The Cumulative Prevalence of Termination of Parental Rights for U.S. Children, 2000–2016

Abstract

Recent research has used synthetic cohort life tables to show that having a Child Protective Services investigation, experiencing confirmed maltreatment, and being placed in foster care are more common for American children than would be expected based on daily or annual rates for these events. In this article, we extend this literature by using synthetic cohort life tables and data from the Adoption and Foster Care Analysis and Reporting System to generate the first cumulative prevalence estimates of termination of parental rights. The results provide support for four conclusions. First, according to the 2016 estimate, 1 in 100 U.S. children will experience the termination of parental rights by age 18. Second, the risk of experiencing this event is highest in the first few years of life. Third, risks are highest for Native American and African American children. Nearly 3.0% of Native American children and around 1.5% of African American children will ever experience this event. Finally, there is dramatic variation across states in the risk of experiencing this event and in racial/ethnic inequality in this risk. Taken together, these findings suggest that parental rights termination, which involves the permanent loss of access to children for parents, is far more common than often thought.

Introduction and Background

A recent series of articles has used a common demographic technique—the synthetic cohort life table—to estimate the cumulative prevalence of having a Child Protective Services (CPS) investigation (Kim, Wildeman, Jonson-Reid, & Drake, 2017), a confirmed maltreatment case (Wildeman et al., 2014), and being placed in the foster care system (Wildeman & Emanuel, 2014) at the national level for American children. These methods have also been used to generate estimates of the cumulative prevalence of foster care placement in Denmark (Fallesen, Emanuel, & Wildeman, 2014) and in Cuyahoga County, OH, to age 10 (Sabol, Coulton, & Polousky, 2004). Birth cohort life table methods, which also shed light not on the proportion who experience a CPS event in any given year but over a childhood, have been used to generate estimates to age 5 in California (Putnam-Hornstein, Needell, King, & Johnson-Motoyama, 2013), to age 6 in the United States (Magruder & Shaw, 2008), and to age 18 for all children in New Zealand (Rouland & Vaithianathan, 2018). The findings from these articles all echo the same theme: CPS contact is extremely common across contexts, and although there is inequality in this contact, it is common even for more advantaged groups.

In this article, we take another step in this new literature by using synthetic cohort life tables and administrative data on the entire population of children who have been in foster care in the United States since 2000 to generate the first estimates of the cumulative prevalence¹ of experiencing termination of parental rights for children in the foster care system. And, in so doing, we answer the following question: What percentage of American children would ever have parental rights terminated if the rates in any given year held constant throughout their entire childhood? Although the study of the termination of parental rights is new to demographic research² and the estimates we present have never been calculated before, the study of the termination of parental rights is hardly new to scholars of child welfare. Some of this emphasis in the child welfare research community is driven by the sheer number of children who are in foster care awaiting adoption, something that can only occur once all living parents have had their parental rights terminated. According to the most recent estimates, roughly 30% of children in the foster care system were awaiting adoption (U.S. Department of Health and Human Services, 2017, p.1). Yet some of this emphasis is driven not by the number of children at risk but by the finality of the event itself. Termination of parental rights, though significantly less common than other stages of CPS contact, is likely far more consequential because it signals the end of attempts to reunify parents and children and, with the exception of the few children for whom this is not the case goal, leads to immediate attempts to place children in adoptive homes. As such, although scholars of CPS focus on many stages, a group of researchers has been especially fixated on this stage (e.g., Bartholet, 1999, 2000; Guggenheim, 2000; Roberts, 2002).

The results from our analyses of this important level of child welfare contact provide support for four core conclusions. First, according to the most recent estimate for 2016, 1 in 100 American children will experience the termination of parental rights. This represents a roughly doubling of the risk of cumulative prevalence from 2000. Second, the risk of experiencing this event is highest in the first few years of life. Third, risks are highest for Native American and African American children. Nearly 3.0% of Native American children and around 1.5% of African American children will ever experience the termination of parental rights based on 2016 estimates. American Indian and Alaska Native children are 2.7 times more likely than White children to ever experience the termination of both parents’ rights, and African American children are 2.4 times more likely than White children to experience the termination of parental rights. Finally, there is dramatic variation across states both in the risk of experiencing this event for the total population of children and in racial/ethnic inequality in this risk; children in some states have 6–7 times the risk of having their parental rights terminated as children who live in states with the lowest rates of termination of parental rights. Taken together, these findings suggest that parental rights termination is sufficiently common that it merits attention not only in the child welfare community but also in the broader social science research community.

