presented by halsey rogers and kathleen beegle world bank june 4, 2009
DESCRIPTION
Fertility, Reproductive Health and Economic Development : Preliminary results of the World Bank research program supported by the William and Flora Hewlett Foundation. Presented by Halsey Rogers and Kathleen Beegle World Bank June 4, 2009. Thematic areas. - PowerPoint PPT PresentationTRANSCRIPT
1
Fertility, Reproductive Health and Economic
Development:Preliminary results of the
World Bank research program supported by the William and Flora
Hewlett FoundationPresented by Halsey Rogers and Kathleen Beegle World Bank June 4, 2009
2
Thematic areas
Fertility and investments in child quality 4 studies
Fertility, poverty, and family welfare in the time of HIV/AIDS 2 studies
Fertility and female labor supply 2 studies
3
Thematic Area 1: Fertility and investments in child quality 4 research projects in this category
Family size Family size and early childhood development:
evidence from Ecuador Declining fertility and rising household
investment in education in Vietnam Gender preference
Development, modernization, and childbearing: The role of family gender composition
Financial incentives for female births and parental investments in daughters in North India
4
Motivating question: How do fertility choices affect investments in children? A large literature documents associations between family size
and children’s outcomes: In developed countries, many studies have documented a
negative association between family size and educational attainment
Research in developing countries has documented negative associations between children’s health nutritional outcomes and family size
Negative associations between family size and child outcomes could be due to a number of factors: Resource dilution (both financial and parental) produces
“quality-quantity” tradeoffs. Son preference adds another dimension to resource dilution, as parents prefer to invest in sons than in daughters
Changes in family dynamics—larger families may have lower “average maturity” of household members
Omitted variables or selection: characteristics of families that result in larger family size also result in poorer child outcomes
This set of studies documents determinants of family size and tests quantity-quality hypothesis in ways that control for selection/omitted variables
5
Family size and early child development: evidence from Ecuador
Christina Paxson (Princeton University) andNorbert Schady (World Bank)
6
Questions and data Research questions:
Part 1: What is the association between (1)cognitive and nutrition outcomes in early childhood and (2) family size?
Part 2: Do children in families that grow between baseline and follow-up experience (relative) declines in their cognitive and nutritional outcomes? Use detailed information on maternal characteristics,
including cognitive ability, mental health and parenting behaviors
Information on multiple children in the households permits within-family estimates, and longitudinal information makes it possible to examine how the presence of “new” children influences the outcomes of their older siblings
Sample 4200 low-income families with about 6700 children aged
0-6 at baseline From rural and urban areas of 6 provinces in Ecuador Longitudinal data on families, with two interviews
spaced approximately 18 months apart, with 1,124 births between baseline and the 1st follow-up
7
Test of family-size effect
If family size merely reflects family-specific unobservables, we expect that children of families that are going to grow will fare worse
If family size has a negative effect on children, we expect that children in families that grow will experience declines in outcomes relative to children in families that remain the same size
8
Result 1: Cognitive and nutritional outcomes are strongly associated with family size (1)
7075
8085
90TV
IP s
core
, M=1
00 S
D=1
5
0 1 2 3 4 or more
05
1015
20lo
ng-te
rm m
emor
y, p
'tile
0 1 2 3 4 or more
010
2030
40sh
ort-t
erm
mem
ory,
p'ti
le
0 1 2 3 4 or more
05
10vi
sual
clo
sure
, p'ti
le
0 1 2 3 4 or more
