Review of Income and Wealth Series 43.Number 4,December 1997 PRIVATE INTERHOUSEHOLD TRANSFERS OF MONEY AND TIME: NEW EMPIRICAL EVIDENCE BY ROBERT F.SCHOENI RAND There is a growing awareness that it is important to understand patterns of family assistance;however, there is still a great deal of information about private transfers that is not known.This study begins to fill this void by presenting results from a new survey and integrating these findings with evidence from recent studies that use other new data sets.It is found that:(i)a large share of households participate in private transfer networks.(ii)a greater amount of financial assistance is provided to lower income family members,(iii)altruism does not fully explain transfer behavior,and (iv)people in their 20s and 30s receive more assistance than people of other ages,even the very old. Within the family workers are born,goods are produced,tastes are formed, decisions to work are made,and resources are redistributed.This paper investi- gates the last of these roles,that is,the transfer of resources (i.e.,money and time help)among family members and friends.Understanding private transfers is important for a number of reasons.First,they provide a means through which individuals can transmit their well-being to others.Second,private transfers have potential consequences for the effectiveness of government redistribution policies (Barro,1974;Becker,1974;Roberts,1984;Andreoni,1988;Bernheim and Bagwell,1988;Laitner,1988;Bergstrom,1989;Andreoni,1989;Bruce and Wald- man,1990;Kotlikoff et al.,1990;Altonji et al.,1994).For example,if publicly provided benefits to an individual become more generous,then that individual's family members and friends may respond by decreasing the amount of private assistance they give to the individual.Third,intergenerational flows of resources within the family have been identified as one determinant of fertility (Caldwell, 1976;Willis,1982).The greater the flows of resources from children to parents, the higher will be desired fertility.As flows begin to reverse direction,going from parents to children,fertility will fall. Private transfers may also be one mechanism through which families transmit inequality across generations(Becker and Tomes,1979;Behrman,Pollack,and Taubman,1990;Menchik,1980;and Tomes,1981).Wealthy families may give larger intergenerational assistance,including bequests,leading to persistent inequi- ties.At the same time,families may give conpensatory transfers to their least wealthy members,which would mitigate inequality.Finally,intergenerational Nore:I thank Ted Bergstrom,Paul Courant,Sheldon Danziger,David Lam,Gary Solon,the editor,and two anonymous referees for valuable comments.Financial support was received from the National Institute on Aging (K01-AG00670). Unless otherwise indicated,throughout the paper private transfers will refer to interhousehold transfers of money and time.There is very little data on intrahousehold transfers. 423
Review of Income and Wealth Series 43, Number 4, December 1997 PRIVATE INTERHOUSEHOLD TRANSFERS OF MONEY AND TIME: NEW EMPIRICAL EVIDENCE BY ROBERT F. SCHOENI RAND There is a growing awareness that it is important to understand patterns of family assistance; however, there is still a great deal of information about private transfers that is not known. This study begins to fill this void by presenting results from a new survey and integrating these findings with evidence from recent studies that use other new data sets. It is found that: (i) a large share of households participate in private transfer networks, (ii) a greater amount of financial assistance is provided to lower income family members, (iii) altruism does not fully explain transfer behavior, and (iv) people in their 20s and 30s receive more assistance than people of other ages, even the very old. Within the family workers are born, goods are produced, tastes are formed, decisions to work are made, and resources are redistributed. This paper investigates the last of these roles, that is, the transfer of resources (i.e., money and time help) among family members and friends.' Understanding private transfers is important for a number of reasons. First, they provide a means through which individuals can transmit their well-being to others. Second, private transfers have potential consequences for the effectiveness of government redistribution policies (Barro, 1974; Becker, 1974; Roberts, 1984; Andreoni, 1988 ; Bernheim and Bagwell, 1988 ; Laitner, 1988 ; Bergstrom, 1989 ; Andreoni, 1989 ; Bruce and Waldman, 1990; Kotlikoff et al., 1990; Altonji et a]., 1994). For example, if publicly provided benefits to an individual become more generous, then that individual's family members and friends may respond by decreasing the amount of private assistance they give to the individual. Third, intergenerational flows of resources within the family have been identified as one determinant of fertility (Caldwell, 1976; Willis, 1982). The greater the flows of resources from children to parents, the higher will be desired fertility. As flows begin to reverse direction, going from parents to children, fertility will fall. Private transfers may also be one mechanism through which families transmit inequality across generations (Becker and Tomes, 1979; Behrman, Pollack, and Taubman, 1990; Menchik, 1980; and Tomes, 1981). Wealthy families may give larger intergenerational assistance, including bequests, leading to persistent inequities. At the same time, families may give conpensatory transfers to their least wealthy members, which would mitigate inequality. Finally, intergenerational Note: I thank Ted Bergstrom, Paul Courant, Sheldon Danziger, David Lam, Gary Solon, the editor, and two anonymous referees for valuable comments. Financial support was received from the National Institute on Aging (K01-AG00670). 'unless otherwise indicated, throughout the paper private transfers will refer to ittterhousehold transfers of money and time. There is very little data on intrahousehold transfers
transfers are important because of the role they may play in determining savings and in the accumulation of wealth (Modigliani,1988;Kotlikoff,1988).The Life Cycle Model of Savings claims that savings are accumulated primarily for retire- ment,not for intergenerational transfers,and the proportion of wealth due to private transfers has been empirically analyzed,with a wide range of estimates being identified.2 Despite the fact that there are a number of reasons for studying private transfers,historically there has been little reliable empirical information on this behavior,at least in the U.S.Indeed,the most basic information such as the frequency and size of private transfers has not been well established.This has changed in the past 5-10 years,with several new surveys providing information on private transfers from nationally representative samples.The objective of this paper is to integrate empirical findings from this new and burgeoning literature and provide additional estimates from one data set,the 1988 Panel Study of Income Dynamics,which arguably has the highest quality information on transfers among all nationally representative data sets in the U.S. The paper begins with a discussion of the behavioral models that have been posited to explain private transfers.This section is followed by a summary of findings from previous empirical analyses.The subsequent sections describe the 1988 Panel Study of Income Dynamics and the empirical results based on these data.A concluding section synthesizes the new results and the earlier findings in an attempt to establish a group of"stylized facts"regarding private interhousehold transfers. 