Data, Analytic Strategy, and Progression of Results

Data

Our analysis is built around 17 distinct years of the Foster Care file from the Adoption and Foster Care Analysis and Reporting System (AFCARS) Data, which are collected by the Children’s Bureau and archived and distributed by the National Data Archive on Child Abuse and Neglect at Cornell University. According to the Children’s Bureau, the “AFCARS collects case-level information on all children in foster care and those who have been adopted with title IV-E agency involvement.” As such, these data include information on every American child who has been placed in the foster care system at any point between the beginning of fiscal year (FY) 2000 and the end of FY 2016 (with the exception of a small number of Native American children who are members of tribes who do not have a Title IV-E agreement with the U.S. government).³

For the purposes of our analysis, these data sets include a series of four variables that are necessary for our analysis: (1) child’s age, (2) date of termination of maternal rights, (3) date of termination of paternal rights, and (4) date of termination of parental rights for the second parent to experience it. For our analyses, it is especially important that we know whether parental rights termination happened in the last year and the child’s age because age-specific first event rates form the core of our life table calculations. Although parental rights can be reinstated in some instances after they have been terminated (e.g., O’Donnell, 2010; Taylor, 2009), reinstatement is sufficiently rare that we can comfortably assume that all terminations of parental rights we observe in the data are first terminations without substantively altering any of our conclusions.

Because our population includes only children who experienced foster care placement at some point, children in the general population who experience the termination of parental rights for only one parent but are not exposed to foster care are excluded from our analysis.⁴ This could include, but is not limited to, parents who voluntarily sign over their parental rights or who lose their parental rights in divorce or child support proceedings. Because family court records are generally sealed, we cannot provide—and found no other research that could provide—estimates of the number of children who experience the termination of parental rights for one parent but never experienced foster care placement. Although this likely represents a nonnegligible number of children, this population does not experience the same magnitude of family disruption as children who experience the termination of both parents’ rights while in state custody.

Analytic Strategy

Following previous research estimating the cumulative prevalence of CPS contact at the national level (e.g., Kim et al., 2017; Wildeman & Emanuel, 2014; Wildeman et al., 2014) and at lower levels of aggregation (e.g., Sabol et al., 2004), we use synthetic cohort life tables to estimate the cumulative risk of termination of parental rights. Synthetic cohort life tables can be used to estimate the cumulative risk of experiencing any event over a lifetime (or childhood) based on age-specific first event rates for a range of ages over any period (e.g., Preston, Heuveline, & Guillot, 2000, 38–68).⁵

Four caveats merit mentioning. First, although synthetic cohort life tables can be sensitive to yearly fluctuations with the risk of experiencing some event, this limitation can be overcome by using many years of data so that years with unusual cumulative prevalence rates are noticeable. In this instance, we overcome this obstacle by producing 17 different synthetic cohort life tables, one for each year from 2000 to 2016. Second, there are some missing data on race/ethnicity in the AFCARS. In order to address this, we construct multiple imputation models to address missing information. Uncertainty intervals are included in the figures below. Because of the extremely large number of cases in the data and the extremely low rate of missing data on race/ethnicity,these uncertainty intervals are small. Third, all pooled analyses are weighted at the child level. Although the results are not sensitive to this decision, it is generally considered a more appropriate method for estimating disparities when populations could be changing rapidly—as is the case for some relatively small populations (e.g., Asians or Native Americans) in some states. Finally, though maybe most importantly, state-year estimates for relatively small population groups (e.g., 2010 estimates for Native Americans in Vermont) are likely to be quite unstable, even though we use population data. We have addressed this issue by (1) presenting pooled results for all 17 years for the state-level analyses and (2) including Table D1 in the online Appendix, which shows what we consider reasonable lower (−1 standard deviation [SD] from the estimate) and upper (+1SD from the estimate) bounds to highlight very sensitive estimates.

Because the identification numbers for children are unique within but not between states and children may experience various CPS events (e.g., investigation) for a first time in multiple states, estimates of the cumulative prevalence of CPS events may be biased.⁶ Although this is a concern with all analyses of state-level data, it is a small concern for these analyses because the risk of experiencing the termination of parental rights in multiple different states is likely small.

Progression of Results

The results from these analyses proceed in three stages. In the first stage, we present national estimates of the cumulative prevalence of termination of parental rights by age 18 for all American children and for five racial/ethnic groups (Native American, African American, Hispanic, White, and Asian American/Pacific Islander) for each year from 2000 to 2016. In the second stage, we present pooled age-specific risks of parental rights termination for the same five racial/ethnic groups. In the third stage, we present pooled state-level estimates of the cumulative prevalence of termination of parental rights by age 18 for all American children and the five racial/ethnic groups, as well as racial and ethnic inequalities in the cumulative prevalence of termination of parental rights.

Results

National Cumulative Prevalence Estimates by Race/Ethnicity

Figure 1 reports estimates of the cumulative prevalence of the termination of parental rights by race/ethnicity for each year from 2000 to 2016. Table A3 provides these estimates in tabular format, as well as including confidence intervals for all estimates. These estimates are presented at the national level, providing the first indication of what share of American children will ever have parental rights terminated. As research indicates that the cumulative prevalence of foster care placement for American children is around 6%, all values for the cumulative prevalence of the termination of parental rights have that as a ceiling (Wildeman & Emanuel, 2014).

Figure 1.

National cumulative prevalence of termination of both parents’ rights by race/ethnicity, 2000–2016 annual synthetic cohorts

As Figure 1 indicates, the cumulative prevalence of having parental rights terminated for both parents was 0.7% in 2000. It then increased to just under 1.0% in 2007 before decreasing between 2007 and 2012, ultimately falling to 0.9%. Starting in around 2012, the rate of the termination of parental rights started to accelerate, reaching a high of around 1.1% by the end of the study period in 2016. This 0.4% increase is equivalent to a 60% increase from 2010 to 2016.

At the beginning of the period, African American children had the highest risks of having parental rights terminated at 1.9% (for both parents) to 2.2% (for either parent; Table A3). The risks then declined for this group, with the exception of an uptick in the risk of experiencing this event from 2014 to 2016. By the end of the study period, African American children had lower risks of having parental rights terminated at 1.7–1.8%.

Native American children experienced a different shift in the risk of experiencing the termination of parental rights. At the beginning of the period, Native American children had risks of 1.1% (for both parents) to 1.5% (for either parent; Table A1). After an increase from 2001 to 2007, a decrease from 2007 to 2012, and another increase from 2012 onward, Native Americans had higher cumulative risks of having parental rights terminated than all other groups. Based on 2016 termination of parental rights rates, Native American children could expect to have anywhere from a 2.7% to a 2.9% risk of having parental rights terminated.

White, Hispanic, and (especially) Asian/Pacific Islander children had lower risks of experiencing the termination of parental rights throughout the period, as well as experiencing smaller changes in the risk of experiencing this event over this period. At the beginning of the study period, Whites and His-panics had risks of 0.5–0.6% (for Whites) to 0.6–0.8% (for Hispanics); by the end of the period, these risks had increased for Hispanics to 0.9% and doubled for Whites to 1.0–1.1 %. The risks for Asian/Pacific Islander children were just over 0.2%. According to our results, therefore, Native American children had 14 times the risk of experiencing the termination of parental rights as did Asian/Pacific Islander children in 2016.

Using similar synthetic cohort designs, Table 1 compares the risk of parental rights termination with other forms of CPS involvement (Kim et al., 2017; Wildeman & Emanuel, 2014; Wildeman et al., 2014) and presents information on racial disparities in cumulative prevalence of CPS contact by type of contact. According to the estimates presented in Table 1, levels of disproportionality in CPS contact between African American children and White children are consistent across levels of contact. This is not the case for Native American children. According to Table 1, Native American children are less likely to experience a CPS investigation than White children but 3 times as likely to experience foster care placement and the termination of parental rights. Thus, while racial disproportionality in CPS contact is consistent across stages for African Americans, it is variable for Native Americans.

Table 1.

Cumulative Prevalence and Racial Disparities in the Cumulative Prevalence of Experiencing Four Levels of CPS Contact for American Children.

	CP			Racial Disparity in CP
	African American (AA)	Native American (NA)	White (W)	AA/White	NA/White
Level of contact (%)
Investigationa	53.0	23.4	28.2	1.88	0.83
Substantiationb	20.9	14.5	10.7	1.95	1.36
Placementc	11.0	15.4	4.9	2.24	3.14
Terminationd	1.7	2.7	1.0	1.70	2.70

Source: Adapted from Yi and Wildeman (2018, p. 44) to include terminations.

