Anemia and Child Education: The Case of Colombia
| Autor | Alejandro Gaviria, Alejandro Hoyos |
| Páginas | 47-77 |
47
Anemia and Child Education:
The Case of Colombia
Anemia y progreso escolar de los niños:
el caso colombiano
Alejandro Gaviria y Alejandro Hoyos*
Abstract
Welfare programs in Colombia have focused on both reducing malnu-
trition and hunger and increasing school attendance. But there is not
much evidence on the hypothesized relationship between nutrition
status and schooling outcomes. Using the National Survey of Nutri-
tional Status in Colombia – 2005 (E N S I N ) and the Demographic and
Health Survey – 2005 (D H S ), this paper estimates the impact of nutrition
on schooling outcomes. The results suggest that anemic children have a
anthropometric measures, does not have an impact on the probability
of being overage. School attendance seems to be unrelated to nutrition
Key words: School attendance, overage, nutrition, micronutrients.
JEL: I12, I20, O15.
* Authors are listed in alphabetical order. Contact information: Alejandro Gaviria, Economics
Department, Universidad de Los Andes, email: agaviria@uniandes.edu.co. Alejandro Hoyos,
World Bank, email: ahoyossuarez@worldbank.org. We thank Profamilia for providing the
E N S I N dataset. We are grateful to Laura Nivia, Holly Kosiewicz, Emily Conover, Adriana
Camacho and Natalia Millán for their comments and suggestions. Any mistake within the
represent the views of the World Bank or its Board of Directors.
-
cepted August 18, 2011.
Revista
Desarrollo y Sociedad
68
II semestre 2011
Anemia and Child Education: The Case of Colombia
48
Resumen
Varios programas sociales en Colombia aspiran a reducir la desnutri-
ción e incrementar la asistencia escolar. A pesar de ello, la evidencia
sobre la conexión entre los indicadores de nutrición y los de progreso
educativo es escasa. A partir de los datos de la Encuesta E N S I N , este
trabajo estudia empíricamente el efecto de una mala nutrición (medida,
alternativamente, a partir del nivel de hierro en la sangre, la estatura y
el peso) sobre la asistencia educativa y el rezago escolar. Los resultados
muestran que los niños anémicos tienen una mayor probabilidad de
estar rezagados, pero no tanto así los niños que sufren de desnutrición
crónica y bajo peso. En general la asistencia escolar no parece estar
afectada por ninguno de los indicadores de desnutrición.
Palabras clave: asistencia escolar, rezago escolar, nutrición, micro-
nutrientes.
JEL: I12, I20, O15.
Introduction
Human capital is amply recognized as a determinant of economic
growth and social wellbeing in general1. Welfare programs have
focused on reducing malnutrition and hunger, and on increasing school
attendance in developing countries. The Copenhagen Consensus
2004 ranked the reduction of malnutrition and hunger through the
provision of micronutrients as the second most desirable intervention
among a large number of social programs and government regulations
(Lomborg, 2006).
report that, in 1995, 9.5% of Colombian children up to 5 years of age
were moderately stunted, and 2.1% severely stunted2. The moderate
1
2 Moderate stunting are children with a height-for-age Z-score between –2 and –3 standard
deviations of the reference standard. Severe stunting are children with a height-for-age
Z-score below –3 standard deviations of the reference standard.
Alejandro Gaviria y Alejandro Hoyos 49
and severe stunting rates in Latin America and the Caribbean were
Colombia is lower than the region’s average, it´s a cause of concern
because of its effects on education, productivity and economic
growth. In the last decade, the Colombian government expanded
social protection programs aimed at reducing malnutrition and hunger
and boosting school attendance.
, a conditional cash transfer program,
grew from 300 thousand families in 2002 to almost three million in
2010. This program imposes two conditions upon families: one based
on controls of weight and height of children up to 6 years of age, and
the other based on school attendance of older children. In 1996 the
-
tion Program. Some of its main objectives were to enrich basic food
products with micronutrients, providing iron and other supplements
to pregnant women and children up to 5 years of age. In 1996, Decree
iron, folic acid and niacin. The Decree also mandated that salt must
be appropriately iodized.
-
tive development, long–term health outcomes, productivity and wages
earnings, income inequality and economic growth.
Taken together these results suggest that interventions which improve
both cognitive skills of individuals and educational and economic
Anemia and Child Education: The Case of Colombia
50
exposure of children to a similar intervention in Colombia. Maluccio,
that, in Guatemala, a nutrition supplement was associated with gains
in cognitive skills and educational attainment after 25 years.
Several studies suggest that poor nutrition and poor health could lead
Sharma (2006) study some interventions in which preschool children
(2006) use a longitudinal dataset to evaluate the impact of hemoglobin
that higher levels of hemoglobin are associated with greater school
as they might be driven by omitted variables.
with nutritional and health status is that these two dimensions are in
many ways parental decisions potentially correlated with unobserved
characteristics, such as the extent of parental involvement and care.
Alderman, Behrman, Lavy and Menos (2001) use an instrumental-
and nutritional outcomes are three times more important for school
enrollment than is suggested by other estimates that do not take into
account the potential endogenity of the main variables of interest.
