SLUMS AND CHILDREN’S DISADVANTAGE:

THE CASE OF INDIA

Valerie A. Lewis

A DISSERTATION

PRESENTED TO THE FACULTY

OF PRINCETON UNIVERSITY

IN CANDIDACY FOR THE DEGREE

OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE

BY THE DEPARTMENT OF SOCIOLOGY

Advisor: Katherine S. Newman, Ph.D.

September 2009

UMI Number: 3364541

Copyright 2009 by Lewis, Valerie A.

All rights reserved

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy

submitted. Broken or indistinct print, colored or poor quality illustrations and

photographs, print bleed-through, substandard margins, and improper

alignment can adversely affect reproduction.

In the unlikely event that the author did not send a complete manuscript

and there are missing pages, these will be noted. Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion.

______________________________________________________________

UMI Microform 3364541Copyright 2009 by ProQuest LLC

All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.

_______________________________________________________________

ProQuest LLC 789 East Eisenhower Parkway

P.O. Box 1346 Ann Arbor, MI 48106-1346

© Copyright 2009 Valerie A. Lewis. All rights reserved.

iii

For Samir.

iv

Abstract The developing world is urbanizing very rapidly; while the average poor person was

once a rural resident, today the average poor person lives in a city. The majority of urban

growth is taking place in the poorest segments of urban society, both because of migration

and high fertility among the urban poor. This has resulted in the growth and proliferation

of slums in the developing world. This dissertation seeks to quantify what disadvantages

are faced by children living in slums in India. As a large, quickly urbanizing, and democratic

country, India is a natural place to undertake a study of slums.

I begin by examining how to best conceptualize and define slums, reviewing slum

definitions from India and around the world. Following this, I use data from the National

Family Health Survey 2005-2006, a national household survey of India that includes

oversamples of slum populations, to examine three sets of outcomes. First, I consider infant

and child mortality. I find that slum infants and children face higher mortality than other

urban children but lower mortality than rural children. Once family background is taken

into account, however, these differences disappear entirely. Second, I examine children’s

health. I find that slum children are no more likely to be malnourished or suffer acute

illness symptoms than other children. Questions are raised about the validity of maternal

reports of children’s health, however. Third and last, I consider children’s school

attendance and work. Contrary to popular belief, I find that rural children are the most

likely to be attending school net of family background characteristics. Slum children are

indistinguishable from other urban children until age 14, at which point they become much

less likely to be in school than either rural or other urban children. Though further

v

research is needed, I attribute this to a combination of a vibrant informal economy in slums

that provides incentives for children to work and a deficit of government run high schools

for children to attend.

Overall, this dissertation provides some of the first rigorous quantitative analysis of

the situation of slum residents in the developing world. Taken together, the results indicate

that slums are a difficult and complex phenomenon to measure, not a homogenous set of

neighborhoods. In some ways slums disadvantage residents, but in other ways slums may

provide advantages compared to rural residence.

vi

Acknowledgements There are many people I would like to thank for their help along the way to this final

product. Kathy Newman has been an amazing chair, and I can confidently say that without

her, this dissertation would not be completed. Doug Massey saw me through every hurdle

of my PhD program, including this dissertation. Several other faculty members helped me

along the way, including Scott Lynch, Jeff Hammer, Devah Pager, Sarah McLanahan, and

Bob Wuthnow. I thank Wayne Appleton, Joyce Lopuh, Elana Broch, and Donna DeFrancisco

for their help with computing, references, and many administrative tasks. In India, I

extended many thanks to Gyan and Nina Badgaiyan, Ashwini Deshpande, Kailash Das,

Laishram Ladu Singh, and the women at the Women’s Democratic Organisation of India, all

of whom were wonderful and hospitable hosts as well as invaluable for my research.

To the special group of friends that was forged out of the fire of our first year of

graduate school (and those we picked up along the way): to you I owe an amazing debt of

gratitude. David, Sheherazade, Kelly, Charles, Elisha, Kevin, Ainsley, Jeff, and January, the

combination of your constant support and mocking is irreplaceable.

My family always supported me in this crazy endeavor for a PhD. And so I could

never offer enough thanks to my parents, my parents in-law, and Monica and Eric. I can

only imagine that they will be happy that I am finally leaving school at the age of 26.

Lastly, perhaps to the dismay of the mysterious graffiti artist from my first spring, I

am married to the most wonderful man in Samir. I can’t enumerate all the ways you have

helped me get to where I am. Thank you.

vii

Table of Contents

Abstract .................................................................................................................................................................... iv

Acknowledgements .............................................................................................................................................. vi

Table of Contents ................................................................................................................................................. vii

List of Tables .......................................................................................................................................................... xi

List of Figures ...................................................................................................................................................... xiii

Chapter 1 : Introduction ...................................................................................................................................... 1

THE URBAN PENALTY AND THE URBAN ADVANTAGE.................................................................... 1

WORLD POVERTY AND URBANIZATION ................................................................................................ 3

THE SOCIOLOGICAL STUDY OF SLUMS ................................................................................................... 6

THE RESEARCH GAP .................................................................................................................................... 10

Chapter 2 : Critical Issues for Indian Slums .............................................................................................. 12

THE FACTS ABOUT INDIAN SLUMS ....................................................................................................... 14

MY RESEARCH QUESTIONS ....................................................................................................................... 16

Defining slums ............................................................................................................................................ 17

Infant and child mortality ...................................................................................................................... 18

Child health .................................................................................................................................................. 19

Children’s education and work ............................................................................................................ 19

DATA AND METHODS .................................................................................................................................. 20

Advantages of the Data ........................................................................................................................... 20

Sampling ....................................................................................................................................................... 21

Response rates and missing data ........................................................................................................ 22

Data Limitations......................................................................................................................................... 23

Measuring income and wealth in the NFHS .................................................................................... 24

viii

Descriptive Statistics ................................................................................................................................ 26

OVERVIEW ....................................................................................................................................................... 27

TABLES AND FIGURES ................................................................................................................................. 31

Chapter 3 : Defining and Measuring Slums ............................................................................................... 32

THE COMPLEXITIES OF OFFICIAL DEFINITIONS ............................................................................. 33

Historical definitions of slums ............................................................................................................. 34

United Nations slum definitions .......................................................................................................... 38

The Indian Case .......................................................................................................................................... 40

CONCEPTUALIZING SLUMS ....................................................................................................................... 43

Slums as places of poor infrastructure ............................................................................................. 44

Slums as squatter settlements ............................................................................................................. 45

Slums as neighborhoods of concentrated poverty ....................................................................... 47

Slums as marginalized or socially excluded communities ........................................................ 49

THE POLITICS OF SLUM DEFINITIONS IN INDIA ............................................................................. 53

MEASURES OF SLUMS FOR ANALYSES ................................................................................................. 55

DISCUSSION ..................................................................................................................................................... 56

Chapter 4 : Infant and Child Death in India ............................................................................................... 61

INTRODUCTION ............................................................................................................................................. 61

THE URBAN ADVANTAGE IN CHILD MORTALITY ........................................................................... 62

DATA AND METHODS .................................................................................................................................. 65

Data ................................................................................................................................................................. 65

Methods ......................................................................................................................................................... 66

RESULTS ............................................................................................................................................................ 68

CONCLUSIONS ................................................................................................................................................. 71

Selection Questions .................................................................................................................................. 73

ix

The more complicated picture ............................................................................................................. 74

TABLES AND FIGURES ................................................................................................................................. 75

Chapter 5 : Child Health .................................................................................................................................... 80

CHILDREN’S HEALTH .................................................................................................................................. 81

HEALTH SEEKING BEHAVIOR .................................................................................................................. 85

The Research Gaps .................................................................................................................................... 87

DATA AND METHODS .................................................................................................................................. 88

RESULTS ............................................................................................................................................................ 91

DISCUSSION ..................................................................................................................................................... 96

TABLES AND FIGURES ............................................................................................................................... 101

Chapter 6 : School Enrollment, Children’s Work, and Idleness ........................................................ 105

SCHOOL FACILITIES, CONTEXTUAL FACTORS, AND SLUMS ..................................................... 108

CHILD LABOR AND IDLENESS ................................................................................................................ 110

BACKGROUND: EDUCATION AND POLICY IN INDIA ..................................................................... 112

POSSIBLE SLUM EFFECTS ........................................................................................................................ 115

METHODS ....................................................................................................................................................... 118

RESULTS .......................................................................................................................................................... 121

CONCLUSIONS ............................................................................................................................................... 129

FIGURES AND TABLES ............................................................................................................................... 133

Chapter 7 : Conclusions .................................................................................................................................. 140

Returning to the definition of slums ................................................................................................ 141

Why stay in slums? ................................................................................................................................. 142

Future directions ..................................................................................................................................... 144

Our urban future ...................................................................................................................................... 147

Appendix A : Infant and Child death interaction models .................................................................... 149

x

Appendix B : Child health interaction models ........................................................................................ 151

Appendix C : Children’s activities interaction models ......................................................................... 156

REFERENCES ...................................................................................................................................................... 162

xi

List of Tables Table 2-1: Descriptive Statistics from National Family Health Survey 2005-2006 ................. 31

Table 3-1: Indicators and Operational Definitions of Slums, from United Nations Human

Settlements Programme (2003) .................................................................................................................. 58

Table 3-2: Descriptive statistics on housing conditions and slum status of households in the

National family Health Survey, 2005-2006 ............................................................................................. 59

Table 3-3: Counts of urban household by census and NFHS-3 surveyor identified slum

status ....................................................................................................................................................................... 60

Table 4-1: Descriptives statistics for births by place of residence, National Family Health

Survey 2005-2006 ............................................................................................................................................. 75

Table 4-2: Results from discrete time logit models predicting infant death, National Family

Health Survey 2005-2006 ............................................................................................................................... 76

Table 4-3: Results from discrete time logit models predicting child death to age 5, National

Family Health Survey 2005-2006 ................................................................................................................ 77

Table 5-1: Descriptive statistics on residence location and percent of children exhibiting

the health condition, National Family Health Survey 2005-2006................................................. 101

Table 5-2: Results from logisitic regressions of child sickness in two weeks prior to survey

on residence, background, and family characteristics, National Family Health Survey 2005-

2006 ...................................................................................................................................................................... 102

Table 5-3: Results from logistic regressions of child wasting and stunting on residence,

background, and housing characteristics, National Family Health Survey 2005-2006 ....... 103

Table 5-4: Results from logistic regression of treatment of cough/fever and diarrhea on

residence, background, and housing characteristics, National Family Health Survey 2005-

2006 ...................................................................................................................................................................... 104

Table 6-1: Descriptions, measurements, and means used in the analysis by place of

residence, National Family Health Survey 2005-2006 ...................................................................... 133

Table 6-2: Proportion of children in work, school, and idleness by place of residence,

National Family Health Survey 2005-2006 ........................................................................................... 134

Table 6-3: Odds ratios and standard errors from logistic regression of school attendance on

residence and background characteristics ............................................................................................ 135

xii

Table 6-4: Relative risk ratios from multinomial logit models of children's activities on

residence, child, and family characteristics, National Family Health Survey 2005-2006 ... 136

Table A-1: Coefficients from discrete time logit models of infant death to 13 months on

residence, background, and residence interactions, National Family Health Survey 2005-

2006 ...................................................................................................................................................................... 149

Table B-1: Results from logisitic regressions of child sickness in two weeks prior to survey

on residence, background, and housing characteristics with residence interactions,

National Family Health Survey 2005-2006 ........................................................................................... 151

Table B-2: Results from logistic regressions of child wasting and stunting on residence,

background, and housing characteristics with residence interactions, National Family

Health Survey 2005-2006 ............................................................................................................................. 154

Table C-1: Logistic regression of school attendance on location and background

characteristics and interactions between location and background characteristics, National

Family Health Survey 2005-2006 .............................................................................................................. 156

Table C-2: Multinomial regression of children's activities on location and background

characteristics with interactions between location and background characteristics,

National Family Health Survey 2005-2006 ........................................................................................... 159

xiii

List of Figures

Figure 4-1: Kaplan-Meier univariate estimates of survival through month 13 by place, NFHS

2005-2006 (N=191321) ................................................................................................................................. 78

Figure 4-2: Kaplan-Meier univariate estimtes of survival to month 60 by place, NFHS 2005-

2006 (N=191321) ............................................................................................................................................. 79

Figure 6-1: Proportion of children attending school by place of residence, National Family

Health Survey 2005-2006 ............................................................................................................................. 137

Figure 6-2: Proportion of children working outside the home by place of residence,

National Family Health Survey 2005-2006 ........................................................................................... 138

Figure 6-3: Proportion of children idle by place of residence, National Family

Health Survey 2005-2006 ........................................................................................................................ 139

1

Chapter 1 : Introduction

“The slum is the measure of civilization.”

Jacob Riis, Dutch American journalist and photographer, 1849-1914

THE URBAN PENALTY AND THE URBAN ADVANTAGE

Who is better off: those who live in cities, or those who live in rural areas? For much

of history, cities have been more dangerous to one’s health than rural areas. The “urban

penalty” was first observed during the Victorian era, when despite greater than average

incomes, city residents died at higher rates than their rural counterparts. The term urban

penalty came from analyses of English mortality data during the 19th century industrial

revolution (Kearns 1988). The data revealed that urban mortality rates were much higher

than rural mortality rates. This disadvantage has been quantified in other countries during

the late 18th century as well. For instance, in 1875 in Prussia, urban infant mortality was

240 per 1000 births, while rural infant mortality was only 190 per 1000 births (Vogele

1994, 2000).

The urban penalty in mortality and morbidity reversed only after two important

developments. First, advancements in the understanding of disease enabled those with

higher incomes to purchase effective treatments, so that on average urban populations

could purchase higher levels of health (Samuel H. Preston and M. R. Haines 1991; Samuel H.

Preston and Walle 1978; van Poppel and van der Heijden 1997; Ewbank and Samuel H.

Preston 1990). Second, these same scientific advancements made possible effective public

health. Governments made public health investments in areas such as clean water and

sanitation, protecting urban populations against the spread of many communicable

2

diseases. When these public health advancements were made in Prussia, for example, by

1905 urban infant mortality was at parity with rural infant mortality at 170 per 1000

births (Vogele 1994, 2000).

Since the beginning of the 20th century, urban residents around the world have

experienced a wealth of advantages when compared to their rural counterparts. Urban

residents across the globe are more likely to be employed, earn higher wages, have higher

levels of completed schooling (or to be in school when children are concerned), have better

health, be more nourished, have better access to health care facilities, and have better

access to health insurance, to name a few (United Nations Centre for Human Settlements

1996).

Since 1980, however, the urban advantage of the 20th century has come into

question. The norm of merely comparing urban and rural mean outcomes covers up the

large array of inequality within each of these settings (van Poppel and van der Heijden

1997). In urban areas in particular, the growth of concentrated poverty brings into

question the relevance of an urban mean compared to a rural mean on any health,

education, or economic outcome. In some cases, the trends have seemed to be reversing—

rural means actually climbing higher than urban means (Montgomery, Stren, and Cohen

2003). This is often surmised as being the result of an increase in the poor segments of the

urban population, often slum residents.

This dissertation seeks to learn whether there is an urban advantage in India. Do

slum residents benefit from any urban advantage in health or educational outcomes? How

do urban poor slum residents compare to the rural poor? And how much of these

3

differences are due to differences in population characteristics, and how much is due to

other factors?

WORLD POVERTY AND URBANIZATION

The United Nations projects that the world’s population will reach 8.27 billion by

2030, an addition of over 2 billion people to the current population. Virtually all of this

increase will take place in developing countries, and the cities will absorb the vast majority.

In essence, the developing world will have moved from being mainly rural (over 80% rural

in 1950) to mainly urban (projected to be 2/3 urban by 2030). Accordingly, the global

portrait of poverty will shift our attention from farmlands and agriculture to the urban

poor living in squatter settlements and slums (United Nations Human Settlements

Programme 2003).

The push and pull factors that dominated past centuries of rural to urban migration

have changed dramatically. During and after the industrial revolution, the pull of industry

led to sharp increases in the size of the urban communities of Europe and the developed

world. Today, traditional pull factors of better jobs in cities and the allure of the city

combined with new push factors, including the mechanization of agriculture in the

developing world and civil unrest, drive rural to urban migration1. In the past,

urbanization went hand in hand with economic development, and with the increase in

urbanization there was a corresponding increase in per capita incomes. This is no longer

universally true. In South Asia and Africa, urbanization is picking up pace without

1 The mechanization of agriculture is certainly an old story in the United States and the rest of the developed

world, but the developing world agriculture is still in the process of becoming mechanized (World Bank

2009b).

4

concomitant industrialization (Cohen 2004; Montgomery et al 2003). Instead, the rural

poor are often left without jobs as agriculture become mechanized, and they move to the

city in search of any work possible. In addition, the urban poor have high fertility rates that

contribute as much as 50% to urban growth (Montgomery et al. 2003).

As a result of rural to urban migration and urban natural increase, urbanization is

taking place at an unprecedented pace in the developing world. The United Nations

projects that the world population will grow by 2 billion by 2030 and an additional 2 billion

by 2050, and 70-90% of this growth will take place in cities of the developing world

(Brockerhoff and Brennan 1998; Montgomery et al. 2003; United Nations Centre for

Human Settlements 1996). While rapid urbanization presents many challenges including

risks to physical environments, health conditions, social cohesion, and individual rights, the

most immediate concerns have been with the large increase in numbers of the urban poor

and the growth of slums.

The facts of developing world slums are important not only to the studies of cities

and urbanization but to the study of poverty in the developing world as well. Where the

typical poor person once lived in the countryside, it will soon be (or may already be) the

case that the typical poor person lives in a city. This will have many important

consequences. The cities of the developing world are currently growing larger and poorer

than any cities the world has ever seen. For example, Mumbai is projected to grow to a

population of 33 million by 2025. Currently, the Indian government estimates that over half

the population of Mumbai lives in slums or on the streets (Bhatt 2000). Estimates also

show that newcomers to cities disproportionately end up in slums, so this proportion will

5

only increase with time. The lack of adequate public health provisions in these areas

including clean water, toilet facilities, and sanitation has caused the long standing urban

advantages in health to come into question. In some cities, researchers suspect the urban

poor are faring worse than their rural counterparts (Montgomery et al. 2003; United

Nations Centre for Human Settlements 1996; United Nations Human Settlements

Programme 2003). For example, in a cross national study done by Brockerhoff and

Brennan (1998) large cities and poor cities often experienced higher infant mortality than

smaller cities and rural areas. It is clear that the extreme poverty of many developing world

cities and the decoupling of development and urbanization is the driving factor behind this

change.

Much of the expected growth will take place in the poorest districts of cities which

are expected to swell in size and density both because of increasing migration of the rural

poor to cities and high rates of natural increase among the poor in cities (Montgomery et al

2003). Despite the many clear disadvantages to life in slums—ranging from insufficient

sanitation, water, and social services to pollution and inadequate housing—slums still

attract migrants from rural areas, often because there is no other affordable place to live

and because people often have networks connecting them to slums (Lobo and Biswaroop

Das 2001). Governments have attempted to relocate slum residents or redevelop slums by

putting up high-quality housing for residents, but these attempts have almost entirely

failed.

The phenomenon of developing world slums and urbanization has received some

attention from journalists and the popular press (Neuwirth 2005; Davis 2006), but it

6

remains understudied in academia, and in particular within the discipline of sociology.

Existing research focusing specifically on slums tends to be mainly descriptive. There is a

need for rigorous sociological research that analyzes poverty and disadvantage in

developing world slums, asking questions about who is disadvantaged, how and why this

disadvantage exists, and consequent policy recommendations.

THE SOCIOLOGICAL STUDY OF SLUMS

Sociology has a rich tradition of studying urban poverty in America and has done so

for decades. Sociologists have established that for most people who will ever experience

poverty, it is a temporary state; people move in and out of poverty. For a small segment of

the population, however, poverty is a constant state that individuals cannot overcome

(Duncan 1984). Poverty has many causes. At the individual level, Americans who have low

education, are in single parent families, and are black are at a higher risk for being poor.

More profoundly however, is the work illuminating the structural causes of poverty.

Deindustrialization and the move to a service economy left unskilled and uneducated

workers with little hope of jobs paying a sustainable wage (Anderson 1999; Massey and

Denton 1993; Wilson 1987, 1996). This shift in the economy has been cited as a cause of

the increasing proportion of female headed households, joblessness in inner cities, the

concentration of poverty, and the creation of high poverty ghettos. Overall, sociologists

have focused on these structural factors as the root causes of poverty.

Sociologists have contributed a vibrant literature on poor neighborhoods in

recognition of the importance of place and social context in determining individual life

chances, and in response to questions about why poor residents of poor neighborhoods

experience very high levels of disadvantage compared to the same kinds of individuals in

7

non-poor neighborhoods. Starting with Du Bois (2007) study of black Americans in

Philadelphia, one answer to this question has been advanced fairly consistently: the

disorderly aspects of public life in poor neighborhoods. By disorder, scholars often point to

high rates of family disruption, high crime and delinquency, and low levels of educational

completion among children among other things.

Despite these correlations, qualitative sociologists have advanced a different

perspective of social disorder. They have argued that there are less observable forms of

social organization that make slums “work” and even dissuade residents from leaving them

when they have the chance. Starting with many of the ethnographies of the Chicago School

and followed by Street Corner Society, sociologists showed how rather than being places of

chaos and disorder, slums, ghettos, and poor neighborhoods have a high degree of social

organization among the residents (Zorbaugh 1929; Park, Burgess, and McKenzie 1925;

Wirth 1928; Gans 1982; Whyte 1943). Suttles (1968) showed that competing groups

organize themselves and create very powerful boundary lines and patterns of interaction

to keep order. Antagonistic groups are able to live in relative peace and with a high level of

order by following informal codes of territoriality, behavior, and interaction. These

patterns of highly organized poor neighborhoods have been found time and again by other

researchers over the years (e.g. Duneier 1999; Hannerz 1969; Anderson 2003, 1990).

Anderson (1990, 1999) looks at how codes of behavior operate and govern social life in the

ghetto. Lee (2002) examines relations between competing urban groups and how stability

(rather than conflict) is the norm for these groups.

8

Social networks of poor neighborhoods have also received a good deal of attention.

Stack (1974) was one of the first to uncover and explore the intricate and complicated kin

networks operating in poor neighborhoods formed to share resources, cope with poverty,

and smooth out the fluctuating resources that define urban poverty. More recently,

Klinenberg (2002) showed how in the Chicago heat wave crisis elderly poor without social

ties were much more vulnerable to health hazards and death than their poor counterparts

embedded in tight social networks. Edin and Lein (1997) probed the lives of poor women

to find out how they made ends meet every month, and found that sharing across friend

and family networks was crucial.

Perhaps most important to public debates on poverty are sociologists’ research on

work, crime, and American values in ghetto neighborhoods. Time and again

ethnographers have found that despite rampant joblessness in ghetto areas (Clark 1989;

Jargowsky 1997; Massey and Denton 1993; Wilson 1987, 1996), poor men and women

want to (and actually do) work and lead productive lives (Anderson 1999; Duneier 1992,

1999; Edin and Lein 1997; Bourgois 1995; Liebow 1967; Newman 1999, 2006). Delinquent

behavior (that perhaps most often leads to public condemnation of the poor and slum

areas) is usually the last resort of a minority when men have failed to obtain legal, gainful

employment where they can garner respect (Anderson 1999; Bourgois 1995; Liebow

1967). Visible as these forms of deviance are, they are not the majority experience even in

very poor neighborhoods (Newman 1999).

Despite a very rich literature on American poverty, sociologists have remained

largely silent on questions of the current, rapid urbanization in the developing world. Many

9

decades ago sociologists and anthropologists conducted seminal studies of rural to urban

migration and urban poverty in the developing world. This tradition, exemplified by

scholars such as Robert Redfield and Oscar Lewis, was focused mainly on Latin America.

These were mostly descriptive studies of poverty in cities and villages, or looking at rural

to urban migration (Redfield 1950, 1977, 1989; Lewis 1975, 1998). More recent work has

focused on survival strategies of the poor in Latin America, most prominently Mexico

(Lomnitz 197; Velez-Ibanez 1983) and the favelas of Brazil (Perlman 2003, 2006, 1976;

Davis 2006; Neuwirth 2005; Kramer 2006; Floris 2006; Garau, Sclar, and Carolini 2005).

While important in the literature, this vein of research has not always occupied a

prominent place in contemporary sociology, where studies of developing world poverty

often have been subordinated to world systems theory, underdevelopment, and other

macro-studies that focus more on the connections between the “metropoles” and the

“satellites” or the more mid-range questions of migration, cultural change, race and health

outcomes. Beyond the rich ethnographic work, most of what we know about slums and

poverty in the developing world is based on economic studies and reports from large

international bodies such as the World Bank and United Nations, interspersed with more

descriptive ethnographies of slum life.

There is a small but growing body of sociological literature describing the daily

experience of slum life in India (Bhatt 2000; Nangia and Sukhadeo. Thorat 2000; Lobo and

Biswaroop Das 2001; Schenk 2001; Verma 2002; Agarwala 2006) and around the world,

with focuses on Latin America and Africa (Perlman 1976, 2003, 2004, 2006). These studies

document slum conditions, economic life, employment, family patterns, threats and fears,

delinquent behavior, and group relations in specific slums in cities around India. Because

10

of a lack of comprehensive quantitative data sets on slums, qualitative researchers have

often taken up the task of documenting the particulars of India’s slum life. Often small scale

surveys or interview projects are undertaken in a single city or a single slum. Researchers

cover descriptive statistics on the sex ratio, literacy and education, living conditions,

migration status, and work of a small number of residents. They often describe a vibrant

social life in slums, including a lively set of religious celebrations, festivals, and social

networks of friends and relatives in the slums (Lobo and Biswaroop Das 2001; Bhatt 2000;

Schenk 2001; Nangia and Sukhadeo. Thorat 2000). This small scale work lies on the

boundaries of quantitative and qualitative work as it provides descriptive statistics but of

very small, non-random samples. Hence, sociologists rarely venture beyond these

descriptive studies to examine mechanisms that might account for the distribution of

critical outcomes: morbidity, mortality, and educational attainment, to name only a few.

THE RESEARCH GAP

Rigorous research is needed into the phenomenon of developing world slums. Slums

are growing quickly in the urban areas of the developing world, and it is these areas where

the next 1 billion world residents are projected to live. Understanding the challenges and

disadvantages slum residents face will be key in targeting programs and policies to the new

world poor and in governing and managing developing world cities.

This dissertation research focuses on two components of human capital: health and

education. These are areas that national and local governments most often turn to when

aiming to improve the lives of the poor. What disparities exist between slum residents,

other urban residents, and rural residents? I focus on four specific areas: child health and

mortality, children’s schooling and work, women’s health, and domestic violence. Are slum

11

residents disadvantaged? In what areas? How much of the disparities do population

characteristics such as wealth and parental education explain? Accordingly, what can we

not explain?

12

Chapter 2 : Critical Issues for Indian Slums

“A stranger could drive through Miguel Street and just say ‘Slum!’ because he could

see no more. But we who lived there saw our street as a world, where everybody

was quite different from everybody else.”

Miguel Street, by V.S. Naipul, British writer

Slums in India are extremely varied. One of the largest slums in Asia is located in

Mumbai. It called Dharavi and is home to around one million people on 538 acres2 in

central Mumbai. Dharavi is a bed of life and activity. A wide street leads through the center

of the slum. Two story buildings line the street, housing everything from health clinics to

families to schools. On a visit to Dharavi, we stop in and visit a government school run

within the slum. It has several buildings and a large open area for recess time. It teaches in

seven different languages: English, Hindi, Urdu, Malayalam, Gujurati, Bengali, and Marati.

Entering the gate, a security guard comes to check who we are. Finding we are affiliated

with a local university and seeing our credentials, he lets us in. We talk with a school

official, who tells us that this school goes through tenth grade. He says there are five such

government schools in Dharavi. There is a separate school for the last grades of high school,

but it is a smaller school.

The streets are filled with hustle and bustle at midday. This main street is wide, and

filled with economic activity. Dharavi is a large exporter of leather around the world: some

exporters send things such as leather belts to American Wal-Marts. Many slum residents

2 For reference, Princeton’s campus is just around 500 acres, with a total of around 7,000 students and 6,000

faculty and staff.

13

are employed in the tanning and leather industry. It is estimated that there are 5,000 one-

room factories where Dharavi residents produce goods, and that the annual size of the

economy of Dharavi is between $500 and $1400 million (depending on the estimate).

The smaller alleys leading off the main street are filled with women and children.

Women are talking, doing laundry, and hanging it to dry in the narrow alleys between

buildings. Children play in the streets. Men and women walk the alleys, carrying goods

from one place to another, often from a home where goods are made to a buyer. Women

and girls stand in line at the public taps throughout the day, getting water for their

dwellings that lack plumbing. Without sewers, open ditches next to the roads carry away

sewage, trash, and murky water. Electrical lines hang in bunches, wires going every

direction.

Dharavi has garnered much attention due to its large size. Many NGOs focus on this

slum, and the government has put much money into redevelopment projects of Dharavi.

Books have even been written on Dharavi, and a popular, classic novel Shantaram was

written based on it. It has been covered in an array of newspapers and magazines in the

west, including National Geographic, The Guardian, The Economist, the Los Angeles Times,

The New York Times, and BBC.

By contrast, Salim Nagar3 is a small Muslim slum in Delhi. There is no main street in

Salim Nagar. To get to it, we go down several small alleys behind a street of shops. The slum

is over forty years old, and some of the residents I speak with have been living here that

3 This name has been changed for confidentiality.

14

long. Though there are people out and on the streets, and the residents are happy to talk

with me, it lacks the vibrancy felt in the streets of Dharavi. A small dukkan (a stall selling

conveniences such as water, candy, and cigarettes) is the only bit of economic activity I see

in my day in the slum.

Two women accompany me to Salim Nagar, activists who work on behalf of slum

residents around the city. They recently have organized the slum residents to write a

petition and host a meeting with the local government officials to try to get a maternity

clinic in the slum, as many women who had complications during labor or childbirth had no

place to go nearby. The local official has agreed, though the project hasn’t started yet. (In

contrast, Dharavi has many health clinics within its borders, run by many different NGOs).

We sit down, and the slum residents are eager to tell me about the various good and bad

points of the slum. They have many worries: getting better water and sanitation, keeping

their children in school, keeping violence out of their slum. The families are poor, and the

men want jobs that pay more. The feeling in this slum is a lack of hope, as older men and

women tell me about living in this slum for decades, raising their children here, and

staying. Still, despite a lack of prospects for change, they tell me the best thing about Salim

Nagar is the strong community—the family ties, the friendships, the unity, pulling together

though things are rough.

THE FACTS ABOUT INDIAN SLUMS

India is a natural place to undertake a study of slums and disadvantage. An

important player in world poverty, China and India combined contain between one-third

and one-half of the world’s absolute poor. The sheer size of India’s total population also

makes it an important case for studying: its population of 1.15 billion is approximately one-

15

sixth of the total world population. India is experiencing high rates of urban growth: the

country is projected to go from approximately one-third to two-thirds urban over the next

thirty years (United Nations Centre for Human Settlements 1996). Additionally, India has a

substantial slum population. In the 2001 census data on slums were collected, and estimate

that 22% of urban residents are slum dwellers, with cities such as Mumbai reaching over

50% of the urban population in slums.

According to the 2001 Indian census, there are 405 cities in India with over one

hundred thousand residents, 26 cities with over one million residents, and five cities with

populations over five million. As a reference, in 2000 the United States had 239 cities over

one hundred thousand, just nine cities with over one million residents, and only one city

over five million. The 2001 census shows great diversity in the proportion of city residents

that live in slums.

The National Sample Survey of India, the country’s largest household survey,

periodically conducts modules on the conditions of urban slums. Their latest report gives

the best national, up-to-date figures on slums. The NSS makes a distinction between

“notified” or “declared” slums, slums recognized by the local city, and non-notified slums,

which have the characteristics of slums, but aren’t officially recognized as such (these

distinctions are discussed in depth in Chapter 3). The 2002 report along with the 2001

Indian census includes good overview statistics of slums in India. According to these

figures, 17.7 million people of the total 73.3 million people living in cities with a population

of over one million live in slums. For India’s five biggest cities: in Mumbai, 54% of the total

16

population lives in slums; in Delhi 19%, in Calcutta 33%, and Chennai 19%, and in

Bangalore 10%.

On average, a notified slum consisted of 205 households and non-notified slums

consisted of 102 households. Most slums have electricity: only 1% of notified and 16% of

non-notified slums lacked and electricity connection. Paved roads are less common: 71% of

notified slums and 37% of non-notified slums had paved (pucca) roads. Latrine facilities

were absent in 17% of notified slums and 51% of non-notified slums. Government agencies

collected garbage from 79% of notified slums and 42% of non-notified slums.

As the world’s largest democracy, Indian urbanization is happening with relatively

few restrictions or government interventions, diverging sharply from China’s unique

political system that puts tight and heavily enforced restrictions on internal migration and

fertility. India’s growing urban masses have a voice in the political system; they cannot be

ignored because they are part of the world’s largest democracy. The eyes of the world are

starting to turn to India, and the India the world we will see will be an urban giant.

