Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS...

47
DEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving the poor - a case study of students’ performance in Indian hostels Author: Rut Lindén Supervisor: Peter Fredriksson

Transcript of Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS...

Page 1: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

DEPARTMENT OF ECONOMICS

Uppsala University

Master Thesis

Spring term 2005

Educational policies serving the poor

- a case study of students’ performance in Indian hostels

Author: Rut Lindén

Supervisor: Peter Fredriksson

Page 2: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

2

Abstract

This study examines the effect on school achievement of a policy such as hostels, aimed at

giving children from a poor socioeconomic background an opportunity to receive education.

Data is collected from two different schools in a district in Andhra Pradesh, India, in which

both hostel students and day-scholar students, having a similar background, are studying.

Exam scores for three different subjects are used as dependent variables in the analysis. The

results indicate that private hostels do have a positive effect on achievement in all subjects,

thereby contributing to reducing the large gap in school achievement between different

socioeconomic groups.

Key words: education – intergenerational mobility – human development – rural area – caste

– India – hostel

Acknowledgement

I would like to express my gratitude to Dr. D Samuel Issmer, director of SAC-CCC in

Bangalore who made it possible for me to conduct this study and who gave great guidance

and support in my work.

I am grateful to Mrs. Mangalam Bahavathsingh at the SAC-CCC office; Head Master Y.E.

Supadham and Mrs Parimala Ravi in Kollagunta; Head Mistress P Suguna Thilagam and

Correspondent Ms Sarawathy in Chittoor for letting me take their time and for sharing their

views with me. I would also like to thank the rest of the staff at the SAC-CCC office as well

as teachers and wardens at the schools and hostels who helped me in practical matters; and

also my supervisor Peter Fredriksson for many good advices.

Finally, I would like to thank all the children included in this study – you brought me so much

joy! I will always keep the memory of you in my heart, hoping that one day all your dreams

come true.

Page 3: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

3

Contents

1 Introduction .................................................................................................................. 4

2 Education in India ........................................................................................................ 6

3 Theories on education and intergenerational mobility ............................................. 7

3.1 Human Capital Theory ............................................................................................... 7

3.2 Human Development Perspective .............................................................................. 7

3.3 Intergenerational mobility .......................................................................................... 8

4 Features of the studied area ........................................................................................ 9

4.1 The educational situation for the poor...................................................................... 10

4.2 The role of the Caste system .................................................................................... 11

4.3 Hostel and admittances............................................................................................. 12

4.4 Background of schools and students included in the study...................................... 13

5 Data collection and Descriptive statistics ................................................................. 15

6 Model ........................................................................................................................... 20

7 Results ......................................................................................................................... 23

8 The effect of staying in a hostel ................................................................................. 34

8.1 Life in the hostels ..................................................................................................... 34

8.2 The relevance of the effect ....................................................................................... 37

9 Conclusions ................................................................................................................. 38

10 References ................................................................................................................... 40

Appendix A: Features of the studied area ........................................................................... 42

Appendix B: Questionnaire ................................................................................................... 43

Appendix C: The distribution of average test scores .......................................................... 44

Appendix D: OLS with distance dummies ........................................................................... 45

Appendix E: Effect of years in hostel ................................................................................... 46

Page 4: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

4

1 Introduction

Education is one of the fundamental objectives of development. It is important from a human

rights’ perspective – to enable people to develop as individuals, raise themselves out of

poverty and to empower them so they can use other rights in the society. From a macro

economic perspective, education is also one of the key ingredients for economic growth in a

society: By raising the educational level of individuals the productivity in the economy will

also be raised (Becker 1964). In 2000, a Millennium declaration was signed by the nations in

the UN. One of the goals in this declaration is to “ensure that, by 2015, children everywhere,

boys and girls alike, will be able to complete a full course of primary schooling” (UN 2004).

Today most developing countries devote a large share of their government budget to

education. Still, there are millions of children around the world who are deprived of

educational opportunities, often due to poverty. In India, with the second largest population in

the world after China, 17 percent of all school-aged children are not enrolled in school and

only 59 percent of the children starting school, reach fifth grade (World Bank 2004). In

addition to this, studies have shown that student achievement in Indian primary schools is

very low, particularly among the poor (World Bank 1997:85). To be able to fulfil the

Millennium goal we need to design policies that raise the educational level and reach the poor

in the society.

In India, social service organisations as well as the government work in several ways to

enable children from poor backgrounds to get an education. One way is by running hostels

where children can stay during the school year while attending school. My aim with this study

is to examine the effects these kinds of policies have on children staying there. The study has

been conducted at two schools in Chittoor district, in the state of Andhra Pradesh. In

particular, I study the effects on school achievement while staying in a hostel. I will compare

high school students who stay in a hostel during the semesters, with day-scholar students

studying in the same school. This will be done by comparing the students’ exam scores, using

them as a dependent variable in a regression function which includes all observed factors that

can have an impact on the grades. My hypothesis is that children who get the chance to stay in

a hostel, instead of growing up under poor conditions, do better in school.

It should be pointed out that there are at least two interesting questions that could be

addressed concerning the effect of these hostel policies. This study attempts to answer the

following question: Given that students attend the same school, what is the effect of staying in

Page 5: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

5

a hostel? Even with similar backgrounds for the hostel students and day-scholars, we can not

say for sure with this kind of study what would have happened to the hostel students had they

not been given the chance to stay in a hostel. If the alternative is to drop out of school, it can

be argued that achievement is of secondary concern for these policies. But fully investigating

the effect of these hostel students not staying in a hostel would require two regions with equal

socioeconomic setting; one with hostel opportunities and one without. Since the setting for

such a study is difficult to find, I have focused on comparing the achievement within a school.

Even if not of primary concern, this question is of great importance in order to give students

more equal opportunities in life; to proceed with their studies beyond the basic schooling, and

reach a higher level of development than the family generation before them. However, both

questions will be addressed in this study, but with the awareness that only one question can be

adequately examined.

In what way does my study contribute to this area of research? Previous studies for

developing countries have analyzed the effects on school attainment and attendance of

policies such as cash transfers to poor families (Inchauste 2000), elimination of costs of

schooling (Deininger 2003), and increasing the supply of schools (Filmer 2004). Furthermore,

many studies have investigated the socioeconomic determinants of schooling (see e.g.

Jayachandran 2002). Even though the policy with hostels is widespread in India1, and similar

policies exists in several other countries, I have not come across any other study analyzing the

effect of staying in a hostel.

This paper begins by giving some background information about the educational situation in

India. Thereafter, the theories within the area of economics of education will be presented and

earlier empirical research will be reviewed. In section 4, focus will be given to the particular

area where the case study has taken place, and the background of the students staying in the

hostels and in the two schools included in the study, will be depicted. The information

collected at the schools, Boyall Memorial High School in Kollagunta and Sherman Memorial

Girls High School2 in Chittoor town, will be described and thereafter the model used for

testing my hypothesis will be presented. Afterwards the results will be displayed and analysed

before concluding the study.

1 In Chittoor district for example, the government alone runs 205 hostels (Sridhar & Sujatha 2002). 2 The schools will hereafter be denoted Boyall HS and Sherman HS.

Page 6: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

6

2 Education in India

Primary education is defined internationally as “providing children with basic reading,

writing, and mathematics skills along with an elementary understanding of such subjects as

history, geography, natural science, social science, art, and music”.3 In India, primary

education covers grades 1 to 8, with upper primary school beginning in grade 5 or 64. In the

Indian constitution (Article 45) it is written that the state shall provide “free and compulsory

education for all children until they complete the age of fourteen”, which is equivalent to

grade 8. Yet, lower secondary school (grades 9 to 10) is often included when discussing basic

schooling in India since many states provide high school education for children from grades 6

to 10. Also, ten years of schooling is increasingly considered a minimum in the society of

today. This view is supported by e.g. Dr. Niranjanaradhya, National Law School of India

University, who argues that the age limit of fourteen was set during a time when

socioeconomic conditions were different: “Now we are in the competitive age of

globalisation, liberalisation and privatisation. The new situation compels the state to provide

minimum ten years of school education to all children.” (Niranjanaradhya 2004:22)

According to the World Bank, there has never been more political support for primary

education in India than now. Still, there is a long way to go before the Millennium goal is

reached or the Article 45 is fulfilled. To reach further, the World Bank stresses the need for

expanding the access – making it possible for more children to go to school – and enhancing

the learning achievement. Empirical studies suggest that students’ achievement is low in the

Indian primary schools. There is a great variation between states and schools, yet, a common

finding for all states is that students with poor socioeconomic background have a lower

achievement than other students. Moreover, the achievement varies by gender, caste and tribe.

These gaps between different groups also exist in the enrolment and retention in schools, and

is thus another important issue pointed out by the World Bank (1997:1-13). These issues are

also in line with the objectives outlined in the Indian National Policy on Education (1992).

The educational situation for the poor will be further discussed in section 4.1, with focus on

the specific region where this study has been conducted. Let us now turn to the theories of

interest for this study.

3 Based on the International Standard Classification of Education, 1976 (ISCED76) and 1997 (ISCED97) (World

Bank 2002). 4 There is no uniform educational system for India, but mostly it follows the same structure in all states.

Page 7: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

7

3 Theories on education and intergenerational mobility

Education is a large area of research and it is explored within many different fields.

Economists have long based their research on human capital theories which emphasize the

economic contribution of education. In the last decade, there has been a move towards a more

holistic perspective of development which focuses more on the individual and other aspects of

education. Though the two perspectives are closely related concepts, and sometimes are

merged into one, I will present each separately since they show the importance of education in

somewhat different ways. Of interest for this study is also the theories of intergenerational

mobility and the empirical findings within this research area.

3.1 Human Capital Theory

The human capital theory5 emerged during the 1960’s as a way of explaining economic

growth in the society. The theory gives education an important role in achieving economic

growth and reducing poverty. Human capital is here seen as the “productive investments

embodied in people” including for example skills, abilities and ideals (Todaro 2003:797).

Education raises this stock of human capital, which leads to improved economic performance

and thus economic growth. Empirically, this has been shown in many studies; for example,

studies from East Asia show that primary education contributes to the largest share to the

growth in recent years (World Bank 1995:22-23). There is, of course, also a direct effect of

education for the individual receiving it. The relationship between education and earnings –

with a higher education leading to higher earnings – has been frequently studied and proven

to hold when studying populations as a whole, as well as sub-groups within the society. The

skills and knowledge that education brings, and the raise in productivity, can for example lead

a farmer to increase his output or a worker to get a more qualified job and thus receive a

higher income (Melin 2002; World Bank 1995:30-39).

3.2 Human Development Perspective

With the human development perspective, development has received a wider meaning. This is

an approach to poverty reduction which, to a larger extent than previously, focuses on the

overall poverty that an individual faces. The leading thinker on human development6 is the

Indian economist Amartya Sen. He argues that an individual’s well-being is not only

determined by his or her income or consumption level: what matters is what he or she can do

with this, i.e. a person’s capabilities. Poverty is a lack of real opportunities – not being able to

5 It was developed by economists like Schultz (1961), Becker (1964) and Mincer (1974). 6 Sen later modified the theory into the Human Capabilities approach

Page 8: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

8

live the kind of life one wants to live. This could be due to social constrains, such as

belonging to a discriminated group, or personal circumstances such as poor health. Education

is, together with health, a fundamental objective for development according to the human

development perspective. It is not only a way for development and to reduce poverty, it is

development. Education improves an individual’s well being and freedom in many ways apart

from only raising incomes; e.g., reading and writing enables people to participate and use

their rights in the society to a greater extent, which can lead to an increased democracy.

Knowledge also contributes to better health, thereby reducing human poverty. This wider,

holistic view gives education an even more important role for reducing poverty. (Sen 1999;

Drèze & Sen 1996:9-16)

3.3 Intergenerational mobility

The differences in how well children succeed in life has been a growing research field within

social science since the 1970’s. Success in this respect can for example be measured by

learning achievement or education, which in turn affects future opportunities and income.

