- Email: [email protected]
Maternal breastfeeding and children's cognitive development
Accepted Manuscript Maternal breastfeeding and children's cognitive development Kanghyock Koh PII:
S0277-9536(17)30380-5
DOI:
10.1016/j.socscimed.2017.06.012
Reference:
SSM 11276
To appear in:
Social Science & Medicine
Received Date: 28 November 2016 Revised Date:
30 May 2017
Accepted Date: June 2017
Please cite this article as: Koh, K., Maternal breastfeeding and children's cognitive development, Social Science & Medicine (2017), doi: 10.1016/j.socscimed.2017.06.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
SSM Manuscript Number: SSM-D-16-03615R1
RI PT
Maternal Breastfeeding and Children’s Cognitive Development∗
AC C
EP
TE D
M AN U
SC
Kanghyock Koh+ Ulsan National Institute of Science and Technology (UNIST)
∗
I thank Anna Aizer, Kenneth Chay, Hyojin Han, Hyunjoo Yang, and the editor and four anonymous referees for valuable comments. Any remaining errors are mine. This work was supported by the New Faculty Research Fund(1.160080) of UNIST(Ulsan National Institute of Science & Technology). + 50 UNIST-gil, Ulju-gun, Ulsan 44919, Republic of Korea. Email: [email protected]
ACCEPTED MANUSCRIPT
ABSTRACT
SC
RI PT
Do children with lower test scores benefit more from breastfeeding than those with higher scores? In this paper, I examine the distributional effects of maternal breastfeeding on the cognitive test scores of 11,544 children who were born in 2000 and 2001 in the United Kingdom using a semiparametric quantile regression model. I find evidence that maternal breastfeeding has larger positive impacts on children with lower test scores. Effects for children below the 20th percentile are about 2–2.5 times greater than those for children above the 80th percentile. I also find that these distributional effects are larger when the duration of breastfeeding is extended. One policy implication is that a public policy aims at promoting breastfeeding might narrow a disparity in children’s cognition.
Keywords: United Kingdom, distributional effects, maternal breastfeeding, children’s development, semiparametric quantile regression
M AN U
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Highlights • Study of the effects of breastfeeding on the distribution of children’s test scores. • A semiparametric quantile regression is used for the estimation. • The effects are greater for children with lower test scores.
TE D
22
1. Introduction
24
Economic theories of human capital have emphasized the importance of parental investment in
25
children (Becker, 1981; Cunha & Heckman, 2007). A large number of empirical studies show
26
that human capital developments in early childhood (e.g., cognitive ability) play a significant
27
role in human capital developments later in life, as measured by educational attainment,
28
employment, wages, etc. (Heckman, 2006; Cunha & Heckman, 2007; Cunha, Heckman, &
29
Schennach, 2010; Almond & Currie, 2011a,b).
AC C
30
EP
23
Maternal breastfeeding has been emphasized as an influential factor in early childhood
31
development (e.g., Kramer et al., 2001, 2008) based on biological mechanisms through which
32
maternal breastfeeding aids children’s development. First, the composition of breast milk is
33
superior to that of formula. Breast milk contains long-chain polyunsaturated fatty acids, such as 1
ACCEPTED MANUSCRIPT
docosahexaenoic acid (DHA) and arachidonic acid (AA), which form the major structures of
35
neuronal membranes and play critical roles in nervous system functioning by positively
36
stimulating development of the human brain (Fernstrom, 1999). Infants require sufficient
37
amounts of these acids during the first few months after birth (Clandinin et al, 1981). Note that
38
the compositional superiority of breast milk may be reduced because DHA and AA are now
39
added to formula. However, Fitzsimons and Vera-Hernandez (2012) reported that DHA and AA
40
were not included in formula in the UK (the context of this paper) until most of the infants who
41
are considered in the empirical analysis were born. Deoni et al. (2013) provided MRI evidence
42
that breastfed infants exhibit better development in specific brain areas associated with language
43
and visual reception abilities compared to those who are fed formula or a mixture of breast milk
44
and formula. Second, skin-to-skin contact between mother and child stimulates maternal
45
hormonal responses such as the production of prolactin and oxytocin, which may indirectly
46
improve cognitive development (Del Bono & Rabe, 2012).
SC
M AN U
TE D
47
RI PT
34
In spite of this clear biological mechanism, empirical evidence for the effects of breastfeeding on children’s cognitive development has been conflicting. A comprehensive
49
review by the Agency for Healthcare Research and Quality (Ip et al., 2007) summarized 400
50
articles (out of 9000 abstracts) and found that breastfeeding has few or small effects on
51
children’s cognitive ability. Many studies are based on observational data, and it is difficult to
52
infer causality due to many potential confounding variables, such as duration of breastfeeding,
53
children’s demographics and birth outcomes (such as birth weight and gestational age), parental
54
health-related behavior, socioeconomic status (SES), intelligence, and family characteristics
55
(Anderson, Johnstone, & Remley, 1999).
AC C
EP
48
2
ACCEPTED MANUSCRIPT
Researchers use various empirical strategies to address the endogeneity issue. First, they
57
estimate the probability of breastfeeding using detailed information regarding the characteristics
58
of children, parents, and families. They then estimate the effects of breastfeeding on cognitive
59
development by comparing children with similar propensities of maternal breastfeeding (Jiang,
60
Foster, & Gibson-Davis, 2011; Belfield & Kelly, 2012; Borra, Iacovou, & Sevilla, 2012;
61
Rothstein, 2012; Cesur et al, 2017). For example, Jian, Foster, & Gibson-Davis (2011) used the
62
Panel Study of Income Dynamics in the U.S. and found that the positive associations between
63
breastfeeding initiation and children’s cognitive development measured by the Woodcock
64
Johnson Psycho-Educational Battery Revised (WJ-R) and Wechsler Intelligence Scale for
65
Children-Revised (WISC-R) were significantly reduced when observational characteristics were
66
controlled for. The estimated associations are between one-tenth and one-fifth of a standard
67
deviation, which is small in magnitude. Second, some researchers use sibling or family fixed
68
effects to control for unobserved family characteristics (Der, Batty, & Deary, 2006; Rees &
69
Sabia, 2009; Colen & Ramey, 2014; Cesur et al., 2017). Recent research points out that maternal
70
intelligence or IQ and home environment explain breastfeeding status better than other SES
71
factors such as income or education, but few previous studies have accounted for these
72
covariates (Der, Batty, & Deary, 2006). Many studies using the propensity score approach or
73
fixed effects analysis have found that a positive association became smaller or disappeared once
74
a large number of confounding factors were controlled for. Researchers pointed out these results
75
as evidence for the endogeneity issue in breastfeeding (Belfield & Kelly, 2012; Cesur et al.,
76
2017). Third, economists use exogenous sources of variation in breastfeeding status as
77
instrumental variables to study the causal relationship between breastfeeding and children’s
78
cognitive development (Del Bono & Rabe, 2012; Fitzsimons & Vera-Hernandez, 2012). For
AC C
EP
TE D
M AN U
SC
RI PT
56
3
ACCEPTED MANUSCRIPT
example, infant feeding support is provided by midwives and nurses in the U.K. Since staff
80
working hours are reduced during the weekend, this support is reduced as well. Mothers are less
81
exposed to support when they give birth on Friday or Saturday. Fitzsimons and Vera-Hernandez
82
(2012) use this institutional fact to estimate the effects of breastfeeding on children’s cognitive
83
development among low-income mothers and their children. Finally, Kramer et al. (2001, 2008)
84
studied the causal effects of the Promotion of Breastfeeding Intervention Trial (PROBIT), which
85
randomly provided health care worker assistance for the initiation and maintenance of
86
breastfeeding, on breastfeeding initiation and duration and consequences for children’s health
87
and cognitive abilities in Belarus.
SC
M AN U
88
RI PT
79
Previous literature mainly focused on the average effects of breastfeeding on children’s cognition. Many measures for cognition, such as test scores, are continuously distributed. These
90
distributions could provide a deeper understanding of the effects of breastfeeding on children’s
91
cognition by evaluating the effects at different quantiles in addition to the average effects. For
92
example, children at lower quantiles could benefit more from breastfeeding than average
93
children or children at higher quantiles, even though the average effects of breastfeeding are
94
small or insignificant based on the existing literature. However, there is little knowledge
95
regarding the effects of breastfeeding on the distribution of cognition. To fill this gap, I
96
investigated the effects of maternal breastfeeding on the distribution of children’s cognitive
97
ability in this study. More specifically, I estimated the effects of breastfeeding at different
98
quantiles of cognitive test scores using a semiparametric quantile regression model. To address
99
the endogeneity of breastfeeding, I used the propensity score as the inverse probability
AC C
EP
TE D
89
100
(Rosenbaum & Rubin, 1983; Hirano, Imbens, & Ridder, 2003). Based on rich information about
101
children’s characteristics, cognitive test scores, and maternal breastfeeding from the UK
4
ACCEPTED MANUSCRIPT
102
Millennium Cohort Survey, I compared cognitive test scores among children with similar
103
propensities for maternal breastfeeding based on observed parental characteristics.
RI PT
104 105
2. Method
107
2.1 Data
108
To investigate the effects of maternal breastfeeding on children’s development, I used data from
109
the Millennium Cohort Study (MCS), a longitudinal study of about 18,500 children who were
110
born in 2000 and 2001 in the United Kingdom. Specifically, I used data from four surveys (at
111
ages 9 months, 3 years, 5 years, and 7 years) and excluded data for multiple births and children
112
who were not living with their biological mothers at the time of the first interview (n = 293).
113
Since I used secondary observational data, ethics approval was not required.
M AN U
TE D
114
SC
106
I collected data on self-reported maternal breastfeeding status from the first MCS survey, which included a question about the duration of maternal breastfeeding. Mothers reported the age
116
of babies when breastfeeding stopped, as measured in days, weeks, or months, which was used
117
for calculating the duration of breastfeeding. It was not feasible to obtain data on exclusive
118
breastfeeding from the MCS because survey questions related to infant feeding were not
119
exhaustive. Since the recommended duration of breastfeeding in the UK was 4 months in 2000, I
120
created a binary breastfeeding variable indicating whether children were breastfed for at least 4
121
months. To study the effects of breastfeeding with different duration cutoffs, I examined the
122
effects of initiation of breastfeeding. Given current policy recommendations [e.g., from the
123
World Health Organization (WHO)], I also examined the effects of extended breastfeeding (i.e.,
124
for at least 6 months).
AC C
EP
115
5
ACCEPTED MANUSCRIPT
125
To assess children’s cognitive development, I used scores from six British Ability Scale (BAS) tests, which are used to measure the cognitive abilities of children aged 2.5 to 8 years old
127
(Elliott, Smith, & McCulloch, 1997). Since these tests are individually administered by trained
128
interviewers, the scores provide a more accurate measure than parents’ self-reported measures
129
(Fernald et al., 2009). Using the age-adjusted scores from the MCS, I calculated average scores
130
within and across ages and created a summary index. I did this to address a concern that one null
131
hypothesis could be rejected simply because I tested it with multiple null hypotheses (Kling,
132
Liebman, & Katz, 2007). This also yields a well-behaved continuous distribution of test scores
133
(Figure 1).
