Introduction

Life-course criminology has made great progress over the last three decades, yielding a wealth of knowledge about the relationship between age and crime, and the different career paths that criminals can follow (Onna et al., 2014). Largely building on Moffitt’s (1993) dual taxonomy, empirical support has been found for adolescence-limited offenders, life-course-persistent offenders, and other groups such as late-onset offenders (Jennings & Reingle, 2012; Piquero, 2008). Scholars have also tried to explain differences in criminal career trajectories by looking at personality characteristics and long-term risk factors (Farrington, 2003; Laub et al., 2018; Le Blanc & Loeber, 1998; Piquero et al., 2003). However, attempts at explaining why some delinquents desist after adolescence while others persist to commit crimes have yielded inconsistent results (Bersani et al., 2009; Sampson & Laub, 2003).

The current paper explores whether differences in criminal trajectories, such as those of adolescence-limited and life-course-persistent offenders, can be explained by considering different types of offences, namely domestic burglaries, street robberies, and residential and commercial robberies.Footnote 1 We focus on these three offences for two reasons. First, in the Netherlands, burglars, street robbers, and residential and commercial robbers are referred to as offenders of high-impact crimes (Ministerie van Veiligheid en Justitie, 2014a, b), to indicate that these offences have a major impact on the victim and society alike. Over the past 20 years, the Dutch government has invested a lot of effort and money into combatting high-impact crimes due to high prevalence rates (Inspectie Veiligheid en Justitie, 2014; Ministerie van Veiligheid en Justitie, 2014a; Taskforce Overvallen, 2009, 2011, 2017; Ministerie van Justitie en Veiligheid, 2019; Ministerie van Veiligheid en Justitie, 2014b). This lends external validity to our unique focus on burglars and robbers, in the Netherlands specifically.

Second, burglars and robbers may differ from the general offender population in ways that are crucial for their career trajectories. Previous research in Western countries has shown that domestic burglars, street robbers, and residential and commercial robbers often engage in criminal behaviour from a young age, commit a relatively large number of offences, and do not limit themselves to one type of crime (e.g. Bernasco, 2009; Blokdijk & Beijersbergen, 2020; Fox & Farrington, 2016; Kros & Beijersbergen, 2021; Maguire et al., 2010; Mesu et al., 2013; Owen & Cooper, 2013; Rovers et al., 2010; Svensson, 2002). These findings suggest that domestic burglars, street robbers, and residential and commercial robbers have extensive criminal careers that likely differ from the criminal careers of other offenders, both in nature and in size.

What is more, the criminal career characteristics of burglars and robbers align rather well with characteristics often associated with life-course persistent offenders (Moffitt, 1993). Life-course-persistent career trajectories could be unique to domestic burglars, street robbers, and residential and commercial robbers. Distinguishing between these three offender groups and other offenders could help to explain why some delinquents restrict criminal behaviour to their adolescent years, while others remain criminally active well into adulthood. This would be in line with studies undertaken in the UK and Sweden that show that delinquents who start their criminal careers with certain debut offences, such as burglaries and robberies, are more likely to become chronic offenders (Owen & Cooper, 2013; Svensson, 2002). Furthermore, recent contributions to the literature on life-course criminology have shown that certain career characteristics may be specific to particular offender groups, such as late-onset for members of mafia or white-collar criminals (Campedelli et al., 2021; Van Koppen et al., 2010).

In this paper, we investigate whether differences in criminal trajectories, for instance between adolescence-limited and life-course-persistent offenders, can be explained by considering different types of offences, namely burglaries, residential and commercial robberies, street robberies, and other offences. Efforts to identify life-course persistent offenders early on, in an attempt to turn young delinquents away from a long life in crime, could then focus on criminals who are involved with specific offences rather than individual risk factors such as psychosocial problems. This would have important policy ramifications, as life-course persistent offenders who commit a large share of relatively serious criminal offences exert tremendous costs on society (Cohen & Piquero, 2009). Being able to identify life-course persistent offenders early and accurately may improve intervention and prevention efforts. However, little is currently known about the long-term career trajectories of burglars and robbers, and how this relates to the career trajectories of life-course persistent offenders. We therefore address two main research questions: (1) What are the characteristics of the criminal careers of domestic burglars, residential and commercial robbers, and street robbers? (2) Do the criminal trajectories of domestic burglars, residential and commercial robbers, and street robbers differ from the criminal trajectories of other offenders?

In order to answer these questions, we make use of a large-scale longitudinal dataset that includes information about all registered criminal cases in the Netherlands from 1997 until 2020. Our data includes offenders who had their first criminal case when they were 12 and offenders who had their last criminal case (provisionally) at age 92. Such a long-term perspective on criminal careers is crucial yet largely missing in the literature on career trajectories to date (Piquero, 2008; Van Koppen et al., 2010). Because we can study criminal activities well into adulthood, we can more accurately capture whether offenders have actually desisted after adolescence or whether they have briefly stopped with delinquent behaviour only to resume their criminal activities later on.

Criminal Careers

There is a long tradition of research on criminal careers, with criminologists studying different dimensions of criminal careers, like the (age of) onset, frequency, seriousness, and (age of) desistance of criminal behaviour (Blumstein et al., 1986; Piquero, 2008; Piquero et al., 2003). The general assumption of the criminal career approach is that criminal careers differ per offender and, therefore, the age-crime curve is not the same for all offenders. The popularity of the criminal career approach and developmental and life course criminology increased with the finding by Wolfgang et al. (1972) that a small group of offenders committed the majority of crimes (Blokland et al., 2005; Laub, 2004). This finding had clear policy implications: being able to identify these chronic offenders at a young age would have significant ramifications for policymakers who try to intervene during childhood and prevent serious offending in adolescence or later on in life (Blumstein et al., 1986; Cohen & Piquero, 2009). To this day, theoretical and empirical attempts are made to identify and characterize these chronic offenders and explain differences in criminal careers (e.g. Bersani et al., 2009; Blokland et al., 2005; Blumstein et al., 1986; LeBlanc & Loeber, 1998; Moffitt, 1993; Piquero, 2008; Piquero et al., 2003; Sampson & Laub, 1993; Thornberry, 1987).

While there have been a number of developmental and life-course theories that try to explain differences in criminal careers, the more influential ones include Moffitt’s (1993) dual taxonomy model and Sampson and Laub’s (1993) age-graded theory of informal social control. First, Moffitt uses a typology approach and assumes that the offender population is composed of subgroups that follow distinctive trajectories of offending, with each group possessing a distinctive developmental aetiology. Moffitt’s (1993) dual taxonomy model distinguishes between adolescence-limited and life-course-persistent offenders. Adolescence-limited offenders (AL) form the largest group and refer to delinquents who limit their involvement in criminal activities to their teenage and adolescent years (Moffitt & Caspi, 2001; Moffitt et al., 2001). For a limited number of years, these delinquents show a peak in anti-social and delinquent behaviour, followed by a rapid decline during early adulthood. The second and much smaller group is labelled life-course persistent offenders (LCP). These offenders show anti-social and criminal behaviour from an early age and remain active as criminals throughout their lives (Moffitt, 1993). Besides distinguishing between these two offender groups based on their respective age-crime curves, Moffitt also argued that the causes of criminal behaviour are different for adolescence-limited offenders than for life-course-persistent offenders. LCP offenders are characterized by psychosocial and neurological difficulties, often in combination with other problems such as complicated relationships with parents. Together, these risk factors spur on problematic and anti-social behaviour, which may in turn result in an amplification of earlier risk factors. Ultimately, this leads to a long-lasting pattern of criminal activities. On the other hand, adolescence-limited offenders are argued not to suffer from the same risk factors during the early stages of their lives. Moffitt (1993) proposes that this offender group tries to gain status by imitating the anti-social and criminal behaviour of their LCP peers. This is the result of the so-called maturity gap: a breach between biological and sociological maturity that results from individuals reaching the age of biological maturity at an earlier age, but not being granted access to the attractive status of adulthood. Once adolescence-limited offenders are old enough to gain status by more conventional means, such as education, marriage, or occupation, they abandon their criminal activities (Blokland et al., 2005). As a result of the different aetiology, Moffitt assumes that the two offender groups also differ in the crimes they commit. Adolescence-limited offenders are expected to engage primarily in crimes that symbolise adult privilege or that demonstrate autonomy from parental control, such as vandalism, public order offences, substance abuse, and theft. Life-course persistent offenders, on the other hand, are expected to engage in a wider variety of offences and in more victim-orientated and serious offences, such as violent crimes (Moffitt, 1993).

Moffitt’s dual taxonomy is widely recognised, and the trajectories of life-course-persistent and adolescence-limited offenders are often found in empirical research (DeLisi & Piquero, 2011). That said, many empirical studies have also found additional career trajectories (Gunnison, 2015; Jennings & Reingle, 2012; Odgers et al., 2008). This eventually led Moffitt to add a third group to her original typology, namely the group of low-level chronic offenders (Moffitt, 2003). These delinquents are persistently involved in criminal activities but at far lower rates than life-course-persistent offenders. Another consistently found and important offender trajectory is that of late-onset offenders, who often start their criminal careers well in adulthood and are typically involved in other criminal activities such as white-collar crime or organised crime (Onna et al., 2014; Piquero, 2008; Van Koppen et al., 2010). Although empirical research finds support for Moffitts’ distinctive offender trajectories, there is little evidence that individual, childhood, or family risk factors can predict differences in long-term trajectories of criminal behaviour (Bersani et al., 2009; Sampson & Laub, 2003, 2005).

Second, Sampson and Laub’s age-graded theory of informal social control (Sampson & Laub, 1993, 2005; Laub & Sampson, 1993) integrates the life-course perspective with social control theory (Hirschi, 1969). It proposes that changes in criminal behaviour are influenced by changes in adult social bonds. According to this theory, social capital and major events such as marriage, parenthood, and taking up permanent employment, are crucial in understanding processes of change in the criminal life course. For instance, getting married or becoming a parent leads to new informal obligations and increased social control, which can lead to desistance from crime. Sampson and Laub (2003) challenge Moffitt’s life-course persistent offender group. They assume a general process of desistance from crime, since their study reveals that offences decline in the middle adult years for all offender groups. Sampson and Laub (2003) introduce the concept of life-course desisters, accounting for the apparent fact that all offenders desist but at different times during their lives. Sampson and Laub’s theory is one of the most empirically tested developmental and life-course theories of crime. There is some evidence for the relationship between turning points and criminal behaviour, although not all prior studies confirm this relationship and it does not seem to hold true for all offenders (e.g. Blokland et al., 2005; Farrington & West, 1995; Laub et al., 2018; Piquero et al., 2002; Uggen & Wakefield, 2008). With regards to Sampson and Laub’s claim that all offenders desist and that life-course persistent offenders do not exist, results are mixed as well. Some studies find support for this proposition (Bersani et al., 2009), while others do not (Blokland et al., 2005).

Contrary to developmental and life-course theories (such as Moffitt’s dual taxonomy model and Sampson and Laub’s age-graded theory of informal social control theory), the general theory of crime by Gottfredson and Hirschi (1990) denies the existence of different criminal careers and different offender groups. According to his theory, offenders lack self-control because it has not developed sufficiently in childhood. There is a single risk factor and cause for criminal behaviour, namely low self-control. This factor explains crime at all ages. Gottfredson and Hirschi (1990) argue that the age-crime curve is the same for all offenders once self-control is developed and that crime declines similarly with age for all offenders.

In line with these three theoretical perspectives, previous research has tried to explain criminal behaviour by focusing on risk factors from early life stages (e.g. neuropsychological deficits, adverse parent–child relationships and low self-control) and turning points (like marriage and employment). However, recent research has taken a different approach to life-course criminology and has shown that certain career characteristics, like age-crime curves, are specific to particular offender groups, such as late-onset for members of mafia, individuals involved in organised crime, or white-collar criminals (Campedelli et al., 2021; Onna et al., 2014; Van Koppen et al., 2010). In a similar vein, we believe it is interesting to have a closer look at the criminal careers of domestic burglars, residential and commercial robbers, and street robbers and to make a comparison with the criminal careers of other offenders. Previous research shows that burglars and robbers are responsible for a large number of crimes and commit their first crime at a relatively young age (e.g. Blokdijk & Beijersbergen, 2020; Fox & Farrington, 2016; Kros & Beijersbergen, 2021; Maguire et al., 2010; Owen & Cooper, 2013; Svensson, 2002). Early age of onset has often been related to long criminal careers (Farrington, 2003). Taken together, these findings suggest that robbers and burglars constitute a serious group of criminals, who are more likely to have extensive criminal careers than other offenders and who are more likely to follow the criminal trajectory of life-course persistent offenders (Moffitt, 1993).

Data and Methods

Data

Dutch Research and Policy Database for Judicial Information (OBJD)

We used data from the Dutch Research and Policy Database for Judicial Information (OBJD). The OBJD is a pseudonymous version of the Justice Documentation System (JDS), the Dutch legal registration system for criminal cases. The JDS features all criminal cases of all persons who came into contact with the justice system in the Netherlands. There is no maximum retention period for data in the OBJD, in contrast to judicial data that is used for criminal justice practices. Therefore, OBJD data is very well suited to investigate criminal careers over a long period of time, as the database contains judicial data about all individuals aged 12 or olderFootnote 2 who came into contact with the justice system since 1997. For this study, we have used judicial information about cases that were registered in OBJD up until June 2020. It is important to note that the use of OBJD also implies that we could only look at crime that comes to the attention of the Public Prosecution Service. Criminal activities and criminals that are not detected and are not reported to the Public Prosecution Service are not included in our data. Furthermore, all our analyses were based on criminal cases instead of offences. OBJD data is processed automatically, and this process is based on analyses at the criminal case level. Analyses at offence level require a new, yet to be developed method for using the OBJD. The fact that we analyse criminal cases instead of offences is important to keep in mind when interpreting our findings. In particular, it is possible that by analysing criminal cases instead of criminal offences, we underestimate the total frequency of criminal behaviour even more than studies that look at registered criminal offences. This is because a criminal case may contain more than one criminal offence.

Time-in-Prison-Table (TIP-Table)

The second dataset we used is the time-in-prison-table (TIP-table) developed by the Research and Documentation Centre (WODC) of the Dutch Ministry of Justice and Security. The TIP-table is a database containing all periods of detention for all offenders. The table is created with information from the Custodial Institutions Agency (DJI) and indicates who was detained for what period of time in a penitentiary institution, juvenile detention centre, or a clinic for detainees under hospital orders. For the purposes of the current study, the TIP-table was used to calculate the time spent in detention by the various offender groups.

Offender Groups

We study all individuals with at least one criminal case from 2002 until 2004,Footnote 3 so it was possible to study criminal careers over a long period of time. This is important because many delinquents become less active when they get older, although this does not necessarily hold true for all offender groups (Moffitt, 1993). The offenders were subsequently divided into four offender groups: domestic burglars, residential and commercial robbers, street robbers, and other offenders. First, we distinguished the group of other offenders from the other three offender groups. The group of other offenders only included delinquents who had never committed a domestic burglary, residential and commercial robbery, or street robbery. Conversely, the other three offender groups only included delinquents who committed at least one domestic burglary, residential and commercial robbery, or street robbery. Some delinquents committed at least two of these offences. For example, someone could have a criminal case with a domestic burglary, a criminal case with a residential and commercial robbery, and another criminal case with a street robbery. In such cases, we looked at which of these offences was most prominent throughout someone’s criminal career and assigned the delinquent to that group. For instance, a delinquent who committed more domestic burglaries than street robberies was assigned to the group of domestic burglars. Whenever someone committed two or three of these offences in equal numbers, we assigned that delinquent to the group of the most severe offence. We considered residential and commercial robbery to be more severe than street robbery, and street robbery to be more severe than domestic burglary. This classification matches the severity of legal punishments in the Netherlands and the UK, which are most severe for residential and commercial robbery, followed by street robbery, and then followed by domestic burglary (Blokdijk & Beijersbergen, 2020; Sentencing Council, 2017; Sentencing Council, 2019).

Furthermore, in order to define the four offender groups, we first had to define the offences and classify the criminal cases in OBJD accordingly. Because OBJD is based on judicial data, we followed the definitions of domestic burglary, street robbery, and residential and commercial robbery used by the Public Prosecution Service (Openbaar Ministerie, 2015). Domestic burglary is defined as ‘theft or attempt at theft, without (threatening with) violence against persons, in combination with illegal entry of a residence for instance by break or entry’ (Openbaar Ministerie, 2015). Street robbery is defined as ‘taking away with (threat of) violence or extortion any good, committed against private persons who are on public road, or the attempt thereat’ (Openbaar Ministerie, 2015). Residential and commercial robbery is defined as ‘taking away with (threat of) violence or extortion any good, committed against persons in a restricted area or building (such as a bank, shop or home), or a planned or organised cash transport, or the attempt thereat’ (Openbaar Ministerie, 2015). These three definitions are mutually exclusive. Domestic burglary, street robbery, and residential and commercial robbery are not directly recognisable in the articles of the Dutch Criminal Code. Therefore, the Public Prosecution Service uses so-called social classifications to register these types of offences (Openbaar Ministerie, 2015). These classifications are included in the OBJD data and give more information about the type of offence and specific characteristics of the offence. We used a total of sixteen social classifications to decide whether a criminal case contained a domestic burglary (e.g. ‘theft from home’ and ‘home burglary’), street robbery (e.g. ‘street robbery including purse robbery’), or residential and commercial robbery (‘business robbery’, ‘shop robbery’, ‘robbery money institutions’, and ‘residential robbery’).

Finally, after we defined the offender groups, we drew a random subsample of other offenders, as this group was very large. This was done to dramatically improve the time our models needed to compute, making our statistical analyses possible, while maintaining the representativity of our data. Ultimately this resulted in n = 69,253 other offenders, compared to n = 9480 domestic burglars, n = 3068 residential and commercial robbers, and n = 7279 street robbers (also see Table 1).

Table 1 Personal and criminal career characteristics of burglars, residential and commercial robbers, street robbers, and other offenders
Full size table

Variables

Personal Characteristics

In order to gain some insight into the demographics of the delinquents included in our data, we also provide information on some personal characteristics (see Table 1). In particular, we list the gender of the delinquents (male or female) and their country of birth (The Netherlands, Morocco, Former Netherlands Antilles, Surinam, Turkey, other countries).

Characteristics of the Criminal Career

Besides analysing the age-crime trajectories of the four offender groups, we also present some descriptive statistics to provide insight in the nature and size of their criminal careers. We looked at six characteristics of the criminal career: age of onset (the age at the time of the first criminal case), age of termination (the age at the time of the last registered criminal case up until June 2020), duration (the difference in years between the age of onset and the age of termination), time in detention (the number of days spent in detention up until June 2020), the frequency of criminal cases (the total number of criminal cases), and specialization.Footnote 4 Specialisation was measured with two indicators. Firstly, we made use of the diversity index (D). This index is based on the number of categories of criminal cases (J), the number of observed criminal cases (N), and the number of criminal cases in a category (nj). The diversity index can be defined as \(\mathrm{D}=1- \sum_{j=1}^{j}{({n}_{j}/N)}^{2}\). A diversity index of 0 indicates complete specialisation. The maximum value on the diversity index depends on the number of categories of criminal cases (J). We looked at ten categories: domestic burglary, residential and commercial robbery, street robbery, violence, sexual offence, property crime with violence (excluding residential and commercial robbery and street robbery), property crime without violence (excluding domestic burglary), drug offence, traffic offence, and finally, vandalism, minor aggression and crimes against public order. Because we looked at ten categories, the maximum value for our diversity index was (10 − 1)/10 = 0.9 (Saramago et al., 2020). Because the diversity index depends in part on the total number of criminal cases, delinquents with fewer criminal cases have a higher chance of being listed as a specialist. We corrected for this bias by adding a weight factor to the diversity index, based on the total number of criminal cases in someone’s career. Essentially, D is made larger for delinquents with few criminal cases, for instance only two, relative to delinquents who have many criminal cases, for instance ten (Francis & Humphreys, 2016).

The diversity index is useful for capturing the extent to which people specialise in certain types of crime or whether they generalise instead. However, it does not tell us anything about the type of criminal behaviour someone is specialised in. Therefore, we used a second measure of specialisation and also calculated the percentage of someone’s total number of criminal cases that fell in each of the ten categories of cases described above.

Finally, individuals who only had one criminal case throughout their criminal career were not included in these two measures of specialisation. These individuals would have otherwise been considered complete specialists. We deemed this to be incorrect, as it is difficult to consider someone a specialist after only committing one crime.

Trajectories

The group trajectory models were fitted using the frequency of criminal cases per person per year of life and age (in years).Footnote 5

Analyses

In order to answer the first research question about the criminal career characteristics of domestic burglars, street robbers, and residential and commercial robbers, we will first present some descriptive statistics about the six criminal career characteristics. Subsequently, and in order to answer the second research question about the crime trajectories, we will present results from the semiparametric group-based trajectory models (SPGM) (Nagin, 2005), based on the age of the individuals and the number of criminal cases they had at that age.

SPGM can be used to identify clusters of individuals that have similar age-crime trajectories (Bersani et al., 2009). A trajectory model is a kind of latent class model, based on maximum likelihood estimation, which divides individuals into mutually exclusive groups (Jennings & Reingle, 2012). It is important to note that SPGM is an explorative method. Choosing the number of trajectories is in part an arbitrary process (Skardhamar, 2010). As such, it is not an appropriate method to test theories about criminal behaviour. Therefore, we have refrained from formulating any strict hypotheses. That said, SPGM can be very informative in identifying and comparing different age-crime trajectories within a given sample.

We have estimated our SPGM models in Stata (15, StataCorp LLC, College Station, TX), using the traj package. The analyses were performed on four subsamples: domestic burglars, street robbers, residential and commercial robbers, and other offenders. Whenever we did not find any record of a criminal case for a given person in a given year, that person received a score of 0 for the frequency of criminal cases for that year. We made use of zero-inflated Poisson models, as there were quite some person-age combinations with no criminal cases. Because theories about age-crime curves typically describe the relationship between age and crime as curvilinear, with a strong increase in criminal activity during adolescence followed by a quick drop, we modelled quadratic relationships between age and crime. The models were fitted in sequence, slowly increasing the number of possible trajectory groups. In the first model, all delinquents were assigned to one group; in the second model, they were divided into two groups, and so on. This process was repeated for each of the four offender groups up until models with six trajectory groups. Scholars have often looked at the Bayesian Information Criterion (BIC) in order to decide what model best matches the data and what number of trajectory groups is optimal. However, this criterion has proven to be less useful when analysing data with a large sample and many time points (Blokland et al., 2005). We have therefore looked at additional selection criteria, following Nagin (2005). We only considered a model to be acceptable if the average of the posterior probabilities of group membership (APP) was at least 0.7, if the odds of correct classification (OCC) was larger than 5, if each of the trajectory groups contained at least 5% of the sample, and if there were no large disparities between the estimated probabilities of group membership and the posterior probabilities of group membership. Whenever more than one model per offender group met these selection criteria, we considered whether each additional trajectory group yielded enough new and substantively interesting information. The three authors made these judgements independently of one another, and afterwards compared and discussed their decisions. Ultimately, this process led to the selection of one model for every offender group.

Results

We first present descriptive statistics to provide insight in the nature and size of criminal careers of the four offender groups: domestic burglars, street robbers, residential and commercial robbers, and other offenders.Footnote 6 Afterwards, we present the criminal trajectories of these offender groups.

Characteristics of the Criminal Career

Table 1 presents personal and criminal career characteristics for the four offender groups. First of all, the large majority of the delinquents in our sample was male. This holds true for all offender groups. That said, there were some differences between the offender groups. The percentage of male offenders was lower for other offenders (87%) than for domestic burglars (91%), residential and commercial robbers (95%), and street robbers (94%). Second, most of the delinquents in our sample were born in the Netherlands. This too holds true for all offender groups. The percentage of Dutch natives was highest for the group of other offenders (76%), followed by domestic burglars (73%), and then by residential and commercial robbers (62%) and street robbers (61%).

With regards to the criminal career characteristics, it can first be said that the age of onset was highest for the group of other offenders (m = 24). Domestic burglars and residential and commercial robbers typically had their first criminal case when they were about 18 years old. Street robbers were younger still when they had their first criminal case (m = 17). Second, street robbers were not only youngest when they had their first criminal case, they were also youngest when they had their last case (m = 32). The (provisional) age of termination was quite similar for other offenders (m = 34), domestic burglars (m = 36), and residential and commercial robbers (m = 35). Third, domestic burglars typically have the longest criminal career (m = 17), followed by residential and commercial robbers (m = 17), street robbers (m = 15), and finally other offenders (m = 11). Most likely, other offenders had such a relatively short criminal career because, compared to the other offender groups, they had a relatively late age of onset. Fourth, other offenders spent (up to now) considerably fewer days in detention than the other offender groups. While the group of other offenders (m = 33) spent, on average, a little over a month in detention, domestic burglars (m = 384) and residential and commercial robbers (m = 389) spent around a year in detention. Street robbers (m = 309) also spent a lot more days in detention than the other offenders. Fifth, on average, domestic burglars (m = 15) had about three times as many criminal cases throughout their career as the other offenders (m = 5), and residential and commercial robbers (m = 12) and street robbers (m = 12) had about twice as many. Sixth, none of the offender groups appeared to specialise into one type of criminal case, as they all scored quite high on the diversity index. Finally, while the four offender groups did not specialise into one type of crime, they did appear to have their preferred types of crime. Property crime without violence was the most common type of offence for all criminal cases of domestic burglars (m = 37), residential and commercial robbers (m = 29), and street robbers (m = 28). Second most common was the group-specific offence, as domestic burglars had a large share of criminal cases with a domestic burglary (m = 18), residential and commercial robbers had a fair amount of cases with residential and commercial robbery as an offence (m = 17), and street robbers were involved in a high percentage of cases with street robberies (m = 15). For other offenders, violence was the offence that was present in the highest percentage of criminal cases (m = 33). The other offenders also had a large share of criminal cases due to vandalism and crimes against public order, property crimes without violence, and traffic offences. While violence was also relatively common amongst the criminal cases of domestic burglars, residential and commercial robbers and street robbers, other offenders mostly differ from these other offender groups due to the comparatively high share of criminal cases with a traffic offence (m = 16) or an offence in the category of vandalism, minor aggression, and crimes against public order (m = 21).

Criminal Trajectories

We have analysed the career trajectories of the four offender groups in order to study what trajectory groups emerge and how they might differ between offender groups. Table 2 shows which models were selected for every offender group, including the number of groups and the quantitative selection criteria (the statistics of all trajectory models can be found in Table 3 in the appendix). Based on the BIC, APP (> 0.7), OCC (> 5.0), and independent judgements of the researchers, models with 4 trajectories fit the data best for burglars and robbers, whereas a model with three trajectories was optimal for the group of other offenders.

Table 2 Statistics of the selected trajectory models for domestic burglars, residential and commercial robbers, street robbers, and other offenders
Full size table

Domestic Burglars

Figure 1a depicts the four criminal career trajectories for domestic burglars. The largest group of domestic burglars (32%) were low-frequency (LF) offenders. These delinquents had relatively few criminal cases throughout their career. Their career trajectory best resembles that of Moffitt’s additional low-level chronics (Moffitt, 2003). There were two groups of moderate frequency offenders who were most active when they were involved with around one criminal case per year. One of these groups, the moderate frequency adolescence limited (MFAL), started with criminal activities during their adolescence, was most active when they were about 20 years old, and quickly reduced their criminal behaviour soon thereafter. A little over a quarter (28%) of the domestic burglars belong to this group. The other group, the moderate frequency late peak (MFLP), was most criminally active at a later age, between 30 and 40 years, decreased the number of criminal cases at a slower rate than the MFAL offenders, and remained involved in criminal behaviour until they were almost 60 years old. This career path was characteristic of 27% of domestic burglars. The fourth career trajectory that was found amongst domestic burglars was also the smallest group: 12% of domestic burglars belonged to the group of high-frequency life-course persistent (HFLCP) offenders. These offenders started with criminal behaviour at a young age, were involved in relatively many criminal cases (an average of 1.5 per year at the peak of criminal activity), and remained criminally active until late in life, some even until they were 80 years old.

Fig. 1
figure 1

Career trajectories of domestic burglars, residential and commercial robbers, street robbers, and other offenders

Full size image

Residential and Commercial Robbers

We found four trajectory groups for residential and commercial robbers (see Fig. 1b). A little less than a quarter (24%) of them followed the career path of low-frequency offenders, who were involved with relatively few criminal cases throughout their career. A bigger portion (35%) of residential and commercial robbers belonged to the group of MFAL offenders. A little less than a third (31%) of residential and commercial robbers were characterised by the age-crime trajectory of MFLP offenders and had a peak in criminal activity around the time they were 35 years old. Furthermore, we also found a small group (10%) of HFLCP offenders amongst residential and commercial robbers.

Street Robbers

The criminal career trajectories of street robbers could be divided into four trajectories. These trajectories can be seen in Fig. 1c. In contrast to domestic burglars and residential and commercial robbers, we did not find a separate group of low-frequency offenders amongst street robbers. Instead, we found two types of moderate-frequency adolescence-limited offenders. One group of MFAL street robbers (27%) showed a peak in criminal activity when they were 20 years old, when they were involved with an average of about 0.75 criminal cases per year. The career path of this group is very similar to that of the MFAL offenders amongst domestic burglars and residential and commercial robbers. The other type of MFAL offender is unique to street robbers and stood apart because these offenders already had their peak in criminal activity when they were relatively young (< 18 years), even compared to the other group of adolescence-limited offenders. About half (48%) of street robbers followed the trajectory of this unique group of young MFAL offenders. Furthermore, we found a comparatively small group (13%) of MFLP offenders amongst street robbers, who had a peak in the number of criminal cases when they were between 35 and 40 years old. Finally, again, we found a small group (12%) of HFLCP offenders amongst street robbers.

Other Offenders

For the population of other offenders, we found three trajectory groups. These are depicted in Fig. 1d. By far the largest group (67%) of other offenders could be characterised as low-frequency offenders. This group is a lot bigger than the LF offenders we found amongst domestic burglars, residential and commercial robbers, and street robbers. The other two trajectory groups that were found amongst other offenders both described career paths of moderate frequency offenders. Twenty percent of the other offenders could be considered MFAL offenders, and 13% were MFLP offenders. It is important to note, however, that the peak in criminal activity of these two groups of MF offenders is considerably lower for other offenders than for burglars and robbers, with an average of 0.5 criminal case per year versus an average of one criminal case per year. Finally, we did not find a group of HFLCP offenders amongst other offenders.

Discussion

The first aim of this paper was to investigate the criminal career characteristics of domestic burglars, residential and commercial robbers, and street robbers in the Netherlands. Compared to other offenders, domestic burglars, residential and commercial robbers, and street robbers start with criminal behaviour at a younger age, commit more crimes throughout their careers, remain criminally active until later in life, and spend more time in detention. These findings are in line with prior research that suggests that domestic burglars, residential and commercial robbers, and street robbers are active criminals, who belong to offender groups that reoffend the most and are responsible for a substantial share of criminal offences in Western countries (e.g. Bernasco, 2009; Durose et al., 2014; Jehle, 2014; Kros & Beijersbergen, 2021; Maguire et al., 2010; Ministry of Justice, 2014; Owen & Cooper, 2013; Rovers et al., 2010; Svensson, 2002). In that light, it seems justified that in the last 20 years, the Dutch law enforcement agencies made extra efforts to combat these offences, commonly grouped together under the name of ‘high-impact crimes’.

While the criminal careers of domestic burglars, residential and commercial robbers, and street robbers appear to be markedly different from that of other offenders, our results also show notable differences between the three groups of offenders. For one, compared to domestic burglars and residential and commercial robbers, street robbers are youngest at the time of their first and their last criminal case, generally have the shortest career, and spend fewest days in detention. Domestic burglars, on the other hand, are oldest at the time of their last criminal case, have the longest careers, are involved with the most criminal cases, specialise the most, and—together with residential and commercial robbers—spend the most days in detention. Residential and commercial robbers tend to fall somewhere in between domestic burglars and street robbers, for nearly all career characteristics.

The second aim of this paper was to explore whether the criminal career trajectories of domestic burglars, residential and commercial robbers, and street robbers differ from the criminal career trajectories of other offenders. By and large, we found four main career trajectories: low-frequency offenders, moderate-frequency adolescence-limited offenders, moderate-frequency late-peak offenders, and high-frequency life-course-persistent offenders. These groups can be distinguished based on two defining characteristics: the shape of the age-crime curve, and the number of criminal cases delinquents had during their most active years. For low-frequency offenders, the average annual criminal cases peaked below 0.25; for moderate-frequency offenders, this average peaked between 0.3 and 1.0; and for high-frequency offenders, the peak was around 1.5 criminal cases per annum. The fact that we found distinct age-crime curves amongst offenders is incongruent with the general theory of crime (Gottfredson & Hirschi, 1990), but is in line with developmental and life-course criminology theories such as the dual taxonomy model (Moffitt, 1993) and the age-graded theory of informal social control (Sampson & Laub, 1993), which assume variation in age-crime curves amongst offenders. Importantly, we did consistently find a group of life-course-persistent offenders, who commit crimes up until the age of 70. This does not support Sampson and Laub’s (2003) idea of a general process of desistance from crime, where all offenders desist in the middle adult years. Instead, the relatively large share of adolescence-limited, life-course persistent, and low-level chronic offenders fits particularly well with Moffitt’s (2003) taxonomy.

We have further shown that some age-crime curves appear to be distinctive of specific offender groups. As such, our first main contribution to the literature on criminal careers is to show that life-course persistent career trajectories may be unique to specific offender groups, like burglars and robbers. We only found high-frequency life-course persistent offenders amongst domestic burglars, residential and commercial robbers, and street robbers, but not amongst other offenders. Furthermore, adolescence-limited (moderate frequency) offenders were found amongst all four offender types. Taken together, these results suggest that one who is interested in identifying future life-course persistent offenders should pay particular attention to delinquents who commit residential and commercial robberies, street robberies, and, in particular, domestic burglaries. This conclusion has important policy ramifications, as life-course persistent offenders who commit a large share of relatively serious criminal offences exert tremendous costs on society (Cohen & Piquero, 2009). Being able to identify life-course persistent offenders early and accurately may improve intervention and prevention efforts and could ultimately help curb their criminal activities. We have shown that domestic burglars, residential and commercial robbers, and street robbers are more likely to be high-frequency life-course persistent offenders than other offenders. This holds particularly true for domestic burglars, as this offender group has the largest share of high-frequency life-course persistent offenders. Street robbers also stand out in this regard, as they have a fairly large share of life-course persistent offenders (12%), but also because we did not find a group of low-frequency offenders amongst them. Additionally, the share of low-frequency offenders is considerably smaller amongst domestic burglars and residential and commercial robbers than amongst other offenders. Life-course persistent offenders were unique to burglars and robbers, whereas adolescence-limited offenders were common in all our offender groups, including general offenders. Our results thus confirm the idea that LCP offenders tend to be serious offenders (Moffitt, 1993) and stress the importance of studying the career trajectories of specific offender groups (also see e.g. Van Koppen et al., 2010).

A second main contribution to the literature is that we consistently found a group of late-peak offenders, next to a group of LCP offenders. These offenders start their criminal career at a young age, are as active as adolescence-limited offenders, and continue to increase their criminal activities after they have transitioned into adulthood. The fact that these offenders are involved with most criminal cases per year in their thirties or even forties seems to defy traditional theories about age-crime curves, including the general theory of crime (Gottfredson & Hirschi, 1990), Moffitt’s (1993) offender taxonomy, and the age-graded theory of social control (Sampson & Laub, 1993). All these theories posit, for one reason or another, that active criminals should start their criminal careers early and peak in offending during adolescence and young adulthood (Liu et al., 2022). Yet we find that desistance does not necessarily start in early adulthood. The group of late-peak offenders also appears to be different from the recently more often studied late-onset offenders (Carlsson & Sivertsson, 2021; Van Koppen, 2018), as the late-peak offenders we found do tend to engage in criminal behaviour from a young age. Some of them have their first criminal case at the age of twelve. If anything, the moderate frequency late-peak offenders we found show most commonalities with ‘late bloomers’. These offenders engage in limited offending during adolescence, only to become involved in serious and escalating criminal behaviour later in life (Thornberry & Matsuda, 2011). Krohn et al. (2013) argue that these late bloomers resemble life-course persistent offenders in anti-social personality traits that make them prone to criminal behaviour, yet differ from them in terms of early-life protective factors, such as supportive family bonds and more economic resources. Yet these social and economic advantages are not always present later in life. For instance, informal social control from parents may become less apparent and latent criminogenic traits may surface, resulting in a late-peak in criminal activity (Liu et al., 2022).

The fact that we consistently found a group of moderate-frequency late-peak offenders could also be a result of the longer timespan in our data. Perhaps, these offenders are would-be adolescence-limited offenders who actually did not desist after adolescence but remained criminally active well into adulthood, but at lower rates than life-course persistent offenders. In line with this idea, Farrington (2019) recently found that reconvictions remained likely even after a gap of 15 years since the previous conviction. Our late-peak offenders could have been easily classified as adolescence-limited offenders if we could not have tracked their presumed desistance for a relatively long period of time. Other studies could investigate whether this group of late-peak offenders is actually more common than currently presumed, by also adopting a long-term perspective with data that tracks adolescent-limited offenders well into their adulthood.

Although this study contains a wealth of valuable information about the long-term criminal careers of domestic burglars, street robbers, and residential and commercial robbers in the Netherlands, it also knows a few limitations. First, this study used register data from the OBJD. As a consequence, we only studied criminal activities that were brought to the attention of the Public Prosecution Service. Offences and offenders that remained under the radar were not taken into account. Second, we studied criminal behaviour at the level of criminal cases instead of the separate offences within these cases. While this approach is not new in research on criminal careers per se (see for example Robert et al., 2018), it is different from similar studies that looked at criminal offences, and this should be taken into consideration when interpreting our results. One consequence of analysing criminal cases instead of criminal offences is that we likely underestimated the total frequency of criminal behaviour even more than studies that look at registered criminal offences. This is because a criminal case may contain more than one criminal offence. This is particularly relevant for the comparison between other offenders and the other three groups of offenders, as the criminal cases of the other offenders typically contain fewer offences (on average 1.4 offence per criminal case and 1.8 offence per criminal case, respectively). Consequentially, the differences we find in terms of criminal activity between other offenders on the one hand and domestic burglars, residential and commercial robbers, and street robbers on the other hand may even be a slight underestimation. That said, the number of offences per criminal cases does not vary over time. Therefore, our longitudinal analyses are not affected by the decision to study criminal cases instead of criminal offences. The internal validity of our results is further guaranteed by relying solely on information about criminal cases recorded in OBJD. Third, our reference group of other offenders is rather heterogenous and includes people who committed minor public order offences as well as people who committed violent offences. Analysing a more homogenous group of offenders could have resulted in a cleaner comparison with burglars and robbers. However, the fact that we only found LCP offenders amongst burglars and robbers may also be all the more striking when considering that the reference group included people who committed more serious crimes. Still, we would encourage other researchers to look at the career trajectories of more detailed and nuanced categorizations within the group of other offenders. Especially since our results highlight the importance of studying the career trajectories of specific offender groups. A fourth limitation is that our data did not include additional information about criminogenic factors. As a result, we could not investigate whether other offenders, domestic burglars, residential and commercial robbers, and street robbers differ from each other in factors that may explain people’s involvement in criminal activities, such as psychosocial problems, substance abuse, socio-economic status, family situations, or criminal networks. This lack of additional criminogenic information also meant that we could not test whether differences in criminal career trajectories between offender groups could in part be explained by specific life-course events, such as marriage or stable employment, nor whether these events affect men and women differently (Giordano et al., 2002; Gunnison, 2014). It could for instance be that our subset of serious offenders—burglars and robbers—are less likely to find a stable job or get married, or that their involvement in crime is impacted less by these forms of social control (Sampson & Laub, 1993). Future research could enrich register data from OBJD with such criminogenic information to gain a better understanding of why delinquents become involved in particular types of offences, such as domestic burglaries.

Notwithstanding these limitations, the current study provides important contributions to the literature on criminal careers in general and the careers of domestic burglars, residential and commercial robbers, and street robbers in particular. We adopted a life-course perspective and used semiparametric group trajectory models to analyse longitudinal data on all criminal cases from 1997 until 2020 for all people in the Netherlands of 12 years or older. This data allowed us to study criminal careers over a longer period of time, making it possible to more accurately capture whether delinquents actually stopped committing crimes altogether or simply took a break from criminal behaviour. All in all, our findings suggest that in order to predict who will follow the career path of a life-course persistent offender, it is important to distinguish between specific groups of offenders. Life-course persistent offenders were found amongst domestic burglars, residential and commercial robbers, and street robbers, but not amongst offenders of other types of crime. Furthermore, the size of the group of life-course persistent offenders varied between the domestic burglars, residential and commercial robbers, and street robbers and is largest for domestic burglars. These findings are not only important for scholars who are interested in life-course criminology, but also for law enforcement agencies who are trying to identify life-course persistent offenders early and accurately.