Transcript Son preference and Early Childhood Investments in China

Douglas Almond
Columbia University & NBER
Hongbin Li
Tsinghua University
Lingsheng Meng
University of Maryland

Do parental investment decisions change when
they are able to know child gender during
pregnancy in China?
◦ Recent research finds big long-term effects of the earlychildhood environment and parental investments
◦ There is son preference in China
 E.G., high sex ratios (excess of male births)
◦ During 1980s, sex no longer revealed at birth but
increasingly revealed during pregnancy
 Rapid diffusion of ultrasound (poorly observed in data until
now)
 Might change prenatal investments: reduce investments
when fetus is a girl?
 Might affect decision to continue pregnancy
1.
Meng (2009)
o
o
o
2.
Collects and analyzes data on county-by-county
diffusion of ultrasound machines across China during
the 1980s
Finds that the local access to ultrasound is strongly
predictive of increased sex ratios at birth (more males).
My job market paper!
Lhila and Simon (2008)
o
o
o
Considers Asian immigrants to the USA
Ultrasound use reported on birth certificate data
(natality data), proxies for knowing sex
Finds no effect on prenatal health investments in US

Consider Postnatal investments
◦ Before ultrasound, knew sex prior to making postnatal
investments (obviously)
◦ But if there is heterogeneity in son preference across
families, then increases in the sex ratio at birth with
ultrasound availability would suggest that following
ultrasound availability, girls are born to parents with a
weaker son preference (relative to parents of girls
prior to ultrasound).
◦ We hypothesize that postnatal investments in girls
increased following ultrasound availability.

Consider Prenatal investments
 Reduction in prenatal investments in girls following
ultrasound availability
 Before ultrasound was available, sex was presumably
unknown until delivery, which would tend to equalize
prenatal investments in girls versus boys.
 Increased preference sorting with ultrasound access
would tend to increase prenatal investments in girls.
◦ The prediction for prenatal investments in girls
following ultrasound availability is ambiguous
yict  1Girlict  2Ultrasoundct  3Girlict Ultrasoundct  c  t  c  t   ict
Girl
= 1 if child is female
Ultrasound
= 1 if ultrasound is available in the county when mother is
pregnant
μc
county fixed effect
νt
year fixed effect
μc×t
county-specific linear time trend
Outcomes
prenatal investment: neonatal mortality (outcome)
postnatal investment: vaccination, breastfeeding, who is
taking care of the child
1. Neonatal mortality
 Usually caused by congenital anomalies, prematurity
and complications of delivery (Grossmand and
Jacobowitz, 1981)
 May capture impacts on child survival through
prenatal investment on which we do not have data
2. Post-neonatal mortality
 Usually caused by post-neonatal infections and
accidents (Grossmand and Jacobowitz , 1981)
 May partly capture impacts on child survival through
postnatal investment
3. Direct postnatal investment measures


From thousands of volumes of Local
Gazetteers of China
Local Gazetteers:
◦ “encyclopedia” of a particular region
◦ compiled by local governments
◦ introduction time of ultrasound machines often
recorded as achievement in the public health
sector

