Link to full report (PDF)

Executive summary

This report presents further results from the 2006 New Zealand Crime and Safety Survey (NZCASS). It looks at what best explains why some social groups face higher victimisation risk than others. One of the purposes of the survey is to understand this better. NZCASS offers the most recent nationally representative data of its kind to do this. It thus offers valuable information for criminal justice practitioners and other researchers.

The sample

Those who took part in NZCASS were interviewed in the first half of 2006. They comprised a nationally representative random sample of 4,229 people aged 15 and over in private households in New Zealand. A Māori ‘booster’ sample of 1,187 was also included. One person per household was interviewed.

Analysing who is at greatest risk

NZCASS showed that in 2005 39% of New Zealanders had been victims once or more of crimes measured by the survey. Some were victimised more than once. To date, differences in victimisation risk have been examined in terms of the characteristics of individuals and households one characteristic at a time. This is bivariate analysis. Results from this in the Key Findings report of NZCASS (Mayhew and Reilly, 2007a) showed differences in risk in terms of age, ethnicity, and area deprivation, for instance. However, bivariate analysis does not account for the overlap of different risk factors. (Māori and those in deprived areas are both at high risk, but Māori more often live in deprived areas.) The current report uses multivariate analysis to deal with the overlap. The multivariate techniques used here produce so-called ‘models’ in which the most predictive combinations of many overlapping factors are selected and their effects estimated.

This report looks at which factors emerged as most important in relation to risk of burglary, vehicle crime (thefts of and from vehicles and vehicle interference), and confrontational (violent) crime. Confrontational crime was broken down into offences by:

• partners
• people well known to the victim
• other offenders (strangers or those known casually or by sight).

The evidence from NZCASS is set alongside a comprehensive review of multivariate analyses of victimisation risk done in other Western countries. There is much consistency in results as to main risk factors, and there is no good reason to think that what determines victimisation in New Zealand is much different. The international analyses give a more rounded picture of what most influences people’s risk than NZCASS results can do on their own.

Main findings

Table A shows factors that were generally associated in the earlier bivariate analysis with higher victimisation risk across a number of offences. The earlier bivariate analyses did not include all the factors examined in the current multivariate analyses; the new factors are marked ‘nt’ (not tested) in Column A.

Columns B to F in Table A show the factors that emerged in the multivariate statistical models as best explaining victimisation risk. These are marked with a . Factors marked ? were not in the best-performing statistical models, but may have some relationship. For burglary and vehicle crime, which are offences against households, it was not appropriate to examine all individual characteristics of respondents; these factors are marked ‘na’ (not applicable) in Columns B and C.

Table A Summary of the NZCASS multivariate analysis results

Note:

nt = not tested

na = not tested because not applicable

What Table A shows is that:

• Being younger is a consistent predictor of higher risk, even considering other factors. Being younger was also a risk factor in bivariate analyses. People aged 15-19 were sometimes at less risk than people aged 20-24, however.
• Living in a neighbourhood rated by survey respondents as high on social disorder (a new measure) was of importance. This measure, defined in Appendix B, may be serving as a general indicator of disadvantage. In bivariate analysis, disadvantage as measured by the New Zealand Index of Deprivation (NZDep) emerged strongly as associated with higher victimisation. After controlling for other factors (including disorder), there was no evidence that NZDep was related to risk of any crime types except burglary, and that finding was tentative.
• The level of crime recorded by the Police in the surrounding area (another new measure) was related to burglary and vehicle crime, but not to confrontational crime. This factor was included as a measure of proximity to offenders. The fact that burglary and vehicle crime are more often reported to the Police than confrontational crime means that Police statistics will reflect them better.
• Living in a more urbanised area was important for vehicle crime and to a lesser extent for burglary. Risk in Auckland was the highest, net of other effects. The same pattern emerged in bivariate analyses.
• The household having difficulties managing on their income (a new measure) was of importance for confrontational crime. Again, this measure may be taking the place of the NZDep indicator of disadvantage.
• Being single, divorced/separated, or in a de facto relationship predicted confrontational crime, as it did in the bivariate analyses. It was not an appropriate factor to test in the multivariate analyses for the property offences of burglary and vehicle crime.
• Māori ethnicity was associated with a higher risk of confrontational crime by people well known and possibly of a higher risk of confrontational crime by partners. It had been of importance in bivariate analyses across a broader range of offences.
• Sole-parent households were at higher risk of burglary. International research, though, suggests that this could be associated with other factors which were not measured in NZCASS. The level of home security that can be afforded by those in the Housing New Zealand Corporation stock, for instance, could be low, and many sole parents live there. Those in sole-parent households seemed more consistently at risk in the bivariate analyses.
• Having more children in the household (a new measure) was associated with risk of confrontational crime by partners.
• Men were more at risk of confrontational crime by other offenders, but gender made no difference to the other two forms of confrontational crime. This is consistent with the results of the bivariate analysis.
  

Table A confirms that both the type of person one is and the type of neighbourhood in which one lives influence risk. Individual characteristics emerged in all five models in some way although, except for age, they were not consistently predictive across the five types of victimisation. Neighbourhood variables (the level of social disorder, the local Police-reported crime rate, and urbanisation) also featured in the models, though again not in a consistently predictive way across the five types of victimisation. The results are consistent with the fact that victimisation will vary within neighbourhoods (with some people at higher risk than others), as well as between neighbourhoods.

Caveats

Some care is required when interpreting these findings.