Economic studies focus on family background to explain these differences (Haveman & Wolf

1995). According to human capital theories developed by Becker and Tomes (see e.g. Becker

Tomes 1986) children’s success is determined by their own abilities (genetic inheritance)

together with the decisions made by their parents to invest in their children. Parents are the

ones making decisions regarding the economic resources in a family and they want to

maximize utility. When doing this, they are concerned about their children’s welfare, whether

for selfish or altruistic reasons. They will therefore maximize utility by choosing between

family consumption and investment in the children. These investments in human capital could

be aimed at for example improving the children’s skills, health, learning and motivation in

order to help the children to succeed in life. When there are no financial limitations,

investments can be made to optimize the level of human capital and abilities of a child. The

degree of inheritability between generations will then be low since, to some extent, the

investments can compensate for bad genetic endowments. For poor families the chances to

invest in their children are limited, so when education is too expensive, the children are not

able to acquire a proper education. The children’s opportunities in life will therefore be

reduced. In other words, they will inherit not only genes and cultural endowments but also

some of their parents’ socioeconomic status. In these cases it is said that the intergenerational

mobility is low. This can be seen as both unfair, from a human rights perspective, as well as

inefficient.

Page 9: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

9

Leibowitz (1974) is another economist who has contributed within this field of research. He

suggests that what determines the family income and the quantity and quality of the

investments in the household is parents’ abilities and education. Many studies on

intergenerational mobility, conducted in industrialised as well as developing countries, have

found a connection between parents’ characteristics and their children’s education (Wolfe &

Behrman 1984; Weinberg 2001). Especially parents’ income and education seems to be

important determinants of how much will be invested in a child. With financial limitations the

number of children will also be a factor affecting the investment in each child. The investment

in a child is not only financial: The time spent with a child, the support, encouragement etc.

also affect a child’s attainments. Likewise, family stability and decisions on where to live

have shown to be relevant factors (see e.g. Ginther & Pollak 2003).

These theories and empirical findings need to be kept in mind when constructing the

regression model further on. However, before proceeding to that, the characteristics of the

studied area and the policy of hostels need to be explored.

4 Features of the studied area

The study was conducted in Chittoor district, situated in the south part of Andhra Pradesh,

bordering the state of Tamil Nadu in the south and the state of Karnataka in the west.7

Approximately 80 percent of the population in this district lives in rural areas, and among the

workers, 30 percent are cultivators and another 36 percent are agricultural labourers. The

agricultural labourers are so called “coolie workers”, i.e. seasonal workers, employed and

paid on a day to day basis depending on the landowners’ need for workers. The literacy

amongst the total population in Chittoor district is estimated to approximately 64 percent in

the rural area and 81 percent in the urban areas. (For females the rates are 52 and 73 percent

respectively.) Chittoor district is special in the sense that it has previously belonged to Tamil

Nadu. Consequently, many people still have Tamil as their mother tongue (the language of

Tamil Nadu), instead of Telugu, which is the Andhra Pradesh state language. (Sridhar &

Sujatha 2002)

7 See map in appendix A:I.

Page 10: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

10

4.1 The educational situation for the poor

In rural areas, one problem when reaching high school is the access. It is sometimes very far

to the closest school and if the village is too remote there are no buses to take the children to

school. Statistics from “The Indian educational survey” (NCERT 2002) show that Chittoor

district has 630 secondary schools (where high schools are included), and out of these 443 are

situated in rural areas. This can be compared to the 4300 primary schools in the same district.

Consequently, all children will not have access to school when reaching upper primary and

secondary school, thus leading them to drop out. Even where buses or other means of

transportation are available, the economic situation in many families makes it difficult for

parents to send their children to school. When it comes to female students, a long distance to

school is not only an economic problem but also a cultural: When girls reach maturity it is no

longer considered safe for them to travel to school and some parents will therefore keep them

in the home.

Even though the government schools are free, there are still direct costs associated with

attending school, apart from possible transportation costs. These are for example costs for

buying school uniform and school supplies, such as pens and notebooks. For poor families the

alternative cost of sending a child to school can also be high: Earning for the family income

might then be considered a better choice for the child.

This argument would be even stronger when the standard of education is very low, which is

often the case in rural areas. The Indian educational survey reveals that many primary and

secondary schools in Chittoor district lack proper facilities: Some schools are without

building and others are made of materials such as mud and grass. Almost a quarter of the

primary schools have only one teacher running the school, and there are also examples of

schools having no teacher employed. This is how the village situation is described in the

neighbouring state of Karnataka:

“..a large number of Government-run primary and secondary schools have already been

set up across the length and breadth of the state. But the challenge is to transform these

schools into functioning schools. Very often these schools have only one or two teachers,

not enough classrooms, no drinking water, no toilets and above all, no real education

going on.” (Niranjanaradhya 2004)

Page 11: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

11

4.2 The role of the Caste system

In India, the caste system still plays an important role in the society both politically and

culturally. Everyone belongs to a caste, regardless of their religion. These castes are divided

into four groups as a way for government to determine which priority certain castes should

have in receiving state aid. Scheduled Tribes (ST) and Scheduled Castes (SC) are the ones

formerly called untouchables. These are the most disadvantaged groups and are thus eligible

to most help from the government. Below these groups in priority are first the Backward

Castes (BC) and finally the group of Other Castes (OC), consisting of the upper castes.

Especially in the rural areas, a person’s caste can give a lot of information about their

socioeconomic status. According to Professor Shah, there is a “positive relationship between

caste and occupational status; the rich and middle farmers belong largely to the upper and

middle castes.” (Shah 2002:394) Among the poor, he also notices a difference between the

poor of the upper and the lower castes. A poor upper caste will have an advantage compared

to the poor lower caste since the upper caste has more contacts among the people possessing

power in the society. Also, the upper caste child has a better “social milieu” than the lower

caste. One reason for that is that the lower castes traditionally have been denied education, so

many are still illiterate, whereas it is considered an exception to find a member of an upper

caste with no education. (Shah 2002:394-397)

Villages in India are divided into smaller habitations, or clusters of households. Just like in

many other parts of India, in the rural villages in Chittoor district the upper castes and

backward castes live separated from the scheduled castes. The upper castes’ standard of living

is in general better, with constructed houses and often more than one room to live in. The

scheduled castes are mostly coolie workers, working in the fields owned by richer upper

castes. Many of the schedule caste people live in small houses constructed with mud and

grass. So, information on the caste is important to include in this study because there should

be a difference between the upper and lower caste even in the case where the parents’

occupation and years of schooling is the same.

Page 12: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

12

4.3 Hostel and admittances

In India, both government and private organisations run hostels in order to help children

receive an education. The hostels primarily included in this study belong to the Church of

South India (CSI) who runs the hostels either by itself or with the help of NGO’s.8 Nearly half

of the hostel students included in this study are sponsored by South Asia Council for

Community and Children in Crisis (SAC-CCC), an organisation running several hostels all

over South India and from whom I received the help to conduct the study. Children eligible to

stay in the CSI hostels are those coming from very poor families, who otherwise would not be

able to complete their high school studies. This means that the majority of the children will

come from scheduled castes9. Most of them also come from rural areas where there is no

opportunity for them to go to high school anywhere close by. To select the most deserving

children, the CSI goes through the following procedure. First, families in poor, rural areas are

informed by a local pastor about the possibility to apply to a hostel after having completed

grade 5. The pastor, as well as child care workers at the hostel, visits the applying families to

investigate their family situation and to determine whether they are needy or not. Priority is

given to orphans, semi-orphans and children coming from large families. Eligible students are

given a test conducted by the CSI diocese. The test is a basic skills test and is a way to select

those children who will most likely be able to continue their higher studies. Often, these rural

children have received a low standard education and, also, they have not been exposed to

things the same way as other children have. Because of that, they will have to strive hard to

catch up with children coming from less rural areas. In some cases their basic knowledge is

too low and thus, they will not be able to follow the syllabus in high school. Those students

are very likely to drop out after some years and will therefore be sorted out with the help of

the test. According to Mrs. M. Bagharavathi, project coordinator at the SAC-CCC office, the

test is “neither intelligence based, nor merit based. It is a simple test to find out whether the

children have grasped the basics in reading and writing”. This is confirmed by Dr. D. S.

Issmer, director of SAC-CCC, and Ms. Saraswathy, CSI Correspondent at the Sherman

Memorial Girls Hostel. They all point out that most of the children admitted to the hostels are

below average when they enter the school, both due to their family background and the

standard of their schools. When high in priority, children will also be admitted even if they

don’t pass the test.

8 See appendix A:II for a list of hostels included 9 Scheduled tribes (ST) are not as many in number as the Scheduled castes (SC) which explains why they are not

represented to the same extent.

Page 13: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

13

Students admitted all have to pay a monthly fee. The fee is set very low, as a symbolic sum,

and is flexible depending on the family situations10. Still, it is considered important to collect

it so that the parents feel their responsibility and take an active role in their children’s

education.

Included in the study conducted at Sherman HS are also 42 students staying in government

hostels. The admission to the government hostels is based on caste and family income. i.e.

only poor children belonging to SC or ST stay there and no entrance test is conducted. No fee

is collected at these hostels; on the other hand, things such as school uniform are not always

provided. I was not able to visit any of these hostels, instead I have based my information of

them from four interviews with children staying there. These hostels will not be in primary

focus in this study; still it can be of interest to compare any differences in the results from the

CSI hostels.

4.4 Background of schools and students included in the study

I conducted the study at two different high schools11 in Chittoor district: Sherman Memorial

Girls High School in Chittoor town, and Boyall Memorial High School, in Kollagunta village.

Sherman is a private-run institution which was founded in 1923 as a boarding school, with the

aim to reach out to girls from rural areas and give them a chance to education and

development. Today 50 percent of the 691 students are day-scholars whereas the rest of them

stay either within the school compound or in nearby, government-run hostels12. Though

private-run, the school receives state aid which makes it possible for them to provide free

education for all students wanting to go there.

At Sherman HS there are two sections – one with Tamil as the medium of instruction and the

other one with Telugu – so students can choose which of the two languages to study in. Still,

many of the children with Tamil background choose to study in Telugu, since that is the state

language and thus more useful to know if they want to live in Andhra Pradesh. The fact that

both Sherman HS and Boyall HS teach in Telugu and Tamil, and not in English implies that

the students studying there all come from poor backgrounds. Nowadays in India, everyone

who can afford to place their child in a private-run, English medium school, will do so.

Obtaining good English skills, and an overall good education, is important in the high

10 The organisations running the hostels under CSI all have different fees, ranging from 50-100 Rs/month. This

can be compared to a daily wage of 25-50 Rs for agricultural labourers, which is roughly 0.55-1.15 U.S. dollars. 11 Including grade six to ten. 12 Boys can study at the school up to 7th grade, therefore a few boys come from a CSI boys hostel nearby.

Page 14: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

14

competition for college seats. English is also becoming increasingly important to master

considering the growing interaction between states, as well as between India and the rest of

the world. Left in the freely provided government schools, or government supported schools,

will thus be children from a poor background who do not have the opportunity to choose

where to study.

The hostel students come from rural villages with a background situation as described above.

According to Ms P. S. Thilagam, Head Master at the Sherman HS, the background of many of

the day-scholars is similar to the hostel students: They belong to the lowest caste with parents

who are illiterate and work for daily wages. The difference is that most of these students have

a shorter distance, or at least easier access, to school which makes it possible for them to stay

in their house. Also some of them are slightly better off due to the fact that they live closer to

Chittoor town and the parents are therefore able to get higher daily wages than the rural

agricultural labourers. In the town area, there are also more opportunities to get work all year

around. Even if better off, these students are also very poor and the difference is marginal

between the two groups.