Many studies have examined the determinants of school attendance,
school achievement and nutritional status in Colombia. However, few
studies have estimated the impact of nutritional status and health on
schooling outcomes3 and as the Colombian government has expanded
social programs in education and health, this paper could provide valu-
able information to understand the complementarities of an interven-
tion focus on these two dimensions (education and health).
3
status on human capital accumulation for adult women in Colombia.
Alejandro Gaviria y Alejandro Hoyos 51
This paper estimates the impact of nutrition on school attendance
and on the probability of being overage4. Two types of measures
of nutrition are used: anthropometric measures (z-scores of height-
for-age, z-scores of weight-for-age and Body Mass Index (B M I ))
and micronutrients measures (hemoglobin). Data is taken from the
National Survey of Nutritional Status in Colombia – 2005 (E N S I N ),
the sample used in the E N S I N is a subsample of the Demographic and
Health Survey – 2005 (D H S
measured nutritional status using micronutrient levels in blood. The
to correct the omitted variables bias. Various robustness checks are
included in order to validate the estimates but causal interpretations
are not always warranted.
The results suggest that anemia appears to increase the probability of
being overage but do not have a discernable impact on school atten-
have an impact either on school attendance or on the probability of
being overage. This paper is structured as follows: section I discusses
the literature on nutrition and its impact on education and economic
growth. Section II describes the database. Section III lays out the
empirical strategy. Section IV presents the main results. Section V
presents some robustness checks and section VI concludes.
I. Literature review
A. Nutrition and child development
According to Stoltzfus (2001), the World Health Organization (W H O )
-
ciency”. DeMaeyer and Adiels-Tegman (1985) identify that iron
4 We consider that a child is overage if his or her years of schooling are strictly less than the
median of the population for the child’s gender and age minus one.
Anemia and Child Education: The Case of Colombia
52
level in the blood.
only causes. Assis, Barreto, Gomes, Prado, Santos and Santos (2004)
identify low levels in other micronutrients, intestinal diseases - which
implied loss of blood -, poor sanitation and environmental conditions
as alternative causes of anemia.
What are the consequences of a poor nutritional status in general?
-
quences of malnutrition. These authors examine the way stunted or
underweight children tend to suffer health disorders that are in turn
is associated with mental retardation and poor cognitive development,
in brain development:
Iron has many functions in the brain. It is necessary for the
production of myelin, the fatty coating around neurons that
speeds the transmission of electrical signs; it facilitates the
production of neurotransmitters […]; it is involved in the func-
tion of neuroreceptors […]; and is essential for the metabolic
processes that provide energy to the brain
p. 65).
The adverse consequences of malnutrition may be mitigated with
interventions that include micronutrient supplements, improvements
in diet, deworming, and others. Some authors have offered a guideline
for successful nutritional interventions (Stoltzfus and Dreyfuss, 1998;
have evaluated the impact of a series of interventions on nutritional
Alejandro Gaviria y Alejandro Hoyos 53
and health status, cognitive development, productivity and the long-
term development of children5.
B. School outcomes and nutrition
Different approaches have been used to examine the relationship
between nutrition and educational outcomes (enrollment and achieve-
ment). Several studies have consistently found that a better nutritional
status leads to a greater chance of school attendance, earlier enrollment
and less grade repetition. Bobonis et al. (2006) carried out a random
evaluation of a health intervention in India, through which preschool
children, between two and six years of age, were provided with iron,
vitamin A supplements and deworming drugs. They found that treated
children experienced a gain of 5.8 percentage points in the probability
-
dren with higher anemia rates at the baseline experienced the greatest
improvements. This evaluation provides strong evidence of the positive
relationship between nutritional status and school attendance.
Ghuman et al. (2006) use a longitudinal dataset to gauge the effect of
nutritional status on school enrollment in early childhood in Philip-
pines. They measure nutrition using both anthropometric measures
-
cies (anemia). They implement an instrumental-variable approach in
order to correct for potential endogeneity problems, and use the quality
of daycare service providers as their instrument. It supposes that the
health and nutritional status of a child should be explained, at some
point, by the quality of the daycare service provider when they were
the instrument used by the authors is not related with school atten-
dance, the quality of daycare service providers is not related with
school attendance directly or at least through other channels different
to nutrition and health.
5
studies that evaluate the impacts of some interventions.
Anemia and Child Education: The Case of Colombia
54
Alderman et al. (2001) use a dynamic model to show that regress school
enrollment on a measure of nutritional status might
lead to incorrect estimates. The reason is that the previous nutritional
status and current school enrollment are driven by contemporaneous
decision of the parents. Thus, the current nutritional status could be
determined by the current school enrollment and the causal effect
of a contemporaneous measure of nutrition on school enrollment
importance of using longitudinal datasets that permit measurement of
nutrition prior to school enrollment. They use a longitudinal dataset
nutritional status is instrumented with price shocks during pre-school
age. Results suggest that the effect of nutrition (height-for-age) on
is three times larger for girls than for boys.
a negative impact on school enrollment. The main reasons are asso-
ciated with parental decisions. Parents perceive smaller children as
both physically and mentally immature and consequentially decide to
wait until children are older to enroll them in school. Besides, healthy
disorders, has a positive impact on children’s schooling. The impact
appears to be associated with improvements in cognition rather than
with reduction in absenteeism due to illness.