MY RESEARCH QUESTIONS

There are many parties interested in slum life. National, state, and local

governments want slums improved for several reasons. The most charitable reasons are

for the slum residents own benefit: these people should not have to endure such poor living

conditions. There are other less charitable reasons. For attracting foreign investment and

tourism, cities want to clean up, be more modern, and be a place where businesses and

tourists want to visit. Slums detract from cities’ overall images, and so cities want to

remove or improve them. In addition to government, many non-governmental

17

organizations take interest in the well-being of slums. In India, countless NGOs focus on

various aspects of slum life, including varied projects such as education, children’s health,

work for slum residents, improving toilet facilities, getting better water supplies, and

obtaining a right to ration cards and bank accounts for slum residents. The many people

who operate and support these organizations want to better the lives of slum residents.

For any plan to improve slums to be successful, the first step is to understand

exactly what challenges and disadvantages slum residents face. Is it simply poor

construction? Are there health hazards? Are slums more socially conservative,

disadvantaging women? Do slum children have adequate access to quality education? Do

women in slums have access to reproductive and pre-natal health? There are literally

hundreds of questions about slums we might ask. Given the absolute paucity of research on

slum life and disadvantage in the developing world, any number of studies could be

undertaken. I focus on several key aspects for understanding slum disadvantage.

Defining slums

The question of what defines a “slum” is debated in various fields around the world,

and few agree. The answer to this question is crucial, however, if rigorous study of slums is

to be undertaken. A thorough review of the literature and prevailing definitions of slums

reveals a focus on housing tenure based definitions and living conditions. Definitions based

on housing tenure hold that slums are illegal or squatter settlements, and as such are

defined by the high proportion of residents who lack any legal claim to their property.

Definitions based on living conditions dwell on the poor physical conditions of slums,

including poor sanitation, overcrowding, and substandard buildings. I also review some

recent literate that posits a continuum of “slumness,” rather than a dichotomous model.

18

After examining current definitions, the question is: what is the best definition? I

will examine what percentage of households is defined as slum under the various

prevailing definitions. How do these definitions overlap? And lastly, can we find an

underlying measure of “slumness” that takes into account the several varying definitions?

Infant and child mortality

One question that often arises is how bad for one’s health living in a slum really is.

Given the poor sanitation in most slums, with open sewers, few toilet facilities, and a lack of

abundant clean water, along with the crowded and poorly constructed homes, it is natural

to think there may be health consequences to living in slums. Problems arise, however,

when comparing adults living in slums to those not living in slums. Many of those living in

slums migrated from rural areas, and their adult health is an accumulation of the

conditions and experiences over the course of their lives. In addition, there may be certain

kinds of people who are more or less likely to migrate to a slum, and this may cloud the

true health consequences of a slum. It is preferable, therefore, to study infants and children

in slums, as they have a cumulative experience confined to the slums they were born and

raised in. Questions of selection are still relevant as healthier parents who migrate may

produce healthier children, but these selection issues are more limited for children than

adults.

The first empirical chapter of the dissertation seeks to examine whether children

living in slums experience higher mortality than urban non-slum children or rural children,

both to age 1 and to age 5. If there are differences, I then set out to determine how much of

raw differences in mortality by place of residence are due to differences in individual and

family characteristics.

19

Child health

Mortality may be the most extreme measure of well-being, and so the second

empirical chapter of the dissertation looks to measures of children’s health. First, I question

whether slum children have higher incidence of acute illness, such as diarrhea, fever, and

cough. Slums are often thought to be breeding grounds of acute illness, as poor sanitation

and contaminated water sources found in slums may spread disease causing germs.

Nutrition is often looked at by public health researchers and epidemiologists as a

measure of well-being, and so I also consider how slum children fare in terms of both acute

and chronic malnutrition. For each of these outcomes, I also question how much of

residence differences can be attributed to family and individual characteristics.

Children’s education and work

Children’s schooling is of fundamental importance to social and economic

development. Typically, urban children have had higher rates of school attendance than

rural children. Does this hold true for slums? I examine whether slum children attend

school at rates on par with non-slum urban children, or the typically lower rates of rural

children.

Since school attendance is not the only activity children may be engaged in, I also

question the risk of children being involved in child labor, or not being involved in any

activity (school or work). Child labor is a topic of concern to international human rights as

well as development. Given the vibrant informal economies found in Indian slums

(Agarwala 2006), it may be that slum children are at a higher risk for working and have

lower chances of being in school than other children.

20

Taken together, these areas of study will paint a picture of slum disadvantage. We

will be able to quantify for the first time in what domains slum residents are

disadvantaged, and how to target anti-poverty programs and public health programs

accordingly. These are vital questions to answer in a rapidly urbanizing world.

DATA AND METHODS

The data for this project come from the third wave of National Family Health Survey,

conducted in 2005-2006. The survey is under the purview of the Ministry of Health and

Family Welfare of the Government of India, and is the Indian component of the

Demographic and Health Surveys. The NFHS interviewed 124,385 women ages 15-49 and

74,369 men aged 15-54. The survey was designed to obtain information on population,

health, and nutrition in India and each of its states, and is representative at the national and

state levels. It includes data on household characteristics, education, fertility, family

planning, infant and child mortality, child health, maternal health, adult health and health

care, mothering behaviors (such as breastfeeding), sexual behavior, HIV/AIDS, women’s

empowerment, and domestic violence.

Advantages of the Data

Despite some shortcomings in the NFHS-3 data, they are by far the best data for this

project. These data are the first to use rigorous data collection across India on slums. While

some other national surveys of India have included some markers of slum residence (such

as two waves of the National Sample Survey), even these surveys present a much less a

detailed picture of living conditions for each household surveyed. Additionally, the very

rich nature of the survey in asking about a vast array of child and woman’s outcomes along

with a large amount of background information on both individuals, families, and

21

households, make this data ripe for analysis of slum life. The comprehensiveness of these

data will allow me to conduct analyses of slum outcomes that no other research has been

able to do. Lastly, because the data are relatively new, little work has been published using

them yet, again making them ripe for analysis.

Sampling

Sample sizes of women in each state were set at between 1,500 and 10,000 based on

the size of the state and adjustments made to allow for estimates of state-level HIV

prevalence. To the best extent possible, the sample within each state was allocated to urban

and rural areas based on the proportion of the population that was urban and rural in the

2001 census.

For the first time, the third wave of the NFHS collected oversamples in eight cities to

obtain slum and non-slum estimates for population and health indicators. These cities are

Delhi, Chennai (Madras) in Tamil Nadu, Hyderabad in Andhra Pradesh, Indore in Madhya

Pradesh, Kolkata (Calcutta) in West Bengal, Meerut in Uttar Pradesh, and Mumbai and

Nagpur in Maharashtra. The oversamples include all slums, both notified and non-notified.

These oversamples of urban areas for slum and non-slum estimates make the NFHS data

unique and very rich in possibilities for looking at slum residents and disadvantage.

The NFHS uses a two-stage sample design in rural areas and a three-stage sample

design for urban areas. In rural areas, the first stage was selecting Primary Sampling Units

(PSUs), typically villages, and the second stage was systematically selecting households

within these PSUs. In urban areas, wards tend to be quite large, making it difficult to

accurately sample households within a ward due to census data restrictions. The urban

22

sampling therefore proceeded in a three stage process. First was the selection of wards;

next, one Census Enumeration Block (CEB) was selected within each ward; and finally,

households were randomly selected within CEBs. For the eight cities that were

oversampled, slum and non-slum Census Enumeration Blocks were selected, and

households were then selected from within the CEBS. Within both rural areas and urban

areas, samples were stratified by geography and female literacy.

The complicated sampling design of the NFHS raises questions about survey

weights. The NFHS calculates weights for several units of analysis: household weights,

woman’s weight, men’s weights, HIV weights for men and women, and domestic violence

weights. These purpose of these weights is to take into account the different selection

probabilities, particularly as samples were adjusted for HIV estimates and oversampling

urban populations. The weights also attempt to take into account differential non-response

rates in urban and rural areas and slum and non-slum areas. There are also state and

national women’s and men’s weights, which are not used in this study as no state or

national level estimates are made. The domestic violence weight was created to account for

the fact that the domestic violence module was administered only to a subsample of the

women’s sample.

Response rates and missing data

Overall, response rates for the survey are very high by any standard. The overall

household response rate was 98%. For women, the response rate overall was 92.4%, and

for men it was 84.9%. In general, urban areas had response rates lower than rural areas by

4 to 7 percent. By American sociological standards these are very high response rates. For a

comparison, we may consider the National Sample Survey, a very large household survey

23

run by the Indian government. The National Sample Survey Organisation (a branch of the

ministry of statistics) does not provide documentation, but reports 100% compliance.

Investigators using the survey have calculated very low non-response rates, at 6% or less

(Angus Deaton 2009). This is on par with the National Family Health Survey.

Missing data on demographic questions (such as age, age at marriage, education)

were very low, ranging from 0.01% missing data to 1.23% missing data. Data on health

questions had much higher proportions missing, ranging as high as 18% missing data on

select questions. Though there is an array of techniques for dealing with missing data

ranging from simple to very complex, none has yet been shown to be any better than

simple listwise deletion, and so that is the method used in this paper.

Data Limitations

There are a few limitations to the NFHS data. First, the main thrust of the survey is

on women (or reproductive ages) and children’s health, rather than on population health

more broadly. This limits the scope of possible outcomes a researcher can analyze using

the data to mostly question of children and women at the childbearing ages. For this

project, given the lack of quantitative research on any aspects of slum life, the

consequences for children and women are a good starting place. For further research

aimed at examining more adverse consequences of slum life, this may prove problematic

and other data will be necessary.

Although oversamples of slums in eight cities were collected, a nationwide slum

census was not. The eight cities vary in size, geographic location, and percent of city

24

residents in slums, and were chosen to be representative. Strictly, however, this is not a

random sample of slum residents nationally.

Currently the Indian Ministry of Health and Macro International are not willing to

release the geocodes collected with all the data. This means that analyses within cities or

between and within particular slums in not possible, making a strict test of neighborhood

effects impossible. Given the paucity of literature on any slum-related outcomes, however,

this study adds to the literature despite these shortcomings.

Lastly, with any data on a typically hidden or difficult to count population one must

question how representative of a sample a survey can achieve. This survey is no different. It

is difficult to know how well the NFHS data cover slum populations, and the organization

that runs the survey has no documentation or discussion of this.

It should be noted that the data on slum populations here is not statistically

representative of all Indian slums since this sample of slums includes only eight cities. It is

difficult to say, however, how one would obtain an accurate random sample of the full slum

population of India since several states did not report slum populations in the 2001 census,

used for creating the sampling strategy for the NFHS. Overall, this is the largest and

arguably most reliable survey of slum residents to date in India.

Measuring income and wealth in the NFHS

An additional problem in the NFHS and all DHS data for this study is the lack of any

income or consumption data. This makes it difficult to accurately distinguish between

income groups. The prevailing method of dealing with this problem is to create indices of

wealth using a principal component analysis a battery of survey items. These include what

25

the family owns, the nature of their house, landholding for rural households, and the

family’s source of light and heating. Often, the factor scores from the principal components

analysis are divided into quintiles for analysis. While this is the most common method used

to deal with wealth (Filmer and Pritchett 1999, 2001; Van de Poel, O'Donnell, and Van

Doorslaer 2007; Montgomery and Hewett 2005), it has many problems that have been

overlooked. In particular, the wealth index cannot distinguish between the very poor and

the very, very poor, often distinguished by food in other consumption surveys. In many

developing countries, particularly in Africa, the bottom three quintiles are more or less

undistinguishable from one another in terms of real wealth (Chaudhury, Hammer, and

Pokharel 2009). This makes the conclusions regarding the relationship of wealth and

outcomes drawn in several papers suspicious, particularly insofar as these papers deal

with health outcomes that may be directly affected by certain items included in a wealth

factor score. Owning a refrigerator, for example, is a mark of wealth but also may have a

more direct impact on health, as refrigeration of food can prevent bacterial growth.

Likewise, having a dirt floor makes it difficult to clean up spills that may harbor germs.

Because of the many given problems inherent in the standard wealth index, in this

study, I will not use an index of household wealth. Instead, I will include separately all of

the items included in a wealth scale, and use likelihood ratio tests to determine if these

items taken together as a measure of wealth are significant predictors of the outcomes. The

specific survey items are if a household owns a clock or watch, a bicycle, a radio, a

television, a sewing machine, a motorcycle or scooter, a refrigerator, a pressure cooker, a

car, a microwave, and a mobile phone; if a household has electricity; the kind of cooking

26

fuel used; if the kitchen is a separate room in a dwelling; and the number of rooms in a

dwelling.

Despite the potential faults of using these measures of assets and wealth as

compared to an income or consumption measure, this is the only survey that is so large and

comprehensive that can examine slum populations in India. Furthermore, some studies

have showed that wealth variables actually perform more reliably than consumption

measures when tested through instrumental variables (Filmer and Pritchett 2001). This is

thought to be because survey questions involving wealth are easier to answer than

questions involving consumption. For example, it is relatively easy for a respondent to

accurately answer “do you own a sewing machine?” compared to “how much rice have you

consumed in the past two weeks?”

Descriptive Statistics

Table 2-1 presents basic descriptive statistics of the NFHS data, broken down by

rural, urban slum, and urban non-slum residence. These descriptive statistics do not

include dependent variables used in the empirical chapters, but include important

independent variables and population characteristics.

Overall, one can see that rural areas have a slightly younger age distribution than

urban areas, and slums are slightly younger than other urban areas. Muslims are

overrepresented in slums compared to non-slum urban areas and rural areas, and

Scheduled Caste/Scheduled Tribes are overrepresented in slums compared to other urban

areas. Scheduled Castes and Scheduled Tribes are officially recognized by the government

India, formerly known officially as “depressed castes” and in everyday language are known

27

as untouchables. Officially, it is illegal to discriminate based on caste. Practically, however,

a great amount of caste discrimination still exists (Newman and Deshpande 2007; Newman

and Jodhka 2007; Newman and Sukhadeo Thorat 2007), and scheduled castes and tribes

lag behind the rest of India on socioeconomic status indicators (Radhakrishna and Ray

2005).

Additionally, the descriptive statistics show several housing indicators and percent

of households owning certain assets. Overall, the pattern that emerges is one of urban non-

slum areas being the most advantaged, rural areas being the most disadvantaged, and

slums falling somewhere in the middle.

OVERVIEW

Urbanization is taking place at an unprecedented pace in the developing world. The

United Nations projects that the world population will grow by 2 billion by 2030 and an

additional 2 billion by 2050, and 70-90% of this growth will take place in cities of the

developing world (Brockerhoff and Brennan 1998; Montgomery et al. 2003; United Nations

Centre for Human Settlements 1996). While rapid urbanization presents many challenges

including risks to physical environments, health conditions, social cohesion, and individual

rights, the most immediate concern is the massive increase in the numbers of urban poor.

The phenomenon of developing world urbanization has received some attention

from journalists and the popular press, but it remains understudied in academia, and in

particular within the discipline of sociology. Existing research, qualitative and quantitative,

focusing specifically on slums tends to be mainly descriptive. There is a need for rigorous

28

sociological research that analyzes poverty and disadvantage across cities, within cities,

and at local levels as small as neighborhoods.

This dissertation aims to fill these gaps in the literature. I focus on India, which

contains one-sixth of the world population and is urbanizing rapidly. Using new

quantitative data, the study aims to take a fresh look at the so called “theory of the urban

advantage.” Using the recently released 2006 National Family Health Survey (NFHS), I

examine on a national level the predictors of human capital outcomes in health and

education, comparing the urban slum residents with both rural residents and urban non-

slum residents. I hope to be able to unpack the urban advantage to discover under what

conditions India’s urban poor fare better than their equally poor counterparts outside of

the cities. I examine the urban advantage in four areas: child health and mortality,

children’s schooling and work, women’s health, and domestic violence.

Overall, the dissertation paints of much more nuanced picture of urban advantages

and penalties and their uneven accrual across groups. In chapter 2, I discuss why India is

the natural starting point for any study of slums. As the world’s largest democracy, rapidly

urbanizing, and home to around 1/6 to ¼ of the total world’s poor, it has at least 22% of

urban residents living in slums. Chapter 3 considers how to best define what a slum is.

Several definitions are examined including the official Indian census and government

definitions, the perception of surveyors of if an area is a slum, and the six point continuum

introduced by the UN-HABITAT definition. Comparisons are made, and a definition is

developed for use throughout the rest of the dissertation.

29

Chapter 4 looks at the case of infant and young child morality. A relatively large

literature has examined urban-rural differences in both child health and mortality. I

examine infant and child mortality as well as child morbidity outcomes to see if there are

differences between slum children and other children. I then examine what proportion of

these differences is due to population characteristics. I find that there are raw differences

by location of residence, with rural infants and children experiencing the highest mortality,

followed by slum children, and finally other urban children experience the lowest.

Multivariate survival models including family and background characteristics, however,

account for all of these differences. This is a new finding that runs in the face of most

conventional wisdom that expects slum residents to be far worse off given their living

conditions.

Chapter 5 examines children’s health, both in terms of nutrition and symptoms of

acute illness. For both chronic and acute measures of malnutrition, rural children are

disadvantaged compared to urban children, and with regard to chronic malnutrition slum

children are disadvantaged compared to their urban counterparts. Like the models of

survival, however, these associations with location disappear once family and background

characteristics are taken into account. Turning to the acute health symptoms of cough,

fever, and diarrhea, I find that residence has no effect on any of these symptoms. However,

I also find confusing effects of standard predictors of health such as income and maternal

education. These findings lead me to conclude that maternal reports of children’s health

are not un-biased indicators, but instead are most likely correlated with aspects of

socioeconomic status in ways that make these models not useful for looking at locational

differences

30

Children’s school attendance and work is examined in Chapter 6. Children’s

schooling is essential to economic and social development, and it has long been better

achieved in cities than in rural areas. Recent evidence suggests this might not be the case

for slum children, however. This chapter aims to see what disparities in school attendance

exist between slum children and other children. I also examine the proportion of children

working outside the home to determine what children are at risk for work and truancy

from school. I find that once family and background characteristics are taken into account,

rural children are actually the most likely to be attending school, and slum children are the

least likely. Slum children are also at a particularly high risk of working outside the home

before age 14. Theories on why these disparities exist are discussed.

Chapter 7 gives an overview and discusses fruitful areas for future research as well

as the future of slums.

31

TABLES AND FIGURES

Table 2-1: Descriptive Statistics from National Family Health Survey 2005-2006

Rural Urban slum Urban non-slum N – households 58805 7233 11297 N – women 67424 7952 12510 N – children 5-18 73859 7257 9135 N – children under 5 17850 1547 2153 Mean age (standard deviation) 27.8 (18.8) 28.2 (17.1) 31.2 (18.2) Percent over 65 4.9 3.3 5.0 Percent under 15 32.3 26.2 22.2 Household head (percent)

Hindu 74.4 71.4 77.8 Muslim 10.2 22.2 13.5 Other 15.4 6.4 8.7

Scheduled caste or tribe 37.6 25.9 15.5 Housing indicators

Percent with safe toilet 0.9 54.8 60.4 Percent with safe water 15.3 56.5 69.5 Roof is durable 34.4 81.7 86.2 Floor is durable 50.7 92.7 96.2 Walls are durable 74.4 72.5 88.7 Number of people per room 3.3 (1.9) 3.7 (1.9) 3.1 (1.9)

Wealth indicators (percent)

Owns a watch 75.0 89.3 94.7 Owns a radio 30.9 36.5 50.3 Owns a television 36.2 72.8 85.4 Owns a sewing machine 16.9 26.0 42.0 Owns a chair 51.8 65.1 83.8 Owns a refrigerator 10.4 27.0 53.1 Owns a pressure cooker 32.0 70.1 87.0 Owns a mobile phone 10.2 32.6 55.1 Owns a bicycle 44.9 37.8 47.1 Owns a motorcycle or scooter 12.1 19.8 41.5 Owns a car 1.8 2.6 12.4 Has electricity 65.2 95.2 98.4 Number of rooms for sleeping 1.8 1.4 1.7

32

Chapter 3 : Defining and Measuring Slums

“To me, the term ‘slum’ means somebody else defined my community in a way that

allowed them to justify destruction of it.”

Mel King, prominent African-American community activist

This chapter considers how to best conceptualize slums. There is little agreement

among scholars over how to define slums, though virtually all would agree that developing

a satisfactory definition is critical for rigorous study of the phenomenon as well as effective

policy for slum development or increased well being for slum residents. I first review the

complex issues that surround the creation of an official definition and then turn to the

competing models in use at present, for example in the Indian census and the continuum

introduced by the UN-HABITAT.

A thorough review of the literature makes it clear that housing tenure is a key

element in most definitions of slums. From this perspective, slums are illegal or squatter

settlements, and as such are defined by the high proportion of residents who lack any legal

claim to their dwellings. Definitions based on living conditions focus on the poor physical

conditions of slums, including poor sanitation, overcrowding, and substandard buildings.

Both of these conceptualizations have merit, but as I explain below, they are also

problematic in a variety of ways. Accordingly, I offer two new ways of conceptualizing

slums as neighborhoods of concentrated poverty and marginalization. My own

conceptualization of the slum builds on American scholarship on neighborhood poverty

and labor market status. My data from the National Family Health Survey is examined to

compare and contrast the application of different models of slums in order to illustrate how

33

the features chosen impact the way in which a community is labeled with all the negative

and positive connotations such a designation sustains.

These definitional debates cannot be understood without recognizing the powerful

political impact of each conceptual configuration. In India, in particular, qualifying for the

designation as a slum area opens the door to government provided benefits, creating

incentives savvy actors work to manipulate for their benefit.

THE COMPLEXITIES OF OFFICIAL DEFINITIONS

National governments, international organizations such as the United Nations, and a

multitude of non-governmental organizations have a stake in this question. Particularly

for organizations and governments working at the national and international level, a

uniform definition of slums is necessary for consistent statistical data, identifying problems

thought to be unique to (or perhaps most extreme in) slum neighborhoods, testing

programs and policies designed to eradicate slums or improve their condition, and

targeting policies so that they impact the areas considered to be slums (as opposed to

neighborhoods that do not belong in the category.

These problems have been with us for over a century as scholars and public officials

have wrestled with what is special (for the most part in a negative sense) about slums and

hence whether they justify special measures in public policy. As the discussion below

makes clear, the early debates revolved around the slum as a source of moral corruption

and public health problems. It took many decades before the health and well-being of slum

dwellers was the issue, as distinct from the unhealthful conditions in slums that could

infect the rest of society.

34

To understand the stakeholders in the current definitional debates, I conducted

fifteen interviews with government officials, leaders of non-governmental organizations

working in the area of slums, and researchers and public intellectuals working on the topic

in Mumbai, Delhi, and Bangalore. In December 2008, January 2009, and July 2009, I

questioned these varied experts on a range of topics including things such as how slum

dwellers think about the idea of slums, the official definitions of slums locally and

nationally, the state of slums in the country, and the variation between slums.

I also visited several slums with these expert informants in both Delhi and Mumbai.

In two slums in Delhi I conducted focus groups with slum residents, working with women

from a local NGO that does activist work to better the lives of slum residents. These women

organized the focus group for me, where I talked with residents on a variety of topics,

including the difficulties of slum life, how the government treats them, their daily life in the

community, and their coping strategies for living in the slums.

Drawing on this interview material as well as census and statistical bureau

documentation, state and municipality codes, and official legislation from various states

and municipalities concerning slums, I am able to consider the wide variety of ways in

which contemporary slum residents and the officials who provide or withhold resources

for them conceptualize the modern slum.

Historical definitions of slums

Adna Weber, writing in 1899 on London and New York, said, “Add to this the

overcrowding, the negligence of the public authorities in regard to the construction of

buildings and their sanitary condition, and you have the city slums, with their sickening

35

odor of disease, vice and crime” (Weber 1899, p. 414). Indeed, the concept of a slum is not a

new one, but in fact dates back to the 19th century cities of Europe and the United States.

Today’s discussions on slums now focus on slums in Latin America, Africa, and Asia, but

looking back historically much of the contemporary discussions mirror those from the

earlier set of slums.

In the United States and England in the first half of the 19th century, the most

extreme examples of moral and environmental degradation of slums set the tone for all

discussions of poverty (Ward 1989). Thus, it is often hard to tell in discourse from the time

if the term “slum” was used in reference to specific notorious places or to vast sections of

American cities where the poor lived. Rarely was the extent of slum conditions specified

with any precision; rather, the term generally referred to the most unsanitary and poorly

drained sections of the city, inhabited by the depraved poor. Closely tied to the idea of

slums was that of immigration, as new immigrants from Europe were often first settling in

slums due to their affordability. Thus, conceptions of moral poverty, filth, and disease

became associated with new migrants.

There was a good deal of sensational publicity about the horrors, depravity, and

squalor of the slums. This sensational publicity implied a complete moral depravity of slum

residents (Ward 1989). In fact, the activity of “slumming” was undertaken by well-to-do

Britons as an evening entertainment, visiting the slums to see astounding poverty in their

own cities (Koven 2006).

Ward (1989) conducts the most thorough review of the changing conceptions and

definitions of the word slum through the 19th century into the first part of the 20th century.

36

In the first half of the 19th century most poverty was attributed to moral defects and a lack

of self-discipline, and slums were seen as vast pits were such moral poverty bred

unchecked. Missions and Sunday schools developed in slums areas to help morally reform

the poor. Eventually, reformers decided that environmental factors were important;

however, bad environments were deemed important not for public health purposes but

because it was thought that living in bad environments and close quarters would breed

immorality. By the second half of the 19th century, many were starting to attribute disease

to public health conditions. Inadequate drainage, the accumulation of filth and excrement,

and the housing environment were all deemed important by the National Quarantine and

Sanitary Convention in the late 1850s, the organization that was the precursor to the

American Public Health Association. A shift started to happen whereby the living

conditions of slums and their impact on public health because the emphasis of most efforts

aimed toward slums. Still, the focus on slums was not devoid of questions of morality, as a

working assumption of many in the upper classes was that crowding and living in small

quarters led to promiscuity.

Assistance to the slums was given through a combination of public and private

sources. The moral aspect of poverty led many to try to determine a “scientific” way of

sorting out those who deserved aid or charity from those who were morally bankrupt and

simply did not want to work or clean up, and of course were not worthy of any aid.

Questions of economic resources were left out of the public discourse on slums almost

entirely, as the Protestant work ethic that dominated mainstream values suggested if slum

residents would simply work hard enough they could get out of their current state.

Therefore, those who did not get out of the slums were seen as simply lazy, and the public

37

and policy discourse on slums and poverty were silent on questions of jobs and economic

problems.

By the 1880s the idea of slums had moved to the image of tenement houses, and

over-crowding (rather than dilapidation) was given the most emphasis in discussions of

slums and tenements. The crowding of tenement houses was thought to breed poverty,

crime, and intemperance among its residence. Debates still raged on whether the poor

conditions caused these vices, or whether the vices caused the poor living conditions, and

aid and policy as such were divided into moral aid policy and housing aid and policy.

At the turn of the century and in the early 1900s, finally the focus of the causes of

poverty and slums shifted from personal causes to economic and social causes.

Overcrowded housing was the dominant measure of the slum, thought to cause unsanitary

and unhygienic living conditions. Policies focused on urban decentralization and the

introduction of public spaces such as playgrounds to de-congest inner cities. After World

War I, the idea of ghettos began to rise and the use of the word slum started to decline.

Ghettos were seen as much more complex, segmented worlds than simply isolated slums.

As immigration continued to rise, immigrants became part of what were seen as ethnic

ghettos rather than slums.

The definition and meanings attached to the word slum have not changed much in

over a century, though rather than looking at slums in Europe and the United States we

now focus on the slums of Asia, Africa, and Latin America. The ideas of overcrowding, poor

sanitation, and (to a lesser extent but still vibrant in public discourse) moral poverty of

slum residents are still very salient today.

38

United Nations slum definitions

The United Nations Millennium Development Goals include targets intended to

encourage attention to the problems of slum dwellers, namely: “By 2020, to have achieved

a significant improvement in the lives of at least 100 million slum dwellers”(United Nations

Human Settlements Programme 2003). To further this cause, the UN convened a

conference of experts in Nairobi, Kenya in 2002 for the purpose of determining measurable

characteristics of slums that could be used as benchmarks for progress. A review of slum

definitions used by governments, statistical offices, and international organizations

involved in slum issues led UN-HABITAT to develop the following list of attributes of slums.

Lack of basic services. A lack of basic services is one of the most frequently

mentioned aspects of slums in definitions worldwide. Safe water and sanitation are the

most important features, along with electricity, waste collection, surfaced roads or

footpaths, rainwater drainage.

Substandard housing or illegal and inadequate building structures. Slums are often

associated with a high proportion of sub-standard or very dilapidated housing, built with

non-permanent materials. These buildings often violate local building codes either because

of the building structure or because of the space or dwelling placement laws.

Overcrowding and high density. Overcrowding can be the result of multiple factors,

including high occupancy rates, low space per person, multiple families sharing one

dwelling, and many single-room units. The high density with slums has been a part of both

current and historical definitions of slums.

39

Unhealthy living conditions. Unhealthy living conditions are often the result of a lack

of basic services. For example, lack of sanitation can cause visible, open sewers that are

breeding grounds for disease.

Hazardous locations. Slums are often built on hazardous locations or places not

suitable for settlement, such as on floodplains and river banks, in close proximity to or built

atop waste disposal sites (such as landfills), or close to industrial plants producing toxic

emissions.

Insecure tenure; irregular or informal settlements. Many definitions include

illegality as a measure of slumness, and a number consider informal or unplanned

settlements as synonymous with slums.

Poverty and social exclusion. Poverty is a central characteristic of slums. The UN-

HABITAT report is careful to note that rather than being an inherent characteristic of

slums, it is an underlying cause of slum conditions. Social exclusion is a concept often used

in Europe, referring to not only poverty but other forms of exclusion such as

discrimination.

Minimum settlement size. Often definitions include something about settlement

size, either physically or in terms of number of households. This indicates the idea of a slum

as a neighborhood, not a single dwelling.

The conference of experts developed a set of criteria and, along with them, goals

intended to spur international concern and investment. They chose five criteria from the

list above, deciding to not use any social criteria (such as poverty) but instead opting for a

40

definition that relies on physical and legal characteristics of a slum. The criteria chosen

were: access to improved water, access to improved sanitation, residences using durable

building materials, housing of adequate space (not overcrowded), and security of tenure in

place of residence. Table 3-1 provides the specific definitions and thresholds decided on by

the conference for measuring each of these conditions.

Researchers have taken these measures and used them to define slums. Most

commonly, scoring low on any of these dimensions qualifies an area as a slum (Mugisha

2006), although some researchers treat these measures as elements of a continuum of

“slumness” along which neighborhoods fall (Weeks 2008).

The Indian Case

The federal government of India has no unified definition of slums. Still, several

units within the federal government have developed definitions of their own. The National

Sample Survey Organization, a division of the ministry of statistics, uses the following

definition to determine what slums are for the purposes of its national survey of

households.

“A compact urban area where at least 20 households live, with a collection of poorly

built tenements, mostly of temporary nature, crowded together usually with

inadequate sanitary and drinking water facilities in unhygienic conditions”

(National Sample Survey Organisation 2003).

The National Sample Survey Organisation uses its definition to collect data on slums in

national surveys every five to ten years. The definition is ambiguous at best, and hence

41

runs the risk of lumping together areas that are quite different and ignoring other areas

that bear many similarities to those that are not officially designated.

The Indian census started collecting data on slums in its 2001 national census. The

census bureau defines slums in the following way:

“For the purpose of Census of India, 2001, the slum areas broadly constitute of:

(i) All specified areas in a town or city notified as “Slum” by State/Local

Government and UT Administration under any act including a “Slum Act”

(ii) All areas recognized as “Slum” by State/Local Government and UT

Administration, Housing and Slum Boards, which may have not been formally

notified as slum under any act;

(iii) A compact area of at least 300 population or about 60-70 households of

poorly built congested tenements, in unhygienic environment usually with

inadequate infrastructure and lacking in proper sanitary and drinking water

facilities” (Census of India 2001 n.d.).

Under Indian law, a “notified” slum is recognized as legal by the government, entitling the

slum and its residents to the privileges of any person residing in a legal residence. For the

slum, these benefits, which are the fiscal and managerial responsibility of the municipality,

may include trash collection, drinking water, storm drains, community latrines, streetlight,

and roads. For residents, benefits include having a permanent address for obtaining a

government sponsored bank account and a ration card. All other slums – those that are not

“notified” – are considered illegal, and their residents are deemed lacking legal addresses.

The process of becoming a notified slum is highly political, and for this reason part ii of the

42

definition is important: it specifies that for census enumeration, all slums, whether they are

notified or not notified, are counted as slums and are not distinguished from one another.

