American Economic Association How Income Transfer Programs Affect Work, Savings, and the Income Distribution: A Critical Review Author(s): Sheldon Danziger, Robert Haveman and Robert Plotnick Source: Journal of Economic Literature, Vol. 19, No. 3 (Sep., 1981), pp. 975-1028 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2724326 Accessed: 16-12-2015 21:11 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic Literature. http://www.jstor.org This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Journal of Economic Literature Vol. XIX (September 1981), pp. 975-1028 How Programs Income Transfer Affect Savings, and Work, the Income Distribution: A Critical Review By SHELDON DANZIGER and ROBERT HAVEMAN University of Wisconsin-Madison and ROBERT PLOTNICK Dartmouth College For helpful comments on earlier drafts, we thank, without implicating, Moses Abramovitz, Yves Balcer, John Bishop, Alan Blinder, Richard Burkhauser, Michael Darby, Irwin Garfinkel, Alan Gustman, Daniel Hamermesh, Martin Holmer, GeorgeJakubson, Robert Lampman, Paul Menchik, Robert Moffitt, Michael Murray,Joseph Quinn, Timothy Smeeding, Eugene Smolensky, Barbara Wolfe and two anonymous referees. The research reported here was supported in part by funds granted to the Institute for Researchon Poverty at the University of Wisconsin-Madison by the Department of Health and Human Services pursuant to the provisions of the Economic Opportunity Act of 1964. I. Introduction pUBLIC SPENDING on income transfers is large and has grown rapidly in recent years as new programs have been enacted, benefit levels in most programs have increased, and eligibility requirements have been loosened. In 1981, expenditures on these programs are estimated to reach almost $300 billion, an amount that is about 10 percent of the Gross National Product. As these programs have absorbed a rising share of the public sector budget and redistributed a sizable fraction of personal income, they have generated substantial controversy. Some have charged that the incentives implicit in income support programs discourage work and private savings and reduce capital accumulation. They suggest that the problems of poverty and income insecurity which motivated the development of these programs are now less serious, and that the efficiency costs of achieving further redistribution are large. The literature on the factor supply and redistributive effects of transfer programs has grown rapidly in recent years. Economists have identified and estimated both the potential behavioral responses to the 975 This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 976 Journal of Economic Literature, Vol. XIX (September 1981) diverse incentives implicit in some of these programs and their redistributive impacts. The empirical literature has steadily improved due to new data sets and econometric advances. However, there is no consensus on the magnitude of the economic and redistributive effects of transfers. Our objective is to review and appraise the literature on transfers' effects on labor supply, private savings, poverty and income inequality. Section II briefly describes the nature and size of those transfer programs on which our evaluation is based. Section III discusses the expected factor supply and redistributive effects derived from the theoretical literature. Section IV assesses the empirical evidence and presents our conclusions regarding the magnitude of effects in the current programs. We also offer suggestions for improving the reliability of the estimates. Section V surveys research on the effects of proposed reforms in existing programs, and Section VI is a conclusion. This review addresses three important and controversial impacts of income transfer programs. However, it does not analyze all of their effects. Taxes to finance these programs have their own factor supply and redistributive consequences. Transfers may also affect demographic choices, including migration, marriage, divorce, childbearing, choice of living arrangements (Maurice MacDonald and Isabel Sawhill, 1978; John Bishop, 1980). In addition, the unemployment rate (Daniel Hamermesh, 1977,1980), cyclical stability (Hamermesh, 1977; U.S. Department of Health, Education and Welfare, 1975), regional income distribution, the level and composition of demand (Frederick Golladay and Robert Haveman, 1977), productivity (Moses Abramovitz, 1981), the position of the short-run Phillips curve (Haveman, 1978), the formation of human capital, the choice of jobs, political stabil- ity and attitudes towards work (Assar Lindbeck, 1980) may all be affected by income transfers. While they may create welfare losses (Edgar Browning, 1978), transfers may also increase efficiencye.g., improved job-worker matches, improved worker health, and increased willingness to accept technological change or to bear risk (Martin Baily, 1977; Frank Stafford, 1977; Peter Diamond and James Mirrlees, 1978). The literature on the effects of transfers on these other variables will not be considered here (but see the partial survey in Sheldon Danziger, Haveman and Robert Plotnick, 1980). II. Overview of Income Transfer Programs TABLE 1 lists the income transfer programs on which our review focuses, and shows their expenditures for 1965 and 1979, as well as estimates for 1981. These programs are divided into two types: social insurance and welfare programs. Within each category are programs providing cash income and others providing in-kind benefits.' The programs have two basic objecincome losses from tives-replacing events that are largely outside an individual's control, and assuring a minimum level of economic support to those who, have little other income. The first objective is largely served by social insurance programs,for which eligibility and benefit levels depend on past contributions and some identifiable problem, such as old age, death of spouse, illness, disability, or unemployment. One does not have to prove financial need to claim benefits. Social inI Transferprogramsare not synonymouswith "social welfare" programs.The latter include cash and in-kind income transfers,but also public outlays on education, job training, employment services, and diverse social services. Thus, this paper does not assess the factor supply and redistributive impacts of the entire "welfare state." This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 977 Danziger, Haveman and Plotnick: Transfers and Their Effects TABLE 1 EXPENDITURES ON MAJOR INCOME TRANSFER PROGRAMS PUBLIC EXPENDITURES DATE ENACTED Social Insurance Cash benefits: Social Security (OASDI) Unemployment Insurance Workers'Compensation Veterans' Disability Compensation RailroadRetirement Black Lung In-kind benefits: Medicare Public Assistance (Welfare) Cash benefits: Aid to Families with Dependent Children (AFDC) Supplemental Security Income (SSI)' Veterans' Pensions General Assistance In-kind benefits: Medicaid2 Food Stamps Housing Assistance Total Expenditures Total Expenditures as a percentage of GNP (Billionsof Current Dollars) 1981 (ESTIMATE) 1965 1979 1935 1935 1908 1917 1937 1969 $16.5 2.5 1.8 2.2 1.1 NE $102.6 11.2 9.9 6.8 4.3 0.6 $137.0 18.7 14.8 7.5 5.2 0.9 1965 NE 29.1 38.4 1935 1972 1933 NA 1.7 2.7 1.9 0.4 10.8 6.8 3.6 1.2 12.8 8.5 4.1 1.5 1965 1964 1937 0.5 0.04 0.3 $31.5 4.6 21.8 6.8 4.4 $219.9 9.1 27.6 9.7 6.6 $293.2 10.0 Sources:TheBudget of the United States Government,Fiscal Year,1981, and its Appendix, for 1979 expenditures and 1981 estimates. Robert Plotnick and Felicity Skidmore (1975) for 1965 data. Notes:Social Insurance programs condition benefits on contributionsbased on previous employment; public assistanceprograms condition benefits on current income and assets (means-tested). NA = not applicable, varies by states. NE = nonexistent I Aid to the Blind, Aid to the Permanently and Totally Disabled, and Old Age Assistancein 1965. 2Medical Aid to the Aged in 1965. surance accounts for nearly three-quarters of the expenditures. The second objective is served by the public assistance (welfare) programs, for which inadequate economic means is the chief eligibility criterion. Receipt of welfare benefits is not conditioned on past contributions. Benefits are asset- and income-tested-they vary inversely with income from private sources and social insurance. Transfer programs have grown rapidly in recent years with expenditures increasing from 4.6 percent of GNP in 1965 to 10.0 percent in 1981. TABLE 2 shows the increase in both the number of beneficiaries of the cash transfer programs and the size of the average benefit, and compares this growth with that of Census money income. In 1965, 37 percent of all households received a cash transfer; by 1978, 42 percent were recipients. The last col- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 978 Journal of Economic Literature, Vol. XIX (September 1981) TABLE 2 CASH INCOME TRANSFERS AND CENSUS MONEY INCOME OF HOUSEHOLDS IN CONSTANT 1965 AND 1978a Social Security and RailroadRetirement Public Assistanceb Other Cash Government Transfersc One or More Cash Transfersd Total CensusMoney Income, all households CashTransfersas a Percentage of Money Income 1965-1978 1978 1965 MEAN FOR RECIPIENT HOUSEHOLDS PERCENT OF HOUSEHOLDS RECEIVING $ 2407 2006 22% 5 MEAN FOR RECIPIENT HOUSEHOLDS $ 3747 2079 1978 DOLLARS, PERCENT OF HOUSEHOLDS RECEIVING REAL GROWTH OF MEAN 26% 8 55.7% 3.6 1801 18 2973 17 65.1 2532 37 3931 42 55.3 13767 16518 6.8% 10.0% 20.0 Source: Computations by authors from 1966 Survey of Economic Opportunity and March 1979 Current Population Survey. a Households include families and unrelated individuals.The programs represented here differ from those in TABLE 1. The Census does not gather data on in-kind transfers, nor does it disaggregate cash transfers by program to the extent shown in TABLE 1. bIncludesAid to Families with Dependent Children, Supplemental Security Income (Old Age Assistance, Aid to the Blind, and Aid to the Permanently and Totally Disabled in 1965), and General Assistance. cIncludes unemployment compensation,workers'compensation,government employee pensions, and veterans' pensions and compensation. dThe mean value in this row exceeds the mean for any individual category, and the percentage receiving one or more transfer is lower than the sum of rows 1-3, because some households receive more than one transfer. umn shows that the average transfer for recipient households increased by 55.3 percent, while Census money income for all households increased by only 20 percent. Since 1965, transfers have increased both in absolute terms and relative to nontransfer sources of income. Simultaneously, labor force participation rates for men have decreased. However, the aggregate participation rate has increased because of the rapid growth in labor supply by women. During the same period, the rate of saving out of disposable income has declined, poverty has decreased, and income inequality has remained relatively constant. Because of these recent trends, researchers have examined the relation- ship between transfers and aggregate labor supply, private savings, poverty and inequality. We now describe and evaluate this research. III. The Labor Supply, Savings and Redistributive Effects of Income Transfers: TheoreticalIssues In this section, theoretical approaches for evaluating the impacts of transfers are reviewed. The literature on factor supply effects is discussed first;that on redistributive effects follows. A. Factor Supply Effects Individuals respond to changes in expected net income or wealth or the net This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects prices of working and saving brought about by transfer programs. The magnitudes of these changes are difficult to measure. Without these programs, both private behavior (e.g., labor force participation, retirement, savings, living arrangements, and charitable giving) and institutions (e.g., insurance and pension funds, employer-provided fringe benefits, accepted retirement ages) might be different in scope and nature. As indicated below, our evaluation is based on the assumption that the work and savings incentives and the redistributive mechanisms in the private economy with public transfer programs in place are similar to those that would exist without the programs. 1. Labor Supply. The effect of income transfers on labor supply is generally analyzed on the basis of standard consumer theory. An individual maximizes utility by choosing among income-leisure options, given the budget constraint. Because most transfer programs provide income support without requiring work and reduce benefits when earnings increase, the recipient's budget constraint is shifted in position and slope. Hence, two program parameters-the income guarantee and the marginal tax (or benefit reduction) rate-are the most important in analyzing work incentive impacts.2 The guarantee produces an income effect; the tax rate, both an income and a substitutioneffect. If leisure is a normal good, a transfer with a positive guarantee and tax rate creates income and substitution effects which unambiguously reduce labor supply. A program with a 2 Other characteristicsof actual programsmay also affect labor supply. A limitation on earnings for retention of program eligibility, as in Medicaid, can be interpreted as an infinite tax rate at the earnings limit. A work test associatedwith a transferprogram tends to increase labor supply relative to its absence and may induce lower net market wages and higher employment. See Robert Moffitt, 1978. 979 zero tax rate has only an income effect that reduces supply.3 Transfer programs may also induce an intertemporal labor supply response. Martin Feldstein (1974) shows that a permanent age-related transferprogram may induce earlier retirement or a decrease in labor supply in pre-retirement years because potential beneficiaries will need less private pre-retirement savings to support consumption in retirement years.4 This standard model has been extended to account for the nature of real world constraints, the complexity of actual income transfer programs, and the nature of utility functions. Giora Hanoch and MarjorieHonig (1978) show that when the budget constraint is kinked, either because of peculiarities in program design or the simultaneous existence of several transfer programs, labor supply functions may be discontinuous, have both forward and backward-bending segments, or have kinks. As a result, increases in the guarantee or tax rate may bring about aggregate responses different from those suggested above. While the standard model treats work effort as a continuous variable, an individual's response may be discontinuous (e.g., withdrawal from the labor force, as in early retirement). Constraints on work time flexibility (e.g., employer-imposed constraints on work hours) also produce discontinuous responses. Mark Killingsworth (1976) suggests that the standard model does not yield predictions about any individual family member's labor supply if this decision is made jointly with those of other household members. 3In addition to these conclusions, search theory also implies that transfers, particularly unemployment insurance,may increasejob search and reduce employment. 4This reduction in pre-retirement work effort would, itself, tend to induce a delay in retirement age. In effect, the reduction in pre-retirement earnings would have its own "income effect" on the retirement decision. The following discussionof Social Security covers other intertemporal effects. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 980 Journal of Economic Literature, Vol. XIX (September 1981) The behavior of any such family member depends on cross-substitution effects involving other members' wages and work decisions, the signs of which are indeterminate. Finally, "entitlement effects" may induce some persons to enter the work force to qualify for social insurance programs, or to work more to raise their future benefits from such programs (Hamermesh, 1979). In spite of these reservations about or extensions of standardtheory, aggregate work effort is, on balance, expected to be lower with than without transfers,but the size of this effect remains an empirical matter.5 2. Private Savings. The price and wealth effects of public transfers, especially those which are agerelated, may also affect consumption behavior. In the life-cycle framework of Roy Harrod (1948), Albert Ando and Franco Modigliani (1963), and Modigliani and RichardBrumberg (1955), saving is undertaken during working years to support consumption during retirement. Within this framework, compulsory Social Security benefits provide an alternative to private savings. If private savings were reduced by the same amount as public savings from Social Security tax revenues were increased, there would be no net reduction in aggregate savings.6However, 'Dynamic models of labor supply with life cycle considerationsand possibilities for lending and borrowing have recently been developed (JamesHeckman and Thomas MaCurdy, 1980; MaCurdy,1981). These models incorporate the dependence of current work effort on future wage rates, asset expectations, time preferences, and discount rates (Heckman, Killingsworth,and MaCurdy, 1979), but have not yet analyzed the effects of existing income transfers. 6 This condition will obviously hold when the retirement decision is unaffected by the transfer program. However, even when revenue raised from an individual to finance the program is repaid to the same individual as benefits, retirement decisions may be affected (Eytan Sheshinski, 1978). Yves Balcer (1981) analyzes this issue and its relationship with actuarialfairness and retirement neutrality. because this program operates on a payas-you-go basis, there is no public saving, only a transfer from current workers to current retirees. Hence, if the present value of benefits appears larger than the present value of Social Security taxes in such a system, private (and therefore aggregate) savings may decrease.7 In an important paper, Feldstein (1974) extended the life-cycle framework and analyzed the simultaneousimpacts of an agerelated transfer program on labor supply and savings. The potential negative savings impact due to life-cycle considerations is isolated from the incentive for retirement at younger ages. As a result of the retirement effect, saving during working years may increase in order to support consumption over a now-longer retirement period. This effect offsets the reduced saving impact from the simple life-cycle model, leaving the net impact indeterminate. Age-related transfers may induce a reduction in saving through another channel. Ando and Modigliani (1963) suggest that an increase in expected retirement transfers may be perceived as an increment to net wealth, resulting in increased consumption and less saving. While the future tax liabilities required to finance an actuarially fair program would appear to offset this benefit-related wealth effect, the offset may not be equivalent. This could occur if, because of finite lives, the time horizon relevant for appraising the wealth effects of the expected taxes is shorter than that for appraising future benefits or if the discount rates applied to future expected benefits and taxes tdiffer due to imperfect capital markets. With a less than equivalent offset, a positive net wealth effect could exist, leading to reduced private savings. The net wealth basis for expecting a 7This discussionrelates to both the taxes and transfers implicit in a pay-as-you-goSocial Security program. If only the transfers are considered, Social Security would substitute for private savings. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects negative effect on savings has been analyzed by Robert Barro (1974) for individuals living in overlapping generations. He demonstrates that, in an effort to maintain the size of bequests, individuals experiencing a net wealth increase from age-related transfersmay increase private transfers to the next generation. The implied increase in required savings to meet such bequests may offset the negative wealth effect on private savings in pre-retirement years. Also, if children would have supported their parents in old age in the absence of Social Security, its provision would operate to reduce these negative bequests. Hence, the imposition of Social Security would have no net effect on lifetime consumption, but rather could be exactly offset by changes in the flow of intergenerational transfers. The private transfer offset to age-related public transfers may not exist because some individuals may want to leave negative bequests but are constrained from doing so. Similarly,if the intergenerational transfer is in the form of human capital rather than cash, and organized human capital markets do not exist, a net wealth effect attributable to age-related transfers would remain (Alan Drazen, 1978). The conditions required for the absence of a net wealth effect have been debated in Feldstein (1976 and 1978b), and Barro (1978). Other perspectives on the potential savings effects of age-related transfers have also been suggested. Phillip Cagan (1965) and George Katona (1965) hypothesize an increased saving impact due to either a "recognition effect" (once having been alerted to retirement needs by the program, private savings are stimulated) or a "goal gradient effect" (the closer one is to achieving desired provision for retirement, the greater is the effort to secure the goal). And Feldstein (1974) suggests that an income effect could increase savings if the consumer's perceived budget constraint is expanded by the 981 program'sassuranceof retirement support. With constrained or distorted markets, the level of private savings may exceed that which is socially optimal. In this case, reduced savings induced by Social Security may correct private market failure. Examples of market distortions which may encourage excessive private savings include: institutional constraints on bequest possibilities, imperfect human capital markets, collectively-bargainedprivate pension programs, tax incentives for retirement savings, or privately relevant but socially irrelevant uncertainties in future income flows (James Davies, 1981). If public transfers are endogenous to the economy, they may be the mechanism for offsetting institutional constraints and reducing private uncertainty, hence moving private saving toward its optimum level (John Hagens, 1979). Or again, considered in the context of a neoclassical general equilibrium growth framework, age-related transferscan lead to long-run increases in employment, retirement age, and per capita savings. Partial equilibrium analysissuggests short-run decreases in these variables as individuals respond to the labor supply incentives in Social Security programs. In the long-run, however, a pay-as-you-go system can induce additional public capital formation and demands for labor, resulting in increased work effort, employment, and economic growth. These dynamic efficiencies, operating through savings and bequest elasticities and the economic growth process, may more than offset any decrease in static efficiency brought about by Social Security (Sheng Hu, 1979). While the above discussion has focused on the savings effects of age-related transfers, other redistributive transferscan also affect savings. Consider, for example, transfers such as unemployment benefits, workers' compensation, or survivors'benefits which reduce temporary losses of income. If savings are undertaken for precautionary motives in addition to life- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 982 Journal of Economic Literature, Vol. XIX (September 1981) cycle considerations, these transfers may reduce private savings. Another channel of impact may be through differing consumption patterns of the net beneficiaries and cost bearers of transfers.At the margin, the poor appear to save proportionately less than do the rich, notwithstanding the models of consumer behavior of Milton Friedman (1957) and Modigliani and Brumberg (1955) which imply marginal propensities to consume invariant with permanent income. Thus, increases in transfers to the poor may increase consumption and reduce savings. This inconsistency between the "obvious" and life-cycle models led Alan Blinder (1975) to add a bequest motive to standard optimal life-cycle consumption models. He shows that if consumption is a luxury good (relative to bequests) with respect to lifetime income, transfers to poor individuals (identical in all respects except income) will increase aggregate' savings.8 Conversely, if bequests are a luxury good, net transfers to the poor will decrease saving.9Again, the predictions of the theory are ambiguous. Finally, the theoretical models generally assume a fully employed economy. In a slack economy, private savings need not decrease if transfers induce an increase in aggregate demand. Indeed, in the absence of evidence that transfers "crowd out" investment spending when unem8Phillip Musgrove (1980) presents a household model which suggests that aggregate consumption is positively related to the level of income in excess of that necessary to meet basic needs, and also to its distributionand its asymmetry. 9 Based on a small sample (less than 200) of children of rich people, Paul Menchik (1980) finds bequests to be a luxurygood, having a lifetime income elasticity in excess of two. Employing a larger,more representative sample, Menchik and Martin David (1979) find that bequests are supplied inelastically over the bottom 80 percent of the lifetime income distributionand elastically over the top 20 percent. The implication is that equalizing redistribution within the bottom four quintiles does not reduce saving but redistribution from the top quintile to the bottom four does. ployment exists, the expected increase in aggregate consumption demand may stimulate investment. In sum, economic theory provides no clear prediction on the direction of the private savings response to transfer programs. While theoretical considerations guide empirical research on behavioral responses, the uncertainty can only be resolved by examining the empirical evidence. B. Redistributive Effects Any analysis of the effects of transfers on the size distribution of income must begin with a choice from each of four constructs: the pretransfer and posttransfer income concepts, the income unit, and the income accounting period. For any time period, size distributions are constructed by measuring incomes of living units and then arrayingthe units in order of increasing incomes. Although few principles guide the choice of these constructs, there is agreement that an ideal study should measure inequality in command over resources, among income-sharing units, over some specified time period, adjusting for "needs" and life-cycle differences. Several methods for both refining the income concept (Browning, 1976; Irwin Garfinkeland Haveman, 1977; Marilyn Moon and Eugene Smolensky, 1977; Timothy Smeeding, 1975; and Michael Taussig, 1973) and adjusting for changes in the income unit (Danziger and Plotnick, 1977; Danziger and Taussig, 1979) have been suggested. Research has also emphasized the inadequacies of analyses based on the distribution of annual well-being, relative to a multiyear or even lifetime construct (Jacob Benus and James Morgan,1975; Roger Gordon, 1977; Richard Layard, 1977; Lee Lillard, 1977; David and Menchik, 1979). The redistributive effect of transfers is generally measured by comparing pretransfer and posttransferincome distribu- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects tions. This comparison assumes that transfers elicit no behavioral responses that would cause income without transfers to deviate from observed pretransfer income. However, the potential savings and labor supply effects discussed above suggest that recipients' net incomes may not be increased by the full amount of their transfer benefits.10Also, studies of the redistributive effects of transfers generally assume that the observed posttransfer income-sharing unit and the period of income receipt are not affected by the transfer. To the extent that the availability of transfersinduces members of a household to alter their living arrangements or the timing of the income flows, estimates of the redistributive effects of transfers will be biased. Most income transfers are expected to reduce measured poverty and inequality. Social insurance transfers are not restricted to those households or persons with low incomes, but their objective of cushioning earnings losses associated with unemployment, disability, retirement, or death causes them to be negatively correlated with current income or employment status. Public assistance transfers are generally received only by persons and households at the lower end of the income distribution. IV. The Labor Supply, Savings and Redistributive Effects of Income Transfer Programs:Empirical Results Recent theoretical and policy interests have led to a large applied literature on 10 Consider an individual who earns $3000. After the passage of a public assistance transferprogram, with an income guarantee of $3000 and a tax rate of 50%, the person reduces hours of work and earns $2500. A transferof $1750 is now received and total income is $4250. But the individual's final income is only $1250 higher. Because pretransfer income in the absence of transfers is not observed, most studies measure the redistributive effect as the difference between observed pretransfer and posttransfer income ($4250-$2500), not as the increase in final income in the two states. 983 the factor supply and redistributive effects of transfer programs. The most important empirical studies are summarized, compared, and assessed in this section. A. Labor Supply Effects 1. Old Age and Survivors Insurance. Because of its size, the Social Security retirement program-Old Age and Survivors Insurance (OASI)-has a potentially large labor supply effect. Almost all citizens are actual or potential beneficiaries and are thus affected by several work incentives implicit in the program. First, all covered workers, irrespective of age, view it as both an asset yielding future benefits and a liability requiring current and future tax payments. If the program is appraised as a net asset (the present value of the stream of expected net benefits is positive), a wealth effect may reduce work effort prior to retirement. The receipt of benefits after retirement creates an income effect that discourages work effort during those years. Second, OASI's earnings test on recipients under age 72, whereby benefits are reduced by 50 percent of earnings in excess of $5000 (in 1980), may discourage work effort among some eligibles." If the earnings test does reduce work among retirees, it may induce an increase among younger workers who substitute work when young to offset the constraint on work whe-n older (Buirk- 11After 1981, the work test will cease at age 70. For retirees between age 62 and 65 who lose some current benefits because of the test, there is an upward adjustmentof benefits after age 65. This adjustment reduces the work effort impact of the test on younger retirees, but does not eliminate it if the adjustmentis less than actuariallyfair (RichardBurkhauser, 1980). In some cases, the adjustmentis more than fair and creates a net work incentive (Blinder, Gordon and Donald Wise, 1980). Burkhauser and John Turner (1981) claim that Blinder, Gordon, and Wise miscalculate and, hence, overstate this adjustment. For those older than 65, there is also an adjustment, but it is always actuariallyunfair. As a result, the test creates a disincentive. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0 ,*. 0 > +- 0 4) 4) $., 0 U,t 00~~~~~~0 0~~~~~~~~~~~~~~0 ~ ~ '54) C cj 0 0 0~~~~~~~~~~~~ o ~~~~~~~~~~~~~~~~~~I z 4 0 CA 0 0~~~~~~~~~~ 0 - -4 1- 0 0 Lo. 4) 1-4lcd bO LO. , C)~~~~~~~~~~~~~~~~~~~~~~~~~~~)C This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions ~ )0 0 o ~. 7, o C)- w ~ ~ ~ ~ ~ ~ ~ ~ ~ o 0 4~ 00 0 0~~~~~~ 0 0~~~ 0~~~~~~~~~~~~~~~c - J34 o 2 4-- C) ~ ~ o .d -U z ) 4.4 - -,- 00 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ;- 0 This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 986 Journal of Economic Literature, Vol. XIX (September 1981) of retiring. Reimers finds no evidence of such a relationship-indeed, for some specifications, the impact was negative and significant. Burkhauser and Joseph Quinn (1980) examine the effect on the retirement decision of the level of and change in OASI'sasset value, which is the conceptually correct program variable in a life cycle model. They report that the 1980).12 TABLE 3 summarizes the principal emlevel of Social Security wealth affects neipirical studies of the labor supply rether the probability of leaving one's most recent job, nor, for those who did leave sponses to these incentives.13The firstfour it, the probabilityof retiring instead of takstudies estimate the probability of fully or partially retiring over some time period.14 ing another job. The authors suggest that Michael Boskin (1977), Boskin and Mi- this insignificant effect may be due to an chael Hurd (1978) and Cordelia Reimers unobserved correlation between taste for (1977) use annual benefits or OASIeligibilwork and wealth, an argument similar to ity as program variables. The first two that advanced by David Greenberg and MarvinKosters (1973), who discuss the difanalyses show a positive, statistically significant effect of OASI on the likelihood ficulty of identifying the effect of property income on work hours.15However, if a delay in leaving a job reduces OASI wealth 12This provision is not widely understood and, thus, may not affect the behavior of most eligibles. by $1000, the probability of leaving rises 13We do not discuss international cross-section by 0.03 (about 11 percent at the mean). work (e.g.,Joseph Pechman, Henry Aaron,and TausAmong men who do leave their jobs, a Robert sig, 1968). Clark,Juanita Kreps, and Joseph Spengler (1978), Thomas Gustafson(1979) and Gary similar change has an insignificant effect Fields and Olivia Mitchell (1980) also provide reon the chance of retiring.16 views for the U.S. We do not discusssurvivor'sinsurThe second four studies estimate the no of have ance because laborsupply analyses it been conducted. probability of being retired in a particular 14These studies and many of the others reviewed year-a static approach. Quinn (1977) are subject to methodological criticisms concerning finds that OASI eligibility significantly their specificationof variables.Glen Cain and Harold Watts (1973) display 18 different measures of labor raises this probability, especially for men supply,some of which may sufferfrom several errors. with health problems. The studies by GorGiven involuntary unemployment and fixed workdon and Blinder (1980) and Anthony hour employment packages, observed work hours may not equal desired work effort (Orley AshenfelPellechio (1978a) adopt an explicit market ter, 1980;Moffitt,1980b).The "weeksworked"variawage-reservation wage decision frameble used to construct annual work time often means work and more advanced econometric "weeks employed" and includes vacation time, and, hence overstates work effort (Heckman, 1980). Dimethods. Retirement occurs when the viding annual earnings by an observed or imputed market wage is less than the individual's wage (Boskin, 1973; Robert Hall, 1973; Heckman, reservation wage. The latter is estimated 1980) yields an oft-used alternative value for hours hauser and John Turner, 1978). However, the provision for benefit recomputation (which occurs when a recipient's current earnings exceed the lowest year in the individual's earnings history) serves as an implicit wage subsidy that may blunt (or totally offset) both of these effects of the earnings test (Blinder, Gordon, and Wise, worked which may be biased due to either reporting or imputation errors. Heckman, Killingsworthand MaCurdy(1979) and James P. Smith (1980) discuss how measurement issues may bias labor supply estimates. A majorproblem concerns the proper measurement of wage rates, which often must be imputed or constructed from earningsand hours information(GeorgeBorjas, 1980). Also, reported nonemployment income omits returns from homes and durable goods, often ignores in-kind transfers,and is understated by respondents (U.S. Bureau of the Census, 1978). 15 Similar problems may apply to OASI wealth variablesused in savings analyses.High OASIwealth reflects high past earnings, which may stem from tastes for work. These tastes may be correlated with tastes for assets and lead to greater savings. 16 Among men who leave their jobs, the insignificant effect of a change in OASI wealth may occur if many quitters take part-timejobs (to keep earnings under the earnings test limit) instead of completely retiring. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects as a function of economic and demographic variables, expected private pension benefits, and the asset value of OASI. In Gordon and Blinder, the OASIvariable is insignificant for workers younger than 62 and positive and significant-though not large economically-for older workers. At the mean, a 14 percent increase in the OASIvariable raises the probability of being retired by less than 0.005. Pellechio's results stand in sharp contrast: among men eligible for OASI (over 62 years old) and with low or average wage rates ($3-7 per hour), a 20 percent increase in OASIwealth raises the probability of being retired by 0.06-0.12 above the mean of 0.78. Clark and Thomas Johnson (1980) provide the only analysis of the joint retirement choices of wives and husbands. For husbands,eligibility for OASI significantly raises the probability of being retired by about 0.12. OASI wealth has a statistically significant, but small negative effect on this probability(perhapsfor the reason advanced by Burkhauser and Quinn). A wife's eligibility and asset value have insignificant effects on a husband's choice. Among wives, their eligibility has a work disincentive roughly similar to that for husbands,while husband's asset value also has a significant, but smaller impact on discouraging work. This study as well as -those of Gordon and Blinder and Burkhauser and Quinn find private pensions to have a larger effect than OASI on male retirement choices. Pellechio's second paper (1978b) focuses on aged men who are still working to examine the disincentives of the earnings test. The kink in the budget constraint created by this test requires special estimation methods.17 Using them, he 17Transferprogramscan create kinked and possibly nonconvex budget constraints.In such cases, the net wage and hours worked are jointly determined and the entire budget set must be considered in assessing labor supply choices. Gary Burtless and JerryHausman(1978),Hausman(1980, 1981),Moffitt 987 finds that the test reduces work by 151 hours (8 percent). OASI wealth has a negative, but statistically insignificant coefficient.18 The last three studies examine the effect of OASI incentives on surrogatesfor labor supply-planned retirement age (Arden Hall and Terry Johnson, 1980; and Laurence Kotlikoff, 1979b) and the decision to accept OASI benefits (Burkhauser, 1980). Disparate effects are obtained. Hall and Johnson find that eligibility for the program significantly accelerates retirement plans. Burkhauser finds that, at the mean, increasing OASI asset value by $3,000 raises the probability of OASI acceptance at age 62 by 0.027 (from 0.212 to 0.239). Kotlikoffreports statistically insignificant effects. While most studies suggest that the labor supply and retirement decisions of the aged are negatively affected by OASI, the size of the disincentive is still in doubt.19 Several reasons can account for the disparities-differences in data sets, regression techniques, and differences in the dependent variable (e.g., the retirement decision defined as zero hours worked, planned retirement age, work hours, acceptance of OASI benefits). Important exand Walter Nicholson (1982, forthcoming),Pellechio (1979) and Terrence Wales and Alan Woodland (1979), apply maximum likelihood estimation techniques in such situations. 18 Pellechio overlooks the benefit recomputation rule, which offsets this kink. However, little bias will result if few people understand the rule and, thus, alter behavior because of it. In contrast, the operation of the work test appears to be widely known. 19Burkhauserand Turner (1978) estimate that the effect of the earnings test in inducing a substitution of work when younger for work when older dominates the wealth effect of OASIin discouragingwork effortwhen young, and induces a 2-3 hoursper week increase in market work among younger workers for the 1929-71 period. Using only post-1946 data, they find no significant effect. Estimates based on the replacement of the erroneous Social Security variable computed by Feldstein (see below) with a correct version give similar results. As we argue in our discussionof the applied work on Social Security and savings, such time series evidence must be regarded as very tentative. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 988 Journal of Economic Literature, Vol. XIX (September 1981) planatory variables (e.g., private pension wealth-the omission of which will bias the OASI coefficient upwards-health status, local labor market conditions) are absent in some studies and measured poorly in others. Some studies lack information on OASI asset value, and employ weak surrogates such as OASI eligibility or annual benefits. While Pellechio appropriately estimates the impact of the kinked budget constraint created by the work test, others do not. Finally, wages, earnings, and potential OASI are estimated without correcting for selectivity bias in some of the studies.20Of the 12 studies in TABLE 3, those by Burkhauser and Quinn, Gordon and Blinder, Pellechio (1978a) and Clark and Johnson appear most satisfactory to us and probably bracket the true retirement effect of OASI. These four all use the appropriate OASI variables and adequate statistical methods, though none allows for the possibility of partial retirement. Pellechio's data lacked health and pension variables, so his results, which suggest the largest impact by far, are likely to be upward biased.21 Future research on retirement should 20Selection bias appears when non-randomly selected samples are used to estimate behavioralrelationships. Important early papers include Cain and Watts(1973),Reuben Gronau(1974),H. Gregg Lewis (1974) and Heckman (1974). Heckman (1976, 1979, 1980); Hausman and David Wise (1977); Hausman (1979) and others further develop methods for dealing with the issue. Biasmay be due to self-selection by the individuals being studied. For example, the wages of aged men who choose to work do not necessarily give valid estimates of potential wages of nonparticipants.Sample choices, such as restricting analysisto low income families, are a second cause of bias. Attrition is a third. 21 Pellechio shows that decreasing SSW from its mean to zero would raise the probability of labor force participation in 1973 by over 34 percentage points for most 65 year old men. Yet from 1940 to 1973, the participation rate for aged men fell only 22 percentage points. It is unlikely that, in the absence of OASI,participationin 1973 would have exceeded its 1940 value. exploit the advantages of panel data, which current studies have failed to do.22 Better data on potential pension income, assets, and property income are needed. In addition, the endogeneity of these variables assumes greater importance for the elderly than other groups, but has not been adequately treated. More attention is warranted to the transition from fulltime work to part-time work to complete retirement, and to female retirement behavior, particularlyin view of the increasing importance of women in the labor market. And, the response of younger workers to the Social Security system has not been adequately researched. 