Other studies have focused on estimating the effect of nutrition on
the impact of nutritional interventions on cognitive abilities. His own
studies show that, on average, nutritional interventions in early child-
hood have a positive impact of two thirds of a standard deviation on
preschool children who received some nutritional support experience
-
tions. Suffering from anemia may be associated with a loss in IQ score
Alejandro Gaviria y Alejandro Hoyos 55
health and nutritional interventions are highly cost-effective and have
a consistently large impact on cognitive development.
This paper is not an evaluation of a nutritional intervention upon
cognitive abilities or educational outcomes. We use a cross-sectional
database in an attempt to estimate the impact of malnutrition on educa-
tional outcomes. The characteristics of the database limit the range of
methodologies that might be applied and therefore cast some doubts
on the causal inferences we can make. Despite its methodological
This study differs from most of the literature in the use of the prob-
ability of being overage as an educational outcome. This variable
combines grade repetition, non-attendance for one or more periods
(years) and late school enrollment. Most previous studies have focused
exclusively on school enrollment, school attendance, and test scores.
-
ized test scores or school grades which might allow us to evaluate
the impact of nutrition on school achievement. This study is similar
to the evaluation of Ghuman et al. (2006), however, we focus more on
the overage and not as much on the school attendance. Moreover, the
study of Ghuman et al. uses a longitudinal data set which is a better
approach to explore the relation between nutrition and educational
outcomes. Because of the limitation of the data used in this paper, the
results should not be interpreted as casual effects.
II. Data
The dataset used in this paper is the National Survey of Nutritional
Status in Colombia – 2005 (E N S I N ), which is based on a subsample of
the Demographic and Health Survey – 2005 (D H S ). In addition to the
information provided by the D H S —anthropometric measures of children,
socioeconomic and health characteristics of mothers, among others,
the E N S I N contains information of micronutrients (hemoglobin, ferritin,
vitamin A, and Zinc), food safety and physical activity of individuals.
Anemia and Child Education: The Case of Colombia
56
The E N S I N
1,920 segments and 209 primary sample units (mainly municipali-
ties). It is representative of urban and rural areas, of six regions and
fourteen sub-regions in Colombia. After restricting the sample to chil-
dren between one to twelve years of age, the number of observations
areas and 25% to the rural areas.
We used three measures for children’s education: school attendance,
schooling gaps (years) and overage (dichotomous variable). As
mentioned earlier, the last two variables implicitly combine grade
repetition, later enrollment and periods of non school attendance (three
outcomes of interest). Table 1 shows the median years of education by
gender and age using the 10% of the 1993 Colombian Census from
I P U M S -International. Schooling gaps are calculated as the maximum
between zero and the actual schooling minus the median schooling
of the population for the child’s same gender and age minus one. It
takes positive values for overage children and zero otherwise. A child
is considered as overage in school if his or her years of schooling are
strictly less than the corresponding median.
Table 1. Median of years of education in the 1993 Colombian Census
Age Gender
Male Female
5 0 0
6 0 0
1 1
8 1 2
9 2 2
10 3 3
11 4 4
12 4 5
: National Administrative Department of Statistics 1993 C olombian census. Minnesota
Population Center. Integrated Public Use Microdata Series.
The sample used in this paper is limited to boys between 9 and 12
years, and girls between 8 and 12 years of age, the resulting sample has
2.965 observations. All the descriptive statistics and results correspond
to this sample. The school-attendance rate is pretty high (95.9%), the
Alejandro Gaviria y Alejandro Hoyos 57
gap is 0.28 years. The nutritional status of children was measured using
micronutrients levels in the blood. In E N S I N , levels of hemoglobin,
ferritin, vitamin A and zinc were measured by taking blood samples
during the visit to households. Levels of hemoglobin were obtained
using the HemoCue system, which enabled the interviewers to obtain
a measure in a matter of seconds. In contrast, ferritin, vitamin A and
Zinc had to be measured later on in the Colombian National Institute
of Health laboratories.
The World Health Organization has established a hemoglobin level
below 12 grams per deciliter of blood among children from 5 to 12
years as the cut-off level for assessing whether a child is anemic or not.
According to this standard, about 21% of the children in the sample
in the anemia rate between girls and boys (20.1% and 22.1%, respec-
tively). Table 2 presents descriptive statistics of the main variables
and covariates used in the analysis.
