The sins of the child: Public opinion about parental responsibility for juvenile crime

Previous research indicates that people ascribe less responsibility to juvenile offenders than adult offenders for the crimes they have committed. Although assignment of responsibility to parents varies depending upon the youth’s age, we know little about the role of other factors. The current study used randomized factorial vignettes to examine whether the seriousness of the offense, peer involvement, and the offender's race, age, sex, and prior record influence support for parental responsibility. We found that participants in our study placed substantial responsibility on parents for dealing with juvenile offenders; however, support for punishing the parents was low, while the importance attached to enrolling parents in training was high. Views on parental responsibility were influenced by the age of the juvenile and the seriousness of the offense. The implications of these findings and how they fit in with previous work are discussed.


Introduction
In the late 19th century, a separate justice system for juveniles emerged in the United States. Following the parents patriae ("the state as parent") doctrine, the state assumed responsibility for re-educating children, who were no longer tried and punished as adult offenders. The underlying notion was that parents of court-involved children were not fulfilling their responsibilities, requiring judicial authorities to act in their stead (Tanenhaus, 2011). Another core assumption guiding the foundation of the juvenile court was that children were not entirely responsible for their criminal behavior and could be rehabilitated (McShane & Williams, 2003). In the second half of the 20th century, however, juvenile justice experienced an upheaval, first with the introduction of procedural protections that shifted the court toward a more adversarial process and later with an explicitly punitive turn (Feld, 2017). To different degrees, states across the U.S. lowered the age of criminal responsibility, created mandatory minimum sentences, and enacted policies aimed at treating youths as adults (Muncie, 2008). This renewed focus on punitive values and increased responsibility overshadowed the principles of protection and treatment that had guided the creation of the juvenile justice system (Junger-Tas, 2008). Parallel to this trend, increased emphasis on parental responsibility resulted in the promulgation of laws holding parents responsible for their children's criminal behavior in the U.S. and elsewhere (Arthur, 2009;Brank, Kucera, & Hays, 2005). In contrast to the widespread literature examining public views on other juvenile justice policies (e.g., Applegate & Davis, 2006;Mears, Pickett, & Mancini, 2015;Moon, Sundt, Cullen, & Wright, 2000;Pickett & Chiricos, 2012), few studies to date have analyzed perceptions of parental responsibility in cases of juvenile offenses. The current study contributes to filling this gap by examining attributions of parental responsibility. Using a randomized experimental design, we analyzed how multiple characteristics of young offenders, and the offenses committed by them, affect perceptions of parental responsibility and preferences with regards to interventions aimed at their parents. behavior of their children (Harris, 2006). In the context of the United States, Brank et al. (2005) identify three different types: civil liability laws, criminal liability laws, and parental involvement statutes. Civil liability laws allow victims to sue the parents of juveniles whose behavior results in property damage and/or personal injury. The second type, criminal liability, allows criminal charges to be brought against parents who contribute to the delinquency of their children. Under parental involvement statutes, parents are held accountable for the juveniles' behavior through different strategies such as paying fines or attending mandatory parenting classes. Therefore, parents can be held responsible for acts of commission (i.e., when they contribute to the offense) as well as acts of omission (i.e., when they fail to prevent the offense from happening).
Despite the profusion of these laws across Western countries (Arthur, 2009), little is known about public support for this form of vicarious responsibility. One of the early studies conducted in the United States showed considerable support for parental responsibility (Brank & Weisz, 2004). The study, conducted in 1999 using a national telephone survey, found that, after the juveniles, parents were identified as the most responsible party for the crime, preceding peers, schools, and the media. Despite placing significant responsibility on parents, respondents were less supportive of blaming and punishing them. A later study combined global questions (including the ones presented in the 1999 study) with specific cases presented in the form of vignettes, to compare global and specific attitudes toward parental responsibility (Brank, Hays, & Weisz, 2006). Consistent with previous research in the area of punitive attitudes (Applegate, Cullen, & Fisher, 2002;Applegate, Cullen, Turner, & Sundt, 1996;Campregher & Jeglic, 2016), it was found that support for parental responsibility was higher in response to global questions but diminished greatly when respondents were presented with specific cases (Brank et al., 2006). This was true for the degree of responsibility and blame attributed to the parents, as well as the punishment that was considered appropriate for them, with less endorsement of all forms of punishment in response to the vignettes.
Studies conducted specifically with parents and juvenile offenders suggest modest endorsements of parental accountability. In the study conducted by White, Augoustinos, and Taplin (2007) in Australia, parents considered that responsibility for the crime lay primarily within the children, and not the parents, even when the child was in his or her pre-teenage years (10 years of age). This finding has been echoed in the United States, with parents holding juvenile offenders mainly responsible, regardless of their age (Brank, Greene, & Hochevar, 2011). In addition, Brank and Lane's (2008) study with juveniles in post-adjudication residential facilities revealed that juveniles generally did not consider their parents to be responsible for their delinquent behavior.
Beyond legal accountability, there is evidence that parents are considered responsible for their children's transgressions. Studies in multiple countries have found that lack of parental discipline is considered an important contributor to juvenile crime (Crime Survey for England and Wales, 2016; Gabbidon & Boisvert, 2012) and improving parenting an effective strategy to reduce juvenile delinquency (Haines, 2007;Hough & Roberts, 2004). In contrast, the public has expressed skepticism about the ability of the juvenile justice system to fulfill this task. In a study conducted in the UK, only 4% of respondents were very confident that the juvenile justice system helps parents become more responsible for their children (Hough & Roberts, 2004).