2. Disability Insurance. Outlays for the Disability Insurance (DI) component of the Social Security system have increased rapidly in recent years. While DI has a stringent definition of disablement, higher benefit levels and relaxed eligibility determination may have induced some persons to accept DI and leave the labor force. This may account for some of the recent decline in male labor force participation rates, especially for older workers (Frederic Siskind, 1975). The three studies in TABLE 4, as well as a number of earlier studies employing less adequate data and techniques have addressed this issue, (Harold Luft, 1975; Richard Scheffier and George Iden, 1974; Monroe Berkowitz, William Johnson, and Edward Murphy, 1976).23 Donald Parsons (1980a, 1980b) finds that the probability of labor force partici22Panel data (containing similar information on each of a sample of individualsfor each of a number of time periods) enable the estimation of dynamic models. However, panels experience attrition and are often of limited size. Such data allow better control for unobserved heterogeneity (person specific effects) and may partially mitigate omitted variable problems. 23Haveman,Jennifer Warlick,and BarbaraWolfe (1981) review past research and discuss the issues in modeling disability transfer-laborsupply relationships. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects 989 TABLE 4 LABOR SUPPLY ANALYSES OF DISABILITY INSURANCE STUDY Population Analyzed Data Used' Dependent Variable Program Variables Specification Results PARSONS LEONARD (1980a, 1980b) (1979) Men, 48-62(a) or 4559(b) NLS, 1969(a) or 1966(b) Participationin work force Potential DI/wage Probit Elast. of participation w.r.t. replacement rate -1.8(1966) or -0.63(1969) Men, 45-54 1972 Soc. Sec. Sur. of Health and Work Char. merged with benefit and earnings records DI recipiency Expected DI benefits Logit Elast. of recipiency w.r.t. expected benefits = 0.35 All are cross-sectionstudies NLS = National Longitudinal Survey pation for men, age 45-62, significantly falls as the ratio of potential DI benefits to the wage increases. He omits several standard explanatory variables (e.g., local labor market conditions, private pensions, property income) and his measure of expected non-DI transfer income-cash public assistance per poor family in the state of residence-is a weak proxy. While one of the elasticity results is triple the other, both are obtained from the same specification (for different years). Jonathan Leonard (1979) finds a strong relation between expected DI benefits and the probability of receiving DI (and, thus, reducing labor supply). He estimates that increased DI benefits reduce the participation rate of 45-54 year-old men by 1.8 percentage points, or about 40 percent of the observed decline since 1957, the year DI outlays began. While the studies concur that DI benefits are inversely related to labor force participation, the size of the effect has not been reliably established. Improved mea- sures of "true" disability status, expected labor market income, and non-DI transfers are needed, as are more detailed structural models of the relations among these variables and labor force activity. 3. Unemployment Insurance. Unemployment Insurance (UI) benefits typically exceed 50 percent of a worker's previous net earnings; they are revoked when work is resumed. These benefits may induce three potential labor supply effects. The incidence of unemployment may increase if workers who leave jobs (quit) are able to collect UI, or if workers more often enter temporary or cyclically or seasonally sensitive jobs because of UI benefits. Second, the duration of unemployment may be extended due to UI's income and substitution effects.24Finally, 24 Duration of unemployment of new entrants may be reduced if they take jobs more quickly to qualify for UI (Dale Mortenson,1977).John Barronand Otis Gilley (1979) fail to find empirical support for this hypothesis. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0 I t4.4 .0 (M O (=;.D (D O 0 A 0 -140 4--j "m .0 oo > o JQ "O. i w 'o -Q (D ; 56) 0 0 cq cn 0 cn z R >, L.O. Lo0 >4 0 bo A.; 0 04 0 '00 04 -W 04 z 0,5 0 0 bo 0 9 - 0 0 L4.4 -,ad 0 z, 0 L4.4 0 0 0 0 04 --' 04 .8. 04 0 O.0 Cd '5 >4 16.0 0 ol o -0 o 0 40),O' 0.0 0. 14 10, 04 o04 0 04 ol 04 04 04 1.4 0 0 0 04 04 A.; 04 U) bO >4 z 04 1410 0 04 0 r-L 4) 0 -4 z 0 0 >' 04 r. -04 0 4i 0 1.4 0 Z)g 4 0, W>, r, 0 0114 9 4) .1 >4 'o 4i C.) 4) 4i 0 P4 1*01 ci 0 Clu 0 4o 0 0 0 0 0 00 4 00 z This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions A Danziger, Haveman and Plotnick: Transfers and Their Effects "entitlement effects" may induce some persons to enter the work force to qualify for future UI benefits or work more hours to raise the benefits to which they are entitled. This effect may offset the two sources of supply reductions. The recent empirical studies of these three relationshipsappear in TABLE 5.25 Stephen Marston (1979) provides the only results on the effect of UI on the incidence of quits leading to unemployment. Using both micro and aggregate cross-section data, he fails to find a positive relationship between benefit levels and quit rates. The effect of UI on duration of unemployment has been assessed in several recent papers, which are summarized in the table.26Typically, duration is estimated as a function of the weekly UI benefit (or a close variant),the number of weeks of eligibility, demographic variables, and labor market conditions. Kathleen Classen (1979, extending earlier work: Classen, 1977), and Floyd Newton and Harvey Rosen (1979) analyze administrative records of UI claimants. Because such data truncate spells of unemployment at the maximum number of weeks of eligibility, both use Tobit estimates. Significant impacts of the UI parameters on duration are found. Newton and Rosen report that a rise in the replacement rate from 40 to 50 percent increases the expected duration of a spell by 1.8 weeks; Classen'scoefficients suggest an increase of about 1.0 week. 25Thisreview addresses only the labor supply effects of Ul. Demand side factors, such as the incentives for layoffsoffered by imperfect experience rating of the UI tax, emphasized by Feldstein (1976a), may also affect the level of unemployment. Analyses of the relation of Ul to unemployment which do not identify explicit supply side responses, such as in Feldstein (1978a), are not discussed. Alan Gustman (1980) reviews the relation between Ul and unemployment. Hamermesh (1978) reviews and appraises studies that attempt to measure the direct relation between UTand unemployment rates. 26Fields (1977), Hamermesh (1977), and Finis Welch (1977) discuss earlier studies. 991 Neither of these duration studies measures UI's effect on employment, which is more relevant for assessing the efficiency cost of the program (Gary Solon, 1979). Recipients may, in part, extend their duration of recorded joblessness by declaring themselves in the labor force to retain eligibility for benefits. To compare employment and unemployment effects, Solon analyzes a sample of unemployed workers who had exhausted regular UI benefits. For this atypical group, extended UI benefits increase the recorded duration of unemployment. However, two-thirdsof the increase results from a reduction in weeks out of the work force. Only onethird of the increased duration represents actual loss of employment and earnings. Barron and Wesley Mellow (1981) obtain similar results for a national sample of unemployed workers. Receipt of UI raises the probability of remaining unemployed, but about half of this increase reflects a reduced probability of leaving the labor force. These two studies suggest that earlier analyses of duration may have overstated UI's true work disincentive (the net loss of employment time) by confounding it with a labor force participation effect.27 Moffitt and Nicholson (forthcoming, 1982) focus directly on the employment impact of UI. They observe that UI eligibles face a kinked budget constraint because the net wage rate shifts abruptly when eligibility is exhausted. The constraint's shape depends upon both the replacement rate and potential weeks of benefits. They devise a maximum likelihood estimator and show that failing to model properly this constraint leads to serious bias. Their data include direct infor27The results also suggest that Ul not only subsidizes search time (which extends unemployment duration but may improve job matches),but also subsidizes non-labor market activity. As Barron and Mellow note, while such recipients gain income, it is not clear what social purpose is served by this latter subsidy. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 992 Journal of Economic Literature, Vol. XIX (September 1981) mation on replacement rates and weeks of eligibility and extensive control variables. They find that raising the replacement rate by 10 percentage points or extending benefits an extra ten weeks decreases employment by about a week. No duration paper is without problems. Variables to capture administrative behavior are usually omitted, or cannot be entered when all observations are taken from one or two states' UI records. Administrative records used by Classen, and Newton and Rosen offer few control variables.28Information on a recipient's nonearned income or earnings, both of which may influence duration, is usually unavailable (except in Moffittand Nicholson, and Barron and Mellow). Classen and Barron and Mellow omit the unemployment rate. Measures of the replacement rate are biased by the omission of fringe benefits and work expenses from net earnings. Solon's findings are qualified by an unrepresentative sample, lack of data on replacement rates, the need to use a rough proxy for wage rates, and use of OLS for a bounded dependent variable. Analyses restricted to UI claimants (all but Barron and Mellow) may harbor selection bias due to the exclusion of the behavior of eligibles who choose not to collect benefits. Finally, states with economies that have high or strongly cyclical levels of unemployment are likely to legislate more generous UI programs. This possible endogeneity between UI benefits and duration of unemployment has not been addressed and may bias the results. 28 Ronald Ehrenberg and Ronald Oaxaca (1976), using nationalsurvey data for similarresearch,found few control or administrativevariablesto be significant, so the omissions may not be serious. Hamermesh (1978) notes that because higher benefits may shorten filing delays, and some insured persons with short spells may be missed, analyses of Ul records may impart an upward bias to estimates of UT'simpact on compensated duration. Despite the problems, a positive relation between UI and duration of unemployment appears robust. The precise impact varies due to the choice of data, model specification, and estimation method. Of the studies considered, Moffitt and Nicholson avoid most of the problems and probably offer the most reliable estimates.29 Only Hamermesh (1979) investigates entitlement effects. For prime-age women, a group sensitive to work incentives, a 20 percent increase in statutory minimum and maximum weekly Ul benefits is estimated to raise the participation rate by 1.1 percentage points and annual work time by 17 hours. He also finds that the disincentive effect of UI for those who qualify for potential benefits is comparable to that found in the duration studies and almost exactly offsets the entitlement effect. Hamermesh concludes that for prime-age females UI's "net effects on employment and thus presumably production appear to be small" (p. 331). Hamermesh did not analyze adult males. The duration effect probably dominates the entitlement effect for such men, who tend to be regularly in the work force. Hamermesh's data require many simplifying assumptions and imputations that may introduce error. Future research on the labor supply effects of UI might profitably move in several directions. The importance of Ul's incentives to accept temporaryjobs, and the incentives from entitlement effects and work tests merit more attention. Duration analyses should build upon recent econometric advances in handling duration data (e.g. Stephen Nickell, 1979) but with better control variables, especially for labor market conditions, and consideration of 29 NicholasKiefer and George Neumann (1979)analyze reservation wages andjob search and find that Ul benefits have a statisticallysignificantpositive effect on the reservation wage. They do not identify the resulting effect on unemployment duration. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects the likely simultaneity between unemployment and UI parameters.30 4. Aid to Families With Dependent Children. In some states, the Aid to Families with Dependent Children (AFDC) program, augmented by Food Stamps and Medicaid, may provide a larger net income for women with several children than fulltime work at the minimum wage. The program also imposes a high tax rate on earned income. Recent studies of AFDC's labor supply effects are summarized in TABLE 6. (For early related analyses, see Carl Brehm and Thomas Saving, 1964; Hirschel Kasper, 1968; and Leonard Hausman, 1970). All but one find statistically significant negative impacts, but the coefficients vary widely. The first three studies analyze cross-section data from the late 1960s. In Garfinkel and Larry Orr (1974) and Williams (1975), the elasticity of the employment rate of female heads with respect to the guarantee is about -0.7; with respect to the tax rate, -0.5 to -0.7. In terms of actual values, increasing the annual guarantee by $500 or the tax rate by 10 percentage points is predicted to lower employment rates respectively by 2.4 and 1.4 percentage points (Garfinkeland Orr) or 5.8 and 2.1 points (Williams). Williams also finds larger elasticities for participation and hours decisions. Daniel Saks (1975) examines participationrates and finds a guarantee elasticity of -0.94. Raising the guarantee by $500 lowers participation by 1.7 percentage points. While the estimated tax rate coefficient is similar to Williams', the elasticity is much larger due to a very 30Inpreliminarywork, Arnold Katz andJackOchs (1980) use such methods and find, relative to Moffitt and Nicholson, a smaller effect of the replacement rate, but a larger effect of potential durationof benefits. Gustman (1980) offers other suggestions for improving the research. 993 low average participation rate in the sample used. These papers focus only on recipients and neglect the effect of changes in program parameters on eligibility. A decrease in the tax rate increases the work effort of current recipients. However, it also allows more families to qualify for assistance and, thus, has a likely negative effect on their labor supply. Five "second-generation" studies analyze all female heads to capture this eligibility effect. Stanley Masters and Garfinkel (1977) find no consistent labor supply effect of program parameters. Frank Levy (1979) finds that higher guarantees reduce work effort and that higher tax rates increase work. While higher tax rates reduce the work efforts of eligibles, they also reduce the number of people who receive benefits. The additional work from these people more than offsets the reduced work of those who remain on the rolls. Nicholas Barr and Hall (1981, forthcoming) use a measure of welfare "dependence" as a proxy for labor supply and find significant effects consistent with Levy's for both guarantees and tax rates. Hausman (1981) and Moffitt (1980a) assess the effect of AFDC's kinked budget constraint (as well as federal income and payroll taxes) with state-of-the-art econometric methods. Hausman finds that raising the annual guarantee by $1000 reduces work by 120 hours. Moffitt's estimated income effect implies a similar response of 90 hours. While Hausman shows that lowering the AFDC tax rate would induce a large increase in work effort of current recipients (and a corresponding decline by other female heads via the eligibility effect), Moffittestimates a statistically insignificant and economically trivial impact of the tax rate. The large disparity in the estimated tax rate effect-Hausman's coefficient is 37 times larger than Moffitt's-is disturbing. Moffitt modeled the labor supply choice This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Q~~~~~~~~~~~~~~~~~nc 42 4 4 ZA.6 . cn~~~ z 0 . 0 C6 0~~~~~~ 0~~~~~~~~~~ o 4 0 0 0 - 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c U 0 ~~~~~~~~~~~~~~~0 00 0 .9~~~~~~~.~ . 0 ~ ~~~~~~~Th 0~~~~~~~~~~~~~~~~~~~~~~~~O0 0r,~~~~~~~~~~~~~~~~~~~~0 0 0 & 0 0 0~~~~~ 0 ~ ~ II. ~ ~ ~ U 0 04 bO~~~~~~~~~~~~~~~~~~~~~~~ 4-4 04~ ~ 04 ~~ o1-li u,a,~~~~~~~~0 04~~~~~~~~~~~~~~~~ c cq 0 -~~~~toa -~~~~~~ ~~~~~~~1 ~~~~ ~~~~~-~r 0~~~~~~~~~0 This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects jointly with the choice to accept AFDC, included a welfare stigma effect, and estimated effective AFDC guarantees and tax rates (which often differ greatly from statutory values), but ignored the fixed costs of working. Hausman did the reverse. For unknown reasons sample sizes differ substantially-i 19 women in Hausman, 565 in Moffitt-even though both use the same data source. That these and other technical differences can lead to such disparate results on a key coefficient suggest that neither analysiscan be regarded as definitive. A variety of problems are present in the AFDC studies: the use of aggregate data (Garfinkeland Orr), reliance on statutory instead of effective tax rates (Garfinkel and Orr, Levy),31poor or missing measures of unearned non-AFDC income (Garfinkeland Orr, Saks, Barr and Hall), use of ordinaryleast squares with a dichotomous dependent variable (Saks, Williams, Masters and Garfinkel), neglect of administrative and/ or local labor market variables (all but Garfinkel and Orr), and poor measures of the dependent variable (Barr and Hall).32 Only Hausman and Moffitt test for selectivity bias and properly model kinked budget constraints. Food stamps and Medicaid, for which nearly all AFDC recipients are eligible, also create work disincentives but were uniformly ignored. Panel data have not been fully used. 31 Actual benefit r6duction rates in AFDC differ markedly from statutory rates due to administrative procedures. Douglas Bendt (1975), Robert Hutchens (1978), Irene Lurie (1974) and Moffitt (1980a) attempt to estimate the "true" rates. 32With the Barr and Hall definition of dependence, D (the ratio of welfare to welfare plus earnings), zero dependence is consistent with any level of work effort. Moreover, since AFDC benefits = G - T * (earnings) after earnings exceed a small exclusion, ,D > 0, 8D < 0, by definition, even if there is no behavioral response. 995 5. Other Transfer Programs. The studies discussed above cover four majorincome support programswhich account for about 60 percent of total transfer spending. The labor supply impacts of the remaining programs have not been estimated. Veterans' disability compensation, veterans' pensions, workers' compensation, and Supplemental Security Income all induce income effects. The latter three also induce substitution effects. In-kind transfers have parallel effects.33 In addition, the "notch" in Medicaid, by which eligibility for medical benefits is terminated if income passes a predetermined threshold, reduces work. Directly -assessing these effects and those stemming from work tests and administrative practices remains on the research agenda. So, too, does analysis of the labor supply effects of multiple program recipiency. 6. Labor Supply Effects: A Synthesis. It is difficult to draw precise conclusions about the overall labor supply impact of current transfer programs. Most analyses concentrate on marginal responses of transfer recipients to changes in the characteristics of a specific program. Thus, extrapolation to derive an estimate of the program'stotal effect becomes difficult.As noted, the impact of some programs has not been estimated. Nonetheless, we have drawn upon the better studies to offer cautiously, in TABLE 7, a "guesstimate"of how much higher total labor supply during the late 1970s would have been if all income transfer benefits were eliminated. The 33Michael Murray's (1980b) theoretical framework enables him to roughly measure the labor supply effects of subsidized public housing from utility function parameters drawn from the literature. He finds the program reduces work among recipients by about 4 percent. Since about 3 percent of all families are recipients, the aggregate impact would be roughly 0.1 percent. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 996 Journal of Economic Literature, Vol. XIX (September 1981) TABLE 7 REDUCTION IN THE LABOR PROGRAMS SUPPLY AS A PERCENTAGE OF RECIPIENTS OF TOTAL LABOR OF MAJOR SUPPLY INCOME TRANSFER OF ALL WORKERS REDUCTION OF WORK BY TRANSFER AS A PERCENTAGE TOTAL PROGRAM Social Insurance: Old Age and Survivors Insurance Disability Insurance Unemployment Insurance Workers' Compensation and Black Lung Railroad Retirement Veterans' Disability Compensation Medicare Public Assistance: AFDC SSI and Veterans' Pensions Food stamps and Housing Assistance Medicaid Total: HOURS RECIPIENTS WORK OF HOURS OF ALL WORKERS 1.2 1.2 0.3 0.7 * 0.4 * 0.6 0.1 0.3 * 4.8% * Denotes under .05% figures exclude any labor supply responses of nonrecipients to the taxes used to finance these transfers.34 This exercise requires a counterfactual. In this, and later similar discussions, we adopt a zero public transfersbasisfor making the relevant comparisons. We ask what the level of labor supply (and, later, savings or income inequality) would have been in the absence of these programs (and assuming no change in private transfer activities) relative to the level existing with the programs in place. For example, interhousehold private transfers and employer-provided substitutes for social insurance might be larger if public transfers were unavailable and might themselves 34 Hausman(1981) provides estimates of the labor supply effect of federal and state income taxes.Isolating the impact on work of the particularcombination of federal and state taxes used to finance transfers would be extremely difficult and has not been addressed in the literature. lead to labor supply losses (Robert Lampman and Smeeding, 1980). Thus, our counterfactual might overstate the net work impact of the public programs on that account. Consider OASI. Most studies conclude that the program induces earlier retirement and, thus, explains part of the 25 percentage point decline since 1950 in the labor force participationrate of older men (65+). Among the leading analyses, Pellechio finds that the program can account for the entire decline; Clark and Johnson report that eligibility for OASIreduces the chances of participating by 12 percentage points (about half the observed decline), and Gordon and Blinder obtain negligible effects. Since we judge the latter two studies to be the most reliable among those discussed, we attribute as an upper bound one-half of the decline in the older male labor force since 1950 to Social Security's work disincentives. This implies a reduc- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects tion in the total work force of about 1.2 percent.35 In a review of pre-1978 studies of UI's impact on duration of unemployment, Hamermesh (1977) concludes that "the best estimate is that a 10 percentage point increase in the gross replacement rate . leads to an increase in the duration of insured unemployment of about half a week when labor markets are tight," (p. 37). More recent studies, using better methods, find impacts of one week or longer in both tight and slack labor markets. Thus, we raise Hamermesh's figure to one week. Since the typical gross replacement rate for UI is 50 percent, we assume the program extends mean duration by five weeks per recipient. This implies a loss in labor supply to the economy of about 0.3%.36 Entitlement effects partiallyoffset this, but are not reflected in our estimate. The three Disability Insurance studies examine male age groups that include only 40 percent of DI recipients. Leonard finds that DI lowers the participation rate for 45-54 year old men by 1.8 percentage 35 For men 65+, one-half of the change in the participation rate between 1950 and 1977 is .128. The 1977 rate was .201 (representing 1.845 million in the labor force). Hence, without OASI,there would have been (.128/.201) * 1.845 = 1.175 million more men in the work force. Given a total work force in 1977 of 97.4 million, the reduction is 1.175/97.4 = 1.2%. This result is biased upwards because we implicitly assume that men who return to the work force would supply the mean number of hours of all workers, when in fact they are likely to supply less. On the other hand, we do not account for the reduction of hoursworked (due to the OASIearnings test and income effect) by men who reduce their hours of work but remain in the laborforce. Because there is only one estimate of OASI'seffect on the labor supply of older women, we have assumed a zero impact of OASI on the work effort of older women. 36 Let U = number of unemployed persons, INS = ratio of insured unemployed to total unemployed, W = weeks lost per insured unemployed worker and T = total weeks worked by all workers. Then, the percentage loss of work time due to UI is 100 * U * INS W/T. In 1977, U = 6.855 million, INS = 0.4, W= 5 (see text) and T = 4.494 billion, yielding an estimate of a 0.3 percent loss of work time. 997 points in 1975. Because this group accounts for one-ninth of all hours worked, we estimate a reduction in total work effort of about 0.22 percent due to this group. Because men induced to leave the labor market probably had weaker commitments to it in the first place, this is probably an upper bound. Men 45-54 years of age comprise about 18 percent of all DI recipients. If others respond to the same extent, an upper bound to the total labor supply effect of DI would be approximately 1.2 percent.37 Based on Hausman's and Moffitt'sestimates, and data on average guarantees and tax rates in AFDC, we estimate that AFDC reduces work effort of the average recipient by roughly 600 hours per year. For 1977, this would imply a reduction in total hours supplied by all workers of about 0.6 percent.38 37Let M = total number of men 45-54, R = percentage point change in the participation rate of men 45-54 (= change in the number of men 4554 in the labor force divided by M), and LF = size of total labor force. Then the percentage change in the aggregate work force due to DI's impact on men 45-54 is M R/LF. In 1975, M = 11.32 million, R = 1.8 and LF= 92.61 million, which, by the above calculation, yields an estimated reduction in total work effort of .22 percent. Since men 45-54 equal about 18 percent of all DI recipients, 0.22 is multiplied by 1/.18 = 5.556 to find DI's effect over all groupsof recipients. Thisproductyields our estimate of 1.2 percent. 38Let H = hours lost per AFDC recipient in the labor force, REC = ratio of recipients to all female heads, F = number of female heads, and T = total hours worked by all workers.Then, the percentage loss of labor supply due to AFDC is 100 H REC * F/T. In 1977, REC was about .3, F = 5.414 million and T = 164.2 billion. To determine H, we observed that in 1975, the mean annual guarantee was about $2400. The mean wage of AFDC recipients was about $3. If the tax rate were 67% (the statutory rate, which is an upper bound), the net wage is reduced $2. Using these together with Hausman's income and wage coefficients for his 1975 sample suggests that a typical recipient worked 1024 hours less because of AFDC. Because Moffitt'stax rate effect is so small, his coefficients suggest a change of only 103 hours. We roughly split the difference and assume a change of 600 hours per recipient (implying an average reduc- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 998 Journal of Economic Literature, Vol. XIX (September 1981) Finally, we offer some speculations on the work effects of transfer programs for which no statisticalanalyses exist. Railroad Retirement is similar to OASIin operation and benefits. Because it is one-twentieth of OASI's cost, we assume its supply impact is one-twentieth of OASI's.Workers' Compensation and Black Lung together cost about the same as Ul or DI. They provide both short and long term benefits. Thus, their joint impact probably lies between that of Ul, a short term program, and DI, which offers long term benefits. We assume the joint effect is 0.7 percent. Veterans' disability compensation is half the size of DI. Its supply impact is probably much less, because persons can receive benefits without withdrawing from the labor market and the basis for compensation may be unrelated to their current job requirements. This leads us to a rough judgment that this program cuts work hours by 0.4 percent. The supply effect of Medicare is likely to be nil because it has no earnings test and its marginal income effect, beyond that already due to OASI is probably small. Among the income tested programs,SSI and veterans' pensions are somewhat similar to OASI and DI in terms of recipient population. Together, they are 11 percent of Social Security's size. Thus, their joint supply effect is assumed to be on the same order. Food stamps and housing assistance reach a wider population than AFDC. Many participantsare two-parent families whose labor supply disincentives are not as large as those of female household heads. Since many AFDC recipients also receive these in-kind benefits, some of tion of 180 hours per female household head). Because real guarantees and wage rates were roughly constant between 1975 and 1977, the same loss is assumed to exist in 1977. The resulting percentage loss of labor supply of 0.6% may be an overestimate since it assumes that recipients not currently in the work force would also work 600 more hours even though their response is likely to be smaller than that of women currently in the labor force. their work effects may have been attributed to AFDC by the AFDC studies. As we noted, these studies do not address multiple program recipiency. Hence, the additional net effect of these two programs is assumed to be about half that of AFDC's, or 0.3 percent. And, because all AFDC recipients are also entitled to Medicaid, part of its effect may already have been counted. A portion of Medicaid assists persons in nursing homes or with severe health problems, whose labor supply can plausibly be assumed invariant with respect to most economic incentives. Thus, the supply effect of this program not already captured in the AFDC results is assumed to be negligible. The sum of our guesstimates for the four programs for which an empirical literature exists suggests that because of labor supply reductions by transfer recipients, total work effort in the economy was 3.3 percent less than it would have been. Adding the reductions due to the other programs, which are even more impressionistic, gives a total reduction of 4.8 percent. (Note that the percentage reduction in work per transfer recipient implied by this estimate is substantially larger than 4.8 percent, while the effect of marginal cutbacks in transfer benefits is substantially smaller.) In an economy with involuntary unemployment, however, not all of this supply would be employed. If the unemployment rate were 7.0 percent, and if the increased labor supply of transfer recipients would find employment at a rate equal to that of other workers, the net loss of employment time would be about 4.5 percent.39Because those receiving transfer benefits tend to have below average wage rates, the loss of total earn39Lampman (1978) "guesstimates"the labor supply effects of all social welfare spending including public education and the taxes required to finance them, to be about 7 percent. Our estimate differs because we examine only the transfersshown in TABLE 1 and we do not examine tax effects. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects ings is probably about 3.5 percent. However, in view of the wide variation of the estimates found in the literature, the true losses of labor supply, work time, and earnings may lie within a wide band around our estimates. B. Effects on Private Savings Numerous recent empirical studies have sought to measure the impact of transfers on private saving. Because private saving is, in part, undertaken for life cycle purposes and because the Social Security program provides the largest amount of transfers, the analysis has focused on this program's effect.40 Several mechanisms by which Social Security affects savings were outlined earlier. First, pay-as-you-go Social Security benefits may substitute for private savings. Second, Social Security tends to induce early retirement, increasing the need for savings. Third, because the program shifts income from children (taxpayers)to parents (beneficiaries),parents may increase their savings to maintain a target level of bequests and offset the taxes their children pay. Finally, if lifetime marginal propensities to consume vary inversely with lifetime income, the equalizing effects of Social Security within generations may reduce private saving. The major studies examining the net savings effect of Social Security are described in TABLE 8.41 Time series work is discussed first, followed by microdata and internationalcross-sectionresults. Unless otherwise noted, all specifications are derived from some variant of the life cycle model. Feldstein's 1974 time series study suggests that each dollar of gross Social Secu40 Throughout this section, "Social Security" means only the retirement income portion of the SocialSecurityprogram,not the disabilityand survivors' components. 41See also Bulent Gultekin and Dennis Logue (1979), Cagan (1965), and Katona (1965). 999 rity wealth (the present value of expected future benefits) increases consumption (and, by inference, depresses savings) by over 2 cents. This effect is statistically significant only if the unemployment rate variable is omitted and only if the entire 1929-71 period (omitting the war years 1941-46) is analyzed. For the 1947-71 period, the effect of Social Security, though positive, is not statistically significant. These results imply that Social Security could have reduced private saving and investment by about 38 percent, with a corresponding decrease in GNP of about 15 percent. The implications of Feldstein's results for investment and economic growth are criticized by Michael Darby (1979), Kotlikoff (1979a), and Balcer (1981). According to Darby, Feldstein's use of gross (rather than the preferred net) Social Security wealth variable overstates the savings reduction by 15 percent. Second, Feldstein's addition of the savings effect of Social Security taxes substantially overstates savings reductions attributable to the transfer. When those adjustments are made, the 38 percent figure falls to 26 percent. Kotlikoffnotes that Feldstein derives partial equilibrium estimates, and, hence, does not allow the implied decrease in saving to affect the rate of return on capital. He estimates that allowing this feedback reduces the savings decline to 20 percent. Finally, Balcer (1981) emphasizes that Feldstein's estimates substantially exceed their true size because he fails to account for the wealth effect on consumption of the Social Security induced reduction in private wealth, and hence measures only the direct consumption impact of Social Security. Feldstein's study triggered a spate of other time series analyses (AliciaMunnell, 1974; Barro, 1978; Feldstein, 1978b; Darby, 1979; and Boskin and Marc Robinson, 1980). These studies differ in the definition of the Social Security wealth variable This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions IL) > (L) -Z: 04 bO 0 (L) bO + (L) I C", >% > rA bO 04 1:1 %i4 (L) --4 CJ >-. (L) CO) -4 .5 A 9 4-1 0 .5 g 0 95 > 4-i 14 ol Cd bO 0 0 0 Cd c4) 78 0 0 "a bO 0 U5 CA cd .4 cn cn i34 cn cn 04 Cd 0 0 0 (L) > Q a - ra 'gb 0 bO U, 0 Iri cd 0, -0 .4cn cn ") .5 i34 0 0 C4-4 0 bO 4-1 .2 C40 4-J C) 1044-i C40) cn C) U5 cd Co, C) bO i34 04 C) bO bO up C) 14 u, 0 cn cn 0 cn cn cn Cn cn bO cn Cd cn 0 cn cn 0 0 0 cn C.) cn > cn :cn Z cn cn 0 0 0 0 U 0 w .1.4 4 0 cd 0 0 i34 4 U, bO 0 t: Cd 0 0 U, cd 4-i 0 in, U, u 0 0 0 4-i U, U, C.) 0 CJ U, Cp Cd > 95 95 0 0 0 0 0 0 0 0 0 cn 0 -ri 0 C.) C.) 0 4 Cd 0 0 0 Cd > U, 0 U, -9 0 ..0 > 0 Co, U, 0 95 U, Cd Cd + C) IT v to tITtC) (M t IT C) C) o t- C) V) v CA v V) v U, U, >-. 00 0 U, U, >4 0 CD >1 U, 0 00 0 4 cn Cd Cd C) ll U, oo 0 C) 00 (m U, 0 U, U, U, oo U, 04 x C) > 0 z cn C) 4-i C.) U, Cd Cd '9 > cn cn cn cn 4-i Cn bO cd cn 0 U, 4-i Cl) cn cn > 0 L4-4 U, 4 0 0 0 1.0 > cn cn 0 > cn 0 rZ4 cn cn 0 4-i U, 0 cn co bO 0 0.4 U cn cn C4-1 C) > r--l 4G& a) C4 z > a.) C) Cd 0 biD Cd 0 bO c) cd Cd U, cd 0 311 0 i34 U, cn cn i > Z bO bO 0 z (L) -Z: ci 04 0 o C.)Cd -w Cd 0 Cd C i34 0 C-4 bO C43 > Cd C) cd 0 'm 4-i Cd -W 04 Cd Cd I > 0 00 U, 0 cn This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0 Iri 0 cn 0 4 95 C,, 04 Co, 04 X 'S cn cn C43 C43 >, 4 t:: 7RI 4-i a) 0 4-1 Cd > .4 o 4G& 0 0 0 4 0 0 -.., 0 "o) a) C.) r0 >-.cn 0 C) Cd Cl) C40. 0 '54 4GO 4 cd Cd a Cd a) cn cn 10 0 biD 0 a) a)'ra C) bO C4 C-P 09. C.) 0 0 14 cn ." ;a Cn '75' icd (M 0 Cd 0 Cd bO EFp C 0 0 9 -a a) ,r 0 0 0 a) 4- in >, 0 'm -z m bO ..4 w 0 6) C40. > bO 0 Cd C43 : cn C4-4 0 95 0 a) > 14 Cl) C4-4 1 a) a) cd CJ 1) 0 L4-4 >% 0 0 cn cn 1.0 r--l 0 0 I "-4 Cd 'n cn 0 cn 4G& 4G& 4-i cn o 0 Cd U5 4-i 0 -w > .9 0 o a cn 0 cn cn Cd U5 C) Jj i34 C4 cn cn -4 o -.4 4-i 04 0 10 c 0 0.) o bO Cd U) L C4 ci C4 0 0 C) a Cd L4.4 4 c Cl 14 cn A o cn cn cn 0 .5 cn 0 cn 0 U5 cn cn 0 4-J C) 4 a) -w C)a) 4)11) 0C)-w 10 0 ll r. '10) 0.) 0C) C4 > C) C4 0 Cd 0 bO Cd o cd -4 C.) Cd 0 0 m 0 14 9 04 0 -4 cn C) bO Cd 14 > Cd - 0 0 0 0 .0 o 1-8 cn :3 Cd 0 U, U, 0 U, > U, U, cn cn U, Cd 0 Cd > Cd bO .5 Cd C) .5 4-J Cd U, U, bO bo U5 0 14 4 0 .,.I Cd Cd Cd 04 o rj > Cd Cd Cd 4 bO cn cn > > A4 Cd 4 0 U, cd > 0 0 Cd > .14 4 0 C) U, 0 0 0 C.) Cd > U, 4 .14 4 U, 0 0 u C4-4 cd 0 0 4 C) 0 o U, a) 0 C) 0 C.) C) U, 4G& 0 m 0U,00 co 0 "-I A -W U, 0 0 0 in U5 0 0 Id 4 Cd > 0 co cq CA cn > 0 o 0 7RI cn 0 Co, 7RI ." >,.o , 0 'o U, 0 Co, C-0, 0 o 0 7il >-. L4-1 0 U, 0 00 14 U, t- -14 ..4 t- 0 .5 Cd 00 oo t> cd 0 00 m 0 Cd cd 0 -z 14 0 o 0 0 cn 0 Ij 6 6 t4 14 0 Cd cn m r--l > 0 0 '-m Cdcn 0 bO Cd 4.4 0 Cd 0 bO 0 10 cn r 0 This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions M bO o bO Cd cn II cn 1002 Journal of Economic Literature, Vol. XIX (September 1981) (some use gross Social Security wealth and others Social Security wealth net of taxes), the specification (some include variables such as the unemployment rate, the government budget surplus, and disposable or permanent income, and others do not), the time period observed, and the definition of the dependent variable. The results vary substantially: they usually imply a negative impact of Social Security on private savings, but in several cases the coefficients are not statistically significant. Only Feldstein (1978b) and Boskin and Robinson (1980) find statistically significant results which suggest a 25-30 percent reduction in private savings due to Social Security. Munnell (1974), using a specification which incorporates the unemployment rate, finds a negative, but not statistically significant,savings effect of Social Security wealth (both gross and net). She does find evidence, however, that Social Security shifts asset holdings away from those designed to support retirement to those held for other purposes. In the Barro-Feldstein debate (1978b), the same basic data as originally used by Feldstein were employed. In Barro'sconsumption function, which includes both the unemployment rate and the government budget surplus, Social Security wealth has a positive, but statisticallyinsignificant, impact on consumption in all periods analyzed. Feldstein challenges Barro'stheoretical specification and extends the period of analysis.While he incorporates Barro'sunemployment variable, he multiplies it by disposable income (assuming that the unemployment rate changes the marginal rather than the average consumption propensity). The budget surplus variable is omitted, on grounds that it varies in response to the tax revenue generated by changes in income, and hence is endogenous. Again, Feldstein finds a statistically significant and sizable effect of Social Security on consumption. While these studies all adopt a current income-based-consumptionfunction, Darby (1979) develops an income variableconsistent with a permanent income-consumption relationship. As with the other studies, except Feldstein's, the effect of Social Security on savings in the 1947-74 period is not significant, though negative. Boskin and Robinson (1980) employ a real consumption (including service flows of durables and excluding durable purchases) instead of a current consumption expenditure variable. The measurement of other relevant variables is also improved. Using their preferred concept of net Social Security wealth (grossSocial Security wealth less the present value of Social Security tax liability), their estimate is about three-fourths of Feldstein's original one. A large number of other specifications and Social Security variables are also tested. Many of the coefficients on the Social Security variables are not significant. In a recent development, Dean Leimer and Selig Lesnoy (1980) discovered an error in the basic Social Security wealth variable constructed by Feldstein and used in most of the pre-1980 time series studies. A replication of Feldstein's analysis with the corrected variable yields no significant negative savings effect if the gross wealth measure is used and a significantand positive effect if the net wealth variable is used. Leimer-Lesnoy also develop alternative Social Security wealth variables with their improved data. Nonetheless, the coefficients are generally insignificant and often of the wrong sign. In response, Feldstein (1980) reestimated the original consumer expenditure equation using the corrected Social Security variable, and estimates a series of other equations, including some with a revised Social Security variable (to take account of the 20 percent benefit increase This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects in 1972 legislation), and additional years of data. These supplemental estimates yield mixed results. The insignificant estimates of Leimer-Lesnoy are verified. Those based on the corrected and revised Social Security variables and additional years of data yield a marginally significant and somewhat smaller coefficient on the Social Security wealth variable than in Feldstein's earlier estimates. In the most recent development in this controversy, Leimer-Lesnoy (1981) note that, while Feldstein incorporates the 1972 benefit increase into his Social Security wealth variable, he neglects the indexing provision of that legislation and the effects on benefit perceptions of all other legislative changes (some of which had benefit increase provisions twice the size of the 1972 change). Leimer-Lesnoy replicate Feldstein's 1980 results, fit other models which presume that consumers adjust to all legislative changes including indexing (and not just the 1972 benefit increase), extend the data series to 1976, test the sensitivity of results to the period of analysis, and employ Social Security wealth variablesbased on alternative benefit perceptions. Their results suggest no significantpositive effect of Social Security wealth on consumption; indeed, a negative impact is indicated for the post-war period. Time series models do not provide a strong base for deriving an estimate of the savings effects of Social Security. Limited observations restrict the number of independent variables that may be employed, and do so in a context in which available aggregate time series variables diverge far from the true price and real value measures suggested by economic theory. Numerous factors that might be expected to influence savings (e.g., employer-contributed private pensions, demographic factors, relative prices of consumption and savings) are not included in the regres- 1003 sions. Specifications are, as a result, crude and ad hoc; biases and their extent are unknown, and judgment and taste play a major role. Feldstein, in his 1980 correction, perhaps states it best: "[I]t is worth emphasizing that the difficulties of measuring expected social security benefits and of separating the effect of social security from the effects of other variables that influence saving will always leave a substantial margin of uncertainty about the precise magnitude of social security's effects" (p. 14). Given the reliance of most studies on the erroneous Social Security wealth variable, the ad hoc adjustments underlying Feldstein's most recent and highly variable estimates, and the careful critique and contrary results reported in Leimer-Lesnoy, we judge that there is little robust time-series evidence of a significant negative relationship between Social Security and private savings. If Social Security has a depressing effect on private savings, analyses of micro-data on households and persons might be expected to disclose this linkage at least as well as time series data.42 The second group of studies in TABLE 8 takes this approach. The first, by Kotlikoff (1979b), analyzes male household heads aged 4559. Two components of each household's net Social Security wealth are distinguished. One is the present value of past Social Security taxes. If these taxes are considered as forced savings earning a market return, they will tend to substitute for private savings. Kotlikoff finds a 70 cent reduction in asset levels per dollar of present value of accumulated taxes. This result, however, says little about the full effect of Social Security on private savings; it does suggest that forced savings 42 However, as Blinder et al. (1981) note, even if the program has no macro effect, those with unusually high (low) benefits may save less (more). Thus, a significantmicro relationshipdoes not assurea similar savings displacement at the macro level. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1004 Journal of Economic Literature, Vol. XIX (September 1981) for a public retirement program is a par- wealth conditional on gross Social Security benefits. No significant alteration of these tial substitute for private savings.43 The second component is a "lifetime profiles in response to gross benefits is found, in contradiction to Feldstein. Bewealth increment" (the present value of cause bequests do not vary significantly expected future benefits less the present value of past and expected future Social with the lifetime wealth increment due to Social Security, Barro's hypothesis is Security taxes) attributable to Social Secualso rejected. rity. This increment is positive and large Another advance in model specification for most older households (Burkhauser and variable construction is present in the and Warlick, 1982, forthcoming). Any recross-section work of Blinder et al. (1981). duction in private savings that it induces would imply a decline in real capital for- Their life cycle model assumes certainty about length of life and number of chilmation. Kotlikoff finds that asset accudren at each point in time, and a tastemulation is not adversely affected by this for-bequests function that depends on the variable-the coefficient is positive, but number of children and the ability to hold statistically insignificant. A somewhat different approach is taken fungible capital. The model specifies total wealth holdings as a function of Social Seby Feldstein and Pellechio (1979), who estimate the effect of Social Security on pricurity wealth, lifetime earnings, and a few other variables. Their Social Security vate wealth accumulation just prior to retirement. They conclude that net Social wealth variable (also imputed, but based Security wealth is close to a perfect substi- on extensive earnings history information) is not statistically significant.The point estute for private savings. These results are open to question, however. Unlike Kotli- timate suggests that an additional dollar of Social Security wealth displaces other koff, they do not distinguish past and fuforms of wealth by $.54. This result seems ture flows contributing to the net wealth to be contradicted by the associated increment, the sample is very small (138 (though again nonsignificant) estimate observations), and (as with Kotlikoff) inthat private pension wealth induces a 33 adequate earnings histories and other information on the determinants of Social percent increment to other wealth holdSecurity wealth required numerous as- ings. A related life cycle estimation, by Diasumptions to impute a wealth variable. David and Menchik (1980) employ data mond and Hausman (1980), exploits panel that include lifetime income from state data observations on wealth in four sepaincome tax records, lifetime contributions rate years. They construct an estimated to Social Security, and bequests from prosavings propensity that is a function of bate records on a sample of Wisconsin permanent income, pension and Social Semales born between 1890 and 1900 to test curity wealth (based on extensive earnings both the Feldstein and Barro hypotheses. history data),and socio-demographiccharThe former is tested by comparing the acteristics. From this model, they estimate age-wealth profiles of the observations the pattern of wealth accumulation up to both without Social Security and with age 65, decumulation after age 65, and expected wealth at age 65. They find that pre-retirement private wealth accumula43 Kotlikoffalso finds that, given the choice to retion is reduced by about $7 for each dollar tire earlier, people accumulate more assets. To the extent that SocialSecurity encourages earlier retireincrement to expected annual Social Sement, the additionalprivate savings so induced tend curity benefits. While this relationship to offset whatever reduction in savings is due to Sois statistically significant, it implies far cial Security taxes. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects 1005 less than a dollar-for-dollarsavings offset. cial Security wealth is not settled, factors Finally, three studies employ cross-na- affecting savings for non-life-cycle reasons are often not measured, general equilibtional data. Feldstein (1977) explains the savings rate for a sample of 15 countries rium type feedback (e.g., through changes for the period from 1954-60 by socio-de- in the return on capital, as in Kotlikoff) mographic variables, income, growth, and or bequest responses (e.g., human capital the replacement rate of Social Security. investment effects, as in Drazen) have reA 25 percent increase in the replacement ceived little consideration, and data problems in measuring real saving as opposed rate implies a 17 percent reduction in the to financial flows are serious. The next average private-savings rate. Barro and Glenn MacDonald (1979) use a similar generation of research should focus on seframework, a larger data base, and a vari- curing improved Social Security wealth and real savings measures, introducing ety of specifications,but find no consistent or statistically significant effect of Social variables affecting savings for other than Security on private saving." The final life cycle purposes, and distinguishing the channels by which the wealth and labor study, by Feldstein (1980b), employs a ''new retiree replacement rate" variable, supply effects of Social Security have an recently developed for 12 countries by the impact on private savings. U.S. Social Security Administration (Leif The studies reviewed here focus on the savings effect of Social Security alone, not Haanes-Olsen, 1978). With retirement age that of all transfer programs. Because assumed to be unaffected by Social Secunone of the other social insurance prority, the replacement ratio is negatively related to the savings variable and mar- grams, except Disability Insurance, is run ginally passes the significance test. Howon a pay-as-you-go basis or is closely tied to the life cycle, there is little reason to ever, when retirement age was treated as endogenous-a key feature of the exanticipate life cycle savings impacts from tended life cycle model-a negative but them. There is, of course, the potential negative savings effect of equalizing transinsignificant relationship between the replacement rate and savings was found. fers if bequests are a luxurygood (Blinder, 1975) or if income-replacing transfers, Moreover,when the average replacement rate used in Feldstein (1977) was included such as unemployment insurance, reduce instead, it was insignificant. savings for precautionary motives. While The cross-section results, then, yield unmeasured in the empirical literature, there is some presumption that savings much the same mixed picture as the time series results. Some evidence of a depress- are smaller because of the safety net proing effect of Social Security on private sav- vided by these programs. On the other hand, the models implicitly assume a fully ing is present, although the coefficientsare often insignificant. Serious model specifi- employed economy and neglect the possication uncertainties exist (e.g., the relative bility that transfers raise current aggreroles of the wealth effect and the induced gate consumption and investment deretirement effect are often not distin- mand and, hence, income and aggregate guished), the appropriate measure of So- savings. Given the small and often statistically insignificant estimates of the impact 44 The sources of the divergence in parameterestiof Social Security on savings, the signifimates between Feldstein (1977) and Barroand Maccant deviations from full employment in Donald (1979) are examined in Charles Horioka (1980). Some responsibility is attributed to (1) the the period under study, the lack of evispecification;(2) the sample; (3) the variable definidence on potential reductions in savings tions; (4) the sources, and (5) the time period, but for precautionary motives, the focus of the the relative importance of each is unclear. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1006 Journal of Economic Literature, Vol. XIX (September 1981) studies on financial as opposed to real savings, and the relative constancy of the income share of the top quintile (David and Menchik, 1980), the case that transfer programs have substantially reduced aggregate savings and, consequently, the growth rate of GNP is not strong. At the same time, since most of the studies do find negative signs on the Social Security variables, there is probably a small, as yet imprecisely measured, negative effect. Given the wide variation in the estimates of savings impacts, we venture the tentative conclusion that income transfer programs have depressed annual private savings by 0-20% relative to their value without these programs, with the most likely estimate lying near the lower end of this range.45This conclusion reflects the general thrust of the microeconomic theory of savings determinants and rests most heavily on the results of the micro-data studies of Blinder et al. (who in our view have specified the determinants of saving most completely) and to a lesser extent Kotlikoff. The results of Feldstein and Pellechio seem less convincing for reasons mentioned above. The time series studies are seriously flawed by reliance on the Social Security wealth variables now known to have been erroneous and by the inherent instability of annual time series estimates. And, as Feldstein (1980b) has stated with respect to his own preferred international cross-section estimates, the "sample is too small and the problems of measurement too great to warrant reliance on the specific parameter estimates" (1980, p. 238). C. Redistributive Effects In this section, those studies that measure the effects of the transfer system on the distribution of income will be described and appraised. We focus on redis45Thisconclusion is consistent with that of Darby (1979), who attributes a 0-10 percent negative savings effect to Social Security alone. tributive studies that utilize three summary indicators-the incidence of income poverty, the share of aggregate income received by the bottom quintile of household units, and the Gini coefficient.46Studies of redistributive effects employ a methodology that is substantiallydifferent from those analyzing labor supply and savings responses. The latter apply multiple regression techniques to cross-section or time series data to estimate behavioral responses to policy-induced price, income, or wealth changes. The redistributive studies rely on more straightforward calculations on aggregate or micro-data bases. As a result, the range of estimated redistributive effects is substantially narrower than the labor supply and savings effects reviewed above. A basic conceptual and empirical problem in measuring redistributive effects concerns the definition of the counterfactual-what would the distribution of income be in the absence of existing transfer programs? A fully accurate definition of the counterfactual would recognize the full set of general equilibrium changes in relative prices and incomes that would occur if transfer programs were removed. Labor supply and savings responses and changes in marital status and living arrangements are among the obvious behavioral responses that should be modeled. However, this ideal has not been implemented. Most studies measure redistribution as the simple difference between a household's final or posttransfer income and its income excluding transfers (pretransfer income).47To the extent that the 46 The Atkinson index (Anthony Atkinson, 1970), Theil's entropy index (Henri Theil, 1972) and other summarymeasuresof inequalityare sometimes used. However, the choice of summarymeasure has little impact on the general evaluation of redistributive effects. 47JeanBehrens and Smolensky(1973),JacobMeerman (1978) and Morgan Reynolds and Smolensky (1977) have addressed the definition of the counterfactualand conclude that "true"pre-transferincome cannot be measured. Irwin Gillespie(1980a)specifies This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects declines in labor supply resulting from transfers are concentrated among those with low incomes, the pretransfer incomes of those at the bottom of the distribution are disproportionately understated. Or if transfers induce some lowincome households to break into smaller units, there will be more such households than in the absence of transfers.Thus, true pretransfer income is likely to be less unequally distributed than measured pretransfer income. Pre-post comparisons, therefore, are likely to provide upperbound estimates of redistributive effects. Most redistributive studies have used either micro-data (available in public use tapes since 1965) or published data on household income (available since 1947) from the Current Population Survey (CPS) (U.S. Bureau of the Census, 1978), although micro-data from the 1960 Census and some earlier CPS surveys are available. Some have employed longitudinal micro-data from the Panel Study of Income Dynamics (PSID), available since 1967 (Morgan et al., 1974). Both the CPS and PSID present problems relating to the income concept, the accounting period, and the income unit.48 Income is best defined as real consumption plus changes in net worth during some accounting period. However, the CPS includes only money income flows during a calendar year, and omits tax liabilities, in-kind receipts from government programs or employer benefit plans, selfproduced goods and services, and capital value changes. Many sources of incomeespecially transfers received by the poor and property income received by the the conditionsunder which government policies will be neutral and hence the observed pre-transferdistribution will coincide with the "true" one. Robert Dalrymple (1980) argues that such neutrality is, in fact, unlikely and cites evidence from Golladayand Haveman (1977) that casts doubt on such neutrality. 48 Taussigand Danziger (1976), Blinder (1980),and Menchik (1981, forthcoming) also review the issues involved in measuring inequality. 1007 rich-are underreported. For many transfer programs-especially those that are age-related-a multi-year or even lifetime accounting period is relevant for analyzing redistribution. The CPS accounting period, the calendar year, is appropriate only if the effect of transfers on current income or poverty status is being measured. The PSID does allow analysis of longer accounting periods since the same sample of families has been followed over time, but few studies of long-run redistribution have been done (Richard Coe, 1978). Finally, the income units in the CPS are the family (individuals in a dwelling unit related by blood, marriage, or adoption), unrelated individuals (those living alone or in a dwelling unit in which they are not related to the other individuals), or households, defined as the sum of the two. Because income-sharing units may span several households and inter-household transfers are generally not recorded, estimates of redistributive effects of public programs may be biased.49 Other data sources (e.g., Consumer Expenditure Survey) share these problems as well. While these measurement problems plague analyses of redistributive impacts during any year, they are compounded when redistributive trends are estimated. For example, over time, in-kind transfers and fringe benefits have grown in relative magnitude, lifetime income profiles have become more peaked, and household composition and patterns of income sharing have changed radically.50 Each of 49Consider, for example, a married couple supporting a non-earning child attending college in another city. Because the CPS considers the student to be a poor unrelated individual with no reported income, a redistributive study would conclude that current transfer programs were not reaching this poor person. 50Morton Paglin (1975) suggests a new measure of inequality to address the shortcomings of annual data. He attempts to remove from the Lorenz-Gini coefficient "changes in the age-income profile and in the age composition of the population" (p. 605). Danziger, Haveman, and Smolensky(1977), Blinder (1980), and Menchik (1981, forthcoming) point out This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1008 Journal of Economic Literature, Vol. XIX (September 1981) these changes means that the difference between the "true" and "measured" redistributive effects of transfers has grown. Given these caveats, we turn to studies of the effect of transfers on income poverty. The most well-known studies of poverty are published annually by the U.S. Census Bureau (1978) and compare Census income (which includes cash but not in-kind transfers) to the poverty line.51 While the series shows a decline in posttransfer poverty from 22.4 percent in 1959 to 11.6 percent in 1979, the effect of transfers cannot be ascertained from published data. TABLE 9 lists the recent studies that measure the antipoverty effects of cash or in-kind transfers. All, except Paglin (1980), use microdata and subtract cash social insurance and cash public assistance transfers from posttransfer income to measure antipoverty effects.52 Smeeding (1975) and G. William Hoagland (1980) also adjust the Census data for underreporting, and impute values for federal personal income and payroll taxes, and for in-kind transfers. Hence, their analyses provide the most comprehensive measures of the incidence of poverty after all major programs. None of the studies, however, account for any behavioral responses that would cause true pretransfer that the Paglin-Gini coefficient does not measure what it purportsto measure. Gillespie (1980b) criticizes the normativeimplicationsof Paglin'smeasure. In any case, Paglin analyzes the trend in posttransfer income inequality and not the redistributive effects of transfers.The Paglin-Giniis insensitive to transfers across cohorts (e.g., transfersfrom the young to the old), so it cannot be used to measure the redistributive effects of programs such as Social Security. 51 The Federal government annually publishes poverty lines for families of varioussizes, structures, and locations.These lines posit minimallyacceptable income levels and are adjusted for increases in the consumer price index. For 1979, the poverty line for a non-farm family of four was $7412; for 1981, it is estimated to be about $9300. 52 Early studies of the antipoverty effects of transfers include Lampman (1966), Benjamin Okner (1972), and Plotnick and Felicity Skidmore (1975). income to differ from measured pretransfer income.53 Each study in TABLE 9 finds that transfers significantly reduce poverty and that this redistributive effect has grown over time as the amount of transfers has increased. The results of the studies can be summarized as follows. First, consider cash transfers (Danziger and Plotnick, 1981). In 1965, about 21 percent of all persons would have been poor in the absence of transfers. Cash transfers reduced poverty to 15.6 percent of the population, a reduction of 27 percent. By 1978, cash transfers reduced poverty from 20.2 to 11.4 percent, a decline of 44 percent. Cash transfers have a much larger antipoverty effect for the aged than for the nonaged (75 percent vs. 26 percent). Next, consider in-kind transfers (Smeeding, 1975; Hoagland, 1980; Paglin, 1980). When they are added to Census income (post-cash transfer), poverty falls to about 4 percent in 1980 (Hoagland).54 The variance in the estimates in the table result from differences in the unit of analysis (households or persons), techniques for assigning in-kind transfers to poor households, the specific transfers included, assumptions for valuing in-kind 53 Plotnick (1980) attempts to move from the standard pre-post comparison to one that adjusts pretransfer income for transfer-induced labor supply and earnings effects. He finds that pretransfer poverty, and hence, the antipoverty effects of transfers, would be lower in the absence of transfers. 54 Ofthe total reduction due to transfers,cash social insurance accounts for about 50 percent, cash public assistance, about 10 percent, and in-kind transfers,about 40 percent. While public assistance programstarget a greater percentage of their benefits on the pretransfer poor than do social insurance programs, the much larger volume of social insurance transfersmakes them more effective in reducing poverty. Smeeding (1981) points out that the 4 percent figure is based on a 1978 data base which is projected to 1980. The macroeconomic assumptionsused for this projection were more favorable than the 1980 reality. He suggests that an adjustmentto actualconditions would yield an estimate of poverty after the receipt of cash and in-kind transfers of about 6.1 percent. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions zz n o Htem>t>F ~ g~ l S l ~ ~ XFt0qO ~ -CO'z K -1 tHmett V c'i I- - -0 OXs0 1 - mm ) 00 0 l c- in _-qin CO)I" cOCO .P --4 -- r- - t~~~~~~~I Ns I r- -- "- .-q- I4 z v ?~~~~~~~~~t U 0 "0t ms CC)r0 OC) m 1-0 mcqCt t-t mO CC o o~~~~~~~~~C a; 0M cOFOF UC/c= m 4o 34 ^4 ^4C;-;t , i-;c r-Ct t )OD c%% 00CC) 0c N) 4 ic 0~~~~~~ z - z - EvQ~~~~~~~~~m z mv m:vvvm r 0 v s?