t I.
. . 7..
.v. 3:...

on! . V I ‘v‘.
. a .53..
.fi 3?.“ $3..

 

» . 1w.

1.1“ 4.;‘1
. 3' .1...«....... 1 .
x... n... In}. 4.,

 

Mini-ta... .. , ,, . .134 . . ‘
. I ‘ . th5n 1 .L.....v~ .2“... ‘ ‘ , . .
. . . 5 . . , on . v. t,

.121. wagfim

 

1&7

 

l LlBRARY l

MiCthan State
University l

 

 

 

This is to certify that the
dissertation entitled

THREE ESSAYS ON THE ECONOMICS OF HEALTH

presented by

YLEANA PAMELA ORTIZ AREVALO

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Economics

 

 

4 6 0/4

 

r Professor’s Signature

W/S/JOO‘]
Date

MSU is an affinnative-action, equal-opportunity employer

 

----—--.-»—-.--i-.-.--—.—.—.--—.-.-.-.-—.--._.-«-.-.—.-.-.-.-u-.—c-I—‘u-o-n—o-u-«-.—--.-.—.—.--.-.-.—.----—

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATEDUE

DAIEDUE

DAIEDUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 pthIRC/DateDue indd-pt1

 

THREE ESSAYS ON THE ECONOMICS OF HEALTH

By

Yleana Pamela Ortiz Arevalo

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Doctor of Philosophy

Department of Economics

2007

ABSTRACT

THREE ESSAYS ON THE ECONOMICS OF HEALTH

By

Yleana Pamela Ortiz Arevalo

This thesis consists of three essays that investigate issues related to the economics of
health. All three studies examine the case of Mexico and draw on the first wave of the
Mexican Family Life Survey (2002). The first two essays focus on school age children

and the second on women.

In the first essay, I look for evidence of increasing overweight—evaluated in terms of
body mass index (BMI)—and examine possible contributing factors. I find that
prevalence of overweight increased for both boys and girls in recent years. To look for
factors that may contribute to this increase, I propose a theoretical model of a mother’s
allocation of food and time, from which I obtain her reduced-form demand for weight
adequacy of her child. I empirically estimate this demand measured as the probability that
a child is overweight. Results indicate that contributing factors to the growth of
overweight in children are: (i) increases in mother’s BMI and in community median BMI
of male adults; and (ii) particularly for boys, less presence of father at home and an

increase in maternal labor supply.

In the second essay, I study the effects of maternal education on hemoglobin levels

and possible consequences of anemia among children. Childhood anemia may impair

physical growth, weaken the ability to fight infections, affect the development of motor
and learning capacities, and reduce physical performance and endurance. All of these
have potential negative consequences on school functioning, accumulation of human
capital, and future labor productivity. Results point out to positive effects of maternal
education. I find evidence of maternal education being endogenous and possibly
contaminated with measurement error. There is evidence suggesting that maternal
education might be related to public health services through a substitution effect. Results

also reveal that mother’s labor supply is negatively associated with hemoglobin levels.

In the third essay, I examine whether sexual and reproductive behavior of Mexican
women changed due to public initiatives promoting birth control and family planning
launched in the 19705. I use a variety of indicators to characterize sexual and
reproductive behavior: first sexual encounter, first and second pregnancies, and
utilization of contraceptive methods. To identify the effects of public programs, I make
use of individual exposure to in-school sexual education and to birth control promotion in
public health facilities. Results suggest that in-school sex education increased the hazard
of first sexual intercourse but did not affect the hazard of first pregnancy. Both sex
education in schools and exposure to campaigns in public health facilities decreased the
hazard of a second pregnancy. There is also some evidence supporting positive effects of

sex education on the use of contraception methods.

Copyright by
YLEANA PAMELA ORTIZ AREVALO
2007

With my love to Jechu, Chicharito, Marifer, Lauri, Merve, and Alex.

ACKNOWLEDGEMENTS

I am very grateful to Professor Jeff Biddle for his guidance and dedication that made

possible this thesis. I also want to thank him for providing refreshing perspectives in his

course on History of Economic Thought.

Thanks to Professors Stephen Woodbury, John Giles, and Thomas DeLeire for their great

classes and their support during these years.

I greatly appreciate useful comments by Professor Scott Loveridge and his interest in my

research.

Thanks to Anita, Lourdes, Luciana, and Yayis for encouraging and lighting up my life.

Thanks to CONACYT for financially supporting my PhD. studies.

Thanks to the IN SP and the INNSZ for providing the databases of the ENN 1999 and of

the ENAL 1996.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES ........................................................................................................... xi
INTRODUCTION .............................................................................................................. 1
CHAPTER 1
OVERWEIGHT IN MEXICAN CHILDREN .................................................................... 4
1 .1. Introduction .................................................................................................................. 4
1.2. Background .................................................................................................................. 8
1.2.1. International dimension .................................................................................... 8
1.2.2. Definition and measurement ........................................................................... 11
1.2.3. Non-economic studies ..................................................................................... 12
1.3. Economic studies ....................................................................................................... 13
1.3.1. The role of maternal characteristics: education and labor supply ................... 13
1.3.2. Economic studies of child overweight and obesity ......................................... 14
1 .4. Theoretical model ...................................................................................................... 21
1.5. Empirical model, data, and results ............................................................................. 25
1.5.1. Empirical model and data ............................................................................... 25
1.5.2. Results ............................................................................................................. 41
1.6. Conclusions ................................................................................................................ 49
CHAPTER 2
THE EFFECTS OF MATERNAL SCHOOLING ON CHILD HEALTH: THE CASE OF
ANEMIA. ......................................................................................................................... 5 1
2. 1 . Introduction ................................................................................................................ 5 1
2.2. Conceptual framework ............................................................................................... 54
2.2.1. Hemoglobin levels as an indicator of nutrition and health ............................. 54
2.2.2. The role of maternal education in the production of child health outcomes .. 59
2.3. Empirical strategy ...................................................................................................... 65
2.3.1. Model .............................................................................................................. 65
2.3.2. Data ................................................................................................................. 71
2.4. Results ........................................................................................................................ 74
2.4.1. OLS ................................................................................................................. 74
2.4.2. ZSLS ................................................................................................................ 80
2.4.3. Interactions ...................................................................................................... 89
2.5. Conclusions ................................................................................................................ 91

vii

CHAPTER 3
DO PUBLIC INITIATIVES CHANGE SEXUAL AND REPRODUCTIVE BEHAVIOR

OF WOMEN? ................................................................................................................... 93
3.1. Introduction ................................................................................................................ 93
3.2. Background ................................................................................................................ 95
3.3. Literature Review ....................................................................................................... 99
3.4. Econometric Models ................................................................................................ 107

3.4.1. Duration-hazard models ................................................................................ 107

3.4.2 Multinomial logit model ................................................................................ 109
3.5. Data and empirical analysis ..................................................................................... 112

3.5.1 Data ................................................................................................................ 112

3.5.2 Empirical analysis .......................................................................................... 117
3.6. Results ...................................................................................................................... 121

3.6.1. Duration models ............................................................................................ 121

3.6.2 Multinomial logit models ............................................................................... 131
3.7. Conclusions .............................................................................................................. 137
CONCLUSIONS ............................................................................................................ 140
APPENDIX 1 .................................................................................................................. 144
APPENDIX 2 .............................. . .................................................................................... 146
APPENDIX 3 .................................................................................................................. 148
APPENDIX 4 .................................................................................................................. 157
REFERENCES ............................................................................................................... 164

viii

Table 1.1.

Table 1.2.

Table 1.3.

Table 1.4.

Table 1.5.

Table 2.1.

Table 2.2.

Table 2.3.

Table 2.4.

Table 2.5.

Table 2.6.

Table 2.7.

Table 2.8.

Table 3.1.

LIST OF TABLES

Previous studies analyzing the child obesity. .................................................. 16
Summary statistics. .......................................................................................... 38
OLS and two-stage least squares results for girls. ........................................... 42
OLS and two-stage least squares results for boys ............................................ 44
Identifying instruments first stage results. ....................................................... 48
Selected studies on the effects of maternal education since the 19905 ............ 60
Summary statistics ........................................................................................... 72
OLS Results. Dependent variable: hemoglobin levels. ................................... 76
Endogeneity and Overidentification Tests ....................................................... 82
Identifying instruments first-stage results. Dependent variable is education. . 83

ZSLS Results. Dependent variable: hemoglobin levels. .................................. 85
Education effects on child height, 2SLS results. ............................................. 89
Interactions results. Dependent variable: hemoglobin levels. ......................... 90
Previous studies analyzing sexual and reproductive behavior ....................... 100

ix

Table 3.2. Contraception methods use matrix. ............................................................... 111

Table 3.3. Summary Statistics. ....................................................................................... 115
Table 3.4. Mexican education system: elementary thru high school. ............................. 118
Table 3.5. Cox-proportional hazard models. Hazard ratio estimates .............................. 126
Table 3.6. Cox-proportional hazard models. Hazard ratio estimates. ............................. 129
Table 3.7. Multinomial logit: types of contraception methods. ...................................... 133

LIST OF FIGURES

Figure 1.1. Distribution of BMI in rural areas (school age children). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey
(MxFLS-l, 2002). ............................................................................................................. 27

Figure 1.2. Distribution of BMI in rural areas (school age girls). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey
(MxFLS-l, 2002). ............................................................................................................. 28

Figure 1.3. Distribution of BMI in rural areas (school age boys). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey
(MxFLS-l, 2002). ............................................................................................................. 29

Figure 1.4. Distribution of BMI national (school age children). Kernel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican
Family Life Survey (MxFLS-l, 2002). ............................................................................. 30

Figure 1.5. Distribution of BMI national (school age girls). Kernel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican
Family Life Survey (MxFLS-l, 2002). ............................................................................. 31

Figure 1.6. Distribution of BMI national (school age boys). Kemel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican
Family Life Survey (MxFLS-l, 2002). ............................................................................. 32

Figure 3.1. Timetable of major public initiatives to promote birth control and family
planning in Mexico. ...................................................................................... . ................... 98

Figure 3.2. Potential exposure to sexual education program at different education years
by cohort groups. ............................................................................................................ 118

Figure 3.3. Exposure to 1974-Program, average by birth cohort .................................... 119

xi

Figure 3.4. Sexual and reproductive behavior, average durations by birth cohort. ........ 119

Figure A2. 1. Prevalence of anemia among Mexican children by age groups. ............... 146
Figure A2.2. Prevalence of anemia among Mexican children by regions. ..................... 147
Figure A3.1. Education attainment in Mexico (1970 and 2005). ................................... 148

Figure A3.2. Attitudes towards role of women as mothers in Mexico: % who thinks a
woman needs to have children to be fulfilled by age groups. ......................................... 149

Figure A3.3. Attitudes towards women deciding to become single parents in Mexico: %
of approval by age groups. .............................................................................................. 150

Figure A3.4. Attitudes toward women wanting a home and children in Mexico (%). 151

Figure A3.5. Demand for contraception of Mexican users, 1976-1997. ........................ 152

xii

INTRODUCTION

The present dissertation contains three analyses of the economics of health. The first
two essays focus on school age children and the third on women in their childbearing
years. The dissertation links various topics in health and development. All three essays
use data drawn from the first wave of the Mexican. Family Life Survey collected in 2002

(MxFLS-l) '.

Health experts agree that developing countries are currently facing health and
nutrition transitions: overweight and obesity and non-communicable diseases are
emerging to join traditional problems of under-nutrition and communicable diseases. For
example, in Brazil, China, and Russia, 8%-11%2 of the households have both
underweight and overweight members (Doak et al., 2000). These concerns motivate the

3 among school age children.

first two chapters, in which I study overweight and anemia
Economically relevant aspects arise regarding the health of children. A child’s ill-health
may translate into low human capital accumulation and low labor productivity, and in the
context of developing countries, these could become serious constrains to development
and attempts to reduce inequality. When treating child health both economic and non-
economic studies tend to emphasize the analysis of parental characteristics since it is

generally believed that children, especially younger ones, play no role or at most a

minimum role in making decisions related to their lives (Haveman and Wolfe, 1995;

 

' For details see Appendix 4.

2 l 1% in Brazil and 8% in China and Russia.

3 Anemia is the deficiency of hemoglobin in blood, the protein in charge of transporting oxygen through the
body. Anerrria may be caused by deficiency of micronutrients, such as iron, folic acid, Vitamin B12, or
Vitamin A, but one of its main causes is iron deficiency. It may also be caused by non-nutritional factors
such as infections, e. g. worms.

Strauss and Thomas, 1995; Paxson and Waldfogel, 1999; and, Burton, Phipps and Curtis,

2002)

In the first chapter, I develop a theoretical model that provides some insights into the
relationships between maternal nutritional choices, in terms of monitoring time and food,
and child’s weight. I empirically test the model’s implications with particular attention on
the effects of maternal labor supply and time allocation. This analysis reveals no evidence
of maternal labor supply being correlated with the likelihood of children being
overweight. However, evidence suggest that other factors, such as child’s age, mother’s
BMI and education, and number of siblings at home, have effects on the probability of
overweight. Finally, results point to effects of community-level variables on the

likelihood of boys being overweight.

In the second chapter, I empirically estimate the effects of maternal education on
child health as measured by hemoglobin concentrations in blood. Since maternal
education could affect child’s health directly or indirectly through other factors, I analyze
how it interacts with household and community infrastructure and access to information.
Results suggest significant positive but small effects of maternal education on children
hemoglobin levels and that maternal education might be operating with public health

services (clinic/hospital) through a substitution effect.

The third chapter is motivated by the public initiatives promoting birth control and
family planning launched in the 19703 by the Mexican government in education and
health sectors. As in other countries, these public policies were introduced in order to

reduce the pressure of population growth and were effective in doing so. In the last three

decades, the average population growth rate dropped from 3.4%, in the period 1960-70,
to 1.0%, in the period 2000-05, while fertility rate fell from 5.7 in 1976 to 2.2 in 20064. I
examine how sexual and reproductive behavior of Mexican women changed due to the
exposure to sex education in schools and family planning in public health facilities. I find
that public initiatives had several effects on the behavior of Mexican women. I find that
not only exposure to such initiatives influenced changes on sexual and reproductive
behavior, but also that private schooling, completed education, and birth cohort play

important roles on changing behavior of Mexican women.

 

" National Institute of Geography, Statistics, and Informatics (INEGI).

CHAPTER 1
OVERWEIGHT IN MEXICAN CHILDREN

1.1. Introduction

Obesity is today regarded as a “pandemic” phenomenon: “[It] has reached epidemic
proportions globally. . .is a major contributor to the global burden of chronic disease and
disability. . .is a complex condition, with serious social and psychological dimensions,
affecting virtually all ages and socioeconomic groups.” 5 Overweight and obesity are
concerns in both developed and developing countries, particularly because of their
association with non-communicable chronic diseases (NCCDs). In developing countries,
these add up to join the traditional public health problems: malnutrition, under-nutrition,
and communicable diseases. Weight problems have also become a concern in the case of
younger population groups, e. g. in the United States, prevalence of child overweight and

obesity—starting in the 1980s—has been rising.

Recently evidence has been presented of a rise in the prevalence of overweight and
obesity in Mexico, particularly among women and preschool-age children. However, this
trend has not yet been documented rigorously nor approached from an economic
perspective, and the purpose of this paper is to do so. The trend is relevant for economists
due to its potential negative consequences—current and future—on labor productivity,
earnings capacity, and health systems and/or budgets. Public policies, such as food
assistance programs or available diets in schools, may also be affected. This paper
addresses the question of whether childhood overweight and obesity have really

increased in Mexico by examining body mass index (BMI) distributions at two points in

 

5 World Health Organization (WHO, 2003).

time, and looks for possible clues to the factors contributing to the increase by analyzing

cross section variation in current childhood obesity.

Child overweight and obesity, like other child outcomes, has been analyzed under the
assumption that decisions are made by the parents, not directly by the child". Both
economic and non-economic studies tend to emphasize the role of parental, household
and other environmental characteristics when exploring factors associated with child
overweight and obesity. Some ideas that have been proposed to explain trends of children
overweight and obesity in the US are that children follow eating and exercising patterns
of their parents, they watch more television, and are monitored less than before since

parents work more in the labor market (Philipson and Posner, 1999).

First, I propose a simple theoretical model to derive the mother’s behavior when
allocating food and time. This provides some insights into the relationships between
maternal nutritional choices, in terms of monitoring time and food, and child’s weight.
The mother is assumed to derive utility from food consumption, non-food consumption,
and the closeness of her child’s weight to what she considers the child’s “ideal” weight.
Her optimization is constrained by a budget constraint containing non-labor and labor
income, and a technology for the “closeness to ideal weight” as a function of food and
monitoring time. Two basic predictions are derived from this model: weight increases

with mother’s wages or labor market opportunities and is decreasing in food prices.

 

6 Note that Burton, Phipps and Curtis (2002) argue that the causality is usually taken as going from parents
to children, that the focus has been more on “production inputs” than on “production process”. As they
state, “More praise may encourage a child to do well, but parents may also praise more when a child does
well.” According to their theoretical model, parental behavior should not be taken as exogenous and the
empirical evidence they present supports this idea. Their results suggest that socioeconomic factors and
parenting style play a very important role in determining child behavior, and that parenting behavior is. in
turn, influenced by stresses in parents’ lives and by children’s behavior.

The empirical strategy gives particular attention to analyzing the relationship between
mother characteristics—such as education, labor supply, and time allocation—and child
weight. The underlying idea follows from the arguable7 assumption that children do not
make nutrition choices by themselves, at least not for the most part. In Mexico, children
are mostly cared for by their mothers or other family members; a market for childcare is
virtually non-existent. A first analysis focuses on mother characteristics without taking
into account whether the child is supervised by someone else. Additionally, I address the
question of how own, parental, household, community characteristics, and food prices
affect a school child’s weight, and how these characteristics contribute to the risk of a

child to becoming overweight or obese.

I next examine evidence of increasing overweight and obesity prevalence in Mexican
children. Unfortunately, large representative surveys have focused on collecting
information on preschool children. Nutrition and health surveys have been carried out in
Mexico since the 19703 but only in a few cases were school children consideredg, thus it
is difficult to even establish the existence of a trend for this age group. But it is most
likely that, like women and preschool children, school children experienced an increase

in their weight.

 

7 Variyam et al. (1999) provide evidence on maternal health and nutrition knowledge improving overall
dietary quality of preschoolers more than that of older children. This perhaps in part because the older the
child the more choices she or he makes on her/his own, like intake of food not supervised by the mother.
8 Only the Urban Nutrition Survey (Avila. et al., 2003), the ENN (1999), the National Health Survey
(ENSA, 2000), and the National Health and Nutrition Survey (ENSANUT, 2006) contain information on
nutritional status of school-age children.

The main cross-sectional analysis makes use of the first wave of the Mexican Family
Life Survey (MxFLS-l)9, which contains extensive information on socioeconomic,
demographic, and health characteristics at individual and household levels. The analysis
reveals evidence of mother’s working hours being correlated with the likelihood of being
overweight of boys; however, this correlation is small and weakly significant. Other
factors, such as child’s age, mother’s BMI and education, number of siblings at home,
and some community variables, are found to have important effects on the probability of
overweight. Number of siblings is negatively correlated with overweight, somehow
reflecting an overall effect of resource availability. Mother’s BMI has a positive effect
that might be representing both genetic and behavioral components. Mother’s education
is also positively correlated with child’s probability of being overweight. The presence of
father at home is positively and largely correlated with overweight, but not significant for
girls. This may be supporting evidence to the idea of fathers “favoring” their sons
comparatively to their daughters (e. g. see Thomas, 1994); however, it is difficult to
disregard the possibility that this result follows from a “role model” effect. The effect of

household income is statistically insignificant.

At the community level, effects are found for boys: median community BMI of 25-50
year-old men and median education (in years) of mothers are positive; while median
education of fathers is negative. This suggests that school boys are heavier, and more
likely overweight, the heavier adult males are in their community and the better educated
their mothers are. The opposite occurs if boys live in communities where fathers are

better educated—boys weigh less and are less likely to be overweight.

 

9 The MxFLS-l is a multi-topic survey that follows the design of the Indonesian Family Life Survey
(IFLS). It is representative at all national, regional, and urban-rural levels.

The chapter is organized as follows. Section 1.2 portrays some background on the
definition, measurement, and various current concerns about overweight and obesity.
Section 1.3 provides a brief appraisal of works on the role of maternal characteristics in
the production of child health outcomes and review of the economic literature on
childhood overweight and obesity. Section 1.4 presents the theoretical model. The

empirical strategy and results are described in section 1.5 and Section 1.6 concludes.

1.2. Background

1.2.1. International dimension

Overweight and obesity have become important concerns both in developed and
deve10ping countries due to their increasing prevalence and their association with as
important risk factors of non-communicable diseases. WHO sums up various facts about
obesity in the world: (i) more than 1 billion overweight adults—at least 300 million
obese; (ii) obesity and overweight represent a major risk for NCCDs—type 2 diabetes,
cardiovascular disease, hypertension and stroke, and certain forms of cancer; and, (iii)
that these are mostly consequences of an increase in the consumption of energy-dense
foods—high contents of saturated fats and sugars—and a decrease of physical activity”).
In the case of children, overweight and obesity have also been related to chronic health
problems, such as type II diabetes, high blood lipids, hypertension, sleep apnea,
orthopedic problems, early malnutrition, and a higher risk of becoming overweight or

obese when adult (Anderson, Butcher and Levine, 2003).

 

'0 WHO (2003).

In the United States, using the National Health and Nutrition Examination Surveys
(NHANES) to estimate overweight and obesity prevalence rates, the Centers for Control
Disease and Prevention (CDC)ll report that, between 1976 and 2002, combined
overweight and obesity prevalence increased by 38% (from 47% to 65%) and obesity
doubled (from 15% to 31%). In the case of children, evidence from the NHANES reveals
that between the 19605 and 1980 rates of overweight children (6-11 years old) and
adolescents (12-19 years old) remained relatively stable. But, an important rise has been
observed since the early 805. Overweight was 7% for children and 5% for adolescents, in
1976-80; 11% for both groups, in 1988-94; and 16% for both groups, in 1999-02. Recent
estimates indicate that Europeans12 have a high average BMI (26.5). Prevalence of
obesity ranges from 5% to 20% in men and up to 30% in women among European
countries. Even countries that traditionally had low rates of overweight—such as France,
Norway, and Netherlands—have experienced rises in adult obesity during the last decade.
Evidence also suggests that 10—30% of the European children (aged 7—11 years) and 8—

25% of adolescents (14—17 years) are overweight.

Experts maintain that developing countries are currently facing health and nutrition
transitions: under-nutrition coexisting with overweight and obesity, and the association of
the latter with a rise in NCCDs, i.e. that overweight, obesity, and NCCDs are emerging to
join and combine with the traditional health problems of developing countries”. Doak et

al. (2000) use data from large national surveys in Brazil, China, and Russia. They find

 

” Centers for Disease Control and Prevention: Prevalence of Overweight among Adults.
http:/wawxcdc.gov/richsr’productsfpubs‘pubd--’lrcstats./obese/obse99.htm. and Prevalence of Overweight
among Children and Adolescents. h_t_tp:.f/wwwxcdcgovinchjfproducts"pubs/pubd/hcstatsx’o\t'erwght99.htm
(accessed April 18, 2006).

'2 WHO (2005 ).

'3 Zambrano and Guzman (2002); Popkin, Richards and Monteiro (1996); and, Hoffman (2001).

that the proportions of households with both underweight and overweight members were
11% in Brazil and 8% in China and Russia; and, that these households account for 45%,
23% and 58% respectively, of all households with an underweight member. Rivera and
Sepr'rlveda (2003) present results from Mexico revealing that between 1988 and 1999 for
women 18-49 years old overweight prevalence increased by 46%, (from 24% to 35%)
and obesity in this group increased by 160% (from 9% to 24%). They also find that

prevalence of obesity in preschool children (under 5 years old) grew by 26%.

Commonly, nutritional and health transitions observed in most countries in Asia,
Latin America, Northern Africa, Middle East and urban areas of sub-Saharan Africa,
have been related to changes in the population’s lifestyles, particularly changes in diet
and physical activity: increased consumption of fat, added sugar, and animal products,
i.e. high fat/high carbohydrate energy-dense foods, in contrast to lower total intakes of
cereal, fiber, fruits and vegetables (Uauy, Albala and Kain, 2001 and Popkin, 2001); and
reduced physical activity as a consequence of more sedentary work and leisure—driven
by technological changes (Rivera et al., 2004). Despite the fact that such changes in
lifestyles have also been observed in developed countries”, according to WHO (2004),
developing countries are those currently more and increasingly affected by this problem
in terms of mortality, morbidity, and disability—where 66% of deaths relate to non-
communicable diseases and on average involve relatively younger people than in

developed countries.

 

'4 See below for an exposition of studies on the causes of obesity trends in the US.

10

1.2.2. Definition and measurement

Obesity is defined as a condition of abnormal or excessive fat accumulation in
adipose tissue or as a disease in which excess body fat has accumulated to such an extent
that health may be adversely affected or impaired (WHO, 2000). The most widespread
technique to evaluate weight adequacy is the body mass index (BMI)15 . The BMI is
defined as weight (in kilograms) divided by height squared (in meters). A simple
classification for adults is the following: overweight if BMI 2 25; preobese if 25 < BMI_>_
29.99, with an increased risk of comorbidities; obese class I if 30 5 BMI 2 34.99, with a
moderate risk of comorbidities; obese class 11 if 35 5 BMI 2 39.99, with a severe risk of
comorbidities; and, obese class 111 if BMI Z 40, with a very severe risk of
comorbidities’b. The BMI has been usually used for adult populations, but it has been

adopted for children populations as well (Rolland-Cachera et al., 1991).

BMI was not adopted until recently to evaluate weight adequacy of children
populations. Unlike case of adults, there is not broad agreement for defining overweight
and obesity cut-offs. In part, this is because BMI changes considerably with age.
However, researchers in different countries have proposed some reference options. In the
United States the 85th and 95th centiles of age and sex based BMI distributions are used as
reference points (Must, Dallal and Dietz, 1991). In France, values of cut-offs are
identified by using Z—scores (Rolland-Cachera et al., 1991 ). Cole et a1. (2000) established

an age and sex based standard definition for child overweight and obesity using

 

'5 Note that BMI does not really measure body fatness, but body mass or weight adequacy to height.

'6 Notice that this classification is age and gender-independent and that WHO acknowledges that “. . .the
relationship between BMI and body fat content varies according to body build and proportion, and. . .a
given BMI may not correspond to the same degree of fatness across populations”.

11

nationally representative cross sectional growth studies from Brazil, Great Britain, Hong
Kong, the Netherlands, Singapore, and the US, which makes it comparable

internationally, and can be used to identify changes over time.

1.2.3. Non-economic studies

Both economic and non-economic studies tend to emphasize the role of parental,
household and other environmental characteristics when exploring factors associated with
child overweight and obesity. For instance, Hernandez et al. (2003) find that risks of
overweight and obesity in Mexican school children were positively associated with
mother’s education and socioeconomic status. Additionally, they find that risk is also
positively associated with age and that prevalence is higher for girls, but found no
significant differences between regions or urban and rural areas. Padron (2002) argues
that considering environmental, social, and genetic factors separately cannot explain the
trends observed in the last decades. She points to an excess of sedentary aetivities, such
as watching television, videogarne and computer use, and to a reduction in relative prices
of low nutritious-quality food. She also suggests that children should not be rewarded (by
parents) with food and that healthy food should not be substituted with snacks. A study
among children in Mexico City by Hernandez et al. (1999), using a random sample of
schools from two communities (delegaciones), reveals that risk (odds ratio) of obesity
was higher by 12% for each additional hour per day of watching television, 10% lower
for each hour per day of moderate or vigorous physical activity, and 20% lower
considering only vigorous activity. Brewis (2003) investigates obesity among middle-

class children from a school in the city of Xalapa, finding that children from households

12

with fewer children, those with more permissive, less authoritarian parents, and boys
were more likely to be obese. She argues that food treats are a cultural index of parental
caring, and as a consequence can result in indulgent feeding especially for sons in smaller

middle-class families.

1.3. Economic studies

1.3.1. The role of maternal characteristics: education and labor supply

There is a large literature on the effects of parental/household characteristics on child
health outcomes. Here I concentrate on some works that analyze maternal labor supply
and education. Thomas et al. (1991) and Glewwe (1999) study possible mechanisms
through which mother’s education affects child’s health in terms of height. Thomas et al.
(1991), in Brazil, find evidence of access to information being such a mechanism; while
Glewwe, in Morocco, finds that it is health knowledge acquired through literacy and
numeracy skills learned in school. Kassouf and Senauer (1996) analyze parental
education direct and indirect effects via wages and full income on three anthropometric
measures for Brazilian children and, using regression results, they simulate effects by
education levels. In particular, they find positive direct effects of the role of mother’s
education, negative small indirect wage effects, and large positive indirect full income
effects on weight for age and weight for health. Total effects are positive and larger as
education increases. Variyam et al. (1999) study the effect of maternal nutrition and
health knowledge on dietary intakes of children 2-17 years of age in the US. They find
that maternal knowledge influences children’s diet especially for preschool children, but

that this effect becomes weaker for older children (> 5 years old). They also include

13

mother's labor participation status (not, part, and full employed), finding that total and
saturated fat intakes decrease if mother is not employed or part-time employed as
compared to being full-employed, only for school-age children. Effects of employment
status for preschoolers seemingly occur through maternal nutrition and health knowledge.
Blau, Guilkey and Popkin (1996) examine the relationship between mother’s labor
supply, infant feeding, and infant health. They find that after accounting for possible
endogeneity there is little evidence of direct causal effects of mother’s labor supply on
weight and height of infants in the Philippines. They also find that mothers with higher
wages have healthier children and that mothers who face higher food (infant formula and
corn) prices have less healthy children. Glick and Sahn (1998) investigate the effect of
mother’s labor supply and income in preschool children of Guinea, as Blau et al. (1996),
accounting for possible labor supply endogeneity. They find that on average positive

effects of labor income on children height are offset by large negative effects of labor

supply.

1.3.2. Economic studies of child overweight and obesity

Works studying adult obesity in the US have focused on distinguishing the effects of
different factors on the trends of adult obesity including technological change (TCH) and
its impact on food prices and more sedentary home and market work (Philipson and
Posner, 1999 and Lakdawalla and Philipson, 2002); change in lifestyle and its association
with changes in value of time, especially for women; increased real cost of cigarette
smoking; more availability of fast food——which reduce costs of search and travel and

relative cost of fast food meals (Chou, Grossman and Saffer, 2002; Rashad, Grossman

l4

and Chou, 2005); and, decrease in time cost of food—due to a rise in mass preparation of
food, permitting the production of food in one location and its consumption in another—
with an effect on the frequency of food consumption of greater variety (Cutler, Glaeser

and Shapiro, 2003).

Economic literature on children, as in the case of adults, mainly addresses the
question of possible causes of the rise in overweight and obesity prevalence. In the
United States, existing works are few and recent; the list would include Anderson et al.
(2002); Anderson et al. (2003); Ruhm (2004); Anderson and Butcher (2005); and,

Anderson and Butcher (2006). Table 1.1 presents a brief synopsis of these works.

Anderson et al. (2002; 2003) study whether the rise in the prevalence of working
mothers observed during the last decades has had a causal effect on the prevalence of
overweight children. In particular, they find a positive relationship between the average
hours per week worked over a child’s lifetime and his/her probability of being
overweight—e. g. changing from a part (20 hours per week) to a full-time (40 hours per
week) employment would cause an increase by 1 to 2 percentage points on such

probability.

15

 

.mEoZBQ bmocowouco 036mg

com 35:00 8 E88 E 29%? 308333
-Smoa a mo comma—05 2F .CcoEmmommm
-508 SA: moms Baa can >28:on
88mg meson—2: .mowm EEQ Sesame

6 x33 com 850: cwwcoaw 2w meoEccow
SEES—mam .znoa 98 83:?

