Kindergarten conduct problems are associated with monetized outcomes in adolescence and adulthood
Conflict of interest statement: No conflicts declared.
Abstract
Background
Across several sites in the United States, we examined whether kindergarten conduct problems among mostly population-representative samples of children were associated with increased criminal and related (criminal + lost offender productivity + victim; described as criminal + victim hereafter) costs across adolescence and adulthood, as well as government and medical services costs in adulthood.
Methods
Participants (N = 1,339) were from two multisite longitudinal studies: Fast Track (n = 754) and the Child Development Project (n = 585). Parents and teachers reported on kindergarten conduct problems, administrative and national database records yielded indexes of criminal offending, and participants self-reported their government and medical service use. Outcomes were assigned costs, and significant associations were adjusted for inflation to determine USD 2020 costs.
Results
A 1SD increase in kindergarten conduct problems was associated with a $21,934 increase in adolescent criminal + victim costs, a $63,998 increase in adult criminal + victim costs, a $12,753 increase in medical services costs, and a $146,279 increase in total costs. In the male sample, a 1SD increase in kindergarten conduct problems was associated with a $28,530 increase in adolescent criminal + victim costs, a $58,872 increase in adult criminal + victim costs, and a $144,140 increase in total costs. In the female sample, a 1SD increase in kindergarten conduct problems was associated with a $15,481 increase in adolescent criminal + victim costs, a $62,916 increase in adult criminal + victim costs, a $24,105 increase in medical services costs, and a $144,823 increase in total costs.
Conclusions
This investigation provides evidence of the long-term costs associated with early-starting conduct problems, which is important information that can be used by policymakers to support research and programs investing in a strong start for children.
Introduction
Conduct problems comprise oppositional and antisocial behaviors that violate the rights of others and conflict with societal norms (American Psychiatric Association, 2022; McMahon & Frick, 2019). In children and adolescents, conduct problems are among the most common mental health difficulties experienced. Meta-analytic research (k = 41) has found that the worldwide prevalence of childhood conduct problems diagnostic categories (i.e., disruptive behavior disorders) is estimated to be 5.7% (Polanczyk, Salum, Sugaya, Caye, & Rohde, 2015). Other burden estimates suggest conduct disorder accounts for 5.75 million years lived with disability globally (Erskine et al., 2014). Left untreated, conduct problems confer risk for elevated mental and physical health problems, greater public services utilization, as well as increased likelihood of participation in violent and criminal behavior throughout the lifespan (Bevilacqua, Hale, Barker, & Viner, 2018; Goulter, Oberth, et al., 2022; Odgers et al., 2008; Rivenbark et al., 2018). A distinction is also made between childhood-onset versus adolescent-onset conduct problems. Compared with conduct problems that first manifest in adolescence, early onset childhood conduct problems are linked with more severe risk factors and a chronic trajectory of conduct problems and related adverse outcomes (Fairchild, Van Goozen, Calder, & Goodyer, 2013; Moffitt, 2018). Consequently, investment in prevention of early conduct problems represents an important public health and policy priority.
To increase support for intervention development and implementation for children presenting with conduct problems, policy makers use economic evaluations to determine return on investment to society (Levin, McEwan, Belfield, Bowden, & Shand, 2018). Analyses that monetize outcomes of interest provide a metric for interpreting the fiscal burden of such outcomes, which can be used for planning budgets to support programs. Crime has been one area of interest over the past decade. Using offending populations, several studies have monetized criminal outcomes, consistently finding that a small number of high-rate offenders are responsible for most incurred costs (Allard et al., 2014; Cohen, Piquero, & Jennings, 2010; Day & Koegl, 2019). Despite several studies examining monetized criminal outcomes (for a systematic review, see Wickramasekera, Wright, Elsey, Murray, & Tubeuf, 2015; k = 21), this research has typically sampled offending populations and it is currently unclear how conduct problems among community-representative samples contribute to criminal and related costs.
Studies with community-representative populations have just begun to map the costs of childhood behavior problems. One United States study using the high-risk subsample of children from the present study found that meeting diagnostic criteria for oppositional defiant disorder in 6th or 9th grade was associated with (USD 2000 prices) $32,718 total costs (including health care costs) and $3,540 juvenile legal costs during a 7-year period into late adolescence (Foster et al., 2005). Meeting criteria for the more severe diagnosis of conduct disorder was associated with $76,795 total costs and $15,111 juvenile legal costs. Another study of 83 6- and 7-year-old boys in Britain found that childhood conduct problems were associated with (GBP 2010 prices) £2,670 health and social care costs and £3,561 in adult criminal legal costs 20 years later, representing a three-fold increase compared to control children (d'Amico et al., 2014). Most recently, Rissanen et al. (2021) found in a Finnish sample that high levels of conduct problems in 8-year-old children were associated with (EUR 2016 prices) €44,348 total costs (including health care costs) and €19,288 criminal offense costs through age 30 years. These studies provide a significant advance to our understanding of the costs of conduct problems; however, no studies have tested the association between early-starting conduct problems, such as those in kindergarten, and long-term criminal and related costs.
