Abstract
Every year, a large number of children in the United States enter the foster care system. Many of them are eventually reunited with their biological parents or quickly adopted. A significant number, however, face long-term foster care, and some of these children are eventually adopted by their foster parents. The decision by foster parents to adopt their foster child carries significant economic consequences, including forfeiting foster care payments while also assuming responsibility for medical, legal, and educational expenses, to name a few. Since 1980, U.S. states have begun to offer adoption subsidies to offset some of these expenses, significantly lowering the cost of adopting a child who is in the foster care system. This article presents empirical evidence of the role that these economic incentives play in foster parents’ decision of when, or if, to adopt their foster child. We find that adoption subsidies increase adoptions through two distinct price mechanisms: by lowering the absolute cost of adoption, and by lowering the relative cost of adoption versus long-term foster care.
Introduction
In 2008, approximately 463,000 children were in the U.S. foster care system, with 273,000 of them entering in that year alone (U.S. Department of Health and Human Services, Children’s Bureau 2009). Many children entering foster care are reunited with their biological parents or are quickly adopted. However, a significant number of these children face long-term foster care. For these children, being adopted by their foster parents represents their best chance of leaving the foster care system. Unfortunately, of the 123,000 foster children awaiting adoption,1 only 54,284 were legally adopted in 2008. On average, these children had been waiting to be adopted for nearly four years. Although adoption rates for boys and girls were similar, the adoption rate among black children was significantly lower than that among white children. Adoption rates were also significantly lower for older children, those placed with single foster parents, and those placed with relatives.
The decision by foster parents to adopt their foster child carries significant economic consequences: forfeiting foster care payments while assuming responsibility for medical and educational expenses, to name a few. Recently, U.S. states have at least partly offset these economic consequences by offering adoption subsidies and credits for adoption-related expenses. This effort began in 1980 with the passage of the Adoption Assistance and Child Welfare Act (AACWA, P.L. 96–272) aimed at reducing the length of stay in foster care and promoting adoptions. This Act requires states to provide adoption assistance payments to families who adopt children with special needs. Although the AACWA provides an outline for the definition of special needs (including children with a diagnosed disability, children in a sibling group, minority children, and older children), the details of the adoption assistance program as well as payment amounts are left to each state. As a result, some states began to offer adoption assistance equal to the payments foster care parents would have received if the child were not adopted, and other states offered a lower payment to adoptive parents than to foster parents. In all cases, these state policies significantly lowered the cost to foster parents of adopting their foster child and provide us an opportunity to study the impact that economic incentives have on family decisions.2
Because many of the specifics of a state’s adoption assistance program are under the control of state and local governments, policy makers have the potential to alter the probability and timing of the adoption of foster children. Ideally, we would like to analyze foster care adoptions before and after states enacted adoption subsidies in 1980. Unfortunately, reliable data on large numbers of foster children do not exist for years prior to 1998. Instead, we use data from the 1998 and 2006 Adoption and Foster Care Analysis and Reporting System (AFCARS) to analyze the role that economic incentives, in the form of foster care and adoption subsidies, play in a foster parent’s decision to adopt a foster child. The AFCARS data contain information on individual children in the foster care system, including the child’s age, gender, race and ethnicity, case goal, reason for removal, special needs status, length of stay, the termination of parental rights, and date of adoption. The data also contain information about the child’s foster caregivers.3 We link the AFCARS data on individual children to state foster care and adoption subsidy rates by the child’s age and state of residence to examine the effects of these policies on the adoption rates of children waiting to be adopted.
A standard cross-sectional estimation approach would not be appropriate if state-level foster care and adoption policies reflect differences in attitudes toward children in general or toward foster care and adoption specifically. To address this possibility, we control for unobserved differences between states and between children of different ages by exploiting the variation in basic monthly foster care and adoption payments within each state-age group over time. For example, many states made significant changes in the structure of their foster care and adoption subsidies between 1998 and 2006. These changes affected both the overall level of foster care and adoption payments, and the difference between the two payments. Furthermore, within individual states, the magnitude of these changes often depended on the age of the child. Our analysis uses the variation in payments over time within each age group in each state to identify the effect that lowering the cost of adoption has on a parent’s decision to adopt a foster child. We find that adoption subsidies significantly increase the adoption rate of children waiting to be adopted. Specifically, we find that foster parents respond to both the level of payments and to the difference between the foster care and adoption payments in predictable ways. For example, logit model estimates suggest that, holding constant foster care payments, a $100 increase in monthly adoption payments is associated with a 4.6 percentage-point increase in the adoption rate of boys and a 5.9-percentage-point increase in the adoption rate of girls. These percentage-point increases translate to an increase in the overall adoption rate of 27 % for boys and 34 % for girls over the 1998 rate. Moreover, increasing the generosity of payments by increasing both foster care and adoption payments by $100 per month is associated with a 3.7-percentage-point increase in the adoption rate of boys and a 6.2-percentage-point increase for girls. This corresponds to an increase in the overall adoption rate of 17 % for boys and 35 % for girls over the 1998 rate. These findings are consistent with the theory that lowering the cost of adoption increases the number of adoptive parents through two mechanisms. First, more generous payments attract additional family foster care providers to the child welfare system. Second, lowering the cost of adoption relative to the cost of foster care encourages more foster parents to choose adoption over long-term foster care.
