by eugenie w. h. maïga african center for economic transformation (acet)

The Impact of Mother’s Education on Child Health in Developing Countries: Evidence

from a Natural Experiment in Burkina Faso

By Eugenie W. H. Maïga

African Center for Economic Transformation (ACET)

Background on Burkina FasoLand area = 274,000 km2, Population

=15.21 million (2008), GNI per capita =$480 (2008).

In Burkina Faso, the prevalence rate for stunting (height-for-age) was 39% in 2003.

In 2009 the gross school enrolment rate for primary, secondary, and higher education were 78%, 20%, and 3%, respectively.

Gross enrollment rates in primary school by province in 2004-2005

Source: Ministere de l’Enseignement de Base et de l’Alphabetisation, 2005

Research problemChild health and nutrition outcomes in

developing countries are disappointing (prevalence of stunting in preschool children was 33% in 2000, De Onis et. al., rates between 30% and 39% are considered high).

Link between child health and child nutrition: malnourished children are more likely to develop illnesses that can have long lasting effects throughout their lives.

Mothers play an important role in child nutrition. How does her education come into play?

Why are child health outcomes important?Costs of related illnesses (physical suffering,

time costs and monetary costs).Impacts on ability to learn i.e. on schooling

outcomes.Impacts on productivity later in life which

could adversely affect economic growth

Objectives General objective:

• Examine the relationship between maternal education and child health in Burkina Faso

Specific objectives:• Verify whether a strong causal relationship

exists• Understand the channels through which

mother’s education affects child health• investigate whether threshold effects exist,

that is, whether specific years or levels of mothers’ education have unusually large impacts on children’s health.

Change in education policy in Burkina FasoPrimary education: construction of 3-year

primary schools Ecoles Satellites, ES) in areas of 28 provinces (out of 45) where enrolment rates are low. As of 2008, there were 304 ES.

Became effective in the school year 1995-1996. Instruction is done in two languages, French and language spoken in the area.

Trying to favor girls enrollment in those schools

Evolution of enrollment rates in provinces with and without the program

Provinces with program

Provinces without program

Previous EvidenceEmpirical studies without pathways (did

not use natural experiment): Baya (1998), Appoh and Krekling(2005).

Empirical studies without pathways (used natural experiment): Breierova and Duflo (2004) and Chou et. al. (2007).

Empirical studies with pathway(s): Thomas et al. (1991), Glewwe (1999), Handa (1999), and Webb and Block (2004).

ContributionMany have studied the subject but few

studies have used natural experiments (change in policy) in identifying the relationship between mother’s education and child health in developing countries

Use of distance to school as IV for mother’s education in a study of the impact of mother’s education on child health

Use of both natural experiment and pathways

Conceptual frameworkDraws on Becker’s model of the family, and

Rosenzweig and Schultz (1983) health production function

Consider the following equations:

Where X is market goods, Z represents goods that affect child health, H is child health, I represents inputs that affect utility only through their effect on H and μ is child health endowment.

IZX

IZ

X

321

),,(

),(

pppM

hH

HuU

Conceptual frameworkEstimating H requires detailed data on health

inputs, data which are not available to this study. A way out of that, is to use reduced form equations.

Maximizing U subject to the health production and the budget constraint yields reduced-form demand functions for X, Z, H and I.

DataMain source: Burkina Faso 2007 Enquete sur les

conditions de vie des menages (survey of living standards).

Additional data sources include the ministry of primary education (policy change data: number, location, and year of construction of the schools, and enrollment information) and Internet sources (distance calculator).

Distance between the closest school (village/city name) and the household village/city of residence computed using distance calculator at www.infoplease.com/atlas/calculate-distance.html

Descriptive Statistics: Child Characteristics

Variable Obs MeanStd. Dev. Min Max

Child’s age in months 2,003 27.3 16.8 0 59

Male 2,007 52.7% 0.50 0 1

HAZ 2,003 -1.8 2.05 -5.98 5.8

Stunting 2,003 47.6% 0.50 0 1

Moderate stunting 2,003 19.4% 0.40 0 1

Severe stunting 2,003 28.3% 0.45 0 1

WHZ 1,932 -0.3 2.04 -4.96 5

Wasting 1,932 20.7% 0.40 0 1

Moderate wasting 1,932 9.5% 0.29 0 1

Severe wasting 1,932 11.2% 0.32 0 1

Descriptive Statistics: Regressors

Variable Obs Mean Std. Dev. Min MaxZero to 5km of program school 2,007 6.9% 0.25 0 1Five to 10km of program school 2,007 10.1% 0.30 0 1Number of years ES was available 2,007 0.4 1.05 0 5Mother’s years of education 2,007 1.4 3.12 0 17Mother’s age 2,007 24.3 3.29 14 29Father’s years of education 2,007 2.1 4.15 0 17Father’s age 2,004 35.7 11.44 15 97Monthly expenditures (1,000 FCFA) 2,005 8.2 12.44 0 225Urban residence 2,007 24.5% 0.43 0 1Television 2,007 20.1% 0.40 0 1Radio 2,007 74.9% 0.43 0 1Refrigerator 2,007 5.4% 0.23 0 1Bicycle 2,007 86.8% 0.34 0 1Moped 2,007 38.5% 0.49 0 1Car 2,007 2.4% 0.15 0 1Electricity 2,007 15.8% 0.37 0 1Cell phone 2,007 25.0% 0.43 0 1Mother’s health knowledge 2,007 3.7 1.80 0 8Mother’s bargaining power 2,007 65.4% 0.48 0 1

