Intervivos Transfers and Bequests in three OECD Countries 1.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 1 Intervivos Transfers and Bequests in three OECD Countries 1. Ernesto Villanueva 2 Department of Economics, Universitat Pompeu Fabra, Barcelona 1. INTRODUCTION Private monetary transfers from parents to adult children living in a separate household constitute an important share of aggregate net worth in the economies of major economies like the United States or Germany. Evidence from the United States from the middle eighties suggests that parental intervivos transfers - gifts given from living parents to adult children- are the source of at least 11 percent of aggregate net worth. On the other hand, accumulated bequests - monetary transfers received after the death of parents - amount to 30% of aggregate net worth in the economy of the United States. Comparable estimates for West Germany for the late eighties, suggest that accumulated bequests constitute a 10% share of net worth. Potentially, these flows of resources may affect the distribution of resources in these economies. From a public policy perspective, private monetary gifts between generations can affect the effectiveness of public programs. Firstly, population aging in Europe and the 1 Second draft for Economic Policy, submitted for consideration for the 40 th October Panel in Amsterdam. 2 Department of Economics, Universitat Pompeu Fabra. Ramon Trias Fargas 25-27 08005 Barcelona, SPAIN. This paper has benefitted from the comments of Michael Hurd, Tullio Jappelli, AnnaMaria Lusardi, Michael Reiter, and Joachim Voth and from participants at the fourth meeting of the Research Training Network Ageing and Retirement in Europe, and the Applied Economics Lunch at Universitat Pompeu Fabra. Special thanks to Giuseppe Bertola for very detailed comments. Funding from the Fundación Ramón Areces is greatfully acknowledged.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 2 United States has sparked a debate about the need of a reform of Social Security systems and the possibility of reducing the replacement rates guaranteed by post-retirement income. The distributional effect of these policy changes depends on the motivation of intergenerational transfers. If individuals care about the felicity of their descendents, and provide them with monetary transfers, a reduction of old age income may affect not only the elderly, but also their descendents. Conversely, if individuals derive utility neither from leaving a bequest nor from providing transfers, a reduction of old age income will affect mainly the elderly. Aggregate statistics about accumulated intergenerational transfers related to net worth are suggestive of their relevance, but by themselves give little information about their interaction with public policy aimed at redistributing income. This paper assesses whether or not intergenerational transfers affect public policy by examining the impact of lifetime income on intergenerational transfers (bequests and intervivos transfers) in three OECD countries, using the methodology defined in Altonji and Villanueva (2003). We focus on the relationship between intergenerational transfers and the lifetime income of individuals for two reasons. The first is that economic theory predicts that intergenerational transfers are lifetime decisions, determined by the entire income histories of the donor. The second is that intergenerational transfers can undo the effects of public policy that redistributes income across generations only if they respond to income changes. If intergenerational transfers do not respond to the income of the donor, an increase in old age income is unlikely to be passed on to the next generation. In a first step, we assess the impact of lifetime income on bequest giving in the United States, West Germany and the United Kingdom. We document that this estimate is similar in West Germany and the United Kingdom, about 1 cent, and about 2 cents in the United States. Further, we use these estimates to assess the presence of a bequest motive. The assessment is done by comparing the fraction of an extra dollar of lifetime income that is devoted to wealth late in life by elderly with and without children. Our approach is based on the assumption that bequests of the elderly without children are involuntary, and mainly driven by uncertainty about longevity. The results suggest that a bequest motive is not likely to be operative for most households in the income distribution. Secondly, we compare the impact of lifetime income on expected intervivos transfers for West Germany and the United States. The results suggest that intervivos transfers may displace public help. We find that the impact of lifetime income on intervivos transfers is higher in the United States than in West Germany. We pose that that difference is due to the generosity of the Unemployment Insurance system, and crosscheck the result by estimating the displacement effect of unemployment benefits on transfers intervivos using a sample of unemployed young. The paper is organized as follows. Section 2 provides aggregate evidence on the relevance of bequests and intervivos transfers in the United States, West Germany and the United Kingdom. Section 3 describes the alternative economic models of bequests and transfer giving, as well as their implications for policy. Section 4 presents the