cognitive outcomes by #siblings
9
Result 1: Cognitive and nutritional outcomes are strongly associated with family size (2)
10.6
10.8
1111
.211
.4he
mog
lobi
n, g
/dl
0 1 2 3 4 or more
4050
6070
80an
emic
, per
cent
0 1 2 3 4 or more
-1.5
-1-.5
0he
ight
-for-a
ge z
-sco
re
0 1 2 3 4 or more
010
2030
40st
unte
d, p
erce
nt
0 1 2 3 4 or more
nutrition outcomes by #siblings
10
Result 2
However, there are large differences in the characteristics of large and small families Select one- and two- child families at baseline Examine outcomes that were measured at both
waves: TVIP score, hemoglobin and height Use panel data to:
Examine whether the children in families that grow between baseline and follow-up have worse outcomes at baseline
Examine whether children in families that grow between baseline and follow-up experience declines in their cognitive and nutritional outcomes
11
Table 5a – Child outcomes at baseline and changes in outcomes between baseline and follow-upOne-child families
TVIP Hemoglobin Height ∆TVIP ∆Hemoglobin ∆Height
One-child families at baseline; no family controls
Indicator: New child between baseline and follow-up
–0.103 (0.065)
–0.118 (0.048)
–0.198 (0.074)
0.001 (0.066)
0.089 (0.072)
0.069 (0.065)
One-child families at baseline; family controls
Indicator: New child between baseline and follow-up
0.082 (0.065)
0.005 (0.050)
–0.064 (0.076)
–0.012 (0.071)
0.067 (0.076)
0.068 (0.069)
Observations 1029 1829 1924 965 1484 1835
12
Table 5a – Child outcomes at baseline and changes in outcomes between baseline and follow-upOne-child families
TVIP Hemoglobin Height ∆TVIP ∆Hemoglobin ∆Height
One-child families at baseline; no family controls
Indicator: New child between baseline and follow-up
–0.103 (0.065)
–0.118 (0.048)
–0.198 (0.074)
0.001 (0.066)
0.089 (0.072)
0.069 (0.065)
One-child families at baseline; family controls
Indicator: New child between baseline and follow-up
0.082 (0.065)
0.005 (0.050)
–0.064 (0.076)
–0.012 (0.071)
0.067 (0.076)
0.068 (0.069)
Observations 1029 1829 1924 965 1484 1835
13
Table 5b – Child outcomes at baseline and changes in outcomes between baseline and follow-upTwo-child families
TVIP Hemoglobin Height ∆TVIP ∆Hemoglobin ∆Height Two child families at baseline; no family controls
Indicator: Second child at baseline
0.093 (0.066)
–0.081 (0.052)
–0.180 (0.086)
0.029 (0.071)
0.146 (0.075)
0.080 (0.075)
Indicator: New child between baseline and follow-up
–0.194 (0.050)
–0.102 (0.042)
–0.179 (0.069)
–0.105 (0.057)
0.030 (0.065)
0.012 (0.063)
Two child families at baseline; family controls
Indicator: Second child at baseline
0.058 (0.062)
–0.114 (0.052)
–0.239 (0.086)
0.012 (0.073)
0.112 (0.078)
0.040 (0.078)
Indicator: New child between baseline and follow-up
–0.029 (0.049)
0.005 (0.043)
–0.140 (0.070)
–0.093 (0.059)
0.018 (0.068)
0.012 (0.066)
Observations 1277 2540 2864 1182 2062 2588
14
Summary and conclusions
Large negative associations between family size and children’s cognitive and nutritional outcomes These results indicate that associations documented
later in life, for education and earnings, are evident in early life
However, little evidence of deterioration in children’s outcomes or in parenting quality with the addition of a new child Children in families that are going to become larger
have poorer outcomes (Table 5) This evidence is consistent with selection stories--
common factors drive family size and child outcomes Current analysis (still preliminary) analyzes outcomes
using a third round of data Do negative causal effects of family size manifest
themselves after a longer time period? Are they more likely at larger family sizes than the ones
we observe in waves 1 and 2 of our data?
15
The decision to invest in child quality over quantity: Declining fertility and rising household investment in education in Vietnam
Hai-Anh Dang and Halsey Rogers (World Bank)
16
Methodology and data
Approach: Use data from Vietnam to investigate the hypothesized child quantity-quality tradeoff: Question: Are lower fertility levels making it possible
for households to invest more in their children’s human capital?