Descriptive Statistics: Histogram of parents education

0.2

.4.6

.8Fr

actio

n

0 5 10 15 20momedu_years

0.2

.4.6

.8Fr

actio

n

0 5 10 15 20dadedu_years

Selected child health outcomes by parents’ education

Stunting Wasting

Mother schoolingNo schooling 50.7% 21.5%

Some schooling

36.1% 17.5%

Father’s schoolingNo schooling 51.2% 21.2%

Some schooling

38.4% 19.2%

Total 47.6% 20.7%

Methods and ProceduresParents’ education and child health maybe

affected by many ‘unobservables’. To identify the effect parents education on child health, I will use a system of equations:

iiki

iki

SH

S

X

XZ

Methods and procedures Use two-stage –least –squares to estimate

the system (education equation and child health equation) with the primary education program as an instrument for mother’s education.

Add income, mother health knowledge, mother’s bargaining power, etc. to the child health equation and estimate it.

RESULTS

Table 1: OLS Estimates of Mother’s Health Knowledge, Bargaining Power, and Household Expenditures on Mother’s Education and other Variables.

  OLS OLS OLS

VARIABLESPer capita

expendituresHealth

knowledge Bargain power       Mother’s education (log) 0.300*** 0.541*** -0.023

(0.031) (0.064) (0.018)Per capita expenditures (log) 0.084* -0.022

(0.049) (0.014)Mother’s age -0.001 0.041*** 0.018***

(0.007) (0.013) (0.004)Father’s education (log) -0.125*** -0.158*** -0.028**

(0.021) (0.043) (0.013)Father’s age -0.007*** -0.001 -0.000

(0.002) (0.004) (0.001)Urban residence 0.370*** 0.499*** -0.137***

(0.052) (0.117) (0.037)Has a television 0.225* -0.064

(0.128) (0.041)Has a radio 0.416*** -0.016

(0.094) (0.026)

Observations 1,934 1,934 1,934R-squared 0.164 0.173 0.056

Table 2: Reduced Form Estimates for Child Height-for-Age (HAZ): Pathways Added

  Base FE IncomeHealth

KnowledgeBargaining

powerCommunity

services All pathwaysVARIABLES HAZ HAZ HAZ HAZ HAZ HAZ             Mother’s education (log) 0.133** 0.083 0.123* 0.139** 0.130* 0.079

(0.064) (0.061) (0.064) (0.065) (0.065) (0.062)Child’s age (months) -0.204*** -0.203*** -0.204*** -0.204*** -0.203*** -0.203***

(0.035) (0.035) (0.035) (0.035) (0.035) (0.035)Father’s education -0.008 -0.005 -0.008 -0.006 -0.008 -0.004

(0.053) (0.054) (0.053) (0.053) (0.054) (0.054)Per capita expenditures (log) 0.169*** 0.170***

(0.060) (0.061)Health knowledge 0.017 0.012

(0.027) (0.027)Bargaining power 0.143 0.152

(0.122) (0.119)Within 30 minutes of a health clinic 0.044 0.035

(0.107) (0.105)

Observations 2,000 1,998 2,000 2,000 2,000 1,998R-squared 0.101 0.104 0.101 0.102 0.101 0.105

Number of provinces 43 43 43 43 43 43

Table 3: Reduced Form Estimates for Child Weight-for-Height (WHZ): Pathways Added

  Base FE IncomeHealth

KnowledgeBargaining

powerCommunity

servicesAll

pathwaysVARIABLES WHZ WHZ WHZ WHZ WHZ WHZ             Mother’s education (log) 0.164* 0.157* 0.165* 0.156* 0.166* 0.154*

(0.086) (0.087) (0.087) (0.086) (0.082) (0.087)Age in months 0.011*** 0.011*** 0.011*** 0.012*** 0.011*** 0.012***

(0.003) (0.004) (0.003) (0.003) (0.003) (0.004)Father’s education -0.140** -0.142** -0.140** -0.143** -0.140** -0.145**

(0.058) (0.059) (0.057) (0.059) (0.058) (0.059)Per capita expenditures (log) 0.008 0.004

(0.072) (0.071)Health Knowledge -0.002 -0.003

(0.038) (0.037)Bargaining power -0.221** -0.218**

(0.097) (0.096)Within 30 minutes of a health clinic -0.021 -0.029

(0.111) (0.117)

Observations 1,925 1,923 1,925 1,925 1,925 1,923R-squared 0.020 0.020 0.020 0.022 0.020 0.022

Number of provinces 43 43 43 43 43 43

Table 4: First Stage IV Regression for Child Weight-for-Height (WHZ)

