Why longer prison terms fail to serve a specific deterrent effect: an empirical assessment on the remembered severity of imprisonment

ABSTRACT

For a prison sentence to exert a specific deterrent effect, the ultimate question is that imprisonment is remembered as aversive once the offender is released, and is contemplating future criminal activities. Drawing on insights from social psychology and cognition, this study assessed (1) how inmates remember the severity of their imprisonment following release, and (2) how the severity as experienced while being incarcerated (e.g. the worst or the last moment) affects its recollected aversiveness among a sample of Dutch inmates who were released for approximately six months (n = 696). The findings indicated that the severity as experienced while being incarcerated is strongly related to the severity as recollected following release, net of the duration of confinement. Strikingly, to the extent that the length of imprisonment affected its recollected aversiveness, it did so in the opposite direction than traditional deterrence research presumes. Implications for correctional policy and future research are discussed.

KEYWORDS:

• Specific deterrence
• imprisonment
• severity
• subjective experience
• memories

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Using registry data from the Dutch Custodial Institutions Agency, we checked to what extent participants (n = 1904) differed from those who refused to participate (n = 933) on a wide range of characteristics. Participants were comparable to non-participants in terms of age, being accused of a violent offence, and being accused of a drugs offence. Participants were however slightly older when they were first registered for a crime (M = 18.59 versus M = 17.02; t(2493) = −6.88, p < .001), had on average less prior convictions than non-participants (M = 8.82 versus M = 11.75; t(1565) = 6.04, p < .001), and were less likely to be accused of a property offence (30.7% versus 35.9%; χ²(1) = 7.67, p < .01).

2. Participants and non-participants were compared across a wide range of variables. At D2, no differences between both groups emerged according to age, education, employment, relationship- parental- and housing status, prior prison experience, and the offence type they were accused of. Participants were, however, less likely to be non-ethnic Dutch (33% versus 49%; χ²(1) = 10.21, p < .01), were more likely to be incarcerated for the first time (30% versus 15%; χ²(1) = 10.71, p < .01), and were slightly older when they first registered for a crime (M = 20.43 versus M = 18.32; t(262) = 2.73, p < .01). Participants at D3 were more likely to be employed prior to entrance into pre-trial detention (53% versus 34%; χ²(1) = 4.67, p < .05). No differences emerged between both groups at D4 (cf. Appendix 1).

3. Inmates who participated at R1 were representative to those who refused to participate in terms of employment, parental, and housing status preceding entrance into pre-trial detention, and in terms of prior prison experience. Participants differed most from those who refused to participate in terms of ethnicity, relationship status, and the number of prior convictions. Participants were less likely to be non-ethnic Dutch (38% versus 49%; χ²(1) = 25.85, p < .001), and to have a partner (45% versus 54%; χ²(1) = 16.30, p < .001). In addition, participants were more likely to be convicted for the first time (22% versus 15%; χ²(1) = 14.01, p < .001) (cf. Appendix 2). 

4. For 63 cases, sentencing data were missing. Data from the Public Prosecutor’s Offices (COMPAS) were used to reduce the number of missing values on this variable.

5. To deal with missing values, mean imputation was performed for age onset (n = 1), housing (n = 3), employment (n = 3), and relationship status following release (n = 18).

6. While the means suggest that the SESI decreases even further after D3, the change from D3 to D4 was, in contrast to the change from D1 to D2, t(355) = −2.52; p < .05, and D2 to D3 t(63) = 3.11; p < .01, not statistically significant. 

7. As will be described further on, almost 50% of our sample were already released by the time the data for the second wave (D2) were collected, and consequently, only a single measure of the SESI was available for them.

8. For 16 inmates (2.3%), the SESI measure at D2 and/or D3 was missing, while the SESI measure at D4 was available. To create a measure of the accumulated severity for these inmates, we imputed the mean of the two surrounding SESI values.

9. The severity as experienced at the end of each subarea was used as inmates were requested to keep in mind the period since the last interview was conducted (e.g. 2.5 months), in answering the questions capturing the SESI.

10. It should be noted however, that, once released, inmates may again serve time in confinement, which is not accounted for in our measure of the length of stay. Inmates may, for instance, be sentenced to prison for a new offence. Alternatively, it is possible that inmates who were pre-trial released received extra time once their case was tried in court. In both cases, our measure of the length of stay is an underestimation of the actual time served once inmates participate in the R1 interview, and recollect the severity of their imprisonment. Additional analyses, excluding inmates who were pre-trial released, yet received extra time (n = 524) did however, not alter the magnitude nor the significance of our findings.

11. The same analyses were also conducted using a linear measure of the length of stay. This did however not alter the direction nor the magnitude of our findings.

12. The age of onset was measured as inmates’ age when they were first registered for a crime. While registration of offences in the Netherlands starts at age 12, imprisonment can be imposed from age 18 onwards.

13. In total, five inmates had a missing value on the index offence according to the GDF. To deal with these missing values, we relied on data obtained from the Public Prosecutor’s Offices (COMPAS).

14. For illustrative purposes, the remembered severity of imprisonment was recoded into discrete categories. In the analyses, however, this variable was included as a continuous one. 

15. According to Cohen (1992), standardized coefficients around 0.1, 0.3, and 0.5 represent weak, moderate, and strong effects, respectively.

16. Note that the correlation between the initial and the end severity is 0.785, which may suggest that multicollinearity is not a problem here. However, the correlations between the initial severity on the one hand, and the end, and the peak-end severity on the other hand, are all higher than 0.8. The same accounts for the relation between the end severity on the one hand, and the peak and the peak-end severity on the other hand. Accordingly, although problems with multicollinearity may not apply to the initial and the end severity, the high correlations between both measures of severity with the end and the peak-end severity, imply that further elaborating on this distinction may not be very useful.

17. Due to the low number of inmates who participated in the fourth wave (n = 10), no further analyses were conducted on this group. Alternatively, we attempted to conduct the same analyses one the entire sample (i.e. without excluding inmates who were not sentenced to imprisonment, and those who were sentenced, or served over two years in prison (N = 906)). This did, however, not result in any substantial differences in our findings. 

Additional information

Funding

The Prison Project is financially supported by the University of Leiden, the Netherlands Institute for the Study of Crime and Law Enforcement (NSCR), the Netherlands Organization for Scientific Research (NWO), and Utrecht University.