Table 2. Summary statistics
Child Obs. Mean Std Dev. Min Max
Boys (%) 2965 45.11 0.0 100.0
Age 2965 1.31 8.0 12.0
Child (%) 2965 45.11 0.0 100.0
Grandchild (%) 2965 19.23 39.42 0.0 100.0
Hemoglobin (g/dl) 2965 12.90 1.22 18.1
Boys 1365 12.90 1.22 18.1
Girls 1600 12.91 1.22 8.9
Anemia (%) 2965 20.99 0.0 100.0
Boys 1365 22.13 41.53 0.0 100.0
Girls 1600 20.06 40.06 0.0 100.0
Chronic (%) 2521 19.54 39.66 0.0 100.0
Boys 1133 21.83 41.33 0.0 100.0
Girls 1388 38.19 0.0 100.0
Global (%) 2521 41.59 49.30 0.0 100.0
Boys 1133 48.45 50.00 0.0 100.0
Girls 1388 36.13 48.06 0.0 100.0
Underweight (%) 2521 9.02 28.65 0.0 100.0
Boys 1133 9.80 0.0 100.0
Girls 1388 8.40 0.0 100.0
School attendance (%) 2965 95.99 19.61 0.0 100.0
Proportion of overage children (%) 2965 34.46 0.0 100.0
Schooling Gap (years) 2965 0.28 0.0 4.0
Anemia and Child Education: The Case of Colombia
58
Table 2. Summary statistics (continued)
Child Obs. Mean Std Dev. Min Max
Mother
Mothers's education (years) 4.41 0 20
Mother works (%) 2485 43.02 49.52 0 100
Household
Proportion of children up to 5 years (%) 2965 8.92 0 50
Proportion of workers (%) 2965 29.18 16.06 0 100
Number of siblings 2812 1.94 0 10
: D H S – 2005, E N S I N – 2005. Author’s calculations.
There appears to be a consistent relationship between the economic
conditions of a household, on the one hand, and the educational
outcomes and the nutritional status of children, on the other. As the
level of household wealth increases, rates of anemia fall as well as
the proportion of overage children. Attendance rates are slightly higher
6. The fraction of anemic children is
Table 3 presents summary statistics by wealth quintiles.
Table 3. Summary statistics by wealth quintiles
Wealth
quintiles %Attendance
(%)
Proportion
of overage
children
(%)
Schooling
gap
(years)
Hemoglo-
bin
(g/dL)
Anemia (%)
1 22.13 95.25 0.36 12.50 33.15
2 23.88 94.63 15.39 0.28 25.20
3 21.86 95.56 9.26 0.23 12.90 20.04
4 18.14 9.89 0.23 13.16
5 13.98 9.26 13.48
: D H S – 2005, E N S I N – 2005. Author’s calculations.
There are six geographical regions in Colombia (Atlántica, Oriental,
differ greatly in their cultural and nutritional habits, as well in their
topographical and agro-climatic conditions and in their levels of
economic development. All of these differences might contribute to
6
quintiles were constructed from existing data on household assets, dwelling characteristics
and access to public services using Principal Components.
Alejandro Gaviria y Alejandro Hoyos 59
differences in educational and nutritional outcomes. The Atlántica
region presents the highest proportion of overage children and anemic
-
by geographical region.
Table 4. Summary statistics by geographical region
Region % Attendance
(%)
Proportion
of overage
children
(%)
Schooling
gap
(years)
Hemoglobin
(g/dL)
Anemia
(%)
Atlántica 25.10 98.13 19.93 0.40 11.98
Oriental 18.56 15.90 0.39 9.09
Central 93.00 10.41 0.19 12.91
16.23 94.49 11.98 13.14
Bogotá 10.89 8.26 0.21 13.93 0.66
Orinoquia-Amazonia 2.48 98.06 0.13 12.66 25.53
: D H S – 2005, E N S I N – 2005. Author’s calculations.
Measures of nutrition based on micronutrients in blood were contrasted
with anthropometric measures of nutrition based on height and weight.
To compare the results of anemia and anthropometric measures, we
restrict the sample to boys between 9 and 12 years, and girls between 8
and 12 years of age, who have information on anemia, weight and height.
After restricting the sample, the sample is reduced to 2.521 children.
malnutrition as a low weight for the age. Similarly, a child is said to
be underweight if he or she has a low Body Mass Index for his or her
age. All measures are based on a set of reference parameters provided
by the National Center of Health Statistics of the United States8.
To determine that a child has some form of malnutrition, a z-score
Body Mass Index is the weight in kilograms divided by the square of the height in me-
ters.
8 Parameters of references may be consulted in the web site of the National Center of Health
Anemia and Child Education: The Case of Colombia
60
has to be constructed using the reference parameters9. Children with
z-scores values below -2 are considered to be suffering from malnutri-
tion (chronic malnutrition, global malnutrition or underweight). In the
sample, the chronic malnutrition rate is 19.5%, the global malnutrition
rate is 41.6% and the underweight rate is 9.0%. Global and chronic
malnutrition are statistically different by gender. Underweight does
not vary by gender.
10
mean hemoglobin and z-Body Mass Index, z-weight and z -height are
0.66, 0.36 and 0.32, respectively. The relationship between z-scores and
rates against anthropometric measures. The corresponding correlation
malnourished children are more likely to suffer from anemia.
Figure 1. Mean hemoglobin and anthropometric nutritional measures
Hemoglobine
Z-Body mass index
-2
10
11
12
13
14
15
-1 012
Z-Weight Z-Height
-2
10
11
12
13
14
15
-1 012 -2
10
11
12
13
14
15
-1 012
9 Z – Scores are constructed as:
z
x
M
LS
L
=
−
×
1
, where x is the anthropometric measure of a
child, M is the median of reference, S is the ge
the power in the Box-Cox transformation. M, L and S are the parameters of reference.
10
with z-scores in the interval [x - 0.05, x + 0.05]. We do this for all 41 values in this sequence:
-2, -1,9, -1,8,…1,8, 1,9, 2.