Theoretical perspectives on the development of criminality emphasize the importance of early-starting conduct problems. Individuals displaying early onset conduct problems are often marked by early-life maltreatment histories, neurological and structural brain differences, and criminogenic environments (e.g. community, peer, or familial crime) resulting in substantial cumulative disadvantage (Carlisi et al., 2020; Moffitt & Caspi, 2001). By contrast, those who first display conduct problems in adolescence are described as engaging in a developmentally normative reaction to the ‘maturity gap’ (i.e., dissonance experienced when youth identify as mature, but legal and social constraints limit access to independence) and they tend to age-out of such behavior before adulthood (Moffitt, 1993; Moffitt, Caspi, Harrington, & Milne, 2002). Just as the comparison between early- and later-starting conduct problems is important for understanding the development of criminality, it may be important to compare estimated costs of offending at different points in development for understanding within-individual variation in offending over the lifespan (Day & Koegl, 2019). Further, certain categories of crime (e.g. violent vs. substance) may differ across developmental stages. Yet, past research has typically reported aggregate costs across life stages and crime types (e.g. d'Amico et al., 2014) rather than considering more specific developmental periods, which could provide cost–benefit analyses for developmentally focused interventions (Day & Koegl, 2019). Another major limitation of past monetization studies is the focus of criminal legal costs to the exclusion of other costs, such as lost productivity of the offender and harm caused to victims (Foster et al., 2005; Rissanen et al., 2021). Costs to the criminal legal system constitute only one aspect of the financial burden of criminal offending, yet loss of offender productivity and victimization costs represent crucial pieces for policy initiatives (Cohen & Piquero, 2009).
A further important consideration in research on conduct problems and criminal offending is the sex of participants. Sex is a predictor of both the prevalence and expression of conduct problems (Goulter, McMahon, et al., 2022; King et al., 2018; Konrad et al., 2021), as well as criminal behavior frequency (D'Unger, Land, & McCall, 2002; Kassing et al., 2019). For example, female offenders are more likely to have single convictions and less likely to engage in criminal careers of more than 10 years, relative to male offenders (Francis, Soothill, & Piquero, 2007). Male individuals are more likely to experience neuropsychological deficits, which can contribute to later antisocial behavior (Mazerolle, Brame, Paternoster, Piquero, & Dean, 2000). Deviant peer affiliation is also a primary risk factor for antisocial behavior, particularly in male samples (Mears, Ploeger, & Warr, 1998). By contrast, sex and gender norms have been considered as crime deterrents among female individuals (Mears et al., 1998; Steffensmeier & Allan, 1996), although more recent research has found that this gender gap may be closing (Schwartz, Steffensmeier, & Feldmeyer, 2009). Much of the past research estimating crime costs has relied on male samples or failed to differentiate based on sex (e.g. Day & Koegl, 2019; Piquero, Jennings, & Farrington, 2013). Thus, costs among female samples are currently unknown.
Monetization studies provide critical information for policy planning. Further research is needed on the long-term relations between early-starting conduct problems and their criminal, victim, government, and medical fiscal toll. The present study addresses five important methodological gaps in the literature. First, there has been a lack of long-term longitudinal data leveraging administrative court information and national criminal databases. We capitalized on longitudinal data by harmonizing two datasets. Participants (N = 1,339) were drawn from two longitudinal multisite studies: the Fast Track project (n = 754; FT; Conduct Problems Prevention Research Group [CPPRG], 2020) and the Child Development Project (n = 585; CDP; Dodge, Bates, & Pettit, 1990). Second, the present study considered parent- and teacher-reported conduct problems in kindergarten (FT = 4.66–7.79 years; CDP = 4.34–6.92 years) and applied costs to two periods relevant for the development of crime: adolescence (or juvenile offending) and adulthood, classifying crimes during these stages into violent, substance, and public order/property offenses. Third, criminal costing research has traditionally monetized crime outcomes to the exclusion of other associated outcomes. In this study, we monetized both criminal and criminal + lost productivity of the offender + victim (described as criminal + victim costs hereafter) outcomes. Given the established link between conduct problems and other social and health maladies (Bevilacqua et al., 2018; Odgers et al., 2008; Rivenbark et al., 2018), we also monetized government and medical service use in adulthood. Fourth, most research examining criminal costs has sampled from offending populations. This sample issue is critical given these individuals only comprise a small proportion of the population. Fifth, taking advantage of the within-sample heterogeneity in our large, harmonized dataset, we also tested whether effects were differentiated by sex, which represents an advance over past research dominated by male samples. Addressing these methodological gaps enables us to advance understanding of how early conduct problems are related to long-term monetized outcomes stemming from distinct types of criminal offenses and associated problems for males and females in community samples. In addition, we consider the implications of our findings in terms of informing future economic evaluation, as well as practice and policy.
Method
Participants and procedure
Participants (N = 1,339) were from two longitudinal studies: FT (n = 754) and CDP (n = 585). Across both studies, legal guardians provided consent and participants assented to procedures. Parents and participants were provided monetary compensation. All procedures were approved by the relevant Institutional Review Board for each site. Data and code for this study are available by emailing the corresponding author. This study was not preregistered.