IV approach, using instruments from different sources, incl. our own survey
One innovation: Good data on private tutoring expenditures, so we’re not just relying on (e.g.) enrollment or attainment as indicator of parental investment in education
Why Vietnam? Very rapid fertility decline and educational advances
Data sources 2006 household survey (VHLSS) DHS 2002 New survey focused on private tutoring (2008)
17
Fig. 1: Share of children attending tutoring classes
0
10
20
30
40
50
60
1 2 3 4 5
No of children in the household
Perc
enta
ge (%
)
age 0-11age 0-15age 0-18
Correlation between quantity and quality (Tutoring)
18
Table 3: Impacts of family size on school enrolment, Vietnam 2007-2008Probit IV Probit IV Probit IV Probit IV Probit IV Probit
No of children age 0-18 -0.167*** -0.544*** -0.999*** -0.867*** -0.945*** -0.923***(-9.48) (-2.61) (-15.55) (-2.79) (-4.86) (-5.50)
Age -0.213*** -0.212*** -0.090** -0.266*** -0.235*** -0.244***(-25.40) (-11.45) (-2.55) (-2.99) (-2.63) (-3.65)
Male -0.133*** -0.209*** -0.161** -0.297*** -0.300*** -0.301***(-3.90) (-4.94) (-2.35) (-3.21) (-3.17) (-3.37)
Head's years of schooling 0.085*** 0.056*** -0.024 0.053 0.033 0.037(13.94) (2.60) (-0.81) (0.82) (0.59) (0.87)
Ethnic major group 0.049 -0.173 -0.316 -0.351 -0.377 -0.360(0.79) (-1.15) (-1.23) (-1.42) (-1.63) (-1.54)
Log of total hh exp. 0.482*** 0.543*** N/A 0.634*** 0.560** 0.585***(11.40) (11.42) (2.75) (2.47) (3.17)
InstrumentsDistance to fam. center YNo of visits per month by mobile fam. team Y
Government reg. Y YParental siblings Y Y
Overid test (J statistic) 0.04Log likelihood -3467 -16069 -2163 -2195 -2170 -2164N 10797 9052 1259 1371 1350 1350Note 1. Regressions control for regional and urban dummy variables.2. Cluster-robust t statistics in parentheses.3. Overidentification tests are from linear regression.
19
Table 3: Impacts of family size on school enrolment, Vietnam 2007-2008Probit IV Probit IV Probit IV Probit IV Probit IV Probit
No of children age 0-18 -0.167*** -0.544*** -0.999*** -0.867*** -0.945*** -0.923***(-9.48) (-2.61) (-15.55) (-2.79) (-4.86) (-5.50)
Age -0.213*** -0.212*** -0.090** -0.266*** -0.235*** -0.244***(-25.40) (-11.45) (-2.55) (-2.99) (-2.63) (-3.65)
Male -0.133*** -0.209*** -0.161** -0.297*** -0.300*** -0.301***(-3.90) (-4.94) (-2.35) (-3.21) (-3.17) (-3.37)
Head's years of schooling 0.085*** 0.056*** -0.024 0.053 0.033 0.037(13.94) (2.60) (-0.81) (0.82) (0.59) (0.87)
Ethnic major group 0.049 -0.173 -0.316 -0.351 -0.377 -0.360(0.79) (-1.15) (-1.23) (-1.42) (-1.63) (-1.54)
Log of total hh exp. 0.482*** 0.543*** N/A 0.634*** 0.560** 0.585***(11.40) (11.42) (2.75) (2.47) (3.17)
InstrumentsDistance to fam. center YNo of visits per month by mobile fam. team Y
Government reg. Y YParental siblings Y Y
Overid test (J statistic) 0.04Log likelihood -3467 -16069 -2163 -2195 -2170 -2164N 10797 9052 1259 1371 1350 1350Note 1. Regressions control for regional and urban dummy variables.2. Cluster-robust t statistics in parentheses.3. Overidentification tests are from linear regression.