Note. CPS = Child Protective Services; CP = cumulative prevalence; AA = African American; NA = Native American; W = White.

^aEstimates based on 2014 synthetic cohorts (Kim et al., 2017, p. 277).

^bEstimates based on 2011 synthetic cohorts (Wildeman et al., 2014, p. 706).

^cEstimates based on 2011 synthetic cohorts (Wildeman & Emanuel, 2014, p. 3).

^dEstimates based on 2016 synthetic cohorts.

Age-Specific National Estimates by Race/Ethnicity

Figure 2 reports estimates of the age-specific risks of the termination of parental rights by race/ethnicity using pooled data from all 17 years. Table A4 reports the estimates presented in Figure 2. We do not make yearly estimates available in a table because there is sufficient instability in some yearly estimates that could be misleading if interpreted in isolation.

Figure 2.

Age-specific national estimates of probability of termination of both parents’ rights by race/ethnicity, 2016 synthetic cohort.

Consistent with the age patterning of the cumulative risk of having a CPS investigation (Kim et al., 2017), experiencing confirmed maltreatment (Wildeman et al., 2014), and being placed in foster care (Wildeman & Emanuel, 2014), the risk is highest in the first yearof life. According to Figure 2, roughly 0.2% of all American children will experience the termination of parental rights in their first year of life. The risk is higher for children from minority groups; 0.4–0.5%of American Indian/Alaska Native children will experience this event in the first year of their life. This is roughly half as high as the risk White and Hispanic children experience over their entire childhoods and is actually nearly twice as high as the risk Asian/Pacific Islander children experience over their entire childhoods. With the exception of the elevated risks in infancy, especially for American Indian/Alaskan Native children, the only noteworthy point from Figure 2 is that most groups experience a slight increase in the risk around ages 13 or 14.

State Cumulative Prevalence Estimates by Race/Ethnicity

Figure 3 reports estimates of the cumulative prevalence of having parental rights terminated for both parents by race/ethnicity and state using pooled data from all 17 years. We focus on estimates of the cumulative prevalence of having parental rights terminated for both parents in the interest of presenting the more conservative of the two estimates in the body of the article. Table 2 also reports these estimates; Table A5 reports estimates of the cumulative prevalence of having parental rights terminated for either parent in order to provide both an upper bound and a lower bound for the risk. Table D1 shows how unstable these estimates are in specific years.

Figure 3.

Cumulative prevalence of termination of both parents’ rights by race/ethnicity, 2000–2016 synthetic cohort.

Table 2.

Cumulative Prevalence of Termination of Both Parents’ Rights by Race/Ethnicity, 2000–2016 Synthetic Cohort.

	American Indian/AK Native	Asian/PI	Black	Hispanic	White	Total
National	.019	.003	.017	.008	.007	.009
Alabama	.002	.001	.005	.004	.005	.005
Alaska	.052	.005	.033	.008	.010	.021
Arizona	.013	.004	.039	.017	.016	.017
Arkansas	.015	.006	.008	.004	.008	.007
California	.015	.003	.026	.008	.008	.009
Colorado	.007	.002	.015	.007	.005	.006
Connecticut	.005	.001	.019	.013	.005	.008
Delaware	.000	.000	.013	.003	.004	.006
District of Columbia	.002	.002	.014	.002	.000	.010
Florida	.006	.001	.014	.004	.010	.009
Georgia	.002	.000	.006	.002	.005	.005
Hawaii	.101	.015	.019	.004	.009	.014
Idaho	.034	.004	.018	.007	.007	.007
Illinois	.002	.000	.028	.002	.005	.008
Indiana	.023	.001	.022	.005	.007	.009
Iowa	.056	.008	.047	.015	.011	.014
Kansas	.018	.002	.031	.007	.011	.012
Kentucky	.011	.002	.018	.011	.008	.009
Louisiana	.002	.001	.006	.002	.005	.005
Maine	.043	.011	.032	.061	.013	.015
Maryland	.001	.000	.006	.001	.003	.003
Massachusetts	.018	.002	.015	.012	.006	.007
Michigan	.030	.002	.036	.014	.012	.016
Minnesota	.038	.002	.023	.007	.005	.007
Mississippi	.002	.001	.005	.004	.005	.005
Missouri	.008	.002	.019	.007	.010	.011
Montana	.040	.016	.047	.021	.012	.017
Nebraska	.037	.004	.030	.010	.010	.012
Nevada	.012	.004	.029	.007	.013	.012
New Hampshire	.012	.001	.012	.006	.003	.004
New Jersey	.005	.001	.033	.005	.005	.010
New Mexico	.005	.002	.018	.008	.007	.008
New York	.002	.000	.019	.007	.004	.007
North Carolina	.007	.002	.010	.003	.006	.006
North Dakota	.026	.017	.027	.019	.008	.011
Ohio	.015	.001	.020	.008	.006	.009
Oklahoma	.035	.011	.043	.016	.014	.020
Oregon	.039	.004	.043	.008	.012	.013
Pennsylvania	.004	.001	.017	.010	.005	.007
Rhode Island	.022	.007	.035	.016	.010	.014
South Carolina	.001	.001	.007	.004	.005	.006
South Dakota	.033	.006	.023	.012	.006	.011
Tennessee	.005	.001	.007	.006	.007	.007
Texas	.006	.001	.022	.009	.009	.010
Utah	.013	.003	.018	.009	.005	.006
Vermont	.007	.001	.019	.012	.020	.019
Virginia	.003	.001	.006	.003	.003	.004
Washington	.034	.003	.025	.009	.011	.012
West Virginia	.015	.012	.048	.018	.020	.021
Wisconsin	.018	.002	.031	.007	.004	.008
Wyoming	.004	.004	.031	.007	.008	.008