Fast Track
The FT project is a longitudinal, multisite investigation of the development and prevention of child conduct problems (CPPRG, 2020). Across four sites (Durham, NC; Nashville, TN; rural Pennsylvania; and Seattle, WA), 55 high-risk schools were selected based on crime and poverty rates of the school communities. Within each site, clusters of schools were randomly assigned to intervention and control conditions. In 1991–1993, teachers screened 9,594 kindergarteners for classroom conduct problems using the Teacher Observation of Classroom Adaptation-Revised authority acceptance score (Werthamer-Larsson, Kellam, & Wheeler, 1991). Those children scoring in the top 40% within cohort and site were then solicited for the next stage of screening for home behavior problems by the parents, using 22 items from the Child Behavior Checklist (CBCL; Achenbach, 1991a) and similar scales to capture behavior problems at home; 91% agreed (n = 3,274). Teacher and parent screening scores were standardized within site and summed to yield a total severity-of-risk screen score. Children were selected for inclusion into the high-risk sample based on this screen score, moving from the highest score downward until desired sample sizes were reached within sites, cohorts, and groups. This multistage screening procedure resulted in 891 children divided into high-risk control (n = 446) and intervention (n = 445) samples. In addition to the high-risk sample of 891, a stratified normative sample of 387 children was identified to represent the population normative range of risk scores and was followed over time. This stratification procedure ensured that the normative sample would be representative of the population in high-risk schools. The normative sample included 79 high risk, control-group children to the proportional degree that they were represented in the school population. The present study used data from the high-risk control (kindergarten age: M = 5.57 years; SD = .44; range = 4.68–6.94; 65% male; 44% Black, 51% white, 5% other race/ethnicity) and normative (kindergarten age: M = 5.55 years; SD = .44; range = 4.66–7.79; 51% male; 42% Black, 51% white, 7% other race/ethnicity) samples; the intervention sample was not included in the present analyses. Thus, the total final sample included 754 participants (kindergarten age: M = 5.56 years; SD = .43; range = 4.66–7.79; 58% male; 46% Black, 50% white, 4% other race/ethnicity). See Figure S1 for the FT CONSORT flow chart.
Child Development Project
The CDP is an investigation of children's social development and adjustment (Dodge et al., 1990). In 1987 and 1988, a sample of children was identified at the time of kindergarten preregistration and then followed over time. Participants were recruited in each of two annual cohorts at each of three geographic sites (Bloomington, IN; Knoxville, TN; and Nashville, TN). Within each site, public school representatives were consulted to identify neighborhoods and schools that varied in demographic characteristics, and federally subsidized lunch rates and neighborhood housing patterns were also used to identify schools that served a full demographic range of the communities. Children from numerous neighborhoods and schools (i.e., 5 to 10 areas within each city) were included to increase variation yet closely represent the overall city means and variances in demographic characteristics. Families of 5-year-old children were recruited at the time of kindergarten preregistration (April preceding the September of matriculation). Parents registering their children were approached at random by research staff and asked to participate in a longitudinal study of child development. A brief description of the study was given, and parents indicated their interest by scheduling an interview. Approximately 75% of those parents who were approached agreed to participate. During the first week of school, families of children who had not preregistered were randomly solicited to participate through a combination of letters sent home with the child and telephone calls. Interested parents were then visited by research staff who explained the project in detail and obtained parents' informed consent. Overall, 585 children participated (kindergarten age: M = 5.50 years; SD = .33; range = 4.34–6.92; 52% male; 81% white, 17% Black, 2% other race/ethnicity).
Measures
Conduct problems
Conduct problems were assessed in kindergarten (FT: 1991–1993; CDP: 1987–1988) using a mean score of parent reports on the Child Behavior Checklist (Achenbach, 1991a) and teacher reports on the Teacher's Report Form (Achenbach, 1991b) externalizing broadband scales. Items are scored on a 3-point scale (0 ‘not true’, 1 ‘somewhat or sometimes true’, 2 ‘very or often true’). In the present study, raw scores were standardized and used in analyses. Internal consistency for the combined sample was excellent (parent reports; α = .90, ω = .91; teacher reports; α = .96, ω = .96).1
Criminal costs
Administrative court data were collected by study personnel. This collection included juvenile court records when permission was granted by parents. These data were supplemented with data collected from national criminal databases. Study participants were identified based on full name, date of birth, and social security number. Using a national database enabled the collection of court data outside the participant's home state. These data include descriptions of the crimes committed and the outcomes of the arrest. Our analyses examined crimes committed in each developmental period based on the age of the participant at the time of the crime: adolescence (ages ≤18) and adulthood (ages 19–24). Each crime was classified as a violent (e.g. assault, murder), substance (e.g. possession, possession with intent to sell), public order and property (e.g. disorderly conduct, trespassing), or status (e.g. truancy, running away) crime. Due to inconsistencies in reporting across sites, status, minor traffic offenses, probation violations, failures to appear, and crimes committed while incarcerated are excluded.
Cohen and Piquero (2009) reanalyzed costs, and we used this new evidence to estimate the cost of individual crimes. For other crimes not estimated by previous literature, such as drug distribution/possession, we used estimated figures by Federal Drug Control Budget (2019).