Reports the year of the introduction of
ultrasound machines for 1,572 counties

1980

1985

1990

1995
100
80
60
40
0
20
1975
1980
1985
Year
1990
1995

Chinese Children Survey
◦ Conducted by the National Bureau of Statistics of
China in June 1992
◦ Representative national data: 560,000 households
surveyed
◦ Pregnancy history records for women
 pregnancy order, date of conception, use of prenatal
care, gestation
 pregnancy outcome (miscarriage, abortion, birth …)
 live births: gender, DOB
 Infant mortality
◦ Retrospective reports of parental investments
 breastfeeding, and childhood vaccinations.
Table 1: Summary statistics (samples with ultrasound information)
Variables
Observations
Mean
Standard Deviation
289860
289860
289860
289860
289860
289860
0.468
0.366
0.003
0.006
0.007
0.003
0.499
0.482
0.055
0.075
0.083
0.054
95693
95693
95322
95321
95321
95321
95693
92930
95692
0.458
0.367
0.160
0.157
0.193
0.254
0.971
14.855
0.846
0.498
0.482
0.367
0.364
0.395
0.435
0.167
7.854
0.361
Panel A
Subsample with mortality measures
Girl
Ultrasound
Death within 24 hours of birth
Death within 7 days of birth
Death within 28 days of birth
Death within 28 days to 1 year
Panel B
Subsample with postnatal measures
Girl
Ultrasound
Vaccine 1 (BCG)
Vaccine 2 (IPV)
Vaccine 3 (DTPa)
Vaccine 4 (Measles)
Breastfeeding
Duration of breastfeeding (months)
Taken care of by mother
Notes:
The samples are taken from the Chinese Children Survey.
Information on ultrasound access is collected by the authors.
Table 2 : The effect of ultrasound availability on relative mortality of girls: linear probability model results
Death within
24 hours of birth
Death within
7 days of birth
Death within
28 days of birth
Death within
28 days to 1 year†
(1)
(2)
(3)
(4)
Girl
0.0001
(0.0003)
-0.0002
(0.0003)
-0.0002
(0.0004)
-0.0001
(0.0003)
Ultrasound
0.0007**
(0.0003)
0.0013***
(0.0004)
0.0019***
(0.0005)
0.0008**
(0.0003)
Ultrasound × Girl
0.0010**
(0.0004)
0.0015**
(0.0006)
0.0014**
(0.0006)
0.0005
(0.0004)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
289859
0.0032
289859
0.0037
289859
0.0042
287830
0.0038
Dependent variable:
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%;
† among neonatal survivors
Table 2 : The effect of ultrasound availability on relative mortality of girls: linear probability model results
Death within
24 hours of birth
Death within
7 days of birth
Death within
28 days of birth
Death within
28 days to 1 year†
(1)
(2)
(3)
(4)
Girl
0.0001
(0.0003)
-0.0002
(0.0003)
-0.0002
(0.0004)
-0.0001
(0.0003)
Ultrasound
0.0007**
(0.0003)
0.0013***
(0.0004)
0.0019***
(0.0005)
0.0008**
(0.0003)
Ultrasound × Girl
0.0010**
(0.0004)
0.0015**
(0.0006)
0.0014**
(0.0006)
0.0005
(0.0004)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
289859
0.0032
289859
0.0037
289859
0.0042
287830
0.0038
Dependent variable:
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%;
† among neonatal survivors
Table 3 : The effect of ultrasound availability on vaccination: linear probability model results
Dependent variable:
Vaccine 1
(1)
Vaccine 2
(2)
Vaccine 3
(3)
Vaccine 4
(4)
Girl
0.004
(0.003)
0.007**
(0.003)
0.005*
(0.003)
0.012***
(0.003)
Ultrasound
0.001
(0.005)
0.001
(0.005)
-0.001
(0.005)
0.002
(0.005)
Ultrasound × Girl
0.001
(0.004)
-0.003
(0.005)
0.001
(0.005)
-0.008
(0.005)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
95321
0.1816
95320
0.1291
95320
0.1531
95320
0.1590
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 3 : The effect of ultrasound availability on vaccination: linear probability model results
Dependent variable:
Vaccine 1
(1)
Vaccine 2
(2)
Vaccine 3
(3)
Vaccine 4
(4)
Girl
0.004
(0.003)
0.007**
(0.003)
0.005*
(0.003)
0.012***
(0.003)
Ultrasound
0.001
(0.005)
0.001
(0.005)
-0.001
(0.005)
0.002
(0.005)
Ultrasound × Girl
0.001
(0.004)
-0.003
(0.005)
0.001
(0.005)
-0.008
(0.005)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
95321
0.1816
95320
0.1291
95320
0.1531
95320
0.1590
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%
Table 4 : The effect of ultrasound availability on care
Statistical model
Linear probability
Poisson
Linear probability
Breastfeeding
(1)
Duration of breastfeeding
(2)
Taken care of by mother
(3)
Girl
-0.002*
(0.001)
-0.048***
(0.004)
-0.004
(0.002)
Ultrasound
0.002
(0.002)
0.014***
(0.005)
0.001
(0.004)
Ultrasound × Girl
-0.002
(0.002)
0.004
(0.006)
0.002
(0.004)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
95693
0.0379
92930
N.A.
95692
0.1424
Dependent variable:
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%
For results from Poisson regression, marginal effects at the mean are displayed.
Table 4 : The effect of ultrasound availability on care
Statistical model
Linear probability
Poisson
Linear probability
Breastfeeding
(1)
Duration of breastfeeding
(2)
Taken care of by mother
(3)
Girl
-0.002*
(0.001)
-0.048***
(0.004)
-0.004
(0.002)
Ultrasound
0.002
(0.002)
0.014***
(0.005)
0.001
(0.004)
Ultrasound × Girl
-0.002
(0.002)
0.004
(0.006)
0.002
(0.004)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
95693
0.0379
92930
N.A.
95692
0.1424
Dependent variable:
County fixed effects
Year fixed effects
County time trends
Observations
R-squared
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%
For results from Poisson regression, marginal effects at the mean are displayed.