Adoption From Foster Care
To address the needs of foster children, legal scholars and policy makers have debated priorities in foster care policies for the past three decades (Guggenheim 1999). The Adoption Assistance and Child Welfare Act of 1980 addressed some of these concerns by providing federal guidelines for state child welfare agencies by mandating that foster care be the option of last resort and that all “reasonable efforts” be made to preserve or reunify the family (Lowry 2000). The Act also established the adoption assistance program designed to subsidize adoptions of special needs children. The 1980s and 1990s saw a dramatic increase in both the number of children in foster care and the average time that children spent in foster care. A shift in foster care philosophy came with the passage of the Adoption and Safe Families Act of 1997, which replaced the “reasonable efforts” requirement with a “fast track” to permanency requirement. The 1997 legislation expedited the termination of parental rights and mandated that states move children toward adoption after a relatively brief period of time, shifting some of the emphasis away from providing services to families in crisis and working toward reunification (Pagano 1999). The legislation also provided new financial incentives for states to increase adoption rates above targeted levels.
Whether the goal is reunification or adoption, most observers agree that long-term foster care is detrimental to the child. Policy makers and professionals dedicated to ensuring a developmentally beneficial environment for foster children point to the need to avoid foster care “drift” that occurs when children are frequently moved from one foster care setting to another. The American Association of Pediatrics released a statement indicating that “multiple foster home placements can be injurious” to children (USA Today Magazine 2001). Child development researchers suggest that multiple placements result in unstable adult-child relationships that deplete a child’s ability to form attachments to significant others (Usher et al. 1999), and so the focus on adoption of children out of foster care has intensified.
In keeping with the consensus that long-term foster care should be avoided, the goal of current policy is to move foster children into the permanence of adoption if reunification with the biological family cannot be achieved quickly. However, large-scale research designed to identify factors associated with adoption and its timing has been conducted only in the past few decades. For example, Finch et al. (1986) found that only one-quarter of the children placed in out-of-home care in New York in the late 1970s and identified as available for adoption were adopted within two years. Their findings, which are consistent with other studies conducted more recently in other states, suggest that the probability of adoption is related to the child’s race and ethnicity. Finch et al. also found that the probability of adoption declines with time spent in out-of-home care. This is similar to the finding that adoption is most likely for younger children (Barth 1997).
Not only do children of color make up a larger proportion of the out-of-home care population than of the population at large, but they also face a lower probability of adoption than white children (Barth 1997; Brooks and James 2003; Courtney and Barth 1996; Finch et al. 1986; Wulczyn and George 1992). Specifically, Finch et al. (1986) estimated that white children are nearly 11 % more likely to be adopted than children of other racial and ethnic groups, although the racial gap in adoptions may be closing (Wulczyn 2003). Concern about the overrepresentation of children of color among the foster care population and low rates of adoption for these children prompted the passage of Multiethnic Placement Act of 1994 and the Interethnic Adoption Provisions of 1996 to remove barriers to interracial adoption and to move more quickly to permanent families for children of color (Brooks et al. 1999).
The call to move children from temporary foster care to the permanence of adoption is clear; however, the mechanism by which this goal might be achieved is not. Research suggesting that the probability of adoption varies by race, ethnicity, age, and time in foster care provides policy makers with little insight into how to increase adoption. Few studies have been conducted to identify factors that provide policy makers with the leverage to alter the probability and timing of the transition from foster care to adoption. In a report based on evaluation of the child welfare system in Washington state, Thompson et al. (2001:13) suggested that “. . . an increase in the state adoption subsidy ha[s] resulted in a substantial increase in foster family adoptions.” Although it is clear that policy makers recognize that goals may be more easily achieved when “government aligns financial incentives with the outcomes it hopes to achieve” (Children and Family Research Center 2003:2), no large-scale national studies of the effect of these subsidies on adoption have been conducted.