1.BEHAVIORAL MODELS OF TRANSFERS Several models of transfer behavior have been posed,including altruism, exchange,"warm glow,"and insurance.The altruism model (Becker,1974;Barro, 1974)states,in terms of parent-child relations,that the parent's well-being is directly related to the well-being of her child (i.e.,U=U(Y,U),where Y are goods consumed by the parent and U.is the utility of the child).The model predicts that parents will decrease the amount of assistance provided to their children in response to increases in the children's income.Altruism also implies that if parents are initially transferring some positive amount to a child,and the parent's income increases by SI while the child's income decreases by $l,then the parent will transfer an additional dollar to the child.In this framework,the coefficient on the parent's income less the coefficient on the child's income should sum to one (Cox and Rank,1992;Altonji et al.,1994). The altruism model has been extended by Andreoni (1989)to include simul- taneous"warm glow"giving.That is,parents not only care about the well-being of their children,they care about the amount of gifts they give their children. Moreover,if the behavior is only motivated by warm glow,then the amount of the transfer given to the child is independent of the characteristics of the child. A third model,exchange,has been the most widely analyzed alternative to altruism 'Kotlikoff and Summers'(1981)estimates are 45-80 percent,while other estimates are Iss than 25 percent 424
transfers are important because of the role they may play in determining savings and in the accumulation of wealth (Modigliani, 1988; Kotlikoff, 1988). The Life Cycle Model of Savings claims that savings are accumulated primarily for retirement, not for intergenerational transfers, and the proportion of wealth due to private transfers has been empirically analyzed, with a wide range of estimates being identified.2 Despite the fact that there are a number of reasons for studying private transfers, historically there has been little reliable empirical information on this behavior, at least in the U.S. Indeed, the most basic information such as the frequency and size of private transfers has not been well established. This has changed in the past 5-10 years, with several new surveys providing information on private transfers from nationally representative samples. The objective of this paper is to integrate empirical findings from this new and burgeoning literature and provide additional estimates from one data set, the 1988 Panel Study of Income Dynamics, which arguably has the highest quality information on transfers among all nationally representative data sets in the U.S. The paper begins with a discussion of the behavioral models that have been posited to explain private transfers. This section is followed by a summary of findings from previous empirical analyses. The subsequent sections describe the 1988 Panel Study of Income Dynamics and the empirical results based on these data. A concluding section synthesizes the new results and the earlier findings in an attempt to establish a group of "stylized facts" regarding private interhousehold transfers. Several models of transfer behavior have been posed, including altruism, exchange, "warm glow," and insurance. The altruism model (Becker, 1974; Barro, 1974) states, in terms of parent-child relations, that the parent's well-being is directly related to the well-being of her child (i.e., U,>= U,(X,, U,), where X, are goods consumed by the parent and U, is the utility of the child). The model predicts that parents will decrease the amount of assistance provided to their children in response to increases in the children's income. Altruism also implies that if parents are initially transferring some positive amount to a child, and the parent's income increases by $1 while the child's income decreases by $1, then the parent will transfer an additional dollar to the child. In this framework, the coefficient on the parent's income less the coefficient on the child's income should sum to one (Cox and Rank, 1992; Altonji et al., 1994). The altruism model has been extended by Andreoni (1989) to include simultaneous "warm glow" giving. That is, parents not only care about the well-being of their children, they care about the amount of gifts they give their children. Moreover, if the behavior is only motivated by warm glow, then the amount of the transfer given to the child is independent of the characteristics of the child. A third model, exchange, has been the most widely analyzed alternative to altruism '~othkoff and Summers' (1981) estimates are 45-80 percent, while other estimates are Iss than 25 percent
(Cox,1987;Bernheim et al.,1985;Cox and Rank,1992;Cox and Jakubson, 1995).The basic presumption is that,using the parent-child notation again, children provide something to their parents,such as assistance in old age,a sympa- thetic ear,or contemporaneous help in household production,and in return par- ents give their children cash.The parent-child relationship can be viewed as a market transaction where the parent demands services,which perhaps only the child can provide or for which there are no close market substitutes,and the child provides services in return for remuneration.As a result,the relationship between the income of the children and the amount of assistance they receive from their parents is a function of the elasticities of supply and demand for the services provided by the child,and it could be either positive,negative,or zero (Cox, 1987).In sum,one test of these models is to examine the effects of child's income on the amount of transfers they receive,with altruism predicting a negative effect, (strict)"warm glow"predicting no effect,and the exchange model being consistent with any relationship." Transfers may also be used to smooth consumption across time and to over- come liquidity constraints (Kotlikoff and Spivak,1981;Cox,1990;Cox and Jappelli,1990).The family may provide insurance during periods of unemploy- ment or low income,and parents may assist their liquidity constrained children with purchasing a home or financing their schooling.Each of these factors is addressed in the empirical analyses. Although these models have been discussed in terms of income,in general, anything that effects the well-being of the child or parent may alter transfers. Examples of these factors include health status and the existence of grandchildren within an adult child's household.If the adult child has children of their own. and their parent cares about these grandchildren,then the amount of assistance would be altered (i.e.,increased under altruism).Or if a family member is in poor health,which in turn lowers their well-being,then altruistically motivated family members would increase the amount of assistance they give to the person in poor health. The approach in this paper is to investigate the empirical support for these various models.To this end,these models provide a conceptual framework for the empirical specification.The general specification which subsumes each of these models is one in which indicators of the well-being of the potential donor (e.g., parent)and the well-being of the potential recipient (e.g.,adult child)are included. In addition,the exchange model implies that the time value of potential providers of services (i.e.,children)may also be important.The models that we estimate will incorporate various measures of well-being,including income and health, allowing us to discriminate among some of the models. II.PREVIOUS STUDIES Frequency and Magnitude of Private Interhousehold Transfers Table I summarizes the evidence on the frequency and magnitude of inter vivos financial transfers reported in studies that have examined nationally One weakness of the exchange model is that its prediction of the effect of income on transfers cannot be refuted. 425