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 3 empirical methodology. Section 5 uses regression analysis to identify the responses of transfers and bequests to lifetime income in Germany, the United States and the United Kingdom. Section 6 discusses whether differences in wealth taxation and in Social Security replacement rates explain the discrepancy in the responses of transfers and bequests in Germany and the United States, Section 7 discusses the results, and Section 8 lays out the implications for public policy. 2. AGGREGATE COMPARISONS BETWEEN GERMANY, THE UNITED STATES AND THE UK How large are intergenerational transfers like intervivos monetary transfers to adult children and bequests, compared with GDP or with aggregate net worth in a society? This Section reviews several aggregate statistics for the United States, West Germany and the UK. The sources of information about these intergenerational transfers are usually household surveys in which individuals are asked about the amount of money given to adult children living in separate households. Bequests are obtained from responses to direct questions about inheritances received. There are two sources of possible underreporting of these flows in typical household surveys. The first is that survey respondents may find it more difficult to remember that they have received monetary transfers than to remember that they gave. Second, gifts are more likely to be prevalent among high-income households, typically hard to survey. With these caveats in mind, Table 1 presents a comparison of magnitudes of bequests and intervivos transfers to adult children in the United States, Germany and the UK. The main source of information in the US case is Gale and Scholz (1994), using data from the 1983 wave of the Survey of Consumer Finances. The corresponding for West Germany is Reil-Held (1999), who reports data from the Socio Economic Panel (GSOEP). Less information is presented about UK due to the lack of data. The aggregate bequest giving magnitude was taken form the website of the International Longevity Centre. The annual flow of bequests in the United States is 104 billion dollars (431.24 per head, dollars of 1985), almost 2.6% of the GDP in that year. The annual flow of intervivos transfers in the US is about a third of the flow of bequests. In West Germany 16.5 billion dollars were inherited in 1985 (264.46 1985-dollars per head). The corresponding flow of intervivos transfers in 1990 also accounted for about a third of the magnitude of annual bequests. The annual flow of bequests in the UK was 22 Billion Sterling in 1996, (corresponding to 432.7 1985-dollars per head). A possible benchmark to assess the relevance of intergenerational transfers received by individuals is to compare a measure of accumulated net transfers (including bequests) over the life cycle to aggregate net worth in an economy. A large ratio of intergenerational transfers to net worth is hard to reconcile with the notion that the main motivation behind household saving is to consume later in life or in bad circumstances.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 4 It suggests that individuals save to pass on resources to their children. In such case, the effectiveness of public policies that redistribute income across generations, like Social Security can be hindered. The reason is that, if individuals are concerned about the well being of their descendents, public transfers that redistribute income from the young to the old can be partially reversed by private transfers in the opposite direction. Analysing the effect of intervivos transfers and bequests involves making assumptions about the interest rate, population growth, and on the ages in which individuals give and receive transfers. Gale and Scholz (1994) estimate that in the US, the ratio of accumulated bequests to net worth is 30%. The corresponding ratio for intervivos transfers (including life insurance and the value of trusts) is 20%. Merely focusing on direct intervivos transfers to adult children, the corresponding estimate is 10%. At a minimum, it looks that intergenerational transfers are a substantial fraction of net worth in an economy. Reil-Held (1999) uses a similar method for West Germany, correcting for the fact that German potential donors experienced an abnormal poverty period after the Second World War, and documents that accumulated bequests account for 10% of the total net worth, below the corresponding estimate for the United States, but still a significant fraction of household wealth. Aside from historical factors and differences in preferences, several institutional arrangements may underlie these differences, some of which are presented in Table 2. The degree of financial development in the country, public programs that substitute private income in the old age or in low-income transitory periods may affect the need to save for the old age or the prevalence of private sources of help. Here we focus on two main programs: Social Security and Unemployment Insurance. 3 We start with public income provision in the old age. Upon retirement, the German Public Social Security System replaces 66.8% of net income prior to retirement with net benefits after retirement. The corresponding number for the United States was 39.6%, and 27.2% for the UK. Holding everything else constant, the incentive to accumulate wealth for the old age in the United States is thus larger than in Germany. Second, Germany has an unemployment insurance system that provides a much higher replacement rate upon the loss of a job than that in the United States or the United Kingdom. For a married person, with two children, the replacement rate in West Germany in 1998 was 73%, while in the United States is 61%, and 64% in the UK. In addition, the maximum duration of unemployment compensation is 26 weeks in the United States, and 12 months in the United Kingdom, while it can reach 32 months in West Germany. The potential role of private networks of insurance (like transfers intervivos) is possibly larger in economies in which the public sector provides less help when individuals lose a job than in economies with a stronger Welfare State. Unfortunately, the aggregate ratios about transfers and bequests as a proportion of net worth do not permit us to establish to what extent a more generous Social Security or 3 The following discussion draws from Börsch-Supan and Lusardi (2001).