Alejandro Gaviria y Alejandro Hoyos 61
Figure 2. Mean anemia and anthropometric nutritional measures
Z-Weight Z-Height
Anemia (%)
Z-Body mass index
-2
0
10
20
30
40
50
-1 012 -2
0
10
20
30
40
50
-1 012 -2
0
10
20
30
40
50
-1 0 12
attendance rates against all four nutritional measures (rounded to the
nearest tenths). There is a negative relationship between the nutritional
status and the fraction of overage children. On the other hand, school
attendance and nutritional measures appear to be uncorrelated. These
Figure 3. Probability of being overage over nutritional measures
0
20
40
60
80
100
Overage (%)
810121416
0
20
40
60
80
100
-2 -1 012
Hemoglobine Z - Body Mass Index
0
20
40
60
80
100
-2 -1 012
0
20
40
60
80
100
-2 -1 012
Z - Weight Z - Height
Overage (%)
Anemia and Child Education: The Case of Colombia
62
Figure 4. Mean school attendance and nutritional measures
8
70
75
80
85
90
95
100
70
75
80
85
90
95
100
80
85
90
95
100
10 12 14 16 -2 -1 012
-2 -1 012
70
75
80
85
90
95
100
-2 -1 012
Hemoglobine
Z-Weight Z-Height
Z-Body mass index
School attendance (%)School attendance (%)
III. Empirical strategy
This paper uses a similar methodology to Ghuman et al. (2006). But
because the database used in this paper is not longitudinal, we use
contemporaneous measurements of nutrition and education. We assume
that the nutritional status is a cumulative process, which implies, among
other things, that the current nutritional status has a high correlation
with the previous nutritional history. Hence, we used the current nutri-
tional status as a proxy for early nutrition.
EG NIMHR
=+++ +++()
(1)
E is an educa-
tional outcome (overage or school enrollment) of child I, N is the
nutritional status (anemia, global malnutrition, chronic malnutrition
or being underweight), I is a vector of the individual’s characteristics
such as gender, age and relation with the household head, M is a vector
Alejandro Gaviria y Alejandro Hoyos 63
of mother’s characteristics such as years of education, age and whether
the mother has a job, H is a vector of a household’s characteristics
such as the proportion of children up to 5 years of age, the proportion
of household members actually working, the number of siblings, and
R is a vector
of the region’s characteristics.
Two dependent variables are used in the analysis: whether the child
is overage and whether he or she is currently enrolled (school atten-
dance). Overage is a dichotomous variable that takes the value of
one if the years of schooling are strictly less than the median of the
population for the child’s same gender and age. School attendance
is also a dichotomous variable that takes the value of one if the child is
currently attending a school. We estimate equation 1 using a Probit
model and report the marginal effects. As a robustness check, we use
the overage measured in years as an alternative dependent variable.
This variable can take values between zero and four. In this case we
use an ordered Probit model instead.
IV. Analytical results
This section presents the main results of the paper. All the regressions
are weighted to account for the sampling scheme. Standard errors are
corrected for clustering at the municipality level; because of the design
of the survey, the municipality is the primary sample unit. Table 5
equation 1. The results show the changes in the probability of being
overage. Children who suffer from anemia are 4.6 percentage points
the expected sign. Children with mothers who are more educated are
less likely to be overage: an increase in one year of the education of
the mother reduces this probability by 0.6 percentage points. The
of anemia is substantial: it is comparable to an increase of eight years
in the mother’s education.
Anemia and Child Education: The Case of Colombia
64
Table 5. Regressions on the probability of being overage
Probit estimation
Dependent
variable
Overage
(1) (2) (3) (4) (5) (6) (7) (8)
Anemia 0.085*** 0.059*** 0.053** 0.046**
[0.024] [0.021] [0.022] [0.022]
Hemoglobin
(g/dL)
-0.028*** -0.014** -0.012** -0.010
[0.008]
Male 0.013 0.009 0.008 0.008 0.014 0.010 0.009
[0.018] [0.018] [0.018] [0.018] [0.018] [0.018] [0.018]
Age 0.052*** 0.049*** 0.050*** 0.049*** 0.053*** 0.049*** 0.049*** 0.049***
[0.008]
Child -0.028 0.032 0.026 0.020 -0,026 0.034 0.026 0.021
[0.028] [0.041] [0.042] [0.028] [0.041] [0.042]
Grandchild -0.040 0.009 0.014 0.005 -0.039 0.011 0.015 0.006
[0.026] [0.045] [0.046] [0.026] [0.046] [0.048] [0.046]
Mother's
education
-0.005 -0.006* -0.005 -0.006*
[0.003] [0.004] [0.003] [0.003] [0.004] [0.003]
Mother works 0.028 0.031 0.006 0.031
[0.018] [0.022] [0.022] [0.018] [0.023] [0.022]
Proportion of
children up to 5
years
0.036 0.043 0.043 0,049
Proportion of
workers
-0.053 -0.053
Number of
siblings
0.010* 0.010* 0.011* 0.010*
[0.006] [0.006] [0.006] [0.006]
Wealth quintiles
Yes Yes Yes Yes
effects Yes Yes
Observations 2965 2334 2334 2965 2334 2334
Pseudo
R-squared 0.061 0.081 0.094 0.102 0.063 0.091 0.100
Clustered standard error at the municipality level in brackets.