Previous studies examining the impact of foster care subsidies on placement within the foster care system found that more generous foster care payments increase the overall supply of foster care providers (Doyle and Peters 2007; Simon 1975), improves the retention of foster families (Chamberlain et al. 1992), increases kin placements (Doyle 2007), and increases the probability of placement with a foster family rather than in a group setting (Duncan and Argys 2007).
A few studies have addressed the impact of state subsidies on adoptions. Avery and Mont (1992) collected data from counties in New York to identify the impact of subsidy levels on the timing and probability of adoption of children with special needs. They found that children with mental disabilities who qualified for greater adoption subsidies (based both on their own characteristics and practices within their county of residence) faced a greater probability of adoption. Their results also suggest that subsidies have no effect on adoption for other special needs children. Hansen and Hansen (2006) used aggregate data from the 1996 AFCARS to conduct cross-sectional analysis of the association between a state’s monthly adoption assistance payments and the number of children adopted out of foster care per 100,000 state populations. They found a positive association between the monthly adoption subsidy for 9-year-olds and the total number of children adopted from state foster care. Hansen (2007) extended her previous work to include AFCARS data from 1996–2003. Her state fixed-effects estimates confirm earlier findings that state assistant payments are positively associated with adoptions of foster children. Finally, Buckles (2009) examined the impact of adoption subsidies on the likelihood of adoption by exploiting variation between states in the age at which children qualify for adoption assistance. Buckles tracked individual children in the 2000–2006 AFCARS based on the child’s state, date of birth, and date of removal, allowing her to estimate the time-to-adoption in a hazard model framework. She found that the likelihood of adoption increases as children become eligible for adoption subsidies. Moreover, she found that adoption subsidies have a larger impact on the likelihood that a child is adopted by an older relative. Both of these results are consistent with our findings.
Theoretical Framework
From the state child welfare agency’s point of view, adoption—because it removes the child from the foster care system and provides a more stable environment—is more desirable than long-term foster care. We assume this point of view even though the foster care agency may lose its maintenance payment when the child is adopted. From the parents’ and child’s point of view, adoption may be emotionally and psychologically desirable (Mulligan 2003). However, parents take on increased financial and legal responsibilities when they adopt, and they must weigh these factors against the potential benefits of adoption.4
For the purposes of this study, we assume that parents act as rational agents with preferences that value their own consumption and the overall well-being of their foster child. This implies that a foster parent will consider both the emotional and financial impacts that an adoption will have on the foster child’s well-being and home environment. Many financial benefits and obligations are altered when a foster child is legally adopted. For instance, although basic expenditures for food, clothing, and other provisions for the child are made by parents regardless of whether the child is adopted, foster parents do receive a basic monthly subsidy for providing care for the foster child. In many states, though, adoptive parents receive a reduced monthly subsidy when they adopt their foster child (Barth 1997). There are also legal costs of adoption, although some states have implemented one-time adoption transfers to help offset these legal expenses. Furthermore, adoptive parents may become financially responsible for the child’s future medical expenses and may be liable for any legal costs or damages caused by the actions of the child.
In light of this, foster parents must weigh the benefits to the child and to themselves that come from adoption against the increased risk and financial obligations. We do, however, assume a downward sloping demand for adoptions: all else being equal, a parent will be more likely to adopt a foster child as the cost of adoption decreases. Although there are many components to the cost of adoption, in this article we focus on the monthly foster care subsidy and the monthly adoption subsidy. Before 1980, foster parents typically forfeited their monthly foster care payments entirely when they adopted their foster child. After passage of the Adoption Assistance and Child Welfare Act of 1980, states began to provide a monthly adoption subsidy, effectively lowering the cost of adoption.
Calculating exactly how much the adoption subsidy lowers the cost of adoption is complicated by the nature of adoptions and by the structure of foster care and adoption payments. At any point in time, the adoption decision faced by foster parents is a dichotomous one: either to continue with the permanent foster care arrangement or to legally adopt the child. However, the timing of the adoption is not dichotomous. In many states, both foster care payments and adoption payments vary by the age of the child. Other states pay a flat rate regardless of the age of the child. Moreover, some states match their adoption subsidies to their foster care payments, and other states set their adoption subsidies below the foster care payments. Our analyses address the following questions: (1) do parents weigh current benefits against current costs, or are they more forward-looking?, and (2) do parents respond to the size of adoption payments, to the difference between adoption and foster care payments, or to both?