Among the day-scholars in Sherman HS there is also a group of better-off students. They

belong to the upper caste and live within the city limits. Their parents receive monthly salaries

and are in general better educated. Still, compared to students attending fee levied English

medium schools, these students are also considered poor.

Boyall Memorial High School is a government-run, Telugu medium school located in a rural

area right on the boarder of Tamil Nadu. The school is co-educated (i.e., both boys and girls

study there) and has 291 students. For each grade from 6 to 10, there is one class. About 50

percent of the students stay in a hostel and the rest are day-scholars. There are two CSI hostels

adjacent to the school; one for boys and one for girls. The majority of the students have Tamil

as their mother tongue; still, all have studied their lower primary school years in Telugu

medium schools.

The socioeconomic background of the hostel students compared to the day-scholars is more

uniform than that of Sherman HS due to the location of Boyall HS. The majority of the hostel

students live more than five kilometres from the school, whereas it is the other way around for

the day-scholars. Still, since all come from rural villages, staying farther away from school

will not reflect having a poorer socioeconomic background to the same extent as when the

Page 15: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

15

school is located in a town. Even with the same village setting, there are still differences in

caste belonging, parents’ education and occupation that makes some of the day-scholars better

off than hostel students.

5 Data collection and Descriptive statistics

To collect the necessary data, I distributed a questionnaire to the students at the two schools.13

At Boyall HS all students present during the time of the distribution filled out the

questionnaire (i.e. 231 out of 291). At Sherman HS only the present students in 7th to 9

th grade

participated in the study (i.e. 360 out of 423). The questionnaire was distributed class wise in

the presence of me and some teachers, all questions being carefully explained to the students.

The names of the students were initially registered on the questionnaire to enable a match

with the students’ school achievement. Also, with the help of their names, I was able to revisit

the school and get clarifications on some of the answers. Afterwards, all names were

removed.

To measure the achievement, I used the results on the quarterly exams in Maths, English and

Telugu (Tamil in the Tamil section), written in October 2004. Every year students write

quarterly, half-yearly and annual exams for each subject. The results are presented in percent,

where 35 percent is the passing limit. These exams are provided by the state and are thus the

same for all schools. Data on the students’ initial level of performance, i.e. when entering the

school, was not available and therefore can not be controlled for.

Apart from the students’ achievement the following data were collected (table 1 and 2 shows

the descriptive statistics):

Child characteristics include age and sex. Since Sherman HS is a girls’ school and boys are

only allowed up until 7th grade there are only 16 boys included in the study conducted there.

Boyall HS, on the other hand, has an even distribution between boys and girls. Within the

group of hostel students though, only 27 percent of the participants are boys.

13 The questionnaire was given to them in the language in which they study, i.e. Telugu for the students in

Telugu medium and Tamil for the students in Tamil medium. See Appendix B for all questions given.

Page 16: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

16

Table 1: Descriptive statistics, Sherman Memorial Girls High School, Chittoor

Variable name

Definition Hostel students

(195 obs)

Mean S.D.

Day-scholar

students

(165 obs)

Mean S.D.

Achievement:

Tamil/Telugu

45.815

119.560

43.794

23.258

English 43.492 112.551 40.860 14.527

Maths

Score on the state provided quarterly exam

38.985 114.910 39.903 19.535

Child

characteristics:

Boy

.021

.142

.073

.260

Age 12.938 .865 13 .890

School

characteristics:

Grade

Grade, transformed into dummy variable when used in

regression.

7.918

.808

7.958

.844

Telugu section =1 if student is in Telugu section .508 .501 .836 .371

Different tongue =1 if student is studying in a different tongue than their

mother tongue

.359 .481 .309 .464

Family

characteristics:

Christian

=1 if Christian

.764

.426

.285

.651

SC/ST =1if student belongs to a schedule caste or tribe .810 .393 .339 .475

BC =1 if student belongs to a backward caste .138 .346 .394 .490

OC =1 if student belongs to an upper caste .051 .221 .267 .444

Family Average number in the household 6.467 3.234 5.679 1.566

Joined family =1 if family has more than nine members .072 .259 .042 .202

Father’s

education

Average years of schooling for father 6.627 4.145 6.307 3.472

Mother’s

education

Average years of schooling for mother 5.340 4.214 4.706 3.922

Father coolie =1 if father works as coolie .841 .367 .764 .426

Father business =1 if father is employed or has a business .005 .072 .127 .334

Father dead/

not supporting

=1 if father is not alive or for other reasons is not supporting

the family

.154

.362 .109

.313

Mother working =1 if mother works .574 .496 .315 .466

Mother dead =1 if mother is dead .010 .101 .006 .078

Mother not

working

=1 if mother does not work .415 .494 .673 .471

Village (km) Average distance from home to school (in km) 20.389 11.915 5.969 6.511

Hostel students:

Private hostel

=1 if student stays in a CSI hostel .785 .412

0

0

Government

hostel

=1 if student stays in a government hostel .215 .412 0 0

Private hostel

years

Average years in private hostel 2.415 1.612 0 0

Government

hostel years

Average years in government hostel 1.236 .549 0 0

The school characteristics collected are grade, and whether students in Sherman HS belong to

the Tamil or Telugu section. Since there is only one class in each grade at Boyall HS, all

teacher and peer group effects will be reflected in the grade variable. At Sherman HS some of

these effects will also be included in the dummy variable “Telugu section” which separates

the two sections at the school. The proportion of Tamil students is higher in the hostels than

for the Sherman school as a whole. The reason for this is that the school is registered as a

Page 17: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

17

Tamil medium with a Telugu section, so in order to keep the strength of Tamil students up,

the CSI turns to many Tamil speaking communities to inform about the possibility to stay in a

hostel. When studying the effects on exam scores in Telugu, a variable has been included for

students having a different mother tongue than the language studied.

Several variables have been included that concerns the Family background: OC, BC and

SC/ST are the caste belongings. Since I won’t be able to control fully for poverty, the caste

variable will reflect this to some extent. In Boyall HS, all hostel students but one belongs to

the SC/ST group and among the day-scholars 51 percent are SC/ST. At Sherman HS 81

percent of the hostel students are SC/ST and 44 percent of the day-scholar students.

Information on religion is also included to see whether Christians, a minority group in the

society with fewer privileges, perform differently from students having a Hindu background.

The majority of the hostel students in both schools are Christian which suggests that they are

prioritised, even though this is not an official priority.

The size of the household is measured in the variable “family”. This variable can be expected

to have a negative impact on achievement based on the theories on investment in children

presented above. Also, in large families a child (especially a girl) can be expected to

participate more in the household work. A matter of concern is that some families are joined

families, i.e. two or more related families living together, which could affect the achievement

in a different way. To correct for this I will include a dummy variable that equals one for

families having more than nine members.14

I have collected information both about the fathers’ as well as the mothers’ years of schooling

with the expectation that their schooling will have a positive effect on the children’s

achievement. Most of the students at these schools are the first literate generation in their

family. Even though many parents have had some education, very few have completed

primary school. Nearly all children knew the level of their parents’ education when filling out

the questionnaire. Only a few who had lost a parent left the question unanswered. In those

rare cases, the educational level has been set to zero with the assumption that the parent died a

long time ago, and thus has not affected the child’s achievement in school.

14 This size has been chosen after consultation with staff at the hostels.

Page 18: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

18

Table 2: Descriptive statistics, Boyall Memorial High School, Kollagunta

Variable name

Definition

Hostel students

(106 obs)

Mean S.D.

Day-scholar

students

(124 obs)

Mean S.D.

Achievement:

Telugu

32.792

14.006

36.379

20.169

English 28.406 12.020 28.952 12.253

Maths

Score on the state provided quarterly exam

24.123 15.098 28.339 19.039

Child

characteristics:

Boy

.274

.448

.540

.500

Age

13.396 1.277 13.637 1.558

School

characteristics:

Grade

Grade, transformed into dummy variable when used in

regression.

8.311

1.35

8.403

1.396

Different tongue =1 if student is studying in a different tongue than their

mother tongue

.943

.232 .780

.456

Family

characteristics:

Christian

=1 if Christian

.972

.167

.363

.483

SC/ST =1if student belongs to a schedule caste or tribe .991 .097 .508 .502

BC =1 if student belongs to a backward caste .009 .097 .290 .456

OC =1 if student belongs to an upper caste 0 0 .202 .403

Family Average number in the household 5.915 2.048 5.298 1.525

Joined family =1 if family has more than nine members .066 .250 .024 .154

Father’s

education

Average years of schooling for father 3.924 4.160 4.789 3.733

Mother’s

education

Average years of schooling for mother 2.495 3.627 2.815 3.450

Father coolie =1 if father works as coolie .840 .369 .782 .414

Father business =1 if father is employed or has a business .009 .097 .105 .308

Father dead/

not supporting

=1 if father is not alive or for other reasons is not supporting

the family

.151 .360 .113 .318

Mother working =1 if mother works .877 .330 .565 .498

Mother dead =1 if mother is dead .009 .097 .024 .154

Mother not

working

=1 if mother does not work .113 .318 .411 .494

Village (km) Average distance from home to school (in km) 20.066 16.736 2.39 3.435

Hostel students:

Hostel

=1 if student stays in a CSI hostel

1

0

0

0

Hostel years Average years in hostel 3.755 1.358 0 0

Information on whether the parents are alive, living in the household and supporting

financially was also collected. Where the above criteria were fulfilled, information on

occupation was also included. In India, the man in the house has the responsibility to support

the family, but often both parents have to work. The working conditions, especially for coolie

workers, are hard and many are worn out in an early age or are unable to work due to

sickness. A recurrent problem among poor communities all over the world is also male

drinking problems, abuse and abandonment. The villages in Chittoor district are no

exceptions. The data collected are therefore important in determining the students’ social

Page 19: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

19

background. When the father does not support the family, the child might have to take a larger

responsibility in the house, giving them less time to study.

A working mother might also lead to more responsibility for the child, since household work

normally is the mother’s duties. It could be argued though, that female labour participation

could have a positive impact on the child’s education under the assumption that the mother

gets higher bargaining power in household decisions (see e.g. Jayachandran 2002). Studies

show that mothers care more about the education of children, so this being true, the effect

could be positive with the child receiving more moral support in his or her studies. However,

staff at the SAC-CCC, as well as managers at the hostels, point out that many uneducated

parents do not know the value of education, and will not give priority to their children’s

education.

With the information above, three variables have been created for the fathers and the mothers

respectively. For fathers, one variable has been created for if they are not alive, or if they are

not supporting the family15. The occupations have been divided into two variables; one for

coolie work and one for if the fathers have their own business, alternatively are employed.16

As seen in table 1 and 2, about 80 percent of the fathers are labelled coolie workers (the

percentage is higher in the group of hostel children in both schools, but the difference is less

noticeable in the rural setting). For most of them, especially if they come from rural areas,

coolie work is synonymous with agricultural labour.

Since females don’t have the economic responsibility and many stay in the house, only one

variable has been created for if the mother was working – regardless of the type of work – and

one variable for doing housework.17 A third variable has been created for mothers no longer

alive.

Data on the village name has also been collected and with this information the distance to

school has been estimated.18 With a school located in a central place (such as a town), the

assumption is that the farther away from the school you live, the lower your socioeconomic

15 The two groups showed similar effects on the dependent variables, making it possible to use only one

variable. 16 See appendix A:III for a list of occupations belonging to each variable 17 Only eight women were employed; thus these were included in the same variable as the coolie working

women. 18 For students living in the same area as the school, one kilometre has been set.

Page 20: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

20

status. This might not be true in the sample from Boyall HS since Kollagunta is a rural area

and no such central place exists.