<0 2U U OCb O mx gm(M M Z Cb m 0 r mmtctIo U >~~~~~~~~~~~~~~~~~~ 0 > o i 00- 00 E o4 So 4i~~ 0 cO v 0U 100) 1>,-4 C) t- o ? 0 C.) 0 -n cn - OC ^ O4=SO e te 0 $4 Xm ~~~~~~~~~~~~ r 0 u, . ?=vz< CC)1~~~~~~~~00)00 0)0 0 3W; "00) , 5C'C 0)~~~~~~ 0)0 0~~~~~~~~$-.4 ~ . $-4 so >, 01 0~~~~~~~ 00 U' 0 u, 0 Cd "0 , c 0 00"~ P4* ~, 00 4.0 "'CC) ~ ~ 0) ~ ~ 0)~ ~ ~ ~ ~~~~~~~~~~~~~~~ 0 U' - 1 .0 C & ~~~~~~~~~~~~--0~~U ii? "~~~~~~~~~ C) 00 0) 14. 0 C-tP , n s.00~~~~~CdOC $_4 bO HM ~~~~~~~~~~~~~~0~~~1 Cd ~-4 ~u Cd u0 C - 0 "0 a0~H Cd 0 C > $4$4C.) C#) This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions P1I "0 () c C 1010 Journal of Economic Literature. Vol. XIX (Sevtember 1981) transfers, corrections for underreporting or misreporting, the data base used, or the year of the analysis. For example, Smeeding includes Food Stamps, Medicare, Medicaid, and Public Housing Benefits. Hoagland adds Child Nutrition Benefits to these. While both subtract Social Security and Federal income taxes, Paglin does not. Smeeding values in-kind transfersat their cash equivalent value, whereas Hoagland and Paglin value them at costs to taxpayers (including administrative costs).55 Paglin's study, which shows the greatest redistributive impact of in-kind transfers, rests on questionable assumptions for distributing some program benefits, does not use microdata (and thus fails to distinguish eligible from ineligible households in an income class), and does not account for taxes paid. All of these procedures tend to overstate the redistributive impacts. Nevertheless, estimates of poverty from studies that account for cash and in-kind transfersare closer to each other than they are to the estimates of studies using Census data excluding in-kind transfers. The change in poverty refers to the absolute impact of transfers on the incomes of the poorest units. A relative dimension 55 Controversysurroundsthe choice of the appropriate manner in which to value in-kind transfers. Browning (1976, 1979), Hoagland (1980), and Paglin (1980) argue that in-kind transfersshould be valued at their costs to taxpayers.Paglin even suggests that an in-kindtransferof $1.00 may be worth more than $1.00 from earnings because it is tax exempt and more than a $1.00 (also tax exempt) cash transfer, because in-kind transfers are more likely to be indexed for changes in the cost of living. In opposition are Smolensky et al. (1977), Kenneth Clarkson (1976), MacDonald(1977),John Kraftand Edgar Olsen (1977), Smeeding (1975), Smeeding and Moon (1980), and Murray(1980a), who argue that the recipient is constrained by an in-kind transfer in that purchase of the subsidized good is required. Thus, a dollar in-kind is worth less than a dollar of cash. Smeeding and Moon (1980) contrast the results from using taxpayer cost, the Hicksian equivalent variation measure of utility theory, and the funds released to purchase nonsubsidized commodities (which does not allow commodity substitution).They find that the choice of technique has little effect on the evaluation of redistributive impacts. examines how transfers affect the share of total income received by those in the lowest quintile. TABLE 10 describes and summarizes these quintile share studies. Two issues arise. First, some studies include only family units (Smeeding, 1979; Browning, 1979), while others include both families and unrelated individuals (Danziger and Plotnick, 1977; Hoagland, 1980). One study examines the distribution of income among persons (Browning and William Johnson, 1979). Second, different procedures are used for ranking individuals in the pretransfer and posttransfer distributions. Most rank the income units by the income definition being analyzed (i.e., the units in the lowest quintile in the pretransfer and posttransfer distributions are not the same). This reflects the "anonymity principle"-the degree of inequality is invariant to a reversal of positions by any two units. Browning and Johnson (1979), however, prefer to maintain units at their pretransfer ranking when measuring posttransfer inequality. Because substantial differences in rankings arise because of transfers, estimates of redistributive effects are sensitive to the procedure chosen.56 In TABLE 10, results from research using microdata are based on reranked distributions. Studies that use grouped data (Smeeding, 1979; Browning, 1979) cannot rerank units. All of the studies show a large impact of transferson the share of the low56 Browning and Johnson (1979) show that the two methods of ranking yield identical results only if the transfer system is "horizontallyequitable"-if transfers do not alter rankings (Atkinson,1980; Plotnick, 1981). However, Plotnick and Skidmore(1975) show that existing transferprogramsproduce a substantial degree of reranking. In a transfer program such as a negative income tax, benefits would decline with income and no units would be reranked.Thus, a procedure of not reranking indicates the potential equalizing effects of improved targeting of current transfers on the pretransfer poor. However, units must be reranked if the actual redistributive effects of existing transfers and the current degree of posttransfer inequality are to be gauged. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0 Cl) C) 00 C's C.) C.) 0 V 00 t- (M Cd in C6 0 PL4 u z 0 E-4 z O CA z 9 cd 04 .< z 4-i C), Cd Cd Cd 0 0 u : C) C.2 4-0, bO 0 CA 1:1 0 1-.4 U) 0.4 u CA z (M 04 O 0 (M 04 0 ") 1.4 0 0 00 V 4 0 4-i Cd -0 1.0 M II- I00 (M (M > rT4 1-4 04 -4 04 Oi 0 (M (M o 4-i o 4-i 0 C) C) Cd -o C's 1.4 11) 0 1.4Cd 0 04 O q 04 0 N > Cd Cd 4 0 P o O 04,0 O o " 0c) a) o > o 04 cn cd 6 U 0 10-0 0 (mt Cd 0 04 0 00 0 0 0 00 0 bO 4 mi- -0 =0 N bO .5 0 0 cn cn 4 = low > o ".I 1.4 Cd 42 :Z Cd 4 O 04 bO 0 .4 d , .5 I- -1-6 (M bO Cd o C) 'Cl 0 cn 0 Z This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 144 0 4 1012 Journal of Economic Literature, Vol. XIX (September 1981) est quintile; the impact increases over time. Danziger and Plotnick (1977) find that cash transfers alone add at least 3 percentage points to this share (this estimate would probably rise to about 4 percentage points if the income data had been corrected for income underreporting). Browning (1979), valuing in-kind transfers at taxpayer costs, concludes that for families only, in-kind transfers increase the share of the bottom quintile by 1.59 percentage points; Smeeding (1979), valuing benefits at recipient values, finds a gain of 1.06 percentage points.57 The most comprehensive estimates are Browning and Johnson's (1979) for 1976 and Hoagland's for 1980 which use microdata corrected for income underreporting. Browning and Johnson estimate that all transfers increase the income share of the bottom quintile by 4.6 percentage points; Hoagland estimates an increase of 5.9 percentage points. Studies of the effect of transfers across the entire income distribution typically evaluate their impact on the Gini coefficients. They are shown in TABLE 11. Again the studies vary in terms of income units, data sets, and transfers analyzed. All of the studies that use data from more than one year (Danziger and Plotnick, 1977; Reynolds and Smolensky, 1977; Smeeding, 1977; and Hoagland, 1980) find that transfers reduce the Gini more in later years than in earlier years. However, while transfers have increased rapidly in recent years, the redistributive impacts have increased only slightly. For the reasons mentioned above, Hoagland's study is the most comprehensive. For 1980, he finds a 15 percent reduction in the Gini due to cash transfers and a further 4 per57 Hoagland also values in-kind transfers at taxpayer cost. Higher estimates of quintile shares in the table are due to the exclusion of unrelated individuals or the failure to rerank units. Browning and Johnsonderive an even larger estimate for the quintile share (not shown in table) by imputing values for educational benefits and leisure. cent decline due to in-kind transfers.58 This redistributive impact may appear smaller than the impact on poverty or the quintile share. However, the Gini coefficient is a relatively insensitive measure that has varied little in the U.S. for the past 30 years. The impact of transfers in any year is large relative to that of other policies. For example, Reynolds and Smolensky (1977) find that Social Security benefits alone produce four times the reduction in the Gini coefficient as does the federal personal income tax. Most studies in TABLE 11 employ families or families and unrelated individuals as the income unit when estimating redistributive impacts. This procedure assigns the same weight to each unit, irrespective of its size. An alternative procedure weights each unit's income by the number of individuals it contains (Danziger and Taussig, 1979). An exercise using 1979 CPS data to compare the redistributive effects of these variations in the income unit reveals a slightly smaller decline in the Gini coefficient when units are weighted by their unit size (13 percent) than when weighted equally (15 percent).59 Most studies also assume that even though income is shared within the household unit, the unit's well-being is invariant with respect to household size. Stanley Lebergott (1976) argues that the use of household income is the appropriate income concept because any concept which varies with household size will show an 58 The redistributive effect varies significantly by program and by demographic group. For example, Danziger (1977) estimates a Gini coefficient reduction of 9 percent due to Social Security benefits and 3 percent due to cash public assistance for 1974. By implication, the remaining 3 percent is due to other cash social insurance transfers.Social security reduced the Gini coefficient by about 2 percent for the nonaged, but by almost 30 percent for the aged. 59This effect is to be expected as income and family size are positively correlated in the U.S. This is not the case in other countries, however, as Jan Tinbergen (1975) emphasizes. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions (6- 0 10 co Z 10 1.4 z 0 1--4 eq 00 oi rT4 w u 0 .4 Z 00 "-4 q (3it6 4 C) V V to to 0 c6 6 c, "4 4 o6 o6 CD r- r- 00 C) m C'sC's t--:c4 6 t6 v Cd 'd 'd 0> 0 bO rT4N 0 z 0 Cd Cd u Cd u u u uuu C>t S' o 6a ,O Cd u u 0 o, z 0 0 O O 0 0 > > 0 o- 0 4 U) 0 4-1 04 .6 z O z CA tZ la) 44 ed in'" t0 mmmm P--4 biD W 00 0 Cd 10 C C43 t- a 04.0 m " 0 C) 1-4 z 0 4 4-1 4-1 04 I z C4, - cd 1m OC) 12) CIJ 0 1-4 u U U 44 0 E-4 u 04 0 >! 4-i cn Cd 1-4 14 z + )4 cn Ca z 1-4 0 la) 0O z 14 >, :t4 1-4 a) cn -4 > 0 >.:z cn la) la) la) 0 cn 1-4 0 > u rX4 00 1-4 P--4 1-.4 --4 0 0 0 O -4 1: -r4 0 cn ._q oo > 0 o Q4 0 cli 0 0 cn bO I.., -',4 U 0 I cn to -4 N bO .,O C'n, cu 0 0 0 0 C) 00 C) OC) m P--4 1-1 m bO cn 0 -W A., cn Z 'a) P4 zo 0 la)>.--4 12) > 1-4 o u o 1-4cn cn '15 0 This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0 ci 1-4 la) Z O cli cli 12) >, ..4 0 > Cd 04 1014 Journal of Economic Literature, Vol. XIX (September 1981) increase in well-being if one household member dies and income remains constant. Simon Kuznets (1950) and Browning (1976) argue that per capita household income better reflects "true" living standards because, if two households have the same income, the larger one will have a lower standardof living. A third approach employs equivalence scales that account for intra-household economies of scale (Morgan and James D. Smith, 1969; Jacques van der Gaag and Smolensky, 1980). Previous studies have shown that inequality in per capita or per equivalent adult income is less than in household income. However, they have not analyzed the sensitivity of transfers' redistributive impact to changes in the income concept. Our calculations using 1979 CPS data suggest a small difference: cash transfers reduce the Gini coefficient of household income by 15 percent, but that of per capita income by 18 percent. The choice of an income accounting period may also affect an evaluation of redistributive impacts. Most studies use annual data, which include both transitory and life cycle income components. Annual income inequality exceeds that of lifetime income (Lillard, 1977; Blinder, 1974) by more than it exceeds inequality in multiyear income (Benus and Morgan, 1975; Saul Hoffman and Nripseh Podder, 1976). However, no studies measure the redistributive effect of all transfersusing a lifetime or even a multiyear accounting period. Because some social insurance benefits are age-related and other transfers respond directly to transitory income fluctuations, studies employing an annual reporting period overstate the redistribution in lifetime income caused by transfers. Consider an actuariallyfair Social Security program in which individuals are taxed when young, and receive benefits when old equal to those from a private pension to which contributions equal to taxes were made. If one compared current pretransfer and posttransferincome of the aged during a single year, one would find substantial redistribution. In this case, however, the effect of the program on lifetime income would actually be zero because the Social Security payments would merely represent a transfer of resources by each individual from an earlier to a later period, and the payment should be treated in the same manner as private savings. Burkhauser and Warlick (1982, forthcoming) test the extent to which pre-post comparisons in a single year overstate the redistributive effects of Social Security. They find that private annuity-type benefits in the existing system account for over 25 percent of the total, with about 75 percent of benefits appropriately treated as redistribution from young to older workers, rather than as annuity benefits. Over time the share of the benefit that would be reflected in an annuity is increasing, so the upward bias should decline. Burkhauser and Smeeding (1981) find the redistributive component of Social Security benefits to be more pro-poor than that of total benefits. For the lowest quintile almost 90 percent of the benefits represent "true" redistribution. What, then, can we conclude? Despite substantial variation-indeed controversy-with respect to the income unit, income definition, income accounting period, valuation of in-kind benefits, and choice of ranking methodology, the studies are in basic agreement. Cash and inkind transfers reduced the percentage of persons living in households with incomes below the poverty line by 53 percent in 1968 (Smeeding) and by 78 percent in 1980 (Hoagland). They increased the income share of the lowest quintile in recent years by about 4.6 (Browningand Johnson) to 5.9 (Hoagland) percentage points, and reduced the Gini coefficient by about 19 percent (Hoagland). These impacts have This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects increased over time, but not as rapidly as the increase in transfer expenditures-diminishing redistributive returns60are in evidence. Improvements in our understanding of the redistributive effects of transfersawait several developments. A comprehensive data set, with information on all forms of cash and public and private in-kind income, assets and liabilities, and other components of economic welfare, would permit more thorough analysis than do current data.61 Further attention to the effect of transfers on the distribution of permanent income over the life cycle is in order. Most important, studies that model how behavioral responses to transfers affect the pretransfer income distribution are needed. IV. The Labor Supply and Redistributive Effects of Transfer Program Reforms A few recent studies have evaluated the effect of proposed reforms in transferprograms, as opposed to the impact of existing programs. Most of these analyses have considered welfare reform options (e.g., negative income tax or demogrant proposals), have used the labor supply parameter estimates from the negative income tax experiments, and have employed a microdata simulation methodology. This approach typically begins with a nationally weighted microdata base. Labor supply functions with income and substitution parameters that vary by demographic group are assigned to each individual. Then, the changes in both net wage rates and nonemployment income resulting from a policy proposal are determined for each household unit by the ap60 This term is due to Reynolds and Smolensky (1978). 61 Data are particularly weak for in-kind income of all sorts, including fringe benefits and perquisites, interhousehold private transfers, and wealth. The Census Bureau plans new surveys which will overcome some of these deficiencies. 1015 plication of the eligibility rules and benefit schedules of the proposed reforms. These changes are combined with labor supply parameters to predict the effect of the policy on each household'slabor supply, earnings, and total income. The resulting estimates are aggregated to national totals to ascertain the effect on poverty levels, labor supply, income inequality, and budgetary costs. The labor supply parameters have been drawn primarily from the Seattle-Denver income maintenance experiments.62 In addition to improved reliability relative to nonexperimental research, the experimental design mimics the reality of a change in transfer policy. A. Labor Supply Effects In evaluating the labor supply effects of a policy change, the responses of three groups must be considered-persons receiving benefits from existing programs (who may experience either positive or negative net wage rate and income changes), persons who become eligible for transfers because of the program reform (for whom net work incentives would fall), and those who help finance the reform through their taxes (forwhom net changes 62 The experimental parameter estimates have a narrowerrange than those estimated from cross-section data, and are somewhat lower. For males, the experimental income and substitution elasticities range from zero to 0.09 and zero to 0.16, respectively. The intervals for wives are zero to 0.14 and .11-.42 and for female heads are .07-.34 and .08-.18. The resultsof the experimentsarejudged more reliable than crosssectionresultsbecause of the reduction in potential unobserved variables bias due to the experimental design. Experimental results are subject to problems of their own, however. These include time horizon problemsdue to limited duration, Hawthorne effects, design problems due to sample truncation,allocation decisions, and attrition and income underreporting problems. While methods to reduce some of these problems have been developed, other problems-in particular,underreporting and the difficulties of generalizing to the national population-remain. Moffitt and Kenneth Kehrer (1981) review these issues. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions e0 - u u 0 60 00, 0 'a)U 0 C 0 D .0 0 _ ~~~~~Z4 ~~~~~~~~~~~~~~4 4 & 0 ~~~~~~~~~~~~Z4 ~ ~ ~~ ~ ~ ~ 0)~~~~ ~ ~ ~ ~ ~ ~ ~ ~~ - 'S 4 , 0 ~~~~~~~~~~~~~~~Z4 ~ ~ ~ ~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~4iC " 0~~~~~~~~~~~~~~c 4):0) ) cq 0)0) t 0) bo a)bo m ~~~~~o~~~~~~~~~~Z 0~~~~~~~~~C OD' v bo CO C This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 0~~~~~~~~~ 0 0c, 00~~~~ - 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0c 10 ~~~\0 ~~~00 0 0 0 - C~~0 0 0 ~ ~ ,0 C00 00> 1?-~~~~~~~~~~~~~~~10 0>~ ~ ~ (4-4 0 -- 0 ~ ~ 0 ~ ~ ~ 0 0 da ~ " >~~~~~~~~~~~~~~~* 5~~~~~~~~~ 00 0 00 m SM 0 0 cd o - -1 0 c;T 0 A?" ~ ~ ~ 0 i u I .0b1- U This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1018 Journal of Economic Literature, Vol. XIX (September 1981) in work incentives are ambiguous). The net change in hours worked, earnings, and incomes for the entire population is thus theoretically indeterminate. While the main differences in the primary simulation studies incorporating behavioral changes are summarized in TABLE 12, two points should be noted.63 First, only the David Betson, Greenberg and Richard Kasten (1981, forthcoming) study indicates an increase in private sector earnings due to negative income tax (NIT) or demogrant plans.64This results from their simulation of the behavioral responses of both taxpayers and transfer recipients: the increased transfers induce a decline in earnings among recipients, but this is offset by the increased work of taxpayers (whose labor supply curve is estimated to be backward-bending).A second paper by these authors (Betson, Greenberg, and Kasten, 1980) also suggests a small earnings increase from a welfare reform, due to the inclusion of a public jobs component. The remaining studies, none of which include taxpayer responses, usually find that relative to current income transfer programs, welfare reforms would have only small additional negative hours and earnings effects in the aggregate. A different simulation approachis taken in Haveman et al. (1980) who model both labor supply and consumption expenditure changes. The latter effects cause changes in gross outputs, employment and earnings. They find that the induced production effects lead to increased output, employment, and earnings. These simulation results convey a clear simulation studies which use non63Early experimental labor supply parameters include Greenberg and Kosters (1973), Samuel Rea (1974) and Mastersand Garfinkel(1977). 64 Philip Robins et al. (1980), extending work in Michael Keeley et al. (1978), also find an increase in private sector earnings,but only in the least generous NIT analyzed. However, this "reform" would increase workby making most current welfare recipients worse off than currently. message: fairly major reforms and extensions of current transfer programs can be undertaken without larger additional adverse impacts on aggregate labor supply. The studies generally find declines in work hours of under one percent. With some reform options, especially those which include a public employment component, total work effort increases. B. Redistributive Effects Several studies in TABLE 12 also simulate the distributional impact of welfare reforms. The NIT simulated by Betson, Greenberg and Kasten (1981, forthcoming) increases budgetary costs by 83 percent over the welfare programsit replaces (an increase of $17.5 billion over status quo expenditures) and reduces poverty by 16 percent and the Gini coefficient by about 10 percent. A credit income tax with the same budgetary costs would reduce the Gini coefficient by about 8 percent, while one that increased budgetary costs by 50 percent over that of the NIT would lower the Gini by about 13 percent.65 When the same plan is simulated by Robins, Richard West, and Michael Lohrer (1980), substantially different poverty reduction effects are estimated due to their omission of single-person households and the aged. The Congressional Budget Office (1979) simulation of the distributional (but not labor supply) impact of the 1977 Program for Better Jobs and Income finds a 39 percent increase in budgetary costs, and a reduction in poverty of about 30 percent. These results indicate that while further reductions in inequality and poverty are possible by well-targeted reforms and extensions of current welfare programs, most of the easy gains have already been achieved by current programs. 