VARIABLESMother’s education

Mother’s education

Per capita expenditures

Zero to 5km of ES (d1) -0.041 -0.008 0.068(0.066) (0.049) (0.092)

Five to 10km of ES (d2) 0.157** 0.163** 0.010(0.076) (0.071) (0.079)

Number of years ES was available (ESyears) -0.016 -0.003 0.007(0.020) (0.019) (0.027)

Interaction d1*ESyears 0.349*** 0.407*** -0.168**(0.107) (0.124) (0.066)

Interaction d2*ESyears -0.049 -0.020 0.048(0.045) (0.042) (0.045)

Health knowledge 0.072*** 0.028**(0.010) (0.013)

Bargaining power -0.023 -0.035(0.030) (0.044)

Within 30 minutes of a health clinic 0.031 -0.007(0.035) (0.041)

Wealth index 0.368*** 0.320***(0.027) (0.026)

Observations 1925 1923 1923R-squared 0.117 0.299 0.161

Table 5: Second Stage IV Regressions for Child Weight-for-Height (WHZ)

  Basic model Full model without income Full model with incomeVARIABLES WHZ WHZ WHZ       Mother’s education (log) 1.735*** 1.736*** 1.287***

(0.579) (0.551) (0.326)Age in months 0.016*** 0.016*** 0.012***

(0.004) (0.004) (0.004)Father’s education -0.163** -0.157** -0.177**

(0.071) (0.071) (0.070)Health knowledge -0.162*** -0.065

(0.062) (0.040)Bargaining power -0.163 -0.287**

(0.129) (0.126)Within 30 minutes of a health clinic -0.145 -0.033

(0.149) (0.130)Per capita expenditures (log) -1.113***

(0.394)

Observations 1,925 1,925 1,923R-squared -0.244 -0.224 -0.227Number of provinces 43 43 43

Threshold effects

-2-1

01

23

Para

met

er e

stim

ate

0 _Imomedu_ye_5 _Imomedu_ye_10 log_pceMother's years of education

WHZ threshold regression coefficients

-2-1

01

2Pa

ram

eter

est

imat

e

0 _Imomedu_ye_5 _Imomedu_ye_10 log_pceMother's years of education

HAZ threshold regression coefficients

Summary of results

• Strong positive coefficient of mother’s education on WHZ (IV/FE estimation)

• Positive but insignificant coefficient of mother’s education for HAZ (OLS estimation).

• Household per capita expenditures are a pathway for impact on HAZ. Counterintuitive sign of father’s education, bargaining power and household expenditures in WHZ regression.

Summary of resultsEvidence of existence of threshold effects but

need to consider the cost.Keep girls in school as long as possible;

design nutrition programs targeted at girls/women.

Limitations: could not IV mother’s health knowledge, quality of bargaining power and health knowledge variables, no information on religion and ethnicity.

Questions?

Thank you!

Table A1: First Stage IV Regressions for Child Height-for-Age (HAZ)

VARIABLESMother’s education

Mother’s education

Mother’s education

Per capita expenditures

Zero to 5km of ES (d1) -0.045 -0.001 -0.006 0.025(0.063) (0.059) (0.047) (0.091)

Five to 10km of ES (d2) 0.155** 0.169** 0.156** 0.023(0.074) (0.074) (0.069) (0.076)

Number of years ES was available -0.013 -0.007 -0.001 0.001(0.020) (0.019) (0.019) (0.027)

Interaction d1*ESyears 0.326*** 0.329*** 0.380*** -0.165**(0.106) (0.112) (0.123) (0.067)

Interaction d2*ESyears -0.051 -0.042 -0.019 0.049(0.044) (0.044) (0.042) (0.045)

Health knowledge 0.095*** 0.070*** 0.030**(0.011) (0.010) (0.013)

Bargaining power -0.084** -0.026 -0.029(0.033) (0.030) (0.044)

Within 30 minutes of a health clinic 0.108*** 0.030 -0.004(0.037) (0.034) (0.040)

Wealth index 0.362*** 0.314***(0.027) (0.026)

Observations 2,000 2,000 1,998 1,998R-squared 0.116 0.177 0.296 0.157F (test of excluded instruments) 3.11 2.91 32.45 27.03P-value of F-test 0.008 0.013 0.000 0.000

Table A2: Second Stage IV Regressions for Child Height-for-Age (HAZ)

  Basic model Full model without income Full model with incomeVARIABLES HAZ HAZ HAZ

Mother’s education -0.637 -0.739 -0.355(0.514) (0.521) (0.301)

Child’s age (months) -0.206*** -0.207*** -0.199***(0.029) (0.029) (0.028)

Father’s education 0.028 0.024 0.034(0.061) (0.061) (0.061)

Urban residence 0.726* 0.713** 0.030(0.387) (0.331) (0.204)

Health knowledge 0.093 0.020(0.058) (0.034)

Bargaining power 0.130 0.217**(0.114) (0.106)

Within 30 minutes of a health clinic 0.118 0.030

(0.128) (0.111)

Per capita expenditures 0.713*(0.364)

Observations 2,000 2,000 1,998R-squared 0.047 0.038 0.054