The first source of ambiguity in this official definition stems from the fact that every

state and local government has the ability to define or identify slums in its own way, and

the federal government uses these designations in determining slum statistics at the

national level. While this decentralization is useful for state and local governments dealing

with problems that may be quite different from one region to another, it makes the

measurement of slums at the national level very problematic. For instance, in the 2001

Indian census five states reported having no slum populations. These states are by no

means richer or more developed than the states with abundant slum populations. Rather,

these are small states have not defined or recognized slums, and therefore for the purposes

of the census they have no slums. Even official census reports recognize this inconsistency,

often reporting the proportion of residents that live in slums excluding states that reported

no slum populations. This is an extreme case that illustrates the problem of deferring to

local and state definitions of slums when attempting to analyze slum conditions across the

country.

Part iii of the Indian census’s definition of slums is equally ambiguous, giving a

portrait of what a slum would look like without being specific. Words like “about”,

“usually”, and “normally” in the definition indicate a degree of flexibility in determining if a

place is a slum or not. In theory, it allows the enumerator to use his or her good judgment

on whether a place is a slum, given the difficulty of defining exactly what constitutes a slum.

In practice, however, such an ambiguous definition makes for difficulty in consistent

43

measurement across census enumerators, who may have slightly varied interpretations of

these guidelines.

TABLE 3-2 shows descriptive data on households defined as living in slums based on

the 2001 census and their living conditions. A full 8% of households defined as slum

households in the census exhibit none of the slum conditions defined by the United

Nations. Alternatively, 29% of urban households not classified as in slums based on the

2001 census exhibited three or more slum characteristics. This shows some of the

ambiguity or noise that may be produced by the census definition of slums.

Lastly, it is worthwhile to note here that slums are known by many names in

different parts of India. A jhuggi jhopri cluster is a set of small, roughly built house or

shelter, occupied by laborers working on a construction project but often remaining long

after the project is completed. In Calcutta, slums are known as katchi bastis or “squatter

settlements”, described by the Calcutta Municipal Act as a collection of huts that are small,

temporary, and flimsy houses or huts covering an area of at least one sixth of an acre, built

without permission. Slums are called zoppad-patti in Marathi, referring to hutment

colonies. In Madras, slums are called cheris, a Tamil world meaning “hamlet”. In Delhi and

Kanpur, one kind of slum are katras, referring to a group of poorly built, single room huts

facing a small courtyard with a single door to the street (Banerjee 2002; Agnihotri 1994).

CONCEPTUALIZING SLUMS

The two current dominant ways of identifying or conceptualizing slums is by

physical infrastructure or by legality, while new alternatives might consider social aspects

of slums as central to the definition of a slum.

44

Slums as places of poor infrastructure

Many policymakers, researchers, and non-governmental organizations tend to think

and talk about slums as urban neighborhoods characterized by very poor infrastructure.

This conception of slums is present in the Indian census, the Indian ministry of statistics,

and the United Nations definitions of slums. These definitions tend to rely heavily on access

to basic services that require infrastructure (such as piped water, toilets and sewers) as

well as building materials of residences when determining whether or not a place is a slum.

While the infrastructure of slums is indeed a very salient and important part of slum

life that demands public attention, defining slums by the physical conditions is largely

unsatisfying. Rather than identifying the key component that distinguishes slums from non-

slums, I contend that poor infrastructure is one piece or symptom of slumness. Defining

slums based on physical characteristics would be akin to defining schools by the presence

of school buildings and playgrounds, rather than by the relations of students, teachers, and

learning that define schools. Additionally, there are many places with poor infrastructure

that would not be considered slums.

There are several reasons to believe poor infrastructure and physical characteristics

alone do not distinguish slums from other urban areas. Slums come to be known as much

more than simply neighborhoods with poor infrastructure. In fact, sociological research in

India has shown that slums come to have stigmas associated with them. For instance,

studies show that slum residents often have difficulty obtaining employment when

employers discover they are from the slums (Bhatt 2000; Lobo and Biswaroop Das 2001).

This clearly implies that slums are far more than simply areas with poor infrastructure, and

45

furthermore, the lack of infrastructure alone is not sufficient for determining what is a

slum, although many current definitions identify slums on infrastructure.

Slums as squatter settlements

A second major way policymakers and researchers conceptualize slums is as urban

squatter settlements. In this view of slums, slums are illegal communities residing on land

to which the residents have no right, such as land owned by the government, a university, a

corporation, or other private owner. This is evident in the United Nations definition of

slums as the fifth measure of slum improvement: security of tenure. When illegal,

settlements constantly face the threat of eviction and residents have no security of tenure.

The question of whether or not this is a defining characteristic of slums is debatable.

The notion of slums as squatter settlements is appealing because of its inherent

implication of illegality. Indeed, many slums are illegally built on public and private

properties, such as university land, airports, and land adjoining railroads and highways.

When this is the case, slum residents often fear the demolition of their slum when either

the government or the private owner wishes to rid their property of the blight of slums.

These demolition projects are not solely threats, but do often happen, and slum residents

often have little recourse as they have no legal right to the land they are on. In India,

sometimes court cases are brought on behalf of slum residents seeking an injunction

against demolition, but typically these cases fail, if they are brought at all.

The illegality of slums is evident in several ways. Most slums around the world

spring up as settlements without any municipal services such as garbage removal. Slum

residents often campaign and advocate to have their slum recognized as a neighborhood

46

and to receive the according services (Angotti 2006; Davis 2006; Neuwirth 2005). This is

certainly the case in India. Until a slum is notified, neither the residents nor the community

are entitled to many of the services residents of the rest of a city receive. An important

aspect of slum life is the lack of public resources, often (though not always) a product of

illegality. Clearly, illegality can be and often is a very salient aspect of slum life.

Defining slums as squatter settlements is appealing, but again ultimately

unsatisfying. Many slums are eventually recognized as legal by their municipality and being

receiving municipal services. In India there is an official recognition called declaration or

notification, by which a slum is declared legal and eligible for municipal services. Defining

slums as squatter settlements would suggest that these neighborhoods would cease to be

slums once they are legal and no longer in danger of demolition. Realistically, the line

separating slums and non-slum neighborhoods is not the line between legal recognition

and ignorance.

In addition, even legal slums are not immune from “redevelopment”, a more

common euphemism for demolition. The extension of land tenure rights over government

land is locally known as patta in India, and often patta is extended to squatters and slum

settlements as a welfare measure (Banerjee 2002). These tenure rights are often granted as

part of a Slum Act or when a slum is notified and becomes legal. Still, an individual or family

having a patta is not safe from redevelopment or eviction; rather, having patta simply

entitles the person to a sum of money or an alternative plot of land if the government takes

over the land for redevelopment. For instance, the city of Mumbai has long been

attempting to redevelop its slums in efforts to modernize and internationalize the city. The

47

city has begun partnerships with private developers, encouraging private developers to

build new, modern apartment buildings, of which a portion of apartments are allocated for

the current slum residents, while the rest become the property of the developer to use for

profit.

Many Indian states have official Slum Clearance Boards charged with (among other

things) clearance and redevelopment of slums, both legal and illegal. Though not facing the

same immediate threat as smaller slums built alongside railway tracks or highways, all

slums are viewed to some extent as a blight upon the face of modern India, and particularly

on the face of individual cities competing to attract foreign investment and foreign

businesses. Thus, though illegality is often an easily recognizable attribute of slums that

may contribute to fears of eviction or demolition, illegality alone does not produce slum

conditions, and nor is it universal to all slums. Thus, defining slums as illegal or squatter

settlements is not a broad enough definition to encapsulate slums.

Slums as neighborhoods of concentrated poverty

An as yet unexplored but potentially fruitful way of conceptualizing slums is as

neighborhoods of concentrated poverty. A large body of literature in American sociology

has focused on the concentration of poverty in American cities, noting that neighborhoods

of concentrated poverty differ considerably from non-poor neighborhoods. Though there is

much debate over the specific dimensions along which high-poverty and other

neighborhoods differ, there is agreement over several aspects of life in high poverty

neighborhoods. First, there is often a dearth of legal economic activity, evidenced by the

lack of businesses and jobs that are available in poor neighborhoods. Second, these high

poverty areas tend to have fewer public resources and get less public attention than other

48

neighborhoods. For instance, politicians often ignore high poverty, segregated

neighborhoods unless as a result of gerrymandering they have incentives to campaign

there. Schools and other public services tend to have far fewer resources to tackle larger

problems, and therefore provide lower quality education and other public services. High

poverty neighborhoods are beset by a host of problems as a result of the concentrated

poverty.

There are several reasons slums may be better conceptualized as neighborhoods of

high poverty. Many researchers and policymakers talk about slums with poverty being the

underlying cause of the many bad living conditions residents face. The United Nations in

fact states that “Income or capability poverty . . . is not seen as an inherent characteristic of

slums, but as a cause of slum conditions” (United Nations Human Settlements Programme

2003, p. 11). If in fact we consider slum conditions to be the result of highly concentrated,

abject poverty, then it makes much more sense to define and measure slums based on

poverty, rather than the outcomes of slum conditions such as poor buildings or lack of

services.

In terms of bringing this concept down to the level of measurement, here we may

consider importing the commonly used definition of a high-poverty neighborhood from the

US context, where a neighborhood where over 40% of the residents are poor is considered

high-poverty. In the developing world context, a consistent poverty line would be needed

to operationalize this definition. There are two good candidates for a poverty line. First,

India has an official poverty measure, though there is considerable debate over a change in

the method of measurement in the 1990s (see Deaton and Kozel 2005; 2005 for a review).

49

According to the Indian poverty line, around 27% of Indians were below the poverty line in

2005 (Planning Commission 2007). This official poverty line is adjusted by state and urban

versus rural residence. The second possible poverty line to use for measurement would be

the World Bank’s international absolute poverty line of $1/day in 1998 dollars,

approximately $1.25 today. The best estimates are that in 2005 approximately 54% of

India fell below this line (Beinhocker, Farrell, and Zainulbhai 2007). There are arguments

for using either of these poverty lines in a high-poverty neighborhood measure, and most

likely looking at both would provide insight into which one is more appropriate for

measuring the highly concentrated poverty of slums.

This social (as opposed to physical or legal) conceptualization of slums may help

policymakers and researchers better measure and identify slums and their attendant

problems.

Of course, this concept, definition, and operationalization would require testing and could

prove problematic for measurement. Many household surveys and even the Indian census

collect no income data, so until this is done it may be difficult if not impossible to identify

areas of concentrated poverty in India. However, it is likely this definition would prove

extremely useful and much more consistent if implemented.

Slums as marginalized or socially excluded communities

Related to conceptualizing slums as neighborhoods of concentrated poverty is the

idea of slums as marginalized communities, cut off from many public and social benefits

enjoyed by the more mainstream or dominant areas. The related concept often used by

European social scientists is social exclusion. Using this framework of conceptualizing

50

slums, things like poor infrastructure and illegality are not defining characteristics of

slums, but rather visible symptoms of slums being marginalized from public goods. There is

no question that slums are cut off from many municipal services. Additionally, stigmas are

often attached to slums. Slum residents are discriminated against; for example, in one study

in India slum residents had difficulty getting a job when employers found out they resided

in slum (Bhatt 2000; Lobo and Biswaroop Das 2001).

While some argue that the marginalization of slums is largely due to the illegal

nature of slums, this is most likely not the real story. First, in India, many slums are actually

legal, notified developments, making the connection between marginality and legality

tenuous at best. For instance, Mumbai’s best-known slum of Dharavi is a notified, legal

slum. Even slums that are legal are often neglected by municipal governments and do not

receive municipal services that non-slum neighborhoods typically receive, such as garbage

collection, and as discussed previously, even legal slums are not immune from

redevelopment or clearance.

Sociological research has long shown that marginalization of neighborhoods can

happen independently of legality. In western countries such as the United States, for

example, predominantly black communities with high rates of poverty are by all means

legal and entitled to all the public services accorded residents of any other kinds of

neighborhoods. Still, this is far from the daily realities these segregated neighborhoods

faces. Schools in segregated, high poverty communities tend to be underfunded and have

few resources. Public resources such as parks and libraries are often either located outside

of high poverty neighborhoods or are in large part neglected. In addition, research has

51

shown how residents of some high poverty or predominantly minority areas are

discriminated against when job searching as a consequence of their residential location

(Newman 1999). It is clear from the American literature that the distinction of legality is

not the mechanism producing public neglect and marginalization of a neighborhood.

From my focus groups with slum residents as well as interviews with local activists

working for NGOs, it is marginalization that underlies their feelings about living in slums. I

conducted a focus group with women in a small, Muslim slum in Delhi. The residents were

quick to enumerate the many problems they faced, from problems with water, buildings,

and children attending school, the feelings that undergirded all of these were feelings of

marginalization. They were not simply upset that they had bad water or no nearby medical

clinic: they were upset that public officials for decades had not given them a second glance.

In fact, it took the aid of an outside NGO to get the local government official to respond to

any of the problems slum residents had. Clearly, marginalization was the most salient

aspect of slum life to these women residents.

The distinction between developing countries and the developed world when

thinking of marginalization is the degree of deprivation in slums. Given the overall lower

level of development in India and much of the developing world, marginalization has much

more dire consequences than in a developed country. Municipal governments are often

overwhelmed by rapid growth and particularly rapid growth of the poor segments of a

population. These governments often lack fiscal and other resources, making

marginalization much more disastrous for slum residents. For instance, Bangalore, India

(the country’s third largest city at 5.1 million, and a relatively prosperous city from the high

52

tech industry) in 1998 had a per capita municipal government revenue of US$25.31 and

per capita expenditures of $26.35, compared to developed world cities of comparable sizes

such as Washington, DC with per capita revenues of US$2,379 and expenditures of $658, or

Madrid, Spain with per capita expenditures of US$547 (United Nations Center for Human

Settlements 1999). As Montgomery, Stren, and Cohen (2003) note, developing world cities

lack not just financial resources but also capacity for urban services and service delivery,

due to a lack of services as well as a generally severe shortage of trained, professional staff

to deal with complex municipal government problems.

The reasons why slums are marginalized could be the subject of interesting and

fruitful further studies. There are several overlapping and related factors that may

contribute to the marginalization of slum areas. For instance, poverty is concentrated in

slums, as are members of backwards castes, Muslims (who face even greater

discrimination (Attewell and Madheswarank, Forthcoming)), and more generally workers

in informal, low-skilled jobs. Just as high poverty, predominantly black neighborhoods tend

to become economically and socially marginalized, we may hypothesize that a similar

process happens with slum settlements.

Once a place becomes known as a slum, the label alone carries a stigma. As

Zerubavel (1996) notes, when official classifications are internalized, the categories are

perceived to be internally homogenous. This clearly happens with the label of “slum” in

India, as all slums are perceived to be homogenous, and often (as in the 19th century), the

worst slums become the model of a slum in the minds of the public. Slums are often

characterized as places that are unsafe, havens of crime and backwards thinking. As a white

53

foreigner doing fieldwork in India, I was often warned not to venture to these “dangerous

places”. I was told by one upper class Indian man, “I was so worried about you going to [the

slums] . . . Then I heard what your research was about, and I knew you must be a very brave

woman.” Few I spoke with knew the specific slums I was visiting, making clear that the

perceptions were not specific to certain areas, but rather specific to the label of “slum”. The

label of slum certainly carries many connotations with it. No rigorous research to date has

examined whether crime is actually higher in slums than non-slums, but perception is

widespread anyway.

In terms of measurement, it may be relatively difficult to measure marginalization.

One more simple way may be to return to the idea of concentrated poverty previously

discussed. It perhaps the very concentration of poverty in slums that allows politicians and

other citizens to ignore and overlook slums, leaving them isolated and marginalized. As

Doug Massey discussed in his Presidential address to the Population Association of

America, the rich and poor are becoming increasingly segregated in the cities of the

developing world, allowing the affluent to remove and separate themselves from the poor

(Massey 1996). In this case, neighborhoods of concentrated poverty may be at the heart of

marginalization, and may be the best way at measuring slums.

THE POLITICS OF SLUM DEFINITIONS IN INDIA

Once there are top-down, official definitions of phenomena such as slums, various

consequences follow. In the case of slums in India, there is maneuvering and politicking

that goes on around two aspects of the slum definitions: the legal vs. illegal slums, and

slums vs. non-slum building construction. Starting at the larger level of slum versus non-

slum building, savvy actors do all they can to take advantage of being considered part of a

54

slum instead of part of a non-slum neighborhood. As Starr (1992) discusses, official

categories, such as slums, become frameworks of incentives as states organize benefits

around categories; those who can qualify for incentives or benefits may adjust their self-

descriptions and self-conceptions to match the official definitions and classifications. This

is the process we see going on among owners and builders in Mumbai’s slums.

For instance, businesses located in slums do not have to pay certain taxes, creating

an incentive for businesses to agitate for the areas in which they are located to be declared

official slums. Within the municipality of Mumbai, buildings cannot be considered part of

slums if they are more than two stories tall. As such, savvy builders purposely erect

buildings only two stories tall, though the buildings are sound enough to support more

stories. A similar form of boundary work goes on not just with height of buildings but with

building materials. In India, kachha building materials are literally “raw” building materials,

including non-durable materials such as mud, untreated wood, bamboo, and thatching.

Durable buildings are called pukka building, because they are made of stone, burnt brick,

and cement. Again, savvy businessmen may leave portions unfinished, most often roofs, or

use kachha building materials so that the buildings and owners are not subject to regular

taxation.

While businesses derive benefits from slum location, residents suffer disadvantages

when their neighborhoods are declared slums. Previous research in India has found that

employers look askance at slum residents when they compete for jobs, and children from

slums are labeled as delinquent or unmotivated to work in school by their teachers (Bhatt

2000; Lobo and Biswaroop Das 2001). Slums are considered havens of crime and poor

55

morals (Bhatt 2000; Lobo and Biswaroop Das 2001). Slumlords and business owners

generally have the upper hand when it comes to the notification process and the

disadvantages accruing to ordinary residents are swept aside.

Yet the ordinary residents will not be entirely opposed to living in a declared or

notified slum because they will be eligible for many services and benefits once this

happens. Activist groups organize and agitate for notification because it will result in the

extension of these vital services. The process of notification has as of yet been little

studied, and the topic warrants an entire book. Suffice to say here that the very existence

of the boundaries defining slums has led to quite a bit of maneuvering, politicking, and

boundary work – meaning attempts to influence the very definition of slums – by groups

that stand to benefit or lose based on the classification of their slum.

MEASURES OF SLUMS FOR ANALYSES

If the conceptual problems with the definition of slums are legion, the availability of

quantitative data makes the empirical task of identifying slums even more problematic. In

the National Family Health Survey, there are three separate ways to identify slums. First,

the survey includes an indicator of whether or not a neighborhood was recorded as a slum

in 2001 census. Second, the surveyor for the NFHS recorded whether he or she would

consider the place a slum. Lastly, four of the five United Nations slum characteristics are

included in the survey data (housing materials, water source, toilet, and crowding).

To compare the various definitions of slums in the National Family Health Survey

data, I ran many analyses to determine how these definitions of slums compare to one

another. Table 3-2 shows descriptive statistics on living conditions and slum factors for

56

both census defined slums and surveyor defined slums. As is clear from these descriptive

statistics, the neighborhoods identified as slums by the census and surveyors are very

similar in housing characteristics, usually within 1 percentage point difference on each of

the characteristics. Table 3-3 shows bivariate relationships between the census and

surveyor identified slums.

Throughout the rest of the dissertation, I present analyses using the census defined

slums as the official identifier and, as well, ran all of my analyses using the surveyor

identified slums. The results were substantively the same. Using the census definition is

most useful mainly because it is the most widely used definition of slums among

researchers working in India and is most easily replicated in future studies. I also include

the UN housing characteristics in my models separately, so as to look at the effects that

each of these conditions might have on my various outcomes.

DISCUSSION

There are many reasons that finding a unifying concept or definition of slums is

difficult. Slums are a complex, multidimensional phenomenon. They are in some ways local

and relative, as the appearance a slum in India may differ substantially from a that of slum

in sub-Saharan Africa. Furthermore, slums today are rapidly changing, as urbanization is

taking place before our eyes.

I contend that conceptualizing of slums as places of poverty and social exclusion is

both closer to the fundamental causes of slums as well as a more consistent way of

measuring of slums. This is not to say that the physical and legal factors should no longer

be measured, monitored, or a focus of policy concern. On the contrary: governments,

57

international organizations, and NGOs must continue to look at poor building materials,

poor sanitation, water supply, hazardous locations, and issues of land tenure. All of these

have important negative consequences, and considerable public effort must continue to

focus on improving these physical and legal conditions for India’s cities to attain better

public health, fewer social inequalities, better safety for all their residents, and security for

all their residents. Yet we must be careful not to equate these negative factors with one

another or conflate the consequences and causes. While physical factors have many

negative consequences, the social exclusion and stigmatization of slums indicates that

slums are far from simply dense places of poor infrastructure. Rather, slums are complex

social spaces that arise from crowded, abject urban poverty and become known as far more

than a set of crowded, poorly built huts.

58

Table 3-1: Indicators and Operational Definitions of Slums, from United Nations Human Settlements Programme

(2003)

Characteristic Indicator Definition Access to water Inadequate

drinking water supply

A settlement has inadequate drinking water supply if less than 50% of the households have an improved water supply:

Household connection; Access to public stand pipe; Rainwater collection,

with at least 20 litres/person/day available within an acceptable collection distance Access to sanitation Inadequate

sanitation A settlement has inadequate sanitation if less than 50% of households have improved sanitation:

Public sewer; Septic tank; Pour-flush latrine; Ventilated improved pit latrine.

The extreta disposal system is considered adequate if it is private or shared by a maximum of two households

Structural quality of housing

a. Location Proportion of households residing on or near a hazardous site. The following locations should be considered:

Housing in geologically hazardous zones (landslide/earthquake and flood areas); Housing on or under garbage mountains; Housing around high-industrial pollution areas; Housing around other unprotected high-risk zones (e.g. railroads, airports, energy

transmission lines). b. Permanency of

structure Proportion of households living in temporary and/or dilapidated structures. The following factors should be considered when placing a housing unit in these categories: Quality of construction (e.g. materials used for wall, floor and roof); Compliance with local building codes, standards and bylaws.

Overcrowding Overcrowding Proportion of households with more than two persons per room. The alternative is to set a minimum standard for floor area per person (e.g. 5 square metres).

Security of tenure Security of tenure Proportion of households with formal title deeds to both land and residence. Proportion of households with formal title deeds to either one of land or residence Proportion of households with enforceable agreements or any document as proof of a tenure

arrangement.

59

Table 3-2: Descriptive statistics on housing conditions and slum status of

households in the National family Health Survey, 2005-2006

Rural Census defined slum

Survey defined slum

Other urban (census)

Water Piped 15.3% 55.7% 56.5% 69.5% Public tap 17.1 24.7 25.8 12.1 Other 67.6 19.6 17.8 18.4

Toilet

Flush 28.4 91.9 91.5 96.7 Pit latrine 10.3 0.8 0.8 0.5 No toilet 61.3 7.4 7.8 2.8

Electricity 65.2 95.8 95.2 98.4 Mean people per room 3.3 (1.9) 3.6 (1.9) 3.7 (1.9) 3.1 (1.7) Building materials

Durable floors 34.4 81.2 81.7 86.2 Durable walls 50.7 93.1 92.7 96.2 Durable roofs 74.4 76.0 72.5 88.7

Total slum factors

0 0.2 8.1 7.5 19.9 1 5.6 28.0 29.1 37.0 2 29.7 34.4 34.1 26.6 3 47.5 23.6 22.9 14.2 4 17.0 5.8 6.4 2.4

Secure from eviction1 73.5 70.6 84.1 N 58,799 8,669 7,233 9,906 1 Questions about if respondents felt secure from eviction were only asked in Mumbai and Kolkatta, N= 4,478

60

Table 3-3: Counts of urban household by census and NFHS-3 surveyor identified

slum status

Census defined

Slum Non-slum Total

Surveyor

defined

Slum 8,234 3,063 11,297

Non-slum 1,627 5,606 7,233

Total 9,861 8,669 18,530

61

Chapter 4 : Infant and Child Death in India

INTRODUCTION

Levels of child mortality in the developing world have declined dramatically over

the past few decades; however, great variation across geographic and population groups

persists. Urban-rural differentials in infant and child mortality have been found to be large

and significant in many cross country studies (Puffer and Serrano 1973; Behm and

Primante 1977; Hobcraft, J. W. McDonald, and Rutstein 1984a; Mensch, Lentzner, and S.

Preston 1985).

The increasing share of slum residents in the cities of India complicates questions of

the urban advantage. As Indian cities grow larger and poorer, slums become more and

more central to questions of urban and rural mortality differentials. To date, no

researchers have quantified slum children’s mortality rates and how these compare to

rural and urban non-slum residents. I examine infant and child mortality, first determining

the raw differences in mortality by place, and then looking deeper to determine to what

extent differences by place of residence are explained by individual and family

characteristics that are unequally distributed by place of residence.

Results show that slum children face better chances of both infant and child survival

as compared to rural residents, but worse chances compared to urban non-slum residents.

Reasons behind these patterns are examined through sets of other population factors may

be affecting infant health, including families’ wealth, parents’ education, maternal age at

birth, sex of the child, short birth intervals, birth order, and family’s caste and religion. The

effects of residence disappear entirely once holding constant the individual and family

characteristics, indicating that residence itself does not matter, but instead mortality

62

differences by place of residence reflect the underlying, uneven distribution of

socioeconomic status across rural areas, slums, and urban non-slum areas.

THE URBAN ADVANTAGE IN CHILD MORTALITY

Historically, infant mortality in the United States and Europe was higher in urban

areas than rural areas. For example, 1860 Swedish infant mortality was 100 deaths per

1000 live births in rural areas, while it was 290 deaths per 1000 live births in the largest

city, Stockholm (Hofsten and Lundström 1976). In the United States in the 1890s, infant

mortality was 22% higher in urban areas (Samuel H. Preston and M. R. Haines 1991), and

similar patterns were true of France, Germany, and England (Samuel H. Preston and van de

Walle 1978; Woods and Hinde 1987; Kintner 1988). Infant and child mortality in Europe

and the United States began to decline in the same pattern as overall mortality with the

advent of investments in public health and improved health care, as discussed in the

introduction, reaching the point of urban infant mortality being significantly lower than

rural infant mortality.

In the contemporary developing world, urban residence has typically been

associated with lower mortality. This finding of an urban advantage in infant mortality is

consistent across many countries and data sources, including the World Fertility Survey

(Hobcraft et al. 1984a; Rutstein 1984), the Demographic and Health Surveys (Sullivan,

Rutstein, and George Bicego 1994; Van de Poel et al. 2007), and national census data

(Mensch et al. 1985). Urban-rural differentials in infant and child mortality have been

found to be large and significant in many cross country studies (Puffer and Serrano 1973;

Behm and Primante 1977; Hobcraft et al. 1984a; Mensch et al. 1985; Brockerhoff 1995;

Brockerhoff and Brennan 1998; Cai and Chongsuvivatwong 2006; Cleland, George Bicego,

63

and Fegan 1992; Gould 1998; Sastry 1997; Wang 2003; Van de Poel et al. 2007;

Montgomery et al. 2003). Levels of child mortality in the developing world have declined

dramatically over the past few decades; however, great variation across geographic and

population groups persist.

Early studies of infant and child mortality hypothesized that the urban advantage

was a result of differences in infrastructure and environment, particularly health facilities

and medical care (Johnson 1964; Puffer and Serrano 1973). Later studies downplayed the

significance of environmental factors, and instead focused on social status differences

between urban and rural areas (Trussell and Samuel H. Preston 1982; Trussell and

Hammerslough 1983; Behm and Primante 1977). Maternal education has also received

much emphasis in the literature (Cleland et al. 1992; Mensch et al. 1985; Sastry 1997).

Several studies have examined the effects of place of residence on infant and child

mortality. Often, the magnitude of urban-rural differences declines and significance of

differences disappears when other covariates such as family background are held constant

(Behm and Primante 1977; Caldwell 1979; Caldwell and P. F. McDonald 1982; Trussell and

Samuel H. Preston 1982; Mensch et al. 1985; Martin et al. 1983; Casterline, Cooksey, and

Ismail 1989; Hobcraft et al. 1984a). Some studies suggest that place of residence

differentials are simply manifestations of underlying differences in individual and family

socioeconomic characteristics (Behm and Vallin 1982; Gilbert and Gugler 1992). Other

studies find that urban-rural differences in child mortality remain, even after holding

constant the effects of family and individual background (Adlakha and Suchindran 1985;

Brockerhoff 1990; Caldwell and P. F. McDonald 1982; Michael R. Haines and Avery 1982;

64

Hobcraft et al. 1984a; Mensch et al. 1985; Van de Poel et al. 2007; Sastry 1997), and often

conclude that infrastructure and environmental factors again deserve more of a

consideration in the literature.

A great flaw in the literature on place and mortality is the aggregation into urban

and rural categories. In particular, the massive growth of the poorest segments of cities, in

particular slums, further complicates the question of the urban advantage. As the

population of Indian cities consists of an increasing share of urban poor, the idea of an

urban advantage becomes murky. For example, slum residents may be closer to health care

than their rural counterparts, but just as unlikely to be able to afford it. Additionally, the

very poor and unsanitary living conditions that characterize so many slums may have dire

consequences for its residents. Residents of a rural home without a toilet may urinate and

defecate outdoors, and have no poor health consequences as this is away from their home

and living space. A slum family without a toilet, in contrast, will have to defecate

somewhere in the tight residence of the slum, leading to much worse health consequences.

These poor living conditions would have particularly dire consequences for the most

vulnerable populations of infants and children. It is unclear theoretically if slum residents

experience an urban advantage or perhaps an urban penalty when it comes to infant and

child mortality.

No studies to date have examined infant or child mortality differences between slum

residents and non-slum residents. This chapter aims to fill that gap by answering two

questions. First, do differences in infant and child mortality exist between slum residence

and non-slum residents? Second, if such differences do exist, to what extent are they the

65

result of underlying differences in the distribution of individual and family characteristics,

such as wealth and education?

DATA AND METHODS

Data

For this study I use data from the National Family Health Survey round 3, conducted

during 2005 and 2006. All women ages 15-49 are asked many retrospective questions

about their pregnancies and births. Each woman is asked about all of her births, including

the month and year of the birth, and if the child died the month and year of the child’s

death.

I limit my analyses to births in the five years leading up to the interview. This is

important for several reasons. First, the covariates available are only measured at the time

of the interview, and are not asked retrospectively. By limiting the period to the five years

prior to the interview, we should limit (though not eliminate) any measurement error

incurred by conditions changing in the past five years. This is particularly important for

living conditions and slum residence, as these are factors that could easily change. The

smaller the time interval we can use, the more likely the covariates accurately reflect

conditions of the child’s first years of life. Additionally, limiting the recall period should also

reduce measurement error due to a woman’s poor recall of specific months of birth and

death.

66

I also limit my analyses to those who have not moved since the child’s birth. This

allows us to eliminate possible biases or errors introduced by people moving between rural

areas, slums, and non-slum urban areas.

Methods

Several methods are used to model infant and child death. First, Kaplan-Meier

univariate models are used to estimate survival curves for rural, slum, and non-slum births.

These models can show us if there are indeed significant differences in child and infant

survival across places.

Next, I use discrete time logit models to predict both infant and child death. These

models allow for a discrete measurement of time. The NFHS data records the month of

death of a child. Ideally, we would have the exact day of death, but instead we have only

months. In this case, the discrete time model is an appropriate choice because the discrete

month categories are being used to approximate the underlying continuous time. A discrete

time model is preferable to standard proportional hazards models because the underlying

assumptions in proportional hazard models are very often violated. Lastly, the discrete

time logit allows a researcher to model the baseline hazard in a variety of forms, including

exponential, Gompertz, Weibull, and piecewise forms. I model both infant and child death

using several different baseline hazard specifications and also compare them to a Cox

proportional hazard model to see how robust the findings are.

To determine how much differences in death are due to location as opposed to

unevenly distributed family and individual characteristics, I include a set of variables that

the literature has shown to be predictive of infant and child death area (Van de Poel et al.

67

2007; Sastry 1997; Smith, Ruel, and Ndiaye 2005). Lower survival (or higher hazard of

death) is commonly associated with mothers having little education, a young age at birth, a

short preceding birth interval (commonly defined as less than 24 months), higher birth

order (meaning being one of the younger siblings in a family), and few family resources.

Additionally, those of the formerly untouchable castes (known as being a member of a

scheduled caste, scheduled tribe, or other backward caste) have long been disadvantaged.

Despite the slow dismantling of the caste system, these castes remain very disadvantaged

compared to non-backwards castes (Newman and Deshpande 2007; Newman and Jodhka

2007; Newman and Sukhadeo Thorat 2007; Radhakrishna and Ray 2005). Muslims are also

disadvantaged compared to Hindus.

By including this set of predictor variables, I can determine what proportion of the

disparities in child survival are due to individual and family characteristics that are

unequally distributed among rural, slum, and non-slum urban households. For instance,

infant death may be a problem of family poverty, not concentrated poverty in slums. If this

is the case, then by including a household’s wealth, we should eliminate the association

between slum residence and infant death.