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 5 Unemployment Insurance System interact with the patterns of giving through intervivos transfers and bequests. Following Altonji and Villanueva (2003) or Kotlikoff (1988), we propose to examine the impact of an extra dollar of the lifetime income of an individual on expected lifetime intervivos transfers and bequests that the individual passes on to the next generation. Bequests are lifetime decisions and depend among other variables on the whole story of income realizations of an individual. Hence, the lifetime income of the donor is a natural determinant of the decision to give. In addition, the response of bequests and transfers to lifetime income is useful to examine the role of intergenerational transfers in the distribution of resources from two points of view. First, as we discuss in Section 3, bequests may undo the distributional effects of public transfers if a significant fraction of households pass on a positive fraction of their income to the next generation and if donors intend to give them. Knowing how much of an extra dollar of lifetime income is passed to the next generation through bequests allows us to compare the magnitude of the response of intergenerational transfers to lifetime income across groups of the population with different incentives to give to the next generation, and to assess whether or not bequests are intended. Second, intergenerational transfers have potentially an impact in the process of transmission of economic inequality from one generation to the next one. It is well documented that inequality in income is transmitted across generations, in the sense that parents with labour incomes over the mean have children with labour income over the mean of the society. The extent to which higher-income parents give also higher intergenerational transfers to their adult children determines whether or not these transfers add to the process of transmission of inequality in lifetime resources. 3. ECONOMIC MODELS OF BEQUEST GIVING AND INTERVIVOS TRANSFERS 3.1. Alternative models on the determinants of bequest giving There are two main explanations for the existence of bequests. The first, usually termed as the pure life-cycle model, states that bequests are accidental. The second assumes that bequests are intended. The distinction matters for policy purposes. According to the pure life-cycle model, individuals are uncertain about their date of death, and their saving decisions are not affected by the welfare of the next generation. Individuals plan their savings taking into account that resources must be provided for their old age. In the absence of annuity markets, or of markets that perfectly insure individuals from uncertain medical expenses, savings act as a buffer stock against the risks of living too long or facing large medical expenses in the old age. Egoistic

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 6 individuals may leave unintended bequests if they die earlier than expected, without having exhausted their wealth. According to the intended bequest model, aside from their own consumption, individuals are concerned about either the felicity of the next generation (altruistic motive) or about the mere fact of giving (warm glow motive). The motivation for bequests has consequences both on the allocation of resources and on the distributional impact of public programs that transfer income to the elderly. Implications for the allocation of resources. Whether or not bequests are intended determines the amount of capital saved in an economy. Consider the case of unintended bequests first. Individuals save during their working life to provide resources for their old age. The introduction of Pay-As-You-Go Social Security system will distort the savings decisions of households, so that in the aggregate, the capital stock will be lower than before the introduction. The reason is that a Pay-As-You Go Social Security System redistributes income from the young, who have a relatively high marginal propensity to save, to the elderly, who have a lower marginal propensity to save (see Fuster, Imrohoroglu and Imrohoroglu (2004)). Computations for the United States done by Kotlikoff and Auerbach (1984) suggest that this effect reduces the capital stock in an economy by about 20%. Considering the case of altruistic motives for bequests, macroeconomic simulations calibrated to the United States, assuming that intervivos transfers and bequests are motivated by altruism, find that Pay-As-You-Go Social Security systems displace only about 6%-8% of the total stock of capital. In the presence of an altruistic motive, individuals save for bequest motives, in addition to other life-cycle motives. Therefore, elderly individuals do not necessarily have a lower marginal propensity to save than young individuals. Implications for the distribution of resources. The presence of a bequest motive also determines which groups benefit from public programs that redistribute income across generations. Assume that a public program redistributes income towards the elderly. If bequests are accidental, an unexpected increase in the income of the elderly will result in an increase in their consumption. Conversely, if bequests are intended, a fraction of that increase will be passed on to the next generation, effectively increasing the welfare of the adult children of the recipients. An extreme version of this result is the Ricardian Equivalence hypothesis. If all individuals in a society derive utility from the welfare of their descendents, and all individuals provide support to their children, a public program that redistributes income from the young to the old will not affect the consumption of any of the cohorts involved. The reason is that a dollar taken from the young and given to the elderly through a public program will be given back by the elderly through bequests. Possibly, bequest motives are mixed in the population. That is, some individuals may derive utility from leaving a bequest to their children, and other individuals may not (see Light and McGarry (2004)). On the other hand, even if all individuals have a bequest

