Money and Marriage: A Fresh Look at Marriage Transactions in Rural India
BY
AFRA RAHMAN CHOWDHURY
B.S.S., DHAKA UNIVERSITY, BANGLADESH, 1997
M.S.S, DHAKA UNIVERSITY, BANGLADESH, 1999
M.A., BROWN UNIVERSITY, 2004
A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
IN THE DEPARTMENT OF SOCIOLOGY AT BROWN UNIVERSITY
PROVIDENCE, RHODE ISLAND
MAY 2008
© Copyright 2008 by AFRA RAHMAN CHOWDHURY
This Dissertation by AFRA RAHMAN CHOWDHURY is accepted in its present form by the Department of Sociology as satisfying the dissertation requirement for the degree of Doctor of Philosophy.
Date ______Dennis P. Hogan, Director
Recommended to the Graduate Council
Date ______David P. Lindstrom, Reader
Date ______Nancy Luke, Reader
Approved by the Graduate Council
Date ______Sheila Bonde Dean of the Graduate School
iii
VITA
Afra Rahman Chowdhury was born on December 22, 1975 in Sylhet, Bangladesh.
She completed her Bachelor of Social Science degree in 1997 from Dhaka University majoring in Economics, where she also earned a Masters of Social Science degree in
Economics in 1999. She then worked at Bangladesh Institute of Development Studies
(BIDS) as a research assistant in projects pertaining to important social issues. While working as a research assistant at BIDS, her interest in Sociology and Social
Demography sparked. She obtained valuable first hand experience in empirical research including fieldwork, data management and analyses while she worked at BIDS. She began her graduate studies in Sociology at Brown University in September 2002 and earned a Masters of Arts degree in May 2004. Her Masters thesis analyzed the effect of parent-daughter relationship quality on American teenage daughters’ sexual behavior contributing to research on important issues like family relations and teenage pregnancy.
While at Brown, she received fellowship and training in social demography through the
Population Studies and Training Center, which significantly shaped her research interest and current course of research. She has also received extensive methodological and technical training working as a research assistant to professor David P. Lindstrom on a project that studies internal migration in Guatemala. During the course of her graduate life, she received fellowships funded by Hewlett Foundation and Population Reference
Bureau. She has also worked as a teaching fellow and taught data management to graduate students. She earned a prestigious dissertation fellowship from the Graduate
School at Brown to pursue her dissertation research. She has presented her work at the
iv
Annual Meeting of Population Association of America, International Union for the
Scientific Study of Population Conference, Population Reference Bureau, and Brown
University. She is expected to receive a Doctor of Philosophy degree in May 2008. She is looking forward to start a new phase in her life and continue to contribute to research in
Social Demography.
v
ACKNOWLEDGEMENT
Without the encouragement and support of some extra ordinary people this dissertation would not have been possible. First, I want to thank my main advisor,
Dennis P Hogan, who provided valuable advice not only for this dissertation but also throughout my graduate student life. He inspired me with his encouraging words and at the same time asked critical questions that made me think and re-think in a very constructive way. He not only made me work hard but also supported me with great patience at times when I showed signs of frustrations. I also want to thank my other advisors –David P Lindstrom and Nancy Luke with great admiration. They supported me with important thoughtful critiques and suggestions and I have learned important skills from each of them. I am proud and honored to have worked with such caring yet professional advisors for my dissertation. In their own ways, each of them showed me the importance of critical thinking, organization and determination that is required to complete a dissertation. A special thanks to Andrew Foster who kindly shared the dataset with me.
I owe thanks to my fellow graduate students, especially, to Rebecca Altman,
Blessing Mberu, Holly Reed, Daniel Schensul, Adriana Lopez and Gabriella Sanchez-
Soto for sharing the moments of joy and pain of graduate student life. Heartfelt thanks go to Kelley Smith for becoming such a good friend and also for reading drafts of this dissertation and providing her thoughts. Especial thanks go to Isaac Mbiti for sharing data tricks, laughter and joy and for providing sugar support in the form of cookies
vi
mailing through FedEx when I was down with work. I also want to thank Muriel
Bessette for her tremendous support in administration.
Most importantly, I am indebted to my parents, Habibur Rahman Chowdhury
and Sajeda Choudhury, for all the support they provided during the entire course of my
life. They taught me how to dream and how to make those dreams come true. You are
my first teachers and I would not have been the person that I am today without your
constant love, care, support and guidance. Abbu, you are not with us today, but I know
how proud you would have been to see my accomplishment. You are always with me in
spirit and I dedicate this dissertation to you, the person with the biggest heart I have ever
seen. I thank my sisters Nusrat R Chowdhury and Saika Ahsan and my brother Sanjid R
Chowdhury for being such supportive and caring siblings. Your love and affection
inspire me to overcome each and every barrier I face in my life. I also want to thank my
parents-in-law, Eunus Ali and Safia Khatun for their encouragement in earning the
doctorate degree.
Finally, I am most grateful to Ali Ehsan Protik, my life partner, husband and best
friend. You always believed in me and supported me in every possible way from reading and re-reading the draft, solving STATA problems to feeding our son in the middle of the night, so I could concentrate on the dissertation. I am so proud, happy and grateful
to share my day-to-day struggles, accomplishments and dreams for the future with you.
It gives me tremendous pleasure to share my life with you. Our son, Ayaan Chowdhury
Ali, has become a constant source of immense inspiration, limitless joy and a sense of completeness. I want him to know how proud and content I am to be his mother.
vii
Contents VITA...... IV ACKNOWLEDGEMENT ...... VI LIST OF TABLES...... IX LIST OF FIGURES...... X CHAPTER 1. INTRODUCTION...... 1 CHAPTER 2: BACKGROUND, LITERATURE, THEORY AND HYPOTHESIS ...... 9 2.1 Introduction...... 10 2.2 Cultural Background of Marriage ...... 10 2.3 Marriage Transaction in India: Dowry and Brideprice...... 14 2.4 Theories and Research on Dowry and Brideprice...... 20 2.5 Hypotheses...... 29 2.6 Summary...... 35 CHAPTER 3: DATA AND METHOD ...... 37 3.1. Data...... 38 3.2 Sample...... 40 3.3 Empirical Model and Estimation...... 46 3.4 Key Variables...... 48 CHAPTER 4: PREVALENCE OF MARRIAGE TRANSACTIONS: TRENDS, CHANGES AND FACTORS AFFECTING THE PRACTICES ...... 58 4.1 Introduction...... 59 4.2 Prevalence of Dowry and Brideprice...... 59 4.3 Change in Prevalence Over Time...... 62 4.4 Who are More likely to Pay Dowry and Brideprice...... 65 4.5 Summary...... 73 CHAPTER 5: FACTORS AFFECTING DOWRY AND BRIDEPRICE...... 81 5.1 Introduction...... 82 5.2 Magnitude of Dowry and Brideprice ...... 82 5.3 Change in the Magnitude of Marriage Transactions Over Time...... 85 5.4 Determinants of the Size of Dowry and Brideprice...... 88 5.5 Summary...... 102 CHAPTER 6. DISCUSSION AND CONCLUSION...... 115 6.1 Introduction...... 116 6.2 Summary of Findings ...... 117 6.3 Limitations ...... 121 6.4 Direction for future research...... 123 6.5 Conclusion ...... 124 REFERENCES...... 126 APPENDIX A ...... 133
viii
List of Tables
Table 2.1: Key Hypotheses to be Tested:...... 29 Table 3.1: Summary Statistics for All India and by Region...... 57 Table 4.1: Regional Prevalence of Dowry and Brideprice...... 74 Table 4.2: Probability of Paying a Dowry: Logistic Estimates ...... 75 Table 4.3: Probability of Paying a Brideprice: Logistic Estimates ...... 76 Table 4.4: Probability of Paying a Positive Net-dowry: Logistic Estimates ...... 77 Table 5.1: Mean and Median Dowry by Region ...... 105 Table 5.2: Mean and Median Brideprice by Region ...... 105 Table 5.3: Mean and Median Net-dowry by Region ...... 105 Table 5.4: OLS Estimates for the Size of Dowry...... 106 Table 5.5: OLS Estimates for the Size of Brideprice...... 107 Table 5.6: Bride and Groom Characteristics by Bride’s Educational Status ...... 108 Table 5.7: OLS Estimates of the Determinants of Dowry Size by Bride’s Education Status ...... 108 Table 5.8: Mean and Median Dowry by Education Status of Groom ...... 109 Table 5.9: OLS Estimates of the Determinants of Dowry Size by Groom’s Education Status ...... 109 Table 5.10: Estimates of Factors Affecting the Size of Dowry and Brideprice Using SUR and Net-dowry Using OLS ...... 110 Table 5.11: Assortative Matching of Bride and Groom by Education ...... 111 Table 5.12: OLS Estimates of the Size of Dowry for the North and the South ...... 111 Table 5.13: OLS Estimates of the Size of Brideprice for North and South...... 112 Table 6.1: Support for key hypotheses...... 120 Table A1: Comparison Between Cases with Available Dowry Information, Zero Dowry and Missing Dowry...... 135 Table A2: Comparison Between Cases with Available Brideprice (BP) Information, Zero Brideprice and Missing Brideprice...... 136
ix
List of Figures
Figure 3A: Distributions of Women’s and Men’s Age at Marriage ...... 55 Figure 4A: Regional Prevalence of Dowry Over Time ...... 78 Figure 4B: Regional Prevalence of Brideprice Over Time...... 78 Figure 4C: Predicted Odds Ratio of Paying Dowry Over Time by Region ...... 79 Figure 4D: Predicted Odds Ratio of Paying Brideprice Over Time by Region...... 79 Figure 4E: Predicted Odds Ratio of Paying Dowry Over Time by Caste Affiliation...... 80 Figure 5A: Trend Line of Predicted Value of Dowry Over Time ...... 113 Figure 5B: Trend Line of Predicted Value of Brideprice Over Time...... 113 Figure 5C: Bride’s Average Years of Schooling Over Time ...... 114
x
Chapter 1. Introduction
1 2 Marriages in the Indian sub-continent are characterized by, among other things, transfers made at the time of marriage between families involved. These transfers of money or goods can go in both directions – from the bride’s family to the groom’s or vice versa. The former is known as dowry, while the latter is known as brideprice.
Although brideprice is a common practice in many parts of Africa, dowry is the dominant form of marriage transaction in the Indian sub-continent.
Dowry, which used to be a common practice among many communities in
Europe, East Asia and some other parts of the world, has disappeared in those regions with modernization1. However, in South Asia, it not only remains dominant but also has
flourished with modernization. According to one recent study, an average dowry in India
amounted to over two-thirds of a household’s assets, or about six times a household’s
annual income (Rao 1993a). This disproportionate amount of dowry can lead to severe
impoverishment and debt to the bridal family, especially if there are more daughters than
sons. As a result, unmarried young daughters can be viewed as burden in the family and
are likely to face sex-based negligence in parental households (Edlund 2006). Dowry not
only brings financial hardship to the bride’s family, but can also shape the destiny of a
bride’s life (Kumari 1989, Menski 1998, Bloch and Rao 2002). The status of a bride in
her husband’s family typically depends on the amount of dowry she brings along with
her at the time of marriage, and unmet demands of dowry may result in mental and
physical abuse. According to the National Crime Bureau of the Government of India
there are approximately 6,000 dowry-related deaths every year. According to Menski
1 For a brief historical profile of marriage transaction, see Botticini and Siow, 2002.
3 (1998), this number would be 25,000 considering both dowry deaths and other dowry- related violence.
Dowry is thus a pressing social issue in India, drawing much media attention and research interest by social scientists. In this dissertation, I analyze the practice of marriage transactions – dowry and brideprice in rural India by using data from Rural
Economic and Development Survey (REDS) 1999. Data from REDS is representative of rural areas of all Indian states. I investigate financial transactions associated with Hindu marriages from 1975 to 1999.
Much of the literature on marriage transactions has tried to determine the role of dowry in the process of marriage. There are mainly two opposing views – in one, dowry is paid as a pre-mortem bequest to daughters who may not be legally or culturally capable of inheriting parental property (Tambiah 1973, Zhang and Chan 1999, Botticini and Siow 2002). In the other, the price model, dowry is viewed as a pecuniary transfer to attract better match (Becker 1991, Rao 1993a, Anderson 2003). Most of the theories of dowry existence or inflation can be posited under these two motives of dowry payments.
What factors affect dowry can critically depend on what the role of dowry is in the process of marriage negotiations. However, researchers do not agree on the underlying motives of dowry. This dissertation does not attempt to analyze the motives of financial transaction in the process of marriage, which by itself is a significant research effort.
Rather, I analyze the determinants of dowry and explain them as evidence of either the bequest or the price theory. Arunachalam and Logan (2006) have argued in a recent
4 paper that both motives can actually co-exist in the same society for different groups of people.
In addition to the roles dowry plays in the process of marriage, there is an issue of measuring dowry. Most of the empirical works have treated dowry either as a one- sided or a net transaction from the bride’s family to that of the groom. Dowry, viewed as a unidirectional flow of cash or kind from the bride’s family to the groom, ignores the fact that brideprice can be paid at the same time for the same marriage. Although some researchers use net-dowry, dowry net of brideprice, as a measure of marriage transaction
(e.g., Rao 1993a 1993b, Dalmia 2004), they implicitly view the two processes as one. In this dissertation I analyze dowry and brideprice not only as independent institutions but also as interdependent practices while determining the factors affecting economic exchanges between families. This allows for examining the complete picture of marriage transactions, as one is likely to affect the other.
One sociological explanation of dowry existence at individual level is marriage hypergamy, where the bride’s family pays dowry to marry off their daughter to a man of higher status, caste or class of their own (Caldwell, Reddy and Caldwell 1983). This explanation of dowry existence falls under the price theory of dowry. Most studies proposing hypergamy as an explanation for dowry are theoretical in nature, and there has not been much systematic empirical inquiry about it. One reason for this is that there are very few cross-caste marriages in India (Driver 1984, Bradford 1985, Deolalikar and Rao
1998, Munshi and Rosenzweig 2005). Thus, caste hypergamy cannot be a valid reason
5 behind dowry existence at a broader level. Class or wealth hypergamy has been empirically tested, and there is some evidence supporting this theory2. This dissertation makes another attempt to explore this theory empirically with respect to wealth.
There is a sizable literature on the expansion and inflation of dowry in India. I use the term ‘expansion’ to refer to the incidence of dowry practice becoming more
prevalent in society over time. The term ‘inflation’ is used to refer to the increase in size
of dowry over time. While the terms ‘expansion’ and ‘inflation’ are dynamic in nature as
these refer to the changes over time, ‘prevalence’ and ‘size’ of dowry refer to the static
status of dowry or brideprice at a certain point in time. Empirical evidence on dowry
inflation is mixed. Epstein (1973), Srinivas (1984), Paul (1985), Upadhya (1990) and Billig
(1992) argued in favor of dowry inflation. However, their studies were based on very
small non-random samples. Examining village-level data representative of South (central)
India, Rao (1993a) and Anderson (2003) found evidence in favor of dowry inflation.
Edlund (2000) and Dalmia (2004), among others, did not find any evidence in favor of
dowry inflation. In this dissertation I also find no evidence of dowry increase over time.
In fact the real size of dowry and net-dowry are found to have declined between 1975
and 1999. The real value of brideprice has also declined in the sample period.
Empirical studies on dowry expansion are mostly ethnographic. Thus, it is not
clear whether the expansion of dowry is restricted within specific castes and
2 Botticini and Siow’s (2002) interpretation of Indian dowry is close to hypergamy. Measuring hypergamy by the difference between bride and groom’s parental landholding, Rao (1993) also found some evidence supporting the theory.
6 communities or applicable throughout India. On the other hand, based on anthropological research, it is argued that the practice of brideprice has shrunk over time. Lack of large-scale empirical analysis restricts us from making any conclusions about these trends at the country level. I explore the diffusion of these two practices over time. I provide empirical evidence that there has been an expansion of dowry practice over time in India. Prevalence of dowry, as defined by the proportion of marriages paying any dowry, has gone up in the same period when the size of dowry was going down. However, there has not been any significant change in net-dowry and brideprice in general with respect to prevalence. One interesting finding about the prevalence of dowry and brideprice is that both the practices have expanded over time among the lower caste population.
One explanation of dowry expansion and inflation at the macro level among the
lower castes is known as sanskritization. This theory suggests that the inclination of the
lower castes to associate themselves with high castes is the reason for them to adopt the
practice of dowry (Epstein 1973, Srinivas, 1984). It should be mentioned here that dowry
originally used to be a high caste phenomenon (Srinivas 1984, Miller 1981). This
behavioral explanation, which does not fall under either price or bequest motive of
dowry existence, comes mostly from anthropological observations based on non-random
small samples from different communities or villages in India. Though sanskritization
may explain the expansion of dowry among new groups of people, it may not explain
dowry inflation in general. In any case, I have not encountered any research that has
empirically tested this theory using large-scale sample. Thus, sanskritization as an
7 explanation for the increase in dowry prevalence and inflation remains largely an untested empirical question. This dissertation makes the first attempt, to my knowledge, to address this issue with data from a national sample representative of rural India.
Given the context and literature on dowry and brideprice practice in India, this dissertation is built on an attempt to answer a series of relevant research questions. The aim of this dissertation is to provide a clear picture about the prevalence, expansion and inflation of dowry and brideprice practices in India for the sample period from 1975 to
1999. Using logistic, ordinary least square and seemingly unrelated regression models, this dissertation investigates two aspects of determinants of marriage transactions – the probability of paying a dowry and brideprice and the size of the two practices. Dowry and brideprice are analyzed independently and interdependently to examine how and whether these practices interact with each other. I also test ‘marriage squeeze theory’3,
the demographic explanation of dowry existence and inflation and the ‘theory of income diversification’4 on existence and expansion of dowry at individual level. Both ‘marriage squeeze theory’ and the ‘theory of income diversification’ fall under the price theory of dowry.
This first chapter of the dissertation presents an overview of the dissertation as a whole. It provides the motivation behind the project, research objective and organization of the dissertation. The rest of the dissertation is organized as follows. Chapter Two
3 Marriage squeeze theory asserts that marriages are associated with dowry if there are more women in the marriage market than men. 4 According to the theory of income diversification, parents marry off their offspring to locationally distant households in order to extend the informal support network to diversify household income risk.
8 serves to present a description of cultural context for Indian marriages and marriage transfers, and provides an overview of the background literature and theories. Special attention is paid to the definition and evolution of marriage transfer. The chapter then explores the literature on marriage transactions focusing on expansion, inflation and factors affecting dowry and brideprice. I then synthesize the bodies of literature presented in the chapter to generate testable hypotheses. Chapter Three explains the data and methods to be used in this study. It reviews the data source and discusses the sample of the analyses. It defines and explains the construction of the variables to be used in the analyses. Empirical models for estimation are also discussed in this chapter. Chapter
Four is the first of two result chapters. After providing a descriptive picture of the regional prevalence of dowry and brideprice, this chapter analyzes the prevalence of dowry and brideprice practices and their change over time. Hypotheses with respect to prevalence and factors affecting probability of paying dowry and brideprice are tested in this chapter using logistic regression models. Chapter Five focuses on the magnitude of marriage transactions and factors affecting the size of dowry and brideprice. It provides a picture of regional variation in magnitude of the two types of financial exchanges and systematically tests the relationships between different factors and marriage transactions posited by theory. In addition to independent analyses of dowry and brideprice, these two practices are also analyzed as interdependent institutions in this chapter by using the concept of net-dowry and using seemingly unrelated regression models. In the final chapter of this dissertation, I review and synthesize the findings from previous chapters.
I discuss the limitations of this study, offer policy implications and suggest directions for further research.
Chapter 2: Background, Literature, Theory and hypothesis
9 10 2.1 Introduction
In order to understand the institution of marriage and marriage transaction in the context of India, first we need to understand the complex pattern, practice, and cultural norm surrounding marriage. In this chapter, the socioeconomic and cultural background of marriage and its changing nature in India is discussed. Then, I provide definitions of dowry and brideprice, briefly discuss the evolution and social consequences of these practices in India and outline theories with respect to dowry existence and expansion.
Finally, based on these discussions I set up hypotheses for analyses.
2.2 Cultural Background of Marriage
Marriage is one of the most important events in the life course of Indian men and women, marking the transition to adulthood. India is a country with heterogeneous culture with respect to marriage practice, marriage and kinship structure, norms, ideology, gender roles and economic transaction surrounding marriage. The seminal study on kinship and gender in India by Dyson and Moore (1983) spotlighted the regional, especially the North-South differences on gender, kinship and marriage related issues. Despite differences, there are some common threads across region, states and
Hindu communities about the practice that allows us to elaborate on Indian culture of marriage under one heading.
11 Importance of marriage in Hinduism
Marriage is considered as a sacramental union in the Hindu faith and is almost universal for both men and women throughout India. According to Hindu faith "One is incomplete and considered unholy if they do not marry" (Prakasa, pg. 14, 1982). Thus, an unmarried girl of marriageable age may face divine sanction and her family may be subjected to social stigma for not arranging her marriage at the socially preferable age.
Using 1981 census data, Rao (1993b) reports that ninety-nine percent of men are married by the age of 25 and for women this proportion is achieved by the age of 20 in South-
Central India.