SC
M AN U
134
RI PT
126
I also collected information on maternal demographics, maternal prenatal characteristics, and spouse and family characteristics as control variables for the propensity score approach:
136
mother’s age at birth, race, marital status, and education; planned pregnancy; dummy variables
137
for receiving any prenatal care and attending any prenatal classes; timing of initial care; dummy
138
variables for any complications during pregnancy (including bleeding or threatened miscarriage
139
in early pregnancy; bleeding in later pregnancy; pregnancy diagnosed as twins, triplets, or more;
140
persistent vomiting; elevated blood pressure; eclampsia/preeclampsia or toxemia; urinary
141
infection; diabetes; too much fluid around the baby; suspected slow growth of baby; and other
142
suspected problems) and labor (including breech birth, shoulder first, very long (or rapid) labor,
143
fetal distress – heart rate sign, meconium sign, and other complications); smoking status during
144
pregnancy; working status during and after pregnancy; frequency of alcohol consumption;
145
dummy variables for countries (England, Wales, Scotland, and Northern Ireland); father’s age at
146
birth, education level, and job status; dummy variables for joint job status of a couple indicating
147
if both in work, father (mother) in work but mother (father) not in work, both not in work, father
AC C
EP
TE D
135
6
ACCEPTED MANUSCRIPT
(mother) in work or on leave but mother (father) not in work, and father (mother) in work but
149
mother (father) is unknown; number of household members and smokers in the household; and
150
dummy variables for government support and income level. The mothers’ and fathers’ education
151
levels were categorized as NVQ level 1 to 5 and overseas. I defined a dummy variable of low
152
education level if the NVQ level was 3 or lower to construct interactions with other covariates.
153
Due to missing values in these control variables, I used 11,540 children in the main analysis.
RI PT
148
SC
154 155
2.2 Empirical Strategy
156
If
157
and
158
effect on the treated (ATT) is
159
being treated based on observed characteristics be the propensity score (Rosenbaum & Rubin,
160
1983),
161
as maternal characteristics, parental characteristics, and household environment. I include control
162
variables mentioned in the previous section. I also include polynomials and interaction terms
163
among them. I estimated the propensity scores using the logit regression specification and used
164
their inverse probability weights to nonparametrically estimate the ATT (Hirano, Imbens, &
165
Ridder, 2003). I used ̂
166
be constant, I also estimated average treatment effects (ATE) using 1/ ̂
167
sampling weights for treatment and control groups, respectively. For statistical inferences, I
168
calculated robust standard errors to correct for heteroscedasticity in test scores by breastfeeding
169
status. The propensity score approach has some advantages over the ordinary least squares (OLS)
170
approach with many control variables. First, I can alleviate misspecification errors by not
is the outcome if treated (breastfed for more than 4 months),
M AN U
is the treatment indicator,
is the outcome if untreated (breastfed for less than 4 months), then the average treatment
| =
, where
| = 1 . Let the conditional probability of
is a vector of observed pretreatment characteristics such
EP
TE D
= Pr
| =1 −
AC C
/(1 − ̂
) as a weight for the control group. Since the ATT may not
7
and 1/ 1 − ̂
as
ACCEPTED MANUSCRIPT
assuming any parametric relationship between breastfeeding and children’s cognitive
172
development (Zhao, 2008). Second, I can address the curse of dimensionality issue associated
173
with OLS when many control variables are included (Dehejia & Wahba, 1999) by summarizing
174
all information from the control variables into a single index (i.e., the propensity score). Third, I
175
can graphically examine the nonparametric relationship between maternal breastfeeding and
176
children’s cognitive development by using the propensity score.
Then I investigated the distributional impact of maternal breastfeeding using a
SC
177
RI PT
171
178
semiparametric quantile regression model. Following Firpo (2007), I estimated the quantile
179
treatment effect for quantile using ∆ ≡
−
( = 0, 1 , where
M AN U ,
,
,
is defined as
180
!"#$% ∑, ( ∙* +- '
181
− 1/. ≤ 01 , and ' ( is the inverse probability of the estimated propensity score for the ATT, 2
+
−
, the check function * ∙ at a real number . is * . = . ∙
which are ' ( = ,⋅4
183
propensity scores using local logit regression. To estimate treatment effects at different quantiles,
184
the outcome variable should be continuous and well behaved without spikes or gaps, which is
185
again confirmed by Figure 1.
188 189
TE D
where 7 is the sample size. I estimated the
EP
187
64 5
AC C
186
5
and ' ( = ,⋅
62
182
3. Empirical Results
Table 1 shows summary statistics by breastfeeding status. The first three columns show
190
that in general, a statistically significant proportion of children who were breastfed for less than 4
191
months came from families with lower SES. The last three columns show average values and
192
their differences after the propensity score adjustment. I used ̂
193
the control group. Compared to the unadjusted differences, the adjusted differences are smaller
8
/(1 − ̂
) as a weight for
ACCEPTED MANUSCRIPT
and statistically insignificant, implying that treatment and control groups appear to be
195
comparable based on observable characteristics. Figure 2 shows the distributions of estimated
196
propensity scores by breastfeeding status, which are virtually the same. This also implies that the
197
treatment and control groups have similar observable characteristics conditional on the
198
propensity score.
199
RI PT
194
In Figure 3, I present a plot of the estimated propensity scores against the actual
proportion of breastfeeding in 100 equally sized cells. The graph shows that the 100 cells are
201
generally clustered around the 45-degree line, which implies that the estimated propensity score
202
predicts the actual proportion of breastfeeding fairly well.
M AN U
203
SC
200
To show the effects of maternal breastfeeding graphically, I plotted the averages of children’s cognitive test scores against the estimated propensity scores separately in 20 equally
205
sized cells by treatment status (Figure 4). The positive slopes indicate a general positive
206
relationship between maternal breastfeeding and children’s cognitive test scores. Among
207
children with similar propensity scores, children who were breastfed generally had higher test
208
scores than their counterparts who were not breastfed. This implies that maternal breastfeeding
209
has a positive impact on cognitive development among children with similar propensity scores.
EP
210
TE D
204
In panel A and B of Table 2, I present the estimated ATT and ATE. Maternal breastfeeding increases cognitive test scores by 1.04 and 0.94 points, respectively. Compared to
212
the standard deviation of the dependent variable, the magnitudes of the impact of maternal
213
breastfeeding on children’s cognitive development are around 10% of the standard deviation.
214
The estimated effects of maternal breastfeeding are also robust under other regression
215
specifications. In columns (1) and (2) of Appendix Table 1, instead of nonparametric estimations
216
with inverse probabilities, I used ̂
AC C
211
and ̂
8
9
as parametric controls and included an
ACCEPTED MANUSCRIPT
217
interaction term between maternal breastfeeding and ̂
218
between the propensity of maternal breastfeeding and children’s cognitive test scores,
219
respectively. Finally, in column (3), I added the control variables used in the propensity score
220
estimation to the parametric specifications.
RI PT
221
to capture differential relationships
In panel C of Table 2, I present the quantile regression coefficient values for cognitive test scores. The impact of maternal breastfeeding is around 1.1 to 1.2 points for children below
223
the 20th percentile of the cognitive test score distribution and around 0.5 points for children
224
above the 80th percentile of the cognitive test score distribution. The impacts of maternal
225
breastfeeding on children’s development are greater among children with lower cognitive test
226
scores.
M AN U
227
SC
222
Table 3 shows estimated effects of breastfeeding with different cutoffs. I used ever breastfeeding status and extended breastfeeding status (i.e., whether a child was breastfed for at
229
least 6 months). I used the same regression specification for the propensity score estimation.
230
Columns (1) and (2) show ATTs, ATEs, and quantile regression coefficient values. The
231
estimated ATTs and ATEs on cognitive test scores do not change. One thing to note is that the
232
distributional effects become larger for extended duration of breastfeeding. The impacts of ever
233
breastfeeding for those below the 20th percentile are about one and a half times larger than for
234
those above the 80th percentile. The impacts of extended maternal breastfeeding for those below
235
the 20th percentile are about five times larger than for those above the 80th percentile.
EP
AC C
236
TE D
228
Table 4 shows results of robustness checks. To begin with, I included multiple birth
237
sample because it has different characteristics compared to the baseline sample (Appendix Table
238
A2). Panels A, B, and C of column (1) show ATT, ATE, and quantile regression coefficient
239
values. The estimated effects of breastfeeding did not change much from the baseline results.
10
ACCEPTED MANUSCRIPT
Next, the propensity score approach could not infer the causal effects of breastfeeding because of
241
unmeasured confounding (Nickel, 2015). To examine any bias due to this, I included variables
242
about children and maternal health that were not accounted for from the baseline propensity
243
score estimation. First, I used birth weights and gestational age as measures for children’s health
244
because children’s birth weights and birth outcomes are important determinants of children’s
245
cognitive development (Almond & Currie, 2011b) and may affect maternal breastfeeding
246
decisions (Cunha & Heckman, 2007). Children with better birth outcomes as measured by birth
247
weight and gestational age are more likely to be breastfed (Appendix Table A3). Second,
248
maternal health status is also considered an important factor for breastfeeding (Jiang, Foster, &
249
Gibson-Davis, 2011). I used self-evaluated health status and a dummy variable of having any
250
chronic disease from the first wave of the MCS as measures of maternal health. Since there was
251
no direct information on maternal health status during breastfeeding, I used those variables as
252
proxies. Panels A and B of columns (2) and (3) show that the magnitudes of the ATTs and ATEs
253
are slightly smaller than my baseline regression results: the effects are around 0.9 and 0.8 points,
254
respectively, which again equals approximately 10% of the standard deviation of the cognitive
255
score distribution. Panel C of both columns shows that the impacts of maternal breastfeeding on
256
children’s development are still greater among children with lower cognitive test scores: the
257
impact of maternal breastfeeding is around 1.2 (1.0) points for children below the 20th percentile
258
of the cognitive test score distribution and around 0.3 (0.3) points for children above the 80th
259
percentile of the cognitive test score distribution.
