University of Southern California Dissertations and Theses

Inter-generational tranfsers and bargaining power within the family in South Korea

(USC Thesis Other)

Inter-generational tranfsers and bargaining power within the family in South Korea

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INTER-GENERATIONAL TRANFSERS AND BARGAINING POWER WITHIN THE FAMILY IN SOUTH KOREA by Heonjae Song ________________________________________________________________________ A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ECONOMICS) August 2009 Copyright 2009 Heonjae Song ii Dedication I dedicate my dissertation work to my wife Jiyoung Kim and my two sons Joshua and Sean for being there for me throughout the entire doctorate program. You have been my best cheerleaders. If it had not been for your support, I could not have finished my long journey of doctorate program. iii Acknowledgements I would like to thank John Ham, Maurizio Mazzocco and John Strauss for excellent advice and guidance. Jay Gwon, Bomin Kim, Roger Hyungsik Moon, Jeffery Nugent, Gary Painter, Geert Ridder and Aloysius Siow provided valuable comments. Harounan Kazianga generously provided Stata program. My special thanks go to Hyunsoo Joo. He provided me valuable help with programming. The remaining errors are mine. iv Table of Contents Dedication ii Acknowledgements iii List of Tables vi List of Figures vii Abstract viii Introduction 1 Chapter One: Bargaining Power within the Family in South Korea: Transfers to Parents from Adult Children and Vice-versa 2 1.1 Introduction 2 1.2 Basic Facts for South Korea 4 1.3 Literature Review of Economic Models 9 1.3.1 Intergenerational Transfers 9 1.3.2 Intrahousehold Bargaining 11 1.4 Economic Models of Transfers from Adult Children to Parents and Vice-versa 12 1.4.1 A Model of Bargaining between Adult Children 14 and their Parents 1.4.2 A Dynastic collective model of transfers 17 between adult children and their parents 1.5 Estimation Strategy 19 1.5.1 Estimation of Models in 1.4.1 and 1.4.2 19 1.5.2 Comparison to Kazianga’s (2006) approach 24 1.5.3 Comparison to Lee, Parish, Willis (1994), 25 Lillard & Willis (1997) and Khemani (1999) 1.6 Data and Institutional Background of South Korea 27 1.6.1 Data Description 27 1.6.2 Institutional Background: Public Support for 30 the Elderly in Korea 1.7 Estimation Results 32 1.8 Conclusions and Future Research 38 Chapter Two: Structural Estimation of Bargaining Power in South Korea: Transfers to Parents from Adult Children and Vice-versa 40 2.1 Introduction 40 v 2.2 Economic Models of Transfers from Adult Children to Parents 41 2.2.1 One-way Transfers 41 2.2.2 A Forward Looking Two-way Transfers model 47 2.3 Estimation Strategy 60 2.3.1 Estimation Approach for the Model of One-Way 60 Transfers 2.3.2 Estimation Approach for the Model of Two-Way 69 Transfers 2.4 Estimation Results 75 2.4.1 Results for the Model of One-Way Transfers 75 2.4.2 Results for the Model of Two-Way Transfers 76 2.4.3 Estimation of the Effect of the Education on 78 Bargaining Power 2.5 Conclusions and Future Research 82 Conclusion 83 Bibliography 84 vi List of Tables Table 1: Income Source for the elderly in 1995 (Age ≥60) (%) 7 Table 2: Distribution of Household by Type of Net Transfers (%) 8 Table 3: Transfer amounts for the households with transfers to both parents 9 Table 4: Summary statistics 30 Table 5.1: Regression Estimation of the Rosett’s Friction Model in 33 Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩100,000 (1,2 , ) ijhpwp = = ) Table 5.2: Regression Estimation of the Rosett’s Friction Model in 34 Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩50,000 (1,2 , ) ijhpwp = = ) Table 6: Structural Results for Two-Way Transfers (Model 1.4.1) 35 Table 7.1: Regression Estimation of the Rosett’s Friction Model in 36 Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩100,000 ( 1,2 , ) ijhpwp = = ) Table 7.2: Regression Estimation of the Rosett’s Friction Model in 37 Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩50,000 (1,2 , ) ijhpwp = = ) Table 8: Structural Results for Two-Way Transfers by Education Group 38 (Model 1.4.1) Table 9: Structural Results for Models for the Transfers. 77 Table 10 (a): Structural Results for mu ( ) ˆ μ <Function 1> 79 Table 10 (b): Structural Results for mu ( ) ˆ μ <Function 2> 79 Table 10 (c): Structural Results for mu ( ) ˆ μ <Function 3> 80 vii List of Figures Figure 1: Inter-vivos transfers to and from the elderly (Age ≥50) 5 Figure 2: Transfers to/from married parents and their children by age 6 in the U.S. (Transfers include time, money, and co-residence.) Figure 3: Plot for ˆ μ with One-way Model 81 Figure 4: Plot for ˆ μ with Two-way Model 81 viii Abstract My dissertation extends the literature on intergenerational transfers between adult married children and their parents in two ways. First, for a given couple, both sets of parents enter the optimization problem. Second, I develop and estimate models where amount of transfers to the husband’s (wife’s) parents is assumed to depend on the couple’s income, the husband’s parents’ income, the wife’s parent’s income, and the bargaining power of husband (wife) within the family. Transfers from adult children to parents are quite important in developing countries; for example in Korea in 1995 they made up approximately 60% of the income of those over 60. Further, I argue that it is plausible that each spouse cares more about their parents than their in-laws, and thus such transfers are a form of semi-private consumption, and some have argued that observable semi-private consumption is very useful for looking at bargaining power within the family. In Chapter 1 I consider two models for these transfers. I derive a static collective model to explain couple’s joint decision on these transfers to both sets of parents and estimate this model on data from Korea (2001–2005). These models consider two-way transfers. The first model allows for bargaining between husband and wife, and between each spouse and their parents. To obtain tractable estimating equations I assume that the couple is myopic in the sense that they do not consider potential transfers from parents when dividing their household income. The second model is a dynastic collective model involving the couple and both sets of parents. I find that the data supports the first model ix but not the second, and I find that I cannot reject equal bargaining power hypothesis between husbands and wives. In second Chapter I consider two models for intergenerational transfers and estimate structural parameters in the models. In the first model I derive a static collective model to explain couple’s joint decision on these transfers to both sets of parents. The estimation of this model takes into account that the functional form of the transfer to one set of parents will change if the optimal transfer to the other set of parents is zero. I find that husbands and wives have equal bargaining power in South Korea. The second model extends the first model to allow the couple to be forward looking in the sense that they take potential transfers from parents into account when deriving their sharing rule; this extension comes at the cost of greatly increasing the complexity of the estimation procedure. Using this model I estimate that husbands and wives have equal bargaining power. Overall the results suggest that the husband’s preferences and the wife’s preferences have equal weight. Lastly, I investigate the role of education in determining bargaining power. I suggest that bargaining power is a function of third polynomial of relative education of spouses. 1 Introduction The flow of intergenerational transfers shows different patterns between developed counties versus developing countries. In most developed countries the main flow of transfers is to children from parents while the counter part in many developing countries is the other way around. This different pattern is manly because of public pension system and economic growth. In developed countries, government pensions substitute for, and supplement, support for older family members by younger ones. However in many developing countries, adult children play a major role in their parents’ financial welfare because public pension plans are very recent phenomena, and younger generations are usually richer than older generations because of rapid economic growth. When we analyze transfers from adult children to old parents, we need to note that a married couple usually has two sets of parents to support. If husband and wife have different preferences for their parents and in-laws respectively, they need to do some bargaining. My dissertation uses models of family bargaining to extend the literature on intergenerational transfers between adult married children and their parents. This dissertation consists of two chapters. In first chapter I develop two models to explain transfers between a married couple and their both sets of parents. I mainly focus on testing models with data. In the second chapter I also develop two models and focus on estimating structural parameters in the models. In the first model I consider budget constraint faced by a couple in the estimation. In the second model I consider that a couple may expect potential transfers from parents. In both models I investigate the role of education in determining bargaining power. 2 Chapter One Bargaining Power within the Family in South Korea: Transfers to Parents from Adult Children and Vice-versa 1.1 Introduction Private intergenerational transfers have been extensively studied by economists. The altruism model made famous by Barro (1974) and Becker (1974, 1991), and the exchange model introduced by Cox (1987), are examples of theoretical models that explain intergenerational transfer behaviors. In more recent work they have been analyzed in a vast literature that included a series of papers by, for example, Altonji et al (1997), Lundberg et al (1996), Duflo (2003) and Thomas (1994). In the U.S. and other developed European countries, market institutions and government pensions substitute for, and supplement, support for older family members by younger ones. As a result most work on transfers has focused on transfers from parents to children. At the same time there have been a series of papers by Chiappori (1988, 1992) and by Mazzocco (2006, 2007) on bargaining within the family. This paper uses models of family bargaining to extend the literature on intergenerational transfers between adult married children and their parents in two ways. First, for a given couple, both sets of parents enter the optimization problem. Second, I develop and estimate models where amount of transfers to the husband’s (wife’s) parents is assumed to depend on the couple’s income, the husband’s parents’ income, the wife’s parents’ income, and the bargaining power of husband (wife) within the family. The motivation for my work is threefold. First, transfers from children to parents are very 3 important in determining the parents’ income in developing countries. Second, I argue that it is plausible that each spouse cares more about their parents than their in-laws, and thus such transfers are a form of semi-private consumption, and Behrman and Rosenzweig (2006) have argued that observable semi-private consumption is very useful for looking at bargaining power within the family. Third, understanding upstream intergenerational transfers are important to help policy makers in developing countries design better policies toward the low income elderly who are not covered by recently introduced pension systems. In this paper I consider two models for these transfers. I derive static collective models to explain couple’s joint decision on these transfers to both sets of parents and estimate these models on data from Korea (2001–2005). The first model allows for bargaining not only between husband and wife but also between each spouse and their parents. To obtain tractable estimating equations I assume that the couple is myopic in the sense that they do not consider potential transfers from parents when dividing their household income. The second model is a dynastic collective model involving the couple and both sets of parents. I find that the first model of transfers fits the data while the second does not. Using the first model I find that I cannot reject equal bargaining power hypothesis between husbands and wives. The chapter is organized as follows. In section 1.2, I present some basic stylized facts on intergenerational transfers in Korea. In section 1.3, the existing literature on the intergenerational transfers focusing on developing countries and the some of the large and growing literature on family bargaining is reviewed. In section 1.4 I present four 4 theoretical models of transfers. In section 1.5 I discuss the respective estimation strategies for the models and how my econometric approach compares to recent work by Kazianga (2006). I then discuss early work on transfers to parents by Lee, Parish, and Willis (1994), Lillard & Willis (1997) and Khemani (1999). Section 1.6 discusses institutional features in Korea and the data used here. In section 1.7 I present estimation results for each model. Section 8 concludes the chapter. 1.2 Basic Facts for South Korea In this section I provide some basic stylized facts for South Korea on transfers between adult children and their parents; see Cox and Fafchamps (2008) for a discussion about these transfers in other developing countries. To show the importance of transfers from children to parents in Korea as compared to many Western economies consider Figure 1, which shows how inter-vivos transfers involving the elderly in South Korea are different from ten Western countries. Interestingly, from Figure 1 one sees that Korean parents are 50% more likely to receive a net transfer from their children than to provide their children with a net transfer. On the other hand, in nine of the ten Western countries, children are five times more likely to receive a net transfer from their parents than to give one to their parents. In the remaining country, Spain, children are likely as twice as to receive a net transfer from their parents than vice-versa. Further, on average children in the ten developed countries are likely as five times as to receive a net transfer from their parents than to give one. 5 0% 5% 10% 15% 20% 25% 30% 35% 40% Austrailia Germany Sweden Netherland Spain Italy France Denmark Greece Swiss Total Korea receiving from family giving to family Source: Study of Health, Ageing and Retirement in Europe (SHARE 2004) and Korean Longitudinal Study of Ageing (KLoSA 2006). Figure 1: Inter-vivos transfers to and from the elderly (Age ≥ 50) Figure 2 presents the overall pattern of transfers between elderly people in the U.S. and their children in 2002, that is, whether there are any exchanges and, if so, in which direction they flow. It demonstrates that down stream transfers dominates the direction of the flow. For example, 38% of those ages 65-79 give to their children but do not receive anything from the children, while only 3% of them report that they do not give to the children but receive transfers from them. 6 0% 10% 20% 30% 40% 50% 60% Neither Give Nor Receive Receive Only Give Only Give and Receive Age 64 and Under Age 65-79 Age 80+ Source: The Health and Retirement Study (HRS 2002). Figure 2: Transfers to/from married parents and their children by age in the U.S. (Transfers include time, money, and co-residence.) Table 1 shows the importance of transfers from their children in the total income of parents in Korea, as compared to the experience in Japan, the US, and Germany. The difference between Korea and these other countries is dramatic: transfers from children make up over half the total income of elderly Koreans, while these transfer constitute less than ten percent of this income of the elderly in the other three countries. 