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 7 motive, lower-income individuals may derive a higher marginal utility from devoting an additional dollar of resources to their own consumption than to the consumption of others, and bequest motives would not be operative. Still, it is important to know whether or not if, in the aggregate, bequests motives are operative. Can we use our estimates to assess the presence of a bequest motive? This paper identifies the presence of bequest motives by comparing the response of expected bequests to lifetime income between individuals with and without children. The identifying assumption is that the savings of the elderly without children are mainly held because of cautionary motives. If individuals with children have a similar marginal propensity to bequeath out of lifetime income than otherwise identical elderly without children then cautionary motives are likely to be main determinant of wealth accumulation for both groups. Only if we find a positive difference in marginal propensities to bequeath between the elderly with and without children we infer that a bequest motive is operating. Bequest motives are less likely to operate for individuals in the bottom of the income distribution than for individuals at the top. Holding everything else constant, even if everyone in the economy cares about the welfare of their descendents, those individuals with lower lifetime income have a higher marginal utility from their own consumption than from the consumption of their descendents, and will prefer consuming an extra dollar of lifetime income than bequeathing it to the next generation. 4 To account for this kind of behaviour, we compute the marginal propensity to bequeath at several points of the income distribution. The comparison between groups of the elderly with and without children is not always indicative of the presence of an operative bequest motive. Box 1 details the conditions under which that comparison detects a bequest motive. 3.2. Models of intervivos transfers Unlike bequests, intervivos transfers are clearly intended by donors. Economic theory stresses two main motivations for intervivos transfers. The first is the altruistic model, described in the previous subsection. The second are exchange models, in which donors transfer resources to their children in exchange of the provision of some (costly) action from the recipient, also called services. Under the altruistic model, donors provide transfers to equalize the marginal utility of own consumption to that of the recipient of help. In that case, private help may be displaced by public support. The reason is that an extra dollar of public help improves 4 On top of this, public programs can disincentive the accumulation of wealth among the poor. Hubbard, Skinner and Zeldes (1995) provide evidence that several public programs in the United States effectively place high disincentives to save for the poor.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 8 the economic condition of the recipient, diminishing the incentive of the donor to provide resources. Conversely, if private transfers are meant to compensate services received, those may in some cases reinforce the redistribution made by the government. The reason is that an increase in public help improves the economic condition of the recipient. If the action that parents require in exchange of transfers is costly, the child may require higher transfer from the parent to be convinced to continue providing services (see Cox, 1987 and Cox and Jakubson, 1995). Therefore, private transfers will reinforce the redistribution of resources made by the government. To assess the extent to which intervivos transfers interact with public programs, we will take two alternative routes. The first is to compare the marginal propensity to give transfers out of lifetime income between West Germany and the United States. The second is to examine whether a particular public program in Germany (unemployment insurance) increases or decreases intervivos transfers, holding constant the lifetime income of both the parent and the child. To advance the findings: In the three countries considered, for most households in the income distribution, the empirical results are consistent with a model of unintended bequests than with a model of intended bequests. Second, to a limited extent, we find that programs like unemployment insurance displace private transfers. Our estimates from West Germany suggest that an extra dollar of unemployment benefits displace private help by an amount that is about 10 cents. 4. DATA SOURCES AND DEFINITIONS Our analysis is based on three samples drawn from household level data for West Germany (the German Socio Economic Panel), the United States (Panel Study of Income Dynamics) and the United Kingdom (British Household Panel). The construction of the samples is detailed in Box 2. To measure the impact of lifetime income on bequests and on intervivos transfers from parents to children, we need information on intervivos transfers, on bequests received by a person and on the income of their parents. Intervivos transfers and measures of income are available in both surveys, but measures of bequests are harder to obtain. Hence, instead on using data on bequests, we analyse wealth held late in life, under the assumption that the bequest left to children is close to the wealth held by an elderly individual. In other words, we infer the relationship between bequests and income from the relationship between wealth late in life and income weighted by average mortality rates. This measure is called potential bequests, and can overstate bequests for various reasons. The first is that there may be substantial medical costs that drain a substantial amount of resources in the years prior to death. At least for the case of the United States, this may not be a major problem, based on recent evidence of Hurd and Smith (2002), who use panel data to compare wealth holdings late in life to actual bequests (as reported by recipients), and document that out-

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 9 of-pocket medical costs account for less than 8% of household wealth prior to death. A second reason for overstating bequests to children is that actual wealth may be bequeathed to charities or other relatives. In the United States, Hurd and Smith (2002) document that, among decedents that leave a positive estate, at most 8% of the estate is given to persons or institutions who are not children or the spouse. 5 We are not aware of evidence from micro data for Germany or the United Kingdom. Aggregate statistics suggest that the amount given to charities in the United Kingdom is a lower fraction of the GDP than in the United States (0.55% in the United Kingdom and 2 % in the United States). 