Penn Majid Ramadan June12

The Persistent Eﬀects of in Utero Nutrition Shocks
over the Life Cycle: Evidence from Ramadan
Fasting in Indonesia
By Farhan Majid
University of California, Riverside
Farhan Majid (UCR) June 11, 2013 1 / 44

Introduction
Outline
Introduction
Conceptual Framework
Data
Empirical Methodology
Results
Conclusion

Introduction
Big Picture: Why are some individuals poorer than others?
Fetal Origins Hypothesis: Inequalities during adulthood have
their origins in the fetal environment
Pandemics, twin births, extreme droughts and civil wars during
early childhood can have potentially long-term effects
Little is understood about the effects of nutritional shocks from
day-to-day life activities based on norms and informal institutions
Mothers fasting during pregnancy may pose a serious concern for
the unborn child
Does exposure to Ramadan in utero have persistent effects on
the unborn child?

Introduction
Background on Ramadan fasting
Ramadan: Holiest month of the Islamic calendar lasting 29-30 days
Does not follow the seasons
Muslims required to fast during daylight hours for spiritual
development
Although pregnant women are exempt, many do fast:
Singapore: 74% fasted at least 1 day (Joosoph et al. (2004))
Sanaa City, Yemen: 90% fasted over 20 days (Makki (2002))
In 2010 alone, more than 1.2 billion Muslims were potentially aﬀected

Introduction
Brief literature review
Almond and Mazumder (2011) find
Lower birth weights and lower sex ratios in the US
Learning disabilities in Uganda and Iraq, and
Negative effects on crude wealth measures
Ewijk (2011) finds
Worse general health and symptoms of coronary heart problems
Type 2 diabetes in old age and lower sex ratios
Contemporaneously, AME (2011) find
Lower math and reading test scores for children age 7

Introduction
Contribution
Question: Does Ramadan exposure in utero affect children’s adult
labor market behavior?
Channels
Persistence over the life course
Exploit panel feature
Better Identification: Fixed effects; Religiosity Placebo
Better Data: Indonesian Family Life Survey
Less measurement error: Exact date of birth and religion is known
Local polynomial smooth plots: Monotonicity of Ramadan effects

Introduction
Results overview: Ramadan exposure in utero worsens
labor market outcomes
Those potentially exposed work 4.5% fewer hours and are 3.2% more
likely to be self-employed
Results robust to not only household fixed effects, but also to
biological sibling fixed effects for a sub-sample of adults
No effects on non-Muslims
Strongest effects on Muslims from most religious families

Introduction
Results overview: Ramadan exposure in utero worsens
outcomes over the life course
Fasting lowers Raven’s CPM cognitive test scores by 5.9% and math
scores by 7.8% for children age 7-15
Test scores are robust to biological sibling ﬁxed eﬀects
Lower schooling inputs: Children (age 7-15) are 3.3% more likely to
be involved in child labor and study 3.4% fewer hours during
elementary school
Lower birth weights for those 0-5

Fasting during pregnancy can worsen children’s health
Child
Health
Fasting
During
Pregnancy

Worse childhood health can lead to lower skills
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling

Lower adult human capital can lead to fewer hours worked
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling
Hours Worked

Low skilled workers work in less skill-intensive sectors
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling
Hours Worked
Self-Employment
Endogenous variables OutcomesIn Utero Shock

Does fasting during pregnancy worsen labor market
outcomes?
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling
Hours Worked
Self-Employment
Endogenous variables OutcomesIn Utero Shock

Data
Data: Indonesian Family Life Survey
Indonesia: 88% Muslim and largest Muslim country
Wave 4 ﬁelded in 2007-2008 and Wave 1 in 1993-1994
Nationally representative
Detailed individual and household data
Rich data for my use: date of birth, labor market outcomes,
religiosity, test scores, schooling, siblings
Adult sample restricted to Muslims of 15-65 years of age in Wave 4
(2007)
IFLS used previously in many other papers such (Strauss & Thomas
(1998), Duﬂo (2001), Maccini & Yang (2009))

Data
Key dependent variables: Adulthood
Log Hours: Natural logarithm of hours worked at primary job in a
normal week
Self-employment: Dummy (=1) for being self-employed and (=0) for
not
Most self employed don’t hire any permanent workers and have
extremely small businesses
Low skill sector (have fewer years of schooling and worse health than
wage workers)

Data
Key dependent variables: Childhood
Cog. Scores: Raven’s Colored Progressive Matrices (CPM)
assessment, a measure of general intelligence, in percentage terms
(age 7-15)
Math Scores: In percentage terms (age 7-15)
Child Labor: Dummy (=1) for working as a child and (=0) for not
(age 7-15); Measured in Wave 1
Log Hrs. School: Hours spent studying while at elementary school
(age 7-15); Measured in Wave 1
Birth Weight: Reported birth weight (age 0-5); Measured in Wave 1

Empirical strategy: Compare children exposed to
fasting in utero vs those not
Potentially exposed to Ramadan if it occurs between birth and
conception date. Otherwise they are not exposed
Assume mothers don’t time their pregnancy around Ramadan
Estimates are lower bounds
Measure exposure in two alternative ways:
Partially exposed: Proportion of days Ramadan overlaps with in
utero period
Fully exposed: Dummy (=1) if an entire Ramadan overlapped with
in utero period.

