Belief Systems and Wealth Outcomes

Feler Bose Abstract: One large area of research in which much work remains to be done is the role of belief systems and how they affect economic performance and wealth accumulation. The problem with any research along these lines is that it is difficult to isolate belief systems as a causal variable in any social dynamic. When looking at populations that are otherwise demographically equal, however, it becomes possible to connect specific outcomes to specific beliefs. This paper will look at how belief systems affect wealth distribution by testing whether Robert Caldwell’s observations in South India are still valid after 150 years. Key words: Principal component analysis, Socio-economic status, fieldwork, India, beliefs, economic outcomes JEL Code: C43, C83, O10, Z12 Journal of Development Economics (see Volume 98, Issue 1 and quote from there) Review of Development Economics Journal of Economic Development I would like to thank Laurence Iannaccone for his advice on this project. I would also like to thank the participants of the Global Prosperity Initiative. Additional thanks to K. Bose, GRD Rajasingh and others in the schools and villages who took time out to help me. Financial assistance from the Mercatus Center, from the Acton Institute, and from an Alma College grant is gratefully acknowledged. Additional funding from a Charles Koch grant for student research is acknowledged. Thanks to student Grant Isley for excellent research assistance. Thanks to support from the Michigan State University visiting scholar program.

2

1. Introduction

To understand the role of beliefs and how they affect economic performance is a difficult

and controversial task. The literature in economics regarding beliefs and economic outcomes has

been sparse but growing. For example, one study tried to determine why Europe and United States

are so different in the area of government’s role in production and redistribution of income

(Rotemberg 2002) and concluded that beliefs tied to the fairness of the distribution of income have

played a role. Rafael Di Tella et. al. (2007) wrote a paper on how property rights in Argentina have

affected people’s beliefs when controlling for other factors. For example, squatters who got legal

title to the land favoured the free market more than squatters who did not get title to the land.

Douglas North (2005) has written about how belief systems shape institutional design. Richardson

(2005) writes about how beliefs in purgatory resulted in development of religious cartels (guilds) and

when the belief in purgatory declined, different economic systems arose. Work by others has shown

that religion/culture does affect economic outcomes (Guiso, et al. 2006). Guiso et al., focus on prior

beliefs and values or preferences to show the causal effect from culture to economic outcomes.

Kuran (2004) writes about how certain Islamic beliefs, especially those tied to inheritance laws,

resulted in stagnant institutions that hindered economic performance in the Middle East. Arrunada

(2009) writes about Catholics, who believe in salvation by works, and the church as the intermediary

and enforcement agent, and the Protestants, who believe in salvation by grace and enforcement

through social interactions, has resulted in Protestants favouring anonymous trade and markets and

Catholics favouring personalized trade. The research I have done hopes to add to the literature on

belief systems and wealth outcomes.

<<<<<<<<<<<<Insert Figure 1-1 Here>>>>>>>>>>

To answer my research question as to whether beliefs affect wealth outcomes, I had

proposed visiting a number of villages in Southern India. Over 150 years ago, Rev. R. Caldwell, a

missionary and linguist, visited the region that I chose to work in. He noted that there were

3

differences when comparing Hindu and Christian villages. Caldwell writes (1857:116) Christian

villages were neater, had better built houses, more trees, “greater prosperity and comfort” and

improvements in certain psychological traits. In this paper, I want to see if there is still a wealth

disparity between villages. On the other hand, do we see convergence in the villages? Since

knowledge of success leaks (Easterly 2001:146ff) have the Hindu households and villages taken

advantage of the new knowledge and adapted it to increase their success. Some historians think

there has been convergence (Grafe 1990).

