dex
pans
ions
onlo
w-i
ncom
ech
ildre
n(C
ard
and
Shor
e-Sh
eppa
rd,
2004
).D
ata
from
SIPP
for
1990
-93,
Mar
chC
PSfo
r19
90-9
6,an
dH
ealth
Inte
rvie
wSu
rvey
for
1992
-199
6on
child
ren
unde
r18
year
sold
.n=
10,2
68to
16,1
96ac
ross
the
differ
enty
ears
inSI
PP.
Two
sour
ceso
fide
ntifi
catio
n:“T
he13
3%ex
pans
ion”
(chi
ldre
nun
der
age
6liv
ing
infa
mili
esw
ithin
com
esbe
low
133%
ofth
epo
vert
ylin
ebe
cam
eco
vere
din
1989
)and
“the
100%
expa
nsio
n”(c
hild
ren
born
afte
rSe
ptem
ber
30,19
83in
fam
ilies
with
inco
mes
belo
wth
epo
vert
ylin
ebe
cam
eco
vere
d).D
iffer
ence
-in-
differ
ence
desig
nco
mpa
ring
age-
6an
dag
e-5
child
ren
infa
mili
esw
ithin
com
esbe
twee
n10
0%an
d13
3%of
the
pove
rty
line
for
the
133%
expa
nsio
n,an
dco
mpa
ring
child
ren
born
befo
rean
daf
ter
Sep.
30,19
83fo
rth
e“1
00%
expa
nsio
n”.R
egre
ssio
nof
Med
icai
den
rollm
ento
ndu
mm
yfo
rbe
ing
belo
wpo
vert
yle
vel,
dum
my
for
bein
gbo
rnaf
ter
9/30
/198
3,th
eir
inte
ract
ion,
dum
my
for
age<
6ye
arso
ld,in
tera
ctio
nbe
twee
ndu
mm
yfo
rag
e<
6ye
arso
ldan
ddu
mm
yfo
rbe
ing
betw
een
100%
and
133%
ofth
epo
vert
ylin
e,a
flex
ible
func
tion
ofag
ean
dfa
mily
inco
me,
and
othe
rba
ckgr
ound
char
acte
rist
icsa
swel
las
year
fixe
deff
ects
.
The
100%
expa
nsio
nle
dto
7%-1
1%ta
ke-u
pra
tes,
whi
leth
e13
3%ex
pans
ion
had<
5%ta
ke-u
pra
tes.
No
evid
ence
for
othe
rin
sura
nce
crow
d-ou
tin
SIPP
data
.R
esul
tsfr
omC
PSda
tasu
gges
ttha
tthe
133%
expa
nsio
nle
dto
decl
ine
inot
her
heal
thin
sura
nce
cove
rage
byap
prox
imat
ely
the
sam
eam
ount
asth
eta
ke-u
pin
Med
icai
d.Si
mila
rre
sults
usin
gH
ealth
Inte
rvie
wSu
rvey
data
.
(con
tinue
don
next
page
)
1458 Douglas Almond and Janet Currie
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Effec
tsof
Med
icai
dm
anag
edca
reon
pren
atal
care
and
birt
hou
tcom
es(A
izer
etal
.,20
07).
Dat
aon
birt
hou
tcom
esfr
omth
eC
alifo
rnia
Bir
thSt
atist
ical
Mas
ter
File
1990
-200
0an
dB
irth
Coh
ortfi
lesf
orth
esa
me
time
peri
od.H
ospi
tal-
leve
lda
tafr
omV
italS
tatis
tics
reco
rds.
n=
55,0
00bi
rths
.
Exp
loite
dth
eco
unty
-by-
coun
tyva
riat
ion
inim
plem
enta
tion
ofM
edic
aid
man
aged
care
,w
hich
resu
lted
from
aph
ase-
inpo
licy
inC
alifo
rnia
that
requ
ired
wom
enen
rolle
din
Med
icai
dto
switc
hto
man
aged
care
plan
s.So
me
coun
tiess
witc
hed
toC
OH
Spl
an,ot
hers
switc
hed
to2
orm
ore
plan
syst
em.M
ultiv
aria
tere
gres
sion
with
coun
tyfixe
deff
ects
,an
dm
othe
rfixe
deff
ects
,as
wel
laso
ther
obse
rvab
lech
arac
teri
stic
s.R
obus
tnes
sche
cks:
Est
imat
ing
sam
em
odel
for
mar
ried
wom
enon
ly(u
nlik
ely
tobe
onw
elfa
re,
henc
eun
likel
yto
besu
bjec
tto
MM
C);
cont
rolli
ngfo
rm
obili
tyb/
nco
untie
s;re
gres
sion
disc
ontin
uity
desig
nto
elim
inat
etim
etr
end
effec
ts;ad
ding
inte
ract
ion
term
sb/n
time
tren
dan
dse
vera
ldum
mie
s;“i
nten
t-to
-tre
at”
mod
els
toco
ntro
lfor
the
MM
Cad
optio
nbe
ing
note
xoge
nous
toco
untie
s.
Prob
abili
tyof
star
ting
pren
atal
care
in1s
ttri
mes
ter:
decr
ease
dby
9%-1
0%in
both
CO
HS
and
2-pl
anco
untie
s.U
seof
indu
ctio
n/st
imul
atio
nof
labo
r:in
crea
sed
by43
.8%
inC
OH
Sco
untie
s.U
seof
feta
lmon
itors
:in
crea
sed
by25
.9%
inC
OH
Sco
untie
s.In
cide
nce
oflo
wbi
rth
wei
ght:
incr
ease
dby
15%
inbo
thC
OH
San
d2-
plan
coun
ties.
Inci
denc
eof
shor
tges
tatio
n:in
crea
sed
by15
%in
both
CO
HS
and
2-pl
anco
untie
s.In
cide
nce
ofne
onat
alde
ath:
incr
ease
dby
50%
in2-
plan
coun
ties.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1459
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Effects
onlaterc
hild
outcom
es
Med
icai
del
igib
ility
and
the
inci
denc
eof
ambu
lato
ryca
rese
nsiti
veho
spita
lizat
ions
forc
hild
ren
(Kae
stne
ret
al.,
2001
).D
ata
from
the
Nat
ionw
ide
Inpa
tient
Sam
ple
ofth
eH
ealth
care
Cos
tand
Util
izat
ion
Proj
ectf
or19
88an
d19
92.n=
36,0
00.
Diff
eren
ce-i
n-di
ffer
ence
,co
mpa
ring
the
chan
gein
ambu
lato
ryca
rese
nsiti
ve(A
CS)
hosp
italiz
atio
nsbe
fore
and
afte
rM
edic
aid
expa
nsio
nsb/
npo
oran
dno
n-po
orch
ildre
n.Po
vert
yst
atus
dete
rmin
edby
med
ian
fam
ilyin
com
ein
child
’szi
pco
deof
birt
h.Tw
otr
eatm
entg
roup
s:in
com
e<
$25,
000
and
$25,
000<
inco
me<
$35,
000.
Con
trol
grou
p:in
com
e>
$35,
000.
Sepa
rate
estim
ates
for
child
ren
aged
2-6,
7-9.
Con
trol
led
for
hosp
ital-
spec
ific,
year
,an
din
divi
dual
fact
ors.
Inci
denc
eof
AC
Sho
spita
lizat
ions
calc
ulat
edus
ing
both
non-
AC
Sho
spita
lizat
ions
and
tota
lbir
ths
inth
ede
nom
inat
or.
For
child
ren
aged
2-6
infa
mili
esw
ith<
$25,
000
inco
me,
inci
denc
eof
AC
Sho
spita
lizat
ions
due
tode
hydr
atio
n,co
nvul
sions
and
non-
asth
ma
illne
sses
decl
ined
by10
%-2
0%.
Est
imat
edeff
ects
izes
of40
%-8
0%fo
rth
ose
affec
ted
byM
edic
aid
expa
nsio
ns.
For
child
ren
aged
2-6
infa
mili
esw
ith$2
5,00
0<
inco
me<
$35,
000,
inci
denc
eof
AC
Sho
spita
lizat
ions
due
tono
n-as
thm
aill
ness
esan
dpn
eum
onia
decl
ined
by10
%-1
4%(o
nly
whe
nde
nom
inat
oris
tota
lbi
rths
).Fo
rch
ildre
nag
ed7-
9in
fam
ilies
with
$25,
000<
inco
me<
$35,
000,
inci
denc
eof
AC
Sho
spita
lizat
ions
due
toas
thm
ade
clin
edby
22%
-30%
;ho
spita
lizat
ions
due
toey
e,no
sean
dth
roat
illne
sses
decl
ined
by>
100%
.N
osig
nifica
nteff
ects
onch
ildre
nag
ed2-
6in
$25,
000<
inco
me<
$35,
000
grou
por
onch
ildre
nag
ed7-
9in<
$25
,000
inco
me
grou
p.
(con
tinue
don
next
page
)
1460 Douglas Almond and Janet Currie
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Low
take
-up
inM
edic
aid:
does
outr
each
mat
ter
and
for
who
m?
(Aiz
er,20
03).
Dat
aon
mon
thly
Med
icai
den
rollm
entf
orch
ildre
nag
ed0-
15in
CA
from
1996
to20
00lin
ked
with
info
rmat
ion
onth
enu
mbe
rof
com
mun
ity-b
ased
appl
icat
ion
assis
tant
sin
each
zip
code
inth
em
onth
ofap
plic
atio
nby
age
and
race
.n=
324,
331
for
anal
ysis
ofta
ke-u
p.n
=12
1,80
6fo
ran
alys
isof
hosp
italiz
atio
ns.
Exa
min
edeff
ects
ofC
Aou
trea
chca
mpa
ign
laun
ched
inJu
ne19
98,w
hich
cons
isted
ofco
mm
unity
-bas
edap
plic
atio
nas
sista
ntsa
nda
med
iaca
mpa
ign
tora
iseaw
aren
essa
bout
Med
icai
d,on
Med
icai
den
rollm
ent.
Mul
tivar
iate
regr
essio
nw
ithke
yex
plan
ator
yva
riab
lebe
ing
the
num
ber
ofco
mm
unity
-bas
edap
plic
atio
nas
sista
nts
inea
chzi
pco
dein
the
mon
thof
appl
icat
ion
byag
ean
dra
ce.In
clud
edzi
pco
dean
dm
onth
fixe
deff
ects
asw
ella
sco
ntro
lsfo
rch
ange
sin
busin
essc
ycle
and
the
dem
ogra
phic
com
posit
ions
ofth
est
ate.
Also
exam
ined
effec
tofe
arly
Med
icai
den
rollm
ento
nA
CS
hosp
italiz
atio
nsby
inst
rum
entin
gM
edic
aid
enro
llmen
twith
outr
each
mea
sure
s.
Acc
esst
obi
lingu
alap
plic
atio
nas
sista
ntsi
ncre
ases
new
mon
thly
Med
icai
den
rollm
entb
y4.
6%am
ong
Hisp
anic
child
ren,
and
by6%
amon
gA
sian
child
ren
(rel
ativ
eto
othe
rch
ildre
nin
sam
ene
ighb
orho
od).
A10
00-c
hild
incr
ease
inM
edic
aid
enro
llmen
tsde
crea
sesA
CS
hosp
italiz
atio
nsby
3.26
hosp
italiz
atio
ns(m
ean
notr
epor
ted,
soca
n’tc
alcu
late
rela
tive
effec
tsiz
e).
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1461
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Publ
icin
sura
nce
and
child
hosp
italiz
atio
ns:A
cces
sand
effici
ency
effec
ts(D
afny
and
Gru
ber,
2005
).D
ata
from
NH
DS
onch
ilddi
scha
rges
for
1983
to19
96.C
ells
define
dfo
r4
age
cate
gori
es(<
1,1-
5,6-
10,11
-15)
for
each
stat
ean
dye
ar.n=
2308
cells
.U
sed
age-
stat
e-ye
arpo
pula
tions
from
the
Cen
susB
urea
uto
calc
ulat
eho
spita
lizat
ion
rate
sfor
each
cell.
Inve
stig
ated
impa
ctof
Med
icai
dex
pans
ions
onho
spita
lizat
ions
usin
gth
eva
riat
ion
inM
edic
aid
elig
ibili
tyb/
nst
ates
,ov
ertim
e,an
dby
age.