20
Table 4: Impacts of family size on children attendance in private tutoring, Vietnam 2007-2008
Probit IV Probit IV Probit IV Probit IV ProbitNo of children age 0-18 -0.131*** -0.932*** -0.286 -0.525 -0.389
(-6.62) (-5.85) (-0.57) (-1.17) (-1.09)Age 0.067*** 0.001 0.060 0.039 0.050
(13.90) (0.04) (1.57) (0.91) (1.62)Male -0.094*** -0.208*** -0.223** -0.235*** -0.220**
(-3.11) (-7.79) (-2.17) (-2.66) (-2.48)Head's years of schooling 0.036*** -0.022 0.022 0.007 0.016
(6.33) (-0.98) (0.65) (0.21) (0.59)Ethnic major group 0.906*** 0.021 0.925** 0.762* 0.853**
(12.73) (0.05) (2.53) (1.78) (2.51)Log of total hh exp. 0.217*** 0.337*** 0.188 0.221* 0.215*
(5.42) (8.44) (1.51) (1.87) (1.79)InstrumentsDistance to fam. center YGovernment reg. Y YParental siblings Y Y
Overid test (J statistic) 0.08Log likelihood -4625 -14547 -2204 -2177 -2172N 8844 7467 1149 1133 1133Note 1. Regressions control for regional and urban dummy variables.2. Cluster-robust t statistics in parentheses.3. Overidentification tests are from linear regression.
21
Summary of findings Larger number of siblings predicts lower educational
investment in Vietnam, in un-instrumented regressions Result holds for both school enrolment and use of private
tutoring IV analysis partially confirms this quality-quantity
tradeoff Impact of sibship size on school enrolment is strongly
negative (from -0.5 to -1.0 per sibling) and significant across instruments
Impact on tutoring investment is not robustly significant, though always negative
Distance to family planning center seems the most promising instrument; other instruments yield mixed results, perhaps due to small N and data issues
Coefficients are generally larger in IV than in un-instrumented regressions
Implications Better availability of family planning may increase
investment in education Two-child policy may have led to more education in
Vietnam
22
Development, modernization, and childbearing: The role of family gender composition
Deon Filmer, Jed Friedman, and Norbert Schady (all World Bank)
23
Research question and methodology Research question: What is the relationship
between continuing fertility and the gender make-up of existing children? Focusing on one indicator of preference for sons
over daughters: son preference in fertility decisions
Note that differential gender-related behavior could be the result of “taste-based” gender discrimination, but also other causes
Focus here is on measuring the extent of son-preferred differential stopping behavior (DSB), regardless of its causes
Methodology Calculate probability of additional birth if zero sons
vs. zero daughters in family already Data: 158 DHS surveys, covering 1.3m women
from 64 countries
24
Results: Differential Stopping Behavior largest
in Central Asia (9.4 percentage
points) South Asia (7.8 percentage
points) Middle East/North Africa (5.8
percentage points) No clear evidence of DSB in
Sub-Saharan Africa Latin America and the Caribbean
0 0.05 0.1 0.15
Sub-Saharan Africa
Southeast Asia
South Asia
Central Asia
Middle East/NorthAfrica
LatinAmerica/Caribbean
After zero sons After zero daughters
Probability of an additional birth
DSB
Where and when do we see the greatest differentials in stopping behavior?
25
Where and when do we see the greatest differentials in stopping behavior? Son preference increases at higher birth orders
Mean number of children per family is 4.1 in Eastern Europe and Central Asia (ECA), and 4.9 in South Asia.
In such high-fertility settings, the gender composition of lower-parity children is less important in determining future fertility.
But once parents are closer to achieving their total desired number of children, the gender composition of children already born becomes an important determinant of whether parents have another child. For example, families with 4 or 5 children in South Asia
are approximately 14 percentage points more likely to add another child if all of the children up to this point are girls rather than boys.
26
Does “modernization” reduce differential stopping behavior?
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
Sub-Saharan Africa
Southeast Asia
South Asia
Central Asia
Middle East/North Africa
Latin America/Caribbean
Six or more years of schooling Less than six years of schooling
-0.02 0 0.02 0.04 0.06 0.08 0.1 0.12
Sub-Saharan Africa
Southeast Asia
South Asia
Central Asia
Middle East/North Africa
Latin America/Caribbean
Urban Rural
Differential Stopping Behavior
Urbanization and female education are often associated with higher, not lower, son preference in continuing fertility
For example, in South Asian countries, son preference is significantly greater for women in urban areas or with more education--and this pattern seems to have increased over time.
It’s possible that latent son preference manifests itself when fertility levels are low—that is, when families are closer to desired fertility at low parity—and indeed fertility has fallen among women in urban areas or with more education.
