Does pre-school improve cognitive abilities among children with early-life growth faltering? A longitudinal

study for Peru

Santiago Cueto, Juan León, Alejandra Miranda (GRADE), Kirk Dearden (Boston University), Benjamin Crookston (Brigham Young University) & Jere Behrman

(University of Pennsylvania)

March 2015

March, 2015

Introduction

• Many empirical studies have shown negative associations between low height-for-age (HAZ), and school achievement.

• Empirical findings suggest that schooling (or other formal educational programs) and nutrition may have independent but also possibly interactive effects in promoting children.

• In Peru, stunting prevalence has gone down, currently about 18% of children under 5 years of age are stunted (UNICEF, 2011).

March, 2015

Types of pre-school in Peru

• For ages three to five years, there are two types of pre-school:• PRONOEI (Programas no Escolarizados de Educación Inicial) :

– provided in marginal urban and rural area– non certified teachers– 3 to 4 days a week of classes• Jardines:– Set up in more populated areas– Teachers are usually certified – 5 days a week of classes• Some research on these: Cueto & Diaz, 1999; Diaz, 2006

March, 2015

Objectives

• The present study seeks to determine whether there is an interaction between low HAZ at age one year and attending pre-school from ages three to five years on abilities by age five (PPVT, receptive vocabulary; CDA, test on notions of quantity).

• Further, we explore if the type of pre-school is associated with stunted children’s performance on these tests.

• Hypothesis: children who were stunted at one year of age and attended Jardines are more likely than children who were stunted at one year of age and attended PRONOEI to perform well on tests by age five.

March, 2015

Methods

• Young Lives is a longitudinal study which follows 12,000 children in Vietnam, India, Ethiopia and Peru over 15 years.

• Two cohorts: born around 1994 (older cohort) and 2001 (younger cohort). This study uses data from the YC in Peru.

• Household and children´s information from 2002, 2006, 2009 and 2013 (two first rounds included here).

• From the sample of 1963 childrenin R2, we only included the 72.5 percent who attended either a Jardin or a PRONOEI pre-school or no pre-school. Mixed cases (i.e. children who attended both) as well as children who attended other types of childcare programs from ages three to five years were excluded from the analysis.

March, 2015

Threats to validity

• Selection bias: we performed corrections for excluded children.• Endogeneity: we performed both OLS and Two Stage Least Squares (2SLS)

regression models (instrumental variables). • Good instruments predict well the endogenous variables (nutritional status,

number of years the child attended a Jardin and number of years the child attended a PRONOEI) in the model and should have no direct impact on the main dependent variable (cognitive abilities).

March, 2015

Instrumental variables

• For height-for-age z scoresMaternal height

• For years the child attended a Jardin pre-school Education of the household head (proportion of household heads that had completed at least secondary schooling in the district where the Young Lives child lived; data from National Census of 2007).

• For years the child attended a PRONOEI pre-school Ratio of PRONOEI pre-schools over total number of pre-schools in the district (from the School Census 2006 administered by the Ministry of Education).

March, 2015

Results

CDA and PPVT mean scores by nutritional status and years of pre-school attendance CDA PPVT Not-

stuntedStunted Difference

Not-stunted

Stunted Difference

0 years 7.79 7.48 0.31 21.13 15.09 6.04*(183) (125) (183) (125)

1 year 8.39 7.82 0.57* 28.44 21.45 6.99*(349) (130) (349) (130)

2 years 8.74 7.62 1.12* 33.01 21.39 11.62*(367) (105) (367) (105)

3 years 9.32 8.59 0.73 39.74 30.24 9.50*(135) (29) (135) (29)

*Differences are statistically significant at 5% according to the t test for independent samples.

Note: Number of children reported in parentheses.

Source: Young Lives Study, Rounds 1 and 2.Own Elaboration.

March, 2015

Results – MCO Model

Effect of attending pre-school and nutritional status on cognitive abilities using OLS, standardized coefficients (N=1,423)

CDA PPVTM1 M2 M1 M2

Main effects HAZ adjusted to age one 0.03 0.03 0.05 * 0.05 ** Years attended JARDIN 0.08 * 0.08 * 0.09 ** 0.08 ** Years attended PRONOEI 0.00 0.00 0.05 + 0.05 Interaction effects

Years attended JARDIN*HAZ adjusted to age one 0.01 0.05 *

Years attended PRONOEI*HAZ adjusted to age one 0.01 0.01 R-squared 0.20 0.20 0.51 0.52

**p<0.01, *p<0.05, +p<0.10

Note: Standard errors are adjusted by possible covariance among children living in the same district. All models include as control variables: child´s age, child´s mother tongue, child´s sex, maternal schooling attainment, mother´s age, place of residence, and household wealth index. Standard errors were calculated using bootstrapping with 100 replications.

March, 2015

Results – 2SLS Model

Effect of attending preschool and nutritional status on cognitive abilities using 2SLS, standardized coefficients (N=1,423)

CDA PPVTM1 M2 M1 M2

Main effects

HAZ adjusted by age one 0.17 ** 0.15 * 0.34 ** 0.28 ** Years attended JARDIN 0.21 ** 0.20 ** 0.33 ** 0.25 ** Years attended PRONOEI -0.04 -0.04 + 0.00 0.00

Interaction effects

Years attended JARDIN*HAZ adjusted by age one

0.11 0.46 *

Years attended PRONOEI*HAZ adjusted by age one

-0.13 -0.43 **

R-squared 0.18 0.18 0.44 0.45

**p<0.01, *p<0.05, +p<0.10 Note: Standard errors are adjusted by possible covariance among children living in the same district. All models include as control variables: child´s age, child´s mother tongue, child´s sex, maternal education, mother´s age, place of residence, and wealth index. Standard errors were calculated using bootstrapping with 100 replicates.

March, 2015

Discussion

• Preschool may be crucial for achievement, but we need to know more about their quality.

• Targeting children at risk of poor achievement (e.g. stunted) could be a priority to consider.

• Comprehensive interventions (i.e. health and nutrition, plus work with their parents) might also be a priority, even from an early age.

• What are the interactive effects on other aspects of children´s development (e.g. motor, social, emotional)?

• What interactions may we find between stunting at an early age, catching up and quality of primary school?

March, 2015

Results

Correlation among cognitive measures and main independent variables (N=1,423)

PPVT score CDA score

HAZ adjustedto age one

CDA score0.56 - -

(0.00)

HAZ adjusted to age one0.32 0.19 -

(0.00) (0.00)

Years attending a JARDIN0.37 0.25 0.22

(0.00) (0.00) (0.00)

Years attending a PRONOEI-0.11 -0.10 -0.06(0.00) (0.00) (0.02)

Note: P-values in parenthesis

Source: Young Lives Study, Rounds 2 and 3

Own Elaboration.