PLOS ONE

RESEARCH ARTICLE

Stunting and its association with education

and cognitive outcomes in adulthood: A
longitudinal study in Indonesia
Esta Lestari ID1,2*, Adiatma Siregar1,3, Achmad K. Hidayat1,3, Arief A. Yusuf1,4

1 Doctorate in Economics Program, Department of Economics, Faculty of Economics and Business,

Universitas Padjadjaran, Bandung, West Java, Indonesia, 2 Research Center for Behavioral and Circular
Economics, National Research and Innovation Agency (BRIN), Jakarta, Indonesia, 3 Center for Economics
and Development Studies, Department of Economics, Faculty of Economics and Business, Universitas
a1111111111 Padjadjaran, Bandung, West Java, Indonesia, 4 SDGs Center, Universitas Padjajaran, Bandung, West Java,
a1111111111 Indonesia
a1111111111
a1111111111 * [email protected], [email protected]
a1111111111

Abstract

OPEN ACCESS

Citation: Lestari E, Siregar A, Hidayat AK, Yusuf AA

Background
(2024) Stunting and its association with education Stunting is associated with adverse outcomes in adulthood. This article specifically aims to
and cognitive outcomes in adulthood: A analyse the relationship between childhood stunting and education as well as cognitive out-
longitudinal study in Indonesia. PLoS ONE 19(5):
e0295380. https://doi.org/10.1371/journal.
comes for adults in Indonesia.
pone.0295380

Editor: Susan Horton, University of Waterloo, Methods

CANADA
Pooled data from wave one (1) and two (2) of the Indonesia Family Life Survey (IFLS) in
Received: October 24, 2022
1993 and 1997 identified a sub-sample of 4,379 children aged 0–5 by their height-for-age
Accepted: November 21, 2023 (HAZ) to be compared for their differences in educational outcomes and cognitive abilities in
Published: May 6, 2024 2014. HAZ was used to proxy relative height to determine stunting status based on 2006
Copyright: © 2024 Lestari et al. This is an open WHO child’s growth standards. Education and cognitive abilities outcomes include years of
access article distributed under the terms of the schooling, age of school entry, grade repetition, and scores for cognitive and math tests.
Creative Commons Attribution License, which The study employs estimation models of pooled regressions and instrumental variable (IV)
permits unrestricted use, distribution, and
to address problems of endogeneity and bias from omitted variables.
reproduction in any medium, provided the original
author and source are credited.

Data Availability Statement: All relevant data are Results

within the manuscript and its Supporting
Information files. The database is freely accessible
Stunting and relatively small stature had significant associations with cognitive develop-
only to those researchers who have registered their ment, and they worked as intermediaries to cognitive developmental barriers as manifested
interest on the RAND website from http://www. in reduced educational outcomes. A lack of one SD in HAZ was associated with 0.6 years
rand.org/labor/FLS/IFLS/access.html.
shortened length of the school, 3% higher chances of dropouts from secondary school, and
Funding: This research was conducted as part of 0.10–0.23 SD lowered cognitive and numerical scores. Similarly, stunting is associated with
Esta Lestari’s PhD studies, supported by a
decrease cognitive test scores by 0.56–0.8 SD compared to non-stunting, two years less
scholarship from the Ministry of Research and
Higher Education - National Research and schooling, and 0.4 years of delayed entry to school. As for cognitive abilities, stunting is
Innovation Agency (BRIN). Neither institution associated with lower cognitive and numerical abilities by 0.38–0.82 z-scores.

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 1 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

played a role in the study design, data collection Conclusion

and analysis, decision to publish, or preparation of
the manuscript. Growth retardation during childhood in Indonesia was associated with lower cognitive abili-
ties, particularly during school age, and this correlation faded as individuals grew up. Subse-
Competing interests: Authors have declared that
no competing interest exist. quently, growth retardation is significantly linked to lower educational outcomes. Impaired
growth has implications for reduced lifetime earnings potential mediated by diminished cog-
Abbreviations: IFLS1-5, Indonesia Family Life
Survey waves 1 to; HAZ, Height-for-age z-score; nitive capacity and lower educational attainment. The finding suggests that development in
WHO, World Health Organization; PCA, Principal Indonesia during recent decades has not provided an adequate environment to enable chil-
component analysis; IVs, Instrumental variables; dren to achieve their potential educational outcomes.
OLS, Ordinary least squares; SD, Standard
deviation; 2SLS, two stages least square.

Introduction
Stunting is one of six forms of malnutrition prioritised to be eradicated by 2025 [1]. That hap-
pened to more than 149.2 million children under five in 2020, whereas Asia and Africa shared
the largest burden at 54% and 40%, respectively [2]. Despite being a middle-income country
and a member of the G20, Indonesia continues to struggle with malnutrition, with 30.8% of
children under five experiencing stunted growth in 2018 [3]. According to the World Health
Organization (WHO), Indonesia was among the countries with the highest rates of stunting in
the world, with a rate slightly better than Cambodia (32.4% in 2014) and Lao PR (33.1% in
2017) in the region [2].
As a marker of long-term chronic malnutrition, stunting has numerous adverse conse-
quences, including impaired physical and cognitive development, low educational achieve-
ment [4–9], a reduction in lifetime income [10], an increased risk of non-communicable
diseases and poor birth outcomes for the future generation [11]; leads to decrease chances of
escaping poverty [12]. Therefore, addressing stunting is crucial for improving health outcomes
and building a supportive socio-economic environment that allows children to reach their full
potential [13, 14].
Specifically, empirical studies showed that growth retardation or stunting has been associ-
ated with adverse educational outcomes such as shorter years of schooling, lower cognitive
abilities, delayed school entry, and a higher risk of failing grades in countries such as Brazil,
Guatemala, India, South Africa, the Philippines, Vietnam and Ethiopia [15–18]. Grantham-
McGregor et al. (2007) found that stunting at 24 months of age was linked to a 0.9 years delay
in school entry and a 16% increased risk of failing grades [19]. Another study by Hoddinott
et al. (2013) reported that an increase in height-for-age z-score (HAZ) was related to more
extended schooling, higher test scores for reading, and nonverbal cognitive abilities [20].
From the mid-1990s until the end of the New Order regime in 1990, Indonesia was one of
the most rapidly growing Asian countries. However, in the following two decades, the high
economic growth was not accompanied by sufficient human capital investment. The World
Bank (2020) estimated that Indonesia’s productive labour would only reach 53% of its poten-
tial, partly due to poor childhood nutritional status [21]. This indicates that the high economic
growth experienced by Indonesia until recently may not necessarily be accompanied by
improved human capital quality.
To the best of our knowledge, studies on the impact of stunting in Indonesia are still rela-
tively limited, particularly with regard to educational outcomes, despite a relatively high num-
ber of studies on determinants and interventions of stunting [22–29]. The challenges come
from the requirement for reliable longitudinal data capable of capturing the developmental

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 2 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

trajectory of children into adulthood. Another challenge in studies in this area is the estimation
techniques that can minimise measurement errors and endogeneity commonly found in
health and education studies. Referring to Behrman (1997) [30], one of the most common fail-
ures is the exclusion of the possibility that health might be endogenous, which leads to mea-
surement problems and potentially biased estimates of the relationship between the child’s
health and education. The bias may arise primarily from unobserved household or community
characteristics omitted in the estimations.
This study aims to fill the gap in Indonesian literature on the consequences of growth fail-
ure and provide a robust estimation of the linkage between chronic childhood malnutrition
(stunting) and adults’ educational outcomes. By applying instrumental variables, this study
demonstrates a reliable estimate that controls bias from heterogeneity in households and com-
munities and endogeneity problems between childhood health measurement and academic
outcomes. The goal is to identify potential interventions that may improve educational out-
comes for stunted individuals and quantify the strength of the relationship between stunting
and academic outcomes.