Data
Our primary data are from the 1998 and 2006 AFCARS, which contains basic information on all children in foster care in 42 states and the District of Columbia.5 Although AFCARS began in 1995, it was not until 1998 that states faced financial penalties for failure to submit AFCARS data. Thus, the 1998 AFCARS is unique, not only because it is the first year in which we have data on large numbers of foster children, but also because it comes on the heels of the Adoption and Safe Families Act of 1997. We use the 2006 AFCARS data for two reasons. First, many states altered their adoption payments in the intervening eight years (from 1998 to 2006), and our long-difference estimation captures changes in this relatively slow-moving process. Second, as we describe in the following paragraph, information is available for both 1998 and 2006 on the monthly state foster care and adoption maintenance payments for children of all ages in each states. From these data sets, we extract all children between the ages of 4 and 16 who were eligible for adoption in fiscal years 1998 and 2006.6 Following AFCARS guidelines, we consider a child to be eligible for adoption if both parents’ rights have been terminated or if the child’s stated case goal is adoption. Our primary purpose is to understand the factors that influence foster parents’ decision to adopt their foster child, and so we exclude children who are in supervised independent living, are participating in trial home visits with their parents, are in group homes or institutions, or have run away. Only a small percentage of foster children eligible for adoption fall into these placement settings, and many of the records for these children contain incomplete information. Children in an additional 14 states were dropped because the state failed to report key information in 1998, such as the year when the child entered foster care, or because the state does not set a basic monthly foster care payment.7 A small number of additional children were excluded because of missing or impossible age entering foster care, gender, or race. The resulting sample includes 118,452 children (60,038 boys and 58,414 girls) living in 29 states in 1998 and 2006.
Although the AFCARS data identify the state in which the child resides, it contains no information about state policy. Therefore, to measure the effect of economic incentives on the adoption rates of foster children, we link the child-level AFCARS data with measures of the basic monthly foster care and adoption subsidy rates by the child’s state, age, and year. We obtained these data from a variety of sources, including the Child Welfare League of America, the North American Council on Adoptable Children, state statues and publications, and direct contact with state employees. These subsidy rates are basic minimum guidelines, which can be supplemented depending on the child’s needs. They do not include any special subsidies for care by relatives or adoption-related expenses. By 1998, every state had some form of adoption subsidy program. Some states simply match their adoption subsidy payments to their foster care payments. In these states, foster parents will generally receive the same monthly payment if they choose to adopt their foster child, provided that the child meets the state definition of special needs. However, several states negotiate adoption payments with prospective adoptive parents under the condition that the adoption subsidy cannot exceed the foster care payment. In other states, foster parents could, according to the basic guidelines, forgo up to $302 per month in 1998 or $343 per month in 2006 in basic foster care payments by adopting their foster child. This difference varies by state, year, and child’s age.
Figure 1 shows the average monthly foster care payments and adoption subsidies across all states in our sample in 1998 and 2006, by the child’s age. Payments in 1998 have been converted into 2006 dollars using the Consumer Price Index for All Urban Consumers (CPI-U). As shown in Fig. 1, states typically designate age ranges and pay one monthly subsidy for children under the age of 5 or 6, a different (usually higher) rate for children between the ages of about 6 and 12, and a higher rate yet for children over the age of 11 or 12. However, not all states increase their payments with age. Comparing the 1998 rates with the 2006 rates, it appears that, on average, foster care payments kept pace with inflation. In contrast, average adoption payments, in real terms, fell between 1998 and 2006, widening the gap between foster care and adoption payments.
The averages shown in Fig. 1 obscure a significant amount of policy variation between states. Some examples of this variation are shown in Fig. 2, which shows the 1998 and 2006 monthly foster care and adoption payments in six selected states. Some states, such as California, set the adoption payment equal to the foster care payment, indexing both to inflation. Eight additional states (not shown) set the adoption payment equal to the foster care payment in both 1998 and 2006, but unlike California, adjusted the overall level of these rates either up or down. Other states reformed the value of their adoption payments relative to foster care payments between 1998 and 2006. Some of these reforms were modest, such as in Connecticut, where real payments were equal in 1998 but both were reduced across the board by 2006 (although adoption subsidies were reduced more than foster care payments); or in South Carolina, where real payments started off equal but were increased across the board (although foster care payments were increased more than adoption payments). Other states, such as Georgia, moved from a system of equal payments for all children to the more common system that pays different amounts based on the age of the child, with lower payments for adopted children. In all, 14 states in our sample in 1998 matched their adoption payments to their foster-care payment, with five of these states offering different rates by 2006. Although most states lowered adoption subsidies relative to the foster care payments between 1998 and 2006, others, such as New Jersey, increased adoption subsidies relative to foster care payments. In the upcoming empirical analysis, we exploit the within-state variation in payments over time to examine the effect that both the level of foster care and adoption payments and the difference between two has on the adoption rate of eligible children.