Overage is a dichotomous variable that takes the value of one if the years of education are stricly
less than the median of the population for the same gender and age.
of siblings. An additional sibling increases the probability of being
overage by one percentage point. Individual controls, such as age and
are apparent but the probability of being overage increases with age.
Alejandro Gaviria y Alejandro Hoyos 65
The estimate of the effect of anemia on the probability of being overage
result suggests that anemia (and malnutrition in general) may be highly
-
ence on educational outcomes. If we were to omit relevant variables at
the regional or parental level, we would arrive at an incorrect estimate
of the marginal effects. Some researchers have suggested in many studies
that estimates of the effect of nutrition on educational outcomes may be
biased due to omitted variables11. The database used in this paper does not
allow us to fully correct for this problem12. We include in the empirical
order to reduce the potential bias. Section 6 presents robustness checks
to validate the results presented in this section.
Table 6 shows the estimated effects of anemia on school attendance.
The probability of attending school does not depend on anemia or
Table 6. Regressions of school attendance
Probit estimation
Dependent variable School attendance
(1) (2) (3) (4) (5) (6) (7) (8)
Anemia 0.005 0.010 0.008 -0.001
[0.012] [0.011] [0.010] [0.009]
Hemoglobin (g/dL) -0.001 -0.002 -0.002 -0.001
[0.004] [0.004] [0.003] [0.003]
Individual controls Yes Yes Yes Yes Yes Yes Yes Yes
Mother controls Yes Yes Yes Yes Yes Yes
Household controls Yes Yes Yes Yes
effects Yes Yes Yes Yes
Yes Yes
Observations 2965 2334 2334 2965 2334 2334
Pseudo R-squared 0.049 0.094 0.159 0.048 0.093 0.159
Clustered standard errors at the municipality level in brackets.
11
12 We attempt to implement an instrumental variables approach. However, instruments like
price shocks and quality of pre-school service providers proposed by other authors are not
available for us to use.
Anemia and Child Education: The Case of Colombia
66
of nutritional status on school attendance is that school attendance
is almost universal (96 percent of students in the sample attend to
school). As mentioned earlier, a child is overage either because of grade
repetition, non-attendance during one or more periods, late school
enrollment, or a combination of these problems. Results suggest that
anemia has an impact on the probability of being overage, but not on
school attendance. Thus, the impact of anemia on the probability of
being overage must be associated with either late enrollment or with
grade repetition. We are not able to distinguish between these two
potential explanations.
Anthropometric measures of nutrition, such as global and chronic
malnutrition, were used in order to compare and complement the
being overage of various anthropometric measures of nutrition. No
in the data used in this paper, the effect of anthropometric measures
is not apparent. Household and individual characteristics preserve the
13.
The different results obtained for the two types of nutrition measures
used in this paper (anemia and anthropometric variables) do not deny
the possibility of omitted variables bias. However, these results suggest
some interesting possibilities. As stated earlier, anemia seems to
increase the probability of being overage, but anthropometric measures
do not. This will occur, for example, if being overage is the result of
grade repetition, caused by poor cognitive development related in
measures could incorporate other characteristics such as genetic and
demographic background or ethnicity, and for this reason, it could not
sense, measures based on micronutrients (anemia for example) are
probably a better indicator of nutrition.
13 The sample was restricted to children with information of anemia in order to compare
the results of anemia and anthropometric measures. When the sample is not restricted, the
results are very similar.
Alejandro Gaviria y Alejandro Hoyos 67
Table 7. Regressions on the probability of being overage using anthro-
pometric measures of nutrition
Probit estimation
Dependent variable Overage
(1) (2) (3) (4)
Chronic 0.033 0.020 0.018
[0.024] [0.025] [0.024] [0.024]
Global 0.043** 0.034* 0.031
[0.019] [0.020] [0.020] [0.020]
Underweight 0.069** 0.006
[0.034] [0.022] [0.022] [0.019]
Individual controls Yes Yes Yes Yes
Mother controls Yes Yes Yes
Household controls Yes Yes
Yes Yes
Yes
Observations 2522 2048 2048
Clustered standard errors at the municipality level in brackets.
Table 8 presents the estimation results of a regression that combines
anemia and the anthropometric measures of nutrition. If the percep-
tion of parents about the nutritional status of their children is based
on a visual examination of weight and height, we could expect those
perceptions to be captured by the anthropometric measures. Hence
probability of being overage controlling for the psychical appearance
of the child. The results suggests that the effect of anemia on the
probability of being overage is not driven by a pessimistic parental
assessment of the children´s nutrition and health that in turn affect
their decisions about whether and when to send the child to school.
the one estimated in Table 5.
anthropometric measures of malnutrition. The percentage of anemic
malnutrition and being underweight, the corresponding percentages
-
lated with the other forms of malnutrition. Anthropometric measures,
however, might be distorted by genetic and ethnic characteristics of
the populations under analysis.