Methods
This study utilises The Indonesian Family Life Survey (IFLS) data, a longitudinal data spanned
over 21 years from 1993 to 2014 covering a representative sample of 83% of the population
and over 30,000 individuals across 13 provinces in Indonesia. The survey collected data on
individual respondents, their families, their households, and the communities in which they
live on various aspects of social, economic, and health issues, including educational achieve-
ments and cognitive abilities Specifically, this study employs data from waves 1–2 (1993 and
1997) [31, 32] as the baseline and waves 3–5 (covering the years beyond 1997 up to 2014) for
the outcome variables [33–35]. Data from the IFLS surveys are publicly available for those who
have registered their interest on the RAND Corporation website [36].

Study population
The sampling frame for this study consists of combined data for children aged 0–5 years from
IFLS wave 1 (1993) and wave 2 (1997) who had complete records of height, weight and age.
The follow-up data encompassed multiple waves and allowed us to examine the outcome of
interest in 2000, 2007, and 2014 (wave 3–5). Pooled data identified sub-sample of 4,379 chil-
dren below five years old to be estimated.

Variables characterisation
Characterisation of stunting and relative height. We calculated the child’s height-for-
age (HAZ) as a z-score using the age and sex-specific references from the WHO growth stan-
dard based on their height for children <5 years [37]. Children with a height-for-age-z-score
of <-2 were categorised as stunted [38] and used as the reference for dummy variables of status
for stunting. For this purpose, one was assigned as the value when the HAZ was <-2, or stunt-
ing, and zero was assigned when the HAZ was >-2, or not stunting. Children above two years
old were measured in a standing position, and those below the age were measured in lying
down otherwise, there was a 0.7cm adjustment following the WHO measurement standards
[39].
The dataset is pooled data of height-for-age collected in 1993 and 1997 for children under
the age of five with a total sample of 5,224. Some of the children initially identified in wave 1
were still below the age of 5 in wave 2, two years later. To avoid counting them twice, we
excluded them in the sample (n = 349). By this, we have ensured that the count reflects unique

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 3 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Fig 1. The distribution of HAZ data for the original sample.

https://doi.org/10.1371/journal.pone.0295380.g001

children, and they were not double-counted, resulting in 4,875 children. The original dataset
distribution can be seen in Fig 1.
However, due to significant measurement errors associated with height mismeasurement,
we excluded 496 children from the sample. This outlier specifically comes from the inconsis-
tencies in height measurements relative to the subject’s ages. Following Alderman’s study
(2006), we utilized samples with HAZ values falling within the range of -6 to +6 to ensure data
quality. The children were then tracked down until they were aged 17–26 years in 2014, and
there remained a total of 4,379 respondents. The distribution of adjusted sample is outlined in
Fig 2.
The follow-up surveys in Waves 3 (2000), 4 (2007), and 5 (2014) aimed to assess the educa-
tional and cognitive outcomes of the participating children. Child samples of 4,379 may not
always be included in the last three waves, primarily because some outcome variables require
specific age criteria or they are no longer traceable in a particular wave. However, excluded
children in a wave can be included again into the estimates if they become traceable and their
outcome data is found in subsequent waves.
Using the methods described in Fitzgerald, Gottschalk, and Moffitt (1998) [40] and Alder-
man et al. (2001) [41], we estimated a probit to determine if there was attrition based on
observable characteristics, as presented in Table 1. As part of this, a dependent variable equal
to 1 if the school achievement is observed in 2004 and 0 otherwise is regressed along with
height-for-age and a variety of child and family characteristics. Because there was no statisti-
cally significant link between height-for-age and attrition for all outcomes except for adults’
cognitive abilities thus, there is not enough evidence of attrition bias.
Characterisation of educational outcomes. In the 2014 IFLS, educational outcomes are
measured by several variables: years of schooling, grade repetition, binary variable for school

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 4 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Fig 2. The distribution of HAZ data for the original sample.

https://doi.org/10.1371/journal.pone.0295380.g002

dropout, and age of first enrolment. The variable "years of schooling" represents the total num-
ber of completed years of education, with a minimum value of 0 for individuals who have never
attended school or did not complete primary education and a maximum value of 22 years for
those who have completed college or university. The school dropout variable is a binary variable
set to one if an individual only completed elementary school and did not continue to secondary
school and zero otherwise. The grade repetition is also a binary variable, with a value of one if
the individual had any experience of grade repetition until 2014. Finally, the age of the first
enrolment to school is the age at which a child was first admitted to elementary school.
IFLS provides a cognitive capacity section to measure the level of intellectual development
using Raven’s Progressive Colored Matrices (RPM) method and mathematics test.
The Raven test is considered a valid measure of cognitive ability due to its strong theoretical
foundation, robust psychometric properties, and demonstrated correlations with other intelli-
gence measures, making it suitable for diverse populations [42]. The RPM has been widely
used as a cognitive ability indicator for studies conducted in Indonesia [26, 43, 44], as well as
in various other countries, including [45]. RPM measures fluid intelligence from non-verbal
cognitive scores and mathematical tests to measure numerical abilities. The levels of tests given
to the respondents were divided into an easier version for all respondents aged 7–14 and a
more complex version for all respondents aged 15–24 years old. The cognitive and mathemat-
ics scores are measured based on the number of correct responses to a set of questions, which
are subsequently standardised to obtain the final scores. The scores were retrieved from the
2000, 2007, and 2014 waves to represent the school-age, adolescence, and adulthood phases.
Potentially confounding variables. We controlled for individual, parental, and house-
hold variables in the base years (1993 and 1997). Individual variables consisted of gender and
age in 2014. Parental characteristics are represented by their age in 1993/1997, and a dummy

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 5 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Table 1. Testing for selective attrition using Fitzgerald, Gottschalk, and Moffitt method.
Outcomes measured in 2014
Exposures Childhood cognitive z- Adolesc. cognitive z- Adult. cognitive z- Age started school Grade Dropout Years of school
score score score (years) retention school (years)
(1) (2) (3) (4) (5) (6) (7)
Initial HAZ 0.02 0.01 -0.04*** -0.01 -0.01 -0.01 -0.00
(0.02) (0.02) (0.01) (0.01) (0.01) (0.04) (0.01)
Child is boy -0.14** -0.05 -0.25*** -0.19*** -0.10** -0.17 -0.18***
(0.06) (0.07) (0.05) (0.05) (0.04) (0.13) (0.05)
Household size -0.04*** -0.00 -0.01*** -0.01 -0.00 0.00 -0.00
(0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.01)
Asset index 0.18*** 0.08** 0.11*** 0.04 0.12*** 0.08 0.04
(0.041) (0.04) (0.03) (0.03) (0.03) (0.09) (0.03)
Rural/Urban -0.04 0.02 0.07 0.11** 0.21*** 0.15 0.10**
(0.07) (0.06) (0.05) (0.05) (0.05) (0.13) (0.05)
Child from Java 0.11 0.03 -0.05 0.00 -0.10* -0.31* 0.00
(0.08) (0.07) (0.06) (0.06) (0.06) (0.18) (0.06)
Child from -0.08 -0.06 -0.02 -0.07 -0.07 -0.23 -0.08
Sumatra
(0.09) (0.08) (0.07) (0.07) (0.06) (0.21) (0.07)
Maternal educ. -0.12* -0.21*** -0.11** -0.06 0.13*** 0.24 -0.08
Level
(0.07) (0.06) (0.05) (0.05) (0.05) (0.18) (0.05)
Recurrent age 0.62*** -0.10*** -0.25*** -0.05*** -0.17*** -0.01 -0.05***
(0.02) (0.01) (0.01 (0.01) (0.01) (0.03) (0.01)
Constant -12.88*** 3.69*** 5.90*** 1.93*** 3.81*** 2.46*** 2.02***
(0.43) (0.33) (0.27) (0.26) (0.26) (0.7) (0.26)

Standard errors in parentheses

*** p<0.01
** p<0.05
* p<0.1

https://doi.org/10.1371/journal.pone.0295380.t001

variable is whether the mother was working in the base year. The household characteristics
consist of the household size and households’ welfare index in the base year. The welfare index
is the assets-based indicator according to the households’ assets as the representative figure of
the socio-economic status in the community [46] and is divided into three terciles (poorest,
middle, and richest).
Meanwhile, household infrastructure consists of binary electricity, safe drinking water, and
sanitation variables. Fixed effects on regional variations are captured by dummy variables of
rural/urban and major islands: Java, Sumatra, Borneo (Kalimantan), Sulawesi and Nusa Teng-
gara. The last three regions are combined into one variable due to their relatively smaller num-
ber of respondents than Java and Sumatra. These three categories also represent the
population density in Indonesia between the western and eastern parts of Indonesia to repre-
sent regional disparity in development.

