Inter-American Development Bank
Banco Interamericano de Desarrollo
Latin American Research Network
Red de Centros de Investigación
Research Network Working Paper #R-492
Undernutrition in Bolivia:
Geography and Culture Matter
Ana María Aguilar
Cataloging-in-Publication data provided by the
Inter-American Development Bank
Felipe Herrera Library
Morales Anaya, Rolando.
Undernutrition in Bolivia: geography and culture matter / Rolando Morales, Ana María
Aguilar, Alvaro Calzadilla.
p. cm. (Research Network Working papers ; R-492)
Includes bibliographical references.
1. Malnutrition in children—Bolivia. 2. Malnutrition in children—Bolivia—Social aspects. I.
Aguilar, Ana María. II. Calzadilla, Alvaro. III. Inter-American Development Bank. Research
Dept. IV. Latin American Research Network. V. Title. VI. Series.
Inter-American Development Bank
1300 New York Avenue, N.W.
Washington, DC 20577
The views and interpretations in this document are those of the authors and should not be
attributed to the Inter-American Development Bank, or to any individual acting on its behalf.
The Research Department (RES) produces a quarterly newsletter, IDEA (Ideas for Development
in the Americas)
, as well as working papers and books on diverse economic issues. To obtain a
complete list of RES publications, and read or download them please visit our web site at:
3. Modeling Child Health
4. Building the Model: Step by Step
20 List of Tables
Table 1. Percentage of Children with Two Standard Deviations below the Median
Table 2. Descriptive Statistics of the Determinants of Child Nutrition
Table 3. Taxonomy of the Determinants of Child Nutrition
Table 4. Model 1. Estimation with the Control Variable (Age)
Table 5. Height and Weight Z-Scores (Means) by Mother’s Language
Table 6. Mother’s Height and Mother’s Years of Schooling, by Language
Table 7. Model 2: Adding Quechua Culture
Table 8. Model-3: Adding Altitude
Table 9. Model 4. Adding Mother’s Height
Table 10. Model 5. Adding Mother’s Education
Table 11. Relation between Access to Running Water and Access to Public Sewerage
Table 12. Model 6. Adding Household Assets and Public Services. SUR Estimates
Table 13. Estimation Summary
Table 14. Tests of Equality of the Coefficients in Both Equations
18 List of Figures
Figure 1. The “Age” Effect on Nutrition
Figure 2. Partial Correlations with Height and Weight Z-Scores,
The prevalence of health problems and malnutrition in Bolivia is shockingly high,
even relative to other developing countries. This study analyzes the association
between a bidimensional measure of child health—composed of height and
weight z-scores—and a set of child nutrition determinants related to physical and
cultural contexts, the mother’s characteristics, household assets and access to
public services. The paper seeks to identify the main determinants of child health
and to measure the impact of each factor related to the bidimensional indicator. A
sequential strategy is adopted in order to estimate a two-equation linear model
with correlated error terms. A major finding is that geographical and cultural
variables are significant determinants of nutritional status, and that the role of the
mother’s anthropometrical characteristics is substantial. This study uses data from
a Demographic and Health Survey (DHS) on over 3,000 children. JEL classification
: O12, I12, I38
: Child malnutrition, height z-score, weight z-score, SUR estimation.
1 The authors would like to thank Jere Behrman and Emanuel Skoufias for valuable comments that helped improve
an earlier version of this paper.
The aim of this study is to identify the main determinants of child health in Bolivia. The authors
suggest the adoption of a bidimensional anthropological measure as an indicator of child health.
This measure comprises the z-scores for height and weight in children aged less than 36 months.
Among the determinants of child health, the paper considers the physical and cultural context,
the mother’s characteristics, household assets and access to public services. The modeling
strategy involves the introduction of sets of variables from each category in a sequential
approach. In the initial steps, ordinary least squares (OLS) models were estimated for height and
weight z-scores. Thereafter, the authors applied an algorithm for the simultaneous estimation of
both equations, since they are correlated. 1.1 Undernutrition in Bolivia
Undernutrition is an acknowledged problem in Bolivia, although its complex causes are not
completely understood. Data from three DHS surveys in the decade 1989-1998 reveal an overall
downward trend in child undernutrition. Unfortunately, this trend has been accompanied by
increases in some deprived areas. In the department of Potosí, for example, the percentage of
children under two standard deviations for height-for-age increased from 33.2 percent in 1994 to
49.2 percent in 1998. Table 1 shows the movement of z-scores of height-for-age, weight-for-age
and weight-for-height indicators between 1994 and 1998. It also shows the differences in
nutritional status between males and females, and between urban and rural children. Table 1. Percentage of Children with Two Standard Deviations below the Median
(Z-scores. Children between 3 and 35 months old) Characteristics Height-for-age Weight-for-age Weight-for-height
94 98 94 98 94 98
28.2 27.1 16.5 9.9 5.5 2.1
28.3 24.0 15.2 9.0 3.2 1.4
20.9 18.3 11.6 6.1 3.3 1.3
36.6 35.6 20.4 14.1 5.6 2.4 Source:
Instituto Nacional de Estadística, Bolivia.
