National Drug Strategy
National Drug Strategy

The costs of tobacco, alcohol and illicit drug abuse to Australian Society in 2004/05

Appendix B, Drugs and crime: calculating attributable fractions from the DUMA and DUCO projects

prev pageTOC |next page

This appendix was written by Dr Toni Makkai, Director, Australian Institute of Criminology and Dr Jeromey Temple, Research Fellow, Research School of Social Sciences, Australian National University.

Acknowledgments

Numerous agencies and individuals have contributed to the DUMA and DUCO collections and their support is gratefully acknowledged, as well as those many individuals who have voluntarily provided personal information on their drug and crime activities. In addition to the considerable 'in-kind' support provided to the projects by various police and juvenile and corrective services agencies, significant financial funding has been provided by the Australian and South Australian Governments. Neither the funding agencies, collectors, nor the various agencies involved bear any responsibility for the analyses or interpretations presented here. We would like to acknowledge the earlier work undertaken by Institute staff, Kiah McGregor and Paul Williams, upon which this update draws.

Introduction

In 2002 the Australian Institute of Criminology produced a series of fractions for crime that could be attributed to drug use (see Makkai and McGregor, 2002 and Williams, 2002). This paper updates that information and as a result draws on the earlier work produced by Institute staff at that time. There are significant limitations in the current national crime and justice collections which do not allow us to produce attributable fractions that could be deemed as being 'true'. In particular Australia does not produce annual national data on:
Administrative data sources such as police and prison crime statistics have significant limitations and are often not regularly published at a national level (see Makkai, 1999; Carcach, 1997; Carcach and Makkai, 2002; Pernanen, et al., 2002). This is compounded by measurement issues of intoxication and dependency (see Makkai, 2002; Pernanen et al., 2002). There is currently no reliable drug test that can determine levels of intoxication (see Poysner, Makkai, Norman and Mills, 2002) or dependency for illegal drugs such as heroin and cocaine.

There have been a number of studies in criminology that have shown discrepancies between self-reported use and drug testing results amongst police detainees and incarcerated detainees (Harrison and Hughes, 1997; Committee on Data and Research for Policy on Illegal Drugs, 2001). These studies have found that concordance between self-report and chemical testing for illegal drug use varies by socio-demographic characteristics and the particular drug involved (see McGregor and Makkai, 2003). There is also debate over whether self-reported attributions for drugs and offending are reliable (see Davies, 1992 and Dalrymple, 2006).

Top of Page

There are essentially three models or ideal types that are used to explain the causal links between drugs and crime (see Pernanen et al., 2002):
  1. Psychopharmacological—this postulates that the person was intoxicated and the intoxication resulted in antisocial and criminal behaviour. This requires data on the level of intoxication at the time of offending and that the intoxication caused the behaviour.
  2. Economic compulsive—this postulates the person has a drug dependency problem that 'compelled' the person to commit crimes to support the drug habit. Again this model requires that a causal link be demonstrated.
  3. Systemic—the crimes result from engagement in 'drug market' activity such as establishing and maintaining an illicit drug market or drug-defined crimes.

Model 1 is usually applied to violent and disorderly behaviour, most notably in the case of alcohol and stimulants such as amphetamines and cocaine.

Model 2 is usually applied to property crime most notably in the case of heroin and other illicit substances, but not usually alcohol or cannabis.

Model 3 involves two components – offending behaviour associated with a drug market and drug-defined crimes. The former is not of relevance to estimating attributable fractions as this requires a causal component (see Pernanen et al., 2002, p. 82 for more detailed discussion). For the latter, drug-defined crimes can theoretically be attributed a fraction of 100 percent on the basis that the crime would not have occurred if the activity had been defined as legal. However, even where drugs are legally available to adults, such as alcohol and tobacco, there continues to be illegal activity. Further, different forms of the substance that are illegal are still trafficked, such as 'chop-chop' tobacco. In this context some offenders will continue to traffic and use illegal drugs so that the legal status of specific drugs will not make any difference to their illegal activity.

There are complications with these ideal types. A person may commit an armed robbery to acquire money for a drug dependency problem, yet armed robbery is classified as a violent offence. Police arrest people for a wide range of infractions of the law that these theories do not cover. For example, driving without a licence or a breach of bail conditions. Determining the extent to which 'crime' is drug-related is complex and requires data at such a level of specificity that it may never be possible to collect on every individual. Until data collection and measurement are advanced in the criminal justice sector it remains necessary to rely on samples and to a large extent self-report data by offenders of their behaviour.

Top of Page

Data sources and key limitations: DUMA and DUCO

The data used in this paper come from the Drug Use Monitoring in Australia (DUMA) (see Mouzos, Smith and Hind, 2006) and the Drug Use Careers of Offenders (DUCO) projects (see Makkai and Payne, 2003; Johnson, 2004). DUMA is a regular quarterly monitoring system that has been operating since 1999. It surveys adult male and female persons brought to selected police stations for arrest purposes. DUCO was a one-off large scale survey of adult male and female offenders. Both collections have limitations. They are voluntary and largely rely on self-report information. However, DUMA has the added benefit of urinalysis testing. Neither data set is national, however both have the largest samples of their respective populations that are available for secondary analysis. DUMA data are available for 2005/06 financial year while DUCO relies on the male survey undertaken in 2002 and the female sample undertaken in 2003.

This paper uses a sample from DUMA of 3,623 adults for the financial year ended July 2006. DUCO relies on 2,135 sentenced adult male inmates in four jurisdictions and 467 sentenced adult female inmates in six jurisdictions.

Measuring offending

There are three measures of offending—the number of offenders, the number of offending episodes and within that episode the number of offences. In any one year an offender may commit more than one offending episode. In addition, the number of charges can vary for the same offending episode depending on the arresting officer (see Makkai et al., 2004). Further, not all charges will go to court, and not all charges that do go to court will result in a guilty verdict. This filtering in the criminal justice system means that administrative and survey data are highly specific to the point in the system where the data are collected.

Charges are coded according to the ABS Australian Standard Offence Classification system. As there are hundreds of offences, for ease of interpretation these have been collapsed into
8 categories—violent, property, drug offences, driving under the influence, traffic offences, disorder, breaches and other offences. A most serious offence hierarchy, which ranges from violent to other offences, has been calculated. For DUMA the decision was made to take into account all arrests reported over the past 12 months, not just the most current offence. Potentially this overcomes the problem of relying on one arrest occasion as a measure of the 'typical' offending profile. However, criminological research has consistently demonstrated that drug-using offenders report higher rates of offending than non-drug using offenders (see Makkai, 2002). As a result, these figures will under-estimate the total volume of crime that is drug-related, unless adjustments for multiple offences are made.

Increasingly police are issuing 'street'-level cautions or notices to appear in court which do not involve bringing people to the police station or watchhouse. As a result the DUMA sample is likely to be skewed towards the more serious crimes. In the DUCO surveys only selected criminal histories were collected, so it is not possible to adjust for offending in the 12 months prior to the current offending episode that resulted in a term of imprisonment.

Top of Page

Measuring intoxication with DUMA data

If a person has used a drug it does not automatically mean they are intoxicated, although clearly use is a prerequisite for intoxication. The DUMA study does not ask detainees if they were intoxicated at the time of arrest; it asks if they had been using any drugs at the time of arrest. Similarly, detainees are asked if they were using alcohol at the time of the arrest but not whether they were intoxicated. Both of these measures are problematic but they are the best available and are used as surrogates for intoxication. In all likelihood they overestimate the level of intoxication, particularly for alcohol.

DUCO did ask offenders if they were intoxicated at the time of committing the most serious offence for which they were incarcerated and this measure is used in the DUCO calculations.

Measuring causation

There is no measure or even approximation for causal behaviour for alcohol in DUMA. In terms of illicit drugs, detainees are asked to indicate in the past 12 months how many of their offences were drug-related. They were specifically told to exclude alcohol. They were presented with five possible responses—all of it, most of it, about half of it, some of it and none of it. Previous experience with asking detainees to provide more detailed information, such as in percentage terms, resulted in unreliable data. The most liberal estimate is taken by assuming that if the detainees indicated some or more of their offending was drug-related they were assumed to be drug-related. DUCO relies upon individual offenders' accounts of why they committed their crimes.

Measuring dependency

Dependency is a clinical term that is difficult to measure outside a clinical setting. Furthermore, there has been relatively little work on validating standard dependency assessment tools amongst police detainees. Dependency has been defined as 'a cluster of physiological, behaviour and cognitive phenomena of variable intensity in which the use of a psychoactive drug (or drugs) takes on a high priority' (Ghodse, 1995: 3). In the previous calculations the DUMA attributions relied on a single item that asked detainees whether they felt they needed or were dependent on [drugs] in the past 12 months.

In 2004, a six item scale to measure alcohol and illegal drug dependency was included. This scale had been developed for use with police detainees in the United States and the questions reflect each of the diagnostic criteria for abuse and dependence defined by the DMS-IV (see Hoffman et al., 2003). If individuals answer yes to three of the six items they are considered to be dependent (Mouzos et al., 2006). This same scale was also replicated in the female DUCO sample and is used to measure dependency. The DUCO male sample relies on offenders' self-reported motivations for their most serious offence. Such motivations may not reflect the general pattern for all offences.

Table 54 provides two pieces of information – the upper and lower estimates for attributions of offending activity to drug dependency and two sets of attributions using a single item versus a scale to measure dependency. Regardless of whether the single item or the scale is used to measure dependency, the distribution of reported attributions remain similar for the upper estimates. However, with the lower bound estimates the dependency scale results in more detainees being assigned to the alcohol category and fewer detainees not attributing any of the offending to drug dependency.

As the dependency scale has been developed and validated for use amongst police detainees, it is probably more appropriate to use this measure. Taking the more conservative measure the lower bound estimates indicate that 25 per cent of detainees were classified as either dependent or intoxicated and attributed some or more of their offending to illicit drug use, 12 per cent were only dependent on alcohol and using alcohol at the time of the arrest and 5 percent were using both alcohol and illicit drugs. Overall 59 per cent of detainees did not fall into of the above categories. For the upper bound estimates the proportion using alcohol almost doubles to 22 per cent and the other estimates are adjusted downwards.

Overall the percentage of detainees that attribute their offending to drugs is 41 per cent for the more conservative lower estimates and 51 per cent for the more liberal upper estimate.

Top of Page

Table 54, Comparing the single item and new dependency measure on self-reported attributions for police detainees, percentages
Lower single item (per cent)
Upper single item (per cent)
Lower dependency scale (per cent)
Upper dependency scale (per cent)
Illicit drugs
24
20
25
22
Alcohol
6
22
12
22
Both
2
7
5
7
None of the above
67
51
59
49

Source: AIC, DUMA collection, n= 3,623 [computer file]

The cut-off point on the dependency scale does have an effect on the attributable fractions. If the cut off is increased to 4 then considerably fewer persons are classed as 'dependent'. As a result it is likely that our results, even the lower bound estimates, are liberal estimates.

Top of Page

Developing multiple offence-adjusted attributable fractions

Essentially the same rationale as outlined in the earlier publication was used to calculate the attributable fractions. For the details readers are referred to that work. As part of this project, the earlier method with the new data was replicated and found to be highly comparable. However, using the DUMA data, a methodological change has been made through the development of parity-adjusted attributable fractions for the estimates provided in this report. The previous method attributed illicit and licit substances to the reported most serious offence (MSO). In reality, a substantial proportion of persons are charged with multiple offences. Apportioning the attribution to the MSO only in this instance biases the true level of attribution. The new method used here entails calculating a set of unique or mutually exclusive parity measures for each of the MSO offences over the last 12 months.

These are then combined to take into account parities across the offence categories producing multiple offence-adjusted attributions. Effectively, attributions for violent crime (the most serious offence category) remain unchanged but all the other attributions are adjusted by parities for lower order offending episodes.

Table 55 provides multiple offence-adjusted attributions by crime type. It is important to note that these attributable fractions are not for individual offenders. The 95 per cent binomial confidence intervals are included so that upper and lower bound estimates can be calculated. The range of the estimates can vary by as much as 11 percentage points for traffic offenders to a low of five percentage points for a number of the crime types.

In the 2002 calculations a theoretical decision was taken to attribute 100 per cent of drug and drink driving behaviour to substance use. In these calculations we have estimated attributions for crime type based on what offenders reported, which is consistent with the DUCO calculations.

Overall attribution tends to be highest for illicit drugs only, followed by alcohol only and then both. When the fractions are adjusted for levels of dependency, there are noticeable differences for the more liberal estimates, with alcohol being the highest category for violent, drink driving and disorder crime types.

Top of Page

Table 55, Self-reported causal attributions by crime type, multiple offence-adjusted attributions, DUMA, percentages


Violent
Property
Drug
Drink driving
Traffic
Disorder
Breaches
Other
Total
(n)
(1128)
(1050)
(255)
(202)
(352)
(177)
(325)
(99)
(3623)
Upper bound estimates
Illicit drugs
21
36
43
9
25
23
23
26
27
Alcohol
25
11
9
51
17
31
23
23
20
Both
11
10
13
9
6
10
8
10
10
(Any)
56
57
65
69
48
64
54
59
57
Bin 95% C.I.
54, 59
54, 59
61, 68
64, 74
45, 51
60, 68
51, 57
56, 62
n.a.
No substance
44
44
35
31
52
36
46
41
43
Lower bound estimates
Illicit drugs
24
40
49
12
27
25
26
30
30
Alcohol
15
6
5
24
8
20
13
13
11
Both
7
6
7
6
4
8
6
6
6
(Any)
47
52
60
43
39
53
44
49
47
Bin 95% C.I.
44, 50
49, 54
57, 64
37, 48
36, 43
49, 57
42, 47
46, 53
n.a.
No substance
53
48
40
57
61
47
56
51
53

Source: AIC, DUMA collection, n= 3,623 [computer file].

Top of Page

Weighting DUCO

Neither DUCO nor DUMA were nationally representative samples although they may indeed be representative. In the case of DUCO the sample excludes South Australian and Victorian males and New South Wales males and females. Given that DUMA shows that there is considerable variation in drug markets between NSW and the other jurisdictions we have provided both unweighted and weighted DUCO data to reflect the age, gender and state profile to the ABS 2005 prison census data. Due to different variable definitions and low cell sizes, it was not possible to implement a more comprehensive procedure. For DUCO males the fractions were adjusted by using census data to weight the attributable fractions within offence cells by each age group and state. For DUCO females, an overall distribution within each age category and state was used because of the small cell sizes for many 'less serious' offences.

Table 56 provides the weighted estimates and Table 57 the unweighted estimates. For males, the unweighted calculations are those developed by Williams (2005). For females the calculations use the dependency scale to determine whether the person was addicted to drugs at the time of committing the MSO and then adjusts the proportion by whether or not the offence was committed because of drug dependency. These figures are slightly different from those produced by Johnson (2004) as the measures adjust for both the first and second charges reported by the offender1. There is effectively little difference in the unweighted and the weighted attributions. Because of the weighting, the confidence intervals are not calculated.

It is important to note that sample sizes for minor offending categories become very small. This reflects the nature of the criminal justice population—incarcerated offenders are usually sentenced to prison for serious offences, although their offending history will frequently contain a plethora of sentences for minor offences. As is consistent with the criminological literature, females report higher rates of drug dependency and attribution.

Top of Page

Table 56, DUCO males and females, weighted estimates, percentages

Violent
Property
Drug
Traffic (a)
Breaches (a)
Disorder (a)
DUI (a)
Other (a)
Total
Females
High only
27
44
38
12
24
0
100
0
32
Drunk only
15
2
0
3
15
44
0
40
9
High and drunk
2
1
0
4
0
0
0
9
2
(Any)
44
46
38
19
39
44
100
49
43
No substance
56
54
62
81
61
56
0
51
57
Males
High only
11
23
26
8
15
7
0
12
14
Drunk only
11
4
1
13
13
13
10
11
9
High and drunk
12
9
4
6
10
6
14
15
11
(Any)
34
37
31
27
38
26
23
38
34
No substance
66
63
69
73
62
74
77
62
66

Notes: (a) For females based on less than 50 observations. Treat with caution.
Source: AIC, DUCO Male 2002 (n=2,135) and Female 2003 (n=467), unweighted data [computer file].


Top of Page

Table 57, DUCO males and females, unweighted estimates, percentages

Violent
Property
Drug
Traffic (a)
Breaches (a)
Disorder (a)
DUI (a)
Other (a)
Total
Females
High only
25
41
36
13
21
0
100
0
31
Drunk only
15
2
0
4
14
33
0
40
9
High and drunk
2
1
0
4
0
0
0
10
2
(Any)
42
43
36
21
36
33
100
50
41
CI 950
35, 49
36, 51
24, 49
4, 38
10, 62
n.a
n.a
17, 83
36, 45
No substance
58
57
64
79
64
67
0
50
59
(n)
(196)
(161)
(58)
(24)
(14)
(3)
(1)
(10)
(467)
Males (b)
High only
11
23
26
8
15
6
0
16
14
Drunk only
11
4
1
13
13
13
10
11
9
High and drunk
13
9
4
7
11
6
14
17
11
(Any)
35
37
31
28
39
25
24
45
35
CI 95
32,37
32,42
24,39
20,36
28,49
4,46
11,36
34,55
33,37
No substance
66
63
69
72
61
75
77
55
66

Source: AIC, DUCO Male 2002 (n=2,135) and Female 2003 (n=467), unweighted data [computer file].
Notes: (a) For females based on less than 50 observations. Treat with caution. (b) specific (n) not available from publication.


Top of Page

Conclusion

This chapter has updated earlier work that attempted to calculate the proportion of adult detainees' and prisoners' offending that could be 'causally' linked to illicit drugs and alcohol. The limitations of the data have been identified throughout this Appendix and should be kept in mind when using or applying these estimates to the whole of the offender population. Four key innovations have been used in an attempt to improve estimations. These are:
  1. A scale rather than a single item has been used to measure dependency in the DUMA and DUCO female samples.
  2. The DUMA data have used a technique to adjust the attributable fractions so that they take into account multiple offences. Previously, the attribution had been applied to the most serious offence (MSO) only.
  3. 95 per cent confidence intervals have been produced for the DUMA data.
  4. Weighted estimates for DUCO males and females have been provided. These take into account the age, sex and state profile of prisoners based on the ABS 2005 Prisoner Census.
Overall, the estimates range from 41 per cent to 51 per cent for police detainees with variation occurring for different offence types, when classified by most serious offence. This is consistent with the DUMA estimates derived in 2002. When adjustments are made for multiple offences, the estimates increase slightly from a range of 47 to 57 per cent. For DUCO females 43 per cent are estimated to have committed their most serious offence because of substance abuse and for males the estimate is 34 per cent. There is far less variability across offending types than occurs for the DUMA sample.

Top of Page

References

Carcach, C. (1997), 'Reporting crime to the police', Trends and Issues in Crime and Criminal Justice, no 68, Australian Institute of Criminology, Canberra.

Carcach, C. and Makkai, T. (2002), Review of Victoria Police Crime Statistics, Report to the Chief Commissioner, Victoria Police, October.

Committee on Data and Research for Policy on Illegal Drugs (2001), Informing America's Policy on Illegal Drugs: What We Don't Know Keeps Hurting Us, Committee on Law and Justice and

Committee on National Statistics, National Research Council.

Dalrymple, Theodore (2006), Romancing opiates: pharmacological lies and the addiction bureaucracy, New York: Encounter Books.

Davies, John Booth (1992), The Myth of Addiction, Harwood Academic Publishers.

Ghodse, Hamid (1995), Drugs and Addictive Behaviour: A Guide to Treatment, second edition, University Press, Cambridge.

Harrison, L. and Hughes, A. (eds.) (1997), The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates, US Department of Health and Human Services,

National Institutes of Health, NIDA Research Monograph 167.

Hoffman, M., Hunt, D., Rhodes, W., and Riley, J. (2003), 'UNCOPE: A brief substance dependence screen for use with arrestees', Journal of Drug Issues, vol.33, 29-44.

Johnson, Holly (2004), 'Drugs and Crime: A study of incarcerated female offenders', Research and Public Policy Series, No 63, Canberra: Australian Institute of Criminology.

Makkai, T. and Payne, J. (2003), 'Drugs and Crime: A study of incarcerated male offenders' Research and Public Policy Series, No 52, Canberra: Australian Institute of Criminology.

Makkai, Toni and McGregor, Kiah, (2002), 'Drugs and crime: calculating attributable fractions from the DUMA project' in David Collins and Helen Lapsley (2002), Counting the Cost: estimates of the social costs of drug abuse in Australia in 1998–9, Canberra: Commonwealth Department of Health and Ageing.

Makkai, T., Radcliffe, J., Veraar, K., and Collins, L. (2004), 'ACT Recidivist Offenders', Research and Public Policy Series, 54, Canberra: Australian Institute of Criminology.

Makkai, Toni (1999), 'Linking Drugs and Criminal Activity: Developing an Integrated Monitoring System', Trends & Issues in Crime and Criminal Justice, No. 109, Canberra: Australian Institute of Criminology.

Makkai, Toni (2002), 'Illicit Drugs and Crime', in Graycar, Adam and Grabosky, Peter (eds.) (2002), Cambridge Handbook of Australian Criminology, Cambridge University Press.

McGregor, K. and Makkai, T. (2003),'Self-reported drug use: how prevalent is under-reporting?', Trends and Issues in Crime and Justice, No. 260, Canberra: Australian Institute of Criminology.

Mouzos, Jenny; Smith, Lance; and Hind, Natalie (2006), 'Drug Use Monitoring in Australia (DUMA): 2005 Annual Report on Drug Use Among Police Detainees', Research and Public Policy Series, No 70, Canberra: Australian Institute of Criminology.

Pernanen, K; Cousineau, M.; Brochu, S.; and Sun, F. (2002), Proportions of crimes associated with alcohol and other drugs in Canada, Toronto: Canadian Centre on Substance Abuse.

Posyer, Carmel; Makkai, Toni; Norman, Louise; and Mills, Leesa (2002), Drug Driving Among Police Detainees in Three States of Australia, Report to the National Drug Law Enforcement Research Fund, April.

Williams, Paul (2002), 'Aetiological fraction estimates of drug-related crime', in David Collins and Helen Lapsley (2002), Counting the cost: estimates of the social costs of drug abuse in Australia in 1998–9, Canberra: Commonwealth Department of Health and Ageing.

Top of Page

prev pageTOC |next page