Sign the Resolution
Contents | Feedback | Search
DRCNet Home | Join DRCNet
DRCNet Library | Schaffer Library | Drugs and Driving
Swedish Road and Transport Research Institute (VTI), S-581 95, Linköping, Sweden
Random Breath Testing was introduced in Sweden in the seventies. The aim of the present study was to evaluate a Random Breath Testing campaign in parts of southern Sweden, which was started during the fall of 1992 and lasted less than a year. The RBT units were made highly visible, and these activities were accompanied by media coverage. Comparisons were made with a control area in southern and middle Sweden. A questionnaire was sent to 1500 randomly selected car owners in each of the two comparison areas on two occasions, before the start of the campaign and a year later.
Routine breath testing as a screening device in traffic law enforcement was introduced in Sweden in 1977. The law allows the police to test drivers involved in accidents, drivers apprehended for violating traffic laws, or at random in preplanned road checks. Before the law the number of breath tests was less than 100.000 per year. In 1980 this number has increased to about one million and in the beginning of the nineties to about 1.5 million.
Random breath testing (RBT) is supposed to work by increasing the drivers' subjective risk of detection. Thereby, the drivers are supposed to be less inclined to drive while intoxicated and the accident figures will go down (Åberg, 1987). Drivers can obtain information about RBT from different sources; they may be subjected to enforcement in road checks, they may observe other drivers being stopped, hear about other persons' experiences, via mass media, etc.
The intentions of the present study were to study effects in terms of subjective risk of detection, self-reported driving after alcohol consumption, attitudes to RBT, and attitudes to driving above the BAC limit. Accident analyses were also to be performed.
A campaign with intensified RBT was started in southern Sweden in the fall of 1992, initiated by the National Society for Road Safety in cooperation with the police authorities. The campaign lasted for slightly less than a year. In line with experiences from Australia (Homel, 1988; Homel et al, 1988), the RBT units were made more visible than is usually the case (the road checks were marked with traffic signs reading POLICE: ALCOHOL CONTROL), and attempts were made to attract the attention of mass media (press, local radio, local TV). However, there was no paid publicity.
A before-and-after design with a control condition was used. In the test area RBT was to be increased substantially whereas no similar increase was expected to take place in the control area.
The test area consisted of the following counties: Kronobergs län (G county), Kalmar län (H county), Blekinge län (K county), Kristianstads län (L county), and Malmöhus län (M county). The control area consisted of Uppsala län (C county), Södermanlands län (D county), Östergötlands län (E county), Skaraborgs län (R county), and Göteborgs-Bohus län (O county).
A questionnaire was sent to 1.500 randomly selected car owners, aged 20 to 70, in each of the two comparison areas on two occasions; one month before the beginning of the campaign and 13 months later. Two reminders were used. The response rate in all four cases was 87-88%.
The content of the questionnaire was: gender, chronological age, driver licence age, driving experience, experiences of RBT, perceived probability of breath test, attitudes, alcohol habits, and driving after alcohol consumption.
An analysis was also performed on accidents with suspected alcohol involvement (according to police reports).
In all analyses a significance level of = .05 was adopted.
In the before measurement some significant differences were found between the two comparison areas (as tested by X2). Thus, a larger percentage of the respondents in the control area stated that they had been breath tested during the last year. They also expressed a more negative opinion about driving above the legal BAC limit, .02 percent. Because of these differences before the campaign, it was decided to study interactions between study area and time period, tested by logistic regression analysis (Everitt, 1992), instead of comparing the two areas before and after the campaign.
Question: "How many times during last year were you Random Breath Tested?" - This question was regarded as a measure of objective risk of detection. The percentage of subjects who stated that they had been breath tested increased from 12.8% to 21.3% in the test area, compared to an increase from 17.3% to 24.5% in the control area. The increase was found to be significantly greater in the test area. Data from police files present a similar picture with a somewhat greater increase in the test area. The relation between police data and questionnaire data was fairly good (r=.75; p=.000; based on percentage figures for the different counties and time periods).