(Cox, 1987; Bernheim et ul., 1985; Cox and Rank, 1992; Cox and Jakubson, 1995). The basic presumption is that, using the parent-child notation again, children provide something to their parents, such as assistance in old age, a sympathetic ear, or contemporaneous help in household production, and in return parents give their children cash. The parent-child relationship can be viewed as a market transaction where the parent demands services, which perhaps only the child can provide or for which there are no close market substitutes, and the child provides services in return for remuneration. As a result, the relationship between the income of the children and the amount of assistance they receive from their parents is a function of the elasticities of supply and demand for the services provided by the child, and it could be either positive, negative, or zero (Cox, 1987). In sum, one test of these models is to examine the effects of child's income on the amount of transfers they receive, with altruism predicting a negative effect, (strict) "warm glow" predicting no effect, and the exchange model being consistent with any re~ationship.~ Transfers may also be used to smooth consumption across time and to overcome liquidity constraints (Kotlikoff and Spivak, 1981; Cox, 1990; Cox and Jappelli, 1990). The family may provide insurance during periods of unemployment or low income, and parents may assist their liquidity constrained children with purchasing a home or financing their schooling. Each of these factors is addressed in the empirical analyses. Although these models have been discussed in terms of income, in general, anything that effects the well-being of the child or parent may alter transfers. Examples of these factors include health status and the existence of grandchildren within an adult child's household. If the adult child has children of their own, and their parent cares about these grandchildren, then the amount of assistance would be altered (i.e., increased under altruism). Or if a family member is in poor health, which in turn lowers their well-being, then altruistically motivated family members would increase the amount of assistance they give to the person in poor health. The approach in this paper is to investigate the empirical support for these various models. To this end, these models provide a conceptual framework for the empirical specification. The general specification which subsumes each of these models is one in which indicators of the well-being of the potential donor (e.g., parent) and the well-being of the potential recipient (e.g., adult child) are included. In addition, the exchange model implies that the time value of potential providers of services (i.e., children) may also be important. The models that we estimate will incorporate various measures of well-being, including income and health, allowing us to discriminate among some of the models. Frequency and Magnitude of Private Interhousehold Trun~fers Table 1 summarizes the evidence on the frequency and magnitude of inter vivos financial transfers reported in studies that have examined nationally 'one weakness of the exchange model is that its prediction of the effect of income on transfers cannot be refuted. 425
representative samples of individuals or households.Despite differences in survey design,some common patterns emerge.'Approximately 8 to 20 percent of house- holds receive financial assistance in a given year,while a slightly higher share of households report giving financial help.Among all households,not just those receiving assistance,the average amount of help received annually is $300 to $500 [expressed in 1987 dollars,which is the year of the transfer supplement to the Panel Study of Income Dynamics (PSID)].Moreover,the distribution of mone- tary transfers is highly skewed (Rosenzweig and Wolpin,1990;Cox and Raines, 1985;Gale and Scholz,1991) TABLE I FINDINGS FROM PREVIOUS STUDIES Reporting Mean Amount Received Period for Share Receiving (Given)Conditional on Authors Data Transfers (Giving) Positive Transfer* Moon (1983) 1978 Panel Study of Past Year 8.4%(98%) s3.753($2.977) Income Dynamics Cox and 1979 President's Past Year 12.3%(15.8% 53.144(S3,256) Raines (1985) Commission on Pension Policy Morgan (1985) 1980 Panel Study of Past Five 22.0%(29.0%0 NR Income Dynamics Years MacDonald 1988 National Survey of Past Five 16.8%(19.5%)Gifts s5,592(S7.081)Gifts (1990) Families and Years 11.5%(19.4%)L0ans s6,334(S6,157)L0ans Households 2.8%Home Assist. S11.381 (NR)Home Assist. Gale and 1983 86 Survey of Past Three 5.3%9.4%) sf6.247(17.714) Scholz (1991) Consumer Finances Years Cox and 1987/88 National Past Five 25% S6.46I Rank(1992) Survey of Families Years and Households Altonji et al. 1988 Panel Study of Past Year 19.49% $1.459 (1996) Income Dynamics Nores:*Expressed in 1987 dollars and excluding bequests.~Includes only transfers of at least S3.000 over the three year period.NR-not reported.Morgan's estimates are for"emergency help,"and Altonji eal.'s estimates are for transfers from parents to children. An alternative form of transfer is bequests,and in Cox and Raines'(1985) data bequests are received by just 0.8 percent of the respondents in the single year,and they account for 25 percent of the total amount of transfer dollars received.MacDonald (1990)reports similar magnitudes with the NSFH;bequests are received by 1.4 percent of the respondents over the five year period,and they account for 19.2 percent of the total amount of private transfers received. 'Analyses using the National Longitudinal Survey.Health and Retirement Survey,and Asset and Health Dynamics Survey are not included in Table I because these data are not representative of all ag℃s. For example,the analyses reported by Gale and Scholz using the Survey of Consumer Finances is restricted to transfers of at least S3,000 over the previous three years.Moreover,while the National Survey of Families and Households collects information on transfers made in the previous five years. most other studies collect information on transfers in the previous one year. "The average amount of assistance received from all three types of transfers in the National Survey of Families and Houscholds is S1.986 over the five year period,which,if annualized by dividing by five.is $397. 426
representative samples of individuals or housel~olds.~ Despite differences in survey design, some common patterns emerge.5 Approximately 8 to 20 percent of households receive financial assistance in a given year, while a slightly higher share of households report giving financial help. Among all households, not just those receiving assistance, the average amount of help received annually is $300 to $500 [expressed in 1987 dollars, which is the year of the transfer supplen~ent to the Panel Study of Income Dynamics (PSID)].~ Moreover, the distribution of monetary transfers is highly skewed (Rosenzweig and Wolpin, 1990; Cox and Raines, 1985; Gale and Scholz, 1991). TABLE 1 Authors Data Reporting Mean Amount Received Period for Share Receiving (Given) Conditional on Transfers (Giving) Positive Transfer* Moon (1983) 1978 Panel Study of Past Year 8.4% (9 8%) $3,753 ($2,977) Income Dynam~cs Cox and 1979 President's Past Year 12.3% (15.8%) $3,144 ($3,256) Raines (1985) Commission on Pension Policy Morgan (1985)A 1980 Panel Study of Past Five 22.0% (29.0%) NR Income Dynamics Years MacDonald 1988 National Survey of Past Five 16.8% (19.5'%) Gifts $5,592 ($7,081) Gifts (1990) Families and Years 11.5'%, (19.4%) Loans $6,334 ($6,157) Loans Households 2.8% Home Assist. $1 1,381 (NR) Home Assist. Gale and 1983 86 Survey of Past Three -5.3%(9.4%) -$16,247 ($17,714) Scholz (1991) Consumer Finances Years Cox and 1987/88 National Past Five 25.9% $6,46 1 Rank (1992) Survey of Families Years and Households Altonji el 01. 