Statistical analyses
To determine the association between the potential effect of stunting and relative height at the
individual level, multivariate regression (OLS) and instrumental variables estimates were

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 6 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

applied. Multivariate least squares were conducted for outcomes with continuous values such
as years of schooling, standardised cognitive and math scores, and age-started schooling.
Meanwhile, linear probit regressions addressed discrete grade retention and dropout school
outcomes. The basic model is as follows:

Xp
Yi ¼ b0 þ b1 Si þ m¼14
bm Xi þ εi ð1Þ

In which the dependent variable Yi consists of the set educational outcomes, β1 is the main
effect of nutritional status divided into relative height (height-for-age z-score) and the dummy
for stunting, and Xi is the covariate vector consisting of characteristics of children, parents,
families and regions.
One problem with using OLS regression is the possible endogeneity of nutritional status
(proxied by HAZ and stunting status) with residuals confounded by other unobserved factors
correlating with stunting and outcomes [20, 47]. Applying OLS estimation would likely pro-
duce omitted variable bias, for example, if unobserved variables (such as parenting skills and
parenting time spent) may positively affect early-life nutrition, cognitive skills, and educational
outcomes. If there was the case, thus in our model, the unobserved variables for the cognitive
abilities and education outcomes would be included in the error term and, with the assump-
tions specified in the previous sentence, the error term would be positively correlated with
nutritional status and other exogenous variables in our model. As a result, the estimated
parameters tend to be upward biased [48]. Another approach that leaves the unobserved vari-
able in the error term is using instrumental variable as an estimation method that recognizes
the presence of the omitted variable and treats the HAZ and stunting as endogenous. The
instrumental variables of nutritional status needs to satisfy two conditions: (1) it should have
no partial effect on educational and cognitive outcomes, and it should not be correlated with
other factors that affect outcomes. (2) It must be related, either positively or negatively, to the
endogenous explanatory variable (relative height and stunting) [49].
Selection of instrumental variables. Several instrumental variables (IVs) are commonly
used to study the relationship between childhood health and adults outcomes, including the
mother’s height [48], the status of being twins [20], environmental variables such as rainfall
and vegetation [50], specific randomised interventions [20], or regional variables such as food
prices and access to health facilities [51]. In Indonesia, no targeted nutritional interventions
were aimed at addressing specific malnutrition conditions prior to 1993. The 1990s were also
marked by the New Order regime, during which the production and prices of staple foods
were controlled by the central government to achieve food self-sufficiency. Therefore, fluctua-
tions in food prices and nutritional interventions could not be considered potential instru-
mental variables. Additionally, there is a lack of data on the status of twins in IFLS and no
information available on the environment over the past two decades.
In this study, the instrumental variables employed was the birth month dummy variable,
divided into August to January, and February to July. The rationale behind using both of these
months was grounded in the consideration of two distinct seasons (rainy and dry) and the
planting cycle prevalent in Indonesia. August to January is commonly a rainy season and
planting period, while in February is the start of dry season. Maccini and Yang (2009) [52] cor-
relate seasonal factors, specifically rainfall, and a range of individual outcomes, such as health,
education, and asset indices. Their findings indicated a correlation between rainfall and health,
with higher early-life rainfall having substantial positive effects on adult outcomes for women
but not for men. These observed patterns are most plausibly attributed to the favorable impact
of rainfall on agricultural productivity, resulting in increased household incomes, enhanced

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 7 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

food availability, and improved health for infant girls. Building upon this concept, this study
adopts the birth month as an instrumental variable for assessing children’s nutritional status.
Specifically, children born between August and January are assigned a value of one (1),
while those born between February and July are assigned a value of zero (0). This variable was
assumed to be correlated with nutritional status and may be exogeneous with the educational
outcomes and therefore meet the conditions as a appropriate instrument. Furthermore, in
addition to the birth season, mother heights were also treated as instruments considering the
variable brings genetic variation in an individual’s height. Although parental heights were pos-
itively correlated with the child’s nutritional status and might also be associated with the child’s
cognitive abilities, this relationship is often indirect [48, 53].

Ethics approval and consent to participate

This study is based on a survey conducted by the Rand Corporation, which has been designed
and executed in accordance with the principles of ethical research. The survey has obtained
ethical clearance from Institutional Review Boards (IRBs) in the United States and Indonesia
at the University of Indonesia (IFLS 1 and 2) and the University of Gadjah Mada (IFLS 3, 4,
and 5). The protocol approval number (i.e., ethical clearance number) that RAND’s Human
Subjects Protection Committee (RAND’s IRB) gave IFLS5 was s0064-06-01-CR01. All neces-
sary consent requirements for participants, both adults and children, were fulfilled and autho-
rised by the IRBs before the start of data collection. All participants provided their written
consent to take part in the survey.
Additionally, consent was obtained from the legal representatives, such as next of kin, care-
takers, or guardians, of any children included in the survey on their behalf. This ethical clear-
ance and participant consent process ensures that the data collected is valid and reliable and
that the rights and well-being of the participants are protected throughout the survey. Data in
this study is de-identified by unique ID numbers, enabling longitudinal tracking while ensur-
ing security and confidentiality.

Results
Table 2 reports the summary statistics of all variables of interest. Approximately 47% of chil-
dren aged 0–60 months were classified as stunted based on their height-for-age (HAZ), with
an average HAZ of –3.08 (nearly severe stunting). The average height for children in this age
group is 81.3cm, with a difference of over four cm between the stunted and non-stunted
groups (85.95cm). The differences in children’s characteristics between the stunting and non-
stunting group are statistically significant.
Stunted children are more likely to live in rural areas (67%) and in households with less
adequate sanitation (73%). Regarding the economic background, families with stunted chil-
dren tend to have lower per capita expenditure (IDR 62,000) than those without stunting (IDR
99,000). Almost all measured parental characteristics between the stunted and non-stunted
groups, such as parental age and height, maternal employment status, and maternal education,
also showed statistically significant differences. Household size does not show a statistically
significant difference between the two groups.
When reaching adulthood, individuals with a history of stunting are also found to exhibit
statistically significant differences in their academic achievements. These individuals tend to
have lower cognitive and math test scores during various stages of growth, including school
age, adolescence, and adulthood. Additionally, they are inclined to achieve shorter educational
attainment, being placed in higher grade levels and having a shorter duration of schooling
(10.7 years of education compared to 11.12 years for those without childhood stunting).

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 8 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Table 2. Descriptive statistics for children’s characteristics and outcomes.

