86265 AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION The Economics of Consanguineous Marriages The definitive version of the text was subsequently published in Review of Economics and Statistics, 95(3), 2013-07 Published by MIT Press THE FINAL PUBLISHED VERSION OF THIS ARTICLE IS AVAILABLE ON THE PUBLISHER’S PLATFORM This Author Accepted Manuscript is copyrighted by the World Bank and published by MIT Press. It is posted here by agreement between them. Changes resulting from the publishing process—such as editing, corrections, structural formatting, and other quality control mechanisms—may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a CC BY-NC-ND 3.0 Unported license http://creativecommons.org/licenses/by-nc-nd/3.0/. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript of an Article by Do, Quy-Toan; Iyer, Sriya; Joshi, Shareen The Economics of Consanguineous Marriages © World Bank, published in the Review of Economics and Statistics95(3) 2013-07 http:// creativecommons.org/licenses/by-nc-nd/3.0/ © 2014 The World Bank THE ECONOMICS OF CONSANGUINEOUS MARRIAGES Quy-Toan Do, Sriya Iyer, and Shareen Joshi* Abstract—This paper provides an economic rationale for the practice of transfers of assets from the bride’s family to the groom’s consanguineous marriages observed in parts of the developing world. In a model of incomplete marriage markets, dowries are viewed as ex ante family, and second, enforcement mechanisms for informal transfers made from the bride’s family to the groom’s family when the contracts are stronger within kinship networks than outside promise of ex post gifts and bequests is not credible. Consanguineous unions such networks. The crux of our model is that a marriage is join families between whom ex ante pledges are enforceable ex post. The model predicts a negative relationship between consanguinity and dowries a contract in which two families make a long-term commit- and higher bequests in consanguineous unions. An empirical analysis based ment to support their offspring through gifts, bequests, and so on data from Bangladesh delivers results consistent with the model. forth. These also enhance the value of the match and, conse- quently, the social status of each family. However, once links I. Introduction have formed and are costly to sever, each family may now pre- fer to invest in alternative opportunities while free-riding on C ONSANGUINEOUS marriage, or marriage between close biological relatives who are not siblings, is a social institution that is, or has been, common throughout human the other family’s investments. To overcome this time incon- sistency, early transfers between families are viewed as an ex ante alternative when ex post investment commitments are history (Bittles, 1994; Bittles, Coble, & Rao, 1993; Hussain not credible. In South Asia, where marriage is characterized & Bittles, 2000). Although in the Western world, consan- by patrilocal exogamy, we postulate the commitment problem guineous marriages constitute less than 1% of total marriages, to be on the bride’s side so that these early monetary transfers this practice has had widespread popularity in North Africa, correspond to dowries. To this aspect, we add two features. the Middle East, and South Asia (Maian & Mushtaq, 1994; First, the extent to which agents are time inconsistent depends Bittles, 2001).1 Scientific research in clinical genetics doc- negatively on how closely related the partners are. Between uments a negative effect of inbreeding on the health and cousins, ex ante commitments are more credible arguably mortality of human populations and the incidence of dis- because informal contracts are easier to enforce within the orders and disease among the offspring of consanguineous extended family. Second, dowries are costly since they imply unions (Banerjee & Roy, 2002). To economists, therefore, borrowing on the credit market in order to make payments this contemporary incidence of consanguineous marriage and at the time of marriage. Consequently, our model predicts its persistence in some societies is puzzling. that consanguinity and dowries substitute as instruments to It is in this setting that this paper makes its contribution: to overcome or mitigate the time-inconsistency problem. argue that consanguinity is a rational response to a marriage Our data test the central idea that consanguinity may be a market failure rather than simply a consequence of culture, relatively cheaper way for families to deal with the problem of religion, or preferences. The starting points of our analysis dowry costs in rural marriage markets. We use data on 4,364 are the following two stylized facts commonly observed in households from the 1996 Matlab Health and Socioeconomic large parts of South Asia and elsewhere: first, marriage cel- Survey conducted in 141 villages in Bangladesh. We find that ebrations are often associated with significant dowries, or women in consanguineous unions are, on average, 6% to 7% less likely to bring a dowry at marriage, after controlling Received for publication June 1, 2009. Revision accepted for publication for other attributes at the time of marriage, suggesting that January 12, 2012. consanguinity and dowry are substitutes.2 We also find that * Do: World Bank; Iyer: University of Cambridge; Joshi: Georgetown women marrying their cousins are on average 4% more likely University. to receive any form of inheritance. The negative relationship We thank Gary Becker, Alan Bittles, Francis Bloch, Michael Bordo, Jishnu Das, Partha Dasgupta, Timothy Guinnane, Geoffrey Harcourt, Hanan between dowry and consanguinity, on the one hand, and the Jacoby, Saumitra Jha, Lakshmi Iyer, Matthew Mason, Mushfiq Mobarak, positive relationship between bequests and consanguinity, on David Newbery, Hashem Pesaran, Biju Rao, Hamid Sabourian, Azusa Sato, the other, is strongly suggestive of consanguinity affecting the Paul Schultz, and Chander Velu for helpful comments and discussions. All remaining errors are ours. The findings, interpretations, and conclusions timing of marital transfers. In further analysis, we corroborate expressed in this paper are entirely our own. They do not necessarily reflect the hypothesis that more stringent credit constraints and lower the view of the World Bank, its executive directors, or the countries they wealth levels will lead to lower dowry payments and a higher represent. 1 In Iraq, for example, 46.4% of marriages are between first or second prevalence of consanguineous marriages. cousins (Hamamy, Zuhair, & Al-Hakkak, 1989). In India, consanguineous The study of consanguineous unions allows us to shed marriages constitute 16% of all marriages, but this varies from 6% in the some light on specific agency problems in marriage markets north to 36% in the south (IIPS & ORC Macro International, 1995; Baner- jee & Roy, 2002). The highest level of inbreeding has been recorded in the and institutions created to mitigate them. It is therefore of South Indian city of Pondicherry, in which 54.9% of marriages were con- the same class of models as Becker (1981), Bloch and Rao sanguineous, corresponding to a mean coefficient of inbreeding of 0.0449, (2002), Botticini and Siow (2003), and Jacoby and Mansuri considered very high by the standards of other populations (Bittles, 2001). Among immigrant populations in the United Kingdom, those of Pakistani origin display a preponderance of consanguineous marriage, estimated to 2 These findings are entirely consistent with earlier observations made by be as high as 50% to 60% of all marriages in this community (Modell, sociologists and demographers (Centerwall & Centerwall, 1966; Reddy, 1991). 1993). The Review of Economics and Statistics, July 2013, 95(3): 904–918 © 2013 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 905 (2010). Our paper is also naturally related to the literature on countries like France, Germany, the Netherlands, and the dowry payments (see Anderson, 2007a, for a thorough dis- United Kingdom, is now likely to be of the order of 1% to 3% cussion). However, unlike Becker (1981) and Botticini and or more. Consanguineous marriage is particularly popular in Siow (2003), our model does not provide a theory of dowry Islamic societies and among the poor and less educated pop- per se but rather links dowries, bequests, and consanguinity ulations in the Middle East and South Asia (Hussain, 1999; in a theory of optimal timing of marital transfers. Finally, by Bittles, 2001). looking at marital sorting, our analysis relates to Lee (2009) There is also evidence that different kinds of consan- and Banerjee et al. (2009) in that we find a preference to guineous unions are favored by different subpopulations. choose mates of similar family background. For example, while Hindu women in South India typically We review important facts and findings related to consan- marry their maternal uncles, Muslim populations favor first- guinity in section II. In section III, we present and solve cousin marriages (Dronamaraju & Khan, 1963; Centerwall a model of incomplete marriage markets and discuss alter- & Centerwall, 1966; Reddy, 1993; Iyer, 2002). native explanations of consanguinity. Section IV uses data The acceptability of consanguineous unions differs across from Bangladesh to test the main predictions of the theory. religions. In Europe, Protestant denominations permit first- Section V concludes. cousin marriage. In contrast, the Roman Catholic church requires permission from a diocese to allow them. Judaism II. Historical and Religious Background permits consanguineous marriage in certain situations, for example, uncle-niece unions. Consanguinity is also permit- In the field of clinical genetics, a consanguineous marriage ted in Islam. According to the institutional requirements of is defined as “a union between a couple related as second Islam in the Koran, and the Sunnah, “a Muslim man is prohib- cousins or closer, equivalent to a coefficient of inbreeding in ited from marrying his mother or grandmother; his daughter their progeny of F ≥ 0.0156” (Bittles, 2001).3 This means or granddaughter, his sister whether full, consanguine or uter- that children of such marriages are predicted to inherit copies ine, his niece or great niece, and his aunt or great aunt, paternal of identical genes from each parent, which are 1.56% of or maternal” (Azim, 1997).10 However, the Sunnah depict all gene loci over and above the baseline level of homozy- that the Prophet Mohammad married his daughter Fatima to gosity in the population at large; the closer are the parents, Ali, his paternal first cousin; this has led researchers to argue the larger is the coefficient of inbreeding. A common con- that for Muslims, first-cousin marriage follows the Sunnah cern is that consanguinity leads to higher levels of mortality, (Bittles, 2001; Hussain, 1999). morbidity, and congenital malformations in offspring due to the greater probability of inheriting a recessive gene (Schull, 1959; Bittles, 1994). III. The Economics of Consanguineous Marriages Historically in Europe, consanguineous marriage was prevalent until the twentieth century and was associated with The model we present belongs to the class of agency mod- royalty and landowning families (Bittles, 1994).4 During the els of marriage. Families are viewed as agents that invest nineteenth and twentieth centuries, consanguinity was prac- in a joint project: the marriage of their offspring. However, ticed more in the Roman Catholic countries of southern the institution of marriage is characterized by two features: Europe than in their northern European Protestant counter- (a) dissolution (divorce) is costly, and (b) marriage contracts parts (McCollough & O’Rourke, 1986). Since the sixteenth are incomplete. The combination of these two features under- century in England, marriage between first cousins has been mines the credibility of some ex ante commitments on the part considered legal. But close-kin marriages are not always of the families. For example, in Jacoby and Mansuri (2010), legally permitted elsewhere. For example, in the United the marriage contract is incomplete because the groom can- States, different states have rulings on unions between first not commit ex ante not to be violent toward his wife. Once cousins: in some states, such unions are regarded as illegal; the marriage takes place, he has incentives to engage in vio- others go so far as to consider first-cousin marriage a crimi- lent behavior to, among other things, extract rents from his nal offense (Ottenheimer, 1996). The overall prevalence of in-laws (Bloch & Rao, 2002). The institution of watta-satta, consanguineous marriage, especially in western European or exchange marriages, then emerges to alleviate this market failure; when grooms cannot commit ex ante not to be violent 3 The coefficient of inbreeding is the probability that two homologous toward their bride-to-be, marrying the groom’s sister to the alleles in an individual are identical by descent from a common ancestor. 4 Intermarriage among the aristocracy that occurred in Europe in previ- bride-to-be’s brother provides a credible retaliation threat that ous centuries was not always in order only to retain land. Annan’s (1999) makes the initial nonviolence claim incentive compatible. In social history of academic dons in Cambridge, Oxford, and elsewhere in the same class of models, Botticini and Siow (2003) argue the United Kingdom documents that for several centuries, British academia was dominated by a handful of families, suggesting that there were lev- that in patrilocal societies, daughters cannot commit to man- els of intermarriage on an unprecedented scale in order to preserve power age parental assets with the same care as their male siblings and influence on intellectual ideas. In particular, the author suggests this do once they get married. This implies that parental transfers was responsible for sustaining an intellectual aristocracy that gave not only social benefits to its members but power and influence over the history of will optimally take the form of dowries for daughters and ideas. bequests for sons. 906 THE REVIEW OF ECONOMICS AND STATISTICS In our model, altruistic parents make transfers to their chil- Equation (1) puts a limit on how credible it is for parents to dren once they are married, and this also enhances the value pledge more than a given fraction of their wealth for their off- of the match. At the time of marriage, however, they are spring. That fraction is assumed to be decreasing with social unable to contract on such future transfers, and as marriages distance, capturing the idea that social distance is negatively are costly to dissolve, ex ante commitments are no longer associated with families’ abilities to write binding contracts credible. In a patrilocal society, where a bride migrates to the with each other. home of her husband after marriage, the incentive to renege In addition, family transfers cannot exceed their net worth, is likely to be particularly strong for the bride’s parents, since which translates into a budget constraint on ex ante transfers: they may prefer to direct their transfers to coresident sons. As for k ∈ {m, f }, in Botticini and Siow (2003), dowries (or bride-prices) then become the second-best solution to this time-inconsistency Dk ≤ wk , (2) problem. Social distance between the families of the bride as well as ex post transfers; and the groom can significant ly influence the terms of the marriage contract. On the one hand, we assume that ceteris zk ≤ wk − Dk + D−k , (3) paribus social distance enhances the outcomes of marriage: families can diversify genes, hedge risks, smooth consump- where −k refers to k ’s in-laws. tion, or simply integrate their social networks (Rosenzweig Finally, we make two additional assumptions: (a) marriage & Stark, 1989; La Ferrara, 2003). On the other hand, shorter is always preferred to remaining single, and (b) once cele- social distance acts as social capital by making ex ante con- brated, marriage is indissoluble. These two assumptions are tracting between families easier: close relatives have more certainly not innocuous. However, in the context of South (verifiable) information about each other, are more likely to Asia, social pressure for a woman to get married and the exert effort in economic activities, are less likely to engage stigma associated with divorce are arguably quite strong. in opportunistic behavior, and are likely to show higher lev- We therefore believe these assumptions to be reasonable els of trust, cooperation, and altruism to both their natal and first-order approximations. marital families (Putnam, 2000). We now proceed to a formal description of the forces at Marital technology. At T = 1, once brides and grooms play. Proofs are left to the appendix. have celebrated their marriage, they receive transfers tm and tf from their parents. These transfers enter a marital production function: A. Model Setup Y tm , tf |wm , wf , dmf = A wm , wf , dmf × tm + tf . Individuals are assimilated to their families and labeled m ∈ M and f ∈ F for male and female, respectively. A pair The marital production function has constant returns to (m, f ) ∈ M × F is characterized by wealth levels wm , wf and scale A wm , wf , dmf . The coefficient of returns A (.) is social distance dmf ∈ [0, 1] . More specifically, families are assumed to be continuously differentiable and increasing in 1 points on the surface of a cylinder of radius π . Each family is wm and wf with nonnegative cross-partial derivatives (Becker, thus characterized by its cylindrical coordinates (w, α) where 1981), and increasing concave in social distance dmf . Fur- height w ∈ h, h ¯ measures wealth with 0 < h ≤ h ¯ , and social thermore, we assume that for every (m, f ) ∈ M × F , and for distance between two individuals m and f is defined by the k ∈ {m, f }, difference in azimuths dmf = π 1 |αm − αf |, and thus takes values in [0, 1] . To abstract from marriage market squeeze ∂ 2 A wm , wf , dmf ∂ A wm , wf , dmf wm + w f + > 0. issues (see Rao, 1993; Anderson, 2007b), we assume brides ∂ wk ∂ dmf ∂d and grooms to be uniformly distributed over the cylinder. (4) The positive dependence of A (.) on parental wealth cap- Timing and the institution of marriage. The economy tures the fact that in addition to monetary transfers, parents consists of two periods. At T = 0, couples (m, f ) form, and transmit social status to their children and share their net- marriages are celebrated with ex ante transfers Df made from works. Second, A (.) is also assumed to be positively cor- the bride’s to the groom’s family (a dowry) /or Dm made from related with social distance: when spouses are not related, the groom’s to the bride’s (a bride-price), or both. At that time, they can diversify genes, hedge risks, integrate their social parents also commit to make ex post transfers zf and zm to networks, and so forth. Furthermore, with the additional the married couple at future date T = 1. These transfers can assumption made with equation (4), the marital production be thought of as parental gifts or bequests. However, parents function Y wm , wf , dmf is supermodular. cannot fully commit to these transfers and are subject to the Finally, agents have access to a storage technology with following limited-liability condition: for k ∈ {m, f } , returns normalized to 1. This storage technology proxies for investment and consumption opportunities available to zk ≤ 1 − dmf wk (1) parents outside their offspring’s marital production. THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 907 Parental preferences. At the heart of the model is the 4. Gale-Shapley stability: there does not exist two cou- asymmetry between the bride’s parents and the groom’s. We ples (m, f ) and (m , f ) and payments {(D ˜m ,D ˜ f ), thus assume that while the groom’s parents internalize the (z ˜ f ), (tm , t f )} that satisfy, equations (8) to (11), and ˜m , z ˜ ˜ full marital product, so that Um ˜ t f |wm , wf , dm f ≥ Um tm , tf |wm , wf , dm f , tm , ˜ Um tm , tf , Dm , Df |wm , wf , dmf (13) = Y tm , tf |wm , wf , dmf + wm + Df − Dm − tm , (5) Uf ˜ t f |wm , wf , dm f ≥ Uf tm , tf |wm , wf , dmf (14) tm , ˜ the bride’s side discounts marital output by a factor θ < 1: with one inequality holding strictly. Uf tm , tf , Dm , Df |wm , wf , dmf B. Equilibrium Characterization = θY tm , tf |wm , wf , dmf + wf + Dm − Df − tf . (6) First, condition (11) implies that the groom’s family has the incentives to invest in their child’s marriage, while the storage This discrepancy can be interpreted as coming from virilocal- technology is a more attractive option to the bride’s family ity, an institution exogenously given to the model. If the bride once marriage has been celebrated. Thus, the use of dowries leaves her parents to live with her in-laws (which is empiri- allows the bride’s family to overcome this time inconsistency. cally the case in our setting), her parents might not capture the Their incentives to do so hinge on the willingness to secure full marital product of the couple. Botticini and Siow (1993) a wealthier groom. rely on similar asymmetry arguments to construct a theory of dowry and inheritance. To make the analysis relevant, we Proposition 1: Marriage market equilibrium with no further assume that for every possible couple (m, f ) ∈ M × F , credit constraints. If A h ¯ , 1 < 2A h, h, 0 , there exists ¯, h a marital discount factor θ < θ ¯ , ¯ , such that for every θ ∈ θ, θ θA wm , wf , dmf < 1 < A wm , wf , dmf . (7) a profile (m, f )m∈M , f ∈F characterized by We henceforth define θ ¯ = 1 • positive assortative matching: ¯ ,1) , the supremum of all ¯ ,h A(h possible values of θ that satisfy equation (7) and restrict the ¯. analysis to θ < θ wm = wf , and Equilibrium concept. As stated earlier, we restrict our- selves to cases in which every individual finds a match. Then dmf = 1 a profile (m, f )m∈M , f ∈F with associated payments • payments equal to Dm , Df , z m , z f , t m , t f m∈M , f ∈F tm , tf = zm , zf is an equilibrium if it satisfies the following conditions: and 1. Feasibility: Dm , Df = 0, wf and zm , zf = wm + wf , 0 Dk ≤ wk , (8) zk ≤ wk + D−k − Dk , (9) is an equilibrium. where k ∈ {m, f } and −k refers to k ’s in-laws Supermodularity is driving positive assortative matching. 2. Limited commitment: However, stability requires a high enough discount factor θ so that brides are willing to invest in the marital production func- zk ≤ 1 − dmf wk (10) tion. In this setting, dowries are perfect substitutes for social ties. Couples can be socially distant and lower the costs of consanguinity; they then overcome the commitment problem 3. Incentive compatibility: by pledging payments upfront in the form of dowries. tk ∈ arg max Uk t , t−k |wm , wf , dmf , (11) t C. Equilibrium with Credit Constraints subject to We now assume that raising funds to pay for dowries is costly. When family k transfers Dk to family −k , −k receives zk ≤ tk ≤ wk + D−k − Dk (12) only a fraction 1−γ, the rest being lost. With costly payments, 908 THE REVIEW OF ECONOMICS AND STATISTICS dowries are no longer perfect substitutes of social proximity, stringent. Applying the implicit function theorem to equation which implies a trade-off between social distance and the cost (15) shows that for every wealth level wf , the equilibrium ∗ of funds. distance dmf verifies Proposition 2: Marriage market equilibrium with credit ∗ ∂ dmf ∗ ∂A ¯, h constraints. If A h ¯ , 1 < 2A h, h, 0 , there exists γ ˆ <1 sgn = −sgn dmf + A ≤ 0. (16) ∂γ ∂d ¯ and θ < θ such that for any parameter configurations such ˆ that γ < γ ˆ and θ ∈ θ ¯ , a match profile characterized by ˆ, θ The intuition underlying equation (16) is straightforward. When credit constraints are more stringent, dowries are more • positive assortative matching costly relative to close-kin marriage, so that the equilibrium social distance decreases with the cost of equity. wm = wf , A second implication of the analysis conducted so far is a comparative statics exercise with respect to wealth. We have, and on the one hand, ∗ dmf = dmf ∗ ∂ dmf ∂ 2A ∗ ∗ sgn = sgn wm + wf − γdmf wf where dmf satisfies ∂ wm ∂ wm ∂ d ∂A ∂A ∂ + − γ wf , (17) ∗ ∗ ∂d ∂ wm A wm , wf , dmf wm + wf − γdmf wf ∂d ∗ = A wm , wf , dmf γwf (15) while on the other hand, we have ∗ • payments equal to ∂ dmf ∂ 2A ∗ ∂A sgn = sgn wm + wf − γdmf wf + ∂ wf ∂ wf ∂ d ∂d tm , tf = zm , zf ∂A ∗ ∂ A − γwf − γ dmf +A . (18) and ∂ wf ∂d ∗ Dm , Df = 0, dmf wf and An increase in the groom’s wealth has two effects on ∗ ∗ spousal distance. Distance positively affects marital produc- zm , zf = wm + (1 − γ) dmf wf , 1− dmf wf tivity, the first two terms on the right-hand side of equation is an equilibrium. (17), but this also translates into a higher opportunity cost of dowry transfers—the third term. The effect of the bride’s The left hand side of equation (15) measures the marginal wealth comes with an added cost of distance, since higher cost of consanguinity; by construction, marrying farther away wealth implies larger dowry payments ceteris paribus, which has a direct and positive effect on payoffs because families is captured by the last term on the right-hand side of equation can diversify their pools of genes, hedge risks across families (18). In both cases, assuming that equation (4) holds, and for (Rosenzweig & Stark, 1989), or merge social networks for sufficiently low dowry costs γ, equilibrium spousal distance better access to credit or labor markets (La Ferrara, 2003). increases with either spouse’s wealth. On the other hand, the right-hand side of equation (15) mea- sures the agency cost. Increasing distance between spouses D. Summary of Testable Implications and Alternative increases the agency problem, requiring a larger dowry to Explanations be paid. This implies a larger transaction cost, which trans- lates into a larger opportunity cost of investment. Since A (.) We have presented and analyzed a general equilibrium is increasing and concave in d , the second-order condition model of the marriage market characterized by positive assor- ∗ holds, making equation (15) necessary and sufficient and dmf tative matching. In this market, parents commit wealth, which unique for any given levels of wealth wm and wf . determines spousal “market value.” However, marriage con- In summary, we have so far described a marriage mar- tracts are incomplete, so that wealth commitments might not ket failure for which consanguinity and dowries are two be credible. Thus, the institution of dowry emerges as a solu- distinct mitigating practices that act as substitutes. Dowries tion to this time-inconsistency; parents pay ex ante in the are viewed as ex ante transfer of control over assets to pal- form of dowry what they cannot commit to transfer ex post liate a lack of ex post incentives to invest. Consanguinity in the form of gifts or bequests. Our agency theory delivers is a practice that directly reduces the agency problem. In predictions not as much on the size of transfers between fam- so doing, equation (15) determines the optimal trade-off ilies as on the timing of such transfers. Because we stipulate between the two. One immediate implication relates to the that contract incompleteness is less severe among close kin, prevalence of consanguinity when credit contraints are more consanguineous marriages are a viable alternative when the THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 909 payment of dowries comes at too high a cost. Thus, our model significantly reduce the costs of searching for a suitable part- predicts that dowry and consanguinity are substitutes: ner. The central idea is that since the bride and groom are generally known to each other prior to marriage, consan- • Prediction 1: Dowry levels are lower in consanguineous guineous marriages do not require families to screen each marriages. other in order to assess the quality of the upcoming match. In many rural societies, this process can take considerable time Symmetrically, if ex ante payments are lower in consan- as well as resources (Sander, 1995). Moreover, parents in guineous marriages, we should expect larger ex post consanguineous unions know their future selves-in-law and transfers. their families, reducing the uncertainty about the compatibil- ity of spouses and families. While all these alternative stories • Prediction 2: Bequests or gifts to daughters are larger have some appeal in explaining the prevalence of consanguin- when they marry close kin. ity, they do not speak to the relationship of consanguinity, bequests, and dowries, which is central to this paper. We How well consanguinity substitutes for dowries depends in next move to the empirical section to empirically test the part on the cost of dowry transfers. If credit constraints are predictions of our model. stringent, one might expect the consanguinity option to be more attractive. IV. Empirical Evidence from Bangladesh The data used in the analysis are drawn from the 1996 • Prediction 3: Consanguinity is more prevalent in envi- Matlab Health and Socioeconomic Survey (MHSS).6 We sup- ronments with more severe credit constraints. plement these data with those on climate data on annual rainfall levels in the Matlab area for the period 1950 to 1996.7 We concluded our analysis with an investigation of the rela- The 1996 MHSS contains information on 4,364 households tionship between consanguinity and wealth. Since higher spread over 2,687 baris, or clusters, in 141 villages. Mat- spousal wealth implies that more is at stake, consanguinity lab is an Upazila (subdistrict) of Chandpur district, which is comes at a higher cost: about 50 miles south of Dhaka, the capital of Bangladesh. Eighty-five percent or more of the people in Matlab are • Prediction 4: Consanguineous unions are less prevalent Muslims, and the others are Hindus. Although it is geograph- among wealthier unions. ically close to Dhaka, the area has been relatively isolated and inaccessible to communication and transportation. The Prediction 4 depends on assumptions about the marital pro- society is predominantly an agricultural society, although duction function. It has, for example, been argued that 30% of the population reports being landless. Despite a consanguinity provides a means to consolida te and maintain growing emphasis on education and increasing contact with family assets and resources (Goody, 1973; Agarwal, 1994; urban areas, the society remains relatively traditional and Bittles, 2001). Cross-cousin marriages, wherein an individ- religiously conservative (Fauveau, 1994). ual marries a mother’s brother’s (or father’s sister’s) offspring, For the purpose of understanding the incidence of con- are able to unite individuals in different patrilines both in sanguineous marriage in the MHSS data, we rely on the the same bloodline, ensuring a consolidation of resources. section of the survey that asked men and women retrospec- This suggests that even at higher levels of wealth, marriage tive information about their marriage histories. Information contract incomplet eness may come at too high a cost, so on first marriages only was considered.8 Our working sample that consanguineous marriages are preferred. 5 Consanguinity consists of 4,087 married women and 3,358 married men, might yet be driven by multiple other factors. A first alter- once we require complete information on age and educa- native explanation for consanguinity is that it is the outcome tion, marriage (including age at marriage, relationships to of personal preference that is mediated by the influence of their spouses, and payments of dowry), parental character- religion or cultural practice. As discussed in section II, much istics, parental assets, inheritances and inherited assets, and of the literature from sociology and biological anthropology household demographics. Descriptive summary statistics of is predicated on this assumption about consanguinity. The the variables of interest are provided in table 1. A quick glance argument here is that because consanguinity is a practice that 6 This survey is a collaborative effort of RAND, the Harvard School of has enjoyed much support historically in certain populations, Public Health, the University of Pennsylvania, the University of Colorado it continues to be popular among these communities to the at Boulder, Brown University, Mitra and Associates, and the International present day. A second explanation for consanguinity is that Centre for Diarrhoeal Disease Research, Bangladesh. 7 The University of Delaware Air and Temperature Precipitation Data it may be a favored form of marriage simply because it can are provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado, USA. 5 The framework of our model also allows thinking of consanguineous 8 About 15% of men and about 7% of women reported that they have had marriages as allowing the enforcement of insurance contracts between fam- more than one marriage. This gender difference is driven by the fact that ilies when an insurance motive is driving marriage (Rosenzweig & Stark, while divorced and widowed men typically remarry, most women in these 1989). same circumstances do not (Joshi, 2004). 910 THE REVIEW OF ECONOMICS AND STATISTICS Table 1.—Summary of Key Variables for Female and Male Samples Female Sample Male Sample Variable Mean SD N Mean SD N Married cousin (1 = Yes, 0 = No)a 0.107 0.309 4,095 0.107 0.309 3,358 Married relative (1 = Yes, 0 = No)b 0.079 0.269 4,095 0.081 0.273 3,358 Married non-relative within village (1 = Yes, 0 = No) 0.143 0.350 4,095 0.140 0.347 3,358 Dowry (1 = Yes, 0 = No) 0.367 0.482 4,095 0.14 0.347 3,358 Log of dowry value (in thousands of taka) −1.694 1.025 4,095 −2.015 0.759 3,358 Age (in years) 36.652 10.135 4,095 47.482 14.089 3,358 Father’s age (in years) 64.786 15.475 4,015 67.984 15.849 3,297 Muslim (1 = Yes, 0 = No) 0.887 0.316 4,095 0.892 0.311 3,358 Years of schooling (in years) 2.204 3.002 4,095 3.356 3.887 3,358 Birth order 2.920 1.801 4,095 2.512 1.566 3,358 Mother attended school (1 = Yes, 0 = No) 0.007 0.082 4,095 0.027 0.162 3,358 Father attended school (1 = Yes, 0 = No) 0.391 0.488 4,095 0.285 0.452 3,358 Mother alive at marriage (1 = Yes, 0 = No) 0.933 0.251 4,095 0.862 0.345 3,358 Father alive at marriage (1 = Yes, 0 = No) 0.842 0.365 4,095 0.656 0.475 3,358 Number of brothers alive at marriage 2.211 1.469 4,095 1.699 1.390 3,358 Number of sisters alive at marriage 1.904 1.368 4,095 1.743 1.338 3,358 Parents farmland (100000) 344.50 2,760.72 4,033 467.475 4,008.87 3,311 Log of parents farmland 0.854 5.085 4,033 1.437 4.782 3,311 Inherited anything from parents (1 = Yes, 0 = No) 0.097 0.296 4,095 0.694 0.461 3,358 Received a transfer from parents (1 = Yes, 0 = No) 0.828 0.377 4,034 Received an inheritance or a transfer from parents (1 = Yes, 0 = No) 0.848 0.359 4,034 Rainfall dev when mother aged 13 (in millimeters)d 2.079 0.385 4,095 a Includes marriages to first cousins. b Includes marriages to all relatives other than first cousins. c Calculated as the deviation of rainfall from the average (of the area) at the average age of a woman’s marriage. at the table does not reveal systematic differences in character that compared to women who marry non-relatives, women istics between male and female samples. Admittedly, when who marry their first cousins 5% to 6% points less likely to we look at individual characteristics, males are more educated bring a dowry, and this effect is robust to controlling for indi- and, by construction of the data set, are older. vidual characteristics (age, years of schooling, religion, and Of all the female respondents in our sample, 10% married a birth order), family characteristics (mother and father were first cousin, 8% married a relative other than a first cousin, and alive at the time of marriage, number of brothers and sisters at 14% married a nonrelative in the same village. It is interesting the time of marriage, and father’s landholdings), and rainfall that 36% of women and only 18% of men report the payment at the time that a woman was of marriageable age. Consid- of a dowry at the time of marriage. We believe the difference ering that in this population, about 35% of all women report is in psychological biases in the interpretation of gifts and the payment of a dowry at the time of marriage, this is a sub- transfers as “dowries” between the the bride and the groom: stantial and important difference. The results are similar if we since women want their dowries to improve their status and expand the definition of consanguinity to include marriages acceptance in their new home, they will have a tendency to between second cousins and other types of marriages between interpret all gifts given at marriage as dowry. We therefore relatives. Marriage to other kin as well as marriages to nonkin use the female sample to carry out our analysis and include within a village are also associated with a 3.5 percentage point in our definition of the dowry all transfers that were paid at lower likelihood of dowry payment. The relationship between the time of marriage. The dowry question was asked in two dowry and social distance is strongest in the case of cousins. ways: respondents were asked whether they paid a dowry at This is consistent with our theory: dowries are predicted to the time of marriage—a binary variable—and whether this become more likely as social distance between the families took the form of bride-wealth or gifts and transfers to the of a bride and groom increases. woman’s in-laws. Respondents were also asked to provide In an additional test, we use the logarithm of the dowry an estimate of the dowry’s value. In our analysis, we use both values as a dependent variable and obtain similar results for the binary indicator and dowry values. marriages at different social distances (table 2, columns 4–6). After controlling for individual, household characteristics, and year of marriage fixed effects, the results show 6.9%, A. Consanguinity and Dowries 12.2%, 7.7% lower dowry values when the two spouses are first cousins, a relative other than first cousin, and nonrela- Prediction 1 of the model can be tested by examining tive in the same village, respectively.9 The coefficients are the simple correlations between the payment of dowries and significant at the 5% level for two of the three variables, pre- first-cousin marriages. We look at the conditional correlation sumably because the sample of dowry values is larger for between dowry payment and consanguinity by correlating the these variables. dummy variable Dowry with the various measures of con- sanguinity that were considered previously. The results are 9 Since age at marriage and year of marriage are often not remembered presented in the first three columns of table 2. They indicate with great precision, we define fixed effects over five-year windows. THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 911 Table 2.—Dowry and Consanguinity: Partial Correlations Paid Dowry Log Dowry Value (1) (2) (3) (4) (5) (6) Married cousin −.0507 −.0550 −.0613 −.0454 −.0557 −.0692 (.0205)∗∗ (.0206)∗∗∗ (.0208)∗∗∗ (.0439) (.0441) (.0447) Married relative −.0476 −.0542 −.1095 −.1226 (.0232)∗∗ (.0234)∗∗ (.0498)∗∗ (.0502)∗∗ Married nonrelative within village −.0347 −.0771 (.0182)∗∗ (.0391)∗∗ Age −.0295 −.0289 −.0301 −.0545 −.0536 −.0560 (.0085)∗∗∗ (.0085)∗∗∗ (.0085)∗∗∗ (.0183)∗∗∗ (.0183)∗∗∗ (.0183)∗∗∗ Age squared .0020 .0019 .0021 .0033 .0032 .0035 (.0012)∗∗ (.0011)∗∗ (.0012)∗∗ (.0025) (.0025) (.0025) Years of schooling −.0117 −.0118 −.0120 .0106 .0105 .0100 (.0024)∗∗∗ (.0024)∗∗∗ (.0024)∗∗∗ (.0052)∗∗ (.0052)∗∗ (.0052)∗∗ Muslim −.2484 −.2459 −.2463 −.5866 −.5792 −.5818 (.0204)∗∗∗ (.0204)∗∗∗ (.0204)∗∗∗ (.0437)∗∗∗ (.0438)∗∗∗ (.0438)∗∗∗ Birth order .0012 .0015 .0013 .0046 .0053 .0050 (.0043) (.0043) (.0043) (.0091) (.0091) (.0091) Mother alive at marriage −.0305 −.0316 −.0313 −.0399 −.0420 −.0416 (.0255) (.0255) (.0255) (.0547) (.0547) (.0547) Father alive at marriage .0201 .0202 .0199 .0089 .0087 .0084 (.0181) (.0180) (.0180) (.0387) (.0387) (.0387) Brothers at marriage .0124 .0124 .0127 .0189 .0187 .0195 (.0050)∗∗ (.0050)∗∗ (.0050)∗∗ (.0108)∗∗ (.0108)∗∗ (.0108)∗∗ Sisters at marriage .0025 .0022 .0022 .0068 .0063 .0061 (.0052) (.0052) (.0052) (.0112) (.0112) (.0112) Mother attended school −.1573 −.1553 −.1563 −.2322 −.2261 −.2297 (.0764)∗∗ (.0764)∗∗ (.0764)∗∗ (.1639) (.1638) (.1638) Father attended school −.0351 −.0346 −.0350 −.0867 −.0864 −.0866 (.0143)∗∗ (.0143)∗∗ (.0143)∗∗ (.0307)∗∗∗ (.0307)∗∗∗ (.0307)∗∗∗ Log of parents’ farmland .0036 .0036 .0035 .0104 .0102 .0101 (.0013)∗∗∗ (.0013)∗∗∗ (.0013)∗∗∗ (.0028)∗∗∗ (.0028)∗∗∗ (.0028)∗∗∗ Parents’ farmland missing −.0433 −.0456 −.0446 −.0922 −.0962 −.0951 (.0322) (.0322) (.0322) (.0691) (.0691) (.0690) Rainfall controls Yes Yes Yes Yes Yes Yes N 4,015 4,015 4,015 4,015 4,015 4,015 R2 .3394 .3396 .3407 .3308 .3316 .3323 The variable Log Dowry Value assumes a dowry value of 1 taka if no dowry was paid. All regressions have year of marriage fixed effects where year of marriage is coded as five-year intervals. Standard errors, shown in parentheses, are clustered at the bari-level. Significant at *10%, **5%, and ***1%. The relationship between dowry and consanguinity over Figure 1.—Prevalence of Cousin Marriage (Solid Line) and Dowry (Dashed Line) Over Time (1935–2000). Results are Based on the time can be observed in figure 1. Dowries in Matlab have been Sample of Adult Women increasing as the practice of first-cousin consanguineous mar- riages has been falling. Our model tells a story consistent with the observed trends: in a setting where improvements in trans- portation and communication allow individuals to search over greater geographic distances for matches at larger social dis- tances, the problem of ex ante commitment becomes greater, calling for the payment of higher levels of dowry in marriages between individuals who are outside the family network. B. Consanguinity and Bequests At the heart of our model is an intertemporal shift of trans- fers: a bride who marries into a socially distant family must bring a dowry at the time of marriage because an ex ante com- mitment by her natal family to pay future gifts and bequests is not credible. Her natal family will have incentives to free-ride on the investments of the groom’s parents and direct their own investments elsewhere. In a consanguineous or socially close higher marriage transfers and bequests after marriage (pre- union, however, the bonds of trust are likely to be stronger, diction 2). To test this prediction, we examine the relationship and interests of the bride and groom’s families are less likely between inheritance or gifts and social distance within mar- to diverge. In this instance, we expect lower dowries and riages. We define inheritance as a binary variable that takes 912 THE REVIEW OF ECONOMICS AND STATISTICS Table 3.—Correlates of Inheritances and Transfers Female Sample Male Sample Inheritance Transfers Inheritance (1) (2) (3) (4) (5) (6) Married cousin .0285 .0326 −.0232 (.0165)∗∗ (.0196)∗∗ (.0223) Married relative .0283 .0046 .0246 (.0180) (.0231) (.0231) Married nonrelative within village .0249 .0228 −.0256 (.0135)∗∗ (.0169) (.0198) Log of dowry value −.0106 −.0471 .0261 (.0200) (.0272)∗∗ (.0360) Log of dowry × age .0007 .0012 −.00007 (.0006) (.0009) (.0010) Age −.0004 −.0051 −.0166 −.0193 .0384 .0280 (.0031) (.0061) (.0041)∗∗∗ (.0083)∗∗ (.0030)∗∗∗ (.0048)∗∗∗ Age squared .0007 .0016 .0020 .0026 −.0027 −.0021 (.0004) (.0008)∗∗ (.0006)∗∗∗ (.0012)∗∗ (.0003)∗∗∗ (.0005)∗∗∗ Father’s age −.0006 −.00009 −.0003 −.0003 −.0004 .0002 (.0004) (.0003) (.0005) (.0005) (.0005) (.0004) Muslim .0663 .0787 −.0721 −.0731 −.0360 −.0027 (.0103)∗∗∗ (.0146)∗∗∗ (.0172)∗∗∗ (.0200)∗∗∗ (.0240) (.0191) Years of schooling −.0015 −.0010 −.0077 −.0081 −.0043 −.0070 (.0015) (.0017) (.0024)∗∗∗ (.0024)∗∗∗ (.0019)∗∗ (.0017)∗∗∗ Birth order .0151 .0075 −.0016 −.0019 .0411 .0116 (.0030)∗∗∗ (.0032)∗∗ (.0041) (.0044) (.0051)∗∗∗ (.0045)∗∗ Mother attended school −.0569 −.1305 .1418 .1249 .0174 −.0738 (.0129)∗∗∗ (.0547)∗∗ (.0415)∗∗∗ (.0746)∗∗ (.0411) (.0368)∗∗ Father attended school .0054 .0053 .0244 .0223 .0036 .0002 (.0099) (.0101) (.0135)∗∗ (.0138) (.0165) (.0141) Mother alive at marriage .0086 .0120 .0105 .0101 .0733 .0443 (.0222) (.0180) (.0250) (.0246) (.0201)∗∗∗ (.0172)∗∗ Father alive at marriage −.0723 −.0381 .0168 .0175 −.1905 −.1168 (.0176)∗∗∗ (.0145)∗∗∗ (.0191) (.0199) (.0150)∗∗∗ (.0142)∗∗∗ Brothers alive at marriage −.0288 −.0247 .0014 .0016 −.0315 .0017 (.0039)∗∗∗ (.0036)∗∗∗ (.0047) (.0049) (.0058)∗∗∗ (.0048) Sisters alive at marriage −.0142 −.0115 .0006 .0005 −.0080 .0063 (.0036)∗∗∗ (.0037)∗∗∗ (.0048) (.0050) (.0061) (.0048) Log parents’ farmland .0037 .0037 .0018 .0018 .0271 .0293 (.0009)∗∗∗ (.0009)∗∗∗ (.0013) (.0013) (.0017)∗∗∗ (.0013)∗∗∗ Parents’ farmland missing −.0083 −.0135 .0034 .0043 −.2714 −.2249 (.0247) (.0228) (.0306) (.0311) (.0700)∗∗∗ (.0523)∗∗∗ N 4,015 4,015 4,015 4,015 3,297 3,296 R2 .0845 .1077 .0136 .0149 .3306 .4944 F statistic 15.6947 14.7908 3.6779 2.2383 124.7328 131.3549 Table 2 notes apply here. the value 1 if the respondent reports that he or she has inher- coefficients for marriage to any relative and marriage to non- ited, or expects to inherit, anything from his or her parents. relatives within the same village are not always statistically We define transfers as a binary variable that takes the value significant in these regressions. This is consistent with the 1 if the respondent reports that she has received a transfer predictions of our theory: it predicts that the commitment to from her parents in the year preceding the survey. Among our bequeath their assets to their daughters is likely to be weaker key independent variables, we then consider three mutually when she marries more distant relatives or nonkin in a village, exclusive forms of marriages in decreasing order of social since the costs or consequences of nonpayment are likely distance: marriages to first cousins, marriages to relatives to be smaller in such relationships.10 The negative coeffi- other than first cousins, and marriages to nonrelatives from cient for the variable Log of Dowry Value in columns 2 and within the village. Our regressions control for our standard 4 also confirm our views that transfers are being optimally set of variables of individual, household, family, and climate timed; women who receive large dowries do not subsequently characteristics. Table 3 reports the results of the regressions. The positive 10 This result, in isolation, could also be explained by a higher propen- and significant estimates for the variable Married a Cousin in sity for parents to make transfers to a consanguineous couple, since they columns 1 and 3 are consistent with prediction 2: women in share more genes with them or have better information about in-laws. In our model, this would correspond to having a higher θ in consanguineous consanguineous unions are more likely to receive or expect unions. However, this assumption would not be able to explain the inverse to receive inheritances or transfers from their parents. The correlation identified in section IVA. THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 913 Table 4.—Correlates of Consanguinity and Dowry Married Cousin Paid Dowry Log Dowry Value (1) (2) (3) (4) (5) (6) Age at marriage −.0060 −.0032 −.0088 −.0001 −.0176 −.0021 (.0013)∗∗∗ (.0016)∗∗ (.0018)∗∗∗ (.0029) (.0038)∗∗∗ (.0062) Brothers at marriage −.0110 −.0088 .0109 .0130 .0190 .0186 (.0034)∗∗∗ (.0039)∗∗ (.0045)∗∗ (.0051)∗∗ (.0097)∗∗ (.0108)∗∗ Sisters at marriage −.0061 −.0041 −.0023 .0029 .0021 .0074 (.0038) (.0041) (.0048) (.0052) (.0102) (.0112) Parents’ farmland −.7509 −.7857 −.5329 .2510 1.8341 2.5783 (.3152)∗∗ (.3280)∗∗ (.6335) (.6226) (1.3523) (1.3329)∗∗ Parents’ farmland squared .4116 .4332 .3054 −.1599 −1.1520 −1.5767 (.1982)∗∗ (.2060)∗∗ (.4337) (.4255) (.9258) (.9109)∗∗ Age −.0060 −.0123 −.0088 −.0270 −.0176 −.0507 (.0013)∗∗∗ (.0074)∗∗ (.0018)∗∗∗ (.0095)∗∗∗ (.0038)∗∗∗ (.0203)∗∗ Age squared .0012 .0018 .0032 (.0009) (.0012) (.0025) Father’s age .0007 −.0003 −.0003 (.0004)∗∗ (.0005) (.0010) Years of schooling −.0025 −.0110 .0119 (.0018) (.0024)∗∗∗ (.0052)∗∗ Muslim .1188 −.2468 −.5718 (.0065)∗∗∗ (.0201)∗∗∗ (.0429)∗∗∗ Mother alive at marriage .0067 −.0288 −.0313 (.0197) (.0256) (.0548) Father alive at marriage −.0317 .0239 .0071 (.0160)∗∗ (.0200) (.0428) Birth order −.0067 .0023 .0063 (.0032)∗∗ (.0044) (.0094) Mother attended school −.0579 −.1710 −.2724 (.0526) (.0763)∗∗ (.1634)∗∗ Father attended school −.0089 −.0293 −.0724 (.0110) (.0143)∗∗ (.0306)∗∗ Rainfall controls Yes Yes Yes Yes Yes Yes Observations 4,007 4,007 4,007 4,007 4,007 4,007 R2 .0126 .0312 .2998 .3373 .2952 .3291 Notes i–iii of table 2 apply. receive inheritances or transfers. The correlation is negative and the negative relationship between bequests and inher- although statistically significant in the case of transfers to itances (for women)—together lend support to our model younger couples only.11 rather than alternative explanations of consanguinity. Next, we run the same regressions for the male sample (table 3, columns 5 and 6). We do not run the regression C. Consanguinity, Credit Constraints, and Wealth for transfers for this sample because most men live either with their parents or in close proximity to their parents, mak- Our next step examines the determinants of consanguinity ing transfers between households very difficult to measure. and dowry payment. Since our survey lacks direct informa- Note that in column 5, the coefficients for the variable Mar- tion on parents’ financial state at the time of a woman’s ried a Cousin are negative, although the coefficients are not marriage, we rely on proxy variables. The first of these is statistically significant. This lack of significance is consis- the value of the father’s landholdings. To the extent that land tent with our predictions; we expect no association between markets in rural South Asia are thin (Griffin, Khan, & Ick- inheritance and consanguinity for the male sample, that is, owitz, 2000), current landholdings (or landholdings at the the woman’s husband and her in-laws, since patrilocal resi- time of father’s death) may be regarded as a proxy for past dence frees this group of the commitment problem faced by landholdings. The results in table 4 confirm that the incidence the bride’s family. Similarly, in column 6, we observe no sig- of consanguineous marriage decreases with the increase in nificant effect of the the magnitude of a dowry on inheritances the extent of father’s farmland and the coefficient is signifi- among men. cant at the 5% level in the consanguinity regression (column Taken together, the the three results obtained thus far— 2). In other words, consanguineous marriages are more com- the negative relationship between dowry and consanguinity, mon among poorer (and likely credit constrained) households the positive relationship between bequests and consanguinity, (prediction 3). This relationship however, is nonlinear. After a point, greater landholdings are actually associated with 11 Since information on transfers consists of five-year recalls, it is not positive dowry payments, consistent with the discussion in surprising that older couples no longer receive transfers from their parents section III. The opposite signs of the coefficients on con- when these are either deceased or no longer income earners. sanguinity and dowry variables are once again consistent 914 THE REVIEW OF ECONOMICS AND STATISTICS with prediction 1. These results are also consistent with the REFERENCES results of Mobarak, Kuhn, and Peters (2006), who use a Agarwal, Bina, A Field of One’s Own: Gender and Land Rights in South difference-in-differences framework to postulate that the con- Asia (Cambridge: Cambridge University Press, 1994). struction of an embankment in Matlab several years prior Anderson, Siwan, “The Economics of Dowry and Brideprice,” Journal of Economic Perspectives 21:4 (2007a), 151–174. to the 1996 data created a positive wealth shock for some ——— “Why the Marriage Squeeze Cannot Cause Dowry Inflation,” households, who were then able to pay higher dowries for Journal of Economic Theory 137:1 (2007b), 140–152. their daughters and were less likely to enter into consan- Annan, Noel G., The Dons: Mentors, Eccentrics and Geniuses (Chicago: University of Chicago Press, 1999). guineous marriages. Their findings thus reinforce the view Azim, Saukath, Muslim Women: Emerging Identity (Jaipur and New Delhi: that as families get wealthier (starting from an initial condi- Rawat Publications, 1997). tion of low levels of wealth), credit constraint s may weaken Banerjee, Abhijit, Esther Duflo, Maitreesh Ghatak, and Jeanne Lafortune, “Marry for What? Caste and Mate Selection in Modern India,” and the family is able to search for a groom outside the kin- NBER working paper 14958 (2009). ship network by providing higher levels of dowries for their Banerjee, Sushanta K., and T. K. Roy, “Parental Consanguinity and Off- daughters. spring Mortality: The Search for Possible Linkage in the Indian Context,” Asia-Pacific Population Journal 17:1 (2002), 17–38. Finally, we look at alternative proxy variables to test our Becker, Gary S., A Treatise on the Family (Cambridge, MA: Harvard hypothesis. For example, we could postulate that when mar- University Press, 1981). riage takes place at an early age, parents might face steeper Bittles, Alan H., “The Role and Significance of Consanguinity as a Demo- graphic Variable,” Population and Development Review 20:3 (1994), cash constraints as they have had less time to accumulate 561–584. assets. In this case, prediction 3 suggests that a consan- ——— “Consanguinity and Its Relevance to Clinical Genetics,” Clinical guineous union might be chosen instead. We therefore look Genetics 60:2 (2001), 89–98. at the correlation between age of marriage and consanguin- Bittles, Alan H., J. Michael Coble, and N. Appaji Rao, “Trends in Consan- guineous Marriage in Karnataka, South India, 1980–1989,” Journal ity (table 4, columns 1–2) as well as age of marriage and of Biosocial Science 25:1 (1993), 111–116. dowry payment (columns 3–6). We find that consanguineous Bloch, Francis, and Vijayendra Rao, “Terror as a Bargaining Instrument: A marriages are more likely to be early marriages, while no Case Study of Dowry Violence in Rural India,” American Economic Review 92:4 (2002), 1029–1043. such asssociation is found with respect to dowry payment. Botticini, Maristella, and Aloysius Siow, “Why Dowries?” American Eco- Admittedly, age at marriage, dowry, and consanguinity are nomic Review 93:4 (2003), 1385–1398. joint decisions, so that the results need to be interpreted with Centerwall, Willard R., and Siegried. A. Centerwall, “Consanguinity and Congenital Anomalies in South India: A Pilot Study,” Indian Journal caution. Another source of variation in credit is the number of Medical Research 54 (1966), 1160–1167. of brothers and sisters alive at the time of marriage: a larger Dronamaraju, Krishna, R., and P. Meera Khan, “The Frequency and number of sisters (or, symmetrically, a smaller number of Effects of Consanguineous Marriages in Andhra Pradesh,” Journal of Genetics 58 (1963), 387–401. brothers) would increase the financial burden on the child Fauveau, Vincent (ed.), Matlab: Women, Children and Health (Bangladesh: to be married, so that consanguineous marriage is a more International Centre for Diarrhoeal Disease Research, 1994). likely option to large dowry payments (prediction 3). The Goody, Jack, “Bridewealth and Dowry in Africa and Eurasia” in Jack Goody and Stanley Tambiah (eds.), Bridewealth and Dowry (Cambridge: results presented in table 4 are consistent with that hypothesis: Cambridge University Press, 1973). an additional male sibling at time of marriage is associated Griffin, Keith, Azizur R. Khan, and Amy Ickowitz, Poverty and Distribution with a lower likelihood that a girl marries a first cousin, of Land (United Nations Development Programme, 2000). Hamamy, Hanan, A. Zuhair, and S. Al-Hakkak, “Consanguinity and Repro- though the magnitude of the effect is small, at about 1%. ductive Health in Iraq,” Human Heredity 39 (1989), 271–275. The effect is symmetric when we look at dowry payments Hussain, Rafat, “Community Perceptions of Reasons for Preference for (columns 3 and 4) as well as actual dowry values (columns Consanguine ous Marriages in Pakistan,” Journal of Biosocial Science 31 (1999), 449–461. 5 and 6). Hussain, Rafat, and Alan H. Bittles, “Sociodemographic Correlates of Con- sanguineous Marriage in the Muslim Population of India,” Journal of Biosocial Science 32 (2000), 433–442. V. Conclusion International Institute of Population Sciences, Mumbai, and ORC Macro International, National Family Health Survey 1992–93 (Mumbai: This paper has argued that consanguinity is a response to IIPS, 1995). a marriage market failure in developing countries. The start- Iyer, Sriya, Demography and Religion in India (Delhi: Oxford University ing point of our analysis is the recognition that dowries exist Press, 2002). Jacoby, Hannan G., and Ghazala Mansuri, “Watta Satta: Exchange Mar- across many societies and that consanguinity is also perva- riage and Women’s Welfare in Rural Pakistan,” American Economic sive across many parts of the world. We propose a theoretical Review 100:4 (2010), 1804–1825. model of the marriage market to reconcile the existence of Joshi, Shareen, “Female Household-Headship in Rural Bangladesh: Inci- dence, Determinants and Impact on Children’s Schooling,” Growth these two facts. We argue that these two social practices Center discussion paper 94 (2004). together address an agency problem between spouses’ fam- La Ferrara, Eliana, “Kin Groups and Reciprocity: A Model of Credit ilies and then provide empirical evidence that are consistent Transactions in Ghana,” American Economic Review 93:5 (2003), 1730–1751. with the central predictions of the model. By focusing on Lee, Soohyung, “Marriage and Online Mate-Search Services: Evidence the economic underpinnings of consanguineous marriage, we from South Korea,” unpublished manuscript, University of Maryland identified agency problems in marriage markets and docu- (2009). Maian, Afsar, and Rehana Mushtaq, “Consanguinity in Population of Quetta mented the existence of institutions designed to overcome (Pakistan): A Preliminary Study,” Journal of Human Ecology 5 them. (1994), 49–53. THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 915 McCullough, John M., and Dennis H. O’Rourke, “Geographic Distribution Before focusing on the interesting case (case 4), we first dismiss the obvious of Consangui nity in Europe,” Annals of Human Biology 13 (1986), cases. Since grooms benefit from in-marriage investments, they are not 359–368. benefiting from a deviation with a poorer bride (case 1) or a bride with Mobarak, A. Mushfiq, Randall Kuhn, and Christina Peters, “Marriage equal wealth (cases 2 and 3): Market Effects of a Wealth Shock in Bangladesh,” unpublished manuscript, Yale University (2006). Case 1: If wm > wf , then wf > wf , which implies Modell, Bernadette, “Social and Genetic Implications of Customary Con- sanguineous Marriage among British Pakistanis,” Journal of Medical Vm wm , wf , 1 > Vm wm , wf , dm f . Genetics 28 (1991), 720–723. Ottenheimer, Martin, Forbidden Relatives: The American Myth of Cousin Since Vm wm , wf , dmf is the highest utility an individual m can obtain by Marriage (Chicago: University of Illinois Press, 1996). marrying a spouse f , the proposed deviation makes m strictly worse off. Putnam, Robert D., Bowling Alone: The Collapse and Renewal of American Thus, there is no deviation from the equilibrium such that wm > wf . Community (New York: Simon and Schuster, 2000). Rao, Vijayendra, “The Rising Price of Husbands: A Hedonic Analysis of Case 2: If wm = wf and dm f < 1, then Dowry Increase in Rural India,” Journal of Political Economy 101:3 (1993), 666–677. Vm wm , wf , 1 > Vm wm , wf , dm f . Reddy, P. Govinda, Marriage Practices in South India (Madras: University of Madras, 1993). In such case, m is made strictly worse off. Thus, there is no deviation from Rosenzweig, Mark R., and Oded Stark, “Consumption Smoothing, Migra- the equilibrium such that wm = wf and dm f < 1. tion, and Marriage: Evidence from Rural India,” Journal of Political Economy 97:4 (1989), 905–926. Case 3: If wm = wf and dm f = 1, then Sander, William, The Catholic Family: Marriage, Children, and Human ∗ Capital (Boulder, CO: Westview Press, 1995). Vm wm , wf , dm f ≥ A wm , wf , dm f ˜f + D wm + z ˜f Schull, William J., “Inbreeding Effects on Man,” Eugenics Quarterly 6 (1959), 102–109. ˜ f = dm f wf so that ˜ f = 1 − dm f wf and D implies that z ∗ Vm wm , wf , dm f = A wm , wf , dm f ˜f + D wm + z ˜f , APPENDIX: PROOFS ∗ Vf wm , wf , dm f = θA wm , wf , dm f ˜f + D wm + z ˜f . Proof of Proposition 1 Thus, there is no deviation from the equilibrium such that wm = wf and To simplify future discussions, we first restrict strategies to payoff- dm f = dm f and either m or f is made strictly better off. relevant strategies: Case 4: The final case is the case wm ≤ wf . Without loss of generality, we can suppose that D˜f,z˜ f is such that Lemma 1: Payoff-relevant strategies. Any match profile (m, f )m∈M , f ∈F it leaves m indifferent between his equilibrium payoff and the deviation with associated payments payoff: (Dm , zm , tm ) , Df , zf , tf m∈M , f ∈F A wm , wf , 1 wm + wf = A wm , wf , dm f ˜f + D wm + z ˜f , (A1) has an identical payoff profile to match profile (m, f )m∈M , f ∈F with associated while making f strictly better off: payments θA wm , wf , 1 wm + wf (A2) 0, 0, wm + Df − Dm , Df − Dm , zf , zf m∈M , f ∈F . < θA wm , wf , dm f wm + z ˜ f + wf − z ˜f + D ˜f − D ˜f . Subgame perfection implies that grooms will always invest whatever Equation (A1) implies funds are available to them, or wm + Df − Dm , while brides will not exceed their initial commitment zf . We can henceforth limit ourselves to profiles A wm , wf , 1 wm + wf − wm = z ˜f + D ˜f, Df , zf f ∈F to characterize payments associated with a given match profile A wm , wf , dm f (m, f )m∈M , f ∈F . which can be plugged into equation (A2) and, after rearranging, yields Proof of Lemma 1. Since, by assumption, separation is not a credible threat, equation (7) implies that constraint (12) is binding for the bride’s A wm , wf , 1 wm + wf − A wm , wf , dm f wm + wf family only, so that tm = wm + Df − Dm and tf = zf are the unique (A3) solutions to maximization of equation (11) subject to equation (12). It is A wm , wf , dm f wf + wm − A wm , wf , 1 wm + wf then straightforward to verify that 1 < − 1. θA wm , wf , dm f 0, 0, wm + Df − Dm , Df − Dm , zf , zf Since dm f ≤ 1, we have also satisfies equations (8–14) and gives the same payoffs to both m ∈ M and f ∈ F as (Dm , zm , tm ) , Df , zf , tf does. A wm , wf , 1 wm + wf − A wm , wf , dm f wm + wf Consider a deviation from two individuals (m , f ) , from equilibrium A wm , wf , dm f wf + wm − A wm , wf , 1 wm + wf matches (m , f ) and (m, f ) . D ˜f,z˜ f denotes the associated (payoff- A wm , wf , 1 wm + wf − A wm , wf , 1 wm + wf relevant) transfers. ≥ On the equilibrium path, indirect utilities for m and f are given by A wm , wf , 1 wf + wm − A wm , wf , 1 wm + wf Vm wm , wf , 1 = A wm , wf , 1 wm + wf Since wm < wm , supermodularity implies A wm , wf , 1 wm + wf − A wm , wf , 1 wm + wf and ≥ A wm , wf , 1 wf + wm − A wm , wf , 1 wm + wf Vf wm , wf , 1 = θA wm , wf , 1 wm + wf . ≥ 0. 916 THE REVIEW OF ECONOMICS AND STATISTICS By transitivity, we thus have Case 4: The final case is the case wm < wf . Suppose that D ˜f,z ˜ f is the underlying transfers. Before going further, we state a first preliminary result: A wm , wf , 1 wm + wf − A wm , wf , dm f wm + wf ≥1 A wm , wf , dm f wf + wm − A wm , wf , 1 wm + wf Lemma 2: Subgame perfect strategies. A match profile (m, f )m∈M , f ∈F with associated payments (Dm , zm , tm ) , Df , zf , tf m∈M , f ∈F is an equilib- and rium only if 1 = wm + (1 − γ) Df − Dm ≤2 (A4) tm (A5) θA wm , wf , dm f tf = zf for any value of θ ≥ θ, where θ = 2A h1,h,0 and θ < θ ¯ if condition and ( ) A h ¯ , 1 < 2A h, h, 0 is satisfied. ¯, h ¯ , inequalities (A3) and (A4) form a contradiction; zf < 1 − dmf wk ⇒ Df = 0 For any θ ∈ θ, θ . (A6) the proposed deviation cannot make f strictly better off while leaving m Dm = 0 indifferent. This concludes the proof of proposition 1. Proof of Lemma 2. Equalities (A5) are subgame perfect transfers, since Proof of Proposition 2 θA wm , wf , dmf < 1 < A wm , wf , dmf . The spouses’ utilities are given by We follow the same strategy as for proposition 1. Consider a deviation Um tm , tf = A wm , wf , dmf wm + (1 − γ) Df − Dm + zf , from two individuals (m , f ) , from equilibrium matches (m , f ) and (m, f ) . ˜f,z Similarly, D ˜ f denotes the associated (payoff-relevant) transfers. Uf tm , tf = θA wm , wf , dmf wm + (1 − γ) Df − Dm + zf On the equilibrium path, indirect utilities for m and f are given by + wf + (1 − γ) Dm − Df − zf . ∗ Wm wm , wf , dm f ∗ ∗ ∗ = A wm , wf , dm w m + 1 − dm f wf + (1 − γ) dm f wf f (i) Suppose that Dm > 0. Then take η = min Dm , zf . Transfers z ˜f = zf − η and D˜ m = Dm − η leave m indifferent while making f strictly and better off. (ii) Suppose that Df > 0 and zf < 1 − dmf wf . Then take ∗ Wf wm , wf , dmf ∗ = θA wm , wf , dmf ∗ wm + 1 − dmf ∗ wf + (1 − γ) dmf wf . = min Df , 1 − dmf wf − zf . ∗ Transfers z ˜ f = Df − make both m and f strictly ˜ f = zf + and D Case 1: wm > wf Optimality of dm f implies that better off. ∗ Wm wm , wf , dm f ≥ Wm wm , wf , dm f . Therefore (i) and (ii) imply that equation (A6) holds. Conditions (A6) reflect the fact that in equilibrium, dowry payment is Since wm > wf , hence wf > wf , so that kept to a minimum and thus used only when equation (10) is binding on the bride’s side, while grooms do not pay bride-prices. Finally, equalities (A5) Wm wm , wf , dm f > Wm wm , wf , dm f . are subgame-perfect conditions for ex post transfers. By transitivity, Lemma 3: Supermodularity. Equation (15) defines a unique function ∗ ∗ d (.) such that for every (m, f ) ∈ M × F, dmf = d wm , wf . Furthermore, Wm wm , wf , dm f > Wm wm , wf , dm f . the indirect utility defined by Consequently, m is always worse off when deviating with some f such W ∗ wm , wf = A wm , wf , d wm , wf wm + wf − γd wm , wf wf that wm > wf . Thus, there is no deviation from the equilibrium such that wm > wf . is supermodular. ∗ ∗ Case 2: If wm = wf and dm f = dm f , then optimality of dm f implies ∗ ∗ Wm wm , wf , dm > Wm wm , wf , dm f . Proof of Lemma 3. dmf , when not a corner solution, is implictly defined f by first-order condition Consequently, m is always worse off when deviating with some f such that ∗ ∂ wm = wf and dm f = dm f . Thus, there is no deviation from the equilibrium f wm , wf , d = ∗ A wm , wf , dmf ∗ wm + wf − γdmf wf ∗ such that wm = wf and dm f < dm f . ∂d ∗ ∗ − A wm , wf , dmf γwf = 0 Case 3: If wm = wf and dm f = dm f , then ∗ and therefore depends on wm , wf only. Since A (.) is increasing and Wm wm , wf , dm f ≥ A wm , wf , dm f ˜f + D wm + z ˜ f − γD ˜f concave in the second-order derivative, ˜ f = dm f wf so that ˜ f = 1 − dm f wf and D implies that z ∂ 2 A wm , wf , d ∂ A wm , wf , d wm + wf − γdwf − 2γwf < 0, ∗ ∂ d2 ∂d Wm wm , wf , dm f = A wm , wf , dm f ˜f + D wm + z ˜ f − γD ˜f , Wf ∗ wm , wf , dm = θA wm , wf , dm f wm + z˜ f + Df − γD ˜ ˜f . for every d , so that for every wm , wf ∈ h, h ¯ 2 , f wm , wf , d is continu- f ∗ ous over [0, 1] and decreasing in d . If the solution is interior, then dmf is ∗ Thus, there is no deviation from the equilibrium such that wm = wf and uniquely defined by f wm , wf , d = 0. When dmf is a corner solution and ∗ dm f = dm f that makes either m or f strictly better off. f wm , wf , d > 0 (resp. f wm , wf , d < 0) for every d ∈ [0, 1], then we THE ECONOMICS OF CONSANGUINEOUS MARRIAGES 917 ∗ ∗ define dmf = 1 (resp. dmf = 0). We can apply the implicit function theo- to be supermodular is ∗ rem to determine the derivatives of dmf ≡ d wm , wf with respect to wf , respectively: ∂2 ∂ ∂A W ∗ wm , wf = wm + wf − γdwf + A (A8) ∂ wm ∂ wf ∂ wf ∂ wm ∂A = wm + wf − γdwf ∂ d wm , wf ∂ wm ∂ wf ∂ wf ∂A ∂d ∂A + 1 − γd − γ wf + >0 ∂2A wm + wf − γdwf + ∂A − ∂A γwf − γ d ∂ A +A ∂ wm ∂ wf ∂ wf =− ∂ wf ∂ d ∂d ∂ wf ∂d , (A7) (A9) ∂2A ∂ d2 wm + wf − γdwf − 2γwf ∂A ∂d The first line, equation (A8), is obtained applying the envelope theorem. 2 A Given that ∂∂wk > 0, k ∈ {m, f } , and ∂ w∂m ∂ A wf > 0, a sufficient condition for ∗ W (.) to be supermodular is for every wm , wf ∈ h, h ¯ 2, where we simplified notations for clarity. A necessary and sufficient condition for ∂ d wm , wf 1 − γd wm , wf − γ wf ≥ 0. (A10) ∂ wf W ∗ wm , wf = A wm , wf , d wm , wf wm + wf − γd wm , wf wf We consider function ∂ 2 A(wm ,wf ,d ) ∂ A(wm ,wf ,d ) ∂ A(wm ,wf ,d ) −γ d ( ∂ ∂ A wm ,wf ,d ) ∂ wf ∂ d wm + wf − γdwf + ∂d − ∂ wf γwf d + A wm , wf , d g γ, d , wm , wf = − . ∂ 2 A(wm ,wf ,d ) ∂ A(wm ,wf ,d ) ∂ d2 wm + wf − γdwf − 2γwf ∂d Since g (.) is well defined over the compact set [0, 1]2 × h, h ¯ , there while making f strictly better off, or ¯ ] g (.) ≤ L . Consequently, there exists a real number L such that sup[0,1]2 ×[h,h ∗ ∗ ¯ > 0 such that () holds for every γ < γ exists γ ¯ , uniformly with respect to θA wm , wf , dmf wm + wf − γdmf wf wm , wf . This concludes the proof of W ∗ (.) supermodularity property. < θA wm , wf , dm f ˜f + D wm + z ˜ f + wf − z ˜ f − γD ˜f − D ˜f . (A12) From equation (A11), we have Consider supporting transfers z ˜ f that leave m indifferent between ˜f , D his equilibrium payoff and the deviation payoff, ∗ A wm , wf , dm f ∗ ˜f = ˜ f + (1 − γ ) D z wm + wf − γdm f wf − wm . A wm , wf , dm f ∗ ∗ A wm , wf , dm f wm + wf − γdm f wf ˜ f + (1 − γ) D First, we can substitute for z ˜ f into equation (A12) and obtain = A wm , wf , dm f ˜ f + (1 − γ ) D wm + z ˜f , (A11) after rearranging (the same way we did for the proof of proposition 1), ∗ ∗ A wm , wf , dmf wm + wf − γdmf wf − A wm , wf , dm f wf + wm − γD ˜f 1 < − 1. (A13) A wm , wf , dm f ∗ ˜ f − A wm , wf , dm wf + wm − γD wm + wf − ∗ γdm wf θA wm , wf , dm f f f ˜ f ≤ 1 − dm f wf , and without loss of Second, limited liability implies z The numerator on the left-hand side of equation (A13) is therefore generality, we can assume that D˜ f ≤ dm f wf ; hence, bounded by ⎡ 1 ⎣ ∗ ∗ dm f wf − wm + wf − γdm f wf A wm , wf , dmf wm + wf − γdmf wf 1−γ ⎤ − A wm , wf , dm f wm + wf − γdm f wf ∗ A wm , wf , dm γ f ∗ ⎦ − A wm , wf , dm f wm + wf − γdm f wf − wm + wf − γdm f wf 1−γ A wm , wf , dm f ∗ ∗ −A wm , wf , dm f wm + wf − γdm f wf ≤D˜f ∗ ≤ A wm , wf , dmf ∗ wm + wf − γdmf wf ≤ dm f wf . (A14) − A wm , wf , dm f wf + wm − γD ˜f ∗ ∗ ≤ A wm , wf , dmf wm + wf − γdmf wf The left-hand side of inequality (A13) is strictly positive. To see that, we look at the numerator and denominator separately. − A wm , wf , dm f wm + wf − γdm f wf . (A15) 918 THE REVIEW OF ECONOMICS AND STATISTICS ∗ Given the optimality of dm f , we have Since m is indifferent between his equilibrium payoffs and the deviation payoffs, ∗ ∗ A wm , wf , dmf wm + wf − γdmf wf ∗ ∗ A wm , wf , dm f wm + wf − γdm f wf − A wm , wf , dm f wm + wf − γdm f wf ∗ ∗ = A wm , wf , dm f ˜ f + (1 − γ) D wm + z ˜f , ≥ A wm , wf , dmf wm + wf − γdmf wf ∗ ∗ − A wm , wf , dm f wm + wf − γdm f wf ˜ f ≤ wf − γdm f wf , we have ˜ f + (1 − γ) D and given that z ∗ ∗ and A wm , wf , dm f wm + wf − γdm f wf ≤ A wm , wf , dm f wm + wf − γdm f wf . A wm , wf , dm f wm + wf − γdm f wf ∗ ∗ We have shown that the left-hand side of equation (A13) is nonnegative. − A wm , wf , dm f wm + wf − γdm f wf To conclude the proof, we remark that optimality implies ∗ ∗ ≤A wm , wf , dm f wm + wf − γdm f wf −A ∗ wm , wf , dm f wm + wf − γdm ∗ f wf . A wm , wf , dm f wm + wf − γdm f wf ∗ ∗ ≤ A wm , wf , dm f wm + wf − γdm f wf , ∗ Lemma 3 states that dmf can be written as a function of wm , wf only and ∗ ∗ ∗ and supermodularity yields the following inequality: the indirect utility W wm , wf = A wm , wf , dmf wm + wf − γdmf wf is supermodular for low enough values of γ. Thus, we have the following ∗ A wm , wf , dmf ∗ wm + wf − γdmf wf inequality: ∗ ∗ − A wm , wf , dm f wm + wf − γdm f wf ∗ ∗ ∗ A wm , wf , dm ∗ wm + wf − γdm ≥A wm , wf , dm wm + wf − γdm f wf f wf f f ∗ ∗ ∗ − A wm , wf , dm ∗ wm + wf − γdm − A wm , wf , dmf wm + wf − γdmf wf . (A17) f f wf ∗ ∗ We therefore have ≤A wm , wf , dmf wm + wf − γdmf wf ∗ ∗ −A ∗ wm , wf , dm f w m + wf − γ ∗ dm f wf . A wm , wf , dmf wm + wf − γdmf wf − A wm , wf , dm f wf + wm − γD˜f γ By transitivity, assuming that 1− γ < 1, the left-hand side term of inequality 1 − 2γ ≥ A wm , wf , dm f wf + wm − γdm f wf (A15) is nonnegative: 1−γ Similarly, the denominator on the left-hand side of equation (A13) can ∗ ∗ be bounded above and below by −A wm , wf , dm f wm + wf − γdm f wf and from inequality (A16) we have A wm , wf , dm f wf + wm − γdm f wf ∗ − A wm , wf , dm ∗ wm + wf − γdm A wm , wf , dm f wf + wm − γD ˜f f f wf ∗ ∗ ≤ A wm , wf , dm f wf + wm − γDf ˜ − A wm , wf , dm f wm + wf − γdm f wf ∗ − A wm , wf , dm ∗ wm + wf − γdm 1 f f wf ≤ A wm , wf , dm f wf + wm − γdm f wf 1 1−γ ≤ A wm , wf , dm f wf + wm − γdm f wf −A wm , wf , dm∗ ∗ wm + wf − γdm 1−γ f f wf , ∗ ∗ −A wm , wf , dm f wm + wf − γdm f wf . (A16) so that the ratio ∗ ∗ A wm , wf , dmf wm + wf − γdmf wf − A wm , wf , dm f wf + wm − γD ˜f ≥ 1 − 2γ. ∗ ∗ A wm , wf , dm f ˜ f − A wm , wf , dm wf + wm − γD f wm + wf − γdm f wf ¯ ,1 ¯ ,h Assuming that γ < 2 1 , the same argument as in proposition 1 applies, {γ ∈ [0, 1] |γ < 1 − 2A( h,h,0) } is nonempty. This in turn yields θ A h ˆ<θ ¯ so that ( ) the set of possible parameters θ verifying equation (A18) uniformly with 1 ≥ 2 (1 − γ ) , (A18) respect to (m , f ) is nonempty. θA wm , wf , dm f By choosing γ ˆ = inf 2 1 ¯ ,h ¯ ,1 , 1 − 2A( h,h,0) , γ A h ¯ has been defined in ¯ , where γ ( ) lemma 3, we have shown that for any γ < γ and any θ ∈ θˆ ¯ , there does ˆ, θ for every θ ≥ 1 ≡ θˆ , so that there is no profitable deviation 2(1−γ)A(h,h,0) not exist a deviation that makes both spouses weakly better off while making from the equilibrium. If condition A h ¯ , 1 < 2A h, h, 0 is satisfied, ¯, h one of the two strictly better off. This concludes the proof of proposition 2.