• It is very difficult to establish causation based solely on cross-sectional surveys, and the NZCASS is no exception. While some of the findings may reflect true causal mechanisms, many of the risk factors identified could be marker variables – i.e. they may only indicate effects due to underlying causes that have not been measured. For example, as noted above, low home security might explain at least part of the higher risk of burglary among sole-parent households.
• Many explanatory variables are themselves correlated. The statistical modelling process used for this analysis gives an optimal measure of the contribution of a particular variable, assuming that it is not correlated with any of the other explanatory variables. Often this is not a sensible assumption, and it may not be possible to completely disentangle the effects of correlated explanatory variables. Related to this, several correlated variables may be represented in a statistical model by just the one variable that is more closely related to victimisation risk. This does not mean that the other variables have no importance at all.
• Where the sample size for a group is small, the importance of some risk factors could be obscured. This may affect findings for Pacific peoples, for instance.
• Some groups may be less willing or able to report victimisation in the survey. This, too, could affect the results.
• The statistical models do not provide a perfect explanation of what determines victimisation risk. This can be because of unknown or unmeasured variables. Burglary and vehicle crime were predicted less well than the three forms of confrontational crime that the analysis examined.

The findings on ethnicity

Māori and Pacific peoples emerged at higher than average risk in bivariate analyses of NZCASS data across a wide range of offences. (Asians did not do so.) The present analyses help explain why this was so by taking account of demographic structure and social position, which are related both to overall victimisation risk and to ethnicity.

In the present analyses, ethnicity per se was not related to victimisation by burglary or vehicle crime, or to confrontational crime by other offenders. Māori ethnicity was a risk factor for offences by people well known, and a possible risk factor for offences by partners – though the effect was small relative to other predictive factors.

Pacific peoples were not identified as being at greater risk of victimisation, although they had been for many offences in bivariate analyses. The lack of evidence of higher risk for Pacific peoples in the multivariate analyses may be due to the small sample size for this group. Alternatively, it may be because other social and demographic factors are at issue.

Other groups

Other sectors of the population emerged at high risk in the previous bivariate analyses. Those unemployed and/or on benefits were at higher risk, but this is now better explained by other correlated factors, such as low income, where they live, and whether they are married or in less stable relationships.

Students, flatmates, social renters and private renters were also at higher risk in the bivariate analyses, but this again now seems to be explained by other factors – age in particular.

International evidence

The international analyses were generally in accord with what NZCASS found about the best demographic and social predictors of risk. However, some other factors, especially behavioural factors, were of importance too. Most of these were not measured in NZCASS. The most important are below.

• Burglary – Risk of burglary is related to:
  – occupancy – homes left empty more often are at higher risk
  – accommodation type – houses are more at risk than flats, irrespective of other effects
  – household security – poorly secured homes are more at risk.
• Vehicle crime – Risk is related to parking patterns (parking on the streets, as many flat dwellers may have to do, is risky). None of the international analyses measured the level of vehicle security, but other research shows this may be crucial.
• Confrontational crime – Risk has been shown to be higher for those:
  – living in a disadvantaged area – this may not have appeared in the NZCASS results because of a strong interrelationship between disorderly and disadvantaged neighbourhoods. Not all the international analyses considered them both.
  – having an active and risky lifestyle – this is a particularly strong predictor of risk of violence outside the home.
  

The import of the current results

The findings in this report provide sharper focus on the types of people in New Zealand who are at greatest risk of victimisation. The report uses some new measures to make the picture clearer.

Previous bivariate analyses identified a number of risk factors that still emerge as important in the more sophisticated multivariate analyses reported here. Being younger is one factor. Being in a less stable partnership (or no partnership at all) is another.

Previous analyses also strongly suggested that New Zealanders were at higher risk if they were less well placed economically and socially – even if this was simply by virtue, for instance, of age, ethnicity, or living in more densely populated and deprived areas. The current results indicate that social disadvantage remains important in explaining vulnerability to victimisation, although new measures of disadvantage – in particular high levels of local social disorder and high local levels of crime as recorded by the Police – proved more illuminating than the measures of disadvantage previously used.

Importantly, the current results indicate that many high-risk groups previously identified on the basis of bivariate analysis are at high risk because of a nexus of associated risk-related factors, such as a younger age profile, less stable partner relationships, and more personal and neighbourhood economic strain. The apparent elevated risk from certain individual characteristics (for instance, being Māori, living in a sole-parent household, being a student, or living with flatmates) is explained by the way other associated factors increase risk. Associated behavioural factors, which are themselves related to some of the other socio-demographic factors, may also be playing a part.

The implications of the current results in terms of policy are not particularly clear-cut. Some points are:

• There is probably not much that can be done to address the fact that younger people are more at risk. Altering the way in which younger people conduct their lives is clearly difficult. Younger people can perhaps be made more aware of their patterns of risk, although other analyses of NZCASS results indicate that they are already aware of their heightened risk, as evidenced by their replies to a number of NZCASS questions on concern about crime (Mayhew and Reilly, 2007b).
• The corollary of higher risk for younger people is lower risk for the elderly. This has failed to come across as a clear public message, especially as media coverage tends to focus on victims who are vulnerable, such as the elderly, in spite of the fact that their experiences of crime are uncommon.
• The current results suggest that it is inherent disadvantage rather than factors related to it (such as Māori and Pacific ethnicity, or sole parent status) which underlies heightened risk of victimisation. A key factor here is living in a disorderly (and probably deprived) neighbourhood. Some social groups are more characterised by disadvantage than others are, making them more vulnerable to crime. Māori are an example here. Their collective experiences, then, are worse. From a policy perspective, this means that crime prevention and victim service initiatives targeted at more socially and economically deprived areas as a whole are the most worthwhile. Such initiatives will encompass the high-risk groups most likely to be represented there, including Māori. This is likely to apply to both victims and offenders.
  

Full report