~32 >85ch ”333.508 oEoEocoom
.wom: m_ 32:33 :EBE: mo seepage

an £85053“ 9% 8 26 .mmmbmca com
62.650 mm @605. 532695 was $2.5
mo :oaoefi vcmfiow "586832 a £053
80¢ 32:85:00 0602: can 2:: 8
Samoa 53> 226:3 2032689880330
250 282% a mo 8:38:me

56:85 5 x83 com 2:0:

98 So» 5d 383 EoE:m%-Eoumm
085$: FREQ ao>o can end: 923:0
memo» m 65 macaw JESS ”ScoEIAo—mfic
mo mcoumecocofii .mozatg _ScoEEHmE

450$:me 333383 388 can
33QO ABE—Eu 393533 0.58
wcofim case.“ 0.8 x83 ESE
559: we mecca—63:8 03332
.3305

manage e E2 8:253 e eee 883%: >32
.mocmcouomaso 238323 358m
.8 fisEZwE £50: wept?» me
.8855 cozosuoa m we 33on
can 38880 230 descended
couwfiemxmfi a an BZSQ

.AmEmom owficoocom

S l we 32 ea $2 5238
bozmcflfi 808 watt?» 20:88
E 852on on :8 22220
332596 mo coco—«>05 E £38m

932 .eiae 558
396365 E moxflfi teem
mo zogm 9:25:00 2: was

2.02.”? .E mmzézv
autism cosmEwam

Aeoomv
Essa

Amoomvaeae

 

can .moococoecmc wEBfi £503 32030 2: («o :3 :95 a 3:0 newcouexx
”com—mam 08 323538 oEoEOCOoo 85H .Ewmozgo use 3003 «so. a Womwgswfimmcwwflwm
.fioBoEEm :Eflwo—ommb—Q: 98 28988 5a mason owmcogv “583380 ._A .H .2 5 . n a N6 3w w
$3338 a Set 32.5w wag—owes: 8:88 5952— 9520522 32 £2 >mq 5 > m
. . . . . 1 . autism _mfigéweoq 3:0th
oEoEQSom $on 285885 oz 338 bammmom was o>Emom .
$5500 338% _moEmEE—moueoosh Smeoafimm cofis<

.bfiopo 230 05 mafia—dew 36:5 3255 A; 2an

16

 

550:0 0: 8:08:88

80:00 000: 80:00 :0 000: 0:: 80::
030:0: 30:00.0 808080 8: n003:00

m: _0008 00000805 02 0038:? 5:00
000: :0 b:0:0w00:0 00:80:00 :83 _000

0: 00:0: 8 8: 8808800 0:00 0: 0:0 0:00
0: 0800080 8:: 388000”: 80008000
_008080 :0: >: 2080003: :0 :08

0: 0:87. :000:000 000.003: 808308:

.088: 000: :0000 Noam 80:: 80:00:
08088000 a88:50:80 w8:00800 In:
00:0: :35 :0 080030 _008080 080080
.00>::00 8 _0008 00000805 02 0:00:00
:0 0:0: 0:: 0:0 8:0: 0:08 88:03
80:88 :0 0:0: 0:: 0:0 .mD 0:: 8 00:00.
0:: :0 3030: 0 n560:0 :0 :0:00:0:0 83:0
0:: ”00:08 _0:0>0m :0 808800 0 000380
:: .:000:000 0:80:80 :0 80:: b:m0:0
00:0 80:0 8:808 m:0>00 :0000 08%

580:0 0w0:0>0
.08080—000 8 0000:08 0:: :0 Xeom
:0: 800000 0: 0:00: 0:0 0:00:00

8 0000: 088. 50380000: 00850 0:00 0:00 :08800
8:8 m.m 0:0 m.: 0: 80:03:00 AmmOZ: m0::m::0:m :000000m Goom:
0:0 000:H Afifld 8 000::0 :0: 088:0:m 8:00:01”: :0:0::& 0:0
000:3: 800:0 :0: 880008080 :0: :0800 080002 Soon 0:0 8800:»:

2050 e 5:3 0:88:00 08: 0:80 :32: >82
3 520 a: 05 as .05.: 0:38
:0 _00:0m :0 000: 080.5 0: 000000
:0 000::0 0w0:0>0 088000 08 0

0:800:00 0: 8880800 0: 080

as 2808 a 08: 028:3 :e 808 20 08:
0:50:80 see 05 0:02.02: .880: :8: game:
a: 0000808 0:800:00 :50 0:: 0:0 080200 80002 _00:0m
:0 :2: 20 as? .3 00888: :2: $8: 0:0 03: .0802 280 ._e e
:8008 0:: 00:3 S03 0800 0:: 8 -m0UV >0>8m :0008000 880050

09800 sea 88: 2-0 800.0 22.50 602.0020 >002
=e 82 .23: a: eases a eewee 880-80: 05 ”8-002

800:0 .303: 0:0 3.43: “3.0.3: €0-:3$ mmZ<EZ
8 00:00 :0080 8:88:80 :35

 

0808800

0:003: 0080890000805. 0:0Q\0_080m :0:8<
.528: .2 030:

17

 

0:00:00 :0 20: 0::
0:0 .0:00 00:0 0:0 _00:00 80:80:80

88030—00.
:0 8000M 0000002 5882

.mD 0::

.mD 0:: 8 00>:00:0 00:0:: :0 8080:»: _00::0_:0:m 80::

 

:8: 8 00w:0:U £5 .Aw:_000::000:: :0 000000 3:00:00 :0 :0_:0880xm_ 00:0: 8000:8800 00:0: 8:01: Good

00 00:20:00 _0:0:::000 0:0 6880:0008 .mD 0:00 :0:0:0m 0:0

a0:0:8: 08030 00:80: 3:08,: 0:: 8 0:0 a:0:8:0::0:8 00:0:: :0::0800:00 :88 ::00 m3 :00:00:<
a: 00:2: 0:0 000388020800 0:38 :o 00020 3000080 .8833: 0:0 00-002
0030 3 :5 0325: 0:203 0038: 50-03: 0053::
007202 .030: 0:::
58030 :0 REES/8Q Go .0wa

3:082:00 0::—000% _mom:_mem\~mofio:oo£rw Sago—QFEW .55st

A0800: . _ . _ 030%

18

However, the magnitude of the estimated effects of maternal employment only
account for 6 to 11 percent of the observed growth in the prevalence of overweight
children. Additionally, using their estimates the increase in adult women overweight and
obesity can “explain” 11 percent of the phenomenon.17 Anderson et al. (2003) analyze
child obesity from a broader economic perspective, though they devote specific attention
to the role of working mothers. They discuss several caveats regarding the particularities
of the child overweight and obesity issue, from the pertinence of using a standard
economic model, to the role of parental supervision, schools, fast food and television, to

the public policy implications due to social health costs and extemalities.

Maternal employment is also analyzed by Ruhm (2004), who assesses the question of
how maternal employment relates to various developmental outcomes of 10 and 11 year-
old children—cognitive development, obesity, and potential risky behaviors. He finds
that 20 additional hours per week are associated with a 22 to 27 percent rise in obesity
and a 19 to 20 percent increase in the risk of being overweight. Anderson and Butcher
(2005) study the role of schools, arguing that schools facing financial pressures fund
themselves through permitting more access to “junk” foods. They look at the likelihood
of schools under financial difficulties implementing “potentially unhealthful food
policies”. They find a positive effect of access to junk food at school on obesity: if the
proportion of schools allowing junk food in a county increases by 10 percentage points,
BMI increases by 1% on average—equivalent to 1.5 extra pounds. But they notice that
this is driven by adolescents with a genetic predisposition for obesity—an effect of 2.2%

or 3.3 extra pounds. Using these figures, they estimate that junk foods in schools are

 

‘7 Notice that even if taken together, these aspects fail to explain the largest part of the observed trend
observed in the US.

19

found to account for 20% of the observed increase in adolescents’ average obesity.
Anderson and Butcher (2006) provide a descriptive analysis of childhood obesity trends
and potential causes in the US. They assess the definition of obesity and emphasize the
validity of using the BMI for measuring children obesity. They review the energy balance
literature (intake, expenditure, and additional correlates as breastfeeding) to address
possible causes of childhood obesity and look at the consistency between the time
patterns of obesity and its “causes”. They present evidence pointing out to, on the side of
energy intake, increased availability of (calorie dense) convenience foods and soft drinks
and increased demand for away-from-home or pre-prepared foods, due to more time
devoted to market work by parents. On the side of expenditure and also related to less
available monitoring time by parents, they identify a more sedentary children’s

lifestyle—i.e. lighter physical activity practices.

Authors use a variety of techniques to overcome biases that may arise because of
unobservables being correlated with maternal employment. Although they try to control
for as many factors as possible, problems such as heterogeneity or endogeneity are
always a concern. Anderson et al. (2002; 2003) use sibling difference models to reduce
the bias due to unobserved heterogeneity. They compare siblings in two ways: the
different ages, using same survey year; and, at same age, using different survey years. To
correct potential bias as a consequence of other unobserved variables, they apply an
instrumental variable procedure for maternal employment. Ruhm (2004) includes a
variable of maternal employment measured at a period following that under study, i.e.
maternal employment after children are 10-11 year-old, to account for possible reverse

causation that might bias his estimates. The interpretation for doing so is that if

20

coefficients of post-assessment employment are different from zero, then not only do
maternal labor decisions affect children development—e. g. their risk of being
overweight—but also that mothers decide whether to work or not, and how many hours,

depending on their children’s situation.

1.4. Theoretical model

To grasp how a child’s weight is influenced by the decisions of his or her mother, I
propose to model her behavior when allocating food and time. She allocates food and
time by maximizing a utility fiinction that depends on food consumption (F), non-food
consumption (X), and the difference between her child’s actual weight and what she
considers an “ideal” weight for her child (AB). The mother is assumed to be her
children’s only caregiver and that she successfully manages her children into consuming
nutritious food, i.e. maternal monitoring time is effective for improving her child’s food

intake. 1 8

Her optimization is constrained by a time and money-budget constraint and by a
technology or production function determining the “closeness to ideal weight”. Also, I
assume that children enjoy food but they are not concerned with nutrition or health

issues. Thus, children must be monitored for them to eat a more nutritious and healthy

 

'8 The model and its assumptions are largely compatible with some results on eating behavior found by
psychologists and child development experts. For example, there is evidence indicating that parental
monitoring and even the threat of it influences food selection of children (4-7 years old): reducing choices
of non-nutritious foods and the caloric intake of the meal (Klesges et al., 1991). Another piece of evidence
on 5 years-old girls supports that “parenting practices are shaped and influenced by the child. ..
[Specifically] that child-feeding practices are influenced by the weight status of the child: mother’s child-
feeding practices and perceptions of their daughters’ risk of overweight are found to have effects on her
daughters’ eating and relative weight—effects that comparable to those of maternal weight (Birch and
Fisher, 2000).

21

diet. Price of the non-food good is assumed to be one. The mother has a utility function

given by:
max U (X,F,AB) = u[X,F] + V[AB]

s.t. Y+w(T—t)—pF—X=O (1.1)
AB =| BMI — BMI* 1: f(t,F) .—. flF[1—%)

where Y is non-labor income, w is the market wage, T is total time, t is monitoring
time, p is food price, and price of X is 1. BMI is child’s actual body mass index, and

BMI“ is mother’s ideal body mass index standard. Assumptions about fiinctional form of

the mother’s utility are the following: “X > 0 ,uF > 0 WM < 0 ’uFF < O and
u XF = u FX > O . The mother experiences disutility if their child’s weight differs from

her ideal standard, so 12' <0, v">0. Mother’s technology of weight adequacy—with
respect to her ideal standard |BMI-BMI*|—is assumed to be linear in all arguments,
increasing in F and decreasing in t. I assume standard BMI“ is fixed and determined by
mother’s preferences, so it may or may not coincide with experts’ recommendations. For
simplicity, this function is defined only for the case of BMI >BMI*, i.e. for the case in

which mothers consider their children to be overweight. In other words, more food in the

household would contribute to increase actual child’s BMI ( f F > 0 )—moving away

22

from BMI*——and more time devoted to monitoring would decrease it ( ft < O )—

approaching towards BMI“ . 19

To solve (1.1), the mother allocates t, consumes X, and allocates and consume F. The

first-order conditions are:

v'ft —7tw=0
11F +v'fF —}tp=0 (1.2)
uX —)\.=O

Y+w(T—t)—pF——X=O

where k is the Lagrangean multiplier. To identify effects of wage, food price, and
income on child’s weight, I totally differentiate expressions in (1.2) and constraint AB
(see Appendix 1). Then I use the results to assess: (i) how higher wages and thus market
opportunities affect children’s weight and (ii) what effect a change in food price has on

weight.

dAB
and
dw dp

, and it can be shown that (see Appendix 1):

 

 

Formally, these are

 

'9 Note that if the child is underweight more of both food and monitoring time would increase BMI, making
it approach towards BMI*.

23

dAB
“_ziftiptt +fF¢in+(T—t)[ftwt +waF]

dw

dAB

“Eb—=[fz90tF +fF€0FFi—Fifzwz +waFi (1-3)
(1218

71—},— = ft“): + waF

where (art , (p FF , and got}: = (0 Ft are own- and cross-compensated price effects

(substitution effects), a), and a) F are income effects, and ft and f F are the partial
derivatives of AB with respect to monitoring time and food. Own-compensated effects are

negative and as stated above ft < 0 and f F > 0. If it is assumed that U tF > 0 and

“small enough”, together with the other functional assumptions of U , then

(”tF , 90 Ft > 0 (see Appendix 1). F is assumed to be a normal good, so a) F > 0 ; and, as

long as AB is a normal good, t is also “normal”, so cut > O .

Under these conditions, signs of expressions in (1.3) cannot be identified accurately

unless further assumptions about “total income” effect, ? = ftwt + f F a) F , are

be either positive or negative but assume (T — t )[ft a), + f F a) F ]

 

stated. Let

24

and — F [ f t cot + f F a) F ] are relatively small with respect to price effects, so that it

can be shown that20

14—8—>0 and dAB

< 0 1.3’
dw dp ( )

 

According to these predictions, the model suggests that an increase in wages for
women, which may be interpreted as increasing market opportunities for mothers and
higher labor force participation, have a positive effect on children’s weight (AB). It also
suggests a negative relationship between food prices and weight, implying that a decrease
in food prices that may translate into reduced price of calories would contribute to

increasing children’s weight.

1.5. Empirical model, data, and results

1.5.1. Empirical model and data

Ideally the empirical strategy would aim to answer whether an increasing trend in
overweight and obesity exist in Mexico and what its causes are. Repeated cross sections
or panel data would be needed for accomplishing these tasks. However, such data are not
available, so I instead focus on looking at possible factors that might be contributing to

the alleged rise in obesity in school-age children by analyzing cross-section variation.

Evidence supports the idea of the existence of an increasing trend. Prevalence of

obesity in preschool children increased from 4.2% in 1988 to 5.3% in 1999; combined

 

2” Note that the same results are obtained if the total income effect is assumed to be positive, which would
imply that jfia),|<fpa)p is assumed.

25

prevalence of overweight and obesity of women (18-49 years old) was 33.4% in 1988
and 59.6% in 1999, and prevalence of obesity increased from 9% to 24%”. Data on
school age children are not available for 1988. Using data from the National Survey of
Food Consumption and Nutrition in Rural Areas 1996 (ENAL, 1996)”, from Rivera et al.
(2001), and from the rural sub-sample of the MxFLS-l, I calculate the combined
prevalence of overweight and obesity of school age children: 15.5%, in 1996; 19.7%, in
1999; and, 18.5% in 200223. Finally, at a national level, prevalence of overweight school-
aged children was 19.5%, in 1999, and 23.2%, in 200224. Overall, these estimates suggest
that there has been an increase in the prevalence of overweight and obesity in Mexico for

both female adults and children.

Comparisons between school children BMI distributions provide additional
supporting evidence. I analyze BMI distributions in rural areas (Figures 1.1-1.3) and
national BMI distributions (Figures 1.4-1.6).25 Rural data are drawn from the ENAL
(1996) and national level data are drawn from the National Nutrition Survey (ENN,
1999)26 and from data from the MxFLS-l. The vertical line (approximately) indicates the
minimum age—sex reference threshold above which a child may be considered

overweight: BMI=17.15 for girls and BMI=17.42 for boys (Cole et al., 2000).One can

 

2‘ Rivera et al. (2004) using National Nutrition Surveys (ENN, 1988 and ENN, 1999).

3: Encuesta Nacional de A limentacio'n y Nutricién en el Media Rural collected by the Instituto Nacional de
la Nutricio'n Salvador Zubiran. For details on this survey see Avila et al. (1997).

23 Estimate for 1999 was calculated using US population as standard reference (Rivera et al., 2001). For
1996 and 2002 estimates, I used Cole et al. (2000) as standard reference. Thus, comparisons should then be
taken with caution.

2’ Both estimates were calculated using Cole et al. (2002) standard references. 1999 figure was taken from
Hernandez et al. (2003).

25 Rural baseline is drawn from the ENAL (1996). National baseline is drawn from ENN (1999). Data of
2002 are drawn from MxFLS-l. The vertical line (approximately) indicates the minimum age-sex reference
threshold above which a child may be considered overweight: BMI=17.15 for girls and BMI=17.42 for
boys (Cole et al., 2000).

26 Encuesta Nacional de Nutricio'n collected by the Instituto Nacional de Salud Pziblica. For details on this
survey see Rivera et al. (2001).

26

observe that the rise in prevalence discussed above derives from a shift of the
distributions to the right. In the case of rural areas, this shift is particularly important at
the left tail and the center part of the distribution, suggesting that prevalence in
overweight and obesity are rising not because the density of the right tail is increasing but

that lower and medium levels of BMI are increasing.

Distribution of BMI 5-11 years old
Rural Areas (1996 and 2002)

 

 

 

 

 

T I l I l
10 15 20 25 30
BMI

MxF LS (2002) — — — ENAL (1996)

Figure 1.1. Distribution of BMI in rural areas (school age children). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey

(MxFLS-l, 2002).

27

Distribution of BMI 5-11 years old (Girls)
Rural Areas (1996 and 2002)

KDensity

 

 

 

F I I I I
10 15 20 25 30
BMI

MxFLS (2002) _ _ _ ENAL (1996)

 

Figure 1.2. Distribution of BMI in rural areas (school age girls). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey

(MxFLS-1, 2002).

28

Distribution of BMI 5-11 years old (Boys)
Rural Areas (1996 and 2002)

 

 

 

I I I I F
10 15 20 25 30
BMI

MxF LS (2002) — _ ._ ENAL (1996)

 

Figure 1.3. Distribution of BMI in rural areas (school age boys). Kernel density
estimation using data drawn from the National Survey of Food Consumption and
Nutrition in Rural Areas 1996 (ENAL, 1996) and the Mexican Family Life Survey

(MxFLS-l, 2002).

29

Distribution of BMI 5-11 years old
National (1999 and 2002)

 

 

 

 

I
10 15 20 25 30
BMI
MxF LS (2002) _ _ _ ENN (1999)

Figure 1.4. Distribution of BMI national (school age children). Kernel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican

Family Life Survey (MxFLS-1, 2002).

30

Distribution of BMI 5-11 years old (Girls)
National (1999 and 2002)

 

 

 

10 15 20 25 30
BMI
MxFLS (2002) __ _ ENN (1999)

 

Figure 1.5. Distribution of BMI national (school age girls). Kernel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican

Family Life Survey (MxFLS-1, 2002).

31

Distribution of BMI 5-11 years old (Boys)
National (1999 and 2002)

/\

 

 

 

I I I T I
10 15 20 25 30
BMI

MxFLS (2002) ___ ENN (1999)

 

Figure 1.6. Distribution of BMI national (school age boys). Kernel density estimation
using data drawn from the National Nutrition Survey (ENN, 1999) and the Mexican

Family Life Survey (MxFLS-l, 2002).

At the national level, the shift is mainly observed at the right tails and only for girls
can one observe some increases at the center-left of the distribution. Another interesting
aspect of the national distributions is that they all become relatively flatter around the
center. Overall, at the national level, it appears that average BMI is raising because of

more overweight; however, at the rural level, BMI seems to be increasing at all levels.

To some extent, these pieces of evidence sustain the idea of nutrition facing a

transition in developing countries, as children are becoming less undernourished in terms

32

of weight, but also more overweight. Other pieces of evidence reveal changes in dietary
patterns of Mexicans. The contribution of fat intake to energy increased between 1988
and 1999. The percentage of total energy from fat increased from 23.5% to 30.3% in this
period. Between 1984 and 1998, mean purchases of different food types also changed:
fruit and vegetables decreased by 29.3%, milk and derivatives decreased by 26.7%, meats
decreased by 18.75; while refined carbohydrates increased by 6.3% and soft drinks

(soda) increased by 37.2%.27

The static model of mother’s behavior in section 4 provides some guidance as to the
expected roles of mother’s labor market opportunities and food prices. It is possible to
identify a demand for the weight (BMI) adequacy of each child that is a function of

prices, wages, and non-labor income:

ABi=g1(PiW:Y) (14)

Note that wages are only observable for those women participating in the market.
Thus, to account for labor market opportunities and decisions, I instead use labor supply.
And, consistently with the unitary household model, I assume that all income sources
affect the child outcome similarly, so I combine mother’s wages and all other sources of
income within the household into a total income measurement”. For the empirical

estimation, I allow for possible differences in BMI“ across mothers using covariates such

 

2’ Rivera et al. (2004).

28 In the cases where individual labor income is missing but should be observed, I input it by matching
individuals with a probability score procedure estimated from socio-demographic characteristics and
information availability at an individual level. While this improves observable data on income, it does not
alter main estimation results on the effects of household income on the probability of overweight.

33

as mother’s education and community-level adult BMI. I estimate a linear approximation

of (1 .4) in terms of the probability of a child being overweight:

P(BMI,- > BMI*|X,e) :30 + ,Bl-Xl- +,6me + ,BhXh + ,BCXC +6 (1.5)

for all i=1,. . ..n school-age children, where X;, X,,,, Xh, and Xc are vectors containing

own, mother’s, household, and community variables. X,- includes only child’s age and its

square to account for possible non-linearities. I exclude children time allocation due to

potential endogeneity led by reverse causality—it is not only that a more sedentary child
might gain weight, but also that an overweight child might become more sendentary. X,”

contains mother’s BMI and a dummy variable if pregnant or breastfeeding interacted with
BMI; mother’s education, labor hours, and hours devoted to other non-sedentary and

sedentary activities. To account for omission of unobservable skills, I include scores from
the Raven Test29 for adults as a proxy. Household characteristics included in Xh are

number of siblings, a dummy variable if father is present in the home, a dummy variable

indicating if occupation of household head is agriculture, and total income. Community
level variables in Xc are median BMI of adults, median years of schooling of mothers and

fathers, median family income, proportion of agricultural households, median % of

expenditure in different types of food, dummy variables for region and population size.

Mother’s education serves as both an indicator of maternal nutrition and health

knowledge and as a proxy of socioeconomic status even for non-working mothers. In its

 

2” The Raven Test consists of progressive matrices, for which an individual is asked to complete a pattern
by selecting a piece that fits a missing part. It provides results on a person’s abilities of observation,
problem solution, and learning.

34

former interpretation, education acts through direct channels if nutrition and health
knowledge are in fact acquired at school and through indirect channels by providing
skills that women use when investing in their children’s health. Cognitive ability in terms
of Raven Scores serves as a control for unobservable skills likely correlated with both
education attainment and dietary decisions by mother affecting, in turn, the probability of
her child to be overweight. Labor supply provides an inverse proxy for monitoring time
(Glick and Sahn, 1998). Time devoted to “active” and “passive” activities may be
regarded as reflecting other aspects of mother’s “life-style”. Maternal BMI serves as an
indicator of current nutrition/health status, as well as of genetics; the interaction term that
indicates if pregnant or breastfeeding is designed to account for possible temporary

increased weight.

Number of siblings may affect availability of nutrition resources and therefore child’s
BMI. Less than 80% of children in the sample live with their fathers, so instead of
including father’s characteristics, I include a dummy variable indicating if father is at
home. Results of nutrition surveys emphasize that children in rural areas tend to be less
overweight. Between 20 and 26% of the sampled children live in households with the
head working in agricultural activities, so I include a dummy variable to account for a

potential correlation.

Community variables are incorporated as proxies for current health, social and
economic environment (adult BMI, schooling, income medians and proportion of

agricultural households) and possibly for diet—cultural30 differences (household % of

 

30 Mexican regions are characterized to some extent by different diets, for example northern states tend to
consume more meat than southern and wheat tortillas instead of corn tortillas.

35

expenditures on different foods respect to household food expenditure). Dummies for

regions and population size are included to account for additional potential effects.

Table 1.2 presents summary statistics of relevant variables for the sub-sample of
school children (ages 5-1 1 years) who live in households where only mother or both
mother and father are present3 1. Overweight and not overweight children differ in many
dimensions”. Compared to not overweight children, overweight children tend to be older,
a higher proportion of overweight children attend school, and, in the case of overweight
girls, a smaller proportion attends public schools. Overweight children sleep fewer hours
and spend more time watching TV. All girls devote more time to household work and
taking care of others, but play less than boys; those girls who are overweight spend more
time on household work and less time playing and taking care of others than those not

overweight, for boys time allocated to these activities is not statistically different.

Mothers of overweight children are older—the difference is weakly significant for
girls. They also have a higher average BMI, are more educated, and obtained higher
Raven scores. Mothers of overweight boys participate more in the labor market than
mothers of not overweight group—a difference of 23%. Mothers of overweight girls are
not more likely to participate in the labor market. Overall, maternal labor participation
seems to be more important for boys than for girls. However, there are no statistical
differences in terms of maternal labor supply for any gender. Mothers may be role

models more for girls than for boys; overweight girls have less “active” mothers who

 

3 l I use data drawn from the Mexican Family Life Survey (MxFLS-1), described in the introduction section.
I exclude children with no parents at home and children living with fathers only, because data on their
mothers is not available.

32 This sub-sample consists of children both parents or only mother present at home. Statistical significance
not presented but checked.

36

dedicate more time to sedentary activities and less time to non-sedentary activities, but

for boys there is no evidence of this.

If father is at home, his characteristics also seem to be associated with children being
overweight. Overweight children have heavier fathers and who are more educated.
Paternal labor decisions do not differ between overweight and not overweight children.
Overweight children also have fathers who devote more time to sedentary activities. Only
in the case of girls, those overweight also have more “active” fathers spending more time
on non-sedentary activities. That fathers are “active” interestingly does not make a
difference for boys, but it does if fathers are more sedentary, which may also contribute

to the transmission of models and habits.

Additionally, evidence suggests that overweight children have fewer siblings, live in
non-agricultural households—supporting the idea that prevalence in obesity is higher in
urban communities—and belong to wealthier families. They live in households where
food expenditure is higher; and, where larger percentages of food expenditure are
devoted to animal origin food and soft drinks—-evidence on soft drinks is weak for
boys—, but smaller percentages are spent on fruit, vegetables, and legumes. A striking
piece of evidence is that households of overweight children spend relatively less in
cereals, bread, and tortillas than do those of not overweight children; while the same is
observed for snacks and sugar the differences are not statistically significant. These
results may be indicating some endogeneity of food consumption, as when mothers

concerned about their children’s weight purchase less high-caloric content food.

37

 

 

38

 

3: v9: ”2 v53 3.9 3.: v9? 3.: 833:8 Ecodom
mmdm 34m modm mdém adm dmdm Ixmm gem 853:3 geodomicoz
OM: wddm 3.3 dwdm cdm cm 5.3 voém C82 “8.88 M: £30: mama—83
no as as one no as 23 to assessed gas .83
mw.m Sh wad £6 mad wwd ww.m cod 35% o>E=moU
mod No.5 5d dad odd ZS cage. 86 :oEwosdm
:sd mud wvd mod dmd _wd Sad odd EwBEgO
vmd mddm 3d wvfim w_.m Edm Nod v03 =2m
wd md.mm \iwd mmdm mod 86m dwd omém ow<
@834.
Sam mm; 86 mm; wed mm; dd md.m £0580 23 83E
No.2 ~02 mm.m_ m3: no.2 5.3 v.3 v.3 BE
3% mm; mogm NM: $6 de Nmé mm €03 223983
mdd~ 3.: mod_ v0.2 mmdfi ma; ad wig NC.
m; :d E; mmd ma; Ed 3.. 2d 8884
Rd om; wed 24 mod and va 21d 0:8: So moE>to<
ow; cod 3.; Cd em; mod 31 md 82m
”cocmoozae. 08:.
w _ d cod Nd vod m _ d mod m ~ d mod 6:03
Mad mod Nd odd Rd Nod Nd odd _oonom 23:;
Ed 3d mmd mod 2d mod omd mod 338 8 00
3: dvd do; 88 g; _m.w odd dds ow<
EEO
Son .9m :32 .>oQ .9m :82 .>oQ .3m :82 .>oQ .9m :32
539530 EwEEgO HoZ £32530 Emmozcgc 82
whom 2:0

.mocmcfim bwgm .m._ 6331

 

 

 

vohom Ndmm $68 3di mum mddmm _ddm 3.9% $82: x2 mddmv 056898 30h
mmd Rd dad 31d Sd dd Ed 3% 8:56 tom
36 5.». 5m wwd and 5m 5m mad Swsm new 308m
NY: hmdm $42 5 : Hm 3.: 3.3. m: 2.0m Ewto ESE/x
3.3 v9: 59: wwd: mNd. 5.: mm: mod: 8:98 d8 985 £880
dvd: oddm MON: modm om: Noam mwd: :Ndm 88st— dcm £05988,» £3“:
:88 mo o\ev Bax—ocomxm. doom:
mwm.$ www.md weds? www.mm mmd.~d www.mo :qmm Sfimm A882 x2 mddm 4852a 2:85 130%
dmd w _ d mvd vmd m md m 2 d mvd mmd Razzsotwae.
wmd mwd vd wd dmd mwd vd md 8:8: “a 85mm
mwd m5: 5d 3.: and 5.: 3d do: 3536
Bleedsl:
mm. : 3.2 me— 3.: 3.: 5.2 mwd: mmd: 82>ch ~088de
Q»: 3.: mod: mm. : wmd: 3.: :V. _N 5.2 833:8 bmdaodoméoz
mmd: _mdm ~05 vcdv Rd: «NE 8.: mddm Cona— BVEWE we meson was—83
Ed wad m :d mod m :d mod m _ d mod 20:88:58 8x88 cones:
8N 2e am one 3m 3 SN Sim base 2,288
N2.» mfiw 3% mod 21‘ mod em...“ mmd c0385”:
dmd _wd Dad cod Sad and wvd mod £99530
Ed wwwm mud dmdm :s odm mod mvdm 22m
no.2 Smmm NW: mdwm mod: nmdm 5m: mdm ow<
8.3
.8Q .9m :82 .89 .9m :82 Son: .9m :82 .83 .9m :82
Ewmozcgo £32530 “oz 2363.330 339530 82
mxom

$.88 .3 use...

39

.EwEB :8 Ewfi: :o 9:: 5:5 :8 E5 850:: :0 85mm :8 8:8: 5:3 w:_>= A20 :8» :-mv 8:250 _

 

 

$3 $3: 23$ :33
a; 0%.: n3 R3 5% macaw
MEN 9d mad :m.m NNN v.m mm.m de 35.6 t8 8:82
Nm nn.wm mm: omnm nm.m 8.2“ we: vwdm 53:0 FEB: 8:82
mod NnN no: mw.m mo: :nN mo: cad Smsm :8 8:85 8:82
cmd own NnN a: 6N mmn wnd no: 805:: :0 5:838 5:82
9mm omn cod m6 cmd N _ .n mn.m 2.0 3050:: no 8:838 8:82
«m: m_.mm nn: wwnm mm: no.wm 3.: vnnm 83:8,: .8 22m 8:82
mm: 2 .nm NY: nném vm: wonm mm: No.8 83:: :0 22m 8:82
gala
.89 .9m :82 .>oD .9m :82 .5: .9m :82 .>oQ .9m :82
EwEEgO E32530 HoZ EmmoEgO Ewmgcgo “OZ
mnom $5

.5308 .2 2%:

4O

Finally, in terms of community characteristics, overweight children tend to belong to
communities where adults are heavier; parents are more educated; and where the median
household spends more in animal origin food, but less on snacks and sugar. There is less
evidence on overweight children living in communities where median expenditure on soft

drinks is larger—a weak difference among girls and none among boys.

1.5.2. Results

I estimate equation (1.5) using a linear probability model. Tables 1.3 and 1.4 present
results for girls and boys, respectively: columns (1) contain ordinary least squares
estimates of a basic specification, which excludes cognitive ability and time allocation of
the mother; columns (2) incorporate cognitive ability; columns (3) add labor hours and
time devoted to sedentary and non-sedentary activities; and, columns (4) contain two
stage least squares (ZSLS) to account for potential endogeneity of maternal labor supply,
using same specification as (3) but excluding regional dummies to avoid collinearity with
state-level identifying instruments. Identifying instruments for maternal labor supply are
mother’s age”, median community working hours and labor participation of men and

women, and state dummy variables.

 

33 Mother’s age does not seem to belong into the overweight equation, in none of the specifications was
found to be economically or statistically significant when directly introduced—results not presented.

41

Table 1.3. OLS and two-stage least squares results for girls.