1988 Panel Study of Past Year 19.4% $1,459 (1996)A Income Dynamics Notes: *Expressed in 1987 dollars and excluding bequests. -Includes only transfers of at least $3,000 over the three year period. NR =not reported. organ's estimates are for "emergency help," and Altonji et al.'s estimates are for transfers from parents to children. An alternative form of transfer is bequests, and in Cox and Raines' (1985) data bequests are received by just 0.8 percent of the respondents in the single year, and they account for 25 percent of the total amount of transfer dollars received. MacDonald (1990) reports similar magnitudes with the NSFH; bequests are received by 1.4 percent of the respondents over the five year period, and they account for 19.2 percent of the total amount of private transfers received. 4Analyses using the National Longitudinal Survey, Health and Retirement Survey, and Asset and Health Dynamics Survey are not included in Table 1 because these data are not representative of all ages. 5~or example, the analyses reported by Gale and Scholz using the Survey of Consumer Finances is restricted to transfers of at least $3,000 over the previous three years. Moreover, while the National Survey of Families and Households collects information on transfers made in the previous five years, most other studies collect information on transfers in the previous one year. he average amount of assistance received from all three types of transfers in the National Survey of Families and Households is $1,986 over the five year period, which, if annualized by dividing by five, is $397
Multivariate Models The direction of the relationship between recipient's income and the amount of money received has been identified as a test of the altruism model,with the altruism model predicting that as the income of a recipient increases,ceteris par- ibus,the amount of private transfers received will decrease (Becker,1981).Cox (1987)and Cox and Rank(1992)analyzed this relationship empirically and find a positive relationship between the amount of assistance received and the(potential) recipient's income.The findings of Cox (1987)and Cox and Rank (1992)are corroborated by those of MacDonald (1990).3.8 However,other studies found a negative relationship between the amount (or probability)of transfers received and income.Rosenzweig and Wolpin's(1990) point estimates imply that a $5,000 increase in the adult child's earnings reduces the probability of co-residing by 11.1 percent and reduces the probability of receiving a monetary transfer while not residing at home by 10.9 percent.In a recent paper using the 1988 PSID,Altonji et al.,(1996)specify a Tobit model and find that the respondent's income has a negative effect on the amount of transfers received from parents.Shelton and Sueyoshi(1993),using information on private transfers collected annually in the PSID,finding that having the lowest household income among the households in the family increases the probability of receiving private transfers from 11 percent to 20 percent.Work by McGarry and Schoeni (1995,1996)using the Health and Retirement Survey and the Asset and Health Dynamics Survey finds that larger financial transfers are given to adult children with lower income,and this result holds when they look within families by controlling for family fixed effects.Using the Asset and Health Dynam- ics Survey,Dunn and Phillips(1995)also find that inter vivos transfers are more likely to be given to poorer children within a family,but that children of different income levels are equally likely to receive parental transfers at the time of the death of a parent. Most studies have found that individuals with more years of schooling both give and receive greater amounts of money transfers (McDonald,1990;Cox and Raines,1985).However,McGarry and Schoeni (1995,1996)demonstrate that the effects of schooling are mitigated substantially when family fixed effects are controlled for by examining differences in transfers among siblings.This pattern is consistent with the hypothesis that parents who make larger transfers to their children are also parents who invest more in their children's education. A popular belief about black families is that they have a more active support network than white families.However,with regard to interhousehold assistance, most recent studies have not found support for this belief(MacDonald,1990; Silverstein and Waite,1992).For example,MacDonald (1990)shows that,among MacDonald(1990)uses a two-step estimation procedure,in which he excludes life-course events (divorce,marriage,births,home-leaving,non-work and non-school spells)from the Probit model. Furthermore,he estimates a Tobit regression as an alternative to the two-step procedure and does not find consistently positive effects of recipient's income. "In Cox's(1987)analyses of the PCPP,only characteristics of the respondent are available.Using the NSFH,Cox and Rank (1992)are able to control for parent's income.Both studies use a two- step estimation procedure. 427
Multivariate Models The direction of the relationship between recipient's income and the amount of money received has been identified as a test of the altruism model, with the altruism model predicting that as the income of a recipient increases, ceteris puribus, the amount of private transfers received will decrease (Becker, 1981). Cox (1987) and Cox and Rank (1992) analyzed this relationship empirically and find a positive relationship between the amount of assistance received and the (potential) recipient's income. The findings of Cox (1987) and Cox and Rank (1992) are corroborated by those of MacDonald (1990).~,' However, other studies found a negative relationship between the amount (or probability) of transfers received and income. Rosenzweig and Wolpin's (1990) point estimates imply that a $5,000 increase in the adult child's earnings reduces the probability of co-residing by 1 1.1 percent and reduces the probability of receiving a monetary transfer while not residing at home by 10.9 percent. In a recent paper using the 1988 PSID, Altonji et al., (1996) specify a Tobit model and find that the respondent's income has a negative effect on the amount of transfers received from parents. Shelton and Sueyoshi (1993), using information on private transfers collected annually in the PSID, finding that having the lowest household income among the households in the family increases the probability of receiving private transfers from 11 percent to 20 percent. Work by McGarry and Schoeni (1995, 1996) using the Health and Retirement Survey and the Asset and Health Dynamics Survey finds that larger financial transfers are given to adult children with lower income, and this result holds when they look within families by controlling for family fixed effects. Using the Asset and Health Dynamics Survey, Dunn and Phillips (1995) also find that inter vivos transfers are more likely to be given to poorer children within a family, but that children of different income levels are equally likely to receive parental transfers at the time of the death of a parent. Most studies have found that individuals with more years of schooling both give and receive greater amounts of money transfers (McDonald, 1990; Cox and Raines, 1985). However, McGarry and Schoeni (1995, 1996) demonstrate that the effects of schooling are mitigated substantially when family fixed effects are controlled for by examining differences in transfers among siblings. This pattern is consistent with the hypothesis that parents who make larger transfers to their children are also parents who invest more in their children's education. A popular belief about black families is that they have a more active support network than white families. However, with regard to interhousehold assistance, most recent studies have not found support for this belief (MacDonald, 1990; Silverstein and Waite, 1992). For example, MacDonald (1990) shows that, among '~ac~onald (1990) uses a two-step estimation procedure, in which he excludes life-course events (divorce, marriage, births, home-leaving, non-work and non-school spells) fiom the Probit model. Furthermore, he estimates a Tobit regression as an alternative to the two-step procedure and does not find consistently positive effects of recipient's income. '1n Cox's (1987) analyses of the PCPP, only characteristics of the respondent are available. Using the NSFH, Cox and Rank (1992) are able to control for parent's income. Both studies use a twostep estimation procedure
