Presentation on theme: "Association Between Parental Resources and Child Development in Peru"— Presentation transcript:
1 Association Between Parental Resources and Child Development in Peru From Parent to ChildAssociation Between Parental Resources and Child Development in PeruMilo VandemoorteleDSA ConferenceNovember, 2013
2 Overarching Research Question How and to what extent are parental resources associated with child development? A case study of Peru
3 Why do research on this topic? A “conservative estimate that more than 200 million children under 5 years fail to reach their potential in cognitive development because of poverty…” and its attendant problems.1Long term consequences2Poor children are more likely to remain poorLimited research on the topic1 Grantham-McGregor et al., 2007:60; 2 Walker et al., 2011
4 Research QuestionsHow and to what extent are parental resources associated with child development? Particularly focusing on:Parental resources at birth and each subsequent stages of lifeChanges in parental resourcesDifferential effect of parental wealth versus parental expenditureShape of the association – is it linear?Does the effect of parental resources differ for poorer children versus wealthier children
5 Analytical FrameworkSource: Adapted from Haveman and Wolfe (1995), Grantham-McGregor et al. (2007)
6 Data: Young LivesFour countries: Ethiopia, Andhra Pradesh (India), Peru and VietnamTotal sample of 12,000 children followed over a period of 15 yearsSample size used here, younger cohort in Peru ~2000 children born in 2001/02Three survey rounds (birth, 5/6 yrs and 7/8 yrs)Extremely low attrition rates (7% in Peru)Limitations
7 Analytic Approach Models Challenge: Omitted variable bias and endogeniety leads to biased estimatesSeveral approaches available to address thisLatent Trait Modeling to recalculate the wealth indexModels1) OLS with Community Fixed Effects2) Spline with Community Fixed Effects3) Child Level Fixed Effects (First-Differences)
8 Control Variables Child specific: Household level: gender race/ethnicitybirth weightchronic health problemsstuntingagesubjective socio-economic status (at age 7/8 yrs)Household level:siblingscaregivers level of educationproxies for household non-cognitive environment:maternal depressioncaregiver’s educational aspirations of childnegative child rearing experiencenetwork/social capitalmain language spoken at home
9 Model 1: OLS with Community Fixed Effects 𝑆 𝑖𝑗,𝑡 𝐶 = 𝑎 0 + 𝑛=0 2 𝛽 𝑛 log(𝑌) 𝑖𝑗,𝑡−𝑛 + 𝑛=0 2 𝜏 𝑛 W 𝑖𝑗,𝑡−𝑛 + 𝑘=1 𝐾 𝜆 𝑘 𝑥 𝑖𝑗𝑡 𝑘 + 𝛼 𝑗 + 𝑙=1 𝐿 𝛾 𝑙 𝑦 𝑖𝑗𝑡 𝑙 + 𝜇 𝑖𝑗,𝑡 1Where:𝑆 𝑖𝑗,𝑡 𝐶 is the cognitive outcome observed in Round 𝑡 for child 𝑖 in community 𝑗log(𝑌) 𝑖𝑗,𝑡−𝑛 is expenditure (logged) in Round 𝑡−𝑛 (𝑛=0,2)W 𝑖𝑗,𝑡−𝑛 is household wealth𝑥 𝑖𝑗𝑡 𝑘 are observed child-level factors, where there are 𝐾 number of variables (𝑘=1,𝐾)𝛼 𝑗 represents the community fixed effects.𝑦 𝑖𝑗𝑡 𝑙 are the unobserved exogenous factors affecting child development, where L is the number of variables (𝑙=1, 𝐿).𝜇 𝑖𝑗𝑡 1 captures measurement error