7 Table 1: Income Source for the elderly in 1995 (Age ≥ 60) (%) Income source Korea Japan U.S. Germany Labor income 26.6 21.6 15.5 4.6 Financial income 9.9 6.6 23.3 13.7 Private transfer 56.6 6.6 1.6 1.9 Public pension 6.6 57.4 55.8 77.6 Source: Seok & Kim (2000). Korea Institute for Health and Social Affairs Table 2 shows the percentage distribution of adult children across different transfer behaviors towards the parents for the years 2001-2005 in the Korea Labor and Income Panel Study (KLIPS). From this we see that approximately 57% of families make net transfers to at least one parent, while only approximately 21% receive a net transfer from their parents. Further, approximately 14% of the households give only to the husband’s parents while only 3% give only to the wife’s parents. 8 Table 2: Percentage Distribution of Household by Type of Net Transfers (%) Year To Both sets of parents Only to Husband’s parents Only to Wife’s parents To Neither parents Receive Net Transfer from parents 2001 37 16 3 24 20 2002 38 18 3 20 19 2003 37 12 3 24 24 2004 45 15 2 18 20 2005 44 11 3 22 20 Total 40 14 3 22 21 Source: Calculated by the author using KLIPS (2001-2005). Note: First-born sons and head’s age over 40 excluded. Table 3 indicates that when a couple makes a transfer to both sets of parents, the transfer to the husband’s parents is 50% more than the transfer to the wife’s parents. The last column of Table 3 shows how much of couple’s household income is allocated toward financial transfers to both sets of parents is calculated. It is approximately 6%. The stake is high enough to consider decision of transfers to each set of parents as bargaining outcome between husband and wife. While these results cannot be considered definitive, these certainly raise the possibility that husbands have greater bargaining power, thus motivating the derivation and estimation of the models below. Further, as shown in the last column of the Table 2 the fact that 21% of couples receive a net transfer from their parents suggests the need to also consider models to allow transfers from parents. 9 Table 3: Transfer amounts for the households with transfers to both parents (1) (2) (3) (4) Transfer to Transfer to Couple's Year N Husband's parents Wife's parents Household income Ratio* 2001 127 102.66 61.67 3025.53 6.39% (10.88) (5.69) (127.26) (0.0074) 2002 131 130.99 86.94 3581.42 6.29% (14.79) (12.86) (180.53) (0.0054) 2003 147 117.17 71.67 3518.00 6.14% (9.56) (6.10) (153.15) (0.0049) 2004 166 118.45 66.14 3859.79 5.04% (11.40) (5.30) (168.22) (0.0033) 2005 158 124.44 87.47 3685.18 5.65% (13.02) (13.58) (146.23) (0.0042) Total 729 118.99 74.84 3557.66 5.85% (7.06) (5.07) (105.11) (0.0022) Source: Calculated by the author using KLIPS (2001-2005). Notes: (1) First-born sons and head’s age over 40 excluded. (2) Robust Standard errors are in parentheses. (3) Transfer amount is measured in tens of thousands of Korean Won ( ₩). ₩10,000 is approximately U$10 in 2004. * Ratio: (Column 1 + Column 2) / Column 3. 1.3 Literature Review of Economic Models 1.3.1 Intergenerational Transfers Cox and Fafchamps (2008) made a thorough review of the literature on the intergenerational transfers. The intergenerational transfers in developing countries are likely to focus on old age support because social security consists of private old-age 10 support from adult children. Ravallion and Deardon (1988) estimated transfer equations with Indonesian data and found significant targeting on the elderly people. More recently Cox, Galasso, and Jimenez (2006) studied private inter-household transfers in a diverse cross section of developing countries for which nationally representative surveys for Albania, Bulgaria, Jamaica, Kazakhstan, the Kyrgyz Republic, Nepal, Nicaragua, Panama, Peru, Russia and Vietnam. They find that transfers from young to old are greater than those going from old to young in both the Latin American countries in their sample and in Vietnam and Nepal as well, whereas the opposite is true for Russia and Bulgaria. Since Cox (1987) introduced exchange model, there has been an issue on the motives on the upstream intergenerational transfers whether these transfers are altruistically motivated or whether the elderly receive reward for the service they provided to the children. Raut and Tran (2005) proposed two alternative models of intergenerational transfers linking parental investment in human capital of children to old-age support. The first model formulates these transfers as a pure loan contract and the second model as self-enforcing two sided altruism. Interestingly in the second model they developed a Nash equilibrium concept and found that parents and children are altruistic in a manner consistent with the second model. Another empirical issue is the test of “crowding out” effect. Altruism model predicts that government income redistribute program will be ineffective by adjustments in private intergenerational transfers. On the other hand, exchange model can prevent the crowding out. Cox and Fafchamps (2008) said numerous studies do suggest partial crowding out, on the order of a 20 to 30 cent reduction in private transfers per dollar 11 increase in public transfers. However, the range of estimated effects is exceedingly wide, with many studies suggesting little private transfer response at all. Kazianga (2006) thoroughly studies possible explanations for the weak transfer response found in numerous empirical studies after taking a careful econometric approach that inquires about a variety of estimation issues at once, including selection bias by making use of the Altonji–Ichimura-Otsu estimator, potential endogeneity of income, and non-linearities in income effects. These papers, as well as the work for the United States by Altonji et al (1997), focus on testing the implications of altruism in estimating equations which are based on theory but do not directly allow for recovering structural parameters. They do not allow for bargaining within the family, although Altonji et al (1993) suggested that a bargaining model could be useful in analyzing these transfers. Finally, none of these studies consider transfers between an adult couple and both sets of parents. I address all three of these issues below. 1.3.2 Intrahousehold Bargaining In this section I give a brief overview of some of the papers in the large and growing literature on household bargaining; see Xu (2007) for a more thorough review. Manser and Brown (1980), and McElroy and Horney (1981) characterized the household as a group of agents making joint decisions. In these papers the household decision process is modeled as a Nash bargaining problem. Chiappori (1988; 1992) extended their analysis to allow for any type of efficient decision process by developing the static 12 collective model. This model has been extensively studied, tested, and estimated in the literature, and numerous empirical papers have shown that the distribution of bargaining power among parents is important to their children’s human capital investment decisions. (See, e.g., Thomas, Contreras, and Frankenberg 2002 and Rubalcava and Thomas 2000). Further, Blundell, Chiappori and Meghir (2005) extended this collective model to allow for the existence of public consumption (which is interpreted as children’s consumption). Mazzocco (2006) extended the Blundell, Chiappori and Meghir (2005) approach by developing a dynamic collective model. He used this model to recover parents’ preferences for expenditure on children using variables available in commonly used datasets when at least one parent works. 1 There have been relatively few studies to consider bargaining with respect to upstream intergenerational transfers. This study is to extend the collective model to this area. 1.4 Economic Models of Transfers from Adult Children to Parents and Vice-versa Here I consider two models of transfers between a couple and their parents. For simplicity I assume here that the parents have only one adult child and there is no difference in the expected role of sons and daughters towards their parent. Further I assume that both of the husband’s parents and both of the wife’s parents are alive and live together. Finally I assume that old parents do not provide any services to their adult 1 Two other papers in this literature are Brown (2008), who studied the positive relationship between dowries and women’s welfare, and Schoeni (2000) who examined the case where altruistic parents and parents-in-law make transfers to their adult children. 13 children in exchange for the transfers. Hence, the exchange motive is left for the future research. Each spouse has his/her own consumption (not observed in my data) and is assumed to care about only his or her own parents’ utility. 2 Each spouse treats their parents collectively. That is, only total consumption of his/her parents matters and how surviving parents allocate the transfers from their child does not affect the couple’s transfer decisions to their parents. Given these assumptions, each partner’s parent(s) is treated hereafter as one entity. Further, I assume that children’s joint utility function does not contain a public good. 3 To make this problem more tractable, I employ an additively separable logarithmic utility function for each spouse’s utility and introduce an altruism parameter i α ) , ( w h i = for the strength of each child’s altruistic feelings toward their parents’ consumption. Further I assume that husband and wife care about their respective parents equally ( hw α α = ). 4 Each parent’s altruism toward their child is given by (, ) i ihpwp β = , where for simplicity I assume that β is the same for sons and daughters. Note that one may want to adjust the adult children’s respective consumption by an equivalence scale reflecting the number of children they have, or adjust their parents’ consumption by an equivalence scale reflecting the number of surviving parents. 2 Since I will assume that the parameter determining altruism from the wife to her parents and the parameter determining altruism from the husband to his parents are equal, allowing for caring in the sense of Becker (1981) will not change my results. 3 Adding a public consumption which is separable to transfer to parents (for example, couple’s own children’s consumption) does not change the result. 4 This assumption can be supported by the assortive mating hypothesis which predicts that prospective spouses are sorted by similar characteristics in the marriage market. I will explore relaxing it in future work. 14 However, this will not affect the optimization problem given the choice of logarithmic utility. 5 From section 1.2 and Table 2, it is clear that a non-trivial fraction of adult children receive net transfers from their parents. Therefore, we need to allow for two-way transfers (transfers to/from parents). Once we introduce transfers from the parents, it is unreasonable to assume that the children always control the decision making. For example, if the couple receives net transfers from their parents, it is more plausible to assume that altruistic parents play a role in determining such transfers. I propose two models which allow both directions of transfers to/from parents. 1.4.1 A Model of Bargaining between Adult Children and their Parents It is plausible to allow for bargaining between the adult children and their parents as Altonji et al (1993) suggest, where the final transfers depend on each child’s intra- household bargaining power as well as inter-household bargaining power. I use a two- step transfer collective model. At the first step, the husband and wife bargain only over the division of household income between them, and their shares are determined by their respective bargaining power. In the second step each child bargains with his parents and after this consumption takes place. The first step for the children is to solve for i ρ ) , ( w h i = which denotes each spouse’s monetary share of the family’s joint income Y . They do this through the optimization problem 5 In future I will consider other functional forms for the utility function that will not have this property. 15 () , ln 1 ln . . , hw hh h w hw Max st Y ρρ μ ρμ ρ ρρ +− += (1) where h μ is the husband’s bargaining power. The solution for i ρ ) , ( w h i = is not surprisingly () 1 hh wh Y Y ρμ ρμ = =− , . (2) Note that the children are myopic in the sense that they do not take into account transfers from the parents in determining their sharing rule. At the second step, each spouse and his/her parents bargain over total all available resources (each parents’ income and each spouse’s share of the couple’s income). It is further assumed that the husband and his parents pool their income, as do the wife and her parents. With this assumption we can write the husband and his parents’ problem as () ( ) ( ) , ln ln 1 ln ln .. , , hhp hp hp h hp h hp CC hp h hp h hp hp hp Max C C C C st C C Y Y TC Y μβ μ α μ ++− + += + =− (3) where hp μ is husband’s parents’ bargaining power over their son, h C is the husband’s private consumption, hp C is the total consumption of the husband’s parents, hp Y 16 denotes the husband’s parents’ before-transfer income and hp T denotes the transfers made to the husband’s parents. Note that hp T can be greater than zero (the net transfer to the parents is positive) or less than or equal to zero (the net transfer to the parents is zero or negative). The final transfer is determined by the respective bargaining power between husband and his parents and their respective altruism towards each other. It is straight forward to show that: () () () () () 1 11 . 11 hp h hp hp hp hp hp TY Y αμ αμ βμ βα μ α βα μ α −+ −− =− −++ −++ (4) Note that the transfer to/from the husband’s parents does not depend on the income of the wife’s parents in this model. The same analysis applies to the wife and her parents, who carry out the following optimization () ( ) ( ) () , ln ln 1 ln ln .. 1 , , wwp wp wp w wp w wp CC wp w wp h wp wp wp Max C C C C st C C Y Y TC Y μβ μ α μ ++− + += + − =− (5) where wp μ is wife’s parents’ bargaining power over their daughter, w C is the wife’s private consumption, wp C is the total consumption of the wife’s parents, wp Y denotes 17 the wife’s parents’ before-transfer income and wp T denotes the transfers made to the wife’s parents. Again the transfer can be negative or positive, and is equal to () ()() () () () 11 11 . 11 wp h wp wp wp wp wp TY Y αμ α μ βμ βαμα βαμα −+ − −− =− −++ −++ (6) Again note that the transfer to/from the wife’s parents does not depend on the income of the husband’s parents. The exclusion restrictions for (4) and (6) will be used to test the model. Finally, since this is a collective model, the outcome will be Pareto efficient (see Chiappori 1992). 1.4.2 A Dynastic collective model of transfers between adult children and their parents I now consider collective bargaining among the adult couple, the husband’s parents and the wife’s parents, over the total resources available ( hp wp YY Y ++ ). Each pair of married child and his/her parents care about each other, but children do not care about their in-laws and parents do not care about their children’s spouses. In this model h μ denotes the husband’s bargaining power with regard to his wife, hp μ denotes the husband’s parents’ relative bargaining power over the married couple and the wife’s parents, and c μ denotes the couple’s bargaining power over the both sets of parents. Thus the bargaining power of the wife’s parents over the couple and the husband’s parents is 1 hp c μ μ −− . The couple and two sets of parents solve 18 () ( ) ( )( ) ( ) ()( ) ,, , ln ln ln ln 1 ln ln 1 ln ln . . , = , = . hw hp wp hp hp h c h h hp h w wp CC C C hp c wp w hw hp wp hp wp hp hp hp wp wp wp Max C C C C C C CC st C C C C Y Y Y TC Y TC Y μβ μμ α μ α μμ β ++ + +− + +− − + ++ + = + + − − (7) The optimal transfers are given by () () 1 hp c h hp hp wp hp c TYYYY μαμμ βα βμ + =++− +− + , (8) () ( ) () () 11 1 hp c c h wp hp wp wp c TYYYY μμ αμ μ βα βμ −− + − =++− +− + . (9) Equations (8) and (9) show that the income of the wife’s parents affects the transfer to the husband’s parents and Vice-versa; recall that the income of the wife’s parents does not affect the transfer to the husband’s parents (and vice-versa) in the model immediately above in section 1.4.1. Thus if there is no effect of the in-law’s income on the transfer to own parents, this would cast doubt on the dynastic collective model, while if there is an effect this will cast doubt on the model in 1.4.1. 