recipients,whites receive $3,500 more than Mexican-Americans,and Mexican- Americans receive $700 more than blacks. Most studies have found that monetary transfers flow primarily from the old to the young (Cox and Raines,1985;MacDonald,1990;Gale and Scholz,1991). Cox and Raines(1985)find that monetary transfers given to younger generations account for 64 percent of the total dollar amount of transfers.And Gale and Scholz (1991)also find that the probability of giving money increases with age, peaking at ages 55-64.The probability of receiving peaks at ages 35-44,with the elderly very unlikely to receive money transfers. Several studies restrict analyses to transfers from parents to children.Mac- Donald(1990)reports that the individual from whom the respondent most com- monly receives transfers is a parent.Similarly,using the Survey of Consumer Finances,Gale and Scholz (1991)estimate that monetary transfers received from parents account for 84 percent of the total amount of transfer dollars received. Most of the theories of private transfers posit that the characteristics of both the potential donor and potential recipient are important,and some studies have examined the effects of parental characteristics since most transfers are between parents and their children.Cox and Rank (1992)find that parent's income posi- tively influences the amount of money received by adult children even when con- trolling for the adult child's income.Similarly,Rosenzweig and Wolpin (1990) find that a rise in parental income by $5,000 increases the probability that the adult child will receive a monetary transfer while living outside the home by 2.2 percent and decreases the probability of co-residence by 2.5 percent.Using the 1988 PSID,Altonji et al.(1996)also find that parent's income positively influences the amount of money received by adult children. These studies have begun to provide some information on private transfers, but there are still many unresolved issues.The various surveys report a wide range of estimates of the share of households receiving financial transfers,from about 8 to 20 percent.The extent to which households are connected through any type of private transfer,whether it be cash assistance or time help,and whether it be through giving or receiving assistance,is not known,yet is important for evaluat- ing the implications of various models (Bernheim and Bagwell,1988).The age pattern of transfers is also important as the age structure of the population contin- ues to change.One of the focal relationships is the effect of income on the amount of transfers received,and the evidence is still mixed.Our analyses of the 1988 PSID will help address some of these issues. III.THE DATA:1988 PANEL STUDY OF INCOME DYNAMICS The data that receive primary attention come from a supplement to the 1988 PSID that investigates private interhousehold transfers.The question regarding private parental monetary transfers asks,"During 1987,did (you or your family living there)receive any loans,gifts,or support worth $100 or more from your Throughout the paper the term"household"will refer to the nuclear family which consists of the PSID respondent and his/her family living there. 428
recipients, whites receive $3,500 more than Mexican-Americans, and MexicanAmericans receive $700 more than blacks. Most studies have found that monetary transfers flow primarily from the old to the young (Cox and Raines, 1985; MacDonald, 1990; Gale and Scholz, 1991). Cox and Raines (1985) find that monetary transfers given to younger generations account for 64 percent of the total dollar amount of transfers. And Gale and Scholz (1991) also find that the probability of giving money increases with age, peaking at ages 55 -64. The probability of receiving peaks at ages 35-44, with the elderly very unlikely to receive money transfers. Several studies restrict analyses to transfers from parents to children. MacDonald (1990) reports that the individual from whom the respondent most commonly receives transfers is a parent. Similarly, using the Survey of Consumer Finances, Gale and Scholz (1 991) estimate that monetary transfers received from parents account for 84 percent of the total amount of transfer dollars received. Most of the theories of private transfers posit that the characteristics of both the potential donor and potential recipient are important, and some studies have examined the effects of parental characteristics since most transfers are between parents and their children. Cox and Rank (1992) find that parent's income positively influences the amount of money received by adult children even when controlling for the adult child's income. Similarly, Rosenzweig and Wolpin (1990) find that a rise in parental income by $5,000 increases the probability that the adult child will receive a monetary transfer while living outside the home by 2.2 percent and decreases the probability of co-residence by 2.5 percent. Using the 1988 PSID, Altonji eta/. (1996) also find that parent's income positively influences the amount of money received by adult children. These studies have begun to provide some information on private transfers, but there are still many unresolved issues. The various surveys report a wide range of estimates of the share of households receiving financial transfers, from about 8 to 20 percent. The extent to which households are connected through any type of private transfer, whether it be cash assistance or time help, and whether it be through giving or receiving assistance, is not known, yet is important for evaluating the implications of various models (Bernheim and Bagwell, 1988). The age pattern of transfers is also important as the age structure of the population continues to change. One of the focal relationships is the effect of income on the amount of transfers received, and the evidence is still mixed. Our analyses of the 1988 PSID will help address some of these issues. The data that receive primary attention come from a supplement to the 1988 PSID that investigates private interhousehold transfer^.^ The question regarding private parental monetary transfers asks, "During 1987, did (you or your family living there) receive any loans, gifts, or support worth $100 or more from your '~hrou~hout the paper the term "household" will refer to the nuclear family which consists of the PSID respondent and his/her family living there