Government services costs
Government service use was self-reported by participants (at approximately age 25 for FT [FT: M = 24.49 years; SD = .64; range = 23.11–27.42; FT high risk: M = 23.34 years; SD = .63; range = 23.11–27.17; FT normative: M = 24.71 years; SD = .58; range = 23.77–27.42] and age 28 for CDP [M = 28.40 years; SD = .48; range = 27.09–30.40]). Services were summed across the following categories for which respondents were given categorical ranges to indicate how much they received: food stamps; housing assistance; Supplemental Security Income/Disabilities/Social Security; Unemployment assistance; Aid to Families with Dependent Children (AFDC)/Temporary Assistance for Needy Families (TANF)/Welfare; Special Supplemental Nutrition Program for Women, Infants, and Children (WIC); and Energy assistance.
Medical services costs
Medical costs were calculated by the self-reported questionnaire (at approximately age 25 for FT and age 28 for CDP) about Medicaid or other medical care assistance and average Medicaid spending data from National Health Expenditure (NHE, 2022). We assigned the total cost of Medicaid spending by age and gender from NHE fact sheet for those who answered they received Medicaid service in past 12 months.
Total costs
We also calculated a total costs score summing across criminal legal costs, offender productivity costs, victim costs, government services costs, and medical services costs.
Study harmonization
To harmonize data across FT and CDP, we determined the measures (and specific measurement items) that were given in both studies when the participants were the same age. Administrative court data were collected by overlapping personnel across studies, and the same search engines were used to collect court data.
Analytic approach
Analyses proceeded in four main steps. First, we calculated correlational and descriptive statistics of priced costs. Second, using Mplus 8 (Muthén & Muthén, 2017), we tested for significant associations between standardized kindergarten conduct problems and priced outcomes using negative binomial regression models given the outcomes were overdispersed. Due to large variances, all priced costs were transformed by a division of 1,000. Given multiple comparisons, we applied a Bonferroni corrected alpha of 0.003. Third, where significant associations were found, we report incidence rate ratios (IRR) and calculated the mean of estimated costs for those with standardized conduct problems score equals to zero and for those with standardized conduct problems score equals to one. Each costed outcome was calculated separately based on its intercept and betas for each predictor (i.e., CP and covariates). Amounts were factored by 1.5% to represent the deadweight loss associated with administrative costs necessary to process criminal activity (Kuklinski, Fagan, Hawkins, Briney, & Catalano, 2015). All monetized associations were multiplied by their transformed values and adjusted for inflation to USD 2020 prices using a consumer price index conversion. Full-information maximum likelihood techniques were used to attain model estimates, which provides full-sample estimates that accommodate all observations regardless of whether missing data occurs for certain variables (Rubin & Little, 2002). The level of missingness was 14.5% for FT and 16.4% for CDP. In these analyses, sampling weights were also used to adjust for the oversampling of higher-risk FT participants and enabling a more representative sample of the communities from which data were collected (we used the Mplus ‘WEIGHT’ command in analyses, and we describe how sampling weights were derived in further detail in Appendix S1). Finally, we conducted multigroup models with sex as the known class variable (0 = female; 1 = male) and corresponding t-tests were used to determine whether parameters related to costs significantly differed. We then estimated costs for the male and female samples as per the above. All models controlled for main effects of sex (in the combined sample models), race/ethnicity, single parenthood status, socioeconomic status, and study site.
Results
Descriptive criminal, victim, government, and medical costs
Percentage of participants with costs are shown in Table S2. Descriptive statistics and correlations of main study variables are presented in Tables S3–S5. (We also report descriptive statistics and correlations for FT, FT high risk and normative samples, and CDP in Tables S6–S9.)
Mean (and standard error [SE]) costs are presented in Figure 1 for the combined sample and Figure 2 for the male and female samples. These figures depict mean costs prior to inflation adjustments for adolescent and adult criminal costs, adolescent and adult criminal + victim costs, adult government and medical services costs, and total costs. As shown in Figure 1 (combined sample), violent criminal and criminal + victim costs represented the largest costs at $1,406 and $5,184, respectively, in adolescence, and $5,234 and $42,380, respectively, in adulthood. Adolescent total criminal and criminal + victim costs were $2,258 and $7,255, respectively. By adulthood, substance criminal and criminal + victim costs were $4,106 and $12,664, respectively. Adult total criminal and criminal + victim costs were $10,685 and $57,102, respectively. Government services were represented by $891 in costs, and medical services by $20,247 in costs. Total costs across all categories equated to $145,268.
As shown in Figure 2, among the male sample, violent criminal and criminal + victim costs were $1,997 and $7,702, respectively, in adolescence. Adolescent total criminal and criminal + victim costs were $3,253 and $10,770, respectively. In adulthood, violent criminal and criminal + victim costs were $7,673 and $52,125, respectively, and substance criminal and criminal + victim costs were $6,436 and $19,851, respectively. Adult total criminal and criminal + victim costs were $16,003 and $74,871, respectively. Government services were represented by $702 in costs, and medical services by $15,203 in costs. Total costs across all categories for the male sample equated to $182,904. Among the female sample, violent criminal and criminal + victim costs were $674 and $2,063, respectively, in adolescence. Adolescent total criminal and criminal + victim costs were $1,024 and $2,899, respectively. Adult violent criminal and criminal + victim costs were $2,213 and $30,305, respectively, and total criminal and criminal + victim costs were $4,095 and $35,084, respectively. Government services were represented by $1,106 in costs, and medical services by $25,619 in costs. Total costs across all categories for the female sample equated to $98,634.
