Title Banner - Alcohol and Highway Safety 2001: A Review of the State of Knowledge

5 - Dealing with the
Alcohol-Crash Problem


Conceptual Frameworks

The most widely used framework for analyzing alcohol-crash countermeasures is the two-dimensional matrix first proposed by Haddon and Brenner in the 1960s for traffic-crash countermeasures in general (Haddon, 1968). The framework is etiologically based, specifying three major crash-related factors along one dimension, and three major phases of traffic crashes along the other dimension (Figure 5-1). Countermeasures are analyzed and generated by the types of causes they address as defined by the nine cells of the matrix. The framework was an outgrowth of an earlier characterization of traffic crashes as a public health problem which was to be addressed by actions directed at the driver (host), the vehicle (agent), and the highway (environment) (Gordon, 1949).

Figure 5-1: The Haddon-Brenner Framework 
for Analyzing Traffic-Crash Countermeasures
Human Factor   
Vehicle & Equipment Factor   
Environment Factor    

Joscelyn and Jones (1981) observed that the Haddon-Brenner framework was too simplistic to justify its place as the dominant framework for analyzing highway safety problems and for generating solutions. They argued that the framework's focus on the crash problem alone is too narrow and that many events not immediately associated with a traffic crash can influence crash causation. Specifically, they noted, the framework does not explicitly include societal aspects, implying that the traffic crash problem will be dealt with by some undefined external forces produced by systems not represented in the framework.

Joscelyn and Jones proposed a new framework relating the loss-generating elements of highway transportation to the elements of society that attempt to control those losses. Specifically, their framework represents a process to control the disutility in the Highway Transportation System, a process in which disutility is monitored by society, and actions to reduce disutility are taken by risk management systems (such as the traffic law system) created for that purpose. By such a process disutility is maintained at a level that society (public and private organizations as well as individuals) will tolerate. Clearly, the absolute level of disutility maintained by the process at any point in time is dependent not only upon the capabilities of the risk management systems that generate the forces to control risk, but also upon societal pressures on those systems to reduce risk. Viewed in this context, the role of accurate and timely information about risk and risk management in highway safety is obvious.

Both frameworks have their advantages, the Haddon-Brenner framework being easy to understand and visualize, and the Joscelyn-Jones risk management framework incorporating elements that deal with crashes, in addition to elements that influence the etiology of the crashes themselves. We will use both frameworks in discussing the nature and effects of alcohol-crash countermeasures in this chapter.


Possible targets of alcohol-crash countermeasures exist in each of the nine cells of the Haddon-Brenner matrix, from human factors in the pre-crash phase to environmental factors in the post-crash phase. However, countermeasures designed explicitly for alcohol-related crashes are inherently more suited to the pre-crash phase, particularly if the pre-crash phase is extended far enough in advance of the crash to where perceptions of risk and risk-management actions are formed. This implies a policy of crash-prevention for alcohol-crash countermeasures, not only by preventing drinking-driving, but also by preventing crashes resulting from drinking-driving.

All three classes of factors (human, vehicular, and environmental) are potential targets for pre-crash alcohol countermeasures. However, another distinction must be made as to whether drinking or drinking-driving is the targeted behavior. This leads to the creation of another matrix devoted only to the pre-crash phase (Figure 5-2).

Figure 5-2: Matrix of Target Types for
Pre-Crash Alcohol Countermeasures
 Drinking-Driving BehaviorDrinking
Human Factor  
Vehicle & Equipment Factor  
Environment Factor   

Here, there are only six cells, containing such countermeasures as:

Countermeasure ClassExample
Human / Drinking-DrivingDWI legal-system deterrence and incapacitation countermeasures such as jain and license revocation
Human / DrinkingCountermeasures aimed at alcohol availability and consumption such as server liability laws
Vehicle & Equipment / Drinking-DrivingAlcohol interlocks
Vehicle & Equipment / DrinkingNone Known
Environment / Drinking-DrivingHighway edge markings
Environment / DrinkingControl of drinking locations


The cells in Figure 5-2 categorize countermeasure targets at the most general level of detail. Classifications at any reasonable level of detail can be defined for each cell by combining sets of drinking-driver characteristics such as those discussed in Chapter 4. Note that a similar typology can be developed for pedestrians and bicyclists as well as drivers.

Driver-Oriented Countermeasure Actions

After the target of a countermeasure has been defined, the actions against that targets must be taken. Joscelyn and Jones (1981) defined two general categories of driver-oriented actions:

These categories are an extension of the concept of specific and general deterrence, where specific deterrence imposes penalties on those who have been caught engaging in a proscribed behavior, and general deterrence creates a threat of punishment for those who have not necessarily been caught engaging in a such a behavior. The concept of incapacitation has often been used in conjunction with deterrence-based approaches. If achieved, incapacitation reduces alcohol-crash risk by removing one's ability, means, or authorization to drive at all, through such means as physical confinement, confiscation of one's vehicle, or removal of one's driving license. In theory, incapacitation establishes a deterrent threat as well as prevention of driving during the period of incapacitation.

Finally, a third type of pre-crash action can also be conceived, an action to reduce the probability of a crash after drinking-driving has occurred. However, we found no evaluated countermeasures with this objective. As will be seen later in this chapter, a large percentage of evaluated alcohol-crash countermeasures have been aimed at deterring drinking-drinking driving through the threat of unpleasant consequences, incapacitating drivers, or both.

Policies and Programs

We found no evidence of a multi-agency national policy for dealing with the alcohol-crash problem. The policy of the National Highway Traffic Safety Administration (NHTSA) seems to stress the direct control of drinking-driving (rather than drinking) with little emphasis on environmental factors (U.S. Department of Transportation NHTSA, 1998) which are the domain of the Federal Highway Administration (FHWA). The National Institutes of Health (NIH) seem most concerned with drinking-oriented countermeasures that stress human factors, while the National Institute of Justice (NIJ) focuses on DWI deterrence. Cooperative agreements exist between NHTSA and NIH, and between NHTSA and NIJ, on some alcohol-countermeasure research and evaluation programs.

There is also evidence of increased interest in partnership arrangements between federal governmental agencies and various local government and private-sector organizations. For example, NHTSA began a "Partners in Progress" program in 1997 that worked with state and local agencies, and their private-sector partners, to develop innovative solutions to the alcohol-crash problem. NHTSA is currently exploring ways of increasing cooperation among public-sector and private-sector organizations.


In this update we are most interested in alcohol-crash countermeasures that have been evaluated by their effectiveness in reducing alcohol-crash risk. Such evaluations are called "impact evaluations" in the literature, in contrast to evaluations that measure effectiveness only in terms of increased activity or improved processing. The number, variety, and quality of impact evaluations have increased dramatically since the 1968 state of knowledge review. We found 107 references to "drinking drivers" in the library of The University of Michigan Transportation Research Institute (UMTRI) prior to 1968, most of which addressed the experimental and epidemiologic aspects of the alcohol-crash problem. Since 1968, the UMTRI library has added more than 3,000 documents on this subject, many of which report the results of impact evaluations of alcohol-crash countermeasures.

Designs. There is an extensive literature on the evaluation of societal programs, a review of which would be far beyond the scope of this update. Some of this literature deals explicitly with the evaluation of highway safety countermeasures, and generally reflects concepts in use today in evaluating alcohol-crash countermeasures (Griffin III, Powers, and Mullen, 1975; Hall and O'Day, 1971; Streff, 1991; U.S. Department of Transportation NHTSA, 1981; Vilardo, Davis, Jones et al., 1977).

A central requirement in evaluating any given countermeasure is to account for factors other than the countermeasure that might affect the outcome. Such factors are termed a threat to the internal validity of the evaluation (Campbell and Stanley, 1963). Threats to external validity (situational factors that may prevent generalizing the results of the countermeasure to other settings) are also present, but, usually, internal validity is the primary goal.

The treatment-and-control design attempts to get internal validity by comparing the experience of different groups of people. Such comparisons are appealing because they are relatively simple in concept. They tend to be more valid if they are used to compare groups of people, one to which a countermeasure is applied (the treatment group), and one to which it is not applied. In the simplest case, the groups are of the same size and are studied for the same period.

One common way of establishing the suitability of a control group is to use a before-and-after × treatment-and-control design. If the basic approach is to compare the treatment group before and after the intervention, then the control group is used for making a corresponding comparison. Typically, in this approach, it is assumed that extraneous factors affect treatment and control groups in the same way, and a relation (usually proportional) is assumed between the measure in the treatment group and the control group. Then, one uses the change in the control group to adjust the change in the treatment group for the influence of extraneous factors.

Time series designs are a conceptual extension of before-and-after comparisons. Here, not just two periods, before and after, but a number (usually 40 or more) of periods are used. Nearly always, a large percentage of the periods are in the "before" phase, and usually several are in the "during" phase of the countermeasure. Such a series of data can be analyzed more thoroughly, and effects of confounding factors estimated and eliminated.

There are basically two approaches to time series analyses. One uses only the time series of the measure of effectiveness (or "criterion variable") studied, and the other also uses other data, either as "control" or as "explanatory" variables. In the first approach, an internal structure is determined for series of the measure of effectiveness. In the second approach, additional time series are used.

Another, even more basic, evaluation design requirement that is often neglected or not provided in evaluation studies is a detailed description of the nature of the countermeasure and what it is intended to accomplish. Such a description should:

The relationship between the input to the countermeasure and the output of the countermeasure is especially critical. The relationships between system input and system output have been described as a "chain of action" in the evaluation literature (Vilardo, Davis, Jones et al., 1977). This literature speaks of various levels of objectives which are related in a hierarchical fashion to the countermeasure's ultimate highway safety objective. For example, countermeasure resources (such as money for police overtime) are related to countermeasure performance (such as more patrol activity and more citations for DWI) which is related to immediate objectives (such as fewer DWI violations) which are related to ultimate objectives (such as fewer alcohol-related crashes involving a fatality or serious injury).

A final aspect of these descriptive preliminaries to the evaluation is describing the operational environment of the countermeasure. Such information is important for identifying possible confounding events or activities that might have affected the impact of the countermeasure on its target. It is also important for comparing the results of the countermeasure as implemented in this jurisdiction with those of similar countermeasures that have been or may be implemented in other jurisdictions.

Limitations. As suggested above, countermeasure evaluations are inherently limited by one's ability to filter out the effects of other factors. The better the confounding effects of non-countermeasure factors can be controlled for, the more convincing the evaluation. True experimental designs of the type used in the laboratory are not possible; instead, quasi-experimental designs have to be used for these "natural experiments."

A potential problem that can severely limit the credibility of an evaluation is the use of an inappropriate criterion variable for measuring the impact of the countermeasure. Alcohol-crash countermeasures following the general risk-management strategy aim to reduce the number of crashes that are likely to have involved alcohol as a causal factor. An objective way is needed for identifying such crashes from available data. At present, driver BAC is the best objective measure of alcohol-related crashes, but it will usually only be available for fatal crashes, and only for the fatally injured drivers. In such cases, a BAC of .10+ can reasonably be expected to be associated with alcohol as a causal factor. Evaluations in jurisdictions having a small number of fatal crashes will have to rely on surrogates based on other attributes that have been found to be closely correlated with BAC. One such surrogate that often has been used is nighttime single-vehicle injury crashes. Subjective measures such as crashes in which a driver is judged by the police officer investigating the crash as "had been drinking" have also been used, but may give spurious results because of non-crash related factors that may have influenced the officer's opinion (e.g., heightened awareness of the alcohol-crash problem stemming from a highly-publicized crash).

Another basic limitation is the quality of the data used in the evaluation. The most commonly used data on traffic crashes are collected by police agencies in the course of investigating crashes. When using such data, a number of points need to be watched to make sure that the data are compatible over time, or across agencies. Examples of such potential problems are:

Similar problems can occur in other types of archival data, including driver records data of prior violations and crashes.

Surveys that have collected information from (or about) individuals are often used as a data source in evaluations of alcohol-crash countermeasures. Such data are used for such purposes as determining the awareness of some group of a countermeasure or of a message related to a countermeasure, and obtaining self-reports of the respondents' driving behavior which is to be affected by the countermeasure. In rare cases, all members of a population are surveyed, for example, all persons arrested for DWI in the study area during a certain time period, or all licensed drivers in an area may receive a questionnaire by mail. Usually, however, only a sample of the population is contacted, and one extrapolates from the sample to the entire target population.

Survey data present their own set of problems, ranging from improperly phrased questions that bias the responses, to the selection of a survey sample theat does not represent the population from which information is sought (e.g., male drivers of age 21-34 years). Such problems are discussed at length in textbooks on survey research. A more fundamental concern is the very nature of the data collected, which is self-reported. Respondents may not remember their behavior or events that may have influenced their behavior, may not wish to report incriminating or embarrassing behavior, or may slant their responses to please the surveyor or to comply with perceived social norms.

A third class of data that are sometimes used in evaluating alcohol-crash countermeasures are obtained through field observations. The BAC of non-crash involved drivers is one common type of such data. Determination of BAC requires that a driver be stopped and tested using either an "active" or a "passive" testing device. Data may be collected at a parking lot, near a restaurant, bar, etc., where drivers can be contacted, or from drivers stopped in traffic.

While quite high response rates have been achieved with such roadside surveys (over 90% in some instances), a variety of problems can occur in making field observations. Problems can include poor selection of times and locations of observations, lack of training of observers, lack of quality control procedures, and poorly designed forms and protocols. These problems can result in data that do not accurately reflect the BACs of the drivers targeted by the countermeasure being evaluated, or the BACs of some comparison group of drivers.


This section has discussed the general nature of evaluations of alcohol-crash countermeasures to provide background and a framework for the more detailed review of the literature to follow. Two conceptual frameworks for analyzing alcohol-crash countermeasures were discussed, and ways of classifying countermeasures programs were suggested. The general requirements for designing an evaluation were examined briefly, along with some of the problems and limitations of traffic-crash evaluations.

Our review deals almost entirely with countermeasures whose impact on the alcohol-crash problem has been evaluated. It incorporates the above considerations and is structured at the top level along the lines they suggest, viz.:

A third class of countermeasure, containing those that would attempt to prevent crashes after the occurrence of drinking driving, is discussed briefly. Countermeasures for alcohol-related crashes involving pedestrians and bicyclists are discussed separately.


Deter and Incapacitate Drinking-Drivers

The Traffic Law System. As indicated above, these countermeasures involve the Traffic Law System (TLS) as the primary risk-management system. This review is concerned with a particular type of traffic-crash risk, that which is created by alcohol-impaired drivers. At the highest level, the formal functions of a DWI enforcement system are law generation, law enforcement, adjudication, and sanctioning, defined broadly by Joscelyn and Jones (1981) as follows:

Law Generation

Law Enforcement



Besides the traditional functions listed above, a fifth, less formal, function is concerned with the dissemination of information among the components of the system and to potential DWI violators, among others.

Interestingly, few researchers have questioned the underlying hypothesis of deterrence as applied to the alcohol-crash problem. However, one study (Evans, Neville, and Graham, 1991) used state-level data over the 1975-1986 period to "test the predictions of deterrence theory" with respect to anti-drunk driving policies and legislation during the years 1982-1986. An econometric regression-analysis approach was used with various alcohol-related and non-alcohol measures as dependent variables, and 0-1 variables for the year during which a given law (ten laws were included) was in effect.

