Since you are looking at this site with an older browser, you will not be able to see any graphics or formatting. For better results, please upgrade your browser.

The Dynamics of Child Care Subsidy Use
A Collaborative Study of Five States

Authors: Ann M. Collins, J. Lee Kreader, and Annie Georges
Publication Date: August 2002

Executive Summary

Background and Research Questions

The federal welfare-reform legislation enacted in 1996, the Personal Responsibility and Work Opportunity Reconciliation Act, gave states and communities new challenges and new opportunities for meeting the needs of low-income families and children. The intensity of these challenges and the importance of these opportunities are especially striking in policies for subsidized child care. In developing subsidy policies, both federal and state policy officials have had to act in the absence of good information about subsidy use among low-income families. Although some (but not all) states are able to use administrative data to obtain basic information about the population using subsidies at a point in time, they have been unable to extend these analyses to examine in great depth the characteristics of families or their patterns of child care use.

Of particular importance, cross-sectional data have not provided information about the dynamics of subsidy use. Given persistent low earnings, parents leaving welfare (and other low-skilled working parents) are likely to remain eligible for means-tested child care assistance for a relatively long period. At the same time, however, their participation in short-term employment preparation activities, turnover in employment, and variable earnings may make it difficult for these families to remain continuously eligible for subsidy assistance. Burdensome application and recertification processes may create additional barriers to continuous subsidy receipt. For parents, instability in subsidy receipt may mean the difference between keeping and losing a job, and, for those employed, between self-sufficiency and poverty. For children, instability in subsidy receipt may contribute to instability in care arrangements, which developmental experts identify as a risk to healthy socio-emotional development.

To advance knowledge and understanding about the dynamic use of child care subsidies, this study used data from five states (Illinois, Maryland, Massachusetts, Oregon and Texas) to address the following questions about child care subsidy use and cross-state variation:

  • What are the characteristics of children and families who receive subsidies?
  • What services do subsidized children and families in these states receive?
  • How continuous is subsidy receipt; i.e., how long do spells of subsidy receipt last?
  • What is the duration of subsidy use; i.e., how likely is it that children who end a spell of subsidy receipt subsequently begin another?
  • How stable are children’s care arrangements while they are in the subsidy system?

Data and Methods

This study used data collected and analyzed by a team of policy and methodological experts brought together through the Child Care Policy Research Consortium. The Child Care Policy Research Partnerships include child care policymakers, practitioners, and researchers from each of several states. These teams are charged with developing a research agenda in response to pressing policy questions in their states. For this study, consortium members representing Illinois, Maryland, Massachusetts, and Oregon joined together to build a collaborative, cross-state research design. Researchers in Texas, who were already engaged in similar research, agreed to join the project. Researchers from each of the states met together several times over a two-year period to develop comparable data sets and to design and interpret data analyses. The research team included policy experts familiar with each of the five states and analysts familiar with administrative data and analytic methods. State agency partners played a critical role in explaining data elements and policies in each state, and provided valuable feedback on the study results and interpretations.

State policy data were collected through document reviews and interviews with key informants in each state. Micro-data were obtained from state child care subsidy administrative systems. Payment record files were obtained for services during 24 calendar months in each state, typically for the period of July 1997 to June 1999. In four of the five states, these data sets included information on all families that were served in the states’ voucher-based subsidy programs during this period. (The data set for Massachusetts included information on about one-half of voucher recipients).

Analysts first constructed comparable variables in each of the five state data files and then reconfigured the data sets into longitudinal files. Analysis samples were constructed by randomly selecting one child from each family that participated in the subsidy system during the observation period. A “spell” of subsidy receipt was measured as the number of continuous months of receipt by the selected child, preceded and followed by a month of nonreceipt. The first analyses use descriptive statistics to compare the characteristics of children and families served in the five states and of the services provided. Researchers next compared the continuity of subsidy receipt across the states, and across groups within states, by estimating the length of the first observed spell of subsidy receipt. The Kaplan-Meier survival analysis procedure was used to adjust for incompletely observed or right-censored periods of receipt (that is, cases in which the child received a subsidy in the last month of the observation period, because researchers do not know whether the child continued to receive a subsidy). The study team compared the likely duration of total receipt by examining rates of reentry to the subsidy system following the end of a subsidy spell. The stability of child care arrangements while children are subsidized was assessed by comparing the number of different providers that children experienced during their time in the subsidy system.

The strength of the study is that it uses child care administrative data for all or a significant proportion of all children who ever used child care subsidies during a two-year period in each state. Analyzing the universe of subsidy cases allows researchers to describe the characteristics of subsidy recipients and the dynamics of their subsidy use with great accuracy. Comparing these characteristics and dynamics across states allows researchers to observe state-level variation that may result from child care, TANF (Temporary Assistance for Needy Families—the federal cash assistance program that replaced Aid to Families with Dependent Children in 1996), and other public policies.

