Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 177
7 The Effect of Welfare on Child Outcomes Janet Currie There is broad support for the idea that welfare should benefit poor children. Yet most research on welfare programs, as well as much of the debate about welfare reform, has focused on the way that parents respond to incentives created by welfare, rather than on its effects on children. Less work has been devoted to the fundamental question of whether any of the web of programs supporting poor families benefit children. If it can be shown that they do, then there are many other questions to be addressed: First, are the benefits short or long term? Second, which types of programs or combinations of programs are most effective; for example, do cash or in-kind programs produce bigger benefits for children? Third, do welfare programs have different effects on different groups, and if so why? Fourth, how exactly do successful programs work? And finally, can efficacious programs pass the more stringent test of cost-effectiveness? This review focuses on the eight large federal programs shown in Table 7-1: Aid to Families with Dependent Children (AFDC), which has been replaced with the new Temporary Aid for Needy Families program (TANF); the Earned In- come Tax Credit (EITC); housing assistance; Food Stamps; the Supplemental Feeding Program for Women, Infants, and Children (WIC); school nutrition pro- grams; Medicaid; and Head Start. The programs are evaluated with respect to their effects on the health and educational achievement of children. Where possible, documented effects on long-term outcomes are noted. The first section of this chapter is a brief discussion of how we know what we know about these programs. The evidence regarding the effects of cash programs and in-kind programs, respectively, is then reviewed in the next two sections. 177
OCR for page 178
178 THE EFFECT OF WELFARE ON CHILD OUTCOMES TABLE 7-1 Trends in Program Expenditures (billion 1995 $) 1975 1980 1990 1995 Cash Transfers AFDC Total 23.8 21.8 21.822.0 Federal only 13.1 11.8 11.912.0 Earned Income Tax Credit Total 3.4 3.7 8.122.2 Refunded portion of credit 2.5 2.6 6.219.0 In-Kind Transfersa Housing assistance 7.0 10.0 18.223.7 Food Stamps 13.5 17.4 19.425.7 WIC 0.7 1.4 2.53.5 School nutrition School lunch 5.4 5.8 4.35.3 School breakfast 0.4 0.5 0.71.2 Medicaid Totalb 35.1 46.3 76.3111.2 Federal only 20.1 27.0 47.886.6 To dependent children 6.3 6.0 10.717.8 To adults in families with dependent children 5.9 6.4 10.114.0 Head Start 1.1 1.3 1.93.5 aAll but the Food Stamps figure for 1975 are actually from 1972. bThe Medicaid figures for 1980 are actually from 1981. SOURCE: U.S. House of Representatives (1993, 1994, 1996). The evidence indicates that contrary to much current publicity, the system is not entirely "broken" when judged using the metric of child well-being: there are specific programs that produce important benefits for children. Nevertheless, not all programs are equally effective, and benefits are not equally distributed across children. Hence, a review of what we know about these programs can provide a useful starting point for welfare reform, as well as highlighting gaps in what we need to know in order to carry out intelligent reform. The last section of the paper discusses fruitful directions for future research and the importance of enhanced data collection efforts. HOW WE KNOW WHAT WE KNOW A comprehensive review of the program evaluation literature is far beyond the scope of this chapter. However, since several different methods are used in the studies discussed here, some comment on methodology is in order. A some- what fuller, nontechnical discussion can be found in Currie (1995a) or Heckman (1990).
OCR for page 179
JANET CURRIE 179 The fundamental problem facing researchers and policy makers is that the children of welfare recipients may have bad outcomes for reasons that have nothing to do with the receipt of assistance per se. It is possible that a program could have substantial benefits for poor children and still leave many children disadvantaged relative to better-off peers. Evidently, parents of children on welfare are worse off than other parents in observable ways: they are poorer, likely to have less education, and may also have health problems. Many datasets available to researchers contain at least crude measures of these observable variables so that observed differences be- tween parents on welfare and other parents can be accounted for using standard regression models. To take a simple example, suppose that children of high school dropouts have lower scores on standardized tests than children of college graduates. Then if mothers on welfare are more likely to be high school dropouts than college graduates, a simple comparison of the two group' s average scores might tell you more about the effects of maternal education than about the effects of welfare. A simple way to "control" for the effects of education in order to focus on the effects of welfare might involve drawing a sample of high school dropouts and comparing children of welfare mothers to other children within this group. Any differences between the welfare children and the others could then be attributed to welfare use and not to maternal education. Multiple regression techniques simply allow one to control for the effects of several observable variables at the same time. The problem becomes much more difficult however if parents on welfare also differ from other parents in ways that are not observed. For example, they may lack motivation or be discouraged by previous misfortune. Failure to prop- erly control for these differences could lead one to incorrectly infer that it was being on welfare that was associated with negative child outcomes, rather than these underlying conditions. Some underlying problem, such as maternal depres- sion, might cause both welfare dependence and negative child outcomes. There are basically two approaches to this issue of unobserved characteris- tics. First, one may design a social experiment, randomly assigning eligibles to a "treatment" group and a "control" group. Random assignment ensures that, on average, the two groups will have the same observed and unobserved characteris- tics. In principle, one can then assess the effect of the treatment simply by comparing mean outcomes for the two groups, just as one would do in a drug trial. The key advantage of an experimental evaluation is its transparency. One disadvantage of social experiments is that they may be very expensive. But there are several disadvantages in addition to high cost (Heckman, 1990~. These include differential attrition between treatments and controls (which causes the treatment group to become less and less like the comparison group over time); the fact that subjects assigned to the control group may not accept their fate passively (for example, subjects denied training in a government program might
OCR for page 180