6 Wealth holding are obtained by aggregating the assets reported by a household that can be bequeathed to children. These include the value of the house, other real state, financial and tangible assets. There are categories that should also be included, like the value of business held by the elderly. Nevertheless, the British Household Panel Survey does not contain information on this category. Hence, we also exclude business wealth from the wealth measure in the United States and in West Germany (perhaps surprisingly, the results for these two countries are not sensitive to that omission). An additional component worth mentioning is life insurance. West German households invest more than Americans in this component, that is possibly related to a bequest motive (Walliser and Winter, 1998). The German Socio-Economic Panel asks individuals about the cash surrender value of the life insurance policy, together with other private pensions and building savings accounts. Kotlikoff (1988) argues that net worth plus the face value of life insurance is the proper measure of what the estate of a person would be upon the event of death. We only have information on cash surrender value, together with other components, that may underestimate wealth holdings in West Germany. On the other hand, the report of life insurance is very similar in the Panel Study of Income Dynamics. In the case of the United States, we omit claims to future income that cannot be bequeathed, like annuities. To analyse bequests we restrict our samples to (a) years in which wealth is asked in the surveys, (b) heads of households to whom we observe after age 50 and (c) who have descendents. As mentioned above, we also examine a control group of elderly without descendents. In the case of the United States, we have four years of data, 1984, 1989, 1994 and 1999. In the West German and British cases, we have one year of information on wealth: 2000 for the United Kingdom and 2002 for West Germany. Intervivos transfers are obtained from questions asking respondents how much money did they give to children who lived outside of the household of the respondents. Given that, under some circumstances parents may want to give more money to children who are in a worse financial situation, we also collect demographic and income data on the 5 Hurd and Smith (2002) document that the fraction of the estate that is given to charities among elderly without children is 17%. The rest goes to siblings or other relatives. 6 The Independent Sector (2001): www.pbs.org

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 10 persons who are the potential recipients of these gifts. In the United Kingdom, the information on transfers is available from the reports of the child. Nevertheless, while about 5% individuals reported having received a transfer, very few respondents gave actual amounts. Hence, we do not use the information on the United Kingdom. 4.1. Comparing measures of wealth and patterns of giving across countries Table 3 documents the average wealth holding in the sample used for bequests in the three countries under consideration. All monetary magnitudes are expressed in 1995 dollars. 7 The average wealth holdings are similar across countries, about 180,000 dollars. Nevertheless, a direct comparison of those measures is misleading, given that the data from the United States contains several years, and that household wealth has been subject to considerable time shocks between 1984 and 1999. The average household wealth in 1999 is 229,586 dollars (of 1995), much larger than the corresponding averages for the United Kingdom and West Germany (around 170,000 dollars in both countries). To account of a measure of income that reflects the lifetime economic status of an individual, we construct a measure of the household s permanent annual income. It is an average across several years of the yearly labour earnings of the husband and wife, normalized so that its average represents the average earnings of a 40-year-old person. As the working lives of many of the individuals we study cover post-war years, in which some cohorts with strong growth of average earnings (specially important in the German case) our measure of permanent income includes an adjustment for these years, not covered by the panels we use. The exact details are contained in Box 4. The measure of permanent yearly labour income is highest in the United States, 42,927 dollars in the US, and lowest in West Germany: 33,236 dollars. We start by checking whether the data we are using are consistent with the widely documented differences in savings rates across the three countries. In 2001, the aggregate savings ratio in the United States was 2.5%, and the corresponding statistics for Germany and the United Kingdom were 10% and 6.2%, respectively (estimates from Börsch-Supan and Lusardi, 2001). The first row of Table 4 presents wealth-income ratios over the distribution of permanent income in United States, West Germany and United Kingdom. The median wealth-income ratio is 3.76 in West Germany, higher than the corresponding estimate of 3.00 for the United States. The wealth ratio in the United Kingdom is 3.56. Our samples are thus consistent with smaller wealth-income ratios in the United States than in West Germany and the United Kingdom, as documented by many other researchers. 7 In what follows we convert DM into euros using the a exchange rate of 2, and then into US dollars using the purchasing power parity of 1 documented by the historical series of the OECD between 1992 and 1999. We convert british pounds into dollars by using the OECD PPP exchange rate of 1.53 dollars per pound These assumptions affect the levels of the variables shown, but not the derivative of transfers and bequests with respect to income.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 11 Nevertheless, the pattern of wealth-income ratios varies over the income distribution. In rows 2 through 6 of Table 4, we condition median wealth-income ratios by income quintile. The first column of Table 4 documents that wealth-income ratios in the United States uniformly increase with income. At the bottom of the income distribution, the wealth-income ratio is slightly below 2. At the top of the income distribution, the wealthincome ratio is 4.5 (see Dynan, Skinner and Zeldes, 2004 for a similar result). In the German case wealth-income ratios are fairly flat, but for the bottom quintile. In United Kingdom, we find declining wealth-income ratios for the bottom two quintiles, but these estimates are somewhat noisy. The increasing pattern of wealth-income ratios over the income distribution in the United States, against the relatively flat for the West Germany and the United Kingdom points at a larger increase of wealth with income in the United States than in the other two economies. That pattern of results is consistent with our findings below that the marginal propensity to bequeath is higher in the United States than in the European countries we examine. Table 5 shows the average transfers given by American and German donors, their own incomes and the income of their descendents. 