For students staying in a hostel, questions have been asked about the name of the hostel and

the number of years they have stayed there. The hostels in Kollagunta sometimes allow needy

children all the way down to 4th grade to stay in the hostel. This is also true for the

government hostels in Chittoor. Since the focus of this study is on private hostels, and I only

have second hand information about the way government hostels are run, I will separate the

effects of the government hostels from the private hostels when running regressions for the

Sherman sample.

6 Model

To measure the effects of staying in a hostel, the students’ exam scores will be used as the

dependent variable, y, in an OLS regression model which includes the variables presented

above – all which I believe can have an impact on the grades.19 Considering the different

settings of the two schools where the study has been conducted, I will run regressions

separately for each school. For purposes of presenting the model, let me here collect all the

information about the child characteristics, as well as the school and family characteristics, in

a variable X. To separate the hostel students from the day-scholars – i.e. the comparison group

– a dummy, H, will be used where H=1 for a hostel child and H= 0 for a child in the

comparison group. (In the sample from Sherman HS two dummies will be used; one for

private and one for government hostel students.)

The basic regression model is then:

iiii XHy 1

'

111 εγβα +++= (1)

The difficulty with this regression function is collecting all necessary data that should be

included in X. Since the information on family background comes from interviewing

children, some information will be left out of the regression either due to measurement error

or that some variable is omitted. One way of trying to capture additional information about

the socioeconomic background that is not observed in the other variables, is to include a

19 The variables father coolie, mother coolie, SC/ST and one grade will be left out to avoid perfect

multicollinearity.

Page 21: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

21

distance variable in the model. This will be done as a second step in the analysis. As

explained earlier, the distance variable contains information about the distance from the

child’s village to the school. The assumption is that school achievement is correlated with the

distance to school, i.e. the farther away from the centrally placed school you live, the lower

your socioeconomic status. According to the human capital theories your school achievement

may thus be lower. It is reasonable to expect such a pattern, at least when it comes to the

study conducted in Chittoor town where differences in coolie wages was confirmed by people

interviewed. The following model, including distance to school, is therefore run in a second

stage:

iiiii XdHy 2

'

2222 εγφβα ++++= (2)

Statistical tests have been conducted to make sure that the model is correctly specified. The

assumption is that the relationship between the dependent variable and the explanatory

variables is linear, and no evidence has been found that goes against this assumption of

linearity. In the regressions, Huber-White robust estimates of standard errors have been used.

Compared with ordinary standard errors, these robust standard errors can better deal with

concerns about failure to meet assumptions, such as problems about normality,

heteroscedasticity, or unusual single observations.

In a general case, with hostel admittances partly based on a knowledge test, we would be

concerned about the hostel variable being endogenous in the regression function (since

children who are likely to do well in school then would be overrepresented in the group of

hostel students). Based on the description of admittances and the differences between the

hostel students and the control group (section 4.3-4.4), I claim that this admittance test, rather

than biasing the estimates, makes the two groups better suited for comparison. Without the

selection done by the CSI, the average achievement of the hostel students when entering the

school would be noticeably lower than that of the day-scholars. Now, teachers and head

masters observe no difference in initial achievement between hostel students and day-scholars

coming from semi-urban or rural settings. Another factor that could make the hostel variable

endogenous is “family motivation”, i.e. that families applying for a hostel seat are more

concerned about their children’s education than other parents, and by that having a positive

impact on the children’s achievement. Again, I believe that this is not the case. Even for

families living within walking distance to school, going to school is accompanied with

expenses such as school material, uniform and school lunch; expenses that can be difficult for

Page 22: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

22

families to pay even if better off than the families eligible for hostel. The day-scholars are

therefore likely to be equally motivated to study.

Still, there is a chance that students with comparably very low academic achievement are

overrepresented in the group of day-scholars; this since there is no entrance test that sort out

these students from the comparison group. Appendix C shows the distribution of exam scores

for hostel students and day-scholars respectively at both schools. Since there is a larger group

of day-scholars receiving the lowest exam scores, this could be an indication of initial

differences between the students’ achievement. Therefore, as a sensitivity check, I have

chosen to run a regression excluding the observations below the 10th percentile with regard to

the exam scores for each subject respectively. Any large differences between the two groups

concerning low academic achievement will then be eliminated.

If staying in a hostel has an effect on the students’ achievement it will also be of interest to

see whether this effect is noticeable in the number of years a student has stayed in the hostel.

If hostels are good for student performance, it seems plausible to assume that the achievement

will be better the longer a child has stayed in a hostel. By running a regression including

information on the years, I will test for this effect.

In table 3 the correlations between the three subjects used as dependent variables is presented.

Since the correlations are substantially below unity, it is relevant to run regressions for all

three subjects since the result could turn out differently on different subjects.

Table 3: Correlation between the dependent variables

Boyall HS (230 obs)

Telugu

English Maths

Telugu 1.00

English 0.42 1.00

Maths 0.72 0.56 1.00

Sherman HS (360 obs)

Tamil/

Telugu

English Maths

Tamil/Telugu 1.00

English 0.80 1.00

Maths 0.74 0.69 1.00

Page 23: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

23

7 Results

The main results from the first OLS regression on the exam scores for students at Sherman

HS and Boyall HS are presented in table 4. The model being used here suffers from high

standard errors. This is due to the small sample size in relation to the many explanatory

variables included in the model and also to the fact that some of the explanatory variables,

especially concerning the socioeconomic background, are only approximations of the actual

phenomena.

For all three subjects in both Sherman HS and Boyall HS, private hostels have a positive

effect on the achievement. The result is statistically significant at the one percent level in all

cases except for Maths and English at Boyall HS where it is significant at the five percent

level. In the Boyall HS sample, the results suggest an increase in exam scores with around 5-6

percentage points when staying in a hostel, whereas the effect seems to be considerably

higher at Sherman HS with as much as 16 percentage points effect for Tamil and Telugu.

While being in a private hostel seems to promote student performance, there is no significant

effect for government hostels. This indicates that these hostels – or their student selection –

differ from the private hostels. This will be given more attention further on.

Children from the highest group of castes, OC, have significantly better test results compared

to SC and ST. The extent of the association varies between 7 and 17 percentage points, with

English being the subject least affected. The same pattern can be seen for the backward castes,

even though the coefficients are not as large and not as significant (at Boyall HS there is no

significance for English). Religious belonging also seems to be associated with school

achievement, suggesting that the socioeconomic background differs depending on not only

caste but also which religious community a person belongs to. The group of Christians has a

lower achievement in Telugu and English at Boyall HS, than the majority group of Hindus.

This could be explained by the fact that they belong to an underprivileged group in the

society.

The variable for family size is significant only in two cases, but an additional family member

seems to have a negative effect as expected. This is not true when it comes to extended

families as seen in the variable “Joined family”. Having a family with more than nine

members is associated with a better achievement for both Telugu and Maths at Boyall HS.

Page 24: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

24

Parents’ education only has a significant effect on English and Maths (the effect is seen for

father’s education except for Maths at Boyall HS where the mother’s educational level seems

to matter). Though significant, the effect is very small suggesting on average a half

percentage point increase of the test results with every additional year of parent schooling.

This is perhaps not surprising considering the low level of parents’ education at the two

schools. Instead the occupation seems to explain more of the differences in test result,

implying that students with employed fathers, or fathers running their own business, receive

significantly higher results in Telugu or Tamil than students having coolie working fathers.

The variable for mothers not working is only significant in one case (for English in Boyall

HS) and is there showing a negative effect compared to working mothers.

The variable for fathers no longer alive or not supporting the family is not significant in any

of the regressions. This does not change if these two “events” are defined as separate

variables, suggesting that there is little difference between those families and families with

coolie working fathers. Having lost a mother seems to affect the achievement negatively,

showing significant and rather large effects for Maths at Boyall HS. It needs to be stressed

though that the number of observations where this variable equals one is small and therefore

too strong conclusions should not be drawn.

As mentioned earlier, distance to school can proxy unobserved family and socioeconomic

characteristics. The results from the extended model, with distance to school included, are

presented in table 5. The expectation would be a negative correlation between the distance to

school variable and the dependent variable if students living farther away from the central

areas have poorer socioeconomic background. In the sample from Sherman HS, distance

seems to capture factors in the social background of the children that have not been possible

to control for with the other variables. The result is very modest in magnitude, but significant

for Maths and Tamil/Telugu. At the same time, the effect of private hostel on the test scores

increases. In the Boyall HS sample there is no sign of correlation between the distance and

test scores. This is not too surprising considering that Kollagunta, where the school is

situated, is defined as a rural area. Hence, there is no reason to expect, e.g., a higher income

for people staying in Kollagunta, compared to people staying in rural areas surrounding this

village.

Page 25: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

25

Table 4: OLS estimates of quarterly exam percentage scores for students in both schools

Sherman HS, Chittoor Boyall HS, Kollagunta

Tamil/Telugu English Maths Telugu English Maths

Age -.716

(2.219)

.357

(1.494)

2.479

(2.002)

-1.561

(1.292)

.211

(.892)

-2.257

(1.736)

Boy 1.592

(5.950)

4.354

(4.201)

8.004*

(4.637)

-.680

(1.836)

1.624

(1.626)

3.652

(2.269)

Different tongue -3.669

(3.072)

-

-

-.615

(2.707)

- -

Grade 6 - - - -31.688***

(6.015)

10.272**

(5.113)

-18.958**

(7.608)

Grade 7 -1.101

(5.071)

-1.982

(3.194)

12.808***

(4.160)

-19.830***

(5.107)

7.509*

(3.984)

-11.757*

(6.258)

Grade 8 -8.256**

(3.230)

-2.792

(2.148)

5.322*

(2.723)

-17.218***

(3.206)

-.350

(2.590)

-12.915***

(4.053)

Grade 9 - - - -4.361

(2.932)

-1.011

(2.295)

-2.318

(3.679)

Private hostel 16.139***

(3.011)

8.926***

(1.950)

9.312***

(2.506)

6.057***

(2.234)

5.405**

(2.411)

5.907**

(2.708)

Government

hostel

.026

(3.267)

-.541

(1.936)

-1.654

(2.919)

- - -

Telugu section 4.713

(2.993)

-.955

(1.615)

5.271***

(2.039)

- - -

BC 10.739***

(2.714)

4.450**

(1.754)

6.104***

(2.286)

6.211*

(3.417)

3.791

(2.899)

7.578*

(4.468)

OC 13.008***

(3.589)

6.932***

(2.259)

11.894***

(3.156)

14.293***

(3.913)

9.022***

(3.134)

17.07***

(4.917)

Christian -2.070

(1.963)

-1.188

(1.225)

-1.639

(1.967)

-6.317**

(2.747)

-6.022**

(2.874)

-4.975

(3.961)

Family -1.180*

(.686)

-.616*

(.338)

-.273

(.494)

-.799

(.698)

.2139

(.628)

-1.046

(.790)

Joined family 10.449

(7.989)

1.67

(4.751)

1.716

(6.175)

11.347**

(5.768)

1.499

(5.131)

16.433**

(6.630)

Father’s education .502

(.3060)

.420**

(.185)

.556**

(.239)

.298

(.235)

.497**

(.235)

.125

(.282)

Mother’s

education

.007

(.290)

.030

(.193)

-.331

(.241)

.328

(.278)

.145

(.244)

.549*

(.311)

Father business 12.753**

(6.289)

5.258

(4.305)

5.459

(5.166)

7.341*

(4.287)

3.264

(2.778)

7.459

(4.593)

Father dead/not

supporting

2.522

(2.880)

2.105

(1.969)

.928

(2.555)

1.428

(2.277)

2.718

(2.248)

2.283

(2.753)

Mother not

working

1.427

(2.211)

.930

(1.504)

.311

(1.894)

-.512

(2.156)

-4.770**

(2.071)

.1714

(2.604)

Mother dead -8.728

(7.453)

-8.716

(7.113)

-14.654

(10.270)

-5.881

(4.918)

-1.922

(5.765)

-13.854**

(6.107)

Constant 46.549

(32.107)

34.396

(21.403)

-9.984

(28.508)

68.834***

(20.393)

20.053

(13.955)

63.120**

(26.453)

Observations 356 356 356 228 228 228

R-squared 0.21 0.16 0.18 0.54 0.23 0.32

F statistic 5.72 4.59 3.98 14.39 3.66 6.92

Prob>F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9 for Sherman HS and

10 for Boyall HS.