The villages in India that I visited had similar demographic features. For example, all the

villages are located in the same district in India (Tuticorin/Thoothukudi) (see Figure 1-1), the

inhabitants all speak the same language (Tamil) and belong to the same caste (Nadars), the villages

are geographically located in similar terrains (e.g. receive similar amounts of rainfall, use mainly well

based irrigation etc.,) and the villages are similar in size. Nevertheless, the villages chosen have very

different religious beliefs: Arokiapuram is a Christian village, Melaramasamiapuram is a Hindu

village, and Arulanandapuram/Vallivillai is a mixed village of approximately 60% Hindu and 40%

Christian1. Since I control for the extraneous variables, it becomes possible to connect specific

outcomes to specific beliefs in these villages.

This article proceeds as follows. Section 2 shows the research methodology used, section 3

is the history of Nadars, section 4 develops the socio-economic score using Principal Components

Analysis, section 5 tests Caldwell’s observations and section 6 contains preliminary conclusions and

suggestions for follow-up work.

1 Christians in these villages are all Protestant. All Protestant Christians in these villages belonged to the

Church of South India.

4

2. Methodology

I went on two field trips to India to study these villages. The first trip was in 2005, which

used one type of methodology. The second trip was in 2011, which used another, but

complementary methodology.

2.1 Data Collection--Field Trip 2005

My trip to India during the summer of 2005 was designed to be exploratory and descriptive

in nature with some hypothesis testing. I conducted a systematic study using field study techniques.

A field study is useful for discovering the relations and interactions of variables that are ex-post facto

in nature. No manipulations of variables are done in a field study. I used three techniques in my

field study in India. First, I used exploratory fieldwork to gain familiarity with the villages,

understand the context of where the villages were located and to narrow the hypothesis.

Second, I did descriptive fieldwork using photography and participant observation.

Participant observation consists of “simultaneously participating in as many of the activities as

possible at a particular site or in a particular setting, observing what is transpiring and interpreting

what the researcher has participated in and observed” (Druckman 2005:235). In most of my

participant observation, I focused on observing how the school system was organized and worked.

Third, I used semi-structured qualitative interviews using opportunistic sampling (willing

informants) and judgmental sampling (those with direct bearing on topic) for initial hypothesis

testing. Table 2-1 shows the number of people interviewed in each of the villages2. The reason for

using interviews is that interviews can be used to uncover and explore “meanings that underpin

people’s lives, routines, behaviors, feelings etc.” (Arksey and Knight 1999:32).

<<<<<<<<<Insert Table 2-1 Here>>>>>>>>>>

2 The total number of interviews conducted in this field study was much larger. I first conducted numerous

interviews to find these three villages. Further, I completed other interviews to lay the groundwork for another trip in the future.

5

2.2 Data Collection--Field Trip 2011

During the summer of 2011, a second field trip was completed. In this trip, I conducted the

survey method (an ex-post facto method) to allow for direct hypothesis testing. The survey

questions covered areas of health, time preferences, wealth, household particulars, education, and

trust questions. I used some questions to crosscheck other questions.

I used ten different surveys to develop the final survey used in India (see Appendix A for the

list of surveys used). The survey had 84 questions and had an additional module that contained

questions on agriculture if the household was involved in agriculture and livestock issues. Everyone

who was available in the village participated in the survey (exhaustive survey)3.

Using all these tools in the two different field trips, I hope to triangulate information. The

idea of triangulation is using different techniques to find results for a set of research questions.

Triangulation serves two purposes: confirmation and completeness (Arksey and Knight 1999:21).

Further triangulation allows for blending and integrating various methods of research to strengthen

the conclusion.

The main problem in the survey comes from measurement errors and translation errors.

The survey was first compiled in English and then translated into Tamil. In the process of translation,

a handful of questions came out with different meanings. This problem could have been minimized

with retranslation into English (Tripathi 2005:77). Measurement errors can occur from the

interviewer effect.

3. Historical Context

Since my fieldwork was in Nadar villages, I will include in this section a historical overview of

the Nadars to help the reader understand how many Nadars became Christians in an

overwhelmingly Hindu India.

3 The mixed village consisted of 58 households, the Christian village consisted of 36 households, and the Hindu

village consisted of 64 households for a total of 158 households.