Key
expl
anat
ory
vari
able
isth
eel
igib
ility
rate
mea
sure
dby
the
frac
tion
ofch
ildre
nel
igib
lefo
rM
edic
aid
inea
chag
e-st
ate-
year
cell.
Con
trol
led
for
age,
stat
e,an
dye
arfixe
deff
ects
asw
ella
ssta
te-y
ear
inte
ract
ions
.U
sed
the
frac
tion
elig
ible
calc
ulat
edus
ing
afixe
dsa
mpl
eto
inst
rum
entf
orac
tual
elig
ibili
tyin
each
stat
e,ye
ar,an
dag
egr
oup.
Also
exam
ined
effec
tofM
edic
aid
elig
ibili
tyon
leng
thof
stay
inho
spita
land
the
num
ber
ofpr
oced
ures
perf
orm
ed.
A10
ppin
crea
sein
Med
icai
del
igib
ility
lead
sto
a8.
4%in
crea
sein
tota
lhos
pita
lizat
ions
,an
8.1%
incr
ease
inun
avoi
dabl
eho
spita
lizat
ions
,an
dno
stat
istic
ally
signi
fica
ntim
pact
onav
oida
ble
hosp
italiz
atio
ns.A
ssum
ing
acce
ssto
hosp
italc
are
incr
ease
sthe
likel
ihoo
dof
allk
inds
ofho
spita
lizat
ions
equa
lly,th
ese
resu
ltsim
ply
that
incr
ease
dus
eof
prim
ary
care
enge
nder
edby
Med
icai
dex
pans
ions
miti
gate
dth
ein
crea
sein
tota
lhos
pita
lizat
ions
byre
duci
ngth
ein
crea
sein
avoi
dabl
eho
spita
lizat
ions
that
wou
ldha
veot
herw
iseoc
curr
ed.A
10pp
incr
ease
inel
igib
ility
lead
sto
a3.
1%de
crea
sein
the
leng
thof
hosp
itals
tay,
a5%
incr
ease
inth
e#
ofpr
oced
ures
perf
orm
ed,an
da
6.6%
incr
ease
inth
elik
elih
ood
ofha
ving
any
proc
edur
epe
rfor
med
,i.e
.le
adst
om
ore
aggr
essiv
eca
re.
Publ
iche
alth
insu
ranc
e,pr
ogra
mta
ke-u
pan
dch
ildhe
alth
(Aiz
er,20
07).
See
Aiz
er(2
003)
entr
yfo
rde
scri
ptio
nof
data
.
See
Aiz
er(2
003)
entr
y.A
dditi
onal
anal
ysis:
Exa
min
edhe
tero
gene
ityin
effec
tson
take
-up
byag
e.E
xam
ined
nonl
inea
reff
ects
onta
ke-u
p.E
xam
ined
effec
tsof
Eng
lish
and
Span
ishla
ngua
gead
vert
isem
ents
onta
ke-u
p.
Prox
imity
toan
addi
tiona
lbili
ngua
lapp
licat
ion
assis
tant
incr
ease
snew
mon
thly
Med
icai
den
rollm
entb
y7%
-9%
amon
gH
ispan
icsa
ndby
27%
-36%
amon
gA
sians
.Sm
alle
steff
ects
for
infa
ntsw
how
ere
alre
ady
larg
ely
elig
ible
;la
rges
teff
ects
for
ages
6-15
.E
ffec
tisl
inea
rfo
rH
ispan
ics;
sligh
tlyco
ncav
efo
rAsia
ns.E
nglis
hla
ngua
gead
vert
isem
ents
incr
ease
Med
icai
den
rollm
enti
nth
efo
llow
ing
mon
thby
4.7%
amon
gal
lchi
ldre
n,an
dSp
anish
lang
uage
adve
rtise
men
tsin
crea
seM
edic
aid
enro
llmen
tin
the
follo
win
gm
onth
byan
addi
tiona
l2.5
%am
ong
Hisp
anic
child
ren.
Incr
easin
gth
e#
ofch
ildre
nw
/M
edic
aid
by10
%re
sults
ina
2%-3
%de
clin
ein
avoi
dabl
eho
spita
lizat
ions
.
(con
tinue
don
next
page
)
1462 Douglas Almond and Janet Currie
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Has
publ
iche
alth
insu
ranc
efo
rol
der
child
ren
redu
ced
disp
ariti
esin
acce
ssto
care
and
heal
thou
tcom
es?
(Cur
rie
etal
.,20
08).
Dat
afr
omth
ena
tiona
lhe
alth
inte
rvie
wsu
rvey
sfo
r19
86-2
005.
n=
474,
164
child
ren<
18yr
sold
.In
stru
men
tdat
afr
omth
eC
PS.
Iden
tifica
tion
due
toth
efa
ctth
atex
pans
ions
inM
edic
aid/
SCH
IPel
igib
ility
for
olde
rch
ildre
nre
lativ
eto
youn
ger
child
ren
happ
ened
atdi
ffer
entt
imes
indi
ffer
ents
tate
s.In
stru
men
ted
for
indi
vidu
alM
edic
aid/
SCH
IPel
igib
ility
usin
gan
inde
xof
gene
rosit
yof
the
stat
e’s
publ
iche
alth
insu
ranc
epr
ogra
ms
calc
ulat
edus
ing
afixe
dgr
oup
ofch
ildre
n.T
hen,
estim
ated
impa
ctof
publ
iche
alth
insu
ranc
eel
igib
ility
onth
ere
latio
nshi
pb/
nfa
mily
inco
me
and
child
heal
thst
atus
and
onth
ere
latio
nshi
pb/
nfa
mily
inco
me
and
doct
orvi
sitsi
nth
epr
evio
usye
ar.A
lsote
sted
whe
ther
the
rela
tions
hip
b/n
inco
me
and
outc
omes
chan
ged
over
time
with
inag
egr
oups
byru
nnin
gse
para
tere
gres
sions
for
4ag
egr
oups
:0-
3,4-
8,9-
12,13
-17.
Fina
lly,es
timat
edre
latio
nshi
pb/
nhe
alth
outc
omes
and
the
frac
tion
ofch
ildre
nel
igib
lein
child
’sbi
rth
coho
rtin
the
child
’scu
rren
tsta
teof
resid
ence
for
child
ren
aged
9-17
.C
ontr
olle
dfo
rfa
mily
back
grou
ndva
riab
lesa
ndye
aran
dst
ate
fixe
deff
ects
.
For
child
ren
aged
9-12
,th
eco
effici
ento
nin
com
ein
am
odel
ofhe
alth
stat
usde
clin
edby
20%
over
2000
-200
5.Fo
rch
ildre
nag
ed13
-17,
itde
clin
edby
18%
over
1996
-200
0an
dby
25%
over
2000
-05.
For
child
ren
0-3/
4-8/
9-12
the
coeffi
cien
ton
inco
me
ina
mod
elof
doct
orvi
sitsd
eclin
edby
64%
/62%
/50%
betw
een
2000
-200
5.Fo
rch
ildre
nag
ed13
-17
ther
ew
asno
signi
fica
ntch
ange
inth
ein
com
eco
effici
ent.
Sign
ifica
ntde
clin
esin
the
inco
me
coeffi
cien
tove
r19
91-1
995
and
1996
-200
0fo
rch
ildre
nag
ed0-
3an
d4-
8as
wel
l.N
ost
atist
ical
lysig
nifica
ntim
pact
ofco
ntem
pora
neou
she
alth
insu
ranc
eel
igib
ility
onch
ildhe
alth
stat
us.
A10
0pp
incr
ease
inth
efr
actio
nel
igib
leat
age
3w
ould
redu
ceth
epr
obab
ility
that
the
aver
age
child
aged
9-17
isin
less
than
exce
llent
heal
thby
11%
.A
100
ppin
crea
sein
the
frac
tion
elig
ible
atag
es1
and
2w
ould
redu
ceth
epr
obab
ility
that
the
aver
age
child
aged
9-17
had
nodo
ctor
visit
inth
epa
stye
arby
41%
.
(con
tinue
don
next
page
)
Human Capital Development before Age Five 1463
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
The
impa
ctof
child
ren’
spu
blic
heal
thin
sura
nce
expa
nsio
nson
educ
atio
nal
outc
omes
(Lev
ine
and
Scha
nzen
bach
,20
09).
Dat
aon
stat
e-le
vela
vera
gesc
aled
test
scor
esfr
omth
eN
atio
nalA
sses
smen
tof
Edu
catio
nalP
rogr
essf
or19
90to
2003
.(n=
431)
.D
ata
for
simul
ated
inst
rum
ents
from
Mar
chC
PS.D
ata
onch
ildhe
alth
from
Vita
lSta
tistic
sfor
1984
-200
3(n=
1020
).
Exa
min
eth
eim
pact
ofpu
blic
heal
thin
sura
nce
atbi
rth
on4t
han
d8t
hgr
ade
read
ing
and
mat
hte
stsc
ores
.D
iffer
ence
-in-
differ
ence
-in-
differ
ence
usin
gcr
oss-
stat
eva
riat
ion
over
time
and
acro
ssag
esin
child
ren’
shea
lthin
sura
nce
elig
ibili
tydu
eto
Med
icai
dan
dSC
HIP
expa
nsio
ns.In
stru
men
ted
for
publ
iche
alth
insu
ranc
eel
igib
ility
usin
gth
esim
ulat
edfr
actio
nel
igib
le(a
sin
Cur
rie
and
Gru
ber
(199
6)se
eab
ove)
.To
test
whe
ther
chan
ges
ined
ucat
iona
lout
com
esar
edu
eto
impr
ovem
ents
inhe
alth
stat
usor
toad
ditio
nalh
ouse
hold
inco
me
gene
rate
dif
the
avai
labi
lity
ofpu
blic
heal
thin
sura
nce
crow
dsou
tpri
vate
heal
thin
sura
nce,
they
estim
ated
the
dire
ctim
pact
ofhe
alth
atbi
rth
oned
ucat
iona
lout
com
es.C
ontr
olle
dfo
rst
ate
and
year
fixe
deff
ects
and
stat
e-sp
ecifi
ctim
etr
ends
.
A50
ppin
crea
sein
elig
ibili
tyat
birt
hin
crea
sed
read
ing
test
scor
esby
0.09
SDre
lativ
eto
4th
and
8th
grad
eco
mbi
ned
mea
nsc
aled
scor
e.N
oeff
ecto
nm
ath
test
scor
es.
Exp
ansio
nsto
publ
iche
alth
insu
ranc
eel
igib
ility
atbi
rth
asso
ciat
edw
/1.
6%re
duct
ion
inlo
wbi
rth
wei
ghtr
ate
for
who
lesa
mpl
ean
d6.
7%re
duct
ion
inin
fant
mor
talit
yra
tes
amon
gw
omen
with
atle
asta
high
scho
olde
gree
.A
50%
incr
ease
inlo
wbi
rth
wei
ght(
infa
ntm
orta
lity)
rate
wou
ldde
crea
sere
adin
gte
stsc
ores
by0.
12SD
(.07S
D).
(con
tinue
don
next
page
)
1464 Douglas Almond and Janet Currie
Table13
(con
tinue
d)Stud
yan
dda
taStud
yde
sign
Results
Publ
icin
sura
nce,
crow
d-ou
tand
heal
thca
reut
iliza
tion
(Koc
h,20
09).
Dat
afr
omth
em
edic
alex
pend
iture
pane
lsur
vey.
Focu
son
subs
ampl
eof
child
ren<
18yr
swho
sefa
mili
esha
vein
com
esb/
n50
%an
d40
0%of
pove
rty
line.
(n=
32,6
09).
Dat
aon
publ
icin
sura
nce
reim
burs
emen
trat
esfr
omth
eA
mer
ican
Aca
dem
yof
Pedi
atri
csfo
r19
98-2
001.