SC
M AN U
TE D
EP
AC C
260
RI PT
240
261 262
4. Discussion
11
ACCEPTED MANUSCRIPT
263
Previous literature focused on the average effects of breastfeeding on children’s cognition. This study first examined the distributional impacts of maternal breastfeeding on children’s
265
development using a semiparametric quantile regression model. I found that the effects of
266
breastfeeding on children’s test scores were greater among children with lower test scores. The
267
effects were around 12–13% of the standard deviation, which is considered small in magnitude
268
according to Cohen’s effect size rule (Cohen, 1992) and the previous literature (Ip et al., 2007;
269
Jiang, Foster, & Gibson-Davis, 2011). However, the effects could become larger at the
270
population level as the disparity in educational achievement becomes larger within a society
271
(Rose, 1985). Given evidence that breastfeeding is associated with a reduction in several health
272
risks among children and mothers (Ip et al., 2007; Eidelman et al., 2012), the benefits of
273
breastfeeding could be larger when considering such health benefits (Cesur et al., 2017).
SC
M AN U
274
RI PT
264
The method and data suffer from some limitations. First, the propensity score approach is based on the ignorability assumption that there is no other confounding factor once a large
276
number of covariates are controlled for (Rosenbaum & Rubin, 1983). This assumption is too
277
strong to infer causal effects of breastfeeding due to unmeasured or unmeasurable confounding
278
(Nickel, 2015). As an example of unmeasured confounding, the MCS does not provide
279
information on mothers’ intelligence, such as IQ or test scores, or home environment scores,
280
which are considered important confounding factors (Der, Batty, & Deary, 2006; Belfield &
281
Kelly, 2012; Cesur et al., 2017). In addition, it is almost impossible from most observational data
282
to obtain information on unmeasurable factors such as maternal genes. Second, the data do not
283
provide enough information to construct an exclusive breastfeeding variable. Since the treatment
284
group consisted of children who were breastfed at all, this may underestimate the actual benefits
285
of exclusive breastfeeding (Jiang, Foster, & Gibson-Davis, 2011). Finally, breastfeeding duration
AC C
EP
TE D
275
12
ACCEPTED MANUSCRIPT
was based on maternal recall. This implies that there may be errors in treatment status and could
287
cause additional bias in the estimates. Since it is unclear if this type of error is random or
288
systematic (Promislow, Gladen, & Sandler, 2005), it is hard to see the exact direction of the bias.
289
An economic theory may provide a possible explanation for the main findings of this study.
RI PT
286
Let’s assume that children’s human capital production function is concave against maternal
291
breastfeeding. Low SES children usually have a low chance of being breastfed and so achieve
292
low test scores based on the underlying biological mechanism. Because they are less likely to be
293
breastfed, marginal returns on breastfeeding could be larger than those in higher SES children
294
due to the concavity of the human capital production function. These results have a policy
295
implication for the promotion of breastfeeding among individuals with lower SES. Significant
296
SES-related disparities in children’s academic performance are well established in the UK
297
(Banerjee, 2016). Government programs such as Sure Start have been introduced to narrow these
298
gaps (Melhuish et al., 2008). Since a high proportion of children with lower test scores generally
299
come from families with low SES and since there are observed SES-related disparities in
300
maternal breastfeeding incidence and duration, government policies such as the UNICEF UK
301
Baby Friendly Initiative aims at promoting breastfeeding may reduce SES-related gaps in
302
children’s academic development. However, there are at least two reasons why this policy
303
implication should be interpreted with caution. First, the policy implication of this study is based
304
on results with strong statistical assumptions. It is hard to draw causal inferences regarding the
305
effects of breastfeeding. Second, the results may not be generalizable. Public agencies such as
306
the American Academy of Pediatrics and the WHO recommend that mothers exclusively
307
breastfeed their infants for at least 6 months and ideally continue to breastfeed beyond 6 months.
308
Under the Affordable Care Act in the United States, the initiation and extended duration of
AC C
EP
TE D
M AN U
SC
290
13
ACCEPTED MANUSCRIPT
309
maternal breastfeeding are promoted. However, this study is based on data from the UK. The
310
results should therefore be cautiously applied to other countries owing to significant differences
311
in health care and other social policy. As an avenue for future research, the distributional effects on other benefits of
313
breastfeeding, such as the health benefits for children and mothers, could also be considered.
314
Previous literature has mainly focused on the average effects of breastfeeding, but there is little
315
evidence of the distributional effects on children’s and maternal health. To draw a more
316
convincing causal inference on the effects of breastfeeding, it would be useful to exploit a natural
317
experiment that quasi-randomizes breastfeeding status or duration.
M AN U
SC
RI PT
312
AC C
EP
TE D
318
14
ACCEPTED MANUSCRIPT
324 325 326 327 328
Labor Economics, 4, 1315-1486.
RI PT
323
Almond, D. & Currie, J. (2011a). Human capital development before age five. Handbook of
Almond, D. & Currie, J. (2011b). Killing me softly: The fetal origins hypothesis. The Journal of Economic Perspectives, 25(3), 153-172.
Anderson, J.W., Johnstone, B.M., & Remley, D.T. (1999). Breast-feeding and cognitive
development: a meta-analysis. The American Journal of Clinical Nutrition, 70(4), 525-535.
SC
322
References
Banerjee, P.A. (2016). A systematic review of factors linked to poor academic performance of
M AN U
319 320 321
disadvantaged students in science and maths in schools. Cogent Education, 3(1), 1178441.
329
Becker, G. (1981). A treatise on the family. Cambridge, MA: Harvard University Press.
330
Belfield, C.R. & Kelly, I.R. (2012). The benefits of breastfeeding across the early years of
332 333 334
childhood. Journal of Human Capital, 6(3), 251-277.
Borra, C., Iacovou, M., & Sevilla, A. (2012). The effect of breastfeeding on children's cognitive
TE D
331
and noncognitive development. Labour Economics, 19(4), 496-515. Cesur, R., Sabia, J.J., Kelly, I.R., & Yang, M. (2017). The effect of breastfeeding on young adult wages: new evidence from the add health. Review of Economics of the Household, 15(1),
336
25-51.
AC C
337
EP
335
Clandinin, M.T., Chappell, J.E., Heim, T., Swyer, P.R., & Chance, G.W. (1981). Fatty acid
338
utilization in perinatal de novo synthesis of tissues. Early Human Development, 5(4), 355-
339
366.
340
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155.
15
ACCEPTED MANUSCRIPT
341
Colen, C.G. & Ramey, D.M. (2014). Is breast truly best? Estimating the effects of breastfeeding
342
on long-term child health and wellbeing in the United States using sibling
343
comparisons. Social Science & Medicine, 109, 55-65.
346 347
Review, 97(2), 31.
RI PT
345
Cunha, F. & Heckman, J.J. (2007). The technology of skill formation. The American Economic
Cunha, F., Heckman, J.J., & Schennach, S.M. (2010). Estimating the technology of cognitive and noncognitive skill formation. Econometrica, 78(3), 883-931.
SC
344
Dehejia, R.H. & Wahba, S. (1999). Causal effects in nonexperimental studies: Reevaluating the
349
evaluation of training programs. Journal of the American Statistical Association, 94(448),
350
1053-1062.
351
M AN U
348
Del Bono, E. & Rabe, B. (2012). Breastfeeding and child cognitive outcomes: Evidence from a hospital-based breastfeeding support policy (No. 2012-29). ISER Working Paper Series,
353
2012-29, http://hdl.handle.net/10419/91672.
354
TE D
352
Deoni, S.C.L., Dean III, D.C., Piryatinksy, I., O’Muircheartaigh, J., Waskiewicz, N., Lehman, K., Han, M., & Dirks, H. (2013). Breastfeeding and early white matter development: A
356
cross sectional study. NeuroImage, 82, 77-86.
357
EP
355
Der, G., Batty, D., & Deary, I.J. (2006). Effect of breast feeding on intelligence in children: Prospective study, sibling pairs analysis, and meta-analysis. British Medical Journal, 333,
359
945-948.
360
AC C
358
Eidelman, A.I., Schanler, R.J., Johnston, M., Landers, S., Noble, L., Szucs, K., & Viehmann, L.
361
(2012). Breastfeeding and the use of human milk. Pediatrics, 129(3), e827-e841.
362
Elliott, C.D., Smith, P., & McCulloch, K. (1997). British ability scales, second ed. (BAS II):
363
Technical manual. London, UK: NFER-Nelson.
16
ACCEPTED MANUSCRIPT
368 369 370 371 372 373 374 375 376
Lipids, 34(2), 161-169.
RI PT
367
Fernstrom, J.D. (1999). Effects of dietary polyunsaturated fatty acids on neuronal function.
Firpo, S. (2007). Efficient semiparametric estimation of quantile treatment effects. Econometrica, 75(1), 259-276.
Fitzsimons, E. & Vera-Hernandez, M. (2012). The causal effects of breastfeeding on children’s
SC
366
low-income countries. Washington, DC: The World Bank.
development. London, UK: Institute for Fiscal Studies.
Heckman, J.J. (2006). Skill formation and the economics of investing in disadvantaged
M AN U
365
Fernald, L.C., Kariger, P., Engle, P., & Raikes, A. (2009). Examining early child development in
children. Science, 312(5782), 1900-1902.
Hirano, K., Imbens, G.W., & Ridder, G. (2003). Efficient estimation of average treatment effects using the estimated propensity score. Econometrica, 71(4), 1161-1189. Ip, S., Chung, M., Raman, G., Chew, P., Magula, N., DeVine, D., et al., 2007. Breastfeeding and
TE D
364
maternal and infant health outcomes in developed countries. Evidence report/technology
378
assessment no. 153 (Prepared by Tufts-New England Medical Center Evidence-based
379
Practice Center, under Contract No. 290-02-0022). AHRQ Publication No. 07-E007.
380
Rockville, MD: Agency for Healthcare Research and Quality. Jiang, M., Foster, E.M., & Gibson-Davis, C.M. (2011). Breastfeeding and the child cognitive
AC C
381
EP
377
382
outcomes: A propensity score matching approach. Maternal and Child Health Journal,
383
15(8), 1296-1307.
384 385
Kling, J.R., Liebman, J.B., & Katz, L.F. (2007). Experimental analysis of neighborhood effects. Econometrica, 75(1), 83-119.
17
ACCEPTED MANUSCRIPT
386
Kramer, M.S., Chalmers, B., Hodnett, E.D., Sevkovskaya, Z., Dzikovich, I., Shapiro, S., Collet, J.P., Vanilovich, I., Mezen, I., Ducruet, T., & Shishko, G. (2001). Promotion of
388
breastfeeding intervention trial (PROBIT): A randomized trial in the Republic of
389
Belarus. Journal of American Medical Association, 285(4), 413-420.
RI PT
387
Kramer, M.S., Aboud, F., Mironova, E., Vanilovich, I., Platt, R.W., Matush, L., Igumnov, S., et
391
al. (2008). Breastfeeding and child cognitive development: New evidence from a large
392
randomized trial. Archives of General Psychiatry, 65(5), 578-584.
SC
390
Melhuish, E., Belsky, J., Leyland, A.H., Barnes, J., & National Evaluation of Sure Start Research
394
Team. (2008). Effects of fully-established Sure Start Local Programmes on 3-year-old
395
children and their families living in England: a quasi-experimental observational study. The
396
Lancet, 372(9650), 1641-1647.