6

The Nadars were part of the Dravidian race of the Indus Valley Civilization in present day

Pakistan, Afghanistan, and Northern India (3000 to 1500 BC). The Nadars lived in that region until

the “Aryan Invasion” around 1400 to 1000 BC when they were driven out to other parts of modern

India (Immanuel 2002:xxxv).

The early history of the Dravidians is known through palm leaf manuscripts and records in

Hindu temples. While the Nadars in modern Indian society indicate a caste position, the original

Dravidians were a casteless society. History shows that many Nadars were once Jains and Buddhists

(casteless religions). The introduction of Caste in Southern India probably occurred between the 8th

and 11th century AD with the entry to the south by a large group of Brahmins. During this time, the

influence of Buddhism was destroyed (Immanuel 2002:270).

Due to infighting among Nadar groups, this group was weakened, with the result that Nadars

were not allowed to worship in their own temples starting in 1664. The Brahmins also teamed up

with other castes to suppress the Nadars. There were more than 100 taxes imposed on the Nadars

in the 19th century. They included taxes for having legs, for walking, for palm leaf texts, for wedding

chains, for using the plough, for being born, for dying, for treating the sick, etc. (Immanuel

2002:App.1). The tortures of the Nadars were so inhumane that in the 17th through 19th

centuries the Christian missionaries found a community that was receptive to their message.

The designation of Nadars as either high caste or low caste is not without controversy.

Some have said the Nadars would be in the Kshatriya (warrior/king) caste (high caste), and

historically the kings among the Dravidians were also priests (Bergunder 2008:16; Immanuel 2002).

This Kshatriya designation is not without controversy (Grafe 1990:104, 212; Kent 2004:68ff;

Mallampalli 2004:247). Missionaries on the field have designated the Nadars at the high end of the

low caste (Caldwell 1857:45), but they also note the claim of Nadars being of high caste (Caldwell

1857:33)4.

4 It seems that Nadars were called Shanars (a derogatory term) during the time period of the missionaries, but

later took the name Nadars (Lords) when their status improved (Frykenberg 2008: 208).

7

Their mass conversion to Christianity resulted in some villages becoming Christian. This was

seen as a threat to the local landlords, who perceived a loss of control over the Nadars, who with

other castes attacked these new converts. Thousands fled their villages and eventually formed

many new villages as places of refuge (Frykenberg 2008:209f; Grafe 1990:27).

Regardless of where the Nadars stand in the caste hierarchy, their mass conversion to

Christianity and their involvement in trade brought their (including Hindu Nadars via spill over

effects) (Grafe 1990:212) economic transformation (Polgreen 2010). The Christian villages

established various institutions including dispensaries, hospitals, basic educational institutions5,

colleges, seminaries, welfare institutions, etc. These institutions transformed the “political economy

of the Tirunelveli area, thereby also beginning to bring about profound transformations in local

culture and society” (Frykenberg 2008:212). Historian Grafe notes that Christian Nadars emerged

“early as one of the groups who rose from social lowliness to occupy places formerly reserved for

Brahmins’’ (1990:84).

4. Construction of a Socio-Economic Status Index using Principal Component Analysis

To test the observations of Caldwell over 150 years ago and recently by Grafe and

Frykenberg, I developed a socio-economic status index using Principal Component Analysis (PCA).

4.1 Compilation of the wealth index using Principal Component Analysis

I used the PCA method to create a wealth index. Filmer and Pritchett (2001:128) note that

the PCA “provides plausible and defensible weights for an index of assets to serve as a proxy for

wealth” and measures the long-run economic status of a household (2001:116). By using the PCA

method, the researcher avoids having to provide arbitrary weights to each component in the wealth

index.

5 Caldwell (1857: 95) writes that in many of the larger Christian villages about 25% of the population was in

school and overall among all the Christian villages the number was 16%. This was a high proportion at that time.