Reg
ress
ion
disc
ontin
uity
desig
nus
ing
fam
ilyin
com
ecu
t-off
stha
tdet
erm
ine
elig
ibili
tyfo
rM
edic
aid
and
SCH
IP.To
dist
ingu
ishb/
nch
ildre
nw
hoar
eel
igib
lefo
rpu
blic
insu
ranc
ew
/out
acce
ssto
priv
ate
insu
ranc
ean
dth
ose
who
sepr
ivat
ein
sura
nce
may
becr
owde
d-ou
tdue
toex
pans
ions
inpu
blic
heal
thin
sura
nce,
cons
truc
ted
am
easu
reof
priv
ate-
insu
ranc
e“o
ffer
”(=
1if
any
ofth
efo
llow
ing
hold
s:ch
ildis
priv
atel
yin
sure
dat
the
time
ofth
eel
igib
ility
mea
sure
men
tor
afa
mily
mem
ber
ispr
ivat
ely
empl
oyed
atth
etim
ean
dei
ther
hasi
nsur
ance
thro
ugh
the
job
ortu
rnsd
own
the
insu
ranc
eat
the
job)
.E
stim
ated
impa
ctso
fpub
liche
alth
insu
ranc
eel
igib
ility
onm
easu
reso
fcar
eut
iliza
tion
(doc
tor’s
visit
s,et
c.)a
swel
laso
npr
ivat
ein
sura
nce
take
-up.
Also
estim
ated
the
differ
entia
leffec
tofe
ligib
ility
onth
ose
w/
and
w/o
utpr
ivat
ehe
alth
insu
ranc
eby
inte
ract
ing
the
elig
ibili
tyva
riab
lew
ithth
e“o
ffer
”va
riab
le.C
ontr
olle
dfo
rba
ckgr
ound
char
acte
rist
icsa
swel
lass
tate
and
year
fixe
deff
ects
.A
lsolo
oked
atim
pact
sofd
iffer
entia
lrei
mbu
rsem
entr
ates
acro
ssst
ates
and
year
son
acce
ssto
care
for
the
publ
icly
insu
red.
Publ
icin
sura
nce
elig
ibili
tyin
crea
sesp
ublic
insu
ranc
eco
vera
geby
21%
and
decr
ease
spri
vate
insu
ranc
eco
vera
geby
24%
.So
over
alld
ecre
ase
inin
sura
nce
cove
rage
ofab
out
4.8%
.E
ligib
ility
decr
ease
sthe
num
ber
ofoffi
cevi
sits/
self-
repo
rted
heal
thst
atus
/tot
alex
pend
iture
sby
12%
/87%
/23%
.In
crea
sest
henu
mbe
rof
ER
visit
sby
18%
.E
ligib
ility
decr
ease
sthe
num
ber
ofoffi
cevi
sitsb
y18
%fo
rth
ose
with
outs
ide
priv
ate
heal
thin
sura
nce.
No
stat
istic
ally
signi
fica
nteff
ects
onE
Rvi
sits,
pres
crip
tions
,or
refills
.C
hild
ren
just
elig
ible
for
publ
iche
alth
insu
ranc
ear
e4%
less
likel
yto
have
aus
uals
ourc
eof
care
and
4%m
ore
likel
yto
gow
ithou
tcar
e.N
ost
atist
ical
lysig
nifica
nteff
ects
onho
spita
lor
ER
visit
s.E
ligib
ility
incr
ease
schi
ldre
n’sB
MI
by2%
.Pa
rent
sofj
ust
elig
ible
child
ren
are
5%le
sslik
ely
tobe
give
nad
vice
abou
tth
eir
child
’sea
ting
heal
thfu
lly.N
osig
nifica
nteff
ects
onas
thm
adi
agno
sis.C
ondi
tiona
lon
taki
ngm
edic
atio
nfo
ras
thm
a,el
igib
ility
incr
ease
spro
babi
lity
ofta
king
inha
led
ster
oids
(gol
dst
anda
rdtr
eatm
ent)
by32
%.A
$4in
crea
sein
reim
burs
emen
tfor
office
visit
slea
dsto
a5%
incr
ease
inth
enu
mbe
rof
office
visit
s.
Perc
entc
hang
es(d
enot
edby
%in
stea
dof
“pp”
)are
repo
rted
rela
tive
toth
em
ean.
Human Capital Development before Age Five 1465
birth weight, but reduced infant mortality. This result suggests that among women ofsomewhat higher income levels, the expansions did not improve health at birth, but mayhave increased access to life-saving technologies after birth. Currie and Grogger (2002)focus on bureaucratic obstacles to obtaining health insurance by looking at contractionsof welfare (women cut from the rolls lost automatic eligibility for Medicaid) as well asoutreach measures undertaken by different states. They find that changes that reducedbarriers to enrollment increased use of prenatal care and had positive effects on infanthealth outcomes.
Baldwin et al. (1998) use individual-level data and compare expansions inWashington, which included enhanced prenatal care services, to expansions in Coloradowhich did not, in a difference-in-differences design. They find reductions in lowbirth weight among medically high risk infants in Washington. Dubay et al. (2001)conduct a difference-in-difference investigation comparing the outcomes of high andlow socioeconomic status women in the 1980-1986 period and in the 1986-1993 period.
They find overall improvements in the use of prenatal care for low SES women, but findimprovements in birth weight only for some groups of white women. However, thisdesign does not really focus on health insurance per se, since the estimates will be affectedby any other changes in health care markets between the two periods that had differentialeffects by SES.
Studies of the effects of health insurance expansions on children often examinepreventable hospitalizations (also called ambulatory care sensitive hospitalizations).The idea is that certain conditions, such as childhood asthma, should not result inhospitalizations if they are properly treated on an outpatient basis. Hence, hospitalizationsfor these conditions are inefficient and indicate that children are receiving inadequatepreventive care. Kaestner et al. (2001) use a difference-in-differences design comparinglow income and other children before and after Medicaid expansions. They findreductions in preventable hospitalizations of 10% to 20%.
Aizer (2003) examines a California outreach program that increased child enrollmentsinto Medicaid. She finds that an increase in enrollments of 1000 reduces hospitalizationsby 3.26. Dafny and Gruber (2005) use a design similar to Currie and Gruber in whichactual individual eligibility is instrumented using a “simulated eligibility measure” whichis an index of the generosity of the Medicaid program in the state. The reason foradopting instrumental variables estimation is that eligibility for Medicaid is determinedby endogenous variables such as parental labor supply. They find that Medicaid eligibilityincreased hospitalizations overall. However, there was no statistically significant increasein avoidable hospitalizations, suggesting that the increase was mainly due to childrenwith unavoidable conditions gaining greater access to care. They also found increasesin the probability of receiving a procedure and reductions in length of stay, suggestingthat children who were hospitalized were receiving more aggressive care and that it mayhave improved their outcomes.
1466 Douglas Almond and Janet Currie
One difficulty with studying child health is that health today is affected by pastinvestments, including health insurance at younger ages. Currie et al. (2008) thereforecompare the health effects of contemporaneous eligibility for health insurance amongolder children to the effect of having been eligible since birth. They find thatcontemporaneous health insurance coverage has little effect on health status but thateligibility from birth is protective. Levine and Schanzenbach (2009) link health insuranceeligibility at birth to 4th and 8th grade scores on the National Assessment of EducationalProgress. They find that a 50% point increase in eligibility at birth is associated with asmall but significant gain on reading scores at both grades, though there is no effect onmath scores. A difficulty with both studies is that neither income at birth nor state ofbirth are directly observed in the cross sectional data sets that they use, so they must beimputed using current income and state of birth.
Another area of research focuses on the quality of care provided by public healthinsurance programs. Analysis of this issue is complicated by the possibility that expansionsof public insurance cause people to lose private health insurance coverage, a phenomenadubbed “crowdout” (Cutler and Gruber 1996; Dubay and Kenney, 1997; Card andShore-Shepard, 2004; Gruber and Simon, 2008). If the private insurance that is lost (ordropped) in response to expansions of public insurance is of superior quality to the privateinsurance, then people’s health may suffer. Koch (2009) concludes that recent expansionsof public health insurance to children at higher income levels have reduced access todoctor’s office visits and increased reliance on emergency rooms. He also shows someevidence consistent with the idea that this is because children are being crowded out ofsuperior (but obviously more expensive) coverage from private heath insurance plans. Infact, it is quite possible that crowding out has increased over time as the public has becomefamiliar with public health insurance plans for children and private health insurance costshave continued to escalate.
Medicaid managed care has also been shown (in at least some cases) to reduce thequality of care. Sloan et al. (2001) conduct a difference-in-difference analysis of Tennesseeand North Carolina before and after Tennessee switched its Medicaid patients to managedcare. They find that use of prenatal care and birth outcomes deteriorated in Tennesseeafter the switch. Aizer et al. (2007) examine data from California, where Medicaidmanaged care was adopted on a county-by-county basis. They also find that compulsorymanaged care had a negative impact on use of prenatal care and birth outcomes. Thismay be because the California Medicaid managed care program “carved out” care forsick newborns–that is, they were covered by a state fund rather than by the managed carecompanies so that companies had little incentive to take actions to improve newbornhealth.
In summary, health insurance matters for children’s outcomes. But quality of care alsomatters. And it is important to remember that for most children threats to health and
Human Capital Development before Age Five 1467
well being come from sources such as injuries, poor nutrition, and toxins rather thanonly from lack of access to medical care.
6. DISCUSSION AND CONCLUSIONS
There has been an explosion of research into the early determinants of human capitaldevelopment over the past 10 years. The work surveyed in this chapter conclusively showsthat events before five years old can have large long term impacts on adult outcomes. Itis striking that child and family characteristics measured at school entry do as much toexplain future outcomes as factors that labor economists have more traditionally focusedon, such as years of education. Yet evidence for long term effects of early insults shouldnot be a cause of pessimism. While children can be permanently damaged at this age, thedamage can be remediated. The picture that emerges is one of vulnerability but also ofresilience.
Since early childhood is a new area of research for economists, there remain manyunanswered questions. One major question implicit in the structure of this chapteris whether it will ever be possible to estimate human capital production functions.The opening sections of this chapter showed that the production function paradigmprovides an extremely useful way of thinking about the problem, and in particular,that it highlights the importance of actions taken by parents and others in exacerbatingor mitigating the effects of random shocks. However, to actually estimate the impliedproduction functions would place huge demands on the data, demands that are unlikelyto be met in practice.16 Hence, the evidence we have surveyed is largely reduced form.
A second major question is whether shocks at certain key ages matter more thanothers? Much has been written about “critical periods”. The idea is that certainfunctionalities must be acquired at a particular point in life, and if they are not acquiredat that point, they will either not be acquired at all, or will not be acquired properly.There is to date little hard evidence of critical periods in humans. However, the evidencediscussed above certainly suggests that the period while children are in utero is one ofthe most important to their later development. This has important implications for thetiming of social interventions designed to mitigate harms—it may be that interventionsshould be targeted at pregnant women and/or women of child bearing age in addition toyoung children. But there is insufficient evidence at present to be able to say that insults orinterventions at for example zero to one are likely to be more effective than interventionsat age three or four. Moreover, the cohort designs used to establish the importance of thefetal period can tell us more about the comparison between the fetal period and the earlypost natal period than they do about the comparison between the fetal period and sayexposures at age five.
16 See however Appendix D for one thought about how a production function might be estimated.
1468 Douglas Almond and Janet Currie
A related question is whether some types of shocks matter more than others. Thischapter surveyed many different types of shocks including exposure to disease, inadequatenutrition, exposure to pollution, injuries, maternal mental health problems, maternalsmoking, and maltreatment. However, given that studies of the effects of these shocksrely on different populations, time periods, and methods, it is difficult to get any sense ofwhether one type of shock is more of a threat to human capital development than anyother. Similarly, while it is clear that shocks to health have long-term effects on domainssuch as education and earnings, it is not clear whether health shocks have direct effectson cognition or learning, or whether they act mainly by affecting future health.
Several studies we reviewed suggested that both shocks and interventions can havedifferent long-term effects on males and females. But these findings are too new forus to be able to predict when this difference will occur, and we have virtually noevidence about why it occurs. One possibility is that gender differences are biological.For example, boys may be less robust than girls so that the same health shock can “cull”boys while girls survive (e.g., see Kraemer (2000); Almond and Mazumder (2008)). Inthis case, average health of male survivors might be better than that of female survivors.Alternatively, gender differences could reflect differential parental or societal responses toshocks inspired by son preferences or by beliefs about biological gender differences.17
Finally, given all of this evidence of long-term effects of early life outcomes,what is the least costly way to intervene to improve outcomes? This is still an openquestion and our knowledge of the types of programs that are effective (and why)is evolving rapidly. For example, until recently, there was little evidence that incometransfers had much effect, so it was easy to surmise that in-kind programs were a moreeffective way to improve child outcomes. Recent evidence that cash transfers are indeedeffective should cause a re-evaluation of the received wisdom on this point, given theinefficiencies involved in providing transfers in-kind. Similarly, the large literature aboutnegative effects of maternal employment in the early years is thrown into question byrecent studies showing that large changes in maternity leave policies affected maternalemployment without having any detectable impact on long-term child outcomes. Thisrapid development in our knowledge makes the study of human capital developmentbefore five an exciting frontier for research in labor economics.