400 401 402 403 404 405 406 407 408
TE D
399
research? Journal of Human Lactation, 31(2), 209-212.
Promislow, J.H., Gladen, B.C., & Sandler, D.P. (2005). Maternal recall of breastfeeding duration by elderly women. American Journal of Epidemiology, 161(3), 289-296. Rees, D.I. & Sabia, J.J. (2009). The effect of breast feeding on educational attainment: Evidence
EP
398
Nickel, N.C. (2015). Look before you leap: can we draw causal conclusions from breastfeeding
from sibling data. Journal of Human Capital, 3(1), 43-72. Rose, G. (1985). Sick individuals and sick populations. International Journal of
AC C
397
M AN U
393
Epidemiology, 14(1), 32-38. Rosenbaum, P.R. & Rubin, D.B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41-55. Rothstein, D.S. (2012). Breastfeeding and children's early cognitive outcomes. Review of Economics and Statistics, 95(3), 919-931.
18
ACCEPTED MANUSCRIPT
409 410
Zhao, Z. (2008). Sensitivity of propensity score methods to the specifications. Economics Letters, 98(3), 309-319.
AC C
EP
TE D
M AN U
SC
RI PT
411
19
ACCEPTED MANUSCRIPT
Tables and Figures Table 1. Baseline Characteristics of Control Variables No propensity score adjustment Breastfeeding Difference Y N Y-N (1) (2) (3) 30.97 28.31 2.66*** .39 .67 -.28*** .38 .58 -.20*** .56 .53 .03*** .82 .88 -.06*** .97 .95 .02*** .69 .60 .09*** .98 .97 .01*** .43 .34 .09*** .10 .26 -.16*** .35 .39 -.04*** 4.82 5.12 -.31*** .21 .37 -.16*** .05 .14 -.09*** 33.63 31.06 2.57*** .41 .66 -.25***
Propensity score adjustment Breastfeeding Difference Y N Y-N (4) (5) (6) 30.97 30.97 .00 .39 .39 .00 .38 .38 .00 .56 .56 .00 .82 .80 .02 .97 .97 .00 .69 .69 .00 .98 .97 .01 .43 .43 .00 .1 .09 .01 .35 .36 -.01 4.82 4.82 .00 .21 .21 .00 .05 .05 .00 33.63 33.64 -.01 .41 .41 .00
RI PT
412 413 414 415
TE D
EP AC C
416 417 418 419 420 421 422
M AN U
SC
Variables Maternal age at birth Low education Low household income Working (after birth) White Married at pregnancy Planned pregnancy Prenatal care (ever) Prenatal class (ever) Smoking during pregnancy Any complications during pregnancy Alcohol consumption Any government support Number of smokers in household Spouse’s age at delivery Low education (spouse) Data source: The MCS. Note: I report differences in selected characteristics by maternal breastfeeding status. The first three columns show average values and their differences without any adjustment. The last three columns show average values and their differences after the propensity score adjustment. I used ̂ /(1 − ̂ ) as a weight for the control group. ***p < .01, **p < .05, * p < .10.
20
ACCEPTED MANUSCRIPT
Table 2. Distributional Effects of Maternal Breastfeeding on Children’s Cognitive Test Scores Breastfeeding ≥ 4 months (1)
A. ATT 1.04*** (0.22) B. ATE
SC
0.94*** (0.27)
0.3 0.4 0.5
1.08* (0.62) 1.23*** (0.46) 1.17*** (0.38) 1.23*** (0.33) 1.17*** (0.29) 1.00*** (0.28) 0.83*** (0.26) 0.50* (0.27) 0.50* (0.30)
M AN U
C. Quantiles 0.1 0.2
RI PT
Treatment status
TE D
0.6 0.7 0.8
EP
0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
Data source: The MCS. Note: For treatment variable, I use the dummy variable if a child was breastfed for at least 4 months. The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
21
ACCEPTED MANUSCRIPT
Table 3. Distributional Effects of Different Durations of Maternal Breastfeeding on Children’s Cognitive Test Scores
(1) A. ATT 1.26*** (0.31) B. ATE 1.16*** (0.27)
0.2 0.3 0.4
0.7 0.8
0.83** (0.33)
TE D
0.5 0.6
1.33** (0.55) 1.5*** (0.50) 1.17*** (0.38) 1.17*** (0.37) 1.13*** (0.33) 1.33*** (0.31) 1.00*** (0.36) 1.00*** (0.38) 0.93** (0.44)
0.96*** (0.24)
M AN U
C. Quantiles 0.1
Breastfeeding ≥ 6 months (2)
RI PT
Ever Breastfeeding
SC
Treatment status
EP
0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
1.42* (0.82) 1.58*** (0.56) 1.50*** (0.43) 1.00*** (0.39) 1.13*** (0.36) 1.10*** (0.33) 0.70** (0.29) 0.27 (0.29) 0.33 (0.35)
Data source: The MCS. Note: In each column, I estimate the effects of breastfeeding using different cutoffs. For treatment variable, I use the dummy variable if a child was ever breastfed in columns (1), and at least 6 months in column (2). The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
22
ACCEPTED MANUSCRIPT
Table 4. Robustness Checks
Including birth outcomes
(1)
(2)
A. ATT 1.06*** (0.22)
0.90*** (0.23)
B. ATE
0.4 0.5 0.6 0.7 0.8 0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
M AN U
0.3
1.00 (0.65) 1.17*** (0.45) 1.17*** (0.39) 1.17*** (0.33) 1.00*** (0.30) 1.00*** (0.28) 0.73*** (0.26) 0.33 (0.27) 0.50* (0.30)
TE D
0.2
1.08* (0.62) 1.23*** (0.46) 1.17*** (0.38) 1.23*** (0.33) 1.17*** (0.29) 1.00*** (0.28) 0.83*** (0.26) 0.50* (0.27) 0.50* (0.30)
EP
C. Quantiles 0.1
0.81*** (0.26)
SC
0.98*** (0.26)
RI PT
Including multiple births
Including both birth outcomes and maternal health (3)
61.52 (10.05)
0.91*** (0.23)
0.79*** (0.27)
0.83 (0.65) 1.00** (0.46) 1.00*** (0.39) 1.16*** (0.34) 0.83*** (0.30) 0.92*** (0.28) 0.73*** (0.26) 0.33 (0.27) 0.50* (0.30) 61.52 (10.05)
Data source: The MCS. Note: In each column, I estimate the effects of breastfeeding using different specifications. I include multiple births in column (1), and children’s birth weights and gestational age in column (2), and maternal health status as well as birth weights and gestational age in column (3). For treatment variable, I use the dummy variable if a child was breastfed for at least 4 months. The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
23
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 1. Distribution of Children’s Cognitive Development
AC C
EP
TE D
Data source: The MCS.
24
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 2. Distribution of Estimated Propensity Score by Breastfeeding Status
AC C
EP
TE D
Data source: The MCS.
25
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 3. Estimated Propensity Score and Actual Probability of Breastfeeding (≥ 4 Months)
AC C
EP
TE D
Data source: The MCS. Note: This figure shows a plot of the average estimated propensity scores against the actual probability of breastfeeding in 100 equal-sized cells. The red line is the 45-degree line.
26
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 4. Children’s Cognitive Development as a Function of Estimated Propensity Score
AC C
EP
TE D
Data source: The MCS. Note: This figure shows a plot of average cognitive test scores against estimated propensity scores in 20 equal-sized cells by maternal breastfeeding status.
27
ACCEPTED MANUSCRIPT
Appendix Table A1. Effects of Maternal Breastfeeding on Children’s Cognitive Test Scores
Breastfed ≥ 4 months
(2)
(3)
RI PT
(1) 1.022***
1.032**
1.090***
(0.207)
(0.496)
(0.214)
Y
Y Y
Y
SC
Parametric control of ̂ Breastfed4mon× Control variables ( Data source: The MCS.
Y
AC C
EP
TE D
M AN U
Note: Breastfed≥4months indicates whether a child was breastfed for at least 4 months. The cognitive test score measure is the average of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture 8 similarity, and word reading. In column (1), I use ̂ and ̂ as controls instead of using sampling weights. In column (2), I add Breastfed4mon× ̂ to the regression specification of column (1). In column (3), I add 8 ̂ and ̂ and the control variables that were used to estimate ̂ to the regression specification of column (1). Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
28
ACCEPTED MANUSCRIPT
Table A2. Characteristics of Baseline and Excluded Sample
AC C
EP
TE D
M AN U
SC
RI PT
Baseline Excluded Excluded - Baseline Variables (1) (2) (3) Maternal age at birth 28.29 30.81 2.52*** Low education .61 .56 -.05* Low household income .61 .56 -.05 Working (after birth) .47 .44 -.03 White .84 .89 .05** Married at pregnancy .81 .88 .07*** Planned pregnancy .54 .63 .09*** Prenatal care (ever) .96 .97 .01 Prenatal class (ever) .34 .36 .02 Smoking during pregnancy .25 .18 -.07*** Any complications during pregnancy .38 .48 .1*** Alcohol consumption 5.14 4.87 -.27 Any government support .43 .43 .00 Number of smokers in household .14 .09 -.05** Spouse’s age at delivery 31.88 33.60 1.72*** Low education (spouse) .40 .37 -.03 Data source: The MCS. Note: I used the same variables that I used in Table 1. ***p < .01, **p < .05, and * p < .10.
29
ACCEPTED MANUSCRIPT
Table A3. Maternal Breastfeeding and Children’s Birth Outcomes Pr (Birth weight ≤ 2,500 g)
Pr (Birth weight ≤ 1,500 g)
Gestational age (days)
(1)
(2)
(3)
(4)
-0.030***
-0.005**
(0.006)
(0.002)
3199.442
0.094
0.013
12788 0.005
12796 0.003
12796 0.001
(12.932) Average of control group Observations R-squared
M AN U
Data source: The MCS.
1.687***
-0.019***
(0.286)
(0.005)
275.617
SC
77.607***
Breastfed ≥ 4 months
Pr (Gestational age ≤ 37 weeks) (5)
RI PT
Birth weight (grams)
12703 0.004
0.069
12703 0.002
AC C
EP
TE D
Note: Breastfed ≥ 4 months indicates whether a child was breastfed for at least 4 months. I used ̂ /(1 − ̂ ) as a sampling weight for the control group. Robust standard errors are calculated and presented in parentheses. ***p < .01, **p < .05, and * p < .10.
30
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Highlights • Study of the effects of breastfeeding on the distribution of children’s test scores. • A semiparametric quantile regression is used for the estimation. • The effects are greater for children with lower test scores.