19 1.5 Estimation Strategy 1.5.1 Estimation of Models in 1.4.1 and 1.4.2 A number of approaches can be used to estimate the transfer function in both of these models. First, one can employ least squares approach, interpreting the transfer equation as a projection and ignoring the fact that there is considerable bunching of transfers at zero that is difficult for a regression model to handle. Recall that in Table 2 in section 1.2 21% of couple’s households report that they neither make nor receive transfers to/from neither set of parents. Transaction costs associated with transfers are introduced to explain these non-participant households. It would imply that positive transfers are observed only when latent transfers exceed the transaction costs. Following Udry (1994) and Kazianga (2006), I use a Rosett’s friction model (Rosett, 1959) to take this bunching into account. () 11 21 22 * if * ; 0 if < * ; , * if * jj j j it it jjjj it it jj j j it it TK T K TKTKjhpwp TK T K ⎧−> ⎪ ⎪ =<= ⎨ ⎪ ⎪+< ⎩ (10) where * j it T is latent transfer, 1 j K and 2 j K are unobserved transaction costs, and j it T denotes the actual transfer functions as defined in (4), (6), (8) and (9). The latent transfers must be greater than the transaction costs for one to observe any transfer. 20 A natural starting point is to consistently estimate the following equations describing transfers between the couple and the husband’s parents, and the transfers between the couple and the husband’s parents: 11 12 13 1 21 22 23 2 , . hp hp wp it it it it it wp hp wp it it it it it TY Y Y e TY Y Y e ππ π ππ π =+ + + =+ + + (11) The models in 1.4.1 and in 1.4.2 place restrictions on these equations. The parameter restrictions for model 1.4.1 are () () () () () 11 12 13 1 11 ;;0. 11 hp h hp hp hp αμ αμ βμ ππ π βα μ α βα μ α −+ −− ==− = −++ −++ (12) () ()() () () () 21 21 23 11 11 ;0; . 11 wp h wp wp wp αμ α μ βμ πππ β αμ α β αμ α −+ − −− ===− −++ −++ (13) In considering this model, I first test 013 22 :0, 0. H π π = = If I do not reject this null hypothesis, I then solve for the structural parameters from the reduced form parameters. However, there are only four reduced form parameters while there are five structural parameters in this model: children’s altruism parameter ( α ), parent’s altruism parameter ( β ), husband’s relative power over wife ( h μ ), husband’s parents’ relative power over son ( hp μ ), wife’s parents’ relative power over daughter ( wp μ ). Therefore, the 21 structural parameters are not identified without additional assumptions. To address this, I assume hp wp μ μ = , which means the parents have same degree of bargaining power over male and female children in this model. 6 Given this parameter restriction which is imposed in this model, only h μ is identified and it is straight forward to show that h μ is recovered as follows. () 11 11 21 . h π μ ππ = + The parameter restrictions for model 1.4.2 are () () () 11 12 13 1 11 1 hp c h hp c h cc hp c h c μαμμ μαμμ ππ β α βμ β α βμ μαμμ π βα βμ ++ == − +− + +− + + = +− + ;; . (14) () ( ) () ( ) ( ) () () ( ) () 21 22 23 11 11 11 11 1 1 hp c c h hp c c h cc hp c c h c μμ αμ μ μ μ αμ μ ππ β α βμ β α βμ μμ αμ μ π βα βμ − − +− − − +− == +− + +− + −− + − =− +− + ;; . (15) 6 This assumption implies 12 23 π π = . 22 Note that this model implies 13 22 0, 0. π π ≠ ≠ Thus, if I cannot reject 013 22 :0, 0 H π π == , I no longer consider model 1.4.2. If I do reject 0 H , I then test a second implication of model 1.4.2. 111 13 21 22 :, . H π ππ π = = If I do not reject 1 H (this null hypothesis), I then solve for the structural parameters for model 1.4.2 from the reduced form parameters. 7 However, estimating the reduced form transfer functions in (11), and obtaining standard errors for these estimates is not trivial, since the parents’ incomes are not observed. Instead I run following imputation regressions in (16) from another data set KLoSA 8 based on explanatory variables that I observe in both data sets: , , hp it hp it hpit wp it wp it wpit YZu YZu δ δ =+ =+ (16) Note that at least two of the variables in it Z must be excluded from (16) for the model to be identified, but this is not problematic since the variables in it Z primarily refer to the parents. Further, one may worry that the family’s income is endogenous, since if there is a shock to their desired transfers to their parents, the husband and wife may work harder. In 7 There are only two reduced form parameters while there are five structural parameters in this model. Hence, it is implausible to solve for structural parameters in this model without further assumptions. 8 Detail explanation about data used in this paper is presented in section 1.6. 23 this case transfers are causing income while I want the casual effect of family income on transfers. Thus I run a reduced form equation for family income 12 3 ˆˆ hp wp it it it it it YX Y Y ϕ ϕϕ ε =+ + + (17) where it X is an excluded instrument from (17) which identifies the model; following Kazianga I use family net assets as the excluded instrument; this is not an ideal exclusion restriction since one could argue that assets should also affect transfers, hence I also use husband and wife’s education as excluded instruments. I then plug predicted values from (16) and (17) into the transfer functions given by (11) * 11 12 13 1 * 21 22 23 2 ˆˆ ˆ , ˆˆ ˆ . hp hp wp it it it it it wp hp wp it it it it it TY Y Y e TY Y Y e ππ π ππ π =+ + + =+ + + (18) I then maximize the period by period likelihood function for (18) conditional on the predicted values ˆ it Y , ˆ hp it Y and ˆ wp it Y . It should be straight forward to establish the consistency of these estimates using arguments similar to those in Amemiya (1979). However, getting the standard errors analytically is difficult, so I use the bootstrap with 500 replications to obtain standard errors. Each bootstrap replication involves 1. Choosing a new bootstrap cross-section sample in KLoSA and 2. Estimate both parents’ income equations from the KLoSA replication sample. 24 3. Choose a new bootstrap sample of family histories from KLIPS, i.e. sample by families, not by family year observations. 4. Estimate a first stage equation for family income using the new KLIPS replication sample pooling the data by family and year. . 5. Estimate (18) for both the transfers to the husband’s parents and to the wife’s parents by forming a quasi-likelihood consisting of family-year contributions. Store the reduced form parameters in a vector. 6. Repeat 1-5 500 times so that each parameter has 500 bootstrap observations. Get standard errors by taking the standard deviation for the bootstrap observations for each parameter. Obtain covariance between parameters by taking the covariance in the 500 bootstrap replications for both parameters. 1.5.2 Comparison to Kazianga’s (2006) approach Numerous empirical studies have estimated income-transfer patterns but because they use a variety of approaches the results are quite broad. Kazianga (2006), in a very careful empirical study on income transfers, addressed a number of estimation issues in this context. First, as noted above, he used Rosset’s friction model. Second, he allows family income to be endogenous, using family assets as an instrument for permanent income and rainfall for transitory income. I also use this procedure with family assets as the excluded instrument; however I also correct the standard errors for this imputation procedure. 9 9 Kazianga appears to have substituted a predicted value of income in, which will produce consistent parameter estimates but inconsistent standard errors. 25 Kazianga also allows for a very flexible response of transfers to income by considering a spline function in income in the transfer equation. Using a spline function or a polynomial in incomes is straight forward if income can be considered exogenous. If one treats income as endogenous it is better to use the actual values of income in the polynomial and then exploit normality to deal with the endogeneity, analogous to the procedure in Blundell and Smith (1986). I will consider this in future work. Finally Kazianga allows for nonseparable functions for transfers between incomes and the unobservables using the approach in Altonji, Ichimura and Otsu when he analyzes one-way transfers (allowing only for positive transfers); the Altonji et al procedure cannot be used for two-way transfers (allowing both for positive and negative transfers). I should note that I deal with missing parents’ income while Kazianga simply considers biases from omitting it. Moreover I would argue that my approach has a closer to link to theory than his. Finally, and perhaps most importantly, I also allow for a role for both sets of parents while he does not do this. In summary, this paper deals with different issues and thus is a compliment, rather than a substitute, for his important paper. 1.5.3 Comparison to Lee, Parish, Willis (1994), Lillard & Willis (1997) and Khemani (1999) There have been three important papers on transfers from children to parents with bargaining approach in developing countries. Lee, Parish, and Willis (1994, hereafter LPW) were the first to address bargaining power in the children’s families when analyzing upstream transfers. Using data from the 1989 Taiwan Family and Women 26 Survey, they found that wives who earned more income provided more support to their own parents. Lillard & Willis (1997, hereafter LW) also found that the amount being transferred to the wife’s parents depends more strongly on the wife’s income than on the husband’s income, and vice-versa for the size of transfers to the husband’s parents. However, these papers did not focus on family bargaining nor did they consider formal models or estimate structural parameters for the process determining transfers from adult children to parents. Khemai (1999) focused on bargaining model and found that the distribution of assets between husbands and wives affects the likelihood of transfers to their origin families using Indonesia Data. She derived latent variables that determine whether transfers are made to the parents of the husband and the wife respectively from the bargaining model and reported reduced form probit estimates. However, she did not consider the actual amounts to be transferred to the each set of parents in her estimation analysis. The respective bargaining power of each spouse is likely to affect not only the probability of transfers made to the parents of the husband and the wife respectively but also the actual amounts transferred. Hence, her findings from the probit analysis may not be sufficient to support her argument. In this paper I extend their seminal analysis in many directions. First, I use formal bargaining models to derive my estimating equations. Second, I used the parents’ characteristics and a second data set to impute the parents’ income, while LPW and Khemani only use parents’ characteristics as control variables. Third, in LPW and Khemani positive (net) transfers from parents to children are treated as zero transfers, while I also develop and estimate a model of two-way transfers which allows transfers 27 from parents at the same time. Finally, LPW, LW and Khemai ignore the role of tradition in upstream transfers while I allow for first-born husbands to differ in their transfer behavior, since they have traditional duties to take care of the parents. I also allow older husbands and wives to have different structural parameters since they may be more affected by tradition than their younger counterparts. 1.6 Data and Institutional Background of South Korea 1.6.1 Data Description This paper uses data of the “Korea Labor and Income Panel Study” (KLIPS) which is administrated by the Korea Labor Institute (KLI). I briefly introduce the data and emphasize the unique feature of KLIPS to be exploited in this paper. The KLIPS is a longitudinal study of a representative sample of Korean households and individuals living in urban areas and conducted annually to track the characteristics of households as well as the economic activities, labor movement, income, expenditures, education, job training, and social activities of individuals starting from year of 1998. Especially important is the fact that this panel data contains information on financial exchange with parent(s) from 4 th wave (2001) on. Specifically a household is asked whether head’s parents who do not co-reside still survive and who they are, and how much of financial support to and from the head’s parents was made last year. The same questions are asked about spouse’s parents. These financial exchanges with parents are, of course, the focus of interest here. The parents’ income is crucial information in my models. Unfortunately, KLIPS does not have parents’ income. Instead, we can obtain the level of parents’ education of 28 both spouses, as well as the children’s birth order. The KLI has created another panel survey on the middle/elderly population (45 or older) in South Korea: “Korean Longitudinal Study of Ageing” (KLoSA) 10 starting in 2006. The KLoSA contains elderly people’s detailed demographic information such as education, marital status as well as income. I impute parents’ income by using parents’ education and widowed status, and children’s age, education and birth order which are common in both KLIPS and KLoSA, and use the procedure described in section 1.5.1. 11 I pool the waves (2001-2005) and all transfer amounts and incomes are in real (2004) values. Note that KLoSA is not superior to KLIPS for my problem since KLoSA does not have information on the income of the adult couple or on the income of the in-laws. In future work I could consider dynamic models by exploiting the panel nature of KLIPS. Greenhalgh (1985, p.265) states that “Traditional Confucian China and its cultural offshoots, Japan and Korea, evolved some of the most patriarchal family systems that ever existed.” It is fair to say that elderly persons depend on their adult sons (especially first son) for old-age support in the East Asian traditional family system affected by Confucianism. On the other hand, it is also fair to say that the Confucian patriarchal family system is no longer valid to all families in modern Korean societies; many changes have occurred to the Korean family structure partly as a result of the increasing 10 KLoSA is designed to provide basic data on population ageing in Korea for policy making and cross disciplinary studies. The survey deals with social, economic, physical, and mental aspects of life. See http://klosa.kli.re.kr for more detail. 11 To be eligible to be included in the imputation regression, elderly persons in KLoSA should have at least one married child who dose not live together. 29 employment of women and the decreasing gender inequality in socioeconomic status. 12 For example, gender difference has been substantially reduced in years of schooling over time. However, it is probably safe to say that i) patriarchal family systems still work in older generations and ii) the first son has usually greater responsibility to support the parents. I will deal with this by allowing different behavior from first sons and from families over 40. For this version of the paper I only include younger households whose head’s age is less than or equal to 40 and where the husband is not a first son. Table 4 presents some summary statistics of the samples which are used in estimation. 13 It shows that the wife’s parents are richer than the husband’s parents. This arises because husbands are older than their wives on average, so the husband’s parents are older than the wife’s parents and less educated. 