Basic Patterns
Table 1 lists the characteristics of adoption-eligible children, by gender, in 1998 and 2006. Table 2 lists the percentage of these eligible children who were adopted from foster care in 1998 and 2006. In both years, nearly an equal number of boys and girls were waiting to be adopted, and the adoption rates for boys and girls were similar. However, the overall adoption rate increased significantly between 1998 and 2006. The adoption rate grew from little more than 17 % of eligible children in 1998 to more than 29 % in 2006. The adoption rate is lower for older children than for younger ones, falling to about 14 % for children aged 12 to 16 in 1998 and 21 % in 2006. Black children are overrepresented among children waiting to be adopted in both 1998 and 2006, although the gap is smaller in 2006. In fact, in 1998, more than one-half of the children waiting to be adopted were black (56 % of boys and 54 % of girls). By 2006, this number had fallen to less than one-third (30 % of boys and 28 % of girls). The adoption rate among black children, about 15 % for both boys and girls in 1998 and about 26 % in 2006, contributes to this overrepresentation. With an adoption rate of about 21 % for both white boys and girls in 1998 and 32 % in 2006, a white child was nearly 1.4 times as likely in 1998 and 1.2 times as likely in 2006 to be adopted from foster care as was a black child.
A foster child is considered disabled in the AFCARS data if he or she has a diagnosed disability, including mental retardation, visual or hearing impairment, physical disability, emotional disability, or other diagnosed disability. The disability rate among children eligible for adoption was 25 % for boys and 20 % for girls in 1998. The number of foster children identified with a diagnosed disability increased sharply in 2006 to 41 % for boys and 34 % for girls. Whereas disabled children had higher than average adoption rates in 1998, they had lower than average adoption rates in 2006.
Certain children may be less suited for adoption because of difficult-to-measure characteristics of the child. To capture some of these characteristics, we construct a dichotomous variable equal to 1 if the child’s behavior or condition was listed as a reason for removal from his or her home. Child-related reasons for removal include child alcohol or drug addiction, child disability, and child behavior problems. In 1998, a child-related reason for removal was reported for approximately 12 % of boys and 11 % girls. These percentages grew slightly in 2006 to 14 % of boys and 12 % girls. The adoption rate among this group was close to the overall rate in 1998 but was lower than the overall average in 2006.
Table 3 lists the characteristics of the foster parents of adoption-eligible children, by the child’s gender; Table 4 reports the adoption rates for these children. In 1998, black parents, older parents, single parents, and parents related to their foster child were the least likely to adopt. For example, 1998 the adoption rate was 7 % among black foster parents, 9 % among foster parents over the age of 51, 4 % among foster parents related to the foster child, and 8 % among single female foster parents. These patterns suggest a possible explanation for the low adoption rate among black children: black children are more likely to be placed with black foster parents, single foster parents, or relatives, all of whom adopt at lower rates. For instance, in other calculations from our 1998 sample, we found that 74 % of black boys were placed with a single foster parent; comparatively, this rate was only 41 % for white boys. Also in 1998, 30 % of black boys and 12 % of white boys were placed with a relative.