31% of American households report having given a transfer to any of their children, while the estimate for Germany is lower, 17%. In both countries, parents who give tend to have higher levels of permanent income than parents who do not give. These summary statistics point to a positive relationship between intervivos transfers and income of the potential donor in both countries, and to a positive impact of lifetime income of the donor on the transfer amount. 4.2. The calculation of the magnitude of interest: the marginal propensity to give We assume that both intervolves transfers and bequests are lifetime decisions, whose main determinants are the lifetime resources of the potential donor and receiver (parent and children, respectively). Of course, a positive income shock when the parents are old and have lower needs may affect a bequest more than a positive income shock at age 40, when parents are still paying their mortgages. We abstract from those considerations, and focus on the impact on gifts of lifetime resources, discounted to age 70. We assume that lifetime income of a household has two main components: the stream of labour earnings during the working life of the members and post-retirement income. We assume that the earnings stream during the working life of the household is unrelated to their preferences for leaving a bequest or a transfer, ignoring the possibility that parents who want to leave a bequest exert greater effort during their lifetime to earn more income. After retirement, parents receive a flow of post-retirement earnings that lasts until their

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 12 death at age 86. 8 The lifetime resources of a person, valued at dollars at age 70 depend on the interest rate r and can be expressed as follows age= 62 age= 86 = 70 age (1 + r) yage + 70 age r P( alive = age)(1 + r) yage age= 25 Y (1) age= 62 That is, the amount of lifetime resources at age 70 is the discounted value of the lifetime flow of earnings of the household y age (assuming that the individual works r from age 25 until age 62) and the expected flow of post-retirement income y age. The expected flow of post-retirement income received depends on the age-specific probability of survival, P ( alive = age). We assume a common interest rate of 4% for all countries. 4.5% is the real rate of return used for the United States by Gale and Scholz (1994), following historical averages from Kotlikoff and Summers (1981). For Germany, Reil-Held (1999) estimates an interest rate of long-term saving net of inflation of 2.39%. To allow for comparability of the estimates, we use first a 4% for all countries, and also experiment with a 2% real interest rate below. 4.2.1. The marginal propensity to give through bequests The approach we follow examines potential, rather than actual bequests. That is, we infer the potential bequest that an individual may leave from the stock of wealth late in life, weighted by the probability that the individual dies (as we assume in the computations a married household, we assume that the bequest occurs after the death of the last person of the couple). We start by analysing the relationship between lifetime income of an individual, Y, and the wealth holding of that individual late in life, W. 2 W = γ o + γ 1 Y + γ 2Y + γ 3Y ( age 70) + controls + ε (2) In the former expression, γ 1 measures how much of an extra dollar of lifetime income is held as wealth at the age of 70, for the average income household and holding everything else constant. 9 Note that lifetime income does not vary over time. Hence, in the data, γ 1 is identified by comparing how average wealth holding changes across individuals with different levels of lifetime income holding age and other demographic variables constant. As discussed in Section 3, in the absence of annuity and complete health insurance markets, the impact of lifetime resources on wealth should be positive, regardless of whether or not the individual cares about the welfare of descendents. γ 2 8 Age 86 is possibly an early date of death in most Western countries. Nevertheless, our samples do not contain enough elderly individuals to assess properly the relationship between income and age at ages after 90. 9 In all expressions from now on, the average income will be subtracted from Y, that allows to interpret γ 1 as the impact of lifetime resources on wealth at age 70 evaluated at sample mean of estimated lifetime resources.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 13 could be positive for various reasons. Replacement rates implied by the United States Social Security System are negatively related to lifetime income. Also, to maintain the same standard of living after and prior to retirement, a person with higher income must accumulate higher savings for their old age. In the presence of a bequest motive an additional dollar of lifetime resources is more likely to result in additional wealth holding in later ages in life for a person with high lifetime income than for a person with low lifetime income. For the United States, Smith (1995) documents that wealth holdings vary with demographics that are also related with income, such as marital status or the race of the head of the household. Hence, we hold constant the marital status of the head of the household, and allow for different responses of wealth and transfers to lifetime income for whites and nonwhites, focusing on whites for the rest of the paper. In a second step, we use the estimates from (2) to assess how much do expected bequests vary with the lifetime resources of an individual. The strategy we follow is to weight the impact of lifetime income on wealth holding at a given age by the probability that death happens at that age, and then aggregate over all ages. In other words, given that the event of death is much more likely at age 80 than at age 60, in the computation of the impact of lifetime resources on expected bequests, the impact of lifetime income on wealth at age 80, when the conditional probability of mortality of a female is about 13% should contribute more to the computation of expected bequests than the corresponding estimate at age 60, when the corresponding probability is 1%. Nevertheless, in both ages there is a chance of observing a bequest, so both should be included in the computation. In formal terms, the following weighted mean is computed age E B Y age age E W Y age = + r P age death = age Y = 86 (, ) 70 (, ) (1 ) ( _ ) Y age= 60 In this expression, r denotes the interest rate, P ( age _ death = age) is the probability that the second member of a couple dies at a given age, and the last term is the effect of lifetime income on wealth at a given age. The result of this computation tells by how much do expected bequests increase with an additional dollar of lifetime income, for a couple that faces ex-ante the average survival rates. The details are provided in Boxes 2 through 4. A benchmark group: elderly without children: As mentioned above, a variety of economic models, with different implications for public policy are consistent with a positive impact of lifetime resources on expected bequests. To provide a benchmark for the presence of a bequest motive, we make the same computations for childless elderly in the three countries. Under the assumption that individuals are more concerned about the utility of their children than about the utility of other relatives or charities, childless couples are less likely to have a bequest motive than individuals with children and, holding income constant, are equally likely to save for cautionary purposes (Hurd, 1989). In addition, other legal constraints penalize inheritances to individuals other than

