** School of Psychology, University of New South Wales
A deterrence-based random breath testing (RBT) program has been in operation in the state of New South Wales since December 1982 as a countermeasure to drink-driving. The program is characterised by a high level of visible, state-wide enforcement activity throughout the year and extensive state-wide publicity.
The effectiveness of the program is examined in this study in greater detail and over a longer time period than in previous studies, to give a comprehensive profile of its impact. Accidents are classified into "drinking hour" and "non-drinking hour" periods, based on an analysis of crashes in which driver alcohol concentrations were known. The RBT program was associated with a nett reduction in drivers crashing during "drinking hours" of over 20% which has been maintained for 10 years. The results provide strong evidence that a well-structured RBT program can have substantial long-term road safety benefits.
The introduction of a random breath testing (RBT) program in NSW in December 1982 followed a number of activities and initiatives in the area of drink-driving. Much of this activity served to strengthen public demand for further action and maintained drink-driving on the media, public and political agenda. The program was characterised by intensive, visible enforcement activity on a year round, statewide basis, supported by substantial publicity and media attention. The program was based on deterrence theory (Homel, 1988), by which drink-driving behaviour could be reduced by setting up perceptions of a significant risk of being caught and a significant penalty if caught.
The principal characteristics of the original RBT program have largely been adhered to up to the present time. In addition to the actual enforcement, the Government has continued substantial informational and educational campaigns addressing a range of drink-driving issues. Research demonstrated substantial changes in both drinking and driving behaviour immediately following the introduction of RBT (Elliott and Shanahan, 1983; Homel, 1988); and changes in knowledge, attitudes, and community unacceptability of drink-driving (Carseldine and Cairney, 1989; Homel et al., 1988; Span 1995).
Research conducted on crash statistics in NSW in the first few years of RBT demonstrated a significant reduction in the road toll attributable to RBT (Arthurson, 1985; Homel et al., 1988). A simple assessment of the incidence of alcohol in motorists killed shows a reduction which has been maintained up to the present time. The current study assessed in detail the impact of the RBT program and the extent to which the impact has been maintained.
The RTA's mass crash database contains a comprehensive set of crash variables. Detailed variables are available on crashes from 1976, which provided several years' data prior to the introduction of RBT to develop a reliable baseline measure. Data were analysed up to 1992.
Although the optimal measure for assessing the impact of the RBT program would be the blood alcohol concentration (BAC) of drivers in crashes, adequate information is not available consistently over the time period being measured. Crashes from a period in which alcohol information was known with reasonable accuracy were analysed to define a surrogate measure. The type of crash, as defined by the "road user movement", and the time of the crash were found to be useful. Of these, time and day of a crash have been coded consistently since 1976 and were used to define the outcome measure.
Crashes were divided into "drinking hours" and "non-drinking hours". (Over 20% of drivers who crash during "drinking hours" were drink-drivers, compared with around 2% during "non-drinking hours".) Under the definition, "non-drinking hours" were between 6am and 5pm on weekends and holidays; and between 5am and 7pm on weekdays.
The outcome measure was defined as the ratio between the number of drivers involved in crashes in drinking hours and those in non-drinking hours-this co-efficient represents the "odds" of a driver crashing during drinking hours. This co-efficient has the advantage of setting up non-drinking hour crash involvement as the predictor of drinking hour crash involvement and simplifies the regression equation by controlling for a large number of factors. Crashes were measured on a monthly basis, which allowed the inclusion of a number of explanatory variables which were also collected on a monthly basis.
Crashes were analyses at the level of the "driver". This allowed separation by age and sex, and controlled for changes in the driving population. It also allowed a more useful analysis by looking at changes in crashes in a number of age/sex groups.
Four variables were selected for the main analysis: driver age (under 25, and 25 or older); driver sex; crash location (Sydney region, rest of state); crash severity (injury, towaway). These variables formed 16 exclusive groups with a large enough sample size to ensure reasonable power in the analysis to generate valid conclusions.
Variables were included which were determined to have a potential impact on crashes. All of the variables were tested for a general effect on the dependent measure as well as linear and quadratic trends. The interventions covered drink-driving (e.g., reduction of prescribed concentration of alcohol (PCA) from .08% to .05%) and other road safety interventions (e.g., speed camera program). Specific variables regarding RBT were: the RBT program; the number of breath tests; and type of breath testing (stationary and mobile).
The unemployment rate was included as a major economic activity indicator. Unemployment has been established in studies as having an association with crashes.
Publicity campaigns, although a potentially important variable, was not included primarily because the level of publicity could not be quantified on a monthly basis during the whole period being assessed. "Unpaid" publicity-e.g., news stories-are also important, but also could not be easily measured. Publicity is an integral part of the RBT program, and is implicitly included in the analysis by examining RBT as a "program" effect.
Unlike the driver/crash variables, it is not possible to create exclusive combinations of the intervention variables. For example, there is a period in which the .05% PCA is in force without RBT, a period in which .05% is in force with RBT, but no period in which RBT is in force without .05%. Conclusions can only be drawn about the impact of RBT with .05%.
The approach taken in the research provides a powerful measure of the impact of the various interventions on the outcome measure (i.e., the "odds" of a driver crashing during "drinking hours"). Setting up the driver/crash variables allows examination of specific effects of the intervention variables which may be expected to affect only certain groups. The analysis focussed on providing valid results: to ensure that conclusions drawn from the analysis were well justified.
In the first stage, of the 16 combinations of driver/crash variables were considered separately. For each group, a stepwise multiple regression analysis was used to determine which interventions were significant predictors of the "odds". Only those variables which were significant predictors at the .01 level in at least 3 groups were retained (this decision was made prior to the analysis). Linear and quadratic trends were also assessed.
Next, all 16 driver/crash groups were combined; severity of crash was treated as a repeated measure. A repeated measures ANOVA was used to determine which group and intervention variables (and any interactions) were significant at .001 (this decision was made prior to the analysis). The significant variables and interactions were used to predict the "odds" for time periods before and after the interventions.
Only the reduction in the PCA to .05% and the RBT program were significant predictors. There was no linear or quadratic trends associated with these interventions, meaning that the effect of the interventions occurred around the time of their introduction and stayed constant during the time period relevant to the interventions. On this basis, the "odds" were predicted for the following three time periods: the baseline period (prior to Dec. 1980); the .05% PCA period (Jan 81 - Nov 82); and the RBT period (Jan 83 - Dec 92).
The reduction in the PCA to .05% resulted in a nett reduction in "drinking hour" crash involvement of 14% for men and 5% for women (Table 1).
Nett % Change in Drinking Hour Crash Involvement Following .05% PCA
The introduction of the RBT program was associated with a nett reduction in "drinking hours" crash involvement of 26% for men and 21% for women (Table 2). A significant reduction occurred in each age group, for crashes in and outside Sydney, and different severities of crashes. There was a trend for a greater reduction for the older group, outside Sydney and for more serious crashes.
Nett % Change in "Drinking Hour" Crash Involvement Following Introduction of RBT
|Rest of NSW||-28||-23|
The results of the research provides good evidence for a substantial impact of the RBT program on crashes, and the maintenance of this impact over a ten year period. This achievement is outstanding and places NSW in a unique position.
Interventions such as the .05% PCA and RBT should not, however, be seen as isolated events. Community and political support for such interventions were achieved through a build up of media and public attention to the drink- driving problem. Nevertheless, behaviour change is quite clearly associated with these specific interventions.
The introduction of the RBT program was a major catalyst for behaviour change, and one which was much stronger and more broadly influential than the .05% PCA. The current research supports the notion that the impact of RBT was a "program" effect.
The research provides no evidence that the substantial increase in the level of RBT from the late-1980s or the introduction of mobile RBT resulted in further changes in crashes. It is not, however, possible to ascertain which of these changes, or other efforts in the drink-driving area have been necessary to maintain the observed impact on crashes. Survey research has demonstrated a significant long-term change in attitudes and reported behaviour amongst drivers; a shift towards "general prevention" of drink-driving, in which drink-driving is considered the wrong thing to do; and substantial public opposition to any degree of drink-driving (Cairney and Carseldine, 1989; Homel et al., 1988; Span, 1995).
While the current research suggests that a RBT program can only achieve a certain degree of change in drink-driving crashes, such a program is an important foundation on which to build more comprehensive countermeasure programs and gives the potential to achieve further gains in the longer term. The road safety benefits of the laws and enforcement activity have been achieved with a significant but moderate penalty structure with no emphasis on extreme penalties; and substantial publicity focussing on enforcement activity, the risk of being caught and consequences of being caught.
Arthurson, R., (1985). Evaluation of random breath testing. Traffic Authority of New South Wales, Research Note RN 10/85.
Cairney, P. and Carseldine, D. (1989). Drink-driving and random breath testing: A survey of knowledge, attitudes, beliefs and self-reported behaviour. Traffic Authority of New South Wales, Research Note RN 3/89.
Elliott, B. and Shanahan, P. (1983). The NSW motorist and random breath testing. Bureau of Crime Statistics, and Research, Department of the Attorney General and of Justice.
Homel, R. (1988). Policing and punishing the drinking driver: A study of general and specific deterrence. Springer-Verlag: NY.
Homel, R., Kearns, I. and Carseldine, D. (1988). Drink-driving countermeasures in Australia. Alcohol, Drugs and Driving, 4(2), 113-144.
Span, D. (1995). Research on knowledge, attitudes and reported behaviour on drink-driving in New South Wales. Paper presented at the 13th International Conference on Alcohol, Drugs and Traffic Safety, T'95, Adelaide.
Return to Contents
Return to T'95 Home Page