ALL samples Stunted below five y.o Non-stunted below five Diff.
y.o
Obs. Mean Conf. Obs. Mean Conf. Obs. Mean Conf.
interval interval interval
CHILD LEVEL FACTORS
Sex = 1 if male 4,379 0.51 (0.50–0.53) 2,039 0.54 (0.51–0.56) 2,340 0.49 (0.47–0.51) -
2.71***
Age Child’s age in 1993 or 1997 (years) 4,379 2.26 (2.21–2.30) 2,039 2.44 (2.38–2.50) 2,340 2.1 (2.04–2.16) -7.88***
Height Height in 1993 or 1997 (cm) 4,379 84.08 (83.7–84.5) 2,039 81.93 (81.47– 2,340 85.95 (85.37– 10.51***
82.39) 86.53)
HAZa Height-for-age z score at age 0–5 years in 4,379 -1.75 (-1.80 2,039 -3.08 (-3.12 2,340 -0.59 (-0.65 74.35***
1993 or 1997 –-1.70) –-3.04) –-0.54)
Stuntinga = 1 if HAZ at age 0–5 <-2 in 1993 or 1997 4,379 0.47 (0.45–0.48) 2,039 - - 2,340 - -
PARENTAL FACTORS
Maternal education Mother’s categorical value of the highest 4,209 0.2 (1.22–1.25) 1,964 0.15 (1.14–1.18) 2,245 0.25 (1.27–1.32) 9.34***
education level completed in 1993 or 1997
Paternal age Father’s age in 1993 or 1997 (year) 3,829 34.97 (34.72– 1,804 35.36 (34.97– 2,025 34.63 (34.29– -2.81***
35.23) 35.74) 34.97)
Maternal age Mother’s age in 1003 or 1997 (year) 4,212 29.7 (29.51– 1,967 29.98 (29.69– 2,245 29.46 (29.20– -2.63***
29.90) 30.27) 29.72)
Paternal height Father’s height in 1993 or 1997 (cm) 3,047 161.26 (161.05– 1,437 160.18 (159.89– 1,610 162.22 (161.93– 9.74***
161.47) 160.47) 162.51)
Maternal height Mother’s height in 1993 or 1997 (cm) 3,797 150.09 (149.92– 1,789 148.96 (148.72– 2,008 151.08 (150.86– 12.49***
150.25) 149.21) 151.31)
Maternal working = 1 if the mother worked in the base year 3,916 0.36 (0.34–0.37) 1,847 2.59 (0.35–0.39) 2,069 2.38 (0.32–0.36) -1.90**
HOUSEHOLD FACTORS
Number of # 3,907 7.78 (7.63–7.93) 1,801 7.81 (7.60–8.03) 2,106 7.75 (7.55–7.96) -0.4
household members
Per capita Household per capita expenditure in 1993 4,371 82.29 (72.9–91.7) 2,037 62.64 (59.6–65.7) 2,334 99.44 (82.1– 3.83***
expenditure or 1997 (000 IDR) 116.8)
Electricity = 1 if the household utilise electricity in 4,376 0.73 (0.72–0.75) 2,038 0.68 (0.66–0.70) 2,338 0.78 (0.76–0.79) 7.10***
1993 or 1997
Safe drinking water = 1 if the household had safe source of 4,364 0.93 (0.92–0.94) 2,034 0.91 (0.90–0.92) 2,330 0.95 (0.94–0.96) 5.49***
drinking in 1993 or 1997
Sanitation = 1 if the household has safe defecation 4,257 0.34 (0.32–0.35) 1,979 0.27 (0.25–0.29) 2,278 0.39 (0.37–0.41) 8.04***
1993 or 1997
Wealth index Household’s wealth index in 1993 or 1997: 4,379 1.67 (1.64–1.69) 2,039 1.63 (1.60–1.66) 2,340 1.7 (1.67–1.73) 2.96***
Poorest = 1; Middle = 2, Richest = 3
Rural/Urban = 1 if urban 4,377 0.4 (0.39–0.42) 2,038 0.33 (0.31–0.35) 2,339 0.47 (0.45–0.49) 8.99***
EDUCATIONAL OUTCOMES
Childhood Raven Raven test score on the 1–100 scale in 2000 2,323 61.7 (60.70– 1,086 58.69 (57.22– 1,237 64.35 (62.98– 5.55***
test score (wave 3) 62.71) 60.15) 65.71)
Childhood Numerical test score on the 1–100 scale in 2,147 58.02 (56.99– 997 56.81 (55.33– 1,150 59.06 (57.63– 2.14**
numerical score 2000 (wave 3) 59.04) 58.29) 60.49)
Adolescence Raven Raven test score on the 1–100 scale in 2007 3,893 78.22 (77.60– 1,808 76.93 (76.00– 2,085 79.34 (78.50– 3.77***
test score (wave 3) 78.85) 77.86) 80.19)
Adolescence Numerical test score on the 1–100 scale in 3,787 57.05 (56.18– 1,762 55.13 (53.86– 2,025 58.72 (57.52– 4.03***
numerical score 2007 (wave 3) 57.92) 56.40) 59.91)
Adulthood Raven Raven’s test score on the 1–100 scale in 3,194 74.73 (73.98– 1,488 73.04 (71.92– 1,706 76.19 (75.19– 4.13***
test score 2014 (wave 5) 75.47) 74.16) 77.19)
Adulthood Numerical test score on the 1–100 scale in 3,509 38.26 (37.25– 1,618 35.53 (34.15– 1,891 40.59 (39.16– 4.97***
numerical score 2014 (wave 5) 39.26) 36.90) 42.02)
Failing grades = 1 if ever failing grades 2,907 0.16 (0.15–0.17) 1,331 0.17 (0.15–0.19) 1,576 0.15 (0.13–0.16) -1.99**
(Continued )

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 9 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Table 2. (Continued)

Age started school The age first admitted to elementary school 3,304 6.36 (6.33–6.38) 1,534 6.41 (6.37–6.44) 1,770 6.31 (6.28–6.34) -4.04***
(years)
Drop out = 1 if never proceeded to secondary school 4,329 0.13 (0.12–0.14) 2,017 0.15 (0.13–0.16) 2,312 0.11 (0.10–0.12) -3.92***
Years of schooling Years of schooling in 2014 3,362 10.92 (10.82– 1,553 10.69 (10.54– 1,809 11.12 (10.98– 4.03***
11.03) 10.85) 11.26)

95% CIs in parentheses.

a
Based upon the 2006 WHO Child Growth Standards for children <5 years [37].
*** indicates statistically significant at the 1% level, and
** indicates statistical significance at the 5% level.

https://doi.org/10.1371/journal.pone.0295380.t002

Relative height and educational outcomes

Table 3 reported the estimation results of the relationship between relative height (HAZ) and
educational outcomes. The sample size for the estimated outcomes in both OLS and IV regres-
sions varies (S1 Table). The grade repetition and dropout school estimations are performed in
linear probit models. Overall, the OLS estimates (Table 3: 1–3) show that relative height
(HAZ) is significantly associated with higher educational achievements in various life-course
from school to adulthood. Holding other variables constant, one additional z-score would
likely increase cognitive (Raven’s) test scores in school-age (0.04-SD) and adolescence
(0.03-SD), yet it has no relationship with adults’ cognitive abilities. In contrast, relative height
influences individuals’ math ability during adolescence (0.02-SD) and adulthood (0.06), with-
out a relationship during school age.
Concerning educational achievement, a one-SD in HAZ in childhood is also significantly
associated with longer schooling years (0.11 years) and a later age of first enrolment in school

Table 3. Estimation results for relative height and educational outcomes.

Relative height (HAZ)1 when measured prior to age five2
OLSa Instrumental variables estimates
Educational outcomes Coef. ci P Coef. ci P
(1) (2) (3) (4) (5) (6)
Childhood’s Raven (Z-scores) 0.04 (0.01–0.08) 0.01 0.22 (0.06–0.40) 0.01
Childhood’s Numerical (Z-scores) 0.01 (-0.02–0.04) 0.35 0.11 (-0.07–0.28) 0.09
Adolescence’s Raven (Z-scores) 0.03 (0.01–0.05) 0.00 0.10 (0.00–0.20) 0.05
Adolescent’s Numerical (Z-scores) 0.02 (0.00–0.04) 0.08 0.12 (0.01–0.24) 0.03
Adult’s Raven (Z-scores) 0.02 (-0.01–0.04) 0.25 0.02 (-0.13–0.17) 0.81
Adult’s Numerical (Z-scores) 0.06 (0.03–0.08) 0.00 0.17 (0.02–0.32) 0.03
Age started school (years) -0.02 (-0.04–0.01) 0.01 -0.12 (-0.22–0.02) 0.02
Repeated grades (pp) -0.03 (-0.13–0.02) 0.17 -0.03 (-0.08–0.03) 0.35
Dropout (pp) -0.05 (-0.09–0.00) 0.03 -0.03 (-0.07–0.01) 0.09
Years of schooling (years) 0.11 (0.04–0.17) 0.00 0.59 (0.18–1.00) 0.00

All values are marginal effects; 95% of CIs are in parentheses.