Lastly, it should be noted that the sample used to create the data was a clustered

sample. Ideally, one would use hierarchical or fixed effects models to take examine if there

are any discernable neighborhood effects. Unfortunately, my data do not allow for this, as

the Indian government will not release the clustered information. Thus, in none of my

models do I correct for clustering. This does not affect or bias the actual coefficients in any

of the models. It does, however, artificially reduce the standard errors. Typical corrections

68

for clustering account for this by inflating standard errors so as not to bias significance

tests. Instead of correcting in the models, however, I simply note the clustering exists and

use a higher standard for considering an effect significant. Rather than considering p<0.05

a statistically significant effect, I consider only those results where p<0.01. To examine

how coefficients may vary across place of residence, I ran interactions between place and

all of the independent variables. The results of these models are presented in Appendix A.

Unless specifically discussed, the results do not change the substantive conclusions found

in the more parsimonious models.

RESULTS

Table 4-1 shows descriptive statistics on all births to women ages 15-49. There is a

much larger sample size of rural births than urban births because the sample of urban

residents has been restricted to cities where data on slum residence was collected. Still, at

15,566 slum births and 20,954 non-slum births, there is a sufficiently large sample to

conduct analyses.

The descriptive statistics show that births in rural areas are in households with a

lower mean wealth, here a log of the factor score traditionally used in DHS analysis (see

chapter 2 for the fuller discussion on the DHS wealth factor score). Rural mothers also have

a distribution of education falling lower than all urban residents, and slum mothers have

less education than non-slum urban mothers. Maternal age at birth is roughly equal for the

three groups. Between 10 and 13 percent of households in each of the areas are headed by

a female, either due to death of a husband, migration of a husband or wife (usually for

work), or divorce. The higher mean birth order of rural homes reflects that rural homes

have higher fertility than urban homes, and slums higher than non-slums. (This is

69

consistent with (Weeks 2008) who estimates slum TFRs in Ghana are higher than non-

slum TFRs.) Lastly, we can see that the scheduled castes and tribes and other backward

castes are overrepresented in rural and slum areas.

Figure 4-1 shows graphically Kaplan-Meier estimates of survival to 13 months. A

disproportionate number of mothers reported having children die in month 12, a

phenomenon typically known in demography as “age heaping,” whereby people round to

the nearest landmark age. To account for this, I calculated survival until 13 months, so as

not to cut the distribution around a modal point. The figure shows that urban non-slum

children have the highest survival to 13 months, followed by slum children, then rural. The

95% confidence intervals indicate that past the first month, these are real differences in

survival. The figure in the graph shows that urban non-slum babies have the highest

probability of surviving, followed by slum babies and finally rural babies. The 95%

confidence intervals indicate whether these are significant differences.

Table 4-2 shows results from the discrete time logit model of infant death. In these

results a Gompertz baseline hazard was specified; however, models specified using

exponential, Weibull, and piecewise (by month) baseline hazards had nearly identical

numeric results, thus not affecting the substantive results at all. Model 1 reflects the same

results as in the Kaplan-Meier graphs: there are sizeable differences in survival between

rural, urban slum, and urban non-slum areas. Rural babies are the most likely to die,

followed by slum babies, then non-slum urban babies.

In the second and third models of Table 4-2, individual and family characteristics

predictive of infant death in the literature are included. The biggest finding of note here is

70

that the effects of place of residence disappear once background characteristics are taken

into account, indicating that the raw differences in mortality between slums, other urban

areas, and rural areas are actually a function of individual and family characteristics such

as wealth being unevenly distributed across place. The background characteristics have

effects that are significant in the expected direction: wealth, higher maternal education, and

maternal age are protective characteristics. A short preceding birth interval, a higher birth

order, and being of a scheduled caste or tribe are all related to higher chances of infant

death.

In model 3 of Table 4-2 I include housing characteristics that are sometimes used to

define slums. The only characteristic that has any effect on infant death is an infant living in

an overcrowded home. An overcrowded home has a protective effect on infants. One

explanation for this finding is that overcrowded homes may be joint families, where a child

is living not only with his or her parents, but also some combination of grandparents, aunts

and uncles, and cousins. This may prove protective of infant health because of the larger

number of adults who can provide care for an infant. This hypothesis would need more

direct testing to be supported or refuted.

Figure 4-2 shows graphically the results from Kaplan-Meier estimates of survival to

age 5. Again, it is clear that urban non-slum children have the highest probability of

survival, followed by slum children, and finally rural children. Table 4-3 shows results from

the dicrete time logit models, again specified with a Gompertz baseline hazard4. Model 1

4 Again, these models were run using exponential, Weibull, and piecewise baseline hazards. The models were

nearly identical despite the baseline specification.

71

reports the univariate results as in the Kaplan-Meier model: rural children face higher odds

of dying by age 5 as compared to slum children, while urban non-slum children face lower

odds of dying.

Model 2 includes the set of predictive covariates. Again the effects of residence

disappear once the effects of family background are held constant. The effects of family

background are again consistent with the literature and the models on infant mortality.

Higher odds of child death are associated with less wealth, less maternal education,

younger mothers, a short preceding birth interval, a higher birth order, and being of a

scheduled caste or tribe. In Model 3 once again the housing characteristics typical of slums

are included. Again, the only significant predictor is that children in crowded homes are

more likely to survive. The same hypothesis relating to infant death is plausible here:

children living in crowded homes are likely to be living in extended families with more

adult supervision, attention, and support.

Overall, both the infant and child mortality models show the same results. There are

significant differences in the raw means: Urban non-slum children are the least likely to

die, followed by slum children, and lastly rural children. These effects disappear, however,

once the child’s family background and individual characteristics are taken into account.

This suggests that location differences infant and child mortality are not the result of

anything specific to the location, but rather reflect underlying differences in family

socioeconomic status, family background, maternal characteristics, and the child’s place in

terms of birth order and timing of births.

CONCLUSIONS

72

My results show that Indian babies and children living in cities on average have

higher chances of survival. Further, non-slum children have higher chances of survival than

children living in slums. The key word here, however, is “on average”. Once these

categories grouped by residence are further disaggregated by socioeconomic status,

maternal education, family structure, and family background, the effects of place of

residence disappear. In effect, these results indicate that where a child is born and lives

does not matter; what matters is the wealth of his or household, the education and age of

his or her mother, the caste of the family, and the child’s position in the family. These

characteristics are not evenly distributed across place of residence. As we saw in the

descriptive statistics, non-slum urban households have higher wealth, education, and caste.

They also have longer birth intervals and smaller families. Each of these is associated with

higher chances of infant and child survival, and as such urban non-slum children

experience the best survival chances.

While many policymakers focus on the aggregate differences across groups, these

results make clear that it is not residence per say that affects health. This is significant in

terms of directing resources. These results suggest, for instance, that putting money toward

improving women’s education may go farther in reducing infant and child mortality in

rural areas than building more rural health clinics aimed at children. While we know access

to health care is important for improving a population’s health, there are other less obvious

factors that can affect mortality. While in this paper and with these data we are not able to

directly test the effect of maternal education as compared to access to a local health clinic,

this paper does suggest the importance of focusing on improving maternal education, caste

disadvantage, and wealth disparities in the quest to reduce infant and child mortality.

73

As with any study that may contain a large number of migrants, questions of

selection are relevant in this discussion. If the most robust, healthy, and ambitious of rural

residents are the ones who migrate to cities, then we might expect that slums full of these

migrants would be more robust than average and may contend with disadvantage better.

Given that slums are often the first stopping place of rural migrants in cities, we might

expect that a large portion of slums are in fact migrants.

Selection Questions

The analyses presented are of course subject to questions of selection. To fully

examine and answer the question of if migrants are selected along health-related aspects,

rigorous work on both sending and receiving locations is necessary. This data are not

available for this study, and so these questions cannot be completely answered. However,

based on several aspects of my data and the literature I contend that selection questions do

not invalidate the results. First, it is a misperception that migration alone causes urban and

slum growth. In fact, only half of urban growth is due to migration; the other half is due to

natural increase (or fertility), often in the poorer segments of the population. Slums have

existed in Indian cities for decades, and in fact many official state and federal Slum Acts

were passed in the 1970s. Thus, it would be mistaken to assume that migration and related

selection questions are impacting the entirety of slums and slum growth. The data on

women in the National Family Health Survey includes a measure of how long a woman has

lived in her current place of residence (be it a city or the countryside). Nearly half of

women at childbearing ages (46.9% to be exact) report having lived in their current place

of residence for their entire lives. Of those who have moved from another location, 45%

last moved from another city residence. These proportions are nearly identical to the same

74

number for urban non-slum residents. This means a relatively small proportion of the

population of slum women are actually migrants, even given relatively high rates of

urbanization. Second, if selection is in fact happening from rural areas, one would expect

rural areas (sending locations) to have worse outcomes, on average, as the healthiest and

best are leaving. In the multivariate models, however, we see no differences between rural

residents and their urban counterparts. Given all of this information, it would be mistaken

to think that the lack of an effect on mortality in slums is solely due to healthy migrants.

The more complicated picture

Given my results, one interpretation is that residence does not matter. This relies on

a very simplistic view of how residence affects people, however. The much more nuanced

story is that incomes, education, and other markers of socioeconomic status are not truly

independent of residence. In fact, it is likely city residence itself that in various ways causes

people to have higher income and education, among other things. Given this, it is perhaps

more realistic to conceive of a much more complicated kind of path model, where cities

produce both individual socioeconomic characteristics, and these characteristics affect

health, just as cities themselves might affect health. This kind of path model may be quite

difficult methodologically, as it would be akin to something like a structural equation

survival model. Still, future research should consider how to best take into account the

multilayered ways cities are impacting the lives of residents.

75

TABLES AND FIGURES

Table 4-1: Descriptives statistics for births by place of residence, National Family

Health Survey 2005-2006

Rural Urban slum Non-slum

urban

Total births 154801 15566 20954

Deaths to age 1 13205 969 1106

Deaths to age 5 16408 1161 1280

Wealth factor 11.7 (0.55) 12.3 (0.35) 12.5(0.31)

Mother’s education

None 61.5% 44.1 28.2

Primary 16.4% 17.7 13.1

Secondary 20.6 34.2 41.4

Higher 1.4 1.0 17.4

Maternal age at birth 22.7 (5.23) 22.1 (4.6) 22.8 (4.5)

Female headed home 12.1% 13.0% 10.3%

Preceding birth interval <

24 months

27.3% 28.7 24.2

Birth order 2.68 (1.8) 2.47 (1.6) 2.20 (1.4)

Caste

Scheduled caste or

tribe

37.1% 27.0% 17.7%

Other backward caste 33.6 28.5 32.1

All other castes 29.3 44.5 50.7

Religion

Hindu 73.4% 65.6% 72.8%

Muslim 12.6 28.7 20.4

Christian 8.7 2.1 1.9

Other 5.3 3.6 4.9

76

Table 4-2: Results from discrete time logit models predicting infant death, National

Family Health Survey 2005-2006

Model 1 Model 2 Model 3

Coefficient SE Coefficient SE Coefficient SE

Residence

Urban non-slum –0.31 ** -0.10 –0.03 -0.10 –0.04 -0.10

Rural 0.49 *** -0.07 –0.03 -0.08 –0.17 -0.08

[Reference = urban slum]

Family background

Wealth –0.44 *** -0.04 –0.46 *** -0.04

Mother has no education 0.16 ** -0.06 0.16 ** -0.06

Mother has secondary education –0.44 *** -0.07 –0.43 *** -0.07

Mother has higher education –0.99 *** -0.26 –1.03 *** -0.26

Female headed home 0.17 ** 0.06 0.11 0.06

Mother’s age at birth –0.08 *** -0.01 –0.08 *** -0.01

Short preceding birth interval 0.52 *** -0.04 0.52 *** -0.04

Female child 0.02 -0.04 0.03 -0.04

Birth order 0.10 *** -0.02 0.10 *** -0.02

Scheduled Caste or tribe 0.14 -0.06 0.14 -0.06

Other backward castes 0.04 -0.06 0.04 -0.06

Muslim –0.29 *** -0.08 –0.27 *** -0.08

Christian –0.08 -0.15 –0.10 -0.15

Other religion –0.11 -0.12 –0.10 -0.12

Housing characteristics

Poor toilet 0.13 -0.10

Poor water 0.11 -0.07

Overcrowded home –0.34 *** -0.05

Poor floor –0.01 -0.06

Poor roof 0.00 -0.05

Poor walls –0.02 -0.06

N 1194244 1194244 1194244

Pseudo R2 0.01 0.04 0.04

*** p<0.001; ** p<0.01

77

Table 4-3: Results from discrete time logit models predicting child death to age 5,

National Family Health Survey 2005-2006

Model 1 Model 2 Model 3

Coefficient SE Coefficient SE Coefficient SE

Residence

Urban non-slum -0.28 *** -0.08 0.05 -0.08 0.04 -0.08

Rural 0.58 *** -0.06 -0.04 -0.06 -0.19 * -0.06

[Reference = urban slum]

Family background

Wealth -- -0.53 *** -0.04 -0.59 *** -0.04

Mother has no education -- 0.25 *** -0.05 0.24 *** -0.05

Mother has secondary education -- -0.50 *** -0.06 -0.50 *** -0.06

Mother has higher education -- -1.09 *** -0.23 -1.12 *** -0.23

Female headed home -- 0.15 ** -0.05 0.08 -0.05

Mother’s age at birth -- -0.06 *** -0.01 -0.06 *** -0.01

Short preceding birth interval -- 0.50 *** -0.04 0.50 *** -0.04

Female child -- 0.05 -0.03 0.05 -0.03

Birth order -- 0.08 *** -0.02 0.09 *** -0.02

Scheduled Caste or tribe -- 0.22 *** -0.05 0.22 *** -0.05

Other backward castes -- 0.06 -0.05 0.06 -0.05

Muslim -- -0.14 -0.06 -0.12 -0.06

Christian -- -0.17 -0.13 -0.18 -0.13

Other religion -- -0.05 -0.10 -0.03 -0.10

Housing characteristics

Poor toilet -- -- 0.19 -0.08

Poor water -- -- 0.04 -0.05

Overcrowded home -- -- -0.36 *** -0.04

Poor floor -- -- -0.01 -0.05

Poor roof -- -- 0.01 0.01

Poor walls -- -- -0.06 -0.05

N 5026912 5026912 5026912

Pseudo R2 0.05 0.07 0.08

*** p<0.001; ** p<0.01

78

Figure 4-1: Kaplan-Meier univariate estimates of survival through month 13 by

place, NFHS 2005-2006 (N=191321)

79

Figure 4-2: Kaplan-Meier univariate estimtes of survival to month 60 by place, NFHS

2005-2006 (N=191321)

80

Chapter 5 : Child Health

By improving the lives of slum dwellers, we are also combating malnutrition and

diseases, many of which are directly linked to overcrowding and to the lack of clean

water and improved sanitation.

Anna Tibaijuka, Under-Secretary-General of the United Nations,

Executive Director of UN-HABITAT

Children are a highly vulnerable subpopulation. Countless governments and non-

governmental organizations around the world concern themselves with the state of

children’s health and how to improve child health. Child health has been shown to have an

impact not only directly on children’s well-being and educational outcomes, but also to

have effects on adult morbidity, with poor childhood health causing higher morbidity in

adulthood (Blackwell, Hayward, and Crimmins 2001; Brunner et al. 1996; Kuh and

Wadsworth 1993; Power and Peckham 1990; Buck and Simpson 1982; Costa 2000; Barker

1998).

On average, in developing countries children’s health outcomes are better in urban

areas than rural areas. However, understanding the nature and causes of these disparities

is key to projecting the impacts of rapid urbanization on children’s health, and thus it is

also key to targeting resources accordingly. Comparison of simple urban and rural means is

not adequate for understanding the complexity of children’s health predictors and how

location of residence interacts with these to produce outcomes. In particular, the rapid

growth of slums begs the question of how slum residence is related to children’s health. As

a particularly vulnerable population, children may be especially at risk to illness due to the

poor sanitation and living conditions characteristic of many slums.

81

In this chapter, I examine five separate measures of child health status for children

under five. Three are acute symptoms: coughing, fever, and diarrhea. Two are measures of

malnutrition: short term malnutrition called wasting, or low weight for height, and chronic

malnutrition called stunting, or low height for age. Results indicate that the measures of the

acute symptoms are unreliable due to the fact they are maternal reports, and maternal

reports may vary considerably by socioeconomic status, and in particular maternal

education. Although at the bivariate level stunting and wasting appear to have

relationships with location, these relationships disappear once a broad set of individual,

family, and housing characteristics are taken into account. I also look at health care seeking,

and if seeking medical care is related to residence location.

CHILDREN’S HEALTH

The literature on children’s health in the developing world is expansive and spread

across the fields of sociology, medicine, public health, and epidemiology. Although it is

generally acknowledged that urban children are on the whole healthier than rural children,

relatively few studies have rigorously examined urban-rural disparities in child health

outcomes. Even fewer studies have examined slum residence. In this section I review the

relevant literature on child health outcomes in the developing world.

The bulk of the literature on the urban advantage in children’s health focuses on

child nutrition. Overall, the literature shows that urban children are better nourished than

rural children, being less likely to have stunted growth (a mark of chronic

malnourishment) and to be severely underweight, known as wasting (Fotso 2006, 2007;

Menon, Ruel, and Morris 2000; D. E. Sahn and Stifel 2003; Smith et al. 2005; Von Braun et

al. 1993; Van de Poel et al. 2007; Montgomery et al. 2003). A few studies have looked

82

specifically at slums, mostly in Bangladesh. Both stunting and wasting rates are very high in

the slums of Bangladesh, and have not improved over time as rapidly as in poor rural areas

(Bloem 2003; Buttenheim 2008).

Several studies over the past ten years have used Demographic and Health Surveys

data and UNICEF data to look at rural-urban disparities in child nutrition across developing

countries. Several studies examine stunting, defined as a child falling two standard

deviations below the mean of the World Health Organization’s international standard

height-for-age distribution (Van de Poel et al. 2007; Fotso 2006; Menon et al. 2000; Smith

et al. 2005; D. E. Sahn and Stifel 2003). Each of the studies found mean or median rural-

urban odds ratios or relative risks in stunting between 1.4 and 2.2, indicating that rural

children have odds of stunting between one and half and two times that of urban children.

Van de Poel et al. (2007) find that controlling for wealth and other socioeconomic

characteristics make these disparities shrink or, in about half of the 47 countries studied,

disappear entirely. Fotso (2006) and Menon et al. (2000) also examined socioeconomic

inequality and rural-urban differentials. Both found that disparities in stunting between

urban children in the poorest and richest quintiles were larger than the disparities

between the rural children of the poorest and richest quintiles.

Wasting, or low weight-for-height, is considered a marker of acute malnourishment,

such as that which may accompany a long spell of sickness or recent diarrhea (Buttenheim

2008). Wasting is less common than stunting because of its acute nature. It is more

prevalent in rural areas than urban areas (Von Braun et al. 1993), although the living

environment and poor sanitation in slums is thought to cause a high prevalence of

83

diarrheal episodes that can lead to wasting. Looking solely at Bangladeshi slums,

Buttenheim (2008) finds that living in slums with better sanitation predicts lower child

wasting.

Some research has been done on acute child health conditions in both the social

science and medical literature. Research on children’s fever has been the most extensive,

particularly in sub-Saharan Africa where a fever is cause for concern because it may be an

indication of malaria. Several researchers have used the Demographic and Health Surveys

data (of which the National Family Health Survey is the Indian component) to examine the

prevalence of fever and socioeconomic predictors of fever (Woldemicael 2001; Yohannes,

Streatfield, and Bost 1992; Ryland and Raggers 1998; Kandala and Madise 2004; Filmer

2005). Across many sub-Saharan African countries, Filmer (2005) found that fever

incidence was not strongly related to family wealth, with incidence dropping only at the top

of the wealth distribution. Unfortunately, the indicator of family wealth in Africa is

relatively unreliable in distinguishing between the poorest three income quintiles

(Chaudhury et al. 2009), and therefore these results should be interpreted with caution.

Additionally, the debate on whether caregivers can accurately recognize fever is

ongoing, although the bulk of studies in the medical literature show that of caregivers who

claim their child has a fever, only 15-20% are correct (Einterz and Bates 1997, 1998;

Morgan et al. 1997; Kofoed et al. 1998; Dunyo, Koram, and Nkrumah 1997). Unfortunately,

this research has not yet been fully recognized in the social science literature, and most

studies using survey data use maternal reports at face value to calculate prevalence and

incidence of fever.

84

Diarrhea has also been the topic of considerable study. Ryland and Raggers (1998)

looked across 34 countries using Demographic and Health Surveys data, finding diarrhea

prevalence ranged from a low of 8% to a high of 28%, with an average prevalence of 16%.

On average in the 24 countries, rural children’s diarrhea prevalence was 13% higher than

urban children’s, although in 9 countries urban children had a higher prevalence. Ruel et al.

(1999) found that the prevalence of diarrhea among urban children of low socioeconomic

status was greater than the prevalence among rural children of low socioeconomic status in

7 of 11 countries studied. Other socioeconomic factors have been shown to be related to

maternal reports of children’s diarrhea. Mothers’ having a secondary or higher education

often (although not always) is related to lower reports of diarrhea, whereas household

wealth has not been found to have a relationship with diarrhea5.

Much of the public discourse on slums in India focuses on the poor sanitation and

living conditions that could breed disease and illness among slum residents. Diarrhea can

easily be caused by poor sanitation, and fever and coughs are symptoms that can be

indicative of viruses or bacterial conditions spread through bad water, poor sanitation, and

cramped living conditions. One might expect, then, that slum children experience higher

prevalence of these symptoms as compared to other urban children and rural children,

given the poor living conditions of slums. In rural areas, for example, sanitation may often

be better than in slums. Though rural areas rarely have sewers, defecation can take place in

the bush, remote to activities of daily life and therefore causing much less of a health

hazard than lack of sewers in densely populated urban slums. The epidemiology literature

5 Again, this is most likely due to the poor measurement of wealth into quintiles.

85

has examined environmental conditions such as toilet facilities and water supplies and

their impacts on diarrhea (Ryland and Raggers 1998; Buttenheim 2008). Several

researchers have posited that slums in particular have significant disease hazards due to

inadequate sanitation (Ahmed 2005; Hanchett et al. 2003), hazards that are likely to have

the impact of raising the prevalence of diarrhea among children (Buttenheim 2008).

Lastly of acute health conditions, children’s coughing has received some attention.

Cross-country data using maternal reports of children having a cough show prevalence

ranges from 7% to 32% (Ryland and Raggers 1998). Children of young mothers are at a

higher risk, whereas children of more educated mothers and in urban areas have lower

prevalence. Oddly, children in households that own a radio (a crude mark of wealth) have

higher maternal-reported prevalence (Ryland and Raggers 1998).

Overall, a picture emerges of urban children generally faring better than their rural

counterparts in both acute conditions as well as long-term and short-term nutritional

status. Little research has examined the impact of slums on these health conditions,

although the public and academic discourse on slums often emphasizes the unsanitary

living conditions that may negatively affect children’s health.

HEALTH SEEKING BEHAVIOR

In addition to simply looking at predictors of child health, another subject of interest

is who seeks medical care. In relations to slums, the question is how seeking medical care

may be different among those living in rural, urban slum, and urban non-slum areas. A

large literature examines predictors of seeking medical care in the United States and

around the world. The literature relevant to this paper is solely on the developing world

86

and children’s health. Cross country analyses using DHS data (Ryland and Raggers 1998)

show that of children reported to have diarrhea, there is wide variation in the proportion

taken to a medical facility for treatment, ranging from 10% in Niger up to 68% in Namibia,

with an average of 34%. The prevalence of treatment varied by family and individual

background characteristics. Urban children are overall 47% more likely to be taken for

medical care. Children of a higher birth order (such as the third born, fourth born, and later

children) are more likely to be taken for care, as are children of mothers with more

education.

Treatment for fever has also been examined using Demographic and Health Surveys

data. Filmer’s (2005) study of sub-Saharan African countries finds that children in the

richest quintiles are most likely to be taken to a doctor for a fever. Across all DHS countries

in the second and third rounds of DHS studies (Ryland and Raggers 1998), on average 43%

of children reported to have had a fever were taken to a medical facility. The proportion

tends to decline with increasing birth order, whereas it is higher for children in urban and

areas and with more educated mothers.

Models predicting treatment of a cough are similar to the other acute symptoms

Ryland and Raggers (1998) found that across countries, 44% of children with a cough were

taken to a medical facility for care. Male children were more likely to be taken for care, as

are urban children, children of mothers with increasing education, and those in households

with more wealth. None of these studies control for the distribution of health care facilities.

Specific to India, Pillai et al. (2003) found that medical care was more often sought when a

87

child had a more severe illness, when the family was of higher economic status, when the

family was rural, and if the mother had less education.

In a study of Delhi, India, Das and Hammer (2007) find that on average, there are 70

medical providers within a fifteen minute walk of any residence in Delhi6. The authors also

found that the poor reported on average a third more visits to a medical provider than the

rich during the study period of two years. Despite this, there is considerable inequality in

access to quality care, as doctors serving those living in poor neighborhoods are overall

less competent than doctors in middle-income neighborhoods and high-income

neighborhoods (Jishnu Das and Hammer 2007). In the case of slums, then, we may expect

that people are visiting less competent practitioners, though perhaps many living in slums

who are poor are visiting practitioners in proportions on par or even higher than their

richer counterparts.

Overall, the literature on seeking health care shows that urban residents are more

likely to seek care than rural residents. In urban areas, poor residents may even be more

likely to seek care, although the care residents of poor neighborhoods seek is often of very

low quality. In terms of slums, we may expect slum residents to visit providers at rates

similar to those of non-slum residents.

The Research Gaps

This paper first aims to quantify child health disparities across slums, other urban

areas, and rural areas in India. Public dialogue and even some research on slums works

6 The authors defined medical providers as all health care practitioners visited by survey respondents over a

two year period as well as all providers within a 15 minute walking radius of every survey respondent. These

include both private and public providers.

88

under the assumption that slum residents are at a higher risk for disease. While this is

based on the idea that slums have poor sanitary conditions, in fact slums may be a diverse

set of neighborhoods that do not have one overall impact on health. It is important for

rigorous research to examine child health outcomes across places of residence in order for

practitioners and researchers alike to be informed of the true disparities and focus debates

on reducing these disparities accordingly. In addition, it is likewise important to look at

disparities in who seeks medical care by location in order to determine if certain people or

types of neighborhoods have less access to care.

To answer questions of how child health and health seeking behavior differ by place

of residence, I use NFHS data to examine the chances of a child having a fever, cough, or

diarrhea in the two weeks preceding the survey. Each of these three symptoms can be a

result of infectious disease, perhaps more easily spread by poor living conditions

characteristic of many slums. I next examine both acute and long-term malnutrition by

looking at children’s weight-for-height and height-for-age compared to international

standards, trying to determine if slums have any impact on these measures of child well-

being. Health-seeking behavior may also vary across residence location. Accordingly, of

children whose parents reported them to have one of the acute symptoms, I look at who

seeks medical care for their children. I first look at if these outcomes vary by residence

location, and then how much the uneven distribution of individual and family

characteristics explains any variation by residence.

DATA AND METHODS

I use data from the National Family Health Survey Wave 3 (NFHS-3) birth files for

these analyses. A sample of women ages 15-49 are asked many retrospective questions

89

about their pregnancies and births. Each woman is asked about all of her births, including

the month and year of the birth. I limit my analyses to births in the five years leading up to

the interview, consistent with the sample of births used in the infant and child mortality

analyses. Children under five are a highly vulnerable subpopulation, and morbidity at these

ages can have lasting effects on children’s physical and cognitive development as well as

later life health (e.g. Blackwell et al. 2001; Hack et al. 1994; Thomas, Strauss, and Henriques

1990).

For all living children under the age of five, mothers were asked questions about the

child’s recent health. They were asked if the child had, in the two weeks preceding the

interview, had a fever, a cough, or diarrhea. Ideally, there would be an objective measure of

these symptoms rather than a maternal report. However, such objective data are very

costly and time consuming to collect, and so are not included in large scale national surveys

such as the National Family Health Survey. Thus, self-reports or maternal-reports of health

conditions is the best alternative, and the data available for this study.

Additionally, all children had their height measured and were weighed, and two

measures of malnourishment were constructed. The first is a binary measure of chronic

malnourishment: stunting. Measurement of stunted growth is based on the World Health

Organization’s (WHO) child growth standards (World Health Organization 2004; 2006)

that were created based on a multi-country longitudinal study from 1997 through 2003. A

child is considered to have stunted growth if his or her height falls more than two standard

deviations below the WHO’s international standard mean. The second measure is a binary

measurement of short-term malnourishment, called wasting. A child is considered wasting

90

if his or her weight-for-height falls two standard deviations below the WHO’s international

standard mean.

Beyond looking at simple disparities between rural, slum, and other urban

residents, I also present disparities that remain after taking into account differences in

household wealth, parents’ education, mother’s age at birth, the sex of the child, birth

order, whether a family is of a scheduled caste or tribe or other backwards class, and the

religion of the family. These are factors that have been shown to predict health outcomes of

children in the literature (Adlakha and Suchindran 1985; Buttenheim 2008; Filmer 2005;

Ryland and Raggers 1998; van Poppel and van der Heijden 1997). I use logistic regression

to look at how these children’s health measures differ by place of residence, and how

individual and family characteristics predict poor health outcomes. For all of the models, I

control for month of the year the interview was taken, as seasonal variation in sickness is

considerable and needs to be accounted for (Ryland and Raggers 1998).

For those children whose mothers reported them having a cough, fever, or diarrhea

in the two weeks preceding the survey, the mothers were asked if the children were taken

for any treatment. I examine differences in treatment first by place of residence alone, and

then again including the set of individual and family characteristics shown to be predictive

of health seeking behavior (Filmer 2005; Ryland and Raggers 1998). The interviewers did

ask mothers what kind of health care provider they took their child to, such as public

hospitals, private hospitals, rural clinics, pharmacies, public and private dispensaries,

traditional healers, homeopathic doctors, and so on. There are considerable differences in

provider competence across the kind of provider, and so it is tempting to examine what

91

kind of provider a child was taken to. Unfortunately, there is substantial heterogeneity in

provider competence even within the best providers, public and private hospitals. Das and

Hammer (2007) show that provider competence at public and private hospitals varies by

the income and wealth level of the neighborhood a hospital is in, with hospitals in rich

areas being staffed by far more competent providers than hospitals in poor neighborhoods.

Thus, simply looking at what kind of provider a child was taken to is not a good indication if

a child received good care, and therefore not a useful direction to go.

RESULTS

Table 5-1 shows descriptively the differences by location in acute child health

conditions and health care seeking for those with medical conditions with no controls for

family background, individual characteristics, or survey date. Though there are some

differences, they are not large in magnitude. Chi-square tests indicate that there are

significant differences in cough, fever, and stunting, though not in diarrhea (as indicated by

the low chi-square) Slum children experience the highest prevalence of coughing at 18.7%,

followed by rural children (at 17.6%) and other urban children (at 15.3%). There are no

significant differences in the prevalence of diarrhea by place of residence. Rural children

experience the highest prevalence of having a fever, followed by slum children and then

other urban children. Rural children also experience the highest level of stunting, with

15.8% of rural children stunted, as compared to slum and other urban children at 12%.

The bivariate results on health care seeking show considerable differences across

urban and rural residence. Of urban residents, slum and non-slum, 77-82% of children with

a cough or fever were taken to a health care provider, compared to only 60% of rural

residents. For diarrhea, 73% of urban children were taken to a health care provider,

92

compared to only 58% of rural residents. Both these distributions are significantly

different than one would expect, as indicated by the large chi-square values.

Table 5-2 shows results from logistic regressions of a child coughing, having a fever,

or diarrhea on residence and the set of family background and housing characteristics. For

only one of the acute conditions is any residence effect significant in the first set of models.

This would appear to contradict the findings in Table 5-1, which showed significant

differences by place on both coughing and fever. The logistic regressions models control for

the month of the year the survey was done, as health conditions such as a cough and fever

are more likely during colder months of the year. It appears that this alone accounts for the

differences in reported prevalence of fever by residence location. This is an important

finding to note, as it is common practice to report basic descriptive statistics on health

conditions not taking into account date of the survey in both the official NFHS reports and

many research studies using DHS data.

The one significant effect of residence (after taking into account survey month) is

the effect of living in a non-slum urban area on coughing. Those in non-slum urban areas

are less likely to experience a cough than both rural and slum residents, who are

indistinguishable from one another. However, his effect disappears entirely once family

and background characteristics are included in the model. Several of the family background

and housing characteristics have effects that are consistent across models. Having poor

toilet facilities predicts higher chances of fever, cough, and diarrhea in children, and having

poor walls predicts coughing and diarrhea. Muslim children are more likely to have all

three of the conditions, indicating some level of disadvantage that is above and beyond the

93

socioeconomic disadvantage of the Muslim population. Female children are less likely to

experience any of the conditions.