Methods
is a dichotomous variable equal to 1 if child i of age j in state k at year t is adopted in fiscal year t, and 0 otherwise. The vector 
represents a set of control variables, including child characteristics (such as the child’s race/ethnicity, months spent in foster care, disability status, and reason for removal) as well as parent characteristics (such as the foster parent’s marital status, age, race/ethnicity, and relationship to the foster child). The basic monthly adoption subsidy and foster care rates, 
and 
, are measured in hundreds of dollars for children of age j living in state k in year t, respectively. Note that these variables are subscripted by k, t, and j because basic foster care and adoption subsidies vary by state, year, and child’s age. However, they are not subscripted by i because basic state payments do not vary among children who are the same age and live in the same state in the same year.The parameters 
in Eq. (1) are 377 state-age fixed effects (13 age categories in 29 states) capturing the influence of any state-specific factors that may or may not vary by age (within each state) that are unobserved but do not change over time. The advantage of using Eq. (1) is that the relationship between payments and adoption rates are identified using only the variation observed within a state-age group.9 That is, 9-year-olds in Connecticut in 1998 are compared with 9-year-olds in Connecticut in 2006, but not with any other group of children either in Connecticut or in any other state. Another advantage of using this level of fixed effect is that it corresponds to the level at which states set their foster care and adoption payments. Estimating the effects of aggregate policy variables on microdata can lead to standard errors that are biased downward (Moulton 1990). As a result, we correct the standard errors for clustering at the state-age group level in all our logit regressions (Bertrand et al. 2004).10
The interpretation of 
in Eq. (1) is the marginal effect of increasing the adoption subsidy, holding constant foster care payments. This is equivalent to increasing the difference between the adoption subsidy and foster care payment, which reflects the amount foster parents must give up to adopt their foster child. From a financial point of view, this is an important calculation for foster parents because it represents the cost of becoming adoptive parents compared with continuing as foster parents.11 Because increasing the adoption payment while holding constant the foster care payment incentivizes adoptions over long-term foster care, we expect 
to be positive. However, for individuals who are considering becoming foster parents, the overall level of payments, rather than the difference between the two payments, may be the relevant calculation. The effect of increasing the level of adoption payments, holding constant the difference in payments, is captured in Eq. (1) by 
. This reflects the overall generosity of the state child welfare system. In theory, simultaneously increasing the generosity of foster care and adoption payments could increase or decrease the adoption rate. On one hand, it lowers the cost of becoming an adoptive parent. On the other hand, it lowers the cost of becoming a long-term foster parent. In effect, increasing the generosity of payments can attract more individuals planning to be adoptive parents, but it can also attract more individuals planning to be long-term foster parents. Presumably, state child welfare agencies are able to sort this out and identify which parents are more likely to become adoptive parents. If so, we would expect 
to be positive, but this is ultimately an empirical question.
captures the effect of changing the difference in payments, and 
captures the effect of changing the overall generosity of payments.Findings
Marginal Effects
Table 5 lists our estimates of Eq. (2) separately for boys and girls, with the logit coefficients converted into average marginal effects so that they can be compared directly with the summary statistics given in Tables 2 and 4.12 The average marginal effects labeled as “difference in payments” measure the association between a $100 increase in basic monthly adoption subsidy, holding constant foster care payments, and the adoption rate (
). The average marginal effects labeled as “generosity of payments” measure the association between a $100 increase in both the foster care and adoption payments and the adoption rate (
). Increasing the adoption payments by $100 per month, holding constant foster care payments, is associated with a 4.6-percentage-point increase in the adoption rate of boys and a 5.9-percentage-point increase in the adoption rate of girls. This corresponds to a 27 % increase in the adoption rate for boys and a 34 % increase in the adoption rate of girls compared with their 1998 rates. Increasing the generosity of payments by increasing both the adoption and foster care payments by $100 per month is associated with a 3.7-percentage-point increase in the adoption rate of boys and a 6.2-percentage-point increase in the adoption rate of girls. This corresponds to a 17 % increase in the adoption rate for boys and a 35 % increase in the adoption rate of girls compared with their 1998 rates.13 The positive effect of the generosity of payments on the adoption rate is consistent with the idea that lowering the cost of foster care and adoption will increase the number of families willing and financially able to take in a child, making it easier for state child welfare agencies to find adoptive homes. Moreover, the fact that increasing the adoption subsidy, holding constant foster care payments, is associated with an increase in the adoption rate is consistent with the idea that reducing the financial penalty associated with switching from a foster parent to an adoptive parent will make it more likely that foster parents will adopt their foster children.
Foster Child and Parent Characteristics
According to the summary data presented in Table 2, male black children were 6.75-percentage points (21.02 – 14.27) less likely be adopted than male white children in 1998 and 5.85 percentage points less likely to be adopted in 2006. However, the marginal effects presented in Table 5 reveal that the racial differences in adoption disappear when controls for parent and child characteristics are added. Although we use a different sample of foster children, this is consistent with Duncan and Argys (2008), who found that nearly all the difference in the adoption rates between black and white children can be explained not by differences in the characteristics of children but by differences in the characteristics of the foster parents. In fact, a significant reason why black children have relatively low adoption rates is that black children are more likely to be taken in by relatives, and relatives are less likely to adopt their foster child.
Table 5 also reveals that, with the inclusion of controls for other factors, disabled children are less likely to be adopted as are children with a child-related reason for removal. These variables may identify difficult-to-place children. Months spent in foster care, which measures the time since the child’s most recent home removal, is also positive and statistically significant. All three of these variables are potentially endogenous, but omitting them as control variables from the regression does not change the coefficients on the subsidy variables in any meaningful way.