Note. AK Native = Alaska Native; PI = Pacific Islander

As Figure 3 and Table 2 indicate, there is considerable variation across states in the cumulative probability of experiencing the termination of parental rights. Three states have cumulative prevalences of experiencing this event in excess of 2.0%: Alaska (2.1%), West Virginia (2.1%), and Oklahoma (2.0%). The state with the lowest cumulative risk of termination of parental rights is Maryland at 0.3%, although a number of other states have risks just above that level. Thus, children in states with the highest risks of experiencing the termination of parental rights are 6–7 times more likely to experience this event than are children in states with the lowest risk of experiencing this event. Although other regions of the country also have some clustering of states with low cumulative prevalence of the termination of parental rights, the Southeast especially stands out as having consistently low risks of termination of parental rights, with no states breaking the 1.0% cumulative prevalence threshold in this region.

American Indian/Alaska Native children have the most noteworthy variation across states with the risk of having parental rights terminated for both parents. As Figure 3 and Table A5 indicate, there are a number of states with very low cumulative risks for these children; Alabama, Georgia, Illinois, and New York all have cumulative prevalence rates of roughly 0.2%. Yet the high for these children is nearly 30 times as high as the risks in these states at 5.6% in Iowa. And there are three other states in which over 4% of American Indian/Alaska Native children will ever experience the termination of parental rights for both parents: Montana (4.0%), Maine (4.3%), and Alaska (5.2%). This pattern of concentration of risks at extremely high levels and extremely low levels is a unique feature of American Indian/Alaska Native children’s risks.

African American children experience a somewhat similar pattern in terms of the risk of experiencing the termination of parental rights to American Indian/Alaska Native children, although the range of their risks is more muted, especially at the lower end. Very few states have low risks of parental rights termination for African American children, with only Alabama and Mississippi having cumulative risks in the 0.5% range or below. Yet many states have very high cumulative risks of the termination of parental rights for African American children, with five states having prevalence rates above 4.0% (Oklahoma at 4.0%, Oregon at 4.3%, Iowa at 4.7%, Montana at 4.7%, and West Virginia at 4.8%). As we noted above, African American children have low risks of experiencing the termination of parental rights in the Southeast. Indeed, with the exception of Florida (at 1.4%), African American children living in the Southeast do not have risks of parental rights termination in excess of 1.0% in any of the Southeastern states.

Hispanic children experience a somewhat unusual pattern of risk across states, with one state having an extremely high risk (Maine at 6.6%) and the state with the next highest risk falling at just over 2.0% (at 2.1% in Montana). Hispanic children also experience low cumulative prevalence rates in many states, with 11 states having rates of under 0.4%. White children and Asian Pacific Islander children experience relatively high cumulative risks of having parental rights terminated in a very small number of states—White children have risks of 2.0% in West Virginia and Vermont, and Asian/Pacific Islander children have risks of 1.7% in North Dakota, 1.6% in Montana, and 1.5% in Hawaii—but their rates of termination of parental rights are low elsewhere, generally not exceeding 0.3% or 0.4% in most other states.