180 THE EFFECT OF WELFARE ON CHILD OUTCOMES sign up for an alternative program); and the fact that it may be difficult to use the experiment to examine differential effects of the treatment on different groups. Nonexperimental evaluations attempt to control statistically for unobserved variables associated both with participation in the program and with the outcome of interest. One method of doing this is to find a third set of variables, called "instruments," that are associated with participation in the program but not with the important unobserved variables. For example, a researcher interested in the effects of participation in Medicaid on child health might argue that the generos- ity of state AFDC benefits is associated with participation in Medicaid because of the link between AFDC recipiency and Medicaid eligibility, but that the level of AFDC benefits does not have any effect on child health other than through its effect on participation in Medicaid. If this assumption were true, then the level of AFDC benefits would qualify as an "instrumental variable." This instrument would be used (along with other observable characteristics of the mother) to predict Medicaid participation, and predicted participation would be substituted for actual participation in the model explaining child health. The idea is that predicted participation will depend only on observable characteristics and differences in state AFDC benefit levels, and not on the unobserved charac- teristics of the mother. The procedure is analogous to an experiment in which AFDC benefit levels are varied across states, Medicaid participation responds, and only this source of variation in participation rates is used to identify the effects of Medicaid on health. The difficulty with instrumental variables techniques is that the key assump- tions may not be satisfied. Suppose that states with more generous AFDC ben- efits also have higher-income populations and that higher incomes are associated with better child health. Then unless one takes account of this relationship, one will tend to find a spurious positive relationship between participation in Medic- aid and child health. Alternatively, suppose that states raise AFDC benefit levels in response to poor child health. Then one might observe a spurious negative relationship between predicted Medicaid participation and child health. An alternative approach involves assuming that the relevant omitted charac- teristics are fixed within a family or for the same child over time. Suppose for example that the relevant unobserved variable is maternal attitudes towards edu- cation and that this remains fixed over some period of time. Suppose further that one sibling participated in Head Start and one did not. Then comparing the sibling who participated to the one that did not provides a measure of the effect of Head Start that is not affected by the fact that, on average, mothers of Head Start children may have more positive (or negative?) views of education than other similarly situated mothers. Of course, the problem with this approach is that the relevant variable may not be fixed within households or over time. The studies discussed below all rely on one of these methodological ap- proaches. Their conclusions are only as valid as the assumptions underlying the chosen approach. It is in cases where the same result has been obtained using
OCR for page 181
JANET CURRIE 181 different assumptions and data sources that we can be most confident of the conclusions. WHAT WE KNOW ABOUT CASH PROGRAMS Aid to Families with Dependent Children The term "welfare" has usually been identified with the Aid to Families with Dependent Children program. This oldest and largest of the federal welfare programs provided cash transfers to (predominantly female-headed) families with children. This is the program that recent welfare reforms (the Personal Respon- sibility and Work Opportunity Reconciliation Act of 1996 [PRWORA]) effec- tively ended, replacing it with the new Temporary Aid for Needy Families pro- gram. TANF differs from AFDC because it ends the "entitlement" of all needy families to welfare benefits, because it introduces time limits on welfare benefits and because it provides states with much more latitude in developing their own welfare programs. Nevertheless, since most of what we know about cash welfare programs comes from studies of AFDC, and because many states will respond to TANF by only gradually altering their AFDC programs, it is of interest to sum- marize this literature here. Like TANF, AFDC was administered at the state level within federal guide- lines. As a result, program characteristics varied widely from state to state. For example, as of January 1993, the maximum monthly AFDC grant for a one- parent family of four persons varied from $164 in Alabama to $923 in Alaska (U.S. House of Representatives, 1993~. On average the federal government pays 54 percent of benefit costs, as shown in Table 7- 1. The continuous erosion of real AFDC benefit levels over the past 15 years provides compelling evidence of the unpopularity of this program: the average monthly AFDC benefit declined from $483 (1993 dollars) in 1980 to $373 in 1993, even though the average family size remained constant at three persons (U.S. House of Representatives, 1994~. One of the problems involved in evaluating the effects of AFDC on children is that the benefits of a cash transfer program can be expected to be diffuse. Small increases in household expenditures on a wide range of items may produce overall benefits for children without affecting any one indicator a great deal. A second problem is that although income is often used as a shorthand summary of a household's socioeconomic status, it is in practice extremely difficult to sepa- rate the effects of income from the effects of other family background character- istics including neighborhoods (Mayer, 1996~. Most research about the effects of AFDC on children focuses on the fact that daughters of women who participate in AFDC are themselves more likely to participate (cf. Gottschalk, 1990; Murray, 1984~. What is less clear is whether the relationship is causal or whether it merely reflects the fact that the children of the poor are more likely to be poor older studies tended to conclude that the
OCR for page 182
182 THE EFFECT OF WELFARE ON CHILD OUTCOMES relationship was not causal, but studies using more recent data have questioned this conclusion. See Moffitt (1992) for a fuller discussion of this issue. There has been comparatively little research linking maternal AFDC partici- pation to other child outcomes, but the empirical issues are the same. First, it is necessary to control for some measure of income as well as for AFDC status since otherwise the estimated effects of participation are likely to reflect the relative poverty of AFDC mothers. Second, within the group of poor women, one would like to control for the fact that women choose whether or not to go onto AFDC. Blank and Ruggles (1996) show that only 60 percent of eligible women actually take up welfare benefits. Those who do are likely to differ from those who do not in many unobservable respects. Hill and O'Neill (1994) find that, when instrumental variables methods are used to take account of unobserved variables that might be correlated with AFDC status, AFDC participation has no effect on children's scores on a standardized test of vocabulary, given income. Currie (1995a) confirms that their results hold up even when sibling comparisons are used to account for unobserved maternal background characteristics. Currie and Cole (1993) use data from the 1979 to 1988 waves of the National Longitudinal Survey of Youth (NLSY) to examine the effect of AFDC participation during pregnancy on the utilization of prenatal care and birthweight. They use both sibling comparisons and instrumental vari- ables methods to take account of unobserved variables that might be correlated with both participation in the AFDC program and outcomes,] and find that AFDC participation has no additional significant effect on birthweight given income. Together, these studies suggest that income from AFDC has much the same effect on children as family income from any other source. The Earned Income Tax Credit: A Comparison to the Negative Income Tax The slack in the growth of AFDC payments over time has been taken up by the growth in expenditures on the Earned Income Tax Credit, which doubled between 1975 and 1990. The EITC was introduced in 1975 as a means of granting tax relief to low-income tax payers. Because it is administered through the tax system, the EITC is not always viewed as a welfare program. However, unlike most tax credits, the EITC is "refundable," that is, if the amount of the credit exceeds the taxpayer's federal income tax liability, then the difference is refunded. Table 7-1 shows that, in fact, most EITC expenditures are outlays of this kind rather than forgone tax dollars. The EITC differs from traditional cash welfare programs primarily because the majority of recipients work and benefits are available to all kinds of families. Thus, it creates fewer perverse incentives than AFDC. 1They instrument AFDC participation using state-level variation in program characteristics.