8

Principal component analysis reduces the correlated variables with uncorrelated ‘principal

components.’ The first principal component explains the most variance of the total variance in the

data. Hence, the first principal component is used by many to create an index (McKenzie 2005:251).

This can be problematic as the first principal component might only explain a small percentage of

the total variation. For my work, the first principal component measures 11% of the variation of the

variables.

The scree plot for the PCA is shown in figure 4-1. From the scree plot it can be seen that the

first four or five components would be sufficient to explain most of the variations in the original

variables. However, only the first component is needed to create the wealth index (McKenzie

2005:251).

<<<<<<<<<<<<<Insert Figure 4-1 Here>>>>>>>>>>>>

The PCA wealth index was created by extracting data from the survey related to ownership

of durable goods, access to utilities and infrastructure, services, and housing

characteristics/structure. I had to recode many of the survey questions as dichotomous variables.

Therefore, all variables were coded zero or one. One variable was continuous and that was the

question “Excluding bathroom and kitchen how many rooms do you have in your house?” The total

number of variables is forty-two, of which five variables had no variation and were dropped when

doing the analysis. Since the data is not standardized, PCA was completed using the correlation

matrix instead of the co-variance matrix (Vyas and Kumaranayake 2006:463).

4.2 Interpretation of PCA Results

The PCA was constructed for all three villages together. An asset found in all households or

no households is ignored. Component scores can be either positive or negative. A household using

wood/coal to cook, or having an earth/dung floor, or grass thatch roof has the largest negative effect

9

on the overall asset index, whereas having a refrigerator, more rooms in the house, and having a

toilet provides for the largest positive effect to the overall asset index.

Table 4-1, provides the mean, standard deviation, and component scores for all the

variables. There is a wide range of ownership of goods. Larger weights (component scores) are

assigned to assets that vary the most and zero for those assets in all households or in none of the

households. Nearly all households have a color television whereas only a few households have a car.

All component scores for durable goods is positive, indicating possibly that creating one index which

includes all three villages is reasonable (Vyas and Kumaranayake 2006:464).

<<<<<<<<<Insert Table 4-1 Here>>>>>>>>

4.3 Socio-Economic Scores for Households & Internal Consistency Check

The weights (component score) from the first principal component is used to create a score

for each household. This score is the socio-economic score (SES). The distribution of the SES scores

of households shows a normal graph (see Figure 4-2). This score could be used in a regression, but

the interpretation of results is not easy. Therefore, one way to overcome this limitation is to divide

the households into broad categories. I have divided households into quintiles.

<<<<<<<<<<Insert Figure 4-2 Here>>>>>>>>>>

To test for the internal validity of the wealth index, the mean asset ownership

needs to be analyzed for each quintile. From table 4-2, we see that, when comparing

poorest and richest households (and also the other quintiles), the data is coherent.

For example, under cooking fuel, wood/coal is used by 100% of the poorest

households, whereas about half of the richest households use it. Only 9% of the poor

10

households have toilets, whereas 88% of the richest households have toilets. The

trends across all quintiles are generally clear except in a few cases, such as having a

borehole in your house.

<<<<<<<<<<Insert Table 4-2 Here>>>>>>>>>>

Some of the trends from table 4-2 are plotted in figures 4-3 and 4-4. Figure 4-3

plots the trend for the four largest negative component scores and figure 4-4 the four

largest positive component scores (excluding average number of rooms).

<<<<<<<<<<<<Insert Figure 4-3 Here>>>>>>>>>>>

<<<<<<<<<<<<Insert Figure 4-4 Here>>>>>>>>>>>

5. Testing the Hypothesis: Convergence or Divergence

In this section, I will test to see if Caldwell’s observations of divergence are still true or

whether there is convergence as noticed by more recent writers (Grafe 1990). Figure 5-1 shows the

different percentages of each faith6 in each quintile. From figure 5-1, we learn that 11% of the

Christian households and 27% of the Hindu households fall in the poorest quintile, whereas 29% of

Christian households and 14% of Hindu households fall in the richest quintile. This indicates that

there is still divergence between Christians and Hindus overall.