Figure 4 and Table 3 display the ratio of the cumulative risk of parental rights termination for children of color relative to White children across the states using a pooled synthetic cohort. In all but two states (Tennessee and Vermont), African American children are at higher risk of having their parents’ rights terminated than are White children. Washington, DC, has the highest level of Black/White inequality in the nation; African American children are 28 times more likely than White children in Washington, DC, to experience the termination of parental rights. In Wisconsin, New Jersey, Illinois, New York, Minnesota, Iowa, South Dakota, and Wyoming, Black children are at least 4 times more likely than White children to see their parents’ rights terminated. American Indian, Alaska Native, and Native Hawaiian children are at higher risk of parental rights termination than are White children in fully 28 states. In Hawaii, Minnesota, South Dakota, Alaska, and Idaho, American Indian and Alaska Native children are at least 5 times more likely that White children to have their parents’ rights terminated. Inequalities for Black and Native American children are clustered in the Midwest and Mountain West; inequalities tend to be lowest in the Southeast. In general, Asian/Pacific Islander and Hispanic children are at or below the risk faced by White children in nearly all of the states.

Figure 4.

Inequality in cumulative prevalence of termination of both parents’ rights by race/ethnicity relative to White children, 2000–2016 synthetic cohort.

Table 3.

Inequality in the Cumulative Prevalence of Termination of Both Parents’ Rights by Race/Ethnicity, 2000–2016 Synthetic Cohort.

	American Indian/AK Native	Asian/PI	Black	Hispanic
National	2.71	0.43	2.43	1.14
Alabama	0.40	0.27	1.18	0.95
Alaska	5.02	0.52	3.16	0.81
Arizona	0.80	0.27	2.35	1.02
Arkansas	2.03	0.83	1.13	0.55
California	1.80	0.30	3.11	1.00
Colorado	1.52	0.35	3.03	1.39
Connecticut	1.05	0.21	3.77	2.60
Delaware	0.00	0.07	3.26	0.81
District of Columbia	3.14	3.13	28.36	3.78
Florida	0.54	0.14	1.32	0.37
Georgia	0.38	0.10	1.25	0.49
Hawaii	10.82	1.56	2.06	0.45
Idaho	5.01	0.67	2.72	0.99
Illinois	0.32	0.07	5.69	0.38
Indiana	3.14	0.17	2.96	0.67
Iowa	4.96	0.74	4.13	1.35
Kansas	1.55	0.18	2.75	0.63
Kentucky	1.37	0.23	2.24	1.45
Louisiana	0.40	0.13	1.06	0.39
Maine	3.28	0.80	2.45	4.67
Maryland	0.49	0.14	2.29	0.45
Massachusetts	3.06	0.31	2.56	2.04
Michigan	2.54	0.20	3.08	1.20
Minnesota	7.67	0.47	4.67	1.46
Mississippi	0.46	0.28	1.02	0.81
Missouri	0.75	0.24	1.90	0.67
Montana	3.24	1.30	3.83	1.71
Nebraska	3.79	0.45	3.04	1.03
Nevada	0.97	0.31	2.29	0.55
New Hampshire	3.39	0.22	3.59	1.80
New Jersey	0.88	0.11	6.17	0.97
New Mexico	0.83	0.23	2.74	1.27
New York	0.49	0.11	4.99	1.82
North Carolina	1.20	0.36	1.63	0.59
North Dakota	3.37	2.18	3.52	2.50
Ohio	2.39	0.22	3.10	1.24
Oklahoma	2.46	0.77	3.00	1.11
Oregon	3.34	0.32	3.67	0.68
Pennsylvania	0.87	0.20	3.77	2.10
Rhode Island	2.15	0.66	3.45	1.64
South Carolina	0.13	0.20	1.42	0.70
South Dakota	5.78	1.13	4.07	2.06
Tennessee	0.67	0.11	0.88	0.82
Texas	0.69	0.17	2.46	0.99
Utah	2.78	0.59	3.82	1.89
Vermont	0.34	0.07	0.94	0.59
Virginia	1.07	0.19	1.94	0.88
Washington	2.99	0.30	2.25	0.83
West Virginia	0.77	0.61	2.40	0.92
Wisconsin	4.24	0.55	7.17	1.74
Wyoming	0.48	0.53	4.04	0.92

Note. AK Native = Alaska Native; PI = Pacific Islander.