1
Height-for-age according to WHO Standard Growth Reference for School-aged Children and Adolescents [37]. Control variables included but not reported are sex,
age, parental age, mother working, household size, household’s wealth index, household’s infrastructure (electricity, drinking water, and sanitation) and region (rural/
urban areas and fixed effect for main island). We have reported the results using a linear probability model for dichotomous outcomes (repeated education and dropout
school). The coefficient x 100 was the marginal effect in percentage points. HAZ: height-for-age z score; IV: instrumental variable; OLS: ordinary least squares.
2
The details regarding the sample size can be found in S1 Table.

https://doi.org/10.1371/journal.pone.0295380.t003

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 10 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

(0.02 years). Moreover, based on linear probit model, being relatively shorter by 1-SD might
be related to a higher probability of dropout secondary school (5%), while the relationship
between the HAZ and grade repetition was not significant.
The IV result for relative height (Table 3: 4–6) suggests that being relatively shorter is signif-
icantly associated with lower educational achievements. Specifically, a one SD decrease in
height is associated with a 3% increase in the probability of dropping out of school, a reduction
of approximately 0.6 years in schooling, and a 0.12-year delay in the age of the first enrolment
in school. For cognitive skills, a one-SD increase in HAZ is related to an increase of 0.23 and
0.13 z-score points in cognitive and numerical scores in childhood, and the influence persists
and declines until adolescence. In contrast, in line with the OLS results, relative height influ-
ences numerical abilities more than cognitive ones during adulthood.
We measure two test statistics assessing the strength of the instruments: the Kleibergen-
Paap Lagrange Multiplier (LM) and Kleibergen-Paap F-tests of weak devices. The Kleiber-
gen-Paap LM-tests the null hypothesis that the excluded instruments are correlated with the
endogenous variable, and The Kleibergen-Paap F-test examines a different null hypothesis
relating to weak devices, where weak means having bias relative to the bias in the OLS esti-
mates [20]. The values of the LM-test show that we reject the null hypothesis that the
excluded instruments are not correlated with the endogenous variable at the P<0.001 level.
Similarly, based on the tabulations found in Stock and Yogo (2005) [54], the critical value
for the Kleibergen-Paap F-test statistic at the 5% significance level is 19.93 for rejecting the
null hypothesis of weak instruments, when weak is defined as having a bias in the IV results
that is larger than 10% of the bias in the OLS results. Accordingly, we conclude that our
instruments have strong explanatory power (S2 Table). In addition, S2 Table reports the P-
values for the Hansen J-statistic for overidentification, where the null hypothesis is that the
overidentification constraint is valid, meaning that the model is well-defined. The instru-
ment is not included in the second-stage equation. Failure to reject the null hypothesis for
Hansen’s test suggests that all instruments are valid. The instrument set includes birth sea-
son and mother height, which gives us some confidence in the power of this specific test. In
all cases, we failed to reject null at P<0.05.

Stunting and educational outcomes

The results for the relationship between stunting and educational achievement are shown in
Table 4. As an indicator of chronic malnutrition, the relationship between stunting and educa-
tional attainment increases to nearly three folds the effect of relative height. However, stunting
is associated with fewer educational outcomes than relative height.
The OLS estimates (Table 4: 1–3) show that stunting would likely decrease children’s non-
verbal abilities (Raven’s test) but not significantly related to numerical scores, except when
they grow into adulthood. Stunted children would probably have lower cognitive scores in
childhood (0.13-SD) and maturity (0.06-SD). In this phase, being stunted would not likely
reduce mathematical scores. However, the influence of stunting on non-verbal cognitive abili-
ties diminishes in adulthood and has a more significant influence on mathematical profi-
ciency. Stunting is also linked to shorter years of schooling (0.3 years) and 0.06 years of later
enrolment in school. However, linear probit model shows that stunting has no significant rela-
tionship with the probability of repeated grades and dropping out of school.
The IV estimation results suggest that stunting tends to reduce children’s educational
achievement and opportunities for higher education (Table 4: 4–6). Childhood stunting is
linked to an around 0.42-year delay in entering school (or five months) and a 2-year reduction
in schooling. Regarding cognitive abilities, stunting influences non-verbal skills (Raven test

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 11 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Table 4. Estimation results for stunting and educational outcomes.

Stunting1 when measured prior to age five2
Educational outcomes OLS ci P IV ci P
(1) (2) (3) (4) (5) (6)
Childhood’s Raven (Z-scores) -0.13 (-0.23–0.04) 0.01 -0.82 (-1.42–0.21) 0.01
Childhood’s Numerical (Z-scores) -0.02 (-0.10–0.06) 0.64 -0.49 (-1.03–0.05) 0.08
Adolescence’s Raven (Z-scores) -0.06 (-0.12–0.01) 0.08 -0.38 (-0.72–0.03) 0.03
Adolescent’s Numerical (Z-scores) -0.02 (-0.09–0.05) 0.63 -0.44 (-0.82–0.06) 0.02
Adult’s Raven (Z-scores) -0.06 (-0.14–0.03) 0.18 -0.07 (-0.63–0.49) 0.80
Adult’s Numerical (Z-scores) -0.17 (-0.26–0.09) 0.00 -0.63 (-1.17–0.08) 0.03
Age started school (years) 0.06 (0.00–0.12) 0.04 0.42 (0.08–0.77) 0.02
Repeated grades (pp) 0.14 (0.00–0.27) 0.05 0.09 (-0.10–0.28) 0.36
Dropout (pp) 0.13 (0.00–0.26) 0.04 0.11 (-0.02–0.23) 0.10
Years of schooling (years) -0.3 (-0.53–0.08) 0.01 -2.06 (-3.47–0.65) 0.00

All values are marginal effects; 95% of CIs are in parentheses.

1
HAZ below -2 Standard Deviation according to WHO Standard Growth Reference for School-aged Children and Adolescents [37]. Control variables included but not
reported are sex, age, parental age, mother working, household size, household’s wealth index, household’s infrastructure (electricity, drinking water, and sanitation)
and region (rural/urban areas and fixed effect for main island). We have reported the results using a linear probability model for dichotomous outcomes (repeated
education and dropout school). The coefficient x 100 was the marginal effect in percentage points. HAZ: height-for-age z score; IV: instrumental variable; OLS: ordinary
least squares.
2
The details regarding the sample size can be found in S1 Table.

https://doi.org/10.1371/journal.pone.0295380.t004

scores) in childhood, but the influence of stunting is strengthened for numerical skills in all
three life stages, with the highest impact in adulthood.
We performed test statistics used to assess the strength of two stunting instruments, namely
birth season and mother’s height, indicate that both the LM test for endogeneity and the F test
for weak devices yield values that lead us to reject the null hypothesis. This means that the
excluded instruments are not correlated with the endogenous variable, and the instruments
have a strong explanatory power (S3 Table).