Page 26: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

26

Table 5: OLS with Distance included

Sherman HS, Chittoor Boyall HS, Kollagunta

Tamil/Telugu English Maths Telugu English Maths

Age -1.189

(2.214)

.174

(1.521)

2.137

(2.063)

-1.622

(1.295)

.213

(.891)

-2.320

(1.756)

Boy 2.230

(6.229)

5.113

(4.515)

9.168*

(4.840)

-.495

(1.837)

1.619

(1.646)

3.841*

(2.278)

Different tongue -3.885

(3.062)

- - -.632

(2.711)

- -

Grade 6 - - - -32.001***

(6.029)

10.280**

(5.136)

-19.279**

(7.687)

Grade 7 -2.172

(5.035)

-2.318

(3.211)

12.104***

(4.235)

-20.261***

(5.117)

7.520*

(3.971)

-12.197*

(6.297)

Grade 8 -9.326***

(3.229)

-3.121

(2.205)

4.719*

(2.800)

-17.489***

(3.210)

-.343

(2.587)

-13.192***

(4.126)

Grade 9 - - - -4.723

(2.905)

-1.002

(2.289)

-2.688

(3.700)

Private hostel 18.688***

(3.296)

10.374***

(2.148)

11.172***

(2.626)

4.809**

(2.429)

5.438**

(2.681)

4.633

(3.026)

Government hostel 1.284

(3.409)

.523

(2.122)

-.493

(3.076)

- - -

Telugu section 4.545

(3.006)

-1.130

(1.640)

5.072**

(2.037)

- - -

BC 11.303***

(2.749)

4.695***

(1.810)

6.568***

(2.350)

6.126*

(3.387)

3.794

(2.914)

7.492*

(4.466)

OC 13.110***

(3.661)

7.049***

(2.326)

11.964***

(3.183)

14.265***

(3.901)

9.023***

(3.143)

17.039***

(4.919)

Christian -1.876

(1.949)

-1.181

(1.232)

-1.623

(2.008)

-6.504**

(2.747)

-6.017**

(2.899)

-5.167

(3.958)

Family -1.326*

(.678)

-.652

(.351)

-.361

(.501)

-.720

(.705)

.212

(.625)

-.964

(.789)

Joined family 11.277

(8.270)

1.857

(5.021)

2.717

(6.552)

11.412*

(5.832)

1.497

(5.156)

16.499**

(6.655)

Father’s education .610**

(.305)

.454**

(.187)

.609**

(.242)

.271

(.238)

.498

(.237)

.098

(.282)

Mother’s education -.032

(.289)

.020

(.194)

-.347

(.242)

.366

(.275)

.144

(.245)

.589*

(.306)

Father business 12.111*

(6.296)

5.072

(4.319)

5.093

(5.177)

7.430

(4.305)

3.262

(2.788)

7.556

(4.619)

Father dead/not

supporting

1.930

(2.959)

1.905

(2.001)

.488

(2.609)

1.468

(2.254)

2.716

(2.256)

2.324

(2.76)

Mother not working 1.396

(2.237)

.876

(1.529)

.308

(1.925)

-.313

(2.184)

-4.775**

(2.089)

.377

(2.632)

Mother dead -9.899

(7.870)

-9.377

(7.388)

-15.451

(10.619)

-6.233

(4.952)

-1.913

(5.786)

-14.213**

(6.195)

Distance in km -.193**

(.097)

-.105

(.071)

-.138*

(.077)

.078

(.066)

-.002

(.059)

.079

(.092)

Constant 55.239

(32.051)

37.830*

(21.842)

-3.816

(29.491)

69.245***

(20.415)

20.042

(13.980)

63.537**

(26.680)

Observations 349 349 349 228 228 228

R-squared 0.23 0.17 0.19 0.55 0.23 0.33

F statistic 6.83 4.49 4.10 13.57 3.45 6.53

Prob>F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9 for Sherman HS and

10 for Boyall HS.

Page 27: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

27

The modest effect seen in the Sherman sample could possibly be explained by the fact that

many hostel students come from very far

distances. The correlation between the

distance from town and the socioeconomic

background can only be assumed to hold for a

certain distance. When describing the students

at Sherman HS in section 4.4, the urban and

semi-urban areas were pointed out as

supposedly better off areas with for instance

opportunities for higher wages. In the rural

areas, outside of these two areas we can no

longer assume that distance to town will

explain possible differences in the socioeconomic background. Some students staying very far

away might even have another town closer by their village, but with a limited number of

hostel opportunities they are bound to study farther away. In an attempt to capture more of the

distance effect, I have thus in table 6 substituted the distance variable in the Sherman HS

sample for two new dummy variables reflecting the two areas “urban” and “semi-urban”.

“Urban” citizens will be families living within three kilometres from the school whereas

“semi-urban” citizens will be families living between four and ten kilometres from the school

(see figure 1).

Table 6: Part of OLS regression for Sherman HS, showing distance dummies.20

Tamil/Telugu English Maths

… … … …

Private hostel 18.279***

(3.111)

10.562***

(2.013)

11.045***

(2.507)

Government hostel 2.084

(3.469)

1.082

(2.073)

-.371

(2.982)

… … … …

Urban area 5.744*

(2.949)

4.399**

(2.061)

5.080**

(2.492)

Semi-urban area -1.318

(2.679)

-.740

(1.786)

-3.192

(2.165)

Observations 356 356 356

R-squared 0.22 0.18 0.20

Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%.

20 For the full regression; see appendix D.

Urban area,

< 4 km to

school

Semi-urban

area,

4-10 km to

school

Rural area,

> 10 km to

school

Figure 1: Distance to Sherman HS

Page 28: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

28

The results show that the positive coefficients for the urban area are significant for all three

subjects suggesting that the assumption on distance holds to some extent. For the semi-urban

area there is no significant effect. This could be due to measurement errors making the

variable not well defined; several villages have received an approximate distance to school

and also the 10 kilometre limit could be too broad to catch any semi-urban effect. Still, adding

these dummy variables to the Sherman HS gives a higher explanatory power to the model.

As pointed out previously, part of the positive effects seen for the private hostel variable

could be a result of differences in the comparison groups that have not been taken into

consideration. To examine this, all observations below the 10th percentile have been excluded

for each of the three subjects. The aim is to exempt students with very low academic

achievement that might not have passed the entrance test if applying to a hostel, thereby

making the groups better for comparison. The results are presented in table 7 (for Boyall HS

the first regression equation has been used since the information on distance was

insignificant). For all subjects there is a reduced positive effect of staying in a private hostel,

compared to previous results. The largest change in percentage points can be seen for

Tamil/Telugu in Sherman HS where the effect of staying in private hostel decreases with

about 3 percentage points. In Boyall HS the positive effect on English and Maths turns

insignificant. Perhaps the chosen level of reduction in observations is rather large considering

that the entrance test is a very basic test; the reason for the smaller number of low achieving

students in the hostels, compared to the day-scholars, could just be reflecting the positive

effects of these hostels. Still, the main results are not considerably changed, making it

reasonable to believe the arguments laid out in section 6 that endogenity is not of major

concern in the model.

Page 29: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

29

Table 7: OLS excluding observations with an exam score below the 10th percentile

Sherman HS, Chittoor Boyall HS, Kollagunta

Tamil/Telug

u

English Maths Telugu English Maths

Age -1.765

(2.091)

.018

(1.288)

.011

(2.033)

-1.529

(1.359)

-.279

(.878)

-1.014

(1.686)

Boy 3.170

(5.193)

6.065

(3.871)

7.289

(4.516)

-.396

(1.897)

2.376

(1.589)

3.534

(2.353)

Different tongue -1.127

(2.896) - -

-.200

(2.680) - -

Grade 6 - - -

-24.458***

(6.252)

8.246*

(4.882)

-14.456*

(7.483)

Grade 7 -.123

(4.664)

-.400

(2.687)

5.026

(4.103)

-17.685***

(5.085)

3.812

(3.914)

-8.807

(6.102)

Grade 8 -5.415*

(2.943)

-.270

(1.840)

3.627

(2.636)

-17.366***

(3.274)

-1.636

(2.435)

-10.429***

(3.892)

Grade 9 - - -

-4.181

(2.958)

-1.549

(2.238)

-2.244

(3.674)

Private hostel 14.869***

(3.077)

9.290***

(1.803)

8.355***

(2.423)

5.127**

(2.407)

3.200

(2.381)

2.408

(2.774)

Government hostel 1.014

(3.583)

.1571

(1.794)

-.362

(2.988) - - -

Telugu section 4.015

(2.851)

.138

(1.422)

4.908***

(1.871) - - -

BC 7.782***

(2.620)

1.920

(1.550)

4.958**

(2.101)

5.728*

(3.397)

2.999

(2.785)

3.761

(4.626)

OC 10.887***

(3.216)

3.988*

(2.093)

10.924***

(2.928)

13.012***

(4.057)

9.038***

(2.814)

13.117***

(5.067)

Christian -1.371

(2.037)

-1.121

(1.015)

-.457

(1.915)

-5.604*

(2.911)

-2.84

(2.644)

-3.669

(4.114)

Family -1.023

(.701)

-.704**

(.290)

-.265

(.462)

-.704

(.694)

.572

(.572)

-.943

(.822)

Joined family 12.413

(8.351)

5.026

(4.214)

3.408

(6.303)

11.322**

(5.663)

.619

(4.806)

19.461***

(6.481)

Father’s education .372

(.287)

.235

(.172)

.299

(.233)

.146

(.247)

.365

(.234)

.000

(.284)

Mother’s education -.145

(.265)

.091

(.174)

-.249

(.224)

.623**

(.293)

.188

(.241)

.682**

(.326)

Father business 16.150***

(5.645)

7.453**

(3.497)

8.196*

(4.828)

9.340**

(3.992)

3.474

(2.893)

8.371*

(4.815)

Father dead/not

supporting 2.773

(2.699)

1.830

(1.853)

1.271

(2.397)

1.395

(2.171)

3.414

(2.193)

4.146

(2.655)

Mother not working 3.499*

(2.110)

2.384*

(1.360)

.7663

(1.824)

.770

(2.322)

-4.168**

(1.978)

1.031

(2.739)

Mother dead -7.003

(8.224)

.553

(7.809)

-5.599

(11.609)

-7.911*

(4.717)

-3.120

(4.715)

-15.629***

(5.273)

Urban area 5.094*

(2.757)

5.856**

(1.911)

5.095**

(2.458) - - -

Semi-urban area -1.781

(2.589)

-.469

(1.677)

-2.884

(2.173) - -

-

Constant 61.097**

(30.383)

38.658**

(18.334)

28.285

(28.666)

67.667***

(21.182)

26.519*

(13.648)

47.580*

(25.553)

Observations 320 322 322 204 206 206

R-squared 0.20 0.20 0.15 0.48 0.22 0.30

F statistic 4.40 5.66 3.07 9.88 3.00 6.43

Prob>F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9 for Sherman HS and

10 for Boyall HS.