Question: "Do you know anyone who was Random Breath Tested during last year?" - The percentage of subjects who stated that they knew such a person increased from 37.1% to 44.9% in the test area and from 39.1% to 50.4% in the control area. The change over time was not significantly different in the two comparison areas.
Similarly, for the questions: "Has the probability of being Random Breath Tested changed in your home district since last year?", and: "Has there been a change in police enforcement directed at drunk driving since last year?", no significant interaction appeared.
Question: "Have you seen a road check with a POLICE: ALCOHOL CONTROL sign during last year?" The percentage of subjects who stated that they had seen such a road check increased in the test area from 1.5% to 4.8% and in the control area from 4.8% to 6.8%, a significant interaction. The increase was greater in the test area (although the obtained figures obviously cause problems of interpretation).
Question: "Suppose police enforcement directed at drunk driving would remain at the present level during the next twelve months. Imagine you drove a car or a motorcycle through your home district three times a day at different hours of the day (i.e. a total of 1000 times) during that time. How many times do you think you would be stopped by the police to provide a breath test?" This question was regarded as a measure of subjective risk of detection.- The percentage of subjects who estimated that they would be breath tested at least once during the specified time period increased from 50.3% to 58.7% in the test area and from 55.3% to 58.6% in the control area. The increase over time was, however, not significantly different in the two areas.
On one of the attitudinal items the percentage who stated that more breath tests should be made decreased from 82.0% to 78.4% in the test area and from 83.7% to 81.5% in the control area; a non significant difference regarding change over time.
Before the campaign 58.0% reported in the test area that they considered driving with a BAC above the legal limit a very irresponsible act. A year later this figure had increased to 62.4%. In the control area these percentage figures were 68.4% and 65.9%, respectively. The interaction was significant. A change in the desired direction thus appeared only in the test area.
Figures for self reported driving after alcohol consumption during the last 12 months changed very little. In the test area, before the campaign 23.7% reported that they had been driving at least once during last year after consuming small amounts of alcohol (the person made the assumption that the BAC level had not been above the legal limit), whereas 4.3% reported that they had been driving at least once during this time period after consuming more alcohol (the person assumed that the BAC level might have been above the legal limit). The corresponding figures a year later were 22.8% and 4.8%. In the control area the figures for driving after intake of small amounts of alcohol were 22.7% before and 18.1% after the campaign. For driving after having consumed more alcohol these figures were 4.5% and 3.6%, respectively. In both cases the interaction between comparison area and time period was not significant.
In spite of no effect of the campaign on subjective risk of detection, some support for the possible relation between objective and subjective risk of detection may still be obtained by comparing figures across the different counties in the study. To this end, the relation between objective and subjective risk of detection (based on percentage figures for the different counties at the two time periods) was analysed; for objective risk (according to questionnaire data) the correlation was r=.86 (p=.000), and for figures according to police data the correlation was r=.74 (p=.000).
In order to better understand the outcome regarding the subjective risk measure, a stepwise regression analysis was made, based on individual level data (Pedhazur, 1982). All the other factors (questions) were used as regressors in the analysis. The time factor and the area factor were included as well. A significance level of a = .05 was adopted for factors entering and leaving the model.
In the analysis nine factors were selected which together accounted for 22.8% of the variance in subjective risk of detection. One factor dominated, the question "How many times during last year were you Random Breath Tested?"; this factor accounted for 12.9% of the variance. The factor "Has the probability of being Random Breath Tested changed in your home district since last year?" accounted for an increment of 5.2% to the variance. The factor "Do you know anyone who was Random Breath Tested during last year?" accounted for an additional increment of 1.7%, whereas distance driven during the last 12 months accounted for another 1.5%. The remaining five factors accounted for increments less than one percent each.
Similar analyses were performed for the two questions regarding self-reported driving after alcohol intake (after excluding the other factor regarding driving after alcohol intake). For the question regarding driving after consuming small amounts, nine factors were selected which together accounted for 17.8% of the variance in the criterion measure. One factor dominated, the question "How often do you drink alcohol?"; this factor accounted for 12.1% of the variance. The attitudinal factor "What is your opinion of driving with BAC a little bit above the legal limit?" accounted for an increment of 3.0% to the variance, whereas the age factor accounted for an additional increment of 1.2%. The remaining six factors accounted for increments less than one percent each.
On the question regarding driving after consuming larger amounts of alcohol, a similar result appeared, with the same factor dominating; the question "How often do you drink alcohol?" accounted for 2.1% of the variance (total variance accounted for by the model was 4.6%). The attitudinal factor "What is your opinion of driving with BAC a little bit above the legal limit?" accounted for an increment of 1.1% to the variance. The remaining five factors contributed with less than one percent each.
Regarding the attitude to driving with a BAC above the legal limit, a similar analysis was performed (after excluding the other attitude factor). One factor, driving after intake of small amounts of alcohol, accounted for 4.6% of the variance (total variance accounted for by the model was 6.7%). Seven other factors entered as factors in the model; all of these gave increments below one percent each.
No meaningful relations were found with any of the factors related to experience of RBT for this attitude measure or for driving after alcohol intake.
Accidents with suspected alcohol involvement were compared between the two comparison areas for two time periods, from 1985 until the beginning of the campaign and for the campaign period. The proportion of such accidents decreased in both areas; in the test area from 6.5% to 5.2% and in the control area from 6.3% to 5.6%. These data were analysed by logistic regression analysis. The changes in the two areas were not significantly different.
The lack of a demonstrated effect of the campaign on subjective risk of detection can probably be attributed to the fact that the difference between the two comparison areas regarding increase in objective risk of detection was too small. Furthermore, the discrepancy between the figures for experiences of RBT and experiences of road checks with the special sign indicates that the campaign was implemented only to a limited extent in the test area.
Support was, however, found for the relation between personal experience of RBT and subjective risk of detection in the form of a correlation across counties, and, more importantly, from a stepwise regression analysis, a result which rather well replicates findings by Åberg (1987): those who had been breath tested experienced a greater subjective risk of detection than those who had not. Those who knew another person who had been breath tested also had a greater risk of detection than others, even if this factor accounted for a much smaller part of the variance in the criterion measure.
The results for self-reported driving after alcohol consumption showed no effect of the campaign. This behavior was not, according to regression analyses, related to experiences of RBT. Instead, alcohol habits was the factor with the strongest connection to this criterion measure (the more people drank, the more they drove after alcohol consumption). The attitude to driving with a BAC above the legal limit was also related to this behavior, although to a smaller extent. Similar results were found by Norström (1983). There are, however, problems of interpretation regarding the present data. For alcohol habits to be treated as a background variable, a definite requirement is that these changed only marginally during the study period, an assumption which does not seem unreasonable. For the attitudinal factor, which was also related to this criterion measure, this assumption seems much less certain. For that reason this factor can not be treated as a background variable in this analysis.
The only effect in the questionnaire study was regarding the attitude to driving above the BAC limit. The only regressor with a reasonably strong connection with this criterion measure was driving after alcohol intake; those who had done this showed a more lenient attitude to such behavior than others. The variations in this attitude measure was not, according to a regression analysis, related to experiences of RBT.
It deserves to be mentioned that the matter of causality is unresolved in all of these regression analyses.
The lack of effect for the accident measure can probably be attributed to two factors, the campaign was far too limited in scope and duration, and RBT increased significantly in the control area as well.
Everitt BS (1992): The analysis of contingency tables. Second edition. Chapman & Hall, London.
Homel R (1988): Policing the drinking driver. A study of general and specific deterrence. Springer Verlag, New York.
Homel R, D Carseldine & I Kearns (1988): Drink-driving countermeasures in Australia. Alcohol, Drugs and Driving, 4, 2.
Norström T (1983): Law enforcement and alcohol consumption policy as countermeasures against drunken driving: Possibilities and limitations. Accident Analysis and Prevention, 15, 6.
Pedhazur EJ (1982): Multiple regression in behavioral research. Explanation and prediction. Second edition. Holt, Rinehart and Winston Inc., Fort Worth,
Åberg L (1987): Routine breath testing and drivers' perceived probability of breath test. In Alcohol, Drugs and Traffic Safety - T86 (Ed PC Noordzij & R Roszbach), Excerpta Medica, Amsterdam.