S0277-9536(17)30380-5
DOI:
10.1016/j.socscimed.2017.06.012
Reference:
SSM 11276
To appear in:
Social Science & Medicine
Received Date: 28 November 2016 Revised Date:
30 May 2017
Accepted Date: June 2017
Please cite this article as: Koh, K., Maternal breastfeeding and children's cognitive development, Social Science & Medicine (2017), doi: 10.1016/j.socscimed.2017.06.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
SSM Manuscript Number: SSM-D-16-03615R1
RI PT
Maternal Breastfeeding and Children’s Cognitive Development∗
AC C
EP
TE D
M AN U
SC
Kanghyock Koh+ Ulsan National Institute of Science and Technology (UNIST)
∗
I thank Anna Aizer, Kenneth Chay, Hyojin Han, Hyunjoo Yang, and the editor and four anonymous referees for valuable comments. Any remaining errors are mine. This work was supported by the New Faculty Research Fund(1.160080) of UNIST(Ulsan National Institute of Science & Technology). + 50 UNIST-gil, Ulju-gun, Ulsan 44919, Republic of Korea. Email: [email protected]
ACCEPTED MANUSCRIPT
ABSTRACT
SC
RI PT
Do children with lower test scores benefit more from breastfeeding than those with higher scores? In this paper, I examine the distributional effects of maternal breastfeeding on the cognitive test scores of 11,544 children who were born in 2000 and 2001 in the United Kingdom using a semiparametric quantile regression model. I find evidence that maternal breastfeeding has larger positive impacts on children with lower test scores. Effects for children below the 20th percentile are about 2–2.5 times greater than those for children above the 80th percentile. I also find that these distributional effects are larger when the duration of breastfeeding is extended. One policy implication is that a public policy aims at promoting breastfeeding might narrow a disparity in children’s cognition.
Keywords: United Kingdom, distributional effects, maternal breastfeeding, children’s development, semiparametric quantile regression
M AN U
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Highlights • Study of the effects of breastfeeding on the distribution of children’s test scores. • A semiparametric quantile regression is used for the estimation. • The effects are greater for children with lower test scores.
TE D
22
1. Introduction
24
Economic theories of human capital have emphasized the importance of parental investment in
25
children (Becker, 1981; Cunha & Heckman, 2007). A large number of empirical studies show
26
that human capital developments in early childhood (e.g., cognitive ability) play a significant
27
role in human capital developments later in life, as measured by educational attainment,
28
employment, wages, etc. (Heckman, 2006; Cunha & Heckman, 2007; Cunha, Heckman, &
29
Schennach, 2010; Almond & Currie, 2011a,b).
AC C
30
EP
23
Maternal breastfeeding has been emphasized as an influential factor in early childhood
31
development (e.g., Kramer et al., 2001, 2008) based on biological mechanisms through which
32
maternal breastfeeding aids children’s development. First, the composition of breast milk is
33
superior to that of formula. Breast milk contains long-chain polyunsaturated fatty acids, such as 1
ACCEPTED MANUSCRIPT
docosahexaenoic acid (DHA) and arachidonic acid (AA), which form the major structures of
35
neuronal membranes and play critical roles in nervous system functioning by positively
36
stimulating development of the human brain (Fernstrom, 1999). Infants require sufficient
37
amounts of these acids during the first few months after birth (Clandinin et al, 1981). Note that
38
the compositional superiority of breast milk may be reduced because DHA and AA are now
39
added to formula. However, Fitzsimons and Vera-Hernandez (2012) reported that DHA and AA
40
were not included in formula in the UK (the context of this paper) until most of the infants who
41
are considered in the empirical analysis were born. Deoni et al. (2013) provided MRI evidence
42
that breastfed infants exhibit better development in specific brain areas associated with language
43
and visual reception abilities compared to those who are fed formula or a mixture of breast milk
44
and formula. Second, skin-to-skin contact between mother and child stimulates maternal
45
hormonal responses such as the production of prolactin and oxytocin, which may indirectly
46
improve cognitive development (Del Bono & Rabe, 2012).
SC
M AN U
TE D
47
RI PT
34
In spite of this clear biological mechanism, empirical evidence for the effects of breastfeeding on children’s cognitive development has been conflicting. A comprehensive
49
review by the Agency for Healthcare Research and Quality (Ip et al., 2007) summarized 400
50
articles (out of 9000 abstracts) and found that breastfeeding has few or small effects on
51
children’s cognitive ability. Many studies are based on observational data, and it is difficult to
52
infer causality due to many potential confounding variables, such as duration of breastfeeding,
53
children’s demographics and birth outcomes (such as birth weight and gestational age), parental
54
health-related behavior, socioeconomic status (SES), intelligence, and family characteristics
55
(Anderson, Johnstone, & Remley, 1999).
AC C
EP
48
2
ACCEPTED MANUSCRIPT
Researchers use various empirical strategies to address the endogeneity issue. First, they
57
estimate the probability of breastfeeding using detailed information regarding the characteristics
58
of children, parents, and families. They then estimate the effects of breastfeeding on cognitive
59
development by comparing children with similar propensities of maternal breastfeeding (Jiang,
60
Foster, & Gibson-Davis, 2011; Belfield & Kelly, 2012; Borra, Iacovou, & Sevilla, 2012;
61
Rothstein, 2012; Cesur et al, 2017). For example, Jian, Foster, & Gibson-Davis (2011) used the
62
Panel Study of Income Dynamics in the U.S. and found that the positive associations between
63
breastfeeding initiation and children’s cognitive development measured by the Woodcock
64
Johnson Psycho-Educational Battery Revised (WJ-R) and Wechsler Intelligence Scale for
65
Children-Revised (WISC-R) were significantly reduced when observational characteristics were
66
controlled for. The estimated associations are between one-tenth and one-fifth of a standard
67
deviation, which is small in magnitude. Second, some researchers use sibling or family fixed
68
effects to control for unobserved family characteristics (Der, Batty, & Deary, 2006; Rees &
69
Sabia, 2009; Colen & Ramey, 2014; Cesur et al., 2017). Recent research points out that maternal
70
intelligence or IQ and home environment explain breastfeeding status better than other SES
71
factors such as income or education, but few previous studies have accounted for these
72
covariates (Der, Batty, & Deary, 2006). Many studies using the propensity score approach or
73
fixed effects analysis have found that a positive association became smaller or disappeared once
74
a large number of confounding factors were controlled for. Researchers pointed out these results
75
as evidence for the endogeneity issue in breastfeeding (Belfield & Kelly, 2012; Cesur et al.,
76
2017). Third, economists use exogenous sources of variation in breastfeeding status as
77
instrumental variables to study the causal relationship between breastfeeding and children’s
78
cognitive development (Del Bono & Rabe, 2012; Fitzsimons & Vera-Hernandez, 2012). For
AC C
EP
TE D
M AN U
SC
RI PT
56
3
ACCEPTED MANUSCRIPT
example, infant feeding support is provided by midwives and nurses in the U.K. Since staff
80
working hours are reduced during the weekend, this support is reduced as well. Mothers are less
81
exposed to support when they give birth on Friday or Saturday. Fitzsimons and Vera-Hernandez
82
(2012) use this institutional fact to estimate the effects of breastfeeding on children’s cognitive
83
development among low-income mothers and their children. Finally, Kramer et al. (2001, 2008)
84
studied the causal effects of the Promotion of Breastfeeding Intervention Trial (PROBIT), which
85
randomly provided health care worker assistance for the initiation and maintenance of
86
breastfeeding, on breastfeeding initiation and duration and consequences for children’s health
87
and cognitive abilities in Belarus.
SC
M AN U
88
RI PT
79
Previous literature mainly focused on the average effects of breastfeeding on children’s cognition. Many measures for cognition, such as test scores, are continuously distributed. These
90
distributions could provide a deeper understanding of the effects of breastfeeding on children’s
91
cognition by evaluating the effects at different quantiles in addition to the average effects. For
92
example, children at lower quantiles could benefit more from breastfeeding than average
93
children or children at higher quantiles, even though the average effects of breastfeeding are
94
small or insignificant based on the existing literature. However, there is little knowledge
95
regarding the effects of breastfeeding on the distribution of cognition. To fill this gap, I
96
investigated the effects of maternal breastfeeding on the distribution of children’s cognitive
97
ability in this study. More specifically, I estimated the effects of breastfeeding at different
98
quantiles of cognitive test scores using a semiparametric quantile regression model. To address
99
the endogeneity of breastfeeding, I used the propensity score as the inverse probability
AC C
EP
TE D
89
100
(Rosenbaum & Rubin, 1983; Hirano, Imbens, & Ridder, 2003). Based on rich information about
101
children’s characteristics, cognitive test scores, and maternal breastfeeding from the UK
4
ACCEPTED MANUSCRIPT
102
Millennium Cohort Survey, I compared cognitive test scores among children with similar
103
propensities for maternal breastfeeding based on observed parental characteristics.
RI PT
104 105
2. Method
107
2.1 Data
108
To investigate the effects of maternal breastfeeding on children’s development, I used data from
109
the Millennium Cohort Study (MCS), a longitudinal study of about 18,500 children who were
110
born in 2000 and 2001 in the United Kingdom. Specifically, I used data from four surveys (at
111
ages 9 months, 3 years, 5 years, and 7 years) and excluded data for multiple births and children
112
who were not living with their biological mothers at the time of the first interview (n = 293).
113
Since I used secondary observational data, ethics approval was not required.
M AN U
TE D
114
SC
106
I collected data on self-reported maternal breastfeeding status from the first MCS survey, which included a question about the duration of maternal breastfeeding. Mothers reported the age
116
of babies when breastfeeding stopped, as measured in days, weeks, or months, which was used
117
for calculating the duration of breastfeeding. It was not feasible to obtain data on exclusive
118
breastfeeding from the MCS because survey questions related to infant feeding were not
119
exhaustive. Since the recommended duration of breastfeeding in the UK was 4 months in 2000, I
120
created a binary breastfeeding variable indicating whether children were breastfed for at least 4
121
months. To study the effects of breastfeeding with different duration cutoffs, I examined the
122
effects of initiation of breastfeeding. Given current policy recommendations [e.g., from the
123
World Health Organization (WHO)], I also examined the effects of extended breastfeeding (i.e.,
124
for at least 6 months).
AC C
EP
115
5
ACCEPTED MANUSCRIPT
125
To assess children’s cognitive development, I used scores from six British Ability Scale (BAS) tests, which are used to measure the cognitive abilities of children aged 2.5 to 8 years old
127
(Elliott, Smith, & McCulloch, 1997). Since these tests are individually administered by trained
128
interviewers, the scores provide a more accurate measure than parents’ self-reported measures
129
(Fernald et al., 2009). Using the age-adjusted scores from the MCS, I calculated average scores
130
within and across ages and created a summary index. I did this to address a concern that one null
131
hypothesis could be rejected simply because I tested it with multiple null hypotheses (Kling,
132
Liebman, & Katz, 2007). This also yields a well-behaved continuous distribution of test scores
133
(Figure 1).
