Sentencing in New Zealand: a
statistical analysis
Foreword | Acknowledgements | Executive Summary | Introduction | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Chapter 7 | Chapter 8 | Chapter 9 | Chapter 10 | Chapter 11 | References | Appendix I | Appendix II | Tables | Figures
9.1 Background
Probation was introduced in 1954 and was replaced by supervision in 1985. As the name suggests, the sentence involves the supervision of the offender by a probation officer on a regular basis, with certain standard conditions (e.g. notifying a change of address). Supervision may also be imposed with additional special conditions (e.g. attending a work or social skills training programme).
Supervision is frequently imposed as cumulative to a short sentence of imprisonment or in combination with periodic detention. As the primary sentence in such cases is taken as the prison or periodic detention sentence, these cases are not included in this analysis of supervision.
This chapter examines which statistical factors most influence the use of supervision sentences, looking first at the variables individually and then at their combined effects using multivariate modelling techniques. Changes in the use of supervision are analysed in the final section of this chapter.
9.2 Single variable analysis of current factors influencing sentencing
Overall, 6.3% of the offenders who had a proved case in 1995 received supervision as their primary sentence.
Compared to imprisonment, the use of supervision shows only fairly weak relationships with the current seriousness of offending and the previous criminal history of the offender (Figure 9.1).
Offenders who have committed offences of very low or very high seriousness have a relatively low probability of receiving a supervision sentence, as do those convicted on only one or two charges. The use of supervision is fairly constant over moderate levels of seriousness (e.g. seriousness scores in the range 20-200) and peaks for the relatively small number of offences with a seriousness of around 300-360 (mainly serious offences against the person).
There is relatively little relationship between the use of supervision and the criminal history characteristics of the offender.
Figure 9.1: Percentage of offenders receiving a supervision sentence in 1995 for selected variables
Nor are there very clear distinctions in the probability of receiving a supervision sentence between demographic groups. Females have a slightly higher probability of receiving supervision (9%) than males (6%), while Pacific peoples have a slightly higher probability (9%) than Māori and Pakeha/Other ethnic groups (7%). Age makes relatively little difference, except for a lower probability for youth offenders (3%). The lower use of supervision for youth offenders is not surprising, given the availability of youth supervision and community work orders as alternative sentences.
Offence type is a more significant factor. Cases involving domestic violence have by far the highest probability of a supervision sentence (27% of proved cases for this offence type result in a supervision sentence). Serious offences against the person (11%), minor offences against the person (9%) and property offences (8%) are also more likely than other offences to result in supervision.
The percentage of offenders receiving a supervision sentence is higher if the most recent sentence prior to the current case is also supervision (14%), than if the previous sentence was a prison sentence (6%), periodic detention (5%), community service (8%), community programme (10%) or a monetary penalty (5%).
9.3 Multivariate analysis of current factors influencing sentencing
9.3.1 The fit and accuracy of the 1995 model
The logistic regression models for supervision achieved a significant overall fit to both the full and half 1995 data, as indicated by log likelihood ratios significant at the 0.0001 level of probability and non-significant residual (unexplained variation) terms. The model not only fits the data it was developed on, but fits equally well to other data. Thus, the results of the test phase (Figure 9.2) show that the model developed using one half of the 1995 data (the 'fitted data') fits equally well to the other, unseen half of the 1995 data (the 'test data').
The supervision model differs from the prison model in that the results are only close to the ideal line up to a probability of about 0.3 and there were very few offenders with a probability more than 0.4. Thus, there were no factors or combination of statistical factors that adequately predicted a high probability of receiving a supervision sentence. This is perhaps not surprising, given that community-based sentences are targeted at an intermediate group of offenders (those who have neither a high probability of imprisonment nor a high probability of receiving a fine or no sentence).
Also, contextual factors relevant to the specific offence and offender, but not measured by the statistical factors available for this study, presumably make a substantial contribution to the sentencing decision. For example, supervision is appropriate when some form or counselling or therapy is indicated (e.g. for drug addiction or violent behaviour) and where closer supervision of the offender is required than can be provided by other sentences.
Figure 9.2: Plot of the predicted probability versus the actual proportion of offenders receiving supervision, 1995 fitted and test data
9.3.2 Results of the 1995 model
When all of the statistical variables are compared in the logistic regression model, the most significant factor is having a current offence of domestic violence, at more than five times the odds ratio or relative risk of property offences (Table 9.1). Other types of violent offences (minor and serious offences against the person) are also more likely than property offences to result in a supervision sentence. Conversely, the probability of receiving a supervision sentence is reduced if the current offence is a traffic offence, drug offence or breach of a periodic detention sentence.
Offences of moderate seriousness (>20-180 and >180-365) have the highest odds ratios of the seriousness groups. People who have been charged with more than one offence are more likely to receive supervision than those charges with only one offence.
Women are almost twice as likely to get supervision as men, once the effects of other factors are taken into account (such as the higher rate of violent offending by men). As for other sentences, youth offenders (aged under 17) have a low relative risk of supervision. Māori offenders appear to be slightly less likely to receive supervision. A guilty plea appears to increase the probability of supervision.