Age at marriage
Traditionally, early marriage for girls is supported by social and cultural factors though there has been slow but persistent rise in the age at marriage since 1928 after the passing of Child Marriage Restraint Act5. The Act was passed in an attempt to abolish
child marriage by providing a legal age at marriage, which was 14 years for girls and 18
years for men. The mean age at marriage for women rose from 13 to 17 years between
the first quarter of the twentieth century and 1971. For men, the mean age climbed to 22 years from 20 years within the same time period (Caldwell, Reddy and Caldwell 1983). At present since 1978, the legal age for marriage in India is 18 years for females and 21 years for males. According to Census 2001, the mean age at marriage for women and men are
18.3 and 22.6 years, respectively. In the sample population of this dissertation, forty-six percent of all brides and thirty-five percent of all grooms who married after 1978 are less
5 Also known as Sarda Act. This Act was passed in 1929 and came into force in 1930.
12 than 18 and 21 years old respectively. This indicates how pervasively the law has been ignored in rural areas. There exist high regional variations at age at marriage, Southern states exhibiting a comparatively higher age at marriage.
Anthropological studies claim that there is a preference for younger bride in the marriage market, especially in rural areas. While addressing the existing preference for early age at marriage for girls in the society Caldwell, Reddy and Caldwell (1983) state the following reasons behind that preference in rural Karnataka6,
“In many societies a young bride is preferred, so that her personality can be moulded by both her husband and his parents. This is important in India, too, but traditionally it has not provided the main motivation for early marriage of women in the study area (rural Karnataka). That motivation was provided by divine sanctions against girls who failed to marry before menarche, and against the family that erred in this way.” (p. 345)
Marriage arrangement and expectation
The ideal timing and age of marriage, be it early or late, is typically decided by the parents, especially in rural areas. Parents not only decide the age at which to get married, but also choose the appropriate partners for their offspring. Selecting a partner for a son
also means selecting a daughter-in-law who will play the vital role in reproducing next
generation for the family. Thus, finding a perfect partner of desirable qualities has its
own importance and can be a challenging task as perceived by people, especially in rural
6 Note: Karnataka is not an early marrying state, average age at marriage for women is 18.9 years (Census, 2001)
13 areas. It is common for people to use social networks or matchmakers to locate potential bride or groom of appropriate match. Socio-economic status and most importantly religion and caste background are frequently used as the basis for matching in the marriage market. In the Indo-Aryan-speaking north, a family seeks marriage alliances with people to whom it is not already linked by ties of blood. On the other hand, In the
Dravidian-speaking south, a family seeks to strengthen existing kin ties through marriage, preferably with blood relatives. In South India, marriages are preferred between cousins and even between uncles and nieces though this trend has been changing slowly over time (Caldwell, Reddy and Caldwell, 1983).
Generally, in those rural areas where within family marriages are not common, arranging matches between strangers without the couple meeting each other is not uncommon. Parents and other relatives generally come to an agreement on behalf of the couple.7 Because parents do the matching, family traits rather than individual traits are
likely to be given more importance. “Marriage is treated as an alliance between two
families rather than a union between two individuals” (Prakasa, pg-15, 1982). There are
reasons for why family characteristics may play important roles in marriage matching.
One big reason is the social norm of intergenerational co-residence, involving parents and sons, together with the norm of patrilocal exogamy on the part of daughters in
almost all regions of India. Sons are the preferred source of old-age support for parents
and are generally expected to co-reside with the parents even after marriage. This norm is particularly binding for the eldest son in most areas. A daughter, on the other hand,
7 For details in how a marriage is arranged in South India, see May 1986.
14 moves out of her patrilocal residence after marriage and starts co-residing with her parents-in-law in case her husband co-resides with them. Daughters-in-law are the preferred source of old-age personal care (Kuhn and Protik 2007).
Hindu society is stratified by jati or caste system and within each broad caste there are sub-castes. Prevalence of cross-caste marriages is very low in rural areas. Using a representative sample of rural Indian households of 16 major states of India, Munshi and Rosenzweig (2005) found that inter-caste marriage was about nine percent for the rural Indian population in 1999. Using data from different surveys, they also found that inter-caste marriage was a little less than eight percent in Bombay city in 2001 and six percent in South Indian plantations in 20038.
2.3 Marriage Transaction in India: Dowry and Brideprice
One very important aspect of Indian marriage is that families transfer goods and
services on the occasion of marriage. Dowry and brideprice are the two types of transfer
that are present in Indian marriages. These two types of payments are not mutually exclusive, rather can occur simultaneously (Rao 1993b). Dowry, which is the dominant
practice between the two, has received wide attention from social scientists at large,
whereas, brideprice is mostly neglected by researchers other than anthropologists.
8 For more example on this, see Reddy and Rajanna 1984, Driver 1984, Bradford 1985, Deolalikar and Rao 1998.
15 Conceptualizing Dowry and Brideprice – Definition
Marriage transactions either in form of cash or kind made by the bride’s family to either the bride or the newly married couple or the groom’s family are broadly classified as dowries. The definition of dowry is important for research and policy purpose as policy implications may vary depending on the definition of the practice. Menski (1998) categorized dowry into three types based on its nature and the way it is presented and argued in writings in Indian context. The first is known as Stridhana, which is gifts, jewelry, household goods and or other property that is given to the bride by her family during the marriage rituals. These payments are voluntary in nature and often viewed as pre-mortem inheritance. The bride enjoys the ownership right over these payments though this may not be the case if the bride herself is perceived as property by the groom and his family. The second form of dowry constitutes the expenditure that is made on the occasion of marriage celebration. Indian marriages are recognized for their conspicuous spending and families view this spending as a way to maintain their status.
Neither the bride or the groom or his family directly benefits from this second form of dowry. The third form of dowry, which generated a lot of media attention, is the payment of property in form of cash or kind that is expected or demanded by the groom and his family as a condition of marriage. This third form of dowry can be noted as
‘groomprice’.
Like dowry, brideprice can also take different forms. Brideprice is the transfer of goods, livestock or money paid by groom or his family to the bride’s family. Thus, considering who gets the payment, brideprice can be converse of groomprice but not
16 dowry in general. In most economics literature of marriage transaction, where dowry is defined as the payment made by bride’s family net of the transfer made by the groom, dowry and brideprice are considered as reverse of each other9. Viewing negative dowry
as brideprice in Indian context can be problematic, as data on dowry does not distinguish
whether the amount paid is stridhana or groomprice or a combination of both. If dowry is
consisted of stridhana in addition to groomprice, the measure of dowry will be positively
biased against brideprice. In this dissertation, I use the term dowry to refer to any
payment made by bride’s family and the term net-dowry is used to refer to payments
made by bride’s family net of what is paid by the groom or his family.
Evolution from Brideprice to Dowry
Originally, in Indian societies, the part of dowry, which is demanded by groom
and his family, was absent and the payment of dowry was voluntary as it was considered meritorious in dominant Hindu religion. According to the holy text “The Laws of
Manu”, one of the ten paths to reach moksha or enlightenment in Hinduism is kanyadana, the act of giving a virgin bride to the groom along with financial and/or other gifts that is known as dakhshina or dowry. Paying dowry used to be widely practiced only among the high caste Hindus of Northern states. As opposed to the Northern states, the direction of transfer used to be from the groom and his family to the bride and her family in most of the South Indian societies in the past. In the Southern region, the custom of paying brideprice was widely practiced even among high caste Brahmins
(Srinivas 1989). These regional differences in the practice of marriage transactions are
9 See Deolalikar and Rao 1990, Rao 1993a, Rao, 1993b.
17 widely recognized by social scientists (Miller 1981, Kolenda 1987). However, researchers have documented a substantial reduction in the gap with respect to marriage transaction between the South and the North in recent decades as the practice of dowry replaces the practice of brideprice in Southern states (Epstein 1973, Caldwell, Reddy and Caldwell
1982,1983, Caplan 1984, Bradford 1985, Ifeka 1989 and Rao 1993b).
Dowry, which used to be a divine part of marriage ritual became a serious social problem when grooms and their families started to demand certain amount of dowry at the time of marriage negotiation, making dowry somewhat mandatory for the families of the bride. Srinivas (1996) summarized the sharp distinction between modern dowry (the third form of dowry) and the traditional respected custom though this distinction is absent in empirical dowry studies due to lack of data availability on dowry composition.
In addition to the change in the meaning of dowry, the ownership rights have also changed as groom’s family rather than bride enjoys the rights over the payments
(Paul 1985). It is difficult to pinpoint the time when these transitions have actually begun due to lack of systematic large-scale research on this topic. Anthropological studies suggest that the change occurred in the middle of nineteenth century. Around this time the lower caste also adopted the practice of dowry (Alexander 1968). Due to the potential adverse effects of modern dowry on brides, their families and society at large, receiving or paying dowry has been made illegal since 1961 under Indian Civil Law. But this act provided very little or no support at all to control the spread of dowry prevalence.
18 Dowry and the Civil Law:
The first nationwide systematic legal attempt to control and prohibit the practice of dowry was taken in 1961, as noted above, by introducing “Dowry Prohibition Act”.
However, the concern about dowry and the use of the mechanism of law dates back to
1939 when Sind Leti-Deti Act was introduced. The “Dowry Prohibition Act” decrees,
“to give, take, or demand a dowry is an offense punishable by imprisonment and fines”
(Diwan, pg.77, 1983). In the act the dowry is defined as “any property or valuable security given or agreed to be given either directly or indirectly by one party to a marriage to the other party to the marriage, or by the parents of either party to a marriage or by any other person, to either party to the marriage or to any other person at or before or after the marriage as consideration for the marriage of the said parties”
(Diwan, pg.77, 1983). The explanation to the section states that “any present made at the time of marriage to either party to the marriage in the form of cash, ornament, clothes or other articles, do not count as a dowry” (Diwan, pg.77, 1983). This act apparently failed to control the expansion of dowry prevalence due to its inherent weaknesses in the definition of dowry. Under this definition, it was difficult to prove any transaction as dowry. With an aim to overcome the inherent weaknesses of the Dowry Act, the Dowry
Prohibition (Amendment) Act was passed in 1984 and brought into force in October
1985. The Dowry Prohibition (Amendment) Act 1984 defines dowry as “any property or valuable security given or agreed to be given …. in connection with the marriage” instead of
‘as consideration of marriage’. Another significant change that this Act brings is higher penalties for violation of the law. Any person gives, takes or supports the giving or
19 taking of dowry is punishable with imprisonment for not less than six months and the person also has to pay a fine which may extend to ten thousand rupees (Menski, 1998).
Consequences of Dowry Practice in India
The status of a bride in her husband’s family typically depends on the amount of dowry she brings along with her at the time of marriage. A bride whose family fulfills the dowry demand at the time of marriage usually enjoys better status in her husband’s family and better treatment from her in-laws10. Consequences faced by the bride if dowry demands are not met include mental and physical abuse of young wives, ill-treatment and neglect. The extreme outcome of dowry related violence is suicide of the bride as a
reaction to dowry related abuse and accidental burning of the bride. Although these outcomes are not accepted in the society, they are not totally uncommon either. This so- called ‘accidental death’ of a bride by burning is commonly termed as “bride-burning”.
From a field survey of dowry victims in Delhi who are essentially the brides, Kumari
(1989) found that the dowry demands are so disproportionately large compared to the
incomes of the brides’ parents that only about eleven percent parents in the sample
could fully give the demanded dowry. The rest of them could pay only a portion of that
and promised to provide the rest in future. Nevertheless, Kumari (1989) looks at only dowry victims, so it is difficult to anticipate what is happening in the entire population.
The adverse outcome of dowry practice has received much media attention and has been highlighted by NGOs. However, there is a lack of actual rigorous research on dowry
10 For positive association between dowry payment and bride’s status, see for example, Amin, Suran, Huq and Chowdhury (2004), Zhang and Chan (1999)
20 related violence to capture the real picture at a national level. One study shows that in recent period, average dowry can amount to over two-third of a household’s asset or to about six times a household’s annual income (Rao 1993a). This disproportionate amount of dowry can bring severe impoverishment and debt to bridal family especially if there are more daughters than sons. As a result, unmarried young daughters can be viewed as burden in the family and are likely to face sex-based negligence in parental household.
Dowry not only brings economic hardship to the bride’s family, it can shape the destiny of a bride’s life too11. According to the National Crime Bureau of the
Government of India there are approximately 6,000 dowry related deaths every year.
According to Menski (1998), this number would be 25,000 considering both dowry
deaths and other dowry related violence. Another study has shown that in Mumbai one
quarter of deaths among females of age 15 to 30 years are linked with dowry violence
(Karlekar 1985). Normative judgment of dowry or brideprice is not the goal of this
dissertation, but the potential outcome of dowry practice is worth mentioning.
2.4 Theories and Research on Dowry and Brideprice
A large body of literature exists on the institution of marriage transaction. The
main focus of this literature in Indian context is on dowry since dowry seems to be the more common practice and attracted immense attention in the media for its potentially adverse outcome.
11 For examples, see Kumari, 1989, Menski 1998, Bloch and Rao 2002.
21 There are mainly two opposing views – in one, dowry is paid as a pre-mortem bequest to daughters who may not be legally or culturally capable of inheriting parental property (Tambiah 1973, Goody 1976, 1990, Botticini and Siow 2002, Zhang and Chan
1999). In the other, the price model, dowry is viewed as a pecuniary transfer where the party with more bargaining power receives dowry and the other party pays dowry as to attain a better match or sometimes a match at all. Dowry according to this definition is similar to groomprice (Becker 1991, Rao 1993a, Anderson 2003). This economic model of marriage by Becker (1991) derives dowry and brideprice (reverse of dowry) as shadow prices related to the surplus of the joint value of the marriage over the single utility gained by one of the spouses. According to this model, the spouse, who gains less from marriage, is compensated by the other spouse or his/her family. These two opposing explanation of dowry existence are not mutually exclusive. In a recent empirical research on dowry motives, Arunachalam and Logan (2006) have shown that bequest and price motives of dowry payment are very much likely to co-exist in explaining dowry existence.
There exist several theories to explain dowry existence and expansion that we can consider under price theory. These are women’s economic role, marriage squeeze and hypergamy. Bequest theory, women’s unproductive economic role, and hypergamy are the explanations available for dowry existence. On the other hand, marriage squeeze and change in women’s economic role are the theories that have been used to explain dowry expansion, rise and disappearance of brideprice as the dominant practice in the South.
Other than bequest and price theory there exist another argument to explain dowry expansion and disappearance of brideprice, which is known as Sanskritization.
22 The bequest theory provides the rationale for the existence of dowry in pre- modern society with gender inequality arguing that dowry acts as a form of pre-mortem inheritance. While sons obtain their inheritance upon death of the parents, daughters receive their share when they marry. The underlying assumption here is that daughters do not have legal rights to inherit family property. This theory is applicable to dowry explanations both at the micro and at the macro level. According to this theory, dowry is one way of receiving proper share of parental family property. This is particularly applicable for Hindu women of past generations who did not have legal or practiced rights on their ancestral or joint family property.
However, with current practiced form of dowry, the above-mentioned argument is not sufficient to explain dowry existence. The reason behind this is two fold. Firstly,
Indian women have legal inheritance rights since 1956 though the applicability of these rights is questionable. Secondly, the dowry argument of the bequest theory takes it for granted that dowry is paid to the bride. However, this is less likely to be the case in current Indian scenario, where in most of the cases dowry is a demanded gift from the bride-giver. Therefore, even if parents spend disproportionately large amount of money or property to pay dowry, it does not necessarily go under the bride’s ownership. One general exception to this is jewelry and personal items given to the bride. Following this line of argument, dowry cannot be viewed as a substitute for inheritance of property.
Thus, it apparently fails to be the main explanation of existence of the current form of
Indian dowry practice.
23 Unlike bequest theory, price theory is more market oriented, where both men and women have certain prices in the marriage market. Dowry is paid if men are attached to higher prices compared to women. In the alternative scenario, with women being highly valued, brideprice will be the dominant practice.
The explanation of marriage transaction with respect to women’s economic role is the first argument to link dowry and brideprice clearly, which falls under price theory and views dowry solely from economic perspective. This hypothesis argues that when women produce or earn little or nothing tangible, they are viewed as a burden. So, the bride’s family pay dowry to shift this burden to the groom’s family. The lower the return from women’s work, the higher the bride’s family pays to compensate. This notion of dowry was first suggested by Boserup (1970). She observed that women’s economic role has direct connection with marriage system and transactions. Polygyny and brideprice are generally associated with female farming systems, while monogamy and dowry are affiliated with male farming systems. Under this line of argument women are viewed either as asset that brings brideprice to the natal family during marriage or a liability for which her natal family needs to compensate at marriage.
Following the explanation with respect to women’s economic role for dowry existence, Rajaraman (1983) suggests that dowry has increased in South Asia as the process of development and introduction of new technology in the agricultural field have diminished the economic role of women and their rates of (informal) labor-force participation. The change in economic role of women is responsible for the switch from
24 the practice of brideprice to the practice of dowry. This change happened since work traditionally done by women has been displaced by men with technological adaptation.
Therefore, daughters who used to be assets of the family are now considered as burden and parents are willing to pay dowry to switch this burden to another family.
Opposing this view, Nair (1986) made the argument that it is not women who were affected by technological adaptation in agricultural fields in India. This is because women were only involved in lowest strata jobs like weeding, transplanting, or harvesting various crops that were not displaced by adaptation of new technologies. Moreover, in
Indian agriculture, the participation rates for women workers remained constant at around twenty-nine percent between 1959 and 1979 that goes against the claim of
Rajaraman (1983)12. Thus, it is unlikely to be the case that women’s economic
participation has reduced over time and that has indirectly increased the amount of
dowry.
The theory of ‘Marriage squeeze’ is a macro level demographic argument
asserting that since in a population with declining mortality younger cohorts are larger
than the older cohorts, there will be a relative scarcity of eligible men over women in the
marriage market as women tend to marry older men leading to escalation of dowries.
This explanation of dowry inflation was first suggested by Caldwell, Reddy, and Caldwell
(1983). The key assumption here is that dowry is the price of grooms and that females
drive up the price of men through competing for eligible men. Billig (1991), Rao (1993a,
12 For more on work participation of women in India see Randeria and Visaria 1984; Srinivas 1989.
25 1993b and 2000), Deolalikar and Rao (1998), Bhat and Halli (1999), and Edlund (2000) focus on marriage squeeze in India to explain the expansion of dowry practice. Rao
(1993b) first published a study that empirically tested the marriage squeeze hypothesis.
Using data from six villages in rural South-central India, he claims to find significant support for the argument of marriage squeeze to explain the rise of dowry. This finding is controversial since using the same data Edlund (2000) fails to replicate the same result, that is, the sex ratio of marriageable women and men has a positive impact on dowry.
Using detailed sex-ratio measure, Bhat and Halli (1999) find evidence of the existence of a large marriage squeeze against females in contemporary India which shows that marriage squeeze is not a felt scarcity rather a real numerical deficit in the availability of eligible marriageable men.
A competing sociological explanation of dowry existence and expansion is marriage hypergamy, where women marry men of higher status. When marriage hypergamy is present, dowry acts as a vehicle through which a family of a bride can marry off their daughter to a higher-status or higher-class family. The ideal for the parents of a girl is to marry their daughter into a family of greater prestige, wealth, or reputation compared to their own. Marrying the daughter into a lower status family not only binds their daughter to an inferior family but also lowers their own prestige in the eyes of their caste fellows. Thus, this micro-level behavior and the competition among the families of the brides for high quality grooms are responsible for the macro-level
26 existence of dowry and its rise13. Hypergamy is most applicable for those closest to the top of society’s status pyramid. Those who are closer to the bottom have fewer restrictions with respect to the status of marriage partners. Low caste and poor families are not capable of ‘buying’ high caste, high quality groom in any case. Some argue that this is the reason why marriages are arranged among status equals and brideprice or some other form of marriage transaction than dowry is more common among the poor and the lower castes (Miller 1981, Srinivas 1984).
The marriage hypergamy argument fails to explain why the practice of dowry has expanded beyond high caste, wealthy families and over time became a common practice among low caste, poor families replacing the practice of brideprice. Recognizing Indian marriage hypergamy, Billig (1991) made the argument that the causation of marriage squeeze can go beyond demographic sex-ratio based understanding of marriage market and marriage squeeze can be perpetuated by status hypergamy and one of the outcomes of that is dowry inflation.
Apart from bequest and price motives, social anthropologists propose a group- level behavioral explanation of Indian dowry expansion beyond high caste and among those who formerly used to practice brideprice (Epstein 1973, Tambiah 1973, Srinivas
1984, 1989). The idea is known as sanskritization, which was first proposed by Srinivas
(1952). According to this hypothesis, the lower castes adopt the behavior of upper castes
13 For evidence on the positive relation between dowry size and quality of the groom see, for example, Hooja (1969), Rao and Rao (1980), Caldwell, Reddy and Caldwell (1983), Caplan (1986), Billig (1992).
27 as a mean of acquiring higher social status. Dowry, which is a traditional practice of the highest caste (Brahmin), is a reflection of upwardly mobile imitative behavior by all other castes. In the view of this argument, the lower castes, former bride-price givers, adopted dowry to demonstrate their adherence to high caste notion of spiritual and social propriety in an attempt to gain social and economic advancement as high caste cultural norm became the basis of cultural homogenization.
Two weaknesses of the Sanskritization theory have been stated by Rao (1993b).