parents?These transfers do not include court ordered assistance such as child support or alimony.Furthermore,the PSID asks about transfers with each of the respondent's parents and parents-in-law.With regard to time help,the question asked is:"About how many hours in 1987 did they [your parents]spend helping (you/your family living there )?"Respondents are also asked to report the amount of transfers given to parents in both time and money.Finally,transfers of money and time with other relatives and with friends are each reported. In addition to the information on private transfers,the households are asked to provide information regarding each of the head's parents and,if there is a spouse,each of the spouse's parents.This information includes the parents'net wealth,education,distance in miles from respondent's residence,and marital status. Combined with the information collected annually,the PSID data on private transfers have several advantages over data available from most other surveys: Demographic and income characteristics of both the donor and the recipi- ent are available for parental transfers.Furthermore,the information regarding the parent is more extensive than in other surveys. Data on transfers of money and time assistance,both given and received, are collected. Data on both where a transfer was made and the magnitude of the transfer are collected. In households where there is a spouse,transfers are recorded to and from the spouse's parents.Furthermore,characteristics of the spouse's parents are ascertained. The PSID has an extensive set of socioeconomic information on the house- hold being interviewed and the individuals within the household. Since it is a panel study,the PSID has information for more than one year,although the reliable data on private transfers are only available in 1988.12 Two sample selections were made for the analyses below.First,households in which the head changed between 1987 and 1988 were eliminated,which con- sisted of 492 cases;this is done to insure that private transfers that were made in 1987 and reported in 1988 are attributed to the correct household head.Second, if the head of the household and the head's parents or parents-in-law live in the same household,the observation is dropped.This reduces the sample size by 420,leaving 6,202 households.The question regarding transfers with non-parents conditions on the transfer being with someone outside the household,i.e.,it asks about interhousehold transfers.The question regarding transfers with parents does not make this condition.Thus,in order to restrict attention to interhousehold transfers,this second selection is made.'3 Additional selections are made for some Underscore included in questionnaire. Note that the PSID interviews the head of the household,and they assume that the head of the household is the male in two-parent households.We follow their convention. Two surveys that also have some of these attractive qualities are the Health and Retirement Survey and the Asset and Health Dynamies Survey,although both are restricted to older populations. Among the 912 houscholds that were dropped due to these two selections,3.2 percent received AFDC,which is very similar to the share receiving AFDC in the retained sample(2.8 percent.see Table 2). 429
parents?"'0 These transfers do not include court ordered assistance such as child support or alimony. Furthermore, the PSID asks about transfers with each of the respondent's parents and parents-in-law. With regard to time help, the question asked is: "About how many hours in 1987 did they [your parents] spend helping (you/your family living there )?" Respondents are also asked to report the amount of transfers given to parents in both time and money. Finally, transfers of money and time with other relatives and with friends are each reported." In addition to the information on private transfers, the households are asked to provide information regarding each of the head's parents and, if there is a spouse, each of the spouse's parents. This information includes the parents' net wealth, education, distance in miles from respondent's residence, and marital status. Combined with the information collected annually, the PSID data on private transfers have several advantages over data available from most other surveys: Demographic and income characteristics of both the donor and the recipient are available for parental transfers. Furthermore, the information regarding the parent is more extensive than in other surveys. Data on transfers of money and time assistance, both given and received, are collected. Data on both where a transfer was made and the magnitude of the transfer are collected. In households where there is a spouse, transfers are recorded to and from the spouse's parents. Furthermore, characteristics of the spouse's parents are ascertained. The PSID has an extensive set of socioeconomic information on the household being interviewed and the individuals within the household. Since it is a panel study, the PSID has information for more than one year, although the reliable data on private transfers are only available in 1988.12 Two sample selections were made for the analyses below. First, households in which the head changed between 1987 and 1988 were eliminated, which consisted of 492 cases; this is done to insure that private transfers that were made in 1987 and reported in 1988 are attributed to the correct household head. Second, if the head of the household and the head's parents or parents-in-law live in the same household, the observation is dropped. This reduces the sample size by 420, leaving 6,202 households. The question regarding transfers with non-parents conditions on the transfer being with someone outside the household, i.e., it asks about interhousehold transfers. The question regarding transfers with parents does not make this condition. Thus, in order to restrict attention to interhousehold transfers, this second selection is made.I3 Additional selections are made for some 10 Underscore included in questionnaire. "~otc that the PSID interviews the head of the household, and they assume that the head of the household is the male in two-parent households. We follow their convention. I2 Two surveys that also have some of these attractive qualities are the Health and Retirement Surve and the Asset and Health Dynamics Survey, although both are restricted to older populations. lGmong the 912 households that were dropped due to these two selections, 3.2 percent received AFDC, which is very similar to the share receiving AFDC in the retained sample (2.8 percent. see Table 2)
of the analyses,and these selections are identified when the results of those analyses are discussed.Robustness to these selections is also examined. IV.FREQUENCY AND MAGNITUDE OF PRIVATE INTERHOUSEHOLD TRANSFERS:NEW EVIDENCE FROM THE PSID Table 2 reports household income in 1987 broken down by source of income as reported in the 1988 PSID.The average household income from all sources is $35,414.Fifty-two percent of total household income is derived from labor income of the head of the household,while 15 percent is attributable to labor income of the spouse.The average amount of private inter vivos transfers received in 1987 is $398 for the entire sample and $2,104 for the 19 percent receiving them.This compares favorably with MacDonald's (1990)annualized estimate of $397 reported in the NSFH and Moon's(1983)estimate of $316 with the PSID. Private transfers are small relative to labor income.However,relative to public transfers (excluding Social Security),private transfers are received by a TABLE 2 DESCRIPTIVE STATISTICS OF HOUSEHOLD INCOME BY SOURCE OF INCOME FOR ALL HOUSEHOLDS (N=6,165). Mean for Entire Coefficicnt of Percent Percent of Sample Variation Receiving Total Income Income Source (1) (2) (3) (4) Totul Household Income: $35.414 0.963 100 100 Market Income: Lavor income-Head 18.346 1.339 69.1 51.80 Labor income-Spouse 5,150 1.853 36.6 14.84 Other income of Head and Spouse 5,140 3.479 58.2 14.51 Income of Others in the Family 2.259 3.073 23.4 6.41 Unit Private Transfers: Inter vivos transfers 398 6.299 18.9 1.12 Inheritance 312 15.667 1.8 0.88 Social Insurance: Social Security 1.874 1.967 26.2 5.29 Unemployment Compensation 94 5.617 5.4 0.27 Worker's Compensation 79 10.456 2.2 0.22 Public Assistance: Aid to Families with Dependent 88 7.091 2.8 0.25 Children Supplemental Security Income 87 7.184 2.8 0.25 Veteran's Pensions 167 7.964 3.8 0.47 Other Assistance: Other welfare 27 11.593 1.1 0.08 Other retircment 1,141 3.535 15.2 3.22 All other transfers 11 8.853 5.2 0.31 Child support 130 6.335 44 0.37 Note:The statistics within the table are calculated using the 1988 PSID family weights.House- holds with non-positive total household income are excluded. Note that some households do not have a spouse,and income from this source is zero for these households. 430
