“Ex-Imprisoned Homicide Offenders: Once Bitten,Twice Shy?” The Effect of the Length of Imprisonment on Recidivism for Homicide Offenders

Abstract

This study aims to examine recidivism patterns and the influence of imprisonment length for all homicide offenders who have been convicted in the Netherlands between 1996 and 2004. In addition, we tested whether imprisonment effects differed between homicide offenders with different characteristics. Analyses on 621 homicide offenders indicate that longer imprisonment systematically increases recidivism frequency, not recidivism speed. We find some indications that imprisonment length increases recidivism to a greater extent for offenders with an intimate partner, with a Western ethnic background and for offenders with a relatively shorter detention history prior to the homicide.

Keywords

nonviolent recidivism violent recidivism length of imprisonment homicide offenders The Netherlands

Introduction

The extreme nature of homicide makes us curious about the offender’s (criminal) future and whether institutional intervention affects this future. A great deal of public and scientific interest in homicide offenders, however, largely focuses on criminal history rather than on criminal recidivism of homicide offenders. Hence, so far only limited longitudinal research has shed light on offending patterns, typologies and recidivism patterns of different types of homicide offenders (Delisi & Scheerer, 2006; Roberts, Zgoba, & Shahidullah, 2007).

So far, few studies have assessed the effects of imprisonment on recidivism for homicide offenders. Previous studies assessing the influence of imprisonment mainly relied on general delinquents (Smith, Goggin, & Gendreau, 2002) and those who have been punished with relatively short prison sentences (Villettaz, Killias, & Zoder, 2006). The paucity of research on homicide offenders is especially relevant given the long prison terms these individuals serve. One of the significant challenges for release from prison is the lack of predictability of recidivism of this special group (Roberts et al., 2007).

To overcome these shortcomings, we investigate to what extent the effect of length of imprisonment affects criminal recidivism for Dutch homicide offenders. In addition, we aim to differentiate between several groups of homicide offenders in order to further understand the (lack of) homogeneity of imprisonment effects.

Theoretical Background and Hypotheses

From research on criminal offenders, we know that several mechanisms are at work when the effect of the length of imprisonment on recidivism is considered (Nagin, Cullen, & Jonson, 2009; Snodgrass, Blokland, Haviland, Nieuwbeerta, & Nagin, 2011). We will discuss two commonly used theories, based on which we will derive hypotheses on the general effect of length of imprisonment as well as on the effect of length of imprisonment for specific groups of homicide offenders.

First, social control theory offers various reasons why social control is a key factor in preventing crime (Hirschi, 1969). When someone is socially tied to others, he will live up to the norms of these others. Hence, attachment to people who share norms of society that condemn committing crimes leads to the internalization of these norms, which decreases the likelihood of committing a crime. After all, people who are attached to others do not want to disappoint those around them and criminal behavior would lead the individual feeling guilt and shame. These bonds are social capital that the offender risks losing if they engage in this behavior. Moreover, contacts with others also reflect the commitment to society. This commitment, and the potential to lose it, can consist of social bonds (e.g., marriage) but one can also think of promising educational and occupational careers that are at stake (DeJong, 1997). Furthermore, involvement in society can cause conformity to norms against committing crimes because there is less time that one can allocate to deviant behavior. Apart from the presence or absence of these forms of social control, its influence on preventing crimes also depends on the strength and quality of the relationships (Gottfredson & Hirschi, 1990; Sampson & Laub, 1990). In sum, when people are more committed to society, the costs of losing this commitment become larger, so they are less likely to break the law.

As to the effect of a prison sentence, Sampson and Laub (1993, 2003) hold that the quality of contacts might suffer from imprisonment. Both the reduced opportunities to keep in contact while someone is imprisoned, and the fact that friends and relatives might not want to keep contact with someone who has committed a crime, can worsen prisoners’ contact with conventional others. Besides someone’s relations, employment chances are deteriorated because of imprisonment (Kling, 2006; Pager, 2003; Waldfogel, 1994). When these forms of social control are reduced, offenders have a smaller incentive to abstain from reoffending. After all, the attachment and commitment that are at stake when committing a crime have already been lost due to imprisonment. This reduction of social control is expected to increase recidivism chances. In summary, social control theory predicts that longer prison sentences increase the likelihood of recidivism because of the deterioration of offenders’ commitment to society, in terms of relations and occupational possibilities.

Second, social learning theorists have referred to prisons as “schools of crime” (Gendreau, Cullen, & Goggin, 1999; Sampson & Laub, 1993; Wermink, Blokland, Nieuwbeerta, Nagin, & Tollenaar, 2010) According to this perspective, having criminal friends makes people more likely to develop norms that favor crime and to be reinforced for this behavior, which, consequently, makes people more likely to actually commit a crime. Moreover, contact with these “negative social bonds” leads to acquiring skills that facilitate (future) crime (Gendreau et al., 1999; Sampson & Laub, 1993; Wermink et al., 2010). As for the effect of imprisonment, it is notable that prisoners are surrounded by other convicts. Depending on whether imprisonment regulations facilitate contact, interaction between convicts takes place, which may lead to the exchange of crime-related information. As such, social learning theory predicts a longer prison sentence to increase the likelihood of recidivism because during a longer prison sentence criminal knowledge and skills are acquired and norms that favor criminal behavior are internalized over a longer period of time.

In summary, in line with social control theory and social learning theory, we hypothesize that length of imprisonment increases recidivism (Hypothesis 1).

Apart from testing the direct effect of imprisonment length on recidivism, we propose interaction effects that more closely assess the mechanisms of social control and social learning theory. First, we use social control theory to hypothesize that the effect of imprisonment length depends on partner status and ethnicity. Second, we apply social learning theory to predict the effect of imprisonment length to vary between people with different detention histories.

As outlined stated above, social control theory assumes imprisonment to damage conventional social ties, in turn influencing criminal behavior. More specifically, the longer offenders are imprisoned, the lower offenders’ quality of conventional relationships and hence the more likely offenders are to recidivate.¹ We examine this assumption by focusing on two groups who are likely to have relatively high levels of social control: those with an intimate partner and those of Western origin. Offenders with an intimate partner are more likely to be married and to have children than singles (Statistics Netherlands, 2009). Moreover, research shows that ethnicity is a proxy for social and economic segregation (Cale, Plecas, Cohen, & Fortier, 2010; Eitle, D’Alessio, & Stolzenberg, 2006; Kubrin, 2003). Also in Europe, previous research has shown that those with a non-Western ethnic background generally have a disadvantaged position with regard to labor force outcomes (e.g., Statistics Netherlands, 2009; Van Tubergen, Maas, & Flap, 2004).²

If social control theory applies, offenders experiencing more social control (i.e., those with an intimate partner and those with a Western origin) are less likely to recidivate. More important to the current study, social control theory implies that offenders experiencing higher levels of social control before imprisonment experience a greater loss of social control due to imprisonment, which makes imprisonment length to increase recidivism chances more for these offenders. We therefore hypothesize that length of imprisonment increases recidivism to a greater extent for offenders with an intimate partner compared to offenders without an intimate partner (Hypothesis 2a) and that length of imprisonment increases recidivism to a greater extent for Dutch offenders and Western immigrants compared to non-Western immigrants (Hypothesis 2b).

As to social learning theory, it is expected that length of imprisonment increases knowledge of and positive attitudes towards criminal behavior, which increases the likelihood of recidivism. This reasoning is examined by comparing offenders who might experience imprisonment differently. We expect that the learning effect is weaker for offenders with a longer criminal history. After all, when experienced offenders go to prison, they cannot learn as much from their inmates and they most likely already have positive attitudes towards crime (Polaschek, Collie, & Walkey, 2004; Walters, 2003).

In the contexts of social learning theory offenders with more criminal experience are more likely to recidivate. More important to the current study is the expectation that length of imprisonment increases recidivism to a smaller extent for offenders who have more prison experience. Using the offender’s detention history as an indicator of the criminal experience, we hypothesize that the shorter someone’s detention history is, the more length of imprisonment increases recidivism (Hypothesis 3).

Previous Studies

Previous studies on the effect of imprisonment length on recidivism have mainly focused on criminals in general. We discuss some key references in this area before elaborating on the literature on homicide offenders in particular. In an early narrative review, Paulus and Dzindolet (1992) found no support for the deterrent effect of imprisonment. Evidence for a crime-enhancing effect of imprisonment was scarce as well, as only one study reported a positive relation between imprisonment length and recidivism (Farrington, 1977). A meta-analysis by Smith et al. (2002) similarly found little support for the social learning hypothesis, as length of imprisonment and recidivism were weakly positively related both for high risk (ϕ = .03) and low risk (ϕ = .04) offenders. The deterrence hypothesis was not supported as an incarceration versus noncustodial comparison showed similar recidivism rates. A comparison of custodial versus noncustodial sentences for randomized controlled trials by Villettaz et al. (2006) likewise showed that reoffending patterns did not differ. Lastly, while applying a dose-response model to more rigorously control for selection effects, Snodgrass et al. (2011) present evidence that imprisonment length has a small positive effect on imprisonment (cf. Nagin et al., 2009). That is, after matching offenders on a range of indicators, longer imprisonment increases the subsequent period of confinement in three years following release, but not the post imprisonment felony conviction rate. Notably, Villettaz et al. (2006) only include custodial sentences up to a few months, and Snodgrass et al. (2011) use more than one year of imprisonment as the highest category, which raises the question whether long-term custodial sentences have different effects.

To narrow our focus down to homicide offenders, we searched through electronic databases using combinations and variations of the keywords “homicide offenders,” “length of imprisonment,” “incarceration,” and “recidivism.” Studies were included if information on length of imprisonment and on a certain measurement of recidivism was given, and if the focus of the study was on homicide offenders. Seven studies were detected, of which the study characteristics can be found in Table 1.

Table 1.

Overview of Prior Studies on the Effect of Length of Imprisonment on Recidivism for Homicide Offenders

Authors (Year)	Country	N	Type of homicide offenders	Years of follow-up	Measurement imprisonment	Measurement Recidivism	Recidivism (%)	Effect imprisonment length on recidivism
Gottlieb and Gabrielsen (1990)	Denmark	52	Homicide committed	9	Length of prison stay	Post release crime/violence: yes/no	54	Increase
			1. Within family
			2. Outside family
Hagan (1997)	USA	20	No distinction	5-15	Not reported	Reoffense: yes/no	60	Decrease
Heide (1998)	USA	59	Juveniles, no distinction	Not reported	Qualitative interviews, no measurements	Qualitative interviews, no measurements	60	Decrease
Heide et al. (2001)	USA	59	Juveniles, no distinction	10 (1-16)	Being longer at risk indicates a shorter sentence	“Failed”	60	Increase
Loeber et al. (2005)	USA	11	No distinction	10	Not reported	Violence trajectories	82	No effect
Roberts et al. (2007)	USA	336	1. precipitated by general argument	> 5	Time served	Time to recidivism	51	Increase
			2. during felony
			3. domestic violence related
			4. after accident
Hill et al. (2008)	Germany	139	Sexual homicide offenders, no distinction	20	Duration of detention (0-14 years vs. >14 years)	Recidivism rate	Violent 29	Decrease
							Nonviolent 48

Based on these seven studies, evidence proves inconclusive whether or not there is a preventive effect of imprisonment on recidivism for homicide offenders. Three out of seven detected studies indicated that imprisonment length increases recidivism (see Table 1: Gottlieb & Gabrielsen, 1990; Heide, 1998; Roberts et al., 2007), while three others observed a decrease in the recidivism rate when the imprisonment length is longer (Hagan, 1997; Heide, Spencer, Thompson, & Solomon, 2001; Hill, Habermann, Klusmann, Berner, & Briken, 2008). One study found no effect of length of imprisonment, as imprisonment length was not related to the seriousness of the violence trajectory (i.e., low, early desist, late desist, chronic; Loeber, Lacourse, & Homish, 2005). Hence, previous research has not clearly confirmed or rejected Hypothesis 1.

The aforementioned studies on the effect of the length of imprisonment on recidivism for homicide offenders have been limited in several respects. First, some studies only provide descriptive evidence of the relationship between imprisonment and recidivism (e.g., Hagan, 1997; Loeber et al., 2005). Second, most studies perform their analyses with considerably small samples (Hagan, 1997; Loeber et al., 2005), not allowing for controlling for selection effects or testing the effect of imprisonment on different types of homicide offenders. Hence, Hypotheses 2 and 3 have not been tested before.

This Study

This study aims to improve upon previous research in a number of ways. First, we use the large scale Dutch Homicide Monitor data set (Nieuwbeerta & Leistra, 2007) that contains information on approximately 1,400 homicide offenders who committed a homicide in the Netherlands between 1996 and 2004. This enables us to statistically model the effect of length of imprisonment—with a considerable follow-up period until 2008—and on different recidivism indicators. Furthermore, the rich data set provides us with a precise measurement of the length of imprisonment and includes control variables to diminish a potential selection effect. Last, we challenge the assumption that all offenders experience the same effect of imprisonment and distinguish between different types of homicide offenders to test whether these differ with regard to the effect of length of imprisonment.

Data and Method

Data

To test our hypotheses, we use population data that consist of four databases. First, we utilize the Dutch Homicide Monitor, which includes all registered homicide offences (i.e., cases of murder or manslaughter) that have been committed in The Netherlands since 1992 (Nieuwbeerta & Leistra, 2007). This data set includes characteristics of the homicide event, the offender and the victim. Second, we include details on imposed sentence and prosecution based on information from the Public Prosecution Service. Third, we use TULP³ data that allows us to precisely measure the length of imprisonment by looking at the exact date of incarceration and release, rather than the sentence as imposed by a judge. Fourth, to examine the extent to which offenders recidivate and to control for their criminal history, the offenders’ criminal records until June 2008 were obtained from the Dutch Offenders Index (DOI; Onderzoeks-en Beleidsdatabase Justitiële Documentatie [OBJD]), a database that includes criminal antecedents for all Dutch citizens.

Sample Selection

Figure 1 portrays the steps we took in the selection procedure. First, we selected all homicides between 1996 and 2004, as the TULP, which administers the date of imprisonment and release, began in 1996. Furthermore, information about homicide offences after 2004 had not yet been digitized.

Figure 1.

Flowchart of sample selection procedure

In the selected time period, 2389 individuals were suspected of having committed a homicide. Due to the pronounced influence of mental disorder on the homicide, we excluded those who received a prison sentence in combination with TBS measure, a mandatory treatment measure for mentally disordered offenders that follows after a prison sentence (Van Marle, 2002; see Figure 1). These and other exclusion criteria resulted in a final sample size of 621 homicide offenders. This includes the entire population of homicide offenders in The Netherlands who committed a homicide between 1996 and 2004, who received a prison sentence, and who were released before June 2008.

Dependent Variables

We did not measure specific recidivism (i.e., committing another homicide) as the likelihood of that event occurring is low. Rather, we examined violent recidivism (including sexual offences, threat, homicide, assault, violent theft and extortion) as well as nonviolent recidivism as this has a greater possibility of occurring and still has detrimental effects on society. Although information on all offences that have led to judicial action is available, we only include those criminal law offences that are followed by a conviction or a prosecutorial disposition due to policy reasons, which excludes noncriminal law offences (e.g., traffic and economic offences) and cases that resulted in acquittal or a prosecutorial disposition because of technicalities (Blokland, Nagin, & Nieuwbeerta, 2005).

Both nonviolent and violent recidivism are operationalized in two measures. First, we examine recidivism speed by considering the number of months after prison release until someone recidivates. Second, we examine the recidivism frequency, operationalized into the number of offences committed after release when controlling for the number of years at risk (Loughran et al., 2009).

Independent Variables

Length of imprisonment is measured in days between the date of imprisonment and the date of release, capturing the actually executed prison sentence instead of the sentence someone was convicted to. For interpretation purposes, we have divided this by 365, to include the length of imprisonment in years.

In order to examine whether imprisonment effects differ between groups of offenders, we first consider whether the offender, before committing the homicide, had an intimate partner (i.e., married and/or cohabitating) or not (i.e., single, widowed, divorced).⁴ People who killed their (ex-)partner are considered not to have an intimate partner, because the mechanisms that we examine theorize about intimate partner effects that especially apply after release. Western immigrants are defined as born in Europe, North America, Israel, Australia or New Zealand (see also Note 2). The offender’s conviction history is measured in four categories (0 crimes, 1-5 crimes, 6-10 crimes, > 10 crimes). Detention history is measured in years of imprisonment before the homicide was committed.⁵

Control Variables

An important challenge in research on sentencing is the occurrence of a selection effect. When background characteristics affect both the sentence length as well as recidivism, spurious relationships between the treatment and the outcome might be interpreted as causal relations. We therefore include control variables that have shown to influence both the sentence and the recidivism pattern (Johnson, Van Wingerden, & Nieuwbeerta, 2010).