Include mother FE into the model
Use the sub-sample of families with 2 or
more kids and where the births “straddle”
the introduction of ultrasound

Advantage: control for unobservable factors
that are common to siblings

Caveat: “straddling” may also be endogenous!
Table A1 : Mother fixed effects estimates of the effect of ultrasound availability on relative mortality of
girls: linear probability model results
Death within
24 hours of birth
(1)
Death within
7 days of birth
(2)
Death within
28 days of birth
(3)
Death within
28 days to 1 year†
(4)
0.0001
(0.0005)
-0.0008
(0.0006)
-0.0005
(0.0007)
0.0000
(0.0005)
Ultrasound
-0.0027***
(0.0009)
-0.0070***
(0.0013)
-0.0087***
(0.0014)
-0.0033***
(0.0008)
Ultrasound × Girl
0.0026***
(0.0010)
0.0043***
(0.0014)
0.0039**
(0.0016)
-0.0001
(0.0010)
Mother fixed effects
County fixed effects
Year fixed effects
County time trends
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Observations
Groups
R-squared (between)
130376
43734
0.0005
130376
43734
0.0000
130376
43734
0.0000
129217
43732
0.0005
Dependent variable:
Girl
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%;
† among neonatal survivors
Table A2 : Mother fixed effects estimates of the effect of ultrasound availability on vaccination: linear
probability model results
Dependent variable:
Vaccine 1
(1)
Vaccine 2
(2)
Vaccine 3
(3)
Vaccine 4
(4)
Girl
0.009
(0.005)
0.006
(0.006)
-0.001
(0.006)
0.002
(0.007)
Ultrasound
-0.018
(0.011)
-0.014
(0.011)
-0.017
(0.012)
-0.008
(0.015)
Ultrasound × Girl
0.024**
(0.012)
0.010
(0.013)
0.015
(0.013)
0.007
(0.015)
Mother fixed effects
County fixed effects
Year fixed effects
County time trends
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Observations
Groups
R-squared (between)
43609
29899
0.0005
43610
29899
0.0002
43610
29899
0.0003
43610
29899
0.0009
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%
Table A3 : Mother fixed effects estimates of the effect of ultrasound availability on care: linear model
results
Dependent variable:
Breastfeeding
(1)
Duration of breastfeeding
(2)
Taken care of by mother
(3)
Girl
-0.003
(0.003)
-0.706***
(0.133)
-0.008
(0.005)
Ultrasound
0.005
(0.006)
0.205
(0.302)
0.001
(0.013)
Ultrasound × Girl
-0.013*
(0.008)
0.166
(0.326)
-0.009
(0.013)
Mother fixed effects
County fixed effects
Year fixed effects
County time trends
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Observations
Groups
R-squared (between)
45264
32939
0.0035
43946
32207
0.0129
45263
32939
0.0019
Notes: Standard errors clustered at the county level are reported in parentheses;
* significant at 10%; ** significant at 5%; *** significant at 1%


Postnatal investments do not seem to
change as a result of preference-sorting
induced by the availability of ultrasound
Female neonatal mortality increases
following ultrasound availability
◦ Effect concentrated soon after birth
 Suggests that parents withhold investment in female
fetuses after prenatal sex determination became
available
◦ “Inframarginal” effect on prenatal investments
outstrips the potential increase in prenatal
investments from preference sorting