 

Dependent variable: OVERWEIGHT (1) (2) (3) (4)
OLS IV 7
Child:
Age 0.012 0.020 0.028 0.029
(0.042) (0.042) (0.044) (0.044)
Age squared 0.001 0.000 -0.000 -0.000
(0.003) (0.003) (0.003) (0.003)
Mother:
BMI 0.017 0.017 0.016 0.016
(0.002)*** (0.002)*** (0.002)*** (0.002)***
BMI"‘Dummy if pregnant
or breastfeeding -0.001 -0.001 -0.002 -0.001
(0.001) (0.001) (0.001) (0.001)
Education 0.009 0.008 0.007 0.008
(0.003)*** (0.003)” (0.004)" (0.004)“
Cognitive ability 0.009 0.009 0.009
(0.004)M (0.004)" (0.004)"
Labor Hours -0.000 -0.003
(0.001) (0.002)
Non-sedentary
(Activities in hours) ~0.000 -0.000
(0.000) (0.000)
Sedentary
(Activities in hours) 0.000 -0.000
(0.001) (0.001)
Household:
Siblings -0.042 -0.041 -0.045 -0.042
(0.011)*** (0.011)"‘** (0.011)*** (0.011)***
Father at home 0.006 0.006 0.005 -0.026
(0.023) (0.024) (0.025) (0.037)
Agricultural ' -0032 -0029 -0029 -0031
(0.024) (0.025) (0.026) (0.026)
Total Income 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)
Community:
Median BMI of males 2 0.021 0.021 0.019 0.017
(0.009)" (0.009)" (0.009)" (0.009)*
Median BMI of females 2 -0015 -0.016 -0012 -0013
(0.007)“ (0.007)M (0.008) (0.008)*
Median years of schooling
(mothers) 0.008 0.008 0.004 0.004
(0.007) (0.007) (0.008) (0.008)
Median years of schooling
(fathers) -0.002 -0.002 0.001 0.001
(0.007) (0.007) (0.008) (0.007)
Median family income 3 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)

 

42

Table 1.3. (cont'd).

 

Dependent variable: OVERWEIGHT (1) (2) (3) (4)
OLS 1v 7
Proportion of agricultural
households 0.012 -0.005 0.018 -0.01 1
(0.081) (0.081) (0.086) (0.085)
Median snacks and sugar 4 0.005 0.006 0.006 0.007
(0.012) (0.012) (0.012) (0.011)
Median animal origin 4 0.001 0.001 0.000 -0001
(0.002) (0.002) (0.002) (0.002)
Median soft drinks 4 0.006 0.005 0.006 0.003
(0.005) (0.005) (0.005) (0.005)
Region and
population size 5":
Northwest region -0.040 -0.042 -0.048
(0.036) (0.036) (0.037)
Northeast region 0003 -0.008 0.002
(0.038) (0.038) (0.039)
Central-west region 0.009 0.011 0.012
(0.032) (0.032) (0.033)
Central region 0.022 0.020 0.024
(0.031) (0.031) (0.032)
Pepulation > 100,000 -0003 -0014 -0013 0.004
(0.040) (0.040) (0.041) (0.041)
Population 15,000 - 100.000 0.046 0.039 0.028 0.041
(0.046) (0.046) (0.049) (0.049)
POPU~1afion 2,500 ' 15,000 0.020 0.022 0.030 0.051
(0.037) (0.038) (0.040) (0.039)
Constant -0.611 -0.642 -0.674 -0.513
(0.253)" (0.255)" (0.263)" (0.267)*
R-squared 0.10 0.10 0.10 0.09
Observations 2,051 2,017 1,887 1,887

 

Standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%
I Dummy indicating if occupation of household head (or other member) is agriculture.
2 Calculated at the community level considering only adults 25 to 50 years old, by sex.

3 Median household total income at the community level.
4 Median % of expenditure on each food type.

5 Dummy variables with baseline: South region.
6 Dummy variables with baseline: Community with population < 2,500 (officially considered as rural).
7 Identifying instruments for maternal labor supply: mother's age, labor characteristics in community

(median working hours at community level for men and women, labor participation at community level for
men and women), state dummy variables.

43

Table 1.4. OLS and two—stage least squares results for boys.

 

Dependent variable: OVERWEIGHT (1) (2) (3) (4)
OLS 1v 7
Child:
Age 0.088 0.090 0.083 0.079
(0.041)" (0.041)" (0.043)* (0.043)*
Age squared -0.004 -0.004 -0.004 -0.004
(0.003) (0.003)* (0.003) (0.003)
Mother:
BMI 0.009 0.009 0.009 0.009
(0.002)*** (0.002)*** (0.002)*** (0.002)***
BMI*Dummy if pregnant
or breastfeeding -0.004 -0.004 -0.003 -0.003
(0.001)*** (0.001)*** (0.001)*** (0.001)”
Education 0.009 0.007 0.007 0.006
(0.003)*** (0.003)M (0.003)" (0.003)*
Cognitive ability 0.004 0.005 0.005
(0.003) (0.004) (0.004)
Labor Hours 0.001 0.003
(0.000) (0.002)*
Non-sedentary
(Activities in hours) -0.000 -0.000
(0.000) (0.000)
Sedentary
(Activities in hours) -0.001 -0.000
(0.001) (0.001)
Household:
Siblings -0.028 -0.027 —0.030 -0.033
(0.010)*** (0.010)*** (0.010)*** (0.011)***
Father at home 0.044 0.042 0.061 0.091
(0.023)* (0.023)* (0.024)" (0.032)***
Agricultural ' -0014 -0013 -0009 -0014
(0.025) (0.025) (0.026) (0.026)
Total Income 0.000 0.000 0.000 -0.000
(0.000) (0.000) (0.000) (0.000)
Community:
Median BMI of males 2 0.018 0.019 0.023 0.026
(0.009)* (0.009)M (0.010)" (0.009)***
Median BMI of females 2 -0007 -0.008 -0009 -0011
(0.007) (0.007) (0.008) (0.007)
Median years of schooling (mothers) 0.016 0.016 0.019 0.017
(0.007)" (0.007)" (0.007)M (0.007)“
Median years of schooling (fathers) -0.009 -0.007 -0.013 -0.013
(0.007) (0.007) (0.007)* (0.007)"‘
Median family income 3 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)

 

44

Table 1.4 (cont'd).

 

Dependent variable: OVERWEIGHT (1) (2) (3) (4)
OLS 1v 7
Proportion of agricultural
households 0.119 0.123 0.104 0.112
(0.082) (0.082) (0.084) (0.080)
Median snacks and sugar 4 0.003 0.003 0.005 0.008
(0.011) (0.011) (0.012) (0.011)
Median animal origin 4 -0000 -0001 -0002 -0002
(0.002) (0.002) (0.002) (0.002)
Median soft drinks 4 -0.006 -0005 -0004 -0004
(0.005) (0.005) (0.005) (0.004)
Region 5 and population size 5‘6:
Northwest region 0.000 0.000 -0.010
(0.038) (0.038) (0.039)
Northeast region -0.037 -0.042 -0.051
(0.035) (0.035) (0.036)
Central-west region -0.008 -0.004 -0.018
(0.031) (0.032) (0.033)
Central region 0020 -0.016 ~0.021
(0.032) (0.032) (0.033)
Population > 100,000 0.038 0.034 0.038 0.036
(0.037) (0.037) (0.038) (0.039)
Population 15,000 r 100,000 0.005 0.011 0.011 -0.003
(0.040) (0.040) (0.042) (0.042)
Population 2.500 - 15,000 0.032 0.036 0.036 0.034
(0.037) (0.037) (0.039) (0.039)
Constant -0.875 -0.888 -0.862 -0927
(0.263)*** (0.264)*** (0.275)*** (0.272)***
R-squared 0.06 0.06 0.06 0.05
Observations 1,947 1,917 1,796 1,796

 

Standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%
l Dummy indicating if occupation of household head (or other member) is agriculture.
2 Calculated at the community level considering only adults 25 to 50 years old, by sex.

3 Median household total income at the community level.
4 Median % of expenditure on each food type.

5 Dummy variables with baseline: South region.
6 Dummy variables with baseline: Community with population < 2,500 (officially considered as rural).
7 Identifying instruments for maternal labor supply: mother's age, labor characteristics in community

(median working hours at community level for men and women, labor participation at community level for
men and women), state dummy variables.

Variables affecting the probability of being overweight are almost the same for girls
and boys, but some distinctive features are observed. Own age has a positive effect only

for boys (approximately, 008-009), i.e. older boys are more likely to be overweight, but

45

such effect becomes weakly significant as mother’s time allocation variables are included
and labor hours are treated as endogenous; additionally, there is no much evidence of

non-linearity of age.

Mother’s BMI has positive effects for both girls and boys that are robust and strongly
significant across specifications: the probability of being overweight increase by 0.09 and
0.05, for girls and boys, respectively, if mother’s BMI rises by 5 points. In the case of
boys, if mother is pregnant or breastfeeding the effect of maternal BMI decreases a
little—the effect is 0.03. The effect of mother’s education is positive and similar for girls
and boys (between 0.021-0.027 for every three additional years of schooling), but it
becomes slightly smaller and weakly significant for boys when labor supply is considered
as endogenous. Maternal cognitive ability is important only for girls, when included, it
practically does not affect education estimated coefficient; moreover, its sign and
magnitude is comparable to that of education. Neither labor supply nor time allocation to
other activities have effects on girls. In the case of boys, there is only weak evidence of a
positive effect of labor hours, which is found when maternal labor supply is treated as
endogenous: a ten hour/week rise in the supply of mother’s labor increases by 0.03 the

probability for a boy of being overweight.

Among the household characteristics that may influence the likelihood of overweight,
number of siblings is important for both girls and boys; it remains robust, slightly
increases for boys, significant across specifications, and a little larger for girls than for
boys—the presence of an additional child at home reduces the probability of being

overweight by 0.04 for girls and 0.03 for boys. The presence of father at home affects

46

boys but not girls; this effect grows and becomes more significant across specifications,
the presence of his father increases the probability of a boy being overweight by 0.09.
This result may well be associated with the “preference” of fathers for their sons and/or
the “transmission” of habits and lifestyle from fathers to sons. Interestingly, there is no
evidence suggesting that household income is important either economically or
statistically. The household being agricultural has the expected negative sign; however,

this is not statistically significant in any case.

At the community level median BMI of adult males has positive effects for both girls
and boys (0.10 and 0.10-0.15, respectively, for every 5-point increase in median BMI),
which become weakly significant for girls but more significant and larger for boys.
However, median BMI of adult females is negatively correlated only with girls’
probability of being overweight, becoming not or weakly significant and smaller across
specifications (from -0.10 to -0.05, for every 5-point increase in median BMI). Education
of parents at the community level has ambiguous effects only for boys, the effect of
median years of schooling of mothers is robust and positive across specifications (about
0.06 for 3 additional years of schooling); however, this is not the case of median years of
schooling of fathers, which is negatively and weakly correlated with the probability of
boys being overweight (approximately -0.04 for 3 additional years of schooling). There is
no evidence of dietary-cultural habits being important, as none of the community-level
food expenditures (median proportion spent in each food type relative to total food
expenditure) is found to be significant. No effects of income or agricultural-type are

obtained at the community level, just as in the case of the household. Finally, neither the

47

regional variables nor the size of the community are found to have a significant influence

on children’s probability of being overweight.

First stage results are in Table 1.5, providing a fairly good support for identifying
instruments. Mother’s age and state dummies are weakly significant at the 10% level in
the first stage for boys and only men labor participation is significant in the case of girls.
Taken altogether instruments seem to be doing well especially in the absence of prices,
which in principle would provide a much better set of identifying instruments.

Table 1.5. Identifying instruments first stage results.
Dependent variable is mother's labor hours.

 

 

 

 

(1) (2)
Boys Girls
Mother's age -0.113 -0.030
(0.073) (0.074)
Community labor characteristics
Median labor hours of women 0.164 0.001
(0.071)** (0.063)
Women labor participation 38.057 36.066
(5.285)*** (5.055)***
Median labor hours of men 0.036 0.132
(0.158) (0.137)
Men labor participation 9.604 17.81 1
(8.973) (8.752)"
17.30 18.19
F Tes’ [ 0.000] [0.000]
State dummies (F Test) 1.65 1.85
[0.056] [0.024]
. 6.33 6.10
All vanables F Test [0000] [0.000]
R-squared 0. 19 0.24
Observations 1,796 1,887

 

Robust standard errors in parentheses. P-values in brackets.
* significant at 10%; ** significant at 5%; *** significant at 1%
1.6. Conclusions

48

This paper examines overweight in Mexican school children. The motivation arises
from: (i) the increase in prevalence of overweight that has been observed in recent years
among Mexican population; and, (ii) the little attention that school age children have
been given—relative to preschoolers and infants—by nutrition and health national
surveys in Mexico“. Evidence I present here supports the growth in the prevalence of
overweight among school children; moreover, evidence suggests that Mexican children

are heavier than before as observed by the right-shifts of the BMI distributions.

1 develop a simple theoretical model to portray the mother’s behavior when allocating
food and time and to provide insights into the relationships between maternal nutritional
choices, in terms of monitoring time and food, and child’s weight. Basic predictions
suggest that an increase in wages for women, which may be interpreted as increasing
market opportunities for mothers and higher labor force participation/supply, have a
positive effect on children’s weight. It also suggests a negative relationship between food
prices and weight, implying that a decrease in food prices that may translate into reduced
price of calories would contribute to increasing children’s weight. Using this model, I
obtain a demand for weight adequacy of the child and estimate a linear approximation of

its reduced form.

Risk of overweight in Mexican children has been positively associated with, among
other factors, maternal education, socioeconomic status, and fewer siblings. Economic
studies identify the growth of mother labor supply as being a plausible cause of

increasing overweight in American children and adolescents. My results tend to confirm

 

3’ This group has been incorporated into nationally representative surveys of nutrition and health since
1999.

49

these findings and provide evidence of maternal BMI and BMI of adults at the
community level influencing the probability of school children being overweight, effects
that are difficult to disentangle between genetic and environmental factors, although the
negative sign of median BMI of females in the girls’ estimation may be pointing to some
environmental effects. In sum, I obtain some clues into what might be possible
contributing factors to the growth in prevalence of overweight among school children: (i)
increase mother’s BMI, specially for girls; (ii) an increase of maternal labor supply in the
case of boys; (iii) increasing adult males BMI at a community level; and (iv) possibly an
increase in father’s BMI or sedentary habits. Future research should investigate the
channels through which community level characteristics affect the behaviors of

individual children with respect to diet and activity levels.

50

CHAPTER 2
THE EFFECTS OF MATERNAL SCHOOLING ON CHILD HEALTH: THE
CASE OF ANEMIA.

2.1. Introduction

Parental characteristics and behavior are essential inputs into the child health
production function, since decisions regarding most aspects of a child’s life are believed
to be made by her/his parents (see Haveman and Wolfe, 1995; Strauss and Thomas, 1995;
Paxson and Waldfogel, 1999; and, Burton, Phipps and Curtis, 2002). Special attention
has been placed on characteristics and behavior of the mother due to her role as a
caregiver (for example, Thomas et al., 1991; Blau et al., 1996; Glewwe, 1999; and,
Handa, 1999). In particular, empirical studies have shown evidence of important effects
of maternal education on child health, as well as possible ways or mechanisms through
which maternal education affects child health relate (see review by Handa, 1999). Much
of this research has used mortality, morbidity, anthropometric, and/or dietary intakes as
indicators of child health, with an emphasis on infants and younger children, given the
vulnerability of these age groups. In this paper, I examine the effect of maternal
education on child health using a biochemical health indicator and concentrating on
relatively older children. In particular, I analyze the effects of maternal education on

hemoglobin concentrations in blood among school age children.

Hemoglobin is the protein in the red cells that carries oxygen to tissues and organs.
The deficiency of hemoglobin in blood is known as anemia. Anemia may be caused by

both nutritional and non-nutritional factors. Iron deficiency is the most common of its

51

nutritional causes. The World Health Organization (WHO) estimates that anemia affects
2 billion people around the world and that about 50% of all anemia can be attributed to

iron deficiency—iron deficiency anemia or IDA (WHO/UNICEF, 2004).”.

Prevalence of anemia varies widely among regions and countries. For example,
prevalences of anemia and [DA are, respectively, 53% and 39% in Southeast Asia but 9%
and 7% in Europe. The prevalence of iron deficiency in Mexico is high compared to the
U336. According to a recent nationwide nutrition survey (ENN, 1999) prevalence was
estimated between 20% and 27% among children (12 or younger). The ENN99 results37
suggest that prevalence of anemia was 21.1% in women, 27.2% in 6-59 months-old
children, and 19.5% in 5-11 years old children. The most vulnerable groups were
pregnant, women with a prevalence of 27.8%— jointly with their unborn babies—and 1
to 2 years old children among whom prevalence was 48.9%. Using three different
procedures to specifically evaluate iron adequacy’g—one for recent iron ingestion and
two for body reserves—estimated prevalence of iron deficiency associated with anemia
were 17.8% (recent ingestion), and 29.7% or 41.4% when measured as body reserves for
all children in sample (0-11 years old). Overall, the combined prevalence of “severe”
(associated with anemia) and “less severe” (associated with a deficient production of
blood cells) iron deficiencies were estimated to be 43% and 55% or 60%, respective to

each procedure.

 

35 It is worth noting some delicate distinctions: “In the absence of international agreement on how to assess
the iron status of populations, the prevalence of iron deficiency has often been derived from the prevalence
of anaemia using measurements of blood haemoglobin concentration. However not all anaemic people are
iron deficient and iron deficiency may occur without anaemia.” (WHO/CDC, 2005)

’6 Estimates of prevalences of IDA in the US (third National Health and Nutrition Examination Survey,
1988-94) are: (i) 3% in 1-2 years old; (ii) <1% in 3-5 years old; (iii) <1% in 6-11 years old; (iv) 2-5%
among non-pregnant females (aged 12 and older); and, (v) <1-2% among adult males (aged 12 and older).
’7 Villalpando et al. (2003) and Shamah-Levy et al. (2003).

3" Rivera et al. (2001).

52

For years evidence has revealed that child health and parental education are positively
correlated. An extensive appraisal of the literature on market and non-market returns to
education by Wolfe and Haveman (2001 and 2002) shows that parental education, mainly
maternal, improves child health—as measured by infant mortality rates, birth weight, and
vaccination rates. The intergenerational connection between parental inputs and child
outcomes that works through parental education has two components”: (i) direct,
“improvements in the choices and investments made by parents”; and, (ii) indirect,
“quality of the human/social capital of the neighbourhoods in which children grow up”.
Or as Case et al. (2005) assert: “cohort members born into poorer families experienced
poorer childhood health, lower investments in human capital and poorer health in early
adulthood, all of which are associated with lower earnings in middle age—the years in

which they themselves become parents.”

The present research contributes to the existing literatures in health, human capital,
and developing countries by providing insights into the non-market returns to education
and the role mothers play in the etiology of anemia in Mexico. First, my results
complement knowledge of the function of mother characteristics as inputs into child
nutrition and health production. Maternal education may affect child’s health directly or
it may interact in different ways with other factors—income, labor supply, access to
information, health services, community infrastructure, and/or family background.
Second, my estimates also provide information on the fill returns to schooling of women.

Since in Mexico labor market participation of adult women is relatively low, 49.5% for

 

’9 Wolfe and Haveman (2001).

53

25-54 year old women“, market returns to schooling fail to fully capture the impact of
educating Mexican women. Thus, estimates of non-market returns to schooling, e. g. in
terms of child health, are needed to correctly assess the potential impact of women’s

education.

The paper is organized as follows. Section 2.2 presents the conceptual framework,
consisting mainly of a review of the literature on two aspects: hemoglobin levels as an
indicator of health and nutrition status and the role of maternal education in the
production of child health outcomes. The empirical strategy is described in section 2.3.

Section 2.4 presents the results. And, Section 2.5 concludes.

2.2. Conceptual framework

2.2.1. Hemoglobin levels as an indicator of nutrition and health

Among the reasons for anemia, iron deficiency is the most common, contributing to
50% of cases worldwide4l'42. Iron is essential to body functioning because it is necessary
for multiple metabolic processes—such as oxygen transport, oxidative metabolism, and
cellular growth. Low hemoglobin levels showing anemia due to iron deficiency indicate

absence of iron stores, inadequate supply of iron to the bone marrow and other tissues,

 

’0 INEGI (Mexican National Institute of Statistics, Geography, and Informatics). Figure is of 2004, may be
found in http://www.inegi.goanxx’est/contenidosx’espanol/rutinasfggt.asp?t=mtraO6&c=-3655.

4' WHO/UNICEF (2004).

’2 Deficiencies of other micronutrient—vitarnin A, riboflavin, folic acid, and vitamin Ely—may also
contribute to the presence of anemia. Iron deficiency is frequently associated with other deficiencies: “Diets
that are low in animal products are low in absorbable iron as well as retinol, riboflavin, folic acid, vitamin
BIZ, and other micronutrients. . .Intestinal parasites can cause the simultaneous malabsorption of iron,
retinol [vitamin A], folic acid, and vitamin Biz, as can blood loses due to hookworm infestation.” However,
most prevalence of anemia in the world is related with iron deficiency (Allen and Casterline-Sabel, 2001).
Anemia may have other non-nutritional causes like exposure to infections (e. g. hookworm,
schistosomiasis), genetic disorders (e.g. thalassernias) or even lead poisoning (e.g. lead may be found in
paint, toys, furniture, plumbing, and soil, among others).

 

54

and impaired oxygen delivery.43 The presence of anemia mostly indicates “both
functionally significant tissue iron deficiency and impaired oxygen delivery” (Lynch and
Green, 2001). Even if anemia is present without iron deficiency, it might still be related

to deficiencies of other micronutrients and indicative of potential or actual ill-health“.

Anemia may produce symptoms such as fatigue, weakness, decreased appetite,
headache, and shortness of breath, among others.45 In the case of school-age children”,
anemia may impair physical growth, ability to fight infections, and development of motor
and learning capacities, and may also reduce physical performance.47 Moderate forms of
anemia may reduce endurance—i.e. increase time a person takes to complete a task—and
even mild iron deficiency without anemia may decrease work capacity and immune
response (Passi and Vir, 2001). As a result school functioning declines, with potential
negative effects on future accumulation of human capital and productivity. For example,
Horton and Ross (2003) provide an extensive review of the evidence on effects of iron
deficiency and anemia on children’s cognitive development and adult productivity, as
well as calculations of annual losses in physical productivity due to iron deficiency using

data from ten developing countries.

 

’3 “The earliest evidence of iron deficiency is the absence of iron stores. . .[with] no functional
consequences, but there are not iron reserves to meet physiological or pathological requirements. . .[Early
functional iron deficiency consists of] a suboptimal rate of iron delivery to the bone marrow and other
tissues although there is no detectable anemia. ..[The] final stage is characterized by a decrease in
functional iron with an onset of anemia and its associated clinical symptoms (iron deficiency anemia)”
(Lynch and Green, 2001)

’4 However, “[unlike] deficiencies of vitamin A, folic acid, and vitamin BIZ ..., functionally significant
tissue iron deficiency does not appear to occur in the absence of anemia.” (Lynch and Green, 2001)

’5 But symptoms may not show if anemia is mild (see Medline Plus
hug/Mwwnlm.nih.gov/medlineplus/encv/article/000584.htm).

4” Thomas (2001) provides a useful and comprehensive discussion about the relationship among labor
productivity and hemoglobin concentrations and iron stores.

’7 WHO/UNICEF (2004) and Passi and Vir (2001).

 

55

Indicators typically used to estimate the effects of different inputs into the production
of child health are height for age, weight for height, body mass index, birthweight,
parental reports of child health status and mortality. Hemoglobin levels have been mainly
used for evaluation of health interventions that provide micronutrient supplementation
and/or deworming drugs. For example, Thomas et al. (2004) present evidence from an
experimental setting in Indonesia, finding that males treated with iron supplements
improved their physical, psycho-social health, and productivity; for those self-employed,
earnings rose significantly. Bobonis et al. (2004) find that weight and school participation
increased substantially among 2-6 year old children supplemented with iron and
deworming drugs in India. Using results fi'om health interventions carried out in
Indonesia, Thailand, Egypt, and Guatemala, Pollitt (1997) finds evidence that suggest
that IDA constitutes a risk factor for poor educational performance of school age
children. To my knowledge there are few works that analyze socioeconomic factors
associated with anemia and hemoglobin concentrations—Bhattacharya et al. (2002 and

2004) using US data and Villalpando et al. (2003) using Mexican data.

Health indicators are likely correlated.48 Height is closely associated with hemoglobin
concentrations due to its importance for child growth, as iron deficiency anemia
contributes to growth retardation. Hemoglobin and weight relate intimately through
nutrition and diet quality.49 Stunting, wasting, and anemia are all essential indicators of

different types of malnutrition. While wasting and anemia occur due to energy-protein

 

’8 “Children suffer from anemia, either as a result of low iron intakes or poor absorption, or as a result of
illness. Severe protein-energy malnutrition and vitamin BlZ/folate deficiency can also lead to anemia.”
(Alderman et al., 2003).

49 It is worth noting that anemia may be caused by lead poisoning and genetic disorders. But as pointed out
above iron deficiency is its most common cause.

56

and/or micronutrient malnutrition, stunting may be the result from a cumulative effect of
either or both types of malnutrition. Such correlations are not perfect (e. g. an individual
may be anemic but show no signs of nutritional problems in terms of height/weight; or an
individual may be stunted/wasted due to other causes than anemia) and, given data
availability, a more complete assessment should probably analyze them together (e. g.

Ghuman et al., 2005).

The most recent nationwide nutrition survey in Mexico (ENN, 1999) reveals the
existence of high rates of anemia among children; prevalence was estimated to be
between 20 and 27%. Estimates of micronutrient ingestion from ENN-99 24-hour recall
data reveal that iron ingestion by Mexican children (0-11 years old) is 32% to 58% of
adequacy recommendations; 61% to 80.8% in the case of the school age group (5—11
years old); and for preschoolers (12-59 months old) median iron adequacy is between
23% and 565%.“). Authors of this survey suggest that at least part of the explanation is
diet: consumed foods mainly either supply non-heme iron (found in vegetables, legumes,
and cereals), which is more difficult to absorb than heme iron (found in meat), or,
particularly in rural areas, contain substances that inhibit iron absorption. Since for
children dietary decisions take place mostly at home and for Mexican children, in
particular, such decisions are made mainly by mothers, a most pertinent matter seems to
be the role of a mother in the production of hemoglobin level of her child. Given that

anemia of school age children has received little attention relative to preschoolers and

 

50 Children aggregate figures by Rivera et al. (2001) and age-group figures by Villalpando et al. (2003).

57

women of fertile age5 I, and due to the importance of an adequate level of hemoglobin
both in the physical and cognitive development of children aged 5 to ll—not only infants

and younger ones.

Using data from the ENN-99, Villalpando et al. (2003), surprisingly find that more
educated mothers have children with smaller concentrations of hemoglobin. After
controlling for anthropometric indicators (height for age or weight for height), sex,
socioeconomic levels, whether household is indigenous, intake of dietary supplements,
beneficiary of food assistance programs, and altitude, they obtain a negative effect of
maternal education (one additional year has a 7% decrease on hemoglobin levels) for 2-
years or older children, but not for younger children (<2 years old). This result is
intriguing because one would expect ceteris paribus that more educated mothers have
healthier children, unless they work more and monitor their child’s diet less. Several
drawbacks may be causing such a result, including the omission of maternal labor; the
possible bias that arise from a simultaneity or/and ambiguous causality on the production
of hemoglobin levels and other nutrition outcomes as height and weight; and,
“overcontrolling”, i.e. the analysis controls for dietary supplements, food assistance
benefits, and anthropometric indicators, all of which are avenues through which maternal
education effects might operate. Considering these and other concerns discussed below, I
obtain new results about the effects of maternal education on hemoglobin levels and

anemia using data drawn from Mexican Family Life Survey (MxFLS).

 

5‘ Passi and Vir (2001). Note that, given that most attention—by nutritionists, psychologists, and even
economists—has focused on younger children and infants, the works of Horton and Ross (2003) and Pollit
(1997) (see above) are particularly relevant.

58

2.2.2. The role of maternal education in the production of child health outcomes

It is typically believed that children have a minor role when making decisions on
allocation of household resources. It is parental behavior and characteristics that are
pointed out as major (causal) factors determining health, among other, outcomes of
children. In a recent work by Currie and Moretti (2005) evidence suggests the importance
of intergenerational transmission of health and income—parent’s income would affect
child’s health, which would in turn affect her/his future income. Researchers have also
commonly looked for the paths and mechanisms through which education affects child
health, introducing covariates that may be correlated or interacting with education— 4
particularly mother’s education—, such as income, household and community
characteristics, and parental background and endowments. Literature examining effects
of maternal education on child outcomes is very rich; here I review only a small set of

this research done since the 19905.

Table 2.1 contains a summary of these works with information on samples and
empirical approaches used by the authors. Strauss (1990) finds positive and jointly
significant effects of mother’s and father’s education on weight for height and only a
positive maternal education effect on height. He finds that community characteristics are
important as measured by health environment and quality of health infrastructure.
Thomas et al. (1991), in Brazil, find evidence that access to information (newspapers,
television, and radio) explains most of the effect of maternal education on child height. In

urban areas, semi literate status of the mother accounts for some of the education effect.

59

 

.0888 :8 .80 8008.20

 

8880300 88880808 888 08888 0:83
.80 $008.80 80888 888 80088 .80 A088 .80 088> E80883 883 8» Wm
888888 8:08 8008088 .8888 .80 80.8.80 0888 2 8008.80 800.808 08 88 £803 3me >0>8m 63:
88 8880808880 .888 .8 0088.60 88 838 088 882 80:88 888 .880: 80.8 808.88 2 888 808880m 88
8050 $80» no 8088 0080880 300.880 80088 8808.88me 8 m8: 880» 8903 .880: 580: 88882 80832
8880800 688080203 888 08080w 88 w-.. .80 8880800 380 .0888 88:85
88888 .8 08908 88088 0.? ”mac :8 888 883 8> $00.28 800888
888 80088 888088 88088
3...... 3 .3 33...”... ........ “...,...R ..M. 3.38%.”...
80808088 80883283 888 8058—08 8 a .5 88.8 a m H t 32 ._. m
. . . . . . . 8880880 08 .80 088 80.8 888800 8888.2
8080 8088860 888888 $008.80 .
. . 80808 05 .«o 888 08808 880m 08088 H0883 £8: 8 8
088880 8 mAO AS 888 8888.80.88 . . 0mm 88 8902
808 8888 82 8880800 88088 0 2888 0.880.;
9 $0008 .20 808068 888 0888
80 80.8.80 08 .80 808 88—88 08082 088m
88 8088888 3 008800088 mqmm . .
A V .0888 80883 88.. 88>» mAO CV A888 888 88030—08 808889505 08 888002
N . . . . 8888888 9 $0008 80 808088 888888
.0388 80:88; 8 808 $088 888 8888.888 2080808
.3888 80m: 880: $0802 .8888 -888 80 800880 888 mo 0080E>m
580mm 280 :83 8888088880 088088888 £808 .80 .8888 888 8» c 80883
:8 888 88088 88888080 80880880 5808 888.888_ 08 880: >0>8m
80 8008.80 20:00:08 0888000808 888088.880 888800 .8808 88 8803 88 808888m 808$ mmsmbm
80.8 808088 .80 8080805 80 80.8.80 8880800 .8 08888 0mm 80.2 8803 w8>5 0803.8
8003800. 8 00888:. ”0.03888 ”8908 80.8 8803 80 8880800 08.00 08%
888888 :83 80808088 .mm ..8 888 8880880 .8 80 300880
2080080: .mE 80508 888 2080802 8808.886 .383 888 08:88
30885 _80888m 888% 8080882 038% 8083..

.893 05 0080 8880800 88.8088 .80 300.880 08 80 8688 80800—0m ._.~ 038,—.

60

 

80.3880

:8 .888 .808 ”8888 888808.88
8082 8.80808 .80 808088088
0900—3080— 80888 888

8808 808 888880 05888.» 8084

8880830880 80208088 80.“ 880008
08 8808 88088 mo 08: 8083
8088.888 :8 m8 888 8888880 8
38080 808 8088 888088. .8808
280 80 8880880 .8 .80 8008.8 05
80 8088088822 AC ”20300—3080.
8808 888 .m0w88w88 .888v 8:88
888 08008 ”880808880 .95qu
Q .88 8.3.8.88 8... 30 3

.Amm 8 80808 85 80:08

888 8083 888 808888 8.5 80508
8880880 8.8088 60808808 8888
68.8 8.280 ”803889 8808808088
.280 .88 8.5.68.8 2.... :82 .30

8880028
888 8808 m0 8808050

80888 8880880 88 808 08.
80895 .08 8080—80 88 808808008
80.8.8 85 :8 ”090085080. 8088.88

888 5808 888088 88 8080—80
800808228 8880800 88082

80808 8 808880
2.8880888 88 .8808: .8808:
8800888 8088808 :88 888088
808 08 m8 @5088 080015080—
8803 8880800 8888

.80 8008.80 08 :8 880888 0800230802
5808 088 808.888 .888 8:05

8088808088880 @8888

08 0880800008008 88 88880008
808.5 .88808 888 8808 80
8008808 8 888 8 85 808088
:8 80 “00.8.80 038808 88058me

8 888 80880880 88082

800888 .0885

08 88 880:

88888888
08 80.8 8880:

m; :-N

8.32 ”<88.
>0>8m 0800—3052
580: 888

805 888 8888388
.3 800882 800m .80
5.8% 888800

88: .8 8
8.8....»