<<<<<<<<<<<<<Insert Figure 5-1 Here>>>>>>>>>>>>>

Another way to look at the data on divergence is to see wealth disparities on a village basis.

From figure 5-2, it is clear that in the Christian village, only 3% of households fall in the poorest

category and 39% of households fall in the richest quintile. The Christian village has a positive trend

6 Percentage of Hindus = (# of Hindu households in quintile/Total number of Hindu households in all 3 villages)

multiplied by 100

11

as one moves across quintiles. For the Hindu village, each quintile contains about 20% of

households, therefore we see a flat trend across quintiles. However, in the mixed village, we see a

large number of households in the poorest category (34%) and only 12% in the richest category, with

the trend being negative across quintiles. This data also indicates that there is no convergence.

<<<<<<<<<<<Insert Figure 5-2 Here>>>>>>>>>>>

I further disaggregated the villages and looked at the mixed village more closely. If the

convergence argument is valid, we should see a similar profile for both Christian and Hindu

households in the mixed village. In figure 5-3, again we instead see a large divergence between

Hindus and Christians. Fifty- two percent of the Hindu households fall in the poorest quintile

whereas twenty-one percent of Christian households fall in the poorest quintile. There is also a large

gap in the middle quintile; however, in the richest quintile the percentage of each faith is similar.

This strengthens one observation that there still exists divergence between Christian and Hindu

households and very little spill over effects.

<<<<<<<<<<<Insert Figure 5-3 Here>>>>>>>>>>>>>

To triangulate information, the observations from the first and second field trips indicate

that the Christian village did in fact look more prosperous in terms of houses built, and their school

looked better (see table 5-1 for information on the local school). However, it was difficult to

determine whether the mixed village or Hindu village was poorer.

On the issue of spill over effects, Christians managed the school in the Hindu village. When

visiting another Hindu village, I found that their school was managed by Hindus albeit staffed by

12

Christian teachers7. The importance of providing education for their children had spilled over to the

Hindu village. Therefore, while there are spill over effects, the wealth disparities remain.

<<<<<<<<Insert Table 5-1 Here>>>>>>>>>>>>

6. Preliminary Conclusion and Future work

PCA was used to create a wealth index to test Caldwell’s observations and to test for

convergence. The support for the observations of Caldwell on divergence is still true today.

Christian villages are better off and Christian households are better off in the mixed village. There is

no convergence as of yet. However, have the rural Hindu Nadars benefitted from the growth of the

rural Christian Nadars? Has there been spill over effects? The answer to these questions is yes, as

documented by many sources. The Nadars as a whole have moved ahead of their peers (Polgreen

2010).

While in the villages chosen, we have a case of ceteris paribus, there could be an over-

looked confounding factor that has not been taken into account. Further, instead of using a wealth

index alone to test for convergence, a consumption based index would also add clarity to the

conclusion. However, developing a consumption index is difficult and prone to error.

In this research, I have shown that beliefs affect economic (wealth) outcomes in these

villages, and there is no convergence. However, for the conclusion to be broadened, further

research is needed. The next step would be to increase the number of villages albeit within the

Nadar community. Further, I would need to control for the idiosyncratic effects of the village.

Community level questionnaires with household surveys need to be completed if one expands the

number of villages. Further, the sample size needs to be increased and instead of completing an

7 Christians in India have generally been active in education for a long time. Therefore Christian schools and

Christian staff seem to be a brand name for high quality. Further, the hiring of Christians in a Hindu school could also be due to more Christians in the education field and hence available to teach when there are job openings.

13

exhaustive survey, cluster sampling of each village can be done. Retrospective surveys of

households can be done to get more data points on wealth.

14

Appendix A: Surveys used for creating the Survey in India.

1. Okello, Julius J. “Compliance with International Food Safety Standards: The Case of Green Bean Production in Kenyan Family Farms.” Michigan State University Dissertation.