Page 30: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

30

So far we have seen the effect of staying in a hostel measured as an average of the number of

years that the students have stayed there. Under the assumption that there is no difference

between students based on the grade in which they were admitted, we should see not only a

positive effect of staying in a hostel, but also an effect that is increasing with the length of the

hostel stay in the two samples.21 To estimate the effect of the number of years in a hostel, an

OLS regression has been run, including a variable with this information. Table 8 shows that

the effect of an additional year in a hostel is only statistically significant in one case, i.e. for

Maths in the Sherman HS sample where an additional year raises the scores with around 2

percentage points. The coefficient for hostel, which previously was showing the full effect of

staying in a hostel, is here showing the initial effect, i.e. the effect of the hostel given that the

student has stayed there less than one year. The effect is positive for all subjects and

statistically significant in all cases except for Maths in Sherman HS. Since the students are

admitted in the beginning of July (when the school year begins) and this study was conducted

during January and February, all students have included the ongoing year when stating the

length of their stay. With the exam scores written in October, this means that the initial effect

is not a full year, but only around four months. It is not possible to say whether the effect seen

is only the result of these first months, or if it is an indication of selection bias, (i.e. that

students admitted to the hostel perform better initially than day-scholars having a similar

socioeconomic background).

The insignificant effects in the number of years stayed in a hostel could be due to several

factors which are difficult to detect when only observing the average annual effect. Therefore,

in table 9 the years stayed in a private hostel have been included as dummy variables in the

model, showing the effect for each cohort of students admitted the same year.22 The effects

seen in table 8 and 9 are also illustrated in figures I-VI in appendix E.

21 The majority of the students are admitted when beginning grade 6, but a few have begun in grade 7 or 8. In

Kollagunta some students are also admitted in grade 4 or 5. 22 Observe that each year is still reflecting an ongoing year; e.g. the variable “hostel 1 yr” shows the effect of the

first 4 months, and “hostel 2 yrs” shows the effect of having begun the second year.

Page 31: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

31

Table 8: OLS including hostel years

Sherman HS Boyall HS

Tamil/Telugu English Maths Telugu English Maths

Age -1.155

(2.155)

.115

(1.479)

2.030

(2.012)

-1.619

(1.325)

.137

(.918)

-2.3174

(1.744)

Boy 1.649

(5.986)

4.623

(4.219)

8.272*

(4.660)

-.776

(1.842)

1.513

(1.620)

3.561

(2.262)

Different tongue -3.742

(3.032)

- - -.689

(2.741)

- -

Grade 6 - - - -33.828***

(6.368)

7.686

(5.629)

-21.082***

(8.035)

Grade 7 -.261

(5.084)

-2.405

(3.201)

13.918***

(4.312)

-21.201***

(5.223)

5.858

(4.428)

-13.113**

(6.392)

Grade 8 -7.872**

(3.297)

-2.961

(2.204)

5.991**

(2.774)

-17.843***

(3.265)

-1.106

(2.723)

-13.536***

(4.124)

Grade 9 - - - -4.787

(2.928)

-1.528

(2.354)

-2.742

(3.714)

Private hostel 12.728**

(5.996)

11.586***

(3.709)

4.657

(4.889)

9.631**

(4.764)

9.699*

(5.548)

9.433*

(4.937)

Private hostel years 1.940

(1.571)

-.379

(.985)

2.209*

(1.254)

-1.053

(1.253)

-1.268

(1.284)

-1.041

(1.306)

Government hostel 3.796

(5.251)

-1.411

(3.339)

-1.820

(5.074)

- - -

Government hostel years -.529 (1.331)

.9337

(.807)

.716

(1.275)

- - -

Telugu section 4.614

(2.985)

-1.243

(1.586)

5.002**

(1.989)

- - -

BC 9.901***

(2.721)

3.729**

(1.736)

5.347**

(2.229)

6.005*

(3.420)

3.550

(2.867)

7.379

(4.521)

OC 13.501***

(3.556)

7.250***

(2.278)

12.533***

(3.133)

13.848***

(3.928)

8.470***

(3.079)

16.617***

(5.016)

Christian -2.381

(1.897)

-1.110

(1.206)

-2.037

(1.952)

-5.967**

(2.717)

-5.613*

(2.966)

-4.639

(3.973)

Family -1.272*

(.6595)

-.659*

(.340)

-.390

(.473)

-.829

(.711)

.179

(.646)

-1.074

(.802)

Joined family 11.684

(8.021)

2.786

(4.765)

3.172

(6.123)

11.703**

(5.835)

1.927

(5.125)

16.785

(6.694)

Father’s education .443

(.306)

.408**

(.184)

.483**

(.236)

.288

(.238)

.486**

(.233)

.116

(.284)

Mother’s education .023

(.286)

.034

(.192)

-.317

(.234)

.335

(.278)

.154

(.241)

.557*

(.310)

Father business 12.677**

(6.301)

5.003

(4.264)

5.561

(5.049)

7.081*

(4.268)

2.983

(2.778)

7.2281

(4.601)

Father dead/not

supporting

2.264

(2.918)

1.805

(1.993)

.466

(2.577)

1.482

(2.313)

2.784

(2.293)

2.338

(2.776)

Mother not working 1.471

(2.198)

.713

(1.488)

.474

(1.896)

-.446

(2.178)

-4.675**

(2.099)

.249

(2.625)

Mother dead -6.678

(7.845)

-6.945

(7.552)

-11.134

(11.084)

-6.153

(4.917)

-2.249

(5.563)

-14.123**

(5.946)

Urban area 5.944**

(2.987)

4.270**

(2.082)

5.176**

(2.516)

- - -

Semi-urban area -1.216

(2.685)

-.739

(1.783)

-3.035

(2.167)

- - -

Constant 50.365

(30.983)

36.674

(21.009)

-5.031

(28.386)

70.640***

(20.871)

22.216

(14.246)

64.896**

(26.613)

Observations 356 356 356 228 228 228

R-squared 0.22 0.18 0.20 0.55 0.23 0.32

F statistic 5.76 4.33 4.41 14.31 3.62 6.67

Prob>F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9 for Sherman HS and

10 for Boyall HS.

Page 32: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

32

For the hostels in Kollagunta the pattern is similar for all subjects: Dummy coefficients for

the second as well as the fifth and sixth year show no significance and seem to deviate from

the expected positive relationship between an additional year in hostel and an improved

achievement. The dummy variable for the second year only includes seven observations, out

of which three are rather extreme, which could explain this deviation. Even the variable for

the fifth and sixth year seems to hold many extreme observations. One explanation could be

that children admitted earlier than grade 6 have a relatively poorer unobserved family

background than children admitted later on (nearly half of the observations in this cohort of

students were admitted early, which is not the case for the other cohorts). When only

observing the significant effects for having stayed three and four years in hostel, the expected

pattern can be seen for both Telugu and Maths.

The positive effect of an additional year in hostel is more evident in the Sherman sample.23

When observing the results for Maths, the positive effect of an additional year seems to still

hold; even though the cohort of students who have stayed four years in a hostel don’t perform

as well as would be expected. This cohort differs from the others by having a larger number

of students lacking father support (25 percent of the group, compared to around 10 percent for

the others). Even though this has been controlled for, there could still be unobserved factors

that make the students who lack father support worse off than others. The effects seen for

Tamil and Telugu also follow the expected pattern, with the exception for the cohort of

second year students, where the effect is very high. Taking into consideration the possibly

worse off students in the cohort having stayed four years in hostel, and the higher performing

students in the cohort for two years in hostel, the results suggest that additional years benefit

the achievement in English as well. Thus, the lack of a significant effect for private hostel

years (in all cases but one) in the regression shown in table 8, seems possible to explain with

the deviation in some of the cohorts of students.

23 The yearly effect of government hostels have been left out here since the variable shows no significance and

consists of relatively few observations.

Page 33: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

33

Table 9: OLS using dummy variables for hostel years

Sherman HS Boyall HS

Tamil/Telugu English Maths Telugu English Maths

Age -1.078

(2.163)

.142

(1.473)

2.059

(1.979)

-1.862

(1.326)

.0758

(.918)

-2.516

(1.761)

Boy 2.378

(6.027)

5.114

(4.212)

8.573*

(4.695)

-1.438

(1.902)

1.159

(1.625)

3.085

(2.298)

Different tongue -4.039

(3.047)

- - -1.304

(2.725)

- -

Grade 6 - - - -34.693***

(6.374)

7.973

(5.581)

-21.750***

(8.163)

Grade 7 -1.155

(4.997)

-3.406

(3.182)

12.539***

(4.197)

-23.605***

(5.154)

5.584

(4.383)

-14.405**

(6.595)

Grade 8 -7.051**

(3.491)

-3.088

(2.461)

4.661

(2.972)

-20.125***

(3.359)

-2.218

(2.764)

-15.233***

(4.180)

Grade 9 - - - -7.181**

(2.965)

-2.261

(2.347)

-4.403

(3.804)

Hostel, 1 yr 14.108

(15.149)

10.511

(7.503)

8.380

(15.042)

5.917

(4.639)

7.373

(6.326)

7.171*

(4.209)

Hostel, 2 yrs 19.133***

(4.202)

12.426***

(2.699)

10.145***

(3.135)

3.002

(6.236)

1.761

(4.313)

1.710

(7.594)

Hostel, 3 yrs 15.788***

(3.999)

9.463***

(2.851)

12.031***

(3.045)

8.000**

(3.490)

8.660**

(4.064)

7.872*

(4.679)

Hostel, 4 yrs 18.385***

(4.517)

8.092***

(2.885)

8.683**

(3.669)

11.471***

(3.056)

7.041***

(2.674)

9.656**

(3.971)

Hostel, 5-6yrs 25.504***

(4.432)

11.219***

(3.060)

18.630***

(3.121)

1.472

(3.003)

2.362

(2.778)

2.221

(3.741)

Government hostel 2.339

(3.473)

.924

(2.093)

-.019

(2.998)

- - -

Telugu section 5.033*

(3.005)

-1.047

(1.598)

5.197***

(1.941)

- - -

BC 9.786***

(2.746)

3.684**

(1.761)

5.291**

(2.252)

5.732*

(3.399)

3.167

(2.922)

6.954

(4.636)

OC 13.638***

(3.460)

7.316***

(2.240)

12.449***

(3.108)

12.877***

(3.952)

7.942**

(3.129)

15.660***

(5.069)

Christian -2.228

(1.938)

-1.090

(1.239)

-2.065

(1.987)

-6.803**

(2.899)

-6.515**

(3.139)

-5.638

(4.145)

Family -1.356**

(.654)

-.692**

(.344)

-.442

(.472)

-.860

(.692)

.103

(.655)

-1.106

(.811)

Joined family 11.616

(8.096)

2.601

(4.828)

2.895

(6.170)

12.987**

(5.589)

2.944

(5.193)

17.840***

(6.880)

Father’s education .477

(.314)

.413**

(.188)

.464*

(.242)

.281

(.233)

.466**

(.231)

.099

(.287)

Mother’s education -.012

(.283)

.031

(.192)

-.305

(.232)

.496*

(.285)

.232

(.247)

.690**

(.317)

Father business 12.322*

(6.363)

4.774

(4.318)

5.388

(5.083)

6.706

(4.176)

3.177

(2.730)

7.260

(4.578)

Father dead/not

supporting

2.202

(2.930)

2.028

(2.028)

.878

(2.551)

.667

(2.334)

2.593

(2.320)

1.924

(2.860)

Mother not working 1.176

(2.209)

.632

(1.496)

.509

(1.919)

-.496

(2.118)

-4.732**

(2.095)

.2567

(2.631)

Mother dead -6.500

(7.230)

-7.399

(7.290)

-11.600

(10.890)

-4.858

(5.788)

-.808

(5.627)

-12.582**

(6.330)

Urban area 5.939**

(2.953)

4.356**

(2.075)

5.246**

(2.510)

- - -

Semi-urban area -1.386

(2.664)

-.883

(1.787)

-3.111

(2.161)

- - -

Constant 49.861

(30.890)

36.786*

(20.833)

-4.313

(27.795)

76.625***

(20.670)

24.436*

(14.093)

69.417***

(26.742)

Observations 356 356 356 228 228 228

R-squared 0.23 0.18 0.21 0.56 0.25 0.34

F statistic 5.35 4.05 4.25 12.98 3.51 5.98

Prob>F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9 for Sherman HS and

10 for Boyall HS.