OCR for page 183
JANET CURRIE 183 If it is difficult to identify the effects of cash transfers under AFDC, the problems involved in identifying the effects of the EITC are even more formi- dable. The fundamental problem is that the amount of the credit depends on the parents' earnings, and earnings are likely to reflect many unobserved factors relevant to child well-being. However, the EITC is in many respects similar to the negative income tax (NIT), an income guarantee program that was subjected to exhaustive scrutiny through four large-scale social experiments, although it was never implemented.2 The four experiments were conducted in New Jersey and Pennsylvania; Seattle and Denver; Gary, Indiana; and rural areas of North Carolina and Iowa. It is important to note that the North Carolina and Gary samples were much poorer than the others. The income guarantees paid out under the NIT program were large relative to cash transfers that have been made under the EITC. The average payments in the Seattle-Denver experiment, for example, ranged from $919 to $2,031 (1972 dollars), depending on the treatment group. By way of comparison, the poverty line for a family of three persons was $3,099 in 1972. In 1992, the maximum EITC was $1,384 and the poverty line $11,280. Since NIT participants were randomly assigned to "treatment" and "control" groups, the NIT experiments provide a unique opportunity to assess the effects of income transfers per se on the well-being of children in poor families. Despite the large transfers, findings about the effects of the NIT are inconsis- tent across studies and experimental populations. In addition, econometric esti- mates are sometimes at odds with those derived from simple comparisons of treatments and controls. For example, Kehrer and Wolin (1979) find that the mean birthweight of infants born to the treatment group in the Gary experiment was actually lower than the birthweight of the controls. Yet estimates from their structural model suggest that the infants of treatments had higher birthweights in 9 out of 12 maternal age groups. O'Conner et al. (1976) examine the effect of the NIT on child nutrition using data from the rural experiment. Among subjects in North Carolina, they found positive and significant treatment effects on nutrient intakes. However, the treat- ment did not appear to have any significant effect in Iowa, a finding that the authors attribute to the relative poverty of the North Carolina sample. 2Under a NIT, a family that earns no income is guaranteed a minimum income G. Families with earnings Y receive a payment D, where D = G - to Y. The quantity B = G/t~ is referred to as the break- even level of income since workers who earn more than B receive no payments. If income is equal to the wage multiplied by hours worked, and workers face a tax rate t, then workers on the NIT earn w (1 - t- to) for every hour of work, whereas workers with incomes above B earn w(1 - I). That is, workers on the NIT face a higher tax rate. The EITC differs from the NIT in that the EITC has no income guarantee. Also, since at first the size of the credit increases with earnings, the EITC lowers effective marginal tax rates for the poorest rather than raising them. After a certain level of income, the credit begins to be phased out, creating a higher implicit tax rate.
OCR for page 184
184 THE EFFECT OF WELFARE ON CHILD OUTCOMES Maynard and Crawford (1976) found that elementary school children from NIT families in North Carolina showed statistically significant improvements in attendance, standardized tests, and grades. However, there were no effects for elementary school children in Iowa. Once again, this pattern of results is attrib- uted to the fact that the children in North Carolina were more disadvantaged than those in Iowa. Maynard and Murnane (1979) found that in the Gary experiment the NIT treatment had positive effects on reading scores of young children but that these effects were statistically significant only among children whose fami- lies had been in the program for 3 or more years. Finally, in an analysis of data from the New Jersey experiment, Mallar (1977) found that teenagers whose parents were enrolled in NIT were 20 percent to 90 percent more likely to complete high school depending on the NIT plan. However, Venti (1984) found only an 11 percent increase in the probability of completing high school for youth in the Seattle-Denver experiment. This lower estimate seems more probable in view of the relatively short duration of the experiments and the many long-term factors (such as achievement in early grades) that have been linked to educational attainment. These results may also be related to the fact that, in all four experiments, youths in treatment households were less likely to be employed than controls (Robins, 1985~. These studies suggest that the relatively large income transfers made to families under the NIT had a positive effect on the nutritional status and educa- tional attainment of children in the poorest families. However, the magnitudes vary greatly from study to study. Perhaps unsurprisingly, studies of the effects of the NIT on consumption also show that families spent much of the subsidy on goods that may not have been directly related to the well-being of their children. For example, the NIT appears to have had a negative effect on the labor supply of married women,3 and positive effects on housing expenditures and purchases of consumer durables (Robins, 1985; Michael, 1978~.4 WHAT WE KNOW ABOUT IN-KIND PROGRAMS A parallel "in-kind" welfare system has grown up alongside the cash system. This system aims to directly provide for a child's "basic needs": decent housing, food, medical care, and quality early education. Table 7-1 shows that expendi- tures on virtually all of these programs have shown steady growth over time (the exception being the School Lunch Program). Table 7-2 indicates that in contrast to stagnant AFDC caseloads, caseloads for most in-kind programs have been . . Increasing. 3No convincing evidence of a link between maternal employment and children's well-being has been found. See Blau and Grossberg (1990) and Desai et al. (1989). 4The NIT may also have increased the probability of marital dissolution, although this finding remains controversial (cf. Cain and Wissoker, 1990; Hannan and Tuma, 1990).
OCR for page 185
JANET CURRIE TABLE 7-2 Trends in Caseloads (millions) 185 1975 1980 1990 1995 Cash Transfers AFDC Total recipients11.1 10.6 11.5 13.6 Child recipients7.8 7.2 7.8 9.3 Earned Income Tax Credit No. of families6.2 7.0 12.6 17.4 In-Kind Transfers Housing Assistance No. of households3.2a 4.0 5.4 5.8 Food stamps Total recipients16.3 19.2 20.0 26.6 WIC No. of womeno.2a 0.4 1.0 1.6 No. of infantso.2a 0.5 1.4 1.8 No. of children0.5a 1.0 2.1 3.5 School nutrition School lunch No. any meals26.3a 26.6 24.1 25.6 No. free meals10.5a 10.0 10.3 12.4 School Breakfast2.5a 3.6 4.0 6.3 No. free meals2.oa 2.8 3.3 5.1 Medicaid Total recipients22.0 21.6 25.3 35.1 Child recipients9.6 9.3 11.2 17.2 Head start0.3 0.4 0.5 0-7 aThese figures are for 1977. SOURCE: u.s. House of Representatives (1993,1994,1996). Initial evaluation of these in-kind programs is more straightforward than the evaluation of cash transfer programs because we can ask whether the program has an impact on the specific child outcome it was designed to affect. For example, we can ask whether receipt of housing assistance is associated with improvements in housing or whether household participation in the Food Stamps program improves a child' s diet. We might then wish to ask whether the program has additional effects on related child outcomes. For example, better nutrition could influence a child's cognitive abilities. Also, subsidies to food and housing may influence child outcomes more generally by relaxing the family's budget constraint (see Moffitt, 1989, and Citro and Michael, 1995, for discussions of the valuation of in-kind benefits).5 However, since the effects of income transfers are discussed above, 5The National Research Council (vitro and Michael, 1995) concludes that for simplicity~s sake, ``near-cash,, benefits such as Food stamps and housing assistance should be counted at their dollar
OCR for page 186
186 THE EFFECT OF WELFARE ON CHILD OUTCOMES the focus in this section is on any effects of participation in in-kind programs on the specific outcomes that the programs were designed to affect. In practice, this restriction eliminates very few studies from consideration.6 Housing Assistance In contrast to AFDC and Food Stamps, housing assistance is not an entitle- ment: when funds allocated to the program run out, people who are eligible must be wait-listed. It is estimated that about half of federal expenditures on housing assistance directly benefit children, while the elderly are the other large group of beneficiaries. Most expenditures are on rental assistance programs rather than on low-rent public housing (which is what many people think of as "public housings. And since 1982, most new authorizations for rental housing assistance have been for Section 8 programs (Pedone, 1988~. The Section 8 existing housing program provides rent subsidies to families who find an apartment of their own choosing, as long as the rent is below the "Fair Market Rent" established by the Depart- ment of Housing and Urban Development (HUD) and the unit meets minimum quality standards. Rental assistance typically reduces a family's rental pay- ments to 30 percent of its income, after deductions for certain expenses are taken into account. Deficient housing is hazardous to children. For example, lead poisoning is three times more common among poor children than among nonpoor children and is directly related to housing conditions. The risk of accidental death is also three times higher for poor children, and some of this increased risk may be due to hazards in the home (Starfield, 1985~. In 1989,18 percent of poor households (2.2 million households) lived in housing with severe or moderate physical prob- lems compared to 7 percent of nonpoor households.7 It is not known whether, in general, housing assistance enables families in deficient housing to move to adequate housing. A 1988 HUD study found that more than half of public housing households lived in projects that needed moder- ate to substantial rehabilitation just to meet HUD's own standards. The estimated cost of bringing these units up to standard would have exceeded $20 billion 1986 dollars (Lazere et al., 1991~. value when comparing the resources available to different households, and various procedures for valuing housing benefits are discussed. However, the panel also recommends that health insurance be excluded from these comparisons because it is too hard to come up with a meaningful estimate of its value to households in different circumstances. 6An exception that deserves mention is Meyers et al. (1993) who found that in a sample of poor children in Boston, those who received housing assistance were less likely to be anemic. The study did not control for selection into public housing. 7Problems that HUD classifies as severe include lack of basic plumbing facilities, serious heating breakdowns, and rat infestations. An example of a moderate deficiency is the use of unvented gas, oil, or kerosene heaters as primary heating equipment.
OCR for page 187
JANET CURRIE 187 Section 8 programs require families to locate a landlord willing to participate and to arrange with the landlord for inspections and repairs within a fixed period of time. One case study of 56 single mothers in eastern Massachusetts in 1985 and 1986 found that after waiting an average of 2 years to receive a certificate, 24 women returned them unused because they were unable to find housing that met program requirements within the allotted time (Mulroy, 1988~. On the other hand, there is some evidence that recipients of vouchers pay higher rent (Kennedy and Finkel, 1987; Apgar, 1990) and move to better neighborhoods (Johnson, 1986~. The often dismal social conditions in many public housing projects must be weighed against any improvements in the physical housing stock. However, it is very difficult to identify the effects of neighborhoods and schools because any relationship we observe between neighborhood characteristics and individual outcomes could reflect the characteristics of the individual or of his or her family that placed them in these neighborhoods in the first place. The Gautreaux program sheds light on this issue. Under the program, resi- dents in public housing projects can apply for Section 8 housing certificates and move to private apartments. Some apartments are in predominantly white sub- urbs, while others are in the inner city. Although the persons admitted to the program are not a random sample of public housing residents,8 Rosenbaum (Rosenbaum et al., 1986; Rosenbaum, 1992) asserts that the program assigns apartments in an approximately random manner, since people get whatever is available when they reach the top of the waiting list. He finds that 7 years after their move, children who had moved to the suburbs were 15 percent less likely to have dropped out of school, 16 percent more likely to be in a college-track program, and 34 percent more likely to be employed than those who had moved within the inner city. All of these differences are statistically significant at the 90 percent level of confidence. These findings suggest that voucher programs can have a positive effect on the life chances of children if they enable families to find housing in better neighborhoods. On the other hand, they suggest that the disamenities associated with large public housing projects may have significant negative effects. How- ever, the study is marred by high rates of attrition from the sample. HUD is currently conducting an experimental evaluation of a program similar to Gau- treaux in four cities.9 An experimental evaluation that took care to minimize attrition could shed great light on the possible beneficial effects of housing vouch- ers, and on the issue of the effects of neighborhoods more generally. Despite their bad reputations, housing projects may be better than much of 8Applicants are screened to make sure that they have paid their rent regularly and that they have adequate housekeeping abilities. The program does not serve families with more than four children because few large housing units are available in the suburbs. In addition, the act of applying for an apartment in an unknown location may indicate that a person is strongly motivated to improve his or her circumstances. 9Personal communication, Lawrence Katz, Department of Economics, Harvard University, 1997.
OCR for page 194
94 THE EFFECT OF WELFARE ON CHILD OUTCOMES is inappropriate, given that Head Start is intended to affect a range of outcomes (see McKey et al., 1985~. Evidence from the Perry Preschool Project, which found that program children were less likely to drop out of high school, engage in crime, or become pregnant as teenagers, is often cited. However, since the project included only 58 treatments and 65 controls, was funded at about twice the rate of a typical Head Start program, and did not involve a national sample, it is not clear that the findings generalize. Currie and Thomas (1995b) examine sibling comparisons from a national sample and find that children who were in Head Start have higher test scores at the end of the program than either stay-at-home siblings or siblings who went to other preschools. The effects are of the same magnitude for both black and white children and indicate that Head Start closes one-third of the gap between these children and others. But consistent with the experimental studies, they find that the effects on black children fade out rapidly. These results suggest that the positive effects of Head Start may be undermined by subsequent deprivation among these children. In contrast, the effects on the test scores of white children do not fade out. Moreover, white children 10 and over are significantly less likely to have re- peated a grade if they attended Head Start and are thus less likely to have experi- enced the age/grade delay that often leads to high school noncompletion. Both black and white children who attended Head Start were more likely to be immu- nized than stay-at-home siblings, although there was no effect on height-for-age, a measure of long-term nutritional status. In related work, Currie and Thomas (1996a) find that Head Start has large and lasting effects on the test scores of Latino students. A closer inspection of the data reveals that these positive effects are largest for Mexican-origin children and smallest for Puerto Rican children. However, due to sample size limitations it is not possible to sort out the effects of ethnicity and the effects of region. It is possible, for example, that the ethnic differences reflect differences in the pro- grams available in New York, where Puerto Rican children tend to be located, and California and Texas, where Mexican-origin children are concentrated, rather than any independent effect of ethnicity per se. Currie and Thomas (1996b) ask whether differences in school quality can explain differences in the pattern of "fadeout" in test scores between whites and blacks. Specifically, the initial positive effects of the Head Start program may be undermined if Head Start children were subsequently exposed to inferior schools. And since we see fadeout for blacks but not for whites, it would have to be the case that black Head Start children are attending worse schools than other black children but that the same was not true among whites. Currie and Thomas test this hypothesis using a sample of eighth graders from the National Educational Longitudinal Study of 1988 (NELS). Their work builds on earlier research by Lee and Loeb (1995) who showed, using these data, that the schools attended by Head Start children are of worse quality in some
OCR for page 195
JANET CURRIE 195 observable dimensions than the schools attended by other children. Even if family income and parent's education are controlled for, children who attended Head Start have lower test scores than other children. This result is to be ex- pected if Head Start does not entirely compensate for early disadvantages. However, among black children, the gap between Head Start children and other children is virtually eliminated when we compare children within the same school. That is, within schools, black Head Start children do no worse than other black children. But since they perform more poorly than other children on average, they must be attending schools in which all black children do badly. If a "quality" school is defined as one in which children do well, then these results suggests that black children who attend Head Start go on to attend schools of significantly worse quality than other black children. In contrast, among non- Hispanic white children there appears to be little difference in the schools at- tended by Head Start and other children. WHAT WE NEED TO KNOW The preceding discussion is summarized in Table 7-3. The table presents a matrix of programs and effects. Differences in the effects of programs across groups have been suppressed, although one theme that has emerged from the discussion so far is that they are important. The most striking feature of Table 7- 3 is that there are many empty cells we clearly need to learn a great deal more about the effects of welfare before we can make informed public policy. In some cases, research has been limited by lack of appropriate data. In others, existing information has not yet been fully exploited. This section highlights some unan- swered research questions and discusses the extent to which better data collection efforts could help. Effects of Welfare on Long-Run Outcomes Ultimately, what many people care about is whether investments in children today will produce productive, well-socialized adults tomorrow. However, Table 7-3 highlights the fact that little is known about the effects of welfare on long- term outcomes. Lack of data places major limitations on this type of research. Many important outcomes can only be examined 10 to 15 years after childhood participation in welfare programs. There are few existing datasets that combine information about childhood participation in welfare, other family background characteristics, and the outcomes of interest. One exception is the National Longitudinal Survey's Child-Mother file (NLSCM). The NLSCM contains information about the children of a sample of approximately 6,300 women who were between the ages of 14 and 21 in 1978. Information about childhood participation in AFDC, the Food Stamp Program, Medicaid, Head Start, and WIC is available. By the time the 1994 wave is
OCR for page 196
196 4= so _4 o 8 ~ V) z be a' a' · _4 o Cq a' Cq a' o Cq so be s°- a' a' o Cq C) a' I. . o VO v · ~ C) · · ~ ca o ca to ~ EVER E-4 ~ z v ¢ ¢ ~ EM ~ + + + - . - . - . + + + + + - ~- + + - - - . O + + + 1 o o o ~° ~ =~ ~ 0 ~ e ~ ~ ~ e e ~ .~ C) 11 - .^ 4= .~ 11 1 .^ 4= := ca O 11 + .^ 4= ~0 11 O En 0 z
OCR for page 197
JANET CURRIE 197 released, there will be more than 800 children over 16. Of course, since these children will have been born to young mothers, they will not be a nationally representative sample of 16 year olds. Still, this sample is a valuable resource. If future waves of the survey continue to be funded, it will grow in size and in representativeness and allow us to address many questions about the relation- ship between welfare and long-term outcomes such as schooling attainment, teen parenthood, and crime. A second exception is a special supplement to the Panel Study of Income Dynamic (PSID) that was fielded in 1995. This module contains retrospective information about early childhood education and criminal activity that can be linked to data about welfare participation from the original PSID file. The PSID is currently undertaking an even more ambitious data collection effort, the 1997 Child Development Supplement. The survey of 3,500 0 to 12-year-old children will have assessments of cognitive, behavioral, and health status. Data are being collected from the mother, a second caregiver, the absent parent (if relevant), teachers, school administrators, and the children themselves. The survey will also include time diaries for caregivers, children, and teachers, to examine inputs into child development. Finally, other inputs such as resources in the home and neighborhood will also be measured. Once again, this information can be linked to data about welfare participation from the main files, and follow-up on these children may help to identify long-term effects of participation. Fielding this type of supplement to existing data sources promises to be a cost-effective method of providing information on the link between the current outcomes of young adults and their participation in various programs as children. An additional issue that can be addressed is whether there are links between the short-term outcomes that have been examined in previous research and longer- term outcomes. If it is found that particular short-term outcomes are reliable "markers" for longer-term outcomes, then future evaluations of welfare programs may not require as much costly long-term follow-up of the participants. Why Do Effects Appear to Vary with Race, Ethnicity, and Natality? The PSID and NLSCM datasets will both support analyses stratified by race, ethnicity, and nasality. However, in many cases the sample sizes are very small. In order to properly document differences in outcomes, or even in utilization, it will be necessary to add questions to existing large-scale datasets. For example, the Census asks questions only about the use of cash welfare, even though expen- ditures on in-kind programs constitute the largest and fastest-growing share of the welfare bill. A second problem is that large-scale, individual-level datasets typically lack information about neighborhoods and administrative procedures that could be used to test specific hypotheses about group differences. For example, one might believe that black children on Medicaid receive fewer visits for illness than white
OCR for page 198
198 THE EFFECT OF WELFARE ON CHILD OUTCOMES children because the providers that serve them are overcrowded and it is more difficult to get additional appointments. It would be very useful to know the extent to which group differences are associated with the administration of wel- fare programs, rather than with differences in parental tastes or circumstances. It is unlikely that many detailed questions of this type will be added to large- scale surveys, but it would be possible to match data from other sources to the surveys if finer geographical information were made available to researchers. While issues of confidentiality are important, the amount of information that could be gained if it were routinely possible to match survey data to, say, zip- code-level data from other sources can hardly be underestimated. This type of matching is also greatly facilitated by the existence of a central agency that collects program information (and is willing to give it to researchers). There is a real danger that further Revolution of responsibility for welfare to the states will result in a loss of information about the administration of programs, making it more difficult to identify program effects using state-level variation in the programs. How Do Programs Interact? One glaring omission from this survey is that there has been no discussion of multiple program participation. Many children are covered by more than one program. For example, AFDC participants are covered by Medicaid and are automatically eligible for Food Stamps. As of 1990, half of AFDC children received free school lunches, 35 percent lived in public or subsidized rental housing, and 19 percent participated in WIC. Conversely, half of all Food Stamp recipients, 42 percent of Medicaid recipients, 38 percent of WIC recipients, and 24 percent of those in public housing also received AFDC. Moffitt (1992) esti- mates that in 1984, 26.4 percent of nonelderly single-parent families received AFDC, Medicaid, and Food Stamps, and 11 percent received at least one benefit in addition to AFDC. It is impossible to say how multiple program participation affects the child outcomes discussed above since there has been little research on this topic. Some programs may be duplicative, while others may interact to produce more positive outcomes. For example, Currie and Thomas (1995b) found that chil- dren in Head Start were more likely to be immunized than other children, even though many Head Start children would have been eligible for free vaccinations under the Medicaid program in any case. Head Start may help families to enroll in Medicaid, may help them locate a Medicaid provider, or may bypass Medic- aid altogether by arranging for children to be immunized at the Head Start center. An analysis of multiple program participation would assist us in answering the question of whether the current patchwork system of programs is an efficient way to provide welfare. The proliferation of programs increases possibilities for
OCR for page 199
JANET CURRIE 199 fraud, waste, and mismanagement. On the other hand, the evidence surveyed here suggests that targeting specific benefits directly to individual children has advantages in terms of ensuring that specific benefits are received. We need to know more about the balance between these benefits and costs. How Do Successful Programs Work? Data limitations place severe restrictions on our ability to look inside the "black box" of welfare programs. For example, we can show that expansions in Medicaid eligibility have been related to reductions in child mortality rates at the state level, but we do not know why. It could be due either to increased use of preventive care or to more intensive palliative care for sick children. The two possibilities have quite different implications for child well-being as well as for efficiency and program costs. Better information about what goes on during doctor visits and about objective measures of child health status (short of mortality statis- tics) could help us to address this question. It might be possible, for example, to add questions about anemia, lead poisoning, and anthropometrics (e.g., height-for-age, weight-for-height) to the next National Health Interview Survey. Still, the most likely scenario is one in which we chip away at these questions using an interactive, multidisciplinary approach: analysis of large-scale surveys can be used to develop broad hypotheses, which can then be tested using case studies. The case studies can then be used to develop more precise hypotheses about the survey data and to suggest supplemental survey questions. Cost-Effectiveness Evidently, if a program has no effect at all on a desired outcome, then it cannot be considered cost-effective. Many of the programs discussed above have passed this initial test they can be shown to have positive effects. The question remains however, of whether they are cost-effective, that is, whether the benefits outweigh the costs. The figures discussed above for WIC are quite impressive in this regard. Cost-effectiveness studies exist for other small-scale early intervention programs (not reviewed here) but have not generally been conducted for large-scale federal programs. Although it is unlikely that there will be agreement on all of the costs and benefits that should be included in such an analysis, some rough calculations under varying assumptions would no doubt be useful to policy makers. CONCLUSIONS This survey chapter discusses eight large federal welfare programs that af- fect children. The available evidence is incomplete but suggests a consistent story: programs that target services directly to children have the largest measured
OCR for page 200
200 THE EFFECT OF WELFARE ON CHILD OUTCOMES effects, while unrestricted cash transfer programs have the smallest, perhaps because their benefits are more diffuse or because the amounts of money in- volved are typically quite small. There are also sinking and largely unexplained differences in the effects of some programs by race, ethnicity, andlor nasality. These differences could reflect nonlineanties in the effects of programs that is, one might expect larger effects for poorer than for richer children, and children from some groups are more likely to be poor. Alternatively they may reflect differences in the programs available to children of different origins or unobserved differences between participants from different groups that have not been adequately accounted for. This survey concludes with five questions for future research: (1) Do wel- fare programs have long-term effects on children? (2) Why do programs have differential effects by race, ethnicity, and nasality? (3) How do programs interact? (4) How exactly do successful programs work? (5) Are programs cost-effective? These questions indicate that though we know much more than we did even 5 years ago about the effects of welfare on children, there is still much work to be done if we are to make informed decisions about public policy. ACKNOWLEDGMENTS The author is grateful to Lindsay Chase-Lansdale, Greg Duncan, Bentley MacLeod, and Robert Moffitt for helpful comments. Support from the Alfred P. Sloan Foundation, the National Science Foundation under grant #SBR-9512670, and the National Institute of Child Health and Human Development under grant #ROlHD31722-OlA2 is gratefully acknowledged. The author is solely respon- sible for the opinions expressed. REFERENCES Apgar, W. 1990 Which housing policy is best? Housing Policy Debate 1(1):1-32. Blank, R., and P. Ruggles 1996 When do women use AFDC and food stamps? The dynamics of eligibility vs. participa- tion. Journal of Human Resources (Winter). Blau, F., and A. Grossberg 1990 Maternal Labor Supply and Children's Cognitive Development. NBER Working Paper No. 3536. Cambridge Mass.: National Bureau of Economic Research. Bloom, B. 1990 Health insurance and medical care. In Advance data from vital and health statistics of the National Center for Health Statistics No. 188. October. Washington D.C.: U.S. Public Health Service. Brien, M., and C. Swann 1997 Prenatal WIC Participation and Infant Health: Selection and Maternal Fixed Effects. Uni- versity of Virginia Department of Economics. Discussion Paper 295. June.
OCR for page 201
JANET CURRIE 201 Caan, B., D. Horgen, S. Margen, et al. 1987 Benefits associated with WIC supplemental feeding during the interpregnancy interval. The American Journal of Clinical Nutrition 45:29-41. Cain, G., and D. Wissoker 1990 A reanalysis of marital stability in the Seattle-Denver income-maintenance experiment. American Journal of Sociology 95:1235-1269. Citro, C.F., and R.T. Michael, eds. 1995 Measuring Poverty: A New Approach. Washington, D.C.: National Academy Press. Clarkson, K. 1975. Food Stamps and Nutrition. Washington, D.C.: American Enterprise Institute. Currie, J. 1995a Welfare and the Well-Being of Children. Fundamentals of Pure and Applied Economics No. 59. Chur, Switzerland: Harwood Academic Publishers. 1995b Do Children of Immigrants Make Differential Use of Public Health Insurance? NBER Working Paper No. 5388, December. Cambridge, Mass.: National Bureau of Economic Research. 1995c Socio-economic status and health: Does universal eligibility for health care reduce the gaps? Scandinavian Journal of Economics 4 Currie, J. and N. Cole 1993 Welfare and child health: The link between AFDC participation and birth weight. Ameri can Economic Review 283(3). Currie, J., and J. Gruber 1996a Saving babies: The efficacy and cost of recent expansions of Medicaid eligibility for pregnant women. Journal of Political Economy (December). 1996b Health insurance eligibility, utilization of medical care, and child health. Quarterly Jour- nal of Economics (May). Currie, J., and D. Thomas 1995a Medical care for children: Public insurance, private insurance, and racial differences in utilization. Journal of Human Resources (Winter). 1995b Does head start make a difference? American Economic Review (June). 1996a Head Start and Cognition Among Latino Children. NBER Working Paper No. 5805. Cambridge, Mass.: National Bureau of Economic Research, December. Could Subsequent School Quality Affect the Long Term Gains from Head Start. Cam- bridge, Mass.: National Bureau of Economic Research, December. Currie, J., and A. Yelowitz 1997 Are Public Housing Projects Good for Kids? NBER Working Paper, October. Cambridge, Mass.: National Bureau of Economic Research. Currie, J., J. Gruber, and M. Fischer 1995 Physician payments and infant mortality: Evidence from Medicaid fee policy. American Economic Review (May). Cutler, D., and J. Gruber 1996 Does public insurance crowd out private insurance? Quarterly Journal of Economics 111(2):391-430. Desai, S., L. Chase-Lansdale, and R. Michael 1989 Mother or market? Effects of maternal employment on the intellectual ability of 4-year- old children. Demography 26(4):545-561. Devaney, B., P. Haines, and R. Moffitt 1989 Assessing the Dietary Effects of the Food Stamp Program, Volume 2: Empirical Results. Washington, D.C.: Mathematica Policy Research Inc. Devaney, B., L. Bilheimer, and J. Schore 1990 The Savings in Medicaid Costs for Newborns and Their Mothers from Prenatal Partici- pation in the WIC Program. Washington, D.C.: Mathematica Policy Research Inc.
OCR for page 202
202 THE EFFECT OF WELFARE ON CHILD OUTCOMES Dubay, L., and G. Kenney 1997 Did Medicaid expansions for pregnant women crowd out private coverage? Health Af- fairs (Jan./Feb.):185-193. Ellwood, D., and M. Bane 1985 The impact of AFDC on family structure and living arrangements. Pp. 137-207 in R. Ehrenberg, ea., Research in Labor Economics, Greenwich, Conn.: JAI Press. Ensminger, M., and A. Slusarcick 1992 Paths to high school graduation or dropout: A longitudinal study of a 1st grade cohort. Sociology of Education 65(April):95-113. Fraker, T. 1990a The Effects of Food Stamps on Food Consumption: A Review of the Literature. Washing- ton, D.C.: USDA Food and Nutrition Service. 1990b Analyses of the 1985 Continuing Survey of Food Intakes by Individuals, Volume II- Estimating the Effects of the WIC and Food Stamp Programs on Dietary Intake by Women and Young Children. Washington, D.C.: USDA Food and Nutrition Service. Fraker, T., A. Martini, and J. Ohls 1995 The effect of food stamp cashout on food expenditures. Journal of Human Resources 30(4):633-649. Gordon, A.R., B.L. Devaney, and J.A. Burghardt 1995 Dietary effects of the national school lunch program and the school breakfast program. American Journal of Clinical Nutrition 61(January):221-232. Gottschalk, P. 1990 AFDC participation across generations. The American Economic Review 80(2):367-371. Hannan, M., and N. Tuma 1990 A reassessment of the effect of income maintenance on marital dissolution in the Seattle- Denver experiment. American Journal of Sociology 95(5):1270-1298. Hanes, S., J. Vermeersch, and S. Gale 1984 The national evaluation of school nutrition programs: Program impact on dietary intake. American Journal of Clinical Nutrition 40(August):390-413. Heckman, J. 1990 Alternative approaches to the evaluation of social programs: Econometric and experimen- tal methods. Address to the World Econometric Society Meetings, Barcelona, Spain. Hicks, L., R. Langham, and J. Takenaka 1982 Cognitive and health measures following early nutritional supplementation: A sibling study. American Journal of Public Health 72:1110-1118. Hill, A., and J. O'Neill 1994 The transmission of cognitive achievement across three generations. Journal of Human Resources 29(4). Johnson, G. 1986 Rent paying ability and racial settlement patterns: A review and analysis of recent housing allowance evidence. American Journal of Economics and Sociology 45(1): 17-26. Jones, J. 1992 The WIC program: Eligibility, coverage, and funding. Washington, D.C.: Congressional Research Service, January 10. Kehrer, B., and C. Wolin 1979 Impact of income maintenance on low birth weight: Evidence from the Gary experiment. Journal of Human Resources 14(4):434-462. Kennedy, S., and M. Finkel 1987 Report of First Year Findings for the Freestanding Housing Voucher Demonstration. Cambridge, Mass: ABT Associates, Inc.
OCR for page 203
JANET CURRIE 203 Korenman, S., and J. Miller 1992 Food stamp program participation and maternal and child health. Draft Report to the Food and Nutrition Service of the U.S. Department of Agriculture, May. Lazere, E., P. Leanard, C. Dolbeare, and B. Zigas 1991 A Place to Call Home: The Low Income Housing Crisis Continues. Washington, D.C.: The Center on Budget and Policy Priorities. Lee, V. and S. Loeb Where do Head Start attendees end up? One reason why preschool effects fade out. Educational Evaluation and Policy Analysis 17(1):62-82. Mallar, C. 1977 The educational and labor supply responses of young adults on the urban graduated work incentive experiment. In Harold Watts and Albert Rees, eds., New Jersey Income Mainte- nance Experiments, New York: Academic Press. Mayer, S. 1996 Can More Money Buy Better Kids? Cambridge, Mass.: Harvard University Press. Maynard, R., and D. Crawford 1976 School performance. Rural Income Maintenance Experiment: Final Report. Madison, WI: Institute for Research on Poverty. Maynard, R., and R. Murnane 1979 The effects of a negative income tax on school performance: Results of an experiment. Journal of Human Resources 14(4). McKey, R., L. Condell, H. Ganson, B. Barrett, C. McConkey, and M. Planz 1985 The Impact of Head Start on Children, Families and Communities: Final Report of the Head Start Evaluation, Synthesis and Utilization Project. Washington, D.C.: CSR, Inc. Metcoff, J., D. Rubin, M. Napoleone, and K. Nichols 1985 Effect of food supplementation (WIC) during pregnancy on birth weight. American Jour- nal of Clinical Nutrition 41 (May). Meyers, A., A. Sampson, M. Weitzman, and H. Kayne 1989 School breakfast program and school performance. American Journal of Diseases of Chil- dren 143(10):1234-1239. Meyers, A., D. Rubin, M. Napoleone, and K. Nichols Public housing subsidies may improve poor children's nutrition. American Journal of Public Health 83(1). Michael, R. 1978 The consumption studies. In John Palmer and Joseph Pechman, eds., Welfare in Rural Areas: The Iowa Income Maintenance Experiment. Washington, D.C.: The Brookings Institution. Mitchell J., and R. Shurman 1984 Access to private obstetrics/gynecology services under Medicaid. Medical Care 22(No- vember): 1026-1037. Moffitt, R. 1992 Incentive effects of the U.S. welfare system: A review. Journal of Economic Literature 30(March): 1-61. 1989 Estimating the value of an in-kind transfer: The case of food stamps. Econometrica 57(2):385-410. Moore, K. and S. Caldwell The effect of government policies on out-of-wedlock sex and pregnancy. Family Plan- ning Perspectives 9(4):164-169. Mulroy, E. 1988 The search for affordable housing. In E. Mulroy, ea., Women as Single Parents: Con- fronting the Institutional Barriers in the Courts, the Workplace and the Housing Market. New York: Auburn House.
OCR for page 204
204 THE EFFECT OF WELFARE ON CHILD OUTCOMES Murray, C. 1984 Losing Ground. New York: Basic Books. O' Conner, F., P. Madden, and A. Prindle 1976 Nutrition. Rural Income Maintenance Experiment: Final Report. Madison, Wisc.: Insti- tute for Research on Poverty. Pedone, C. 1988 Current Housing Problems and Possible Federal Responses. Washington, D.C.: Congres- sional Budget Office. 1988. Robins, Philip 1985 A comparison of the labor supply findings from the four negative income tax exper meets. Journal of Human Resources 20(4):567-582. Rosenbaum, J., L.S. Rubinowitz, and M.J. Kulieke 1986 Low Income African-American Children in White Suburban Schools. Evanston, Ill.: Cen- ter for Urban Affairs and Policy Research. Rosenbaum, J. 1992 Black pioneers Do their moves to the suburbs increase economic opportunity for moth- ers and children? Housing Policy Debate 2(4):1179-1213. Schramm, W. 1985 WIC prenatal participation and its relationship to newborn Medicaid costs in Missouri: A cost/benefit analysis. American Journal of Public Health 75(8). Starfield, B. 1985 Effectiveness of Medical Care: Validating Clinical Wisdom. Baltimore, Md.: Johns Hopkins University Press. Stewart, A. 1992 Head Start: Funding Eligibility, and Participation. Report for Congress. Washington, D.C.: Congressional Research Service. July 22. U.S. Department of Health and Human Services Various Vital Statistics of the United States: Natality. Washington, D.C.: Government Printing years Office. U.S. Department of Health and Human Services, Centers for Disease Control, Public Health Service 1978 CDC Analysis of Nutritional Indices for Selected WIC Participants. ENS- 176, June. U.S. House of Representatives, Committee on Ways and Means 1993 1993 Green Book: Background Material and Data on Programs Within the Jurisdiction of the Committee on Ways and Means. WMCP-103-18. Washington, D.C.: U.S. Govern ment Printing Office. 1994 1994 Green Book: Background Material and Data on Programs Within the Jurisdiction of the Committee on Ways and Means. WMCP-103-27. Washington, D.C.: U.S. Govern ment Printing Office. 1996 1996 Green Book: Background Material and Data on Programs Within the Jurisdiction of the Committee on Ways and Means. Washington, D.C.: U.S. Government Printing Office. Venti, S. 1984 The effects of income maintenance on work, schooling, and non-market activities of youth. Review of Economics and Statistics 66(1):16-25.
Representative terms from entire chapter: