Introduction

Most research on the continuity of criminal behavior has been done within individuals. However, few studies looked at the intergenerational continuity of crime. It has long been recognized that crime runs in families. Studies on the genetic nature of crime are old and sometimes controversial (Rafter 1988). For instance, the famous anthropological study of the criminal Kallikak family (Goddard 1912) was very popular when it was published but it appeared, after closer examination, to have been fabricated to some extent (Carlson 1980; Fancher 1987). Today, this specific study usually is described as an “example of eugenic propaganda” (Elks and O’Brien 2005, p. 279). After the second World War, the intergenerational continuity of crime became the subject of serious quantitative research in criminology (Glueck and Glueck 1950; Farrington et al. 1996; Robins and Lewis 1966; van de Rakt et al. 2008; Thornberry et al. 2003b).

Why would this subject fit with the theme of compassionate criminology, which is the focus of this special issue of the European Journal of Criminal Policy and Research? We believe that compassionate policies should be based on the best available research. The better we understand the steps that lead to delinquent behavior, the better as a community, we will be able to help those at risk.

It is important to note that preventing crime is helping society at large but also those with delinquent tendencies. Conduct disorders and aggressive behavior at early ages are predictive of many negative outcomes. Children with conduct disorders have a higher chance of psychiatric problems in adulthood, such as depression, addiction, personality disorders (Matthys 2003), substance abuse and smoking during pregnancy (Elliott 1993; Kodl and Wakschlag 2004), health problems (Laub and Vaillant 2000; Bardone et al. 1998), a lower level of cognitive functioning (Huesmann et al. 1987) and a higher likelihood of becoming a victim of crime and violence (Shepherd and Farrington 1995). Furthermore, they have a lower likelihood of graduating from school and a higher likelihood of becoming unemployed (Kokko and Pulkkinen 2000; Robins 1966). Fortunately, experimental studies show that early interventions help mothers and their children in avoiding these negative consequences, suggesting there is hope for better outcomes (Olds et al. 1998, 2010; Thomas and Zimmer-Gembeck 2007).

Few studies have documented the nature and degree to which intergenerational continuity actually occurs and even fewer have used officially registered crime across three generations to support such propositions. Cross-generational information is important for understanding the degree of risk children born into offending families face and for determining whether these families might be selected for intervention and prevention programs.

Previous research showed that crime is heavily concentrated within families (Beaver 2012; Besemer and Farrington 2012; Hjalmarsson and Lindquist 2012; Thornberry 2009). For example, in London, 5 % of the families participating in Cambridge Study of Delinquent Development (CSDD) were responsible for 24 % of offenders, and 46 % of convictions (Farrington et al. 1996 (p. 9)). In Pittsburgh 4.7 % of the families contained 30.4 % of offenders (Farrington et al. 2001). Indeed, the criminal involvement of family members is one of the best predictors of the crime involvement of any individual within that family (Farrington 2007).

The likelihood of criminal convictions for an individual seems to increase with the number of convicted family members (Farrington et al. 1996, 2001), although some studies could not find this dose–response effect (Robins and Lewis 1966; Murray et al. 2012; Besemer and Farrington 2012). In the Cambridge Study of Delinquent Development (CSDD) the criminal involvement of the father was the most important predictor of crime of his son (Farrington 2007, 2001). Alternatively, Robins and Lewis (Robins and Lewis 1966) reported that grandfathers were most influential on children (men and women) within the family.

Apart from parental crime, two aspects of family violence were found to be predictive of future criminal behavior. First, child abuse is related to future crime. A number of studies showed that physical child abuse and neglect predict future criminal involvement (Widom and Maxfield 2001; Cicchetti and Toth 2005) and some authors argue that child abuse explains the intergenerational transmission of externalizing behavior (Verona and Sachs-Ericsson 2005). Second, partner violence is also predictive of future crime in children (Margolin and Gordis 2000).

Many studies investigating the intergenerational transmission of criminal behavior were characterized by methodological limitations (Parsons-Pollard 2011). Such limitations take many forms. First, as far as could be assessed, all previous research is based on selected samples that may lead to biased estimates of effects. Several studies focused on high-risk groups (Farrington et al. 1996, 2001; Robins and Lewis 1966) which most likely inflates the degree of effects and thus limits the generalizability of the findings.

A second problem is that many studies used self-reported offending to measure criminal involvement (Parsons-Pollard 2011). Self-report measures of crime have generally been validated for adolescents but have been judged to be less valid and reliable among adults (Junger-Tas and Haen-Marshall 1999).

Third, self-report studies are generally troubled with significant drop-out and non-response rates (Junger-Tas and Haen-Marshall 1999).

Fourth, most studies using official records tend to use relatively small samples ranging from N = 67 in the Robins study (Robins and Lewis 1966) to N = 411 in the Cambridge Study of Delinquent Development (CSDD) (Farrington et al. 1996).

Fifth, many previous studies (Robins and Lewis 1966; Farrington et al. 1996, 2001) focused on males only, much less attention has been given to females.

Finally, although several studies examined parenting as well as parental criminal behavior (Farrington et al. 2001; Johnson et al. 2004; Dogan et al. 2007; Thornberry et al. 2003a), few studies examined crime as well as family violence. This is an important omission as intra-family violence is seen as significantly associated with future anti-social and criminal behavior of other family members.

In the present study, the official criminal records of the grandparents, parents and siblings of an entire birth cohort were used to estimate risk, increasing the representativeness of the sample, while eliminating both the non-response and self-report biasing problems previously mentioned. Four questions are addressed:

  1. 1.

    What is the prevalence of offending, as registered by the police, across three generations of families in a medium-sized city in the Netherlands?

  2. 2.

    What is the concentration of offenders and arrests within families?

  3. 3.

    Are there interrelationships between offending among the relatives across three generations?

  4. 4.

    Is family violence related to offending in family members?

Method

Sample

The sample of the study consists of all 1681 families that had a child born in 2006 in a Dutch city of approximately 152,000 inhabitants. This is a medium-sized university town in the east of the Netherlands. About one quarter of the population is of non-Dutch decent.

Measures

Arrests

Information on involvement with the police for criminal offenses was collected from an electronic database that is used by the Dutch police. This electronic database contains information about persons suspected of a crime as well as about incidents. It contains information on all persons that are suspected of a crime by the police. The present study focused on acts falling under the criminal law and did not include misdemeanors and traffic violations.

For the present study information was retrieved on all arrests as a result of being suspected of a crime, as defined by the criminal law.

Arrests were recorded for mothers, fathers, grandparents and siblings. For approximately 2/3 of the sample grandparents addresses were not known to the municipality. In such cases they could not be searched for in the police records. The smaller number of available grandparents included in this study obviously influences the conclusions that we may draw. However, having one-third of these family members in the sample (2184 grandparents, see Table 1, below) provides a large number of individuals of the third generation under study. All contacts were summed, therefore the measures used are life-time arrests as registered by the police.

Table 1 Prevalence of arrests for family members among cohort children, in % (N = 1674)
Full size table

For every person registered by the police information goes back until 1985 and for every incident registered by the police information goes back until 1998. The data were collected during the first semester of 2007. Accordingly, this study collected information on 22 years of offender registration and 9 years of incident registration. This means that there is a fixed identical period for all family members for which information is available. Because older family members may have been at risk for a longer period than younger family members, odds ratio’s between the prevalence of arrests of family members have been computed, with and without controlling for age.

At the time of the data-collection, only 15 of the 26 Dutch regional police forces were participating in this electronic database. These 15 police forces covered about 57 % of the Dutch population, including the police region that covers the city in this study. For all the relatives known to the municipality, a search was done in the database to investigate if that particular family member had been arrested. It is possible that a family member or family member’s arrests was missed because of the nine Dutch police forces that were not involved in this project at the time of the study. Nonetheless, it is not very likely that a significant number of arrests of family members have been missed, because research shows that offenders tend to commit most crimes close to home (Rengert et al. 1999), and this also holds for the Netherlands (Rengert et al. 1999; Bernasco 2006).

For ease of presentation a family member is referred to as a ‘suspect’ whenever that person has had contact with the police.

Family Violence

Information on family violence consisted of a combination of two types of violence, namely partner violence and physical child abuse. The police do not code information on family violence systematically. However, every time a family member of a selected child was registered in the database, the description of every incident was read carefully by the researchers. When this description mentioned violence between partners or mentioned that a child in the household was physically abused, or if the police were notified by others (e.g., neighbors) about possible physical abuse or partner violence, this information was recorded. A dichotomous variable was constructed (1 = family violence, 0 = no family violence). Incidents of family violence were not coded twice, under a different code. The combination of information on partner violence and physical child abuse was done because of practical reasons: in the present sample numbers were too small to use both separately. Although this is not ideal, studies showed that there is a relatively strong relationship between partner violence and child physical abuse. It was estimated that in violent homes the co-occurrence is about 40 % (Appel and Holden 1998).

Information on sex and age was available for all family members. Age of eldest sibling was coded as: age 1–11 = 1, age 12–14 = 2, age 15–17 = 3, age 18–20 = 4, and 21 or more = 5.

Information on the demographic characteristics of the families was collected at the municipality.

Analyses

Analyses were done by computing bivariate odds ratios. When looking at the concentration of offending, two generation families (siblings and parents) and three generations families (siblings and parents and grandparents) were analyzed separately. Relationships between family members were analyzed by computing odds ratios (OR). These odds ratios are presented without and with controlling for age, as a proxy for time at risk, or risk exposure.

The study was approved by the Police Force Gelderland-Zuid whose rules conform to the Dutch law on the protection of databases (Wet Bescherming Persoonsgegevens, WBP) and the law on Police Registrations (Wet Politie Gegevens, WpolG). Furthermore, research in the Police Force is done according to the Regional Research Protocol (onderzoeksprotocol Gelderland-Zuid) which is based on the Code of Scientific Research developed by the Organization of Research Institutes (Vereniging van Onderzoeksinstituten).

Results

The cohort for this study was composed of 1681 families with a child born in 2006. For seven children no information was available about any relative; they were excluded from further analysis, reducing the cohort to 1674 children/families. There were older siblings present in 52 % of the families; 3.5 % of the families 3 children, except for the newborn; 8.2 % of the children were born to a single parent family; 2.7 % of the mothers were 20 years old or younger when they gave birth to the study child; 23.5 % of the children were born in a disadvantaged neighborhood; and 64.6 % of the fathers were of Dutch nationality. The nationality of other fathers included Moroccan (4.3 %), Turkish (3.8 %), Surinamese or Dutch Antilles (1.9 %), Eastern Europe (2.1 %), Middle East (2.1 %), Western Europe (4.9 %) or other (9.2 %) or the father was unknown (8.2). In 1.6 % of the families family violence was present.

Prevalence of Arrests in Three Generations

Table 1 presents the number of relatives identified for the children in this cohort as well as the number of relatives with arrests. Mean and range information are also provided for the age data. 7.2 % of the mothers and 18 % of the fathers of the newborn children had arrests. For 3.3 % of this cohort of newborns, both mother and father have arrests and in 20.3 % of the families, either the father or the mother had arrests. 1.7 % of all newborn children were born into a family where an older sibling already had a registered arrest. Collectively, slightly over one-quarter (25.2 %) of the children were born in families were at least one family member had prior involvement with the police for criminal matters.

A slightly higher percentage of mothers’ parents (both grandmother and grandfather) had arrests in comparison to fathers’ parents. Of maternal grandmothers 8.1 % and 4.9 % of paternal grandmothers had been arrested, and 10.7 % and 8.7 % of the grandfathers on mother’s and father’s side, respectively, had an arrest record. From additional analysis (not shown) it appeared that 1.5 % of the families were known to the police because of family violence.