27
Possible implications of DSB for investment in girls Differential stopping behavior driven by son
preference is likely to exacerbate other forms of gender discrimination Mean number of siblings of girls exceeds boys’ in
regions where DSB is high (sons are preferred). Girls in South Asia have about 0.13 more siblings than boys in the Central Asian countries, the comparable number is
0.10 in Sub-Saharan Africa, boys and girls have the same
number of siblings Studies on the association between family size and child
outcomes usually show that more siblings dilute household and parental resources devoted to each child, a “quantity-quality” tradeoff.
If this association is causal, son preference, as manifested in gender-specific fertility choices is likely to have adverse consequences for girls since they will grow up in larger families.
28
Long-Term Financial Incentives And Investment In Daughters: Evidence From Conditional Cash Transfers In North India
Nistha Sinha (World Bank) and Joanne Yoong (RAND)
29Sinha and Yoong (2009)
Background and Program Description Gender Bias in Haryana State, North India
One of India’s richest states, but among the worst in terms of female disadvantage 1990s: evidence of consistent gender gap in
sex ratios at birth (Sudha and Rajan,1999) early childhood mortality (Filmer, King and Pritchett,1998) school enrollment for 6-14 year olds (Filmer and Pritchett,
1998)
October 1994: Haryana State Government introduced Apni Beti Apna Dhan (ABAD), a conditional cash transfer program to address these issues Upon the birth of a daughter, families receive
Immediate cash grant of Rs. 500 to cover post-delivery needs
Government savings bond in daughter’s name, redeemable for Rs 25,000 (about $550) only on 18th birthday if still unmarried
Additional bonuses for completed education or claim deferral
Subject to belonging to poor or low caste households
30Sinha and Yoong (2009)
Program Evaluation Strategy and Empirical Challenges
Full evaluation some years away: first beneficiaries turn 18 in 2012
Empirical challenge 1: No systematic data collection; uniformly implemented across Haryana in October 1994 without piloting
Solution: Use data from India’s National Family Health Surveys (NFHS); repeated cross-sections
Empirical challenge 2: No measures of actual participation in NFHS Solution: Program evaluation is limited at best to an intent-to-treat analysis
(measuring effects of being eligible ); use data on eligibility criteria to identify “eligible individuals” among poor households or households belong to certain castes
Empirical method: Basic Difference in Difference Specification (i.e. before and after program, eligible non-eligible girls or households)
1992 NFHS 1 Before program
1998/9 NFHS 2
1994 Program rolled out statewide
2005-06 NFHS 3
31
Results: Impact of the program (but recall the challenges of the program evaluation) Increased girl child survival
Positive, significant estimated effects on sex ratio of living children for individual women
Perhaps due to less sex-selective abortion, since insignificant (but consistently positive) estimated effects on survival rates in early childhood
However, no effects on expressed preferences for girls
Increased health investment in children Positive, significant effects on childhood vaccinations
Effects on education and marriage are limited by time horizon of data, but early results for education suggest positive relationship
32
Thematic Area 2: Fertility, Poverty and Family Welfare in the time of HIV/AIDS HIV/AIDS is the leading cause of prime-age death in
Africa. Early sexual initiation, early marriage, risky sexual practices, and commercial sex work have all contributed to the transmission of the pandemic—with consequences for the wellbeing not only of the person who has AIDS, but also for others in their household.
These studies seek to understand the socio-economic consequences of early marriage and non-marital sexual relations, and of efforts to reduce premature adult mortality through the use of anti-retroviral therapy
2 research projects in this category which are entail longitudinal surveys: Marriage Transitions and HIV/AIDS in Malawi HIV/AIDS and the impact of treatment on family and
individual welfare
33
Marriage Transitions and HIV/AIDS in Malawi
Kathleen Beegle (World Bank), Berk Ozler (World Bank) and Michelle Poulin (Brown University)
34
Research question and methodology Research question: What is the relationship
between socioeconomic characteristics of young people, economic shocks they experience, their partnership choices, sexual behavior, and risk of HIV infection? To explore in detail young people’s transition into
marriage and the effect of these transitions on their subsequent outcomes, such as health, fertility, labor market participation and important outcomes for their young children, such as anthropometrics, nutrition, cognitive ability, etc
Methodology: new data collection effort Surveys integrated over topics not normally
covered in traditional household surveys. Specific sample of young adults. The study is
following an initially never-married sample of 1,185 young Malawians for at least 3 years, using an array of panel data collection methods.
35
Details of the data effort
Annual household survey started in summer 2007. Modified LSMS-style household questionnaire, accompanied by a detailed individual component on marital aspirations and sexual behavior .
Interim in-depth partnership interviews (PIs) collected from Feb-March between rounds of the annual household survey.
HIV testing introduced in summer 2008 on a random sub-sample (to address concerns of the influence of testing itself on subsequent behaviors/outcomes).
Tracking individuals who move: This is a highly mobile population.
Last round of data planned for summer 2009.
36
Preliminary findings
28 of the 596 young women, aged 14-21 in our sample (5%) have ever given birth and all of them have given birth only once.
The mean (and median) age at birth for these young women is 17 (youngest 14 and oldest 20).
35 of the 583 men, aged 17-25 in our sample (6%) reported than a women has given birth to their child at least once.
The mean (and median) age for these young men when the at the time of the first birth was 19 (youngest 16 and oldest 22).
Approximately, one third of the women and three quarters of the men reported ever having sex.
37
HIV/AIDS and the impact of treatment on family and individual welfare
Damien de Walque (World Bank), Harounan Kazianga (World Bank) and Mead Over (CGD)
38
Research question and methodology Research question: What is the impact of
HIV treatment on… Lives saved and health outcomes Labor supply of patient and family members Schooling and welfare of children Other welfare indicators
Methodology: new data collection effort Biomedical follow-up including data on treatment
regimen and treatment success (CD4 counts) Household surveys (HIV patients and general
population) including health, schooling, labor force. 7 countries: Burkina Faso, Ghana, Kenya, India,
Mozambique, Rwanda and, South Africa
39
Methodological challenges
It is not possible to randomize ARV treatment!
But in some countries, can evaluate some experiments on the conditions of ARV delivery.
Rwanda: performance-based contracting for HIV/AIDS services in health facilities
South Africa: food and counseling intervention as adherence support.
Kenya: text messaging intervention as reminders for adherence
40
Preliminary findings from baseline surveys Access to pediatric ART appears limited
(evidence from Mozambique and Ghana) Parents and family might not identify weak or sick
children as suffering from HIV/AIDS HIV/AIDS affects not only the mental health
of persons with AIDS but also affects the mental health of family members in these households (evidence from Ghana)
Compared to other patients, HIV/AIDS patients seem to receive better health services (evidence from Burkina Faso): They wait less They receive higher quality care
41
Thematic Area 3: Fertility and female labor supply
Motivating question: What is the relationship between fertility outcomes and women’s labor market participation?
2 research projects using household survey data Fertility and women’s labor force
participation (96 DHS surveys) Fertility and Women’s Labor Supply in A
Low Income Rural Economy (the case of Matlab, Bangladesh)
42
Fertility and women’s labor force participation
Elizabeth King (World Bank) and Maria Porter (University of Chicago)
43
Research question and methodology Research question: This study focuses on the
relationship between fertility outcomes and women’s labor market behavior. As fertility declines around the world, childbearing
patterns change in three ways: women may delay their first birth, space their births, or stop having children at an earlier age than previous cohorts. Each of these changes is likely to have a different impact on the ability of women to work outside the home and on the decisions they make regarding work and child-bearing
Methodology: Analysis of 96 Demographic & Health Surveys in 59
countries
44
Methodological challenges
Endogeneity between fertility and labor market behaviors of women. Most previous studies of the relationship between fertility
and labor force participation have relied on cross-sectional data, but with cross-sectional data, it is difficult to correct for both the endogeneity of fertility and the impact of unobserved heterogeneity among women.