Who is fully exposed?
1960 1961 1962
Ramadan
Exposed
Year (Gregorian Calendar)
Ramadan

Who is partially exposed?
1960 1961 1962
Exposed
Ramadan Ramadan

Who is not exposed?
1960 1961 1962
Not Exposed
Ramadan Ramadan Ramadan

How full, partial and no exposure is determined
1960 1961 1962
Fully Exposed
Not Exposed
Fully Exposed
Partially Exposed
Ramadan Ramadan Ramadan

Regression equations
OLS formulation is as follows:
Yif = α + β1exposureif + β2Xif + FCf + Uif , (1)
where
Yif are human development outcomes individual i belonging to family f .
Exposureif is a dummy for potential exposure to Ramadan for a full
month in utero. Xif are controls
FC- family level time invariant unobservables

Regression equations
Fixed Eﬀects formulation is as follows:
∆Y[t1,t2] = α + β1∗exposure[t1,t2] + β2∗∆X[t1,t2] + ∆V[t1,t2]. (2)
where
t1 and t2, the date of births of members of the family ‘f’.
Identifying assumption: timing of birth and timing of Ramadan is
exogenous and relevant family level covariates are time-invariant.

Results
Longer exposure reduces hours worked for overall
sample 3.563.583.63.623.64
LogHoursAtPrimaryJob
0 .2 .4 .6 .8 1
Proportion of Days Potentially Exposed
95% CI Log Hours
Exposed Work Fewer Hours

Results
Longer exposure increases probability of
self-employment for overall sample.26.28.3.32.34
Self−employed
0 .2 .4 .6 .8 1
Proportion of Days Potentially Exposed
95% CI Self−employed
Exposed Are More Likely To Be Self−employed

Results
Table: Summary of Key OLS Estimates
Muslims Non-Muslims
VARIABLES Exposed Exposed
Log Hours -0.045** -0.027
(0.020) (0.071)
Observations 8,051 1,035
Self-employed 0.032** -0.089**
(0.014) (0.041)
Observations 8,373 1,069
Cognitive Scores -0.059*** 0.083
(0.018) (0.054)
Observations 3,514 379
Math Scores -0.078*** 0.036
(0.023) (0.048)
Observations 3,521 380
Birth Weight -0.271* 0.449
(0.153) (0.369)
Observations 828 144
Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Robust standard errors are clustered at current household level.

Results
Table: Estimates From IFLS 4 for Adult Labor Market Behavior
OLS OLS-Rest. Fixed Effect OLS OLS-Rest. Fixed Effect
VARIABLES Log Hours Log Hours Log Hours Self-employed Self-employed Self-employed
Muslims
Exposed -0.045** -0.088*** -0.103*** 0.032** 0.079*** 0.078**
(0.020) (0.033) (0.040) (0.014) (0.023) (0.036)
Observations 8,051 2,859 2,859 8,373 2,968 2,968
The OLS-restricted and fixed effect estimates are clustered at household level.

Results
Muslims
Exposed -0.045** -0.088*** -0.103*** 0.032** 0.079*** 0.078**
(0.020) (0.033) (0.040) (0.014) (0.023) (0.036)
Observations 8,051 2,859 2,859 8,373 2,968 2,968
Religious Muslims
Exposed -0.054** -0.131*** -0.133* 0.040** 0.097*** 0.119***
(0.026) (0.043) (0.069) (0.018) (0.033) (0.041)
Observations 5,031 1,412 1,412 5,232 1,465 1,465

Results
Muslims
Exposed -0.045** -0.088*** -0.103*** 0.032** 0.079*** 0.078**
(0.020) (0.033) (0.040) (0.014) (0.023) (0.036)
Observations 8,051 2,859 2,859 8,373 2,968 2,968
Religious Muslims
Exposed -0.054** -0.131*** -0.133* 0.040** 0.097*** 0.119***
(0.026) (0.043) (0.069) (0.018) (0.033) (0.041)
Observations 5,031 1,412 1,412 5,232 1,465 1,465
Less Religious Muslims
Exposed -0.029 -0.049 -0.073 0.016 0.063** 0.051
(0.034) (0.049) (0.068) (0.023) (0.032) (0.041)
Observations 3,015 1,447 1,447 3,134 1,503 1,503