The type of homicide was operationalized into “criminal,”⁶ “robbery,” “family sphere,” “arguments” (not with family) and “else and unknown” (Smit, Bijleveld, & Van Der Zee, 2001). Moreover, we include whether an offender was convicted for murder or manslaughter. In addition, we control for the offender’s conviction history and country of birth (The Netherlands, Dutch Antilles, Suriname, Turkey, Morocco—being the four largest minority groups in The Netherlands (Statistics Netherlands, 2009), remaining Western countries, and non-Western countries). In analysing nonviolent recidivism, we also control for the offender’s gender. In analysing violent recidivism, we did not include the offender’s gender because women did not recidivate violently. Lastly, in order to not confuse an imprisonment effect with an aging effect, we control for age at time of homicide.⁷

In the analyses on frequency of recidivism we first controlled for the offender’s exposure time, computed as the number of months between an offender’s release from prison and the end of our follow-up, from which we subtract the number of months that an offender spent in prison due to crimes after release (Piquero et al., 2001).⁸ This measure simultaneously accounts for the notion that those who have been released more recently (i.e., those with a shorter exposure time) are more likely to have a high relative recidivism frequency as offenders are more likely to recidivate shortly after release (DeJong, 1997; Heide et al., 2001; Hill et al., 2008; Spohn & Holleran, 2002). Second, we control for the total length of imprisonment during follow-up, as our expectation is that the length of imprisonment affects the recidivism rate.⁹ For interpretation purposes, we have included these measures in years.

The descriptive statistics of the variables used in the analyses are displayed in Table 2.

Table 2.

Descriptive Statistics of Independent and Control Variables

	N	Range	Mean	Std. Deviation
Length of imprisonment for homicide (years)	621	.02-11.60	4.25	2.15
Exposure time (years)	621	.00-11.44	4.17	2.65
Length of imprisonment before homicide (years)^a	547	0-13.20	.51	1.44
Length of imprisonment during follow-up (years)	621	0-8.91	.23	.76
Age at time of homicide (years)^b	621	17-71	31.95	10.33
	N	Range	Percentage
Country of birth
The Netherlands	317	0/1	51.05
Dutch Antilles	62	0/1	9.98
Suriname	56	0/1	9.02
Turkey	56	0/1	9.02
Morocco	37	0/1	5.96
Western	31	0/1	4.99
Non-Western	62	0/1	9.98
Conviction history
0 crimes	205	0/1	33.01
1-5 crimes	167	0/1	26.89
6-10 crimes	81	0/1	13.04
> 10 crimes	168	0/1	27.05
Type of homicide
Criminal	118	0/1	19.00
Robbery	50	0/1	8.05
Family sphere	198	0/1	31.88
Arguments	168	0/1	27.05
Unknown and else	87	0/1	14.01
Prosecuted for murder	453	0/1	72.95
Woman	37	0/1	5.96
Having a partner before homicide^a	152	0/1	29.06

Note: All dichotomous variables are coded such that “1” refers to the variable name and “0” to the rest. For example, offenders with a “1”on the variable “having a partner before homicide” had a partner before the homicide, whereas offenders with a “0” did not have a partner before the homicide.

Partner and detention history have 98 and 74 missings, respectively.

Age at time of homicide is grand-mean centered in the analyses.

Analytic Strategy

For both violent and nonviolent recidivism, we use Cox regression survival analysis to analyze recidivism speed and Poisson regression analysis for the recidivism frequency. In Poisson regression models it often occurs that the model is overdispersed. This means that the variance of the dependent variable is larger than the mean. To correct for this, a weight factor is included (Gardner, Mulvey, & Shaw, 1995).¹⁰

We include those offenders who have been released at least one day before the observation period ends. This means we could have follow-up data of only one day for some offenders. Both analytical models can account for the offenders’ varying follow-up period. Survival analysis considers the period in which an offender is “at risk,” after which someone either recidivated or is censored (i.e., regarded as someone who has not recidivated during the follow-up period; Hosmer, Lemeshow, & May, 2008). Poisson regression analysis uses the exposure time to compute the recidivism frequency relative to the period “at risk” (Agresti, 1996).¹¹

Results

Descriptives

The general recidivism rate for the 621 homicide offenders under study was 51%. 36% (N = 226) recidivated nonviolently; 16% (N = 100) recidivated violently. However, as not every offender has been followed for the same period, we examine the indicators of recidivism by comparing offenders with different periods of imprisonment.

Table 3 shows the number of offenders who have been followed for different periods, split by their imprisonment length. Note that offenders can appear multiple times in this table: of the 106 offenders with a sentence shorter than 2 years who were followed 1 year, 99 were also followed 3 years. Table 4 indicates the percentage of offenders who have recidivated nonviolently and violently. For nonviolent recidivism, more than 30% have recidivated after 3 years. Near the end of the follow-up, after 9 years, more than 50% have nonviolently recidivated. Most notable, imprisonment length does not seem to relate to recidivism rates. For violent recidivism, approximately 10% to 15% have recidivated after 3 years. After 9 years, around 25% have recidivated. Again, imprisonment length is not systematically related to recidivism rates (note that the higher recidivism rates for 6 or more years of imprisonment are based on small numbers of offenders).

Table 3.

Number of Observed Offenders After Different Follow-Up Periods by Sentence Length

	< 2 years of imprisonment	2-4 years of imprisonment	4-6 years of imprisonment	6-8 years of imprisonment	> 8 years of imprisonment
1 year follow-up	106	191	170	62	16
3 years follow-up	99	147	128	24	2
5 years follow-up	81	104	65	1	1
7 years follow-up	60	48	14
9 years follow-up	35	3

Table 4.

Nonviolent/Violent Recidivism Rates After Different Follow-Up Periods by Sentence Length

	< 2 years of imprisonment	2-4 years of imprisonment	4-6 years of imprisonment	6-8 years of imprisonment	> 8 years of imprisonment
1 year follow-up	19.8/5.7	18.3/6.3	21.2/5.9	29.0/3.2	6.3/18.8
3 years follow-up	32.3/11.1	34.7/14.3	37.5/14.1	33.3/25.0	50.0/50.0
5 years follow-up	43.2/17.3	38.5/19.2	47.7/20.0	100.0/100.0*	0.0/0.0*
7 years follow-up	48.3/21.7	50.0/20.8	71.4/21.4
9 years follow-up	57.1/25.7	0.0/0.0*

Note: Numbers with an asterisk should be interpreted cautiously, as there are based on fewer than five cases.

Table 5 shows that the mean recidivism frequency and length of imprisonment are seemingly not systematically related. However, the maximum frequency per year is higher among offenders who have been imprisoned for more than 4 years than those imprisoned less than 4 years (respective maximum numbers are 8 vs. 15). For violent recidivism the same pattern applies, with comparable mean violent recidivism frequencies and a higher maximum violent frequency per year for the long-term inmates (respective maximum numbers are 3 vs. 4).

Table 5.

Mean Number of Recidivism Offences Per Year (i.e., Corrected for Follow-Up Period and Incapacitation) by Sentence Length

	< 2 years of imprisonment	2-4 years of imprisonment	4-6 years of imprisonment	6-8 years of imprisonment	> 8 years of imprisonment
Nonviolent offences per year	.45	.49	.57	.54	.24
Violent offences per year	.11	.19	.21	.08	.17

So, based on bivariate analyses we have to conclude that differences with regard to imprisonment length are barely related to the differences in recidivism between individuals. It is, however, possible that potential selection effects are at work, and that whether people were imprisoned for less or more than four years provides insufficient information regarding the length of imprisonment. To investigate this, we continue by performing multivariate analyses that include covariates accounting for individual differences. Moreover, by including interaction effects in our models, we test more precise hypotheses on the effects of length of imprisonment.

Multivariate Analyses

Table 6 shows the results for violent and nonviolent recidivism speed and frequency. For the main effect of imprisonment length, we find that recidivism frequency increases with longer imprisonment. Recidivism speed however is not affected by length of imprisonment. Our results thus show some support for Hypothesis 1, but only with regard to recidivism frequency not for recidivism speed.¹²

Table 6.

Cox regression Survival Analysis on Recidivism Speed and Poisson Regression Analysis on Recidivism Frequency: Main Effect (N = 621)

	Nonviolent speed		Nonviolent frequency		Violent speed		Violent frequency
	B	SE (B)	B	SE (B)	B	SE (B)	B	SE (B)
Length of imprisonment	−.032	.033	.096**	.028	.042	.051	.178***	.035
Woman	−.189	.381	.443	.330
Age at time of homicide^a	−.029***	.008	−.049***	.008	−.035**	.013	−.040***	.010
Born in The Netherlands
Dutch Antilles	.242	.208	.066	.163	.381	.277	.445*	.182
Suriname	.023	.245	−.204	.256	−.544	.446	−.717	.382
Turkey	−.198	.302	−.485	.310	−.292	.444	.572*	.264
Morocco	−.291	.401	.035	.318	−1.867	1.024	−.748	.542
Western	.617*	.286	.142	.274	.014	.487	.129	.353
Non-Western	−1.200*	.517	−1.547*	.628	−1.808	1.024	−.730	.609
No conviction history
1-5 crimes	1.514***	.256	.992***	.242	.651	.381	2.011***	.328
6-10 crimes	1.427***	.290	.796**	.285	1.091**	.402	1.973***	.337
> 10 crimes	2.169***	.262	2.022***	.230	1.507***	.363	2.331***	.340
Criminal homicide
Robbery	.151	.289	.745***	.208	.034	.417	.371	.259
Family sphere	.071	.215	−.025	.201	−.409	.339	−.550*	.257
Arguments	.030	.208	.146	.181	−.029	.297	−.322	.216
Unknown/else	.724*	.225	.422*	.200	.220	.332	.000	.238
Prosecuted for murder	.062	.149	−.293*	.117	−.162	.215	−.008	.159
Having no partner
Having a partner	−.201	.181	.070	.145	−.184	.276	−.287	.203
Partner information missing	.033	.193	.070	.178	−.185	.300	−.232	.249
Length of detention history	.141	.042	.135***	.029	.042	.063	.111*	.044
Length of imprisonment during follow-up			.330***	.039			.564***	.043

Age at time of homicide is grand-mean centered in the analyses.

p < .05. **p < .01. ***p < .001.

When examining the control variables, younger offenders and those with a longer conviction history recidivate more quickly and more often, as expected. Furthermore, those who have committed a homicide during robbery are more likely to nonviolently recidivate than criminal homicide offenders. Homicide offenders in the family sphere violently recidivated less frequently than criminal homicide offenders.

We continue by distinguishing between groups of homicide offenders, to test our second and third hypothesis. The interaction terms are presented in Table 7. To avoid interaction terms that include four variables (e.g., Length of imprisonment * partner * Western offender * detention history), interaction terms have been added in separate models, which facilitates comparisons between groups.

Table 7.

Cox Regression Survival Analysis on Recidivism Speed and Poisson Regression Analysis on Recidivism Frequency: Interaction Effects (N = 621)

	Nonviolent speed		Nonviolent frequency		Violent speed		Violent frequency
	B	SE (B)	B	SE (B)	B	SE (B)	B	SE (B)
M1: Partner	−.613	.369	−.540	.309	−.611	.579	−.295	.421
M1: Partner information missing	−.060	.397	.079	.274	−.232	.644	.534	.374
M1: Length of imprisonment	−.057	.041	.058	.035	.020	.062	.207***	.041
M1: Length of imprisonment * Partner	.104	.079	.139*	.063	.096	.122	−.008	.082
M1: Length of imprisonment * Partner missing	.023	.089	.000	.011	.014	.147	−.040*	.018
M2: Western	−.391	.298	−.547*	.241	−.688	.449	−1.520***	.295
M2: Length of imprisonment	−.123*	.054	−.038	.049	−.116	.089	−.015	.052
M2: Length of imprisonment * Western	.134*	.067	.196**	.057	.232*	.105	.346***	.066
M3: Detention history	.273**	.087	.330***	.058	.215	.124	.355***	.080
M3: Length of imprisonment	−.003	.038	.164***	.033	.085	.059	.245***	.040
M3: Length of imprisonment * Detention history	−.028	.017	−.050**	.015	−.040	.028	−.059**	.019

Note: Control variables are omitted from the table.

p < .05. **p < .01 ***p < .001.

From the first model (M1), we see that the effect of length of imprisonment increases nonviolent recidivism frequency only for offenders with an intimate partner. This finding, in line with Hypothesis 2a, is not replicated for the other recidivism indicators. Hence, Hypothesis 2a is only partially supported.

The second model shows that Western and non-Western offenders are affected differently by the length of imprisonment. For non-Western offenders, as the main effect indicates, length of imprisonment does not affect recidivism frequency but decreases nonviolent recidivism speed. For Western offenders, as the interaction effect indicates, length of imprisonment increases recidivism, both speed and frequency. This provides support for Hypothesis 2b.

The third model indicates that imprisonment length increases recidivism frequency of offenders without detention history (i.e., the main effect of length of imprisonment). The longer the offenders’ detention history, the less imprisonment length increases recidivism frequency (i.e., the interaction term). This holds for both nonviolent and violent recidivism frequency but does not apply to recidivism speed. For nonviolent recidivism speed, longer detention history amplifies the negative effect of length of imprisonment; for violent recidivism speed, length of imprisonment shows no effect on recidivism. This pattern provides support for Hypothesis 3 with regard to recidivism frequency but not for recidivism speed.

Discussion

This study focused on the recidivism patterns and the influence of the length of imprisonment on recidivism for Dutch offenders who have been convicted of homicide between 1996 and 2004. We contributed to previous research in several important ways. The few studies on homicide offenders used small-scale data and have generally not applied a theory-driven approach. Moreover, previous research on homicide offenders has rarely tested for the imprisonment effect and has never distinguished between groups of homicide offenders. The large-scale data enabled us to acquire deeper insight into the effect of length of imprisonment: Longer imprisonment led to a higher recidivism frequency and imprisonment length effects differed considerably between groups of offenders. These findings also shed light on the theories we have applied.

Overall, 51% of the homicide offenders recidivated during the follow-up period. This finding is comparable to previous studies on homicide offenders, which found recidivism rates between 51% and 60% (e.g., Heide et al., 2001; Hill et al., 2008). Our results show that longer imprisonment systematically increases recidivism frequency, not recidivism speed.

This study has provided some indication for differences that exist between groups of homicide offenders. Western offenders were more likely to recidivate when imprisonment was longer than those of non-Western origin. This is in line with the proposed mechanism that people with higher status lose more (valuable) social control and are therefore more likely to recidivate. This social control mechanism has also been reported in a study that found reduction of family and community support after prison release to explain why some homicide offenders recidivated and others did not (Cale, 2010). However, we were not able to replicate this finding with the partner effect. Whether or not people had an intimate partner before committing the homicide was only important for the effect of imprisonment length on nonviolent recidivism frequency. The suboptimal way of measuring partner effects—only partner status before homicide, not after release was available—provides one explanation for this partial null finding. In sum, the predictions derived from social control theory as stated in Hypotheses 2a and 2b were partially confirmed.

With regard to social learning theory, we found that imprisonment length increased recidivism to a greater extent when the offenders’ detention history was shorter. This is in line with the prediction that offenders learn criminal skills from being in prisons. Whereas offenders without detention history recidivated more often with longer imprisonment, offenders who already had been imprisoned did not experience this imprisonment effect, as was predicted in Hypothesis 3. Hence, although not explicitly tested, the idea from social learning theory that prisons lead to the acquisition of crime-related knowledge and skills, especially for first-time prisoners, is supported.

Our results further show that imprisonment has a deterrent effect on nonviolent speed for non-Western offenders, which was not anticipated. One explanation for the finding that non-Western offenders are deterred whereas Western offenders are not, is that offenders from non-Western, collectivistic cultures respond with more shame to sanctions than people from individualistic cultures (Bierbrauer, 1992). Alternatively, non-Western immigrants might return to their home country after release from prison, which is why we could not follow up on their criminal career. This, however, would also have resulted in lower recidivism rates for non-Western offenders, which was not observed. Therefore, the first explanation seems more realistic.

Because, according to our study, length of imprisonment does not generally increase or decrease recidivism, it is tempting to conclude that imprisonment is not worth its costs. One should, however, take into account other reasons to imprison homicide offenders (Nunes, Firestone, Wexler, Jensen, & Bradford, 2007). First, society is not—or is substantially less—disturbed by the behavior of offenders when they are imprisoned (Blokland & Nieuwbeerta, 2007). Given the relatively high recidivism rate of homicide offenders, this incapacitation seems important. Second, society might wish to punish the criminal severely to exact retribution. Third, long imprisonment sentences in a society might prevent other citizens from committing a first homicide (McCall, Parker, & MacDonald, 2008; Snodgrass et al., 2011), however the effect of this general deterrence is difficult to assess (Defina & Arvanites, 2002; Levitt, 1996; Marvel, 2009).

Although this study has improved upon previous studies, there are several limitations. First, we have derived predictions from criminological theories but we were not able to test the mechanisms. Although we show, for example, that offenders of Western origin become more likely to recidivate with longer imprisonment, this is only indirect evidence for social control theory. Similarly, the finding that for people with a detention history, imprisonment length increases recidivism less is only suggestive evidence for social learning theory. Although we made a serious attempt to put these theories to a test in the context of the effect of imprisonment on recidivism for homicide offenders, progress can be made in this respect. To further investigate the social control mechanism, for example, future research should focus on the effect of length of imprisonment on divorce (see, for example, Apel, Blokland, Nieuwbeerta, & Van Schellen, 2010) and on labor market outcomes (Kling, 2006; Pager, 2003; Waldfogel, 1994), which in turn could influence recidivism rates. Focusing on different measures of social control, future research could also use indicators such as the presence of children, religious involvement, and neighborhood cohesion.

Another suggestion for future research is to focus on homicide offenders with a different type of punishment. We have included those homicide offenders who received an imprisonment sentence only but there is a relatively large group of homicide offenders who receive an imprisonment sentence combined with a mandatory treatment measure for mentally disordered offenders (TBS). This group differs from other homicide offenders in many respects (Van Marle, 2002); future research should assess whether they follow similar recidivism patterns as the group under study.

In order to further understand the recidivism patterns of homicide offenders, it would be interesting to tease out the differences within the categories of violent and nonviolent recidivism. Especially based on the specific nature of the homicide, differences with regard to recidivism patterns could be expected.

Most importantly, further study is needed on the side effects of imprisonment that lead to higher recidivism rates. When the side effects of losing a job, losing social contacts or learning criminal skills from inmates can be tempered, society as a whole might be better off.

Footnotes

Acknowledgements

The authors wish to thank four anonymous reviewers and the editor for their helpful suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Pieter E. Baay studied sociology at Utrecht University. He is a doctoral researcher at the department of developmental psychology and the department of sociology at Utrecht University. His current research focuses on factors that influence the transition from school to work for vocational training graduates, such as personality, social capital, ethnic identity, and work motivation.

Marieke Liem received her PhD in forensic psychology. She is assistant professor at the department of criminology at Leiden University and is currently a Marie Curie fellow at the Harvard Kennedy School of Government. Her research focuses on criminal careers of homicide offenders.

Paul Nieuwbeerta is professor in criminology at Leiden University. His research interests include the development of crime over the life course, the determinants and consequences of judicial interventions, incarceration and homicide research.

References

Agresti

(1996). An introduction to categorical data analysis. New York, NY: Wiley Interscience.

Apel

Blokland

A. J.

Nieuwbeerta

Van Schellen

(2010). The impact of imprisonment on marriage and divorce: A risk set matching approach. Journal of Quantitative Criminology, 26, 269-300.

Bierbrauer

(1992). Reactions to violation of normative standards: A cross-cultural analysis of shame and guilt. International Journal of Psychology, 27, 181-193.

Blokland

A. J.

Nagin

Nieuwbeerta

(2005). Life span offending trajectories of a Dutch conviction cohort. Criminology, 43, 919-954.

Blokland

A. J.

Nieuwbeerta

(2007). Selectively incapacitating frequent offenders: Costs and benefits of various penal scenarios. Journal of Quantitative Criminology, 23, 327-353.

Cale

Plecas

Cohen

I. M.

Fortier

(2010). An exploratory analysis of factors associated with repeat homicide in Canada. Homicide Studies, 14, 159-180.

DeJong

(1997). Survival analysis and specific deterrence: Integrating theoretical and empirical models of recidivism. Criminology, 35, 561-576.

Defina

R. H.

Arvanites

T. M.

(2002). The weak effect of imprisonment on crime: 1971-1998. Social Science, 83, 635-653.

Delisi

Scheerer

A. M.

(2006). Multiple homicide offenders: Offense characteristics, social correlates, and criminal careers. Criminal Justice and Behavior, 33, 367-391.

10.

Farrington

D. P.

(1977). The effects of public labelling. British Journal of Criminology, 17(2), 112-125.

11.

Gardner

Mulvey

E. P.

Shaw

E. C.

(1995). Regression analyses of counts and rates: Poisson, overdispersed Poisson, and negative binomial models. Psychological Bulletin, 118, 392-404.

12.

Gendreau

Cullen

F. T.

Goggin

(1999). The effects of prison sentences on recidivism. Ottawa, Ontario, Canada: Solicitor General Canada.

13.

Gottfredson

Hirschi

(1990). A general theory of crime. Cambridge, MA: Harvard University Press.

14.

Gottlieb

M. R.

Gabrielsen

(1990). Results from a homicide project in Copenhagen. International Journal of Law and Psychiatry, 13, 191-205.

15.

Hagan

(1997). An analysis of adolescent perpetrators of homicide and attempted homicide upon return to the community. International Journal of Offender Therapy and Comparative Criminology, 41, 250-259.

16.

Heide

K. H.

(1999). Young killers: The challenge of juvenile homicide. Thousand Oaks, CA: Sage.

17.

Heide

K. H.

Spencer

Thompson

Solomon

E. P.

(2001). Who’s in, who’s out and who’s back: Follow-up data on 59 juveniles incarcerated in adult prison for murder or attempted murder in the early 1980s. Behavioral Sciences and the Law, 19, 97-108.

18.

Hill

Habermann

Klusmann

Berner

Briken

(2008). Criminal recidivism in sexual homicide perpetrators. International Journal of Offender Therapy and Comparative Criminology, 52(1), 5-20.

19.

Hirschi

(1969). Causes of delinquency. Berkeley: University of California Press.

20.

Hosmer

D. W.

Lemeshow

May

(2008). Applied survival analysis: Regression modeling of time to event data. New York, NY: Wiley.

21.

Johnson

B. D.

Van Wingerden

Nieuwbeerta

(2010). Sentencing homicide offenders in the Netherlands: Offender, victim, and situational influences in criminal punishment. Criminology, 48, 981-1018.

22.

Kling

J. R.

(2006). Incarceration length, employment, and earnings. American Economic Review, 96, 863-876.

23.

Levitt

(1996). The effect of prison population size on crime rates: Evidence from prison overcrowding litigation. Quarterly Journal of Economics, 111, 319-351.

24.

Loeber

Lacourse

Homish

D. L.

(2005). Homicide, violence, and developmental trajectories. In Tremblay

R. E.

Hartup

W. W.

Archer

(Eds.), Developmental origins of aggression (pp. 202-220). New York, NY: Guilford.

25.

Loughran

T. A.

Mulvey

E. P.

Schubert

C. A.

Fagan

Piquero

A. R.

Losoya

S. H.

(2009). Estimating a dose-response relationship between length of stay and future recidivism in serious juvenile offenders. Criminology, 47, 699-740.

26.

Marvel

T. B.

(2009). Prison population and crime. Retrieved from http://ssrn.com/abstract=1452427

27.

McCall

Parker

MacDonald

(2008). The dynamic relationship between homicide rates and social, economic, and political factors from 1970 to 2000. Social Science Research, 37, 721-735.

28.

Nagin

Cullen

F. T.

Jonson

C. L.

(2009). Imprisonment and reoffending. Crime and Justice: A review of research, 38, 115-200.

29.

Nieuwbeerta

Leistra

(2007). Dodelijk geweld: Moord en doodslag in Nederland [Deadly violence: Murder and manslaughter in The Netherlands]. Amsterdam, Netherlands: Uitgeverij Balans.

30.

Nunes

K. L.

Firestone

Wexler

A. F.

Jensen

T. L.

Bradford

J. M.

(2007). Incarceration and recidivism among sexual offenders. Law and Human Behavior, 31, 305-318.

31.

Pager

(2003). The mark of a criminal record. American Journal of Sociology, 108, 937-975.

32.

Paulus

P. B.

Dzindolet

M. T.

(1992). The effects of imprisonment. In Suedfeld

Tetlock

P. E.

(Eds.), Psychology and social policy (pp. 327-341). New York, NY: Hemisphere.

33.

Piquero

A. R.

Blumstein

Brame

Haapanen

Mulvey

E. P.

Nagin

D. S.

(2001). Assessing the Impact of Exposure Time and Incapacitation on Longitudinal Trajectories of Criminal Offending. Journal of Adolescent Research, 16(1), 54-74.

34.

Roberts

Zgoba

Shahidullah

(2007). Recidivism among four types of homicide offenders: An exploratory analysis of 336 homicide offenders in New Jersey. Aggression and Violent Behavior, 12, 493-507.

35.

Sampson

R. J.

Laub

J. H.

(1990). Crime and deviance over the life course: The salience of adult social bonds. American Sociological Review, 55, 609-627.

36.

Sampson

R. J.

Laub

J. H.

(1993). Crime in the making. Cambridge, MA: Harvard University Press.

37.

Smit

P. R.

Bijleveld

C. J. H.

Van Der Zee

(2001). Homicide in the Netherlands: An Exploratory study of the 1998 cases. Homicide Studies, 5, 293-310.

38.

Smith

Goggin

Gendreau

(2002). The effects of prison sentences and intermediate sanctions on recidivism: General effects and individual differences. Ottawa, Ontario, Canada: Solicitor General Canada.

39.

Snodgrass

G. M.

Blokland

A. A. J.

Haviland

Nieuwbeerta

Nagin

D. S.

(2011) Does the time cause the crime? An examination of the relationship between time served an reoffending in the Netherlands. Criminology, 49, 1149-1194.

40.

Spohn

Holleran

(2002). The effect of imprisonment on recidivism rates of felony offenders: A focus on drug offenders. Criminology, 40, 329-358.

41.

Statistics Netherlands. (2009). Statline. Retrieved from http://statline.cbs.nl

42.

Van Marle

H. J. C.

(2002). The Dutch Entrustment Act (TBS): Its principles and innovations. International Journal of Forensic Mental Health, 1(1), 83-92.

43.

Villettaz

Killias

Zoder

(2006). The effects of custodial vs. non-custodial sentences on re-offending. Lausanne, Switzerland: University of Lausanne.

44.

Waldfogel

(1994). Does conviction have a persistent effect on income and employment? International Review of Law and Economics, 14(1), 103-119.

45.

Wermink

Blokland

Nieuwbeerta

Nagin

Tollenaar

(2010). Comparing the effects of community service and short-term imprisonment on recidivism: A matched samples approach. Journal of Experimental Criminology, 6, 325-349.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

10.28 MB