Monetization of kindergarten conduct problems
Table 1 provides estimates from negative binomial models of kindergarten conduct problems predicting costs for the combined sample and for the male and female samples. Although there were significant differences between the male and female samples on adolescent substance criminal and criminal + victim costs, we have not monetized adolescent substance crime in the female sample because of the very low count (see Table S2).
| Variable | Combined sample | Male sample | Female sample | Male versus female | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| B (SE) | β | p | B (SE) | β | p | B (SE) | β | p | t (SE) | p | |
| Adolescent criminal costs | |||||||||||
| Violent | 0.83 (.17) | 0.72 | <.001 | 0.96 (.21) | 0.81 | <.001 | 1.01 (.26) | 0.85 | <.001 | 0.05 (.33) | .881 |
| Substance | 1.08 (.23) | 0.94 | <.001 | 0.53 (.23) | 0.45 | .019 | 2.17 (.45) | 1.83 | <.001 | 1.64 (.48) | .001 |
| Public order and Property | 0.34 (.13) | 0.30 | .007 | 0.22 (.16) | 0.19 | .170 | 0.57 (.19) | 0.49 | .002 | 0.35 (.25) | .159 |
| Total | 0.79 (.14) | 0.68 | <.001 | 0.70 (.15) | 0.60 | <.001 | 0.81 (.22) | 0.68 | <.001 | 0.10 (.27) | .702 |
| Adolescent criminal + Victim costs | |||||||||||
| Violent | 1.08 (.20) | 0.93 | <.001 | 1.11 (.27) | 0.94 | <.001 | 1.56 (.31) | 1.32 | <.001 | 0.46 (.41) | .259 |
| Substance | 1.51 (.29) | 1.31 | <.001 | 0.65 (.26) | 0.55 | .014 | 3.31 (.57) | 2.80 | <.001 | 2.67 (.61) | <.001 |
| Public order and Property | 0.37 (.13) | 0.32 | .006 | 0.20 (.18) | 0.17 | .268 | 0.62 (.23) | 0.53 | .008 | 0.43 (.29) | .145 |
| Total | 1.06 (.16) | 0.91 | <.001 | 0.91 (.19) | 0.77 | <.001 | 1.08 (.27) | 0.92 | <.001 | 0.18 (.33) | .589 |
| Adult criminal costs | |||||||||||
| Violent | 0.74 (.13) | 0.64 | <.001 | 0.52 (.11) | 0.44 | <.001 | 1.18 (.22) | 1.00 | <.001 | 0.66 (.25) | .007 |
| Substance | 0.15 (.10) | 0.13 | .151 | 0.23 (.11) | 0.19 | .045 | 0.15 (.19) | 0.13 | .414 | −0.08 (.22) | .722 |
| Public order and Property | 0.27 (.10) | 0.23 | .007 | 0.16 (.12) | 0.13 | .183 | 0.44 (.16) | 0.37 | .006 | 0.28 (.20) | .161 |
| Total | 0.40 (.10) | 0.35 | <.001 | 0.33 (.09) | 0.28 | <.001 | 0.52 (.16) | 0.44 | .001 | 0.18 (.19) | .321 |
| Adult criminal + Victim costs | |||||||||||
| Violent | 0.86 (.20) | 0.74 | <.001 | 0.56 (.11) | 0.48 | <.001 | 1.34 (.36) | 1.13 | <.001 | 0.78 (.38) | .040 |
| Substance | 0.14 (.11) | 0.12 | .184 | 0.23 (.11) | 0.20 | .043 | 0.16 (.20) | 0.13 | .428 | −0.08 (.23) | .733 |
| Public order and property | 0.27 (.10) | 0.23 | .008 | 0.16 (.12) | 0.13 | .194 | 0.41 (.17) | 0.35 | .017 | 0.25 (.21) | .230 |
| Total | 0.62 (.14) | 0.53 | <.001 | 0.42 (.09) | 0.36 | <.001 | 0.88 (.27) | 0.74 | .001 | 0.45 (.29) | .117 |
| Other services costs | |||||||||||
| Government services | 0.37 (.16) | 0.32 | .020 | 0.21 (.25) | 0.17 | .407 | 0.67 (.18) | 0.57 | <.001 | 0.47 (.31) | .127 |
| Health services | 0.31 (.10) | 0.27 | .001 | 0.55 (.20) | 0.46 | .005 | 0.40 (.12) | 0.34 | .001 | −0.15 (.23) | .531 |
| Total costs | 0.55 (.01) | 0.47 | <.001 | 0.43 (.08) | 0.36 | <.001 | 0.70 (.19) | 0.59 | <.001 | 0.28 (.21) | .190 |
- Controlling for sex (in the combined sample models), race/ethnicity, single parenthood status, socioeconomic status, and study site.