12 I rely on Xie and Zhu (2006) for this characterization. 13 When estimating upstream transfer model, if total transfers made by household is greater than household income, those observations are excluded from the estimation. Further, if transfer made by either set of parents is greater than the parents’ imputed income, those households are additionally ruled out from the estimation of two-way model. 30 Table 4: Summary statistics Variables Transfer to husband's parents 73.18 (4.16) Transfer to wife's parents 40.33 (4.27) Transfer from husband's parents 16.66 (2.07) Transfer from wife's parents 16.13 (2.12) Couple's household income 3136 (68.55) Husband's parents' imputed income 1321 (33.53) Wife's parents' imputed income 1621 (36.36) Husband’s education 13.68 (.0996) Wife’s education 13.12 (.0850) Observation 1800 Notes: (1) First-born sons and head’s age over 40 excluded. (2) Zero transfers are included. (3) Transfer amount is measured in tens of thousands of Korean Won ( ₩). ₩10,000 is approximately U$10 in 2004. 1.6.2 Institutional Background: Public Support for the Elderly in Korea As I show in section 1.2 and Table 1, adult children play a major role in their parents’ financial welfare in South Korea because public pension plans are very recent 31 phenomena. The compulsory coverage of social security system had not been extended to all residents until 1999. 14 In addition to the National Pension Program various types of assistance under the National Basic Livelihood Security System are currently provided to low-income citizens who meet the criteria in South Korea. To be eligible to be a recipient of government support citizens should show that their imputed total income 15 is lower than the minimum living cost as defined by the government guideline. A certain level of financial support from children is assumed to take place and is included in imputed total income. That is, under Korean law there is a legal family responsibility that obligates adult children to support for their parents, and the government assumes that children provide a certain level of such support regardless of the amount actually transferred by the children. Hence, even though children do not provide any transfers, low income elderly citizens can be excluded from the public assistance program if their children are presumed to be capable of support. Those responsible for financial support include the married daughter and her husband; daughters have same degree of responsibility as their male siblings under Korean law. Note that if men have considerable bargaining power and limit the families’ contribution to their wives’ parents, the public safety net may be inadequate for the wives’ parents. 14 The National Pension Act came into effect in January 1988 in Korea. It covered only those who were working in firms with more than 10 full-time employees. The National Pension has extended coverage to workplaces with more than 5 full-time employees (January 1992), and farmers and fishermen (July 1995) and April 1999, the National Pension Program extended compulsory coverage to all residents aged 18 to 60 in Korea. The number of insured persons increased from about 6.5 million in 1998 to about 16 million in 1999. 15 Imputed total income consists of actual income and appraised income from assets. 32 1.7 Estimation Results I estimated consistent parameters and standard errors for the reduced form conditional on the predicted values in (18), which underlies the model in 1.4.1 and 1.4.2. I use a sample of non-first sons under 40 years. Instead of estimating 1 j K and 2 j K (,) jhpwp = which are unobserved transaction costs in the Rosett’s Friction Model in (10) in section 1.5.1 16 , I assume that there is a fixed minimum transaction cost of ₩100,000 in any transfers. That is, I assume j i K = ₩100,000 (1,2 , ) ijhpwp == . 17 Columns 1 and 2 of Table 5.1 reports estimation results for the Rosett’s friction model under this fixed cost assumption. First, note both the coefficient on the wife’s parents’ income in the transfer equation to the husband’s parents, and the coefficient on the husband’s parents income in the transfers equation to the wife’s parents, are very insignificant, casting serious doubt on the model in section 1.4.2. The model in 1.4.1 predicts i) both the coefficient on the wife’s parents’ income in the transfer equation to the husband’s parents and the coefficient on the husband’s parents income in the transfers equation to the wife’s parents are zero, and ii) the coefficient on the husband’s parent’s income in their transfer function should equal that on the wife’s income in their transfer equation. As noted above i) is clearly satisfied by the data. Thus I have re-estimated the transfer equations with this constraint imposed, and placed the results in columns 3 and 4 16 I estimated 1 j K and 2 j K (,) jhpwp = and found that the estimated friction coefficients are much bigger in absolute value than many transfers, which is not consistent with a fixed cost interpretation. 17 ₩100,000 is approximately U$100 in 2004. The choice of benchmark cost K is arbitrary but it reflects a social sanction in South Korea. Koreans have some standard expenses for congratulations and condolences. U$100 works as widely acceptable amount. If reported transfers are less than ₩100,000, I recoded those as 0’s and estimated the transfer equations. 33 of Table 5.1. In both columns 1 and 2, and in columns 3 and 4, prediction ii) is satisfied. For either set of results the coefficient estimates are very close and, not surprisingly, I cannot reject their equality at standard confidence levels. I then solve for the measure of the husband’s bargaining power. Focusing on the estimates in columns 3 and 4, I estimate this to be .60 with a standard error of .0707 as in Table 6. I then test the null hypothesis that this parameter equals .5, i.e. equal bargaining power between the husband and wife. I find that the null hypothesis that husbands and wives have equal bargaining power is not rejected at the standard size of test. Table 5.1: Regression Estimation of the Rosett’s Friction Model in Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩100,000 ( 1,2 , ) ijhpwp = = ) (1) (2) (3) (4) Variables Transfer to husband's parents Transfer to wife's parents Transfer to husband's parents Transfer to wife's parents Couple's predicted income .0274*** .0179*** .0269*** . 0180*** (.0048) (.0055) (.0040) (.0054) Husband's parents' income -.0158 .0004 -.0172* (.0101) (.0088) (.0093) Wife's parents' income -.0021 -.0166* -.0164* (.0089) (.0097) (.0086) Observations 1800 1800 1800 1800 Notes: (1) Bootstrapped Standard errors are in parentheses. (2) Resamplings are based on households. (3) Bootstrap replications (500). (4) Couple's income is predicted from couple's net assets, husband’s education and wife’s education. *** p<0.01, ** p<0.05, * p<0.1 34 I restimate reduced form model in (18) for j i K = ₩50,000 ( 1,2 , ) ijhpwp == for robustness check. Table 5.2 shows the results which are very similar to those in Table 5.1. The second column of Table 6 shows the structural results with this fixed cost. Again I estimate h μ at .60 with a standard error of .0729. Generally, the test results support equal bargaining power between husbands and wives. Table 5.2: Regression Estimation of the Rosett’s Friction Model in Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩50,000 (1,2 , ) ijhpwp = = ) (1) (2) (3) (4) Variables Transfer to husband's parents Transfer to wife's parents Transfer to husband's parents Transfer to wife's parents Couple's predicted income .0266*** .0173*** .0261*** .0173*** (.0047) (.0055) (.0039) (.0052) Husband's parents' income -.0156 .0002 -.0171* (.0100) (.0087) (.0089) Wife's parents' income -.0022 -.0161* -.0160** (.0088) (.0096) (.0082) Observations 1800 1800 1800 1800 See notes to Table 5.1. 35 Table 6: Structural Results for Two-Way Transfers (Model 1.4.1) Parameter K = ₩100,000 K = ₩50,000 h μ .5992 .6020 (.0707) (.0729) Notes: (1) Standard errors are in parentheses. (2) Imputation error is corrected by bootstrapping. Next, I investigated whether husband’s bargaining power varies across education groups. I divided 4 groups of households according to different level of husband’s and wife’s education. The Group1 includes households that both levels of husband and wife’s education are greater than high school. In Group2 households husband’s education is greater than high school while wife’s education is less than or equal to high school. The opposite case of households is included in Group3. The last group of households (Group4) consists of households where both spouses’ education level is below or equal to high school. Table 8 shows the structural results for each education group which are calculated from Table 7.1 and Table 7.2. I tested whether husband’s bargaining power vary across groups and I found that I cannot reject equality of h μ across groups. 18 18 The last row of Table 7 shows results of the test 01 2 3 4 : hhh h H μ μμμ === . 36 Table 7.1: Regression Estimation of the Rosett’s Friction Model in Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩100,000 ( 1,2 , ) ijhpwp = = ) (1) (2) (3) (4) Variables Transfer to husband's parents Transfer to wife's parents Transfer to husband's parents Transfer to wife's parents Group1's predicted income .0288*** .0193*** .0276*** .0195*** (.0048) (.0068) (.0042) (.0068) Group2's predicted income .0257*** .0150*** .0247*** .0153*** (.0060) (.0045) (.0053) (.0041) Group3's predicted income .0419*** .0168** .0400*** .0170** (.0077) (.0072) (.0068) (.0072) Group4's predicted income .0255*** .0135*** .0240*** .0138*** (.0055) (.0049) (.0048) (.0044) Husband's parents' income -.0131 .0016 -.0166* (.0103) (.0092) (.0094) Wife's parents' income -.0055 -.0152 -.0144* (.0089) (.0095) (.0082) Observations 1800 1800 1800 1800 See notes to Table 5.1. Group1: Husband’s education > 12 and Wife’s education > 12 Group2: Husband’s education > 12 and Wife’s education ≤ 12 Group3: Husband’s education ≤ 12 and Wife’s education > 12 Group4: Husband’s education ≤ 12 and Wife’s education ≤ 12 37 Table 7.2: Regression Estimation of the Rosett’s Friction Model in Sections 1.4.1 and 1.4.2 with fixed costs ( j i K = ₩50,000 ( 1,2 , ) ijhpwp = = ) (1) (2) (3) (4) Variables Transfer to husband's parents Transfer to wife's parents Transfer to husband's parents Transfer to wife's parents Group1's predicted income .0280*** .0186*** .0268*** .0189*** (.0048) (.0067) (.0042) (.0068) Group2's predicted income .0249*** .0142*** .0238*** .0145*** (.0059) (.0045) (.0053) (.0041) Group3's predicted income .0409*** .0162** .0388*** .0164** (.0076) (.0071) (.0067) (.0071) Group4's predicted income .0246*** .0127*** .0231*** .0130*** (.0054) (.0048) (.0047) (.0044) Husband's parents' income -.0129 .0015 -.0163* (.0102) (.0091) (.0093) Wife's parents' income -.0055 -.0147 -.0140* (.0088) (.0094) (.0082) Observations 1800 1800 1800 1800 See notes to Table 7.1. 38 Table 8: Structural Results for Two-Way Transfers by Education Group (Model 1.4.1) Parameter K = ₩100,000 K = ₩50,000 Group1 ( 1 h μ ) .5857 .5873 33%* (.0814) (.0837) Group2 ( 2 h μ ) .6171 .6213 18%* (.0686) (.0707) Group3 ( 3 h μ ) .7016 .7032 6%* (.0907) (.0927) Group4 ( 4 h μ ) .6353 .6398 42%* (.0770) (.0782) Prob >chi2(3)** (.5983) (.6145) See Notes to Table 6. *Percentage of each group in the samples is shown under each group **Test :. 01 2 3 4 hh hh H μ μμ μ === 1.8 Conclusions and Future Research In this paper, two models of intergenerational transfers between adult children and old parents are derived assuming a formal collective model framework. This allows me to investigate the respective bargaining power between husbands and wives. My work differs from previous work in that I include both sets of parents, focus on bargaining over a form of semi-private consumption, and estimate a structural bargaining parameter. Overall the results suggest that the husband’s preferences and the wife’s preferences are equally weighted. 39 For my future research I can consider the following extensions. First, I will consider more structural models where the husband’s weight depends positively on his education, and negatively on the wife’s education. 19 Second, I will consider alternative specifications of preferences where equivalence scales play a role in transfers. Third, Mazzocco (2007) found that household members cannot commit to future plans and the individual participation constraints bind frequently, which implies that households must renegotiate their decisions over time. This paper use pooled cross section data for 5 years. This time period may not be long enough to detect the variation in individual decision power. However, it may be possible to apply dynamic collective model in the long run as we have more time periods in the panel data set. If we incorporate the dynamic features of the collective model, we may be able to allow for bargaining power between husband and wife to be affected by transfers made by each set of parents. 19 If the model of Raut and Tran is correct, the education coefficients will also reflect the son’s and daughter’s obligation to their parents varying positively with the child’s obligations. 40 Chapter Two Structural Estimation of Bargaining Power within the Family: Transfers to Parents from Adult Children and Vice-versa 2.1 Introduction In this paper I consider two models for intergenerational transfers and estimate structural parameters in the models. The motivation of my work in this paper is fourfold. First, I introduce bargaining between husband and wife in the intergenerational transfers and consider transfers from a married couple to both spouses’ parents at the same time because one married couple usually face two sets of parents to support. Second, I consider Kuhn-Tucker problem in estimation. In the first I allow only for transfer to the parents (and not from the parents). I derive a static collective model to explain couple’s joint decision on these transfers to both sets of parents. This setting of collective model generates a typical example of Kuhn-Tucker problem. The estimation of this model takes into account that the functional form of the transfer to one set of parents will change if the optimal transfer to the other set of parents is zero. Even though micro theory models deal with Kuhn-Tucker problem pretty well, it is rare to directly consider Kuhn-Tucker solutions in the estimation process. I estimate this model on data from Korea (2001–2005) and I find that husbands and wives have equal bargaining power in South Korea. Third, I extend this model to allow for transfers from parents. As shown in Table 2 in section 1.2 approximately 57% of families make net transfers to at least one parent. Even though the main flow of transfers is upstream, 21% of couples rather receive a net transfer from their parents. This 41 suggests the need to also consider models to allow transfers from parents. The second model extends the first model to allow the couple to be forward looking in the sense that they take potential transfers from parents into account when deriving their sharing rule. I extend Raut and Tran (2005) by applying Nash Equilibrium concept to the collective model; this extension comes at the cost of greatly increasing the complexity of the estimation procedure. Using this model I estimate that husbands and wives have equal bargaining power. Overall the results suggest that the husband’s preferences and the wife’s preferences have equal weight. Lastly, I investigate the role of education in determining bargaining power. I suggest that bargaining power is a function of third polynomial of relative education of spouses. This paper is organized as follows. In section 2.2, I present two theoretical models of transfers. In section 2.3 I discuss the respective estimation strategies for the models. Section 2.4 present estimation results for each model. Section 2.5 concludes the paper. 2.2 Economic Models of Transfers from Adult Children to Parents 2.2.1 One-way Transfers When studying one-way transfers from the children to the parents, I assume that the children determine the size of the transfer – parents are completely passive in this model. Each spouse has his/her own consumption (not observed in my data) and is assumed to care about only his or her own parents’ utility. Each spouse treats their parents collectively. That is, only total consumption of his/her parents matters and how surviving parents allocate the transfers from their child does not affect the couple’s transfer 42 decisions to their parents. Given these assumptions, each partner’s parent(s) is treated hereafter as one entity. Further, I assume that children’s joint utility function does not contain a public good. The family’s maximization problem that determines all consumption can be written as ( ) () () ( ) ( ) ,, , ,1 , . . , , , 0, 0, hw hp wp h h hp hp w w wp wp CC T T hw hp wp hp hp hp wp wp wp hp wp Max U C V C U C V C st C C T T Y CY T CY T T T μμ +− ++ + = =+ =+ ≥ ≥ (19) where i U ) , ( w h i = is each marriage partner’s own utility function while 1 0 < < μ is the Pareto weight on the husband’s utility, which is interpreted as husband’s relative bargaining power. Further, ( , ) i Ci hw = is each spouse’s private consumption, (,) ip Ci hw = is the total consumption of each set of parents, ( , ) ip Yi hw = denotes the parents’ before-transfer income and ( ) , ip Ti hw = denotes the transfers made to the parents. To make this problem more tractable, I employ an additively separable logarithmic utility function for each spouse’s utility and introduce an altruism parameter 1 0 < < i α ) , ( w h i = for the strength of each child’s altruistic feelings toward their parents’ consumption. Husband and wife are assumed to care about their respective parents equally ( hw α α = ). 43 The optimization problem is ( ) () () ( ) ( ) ,, , ln ln 1 ln ln . . , 0, 0. hw hp wp h hp hp w wp wp CC T T hw hp wp hp wp Max C Y T C Y T st C C T T Y T T μα μ α ++ +− + + ++ + = ≥ ≥ (20) This is a Kuhn-Tucker problem () () () ( ) ( ) () 123 ln ln 1 ln ln . h hp hp w wp wp hw hp wp hp wp LC Y T C Y T CC T T Y T T μα μ α λλλ =+ + +− + + −+ + + − + + (21) The first order conditions are () () () 1 1 12 13 123 0, 1 0, 0, 1 0, 0, 0, 0. hh ww hp hp hp wp wp wp h w hp wp hp wp L CC L CC L TY T L TY T CC T T Y T T μ λ μ λ μα λλ μα λλ λλλ ∂ =− = ∂ − ∂ =−= ∂ ∂ =−+= ∂+ − ∂ =−+= ∂+ ++ + − = = = (22) 44 (The multiplier 0 1 > λ because the couple’s budget constraint is always binding.) The first order conditions for 2 λ and 3 λ can be simplified as follows. () 21 31 1 , 1 1 . hp hp hp wp wp wp hp wp YT YT T YT YT T μα λλ μα λλ += = +−− − += = +−− (23) We have four solutions depending on whether couple’s corner solutions: Case 1: 0 hp T = and 0. wp T = In this case the solution for i λ is 20 () 1 2 3 1 , , 1 . hp hp wp wp Y YY YY YY YY λ μα λ μ α λ = − = −− = (24) Equation (24) shows when the incomes of both spouses’ parent are greater than each partner’s composite altruism (product of bargaining power and altruism); no transfer will be made to any parents since 2 λ and 3 λ are greater than zero. 20 Recall that when 0 hp T = and 0, wp T = 2 0 λ > and 3 0. λ > 45 Case 2: 0 hp T = and 0. wp T > Now the solution for 12 3 ,, λ λλ and wp T is obtained from (22) as follows. () () () () () ()() () 1 2 3 11 , 11 , 0, 1 . 11 wp hp wp hp wp wp wp wp YY YYY YY Y YY TY μα λ μα μα λ λ μα μα +− = + +− − + = + = −+ =− +− (25) Equation (25) shows that if () () , 11 hp wp YYY μ α μα >+ +− then 0. hp T = This is consistent with our intuition that couple does not make a positive transfer if parents’ income is high enough. However, it can be shown that as husband’s bargaining power improves and/or couple’s altruism increases, the threshold income of the husband’s parents hp Y at which no transfer will be made to the husband’s parents becomes higher. Case 3: 0 * > hp T and * 0. wp T = In the same way, the solution for 12 3 ,, λ λλ and hp T is given by 46 () ( )() () () 1 2 3 1 , 0, 11 , . 1 wp wp hp wp hp hp hp hp YY YYY YY Y YY TY μ α λ λ μα μ α λ μα μα + = + = +−− + = + + =− + (26) From (26) we see that if ( ) () 1 , 1 wp hp YYY μα μα − >+ + then 0. wp T = It can be shown that as couple’s altruism increases, the threshold income of the wife’s parents wp Y at which no transfer will be made to the wife’s parents becomes higher while the threshold becomes lower as husband’s bargaining power improves. Case 4: 0 * > hp T and * 0. wp T > In this case we have simple form of interior solution. () , 1 hp hp wp hp TYYYY α μ α =++ − + (27) () () 1 . 1 wp hp wp wp TYYYY αμ α − =++− + (28) The interior solution is derived from the first order condition exploiting the fact that the marginal utilities of the goods must be equalized. This property is used in estimation strategy of this model in section 2.4. 47 2.2.2 A Forward Looking Two-way Transfers model In this section I consider a two-way transfers model that allows the couple to consider possible transfers from the parents when deriving the sharing rule. To do this I assume that the outcome is the result of a game between the married couple and their parents, and focus only on the Nash Equilibrium for this game. 21 Formally there are three players in the game: i) the husband and wife who are assumed to play jointly; ii) the husband’s parents and iii) the wife’s parents. I assume that each spouse’s parents care only about their own child and neither set of parents cares about either their child’s spouse or the in-laws. To give an overview, I assume the children play first, and derive the optimal transfers assuming that transfers only go to the parents. If both transfers are positive, then the children make those transfers. If both optimal transfers to the parents are zero, the couple will consider the possibility they will get transfers from both sets of parents. In this case they will need to derive the sharing rule incorporating possible transfers from the parents. For the case in the one-way transfers model where the optimal transfer to the wife’s parents is zero, but the optimal transfer to the husband’s parents is positive, they assume they may receive a positive transfer from the wife’s parents, but that they will make a positive transfer to the husband parents. In other words, the possibility of receiving a transfer from the wife’s parents does not cause them to eliminate their transfer to the husband’s parents, but may affect their size of the transfer to the husband’s parents. The case where the optimal transfer in the one-way model to the husband’s parents is 21 This model can be viewed as an extension of Raut and Tran (2005)’s two-sided altruism reciprocity model, who also focus on the Nash Equilibrium. 48 zero and the optimal transfer to the wife’s parents is positive is analogous. Revisit the couple’s problem for the one-way transfer model from section 2.1.1 above () ( ) ( ) () ( ) ( ) ( ) ,, , ln ln 1 ln ln . . hw hp wp hhphp w wpwp CC T T hw hp wp Max C Y T C Y T st C C T T Y μα μ α ++ +− + + ++ + = (29) Note that there is no constraint that 0, 0 hp wp TT >> in (29), which makes difference of two-way model from one-way model. If the optimal transfers are both positive in (29), couple make those optimal transfer amounts to both sets of parents. The functional from of each transfer is as in (27) and (27) in the previous section. If one of the transfers is zero in (28), this will affect the functional form of the other transfer and the sharing rule. This will be analyzed as are the cases when one of the transfers from the couple to parent is positive in the one-way model. I consider three different cases for this possibility. Case 1: Suppose 0 hp T < in (27) and 0 wp T > in (28). 22 This means couple wants to make a positive transfer to the wife’s parents while expects some transfer from the husband’s parents. In this case I assume that the preference of the husband’s parents is taken into account by the couple when it carries out maximization. I introduce the husband’s parents’ altruism toward their son which is given by hp β (01 hp β << ). 22 If () () 11 hp wp YYY μα μα >+ +− and ( ) () 1 1 , wp hp YYY μα μα − <+ + then 0 hp T < and 0. wp T > 49 Specifically, I assume that the husband’s parents consider the optimization problem. ln ln .. , , 0, ph hp h hp T hp ph hp hh ph pw Max C C st C T Y CT T β ρ + += =+ ≥ (30) where ph T represents net transfer to the husband made by his parents, and h ρ is the husband’s income share out of the couple’s total household income . Y The solution is , . 1 if RHS>0 11 0 otherwise hp ph hp h hp hp TY β ρ ββ =− ++ = (31) From (31) we can see that the solution of the optimization of the husband’s parents is a function of h ρ which is husband’s share out of couple’s family income. That is, (31) is the best response of the husband’s parents to the couple’s sharing rule. In this case where 0 hp T < in (27) and 0 wp T > in (28), the couple’s problem is to determine optimal transfer to the wife’s parents ( wp T ) and the sharing rule ( , hw ρ ρ ) for their family income Y given arbitrary transfer amount of 0. ph T ≥ 50 () ( ) () ( ) ( ) ( ) ( ) ,, ln ln 1 ln ln .. , , , . hw wp h hp w wp wp T hw wp ph hw hh ph hp hp ph Max C C C Y T st C C T Y T Y CT CY T ρρ μα μ α ρρ ρ ++− + + ++ = + += =+ =− (32) Last two expressions in the constraints come from the fact that the husband’s parents make transfer only to their son. We can rewrite Problem (32) by substituting all the constraints into the objective function. () ( ) () () ( )( ) ( ) , ln ln 1 ln ln . hwp hph hp ph wp h wp wp T Max T Y T Y T Y T ρ μρ α μ ρ α ++ − +− − − + + (33) The solution of (33) is a function of ph T as ()() () ()()( ) () 1 , 11 11 . 11 ph wp wp wp ph h YT Y T YY T μα μα μμα ρ μα −+ − = +− +−− + = +− (34) Two equations of wp T and h ρ in (34) are couple’s best response functions to . ph T We achieve the Nash equilibrium by solving (31) and (34) simultaneously. 51 () () ( ) () ()()()( ) () () ()() () 1 , 1 111 , 1 1 . 1 hp wp hp hp ph hp hp wp hp hp hp h hp hp wp wp hp hp wp hp hp YYY T YY Y YY Y T βμαβ μ βμαβ μ μβ μ αβ ρ βμαβ μ μαβ β μ βμαβ μ +− − + = +− + ++ −− + = +− + −+−+ = +− + (35) We have two sets of transfers in this case depending on . ph T 23 (i) If 0 ph T > in (35), () ()() () () () () () 0, 1 , 1 1 . 1 hp wp wp hp hp wp hp hp hp wp hp hp ph hp hp T YY Y T YYY T μαβ β μ βμαβ μ βμαβ μ βμαβ μ = −+−+ = +− + +− − + = +− + (36) (ii) If 0 ph T = in (35), ()() () 0, 1 , 11 0. hp wp wp wp ph T YY TY T μα μα = −+ =− +− = (37) 23 Recall I assume that there is no role of the wife’s parents in case 1. Hence, I do not consider transfer behavior of the wife’s parents. 52 Note that (36) implies that the husband’s parents make no transfer to the husband unless their income is greater than the threshold income , hp Y i.e. () () If ,then 0. 1 hp wp hp ph hp hp YYYYT μ βμαβ <+= = +− (38) Recall that if the husband’s parents’ income is greater than the threshold income hp Y , no transfer will be made by couple to the husband’s parents. i.e. () () If ,then 0. 11 hp wp hp hp YYYYT μ α μα >+= = +− (39) It can be shown from (38) and (39) that there are income zones where a transfer from the husband’s parents to the husband, or from couple to the husband’s parents, and also zones where no transfers are made. 24 However, a case with two transfers, from couple to the husband’s parents and from the husband’s parents to their son, can never occur. Further, wp T in (37) is the same as one in case 2 in section 2.1.1. This implies that this two-way transfers model goes back to the one-way transfers model if optimal ph T is zero. Case 2: Suppose 0 hp T > in (27) and 0 wp T < in (28). This means the couple wants to make a positive net transfer to the husband’s parents while expects some transfer from the 24 () () 111 . hp hp μμα β μαβ μ α > +− + − Hence, . hp hp YY > If hp Y is between these two thresholds, there will be no transfer between couple and the husband’s parents. 53 wife’s parents. The same analysis is applied here because the problem is totally symmetric to the case 1. I assume that the preference of the wife’s parents is important when couple carries out maximization. The wife parents’ altruism toward their daughter which is given by wp β (0 1 wp β << ) is introduced. The wife’s parents have the optimization problem. ln ln .. , , 0, pw wp w wp T wp pw wp ww pw pw Max C C st C T Y CT T β ρ + += =+ ≥ (40) where pw T stands for net transfer to the wife made by her parents and w ρ is the wife’s income share. The solution of problem (40) is . 1 if RHS>0 11 0 otherwise wp pw wp w wp wp TY β ρ ββ =− ++ = (41) Likewise, (41) is the best reaction function of the wife’s parents to the couple’s sharing rule. In this case where 0 hp T > in (27) and 0 wp T < in (28), couple determines optimal transfer to the husband’s parents and the sharing rule given arbitrary transfer amount of 0. pw T ≥ 54 () ( ) () ( ) ( ) ( ) ( ) ,, ln ln 1 ln ln . . , , , . hw hp hhphp w hp T h w hp pw hw ww pw wp wp pw Max C Y T C C st C C T Y T Y CT CY T ρρ μα μ α ρρ ρ ++ +− + ++ = + += =+ =− (42) We can rewrite (42) by substituting all the constraints into the objective function as () ( ) () () ( )( ) ( ) , ln ln 1 ln ln . whp hp w hp hp w pw wp pw T Max Y T Y T T Y T ρ μρα μρ α −− + + + − + + − (43) In the same way we have the Nash equilibrium as () ( ) ( ) () ()() () ()()() () () 1 , 1 1 , 1 11 1 . 1 wp hp wp wp pw wp wp hp hp wp wp hp wp wp hp wp wp hp w wp wp YYY T YY Y T YY Y βμαβ μ βμαβ μ μα β β μ βμαβ μ μβ μ αβ ρ βμαβ μ +−− + = ++− +− −+ = ++− −+ + − + = ++− (44) We have two sets of transfers in this case depending on . pw T 55 (iii) If 0 pw T > in (45), () ( ) () () ()() () 1 , 1 0, 1 1 hp hp wp wp hp wp wp wp wp hp wp wp pw wp wp YY Y T T YYY T μα β β μ βμαβ μ βμαβ μ βμαβ μ +− −+ = ++− = +−− + = ++− (45) (iv) If 0 pw T = in (45), () , 1 0, 0. hp hp hp wp pw YY TY T T μα μα + =− + = = (46) We can see from (45) that the wife’s parents make no transfer to the wife unless their income is greater than the threshold , wp Y i.e. If () () 1 , wp hp wp hp hp YYYY μ βμαβ − <+= + then 0. pw T = (47) Recall that if the wife’s parents’ income is greater than the threshold wp Y , no transfer will be made by couple to the wife’s parents, i.e. 56 If () () 1 , 1 wp hp wp YYYY μα μα − >+= + then 0. wp T = (48) In case 2 we can also show that there are three income zones where a transfer from the wife’s parents to the wife, or from couple to the wife’s parents, and where no transfers are made. 25 However, a case with two positive transfers, from couple to the wife’s parents and from the wife’s parents to their daughter, can never arise. It is also the case that if 0, pw T = hp T has the same functional form in case 3 in section 2.1.1. Case 3: Suppose 0 hp T < in (27) and 0 wp T < in (28). This means that the couple does not want to make a transfer to either set of parents in Model 1. Instead, the couple expects some positive transfers from both sets of parents and I assume that both sets of parents enter the game. The couple solves following optimization problem given arbitrary transfer amount of 0, 0 ph pw TT >> . 25 () () 11 1 . hp hp μ μα βμαβ μα −− > ++ Hence, . wp wp YY > If wp Y is between these two thresholds, there will be no transfer between couple and the wife’s parents. 57 () ( ) () () ( ) ( ) ( ) , ln ln 1 ln ln .. , , , , , . hw hhp w wp h w ph pw hw hh ph hp hp ph ww pw wp wp pw Max C C C C st C C Y T T Y CT CY T CT CY T ρρ μα μ α ρρ ρ ρ ++− + += + + += =+ =− =+ =− (49) Substituting all the constraints into the objective function yields () ( ) () () ( )( ) ( ) ln ln 1 ln ln . h hph hp ph hpw wp pw Max T Y T Y T Y T ρ μρ α μ ρ α ++ − +− − + + − (50) The solution of h ρ as a function of hp T and wp T is () . hphpwph YT T T ρμ =+ + − (51) The husband’s parents’ reaction function is still given by (31) and the wife’s parents’ reaction function is also given by (41). We obtain the Nash equilibrium by solving (31), (41) and (51) simultaneously as 58 () () ()() , 1 , 1 1 1. 