One of the principal limitations of the study is the exclusive reliance on one source of administrative data. Administrative data are collected by systems designed to determine eligibility and process payments; they are not designed with research purposes in mind. The child care subsidy payment records used as data for this study did not include information on families once they left the subsidy system or the reason for their exit; this limited the team’s ability to evaluate transitions out of the subsidy system. These data also lacked individual-level variables that could be used for multivariate analyses of subsidy dynamics. The analysis team experimented with both single-state and pooled, hierarchical regression analyses. The main finding from these analyses is that the effect of predictor variables varied across states; i.e., the association between the length of subsidy receipt and factors such as child and service characteristics is not constant. The lack of exogenous measures of family and policy characteristics precluded estimating reliable multivariate models. Families’ decisions to participate in the subsidy program and their employment and child care decisions are so closely intertwined that separating out causal factors is a challenging task and requires data that were not available in this study.

Summary of Major Findings

  • The exercise of policy discretion at the state level has produced very different child care subsidy programs in different states.
  • States served different populations in their subsidy systems.
  • –Median incomes among subsidized families in different states ranged from 12 percent of the state median income in Texas to 24 percent in Illinois.

    –The proportion of subsidized families that was working and not receiving TANF ranged from 15 percent in Illinois to 55 percent in Texas.

    –The proportion of subsidized families that was mixing work and welfare ranged from 5 percent in Oregon to 71 percent in Illinois.

  • States provided different services to subsidized families and children.
  • –The median incomes of families receiving subsidies ranged from only 16 percent of the state eligibility ceiling in Texas to 51 percent of the ceiling in Illinois.

    –The proportion of subsidized children whose main care arrangement was center-based ranged from 18 percent in Oregon to 79 percent in Texas.

    –The proportion of subsidized families who were exempted from copayments ranged from 10 percent in Illinois to 85 percent in Massachusetts.

    –The median value of copayments (among families who paid them) ranged from $29 per month in Maryland to $67 per month in Oregon.

    –The median net value of child care subsidies (provider payments minus family copayments) ranged from 21 percent of the median income of subsidized families in Illinois to 76 percent in Massachusetts.

  • The length of children’s spells of subsidy receipt was short in all states and varied across the states.
  • –In all states, spells of subsidy receipt ended for one-half or more of the children within seven months.

    –The median length of subsidy spells ranged from a low of three months in Oregon to a high of seven months in Texas.

  • Within states, the length of subsidy spells varied with some family characteristics.
  • –In four of the five states, children whose parents were employed and not receiving TANF at the start of the spell had longer spells of child care subsidy receipt than children with nonworking TANF-recipient parents.

    –In all state, the length of subsidy spell varied by one to two months for children who started the spell in alternative care arrangements. No consistent pattern emerged in the type of care arrangement associated with shorter or longer spells across states.

  • There was considerable reentry to the subsidy system.
  • –In all states, at least one-third of children who exited a spell of subsidy receipt began a subsequent spell of assistance within 12 months; the proportion of children returning to the subsidy system within 12 months ranged from 35 percent in Texas to 58 percent in Maryland.

    –For children who returned to the subsidy system within 12 months, subsequent spells of receipt were no longer than their first observed spells.

  • Most children who had short spells of subsidy receipt had a consistent provider during their months of receipt. Approximately one-half of children receiving subsidies for at least one year had one or more transitions in their primary provider.
  • Spell length appears to be affected by interactions of child care subsidy, TANF, and regulatory policies rather than by any single child care policy. Some policies were associated with variation in spell length in the direction that would be predicted by theory and child care research.
  • –Spells of subsidy receipt were shorter in states that required some or all families to recertify eligibility more often than every six months.

  • Other policies did not have the anticipated association with child care subsidy continuity. This may be due, in part, to the interaction of multiple state policies.
  • –The generosity of the ceiling for continuing eligibility was not consistently associated with the length of subsidy spells, in part because the population of families served in several states had incomes well below the eligibility cutoff.

    –The level of copayments required of families was not consistently associated with length of subsidy spells across the five states. Families may leave the subsidy system before reaching the high levels of copay, as few of the observed families had incomes high enough to be subject to the highest levels of copay imposed by the states.

    –The generosity of payments to providers was not consistently associated with continuity, which may be due to the reduction in the “net” value of assistance to families once copayments were imposed.

  • Cross-state variation in subsidy continuity may also be due to policies that influence the mix of families served in each state.
  • –The observed length of subsidy spells was longer in states that had a higher proportion of working families in their subsidized populations, in part because children in these families had longer spells of receipt than children in fully TAF-reliant families.

    –Although variation in state TANF and other policies may explain a portion of the overall cross-state variation in length of subsidy spells (through their influence on the mix of families in the subsidy system), it cannot fully explain cross-state variation in subsidy continuity among families with similar characteristics.