Table 2 depicts monetized outcomes for the combined sample, male sample, and female sample. For the combined sample, a one standard deviation (SD) increase in kindergarten conduct problems was associated with increases of $2,587, $1,925, and $3,788 in adolescent violent, substance, and total criminal costs, respectively, as well as increases of $18,191, $13,899, and $21,934 in adolescent violent, substance, and total criminal + victim costs, respectively. A 1SD increase in conduct problems was associated with increases of $7,408 and $7,374 in adult violent and total criminal costs, respectively, as well as increases of $72,059 and $63,998 in adult violent and total criminal + victim costs, respectively. Finally, a 1SD increase in conduct problems was associated with increases of $12,753 and $146,279 in medical services costs and total costs, respectively.
| Variable | Combined sample | Male sample | Female sample | ||||||
|---|---|---|---|---|---|---|---|---|---|
| IRR | Dollars per 1SD unit | 95% CI | IRR | Dollars per 1SD unit | 95% CI | IRR | Dollars per 1SD unit | 95% CI | |
| Adolescent criminal costs | |||||||||
| Violent | 2.29 | $2,586.60 | [$1,321.00, $4,340.43] | 2.55 | $4,214.04 | [$1,994.48, $7,486.24] | 3.17 | $3,692.91 | [$1,421.42, $7,605.70] |
| Substance | 2.95 | $1,925.23 | [$881.24, $3,556.14] | – | – | – | – | – | – |
| Public order and Property | – | – | – | – | – | – | 1.83 | $148.89 | [$45.01, $301.25] |
| Total | 2.21 | $3,788.13 | [$2,179.74, $5,881.30] | 2.00 | $5,052.00 | [$2,475.80, $8,495.60] | 2.46 | $2,534.66 | [$929.72, $5,105.67] |
| Adolescent criminal + Victim Costs | |||||||||
| Violent | 2.93 | $18,191.00 | [$9,292.01, $31,283.30] | 2.99 | $29,569.26 | [$11,852.60, $58,968.80] | 5.49 | $52,128.39 | [$22,465.60, $107,602.00] |
| Substance | 4.54 | $13,899.10 | [$6,172.16, $27,503.80] | – | – | – | – | – | – |
| Public order and Property | – | – | – | – | – | – | – | – | – |
| Total | 2.88 | $21,934.10 | [$12,801.40, $34,517.40] | 2.45 | $28,530.46 | [$13,902.50, $49,451.80] | 3.25 | $15,481.49 | [$5,682.20, $32,929.80] |
| Adult criminal costs | |||||||||
| Violent | 2.09 | $7,408.33 | [$4,160.37, $11,615.40] | 1.67 | $7,323.49 | [$3,836.61, $11,612.00] | 3.40 | $8,673.93 | [$4,219.23, $15,677.10] |
| Substance | – | – | – | – | – | – | – | – | – |
| Public order and Property | – | – | – | – | – | – | – | – | – |
| Total | 1.50 | $7,373.70 | [$3,767.36, $11,705.50] | 1.40 | $11,018.14 | [$4,610.51, $18,743.30] | 1.68 | $3,900.67 | [$1,269.06, $7,505.28] |
| Adult criminal + Victim costs | |||||||||
| Violent | 2.35 | $72,059.40 | [$31,598.40, $131,821.00] | 1.77 | $49,782.45 | [$27,431.10, $77,691.40] | 3.83 | $134,374.90 | [$41,629.60, $323,956.00] |
| Substance | – | – | – | – | – | – | – | – | – |
| Public order and property | – | – | – | – | – | – | – | – | – |
| Total | 1.85 | $63,998.30 | [$30,434.90, $108.228.00] | 1.53 | $58,871.89 | [$31,534.30, $91,268.40] | 2.40 | $62,916.26 | [$18,054.20, $139,853.00] |
| Other services costs | |||||||||
| Government services | – | – | – | – | – | – | 1.91 | $1,530.49 | [$592.11, $2,851.49] |
| Medical services | 1.37 | $12,752.90 | [$4,603.59, $22,541.80] | – | – | – | 1.50 | $24,105.20 | [$8,542.35, $44,005.00] |
| Total costs | 1.73 | $146,279.00 | [$84,568.20, $220,932.00] | 1.53 | $144,139.50 | [$82,816.10, $216,590.00] | 2.02 | $144,823.40 | [$54,162.70, $277,682.00] |
- Controlling for sex (in the combined sample models), race/ethnicity, single parenthood status, socioeconomic status, and study site.
For the male sample, a 1SD increase in kindergarten conduct problems was associated with increases of $4,214 and $5,052 in adolescent violent and total criminal costs, respectively, and increases of $29,569 and $28,530 in adolescent violent and total criminal + victim costs, respectively. A 1SD increase in male kindergarten conduct problems was associated with increases of $7,323 and $11,018 in adult violent and total criminal costs, respectively, as well as increases of $49,782 and $58,872 in adult violent and total criminal + victim costs, respectively. In addition, a 1SD increase in male kindergarten conduct problems was associated with an increase of $144,140 in total costs. Finally, for the female sample, a 1SD increase in kindergarten conduct problems was associated with increases of $3,693, $149, and $2,535 in adolescent violent, public order/property and total criminal costs, and $52,128 and $15,481 in adolescent violent and total criminal + victim costs, respectively. A 1SD increase in female kindergarten conduct problems was associated with increases of $8,674 and $3,901 in adult violent and total criminal costs, respectively, as well as increases of $134,375 and $62,916 in adult violent and total criminal + victim costs, respectively. In addition, a 1SD increase in female kindergarten conduct problems was associated with increases of $1,530, $24,105, and $144,823 in government services, medical services, and total costs, respectively.