Anemia and Child Education: The Case of Colombia
68
Table 8. Regressions on the probability of being overage using anemia
and anthropometric measures of nutrition
Probit Estimation
Dependent Variable Overage
(1) (2) (3)
Anemia 0.046* 0.046**
[0.024] [0.024] [0.022]
Chronic 0.014
[0.024]
Global 0.029
[0.020]
Underweight 0.041
[0.033]
Individual Controls Yes Yes Yes
Mother Controls Yes Yes Yes
Household controls Yes Yes Yes
Yes Yes Yes
Yes Yes Yes
Observations 2049 2048 2334
Pseudo R-squared 0.098 0.100 0.104
Clustered standard errors at the municipality level in brackets.
Table 9. Comparison of anemia and anthropometric
measures of nutrition
Panel A
Chronic malnutrition?
% No Yes Total
Anemia? No 59.4 12.6
Yes 21.2 6.8 28.0
Total 80.6 19.4 100.0
Panel B
Global malnutrition?
% No Yes Total
Anemia? No 44.2
Yes 15.1 12.9 28.0
Total 59.3 100.0
Panel C
Underweight?
% No Yes Total
Anemia? No 68.3 4.3
Yes 25.2 2.2
Total 93.5 6.5 100.0
: D H S – 2005, E N S I N – 2005. Author’s calculations.
Alejandro Gaviria y Alejandro Hoyos 69
V. Robustness checks
a dichotomous variable that takes the value one if the years of schooling
are strictly less than the median of the population for the child’s same
-
ence between the years of schooling and the median of the relevant
population minus one. We refer to this variable as the schooling
gap. The rationale of this exercise is to give more variation to the
dependent variable and to verify whether or not the effect of anemia
on schooling remains after that. Table 10 presents the estimate of an
ordered Probit where the dependent variable is the schooling gap in
-
ability of not being overage. The results are almost the same as those
presented in Table 5.
As mentioned earlier, the correlation between schooling and nutrition
might be driven by the quantity and quality of social infrastructure
(hospitals, schools, sewer systems, etc). If this is so, the results will
overage children at the municipality level. A positive correlation is
apparent. To check whether or not the results are sensitive to omitted
municipal variables or city characteristics, we perform two different
of the country and the second to rural areas only; and second, we add
proxy for social infrastructure.
Table 11 reports the estimation results for the restricted samples. As
-
itan areas of the country (Medellín, Barranquilla, Bogotá, Cartagena,
Manizales, Montería, Neiva, Villavicencio, Pasto, Cúcuta, Pereira,
Bucaramanga and Cali). Because all these cities have a
similar level of social infrastructure, the problem of omitted municipal
variables is arguably less serious in this case. As shown, the magnitude
overage does not change substantially. Additionally we restrict the
larger than in the full sample. Probably some characteristics of rural
Anemia and Child Education: The Case of Colombia
70
areas, lack of infrastructure, for example, aggravate the supposed
impact of anemia on schooling outcomes.
Table 10. Estimation of the ordered probit model
Ordered probit estimation
Dependent
variable
No schooling gap (years)
(1) (2) (3) (4) (5) (6) (7) (8)
Anemia -0.051*** -0.048*** -0.044**
[0.022] [0.020] [0.020] [0.020]
Hemoglobin
(g/dL)
0.022*** 0.008 0.005
[0.008]
Male 0.005 -0.001 0.002 0.002 0.003 -0.002 0.001 0.002
[0.019] [0.020] [0.019] [0.019] [0.019] [0.020] [0.019] [0.019]
Age -0.061*** -0.056*** -0.056*** -0.061*** -0.056*** -0.056*** -0.056***
Child 0.021 -0.031 -0.021 -0.011 0.019 -0.034 -0.022 -0.012
[0.035] [0.042] [0.043] [0.026] [0.035] [0.043] [0.044]
Grandchild 0.032 -0.006 -0.003 0.008 0:031 -0.009 -0.004
[0.026] [0.040] [0.043] [0.041] [0.026] [0.041] [0.044] [0.042]
Mother's
education
0.004 0.003 0.004 0.004 0.003 0.004
[0.003] [0.003] [0.003] [0.003] [0.003] [0.003]
Mother works -0.009 -0.028 -0.031 -0.030
[0.019] [0.022] [0.021] [0.019] [0.022] [0.021]
Proportion of
child up to 5
years
-0.026 -0.034 -0.032 -0.040
Proportion of
workers
0.059 0.050 0.060 0.051
Number of
siblings
[0.005] [0.005] [0.005] [0.005]
Wealth quin-
effects
Yes Yes Yes Yes
effects Yes Yes
Observations 2965 2334 2334 2965 2334 2334
Pseudo R-
squared 0.054 0.063 0.068 0.054 0.060 0.065
Robust standard errors in brackets.
Alejandro Gaviria y Alejandro Hoyos 71
Figure 5. Anemia and proportion of overage children
rate by municipality
0
20
40
60
80
100
Proportion of overage children(%)
020406080 100
Anemia (%)
Table 11. Regressions on the probability of being
overage using different samples
Probit estimation
Dependent variable Overage
Sample: All 13 cities Rural
Anemia 0.053** 0.068* 0.088*
[0.022] [0.044] [0.053]
Individual controls Yes Yes Yes
Mother controls Yes Yes Yes
Household controls Yes Yes Yes
Yes Yes Yes
Observations 2334 411
Pseudo R-squared 0.094 0.151 0.122
Clustered standard errors at the municipality level in brackets.