First, it is not very believable that the benefits gained by lower castes in behaving like dowry practicing higher castes (Brahmins) is greater than the immense destitution they often have to suffer or face by paying dowries. Secondly, using data from villages of two
South Indian states, Rao (1993) found that the lower castes have not only switched from brideprice to dowry, rather there has been an upward shift in real dowry payments even in castes, which historically have been paying brideprice. According to him the argument of sanskritization fails to explain the rise of dowry among high castes. Analyzing all-India rural data I found the real size of dowry has declined at both country and regional level.
Thus the relevance of sanskritization in this case is irrelevant here.
The availability of resources can produce real constraints in the size of marriage transactions paid at marriages. Number of sisters a bride has can reduce the size of dowry that the bride’s parents are willing to pay. This could be either because of cash constraints or parents’ desire to pay equal size of dowry to each daughter especially if the dowry is paid as a bequest (Botticini 1999, Dalmia 2004). For brideprice practicing
28 societies, this relation will be opposite. Brideprice is paid to the bride’s family; hence, men with more sisters will have more resources attained from their sisters’ marriages to pay for their own brideprices. Therefore, number of sisters the groom has may increase the size of brideprice in a brideprice practicing community.
Rural households are likely to marry off their daughters to locationally distant households in order to broaden their safety nets and ensure informal insurance to mitigate income risks from external economic shocks. This is especially the case with agricultural households for which natural climatic shocks like rainfall variations, flood and the like are important sources of income shocks that differ by location. Households from distant places are less likely to share similar shocks at the same time. Households linked by marriages belong to the same informal support network by cultural norms.
Thus, it is advantageous to parents to marry off their offspring to distant areas. This argument of income diversification was made by Rosenzweig and Stark (1989). The preference to diversify income risk by marrying off the offspring to a distant place may provide a household-level incentive to pay larger dowries to grooms of far away places.
Similarly, grooms are also likely to pay larger brideprices the farther away the bride lives.
In this dissertation, I empirically test several theories of dowry existence, expansion and inflation– marriage squeeze theory, hypergamy with respect to wealth, sanskritization and theory of income diversification. I test some indirect aspects of bequest theory and theory of women’s economic role.
29 2.5 Hypotheses
Review of marriage transaction related theories and previous literature suggest several hypotheses. In this dissertation, I explore a number of selective hypotheses that stem from theory to be tested statistically. These hypotheses are presented in Table 2.1.
The empirical measurement of these hypotheses and their testing are explained in the next chapter.
Table 2.1: Key Hypotheses to be Tested: Hypothesis Level Basis A Prevalence Macro Anthropological observation A.1 Prevalence of dowry is increasing over time. A.2 Prevalence of brideprice is decreasing over time. A.3 Expansion of dowry practice is higher among Sanskritization the lower caste.
B Magnitude Macro Anthropological observation B.1 The size of dowry is increasing over time. B.2 The size of brideprice is decreasing over time. B.3 Dowry inflation is higher among the lower Sanskritization caste.
C Factors affecting dowry and brideprice
C.1i More marriages are associated with dowries if Macro Marriage squeeze there are fewer men than women in the marriage market. The opposite is the case with brideprice.
C.1ii The size of dowry is larger if there are fewer Macro Marriage squeeze men than women in the marriage market. The opposite is the case with brideprice.
C.2 Dowry is larger if the groom’s family is Micro Hypergamy wealthier than the bride’s family.
30 C.3 Better quality grooms receive larger dowries Micro Female competition and pay smaller brideprices. (quality of groom)
C.3i Education: Grooms with higher education receive larger dowries and pay smaller brideprices.
C.3ii Wealth: Grooms with more parental landholdings receive larger dowries and pay smaller brideprices.
C.4 Brides, who are more attractive in the marriage Micro Female competition market, pay smaller dowries and receive larger (quality of bride) brideprices.
C.4.i Education: Brides with higher education pay smaller dowries and receive larger brideprices.
C.4.ii Age: Younger brides pay smaller dowries and receive larger brideprices.
C.5 Brides from wealthier families pay larger Micro Bequest or Family dowries. status
C.6. Brides with more sisters pay smaller dowries Micro Resource constraint and grooms with more sisters pay larger (replacement) brideprices.
C.7 The size of dowry and brideprice are larger the Micro Income farther away the bride and groom’s pre- diversification marriage locations are.
Hypothesis A.1: Prevalence of dowry practice is increasing over time.
Hypothesis A.2: Prevalence of brideprice practice is decreasing over time.
These two hypotheses about the prevalence of the two types of marriage
transactions are based on anthropological observations. There is a lack of statistical
analysis in the literature due to the shortage of appropriate large-scale data to qualify
these observations.
31 Hypothesis A.3: Expansion of dowry practice is higher among the lower caste.
According to Sanskritization argument, the lower castes, former bride-price givers, adopted dowry to demonstrate their adherence to high caste notion of spiritual and social propriety in an attempt to gain social and economic advancement as high caste cultural norm became the basis of cultural homogenization. So, dowry expansion will be higher among the lower caste groups. This theoretical argument was made by social anthropologists. However, there is a lack of empirical evidence in the literature to support this theory.
Hypothesis B.1: The size of dowry is increasing over time.
Hypothesis B.2: The size of brideprice is decreasing over time.
These two hypotheses about the vertical magnitude of dowry and brideprice are also based on anthropological observations. As mentioned before, empirical findings from previous research based on statistical analysis are inconclusive. Village or community based ethnographic research suggest decline in brideprice, but this observation is not yet tested with large-scale data set.
Hypothesis B.3: Dowry inflation is higher among the lower cast.
As mentioned before, according to ‘Sanskritization’ argument, in the caste- stratified society of India, the lower castes follow the higher caste rituals and way of living to gain social status. Thus, as a marriage ritual they not only adopt dowry practice but also pay higher dowries. As a result, dowry inflation will be higher among the lower caste. Again, empirical evidence is not available to support this theory.
32 Hypothesis C.1i: More marriages are associated with dowries if there are fewer marriageable men in the marriage market. The opposite is the case with brideprice.
Hypothesis C.1ii: The size of dowry is larger if there are fewer marriageable men in the marriage market. The opposite is the case with brideprice.
These two hypotheses are based on marriage squeeze theory. These suggest if there are more women at marriageable age than men in the marriage market, on one hand, there will be high competition among the parents to marry off their daughters, hence, it is very much likely that the bride’s parents first will agree to pay dowry and secondly will pay larger dowry to remain in the competition. On the other hand, it is also likely that the groom’s family will have higher bargaining power in the dowry negotiation process. However, regardless of underlying motives, there will be an increase in dowry practice and amount paid. Similarly, in the opposite situation, with fewer women compared to men, brideprice will be paid and the size of brideprice will also be affected negatively. Empirical evidence supporting the marriage squeeze hypothesis in explaining dowry existence and inflation is inconclusive.
Hypothesis C.2: Dowry is larger if the groom’s family is wealthier than the bride’s family.
This hypothesis stems from the theory of marriage hypergamy with respect to wealth. Bride’s family pay larger dowry to attract wealthier groom in order to attain higher status by marrying off their daughter into a wealthier family.
33 Hypothesis C.3: Better quality men receive larger dowries at marriage and pay smaller brideprices.
Dowry is paid as a result of competition among marriageable women whose families compete with each other to marry off their daughters to better quality grooms.
In rural areas, where agriculture is the main occupation, land represents wealth.
Education is another trait that increases a groom’s quality by providing him with more opportunities in the labor market and income earning. Thus, men with higher education and larger landholdings are perceived as ‘better grooms’. This leads to the following two hypotheses:
Hypothesis C.3i: Men with higher education receive larger dowries and pay smaller brideprices.
Hypothesis C.3ii: Men with larger parental landholdings receive larger dowries and pay smaller brideprices.
Parents of brides pay larger dowries to the grooms who are more educated and come from wealthier families. Again, larger dowry can be paid to these grooms as they enjoy higher bargaining power in the marriage negotiation process and demand accordingly. Similarly, in a brideprice paying society, ‘better grooms’ pay less brideprice for similar reasons.
34 Hypothesis C.4: Women, who are more attractive in the marriage market, pay smaller dowries and receive larger brideprices.
Better quality brides are valued highly in the marriage market and hence, their families have to compensate less to marry off their daughters. In brideprice practicing communities, the better the bride is, the higher the brideprice is. Education provides additional opportunities in the labor market and it also provides knowledge and information especially about better child-care. Again, as mentioned before in this chapter, in India, women marry at young ages, which might imply that there is a demand for younger brides who are easily controllable. Thus, educated younger brides are likely to be considered as better quality brides. This provides us with the following two hypotheses.
Hypothesis C.4i: Brides with higher education pay smaller dowries and receive larger brideprices.
Hypothesis C.4i: Younger brides pay smaller dowries and receive larger brideprices.
Hypothesis C.5: Brides from wealthier families pay larger dowry.
Wealthier families possess more resources and if dowry is paid as bequest then the wealthier the family is the larger the share of pre-mortem inheritance will be and therefore, the larger the size of dowry will be. Alternative explanation is that wealthier families pay dowry to signal their wealth status to the community where the size of dowry is likely to be well publicized. As a result, the size of dowry increases with bride’s family wealth status.
35 Hypothesis C.6: Brides with more sisters pay smaller dowries and grooms with more sisters pay larger brideprices.
Due to resource constraint of the family, as the number of daughters increase, the amount paid as dowry for each daughter is likely to go down. In the case of brideprice, the more sisters a man has, the more brideprice his family receives from his sisters’ marriages and the more he is able to pay as brideprice for his own marriage.
Hypothesis C.7: The size of dowry and brideprice are larger the farther away the bride and the groom’s pre-marriage locations are.
The income diversification argument asserts that, in rural agricultural communities, to mitigate sudden income shocks originated from natural calamity such as flood, draught and the like, households diversify their income and sources of informal support. One way to extend the informal safety net is by marrying off the offspring to geographically distant places that are less likely to share similar shocks. Therefore, to attract a marriage partner from a distant place parents might offer larger financial exchange.
2.6 Summary
This chapter presents a review of the relevant literature from sociology, anthropology, economics and social demography. Discussions on the cultural background of marriage, its characteristics and the practice of marriage transactions in detail reveal the profound impact of marriage transaction on Indian marriages. Taken as
36 a whole, we see that the practice of marriage transaction has gone through a substantial change over time. We also see the theoretical importance of individual, household and community level characteristics of both bride and groom on affecting the prevalence and magnitude of both dowry and brideprice practice. Based on those theoretical arguments and previous literature a number of hypotheses are posed in this chapter. In the next chapter, I review the data sources and empirical model estimation that will be utilized to test these hypotheses.
Chapter 3: Data and Method
37 38 3.1. Data
I use data from the Rural Economic and Demographic Survey (REDS), a panel survey conducted by the National Council of Applied Economic Research (NCAER) since the early 1970s. The first round of REDS was conducted in 1971 and included complete village and household information from 4,527 households spread over 259 villages from 17 major states14 of India. The sample is representative of the entire rural
population of India though the middle and upper income households were slightly over-
sampled. The second round of the survey took place in 1982 and 250 of the 259 original
villages were revisited15. Households from the original 1971 survey, with at least one
member of the household remaining in the village were resurveyed, which totaled to
4,979 households. Approximately two thirds of the households were the same
households as in 1971. A third round of the survey was carried out in 1999. All 1971
villages were surveyed this time excluding the sample villages in Jammu and Kashmir16,
thus making a total of 242 villages. In this survey round, all surviving households from
the 1982 survey living in these 242 villages were surveyed again, including all split-off
households residing in the same villages as the original household. In addition, a small random sample of new households was also added. Because of household division and this new sample, number of households in the 1999 round increased to 7,474. I use data from the 1999 survey round for this dissertation.
14 These 17 states are Andhra Pradesh, Bihar, Gujarat, Haryana, Himachal, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh, West Bengal, Assam, Jammu and Kshmir. 15 A total of 9 villages from Assam were excluded for administrative reasons. 16 Jammu and Kashmir were excluded due to the political unrest prevailing in that area.
39 The 1999 household survey provides detailed information on asset ownership, incomes, and financial transaction at the household level. It also provides information on individual characteristics of household members. My principal use of REDS household data is to analyze the practice of marriage transaction in rural India and also to portray changes Indian marriages have experienced with respect to marriage transaction over time. For this purpose, it is important for the data to be representative of rural India.
However, although the original sampling scheme for the REDS panel data set was meant to create a representative sample of the entire rural population in the late 1960's, it is not clear how representative the sample is in 1999. This is because of the transition of some villages into towns and growth in new villages, which would not be included in the
REDS sampling.
To test the level of national representiveness of this survey Foster and
Rosenzweig (2003) compared data on rural households from the 1999 REDS with those on rural households from the 55th round of the National Social Surveys (NSS) of India, which was carried out in 1999-2000. The National Social Surveys of India (NSS) are large cross-sectional surveys that are designed to provide a comprehensive characterization of
Indian households. They found little or no overlap with respect to villages in the two surveys and strong cross-state relationships between variables constructed from the two data sets.
Additional Data: I also use data from Indian censuses in 1981, 1991 and 2001. Indian census is conducted in every ten years covering the whole population of the country.
40 Census provides individual level data along with district and state level information. For this dissertation, district17 level information such as sex ratio of marriageable men and women are constructed from census data. The most proximate decadal sex ratio in time is used as the proxy sex ratio of the year of marriage.
3.2 Sample
The sample for this study includes 2,154 Hindu marriages that were conducted within the time period 1975 to 1999. Muslims and others are different from Hindus on many observable and unobservable characteristics and dowry dynamics in Muslim marriages are expected to be very different from these of Hindu marriages. In the REDS data set, the religious background of eighty-nine percent of the households is Hinduism and unfortunately, there are not enough Muslims in our sample for any meaningful comparison. Most of the previous studies on dowry in India have also excluded Muslims.
For this dissertation, only household heads and their spouses are included in the sample. Marriage and dowry information were collected retrospectively from the head of the household. Respondents were asked how much they received or paid during marriage as dowry or brideprice. If the respondent is a male then the value of marriage transaction he received at the time of marriage from his in-laws is coded as dowry and the amount paid by him is coded as brideprice. For a female respondent, it is the opposite - amount
17 There are 28 states in India. This survey was conducted in 17 states. Each state is divided into basic units of government and administration called districts. Districts in turn are further divided into tehsils and eventually into villages.
41 paid by her family is coded as dowry and the value of marriage transaction received by her family is coded as brideprice. Demographic and economic Information on spouse’s natal household at the time of marriage were also collected retrospectively at the same survey18.
Retrospective data always has potential to be affected by recall bias. But as
Deolalikar and Rao (1990) mentioned, marriage in India is one of the major events in a person’s life, especially for women. Hinduism does not allow polygamy or informal unions and marriage is viewed as the most prominent way to enter adulthood for both men and women. Besides, marriage transaction represents a very large proportion of household income and asset, and is a factor that plays a very significant role in marriage
negotiation and decision-making. So, it is less likely to be subject to recall bias. Both
dowry and brideprice are measured in rupees19.
The marriages in the sample are spread over twenty-four years starting in 1975 and ending in 1999. Prices or value of things changes over time due to inflation of currency, which also changes cost of living. As a result, a thousand rupee in 1975 will be valued differently than the same thousand rupee in another year. Hence, marriage transactions paid in the year of 1971, for example, is not readily comparable to the dowry or brideprice paid in 1999 or any other year. To adjust for inflation and to make dowries and brideprices comparable between years, nominal dowries and brideprices from each
18 It is not clear from REDS documentation, whether the respondent or his spouse provided the information on spouse’s natal household. 19 1 Dollar = 40 Rupee (approximately) in 2008.
42 year needs to be converted to the value of a certain year, known as the base year. By converting nominal values of dowry and brideprice to the value of this base year one obtains the inflation adjusted real values of dowry and brideprice.
Researchers have used different indices to convert nominal values of dowry and brideprice to real ones. Amin and Cain (1998) and Arunachalam and Logan (2006) used price of rice to adjust the change in cost of living over time while investigating the rise of dowry and dowry motives in Bangladesh respectively. Rao (1993a, 1993b) and Deolalikar and Rao (1990) used price of gold to analyze dowry practice in India. Dalmia (2004) used consumer price index for the conversion to assess the determinants of dowry and demand for groom characteristics in India. Consumer price index usually includes prices of a collection of commodities that have significant use in everyday life. Thus, consumer price index is preferable than using one commodity such as the price of rice or gold to do the adjustment of cost of living over time. Using rural price index, which is a price index constructed using the price of commodities at rural areas, is even better in this context as the sample for the analysis in this dissertation comes from a rural population.
Thus In this dissertation, I use rural consumer price index20 to adjust for inflation over
the sample period.
It is a sensible practice to choose a base year that is ordinary in the economic
sense, a year in which events like flood, war or anything that can dramatically influence
prices did not occur. I use the price of 1999 as the base price. Although the choice of
20 Source: Price Indices 1999-2003, Economic Intelligence Division, Directorate of Economics and Statistics, Govt. of Gujarat. The base year of these price indices is 1960-61.
43 1999 as the base year is somewhat arbitrary, nothing extraordinary happened during this year nonetheless. Hence nominal values of marriage transactions (MT) of each year from
1975 to 1999 were transformed into the real values using the following formula:
Real value of MTi = (Nominal value of MTi) x (CPI1999/CPIi)
where, MTi = Marriage transaction of year i and i ∈ {1975, 1999}, CPI1999 is consumer
price index of year 1999 and CPIi is consumer price index of year i.
Key characteristics of the sample:
As mentioned before, the sample includes information about 2,154 marital
unions. Table 3.1 provides summary statistics of key variables for the sample cases used
for the empirical analysis in the next two chapters. The first column of table 3.1 shows the mean and standard error for all cases included in the sample. Dowry is highly prevalent in the sample – a dowry was paid in ninety percent of the marriages. On
average, in cases where dowry was paid, each dowry was equivalent to 39,035 rupees in
1999 prices. Unlike dowry, brideprice is much less practiced – a brideprice was paid in only eighteen percent of the marriages and the average size of brideprice is 4,941 rupees in 1999 prices, though the average approaches to 26,948 rupees if we consider only those
who paid brideprice.
As is evident from table 1, average age at marriage is quite low in the sample for
both males and females. Groom’s average age at the time of marriage is 23.76 year, one
44 year higher than the country average. For brides, this average age is 17.95 years, very close to the country average, which is 18.3 years (Census 2001).
Levels of education are also low among both males and females in the sample cases, which is not unexpected at all given the over all education level in India. A little less than three quarter of the grooms and a little less than half of the brides have attended formal schooling for more than one year. In the sample, average years of schooling for the bride and the groom are about 3 and 5 years respectively.
Grooms in the sample come from slightly wealthier families with respect to landholdings compared to the brides. Though high and middle income families are over- sampled in the REDS data. However, landlessness is not uncommon among this sample cases – about 23 percent of groom’s families and about 27 percent of bride’s families do not own any land of their own.
On average, the physical distance between a bride’s natal household and the new household where she moves after marriage is about 33 kilometers. There is significant caste variation in the sample with a majority being lower caste, about 43 percent. Thirty- three percent of the sample has high caste background and the rest of 21 percent are from middle caste. In the sample, for each marriage, at the district level there are 106 women at marriageable age available for each 100 men.
45 Regional variation in characteristics within the sample population:
There are four major regions in India - North, South, East and West. All four of these regions are represented in the sample with the North and the South being over represented compared to the East and the West, 55 and 24 percent respectively21.
Information on dowry and brideprice are missing for a large number of cases of Western region. The reasons behind these missing information are unknown and thus are excluded from the data set. Therefore, data for Western region will be subject to large sampling error. The second, third, fourth and fifth column of Table 3.1 presents the mean and standard error of the variables for the four regions. The sample shows very interesting regional pattern. The average size of dowry and brideprice are the largest in the West and the smallest in the North and the East respectively. The size of dowry is more than nine times larger in the West compared to the East. Brideprice, on the other hand, is almost six times larger in the West than that of the North. Net-dowry reflects the imbalance in marriage transaction, which is the highest in the Northern region and lowest in the West.
Between the two major regions of the sample - North and South, age at marriage for both bride and groom are higher in the South. Although there is no difference at average education level between grooms, brides’ average years of schooling is double in the South compared to the North. The distance between bride’s natal household and the household where she moves after marriage is also much shorter in the South. The district
21 States within the North region includes Bihar, Haryana, Himachal, Madhya Pradesh, Punjab, Rajasthan and Uttar Pradesh. South region includes Andhra Pradesh, Karnataka, Kerala and Tamil Nadu, Gujarat and Maharashtra falls under the Western region, while West Bengal, Assam and Orissa falls under the Eastern region.
46 level sex ratios (female/male) are 102 and 112 in the North and the South respectively suggesting Southern region having better female outcomes.
3.3 Empirical Model and Estimation
To determine the desirable characteristics, that are important in assessing the amount of marriage transaction from one family to the other, I use Ordinary Least
Square (OLS) regression method. I use dowry and brideprice as dependent variables and individual, household and community level variables as independent variables to examine which of these variables are important in determining dowry and brideprice. In particular, the following OLS regression models are used:
Di,j = β0 + β1Xi + β2Xj + β3Hi + β4Hj + β5Ci,j + β6Mi,j
+ β7Yi,j + β8SRi,j + β9RDi,j + ui,j (1)
BPi,j = γ0 + γ1Xi + γ2Xj + γ3Hi + γ4Hj + γ5Ci,j + γ6Mi,j
+ γ7Yi,j + γ8SRi,j + γ9RDi,j + vi,j (2)
where Di,j and BPi,j are dowry and brideprice transferred at the time of marriage between
bride i and groom j. Matrices Xi and Xj, contains individual characteristics like age and
years of education of bride i and groom j respectively. Similarly, Hi and Hj contains
household level variables like parental landholding and number of sisters of bride i and
groom j, respectively. Ci,j is a dummy variable which takes a value of one if bride i and
47 groom j belong to high caste22. Distance of marriage migration for bride i when married
to groom j is denoted by Mi,j, year of marriage by Yi,j and District level sex-ratio by SRi,j.
RDi,j includes two region dummy variables, one for the West and the other for the
North. Finally, ui,j and vi,j are random errors.
In the sample population, twenty four percent of all marriages, for which I have
dowry and brideprice information available, the financial transaction went in both
directions simultaneously. Therefore, it is reasonable to assume that the amount paid by
the groom’s family is not independent of the amount paid by the bride’s family. In other
words, dowry and brideprice for a particular marriage is jointly determined. In this case,
equation (1) and (2) may seem unrelated but there will be correlations between ui,j and vi,j.
This means that Cov (ui,j, vi,j) ≠ 0. Estimated coefficients in (1) and (2) will still be
unbiased and consistent, but will be inefficient.
To obtain efficient estimates, I use Zellner’s seemingly unrelated regression
model (SUR) to estimate equation (1) and (2) jointly. This provides efficient estimates
through the use of generalized least-square estimation (Greene, 2000). If the equations
are actually unrelated, then there is no payoff to generalized least-square estimation and
it will be the same as ordinary least squares. However, there is no cost either in the
econometric sense in using the SUR model when the equations are actually unrelated, i.e.
Cov (ui,j, vi,j) = 0.
22 Note that ideally a dummy for bride’s caste affiliation and a dummy or groom’s caste affiliation should be included. However, there are no cross caste marriages in the sample. So, a single dummy for caste affiliation for both bride and groom suffices.
48 To estimate the prevalence rate of dowry and brideprice over time, I use logistic regression model. The following logistic regression models are used to predict the odds of paying a dowry, equation (3) and the odds of paying a brideprice, equation (4):
(β0 +β1Xi + β2Xj + β3Hi + β4Hj + β5Ci,j + β6Mi,j + β7Yi,j + β8SRi,j + β9RDi,j + ui,j) Pdi,j / (1-Pdi,j ) = e (3)
(γ0 + γ1Xi + γ2Xj + γ3Hi + γ4Hj + γ5Ci,j + γ6Mi,j + γ7Yi,j + γ8SRi,j + γ9RDi,j + vi,j) Pbpi,j / (1-Pbpi,j ) = e (4)
where Pdi,j is the probability of paying a dowry in the marriage between bride i and
groom j and, (1- Pdi,j ) is the probability of not paying a dowry. Similarly, Pbpi,j is the
probability of paying a brideprice to bride i’s family by groom j and, (1- Pdi,j ) is the
probability of not paying a brideprice. Matrices and variables Xi, X,j, Hi, Hj, Ci,j, Mi,j, Yij,
Ri,j, RDi,j contains the same variables - individual, household and community level, that
were included in regression model (1) and (2). Coefficient of the year of marriage
variable should be positive if the prevalence rate of dowry (brideprice) is expected to
increase over time.
3.4 Key Variables
Dependent Variable:
The dependent variables – dowry and brideprice – are measured in two different
ways for the two different sets of equations above. First, to identify factors determining
dowry and brideprice, I use the amounts of dowry and brideprice as dependent variables
49 for OLS regressions (1) and (2) respectively. Real values of dowry and brideprice are used as amounts paid by households. As mentioned above, I obtain these real values of dowry and brideprice by inflating nominal values using rural consumer price index and using 1999 as the base year. Naturally, real amounts of dowry and brideprice are also used as dependent variables in the SUR model.
Secondly, for the multivariate logistic regression models23 (3) and (4), where I test
the prevalence of dowry and brideprice, two dummy variables are created as dependent
variables. For regression model (3), the dowry regression, the dependent variable takes a
value of one if a dowry was paid and zero otherwise. Similarly, for regression model (4),
the brideprice regression, the dependent variable is coded one if a brideprice was paid
and zero otherwise.
Independent Variables:
Individual characteristics of bride and groom: Two individual level characteristics of
both bride and groom are included in the regression models – age and education. Level
of education is measured in years of schooling. However, since it is common for people
in rural areas to drop out of school, attendance and completion may be different. In this
study I use completed years of schooling as the level of education. In the survey age (in
years) at the time of marriage were asked to the respondents and so it was readily
available to use in the analysis. However, I re-scaled age by centering it to the mean age
23 To parallel the SUR equations, bi-variate Probit model could have been used. Using bi-variate Probit would have given similar results as of Logistic regression, thus I decided to use Logistic regressions instead.
50 of the sample to achieve more meaningful estimates of the intercepts. According to the hypotheses (discussed in the previous chapter), a younger bride with education is more preferred in the marriage market, thus is expected to pay a lower dowry. Therefore, dowry is likely to increase with bride’s age and decline with her level of education. The opposite will be the case for brideprice – brideprice is likely to decrease with bride’s age and increase with more years of schooling.
For dowry: Bride’s age ↑ ⇒ Dowry ↑
Bride’s years of schooling ↑ ⇒ Dowry ↓
For brideprice: Bride’s age ↑ ⇒ Brideprice ↓
Bride’s years of schooling ↑ ⇒ Brideprice ↑
As opposed to ‘high quality bride’ who is expected to pay smaller dowry and bring larger brideprice, a ‘high quality groom’ is expected to receive larger dowry and pay smaller brideprice. Based on previous literature, as young and educated groom is viewed in the rural areas as a better groom, dowry is expected to increase with groom’s years of schooling and decrease with age. On the other hand, brideprice is expected to decrease with years of education and increase with groom’s age.
For dowry: Groom’s age ↑ ⇒ Dowry ↓
Groom’s years of schooling ↑ ⇒ Dowry ↑
For brideprice: Groom’s age ↑ ⇒ Brideprice ↑
Bride’s years of schooling ↑ ⇒ Brideprice ↓
51 Household characteristics: The matching household level variables are - parental landholding at the time of marriage, comparative wealth status of the groom, number of sisters, distance of marriage migration, caste affiliation and timing of marriage. The data provides information on the amount of land, in cents24, owned by parents of both bride
and groom at the time of marriage. One limitation of the data is that it does not provide
any information about household income during the marriage year. Thus, parental
landholding is the only variable that can be used to assess wealth effect. The expected
relation between dowry and parental landholding of both bride and groom is positive.
For brideprice, there is expected to be a negative correlation between brideprice and
groom’s parental landholding.
For dowry: Groom’s parental landholding ↑ ⇒ Dowry ↑
Bride’s parental landholding ↑ ⇒ Dowry ↑
For brideprice: Groom’s parental landholding ↑ ⇒ Brideprice ↓
To test the hypergamy hypothesis, a dummy variable is created based on the difference between the amount of land owned by the parents of both bride and groom at the time of marriage. This variable is coded one if groom’s parents have more land than the bride’s parents, and zero otherwise. In presence of wealth hypergamy, the coefficient of this dummy variable is expected to be positive and significant.
For dowry: Groom’s parental land > Bride’s parental land ⇒ Dowry ↑
24 1 cent = 0.01 acre
52 Number of sisters of the bride and number of sisters the groom had are included in the dowry and brideprice models respectively for reasons explained in previous chapter. The number of sisters the bride had when the marriage took place is expected to be negatively correlated with dowry. On the other hand, brideprice is expected to increase with number of sisters the groom had at the time of marriage, that is, a positive correlation between brideprice and groom’s number of sisters is expected in this case.
For dowry: Bride’s no. of sisters ↑ ⇒ Dowry ↓
For brideprice: Groom’s no. of sisters ↑ ⇒ Brideprice ↓
The survey collected data on distance, in kilometer, between respondent’s
household and his spouse’s parental household. This information is used as a measure of
marriage migration. Based on the argument25 about parents’ preference on income diversification by marrying off their offspring far from home, I expect positive correlation between marriage transaction and distance of marriage migration. This distance variable is expected to differ by region due to cultural differences between regions. In the South, village endogamy is more common and there is also much marriage between relatives. On the other hand, village exogamy is more practiced in the
North. Distance between spouses’ prenatal household reflects this cultural pattern.
However, controlling regional effects by using region dummy should take care of these cultural differences.
25 For detail see, Rosenzweig and Stark (1989)
53 For dowry: Distance between bride and groom’s household ↑ ⇒ Dowry ↑
For brideprice: Distance between bride and groom’s household ↑ ⇒ Brideprice ↑
Caste affiliation is added as a control variable as dowry and brideprice practice is expected to be highly influenced by caste background. Though both bride and groom’s caste affiliation should be added in the model, there is no inter-caste marriage found in the sample, thus including one caste variable indicating both bride and groom’s background will suffice. To control the effect of caste, I use a dummy variable, which is coded one if the couple belongs to low caste. For everyone else the variable is coded as zero.
I have inflated nominal value of marriage transaction to 1999 constant price to make transactions from different years comparable to each other. Still marriages conducted at different years can create bias due to presence of cohort specific trends in marriage transactions and matching. I have included re-scaled year of marriage as an independent variable in the regression models to take care of this potential bias. The first year in the sample - 1975, is coded as 1; the second year 1976 is coded as 2 and the rest of the years are coded accordingly. Other than controlling for unobserved time factors, this year of marriage variable is important as it indicates the direction of the practice of a specific financial exchange over time. A positive correlation with amount of dowry is expected which will imply dowry inflation over time. For brideprice, the expected relation is negative. Again, while testing for the likelihood of paying dowry, the expected relation with the dummy of dowry variable and year of marriage is positive implying
54 marriages are more likely to be associated dowry over time. Unlike dowry, the expected relation with year of marriage for brideprice is negative.
For dowry: Year of marriage ↑ ⇒ Dowry ↑
For brideprice: Year of marriage ↑ ⇒ Brideprice ↓
To test the Sanskritization hypothesis about dowry practice, I use an interaction variable by multiplying low caste dummy and year of marriage. The positive relation and
significance of this variable will indicate that over time dowry practice have expanded
among low caste groups who adopted that practice to demonstrate their adherence with
high caste culture.
For dowry: Low caste* Year of marriage↑ ⇒ Dowry ↑
Community level variables: To assess the effect of marriage squeeze, I use district
level ratio of marriageable women to marriageable men26. Figure 3A shows the distribution of age at marriage for men and women in the sample 27. In Figure 3A Panel
1, a significant portion of the bell shaped density function lies within the age range 15 to
25 years suggesting a large portion of women of the sample got married within that age
range. Similarly, Figure 3A Panel 2 depicts that a large proportion of men got married at
age 20 to 30 years. Hence, these age groups are good indicators of marriageable age for
26 There are 79 districts in 16 states. 27 The densities shown in the two panels of Figure 3.1 are Kernel densities. Kernel density estimators approximate the density f(x) from observations on x. For discussions on kernel estimators see Fox (1990).
55 the particular cohorts of the sample. Therefore, I use the ratio of women in the 15 to 25 age group to men in the 20 to 30 age group in a certain district to construct district level sex ratio. This information comes from district level census data closest to the year of marriage.
Figure 3A: Distributions of Women’s and Men’s Age at Marriage
0.14 0.1
0.12 0.08 0.1
0.06 0.08
0.06 Density Density 0.04
0.04 0.02 0.02
0 0 10 15 20 25 30 35 40 10 15 20 25 30 35 40 Women's age at marriage (year) Men's age at marriage (year)
Panel 1 Panel 2
Indian census of 1981 is used to construct district level sex ratios for all marriages conducted between the years 1975 to 1985. For marriages conducted between the years 1986 to 1995, information from the Indian census of 1991 is used. Similarly,
Indian census of 2001 is used for marriages conducted after the year 1995. For dowry, the expected relationship between sex ratio of marriageable women to marriageable men is positive and a reverse relationship is expected in the case of brideprice.
For dowry: District level sex ratio (F/M) ↑ ⇒ Dowry ↑
For brideprice: District level sex ratio (F/M) ↑ ⇒ Brideprice ↓
56 There exist high regional variation with respect to marriage and marriage related practices and norms in India, which is reflected in the individual and community level traits by region in Table 3.1. To control for regional variation in the data, two regional dummy variables are included. Dummy for Western region takes the value of one if a couple resided in that region of India, if not the dummy variable is coded as zero. The dummy for Northern region is created in a similar fashion. The base category for these two region dummy variables are the South and the East.
57 Table 3.1: Summary Statistics for All India and by Region All India North South East West Variable Mean (s.e.) Mean (s.e.) Mean (s.e.) Mean Mean (s.e.) (s.e.) Groom’s family paid brideprice+ 0.18 0.16 0.35 0.01 0.18 (0.30) (0.36) (0.48) (0.11) (0.39) Bride’s family paid dowry 0.90 0.90 0.96 0.78 0.94 (0.30) (0.30) (0.19) (0.41) (0.23) Amount paid as brideprice, if 26,948 16,801 27,901 10,883 103,284 paid (constant 1999 rupee)* (53,230) (52,806) (34,601) (13,495) (102,758) Amount paid as dowry, if paid 39,035 20,931 57,191 41,829 118,238 (constant 1999 rupee)* (70,679) (35,496) (80,905) (68,073) (149,227) Amount paid as net-dowry++ 15,741 16,270 14,832 15,543 2,329 (constant 1999 rupee)* (35,622) (32,696) (39,636) (36,374) (83,967) Groom’s age at marriage* 23 21 26 27 22 (6) (6) (6) (7) (4) Bride’s age at marriage* 18 17 19 20 18 (5) (5) (4) (5) (3) Groom continued school for at 0.73 0.69 0.70 0.80 0.90 least more than a yr (0.45) (0.46) (0.46) (0.40) (0.31) Bride continued school for at 0.45 0.31 0.59 0.64 0.77 least more than a yr (0.50) (0.46) (0.49) (0.48) (0.42) Groom’s schooling (year)* 5 5 5 6 5 (4) (4) (4) (4) (3) Bride’s schooling (year)* 3 2 4 4 3 (3) (2) (3) (3) (2) Groom’s parents owns land 0.77 0.76 0.76 0.75 0.84 (0.42) (0.42) (0.43) (0.43) (0.37) Bride’s parents owns land 0.73 0.75 0.67 0.71 0.85 (0.44) (0.43) (0.47) (0.45) (0.36) Groom’s parents’ landholding at 701 746 590 485 1,310 the time of marriage (cent)* (1,061) (1,022) (789) (1,211) (1,594) Bride’s parents’ landholding at 553 613 356 285 1,507 the time of marriage (cent)* (892) (969) (556) (356) (1,401) Year of marriage* 1985 1984 1986 1986 1984 (6) (6) (6) (6) (6) Distance of marriage migration 33 37 22 22 67 (km)* (72) (85) (39) (44) (96) High caste 0.36 0.40 0.16 0.45 0.57 Middle caste 0.21 0.27 0.11 0.20 0.15 Low caste 0.43 0.33 0.73 0.35 28 District marriageable sex ratio 106 102 112 111 105 (F/M) (12) (11) (9) (12) (13) Region: East 0.11 West 0.06 North 0.55 South 0.24 N 2154 1178 520 330 126 + The missing cases are considered as zero. ++ Includes only the cases where both dowry and brideprice information are available. * The values are rounded off to the nearest integer.
Chapter 4: Prevalence of Marriage Transactions: Trends,
Changes and Factors Affecting the Practices
58 59 4.1 Introduction
It is important to understand the underlying factors that determine the likelihood of paying a dowry or brideprice at marriage to broaden our knowledge of marriage transaction. In this chapter, I first explore the regional prevalence rate of the two practices followed by a discussion on changes in the prevalence rates in the twenty-five years from 1975 to 1999. Then I analyze the factors that affect the probability of being involved in exchanging marriage transactions between families.
There exists a large body of literature on the determinants of dowry phenomenon, but surprisingly very little or almost no research has been done on brideprice in the Indian context. Most of the works have treated dowry either as a one- sided or a net transaction from bride’s family to that of the groom. Dowry, viewed as a unidirectional flow of cash or kind from bride’s family to the groom, ignores the fact that brideprice can be paid at the same time for the same marriage. In this chapter I take dowry, brideprice and net-dowry, which is constructed using both dowry and brideprice information, into account, in determining the factors affecting the practices. I use logistic regressions to do the analyses.
4.2 Prevalence of Dowry and Brideprice
The Prevalence of dowry for a certain year is measured as a ratio of marriages in that year in which a dowry was paid over the total number of marriages conducted in the same time period. Prevalence of brideprice is measured in the same fashion. If dowry or
60 brideprice is denoted by x, then the prevalence of x, PRx, is measured in the following
way:
PRx = (Number of marriages with x )/( Total marriages).
Dowry is a highly prevalent practice in the sample cases – a dowry was paid in
eighty-nine percent of all marriages. The prevalence of dowry from 1975 to 1999 for
rural areas of all four regions of India and the country as a whole is shown in Figure 4A.
The prevalence of practices by year presented in Figure 4A can be a little under
estimated as the missing cases of dowry are coded as zero28. However, it is important to
note that the prevalence of marriage transactions shown in the figure comes from un-
weighted sample cases and therefore, suggestive but not true representative of either the
whole country or the regions. From the bi-variate presentation, it is evident that dowry
practice has become more widespread in twenty-four years in all the regions except for the East. The practice is almost universal in the Southern and Western regions of the sample. Prevalence of dowry is the lowest in the East, which is still very high. Last five years weighted average shows seventy-eight percent of all marriages in the East are associated with dowry.
Unlike dowry, brideprice is much less practiced – a brideprice was paid in only thirty percent of all marriages of the sample. We have to be cautious while stating the prevalence of brideprice as brideprice information is missing for a significant number of
28 There are 38 missing cases. I have coded these as zero dowry because these cases are not significantly different from the cases where no dowry was paid by any individual or household level characteristics. For details of the comparison, see, appendix A.
61 marriages and are excluded from the sample29. Thus, for brideprice, I use a smaller
sample of 1329 cases for the analysis. The missing cases are significantly different from
non-missing positive brideprice cases and also from zero brideprice cases regarding individual and household level characteristics. Even though these missing cases are
significantly different, these are less different from the cases where some brideprice were
paid than none. Therefore, excluding those cases might have created some negative bias
in the measured prevalence of brideprice. However, it is also very much likely that a
majority of the missing information cases are random. Thus, the expected negative bias
should not be severe.
Trends in the practice of brideprice are shown in Figure 4B and the sharp
regional differences are evident immediately. Table 4.1 also highlights the differences at
regional level – the practice is universal in the South and in the West, and is almost non-
existent in the East, where only two percent of all marriages paid brideprice. According
to last five years of marriage data in the sample, none of the marriages in the East were
associated with brideprice. In the North, on average, a brideprice was paid in twenty
percent of all marriages. These prevalence of brideprice are suggestive and has the
potential to be over estimated as the missing brideprice cases, which are more likely to
be zero than any other value, are excluded from the sample in this case.
In most of the cases when brideprice is paid in a marriage, a dowry is also paid;
specially, in the Western and Southern region, marriage transaction flows in both
29 If we consider the missing brideprice cases as zero brideprice then marriages involving brideprice becomes 18 percent.
62 directions. Where as, in the East and the North, there exist high imbalance in the prevalence of dowry and brideprice practice. In the smaller sample (1329 cases) fifty- eight percent of all marriages in the East are associated with dowry and only two percent from the same sample paid brideprice. Northern region also exhibit similar trend with eighty-eight percent of all marriages paying dowry in contrast to nineteen percent paying brideprice.
4.3 Change in Prevalence Over Time
The goal of this section is to demonstrate expansion of dowry practice in the sample cases of India over twenty-four years while holding other factors that might affect the practice, constant. Graphical presentation of prevalence of brideprice does not signal much change in the practice over time. However, bi-variate relation may not capture subtle changes. Therefore, in this section, I further explore the direction of changes in dowry and brideprice practice over time holding other variables constant.
Table 4.2 provides logistic regression results for the likelihood of paying a dowry at marriage. Effect of time on the likelihood of paying a dowry or the prevalence of dowry, is captured by the coefficient of “year of marriage” variable. Any coefficient larger than one would imply a positive correlation between the associated variable and the odds ratio in favor of paying a dowry. As is evident from Table 4.2, the odds in favor of paying a dowry at marriage has increased over time and this expansion of dowry practice over time has occurred among the low caste group. But there has not been any
63 statistically significant change in the practice of brideprice, showing no support for the anthropological argument of declining brideprice practice. The logistic regression results for the odds of paying a brideprice are shown in Table 4.3.
The graph of the prevalence of dowry and brideprice over time, presented in the previous section, also showcase the change in prevalence of dowry and brideprice in twenty-four years from 1975 to 1999 in the sample cases. The limitation of this bivariate relation between marriage transactions and time is that the relation might have been affected by other trends internalized in the practices. Thus, to observe a complete picture, we need to control the effects of other factors influencing the prevalence of the practices. Logistic regression provides us with the odds ratio of paying a marriage transaction while holding other factors constant. The time trend of predicted odds ratio of paying a dowry is graphed in Figure 4C. Even after controlling for other important factors, we find that the regional variation in the odds of paying dowry holds. The odds of paying a dowry have inclined in all the regions, Northern region exhibiting the most moderate incline. As mentioned before, time is not statistically significant to influence the odds of paying a brideprice. The predicted odds ratio of paying a brideprice is graphed in Figure 4D.
One of the hypotheses about dowry expansion is sanskritization, which argues that low caste follows the high caste practices to achieve higher status in the caste and status stratified Indian society. The effect of sanskritization is captured by an interaction between two variables – low caste and year of marriage. According to sanskritization
64 argument, low caste adopt the practice of dowry but not brideprice since dowry is viewed as a high caste phenomenon contrary to brideprice, which is more common among low caste population. If sanskritization is the underlying reason or one of the reasons for dowry expansion then the interaction variable is expected to be positive for dowry practice but not for brideprice. Interestingly, when sanskritization is controlled
(Model 3, Table 4.2), not only the interaction variable appears to be positive and significant (though marginally, at ten percent level), but also at the same time the year of marriage variable becomes insignificant. This indicates that the increase in dowry prevalence is due to the expansion of the practice among the low caste population over time. The predicted odds ratio of paying a dowry over time by low caste and high caste groups are presented in Figure 4E, which shows a very intriguing trend. Not only that the likelihood of paying a dowry is at the rise among the low caste, at the same time the odds are declining among the high caste group, thus broadening the gap with respect to dowry practice between these two groups.
Nonetheless, the increase in the participation in dowry practice among the low castes is not enough to prove sanskritization is responsible for this expansion. I included the same interaction variable in the logistic regression for brideprice to confirm the argument and found that over time the participation in brideprice practice have also increased among low castes population of the sample (Table 4.3, Model 2). Once included, the odds of paying a brideprice increases among low caste population but actually declines in general. Both dowry and brideprice practice have increased among
65 low caste population and therefore, sanskritization is less likely to be the reason behind the expansion30.
Possible explanation behind this dowry and brideprice expansion could be the overall escalation of household income among the low caste. To test this, I have used an interaction between low caste, year of marriage and groom’s status of landholding (not shown in the table). The estimate of this interaction variable is statistically insignificant.
But this does not rule out the possibility of income rise among the low caste as the explanation for dowry expansion among them. This is because landholding is a good proxy for wealth; however, it is not a perfect proxy for income especially if income is generated from sources other than land, which is very likely to be the case with low caste groups. But it is beyond the capacity of this dissertation to test that further since household income for the sample population at the time of marriage is not available.
4.4 Who are More likely to Pay Dowry and Brideprice
The previous section explores and attempts to explain the direction of changes of financial transactions over time. In this section, I analyze the individual, household and community level factors that influence the likelihood of being involved in financial transaction at marriage. Logistic regression model is used to estimate the likelihood of
30 One reader has raised a concern that sanskritization might be occurring with the rise in brideprice practice to compensate poor households for income loss. My response is that, sanskritization refers to the following of upper castes behavior and rejecting those of the lower castes. Thus, expansion of both dowry and brideprice is not consistent with the idea of sanskritization. Besides, given the higher status the groom’s family enjoys in marriage negotiations, it is unlikely that groom’s family will offer to pay brideprice to compensate the income loss of bride’s family.
66 paying or receiving a marriage transaction. As mentioned before, the results for dowry and brideprice are presented in Table 4.3 and Table 4.4 respectively.
4.4a Factors Affecting the Probability of Paying or Receiving a Dowry:
The only individual characteristic that is important in affecting the likelihood of paying a dowry is bride’s education. The odds of paying a dowry is higher for parents of educated brides. In this analysis, a bride who attended school for more than one completed year is considered as educated. For educated brides average years of schooling is about 6 years.
The educated brides, who have continued schooling for more than one year, are one and half times more likely to pay dowry compared to those who did not attend school for more than a year. This positive relation between bride’s education status and odds of paying a dowry goes against the hypothesis that a bride with more education is preferred in the marriage market and thus is less likely to pay a dowry. One explanation for this could be that dowry is paid as a bequest for educated brides. It is possible that parents, who educate their daughters, are also more likely to pay dowry as a bequest at marriage.
Bride’s parental landholding status is used as a proxy for bride’s parental income and wealth. Landholding though is a good proxy for wealth; it definitely is not a perfect one.
Thus, bride’s education may display some of the wealth effect that is not captured by landholding and in that case will be over estimated.
I have not found any statistically significant effect of either bride’s age at marriage or groom’s characteristics on the probability of paying or receiving a dowry at marriage. Both bride and groom’s parental landholding status, which is a proxy for their
67 wealth, are not statistically significant either. The insignificant relation between bride’s parental landholding and odds of paying dowry is at odds with the argument that wealthier parents are more likely to pay dowry. If dowry is paid as a bequest then it is expected that wealthier parents will pay it. Thus, respective insignificant relation suggests at least for wealthier parents dowry is not a bequest.
There is a significant effect of marriage migration on the incidence of dowry payment. A dummy variable is used to capture the distance between the two households.
The median distance between bride and groom’s parental household in the sample is twenty kilometers. I use twenty kilometer as the cut-off point to create the distance dummy. Distance more than twenty kilometers significantly increases the probability of paying or receiving dowry by forty percent. This supports the argument that households prefer to diversify their sources of support by marrying off their offspring to a distant place to avoid similar types of income shocks. The underlying assumption is that households in distant places do not experience similar income shocks. This is specifically true in rural areas where most households are primarily agricultural and rainfall variation in different areas imposes different income shocks (mostly opposite) to households in different areas. Because households linked by marriages of their offspring belong to the same informal support network by cultural norms, it is advantageous to parents of daughters to marry them off to distant areas to diversify income shocks (Rosenzweig and
Stark, 1989). Thus, parents of the bride are more likely to pay dowry to attract a groom from a distant place for establishing informal kin networks.
68 Anthropological literature on marriage and caste argues that dowry is practiced more among high caste population. There has been some empirical support for this claim. Although dowry is found to be larger in size among highest caste members, it is sometimes found to be smaller among medium caste members compared to the lowest caste. The sample that I am using for analysis does not support that expectation. In this sample, the lowest caste marriages are three fourth times more likely to be associated with dowry compared to middle and high caste marriages. In this analysis, I have divided the sample between high and low caste only and so these results are not directly comparable to those previous studies that use a more detailed categorization of caste.
As expected, marriage squeeze, which means the existence of more marriageable females compared to marriageable males, increases the probability of paying a dowry.
For one percent increase in the sex ratio, the probability of paying dowry increases by only four percent.
After controlling for other factors affecting dowry prevalence, I found that marriages conducted in the West and the North are significantly more likely to be associated with dowry practice compared to Southern and Eastern region. Western and
Northern marriages are two and half and one and half times more likely to pay dowry at marriage respectively. The pioneer research done by Dyson and Moore (1983) highlights the dichotomy between Northern and Southern region of India with respect to kinship structure, marriage, female autonomy and demographic behavior. The existence of significant regional variations even after controlling for factors that influence the
69 probability of dowry payment suggests that the regional diversity is much deeply rooted and the variables included in the analysis are not sufficient to explain these effects away.
4.4b Factors Affecting the Probability of Paying or Receiving a Brideprice:
Individual characteristics of bride and groom influence the likelihood of paying or receiving brideprice in the expected directions. Parents of an older bride are half times less likely to receive brideprice from the groom or his family and this effect is highly statistically significant. This indicates that older brides are less desirable in the marriage market, at least to those who have the potential to pay brideprice.
Older grooms are almost twice as much to pay brideprice compared to the younger ones. In the sample population, the median age at marriage for men is twenty- two years. Going beyond that age might reflect a man’s poor level of personal, familial or financial ability in the marriage market. Thus, it is very much likely that he compensates for that negative reputation by paying a brideprice. On the other hand, an educated groom, who is a very desirable candidate in the marriage market, does not need to pay brideprice to get married. An educated groom is one-third times less likely to pay brideprice at marriage.
Household characteristics include bride and groom’s parental landholding, their caste background, timing of marriage and marriage migration. None of these matching household level variables have any statistically significant effect on the likelihood of paying a brideprice.
70 Availability of more women at marriageable age compared to marriageable men increases the likelihood of paying or receiving a brideprice. The size of this effect is very small but highly statistically significant. This goes against the marriage squeeze argument if brideprice is paid as a price rather than as a marriage ritual. If brideprice is not paid as a price of a bride then it is very likely that this practice would be more prevalent in communities where gender discrimination is comparatively lower. Such a community will have higher female to male ratio as gender based negligence and gender biased mortality in the community will be lower as well. In the sample of brideprice, about fifty percent of marriages that paid brideprice are conducted in the South. Southern region is historically known for having more gender egalitarian societies compared to the rest of the country and thus have a lower level of imbalance in sex ratio, which means the sex ratio at birth is closer to biological ratio of male to female in the Southern region compared to the rest of the country31. The dominance of marriages taking place in the
Southern region with more females compared to males could be the reason for having
marriage squeeze as a significant variable in the regression result of paying a brideprice.
As found in dowry practice, the regional variation in bridepice practice also holds
even after controlling for individual, household and other community level variables. A
marriage in the East and North are ninety-nine percent and ninety percent less likely to
be associated with brideprice compared to the Western and the Southern region.
31See for example, Rao (1993a, 1993b), Dalmia (2004), Dasgupta and Mukherjee (2003).
71 4.4c When Dowry Exceeds Brideprice:
Dominance of one type of marriage transaction has the potential to create sex- based discrimination in the society and can be viewed as a burden in a community.
Clearly, in this sample population of rural India, dowry is a much more common practice compared to brideprice. However, for one-fourth of all marriages, for which I have dowry and brideprice information, marriage transactions went in both directions. To strengthen our understanding of financial exchanges at marriage in the Indian context and to bring dowry and brideprice in the same picture, I have also analyzed the factors that affect the probability of having a positive net-dowry, which means having a larger dowry compare to brideprice. Out of 1,329 marriages, in 1,039 cases, which is about seventy-eight percent of this sample, the size of dowry is larger than brideprice.
Table 4.4 presents the results from a logistic regression where the dependent variable is one if the size of dowry is larger than brideprice and zero otherwise. Dowry is paid in larger amounts compared to brideprice in the North and the West, and these associations are statistically significant at the one percent and five percent level respectively. District level sex ratio of females to males is also highly significant confirming, once again, the demographic claim that as men become relatively scarce, females will compete for them by paying higher dowry. The opposite is true for brideprice. Although I did not find evidence in the previous section on the proposed effect of sex ratio on brideprice, the results here show that the availability of more females in the marriage market results in larger dowry compared to brideprice.
72 Table 4.4 also shows that the distance of marriage migration is more likely to result in a larger dowry to be paid and the effect is statistically significant at the one percent level. As we discussed before, this confirms the argument that households are willing to pay up front for establishing informal kin networks to diversify their income risks. Why the bride’s household is paying more (dowry) than the groom’s household
(brideprice) at the time of establishing this link (through marriage) may be cultural. There could be cultural norms regarding how these informal kin networks operate regarding their financial transactions.
Individual characteristics like groom’s age have a negative effect on the likelihood of a larger dowry being paid, but the effect is weakly significant at the ten percent level.
Both landholding of the bride’s household and the groom’s household have positive effect on the likelihood of a higher dowry being paid, but the former is only significant at the ten percent level while the later at the five percent level.
The fact that variables like sex ratio and distance of marriage migration are highly significant in this regression rather than individual and household characteristics imply that demographic and economic reasons are more likely to result in a relatively larger dowry. This is to say that there is less to worry about sex-based discrimination based on individual or household characteristics.
73 4.5 Summary
There is substantial evidence that dowry, which is the dominant practice between the two types of marriage transaction, is becoming more prevalent over time, but we do not see much change in brideprice practice within the twenty-four years time period.
Among the low caste marriages, both dowry and brideprice have become more prevalent over time. Nonetheless, the idea of Sanskritization fails to explain the dowry expansion.
There is evidence that marriage squeeze has significant role in explaining dowry practice.
We see that there exists strong regional variation in the practice of marriage transaction showcasing diverse cultural heritage surrounding marriages in India. Both practices are highly prevalent in the South and the West. In the North and the East, dowry is the central practice. In many cases both dowries and brideprices are paid simultaneously. But in all four regions, the amount of dowry exceeds that of brideprice on average.
Individual and community level factors are more important in affecting the probability of paying either a dowry or a brideprice. On the other hand, economic and community level variables are more significant for the probability of paying a larger dowry than brideprice. In the next chapter, I explore the magnitude of dowry and brideprice. That will be followed by a detailed analysis of determinants of dowry and brideprice focusing on the size of the transactions.
74 Table 4.1: Regional Prevalence of Dowry and Brideprice32 Region Total Paid % Paid Paid % Paid Paid % Paid marriages dowry dowry brideprice brideprice both both East 161 93 58 4 2 3 2 West 23 18 78 23 100 18 78 North 961 848 88 184 19 166 17 South 184 165 90 184 100 165 90 All India 1329 1124 85 395 30 352 26
32 Table includes only those cases for which both dowry and brideprice information are available. Therefore, dowry prevalence rates do not reflect that of the full sample.
75 Table 4.2: Probability of Paying a Dowry: Logistic Estimates Variables Model 1 Model 2 Model 3 (Basic) (Hypergamy) (Sanskritization) Co-ef Odds Co-ef Odds Co-ef Odds (s.e) Ratio (s.e) Ratio (s.e) Ratio Matching bride & groom characteristics Bride’s age at marriage (in years, -0.04 0.96 -0.04 0.96 -0.04 0.97 centered at the mean) (0.02) (0.96) (0.02) Groom’s age at marriage (in years, -0.01 0.99 -.01 .99 -0.01 0.99 centered at the mean) (0.02) (0.02) (0.02) Bride attended school for more 0.38 1.47* 0.38 1.47* 0.38 1.46* than 1 year (0.18) (0.18) (0.19)
Groom attended school for more 0.07 1.07 0.07 1.07 0.09 1.09 than 1 year (0.18) (0.18) (0.18) Matching h/h characteristics
Bride's parents has land 0.22 1.25 0.22 1.24 0.21 1.23 (0.19) (0.20) (0.19) Groom's parents has land 0.12 1.13 0.13 1.14 0.12 1.13 (0.20) (0.25) (0.20) Distance of marriage migration is 0.33 1.39* 0.33 1.39* 0.34 1.40* more than 20 km (median dist.) (0.16) (0.16) (0.16) Low caste 0.60 1.82*** 0.60 1.81*** 0.12 1.13 (0.16) (0.16) (0.31) Year of marriage 0.03 1.03* 0.03 1.03* 0.01 1.01 (0.01) (0.01) (0.02) Groom’s parents wealthier than 0.01 0.99 bride’s (Hypergamy) (0.19)
Low caste*Year of marriage 0.05 1.05+ (Sanskritization) (0.03)
Community level variables District marriageable sex ratio 0.04 1.04*** 0.04 1.04*** 0.04 1.04*** (F/M) (0.01) (0.01) (0.01) West 1.01 2.75* 1.01 2.74* 0.98 2.66* (0.44) (0.44) (0.44) North 0.64 1.90* 0.64 1.90* 0.62 1.87* (0.20) (0.20) (0.20) Intercept -3.71 -3.71 -3.5 (0.92) (0.92) (0.93) Total 2154 2154 2154 *** p<.001, ** p<.01, * p<.05, + p<.1
76 Table 4.3: Probability of Paying a Brideprice: Logistic Estimates Variables Model 1 Model 2 (Basic) (Sanskritization) Co-ef Odds Ratio Co-ef Odds (s.e) (s.e) Ratio Matching bride & groom characteristics Bride’s age at marriage (in years, -0.09 0.92*** -0.08 0.92*** centered at the mean) (0.02) (0.03) Groom’s age at marriage (in years, 0.06 1.06*** 0.06 1.06*** centered at the mean) (0.02) (0.02) Bride attended school for more than 1 yr -0.002 0.98 0.01 0.99 (0.18) (0.18) Groom attended school for more than 1 yr -0.40 0.67* -0.39 0.68* (0.17) (0.17) Matching h/h characteristics Bride's parents has land 0.21 1.23 0.20 1.22 (0.20) (0.20) Groom's parents has land -0.33 0.72 -0.32 0.73 (0.20) (0.20) Distance of marriage migration is more 0.07 0.94 -0.05 0.95 than 20 km (median dist.) (0.15) (0.15) Low caste (0.12) 1.13 -0.65 0.52 * (0.15) (0.32) Year of marriage 0.01 1.01 -0.02 0.98 (0.01) (0.02) 0.07 1.08** Low caste*Yr of marriage (Sanskritization) (0.02)
Community level variables 0.02 1.02*** 0.02 1.02*** District marriageable sex ratio (F/M) (0.01) (0.01)
Region: East -4.24 0.01*** -4.21 .01*** (0.55) (0.55) North -2.25 0.10*** -2.29 0.10*** (0.20) (0.20) Intercept -1.65 -1.34 (0.55) (0.89) Total 1329 1329 *** p<.001, ** p<.01, * p<.05, + p<.1
77 Table 4.4: Probability of Paying a Positive Net-dowry: Logistic Estimates33 Variables Co-efficient s.e Odds Ratio
Matching bride & groom characteristics Bride’s age at marriage (in yrs, centered at the mean) -0.05 0.02 0.95* Groom’s age at marriage (in yrs, centered at the mean) -0.01 0.02 0.99 Bride attended school for more than 1 yr 0.20 0.18 1.22 Groom attended school for more than 1 yr 0.05 0.17 1.06
Matching h/h characteristics Bride's parents has land 0.29 0.18 1.34 Groom's parents has land 0.40 0.19 1.49* Distance of marriage migration is more than 20 km 0.39 0.15 1.47** (median dist.) Low caste 0.19 0.15 1.21 Year of marriage 0.02 0.01 1.02+
Community level variables District marriageable sex ratio (F/M) 0.02 0.01 1.01* Region: West -1.10 0.46 0.33* North 1.08 0.18 2.95*** Intercept -2.21 0.84
Total 1329 *** p<.001, ** p<.01, * p<.05, + p<.1
33 Sanskritization and hypergamy do not have any significant effect on net-dowry. Thus, the results for those analyses are not shown in the table.
78
Figure 4A: Regional Prevalence of Dowry Over Time34
1.05
1
0.95
0.9
0.85
0.8
0.75 Prevalence of Dowry 0.7 East West South North 0.65 India 0.6 1975 1980 1985 1990 1995 2000 Yr of Marriage
Figure 4B: Regional Prevalence of Brideprice Over Time
1
0.8
India 0.6 East North 0.4 South and West
Prevalence of Brideprice 0.2
0 1975 1980 1985 1990 1995 2000 Yr of Marriage
34 Trends shown in Figure 4A and Figure 4B are not net of other effects.
79
Figure 4C: Predicted Odds Ratio of Paying Dowry Over Time by Region
1
0.98
0.96
0.94
0.92
0.9
0.88 North South Predicted Odds Ratio 0.86 East West 0.84 India
0.82 1975 1980 1985 1990 1995 2000 Yr of Marriage
Figure 4D: Predicted Odds Ratio of Paying Brideprice Over Time by Region
0.9
0.8
0.7
0.6 India North 0.5 East South West 0.4
0.3
Predicted Odds Ratio 0.2
0.1
0 1975 1980 1985 1990 1995 2000 Yr of Marriage
80
Figure 4E: Predicted Odds Ratio of Paying Dowry Over Time by Caste Affiliation
0.98
0.96
0.94
0.92
0.9
0.88
High Caste Predicted Odds Ratio 0.86 Low caste 0.84
0.82 1975 1980 1985 1990 1995 2000 Yr of Marriage
Chapter 5: Factors Affecting Dowry and Brideprice
81 82 5.1 Introduction
In chapter 4, I found that dowry practice has expanded over time and this expansion has taken place among the lower caste. However, there has not been any significant change in the prevalence of brideprice over time. In addition to that, I analyzed the factors that influence the likelihood of being involved in a financial exchange. To extend our knowledge beyond the dichotomous scenario, in this chapter I focus on the size of dowry and brideprice and how the size of these two payments have changed in twenty-five years from 1975 to 1999. Here, I assess the relationship between individual, household and community level variables with the amount paid as dowry and brideprice.
5.2 Magnitude of Dowry and Brideprice
On average, each dowry is equivalent to 35,194 rupees in 1999 prices (Table
5.1)35. Regional variation in the mean and median value of dowry paid at the time of
marriage in the sample is shown in Table 5.1. Average dowry is more than three times
higher in the West (about 111,670 rupees) compared to the sample national average of
about 35,193 rupees. On the other hand, average dowry is almost half of the sample
national average in the North (about 18,888 rupees). Median dowry, however, is much
35 The average dowry in this table is different from that of Table 3.1. The average dowry presented in Table 3.1 is calculated using only those cases where dowry is paid. On the other hand, the average dowry presented in Table 5.1 includes all the cases in the sample.
83 lower than the average in all four regions and the country as a whole36. This indicates
that a small group of people pays a very high amount of dowry driving the average
upward. Similar results were found by Dalmia (2004).
Regional variation in the mean value of brideprice paid at the time of marriage
for the sample cases is shown in Table 5.237. The mean value of brideprice is much
smaller than that of dowry, 8,616 rupees in 1999 prices. Like dowry, average brideprice is
also the highest in the West, almost 12 times higher than the national average (about
103,000 rupees). The East has the lowest average brideprice, one-third of the national
average (about 281 rupees). In the East and the North, median brideprice is zero
suggesting paying a brideprice is not a common practice in those regions. In the West
and South, average brideprice is also much higher than the median amount, meaning that
the average is possibly driven up by a small group of people. The mean brideprice is also
very low compared to dowry in all four regions. Once controlled for other factors
affecting marriage transaction, the regional variation in the magnitude of dowry and
brideprice still holds – compared to the South and the East, the West experiencing larger
dowry and brideprice (Table 5.4), and the North experiencing smaller dowry (Table 5.5).
From the prevailing trends in dowry and brideprice discussed in the previous
chapter we have seen a contrasting culture of marriage transaction in India across region.
The regional mean and median dowry and brideprice presented in Table 5.1 and Table
36 The national amounts are suggestive but should be read with caution as the sample cases are un- weighted. Also for dowry, the amounts can be underestimated since the missing dowries are coded as zero. 37Like dowry the national amounts for brideprice are also suggestive. The average shown in Table 5.2 are different from that of Table 3.1 for similar reason as of dowry.
84 5.2 provides additional support to that finding. Considering dowry and brideprice together, we have seen that transfers flow from both directions in the South and the
West, although magnitudes of these two types of transfers can be very unequal. In the
North and the East, dowry is the more common and dominant practice. Net-dowry, which is dowry net of brideprice for the same marriage, will be helpful to further dissect the flow of marriage transaction for better understanding of the dynamics. The mean and median net-dowry is presented in Table 5.3.
Positive net-dowry reflects the dominance of dowry practice over brideprice with respect to the size of marriage transaction. Net-dowry is measured for all the cases for
which both dowry and brideprice information is available. For example, if in a marriage,
both dowry and brideprice are paid then dowry net of brideprice is considered as net- dowry. But, if only dowry is paid and the amount of brideprice is zero, in that case, dowry is considered as net-dowry. Similarly, with positive brideprice and zero dowry, the negative value of brideprice is taken as net-dowry.
Since net-dowry consists of dowry and brideprice, the regional level of net-dowry is different from both dowry and brideprice. In all four regions, the average net-dowry is positive and large in size. Average net-dowry is the largest in the North, where both average dowry and brideprice are comparatively smaller in size. This reflects higher
imbalance in financial flow between families of bride and groom in that region. Despite
larger size of both dowry and bridepirce, West exhibit the lowest average net-dowry
reflecting a balance between dowry and brideprice paid in that region. Moreover, median
85 net-dowry in the West is negative suggesting for at least fifty percent of all marriages, the size of brideprice exceeds that of dowry. Once controlled for other factors, the West still exhibits significantly lower net-dowry than the South and the East (Table 5.10, column 3).
5.3 Change in the Magnitude of Marriage Transactions Over Time
The most consistent but somewhat surprising finding is the decline in the size of dowry, despite the expansion of the practice. Ordinary least square regression is used to explore the direction of change in the magnitude of dowry and brideprice over time holding other variables constant. The OLS estimates for determinants of dowry and brideprice are presented in Table 5.4 and Table 5.5. Among other variables, I have included rescaled “year of marriage” as an explanatory variable. A positive coefficient for the year of marriage variable will mean increase in the size of marriage transaction over time. On the other hand, a negative coefficient will mean a decline in size. Table 5.4, the table of determinants of dowry presents results from five regression models. Model 1 is the basic model and model 2 and 3 are additional models for testing the marriage hypergamy and sanskritization hypotheses. Model 4 and 5 include two interaction terms - between bride’s years of schooling and year of marriage and between North region dummy and year of marriage, respectively. Results from all five models show that the magnitude of dowry has declined significantly (p-value<0.001) over the sample period38.
This finding goes against a large body of literature arguing in favor of the inflation of
38 I explore the real size of dowry and brideprice not their proportion with respect to household income. Thus, the stated deflation of dowry over time corresponds to its size rather than its proportion to household income.
86 dowry in the Indian marriage market (Rao 1993a; Anderson, 2003; Edlund, 2001; Billig,
1992; Upadhya 1990; Deolalikar and Rao 1990; Paul, 1985; Srinivas 1984; Lindenbaum,
1981; Epstein 1973)39. Results from OLS regression on determinants of brideprice (Table
5.5) show that the size of brideprice has also declined over time.
In the previous chapter, we have seen that dowry has expanded among the lower
caste. One of the anticipations about the decline in dowry size could be that these new
groups, among which dowry practice has expanded, are paying smaller dowry and as a
result the average size of dowry is declining over time. Insignificance of the dummy
variable for low caste and the interaction between low caste and year of marriage imply
that the size of dowry at either one point of time or over time does not vary significantly
by caste affiliation. Thus, dowry expansion among the low caste is not responsible for
dowry deflation.
Interestingly, the size of dowry has declined with bride’s level of education over
time. The significance of the interaction term constructed using bride’s level of
education and year of marriage provides evidence supporting that. In general, the size of
dowry increases with each year of additional schooling for the bride. Nevertheless, this
increase with level of education has been declining over the years. Although the decline
in dowry size with bride’s years of schooling cannot completely explain away the overall
decline of dowry size over time, it definitely explains part of it.
39 For articles arguing for decline in real dowry, see Dalmia, 2004.
87 Again, in the previous chapter we have seen that there exists high regional variation in the prevalence of dowry and brideprice. It is also likely that the decline in magnitude of dowry and brideprice are likely to vary among different regions of India.
The size of dowry is much smaller in the North compared to the base regions and in general the size of dowry is declining over time. The interaction term between Northern region and year of marriage in Model 5 of Table 5.4 compares the direction of dowry trend of the North with the rest of the regions. The positive and significant estimate of the interaction term suggests that compared to the rest of the country the decline in the size of dowry is much slower in the North. Unlike North, in the West the size of dowry has declined at a much higher rate than the rest of the regions over time.
Like dowry, brideprice has also declined in all the regions except for North. To capture time trajectory of dowry magnitude at regional level, I use two interaction variables - one created by multiplying Northern region dummy and year of marriage and the other by multiplying Western region dummy and year of marriage. The estimates of these interaction terms are presented in Model 2, Table 5.5. In twenty-five years, the size of brideprice has declined at a significantly higher rate in the West compared to the rest of the regions. On the other hand, there has been inflation with respect to the size of brideprice in the North relative to other regions, though later we will see that within the
Northern region there has not been any significant change in the size of brideprice over time.
88 In Figure 5A and Figure 5B the trends in predicted values of dowry and brideprice obtained from regression results presented in Table 5.4 (Model 5) and Table
5.5 (Model 2) are plotted40. It is evident that in twenty-five years the size of both dowry
and brideprice has declined significantly not only at the country level but also within
each region. The only exception to that is the size of brideprice in the North where
brideprice is not a common practice. The Northern region shows the most moderate
level of change with respect to the size of marriage transactions. On the other hand, the changes are the most dramatic in the West.
5.4 Determinants of the Size of Dowry and Brideprice
To determine the desirable characteristics those are important in assessing the amount of marriage transaction from one family to the other, ordinary least square
(OLS) and seemingly unrelated regression (SUR) models are used. The rationales for using both of these models are explained in detail in Chapter three.
5.4.i Factors Affecting Dowry:
The results from ordinary least square regression model are presented in Table
5.4. The second column shows the results from the basic model (Model 1) of dowry
determinants. Third and fourth columns show results from additional regressions for
testing hypergamy and sanskritization hypotheses.
40 The lines in the graphs are trend line drawn through the predicted values of dowry and brideprice over time.
89 Individual characteristics of bride and groom:
Findings show that there exists a significant positive correlation between dowry and bride’s age after controlling for other factors affecting dowry. Parents of older brides pay a larger amount of dowry, for each additional year of their daughters’ age they pay about 1,221 rupees more as dowry. This is quite a large premium given that the median value of dowry is about 12,000 rupees.
There are two factors that might explain the age affect. First, because parents generally arrange marriages and sons are responsible for supporting parents in their old ages, groom’s parents might prefer a daughter-in-law who will conform to their expectations and are likely to be more controllable. Younger brides are more attractive in the marriage market because they are comparatively more vulnerable and are easily malleable. Parents of older brides may therefore have to pay larger dowries to compete with younger brides, ceteris paribus. Second, social norm in most rural areas dictates marrying off daughters at young ages and larger dowry paid by older brides may just be a premium for deviating from social norm.
Unobserved bridal characteristics, however, can bias the estimated age effect. It is possible that certain unobserved traits of brides (e.g. beauty) results in a shorter stay in the marriage market for some brides, because those traits are more desirable to the grooms. Larger dowries paid by older brides in that case could partially be a compensation for their shortage in those desirable traits. In such a case, age effect on dowry will be overestimated. The association is still positive nonetheless.
90 Unlike the positive correlation between bride’s age and dowry, groom’s age is negatively correlated with dowry, and the correlation is significant (at 5 percent level).
For each additional year of groom’s age, bride’s parents pay about 688 rupees less for dowry. This negative correlation indicates that older grooms are less preferred in the marriage market. The average age at marriage for men in the sample is 23. Since marriage is almost universal for both men and women in India, men marrying later than the average age provide a negative signal about their ability, which can be financial, familial or personal. Therefore, older grooms fail to attract or be in a position to demand larger dowries.
One surprising but not unusual result from Table 5.4 is that bride’s level of education has a significantly positive correlation with the amount paid by the bride’s family. This result is consistent with the finding presented in Chapter 4, that brides who attended school for more than a year is almost one and half times more likely to pay any dowry. For each additional year of schooling, the value of dowry increases by 4,639 rupees. Other studies have also found similar results41. Education increases the ability to
generate more income given the availability of formal sector job opportunities. Thus,
educated brides are supposed to be more attractive in the marriage market. But, lack of
formal employment opportunities that requires schooling in rural areas or the prevailing
cultural norm supporting the idea that wives are responsible for home management not
income generation can put negative pressure on the demand for educated brides.
Besides, educated women have their own opinion and ideas, which might contradict with
41 For example, see Dasgupta and Mukherjee 2003, and Dalmia 2004.
91 groom’s parents’ interests because of the reasons we explained above. Thus, despite their higher earning potentials, groom’s parents may not find educated brides to be attractive for their son. In that case, dowry will not be smaller for educated brides as expected
(Dasgupta and Mukherjee, 2003).
The problem with the above mentioned line of argument is that in a dynamic environment, parents will not have incentives to educate their daughters and that will result in a decline or stagnation in girls’ education level, which is not the case in rural
India. Average years of schooling by birth cohort of brides of this sample are graphed in
Figure 5C, which shows an increase in years of school attendance over time. Therefore, there has to be some other explanations.
Table 5.6 breaks down marriage market characteristics of brides and grooms by bride’s educational status to assess the effect of bride’s level of education on dowry in further detail. The mean value of dowry for educated brides is almost three fold of the mean value of dowry for uneducated brides, and the difference is statistically significant.
This indicates that these two groups might be very distinct and therefore, might face different marriage market choices. Educated brides are married to grooms who have on average more than 4 years of schooling than grooms of uneducated brides, showing strong positive assortative matching on education. Educated brides also come from wealthier households in terms of landholding and are also married to wealthier grooms on average. This means that wealthier households are more likely to educate their children more and that they are more likely to arrange marriages among themselves. In
92 that case, marriage transfer dynamics within this wealthy class of people could be very different and a larger dowry can be paid for status and/or as bequest from parents to their daughters.
I run separate OLS regressions for educated and uneducated brides. Estimates of these two regression models are presented in Table 5.7; Model A includes only the uneducated brides and model B includes only the brides who have completed more than one year of schooling. The results show that for these two groups of women, underlying mechanisms determining the value of dowry are different.
For the educated group (Model B), each year of schooling of the bride significantly increases the level of dowry. After controlling for other individual, household and community level variables, I found that for each year increase in schooling, dowry increases by 6,721 rupees. The level of education is the only bridal characteristic that matters in determining dowry for those educated brides. Neither the bride nor the groom’s age or even the groom’s level of education has any significant effect. This indicates that for educated brides, dowry could be a bequest from their parents. Parents who are likely to provide bequest for their daughters are also more likely to educate their daughters. Again, bride’s education could be an imperfect proxy of wealth, which may not be completely captured by landholding. In that case, we can say that wealthier parents are more likely to provide dowry as bequest42.
42 One reader made a comment that it could be the case that parents must pay for educated grooms and usually, women marry more educated grooms than themselves. Thus, if girls are highly educated, parents must pay more to get even more educated grooms. My response in this regard is that, if
93 Unlike the educated brides, for the uneducated group (Model A), bride’s age significantly increases dowry. Dowry is negatively related with groom’s age but the relation is positive with groom’s education, suggesting uneducated bride’s parents pay less if the groom is older but they pay more for educated grooms. Thus, for uneducated brides, dowry is less likely to be bequest, rather it is paid either to attract better quality groom or because dowry is demanded by the groom’s family as a condition of marriage.
Like bride’s level of education, groom’s years of schooling is also positively correlated with dowry and this relationship is highly statistically significant at p<0.01
(Table 5.4). One additional year of groom’s schooling is associate with 1,310 rupees
more in dowry at the time of marriage. An educated groom is more attractive in the
marriage market compared to his uneducated counterpart for his better capacity to earn a
regular and higher income, higher level of social standing and accumulation of social
capital. So, on the one hand, bride’s family will be willing to pay larger dowry to marry
off their daughter to an educated groom, and on the other hand, groom’s family will
have more bargaining power to demand higher dowry. In either case, a positive
correlation is expected. It is however not possible to conclude which one of the above
two effects dominates from the regression I present here.
To examine the effect of groom’s education level on dowry, I have also divided
the grooms into two groups based on their literacy status. Table 5.8 presents the mean
parents are forced to pay more dowry for their educated daughters, then they will not have any incentive to educate their daughters to begin with. And we would have sees a decline in female education over time, which is not the case in rural India. Thus, parents who decide to educate their daughters are also more likely to pay dowry voluntarily.
94 and median dowry for these two different groups as well as for all marriages. Like educated brides, educated grooms are also associated with about three times of the dowry compared to their uneducated counterpart displaying clear distinction between these two groups. I ran separate OLS regressions to identify which characteristics of bride and groom are associated with higher dowry for these two different sets of grooms.
The results of the regressions are shown in Table 5.9. Model A corresponds to uneducated grooms and Model B includes only the educated grooms.
For the educated grooms, both the bride and groom’s attributes play significant role in determining the value of dowry. Dowry increases both with bride’s age and level of education. Dowry increases with groom’s years of education but declines with age.
Unlike their educated counterpart, for uneducated grooms none of the individual characteristics of the bride or the groom has any significant effect on the determination of dowry except for bride’s age, which is marginally significant. Thus, for this group, dowry does not significantly vary much by personal attributes of its agents. These findings indicate the existence of multiple marriage markets and interesting implication of dowry and its nature.
Bride’s parents pay larger dowry either to attract young educated grooms, whom I am referring to as ‘high quality grooms’ or they pay more because the high quality grooms have higher bargaining power to demand larger dowries. As there is a positive association between bride and groom’s level of education43 and dowry increases with
43 Table 5.11 shows positive association between bride and groom’s education in detail.
95 bride’s education level, larger dowry can be paid as a result of both providing bequest and a method to attract better quality groom or meeting the demand of the groom’s family.
Household characteristics:
There is no significant correlation between dowry and parental landholding of the bride ceteris paribus at the time of marriage (see Table 5.4). This finding indicates that dowry is not a pre-mortem bequest or wealth effect in general as is argued by some authors (e.g. Edlund, 1997). Like education, wealth also increases a groom’s quality. On one hand, a better groom has higher bargaining power to demand larger dowry; and on the other hand, bride’s family is more likely to pay a higher dowry to marry off their daughter to a better groom. As a result, grooms with parental landholding will be associated with higher dowry. There is positive and significant correlation between dowry and landholding of groom’s parents, although the magnitude of the relationship is low. The significance of groom’s parental landholding is consistent with the above mentioned two possible scenarios: price and demand argument.
Given the findings, it is thus more likely that dowry in my sample, in general, is more of a price paid for grooms rather than a bequest or status except for the educated brides. Whether this price is settled by market forces or is demanded by these higher quality grooms exploiting their better bargaining position remains inconclusive.
96 There is a statistically significant positive correlation between distance of marriage migration and amount of dowry. This provides support for Rosenzweig and
Starks’ (1989) argument that parents are willing to pay larger dowry if they can diversify income risks through informal credit provided by their in-laws living in distant areas characterized by different income risks. Recall that I explained this argument in details in the previous chapter, where I presented the finding that the likelihood of paying a dowry also increases with the distance of marriage migration.
The number of sisters a bride has does not have any significant effect on the amount paid as dowry. The final household characteristic that I examine is caste and low caste dummy is included in the regression model to measure caste effect. Without controlling for household wealth, low caste marriages are significantly correlated with smaller dowry (not shown in the table). Nonetheless, households from higher castes are usually wealthier than those from lower castes. In effect, once I control for wealth, as measured by landholding, caste does not have any significant effect on dowry.
In addition to the base model, I have run two models to test the hypergamy and sanskritization hypotheses. The results of these two additional models are presented in column three and column four of Table 5.4. I have used a dummy variable to measure hypergamy, which takes the value one if groom’s family is wealthier than bride’s family and zero otherwise. Wealth is measured by the amount of landholding. To measure the effect of sanskritization, I have included an interaction variable by multiplying year of marriage with low caste dummy. I have not found any evidence in favor of either
97 marriage hypergamy or sanskritization. Whether groom’s family is wealthier than the bride’s family or not does not have any significant effect on the amount paid as dowry.
Similarly, there is no evidence of dowry inflation among the lower caste. Thus, the effect of sanskritization is irrelevant here.
Community Level Effects:
One of the major explanations provided in the demographic literature to explain the existence and inflation of dowry is the excess supply of marriageable women compared to men in the marriage market. This is generally referred to as ‘marriage squeeze’ (Caldwell at. el. 1983; Rao 1993a, 1993b; Bhat and Halli 1999; Billig 1992) and has gone under much scrutiny in the empirical literature. The result of the analysis provides strong support for marriage squeeze argument. I found that dowry is significantly larger if there are more women at marriageable age in a district than the number of marriageable men in the same district.
The effect of household and community level factors on the magnitude of dowry provides support to the price argument of dowry motive. Parents of a bride pay larger dowry if the groom comes from a wealthier background or reside farther away from the bride’s natal household. Dowry is also larger when there are more women at the district level marriage market. Dowry in this situation is more likely to be paid as a price to attract limited potential grooms in the market by competing potential brides.
98 5.4.ii Factors Affecting Brideprice
Individual characteristics of bride and groom:
The only individual characteristic that has any significant effect on the amount paid as brideprice is bride’s education. Brideprice is positively associated with bride’s years of schooling. Groom’s family pays about 1,004 rupees more with each additional year of bride’s education. There is no significant effect of either bride’s age or parental landholding or even groom’s age or education on brideprice.
Household and matching characteristics:
There is a significant positive correlation between groom’s number of sisters and brideprice. Brideprice increases by 2,190 rupees for each additional sister that the groom has. It provides support to the hypothesis that a groom with more sisters has more resources available that is obtained from his sisters’ marriages to pay brideprice at his own marriage. Bride or groom’s parental landholdings, distance of marriage migration and caste are not significant variables that affect brideprice. The significance of year of marriage and its negative correlation indicates that over time the real value of brideprice has declined.
Community level effect:
According to the marriage squeeze hypothesis, with high sex ratio of marriageable women to men (F/M), it is the dowry that is expected to be affected positively not brideprice. And as expected, the result does not display any significant
99 effect of sex ratio on the amount paid as brideprice. But like dowry, we see significant regional variation in case of brideprice too. These regional variation holds even after controlling for individual, household and other community level variables. The magnitude of brideprice is significantly larger in the West compared to the base regions of the South and the East.
5.4.iii Considering Dowry and Brideprice Together:
There are two ways to analyze dowry and brideprice together: first, dowry and brideprice can be viewed as different institutions but not independent of each other; secondly, these two types of marriage transactions can be viewed as part of the same institution where they are exactly reciprocal of each other. According to this definition, dowry is the amount paid by bride’s family net of payments made by groom’s family.
This definition of dowry is generally used by economists44.
I did two different analyses to investigate dowry and brideprice together and
found similar results. Results for these two analyses are presented in Table 5.10. The
second and third columns contain the estimates for dowry and brideprice where they are
viewed as independent institutions but not independent of each other by using a seemingly unrelated regression (SUR) model. The fourth column presents the estimates of net-dowry obtained from using ordinary least square regression model. Details of both
SUR and OLS regression models are discussed in Chapter Three.
44For example see, Rao 1993(a) 1993(b), Dalmia (2004), Dasgupta, and Mukherjee,(2003).
100 SUR results show that as expected the error terms of the two regression models with outcome variables brideprice and dowry are significantly correlated at 0.2154 level of correlation. This suggests that the error terms in the dowry and the brideprice regressions are not totally independent of each other and there are unobserved factors that simultaneously affect both the amount of dowry and the amount of brideprice.
Thus, the size of brideprice increases with any increase in the size of dowry and vice versa indicating marriages are generally conducted within social classes and marriage transactions are indicators of social status. However, this does not rule the price motive of marriage transactions.
Dowry estimates from SUR model (2nd column of Table 5.10) and net-dowry
estimates from OLS model (4th column of Table 5.10) are similar to the OLS dowry estimates (Table 5.4) discussed in the previous section. The direction of correlations between dependent and independent variables are the same, though the magnitudes are different in many cases. In case of dowry, age of bride and groom lost their significance in both SUR model and OLS of net-dowry model45. Nonetheless, similar results of the
net-dowry regression and the dowry regression suggest dowry is the dominant marriage
transaction in terms of magnitude and influence of the practices. In the next section, I
focus on the regional differences in determinants and nature of dowry and brideprice.
45 Dowry sample is different from the sample of brideprice and net-dowry. For detail about differences in sample size and missing cases, see appendix A.
101 5.4.iv Regional Variation in Determinants of the Size of Dowry and Brideprice:
Previous literature highlights the existing disparity in marriage practices in the
North and the South. I have also found interesting distinctions between these two regions regarding the factors influencing the magnitude of financial transactions at marriage reflecting divergent motives behind marriage transactions, especially dowry practice in those regions. Table 5.12 and Table 5.13 display the OLS estimates of the determinants of dowry and brideprice size for the two contrasting regions – the North and the South, respectively.
Factors that reflect the price motive are more influential in the Northern cases, whereas, factors that are more likely to represent dowry as a status symbol are significant in the Southern cases. Bride’s age, distance of marriage migration, number of females available in the marriage market - are the variables that reflect price component and have significant effect in determining dowry size in the North. On the other hand, Bride’s parental landholding, the only matching household characteristic reflecting status or bequest is significant in the South. Bride and groom’s education that has significant effect in both of the regions may reflect a combination of both price and status motive.
In general, in the South there are more women available in the marriage market than men at district level compared to the North46. This outcome can be a result of
relatively higher status that women possess in the South compared to the North.
Egalitarian attitude and higher status of women in the South, which is reflected by the
46 See Table 3.1 for a comparative description between the Northern and Southern cases of the sample.
102 higher sex ratio of women to men at marriageable age, implies lower gender biased negligence and mortality rate at early stage of life47. Interestingly, even though there are
more women in the marriage market in the South, the insignificance of sex ratio in
determining the size of dowry shows that dowry is not paid as a result of female competition in that region. This is not the case with the Northern region.
Like dowry, the practice is brideprice is also regionally heterogeneous. Interesting difference is that in the South educated grooms pay significantly larger brideprice. Again,
in the South, brideprice is larger if there are more women than men at marriageable age.
Thus, evidently there are different underlying motives behind the practice of marriage
transactions at regional level, which can be very important for policy implications.
Marriage transactions are more of a reflection of status or bequest in the South. On the
contrary financial transactions are made more as a price either to attract a better quality bride or groom, or to satisfy the demand of groom’s family to marry off the daughter in
the North.
5.5 Summary
This chapter analyzes dowry and brideprice as independent institutions and also
sheds light on how and whether these two institutions interact with each other. Thus,
this chapter provides a more comprehensive picture of the determinants of marriage
47 It can also indicate more rural-urban migration in the South
103 transaction and fills this gap in the literature. The findings hold regardless of how these two practices are defined or related to each other.
Bride and groom’s age, groom’s education, groom’s wealth status measured by landholding, distance of marriage migration are the individual and matching characteristics that significantly influence the size of dowry in the expected directions.
One puzzling finding is that bride’s years of schooling positively increases the amount of dowry even after holding everything else constant. I argue that this is not because there is a demand for uneducated brides in the rural marriage market, rather because for educated brides, a portion of dowry might be paid as bequest from her natal family. For uneducated brides, bride’s age and groom’s quality significantly affect the amount paid which is not the case with an educated bride. However, for educated grooms both bride and groom’s age and education play significant role in determining the value of dowry.
But for uneducated grooms, none of the individual characteristics of either the bride or groom matters. For brideprice, bride’s education and groom’s number of sisters are the only two individual or matching level factors that have any significant effect on the size of the payment.
I have not found any evidence of wealth hypergamy or sanskritization influencing dowry practice. In Chapter Four, we have seen that the prevalence of dowry has expanded among the lower caste but sanskritization fails to explain that expansion. In this chapter, I have found that the size of dowry has deflated over time. Thus, the explanation of sanskrtization is not only insignificant also irrelevant. On the other hand,
104 marriage squeeze is highly significant in affecting both the probability of paying a dowry and its magnitude. I also find strong evidence that the size of real value of dowry has declined over time at both regional and country level as oppose to popular belief and this is also the case with brideprice.
Indirect evidence implies the existence of heterogeneous dowry motives at regional level. I find that the price and bequest or status motives are not mutually exclusive rather can co-exist together. Dowry acts more as a tool of market clearing price in the Northern region, while in the Southern region dowry and brideprice are paid more as a symbol of family status or to provide bequest to the bride.
105 Table 5.1: Mean and Median Dowry by Region Region N Mean Dowry Median Dowry All 2154 35,193 12,254 East 330 32,830 10,634 West 126 111,669 69,966 North 1178 18,887 6,367 South 520 55,101 28,719 Note: All mean and median dowries are in 1999 Rupees
Table 5.2: Mean and Median Brideprice by Region Region N Mean Brideprice Median Brideprice All 1329 8,616 0 East 161 270 0 West 23 103,284 72,379 North 961 3,216 0 South 184 27,901 17,476 Note: All mean and median brideprices are in 1999 Rupees
Table 5.3: Mean and Median Net-dowry by Region Region N Mean Net-dowry Median Net-dowry All India 1329 15,741 5,790 East 161 15,542 4,101 West 23 2,328 -200 North 961 16,270 5,668 South 184 14,831 8,650 Note: All mean and median net-dowries are in 1999 Rupees
106 Table 5.4: OLS Estimates for the Size of Dowry Variables Model 1 Model 2 Model 3 Model 4 Model 5 (Basic) (Hypergamy) (Sanskritization) Matching bride & groom characteristics 1221** 1233** 1253** 1189** 1008* Bride's age (455) (455) (456) (455) (454) (in yrs, centered at mean) Groom's age -688* -701* -703* -662+ -632+ (in yrs, centered at mean) (349) (350) (350) (350) (347) Bride's year of schooling 4639*** 4628*** 4629*** 6577*** 4705*** (624) (625) (625) (1056) (621) Groom's year of schooling 1310** 1309** 1311** 1248** 1224** (454) (455) (455) (455) (452) Matching household charac- teristics at the time of marriage Bride's parental landholding 0.87 1.46 0.92 0.91 1.09 (cent)48 (1.67) (1.77) (1.67) (1.67) (1.66) Groom's parental landholding 5.69*** 4.94** 5.69*** 5.67*** 5.66*** (cent) (1.37) (1.56) (1.37) (1.37) (1.37) Bride's number of sisters -1388 -1383 -1376 -1532 -1533 (948) (948) (948) (950) (943) Distance of marriage 42* 42* 42* 44* 45* migration (km) (18) (18) (18) (18) (18) Low caste -1856 -1981 -8053 -1816 -2278 (2712) (2715) (5323) (2709) (2697) Year of marriage -1274*** -1266*** -1528*** -759* -2418*** (228) (229) (296) (321) (317) Groom’s parents wealthier 3028 than bride’s (Hypergamy) (3012) Low caste*Year of marriage 579 (Sanskritization) (428) Bride’s yrs of education* Year -156* of marriage (69) 2230*** North*Year of marriage (431)
Community level variables District marriageable sex ratio 755*** 756*** 751*** 756*** 790*** (F/M) (120) (120) (120) (120) (119) Region: West1 60740*** 61131*** 60507*** 60685*** 58462*** (6306) (6318) (6308) (6300) (6284) -17601*** North -17481*** -17738*** -17043*** -41665*** (3481) (3483) (3482) (3487) (5798) Intercept -45188** -46606** -41995** -51137** -35201* (14893) (14959) (15075) (15106) (14929) Total 2154 2154 2154 2154 2154 ***p<0.001, **p<.01, *p<.05, +p<.10 1 The estimate of West*Year of marriage (not shown) is -8460 (989), which is significant at p <0.001.
48 1 cent = .01 acre
107 Table 5.5: OLS Estimates for the Size of Brideprice Model 1 Model 2 Variables Co-eff (s.e.) Co-eff. (s.e.) Matching bride & groom characteristics Bride's age at marriage (in yrs, centered at mean) 339 (261) 276 (261) Groom's age at marriage (in yrs, centered at mean) 83 (200) 86 (199) Bride's year of schooling 1004** (397) 954* (395) Groom's year of schooling 24 (262) 37 (261)
Matching household characteristics Bride's parental landholding at marriage (cent) 0.09 (0.96) 0.05 (0.95) Groom's parental landholding at marriage (cent) -1.01 (1.00) -0.84 (0.99) Groom's number of sisters 2191*** (526) 2201*** (524) Distance of marriage migration (km) 3.53 (9.71) 3.70 (9.67) Low caste 1753 (1651) 1349 (1648) Year of marriage -347* (140) -846** (248)
Community level variables District marriageable sex ratio (F/M) 141+ (73) 153* (73) Region: West 88,996*** (6186) 111,840*** (13391) North -8,976 (2128) -17,712*** (3937) West*Year of marriage -2,654* (1278) North*Year of marriage 764** (289)
Intercept -5360 (9010) -535 (9297) Total 1329 1329 *** p<.001, ** p<.01, *<.05, +p<.1
108 Table 5.6: Bride and Groom Characteristics by Bride’s Educational Status Uneducated Bride Educated Bride Mean Dowry (Rs) 18,792 55,027 Median Dowry (Rs) 7,003 26,781 Groom’s years of schooling 3 7.43 Groom’s age (year) 22 24 Groom’s parental landholding (cent) 624 784 Bride’s years of schooling <2 5.62 Bride’s age (year) 17.25 18.80 Bride’s parental landholding (cent) 523 582 Note: Mean and median dowries are in 1999 Rupees. The difference between mean dowries by bride’s education status is significant at p> |t| = 0.000.
Table 5.7: OLS Estimates of the Determinants of Dowry Size by Bride’s Education Status Model A Model B (Uneducated Bride) (Educated Bride) Variables Co-eff (s.e.) Co-eff (s.e.)
Matching bride & groom characteristics Bride's age (in yrs, centered at mean) 1016** (369) 1138 (953) Groom's age (in yrs, centered at mean) -616* (291) -908 (700) Bride's year of schooling 6721*** (1257) Groom's year of schooling 1589*** (382) 808 (905) Matching household characteristics at the time of marriage Bride's parental landholding (cent) -1.48 (1.50) 1.80 (3.22) Groom's parental landholding (cent) 10.17*** (1.58) 4.12+ (2.14) Bride's number of sisters -903 (782) -2468 (1920) Distance of marriage migration (km) 28+ (15) 64+ (35) Low caste 3126 (2280) -9684+ (5358) Year of marriage -698** (200) -1853*** (432) Community level variables District marriageable sex ratio (F/M) 205+ (108) 1321*** (219) Region: West 58750*** (7565) 59495*** (10553) North -18165*** (2968) -20687** (6866) Intercept 7272 (13290) -101694*** (28257)
N 1179 975
*** p<.0001, ** p<.01, *<.05, + p<.1
109 Table 5.8: Mean and Median Dowry by Education Status of Groom Uneducated Groom Educated Groom Mean Dowry (Rs) 15,041 42,902 Median Dowry (Rs) 6,897 16,069 Bride’s schooling (year) 1.11 3.78 Bride’s age (year) 17 18 Bride’s parental landholding (cent) 386.7 616.57 Groom’s years of schooling (year) <1 6.82 Groom’s age (year) 22.2 23.5 Groom’s parental landholding (cent) 509.896 774.47 Note: Mean and median dowries are in 1999 Rupees, The difference between mean dowries by groom’s education status is significant at p> |t| = 0.000
Table 5.9: OLS Estimates of the Determinants of Dowry Size by Groom’s Education Status Model A Model A Uneducated Grooms Educated Grooms
Variables Co-eff (s.e.) Co-eff
Matching bride & groom characteristics Bride's age (in yrs, centered at mean) 702+ (369) 1360* (604) Groom's age (in yrs, centered at mean) -385 (267) -854+ (474) Bride's year of schooling 1449 (1197) 4614*** (746)
Groom's year of schooling + 1278 (665) Matching household characteristics at the time of marriage Bride's parental landholding (cent) 1.37 (2.21) 0.55 (2.03) Groom's parental landholding (cent) 3.58+ (1.96) 5.78** (1.65) Bride's number of sisters -909 (758) -1761 (1264) Distance of marriage migration (km) 33** (10) 55 (932) Low caste 5610** (2094) -4735 (3656) Year of marriage -850** (180) -1448*** (306) Community level variables District marriageable sex ratio (F/M) 33 (103) 929*** (157) Region: West 11729 (7193) 64487*** (7947) North -15351*** (2567) -18894*** (4809) Intercept 25507* (12531) -58578** (19975)
N 596 1558 *** p<.001, ** p<.01, *<.05, + p<.1
110 Table 5.10: Estimates of Factors Affecting the Size of Dowry and Brideprice Using SUR and Net-dowry Using OLS Variables SUR SUR estimates of OLS estimates estimates of Brideprice (s.e.) of Net-dowry Dowry (s.e.) (s.e.) Matching bride & groom characteristics Bride's age (in yrs, centered at mean) 343 332 26 (366) (260) (319) Groom's age (in yrs, centered at mean) -65 87 -160 (280) (199) (245) Bride's year of schooling 2280*** 1000** 1288** (555) (395) (484) Groom's year of schooling 1473*** 39 1409*** (366) (261) (320) Matching household characteristics at the time of marriage Bride's parental landholding (cent) -0.65 0.11 -0.82 (1.34) (0.95) (1.17) Groom's parental landholding (cent) 3.60** -0.92 4.38*** (1.39) (0.99) (1.22) Bride's number of sisters 211 355 (682) (700) Groom’s number of sisters 1738*** -990 (444) (642) Distance of marriage migration (km) 35** 4 31** (14) (10) (12) Low caste -1114 1731 -2827 (2312) (1643) (2018) Year of marriage -682*** -340* -356* (195) (139) (171) Community level variables District marriageable sex ratio (F/M) 453*** 137 322*** (102) (73) (89) Region: West 71901*** 88872*** -16738* (8654) (6153) (7552) North -7660** -8987*** 1344 (2977) (2117) (2598) Intercept -27768* -4338 -25355* (12548) (8942) (11054) Total 1329 1329 1329
Correlation matrix of residuals from SUR: Brideprice Dowry Brideprice 1 Dowry 0.2154 1
Breusch-Pagan test of independence: chi2(1) = 99.914, Pr = 0.0000 *** p<.0001, ** p<.01, *<.05
111 Table 5.11: Assortative Matching of Bride and Groom by Education Bride: 1 or less yr of Bride: More than 1 yr of Total schooling schooling Groom: 1 or less yr 561 35 596 of schooling (47.58%) (3.59%) Groom: More than 618 940 1558 1 yr of schooling (52.42 %) (96.41 %) Total 1179 975 2154 (100 %) (100 %) t= 26.0395 P > |t| = 0.000
Table 5.12: OLS Estimates of the Size of Dowry for the North and the South Region: North Region: South Variables Co-eff (s.e.) Co-eff (s.e.) Matching bride & groom characteristics Bride's age (in yrs, centered at mean) 825* (352) 139 (1181) Groom's age (in yrs, centered at mean) -105 (299) -472 (773) Bride's year of schooling 2576*** (478) 6945*** (1470) Groom's year of schooling 1396*** (304) 2206+ (1209)
Matching household characteristics at marriage Bride's parental landholding (cent) -1.15 (1.09) 21.47*** (6.51) Groom's parental landholding (cent) 0.78 (1.07) 2.47 (4.74) Bride's number of sisters 127 (675) -2886 (2502) Distance of marriage migration (km) 30** (11) 126 (84) Low caste -3796+ (2006) -4117 (7152) Year of marriage -453** (173) -1915*** (567)
Community level variables District marriageable sex ratio (F/M) 349*** (86) 390 (407)
Intercept -23919* (9639) -6230 (48973)
Total 1178 520 ***p<0.001, **p<.01, *p<.05, +p<.10
112 Table 5.13: OLS Estimates of the Size of Brideprice for North and South Region: North Region: South
Variables Co-eff (s.e.) Co-eff. (s.e.)
Matching bride & groom characteristics Bride's age at marriage (in yrs, centered at mean) 193 (278) -991 (876) Groom's age at marriage (in yrs, centered at mean) 289 (236) 539 (516) Bride's year of schooling 798* (394) 2220+ (1201) Groom's year of schooling -12 (249) 2117* (877)
Matching household characteristics at marriage Bride's parental landholding (cent) -0.79 (0.86) 1.52 (3.82) Groom's parental landholding (cent) -2.20* (0.94) -5.76+ (3.11) Groom's number of sisters 2422*** (511) -1446 (1558) Distance of marriage migration (km) -8 (9) 49 (96) Low caste -2212 (1667) 5344 (5681) Year of marriage -182 (141) -1101** (417)
Community level variables District marriageable sex ratio (F/M) 12 (70) 669* (275)
Intercept -405 (7894) -52828 (33875)
Total 961 184 *** p<.001, ** p<.01, *<.05, +p<.1
113 Figure 5A: Trend Line of Predicted Value of Dowry Over Time
120000
100000
80000 West South Eas t India 60000 North
40000 Predicted Dowry (Rp)
20000
0 1975 1980 1985 1990 1995 2000 Yr of Marriage
Figure 5B: Trend Line of Predicted Value of Brideprice Over Time
140000
120000 West South Eas t India 100000 North
80000
60000
40000 Predicted Brideprice (Rp) 20000
0 1975 1980 1985 1990 1995 2000 Yr of Marriage
114 Figure 5C: Bride’s Average Years of Schooling Over Time
9
8
7
6
5
4
3
2 Average Yrs of Schooling 1
0 1930 1940 1950 1960 1970 1980 1990 Year of Birth
Chapter 6. Discussion and Conclusion
115 116 6.1 Introduction
One of the key objectives of development research is to broaden our understandings on pressing social issues that threaten the welfare of individuals in a society. One such pressing issue is the marriage transaction in India. Marriage transaction is a burning topic in Indian context because of its potential adverse effect on women and on union formation in general. Marriage transaction can be studied from a multidimensional perspective. In this dissertation I examined the institution of dowry and brideprice, their prevalence in the community and determinants that influence practices and their magnitudes. This study is guided by a series of research questions –
Are the prevalence of dowry and its magnitude inflating over time in India? What is the course of brideprice over time with respect to prevalence and size? Who are more likely to either pay or receive dowry? Similarly, which factors affect the probability of paying or receiving brideprice? What individual, household, and community level factors affect the size of dowry and brideprice? What role does regional heterogeneity in terms of women’s status and marriage rituals and norms play in determining the underlying role of these practices? The importance of these questions lie on broadening our sociological knowledge about the thriving practice of dowry at current times in a society when the practice has become obsolete in many parts of the world where dowry used to be commonly practiced. Besides, from policy perspective, these questions and an attempt to answer those provide us with valuable knowledge much needed to take effective measures to control the adverse effects of dowry in the society.
117 In the literature, there exist mainly two measures of dowry – real dowry and real dowry net of brideprice, which is also known as net-dowry. Net-dowry views dowry and brideprice as opposite to each other. My analyses show that the underlying dynamics and determinants are different for dowry and brideprice suggesting that these two are separate institutions. Therefore, dowry and brideprice should be analyzed separately.
However, using net-dowry has its benefits too as it not only depicts the direction of the flow of marriage transactions but also highlights the existing imbalance in financial exchanges between families involved. Previous studies on dowry have used either real dowry or net-dowry as a measure of analysis. I have used both measures to analyze dowry and brideprice and thus have provided a more complete picture of marriage transaction in rural India.
6.2 Summary of Findings
I use multivariate regression analyses to explore and analyze the dynamics of dowry and brideprice empirically. Logistic regressions are used to explore the direction of dowry and brideprice expansion over time and to examine the factors that affect the probability of paying a dowry, brideprice and a positive net-dowry as well. Ordinary
Least Square regressions are used to investigate not only the change in the magnitude of dowry, brieprice and net-dowry over time but also the factors that influence the size of these practices. To assess the inter-dependence of dowry and brideprice practice, I have used Seemingly Unrelated regression model. The findings from these analyses are presented in figure 6.1 along with the corresponding hypotheses.
118 As claimed by anthropological studies, I found evidence that dowry practice has become more prevalent over time although the real value of dowry has declined over the sample period of twenty-five years from 1975 to 1999. Similarly, the real value of brideprice has also declined over time though there has not been any significant change on the prevalence rate. The value of net-dowry has also declined with time suggesting a larger decline in dowry over brideprice.
Even though dowry practice has expanded over time and the expansion occurred among the low caste groups, I have not found any evidence supporting sanskritization as expected. Recall that dowry originally used to be a high caste phenomenon and sanskritization refers to low castes internalizing the high caste custom as a mean of acquiring higher social status. Interestingly, brideprice practice has also become more prevalent among the lower castes though there has not been any significant change on the prevalence rate of brideprice in general in the sample population. I also find no support to marriage hypergamy, women marrying higher status men, with respect to wealth measured by landholding.
Previous empirical studies show mixed results about marriage squeeze theory in explaining dowry inflation. In general, I have found a positive association between a higher female-to-male ratio and higher dowry and net-dowry, which is what is predicted by the marriage squeeze hypothesis. But, this dissertation has not found any evidence on dowry inflation. Rather there has been a decline in the real value of dowry at both country and regional level as mentioned before. The resolution in these apparent
119 contradictory findings is in the dynamic nature of the marriage market – only a positive change in the sex ratio rather that the amount of sex ratio may explain an inflation of dowry (Foster, Protik and Khan, 2007). Female-to-male ratio has no effect on the amount paid as brideprice as expected.
Strong support is found in favor of income diversification theory for probability of paying and magnitude of dowry and net-dowry but not for brideprice practice. These findings suggest parents of brides are more inclined to marry off their daughters to grooms of distant places to diversify the geographic and climatic income shocks that the households face. But the insignificant relationship between brideprice and marriage migration possibly suggest no such preference is present on part of grooms or their parents.
One interesting finding is that parents of educated brides are more likely to pay dowry and the size of dowry increases with bride’s level of education. At first it might seem that there is a negative demand for educated brides in the rural marriage market.
But, with the support of findings from some additional analyses, I establish that parent’s of educated bride do not pay larger dowry to compensate for their daughter’s lack of attractiveness in the marriage market rather they pay at least a part of dowry as bequest.
Parents who educate their daughters are also more likely to pay bequest at their daughter’s marriage. For uneducated brides, dowry is more likely to be paid as a price.
120 The role of marriage transactions in the Southern region is very distinct than that of the North. Both dowry and brideprice are paid as symbols of family status in the
South. Nevertheless, this does not completely rule out the possibility of price component of dowry for marriages in the South. On the other hand, marriage transactions in the
North follow the price theory more sincerely.
Table 6.1: Support for key hypotheses
Hypothesis Theory Finding A Prevalence Anthropological observation A.1 Prevalence of dowry is increasing over time Supported A.2 The expansion of dowry practice over time is Sanskritization Dowry has higher among the lower caste. expanded among the lower caste, but there is not enough evidence to support sanskritization. A.3 Prevalence of brideprice is decreasing over No support time B Magnitude Anthropological observation B.1 The size of dowry is increasing over time No support. Found strong evidence that the size decreased over 24 years. B.2 Dowry inflation is higher among the lower Sanskritization No support caste. B.3 The size of brideprice is decreasing over time Supported C Factors affecting dowry and brideprice
C.1i The likelihood of paying a dowry is higher if Marriage squeeze Dowry: Supported there are fewer men than women in the Brideprice: No marriage market. support The opposite is the case with brideprice. C.1ii The size of dowry is larger if there are fewer Marriage squeeze Dowry: Supported men than women in the marriage market. Brideprice: No The opposite is the case with brideprice. support C.2i The likelihood of paying a dowry is higher if Hypergamy No support the groom’s family is wealthier than the bride’s family.
121 C.2ii Dowry is larger if the groom’s family is Hypergamy No support wealthier than the bride’s family. C.3i Grooms with higher education receive larger Female competition Dowry: Supported dowry and pay smaller brideprice. (quality of groom) Brideprice: No support C.3ii Grooms with more parental landholding Female competition Dowry: Supported receive larger dowry and pay smaller (quality of groom) Brideprice: No brideprice. support C.4.i Brides with higher education pays smaller Female competition Dowry: No support dowry and receives larger brideprice. (quality of bride) Found strong evidence that educated brides pay larger dowry. Brideprice: Supported C.4.ii Older brides pay larger dowry and receive Female competition Dowry: Supported smaller brideprice. (quality of bride) Brideprice: No support C.5 Brides from wealthier families pay larger Bequest or Family No support dowry. status C.6.i Brides with more sisters pay smaller dowry. Resource constraint No support (replacement) C.6.ii Grooms with more sisters pay larger Resource constraint Supported brideprice. (replacement) C.7 The size of dowry and brideprice are larger Income Dowry: Supported the farther away the bride and groom’s pre- diversification Brideprice: No marriage locations are. support
6.3 Limitations
At the macro level, there are three main indicators of dowry that showcase changes in the practice over time. These are expansion, inflation and burden of dowry on the household. According to Rao (1993a), an average dowry in India can amount to over two-thirds of a household’s assets, or about six times a household’s annual income. This is a huge burden on poor households and it is important to know whether this burden has worsened over time from a social welfare point of view. Since the size of dowry has
122 declined over time, it is very likely that dowry has become less of a burden, on average, over time, given that households have not experienced a decline in real income over time. However, due to lack of data on household income at the time of marriage, I could not determine the level of and change in the financial burden that households face to pay dowries. This remains as a limitation of this dissertation.
Empirical analysis of marriage transaction is very much likely to be affected by unobserved characteristics that are not included in the regression models. The analyses I have done have the potential to be affected by three different levels of unobservable characteristics – community, household and individual. It is possible that some community level traits such as village norms can systematically affect the estimates that I have measured. Similarly there can be household level characteristics that are unobserved and can systematically determine not only who pays and who does not but also how households determine investments in female children or the like. For example, some households can be more progressive than others. It could be the case that those households are less likely to be involved in marriage transactions, more likely to educate their daughters and marry them late. Individual level traits such as beauty can affect the size of marriage transaction and timing of marriage at the same time. In this dissertation,
I have ignored unobserved variables that might affect marriage transactions. However, this does not posit any problem in establishing association between various variables and marriage transactions.
123 6.4 Direction for future research
Literature and research on marriage transactions are based on mainly four different aspects of the transactions. These are motives, determinants, inflation (or deflation) and effect of the practices on society in general. Not all these sub-fields have attracted equal attention from social scientists. Research on motives behind marriage transactions is not established yet. Studies suggest mixed results and coexistence of the two main motives discussed before. The underlying motive is very important not only to explain the determinants of marriage transactions but also to analyze its effect on the society and for policy implications. Thus, there is a need to systematically link determinants, motive and outcome of marriage transactions as the finding of one affects the explanation of the other.
Another important aspect of dowry is its measurement. Conclusions can be very different depending on how it is measured. Gifts given at marriage are not comparable with dowry demanded as a condition of marriage. Besides, dowry paid as voluntary price is also different from demanded dowry. Note that though voluntary price is a price, the size of this is likely to be decided by bride’s parents to attract better quality groom. On the other hand, bride’s family does not have much say about dowry that is demanded from groom’s family. The easiest solution to solve this ambiguity is to identify gifts from demanded dowry. Nonetheless, the difficult part of this easiest solution is collecting reliable information on demanded dowry. As demanding dowry at marriage is illegal in
India this information is likely to be under reported or not reported at all. Thus, a
124 consensus about dowry measurement is needed to make dowry research more relevant for policy implication.
6.5 Conclusion
Historical knowledge points toward decline in dowry practice with modernization of society, whereas India is an exception. There is anecdotal evidence that in India parents of a girl start to save from her birth so that they can provide a decent dowry at her marriage. This does not posit any concern as long as this dowry payment is voluntary and free of social pressure. Nonetheless, the same dowry payment can be associated with completely different meaning and have the potential to affect the society in a very adverse way if dowry is paid to meet the demand of groom’s family or out of social pressure. When marriage is associated with demanded dowry it has implications for bride’s position within her husband’s family and the society in general. Dowry paid as groomprice has the potential to increase gender discrimination by worsening sex ratios through sex selective abortion, infanticide and gender biased child mortality. Given the negative influence of dowry in the society, there should be social sanctions against dowry demands. The Indian government has taken a few appreciable steps to address the issue.
Paying or receiving dowry as a part of marriage contract is illegal in India since 1961.
The Dowry Prohibition (Amendment) Act was passed in 1984 in an attempt to deal with the weaknesses of the dowry act. The findings of horizontal expansion over time suggest the inadequateness of these policies. Lack of proper implementation of policies at the local community level can also be responsible for the ineffectiveness of the policies.
125 In conclusion, it is important to understand how certain institutions and processes work in order to design effective solution delivered to the society through social policies. This dissertation, I believe, has resolved some of the uncertainties that existed in the literature. Therefore, it has taken an important step to further our knowledge on the dynamics of marriage transactions in India and some of its key determinants. There are certain limitations of this study, which I discuss in this chapter and also point out directions for future research to overcome them. This dissertation thus contributes to the broader agenda of improving our understanding of an important social institution that is too often responsible for violating the basic human rights of a traditionally disadvantaged gender group and their families. It is this author’s sincere hope that the evidences presented in this study not only enriches our knowledge about marriage transactions but also adds to the urgency of social policy actions required to address the issue.
References
126 127 Alexander, K. C. 1968. Social Mobility in Kerala. Deccan College, Poona.
Amin, Sajeda and M Cain. 1998. “The Rise of Dowry in Bangladesh,” in The Continuing Demographic Transition, ed. By G. W. Jones, R. M. Douglas, J. C. Caldwell, and R. M. D’Souza. Oxford University Press.
Amin, Sajeda, Luciana Suran, Lopita Huq and Kobita Chowdhury. 2004. “Does dowry make life better for brides? A test of bequest theories of dowry”. Population Council Research Division Working paper 195.
Anderson, Siwan. 2003. “Why Dowry Payments Declined with Modernization in Europe but are Rising in India”, Journal of Political Economy, Vol. 111, No. 2, (269-308).
Arunachalam, Raj and Trevon D. Logan. 2006. “On the Heterogeneity of Dowry Motives” NBER Working Paper 12630, October 2006.
Becker Gary S. 1991. A Treatise on the Family, Harvard University Press.
Bhat, P. N. Mari and Shiva S. Halli. 1999. “Demography of Brideprice and Dowry: Causes and Consequences of the Indian Marriage Squeeze” Population Studies, Vol. 53, No. 2. (Jul., 1999), (129-148).
Billig, M S. 1992. “The Marriage Squeeze and the Rise of Groomprice in India’s Kerala State”, Journal of Comparative Family Studies, Vol.23, No.2.
Bloch, F and V. Rao. 2002. “Terror as a bargaining instrument: A case study of dowry violence in rural India”, American Economic Review, Vol. 92(4), P. 1029-1043.
Boserup, E. 1970. Women’s Role in Economic Development. London: Allen and Unwin.
Botticini, Maristella and Aloysius Siow. 2002. “Why Dowries?” mimeo, University of Toronto, Canada.
128 Botticini, Maristella. 1999. “A loveless economy? Intergenerational altruism and marriage market in a Tuscan town, 1415-1436” Journal of Economic History, Vol. 59(1), (104 - 21).
Bradford, N J. 1985. “From Bridewealth to Groom-fee: Transformed Marriage Customs and Socioeconomic Polarisation Amongst Lingayats”, Contributions to Indian Sociology (n.s.), Vol.19, No.2, (369-302).
Caldwell, J C, P Reddy and P Caldwell. 1983. “The causes of marriage change in South India”, Population Studies, Vol. 37, (343-361).
Caldwell, J C, P Reddy and P Caldwell. 1982. “The causes of demographic change in rural South India: A micro approach”, Population and Development Review, Vol. 8(4), (689-727).
Caplan, Lionel. 1984. “Bridegroom Price in Urban India: Class, Caste and ‘Dowry Evil’ among Christians in Madras” Man, n.s., 19 (June 1984), (216-33).
Census, 2001. Office of the Registrar General and Census Commissioner, India.
Dalmia, Sonia. 2004. “A hedonic analysis of marriage transactions in India: estimating determinants of dowries and demand for groom characteristics in marriage” Research in Economics, Vol. 58, (235-255).
Dasgupta, Indraneel and Mukherjee, Diganta, "'Arranged' Marriage, Dowry and Female Literacy in a Transitional Society" (August 2003). CREDIT Research Paper No. 03/12.
Deolalikar , Anil B. and Vijayendra Rao. 1990. “The demand for Dowries and Bride Characteristics in Marriage: Empirical Estimates for Rural South-Central India”, in Gender, Population and Development, K Maithreyi, R Sudershan and A Shariff (eds), Oxford University Press, Oxford, (122-140).
Diwan, Paras. 1983. Family Law: Law of Marriage and Divorce in India New Delhi: Sterling Publishers Private Limited, 1983.
129 Driver, Edwin D. 1984. “Social Class and Marital Homogamy in Urban South India” Journal of Comparative Family Studies 15 (Summer 1984), (195-210).
Dyson, Tim; Mick Moore. 1983. “On Kinship Structure, Female Autonomy, and Demographic Behavior in India”, Population and Development Review, Vol. 9, No. 1. (Mar., 1983), (35-60).
Edlund, Lena. 2006. “The Price of Marriage: Net vs. Gross Flows and the South Indian Dowry Debate?” The Journal of European Economic Association, papers and proceedings Vol. 4, No. 2-3, April-May 2006, pp. 542-551.
Edlund, Lena. 2001. “Dear Son-Expensive Daughter: Why Do Scarce Women Pay to Marry?” Manuscript. New York: Columbia University, Dept. of Economics, 2001.
Edlund, Lena. 2000. “The marriage squeeze interpretation of dowry Inflation: A comment?” Journal of Political Economy, Vol. 108, No. 6 (December, 2000), pp. 1327-1333.
Edlund, Lena. 1997. “Dowry to Daughters: Theory and Evidence?” Manuscript. New York: Columbia University, Dept. of Economics, 1997.
Epstein, T. Scarlett. 1973. South India: Yesterday, Today and Tomorrow – Mysore Villages Revisited, London: Macmillan.
Foster, Andrew D. and M. Rosenzweig (2003). “Agricultural Development, Industrialization and Rural Inequality” unpublished manuscript, Brown University.
Foster, Andrew, Ali Protik and Nizam Khan, 2007. “Dynamics of the Marriage Market: Marriage Squeeze and Age at Marriage in Rural Bangladesh” unpublished manuscript, Brown University.
Fox, J. 1990. Describing univariate distributions, in J. Fox and J. S. Long (ed) Modern Methods of Data Analysis, Newbury Park, CA:Sage Publications. (58-125).
Goody, J. 1976. Production and Reproduction, Cambridge University Press, Cambridge.
130 Goody, J. 1990. The Oriental, The Ancient and The Primitive, Cambridge University Press, Cambridge.
Greene, William H. 2000. Econometric Analysis, Prentice-Hall, Inc.
Hooja, S. L. 1969. Dowry system in India: a case study. Asia Press, Delhi.
Ifeka, C. 1989. “Hierarchical women: The dowry system and its implications among Christians in Goa, India”, Contributions to Indian sociology, New Series 23 (2), (261- 284).
Munshi, Kaivan and Mark R Rosenzweig. 2005. “Why is Mobility in India so Low? Socail Insurance, Inequality, and Growth” BREAD Working Paper No. 097.
Karlekar, Malavika. 1998. “Domestic Violence”, Economic and Political Weekly, Vol. 33, No. 27, July 1998, (1741-1751).
Kuhn, R and Ali Protik. 2007. “The Hidden Cost of Migration: Effect of Brother’s Migration on Sister’s Marital Outcome” unpublished manuscript, Brown University.
Kolenda, P. 1987. Regional Differences in Family Structure in India. Jaiur: Rawat Publications.
Kumari, R. 1989. Brides are not for burning. Dowry victims in India. Radiant Publishers, New Delhi.
Lindenbaum, Shirley. 1981. “Implications for Women of Changing Marriage Transactions in Bangladesh” Studies Family Planning, Vol. 12 (November 1981), (394-401).
May, Linda. 1986. Arranging marriages: Negotiation and decision-making in South India, PhD Dissertation, University of Pennsylvania.
Menski, W. 1998. (ed.) South Asians and the Dowry Problem, Vistaar Publications, New Delhi.
131 Miller, B. D. 1981. The Endangered Sex: Neglect of Female Children in Rural North India. Ithaca: Cornell University Press.
Nair, Kusum, 1986. “Impact of New Technologies in Agriculture in South Asia” in James Warner Bjorkman edited The Changing Division of Labor in South Asia: Women and Men in India’s Society, Economy, and Politics. The Riverdale Company Publishers, Riverdale, Maryland, USA. (53-59).
Paul, Madan C. 1985. Dowry and Position of Women in India: A Study of Delhi Metropolis. New Delhi: Inter-India Pubs., 1985.
Prakasa, Rao. 1982. Marriage, the Family and Women in India, Printox: South Asia Books, 1982.
Rajaraman, Indira. 1983. “Economics of Brideprice and Dowry” Economic and Political Weekly, Vol. 18(8), (275-278).
Randeria S. and L. Visaria, 1984. “Sociology of brideprice and dowry”, Economic and Political Weekly, Vol. 19(15) (648-652).
Rao, Vijayendra. 1993a. “The rising price of husbands: a hedonic analysis of dowry increases in rural India”, Journal of Political Economy, Vol. 101 (666-677).
Rao, Vijayendra. 1993b. “Dowry inflation in rural India: a statistical investigation”, Population Studies, Vol. 47 (283-293).
Rao, V V. P and V. N. Rao (1980). “The dowry system in Indian marriages: attitudes, expectations and practices”, International Journal of Sociology of the Family, Vol. 10 (99-113).
Reddy, G. Sreenivas, and Rajanna, Ch. 1984. “Intercaste Marriage: A Study “, Social Welfare, Vol. 31, (April 1984), (10-12).
Rosenzweig, M R and O. Stark. 1989. “Consumption smoothing, migration and marriage: evidence from rural India“ Journal of Political Economy, Vol. 97, (905-926).
132 Srinivas, M N. 1989. The Cohesive Role of Sanskritization and Other Essays, Delhi: Oxford University Press.
Srinivas, M N. 1984. Some Reflections on Dowry, New Delhi: Oxford University Press.
Srinivas, M N. 1952. Religion and Society Among the Coorgs of South India, Oxford Clarendon Press, 1952.
Tambiah, S J. 1973. “Dowry and Bridewealth and the Property Rights of Women in South Asia” in Bridewealth and Dowry, J Goody and S Tambiah (eds.) Cambridge University Press, Cambridge, (59-166).
Upadhya, C B. 1990. “Dowry and Women’s Property in Coastal Andhra Pradesh”, Contributions to Indian Sociology (n.s.), Vol.24, No.1, (29-59).
Zhang, J and W Chan. 1999. “Dowry and Wife’s Welfare: A Theoretical and Empirical Analysis” Journal of Political Economy, Vol. 107, No. 4 (786-808).
Appendix A
133 134 Sample and Missing Cases
A detailed description of data and sample that I am using for the analyses is presented in Chapter 3. The sample includes information of 2,154 marriage unions. Out of these 2154 cases, for 38 marriages dowry information is missing, which is 1.8 % of all marriages. Again, for another 825 marriages that is 38 % of all marriages, brideprice information is not reported.
I replaced the missing dowry information with zero as the cases with missing dowry information are not statistically different from those cases where zero dowry was paid. Table A1 presents a comparison between cases with missing dowry information and non-missing cases and zero dowry cases.
Cases with missing brideprice information are different than cases with available information or cases with zero brideprice. Therefore, I have not included those cases in the analyses. This creates some bias for the individual level attributes as those are the characteristics that are significantly different. For example, bride’s education is significantly higher for the missing cases. Therefore, excluding those cases will result in underestimation of the effect of bride’s education on brideprice if the missing brideprice are high in value. On the contrary, for missing cases if the brideprice is actually zero then bride’s education will be over estimated. But it is not possible to anticipate the direction of the bias with the information available.
135 Table A1: Comparison Between Cases with Available Dowry Information, Zero Dowry and Missing Dowry
Variables Dowry = Dowry = Dowry = Diff bt dowry non- Diff bt dowry zero non zero missing missing and dowry and dowry missing (t) missing missing (t)
Brideprice (Rp) 765949 13850 19904 12245* (2.37) -19766*** (2037) Bride’s age (yr) 17.91 18.69 19.47 1.56* (2.09) - 0.78 (- 0.87) Groom’s age (yr) 23.12 24.06 25.26 2.14* (2.10) - 1.20 (- 0.91) Bride’s yr of 3.04 2.48 3.00 - 0.04 (-0.09) - 0.55 (- 1.14) schooling Groom’s yr of 5.19 4.87 4.97 - 0.22 (-0.35) - 0.11 (-0.15) schooling Bride’s parental 554 525 511 - 43 (-0.29) 14 (0.09) land (acre) Groom’s parental 705 593 498 - 207 (-1.19) 95 (0.59) land (acre) Distance of 33 31 25 - 8 (-0.65) 6 (0.31) marriage migration (km) Total cases 2116 174 38
49 Number of cases with available brideprice information = 1,291 50 Number of cases with available brideprice information = 167
136 Table A2: Comparison Between Cases with Available Brideprice (BP) Information, Zero Brideprice and Missing Brideprice
Variables BP = non BP = zero BP = Diff bt BP non- Diff bt BP missing missing missing and BP zero and BP missing (t) missing (t) Dowry (Rp) 23751 17989 53626 29875*** -35637*** (10.13) (11.06)
Bride’s age (yr) 17.62 17.55 18.45 0.82*** - 0.89*** (4.08) (3.40) Groom’s age (yr) 22.59 21.99 24.06 1.47*** - 2.07*** (5.35) (7.01) Bride’s yr of 2.58 2.56 3.79 1.20*** - 1.23*** schooling (9.53) (8.78)
Groom’s yr of 4.98 5.31 5.52 0.54** - 0.21*** schooling (3.21) (-1.14)
Bride’s parental 582 605 506 - 77 100* land (acre) (-1.95) (2.25)
Groom’s parental 695 725 711 16 14 land (acre) (0.35) (0.26)
Distance of 34 37 31 - 3 6 marriage migration (-0.84) (1.69) (km) Total cases 1329 934 825