651n earlier studies, Watts and Kim Peck (1975) and Danziger and Haveman (1977) simulate the distributional impact of a credit income tax, but do not model labor supply responses. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects While the microdata simulation methodology has advanced rapidly in recent years, the state-of-the-art is still evolving (Haveman and Kevin Hollenbeck, 1980). A serious shortcoming is the absence of labor supply parameters for higher income groups. Also, the analyses have neglected demand-side factors by implicitly assuming a perfectly elastic demand for the services of transfer recipients and, hence, no change in wages due to induced labor supplies or demands. Given less than perfect elasticity, estimates of earnings and hours changes are biased to some unknown extent. Moreover, findings are offered without confidence intervals (except in Robins, West and Lohrer) although the labor supply parameters that underlie the results are estimated with error and there is sampling error in the data sets used to provide national estimates. Finally, current simulation models generally neglect the following possibilities that could result from transfer policy changes: revisions in administrationwhich may influence work responses; the creation of nonlinear budget constraints;and changes in behavioral responses due to associated public jobs programs. VI. Summary and Conclusion This review of the labor supply, savings, and redistributive effects of income transfers substantiatestheir importance and re- 1019 veals the range of their estimated magnitudes. TABLE 13 presents our summary judgments of their total effect based on the empirical studies reviewed. The first column is backward-lookingit appraises the effects of programs as they currently exist relative to the zero-publictransfers counterfactual. The first entry indicates that current programs reduce aggregate labor supply by 4.8 percent. However, our counterfactual ignores any changes in private transfer mechanisms that would exist in the absence of-or be induced by-the withdrawal of public transfers. Because private transferswould have their own work disincentives, and because our estimate does not reflect "entitlement effects" which tend to increase work effort, 4.8 percent probably is an upper bound. The 0-20 percent reduction in private savings reflects the wide variation of estimated effects. There is more consensus on redistributive effects. Our review suggests that the incidence of poverty is about 75 percent lower and the Gini coefficient about 19 percent lower than in the absence of transfers.The redistributive studies, however, adjust for neither the replacement of public by private transfersin the absence of the former, nor for the tendency of transfers to increase pretransfer poverty by both reducing work effort and enabling persons with low market incomes to live independently. As TABLE 13 LABOR EFFECTS SUPPLY, OF INCOME EFFECT INCOME SAVINGS, AND REDISTRIBUTIVE TRANSFER PROGRAMS: OF CURRENT TRANSFER PROGRAMS Labor Supply Private Savings Income Poverty Reduction of 4.8 percent Reduction of 0-20 percent Reduction of 75 percent Income Inequality (Gini Coefficient) Reduction of 19 percent A SUMMARY EFFECT OF MARGINAL PROPORTIONAL OF CURRENT EXPANSION PROGRAMS Negative Neutral or slightly negative Not large, as most easy gains have been made Not large, as most easy gains have been made This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1020 Journal of Economic Literature, Vol. XIX (September 1981) a result, the redistributive effects are also upward-biased. Our conclusion on the labor supply effect of transfers leads us to speculate on the economic cost per dollar of transfersthe leak in Arthur Okun's (1975) income redistribution bucket. The earnings loss resulting from the 4.8 percent labor supply reduction is probably about 3.5 percent because not all of the additionallabor supply would find employment and because persons subject to the work disincentives of current transfers tend to have below average wage rates. Transfers equalled 15.2 percent of total earnings in 1978. Thus, the leak was roughly 23? per dollar of transfer expenditures (3.5/ 15.2 = 0.23). This estimate excludes several other efficiency effects of transferstheir administrative costs, welfare losses from financing transfers and the welfare gains of increased leisure. The second column of TABLE 13 is prospective. It reflects our judgment regarding the effects of a modest proportional increase in current benefit levels. This expansion is not a reform of the type reviewed in TABLE 12, which might change the way current programs target benefits or affect work and savings. Whatever one's opinion as to whether or not the marginal redistributive benefits of the current system exceed its marginal factor supply costs, it appears that such a proportional expansion would yield a less favorable tradeoff than exists currently. Such a change is not likely to produce a sizable reduction in poverty, for most of the additional payments would go to recipients who have already been taken above the poverty line by transfers. Similarly, inequality would decrease very slightly because an increased share of the payments would be received by persons above the second quintile. At the same time, if the marginal labor supply and savings responses are constant (and they may well be increasing), the aggregate loss in earnings or savings per dollar of additional transfers will not fall. Thus, while the redistributive returns from program expansion would diminish, marginal work and savings effects would at best remain constant. Of course, a complete appraisal of the benefits and costs of U.S. income transfer programs must consider not only the three issues treated in this review, but also those effects that we did not analyze. These include the efficiency and redistributive impact of the taxes that finance these programs and the impact of public transfers on demographic choices, the unemployment rate, and the other important economic and social variables noted earlier (sup. 976). Given this caveat, what implications can be drawn from the research findings reviewed? Are the labor supply and savings effects of transfersso large relative to their redistributive effects that the absolute size of these programs should be reduced or their growth controlled? Are these costs small enough to support the conclusion that additional redistribution be pursued without concern for labor supply and savings effects? Can the research results guide a restructuring of current transfer programs to reduce the disincentives associated with the redistribution achieved? Although judgment and interpretation must loom large in answers to these questions, we venture the following hypotheses. First, compared to continued proportional expansion of current benefits, reforms can be designed to reduce work and savings disincentives without sacrificing the distributional effects that have been achieved. Second, reductions in, or the elimination of, current benefits will increase income poverty and achieve only small increases in work effort and savings. Finally, the research findings are too varied, too uncertain, and themselves too colored with judgment to serve as more than a rough guide to policy choices. Perhaps future methodological developments and This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects improvements in data (e.g., panel data with improved information on income sources, transfer program participation, and background characteristics),and estimation techniques (e.g., modeling the true budget constraint, dynamic models, and econometric techniques for analyzing panel data) can decrease the domain over which valiie ilidcmentk now relgn. REFERENCES ABRAMOVITZ,MOSES. "Welfare Quandaries and ProductivityConcerns,"Amer. Econ. Rev., March 1981, 71(1), pp. 1-17. ANDO, ALBERT AND MODIGLIANI,FRANCO. "The 'Life-Cycle' Hypothesis of Saving: Aggregate Implications and Tests," Amer. Econ. Rev., March 1963, 53(1), pp. 55-84. ASHENFELTER,ORLEY."Unemployment as Disequilibrium in a Model of Aggregate Labor Supply," Econometrica,April 1980, 48(3), pp. 547-64. ATKINSON,ANTHONYB. "On the Measurement of Inequality,"J. Econ. Theory,Sept. 1970, 2(3), pp. 244-63. . "Horizontal Equity and the Distribution of the Tax Burden," in The economics of taxation. Edited by HENRYJ. AARONAND MICHAELJ. BosKIN. Washington, D.C.: Brookings Institution, 1980, pp. 3-18. BAILY,MARTINNEIL. "Unemployment Insuranceas Insurancefor Workers,"Ind. Lab. Relat. Rev.,July 1977, 30(4), pp. 495-504. BALCER, YVES. "The Impact of Lifetime Earnings and Pension on Retirement and Savings:A Critical View." Discussion paper, Institute for Research on Poverty, Madison, Wis., forthcoming, 1981. BARR, NICHOLASAND HALL, ROBERT."The Probability of Dependence on Public Assistance,"Economica, forthcoming, 1981. BARRO, ROBERTJ. "Are Government Bonds Net Wealth?"J. Polit. Econ., Nov.-Dec. 1974, 82(6), pp. 1095-1117. . The impact of Social Security on private sav- ing: Evidencefrom the U.S. time series. Washington, D.C.:AmericanEnterprise Institutefor Public Policy Research, 1978. AND MAcDONALD, GLENN M. "SocialSecurity and Consumer Spending in an International Cross Section," J Public Econ., June 1979, 11(3), pp. 275-89. 1021 surance:The Recipients and Its Impact,"Southern Econ. J., Jan. 1981, 47(3), pp. 606-16. BEHRENS, JEAN AND SMOLENSKY, EUGENE. "Alternative Definitions of Income Redistribution,"Public Finance, 1973, 28(3-4), pp. 315-32. BENDT, DOUGLAS. The effects of changes in the AFDC programon effectivebenefit reduction rates and the probability of working. Princeton, NJ.: Mathematica Policy Research, 1975. BENUS, JACOB AND MORGAN, JAMES N. "Time Period, Unit of Analysis,and Income Concept in the Analysisof Income Distribution,"in 7he personal distribution of income and wealth. Studies in Income and Wealth, Vol. 39. Edited by JAMES D. SMITH. New York:NBER;distributedby Columbia University Press, New Yorkand London, 1975, pp. 209-24. BERKOWITZ, MONROE; JOHNSON, WILLIAM G. AND MURPHY, EDWARD H. Public policy toward disa- bility. New York:Praeger Publishers, 1976. "A Simulation Study of the Interaction Between Transfer Policy and Employment Programs."Unpublished paper. Madison, Wis.: Institute for Research on Poverty, 1980. BETSON, DAVID AND GREENBERG, DAVID. GREENBERG, DAVID AND KASTEN, RICHARD. "A MicrosimulationModel for Analyzing Alternative Welfare Reform Proposals:An Application to the Program for Better Jobs and Income," in Microeconomic simulation models for public policy analysis, Vol. 1. Edited by ROBERT HAVEMAN AND KEVIN HOLLENBECK. New York:Academic Press, 1980, pp. 153-88. GREENBERG, DAVID AND KASTEN, RICHARD. "A SimulationAnalysisof the Economic Efficiency and DistributionalEffects of Alternative Program Structures:The Negative Income Tax versus the Credit Income Tax,"in Income-testedtransferprograms: a case for and against. Edited by IRWIN GARFINKEL. New York: Academic Press, forthcoming, 1981. BISHOP, JOHN H. "Jobs,Cash Transfersand Marital Instability: A Review and Synthesis of the Evidence," J. Human Res., Summer 1980, 15(3), pp. 301-34. BLINDER, ALAN S. Toward an economic theory of income distribution.Cambridge,Mass.:MITPress, 1974. . "Distribution Effects and the Aggregate Con- sumption Function," J Polit. Econ., June 1975, 83(3), pp. 447-75. . "The Level and Distribution of Economic BARRON,JOHNM. AND GILLEY,OTIS W. "TheEffect of Unemployment Insurance on the Search Pro- cess," Ind. Lab. Relat. Rev., April 1979, 32(3), pp. 363-66. AND MELLOW,WESLEY."Unemployment In- Well-Being,"in TheAmerican economy in transition. Edited by MARTIN FELDSTEIN. Chicago: University of Chicago Press, 1980, pp. 415-479. GORDON, ROGER H. AND WISE, DONALD E. "Reconsidering the Work Disincentive Effects of This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1022 Journal of Economic Literature, Vol. XIX (September 1981) Social Security," Nat. Tax. J, December 1980, 33(4), pp. 431-442. ; GORDON, ROGER AND WISE, DONALD. "Life Cycle Savings and Bequests: Cross-SectionalEstimates of the Life-Cycle Model." Unpublished paper. Princeton, NJ.: Princeton University, 1981. BORJAS, GEORGE J. "The Relationship Between Wages and Weekly Hours of Work:The Role of Division Bias," J. Human Res., Summer 1980, 15(3), pp. 409-23. BOSKIN, MICHAEL J. "The Economics of Labor Supply," in Income maintenance and labor supply: Econometricstudies. Edited by GLEN G. CAIN AND HAROLD W. WATTs. New York:Academic Press, 1973, pp. 163-81. . "Social Security and Retirement Decisions," Econ. Inquiry,January 1977, 15(1), pp. 1-25. AND HURD, MICHAEL D. "The Effect of Social Security on Early Retirement," J Public Econ., December 1978, 10(3), pp. 361-77. AND ROBINSON, MARC. "Social Security and Private Saving:AnalyticalIssues,Econometric Evidence, and Policy Implications," U.S. Congress, Joint Economic Committee: Special Study on Economic Change, Vol. 8. Social Security and Pensions: Program of Equity and Security. Dec. 4, 1980, pp. 38-64. BREHM, CARL T. AND SAVING, THOMAS R. "The Demand for General Assistance Payments," Amer. Econ. Rev., December 1964, 54, pp. 1002-18. BROWNING, EDGAR K. "The Trend TowardEquality in the Distribution of Net Income," Southern Econ. J, July 1976, 43(1), pp. 912-23. . "The Marginal Welfare Cost of Income Re- distribution,"Southern Econ. J., July 1978, 45(1), pp. 1-17. . "On the Distribution of Net Income: Reply," Southern Econ. J., January 1979, 45(3), pp. 94559. AND JOHNSON, WILLIAM R. "Taxes,Transfers and Income Inequality,"in Regulatory change in an atmosphereof crisis: Current-dayimplications of the Roosevelt years. Edited by GARY M. WALTON. New York:Academic Press, 1979, pp. 12952. BURKHAUSER, RICHARD V. "The Early Acceptance of Social Security: An Asset Maximization Approach," Ind. Lab. Relat. Rev., July 1980, 33(4), pp. 484-92. AND QUINN, JOSEPH. "The Effect of Changes in MandatoryRetirement Rules on the LaborSupply of Older Workers."Presented at NBER Conference on the Economics of Compensation,Cambridge, Mass.November 1980. AND SMEEDING, TIMOTHY. "The Net Impact of the Social Security System on the Poor," Public Policy, Spring 1981, 29 (2), pp. 159-78. AND TURNER, JOHN A. "A Time Series Analysis of Social Security and its Effect on the Market Work of Men at Younger Ages," J. Polit. Econ., August 1978, 86(4), pp. 701-15. AND TURNER, JOHN A. "Can Twenty-Five Million Americans Be Wrong?-A Response to Blinder, Gordon and Wise," Nat. TaxJ., forthcoming, Dec. 1981. AND WARLICK, JENNIFER. "Disentanglingthe Annuity From the RedistributiveAspects of Social Security," Rev. Income Wealth, forthcoming, 1982. BURTLESS, GARY AND HAUSMAN, JERRY A. "The Effect of Taxation on Labor Supply: Evaluating the Gary Negative Income Tax Experiment,"J. Polit. Econ., Dec. 1978, 86(6), pp. 1103-30. CAGAN, PHILLIP. The effect of pension plans on aggregate saving; evidence from a sample survey. New York:NBER; distributed by Columbia University Press, 1965. CAIN, GLEN G. AND WATTS, HAROLD, W., eds. Income maintenance and labor supply; econometric studies. New York:Academic Press, 1973. AND WATTs, HAROLD. "Towarda Summary and Synthesis of the Evidence," in Income maintenance and labor supply; econometricstudies. Edited by GLEN CAIN AND HAROLD WATTs. New York:Academic Press, 1973, pp. 32867. CLARK, ROBERT; KREPS, JUANITA AND SPENGLER, JOSEPH. "Economics of Aging: A Survey,"J. Econ. Lit., Sept. 1978, 16(3), pp. 919-62. AND JOHNSON, THOMAS. Retirement in the dual careerfamily. Final Reportfor the U.S. Social Security Administration.Raleigh,N.C.:North Carolina State University, 1980. CLARKSON, KENNETH W. "Welfare Benefits of the Food Stamp Program," Southern Econ. 1. July 1976, 43(1), pp. 864-78. CLASSEN, KATHLEEN P. "The Effect of Unemployment Insurance on the Duration of Unemployment and Subsequent Earnings,"Ind. Lab. Relat. Rev., July 1977, 30(4), pp. 438-44. ."Unemployment Insuranceand Job Search," in Studies in the economics of search. Edited by STEVEN A. LIPPMAN AND JOHN J. MCCALL. Amsterdam: North-Holland, 1979, pp. 191219. COE, RICHARD D. "Dependency and Poverty in the Short and Long Run,"in Five thousand American families-patterns of economic progress, Vol. VI. Edited by GREG J. DUNCAN AND JAMES N. MORGAN. Ann Arbor: Institute for Social Research, 1978, pp. 273-94. CONGRESSIONAL BUDGET OFFICE. "An Analysis of the Administration's Social Welfare Reform Amendments of 1979." Mimeographed.Washington, October 1979. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects The sensitivity of measured comprehensiveincome inequality to aggregation, reranking, and underreporting. Unpublished Ph.D. Thesis, University of Wisconsin, Madison, Department of Economics, August 1980. DANZIGER, SHELDON. "Income Redistributionand Social Security:Further Evidence," Social Service Rev., March 1977, 51(1), pp. 179-84. ____ AND HAVEMAN, ROBERT. "Tax and Welfare Simplification:An Analysis of Distributional and Regional Impacts," Nat. Tax.J, September 1977, 30(3), pp. 269-83. DALRYMPLE, ROBERT. ; HAVEMAN, ROBERT AND PLOTNICK, ROBERT. "Income Transfer Programs in the United States: An Analysis of Their Structure and Impacts," in Federalfinance: The pursuit of American goals. Special Study on Economic Change, Joint Economic Committee, Vol. 6, Dec. 23, 1980, pp. 225-86. ; HAVEMAN, ROBERT AND SMOLENSKY, EuGENE."The Measurement and Trend of Inequality: Comment," Amer. Econ. Rev., June 1977, 67(3), pp. 505-12. AND PLOTNICK, ROBERT. "Demographic Change, Government Transfers,and Income Distribution,"Mon. Lab. Rev., April 1977, 100(4),pp. 7-11. ~ AND PLOTNICK, ROBERT. Has the war an poverty been won? Madison, Wis.: Institute for Research on Poverty, forthcoming, 1981. ~ AND TAUSSIG, MICHAEL K. "The Income Unit and the Anatomy of Income Distribution," Rev. Income Wealth, December 1979, 25(4), pp. 36575. DARBY, MICHAEL R. The effects of Social Security on income and the capital stock. Washington, D.C.: AmericanEnterprise Institutefor Public Policy Research, 1979. DAVID, MARTIN AND MENCHIK, PAUL. "Aspects of the Lifetime Distribution of Income and Wealth." Unpublished paper. Madison, Wis.: Institute for Research on Poverty, 1979. __ AND MENCHIK, PAUL. "The Effects of Social Security on Bequests." Discussion paper No. 63780, Institute for Research on Poverty, Madison, Wis., 1980. DAVIES,JAMES."Uncertain Lifetime, Consumption, and Dissavingin Retirement,"J Polit. Econ.,June 1981, 89(3), pp. 561-77. DIAMOND, PETER A. AND HAUSMAN, JERRY. "Indi- vidual Savings Behavior." Unpublished paper. Cambridge, Mass.: Massachusetts Institute of Technology, 1980. ____ AND MIRRLEES, JAMES A. "A Model of Social Insurance with Variable Retirement," J. Public Econ., Dec. 1978, 10(3), pp. 295-336. DRAZEN,ALLAN."Government Debt, Human Capi- 1023 tal, and Bequests in a Life-Cycle Model,"J. Polit. Econ., June 1978, 86(3), pp. 505-16. EHRENBERG,RONALDG. AND OAXACA,RONALDL. "Unemployment Insurance, Duration of Unemployment and Subsequent Wage Gain," Amer. Econ. Rev., Dec. 1976, 66(5), pp. 754-66. FELDSTEIN,MARTINS. "Social Security, Induced Re- tirement and Aggregate Capital Accumulation," J Polit. Econ., Sept./Oct. 1974, 82(5), pp. 905-26. . "Social Security and Saving: The Extended Life Cycle Theory,"Amer.Econ. Rev., May 1976a, 66(2), pp. 77-86. . "Temporary Layoffs in the Theory of Unem- ployment," I. Polit. Econ., Oct. 1976b, 84(5), pp. 937-957. . "Social Security and Private Savings: Interna- tional Evidence in an Extended Life-Cycle Model," in The economics of public services. Edited by MARTIN FELDSTEIN AND ROBERT INMAN. Lon- don: Macmillan;New York:distributedby Halsted Press, 1977, pp. 174-205. . "The Effect of Unemployment Insurance on Temporary Layoff Unemployment," Amer. Econ. Rev., Dec. 1978a, 68(5), pp. 834-46. "Reply," in The impact of Social Security on private saving: Evidencefrom US. time series. Edited by ROBERTBARRO.Washington: American Enterprise Institute for Public Policy Research, 1978b, pp. 37-47. . "Social Security, Induced Retirement and Aggregate Capital Accumulation: A Correction and Updating," National Bureau of Economic Research Working Paper 579, 1980a. . "International Differences in Social Security and Saving," J Public Econ., Oct. 1980b, 14(2), pp. 225-44. AND PELLECHIO, ANTHONY. "Social Security and Household Wealth Accumulation:New Micro Econometric Evidence," Rev. Econ. Stat., Aug. 1979, 61(3), pp. 361-68. FIELDS, GARY S. "Direct Labor Market Effects of Unemployment Insurance," Ind. Relat., Feb. 1977, 16(1), pp. 1-14. AND MITCHELL, OLIVIA S. "The Effects of Pensions on Retirement: A Review Essay." Unpublished paper. Ithaca, N.Y.:Cornell University, Sept. 1980. FRIEDMAN,MILTON. A theory of the consumption function. Princeton, NJ.: Princeton University Press, 1957. GARFINKEL, IRWIN AND HAVEMAN, ROBERT H. Earnings capacity, poverty, and inequality. New York:Academic Press, 1977. AND ORR, LARRYL. "Welfare Policy and Em- ployment Rate*of AFDC Mothers," Nat. Tax J., June 1974, 27(2), pp. 275-84. GILLESPIE,W. IRWIN. "Taxes, Expenditures and the This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1024 Journal of Economic Literature, Vol. XIX (September 1981) Redistribution of Income in Canada,"in Reflections on Canadian incomes. Edited by the EcoNOMICCOUNCILOF CANADA. Quebec: Canadian Government Publishing Centre, 1980a, pp. 2750. . "The Lorenz-Gini, the Paglin-Gini, and the Measurement