This study provides new information about the dynamics of subsidy use in these five states. Due to the limitations of the data, the findings are largely descriptive. The administrative data used in this study did not permit more extensive analysis of the factors that explain variation in subsidy dynamics among families or across states. They did not include, for example, family characteristics (such as education) and circumstances (such as prior employment) that are likely to have influenced participation in the subsidy system. Even more importantly, they did not include data on changes in family circumstances in the months following the end of a subsidy spell. The research team could not determine whether a spell of subsidy receipt ended for a positive reason—such as an increase in earnings—or for more problematic such as the loss of a job, the loss of a child care provider, or difficulties with the recertification process. These data constraints sharply limited the team’s ability to model, in a multivariate context, the factors associated with the length of subsidy receipt.

Conclusions and Implications

The diversity of the populations served and the diversity of the services provided across the five states are two of the most striking findings of this study. The devolution of already highly decentralized child care subsidy programs in the 1990s increased opportunities for state policymakers to determine who receives subsidies, what types of providers are subsidized, how much providers are paid, and what portion of costs are paid by families. One consequence of these policy choices is that states serve very different populations of families in their subsidy systems. A second consequence of these policy choices is that states now provide varying types and levels of service to families.

One common characteristic across these states was the low level of continuity in subsidy assistance. Many children also reentered the subsidy system, although the rate of reentry varied and did not exceed 60 percent in any state. The duration and stability of subsidy assistance varied across the five states. The median length of subsidy spells ranged from three to seven months, and the proportion of children returning to the subsidy system within 12 months ranged from 35 to 58 percent.

Because the data for this study did not reveal why children left the subsidy system, it is difficult to interpret these findings. Short subsidy spells and churning in and out of the system may be due to the episodic nature of parents’ employment activities or to problems associated with child care arrangements or subsidy receipt. Regardless of the reason, the lack of continuity and short duration of subsidy assistance is of concern. It is unlikely that parents who were poor enough to qualify for subsidies had achieved a level of self-sufficiency such that they no longer needed subsidies within the few months that their children received assistance. Indeed, the fact that as many as one-half of children returned for a subsequent spell of subsidy assistance within one year suggests that many parents remained eligible for assistance.

It is clear that variation in the dynamics of subsidy use cannot be explained by any single child care policy. Variation in state TANF policies may explain some of the cross-state variation in the length of assistance, through its influence on the mix of families in the subsidized population. Other state policies may also explain some of this cross-state variation in subsidy dynamics by creating incentives or barriers to participation.

These analyses raise a number of questions for future research. The lack of continuity and frequency of reentry to the subsidy system deserve further research, both to describe the dynamics of subsidy receipt and to identify factors that are associated with more stable subsidy experiences.

These analyses also raise a number of cautions and suggestions for future research. This is one of the few studies to use administrative data from the child care subsidy systems and one of the only studies to use data from the child care systems of multiple states. A key methodological insight of the study concerned the potential noncomparability of variables and effects across states. Some of this noncomparability is due to measurement. Other data elements were not fully comparable because they had very different implications in different policy contexts.

These issues of noncomparability raise two important cautions for future research in this area. First, it is critical for researchers using administrative data sets from multiple states to understand what the data elements measure. This requires both a detailed knowledge of the administrative processes through which the data were collected and used, and a detailed knowledge of the relevant policies in each state. Second, there is no reason to assume a priori that variables will have a constant association across states. As both the parallel and pooled, multi-level analyses for this study found, different parameter estimates were obtained in different states for the same variable. This raises important cautions about pooling data for analysis without accounting for the possibility that the true parameters, or underlying associations, vary across states or sites.

The results of this study do not provide specific lessons for the development of child care subsidy policy. The sample of five states was too small, and the administrative data sets too limited, for the study team to identify policies that hinder or support families’ use of subsidies. The results do suggest, however, two areas of concern for future policy.

The first concern relates to equity. Social policy devolution is often praised as a mechanism for increasing local political control and responsiveness. It is just as often criticized because it eliminates national standards and due process protections for applicants and clients who are often socially and economically vulnerable. This tension is apparent in child care subsidy policies. As these results suggest, essentially similar families have different likelihoods of receiving assistance, depending on the state in which they live. Once they are in the system, families have different service options, and face different costs and benefits, depending on where they live. This raises important questions about whether the public child care subsidy system is providing assistance equitably to needy families.

The second concern relates to the specific dynamics observed in this study. A number of studies have documented low and variable rates of participation in child care subsidy programs among the low-income populations of different states. This study suggests that in addition to having trouble accessing subsidy assistance, low-income families may be having trouble retaining that assistance. One of the clearest conclusions from decades of research on welfare dynamics and the employment of low-educated workers is that mothers in the low-wage job sector experience both high levels of job instability and low levels of earnings growth over time. This suggests that low-income families exiting welfare, and other working poor families, are likely to need child care subsidy assistance for a long period of time. The results of this study suggest that, currently, the assistance families receive is not very continuous, does not last very long, and may be associated with substantial turnover in their children’s care arrangements. These dynamics do not bode well either for families’ economic security or for children’s healthy socioemotional development.