2. Code Book: Michigan State of the State 54: 991. Weight Variable: statewt http://www.ippsr.msu.edu/Documents/SOSSArchive/Codebooks%20PDF/SOSS54wt_CBK.pdf accessed February 10, 2010.

3. Records from Polling Nation database. Questions extracted from database. Accessed:

March/8/2011.

4. World Values Survey 2005-2006 Wave, OECD- Split-Version (Ballot A)

5. European Social Survey, 2010. EES Round 5 Source Questionnaire. London: Centre for Comparative Social Surveys, City University, London.

6. Living Standards Measurement Survey. Statistical Office of the Republic of Serbia. Belgrade,

April 2007. Source: The World Bank Group. http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/EXTLSMS/0,,contentMDK:21718493~pagePK:64168445~piPK:64168309~theSitePK:3358997,00.html

7. General Social Survey questionnaire. NORC at University of Chicago. Ballet 1 - Area -

English. Survey. 2010. HYPERLINK <http://www.norc.uchicago.edu/GSS/Templates/BaseTemplate.aspx?NRMODE=Published&NRNODEGUID={21403184-C064-4E20-944F-0CFCABC9BB5E}&NRORIGINALURL=%2fGSS%2bWebsite%2fPublications%2fGSS%2bQuestionnaires%2f&NRCACHEHINT=NoModifyGuest.>

8. World Values Survey, Non-OECD Split Version, 2005-2006

9. Gallup Survey, The Values and Beliefs of the American Public- A National Study. The Gallup Organization 2005, 2007. Princeton, NJ.

10. Women of Chichewa Survey, 2009

15

7. References

Arksey, Hilary, and Peter T. Knight

1999 Interviewing for Social Scientists: An Introductory Resource with Examples. London: Sage Publications.

Arrunada, Benito

2009 Protestants and Catholics: Similar Work Ethic, Different Social Ethic. Economic Journal.

Bergunder, Michael

2008 The South Indian Pentecostal Movement in the Twentieth Century: Studies in the History of Christian Missions. Grand Rapids, MI: Wm. B. Eerdmans Publishing Co.

Caldwell, Rev. R.

1857 Lectures on the Tinnevelly Missions, Descriptive of The Field, the Work, and the Results London: Bell & Daldy.

Druckman, Danies

2005 Doing Research: Methods of Inquiry for Conflict Analysis. Thousand Oaks, CA: Sage Publications.

Easterly, William

2001 The Elusive Quest for Growth: Economists' Adventures and Misadventures in the Tropics. Cambridge, MA: MIT Press.

Filmer, Deon, and Lant H. Pritchett

2001 Estimating Wealth Effects without Expenditure Data-or Tears: An Application to Educational Enrollments in States of India. Demography 38(1):115-132.

Frykenberg, Robert Eric

2008 Christianity in India: From Beginnings to the Present. Oxford: Oxford University Press.

Grafe, Hugald

1990 The history of Christianity in Tamilnadu from 1800 to 1975. 6 vols. Volume 4 (part 2). Bangalore: Church History Association of India.

Guiso, Luigi, Paola Sapienza, and Luigi Zingales

2006 Does Culture Affect Economic Outcomes. Journal of Economic Perspectives 20(2):23-48.

Immanuel, M

16

2002 The Dravidian Lineages A Socio-Historical Study: The Nadars Through the Ages. Nagercoil, India: Historical Research and Publications Trust.

Kent, Eliza F.

2004 Converting Women: Gender and Protestant Christianity in Colonial South India. Oxford: Oxford University Press.

Kuran, Timur

2004 Why the Middle East is Economically Underdeveloped. Journal of Economic Perspectives 18(3):71-90.

Mallampalli, Chandra

2004 Christians and Public Life in Colonial South India, 1863-1937: Contending with Marginality. London: RoutledgeCurzon.

McKenzie, David J.