Concentration of Suspects and of Arrests Within Families

The concentration of offending was studied by examining the number of suspects within families (Table 2). Table 2 shows the cumulative number and percentage of suspects accounted for in the three-generation families, for families without suspects, families with one suspect, families with two suspects, and so forth. The vast majority of families, 74.8 %, contained no suspects, while 17.4 % of the families had one relative with police contact, and 5.3 % of families had two relatives with arrests (column b). As the data in Table 2 reveal, approximately one quarter of the families (25.2 %) are responsible for all crime suspects. Columns d and e show the number of suspects for each category of families; 7.8 % (column (c)) of the families accounted for 52.3 % (column (e)) of the suspects.

Table 2 Concentration of suspects within families: three generations: grandparents, parents and siblings
Full size table

A similar analysis is presented in Table 3, for the two generation-families (parents and siblings). Of the families of this birth cohort (column (c)) 21 % are responsible for all the suspects (column (e)), and 4.3 % (column (c)) of the families are responsible for 36 % (column (e)) of the suspects.

Table 3 Concentration of suspects within families: two generations: parents and siblings
Full size table

The concentration of arrests within families is presented in Table 4. Among three generation-families, 1330 children (column (a)) are born into a family where no one has ever been arrested. Of the children (column (c)) 5.1 % are born into families that are responsible for 73.5 % (column (e)) of all arrests. When studying two generations (parents and siblings, table not shown) 5.4 % of the families were responsible for 81.8 % of the arrests.

Table 4 Concentration of arrests within families: three generations: grandparents, parents and siblings
Full size table
Table 5 Inter-relations among arrested relative: odds Ratio, without and with control for age of both relatives (in bold)
Full size table

Interrelationships Between Offending Among the Relatives in Three Generations

Within the families of the cohort children, arrests of one relative are usually related to arrests of another relative. The figures corrected for age are used below to discuss the findings (Table 5). On the mother’s side, for example, grandparent’s arrests increase the likelihood of maternal arrests by a factor three (OR = 3.3 for both the maternal grandmother and grandfather). The same is true on fathers’ side: when father’s parents have been arrested, the likelihood that their son was arrested increases, with an OR of 2.9 and 3.9 for his father and mother, respectively.

Interestingly, maternal arrests are related to arrests of her parents in-law, with OR of 2.9 and 3.9 for her mother and father-in-law, respectively. Similarly, paternal arrests are related to the likelihood of arrests of mother-in-law, with OR of 2.7 but not his father-in-law.

Within generations, the relationship of arrests between partners is relatively strong, with an OR = 7.1 for fathers and mothers, OR = 3.5 for the mothers’ parents and OR = 4.9 for the fathers’ parents.

The likelihood that the family has been reported for family violence increases with the likelihood of arrests of the mother (OR = 3.2), of the father (OR = 3.9) and also the maternal grandfather (OR = 3.9).

Arrests of siblings of the cohort child were related to mothers arrests (OR = 3.8) but not to arrests of other relatives.

Discussion

The present study investigated arrests within a complete 2006 birth cohort of children and their families in a middle-sized city in the Netherlands. By using a full birth cohort and focusing on arrests by the police, the present study overcomes some of the limitations of prior studies about the concentration of crime in families.

The study shows that 7.2 % of the mothers of the cohort children and 18 % of the fathers have recorded arrests. Perhaps the most surprising finding in the present study is that 3.3 % (N = 56) of the infants born in 2006 in this region of the Netherlands had two parents who were arrested for a crime, and 20.3 % had at least one parent who had a prior arrest. However, these findings are within the range of what was found elsewhere in the Netherlands (Junger et al. 2001).

The present study finds that arrests are heavily concentrated in families; 7.8 % of the families (parents, siblings and grandparents) are responsible for 52.5 % of all arrests. This figure shows that suspects are slightly more concentrated within families in this city in the Netherlands as compared to Pittsburgh, where 8.4 % of the families contained 42.6 % of the suspects (Farrington et al. 2001).