Table 9.1: Logistic regression model of the probability of supervision, 1995
Note: An odds ratio of >1.0 indicates a high relative risk (i.e. more likely to receive this sentence than the reference group). The most significant variable (highest Wald Chi-square, lowest probability P), is rank '1'.
As for other community-based sentences, the current sentence is more likely to follow a sentence of the same type. That is, if the most recent sentence is supervision, or a supervision sentence has been served in the past, then the probability of a further supervision sentence is increased. Previous community programme and community service sentences also increase the probability of supervision.
Other than having a previous supervision sentence, the criminal history variables are amongst the least significant of all variables. A low to moderate number of previous cases and previous seriousness and a moderate gap since the last case tend to increase the odds of a supervision sentence.
9.4 Changes in the use of supervision
The use of supervision has increased over the 1982 to 1997 period, although the magnitude of the increase has been much less than for either community service or periodic detention. In 1982, 4.5% of proved cases involving imprisonable offences resulted in a supervision sentence, compared to 6.3% in the peak year of 1995 (Figure 9.3). In contrast to other community-based sentences, which increased in the late 1980s and early 1990s, the main period of growth for supervision was in the mid 1990s.
Figure 9.3: The percentage of proved cases resulting in a supervision sentence for imprisonable offences, 1982-1997
A change in the use of a particular sentence may be due to one or both of the following: (i) a change in the type of offence or offender being dealt with by the courts or (ii) a change in sentencing practice with respect to these characteristics (such that offenders with the same statistical characteristics are now more or less likely to receive a particular sentence than in the past). Given the use of supervision for offences of moderate seriousness and particularly its use for violent offences, the significant increases in the average seriousness of offences and the number of violent offences dealt with by the courts would be expected to result in an increase in the use of supervision.
To test whether sentencing practice has changed with respect to supervision, the logistic model developed using 1995 data was used to predict the probability of receiving a supervision sentence for offenders in earlier years. This means that an offender in 1983 who has the same criminal history and current case characteristics as a person in 1995 would have the same predicted probability of supervision. The predicted probabilities are then compared to the actual proportion of people receiving a supervision sentence in each year. If no change in sentencing practice has occurred with respect to the statistical variables, then the predicted probability should be the same as the actual proportion receiving supervision.
The results indicate that much of the increase in the use of supervision between 1983 and 1995 can be explained by changes in statistical factors. Overall, 4.5% of offenders in 1983 received supervision, compared to 6.3% in 1995 (Table 9.2). The similarity of the actual 1983 percentage (4.5%) to the predicted percentage for 1983 (5.0%) based on the 1995 model indicates that, overall, there has been relatively little change in sentencing practice between 1983 and 1995. However, there does appear to have been a decrease in the use of supervision between 1987 and 1991, followed by an increase between 1991 and 1995. Thus, the actual use of supervision is lower in 1991 (3.4%) than predicted (5.2%) given the types of offenders and offences dealt with by the courts that year.
Changes in sentencing practice over the range of probabilities are shown in Figure 9.4.
Figure 9.4: Comparison among years of the actual and predicted probabilities of supervision, with predictions made using the 1995 model
Table 9.2: Actual percentage receiving supervision for each variable and year and predicted percentage based on 1995 sentencing practice
Note: The predicted percentage is the percentage of offenders who would have received the sentence had the sentencing practices of 1995 been applied, as predicted by the 1995 logistic model. The difference between the predicted percentage for 1995 and other years indicates the proportion of the total change due to changes in statistical factors (e.g. the increase in average seriousness) while the difference between the actual and predicted percentage for each year indicates the proportion of total change due to changes in sentencing practice.
The lower than predicted use of supervision in 1991 appears to occur over most levels of probability. Differences between 1995 and the 1980s occur mainly at probabilities of 25% or more, for which the number of offenders is relatively small (and therefore this difference does not make much impact on the overall probability).
A significant increase in the use of supervision, especially between 1991 and 1995, has occurred for all the violent offences (particularly domestic violence, but also minor and serious offences against the person). For example, 11% of offenders convicted of domestic violence in 1991 received a supervision sentence compared to 27% in 1995. These increases cannot be explained by changes in other factors, as indicated by the much smaller increase in the predicted percentages for these offences. The number of domestic and other violence cases prosecuted increased rapidly between 1991 and 1995. This trend, combined with an increase in the use of supervision for violent offences, has made a significant contribution to the overall increase in the number of people sentenced to supervision over this period.
Figure 9.5: Actual percentage receiving supervision and predicted percentage based on 1995 sentencing practice, by offence seriousness and previous conviction history, 1983-1995
There has also been an increase in the use of supervision for offences of high seriousness and for offenders with several previous proved cases (Figure 9.5 and Table 9.2). Pacific peoples also show a relative increase in the use of supervision, reflecting the greater proportion of domestic violence relative to other types of offence for Pacific peoples. The relative use of supervision for youth offenders (aged under 17) has decreased, as has the use of other community-based sentences and prison sentences for this age group.