2005 Measuring inequality with asset indicators. Journal of Population Economics 18(2):229-260.

North, Douglas C.

2005 Understanding The Process of Economic Change. Princeton, NJ: Princeton University Press.

Polgreen, Lydia

2010 Business Class Rises in Ashes of Caste System. In The New York Times. Manhattan: The New York Times Company.

Richardson, Gary

2005 Craft Guilds and Christianity in Late-Medieval England: A Rational Choice Analysis. Rationality and Society 17(2):139-189.

Rotemberg, Julio

2002 Perceptions of Equity and the Distribution of Income. Journal of Labor Economics 20(2):249-288.

Tella, Rafael Di, Sebastian Galiani, and Ernesto Schargrodsky

2007 The Formation of Beliefs: Evidence from the Allocation of Land Titles to Squatters. The Quarterly Journal of Economics 122(1):209-241.

Tripathi, P.C.

2005 A Textbook of Research Methodology in Social Sciences. New Delhi: Sultan Chand & Sons.

Vyas, Seema, and Lilani Kumaranayake

17

2006 Constructing socio-economic status indices: how to use principal components analysis. Health Policy and Planning 21(6):459-468.

18

Figure 1-1: Map of Tamil Nadu, India (http://en.wikipedia.org/wiki/Tamil_Nadu accessed August 2005). The District of Thoothukudi (number 22) is where these villages are located.

19

Table 2-1: The total number of people interviewed8 in the three villages is shown below. M/F stands for Male/Female.

Village Population in Village9 M/F

Total interviewed

Arulanandapuram/Vallivillai 180/194 14

Melaramasamiapuram 137/158 12

Arokiapuram 90/103 17

8 The small number of people interviewed was due to the fact that I focused mostly on educational issues

during the first trip. 9 These numbers have shrunk when compared to the second trip in 2011 as people are migrating to cities.

20

01

23

4

Eig

enva

lues

0 10 20 30 40Number

Scree plot of eigenvalues for all villages

Figure 4-1: The Scree plot of eigenvalues with the turning point occurring between fourth and fifth component.