Arrests were even more concentrated than suspects: 5.1 % of the families were responsible for 73.5 % of the arrests. In London, for example the concentration was strong but not as strong as in this city in the Netherlands: 5 % of the London families, containing 8 % of the family members accounted for 46 % of the convictions. It should be noted that it will always be the case that offenses are more concentrated than offenders. Since police reports are based on contacts with suspected persons who often, committed more than one crime, there usually are more offenses than offenders and this leads to stronger concentration of offenses than offenders within families.

Several differences between the current sample and the Cambridge Study of Delinquent Development could account for these differences. The Cambridge study was conducted on a high-risk sample and estimates of concentration are based on lifetime convictions whereas the current study focuses on a full cohort and arrests, each study focusing on different aspects of the justice process—arrest versus conviction. Taken together the current findings suggest that by using a full cohort of children, similar estimates of the concentration of suspects in families are found but there is a much higher concentration of arrests in the present study, suggesting that studying a complete cohort leads to higher percentages of concentration. Still some of the differences may be associated in part with the size and complexities of the samples studied.

Important for this study, however, there are relatively strong interrelationships between arrests among the relatives of different generations. The odds ratios between grandparents and parents vary between 2.9 and 3.9. Mother and father arrests were also related to the likelihood of arrests with their respective parents-in-law with one exception: the relationship between father and his father-in-law was not significant.

That is, there is clear evidence for assortative mating, with OR’s varying between 3.5 and 7.1. This corresponds with previous findings (Farrington et al. 1996, 2001). Partners tend to resemble each other with respect to various behavioral characteristics, including antisocial behavior (Krueger et al. 1998) but less so on more general psychological characteristics (Krueger et al. 1998; Lykken and Tellegen 1993).

The likelihood of an arrest of the eldest sibling of the cohort child was related to maternal arrests (OR = 3.8) but not to arrests of other family members. This is remarkable as the siblings were still quite young, with 92 % younger than 12, which is the age of criminal responsibility in the Netherlands. This corresponds with previous research that reported that officially registered crime is one of the strongest and most consistent predictors of future crime and antisocial behavior, more generally, even after controlling for individual characteristics and parenting (Farrington. et al. 2006, 2001; Besemer and Farrington 2012; Thornberry 2009; Hjalmarsson and Lindquist 2012; Beaver 2012; van de Rakt et al. 2008). The present study confirms this general finding that crime in one generation predicts crime in the next.

Family violence was a correlate of maternal, and paternal arrests as well arrests of maternal grandfather. This corresponds with previous research investigating the relationship between family violence and criminal behavior (Farrington 2007; Widom and Maxfield 2001).

The intergenerational continuity of criminal behavior is often explained by two global and likely complementary approaches: it may have a genetic or a parenting explanation. There is a large body of research supporting the claim that there is a strong genetic component (Rutter et al. 1999; Mason and Frick 1994; Rhee and Waldman 2002), primarily resulting from the inheritance of individual characteristics such as intelligence, impulsivity or aggression (Farrington et al. 1996). A meta-analysis conducted by Rhee and Waldman (2002) concluded that 32 % of the variance in measures of antisocial behavior is attributable to additive genetic effects and 9 % of the variance is attributable to nonadditive genetic effects. Environmental effects are responsible for 16 % of the shared environmental and 43 % of the non-shared environmental influences on antisocial behavior.

However, the transmission of crime from one generation to the next can be explained by environmental factors as well. For example, several authors found that the strong similarity between parents and children on crime involvement was mediated through parenting and family functioning (Dogan et al. 2007; Johnson et al. 2004). Surveys, however, have been criticized for not taking into account genetic influences on behavior (Wright and Beaver 2005). Stronger evidence supporting the importance of parenting comes from experimental studies showing that improved parenting leads to significant improvements in child behavior (Thomas and Zimmer-Gembeck 2007) (see also point four, below).

Regardless of the specific mechanism by which crime becomes concentrated in families, our data as well as the previous research described above show that we can identify children who are at risk even in their first year of birth. As maternal involvement in crime and family violence is related to children’s involvement in crime, information on parental antisocial behavior and family violence can be tracked on the basis of mother’s characteristics and behavior already during pregnancy, and can help locate families in which children are at a higher risk for poor developmental outcomes. Police registers contain information that is relevant for children’s actual and future welfare (Hagell and Jeyarajah-Dent 1999; Fantuzzo and Fusco 2007). Police officers have been described as ‘public health sentinels’ (Fantuzzo and Fusco 2007).

Information in police records is also important because this can be checked early in life. Early detection and intervention are important for several reasons.

Importance of Early Detection and Intervention

First, children and adolescents involved in criminal behavior are at risk for many other negative outcomes later in life beyond criminal behavior (Campbell 1995). For example, they have an increased risk for psychiatric problems in adulthood, such as depression, addiction, personality disorders (Matthys 2003), substance abuse and smoking during pregnancy (Kodl and Wakschlag 2004), spouse assault (Farrington et al. 1996), a lower level of cognitive functioning (Huesmann et al. 1987), and a lower likelihood of graduating from school and a higher likelihood of becoming unemployed (Kokko and Pulkkinen 2000). Serbin and her colleagues (Serbin et al. 1998) found that aggressive behavior and depressive symptoms in girls measured during primary school (when they were between the ages 5 and 13 years of age) were predictive of aggressive as well as non-responsive behaviors in their children. This finding is noteworthy in that there was approximately 20 years between the measurements of mothers and their children.

Second, it seems fair to say that there is no cure for involvement in crime. After the fact, few reactions to crime among adolescents and adults, seem to help to prevent it from happening again. Investing in more criminal justice seems a waste of time and money. Studies indicate that recidivism rates among those convicted to a prison sentence in the Netherlands are usually around 75 % (Wartna et al. 2005). A recent meta-analysis looked at 111 studies which studied the association between type of sanction and recidivism (Smith et al. 2002). Overall, 442,000 offenders were involved in these studies. The overall effect size was zero. The authors conclude that harsher sanctions had no deterrent effect on recidivism. These findings were consistent across subgroups of offenders, namely adult—youth, male—female, and white—minority. Compared to community sanctions, a prison sentence produced a slight (3 %) increase in recidivism. Smith, Goggin, & Gendreau (Smith et al. 2002) conclude that ‘getting tough on crime’ by harsher punishment will not reduce recidivism, and that this holds for all categories of offenders. Measures aiming at providing work after prison usually also found discouraging results with no notable effect on recidivism (Van Dijk 1994; Freeman 1983; Terlouw 1991; Wilson and Herrnstein 1985; Visher et al. 2006).

Third, early intervention is more likely to be successful than interventions later in life. We believe this point has not received the attention that it deserves. ‘The earlier the better’ seems a wise guideline for policymakers. Several points strongly support this. a) Recent research has emphasized the importance of the early years in the healthy development of children (Shonkoff and Meisels 2000). b) As reported earlier, conduct disorders early in life are related to many negative life outcomes and intervening early may help to avoid not just future crime but some of the other associated negative outcomes. c) An economic analysis of experimental studies of interventions beginning at different ages concluded that the most effective way to allocate recourses, given the expected outcomes, is to invest heavily in the first years of life and those interventions later in life produce increasingly less benefit (Carneiro and Heckman 2003; Heckman 2006). d) Experimental evidence shows that some effective interventions, such as the Nurse-Family Partnership (see below) are effective as a preventive intervention but not as a curative intervention. For instance, a series of studies done by Olds (Olds et al. 1999) indicates that the Nurse-Family Partnership program, conducted by nurses, that starts during pregnancy and continues until the child is 2 years of age can significantly reduce the likelihood of child maltreatment as well as provide other beneficial effects to mother and child. In comparison, other research has found that an intensive home visitation program starting after child maltreatment had appeared was not effective (McMillan et al. 2005). Again, though not conclusive, these two studies taken together suggest that preventive programs aimed at high-risk groups may be more effective than are interventions targeting those already showing problem behaviors.