Perhaps the most puzzling coefficients in the models are those showing the effect of

maternal education on children’s health conditions. A mother having no education predicts

less of a chance of a child coughing or having a fever. This runs contrary to nearly all the

literature on the impacts of maternal education, and will be discussed in more depth in the

discussion section of the paper. In addition to this puzzling finding, many of the results

reported in these models are inconsistent across models or surprising. Being of a scheduled

caste or tribe predicts less of a chance of coughing, but nothing else, and being of another

backward class predicts higher chances of diarrhea, but nothing else. Children in Christian

families are less likely to experience diarrhea, but nothing else. While these three

symptoms are certainly distinct from one another, the inconsistency of common

socioeconomic predictors across models suggests that the findings may not be robust.

Table 5-3 shows the results for models predicting stunting and wasting. Rural

children are more likely to be stunted than all urban children. This difference, however,

disappears once family background is taken into effect. In particular, higher wealth

predicts lower chances of child stunting. Maternal education has no effect. Older mothers

predict higher chances of stunting, as does being a member of scheduled caste or tribe.

Female children and children who are close in age to their next older sibling are less likely

to be stunted, although birth order has no effect. Lastly, Muslim and Christian children are

less likely to be stunted than Hindu children or children of other religions. The inclusion of

housing characteristics does not change the model considerably (the pseudo r-square

94

increases from 0.012 to 0.013, such a small increase that it may be the function of simply

adding more variables into the model). However, having poor walls is positively related to

child stunting; it is unclear what the mechanism behind this relationship may be.

Wasting is a more unusual phenomenon than stunting, as seen in the simple

descriptive statistics in Table 5-1. The predictors wasting are quite different than those of

stunting, indicating that long-term and short-term malnutrition are distinct problems with

unique factors affecting them. Wealth is protective against both kinds of malnutrition,

while maternal education is only protective against stunting. It appears that the structure

of siblings in the family has an impact, though complicated, on both childhood wasting and

stunting. Children close in age to their next oldest sibling are less likely to be acutely

malnourished, but are more likely to be chronically malnourished. Female children are less

likely to be wasting. Lastly, children of a high birth order are more likely to be stunted,

perhaps indicating that families with more children are forced to spread their resources

thinner. Christians and Muslims are less likely to suffer wasting.

Members of scheduled castes are more likely to suffer chronic malnutrition,

whereas members of scheduled tribes are more likely to suffer acute malnutrition. Children

from other backwards castes are more likely to be stunted, but not wasted. Scheduled

tribes usually live in their own communities, often in rural areas. Members of scheduled

castes and other backward castes live in diverse communities. When there is a community

resource, members of scheduled castes and other backwards castes are typically

discriminated against and excluded from those resources, likely causing long term

disadvantage, as seen in stunting. Isolated communities are more likely to fall victim to

95

wasting, particularly if they have inconsistent trade relations with the outside world, likely

common of schedule tribes.

TABLE 5-4 shows results from models predicting the chances of a child being taken to

a health care facility for treatment. For cough/fever and diarrhea, rural children were less

likely to be taken for care than urban children of both slums and other urban areas. For

treatment of cough/fever, this disparity remains once controlling for family and

background characteristics, while as for diarrhea treatment the effect disappears. Wealth is

a positive predictor of both kinds of treatment, whereas children of a higher birth order,

female children, and Christian children are less likely to be taken for care. Maternal

education predicts higher chance of treatment, though the specific levels vary by kind of

sickness.

DISCUSSION

The contribution of this paper is for the first time to quantify child health disparities

across urban slums, other urban areas, and rural areas. The effects of residence on child

health outcomes appear to be non-existent based on the results of this paper. This is a

somewhat surprising result, given that the poor sanitation, bad water, and overcrowding in

slums seems like a situation that would breed infectious diseases that would cause such

symptoms such as coughs, diarrhea, and fevers. One explanation of the lack of relationship

between slum residence and child symptoms is that slums are a very heterogeneous group

of neighborhoods, and therefore overall do not appear to have any impact on child health.

Particular studies that have sampled only within slums found that various environmental

predictors had strong relationships with child health outcomes (Buttenheim 2008),

indicating that even within slums there is great variation in the factors that impact child

96

health. The heterogeneity of slums may mask the more direct impacts of specific living and

sanitation conditions on health outcomes.

Another possibility in explaining the (lack of) relationship of residence and

children’s health outcomes is measurement error in the dependent variables. The child

health symptoms are maternal reports and not objectively measured, and it is likely that

specific background and family characteristics may be correlated with reporting errors. For

instance, the results from the regressions on symptoms show that mothers having no

education predicts lower prevalence of children coughing and having a fever, a finding that

runs contrary to much literature that has found positive relationships between maternal

education and child well-being (Mensch et al. 1985; Hobcraft, J. W. McDonald, and Rutstein

1984b; G. T. Bicego and Boerma 1991; Boerma, Sommerfelt, and Rutstein 1991; Victoria et

al. 1992; Cleland and van Ginneken 1988).

This is perhaps less troubling when considered in the broader literature of

morbidity in the developing world. Overall, the rich in the developing world are more likely

to report being sick than are the poor. Researchers have posited several explanations for

this phenomenon. Perhaps the rich are less tolerant of discomfort (Murray and Chen

1992); perhaps the poor downplay their illnesses because they know they cannot do much

about it, a sort of cognitive dissonance; and lastly, perhaps the poor are less aware of

certain illnesses, particularly chronic illnesses, that the rich are aware of (Jishnu Das and

Hammer 2008). While the data used in this paper do not show the relationship of more

wealth and more sickness, it is possible that the maternal education and reports of

children’s illness are behaving in a similar way. For example, it may be that mothers with

97

no education are less likely to consider their children’s symptoms out of the ordinary, and

therefore are less likely to report them.

Few studies to date have rigorously examined maternal reports of children’s health

in the developing world. Some work has been done looking at maternal reports of child

health in the United States, finding that some social characteristics affect how a parent or

mother reports a child’s health, such as maternal education and socioeconomic resources

(Waters et al. 2000; Arcia 1998; Monette et al. 2007). As discussed in the introduction, the

studies that have examined caregivers’ reporting of children’s fever find that overall,

caregivers do not accurately report when a child has a fever. Nonetheless, official

Demographic and Health Surveys and National Family Health Survey reports as well as

social science research using these surveys continues to treat maternal reports of

children’s fever, cough, and diarrhea as reliable and valid measures of symptom

prevalence. This is highly problematic, particularly given that there are many plausible

reasons that the errors in maternal reports of children’s symptoms may be correlated with

various socioeconomic factors.

Despite the problems in maternal reports of child health symptoms, one consistent

finding across models is that mothers’ reports of fever, cough, and diarrhea are all higher in

households with access to bad toilets. This finding is consistent with the epidemiological

literature on childhood diarrhea and its strong relationship with household toilet facilities

and practices.

Another minor, but important, finding of this paper is the absolute necessity in

taking into account time of year a survey was taken when looking at health conditions.

98

Several papers published using the DHS data on children’s symptoms fail to account for

seasonality of the individual interviews, although official DHS reports emphasize the

importance of doing so (Ryland and Raggers 1998). In addition, new work looking at

spatial correlation of symptoms such as fever, cough, and diarrhea (e.g. Kandala and

Madise 2004) may also have biased results as geographic correlations of disease may also

reflect geographic correlations in the month the interviews were completed.

Wasting and stunting are not as problematic in terms of measurement as child

health symptoms, as they are calculated using objective measures of children’s height and

weight. The results from this paper indicate that residence location has no independent

effect on either stunting or wasting, although several other individual and family

characteristics do.

The role of caste should be highlighted in these results. Throughout the models,

caste continues to play a role in child well-being and health care. Members of scheduled

castes and other backward castes are more likely to be chronically malnourished, whereas

members of scheduled tribes are more likely to suffer short term malnutrition. Scheduled

tribes are also less likely to suffer a cough or fever, perhaps because of their remoteness to

other groups having a protective effect.

Lastly, it appears that rural children are less likely to be taken to a health care

provider when mothers think they have a cough or fever, though they are not less likely to

be taken for diarrhea. Slums have no effect. Various family and background characteristics

have effects on treatment, and this is an area ripe for more investigation. As mentioned

earlier, the National Family Health Survey has no data on provider competence, so we

99

cannot know from our results if those who sought medical treatment actually had better

outcomes. Das and Hammer (2007) found that the average health care provider in poor

neighborhoods was more likely to administer a harmful treatment than a non-harmful

one7, indicating that simply taking a child to get care is not synonymous with receiving

good care.

7 In the case of diarrhea, for instance, a harmful treatment would be a treatment that recommends

withholding fluids from a child, as children with diarrhea should receive more fluids to counteract

dehydration.

100

TABLES AND FIGURES

Table 5-1: Descriptive statistics on residence location and percent of children

exhibiting the health condition, National Family Health Survey 2005-2006

Rural Urban slum Other urban N Chi-square

Cough 17.6% 18.7% 15.3% 36269 14.9**

Diarrhea 9.2 8.5 8.3 36285 4.5

Fever 15.0 13.8 12.3 36273 20.4***

Stunting 38.3 29.8 26.0 39366 289.6***

Wasting 15.8 12.2 12.3 39366 53.9***

Medical care for

cough/fever

59.9 77.3 81.7 7893 185.1***

Medical for diarrhea 58.1 72.6 72.5 3280 38.7***

*** p<0.001; ** p<0.01

101

Table 5-2: Results from logisitic regressions of child sickness in two weeks prior to survey on residence, background,

and family characteristics, National Family Health Survey 2005-2006 Coughing Diarrhea Fever

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE

Residence

Urban non-slum -0.24 *** 0.07 -0.11 0.07 -0.02 -0.09 0.03 0.09 -0.13 -0.08 -0.07 0.08

Rural -0.07 0.05 -0.07 0.06 0.09 -0.07 -0.06 0.09 0.10 -0.06 0.06 0.07

[Reference = slum] 0.00 0.00 0.00 0.00 0.00 0.00

Family background

Wealth 0.02 0.04 0.00 0.06 0.06 0.05

Mother has no education -0.21 *** 0.04 -0.08 0.06 -0.14 ** 0.05

Mother has secondary education 0.04 0.04 0.09 0.06 0.07 0.05

Mother has higher education -0.04 0.08 -0.05 0.11 -0.10 0.09

Female headed home 0.00 0.05 0.04 0.06 0.06 0.05

Mother’s age at birth 0.00 0.00 -0.01 0.01 0.00 0.00

Short preceding birth interval -0.09 0.04 -0.08 0.05 -0.07 0.04

Female child -0.09 ** 0.03 -0.12 *** 0.04 -0.09 ** 0.03

Birth order -0.03 0.01 0.04 0.02 0.01 0.01

Scheduled Caste 0.02 0.04 0.05 0.06 0.04 0.05

Scheduled tribe -0.27 *** 0.06 0.01 0.07 -0.16 ** 0.06

Other backward castes -0.02 0.04 0.13 ** 0.05 0.03 0.04

Muslim 0.31 *** 0.04 0.14 ** 0.05 0.42 *** 0.04

Christian 0.14 0.06 -0.19 ** 0.09 -0.01 0.07

Other religion 0.08 0.07 0.21 0.09 0.06 0.07

Housing characteristics

Poor toilet 0.37 *** 0.07 0.43 *** 0.09 0.28 *** 0.07

Poor water 0.07 0.04 -0.09 0.05 0.06 0.05

Overcrowded home -0.07 0.05 0.04 0.06 -0.02 0.05

Poor floor 0.02 0.04 -0.09 0.05 0.05 0.04

Poor roof -0.04 0.04 0.01 0.05 0.04 0.04

Poor walls 0.15 *** 0.04 0.19 *** 0.05 0.08 0.04

N 36237 36237 36253 36253 36241 36241

Pseudo R2 0.02 0.02 0.00 0.01 0.00 0.01

*** p<0.001; ** p<0.01

Note: All models control for the month of the year the interview took place

102

Table 5-3: Results from logistic regressions of child wasting and stunting on residence, background, and housing

characteristics, National Family Health Survey 2005-2006

Wasting Stunting

Model 1 Model 2 Model 1 Model 2

Coef. SE Coef. SE Coef. SE Coef. SE

Residence

Urban non-slum 0.01 -0.08 0.06 0.08 -0.19 *** -0.06 0.04 0.06

Rural 0.30 *** -0.06 0.04 -0.07 0.38 *** -0.04 0.09 0.05

[Reference = slum] 0.00 0.00 0.00 0.00

Family background

Wealth -0.30 *** 0.04 -0.21 *** 0.03

Mother has no education 0.03 0.04 0.03 0.03

Mother has secondary education -0.08 0.05 -0.19 *** 0.03

Mother has higher education -0.07 0.09 -0.79 *** 0.07

Female headed home 0.06 0.05 -0.01 0.04

Mother’s age at birth 0.01 *** 0.00 -0.02 *** 0.00

Short preceding birth interval -0.10 ** 0.04 0.12 *** 0.03

Female child -0.10 *** 0.03 -0.04 0.02

Birth order -0.01 0.01 0.06 *** 0.01

Scheduled caste 0.07 0.05 0.15 *** 0.03

Schedule tribe 0.27 *** 0.05 0.03 0.04

Other backward castes 0.09 0.04 0.11 *** 0.03

Muslim -0.14 ** 0.04 -0.02 0.03

Christian -0.53 *** 0.07 -0.05 0.05

Other religion -0.16 0.07 -0.10 0.05

Housing characteristics

Poor toilet -0.00 0.07 0.03 0.05

Poor water 0.00 0.05 0.02 0.03

Overcrowded home 0.03 0.05 0.34 *** 0.04

Poor floor -0.07 0.04 0.22 *** 0.03

Poor roof -0.08 0.04 -0.05 0.03

Poor walls 0.13 *** 0.04 -0.02 0.03

N 39,330 39,330 39,330 39,330

Pseudo R2 0.002 0.013 0.006 0.031

*** p<0.001; ** p<0.01

103

Table 5-4: Results from logistic regression of treatment of cough/fever and diarrhea on residence, background, and housing

characteristics, National Family Health Survey 2005-2006

Cough or fever treatment Diarrhea treatment

Model 1 Model 2 Model 1 Model 2

Coef. SE Coef. SE Coef. SE Coef. SE

Residence

Urban non-slum 0.27 0.10 0.19 0.14 -0.01 0.20 -0.16 0.21

Rural -0.83 ** 0.14 -0.31 ** 0.12 -0.65 0.15 0.04 0.18

[Reference = slum] 0.00 0.00 0.00 0.00

Family background

Wealth 0.36 *** 0.07 0.26 0.11

Mother has no education 0.04 0.07 -0.03 0.11

Mother has secondary education 0.23 ** 0.08 0.14 0.12

Mother has higher education 0.28 0.15 0.71 ** 0.25

Female headed home 0.06 0.08 0.08 0.12

Mother’s age at birth 0.02 0.01 0.02 0.01

Short preceding birth interval 0.03 0.06 0.04 0.10

Female child -0.17 *** 0.05 -0.16 0.08

Birth order -0.07 *** 0.02 -0.09 ** 0.03

Scheduled caste 0.23 ** 0.08 0.25 0.12

Schedule tribe -0.18 0.09 0.09 0.14

Other backward castes 0.15 0.06 0.06 0.10

Muslim 0.24 *** 0.07 0.14 0.11

Christian -0.78 *** 0.11 -1.17 *** 0.18

Other religion 0.15 0.12 -0.12 0.17

Housing characteristics

Poor toilet 0.11 0.14 -0.34 0.20

Poor water -0.01 0.08 -0.05 0.11

Overcrowded home 0.10 0.08 0.26 0.13

Poor floor -0.20 ** 0.07 -0.08 0.11

Poor roof 0.16 ** 0.06 -0.04 0.09

Poor walls -0.14 0.07 -0.23 0.11

N 7887 7887 3279 3279

Log likelihood -5097 -4881 -2182 -2078

Pseudo R2 0.019 0.058 0.009 0.056

*** p<0.001; ** p<0.01

104

Chapter 6 : School Enrollment, Children’s Work, and Idleness

The foundation of every state is the education of its youth.

Diogenes Laertius, biographer of Greek philosophers

Education is a fundamental component of human capital, and population education is a

fundamental component of development. Education has long been seen as an effective and

powerful means for achieving economic growth, reducing poverty, improving individuals’

earning potential and empowerment, promoting healthy populations, and building

competitive economies (Hanushek and Wossman 2007; UNESCO 2007; World Bank 2006).

Universal children’s schooling is considered so important that World Bank specifically

included it as one of the Millennium Development Goals. School enrollment of developing

world children has been studied by both academics and policy practitioners interested in

increasing enrollment as both a means and an end of development.

A wide body of literature has shown that a variety of family background

characteristics impact a child’s chances of being enrolled in school, including economic

resources (Basu 1999; Huisman and Smits 2009), parents’ education and occupation

(Breen and Goldthorpe 1997; Huisman and Smits 2009; Buchmann and Brakewood 2000;

Colclough, Rose, and Tembon 2000; UNESCO 2004; Ersado 2005; Smits and Hosgör 2006),

birth order and number of siblings (Buchmann and Hannum 2001; Emerson and Souza

2008; Pong 1997; Chernichovsky 1985), and whether or not a child is a biological child

(Fafchamps and Wahba 2006). Additionally, various contextual characteristics have been

shown to be important factors predicting a child’s chances of attending school, including

105

rural versus urban residence (Huisman and Smits 2009); distance to the nearest school;

quality of the local schools (Buchmann and Hannum 2001; Colclough et al. 2000; Ersado

2005; Handa 2002; Michaelowa 2001; Vasconcellos 1997); the makeup of the local labor

market and the associated expectations of the returns to schooling (Buchmann and

Brakewood 2000; Colclough et al. 2000; Smits and Hosgör 2006); modernization; and

percent of female teachers, which is particularly important for girls’ school attendance

(Thomas S. Dee 2006; Leach 2006; Michaelowa 2001; Colclough et al. 2000).

Of those children not enrolled in school, many are engaged in work for pay outside

their homes. Despite various public outcries against it, child labor remains widespread

throughout the developing world (Bacolod and Ranjan 2008). The UN’s International Labor

Organization estimates that one in six children around the world between the ages of five

and seventeen work, and this proportion is higher in the poorer parts of Africa and Asia,

including India (International Labour Office 2006). Research has shown that working as a

child laborer has a negative effect on both education achievement and adult wages

(Emerson and Souza 2003, 2007a, 2007b). Child labor is also many times an

intergenerational poverty trap: children who work have less education, and so grow up to

be poor as an adult, necessitating that they send their children to work instead of to school.

A growing body of literature, particularly in economics, has looked at the economic causes

of child labor (Starting with Basu and Van 1998 and Grootaert and Kanbur 1995; For

overviews of the literature see Basu and Tzannatos 2003; Basu 1999; Brown, Deardorff,

and Stern 2003).

106

Lastly, some children who are not enrolled in school are also not engaged in work.

This group is most often ignored by the literature, save a few recent studies that have

begun to explicitly include this group (Rosati and Tzannatos 2006; Biggeri et al. 2003;

Bacolod and Ranjan 2008). Less than half of these children are engaged in significant

household chores (such as caring for younger siblings), are disabled, or are job searching

(Biggeri et al. 2003). In fact, a research team in India found that children not attending

school were mostly playing hopscotch, not working (De and Dreze 1999). This group of

children under age 18 has come to be called “idle children”, indicating their lack of

participation in both school and work.

While the literature on school enrollment and attendance has often focused on rural

and urban divides, the literature on child labor and idleness almost completely ignore the

urban or rural context. While the academic literature on American poverty has experienced

a surge in the study of neighborhood effects (Wilson 1987; Sampson, Morenoff, and

Gannon-Rowley 2002), the literature on the developing world has yet to focus on any local

contextual factors such as neighborhoods in studies of poverty (Montgomery et al. 2003).

Scores of academic and policy studies have shown that children are less likely to attend

school in rural areas than in cities (Montgomery et al. 2003), and this is often considered

the result of poor quality schools or long distances to the nearest school in rural areas

(Huisman and Smits 2009). These explanations implicitly assume that urban children

automatically are advantaged because they have better access to schools. This is a shaky

assumption in the cities of the contemporary developing world. As the world’s poor

become more urban and the poor are concentrated and segregated from the non-poor in

107

urban centers (Massey 1996), it is unclear whether traditional benefits of cities extend to

the growing urban underclass and those living in slums.

The only study to date that explicitly examines slum residence and child activities

was a study of Kenya, finding that slum children8 are less likely to attend school than their

urban counterparts (Mugisha 2006). No literature has looked at children’s work and

idleness across urban contexts. The paucity of literature on the urban context on child

school, work, and idleness is of importance particularly as more and more poor children in

the developing world are living in cities, and in particular slums. The 2001 Indian census

reports that 16.4% of urban children ages 0-6 live in slums. This means that every sixth

urban child is a slum resident, a total of over 6 million children. In cities of over 1 million

residents, the proportion of urban children in slums jumps to 27.3%. As Indian cities like

cities throughout the developing world continue to experience high growth both from

migration and natural increase, often in the poorest segments, it is probable that the

proportion of urban children in slums will stay at least as high, if not grow over the next

few decades. What are the risks the new developing world urban poor children face with

regards to schooling?

SCHOOL FACILITIES, CONTEXTUAL FACTORS, AND SLUMS

While older literature on school enrollment in the developing world focused mainly

on family and background characteristics that predict school enrollment, the current

8 Mugisha (2006) uses a version of the United Nation’s definition of slums drawn from the Millennium

Development Goals. That is, to be considered a slum you meet one of the following five criteria: lack of access

to improved water, lack of access to improved sanitation, nondurable building materials, overcrowding of a

residence, and insecurity of tenure.

108

literature has broadened to include a focus on contextual factors affecting the likelihood of

a child being in school. The first set of factors often considered concern available

educational facilities. Both the quantity and quality of available schools are important to

education participation, and in particular are important for disadvantaged groups such as

girls and the poor (Buchmann and Hannum 2001; Colclough et al. 2000; Ersado 2005;

Handa 2002; Michaelowa 2001; Vasconcellos 1997; Huisman and Smits 2009). Throughout

the developing world, it has been shown that people don’t want to send their children to

low quality schools because they realize that their children will learn less in poor quality

schools (Buchmann and Brakewood 2000; Colclough et al. 2000). Several studies have

shown that distance plays a role in school attendance, with shorter distances to schools

associated with higher chances of enrollment (Colclough et al. 2000; Peter Glick and David

E. Sahn 2006). Additionally, the percent of female teachers at the local school is important

for educational participation, and in particular the participation of girls (Colclough et al.

2000; Peter Glick and David E. Sahn 2006).

Context factors apart from educational facilities also can matter. Rural residence has

long been associated with lower levels of school enrollment (Huisman and Smits 2009).

Several economists have argued that local labor markets that reward schooling (as

opposed to heavily agricultural local labor markets) make parents more likely to put

children in school because the parents recognize the delayed benefits of more education for

later earnings (Buchmann and Brakewood 2000; Colclough et al. 2000; Smits and Hosgör

2006). Modernization, including better roads and trains for transportation and more

modern pressures to put children in school may also have an impact. Additionally, some

studies have showed that various family and demographic factors vary depending on the

109

context one is in. Most recently, Huisman and Smits (2009) showed that in 30 developing

countries, household effects varied widely by the local context.

Again, no literature has covered the context of slums explicitly. The heavy bias in the

literature is a focus on context in explaining disparities in rural and urban school

enrollment, thus ignoring questions of how varying urban contexts may affect school

enrollment.

CHILD LABOR AND IDLENESS

Child labor is the source of much moral and public outrage, particularly in the

developed world. Although one can get mired in the specifics of defining and measuring

child labor, by any definition both the number and proportion of children around the world

engaged in economic work remains substantial. The proportion of children engaged in

child labor is declining, but the United Nations International Labor Office estimates that

still one in six children in the world are engaged in economic activities (International

Labour Office 2006).

A growing body of literature, particularly in economics, has looked at the causes of

child labor (Starting with Basu and Van 1998 and Grootaert and Kanbur 1995; For

overviews of the literature see Basu and Tzannatos 2003; Basu 1999; Brown, Deardorff,

and Stern 2003). Many contemporary economists argue that child labor is not simply child

abuse resulting from parental selfishness, but instead is a reflection of stark poverty that

compels parents to send children to work for household survival. This is supported by

evidence such as the strong relationship between child labor and GDP per capita (Krueger

110

1998; Moehling 1999), and the decline of child labor as a result of economic growth

(Edmonds 2005).

There are, of course, many negative outcomes associated with child labor. First and

foremost, children who are working are less likely to be enrolled in school, and with less

education these children end up as poor adults (Basu 1999). This creates an

intergenerational cycle of child labor, as poor parents often are forced to send their own

children to work, reproducing the cycle. There are other negative effects, as well. Child

labor also often has physical dangers associated with it (Edmonds and Pavcnik 2005), and

the child labor market often involves coercion and psychological pressures (Silvers 1996).

Most research on children’s labor implicitly assumes that the only alternative to

child labor is schooling. In fact, a substantial proportion of children throughout the

developing world are neither enrolled in school nor engaged in economic activities. These

children are typically called “idle” children, referring to their idleness with relation to

educational and economic activity. The extent of idleness varies considerably across

countries, ranging from as low as 2% to as high as 35%. Many countries have over 20% of

school age children idle, often outnumbering children engaged in economic work (Biggeri

et al. 2003). The level of idleness also differs noticeably within countries, typically with

rural children, girls, and young children experiencing more idleness than urban children,

boys, and older children.

It may seem logical that many children not in school or working are needed for

domestic work, such as caring for younger siblings, getting water from public water

supplies, and cooking. However, the best cross-national study by Biggeri et al. (2003)

111

indicates that less than half, and often as low as one fifth, of idle children are engaged in

significant domestic work. In fact, in this study idle children overall performed only

marginally more domestic work than children in school and working. Some idle children

are unemployed and looking for work, but in most countries this accounts for less than

10% of idleness. Lastly, some idle children are chronically ill or disabled, but again this

accounts for less than 10% of idleness. Overall, then, no more than half of children on

average are engaged in any significant domestic work, job searching, or are chronically ill.

Idleness is generally considered to have negative impacts on children, as current absence

from schooling predicts future absence from schooling and lower lifetime educational

attainment, resulting in lower earnings later in life. Generally, this work on idle children

does not include homeless children living without their families.

There are some theoretical explanations for idleness as rational household behavior.

If costs for schooling are high and there are high fixed costs associated with sending a child

to work (such as high transportation costs), it may be more efficient or logical to let a child

sit idle (Biggeri et al. 2003). Others have brought child ability into the equation, showing

that children of lower cognitive ability are more likely to be idle, suggesting that parents

may perceive a higher benefit for schooling costs for talented children (Bacolod and Ranjan

2008).

BACKGROUND: EDUCATION AND POLICY IN INDIA

India is committed to increasing enrollment and access to education for its entire

population. It is a right guaranteed to citizens in the Constitution: “The State shall provide

free and compulsory education to all children of the age of six to fourteen years in such

manner as the State may, by law, determine” (CITATION). A large network of government

112

schools fills this need, but they have been of varying quality and accessibility for decades

following the country’s independence.

Starting in 1986, India’s New Education Policy made primary education a national

priority and committed increasing resources to this end, up to at least six percent of the

GDP per year. Several federally sponsored programs were launched to improve primary

education. In 1994 the District Primary Education Programme (DPEP) was started by the

federal government, aimed at universalizing primary education by reforming and

revitalizing the primary education system, particularly in rural areas which tend to suffer

from low enrollment. The program opened 160,000 new schools, delivering education to

approximately 3.5 million children (Ministry of Information and Broadcasting 2009). In

2002, the Indian government in conjunction with the World Bank, UNICEF, and over 7,000

small NGOs launched a new education program called Sarva Shikshka Abhiyan, which

builds on the initiatives of the earlier DPEPs to improve access to education with more

local schools, train teachers, develop teaching materials, and monitor learning outcomes. It

was the first to support upper primary education for grade 6-8 (World Bank 2009a).

In the same year, the Supreme Court of India announced a groundbreaking ruling

on the right to food, directing the federal government to fully implement a scheme to

provide cooked, mid-day meals to all children in primary schools. This has not only had

nutrition benefits for children, but also has attracted students to schools and thereby

increased enrollment.

Secondary education has only recently entered the public dialogue as an important

and necessary next step in India’s development. Traditionally, the low rates of secondary

113

school attendance were seen as a compounded consequence of low primary school

enrollment. The view clearly has merit since as primary school enrollment rose, so did

demand for secondary school. Prime Minister Manmohan Singh in his 2007 national

Independence Day address recognized the trend and urged it forward. “As our primary

education programmes achieve a degree of success,” he noted, “there is growing demand

for secondary schools and colleges. We are committed to universalizing secondary

education. An extensive programme for this is being finalized.” Progress on this frontier

remains elusive, however. No major national programs have yet tackled the problems of

low secondary school enrollment. In 2004 the Ministry of Human Resource Development of

the Government of India commissioned a report on the universalization of secondary

education (CABE Committee 2004). Though government schools are legally supposed to be

free, recent government studies show that secondary school students (even those

attending free government schools) incur expenses for schooling, including for books,

uniforms, transportation, and testing fees (Mehta 2002). The Commissions (Valerie: what

commissions are these?) and others have called for increased financial support for

secondary education (such as free books), as well as vastly expanded secondary education

infrastructures, particularly to serve traditionally disadvantaged groups.

India is a country of great extremes and inequality in education. It is worthwhile to

note that at the high end of Indian education, there are the world class Indian Institutes of

Technology, with ferociously competitive admission and a physics establishment that rivals

many developed nations. At the same time, the vast majority of Indian children never see

the inside of a high school. India also has the longest track record of any country with

114

affirmative action; in efforts to reduce caste inequality, in each entering class for college a

certain number of spots are reserved for children of scheduled castes and tribes.

POSSIBLE SLUM EFFECTS

There are several potential effects slum may have on the activities of children. I will

first go through four possible effects on school attendance, then effects on child work and

idleness. First, slums that are deemed “non-notified” by the local, state, and national

governments are considered illegal slums, and the slums and residents of those slums are

not eligible for certain benefits, such as having a ration card, being able to open a bank

account, or receiving trash collection. In some places, government schools may not enroll

students who live in non-notified slums. Other slums in typically non-populated areas, such

as along highways, riverbanks, on airports, or in heavily industrial areas, may not be

included in the catchment areas for any government school, leaving children with no school

to attend. Practices and laws regarding the administration and handling of slums vary

greatly both between cities and municipalities and even within cities, as large cities are

often broken up into local districts with their own sets of practices, making this difficult to

test.

Second, slum children may not have good access to government schools. Distance

from schools has often been shown to be a predictor of school enrollment of children

(Huisman and Smits 2009). Slum children may face one or multiple variations on the idea

of distance. First, particularly in very large slums, children may actually be a considerable

physical distance from a school. Even if there is not a large physical distance from a home

to a school, covering that distance may be difficult making it akin to a long distance. Slums

are often built on pieces of land that are undesirable for higher income residents, such as

115

areas directly abutting highways and railroads, atop very steep slants, or along the banks of

rivers. Additionally, some slums have higher crime, for example, perhaps making it difficult

for children to get to school on their own. Lastly, many slums are segregated by ethnicity or

religion, and in places or times of ethnic and religious tension, children may not want to

venture across the boundaries of these groups. Thus, it is possible that slum children may

have difficulty overcoming these physical and human boundaries to get to school.

Third, in middle- and upper-class popular Indian dialogue, slums are often

considered to be refuges and outposts of poverty and “backward”, conservative attitudes

characteristic of rural areas. If this is true, it is possible that social norms develop in slums

that do not promote schooling among children, but promote more traditional ideas about

education and work, such as ideas that girls do not need or should not be highly educated,

and that boys should be working to contribute to family income rather than wasting time in

school. No research has rigorously examined whether or not those in rural areas actually

hold more such conservative attitudes than those in cities. Economists would argue that

attitudes toward children’s schooling and work are directly affected by the local labor

markets, and such attitudes would not become sui generis, taking on a life of their own

divorced from economic conditions. These factors make this a tenuous argument, though

publicly popular.

Lastly, related but distinctly different from the previous hypothesis, it is possible

that slums function as many researchers argue neighborhoods of concentrated poverty

function in the United States. Though there are heated debates on the proximate causes and

mechanisms creating high poverty neighborhoods in the United States, starting with

116

William Julius Wilson a body of literature has argued that the residential concentration of

poverty isolates the poor from the non-poor and the corresponding resources, networks,

and role models, leading to the development of a different set of norms and beliefs (often

termed oppositional culture) in poor neighborhoods and related negative outcomes.

Various other researchers argue that high poverty neighborhoods have negative effects due

to one or many other factors including high stress, social disorganization, lack of

attachment to neighborhood institutions, lack of connections to information networks on

jobs, lack of neighborhood resources, and difficulty creating political alliances to attract

public resources. Specifically with regards to education, high poverty neighborhoods are

related to lower education attainment or higher rates of dropping out (Ainsworth 2002;

Small and Newman 2001). If we consider slums akin to high poverty neighborhoods in the

American context, we might expect that living in slums would have similar effects to living

in a high poverty neighborhood in the United States.