Our analytic approach of examining crime classifications separately and established differences in IRR explain why specific classifications of crime resulted in higher priced outcomes relative to total crime in some models (e.g. adult violent criminal costs vs. adult total criminal costs).
Discussion
Multi-informant kindergarten conduct problems were associated with USD 2020 priced criminal offending and government and medical services use. Monetized associations differed according to whether the crime occurred in adolescence or early adulthood and the type of criminal offense. By examining cost outcomes in both developmental periods, we provide richer information that can be used by policy makers for targeted budget plans. In addition, we add to a limited literature using community samples aggregating crime costs across classifications, and we demonstrate that among mostly population-representative children elevated conduct problems can result in financial toll related to offending and medical care. The novel information gleaned from our study can also be used in benefit–cost analyses for prevention and intervention programs aiming to target specific forms of criminality.
Few studies have examined the costs associated with childhood behavior problems. In contrast to other United States based research, we elected to test associations based on a 1SD increase in conduct problems to advance understanding of monetized associations across the distribution (rather than for specific diagnostic criteria; Foster et al., 2005). Because of this, and other methodological strengths of our study (e.g. examining conduct problems as early as kindergarten [FT = 4.66–7.79 years; CDP = 4.34–6.92 years]), direct comparisons cannot optimally be made to past studies. However, our findings do align with some past research. In our study, adolescent total criminal costs were $3,788. Among a subgroup of the present study (i.e., FT) meeting diagnostic criteria for oppositional defiant disorder in 6th or 9th grade, juvenile legal costs equated to $3,540 across a 7-year period into late adolescence (Foster et al., 2005). Our findings also extend past research on the cost of conduct problems, which has typically examined criminal costs in either adolescence (Foster et al., 2005) or adulthood (d'Amico et al., 2014), but not both.
An advance of the present study was the examination of cumulative criminal and related costs. Our findings provide a dollar estimate for criminal legal costs, the loss of offender productivity, as well as costs to the victims of crimes committed by our participants. Specifically, a 1SD increase in kindergarten conduct problems was associated with a $21,934 increase in adolescent criminal + victim costs, a $63,998 increase in adult criminal + victim costs, and a $146,279 increase in total costs across all categories. Past monetization studies with nonoffending populations have omitted these other related costs (e.g. d'Amico et al., 2014); thus, our findings inform current understanding of expenses associated with criminality beyond the legal sector.
We also examined costs for male and female participants. Notably, in the male sample, a 1SD increase in kindergarten conduct problems was associated with a $49,782 increase in adult violent criminal + victim costs, a $58,872 increase in adult total criminal + victim costs, and a $144,140 increase in total costs. In the female sample, a 1SD increase in kindergarten conduct problems was associated with a $134,375 increase in adult violent criminal + victim costs, a $62,916 increase in adult total criminal + victim costs, a $1,530 increase in government services costs, a $24,105 increase in medical services costs, and a $144,823 increase in total costs. Prior research has mostly concerned predictors of criminality costs among male samples; the present study advances the field by studying prediction of female crime and related costs.
Methodological strengths of the present study include its multisite longitudinal design, large sample of community-representative participants, assessment of parent- and teacher-reported conduct problems in early childhood, and use of administrative records and national databases. Our analytic approach took advantage of information that had been excluded from many past studies, including costs by distinct categories of crime, other costs associated with criminality (i.e., offender productivity and victim costs), and costs distinguished by sex. Despite these strengths, findings must be considered within the context of some limitations. Costs and monetized associations are not authoritative, and rather represent estimates. Monetized associations are also not causal; however, given the parallel nature of early conduct problems and adverse outcomes, our findings do reflect important associations that may be expected. In addition to monetizing the association between conduct problems and several criminal outcomes, we also monetized government and medical service use. However, we were not able to differentiate physical from mental health service use, thus representing an important avenue for future research. A further limitation is that our sample is not representative of many diverse populations. Our combined multisite sample was comprised of higher-risk (FT) and community-representative (CDP) samples.2 In our analytic approach, we applied sampling weights to adjust for the higher-risk oversampling; however, our findings may not generalize to other communities. It is important for future research to conduct a broader sampling of participants. In addition, although our longitudinal data extend nearly 20 years into adulthood (age 24/28), further research is needed throughout adulthood. For example, established adulthood (ages 30–45) has recently been classified as a distinct developmental period characterized by some of the most demanding years due to a “career-and-care crunch” (Mehta, Arnett, Palmer, & Nelson, 2020, p. 436); thus, future research should assess associated costs of criminal offending in other adult life stages. Finally, various integrated, cascade, and systems theories of crime all emphasize the multifaceted complexities of contributors to conduct problems and crime initiation and desistence (Dodge et al., 2008; Farrington, 2017). Our study provides information on one critical piece of the expenditures puzzle, and it would be important for future research to monetize other theoretically relevant predictors.