Marginal Effects by Child and Parent Characteristics
In addition to estimating the overall effect of lowering the cost of adoption on the adoption rate, we also estimate models that allow these effects to vary by the characteristic of the child and foster parent. This is accomplished by estimating separate logit regressions of Eq. (2) that include interaction terms between the specific child or parent characteristic and the foster care and adoption payment variables. For example, we estimate a model that interacts the difference in payment and the generosity of payment variables with three age categories to examine whether lowering the cost of adoption has different effects on older children than on younger ones. We also estimate models that interact the payment variables with the child’s race/ethnicity, disability status, and reason for removal, as well as with the foster parents’ marital status, age, race/ethnicity, and relation to the child.14
Tables 6 and 7 present marginal effects calculated separately for various child and parent characteristics, respectively. All regressions include state-age fixed effects and the same control variables as those presented in Table 5.
The first row in Table 6, labeled “overall,” is the same regression model presented in more detail in Table 5. The remaining marginal effects reported in Table 6 suggest that lowering the cost of adoption increases the adoption rate of all children by similar amounts, regardless of the child’s observed characteristics. Possible exceptions include the adoption rates of black girls and disabled boys. For example, a $100 increase in the adoption subsidy, holding constant the foster care payment, is associated with a 5.9-percentage-point increase in the adoption rate of white girls and a 7.6-percentage-point increase in the adoption rate of black girls. With a t value of 2.15, this difference is statistically significant at the 5 % significance level. A $100 increase in the generosity of payments is associated with a 1.9-percentage-point increase in the adoption rate of disabled boys but a 4.5-percentage-point increase in the adoption rate of nondisabled boys. The difference is also statistically significant at the 1 % level (t = 4.46). However, the remaining marginal effects presented in Table 6 show that lowering the cost of adoption has fairly consistent effects across children with different child characteristics. For example, increasing the adoption subsidy by $100, holding constant the foster care payment, has a slightly larger effect on girls aged 4 to 6 (6.8 percentage points) than on girls aged 12 to 16 (5.1 percentage points). However, this difference is not statistically significant. Reducing the cost of adoption has a slightly smaller effect on troubled children than those without a child-related reason for removal, but again, these differences do not reach conventional levels of statistical significance.
Conversely, the marginal effects presented in Table 7 suggest that lowering the cost of adoption affects some foster parents differently than others. Specifically, reducing the cost of adoption has a larger effect on older foster parents, black foster parents, and those related to the foster child. For example, a $100 increase in the adoption subsidy is associated with a 5.2-percentage-point increase in the adoption rate of boys placed with black foster parents, compared with a 2.9-percentage-point increase in the adoption rate of boys placed with white foster parents. The difference, 2.2, is statistically significant at the 5 % level. A similar pattern is seen for girls. The last model in Table 7 indicates that a $100 increase in the adoption subsidy is associated with a 15.8-percentage-point and a 16.5-percentage-point increase in the adoption rates of children placed with related individuals for boys and girls, respectively. These are the largest marginal effects we calculate for any parent or child characteristic.
Conclusions
Beginning with the Adoption Assistance and Child Welfare Act of 1980, states began to offer adoption subsidies designed to lower the cost of adopting a child out of foster care. Some states went so far as to offer a monthly adoption subsidy that is equal to the child’s foster care payment. Other states set adoption subsidies below the foster care payment. Over the years, many states have revised their foster care and adoption payment rates, with some increasing or decreasing both rates overall, and others increasing or decreasing one rate relative to the other. In this article, we examine the effect that lowering the cost of adoption has on the adoption rate of foster children waiting to be adopted. We consider the effect of lowering the cost of adoption in two ways: by increasing the adoption payments while holding constant the foster care payments, and by increasing the generosity of both payments. Using the 1998 and 2006 AFCARS data, we find that lowering the cost of adoption increases adoption rates through both channels. First, increasing adoption payments, holding constant foster care payments (effectively reducing the difference between foster care and adoption payments), makes adoption more attractive to a current foster parent. Second, increasing the generosity of both payments makes foster care and adoption more attractive to prospective parents, making it easier for child welfare agencies to place children in homes that are more likely to adopt.
Our difference-in-differences estimation strategy identifies the relationship between adoption rates and foster care/adoption subsidies using the variation observed between two points in time within a state-age group. Any unobserved time-invariant factors specific to children of a particular age in a given state are implicitly accounted for. However, an important limitation of this study is that we are unable to control for state-specific trends in the adoption rate and in the foster care/adoption subsidies. Furthermore, although we find that lowering the cost of adoption has a slightly larger effect on foster parents who are related to their foster child compared with those who are unrelated, states differ in their policies toward kinship care. Further research is needed to investigate the role that such policies have on the adoption rates of children from kinship care.