Discussion
Using a longitudinal survey from 1993 to 2014, we examine the correlation between the rela-
tive height and stunting status of children under five years old in 1993 or 1997, and their cog-
nitive and educational outcomes at ages 17 to 26 in 2014. This cohort consists of individuals
initially included in the study as children and successfully traced into adulthood in at least one
wave. During this 21-year period, which encompasses three life phases—school age, adoles-
cence, and adulthood—most participants had either completed high school or were engaged
in employment. We apply multivariate regression and instrumental variables models to inves-
tigate the potential impact of childhood stunting.
These findings are consistent with previous studies that showed that short stature of chil-
dren due to childhood stunting at an early age was associated with poor cognitive development
later in life, leading to reduced educational outcomes. Specifically, stunting is negatively linked
to cognitive abilities [17, 18, 20], lower educational outcomes [15, 20], and delays in enrolling
in primary school [20]. The results are significant after adjusting for the confounding effects of
age, sex, parental and household characteristics, and regional aspects.
Relative height (HAZ) and stunting significantly influence fluid intelligence during school
age. It can be concluded from the study that relative height and stunting have a greater

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 12 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

influence on fluid intelligence from school age to adulthood yet, as the individual grows, some
abilities, such as numerical ones, could be improved along the way. However, these altered
abilities may not compensate for the lagging of educational attainment.
This indicates that stunting and relative height work as an intermediary for cognitive defi-
cits. This is because poor health during childhood potentially contributes to difficulty follow-
ing formal education; thus, children might have difficulties attending lessons, increasing
absenteeism and lacking the energy to learn in the classroom [55].
The results also suggest that the adverse consequences of chronic undernutrition in early
life on children’s intellectual development may be exacerbated by environmental factors in the
family and/or community, such as the care and affection received from parents. Stunted chil-
dren may be treated differently from non-stunted children because of their smaller stature and
often appearing younger than their age (Rosenthal effect) [6, 55], which can affect their abilities
and interest in exploring their environment.
Furthermore, this study suggests that chronic malnutrition indirectly correlates with
schooling outcomes through decreased cognitive abilities. The effects were significant and rela-
tively stronger in the relationship between stunting and education. Stunted children in this
cohort had a marked delay in the first enrolment and a shorter length of schooling, and the
magnitude of the relationship between undernutrition and educational achievement in Indo-
nesia is relatively higher compared to other studies. The study found a 0.06-year delay in ele-
mentary school enrolment, compared to Victora’s (2008) finding (0.9 years) [15]. As for the
length of schooling, by applying a similar method, Hoddinott (2013) reported up to 4.6 years
of reduced schooling, while this study found two years shorter [20].
The declining relationship between stunting and cognitive education outcomes with
increasing age is likely a result of inadequate policies aimed at reducing stunting and poor edu-
cation sector performance that fails to provide optimal cognitive development opportunities
for non-stunted children [56, 57].
Bogin (2021) [58] argues that social-economic-political-emotional (SEPE) factors influence
community views towards adults based on their height which is more visible compared to the
more intangible assessment of cognitive ability, which explains the variation of the implication
of stunting in the different life course. Another argument stems from the possibility that
stunted children may experience growth delays after catching up on their height growth deficit
as an opportunity for extended growth due to delayed maturity after puberty [59, 60]. Third,
there is debate over the reference data used in measuring stunting. Scheffler and Hermanussen
(2021) [61] conducted a historical study on the Indonesian population that showed that the
height of Indonesians has never been equal to the European population, which is used as the
“normal value” for measuring stunting, making stunting as normal cognition on human
height. Therefore, shorter height does not imply a difference in physical fitness among chil-
dren with stunting [62]. Some studies provide alternative measurements of malnutrition that
might comply with the standards in the Indonesian population, such as height-for-difference
(HAD) and thus considered a more representative measure [63–66].
Therefore, this study highlights the importance of early intervention, particularly for chil-
dren nutritionally disadvantaged at age five, mainly due to its detrimental effect on child devel-
opment. Heckman (2007) suggests that interventions for disadvantaged young children are
more effective than those later in life, and remediation at a later age might be costly [67]. How-
ever, considering stunting is associated with various socio-cultural and economic disadvan-
tages thus, variations in environments and parenting practices may provide schooling and
other learning experiences which may mitigate the effects of early undernutrition on cognition
[17, 54]. Ensuring children with early stunting receive schooling comparable in quantity and
quality to non-stunted children could help improve their educational outcomes.

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 13 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

This implies policy responses that require the involvement of various parties at different
levels and the identification of actors needed to encourage changes at the community and
household levels, particularly during children’s early years. Indonesia has committed to invest-
ing significant resources, equivalent to USD51.9 trillion, in cross-sectoral strategies to address
stunting [68]. Food policy, equitable distribution of health provision at the village level, condi-
tional social assistance, clean water and sanitation infrastructure have been identified as the
most effective strategies for improving stunting rates and overall health quality in Indonesia
especially for the poor [24, 26, 69, 70]. The study also indicates the importance of household
wealth and parental education in children’s nutritional status and educational outcomes,
implying that policies to improve households’ livelihood would positively affect children’s
nutrition and education.
Our findings may be limited by the substantial level of attrition and exclusion of the vari-
ables that may affect relative height and education. Even though applying 2SLS with instru-
ment variables is considered the optimal effort to encounter the potential problems, it can be
challenging and hard to verify. Despite these challenges, the instrumental variable tests con-
ducted in this study are suitable. Meanwhile, this study’s strength is filling the literature gap on
the implications of stunting in Indonesia. Studies in similar areas have been conducted in vari-
ous countries, but none have been done specifically for the case of Indonesia. Yet, Indonesia is
a country with one of the largest populations and an economy that is considered globally
significant.

Conclusion
Our study shows a strong relationship between stunting and lower cognitive abilities that is
likely to persist and lead to lower educational outcomes over the long term. However, the rela-
tionship appears to weaken as individuals enter adulthood, potentially indicating the influence
of environmental factors. This finding suggests that recent development has not provided an
adequate environment for children to reach their academic potential, potentially leading to a
decline in future labour quality. To address this issue, it is necessary to prioritise addressing
stunting and its underlying determinants, including social and economic factors. This will
require collaborative efforts from various parties to address the causes of stunting and reduce
its prevalence.

Supporting information
S1 Table. Sample sizes for the regression results of the relationship between HAZ and
stunting on educational and cognitive achievements.
(DOCX)
S2 Table. Instrumental variables test statistics by domain for HAZ. Note: 1. *** is significant
at 95%. 2. Stock-Yogo critical values alpha = 5%; Bias 10%, two instruments: 19.93; Bias 15%,
two instruments: 11.59.
(DOCX)
S3 Table. Instrumental variables test statistics by domain for STUNTING. Note: 1. *** is
significant at 95%. 2. Stock-Yogo critical values alpha = 5%; Bias 10%, two instruments: 19.93;
Bias 15%, two instruments: 11.59.
(DOCX)

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 14 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

Acknowledgments
Part of the paper was presented at the 16th Indonesia Regional Science Association Conference
(IRSA) and the 7th Indonesia Health Economics Association (InaHEa) conferences in 2021.
We want to thank Firman Witoelar, M. Purnagunawan, and Marthin D. Siyaranamual for the
valuable feedback on the study design; and Siwage Dharma Negara for reviewing the manu-
script draft. We are also sincerely thankful to the anonymous referees and editors for their
insightful comments and constructive feedback, which significantly contributed to improving
this paper.

Author Contributions
Conceptualization: Esta Lestari, Adiatma Siregar, Achmad K. Hidayat, Arief A. Yusuf.
Data curation: Esta Lestari, Adiatma Siregar, Achmad K. Hidayat, Arief A. Yusuf.
Formal analysis: Esta Lestari, Adiatma Siregar, Arief A. Yusuf.
Investigation: Esta Lestari, Adiatma Siregar, Arief A. Yusuf.
Methodology: Esta Lestari, Arief A. Yusuf.
Software: Esta Lestari, Arief A. Yusuf.
Supervision: Adiatma Siregar, Achmad K. Hidayat, Arief A. Yusuf.
Validation: Esta Lestari, Achmad K. Hidayat, Arief A. Yusuf.
Visualization: Esta Lestari.
Writing – original draft: Esta Lestari, Adiatma Siregar.
Writing – review & editing: Esta Lestari, Adiatma Siregar, Achmad K. Hidayat, Arief A.
Yusuf.