APPENDIX AThe following acronyms are used in this chapter:
AFDC = Aid to Families with Dependent Children
BCS = British Birth Cohort Study of 1970.
17 For example, advances in ultrasound technology could have changed the average human capital endowments of boysand girls by allowing parents who prefer sons to invest differentially prenatally (and not only by allowing them to abortfemale fetuses). See Lhila and Simon (2008) for recent work on this topic.
Human Capital Development before Age Five 1469
BPI = Behavioral Problems Index
BW = birth weight
CESD = Center for Epidemiological Depression scale
CCT = Conditional Cash Transfer
COHS = County Organized Health System
CPS = Current Population Survey
DDST = Denver Developmental Screening Test
ECLS-B = Early Childhood Longitudinal Study—Birth Cohort
ECLS-K = Early Childhood Longitudinal Study—Kindergarten Class of 1998-1999
EITC = Earned Income Tax Credit
EPA =US Environmental Protection Agency
FSP = Food Stamp Program
HAZ =Height for age z-score
HOME =Home Observation for Measurement of the Environment Score
IHDP = Infant Health and Development Project
IPUMS = Integrated Public-Use Microdata Samples of the US Census
IV = instrumental variables
LBW = Low Birth Weight (birth weight less than 2500 g)
MMC =Medicaid Managed Care
NBER =National Bureau of Economic Research
NCDS =National Child Development Survey (1958 British Birth Cohort)
NELS =National Education Longitudinal Study
NHANES =National Health and Nutrition Examination Survey
NHDS =National Hospital Discharge Survey
NLSY =National Longitudinal Survey of Youth, 1979 cohort
NLSY-Child = Children of the NLSY 1979 cohort
1470 Douglas Almond and Janet Currie
NLSCY =National Longitudinal Survey of Children and Youth (Canadian)
PIAT = Peabody Individual Achievement Test
pp = percentage points
PPVT = Peabody Picture Vocabulary Test
NSLP =National School Lunch Program
OLS =Ordinary Least Squares
PNM = Post Neonatal Mortality (death after 28 days and before 1 year)
PSID = Panel Study of Income Dynamics
RDA =Recommended Dietary Allowance
REIS =Regional Economic Information System
SCHIP = State Child Health Insurance Program
SD = Standard Deviation
SES = Socio-economic Status
SGA = Small for Gestational Age
SIPP = Survey of Income and Program Participation
SNAP = Supplemental Nutrition Assistance Program (formerly, Food Stamps)
TSIV = Two Sample Instrumental Variables
TVIP = Spanish-speaking version of the Peabody Picture Vocabulary Test
TSP = Total Suspended Particles
USDA =US Department of Agriculture
VSDN = Vital Statistics Detailed Natality files (birth certificate data for US)
WIC = Special Supplemental Nutrition Program for Women, Infants, and Children
WPPSI =Wechsler Preschool and Primary Scale of Intelligence
Human Capital Development before Age Five 1471
APPENDIX BHuman capital of a child is produced with a CES technology:
h = A[γ ( I1 + µg)
φ+ (1− γ )Iφ2
]1/φ, (9)
where µg is an exogenous shock to (predetermined) period 1 investments. Parents valuetheir consumption and the human capital of their child:
Up = U (C, h) = B[θ(C)ϕ + (1− θ)hϕ
]1/ϕ, (10)
and have the budget constraint:
I1 + I2 + C = y.
Absent discounting, the marginal utility from consuming equals the marginal utilityfrom investing:
δU
δC∗=δU
δh
δh
δ I ∗2.
θCϕ−1= (1− θ)hϕ−1 A[· · ·]
1φ−1(1− γ )I ∗φ−1
2 (11)
θ(y − I1 − I ∗2 )ϕ−1= (1− θ)Aϕ−1
[· · ·]ϕ−1φ A[· · ·]
1φ−1(1− γ )I ∗φ−1
2 (12)
θ(y − I1 − I ∗2 )ϕ−1= (1− θ)(1− γ )Aϕ [· · ·]
ϕ−φφ I ∗φ−1
2 (13)
G(ug, I ∗2 ) ≡ θ(y − I1 − I ∗2 )ϕ−1− (1− θ)(1− γ )Aϕ [· · ·]
ϕ−φφ I ∗φ−1
2 = 0. (14)
δ I ∗2δµg= −
δGδµg
δGδ I ∗2
=
a(I ∗2 )φ−1[· · ·]
ϕ−2φφ
(ϕ−φφ
)γφ( I1 + µg)
φ−1
−(ϕ − 1)θ(y − I1 − I ∗2 )ϕ−2− a
[[· · ·]
ϕ−φφ (φ − 1)I ∗φ−2
2 +ϕ−φφ[· · ·]
ϕ−2φφ φ(1− γ )I ∗φ−1
2 I ∗φ−12
] ,(15)
using the implicit function theorem and defining a to be (1− θ)(1− γ )Aϕ ≥ 0.
=(ϕ − φ)a(I ∗2 )
φ−1[· · ·]
ϕ−2φφ γ ( I1 + µg)
φ−1
(1− ϕ)θ(y − I1 − I ∗2 )ϕ−2+ a[· · ·]
ϕ−φφ I ∗φ−2
2
[(1− φ)+ (ϕ − φ)(1− γ )I ∗φ2 /[· · ·]
] . (16)
For ϕ > φ, (16) is positive, so negative shocks in the first period should be reinforced.Accommodation through preferences (i.e., more consumption and less investment,which lowers h in addition to that caused by µg) is optimal.
1472 Douglas Almond and Janet Currie
APPENDIX C
Sibling a has human capital ha , which is affected by a period 1 investment shock of µg:
ha = A[γ ( I1a + µg)
φ+ (1− γ )Iφ2a
]1/φ. (17)
Sibling b does not experience a shock to first period investments:
hb = B[γ Iφ1b + (1− γ )I
φ
2b
]1/φ. (18)
Assume further that first period investments do not distinguish between the twosiblings (absent the shock experienced by sibling a):
I1a = I1b = I1.
Parents have Cobb-Douglas utility that cares only about the human capital of theirtwo children:
Up = U (ha, hb) = (1− α)log ha + αlog hb. (19)
The parents exhaust their budget on investments in their children:
y = 2 I1 + I2a + I2b.
Denoting Y = y − 2 I1 as the budget for second period investments, I2b = Y − I2a.
To maximize utility, the marginal utilities from investing in siblings a and b should beequal:
δUp
δha
δha
δ I2a=δUp
δhb
δhb
δ I2b(1− α
ha
)A
φ
[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
](1/φ)−1φ(1− γ )Iφ−1
2a
=
(α
hb
)B
φ
[γ ( I1)
φ+ (1− γ )Iφ2b
](1/φ)−1φ(1− γ )Iφ−1
2b (20)
(1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1Iφ−12a
= α[γ ( I1)
φ+ (1− γ )Iφ2b
]−1Iφ−12b (21)
G(µg, I2a) ≡ (1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1Iφ−12a
−α[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−1
(Y − I2a)φ−1= 0 (22)
Human Capital Development before Age Five 1473
using budget the constraint: I2b = Y − I2a . By the implicit function theorem:
δ I ∗2a
δµg=
−δGδµg
δGδ I ∗2a
(23)
δG
δµg= −(1− α)Iφ−1
2a
[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−2φγ ( I1 + µg)
φ−1 (24)
⇒ signum
[−δG
δµg
]= signum[φ].
δG
δ I ∗2a= (1− α)
[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1(φ − 1)Iφ−2
2a
+ (−1)(1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−2φ(1− γ )Iφ−1
2a Iφ−12a
−
[α[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−1
(φ − 1)(Y − I2a)φ−2(−1)
]−
[(−1)α
[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−2
× φ(1− γ )(Y − I2a)φ−1(−1)(Y − I2a)
φ−1]
(25)
= (1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1(φ − 1)Iφ−2
2a
− (1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−2φ(1− γ )I 2φ−2
2a
+
[α[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−1
(φ − 1)(Y − I2a)φ−2
]−
[α[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−2
φ(1− γ )(Y − I2a)2φ−2
](26)
= (1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1
× Iφ−22a
[(φ − 1)−
φ(1− γ )Iφ2a
γ ( I1 + µg)φ + (1− γ )Iφ
2a
]+α
[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−1
(Y − I2a)φ−2
×
[(φ − 1)−
φ(1− γ )(Y − I2a)φ
γ ( I1)φ + (1− γ )(Y − I2a)φ
](27)
= (1− α)[γ ( I1 + µg)
φ+ (1− γ )Iφ2a
]−1Iφ−22a
×
[φ
(1−
(1− γ )Iφ2a
γ ( I1 + µg)φ + (1− γ )Iφ
2a
)− 1
]
1474 Douglas Almond and Janet Currie
+α[γ ( I1)
φ+ (1− γ )(Y − I2a)
φ]−1
(Y − I2a)φ−2
×
[φ
(1−
(1− γ )(Y − I2a)φ
γ ( I1)φ + (1− γ )(Y − I2a)φ
)− 1
]. (28)
Because φ ≤ 1 and:
(1− γ )Iφ2a
γ ( I1 + µg)φ + (1− γ )Iφ
2a
< 1 and(1− γ )(Y − I2a)
φ
γ ( I1)φ + (1− γ )(Y − I2a)φ< 1,
Eq. (28) is always negative. Therefore:
signum
[δ I ∗2a
δµg
]= − signum[φ].
We consider three cases for the substitutability of period 1 and period 2 investments(as captured by φ):1. Good Substitutability Between Periods 1 and 2 When φ > 0, the optimal I2a
moves in the opposite direction from µg and parents should compensate a negativeshock to child a by reducing second period investments in child b. Intuitively, itis easier to substitute though the production function for human capital than it isthrough the Cobb-Douglas utility function.
2. Cobb-Douglas Substitutability Between Periods 1 and 2 For φ = 0, theelasticity of substitution between periods is the same as the elasticity of substitutionin preferences between the children (both Cobb-Douglas). Here, there is no winning
investment response to the shock to child a, i.e.δ I ∗2aδµg= 0, so period 2 investments
should be left unchanged.
3. Poor Substitutability Between Periods 1 and 2 For φ < 0, it is difficult to repairdamage from a negative µg shock in the second period, so the return to period 2investments in sibling a is below that for sibling b. Therefore, parents should reinforcethe first period shock by allocating second period investments away from sibling a.
Importantly, the direction of these investment responses did not depend on α, therelative weight parents place in their utility function on the human capital of child aversus child b. Favoring the human capital formation of a particular child—even thechild that experiences the negative endowment shock—does not affect the directionof the optimal investment response. Nor do differences in the “overall” productivityof the child, i.e. efficiency parameter A 6= B in Eqs (17) and (18), alter the directionof the optimal investment response to µg. Thus, empirical evidence suggesting eitherreinforcing or compensating investments within the family does not reveal information
Human Capital Development before Age Five 1475
on parental preferences absent additional information on the production function forhuman capital.
APPENDIX DIn general, we need to observe the baseline investments I1 and I2 to estimate parametersof the production function φ and γ . However, nearly all datasets with measures ofhuman capital h and an observable investment shock µg lack measures of human capitalinvestments I1 and I2. We can still make progress in estimating parameters of theproduction function despite not observing I1 and I2, so long as we expect baselineinvestment levels to be similar: I1 ∼= I2. For µg = µ
′g,18 Eq. (4) reduces to:
1− γγ
. (29)
That is, we can observe damage to h from the shock µ′g in second period investmentsrelative to the damage from the first period shock µg, which isolates γ (while remainingsilent on the magnitude of φ).
With an estimate of γ in hand, we can then estimate φ by using the total derivativein investment shocks, i.e. how human capital changes as we change both first and secondperiod investments (by an equal amount). In an overlapping generations framework, thiswould require a shock lasting two childhood periods (or longer), and “half-exposed”cohorts on either end of the shock. Damage to the fully exposed cohort relative to thecohort exposed in period 1 alone is:
=1γ
[γ ( I1 + µg)
φ+ (1− γ )( I1 + µg)
φ
γ ( I1 + µg)φ + (1− γ ) Iφ
1
] 1−φφ
, (30)
using the assumption that I1 = I2 and µg = µ′g,
=1γ
[( I1 + µg)
φ
γ ( I1 + µg)φ + (1− γ ) Iφ
1
] 1−φφ
, (31)
=1γ
1
γ + (1− γ ) I1I1+µg
φ
1−φφ
, (32)
=1γ
[γ + (1− γ )
(1+
µg
I1
)−φ] φ−1φ
. (33)
18 The assumption that µg = µ′g simplifies the algebra, but φ and γ can still be estimated for µg 6= µ
′g so long as µg
and µ′g are observed. We thank Christine Pal for pointing this out.
1476 Douglas Almond and Janet Currie
Similarly, damage to the “doubly exposed” cohort relative to that experiencing ashock in just the second childhood period is:
=1
1− γ
[(1− γ )+ γ
(1+
µg
I1
)−φ] φ−1φ
, (34)
Eqs (33) and (34) constitute two equations in the two unknowns I1 and φ, with γknown from Eq. (29).
REFERENCES2001. An overstretched hypothesis? The Lancet 357 (9254), 405.Aguero, Jorge M., Carter, Michael R., Woolard, Ingrd, 2009. The impact of unconditional cash transfers on
nutrition: the South African child support grant. Manuscript, February.Aizer, Anna, 2003. Low take-up in Medicaid: does outreach matter and for whom? American Economic
Review Papers and Proceedings 93 (2), 238–241.Aizer, Anna, 2007. Public health insurance, program take-up and child health. The Review of Economics
and Statistics 89 (3), 400–415.Aizer, Anna, Currie, Janet, Moretti, Enrico, 2007. Does managed care hurt health? Evidence from Medicaid
mothers. The Review of Economics and Statistics 89 (3), 385–399.Aizer, Anna, Stroud, Laura, 2009. Education, medical knowledge and the evolution of disparities in health.
manuscript, Brown University, September.Aizer, Anna, Stroud, Laura, Buka, Stephen, 2009. Maternal stress and child well-being: evidence from
siblings. manuscript. Brown University, February.Almond, Douglas, 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.Almond, Douglas, Chay, Kenneth Y., Lee, David S., 2005. The costs of low birth weight. The Quarterly
Journal of Economics 120 (3), 1031–1084.Almond, Douglas, Edlund, Lena, Palme, Marten, 2009. Chernobyl’s subclinical legacy: prenatal exposure to
radioactive fallout and school outcomes in sweden. Quarterly Journal of Economics 124 (4), 1729–1772.URL http://www.mitpressjournals.org/doi/abs/10.1162/qjec.2009.124.4.1729.
Almond, Douglas, Hoynes, Hilary W., Schanzenbach, Diane Whitmore, 2010. Inside the war on poverty:the impact of food stamps on birth outcomes. The Review of Economics and Statistics (forthcoming).
Almond, Douglas, Mazumder, Bhashkar, 2005. The 1918 influenza pandemic and subsequent healthoutcomes: an analysis of SIPP data. American Ecnomic Review, Papers and Proceedings 95 (2).
Almond, Douglas, Mazumder, Bhashkar, 2008. Health capital and the prenatal environment: the effect ofmaternal fasting during pregnancy. Working Paper 14428, National Bureau of Economic Research,October, URL http://www.nber.org/papers/w14428.
Anderson, Kathryn H., Foster, James E., Frisvold, David E., 2009. Investing in health: the long-term impactof Head Start on smoking. Working Paper 1465-7295, Western Economic Association International.
Anderson, Michael, 2008. Multiple inference and gender differences in the effects of early intervention: areevaluation of the Abecedarian, Perry Preschool, and Early Training projects. Journal of the AmericanStatistical Association 103 (484), 1481–1495.
Angrist, Joshua D., Pischke, Jorn-Steffen, 2009. Mostly Harmless Econometrics: an Empiricist’sCompanion. Princeton University Press, Princeton, New Jersey.
Aos, Steve, Lieb, Roxanne, Mayfield, Jim, Miller, Marna, Pennucci, Annie, 2004. Benefits and costs ofprevention and early intervention programs for youth. Technical appendix, Washington State Institutefor Public Policy, September.
Human Capital Development before Age Five 1477
Ashenfelter, Orley, Rouse, Cecilia, 1998. Income, schooling, and ability: evidence from a new sample ofidentical twins. Quarterly Journal of Economics 113 (1), 253–284.URL http://www.mitpressjournals.org/doi/abs/10.1162/003355398555577.
Baker, Kevin, 2008. Do cash transfer programs improve infant health: evidence from the 1993 expansion ofthe earned income tax credit. Manuscript, University of Notre Dame.
Baker, Michael, Gruber, Jonathan, Milligan, Kevin, 2008. Universal child care, maternal labor supply, andfamily well-being. Journal of Political Economy 116 (4), 709–745.URL http://www.journals.uchicago.edu/doi/abs/10.1086/591908.
Baker, Michael, Milligan, Kevin, 2010. Evidence from maternity leave expansions of the impact of maternalcare on early child development. Journal of Human Resources 45 (1), 1–32.
Baldwin, Laura-Mae, Larson, Eric H., Connell, Frederick A., Nordlund, Daniel, Cain, Kevin C., Cawthon,Mary L., Byrns, Patricia, Rosenblatt, Roger A., 1998. The effect of expanding Medicaid prenatal serviceson birth outcomes. American Journal of Public Health 88 (11), 1623–1629.
Banerjee, Abhijit, Duflo, Esther, Postel-Vinay, Gilles, Watts, Timothy, 2010. Long run health impacts ofincome shocks: wine and phylloxera in 19th century France. The Review of Economics and Statistics(forthcoming).
Banzhaf, Spencer, Walsh, Randall P., 2008. Do people vote with their feet? an empirical test of tiebout.American Economic Review 98 (3), 843–863.
Barker, D.J.P. (Ed.), 1992. Fetal and infant origins of adult disease. British Medical Journal.Barnett, W.S., Masse, Leonard N., 2007. Comparative cost-benefit analysis of the Abecedarian program and
its policy implications. Economics of Education Review 26, 113–125.Barnett, Steven, Belfield, Clive R., Nores, Milagros, Schweinhart, Lawrence, 2006. The High/Scope Perry
Preschool program: cost-benefit analysis using data from the age-40 follow-up. The Journal of HumanResources 41 (1), 162–190.
Barreca, Alan, 2010. The long-term economic impact of in utero and postnatal exposure to malaria. Journalof Human Resources 45 (4), 865–892.
Basiotis, Peter, Kramer-LeBlanc, Carol S., Kennedy, Eileen T., 1998. Maintaining nutrition security anddiet quality: the role of the Food Stamp program and WIC. Family Economics 11 (1-2), 4–16.
Baten, Jorg, Crayen, Dorothee, Voth, Joachim, 2007. Poor, hungry and stupid: numeracy and the impact ofhigh food prices in industrializing Britain, 1780-1850. Economics Working Papers 1120, Department ofEconomics and Business, Universitat Pompeu Fabra, October.URL http://ideas.repec.org/p/upf/upfgen/1120.html.
Baum, Charles L., 2008. The effects of food stamp benefits on weight gained by expectant mothers.Discussion Paper 1343-08, Institute for Research on Poverty, September.
Becker, Gary S., Tomes, Nigel, 1976. Child endowments and the quantity and quality of children. TheJournal of Political Economy 84 (4), S143–S162.
Behrman, Richard E., Butler, Adrienne S. (Eds.), 2007. Preterm Birth: Causes, Consequences, andPrevention. Institute of Medicine Report. National Academies Press.
Behrman, Jere R., Pollak, Robert A., Taubman, Paul, 1982. Parental preferences and provision for progeny.Journal of Political Economy 90 (1), 52–73.
Behrman, Jere R., Rosenzweig, Mark R., 2004. Returns to birthweight. The Review of Economics andStatistics 86 (2), 586–601.
Behrman, Jere R., Rosenzweig, Mark R., Taubman, Paul, 1994. Endowments and the allocation ofschooling in the family and in the marriage market: the twins experiment. Journal of Political Economy102 (6), 1131–1174.
Van Den Berg, Gerard J., Lindeboom, Maarten, Portrait, France, 2006. Economic conditions early in lifeand individual mortality. American Economic Review 96 (1), 290–302.
Van Den Berg, Gerard J., Lindeboom, Maarten, Portrait, France, 2010. Long-run longevity effects of anutritional shock early in life: the Dutch potato famine of 1846-1847. Journal of Health Economics (inpress).
Berger, Lawrence M., Hill, Jennifer, Waldfogel, Jane, 2005. Maternity leave, early maternal employment,and child health and development in the us. The Economic Journal 115, F29–F47.
Berger, Lawrence M., Paxson, Christina, Waldfogel, Jane, 2009. Income and child development. Childrenand Youth Services Review 31 (9), 978–989.
1478 Douglas Almond and Janet Currie
Bingley, Paul, Walker, Ian, 2007. There’s no such thing as a free lunch: altruistic parents and the responseof household food expenditures to nutrition program reforms. Discussion Paper 2007-19, UniversityCollege Dublin Geary Institute, June.
Bitler, Marianne, Currie, Janet, 2005. Does WIC work? the effect of WIC on pregnancy and birthoutcomes. Journal of Policy Analysis and Management 24 (1), 73–91.
Black, Maureen M., Cutts, Diana B., Frank, Deborah A., Geppert, Joni, Skalicky, Anne, Levenson, Suzette,Casey, Patrick H., Berkowitz, Carol, Zaldivar, Nievves, Cook, John T., Meyers, Alan F., Herren, Tim,2004. Special supplemental nutrition program for women, infants, and children participation and infants’growth and health: a multi-site surveillance study. Pediatrics 114 (1), 169–176.
Black, Sandra E., Devereux, Paul J., Salvanes, Kjell G., 2005. The more the merrier? The effect of familysize and birth order on children’s education. Quarterly Journal of Economics 120 (2), 669–700.
Black, Sandra E., Devereux, Paul J., Salvanes, Kjell G., 2007. From the cradle to the labor market? The effectof birth weight on adult outcomes. Quarterly Journal of Economics 122 (1), 409–439.
Blanden, Jo, Gregg, Paul, MacMillan, Lindsey, 2006. Explaining intergenerational income persistence: non-cognitive skills, ability and education. CMPO Working Paper Series 06/146, University of Bristol, TheCentre for Market and Public Organisation, April. URL www.bris.ac.uk/Depts/CMPO/.
Blau, David, Currie, Janet, 2006. Pre-school, day care, and after-school care: Who’s minding the kids? 2:1163–1278.
Bleakley, Hoyt, 2007. Disease amd development: evidence from hookworm eradication in the south. TheQuarterly Journal of Economics 122 (1), 73.
Blow, Laura, Walker, Ian, Zhu, Yu, 2005. Who benefits from child benefit. manuscript, February.Blundell, Richard, 2006. Earned income tax credit policies: impact and optimality: the Adam Smith
lecture, 2005. Labour Economics (ISSN: 0927-5371) 13 (4), 423–443. European Association of LabourEconomists, 2nd World Conference SOLE/EALE, Fairmont Hotel San Francisco, USA, 2-5 June 2005.
Boyer, George R., 1990. An Economic History of the English Poor Law. CUP, Cambridge, 1750–1850.Bozzoli, Carlos, Deaton, Angus S., Quintana-Domeque, Climent, 2010. Child mortality, income and adult
height. NBER Working Paper No. 12966, March. Demography (forthcoming).Brooks-Gunn, Jeanne, Markham, Lisa B., 2005. The contribution of parenting to ethnic and racial gaps in
school readiness. The Future of Children 15 (1), 139–168.Butler, J.S., Raymond, Jennie E., 1996. The effect of the Food Stamp program on nutrient intake. Economic
Inquiry 34, 781–798.Card, David, Shore-Sheppard, Laura, 2004. Using discontinuous eligibility rules to identify the effects of
the federal Medicaid expansions on low-income children. Review of Economics and Statistics 86 (3),752–766.
Carneiro, Pedro, Ginja, Rita, 2008. Preventing behavior problems in childhood and adolescence: evidencefrom Head Start. manuscript, March.
Cascio, Elizabeth U., 2009. Do investments in universal early education pay off? long-term effects ofintroducing kindergartens into public schools. Working Paper 14951, National Bureau of EconomicResearch, May.
Case, Anne, Fertig, Angela, Paxson, Christina, 2005. The lasting impact of childhood health andcircumstance. Journal of Health Economics 24, 365–389.
Case, Anne, Lee, Diana, Paxson, Christina, 2008. The income gradient in children’s health: a comment onCurrie, Shields, and Wheatley price. Journal of Health Economics 27 (3), 801–807.
Case, Anne, Lubotsky, Darren, Paxson, Christina, 2002. Economic status and health in childhood: theorigins of the gradient. American Economic Review 92 (5), 1308–1334.
Case, Anne, Paxson, Christina, 2008a. Height, health, and cognitive function in old age. AmericanEconomic Review: Papers & Proceedings 98 (2), 463–467.
Case, Anne, Paxson, Christina, 2008b. Stature and status: height, ability, and labor market outcomes. Journalof Political Economy 116 (3), 499–532.
Case, Anne, Paxson, Christina, 2009. Early life health and cognitive function in old age. American EconomicReview Papers and Proceedings 99 (2), 104–109.
Case, Anne, Paxson, Christina, 2010a. Causes and consequences of early life health. Working Paper 15637,National Bureau of Economic Research, January. URL http://www.nber.org/papers/w15637.
Human Capital Development before Age Five 1479
Case, Anne, Paxson, Christina, 2010b. The long reach of childhood health and circumstance: evidence fromthe Whitehall II study. Working Paper 15640, National Bureau of Economic Research, January. URLhttp://www.nber.org/papers/w15640.
Chatterji, Pinka, Brooks-Gunn, Jeanne, 2004. Wic participation, breastfeeding practices, and well-child careamong unmarried, low-income mothers. American Journal of Public Health 94 (8), 1324–1327.
Chatterji, Pinka, Markowitz, Sarah, 2000. The impact of maternal alcohol and illicit drug use on children’sbehavior problems: evidence from the children of the national longitudinal survey of youth. WorkingPaper 7692, National Bureau of Economic Research.
Chatterji, Pinka, Bonuck, Karen, Dhawan, Simi, Deb, Nandini, 2002. WIC participation and the initiationand duration of breastfeeding. Discussion Paper No. 1246-02. Madison, WI: Institute for Research onPoverty.
Chay, Kenneth Y., Greenstone, Michael, 2003a. Air quality, infant mortality, and the Clean Air Act of 1970.Working Paper 10053, National Bureau of Economic Research, October.URL http://www.nber.org/papers/w10053.
Chay, Kenneth Y., Greenstone, Michael, 2003b. The impact of air pollution on infant mortality: evidencefrom the geographic variation in pollution shocks induced by a recession. Quarterly Journal of Economics118 (3), 1121–1167.
Chay, Kenneth Y., Guryan, Jonathan, Mazumder, Bhashkar, 2009. Birth cohort and the black-whiteachievement gap: the roles of access and health soon after birth. Working Paper 15078, National Bureauof Economic Research, June. URL http://www.nber.org/papers/w15078.
Chiang, Alpha C., 1984. Fundamental Methods of Mathematical Economics, third edition, McGraw-Hill,New York.
Condliffe, Simon, Link, Charles R., 2008. The relationship between economic status and child health:evidence from the United States. American Economic Review 98 (4), 1605–1618.
Conley, Dalton, Bennett, Neil G., 2001. Birth weight and income: interactions across generations. Journalof Health and Social Behavior (ISSN: 00221465) 42 (4), 450–465.URL http://www.jstor.org/stable/3090189.
Conley, Dalton, Pfeiffer, Katherine M., Velez, Melissa, 2007. Explaining sibling differences in achievementand behavioral outcomes: the importance of within- and between-family factors. Social Science Research36 (3), 1087–1104.
Conley, Dalton, Strully, Kate W., Bennett, Neil G., 2006. Twin differences in birth weight: the effectsof genotype and prenatal environment on neonatal and post-natal mortality. Economics and HumanBiology 4 (2), 151–183.
Costa, Dora L., Lahey, Joanna N., 2005. Predicting older age mortality trends. Journal of the EuropeanEconomic Association 3 (2–3), 487–493.
Crimmins, Eileen M., Finch, Caleb E., 2006. Infection, inflammation, height, and longevity. Proceedingsof the National Academy of Sciences 103 (2), 498–503.
Cunha, Flavio, Heckman, James, 2007. The technolgy of skill formation. American Economic Review 97(2), 31–47.
Cunha, Flavio, Heckman, James J., 2008. Formulating, identifying and estimating the technology ofcognitive and noncognitive skill formation. Journal of Human Resources 43 (4), 738–782. URLhttp://jhr.uwpress.org/cgi/content/abstract/43/4/738.
Cunha, Flavio, Heckman, James, Schennach, Susanne, 2010. Estimating the technology of cognitive andnoncognitive skill formation. Econometrica 78 (3), 883–931.
Currie, Alison, Shields, Michael, Price, Stephen W., 2007. Is the child health/family income gradientuniversal? Evidence from England. Journal of Health Economics 26 (2), 213–232.
Currie, Janet, 2003. US food and nutrition programs. In: Moffitt, Robert (Ed.), Means-tested TransferPrograms in the United States. University of Chicago Press for NBER, Chicago, IL, pp. 199–290.
Currie, Janet, Cole, Nancy, 1993. Welfare and child health: the link between AFDC participation and birthweight. The American Economic Review 83 (4), 971–985.
Currie, Janet, Gahvari, Firouz, 2008. Transfers in cash and in-kind; theory meets the data. Journal ofEconomic Literature 46 (2), 333–383.
Currie, Janet, Grogger, Jeffrey, 2002. Medicaid expansions and welfare contractions: offsetting effects onprenatal care and infant health. Journal of Health Economics 21 (2), 313–335.
1480 Douglas Almond and Janet Currie
Currie, Janet, Gruber, Jonathan, 1996a. Health insurance eligibility, utilization of medical care, and childhealth. The Quarterly Journal of Economics 111 (2), 431–466.
Currie, Janet, Gruber, Jonathan, 1996. Saving babies: the efficacy and cost of recent changes inthe Medicaid eligibility of pregnant women. Journal of Political Economy 104 (6), 1263. URLhttp://www.journals.uchicago.edu/doi/abs/10.1086/262059.
Currie, Janet, Hyson, Rosemary, 1999. Is the impact of shocks cusioned by socioeconomic status? The caseof low birth weight. American Economic Review 89 (2), 245–250.
Currie, Janet, Moretti, Enrico, 2007. Biology as destiny? short- and long-run determinants ofintergenerational transmission of birth weight. Journal of Labor Economics 25 (2), 231–264. URLhttp://www.journals.uchicago.edu/doi/abs/10.1086/511377.
Currie, Janet, Moretti, Enrico, 2008. Short and long-run effects of the introduction of food stamps on birthoutcomes in California. In: Schoeni, R.F., House, J., Kaplan, G., Pollack, H. (Eds.), Making AmericansHealthier: Social and Economic Policy as Health Policy. Russell Sage, New York.
Currie, Janet, Neidell, Matthew, 2005. Air pollution and infant health: what can we learn fromCalifornia’s recent experience? Quarterly Journal of Economics 120 (3), 1003–1030. URLhttp://www.mitpressjournals.org/doi/abs/10.1162/003355305774268219.
Currie, Janet, Stabile, Mark, 2003. Socioeconomic status and child health: why is the relationship strongerfor older children? American Economic Review 93 (5), 1813–1823.
Currie, Janet, Stabile, Mark, 2006. Child mental health and human capital accumulation: the case of ADHD.Journal of Health Economics 25 (6), 1094–1118.
Currie, Janet, Tekin, Erdal, 2006. Does child abuse cause crime? Working Paper 12171, National Bureau ofEconomic Research, April. URL http://www.nber.org/papers/w12171.
Currie, Janet, Thomas, Duncan, 1995. Does Head Start make a difference? American Economic Review 85(3), 341–364.
Currie, Janet, Thomas, Duncan, 1999a. Does Head Start help Hispanic children? Journal of PublicEconomics (ISSN: 0047-2727) 74 (2), 235–262.
Currie, Janet, Thomas, Duncan, 1999b. Early test scores, socioeconomic status and future outcomes.Working Paper 6943, National Bureau of Economic Research, February.URL http://www.nber.org/papers/w6943.
Currie, Janet, Walker, Reed, 2009. Traffic congestion and infant health: evidence from E-Zpass. WorkingPaper 15413, National Bureau of Economic Research, October.URL http://www.nber.org/papers/w15413.
Currie, Janet, Widom, Cathy S., 2009. Long-term consequences of child abuse and neglect on adulteconomic well-being. Manuscript, September.
Currie, Janet, Yelowitz, Aaron, 2000. Are public housing projects good for kids? Journal of Public Economics(ISSN: 0047-2727) 75 (1), 99–124.
Currie, Janet, Decker, Sandra, Lin, Wanchuan, 2008. Has public health insurance for older children reduceddisparities in access to care and health outcomes? Journal of Health Economics 27 (6), 1567–1581.
Currie, Janet, Neidell, Matthew, Schmieder, Johannes F., 2009. Air pollution and infant health: lessons fromNew Jersey. Journal of Health Economics 28 (3), 688–703.
Currie, Janet, Stabile, Mark, Manivong, Phongsack, Roos, Leslie L., 2010. Child health and young adultoutcomes. The Journal of Human Resources 45 (3), 517–548.
Cutler, David M., Miller, Grant, Norton, Douglas M., 2007. Evidence on early-life income and late-life health from America’s Dust Bowl era. Proceedings of the National Academy of Sciences 104 (33),13244–13249. URL http://www.pnas.org/content/104/33/13244.abstract.
Dafny, Leemore, Gruber, Jonathan, 2005. Public insurance and child hospitalizations: access and efficiencyeffects. Journal of Public Economics 89 (1), 109–129.URL http://ideas.repec.org/a/eee/pubeco/v89y2005i1p109-129.html.
Dahl, Gordon, Lochner, Lance, 2008. The impact of family income on child achievement: evidence from theearned income tax credit. Working Paper 14599. National Bureau of Economic Research, December.URL http://www.nber.org/papers/w14599.
Dahl, Gordon B., Lochner, Lance, 2005. The impact of family income on child achievement. Working Paper11279, National Bureau of Economic Research, April. URL http://www.nber.org/papers/w11279.
Human Capital Development before Age Five 1481
Dalenius, Tore, Reiss, Steven P., 1982. Data swapping: a technique for disclosure control. Journal of StatisticalPlanning and Inference 6 (1), 73–85.
Datar, Ashlesha, Kilburn, Rebecca, Loughran, David S., 2010. Endowments and parental investments ininfancy and early childhood. Demography 47 (1), 145–162.
Dearden, Lorraine, 1998. Ability, family, education, and earnings in Britain. Working Paper 98/114,Institute for Fiscal Studies, July.
Del Bono, Emilia, Ermisch, John, Francesconi, Marco, 2008. Intrafamily resource allocations: a dynamicmodel of birth weight. IZA Discussion Papers 3704, Institute for the Study of Labor (IZA), September.URL http://ideas.repec.org/p/iza/izadps/dp3704.html.
Deming, David, 2009. Early childhood intervention and life-cycle skill development. American EconomicJournal: Applied Economics 1 (3), 111–134.
Deschnes, Olivier, Greenstone, Michael, 2007. Climate change, mortality, and adaptation: evidence fromannual fluctuations in weather in the US. NBER Working Paper No. 13178, June.
Doblhammer, Gabriele, Vaupel, James W., 2001. Lifespan depends on month of birth. Proceedings of theNational Academy of Sciences 98 (5), 2934–2939.
Doyle Jr., Joseph J., 2008. Child protection and adult crime: using investigator assignment to estimate causaleffects of foster care. Journal of Political Economy 116 (4), 746–770.URL http://www.journals.uchicago.edu/doi/abs/10.1086/590216.
Doyle, Orla, Walker, Ian, Harmon, Colm P., 2005. The impact of parental income and education on thehealth of their children. Discussion Paper 5359, Centre for Economic Policy Research, December.
Dubay, Lisa, Joyce, Ted, Kaestner, Robert, Kenney, Genevieve M., 2001. Changes in prenatal care timingand low birth weight by race and socioeconomic status: implications for the medicaid expansions forpregnant women. Health Services Research 36 (2), 373–398.
Duggan, Anne K., McFarlane, Elizabeth, Fuddy, Loretta, Burrell, Lori, Higman, Susan M., Windham, Amy,Sia, Calvin, 2004. Randomized trial of a statewide home visiting program: impact on preventing childabuse and neglect. Child Abuse and Neglect 28 (6), 597–622.
Duggan, A.K., McFarlane, E.C., Windham, A.M., Rohde, C., Salkever, D.S., Fuddy, L., Rosenberg,L.A., Buchbinder, S.B., Sia, C.C.J., 1999. Evaluation of Hawaii’s Healthy Start Program. The Futureof Children 9 (1), 66–90.
DuMont, Kimberly, Mitchell-Herzfeld, Susan, Greene, Rose, Lee, Eunju, Lowenfels, Ann, Rodriguez,Monica, 2006. Healthy families New York (hfny) randomized trial: impacts on parenting after the firsttwo years. Working Paper Series: evaluating Healthy Families New York 1, New York State Office ofChildren & Family Services, June.
Dustmann, Christian, Schonberg, Uta, 2009. The effect of expansions in maternity leave coverage onchildren’s long-term outcomes. SSRN eLibrary.
Eissa, Nada, Liebman, Jeffrey B., 1996. Labor supply response to the earned income tax credit. The QuarterlyJournal of Economics 113 (3), 653–691.
Farrelly, Matthew C., Nimsch, Christian T., James, Joshua, 2003. State cigarette excise taxes: Implicationsfor revenue and tax evasion. Report 08742.000, RTI International, Research Triangle Park, NC, May.
Fergusson, David M., Grant, Hildegard, Horwood, L. John, Ridder, Elizabeth M., 2006. Randomized trialof Early Start program of home visitation. Pediatrics 117 (3), 803–809.
Fertig, Angela R., Reingold, David A., 2007. Public housing, health, and health behaviors: Is there aconnection? Journal of Policy Analysis and Management 26 (4), 831–859.
Field, Erica, Robles, Omar, Torero, Maximo, 2009. Iodine deficiency and schooling attainment in Tanzania.American Economic Journal: Applied Economics 1 (4), 140–169.
Figlio, David, Hamersma, Sarah, Roth, Jeffrey, 2009. Does prenatal WIC participation improve birthoutcomes? new evidence from Florida. Journal of Public Economics 93 (1–2), 235–245.
Fletcher, Jason M., Lehrer, Steven F., 2009. Using genetic lotteries within families to examine the causalimpact of poor health on academic achievement. Working Paper 15148, National Bureau of EconomicResearch, July. URL http://www.nber.org/papers/w15148.
Frank, Richard G., Meara, Ellen, 2009. The effect of maternal depression and substance abuse on childhuman capital development. Working Paper 15314, National Bureau of Economic Research, September.
Frisvold, David E., 2006. Head Start participation and childhood obesity. Working Paper 06-WG01,Vanderbilt University, February.
1482 Douglas Almond and Janet Currie
Frisvold, David E., Lumeng, Julie C., 2009. Expanding exposure: can increasing the daily duration of HeadStart reduce childhood obesity? manuscript, April.
Garces, Eliana, Thomas, Duncan, Currie, Janet, 2002. Longer-term effects of Head Start. AmericanEconomic Review 92 (4), 999–1012.
Gertler, Paul J., 2004. Do conditional cash transfers improve child health? Evidence from Progresa’s controlrandomized experiment. Health, Health Care, and Economic Development 94 (2), 336–341.
Gluckman, Peter D., Hanson, Mark A., 2006. Adult disease: echoes of the past. European Journal ofEndocrinology 155 (Suppl. 1), S47–S50.
Gormley Jr., William, 2008. The effects of Oklahoma’s pre-K program on Hispanic children. Social ScienceQuarterly 89 (4), 916–936.
Gormley Jr., William, Gayer, Ted, 2006. Promoting school readiness in Oklahoma: an evaluation of Tulsa’spre-K program. The Journal of Human Resources 40 (3), 533–558.
Gormley Jr., William, Gayer, Ted, Phillips, Deborah, Dawson, Brittany, 2005. The effects of universal pre-Kon cognitive development. Developmental Psychology 41 (6), 872–884.
Goux, Dominique, Maurin, Eric, 2005. The effects of overcrowded housing on children’s performance atschool. Journal of Public Economics 89 (5-6), 797–819.
Griliches, Zvi, 1979. Sibling models and data in economics: beginnings of a survey. Journal of PoliticalEconomy 87 (s5), S37. URL http://www.journals.uchicago.edu/doi/abs/10.1086/260822.
Grossman, Michael, 1972. On the concept of health capital and the demand for health. Journal of PoliticalEconomy 223–255.
Gueorguieva, Ralitza, Morse, Steven B., Roth, Jeffrey, 2009. Length of prenatal participation in WIC andrisk of delivering a small for gestational age infant: Flordida, 1996-2004. Journal of Maternal and ChildHealth 13 (4), 479–488.
Gundersen, Craig, Ziliak, James, 2004. Poverty and macroeconomic performance across space, race, andfamily structure. Demography 41 (1), 61–86.
Hanratty, Maria J., 1996. Canadian national health insurance and infant health. American Economic Review(ISSN: 00028282) 86 (1), 276–284. URL http://www.jstor.org/stable/2118267.
Harding, Kathryn, Galano, Joseph, Martin, Joanne, Huntington, Lee, Schellenbach, Cynthia J., 2007.Healthy Families America effectiveness: a comprehensive review of outcomes. Journal of Prevention andIntervention in the Community 34 (1-2), 149–179.
Havnes, Tarjei, Mogstad, Magne, 2009. No child left behind: universal child care and children’s long-runoutcomes. Discussion Paper 582, Statistics Norway, Research Department, May.
Heckman, James, 2007. The economics, technology, and neuroscience of human capability formation.PNAS 104 (33), 13250–13255.
Heckman, James, Moon, Seong H., Pinto, Rodrigo, Savelyev, Peter A., Yavitz, Adam, 2009. The rateof return to the High/Scope Perry preschool program. Working Paper 15471, National Bureau ofEconomic Research, November.
Heckman, James J., Rubinstein, Yona, 2001. The importance of noncognitive skills: lessons from the GEDtesting program. American Economic Review: Papers & Proceedings 91 (2), 145–149.
Heider, Fritz, 1934. The influence of the epidemic of 1918 on deafness: a study of birth dates of pupilsregistered in schools for the deaf. American Journal of Epidemiology 19, 756–765.
Howard, Kimberly S., Brooks-Gunn, Jeanne, 2009. The role of home visiting programs in preventing childabuse and neglect. The Future of Children 19 (2), 119–146.
Hoynes, Hilary W., Page, Marianne, Stevens, Ann H., 2009. Is a WIC start a better start? evaluatingWIC’s impact on infant health using program introduction. Working Paper 15589, National Bureau ofEconomic Research, December.
Hoynes, Hilary W., Schanzenbach, Diane, 2009. Consumption responses to in-kind transfers: evidence fromthe introduction of the Food Stamp program. American Economic Journal: Applied Economics 1 (4),109–139.
Hsin, Amy, 2009. Is biology destiny? Birth weight and differential parental treatment. Manuscript, PopulationStudies Center, University of Michigan.
Hustedt, Jason T., Barnett, W. Steven, Jung, Kwanghee, Figueras, Alexandra, 2008. Impacts of new Mexicopre-K on children’s school readiness at kindergarten entry: Results from the second year of a growinginitiative. Research report, National Institute for Early Education, Rutgers University, June.
Human Capital Development before Age Five 1483
Jacob, Brian, 2004. Public housing, housing vouchers, and student achievement: evidence from publichousing demolitions in Chicago. American Economic Review 94 (1), 233–258.
Johnson, Rucker C., Schoeni, Robert F., 2007. The influence of early-life events on human capital, healthstatus, and labor market outcomes over the life course. Working Paper 07-05, National Poverty Center,February.
Joyce, Theodore, Gibson, Diane, Colman, Sylvie, 2004. The changing association between prenatalparticipation in WIC and birth outcomes in New York city. Working Paper 10796, National Bureauof Economic Research, September.
Joyce, Theodore, Racine, Andrew D., Yunzal-Butler, Cristina, 2007. Reassessing the WIC effect: evidencefrom the pregnancy nutrition surveillance system. Journal of Policy Analysis and Management 87 (2),277–303.
Kaestner, Robert, Joyce, Ted, Racine, Andrew, 2001. Medicaid eligibility and the incidence of ambulatorycare sensitive hospitalizations for children. Social Science and Medicine 52 (2), 305–313.
Karoly, Lynn A., Kilburn, Rebecca M., Cannon, Jill S., 2005. Early Childhood Interventions: ProvenResults, Future Promise. RAND Corporation.
Karoly, Lynn A., Greenwood, Peter W., Everingham, Susan S., Hoube, Jill, Kilburn, M. Rebecca, Rydell, C.Peter, Sanders, Matthew, Chiesa, James, 1998. Investing in our children: what we know and don’t knowabout the costs and benefits of early childhood interventions. Consensus report, Rand, July.
Kehrer, Barbara H., Wolin, Charles M., 1979. Impact of income maintenance on low birth weight: evidencefrom the Gary experiment. The Journal of Human Resources 14 (4), 434–462. autumn.
Kelly, Elaine, 2009. The scourge of Asian flu: in utero exposure to pandemic influenza and the developmentof a cohort of British children. Institute for Fiscal Studies, Working Paper 09/17, September. UniversityCollege London.
Kermack, W.O., McKendrick, A.G, McKinlay, P.L., 1934. Death-Rates in Great Britain and Sweden. TheLancet 698–703.
Khanam, Rasheda, Nghiem, Hong S., Connelly, Luke B., 2009. Child health and the income gradient:evidence from Australia. Journal of Health Economics 28 (4), 805–817.
Kling, Jeffrey R., Liebman, Jeffrey B., Katz, Lawrence F., 2007. Experimental analysis of neighborhoodeffects. Econometrica 75 (1), 83–119.
Kling, Jeffrey R., Ludwig, Jeffrey, Katz, Lawrence F., 2005. Neighborhood effects on crime for femaleand male youth: evidence from a randomized housing voucher experiment. The Quarterly Journal ofEconomics 120 (1), 87–130.
Knittel, Christopher R., Miller, Douglas L., Sanders, Nicholas J., 2009. Caution drivers! children present:Traffic, pollution and infant health. Manuscript, UC Davis.
Koch, Thomas G., 2009. Public insurance, crowd-out, and health care utilization. Manuscript, August.Kowaleski-Jones, Lori, Duncan, Greg J., 2002. Effects of participation in the WIC program on birthweight:
evidence from the National Longitudinal Survey of Youth. Am J Public Health 92 (5), 799–804. URLhttp://ajph.aphapublications.org/cgi/content/abstract/92/5/799.
Kraemer, Sebastian, 2000. The fragile male. BMJ 321 (7276), 1609–1612. URL http://www.bmj.com.Lee, Bong J., Mackey-Bilaver, Lee, 2007. Effects of WIC and Food Stamp program participation on child
outcomes. Children and Youth Services Review 29 (4), 501–517.Lefebvre, Pierre, Merrigan, Philip, Verstraete, Matthieu, 2006. Impact of early childhood care and education
on children’s preschool cognitive development: Canadian results from a large quasi-experiment. WorkingPaper 06-36, Centre interuniversitaire sur le risque, les politiques conomiques et lemploi (CIRPEE),October.
Levine, Philip B., Schanzenbach, Diane W., 2009. The impact of children’s public health insuranceexpansions on educational outcomes. Forum for Health Economics and Policy 12 (1), Article 1.
Levine, Phillip B., Zimmerman, David J., 2000. Children’s welfare exposure and subsequent development.Working Paper 7522, National Bureau of Economic Research, February.URL http://www.nber.org/papers/w7522.
Lhila, Aparna, Simon, Kosali I., 2008. Prenatal health investment decisions: does the childs sex matter?Demography 45 (4), 885–905.
Lien, Diana S., Evans, William N., 2005. Estimating the Impact of large cigarette tax hikes: the case ofmaternal smoking and infant birth weight. J. Human Resources XL (2), 373–392.URL http://jhr.uwpress.org/cgi/content/abstract/XL/2/373.
1484 Douglas Almond and Janet Currie
Lin, Ming-Jen, Liu, Jin-Tan, 2009. Do lower birth weight babies have lower grades? twin fixed effect andinstrumental variable evidence from Taiwan. Social Science & Medicine 68 (10), 1780–1787.
Logan, W.P.D., Glasg, M.D., 1953. Mortality in London fog incident, 1952. Lancet 1 (1), 336–338.Love, John M., Kisker, Ellen E., Ross, Christine, Constantine, Jill, Boller, Kimberly, Chazan-Cohen,
Rachel, Brady-Smith, Christy, Fuligni, Allison S., Raikes, Helen, Brooks-Gunn, Jeanne, Tarullo, LouisaB., Schochet, Peter Z., Paulsell, Diane, Vogel, Cheri, 2005. The effectiveness of Early Head Start for3-year-old children and their parents: lessons for policy and programs. Developmental Psychology 41 (6),885–901.
Ludwig, Jens, Miller, Douglas L., 2005. Does Head Start improve children’s life chances? evidence from aregression discontinuity design. Discussion Paper 1311-05, Institute for Research on Poverty, November.
Ludwig, Jens, Miller, Douglas L., 2007. Does Head Start improve children’s life chances? evidence from aregression discontinuity design. Quarterly Journal of Economics 488 (1), 159–208.
MacDorman, Marian F., Martin, Joyce A., Mathews, T.J., Hoyert, Donna L., Ventura, Stephanie J., 2005.Explaining the 2001-2002 infant mortality increase in the United States: data from the linked birth/infantdeath data set. International Journal of Health Services 35 (3), 415–442.
MacMillan, Harriet L., Thomas, B. Helen, Jamieson, Ellen, Walsh, Christine A., Boyle, Michael H.,Shannon, Harry S., Gafni, Amiram, 2005. Effectiveness of home visitation by public health nurses inprevention of the recurrence of child physical abuse: a randomized controlled trial. Lancet 365 (9473),1786–1793.
Macours, Karen, Schady, Norbert, Vakis, Renos, 2008. Cash transfers, behavioral changes, and cognitivedevelopment in early childhood: evidence from a randomized experiment. Working Paper 4759, TheWorld Bank: Development Research Group, Human Development and Public Services Team, October.
Magnuson, Katherine A., Ruhm, Christopher, Waldfogel, Jane, 2007. Does pre-kindergarten improveschool preparation and performance? Economics of Education Review 26 (1), 33–51.
Masse, Leonard, Barnett, W. Steven, 2002. A Benefit-cost Analysis of the Abecedarian Early ChildhoodIntervention. National Institute for Early Education Research, Rutgers, NJ.
McCormick, Marie C., Brooks-Gunn, Jeanne, Buka, Stephen L., Goldman, Julie, Yu, Jennifer, Salganik,Mikhail, Scott, David T., Bennett, Forrest C., Kay, Libby L., Bernbaum, Judy C., Bauer, Charles R.,Martin, Camilia, Woods, Elizabeth R., Martin, Anne, Casey, Patrick H., 2006. Early intervention in lowbirth weight premature infants: results at 18 years of age for the infant health and development. Pediatrics117 (3), 771–780.
McIntosh, Emma, Barlow, Jane, Davis, Hilton, Stewart-Brown, Sarah, 2009. Economic evaluation of anintensive home visiting programme for vulnerable families: a cost-effectiveness analysis of public healthintervention. Journal of Public Health 31 (3), 423–433.
McLeod, Jane, Kaiser, Karen, 2004. Childhood emotional and behavioral problems and educationalattainment. American Sociological Review 69 (5), 636–658.
Meyer, Bruce D., Rosenbaum, Dan T., 2001. Welfare, the earned income tax credit, and the labor supply ofsingle mothers. Quarterly Journal of Economics 116 (3), 1063–1114.URL http://www.mitpressjournals.org/doi/abs/10.1162/00335530152466313.
Milligan, Kevin, Stabile, Mark, 2008. Do child tax benefits affect the wellbeing of children? evidence fromCanadian child benefit expansions. Working Paper 14624, National Bureau of Economic Research,December. URL http://www.nber.org/papers/w14624.
Morris, Pamela, Duncan, Greg J., Rodrigues, Christopher, 2004. Does money really matter? estimatingimpacts of family income on children’s achievement with data from random-assignment experiments.Manuscript.
Newman, Sandra J., Harkness, Joseph M., 2002. The long-term effects of public housing on self-sufficiency.Journal of Policy Analysis and Management 21 (1), 21–43.
Nilsson, Peter, 2008. Does a pint a day affect your child’s pay? the effect of prenatal alcohol exposure onadult outcomes. Centre for Microdata Methods and Practice Working Paper CWP22/08, The Institutefor Fiscal Studies, August.
Nilsson, Peter, 2009. e long-term effects of early childhood lead expsoure: evidence from sharp changesin local air lead levels induced by the phase-out of leaded gasoline. manuscript, Uppsala University,September.
Human Capital Development before Age Five 1485
Oaxaca, Ronald L., 1973. Male-female wage differentials in urban labor markets. International EconomicReview 14, 693–709.
Olds, David L., Henderson, Charles R., Kitzman, Harriet J., Eckenrode, John J., Cole, Robert E.,Tatelbaum, Robert C., 1999. Prenatal and infancy home visitation by nurses: recent findings. The Futureof Children 9 (1), 44–65. Spring-Summer.
Olds, David L., Kitzman, Harriet, Hanks, Carole, Cole, Robert, Anson, Elizabeth, Sidora-Arcoleo,Kimberly, Luckey, Dennis W., Henderson Jr., Charles R., Holmberg, John, Tutt, Robin A., Stevenson,Amanda J., Bondy, Jessica, 2007. Effects of nurse home visiting on maternal and child functioning: age-9follow-up of a randomized trial. Pediatrics 120 (4), 832–845.
Olds, David L., Robinson, JoAnn, O’Brien, Ruth, Luckey, Dennis W., Pettitt, Lisa M., Henderson Jr.,Charles R., Ng, Rosanna K., Sheff, Karen L., Korfmacher, Jon, Hiatt, Susan, Talmi, Ayelet, 2002. Homevisiting by paraprofessionals and by nurses: a randomized controlled trial. Pediatrics 110 (3), 486–496.URL http://pediatrics.aappublications.org/cgi/content/abstract/110/3/486.
Oreopoulos, Phil, Stabile, Mark, Walld, Randy, Roos, Leslie, 2008. Short, medium, and long-termconsequences of poor infant health: an anlysis using siblings and twins. Journal of Human Resources43 (1), Winter.
Oreopoulos, Philip, 2003. The long-run consequences of living in a poor neighborhood. Quarterly Journalof Economics 118 (4), 1533–1575.URL http://www.mitpressjournals.org/doi/abs/10.1162/003355303322552865.
Orr, Larry, Feins, Judith D., Jacob, Robin, Beecroft, Erik, Sanbonmatsu, Lisa, Katz, Lawrence F., Liebman,Jeffrey B., Kling, Jeffrey R., 2003. Moving to opportunity: interim impacts evaluation. Final report, USDepartment of Housing and Urban Development, Office of Policy Development and Research, June.
Pitt, Mark, Rosenzweig, Mark, Hassan, Nazmul, 1990. Productivity, health, and inequality in theintrahousehold distribution of food in low-income countries. American Economic Review 80 (5),1139–1156.
Propper, Carol, Rigg, John, Burgess, Simon, 2007. Child health: evidence on the roles of family income andmaternal mental health from a UK birth cohort. Health Economics 16 (11), 1245–1269.
Puma, Michael, Bell, Stephen, Cook, Ronna, Heid, Camilla, Shapiro, Gary, Broene, Pam, Jenkins,Frank, Fletcher, Philip, Quinn, Liz, Friedman, Janet, Ciarico, Janet, Rohacek, Monica, Adams, Gina,Spier, Elizabeth, 2010. Head start impact study: Final report; executive summary. Technical report,Administration for Children and Families. US Department of Health and Human Services, January.
Rasmussen, Kathleen Maher, 2001. The ”fetal origins” hypothesis: challenges and opportunities for maternaland child nutrition. Annual Review of Nutrition 21, 73–95.
Reyes, Jessica Wolpaw, 2007. Environmental policy as social policy? the impact of childhood leadexposure on crime. Working Paper 13097, National Bureau of Economic Research, May. URLhttp://www.nber.org/papers/w13097.
Reznek, Arnold, 2007. Methods of confidential data protection. VirtualRDC powerpoint presentation,Cornell University, April. URL http://www.vrdc.cornell.edu/news/.
Rosenzweig, Mark R., Paul Schultz, T., 1982. Market opportunities, genetic endowments, and intrafamilyresource distribution: child survival in rural India. American Economic Review 72 (4), 803–815.
Rosenzweig, Mark R., Wolpin, Kenneth I., 1988. Heterogeneity, intrafamily distribution, and child health.Journal of Human Resources 23 (4), 437–461.
Royer, Heather, 2009. Separated at girth: US twin estimates of the effects of birth weight. AmericanEconomic Journal: Applied Economics 1 (1), 49–85.
Ruhm, Christopher J., 2004. Parental employment and child cognitive development. The Journal of HumanResources 39 (1), 155–192.
Rush, David, Stein, Zena, Susser, Mervyn, 1980. Diet in Pregnancy: a Randomized Controlled Trial ofNutritional Supplements. Alan R. Liss, Inc., New York.
Salm, Martin, Schunk, Daniel, 2008. The role of childhood health for the intergenerational transmission ofhuman capital: evidence from administrative data. Discussion Paper 3646, The Institute for the Study ofLabor (IZA), August.
Sanbonmatsu, Lisa, Kling, Jeffrey R., Duncan, Greg J., Brooks-Gunn, Jeanne, 2006. Neighborhoods andacademic achievement. Journal of Human Resources 41 (4), 649–690.
1486 Douglas Almond and Janet Currie
Schanzenbach, Diane W., What are food stamps worth. American Economic Journal - Economic Policy(forthcoming).
Scherjon, Sicco A., Oosting, Hans, Ongerboer de Visser, Bram W., deWilded, Ton, Zondervan, Hans A.,Kok, Joke H., 1996. Fetal brain sparing is associated with accelerated shortening of visual evoked potentiallatencies during early infancy. American Journal of Obstetrics and Gynecology 175 (6), 1569–1575.
Sloan, Frank A., Conover, Christopher J., Rankin, Peter J., 2001. Effects of Tennessee Medicaid managedcare on obstetrical care and birth outcomes. Journal of Health Politics, Policy, and Law 26 (6), 1291–1324.
Smith, James P., 2009. The impact of childhood health on adult labor market outcomes. The Review ofEconomics and Statistics 91 (3), 478–489.
Smith, James P., Smith, Gillian C., 2008. Long-term economic costs of psychological problems duringchildhood. Manuscript, October.
Smolensky, E., Gootman, J. Appleton, 2003. Working Families and Growing Kids: Caring for Children andAdolescents. The National Academies Press, Washington, DC.
Solon, Gary, 1999. Chapter Intergenerational mobility and the labor market. In: Ashenfelter, Orley,Card, David (Eds.), Handbook of Labor Economics, vol. 3. Elsevier, pp. 1761–1800.
St. Pierre, Robert G., Layzer, Jean I., 1999. Using home visits for multiple purposes: the comprehensivechild development program. The Future of Children 9 (1), 134–151.
Stein, Zena, Susser, Mervyn, Saenger, Gerhart, Marolla, Francis, 1975. Famine and Human Development:the Dutch Hunger Winter of 1944-1945. Oxford University Press, New York.
Tominey, Emma, 2007. Maternal smoking during pregnancy and early child outcomes. CEP DiscussionPapers dp0828, Centre for Economic Performance, LSE, October.URL http://ideas.repec.org/p/cep/cepdps/dp0828.html.
Vujic, Suncica, Konig, Pierre, Webbinnk, Dinand, Martin, Nick, 2008. The effect of childhood conductdisorder on human capital. Discussion Paper 3646, CPB Netherlands Bureau for Economic PolicyAnalysis, November.
Wong, Vivian C., Cook, Thomas D., Barnett, W. Steven, Jung, Kwanghee, 2008. An effectiveness-basedevaluation of five state pre-kindergarten programs. Journal of Policy Analysis and Management 27 (1),122–154.
Zweifel, Peter, Breyer, Friedrich H.J., Kifmann, Mathias, 2009. Health Economics, second edition, OxfordUniversity Press.