10 Model 2: Spline with Community FE Are there non-linear effects?A prototypical model is represented as follows:𝑆 𝑖𝑗𝑡 𝐶 = 𝑎 0 + 𝑛=0 2 𝑠(log(𝑌) 𝑖𝑗,𝑡−𝑛 ) + 𝑛=0 2 𝑠( W 𝑖𝑗,𝑡−𝑛 ) + 𝑘=1 𝐾 𝜆 𝑘 𝑥 𝑖𝑗𝑡 𝑘 + 𝛼 𝑗 + 𝑙=1 𝐿 𝛾 𝑙 𝑦 𝑖𝑗𝑡 𝑙 + 𝜇 𝑖𝑗𝑡 1The only difference between Equation 2 and Equation 1 is the spline function around the household resources variables 𝑠(log(𝑌) 𝑖𝑗,𝑡−𝑛 ) and 𝑠( W 𝑖𝑗,𝑡−𝑛 )
11 Model 3: First-difference model Controls for differences between households and communitiesAllows to examine the effect of changes in parental resources on child development𝑆 𝑖,𝑡+1 𝐶 − 𝑆 𝑖𝑡 𝐶 = 𝑎 0 + 𝑛=0 2 𝛽 𝑛 log 𝑌 𝑖,𝑡+1−𝑛 − log 𝑌 𝑖,𝑡−𝑛 ) + 𝑛=0 2 𝜏 𝑛 W 𝑖,𝑡+1−𝑛 − W 𝑖,𝑡−𝑛 𝛽 3 log 𝑌 𝑖,0 + 𝜏 3 𝑊 𝑖,0 + 𝛽 4 𝑛=0 2 log 𝑌 𝑖,0 ∗ log 𝑌 𝑖,𝑡+1−𝑛 − log 𝑌 𝑖,𝑡−𝑛 𝜏 4 𝑛=0 2 𝑊 𝑖,0 ∗ W 𝑖,𝑡+1−𝑛 − W 𝑖,𝑡−𝑛 𝛽 5 𝑇 𝑖 + (𝜇 𝑖,𝑡+1 1 −𝜇 𝑖,𝑡 1 )𝑆 𝑖𝑡 𝐶 − 𝑆 𝑖,𝑡−1 𝐶 : changes in the developmental outcome between age 5/6 yrs and 7/8 yrs
12 Results: OLS with Community FE (1) Consumption appears to have a positive and contemporaneous effect. Wealth too, but also a lagged effect.
13 Results: OLS with Community FE (2) Wealth is a better predictor of children’s math and reading skills at 7/8 yrs old than consumption.
14 Results: Spline with Community FE (1) The effect of wealth on vocabulary scores at age 5/6 yrs is non-linear, whereas the effect of expenditure is linear.This is differs with math scores at 5/6 yrs, where the effect of wealth and expenditure in the same time period are broadly linear.
15 Results: Spline and Community FE (3) With maths and reading scores at 7/8 yrs, wealth in the previous period (5/6 yrs) appears to have a nonlinear effect.
16 Results: First-Differences (1) Changes in expenditure between 5/6 and 7/8 years appear to benefit the poorest children more than the better off children.VocabularyMath
17 SummaryConsumption appears to have a contemporaneous effect, while wealth has both a contemporaneous and lagged effect on vocabulary test scores.Data on parental wealth are a better predictor of children’s maths and reading skills at 7/8 yrs old than consumption data.The assumption that wealth is linearly associated with cognitive development does not hold.A change in wealth appears to have a lagged effect, while changes in expenditure an immediate effect on both math and vocabulary test scores.This effect appears to benefit poorer children more than better off children.
19 Mechanisms Goods inputs and under-nutrition Stunting - negative and significant effect generallyLong term health problems – no significant effectBirth weight – no significant effectTime inputs / under-stimulationCaregiver’s depression – positive with reading at age 7/8 yrsParental aspirations – positive and significance with vocabulary at age 7/8 yrsDemographicGender - significant differences in math and vocabulary at age 7/8 yrs, but smallSESSubjective SES – small, positive and significant with reading at age 7/8 yrs only
20 Results: First-Differences (2) Changes in wealth between birth and 5/6 yrs appears to benefit poorer children more than wealthier children in both maths and vocabulary scores.MathVocabulary