Discussion and Conclusion

Using synthetic cohort life tables, we estimated the cumulative prevalence of parental rights termination nationally, at the state level, and by race/ethnicity for U.S. children from 2000 to 2016. As of 2016, 1 in 100 American children experienced the termination of parental rights before their 18th birthday. In comparison, roughly, 4% of children are separated from parents due to parental incarceration (Sykes & Pettit, 2014), 6% are placed in foster care prior to age 18 (Wildeman & Emanuel, 2014), and 4.5 million children live with one or more undocumented parents and under the threat of separation via deportation (Gulbas & Zayas, 2017). Our findings also show that parental separation in the form of parental rights termination is highly racially disparate, as are many other state-induced causes of parental loss (e.g., Wildeman, 2009).

While the causes and consequences of increasing and racially disparate parental rights termination is beyond the scope of this particular paper, future research should investigate both. Related research on other forms of state intervention in the lives of families provide a useful guide and suggest that the prevalence estimates we have detailed here are consequential for a host of outcomes. Parental incarceration, as one example, affects a similarly small percentage of children (though in a year rather than a childhood) but represents the tip of the iceberg of a broader set of criminal justice and institutional intervention in family life (Enns et al., 2019). Indeed, the incarceration of parents may affect the probability of parental rights termination, in much the same way increases in maternal incarceration are linked to foster care placement (Johnson & Waldfogel, 2002).⁷ The risk of parental incarceration and foster care placement are similarly raced and classed, with a robust literature linking these experiences to a cascade of social disadvantages and racial inequality in child well-being (Roberts, 2012; Sykes & Pettit, 2014). The estimates we present here represent a call for similar investigations of the causes and consequences of the increasing and disproportionate use of parental rights termination policies.

Overall, the risk of parental rights termination is sufficiently high, variable across states, and racially disparate to merit significantly more attention. Moreover, as we demonstrate, stage of CPS contact emerges as acutely important for analyzing racial disparities in child welfare engagement. Our results, taken in context with estimates produced by others, show tremendous variation by race and ethnicity in the likelihood of parental separation once an investigation has begun. Native American children, for example, fall at the midpoint between Whites and African American children for early stage CPS contact (investigation and substantiation) but are more likely to experience later stage separation from a parent than either White or African American children (foster care placement and, especially, termination of parental rights). This is intriguing not only because of what it means about the necessity of looking across stages to develop a nuanced opinion on racial/ethnic inequality in the child welfare system. Indeed, it is also interesting because it indicates that the transition probability for each stage is certain to be higher for Native American children than other children, indicating a different process entirely.

The particularly high rates of parental rights termination experienced by American Indian and Alaska Native children and families, who are afforded special protections under the Indian Child Welfare Act (1978), are particularly striking in light of the relatively low risk of having a CPS investigation they face. Explorations of the relationships between historical trauma (Evans-Campbell, 2008) and institutionalized practices that disrupt Native families (Jacobs, 2014) may reveal underlying causes for these persistent inequalities in a way previous research has not.

Provocative as we believe these estimates are, our analyses nonetheless have numerous limitations. First, these data do not allow for analysis of children whose parental rights are terminated in private family court proceedings that do not involve a child in the formal foster care system, for instance. Second, some terminations are reversed, and so our estimates may slightly overstate the permanent termination of parental rights. Third, there are clearly situations in which children reside with other loved ones after termination of parental rights and, as such, these estimates do not always indicate termination of broader family ties. Finally, although many child welfare analyses rely on the state as the unit of analysis, state-level estimates obscure more microlevel differences, calling for analyses at a lower level of aggregation (e.g., Fong, 2019).

Future work should explore how local and state political conditions that structure broader social policies toward lowincome mothers of color relate to the termination of parental rights (Edwards, 2016; Roberts, 2012). Poverty, child maltreatment, social policy, and foster care placement have a complex set of microlevel and place-level relationships (Wulczyn, Gibbons, Snowden, & Lery, 2013); it is likely that the termination of parental rights is also affected by the strength and structure of the local safety net (Cancian, Mi-Youn, & Kristen, 2013; Wildeman & Fallesen, 2017) and the ability of kin and fictive kin to provide support to families in crisis (Pittman, 2015).