Schaffer Online Library of Drug Policy Sign the Resolution
Contents | Feedback | Search
DRCNet Home
| Join DRCNet
DRCNet Library | Schaffer Library | Drugs and Driving 


An Evaluation of Compulsory Breath Testing in New Zealand

John P M Bailey

Institute of Environmental Science & Research Limited, PO Box 50-348, Porirua, New Zealand


Compulsory Breath Testing (CBT) was introduced in New Zealand on 1 April 1993, following nine years of Random Stopping. Passive alcohol detectors were used for the initial screening. The targeted rate of stopping, of 1.5 million drivers annually, was not achieved in the first four months of operation, but was met by the end of the first year. Paid publicity was largely limited to the first three months.

Evaluation was made difficult by the merger of the Ministry of Transport enforcement operations with those of the Police, in July 1992, and an intense drink driving enforcement blitz in December 1992, prior to the introduction of CBT. There was a reduction in reported drink driving injury accidents and in the proportion of all reported injury accidents which were alcohol related, subsequent to the campaign in December 1992. The reduction continued for some months after the introduction of CBT in April 1993. Alcohol related accidents and accidents at night time decreased significantly more than did non alcohol related accidents or accidents during the day. A greater reduction occurred in urban areas than in rural areas.


On 1 April 1993 Random Breath Testing (RBT) was introduced in New Zealand. This followed just over eight years of random stopping. To distinguish the RBT programme from the random stopping programme it was called Compulsory Breath Testing (CBT). At the same time the legal blood alcohol limit was lowered from .08 for drivers under 20 years of age to .03 and raised from .03 to .08 for unlicensed and disqualified drivers over 20 years of age.

The Police operate CBT checkpoints at which all drivers stopped are supposed to be tested. Checkpoints are essentially car based operations, with a target of 1.5 million drivers stopped and tested each year. This translates to 7 in 10 licensed drivers. For the first four months the stopping rate was 85% of the targeted rate. The target has been achieved in most months subsequently. Drivers are screened by using passive alcohol sensors (an Alcotech AR1005). The television launch commenced on 21 March 1993, with a large level of activity through to the first week of May (an average of 325 Target Audience Rating Points per week), followed by a two week burst in early June and again in early August.


The impact of CBT on road accidents is evaluated by studying data on injury accidents reported to the Land Transport Safety Authority (LTSA) and from data on fatally injured drivers obtained by matching LTSA data to postmortem blood alcohol results.

Neither of these sources of data is completely satisfactory. LTSA data are incompletely reported, with about a 50% reporting rate compared with data on road accident victims taken to emergency departments at selected hospitals. The data have largely subjective information on alcohol involvement with, for instance, only 15% of seriously injured drivers in 1993 being assessed for alcohol consumption. Data on fatally injured drivers are based on a relatively small number of cases and are not obtained for about 30% of drivers.

Figure 1 presents changes in the ratio of alcohol-suspected injury accidents to non-alcohol suspected accidents. Alcohol is indicated in accidents by either being measured (in 1993 15% of seriously injured drivers had a chemical test) or the officer recording if he or she suspected alcohol involvement. While there is considerable variability in the data, two major trends are evident. A flat/slightly upward trend to about August 1990, followed by a downward trend. Regression lines through each of these have been fitted.

Figure 1
Reported Injury Accidents: Ratio of Alcohol Suspected to Non-Alcohol Suspected Accidents: Jan 1987 - Sep 1994

Figure 2 presents cusums of the ratio of alcohol-related reported injury accidents to non-alcohol-suspected accidents between January 1987 and September 1994. A cusum is a series of numbers which are the cumulative sum of the differences between an observed series and the corresponding expected series. The major features of this figure are the small variability around the horizontal line to about August 1990 when it increased, the brief decrease about April 1992 and the near steady decrease since December 1992.

Figure 2
Cusum of Ratio of Alcohol-Suspected to Non Alcohol-Suspected Accidents from Jan 87 - Sep 94

Data on injury accidents for the years April-March years 1991/92, 1992/93, 1993/94 are presented in Table 1.

Table 1
Changes in Reported Injury Accidents, April-March years

  1991/92 1992/93 1993/94 % change 91/92-92/93 % change 92/93-93/94
night (8pm-4am) 2857 2762 2441 -3.3 -11.6
day (4am-8pm) 9175 8809 8513 -4.0 -3.4
alcohol-suspected 2731 2429 2143 -11.1 -11.8
sober 9301 9142 8811 -1.7 -3.6
alcohol-suspected, night 1699 1545 1302 -9.1 -15.7
sober, night 1158 1217 1139 5.1 -6.4
9pm-midnight 1438 1319 1113 -8.3 -15.6
midnight-5am 1004 1057 965 5.3 -8.7

In comparing data for 1992/93-1993/4, the differences between night and day, or alcohol-related and non-alcohol related, are both significant at the 1% level. In comparing 1991/2 with 1992/3. the differences between night and day are not significant, whereas those between alcohol related and non-alcohol related, night time alcohol-related and non-alcohol related and 9 pm-midnight compared with midnight-5 am are all significant at the 2.5% level.

A comparison of data for the periods April-July in each of the years 1987 to 1994 in areas with speed limits of 50 or 100 km/hr, given in Figure 3, shows how CBT has performed after the first full year, relative to its performance shortly after its introduction. The period April-July was selected in that July 1994 was the last month for which data on reported injury accidents were believed to be complete at the time of writing. The decrease in drink driving accidents, particularly in urban areas, began after 1990. Both the increase from 1992 to 1993 and the decrease from 1993 to 1994, were statistically significant at the 2.5% and 5% levels, respectively, in the urban areas, whereas only the decrease from 1992 to 1993 was significant in the rural areas, at the 10% level.

Figure 3
Reported Injury Accidents April-July 1987-1994 in 50 and 100 km/hr Zones Taking 1987 as Base 100

Data on BACs of fatally injured drivers need to be corrected for the likely levels of the drivers from whom no blood sample was obtained. This is done using police suspicion of alcohol involvement and information from Coroners' reports, using a method devised by the author. The estimated numbers over the legal limit are given in Table 2.

Table 2
Number of Fatally Injured Drivers with a BAC over the Legal Limit

  1987 1988 1989 1990 1991 1992 1993
measured BAC>.08 151 121 128 132 107 98 87
estimated BAC>.08 170 147 146 157 125 121 100
Total 438 394 413 374 342 361 333
% >.08 38.8 37.3 35.4 42.0 36.5 33.5 30.0

The drop in proportion over .08 was greater between 1990 and 1991 than between 1992 and 1993, though the 1990 figure was unusually high. Furthermore, the numbers of drivers involved are small. The reduction in the proportion over .08 in 1993 is not statistically significantly different from that in the previous year.


Assessing the effects of the CBT programme in New Zealand is difficult as there is not a suitable control or comparison group. Preceding the introduction of CBT were large generalised and local changes in economic conditions, a gradual reduction in the amount of alcohol drunk per person, the removal of numeric and time of day conditions on liquor licences (1 April 1990), the amalgamation of Ministry of Transport traffic officers with the New Zealand Police (1 July 1992) and the introduction of speed cameras (15 October 1993).

It was expected that the CBT programme would result in an immediate and obvious reduction in alcohol-related accidents. It is clear that this did not happen, though the percentage of killed drivers who were estimated to be over the legal limit reached an historical low in 1993. The decrease in 1993 is similar to that in 1991 and 1992, though the decrease in 1991 could be seen as a result of an unusually high figure in 1990.

The question arises as to whether or not the 1993 figure would have been as low as it is without the introduction of CBT. An examination of the injury data provides some assistance, as it is likely (but not inevitable) that trends in the percentage of killed drivers who were estimated to be over the legal limit are reflected in trends in injury accidents. The ratio of alcohol suspected to non-alcohol suspected drivers for injury accidents (Figure 1) shows a gradual reduction from 1990 on, with a dip beginning late 1992. The cusums (Figure 2) indicate a short increase in the ratio (Figure 1) in mid-1990, a short decrease in early 1992, followed by a further decrease in the ratio commencing December 1992, which has been maintained. We are not aware why there should have been an increase in the ratio in 1990 or a reduction in early 1992. However the reduction in the ratio commencing December 1992 may be related to the Christmas alcohol campaign. This campaign, the first where the enforcement responsibility belonged to the Police, is believed to have been unusually intensive, though it is poorly documented. It appears that most of the reduction in the percentage of killed drivers who had been drinking in 1993 is associated with the 1992 Christmas campaign.

The data in Figure 3 are for April to July months and provide a useful comparison of performance up to and including 1994. Overall alcohol-related accidents are lower in 1994 than 1992, the year prior to CBT, especially in rural areas. However the differences are small and it is unclear as to how speed cameras have affected the number of alcohol-related accidents relative to sober accidents. It is possible that CBT has had a positive but small residual effect but further work will be required to determine this.

In exploring why the CBT programme has not lived up to expectations it can be compared to guidelines that Australian authorities have developed for achieving successful random breath testing (RBT) programmes. Homel (1990) listed guidelines as to what needs to be done to achieve a successful outcome. At least 1 in 3 drivers should be stopped each year at checkpoints and all drivers tested, with the programme supported by extensive publicity. In New Zealand the Police reported stopping the equivalent of 7.8 in 10 licensed drivers in the first year of CBT. A household survey of 1,000 adults in May 1994 by the Land Transport Safety Authority found 286 persons who said that they had been stopped at an alcohol checkpoint in the previous 12 months. For the last time they were stopped, 31% of respondents reported not been tested and 40% said that they were stopped between 8.00 am and 8.00 pm. While the CBT programme was launched with "a bang", as recommended by Homel, the intensity of advertising has not being sustained. Similarly coverage of CBT in newspapers was high in March - May 1993 but then declined to low levels (Wyllie, 1993).


The author acknowledges the assistance of Mr W A Perkins in providing survey data and in the preparation of the paper.


Homel, R. (1990). Random Breath Testing that Saves Lives. In Effective Strategies to Combat Drinking and Driving: An Edited Collection of Papers Presented at the International Congress on Drinking and Driving Edmonton, Alberta, Canada, March 28 - 30, 1990. Alberta Solicitor General' Office.

Wylie, S. (1993) Newspaper Media Coverage of Compulsory Breath Testing: A Content Analysis. Research & Statistics, Land Transport Safety Authority, Wellington.