References
1. WHO. WHA Global Nutrition Targets 2025: Stunting Policy Brief. WHO/NMH/NHD/14.3. 2014.
2. UNICEF; WHO; IBRD/ The World Bank Group. Levels and trends in child malnutrition: key findings of
the 2021 edition of the joint child malnutrition estimates. World Health Organization. Geneva; 2021.
3. Kementerian Kesehatan RI. Riskesdas 2018. Laporan Riskesdas Nasional 2018. 2018.
4. Alderman H, Hoddinott J, Kinsey B. Long term consequences of early childhood malnutrition. Oxf Econ
Pap. 2006; 58(3):450–74.
5. Dewey KG, Begum K. Long-term consequences of stunting in early life. Matern Child Nutr. 2011; 7
(SUPPL. 3):5–18. https://doi.org/10.1111/j.1740-8709.2011.00349.x PMID: 21929633
6. Leroy JL, Frongillo EA. Perspective: What Does Stunting Really Mean? A Critical Review of the Evi-
dence. Advances in Nutrition. 2019; 10(2):196–204. https://doi.org/10.1093/advances/nmy101 PMID:
30801614
7. de Onis M, Branca F. Childhood stunting: A global perspective. Matern Child Nutr. 2016; 12:12–26.
https://doi.org/10.1111/mcn.12231 PMID: 27187907
8. Alderman H. Economic drivers and consequences of stunting. Nestle Nutr Inst Workshop Ser. 2013;
71:131–41. https://doi.org/10.1159/000342592 PMID: 23502147
9. Prendergast AJ, Humphrey JH. The stunting syndrome in developing countries. Paediatr Int Child
Health. 2014; 34(4):250–65. https://doi.org/10.1179/2046905514Y.0000000158 PMID: 25310000
10. Chang SM, Walker SP, Grantham-McGregor S, Powell CA. Early childhood stunting and later behav-
iour and school achievement. J Child Psychol Psychiatry. 2002; 43(6):775–83. https://doi.org/10.1111/
1469-7610.00088 PMID: 12236612
11. Özaltin E, Hill K, Subramanian S V. Association of maternal stature with offspring mortality, under-
weight, and stunting in low- to middle-income countries. J Am Med Assoc. 2010; 303(15):1507–16.
https://doi.org/10.1001/jama.2010.450 PMID: 20407060

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 15 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

12. Horton S, Steckel RH. Malnutrition: Global Economic Losses Attributable to Malnutrition 1900–2000
and Projections to 2050. How Much have Global Problems Cost the World? 2014;247–72.
13. Adair LS, Fall CHD, Osmond C, Stein AD, Martorell R, Ramirez-Zea M, et al. Associations of linear
growth and relative weight gain during early life with adult health and human capital in countries of low
and middle income: Findings from five birth cohort studies. The Lancet. 2013; 382(9891):525–34.
https://doi.org/10.1016/S0140-6736(13)60103-8 PMID: 23541370
14. Hoddinott J, Alderman H, Behrman JR, Haddad L, Horton S. The economic rationale for investing in
stunting reduction. Matern Child Nutr. 2013; 9(S2):69–82. https://doi.org/10.1111/mcn.12080 PMID:
24074319
15. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L, et al. Maternal and child undernutrition:
consequences for adult health and human capital. The Lancet. 2008; 371(9609):340–57. https://doi.
org/10.1016/S0140-6736(07)61692-4 PMID: 18206223
16. Hall A, Khanh LNB, Son TH, Dung NQ, Lansdown RG, Dat DT, et al. An association between chronic
undernutrition and educational test scores in Vietnamese children. Eur J Clin Nutr. 2001; 55(9):801–4.
https://doi.org/10.1038/sj.ejcn.1601229 PMID: 11528497
17. Mendez MA, Adair LS. Severity and timing of stunting in the first two years of life affect performance on
cognitive tests in late childhood. Journal of Nutrition. 1999; 129(8):1555–62. https://doi.org/10.1093/jn/
129.8.1555 PMID: 10419990
18. Woldehanna T, Behrman JR, Araya MW. The effect of early childhood stunting on children’s cognitive
achievements: Evidence from young lives Ethiopia. Ethiopian Journal of Health Development. 2017; 31
(2):75–84. PMID: 29249889
19. Grantham-McGregor S, Cheung YB, Cueto S, Glewwe P, Richter L, Strupp B. Developmental potential
in the first 5 years for children in developing countries. Lancet. 2007; 369(9555):60–70. https://doi.org/
10.1016/S0140-6736(07)60032-4 PMID: 17208643
20. Hoddinott J, Behrman JR, Maluccio JA, Melgar P, Quisumbing AR, Ramirez-Zea M, et al. Adult conse-
quences of growth failure in early childhood. American Journal of Clinical Nutrition. 2013; 98(5):1170–8.
https://doi.org/10.3945/ajcn.113.064584 PMID: 24004889
21. The World Bank. The Human Capital Index 2020 Update: Human Capital in the Time of COVID-19.
IBRD/The World Bank. 2020.
22. De Silva I, Sumarto S. Child Malnutrition in Indonesia: Can Education, Sanitation and Healthcare Aug-
ment the Role of Income? J Int Dev. 2018; 30(5):837–64.
23. Putri OQ, Arimbi DQ, Fauzi HD. Study on Stunting Prevention Program in Indonesia: A Literature
Review. ASEAN/Asian Academic Society International Conference Proceeding Series. 2016; 0(0):1–5.
24. Titaley CR, Ariawan I, Hapsari D, Muasyaroh A, Dibley MJ. Determinants of the stunting of children
under two years old in Indonesia: A multilevel analysis of the 2013 Indonesia basic health survey. Nutri-
ents. 2019; 11(5). https://doi.org/10.3390/nu11051106 PMID: 31109058
25. Roemling C, Qaim M. Dual burden households and intra-household nutritional inequality in Indonesia.
Econ Hum Biol. 2013; 11(4):563–73. https://doi.org/10.1016/j.ehb.2013.07.001 PMID: 23890671
26. Cameron L, Chase C, Haque S, Joseph G, Pinto R, Wang Q. Childhood stunting and cognitive effects
of water and sanitation in Indonesia. Econ Hum Biol. 2021; 40:100944. https://doi.org/10.1016/j.ehb.
2020.100944 PMID: 33316572
27. Qureshy LF, Alderman H, Rokx C, Pinto R, Wai-Poi M, Tandon A. Positive returns: cost-benefit analysis
of a stunting intervention in Indonesia. J Dev Effect. 2013; 5(4):447–65.
28. Torlesse H, Cronin AA, Sebayang SK, Nandy R. Determinants of stunting in Indonesian children: Evi-
dence from a cross-sectional survey indicate a prominent role for the water, sanitation and hygiene sec-
tor in stunting reduction. BMC Public Health. 2016; 16(1):1–11. https://doi.org/10.1186/s12889-016-
3339-8 PMID: 27472935
29. Rachmi CN, Agho KE, Li M, Baur LA. Stunting, underweight and overweight in children aged 2.0–4.9
years in Indonesia: Prevalence trends and associated risk factors. PLoS One. 2016; 11(5):1–17. https://
doi.org/10.1371/journal.pone.0154756 PMID: 27167973
30. Behrman J, Lavy V. Child health and schooling achievement: association, causality and household allo-
cations. CARESS Working Papres. 1997.
31. Frankenberg, E; Karoly L. The 1993 Indonesian Family Life Survey: Overview and field report. Santa
Monica, CA., CA.; 1995.
32. Frankenberg, E; Thomas D. The Indonesia Family Life Survey (IFLS): Study Design and Results from
Waves 1 and 2. Santa Monica, CA, CA; Report No.: DRU-2238/1-NIA/NICHD.
33. Strauss, J; Beegle, K; Sikoki, B; Dwiyanto, A; Herawati, Y; Witoelar F. The Third Wave of the Indonesia
Family Life Survey (IFLS): Overview and Field Report. 1993. Report No.: WR-144/1-NIA/NICHD.

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 16 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

34. Strauss, J; Witoelar, F; Sikoki, B; Wattie AM. The Fourth Wave of the Indonesian Family Life Survey
(IFLS4): Overview and Field Report. Report No.: WR-675/1-NIA/NICHD.
35. Strauss J, Witoelar F, Sikoki B. The Fifth Wave of the Indonesia Family Life Survey: Overview and Field
Report: Volume 1. The Fifth Wave of the Indonesia Family Life Survey: Overview and Field Report: Vol-
ume 1. 2016; 1(March).
36. RAND Corporation. Register to download data IFLS [Internet]. Available from: https://www.rand.org/
well-being/social-and-behavioral-policy/data/FLS/IFLS/access.html.
37. World Health Organization. 2006 WHO Child growth standards [Internet]. Available from: https://www.
who.int/toolkits/child-growth-standards/standards.
38. De Onis M. WHO Child Growth Standards based on length/height, weight and age. Acta Paediatrica,
International Journal of Paediatrics. 2006; 95(SUPPL. 450):76–85.
39. Kementerian Kesehatan RI. Peraturan Menteri Kesehatan RI No. 2/2020 tentang Standar Antropometri
Anak. Jakarta; 2020 p. 1–9. Available online from: https://yankes.kemkes.go.id/unduhan/fileunduhan_
1660187306_961415.pdf.
40. Fitzgerald J, Gottschalk P, Moffitt R. An analysis of sample attrition panel data: the Michigan panel
study of income dynamics. Journal of Human Resources. 1998; 33(2):251–99.
41. Alderman H, Behrman JR, Lavy V, Menon R, Alderman H, Behrman JR, et al. Child Health and School
Enrollment: A Longitudinal Analysis. J Hum Resour. 2001; 36(1):185–205.
42. Raven J. The Raven’s Progressive Matrices: Change and Stability over Culture and Time. Cogn Psy-
chol. 2000; 41(1):1–48. https://doi.org/10.1006/cogp.1999.0735 PMID: 10945921
43. Galasso E, Wagstaff A. The aggregate income losses from childhood stunting and the returns to a nutri-
tion intervention aimed at reducing stunting. Econ Hum Biol. 2019; 34:225–38. https://doi.org/10.1016/j.
ehb.2019.01.010 PMID: 31003858
44. Maika A, Mittinty MN, Brinkman S, Harper S, Satriawan E, Lynch JW. Changes in socioeconomic
inequality in Indonesian children’s cognitive function from 2000 to 2007: a decomposition analysis.
PLoS One. 2013; 8(10). https://doi.org/10.1371/journal.pone.0078809 PMID: 24205322
45. Behrman JR, Hoddinott J, Maluccio JA, Martorell R. Brains versus Brawn: Labor Market Returns to
Intellectual and Health Human Capital in a Poor Developing Country. SSRN Electronic Journal. 2011;
(0907).
46. Vyas S, Kumaranayake L. Constructing socio-economic status indices: How to use principal compo-
nents analysis. Health Policy Plan. 2006; 21(6):459–68. https://doi.org/10.1093/heapol/czl029 PMID:
17030551
47. Behrman JR. Nutrition, health, birth order and seasonality. Intrahousehold allocation among children in
rural India. J Dev Econ. 1988; 28(1):43–62.
48. Cueto S, Leon J, Miranda A, Dearden KA, Crookston BT, Behrman JR. Does pre-school improve cogni-
tive abilities among children with early-life stunting? A longitudinal study for Peru. International Journal
Educaation Research. 2017; 176(10):139–48.
49. Wooldridge JM. Introductory Econometrics: A Modern Approach. Economic Analysis. 2003.
50. Kinyoki DK, Berkley JA, Moloney GM, Odundo EO, Kandala NB, Noor AM. Environmental predictors of
stunting among children under-five in Somalia: Cross-sectional studies from 2007 to 2010. BMC Public
Health. 2016; 16(1):1–9. https://doi.org/10.1186/s12889-016-3320-6 PMID: 27464568
51. Schultz TP. Human capital, schooling and health. Econ Hum Biol. 2003; 1(2):207–21. https://doi.org/10.
1016/S1570-677X(03)00035-2 PMID: 15463974
52. Maccini S, Yang D. Under the weather: Health, schooling, and economic consequences of early-life
rainfall. American Economic Review. 2009; 99(3):1006–26. https://doi.org/10.1257/aer.99.3.1006
PMID: 29508969
53. Hoddinott J, Maluccio J, Behrman JR, Martorell R, Melgar P, Quisumbing AR, et al. The consequences
of early childhood growth failure over the life course. IFPRI Discussion Paper #01073. 2011;(March).
54. Stock JH, Yogo M. Testing for weak instruments in Linear IV regression. In: Identification and Inference
for Econometric Models: Essays in Honor of Thomas Rothenberg. 2005. p. 80–108.
55. Brown JL, Pollitt E. Malnutrition, Poverty and Intellectual Development. Sci Am. 1996; 274(2):38–43.
https://doi.org/10.1038/scientificamerican0296-38 PMID: 8560214
56. Suryadarma D, Suryahadi A, Sumarto S. Causes of Low Secondary School Enrollment. The SMERU
Research Institute. 2006.
57. Muttaqin T, Wittek R, Heyse L, Van Duijn M. Why Do Children Stay Out Of School In Indonesia? Jurnal
Perencanaan Pembangunan: The Indonesian Journal of Development Planning. 2017; 1(2):93–108.
58. Bogin B. Social-Economic-Political-Emotional (SEPE) factors regulate human growth. Human Biology
and Public Health. 2021; 1(June).

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 17 / 18

PLOS ONE Stunting and educational outcomes in Indonesia

59. Benny L, Boyden J, Penny M. Early is best but it ‘ s not always too late. Young Lives evidence on nutri-
tion and growth in Ethiopia, India, Peru and Vietnam. 2018. 1–79 p.
60. Soliman A, De Sanctis V, Alaaraj N, Ahmed S, Alyafei F, Hamed N, et al. Early and long-term conse-
quences of nutritional stunting: From childhood to adulthood. Acta Biomedica. 2021; 92(1):1–12. https://
doi.org/10.23750/abm.v92i1.11346 PMID: 33682846
61. Scheffler C, Hermanussen M. Stunting is the natural condition of human height. American Journal of
Human Biology. 2022 May; 34(5):1–13. https://doi.org/10.1002/ajhb.23693 PMID: 34761833
62. Scheffler C, Hermanussen M. Stunting does not impair physical fitness in Indonesian school children.
Human Biology and Public Health. 2021; 2(January).
63. Pulungan AB, Julia M, Batubara JR, Hermanussen M. Indonesian National Synthetic Growth Charts.
Research Article “Indonesian National Synthetic Growth Charts” Acta Scientific Paediatrics. 2018; 1
(August):20–34.
64. Alkaff FF, Flynn J, Sukmajaya WP, Salamah S. Comparison of WHO growth standard and national
Indonesian growth reference in determining prevalence and determinants of stunting and underweight
in children under five: A cross-sectional study from Musi sub-district. F1000Res. 2020; 9(May):1–18.
https://doi.org/10.12688/f1000research.23156.4 PMID: 32551096
65. Leroy JL, Ruel M, Habicht JP, Frongillo EA. Using height-for-age differences (HAD) instead of height-
for-age z-scores (HAZ) for the meaningful measurement of population-level catch-up in linear growth in
children less than 5 years of age. BMC Pediatr. 2015; 15(1):1–11.
66. Himaz R. Stunting later in childhood and outcomes as a young adult: Evidence from India. World Dev.
2018.
67. Heckman JJ, Dimitriy VM, Masterov D. The Productivity Argument for Investing in Young Children.
NBER Working Paper No. 13016. National Bureau of Economic Research. 2007.
68. The World Bank. Spending Better to Reduce Stunting in Indonesia: Findings from a Public Expenditure
Review. 2020.
69. Booth A, Purnagunawan RM, Satriawan E. Towards a Healthy Indonesia? Bull Indones Econ Stud.
2019; 55(2):133–55.
70. Heckman JJ, Dimitriy VM. NBER Working Paper Series: The Productivity argument for investing in
young children. UkpmcAcUk. 2007.

PLOS ONE | https://doi.org/10.1371/journal.pone.0295380 May 6, 2024 18 / 18