According to the 1998 Demographic and Health Survey (DHS), more than 25 percent of
children are below a –2 z-score for height-for-age; 50 percent had a z-score lower than –115; and
the inter-quartile range falls between –209 and –27. Few children have a height-for-age score
above zero. The same survey shows that 10 percent of children are below –2 standard deviations
for weight-for-age; 50 percent have a z-score lower than –53; and the inter-quartile range for this
indicator is located between –128 and 28. Nearly 40 percent of children had a weight-for-age z-
score above zero and in only 1.5 percent of cases was that score above two standard deviations.
The data indicate that height is more compromised than weight. 1.2 Geographical and Ethnic Characteristics
Bolivia is an atypical country in at least four ways. First, it is located between the Tropic of
Cancer (23.45 N latitude) and the Tropic of Capricorn (23.45 S latitude). It is thus classified as
tropical from the geographical perspective, although a significant proportion of its territory is not
tropical from an ecological standpoint. Second, the population is overwhelmingly concentrated
in high and cold areas where agricultural productivity is low. Third, Bolivia is the only country
in the Americas in which indigenous peoples (Aymaras, Quechuas and other ethnic groups) still
account for most of the population. Fourth, human settlements are highly concentrated in a few
cities and are dispersed in rural areas. In general, towns far from the main cities have insufficient
Bolivia’s climate varies widely, from tropical in the lowlands to glacial in the highest
parts of the Andes. Temperatures depend mainly on elevation and vary little by season. Most of
the northern part of the Western Andes rises to about 4,000 meters; the southern part is lower.
Rainfall is scant everywhere. In the southern region, precipitation is limited and useful
vegetation is sparse. Regions of the Western Andes are only lightly populated, and the south is
virtually uninhabited. The Cordillera Occidental
is a high desert with cold and windswept peaks.
The climate of the altiplano
, which is also swept by strong, cold winds, is arid and chilly.
Daily temperatures vary sharply and rainfall decreases with latitude from north to south.
Sheltered valleys and basins throughout the Eastern Andes have mild temperatures, with
moderate rainfall averaging 64-76 millimeters a year. Temperatures drop as elevation rises, and
snowfall is possible above 2,000 meters. The eastern slopes of the Central Andes descend
gradually in a complex and extensive series of flatlands and hills. Rivers, draining to the east, cut
long, narrow valleys that are favorable for crops and settlements. Rich alluvial soils fill the low
areas, but in some places the removal of vegetation has caused erosion.
The adverse geographical environment prompts problems of income-generation and
domestic food production. Inadequate and costly transport, a result of the country’s rocky terrain
and scattered population, severely hinders access to basic services and to the markets of the main
cities. The inhabitants of the Andes are familiar with food crises; in the altiplano
and the valleys,
recurrent cycles of drought, frost and hail affect crops and kill livestock. 2. Data 2.1 The Source of the Data
An extensive health assessment, the Bolivian DHS of 1998, is the main source of information for
this paper, especially its data on the height and weight of children under three. A second source
is the geographical database maintained by CIESS-Econometrica, which has useful information
on conditions at the municipal level. A third source is information on municipal-level services
and facilities, compiled by the Unidad de Análisis de Políticas Sociales y Económicas (UDAPE)
and the Viceministerio de Participación Popular.2
The data on height and weight were converted to z-scores. The correlation between the
height and weight z-scores is 0.66 and significant at the 0.0000 level. The effort to identify child
health-related variables focused on those that might explain the gap between the reference and
the observed distribution of the z-scores (for height or for weight). The large amount of
information available does not cover specific nutritional programs, access to food, regional
eating practices, household consumption or prices. 2.2 Determinants of Child Nutrition
Table 2 provides a statistical overview of the variables identified as the prime determinants of
child nutrition. The list was drawn up following several tests to discard other variables.
2 Estimating econometric models, the authors have shown that some variables in these sets were not significant in
explaining child health.
Table 2. Descriptive Statistics of the Determinants of Child Nutrition Variable Obs.MeanStd. Dev.Min.Max.
Age in months
Mother’s height z-score (multiplied
Mother’s years of education
Possession of refrigerator
Public sewer system
a In this and subsequent tables and figures, “talla” refers to height. 3. Modeling Child Health
The paper assumes that households have a utility function with health, leisure and consumption
as arguments, as explained by Behrman and Skoufias (2004). It adopts a bidimensional measure
of child health, <Y1,Y2>,where Y1 is the z-score for height and Y2 is the z-score for weight for
children under three. <Y1, Y2> are related to the set of variables listed in Table 2. The latter
variables belong to the groups outlined in Table 3. Table 3. Taxonomy of the Determinants of Child Nutrition
Household assets and access
to public services
Access to public
This study pursues a sequential strategy to identify the variables related to the
bidimensional measure of child health. In the first five steps of the sequence, ordinary least
squares are used to estimate each equation related to height and weight z-scores. In the sixth step,
the general framework is assumed under the umbrella of a simultaneous equation system with the
seemingly unrelated regression (SUR) option.
- Undernutrition in Bolivia:
- Geography and Culture Matter
- List of Tables
- List of Figures