8% m 8083
88-82 H08.2.6
808802 m00

08> 08 88082 0_ :5
088882 08888”...—

3%:
0330.0

.588 8-2

888 8808 3&2
8088800 NNV

$0.88 8802
888 088880882

88: .2...
888 8m0Q

 

308.88% 80888m

888m

03.808088

0_888m 8083..

88.800 8.8 28....

61

 

4852on SE 388%

Emma: @5508 09:05 .0885 as ho
@335 oSHmwcoQXo 18883 .8535»
958303 @8508 ”35:55.:
@5252 .E 35868

was .5508 EB hmqmm amAO

A2 5 8508 .«o 8:63 98 858E
nwficfiwap “new ”080: an .8an 550
we «553:3 ”38:0 20530555
”328825 :wsobt v5 bfioohv
”momzmofl omson we muooto 8% £05.80
.bfiaowefio: 330325 Ho.“ 6588
”mm .msocowonco ma 3:35 @885

N 2 quote @885 02:38 how 258%
8 2:85 5933 van 53> ”mqo

.Aofigzgv 3on 28288208 new
.05. cosmos? 28253 323885
5:5 mAO .masohw 28958208
wcofiw 3358855 8508 can 230
089:8 9 3mm: <>OZ< 33-25

$808385 woonEEo

oEaEumno €059: Sm was—8:80

5:3 388% 203835
8 3.58% .330; @859: go.“

wfizobcoo 5:3 mosmmEEE £023

.518: 250 no Soho 2558
a was 55% PEEwoo 3:882

A2 5 6588

go "Summon 98 55mm «o 8535
.bficowobwo: 32030:: 515
vows—2.50 m_ z .080: 8 moomtom
cosfimcmm 5E b25835mnsm mo
98 £888 mafia can soars—8

OH mmooow USN QOEMOSUO .E CMQBHOD

828835 wo 852%. .0885

ac #:0202093 cosmos?“ _mEBaE

.530: 220 :o Soto gamma

a 523260 .E kiwi 8m .0258

>233 .LwE canowfiflm 3368
63383:: sumoEcwaéo:

Sc .32 ”£26— 28308208
mmohom 5&6 €30» m vv
5:835 380me 32 mo 38me

 

388% _moEQEm

338m

m;

Quorum :53
own 5 wa 2:0 SN
358m 98:82

3% m 8ch
8mm:
awn 15m Ewfim mcoEvcoU Moonwv
may: .8 u :
zogm 588:5.
mficoE SW2
:52 £56
£850 68: .3 a
own 8.“ Ems? mo 6ng Box
05 mo >033.
3888-380
ow<t8mo€£ 295m 853x

.5288 AN 2an

62

They also find evidence of substitutability between maternal education and
community services. Glewwe (1999) studies possible mechanisms of influence in
Morocco, finding that skills—math, languages, and health knowledge—absorb all the
effect of formal education. Particularly, it is health knowledge which shows as the most
important skill acquired indirectly, through literacy and numeracy, in school. Using
Jamaican data, Handa (1999) provides a rich analysis of possible mechanisms through
which maternal education affects child height: income effects, interactions with
household and community characteristics, information processing, unobserved household
heterogeneity, and bargaining power. He finds evidence of mother’s education having an
effect independent of income, of complementarities between maternal education and
access to television and private doctors, and of substitutability with respect to house
sanitation services. He also finds maternal education is correlated with unobserved
heterogeneity, and that intrahousehold effects and bargaining are important in the form of

presence of father and position of mother in household.

Kassouf and Senauer (1996) analyze the direct effect of parental education and its
indirect effects via wages and full income on three anthropometric measures—height,
weight for height, and weight for age—for Brazilian children. They find positive direct
effects, negative small indirect wage effects, and large positive indirect fill] income
effects on weight for age and weight for health. Total effects are positive and larger as
education increases. Variyam et al. (1999) study the effect of maternal nutrition and
health knowledge on dietary intakes of children 2-17 years of age in the US. They find
that maternal knowledge influences children’s diet especially for preschool children.

However, this effect becomes weaker for older children (> 5 years old). Rubalcava and

63

Teruel (2004) examine the effect of different sources of maternal human capital—
schooling, cognitive ability and childhood background—on child health measured in
terms of height using data drawn from MxFLSl. Thus, their work provides a reference
for the analysis I present here. They find that maternal cognitive ability has a positive
effect on child health, which diminishes when controlling for mother's height. They also
provide evidence of returns to education decreasing when controlling for mother's

observable childhood endowments.

A further concern is that child characteristics may influence how resources are in fact
allocated within the household5 2—e. g. if a child is sick, a working mother could decide to
stop working or work fewer hours to take care of him/her, or else start working or work
more hours to be able to pay for the expenses of her sick child. Reed et al. (1996) finds
that the effects of maternal education (less than 4 years of formal schooling) are non-

9

which, besides socioeconomic status of a

 

linear across “socio-environmental status’
household, includes economic, ecologic and infrastructural characteristics at a village
level—in rural Benin. At higher levels of education this relationship becomes negative,
suggesting a possible association between higher education and labor market
participation, according to the authors. Variyam et a1. include mother's labor participation
status (not, part, and full employed), finding that total and saturated fat intakes decrease if
mother is not employed or part-time employed as compared to being full-employed, but
only for school-age children. Effects of employment status for preschoolers seemingly
occur through maternal nutrition and health knowledge, while school children are more

affected directly by time allocation of mother.

 

52 Burton et a1, (2002) also discuss how parental behavior may be affected by their children‘s behavior.

64

While some researchers have recognized and accounted for the endogeneity of
mother’s labor participation and supply”, maternal education has been mostly treated as
exogenous. However, Desai and Alva (1998) argue that education acts as a proxy for
socioeconomic status of a family and area of residence. They examine data from the
Demographic and Health Surveys for 22 developing countries and find that a causal
relationship between mother’s education and child health has not really been established
because as other household and community-level characteristics are introduced the
effects of maternal education mostly become statistically insignificant. This provides
evidence indicating that mother’s education may be correlated with other factors that
could influence her decisions and behavior towards her child’s health. Thus, even after
controlling for household and community characteristics, there could be unobservables
correlated with both maternal education and child health, such as mother’s genetics
and/or early life constraints that may also have affected her educational attainment. The

next section contains the empirical strategy I follow to account for these issues.

2.3. Empirical strategy

2.3.]. Model

The estimation follows from an application of the household production function by

Berhman and Deolalikar (1 98 8) to the case of child health and nutrition. Household

 

5’ Blau et al. (1996) examine the relationship between mother’s labor supply, infant feeding, and infant
health. They find that after accounting for possible endogeneity there is little evidence of direct causal
effects of mother’s labor supply on weight and height of infants in the Philippines, but that mothers with
higher wages have healthier children. Glick and Sahn (1998) investigate the effect of mother’s labor supply
and income in preschool children of Guinea. After accounting for possible labor supply endogeneity, they
find that in average positive effects of labor income on children height are offset by large negative effects
of labor supply.

65

(parental) optimization, under normal conditions, leads to a reduced form demand for

health of child 1':

H’: =H(p.w.Y;#.~.m.#P,/IC) (2.1)

where p is the price of a market composite-good C, w is the wage rate and Y is the
household non-labor income; ,u,, [1]), ,uh, and ye: are observed and unobserved technology
parameters determined by characteristics of member/child, parents, household, and
community, e. g. sex, age, endowments, background. For empirical estimation some

points should be considered:

(i) Prices are not directly observed, but to account for price variation across communities,
I use expenditures on five food items as percentages of total food expenditures. I expect
that these variables in part reflect price differentials across local food markets rather than
wealth effects, since I also control for household income and community level
infrastructure. I calculate expenditure variables at a household level and then use the

community—level median as price proxies in the empirical estimation.

(ii) Parental labor income not only translates into more household resources but also may
translate into less time available to spend in non-labor activities, e. g. child monitoring
and care. Ideally mother’s wages would capture market vs. non-market decisions,
however, wages are not observed for mothers not participating in the labor market—
approximately 60% of mothers of school-age children in sample. Therefore, I use labor
supply to account for the effect of maternal labor decisions. Concerns of possible

endogeneity naturally rise, for example, a mother may choose to work, stop working, or

66

work less in the market due to her child’s health condition. I discuss this issue below and

explicitly deal with it in my empirical procedure.

(iii) Labor and non-labor incomes are collapsed into total household income.

(iv) Since ,u,, ,up, ,uh, and ,UC are not fully observed I include observable characteristics,

such as age and sex of child, mother’s education, presence of father, and community

characteristics.

Let vector X contain all observable characteristics—individual, parental, household,

and community—and 1) include all unobservables, so the determinants of #1“, yp,,u;,, and

#C may be reclassified into two categories: observed and unobserved. Then, using these

criteria, I rewrite equation (2.1):

H. = H(X;v) (2.2)

where X contains child, parental, household, and community (observable)
characteristics, and 0 stands for all unobservable heterogeneity that may be thought of as

being captured by a random error term. Now, consider a linear approximation of (2.2):

H :30 + :Bixi +15me +flhxh +flcxc +1) (2'3)

where Xi, Xm, Xh, and Xc are vectors containing child’s, parent’s, household, and

community variables, respectively. X,- consists of age and sex; Xm includes maternal

education, cognitive ability (CA), and labor supply. Mother’s education serves as both an

67

indicator of maternal nutrition and health knowledge and as a proxy of socioeconomic
status even for non-working mothers—more educated mothers are likely to belong to
relatively wealthier households. In its former interpretation, education may act through
direct channels if nutrition and health knowledge are in fact acquired at school and
through indirect channels by providing skills that women use when investing in their
children’s health. Unobserved skills are believed to bias returns to schooling; one of its
forms, CA, may be considered as an innate or an acquired ability or both“, affecting
educational performance and attainment. To overcome this omitted variable problem, I
use Raven Test Scores as a proxy for CA. Labor supply provides both an inverse proxy
for available time devoted to activities that promote child’s health (Glick and Sahn, 1998)

and an indicator of labor market opportunities.

Household-level variables in X], include number of siblings, total income, and

whether main labor activity of head of household is agriculture. Number of siblings may
affect availability of nutrition resources per child and therefore her/his health. Instead of
including father’s characteristics I include a dummy variable indicating whether father is
at home, since about 20% of children in the sample do not live with their fathers. I
include a dummy variable indicating whether the child lives in a household where the

head works in agricultural activities—22% of sampled children—which also serves as an
indicator of rural households. Xc includes community median food expenditure (%) on

different types of food to account for dietary/cultural differences between communities. It

also includes community infrastructure and services—sanitary services, health services,

 

5" “It is not clear that innate ability can be measured: any test that claims to do so (in the sense of measuring
a genetic endowment) almost always reflects environmental factors (American Psychological Association,
1995)” in Glewwe and Kremer (2005).

68

and access to information. Since the higher the altitude the less oxygen pressure and the
higher the hemoglobin level of an individual, a control for the altitude is necessary.
Unfortunately, due to data availability I do not observe community-level altitude, so I
have to use the average altitude at the municipality level—the average altitude across
communities in a municipality”. State and population size dummy variables capture

other effects.

I apply two procedures to estimate (2.3): an ordinary least square (OLS) and a two
stage least square (2SLS) to deal with potential endogeneity of education and labor
supply. As identifying instruments in the 2SLS estimation, I propose to use mother’s
background characteristics and current labor environment variables. I use mother’s height
as a proxy of her genetic endowment and past nutrition constraints, since health
investments during childhood are closely associated with investments in education and
eventually with schooling attainment. Child health and/or nutrition have been considered
and found to be correlated with education outcomes for a long time (Alderman et al.,
2001). For example, Strauss and Thomas (1998) review evidence revealing that taller
men tend to be better educated and participate more in the labor market; Alderman et a1.
(2001) find large effects of preschoolers’ nutrition on school enrollments in rural
Pakistan; and, Miguel and Kremer (2004) find that improvements in current health status,
in terms of deworming, reduce school absenteeism in Kenyan children. Mother’s family
background is proxied by using her birth order with respect to her siblings, which is
likely to be related to her school attendance, affecting education attainment, and her

current labor participation/supply, through past or lifetime labor decisions. To control for

 

55 INEGI: Catalog of States Municipalities, and Localities. Downloaded from
http://mapservcr.inegi.gob.mxx’nign2kf‘?c:646.

 

69

current labor environment, I use median labor hours of female and women labor force
participation at the community level. Additionally, to account for potential cohort and
education policy effects, I draw data on the proportion of women who graduated from
elementary, secondary and high school in the mother’s age cohort and state she lived in at

12 years of age from the 2000 Mexican Population Census (INEGISé).

Child health has been usually measured by anthropometric indicators; in particular,
height (see section 2.2.2), which is a long-run marker of general health and lifetime
nutrition. On the one hand, given that hemoglobin concentrations are calculated by using
an electronic device that counts hemoglobin from a drop of blood, hemoglobin is less
prone to measurement error than it is height, which is obtained by using a height
measuring scale. In this sense, measured hemoglobin is a less noisy indicator of true
hemoglobin and, thus, of anemia, than measured height is for true height and thus, of
general health. On the other hand, it is important to understand better the value of
hemoglobin as an alternative way of assessing how maternal education relates to child
health, so I compare maternal education estimated effects by using hemoglobin with
estimates using height. The idea is to examine how the effects of maternal education on
health compare between these two approaches of measuring health. I replicate the
estimation of the econometric model using height; I apply the same identifying
instruments as in the case of hemoglobin, except mother’s height, which enters directly

into the regression.

 

5" Downloaded from

http://www. inegigob.mx/estx contcnidosr’cvspaflgfiprgyeclog/censosr’cp v2000.."bd/pv2000-"pr 2.asp?c—56,2_Q.

 

 

7O

2.3.2. Data

Table 2.2 presents summary statistics of all variables for the sub-sample of school
children (ages 5-11 years) who live in households where only mother or both mother and
father are present”. Average hemoglobin level is 13.2 g/dl with a standard deviation of
1.5. Using critical values proposed by the WHO of 11.0 g/dl for 5-year old and 12.0 for
6-11 year-old children”, I estimate that 14.5% of girls and 16.4% of boys are anemic.
Taken together this is 15.5% of school children in my sample. Considering the sample
average and standard deviation of hemoglobin together with critical values of anemia
diagnosis, the average child would be anemic if his hemoglobin levels drop by one-
standard deviation (i.e. 1.5 g/dl). These children are 8 years old on average and the
sample is balanced across genders. On average, mothers are 35 years old, they attained
6.5 years of schooling—slightly more than only elementary school, and obtained 44% of
total points in the Raven Test. Forty percent of them participate in the labor market,

working a mean of 35.4 hours/week.

On average, there are around 2 children in a household, 80% of the children live with
their fathers and 20% live in agricultural households. The largest part of food
expenditures goes to that of animal origin (37.5%)—meat, poultry, pork, eggs, and
dairy—followed by fruit, vegetables and legumes (23.9%); and cereals, bread and
tortillas (17%); and the smallest proportions go to soft drinks and snacks and sugar (3.3

and 2.8%, respectively).

 

’7 I use data drawn from the Mexican Family Life Survey (MxFLS-1), described in the introduction section.
I exclude children with no parents at home and children living with fathers only, because data on their
mothers is not available.

5” Rivera et al. (2001). Notice that these criteria are for individuals living at the sea level and my estimates
are rough, since I do not adjust for altitude.

71

Table 2.2. Summary statistics

 

Variable Mean Std. Dev.
EMA
Hemoglobin level (g/dl) 13.2 1.5
Girls 0.5 0.5
Age (years) 8.1 2.0
Mother
Age (years) 34.6 6.8
Education (years) 6.5 4.0
Labor participation 0.4 0.5
Labor hours
(weekly hours, if participating) 35'4 19'7
Cognitive ability (point-score Raven Test) 5.3 2.9
Household
Siblings at home 1.9 0.9
Father at home 0.8 0.4
Total Income (Annual 2002 Mx pesos) 55,723 59,395
Agricultural household 0.2 0.4
Community
Food (% of total food expenditure)
Animal origin 37.5 6.3
Fruit, vegetables and legumes 23.9 5.1
Soft drinks 3.3 2.2
Snacks and sugar 2.8 1.0
Cereals, bread and tortillas 17.0 4.6
Infrastructure
Access to piped water (% of houses) 84 26
Access to drainage (% of houses) 54 41
Access to electricity (% of houses) 92 15
Sanitary services
Private toilet with drainage 0.56 0.5
Private toilet with moat 0.36 0.5
Toilet shared with neighbors 0.01 0.1
Lake/river 0.01 0.1
Yard/field 0.01 0.1
Other 0.05 0.2
Garbage disposal
Basket/picked up 0.69 0.5
Burned in house land 0.15 0.3
Burned outside house land 0.07 0.3
River 0.01 0.1
Buried 0.01 0.1
Other 0.06 0.2

 

72

Table 2.2. (Cont’d).

 

Variable Mean Std. Dev.
State and population size:
Baja California Sur 0.03 0.2
Coahuila 0.06 0.2
Mexico DF 0.02 0.1
Durango 0.06 0.2
Guanajuato 0.07 0.3
Jalisco 0.05 0.2
Mexico 0.09 0.3
Michoacan 0.10 0.3
Morelos 0.04 0.2
Nuevo Leon 0.07 0.3
Oaxaca 0.06 0.2
Puebla 0.06 0.2
Sinaloa 0.08 0.3
Sonora 0.07 0.3
Veracruz 0.09 0.3
Yucatan 0.05 0.2
Population > 100,000 0.35 0.5
Population 15,000 — 100,000 0.09 0.3
Population 2,500 - 15,000 0.11 0.3
Population < 2,500 0.46 0.5
Sample size’: 4,410

 

1 Children (5-11 years old) living with mother and father or mother only and with data on
hemoglobin level.

Access to services in the average community is of 84% piped water, 54% drainage,
and 92% electricity; sanitary services are basically private toilet with either drainage
(56%) or moat (36%) and garbage is mainly removed in baskets to be picked up (69%}—
though in 15% of communities people burn it within their land. Sample distribution
among states and community categorized by sized is shown in the second part of table 2;
46% of the children live in rural areas where population size is less than 2,500, 35% live
in large cities with more than 100,000 inhabitants; and 20% live in “mid-size” cities

(2,500-100,000).

73

2.4. Results

2.4.1. OLS

Table 2.3 presents OLS estimates of equation (3) using pooled sub-samples of girls
and boys. Chow tests in the bottom panel provide support for the pooling assumption.
Column (1) reports the basic specification containing mother’s characteristics, child’s
own characteristics, household variables, state and population size dummies—it excludes
all community level variables. I incorporate municipality-average altitude in Column 2.
In column (3), I include community diet-cultural characteristics. Following the common
practice by authors”, I examine possible mechanisms through which maternal education
may work to improve child health, so in columns (4) thru (6), I add community
infrastructure—sanitary services and health services—and access to information—time
allocation to watching TV—respectively. The OLS results may be summarized as

follows:

(i) Maternal education has a positive but small effect on hemoglobin levels (H); three
additional years of education have an effect of 0.02 standard deviations, and such effect is
weak (significant at 10%). When altitude is included the influence of maternal education
falls by 27% and becomes statistically insignificant. However, this effect remains stable
across specifications; it does not change as variables of food expenditure, infrastructure,

or access to information are added.

(ii) Maternal labor supply becomes statistically significant (at 10%) when attitude is

included and remains equal throughout columns (2) — (6). This effect is negative and

 

’9 l partly base on Handa’s (1999) approach. In this section, I analyze direct effects of infrastructure and
information, below I show results of how education interacts with these variables.

74

small; a ten-hour increase in weekly labor supply would make hemoglobin level diminish

by 0.01 standard deviations.

(iii) Other important results:

Mother’s cognitive ability has an unexpected negative effect, but it is not statistically
significant; and, there is weak evidence of older mothers having healthier children.
There is evidence of older children being healthier than younger ones, which
confirms previous findings of prevalence of anemia decreasing with children’s age
(see Appendix 2, Figure A2.1).

Effects of number of children in household are weak and negative; an additional child
at home would decrease H by 0.02-0.03 standard deviations. The presence of father at
home becomes negatively associated with better health (in terms of H), though
statistically insignificant, when controlling for altitude.

Relative to other (“non—rural”) children, those in agricultural households have lower
hemoglobin levels, the effect is of 0.15 g/dl less H, which is equivalent to one-tenth
of a standard deviation; however, this effect disappears when controlling for altitude.
Wealthier households have healthier children. This effect of income is stable across
specifications but small—a one-thousand pesos increase in monthly income leads to
an increase of about 0.006 hemoglobin standard deviations. I suspect attenuation bias

due to classical measurement error60.

 

60 Note that about 3% of the households in the children sample have a calculated total income of zero even
after irrrputing labor income for those working household members with no information of labor income.
Effects of expenditure—results not presented—are smaller (estimates is around 0.006) than those of
income (0.009), goodness of fit of specifications using income is slightly better than when using
expenditure, so I prefer the former to account for material resources available to household.

75

Table 2.3. OLS Results. Dependent variable: hemoglobin levels.

 

(1) (2) (3) (4) (5) (6)
Mother:
Education 0.01 1 0.008 0.008 0.008 0.008 0.008
(0.006)* (0.006) (0.007) (0.007) (0.007) (0.007)
Labor hours -0.002 -0.002 -0.002 -0.002 -0.002 -0.002
(0.001) (0.001)* (0.001)* (0.001)* (0.001)* (0.001)*
Cognitive ability -0.011 -0.010 -0.010 -0.010 -0.010 -0.010
(0.008) (0.008) (0.008) (0.008) (0.008) (0.008)
Age 0.005 0.006 0.006 0.006 0.006 0.006
(0.003) (0.003)* (0.003)* (0.003)* (0.003)* (0.003)*
Child:
Sex 0.028 0.032 0.031 0.031 0.033 0.030
(0.041) (0.041) (0.041) (0.041) (0.041) (0.041)
Age:l
6 years 0.118 0.129 0.130 0.130 0.135 0.129
(0.078) (0.077)* (0.077)* (0.077)* (0.077)* (0.077)*
7 years 0.233 0.246 0.238 0.237 0.238 0.237
(0.079)*** (0.078)*** (0.078)*** (0.078)*** (0.078)*** (0.078)***
8 years 0.450 0.461 0.461 0.461 0.462 0.460
(0.079)*** (0.077)*** (0.077)*** (0.077)*** (0.077)*** (0.077)***
9 years 0.533 0.541 0.538 0.538 0.537 0.538
(0.078)*** (0.077)**"‘ (0.077)*** (0.077)*"‘* (0.077)*** (0.077)***
10 years 0.594 0.597 0.588 0.588 0.589 0.587
(0.080)*** (0.079)*** (0.079)*** (0.079)*** (0.079)*** (0.079)***
11 years 0.785 0.784 0.784 0.784 0.787 0.785
(0.079)*** (0.077)*** (0.077)*** (0.077)*** (0.078)*** (0.078)***
Household:
Siblings -0.036 -0.039 -0.041 -0.040 -0.039 -0.041
(0.024) (0.023)* (0.023)* (0.024)* (0.023)* (0.024)*
Father at home 0.019 -0.016 -0.013 -0.012 -0.013 -0.012
(0.057) (0.056) (0.056) (0.056) (0.056) (0.056)
Total Income2 0.009 0.009 0.009 0.009 0.009 0.009
(0.005)* (0.004)M (0.005)" (0.005)" (0.005)“ (0.005)"
Agricultural household -0.151 -0.064 -0.071 -0.072 -0.078 -0.075
(0.057)*** (0.057) (0.057) (0.057) (0.057) (0.057)
Community:
Altitude’ 0.051 0.048 0.048 0.048 0.049
(0.005)*** (0.005)*** (0.005)*** (0.005)*** (0.005)***
Food4
Animal origin 0.008 0.008 0.008 0.008
(0.006) (0.006) (0.007) (0.006)
Cereals’ 0.000 0.000 0.002 -0.000
(0.007) (0.007) (0.008) (0.008)
Fruit and vegetables° 0.017 0.018 0.019 0.017
(0.007)" (0.007)" (0.007)*** (0.007)M

 

76

Table 2.3. (Cont’d).

 

(1) (2) (3) (4) (5) (6)
Soft drinks -0.004 -0.003 -0.002 -0.003
(0.013) (0.013) (0.013) (0.013)
Snacks and sugar 0.087 0.087 0.074 0.083
(0.029)*** (0.029)*** (0.029)“ (0.029)***
Sanitary services
Private toilet -0.042
with moat (0.071)
Shared/outdoors/other7 '0-073
(0.109)
Health services8
Public clinic/hospital
Kms. < 0.2 0.002
(0.068)
Kms. 2 0.2 or DK 0068
(0.068)
Private clinic/hospital
Kms. < 0.2 -0.185
(0.094)"
Kms. 2 0.2 or DK 0.016
(0.072)
Traditional healer or
midwife
Kms. < 0.2 0.022
(0.062)
Kms. 2 0.2 or DK -0.040
(0.055)
Access to information
Tv" -o.005
(0.006)
State and population size'o'”
Baja California Sur -2.297 -1.l80 -l.222 -l.218 -l.190 -l.217
(0.205)*** (0.226)*** (0.226)*** (0.226)*** (0.230)*** (0.226)***
Coahuila -1.278 -0.577 -0.598 -0.593 -0.563 -0.577
(O.180)*** (0.188)*** (0.197)*** (0.197)*"‘* (0.200)*** (0.197)***
Durango -l .421 -1.065 -1.017 -1.016 -0.986 -l.004
(0.173)*** (0.176)*** (0.183)*** (0.183)*** (0.188)*** (0.183)***
Guanajuato -0.931 -0.679 -0.741 -0.741 -0.621 -0.721
(0.177)*** (0.179)*** (0.182)*** (0.182)*** (0.195)*** (0.183)***
Jalisco -0.824 -0.256 -0.211 -0.216 -0. 174 -0.204
(0.178)*** (0.183) (0.184) (0.185) (0.192) (0.184)
Mexico ~0.067 -0.062 -0. 100 -0.085 -0.048 -0. 104
(0.169) (0.171) (0.172) (0.172) (0.181) (0.173)
Michoacan -1.067 -0.708 -0.709 -0.684 -0.657 ~0.7 14
(0.172)*** (0.175)*** (0.178)*** (0.179)*** (0.185)*** (0.178)***

 

77

Table 2.3. (Cont’d).

 

 

(1) (2) (3) (4) (5) (6)

Morelos -0.716 -O. 150 -0. 161 -0.165 -0.097 -0. 147
(O.185)*** (0.190) (0.195) (0.195) (0.203) (0.195)

Nuevo Leon -1.562 -O.64O -O.665 -0.665 -0.604 -0.646
(0.174)*** (0.189)*** (0.203)*** (0.203)*** (0.211)*** (0.203)***

Oaxaca -1.108 -0.601 -0.565 -0.561 -O.525 -0.577
(O.173)*** (0.178)*** (0.185)*** (0.185)*** (0.192)*** (0.186)***

Puebla -1.362 -0.976 -1.046 -1.031 -0.972 -1.050
(0.175)*** (0.177)*** (0.179)*** (0.180)*** (O.188)*** (0.179)***

Sinaloa -2. 103 -0.91 1 -0.893 -O.897 -O.777 -0.874
(0.171)*** (O.l98)*** (0.204)*** (0.204)*** (0.214)*** (0.205)***

Sonora -1.728 -O.591 -0.691 -0.691 -O.626 -0.683
(0.172)*** (O.196)*** (0.203)*** (0.203)*** (0.214)*** (0.203)***

Veracruz -l .467 -0.534 -0.692 -0.677 -O.622 -O.689
(0.168)*** (0.184)*** (0.193)*** (0.194)*** (0.198)*** (0.193)***

Yucatan -l .933 -O.73O -0.669 -0.630 -0.646 -O.660
(O.179)*** (O.206)*** (0.216)*** (0.221)*** (O.222)*** (0.216)***

Pop. > 100,000 -0.021 0.008 0.127 0.103 0.134 0.127
(0.057) (0.057) (0.071)* (0.082) (0.083) (0.071)*

Pop. 15,000 - 100,000 -0249 -0144 -0015 -0045 -0031 -0011
(0.079)*** (0.079)* (0.089) (0.103) (0.097) (0.089)

Pop. 2,500 - 15,000 -O.l85 -O.100 -0.014 -0.04O 0.009 -0.011
(0.069)*** (0.069) (0.074) (0.085) (0.079) (0.074)

Constant 13.940 12.689 11.756 11.755 11.604 11.795
(0.217)*** (0.241)*** (0.499)*** (0.500)*** (0.508)*** (0.502)***

R-squared 0.19 0.21 0.22 0.22 0.22 0.22
99$.9fi99999......-.- 4150 4150..._4150-4150 4,150 4,150

Chow testn 0.96 0.99 1.11 1.19 1.22 1.12
p-value [0.464] [0.509] [0.707] [0.807] [0.837] [0.724]

 

Standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%
Column (1): basic independent variables.

Column (2): (1) plus mean altitude at municipality level.

Column (3): (2) plus diet variables at community level; median % food expenditure at community level.
Columns (4)-(6): (3) plus sanitary services, health services and access to information, respectively.
’ Baseline of child's age: 5 years.

’- Monthly, 2002 Mx pesos.

3 Average at municipality-level.
4 % of total food expenditure, median at the community-level.

5 Includes tortillas and bread.

6 Includes legumes.

7 Baseline of sanitary services: Private toilet with drainage. Outdoors may be lake/river or yard/ field.
8 Distance dummy variables, baseline is service not present in community. DK if don't know distance.
9 Median-community weekly hours that mothers' watch TV.
’0 Dummy variables with baseline: Mexico DF (excludes metropolitan area).
1’ Dummy variables with baseline: Community with population < 2,500 (officially considered as rural).

’2 Tests the pooling assumption. Ho: Each coefficient and constant term is the same for girls than for boys.

78

Altitude is positively related to hemoglobin levels as expected and its effect is
statistically significant (at 5%), a one-hundred meters increase in average
municipality altitude will have an effect of 0.03 standard deviations on hemoglobin
level

There is some evidence of community dietary habits affecting hemoglobin levels,
higher consumption of animal origin food, fruits, vegetables and legumes, and
(surprisingly) snacks and sugar, increase H.

At the community level there is no evidence of sanitary services having any
statistically significant effect on H, however the signs of the estimates show that
compared to children in communities where the most used service is private toilet
with drainage, children in communities with other types of services would have less
hemoglobin levels.

Among the effects of health services, only relative closer presence of private
clinic/hospitals has a statically significant effect on H: hemoglobin levels decrease by
almost one-eight of a standard deviation in communities with a private clinic/hospital
within 0.2 kilometers from the community center.

Access to information, in terms of community median weekly hours that mothers
watch TV, has a negative but small and statistically not significant effect.

Children living in other states than Mexico DF have considerable lower hemoglobin
levels, with estimated effects that range between of one-third and 1.5 standard
deviations—-the only exceptions, with not significant effects, are the state of Mexico
whose larger municipalities conform the metropolitan area of Mexico City together

with Mexico DF—confirming previous observations showing lower prevalence of

79

anemia in children living in Mexico City (metropolitan area) relative to other regions
(see Appendix 2 Figure A2.2). Possible explanations to these could be differences in
access and/or quality of health services available in Mexico City and/or the altitude of
Mexico City. Actually, once average altitude is controlled for, the effects of state
variables fall by half and the effects of other two states, J alisco and Morelos,
disappear (statistically); in addition, when health services are accounted for, state
coefficient estimates are the smallest (in absolute terms) of all specifications,
particularly those of the states of Mexico, J alisco, and Morelos have the lowest (not
significant) effects. Finally, note that effects of community sizes are ambiguous and

significant only in few cases.

2.4.2. ZSLS

As discussed above, unobserved endowments or maternal background correlated with
mother’s schooling may affect mother’s current behavior towards her child, as pointed
out by Rubalcava and Teruel. Also, child health may influence maternal work decisions.
This is less likely in the case of anemia, which is difficult for mothers to detect. Both of
these considerations represent endogeneity problems. Table 2.4 presents results of
endogeneity and overidentification tests. The left panel presents results of testing that
both education and labor supply are exogenous, and central and right panels contain
results of testing that education only and labor supply only, respectively, are exogenous.
Evidence suggests that education is endogenous but not labor supply—I cannot reject
exogeneity assumption in any case. To some extent this is not surprising, in that maternal

labor decisions are more likely to be affected by their children’s health problems or

80

diseases that are relatively more evident to mothers, and low hemoglobin levels are not
obvious, except maybe when anemia symptoms become a concern to the mother or until
the child is medically diagnosed with anemia. In what follows, I relax the assumption that

maternal education is exogenous, but keep it for maternal labor supply.

I run overidentification tests in order to verify that the identifying instruments are
uncorrelated with the error. Results of these tests are valid under the assumption that at
least one of the instrumental variables is exogenous. Within my instrument set this
requirement most plausibly holds for graduates in mother’s cohort, a variable that
measures the proportion of women in mother’s birth cohort and state where she lived at
12 years that graduated from each school level—although, of course, I expect that all my
instruments are in fact exogenous. Results of overidentification tests indicate that
identifying instruments perform better when relaxing only the assumption of education
being exogenous. First-stage results are in Table 2.5; columns (1) thru (4) contain four
specifications corresponding to columns (3) thru (6) in table 3. Evidence of instruments

identifying education is consistent across specifications.

Some instruments are strongly correlated with education, in particular, mother’s
height that serve as proxy for genetics and mother’s past nutrition and mother’s birth
order, a proxy for family background. Evidence is slightly weaker in the case of mother’s
cohort graduates at each level (see estimates of elementary and high school graduates)

but overall these seem to be good instruments.

81

.o n 3.3 >00 echo 2: 5:5 noun—otooca 2a m.>_ =< no: N

mzocowoxo A23 2 a acmmuawoz no: _

.bco cosmos? .3 358$ch mama a 98 Sam. 98 5:835 Eon no“. 8285:»? “ES. mum“: 3

.Amoom v.8 Noomv owutvooB wags—om...

88.85 E mos—«>3

 

:83 :83 $8.3 $83

3...: :2 4w: 8.8
$8.3 $8.3 88.3 :83
0: $2 at $3

$33 53.3 :33 33.3
o :.o :3 :3
50.3 35.3 :33 $33
30 :.o :3 So
38.3 .33 8403 $3.3

 

$83 :83 88.3 :83 m 08.3 :33 $8.3 $33
32 2: 3: 2: of 3: 02 mi
$8.3 $8.3 $83 $23 38.3 83.3 $8.3 $33
2: Z: _.: 2: 2.2 S: a: m:

   

 
 

$8.3 $8.3 $8.3 38.33523 32.3 :33 $33
2 R 2 3 S 2 M: mm

$8.3 $8.3 $3.3 $8di$m3 :23 :93 $33
2 Z 2 E m 3 3 0.0 2

$3.3 $3.3 533 5303323 :23 :33 53.3

m 33 :mmfim

2:22 33 82-4%

N 3.8 =e=8$~=~ut§©

“mow-N4 gfimswm-DB-Efiso

62-h 52533-53

 

no.0- mmd no.0 2.0 ON- o.N- ON- ON- 2 2 06 m; *mHmB-H can u LEE .5:
_ 3.2 bmunmueutm
mm>~ NmZ :2 NZ , mm>~ Nm>_ :2 NZ mmZ mmZ ~m>H NZ
a: 2 s a a 6 as 5 A8 @ ca 6 Q 3
Eco Snag baa cogosum 3&3 59$ 28 :oumosvm ” ”msocowowsm

.333. :oumocscoctgo can bficowovcm .vN 033.

82

Table 2.5. Identifying instruments first-stage results. Dependent variable is education.

 

 

 

 

( 1) (2) (3) (4)
Mother's genetics and past nutrition
Mother's height 58.13 56.59 56.95 58.71
(20.346)*** (20.345)*** (20.341)*** (20.331)***
Mother's height square -17.38 -16.93 -l6.96 -17.57
(6.632)*** (6.630)“ (6.629)" (6.627)***
F Test 21.30 19.64 21.35 21.30
Mother's family backgroundI
Mother is not oldest -1.32 -1.32 -1,32 -1.33
(0.265)*** (0.265)*** (0.266)*** (0.265)***
Mother does not know -0.30 -0.31 -0.29 -0.30
siblings ages (0.132)** (0.132)” (0.132)** (0.131)"
12.5 12.4 12.3 12.6
[00001 10.000] [9000] 10-000..1_..._._---
Graduates in mother's cohort2
Elementary graduates 3.04 3.07 3.14 3.02
(1.582)* (1.584)* (1.581)" (1.583)*
Secondary graduates 5.61 5.63 5.87 5.60
(2.444)** (2.448)** (2.445)" (2.443)M
High school graduates -0.84 -0.91 -1.34 -0.83
(2.653) (2.654) (2.666) (2.651)
27.87 27.93 27.40 27.77
- 100001100001 10.0901 10.0001
. 25.83 25.28 25.81 25.84
A“ vanables F Te“ [0.000] [0.000] [0.000] [0.000]
R-squared 0.43 0.43 0.43 0.43
Observations 3,968 3,968 3,968 3,968

 

 

 

 

Robust standard errors in parentheses; P-values in brackets,

* significant at 10%; ** significant at 5%; *** significant at 1%.
Column (1): Basic specification + altitude + median % food-types expenditure.
Column (2): (1) + Sanitary Services.
Column (3): (2) + Health services

Column (4): (3) + Access to information (Median weekly hours devoted to TV watching

by mothers)

’ Family composition: mother's birth order. Baseline indicates mother is oldest sibling.
2 Proportion of women, in mother's cohort and state she lived in at 12 years, that

graduated from each school level (proportion accumulates).

83

Estimates of ZSLS procedure are shown in Table 2.6. Columns (1)-(5) correspond to

columns (2)-(6) of the OLS estimation in table 3. I summarize results as follows:

(i) Effects of education remain positive but become significant and larger, increasing by
approximately 7-10 times relative to their OLS counterparts. This reflects that the 2SLS
procedure may well be also correcting measurement error, which could have produced a
downward biased estimate before. Estimated effects of 3 additional years of maternal
schooling on child’s hemoglobin levels range between 0.22 and 0.23, equivalent to 0.15
of a standard deviation. This is a small effect that may be reinterpreted as follows: it takes
6.8 additional years of maternal education for a child, say with a hemoglobin

concentration of 11 g/dl, to be no longer anemic, i.e. to reach a level of 11.5 g/dl.

(ii) Estimates of labor supply effects become more statistically significant than, but still
as small as, their OLS counterparts in table 3. A lO-hour increase in weekly labor hours
decreases child’s hemoglobin level by 0.03, which is equivalent to 0.02 standard
deviations. This is effect is small, for example, for a child with the average hemoglobin
level it would not represent a threat of becoming anemic, since it would require a mother

to work 600 additional weekly hours!

84

Table 2.6. 2SLS Results. Dependent variable: hemoglobin levels.

 

( 1) (2) (3) (4) (5)
Mother:
Education 0.073 0.074 0.075 0.074 0.074
(0.031)M (0.033)" (0.034)M (0.033)" (0.033)M
Labor hours 0002 -0.003 -0.003 -0.003 -0.003
(0.001)" (0.001)M (0.001)" (0.001)" (0.001)M
Cognitive ability -0.038 -0.038 -0.038 -0.038 -0.038
(0.016)“ (0.016)M (0.016)“ (0.016)" (0.016)”
Age 0013 0.013 0.013 0.013 0.013
(0.005)*** (0.005)*** (0.005)*** (0.005)*** (0.005)***
Child:
Sex 0.064 0.062 0.061 0.063 0.061
(0.043) (0.043) (0.043) (0.043) (0.043)
Age:l
6years 0.112 0.115 0.114 0.121 0.115
(0.081) (0.081) (0.081) (0.081) (0.081)
7 years 0.260 0.251 0.249 0.251 0.250
(0.081)*** (0.081)*** (0.081)*** (0.081)*** (0.081)***
8 years 0.449 0.448 0.447 0.451 0.447
(0.080)*** (0.080)*** (0.080)*** (0.080)*** (0.080)***
9 years 0.540 0.542 0.540 0.539 0.542
(0.079)*** (0.079)*** (0.079)*** (0.079)*** (0.079)***
10 years 0.618 0.611 0.611 0.611 0.610
(0.081)*** (0.081)*** (0.081)*** (0.081)*** (0.081)***
11 years 0.816 0.816 0.816 0.818 0.816
(0.080)*** (0.080)*** (0.080)*** (0.080)*** (0.080)***
Household:
Siblings -0.036 -0.044 -0.043 -0.042 -0.044
(0.025) (0.025)* (0.025)* (0.025)* (0.025)*
Father at home 0.019 0.027 0.026 0.026 0.028
(0.059) (0.058) (0.059) (0.058) (0.058)
Total Income2 -0.000 0.000 0.000 0.000 0.001
(0.006) (0.006) (0.006) (0.006) (0.006)
Agricultural household -0.013 -0.023 -0.025 -0.030 -0.026
(0.063) (0.063) (0.063) (0.063) (0.063)
Community:
Altitude’ 0.047 0.042 0.042 0.043 0.043
(0.005)*** (0.005)*** (0.006)*** (0.006)*** (0.006)***
Food4
Animal origin 0.009 0.008 0.009 0.009
(0.007) (0.007) (0.007) (0.007)
Cereals5 0.007 0.007 0.009 0.006
(0.009) (0.009) (0.009) (0.009)

 

85

Table 2.6. (Cont’d).

 

(1) (2) (3) (4) (5)
Fruit and vegetables6 0.026 0.026 0.028 0.026
(0.008)*** (0.008)*** (0.008)*** (0.008)***
Soft drinks 0.015 0.015 0.017 0.016
(0.015) (0.015) (0.016) (0.015)
Snacks and sugar 0.076 0.082 0.060 0.073
(0.030)M (0.031)*** (0.031)* (0.031)“
Sanitary services
Private toilet -0.047
with moat (0.074)
Shared/outdoors/other7 0-002
(0.114)
Health services8
Public clinic/hospital
Kms. < 0.2 0.001
(0.071)
Kms. 2 0.2 or DK 0.062
(0.070)
Private clinic/hospital
Kms. < 0.2 -0.256
(0.095)***
Kms. 2 0.2 or DK 0010
(0.074)
Traditional healer or
Midwife
Kms. < 0.2 0.016
(0.063)
Kms. 2 0.2 or DK -0.060
(0.057)
Access to information
TV9 0004
(0.007)
State and population size'o'll
Baja California Sur -l.311 -1.420 -1.420 -l.382 -l .416
(0.237)*** (0.243)*** (0.244)*** (0.245)*** (0.243)***
Coahuila -0.61 1 -0.764 -0.754 -0.731 -0.748
(0.195)*** (0.212)*** (0.212)*** (0.214)*** (0.213)***
Durango -1.050 -l.134 -l.119 -l.114 —l.124
(0.183)*** (0.192)*** (0.192)*** (0.195)*** (0.193)***
Guanajuato -0.563 -0.747 -0.738 -0.610 -0.732
(0.198)*** (0.188)*** (0.188)*** (0.201)*** (0.189)*"‘*
Jalisco -0.325 -0.356 -0.358 -0.324 -0.350
(0.191)* (0.193)"‘ (0.194)* (0.200) (0.194)*
Mexico -0.015 0130 -0.112 -0.103 -0.133
(0.182) (0.178) (0.177) (0.187) (0.178)
Michoacan -0.655 -0.764 -0.756 -0.713 -0.768
(0.187)*** (0.184)*** (0.186)*** (0.191)*** (0.185)***

 

86

Table 2.6. (Cont’d).

 

(1) (2) (3) (4) (5)
Morelos -0.204 -0.339 -0.336 -0.271 -0.328
(0.197) (0.209) (0.209) (0.214) (0.210)
Nuevo Leon -0.716 -0.889 -0.883 -0.831 -0.875
(0.198)*** (0.224)*** (0.224)*** (0.230)*** (0.226)***
Oaxaca -0.610 -0.722 -0.705 -0.687 -0.730
(0.188)*** (0.193)*** (0.193)*** (0.198)*** (0.194)***
Puebla -0.880 -1.035 -1.042 -0.949 -1.038
(0.187)*** (0.184)*"‘* (0.186)*** (0.192)*** (0.184)***
Sinaloa -0.963 -1 .063 -1.067 -0.916 -1.049
(0.205)*** (0.216)*** (0.217)*** (0.224)*** (0.218)***
Sonora -0.662 -0.896 -0.896 -0.824 -0.890
(0.203)*** (0.220)*** (0.221)*** (0.228)*** (0.220)***
Veracruz -0.552 -0.827 -0.824 -0.753 —0.824
(0.193)*** (0.201)*** (0.203)*** (0.205)*** (0.202)***
Yucatan -0.795 -0.863 -0.829 -0.840 -0.857
(0.213)*** (0.226)*** (0.233)*** (0.231)**"‘ (0.227)***
Pop. > 100,000 0060 0.097 0.085 0.129 0.097
(0.070) (0.076) (0.087) (0.088) (0.076)
Pop. 15,000 - 100,000 0207 -0.039 -0.062 -0.033 -0.037
(0.088)" (0.093) (0.108) (0.102) (0.094)
Pop. 2,500 - 15,000 -0.104 -0.001 -0.012 0.039 0.001
(0.071) (0.076) (0.088) (0.081) (0.076)
Constant 12.248 11.036 11.041 10.919 11.066
(0.332)*** (0.615)*** (0.617)*** (0.628)*** (0.622)***
R-squared 0.20 0.20 0.20 0.20 0.20
Observations 3,968 3,968 3,968 3,968 3,968

 

Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%

Column (1): basic independent variables.
Column (2): (1) plus mean altitude at municipality level.
Column (3): (2) plus diet variables at community level; median % food expenditure at community level.
Columns (4)-(6): (3) plus sanitary services, health services and access to information, respectively.

’ Baseline of child's age: 5 years.

2 Monthly, 2002 Mx pesos.

3 Average at municipality-level.

4 % of total food expenditure, median at the community-level.
5 Includes tortillas and bread.

6 Includes legumes.

7 Baseline of sanitary services: Private toilet with drainage. Outdoors may be lake/river or yard/field.
8 Distance dummy variables, baseline is service not present in community. DK if don't know distance.
9 Median-community weekly hours that mothers' watch TV.

’0 Dummy variables with baseline: Mexico DF (excludes metropolitan area).

” Dummy variables with baseline: Community with population < 2,500 (officially considered as rural).

87

(iii). Other results:

— The effect of maternal cognitive ability remains negative and, unexpectedly,
becomes 3.5 times larger and more significant than the OLS—~three additional
points in the Raven Score is related with a decrease of 0.1 1012 in hemoglobin
levels“.

- Effects of mother’s age also become larger (and more significant) relative to the
OLS case. The average child with a mother who is 5 years younger would
decrease H by 0.04 standard deviations.

- Total income is no longer significant when applying ZSLS and its effect become
ambiguous across specifications.

— Remaining results are similar to those of OLS.

Table 2.7 shows maternal education effects on child height. These results reveal that
three additional years of maternal education translate into an increase approximately 0.24
of a standard deviation of height—the analogous effect is approximately 0.15 of a
standard deviation of hemoglobin. A relatively larger estimate in the case of height seems
reasonable since general education is more likely to provide basic nutrition and health
knowledge that eventually affects height, than specific instruction on, for example, food
containing supplying iron and effects of iron deficiency. Moreover, height and growth
retardation are easier to observe and “diagnose” by the mother than low hemoglobin

levels and anemia.

 

6| . . . . .
The estimated coefficrent IS also negative when Raven Scores are 1n terms of Z-scores.

88

Table 2.7. Education effects on child height, ZSLS results.

 

(1) Basic + altitude 0.082
(0.025)***
(2) + Diet variables at community level 0.081
(0.026)***
(3) (2) + sanitary services 0.078
(0.025)***
(4) (2) + health services 0.082
(0.026)***
(5) (2) + access to information 0.083
(0.026)***

 

Standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%

Basic specification includes: labor hours, cognitive ability and age of mother; child's
gender and age; number of children in the household; presence of father; whether the
household is agricultural; and, state and population size dummies.

See notes in table 2.6.

2.4.3. Interactions

In this section, I examine possible mechanisms through which maternal education
affects child’s health. Above, I directly included variables of infrastructure and access to
information, where the effect of education remained practically unchanged and only

private health services coefficients were statistically significant.

Now, I incorporate interactions between education and infrastructure and between
education and access to information to analyze possible complementarities or substitution

effects. Results are shown in Table 2.8.

89

Table 2.8. Interactions results. Dependent variable: hemoglobin levels.

 

(1) (2) (3)
Maternal Education 0.048 0.197 0.073
(0.030) (0.043)*** (0.039)*
Interactions:
With sanitary services
Private toilet with moat -0.021
(0.034)
Shared/outdoors/other 0.080
(0.056)
With health services
Public clinic/hospital
Kms. < 0.2 -0.l27
(0.041)***
Kms. 2 0.2 or DK -0.100
(0.045)”
Private clinic/hospital
Kms. < 0.2 -0.066
(0.052)
Kms. _>_ 0.2 or DK -0.006
(0.033)
Traditional healer or midwife
Kms. < 0.2 -0.052
(0.032)
Kms. 2 0.2 or DK -0.028
(0.030)
With access to information
TV 0.000
(0.003)
R-squared 0.21 0.18 0.20
Observations 3,968 3,968 3,968

 

Standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%
All regressions include basic covariates, mean-municipality altitude, and median % food-

types expenditure.

Column (1): Sanitary services and interactions of education and sanitary services
Column (2): Health services and interactions of education and health services
Column (3): Access to information and interactions of education and median-community

weekly hours of TV watching by mothers.

Identifying instruments: mother's height and its square, mother's birth order, proportion of

graduates in mother's cohort by school levels.

90

When including interactions with sanitary services the effect of education decreases
and becomes statistically not significant. In the case of access to information the effect is
similar than before. Only estimates of public clinic/hospital coefficients are statistically
significant. The presence of this health service decreases the effect of education and more
so the closest the service to the community center (downtown); may be reflecting a
substitution effect between education and public health services. The estimated effect of
3 additional years of education when public health service is not present in the
community is equivalent to 0.4 standard deviations of hemoglobin levels, but only 0.14,
if this service is relative close to the community center, and 0.19 standard deviations, if
service is relatively far. In other words, the presence of a public hospital reduces the
effect of maternal education with the total effect remaining similar than before and

becomes slightly higher if the public hospital is relatively distant.

2.5. Conclusions

I find that effects of maternal education on hemoglobin levels are consistently
positive across different specifications and procedures. Evidence suggests that education
is endogenous and possibly contaminated by measurement error, producing biased
estimates from OLS procedures. Most accurate estimates show that an increase of 3 years
in mother’s education affects a child’s hemoglobin level by 0.15-0.16 of a standard
deviation, which are small. However, this does not imply that it is not valuable to educate
women, rather that strategies to improve the quality of education would be valuable. For
example, education programs in Mexican schools and health prevention programs in

facilities could be improved by including specific nutrition-health topics that are of public

91

health relevance, emphasizing the relationship between dietary intakes and nutritional-
health outcomes. These would be particularly important in the case of health problems

that are not obvious or which symptoms may be misleading.

I also find evidence indicating that maternal education may operate with public health
services (clinic/hospital) through a substitution effect. I also find evidence suggesting that
mother’s labor supply may not be endogenous, which is reasonable, since low
hemoglobin levels are not obvious to the mother and her labor decisions are likely to be
independent of this aspect of her child’s health. However, further analysis would be
needed to confirm this. Additionally, I find that younger mothers have less healthy
children and that, although, girls appear to be healthier than boys this difference is not

statistically significant.

92

CHAPTER 3
DO PUBLIC INITIATIVES CHANGE SEXUAL AND REPRODUCTIVE
BEHAVIOR OF WOMEN?

3.1. Introduction

Sexual and reproductive education intends to change sexual/reproductive behavior. It
is expected to translate into better (more conscious) sexual and reproductive health
practices: pregnancy control, prevention of AIDS and other sexually transmitted diseases,
and, in the case of women, prevention of breast and cervical cancers. As Michael (2004)
states “the value of capital associated with sexual behavior. . .involves resources that are
both capabilities and habits that are embedded or become part of the person’s pattern of
behavior”, which would imply an important role of sexual and reproductive education in
the formation of such a capital. Yet opponents of programs of sexual education argue that
instead it encourages sexual activity and increases unwanted pregnancies, particularly

among younger (unmarried) people (Oettinger, 1999).

During the 19705 the Mexican government, like that of other countries, introduced
policies to promote birth control and family planning in order to reduce the pressure of
population growth. These policies were implemented through two essential sectors:
education and health care. In 1974, sexual education programs were implemented in
Mexican schools for the first time and contents upgraded as new social and public
concerns rose: in 1987, with the AIDS epidemic, the emphasis was put on prevention;
and, in 1994-1998, gender issues, sexual, and reproductive rights were included

(Rodriguez, 2001). Along with education programs, in 1972 and 1973, public health

93

institutions incorporated family planning and birth control programs at large. In 1973, a

new Sanitary Code eliminated the prohibition on contraceptives sales and advertising.

The purpose of this study is to examine whether the sexual and reproductive behavior
of Mexican women changed due to public initiatives launched in the 19703. I use a
variety of indicators to characterize sexual and reproductive behavior: first sexual
encounter, first and second pregnancies, and utilization of contraceptive methods. To
identify the effects of public programs, I make use of individual exposure to in-school
sexual education and to birth control promotion in public health facilities. In a previous
work, Miranda (2006) finds that young cohorts of Mexican women are actually
postponing first birth compared to older ones. However, to my knowledge, no study so
far has aimed to explicitly disentangle the effect of birth control and family planning
public initiatives on the reproductive and sexual behavior of Mexican women. My
contribution consists of analyzing changes in various components of sexual and
reproductive behavior, exploiting the fact that exposure to programs varied across women

of different cohorts, schooling attainments, and utilization of health facilities.

The following section contains a brief account of the family planning policies
implemented in Mexico since the 19703. Section 3.3 presents a review of the literature.
Econometric models are described in section 3.4. Data and empirical analysis are in

section 3.5. Results are in section 3.6. Section 3.7 concludes.

94

3.2. Background

In the 1970s, the Mexican government launched several initiatives to promote birth
control and family planning, in line with international trends62, to reduce the pressure of
the population growth: (i) sexual education in schools as part of the natural science
curricula at the elementary and secondary levels; (ii) family planning and birth control
programs in public health institutions; and (iii) foundation of the Mexican National
Council of Population (CONAPO) in charge of demographic plans and research. These
public initiatives consisted of education, informative campaigns, and direct promotion in
schools, hospitals and clinics, and mass media. The principles regarding the new
population policies were introduced in both the Federal Constitution (CPEUM) and the

63.64_

General Population Law (LGP)

“Every individual has the right to freely, responsibly, and well-informed
decide on the number and timing of her/his children.” (CPEUM, Title

First. Chapter 1. Article 4. Modification published in 1974).

“. . .carry out family planning programs through public education and
public health services, and supervise that such programs and those
managed by private organizations are respectful of the human rights and

family dignity. . .”(LGP, Chapter 1. Article 3. Part II. Modified in 1973).

 

62 For example, the United Nations founded the Fund for Population Activities (UNFPA) in 1969.
63 The General Population Law was first promulgated in 1936 and explicitly promoted fertility and
population growth (Palma C., 2005 and Sénchez L., 1997). Note acronyms are in Spanish.

6’ Own translation.

95

“. . .influence the population dynamics through education, public health,
professional and technical training, and children protection systems...”

(LGP, Chapter I. Article 3. Part IV. Modified in 1973).

The aim of the Mexican government was to promote birth control and family
planning“. Yet, the use of contraception methods and the development of a new
perception regarding sexuality always emphasized the role of the family [wedlock] as the
[only] acceptable space for sexual practices (Garcia A., 2001)“. Overall, these public
policies were effective, annual average population growth rate decreased from 3.4 (1960-

70) to 1.0 (2000-05)”7 and global fertility rates dropped from 5.7 (1976) to 2.2 (2006).

Starting in 1974, sexual education was included in the contents of the textbooks”8 at
the elementary level, beginning in fifth grade, and at the secondary level (seventh thru
ninth grades). In 1974, the curriculum included menstrual functioning, changes due to
puberty and adolescence, and reproductive processes (fifth grade textbook); and birth
control, contraceptive use, and prevention of sexually transmitted diseases (secondary
textbooks). Since 1987, with the AIDS epidemic an emphasis was put on prevention in
the secondary courses. Beginning in 1994-1998, from fifth grade on, the program has
added the formation of values: gender-equality issues, sexual and reproductive rights,

sexual relationships, love, and sexual violence, among others. While curricula upgrading

 

(’5 Figure 3.1 presents a timeline indicating basic initiatives in the 19705.

66 At the beginning the idea was based on “biological and neo-malthusian” principles, but then due to the
intervention of the Catholic Church and the National Parents Union (Union Nacional de Padres de F amilia,
UNPF)—a conservative institution that defends religious-related moral principles—the emphasis shifted.
67 Dindmica de la poblacién—population dynamics—at the National Institute of Statistics, Geography and
Informatics (INEGI) webpage (http:/x’wwwinegi.gobmx/esL’delault.aspx‘?c=2343).

68 Textbooks published by the Secretaria de Educacion Publica (Ministry of Education) are free and
mandatory in public schools and while also free for private schools their use is discretionary (Torres S.,
2002)

 

96

tried to adapt to social and environmental needs, many young Mexican people (12-29
years old) still believe that they have received important sexual health information in
places other than school. According to the 2000 Mexican Youth Survey (Rodriguez,
2001): 34.1% of the young people believed that they received important sexual health
information from school; 24.4 % believed that they got it from parents, 19.4% from own

learning; 9.7% from friends; 5.9% from media; 4% other and church.

Along with education programs, other sexual-reproductive policies were launched in
the 1970369. In 1972 and 1973, public health facilities with large coverage70 started to
incorporate family planning and birth control programs. In 1973, a new Sanitary Code
eliminated the prohibition on contraceptives sales and advertising, and between 1973 and
1976 the politics, organization, and agenda were set to integrate family planning into the
health services. The government founded the CONAPO in 1974, and, in 1977, the first
National Program for Family Planning (PNPF). Many important strategies were initiated,
among which was the Family Planning Communication Program, with the purpose to
“inform, make aware, and generate changes in attitudes with respect to reproductive
behavior”“, and an expansion strategy to reach rural communities (population size: 500-

2500). It was not until 1984 that family planning was included in the General Health Law

 

69 The following brief historical account of the sexual/reproductive programs of the health sector is based
on Palm and Rivera (1996), unless otherwise noted.

70 Large coverage institutions: Instituto Mexicano del Seguro Social (Mexican Social Security Institute,
IMSS), Secretaria de Salud y Asistencia (Health Ministry, SSA); and Instituto de Seguridad y Servicios
Sociales para los Trabajadores del Estado (Social Security for State Workers, ISSSTE). Before the public
initiatives of the 19705 two private organizations were founded with the objectives of providing family
planning services (Asociacion Pro Salud Matema, in 1958), doing research, providing education and
supplying medical services for family planning (Fundacion para Estudios de la Poblacién, A.C., in 1965—
now F undacion Mexicana para la Planeacion Familiar, Mexfam). Also, in 1968, previous to other and more
general public initiatives, the Instituto Nacional de la Nutricion Salvador Zubiran (Nacional Nutrition
Institution, INNSZ) started a research program and the provision of family planning services. (All
acronyms are in Spanish).

7' In gQué es el CONAPO? (What is CONAPO?) at www.conapogob.mx.

97

(LGS) under similar principles as those in the LGP. During 19805 other approaches were

adopted as more results of social (surveys) and medical (contraceptive methods and

techniques) research were obtained. Also, policies for the prevention of AIDS were

developed like, in 1988, the creation of the National Council of AIDS in Mexico (now

CENSIDA)”. In the 19905, the PNPF emphasis was set on specific population groups

(teenagers, rural, and men) and it continued to provide planning services, information,

and education.

 

Oldest Youngest

women women

(49 yr-old) (14 yr-old)

1953 1988

l l l | I I _

I I l | I I 7
1972-73 1974 1987 2002

M FLS
Health sector: Education sector: x

1. Introduction of
family planning in
public health
facilities

2. Reform of
sanitary code to
allow sales of
contraceptives.

Introduction of
sexual education
program in
schools.
(5mgrade
elementary thru
3'dgrade
secondary)

Figure 3.1. Timetable of major public initiatives to promote birth control and family

planning in Mexico.

 

’2 Hemandez-Chavez (1995).

98

At the same time, other factors may have also contributed to changes in sexual and
reproductive behavior of the Mexican population. For example, the increase in schooling
attainment of both females and males may be associated with higher opportunity costs of
unwanted consequences of sexual activity, i.e. unwanted pregnancies”. New generations
may develop different religions and social conventions compared to older ones, e. g. the
“correct” role of women with respect to marriage and children". However, it is not
possible to know the extent to which these socio-cultural attitudes have been influenced

by the public initiatives to promote sexual and reproductive health.
3.3. Literature Review

Table 3.1 contains a synthesis of related literature. Most of these works analyze the
relationship between sexual and reproductive behavior and sex education, birth control
promotion, or other similar policies. This literature mostly focuses on teenagers in the
USA and the UK. I also include the paper by Miranda because he examines the
postponement of first birth by younger Mexican women, although he does not explicitly

look at the impact of sex education or family planning policies.

Oettinger (1999) and Tremblay and Ling (2005) examine the association between sex
education and sexual behavior among American teenagers. Oettinger theoretically models
the impact of sexual education on sexual activity. Oettinger’s model predicts different
outcomes resulting from various types of sexual education: risk altering, risk revealing,

and utility altering.

 

7’ Figure A3.1 shows how education attainment increased in Mexico in last decades.
7” Figures A3.2-A3.4 present results from the World Value Survey that suggest how Mexican society has
changed its attitudes toward motherhood and the role of women.

99

 

.:..::m 8::::

0>::_8 30: :0 88:08:
0w:::0 8:08:88

:8 b:>:0< 53:0: 88:8
:0 880080 880:6

m0 8:::: :0:00qx0
888:0 888:: 8:5 .m
.:83:30: :0 :83?

888:8:

m8: 888 :0 800:: 80:88:08: .:0
88:8 80: 3038—8 :::5 8880.: 8.:
:09:— 803 808.8 .880 ::0: E 888w
0880.: 088 :0: 8:08:88 :0 888:0
0:: :0 88:30 80803 080m .8888

8:808:88
8:232: 0: 0:: :0:::.8>
08889? 88:85
.w::0:: b::: :8

 

888: m: 88:88: :0 8.: 53:0: 83:8 8:80 :85 :08808: $8808: :8: 888:: .Amowi
0::: :03080: 850:3 :0 :0::0::0 x8 ”88:8”: 58:80:08: m._.: 80:88.: 0 :.:x8 .280 :8 33:
:88: 888: 8:08:88 88080:: :0: 8:000: 0: 08880.8 0: . . . : _ 8 88:80::
:: :05: . w w - Q .:0 89m: 80:? m0 800.:0 - A000:
: . .:0< 0 8: 88:00 83:: x8 08:: :8: 0: 88 20:3 m8: .b v: 30: w
:: 80:88 85808800 8::_:8:m 3:0: 8:8: 803:0: :0 8:00
: . :0 80800: .80w::00:
:0 8:88:08: .888:8 80:80:: b0>:::::::0 0:308 0 08 z :0: 800 80: 28:38::
88808: :8: .N 08:: .:0 :< 08—08: 0:888:80: ._mhw::0w_ 88:08:“? >m: Z <mD
80:08:88 :00: :0 80:0 : :85 8:82:88 :88: :80: : c5 w:0:w: :0_.>: 0<
80858: :85 53:0: :8 m:00::8_: 8:8: :80 a: 83:8 :8 :0_.::0”:w
8880:: .800: 0>:0: A80: .882: 88 803:0: 0::: 0888::
2888 8.: 88:88: 8:08: 0:: 8:0,: 0: 8388:, :88: 3 . .
:0 0::: 8:: 88 8:: ”5:5 :088:8
8888.8: 888:: :8: .: 8: 8:: 20:08 08: :88: :83
8080300 x8 :0 89?
888800 388% 80:88: 803088: 88:00:00 088m :0:::<

.:0:>::0: 0>:0::088 :8 88:8 88:88 86:8 8038: ._.m 0:::H

100

 

.80::00:00
088:8 8 08808 8
::_3 :08888 8: w88_:

.3 :0 :0mm:0> 38:»: 8 08838 0:
08:08:: 220 c: .:8 ”808880
8 888—: 3:8: 8:88: h8:08

.803:00:00 8:08 :0 8::
80:88 088:8 :0 :8:::

m8: :0 8808 8: 88::

 

038m 8 :0 0:: 000:: C 88: 3 :8 0: 8008 8 :0 3 .8483:
3:8: 0: 8008 .808 .w:::: 33.” 33:”:02 0:WW:M:::0:8 :_0 8w8w58 95:80:“8 88880:
8:: 00:0:30 080m: 88: 8:88 3:8: :0 88: 883 8888: :0 888:8 :0 83888
:0WMMMMWHMMW3MWWMN .32 80% :0: 03883 888: 8 88: 8: 88:8: 2: .3338 83:8 .8330:::< 38$
8 0 00:0 30 0 .mw.8 8 w:::: 33: :0 :00::0 0:: 88808 80880: :0 08: 8 08808 :80: 80:: :08:
: :. Z . .: .: .88: 8:88 :8 83:00:00 8 0: 8:0: w::8_: 2:8: 88: :08:
. :0 £088 :0 800::0 :08: 8:: :0 3:58:88 :08808 3 8878880:
:880M :88: 8:80: 808:0 w88_: >:8: 8 800: ”08 803068: 8038005. v5
””00 88:0 2 3M8: 800: 3: 008:8:8 88:0: 833 .388: 8:08:00:
: 0.088 :2: 8268 8:8 33: 8:2: 25:: a: 08.: 5: 85a: :8: 8:8
. 8:8 855 33: m a: :0:30:: 8083098 888: 8: 0: 8:08 8.88380 m::::<
888800 308:8 80388: ::0308:: 88:00:00 0888 :0:::<

8:50: ._.m 28:.

 

.88 :0 80:

x 85 8803 :0 883:5
8w8_ m 8: 880:8 mm?
888:8 8:88 32::
885:8: 8588: 8:
:8 888 8B 53:8
888 888008 :0: 80:
:8 880088: 88308::
0: 8:88: 8: 80:80

:0 55890:: 88885

888:: m9< ”85:5“:

.8: 8050:: 8:885:00
:8 53:8 83:8
8:058 8888888
:0050538: (:0

80823”: 88 8:885:00
:0 858 3:80:

8 w:_8: 8:83 >58: :0
88:_>m_ .88 8:885:00
:0 b_>:8 83:8

8:058 :0::08 .«0 880
:0 88:38 02 ”85:5:

8880588 88,—.

.m80::00 :0 8:: 5:38:88

.:8 80828888 :58:

:8 88— 888 8 88.8: ::<

.50: :8 6:8: a3058 “880:8
we]: :0 88:8 :8 88:88:88
:0: 20:80 888w 88 :8 888
3 83838 :8 038m 2055
83 80:80 5:5 880088:

:8 .::0::00 5:5 880088

608588 (:0 b25308
8x88 0: :82 85:05:52

88005 852: :8 88:50:

853:0: :0 80:88::

8288> 88780509882

9: 888 5 28. :8800 8:88:
:0 :82 05 .:0::08 0:88:85
:0: 85:5 :8282 a: 5800 :m
8:308 :0::08 :0 80888: 3
”88880 582.8388 >888
.5 mm: 8:885:00 :0 :8 258
>838: 88: m0 8:238:08
888:8 0: 8:08 :32: a 83

8880050 888 :8

8:88 mo? 5222: :5:

o:_8xm $588 b5? :8
8:888: 8: aw:_88 8:

”880:8 mn=< 0: 888::
.:0 882.888 88%: .88

:0: :0 8: 8.: 8: 80:80
88::0 :0 8:0:_::00
:0888: 880088
808588 85 :0: 8:08
@8858: :0 :0:88:<

8:052: 8:885:00
82 mm: :0 83:8
828m 8885 88 :88

85 30:: :0::08 :0 800 .2

.8388: 83:3. 88:8: :0
8:888:88 :0050szw8:
:0 38:8 8 x00. .82

8:052: 8:885:00 88:

:0 :8 8:8 £838: 88:
:0 :88 85 88 8:: 85

m3“: :8 88:0: :8:::8>0w

:0 308,8 :5 83:8

 

$8800

388% 805::5

010328: m 83:80:00

82: 828
26388: 5882
8082 2: 8 $83
a: :5 :8
.:0 ::o:8_ :m
83:85.
:288m
83 58> «db
.:_0 88»
2-2 5:53 88
530:0 3:8”— 8 8
:0 833m 88><
8:082 <mD
298m :05:<

.A:.::0Uv .~.m 28B

102

 

.m:o:m:0:0w

:0w:=o> .8 030 05 :m cm
08:: 0mm :0w::o> 0 :0 220
“ma 02E 8 300:— 0.8::
03 300.00% mso:0w=0£
.205533 w:=oo:0m
8:3: 00 “moo £55250
SawE a 8 : 8:20:
£235: 2 :5 580

E 85:: 0:: 3 03038::
8: fl 5:00:60 00 0:: 0gb

.mA .020

.02 ”:280—3 _002: :00 38%.??—
.:0w:mm :50 525—0020 A350: .0008 3
BEEQEQQE ”1:2 0508058-:8 a
5:5 83030305: 00 830:3 EVE:
wfififi a w::m:::m0 3 b_0:0wo:0:0:
33030:: :8 28:00 .8- :0

“:6: $2: 0 @8205 3 30:30:03?
0:500: :8 $5800: Exozmxm

.020 0:88 8 “we
0:: 33.2: :_w0n £00508
53:08:50 00 55:00:

95 0:: 500:8 .w.0v 80:5

30:00 @6500 («o 250:
:m 0050:: 3m: 0000— 0::

03800.00 50:: 05 m: :0803
50:82 as 003080: E mam

:0 320:5 28. s m a?
:0803 :00 805% 2:00:

Nwlmvo _
:00389 Eon

 

83,5 233 5 mm: 0o 8880a ow“ 2:3: a 683
00:0::QO x00 0:0 .85 mtonoo 0:: 29:3 03:3 mEm: 3 0:80.00 3:35:00 :0603 A30: «0:052
$0508 00 3080:0900: Ema—0:»: 0:28 5:5 0:0 £20550 .m0 000 .D_Q<Zm
9 0:: >530: :8 58:: «N: :0 5:00:00 0050:: v c 00532
n a :92 wESw 0020080: 25508
09:8 05 0005052: :23? 00 885200 :0 38:50:00 @8380: 8 92:0
w:t0::0 («o 0:30: 05 5:5 «3000.00 Eta: 0w0:0>0 0000850 0 _ 3 A 0:0.WHMN_ :0.“
0036800 30333: 08 r: wmmmgma 0.._m.::0
5:00:00 0:0 8:022:00 :38: A5 0:: 01:50:09 0 mm 030> o :oswmw: :3 o 0mh0:0:.m
08098 3 BESS—00,0 :8 0.050008: 0 . . ._. 0 .
. . . . . 0whfi 5:3 Ea $30.00 0:0
8:0 :020 8 03:20: 95 m0mD 530:3 Ewan: 038:8 :0803 50:82 o 3.5 00
5:5 6:0 @6098: 0:0 8 2000:: :osabfi 08: 80:86 m0mD wSS» :0 . 0 300058“:
5an :0w::ox ”me—ufim 5 s . m
3:05:50 30:25 _00EQEm 0.830805 33:00:00 2ng 853<

.5280 .3. 2%:

103

Risk altering consists of information that may alter perceived risk of pregnancy for
sexually active teens. As a result sexual activity increases but the effect on teen
pregnancies is ambiguous. Risk revealing education provides knowledge on
contraception methods and their correct use. Outcomes of sexual activity and pregnancies
depend on whether perceived risk of pregnancy is revised upward or downward but both
outcomes move in the same direction. Utility altering education changes expected utilities
of different outcomes of sexual activity. Predictions of the model with respect to utility
altering education are not conclusive, sexual activity and pregnancies change depending
on how relative utilities shift and both move in same direction. This model suggests that
magnitudes of these predictions vary with individual characteristics: gender and access to
alternative sources of information. Oettinger empirically estimates hazard models to
analyze the effects of reported enrollment in sex education on teen sexual activity and
teen pregnancy. He finds that sex education is associated with earlier sexual activity for
females and (weakly) with pregnancy for some female groups. For both sexual activity
and pregnancy, effects are larger for females with fewer sources of information relative to

males and groups with more sources of information.

Tremblay and Ling adapt Oettinger’s model to analyze the effect of AIDS education
on the abstinence-intercourse-condom use decisions. They adjust the original
classification of education types, i.e. risk altering, risk revealing, and utility altering, for
AIDS education. They empirically estimate a multinomial logit model to analyze
probability of abstinence and intercourse, finding that AIDS education increases

probability of condom use relative to unprotected intercourse but it does not encourage

104

sexual activity. Additionally, they find that AIDS education has a larger influence on

women than it does on men.

Paton (2002) also adapts Oettinger’s model to analyze the effect of family planning
on British teenagers’ sexual behavior. Predictions of the model are extended to examine
aggregate level data. Theoretical predictions of this approach may be summarized as
follows: (i) increased availability of family planning leads to an increase in the rate of
teenagers’ sexual activity and (ii) impact on pregnancy rates and whether access to family
planning has greater or less impact on underage abortions than on overall conceptions are
ambiguous. Paton uses a natural experiment generated by a policy change, the 1984
Gillick Ruling, which prohibited teenagers below 16 from obtaining advice on family
planning without parental consent. He finds no evidence that provision of family
planning information reduces underage conception and abortion rates but finds that child
care services and the unemployment rate are positively associated with pregnancy rates
while proportion of young people staying in post-compulsory education has a negative

effect on both pregnancy and abortion rates.

Averett et al. (2002) analyze the effects of government policies, laws, and
neighborhood characteristics on sexual activity and contraceptive use among American
female teenagers. Policies and laws are included in the form of the existence of an
abortion provider in a county, availability of Medicaid funding for therapeutic abortion,
existence of consent law in state (need of parental consent for abortion), and family
planning clinics in county. Neighborhood effects are measured by the proportions of non-

white populations and median income. Averett et al. do not find any evidence of costs of

105

abortion affecting sexual activity or contraception use. But they find evidence that family
planning clinics have a positive effect on contraception use, and that neighborhood

characteristics affect both sexual activity and contraception methods employed.

Miranda empirically estimates discrete time duration models of first birth, finding that
younger Mexican women are postponing first birth relative to older ones. He also finds
that Catholicism and education are negatively associated with the hazard of entering
motherhood, while, ceteris paribus, indigenous women are more likely to have their first
child at a younger age, more so in the case of younger generations. There is evidence on
effects of religion, education, and ethnicity varying across cohorts. In this paper, I extend
Miranda’s work on Mexican women in three ways. Because I am interested in changes in
sexual and reproductive behavior, I analyze a variety of outcomes seeking to complement
his findings. Miranda indirectly looks at the role of family planning campaigns by
defining cohorts of women who “were already at risk, or about to become at risk, at the
time of the policy innovation [1963-72]”, who became “at risk after the start of the new
policy [1973-82]”, and who were born between 1942 and 1962. Since I am trying to
establish a (causal) connection between differences among cohorts and the
implementation of public education policies, I directly include covariates that capture the
effect of such campaigns besides 5-year cohort dummy variables. Due to data restrictions
Miranda does not observe actual menarche age and has to assume 12 years old. I am able
to calculate durations with respect to actual (reported) menarche ages—except in the
cases of contraception method and condom use in which duration is measured with
respect to age of first sexual encounter, and second pregnancy in which duration is

calculated with respect to first pregnancy.

106

3.4. Econometric Models

I estimate duration-hazard models following econometric treatments by Miranda and
Oettinger (1999). As indicators of sexual and reproductive behavior, I specifically
analyze first sexual encounter, first and second pregnancies, first use of contraception
method, and first use of condom by woman’s partner. I also estimate a multinomial logit
model of contraception method use, i.e. traditional, temporary (local and hormonal), or
permanent—a procedure similar to that used by Tremblay and Ling (2005) for condom
use and intercourse among youths. This analysis of contraception methods becomes very
important in the Mexican context since, in the last decades, the demand for permanent
contraception (female operation) has dramatically increased relative to other methods

(Miranda, 2006; also see Figure A3.5).
3.4.1. Duration-hazard models

I use proportional hazard models that contain both time-varying (16(0) and time-

invariant (Z) covariates.
A[t;x(t),z,,B] = x[x(t),z,,B]/to(t) (3.1)

where 10(1) is the baseline hazard. If K[x(t), z] = exp[x(t)flx(t)+ 2,32] then

l[x(t), z, ,Bx(t): :32 ]= 20 (t)expl:x(t),8x(t) + 2,32] (3.2)

107

where ith coefficient in either vector 6,“) or ,BZ may be interpreted as the proportion by
which the hazard changes due to a one unit increase in the ith covariate (see Wooldridge,
2002 and Cameron and Trivedi, 2005). I estimate several outcomes in order to depict
sexual and reproductive behavior: first sexual encounter, first pregnancy, second
pregnancy, first-time temporary contraception method, and first-time condom use.
Durations of time to first sexual encounter and first pregnancy are calculated with respect
to age of menarche, duration of time to second pregnancy is computed with respect to age
of first pregnancy, duration of time to first use of contraception methods (temporary and
condom) are measured with respect to age of first sexual encounter. The original cross
section sample is converted into a panel-type database expanding each observation by the
corresponding woman’s age. In the panel dataset each observation is a woman-year and
the “history” of each woman is described by her (time-varying) age, durations until

failures, education, and exposure to sexual education.

I am interested in analyzing how the hazard functions shift when covariates change.

Thus estimation of 110 (t) is not necessary and I use a Cox procedure to estimate equation
(3.2). Vector x(t) includes exposure to sex education that captures either the effect of the

introduction of sexual education in schools (1974) or the program upgrading due to AIDS
epidemics (1987). Vector Z contains: (i) exposure to family planning promotion in public

health facilities; (ii) education controls: a dummy indicating whether a woman attended a
private school at highest level of schooling, interactions between private school and
exposure, and highest level attained; (iii) cohort and marital status dummy variables; (iv)

characteristics of community of residence at age 12 that serves to controlling for possible

108

unobserved heterogeneity due to, for example, differences in background like past
relative wealth or access to basic services; and, (v) state and community-level population
size dummy variables. In the cases of first and second pregnancies, I do not include a

marital status variable due to potential endogeneity.

It is worth noting that Miranda explicitly takes into account childless women, i.e.
women who may be sterile or dislike children, by including a mass point. However,
according to MxFLS sample only 2.5% (1.4%) of women have long durations to first
pregnancy—durations of more than 20 years (25 years). Since the sample consists of
women in fertile ages, it is difficult to assume that these women will definitely remain
childless, except perhaps for those over 45. Possibly a group closest to the childless
category would be that with long durations and no sexual intercourse ever: 1.8% (1%) of
women with durations of more than 20 years (25 years). Moreover, if one believes that
average sterility or preferences for children do not really change across generations, one

would not expect that the lack of control for childlessness would affect results.

3.4.2 Multinomial logit model

A woman or her partners could have used more than one birth/disease control
(contraception methods for simplicity). I group these methods in three types, traditional,
temporary, and permanent. Traditional type consists of rhythm (periodic abstinence),
withdrawal (coitus interruptus), and herbs or teas. Temporary type includes pills,
emergency pills, and other hormonal methods, intra-uterine device (IUD), and masculine
condom. Permanent types are female operation (tubal occlusion) and male operation

(vasectomy).

109

None, one, or some of three types of methods may have been used by a woman (or
her partner). Let CM be a set of all possible methods—for convenience let it include a
“no method” possibility—, then CM = {never used any method, used only traditional
methods, used traditional and temporary methods, used traditional and permanent
methods, used only temporary methods, used temporary and permanent methods, and
used only permanent}. I am interested in three possible outcomes: (i) never used a
contraception method or have used only traditional types, (ii) have used temporary types
(may have used traditional but never permanent), or (iii) woman or partner presently use

permanent, i.e. either is operated (may have used traditional or temporary). I construct a

variable y that categorizes such choices, taking a value j (j = O, 1, 2) for each alternative

in (i) thru (iii), respectively. Table 3.2 presents a matrix summarizing CM and y, for

example, if a woman is operated (permanent) then her response is categorized as 2, even
if she used other types of methods during her life; or, a woman’s response would be 1 if
she has used temporary methods but never a permanent one, even if she has used

traditional methods as well. The unconditional probabilities associated with these choices

are P( y = 0) = the probability of not using contraception methods or having used only
traditional types P(y = 1) = the probability of having used temporary contraception
methods but never permanent, and P( y = 2) = the probability of using a permanent

method.

110

Table 3.2. Contraception methods use matrix.

 

 

None Traditional Temporary Permanent
None None — - -
Traditional _ Traditional Trad1t10nal and Tradltlonal and
Temporary Permanent
Temporary and
Temporary - Temporary Permanent
Permanent - Permanent
y= l 2

 

Using a binary response model would not capture all possibilities of contraception

methods potentially available to a woman. Also, since a woman’s decision to use

permanent rather than temporary contraception methods likely reflects a very different

life situation, it seems pertinent to analyze the former apart from the latter. Thus, I use a

multinomial logit model in order to fully understand decisions over contraception:

P(y = jlx)=exp(x§j )/[l + i exp(x§h ):l, j: 0, 1,2 (3.3)
h=l

where x is a vector of covariates and 5 a vector of parameters (Note: elements in

these vectors differ from those in section 3.4.1). I estimate the probabilities of using

temporary and permanent methods with respect to the baseline j = 0, never used

111

contraceptive methods or have used only traditional methods. Variables in vector x are

the following: exposure to sex education program in school, age of first sexual encounter,
education controls (same as in duration models), cohort and marital status dummy

variables, number of children, and state and population size dummy variables.

3.5. Data and empirical analysis

3.5.1 Data

I use data drawn from the module on Reproductive Health of the Mexican Family
Life Survey (MxFLS-1). The Reproductive Health Questionnaire (RHQ) was applied to
all women of fertile age (14-49 years old) within each household. The sample size is
8,500 households and representative at all national, regional, and urban-rural levels. The

RHQ gathered information on history of pregnancies, contraception, and sexual health.

Table 3.3 presents summary statistics of relevant variables. Total sample of women in
fertile age is 8,743; 2,034 of the original sample are women between 14 and 19 years old.
The number of observations after excluding teenagers is 6,709 and this is the final sample
I refer to hereafter. I exclude teenagers because they may not have completed the highest
level of schooling by the time of the survey, so in their case this would not be a good
control of completed education. Average age is 33.1 years and the cohorts are distributed
as follows: 11.5% ofthe women were born in 1953-57; 14.8% in 1958-62; 16.7% in
1963-67; 19.2% in 1968-72; 18.2% in 1973-77; and, 19.6% in 1978-82. In terms of
marital status: 20.3% of the women report being single, 5.5% separated, 58.2% married,

2.5 divorced, 1.6% widow, and 11.9% live in consensual union.

112

Sexual and reproductive indicators reveal that 85.6% of the women reported having
had sexual intercourse, with an average duration to first sexual intercourse of 6.4 years
after menarche. 82.2% of these women have used contraception methods: 4.0%,
traditional methods; 57.4%, temporary methods; and, 38.6%, permanent methods. 23.3%
of the women who have had sex reported that their partners have used (male) condom.
On average, duration from first sexual intercourse to first use of a contraception method
is 3.6 years, duration to first use of a temporary method is 4.1 years, and to first use of
condom is 5.3 years. Of all women, 81.4% have been pregnant and their average time
duration fiom menarche to first pregnancy is 7.4 years. 75.1% of these women had a
second pregnancy (69.1% of all women) and their average duration from first to second

pregnancy is 2.8 years.

Most women were potentially75 exposed to the first in-school sex education program
(1974) for at least one year; on average, they were exposed 3.9 years. 33.3% of the
sample was not exposed to sex education either because they did not attain 5th grade of
elementary school or because they quit or finished school before 1974, and 40.1% of the
sample was exposed to sex education for 5 years. Only 33% of the sample was exposed
to the 1987 program upgrading for at least one year; on average, they were exposed 3.6
years. In total, 15.7% of the sample was exposed to the program upgrading for 5 years.
According to general education statistics, average schooling is 7.3 years; 54.0% attained

less than secondary schooling; 27.9% completed secondary school; 18.0% completed

 

75 Information on whether a woman actually received sexual education or not is not available. I assume that
the program was put into operation (i.e. contents of free textbooks actually taught in class) in all public
schools and probably in private schools since 1974.

113

high school or a higher level; and, relatively few women attended a private school at their

highest level attained (7.6%).

Exposure to birth control and family planning in health facilities began around 1972. I
assume that women who were taken care of during their first pregnancy in public clinics
and hospitals were exposed to these programs: 97.3% of the women had a first pregnancy

in or after 1972 and 59.3% received care in public facilities.

I define characteristics of community of residence at 12 years with respect to
available sources of drinking water and sanitary services. 46.0% of the women lived in a
community where WC was available regardless of the source of drinking water (bottled,
tap, or other); 21.3% lived in a community where sanitary service was a latrine or other
and drinking water was obtained from tap; and, 32.7% lived in a community with a
latrine or other sanitary service and source of drinking water was other than tap (bottled

or other).

Finally, women are distributed by current communities of residence (in 2002) mostly
in large cities and rural areas. 39.7% live in places with more than 100,000 habitants;
9.6% live in smaller cities with 15,000-100,000 habitants; 11.3% live in communities
with 2,500-15,000 habitants; and, 39.4% live in rural communities with population less

than 2,500.

114

Table 3.3. Summary Statistics.

 

Variable Mean Std Dev.
Sexual and reproductive indicators
Sexual intercourse (%) 85.6 35.2
Duration (years) 6.4 3.9
Contraception use (%) 82.2 38.2
Traditional (%) 4.0 19.6
Temporary (%) 57.4 49.5
Permanent (%) 38.6 48.7
Condom (%) 23.3 42.3
Duration any method (years) 3.6 4.5
Duration temporary method (years) 4.1 5.2
Duration condom (years) 5.3 5.5
First pregnancy (%) 81.4 38.9
Duration (years) 7.4 4.2
Second pregnancy (%) 69.1 46.2
Duration (years) 2.8 2.5
Exposure variables
In-school
Exposure-1974 (years) 3.9 1.5
0 years (%) 33.3 47.1
1 year (%) 4.3 20.2
2 years (%) 16.5 37.1
3 years (%) 2.5 15.5
4 years (%) 3.4 18.2
5 years (%) 40.1 49.0
Exposure-1987 (years) 3.6 1.6
0 years (%) 67.0 47.0
1 year (%) 4.3 20.4
2 years (%) 7.5 26.3
3 years (%) 2.4 15.4
4 years (%) 3.0 17.1
5 years (%) 15.7 36.4
In-health facility (First pregnancy) 0.0 0.0
In or afier 1972 (%) 97.3 16.3
Care in public facility (%) 59.3 49.1
Birth cohort (%)
1953-57 11.5 31.9
1958-62 14.8 35.5
1963-67 16.7 37.3
1968-72 19.2 39.4
1973-77 18.2 38.6
1978-82 19.6 39.7

 

115

Table 3.3. (Cont’d).

 

Variable Mean Std Dev.
Education
Years 731.8 385.6
Less than secondary school (%) 54.0 49.8
Secondary school completed (%) 27.9 44.9
High school or higher (%) 18.0 38.5
Attend private school (%) 7.6 26.5
Age (average) 33.1 8.4
Marital status (%)
Single 20.3 40.2
Separated 5.5 22.8
Married 58.2 49.3
Divorced 2.5 15.5
Widow 1.6 12.6
Union 11.9 32.4
Community of residence at 12 years (%)
WC available and any source of drinking water 46.0 49.8
Latrine/other and drinking water from tap 21.3 40.9
Latrine/other and drinking water from 32.7 46.9
State of residence
Baja California Sur 3.9 19.3
Coahuila 5.4 22.6
Distrito Federal 2.4 15.2
Durango 5.8 23.4
Guanaj uato 6.5 24.7
Jalisco 5.3 22.5
Estado de Mexico 7.1 25.6
Michoacan 7.9 27.0
Morelos 4.9 21.5
Nuevo Leon 8.8 28.3
Oaxaca 6.5 24.7
Puebla 5.2 22.2
Sinaloa 8.2 27.5
Sonora 8.4 27.7
Veracruz 9.2 29.0
Yucatan 4.6 20.9
Community of residence population size
> 100,000 39.7 48.9
15,000 - 100,000 9.6 29.5
2,500 - 15,000 11.3 31.7
< 2,500 39.4 48.9
Sample size]: 6,709

 

116

3.5.2 Empirical analysis

Sexual education programs implemented since 1974 begin at the elementary level in
5‘h grade and 6th grades, and continue through all three grades at the secondary level—the
structure of education levels in Mexico is described in Table 3.4. This implies some
women were not exposed to in-school sexual education: (i) those not attaining 5th grade
before and after 1974 and (ii) those that finished secondary school before 1974. I exploit
this fact to identify the effect of exposure to sexual education. I construct specific year-
grades for each individual and an indicator for whether a woman completed targeted
school grades before or in/after 1974. I then interact years of schooling attainment with
the 1974-dummy to obtain years of exposure (YRSEXP) to program (see Figure 3.2).
Figure 3.3 shows the average exposure to the 1974-program by cohort of birth, women
born before 1959 were practically not exposed. Posterior generations were increasingly
exposed but it is the birth cohort of 1960-1965 the one that faced the most important
transition. In Figure 3.4, one can actually observe that, although duration fi'om first
sexual intercourse and contraception use was already declining, a significant

discontinuity shows between the generations born in 1959 and 1960.

To analyze possible effects of program-upgrading due to social and public health
concerns, I design a variable that captures the effect of programs since 1987, when the
AIDS epidemic led to an emphasis on protection and condom use. This variable is
constructed in the same way as YRSEXP, adjusting calendar-school year to 1987. Both

years of exposure-to-education variables are treated as time varying covariates, x(t) in

equations (3.1) and (3.2).

117

Table 3.4. Mexican education system: elementary thru high school.

Grades

School Level (S.ch001-years)

 

 

 

 

 

Elementary 15‘- 6th Starting at 6 years old.
Secondary 1St - 3rd
High school 1St - 3rd Individuals are around 18 years old when graduating.
Required to enroll in college.
1953-62 1963-72
5 o o o o o o o o o o o o o o o
A 0 -< I). O C . O I O . O 0 O C O . C C O C C O . O O .
g 1973-82 1983-88
0 5 - C . O O O O O C . O O . C C O
3)
o
0" O O
X
LL] 0 I
O — C O O O O O O O C O
O 5 10 15 20 0 5 10 15 20

Education (years)

Figure 3.2. Potential exposure to sexual education program at different education years

by cohort groups.

118

 

 

4.5

3.5

2.5 e

1.5 ,.

0.5

 

 

 

1

1961
[£61
{£61
SL61
LL6IC_
6L6I
1861

$961
9961
L961
6961
£961
S961
L961
6961

Figure 3.3. Exposure to 1974-Program, average by birth cohort.

10

 

 

 

 

1 1 7.1 1 1

1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965

—— lst Sexual Intercourse -3K— lst Contraception
lst Temporary Method —A— lst Pregnancy
—B-— 2nd Pregnancy

Figure 3.4. Sexual and reproductive behavior, average durations by birth cohort.

119

 

 

To isolate the effect of exposure from that of schooling, 1 control for the highest level
of completed education (a time-invariant variable). In addition, notice that official
textbooks are not compulsory for private schools, so to control for this potential source of

bias, I include dummy variable indicating whether woman attended a private school.

Public initiatives in the 705 also involved the promotion of birth control and family
planning in the health sector. In order to estimate the potential impact of these policies, I
estimate the hazard of a second pregnancy, conditional on having been pregnant a first
time. This model includes additional variables indicating whether the first child was born
in or after 1972 and whether delivery of first child was in a public facility. For a more
precise assessment of the effect of obtaining information on birth control in public
facilities since 1972, I include an interaction of public facility and year-1972 dummy
variables. It is worth noticing that those women who used public health facilities in the
19705 may have been different from those who did not. For example, wealthier women
may have been less likely to use public facilities. Ideally, one would look at demand
indicators from that period, but they are difficult to obtain. Taking this caveat into
account together with the shortcoming that the following data include men, I examine

recent evidence in order to get an idea of what types of people use public facilities.

Available information from the National Health Survey (ENSA, 2000)76 reveals that
the proportion of insured people in the upper quartile of income that prefer to attend
public facilities is relatively smaller (70.9%) that those in the lower quartiles (80.7%-
84.2%). Also, proportion of insured people in the upper quartile (24.5%) that prefer to

use private services is relatively larger than those in lower quartiles (12.1%-l4.8%).The

 

7" Valdespino et al. (2003 ).

120

same pattern is observed for uninsured people, but differences between quartiles are
bigger. However, actual use of inpatient services77 do not vary as much across income
quartiles: 66.5% of the highest quartile used public health facilities versus 70.5% of the
lowest quartile, and 29.9% of the highest quartile used private facilities versus 25.5% of
the lowest quartile. Thus, bias perhaps is not a serious problem and besides can be
reduced by controlling for other covariates, such as educational attainment and

characteristics of the community of residence at age 12.

All specifications control for cohort, state, and population size (community of
residence) dummies to account for socio-cultural factors that may differ across
generations, regions, and rural/urban areas. In the cases of first sexual encounter, first
contraception method, and type of contraception method used, I also control for marital
status—I do not include marital status in analyses of pregnancies since it may be

endogenous.

3.6. Results

3.6.]. Duration models

Tables 3.5 and 3.6 contain results of the proportional hazard Cox estimation. These
estimates are hazard ratios, exp(,Bi), rather than coefficients (6078. Results in terms of
coefficients can be found in Tables A31 and A32 in Appendix 3. In Table 3.5, columns
(1) thru (3) present estimates of models for first sexual encounter, first pregnancy, and

second pregnancy, respectively. Notice that if the hazard ratio is greater (smaller) than

 

77 Approximately, 71% of total inpatient services were used by women, around 49% of them used these
services due to pregnancy, birth, and related motives. Of all women using inpatient services for pregnancy
and birth attention 71% used public services.

78 Note that ,8, : exp(,6,) — 1 (Cameron and Trivedi, 2005).

121

one, the coefficient is positive (negative): if exp(,Bi) > 1 then ,6, > 0 and if exp(,Bi) < 1

then ,8; < 0.

Exposure to initiatives

Results show that exposure to in-school sex education: (i) increases the hazard of first
sexual intercourse; (ii) does not affect the hazard of first pregnancy; and, (iii) decreases
the hazard of second pregnancy. Exposure to sex education increases the hazard of first

sexual encounter if a woman attended a public school.

For example, for women with no exposure to sex education and who attended a
public school, a year of exposure would increase their hazard of first sexual encounter by
4.1%, other things equal. If a woman attended a private school, exposure to sex education
has a positive (2.3%) but not significant effect on the hazard of first sexual encounter.
Exposure to sex education has a positive but small and statistically insignificant effect
(0.3%, if public school) on the hazard of first pregnancy. The effect on the hazard of first
pregnancy of exposure having attended a private school is larger but still not statistically

significant.

The hazard of second pregnancy is reduced by 2.7%, if a woman attended a public
school but not significant if a woman attended a private school. In any of the cases the
interaction between exposure and private school is significant. Given that the use of free
textbooks is not compulsory for private schools it is likely that actual exposure to sex

education of women who attended private schools differs from that of women who

122

attended public schools. Moreover, many private schools are also religious and may try to

avoid teaching sex education entirely or delay it until students are older (15-18 years old).

Women who had a first pregnancy in or after 1972 are less likely to become pregnant
for a second time (column 3). The exposure to campaigns in public health facilities is
captured by an interaction term between having a first pregnancy in or after 1972 and
having been cared in a public facility during first pregnancy. The exposure to campaigns

reduces the hazard of becoming pregnant for the second time by 31.4%.

Table 3.6 presents estimates of the first time a temporary contraception method was
used and the first time a woman’s partner used a condom. Both columns include the same
covariates except for the exposure variable: In column (1), the exposure variable is
designed to capture the introduction of sex education in schools in 1974, whereas the
exposure variable in column (2) captures the program upgrading in 1987. Evidence
reveals that exposure to sex education increases the hazard of using contraception
methods by 7%. When interacted with private school, the exposure effect becomes
negative and significant, reducing the hazard of using a contraception method by almost
7% (6.6%). The total effect of exposure for a woman who attended a private school
practically disappears. Again, it seems that religiosity in private schools may be playing a
role here by promoting traditional contraception methods over temporary ones. In the
case of condom use, exposure to sex education has a positive but small (1.2%) and not
significant effect, also if a woman attended a private school. Notice that a more complete
analysis of the effects of sex education on condom use would involve using a male

sample as well.

123

Other covariates

So far I have discussed the estimates of the interaction between exposure and private
school. However, notice that the estimated coefficients of the private school dummy are
in all cases, except second pregnancy and condom use, relatively larger and statistically
significant. I mentioned the plausible role of religious schools in providing “sex
education”. Actually, Miranda finds that being a Catholic reduces the risk of first birth
even among younger generations of Mexican women. The Catholic Church prohibits the
use of ‘non-traditional’ contraception methods (e. g. pills and condoms), apparently
increasing risk of pregnancy. But it also discourages sexual activity out-of-wedlock,
which may translate into the delay of first sexual encounter [and potentially first

pregnancy] (Miranda, 2006).

However, other factors associated with attendance to a private school could be wealth
and, with it, unobservables as, for example, perception of better prospects for education
investment that may translate into higher opportunity costs. In turn, a woman with higher
opportunity costs could postpone first sexual intercourse if she perceives that risk of
pregnancy is high, and/or use contraception methods to delay pregnancy if sexually
active. The private school estimated coefficients are negative in Table 3.5 and positive in
Table 3.6. The hazards of first sexual encounter and pregnancy decrease by 16% and
19%, respectively. The total effect on the hazard of first contraceptive use is an increase
of 48% for a woman who attended a private school and had no exposure to sexual
education, while for each additional year of exposure this total effect decreases by around

7%.

124

Attained education level has large effects on sexual and reproductive behavior,
indicating higher opportunity costs of pregnancy. The hazard of first sexual intercourse
decreases by 36%, if the highest level of schooling attained was secondary, and by 53%,
if high school or more was attained. The hazard of first pregnancy diminishes by 19%
and 27%, respectively. The effect of secondary level on the hazard of second pregnancy
is negative but relatively small and statistically not significant, while having attained high
school or more reduces the hazard of second pregnancy by 26%. The higher the school
level attained the larger the hazard of using contraception methods. The hazard of using
temporary contraception methods increases by 16%, if secondary, and 24%, if more than
secondary. The hazard of partner using condom increases by 73% and 151%,
respectively. In the next subsection, I present complementary results on the probability of

using different types of contraception methods.

Finally, estimates of cohort coefficients reveal that relatively younger women are
having first sexual intercourse earlier than did older ones. The effect on hazard of first
sexual intercourse is larger for cohorts born after the early 19705. Younger women, born
since the late 19605, are also delaying first pregnancy. But this effect is only statistically
significant for those born in 1978-82. Also notice that the effect of delaying second

pregnancies is not seen until the late 1978-82 generation.

125

Table 3.5. Cox-proportional hazard models. Hazard ratio estimates.

 

(1) (2) (3)
Exposure to program in school1 1.041 1.003 0.973
(0.013)*** (0.014) (0.014)“
Education controls2
Private school (highest level) 0.838 0.814 0.988
(0.070)** (0.074)“ (0.120)
Exposure * Private school 1.023 1.011 0.999
(0.021) (0.023) (0.030)
Highest level attained
Secondary 0.641 0.726 0.928
(0.029)*** (0.035)*** (0.050)
High school 0.468 0.411 0.744
(0.022)*** (0.022)*** (0.045)***
Exposure to program in health service3
First child born in or after 1972 0.615
(0.077)***
Public health facility birth of first child 1.265
(0.237)
Interaction 0.686
Year 1972 * public health facility (0.130)**
Cohort dummies4
1958-62 1.074 1.138 1.108
(0.053) (0.055)*** (0.061)*
1963-67 1.051 1.053 1.186
(0.059) (0.062) (0.072)***
1968-72 1.039 0.979 1.086
(0.059) (0.059) (0.067)
1973-77 1.177 0.957 1.019
(0.068)*** (0.059) (0.066)
1978-82 1.475 0.783 0.770
(0.089)*** (0.052)*** (0.060)***
Single woman (dummy variable) 0.138
(0.008)***

 

126

Table 3.5. (Cont’d).

 

(1) (2) (3)
Community at 12 years5
Latrine/other and 1.087 1.103 0.970
drinkable water from tap (0.038)** (0_042)*** (0.039)
Latrine/other and 1.074 1.083 1.016
drinkable water from other source (0,037)” (0,040)“ (0,03 8)
State and population size dummies6
Baja California Sur 1.144 1.243 1.106
(0.108) (0.122)M (0.124)
Coahuila 1.129 1.125 1.185
(0.106) (0.117) (0.132)
Durango 1.087 1.226 1.250
(0.097) (0.118)** (0.134)**
Guanajuato 0.709 0.680 1.234
(0.064)*** (0.067)*** (0.138)*
Jalisco 0.774 0.894 1.211
(0.071)*** (0.085) (0.135)*
Mexico 1.080 1.231 1.186
(0.096) (0.116)** (0.126)
Michoacan 0.941 1.029 1.211
(0.083) (0.096) (0.127)*
Morelos 1.070 1.118 1.125
(0.103) (0.114) (0.128)
Nuevo Leon 0.796 0.907 1.054
(0.069)*** (0.082) (0.110)
Oaxaca 1.204 1.183 1.190
(0.114)" (0.118)* (0.131)
Puebla 1.087 1.084 1.250
(0.108) (0.111) (0.138)”
Sinaloa 1.099 1.235 1.262
(0.096) (0.114)** (0.129)**
Sonora 1.088 1.074 1.137
(0.096) (0.099) (0.122)
Veracruz 1.092 1.034 1.030
(0.096) (0.096) (0.107)
Yucatan 0.773 0.844 1.098
(0.070)*** (0.082)* (0.124)

 

127

Table 3.5. (Cont’d).

 

(1) (2) (3)
Population > 100,000 1.054 1.056 0.931
(0.034) (0.036) (0.033)**
Population 15,000 - 100,000 0.974 1.076 0.979
(0.045) (0.055) (0.049)
Population 2,500 - 15,000 0.956 0.959 1.008
(0.041) (0.046) (0.049)
Observations 54,913 62,63 8 20,434
Pseudo R-squared 0.03 0.01 0.00
Log-pseudolikelihood -43,829.96 -42,860.19 -31,002.88
Wald Chi2 1840.25 943.39 298.10

 

Robust standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%

Column (1): First sexual encounter. Duration calculated with respect to menarche.

Column (2): First pregnancy. Duration calculated with respect to menarche.

Column (3): Second pregnancy. Duration calculated with respect to first pregnancy.

' Potential exposure to in-school sexual education programs calculated as the interaction between attained
education and calendar years of school attendance for 1974 or after.

2 Other education covariates: dummy variable indicating the highest level of schooling an individual
attended, representing a higher opportunity cost of becoming pregnant and possibly more access to
information and/or understanding of contraceptive method.

3 Health sector covariates account for public initiatives for promoting family control.

4 Baseline: women born between 1953 and 1957, i.e. who were 40-45 years old in 2002.

5 Baseline: WC available and any source of drinkable water (bottled, tap, or other).

6 Baseline: Mexico, DP and communities with <2,500 habitants. Dummies are of current state of residence,
using state of residence at 12 makes no difference on estimated coefficients.

128

Table 3.6. Cox-proportional hazard models. Hazard ratio estimates.

 

(1) (2)
Exposure to program in school (1974)1 1.070
(0.016)***
Exposure to program in school (1987)2 1.012
(0.029)
Education controls3
Private school (highest level) 1.480 1.166
(0.199)*** (0.143)
Exposure * Private school 0.934 1.018
(0.030)** (0.049)
Highest level attained
Secondary 1.160 1.727
(0.066)*** (0.135)***
High school 1.238 2.506
(0.081)*** (0.228)***
Cohort dummies4
1958-62 1.259 1.807
(0.074)*** (0.253)***
1963-67 1.269 2.437
(0.082)*** (0.332)***
1968-72 1.517 3.869
(0.100)*** (0.522)***
1973-77 1.828 5.940
(0.124)*** (0.872)***
1978-82 2.020 7.182
(0.147)*** (1 .252)***
Marital status5
Separated 1.477 1.023
(0.142)*** (0.169)
Married 1.867 1.328
(0.140)*** (0.153)"
Divorced 2.025 1.296
(0.232)*** (0.245)
Widow 1.912 1.195
(0.243)*** (0.324)
Living with partner (not married) 1.649 1.083
(0.134)*** (0.145)

 

129

Table 3.6. (Cont’d).

 

(1) (2)
Community at 12 years6
Latrine/other and 0.974 0.949
drinkable water from tap (0.039) (0.075)
Latrine/other 0.823 0.936
and drinkable water from other source (0,034)*** (0,078)
State and population size dummies7
Baja California Sur 1.677 1.441
(0.211)*** (0.298)*
Coahuila 1.256 0.795
(0.154)* (0.169)
Durango 1.464 1.034
(0.172)*** (0.210)
Guanajuato 0.743 1.244
(0.096)M (0.254)
Jalisco 1.014 1.424
(0.126) (0.279)*
Mexico 0.948 0.941
(0.114) (0.190)
Michoacan 0.796 0.847
(0.096)* (0.173)
Morelos 1.046 0.878
(0.136) (0.192)
Nuevo Leon 1.128 0.824
(0.132) (0.166)
Oaxaca 0.755 0.681
(0.096)** (0.150)*
Puebla 0.639 0.426
(0.083)*** (0.104)***
Sinaloa 1.938 1.372
(0.220)*** (0.260)*
Sonora 1.288 1.408
(0.151)** (0.270)*
Veracruz 0.950 0.664
(0.113) (0.135)**
Yucatan 0.988 0.788
(0.125) (0.173)

 

130

Table 3.6. (Cont’d).

 

(1) (2)
Population > 100,000 1.101 1.143
(0.042)** (0.084)*
Population 15,000 - 100,000 1.125 1.131
(0.063)** (0.122)
Population 2,500 - 15,000 1.083 1.298
(0.058) (0.127)***
Observations 45,197 80,903
Pseudo R-squared 0,02 0,04
Log-pseudolikelihood -31,655.00 -10,398.62 '
Wald Chi2 1260.60 843.73

 

Robust standard errors in parentheses

* significant at 10%; ** significant at 5%; *** significant at 1%

Column (1): First contraception method. Durations are calculated with respect to first sexual encounter.
Column (2): First condom use. Durations are calculated with respect to first sexual encounter.

' Potential exposure to in-school sexual education programs calculated as the interaction between attained
education and calendar years of school attendance for 1974 or after.

2 Potential exposure to in-school sexual education programs calculated as the interaction between attained
education and calendar years of school attendance for 1987 or after. Due to AIDS epidemic sexual
protection started to be emphasized.

Other education covariates: dummy variable indicating the highest level of schooling an individual
attended, representing a higher opportunity cost of becoming pregnant and possibly more access to
information and/or understanding of contraceptive method.

4 Baseline: women born between 1953 and 1957, i.e. who were 40-45 years old in 2002.
5 Baseline: Single women.

6 Baseline: WC available and any source of drinkable water (bottled, tap, or other).

7 Baseline: Mexico, DP and communities with <2,500 habitants.

3.6.2 Multinomial logit models
Exposure to initiatives

Table 3.7 contains the results of multinomial logit models for contraception method
decisions. Column (1) refers to temporary methods (2,820 observations) and column (2)
refers to permanent methods (1,790 observations), using the “never used or used only
traditional methods” outcome as the baseline for comparison (1,345 observations).

According to these results, exposure to in-school sex education is weakly correlated with

131

use of modern contraception methods. An additional year of exposure to in-school sex
education increases the log-odds of using temporary contraception methods by 0.06 if a
woman attended a public school. But an additional year of exposure to sex education
decreases by 0.06 the log-odds of using permanent contraception methods—weakly
significant. If a woman attended a private school the total effect of exposure to sex
education decreases in the case of temporary methods and increases (in absolute terms) in
the case of permanent methods, but the interaction estimate is not statistically significant

in any specification.

Other covariates

The coefficients of the private school dummy are large, positive and, although
statistically insignificant, provide some indication confirming the findings from the
duration models: women attending private schools may, probably due to unobservables
associated with wealth, face relatively higher opportunity costs. As in the duration model
results, the use of contraception methods is positively correlated with school attainment:
(i) the log-odds of using temporary methods increase by 0.30 if highest education level is
secondary and by 0.40 if high school or more; and, (ii) the log-odds of using permanent

methods increase by 0.59, if secondary, and by 0.43, if high school.

132

 

 

3.38.8 8:528 ...,...828 352.8 ...,..828 £328
:3- $2 ~22- 38 23. 92 83:8

838 .1888 82.8 388.8 82.8 ...tafl8
53 £3 83 :3 $3 $3 Pie

...tac8 1.5.92.8 1.2828 132.8 ...,...928 1288
2:3 mm; 23 8.3 65.0 $3 $88

...2828 ...,...228 $3.328 .........:2.8 1.828 352.8
33 ad £3 om: 33 ago $-88

8.1.82.8 ...1828 1.23.8 3262.8 ........82.8 8188.8
Rod 62 23 $2 8.3 83 Same
«momfigctosoo

13:8 ........:2.8 1.828 ........:2.8 1:828 ...rbfl8
593 32 8.2 M83 33 883 623. Em

.......§._.8 168.8 ILE8 3.68.8 8.2.82.8 168.8
52 83 £2 82 32 32 $288
3583 6%: «magma

£38 :88 £58 :38 $8.8 388
N88 N58 85.? :98 83.8 :3- 628 825 .. 238$

33.8 83.8 :88 638 :38 93.8
82 ”MS was $3 883 $2 :92 58:88 .628 6825
mflobconv Gomumoswm

$88 ...288 .888 ...288 1.688 ...288
898 63¢ $88 $3 $3. 523 Neewoa 2 838$

3.2898 5......288 33.858 ...,...858 1.88.8 1.888
85.8 3.3. 83. 38.8 $3. 3.3- 8:385 .38... 8.8 .8 6w<

52885; $8988. Suzanna Ehomfioh “accustom b88808.

5 S E

_ .mwofioE cosaoombcoo mo womb 8.wa 389:st Sm 2an

133

 

 

.11.: ~ fie ***Am©o.ov .11.: ~ _dV ...aiAmmodv .12.": — _dV ***Am¢o.0v
£2- was- fine- was- :2- 82o- 5&6 865683 638.6% 85858853
22 2.8 82.8 22 _8 23 _ .8 22 2.8 28 _ .8
83- 88.8 23- N38 83. ~38. 383883 63.88% 28552853
0832 S 3 3:58:50
12488 2:88
$3 8.3 28825 223 BB 83
$28 ...888
m :3 ~28 62236 288
12888 1.388 12298 1388
$2 $3 £28 Sod 68226 632
1968.8 1258.8
22 $3. 885822
:8250 6:6 momocdcwoum
:2SN8 23.238 :2SN8 £2328 ._.....2N_N.8 ...._..2§.8 .
68.. 83 EB ES 3.3 83¢ 8258 88 6:88 as, was:
122.88 12688 12.438 1.2288 1.2628 1.2538
82 8: mm: “:3 :3 82 3623
$223.8 $228.8 1.2808 ...,..2me8 1.2808 2993.8
23 68.. 83 5: 23 a2: 8825
3.2888 1288 $2828 .......2§.8 1.258 3.2388
:3 82 83 £2 RE 82 86:82
2.2888 1.28:8 t2~§8 1.26:8 $23.8 1.26:8
$2 $88 3: 89¢ $2 $08 853%
@5255:
2828882 25208th EoSflEoA Ehomfioh accustom 238952.
28 28 28

.3258 SM 2an

134

 

@325 83.8 :dev 2:de Quincy Ao—Mdv
vamd ommd- Nde oomd- acmd ommd- cgwo=>
Amfimdv Amnmdv A208 Amnmdv Am—mdv Amnmdv
waod- 5N6- mN_.ou oomdu cnod- mwmdu Naowuo>
*Awgdv Cwmdv Aw—mdv Cwmdv *Aommdv Cwmdv
Nmmd owed ommd Nwod Sand Nwod Socom
***A©Nm.ov *Lmomdv friwmmdv *Lmomdv ***8mm.ov ...Lmamdv
Goo; v2.6 @504 _ond 2 _._ OObd «CECE
...fLondv *fwcuwmdv *fLommdv ..:_:..Cuwm.ov .31.: mmdv fwivwmdv
oaad- Vmo. T woo. T wmo. T —©o.ou hmo. T @325
...memdv ...Ikwhwtov ...memdv Efflwhmdv ...memdv “I.“Lwhmdv
090.0- vowd- omhd- wood- mood- wood- «0980
:dev Amwmdv Smmdv Amwmdv ANdev Amwmdv
356 05—6 owed onfio N~ _d oh—d Good o>o=Z
Gmmdv Somdv Gmmdv Some Awmmdv Aoomdv
Nde am _ .o- covd Vomdn ommd bofiou m2802
...nIkamdv .Ichwtov ..iLh—WOV ***:hm.ov ***a~m.ov ***A:N.ov
flood- 3%.? ms. T mmwd- mmqou 3%.? 5350:0me
:dev ands :dev AOBNdV Ammmdv ahmdv
mafiou owmd- O_N.o- ahmd- mcfid- mhmd- 02x32
*Ammmdv ASN.8 13“de Comdv ...Gmmdv Candy
oomd- oomd- ~ 3.0.. memd- med- oomdu Comm—E.
***Ao.vm.ov ***Aowm.ov ***amm.ov ***AowN.ov ***on.m.ov “.iiowmdv
NVm. T wOo. T gm. T 05. T owv. T OS. T 0339830
onMdv *AhoNdv onMdv *Ahomdv CV08 Lhomdv
hhvd med Ohio momd momd mowd 33950
Awmmdv some Ahmmdv some onMdV Somdv
fiofid wood 5; .o muod hh—d Nhod Ema—MOO
***C.mm.ov Cmmdv *wammdv :mmdv trimmmdv Cmm.€
OMMA med won; @de m3“; hNNd Sm «€8sz «_mm

Eggnog bflomfioh

“accustom bauomfioh

S

328::ko EEOQEQH

E

%2885 BB cog—smog Ea 83m

.5288 .2 2an

135

3:832 oom.mv FEB moE§EEoo now ad 62on 65395 h
.Conno no .93 .3388 333 038—56 mo 3.58 Ea Ea 2823.6 03 ”eczemam o

5an BwEm noon—ommm n

.moom 5 2o E3» 913 033 on? .o._ .33 can mmi :ooEon Eon ooEoB 6::on v

nonooE 0338838 no wag—Swans: SEEN :2an85 on $88 EoE bnfimoo new =8:on wEEooon
.no “moo bEstoooo anE m wEEomEofl .3233 126365 5 won—oonom .no .32 HmonwE 06 $5865 Ennis.» x883 ”$5238 cosmos? 550 m

3% a $2

on 85223 _oonom go Boo» 895:3 can cosmos? @0583 Season :oBoEBE 05 mm noon—:28 mEEwoa :omomoooo xom _oonom-:_ on 2:898 _mtcooom N

.22» 8 .2862

865 _EooEoo new: 023 .925 moonoofi “accustom ”Eucanm .cESWEoo no: Sn :9.» mm _mcoEoE new: 022 38V :82 now moooEonv moofioE
.2388 new: oz“: “—83th .Amnoon no 88 .oooootmnm ”.6233 moo—=05 _oooEog boo com: 022. .o 2552: 0338838 new: no: o>wI 63m _

:85:on moron ao— oom 68o Eon .o>=m Eon 56:5 (no .382: .5.“ 2828 AC oesoo
63.x Eon 56:5 no gonna: how 2838 AS :an0
.385:on (no Snags: “on 2828 “A C oEEoU

£1 E ESE—Em 1.... ”fan a Emoc_:w_m .2. xxx: on ancEwi ... 385228 E 38.5 32:85 Hmonom

 

$.33 $.83 $.33 ”go 5

2.83. $.83- 3.23- Bo£_8_=-w3

m3 m3 m3 Bangm 85mm

33 a? £3 223530
...,...GSQ some ...t..@$.8 some .3638 some

~37 mod- 3?. 98¢- $2- $3. 3938
$3.8 ANNE: $2.8 :38 $2.8 22.8

83 83 ES N2 .o £3 23 25.2 - 8mm Eggnog
153.8 32.8 2.23.8 $2.8 1.23.8 $2.8

:wd mood N33 83 33 mood ooode - 25.2 833$
152.8 $2898 1.28% 3.38.8 $32.8 $98.8

«one 22 Ed :2 53 23 832 A 833$

%EEEEo onm compo—zoom coo 88m

 

“:28:th $8959

5

EoEwEEom onanH

Q

Eochom .053th

E

.6388 .2 22¢

136

Finally, the higher the age at which a woman experiences her first sexual encounter,
the less likely it is that she uses contraception methods, both temporary and permanent.
Age-cohort variables capture the lifecycle or fertility profiles that change as women get
older. The log-odds of using temporary methods increase among younger cohorts,
ranging from 0.56 to 0.97. This effect becomes smaller among those born in 1978-82 (20-
24 years old in 2002) most likely because they are actually trying to get pregnant.
Similarly, the increase in the log-odds of using a permanent contraception method starts
diminishing for the 1968-72 generation (30-34 years old), becoming negative for the
youngest women. That younger cohorts are less likely to use permanent methods reflects
that they are looking forward to start motherhood or have more children. Finally, non-
single women are more likely to use contraception methods than single women, also an

intuitive result.

3.7. Conclusions

In this chapter, I examine whether public initiatives launched in the 1970s contributed
to change sexual and reproductive behavior of Mexican women. I characterize sexual and
reproductive behavior using information on first sexual encounter, first and second
pregnancies, and utilization of contraceptive methods. I use duration-hazard procedures
to model first sexual encounter, first and second pregnancies, first use of contraception
method, and first use of condom by woman’s partner and multinomial logit models to

estimate the choice of contraception method type.

Results suggest that exposure to in-school sex education increases the hazard of first

sexual intercourse and decreases the hazard of second pregnancy but does not affect the

137

hazard of first pregnancy, regardless of the type of school a woman attended. Also, I find
that exposure to campaigns in public facilities reduces the hazard of becoming pregnant
for the second time. Additionally, evidence reveals that exposure to sex education
increases the hazard of using (all kinds of) temporary contraceptives, but this effect
practically disappears if a private school was attended. And, specifically, sex education
that emphasizes protection seems to have no effect on condom use; however, a more
accurate result would require the analysis of male behavior as well. On the other hand,
according to results on the choice of contraception types, I find evidence of exposure to

in-school sex education influencing the use of modern methods.

Other important factors associated with sexual and reproductive behavior are
completed education, private school attendance, and birth cohorts. Completed education
has negative effects on first sexual encounter and first and second pregnancies, and
positive effects on the utilization of contraception methods. The choice of modern
contraception methods over traditional is also positively correlated with school

attainment. These results may be indicating higher opportunity costs of pregnancy.

Attending a private school is associated with larger durations to first sexual encounter
and first pregnancy. Many private schools in Mexico are religious and thus religion may
be encouraging postponement of first sexual encounter and pregnancy until marriage.
However, it should also be discouraging the use of non-traditional contraception
methods, i.e. temporary; but, attendance to a private school is associated with a smaller
duration from first sexual encounter to first use of temporary contraception methods. So

maybe it is wealth influencing the latter result, since relative wealthier people—with

138

higher expectations of education investment—are more likely to attend private schools.
In this sense, a woman with higher opportunity costs may have higher incentives to use
contraception methods in order to delay pregnancy or to postpone first sexual encounter
if her perceived risk of pregnancy is high enough. Yet, this is only weakly consistent
because attendance at a private school does not have a (statistically) significant effect on

the choice of contraception method.

Finally, age—cohort variables capture the lifecycle or fertility profiles that change as
women get older, as well as generational differences. Relatively younger women are
having first sexual intercourse earlier than did older ones but are postponing first and
second pregnancies. They tend to use contraception methods earlier with respect to first
sexual encounter. Younger women are less likely to use temporary and permanent

methods than older women maybe due to seeking pregnancy.

139

CONCLUSIONS

This study consists of three essays on the economics of health using the first wave of
the Mexican Family Life Survey (2002). The first two chapters focus on school age
children and the third one on women in fertile age. Chapter 1 addresses the questions of
whether overweight has increased among Mexican school age children and, if so, what
factors may have contributed to this. Chapter 2 analyzes whether more educated mothers
have healthier children as measured by hemoglobin concentrations in blood and through
which possible ways may maternal education operate to affect child health. Chapter 3
deals with whether public initiatives launched in the 19705 affected sexual and
reproductive behavior of women. These works contribute to health economics literature

in several ways.

The first proposes a theoretical model of a mother’s allocation of food and time. It
complements existing empirical evidence of an increase in prevalence of overweight in
Mexico in recent years. Moreover, it analyzes the case of school age children, a group
largely overlooked by nutritional and health national surveys. In particular, it provides
evidence supporting that Mexican children are heavier than before, and that some
indications that contributing factors might be an increase in mother’s BMI especially for
girls, an increase of maternal labor supply in the case of boys, less presence of father at
home for boys, and an increase in adult males BMI at a community level. Future
research should investigate the channels through which community level patterns affect

individual children’s behavior with respect to, for example, dietary and activity habits.

140

Second, the estimates of maternal education effects on child hemoglobin levels
complement the extensive research on the function of parental role in child health, mainly
measured by anthropometrics, mortality, immunization, and dietary intakes. Also, since
labor market participation of Mexican women is relatively low, so that market returns to
schooling fail to fully capture the impact of educating Mexican women, estimated effects
of maternal education on child health provide an appraisal of non-market returns to
schooling. Estimated effects of maternal education are positive, although small in terms
of falling into or recovering from anemia. Evidence suggests that maternal education may
be operating with public health services (clinic/hospital) through a substitution effect.
Additionally, there is some indication that mother’s labor supply may not be endogenous,
which is not implausible since low hemoglobin levels are not obvious to the mother. Yet,

further analysis would be needed to confirm this.

Finally, the third essay is distinguished from previous studies because it explicitly
deals with the effect of sexual education, birth control, and family planning public
initiatives on the reproductive and sexual behavior of Mexican women. More precisely, it
analyzes possible changes in sexual and reproductive behavior due to exposure to
programs in schools and in health facilities. Sexual and reproductive behavior is
characterized by various indicators: first sexual encounter, first and second pregnancies,
and utilization of contraceptive methods. Public programs are identified by individual
exposure to in-school sexual education and to birth control promotion in public health
facilities. Results suggest that exposure to in-school sex education is associated with
earlier first sexual intercourse without affecting the hazard of first pregnancy. Both in-

school and in-health facility exposure decreases the hazard of second pregnancy.

141

Evidence reveals that sex education may be effective in promoting an earlier use of
contraception but not specifically condoms, which would also require changes in male
behavior. On the other hand, evidence is weak regarding the influence of exposure to in-
school sex education on the use of modern contraception methods rather than traditional

01168.

142

APPENDICES

143

APPENDIX 1

Total differentiation of (l .2) and AB leads to equations A1.1 and equation A1.2:

—wdt—de—dXdep—(T—t)dw—dY
_W,a+[,/f"+wft2]dz+[v'ftp+v~f,fp]dF=Mw (A1.1)
— pd/i +[v'fF, +v"fFf, ]dt +[uFF +v'fFF +v~fF2]dF+uFXdX = Mp
—dA+uXFdF+uXXdX=O

dAB= f,dt+ deF (A1.2)

Derivation of expressions in (1.3):

From (Al .2), it is easy to see that

dAB t dF

__ : ___}. _

dw f’ dw fF dw
(A1.3)

dAB dt dF

—— = ft —- + fF -—

dp dp dp

Using (Al .1) obtain and substitute the following into (Al .3):

dt dF
—=<0tt +(T—t)cut, —=(0Ft +(T—t)wF,
dw dw

dF dt
—=€0FF —FCUF, —= (”W —sz
dp dp

144

where

 

 

 

2 2
—pu +2 u —U ‘WuH-Utt
40nd XX 1;” FF ,‘PFF=4 — .
l 1 VII
W(PuXX -uXF)+UtF
¢Ft=¢tF=4 ,

lfil

= “LYN? UtF —WUFF)+ WIIXF2 _UtFuXF
lfil

 

3

t

CUF : “Ll/(WUIF “PUtt)+UttuXF
lfil

 

,

2
, r t
UFF =uFF +V fFF +V' f1.g =uFF +flz[1-;) V",

2
1):]

I H 2
Utt=Vftt+V ft =[ T

UtF =UFt =V'ftF +V"fth =1”th +V"fth =-§[V'+flF[1-%]V"], and

[Hi < 0.
Note that U11: = UFt is “small enough” for got]: 2 gap, > 0, if

“’(“XF — PuXX )> V'sz +V"fth.

145

APPENDIX 2

60
50 ._ ~‘"
40.

30%

%

20'

 

10:

6-11 1 2 3 4 5-6 7-8 9-10 11
mos Age (years)

URBAN --$.--— RURAL + NATIONAL

Figure A2. 1. Prevalence of anemia among Mexican children by age groups.

Source: Own graph using estimates by Villalpando et al., 2003.

146

 

20 '
15 .
o\°
10 -
0 é
Central Mexico City Total
5%.: GIRLS BOYS I BOTH

Figure A2.2. Prevalence of anemia among Mexican children by regions.

Source: Own graph using estimates by Rivera et al., 2001.

147

APPENDIX 3

 

1970 I 2005
50

40

20 . ,,

0 I
No schooling Secondary or High school or College or higher
higher higher

 

 

Figure A3.1. Education attainment in Mexico (1970 and 2005).

Source: INEGI. Selected years and indicators of education statistics—Estadxisticas sociodemogrdficas:

educacién. (Downloaded from httgw’.’www inepi nob nix/cstfrlefnnlt nsnx?crr2359)_

148

70 - , , , , 7,, W7
60 ~
50 ~
40 ~
30 .
20 -

 

1 990 2000

15-29 4"“ 30-49 I 50 and more

Figure A3.2. Attitudes towards role of women as mothers in Mexico: % who thinks a
woman needs to have children to be fulfilled by age groups.

Source: Own figure using selected years and indicators of the World Value Survey, WVS. (Online Data

Analysis at http:r’x’www.worldvalucssurvcvorgx). Question: Do you think that a woman has to have children

 

in order to be fulfilled or is this not necessary? Answers: Not necessarily or Needs children.

149

60

 

 

 

 

15—29 30-49 50 and more

Disapprove 1% Approve I Depends

Figure A3.3. Attitudes towards women deciding to become single parents in Mexico: %
of approval by age groups.

Source: Own figure using pooled samples ( 1990, 1996 and 2000) and selected indicators of the World
Value Survey, WVS (Online Data Analysis at http://www.worldvaluessurvevorgh. Question: If a woman
wants to have a child as a single parent but she doesn't want to have a stable relationship with a man, do

you approve or disapprove?

150

100 ~

     

 

 

 

80
60
,\°
40
20
0
15-29 30-49 50 and more
Agree I Disagree

Figure A3.4. Attitudes toward women wanting a home and children in Mexico (%).
Source: Own figure using 1990 sample and selected indicators of the World Value Survey, WVS (Online
Data Analysis at http://www WUI ' J ' veyorw’). Question: Strongly agree, agree, disagree or strongly
disagree with the statement: “A job is alright but what most women really want is a home and children.”

These four answer categories were collapsed into two groups.

151

40:

\\ +Pills
3O ' "A" ' " “ r ' DIU

- rt. .- Operation

 

 

 

 

 

 

20 . , Male operation
——&-— Shots
10 . ’ . -—O— Local
,. —9— Traditional
S’“ ,- -r"’ “x .
0 ,

1975 1980 1985 1990 1995 2000

Figure A3.5. Demand for contraception of Mexican users (%), 1976-1997.

Source: Own figure with estimates of the Mexican National Population Council based on national

demographic surveys. Original Figure may be found in CONAPO (2000).

152

Table A3. 1. Proportional hazard models, Cox estimation. Observation level: woman-year.

 

(1) (2) (3)
Exposure to program in school1 0.040 0.003 -0.027
(0.013)*** (0.014) (0.014)"
Education controls2
Private school (highest level) -0.177 -0.206 -0.012
(0.084)" (0.091)** (0.122)
Exposure * Private school 0.023 0.011 -0.001
(0.021) (0.023) (0.031)
Highest level attained
Secondary -0.445 -0.320 -0.074
(0.045)*** (0.048)*** (0.054)
High school -0.758 -0.889 -0.296

Exposure to program in health service
First child born in or after 1972

Public health facility birth of first child

Interaction
Year 1972 * public health facility

Cohort dummies4
1958-62

1963-67
1968-72
1973-77
1978-82
1983-88
Single woman (dummy variable)

Community at 12 years5

3

Latrine/other and drinkable water from tap

Latrine/other and drinkable water from other

(0.048)*** (0.053)*** (0.061)***

-O.486
(0.125)***
0.235
(0.187)
-0377
(0.190)**
0.072 0.129 0.102
(0.049) (0.049)*** (0.055)*
0.050 0.052 0.171
(0.056) (0.059) (0.061)***
0.039 -0021 0.082
(0.057) (0.060) (0.061)
0.163 -0044 0.019
(0.058)*** (0.062) (0.065)
0.389 -0.245 -0.261

(0.060)*** (0.067)*** (0.078)***

-1.978
(0.056)***

0.084 0.098 -0.030
(0.035)** (0.038)*** (0.040)

0.072 0.080 0.016
(0.035)** (0.037)** (0.038)

*‘r-zit'fl

 

153

 

Table A3.]. (Cont’d).

 

(1) (2) (3)
State6 and population size dummies
Baja California Sur 0.134 0.218 0.101
(0.095) (0.098)** (0.112)
Coahuila 0.121 0.118 0.170
(0.094) (0.104) (0.111)
Durango 0.084 0.204 0.223
(0.090) (0.096)** (0.107)“
Guanajuato -0.344 -0.386 0.210
(0.090)*** (0.099)*** (0.112)*
Jalisco -0.256 -0.112 0.191
(0.091)*** (0.095) (0.112)*
Mexico 0.077 0.208 0.170
(0.089) (0.094)** (0.106)
Michoacan -0.061 0.028 0.191
(0.088) (0.093) (0.105)*
Morelos 0.067 0.1 11 0.118
(0.096) (0.102) (0.1 14)
Nuevo Leon -0.228 —0.098 0.052
(0.087)*** (0.090) (0.105)
Oaxaca 0.186 0.168 0.174
(0.095)" (0.100)* (0.1 10)
Puebla 0.083 0.080 0.223
(0.099) (0.102) (0.1 10)"
Sinaloa 0.094 0.21 1 0.232
(0.087) (0.093)" (0.102)"
Sonora 0.085 0.072 0.128
(0.088) (0.093) (0.107)
Veracruz 0.088 0.034 0.029
(0.088) (0.093) (0.104)
Yucatan -0.257 -0. 169 0.093
(0.090)*** (0.097)* (0.113)
Population > 100,000 0.053 0.055 -0.072
(0.033) (0.034) (0.036)"
Population 15,000 - 100,000 -0.026 0.073 -0.021
(0.046) (0.051) (0.050)
Population 2,500 - 15,000 -0.045 -0.042 0.008
(0.043) (0.048) (0.048)
Observations 54,913 62,638 20,434
Pseudo R-squared 0.03 0.01 0.00
Log-pseudolikelihood -43,829.96 -42,860. 19 -31,002.88
Wald Chi2 1840.25 943.39 298.10

 

Notes in Table 3.5.

Table A3.2. Proportional hazard models, Cox estimation. Observation level: woman-year.

 

(1) (2)
Exposure to program in school (1974)1 0,067
(0.015)***
Exposure to program in school (1987)2 0.012
(0.029)
Education controls3
Private school (highest level) 0.392 0.154
(0.134)*** (0.122)
Exposure * Private school -0.068 0.018
(0.032)“ (0.048)
Highest level attained
Secondary 0. 149 0.546
(0.057)*** (0.078)***
High school 0.213 0.919
(0.065)*** (0.091)***
Cohort dummies4
1958-62 0.230 0.592
(0.059)*** (0.140)***
1963-67 0.238 0.891
(0.065)*** (0.136)***
1968-72 0.417 1.353
(0.066)*** (0.135)***
1973-77 0.603 1.782
(0.068)*** (0.147)***
1978-82 0.703 1.972
(0.073)*** (0.174)***
1983-88
Marital status5
Separated 0.390 0.022
(0.096)*** (0.165)
Married 0.625 0.284
(0.075)*** (0.115)**
Divorced 0.706 0.259
(0.115)*** (0.189)
Widow 0.648 0.178
(0.127)*** (0.271)
Living with partner (not married) 0.500 0.080
(0.081)*** (0.134)
Community at 12 years6
Latrine/other and drinkable water from tap (.8 8:17 ) (.83).;92)
Latrine/other and drinkable water from other (0343:" ((288637)

 

155

 

 

Table A3.2. (Cont’d).

 

( 1) (2)
State and population size dummies7
Baja California Sur 0.517 0.365
(0.126)*** (0.207)*
Coahuila 0.228 -0.229
(0.122)* (0.212)
Durango 0.381 0.033
(0.117)*** (0.204)
Guanajuato -0.297 0.218
(0.129)M (0.204)
Jalisco 0.014 0.354
(0.124) (0.196)*
Mexico -0.053 -0.061
(0.120) (0.201)
Michoacan -0.228 -0.166
(0.121)* (0.204)
Morelos 0.045 -0.130
(0.130) (0.219)
Nuevo Leon 0.121 -0.194
(0.1 17) (0.202)
Oaxaca -0.281 -0.384
(0.127)“ (0.220)*
Puebla -0.447 -0.853
(0.131)*** (0.244)***
Sinaloa 0.662 0.316
(0.113)*** (0.189)*
Sonora 0.253 0.342
(0.117)** (0.192)*
Veracruz -0.052 -0.409
(0.1 19) (0.204)"
Yucatan -0.012 -0.238
(0.127) (0.219)
Population > 100,000 0.097 0.134
(0.038)“ (0.074)*
Population 15,000 - 100,000 0.117 0.123
(0.056)" (0.107)
Population 2,500 - 15,000 0.080 0.261
(0.053) (0.098)***
Observations 45,197 80,903
Pseudo R-squared 0.02 0.04
Log-pseudolikelihood -31,655.00 -10,398.62
Wald Chi2 1,260.60 843.73

 

Notes in Table 3.6.

156

APPENDIX 4

The Mexican Family Life Survey (MxF LS) is representative at the national, rural-
urban, and regional levels. It contains a variety of information of individuals, households,
and communities. This survey adapts the design of the Indonesian Family Life Survey

(IF LS), preserving comparability with Mexican population based surveys.

The first wave of the MxF LS (MxFLS-1) was collected in 2002. The MxFLS-l
sample consists of 8,400 households (approximately, 38,000 individuals) in 150
communities. These communities are located in 16 out of the 32 states, which include
Baja California Sur, Coahuila, Distrito Federal, Durango, Guanajuato, Jalisco, Estado de
México, Michoacan, Morelos, Nuevo Leon, Oaxaca, Puebla, Sinaloa, Sonora, Veracruz,
and Yucatan. The Mexican institute of statistics Instituto Nacional de Estadz'stica,
Geografia e Informdtica (INEGI) was in charge of the sampling design: a probabilistic

sample, stratified, multi-staged, and independent at each study dominion.

Household members of age 15 and older were directly interviewed and information of
children younger than 15 years was collected from the parents or responsible person. The
following list summarizes topics included in the MxFLS-1 at the individual (I) and
household (H) levels. For a detailed account of all sections contained in the

questionnaires refer to tables A41 and A42 below.

1. Expenditure and consumption (H).

157

10.

11.

12.

13.

Savings, credits, loans, assets, labor and non-labor income, public and private
transfers, and participation in social programs. (I), (H)
Family businesses and land use. (H)

School attendance, grade repetition, and completed schooling. (1)

. Time allocation. (1)

Current and past employment. (I)

Socio-demographic and geographic information of non co-resident relatives. (1)
Permanent and temporal migrations. (I)

Self-reported health status: own perception, habits and functioning indicators, chronic
diseases, and morbidity. (I)

Inpatient and outpatient use of health services. (I)

History of marriages and unions. (I)

Sexual and reproductive health of women of fertile age: history of pregnancies and
births; sexual activity; use of contraception methods; and use of related-health
services. (1)

Crime incidence and victimization. (H)

Additionally, the health of all individuals was evaluated in terms of anthropometric

dimensions, blood pressure, and hemoglobin levels. Also, individuals of age 5 and older

answered Raven tests of progressive matrices for adults and children, which measure

abstract reasoning and cognitive ability. Information on general characteristics of the

community, schools, and health providers and facilities was gathered in questionnaires

applied to political authorities and school and health facilities administrators.

158

Table A4.1 MxF LS contents: individual and household information.

Questionnaire Description

Sections

 

Book C Control book

Book 1 Household Consumption

Book 2 Household Economy
Characteristics of Adult

Book 3A Household Members

DF- Definitions

LS- Household Roster

CV- Dwelling characteristics
RC- Information for re-contact
SP- Family planning and health
EH- Schooling of household members
CVO- Direct observation of the
dwelling characteristics

NE- Interview session notes

CS- Consumption

NE- Interview session notes

SU- Land

INR- Rural income

NNA- Non-agricultural business
AH- Household Assets

CRH- Household Credit

IN- Household non-labor income

SE- Household economic shocks
VLH- Household crime
and victimization

NE- Interview session notes

ED- Education

IE- Schooling interruptions

AH- Household assets

IIN- Individual non-labor income
HM- Marital history

DH- Household decision making
MG— Permanent migration

MT- Temporary migration

TB- Employment

ATA- Adult time allocation

SHI- Individual shocks
VLI- Individual crime and
victimization

NE- Interview session notes

 

159

Table A4.1 (Cont’d).

Questionnaire Description

Sections

 

Characteristics of Adult

Book 3A Household Members
Book 4 Reproductive Health
Book 5 Characteristics of Children

ES- Health condition

SM- Emotional well-being

EC- Acute Morbidity

ATS- Self-treatment

CE- Outpatient Utilization

HS- Inpatient Utilization

CA- Insurance condition

CR- Credit

RE- Re-contact information of
relatives in the United States

TP- Non co-resident parents transfers
TH- Non co-resident siblings transfers
THI- Non co-resident child transfers
TO- Transfers of other non-resident
persons

NE- Interview session notes

RES- Pregnancy summary

HE- Pregnancy history

AC- Contraception

NE- Interview notes

EDN- Child’s education

EMN- Child employment

ATN- Child's time allocation

CEN- Child outpatient utilization
ESN- Child’s health condition
HSN- Child inpatient utilization
AUTN- Child self treatment

VAC- Child vaccination

NE- Interview session notes

 

160

Table A4.1 (Cont’d).

Questionnaire Description

Sections

 

Proxy Book

Book S
Book EA
Book EN

Characteristics of
Household Members not
present at interview

Anthropometrics and
biomarkers

Adult Cognitive Ability
Child Cognitive Ability

HM- Marital history

MG- Permanent migration: one year
or more

ED- Education

TB—Employment

CR- Credit

GH- Tastes and habits

ES- Health condition

CE-Outpatient Utilization

HS- Inpatient Utilization
CA-Insurance condition

TP- Non co-resident parents transfers
TH- Non co-resident siblings transfers
THI- Non co-resident child transfers
TO- Transfers of other non-resident
persons

RES- Pregnancy summary

HE— Pregnancy history

AC- Contraception

NE- Interview notes

 

161

Table A4.2 MxFLS contents: community information.

Questionnaire
Community Characteristics

Schools Questionnaire

Sections

PB- Population

ASC- Community social aspects
DN- Natural disasters

HI- Infrastructure history

AS— Social attendance

SED- History of the presence of schools
SM- History of the presence of health facilities
AC- Communitarian activities
MT- Transportation

INF- Infrastructure

EL- Availability of electricity
FA- Water source and sanitation
OC- Credit opportunities

I- Industry

EF- Enterprises and factories
BC- Community welfare

CM- Migrants clubs

OD- Direct observation

NE- Interview session notes
DG- General descriptions

DE— Schools principal

NEl- Interview session notes

F E- School’s firnction

NE2- Interview session notes
AM- Students and teachers
NE3- Interview session notes
CA- Education costs and school aids
NE4- Interview session notes
IN- Infrastructure

NES- Interview session notes
PRM- Questions to teachers
NE6- Interview session notes
PRM- Questions to teachers
NE7- Interview session notes

 

162

Table A4.2 (Cont’d).

Questionnaire
Health Service Infrastructure

Small Health Providers

Sections

RS- Responsible for the services

AG- General aspects of the unit

NEl - Interview session notes

PS- Personnel and services that the unit has
NE2— Interview session notes

SER— Activities and services provided by the unit
NE3- Interview session notes

LAB- Laboratory for clinical tests

LB- Direct observation of laboratory

NE4- Interview session notes

F M- Drugstore/Medicines

NE5- Interview session notes

RM- Direct observation of exploration room
SE- Direct observation of waiting room

GP- General data of provider

POS- Problems when offering services
AGC- General aspects of the town

SV- Services offered
MS- Health material, equipment and set of
instruments

MED- Medicines

PT- Traditional midwife

SRM- Medical exploration room
SA- Waiting room

NE- Interview session notes

 

The MxFLS-1 databases and the information contained in this appendix may be

downloaded from http://www.radix.uia.mx/ennvih/ and http://www.mxfls.cide.edu.

 

 

163

REFERENCES

Alderman, H, Jere R. Behrman, Victor Lavy and Rekha Menon. (2001). Child Health and
School Enrollment: A Longitudinal Analysis. Journal of Human Resources: 36(1),
185-205.

Alderman, H., Jere Behrman and John Hoddinott. (2003). Nutrition, Malnutrition and
Economic Growth. PAHO/IDB Workshop: On Health, Human Development
Potential and the Quality of Life: Towards Biological-Based Index of Human
Development Potential for Assessing the Quality of Life. (Later published in

Health and Economic Growth: Findings and Policy Implications. Guillem Lo'pez-
Casasnovas, Berta Rivera and Luis Currais, eds. (2004) Cambridge, MA: MIT

Press).

Allen, L. and Jennifer Casterline-Sabel. (2001). Prevalence and Causes of Nutritional
Anemias. Nutritional Anemias, Usha Ramakrishnan, ed. Florida: CRC Series in
Modern Nutrition.

Anderson P. M., Phillip B. Levine and Kristin F. Butcher. (2002). Maternal Employment
and Overweight Children. Paper No. 8770, National Bureau of Economic
Research, Cambridge, MA.

Anderson, P. M., Kristin F. Butcher and Phillip B. Levine. (2003). Economic
perspectives on childhood obesity. Economic Perspectives: 3Q, 30-48.

Anderson, P. M. and Kristin F. Butcher. (2005). Reading, Writing, and Raisinets: Are
School Finances Contributing to Children's Obesity? Paper No. 11177, National
Bureau of Economic Research. Cambridge, MA.

Anderson, P. M. and Kristin F. Butcher. (2006). Childhood Obesity: Trends and Potential
Causes. The Future of Children: 16(1), 19-45.

Averett, Susan L., Daniel I. Rees and Laura M. Argys. (2002). The Impact of
Government Policies and Neighborhood Characteristics on Teenage Sexual
Activity and Contraceptive Use. American Journal of Public Health: 92(11),
1773—1778.

Avila, A., Teresa Shamah L. and Adolfo Chavez V. (1997). Encuesta Nacionfide
Alimentacion y Nutricion en el Medio Rural 1996. México, DF.: Instituto
Nacional de la Nutricion Salvador Zubiran.

Avila, A., Teresa Shamah L., Adolfo Chavez V.,and Carlos Galindo G. (2003). Encuesta
Urbana de Alimentacion v Nutricion en la Zona Metropolitana de la Ciudad de

164

 

 

 

México 2002. México, DF: Instituto Nacional de Ciencias Médicas y Nutricion
Salvador Zubiran, Instituto Nacional de Salud Publica.

 

Bhattacharya, J ayanta, Janet Currie and Steven Haider. (2004). Poverty, food insecurity,
and nutritional outcomes in children and adults. Journal of Health Economics: 23,
839—862.

Bhattacharya, J ayanta, Thomas DeLeire, Steven Haider and Janet Currie. (2002). Heat or
eat? Cold weather shocks and nutrition in poor American families. Paper No.
9004, National Bureau of Economic Research, Cambridge, MA.

Birch, L. L. and Jennifer 0. Fisher. (2000). Mothers' child-feeding practices influence
daughters' eating and weight. American Journal of Clinical Nutrition: 71,1054-
1061.

Blau, D. M., David K. Guilkey and Barry M. Popkin. (1996). Infant Health and the Labor
Supply of Mothers. Journal of Human Resources: 31(1), 90-139.

Bobonis, G., Edward Miguel and Charu Puri Sharma. (2004). Iron Deficiency Anemia
and School Participation. Paper No.7, Poverty Action Lab MIT, Cambridge, MA.

Brewis, Alexandra. (2003). Biocultural aspects of obesity in young Mexican
schoolchildren. American Journal of Human Biologv:15, 446-460.

Burton, P., Shelley Phipps and Lori Curtis. (2002). All in the Family: A Simultaneous
Model of Parenting Style and Child Conduct. AEA Papers and Proceedings:
92(2), 368-372.

Cameron, A. C. and Pravin K. Trivedi. (2005). Microeconometrics Methods and
ApplicationsNY: Cambridge University Press.

Case, A., Angela Fertig and Christina Paxson. (2005). The Lasting Impact of Childhood
Health and Circumstance. Journal of Health Economics: 24(2), 365-389.

 

CONAPO. (2000). Cuademos de Salud Reproductiva: Republica Mexicana. Primera
edicion. Mexico, D.F. Mexico. (Downloaded from http://wwwwconapo.gobmx)

 

CPEUM. Constitucion Politica de los Estados Unidos Mexicanos: Articulo 4. Instituto de
Investigaciones J uridicas, UNAM. (Downloaded from
http://info4.iuridicasunam.mx/iiure/fed/9/dcfault.htm‘?s=)

Currie J. and Enrico Moretti. (2005). Biology as destiny? Short and long-run

determinants of intergenerational transmission of birthweight. Paper No.11567,
National Bureau of Economic Research, Cambridge, MA.

165

Chou, SY., Michael Grossman and Henry Saffer. (2002). An Economic Analysis of Adult
Obesity: Results from the Behavioral Risk Factor Surveillance System. Paper No.
9247, National Bureau of Economic Research, Cambridge, MA.

Cole, T. J ., Mary C. Bellizi, Katherine M. F legal, William H. Dietz. (2000). Establishing
a standard definition for child overweight and obesity worldwide international
survey. British Medical Journal: 320, 1240-1245.

 

Cutler, D. M., Edward L. Glaeser and Jesse M. Shapiro. (2003). Why Have Americans
Become More Obese? Journal of Economic Perspectives:17(3), 93-118.

Desai, S. and S. Alva (1998). Maternal education and child health: Is there a strong
causal relationship? Demography: 35(1), 71-81.

Doak, C. M., Linda S. Adair, Carlos Monteiro and Barry M. Popkin. (2000). Overweight
and Underweight Coexist within Households in Brazil, China and Russia. Journal
of Nutrition: 130, 2965-2971.

Garcia A., M. Guadalupe. (2001). La educacion sexual en la reforrna educativa de los
afios setenta. Revista Educar: 17, Abril/Junio. (Downloaded from:
littp://educacion.ialisco.gob.mx/consulta/educar/ l 7/GarciaA.html)

 

 

Ghuman S., Jere R. Behrman, Judith B. Borja, Socorro Gultiano and Elizabeth M. King.
(2005). Family Background, Service Providers, and Early Childhood
Development in the Philippines: Proxies and Interactions. Economic
Development and Cultural Change: 54, 129—164.

Glewwe, P. (1999). Why does mother's schooling raise child health in developing
countries? Evidence from Morocco. Journal of Human Resources: 34(1), 124-159.

Glewwe, P. and Michael Kremer. (2005). Schools, Teachers, and Education Outcomes in
Developing Countries. Paper No. 122, Center for International Development,
Harvard University.

Glick, P. and D. Sahn. (1998). Maternal labour supply and child nutrition in West Afi'ica.
Oxford Bulletin of Economics and Statistics: 60(3),325-355.

Handa, S. (1999). Maternal education and child height. Economic Development and
Cultural Change: 47(2), 421-439.

 

Haveman R. and Barbara Wolfe. (1995). The Determinants of Children’s Attainments: A
Review of Methods and Findings. Journal of Economic Literature: 33(4),1829-
1878.

Hemandez, B., Lucia Cuevas N., Teresa Shamah L., Eric A. Monterrubio, Claudia 1.
Ramirez S., Raquel Garcia R, Juan A. Rivera and Jaime Sepulveda. (2003).

166

F actores asociados con sobrepeso y obesidad en nifios mexicanos de edad escolar:
resultados de la Encuesta Nacional de Nutricion 1999. Salud Publica de México:
45(supl. 4), 8551-8557.

Hernandez, B., S. L. Gortmaker, G. A. Colditz, K. E. Peterson, N. M. Laird, S. Parra-
Cabrera. (1999). Association of obesity with physical activity, television
programs and other forms of video viewing among children in Mexico City.
International Journal of Obesity: 23, 845-854.

Hemandez-Chavez, J .J . (1995). Mexican NGO HIV/AIDS activities. Salud Publica de
México: 37, 654-660.

Hoffman, Daniel. (2001). Obesity in developing countries: causes and implications. 1
Food, Nutrition and Agriculture: 28, 35-42.

Horton, S. and R. Ross. (2003). The economics of iron deficiency. Food Policy: 28, 51—
75.

Kassouf, A. L. and B. Senauer. (1996). Direct and indirect effects of parental education
on malnutrition among children in Brazil: A full income approach. Economic
Development and Cultural Change: 44(4), 817-838.

 

Klesges, R. C., Risa J. Stein, Linda H. Eck, Terry R. Isbell and Lisa M. Klesges. (1991).
Parental influence on food selection in young children and its relationships to
childhood obesity. American Journal of Clinical Nutrition: 53, 859-64.

Lakdawalla, D. and Tomas Philipson. (2002). Technological Change and the Growth of
Obesity. Paper No. 8946, National Bureau of Economic Research, Cambridge,
MA.

 

LGP. Lev Genmrl de PoblacionL1973. Secretaria de Gobemacion. (Downloaded from
http://\4-'ww.ordcniuridico.gob.rnx/Federal/PE/PlUl..eves/070l 1974(1 ).pdf)

 

 

Lynch, S. and Ralph Green. (2001). Assessment of Nutritional Anemias. Nutritional
Anemias. Usha Ramakrishnan, ed., Florida: CRC Series in Modern Nutrition.

 

Michael, Robert T. (2004). Sexual Capital: An Extension of Grossman's Concept of
Health Capital. Journal of Health Economics: 23(4), 643-652.

 

Miguel, E. and M. Kremer. (2004). Worms: identifying impacts on education and health
in the presence of treatment extemalities. Econometrica: 72(1), 159-217.

 

Miranda, Alfonso. (2006). Are young cohorts of women delaying first birth in Mexico?
Journal of Pgmlation Economics: 19, 55-70.

167

 

Must, A., Gerard E. Dallal and William H. Dietz. (1991). Reference data for obesity: 85th
and 95th percentiles of body mass index (wt/ht2) and triceps skinfold thickness.
American Journal of Clinical Nutrition: 53, 839-846.

Oettinger, Gerald S. (1999). The Effects of Sex Education on Teen Sexual Activity and
Teen Pregnancy. Journal of Political Economy: 107(3), 606-644.

Padron, M. (2002). Obesidad infantil: un problema creciente. Nutricion Clinica: 5(4),
258-262.

Palma , Yolanda (2005). Politicas de poblacion y planificacion familiar.Demos Carta
Demografica sobre Mexico: 16. (Downloaded from:
http://www.ejournal.unam.mx/demos/demos_index.html)

 

Palma, Yolanda and Gabriela Rivera. (1996) La planificacion familiar en México. Muj er:
Sexualidad v salud reproductiva en Mexico. Ana Langer y Kathryn Tolbert, eds.
Mexico, DF: EDAMEX.

Passi, S. J. and Sheila C. Vir. (2001). Functional Consequences of Nutritional Anemia in
School Age Children. Nutritional Anemias. Usha Ramakrishnan, ed., Florida:
CRC Series in Modern Nutrition.

Paton, David (2002). The economics of family planning and underage conceptions.
J oumal of Health Economics: 21, 207-225.

Paxson, C. and Jane Waldfogel. (1999). Parental Resources and Child Abuse and
Neglect. AEA Papers and Proceedings: 89(2), 239-244.

Philipson, T. J. and Richard A. Posner. (1999). The Long-Run Growth in Obesity as a
Function of Technological Change. Paper no. 7423, National Bureau of Economic
Research, Cambridge, MA.

Pollitt, E. (1997). Iron deficiency and educational deficiency. Nutrition Reviews: 55(4),
133-141.

Popkin, B. M. (2001). The Nutrition Transition and Obesity in the Developing World.
Journal ofNutrition: 131, 8718—8738.

Popkin, B. M., M. K. Richards and C. A. Monteiro. (1996). Stunting is associated with
overweight in children of four nations that are undergoing the nutrition transition.
Journal of Nutrition: 126(12), 3009-3016.

 

Rashad, 1., Michael Grossman and Shin-Yi Chou. (2005). The Super Size of America: An
Economic Estimation of Body Mass Index and Obesity in Adults. Paper No.
11584, National Bureau of Economic Research, Cambridge, MA.

168

Reed. B.A., J .P. Habicht, C. Niameogo (1996). The effects of maternal education on child
nutritional status depend on socio-environmental conditions. International Journal
of Epidemiology: 25(3), 585-592.

 

Rivera, J ., Teresa Shamah L., Salvador Villalpando, Teresita Gonzalez de Cossio,
Bernardo Hernandez and Jaime Sepulveda. (2001). Encuesta Nacional de
Nutricion 1999: Estado nutricio de nifios y mujeres en México. Cuemavaca,
Mexico: Instituto Nacional de Salud Publica.

 

Rivera, J. and Jaime Sepulveda. (2003). Conclusions from the Mexican National
Nutrition Survey 1999: translating results into nutrition policy. Salud Publica de
México: 45(supl. 4), $565-$575.

Rivera, J ., Simon Barquera, Teresa Gonzalez-Cossio, Gustavo Olaiz and Jaime
Sepulveda. (2004). Nutrition Transition in Mexico and in Other Latin American
Countries. Nutrition Reviews: 62(7), 8149-8157.

Rodriguez, Gabriela (2001). LCuales son los beneficios de la educacion sexual?
Documento de Red Democracia y Sexualidad, A.C., México. (Downloaded from
http://wwwlibertadeslaicas.org.mx/pdlS/educac/beneficiospdl)

 

Rolland-Cachera, M. F ., T. J. Cole, M. Sempé, J. Tichet, C. Rossignol and A. Charraud.
(1991). Body mass index variations: centiles from birth to 87 years. European
Journal of Clinical Nutrition: 45, 13-21.

 

Rubalcava L. and Graciela Teruel. (2004). The Role of Maternal Cognitive Ability on
Child Health. Pger No. R-497, Inter-American Development Bank.

Ruhm, Christopher J. (2004). Maternal Empoyment and Adolescent Development. Paper
No. 10691 , National Bureau of Economic Research, Cambridge, MA.

Sénchez L., Gabriela. (1997). Poblacion, desarrollo y su marco juridico. Boletin
Mexicano de Derecho Comparado Nueva Serie: XXX(90). (Downloaded from
http://www.iuridicas.unam.mx/publica/rev/boletin/cont/90/art/artl4.htm)

 

Shamah-Levy, T., Salvador Villalpando, Juan A. Rivera, Fabiola Mejia-Rodriguez,
Martha Camacho-Cisneros, Eric A Monterrubio. (2003). Anemia in Mexican
women: A public health problem. Salud Publica de México: 45(4), $499-$507.

Strauss, J. (1990). Household, Communities, and Preschool Children’s Nutrition

Outcomes: Evidence from Rural Coté d’lvoire. Economic Development and Cultural
Change: 38(2), 231-261.

Strauss, J. and Duncan Thomas. (1995). Human Resources: Empirical Modeling of
Household and Family Decisions. Handbook of Development Economics 3A. Jere
R. Behrman and T. N. Srinivasan, eds., Amsterdam: North-Holland.

 

169

Strauss, J. and Duncan. Thomas. (1998). Health, nutrition and economic development.
Journal of Economic Literature: XXXVI, 766-817.

Thomas, D. (1994). Like Father, Like Son; Like Mother, Like Daughter. Parental
Resources and Child Height. Journal of Human Resources: 29(4), 950-988.

Thomas, D. (2001). Health, Nutrition, and Economic Prosperity: A Microeconomic
Perspective. Paper No. WGl: 7. Commission on Macroeconomics and Health,
World Health Organization.

Thomas, D., Elizabeth Frankenberg, Jed Friedman, J can-Pierre Habicht, Nathan Jones,
Christopher McKelvey, Gretel Pelto, Bondan Sikoki, James P. Smith, Cecep
Sumantri and Wayan Suriastini. (2004). Causal Effect of Health on Labor Market
Outcomes: Evidence from a Random Assignment Iron Supplementation
Intervention. Paper CCPR-022-O4. California Center for Population Research,
University of California, Los Angeles.

 

Thomas, D., John Strauss and Maria-Helena Henriques. (1991). How Does Mother's
Education Affect Child Height? Journal of Human Resources: 26(2), 183-211.

Torres 8., Valentina. (2002). La Educacion Privada en México. Diccionario de la Historia
de la Educacion en México. Publicaciones digitales, UNAM. (Downloaded from
http://biblioweb.dgsca.unam.mx/diccionario/htm/articulos/scc_20.htm)

 

Tremblay, Carol H. and Davina C. Ling. (2005). AIDS education, condom demand, and
the sexual activity of American youth. Health Economics: 14, 851-867.

Uauy, R., Cecilia Albala and Juliana Kain. (2001). Obesity Trends in Latin America:
Transiting from Under- to Overweight. Journal of Nutritionzl31, 8938—8998.

Valdespino J L, Olaiz G, Lopez-Barajas MP, Mendoza L, Palma O, Velazquez O, Tapia
R, Sepulveda J. (2003). Encuesta Nacional de Salud 2000. Tomo I. Vivienda,
mlacion y utilizacion de servicios de salud. Cuemavaca, México: Instituto
Nacional de Salud Publica.

Variyam, J. N., James Blaylock, Biing-Hwan Lin, Katherine Ralston and David
Smallwood. (1999). Mother's Nutrition Knowledge and Children's Dietary
Intakes. American Journal of Agricultural Economics: 81(2), 373-384.

Villalpando, 8., Teresa Shamah-Levy, Claudia I. Ramirez-Silva, F abiola Mejia-
Rodriguez and Juan A Rivera. (2003). Prevalence of anemia in children 1 to 12
years of age. Results from a nationwide probabilistic survey in Mexico. M
Publica de México: 45(4), S490-S498.

 

170

WHO. (2000). Obesity: Preventing and Managing the Global Epidemic. Technical Report
894, Geneva: World Health Organization. (Downloaded from
http://www.who.int/nutrition/publications/obesity/en/index.html)

WHO. (2003). Information sheet on Obesity and Overweight. Global Strategy on Diet,
Physical Activity and Health: World Health Organization. (Downloaded from
littp://www.who.int/dietphvsicalactivity/publications/facts/obesity/enfl

 

WHO. (2004). Global Strategy on Diet, Physical Activity and Health. Geneva: World
Health Organization. (Downloaded from:
http://wwwvwho.int/dietphvsicalactivity/strategy/cb1 l344/en/)

 

WHO. (2005). The challenge of obesity in the WHO Europgan Region. Fact sheet
EURO/13/05. Copenhagen: World Health Organization. (Downloaded from.
http://wwweuro. who.int/1nediacentre/200206l 7 l)

 

WHO/CDC. (2005). Assessing the iron status ipopulations: report of afiJoint World
Health Organization/Centers for Disease Contrflnd Prevention Technical

Consultation on the Assessment of Iron Status at the Population Level. Geneva,
Switzerland: 6—8 April 2004.

WHO/UNICEF. (2004). Focusing on anaemia: Towards an integrated approach for
effective anaemia control. Joint statement. World Health Organization.
(Downloaded from http://uM-wwho.int/topics/anaemia/enfl.

 

Wolfe, B. and R. Haveman. (2001). Accounting for the Social and Non-Market Benefits
of Education. The Contribution of Human and Social Capital to Sustained
Economic Growth and Well Being. International Symposium Report. John
Helliwell, ed., Organization for Economic Co-operation and Development and
Human Resource Development Canada, Paris and Ottawa.

Wolfe, B. and R. Haveman. (2002). Social and nonmarket benefits from education in an
advanced economy. Education in the let Century: Meeting the Challenges of;
CharLginQVorld. Y. Kodrzyck, ed., Federal Reserve Bank of Boston.

Wooldridge, Jeffrey M. (2002). Econometric Analfiis of Cross Section and Panel Data.
Cambridge, MA: MIT Press.

Wooldridge, Jeffrey M. (2003). Introductory Econometrics: A Modern Approach. Second
edition. Cincinnati, OH: South-Westem College Publishing.

Zambrano, E. and Carolina Guzman. (2002). Desnutricion intrauterina y factores de
riesgo de la obesidad. Nutricion Clinica: 5(4), 241-245.

171

IIIIIIIIIIIIIIIIIIIIIIIIIIII

11111111111111111111111111111