|
K. PERNANEN
National Institute for Alcohol and Drug Research, Oslo, Norway and Uppsala University, Uppsala, Sweden
S. BROCHU
International Centre for Comparative Criminology, University of Montreal, Montreal, Canada
M.-M. COUSINEAU
International Centre for Comparative Criminology, University of Montreal, Montreal, Canada
L.-G. COURNOYER
Université du Québec à Hull, Hull, Canada
FU SUN
International Centre for Comparative Criminology, University of Montreal, Montreal, Canada
Abstract
Introduction
Conceptual background
Individual-level versus aggregate-level data
The empirical studies in the research programme
The computerized Lifestyle Assessment Instrument and the Pathway studies
Constructing the estimates
Internal consistency checks
Questioning the assumptions
Discussion
Bibliography
A research programme in Canada is aimed at estimating attributable fractions for the use of alcohol and illicit drugs in relation to crime. Analyses from two studies of new inmates in federal penitentiaries are presented, the first based on a computer-driven questionnaire completed by 8,598 inmates and the second on interviews with 477 inmates. One method used in the estimation combined the following three models linking psychoactive substances to criminal behaviour: the intoxication model, the economic model and the systemic model. Data pertaining to the first two models were used to illustrate this method. Consistency checks showed that crime events attributed to illicit drugs or alcohol were concordant with the inmate being addicted to a substance, and with the inmates' overall assessments of drugs or alcohol on their criminality. Issues discussed include validity, the extent to which findings can be generalized and the advantages and drawbacks of basing attributable fraction estimates on data from self-reports on individual crime events.
The present article outlines the steps taken to arrive at estimates of attributable fractions for drugs and alcohol in relation to crime in Canada. In particular, the article looks at a method that will enable estimates to be made for important subgroups of the population and different types of drugs and types of crimes.
Some of the empirical studies needed for making overall estimates are still at the fieldwork stage, and therefore no final estimates can be made. However, the following is presented: (a) the method; (b) its conceptual background; (c) some findings that will lead up to the estimates; and (d) some methods used for checking the robustness (the internal consistency) of the estimates. Some preliminary calculations that give an indication of the range of attributable fractions for federal inmates in Canada are also provided.
The empirical material available for the estimates in the present article is limited to information on inmates in federal prisons in Canada. In all probability, that population incurs costs to society far beyond its numerical size. For a full cost accounting, however, data on drug and alcohol use and criminality patterns in other populations is also needed, in particular with regard to individuals arrested for a variety of crimes and inmates in provincial prisons, who generally have committed less serious crimes than federal inmates have.
Any estimation procedure for the causal contribution of psychoactive substances on crime is based on key conceptual assumptions. These assumptions are based on findings from past empirical research in the field. Applying the conceptual frames to social reality requires data that allow assessments regarding how applicable alternative causal models are. Because such conceptually relevant data have been missing, researchers have sometimes refrained from making any estimates on the crime component in the social costs of alcohol and illicit drugs. This has been the case for recent social cost estimates in Canada. In other studies, estimates have been based on questionable conceptual and empirical assumptions.
The starting point for the present calculations is to use the following three models, which assign different causal roles to drugs and alcohol in relation to crime: (a) the pharmacological or intoxication model; (b) the economic means model; and (c) the illegal system model. The models used are a tripartite collection that Goldstein used for classifying drug-related violence. In addition, in the fourth model in the series, some crimes are alcohol-related or drug-related by legal definition. However, they are solidly based on empirical findings from research on a variety of crimes.
The intoxication model attributes a direct causal role to a substance used at the time of a crime. The assumption is that intoxication made, or helped make, the individual commit an illegal act that he or she would not otherwise have committed. In the study of the effects of alcohol, this model is often referred to as a "disinhibition" model. It has been used frequently in various estimations of the role of alcohol on crime, specifically in calculating attributable fractions linked to violent crime. Since no other information has been available regarding the causal role of alcohol, it has been assumed that all crimes in which the perpetrator had been drinking were caused by drinking, that is, that if the perpetrator of the violent crime had not been intoxicated at the time, he or she would not have committed the crime. With one important modification, this model is also used as part of the four model conceptual frame for estimating attributable fractions for drugs and alcohol on crime.