21

Table 4-1: Results from Principal Component Analysis

Variable Mean SD Component

Score N

Cooking Fuel

Wood/coal 0.886 0.319 -0.303 158

Dung 0.000 0.000 158

Kerosene 0.241 0.429 0.0993 158

Gas 0.544 0.500 0.213 158

Electricity 0.013 0.112 -0.0041 158

Roof Material

Grass thatch 0.158 0.366 -0.2678 158

Asbestos 0.038 0.192 -0.0241 158

Cement/Concrete 0.449 0.499 0.226 158

Iron Sheets 0.044 0.206 -0.0095 158

Shingles/Tiles 0.405 0.492 -0.0213 158

Flooring in House

Earth/dung 0.127 0.334 -0.2762 158

Bricks 0.000 0.000 158

Tiles 0.063 0.244 0.1828 158

Cement 0.842 0.366 0.1178 158

Wood/bamboo 0.000 0.000 158

Water Source

Borehole 0.089 0.285 0.106 158

Well 0.025 0.158 0.0524 158

River/Spring 0.000 0.000 158

Community Kiosk 0.285 0.453 -0.0225 158

Stand in pipe/tap 0.658 0.476 -0.0368 158

Durable Goods

Bed with mattress 0.222 0.417 0.2412 158

Television 0.905 0.294 0.1037 158

Radio 0.253 0.436 0.074 158

Cell phone 0.886 0.319 0.1417 158

Land Line Phone 0.070 0.255 0.2392 158

Refrigerator 0.057 0.233 0.2997 158

Bicycle 0.658 0.476 0.0627 158

Motorcycle 0.285 0.453 0.1968 158

Animal drawn cart 0.006 0.080 0.0572 158

Car/truck 0.006 0.080 0.1547 158

Washing Machine 0.025 0.158 0.1312 158

Compound Wall

Mud 0.000 0.000 158

Wire 0.038 0.192 -0.057 158

Brick 0.532 0.501 0.2087 158

Thatch 0.373 0.485 -0.252 158

Limestone 0.006 0.080 -0.0068 158

Stone 0.025 0.158 0.0681 158

Sticks 0.013 0.112 -0.001 158

Thorns 0.013 0.112 -0.0938 158

Miscellaneous

22

Own home 0.829 0.378 0.005 158

Have Toilet 0.462 0.500 0.2534 158

Rooms in House 2.399 1.226 0.2701 158

23

0.1

.2.3

Den

sity

-5 0 5 10Socio-Economic Score

Histogram of Socio-Economic Scores

Figure 4-2: Distribution of socio-economic scores in the three villages.

24

Table 4-2: Checking the validity of the Wealth Index by checking characteristics of households in each quintile.

Variable Poorest Second Middle Fourth Richest

Cooking Fuel

Wood/coal 100% 100% 94% 94% 56%

Dung 0% 0% 0% 0% 0%

Kerosene 9% 19% 13% 39% 41%

Gas 28% 29% 50% 74% 91%

Electricity 0% 3% 0% 3% 0%

Roof Material

Grass thatch 69% 6% 0% 3% 0%

Asbestos 6% 0% 9% 3% 0%

Cement/Concrete 9% 42% 44% 52% 78%

Iron Sheets 3% 13% 0% 0% 6%

Shingles/Tiles 25% 52% 53% 48% 25%

Flooring in House

Earth/dung 56% 6% 0% 0% 0%

Bricks 0% 0% 0% 0% 0%

Tiles 0% 0% 3% 6% 22%

Cement 53% 97% 97% 94% 81%

Wood/bamboo 0% 0% 0% 0% 0%

Water Source

Borehole 9% 0% 3% 3% 28%

Well 0% 0% 0% 6% 6%

River/Spring 0% 0% 0% 0% 0%

Community Kiosk 31% 29% 19% 29% 34%

Stand in pipe/tap 66% 71% 81% 61% 50%

Durable Goods

Bed with mattress 0% 10% 13% 26% 63%

Television 81% 84% 100% 94% 94%

Radio 16% 29% 25% 16% 41%

Cell phone 81% 68% 97% 100% 97%

Land Line Phone 0% 0% 0% 6% 28%

Refrigerator 0% 0% 0% 0% 28%

Bicycle 53% 58% 75% 74% 69%

Motorcycle 13% 10% 19% 42% 59%

Animal drawn cart 0% 0% 0% 0% 3%

Car/truck 0% 0% 0% 0% 3%

Washing Machine 3% 0% 0% 0% 9%

Compound Wall

Mud 0% 0% 0% 0% 0%

Wire 9% 3% 3% 3% 0%

Brick 22% 39% 50% 71% 84%

Thatch 81% 55% 34% 13% 3%

Limestone 0% 0% 3% 0% 0%

Stone 0% 3% 0% 3% 6%

Sticks 3% 0% 0% 0% 3%

Thorns 6% 0% 0% 0% 0%

Miscellaneous

25

Own home 84% 77% 84% 81% 88%

Have Toilet 9% 26% 53% 55% 88%

Avg. Rooms in House 1.469 2.161 2.500 2.355 3.500

26

Figure 4-3: Inequality based on the largest negative component scores.

27

Figure 4-4: Inequality based on the largest positive component scores (excluding average number of rooms).

28

Figure 5-1: The location of Christian and Hindu households in the different quintiles.

29

Figure 5-2: Distribution of households by village.

30

Figure 5-3: Distribution of households in the mixed village.

31

Table 5-1: Basic information regarding the infrastructure found in the schools.

Village Electricity Private Bathroom

School Furnishing

Arulanandapuram/Vallivillai (Mixed)

Yes Yes Good

Melaramasamiapuram (Hindu) No No Minimal

Arokiapuram (Christian)

Yes1 Yes Best

1 The school borrows electricity from the nearby church.