SC
M AN U
134
RI PT
126
I also collected information on maternal demographics, maternal prenatal characteristics, and spouse and family characteristics as control variables for the propensity score approach:
136
mother’s age at birth, race, marital status, and education; planned pregnancy; dummy variables
137
for receiving any prenatal care and attending any prenatal classes; timing of initial care; dummy
138
variables for any complications during pregnancy (including bleeding or threatened miscarriage
139
in early pregnancy; bleeding in later pregnancy; pregnancy diagnosed as twins, triplets, or more;
140
persistent vomiting; elevated blood pressure; eclampsia/preeclampsia or toxemia; urinary
141
infection; diabetes; too much fluid around the baby; suspected slow growth of baby; and other
142
suspected problems) and labor (including breech birth, shoulder first, very long (or rapid) labor,
143
fetal distress – heart rate sign, meconium sign, and other complications); smoking status during
144
pregnancy; working status during and after pregnancy; frequency of alcohol consumption;
145
dummy variables for countries (England, Wales, Scotland, and Northern Ireland); father’s age at
146
birth, education level, and job status; dummy variables for joint job status of a couple indicating
147
if both in work, father (mother) in work but mother (father) not in work, both not in work, father
AC C
EP
TE D
135
6
ACCEPTED MANUSCRIPT
(mother) in work or on leave but mother (father) not in work, and father (mother) in work but
149
mother (father) is unknown; number of household members and smokers in the household; and
150
dummy variables for government support and income level. The mothers’ and fathers’ education
151
levels were categorized as NVQ level 1 to 5 and overseas. I defined a dummy variable of low
152
education level if the NVQ level was 3 or lower to construct interactions with other covariates.
153
Due to missing values in these control variables, I used 11,540 children in the main analysis.
RI PT
148
SC
154 155
2.2 Empirical Strategy
156
If
157
and
158
effect on the treated (ATT) is
159
being treated based on observed characteristics be the propensity score (Rosenbaum & Rubin,
160
1983),
161
as maternal characteristics, parental characteristics, and household environment. I include control
162
variables mentioned in the previous section. I also include polynomials and interaction terms
163
among them. I estimated the propensity scores using the logit regression specification and used
164
their inverse probability weights to nonparametrically estimate the ATT (Hirano, Imbens, &
165
Ridder, 2003). I used ̂
166
be constant, I also estimated average treatment effects (ATE) using 1/ ̂
167
sampling weights for treatment and control groups, respectively. For statistical inferences, I
168
calculated robust standard errors to correct for heteroscedasticity in test scores by breastfeeding
169
status. The propensity score approach has some advantages over the ordinary least squares (OLS)
170
approach with many control variables. First, I can alleviate misspecification errors by not
is the outcome if treated (breastfed for more than 4 months),
M AN U
is the treatment indicator,
is the outcome if untreated (breastfed for less than 4 months), then the average treatment
| =
, where
| = 1 . Let the conditional probability of
is a vector of observed pretreatment characteristics such
EP
TE D
= Pr
| =1 −
AC C
/(1 − ̂
) as a weight for the control group. Since the ATT may not
7
and 1/ 1 − ̂
as
ACCEPTED MANUSCRIPT
assuming any parametric relationship between breastfeeding and children’s cognitive
172
development (Zhao, 2008). Second, I can address the curse of dimensionality issue associated
173
with OLS when many control variables are included (Dehejia & Wahba, 1999) by summarizing
174
all information from the control variables into a single index (i.e., the propensity score). Third, I
175
can graphically examine the nonparametric relationship between maternal breastfeeding and
176
children’s cognitive development by using the propensity score.
Then I investigated the distributional impact of maternal breastfeeding using a
SC
177
RI PT
171
178
semiparametric quantile regression model. Following Firpo (2007), I estimated the quantile
179
treatment effect for quantile using ∆ ≡
−
( = 0, 1 , where
M AN U ,
,
,
is defined as
180
!"#$% ∑, ( ∙* +- '
181
− 1/. ≤ 01 , and ' ( is the inverse probability of the estimated propensity score for the ATT, 2
+
−
, the check function * ∙ at a real number . is * . = . ∙
which are ' ( = ,⋅4
183
propensity scores using local logit regression. To estimate treatment effects at different quantiles,
184
the outcome variable should be continuous and well behaved without spikes or gaps, which is
185
again confirmed by Figure 1.
188 189
TE D
where 7 is the sample size. I estimated the
EP
187
64 5
AC C
186
5
and ' ( = ,⋅
62
182
3. Empirical Results
Table 1 shows summary statistics by breastfeeding status. The first three columns show
190
that in general, a statistically significant proportion of children who were breastfed for less than 4
191
months came from families with lower SES. The last three columns show average values and
192
their differences after the propensity score adjustment. I used ̂
193
the control group. Compared to the unadjusted differences, the adjusted differences are smaller
8
/(1 − ̂
) as a weight for
ACCEPTED MANUSCRIPT
and statistically insignificant, implying that treatment and control groups appear to be
195
comparable based on observable characteristics. Figure 2 shows the distributions of estimated
196
propensity scores by breastfeeding status, which are virtually the same. This also implies that the
197
treatment and control groups have similar observable characteristics conditional on the
198
propensity score.
199
RI PT
194
In Figure 3, I present a plot of the estimated propensity scores against the actual
proportion of breastfeeding in 100 equally sized cells. The graph shows that the 100 cells are
201
generally clustered around the 45-degree line, which implies that the estimated propensity score
202
predicts the actual proportion of breastfeeding fairly well.
M AN U
203
SC
200
To show the effects of maternal breastfeeding graphically, I plotted the averages of children’s cognitive test scores against the estimated propensity scores separately in 20 equally
205
sized cells by treatment status (Figure 4). The positive slopes indicate a general positive
206
relationship between maternal breastfeeding and children’s cognitive test scores. Among
207
children with similar propensity scores, children who were breastfed generally had higher test
208
scores than their counterparts who were not breastfed. This implies that maternal breastfeeding
209
has a positive impact on cognitive development among children with similar propensity scores.
EP
210
TE D
204
In panel A and B of Table 2, I present the estimated ATT and ATE. Maternal breastfeeding increases cognitive test scores by 1.04 and 0.94 points, respectively. Compared to
212
the standard deviation of the dependent variable, the magnitudes of the impact of maternal
213
breastfeeding on children’s cognitive development are around 10% of the standard deviation.
214
The estimated effects of maternal breastfeeding are also robust under other regression
215
specifications. In columns (1) and (2) of Appendix Table 1, instead of nonparametric estimations
216
with inverse probabilities, I used ̂
AC C
211
and ̂
8
9
as parametric controls and included an
ACCEPTED MANUSCRIPT
217
interaction term between maternal breastfeeding and ̂
218
between the propensity of maternal breastfeeding and children’s cognitive test scores,
219
respectively. Finally, in column (3), I added the control variables used in the propensity score
220
estimation to the parametric specifications.
RI PT
221
to capture differential relationships
In panel C of Table 2, I present the quantile regression coefficient values for cognitive test scores. The impact of maternal breastfeeding is around 1.1 to 1.2 points for children below
223
the 20th percentile of the cognitive test score distribution and around 0.5 points for children
224
above the 80th percentile of the cognitive test score distribution. The impacts of maternal
225
breastfeeding on children’s development are greater among children with lower cognitive test
226
scores.
M AN U
227
SC
222
Table 3 shows estimated effects of breastfeeding with different cutoffs. I used ever breastfeeding status and extended breastfeeding status (i.e., whether a child was breastfed for at
229
least 6 months). I used the same regression specification for the propensity score estimation.
230
Columns (1) and (2) show ATTs, ATEs, and quantile regression coefficient values. The
231
estimated ATTs and ATEs on cognitive test scores do not change. One thing to note is that the
232
distributional effects become larger for extended duration of breastfeeding. The impacts of ever
233
breastfeeding for those below the 20th percentile are about one and a half times larger than for
234
those above the 80th percentile. The impacts of extended maternal breastfeeding for those below
235
the 20th percentile are about five times larger than for those above the 80th percentile.
EP
AC C
236
TE D
228
Table 4 shows results of robustness checks. To begin with, I included multiple birth
237
sample because it has different characteristics compared to the baseline sample (Appendix Table
238
A2). Panels A, B, and C of column (1) show ATT, ATE, and quantile regression coefficient
239
values. The estimated effects of breastfeeding did not change much from the baseline results.
10
ACCEPTED MANUSCRIPT
Next, the propensity score approach could not infer the causal effects of breastfeeding because of
241
unmeasured confounding (Nickel, 2015). To examine any bias due to this, I included variables
242
about children and maternal health that were not accounted for from the baseline propensity
243
score estimation. First, I used birth weights and gestational age as measures for children’s health
244
because children’s birth weights and birth outcomes are important determinants of children’s
245
cognitive development (Almond & Currie, 2011b) and may affect maternal breastfeeding
246
decisions (Cunha & Heckman, 2007). Children with better birth outcomes as measured by birth
247
weight and gestational age are more likely to be breastfed (Appendix Table A3). Second,
248
maternal health status is also considered an important factor for breastfeeding (Jiang, Foster, &
249
Gibson-Davis, 2011). I used self-evaluated health status and a dummy variable of having any
250
chronic disease from the first wave of the MCS as measures of maternal health. Since there was
251
no direct information on maternal health status during breastfeeding, I used those variables as
252
proxies. Panels A and B of columns (2) and (3) show that the magnitudes of the ATTs and ATEs
253
are slightly smaller than my baseline regression results: the effects are around 0.9 and 0.8 points,
254
respectively, which again equals approximately 10% of the standard deviation of the cognitive
255
score distribution. Panel C of both columns shows that the impacts of maternal breastfeeding on
256
children’s development are still greater among children with lower cognitive test scores: the
257
impact of maternal breastfeeding is around 1.2 (1.0) points for children below the 20th percentile
258
of the cognitive test score distribution and around 0.3 (0.3) points for children above the 80th
259
percentile of the cognitive test score distribution.
SC
M AN U
TE D
EP
AC C
260
RI PT
240
261 262
4. Discussion
11
ACCEPTED MANUSCRIPT
263
Previous literature focused on the average effects of breastfeeding on children’s cognition. This study first examined the distributional impacts of maternal breastfeeding on children’s
265
development using a semiparametric quantile regression model. I found that the effects of
266
breastfeeding on children’s test scores were greater among children with lower test scores. The
267
effects were around 12–13% of the standard deviation, which is considered small in magnitude
268
according to Cohen’s effect size rule (Cohen, 1992) and the previous literature (Ip et al., 2007;
269
Jiang, Foster, & Gibson-Davis, 2011). However, the effects could become larger at the
270
population level as the disparity in educational achievement becomes larger within a society
271
(Rose, 1985). Given evidence that breastfeeding is associated with a reduction in several health
272
risks among children and mothers (Ip et al., 2007; Eidelman et al., 2012), the benefits of
273
breastfeeding could be larger when considering such health benefits (Cesur et al., 2017).