Although lowering the cost of adoption appears to have similar effects on the adoption rates of children regardless of the individual characteristics of the child, it has a stronger impact on children placed in some homes than in others. For example, reducing the cost of adoption has a relatively larger impact on the adoption rates of older foster parents, black foster parents, and foster parents who are related to their foster child. Moreover, these are the foster parents who, in 1998, had the lowest adoption rates.
Acknowledgments
This study was funded by The National Institute of Child Health and Human Development (1 R03 HD049867-02). We would like to thank Christopher Clark for his excellent research assistance.
Notes
The U.S. Department of Health and Human Services, Children’s Bureau, defines children waiting to be adopted as those whose parents’ rights have been terminated and/or with a stated case goal of adoption.
Previous studies have found that economic incentives alter family decisions, such as fertility (Acs 1996; Grogger and Bronars 2001; Kearney 2004), unwed births (Hoffman and Foster 2000), and abortions (Argys et al. 2000).
The AFCARS data contain information about the foster parents’ age, race/ethnicity, marital status, and whether they are related to the foster child. Unfortunately, the data do not contain any additional demographic information about the foster parents, such as income, education, or number of children (including other foster children) in the household.
In addition to long-term foster care, other possible outcomes for parents who fail to adopt their foster child include the child welfare agency removing the child to place in a pre-adoptive home, or the child running away.
Kentucky, Massachusetts, Nebraska, Nevada, New Hampshire, Ohio, South Dakota, and Tennessee are not included in the 1998 AFCARS data.
We exclude children under 4 because in many states, they may not qualify as special-needs and would therefore not be eligible for the adoption subsidy.
Children in Colorado, Indiana, Kansas, New York, and Pennsylvania are excluded because a basic monthly subsidy rate is not set by the state. Children in Alabama, Alaska, Arizona, Arkansas, Delaware, the District of Columbia, Florida, Michigan, and New Mexico are excluded because of missing data in the 1998 AFCARS data.
Although the adoption rates are similar among boys and girls, we are nevertheless concerned that lowering the cost of adoption, as well as other demographic controls, may have different influences on the adoption rates of boys and girls.
An intermediate step would be to simply include 29 state fixed effects and 13 age fixed effects in Eq. (1). State fixed-effects regressions produce similar estimates to the state-age fixed-effects regressions reported in this article (see footnote 13).
Our policy of interest varies at the state-age level, and so we include state-age fixed effects and cluster the standard errors at the state-age level. As a robustness check, we also estimate models with state-age fixed effects, but with standard errors clustered at the state level. For both boys and girls, clustering at the state level approximately doubles the standard errors reported here. When clustered at the state level, the “difference in payments” marginal effects remain statistically significant, but the “generosity of payments” marginal effects become statistically insignificant.
If parents are forward-looking, it may not be current payments that matter but rather the value of all future payments. To investigate this possibility, we estimated additional regressions that include the net present value of basic foster care and adoption subsidies in place of the monthly rates. These regressions yield similar results as those reported in this article, which is not surprising given that the present values and the monthly rates are highly correlated and that state-age fixed effects are included in the regressions.
Ordinary least squares estimates of Eq. (2) produce nearly identical marginal effects.
As a robustness check, we estimated several alternative models that use different identification strategies. For example, models that include state fixed effects and age fixed effects separately, rather than the state-age fixed-effects models presented here, produce similar results. Specifically, the corresponding “difference in payments” marginal effects are .040 for boys and .050 for girls. The corresponding “generosity of payments” marginal effects are .028 for boys and .040 for girls. In addition, we estimated models that include state-year fixed effects. These models use only the variation in payments within a state-year by child age, and produce marginal effects that are smaller in magnitude and are not statistically significant.
The resulting marginal effects reported in Tables 6 and 7 can be interpreted as if they were produced from 22 separate estimates of Eq. (2): one overall, and 21 more for each child and parent characteristic (i.e., a regression for children aged 4–6, another for children aged 7–11, another for white children, for black children, and so on). Limited by sample sizes, we actually estimated nine regressions: one overall, and eight that include interactions with the child characteristics (age, disability, race/ethnicity, and reason for removal) or the foster parent characteristics (age, marital status, race/ethnicity, and whether related to child). Each regression was estimated separately for boys and girls.