The second causal model used, the economic means model (or, in Goldstein's terms, the economic-compulsive model) pertains mainly to the role of drugs and to a much lesser extent to alcohol, as motivators in predominantly acquisitive crimes. Psychoactive substances serve as incentives for a person to commit a crime so that they will get money or other means for acquiring drugs or alcohol.
The third causal model, the systemic model, concerns crimes that were committed, for example, in the course of selling drugs, collecting drug debts and conflicts over drug territory. If it can be assumed that the individual would not have committed these crimes had he or she not been involved in the illegal economy, the crime can be considered to be caused by the drug being present as a commodity within a system of illegal transactions and enforcement methods.
The fourth, the substance-defined model, is not a causal one, but represents a tautological connection with alcohol and drug use. The crimes in this category are included on the basis of laws regulating alcohol and drugs in society. Drinking and driving is by far the most common of the alcohol-defined crimes. Several drug-related offences, such as the manufacture, smuggling and trafficking of drugs, are included in the category of drug-defined crimes. Possession and use of most illicit drugs are also defined as criminal acts in many countries and would be covered by this model. (The present estimates are limited to the population of federal inmates, therefore few cases of minor drug crimes will appear in the estimates.) These will be added as the fourth stage of the construction of the composite attributable fraction model.
Some overlap can be expected between the positive cases in the four models in any population. A certain proportion of individuals who committed a crime under drug intoxication were also driven by the motive to get more drugs for personal use, for instance (so as to prevent their supply from running out). In a similar way, some individuals who used violence to collect a drug debt for themselves or for someone else in the distribution chain did so in order to get drugs or the means to buy drugs for personal use. If there is a great deal of overlap between the positive cases from the three models, it can be inferred that any one of the models would have provided a good approximation of the attributable fraction. More importantly, however, a great deal of overlap is also an internal confirmation of the validity of the combined measure.
As shown below, it is possible with the right kind of data to use aggregate-level data to calculate attributable fractions for some factors on crime. Time series analyses have been used for this purpose. However, they require relatively valid data that has been collected over lengthy periods of time. Such data exist for alcohol but do not exist for illegal drugs. For illicit drugs, aggregate-level analyses cannot therefore be used for calculating attributable fractions.
Aggregate-level data have certain weaknesses for the purpose of causal attribution. They are dependent on the availability of data on other potential causative factors to be used as controls. Even for alcohol consumption, no statistical series exist on alcohol use in different subgroups. Using this method, it is therefore not possible to calculate attributable fractions for alcohol on, for example, crimes committed by those under the age of 30 and over, or for men versus women.
Using individual-level data for estimation makes it possible to distinguish individual cases from non-cases on attributable fraction variables. This means that attributable fractions can very easily be arrived at, for example, for different types of crimes or for different subgroups of offenders. Given a large enough sample of perpetrators or crime events, estimates can be obtained of what proportion of violent crimes among perpetrators under 30 years of age were attributable to alcohol or drugs.
The greatest drawback of individual-level data is that they must be collected by special studies. In addition, much of the information has to be based on self-reports, and it is known that self-reports may be unreliable. The questions generally concern past behaviour with a risk for memory lapses. In the case of sensitive information, social desirability may affect the validity of the information given.
The present data are based on information about inmates in federal penitentiaries. All information (all the sample units) represents cases on the dependent variable: there are no individuals in the sample who did not commit a crime and who could serve as controls for analyses assigning an explanatory value to independent variables, such as alcohol or drug use, or risk calculations linked to alcohol and drugs. This places restrictions on the type of analyses that can be made.
The ideal type of study, which would allow a flawless estimation of attributable fractions, may be easily specified in theory but, at the present time at least, is impossible to conduct in practice. The choice is between methods that are all lacking in some respects. The choice depends primarily on the data available for the purpose and the modelling preferences of the researchers.
Two studies are used for estimation in the present article: the Computerized Lifestyle Assessment Instrument (CLAI) data made available by the Correctional Service of Canada (CSC) and interviews with 477 male inmates in federal prisons in Ontario and Quebec (the Pathway study). The latter study was specifically conducted for the estimations of the project. In some respects, the studies complement each other and in other respects they support alternative estimation methods. If the alternative estimates are within a tolerable range of differences, confidence in their robustness is strengthened.
Data collection has been completed in a third study relevant to assigning an attributable fraction for alcohol and illicit drugs in relation to crime in Canada: information on arrests made in 24 locations in Canada during a one-month period (1-31 May 2000). These data were collected by police officers and are based on information available at the time of the arrest. Most sites were selected on the basis of a stratification of communities in Canada according to population size: (a) two megacities (with populations of over 1 million); (b) three large cities (with populations of between 500,000 and 1 million); (c) three medium-sized cities (with populations of between 250,000 and 500,000); (d) six small cities (with populations of between 100,000 and 250,000); and (e) ten other communities of interest. Within categories (b), (c) and (d), sites have been selected on the basis of statistical information on their overall crime rate: one characterized by a relatively high crime rate, one with a medium rate and one with a low crime rate. (Such stratification was not possible in the "megacity" category.) Data from the study of arrestees add information on another population of relevance to the connections between drugs, alcohol and crime.
Three smaller-scale studies were later added to the research programme for the purpose of extending descriptive and causal objectives to other key populations: interviews with 100 male inmates in provincial prisons; interviews with 100 male provincial probationers; and interviews with 100 female inmates in provincial prisons.
Information on the crime and substance use patterns of provincial inmates and probationers is important because criminality differs between these two populations and the federal inmates who are the subject of the present article. Convicted criminals with a sentence of at least two years' imprisonment serve their sentences in federal penitentiaries, those with a lesser prison sentence serve their time in provincial custody.
The Computerized Lifestyle Assessment Instrument (CLAI) is both a diagnostic tool and a survey instrument used by the Correctional Service of Canada. It is administered to all federal inmates upon admission to an assessment centre, prior to their being sent to an institution. CLAI helps in taking into account treatment and other individual needs of the inmate. There are detailed questions on alcohol and drug use and criminal activities. Such details make the database uniquely suitable for some of the purposes of the present article.
The data are collected by means of a computer-driven questionnaire: the inmates enter responses to questions that appear on a computer screen. It takes an average of two hours to fill out the questionnaire.
The data collection on the CLAI started in 1990. Over the years, an increasing number of penitentiaries have contributed information to the computer file. In the file to which the authors have access, there is information on close to 17,000 inmates. The best geographical coverage is for the period 1993-1995 and has been selected for the present analyses (N=8598). However, the differences between the total file and the subfile are generally negligible.
In this study, 477 inmates were interviewed at regional reception centres in Ontario and Quebec. The data were collected between September 1999 and January 2000 in Ontario and between February and December 1999 in Quebec. The most central data collection instrument of the study was a calendar used in charting the 36 most recent months in the inmate's life prior to arrest. The focus was on several aspects of the relationship between drug and alcohol use and criminal behaviour. Much of this information was not available in the CLAI data.
The Pathway study also incorporated central questions from the CLAI on the inmate's drug and alcohol use and criminality. It included the same tests for dependence on alcohol (the Alcohol Dependence Scale) and drugs (the Drug Addiction Severity Test), enabling an aggregate-level reliability check of estimates made from the two studies.
A major difference between the studies is that the CLAI drew its population from all five regions of the Correctional Service in Canada (Atlantic, Ontario, Pacific, Prairie and Quebec), while the Pathway study was carried out on inmates in the Ontario and Quebec regions only. The decision to limit that study to the two regions was made for financial reasons. Analyses will be conducted to determine what effect the difference in geographical base of the inmate populations may have for the results and how any biases can be corrected.
The time periods of data collection also differ. As was pointed out above, information from 1993-1995 has been used to arrive at the CLAI estimates in the present article. The Pathway data, on the other hand, were collected an average of five to six years later. Analyses of the CLAI data from the Quebec region (which has been providing data regularly since the early 1990s) showed very small differences in drug-use patterns among inmates over a period of six years.
The selection procedure for federal inmates in the two studies was about the same, although the CLAI study was a census (with some attrition), while the Pathway study used a random sampling procedure to select incoming inmates to the study. A sampling procedure was necessary because the inflow of new prisoners was greater than what could be handled by two interviewers, which was the maximum number possible for logistical reasons. In both studies, the inmates participated in the interview about two weeks into their stay at the reception centres.
Reasons for attrition in the Pathway interviews at the Ontario and Quebec reception centres are shown in table 1. The response rate in the Ontario part of the study, calculated out of those who were contacted for an interview, was 84.8 per cent. The corresponding figure for the Quebec part of the study was 78.9 per cent.
Table 1. Pathway study: sample attrition and response rates in Ontario and Quebec |
||
Response rates |
||
Subjects | Ontario | Quebec |
Total sampled | 342 | 419 |
Not eligible | 10 | 44 |
Not available | 35 | 90 |
Approached | 297 | 285 |
Refused | 44 | 56 |
Interrupted | 1 | 4 |
Interviewed | 252 | 225 |
Response rate (percentage ) | 84.8 | 78.9 |
As was mentioned above, the main data collection instrument of the Pathway study is a 36-month calendar onto which a monthly record of drug-use patterns and criminality was recorded. Whereas the CLAI instrument only asked for crime-specific information on alcohol and drug use in connection with the most serious crime on the present sentence, the calendar instrument asked for this information on all self-reported crimes over a three-year period. Many of these crimes have remained undetected by authorities. The great number of crime episodes and the longer reference period, among several other features, adds to the power of the analyses that can be performed on the Pathway data.
While the CLAI data were entered by the inmate in response to questions and response alternatives appearing on a computer screen, the calendar data were filled in by the interviewer while she (all research assistants on the study were women), together with the respondent, consulted the calendar in order to place occurrences in the correct time period on the calendar. Other data collection instruments in the Pathway study were filled out as a write-in questionnaire by the inmate, although he was allowed to ask questions on the meaning of questions, for example. Most of the interview session was spent on the calendar.
Comparisons between the CLAI and the Pathway studies must take into account that they pertain to somewhat different geographical areas and to time periods that are five to six years apart on average. In the CLAI data from the period 1993-1995, 37 per cent of the new inmates had been admitted in the Ontario region and 40 per cent in the Quebec region. Thus, approximately 23 per cent of inmates in the CLAI data file are from outside the two regions. A correction factor will be used in generalizing findings from the Pathway study to the total federal inmate population in Canada. 1
Three major limitations remain, even with the corrections outlined above:
(a) All the estimates presented in the present article pertain to inmates in federal prisons in Canada only. Generalizations to similar populations should be made with great caution. Separate studies will be carried out on provincial prisoners and individuals who have been arrested for a crime (as was mentioned above);
(b) All the estimates presented here pertain to male offenders only. The Pathway study was carried out on male inmates exclusively. The Correctional Service of Canada collects CLAI data on both male and female inmates. However, the CLAI file to which the authors have access currently does not contain data on female prisoners. Female inmates make up only about 2 per cent of the federal inmate population, and special characteristics of this group would have a negligible effect on the overall estimates pertaining to the federal inmate population;
(c) All the findings in the CLAI and the Pathway studies were based on information provided by the inmates themselves. Giving of false information cannot be ruled out, and memory lapses may also play a part. This, of course, is a shortcoming common to much of the social-survey-type research in sensitive areas of study.
Some of the questions in the CLAI study were used in the Pathway study, partly for the purpose of replicating the questions with a different data collection method.
Crime-specific information on drugs and alcohol use was available for the most serious crime on the inmate's current sentence. The proportions of inmates who were under the influence of a substance when committing such a crime are as follows:
Associative fractions from intoxication model |
||
CSC-CLAI | Pathway-CLAI | |
Drugs | 0.16 | 0.20 |
Alcohol | 0.21 | 0.19 |
Drugs and alcohol | 0.13 | 0.14 |
No substance | 0.50 | 0.47 |
Total | 1.00 | 1.00 |
These figures correspond to the one-model estimates of attributable fractions sometimes used in social cost calculations. If a restricted focus is maintained on the attributional fraction for alcohol on crime, an attributional fraction of 0.34 (0.21+0.13) would be arrived at, on the basis of the CSC study, and 0.33 on the basis of the Pathway study. For drugs, the fractions would be 0.29 and 0.34 respectively. It should be noted, however, that a large portion of the cases for drugs and alcohol overlap. The main difference in the estimates from the two studies is a 25 per cent higher attributable fraction for drugs in the Pathway study.
According to the CSC study, cocaine had been used prior to the crime by 9 per cent of the inmates, cannabis by 3 per cent and heroin by 2 per cent. In addition, cocaine in combination with alcohol had been used prior to the crime by 5 per cent, cannabis with alcohol by 3 per cent and other drugs with alcohol by 5 per cent of the inmates. (Heroin had not been used with alcohol.) These figures will later be used in constructing four-model attributable fractions for cocaine, cannabis and heroin.
Among the individuals who were under the influence of any of these substances, the following proportions of inmates said that they would not have committed the crime if they had not been under the influence:
Intoxication crimes attributed to alcohol and drugs by the perpetrators
|
||
CSC-CLAI | Pathway-CLAI | |
Drugs | 77 | 66 |
Alcohol | 79 | 70 |
Drugs and alcohol | 86 | 74 |
By multiplication, the part of the attributable fraction contributed by the intoxication model is obtained as shown by the following figures:
Corrected associative fractions from intoxication model |
||
CSC-CLAI | Pathway-CLAI | |
Drugs | 0.13 | 0.14 |
Alcohol | 0.16 | 0.13 |
Drugs and alcohol | 0.11 | 0.10 |
No substance | 0.60 | 0.63 |
Total | 1.00 | 1.00 |
These figures pertain to the most serious crime on the inmates' current sentence. They will later be adjusted based on available information regarding all the crimes on the inmates' current sentence.
The comparison between the two sets of estimates gives an indication of the robustness of the estimates in the face of different data collection methods (computer-driven questionnaire in the CLAI versus personal interview situation in the Pathway study). The estimates from the two studies are fairly similar, considering that they are based on partly different regional populations and different time periods.
In response to questions on the role of alcohol and drugs as motivators for the most serious crime, such as "Was this crime committed to get or while trying to get alcohol/drugs for your own personal use?", the following proportions of inmates reported that such was the case:
Associative fractions from economic model |
||
CSC-CLAI | Pathway-CLAI | |
Drugs | 0.12 | 0.15 |
Alcohol | 0.03 | 0.03 |
Drugs and alcohol | 0.06 | 0.05 |
No substance | 0.79 | 0.77 |
Total | 1.00 | 1.00 |
The data show that illicit drugs are greater motivators for crime than is alcohol, as expected: 12 per cent of the inmates in the CSC study and 15 per cent in the Pathway study stated that they had committed the most serious crime on the current sentence in order to get drugs for their personal use, while the percentage for alcohol was 3 per cent. Combined use as a cause of crime was evident also in this context, with 6 per cent and 5 per cent of inmates stating that the crime was committed in order to obtain both alcohol and drugs for personal use. Expressed as a total share, alcohol was at least a partial motivator in 9 per cent in the CSC study (3 per cent + 6 per cent) (compared with 8 per cent in the Pathway study) of the most serious crimes committed, while this was true for double that share (6 per cent + 12 per cent) (compared with 20 per cent in the Pathway study) in the case of illicit drugs. There is considerable agreement between the estimates from the two studies, which adds confidence to the reliability of the estimates.
With a starting point in the intoxication model, the importance of the economic factor will depend on the number of cases it adds to those already identified by the former model. The findings presented in table 2 indicate that the additional contribution of the economic factor is rather modest.
Table 2. Perpetrators who committed the crime in order to get alcohol or drugs according to whether they were on alcohol or drugs at the time of the crime (CSC-CLAI)
|
||||
Subjects | On alcohol |
On alcohol
and drugs |
On drugs | On neither |
To get alcohol | 10.6 | 3.6 | 0.3 | 0.2 |
To get drugs | 1.8 | 12.8 | 56.1 | 1.9 |
To get drugs and alcohol | 6.3 | 25.4 | 9.1 | 0.7 |
To get neither | 81.4 | 58.3 | 34.5 | 97.2 |
Total | 100.0 | 100.0 | 100.0 | 100.0 |
Combining the two models, the corrected intoxication model and the economic model gives the following estimates from the two sets of CLAI data:
Associative fractions from combined intoxication-economic model |
||
CSC-CLAI | Pathway-CLAI | |
Drugs | 0.13 | 0.17 |
Alcohol | 0.15 | 0.13 |
Drugs and alcohol | 0.14 | 0.12 |
No substance | 0.58 | 0.58 |
Total | 1.00 | 1.00 |
The estimates based on the CSC study do not differ greatly from the estimates from the corrected intoxication model: the fraction for alcohol is 0.01 lower and that for the combination of drugs and alcohol is 0.03 higher. The differences in the Pathway data are approximately of the same magnitude.
Calculations show that 93 per cent of cases in the combined model from the CSC study were already included in the corrected intoxication model. The number of drug cases rose by 7 per cent compared with the corrected intoxication model, while 4 per cent were added to alcohol cases and 1 per cent to the combined category of drugs and alcohol.
The information for this segment of the conceptual model was available in the calendar part of the Pathway study. These analyses were not available at the time of writing. Including cases from the systemic model will not change the attributable fraction for alcohol from the two-model estimate because almost all cases in the systemic model will be attributable to drugs. Neither is it likely that the attributable fraction for the combined category of drugs and alcohol will change to any notable degree from the two-model estimate.
In the last stage of the model combinations, drug offences will be added as attributable cases to the drug category, and drinking and driving infractions as cases to the alcohol category.
Consistency and reliability checks were especially needed in material that was based entirely on self-reports.
One internal consistency check relates the attributable fraction status of the crime to the addiction status of the inmate based on two validated and widely used dependency scales: the Alcohol Dependence Scale (ADS) and the Drug Addiction Severity Test (DAST). If there is a strong positive relationship, that is, if the cases on the attributable fraction variable to a large extent overlap with the perpetrator being defined as an addict, this is an additional indicator of the construct validity of the attributable fraction and it will increase confidence in the measure. Consistency check number one is demonstrated in the following figures:
Consistency check between two-model attribution of crime to drugs and alcohol and addiction status of perpetrator (CSC-CLAI)
|
||||
Crime attributable to |
||||
Inmate addiction | Drugs | Alcohol |
Alcohol
and drugs |
No substance |
Drugs | 76 | 9 | 65 | 13 |
Alcohol | 5 | 35 | 42 | 5 |
The figures show that 5 per cent of inmates whose most serious crime was attributable to drugs according to the two-model estimate were addicted to alcohol, while 76 per cent were addicted to drugs. A high proportion of drug-addicted individuals was also found among those whose crime was attributed to the combination of alcohol and drugs. Similarly, 35 per cent of those whose crime was attributable to alcohol were addicted to alcohol and 9 per cent to drugs. Addiction to drugs and to alcohol was relatively high among those whose crime was attributed to both substances. Among those perpetrators whose crime was not attributable to any substance, only 5 per cent were addicted to alcohol and 13 per cent to drugs, both proportions well below the level of addicted inmates in the total population (12 per cent and 29 per cent).
The inmates were asked to rate the effect that alcohol and drugs had on their involvement in crime by answering the question "What do you feel has been the overall effect of your drug use (alcohol use) on your involvement in crime?". This has been used as another internal consistency check, as demonstrated in the following figures:
Consistency check between two-model attribution of crime and the perpetrators' assessment of the influence of drugs and alcohol on their criminality (data from CSC-CLAI)
|
||||
Crime attributable to |
||||
Perpetrators | Drugs | Alcohol | Alcohol and drugs | No substance |
Who felt that drugs had increased involvement in crime | 85 | 28 | 77 | 21 |
Who felt that alcohol had increased involvement in crime | 17 | 76 | 75 | 10 |
More than four fifths of those offenders whose most serious crime was attributable to drugs felt that their drug use generally had increased their involvement in crime. In comparison, less than one fifth felt that alcohol had increased their involvement. The same substance-specific pattern was evident with regard to the inmates whose crimes were attributable to alcohol. The proportion of offenders with a crime attributable to drugs who felt that alcohol had increased their involvement in crime (17 per cent) was relatively close to the proportion of offenders whose crime was not attributable to any substance and who felt that alcohol had increased their involvement in crime. The proportion of offenders with a crime attributable to alcohol (28 per cent) was also relatively close to the proportion of offenders whose crime was not attributable to any substance (22 per cent).
A comparison will be made later between two-, three- and four-model estimates as to how well they discriminate on these and other consistency checks. The aim is to obtain some insight into what was gained in accuracy between estimates of attributable fractions from different combinations of models, and whether it is justified to assign the crime or the individual as a case on the attributable fraction variable.
Another indication of the reliability of the measure was mentioned above. It is the extent of overlap in the cases that the different models bring to the attributable fraction. As shown, the cases included on the basis of the economic model greatly overlapped with the cases obtained from the (corrected) intoxication model. On the other hand, 76 per cent of the intoxication cases for the combined category of alcohol and drugs could not be found in the economic model, mainly because there were many more cases from the intoxication model than from the economic model.
Self-reports on the level of intoxication have been used in numerous surveys on alcohol use in general populations around the world. On the whole, the experience is that the data obtained are sufficiently valid. Self-reports on drinking among alcohol abusers are considered to be even more valid. Emergency-room studies also indicate that self-reports yield valid data on alcohol use in connection with the injury, whether from violence or accident. There is also strong evidence that drug-use surveys in general populations of adults, high school students etc. on the whole provide valid data on drug use.
A fair number of studies have been conducted on the validity of self-reports on substance use among drug users and abusers. Out of 54 reports examined by the present authors, 48 assessed that the self-reports were reasonably valid if certain conditions were met in the execution of such studies.
Many studies on violent crimes have used police or court records to ascertain whether the perpetrator and victim had been drinking at the time of the crime. Results from such studies are used to provide attributable fractions in calculating social costs of alcohol. However, there is often no definition of "being under the influence" in such studies, and circumstantial information (such as the crime having occurred inside or outside a bar) is sometimes used to classify a crime as being alcohol-related. In addition, information on alcohol and drug use in police and court records is to a large extent provided by individuals involved in the crime. It is therefore doubtful if studies using official records of crime events provide estimates that are appreciably more valid than data based on self-reports by inmates.
The assumption that the inmate is in a position to assess whether he would have committed the crime had he not been under the influence of alcohol or drugs can be seriously questioned. As discussed in the CSC and Pathways studies, 77 per cent and 66 per cent of those under the influence of drugs, 79 per cent and 70 per cent of those intoxicated from alcohol and 86 per cent and 74 per cent of those intoxicated from both substances stated that they would not have committed their most serious crime had they not been intoxicated. The correction decreases the sizes of the attributable fractions by 14 per cent to 23 per cent in the CSC study, and 26 per cent to 34 per cent in the Pathway study. Despite severe doubts regarding the validity of such judgements, such a correction seems to provide an improvement compared with accepting the intoxication model without any downward adjustments.
In this case, to get drugs or alcohol for personal use is perhaps easier to assess for the actor than the counterfactual scenario of the previous point.
Another indication on the aggregate validity of the responses to the questions asked can be had from a comparison of the CLAI findings and the subset of identical questions that were included in the Pathway study. Such a comparison also, to some extent, serves as a check of the assumptions above, in the sense that it provides an indication as to the robustness of the estimates in the face of varying situational settings: an interview situation with a female research assistant from outside the prison setting versus responding to a computerized questionnaire by means of punches on a keyboard. It has been found that, although the figures from the two studies are not identical, they fall within a relatively narrow range of estimates.
One important conceptual question concerns what to do with the combined category of crime associated with both alcohol and drugs. It is important to ask whether crimes attributable to this combination can be divided up between attributable fractions for alcohol and fractions for drugs. It may not make sense to ask an inmate or arrested person which of the substances was the most important in causing the crime. To some extent, the perpetrator's history of abuse, treatment and societal reactions may provide sufficiently valid information for assigning a prime role to either of the two substances.
Is it, however, necessary to separate the role of drugs and alcohol in the cases where both had been involved in causing the crime? It is a fact that they exist in the crime episodes simultaneously, either as pharmacologically causal, motivating or systemic factors. Perhaps it is time to take this into account in assigning attributable fractions. However, the advisability of using a combined attributable fraction will naturally depend on what purpose it is used for, and the framework of the costing process in particular.
Another question is whether something is being missed by using the four-component method by comparison, for instance, with the analyses of aggregated data that have been used to arrive at attributable fractions for alcohol on crime (almost exclusively violent crime). Different types of time-series analyses are potentially the most powerful methods available for aggregate-level estimates.
In this context, it is worth remembering that it is not at all possible to conduct time-series analyses for illicit drugs. Some type of individual-level data must be used. Information must be collected from individuals who participated in the crime episodes. Observers or informants can be used in some cases, but the most relevant and detailed data can only be provided by the actors themselves, through self-reports. An interesting question is how comparable attributable fractions for alcohol from, for instance, time-series analyses are to attributable fractions based on self-reports of the role of alcohol as a determinant in individual crime events. Such comparisons are possible for a number of countries or other jurisdictions, but for alcohol only.
It is possible that something further may be missed in the four-model method. There may be countries and cultures where additional models should be taken into account, in addition to the intoxication model, the economic and the systemic model and the substance-defined crimes.
There may be simpler ways of collecting data for the method of estimation used in the present article. The study from Canada presented here has six components. Conducting such studies of inmates in prisons and individuals who are arrested requires a sometimes lengthy process for obtaining access to relevant samples. Having several components also requires added effort and supervision, compared with conducting one larger-scale study. There may be a simpler way of getting the same information or sufficient information of some other kind that would make it possible to calculate reasonably valid attributable fractions.
The final aim of the project is to provide attributable fractions for alcohol, drugs, cocaine, cannabis and heroin on crime in Canada. However, for the purpose of calculating the social costs of keeping inmates in federal penitentiaries (a considerable sum), using attributable fractions for federal inmates is probably more accurate than using an overall fraction for all crimes in Canada. In the same way, estimates from the study of individuals arrested by the police may be more (but not exclusively) relevant for policing costs.
Validity issues in general population studies are in some respects identical to studies with inmates. Under-reporting may be more likely in general population studies because there is no incentive to tell the truth, in particular about drug use and other illegal activities. In some cases, such an incentive does exist for inmates if they want to be free of their drug problem or to spend their prison time in a treatment environment which, in many cases, is more pleasant than a standard prison setting. This of course opens up possibilities for over-reporting the role of alcohol and drugs.
The method outlined above provides easily accessible information on the share of crime contributed by different causal processes and their combinations. From the perspective of prevention, it will also be important to know which type of causal process predominates, what the overlaps between the determinant models are and what changes occur over time in these constellations.
Finally, using attributable fractions that are built up from individual-level causal models regarding the effects of alcohol and drug use on crime may bring the study and estimation of attributable fractions closer to the main body of research on drugs, alcohol and crime.
Adrian, M. Social costs of alcohol. Canadian journal of public health (Ottawa) 79: 316-322, 1988.
Drugs, alcohol and crime: patterns among Canadian federal inmates. By S. Brochu and others. Bulletin on narcotics 51:1-2: 57-73, 1999 (United Nations publication).
Goldstein, P. J. The drugs/violence nexus: a tripartite conceptual framework. Journal of drug issues 14:493-506, 1985.
International guidelines for estimating the costs of substance abuse. By E. Single and others. Ottawa, Canadian Centre on Substance Abuse, 1996.
Stinson, F. S., and S. F. De Bakey. Alcohol-related mortality in the United States, 1979-1988. British journal of addiction, 87:5:777-783, 1992.
1 Preliminary findings indicate that the attributable fraction on crime for alcohol and the combined use of drugs and alcohol is considerably higher in western Canada than in Ontario and Quebec, while the fraction for drugs only is very similar between those regions.
|