How Love Conquered Marriage: Theory and Evidence on the Disappearance of Arranged Marriages
Total Page:16
File Type:pdf, Size:1020Kb
How Love Conquered Marriage: Theory and Evidence on the Disappearance of Arranged Marriages Gabriela Rubio∗ February 2014 University of California, Merced The most up-to-date version of this paper is available at http://gabrielarubio.bol.ucla.edu/ Abstract This paper documents the sharp and continuous decline of arranged marriages (AM) around the world during the past century, and describes the factors associated with this transition. To understand these patterns, I construct and empirically test a model of marital choices that assumes that AM serve as a form of informal insurance for parents and children, whereas other forms of marriage do not. Children accepting the AM will have access to insurance but might give up higher family income by constraining their geographic and social mobility. Children in love marriages (LM) are not geographically/socially constrained; they can look for the partner with higher labor market returns and have access to better remunerated occupations. The model predicts that AM disappear when the net benefits of the insurance arrangement decrease relative to the (unconstrained) returns outside of the social network. Using con- sumption and income panel data from the Indonesia Family Life Survey, I show that consumption of AM households does not vary with household income (while consumption of LM households does), consistent with the model's assumption that AM provides insurance. I then empirically test the main predictions of the model. I use the introduction of the Green Revolution (GR) in Indonesia as a quasi-experiment. First, I show that the GR increased the returns to schooling and lowered the variance of agricultural income. Then, I use a difference-in-difference identification strategy to show that cohorts exposed to the GR experienced a faster decline in AM as predicted by the theoretical framework. Second, I show the existence of increasing divorce rates among couples with AM as their insurance gains vanish. Finally, using the exogenous variation of the GR, I find that couples having an AM and exposed to the program were more likely to divorce, consistent with the hypothesis of declining relative gains of AM. JEL: B52, C72, D13, J11, J12, O15, O53, O55. ∗[email protected], [email protected]. I am truly indebted to my advisor, Adriana Lleras-Muney, for uncountable hours of discussion, guidance and support. I am also particularly grateful to Leah Boustan, Paola Giuliano, Maurizio Mazzocco and Robert Jensen for their encouragement, numerous discussions and guidance. And I am especially indebted to George Georgiadis for all his help and advice throughout all the stages of this project. I would also like to thank Kathleen McGarry, Maria Casanova, Dora Costa, Till von Wachter, Aprajit Mahajan, and Simon Board for helpful comments and suggestions. I have also benefited from participants at the Applied Micro Lunch Meeting, the Albert Family Proseminar at UCLA, the 2013 All-California Labor Economics Conference, the Applied Microeconomics Seminar at University of California at Riverside, and from seminars at Simon Fraser University, Eastern Connecticut State University, University of California at Merced, University of Illinois at Chicago and NERA. I would like to thank Esther Duflo for making available the data used in her paper \Schooling and Labor Market Consequences of School Construction in Indonesia: Evidence from an Unusual Policy Experiment." All remaining errors are mine. 1 1 Introduction At the beginning of the 20th century, 72% (or more) of all marriages in Asia and Africa were arranged by the families of the couple. Throughout the last century, these marriages have decreased by approximately 40%. This paper uses a variety of sources to document this continuous and large decline in arranged mar- riages in several countries of Asia and Africa (Turkey, Saudi Arabia, Israel, Japan, Korea, China, Taiwan, Indonesia, Malaysia, Cambodia, Vietnam, Sri Lanka, Nepal, Togo and Ghana). Although a few countries (India, Pakistan and Bangladesh) do not follow the same patterns (arranged marriages still represent at least 95% of all marriages), young cohorts living in these countries' urban areas have also started to move away from arranged marriages. The goal of this paper is to understand the main driver(s) of the transition by proposing and testing empirically a model of marital choices. I first show that this transition away from arranged marriages in favor of self-choice or \love" marriages is correlated with increases in education, formal employment, urbanization, and declines in agriculture. These trends are common in all the countries where micro-data is available, suggesting that despite having different institutions at work, there is a fundamental economic explanation behind these changes in marriage institutions. Based on these patterns, I build a simple model of marriage choice. I assume that arranged marriages serve as a form of informal insurance (as suggested in the literature of sociology, anthropology and economics, e.g. Rosenzweig and Stark (1989)) whereas other marriages (outside one's networks) do not. In the first period, parents spend resources on educating their offspring and looking for a spouse for an arranged marriage. Both investments have pay-offs in the second period when the child earns an education dependent wage subject to a shock, transfers a share back to her parents and decides to marry the parental arranged spouse or choose her own spouse. Arranged marriages provide insurance (to both parents and children) because parents (but not the children) can observe who within their network faces income shocks that are negatively correlated with their child's shocks (by observing the entire shock histories of other households and by having repeated interaction with them). However, arranged marriages come with a cost: they constrain the choice set and the mobility of the child, thereby reducing her potential income, compared to the case in which she has the option to move geographically, find a more lucrative occupation or a spouse with higher earnings. Thus, there is a trade-off between marrying within the network and marrying outside the network: individuals might be willing to give up potentially higher income in exchange for a risk-sharing agreement. The model predicts that arranged marriages disappear when the net benefits of the insurance arrangement decrease relative to the (unconstrained) returns outside of the social network. When this is the case, parents invest in more education for the child, effectively increasing her outside option and, thus, the probability that she will reject the arranged marriage. In this framework, the decline in arranged marriages and the increase in years of schooling are endogenously determined; more specifically, the theoretical framework predicts that when the returns to schooling increase or the variance of income declines, the demand for insurance decreases | households pay a greater cost or \premium" for the insurance. I extend the base model in two directions that provide additional testable implications. First, motivated by the fact that divorce has a low cost in some of the countries studied (e.g. Indonesia), I introduce the possibility of divorce in a third period. Holding the cost of divorce constant, the model implies that couples in arranged marriages will have higher rates of divorce as their insurance gains vanish. Second, I assume two children within the household and theoretically explore how parents decide which child to offer an arranged marriage. I show that when social networks are small (equivalent to have 2 insurance partners with positively correlated shocks), parents have incentives to arrange the marriage of only one child and they choose to marry only the child with the lowest expected return in the labor market. The second part of the paper uses micro data to test the model. First, I test the main assumption of the model that arranged marriages provide insurance to the couple.1 I implement a standard test of full insurance using the first three waves (1993, 1997 and 2000) of the Indonesia Family Life Survey (IFLS). This is the only data set containing both marriage type, and monthly consumption and income data of couples. For this test, I assume that the relevant unit for risk sharing is the village or town. If there is full insurance, the consumption of each individual household should not depend on its income but should have a one-to-one relationship with the average consumption of the insurance group. Although the test is formally rejected for both types of marriages, the results show that the consumption of arranged marriage couples does not depend on their income.2 In contrast, the consumption of couples in love marriages varies significantly with their income, suggesting that individuals in arranged marriages are better insured. These results are consistent with Rosenzweig and Stark (1989), who show that households in India use the marriage of their children as insurance to mitigate the effect of profits shocks on consumption. I then test the two main predictions of the model, namely that factors that lower the demand for informal insurance (increasing returns to schooling and decreasing income variance) accelerated the transition to love marriages and increased the investment in education. To test these predictions, I use the gradual introduction of the Green Revolution in Indonesia as a quasi-experiment. The Green Revolution refers to a series of technological innovations associated with the diffusion of higher yield variety seeds in developing countries in the late 1960s, which increased the returns to schooling and agricultural income (Foster and Rosenzweig, 1996).3 I collected data from the 1963 and 1983 Indonesian agricultural census (and other printed sources) on the intensity of the Green Revolution by district (% of inputs use, % of land covered, rice yields, among others). I combine this information with the 1976 and 1995 census, the early 1980s socioeconomic surveys and the 1993 wave of IFLS to study the effect on labor market outcomes.