of Income Inequality," in Reflections on Canadian incomes. Edited by the EcoNOMICCOUNCILOF CANADA. Quebec: Canadian Government Publishing Centre, 1980b, pp. 557- 66. GOLLADAY,FREDERICKL. AND HAVEMAN,ROBERT H. The economic impacts of tax-transfer policy: Regional and distributional effects.New York:Academic Press, 1977. GORDON, ROGER H. "EmpiricalComparisonsof Alternative Measures of Economic Well-Being," Research Memorandum 220, Princeton University, 1977. AND BLINDER, ALAN S. "Market Wages, Reservation Wages, and Retirement Decisions," J Public Econ., Oct. 1980, 14(2), pp. 277-308. GorrSCHALK, PETER. "Transfer Scenarios and Projections of Poverty into the 1980s," J Human Res., Winter 1981, 16(1), pp. 41-60. GREENBERG,DAVID H. AND KOSTERS,MARVIN."Income Guarantees and the Working Poor: The Effect of Income-Maintenance Programs on the Hours of Work of Male Family Heads," in Income maintenance and labor supply: Econometricstudies. Edited by GLEN G. CAIN AND HAROLD W. WATTS.New York:Academic Press, 1973, pp. 14101. GRONAU,REUBEN."Wage Comparisons-A Selectivity Bias," J. Polit. Econ., Nov./Dec. 1974, 82(6), pp. 1119-43. GULTEKIN,BULENTN. AND LOGUE,DENNIS E. "Social Security and Personal Saving: Survey and New Evidence," in Social security versusprivatesaving. Edited by GEORGEM. VON FURSTENBERG.Cambridge, Mass.: Ballinger Publishing Co., 1979, pp. 65-133. GUSTAFSON,THOMAS. "Labor Supply of the El- derly," in Work,income, and retirement:Report of a workshop. Edited by THOMASGUSTAFSON. Office of Income Security Policy, U.S. Department of Health, Education and Welfare, Technical Analysis Paper No. 18, 1979, pp. 59-72. GUSTMAN,ALAN. "Analyzing the Relation of Unemployment Insurance to Unemployment." Working Paper No. 512, National Bureau of Economic Research, July 1980. HAANES-OLSEN,LEIF. "Earnings-Replacement Rate of Old-Age Benefits, 1965-75, Selected Countries," Soc. Sec. Bull., Jan. 1978, 41(1), pp. 3-14. HAGENS,JOHN. "A Re-Examination of the Link Between Social Security and Savings." Working Pa- per No. 1, Office of Research and Statistics,Social Security Administration,1979. HALL, ARDEN AND JOHNSON, TERRY R. "The Determinants of Planned Retirement Age," Ind. Lab. Relat. Rev., Jan. 1980, 33(2), pp. 241-54. HALL, ROBERT E. "Wages, Income, and Hours of Workin the U.S. Labor Force," in Income maintenance and labor supply: Econometricstudies. Edited by GLEN G. CAIN AND HAROLD W. WATTS. New York:Academic Press, 1973, pp. 102-62. HAMERMESH, DANIEL S. jobless pay and the economy. Baltimore:Johns Hopkins University Press, 1977. - "Unemployment Insurance and Unemployment in the United States,"in Unemployment insurance: Global evidence of its effects on unemployment. Edited by HERBERT G. GRUBEL AND MICHAEL A. WALKER. Vancouver:The Fraser Institute, 1978, pp. 39-58. - "Entitlement Effects, Unemployment Insur- ance and Employment Decisions," Econ. Inquiry, July 1979, 17(3), 317-32. . "Transfers,Taxes and NAIRU,"National Bureau of Economic Research Working Paper No. 548, 1980. HANOCH, GIORA AND HONIG, MARJORIE. "The Labor Supply Curve Under Income Maintenance Programs,"J Public Econ., February 1978, 9(1), pp. 1-16. HARROD, SIR RoY FORBES. Towardsa dynamic economics. London: Macmillan, 1948. HAUSMAN, JERRY A. "The Effects of Wages, Taxes and Fixed Costs on Women's Labor Force Participation," J Public Econ., October 1980, 14(2), pp. 161-94. - "Labor Supply," in How taxes affect eco- nomic behavior. Edited by HENRY J. AARON AND JOSEPH A. PECHMAN. Washington, D.C.: Brookings Institution, 1981, pp. 27-72. - AND WISE, DAVID A. "Social Experimenta- tion, Truncated Distributionsand EfficientEstimation," Econometrica,May 1977, 45(4), pp. 919-38. - AND WISE, DAVID A. "AttritionBias in Experimental and Panel Data: The GaryIncome Maintenance Experiment,"Econometrica,March1979, 47(2), pp. 455-73. HAUSMAN, LEONARD. "The Impact of Welfare on the Work Effort of AFDC Mothers,"in The President's commission on income maintenance programs: Technical studies. Washington, D.C.: U.S. Government Printing Office, 1970, pp. 83-100. HAVEMAN, ROBERT H. "Unemployment in Western Europe and the United States: A Problem of Demand, Structure, or Measurement?"Amer. Econ. Rev. May 1978, 68(2), pp. 44-50. - AND HOLLENBECK, KEVIN, eds. Microeco- nomic simulation modelsfor public policy analy- This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects sis. 2 Volumes. New York:Academic Press, 1980. ; HOLLENBECK, KEVIN; BETSON, DAVID AND HOLMER, MARTIN. "A MicroeconomicSimulation Model for Analyzing the Regional and Distributional Effects of Tax-TransferPolicy: An Analysis of the Program for Better Jobs and Income," in Microeconomicsimulation modelsfor public policy analysis, Vol. 2. Edited by ROBERT H. HAVEMAN AND KEVIN HOLLENBECK. New York:Academic Press, 1980, pp. 131-62. ; WARLICK, JENNIFER AND WOLFE, BARBARA. "Disability,Income Transfers,and Work Effortof Older Workers." Unpublished paper. Madison, Wis.: Institute for Research on Poverty, 1981. HECKMAN, JAMES J. "ShadowPrices, MarketWages, and LaborSupply,"Econometrica,July 1974, 42(4), pp. 679-94. . "The Common Structure of Statistical Mod- els of Truncation, Sample Selection and Limited Dependent Variables and a Simple Estimatorfor Such Models,"Ann. Econ. Soc. Measur.,Fall 1976, 5(4), pp. 475-92. "SampleSelection Bias as a SpecificationError," Econometrica,January 1979, 47(1), pp. 15361. "Sample Selection Error as a Specification Error with an Application to the Estimation of LaborSupply Functions,"in Female labor supply: Theoryand estimation. Edited byJAMEs P. SMITH. Princeton, NJ.: Princeton University Press, 1980, pp. 206-48. ____ AND MACURDY, THOMAS E. "A Life Cycle Model of Female Labour Supply," Rev. Econ. Stud.,Jan. 1980, 47(1), pp. 47-74. ; MACURDY, THOMAS AND KILLINGSWORTH, MARK. "Recent Theoretical and EmpiricalStudies of Labor Supply: A Partial Survey." Paper presented at Conference on the Labor Market,Magdalen College, Oxford University, 1979. HOAGLAND, G. WILLIAM. "The Effectivenessof Current TransferProgramsin Reducing Poverty."Paper presented at the MiddleburyCollege Conference on Welfare Reform, April 1980. HOFFMAN, SAUL AND PODDER, NRIPESH. "Income Inequality,"in Five thousandAmericanfamiliespatterns of economic progress,Vol. 4. Edited by GREG J. DUNCAN AND JAMES N. MORGAN.Ann Arbor:Institute for Social Research, University of Michigan, 1976, pp. 333-56. HORIOKA,CHARLESY. "International Differences in Social Security and Saving: A Comparisonof the Barro and Feldstein Estimates,"J Public Econ., October 1980, 14(2), pp. 238-44. Hu, SHENGCHENG. "Social Security, the Supply of Labor, and Capital Accumulation," Amer. Econ. Rev., June 1979, 69(3), pp. 274-83. HUTCHENS, ROBERT M. "Changes in AFDC Tax 1025 Rates, 1967-1971," J Human Res., Winter 1978, 13(1), pp. 60-74. KASPER, HIRSCHEL. "Welfare Payments and Work Incentive: Some Determinants of the Rates of General Assistance Payments," J Human Res., Winter 1968, 3(1), pp. 86-110. KATONA, GEORGE. Private pensions and individual saving. Ann Arbor:Survey ResearchCenter, Institute for Social Research, University of Michigan, 1965. KATZ, ARNOLD AND OCHS, JACK. "Implications of Potential Duration Policies in Unemployment Compensation." Prepared for the National Commission on Unemployment Compensation, July 1980. KEELEY, MICHAEL C.; ROBINS, PHILIP K.; SPIEGELMAN, ROBERT G. AND WEST, RICHARD W. "The Estimation of Labor Supply Models Using Experimental Data," Amer. Econ. Rev., December 1978, 68(5), pp. 873-87. KIEFER, NICHOLAS M. AND NEUMANN, GEORGE R. "An Empirical Job-SearchModel, with a Test of the Constant Reservation-Wage Hypothesis," J Polit. Econ., February 1979, 87(1), pp. 89-107. KILLINGSWORTH, MARK R. "Must a Negative Income Tax Reduce Labor Supply? A Study of the Family'sAllocationof Time,"J Human Res.,Summer 1976, 11(3), pp. 354-65. KOTLIKOFF, LAURENCE J. "SocialSecurity and Equilibrium Capital Intensity," Quart. J Econ., May 1979a, 93(2), pp. 233-53. . "Testing the Theory of Social Security and Life Cycle Accumulation,"Amer. Econ. Rev.,June 1979b, 69(3), pp. 396-410. KRAFT, JOHN AND OLSEN, EDGAR. "The Distribution of Benefits from Public Housing,"in The distribution of economic well-being. Studies in Income and Wealth,Vol. 41. Edited by F. THOMAS JUSTER. Cambridge:NBER;distributedby Ballinger, 1977, pp. 51-64. KUZNETS, SIMON SMITH. "Sharesof Upper Income Groups in Income and Savings."National Bureau of Economic Research, Occasional Paper #35, 1953. LAMPMAN, ROBERT. "HowMuchDoes the American System of Transfers Benefit the Poor?" in Economic progress and social welfare. Edited by LEONARD H. GOODMAN. New York: Columbia University Press, 1966, pp. 125-57. . "Labor Supply and Social Welfare Benefits in the United States,"in Conceptsand data needs: Counting the laborforce. Appendix VolumeI. National Commission on Employment and Unemployment Statistics. Washington, D.C.: U.S. Government Printing Office, 1978. ____ AND SMEEDING, TIMOTHY. "On the Conversion of Inter-family Transfersinto Governmental This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1026 Journal of Economic Literature, Vol. XIX (September 1981) Transfers to Persons." Unpublished paper. Madison, Wis.:Institute for Research on Poverty, 1980. LAYARD,RICHARD."On Measuring the Redistribu- tion of Lifetime Income,"in Theeconomicsof public services:Proceedings of a conference held by the International Economic Association at Turin, Italy. Edited by MARTINS. FELDSTEINAND ROBERT P. INMAN.London: MacMillan, 1977, pp. 45- 72. LEBERGOTr,STANLEY.The American economy: In- come, wealth and want. Princeton, NJ.: Princeton University Press, 1976. LEIMER,DEAN AND LESNOY,SELIG."Social Security and Private Saving:A'Reexaminationof the Time Series Evidence Using Alternative Social Security Wealth Variables."Working paper. Office of Research and Statistics, Social Security Administration, Washington,D.C., 1980. AND LESNOY, SELIG. "Social Security, In- duced Retirement, and Aggregate Capital Accumulation:A Correctionand Updating:Comment." Washington,D.C.: Social Security Administration, Office of Research and Statistics, 1981. LEONARD,JONATHAN."The Social Security Disabil- ity Program and Labor Force Participation."National Bureau of Economic Research WorkingPaper No. 392, 1979. LEvY, FRANK. "How Big is the American Under- class?"Workingpaper. Graduate School of Public Policy, University of California, Berkeley, June 1976. . "The Labor Supply of Female Households Heads, or AFDC WorkIncentives Don't WorkToo Well,"J Human Res., Winter 1979, 14(1),pp. 7697. LEWIS, H. GREGG."Comments on Selectivity Biases in Wage Comparisons,"J Polit. Econ., Nov./Dec. 1974, 82(6), pp. 1145-55. LILLARD,LEE A. "Inequality: Earnings vs. Human Wealth,"Amer. Econ. Rev., March 1977, 67(2),pp. 42-53. LINDBECK,ASSAR."Work Incentives in the Welfare State." Institute for InternationalEconomic Studies, University of Stockholm, 1980. LuFr, HAROLDS. "The Impact of Poor Health on Earnings," Rev. Econ. Statist., Feb. 1975, 57(1), pp. 43-57. LURIE, IRENE."Estimates of Tax Rates in the AFDC Program,"Nat. TaxJ, March 1974, 27(1), pp. 93111. MAcDONALD, MAURICE.Food, stamps and income maintenance. New York: Academic Press, 1977. _____AND SAWHILL,ISABELV. "Welfare Policy and the Family,"Public Policy, Winter 1978, 26(1),pp. 89-119. MACURDY,THOMASE. "An Intertemporal Analysis of Taxation and Work Disincentives: An Analysis of the Denver Income MaintenanceExperiment." National Bureau of Economic Research Working Paper No. 624, Jan. 1981. MARSTON, STEPHEN T. "Voluntary Unemployment," in Unemployment compensation: Studies and research,Vol. 2. National Commissionon Unemployment Compensation,July 1980, pp. 43138. MASTERS, STANLEY H. AND GARFINKEL, IRWIN. Estimating the labor supply effects of income maintenance alternatives. New York:Academic Press, 1977. MEERMAN, JACOB. "Do EmpiricalStudies of Budget Incidence Make Sense?," Public Finance, 1978, 33(3), pp. 295-313. MENCHIK, PAUL. "The Importance of Material Inheritance: The Financial Link Between Generations,"in Modelling the distributionand intergenerational transmission of wealth. Studies in Income and Wealth, Vol. 46. Edited by JAMES D. SMITH. Chicago:Universityof ChicagoPress, 1980, pp. 159-85. . "Some Issues in the Measurement of Income Inequality," in Valuejudgments and income distribution. Edited by R. SOLO AND C. ANDERSEN. New York:Praeger 1981 forthcoming. ____ AND DAVID, MARTIN. "The Effect of Income Distribution and Redistributionon Lifetime Saving and Bequests." Discussion paper No. 582-79, Institute for Research on Poverty, Madison,Wis., 1979. MODIGLIANI, FRANCO AND BRUMBERG, RICHARD. "Utility Analysis and the Consumption Function: An Interpretation of Cross-SectionData," in PostKeynesian economics. Edited by KENNETH K. KURIHARA. London: Allen & Unwin, 1955, pp. 388-436. MOFFITT, ROBERT. "The IndividualWorkIncentives and Labor Market Impacts of Transfer Programs with a Work Requirement: The Static Theory." Unpublished paper. New Brunswick,N.J.:Rutgers University, 1978. . "An Economic Model of Welfare Stigma." Unpublished paper. New Brunswick,NJ.: Rutgers University, 1980a. . "Disequilibriumin LaborSupply:The Effect of the Negative Income TaxExperimentson Hours of Work." Unpublished paper. New Brunswick, N.J.:Rutgers University, 1980b. ____ AND KEHRER, KENNETH. "The Effect of Tax and TransferProgramson LaborSupply:The Evidence from the Income Maintenance Experiments," in Research on labor economics. Edited by RONALD EHRENBERG. Greenwich, Conn.:JAI Press, 1981, forthcoming. ____ AND NICHOLSON, WALTER. "The Effect of Unemployment Insurance on Unemployment: This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions Danziger, Haveman and Plotnick: Transfers and Their Effects The Case of Federal Supplemental Benefits,"Rev. Econ. Stat., 1982, forthcoming. MOON, MARILYN L. The measurement of economic welfare: Its application to the aged. New York: Academic Press, 1977. AND SMOLENSKY, EUGENE, eds. Improving measuresof economic well-being. New York:Academic Press, 1977. MORGAN, JAMES N.; DICKINSON, KATHERINE; DICKINSON, JONATHAN, et al. Five thousandAmerican families-patterns of economic progress,Vol.I. Ann Arbor:Institute for Social Research, 1974. AND SMITH, JAMES D. "Measuresof Economic Well-Offnessand Their Correlates,"Amer. Econ. Rev., May 1969, 59(2), pp. 450-62. MORTENSEN, DALE T. "Unemployment Insurance and Job Search Decisions," Ind. Lab. Relat. Rev., July 1977, 30(4), pp. 505-17. MUNNELL, ALICIA H. "The Impact of SocialSecurity on PersonalSavings,"Nat. TaxJ, Dec. 1974, 27(4), pp. 553-67. MURRAY, MICHAEL P. "RealVersusMonetaryTransfers: Lessonsfrom the American Experience."Unpublishedpaper given at the InternationalSymposium on Alternative Measures for Improving the Efficiency and Effectiveness of Public Transfer Policy, University of Augsburg, Germany, 1980a. . "A Reinterpretation of the Traditional In- come-Leisure Model, With Applicationto In-Kind Subsidy Programs,"I Public Econ., Aug. 1980b, 14(1), pp. 69-81. MUSGROVE, PHILIP. "Income Distribution and the Aggregate ConsumptionFunction,"j Polit. Econ., June 1980, 88(3), pp. 504-25. NEWTON, FLOYD C. AND ROSEN, HARVEY S. "Unemployment Insurance, Income Taxation,and Duration of Unemployment: Evidence from Georgia," Southern Econ. J, Jan. 1979, 45(3), pp. 773-84. NICKELL, STEPHEN. "Estimating the Probability of Leaving Unemployment," Econometrica, Sept. 1979, 47(5), pp. 1249-66. OKNER, BENJAMIN A. "Transfer Payments: Their Distribution and Role in Reducing Poverty," in Redistributionto the rich and the poor:Thegrants economicsof income distribution. Edited by KENNETH E. BOULDING AND MARTIN PFAFF. Belmont, Calif.:Wadsworth, 1972, pp. 62-76. OKUN, ARTHUR M. Equality and efficiency, the big tradeoff:Washington:The Brookings Institution, 1975. PAGLIN, MORTON. "The Measurement and Trend of Inequality:A Basic Revision,"Amer.Econ. Rev., Sept. 1975, 65(4), pp. 598-609. . Poverty and transfers in-kind: A reevaluation of poverty in the United States. Stanford, Calif.: Hoover Institution Press, 1980. PARSONS, DONALD. "The Decline in Male Labor 1027 Force Participation,"J Polit. Econ., Feb. 1980a, 88(1), pp. 117-34. ._ "Racial Trends in Male Labor Force Partici- pation," Amer. Econ. Rev., Dec. 1980b, 70(5),pp. 911-20. PECHMAN, JOSEPH A; AARON, HENRY J. AND TAUSSIG, MICHAEL K. Social Security:Perspectivesfor reform. Washington: The Brookings Institution, 1968. PELLECHIO, ANTHONY. "The Effect of Social Security on Retirement." NationalBureauof Economic Research Working Paper No. 260, July 1978a. _ . "The Social Security Earnings Test, Labor Supply Distortions and Foregone PayrollTax Revenue." National Bureau of Economic Research Working Paper No. 272, Aug. 1978b. . "The Estimation of Labor Supply Over Kinked Budget Constraints:Some New Econometric Methodology." National Bureau of Economic Research Working Paper No. 387, 1979. PLOTNICK, ROBERT D. "The Redistributive Effects of Income TransferPrograms:A Better Measure." Unpublished paper. Madison, Wis.: Institute for Research on Poverty, 1980. . "A Measure of Horizontal Inequity," Rev. Econ. Stat., May 1981, 63(2), pp. 283-88. AND SKIDMORE, FELICITY. Progress against poverty. New York:Academic Press, 1975. QUINN, JOSEPH F. "MicroeconomicDeterminants of Early Retirement: A Cross-Sectional View of White Married Men," J Human Res., Summer 1977, 12(3), pp. 329-46. REA, SAMUEL A., JR. "Incentive Effects of Alternative Negative Income Tax Plans,"J. Public Econ., Aug. 1974, 3(3), pp. 237-49. REIMERS, CORDELIA K. W. The timing of retirement of American men. Unpublished Ph.D. Dissertation, Columbia University, 1977. REYNOLDS, MORGAN AND SMOLENSKY, EUGENE. Public expenditures, taxes, and the distribution of income: The United States, 1950, 1961, 1970. New York:Academic Press, 1977. AND SMOLENSKY, EUGENE. "The Fading Effect of Government on Inequality," Challenge, July/Aug. 1978, 21(3), pp. 32-37. ROBINS, PHILIP; WEST, RICHARD AND LOHRER, MI- CHAEL."LaborSupply Response to a Nationwide Negative Income Tax: Evidence from the Seattle and Denver Income Maintenance Experiments." Unpublished paper. Stanford, Calif.:SRI International, Sept. 1980. SAKS, DANIEL H. Public assistance for mothers in an urban labor market.Princeton, N.J.:Industrial Relations Section, Princeton University, 1975. SCHEFFLER, RICHARD M. AND IDEN, GEORGE. "The Effect of Disability on Labor Supply," Ind. Lab. Relat. Rev., October 1974, 28(1), pp. 122-32. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions 1028 Journal of Economic Literature, Vol. XIX (September 1981) "AModel of SocialSecurityand Retirement Decisions,"J. Public Econ., Dec. 1978, SHESHINSKI, EYTAN. distribution of economic welfare. Princeton, N.J.: Industrial Relations Section, Princeton University, 1973. 10(3), pp. 337-60. B. "Labor Force Participation of Men, 25-54, by Race," Mon. Lab. Rev., July SISKIND, FREDERIC 1975, 98(7), pp. 40-42. Measuring the economic welfare of low income households and the antipoverty effectiveness of cash and non-cash transfers programs. Unpublished Ph.D. thesis, Uni- SMEEDING, TIMOTHY M. versity of Wisconsin, Madison, Department of Economics, 1975. . "The Antipoverty Effectiveness of In-Kind Transfers,"J Human Res., Summer 1977, 12(3), pp. 360-78. . "On the Distribution of Net Income: Com- _____AND DANZIGER, SHELDON. Conferenceon the trend in income inequality. Madison, Wis.: Institute for Research on Poverty, Special Report LI, 1976. THEIL, HENRI. Statistical decomposition analysis. Amsterdam: North-Holland, 1972. TINBERGEN, JAN. Income distribution:Analysis and policies. Amsterdam: North-Holland, 1975. U.S. BUREAU OF THE CENSUS. "Money Incoine in 1976 of Families and Persons in the United States," in Currentpopulation reports,Series P-60, #114. ment," SouthernEcon. J, January1979, 45(3),pp. Washington, D.C.: U.S. Government Printing Office, 1978. U.S. DEPARTMENT OF COMMERCE. The budget of 932-44. . "The Antipoverty Effects of In-Kind Transfers: A 'Good Idea' Gone Too Far?" Policy Studies Washington, D.C.: U.S. Government Printing Office, 1980. Journal, forthcoming, 1981. AND MOON, MARILYN. "ValuingGovernment the United States Government,fiscal year 1981. U.S. DEPARTMENT OF HEALTH, EDUCATION AND WELFARE. The cyclical behavior of income trans- Expenditures: The Case of Medical Care Transfers fer programs:A case study of the currentrecession. and Poverty," Rev. Income Wealth, Sept. 1980, Office of Income Security Policy, Technical Analysis Paper No. 7, 1975. 26(3), pp. 305-324. P., ed. Female labor supply: Theory and estimation.Princeton, N.J.:Princeton Univer- SMITH, JAMES, sity Press, 1980. . "Assets and Labor Supply," in Female labor supply: Theoryand estimation. Edited by JAMES P. SMITH Princeton, N.J.: Princeton University Press, 1980, pp. 166-205. SMOLENSKY, SCHMUNDT, EUGENE; STIEFEL, LEANNA; MARIA AND PLOTNICK, ROBERT. "Adding In-Kind Transfers to the Personal Income and Outlay Account: Implications for the Size Dis- tribution of Income." in The distribution of economic well-being. Studies in Income and Wealth, Vol. 41. Edited by F. THOMAS JUSTER. Cambridge, Mass.: Ballinger, 1977, pp. 9-44. SOLON, GARY. "Labor Supply Effects of Extended Unemployment Benefits," J Human Res., Spring 1979, 14(2), pp. 247-55. "More on Unemployment Insurance as Insurance,"Ind. Lab. Relat. Rev.,July STAFFORD, FRANK P. 1977, 30(4), pp. 521-26. TAUSSIG, MICHAEL K. Alternative measures of the VAN DER GAAG, JACQUES AND SMOLENSKY, EU- GENE. "Income, Consumption, True Equivalence Scales and Characteristics of the Poor." Unpublished paper. Madison, Wis.: Institute for Research on Poverty, 1980. WALES, T. J. AND WOODLAND, ALAN D. "Labour Supply and Progressive Taxes," Rev. Econ. Stud., Jan. 1979, 46(1), pp. 83-95. WATTS, HAROLD AND PECK, KIM. "On the Comparison of Income Redistribution Plans," in The personal distribution of income and wealth. Studies in Income and Wealth, Vol. 39. Edited by JAMES D. SMITH. New York: NBER; distributed by Columbia University Press, 1975, pp. 75-118. WELCH, FINIS. "WhatHave We Learned from Empirical Studies of Unemployment Insurance?," Ind. Lab. Relat. Rev., July 1977, 30(4), pp. 45161. WILLIAMS, ROBERT. Public assistance and work effort. Princeton, Princeton N.J.: Industrial Relations Section, University, 1975. This content downloaded from 192.75.12.3 on Wed, 16 Dec 2015 21:11:11 UTC All use subject to JSTOR Terms and Conditions