The second exercise adds the sewage coverage rate to the original speci-
infrastructure, thus reducing the possibility of omitted variables bias.
The effect of anemia on the probability of being overage dropped to 4.1
Anemia and Child Education: The Case of Colombia
72
-
year old if he or she has not passed second grade, and so on. Table 12
presents the new results. The estimated effect is even larger than that
the probability of being overage by 4.6 percentage points; under the
Table 12. Regressions on the probability of being overage
(an alternative definition)
Probit estimation
Dependent variable Overage
(1) (2) (3) (4)
Anemia 0.163*** 0.116*** 0.084*** 0.051**
[0.032] [0.029] [0.028] [0.024]
Individual controls Yes Yes Yes Yes
Mother controls Yes Yes Yes
Household controls Yes Yes
Yes Yes
Yes
Observations 2635 2593 2593
Pseudo R-squared 0.088 0.286 0.314 0.324
Clustered standard errors at the municipality level in brackets.
VI. Conclusions
This paper provides evidence of the relationship between nutrition and
education in Colombia. The evidence suggests that anemic children are
more likely to be overage. Chronically malnourished and underweight
children, on the other hand, are not more likely to be overage. Nutri-
Despite the limitations of the database and the potential endogeneity
problems, the results passed all the robustness checks presented.
The results suggest that educational and nutritional policies in
-
cies. Interventions should include a preventive component, such as the
supplementation of micronutrients for children from 6 to 24 months
Alejandro Gaviria y Alejandro Hoyos 73
of age and for pregnant women. Probably, they should also include
Nutritional programs should be revised in order to incorporate effective
hunger. Even more, preventive strategies should be implemented from
pregnancy to early childhood. Educational outcomes should be part
that for malnourished populations, such as those living in many devel-
oping countries, the gains from investments in better nutrition may
be substantial - perhaps even greater than the gains from schooling
on educational outcomes in Colombia. However, the methodology
employed is susceptible to criticism. This is an observational study and
research is needed to both validate the results presented here and to study
the impact of nutrition on mental development, psychosocial outcomes
and long-term outcomes such as wages and labor productivity.
References
1.
analysis”, 36(1):185-205.
ASSIS, A. M., BARRETO, M. L, GOMES, G. S., PRADO, M. 2.
anemia prevalence and associated factors in Salvador, Bahia,
Brazil”, 20:1633-1641.
BARRO, R. (1996). 3. . Cambridge,
MA. Harvard University.
4.
preschool programs when length of exposure to the program
Anemia and Child Education: The Case of Colombia
74
varies: A nonparametric approach”,
86(1):108-132.
5.
achievement: Association, causality and household allocations”
University of Pennsylvania.
6.
determinants of child anthropometrics in Latin America: Back-
ground and overview of the symposium”,
2(3):335-351.
F” (Documento CEDE
de Economía, Universidad de los Andes.
8.
growth”, 20(3):423- 440.
9.
effect of health on economic growth: A production function
approach”, 32(1):1-13.
10.
Decreto 1944 de 1996. Ministerio de Salud Pública. República 11.
de Colombia. Octubre de 1996.
12.
prevalence of anaemia in the world”,
38:302-316.
Alejandro Gaviria y Alejandro Hoyos 75
-13.
ción. República de Colombia. Mayo de 1996.
14.
1(4): 140-169.
15.
nutrition, human capital and economic growth in Colombia
1995-2000” (Documento C E D E
Universidad de los Andes.
16. .
The University of Chicago. Third International Conference The
Economics of Health.
Children’s nutrition, school quality and primary school enroll-
ment in the Philippines (Working Paper Series Vol. 2006-24). The
International Centre for the Study of East Asian Development,
Kitakyushu.
18.
in children”, 131: 649S-668S.
19.
. Washington,
D. C., The World Bank.
20.
school improvement in economic development” (N B E R Working
Paper 12832). National Bureau of Economic Research.
21.
Anemia and Child Education: The Case of Colombia
76
22. School
.
Cambridge, MA., C A B International Publishing.
LOMBORG, B., (23. Ed.) (2006).
. Nueva York, Cambridge University
Press.
24.
QUISUMBING, A. R., MARTORELL, R. y STEIN, A. D.
(2006).
. Washington, D. C., I F P R I ,
mimeo.
Minnesota Population Center. Integrated Public Use Microdata 25.
Series — International: Version 4.0. Minneapolis: University of
of Statistics.
National Center for Health Statistics. 26.
– Growth Charts. Available at: http://
htm.
Profamilia (2005). Encuesta Nacional de Demografía y Salud.28.
Profamilia (2005). Encuesta Nacional de la Situación Nutricional 29.
en Colombia.
30.
nutrition and health: An integrated human capital approach”,
Alejandro Gaviria y Alejandro Hoyos 77
31.
32.
. Washington, D. C., International Nutritional Anemia
Consultative Group.
33.
economic growth”, The
122(3):1265-1306.
34.
sprinkles to control childhood anemia”, PLoS Med 2 (1): e1.