Page 34: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

34

8 The effect of staying in a hostel

The above results show positive effects of staying in a private hostel in the samples from both

Sherman HS and Boyall HS. It also suggests a positive correlation between exam scores and

the number of years stayed in a private hostel (though this is more evident in the Sherman HS

sample), strengthening the support for the hypothesis of a positive hostel effect. Since I have

not been able to control for initial performance, conclusions from the estimated relationships

revealed in the study must be drawn with some caution. Yet, I have presented additional

information on the background of the students, received through interviews and my own

observations, making it reasonable to believe that there is a causal relationship between hostel

and school achievement. With these results, I will now proceed to discuss the factors behind

them as well as their relevance.

8.1 Life in the hostels

Based on the intergenerational mobility theories and earlier empirical evidence, there are

several reasons why staying in hostel such as the ones included in this study should have a

positive effect of a child’s achievement. First of all, in all hostels included in the study about

3-4 hours is set aside daily for studying - normally one hour in the morning and the rest in the

afternoon. As argued earlier, one reason why a larger family or an absent parent would have a

negative effect on the achievement is that the children’s time spent on duties at home would

increase, giving them less time to study. The children in the hostels spend around one hour

daily on duties such as cleaning and washing. In the survey conducted at Sherman HS many

day-scholars claimed that their daily duties took more than one hour – 27 out of 110 wrote

that they spend two or more hours per day working.24 Eight of the students also wrote in the

questionnaire that they spend five or six hours per day working. A student recently admitted

to a government hostel said in an interview that the main advantage with staying in the hostel

was that she could now focus on the home work instead of spending her time on household

work. The positive effect of staying in a hostel could then to some extent be explained by

giving the children the opportunity to study after school. However, this can far from fully

explain the positive results considering the insignificant effects of staying in a government

hostel. There are also rather large differences between the effects seen in the CSI hostels in

the Sherman HS sample (mainly the effect of one hostel) and the two hostels in the Boyall HS

sample.

24 Grade nine is not included since the question was misinterpreted by those students.

Page 35: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

35

In the CSI hostels, an educated adult (normally a child care worker) would be there to

overlook the studying and be of assistance when needed. Child care workers stay with the

children at all times, which makes it possible for the children to receive more help. Extra

tuition is also given in Maths (and sometimes Hindi) to students in grade 7 and 10, starting in

the spring, to prepare for the public exam conducted for these grades. At Sherman Memorial

Girls hostel, situated on the high school compound, the school and the hostel have good

cooperation; hence the children are able to receive help from the teachers even on other times

than school hours. The hostel does not provide tuition like the other hostels since the school

gives tuition free of charge for all school children as a preparation for the public exam. The

close contact between teachers and child care workers at the Sherman compound helps all

people involved to see the needs of the children and follow up each individual child. For

example, child-care workers from the hostel participate in the regular parent-teacher meetings

conducted by the school. Meetings with the parents are also conducted by the hostel once a

month. By comparison, parents are called for meetings twice a year to the hostels in

Kollagunta to discuss the child’s progress, and the cooperation between the hostel and the

school is not as strongly developed. The built-up network of teachers, hostel staff and parents

within the Sherman compound could explain the stronger effects seen in the Sherman sample

for private hostel students, compared to the effects seen for the private hostels in Kollagunta.

Also, part of the higher effect could be a result of a higher educational level at Sherman HS

compared to Boyall HS; e.g. with better educated teachers on average, and more interaction

with the students. This is noticed not only in the differences in average exam scores, but also

from my own class room and study time observations. It is reasonable to assume that the more

a student has been able to comprehend in the class room, the easier and more benefiting the

self studies will be.

The absence of significant effects in achievement for the government hostels could possibly

be explained by differences in admittances, compared to the private hostels. Since no entrance

test is conducted, it is reasonable to assume that there is a larger number of low performing

students among the government hostel students. Also, the average period of time stayed in the

government hostels is shorter compared to that of the private hostels, which makes a possible

positive effect of staying in these hostels more difficult to measure. There are, however, more

factors that could explain the differences in results between the government and the private

hostels. For the government hostels that I got information from, three out of four have

Page 36: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

36

supervised study time.25 In addition, tuition teachers are brought to the hostels to help with

Maths and Hindi. A difference in comparison with the CSI hostels is that the child care

workers normally do not stay in the hostels, but have regular office hours. Hence, the study

time is not always supervised by educated staff. Moreover, the contact with the school is not

as close as in the case of the Sherman Memorial Girls Hostel. According to the information I

received from interviewing government hostel students, none of the hostels have regular

meetings with parents and students where the school performance is discussed. In addition to

that, the parents do not pay a monthly fee, which is the case for all CSI hostels. According to

the CSI hostel staff, both in Kollagunta and in Chittoor, the parents are more interested in the

children’s performance and how the hostel is being run, when being requested to contribute to

the education – even though the fee is more or less symbolic in size. When they do not

contribute, some parents might be satisfied knowing that their child receives proper food and

shelter. Thus, what seem to be relevant factors for a positive hostel effect is not only giving

time for studying but also to follow up, encourage and motivate the students – both directly

and by involving their parents.

Apart from the study time set aside for the CSI hostel children and the presence of educated

adults, these children get nutritious food every day; something that is not available to the

same extent in the children’s homes26. Though I cannot measure the extent of this effect,

nutritious food can be claimed to have a positive effect on the children’s health and their

learning abilities. Time is also given for developing talents such as drawing, singing, dancing

and acting and learning skills including gardening and cooking. Leadership training is a

natural part of the hostel life since children are appointed as leaders for these activity groups

as well as for the assemblies held on a daily basis. They are also taught basic knowledge such

as taking care of their personal hygiene – knowledge that is often lacking when brought up in

an illiterate community. All these things are examples of development and could be said to

have an influence on the children’s learning abilities by raising the children’s self-esteem and

increasing their basic knowledge by exposing them to various different things27. Hence,

indirectly, the life in these hostels could have a positive impact on the children’s achievement.

25 Totally, six government hostel are included in the study. 26 This is based on interviews with hostel students and hostel staff, as well as my own experience from visiting

rural villages. 27 These hostel routines and activities are similar in most of the government hostels as well.

Page 37: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

37

8.2 The relevance of the effect

When comparing the effects between different subjects, the private hostel effect seems to be

particularly strong for Telugu and Tamil in Sherman HS. This phenomenon is not exclusive

for the hostel variable though; the pattern is to some extent similar if looking at differences in

achievement between castes, suggesting that it is not the hostels which give more attention to

one subject. Instead, part of the explanation could be that it is easier to improve the skills in

Telugu or Tamil than in the other subjects. However, this is not as evident when looking at

the results for the hostel effect in Boyall HS where the effects of the three subjects are rather

similar.

Perhaps more interesting is that in both samples, the CSI hostels seem to affect Maths and

English to a similar extent. The same pattern is not seen in the caste variables where the effect

on English is lower than the effect on Maths in both school samples. This could indicate that

the hostels are more benefiting for the English performance, perhaps due to the fact that these

students are more exposed to English; learning English songs, and being surrounded by adults

frequently conversing in English.

How important is then for instance a five percentage point increase in exam scores in the

sample from Boyall HS? Considering the low average performance at the school, I claim that

even a small effect in the achievement should count as important for raising the level of

knowledge of the child – seen from a human development point of view. Hopefully

performing better could also motivate children to stay in school who would otherwise be

likely to drop out. Hypothetically, the difference could in some cases even be enough for

helping children to pass their grade. It is also possible to compare the increases seen in the

private hostel variable at both schools with the standard deviations (S.D.) of the average exam

scores (presented in table 1 and 2). Relative to this standard deviation the effect for the hostels

in the Boyall HS sample is between .39-.45, whereas for the CSI hostels in the Sherman HS

sample the effect is between .74-.9328. This can be compared to the class size effect found in a

study by Kreuger (1999) where attending a small class has an impact of around .20 on the

students’ scores. Even in the case of some selection bias to these hostels, the magnitude of the

effects can thus still be considered sizable.

28 For the Sherman HS sample the hostel coefficients in table 6 have been used, and divided by the S.D. in table

1. For the Boyall HS sample the hostel coefficients in table 4 have been used, and divided by the S.D. in table 2.

Page 38: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

38

Though the positive hostel effects suggest a reduction of gaps between different

socioeconomic groups’ achievement at the schools, it can still be seen in the results from the

regressions that students with an overall better socioeconomic background – coming from an

upper caste, with a small family and an employed father – perform better in general than the

hostel students. For a real reduction in gaps between different socioeconomic groups to take

place, it is important that the students not only learn more but that this leads to a continued

studying. The children need to complete at least twelve years of schooling in order to receive

an employment, according to the CSI correspondent in Sherman, otherwise they will most

likely be forced to do coolie work. As mentioned earlier, there is a great competition for

college seats, making it important to receive marks high enough to be able to proceed to

college.

9 Conclusions

The overall aim with this thesis was to test for the hypothesis that children who get the chance

to stay in a hostel, instead of growing up under poor conditions, do better in school. This was

based on the economic theories of intergenerational mobility which focuses on family

background to explain differences in children’s success in life (here measured by learning

achievement).

After controlling for as many characteristics as possible – related to school, child and family

background – positive effects on achievement were found for the Church of South India (CSI)

hostels at both schools included in the study. The effect was highest for CSI hostels in

Chittoor, where the students attend Sherman Memorial Girls’ High School (HS). For

example, the hostel students received 18 percentage points higher scores on average on the

Tamil and Telugu quarterly exam, compared to the group of day-scholar students. Also,

positive effects were seen in the number of years stayed in these hostels, reinforcing the

results of the average effect of the hostel stay. In the sample from Boyall Memorial High

School (HS) in Kollagunta, the impact of the CSI hostels was 5-6 percentage points for all

three subjects included in the study. Positive effects depending on the length of the stay in a

hostel could only vaguely be seen, possibly due to some extreme observations or other

unobserved factors influencing two of the cohorts of students, grouped according to the

number of years stayed in the hostel.

Page 39: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

39

The results showed no effect of staying in one of the government hostels in Chittoor. To some

extent, this was assumed to be due to a lower initial performance for this group compared to

the students in the CSI hostels. Furthermore, it could reasonably be explained –linking back to

human capital theories – by the lower investment in these children. The latter could also be

said concerning the differences in effects between the CSI hostels in Chittoor and Kollagunta

– even though not to the same extent. Arguments were laid forward that the CSI hostels in

Chittoor put more emphasis on following up on the children’s performance. This was done for

example by involving parents more, and having a closer cooperation between the school and

the hostels. Thus, apart from factors such as supplying nutritious food, having supervised

study time and educated staff, investments in terms of encouragement and motivation seems

to be of importance for making the children perform better.

Still, comparing to the better-off students at these schools, the hostel students’ achievement is

lower on average. This brings us back to the question of: Where would these students be

without the hostel? Very likely, many of them would have dropped out before completing

grade 10 considering the lack of schooling facilities in the rural areas in this district, the

poverty level of the families, and to some extent the lack of knowledge among illiterate

parents about the importance of education. Thus, even a small or insignificant effect of

staying in a hostel can be regarded as important if the alternative would be to receive no

education. It is also important to keep in mind that from a holistic development view, staying

in a hostel can have several other effects on a child’s development that are not seen in the

school achievement.

Policies such as hostels is just one way of trying to reduce the educational gaps. Whether this

is the most effective (or the best, seen from other aspects than educational) is left unanswered

in this study. Decisively, in order to reach the Millennium goal of having all children

complete primary schooling – and also making them profit from what they are taught in

school – more needs to be done.