Findings from this study also have implications for further economic evaluation research, practice, and policy. The literature is converging on the point that conduct problems are associated with significant economic costs. Although one recent cost-of-illness systematic review using data from 48 countries and categorized into 9 mental illness related groups found that developmental disorders (e.g. autism spectrum disorders) were associated with the highest median societal cost per patient, other categories, such as childhood behavioral disorders, are more prevalent, and thus, may ultimately contribute greater public costs (Christensen et al., 2020). Adding to this literature, the present study represents the first effort to monetize the prediction of kindergarten conduct problems for criminal and victim costs into adulthood, as well as government and medical service use costs.
Childhood conduct problems are also associated with disproportionate burden relative to conduct problems that manifest later, and even individuals classified as ‘childhood-limited’ often do not show complete recovery regarding a host of negative outcomes, including greater service utilization and criminal behavior (Rivenbark et al., 2018). However, multicomponent interventions applied early in childhood have been shown to reduce psychopathology, substance use problems, and criminality through adulthood (e.g. Dodge et al., 2015). By harnessing information from emerging cost evaluations of early intervention programs, we can further contextualize the fiscal benefit of this preventive support in early childhood. For example, Hunter, DiPerna, Hart, and Crowley (2018) estimated the efficacy and cost effectiveness of a universal social–emotional learning program (Social Skills Improvement System-Classwide Intervention Program; Elliott & Gresham, 2007) showing that intervention participants improved in teacher-rated social skills relative to the control group (g = .36) and the average cost to implement the program was approximately only $18.99 per student. In another study testing the cost of Family Check-Up (a comprehensive intervention focused on strengthening the parent–child relationship and reducing conduct problems; Dishion et al., 2008), Kuklinski et al. (2020) found an average cost of $1,066 per family in the first year and $501 per family once training was established. Future research should project cost-benefits of such early interventions through adulthood. However, in the context of the costs captured in our study, these findings suggest that robust efforts to invest in early prevention hold promise for mitigating both long-term economic and societal toll. In sum, by placing a dollar value on early conduct problems, we have provided information that can be used by legislators for evidence-based budgeting. Policymakers should consider placing greater investment in research on childhood conduct problems, as well as clinical prevention initiatives to ensure a strong start for children.
Acknowledgments
This work used data from the Fast Track project (for additional information concerning Fast Track, see http://www.fasttrackproject.org). The authors are grateful to the members of the Conduct Problems Prevention Research Group (in alphabetical order, Karen L. Bierman, Pennsylvania State University; John D. Coie, Duke University; D. Max Crowley, Pennsylvania State University; Kenneth A. Dodge, Duke University; Mark T. Greenberg, Pennsylvania State University; John E. Lochman, University of Alabama; Robert J. McMahon, Simon Fraser University and B.C. Children's Hospital Research Institute, and Ellen E. Pinderhughes, Tufts University) for providing the data and for additional involvement.
The Fast Track project has been supported by National Institute of Mental Health (NIMH) Grants R18 MH48043, R18 MH50951, R18 MH50952, R18 MH50953, R01 MH062988, R01 MH117559, K05 MH00797, and K05 MH01027; National Institute on Drug Abuse (NIDA) Grants R01 DA016903, R01 DA036523, R01 DA11301, K05 DA15226, RC1 DA028248, and P30 DA023026; National Institute of Child Health and Human Development Grant R01 HD093651; and Department of Education Grant S184U30002. The Center for Substance Abuse Prevention also provided support through a memorandum of agreement with the NIMH. Additional support for this study was provided by a B. C. Children's Hospital Research Institute Investigator Grant Award and a Canada Foundation for Innovation award (to Robert J. McMahon). The Child Development Project has been funded by grants MH56961, MH57024, and MH57095 from the National Institute of Mental Health, HD30572 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, and DA016903 from the National Institute on Drug Abuse.
The authors are grateful for the collaboration of the Durham Public Schools, the Metropolitan Nashville Public Schools, the Knox County Schools, the Monroe County Community School Corporation, the Bellefonte Area Schools, the Tyrone Area Schools, the Mifflin County Schools, the Highline Public Schools, and the Seattle Public Schools. The authors appreciate the hard work and dedication of the many staff members who implemented the project, collected the evaluation data, and assisted with data management and analyses.
Data and code for this study are available by emailing the corresponding author. This study was not preregistered. The authors have declared that they have no competing or potential conflicts of interest.
Key points
- Childhood conduct problems are highly prevalent and confer risk for participation in criminal behavior, as well as elevated mental and physical health problems.
- We monetized the association between parent- and teacher-reported kindergarten conduct problems and criminal and victim costs (including violent, substance, and public order/property) across adolescence and adulthood, in addition to government and medical services in adulthood.
- A 1SD increase in kindergarten conduct problems was associated with a $21,934 increase in adolescent criminal + victim costs, a $63,998 increase in adult criminal + victim costs, a $12,753 increase in medical services costs, and a $146,279 increase in total costs. Our findings have important implications by providing policymakers information that can be used to support investment in early childhood research and clinical initiatives.