Results
Muslims
Exposed -0.045** -0.088*** -0.103*** 0.032** 0.079*** 0.078**
(0.020) (0.033) (0.040) (0.014) (0.023) (0.036)
Observations 8,051 2,859 2,859 8,373 2,968 2,968
Religious Muslims
Exposed -0.054** -0.131*** -0.133* 0.040** 0.097*** 0.119***
(0.026) (0.043) (0.069) (0.018) (0.033) (0.041)
Observations 5,031 1,412 1,412 5,232 1,465 1,465
Less Religious Muslims
Exposed -0.029 -0.049 -0.073 0.016 0.063** 0.051
(0.034) (0.049) (0.068) (0.023) (0.032) (0.041)
Observations 3,015 1,447 1,447 3,134 1,503 1,503
Non-Muslims
Exposed -0.027 0.087 -0.008 -0.089** -0.144** 0.000
(0.068) (0.116) (0.154) (0.042) (0.067) (0.101)
Observations 1,035 377 377 1,069 392 392

Results
Does fasting during pregnancy worsen schooling outcomes?
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling
Hours Worked
Self-Employment

Results
Table: Estimates For Test Scores for Children Aged 7-15
VARIABLES Cog. Scores Cog. Scores Cog. Scores Math Scores Math Scores Math Scores
Exposed -0.059*** -0.085*** -0.100*** -0.078*** -0.099*** -0.143***
(0.018) (0.024) (0.035) (0.023) (0.033) (0.052)
Observations 3,514 2,084 2,084 3,521 2,087 2,087
Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.The OLS-restricted and fixed effect estimates
are clustered at the mother level. Standard errors are bootstrapped for fixed effect estimates. ‘Fixed effects’ are biological
siblings fixed effects. Sample is restricted to Muslim children who were 7-15 year old in 2007. The OLS-restricted limits
sample further to those households with two or more household members.

Results
Table: Estimates For Schooling Inputs for Children Aged 7-15 in IFLS1
OLS OLS-Res. Fixed Effect OLS OLS-Res. Fixed Effect
VARIABLES Log Hrs. School Log Hrs. School Log Hrs. School Child Labor Child Labor Child Labor
Exposed -0.034 -0.046 -0.100* 0.033*** 0.039*** 0.016
(0.024) (0.034) (0.055) (0.012) (0.014) (0.023)
Observations 1,815 941 941 2,164 1,117 1,117
Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The OLS-restricted and fixed effect estimates
at the mother level. Standard errors are bootstrapped for fixed effect regressions. ‘Log. Hrs School’ are hours spent studying
while at elementary school. ‘Fixed effects’ are biological siblings fixed effects. Sample is restricted to Muslim children aged
7-15 in 1993. The OLS-restricted limits sample further to those households with two or more members.

Results
Fasting during pregnancy worsens birth outcomes
Child
Health
Fasting
During
Pregnancy
Adult Health
Stock of
Cognitive & Non-
Cognitive Skills
Years of
Schooling
Hours Worked
Self-Employment

Results
Other Results: Estimates vary by trimester
Fasting lowers birth weight particularly in the second and third
trimesters. Low blood glucose can explain such effects (Scholl et al.
(2001)
Fasting lowers math scores in the first and third trimesters. Low
blood glucose can explain the first trimester effects (Rizzo et al.,
1991)
Cognitive scores lower in third trimester. Lower fetal heart rates can
explain this effect (Mirghani et al., 2005)
The labor supply and self-employment estimates suggest that the
first two trimesters are particularly sensitive periods
There is heterogeneity by sample: For large enough families, last two
trimesters are more sensitive

Results
Further robustness exercises
Placebos: No effects found for non-Muslims on the schooling and
birthweight variables
Biological siblings fixed effects for adults
Selective timing of pregnancy: Cohort size does not differ by
exposure to Ramadan
Selective timing of pregnancy: Robust to birth order dummies
No effects found on labor force participation

Conclusion
Results summary: Ramadan exposure in utero worsens
labor market outcomes
Those potentially exposed work 4.5% fewer hours and are 3.2% more
likely to be self-employed
Results robust to sibling fixed effects
No effects on non-Muslims
Strongest effects on Muslims from most religious families

Conclusion
Results summary: Ramadan exposure in utero outcomes
over the life cycle
Fasting lowers Raven’s CPM cognitive test scores by 5.9% and math
scores by 7.8% for children aged 8-15
Test scores are robust to biological sibling ﬁxed eﬀects
Lower schooling inputs: Children in Wave 1 (age 7-14) are 3.3%
more likely to be involved in child labor and study 3.4% fewer hours
during elementary school
Lower birth weights for those 0-5 in Wave 1 (15-20 in 2007)

Conclusion
Future Work
Do parents mitigate or reinforce in utero shocks?
Do public health programs mitigate Ramadan eﬀects?
Do awareness programs make a diﬀerence? An experiment
Demand for maternal fasting- why do pregnant women fast?