SC
M AN U
274
RI PT
264
The method and data suffer from some limitations. First, the propensity score approach is based on the ignorability assumption that there is no other confounding factor once a large
276
number of covariates are controlled for (Rosenbaum & Rubin, 1983). This assumption is too
277
strong to infer causal effects of breastfeeding due to unmeasured or unmeasurable confounding
278
(Nickel, 2015). As an example of unmeasured confounding, the MCS does not provide
279
information on mothers’ intelligence, such as IQ or test scores, or home environment scores,
280
which are considered important confounding factors (Der, Batty, & Deary, 2006; Belfield &
281
Kelly, 2012; Cesur et al., 2017). In addition, it is almost impossible from most observational data
282
to obtain information on unmeasurable factors such as maternal genes. Second, the data do not
283
provide enough information to construct an exclusive breastfeeding variable. Since the treatment
284
group consisted of children who were breastfed at all, this may underestimate the actual benefits
285
of exclusive breastfeeding (Jiang, Foster, & Gibson-Davis, 2011). Finally, breastfeeding duration
AC C
EP
TE D
275
12
ACCEPTED MANUSCRIPT
was based on maternal recall. This implies that there may be errors in treatment status and could
287
cause additional bias in the estimates. Since it is unclear if this type of error is random or
288
systematic (Promislow, Gladen, & Sandler, 2005), it is hard to see the exact direction of the bias.
289
An economic theory may provide a possible explanation for the main findings of this study.
RI PT
286
Let’s assume that children’s human capital production function is concave against maternal
291
breastfeeding. Low SES children usually have a low chance of being breastfed and so achieve
292
low test scores based on the underlying biological mechanism. Because they are less likely to be
293
breastfed, marginal returns on breastfeeding could be larger than those in higher SES children
294
due to the concavity of the human capital production function. These results have a policy
295
implication for the promotion of breastfeeding among individuals with lower SES. Significant
296
SES-related disparities in children’s academic performance are well established in the UK
297
(Banerjee, 2016). Government programs such as Sure Start have been introduced to narrow these
298
gaps (Melhuish et al., 2008). Since a high proportion of children with lower test scores generally
299
come from families with low SES and since there are observed SES-related disparities in
300
maternal breastfeeding incidence and duration, government policies such as the UNICEF UK
301
Baby Friendly Initiative aims at promoting breastfeeding may reduce SES-related gaps in
302
children’s academic development. However, there are at least two reasons why this policy
303
implication should be interpreted with caution. First, the policy implication of this study is based
304
on results with strong statistical assumptions. It is hard to draw causal inferences regarding the
305
effects of breastfeeding. Second, the results may not be generalizable. Public agencies such as
306
the American Academy of Pediatrics and the WHO recommend that mothers exclusively
307
breastfeed their infants for at least 6 months and ideally continue to breastfeed beyond 6 months.
308
Under the Affordable Care Act in the United States, the initiation and extended duration of
AC C
EP
TE D
M AN U
SC
290
13
ACCEPTED MANUSCRIPT
309
maternal breastfeeding are promoted. However, this study is based on data from the UK. The
310
results should therefore be cautiously applied to other countries owing to significant differences
311
in health care and other social policy. As an avenue for future research, the distributional effects on other benefits of
313
breastfeeding, such as the health benefits for children and mothers, could also be considered.
314
Previous literature has mainly focused on the average effects of breastfeeding, but there is little
315
evidence of the distributional effects on children’s and maternal health. To draw a more
316
convincing causal inference on the effects of breastfeeding, it would be useful to exploit a natural
317
experiment that quasi-randomizes breastfeeding status or duration.
M AN U
SC
RI PT
312
AC C
EP
TE D
318
14
ACCEPTED MANUSCRIPT
324 325 326 327 328
Labor Economics, 4, 1315-1486.
RI PT
323
Almond, D. & Currie, J. (2011a). Human capital development before age five. Handbook of
Almond, D. & Currie, J. (2011b). Killing me softly: The fetal origins hypothesis. The Journal of Economic Perspectives, 25(3), 153-172.
Anderson, J.W., Johnstone, B.M., & Remley, D.T. (1999). Breast-feeding and cognitive
development: a meta-analysis. The American Journal of Clinical Nutrition, 70(4), 525-535.
SC
322
References
Banerjee, P.A. (2016). A systematic review of factors linked to poor academic performance of
M AN U
319 320 321
disadvantaged students in science and maths in schools. Cogent Education, 3(1), 1178441.
329
Becker, G. (1981). A treatise on the family. Cambridge, MA: Harvard University Press.
330
Belfield, C.R. & Kelly, I.R. (2012). The benefits of breastfeeding across the early years of
332 333 334
childhood. Journal of Human Capital, 6(3), 251-277.
Borra, C., Iacovou, M., & Sevilla, A. (2012). The effect of breastfeeding on children's cognitive
TE D
331
and noncognitive development. Labour Economics, 19(4), 496-515. Cesur, R., Sabia, J.J., Kelly, I.R., & Yang, M. (2017). The effect of breastfeeding on young adult wages: new evidence from the add health. Review of Economics of the Household, 15(1),
336
25-51.
AC C
337
EP
335
Clandinin, M.T., Chappell, J.E., Heim, T., Swyer, P.R., & Chance, G.W. (1981). Fatty acid
338
utilization in perinatal de novo synthesis of tissues. Early Human Development, 5(4), 355-
339
366.
340
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155.
15
ACCEPTED MANUSCRIPT
341
Colen, C.G. & Ramey, D.M. (2014). Is breast truly best? Estimating the effects of breastfeeding
342
on long-term child health and wellbeing in the United States using sibling
343
comparisons. Social Science & Medicine, 109, 55-65.
346 347
Review, 97(2), 31.
RI PT
345
Cunha, F. & Heckman, J.J. (2007). The technology of skill formation. The American Economic
Cunha, F., Heckman, J.J., & Schennach, S.M. (2010). Estimating the technology of cognitive and noncognitive skill formation. Econometrica, 78(3), 883-931.
SC
344
Dehejia, R.H. & Wahba, S. (1999). Causal effects in nonexperimental studies: Reevaluating the
349
evaluation of training programs. Journal of the American Statistical Association, 94(448),
350
1053-1062.
351
M AN U
348
Del Bono, E. & Rabe, B. (2012). Breastfeeding and child cognitive outcomes: Evidence from a hospital-based breastfeeding support policy (No. 2012-29). ISER Working Paper Series,
353
2012-29, http://hdl.handle.net/10419/91672.
354
TE D
352
Deoni, S.C.L., Dean III, D.C., Piryatinksy, I., O’Muircheartaigh, J., Waskiewicz, N., Lehman, K., Han, M., & Dirks, H. (2013). Breastfeeding and early white matter development: A
356
cross sectional study. NeuroImage, 82, 77-86.
357
EP
355
Der, G., Batty, D., & Deary, I.J. (2006). Effect of breast feeding on intelligence in children: Prospective study, sibling pairs analysis, and meta-analysis. British Medical Journal, 333,
359
945-948.
360
AC C
358
Eidelman, A.I., Schanler, R.J., Johnston, M., Landers, S., Noble, L., Szucs, K., & Viehmann, L.
361
(2012). Breastfeeding and the use of human milk. Pediatrics, 129(3), e827-e841.
362
Elliott, C.D., Smith, P., & McCulloch, K. (1997). British ability scales, second ed. (BAS II):
363
Technical manual. London, UK: NFER-Nelson.
16
ACCEPTED MANUSCRIPT
368 369 370 371 372 373 374 375 376
Lipids, 34(2), 161-169.
RI PT
367
Fernstrom, J.D. (1999). Effects of dietary polyunsaturated fatty acids on neuronal function.
Firpo, S. (2007). Efficient semiparametric estimation of quantile treatment effects. Econometrica, 75(1), 259-276.
Fitzsimons, E. & Vera-Hernandez, M. (2012). The causal effects of breastfeeding on children’s
SC
366
low-income countries. Washington, DC: The World Bank.
development. London, UK: Institute for Fiscal Studies.
Heckman, J.J. (2006). Skill formation and the economics of investing in disadvantaged
M AN U
365
Fernald, L.C., Kariger, P., Engle, P., & Raikes, A. (2009). Examining early child development in
children. Science, 312(5782), 1900-1902.
Hirano, K., Imbens, G.W., & Ridder, G. (2003). Efficient estimation of average treatment effects using the estimated propensity score. Econometrica, 71(4), 1161-1189. Ip, S., Chung, M., Raman, G., Chew, P., Magula, N., DeVine, D., et al., 2007. Breastfeeding and
TE D
364
maternal and infant health outcomes in developed countries. Evidence report/technology
378
assessment no. 153 (Prepared by Tufts-New England Medical Center Evidence-based
379
Practice Center, under Contract No. 290-02-0022). AHRQ Publication No. 07-E007.
380
Rockville, MD: Agency for Healthcare Research and Quality. Jiang, M., Foster, E.M., & Gibson-Davis, C.M. (2011). Breastfeeding and the child cognitive
AC C
381
EP
377
382
outcomes: A propensity score matching approach. Maternal and Child Health Journal,
383
15(8), 1296-1307.
384 385
Kling, J.R., Liebman, J.B., & Katz, L.F. (2007). Experimental analysis of neighborhood effects. Econometrica, 75(1), 83-119.
17
ACCEPTED MANUSCRIPT
386
Kramer, M.S., Chalmers, B., Hodnett, E.D., Sevkovskaya, Z., Dzikovich, I., Shapiro, S., Collet, J.P., Vanilovich, I., Mezen, I., Ducruet, T., & Shishko, G. (2001). Promotion of
388
breastfeeding intervention trial (PROBIT): A randomized trial in the Republic of
389
Belarus. Journal of American Medical Association, 285(4), 413-420.
RI PT
387
Kramer, M.S., Aboud, F., Mironova, E., Vanilovich, I., Platt, R.W., Matush, L., Igumnov, S., et
391
al. (2008). Breastfeeding and child cognitive development: New evidence from a large
392
randomized trial. Archives of General Psychiatry, 65(5), 578-584.
SC
390
Melhuish, E., Belsky, J., Leyland, A.H., Barnes, J., & National Evaluation of Sure Start Research
394
Team. (2008). Effects of fully-established Sure Start Local Programmes on 3-year-old
395
children and their families living in England: a quasi-experimental observational study. The
396
Lancet, 372(9650), 1641-1647.
400 401 402 403 404 405 406 407 408
TE D
399
research? Journal of Human Lactation, 31(2), 209-212.