Page 40: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

40

10 References

Becker, G. (1964): Human Capital, A Theoretical and Empirical Analysis, with Special

Reference to Education, National Bureau of Economic Research, New York.

Becker, G. and N. Tomes (1986): “Human capital and the rice and fall of families”, Journal of

Labor Economics, 4(3), pp.1-39.

Deininger, K. (2003): “Does cost of schooling affect enrollment by the poor? Universal

primary education in Uganda”, Economics of Education Review, 22(03), pp. 291-305

Drèze, J. and A. Sen (1996): India: Economic Development and Social Opportunity, Oxford

University Press, New Delhi.

Filmer, D. (2004): If you build it, will they come? School availability and school enrollment in

21 poor countries, Policy Research Working Paper No. 3340, World Bank

Ginther, D. K. and R. A. Pollak (2003): Does family structure affect children’s educational

outcomes?, National Bureau of Economic Research, Working Paper 9628, Cambridge.

Haveman, R. and B. Wolf (1995): “The determinants of children’s attainments: a review of

methods and findings”, Journal of economic literature, 33(4), pp.1829-1878.

Inchauste, G. (2000): Educational Choices and Educational Constraints: Evidence from

Bolivia, IMF Working Paper No. 42.

Indian Government, Ministry of Human Resource Development (1992): National Policy on

Education: 1986, Department of Education, New Delhi.

Jayachandran, U. (2002): Socio-Economic Determinants of School Attendance in India,

Centre for Development Economics, Delhi School of Economics, Working Paper No.103.

Kreuger, A. B. (1999): “Experimental estimates of education production functions”,

Quarterly Journal of Economics, CXIV, pp. 497-532.

Leibowitz, A. (1974): “Home Investments in Children”, Journal of Political Economy,

82(2 II), pp.111-131.

Melin, M. (2002): Education – a way out of poverty?, New Education Division Documents

No.12, DESO/SIDA/Education Division, Stockholm.

Mincer, J. (1974): Schooling, Experience and Earnings, National Bureau of Economic

Research, New York.

NCERT (2002-09-30): ”7th All India School Education Survey”, State: Andhra Pradesh,

http://7thsurvey.ncert.nic.in/, (2005-04-20).

Niranjanaradhya, V. P. (2004): Universalisation of school education, the road ahead…,

Centre for Child and the Law, National Law School of India University, Books for

Change, Bangalore.

Page 41: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

41

Schultz, T. (1961): “Investment in Human Capital”, American Economic Review,

51(1), pp. 1-17.

Sen, A. K. (1999): Development as Freedom, Oxford University Press, New Delhi.

Shah, G. (2002): Caste and Democratic Politics in India, Permanent Black, New Delhi.

Sridhar, B. V. and B. L. Sujatha (2002): Focus on Chittoor District, Discover District Series

(A.P), Om Art Print, Vijayawada.

Todaro, M. and S. Smith (2003): Economic Development, Eighth edition, Addison-Wesley,

London.

United Nations (2004): “Millennium goals”, http://www.un.org/millenniumgoals/,

(2004-09-27).

Weinberg, B. (2001): “An incentive model of the effect of parental income on children”,

Journal of Political Economy, 109(2), pp.266-280.

Wolfe, B. and J. Behrman (1984): “Who is Schooled in Developing Countries? The Roles of

Income, Parental Schooling, Sex, Residence and Family Size”, Economics of Education

Review, 3(3), pp.231-245.

World Bank (1995): Priorities and strategies for education, Development in practice, World

Bank review, Washington D.C.

World Bank (1997): Primary Education in India, Development in Practice, World Bank

Publication, Washington D.C.

World Bank (2002): “International Standard Classification of Education, 1976 (ISCED76)

and 1997 (ISCED97)”, http://devdata.worldbank.org/idg/IDGProfile.asp?CCODE=IND&

CNAME=India&SelectedCountry=IND, (2005-03-20).

World Bank (2004): India, Data and Statistics, “Millennium Development Goal for India”,

www.worldbank.org/ind, (2004-10-02).

List of people interviewed at several points in time between 8/1-3/2 2005:

Dr. D Samuel Issmer, Ph.D. Psychology, Director of SAC-CCC, Bangalore.

Mrs. Mangalam Bahavathsingh, Project coordinator, SAC-CCC, Bangalore.

Mrs. Parimala Ravi, Manager, CSI Girls Hostel, Kollagunta.

Mr. Y.E. Supadham, Head Master, Boyall Memorial High School, Kollagunta.

Ms. Saraswathy, CSI Correspondent, Sherman Memorial Girls Hostel, Chittoor.

Ms. P. Suguna Thilagam, Head Master, Sherman Memorial Girls High School, Chittoor.

Four female students staying in government hostels in Chittoor.

Page 42: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

42

Appendix A: Features of the studied area

I. Map of Andhra Pradesh and Chittoor district

II. Hostels included in the study

Sherman HS, Chittoor Obs Boyall HS, Kollagunta Obs

CSI Beattie Memorial Boys' Home 3 CSI Boys’ Boarding Home 29

CSI Sherman Memorial Girls’ Hostel 150 CSI Girls’ Home 77

SC Government Hostel (3 hostels)

ST Government Hostel

Social Welfare Government Hostel 42

Government Hostel for Physically Handicapped

III. Work variables for males

Estimated monthly

earning in Rupees

Type of work Included in

variable:

8 000 – 15 000 Coolie*, agricultural labourer, watchman, farmer, dobi, sweeper,

scavenger, rice-mill worker, helper, gardener, line-man, driver,

electrician, pastor**, teacher**, nature doctor, tailor, cook

Coolie

17 000-60 000 Business, shop-keeper, bank attender, petrol bank worker, barber,

typist, weaver, mason, conductor, bus/lorry driver, sub-inspector,

supervisor, teacher, warden, railway worker

Business/

employed

* In India, all labourers receiving daily wages goes under the name coolie. According to people interviewed in

Chittoor district, an agricultural coolie can normally work at the most 15 days per month normally receiving a

daily wage of 25-50 Rupees. A coolie worker can also be employed with constructing houses, roads, selling

vegetables etc. Common for all coolie occupations is a low wage and an unevenly distributed income.

**No academic title

Chittoor District

Area: 15 152 Sq.Kms

Population: 3,7 million

Page 43: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

43

Appendix B: Questionnaire

1) What is your name? …………………. V Boy V Girl

2) How old are you? ……

3) What standard are you in? V V V V V

6th 7

th 8

th 9

th 10

th

4) What is your religion?……………………………………………………..

5) What caste do you belong to?......................................................................

6) What is the name of your village / town? …………………………....

7) How many lives in your home? (including yourself) .….

8) Is your father alive? V Yes. V No.

9) Is your mother alive? V Yes. V No.

10) Is your father staying in your house? V Yes. V No.

11) Is your father supporting the family financially? V Yes. V No

12) Is your mother staying in your house? V Yes. V No.

13) What is your father’s occupation?……………………………………..

14) How many years did he go to school?

V No schooling V Some schooling. …. years in school V I don’t know

15) What is your mother’s occupation?……………………………………..

16) How many years did she go to school?

V No schooling V Some schooling. …. years in school V I don’t know

17) If no parents, please write your guardian’s (e.g. uncle/aunt)

occupation:………………………………..

18) Do you stay in a Hostel? V Yes V No

19) If Yes, name of the Hostel…………………………

20) Number of years stayed in the Hostel…………….

21) How many meals do you normally take per day?

V One meal V Two meals V Three meals

22) Do you have any adult/tuition/warden to assist you with your homework?

V No. V Yes, sometimes VYes, every day.

23) Do you work before and after school at the hostel/home? V Yes V No

24) If Yes, how many hours do you work? …………

Page 44: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

44

Appendix C: The distribution of average test scores

Sherman HS Hostel students Sherman HS Day-scholar students

Boyall HS Hostel students Boyall HS Day-scholar students

0.01

.02

.03

Density

0 20 40 60 80Average score for all three subjects

0.005

.01

.015

.02

.025

Density

0 20 40 60 80 100Average score for all three subjects

0.01

.02

.03

.04

.05

Density

0 20 40 60 80Average score for all three subjects

0.01

.02

.03

.04

Density

0 20 40 60 80Average score for all three subjects

Page 45: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

45

Appendix D: OLS with distance dummies

Sherman HS, Chittoor

Tamil/

Telugu

English Maths

Age -1.099

(2.147)

.081

(1.460)

2.065

(1.986)

Boy 1.945

(5.964)

4.605

(4.226)

8.709*

(4.613)

Different tongue -3.875

(3.041)

- -

Grade 7 -1.644

(4.874)

-2.328

(3.053)

12.073***

(4.090)

Grade 8 -8.548***

(3.1345)

-2.995

(2.094)

4.966*

(2.689)

Private hostel 18.279***

(3.111)

10.562***

(2.013)

11.045***

(2.507)

Government hostel 2.084

(3.469)

1.082

(2.072)

-.371

(2.982)

Telugu section 4.541

(2.955)

-1.171

(1.587)

4.889**

(1.987)

BC 9.790***

(2.722)

3.743**

(1.732)

5.226**

(2.233)

OC 13.406***

(3.605)

7.2457***

(2.266)

12.423***

(3.202)

Christian -2.064

(1.925)

-1.192

(1.191)

-1.732

(2.007)

Family -1.227*

(.681)

-.654*

(.338)

-.319

(.490)

Joined family 11.731

(8.111)

2.642

(4.778)

3.027

(6.233)

Father’s education .468

(.305)

.397**

(.183)

.503**

(.235)

Mother’s education .025

(.285)

.0418

(.190)

-.302

(.234)

Father business 12.448**

(6.286)

5.027

(4.243)

5.284

(5.040)

Father dead/not supporting 2.261

(2.886)

1.918

(1.986)

.648

(2.569)

Mother not working 1.255

(2.193)

.780

(1.488)

.262

(1.887)

Mother dead -6.829

(7.812)

-7.314

(7.514)

-11.833

(10.742)

Urban area 5.744*

(2.949)

4.398**

(2.061)

5.080**

(2.492)

Semi-urban area -1.318

(2.679)

-.740

(1.786)

-3.192

(2.165)

Constant 50.301

(30.962)

36.947*

(20.718)

-4.761

(28.056)

Observations 356 356 356

R-squared 0.22 0.18 0.20

F statistic 5.92 4.75 4.24

Prob>F 0.0000 0.0000 0.0000

Robust standard errors in parentheses. *significant at 10%, **significant at 5%, ***significant at 1%. Reference

categories: father’s occupation, coolie; mother’s occupation, coolie; caste, SC/ST; grade, 9.

Page 46: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

46

Appendix E: Effect of years in hostel

Figure I: Boyall HS, Telugu

0

2

4

6

8

10

12

14

1 2 3 4 5

years in hostel

effect (%-points)

average annual

effect

hostel effect for

each cohort of

Figure II: Boyall HS, English

0

1

2

3

4

5

6

7

8

9

10

1 2 3 4 5

years in hostel

effect (%-points)

Figure III: Boyall HS, Maths

0

2

4

6

8

10

12

1 2 3 4 5

years in hostel

effect (%-points)

Page 47: Educational policies serving the poor - a case study …130289/fulltext01.pdfDEPARTMENT OF ECONOMICS Uppsala University Master Thesis Spring term 2005 Educational policies serving

47

Figure IV: Sherman HS, Tamil/Telugu

0

5

10

15

20

25

30

1 2 3 4 5

years in hostel

effect (%-points)

average annual

effect

hostel effect for

each cohort of

students

Figure V: Sherman HS, English

0

2

4

6

8

10

12

14

1 2 3 4 5

years in hostel

effect (%-points)

Figure VI: Sherman HS, Maths

0

2

4

6

8

10

12

14

16

18

20

1 2 3 4 5

years in hostel

effect (%-points)