In the absence of natural experiments that may affect fertility choice but not otherwise affect other behaviors such as child outcomes, an econometric approach is needed in order to identify and quantify such an effect
Using exogenous shocks to fertility (twins in first birth and sex of first two births), we estimate how fertility affects women’s labor force participation decisions across different regions of the developing world
45
Effect of Sex at 1st Birth on Women’s LFP in Sub-Saharan Africa
Women Ages:15-44 15-24 25-34 35-44
Regression 31st 2 children: same sex -0.003* -0.002 -0.002 -0.004
(0.002) (0.005) (0.003) (0.003)1st child: boy -0.005*** 0.006 -0.007*** -0.008***
(0.002) (0.005) (0.003) (0.003)2nd child: boy -0.003* 0.002 -0.002 -0.005*
(0.002) (0.005) (0.003) (0.003)Regression 4
1st child: boy -0.002 0.005 -0.006 -0.003(0.002) (0.007) (0.004) (0.004)
1st 2 children: both boys -0.006*** -0.001 -0.004 -0.009**(0.002) (0.007) (0.004) (0.004)
1st 2 children: both girls 0.000 -0.004 -0.001 0.001(0.003) (0.007) (0.004) (0.004)
46
Summary of Findings Women are more likely to have worked in
the past year if they have more children in sub-Saharan Africa, and for some older women in South Asia.
Younger women in South Asia face the tradeoff between more children or work in the labor force. These women are less likely to have worked in the past two years as a consequence of having more children.
47
Summary of Findings for Sub-Saharan Africa, where the income effect dominate Women have more children if they had twins in
the first birth, if the first two births were the same sex, or if the first two births were girls.
Women have fewer surviving children if their first or second child was a boy.
Positive effects on the number of surviving children are strongest for women who have completed secondary schooling or higher.
More educated women are also more likely to work when they have twins in the first birth.
Any effect of the sex of first birth(s) does not vary much by education.
48
Fertility and Women’s Labor Supply in A Low Income Rural Economy: the case of Matlab, Bangladesh
Mattias Lundberg (World Bank), Nistha Sinha (World Bank) and Joanne Yoong (RAND)
49
Study Objectives & Context Explore effect of children on women’s labor force
participation using data from rural Bangladesh
Rural labor market characterized by Low female participation in wage labor Cultural practice of female seclusion Home based production by women
Women’s work in rural Bangladesh: Based on a question about primary activity, 58% of women and 82% of men aged 20-55 are “working”: Among those who report earnings (1995),
Men’s mean earnings were 21,370 Takas Women’s mean earnings were 3,005 Takas
Most women’s (87%) location of work is home Women report activities such processing rice, raising
poultry and livestock
50
Data and methodology Data: Matlab Health and Socioeconomic Survey 1996
Survey of 4,363 households in a demographic surveillance area in rural Bangladesh.
Surveillance area is site of a family planning program experiment
Survey covered 142 villages in treatment and control areas Methodology: Identify causal effect of children on
women’s labor supply Unobservables influence both women’s decision to work and
their family size Standard methodology finds a variable that influences
fertility but not labor-force participation (Literature: twins, sex of first-born)
This paper exploits women’s exposure to a family planning program experiment
51
Initial finding Number of children is positively
associated with women’s probability of engaging in home-based work and negatively associated with work outside the home
Finding appears consistent with home-based production technology which allows women to combine child care and work
Next step: Conducting robustness checks
52
Research program summary findings Fertility and investments in child quality
Higher fertility is associated with lower parental investment in children, in some cases reflecting a quantity-quality tradeoff Larger sib-size reduces school enrollment and overall investments
in children in Vietnam Parents who will go on to have larger families invest less in
children today in Ecuador Son preference adds another twist to the problem
Parents stop childbearing earlier if have son(s), so girls are more likely to end up in larger families and therefore suffer more from resource dilution
But this DSB can change: CCTs in North India have some positive effects on improving sex ratios
Fertility, poverty, and family welfare in the time of HIV/AIDS Panel data collection still underway
The link between fertility and female labor supply varies by region Recent births are associated with higher female labor-force
participation in sub-Saharan Africa, but lower in South Asia In Bangladesh, higher fertility is associated with lower labor-force
participation of women outside the home, but higher participation in home-based income-earning activities