Person-level data were assessed across multiple data sources, including CJARS, the decennial census, the Census Bureau’s Numerical Identification file (Census Numident), and the Census Bureau’s American Community Survey, as well as data from the Internal Revenue Service (IRS), Department of Housing and Urban Development (HUD), and Centers for Medicare and Medicaid Services (CMS).^17–22 Protected Identification Keys do not contain personally identifiable information.^23,24 This study did not constitute human subjects research (Common Rule, 45 CFR Section 46).

Study outcome measures were ACEs incidence by type and the estimated average marginal effect of individual ACEs (for example, parental incarceration) and cumulative ACEs (1, 2–3, 4 or more, or any) on young adult socioeconomic status measures. Analysis was conducted using Stata, version 18. The online appendix details sample construction (appendix exhibits A1–A5); comparison of ACEs incidence measures from this study versus from the Behavioral Risk Factor Surveillance System (BRFSS) and National Survey of Children’s Health (NSCH), which are prominent federal self-reported and caregiver-reported survey sources (appendix exhibit A6); timing of exposure and outcome measures (appendix exhibit A7); survey weights; and complete model results (appendix exhibits A8–A17).²⁵

We acknowledge several limitations. In using administrative and criminal justice data sources, this study applied objective but very conservative measures of ACEs exposure. Associations between ACEs and young adult socioeconomic disadvantages were underestimated because children with ACEs who did not come to the attention of the criminal justice system were included in the control sample (no ACEs), as were children with ACEs linked to non-cohabitating parents. This study assumed that parental felony charges for child abuse while the parent and child cohabitated included the child as a victim. This was a limitation of available data (CJARS does not include victim information) but is reasonable, given that nationwide child welfare administrative data indicate that nearly 90 percent of child abuse victims had a parent perpetrator.³² This study’s measure for parental incarceration included only prison sentences, which may be reasonably interpreted as more disruptive to children’s lives than jail sentences as a result of their longer duration. This study relied on data derived from probabilistic parent-child record linkages, which are widely used but have drawbacks.^24,33 This study’s primary regression models controlled for childhood socioeconomic status using a single year of parental income; parental separation was assessed only in 2010, and study subjects’ employment as young adults was assessed only in 2021. Census Bureau Protected Identification Key methods are known to underrepresent Hispanics because this ethnic population has a higher prevalence of undocumented people and more record linkage challenges such as compound family names (see appendix section A).²⁵

ACEs incidence estimates for parental death (1.6 percent of the study sample) and parental separation (26.0 percent), using this study’s administrative and census data source measures (exhibit 1), were comparable in magnitude to the most recent nationwide self-reported BRFSS and parent-reported NSCH incidence estimates (appendix exhibit A6).²⁵ This study’s parental SUD incidence estimate (14.0 percent) was lower than the comparable measure in the BRFSS but higher than in the NSCH. This study’s estimated incidences of parental incarceration (3.7 percent), witnessing intimate partner violence during childhood (1.0 percent), physical or emotional child abuse (0.5 percent), and sexual abuse (1.7 percent), based on parental felony charges, were lower than comparable estimates in the BRFSS and the NSCH, which both used broader definitions of exposure (exhibit 1 and appendix exhibit A6).²⁵

Controlling for childhood socioeconomic status, young adults with ACEs had a higher probability of having felony charges compared with other young adults (from 11 percent [parental death] to 52 percent [parental SUD] higher probability) (appendix exhibit A17, model 1, and exhibit 2, which visualizes statistically significant model 1 results).²⁵ This association ceased to be statistically significant for study subjects with parental death or parental felony charges for physical or emotional abuse when additional childhood socioeconomic controls were included among the smaller analysis sample with available data (appendix exhibit A17, model 3).²⁵ Young adults with any of the assessed individual ACEs except parental death were significantly more likely to have a teenage birth (from 25 percent [parental incarceration] to 73 percent [parental separation] higher probability; exhibit 2), but the association between childhood physical or emotional abuse and this outcome was sensitive to childhood socioeconomic status (appendix exhibit A17, model 3).²⁵ Study subjects who had parental separation or parental felony charges for sexual abuse during childhood were significantly more likely to have household poverty (27 percent and 43 percent higher probability, respectively) (exhibit 2). All ACEs except parental death (27 percent lower probability) were significantly associated with higher probability of being in a household that received housing assistance in young adulthood (up to 67 percent [parental separation] higher probability), although associations with parental incarceration, physical or emotional abuse, and sexual abuse were sensitive to childhood socioeconomic status (appendix exhibit A17, model 3).²⁵

Each assessed individual ACE was significantly associated with Medicaid enrollment in young adulthood (from 23 percent [witnessing intimate partner violence] to 52 percent [parental separation] higher probability) (exhibit 2). Young adults who experienced parental death or parental felony charges for SUD, child physical or emotional abuse, or child sexual abuse during childhood were slightly less likely to be employed (1–3 percent lower probability) in primary models (exhibit 2), although the associations with parental death and abuse charges were sensitive to childhood socioeconomic status (appendix exhibit A17, models 2 and 3).²⁵ Young adults whose parents separated during childhood were slightly more likely to be employed (2 percent higher probability; exhibit 2), regardless of controls for childhood socioeconomic status (appendix exhibit A17, models 2 and 3).²⁵ Young adults with parental death, parental separation, witnessing intimate partner violence, parental SUD, or parental child sexual abuse were less likely to be enrolled in education than young adults in the primary models (from 11 percent [parental death] to 20 percent [sexual abuse] lower probability) (exhibit 2), although each of these associations, except for parental separation, was sensitive to childhood socioeconomic status (appendix exhibit A17, models 2 and 3).²⁵

Young adults with any ACEs had a higher probability of a felony charge (67 percent), having a teenage birth (98 percent), having household poverty (32 percent), having household housing assistance (75 percent), having Medicaid enrollment (70 percent), and being employed (1 percent), but lower probability of being enrolled in education (−18 percent), compared to young adults with no ACEs (exhibit 3). Study subjects’ number of ACEs was correlated with the magnitude of these estimated associations. For example, having one, two to three, or four or more ACEs compared with no ACEs was significantly associated with a 49 percent (one ACE), 109 percent (two to three ACEs), and 182 percent (four or more ACEs) higher probability of a felony charge, respectively, and having one or two to three ACEs was significantly associated with progressively lower probability (−15 percent and −28 percent, respectively) of education enrollment. Having one ACE was associated with a slightly increased (1 percent) probability of employment, but there was no statistically significant association with higher ACEs counts. In sensitivity analyses, dropping the control for parental income during study subjects’ childhoods did not yield substantively different results (appendix exhibit A17, model 2).²⁵ Additional controls for childhood socioeconomic status among the smaller analysis sample with available data did not change the direction or significance of most measured associations from the primary models, except for the association between any ACEs and modestly higher young adult employment, and one ACE and both employment and education enrollment ceased to be statistically significant with the additional controls (appendix exhibit A17, model 3).²⁵ Overall, the added childhood socioeconomic status controls in model 3 modestly reduced the increased probability of young adult socioeconomic disadvantages associated with ACEs, suggesting a protective effect from more socioeconomic advantages in childhood.