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 14 children or parents of the potential donor. In the United Kingdom, the intestacy law places a higher tax on inheritances passed on to siblings or other relatives (12%) than to inheritances passed on to children, parents or grandchildren (4.5%). 10 In Germany, the tax system also places a disincentive on estates and gifts passed on to individuals who are not descendents or ascendants of the donor. 4.2.2. The marginal propensity to give through intervivos transfers To assess how intervivos transfers change with the lifetime income of individuals, we use the relationship between lifetime income Y and intervivos transfers that parents give to all of their adult children living in other households in a given year (R) in a given year, holding age and other demographic variables constant. 2 R = β o + β1 Y + β 2Y + β3y age 70) + β 4Yk + controls + ε it ( (3) In the former expression, β 1 captures the effect of parental lifetime resources ( Y ) on intervivos transfers in a given year. age denotes the age of the donor, and Y k the mean income of the recipient children. Controls includes variables like the age of chidren, marital status indicators for the donor and the receiver of help. β 1 is identified from the variation in average transfers associated to cross-sectional changes in lifetime income. Economic theory places restrictions on the impact of lifetime income on intervivos transfers, holding the ages of the parent and child constant. If intervivos transfers are motivated by the fact that parents want to improve the welfare of their children, higher income parents will have a lower utility from own consumption, and will give higher transfer, leading to a positive relationship. If, instead, parents give monetary transfers to their children in exchange of some service, we should still observe that higher income parents are more able to buy services, resulting again in a positive β 1. Second, if parents are concerned about the economic status of their children, parents with similar incomes will tend to give higher transfers to their less well-off children. Holding parental income constant, and abstracting from issues like envy or competition among siblings, the relationship between the lifetime income of the child and intervivos transfers must be negative ( β 4 should be negative). Specification (3) also includes controls for the age of the parent age and child to capture systematic changes in needs over the life cycle of both donors and recipients. The parameter of interest is the effect of parental lifetime income on lifetime intervivos transfers, that can be expressed as the sum of transfers given to children during all the life of the parent, that we assume starts giving at 45 until age 86, but discounted to age 70 of the parent, and weighted by the probability of survival. 10 See www.wisewills.co.uk, for details.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 15 E( R Y age Y age E R Y Yk age = = 86 ) 70 (,, ) (1 + r) P( alive = age) Y age = 45 (4) In the previous expression, r denotes the interest rate (4% for all countries), age is the age of the donor, P ( alive = age) is the probability that at least a member of a couple is E( R Y, Yk, age) alive at a given age. is the impact of lifetime income on intervivos Y transfers at a given age of the donor, holding constant the mean income of the recipient children. The calculation is made for West Germany and the United States using a married couple (a white one in the US), and assuming that they face the average mortality rates. That magnitude is obtained using the coefficients β 1 and β 3 in specification (3). 11 The result of this computations tells by how much do intervivos transfers over the lifetime increase with an additional dollar of lifetime income. Section 5 provides reduced-form estimates of the impact of lifetime resources on intervivos transfers and bequests in both countries, and discusses the presence of a bequest motive. Second, Section 6 examines the responses of transfers and bequests to variables correlated with policy measures that vary the economic conditions of parent and children. 5. EMPIRICAL RESULTS 5.1. Estimates of the impact of parental lifetime resources on bequeathable wealth Table 6 presents the estimates of an extra dollar of lifetime resources on wealth holding at age 70 for the United States (first column),west Germany (second column) and the United Kingdom (third column). The coefficient of lifetime resources of the parent ( Y ) for the US is 4.6 cents. This means that, in the United States, a white couple saves 4.6% of an extra dollar of lifetime resources at the age of 70. The estimate of the interaction of parental lifetime resources and the age of the parent is small. It implies that, at age 80 of the parent the impact of lifetime resources on wealth holding is even higher, 5.1 cents. In Figure 1, we present a graphic with the evolution of the impact of an additional dollar of lifetime resources between the ages of 70 and 86 implied by the estimates in Table 5. The impact of lifetime resources on wealth is much lower for widows. The results in Figure 1 suggest that, in the United States an extra dollar of lifetime resources increases wealth holding at age 70 by 2.9 cents for a widow person, and it is 3.4 cents at the age of 80, when the probability of death is highest. 12 11 An example of how this is computed for West Germany is presented in Table A.1 12 Two notes are in order. First, given that the patterns of wealth accumulation among non-white American households are not well understood, we chose to report only those of white households. Second, one could claim that the fact that widows have a lower relationship between lifetime income and wealth reflects bequests made upon the death of the first member of the