of the analyses, and these selections are identified when the results of those analyses are discussed. Robustness to these selections is also examined. IV. FREQUENCY AND MAGNITUDE OF PRIVATE INTERHOUSEHOLD TRANSFERS: NEW EVIDENCE FROM THE PSID Table 2 reports household income in 1987 broken down by source of income as reported in the 1988 PSID. The average household income from all sources is $35,414. Fifty-two percent of total household income is derived from labor income of the head of the household, while 15 percent is attributable to labor income of the spouse.I4 The average amount of private inter vivos transfers received in 1987 is $398 for the entire sample and $2,104 for the 19 percent receiving them. This compares favorably with MacDonald's (1990) annualized estimate of $397 reported in the NSFH and Moon's (1983) estimate of $316 with the PSID. Private transfers are small relative to labor income. However, relative to public transfers (excluding Social Security), private transfers are received by a TABLE 2 DESCRIPTIVE STATISTICS OF HOUSEHOLD INCOME BY SOURCE OF INCOME FOR ALL HOUSEHOLDS (N= 6,165). - Mean for Entire Cocfficicnt of Percent Percent of Sample Variation Receiving Total Income Income Source (1) (2) (3) (4) Total Household Income : $35.414 0.963 100 100 Market Itzcome : Lavor income--Head Labor incomc-Spouse Other income of Head and Spouse Income of Others in the Family Unit Private Trcznsfirs: Inter viuos transfers Inhcritancc Social Insurance: Social Security Unemployrncnt Compensation Worker's Compensation Public Assistance: Aid to Families with Dependent Children Supplemental Security Income Veteran's Pensions Other Assistance: Other welfare Other retirement All other transfers Child support Note: The statistics within the table are calculated using the 1988 PSID family weights. Households with non-positive total household income are excluded. 14 Note that some households do not have a spouse, and income from this source is zero for these households
large share of the population and are sizable.Moreover,the average amount received from private transfers is greater than the total amount received from SSI,AFDC,Unemployment Insurance(UI),and Workers Compensation (WC). However,the average amounts of SSI,AFDC,and WC conditional on receipt from the respective program are each larger than the average amount of private transfers received conditional on receipt.Moreover,the most common type of non-market income is Social Security;just over one quarter of all households receive Social Security benefits for an average of S7,152 among recipient households. Several studies (Tomes,1981;Menchik,1980,1988;Wilhelm,1996; Kotlikoff,1988;Modigliani,1988)have analyzed bequests to test theories of private transfers and theories of savings.Table 2 shows that inter vivos transfers are 28 percent larger than bequests.However,although bequests are received by less than 2 percent of respondents in a given year,when a bequest is received it is quite large,with an average of over $17,000. Table 3 reports the proportion of respondents with each type of transfer (i.e., money given,money received,time help given,and time help received)and the average amount transferred by the relationship to the head of the household of the person with whom the transfer was made.Monetary transfers are given by 13 percent of all households for an average of $291 for the entire sample.As was shown in Table 2,monetary transfers are reported being received much more frequently;20 percent of the households received a transfer in 1987 and the mean amount received for the entire sample is about $400.'5.16 Assistance in the form of time help is made more frequently;28 percent of the respondents receive time help for an average of 332 hours per year for those receiving help.Time transfers are reported as being given more often than received(33 percent give),and the amount given,conditional on giving,is higher(354 hours are given). TABLE 3 PRIVATE TRANSFERS BY RELATIONSHIP TO THE HEAD OF THE HOUSEHOLD (N=6,202) Monetary Transfers Time Transfers (Hours) Person with whom Percent Mean Percent Mean Percent Mean Percent Mean transfer was made Giving Given Receiving Received Giving Given Receiving Keceived Any individual 13.3 S291 20.2 S405 33.2 1I7 28.4 93 Parent/parent-in-law 3.1 56 17.6 328 24.0 82 203 66 Child 53 175 0.9 3.8 10 5 6 Sibling 1.7 4 1.7 3 4.0 6 3.8 6 Other relative 1.7 30 1.6 25 2.8 》 1.7 3 Non-relative 20 11 15 6 7.5 12 7.1 10 Note:Means are for the entire sample.The 1988 PSID family weights are used in the calculations above. The calculations reported in Table 3 are slightly different than those in Table 2 because in Table 2 we restrict the analysis to those households with positive household income.This reduces the sample by 37 cases,and the results reported in Table 2 are not sensitive to this selection. In the aggregate for a representative cross-section,the mean amount of money given to other households should equal the mean amount received from other households.Differences in the reported amounts may exist because transfers received include loans and gifts,while transfers given do not explicitly include loans,and loans and gifts cannot be separated in the data.In addition,a separate question regarding monetary transfers received is asked explicitly about each parent and all other relatives,perhaps eliciting a greater amount of assistance than from the one catch-all question about transfers given to others. 431
large share of the population and are sizable. Moreover, the average amount - received from private transfers is greater than the total amount received from SSI, AFDC, Unemployment Insurance (UI), and Workers Compensation (WC). However, the average amounts of SSI, AFDC, and WC conditional on receipt from the respective program are each larger than the average amount of private transfers received conditional on receipt. Moreover, the most common type of non-market income is Social Security; just over one quarter of all households receive Social Security benefits for an average of $7,152 among recipient households. Several studies (Tomes, 198 1 ; Menchik, 1980, 1988 ; Wilhelm, 1996; Kotlikoff, 1988; Modigliani, 1988) have analyzed bequests to test theories of private transfers and theories of savings. Table 2 shows that inter vivos transfers are 28 percent larger than bequests. However, although bequests are received by less than 2 percent of respondents in a given year, when a bequest is received it is quite large, with an average of over $17,000. Table 3 reports the proportion of respondents with each type of transfer (i.e., money given, money received, time help given, and time help received) and the average amount transferred by the relationship to the head of the household of the person with whom the transfer was made. Monetary transfers are given by 13 percent of all households for an average of $291 for the entire sample. As was shown in Table 2, monetary transfers are reported being received much more frequently; 20 percent of the households received a transfer in 1987 and the mean amount received for the entire sample is about $400.~~,'~ Assistance in the form of time help is made more frequently; 28 percent of the respondents receive time help for an average of 332 hours per year for those receiving help. Time transfers are reported as being given more often than received (33 percent give), and the amount given, conditional on giving, is higher (354 hours are given). TABLE 3 Monetary Transfers Time Transfers (Hours) Person with whom Percent Mean Percent Mean Percent Mean Percent Mean transfer was made Giving Given Receiving Received Giving Given Receiving Received Anv individual 13.3 $291 20.2 $405 33.2 117 28.4 93 Parent/parent-in-law 3.1 56 17.6 328 24.0 82 20.3 66 Child 5.3 175 0.9 5 3.8 10 3.5 6 Sibling 1.7 14 1.7 2 1 4.0 6 3.8 6 Other relative 1.7 30 1.6 25 2.8 7 1.7 3 Note: Means are for the entire sample. The 1988 PSID family weights are used in the calculations above. I5 The calculations reported in Table 3 are slightly different than those in Table 2 because in Table 2 we restrict the analysis to those households with positive household income. This reduces the sample by 37 cases, and the results reported in Table 2 are not sensitive to this selection. 16 In the aggregate for a representative cross-section, the mean amount of money given to other households should equal the mean amount received from other households. Differences in the reported amounts may exist because transfers received include loans and gifts, while transfers given do not explicitly include loans, and loans and gifts cannot be separated in the data. In addition, a separate question regarding monetary transfers received is asked explicitly about each parent and all other relatives, perhaps eliciting a greater amount of assistance than from the one catch-all question about transfers given to others
Parents are the most common source of private transfers;over three-fourths of transfer dollars received are received from parents,and almost three-fourths of time help received is received from parents(Table 3).Transfers of either form are received relatively infrequently from siblings,other relatives,and non-relatives. The amount of private transfers varies widely,with a mean of $2,095 and standard deviation of S5,438.The most common transfer reported is $500.Time help also has a wide distribution.The average number of hours received for recipients is 298 hours and the standard deviation is 573. Bernheim and Bagwell (1988)show that if all households are altruistically linked,either directly or indirectly through intermediary households,redistribu- tion may be completely neutralized.Table 4 reports additional evidence on the extent to which households are linked through transfers of money or time help. While 20 percent of the households receive monetary transfers,13.3 percent give money transfers,28.4 percent receive time transfers,and 33.2 percent give time transfers.Over 30 percent participated in some form of money transfer in the single year 1987,with only 2.2 percent simultaneously giving and receiving money transfers.Time transfers were more frequent,with 46 percent either giving or receiving time help.Time help is also more frequently given and received by the same household in a given year;16 percent both give and receive time help.In general,households appear to be fairly well linked,with almost 60 percent of the households either receiving or giving money or time assistance during the single year 1987. Some households give transfers to,or receive transfers from,more than one household within the same time period.This may extend the chains of private TABLE 4 PERVASIVENESS OF PRIVATE INTERHOUSEHOLD TRANSFERS (N=6,202) Percent with Percent with Type of Transfers these Transfers Type of Transfers these Transfers Reccive money 19.5 Received or gave money 30.6 Gave money 13.3 Received money or time 38.7 Received time 28.4 Received money or gave time 44.1 Gave time 33.2 Gave money or received time 38.2 Received and gave money 2.2 Gave money or time 40.5 Received money and time 9.2 Received or gave time 45.6 Received money,gave time 85 Received or gave money,or 47.4 Gave money,received time 3.5 received time Gave money and time 6.0 Received or gave money,or 50.5 Received and gave time 16.0 gave time Received and gave money,and 1.1 Gave money.or received or 51.7 received lime gave time Received and gave money,and 1.3 Participated in any form of 58.7 gave time transfer Gave money,reccived and 2.3 gave time Participated in all forms of 0.9 transters Note:The 1988 PSID family wcights are used in the calculations above 432
Parents are the most conimon source of private transfers; over three-fourths of transfer dollars received are received from parents, and almost three-fourths of time help received is received from parents (Table 3). Transfers of either form are received relatively infrequently from siblings, other relatives, and non-relatives. The amount of private transfers varies widely, with a mean of $2,095 and standard deviation of $5,438. The most common transfer reported is $500. Time help also has a wide distribution. The average number of hours received for recipients is 298 hours and the standard deviation is 573. Bernheim and Bagwell (1988) show that if all households are altruistically linked, either directly or indirectly through intermediary households, redistribution may be completely neutralized. Table 4 reports additional evidence on the extent to which households are linked through transfers of money or time help. While 20 percent of the households receive monetary transfers, 13.3 percent give money transfers, 28.4 percent receive time transfers, and 33.2 percent give time transfers. Over 30 percent participated in some form of money transfer in the single year 1987, with only 2.2 percent simultaneously giving and receiving money transfers. Time transfers were more frequent, with 46 percent either giving or receiving time help. Time help is also more frequently given and received by the same household in a given year; 16 percent both give and receive time help. In general, households appear to be fairly well linked, with almost 60 percent of the households either receiving or giving money or time assistance during the single year 1987. Some households give transfers to, or receive transfers from, more than one household within the same time period. This may extend the chains of private TABLE 4 Percent with Percent with Type of Transfers these Transfers Type of Transfers these Transfers Receive money 19.5 Received or gave money 30.6 Gave money 13.3 Received money or time 38.7 Received time 28.4 Received money or gave time 44.1 Gave time 33.2 Gave money or received time 38.2 Received and gave money 2.2 Gave money or time 40.5 Received money and time 9.2 Received or gave time 45.6 Received money, gave time 8.5 Received or gave money, or 47.4 Gave money, received time 3.5 received time Gave money and t~me 6.0 Received or gave money, or 50.5 Received and gave time 16.0 gave time Received and gave money, and 1.1 Gave money, or received or 51.7 received time gave time Received and gave money, and 1.3 Participated in any form of 58.7 gave time transfer Gave money, received and 2.3 gave time Participated in all forms of 0.9 transfers Note: The 1988 PSlD family weights are used in the calculations above 432