1 hhpwphp ph hp hp wp hp pw wp hp wp wp YY Y Y TY YY Y TY YY Y ρμ μ β μ β =+ +− =− + + + =− − + + + (52) We have four sets of transfers in this case depending on ph T and . pw T (v) If 0 ph T > and 0 pw T > in (52), () ()() 0, 0, 1 , 1 1 1. 1 hp wp ph hp hp wp hp pw wp hp wp wp TT TY YY Y TY YY Y μ β μ β == =− + + + =− − + + + (53) (vi) If 0 ph T > and 0 pw T = in (52), () 0, 0, 1 , 1 0. hp wp ph hp hp wp hp pw TT TY YY Y T μ β == =− + + + = (54) 59 (vii) If 0 ph T = and 0 pw T > in (52), ()() 0, 0, 0, 1 1. 1 hp wp ph pw wp hp wp wp TT T TY YY Y μ β == = =− − + + + (55) (viii) If 0 ph T = and 0 pw T = in (52), 0, 0, 0, 0. hp wp ph pw TT T T == = = (56) We can find the threshold incomes for each set of parents at which no transfer will be made to the couple by parents from (53). If, () () , 1 hp wp hp hp YYYY μ βμ <+= +− then 0. ph T = (57) If, () () 1 , wp hp wp wp YYYY μ βμ − <+= + then 0. pw T = (58) 60 2.3 Estimation Strategy 2.3.1 Estimation Approach for the Model of One-Way Transfers A structural estimation method which is directly derived from Kuhn-Tucker maximization problem is required. I introduce error terms by assuming that preferences are random over population to allow for household differences in tastes. 26 A convenient assumption is to let marginal utilities consist of a deterministic part and a random component. I simply work with observable total family consumption because I do not have to observe private consumption by the husband and wife. The couple allocates its income over family consumption, transfer to the husband’s parent and transfer to the wife’s parent. Given that family consumption is always positive, we observe four possible patterns, one with all transfers positive, two with one transfer positive, and one with all transfers zero. I assume there is a composite consumption good with normalized price unity. The marginal utilities of couple’s family consumption, a transfer to the husband’s parents and a transfer to the wife’s parents are derived. I normalize the variables by dividing couple’s income Y to avoid the problem that the marginal utilities depend on dollar unit of variables in the estimation process. Normalization dose not change couple’s optimization process, and now we focus on utility maximizing expenditure shares instead of actual dollar amount of the consumptions. 27 26 I follow Wales and Woodland (1983)’s Kuhn-Tucker approach so as to derive likelihood function. 27 I use the properties that conditional demand function is homogenous of degree zero in price and income. 61 (),, it it Cit cit cit hp wp it it it Y MUu u YT T =+ −− X (59) () () 1 ,, wp it it it wpit wpit wp wp T it it Y MU u u YT μα − =+ + X (60) () ,, hp it it it hpit hpit hp hp T it it Y MUu u YT μ α =+ + X (61) where it X is a vector of expenditure shares. In my empirical work I will make μ to depend on the husbands and wife’s education or other characteristics and in the same way one can make α a function of the couple’s number of children. However for ease of exposition I treat them as constant in what follows. From the individual household’s perspective marginal utilities are non-stochastic since the household knows it u vector. However, from the econometrician’s perspective the vector it u for each household is random drawing from a known population. We assume it u has a joint normal distribution with zero means and constant covariance matrix . Σ From the utility maximization we know if 0, wp it T > this implies the marginal utility of family consumption is equal to the marginal utility of a transfer to the wife’s parents, and if 0, wp it T = then the marginal utility of family consumption is greater than the marginal utility of a transfer to the wife’s parents. Similarly, if 0, hp it T > the marginal utility of family consumption is equal to the marginal utility of a transfer to the husband’s parents, and if 0, hp it T = then the marginal utility of family consumption is greater than the marginal utility of a transfer to the husband’s parents. First, for the case where 62 0 hp it T = and 0, wp it T = we have following equations from (59), (60) and (61) by comparing the marginal utilities. () 1, 1 1. it hpit cit hp it it wpit cit wp wp it it Y uu Y Y uu YT μ α μα +<+ − +<+ + (62) Second, for the case where 0 hp it T = and 0, wp it T > we have following equations in the same way. () , 1 . it it hpit cit hp wp it it it it it wpit cit wp wp wp it it it it YY uu YYT Y Y uu YT Y T μ α μα +< + − − += + +− (63) Third, for the case where 0 hp it T > and 0, wp it T = we have () , 1 . it it hpit cit hp hp hp it it it it it it wpit cit wp hp it it it YY uu YT Y T Y Y uu YYT μ α μα += + +− − +< + − (64) For the case where 0 hp it T > and 0, wp it T > we have following two equality conditions. 63 () , 1 . it it cit hpit hp wp hp hp it it it it it it it cit wpit hp hp wp wp it it it it it YY uu YT T Y T Y Y uu YT T Y T μ α μα += + −− + − += + −− + (65) Define hpit hpit cit uu ε =− and , wpit wpit cit uu ε = − where ( , hpit wpit ε ε ) is assumed to follow a bivariate normal distribution with zero means and covariance matrix . Ω Given the distributional assumptions, we can determine the contribution to the likelihood function. For the case where no transfer is made to either set of parents, the marginal utility of couple’s family consumption is greater than both the marginal utilities of a transfer to the husband’s parents and a transfer to the wife’s parents. We can derive following expressions from (62). () 1, 1 1. it hpit hp it it wpit wp it Y Y Y Y μ α ε μα ε <− − <− (66) Hence, the probability of this event is () 2 ;0, hpit wpit zz wpit hpit ndd ε ε −∞ −∞ Ω ∫∫ it ε (67) 64 where 1, it hpit hp it Y z Y μ α =− ( ) 1 1, it wpit wp it Y z Y μα − =− ( ) , hpit wpit εε = it ε and () 2 ;0, n Ω it ε is the bivariate normal density with zero means and covariance matrix , Ω where 2 2 hp hpwp hpwp wp σσ σσ ⎡⎤ Ω= ⎢⎥ ⎢⎥ ⎣⎦ . For the case where transfer to the wife’s parents is only made, the marginal utility of family consumption is greater than the marginal utility of a transfer to the husband’s parents and equal to the marginal utility of a transfer to the wife’s parents. We can express this case in terms of hpit ε and wpit ε from (63) () , 1 . it it hpit wp hp it it it wp it it wpit wp wp wp it it it it YY YT Y Y Y YT Y T μ α ε μα ε <− − − =− −+ (68) Hence, the contribution to the likelihood function for this case is () () () 21 2 22 1 2 2 0, ;0, 1 exp / 1 2 2 hpit z wp it it hpit wpit hp hpit wpit hp it wp wp wp fT n J d Fz J ε εσ ρε ρσ σσ πσ −∞ =Ω ⎡⎤ ⎧⎫⎧ ⎫ ⎪⎪⎪ ⎪ =− − − ⎢⎥ ⎨⎬⎨ ⎬ ⎪⎪⎪ ⎪ ⎢⎥ ⎩⎭⎩ ⎭ ⎣⎦ ∫ it ε (69) 65 where , it it hpit wp hp it it it YY z YT Y μ α =− − ( ) 1 , it it wpit wp wp wp it it it it Y Y YT Y T μα ε − =− −+ and F denotes the univariate normal distribution function. Further, 1it J is the Jacobian of the transformation from wpit ε to wp it T and is given by () () () 1 22 1 . it it it wp wp wp it it it it Y Y J YT Y T μα − =+ −+ The last expression of (69) is obtained by expressing the joint normal density as the product of the marginal normal density for wpit ε and the conditional normal density of hpit ε given wpit ε and performing the appropriate integration. For the case where transfer to the husband’s parents is only made, marginal utility of couple’s consumption is greater than marginal utility of a transfer to the wife’s parents and equal to marginal utility of a transfer to the husband’s parents. From (64) () , 1 . it it hpit hp hp hp it it it it it it wpit hp wp it it it YY YT Y T Y Y YT Y μ α ε μα ε =− −+ − <− − (70) The contribution to the likelihood function for this case is () () () 22 2 22 2 2 2 ,0 ;0, 1 exp / 1 , 2 2 wpit z hp it it wpit hpit wp wpit hpit wp it hp hp hp fT n J d Fz J ε εσ ρε ρσ σσ πσ −∞ =Ω ⎡⎤ ⎧⎫ ⎧ ⎫ ⎪⎪ ⎪ ⎪ =− − − ⎢⎥ ⎨⎬ ⎨ ⎬ ⎪⎪ ⎪ ⎪ ⎢⎥ ⎩⎭ ⎩ ⎭ ⎣⎦ ∫ it ε (71) 66 where () 1 , it it wpit hp wp it it it Y Y z YT Y μα − =− − it it hpit hp hp hp it it it it YY YT Y T μ α ε=− −+ and 2it J is the Jacobian of the transformation from hpit ε to hp it T equal to ()() 2 22 . it it it hp hp hp it it it it YY J YT Y T μ α =+ −+ For the last case where two transfers are made to both sets of parents, the marginal utility of couple’s family consumption is equal to both the marginal utilities of a transfer to the husband’s parents and a transfer to the wife’s parents. From (65) () , 1 . it it hpit hp wp hp hp it it it it it it it wpit hp hp wp wp it it it it it YY YT T Y T Y Y YT T Y T μ α ε μα ε =− −− + − =− −− + (72) Hence, the density function is () () () 23 22 3 2 2 ,;0, 11 exp 2 , 21 21 hp hp it it it hpit hpit wpit wpit it hp hp wp wp hp wp fT T n J J εεε ε ρ σσσ σ ρ πσ σ ρ =Ω ⎡⎤ ⎧⎫ ⎡⎤ ⎡ ⎤ ⎡ ⎤ ⎪⎪ ⎢⎥ =− −⋅+ ⎢⎥ ⎢ ⎥ ⎢ ⎥ ⎨⎬ ⎢⎥ − − ⎢⎥ ⎢ ⎥ ⎢ ⎥ ⎪⎪ ⎣⎦ ⎣ ⎦ ⎣ ⎦ ⎩⎭ ⎣⎦ it ε (73) where 3it J Jacobian of the transformation from ( ) , hpit wpit εε to ( ) , hp wp it it TT which is described as 67 ()() () () () ()( ) 3 22 2 22 2 2 2 11 . hpit hpit hp wp it it hpit wpit hpit wpit it hp wp wp hp wpit wpit hp wp it it it it hp wp hp hp wp wp hp hp wp wp it it it it it it it it it it it TT J TT T T TT Y Y YT T Y T Y T Y T Y T εε εε εε εε μμμα α μ ∂∂ ∂∂ ∂∂ ∂∂ ==⋅−⋅ ∂∂ ∂∂ ∂ ∂ ∂∂ ⎡⎤ −− ⎢⎥ =++ ⎢⎥ −− + + + + ⎣⎦ The contribution to the likelihood function may then be expressed as () () 1 1 ,,, NT hp wp hp wp it it it LT T f T T = = Π where () , hp wp it it TT is the th i household’s observation on wp hp T T , at time . t The parameters of this utility function and the covariance matrix Σ can be estimated by maximizing the likelihood function. These maximum likelihood estimates will be consistent, efficient and asymptotically normally distributed. The complete form of likelihood function is expressed by combining equations (67), (69), (71) and (73). 68 () () ( ) () () () () () 10, 0 12 11 10, 0 22 10, 0 21 2 ,;0, ;0, ;0, ;0, hp wp it it hpit wpit hp wp it it hpit hp wp it it wpit TT zz NT hp wp it hpit wpit it TT z it it hpit TT z it it wpit it LT T n d d nJd nJd nJ εε ε ε = = == −∞ −∞ => −∞ >= −∞ ⎡⎤ =Π Π Ω ⎢⎥ ⎢⎥ ⎣⎦ ⎡⎤ ⋅Ω ⎢⎥ ⎢⎥ ⎣⎦ ⎡⎤ ⋅Ω ⎢⎥ ⎢⎥ ⎣⎦ ⋅Ω ∫∫ ∫ ∫ ε ε ε ε () 10, 0 3 . hp wp it it TT it >> ⎡⎤ ⎣⎦ (74) The above likelihood function is based on observing parents’ income. In fact I need to impute it for this model, as well as the models below, from KLoSA. The bottom line is that imputing income for each set of parents and substituting it into the estimating equations is likely to provide consistent estimates but incorrect standard errors unless one uses the bootstrap to calculate the standard errors. Further, one may worry that the family’s income is endogenous, since if there is a shock to their desired transfers to their parents, the husband and wife may work harder. In this case transfers are causing income while I want the casual effect of family income on transfers. Following Kazianga I treat couple’s income as endogenous and predict using family net assets and husband and wife’s age and education as excluded instruments. The couple’s predicted income is used in the estimation. 69 2.4 Estimation Approach for the Model of Two-Way Transfers The appropriate estimation approach for this model is extremely complicated. As described in the previous section I consider nine possible transfer patterns between the couple and both sets of parents. I construct a contribution to the likelihood function for each case by introducing normal optimization error terms and then combine these contributions to build the complete likelihood function. The couple’s predicted income is used in the estimation as one-way transfers model. Case 1: 0, 0. hp wp it it TT >> In this case I assumed parents do not play. Couple decides transfers to both sets of parents. Thus, I have two index functions. () () () , 1 1 . 1 hp hp wp hp it it it it it hpit wp hp wp wp it it it it it wpit TYYYYu TYYYYu α μ α αμ α =++ −+ + − =++−+ + (75) The contribution to the likelihood function for case 1 is attained as ()() 2 ,,;0,, hp wp it it hpit wpit fT T n u u =Σ (76) where () ( ) () 1 , 11 hp hp wp hp wp hp wp wp hpit it it it it it wpit it it it it it u T YY Y Y u T YY Y Y αμ αμ αα − =− + + + = − + + + ++ 70 and () 21 ,;0, hpit wpit nu u Σ is bivariate normal density with zero means and covariance matrix 1 . Σ Case 2: 0, 0, 0. hp wp ph it it it TT T => > I assumed that the husband’s parents enter and play if 0 hp it T = in the model. Thus, I have three index functions. () () ()() () () () () () , 1 1 , 1 1 . 1 hp hp hp wp it it it it it hpit wp wp hp it it hp it wp it wpit hp hp hp wp hp hp it it it ph it phit hp hp TY Y Y Y u YY Y Tu YYY Tu αμ α μαβ β μ βμαβ μ βμαβ μ βμαβ μ =− + + + + −+−+ =+ +− + +− − + =+ +− + (77) The contribution to the likelihood function for case 2 is ()() 3, 2 0, , , ;0, , hpit z wp ph it it hpit wpit phit hpit f TT n u u u du −∞ =Σ ∫ (78) where () ( ) ( )() () 1 ,, 11 wp wp hp it it hp it hp hp wp wp hpit it it it it wpit it hp hp YY Y zY Y Y Y u T μαβ β μ αμ αβμαβμ −+−+ =− + + = − ++−+ () () ( ) () 1 1 hp wp hp hp it it it ph phit it hp hp YYY uT βμαβ μ βμαβ μ +− − + =− +− + and ( ) 3, 2 ,;0, hpit wpit phit nu u u Σ is trivariate normal density with zero means and covariance matrix 2 . Σ 71 Case 3: 0, 0, 0. hp wp ph it it it TT T => = I have the same index functions as in case 2. The contribution to the likelihood function for case 3 is () () 3, 2 0, ,0 , ;0, , hpit phit zz wp it hpit wpit phit phit hpit fTnuuu dudu −∞ −∞ =Σ ∫∫ (79) where () ( ) ( ) () () 1 , 11 hp wp hp hp it it it hp hp wp hpit it it it it phit hp hp YYY zY Y Y Y z βμαβ μ αμ αβμαβμ +− − + =− + + =− ++−+ and () ( ) ( ) () 1 . 1 wp wp hp it it hp it wp wpit it hp hp YY Y uT μαβ β μ βμαβ μ −+−+ =− +− + Case 4: 0, 0, 0. hp wp pw it it it TT T >= > Case 4 is totally symmetric to case 2. I assumed that the wife’s parents enter and play if 0. wp it T = I have three index functions. () ( ) ( ) () () () () ()() () 1 , 1 1 , 1 1 . 1 hp hp hp it it wp it hp it hpit wp wp wp wp hp wp it it it it it wpit wp hp wp wp it it it pw it pwit wp wp YY Y Tu TY Y Y Y u YYY Tu μα β β μ βμαβ μ αμ α βμαβ μ βμαβ μ +− + − =+ ++− − =− + + + + +−− + =+ ++− (80) The contribution to the likelihood function for case 4 is 72 ()() 33 ,0, , , ;0, , wpit z hp pw it it hpit wpit pwit wpit fT T n u u u du −∞ =Σ ∫ (81) where () ( ) ( ) () () () 1 1 ,, 11 hp hp wp it it wp it hp wp hp wp hpit it wpit it it it it wp wp YY Y uT z Y Y Y Y μα β β μ αμ βμαβ μ α +− +− − =− = − + + ++− + () ( ) ( ) () 1 1 wp hp wp wp it it it pw pwit it wp wp YYY uT βμαβ μ βμαβ μ +−− + =− ++− and ( ) 33 ,, ;0, hpit wpit pwit nu u u Σ is trivariate normal density with zero means and covariance matrix 3 . Σ Case 5: 0, 0, 0. hp wp pw it it it TT T >= = The index functions remain same as in case 4. The contribution to the likelihood function for case 5 is () () 33 ,0,0 , , ;0, , wpit pwit zz hp it hpit wpit pwit pwit wpit f T n u u u du du −∞ −∞ =Σ ∫∫ (82) where () ( ) ( ) () () () 1 1 , 11 hp hp wp it it wp it hp wp hp wp hpit it wpit it it it it wp wp YY Y uT z Y Y Y Y μα β β μ αμ βμαβ μ α +− +− − =− = − + + ++− + and () ( ) ( ) () 1 . 1 wp hp wp wp it it it pwit wp wp YYY z βμαβ μ βμαβ μ +−− + =− ++− 73 Case 6: 0, 0, 0, 0. hp wp ph pw it it it it TT T T == > > When couple makes no transfers to either set of parents I assumed that both sets of parents play in the model. Hence, I have four index functions. () () () () () () , 1 1 , 1 , 1 1 . 1 hp hp hp wp it it it it it hpit wp wp hp wp it it it it it wpit ph hp hp wp it it it it it phit wp pw wp hp wp it it it it it pwit wp TY Y Y Y u TY Y Y Y u TY Y Y Y u TY Y Y Y u α μ α αμ α μ β μ β =− + + + + − =− + + + + =− + + + + − =− + + + + (83) The contribution to the likelihood function for case 6 is ()() 44 0,0, , , , , ;0, , hpit wpit zz ph pw it it hpit wpit phit pwit wpit hpit f TT n u u u u du du −∞ −∞ =Σ ∫∫ (84) where () ( ) () 1 ,, 11 hp hp wp wp hp wp hpit it it it it wpit it it it it zY Y Y Y z Y Y Y Y αμ αμ αα − = − ++ = − ++ ++ () ( ) () 1 , 11 ph hp hp wp pw wp hp wp phit it it it it it pwit it it it it it hp wp u T Y Y YY u T Y Y YY μ μ ββ − =− + + + = − + + + ++ and () 44 ,, , ;0, hpit wpit phit pwit nu u u u Σ is four-variate normal density with zero means and covariance matrix 4 Σ which has sub covariance matrix 1 , Σ 2 Σ and 3 . Σ 74 Case 7: 0, 0, 0, 0. hp wp ph pw it it it it TT T T == > = The contribution to the likelihood function for case 7 is () () 44 0,0, ,0 , , , ;0, , hpit wpit pwit zz z ph it hpit wpit phit pwit pwit wpit hpit f T n u u u u du du du −∞ −∞ −∞ =Σ ∫∫ ∫ (85) where () ( ) () 1 ,, 11 hp hp wp wp hp wp hpit it it it it wpit it it it it zY Y Y Y z Y Y Y Y αμ αμ αα − = − ++ = − ++ ++ () 1 ph hp hp wp phit it it it it it hp uT Y Y Y Y μ β =− + + + + and ( ) () 1 . 1 wp hp wp pwit it it it it wp zY YY Y μ β − =− + + + + Case 8: 0, 0, 0, 0. hp wp ph pw it it it it TT T T == = > The contribution to the likelihood function for case 8 is () () 44 0,0,0, , , , ;0, , hpit wpit phit zz z ph it hpit wpit phit pwit phit wpit hpit f T n u u u u du du du −∞ −∞ −∞ =Σ ∫∫ ∫ (86) where () ( ) () 1 ,, 11 hp hp wp wp hp wp hpit it it it it wpit it it it it zY Y Y Y z Y Y Y Y αμ αμ αα − = − ++ = − ++ ++ () 1 hp hp wp phit it it it it hp zY YY Y μ β =− + + + + and ( ) () 1 . 1 wp wp hp wp pwit it it it it it wp uT Y Y Y Y μ β − =− + + + + 75 Case 9: 0, 0, 0, 0. hp wp ph pw it it it it TT T T == = = The contribution to the likelihood function for the last case is () () 44 0,0,0,0 , , , ;0, , hpit wpit phit pwit zz z z hpit wpit phit pwit pwit phit wpit hpit f nu u u u du du du du −∞ −∞ −∞ −∞ =Σ ∫∫ ∫ ∫ (87) where () ( ) () 1 ,, 11 hp hp wp wp hp wp hpit it it it it wpit it it it it zY Y Y Y z Y Y Y Y αμ αμ αα − = − ++ = − ++ ++ () 1 hp hp wp phit it it it it hp zY YY Y μ β =− + + + + and ( ) () 1 . 1 wp hp wp pwit it it it it wp zY YY Y μ β − =− + + + + Hence, the complete form of likelihood function is expressed by combining equations (76), (78), (79), (81), (82), (84), (85), (86) and (87). 2.4 Estimation Results 2.4.1 Results for the Model of One-Way Transfers For now I treat μ , the husband’s bargaining power parameter, and α , the children’s altruism parameter as constants. I have substituted the imputed values for parents’ income and have adjusted the standard errors by bootstrapping. I deal with the fact that I have multiple observations on the same people by taking random sample by families, not by family year observations in the bootstrapping. The results are in column 1 of Table 9. The husband’s bargaining power parameter μ is estimated at .47 while the 76 altruism parameter α is estimated at .66. Note that the test of equal bargaining power ( 0 H: .5 μ = ) is not rejected. One might expect the measure of the husband’s greater bargaining power given the disparity between the amount and frequency of transfers to the husband’s parents and the transfers to the wife’s parents shown in Tables 2 and 3 above. However, recall that Table 4 shows that the wife’s parents are wealthier than the husband’s parents. The model suggests that the numbers in Tables 2 and 3 are being driven by lower parental income for the husband. The one-way model also estimates relatively high altruism by children toward parents ( ˆ .66 α = ). This implies that children value an extra payment to the parents at 66% of the value of these funds to themselves. This high altruism may come from the lack of pension support to their parents. 2.4.2 Results for the Model of Two-Way Transfers As in the estimation of the one-way transfer model, I have simply substituted the imputed values for parents’ income and have adjusted the standard errors by bootstrapping. I only allow error terms to be correlated between each pair of child and parents for computational purpose. That is, the error terms in the transfer equations between husband and his parents and the error terms in the transfer equations between wife and her parents as independent. Otherwise, I have to deal with up to four-variate normal distribution function. I will address these issues in future work. Finally I deal with the fact that I have multiple observations on the same people by taking random sample by families, not by family year observations in the bootstrapping as same way in the one- way model. The results are in column 2 of Table 9. The bargaining power between 77 husband and wife appears to be balanced ( ˆ .48 μ = ) and again the equal power hypothesis is not rejected at the standard 5% significance level. The one-way model and two-way model give consistent estimates for the bargaining power parameter. The level of altruism from parents to their children, and from children to the parents, are both estimated high ( ˆˆ ˆ.79, .70, .77 hp wp αβ β == = ). This is consistent with the fact that 78% of the households in the sample report at least one transfer between couple and parents. Further, I cannot reject 0 : wp hp H β β = , which means no difference of parents’ altruism toward children by children’s gender. Lastly, degree of altruisms between parents and children are not statistically different. Table 9: Structural Results for Models for the Transfers. (1) (2) Parameters One-Way Model Two-Way Model μ .4684 .4797 (.0247) (.0106) α .6641 .7943 (.0532) (.0260) hp β .6989 (.0473) wp β .7697 (.0708) Notes: (1) Standard errors are in parentheses. (2) Imputation error is corrected by bootstrapping. 78 2.4.3 Estimation of the Effect of the Education on Bargaining Power It is straightforward to allow bargaining power parameter ˆ μ to depend on the couple’s characteristics. I assume ˆ μ is a function of the husband and wife’s education and estimate the effect of the education on the bargaining power. First, ˆ μ is assumed to depend on the level of the husband and wife’s education. Second, I assume ˆ μ is a function of husband’s relative education level. Last, it is assumed that ˆ μ is a third polynomial function of the husband’s relative education. Table 10 (a), Table 10 (b) and Table 10 (c) summarize the estimation results per each functional form assumption. The results suggest that partners’ relative education is important to determine bargaining power. Table 10 (b) proposes husband’s relative education linearly affects husband’s bargaining power in one way transfers model. However, it appears that the effect is non- linear in the two way transfers model. For comparison of predictions of two models, I assume ˆ μ is a third degree polynomial function of husband’s relative education, and plot ˆ μ against husband’s relative education based on the estimation results in Table 10 (c). Figure 3 and Figure 4 show the plot for each one way and two way model respectively. In both Figures ˆ μ appears to be less affected by education when the education level is similar between husband and wife while the effect is significant when the level difference is considerable. Further, the effect looks to be more non-linear in two way model as shown in Figure 4. However, I have to stress that the discussion of this section is based on my functional form assumptions. 79 Table 10 (a): Structural Results for mu ( ) ˆ μ <Function 1> (1) (2) Parameters One-way Model Two-way Model 0 γ -.8604 -.0243 (.5124) (.2013) 1 γ .0699 .0142 (.0857) (.0146) 2 γ -.0181 -.0193 (.0883) (.0131) Notes: (1) HE: Husband’s Education, WE: Wife’s Education. (2) Standard errors are in parentheses. (3) Imputation error is corrected by bootstrapping. <Function 1>: 1 ˆ 1 exp( ( )) 01 2 HE WE μ γγ γ = +− + + Table 10 (b): Structural Results for mu ( ) ˆ μ <Function 2> (1) (2) Parameters One-way Model Two-way Model 0 γ -2.0427** -.5941* (.8351) (.3399) 1 γ 3.7299** .9952 (1.5393) (.6491) See notes to Table 10 (a). ** p<0.05, * p<0.01 <Function 2>: 1 ˆ 1 exp( ( )) 01 X μ γγ = +− + where . () HE X HE WE = + 80 Table 10 (c): Structural Results for mu ( ) ˆ μ <Function 3> (1) (2) Parameters One-way Model Two-way Model 0 γ -59.37 -36.57 (31.49) (.36) 1 γ 320.28 204.39 (179.33) (.58) 2 γ -577.31 -380.68 (340.30) (1.19) 3 γ 347.54 235.84 (215.16) (2.92) Prob >chi2(3)* (.0479) (.0000) See notes to Table 10 (a). *Test :0. 01 2 3 H γ γγ == = <Function 3>: 1 ˆ 23 1 exp( ( )) 01 2 3 XX X μ γγ γ γ = +− + + + where . () HE X HE WE = + 81 0 .2 .4 .6 .8 mu .3 .4 .5 .6 .7 X=HE/(HE+WE) Figure 3: Plot for mu ( ˆ μ ) with One-way Model 0 .2 .4 .6 .8 mu .3 .4 .5 .6 .7 X=HE/(HE+WE) Figure 4: Plot for mu ( ˆ μ ) with Two-way Model 82 2.5 Conclusions and Future Research In this paper, two models of intergenerational transfers between adult children and old parents are derived assuming a formal collective model framework. This allows me to investigate the respective bargaining power between husbands and wives. My work differs from previous work in that I include both sets of parents, focus on bargaining over a form of semi-private consumption, and estimate a structural bargaining parameter. Overall the results suggest that the husband’s preferences and the wife’s preferences are equally weighted. For my future research I can consider the following extensions. First, I will consider alternative specifications of functional form of ˆ μ . In the bargaining literature the bargaining power depends on each individual’s own income. Raut and Tran argue that the education will also reflect the son’s and daughter’s obligation to their parents varying positively with the child’s obligations. Hence, income may be a better factor to infer bargaining power. In the same way one can make ˆ α a function of the adult children’s family size. 28 One could also relax the assumption that hw α α = and let each child’s altruism depend on their parents age or health. By exploiting advantage of structural estimation we can better understand how bargaining power and degree of altruism are determined. 28 Allowing α and μ depend on the same chrematistics would raise identification issues. 83 Conclusion In this dissertation I developed and estimated several models about intergenerational transfers to old parents from adult children and Vice-versa. I noted that a married couple usually has two sets of parents to support and couple’s transfer decisions can be affected by husband and wife’s different preferences toward own parents and in-laws. In the estimation I dealt with a bunching of zero transfers. The models in the dissertation did not consider exchange motives in the transfer behaviors. This is left for my future research. 84 Bibliography Altonji, J.G., Hayashi, F. and Kotlikoff, L.J. (1993). “Is the extended family altruistically linked? Direct tests using micro data.” American Economic Review, Vol. 82, pp. 1177–1198. Altonji, J.G., Hayashi, F. and Kotlikoff, L.J. (1997). “Parental altruism and inter- vivos transfers: Theory and evidence.” Journal of Political Economy, Vol. 105, pp. 1121–1166. Altonji, J.G., Ichimura, H. and Otsu, T. (2008). “Estimating Derivatives in Nonseparable Models with Limited Dependent Variables.” Cowles Foundation Discussion Paper No. 1668 Amemiya, Takeshi. (1978). “The estimation of a simultaneous equation generalized Probit model.” Econometrica, Vol. 46, No. 4, pp. 1193–1105. Amemiya, Takeshi. (1979). “The Estimation of a Simultaneous-Equation Tobit Model.” International Economic Review, V ol. 20, No. 1, pp. 169-181. Barro, Robert J. (1974). “Are government bonds net wealth?” Journal of Political Economy, Vol. 82, pp. 1095-1117. Becker, Gary S. (1974). “A theory of social interactions.” Journal of Political Economy, Vol. 82, pp. 1063-1093. Becker, Gary S. (1991). “A Treatise on The Family.” Harvard University Press, Cambridge. Behrman, Jere R. and Mark R. Rosenzweig. (2006). “Parental Wealth and Adult Children's Welfare in Marriage.” The Review of Economics and Statistics, V ol. 88, No. 3, pp. 496-509. Blundell, Richard, Pierre-Andr´e Chiappori and Costas Meghir. (2005). “Collective Labor Supply With Children.” Journal of Political Economy, Vol. 113, No. 6, pp. 1277- 1306. Blundell, Richard and Richard Smith. (1986). “An Exogeneity Test for the Simultaneous Equation Tobit Model.” Econometrica, Vol. 54, pp. 679-685. Brown, Philip H. (2008). “Dowry and Intrahousehold Bargaining: Evidence from China.” The Journal of Human Resources (Forthcoming). 85 Chiappori, Pierre. (1988). “Rational Household Labor Supply.” Econometrica, Vol. 56, No. 1, pp. 63–90. Chiappori, Pierre. (1992). “Collective Labor Supply and Welfare.” Journal of Political Economy, Vol. 100, No. 3, pp. 437–467. Cox, Donald. (1987). “Motives for private income transfers.” Journal of Political Economy, Vol. 95, No. 3, pp. 508-46. Cox, D., Galasso, E., Jimenez, E. (2006). “Private transfers in a cross section of developing countries.” WP 2006-2. Center for Retirement Research, Boston College, Chestnut Hill, MA. Cox, Donald and Marcel Fafchamps. (2008). “Extended Family and Kinship Networks: Economics Insights and Evolutionary Direction.” Handbook of Development Economics, Vol. 4, Chapter 58, pp. 3711-3784. Greenhalgh, Susan. (1985). “Sexual Stratification: The Other Side of ‘Growth with Equity’ in East Asia.” Population and Development Review, Vol. 11, No. 2, pp. 265–314. Duflo, Esther. (2003). “Grandmothers and Granddaughters: Old Age Pension and Intra-household Allocation in South Africa." World Bank Economic Review, V ol. 17, No. 1, pp. 1-25. Kazianga, H. (2006). “Motives for household private transfers in Burkina Faso.” Journal of Development Economics, Vol. 79, 73–117. Khemani, Stuti. (1999). “Inter-generational Transfers and Intra-household bargaining: evidence from Indonesia.” mimeo, World Bank. pp. 1-44. Lee, Yean-Ju, William L. Parish and Robert J. Willis. (1994). “Sons, Daughters, and Intergenerational Support in Taiwan.” American Journal of Sociology, V ol. 99, pp. 1010- 1041. Lillard, Lee A. and Robert J. Willis. (1997). “Motives for Intergenerational Transfers: Evidence from Malaysia." Demography, V ol. 34 No. 1, pp. 115-134. Lundberg, Shelly J., Robert A. Pollak and Terence J. Wales. (1996). “Do Husbands and Wives Pool their Resources? Evidence from the United Kingdom Child Benefit.” The Journal of Human Resources, V ol. 32, No. 4, pp. 463–480. Manser, M. and Brown, M. (1980). “Marriage and Household Decision-Making: A Bargaining Analysis.” International Economic Review, Vol. 21 No. 1, pp. 31–44. 86 Mazzocco, Maurizio. (2006). “Parents’ Preferences for Expenditure on Children When At LeastOne Parent Works and Preferences Are Non-separable.” Mimeographed, Department of Economics, UCLA. Mazzocco, Maurizio. (2007). “Household Intertemporal Behaviour: A Collective Characterization and a Test of Commitment.” Review of Economic Studies, Vol. 74, No. 3, pp. 857-895. McElroy, M. B. and M. J. Horney (1981). “Nash-Bargained Household Decisions: Toward a Generalization of the Theory of Demand.” International Economic Review, Vol. 22, No. 2, pp. 333-349. Raut, L.K., Tran, L.H. (2005). “Parental human capital investment and old-age transfers from children: Is it a loan contract or reciprocity for Indonesian families?” Journal of Development Economics, Vol. 77, pp. 389–414. Ravallion, M., Dearden, L. (1988). “Social security in a moral economy: An empirical analysis for Java.” The Review of Economics and Statistics, Vol. 70, pp. 36–44. Rosett, R., (1959). “A statistical model of friction in economics.” Econometrica 27 (1), 263–267. Schoeni, Robert F. (2000). “Support networks within the family as a public good problem.” Mimeographed, Rand Labor and Population Program Seok, J. and T. Kim (2000). “The Reality of Elderly Income and Policies for the Improvement of Income Security System.” (in Korean) Korea Institute for Health and Social Affairs, p. 35. Thomas, Duncan. (1994). “Like Father, Like Son: Like Mother, Like Daughter: Parental Resources and Child Height.” The Journal of Human Resources, V ol. 24, No. 4, pp. 950-88. Thomas, D., D. Contreras, and E. Frankenberg. (2002). “Distribution of Power Within the Household and Child Health.” Mimeographed, Department of Economics, UCLA. Udry, C. (1996). “Gender, agricultural production and the theory of the household.” Journal of Political Economy 104 (5), 1010– 1046. Xie, Yu and Haiyan Zhu. (2006). “Do sons or daughters give more money to parents? Gender and intergenerational support in contemporary urban China.” Population Studies Center Research Report No. 06-607, University of Michigan. 87 Zeyu, Xu. (2007). “A survey on intra-household models and Evidence.” Munich Personal RePEc Archive, Paper No. 3763, University Library of Munich, Germany.

Abstract (if available)

Abstract My dissertation extends the literature on intergenerational transfers between adult married children and their parents in two ways. First, for a given couple, both sets of parents enter the optimization problem. Second, I develop and estimate models where amount of transfers to the husband's (wife's) parents is assumed to depend on the couple's income, the husband's parents' income, the wife's parents' income, and the bargaining power of husband (wife) within the family. Transfers from adult children to parents are quite important in developing countries

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Measurement error in income and consumption data and economic mobility

PDF

Essays on labor and development economics

PDF

Essays on family and labor economics

PDF

Trade, training, employment, and wages: evidence from the U.S. manufacturing industry

PDF

An empirical analysis of the quality of primary education across countries and over time

PDF

The EITC, labor supply, and child development

PDF

Essays on the empirics of risk and time preferences in Indonesia

PDF

Essays on child labor and poverty in the context of a conditional cash transfer program in Nicaragua

PDF

Couples’ neuroendocrine activity in response to family conflict discussions: the role of self-reported anger and previous marital aggression

PDF

Pathways of local economic development: tales of cities in the United States and South Korea

PDF

Institutional contestation, network legitimacy and organizational heterogeneity: interactions between government and environmental nonprofits in South Korea

PDF

A formative evaluation of the student support services TRIO program for low income and first generation college bound students self-efficacy at Butte-Glenn Community College District

PDF

Dynamic analyses of the interrelationship between mothers and daughters on a measure of depressive symptoms

PDF

Environmental policy reforms in a global and regional world: mimetic, diffusive, and coercive policy adoptions in France and Korea

PDF

The relationship between women's household responsibilities and commute lengths: a study on women in the US and Great Britain

PDF

What leads to a happy life? Subjective well-being in Alaska, China, and Australia

Asset Metadata

Creator Song, Heonjae (author)

Core Title Inter-generational tranfsers and bargaining power within the family in South Korea

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 08/06/2009

Defense Date 06/19/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag bargaining power,inter-generational tranfsers,OAI-PMH Harvest

Place Name South Korea (countries)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ham, John C. (committee chair), Painter, Gary D. (committee member), Strauss, John (committee member)

Creator Email heonjaes@kipf.re.kr,heonjaes@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2507

Unique identifier UC1271815

Identifier etd-SONG-3094 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-175385 (legacy record id),usctheses-m2507 (legacy record id)

Legacy Identifier etd-SONG-3094.pdf

Dmrecord 175385

Document Type Dissertation

Rights Song, Heonjae

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu