States with family-cap public assistance policies deny or reduce incremental welfare benefits to mothers who conceive and give birth to additional children while they are receiving aid. Such policies are both controversial and common. Between January 1992 and the passage of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) in August 1996, a total of 19 states sought and received approval (waivers) from the U.S. Department of Health and Human Services (HHS) for experimenting with family caps of one variety or another (HHS 1997). The same 19 states included some sort of cap in the Temporary Assistance for Needy Families (TANF) plan formulated to meet PRWORA requirements. By July 1999 the number of states with family-cap policies in place had increased to 22 (Rowe in press, table 19).

Oddly, none of the pre-PRWORA waiver applications included estimates of the size of the problem�the welfare conceptions�that family caps are supposed to address. At least two reasons might be cited for this e familiar departure from the rational policy paradigm. One has to do with Washington, the other with the states.

President Clinton�s 1993 invitation to states to increase welfare experimentation produced a tidal wave of applications. Faced with this flood, the federal agency responsible for granting permission for state welfare experiments, HHS�s Administration for Children and Families, did not, for the most part, challenge the procedures used by the states in selecting policies to pursue. Instead, the agency focused its attention on ensuring that approved reform demonstration projects would be evaluated rigorously�that is, by random assignment. On the states� side, assessing the possible impact of caps was impeded by software problems. Developing such estimates requires linking newborns� assistance cases to the circumstances of their parent(s) at the time of conception. Until recently, most state case management systems were designed primarily for dealing with current transactions�assessment of eligibility, calculation of payments, allocating current services, and the like (Wiseman 1999). Cobbling together such point-in-time data to link infants in dependent families to the assistance history of parents was a daunting task. Since the federal government didn�t press, the states remained daunted and avoided the effort.

So, despite the fact that caps are fashionable, we are left with a big cap question: Are conceptions among women on welfare sufficiently common to justify so much attention? This paper uses an extract from California�s Medi-Cal Eligibility Data System to count births among AFDC recipients and to assess how many of those births might have been subject to a cap. The data are from 1995 and 1996, the period immediately prior to PRWORA�s passage. The boundaries that delineate "a few" from "some" and "some" from "many" lie in the disposition of the beholder. At this point, the reader might want to decide for him- or herself just where those boundaries should be, before learning the number.

Before doing the analysis, it would be useful to explain the context and the data.

Context. As elsewhere, in 1995 California families with children qualified for Aid to Families with Dependent Children (AFDC) if their incomes, adjusted for certain expenses, fell below the state�s standard of need and (a) one parent was absent from the household or disabled or (b) the family�s principal earner was involuntarily working less than 100 hours per month. Families qualified on basis (a) were included in AFDC-FG (for "family groups"); families qualified on basis (b) were included in AFDC-UP (for "unemployed parent"). In July 1995, a mother with one child and no other income received a monthly grant of approximately $630 in AFDC and Food Stamps. (The exact amount depended on shelter costs.) An additional child would increase this benefit by $174, or 28 percent.

The decade from 1985 through 1995 was one of exceptional caseload growth in California (MaCurdy, Mancuso, and O�Brien-Strain 2000). Figure 1 shows the aggregate AFDC caseload and the two components, AFDC-FG and AFDC-UP, as well as points of reference for the calculations reported here. Between 1988 and 1995, the AFDC caseload in the state increased by 54 percent. By the end of this period, 21 percent of the state�s children were in families receiving AFDC. The state�s economic recovery from the recession of 1991�1992 lagged behind that of the rest of the country. Elsewhere the caseload turnaround began in late 1993 or early 1994 (the national peak was in March 1994). In California the turnaround began only in 1995, but subsequent reductions have been swift.

Note: Figure 1 also indicates that, by focusing on 1995, we will be counting welfare�s children at the peak of the caseload expansion.

The Longitudinal Database. Although the cap question seems simple enough in abstract, it is especially difficult to answer in California. Unlike most other states, California does not have a centralized management information system for public assistance programs. Instead, the operating systems for AFDC/TANF eligibility and benefits vary across counties. The consequence is that reviewing welfare conceptions is not simply a matter of looking at the state�s AFDC management information system, finding cases with infants, and looking back in the (electronic) file to the family�s welfare status 9 months prior to the new child�s birth. If such an approach were to be attempted, the task would require collection of data from each of the state�s 58 counties.

Fortunately, over the past eight years California has used an alternative data source to develop a longitudinal public assistance database that permits estimation of the number of births that would potentially have been subject to the cap. The alternative is the Medi-Cal Eligibility Data System (MEDS), the supporting system for the state�s Medicaid system. MEDS data are reported by counties in conjunction with claiming reimbursement from the state for poor peoples� health costs. When people qualify for AFDC (and now TANF) benefits, they are automatically eligible for Medi-Cal. The data about them that are entered into the MEDS system include the AFDC case number. When a person leaves AFDC, their Medi-Cal status and MEDS code changes, even in the common situations in which Medi-Cal eligibility endures. The MEDS status codes support study of AFDC case dynamics using the MEDS data.

The Longitudinal Database of Cases (LDB) is a 10 percent sample of all cases reported as being eligible for Medi-Cal for at least one month in the interval 1987�1996 (University of California-Berkeley Data Archive & Technical Assistance [UC DATA] 1997). The data cover the aid history of each person associated at any time with the sampled case from January 1987 or the date of first receipt of public assistance anywhere in the state, whichever is later. The LDB is intended to be part of the state�s data infrastructure for welfare policy research and evaluation of welfare reform initiatives. Public-use copies of the LDB are made available to researchers by the state�s contractor, subject to certain protocols designed to protect the confidentiality of the information contained. The LDB used here includes 1.4 million people and 625,000 cases; of these, 114,000 cases were open at some time during 1995.

Given its size, statewide coverage, and longitudinal character, the MEDS/LDB would appear to be ideal as a resource for evaluating the consequences of the cap. The approach has an important shortcoming, however. New babies typically are added immediately to the AFDC budget unit and the Food Stamps household. In contrast, data often are entered into MEDS with a lag, especially near the end of the year. In many counties, MEDS data are manually entered in a system separate from that used for AFDC/TANF management. Because births are charged to the Medi-Cal account of the mother, there is less pressure for immediate system adjustment. The result is that babies themselves may not appear immediately in the MEDS file, even when they are accounted for in AFDC.

Because of this administrative lag, in the computations that follow infants are identified as born on public assistance if they appear in MEDS data within six months of birth, are recorded as an AFDC recipient at the point they first appear, and are listed in cases in which at least one member was reported to be an AFDC recipient during the month of birth. Infants are identified as conceived by a woman already receiving assistance if they satisfy the criteria for being born on assistance and if they appear in cases in which at least one member was reported to be an AFDC recipient nine months earlier than the birth date. Note that this group includes cases receiving money only on behalf of children, but it does not include cases such as foster care, in which children are not living with at least one parent.

All Cases. The results are summarized in Table 1. The table is divided between information on case and birth counts and information about family composition; separate results are presented for AFDC-FG, AFDC-UP, and the caseload as a whole. The first row lists the average beginning-of-month caseload for the two subcategories and the combined total for 1995. The second row reports the estimated total number of cases ever receiving benefit within the subprogram over the entire year. These estimates are simply 10 times the relevant case counts from the LDB (recall the LDB is a 10 percent sample), so each number in the second row ends with zero. The total number of cases is less than the sum of the parts because 63,760 cases appeared both in AFDC-FG and AFDC-UP over the course of the year. Such transitions occur, for example, when an AFDC-UP family splits or when a disabled adult in a two-parent AFDC-FG family is judged capable of work.

Those readers desperate for the births number should skip to line (c) in the table and to the next paragraph in this paper. Those willing to pause before the prize might consider the following. First, the table provides an unusual insight into the circumstances of families receiving benefits under AFDC-UP. The fact that 63,760 cases appear in both AFDC-FG and AFDC-UP over the course of the year implies that roughly one out of every four mothers observed in AFDC-UP cases at some time during the year will be in an AFDC-FG case at another time in that same year. Second, there is turnover, and these data provide a chance to look at it in an unusual way. One measure of turnover is the ratio of cases ever open during a year to the average monthly caseload. This is the ratio of the program number in line (b) to the number in line (a). As is to be expected, turnover in the AFDC-UP program (ratio 1.49) is greater than in AFDC-FG (ratio 1.30), but both subprogram turnovers give an exaggerated impression of total turnover because some turnover is the product of intraprogram moves, from UP to FG and from FG to UP.

Returning to the cap question, line (c) reports the number of cases with births (a multiple birth counts as one case with birth), and line (d) is the incidence of birthing among cases open during the year. In 1995 slightly more than 1 in 12 cases open at any time during the year is associated with a newborn. This rate is higher for AFDC-UP cases (.096) than for AFDC-FG (.078). Line (e) is the subset of total births that satisfy the welfare conceptions criteria, line (f) expresses those births as a share of all births, and line (g) is the incidence of births potentially subject to caps across all cases open during the year. The answer: Statewide, in 1995 some 62,100 California AFDC cases included births that occurred which would have potentially been subject to a benefit cap had the state implemented such a policy. Almost two-thirds of all births to recipient families were children conceived while the family was already receiving assistance.

To some readers, two-thirds may seem low, not high. It should be recalled, however, that in 1995 California allowed women to open AFDC cases during the last trimester of pregnancy. About 62 percent of the births not subject to the cap appear to be first births; that is, the associated cases include no siblings and the case was not open at the estimated time of conception. The remaining 38 percent of births not subject to the cap were higher birth-order infants appearing in cases that were not open at the presumed time of conception.

Outcomes by Group. Birth rates and the incidence of welfare conceptions differ across race and ethnic groups. The LDB includes an extensive race/ethnic breakdown (UC DATA 1997, p. 17). Some groups�Filipinos, for example�have too few recipients for separate analysis. Data are adequate for separate analysis of six groups. Using the same descriptors applied by the California Department of Social Services, they are (in order from highest to lowest share of caseload), Hispanic, white (non-Hispanic), black, Vietnamese, Laotian, and Cambodian. All other groups, including the relatively few cases for which race/ethnicity was not identified, are combined as "other."

Table 2 reports the caseload share and the equivalent of the data in lines (e) and (g) from Table 1 for each of the six race/ethnic groups. Substantial variation across groups exists in both the incidence of births and the incidence of births potentially subject to the cap. Sample sizes are large, so virtually all the differences that appear to be important are statistically significant using the common .05 standard. Among the larger subgroups, the incidence of births subject to caps is greatest for blacks, followed by Hispanics and non-Hispanic whites. Laotian and Cambodian families, a small percentage of the caseload, would be substantially affected.

In Fiscal Year 1996, AFDC in California cost $6.5 billion in benefits and administrative costs (Committee on Ways and Means 1998, pp. 422, 432). The state actually implemented a cap of sorts in 1997 as part of its TANF program. Suppose that AFDC had always included a cap. How much smaller would this bill be?

It is not possible to answer that question with precision, for at least four reasons. First, the LDB does not extend back far enough in time to establish the status at conception of the parents of each child on assistance. The LDB begins in January 1987; consequently, given the 10-month grace period allowed in California�s new regulations, the applicability of a cap can be judged only for children born after November 1987. In the last year of the LDB case extract, these children are, at most, nine years old.

A second reason for uncertainty about the long-term consequences of a cap is that the literature does not provide enough information to incorporate the behavioral effects of a cap once instituted. Most models of caseload dynamics suggest that, other things being equal, small cases exit faster than large ones and cases close more rapidly in states with low benefits than in those that offer more generous benefits. If a cap reduces fertility, it would therefore also reduce the expected duration of cases. The public would save both through smaller benefit payments and through smaller caseloads.

The third reason has to do with details of the cap actually applied in California. The cap, termed the "Maximum Family Grant (MFG) Rule," was established by law in 1994 on the basis of a previously granted federal waiver (California Department of Social Services 1996). Application began August 1, 1997; notices were sent to recipient families in July 1996 cautioning that the MFG rule could apply to any child born after July 1, 1997. The legislation allowed for aid to be paid on behalf of a child conceived on assistance if the entire family had been off aid for two years at some point after the child�s birth or�and here is the important exemption�if the entire family had left cash assistance continuously for at least two months during the 10 months prior to the child�s birth. The reference to cash assistance is important; it means that should the family choose to separate from assistance for 60 days, the new child would be exempt from the cap. For a single mother with one child, the cost of this strategy would be two months of benefits in return for the substantially greater long-term benefit of inclusion of the new child on the case. Medi-Cal benefits would not be lost over this period, and the food stamp benefit would increase. In California, families returning to assistance after short departures are not subject to the same requirements that are applied on first entry; such "restorations" generally are fast-tracked. Clearly, California�s version of the family cap is much different from the cap as applied in other states, so even if reliable evidence of cap effects on behavior could be obtained from experience elsewhere, its applicability for forecasting the effects of the MFG rule would be questionable.

The fourth reason for uncertainty about what the cap might mean in California practice lies with a feature of the state�s payment system already mentioned. Benefits payments in California are handled by counties. There appears to be no systematic review of application of such rules, and it thus is not clear at all that they are systematically applied.

These are serious problems. Nevertheless, it is useful to estimate the number of children of younger ages who were in families on AFDC at a point near the end of the LDB and who would have been ineligible for benefit had a benefit cap policy similar to that applied in most other states been in effect since 1987. This estimate provides a sense of the magnitude of potential cap savings and the breadth of distribution of cap effects; actual effects would be less, because of the various exemptions and uncertainty about implementation. Because of problems associated with the process of updating the LDB that are detailed elsewhere (Wiseman in press), I chose February 1996 for this estimate. I know of no reason to believe that results for other months would differ significantly. Table 3 reports the outcome of my calculations, both in terms of children under age nine who would potentially be affected and in terms of the number of cases with such children that would lose benefits. The two numbers can differ because some cases may have more than one child potentially subject to a cap. In February 1996 there were 936,000 AFDC cases open in California, which included 1.82 million children.

Here�s how the table should be read. Beginning with the top panel, in February 1996 a total of 1,006,210 children under age 9 were living in California families receiving AFDC cash benefits. This is (looking back at the child count from the previous paragraph) 55 percent of all children in AFDC families. Of these children, 30 percent (the number at the head of the last column), or 297,870, were conceived during a period in which a family member was receiving cash assistance. Children under age 9 resided in 649,780 cases, 69 percent of total cases open. The number in the upper-right hand corner of the bottom panel of the table shows that 34 percent of these 650,000 cases with children less than nine years old would have been affected. The differences across race/ethnic groups are large, with Southeast Asian families much more likely to be affected than blacks, Hispanics, or non-Hispanic whites. In general a higher proportion of cases would be affected than of children. This suggests that conceptions on welfare may be more a characteristic of cases with a small number of children than of those with a large number.

Suppose for simplicity that each of these children was a second child, so that the marginal benefit figure of $174 (actually, $177 by February 1996) reported earlier would apply. Of this change, $117 is attributable to the AFDC payment itself; the rest is food stamps. If February may be taken as representative of all 12 months, total savings from removing all these children from AFDC benefits would amount to 298,000 children multiplied by the marginal benefit of $117 multiplied by 12 months, or $418 million. This is slightly more than 7 percent of total outlays in the state. In 1996, the federal government paid half of California�s AFDC costs, so the state�s saving from this hypothetical exercise would have been roughly $209 million. The TANF block grant raises the ante; since the grant is fixed, the state would keep both the state and federal share. Under either the old or new rules, roughly 30 percent of the reduction in payments to recipients, if applied, would be offset by an increase in food stamps.

The reader is counseled to exercise caution in using these numbers. In practice, some children who meet the definition of "capped" used here would probably not be counted, especially if the case were to be opened and closed during the interval between the child�s birth and the point of enumeration. By law if such departures lasted for two years or more, the cap is set aside. These calculations do not include adjustment for the possibility that women who become pregnant while receiving assistance will choose to forego the TANF cash benefit for two months in order to avoid the penalty. Nevertheless, the calculations indicate that at least in California, a full-fledged family cap could reduce payments by a considerable amount, especially given that such savings grow over time, as the cohort initially subject to the cap ages and reduced benefits lead to higher termination rates for the families involved.

This paper has produced estimates of the number and cost implications of infants conceived by mothers receiving assistance in California in 1995. The estimates are constructed using a longitudinal analytic database derived from administrative records.

The results indicate that such births are a frequent occurrence. In 1995 about 4.8 percent of the AFDC-FG (single-parent) California cases and 6.7 percent of the AFDC-UP (two-parent) cases experienced the birth of a child conceived while the case was open. The absolute number of such births is large, involving some 62,000 families across both subprograms in 1995, the peak year of the caseload.

Under benefit cap policies, children conceived and born while their families were receiving TANF do not produce a change in welfare grant. Although the one-year number of such births is small in comparison with total numbers of children in AFDC families in 1995, such children accumulate over time, and cases that include such births tend to stay on benefit for a long time. Thirty percent of children under age nine in families receiving AFDC benefits in February 1996 were conceived while their case was open; benefits paid on their behalf amounted to roughly 7 percent of total state outlays. The incidence of such births is much higher for some race/ethnic groups than for others. Absent immediate change in birth rates in response to caps, the consequences for family income of such policies will differ substantially across groups. Because the savings resulting from the family cap grow over time, California�s family cap could be contributing to the accelerating decline in the state�s welfare costs.

It is not clear, however, that California�s cap has such effects. The actual amount of savings depends in part on recipient responses to the strategic option that the MFG rule provides to women who become pregnant. What the state has done is to create a sort of two-month "penalty box" that such women can use to avoid the cap. At least on its face, the penalty box seems principally to benefit women who are well counseled and who have the best networks for sustaining their families while foregoing benefits in order to avoid the cap. Women without such options may not be able to do without benefits, but they may be tempted to do so, especially if the consequences of the cap are not well understood. The state�s policy seems to encourage families to reduce their resources at a time�pregnancy�when the consequences of stress and lack of money would seem most severe.

California Department of Social Services. 1996. Maximum Family Grant Notice Implementation in the Aid to Families with Dependent Children (AFDC) Program. (All-County Letter No. 96�37). Sacramento, CA: Author.

U.S. House of Representatives, Committee on Ways and Means. 1998. 1998 Green Book: Background Material and Data on Programs within the Jurisdiction of the Committee on Ways and Means. Washington, DC: U.S. Government Printing Office.

MaCurdy, T.; Mancuso, D.; and O�Brien-Strain, M. 2000. The Rise and Fall of California�s Welfare Caseload: Types and Regions, 1980�1999. San Francisco: Public Policy Institute of California.

Rowe, G. In press. The Welfare Rules Databook. Washington, DC: Urban Institute.

U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation. 1997. Setting the Baseline: A Report on State Welfare Waivers. Washington, DC: Author.

University of California-Berkeley Data Archive & Technical Assistance (UC DATA). 1997. California Work Pays Demonstration Project: Statewide Longitudinal Database�Cases 10% Sample. 1987�95, Public Use Version 2 Dataset Codebook [machine-readable data file]. Berkeley: UC DATA.

Wiseman, M. 1999. A Management Information Model for New-Style Public Assistance. (Assessing the New Federalism Discussion Paper 99�10). Washington, DC: Urban Institute.

Wiseman, M. In press. Welfare�s Children (Discussion Paper). Madison, WI: Institute for Research on Poverty.

Back to Table of Contents