There may also be slum effects on children’s work and idleness. Children in slums

likely have better access to more work than children in rural areas, as well as be in a labor

market with higher demand, perhaps increasing the proportion of children working.

However, the concentrated poverty of slums may isolate slum children from labor markets

when compared to other urban children. It is an empirical question what patterns of child

work are in slums. The small body of work on idleness does not lend to any obvious

impacts slum residence may have on idleness. Again, this is an empirical question that has

yet to be answered.

117

I aim to fill gaps in our knowledge about the disadvantages associated with slum life

in terms of children’s school enrollment, work, and idleness. I investigate simple yet

important unanswered questions: do children in slums attend school at lower rates than

their urban and rural peers? Do they work for pay more? If there are disparities in school

attendance, work, and idleness by location, how much of these disparities are attributable

to differences in family and background characteristics, such as socioeconomic status?

Lastly, for those that do not attend school, are there major differences in why they do not

attend school that can shed some light on the mechanisms at work?

METHODS

Data for this chapter come from the NFHS-3 household survey. The household

survey covers a wide array of household characteristics, as well as asking a set of detailed

questions for up to 35 household residents. I use data on children ages 5-18 at the time of

the study, as information is gathered for children at these ages on their school attendance.

For analyses including information on children’s work activities, the sample is limited to

children ages 5 through 14, as the NFHS surveys ask about work only through age 14.

I examine the effects of residence on children’s activities using two modeling

strategies. First, because school attendance receives the most attention and focus, I predict

the chances of a child being enrolled in school using logistic regression. I look at the effects

of residence alone, as well as including a wide set of family and background characteristics

that have been deemed important to school enrollment in the literature on schools in the

developing world. Next, I use multinomial logit regression to predict children’s activities in

one of four categories: in school only, working only, neither working nor in school (idle),

118

and both in school and working. Again, I examine the raw effects of residence, then

examine the effects of residence over and above family and background characteristics.

Ideally a hierarchical or fixed effects model would be used to allow for effects at the

neighborhood level. Unfortunately, the data do not allow this, as the Indian government

will not release cluster codes at the neighborhood level, rendering such models impossible.

To account for this, I ran models interacting the residence variables with each of the

independent variables; I also ran separate models for each location. On the whole, these

models were not informative, and very few coefficients varied by location. These tables can

be found in the Appendix. To account for clustering, I only consider significance at the

p<0.01 level, rather than the standard p<0.05 level. Since clustering does not bias

estimates, but instead artificially deflates standard errors, using a lower level of

significance accounts for this.

The set of family and background characteristics I include in my models are rooted

in the large literature on children’s school attendance, work, and idleness. Table 6-1 shows

the full set of family and background variables I include in my model and their coding.

Several socioeconomic factors affect children’s school enrollment. Children from families

with more economic resources are more likely to be enrolled in school (Basu 1999;

Huisman and Smits 2009). Economic resources mean parents can pay for direct costs such

as school fees and books. Additionally, families with more economic resources face lower

opportunity costs of their children not being able to work or contribute to family income.

For this paper, the log of NFHS wealth index is used for a measure of wealth.

—Table 6-1 about here—

119

Father’s occupation has been found to have an impact on school enrollment. Fathers

who work in salaried employment, particularly in non-manual labor, are more likely to

have children enrolled in school, perhaps because these men are more aware of the

importance and benefits of education (Breen and Goldthorpe 1997; Huisman and Smits

2009). Conversely, self-employed parents likely face more opportunity costs to sending

their children to school. Here, I code father’s occupation as agricultural, lower non-

agricultural, and upper non-agricultural, consistent with Huisman and Smits (2009)

research on school enrollment using data similar to my own.

Mothers’ employment status could have competing effects on children’s school

enrollment. A working mother may necessitate girl children staying at home to complete

domestic chores. Alternatively, a mother working may increase her power to send her

children to school. The literature is not conclusive on these effects. I include a dummy

variable for whether or not a mother works outside the home.

Lastly, parents’ with more education more often have children who are enrolled and

stay in school (Buchmann and Brakewood 2000; Colclough et al. 2000; Ersado 2005; Smits

and Hosgör 2006; Mugisha 2006; Huisman and Smits 2009). For girls, maternal education

is especially important (Emerson and Souza 2007a). I code parents’ education into four

categories: no education, primary education, secondary education, and higher education.

The second set of family level factors often important for school enrollment are

demographic factors. These include things such as living with extended family, a child’s

birth order and number of siblings, having a missing parent, and being a biological child.

Younger children often have more opportunities to go to school because older children are

120

responsible for domestic or outside work (Buchmann and Hannum 2001; Emerson and

Souza 2008; Huisman and Smits 2009). I use a child’s birth order as a number

corresponding to their birth order, so a first child is coded 1, the second child is coded 2,

and so on9. Often (although not always) the literature often finds that more children in a

family means lower chances of children attending school (Buchmann and Hannum 2001;

Pong 1997; Chernichovsky 1985; Huisman and Smits 2009), perhaps because resources are

being split among more children, leaving less money for each child’s schooling. I include the

number of children in the family.

RESULTS

Table 6-2 shows basic descriptive statistics on the activities of children by place of

residence. These numbers are for children ages 5-14. Although school attendance was

asked of children up to age 17, questions about children’s work outside the home were only

asked of children ages 5 to 14. This is an unfortunate inconsistency in the survey that is

worked around as best as possible in these results.

The first important finding in Table 6-2 is that across place of residence, a large

majority of children are enrolled in school, ranging from 69% in rural areas to 76% in non-

slum urban areas. Non-slum urban children do attend at higher proportions than either

rural or urban slum children, however. Very few children overall are engaged only in work

outside the home, from 2 to 3 percent. There are surprisingly high percentages of children

who are neither working nor in school. Nearly one quarter of rural children are idle, and

9 Though some studies lump birth order together (Huisman and Smits 2009), implying that

higher order births are not distinguishable from one another, I consider each birth on its

own. I ran models both ways, but there were no substantive differences in the results.

121

17-19% of urban children are idle. Lastly, more children are both working and in school

than working alone, and slum children are more represented at 7% in this category than

are either other urban children or rural children.

—Table 6-2 about here—

More descriptive statistics are presented in Figure 6-1, Figure 6-2, and Figure 6-3,

showing the age patterns of school, work, and idleness by place of residence. Figure 6-1

shows school attendance through age 17. Overall, school enrollment rises with age until

around age 8, when in all locations attendance is very high at 85-90%. Around age 11,

attendance starts to drop. The patterns by residence are interesting: slum children’s school

attendance is close to that of non-slum children until age 10 or 11, when it falls off and

becomes more similar to proportions of rural children. By the middle and late teenage

years, there are large differences in proportions of children attending school. At ages 15

and later, the difference in urban non-slum children’s school attendance and both rural and

slum children’s school attendance is roughly 20%. This implies that though the numbers in

Table 6-2 indicate smaller differences in school attendance, these disparities grow

considerably past age 14 (as included in that table).

Were family and background characteristics the sole factor responsible for differing

rates of school attendance by place, the age trends in school attendance should be parallel

to one another. The marked change in slum children’s school attendance by age suggests

that slums have some relationship to school attendance not solely explained by

background.

—Figure 6-1, Figure 6-2, and Figure 6-3 about here—

122

Figure 6-2 shows the proportion of children working outside the home by age,

regardless of school attendance status. Proportions of children working outside the home

increase with age across residence. Slum children are most likely at all ages to be working.

Other urban children and rural children have similar rates of working outside the home.

Figure 6-3 shows the proportion of children idle by place. The proportion of children idle

starts near 70% for all residence locations then drops quickly with age before leveling off

around ages eight and nine. Around age eleven the proportions begin to slightly rise again,

increasing between 5 and 10% by age 14. Still, of children ages eight to fourteen, no more

than 20% of children are idle, and considerably fewer at some ages. Lastly, the bottom right

panel shows the proportion of children both in school and working by age.

Next, I consider the impact of family and background characteristics in explaining

differences in activities by place of residence. First, I examine school enrollment alone

using logistic regression. Table 6-3 shows results from logistic regression models

predicting child school attendance in the year of the survey. Model 1 shows the raw effects

of location. Model 2 includes the set of background characteristics. Because of the patterns

observed by age in the descriptive data, an age squared term is included to capture the

shape of enrollment seen in the graph in Figure 6-1.

—Table 6-3 about here—

Many of the results in the final model are what one would expect given the

literature. The chances of a child being in school rise with age, and then fall, as indicated by

the age and square of age terms. Wealth predicts higher school attendance. Girls, Muslim

children, and children from scheduled castes, scheduled tribes, and other backwards castes

123

are all less likely to be attending school. Parents’ characteristics also predict school

attendance: mothers and fathers having no education predicts lower chances of school

attendance, and fathers having secondary and higher education positively predict school

attendance. Having a missing father results in 20% lower odds of being in school, as does

having a missing mother. Children in larger families also have lower chances of attending

school.

Next we move to our variables of interest, the location of residence. In the bivariate

models, residence has clear effects. Children in non-slum urban areas have odds of being

school that are 1.9 times the odds of both slum and rural children, who are

indistinguishable from one another. Once controlling for family and individual

characteristics in model 2, however, children in slums are at the highest risk for not

attending school.

Theorizing that family and individual characteristics may have varying impacts

across locations, I ran models including interaction terms of residence and individual and

family predictors, and I also ran separate models predicting school attendance in rural,

slum, and non-slum urban areas to compare how coefficients differ across models (results

not shown). Overall, almost all of the predictors had very similar effects in rural, slum, and

other urban areas. Only a few predictors had different impacts across location of residence.

Wealth was the first, and had the clearest relationship with location and schooling. Wealth

has the strongest influence in non-slum urban areas, followed by slums and finally rural

areas. One way of interpreting this is that wealth buys the most advantage in non-slum

urban areas, whereas it buys the least advantage in rural areas, with slums falling

124

somewhere in the middle. Additionally, the interactions show that girls are less likely to

attend school only in rural areas; in slums and non-slum urban areas, girls are as likely to

be in school as boys. Lastly, children of scheduled tribes appear to be less disadvantaged in

rural as opposed to urban areas.

Table 6-4 shows results from a multinomial logit regression of children’s activities

on predictor variables. Children can be in one of four outcome categories: attending school

only, working only, idle (neither in school or working), or both working and in school.

Because school attendance is implicitly considered the ideal activity for children in the

literature, school attendance is the base outcome that other outcomes are compared

against. The first model, including just residence and age, is shown on the top half of the

table. Here, we see some varying patterns of the effect of residence. Rural and slum

children are more likely than urban non-slum children to be idle instead of in school. Slum

children are more likely than any other children to be working instead of in school. And

lastly, all urban children (slum and otherwise) are more likely to be both working and in

school than rural children.

—Table 6-4 about here—

These patterns change once child and family characteristics are taken into account.

All else held equal, rural children are less likely to be idle, working, or working and in

school rather than just in school as compared to all urban children. This is an important

finding, as the literature often considers structural factors (such as the lack of nearby

schools) as the reason rural children have lower rates of school attendance, whereas this

125

research suggests that once family characteristics are taken into account, rural children are

actually more likely to attend school.

Various other interesting patterns appear in these models. Idleness is related to

several factors. Female children are much more likely than male children to be idle instead

of in school. Children of scheduled castes and tribes are particularly likely to be idle, as are

Muslim children. A missing mother or low maternal education predicts higher odds of

being idle or working, even taking into control family wealth. Working has varied

predictive factors. Children with working mothers are much more likely to be working

themselves. Having a father with a secondary or higher education predicts lower chances of

working, and higher wealth predicts lower chances of working compared to being in

school. Lastly, looking to both working and being school, the models show that girls are less

likely to be both working and in school than just in school. Children with working mothers

are more likely to be working while in school than just in school.

Again theorizing that the mechanisms at work may be different across places of

residence, I ran models interacting location of residence with the individual and family

predictors (results not shown). Overall, the vast majority of individual and family factors

have the same impacts across location. Again, wealth had varying impacts. Wealth predicts

higher chances of being in school than being idle, but this effect is almost twice as strong in

urban non-slum areas than in rural or slum areas. This can be interpreted that wealth

protects against a child being idle, but it is much more protective in non-slum urban areas

than in slums or rural areas. Wealth predicts higher chances of being in school as compared

to working, but this impact is less in rural areas than urban areas, either slum or non-slum.

126

Here the interpretation may be that wealth protects against child labor, but it is much more

protective in urban areas (slum or non-slum) than it is in rural areas.

Only three other effects varied by residence. Female children are more likely to be

idle than in school only in rural areas. Children of scheduled tribes are more likely to be

idle in urban areas than rural areas. This is likely because in rural areas, scheduled tribes

typically have their own communities and thus their own community resources such as

schools. Members of scheduled tribes in urban areas, however, much share in community

resources with other groups, and are therefore much more likely to be discriminated

against and excluded than in rural areas where they have their own schools. Lastly, a child

living in a home with more children are more at risk for both working and being in school

than just being in school only in slum areas. In rural and other urban areas, the number of

children in the home does not impact the odds of a child working while being in school.

The logistic model predicting school attendance shows that slum residence is

negatively related to school attendance, all else held equal. The multinomial model reports

that slum residence has no relationship with school attendance when family and individual

characteristics are taken into account. On the surface, this seems to be conflicting reports of

how slum residence may matter. To further investigate the exact nature of this

relationship, I revisited logistic models predicting school attendance. As seen in Figure 1,

the raw proportions of slum children attending school from around age 14 and on drop

more quickly for slum children than other urban children. The multinomial models only

look at children through age 14, as children’s work is only considered through this age.

Thus, it may be that slum residence has a negative impact on school attendance at older

127

ages, and the multinomial model may not be reflecting this given its inclusion of only ages 5

through 14. I revisited logistic models, this time running separate models for children ages

5-13 and children ages 14-18, reported as models 3 and 4 of Table 6-3. Here, we see that

slums are indistinguishable from other urban areas under age 14, but are negatively

associated with school attendance from age 14 and on, even taking into account family and

background characteristics. This explains the discrepancy between the model predicting

school attendance alone and the multinomial models: slum residence is only negatively

related to school attendance in children of secondary school age.

The primary and secondary school separate models indicate that several individual

and family characteristics have very different effects on primary versus secondary school

attendance. Wealth is a stronger predictor of secondary school attendance than primary

school attendance. The impact of parents’ education varies considerably for primary and

secondary school attendance. For primary school attendance, both the mother and father’s

completion of primary school is a positive predictor of a child attending school, while

having a secondary education or higher has no additional benefit. For secondary school

attendance, however, parents having secondary education and higher education have

increasing impacts. For instance, a child whose mother has a secondary education is 1.6

times more likely to be attending secondary school than a child whose mother has only a

primary education, and a child whose mother has higher education is twenty-four times as

likely. The same pattern follows with paternal education: a child whose father has a

secondary education is 1.5 times as likely to be attending secondary school, and a child

whose father has higher education is more than four times as likely to be attending

secondary school.

128

CONCLUSIONS

There are three important results pertaining to residence and children’s activities

found in this research. First, living in slums predicts lower chances of a child being enrolled

in secondary school, but not primary school. This indicates that something is going on that

causes more children in slums to stop their schooling after primary school than in other

urban areas or rural areas, all else held equal. Given the current state of secondary

education in India, and in particular the lack of any centralized effort to reform and

universalize secondary education, it is likely that slum children face a lack of infrastructure.

Observation in various slums in India indicates that primary schools are relatively common

and accessible, but secondary schools are scarce. In addition, as Agarwala (2006) showed

the incredibly abundance of low-skilled, home based informal work in slums, it is likely that

older slum children have many economic options in the informal economy. It is likely that

the scarcity of secondary schools, combined with an abundance of informal work

opportunities create the low rates of secondary school attendance for slum children. As

India moves forward with plans and policies to universalize secondary school attendance,

the disadvantages of slums should be a focus of narrowing the gaps between marginalized

and dominant groups. Further research should be done to better understand these forces at

work amongst older children, including in-depth quantitative and qualitative work

examining school and informal work in secondary school age slum children.

The finding that slum residence is important to secondary school attendance but not

primary school attendance raises a more general issue: the predictors of secondary school

attendance are not the same as the predictors of primary school attendance. The vast

majority of literature on school attendance in the developing world either focuses solely on

129

primary school attendance or lumps all school attendance through age 18 together. This

research indicates that lumping primary and secondary school together would be a

mistake, as the predictors of each are unique. In particular, it seems that more attention

needs to be given to looking at secondary school attendance as primary school attendance

rates climb near saturation. In India, for example, my data show that by age 11 around 90%

of all children are attending school, leaving only room for small improvements. Vast

improvements are needed in secondary school attendance, however. A new line of research

investigating secondary school in the developing world is needed to further international

goals of improved education for the developing world.

The next major finding in this chapter is that for children ages five through fourteen,

rural children are not in fact disadvantaged in terms of school attendance. In fact, once

taking into account family background, rural children are actually more likely to be in

school than their urban counterparts. This finding runs counter to much of the literature as

well as to many of the implicit assumptions of development practitioners that focus on

increasing access to education in rural areas. Many studies of the developing world

continue to show that rural children have lower levels of school attendance, even

controlling for socioeconomic and other family characteristics. This does not mean,

however, that these data are flawed or incorrect. Rather, it is possible and even likely that

the high levels of primary school attendance that India have achieved make it different

from many other developing world countries, such as much as sub-Saharan Africa. Having

reached very high levels of primary school attendance, it is possible that India is facing a

new set of challenges with regard to education. Rather than rural residence being a

detriment, it appears that urban residence may be a detriment to school attendance. Again,

130

further research is needed to investigate why this shift has taken place and why urban

children may be disadvantaged. It may be, for instance, that the more vibrant economy of

cities draws more children to work and out of school. As an increasing proportion of

developing countries reach high primary school attendance, India’s patterns are likely to

become common to many countries.

This research points to the fact that some family and background characteristics

vary by place. Most notably, wealth appears to buy more advantage and protection in some

areas. Family wealth buys children the most advantage in school attendance in non-slum

urban areas, followed by slums, and finally rural areas. This suggests that rather than being

great equalizers where the poor experience greater access, in fact cities may further

entrench inequality by rewarding the rich even more than in rural areas. Again, this is a

fine point that begs for more study into how wealth plays out in various contexts. As India

and the rest of the world continue to rapidly urbanize, the nature of inequality will shift

from being predominantly an urban-rural inequality to an intra-urban inequality. The

finding in this paper that wealth buys more in cities than rural areas suggests that cities

may actually expand inequality by rewarding the wealthy more than in rural areas. In this

research I examined advantages wealth confers for educational attendance. As education is

one of the best tools to reducing poverty and inequality, the greater advantages of wealth

in urban areas are somewhat alarming, particularly given than wealth most certainly

confers additional advantages in not just school attendance but the quality of schools

attended as well.

131

Lastly, this data support the conclusion that India has made phenomenal progress

toward universal primary enrollment. Enrollments of around 90% for all groups around

ages seven through eleven indicate a great deal of success in implementing universal

education. If India can tackle the problems of secondary education with as much force and

gusto as it tackled primary education, universal secondary education is within it reach.

132

FIGURES AND TABLES

Table 6-1: Descriptions, measurements, and means used in the analysis by place of

residence, National Family Health Survey 2005-2006

Variable Description Measurement Rural Slum Urban

non-slum

Wealth Log of DHS wealth factor

score

11.5

(.76)

12.2

(.41)

12.5 (.34)

Mother’s employment Employed = 1 48.0% 33.5% 24.8%

Father’s education

No education

Primary

Secondary

Higher education

Yes = 1

Yes = 1

Yes = 1

Yes = 1

28.0%

14.5

28.2

3.3

25.8%

12.1

31.8

7.2

11.8%

7.3

39.1

22.4

Mother’s education

No education

Primary

Secondary

Higher education

Yes = 1

Yes = 1

Yes = 1

Yes = 1

53.7%

11.8

16.9

1.0

45.2%

7.9

27.3

3.9

24.9%

7.1

35.5

16.9

Total kids in home Number of kids in the home 3.77 3.85 3.18

Female child Female =1 49.7% 47.7 46.5

Missing parent

Father missing Yes = 1 25.8 23.1 19.2

Mother missing Yes = 1 16.5 15.8 15.4

N 38,642 3,324 4,380

133

Table 6-2: Proportion of children in work, school, and idleness by place of residence,

National Family Health Survey 2005-2006

Rural Urban slum Urban non-slum

School only 69.3% 70.8% 76.0%

Work only 2.1 3.3 2.0

Idle (no school or work) 24.5 18.8 17.2

Both school and work 4.1 7.0 4.8

N 71791 7035 9150

134

Table 6-3: Odds ratios and standard errors from logistic regression of school attendance on

residence and background characteristics

Model 1 Model 2:

Full Model

Model 3:

Ages 5-13

Model 4:

Ages 14-18

OR SE OR SE OR SE OR SE

Residence

Rural 1.00 0.04 1.86 *** 0.05 1.60 *** 0.06 2.56 *** 0.08

[Reference = Slum] 1.00 1.00 1.00 1.00

Urban non-slum 1.90 *** 0.06 1.19 ** 0.06 1.03 0.08 1.28 ** 0.09

Child Characteristics

Age 4.01 *** 0.02 4.71 *** 0.02 15.18 *** 0.05 0.04 *** 0.44

Age squared 0.94 *** 0.00 0.93 *** 0.00 0.88 *** 0.00 1.09 *** 0.01

Female child 0.70 *** 0.02 0.77 *** 0.03 0.59 *** 0.04

Family Background

Wealth 2.08 *** 0.02 1.86 *** 0.02 2.94 *** 0.04

Scheduled caste 0.72 *** 0.04 0.74 *** 0.05 0.70 *** 0.06

Scheduled tribe 0.81 *** 0.04 0.77 *** 0.06 0.96 0.07

Other backwards caste 0.73 *** 0.03 0.75 *** 0.04 0.75 *** 0.05

Muslim 0.62 *** 0.03 0.61 *** 0.04 0.59 *** 0.05

Christian 0.99 0.06 0.73 *** 0.08 1.75 *** 0.11

Other non-Hindu religion 1.06 0.07 0.79 *** 0.09 1.82 *** 0.12

Mother has no education 0.59 *** 0.04 0.54 *** 0.05 0.64 *** 0.08

Mother has secondary

education

1.11 0.05 1.03 0.06 1.62 *** 0.1

Mother has higher education 0.92 0.11 0.84 0.13 24.29 ** 1.01

Mother not in home 0.79 *** 0.06 0.55 *** 0.09 0.47 *** 0.11

Mother works outside home 0.94 0.03 1.03 0.04 0.73 *** 0.06

Number of kids in home 0.92 *** 0.01 0.94 *** 0.01 0.92 *** 0.02

Father has no education 0.68 *** 0.04 0.59 *** 0.05 0.73 *** 0.07

Father has secondary

education

1.20 *** 0.04 1.07 0.05 1.52 *** 0.07

Father has higher education 1.46 *** 0.08 1.07 0.1 4.22 *** 0.23

Father not in home 0.80 *** 0.04 0.67 *** 0.06 0.84 0.08

Observations 46346 46346 31282 15064

Pseudo R-squared 0.13 0.23 0.24 0.24

** p<0.01, *** p<0.001

135

Table 6-4: Relative risk ratios from multinomial logit models of children's activities

on residence, child, and family characteristics, National Family Health Survey 2005 -

2006

Idle vs.

Attending School

Working vs.

Attending School

Working and in school vs.

Attending School

Relative risk

ratio SE

Relative risk ratio

SE Relative risk

ratio SE

Model 1

Residence

Rural 1.09 0.06 0.61 *** 0.11 0.67 *** 0.09

[Reference=Slum] 1.00 1.00 1.00

Urban non-slum 0.58 *** 0.08 0.57 *** 0.16 1.02 0.11

Age 0.13 *** 0.04 0.44 *** 0.11 1.07 0.08

Age squared 1.09 *** 0.00 1.05 *** 0.01 1.00 0.00

Pseudo R-squared 0.112

Model 2

Residence

Rural 0.60 *** 0.06 0.30 *** 0.13 0.69 *** 0.09

Urban non-slum 0.90 0.08 1.17 0.16 1.08 0.11

Child characteristics

Age 0.09 *** 0.04 0.28 *** 0.12 1.12 0.08

Age squared 1.12 *** 0.00 1.07 *** 0.01 1.00 0.00

Female child 1.31 *** 0.03 1.03 0.07 0.78 *** 0.05

Family background

Wealth 0.52 *** 0.02 0.52 *** 0.05 1.12 0.05

Scheduled caste 1.34 *** 0.05 1.32 0.12 1.07 0.08

Scheduled tribe 1.28 *** 0.06 1.12 0.13 1.02 0.10

Other backwards caste 1.32 *** 0.04 1.43 *** 0.10 0.87 0.07

Muslim 1.63 *** 0.04 1.36 ** 0.10 0.63 *** 0.08

Christian 1.35 *** 0.08 0.75 0.23 0.44 *** 0.18

other 1.20 0.09 0.88 0.21 0.54 *** 0.18

Mother has no education 1.82 *** 0.05 1.84 *** 0.15 1.08 0.08 Mother has secondary education 0.96 0.06 0.78 0.20 1.17 0.09

Mother has higher education 1.27 0.13 0.15 1.04 0.69 0.20

Mother not in home 1.80 *** 0.09 1.68 ** 0.20 1.20 0.15

Mother works outside home 0.92 0.04 2.16 *** 0.09 1.35 *** 0.06

Number of kids in home 1.06 *** 0.01 1.26 *** 0.03 1.06 0.02

Father has no education 1.67 *** 0.05 1.52 *** 0.11 0.80 ** 0.08

Father has secondary education 0.96 0.05 0.54 *** 0.14 0.89 0.08

Father has higher education 0.90 0.10 0.27 ** 0.47 0.63 *** 0.14

Father not in home 1.42 *** 0.06 1.52 *** 0.12 0.56 *** 0.10

Psuedo R-squared 0.182

N=34006

** p<0.01, *** p<0.001

136

Figure 6-1: Proportion of children attending school by place of residence, National Family Health

Survey 2005-2006

137

Figure 6-2: Proportion of children working outside the home by place of residence, National

Family Health Survey 2005-2006

138

Figure 6-3: Proportion of children idle by place of residence, National Family Health Survey 2 005-

2006

139

Chapter 7 : Conclusions

“One person’s slum is another person’s community.”

May Hobbs

The sum of this dissertation paints a mixed picture of those in slums. These data

show no independent disadvantages faced by slum infants and children in terms of

mortality. This runs contrary to much of the conventional wisdom about slums, which

holds that slum residents face a battery of problems, in particular health problems related

to the poor living conditions of slums. While most of the current literature compares slum

residents to other urban residents, perhaps the more relevant comparison (illuminated

throughout this dissertation) is to rural residents, as rural residents are also poor and are

the potential migrants to slum areas. This works suggests that parents migrating for social

mobility (such as greater economic opportunities) does not put their children at greater

risk for mortality or morbidity, at least as measured in this dissertation. Rather, in so much

as cities produce greater incomes and education for the parents, moving to cities may

actually improve children’s mortality risks.

It is unclear from these results whether or not slum children face higher morbidity

risks than other urban children or rural children. The data suggest that slum children are

not at a higher risk of either chronic or acute malnutrition, a commonly used marker of

child well-being in epidemiology and public health. However, the data on acute symptoms

of illness, more closely akin to infectious diseases thought to be more common in slums, are

inconclusive. The results in fact suggest that maternal reports of children’s health are not

140

particularly reliable or valid, as it appears that maternal reports may be correlated with

socioeconomic factors and therefore not good indicators of actual child health. This is an

interesting finding, if not related to slums per se. Social scientists should be careful in

interpreting results from surveys that rely on maternal reports and in particularly should

take great care not to refer to these as measures of prevalence.

The results of the education and work present a less optimistic picture of slums.

Overall, India has achieved remarkable success in instituting universal primary education.

The problems of upper primary education and secondary education, however, have yet to

be solved. At age 11, the proportion of children attending school begins to drop rapidly, and

in slum areas the decrease is the most drastic. Slum children are also at a greater risk for

working outside the home. While I cannot fully test the reasons behind slum children’s

drop in school attendance, I contend that this is not because parents or children do not

value education, but more likely there is a deficit of public resources in the form of

secondary schools and high quality education as well as a wealth of (mostly informal)

economic activities that may entice older children into work and away from school. Further

research should examine both the availability and quality of upper primary (grades 6

through 9) schools and high schools for slum residents, the economic opportunities of

older children, and how to best encourage continued schooling among slum children.

Returning to the definition of slums

We must return to the question of defining and measuring slums. As discussed at

length in chapter 3, the current definitions of slums in India and around the world are often

vague or poorly measured. This is no different in the National Family Health Survey data or

the Indian census data, both of which use relatively ambiguous definitions that leave much

141

up to the interpretation of the census enumerator or surveyor. Such an ambiguously

defined and measured concept is likely to introduce a considerable amount of noise into

the data. It is possible that the noise in the data cloud some of the true impacts that slums

have on residents, such as the effects on health and mortality reported here.

As some of the descriptive statistics from the second and third chapters show, slums

are not uniform even on basic housing characteristics. Qualitative work by others as well as

my own fieldwork also indicate that the term “slums” incorporates a wide variety of

locations that are in many ways not homogeneous. Slums vary considerable in the size of

population, infrastructure, ethnic and religious composition, legality, and age, to name just

a few of the dimensions of variation. Instead of simply finding a definition of slums to use,

researchers and policy makers should also continue to measure various aspects of housing

conditions and neighborhoods in order to determine exactly what aspects of slum life

produce what kinds of disadvantage. In this way, the most information possible can be used

to target resources appropriately.

Why stay in slums?

Given the overall poor living conditions in slums, one may wonder why slums

residents remain. While often economic circumstances may keep residents in slums, as

they may not be able to afford to move to a better neighborhood, economic necessity is not

the entire story. There are usually great protests from slum residents against

redevelopment or relocation schemes that would in the end give slum residents much nicer

living conditions, often in multistory, modern apartment buildings. There is great interest

among policymakers and government officials in why these plans at best fail to attract any

interest from slum residents and at worst lead to years long court battles between slums

142

residents and the government to keep the slum as it is. This question is a large one that

should be the subject of rigorous research, and I can speak to it only tangentially through

this work.

In my focus groups with women in slums, the women were quick to enumerate the

many atrocities of slum life. However, when asked about if there was any positive side to

living in slums, they all gave the same answer: the community in their slum was very

strong. In one slum where I conducted a focus group, some residents had been in the slum

for over 40 years. The residents spoke of how they were truly had unity as a community, no

matter how difficult their lives may be. These strong social ties made the slum not only

bearable, but a cherished home to many of the residents. Despite the poor infrastructure,

ignorance from the local government, and concerns about their children, the women who

had once lived in rural areas were adamant that they would not want to return there, but

stay in their slum, their home.

Additionally many slum residents had livelihoods that were based in or near the

slum, be it small scale factory type work like that of bidi (cigarette) rollers who work from

their own homes (Agarwala 2006) or as domestic work in the nearby, richer residential

colonies. Relocation plans that would move slum residents were not welcome, even if the

move would provide a clean, modern apartment or ownership of a parcel of land. The social

and economic ties that bind residents to slums suggest that government efforts to simply

wipe out slums by demolition of redevelopment (the most popular two options) will not be

successful. These social and economic aspects of slums deserve considerable more

research, both qualitative and quantitative.

143

Future directions

A great deal more work on slums is needed. There are likely hundreds of salient

research questions to be asked and answered, and more than can be enumerated here. I

specifically discus possible future directions for work on slums and neighborhood effect,

health consequences, economic integration of slums, stigmatization, and social capital in

slums.

Neighborhood effects. Underlying much of the discussion of slums is an idea that

slums as neighborhoods have effects on their residents. Work that can illuminate the

neighborhood effects of very poor urban neighborhoods in the developing world will best

shed light on how concentrated poverty and affluence is affecting the human population

(Massey 1996). In addition, as discussed in chapter 3, it may be that slums effects are truly

the effects of neighborhoods of concentrated poverty. Studying neighborhood effects in

developing world cities would give researchers and policymakers great insight into the

conditions and consequences of the rapidly growing populations of urban poor.

The health consequences of slum life. Though this dissertation provided a first

glimpse into the health consequences faced by slum residents, there are still a great many

questions left to be answered. Public discussions on slum improvements tend to focus on

health consequences and improvements. Epidemiological studies could further our

understanding of the specific, direct impacts of various housing conditions. For instance,

poor sanitation and sewage systems certainly will have health consequences, although at

the scale of national data such as those used in this study the impacts may be diluted in

noisy data. Targeted local studies could unpack the particular pieces of poor sanitation that

are detrimental. For instance, Buttenheim (2008) examined the impacts of child toileting

144

behavior as compared to adult toileting behavior in slums. Further studies like these could

provide a better scope for specific public health campaigns aimed at improving slum health.

In addition to sanitation, other studies on health may be undertaken to examine the

impacts of piped water. Some research has shown that in Indian cities even treated, piped

water is not free of fecal contaminants such as E. coli (Brick et al. 2004). Similarly, one

might expect that certain housing materials may matter more than others for health, such

as a dirt floor from which spills cannot be easily wiped up. This kind of specific research on

slums may provide the best information for public health campaigns that wish to improve

the health and living conditions of slum residents.

Economic activity and integration of slums. A set of pressing questions about slums

regards the employment and economic opportunities of slum residents. Slums are often

considered to be the first stopping place for rural migrants entering cities. Do slums

promote or hinder social mobility of these migrants? Additionally, further work needs to be

done on the employment and employment trajectories of slum residents. How

economically integrated are slum residents in the wider city? Additionally, ethnographic

evidence suggests that some slums have vibrant informal economies functioning within the

slum itself. Work is needed to understand the scope and strength of these slum economies.

These economies may help slum residents in providing local employment. However, as we

have seen in the education chapter, they may also in some ways disadvantage slum

residents by encouraging children to work instead of continue schooling. The results from

this dissertation suggest that economic status has large impacts on both health and

educational outcomes, and therefore the economic activity of slums and slum residents

145

may be key to understanding the place of slums in the broader metropolis and the impacts

of slums on their residents.

Slums and stigmatization. Several pieces of ethnographic research indicate that

slum residents face a disadvantage when seeking employment because of stigmas

associated with living in slums. This finding speaks to the broader stigmatization of slums

and the social meanings of slums that come to carry disadvantage for residents, above and

beyond simple living conditions. Further research could illuminate in what areas of social

these stigmas have consequences. For instance, do these stigmas affect social networks,

marriage markets, and school admissions? Work on the intersections of caste, poverty, and

slum stigmatization could shed light on how multiple dimensions of stigma interact. Given

that in slums the most disadvantaged groups such as former untouchable castes and

Muslims are overrepresented, there may be interesting and complicated relationships

between slum stigmas and other social stigmas.

Social networks, social capital, and slums. My fieldwork and interviews with Indian

officials indicated that while many slums are physically disadvantaged, there are very

strong communities and ties within slums. There is ethnographic evidence that slums have

very tight social networks, with a vibrant street life and very strong ties between residents

that may provide a form of social control (Bhatt 2000; Lobo and Biswaroop Das 2001).

Many slums contain pockets of residents that all come from the same state, region, or even

village. Given the importance of social networks in migration (Massey 1990), studying the

social networks of slum residents and migrants to urban slums could prove fruitful to

understanding the social world of contemporary slums. In addition to the impacts of social

146

capital on health, the broader literature on social capital suggests that the social capital in

slums could affect a variety of other outcomes for slum residents, such as education and

children’s welfare, safe neighborhoods, economic outcomes, health and happiness (Putnam

2000).

Some of this future work can be done using existing data sources, but new efforts to

collect data on developing world cities must be a central effort at future work. Because for

so long developing world poverty has been characterized by rural poor, often working in

the agricultural sector, much of national and international efforts at poverty alleviation

have focused on the rural poor. The time has come for a shift in our focus and efforts

towards cities and the new urban poor. One of the first parts of this effort should be

concerted work on collecting information at the local, national, and international level on

the world’s slums. Currently in most large cities in developing countries, each department

or municipal agency keeps its own database, rarely sharing data and often using different

standards, and the same is true at the national level (Montgomery et al. 2003).

Computerization of these records is relatively uncommon. Still, developing world cities do

have much data, and one step toward better data collection could be efforts to standardize,

computerize, and share these data within and across metropolitan regions and countries.

Our urban future

At the turn of the 20th century, only about 14 percent of the world’s residents were

urban. Around 1950, the figure stood at about 30 percent (Grauman 1976). The world we

live in today may be hardly recognizable to urbanites from that time, as by 2030 it is likely

that 80% of Latin American residents and over 50% of Asian and African residents will be

in cities. In India, the number is projected to be 67% (United Nations 2004). Slums are

147

growing in both number and size. These numbers are not meant to scare us about an

impending “urban apocalypse” (Angotti 2006) as some have called it. Rather, these

projections should encourage both researchers and policymakers to answer a call to

rigorous research and planning on what will happen given the world’s poor, urban future.

Additionally, both researchers and policymakers must be careful that efforts to improve the

lots of slum dwellers not be a paternalistic endeavor. Slum residents around the world have

clearly demonstrated their entrepreneurialism in creating vast informal economies based

in some of the poorest communities in the world. They have demonstrated their social

organization in the many movements of slum residents that have arisen for various causes,

from agitating for notification of Indian slums to organizing protests and lengthy legal

battles to stop the demolition of their slums. They have demonstrated the strength of their

communities in the social networks that have supported slum residents through these

efforts. And they further have demonstrated a strong will to survive in what are often

places at which researchers and policymakers cringe, but these people have made into

lasting homes. The best research will involve slum residents in the creation of knowledge

and policy regarding their situation.

Globalization is touching the far reaches of the urban world, and as globalization

continues slum residents will only be more tied into the world economy. This is already

true of many slums. For instance, the leather work mentioned of Dharavi is exported

around the globe. Our future is in many ways tied to the future of the urban giants of the

developing world, slums and all. Let us accept the research challenge to continue

discovering more about how these slums function and affect residents and us all.

148

Appendix A : Infant and Child death interaction models

Table A-1: Coefficients from discrete time logit models of infant death to 13 months

on residence, background, and residence interactions, National Family Health Survey

2005-2006

Simple Model Interactions Coefficient SE Coefficient SE Residence

Urban non-slum –0.04 0.10 2.65 2.67 Rural –0.17 0.08 6.04 3.77 [Reference = urban slum] 0.00 0.00

Family background Wealth –0.46 *** 0.04 -0.26 0.20 x rural -0.18 0.21 x urban non-slum -0.47 0.29 Mother has no education 0.16 ** 0.06 0.36 0.14 x rural -0.13 0.15 x urban non-slum -0.07 0.21 Mother has secondary education –0.43 *** 0.07 -0.33 0.17 x rural -0.07 0.18 x urban non-slum 0.02 0.24 Mother has higher education –1.03 *** 0.26 -0.89 0.60 x rural 0.11 0.66 x urban non-slum 0.30 0.66 Female headed home 0.11 0.06 0.36 ** 0.14 x rural -0.37 0.14 x urban non-slum 0.07 0.20 Mother’s age at birth –0.08 *** 0.01 -0.07 *** 0.02 x rural 0.00 0.02 x urban non-slum 0.00 0.02 Short preceding birth interval 0.52 *** 0.04 0.57 *** 0.11 x rural -0.03 0.12 x urban non-slum 0.07 0.16 Female child 0.03 0.04 0.04 0.10 x rural 0.04 0.11 x urban non-slum -0.05 0.15 Birth order 0.10 *** 0.02 0.14 ** 0.05 x rural -0.02 0.05 x urban non-slum 0.02 0.07 Scheduled Caste or tribe 0.14 0.06 0.34 0.14 x rural -0.28 0.15 x urban non-slum -0.24 0.21 Other backward castes 0.04 0.06 0.31 0.14

149

x rural -0.31 0.14 x urban non-slum -0.17 0.18 Muslim –0.27 *** 0.08 -0.33 0.14 x rural 0.18 0.15 x urban non-slum -0.04 0.20 Christian –0.10 0.15 0.06 0.34 x rural 0.07 0.35 x urban non-slum -0.10 0.57 Other religion –0.10 0.12 0.04 0.28 x rural -0.17 0.29 x urban non-slum -0.94 0.54

Housing characteristics Poor toilet 0.13 0.10 -0.03 0.12 x rural -0.06 0.24 x urban non-slum 0.22 0.16 Poor water 0.11 0.07 0.35 ** 0.13 x rural -0.30 0.14 x urban non-slum -0.31 0.17 Overcrowded home –0.34 *** 0.05 -0.32 0.17 x rural 0.00 0.18 x urban non-slum 0.15 0.23 Poor floor –0.01 0.06 -0.07 0.14 x rural 0.13 0.15 x urban non-slum -0.17 0.20 Poor roof 0.00 0.05 -0.12 0.12 x rural 0.18 0.13 x urban non-slum -0.04 0.20 Poor walls –0.02 0.06 0.35 0.17 x rural -0.48 ** 0.18 x urban non-slum -0.83 ** 0.31

N 1194244 1194244 Log likelihood Pseudo R2 0.04 0.04 *** p<0.001; ** p<0.01

150

Appendix B : Child health interaction models

Table B-1: Results from logisitic regressions of child sickness in two weeks prior to survey on residence, background,

and housing characteristics with residence interactions, National Family Health Survey 2005 -2006

Coughing Diarrhea Fever Main effects Interactions Main effects Interactions Main effects Interactions Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE Residence

Rural -0.07 0.06 -3.79 2.67 -0.06 0.09 4.48 3.75 0.06 0.07 -2.70 2.99 Urban non-slum -0.11 0.07 -1.29 3.91 0.03 0.09 2.75 5.23 -0.07 0.08 0.74 4.33 [Reference = slum]

Family background Wealth 0.01 0.04 -0.34 0.21 0.00 0.06 0.24 0.29 0.06 0.05 -0.13 0.23 x rural 0.37 0.21 -0.25 0.30 0.22 0.24 x urban non-slum 0.25 0.31 -0.15 0.41 -0.05 0.34 Mother has no education -0.21 *** 0.04 -0.33 0.17 -0.08 0.06 0.15 0.27 -0.14 *** 0.05 -0.18 0.19 x rural 0.13 0.18 -0.27 0.27 0.05 0.20 x urban non-slum 0.23 0.25 0.12 0.36 0.06 0.28 Mother has secondary education 0.04 0.04 0.05 0.16 0.09 0.06 0.55 0.25 0.07 0.05 0.15 0.18 x rural 0.00 0.17 -0.48 0.25 -0.09 0.19 x urban non-slum -0.04 0.23 -0.52 0.33 -0.09 0.26 Mother has higher education -0.04 0.08 0.76 ** 0.25 -0.05 0.11 0.04 0.41 -0.10 0.09 0.37 0.29 x rural -0.79 ** 0.27 -0.02 0.43 -0.50 0.31 x urban non-slum -1.04 ** 0.33 -0.20 0.49 -0.68 0.37 Female headed home 0.00 0.05 0.02 0.17 0.04 0.06 0.12 0.23 0.06 0.05 -0.12 0.20 x rural -0.01 0.18 -0.07 0.24 0.19 0.20 x urban non-slum 0.00 0.24 -0.29 0.33 0.27 0.26 Mother’s age at birth 0.00 0.00 0.01 0.02 -0.01 0.01 0.01 0.02 0.00 0.00 -0.01 0.02 x rural -0.01 0.02 -0.02 0.02 0.01 0.02 x urban non-slum -0.05 0.02 -0.02 0.03 0.02 0.02 Short preceding birth interval -0.09 0.04 -0.08 0.13 -0.08 0.05 -0.32 0.19 -0.06 0.04 -0.09 0.15 x rural 0.03 0.14 0.26 0.20 0.07 0.15 x urban non-slum -0.38 0.19 0.13 0.25 -0.42 0.21 Female child -0.09 ** 0.03 -0.11 0.10 -0.12 *** 0.04 -0.17 0.14 -0.09 ** 0.03 0.00 0.11

151

x rural 0.02 0.11 0.03 0.15 -0.12 0.12 x urban non-slum 0.07 0.14 0.30 0.19 0.04 0.15 Birth order -0.03 0.01 -0.02 0.05 0.04 0.02 0.04 0.07 0.01 0.01 0.00 0.05 x rural -0.02 0.05 0.00 0.07 0.00 0.05 x urban non-slum 0.14 0.07 0.07 0.09 0.06 0.07 Scheduled Caste -0.02 0.04 0.22 0.15 0.04 0.06 -0.22 0.21 0.04 0.05 -0.02 0.18 x rural -0.29 0.16 0.29 0.22 0.08 0.18 x urban non-slum -0.08 0.21 0.19 0.29 -0.12 0.24 Scheduled tribe -0.27 *** 0.05 -0.28 0.36 0.01 0.07 -0.17 0.46 -0.16 ** 0.06 -0.64 0.49 x rural 0.02 0.36 0.20 0.47 0.50 0.49 x urban non-slum -0.72 0.59 -0.26 0.71 0.41 0.64 Other backward castes -0.02 0.04 0.29 0.13 0.13 ** 0.05 -0.06 0.17 0.03 0.04 0.36 ** 0.14 x rural -0.35 ** 0.13 0.19 0.18 -0.37 0.15 x urban non-slum -0.20 0.17 0.25 0.22 -0.21 0.18 Muslim 0.31 *** 0.04 0.30 0.13 0.14 ** 0.05 -0.21 0.18 0.42 *** 0.04 0.42 ** 0.14 x rural 0.06 0.14 0.39 0.19 0.04 0.15 x urban non-slum -0.37 0.18 0.21 0.25 -0.25 0.20 Christian 0.14 0.06 0.42 0.30 -0.19 0.08 0.09 0.42 -0.01 0.07 0.22 0.34 x rural -0.32 0.31 -0.27 0.43 -0.28 0.34 x urban non-slum -0.21 0.44 -1.23 0.84 -0.05 0.48 Other religion 0.08 0.07 0.18 0.33 0.21 0.08 -0.23 0.55 0.06 0.07 -0.13 0.45 x rural -0.15 0.34 0.42 0.55 0.09 0.46 x urban non-slum 0.30 0.40 0.92 0.61 1.15 0.50

Housing characteristics Poor toilet 0.36 *** 0.07 0.42 *** 0.12 0.43 *** 0.09 0.73 *** 0.17 0.28 *** 0.07 0.35 ** 0.14 x rural -0.20 0.23 -0.38 0.33 -0.03 0.26 x urban non-slum -0.22 0.16 -0.53 0.22 -0.20 0.18 Poor water 0.07 0.04 0.19 0.12 -0.09 0.05 0.01 0.17 0.06 0.05 0.22 0.14 x rural -0.12 0.13 -0.12 0.18 -0.16 0.15 x urban non-slum -0.27 0.16 -0.09 0.22 -0.33 0.18 Overcrowded home -0.07 0.05 0.29 0.25 0.04 0.06 0.64 0.41 -0.02 0.05 0.14 0.28 x rural -0.35 0.26 -0.64 0.41 -0.15 0.28 x urban non-slum -0.58 0.30 -0.31 0.47 -0.24 0.33 Poor floor 0.01 0.04 -0.28 0.15 -0.09 0.05 0.25 0.20 0.05 0.04 -0.09 0.16 x rural 0.35 0.15 -0.34 0.21 0.17 0.17 x urban non-slum 0.05 0.21 -0.52 0.27 0.08 0.22 Poor roof -0.05 0.04 -0.40 ** 0.13 0.01 0.05 -0.28 0.19 0.04 0.04 -0.11 0.14 x rural 0.38 ** 0.14 0.33 0.19 0.19 0.15 x urban non-slum 0.40 0.20 0.09 0.28 -0.13 0.23

152

Poor walls 0.17 *** 0.04 0.13 0.20 0.20 *** 0.05 0.07 0.29 0.08 0.04 0.09 0.23 x rural 0.02 0.21 0.12 0.30 -0.01 0.24 x urban non-slum -0.01 0.32 0.41 0.41 -0.31 0.37

N 36237 36237 36253 36253 36241 36241

Pseudo R2 0.02 0.03 0.01 0.01 0.01 0.01 *** p<0.001; ** p<0.01 Note: All models control for the month of the year the interview took place

153

Table B-2: Results from logistic regressions of child wasting and stunting on residence, background, and housing

characteristics with residence interactions, National Family Health Survey 2005-2006

Stunting Wasting Main effects Interactions Main effects Interactions

Coef. SE Coef. SE Coef. SE Coef. SE Residence

Rural 0.09 0.05 1.29 2.18 0.04 0.07 -1.13 3.02 Urban non-slum 0.04 0.06 6.06 3.08 0.06 0.08 -7.46 4.28 [Reference = slum]

Family background Wealth -0.21 *** 0.03 -0.13 0.17 -0.30 *** 0.04 -0.38 0.24 x rural -0.07 0.17 0.07 0.24 x urban non-slum -0.49 0.24 0.63 0.34 Mother has no education 0.03 0.03 0.28 0.14 0.03 0.04 0.11 0.21 x rural -0.26 0.14 -0.09 0.21 x urban non-slum -0.27 0.19 -0.23 0.28 Mother has secondary education -0.19 *** 0.03 0.06 0.13 -0.08 0.05 0.43 0.20 x rural -0.28 0.14 -0.57 ** 0.20 x urban non-slum -0.11 0.18 -0.58 0.26 Mother has higher education -0.79 *** 0.07 -0.71 ** 0.26 -0.07 0.09 0.54 0.30 x rural -0.08 0.28 -0.58 0.32 x urban non-slum 0.24 0.31 -1.00 ** 0.37 Female headed home -0.01 0.04 -0.08 0.14 0.06 0.05 -0.02 0.20 x rural 0.08 0.15 0.10 0.20 x urban non-slum 0.09 0.19 0.03 0.26 Mother’s age at birth -0.02 *** 0.00 -0.02 0.01 0.01 *** 0.00 0.00 0.02 x rural 0.01 0.01 0.02 0.02 x urban non-slum -0.01 0.02 0.02 0.02 Short preceding birth interval 0.12 *** 0.03 0.22 0.10 -0.10 ** 0.04 -0.12 0.15 x rural -0.12 0.10 0.02 0.15 x urban non-slum -0.04 0.14 0.14 0.20 Female child -0.04 0.02 -0.17 0.08 -0.10 *** 0.03 -0.08 0.12 x rural 0.13 0.09 -0.01 0.12 x urban non-slum 0.23 0.11 -0.14 0.15 Birth order 0.06 *** 0.01 0.05 0.04 -0.01 0.01 0.01 0.06 x rural 0.00 0.04 -0.02 0.06 x urban non-slum 0.07 0.05 -0.06 0.08 Scheduled Caste 0.15 *** 0.03 0.14 0.12 0.07 0.05 -0.43 0.18

154

x rural 0.01 0.13 0.52 ** 0.19 x urban non-slum -0.02 0.17 0.66 ** 0.23 Scheduled tribe 0.03 0.04 -0.10 0.26 0.27 *** 0.05 0.23 0.33 x rural 0.14 0.27 0.06 0.33 x urban non-slum 0.01 0.37 -0.01 0.46 Other backward castes 0.11 *** 0.03 0.09 0.11 0.09 0.04 0.18 0.14 x rural 0.02 0.11 -0.09 0.15 x urban non-slum -0.04 0.14 -0.17 0.18 Muslim -0.02 0.03 -0.08 0.11 -0.14 ** 0.04 -0.17 0.15 x rural 0.06 0.11 0.06 0.16 x urban non-slum 0.17 0.14 -0.04 0.20 Christian -0.05 0.05 -0.26 0.29 -0.53 *** 0.07 0.11 0.34 x rural 0.22 0.29 -0.67 0.35 x urban non-slum -0.10 0.43 -0.10 0.48 Other religion -0.10 0.05 0.34 0.26 -0.16 0.07 0.14 0.40 x rural -0.47 0.27 -0.32 0.41 x urban non-slum -0.25 0.34 -0.27 0.48

Housing characteristics Poor toilet 0.03 0.05 0.07 0.10 0.00 0.07 0.24 0.14 x rural -0.36 0.18 -0.08 0.26 x urban non-slum 0.02 0.13 -0.37 0.18 Poor water 0.02 0.03 0.02 0.10 0.00 0.05 -0.05 0.14 x rural 0.04 0.11 -0.02 0.15 x urban non-slum -0.18 0.13 0.42 0.18 Overcrowded home 0.34 *** 0.04 0.36 0.23 0.03 0.05 -0.12 0.25 x rural -0.04 0.23 0.19 0.26 x urban non-slum 0.19 0.28 -0.20 0.30 Poor floor 0.22 *** 0.03 0.07 0.11 -0.07 0.04 -0.32 0.17 x rural 0.18 0.12 0.25 0.18 x urban non-slum -0.10 0.16 0.40 0.22 Poor roof -0.05 0.03 -0.05 0.10 -0.08 0.04 0.11 0.14 x rural -0.01 0.11 -0.20 0.14 x urban non-slum 0.14 0.15 -0.07 0.21 Poor walls -0.02 0.03 0.16 0.17 0.13 ** 0.04 0.46 0.22 x rural -0.20 0.17 -0.35 0.22 x urban non-slum 0.00 0.24 -0.16 0.32

N 39,330 39,330 39,330 39,330 Pseudo R2 0.03 0.03 0.01 0.02 *** p<0.001; ** p<0.01

155

Appendix C : Children’s activities interaction models

Table C-1: Logistic regression of school attendance on location and background characteristics and interactions between location

and background characteristics, National Family Health Survey 2005-2006

All ages All ages: Interactions

Ages 5-13 Ages 5-13: Interactions

Ages 14-18 Ages 14-18: Interactions

OR SE OR SE OR SE OR SE OR SE OR SE

Residence Rural 0.62 *** 0.05 5.76 *** 1.72 0.47 *** 0.06 2.90 2.24 0.94 *** 0.08 7.87 13.52 Urban non-slum 0.17 ** 0.06 -13.24 *** 2.67 0.03 0.08 -18.38 *** 3.73 0.25 ** 0.09 24.51 18.46

Child Characteristics Age 1.55 *** 0.02 1.32 *** 0.08 2.72 *** 0.05 2.63 *** 0.18 -3.27 *** 0.44 -4.16 ** 1.60 x rural 0.25 0.08 0.04 0.19 1.52 1.67 x urban non-slum 0.44 *** 0.11 1.04 *** 0.28 -3.67 2.30 Age squared -0.07 *** 0.00 -0.06 *** 0.00 -0.13 *** 0.00 -0.13 *** 0.01 0.09 *** 0.01 0.12 0.05 x rural -0.01 ** 0.00 0.00 0.01 -0.06 0.05 x urban non-slum -0.01 ** 0.00 -0.05 ** 0.02 0.12 0.07 Female child -0.35 *** 0.02 0.12 0.09 -0.26 *** 0.03 0.09 0.12 -0.52 *** 0.04 0.11 0.14 x rural -0.55 *** 0.09 -0.41 *** 0.12 -0.77 *** 0.15 x urban non-slum 0.01 0.12 -0.01 0.17 -0.02 0.20

Family Background Wealth 0.73 *** 0.02 1.23 *** 0.13 0.62 *** 0.02 0.85 *** 0.16 1.08 *** 0.04 2.44 *** 0.25 x rural -0.52 *** 0.13 -0.24 0.16 -1.41 *** 0.25 x urban non-slum 0.81 *** 0.21 1.06 *** 0.27 0.33 0.36 Scheduled caste -0.33 *** 0.04 -0.18 0.13 -0.30 *** 0.05 -0.17 0.17 -0.35 *** 0.06 -0.44 0.21 x rural -0.16 0.13 -0.16 0.18 0.11 0.22 x urban non-slum -0.19 0.19 0.17 0.27 -0.07 0.29 Scheduled tribe -0.21 *** 0.04 -1.05 *** 0.22 -0.26 *** 0.06 -1.24 *** 0.26 -0.04 0.07 -0.94 0.46 x rural 0.87 *** 0.22 1.00 *** 0.27 0.91 0.46 x urban non-slum 0.60 0.38 0.82 0.50 0.85 0.68 Other backwards caste -0.31 *** 0.03 -0.27 0.12 -0.29 *** 0.04 -0.18 0.16 -0.29 *** 0.05 -0.46 0.20 x rural -0.07 0.12 -0.14 0.16 0.15 0.20 x urban non-slum 0.28 0.17 0.15 0.23 0.58 0.28

156

Muslim -0.48 *** 0.03 -0.45 *** 0.11 -0.49 *** 0.04 -0.44 ** 0.15 -0.53 *** 0.05 -0.73 *** 0.19 x rural -0.03 0.12 -0.05 0.15 0.23 0.20 x urban non-slum -0.30 0.16 -0.31 0.22 0.07 0.26 Christian -0.01 0.06 0.54 0.45 -0.32 *** 0.08 -0.36 *** 0.08 0.56 *** 0.11 0.53 *** 0.11 x rural -0.63 0.46 x urban non-slum 0.28 0.62 Other non-Hindu religion 0.06 0.07 2.45 1.20 -0.23 ** 0.09 -0.27 ** 0.09 0.60 *** 0.12 0.55 *** 0.12 x rural -2.46 1.20 x urban non-slum -2.52 1.26 Mother has no education -0.52 *** 0.04 -0.42 0.17 -0.61 *** 0.05 -0.89 *** 0.16 -0.44 *** 0.08 -0.42 0.21 x rural -0.13 0.18 0.29 0.16 -0.03 0.22 x urban non-slum 0.20 0.25 0.39 0.23 -0.03 0.29 Mother has secondary education 0.10 0.05 0.41 0.19 0.03 0.06 0.00 0.07 0.48 *** 0.10 0.44 *** 0.11 x rural -0.42 0.20 x urban non-slum 0.18 0.27 Mother has higher education -0.08 0.11 0.28 0.34 -0.18 0.13 -0.40 ** 0.14 3.19 ** 1.01 2.81 ** 1.02 x rural -0.81 0.38 x urban non-slum -0.16 0.43 Mother works outside home -0.06 0.03 -0.23 0.11 0.03 0.04 0.04 0.04 -0.32 *** 0.06 -0.31 *** 0.06 x rural 0.19 0.12 x urban non-slum 0.40 0.17 Mother not in home -0.24 *** 0.06 0.09 0.25 -0.59 *** 0.09 -0.91 ** 0.32 -0.76 *** 0.11 -0.48 0.37 x rural -0.34 0.26 0.41 0.33 -0.27 0.38 x urban non-slum -0.30 0.36 -0.84 0.46 -0.46 0.50 Number of kids in home -0.08 *** 0.01 -0.07 0.04 -0.06 *** 0.01 -0.01 0.05 -0.08 *** 0.02 -0.07 0.06 x rural -0.01 0.04 -0.07 0.05 -0.01 0.06 x urban non-slum 0.03 0.05 0.01 0.07 -0.03 0.09 Father has no education -0.39 *** 0.04 -0.54 *** 0.15 -0.52 *** 0.05 -0.55 ** 0.20 -0.31 *** 0.07 -0.76 ** 0.26 x rural 0.14 0.15 0.03 0.20 0.48 0.27 x urban non-slum 0.28 0.23 0.28 0.31 0.75 0.40 Father has secondary education 0.18 *** 0.04 -0.21 0.16 0.07 0.05 -0.34 0.20 0.42 0.07 0.08 0.27 x rural 0.42 ** 0.16 0.43 0.21 0.36 0.28 x urban non-slum 0.37 0.23 0.56 0.31 0.59 0.39 Father has higher education 0.38 *** 0.08 -0.22 0.27 0.07 0.10 -0.33 0.31 1.44 0.23 0.49 0.66 x rural 0.52 0.29 0.35 0.33 0.84 0.72 x urban non-slum 0.47 0.36 0.55 0.42 1.21 0.88 Father not in home -0.22 *** 0.04 -0.64 *** 0.18 -0.40 *** 0.06 -0.61 0.24 -0.18 0.08 -0.91 ** 0.29

157

x rural 0.45 0.18 0.23 0.25 0.83 ** 0.30 x urban non-slum 0.30 0.28 0.27 0.39 0.56 0.44

Observations 46346 46346 31282 31282 15064 15064 Pseudo R-squared 0.23 0.24 0.24 0.24 0.24 0.25

** p<0.01, *** p<0.001 Note: Interactions that are missing in this table could not be calculated in the model, usually due to collinearity

158

Table C-2: Multinomial regression of children's activities on location and background characteristics with

interactions between location and background characteristics, National Family Health Survey 2005-2006

Idle vs.

Attending School

Working vs.

Attending School

Working and in school vs.

Attending School

Relative risk ratio SE Relative risk ratio SE Relative risk ratio SE

Residence

Rural -1.41 2.3 -14.02 *** 3.77 -2.78 3.68 Urban non-slum 20.01 *** 3.84 5.96 6.25 -6.99 5.32

Child characteristics

Age -2.54 *** 0.16 -1.28 *** 0.35 0.02 0.27 x rural 0.19 0.17 0.21 0.37 0.02 0.29 x urban non-slum -0.75 ** 0.24 -1.21 0.5 0.51 0.36 Age squared 0.12 *** 0.01 0.07 *** 0.02 0.00 0.01 x rural -0.01 0.01 -0.01 0.02 0.00 0.01 x urban non-slum 0.03 0.01 0.06 0.02 -0.03 0.02 Female child -0.11 0.12 -0.43 0.23 -0.54 ** 0.18 x rural 0.43 *** 0.12 0.59 0.25 0.30 0.19 x urban non-slum 0.17 0.17 -0.39 0.35 0.40 0.23

Family background

Wealth -0.67 *** 0.17 -1.55 *** 0.26 -0.15 0.27 x rural 0.04 0.17 0.95 *** 0.27 0.22 0.27 x urban non-slum -1.27 *** 0.29 0.10 0.46 0.45 0.40 Scheduled caste 0.27 0.17 0.38 0.35 0.48 0.24 x rural 0.05 0.18 -0.03 0.38 -0.55 0.25 x urban non-slum -0.15 0.28 -0.54 0.51 -0.07 0.31 Scheduled tribe 1.32 *** 0.27 0.83 0.50 -0.82 0.64 x rural -1.10 *** 0.27 -0.70 0.52 0.85 0.65 x urban non-slum -0.96 0.49 -1.23 0.95 0.05 0.99 Other backwards caste 0.15 0.16 0.75 0.31 -0.03 0.24

159

x rural 0.18 0.16 -0.36 0.33 -0.11 0.26 x urban non-slum -0.23 0.24 -0.88 0.45 -0.31 0.32 Muslim 0.50 *** 0.15 0.58 0.30 -0.81 ** 0.25 x rural -0.01 0.16 -0.23 0.32 0.54 0.27 x urban non-slum 0.25 0.23 -0.56 0.44 -0.46 0.37 Christian -0.74 0.80 -20.32 *** 1.06 -20.44 *** 0.18 x rural 1.13 0.81 x urban non-slum -0.51 1.11 other -21.59 *** 0.53 -21.11 *** 1.11 -19.66 *** 0.42 x rural 21.85 *** 0.53 21.01 *** 1.13 18.95 *** 0.47 x urban non-slum Mother has no education 0.70 *** 0.23 0.25 0.47 0.22 0.31 x rural -0.08 0.24 0.39 0.50 -0.23 0.32 x urban non-slum -0.48 0.34 0.26 0.65 0.27 0.43 Mother has secondary education -0.30 0.25 -1.93 0.84 -0.16 0.33 x rural 0.36 0.26 1.83 0.87 0.40 0.35 x urban non-slum -0.19 0.35 1.70 0.97 0.37 0.45 Mother has higher education -0.24 0.40 -22.19 *** 1.04 -19.67 *** 0.39 x rural 0.85 0.44 19.80 *** 0.48 x urban non-slum 0.28 0.52 Mother works outside home 0.23 0.14 0.16 0.29 -0.52 0.24 x rural -0.33 0.15 0.70 0.31 0.99 *** 0.25 x urban non-slum -0.69 ** 0.23 0.41 0.43 -0.15 0.32 Mother not in home 0.81 0.36 -0.01 0.60 0.21 0.46 x rural -0.27 0.38 0.53 0.64 0.03 0.49 x urban non-slum 0.61 0.52 1.02 0.86 -1.22 0.95 Number of kids in home 0.06 0.05 0.25 ** 0.09 0.22 *** 0.06 x rural 0.00 0.05 -0.01 0.09 -0.16 0.07 x urban non-slum -0.09 0.07 -0.09 0.14 -0.29 ** 0.09 Father has no education 0.53 ** 0.19 1.22 ** 0.43 0.58 0.30 x rural -0.02 0.20 -0.88 0.45 -0.87 ** 0.32 x urban non-slum -0.06 0.32 -1.06 0.62 -0.92 0.44 Father has secondary education 0.42 0.20 -0.44 0.58 0.53 0.30

160

x rural -0.49 0.21 -0.30 0.60 -0.73 0.31 x urban non-slum -0.44 0.31 0.24 0.72 -0.66 0.39 Father has higher education 0.46 0.31 -20.56 *** 1.10 -0.74 0.66 x rural -0.55 0.33 19.54 *** 1.22 0.45 0.68 x urban non-slum -0.39 0.45 -0.10 0.75 Father not in home 0.59 0.24 1.50 ** 0.47 0.22 0.39 x rural -0.26 0.25 -1.21 0.49 -0.83 0.41 x urban non-slum 0.03 0.40 -0.74 0.68 -0.89 0.59

Psuedo R-squared 0.191

N=34006

** p<0.01, *** p<0.001

Note: Interactions that are missing in this table could not be calculated in the maximum likelihood model, usually due to collinearity

161

REFERENCES

Adlakha, A. L., and C. M. Suchindran. 1985. “Factors affecting infant and child mortality.”

Journal of Biosocial Science 17:481-496. Agarwala, Rina. 2006. From work to welfare : informal workers' organizations and the state

in India. Princeton, NJ: Princeton University Dissertation Abstracts. Agnihotri, Pushpa. 1994. Poverty amidst prosperity: Survey of slums. Delhi: M.D.

Publications. Ahmed, Rokeya. 2005. Shifting Millions from Open Defecation to Hygienic Practices.

http://www.adb.org/Documents/Events/2005/Sanitation-Wastewater-Management/paper-ahmed.pdf (Accessed February 24, 2009).

Ainsworth, James W. 2002. “Why Does It Take a Village? The Mediation of Neighborhood

Effects on Educational Achievement.” Social Forces 81:117-152. Anderson, Elijah. 2003. A Place on the Corner. 2nd ed. Chicago: University of Chicago Press. Anderson, Elijah. 1999. Code of the Street: Decency, Violence, and the Moral Life of the

Inner City. 1st ed. New York: W.W Norton. Anderson, Elijah. 1990. Streetwise: Race, Class, and Change in an Urban Community.

Chicago: University of Chicago Press. Angotti, Tom. 2006. “Apocalyptic anti-urbanism: Mike Davis and his planet of slums.”

International Journal of Urban and Regional Research 30:961-967. Arcia, Emily. 1998. “Latino parents' perception of their children's health status.” Social

Science & Medicine 46:1271-1274. Attewell, Paul, and S. Madheswaran. “The Price of Globalization: Wage Penalties and Caste

Inequality in 'Liberal' India.” in Discrimination in an Unequal World, edited by Miguel Centeno and Katherine S. Newman. New York: Oxford University Press.

Bacolod, Marigee P., and Priya Ranjan. 2008. “Why Children Work, Attend School, or Stay

Idle: The Roles of Ability and Household Wealth.” Economic Development and Cultural Change 56:791-828.

Banerjee, Banashree. 2002. “Security of Tenure in Indian Cities.” in Holding Their Ground,

edited by Alain Durand-Lasserve and Lauren Royston. London: Earthscan Publications.

162

Barker, D. J. P. 1998. Mothers, Babies and Health in Later Life. 2nd ed. Churchill Livingstone. Basu, Kaushik. 1999. “Child Labor: Cause, Consequence, and Cure, with Remarks on

International Labor Standards.” Journal of Economic Literature 37:1083-1119. Basu, Kaushik, and Zafiris Tzannatos. 2003. “The Global Child Labor Problem: What Do We

Know and What Can We Do?.” World Bank Econ Rev 17:147-173. Basu, Kaushik, and Pham Hoang Van. 1998. “The Economics of Child Labor.” The American

Economic Review 88:412-427. Behm, H., and D. A. Primante. 1977. La mortalidad en los primeros años de vida en países

de la América Latina. Centro Latinoamericano de Demografía. Behm, H., and J. Vallin. 1982. “Mortality differentials among human groups.” Pp. 11-37 in

Samuel H. Preston (ed.), Biological and social aspects of mortality and the length of life. Liege.

Beinhocker, Eric D., Diana Farrell, and Adil S. Zainulbhai. 2007. “Tracking the growth of

India's middle clas.” McKinsey Quarterly 2007:50-61. Bhatt, Manju. 2000. Slums and metropolis : a sociological study. New Delhi: Rajat

Publications. Bicego, G. T., and J. T. Boerma. 1991. “Maternal education and child survival: a comparative

analysis of DHS data.” in Demographic and Health Surveys World Conference, vol. 5. Biggeri, Mario, Lorenzo Guarcello, Scott Lyon, and Furio Camillo Rosati. 2003. The puzzle of

idle children: neither in school nor performing economic activity: Evidence from six countries. Understanding Children’s Work. An Inter-Agency Research Cooperation Project. ILO, Unicef and Work Bank.

Blackwell, Debra L., Mark D. Hayward, and Eileen M. Crimmins. 2001. “Does childhood

health affect chronic morbidity in later life?.” Social Science & Medicine 52:1269-1284.

Bloem, Martin W. Panagides. 2003. Health & Nutritional Surveillance for Development

Moench-Pfanner, Regina. Singapore: Helen Keller International. Boerma, J. T., A. E. Sommerfelt, and S. O. Rutstein. 1991. Childhood Morbidity and

Treatment Patterns. Columbia, MD: Institute for Resource Development/Macro International, Inc.

Bourgois, Philippe I. 1995. In Search of Respect: Selling Crack in El Barrio. Cambridge:

Cambridge University Press.

163

Breen, Richard, and John H Goldthorpe. 1997. “Explaining Educational Differentials:

Towards a Formal Rational Action Theory.” Rationality And Society 9:275-305. Brick, Thomas et al. 2004. “Water contamination in urban south India: household storage

practices and their implications for water safety and enteric infections.” International Journal of Hygiene and Environmental Health 207:473-480.

Brockerhoff, Martin. 1995. “Child survival in big cities: The disadvantages of migrants.”

Social Science & Medicine 40:1371-1383. Brockerhoff, Martin. 1990. “Rural-to-urban migration and child survival in Senegal.”

Demography 27:601-616. Brockerhoff, Martin, and Ellen Brennan. 1998. “The Poverty of Cities in Developing

Regions.” Population and Development Review 24:75-114. Brown, D. K., A. V. Deardorff, and R. M. Stern. 2003. “Child Labor: Theory, Evidence, and

Policy.” in International Labor Standards: History, Theory, and Policy Options. Malden, MA: Blackwell.

Brunner, E et al. 1996. “Childhood social circumstances and psychosocial and behavioural

factors as determinants of plasma fibrinogen.” Lancet 347:1008-13. Buchmann, Claudia, and D. Brakewood. 2000. “Labor Structures and School Enrollments in

Developing Societies: Thailand and Kenya Compared.” Comparative Education Review 44:175-204.

Buchmann, Claudia, and Emily Hannum. 2001. “Education and Stratification in Developing

Countries: A Review of Theories and Research.” Annual Review of Sociology 27:77-102.

Buck, C, and H Simpson. 1982. “Infant diarrhoea and subsequent mortality from heart

disease and cancer.” Journal of Epidemiology and Community Health 36:27-30. Buttenheim, Alison M. 2008. “The Sanitation Environment in South Asian Slums:

Implications for Child Health.” Population and Environment 30:26-47. CABE Committee. 2004. Universalisation of Secondary Education. New Delhi: Department

of Secondary and Higher Education, Ministry of Human Resource Development, Government of India http://www.educationforallinindia.com/universalisation%20of%20secondary%20education%20report%20of%20CABE%20Commuitee.pdf (Accessed April 23, 2009).

164

Cai, L., and V. Chongsuvivatwong. 2006. “Rural-urban differentials of premature mortality burden in south-west China.” International Journal for Equity in Health 5:13.

Caldwell, J. C. 1979. “Education as a Factor in Mortality Decline An Examination of Nigerian

Data.” Population Studies 33:395-413. Caldwell, J. C., and P. F. McDonald. 1982. “Influence of maternal education on infant and

child mortality: levels and causes..” Health Policy and Education 2:251-267. Casterline, John B., E. C. Cooksey, and A. F. Ismail. 1989. “Household income and child

survival in Egypt.” Demography 26:15-35. Census of India 2001. n.d. Metadata and Brief Highlights on Slum Population. New Delhi:

Census of India. Chaudhury, Nazmul, Jeffrey S. Hammer, and Atul Pokharel. 2009. “First things First:

Infectious disease, child mortality, and the poor in India 1992-2005.” Proceedings of the World Bank Annual Bank Conference in Development Economics, Captetown, South Africa.

Chernichovsky, Dov. 1985. “Socioeconomic and Demographic Aspects of School Enrollment

and Attendance in Rural Botswana.” Economic Development and Cultural Change 33:319.

Clark, Kenneth Bancroft. 1989. Dark Ghetto: Dilemmas of Social Power. 1st ed. Middletown,

Conn: Wesleyan University Press. Cleland, John, George Bicego, and Greg Fegan. 1992. “Socioeconomic inequalities in

childhood mortality: the 1970s to the 1980s.” Health Transition Review 2. Cleland, John G., and Jerome K. van Ginneken. 1988. “Maternal education and child survival

in developing countries: The search for pathways of influence.” Social Science & Medicine 27:1357-1368.

Colclough, C., P. Rose, and M. Tembon. 2000. “Gender inequalities in primary schooling The

roles of poverty and adverse cultural practice.” International Journal of Educational Development 20:5-27.

Costa, Dora L. 2000. “Understanding the Twentieth-Century Decline in Chronic Conditions

among Older Men.” Demography 37:53-72. Das, Jishnu, and Jeffrey Hammer. 2007. “Location, Location, Location: Residence, Wealth,

And The Quality of Medical Care In Delhi, India.” Health Affairs 338-351. Das, Jishnu, and Jeffrey S. Hammer. 2008. “Chronic illness, information and behavior

change: An example from Delhi.”.

165

Davis, Mike. 2006. Planet of Slums. New York: Verso. De, Anuradha, and Jean Dreze. 1999. Public Report on Basic Education in India. New York:

Oxford University Press. Deaton, A., and V. Kozel. 2005. “Data and Dogma: The Great Indian Poverty Debate.” The

World Bank Research Observer 20:177-199. Deaton, Angus. 2009. “National Sample Survey Response Rates.”. Deaton, Angus, and Valerie Kozel. 2005. The Great Indian Poverty Debate. Delhi: Macmillan

India Ltd. Du Bois, W. E. B. 2007. The Philadelphia Negro: A Social Study. New York, NY: Oxford

University Press. Duncan, Greg J. 1984. Years of Poverty, Years of Plenty: The Changing Economic Fortunes of

American Workers and Families. Ann Arbor, MI: Survey Research Center, Institute for Social Research, University of Michigan.

Duneier, Mitchell. 1999. Sidewalk. 1st ed. New York: Farrar, Straus and Giroux. Duneier, Mitchell. 1992. Slim's Table: Race, Respectability, and Masculinity. Chicago:

University of Chicago Press. Dunyo, Samuel K, Kwadwo A Koram, and Francis K Nkrumah. 1997. “Fever in Africa and

WHO recommendation.” The Lancet 350:1550. Edin, Kathryn, and Laura Lein. 1997. Making Ends Meet: How Single Mothers Survive

Welfare and Low-Wage Work. New York: Russell Sage Foundation. Edmonds, Eric V. 2005. “Does Child Labor Decline with Improving Economic Status?.” The

Journal of Human Resources 40:77-99. Edmonds, Eric V., and Nina Pavcnik. 2005. “Child Labor in the Global Economy.” The Journal

of Economic Perspectives 19:199-220. Einterz, Ellen M, and Myra E Bates. 1998. “Diagnosis of fever in Africa.” The Lancet

351:373-374. Einterz, Ellen M, and Myra E Bates. 1997. “Fever in Africa: do patients know when they are

hot?.” The Lancet 350:781. Emerson, Patrick M., and Andre Portela Souza. 2008. “Birth Order, Child Labor, and School

Attendance in Brazil.” World Development 36:1647-1664.

166

Emerson, Patrick M., and Andre Portela Souza. 2007a. “Child Labor, School Attendance, and

Intrahousehold Gender Bias in Brazil.” The World Bank Economic Review 21:301. Emerson, Patrick M., and Andre Portela Souza. 2007b. “Is Child Labor Harmful? The Impact

of Working Earlier in Life on Adult Earnings.” http://ssrn.com/abstract=1015498 (Accessed January 22, 2009).

Emerson, Patrick M., and Andre Portela Souza. 2003. “Is There a Child Labor Trap?

Intergenerational Persistence of Child Labor in Brazil.” Economic Development and Cultural Change 51:375-398.

Ersado, Lire. 2005. “Child Labor and Schooling Decisions in Urban and Rural Areas:

Comparative Evidence from Nepal, Peru, and Zimbabwe.” World Development 33:455-480.

Ewbank, Doug C., and Samuel H. Preston. 1990. “Personal health behaviour and the decline

in infant and child mortality: the United States, 1900-1930.” What We Know about Health Transition: The Cultural, Social and Behavioural Determinants of Health 2:116-149.

Fafchamps, Marcel, and Jackline Wahba. 2006. “Child labor, urban proximity, and

household composition.” Journal of Development Economics 79:374-397. Filmer, Deon. 2005. “Fever and its treatment among the more and less poor in sub-Saharan

Africa.” Health Policy Plan. 20:337-346. Filmer, Deon, and Lant Pritchett. 1999. “The Effect of Household Wealth on Educational

Attainment: Evidence from 35 Countries.” Population and Development Review 25:85-120.

Filmer, Deon, and Lant H. Pritchett. 2001. “Estimating Wealth Effects without Expenditure

Data-or Tears: An Application to Educational Enrollments in States of India.” Demography 38:115-132.

Floris, Fabrizio. 2006. Puppets or People?: A Sociological Analysis of Korogocho Slum.

Nairobi: Pauilines Publications Africa. Fotso, Jean-Christophe. 2006. “Child health inequities in developing countries: differences

across urban and rural areas.” International Journal for Equity in Health 5:9. Fotso, Jean-Christophe. 2007. “Urban-rural differentials in child malnutrition: Trends and

socioeconomic correlates in sub-Saharan Africa.” Health and Place 13:205-223. Gans, Herbert J. 1982. The Urban Villagers: Group and Class in the Life of Italian-Americans.

Updated and expanded ed. New York: Free Press.

167

Garau, Pietro, Elliott Sclar, and Gabriella Y Carolini. 2005. A Home in the City: UN

Millennium Project. London: Earthscan. Gilbert, Alan, and Josef Gugler. 1992. Cities, Poverty, and Development. Oxford: Oxford

University Press. Glick, Peter, and David E. Sahn. 2006. “The demand for primary schooling in Madagascar:

Price, quality, and the choice between public and private providers.” Journal of Development Economics 79:118-145.

Gould, W. T. S. 1998. “African mortality and the new `urban penalty'.” Health & Place 4:171-

181. Grauman, J. V. 1976. “Orders of magnitude of the world's urban population in history.”

Population Bulletin of hte United Nations 8:16-33. Grootaert, Christian, and Ravi Kanbur. 1995. “Child labour: An economic perspective..”

International Labour Review 134:187. Hack, Maureen et al. 1994. “School-Age Outcomes in Children with Birth Weights under 750

g.” N Engl J Med 331:753-759. Haines, Michael R., and Roger C. Avery. 1982. “Differential Infant and Child Mortality in

Costa Rica: 1968-1973.” Population Studies 36:31-43. Hanchett, Suzanne, Shireen Akhter, Mohidul Hoque Khan, Stephen Mezulianik, and Vicky

Blagbrough. 2003. “Water, sanitation and hygiene in Bangladeshi slums: an evaluation of the WaterAid- Bangladesh urban programme.” Environment and Urbanization 15:43-56.

Handa, Sudhanshu. 2002. “Raising primary school enrolment in developing countries: The

relative importance of supply and demand.” Journal of Development Economics 69:103-128.

Hannerz, Ulf. 1969. Soulside: Inquiries into Ghetto Culture and Community. New York:

Columbia University Press. Hanushek, E. A., and L. Wossman. 2007. Education Quality and Economic Growth.

Washington, D.C.: The World Bank. Hobcraft, J. N., J. W. McDonald, and S. O. Rutstein. 1984a. “Socio-Economic Factors in Infant

and Child Mortality: A Cross-National Comparison.” Population Studies 38:193-223. Hobcraft, J. N., J. W. McDonald, and S. O. Rutstein. 1984b. “Socio-Economic Factors in Infant

and Child Mortality: A Cross-National Comparison.” Population Studies 38:193-223.

168

Hofsten, E, and Hans Lundström. 1976. Swedish Population History. Huisman, Janine, and Jeroen Smits. 2009. “Effects of Household- and District-Level Factors

on Primary School Enrollment in 30 Developing Countries.” World Development 37:179-193.

International Labour Office. 2006. The End of Child Labour: Within Reach. Geneva:

International Labour Office. Jargowsky, Paul A. 1997. Poverty and Place: Ghettos, Barrios, and the American City. New

York: Russell Sage Foundation. Johnson, Gwendolyn Z. 1964. “Health Conditions in Rural and Urban Areas of Developing

Countries.” Population Studies 17:293-309. Kandala, Ngianga-Bakwin, and Nyovani Janet Madise. 2004. “The spatial epidemiology of

childhood diseases in Malawi and Zambia.” African Population Studies 19:199-226. Kearns, G. 1988. “The urban penalty and the population history of England.” Pp. 213–236 in

Society, Health and Population During the Demographic Transition. Stockholm: Almquist and Wiskell International.

Kintner, H. J. 1988. “Determinants of temporal and areal variation in infant mortality in

Germany, 1871-1933..” Demography 25:597-609. Klinenberg, Eric. 2002. Heat Wave: A Social Autopsy of Disaster in Chicago. Chicago:

University of Chicago Press. Kofoed, P E, F Dias, F Lopes, and L Rombo. 1998. “Diagnosis of fever in Africa.” Lancet

351:373-4. Koven, Seth. 2006. Slumming Sexual and Social Politics in Victorian London. Princeton, N.J:

Princeton University Press. Kramer, Mark. 2006. Dispossessed: Life in Our World's Urban Slums. Maryknoll, NY: Orbis

Books. Krueger, A. B. 1998. “International Labor Standards and Trade.” Developing and Newly

Industrializing Countries. Kuh, D., and M.E.J. Wadsworth. 1993. “Physical health status at 36 years in a British

national birth cohort.” Social Science & Medicine 37:905-916. Leach, Fiona. 2006. “Researching gender violence in schools: Methodological and ethical

considerations.” World Development 34:1129-1147.

169

Lee, Jennifer. 2002. Civility in the City: Blacks, Jews, and Koreans in Urban America.

Cambridge, MA: Harvard University Press. Lewis, Oscar. 1975. Five families : Mexican case studies in the culture of poverty. New York:

Basic Books. Lewis, Oscar. 1998. “The Culture of Poverty.” Society. 35:7. Liebow, Elliot. 1967. Tally's Corner: a Study of Negro Streetcorner Men. 1st ed. Boston:

Little, Brown. Lobo, Lancy, and Biswaroop Das. 2001. Poor in urban India : life in the slums of a western

Indian city. Jaipur: Rawat Publications. Lomnitz, Larissa Adler de. 197. Networks and Marginality: Life in a Mexican Shantytown.

New York: Academic Press. Martin, Linda G., James Trussell, Florentina Reyes Salvail, and Nasra M. Shah. 1983. “Co-

variates of Child Mortality in the Philippines, Indonesia, and Pakistan: An Analysis Based on Hazard Models.” Population Studies 37:417-432.

Massey, Douglas S, and Nancy A Denton. 1993. American Apartheid: Segregation and the

Making of the Underclass. Cambridge, MA: Harvard University Press. Massey, Douglas S. 1990. Return to Aztlan: The social process of international migration

from western Mexico. University of California Press. Massey, Douglas S. 1996. “The Age of Extremes: Concentrated Affluence and Poverty in the

Twenty-First Century.” Demography 33:395-412. Mehta, Arun. 2002. “Status of Secondary Education in India.” Pariperkshaya NIEPA, New

Delhi 9. http://www.educationforallinindia.com/page167.html (Accessed April 23, 2009).

Menon, P., M. T. Ruel, and S. S. Morris. 2000. “Socio-economic differentials in child stunting

are consistently larger in urban than rural areas: analysis of 10 DHS data sets.” Food and Nutrition Bulletin 21:282–299.

Mensch, B., H. Lentzner, and S. Preston. 1985. Socioeconomic Differentials in Child

Mortality in Developing Countries. New York: United Nations. Michaelowa, Katharina. 2001. “Primary Education Quality in Francophone Sub-Saharan

Africa: Determinants of Learning Achievement and Efficiency Considerations.” World Development 29:1699-1716.

170

Ministry of Information and Broadcasting. 2009. India, 2009: A Reference Annual. New Delhi: Publications Division, Ministry of Information and Broadcasting, Govt. of India.

Moehling, Carolyn M. 1999. “State Child Labor Laws and the Decline of Child Labor,.”

Explorations in Economic History 36:72-106. Monette, S. et al. 2007. “Validation of a measure of maternal perception of the childs health

status.” Child: Care, Health and Development 33:472-481. Montgomery, Mark, Richard Stren, and Bamey Cohen. 2003. Cities Transformed.

Washington, D.C.: National Academy Press. Montgomery, Mark R., and Paul C. Hewett. 2005. “Urban Poverty and Health in Developing

Countries: Household and Neighborhood Effects.” Demography 42:397-425. Morgan, Dilys, James AG Whitworth, Amanda Ross, Nicholas Omoding, and Henry Eotu.

1997. “Fever in Africa and WHO recommendation.” The Lancet 350:1549-1550. Mugisha, F. 2006. “School enrollment among urban non-slum, slum and rural children in

Kenya: Is the urban advantage eroding?.” International Journal of Educational Development 26:471-482.

Murray, Christopher J. L., and Lincoln C. Chen. 1992. “Understanding Morbidity Change.”

Population and Development Review 18:481-481. Nangia, Sudesh, and Sukhadeo. Thorat. 2000. Slum in a metropolis : the living environment.

Delhi: Shipra Publications. National Sample Survey Organisation. 2003. Condition of urban slums, 2002 : salient

features, NSS 58th round (July 2002-December 2002). [New Delhi]: National Sample Survey Organisation Ministry of Statistics and Programme Implementation Govt. of India.

Neuwirth, Robert. 2005. Shadow Cities: A Billion Squatters, a New Urban World. New York:

Routledge. Newman, Katherine S. 2006. Chutes and Ladders: Navigating the Low-Wage Labor Market.

New York: Russell Sage Foundation. Newman, Katherine S. 1999. No Shame in My Game: The Working Poor in the Inner City. 1st

ed. New York: Knopf and the Russell Sage Foundation. Newman, Katherine S., and Ashwini Deshpande. 2007. “Where the Path Leads: The Role of

Caste in Post-University Employment Expectations.” Economic and Political Weekly 42:2133-2140.

171

Newman, Katherine S., and Surinder Jodhka. 2007. “In the Name of Globalization:

Meritocracy, Productivity and the Hidden Language of Caste.” Economic and Political Weekly 42:4125-4132.

Newman, Katherine S., and Sukhadeo Thorat. 2007. “Caste and Economic Discrimination:

Causes, Consequences, and Remedies.” Economic and Political Weekly 42:4121-4124.

Park, Robert Ezra, Ernest Watson Burgess, and Roderick Duncan McKenzie. 1925. The City.

Chicago, Ill: Univ. of Chicago Press. Perlman, Janice E. 2003. “Marginality: From Myth to Reality in the Favelas of Rio de Janeiro,

1969–2002.” Urban Informality: Transnational Perspectives from the Middle East, Latin America, and South Asia 105–46.

Perlman, Janice E. 2004. “The Metamorphosis of Marginality in Rio de Janeiro.” Latin

American Research Review 39:189-192. Perlman, Janice E. 2006. “The Metamorphosis of Marginality: Four Generations in the

Favelas of Rio de Janeiro.” The ANNALS of the American Academy of Political and Social Science 606:154.

Perlman, Janice E. 1976. The Myth of Marginality: Urban Poverty and Politics in Rio De

Janeiro. Berkeley: University of California Press. Pillai, Rajamohanan K et al. 2003. “Factors affecting decisions to seek treatment for sick

children in Kerala, India.” Social Science & Medicine 57:783-790. Planning Commission. 2007. Poverty estimates for 2004-2005. New Delhi, India: Planning

Commission http://www.planningcommission.gov.in/news/prmar07.pdf (Accessed May 12, 2009).

Pong, Suet-Ling. 1997. “Sibship Size and Educational Attainment in Peninsular Malaysia: Do

Policies Matter?.” Sociological Perspectives 40:227-242. van Poppel, F, and C van der Heijden. 1997. “The effects of water supply on infant and

childhood mortality: a review of historical evidence.” Health Transition Review: The Cultural, Social, and Behavioural Determinants of Health 7:113-48.

Power, C, and C Peckham. 1990. “Childhood morbidity and adulthood ill health..” Journal of

Epidemiology and Community Health 44:69-74. Preston, Samuel H., and M. R. Haines. 1991. Fatal Years: Child Mortality in Late Nineteenth-

Century America. Princeton, NJ: Princeton University Press.

172

Preston, Samuel H., and E. van de Walle. 1978. “Urban French mortality in the nineteenth century.” Population Studies 32:275-297.

Preston, Samuel H., and Etienne van de Walle. 1978. “Urban French Mortality in the

Nineteenth Century.” Population Studies 32:275-297. Puffer, R. R., and C. V. Serrano. 1973. Características de la mortalidad en la niñez.

Washington, D.C.: Pan American Health Organization. Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community.

New York: Simon & Schuster. Radhakrishna, R., and Shovan Ray, eds. 2005. Handbook of Poverty in India: Perspectives,

Policies, and Programmes. New Delhi: Oxford University Press. Redfield, Robert. 1950. A village that chose progress; Chan Kom revisited. [Chicago:

University of Chicago Press. Redfield, Robert. 1989. The little community and peasant society and culture. Chicago: The

University of Chicago Press. Redfield, Robert. 1977. The primitive world and its transformation. Ithaca, N.Y.: Cornell

University Press. Rosati, Furio Camillo, and Zafiris Tzannatos. 2006. “Child Labour in Vietnam.” Pacific

Economic Review 11:1-31. Rutstein, S. O. 1984. Socioeconomic Differentials in Infant and Child Mortality. Voorburg,

Netherlands: ISI. Ryland, Shane, and Hendrik Raggers. 1998. Child morbidity and treatment patterns.

Calverton, MD: Macro International. Sahn, D. E., and D. C. Stifel. 2003. “Urban-Rural Inequality in Living Standards in Africa.”

Journal of African Economies 12:564-597. Sampson, Robert J., Jeffrey D. Morenoff, and Thomas Gannon-Rowley. 2002. “Assessing

"Neighborhood Effects": Social Processes and New Directions in Research.” Annual Review of Sociology 28:443-478.

Sastry, N. 1997. “What explains rural-urban differentials in child mortality in Brazil?.”

Social Science & Medicine 44:989-1002. Schenk, H. 2001. Living in India's slums : a case study of Bangalore. New Delhi: Manohar

Publishers & Distributors.

173

Silvers, Jonathan. 1996. “Child Labor in Pakistan.” The Atlantic Monthly 17:79-92. Small, Mario Luis, and Katherine Newman. 2001. “Urban Poverty after The Truly

Disadvantaged: The Rediscovery of the Family, the Neighborhood, and Culture.” Annual Review of Sociology 27:23-45.

Smith, L. C., M. T. Ruel, and A. Ndiaye. 2005. “Why Is Child Malnutrition Lower in Urban

Than in Rural Areas? Evidence from 36 Developing Countries.” World Development 33:1285-1305.

Smits, Jeroen, and Ayse Gündüz Hosgör. 2006. “Effects of family background characteristics

on educational participation in Turkey.” International Journal of Educational Development 26:545-560.

Stack, Carol B. 1974. All Our Kin: Strategies for Survival in a Black Community. 1st ed. New

York: Harper & Row. Starr, Paul. 1992. “Social Categories and Claims in the Liberal State.” Social Research

59:263-295. Sullivan, J. M., S. O. Rutstein, and George Bicego. 1994. Infant and Child Mortality. Calverton:

Macro International. Suttles, Gerald D. 1968. The Social Order of the Slum; Ethnicity and Territory in the Inner

City. Chicago: University of Chicago Press. Thomas S. Dee. 2006. “A Teacher Like Me: Does Race, Ethnicity, or Gender Matter?.” The

American Economic Review 95:158-165. Thomas, D, J Strauss, and M Henriques. 1990. “Child survival, height for age and household

characteristics in Brazil.” Journal of Development Economics 33:197-234. Trussell, James, and Charles Hammerslough. 1983. “A Hazards-Model Analysis of the

Covariates of Infant and Child Mortality in Sri Lanka.” Demography 20:1-26. Trussell, James, and Samuel H. Preston. 1982. “Estimating the covariates of childhood

mortality from retrospective reports of mothers.” Health Policy and Education 3:1-36.

UNESCO. 2007. Education for All by 2015: Will We Make It? New York: Oxford University

Press. UNESCO. 2004. Education for All: The Quality Imperative. Paris: UNESCO. United Nations. 2004. World Urbanization Prospects, the 2003 Revision. New York: United

Nations.

174

United Nations Center for Human Settlements. 1999. Global Urban Indicators Database,

1998. New York: United Nations. United Nations Centre for Human Settlements. 1996. An Urbanizing World: Global Report

on Human Settlements 1996. Oxford: Oxford University Press. United Nations Human Settlements Programme. 2003. The challenge of slums: Global

report on human settlements. Sterling, VA: Earthscan Publications. Van de Poel, E, O O'Donnell, and E Van Doorslaer. 2007. “Are urban children really

healthier? Evidence from 47 developing countries.” Social Science & Medicine 65:1986-2003.

Vasconcellos, Eduardo A. 1997. “Rural transport and access to education in developing

countries: policy issues.” Journal of Transport Geography 5:127-136. Velez-Ibanez, Carlos G. 1983. Rituals of Marginality: Politics, Process, and Culture Change in

Urban Central Mexico, 1969-1974. Berkeley: University of California Press. Verma, Gita Dewan. 2002. Slumming India : a chronicle of slums and their saviours. New

Delhi; New York: Penguin Books. Victoria, Cesar G., Sharon R. A. Huttly, Fernando C. Barros, Cintia Lombardi, and J. Patrick

Vaughan. 1992. “Maternal education in relation to early and late child health outcomes: Findings from a Brazilian cohort study.” Social Science & Medicine 34:899-905.

Vogele, Jorg. 2000. “Urbanization and the urban mortality change in Imperial Germany.”

Health and Place 6:41-55. Vogele, Jorg P. 1994. “Urban Infant Mortality in Imperial Germany.” Soc Hist Med 7:401-

425. Von Braun, J., J. McComb, B. K. Fred-Mensah, and R. Pandya-Lorch. 1993. Urban Food

Insecurity and Malnutrition in Developing Countries: Trends, Policies, and Research Implications. International Food Policy Research Institute Washington, DC (USA).

Wang, L. 2003. “Determinants of child mortality in LDCs Empirical findings from

demographic and health surveys.” Health policy 65:277-299. Ward, David. 1989. Poverty, Ethnicity, and the American City, 1840-1925: Changing

Conceptions of the Slum and the Ghetto. Cambridge: Cambridge University Press. Waters, Elizabeth et al. 2000. “Influence of Parental Gender and Self-Reported Health and

Illness on Parent-Reported Child Health.” Pediatrics 106:1422-1428.

175

Weber, Adna Ferrin. 1899. The Growth of Cities in the Nineteenth Century: A Study in

Statistics. New York: MacMillan for Columbia University. Weeks, John R. 2008. “Do Slums Promote High Urban Fertility: Neighborhood Differences in

Accra, Ghana.”. Whyte, William Foote. 1943. Street Corner Society: The Social Structure of an Italian Slum.

Chicago, Ill: University of Chicago Press. Wilson, William J. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and

Public Policy. Chicago: University of Chicago Press. Wilson, William J. 1996. When Work Disappears: The World of the New Urban Poor. 1st ed.

New York, NY: Alfred A. Knopf. Wirth, Louis. 1928. The Ghetto. Chicago, Ill: The University of Chicago Press. Woldemicael, Gebremariam. 2001. “Diarrhoeal Morbidity among Young Children in Eritrea:

Environmental and Socioeconomic Determinants.” Journal of Health, Population, and Nutrition 19:83-90.

Woods, Robert, and P.R. Andrew Hinde. 1987. “Mortality in Victorian England: Models and

Patterns.” Journal of Interdisciplinary History 18:27-54. World Bank. 2009a. “India - Education for All: Sarva Shiksha Abhiyan.”

http://www.worldbank.org.in/WBSITE/EXTERNAL/COUNTRIES/SOUTHASIAEXT/INDIAEXTN/0,,contentMDK:21974928~pagePK:141137~piPK:141127~theSitePK:295584,00.html (Accessed April 23, 2009).

World Bank. 2009b. “India Country Overview 2009.”

http://www.worldbank.org.in/WBSITE/EXTERNAL/COUNTRIES/SOUTHASIAEXT/INDIAEXTN/0,,contentMDK:20195738~menuPK:295591~pagePK:141137~piPK:141127~theSitePK:295584,00.html (Accessed May 15, 2009).

World Bank. 2006. World Development Report 2007: Development and the Next

Generation. Washington, D.C.: World Bank/Oxford University Press. World Health Organization. 2004. “The WHO Multicentre Growth Reference Study (MGRS):

Rationale, planning, and implementation.” Food and Nutrition Bulletin 25:S3-S84. World Health Organization Multicentre Growth Reference Study Group. 2006. WHO child

growth standards: Length/heightfor- age, weight-forage, weight-for-length, weight-for-height and body mass index-for-age: Methods and development. Geneva: World Health Organization.

176

Yohannes, A G, K Streatfield, and L Bost. 1992. “Child morbidity patterns in Ethiopia.”

Journal of Biosocial Science 24:143-55. Zerubavel, Eviatar. 1996. “Lumping and splitting: Notes on social classification.”

Sociological Forum 11:421-433. Zorbaugh, Harvey Warren. 1929. The Gold Coast and the Slum: A Sociological Study of

Chicago's Near North Side. Chicago: University of Chicago Press.