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 16 The coefficient of lifetime income for West Germany implies that for the average income household, one extra dollar of lifetime income increases wealth holding at age 70 by 3.2 cents, about 85% of the corresponding magnitude for the United States. Further, the magnitude of the impact of lifetime resources on wealth levels decreases with age more quickly than it does in the United States. Consider the impact of an additional dollar of lifetime resources on wealth holding at the age 80. The results in Figure 1 suggest that, for a German couple, one extra dollar of lifetime income increases wealth holding by age 80 by about one and a half of a cent. Given that the probability of death (and thus of observing a bequest) increases with age, these estimates suggest that, in Germany, very little of an extra dollar of lifetime resources is passed on to the next generation through bequests. The results for the United Kingdom are very similar to those in West Germany. Summarizing the results in Figure 1, for all ages, West German households save less out of lifetime resources than similar American households headed by a white. The gap between the marginal propensity to save out of lifetime resources increases with age. This is likely to result in a smaller bequest out of an extra dollar of lifetime resources in Germany than in the United States. Next, we take each of the age-specific derivatives of the impact of lifetime resources on wealth, and weight them by the probability that the last member of the couple dies at that specific age. Given that there is a high probability that the death of the second spouse is the death of a widow person, the impact should reflect the impact of an extra dollar of lifetime resources on wealth holdings at ages above 80 for a widowed person. Table 7 shows the impact of lifetime resources on expected bequests. The estimate in row 1, column 2 suggests that, for US white couple with the average level of permanent income, out of an extra dollar of lifetime resources, parents pass on about 2 cent through expected bequests to all their children. The estimate at the 10th percentile of the income distribution is 1.3 cents and at the 90th percentile is 2.3 cents. By themselves, these numbers do not tell us much about the presence or not of a bequest motive. Row 2 in Table 7 presents a benchmark: we compute at several points of the income distribution the marginal propensity to bequeath of elderly who did not have children. At the average level of income, the marginal propensity to spend is remarkably similar between both groups. For elderly with children, our estimate of the impact of an additional dollar of lifetime income on expected bequests is 1.9 cents. The corresponding estimate at the sample mean for elderly without children is 1.5 cents, and the difference is not significantly different from zero. Given the limited sample size, we have experimented with alternative specifications of the wealth regression for the sample of elderly without children. The marginal propensity to bequeath among individuals without children is never different from the corresponding estimate for elderly with couple, not the second (early bequests). In that case, the impact of lifetime income on wealth holding in the case of the US would be larger in reality than we are estimating, as we are missing part of the bequest.

INTERVIVOS TRANSFERS AND BEQUESTS IN THREE OECD COUNTRIES 17 children. A possible criticism of this evidence is that the average household is not the most relevant one to compare the marginal propensity to save, as bequests are most relevant higher up in the income distribution. Column (3) in Table 7 shows the estimate of the marginal propensity to bequeath out of an extra dollar of lifetime income at the 90th percentile of the income distribution. The results are even larger for the sample of elderly without children. 13 To the extent that childless elderly have less of a bequest motive than elderly with children, these results suggest that most of the saving for the old age responds to cautionary motives. In Row 3 of Table 7, we present comparable estimates for West Germany. The point of evaluation of the income distribution is not the same as in the US, as the distribution of permanent income is different across countries. An extra dollar of lifetime income increases wealth holding late in life by 1 cent in West Germany, at least at the average of the income distribution. The point estimate is significantly lower than the corresponding US estimate at the 1 percent confidence level. The estimate for Germany is higher at the 90th percentile of the income distribution: around 1.5 cents. Again, the point estimate is lower than the corresponding estimate for the United States. The comparison with the same magnitude for childless elderly gives another useful benchmark, especially at the average income level, where the estimate is most precise. The computation of the marginal propensity to bequeath in West Germany for elderly couples is 1 cent, very similar to the corresponding estimate for individuals with descendents. In the German case, at the 90 th percentile of the income distribution, the marginal propensity to bequeath is 1.53 cents (Table 7, row 3, Column 3). The point estimate is somewhat larger than the corresponding estimate for elderly without children, (one cent, row 4 Column 3) but the difference is not significantly different from zero. Finally, the results for the United Kingdom are remarkably similar to those in Germany. Parents with children leave a bequest of around 1 cent at the average of the income distribution, an estimate that is identical to the corresponding estimate for individuals without children. Overall, we draw two conclusions from the results in Table 7. The first is that out of an extra dollar of lifetime income, white American households headed by a white pass on a higher fraction of resources through bequests than households in West Germany and in the United Kingdom. 14 Table A4 presents alternative estimates of the marginal propensity to bequeath using an interest rate of 2% A lower interest rate exacerbates the differences between the United States on one hand and Germany and the United Kingdom on the other. Second, in the three countries under consideration, the marginal propensity to bequeath of individuals at the average of the income distribution, and without children 13 The regression of wealth on the lifetime income of individuals without children is in the appendix Table A5 14 Box 5 discusses the implications of these estimates for the process of transmission of income inequality across generations