Promislow, J.H., Gladen, B.C., & Sandler, D.P. (2005). Maternal recall of breastfeeding duration by elderly women. American Journal of Epidemiology, 161(3), 289-296. Rees, D.I. & Sabia, J.J. (2009). The effect of breast feeding on educational attainment: Evidence
EP
398
Nickel, N.C. (2015). Look before you leap: can we draw causal conclusions from breastfeeding
from sibling data. Journal of Human Capital, 3(1), 43-72. Rose, G. (1985). Sick individuals and sick populations. International Journal of
AC C
397
M AN U
393
Epidemiology, 14(1), 32-38. Rosenbaum, P.R. & Rubin, D.B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41-55. Rothstein, D.S. (2012). Breastfeeding and children's early cognitive outcomes. Review of Economics and Statistics, 95(3), 919-931.
18
ACCEPTED MANUSCRIPT
409 410
Zhao, Z. (2008). Sensitivity of propensity score methods to the specifications. Economics Letters, 98(3), 309-319.
AC C
EP
TE D
M AN U
SC
RI PT
411
19
ACCEPTED MANUSCRIPT
Tables and Figures Table 1. Baseline Characteristics of Control Variables No propensity score adjustment Breastfeeding Difference Y N Y-N (1) (2) (3) 30.97 28.31 2.66*** .39 .67 -.28*** .38 .58 -.20*** .56 .53 .03*** .82 .88 -.06*** .97 .95 .02*** .69 .60 .09*** .98 .97 .01*** .43 .34 .09*** .10 .26 -.16*** .35 .39 -.04*** 4.82 5.12 -.31*** .21 .37 -.16*** .05 .14 -.09*** 33.63 31.06 2.57*** .41 .66 -.25***
Propensity score adjustment Breastfeeding Difference Y N Y-N (4) (5) (6) 30.97 30.97 .00 .39 .39 .00 .38 .38 .00 .56 .56 .00 .82 .80 .02 .97 .97 .00 .69 .69 .00 .98 .97 .01 .43 .43 .00 .1 .09 .01 .35 .36 -.01 4.82 4.82 .00 .21 .21 .00 .05 .05 .00 33.63 33.64 -.01 .41 .41 .00
RI PT
412 413 414 415
TE D
EP AC C
416 417 418 419 420 421 422
M AN U
SC
Variables Maternal age at birth Low education Low household income Working (after birth) White Married at pregnancy Planned pregnancy Prenatal care (ever) Prenatal class (ever) Smoking during pregnancy Any complications during pregnancy Alcohol consumption Any government support Number of smokers in household Spouse’s age at delivery Low education (spouse) Data source: The MCS. Note: I report differences in selected characteristics by maternal breastfeeding status. The first three columns show average values and their differences without any adjustment. The last three columns show average values and their differences after the propensity score adjustment. I used ̂ /(1 − ̂ ) as a weight for the control group. ***p < .01, **p < .05, * p < .10.
20
ACCEPTED MANUSCRIPT
Table 2. Distributional Effects of Maternal Breastfeeding on Children’s Cognitive Test Scores Breastfeeding ≥ 4 months (1)
A. ATT 1.04*** (0.22) B. ATE
SC
0.94*** (0.27)
0.3 0.4 0.5
1.08* (0.62) 1.23*** (0.46) 1.17*** (0.38) 1.23*** (0.33) 1.17*** (0.29) 1.00*** (0.28) 0.83*** (0.26) 0.50* (0.27) 0.50* (0.30)
M AN U
C. Quantiles 0.1 0.2
RI PT
Treatment status
TE D
0.6 0.7 0.8
EP
0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
Data source: The MCS. Note: For treatment variable, I use the dummy variable if a child was breastfed for at least 4 months. The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
21
ACCEPTED MANUSCRIPT
Table 3. Distributional Effects of Different Durations of Maternal Breastfeeding on Children’s Cognitive Test Scores
(1) A. ATT 1.26*** (0.31) B. ATE 1.16*** (0.27)
0.2 0.3 0.4
0.7 0.8
0.83** (0.33)
TE D
0.5 0.6
1.33** (0.55) 1.5*** (0.50) 1.17*** (0.38) 1.17*** (0.37) 1.13*** (0.33) 1.33*** (0.31) 1.00*** (0.36) 1.00*** (0.38) 0.93** (0.44)
0.96*** (0.24)
M AN U
C. Quantiles 0.1
Breastfeeding ≥ 6 months (2)
RI PT
Ever Breastfeeding
SC
Treatment status
EP
0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
1.42* (0.82) 1.58*** (0.56) 1.50*** (0.43) 1.00*** (0.39) 1.13*** (0.36) 1.10*** (0.33) 0.70** (0.29) 0.27 (0.29) 0.33 (0.35)
Data source: The MCS. Note: In each column, I estimate the effects of breastfeeding using different cutoffs. For treatment variable, I use the dummy variable if a child was ever breastfed in columns (1), and at least 6 months in column (2). The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
22
ACCEPTED MANUSCRIPT
Table 4. Robustness Checks
Including birth outcomes
(1)
(2)
A. ATT 1.06*** (0.22)
0.90*** (0.23)
B. ATE
0.4 0.5 0.6 0.7 0.8 0.9
61.52 (10.05)
AC C
Mean (SD) of outcome
M AN U
0.3
1.00 (0.65) 1.17*** (0.45) 1.17*** (0.39) 1.17*** (0.33) 1.00*** (0.30) 1.00*** (0.28) 0.73*** (0.26) 0.33 (0.27) 0.50* (0.30)
TE D
0.2
1.08* (0.62) 1.23*** (0.46) 1.17*** (0.38) 1.23*** (0.33) 1.17*** (0.29) 1.00*** (0.28) 0.83*** (0.26) 0.50* (0.27) 0.50* (0.30)
EP
C. Quantiles 0.1
0.81*** (0.26)
SC
0.98*** (0.26)
RI PT
Including multiple births
Including both birth outcomes and maternal health (3)
61.52 (10.05)
0.91*** (0.23)
0.79*** (0.27)
0.83 (0.65) 1.00** (0.46) 1.00*** (0.39) 1.16*** (0.34) 0.83*** (0.30) 0.92*** (0.28) 0.73*** (0.26) 0.33 (0.27) 0.50* (0.30) 61.52 (10.05)
Data source: The MCS. Note: In each column, I estimate the effects of breastfeeding using different specifications. I include multiple births in column (1), and children’s birth weights and gestational age in column (2), and maternal health status as well as birth weights and gestational age in column (3). For treatment variable, I use the dummy variable if a child was breastfed for at least 4 months. The cognitive test score measure is the average score of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture similarity, and word reading. I used different sampling weights for ATTs and ATE: ̂ /(1 − ̂ ) as a sampling weight for the control group in panel A and C, and 1/ ̂ and 1/ 1 − ̂ as sampling weights for treatment and control groups in Panel B. Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
23
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 1. Distribution of Children’s Cognitive Development
AC C
EP
TE D
Data source: The MCS.
24
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 2. Distribution of Estimated Propensity Score by Breastfeeding Status
AC C
EP
TE D
Data source: The MCS.
25
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 3. Estimated Propensity Score and Actual Probability of Breastfeeding (≥ 4 Months)
AC C
EP
TE D
Data source: The MCS. Note: This figure shows a plot of the average estimated propensity scores against the actual probability of breastfeeding in 100 equal-sized cells. The red line is the 45-degree line.
26
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 4. Children’s Cognitive Development as a Function of Estimated Propensity Score
AC C
EP
TE D
Data source: The MCS. Note: This figure shows a plot of average cognitive test scores against estimated propensity scores in 20 equal-sized cells by maternal breastfeeding status.
27
ACCEPTED MANUSCRIPT
Appendix Table A1. Effects of Maternal Breastfeeding on Children’s Cognitive Test Scores
Breastfed ≥ 4 months
(2)
(3)
RI PT
(1) 1.022***
1.032**
1.090***
(0.207)
(0.496)
(0.214)
Y
Y Y
Y
SC
Parametric control of ̂ Breastfed4mon× Control variables ( Data source: The MCS.
Y
AC C
EP
TE D
M AN U
Note: Breastfed≥4months indicates whether a child was breastfed for at least 4 months. The cognitive test score measure is the average of six British Ability Scales (BAS) for naming vocabulary, pattern construction, picture 8 similarity, and word reading. In column (1), I use ̂ and ̂ as controls instead of using sampling weights. In column (2), I add Breastfed4mon× ̂ to the regression specification of column (1). In column (3), I add 8 ̂ and ̂ and the control variables that were used to estimate ̂ to the regression specification of column (1). Robust standard errors are calculated and presented in the parentheses. ***p < .01, **p < .05, and * p < .10.
28
ACCEPTED MANUSCRIPT
Table A2. Characteristics of Baseline and Excluded Sample
AC C
EP
TE D
M AN U
SC
RI PT
Baseline Excluded Excluded - Baseline Variables (1) (2) (3) Maternal age at birth 28.29 30.81 2.52*** Low education .61 .56 -.05* Low household income .61 .56 -.05 Working (after birth) .47 .44 -.03 White .84 .89 .05** Married at pregnancy .81 .88 .07*** Planned pregnancy .54 .63 .09*** Prenatal care (ever) .96 .97 .01 Prenatal class (ever) .34 .36 .02 Smoking during pregnancy .25 .18 -.07*** Any complications during pregnancy .38 .48 .1*** Alcohol consumption 5.14 4.87 -.27 Any government support .43 .43 .00 Number of smokers in household .14 .09 -.05** Spouse’s age at delivery 31.88 33.60 1.72*** Low education (spouse) .40 .37 -.03 Data source: The MCS. Note: I used the same variables that I used in Table 1. ***p < .01, **p < .05, and * p < .10.
29
ACCEPTED MANUSCRIPT
Table A3. Maternal Breastfeeding and Children’s Birth Outcomes Pr (Birth weight ≤ 2,500 g)
Pr (Birth weight ≤ 1,500 g)
Gestational age (days)
(1)
(2)
(3)
(4)
-0.030***
-0.005**
(0.006)
(0.002)
3199.442
0.094
0.013
12788 0.005
12796 0.003
12796 0.001
(12.932) Average of control group Observations R-squared
M AN U
Data source: The MCS.
1.687***
-0.019***
(0.286)
(0.005)
275.617
SC
77.607***
Breastfed ≥ 4 months
Pr (Gestational age ≤ 37 weeks) (5)
RI PT
Birth weight (grams)
12703 0.004
0.069
12703 0.002
AC C
EP
TE D
Note: Breastfed ≥ 4 months indicates whether a child was breastfed for at least 4 months. I used ̂ /(1 − ̂ ) as a sampling weight for the control group. Robust standard errors are calculated and presented in parentheses. ***p < .01, **p < .05, and * p < .10.
30
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Highlights • Study of the effects of breastfeeding on the distribution of children’s test scores. • A semiparametric quantile regression is used for the estimation. • The effects are greater for children with lower test scores.