This is the most commonly used method to reduce drink driving. Police enforcement is only possible when there is a certain legal limit. The police must be able to detect when a driver has exceeded that legal limit and once this is detected, the driver must be punished. The effective element of police enforcement is deterrence and the effectiveness of deterrence depends on the impression the driver has of his chance getting caught when exceeding the limit and on how severe the punishment is. A distinction can be made between general deterrence and specific deterrence. The aim of general deterrence is to motivate all drivers not to breach the rules by creating fear of sanctions and by giving the idea that the chance of getting caught is high. The aim of specific deterrence is to improve the attitudes and behavior of drivers once they are caught in order to prevent recidivism. For this purpose not only severe sanctions like suspension of the driving license are used (."I will never drink and drive again because the temporarily loss of my driving license has been a horrible experience.") but also remedial treatment programmes. Well-known remedial treatment programmes are the what are known as compulsory driver rehabilitation courses for offenders.
Effectiveness of police enforcement
As mentioned earlier, the effect of police enforcement is based on three elements: the level of the legal limit, the chance of getting caught when exceeding the limit and the severity of the sanctions. The effects of these elements will be dealt with separately.
The effect of having low legal limits
According to a meta-analysis carried out by Elvik & Vaa  reducing the existing BAC-limit for all drivers in a country leads to a reduction of 8 % in fatal crashes and a reduction of 4% in injury crashes. Allsop  estimates that in the United Kingdom 65 lives will be saved annually if the legal limit for the general driver population would be reduced from 0.8 g/l to 0.5 g/l. If a reduction of the BAC-limit always leads to a decrease in the number of crashes, a BAC-limit of 0 g/l for all drivers would be the very best to have. From the perspective of getting the clear message across 'one should never combine drinking and driving' a BAC-limit of 0 g/l indeed would be the best solution. If it is 0 g/l, it is clear to everyone that even the slightest amount of alcohol in the blood is forbidden for all road users. When the limit is above zero, there is always the appraisal a driver has to make whether that one glass of wine (or any other alcoholic beverage) can be consumed or not. From a jurisdictional and technical point of view however, a BAC-limit of 0 g/l might not be not such a good idea. For older (more experienced) drivers the crash rate starts to rise from 0.5 g/l onward. This means that up to 0.5 g/l older drivers are no substantial threat to other road users and themselves. Being sanctioned for something that is hardly dangerous is not fair. Another aspect is that with a BAC-limit of 0 g/l a driver also cannot use a mouth spray and the devices to measure the BAC-level are still not accurate enough to detect very low levels. A third drawback is that a very low limit might hamper catching the big fish (the drivers that drive with levels far above the legal limit). If too much time is spent on the small fish (drivers with a BAC between 0 and 0.5) and the enforcement system is not very efficiently organized, this may lower the chance of getting caught for drivers with a high BAC-level. And it is precisely the drivers with high BAC-levels who cause most of the crashes. Although a BAC-limit of 0 g/l for all drivers may cause problems, this is not the case for young drivers. As the crash rate for young drivers significantly starts to rise at very low levels, a BAC-limit of 0 g/l for young drivers is good for road safety . If one takes account of the inaccuracy of the devices and the fact that one can have a presence of alcohol in the mouth without having consumed alcohol, a BAC-limit of 0.1 g/l or 0.2 g/l for young drivers may be more realistic than 0 g/l. After implementing a BAC limit of 0.1 g/l in Austria for novice drivers, there was a 16.8% fall in fatal crashes involving drivers with a BAC-level of 0.8 g/l or more .
The effect of police enforcement
Some countries allow for random roadside breath testing and in others there must be some kind of suspicion (i.e. the smell of alcohol) before a policeman can test a driver. Both systems are effective, but random breath testing (RBT) is twice as effective as selective testing (only testing after suspicion) . After each doubling of the number of RBTs in the Netherlands, the number of drink driving offenders has decreased by approximately 25% . The effectiveness of RBT can be enhanced when it is targeted on the vicinity of places where alcohol is consumed and at times when the prevalence of drink driving is high, i.e. in weekend nights, and when publicity accompanies enforcement campaigns. Research and experience suggest that highly visible RBT (to deter) combined with targeted RBT that is not clearly visible (to detect) is the most effective  .
On page 41 of the final report the ESCAPE-project called 'Traffic enforcement in Europe: effects, measures, needs and future' (Mäkinen et al, 2003) one can read:
"The Finnish police have pursued a systematic DUI (Drinking Under the Influence) surveillance, including random breath testing and extensive use of publicity, for over a quarter of a century. The risk of being caught for drink driving has increased considerably since 1977 when the police were first empowered to carry out random breath testing and were equipped with pocketsize Alcoholmeter breath analysers. Currently, some 40% of drivers are tested annually in Finland. The number of those caught for drink driving has fallen during the past 10 years from 0.33% to 0.14%. The overall positive trend is clear when evaluating the figures together with the results of roadside breath-testing studies. In the course of this process the punishments for drink driving have gradually eased."
The risk of being controlled for alcohol differs substantially between EU-member states. One of the questions in the SARTRE3-questionnaire (2002) was: "In the past 3 years, how many times have you been checked for alcohol?" The results are shown in Figure 10
Figure 10: Self-reported frequency of alcohol controls over the past 3 years. Source: SARTRE3
In 2002 almost none of the drivers in Italy had been checked on alcohol in the past three years whereas in Finland only 36% was not checked in the past three years. In the Worldwide Brewing Alliance's 'Drink and driving- report 2005' the answers are listed that Brewing Trade Associations were able to collect from the authorities in their countries. One of the questions was: "Please give brief details of the level of enforcement and rates of conviction for driving while under the influence of alcohol or drugs?" The answers reflect the percentages mentioned in Figure 10. The answer from Italy is: "The levels of enforcement are low. The road police have been supplied with pocketsize Alcoholmeter breath analysers but in insufficient quantities. Therefore controls of the BAC level are seldom made with the exception of some areas where discotheques are more widespread (Emilia Romagna). Whereas the answer from Finland is: "RBT is used frequently. Fairly high level of enforcement."
The effect of sanctions
Fines have some effect, but these effects don't last very long. In a Canadian case-crossover study concerning police enforcement in general (thus police enforcement regarding drink driving, but also regarding speeding and other violations)  discovered that the fatal crash rate in the month after conviction was about 35% lower than in a comparable month with no conviction However, 3-4 months after the conviction the drivers drove in an as unsafe manner as they did before the conviction. When the severity of the conviction increased (more demerit points), the effect on the relative rate reduction increased, but didn't last longer. However, if the conviction was very severe (two of these types of convictions would be enough to lose one's driving license), the effect on the reduction of the relative crash rate was small again.
According to a meta-analysis by Elvik & Vaa  driving license suspension leads to a reduction of all crashes by 18%. This makes driving license suspension very effective. There is however one drawback. If enforcement is rather weak, drivers who have lost their driving license may start to drive illegally.
Imprisonment seams to be less effective according to Elvik & Vaa . A change in Norway and Sweden from imprisonment to a graduated tariff of fines and license suspension had lead to reduction of all crashes by 4%.