Conclusion
THANK YOU !

Extras
Nothing seems more middle class than the fact of having a steady
well-paying job. While there are many petty entrepreneurs among the
middle class, most of them do not seem to be capitalists in waiting.
They run businesses, but for the most part only because they are still
relatively poor and every little bit helps. If they could only ﬁnd the
right salaried job, they might be quite content to shut their business
down” - Banerjee and Duﬂo (2007, p. 162) write:
”Typically, the better jobs are in wage employment, not
self-employment.” - Gary S Fields (2010)
A simple mean comparison (t-test) of outcomes such as general
health and years of schooling reveals that those self-employed
complete 1.3 less years of schooling, score 0.09 of a st. dev lower on
general health measure.

Extras
First stage effects: Fasting during pregnancy has
heterogeneous effects by trimester
Table: Summary of Select Medical Studies
Outcome Possible Channel Trimester
Birthweight Low Blood Glucose Level (Scholl et al., 2001) Third
Birthweight Shorter Gestation (Siega-Riz et al., 2001) Second/Third
Birthweight Ketones and HPA Axis First
Long-term Outcome
Diabetes Fetal nutrition First
Heart DIsease Fetal nutrition First/Second
Cognitive Function Low Blood Glucose Level (Rizzo et al., 1991) First
Cognitive Function Fetal Heart Rate (Mirghani, 2006) Third
Note: The above table, adapted largely from Almond and Mazumder (2011),
is based on a review of very select studies from the medical literature. Both animal
and human studies are included. In many studies, only a select trimester was studied,
so that the effect in one trimester does not necessarily rule out the effects in other trimesters.

Extras
Table: Summary Statistics by Exposure and Religion: Adults (age 15-65)
Wave 4
Muslims Muslims Non-Muslims Non-Muslims Total
Exposed Not Exposed Exposed Not Exposed Total
Age 33.08 32.82 34.05 34.94 33.16
(12.51) (12.25) (13.43) (13.85) (12.59)
Male 0.495 0.518 0.513 0.560 0.500
(0.500) (0.500) (0.500) (0.498) (0.500)
Religiosity 2.781 2.770 2.940 2.903 2.796
(0.464) (0.468) (0.424) (0.474) (0.463)
Work 0.679 0.678 0.729 0.820 0.685
(0.467) (0.467) (0.445) (0.385) (0.464)
Log Hours 3.581 3.631 3.527 3.541 3.582
(0.666) (0.611) (0.696) (0.734) (0.664)
Self-employed 0.308 0.270 0.309 0.392 0.305
(0.462) (0.444) (0.462) (0.490) (0.460)
Observations 10207 1630 1223 191 13251
Note: Mean of each variable with standard deviation in parentheses. Sample does not include those conceived less than
21 days after the end of Ramadan.

Extras
Table: Summary Statistics by Exposure and Religion: Children
Muslims Muslims Non-Muslims Non-Muslims Total
Exposed Not Exposed Exposed Not Exposed Total
Wave 4 Children: 8-15 years
Cognitive Scores 0.751 0.761 0.735 0.690 0.750
(0.226) (0.216) (0.247) (0.267) (0.227)
Math Scores 0.584 0.596 0.578 0.600 0.585
(0.263) (0.270) (0.261) (0.217) (0.263)
Total Scores 0.697 0.706 0.683 0.662 0.696
(0.209) (0.201) (0.223) (0.230) (0.210)
Observations 3615 543 390 67 4615
Wave 1 Children: 7-14 years
Hours Studied-Elem. 1.441 1.448 1.497 1.532 1.451
(0.263) (0.269) (0.200) (0.154) (0.256)
Child Labor 0.0170 0.00673 0.0242 0.0200 0.0168
(0.129) (0.0819) (0.154) (0.141) (0.128)
Observations 2235 372 432 72 3111
Wave 1 Infants: 0-5 years
Birth Weight 3.087 3.181 3.126 3.374 3.123
(0.550) (0.554) (0.597) (0.584) (0.573)
Observations 477 52 339 53 921
Note: Mean of each variable with standard deviation in parentheses. Sample does not include those conceived less than
21 days after the end of Ramadan.

Penn Majid Ramadan June12

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (7)

Similar to Penn Majid Ramadan June12

Similar to Penn Majid Ramadan June12 (20)

Recently uploaded

Recently uploaded (17)

Penn Majid Ramadan June12