Correlates of public opinion on parental responsibility
Views on parental responsibility are influenced by case and respondent characteristics. We begin by describing previous research examining case characteristics followed by findings regarding respondent attributes.
The case characteristic that has received the most attention in this field of research has been the age of the juvenile. A study conducted with parents in Australia using factorial vignettes found that respondents placed greater responsibility on older children when compared to their younger counterparts (13-year-olds versus 10-year-olds) (White et al., 2007). In a more recent series of studies, Brank et al. (2011) examined the role of various case characteristics on public opinion regarding parental responsibility. Across the studies, they found that attributions of responsibility to the parents varied as a function of the child's age, with parents of younger children (i.e., 9-and 10-year-olds) deemed as more responsible than parents of older children (13-, 16-, and 17-year-olds). Replicating the results of White et al.'s (2007) study in Australia, Brank et al. (2011) found that U.S. respondents placed more responsibility on older youths. These findings are also consistent with previous research indicating that age affects attributions of blame and guilt, perceptions of legal competence, and sentence preferences for young offenders (Ghetti & Redlich, 2001;McPhetres & Hughes, 2016;Scott, Reppucci, Antonishak, & Degennaro, 2006). Using factorial survey designs, previous studies have also found that support for transferring juveniles to adult court and for more severe sentences in general is higher when juveniles are described as being older (Applegate & Davis, 2006;Applegate, Davis, & Cullen, 2009).
The evidence regarding the influence of variables other than age is less consistent. In two of the three studies conducted by Brank et al. (2011), personal injury offenses resulted in higher ratings of parental responsibility when compared to property offenses. In contrast, White et al. (2007) found no differences in responsibility ratings based on whether the vignette described a property or a personal offense. They found, however, that the severity of the offense influenced attributions of responsibility, with high severity offenses resulting in increased responsibility of both juveniles and parents.
Inconsistent findings have also been observed for parental actions, with one of the studies suggesting that parents described as having committed, rather than omitted, actions (purchasing the fireworks or shotguns used in the offense vs. failing to secure them) were considered more responsible, while another study indicated no differences (Brank et al., 2011). In another study (Brank et al., 2006), only the juvenile's premeditation in the case was manipulated, and it showed no effects on attitudes toward holding the parents responsible. In addition, one of the three studies conducted by Brank et al. (2011) manipulated multiple characteristics of the juveniles, finding that ratings of parental responsibility were not influenced by prior record, premeditation, seriousness of the crime, race, or gender of the offender. Despite the number of factors manipulated in this research, only a small number of options were examined within each factor, increasing the need to further investigate the effects of case characteristics on public perceptions about parental responsibility. For example, Brank et al. (2011) reported differences in perceived responsibility for parents of younger (9-yearolds) and older children (13-and 17-year-olds), but they found no differences in responsibility ratings between the parents of the 13-and 17-year-olds, suggesting a non-linear effect.
The few studies that have analyzed how the characteristics of respondents affect attributions of responsibility to parents have shown that demographic variables do not account for much of the variation in public opinion. In the study conducted by Brank and Weisz (2004) using a national telephone survey and asking global questions (i.e., "parents are to blame when their child breaks the law"), respondents' gender, race, and age were unrelated to opinions on blaming or punishing the parents. The only demographic variable that reached statistical significance in their models was education, with higher levels of education being associated with greater support for blaming and punishing parents. Similarly, a study conducted in the UK found limited evidence that education, ethnicity, and age affected beliefs in parental deficiency as a causal factor of juvenile crime (Collins, Cox, & Leonard, 2015). Later studies have shown inconsistent findings regarding gender. In their study with college students using factorial surveys, Brank et al. (2011) found that respondents' gender did not affect attributions of responsibility to the parents or the juveniles. When the researchers replicated the study with parents, however, they found that mothers rated the parents of the hypothetical juvenile offender as more responsible than did fathers and placed significantly less responsibility on the juveniles. Despite many jurisdictions having parental responsibility laws that hold parents accountable for the delinquent acts of their children, the literature examining public views on this topic is scarce. To date, only a handful of studies have been conducted, showing considerable support for parental responsibility, but less endorsement of blame and punishment. The age of the juveniles influences attributions of responsibility to the parents, but the role of other case and offender characteristics remains unclear.

The current study
The current study contributes to previous research by examining how multiple characteristics of juvenile offenders and the offenses committed by them affect perceptions of parental responsibility and preferences for intervening with a young offender's parents. Specifically, it adds depth in this area by examining a wider range of levels within each manipulated factor -juveniles' age and race, prior record, and offense type -and by adding a case characteristic not considered in previous research examining parental responsibility: peer involvement, an important correlate of youth engagement in delinquency (McGloin & Thomas, 2019). This study also expands on prior research by examining three expressions of parental responsibility, including the responsibility to deal with young offenders, as well as support for parental training and punishment. Using an experimental design, this research seeks to answer the following research questions: Research question 1: What is the level of support for parental responsibility, parental training, and punishment?
Research question 2: Do the background characteristics of age, sex, and race of juvenile offenders affect perceptions of parental responsibility and preferences for programs aimed at the parents?
Research question 3: Do the type of offense, its perceived seriousness, and whether it was committed in a group affect perceptions of and preferences for parental responsibility programs?
In addition, the current study contributes to the current knowledge about the impact of personal characteristics on views regarding parental responsibility. To do so, we incorporate a number of respondents' variables not previously examined in studies on parental responsibility, but importantly, which have shown associations with punishment preferences, including fear of victimization and victimization risk (Armborst, 2017), and exposure to prison (Rose & Clear, 2004).

Participants and procedure
The sample consisted of 662 college students attending two universities in the Southeastern United States. 1 The average age of respondents was 20.32 years (SD = 1.53). Most participants identified as women (60.2%) and White (77.5%). Participants included freshman (50.5%), sophomores (26.7%), juniors (15.3%), seniors (7.1%), and others (0.5%). Nearly one in eight (13.3%) were criminology and criminal justice majors. Data were collected during the fall of 2015, using self-administered paper surveys. Questionnaires were distributed in general education classes where students completed the survey independently either at the beginning or end of the class period.

Design and measures
As we stated, one limitation of prior research in this area is examination of a limited set of juvenile offender characteristics that might impact people's views on parental responsibility. To address this shortcoming, we employ the factorial vignette survey technique pioneered by sociologist Peter Rossi (Rossi & Nock, 1982). In this approach, respondents are presented with a brief scenario in which the levels of multiple dimensions of details are randomly varied (Auspurg & Hinz, 2015). The result is a set of vignettes that constitute a random sample of all possible permutations of details, and the dimensions are uncorrelated except by chance (Auspurg & Hinz, 2015). By asking for respondents' reactions to the vignettes, it is possible to model the extent to which each dimension affects opinions. This technique has been applied to many social issues (Wallander, 2009), and prior factorial vignette studies have demonstrated that opinions on sentencing and punishment are influenced by details such as crime type, prior criminal record, and offender characteristics (e.g., Applegate & Davis, 2006;Applegate et al., 2009;Jacoby & Cullen, 1998;Miller, Rossi, & Simpson, 1986;Vuk, Applegate, Ouellette, Bolin, & Aizpurua, 2019).
Using a factorial survey design, participants were randomly assigned to one hypothetical vignette depicting a juvenile offense. In each vignette, the case characteristics of offense type and peer involvement, and the offender characteristics of sex, age, race, and prior record, were assigned randomly, and constant text (identical for all cases) was added. The following is an example of one of these vignettes (manipulated factors are presented in italics): D.R., a 17-year-old Hispanic female youth is charged with harassing someone by uploading embarrassing pictures, threatening them, and spreading rumors to humiliate the person. She committed this crime with some friends. D.R. has been in trouble with the law in the past for serious crimes.

Independent variables
The key independent variables are derived from the six dimensions used in the construction of the vignettes. These include offender age (8, 9, 10, 11, 12, 13, 14, 15, 16, and 17), race (White, Black, Hispanic, and Asian), sex (male, female), offending history (no prior record, minor crimes, and serious crimes), peer involvement (whether the crime was committed alone or with other people), and offense type, which ranged from violent felonies to petty offenses and status offenses. The exact wording for each offense and shortened labels that we use below are provided in Appendix A.
In addition to the vignette dimensions, we also included perceived offense seriousness as a possible predictor. Although we did not directly manipulate crime seriousness as part of the randomized factorial design, it has long been established that perceived seriousness varies by crime type (Stylianou, 2003). As a result, seriousness may be considered an extension of the instant offense, which was randomized. Immediately following the vignette, we asked respondents: "In your opinion, how serious is D.R.'s crime?" The respondents answered this question using a 7-point Likert-type scale, ranging from not at all serious (1) to very serious (7).

Dependent variables
Following the description of the case, participants were asked to indicate how much D.R.'s parents should be held "responsible for dealing with D.R." (general responsibility). Response options ranged from 1 (not at all responsible) to 7 (very responsible). To compare the degree of responsibility attributed to the parents with that assigned to other institutions, participants were also asked to rate the responsibility attributed to schools, child protection services, community organizations, the juvenile justice system, and the adult justice system, using the same 7-point response scale. In addition, participants were asked to rate the importance of two different responses: a) punishing the parents and b) providing them with training. These variables were measured using a 7-point Likert-type scale, with higher scores indicating greater importance (1 = not at all important, 7 = very important).

Control variables
In addition to the independent variables, in the multivariate models several personal characteristics were included as control variables: respondents' sex (male, female), age (measured in years), race (White, non-White), class standing, and major (criminology or criminal justice, other). Since victimization and fear of crime affect criminal justice attitudes (Singer et al., 2019), we also controlled for fear of victimization and perceived victimization risk. Fear of victimization was measured using a 7-point scale representing responses to a question about how much the respondent felt afraid that he or she might become a victim of a crime in their everyday life, where 1 indicated "not at all afraid" and 7 was anchored by "very afraid" (M = 3.20, SD = 1.52). Perceived likelihood of victimization is an index composed of the mean of six items assessing perceived likelihood of being a victim of six different crimes in the next year. Response options for these items ranged from not at all likely (1) to very likely (7). The values of the scale ranged from 1.00 to 6.33 (M = 2.74, SD = 1.10), and the Cronbach's α coefficient was 0.81. The last two control variables assessed vicarious exposure to prison. The first question asked respondents whether anyone in their immediate family ever served time in prison or jail (15.0% yes), while the second asked respondents to estimate what percentage of people from the neighborhood where they attended high school had served time in prison (M = 14.3%, SD = 14.65).

Analytic strategy
The analysis proceeded in three steps. First, descriptive statistics, bivariate correlations, and means for the dependent variables were calculated. These tests provided information about the overall perception of parental responsibility, as well as the perceived importance of punishing and providing training to the parents. Paired t-tests were performed to compare the level of responsibility attributed to the parents when compared with the other five institutions. To estimate the magnitude of these differences, effect sizes were calculated using Cohen's d. Bivariate correlations between the two responses aimed at the parents (punishment and training) shed light on whether the importance assigned to one was related to the importance attached to the other and the extent to which these responses were associated with the responsibility attributed to the parents.
Next, we computed analysis of variance (ANOVA) to examine associations between the manipulated factors and the dependent variables. To control for multiple testing error, a Bonferroni correction was applied. Finally, Ordinary Least Squares (OLS) regression models were estimated to further examine the effects of the explanatory and control variables on the three measures of parental responsibility. Collinearity diagnostics revealed no multicollinearity issues (variance inflation factors for all variables were below 3) and no severe outliers were found in any of the models. However, there was evidence of heteroskedasticity among the residuals. For this reason, the models were estimated using robust standard errors. Post hoc power analyses were conducted to evaluate the statistical power of the regression models. These analyses indicated that, in all three models, power exceeded 0.99 to detect medium effects. Since the dependent variables were Likerttype measures (unipolar, with seven points), these analyses were supplemented by non-parametric tests (Wilcoxon matched-pairs rank test, Spearman correlations, Kruskal-Wallis test, and ordinal regression models). The consistency between the results produced by linear and non-linear approaches and the easier interpretation of the linear analyses led us to report the former in the text and provide the non-parametric analyses in Appendix B.
The loss of information from missing data was small; variables used in the analysis had a relatively small number of missing data (range: 0% to 3.6%, with mean of 0.99% and SD = 1.30). The full sample for the regression analyses resulted in a total loss of 37 observations (5.6%) for the parental responsibility model and 42 observations (6.3%) for the two other models (interventions aimed at the parents).

Perceptions of and preferences for parental responsibility
A majority of respondents believed that parents were responsible for dealing with their child (M = 5.53, SD = 1.64). In fact, parents were attributed greater responsibility than any other institution, including the juvenile justice system (M = 4.53, SD = 1.95), child protection services (M = 3.52, SD = 1.91), schools (M = 3.34, SD = 1.66), community organizations (M = 2.94, SD = 1.58), and the adult justice system (M = 2.53, SD = 1.73). As indicated in Table 1, paired t-tests showed significant differences for each pair of variables (p < 0.001), indicating that participants assigned the parents of juvenile offenders more responsibility than any other institution. Effect sizes were large in all cases, except for the differences between the responsibility attributed to the parents and the juvenile justice system, where the effect size was medium (Cohen's d = 0.47).
When analyzing the reactions to the responses aimed at parents, we found that respondents perceived that providing training (M = 4.92, SD = 1.84) was markedly more important than punishing parents for their child's actions (M = 2.78, SD = 1.70). This difference was statistically significant (t = 26.99, df = 636, p < 0.001), and the effect size was large (Cohen's d = 1.07). Fig. 1 reports the percent of respondents who selected each answer option on the 7-point continua of responsibility and importance, showing the same pattern of favoring training over punishment.
Parental responsibility was positively related to the two responses aimed at the parents: punishing them (r = 0.26, p < 0.001), and providing them with training (r = 0.39, p < 0.001). Moreover, these two policy responses were positively correlated with each other, r = 0.36 (p < 0.001), suggesting that respondents simultaneously support punitive and non-punitive options, but they are distinct domains of parental responsibility.

Parental responsibility and vignette dimensions
As displayed in Table 2, there were no significant differences by sex, race/ethnicity, or peer involvement for any of the outcome variables. The age of the offender was negatively related to the importance assigned to punishing the parents (r = −0.17, p < 0.001), and to providing them with training (r = −0.19, p < 0.001). Paradoxically, the negative association between juveniles' age and parental responsibility was very small and did not reach statistical significance (r = −0.07, p = 0.098). Criminal history resulted in significantly different attributions of responsibility toward the parents, F(2, Not at all 2 3 4 5 6 Very

Table 2
Relationship between vignette dimensions and perceived parental responsibility, importance of parental punishment, and importance of parental training.

Parental responsibility Punishment Training
Dimension Note: M = mean; SD = standard deviation; N = none; MC = minor crimes; DWL = driving without a license; SC = stealing clothes; RP = robbery (pushing); RA = running away. *p < 0.05; **p < 0.01; ***p < 0.001. E. Aizpurua, et al. Children and Youth Services Review 114 (2020) 105023 the offense committed was driving without a license (M = 3.47, SD = 2.01), and least important when it consisted of stealing clothes (M = 2.42, SD = 1.63). Bonferroni post hoc comparisons revealed that this difference was significant (p < 0.05). There was a significant effect of offense type on the importance assigned to training as well, F(9, 628) = 2.89, p < 0.01, n 2 = 0.05. Specifically, participants viewed training as more important when the offense was running away from home or robbing someone without using a weapon compared to stealing clothes (p < 0.05).

Multivariate analyses
The influence of the aforementioned factors on each of the outcome variables was analyzed using OLS regression models, while controlling for the seriousness of the offense and respondents' personal characteristics. The unstandardized regression coefficients, their robust standard errors, and 95% confidence intervals can be found in Table 3. In Model 1, predictors of parental responsibility were investigated. This model was found to be statistically significant (F[30, 594] = 1.88, p < 0.01), although the amount of explained variance was minimal (adjusted R 2 = 0.04). Criminal history remained significant in the multivariate analysis, with participants seeing the parents of juveniles with previous minor offenses as less responsible than parents of juveniles with no prior records (b = −0.36, 95% CI = −0.68, −0.04). The offense type was the only other vignette dimension that influenced respondents' views of parental responsibility. Compared to shoplifting, running away from home increased the responsibility attributed to the parents (b = 0.67, 95% CI = 0.11, 1.24). Notably, offenders' sex, age, race/ ethnicity, and peer involvement did not alter perceptions of parental responsibility. Two other predictors of attributions of parental responsibility emerged as statistically significant. As shown in Table 3, parents were seen as more responsible as the perceived seriousness of their child's offense increased (b = 0.18, 95% CI = 0.07, 0.29). In addition, sophomores, juniors, and seniors attributed more responsibility to the parents than did freshmen (b = 0.60, 95% CI = 0.20, 1.00; b = 0.83, 95% CI = 0.32, 1.33; b = 0.89, 95% CI = 0.15, 1.63, respectively). A second OLS model was estimated to predict the perceived importance of punishing parents for the crimes of their children. This model was significant (F[30, 589] = 3.88, p < 0.01), explaining a larger amount of variance than the previous one (adjusted R 2 = 0.117). Consistent with the bivariate analysis, offenders' age was a significant predictor of the importance attached to punishment. The parents were seen as deserving more punishment when the child was described as younger (b = −0.10, 95% CI = −0.14, −0.05). Moreover, participants saw punishment as more important when the offense was driving without a license or running away, as opposed to shoplifting (b = 0.87, 95% CI = 0.26, 1.48 and b = 0.62, 95% CI = 0.07, 1.17, respectively). The other vignette dimensions (offenders' sex, race/ethnicity, criminal history, and peer involvement), however, failed to achieve a significant association with the outcome. In contrast, a positive relationship was found between the perceived seriousness of the crime and the importance attached to punishing the parents (b = 0.34, 95% CI = 0.23, 0.45). Remarkably, ratings of the importance of punishment did not vary by any of the respondents' personal characteristics.
The final model predicting the emphasis on providing parent training was also statistically significant (F[30, 589] = 3.89, p < 0.001) and explained roughly 12% of the variance in the outcome variable (adjusted R 2 = 0.123). The child's age was negatively associated with the importance attributed to parental training (b = −0.11, 95% CI = −0.16, −0.07). Training ratings were higher for juveniles with serious prior crimes than those with no criminal history (b = 0.35, 95% CI = 0.00, 0.69). There was no significant difference, however, between children with no criminal history and those who had been in trouble with the law for minor crimes. The last vignette dimension to attain statistical significance in this model was offense type. Respondents who read the condition describing the robbery involving pushing were more supportive of parental training than those who read the condition in which the youth committed shoplifting (b = 0.96, 95% CI = 0.30, 1.61). In addition, respondents indicated that parental training was more important when the child ran away from home, when compared with shoplifting (b = 1.23, 95% CI = 0.62, 1.83). Perceived seriousness was also significant, indicating that individuals who considered the offense to be more serious rated training as more important (b = 0.23, 95% CI = 0.12, 0.34). College level and gender were the only control variables that were significantly associated with training. Findings indicate that sophomores, juniors, and seniors tended to rate training as more important than freshmen did (b = 0.50, 95% CI = 0.03, 0.97; b = 0.65, 95% CI = 0.01, 1.29; b = 0.93, 95% CI = 0.11, 1.76). In contrast, male students expressed less support for parental training than their female counterparts (b = −0.35, 95% CI = −0.64, −0.06).

Discussion
Previous research indicates that people believe parents should be held responsible for the wrongdoing of their children and that the age of juvenile offenders affects the degree of responsibility attributed to the parents (Brank et al., 2011;White et al., 2007). The current study adds to the limited research in this area by examining perceptions and preferences regarding parental responsibility using hypothetical scenarios in which multiple characteristics of the juvenile and the offense were randomized. In this way, we were able to assess the independent effect of each characteristic on respondents' views of holding parents responsible for the delinquency of their child.
The answer to Research Question 1 reveals that participants held parents largely accountable for dealing with their children after they had committed a crime or status offense. Lower on the list was the juvenile justice system, followed by child protection services, schools, community organizations, and the adult criminal justice system. That is, across the cases, respondents assigned more responsibility to parents for managing their child than the juvenile justice system. This finding agrees with and builds on previous research indicating that, in addition to juveniles themselves, parents are considered the party most responsible for youths' criminal activity (Brank et al., 2006(Brank et al., , 2011Brank & Weisz, 2004). Despite high attributions of parental responsibility, the proposed interventions involving parents received varying degrees of support, with training parents considered significantly more important than punishing them. This result concurs with previous research indicating that blaming parents, and support for their punishment, is rather low (Brank et al., 2006;Brank & Weisz, 2004). Although individuals place a certain degree of responsibility on parents, this appears to translate into a preference for non-punitive responses.
Our respondents' preference for parental training over punishment parallels a consistent finding that the public generally embraces a balanced approach to youthful offenders and often prioritizes juvenile rehabilitation over punitive sanctions for juvenile offenders (Applegate, 2020;Mears et al., 2015). Moreover, it suggests the existence of policy space for addressing juvenile crime by building parenting competencies. We find this encouraging in light of compelling empirical evidence that parent training programs can effectively prevent delinquency as well as reduce recidivism among previously adjudicated youths (Farrington & Welsh, 2007).
Participants rated both types of actions aimed at parents (punishment and training) less important as children age. Interestingly, attributions of responsibility to parents for dealing with their children did not vary depending on the children's age (Research Question 2). This finding contrasts with the previous literature, which has consistently reported a negative relationship between the age of juveniles and perceptions of parental responsibility for their children's conduct (Brank et al., 2011;White et al., 2007). In our research, the assignment of responsibility to the parents for dealing with their children did not vary across the wide range of ages manipulated in the vignettes (8-17 years E. Aizpurua, et al. Children and Youth Services Review 114 (2020) 105023 old). Respondents, regardless of the juvenile's age, believed that parents were responsible for dealing with their child's behavior. This paradoxical finding resembles previous research pointing to the complex relationship between offenders' ages and public preferences. In an early study, Ghetti and Redlich (2001) found that the ages of young offenders affected perceptions of blameworthiness and legal competence but did not impact sentence recommendations. Similarly, McPhetres and Hughes' findings (2016) reveal that, while more blame and guilt were attributed to older juveniles, endorsements of sentence lengths were unaffected by age. Our results support this complexity by showing that the youth's age did not affect views on parental responsibility but did have effects on preferences regarding the punishment and training of parents. More research is needed to better understand how and under what circumstances the juvenile's age, a key variable in the development of the juvenile justice system, shapes public opinion and preferences. This study supports past findings (Brank et al., 2011) suggesting no differences in attributions of parental responsibility based on the race or the gender of the juvenile depicted in the vignettes (Research Question 2). In contrast, the perceived seriousness of the case consistently affected perceptions and preferences with regard to parental involvement. As the perceived seriousness of the youth's behavior increased, parents were deemed more responsible as well as more deserving of punishment, and more in need of training. These findings corroborate existing research conducted with parents, in which highseverity offenses resulted in increased assignments of responsibility to parents (White et al., 2007). However, other studies conducted with college students have found that the seriousness of the offense did not have a consistent effect (Brank et al., 2011), calling for further research in this area. Our examination of the possible impact of peer involvement revealed that whether the offense was committed alone or with other individuals did not influence responsibility attributions (Research Question 3). Thus, respondents held parents equally responsible Table 3 OLS regression models predicting parental responsibility and interventions.

Vignette dimensions
Sex ( Note: b = unstandardized regression coefficients; RSE = robust standard errors; CI = confidence intervals. *p < 0.05; **p < 0.01; ***p < 0.001. a Reference category for race is White. b Reference category for type of offense is shoplifting. c Reference category for offending history is no prior record. d Reference category for college level is freshmen. Aizpurua, et al. Children and Youth Services Review 114 (2020) 105023 regardless of whether their child acted alone or in a situation that may have involved pressure from their peers. This finding merits consideration when viewed in the light of substantial evidence that delinquent peers are one of the strongest, most consistent predictors of delinquency (McGloin & Thomas, 2019) and suggestions that parenting deficits and association with delinquent peers interact to produce antisocial behavior among adolescents (Patterson, DeBaryshe, & Ramsey, 1989). Our respondents expected parents to govern their children regardless of peer involvement in their delinquent act, but we did not examine the relative responsibility that people attribute to peers versus parents. Such an investigation may be a useful addition to the literature. The type of offense, a final dimension that we randomized in the vignettes, did have some impact on responses (Research Question 3). Parents were seen as being more responsible for their children running away as opposed to the scenario describing shoplifting. Moreover, cases of running away resulted in increased importance being attached to both punishing parents and providing them with training. Other types of crimes did not consistently influence attributions of parental responsibility and preferences for interventions aimed at parents. At this point, it is unclear why people would hold parents more responsible for their child running away than for other crimes. We can speculate, however, that running away may uniquely raise a desire for parental accountability because it implies, in a way that other offenses do not, that the youth had been overtly seeking separation from his or her parents.
Among the respondents' demographic variables, only gender had a significant effect and only on one of our outcomes-the perceived importance of training. Female respondents placed more emphasis on training than did male respondents. Our findings of a limited effect of gender and that no other respondent demographic characteristics were significantly related to opinions is consistent with previous research indicating that demographic variables are not strong predictors of parental blaming or support for their punishment (Brank & Weisz, 2004). Notably, our null results in this area are also consistent with the broader literature on public attitudes toward responses to crime. That is, studies often reveal no significant relationships between opinions about criminal justice policies and respondents' demographic characteristics. When group differences are reported, they are commonly small and a matter of intensity rather than direction (Applegate, 2020;Cullen, Fisher, & Applegate, 2000). Other variables not included in prior studies, such as perceived risk and fear of victimization, were also not significant in any of the models, suggesting that fear is not driving support for parental responsibility. This result comports with recent studies indicating that fear of crime and perceived risk have no effect on views about juvenile-specific sanctions (Welch, Butler, & Gertz, 2019).
Despite the contributions of the study to better understanding factors shaping attributions of parental responsibility, limitations of our research should be noted. The participants comprised a non-random sample of college students from two universities located in the southeastern region of the U.S. Although the use of college student samples is common in experimental social science research, broadly, we should expect our sample to be more homogeneous and differ in relevant aspects when compared to the general population (education level, race/ ethnicity…), which raises questions about generalizability. A more specific limitation that emerges from studying this group has to do with the possible influence of parental status. Under the assumption that we would find little variation, we did not measure whether our respondents had children of their own. Because previous research points to potential differences between parents and non-parents regarding notions and predictors of parental responsibility (Brank et al., 2011), future studies should control for this variable. It would also be productive to seek to replicate our findings using random samples from the general population. While acknowledging the generalizability limitations, the experimental design used guarantees a high degree of internal validity, allowing us to ascertain how the factors manipulated affect attributions of parental responsibility.
Researchers may also consider including additional variables in their models, such as the number of children and their ages, and indicators of social values including political orientation, religiosity, social anxiety, authoritarianism, and racial animus, which have emerged as important predictors across several domains of public opinion toward responses to offending (Brank & Weisz, 2004;Brown & Socia, 2017;Gerber & Jackson, 2016;Mears et al., 2015;Unnever & Cullen, 2010;Unnever, Cullen, & Jonson, 2008;Welch et al., 2019). Finally, we focused solely on people's opinion on parental involvement. Important complementary research should examine support for civil and criminal liability for parents and the factors that impact these views. Throughout this article we characterize training as a non-punitive response. However, we cannot rule out the possibility that respondents viewed it otherwise, particularly if they believed that this initiative may be mandatory. Further, the history of justice interventions reveals that benevolent intentions have not always resulted in supportive practices (Rothman, 2002).
Despite the limitations cited, the current study adds to the scarce literature in this area by analyzing the effects of multiple case and offenders' characteristics on perceptions of parental responsibility. Factors included in previous research were examined in more depth by including additional levels (juvenile's age, race, and criminal history, as well as offense type) and new variables were incorporated into the vignettes (peer involvement). Results indicate that age influences preferences for interventions aimed at parents, while offense severity influences both preferences and levels of responsibility attributed to the parents. This study is the first to examine public support for parental training, a form of parental involvement required in multiple statutes (Brank et al., 2005), and has revealed nuances in views regarding parental responsibility. Considering that parental responsibility laws have been widely embraced in the United States and European countries, it is important to understand public support for such laws as well as the underlying factors for such support. This study is an attempt to improve our understanding in this regard.

Conclusion
Using an experimental design, the current study assessed how the characteristics of the offenses committed by juveniles and their own characteristics affect perceptions of parental responsibility and the importance attached to punitive and non-punitive responses aimed at parents. The findings reveal substantial support for holding parents accountable, favorable views of punishing parents for their children's misbehavior but stronger endorsement for parental education. Although the public largely attributes crime to lack of discipline from parents (Crime Survey for England and Wales, 2016), our research and others indicate that support for parental responsibility does not equate to parental blame and a strong desire for punishment (Brank & Weisz, 2004). Other institutions -the juvenile justice system, child protective services, schools, community organizations, and the adult criminal justice system -were also expected to be somewhat responsible for "dealing with" juvenile delinquents. In this respect, our results suggest that people still believe "it takes a village to raise a child" (Clinton, 1996;Moon, Cullen, & Wright, 2003, p. 32). The responsibility attributed to parents, however, outstripped each of these other groups, showing the cultural primacy of parents in addressing juvenile crime (Brank & Scott, 2012). E. Aizpurua, et al. Children and Youth Services Review 114 (2020) 105023 CRediT authorship contribution statement

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A
See Table 1A.

Table 4B
Ordinal regression analysis predicting parental responsibility and interventions.  Note: b = unstandardized regression coefficient; RSE = robust standard errors; CI = confidence intervals. *p < 0.05; **p < 0.01; ***p < 0.001. a Reference category for race is White. b Reference category for type of offense is shoplifting. c Reference category for offending history is no prior record. d Reference category for college level is freshmen. E. Aizpurua, et al. Children and Youth Services Review 114 (2020)