Fourth, some early parenting trainings that have been found effective: PCIT is a parenting program that has been found effective to combat conduct disorders in children age 2 to 8 and is also one of the few evidence based programs to help parents who physically abuse their children (Schuhmann et al. 1998; Thomas and Zimmer-Gembeck 2007). Several of these programs have been evaluated rigorously in randomized controlled trials and have a variety of positive outcomes for children (Schweinhart 2004; Olds et al. 1998). It is important for policymakers and professionals to know that there are some evidence based treatments.

Fifth, several preventive programs have been studied by means of a randomized controlled trial and have shown to have long term benefits for children. As mentioned above, the Nurse-family Partnership has been studied and replicated twice and provided pre-school programs have been found to help avoid negative outcomes in children (Olds et al. 2010). Also, preschool programs can help children (Karoly 2012). Many positive outcomes were reported from the Perry-Preschool program. A lengthy follow-up indicated that at the age of 40, the now adult children from the experimental group performed significantly better in many and various ways in comparison to the adult children from the control group. For instance, fewer program individuals dropped out of school or experienced grade retention. They also fared better economically than their non-program participating counterparts as indicated by their higher likelihood to have a job, earn a salary and own their own homes and their lower likelihood of being on welfare. Finally, in terms of criminal involvement, participants in the experimental group were arrested less often and where arrested for less violent offences than those who had been in the control group. Those in the experimental group also scored lower on a variety of additional crime indicators (Schweinhart 2004). A replication study has since confirmed the original study’s findings. Again treated individuals fared better overall including on measures of delinquency and crime than did their control counterparts (Schweinhart and Weikart 1997).

Early and sensitive prediction of negative outcomes, such as criminal behavior will rely in some measure on information contained in police reports and contacts with families. For example, in Boston, the Grove Hall Safe Streets Initiative Klofas and Hipple (2006), part of the federal Weed and Seed Initiative is focused, among many things, on identifying those families with the highest amount of criminal indictments, and the children within those families who have yet to embark on serious criminal careers. Such efforts have led to the coordination of police and social services to these families, and most especially to the children in these families, who are clearly at higher risk for criminal involvement. Such efforts create the possibility of linking police knowledge of high risk families with social service agencies capacity to delivery timely and targeted services to those families and children in most need of such services. Such a shift in police intervention highlights the differences between the police acting to prevent crime and community harm, as opposed to merely responding to such conditions and actions after they have occurred (Greene 1998, 141–160). Such a prevention role for the police is consistent with shifts toward community and problem-oriented police strategies.

Limitations and Contributions

To better appreciate our results, some advantages and limitations of the present study should also be addressed. First, this study is based on an analysis of an entire birth cohort in an open population of a middle-sized city, allowing generalization of the findings to a wider population than previous studies that have focused on high risk samples. Second, this study is based on official data and therefore is not subject to reporter bias or non-response problems. Third, the current study is considerably more accurate on the number of arrests per family member, while self-reports are likely less accurate (Farrington et al. 2001).

Limitations of the study include the fact that we did not have information on the timing of arrests and therefore were unable to control for time at risk of offending for each family member. This may be important in the case of non-Dutch families as the degree of prior offending may be underestimated. The electronic database only began in 1998 and at the time of the analyses only covered 57 % of the area in which the Dutch population was living. Although the city in which the families were living was in the center of that area and most crimes occur close to home, we are probably underestimating at least slightly, criminal involvement. Finally, our analyses for female and male siblings separately should be interpreted with caution, as overall, only small numbers of siblings had been arrested.

Despite these limitations, the observed results indicate that crime statistics on an entire cohort could be a useful tool for identifying at-risk children. From a policy perspective the next step would be to implement an early, evidence-based intervention strategy to prevent these children from repeating the mistakes of their predecessors, based on these and others’ findings that families evidencing criminal involvement often yield children at-risk of criminality.