Evans and associates found no conclusive evidence that any specific form of punitive legislation had a measurable effect on motor vehicle fatalities nationwide, but that there was evidence that multiple laws designed to increase the certainty of punishment (e.g., sobriety checkpoints and preliminary breath tests) had a synergistic deterrent effect. They also concluded that mandatory seat belt use laws and beer taxes may be more effective at reducing drunk driving fatalities than policies aimed at general deterrence. We note that this study dealt only with the legislative component of the TLS, rather than all components (e.g., enforcement and public information) that are hypothesized as being necessary to create deterrence. Nevertheless, the study provides some evidence that passage of multiple laws, accompanied by some unspecified further TLS actions by non-legislative components, can deter drinking-driving.

As noted in Chapter 1 of this review, Wagenaar, Zobeck, Williams et al. (1995) attempted a meta-analysis of TLS-based alcohol-crash countermeasures reported during 1960-1991. They found 125 studies containing separate empirical evaluations of the effects of 12 alcohol-crash countermeasures. They found that all of the DWI control efforts were associated with reductions in drink-driving and traffic crashes. They estimated that overall, the mean reductions achieved by such countermeasures in the U.S. reported in the recent literature (i.e., 1980-1991) ranged from 8% to 17%. The authors observed that "DWI control literature is limited by the preponderance of weak study designs and reports that often fail to include basic data required for meta-analysis." (p. 307)

Prior NHTSA-sponsored reviews have classified deterrence and incapacitation countermeasures for drinking drivers by the TLS function most predominant in executing the countermeasure, and our review below follows the same approach. One such prior review was sponsored NHTSA and was conducted by Jones and Lacey in 1991. It deals with scientific impact evaluations conducted in the 1980-1989 time period, and material from that review is discussed briefly here to provide perspective for the evaluations conducted since 1989. A bibliography of the studies examined in the 1991 review is contained in Volume I of the review report (Jones and Lacey, 1991a), and reviews of individual studies are collected in Volume II (Jones and Lacey, 1991b). As indicated earlier in this report, under this contract, we conducted a special review of the recent scientific literature about drivers who have been convicted more than once of DWI and that has been published as a separate report (Jones and Lacey, 2000). Many of the evaluations reviewed also dealt with first offenders. Such evaluations are re-examined here below with respect to their findings on first, as well as repeat, offenders.

Legislative Countermeasures. While all TLS countermeasures have their origin in laws that authorize them, some exist without any explicit, formal program of enforcement, adjudication, and sanctioning. In their meta-analysis, Wagenaar and associates (1995) estimated that such laws were associated with mean reductions in "alcohol-impaired driving and crashes" of 3% to 14% in the U.S., depending on the length of the post-intervention period examined in the evaluations. (The reductions were 9% to 14% for post-intervention periods in the 13-48 months range).

Administrative License Revocation Laws. Several such legislative countermeasures have been evaluated in recent years, including administrative license revocation (ALR) laws (also called administrative per se laws) in the past ten years. Laws calling for administrative license revocation have become widely used in the U.S., existing at this writing in 40 states and the District of Columbia (NHTSA State Legislative Fact Sheet - Administrative License Revocation). The term "administrative license revocation" (ALR) as used here refers to the revocation of an impaired driver's drivers license by an administrative agency (such as a state department of motor vehicles) rather than by a judicial agency. In general, such laws permit a driver's license to be revoked (or suspended) on the basis of failing or refusing to submit to a BAC test. The laws provide that the driver's license shall be revoked within a prescribed time period, based upon a report submitted to the administrative agency by a police officer. Most ALR laws require the police officer to take the offender's license at the time of arrest and to provide the offender with a temporary license which also serves as a notice of revocation and provides information regarding the offender's right to an administrative hearing or review.

McArthur and Kraus (1999) reviewed three evaluations of ALR laws in several states and concluded that the laws were effective in reducing recidivism in some states, but not in others. All three evaluations were termed "dual retrospective cohort studies" in which recidivism rates under the ALR law was compared to recidivism rates not under the ALR law. The first study (Stewart, Gruenewald, and Parker, 1992) found the law did not reduce recidivism in two states, but did reduce it (by about 30% to 40%) in another state. The second study (Lacey, Stewart, Marchetti et al., 1990) found a recidivism reduction of about 37% up to 30 months after the license action, but no reduction at 36 months. The most recent of the three evaluations (Rogers, 1997) examined the odds of recidivating within one year after license action, finding that the ratio of the odds under the ALR condition to the odds under the no-ALR condition was 0.73. Note that the there were differences in the definition of recidivism among the three studies, with Stewart and associates and Rogers using re-conviction as the "failure," and Lacey and associates using re-arrest.

Voas and associates (1998) evaluated the effect of ALR on the recidivism of multiple offenders. They examined the driving records of 45,788 drivers who were convicted of driving under the influence (DUI) in Ohio between July 1, 1990 and August 30, 1995. Several analyses were performed, the most pertinent of which to this review was an analysis of the DUI recidivism of two cohorts of drivers, the first cohort convicted before the ALR law, and the second convicted after the ALR law. The analysis showed that one year after their arrest, about 19% of the before group had recidivated, compared to only about 5% of the after group. However, as the authors indicate, not all of this large reduction in recidivism can be attributed to ALR, since new legislation strengthening and extending the vehicle impoundment and immobilization occurred at the same time as the ALR law.

In an earlier study, Beirness, Simpson, and Mayhew (1997) evaluated both the general deterrence effect and the specific deterrence effect of ALR combined with another sanction (discussed in detail below), vehicle seizure and impoundment (VSI), but did not differentiate between first offenders and repeat offenders. Both programs had been implemented in Manitoba, Canada in 1985. They found that the combined sanctions decreased: drinking-driving fatalities by 12%, nighttime single-vehicle crashes by 26%, repeat DWI offenses within four years by 44%, and traffic crashes among DWI offenders in the 97 days following a DWI offense by 69%.

Voas and Tippetts (1999) also examined the general deterrent effect of ALR in their recent national study. Data from their report indicate that ALR laws reduced alcohol-related fatal crashes in the United States by about 30% over the period 1982-1997. The authors noted that "this long-term trend is not the product of a single law, but the result of the growing impact of several laws over time plus the affect of some factors not included in the model tested, such as the increasing use of sobriety checkpoints and the media's attention to the drinking-and-driving problem."

Two other types of such laws that have been the subject of recent evaluations are of interest in this review, .08 BAC legal-limit laws, and zero tolerance laws for youth.

.08 Laws. In the United States, .10 BAC illegal limit laws were the rule circa. 1980 until the states of Oregon and Maine adopted .08 BAC as their illegal limit in the early 1980s. Since then, 23 states, the District of Columbia and Puerto Rico have adopted new laws setting .08 as the limit, and many more are likely to follow due to recent Federal legislation imposing financial penalties on those who fail to do so. The Federal legislation was supported by several evaluations indicating the lowered illegal limit could reduce alcohol-related crashes.

A report by Research and Evaluation Associates (1991) studied the immediate effects of the passage of a .08 law in California which went into effect in January 1990, followed by implementation of administrative license revocation six months later. Both initiatives received extensive media attention during legislative debate and at implementation. Time series analyses of FARS data revealed a 12% reduction in alcohol related fatalities associated with the implementation of the .08 law. However, the authors were unable to separate out the relative effects of the two laws, so the true magnitude of the effect of .08 remains under debate.

In 1995, the California Division of Motor Vehicle published an evaluation of the law with a longer term (four year) follow-up. They, too, found reductions in alcohol related fatal crashes, and also in serious injury crashes and late night crashes. However, again, because of the temporal proximity of the implementation of the two laws, they were unable to discern the specific effects of the .08 law.

In a preliminary study of FARS data of the first five states adopting .08, Johnson and Fell (1995) reported reductions ranging from 4% to 40% in four of the five states studied.

Hingson, Heeren, and Winter (1996) studied the effects of lowering the BAC limit from .10 to .08 in California, Maine, Oregon, Utah and Washington. They used a design incorporating comparison states which kept their BAC illegal limit at .10. Most of the comparison states were in the same region as their respective .08 states; however in the case of California, Texas was selected, presumably because of its size. The authors reported a 16% reduction in crashes with a fatally injured driver at .08 or higher associated with passage of an .08 law. They also observed similar reductions for high BAC drivers. As with the California studies cited above, these conclusions must be tempered because some of the states also adopted other laws such as ALR during the study period.

Another analysis of a more recent FARS file for the same five target states by Scopatz (1998) used different comparison states that those selected by Hingson and associates as well as the original set. He reported reductions ranging from 4% to 14% (none statistically significant) in drivers in crashes at or above .08 in the .08 states relative to his comparison states and argued that this demonstrated the influence the choice of comparison states could have on studies of this design.

The six states adopting .08 laws in 1993 and 1994 were studied by Hingson, Heeren, and Winter (2000) using techniques similar to those employed in their earlier study and expanding on them. They used meta-analytic techniques to combine the results of single state pair analyses comparing reductions in the proportions of both fatal crashes involving a driver with a BAC of .01 or above and fatal crashes involving a driver at .10 or above. Examining all six .08 states relative to their comparison states, they found a statistically significant 6% reduction in the proportion of fatal crashes with drivers at or above .10 BAC. When they limited the analysis to those four states which had had an ALR law in effect for three or more years before implementation of the .08 law, they also observed a 6% reduction attributable to the .08 law. The analyses of crashes with drivers with a BAC of .01 or above showed a 5% relative decline both for all six .08 states and the four that had longstanding ALR laws.

North Carolina implemented an .08 law on October 1, 1993, long after having implemented other legislation traditionally thought to be effective, such as ALR. Foss, Stewart, and Reinfurt (1998) conducted time series analyses on a number of measures of alcohol-related crashes to assess the effect of the new .08 law. Additionally, they conducted a series of before-and-after comparisons with control states similar to those conducted by Hingson and associates cited above. North Carolina had been experiencing a long term trend of declines on those measures since the early 1980s when significant legislative measures in this area were implemented. These authors did not discern a statistically significant effect coincident with the implementation of the .08 law, but merely a continuation of the downward trend.

As indicated above, Voas and Tippets (1999) conducted a global analysis of FARS data for all 50 states in an effort to discern the effects of ALR, .10 per se, and .08 per se laws. Examining data from 1982 through 1997, they used weighted least squares regression models to test for the effects on both low BAC and high BAC drivers in fatal crashes coincident with the adoption of each of these categories of laws. They reported reductions attributable to .08 laws on the order of 8% on both measures. However, the authors also cautioned that "the attribution of savings to any single law should be made with caution since each new law builds to some extent on existing legislation and on other ongoing trends and activities." (p. iv).

A time series study of fatal crashes in eleven states which had adopted .08 by Apsler, Harding, and Klein (1999) found significant reductions in alcohol-related fatalities in seven of the eleven coincident with the introduction of .08. In five of those states, the reduction was attributed to the introduction of .08 alone, though it is important to note that they all had ALR laws in effect. In the other two states showing reductions, ALR and .08 laws were implemented in close succession.

In 1999, the General Accounting Office conducted a review of available evidence on the effectiveness of .08 laws including reviewing relevant studies and interviewing researchers and traffic safety professionals (McCain, Hollings, Shuster et al., 1999). The authors concluded that "the evidence does not conclusively establish that .08 BAC laws by themselves result in reductions in the number and severity of crashes involving alcohol." and that "A .08 BAC law can be an important component of a state's overall highway safety program, but a .08 BAC law alone is not a ‘silver bullet.' Highway safety research shows that the best countermeasure against drunk driving is a combination of laws, sustained public education, and vigorous enforcement." (pg 23)

Voas, Taylor, Baker, et al. (2000) studied the effects of the passage and implementation on an .08 law in Illinois in July 1997. In this preliminary study, they examined not only the effect of the law on alcohol-related crashes but on the Traffic Law System. Though the total number of DUI arrests increased by 10.8%, they reported no adverse consequences on the operation of the courts and sanctioning agencies. The proportion of arrestees with BACs of .08 and .09 increased, and the average BAC of DUI arrestees Statewide decreased from .18 to .16. BAC test refusal rates were unchanged. Using interrupted time series analyses, they observed a significant reduction of 13.7% in the proportion alcohol-positive drivers in fatal crashes compared to a non-significant 2.5% increase on the same measure in five adjacent comparison states. A follow-up analysis of the effects of the Illinois .08 law incorporating 1999 FARS data, indicated that the effect persisted (Voas, Tippetts, and Taylor, 2001). The revised estimate of the effect attributable to the .08 law was a 12.3% reduction in alcohol-involved drivers in fatal crashes.

Finally, the Centers for Disease Control (CDC) and their Task Force on Community Preventative Services more recently conducted a systematic review of studies addressing the efficacy of .08 laws (Centers for Disease Control, 2001). It found that "these laws are effective in reducing alcohol-related motor vehicle crash fatalities." This resulted in a "strong recommendation" that states adopt such laws. The CDC review estimated that the median decrease in alcohol-related motor vehicle fatalities following implementation of such laws was 7%. A recent review by Dee (2001) arrived at similar conclusions, estimating that nationwide adoption of .08 BAC laws would "generate substantial gains, reducing the annual count of traffic fatalities by at least 1,200."

Zero Tolerance Laws. By the 1980's, the legal minimum drinking age for alcohol consumption had been raised to 21 in all 50 United States; however, in a large fraction of those states, the legal BAC limit for persons under 21 remained at .08 or .10, the same as for adults. What have become to be known as zero tolerance laws attempt to bring the drinking driving laws in concert with the drinking laws by making it an offense for persons under 21 to drive with alcohol in their system. Currently, all 50 states now have laws for youth prohibiting driving at a BAC over .00, .01, or .02, depending on the state. The traffic safety impact of such laws (often referred to as "zero tolerance laws") has also been the subject of several studies.

In an early study of zero tolerance laws, Blomberg (1992) conducted time series analyses of Maryland crash data statewide and in six experimental counties in conjunction with implementation of a .02 zero tolerance law and publicizing its existence. The dependent variable was crash involved drivers under 21 judged "had been drinking" (HBD) by the investigating officer. He found an approximate 11% reduction in HBD crashes statewide associated with implementation of the law. For the smaller subset of six experimental counties where PI&E efforts were mounted, the initial reduction observed was 21%, with a further reduction of 30% subsequent to the PI&E effort which emphasized the sanctions associated with violating the law.

Hingson, Heeren, Howland et al. (1991) examined nighttime fatal crashes in four states which passed zero tolerance laws before 1989. They studied adolescent and adult nighttime fatal crashes in four study states and four nearby comparison states. In the study states they observed a 34% post-law decline in nighttime fatal crashes for those under 21 compared to a 7% reduction for adults. Interestingly, in the comparison states the comparable figures were 26% and 9%. They reported that the states adopting the zero tolerance laws had significantly greater relative reductions, adolescents to adults, than the comparison states. Using a similar method for twelve states adopting low BAC laws for youth, Hingson, Heeren, and Winter (1994) reported that in low BAC law states, a 16% reduction in single vehicle nighttime crashes for youth relative to a 5% decrease in adults occurred. The comparable figures for the comparison states were a 1% increase for youth and a 6% decrease for adults. They also reported that states with .00 BAC levels fared better that states that had .02 or .04-.06 levels for youth.

California implemented its zero tolerance law in 1994. The law created an offense which was an administrative violation. It allowed police officers to test with a preliminary breath tester at the roadside those youthful drivers they suspected of driving after drinking. If the driver refused or had a BAC over .01 the officer could then seize the license and issue a citation which permitted driving for 30 day (to allow a hearing request ) after which time a license suspension would be administratively imposed by the DMV. The law was crafted to make enforcement relatively easy and avoid the need to take youthful offenders into legal custody. Voas, Lange, and Tippetts (1998) examined its effect in terms of level of enforcement, perceived risk of arrest among young drivers and effect on fatal crashes. Data used in the evaluation included driver records, random digit dial telephone surveys, interviews with underage drivers in weekend roadside surveys and the FARS file.

The roadside surveys indicated that before July 1993, 10-20% of youth thought the limit was .01 or less and in the six months before the law went into effect that level increased to 50-60%. The telephone survey taken just before the law went into effect indicated approximately 53% thought the level was zero and two years later about 56% thought so. The authors concluded that there was no evidence the law was better understood two years after it went into effect. However, over half of the affectable population were aware of the zero tolerance provision when it was in effect. The authors found little change in the rate of alcohol-related charges for drivers under 21 but rather found a replacement of the more serious DUI charge with zero tolerance violations. Over half of such violations in the post law period were for the less serious zero tolerance offense. Telephone survey questions on perceived risk of arrest showed little change from the immediate pre-law survey to the survey two years later.

Time series of fatal alcohol-related crashes for young drivers were analyzed for California and four comparison states. California adult alcohol-related crashes were also examined as a comparison. A 20.9% reduction on this measure for California youth was observed compared to a 4.1% reduction for adults and an 11.6% reduction for youth in the comparison states. These differences were not statistically significant. The authors suggest that better publicity about the law and stronger enforcement would have resulted in "more substantial results."

To study the effect of zero tolerance laws in all 50 states, Voas, Tippetts, and Fell (1999) conducted a pooled cross-sectional time-series analysis of fatal crashes involving youthful drivers using FARS data from 1982 through 1997. Using this technique they attempted to account for differences among the states in background factors such as urbanization, number of licensed drivers, vehicle miles traveled, characteristics of the vehicle fleet and the like. They also attempted to factor in changes in demographic factors within states, alcohol consumption and effects of other related laws. The overall analysis indicated a reduction in the proportion of underage drinking drivers in crashes of 24.4% attributable to zero tolerance laws. The authors do caution that in an analysis such as this, other factors related to the outcome may have been left out. An example might be that there may have been greater normative changes about drinking and driving which may have effected youth more than adults. Nonetheless, they conclude that zero tolerance laws are having a beneficial effect.

Lacey, Jones, and Wiliszowski (2000) examined the effects of zero tolerance laws in four states, two of which (Maine and Oregon) had had the law in effect since the early 1980's and had recently modified them in an attempt to make them more effective. The two other (Florida and Texas) had only recently enacted such laws. This study examined issues surrounding how the laws were implemented in the four states and the extent to which implementation of the laws has had an effect on alcohol-related crashes as measured by Nighttime Single Vehicle Injury (NSVI) crashes of youth.

In all four states, the administrative license suspension procedures for zero tolerance violations seemed to be running smoothly. Most states integrated the zero tolerance license suspension process into existing administrative license suspension procedures for the adult DWI offense. Youth seemed to request hearings to contest suspensions and request hardship licenses less often than did adults.

In Florida and Texas, enforcement of the zero tolerance law was low but gradually rising. In both of those states, efforts were made from the outset to ease the paperwork burden for officers taking zero tolerance enforcement action. However, as in California above, this initially did not result in an increase in alcohol related arrests for youth.

In Maine and Oregon, which had longstanding zero tolerance laws, the volume of enforcement actions for zero tolerance violations approximated the rate for adult DWI. It was observed that a number of zero tolerance violations were at BAC levels above the legal threshold for adults. However, there was no evidence on the basis of volume that zero tolerance violations were being used instead of DWI for youth as had been observed in California.

Most officers indicated that a permissible level of .00 was preferable to .02 in that it sent a clear message to youth that no consumption of alcohol was legally compatible with driving. This observation is in concert with the findings of Hingson above.

In the states which had longstanding zero tolerance laws, Oregon and Maine, and where police were generally familiar with basic enforcement procedures for the law, recent changes in the law were associated with further reductions in a proxy of alcohol-related crashes. In Maine, where the permissible BAC level was reduced from .02 to .00, a reduction in nighttime single vehicle injury (NSVI) crashes for youth on the order of 36% was observed. In Oregon, where a change in the age for the .00 limit was made from 18 years to 21 years, a NSVI reduction of 40% was observed. In the two states where the basic law was more recently adopted, a much smaller reduction was observed in Florida (5%), and no reduction was observed in Texas.

Hingson, Heeren, and Winter (1998) evaluated a Maine law lowering the legal BAC limit from .10 to .05 for persons convicted of DWI. The authors used a before-and-after × treatment-and-control design. They calculated changes in the proportions of fatal crashes involving drivers with prior DWI convictions from the six-year period before enactment of the law to the six-year period following enactment of the law, and compared Maine with the other New England states. They found that in Maine, the proportion of fatal crashes involving drivers with recorded prior DWI convictions declined 25 percent following passage of the .05 DWI law, while the proportion rose in the rest of New England during the same years. The proportion of fatal crashes involving drivers with recorded prior DWI convictions and illegal alcohol levels also declined significantly in Maine, as did the proportion of fatal crashes involving fatally injured drivers with recorded prior DWI convictions and illegal alcohol levels. Most of the latter decline was due to a decline in alcohol-related fatalities of previously convicted drivers with very high BACs, of .15 percent or higher, at the time of the fatal crash. Each of these declines in Maine was significant relative to the rest of New England.

Enforcement Countermeasures. The 1991 review by Jones and Lacey concluded that countermeasures that have stressed enforcement can have a significant traffic safety deterrent impact, particularly when used in combination with a strong public information and education (PI&E) component. It noted that some of the stronger studies of the effects of enforcement coupled with public information campaigns found reductions in the number of nighttime crashes ranging from 10 to 30%. However, some other strong studies found only small or "possible," but not significant, effects. Wagenaar, Zobeck, Williams et al. (1995) estimated the average percent reduction in evaluations of U.S. countermeasures classified as "enforcement" (which included sanctions as well as enforcement) to be in the 6% to 12% range, again, depending on the post intervention period examined.

Enforcement strategies employing the concept of sobriety checkpoints appear to have been successful in Australia and France, and earlier research in the U.S. suggested they had been an important factor in some DWI programs that combined enforcement with enforcement-strategy specific PI&E. A new examination of 14 of the better evaluations of random alcohol screening (including sobriety checkpoints) since 1983 concluded that random screening was effective in reducing alcohol-related fatalities by 8% to 71% (Peek-Asa, 1999).

Recently, an evaluation of a large-scale checkpoint program in Tennessee was completed, bolstering these earlier results significantly (Lacey, Jones, and Smith, 1999). In March 1994, Tennessee implemented an extensive statewide sobriety checkpoint program called Checkpoint Tennessee. Checkpoints were scheduled on each weekend of the year in at least four counties in the state. On five weekends checkpoints were scheduled in each of the state's 95 counties. The volume of checkpoints increased from about 15 in the preceding year to nearly 900 in the program year. The checkpoint activity was publicized extensively both through public service advertising and earned media. Interrupted time series analyses were used to evaluate the program, indicating that the program resulted in a 20.4% reduction in alcohol-related crashes, extending at least 21 months after conclusion of the formal program. Extensive checkpoint activity was continued after the formal program completion.

A checkpoint program in New Mexico, operated during and shortly after extensive changes to the State's DWI laws, achieved results that were similar to those in Tennessee. Lacey and Jones (2000) evaluated these New Mexico initiatives, and their findings are discussed on page 130.

Foss, Bierness, Tolbert et al. (1997) examined the effects of an intense statewide sobriety checkpoint program implemented in North Carolina in 1994 and 1995. Two statewide blitzes of three to four weeks duration were implemented--one in late November-early December and the other in late June through most of July. These blitzes, which involved over 1,200 checkpoints for the winter blitz and nearly 2,000 for the summer blitz, were accompanied by extensive paid advertising as well as hard news coverage. The effects of the program were assessed through both roadside surveys and interrupted time series analyses of statewide crash data. Overall, the roadside survey results revealed a significant reduction in drivers with BAC's exceeding .08 (from 1.96% to .90%). Surveys were conducted in four communities, and the within-community results were statistically significant only in the two smaller cities. However, the crash analyses did not reveal a sustained reduction in alcohol-related crashes. The authors hypothesized that the effect of such a program is likely to be more pronounced in smaller communities which are more completely saturated by the effort. They also suggested that periodic blitzes may be less effective that a continuous program, but that it may be difficult to obtain sustained media coverage for such a program since the enforcement effort becomes a more routine part of operations.

Jones, Joksch, Lacey et al. (1995) reported the results of a project in three sites to test the hypothesis that combined speed, alcohol, and seatbelt enforcement strategies, coupled with a strong public information and education program (PI&E), can reduce the incidence of speeding, alcohol-impaired driving, and non-use of seatbelts. The three sites were Knoxville, Tennessee; Wichita, Kansas; and Lexington, Kentucky. This project publicized the enforcement of several highway safety laws in combination, rather than enforcement of one particular law. This approach was designed to make enforcement more efficient in raising perceived risk of arrest for each type of violation and also to achieve increased deterrence by creating a perception of more severe penalties for multiple violations occurring in a single incident. It was hypothesized that, as a result, deterrence for one category of violation may be enhanced by the perceived severity of sanctions for another.

Each program was designed to sequentially emphasize five different combined enforcement strategies during a period of approximately one year. A PI&E campaign focusing on each strategy was to operate for about two months. A general program theme was established for all of these campaigns, stressing the concept of simultaneous enforcement of speeding, DWI, and occupant restraint laws. The first program began in September, 1990, and the last program continued through May, 1992.

The evaluation effort was directed at measuring the effect of the combined enforcement / PI&E program on:

A comparison site was used for each test site to help recognize trends that could affect the test site and confound the effects of the program in the test site. The comparison site was chosen so as to match the test site as closely as possible except that it planned no special traffic law enforcement program.

The evaluation showed that the programs that had increased intensity of enforcement of the target laws as well as a strong public information and education (PI&E) program supporting the enforcement effort and its highway safety benefits were effective for DWI and speeding. Programs that did not have both increased enforcement and supporting PI&E for one or both of these two behaviors did not show an effect. The results for the third target behavior, non-use of seatbelts, were inconclusive in that one site (Wichita) with increased enforcement and PI&E showed no effect, while another site (Lexington) with increased enforcement and PI&E was able to maintain its already high seatbelt usage rate throughout its program period. Of the two sites that showed clearly positive results, one (Wichita) had an effect on DWI, and the other (Lexington) had an effect on DWI and speeding.

In Wichita, a variety of analytic methods were used in the analyses of alcohol-related accidents, ranging from visual examination of the data to several kinds of models, including regression models, general linear models, and, finally, interrupted time series models. Terms accounting for trends and seasonal effects were included in many of the models as were terms that acted as "control" variables to account for non-program effects that might have occurred during the time period studied.

The results of the Wichita analyses using the interrupted time series models of proxies of alcohol-related crashes are summarized in Figure 5-2. They indicate reductions in alcohol-related crashes ranging from 20% to 35%. All of these reductions are highly significant, with the probability p that they could be due to chance alone being less than 0.005.

Table 5-1: The Haddon-Brenner Framework 
for Analyzing Traffic-Crash Countermeasures
Type of AccidentReduction in Accidents
as a % of the Mean
Nighttime20% (30/148)
Nighttime Injury21% (14/65)
Nighttime Single-Vehicle35% (24/69)
Nighttime Single-Vehicle Injury23% (7/30)

Lexington had statistically significant reductions in alcohol-crash proxies in the 10% range and also a 12% reduction in the number of vehicles exceeding the speed limit by at least five mph. Further, Lexington also had a reduction in minor injury accidents of about 17%, a possible reflection of these lower speeds.

The authors found the study results to be encouraging but inconclusive as to the overall traffic safety impact of combined enforcement of DWI, speeding, seatbelt usage laws. In two of the test sites it was not possible to consistently maintain increased enforcement activity and associated publicity for all three of the target behaviors. The one site that was able to maintain increased enforcement activity and a high level of publicity for all three behaviors produced positive results against DWI and speeding. This site did not produce any increase in seatbelt usage, but was able to maintain its already high usage rate. Another site that was able maintain increased enforcement and related publicity against just one of the target behaviors, DWI, showed positive results for that target behavior, but for none of the others.

Adjudication and Sanctioning Countermeasures. The 1991 review found that one adjudicative countermeasure aimed at increasing the probability that a charged drunk driver will be convicted of drunk driving (implied consent) had a traffic safety benefit in itself by suspending refusers' drivers licenses. Another adjudicative sanction, deferring prosecution as an incentive for entering a treatment program, was found to be ineffective.

Of evaluated countermeasures focusing on sanctions, the 1991 review found that those that suspended or revoked a DWI's driver's license were clearly the most effective, particularly when well-publicized and applied administratively. One strong study showed that suspending the license of drivers refusing to submit to an alcohol test reduced their crash involvement during suspension, including alcohol-related crashes by about 70%, presumably because most did not drive while under suspension. Several other studies of different degrees of strength showed that suspending or revoking licenses for DWI reduced all accidents as well as alcohol-related accidents during the period of suspension or revocation. Most of the recent literature dealing with driver's license sanctions has been concerned with applications of ALR laws and were discussed in the section above dealing with legislative countermeasures.

The 1991 review found no strong support for the hypothesis that alcohol-related crashes can be reduced by sanctions aimed at treatment and rehabilitation18, although two strong studies found reductions in the re-arrest rate ranging from 10% to 35%. This conclusion applied to programs that dealt with social drinkers and first offenders as well as to programs that deal with persons with drinking problems and with multiple offenders. Further, the review found a "disturbing tendency" for the better designed and executed evaluations to show little or no impact, and for the less rigorous evaluations to show an impact. Nevertheless, the review found that more recent studies continued to confirm past studies indicating that rehabilitative sanctions can be effective when applied in addition to traditional sanctions such as driver's license suspension or revocation.

A later examination of the pertinent literature involved a meta-analysis19 of the efficacy of so-called "remediation" (i.e., treatment and rehabilitation as traditionally defined) with drinking-driving offenders (Wells-Parker, Bangert-Drowns, McMillen et al., 1995). A total of 215 independent evaluations were studied. The methods used in the studies were rated using scales and protocols developed by expert panels. The authors found that better methodological quality was associated with smaller effect size, and that the better studies suggested that treatment and rehabilitation reduced drinking-driving recidivism by an average of about eight to nine percentage points over no treatment and rehabilitation. (This means, for example, that a two-year recidivism rate of 20% would be reduced, on average, to roughly 18%.) A similar effect size was found for alcohol-involved crashes. Their research also suggested, as had prior reviews, that combinations of treatment modalities were more effective than other evaluated individual modes for reducing drinking-driving recidivism.

The above findings of Wells-Parker and associates apply to DWIs in general, including first offenders as well as repeat offenders. The recidivism of repeat offenders was not analyzed separately, although two of the three risk categories studied ("moderate" and "high") appear to have contained significant numbers of repeat offenders. The authors found "...some evidence that ‘moderate risk' offenders –a category that included multiple offenders–might be more responsive to treatment than either severe or low risk offenders but, because risk type was confounded with treatment type..., this finding is only suggestive" (page 924).

Four recent studies were found that examined the effect of treatment and rehabilitation countermeasures. The nature and results of these studies are presented below.

As indicated above, Langworthy and Latessa (1993) evaluated Turning Point, a program in Cincinnati, Ohio, designed to treat and educate chronic drunk drivers. This program was an attempt to limit the period of incarceration and improve the behavior of "chronic" drunk drivers, therefore easing the strain on jails. Program participants had to have served at least 30 days in jail, and then had to complete a 28-day residential program followed by a six months post-release aftercare program. The evaluation sought to determine whether Turning Point subjects performed better than other chronic drunk drivers did after they were released from custody.

The study group consisted of 531 repeat DUI offenders who participated in the Turning Point program during the first 23 months of project operation. The comparison group consisted of 200 repeat DUI offenders who were adjudicated during the same period, but who did not participate in the Turning Point Project. Random assignment to the two groups was not used, making it necessary to use statistical methods (logistic regression) to control for differences between the groups.

The study found that 33% of Turning Point subjects had new charges within the next 18 months, 8% of which were DUI. By contrast, 40% of the comparison group had new charges, 10% being DUI. From this, the authors concluded that the Turning Point project had its intended effect and that Turning Point subjects were more likely to succeed than comparison group subjects. Note, though, that the observed statistical relations were weak, with the Turning Point subjects doing just marginally better than comparison group subjects.

Langworthy and Latessa (1996) did a follow-up of their original study which extended the tracking data to more than four years. The data dealing with Turning Point participants revealed that about 60% had new arrests since release, and that 25% had further DUI arrests. For the comparison group, it was seen that 58% had a new arrest and that 28% had a new DUI arrest. Subjects with three or more prior DWIs did slightly better in relation to the comparison group.

Peck and associates (1994) analyzed the effect of treatment programs for first-offender and multiple-offender DWIs on a number of criterion measures, including post-treatment DWI recidivism. The subjects studied were 7,316 DWI offenders in Sacramento County, California who were randomly assigned to several treatment and control groups following their conviction in the period September 1977 through January 1981. The subjects included 2,685 repeat offenders. These treatments were represented in the study by a system of 0-1 dummy codes following the procedure used the earlier study cited above (Arstein-Kerslake and Peck, 1985). For first offenders, the dummy variables represented the following treatment conditions: (1) no-treatment control; (2) in-class educational program (four 2.5-hour sessions); and (3) home study program. For multiple offenders, the treatments were: (1) no-treatment control; (2) therapeutic counseling; (3) counseling plus chemical therapy; and (4) bi-weekly contacts without counseling or chemical therapy.

The authors found that none of the treatments affected recidivism for first offenders or repeat offenders, concluding that:

... the present study found no evidence even suggesting a positive impact for the first-offender home study program or the multiple-offender biweekly contact (without counseling) program. In addition, none of the multiple-offender treatments produced effects approaching conventional significance levels. (Page 676)

However, the authors cautioned that their conclusions were limited to the specific data analyzed and should not be interpreted as a general conclusion that all DUI treatment programs are ineffective.

DeYoung (1997a) re-examined the effectiveness of California's treatment programs which had undergone some changes since 1981. In 1997, California had three types of outpatient alcohol education and treatment programs. First-offender programs were typically three months in duration and consisted of a minimum of 10 hours education (e.g., the effects of alcohol on the body and on driving, DWI laws, etc.), 10 hours counseling and 10 additional hours of education/counseling. It was also required that the client maintain "close and regular" face-to-face interviews with program staff.

Second offenders (within 7 years) could be sentenced to attend an 18-month "SB 38" (named after the sponsoring legislation) program. SB 38 programs were 18 months in length and required at least 12 hours of education, 52 hours of counseling and bi-weekly face-to-face interviews. Third and higher offenders were required to participate in a 30-month program consisting of a minimum of 18 hours education, 117 hours counseling, 120-300 hours of community service, and "close and regular" face-to-face interviews. In 1997, there were less than 500 annual enrollments in 30-month programs.

The DeYoung study examined the effectiveness of these three levels of alcohol treatment programs, comparing them to other sanctions, singly and in combination with others, which were typically prescribed for DWI offenders convicted in California. Drivers studied were all licensed California residents who were convicted of a DWI in a California court between July 1990 and June 1991. The sample included 88,552 first offenders and 27,293 repeat offenders.

The study found that combining treatment with drivers license action was associated with reduced recidivism for repeat offenders and first offenders as well. For repeat offenders with one prior, the mean number of subsequent DWI convictions within 18 months of the index conviction was .096 for offenders assigned to the treatment program and receiving driver's license suspension or restriction, and .139 for drivers receiving driver's license suspension or restriction alone. Thus, those receiving license revocation alone were about 1.5 times as likely to recidivate as those receiving license revocation and the 30-month program. For repeat offenders with three or more priors, a similar effect was noted, with those receiving license revocation alone having about 1.7 times the risk of recidivating as those receiving license revocation and the 30-month program.

In recent years there has been increased interest in the potential of treatment programs involving brief interventions by medical care providers, particularly with alcohol positive patients presenting at emergency departments because of accidental trauma, often resulting from traffic crashes. These interventions usually involve a brief screening instrument followed by a short counseling session (15-20 minutes) or referral to an alcohol treatment provider for further assessment and possible treatment. Gentilello, Rivara, Donovan et al. (1999) conducted a randomized trial of such an intervention with patients admitted to a level I trauma center. At twelve month follow up they found reductions in alcohol use, injuries and DWI citations. A study designed to assess the potential of screening of trauma patients in emergency departments conducted by Runge, Garrison, Hall et al. (In Press) found, in a randomized trial, that alcohol problem screening and referral of patients in emergency departments tends to increase the proportion of patients seeking follow up assessment and treatment for alcohol problems. Though this approach shows promise there is still reluctance in the medical community to address alcohol problems with patients presenting for other issues. Rivara, Dunn and Simpson (2000) developed a training program for physicians, residents and medical students to train them in implementing a brief intervention and encourage them to do so. They pilot tested the program in the Pacific Northwest.

The 1991 review by Jones and Lacey concluded that the impact of other, more traditional sanction-directed countermeasures remained unclear. For example, one statewide study of incarceration in a jail reported no effect, while another local study of the same sanction reported a positive effect. In fact, circa 1991, there was still considerable controversy about the effect of sanction severity on drunk driving. Three studies suggested that the severity of a sanction may be less important than the certainty of a sanction, while another study found that certain severe sanctions (including jail) were highly effective. The findings of the first three studies are supported by a later study by Ross and Klette (1995) who examined the effect of abandoning mandatory jail sentences for persons driving with BACs above specific limits in two Scandinavian countries. This law change was a part of legal reforms in 1998 and 1990 in Norway and Sweden, respectively. Interrupted time-series analysis found that in both countries, traffic deaths diminished simultaneously with the reforms, suggesting that Scandinavian success in reducing alcohol-impaired driving does not depend upon mandatory jail.

However, a study in Franklin county, Ohio, by Socie, Wagner, and Hopkins (1994) suggests that jail may be preferable to traditional DWI schools for repeat DWI offenders, but that the opposite may be true for first DWI offenders. The authors compared the DWI / alcohol-related crash recidivism of 124 jailed offenders and to that of 218 offenders assigned to a driver intervention program (DIP). After controlling for gender, age, race, BAC, additional charges filed at the time of arrest, and driving history, they found that, overall, DIP attendees had significantly lower rates of subsequent impaired driving after four years than drivers who were given a jail sentence. Drivers with no priors had a significantly higher four-year recidivism rate when jailed than that of drivers enrolled in a DIP (odds ratio = 2.53), while those with priors fared better after their jail sentence (OR = .56). They also found that drivers younger than 21 years of age had high recidivism rate than other drivers (OR = 2.46).

Recent research suggests that certain alternatives to jail can be successful in reducing recidivism. Such alternative sanctions that have been suggested, and some cases evaluated, include referral of drivers to treatment and education, community service in lieu of or in addition to jail, electronic monitoring, intensive supervision probation, impoundment or forfeiture of vehicles or license plates, victim impact panels, victim restitution, visits to a hospital emergency room that treats traffic crash victims, ignition interlocks, and using license plates that identify the vehicle owner as a DWI offender, among other sanctions. Many of these sanctioning concepts do not necessarily require law changes, but may be available under existing laws.

Jones and Lacey (1998b) evaluated a judge-based program in a small county court near Atlanta. This program used a combination of individually tailored traditional and alternative sanctions packaged so as to meet the needs of individual offenders, be they repeat offenders or first offenders. All of the sanctions were available under existing laws. The judge-based program was more effective by a wide margin (two to one) than was another sentencing program in a nearby jurisdiction that imposed the minimum sanctions.

Another program operating in Milwaukee County, Wisconsin under existing laws used intensive supervision probation for repeat DWI offenders. The evaluation showed that the program reduced a moderate alcohol-related arrest recidivism rate by about one-half, from about 11% to 5.5% after one year (Jones, Wiliszowski, and Lacey, 1996). Yet another program (in Los Angeles County, California) used a combination of electronic monitoring and home detention for repeat DWI offenders (Jones, Wiliszowski, and Lacey, 1996). The program reduced a low alcohol-related conviction recidivism rate by about one-third, from 6% to 4% after one year.

A third, more recent, NHTSA-sponsored study (Jones and Lacey, 1999) examined the effectiveness of a day reporting center (DRC), a highly structured non-residential facility that provides supervision, reporting, employment counseling, education and community resource referrals to probationers who had been convicted of a felony DWI. The study found that the DRC program was no more effective in reducing recidivism than was a standard probation program in use by the study jurisdiction. Both programs had a reconviction recidivism rate of about 8% after two years, quite low for this group of offenders. However, the study found that the DRC program was more helpful than standard probation in assisting in the reintegration of the offenders into society and provided correctional services at a significantly lower cost than jail.

Another alternative sanction that has been proposed but not evaluated for DWI offenders is the use of shock incarceration or "boot camps" in which rigid military discipline is imposed. Cowles, Castellano, and Gransky (1995) reviewed evaluations of such programs for non-violent offenders in general, and that included treatment components. They recommended that maximally effective treatment regimens should "include substance abuse education and treatment programs involving psychotherapeutic-based interventions, such as individual and small group therapies, with a focus on multimodal approaches that are relevant to the offender population."

The use of victim impact panels as an alternative sanction was evaluated by Shinar and Compton (1995) who defined a victim impact panel (VIP) as "a group of three or four persons who were seriously injured or whose loved one was killed in a DWI-related crash." The panel members present their personal stories orally to DWI offenders who are ordered by the court to attend the VIP. This study compared the pre-panel DWI recidivism rates with the post-panel recidivism rates of over two thousand DWI offenders who attended the VIPs in Oregon and California. The study also compared these rates to the rates of age-sex matched control groups of drivers. The control subjects were convicted of DWI in the same states at the same time period, but were not ordered to participate in the VIP. In addition, pre- and post-panel DWI convictions were also studied for 683 drivers who were ordered to attend the VIP but failed to do so (No-Shows).

The results showed that, although in Oregon the VIP attendees had a lower rate of recidivism than their matched control group, the recidivism rate of the attendees was not different than that of either those who were ordered by the court to attend the VIP, but failed to do so (No-Shows), or the age-sex matched control group for the No-Shows. In California, there were no differences in recidivism rates between the VIP group and either the no-show group or the two control groups. Other analyses singling out specific age groups and distinguishing males from females suggested that VIPs may be more effective for offenders of age 35 years or more than for offenders of other ages.

The idea of removal of an offender's vehicle (or access to it) as an alternative sanction has been around for some time, but has not been used to any great extent until fairly recently. Several variations on this basic theme have been studied.

DeYoung (1997b) evaluated the effect of two 1995 California laws which provided for the impoundment / forfeiture of vehicles driven by drivers with suspended or revoked licenses (S/R) and by unlicensed drivers20. Data used in the evaluation were obtained from police and court records in four jurisdictions (Riverside, San Diego, Stockton and Santa Barbara) that had record systems which would allow impoundment data to be linked to driver record data in the state DMV database.

The study compared the 1-year subsequent driving records of subjects whose vehicles were impounded with similar subjects (i.e., S/R and unlicensed drivers) who would have had their vehicles impounded, but who did not because their driving offense occurred in 1994, the year before the impoundment / forfeiture laws were implemented. Statistical controls were used to attempt to control potential biases resulting from pre-existing differences between the groups.

The study examined three measures of recidivism:

The effect of impoundment on subsequent DWI convictions per se was not studied, although a DWI conviction would also trigger a license suspension or revocation. The results showed that repeat DWS/DWU offenders who were impounded had 34.2% fewer DWS/DWU convictions, 22.3% fewer traffic convictions and 37.6% fewer crashes, than did similar drivers whose vehicles were not impounded. By comparison, drivers with no prior convictions for DWS/DWU whose vehicles were impounded had, relative to similar drivers whose vehicles were not impounded, 23.8% fewer (DWS)/(DWU) convictions; 18.1% fewer traffic convictions and, and 24.7% fewer crashes.

DeYoung (2000) also examined the general deterrent effect of impoundment on suspended and revoked (S/R) drivers in California. His analysis involved a comparison of the crash rates of all drivers who were suspended or revoked between January 1992 and January 1997 to a 1% random sample of drivers not suspended or revoked during the same period. An interrupted time series design was used, the intervention occurring at January 1, 1995, when the impoundment law went into effect. DeYoung found that there was a statistically significant reduction in crash rates for both groups when the groups were analyzed separately, but that there was no significant reduction for the S/R group (p=0.099) when the series for the non-S/R group was used as an input series. He concluded that the study "failed to find compelling evidence of a general deterrent impact of vehicle impoundment/forfeiture in California." Note that this study examined the effect of the impoundment law on drivers who been S/R for any reason, not necessarily DWI.

Voas, Tippetts, and Taylor evaluated the effects a variation on the impoundment theme, temporary vehicle impoundment and/or immobilization in two counties in Ohio. The period of immobilization provided by the Ohio law is 30 days for the first DWS offense, 60 days for the second and vehicle forfeiture for the third DWS offense. Second DWI offenders are subject to 90 days, and third DWI offenders to 180 days immobilization--and the vehicles of fourth offenders are subject to forfeiture. The law applies both to the vehicle owned by the offender and, if the offender was driving a vehicle owned by someone else, to that vehicle as well.

The first evaluation (Voas, Tippetts, and Taylor, 1997a) was in Franklin County, Ohio, which includes the city of Columbus. The recidivism of groups of drivers who had their vehicles impounded and/or immobilized were compared to groups of drivers who did not have their vehicles impounded and/or immobilized. Random assignment to the experimental group and the comparison group was not possible in the study. Of particular interest here was the DWI recidivism of DWI offenders. The analysis technique used to study the recidivism of these groups (Kaplan-Meier) did not control for differences between the two groups with respect to such variables as income and employment status.

The study found that during the period of impoundment and/or immobilization, 1.8% had committed another DWI offense by the end of their 90-days period of impoundment and/or immobilization. However, 3.8% of the DWI offenders with one prior DWI in the comparison group had committed another DWI offense after 90 days. This reduction (53%) was statistically significant at the .025 level. A similar effect was found for DWI offenders with two prior DWIs, with the comparable recidivism percentages for the two-priors group after 180 days being 2.4% and 6.6%, respectively, an effect (64%) that was significant only at the .094 level.

After the period of impoundment and/or immobilization, the effect size was much smaller and only significant for DWI offenders with one prior, 5.0% vs. 8.0%, an effect size of 38%.

The second study of vehicle impoundment in Ohio by Voas, Tippetts, and Taylor (1997b) was conducted in Hamilton County which kept the vehicles impounded throughout the applicable sanction period. The applicable sanction period for DWI varied by number of prior DUIs, being 90 days for offenders with one prior and 180 days for offenders with two prior offenses. The principal objective of the Hamilton County study was to provide an independent replication of the results of the evaluation of the immobilization law in Franklin County.

The study found that impoundment decreased recidivism by large percentages both during the period of impoundment and after the period of impoundment. For repeat offenders with one prior DWI, the reduction in DWI offenses was 80% during the impoundment period and 56% after the impoundment period. For repeat offenders with two prior DWI offenses, the reductions during and after the impoundment period were 56% and 58%, respectively.

Rodgers (1994) evaluated a Minnesota law that provided for the impoundment of the license plate of DWI offenders with two prior DWIs in five years or three or more prior DWIs in ten years. The law took effect in August 1998, and required that such drivers surrender for destruction the license plates of all vehicles registered in their name. Further, the law stipulated that the violator could not sell any vehicle with impounded plates without permission from the Department of Public Safety. To protect innocent persons who depended on a vehicle from being deprived of a vehicle, the law allowed the violator to apply for a special license plate with a distinctive pattern of characters that can be recognized by police but not by the general public.

Initially, the license was to be surrendered in court, but less than 5% of the offenders who should have surrendered their plates were required to do so by the court. The law was subsequently changed to provide for the administrative impoundment and destruction of the plates by the arresting officer, and by the Department of Public Safety, if the officer did not perform the impoundment. In the first 21 months of the law, there were 6,993 violations to the law and 4,494 impoundment orders, a percentage of 64%.

The impact of the law as amended to provide for administrative impoundment is of particular interest here, since the court-based version was clearly not successful from a practical standpoint. Three groups were studied in the assessment, all of whom had violated the law during the 21 months from January 1991 through September 1992. The groups were composed of:

The analysis examined the recidivism of all three groups. The study found that violators whose licenses were impounded had much lower recidivism rates that did those whose plates were not impounded. The rates for violators with three recorded DWIs and with four recorded DWIs are shown in Table 5-2. The officer-impounded groups had the lowest rates. Three-time violators with officer-impounded plates had about half the rates than did violators with mail-order impounded plates. However, for four-time violators, the rates of the two impounded groups were about the same, and still less than those of the non-impounded group.

Table 5-2: Recidivism Rates After 12 Months and 24 Months
for Three Groups Studed by Rodgers (1994)
DWIs of RecordMonths After OrderGroup Impounded by Mail OrderGroup Impounded by Police OrdersGroup Not Impounded
312 11%8%16%
324 19%13%26%
412 11%10%18%
424 18%17%26%

Finally, we note a study of the effect of vehicle seizure by Crosby (1995) that may not be readily available to some readers. The study examined the recidivism of drivers sanctioned under Portland Oregon's forfeiture ordinance and found that "perpetrators whose vehicles were seized could reliably expect to be rearrested on average half as often as those whose vehicles were not."

Ignition interlocks that test a driver's BAC and prevent driving for those whose BAC exceeds a specified value have been evaluated in a number of jurisdictions. Coben and Larkin (1999) reviewed six of these evaluations published in the 1990 - 1997 time period, extracting the summary data shown in Table 5-3, below.

Table 5-3: Recidivism Effect of Six Evaluations 
of Ignition Interlocks (After Coben and Larkin, 1999)
StudyOutcome MeasuresRelative RiskPrevented FractionP Value
The EMT Group, 1990DWI recidivism .7130%>.05
Popkin et al., 1992DWI recidivism .3864%<.05
Morse and Elliott, 1992Re-arrest for DWI.33 69%<.05
Jones, 1992Re-arrest for DWI .8516%.05
Weinrath, 1995Impaired driving recidivism.40 66%<.05
Beck et al., 1997Re-arrest for DWI  .36 65%  <.05

Cites are in Coben and Larkin, 1999

All six of the evaluations compared the recidivism of a group of repeat offenders who participated in an interlock program to a comparison group that did not participate in an interlock program. The "prevented fraction" column in the table indicates that the interlock participants had recidivism rates that were 16%-69% less than those of the participants in the comparison groups. Note that these figures were for the period during which the interlocks were attached. One of the studies (Popkin et al., 1992) also examined the recidivism rates after the interlock was removed, finding that the recidivism returned to higher levels after removal. Comparison group participants varied widely – they included a those receiving a conditional license (Popkin et al., 1992), those receiving a matched license suspension (Morse and Elliot, 1992), all other DWI offenders in the state (Jones, 1992), license-suspended impaired drivers (Weinrath, 1995), and those given "usual post-licensure treatment" (Beck et al., 1997).

Only the study by Beck and associates assigned its participants randomly to the interlock group and the comparison group. A total of 1,380 persons participated, and their recidivism was tracked for one year after their assignment to the interlock group or the comparison group. Subjects in the interlock group were required to drive an interlock-equipped vehicle for the entire one-year period. The study found that 2.4% of the interlock group had committed an alcohol traffic offense, compared to 6.7% of the comparison group.

Another study of the use of interlocks in Alberta, Canada, also found them effective for second offenders during the period in which they were required to be installed (Voas, Marques, Tippetts et al., 1999). The Alberta interlock program was introduced in 1990 as a required program for repeat offenders, but, according to the paper, was changed in 1994 to a voluntary program for first offenders. In this study, only about seven percent of repeat offenders eligible chose to participate. In addition, all participants (and all comparison group members who did not participate in the interlock program) were required to have their license suspended prior to entry into the interlock program, with the length of the suspension for the participating second offenders having a mode of two years. Persons assigned to the interlock group were free to drive without restriction during the interlock period, 6 to 12 months for first offenders and 12 to 24 months for second and third offenders. Persons assigned to the comparison group remained fully suspended during the interlock period. After the interlock period, the licenses of the interlock group were reinstated, and the licenses of the comparison group were either reinstated or remained suspended.

The Alberta study used survival analysis techniques to compare the recidivism of the interlock participants with that of the non-participants, finding that during the program period, the recidivism of the interlock group was less than 1% after one year, and the recidivism of the suspended comparisons was about 4%. However, during the period that the interlocks were removed, the recidivism of the participants was about 4% after one year, and the recidivism of the still-suspended comparison group was about 3%. Also, during the removal period the recidivism of the reinstated comparison group was about 5%.

It has also been suggested that, rather than impound or confiscate a convicted DWI's vehicle, one might "confiscate" the driver's vehicle registration. Two states, Oregon and Washington, enacted legislation in 1990 and 1988, respectively, establishing a procedure by which law enforcement officers, upon apprehending a driver whose driver's license has been suspended (for DWI or other applicable offense), could take possession of the driver's vehicle registration. In such cases, the driver was given a temporary registration certificate, and a striped ("Zebra") tag was placed over the annual sticker on the vehicle license plate. Officers were allowed to stop a vehicle with such a tag without additional probable cause, thus providing a means for interdicting offenders who continue to drive after suspension. A new annual sticker could only be obtained by the owner demonstrating that he or she was properly licensed. The Oregon law and the Washington law differed mainly in that the latter applied only to suspended or revoked drivers who own the vehicles they are operating, while the former law applied to all suspended or revoked drivers, regardless of whether the driver owned the vehicle.

The specific and general deterrence effects of this "Zebra" Tag Law for DWIs in general (first offenders and repeat offenders alike) and the general deterrence effects for persons not yet arrested were evaluated by Voas and Tippetts (1995), who found a general deterrent effect in Oregon (on moving violations and crashes) but none in Washington. Available data did not permit an evaluation of the effect of the law on recidivism in Washington, but the Oregon data indicated a reduction of more than 50% in mean number of subsequent DWIs. The effect of the laws on repeat offenders was not determined, but the results suggest that the effect of such a tag would not be limited just to first offenders.

Comprehensive Countermeasures. Evaluations of two early comprehensive programs employing multiple Traffic Law System countermeasures, the British Road Safety Act of 1967, and NHTSA's Alcohol Safety Actions Projects (ASAP), were discussed at length in the 1978 update. Because of their significance to the field, these two programs are revisited briefly here.

The program known as the British Road Safety Act of 1967 was actually much more that just a legislative change. The act limited BAC to .08% and prescribed a mandatory punishment of a one-year license suspension and a fine of £100, or imprisonment for up to four months, or both (Ross 1973). The Act also authorized a preliminary breath test when an officer had "reasonable cause" to suspect that a driver had alcohol in his or her body or had committed a traffic violation, or in the event of a crash. Refusal to take the test was penalized by a fine of £50, and if the test after a given stop were refused a second time, the driver was punishable as though the sample had been given and the test failed (Jones and Joscelyn, 1978). The Act made extensive use of a public information campaign to support law enforcement by increasing the perception of apprehension for drunk driving.

The number of tests per month (in Great Britain) went from the neighborhood of 3,000 in 1967 to around 7,000 or 8,000 by early 1971. In all of 1970, approximately 70,000 breath tests were given in Britain. In the three years following the implementation of the Act, conviction rates averaged more than 90%, compared to 80% before the Act.

In one of the better known post hoc evaluations of its time, Ross analyzed the effect of the Act. He used the technique of the interrupted time series analysis (see page 99), and determined that the Act was effective for a period of three years in decreasing the rate of accident casualties. A later analysis of the Act (Department of the Environment 1976) found indications that the effects of the Act may have been more lasting among drivers in the 40-60 age group and that they wore off most rapidly among those in the under-30 age group. Ross also found evidence that the public engaged in less drinking and driving after the Act. A pre- and post-survey of adults found that more people said they walked to drinking sites and fewer admitted to drinking and driving. In addition, among traffic fatalities, there was a smaller percentage who had illegally high BACs.

Ross attributed the effectiveness of the 1967 Act primarily to the public education campaign which led drivers to believe that the chances of their being apprehended when drinking and driving were great. He concluded that after the driving public learned that there was little increased enforcement of drinking-driving laws, they adjusted their estimate of the chances of apprehension accordingly. Consequently, the effectiveness of the Act decreased (Ross 1973).

NHTSA's Alcohol Safety Action Projects (ASAP) was by far the most ambitious and comprehensive of any alcohol-crash countermeasure program ever conducted. It attempted to apply the principles and some of the methods of the systems sciences used so successfully in the aerospace field to the alcohol-crash problem. It began in 1969 when the National Highway Safety Bureau (later the National Highway Traffic Safety Administration) announced a nationwide ASAP program and established an Office of Alcohol Countermeasures to manage it. The program provided financial assistance to and coordinated the efforts of, initially, nine and, ultimately 35, individual ASAPs around the country.

ASAP was heavily oriented toward the problem drinking-driver who, it claimed, was responsible for two-thirds of the alcohol-involved traffic fatalities in the U.S. and 34% of all traffic fatalities. They believed it useful to distinguish between problem drinkers who must drastically change a behavior over which they have little or no control, and social drinkers who need to make only a relatively minor change in their behavior. ASAP also placed emphasis on drinking drivers who drive at night and on weekends. This policy was implemented at most ASAP sites by fielding more police units (e.g., an enforcement task force) during the hours between 6 p.m. and 4 a.m.

Individual ASAP projects were conducted at the local level, emphasizing improved law enforcement, traffic court procedures, public information, and special efforts to counsel and assist drivers. The local activities were complemented by state-level efforts in driver licensing, motor vehicle registration, traffic records and legislation. The first group of nine ASAPs began operation in 1971. Twenty-six additional ASAPs initiated operations in 1972. The locations were widely distributed around the U.S. and included one site in Puerto Rico.

ASAP used a combination of the legal, health, public information and education, and technological approaches in its attack on the alcohol-crash problem. Eventually, five separate categories of countermeasure activities evolved (U.S. Department of Transportation 1975a):

Very large increases in activity occurred in all of these categories. The increases in activity were not, however, reflected in proportionate decreases in night fatal crashes. On the average, night fatal crashes per 100,000 licensed drivers per year decreased from 12.9 to 12.5, or about 3%. However, NHTSA analysts found no correlation between activity and impact. Further analyses of night fatal crashes using the interrupted time series analysis indicated a statistically significant reduction in such crashes for sites with two years operational experience but no significant reduction for sites with only one year of operation. By contrast, the overall trends of night fatal crashes in the reporting states were said to be upward rather than downward.

This first ASAP evaluation of operations lacked control groups as a basis for comparison of the outcomes experienced at the ASAP sites. Zador (1976) attempted to correct this deficiency "by comparing year-to-year variations in fatality statistics between groups of areas with ASAPs and comparison groups of areas without ASAPs." He concluded the ASAP countermeasures could not have been responsible for the observed reductions in the ratio of night fatal crashes to day fatal crashes, and that "ASAPs, as large-scale social programs, have been ineffective". After a subsequent analysis of crash data, NHTSA (Johnson, Levy, and Voas, 1976) expressed a disagreement with Zador's conclusions, prompting another response by Zador (1977) repudiating NHTSA's criticisms.

The 1991 review examined evaluations of several comprehensive programs conducted since the 1978 update, noting mixed results regarding their impact on alcohol-related crashes. Three evaluations found a positive effects, while four others found no effect. However, as with NHTSA's ASAP evaluations, none of the evaluations of statewide programs used other states as a control.

Roger and Schoenig (1994) evaluated the impact of California's 1982 legislative reforms, including enhanced penalties, greater sentencing uniformity, and the introduction of an illegal per se standard. Interrupted time series analysis was used to evaluate the general deterrent effects of these laws, as measured by alcohol-related fatal and injury accident rates, both statewide and in counties sharing similar demographic and enforcement patterns. The formation of Mothers Against Drunk Driving (MADD) as well as the legislation changes were used as intervention variables, and measures of daytime crashes were used as controls. Both interventions were statistically significant, with the legislation associated with an 8% reduction in nighttime injury crashes, and the MADD intervention with a reduction of about the same size. The effect of the legislation extended for five years beyond the implementation of the legislation.

Neustrom and Norton (1993) used a similar time-series analysis to measure the impact of tougher, deterrence-based drunk driving legislation in Louisiana. The study found a statistically-significant 20% reduction in nighttime fatal and injury accidents that lasted 36 months after the new law was implemented. A smaller (5%) reduction was found in daytime fatal and injury accidents, suggesting a positive effect of the new law on drinking-driving.

Recently, Lacey and Jones (2000) evaluated New Mexico's omnibus anti-DWI legislation which was introduced in the later half of 1993 and the beginning of 1994. New Mexico further initiated an extensive statewide DWI checkpoint initiative in December 1993.

The legislation contained provisions for:

Though the specific effect of each individual component on alcohol-related crashes could not be discerned because of multiple, simultaneous interventions, time series analyses indicated an overall reduction on the order of 19% in drunk driving fatal crashes (a crash where one or more of the involved drivers had a BAC of .10 or more) when the period December 1993 through 1995 is compared with January 1988 through November 1993. Crash trends in five neighboring states were examined as a comparison and no such reduction was observed.

Several researchers have continued to explore the concept of community-based countermeasure programs to prevent alcohol-related injuries in general. These programs are hybrid in nature, employing other risk management systems in addition to the Traffic Law System. Holder and associates (2000) recently completed a five-year community trials program in three experimental communities (two in California and one in South Carolina). The project included a component devoted explicitly to drinking and driving, concentrating on the TLS functions of enforcement and public information and education. This involved training local project coordinators and others in "media advocacy," and assisting local law enforcement staff in obtaining funds to bolster DWI enforcement and in designing an improved enforcement program.

The drinking and driving component was evaluated by Voas, Holder, and Gruenewald (1997). The evaluation took particular care to relate project activity to project impact through a chain of intermediate variables, using statistical techniques to do so (see page 100 of this report). This, along with the use of matched comparison communities, helped insure that any positive impact was due to the drinking and driving component, and not to some other factor.

From their report, it appears that the project resulted in substantial increases in the use of breath-alcohol testing devices, enforcement-related news coverage, and the use (monthly) of sobriety checkpoints. The impact evaluation indicated a positive effect on nighttime single-vehicle crashes in the California sites, but none in the South Carolina site. The report estimated a reduction in the number of nighttime single-vehicle crashes in the two California sites of approximately 50 to 80 over the two-year operational period, or (from the report's Figure 1) about 12% to 20% of all such crashes in that period. An analysis of the contributions of the individual variables indicated that only the breath-alcohol testing devices and the enforcement-related news coverage contributed to this positive effect.

A recent paper by Holder, Gruenewald, Ponicki et al. (2000) summarized the results of an evaluation of the complete five-year program (1992-1996), including the four non-drinking and driving components (that is, community mobilization, responsible beverage service, underage drinking preventions, and alcohol access reduction). In this evaluation, nighttime injury crashes and so-called DUI crashes (crashes in which a citation was issued for DUI) were used as measures of alcohol-crash impact. The evaluation indicated a 10% relative decrease in number of nighttime injury crashes per month, and a 6% relative decrease in the number of DUI crashes per month. Note that these decreases were relative to the rates in the comparison communities. In addition, analyses of data from hospital emergency departments indicated that violent assaults also declined during the program period.

Wagenaar and associates (2000) also conducted a community-based intervention in the mid-1990s, but the program had a more specific objective for a more restricted target group, that is, to reduce the accessibility of alcoholic beverages to youths under the legal drinking age. The program, called Communities Mobilizing for Change on Alcohol (CMCA), was a randomized 15-community trial of a community-organizing intervention designed to reduce the accessibility of alcoholic beverages to youths under the legal drinking age. The communities were located in Minnesota and Wisconsin, seven of which were assigned to the program and the remainder to a control group (Wagenaar, Gehan, Jones-Webb et al., 1999). The 15 participating districts were matched on state, presence of a residential college or university, population size, and on the results of the baseline alcohol purchase survey. The average population of the study communities was 20,836 (range 8,029 to 64,797). Communities that were already addressing the issue of underage drinking, or were participating in other major funded efforts, were eliminated from the pool of communities considered for CMCA. The authors found that, not only were the selected communities not necessarily "ready" for the program, but many were resistant to defining youth drinking as a serious problem in their community.

The program used a community organizing approach involving part-time community organizers over a 2½ year period to accomplish changes in the policies and practices of pertinent local institutions, for example, increased patrol time by enforcement agencies and increased media coverage of alcohol-related issues. Specific interventions varied among the seven test communities. Two communities prepared a report form for merchants to record (and report to police) underage buy attempts. Another community worked with a local legislator to draft state legislation to repeal a law exempting alcohol licensees from punishment for serving underage drinkers. Several communities restricted alcohol availability at major community events, such as university homecomings. Other interventions included:

Activities focusing on community awareness included:

The results of an evaluation of the effect of the CMAC program on alcohol-related crashes are described by Wagenaar, Murray, and Toomey (2000). The authors collected annual arrest and quarterly traffic crash data for the year 1987-1995, providing a 6-year baseline and 3 years of data during the intervention. Data were stratified into two target age groups (16-17 and 18-20) and a control group (age 21 and over). The authors observed net declines in the intervention communities for all arrest and traffic crash indicators. The decline was statistically significant for driving under the influence (DUI) arrests among 18-20-year-olds and approached significance for DUI arrests and disorderly conduct violations among 15-17-year-olds. However, the decline in alcohol-related crash surrogates was not significant, possibly because (as the authors noted) of the small numbers of such crashes involving the target group.

Provide Alternatives to Drinking-Driving

Ride service programs (RSPs) have been used for many years to provide alternative transportation for alcohol-impaired persons. Programs vary from those that are organized in conjunction with a major holiday associated with drinking (e.g., New Year's Eve) to programs that operate on a year-around basis. Harding, Apsler, and Goldfein (1988) examined the literature on such programs and identified 325 that had been operating in the U.S. They found that most ride service programs (two-thirds) used taxis to provide the transportation, but that other types of vehicles were also used, including privately owned vehicles (often used in programs run by or for students), tow trucks, buses, vans and even police cars. Some dispatched two drivers, one to provide the ride home and the other to transport the drinker's vehicle. Harding and associates concluded that ride service programs were widespread and relatively easy to set up and operate, but that "Rigorous (impact) evaluations of RSPs were virtually non-existent" and that, for program operators, "the fact that they are delivering rides (even a small number of rides) is evidence enough for them that the program is working" (p. 56).

Molof, Dresser, Ungerleider et al. (1995) examined the effectiveness of two types of RSPs, those that operate on a year-round basis, and those that operate sporadically during perceived high-risk periods. They studied both a year around program in Syracuse, New York and a holiday program that operated between December 25 and January 1 in Minneapolis and St. Paul, Minnesota. They found that, while these programs functioned quite smoothly and there was good awareness of the programs in both service areas, the study did not detect an effect on alcohol-related crashes from either program.

A recent evaluation of an alternative ride program in Aspen, Colorado had a different result, providing evidence that year-around RSPs can, under the right conditions, have a positive effects on alcohol-related crashes (Lacey, Jones, and Anderson, 2000). The program (named Tipsy Taxi) is administered through the Pitkin County Sheriff's Office as a crime prevention program with assistance from Aspen and Snowmass, Colorado Police Departments and the local restaurant association. The program is designed to be easy for an alcohol-impaired person to use for a free ride home instead of driving.

A Tipsy Taxi ride can be initiated in several ways. Most often, a Tipsy Taxi ride is offered by a bar employee or peace officer who identifies a person as needing help. However, a ride can be requested by a bar patron from his or her bartender. A host of a private party can call the police or sheriff for a voucher for a guest who has over-indulged. In this case, the responding officer congratulates the host for caring for his or her guests and arranges the safe Tipsy Taxi ride home.

Once an individual has been identified as a Tipsy Taxi client, the bartender or peace officer responsible for arranging for the ride follows a few simple guidelines designed to create a safe atmosphere for the rider and to minimize liability to the Sheriff's Office and the director.

The program operates 24-hours a day, 365 days a year. Funding has come from regular fund-raising events, mailed solicitations, grants, alcohol license fees, fees for DUI offenders, and the like.

In addition to occasional publicity events, on-going public information efforts have included advertisements in the local newspaper, radio public service ads (in English and Spanish), flyers distributed in rental cars, and hard news coverage about the program.

Interrupted time-series analyses of crash data indicated that injury crashes decreased by 15% in Pitkin County after the implementation of Tipsy Taxi (p=0.020), and that there was no reduction of injury crashes in the comparison counties. Nighttime and fatal crashes also declined coincident with the implementation of the Tipsy Taxi program, but the declines were not statistically significant.

Designated drivers provide another alternative to drinking-driving, and have been widely promoted in public information campaigns. An interesting discussion of the concept and some impediments to its effective use is provided in a recent NHTSA report (Lange, Baker, and Johnson, 2000). We found no evaluations of the traffic safety impact of the concept, but Caudill, Harding, and Moore (2001) studied the characteristics of designated driver users obtained from 1,391 Computer-Assisted Telephone Interviews (CATIs) and from 902 barroom patron surveys. Their analyses showed that designated driver users, compared to non-users, tended to be at-risk, heavier drinkers. For example, logistic regression using the CATI sample indicated that designated driver users were more likely to drink more often outside the home, to achieve higher BACs when drinking outside the home, to ride with intoxicated drivers, and to be heavy drinkers. Analyses using the barroom sample showed that designated driver users tended to be heavy drinkers, and were more likely to drive after drinking and to ride with intoxicated drivers. Additional analyses showed that designated driver users also were more likely than non-users to engage in other behaviors to avoid DWI, including drinking less, waiting to drive until the effects of alcohol diminish, walking home, and staying overnight.

Educate Drivers and Potential Drivers

Most of the evaluative literature in this area deals with school-based educational programs designed to acquaint students with the seriousness of the alcohol-crash problem and ways of avoiding one's own, or others, involvement in such crashes, either as a driver or a passenger. Often, the curricula are a part of a broader course in alcohol education. As observed by Shope and associates (1992), such programs have been concerned primarily with information dissemination and attitudinal change rather than ways of modifying behavior. These researchers cite reviews concluding that programs stressing information dissemination and attitudinal changes were ineffective and lacking in appropriate evaluation techniques. They report the results of an evaluation of an alcohol misuse prevention study (AMPS) curriculum for fifth- and sixth-grade involving over 5,000 students in Michigan.

The program provided training in the immediate effects of alcohol, the risks of alcohol misuse, and social pressures to misuse alcohol. The students were then taught how to deal with these pressures. Sessions were conducted over two years. In the first year, there were four sessions, four weeks apart; while in the second year, there were three "booster" sessions, also one week apart.

The evaluation of AMPS used random assignment of 49 schools in southeastern Michigan to curriculum, curriculum-plus-booster or control groups, with half of each group pretested and all post-tested. Measures focused on susceptibility to peer pressure, internal health locus of control, understanding of the curriculum material, alcohol use and alcohol misuse. Self-reported data were used in the evaluation, but considerable care was taken to acquire valid self-reports. It was found that, after 26 months, there were significant positive effects with respect to knowledge and attitudes, but none with respect to alcohol use and misuse for the treatment group as a whole. However, there were positive effects on alcohol use and misuse among sixth grade students who had prior unsupervised use of alcohol.

Newman, Anderson, and Farrell (1992) also reported promising behavioral effects of an educational program targeted at ninth graders, and more recently, Shope and associates (1996) implemented and evaluated a program for high school students. The Shope paper is of especial interest here, since it contains specific material on drinking-driving. The program it addresses is an alcohol misuse prevention curriculum for tenth-grade students that was developed, implemented, and evaluated through twelfth grade with 1,041 students from four school districts in southeastern Michigan. The students had participated in the prior AMPS program for elementary school students described above. As with the prior program, the curriculum emphasized social pressures resistance training, immediate effects of alcohol, risks of alcohol misuse, and social pressures to misuse alcohol. The curriculum involved five sessions of 45 minutes each.

The evaluation used self-reported data obtained from surveys of the students. Measures of knowledge, alcohol refusal skills, alcohol use and alcohol misuse were obtained. The study found significant positive program effects on alcohol misuse prevention knowledge (p <0.001), alcohol misuse (p <0.02), and refusal skills (p <0.09). Gender differences over time were found on alcohol use, alcohol misuse, and driving after drinking, with boys' rates increasing more than those of girls. The authors concluded that, despite high levels of alcohol use among high school students, a tenth-grade curriculum can result in some desirable effects, cautioning that "creative approaches are needed, however, especially for boys who tend to use and misuse alcohol at rates that increase more steeply than those of girls." Interestingly, exposure to the sixth-grade program, as well as the tenth-grade program, did not result in better outcomes

In a study published just prior to the publication of his report, Shope, Elliott, Raghunathan, et al. (2001) examined the effects on subsequent driving of the AMPS 10th grade program. The study reported the findings of a randomized test of the effectiveness of program among 4,635 10th-grade students, 1,820 of whom were assigned to the intervention group and 2,815 to the control group. Both groups were followed for an average of 7.6 years after licensure, which typically occurred during or shortly after the 10th grade. Outcomes examined included alcohol-related and other serious offenses, and at-fault, single-vehicle, and alcohol-related crashes. The authors found that only serious offenses (which included alcohol-related offenses) had a significant treatment effect (statistically marginal) after adjustment for sex, age, race, alcohol use/misuse, family structure, presence of prelicense offenses, age of driver licensure, and parental attitudes toward teen drinking. The effect was found only during the first year of licensure. Also, two first-year serious offense interactions were found. The positive effect was strongest among the largest subgroup of students, those who were drinking less than one drink per week on average before the curriculum, compared with those who drank more than one drink per week (p = 0.009). The effect was also stronger for the small subgroup of students whose parents had not expressed disapproval of teens' drinking, compared with those whose parents had disapproved (p = 0.004). The authors concluded that their findings suggested that a high school-based alcohol prevention program can positively affect subsequent driving, particularly that of students who do not use alcohol regularly.


The 1991 review of impact evaluations of alcohol-crash countermeasures (Jones and Lacey, 1991a) found that nearly all such evaluations in this category dealt with countermeasures aimed at regulating the availability of alcohol, observing that, among these, "the evaluations of countermeasures that raised the legal minimum drinking age (LMDA) include many that are among the best in the highway safety field." The review clearly showed that controlling the availability of alcohol by raising the LMDA reduced alcohol-related fatal crashes among the affected age group on the order of 9% to 14%.

The 1991 review found that other attempts at controlling the availability of alcohol had rarely been evaluated to determine their traffic safety effect. Such countermeasures included restricting the sales of alcohol (for example, banning happy hours, and having self-testers in bars for determining blood alcohol concentration) and placing increased taxes on alcoholic beverages. The few evaluations of countermeasures in the former group found that none of them had any significant highway safety impact, either directly or indirectly. The evaluations of the latter group were plagued by methodological difficulties, but did suggest that raising the cost of beverage alcohol through increased taxes may have potential.

This review indicates that, since 1991, there have been very few evaluations of programs aimed at preventing excessive drinking prior to driving, and that the programs evaluated have pursued this objective by controlling the availability of alcohol. Toomey and Wagenaar (1999) discuss a range of policy options for preventing alcohol-related problems, including alcohol-related crashes.

One strategy explored during the 1980s was training servers of alcoholic beverages to be more responsible in their serving practices. Two small-scale programs had obtained positive results, encouraging NHTSA to sponsor a larger scale study in two states. The program, which requires three hours on the part of servers and six hours on the part of managers, dealt with the need for responsible alcohol service, preventing intoxication, identifying signs of intoxication, and preventing driving by intoxicated patrons (McKnight, 1987).

The program was field-tested on 245 servers and managers in Louisiana and Michigan. It produced significant improvement in knowledge, attitudes, and self-reported serving practices at both sites. Improvement in reported policy, as well as increases in observed intervention in patron drinking, occurred only in Michigan. However, the impact of the program on alcohol-related crashes was not determined. The program was later expanded to include 876 servers and 203 managers in eight states, with positive results reported in five of the states (McKnight, 1991). Again, the study did not attempt to measure the alcohol-crash impact of the program.

However, Holder and Wagenaar (1994) evaluated the alcohol-crash impact of an Oregon law (effective January 2, 1987) requiring all new applicants for beverage service permits to successfully complete a state-approved server-training course. The bill also required that all persons holding existing alcohol retail licenses or applying for new licenses must complete a management training program. An amendment required existing server permit holders to complete training only on the five-year anniversary when their permit expired. Approximately 20% of existing permit holders in Oregon were trained each year, with all servers trained by the end of 1991.

Classes in server education were conducted by 20 state-certified providers. The one-day training course covered seven areas:

Students paid $20 tuition and $13 for program administration. A standardized written test had to be passed by all students with a score of at least a 70%.

Approximately 36,000 servers and 6,000 owners / managers of establishments licensed to sell alcohol completed the course by the end of December 1988, and by 1994, approximately 13,000 new servers and existing licensed servers seeking their renewal were completing the required training each year.

The evaluation of the Oregon server training program used an interrupted time-series design in which the effect of the intervention on nighttime single-vehicle crashes was analyzed. Two other confounding interventions occurring during the program period, reducing the legal BAC limit to .08 and the passage of new DWI legislation, were accounted for in the analyses, and nighttime single-vehicle fatal crashes in the other 47 continental states were used as an explanatory series. The analysis found that, by the end of the third year of the program, nighttime single-vehicle crashes had been reduced by 23% (t = -2.40).

The server training programs had been stimulated, in part, by the notion that legal liability of alcoholic beverage servers for damages resulting from patrons being served too much alcohol would help promote more responsible serving practices and thus reduce alcohol-related crashes. In an evaluation design similar to the design of the evaluation of server training described above, Wagenaar and Holder (1991) assessed the alcohol-crash impact of the filing of two highly-publicized server liability cases in Texas. The evaluation found a 6.5% reduction in nighttime single-vehicle crashes after the first case (1983) and a 5.3% reduction after the second case (1984).

The nature of server exposure to legal liability nationwide was examined by Holder and associates (1992). The study developed a conceptual model linking pertinent variables, and an expert legal panel was used to identify and rate the major legal factors contributing to server liability. As a result, each state was ranked according to its relative level of liability exposure. States that ranked highest in server liability were found to have more publicity about such liability, greater awareness and higher concern among licensed establishment owners / managers and different serving practices compared to states with lowest liability exposure. As a result, the authors concluded that server liability had a real potential for reducing alcohol-involved problems.

Our literature search failed to find any pertinent recent evaluative research on server liability or stand-alone programs promoting responsible beverage service. However, the community trials programs discussed in the prior section included components with such objectives. The five-year community trials program reported by Voas and associates (1997) and Holder and associates (2000), had a responsible beverage service component, described in detail by Saltz and Stranghetta (1997), as did the youth-oriented program (Communities Mobilizing for Change on Alcohol) evaluated by Wagenaar and associates (2000). The evaluations of both programs indicated positive results overall in reducing alcohol-related crashes, but could not determine the effects of individual non-TLS components such as responsible beverage service.

The use of behavioral tests for use by social hosts has been suggested to help avoiding over-serving of guests at social gatherings. Streff and Kalsher (1990) explored the efficacy of four sobriety tests for use by individuals with little training (one-leg stand, horizontal gaze nystagmus, backward counting, and ruler drop). The tests predicted BAC with statistically significant accuracy, but prediction accuracy varied. Inter-observer reliability was high for each except the nystagmus test. The ruler drop was found to affect self-reported driving decisions most. We found no studies of the effects of such tests on crashes.

The use of self-testing devices for determining one's own BAC has been studied extensively in Australia. Research by Haworth, Bowland, and associates (Haworth and Bowland, 1995; Haworth, Bowland, Vulcan et al., 1997) has demonstrated potential benefits in Australia, but an early evaluation of such devices in the U.S. found that study participants were no more likely to avoid driving under the influence of alcohol than were individuals not exposed to the breath test (Oates, 1976). Further, there was no evidence that participants in the U.S. study moderated their drinking on subsequent occasions. (It has been suggested that placing such devices in bars could result in BAC "contests" to see who could reach the highest BAC.)

The concept known as "Cops in Shops" for restricting alcohol availability for under-21 youth has been incorporated into several countermeasures programs in recent years. Cops in Shops is a program where officers in civilian clothes are stationed in retail outlets. If they observe underage persons attempting to purchase beverage alcohol, they issue appropriate citations. The most recent reported evaluation of this countermeasure is of a youth-alcohol program in Salt Lake City, Utah (Lacey, Wiliszowski , and Jones, In Press). Cops in Shops was a major component of the Salt Lake City program, which also included working with Peer Leadership Teams engaged in such activities as graduation, ribbon week, December anti-drunk driving month activities, and Teen Courts for "adjudicating" drinking violations. A time series evaluation of the program suggested a possible positive impact gradually increasing to a 14% reduction in youth nighttime crashes after three years.

Some new studies of the effect of alcoholic beverage taxes on traffic crashes have appeared since the last state of knowledge update. Ruhm (1996) performed another econometric analysis, examining the impact of beer taxes (and other alcohol-control policies) on total fatality rates of 18 to 20 year olds per capita and per mile driven, and on per capita death rates of the same age group, from nighttime fatal crashes occurring during 1982 to 1988. The alcohol-related policies included minimum legal drinking age for purchases of beer with an alcohol content greater than 3.2%; preliminary breath test laws; dram shop laws; administrative per se laws; implied consent laws; and mandatory jail or community service for the first DUI conviction. Macroeconomic conditions in the states, local efforts, and enactment of other legislation related to highway fatalities were controlled for. Ruhm concluded that most of the laws have had little or no impact on traffic mortality, with increasing the minimum legal drinking age having only an initial significant impact. On the other hand, the author found that raising beer taxes "was associated with consistently robust reductions in highway vehicle fatalities," and concluded that "further legislative activity related to DUIs is unlikely to produce further declines in traffic fatalities, while preferred estimates indicate that raising alcohol tax rates would continue to have significant effects on drinking and driving and traffic fatalities."

Some of these findings are at odds with many other careful evaluations of alcohol-crash countermeasures, for example, the finding with respect to the minimum legal drinking age. This causes serious concerns about the validity of the conclusions on beer taxes. Dee (1999) questioned the effect of beer taxes on traffic crashes after finding an implausible reduction in daytime crashes (which are much less likely to be alcohol-related than nighttime crashes) with increasing taxes.

Manning, Blumberg, and Moulton (1995) examined the sensitivity of alcohol consumption to price (as might be affected by taxes) among light or moderate drinkers. The study used data on alcohol consumption on the 1983 National Health Interview Survey and found that both light and heavy drinkers are much less price elastic than moderate drinkers. The authors concluded that "we cannot reject the hypothesis that the very heaviest drinkers have perfectly price inelastic demands."

Mast, Benson, and Rasmussen (1999) expressed surprise that many studies of alcohol-related traffic fatalities find beer taxes to be an important policy variable, "since beer taxes only have a small impact on consumption and heavy drinkers are the least responsive to prices." The authors found that the tax relationship is not robust across data periods and that it reflects missing-variable biases. They concluded that failure to include determinants of alcohol consumption other than taxes and drinking age, and / or factors that simultaneously determine drinking behavior and political support for alcohol taxes, bias coefficients in the regression equations.

Thus, despite the emergence of some new literature on the subject, the effect of taxes on alcohol-related crashes remains uncertain, with the need for further research clearly indicated.

The Tenth Special Report to the U. S. Congress on Alcohol and Health provides an extensive discussion of restrictions on advertising as a means for reducing alcohol consumption and alcohol-related problems. In general, the discussion indicates little consistent support for any relationship between advertising and consumption or alcohol-related problems, although a recent study by Saffer (1997) concludes that a ban on broadcast alcohol advertising and eliminating the tax deductibility of alcohol advertising could have a positive effect of traffic crashes. This study differs from other econometric studies of on the subject study in that it considers the market-specific price of advertising and uses metropolitan-level data rather than aggregated national data. Again, more research is needed to establish any basis for a national policy for more restrictions on advertising as a traffic-crash countermeasure.


Countermeasures of this type have rarely been implemented in the U.S., and are seldom even discussed in the literature. One such countermeasure would address the impairment of some modalities of vision at relatively low BACs. Failure of an impaired driver to distinguish the edge of a roadway could lead to a run-off-the-road crash. Epidemiologic data indicate that alcohol-impaired drivers are highly over-represented in single-vehicle crashes, especially at night. (See Chapter 4.) This could be due in part to alcohol-impaired drivers having difficulty in recognizing the edge of the roadway. If this is so, making the roadway edge more visible could reduce the number of run-off-the-road crashes by alcohol-impaired drivers. The 1989 update cited research indicating that, compared to a no-edge line condition, a standard 4-inch edge line reduced the simulated vehicle's lateral position error at high driver BACs by about a foot. No effect was found at low BACs. Wide edge lines were associated with additional, although statistically non-significant, benefits. Another approach to prevent "ran off road" crashes is the use of grooved "rumble strips"at the edge of the road, estimated by the U.S. Federal Highway Administration to reduce the incidence of such crashes by up to 70%. Again, no new research was found on the effect of edge warning techniques on alcohol-related crashes.

Drunk Driving Warning Systems (DDWS) have been proposed for alerting drivers and pedestrians of a nearby vehicle driven by an alcohol-impaired driver. If a person in a DDWS-equipped vehicle does not take a critical tracking test (CTT) for alcohol impairment (or fails the test), emergency flashers are triggered, and if the car is driven above a certain speed, the horn honks intermittently. DDWSs have been tested in the U.S. (Snyder, 1984) and Australia (Bodi, O'Connor, and King, 1986), but have not been used operationally and evaluated with respect to their traffic safety impact. The last version tested in the U.S. had a CTT display unit and an electronics module in the trunk that scored test performance, activated the alarm as required, and recorded data (test scores, ignition on, alarms activated, etc.) Tampering countermeasures were included. Snyder concluded that DDWS use is feasible as a sanction or condition of probation, and that people are unlikely to drive a vehicle when its DDWS alarms are activated.


The sparseness of literature on this subject has been noted in prior updates. The 1989 update did not discuss pedestrian countermeasures at all because of a lack of published literature in this area. It cited a 1989 report of a conference session on alcohol-impaired pedestrians, during which experts in the field were unable to cite any new work in this area.

In our review of the more recent literature on the subject for this update, we found only a single U.S. study documenting the design, implementation, and evaluation of a pedestrian countermeasure program (Blomberg and Cleven, 2000), and no studies at all on bicycle countermeasure programs.

The study by Blomberg and Cleven was well-designed and thorough. It dealt with a program instituted in Baltimore, Maryland in 1995. The program used Walk Smart Baltimore as its slogan, and included an extensive public information and education (PI&E) campaign. Retroreflective caps were distributed to persons at risk in zones with large numbers of nighttime pedestrian crashes. Several improvements were made to the roadway environment in the high-risk zones, including warning signs, improved lighting, refreshed crosswalk markings, and removal of items that prevented drivers and pedestrians from seeing each other.

The evaluation used several measures of effectiveness, the one involving a "surrogate" measure of pedestrian-alcohol crashes being the most appropriate for measuring direct program effects. The surrogate crash was defined as one occurring between 7:00 p.m. and 3:59 a.m. on Thursday through Sunday, and involving a male pedestrian between the ages of 30 and 59. Surrogate crashes in the zones where the majority of countermeasures were concentrated decreased by 22% from the baseline period, and stayed about the same outside the zones. A time series analysis of the surrogate measure did not show any significant reduction, but a time series analysis of crashes occurring on treated roads and involving male pedestrians of age 14 and over did show a significant reduction of 16%.


The most comprehensive study of programs for reducing excessive drinking by college students was performed as a part of the Harvard School of Public Health College Alcohol Study. The School surveyed 734 U.S. college presidents and administrators to learn what colleges were doing to prevent binge drinking among students. Respondents were asked to rate the severity of alcohol abuse problems among students and specify prevention programs used to address this problem.

The study found that prevention practices were widespread regarding general education about alcohol, use of policy controls to limit access to alcohol, restricting alcohol advertising at home-game sporting events, and creating alcohol-free dormitories (Wechsler, Kelley, Weitzman et al., 2000). However, specific programs for alcohol education, outreach, and restrictions on alcohol advertising in campus media were less prevalent. Efforts involving more specific functions (e.g., in-house program evaluation), more personnel (e.g., a task force), and more community involvement (e.g., cooperative agreements and community meetings) were less common, and cooperative agreements were rare. Nevertheless, many colleges reported having task forces, and about half were performing in-house data collection.

Many colleges have established substance-free housing as a part of their efforts to reduce student drinking. The nature of substance-free housing programs at 10 colleges and universities in different geographic regions was investigated by Finn (1996) using information based primarily on extended telephone interviews with college administrators. Students gave a number of reasons for choosing substance-free housing, including: 1) avoid roommate problems associated with drinking or drug use (78%), 2) academic issues such as wanting a quiet study atmosphere (59%), 3) parental influence (26%), 4) religious beliefs or preference (22%), 5) family member with alcohol of other drug problem (6%), and 6) recovery from an alcohol or other drug problem (less than 1%).

College administrators recommended the following guidelines for making substance-free housing a success 1) start new programs small to assist in bonding and commitment; 2) involve students in the planning and operation of the program; 3) keep substance-free areas separate; 4) pick the substance-free locations carefully; 5) consider whether to offer special programming; 6) enforce the rules; 7) utilize peer pressure to enforce conforming behavior; 8) and establish recovery housing for addicted students. Responses to the various concerns and objections from the students concerning the establishment of substance-free residence halls are addressed.

Another survey of college anti-drinking programs investigated the characteristics of alcohol education (Flagstad-Kramer, 1997). Six colleges were identified as having programs with a positive impact, and telephone interviews were conducted with senior student affairs officers, health educators, and campus security officers to define the characteristics of these programs. The characteristics which contributed to effective alcohol education programming were determined to be a broad-based or campus-wide effort, student involvement, policy development and enforcement, top-down commitment, programming specific to campus environment, and the use of "teachable moments."

Finally, Mitka (1998) studied programs at the University of Wisconsin, the University of Vermont, and the University of Iowa, and described them in general terms.

Other studies have described and, in some cases, evaluated specific college anti-drinking programs. Researchers at Northern Illinois University (NIU) evaluated a new strategy to prevent binge drinking that focused on changing students' perceptions of the typical drinking behavior of college students (Haines and Spear, 1996; Haines, 1996). The new strategy involved changing the perception of the norm regarding heavy binge drinking and presented a campaign designed to change this perception. The researchers then collected data (self-reported), comparing students' own binge drinking practices and their perceptions of binge drinking college-wide after implementation of this strategy with a strategy involving a "traditional" approach that included three prominent themes: 1) It's OK to abstain; 2) It's OK to drink in moderation; and 3) Heavy drinking / intoxication causes harm to oneself and others.

They found that there was a significant difference between students' perception of typical drinking behavior and the self-reported drinking behavior. Students significantly overestimated the proportion of their peers who engage in heavy or binge drinking at parties. After implementation of the change-in-perception-of-the-norm strategy, the proportion of students who reported heavy or binge drinking as the norm decreased significantly, and the proportion of students who reported binge drinking dropped significantly. Outcome data on more traditional strategies indicated no significant change in drinking behavior.

The UNC program alluded to above followed a similar strategy to that tested at NIU. It put in place a campaign based on the theme "2 out of 3, .00 BAC." The evaluation of the program is still in progress, but preliminary results are positive, indicating that 71% of the students had heard of the campaign, and that 70% understood its message.

Steffian (1998) evaluated another program using a normative education technique designed to encourage more responsible use of alcohol by college students through correcting misperceptions students have concerning campus drinking norms. This study examined the utility of employing a group program designed to challenge college students' misperceptions of college drinking norms. As in the study at NIU, students were assigned either to a normative education group or to a control group representative of "traditional" alcohol education efforts. Subjects in the normative education groups demonstrated more accurate perceptions of campus drinking norms at follow-up periods relative to baseline perceptions. These subjects also demonstrated a trend toward fewer binges and a significant reduction of consequences due to alcohol use relative to baseline, while those in the control group did not. Discriminant analyses revealed the variables "change in perception of the average student's number of drinks per week," "change in approval of moderate drinkers," and "change in approval of heavy drinkers" to be the strongest contributors to the function discriminating between those who decreased their drinking and those who did not.

So-called brief intervention programs have been implemented in some colleges. Ivy (1999) investigated the effectiveness of a brief intervention using motivational enhancement therapy techniques to motivate subjects to reduce intended and actual alcohol consumption. Information was presented and discussions were held on the relationship between alcohol use and sexual assault and between alcohol use and violence in courtship and dating. An evaluation showed reductions in the number of drinks and in the number of people who could be classified as problem drinkers. Also, the heaviest drinkers showed reductions in comparison with light drinkers. However, episodes of binge drinking and intentions about future drinking were not reduced, and motivation to change drinking decreased after the intervention.

Another study (Borsari and Carey, 2000) of brief intervention consisted of a randomized controlled trial of a 1-session motivational intervention for college student binge drinkers. Sixty students who reported binge drinking two or more times in the past 30 days were randomly assigned to either a no-treatment control or a brief intervention group. The intervention provided students with feedback regarding personal consumption, perceived drinking norms, alcohol-related problems, situations associated with heavy drinking, and alcohol expectancies. At a six-week follow-up, the brief intervention group exhibited significant reductions on number of drinks consumed per week, number of times drinking alcohol in the past month, and frequency of binge drinking in the past month.

Hoy (1996) describes a peer education training program at San Antonio College. A campus survey, which was used to direct the development of the training program, showed that students at the college were involved in binge drinking, intoxicated driving, and marijuana use at a significantly higher level than students in other studies of two-year higher education institutions. The program was found to have a positive impact on excessive drinking.

Several researchers in the field have offered advice about which programs or program components should or should not be tried. Broughton (1997) determined the impact of three information methods on health beliefs among drinking college students: (1) perceived threat, (2) perceived benefit, and (3) perceived barrier. The author concluded that the informational interventions had little effect on health beliefs and behavior change, indicating that the Health Belief Model employing these informational methods may not be an appropriate theoretical model for college students.

Erenberg and Hacker (1997) developed an action guide for dealing with college drinking issues. The report has six chapters. Chapter 1 examines drinking and its consequences for college students and others in the community. Chapter 2 discusses alcohol marketing and promotional practices on campus and in campus media, and strategies to reduce irresponsible marketing and service practices at bars frequented by students. Chapter 3 examines laws and policies that restrict alcohol advertising on college campuses. Chapter 4 discusses laws to limit high-risk alcohol promotions. Chapter 5 presents community-based approaches to reduce problems associated with heavy drinking at bars frequented by college students. Chapter 6 provides suggestions for organizing a coalition, gathering information, developing a strategic plan, and taking action to alter bar marketing and service practices.

Keeling (2000) examined two different approaches for dealing with the problem of collegiate binge drinking: 1) an environmental intervention that depends on community health assumptions and seeks to produce definable population-based goals, and 2) a clinical approach using very personalized, direct feedback to influence the behavior of self-acknowledged heavy drinkers. The author suggests that colleges consider "opportunistic prevention" whereby students might be approached during routine brief clinical visits to identify problems associated with alcohol and to engage in more intense areas of prevention. An integrated approach is recommended, focusing on clusters of related, intertwined behaviors, rather than on separate behaviors in isolation. Love (1998) also stressed the need for multiple strategies to address the complex interplay of drinking behaviors, college culture, and environmental factors as they are related to alcohol use and consequences. Love asserts that programs should strive to strengthen protective processes and factors, but more critically, to reduce the impact of identified risk factors.

Robinson (1998) observed that, while the literature indicates that many male undergraduates suffer numerous negative consequences from alcohol usage, little is known about what, if any, consequences female undergraduates experience. After analyzing 1992 and 1993 United States Department of Education CORE Survey data, the author concludes that, despite their predominately occasional and moderate drinking, females were still similar to men in suffering 15 negative consequences, such as memory loss, thoughts of suicide, arrest for drinking and driving, or missing classes. It is also concluded that with little exception, females use less alcohol than males but suffer similar negative consequences, and that, therefore, male and female undergraduates need similar alcohol intervention and prevention programs.

Turrisi (1999) studied the cognitive and attitudinal factors relevant to binge drinking at colleges using data from a sample of 250 psychology students in a moderately-sized northwestern city. The author suggests that educational efforts to prevent binge drinking include opportunities for new students to interact with groups whose values are compatible with non-binge drinking, having peer groups to help students consider pressures for binge drinking; and individuals who have contact with college-bound teens should inform students of college binge-drinking and avoidance strategies.

Riordan and Dana (1998) examined ways in which student affairs officers and the Greek organizations can address drinking problems. They conclude that the Greek letter organizations must themselves confront the problems of alcohol abuse by their members, recommending collaboration across campuses and throughout student and community groups. They note that there are no simple solutions to the problem, and that as students and campuses change, so must the approaches to student alcohol abuse.

Several researchers have examined the use of designated drivers by college students as a means to reducing the traffic-crash consequences of drinking. DeJong and Winsten (1999) surveyed a representative national sample of students attending four-year colleges in the United States to learn whether they had served as or ridden with a designated driver within the past 30 days and how much alcohol they had consumed the last time they had used this strategy. Among those who had consumed alcohol in the past year, 36 percent said they had served as a designated driver in the past 30 days; of these, 40 percent said they usually binged when they drank but had not done so the last time they served as a designated driver, with the majority either abstaining or having one drink. Among drinkers, 37 percent reported riding with a designated driver in the past 30 days; of these, 22 percent said they did not usually binge but did so the last time they had a designated driver because they had consumed one or more extra drinks. The authors concluded that "among college students, using designated drivers is now a well-established strategy for avoiding impaired driving."

Meier and associates (1998) performed three experiments designed to test methods for increasing participation in designated driver programs at drinking establishments. Experiment 1 involved young adults at two bars in university communities in the western United States. Graduate and undergraduate observers monitored the activities of designated drivers. Incentive programs were established to encourage and reward designated driving, promoted through a print advertising campaign. A significant increase in designated driving was observed in Bar 2, but only a slight increase in Bar 1. Experiment 2 was a systematic replication of Experiment 1, but conducted in a larger urban-suburban setting, and with a significantly smaller proportion of college students. The bars also were considerably different from each other and from those in Experiment 1. Results were very similar to Experiment 1. Experiment 3 incorporated a variation in advertising the program, using television rather than newspaper ads and modeling them on a game show. The incentive program and ads had a statistically significant positive impact on the number of designated drivers. An important and unanticipated result was that the bar had a 350 percent increase in sales of nonalcoholic beverages. However, no binge drinking effect of the program was observed. The authors concluded that designated driving can be increased through incentives and advertising, and has the potential to play an important role in reducing DWI and alcohol-related traffic accidents.

In a third study of designated drivers among college students, Raimondo (1998) investigated the alcohol use behavior of college students who use a designated driver compared with students who do not. Data were collected from a questionnaire survey of 243 undergraduate seniors at a private, four year, liberal arts college in the northeast United States. The author found a significant difference in the quantity of alcohol consumed by male students who utilized a designated driver compared to male students who did not use a designated driver, with the heavier drinkers being less likely to use a designated drivers. White males were found to be most likely to engage in binge drinking but were least likely to use a designated driver.


At the millennium, we find that the literature reporting scientific evaluations of alcohol-crash countermeasures deals overwhelmingly with countermeasures using strategies of deterrence and incapacitation carried out by elements of the criminal justice system, that is, the Traffic Law System (TLS). Whether this is reflective of the extent of the total societal effort directed toward the problem, or is more indicative of the conduciveness of these types of countermeasures to evaluation cannot be said. For example, advocacy groups such as Mothers Against Drunk Driving (MADD) mounted a very strong effort against alcohol-impaired driving during a time when many of the evaluated deterrence / incapacitation countermeasures were in place. A large part of this early MADD effort was publicity-oriented and did not involve activities to influence passing new laws, enforcing laws more effectively, and convicting and sanctioning drunk drivers. The effectiveness of this effort and, no doubt, other efforts, has not been examined quantitatively in the scientific literature21.

At any rate, the evaluations that have been performed clearly indicate that many TLS-based countermeasures have been effective in reducing alcohol-related crashes. These include those seeking general deterrence as well as specific deterrence. Among those that were primarily legislative is nature, laws establishing administrative license revocation (ALR) been found to reduce alcohol-related fatal crashes among arrested DWIs by up to 40%. This adds support for prior research showing the effectiveness of driver license sanctions in general.

A consensus is growing that laws which lower the legal BAC limit from .10 to .08 do result in reductions in alcohol related fatalities. It is clear that such laws work best in conjunction with enforcement, publicity and other laws which are intended to make the TLS work more smoothly--such as ALR laws. Estimates of effects vary but there is mounting evidence that a reduction in alcohol related fatalities on the order of 6-8% might be expected in conjunction with the passage of such a law.

Zero tolerance laws, which in effect prohibit driving with any alcohol in one's system for persons under 21, have been shown to result in reductions in youth alcohol-related fatal crashes ranging from zero to 40%. The key elements for this law are also effective enforcement and increased awareness of the law. States where DUI or DWI cases are essentially diverted into zero tolerance cases and there is no net increase in youth alcohol-related driving citations are less likely to receive the potential beneficial effects of such laws

Recent evaluations of comprehensive changes in State laws that have been accompanied by activities to implement those laws have shown positive results in reducing surrogates of alcohol-related crashes from 8% to 20%.

With respect to TLS countermeasures that are enforcement-oriented, recent research strengthens prior findings that programs using multiple strategies (including sobriety checkpoints and saturation patrols) and supported by strong public information and education efforts can be effective in reducing various surrogates of alcohol-related crashes by some 10% - 30%. Strong evaluations of large-scale checkpoint programs alone indicate positive impacts on alcohol-related crashes of the order of 20%.

Countermeasures focusing on sanctions for drivers convicted of DWI have received considerable attention in the recent evaluative literature. Earlier research on countermeasures that attempted to treat and rehabilitate DWIs had found mixed results, with some programs reporting recidivism reductions of up to 35%, and other programs achieving no reductions. More recent research provides evidence that treatment and rehabilitation combined with driver's license suspension can be more effective than suspension alone, obtaining recidivism reductions in the 30% range.

No recent research was found on the use as jail as a sanction for DWI, but earlier research reported conflicting results, with some evaluations indicating a positive effect and others no positive effect. However, there has been considerable recent research on the effectiveness of sanctions used as an alternative to jail. Three forms of such sanctions have been found to be effective in reducing recidivism for DWI: intensive supervision probation, electronic monitoring, and sanctions tailored expressively for individual offenders. For these sanctions, 33% - 50% recidivism reductions over traditional sanctions have been found. Other alternative sanctions that show promise but need further evaluation are day reporting centers, boot camps with strong treatment programs, and possibly, victim impact panels.

A number of evaluations of vehicle-oriented sanctions have been conducted in recent years. Of those that require the vehicle to be impounded or seized, recidivism reductions of 50% or more have been found. A similar effect has been noted for just seizing and destroying the offender's license plates. Ignition interlocks that prevent an offender from starting his or her car have also been found to reduce recidivism (by up to 69%) during the period in which interlocks are attached, but the effect disappeared after the interlocks were removed. The use of specially marked license plates for DWIs also reduced recidivism in one state.

Other countermeasure programs have not relied totally on the TLS, but have used it in conjunction with other risk managements systems. Two recent community trials programs had TLS components, and one these evaluated the TLS component which was composed primarily of enforcement system support. The evaluation indicated a positive impact on alcohol-related crash surrogates in the 10% - 20% range. The other evaluation dealt with a similar program of enforcement system support and suggested a reduction in alcohol-crash surrogates for the targeted under-21 age group.

Several countermeasures not involving the TLS have been evaluated. One of these was a large-scale, long-range prevention program consisting of school-based education in avoidance of alcohol-related problems. Self-reported data suggest reductions in alcohol use and misuse, and a recent study suggests a positive effect on alcohol-related crashes as well. Another education-oriented program to has just been developed, this time to enable medical practitioners to identify and help patients with alcohol problems, but its impact on alcohol-related crashes is not yet known. Ride-service programs providing a transportation alternative to drinkers have been evaluated recently in two locations, one evaluation indicating no effect on crashes, but the other, an extensive full-time program extending over several years, indicating a 15% reduction in injury crashes. Ride service programs have been found to raise public awareness programs of the alcohol-crash problem even where no direct effect on crashes was found.

Of programs aimed at limiting the availability of alcohol, legislation raising the legal minimum drinking age has been the most extensively evaluated, with results that clearly indicate a reduction of 9% to 14% in alcohol-related fatal crashes for the affected age group. Evaluations of other early attempts to limit alcohol availability have not provided convincing evidence of any positive impact on alcohol-related crashes, but preliminary evaluations of alcohol-server programs have appeared promising. More recent evaluations provide evidence of alcohol-crash impact of server training, especially if used as a component of a broad community-based program or mandated by law. Evaluations of behavioral tests that help servers and social hosts identify alcohol-impaired guests also suggest potential impact on alcohol-related crashes. Finally, there is some hard evidence that more vigorous enforcement of alcohol sales to minors through such programs as "Cops in Shops" can reduce alcohol-related crashes for under-21 drivers. The traffic safety impact of alcohol self-testers in drinking locations has yet to be evaluated in the U.S.

Several countermeasures aimed at reducing the probability of a crash involving an alcohol-impaired driver have been suggested and even tested in field experiments, but their possible impact on crashes has not been determined. Including among such countermeasures are wider and better-marked road edges, rumble strips ro warn drivers they have left the roadway, and drunk driver warning systems to alert roadway users of the presence of an alcohol-impaired driver.

The above discussion has dealt with alcohol-crash countermeasures aimed at alcohol-impaired drivers. However, as indicated in Chapter 2, a significant number of fatalities involve non-drivers such as pedestrians and bicyclists, Unfortunately, there have been very few evaluations of such countermeasures. However, one recent evaluation of a multi-faceted pedestrian countermeasure program in Baltimore, Maryland indicated a positive effect on alcohol-related pedestrians crashes. The program included an extensive public information and education (PI&E) campaign, reflective caps for persons in high-risk zones, and improvements to the roadway environment in the high-risk zones. The evaluation indicated a reduction of 16% -22% in surrogates of pedestrian-alcohol crashes.

The recent scientific literature on programs to reduce excessive drinking among college students (and resultant alcohol-related crashes) does not reveal many program evaluations, although many colleges have such programs in place. We found four types of programs that had been the subject of an evaluation:

Three evaluations of normative programs were identified, all indicating a positive effect on excessive drinking. The evaluations of the alcohol education programs (at six colleges) could be more accurately described as assessments and also indicated a positive effect on excessive drinking. The one evaluation of a peer education program found a positive effect, but the brief intervention evaluations obtained conflicting results, one indicating a positive and the other indicating no effect. While none of these evaluations measured drinking-driving or alcohol-crash impact, the normative programs seemed to offer the most potential. Also, other non-evaluative studies suggest that the programs that have a component that promotes the use of designated drivers could reduce the prevalence of drinking-driving.

In short, the literature on anti-drinking programs for college students provides considerable information on the nature and location of such programs, but every little information on the effects of such programs, especially their effects on the alcohol-crash problem. In addition, many suggestions are offered for the content of programs that could be of value in selecting programs to evaluate in future projects, including:

Some of the limitations to alcohol-crash countermeasures discussed at the beginning of this chapter should be kept in mind when interpreting the above findings on effectiveness. It is especially important to note that many evaluations have been conducted against a background of other anti-DWI activity, making it difficult to ascribe an observed effect to any single countermeasure. This is especially true of multi-state evaluations of legislative countermeasures where the evaluations often not have analyzed the level of activity supporting the implementation of the countermeasure in the various States. Clearly, when considering the simultaneous application of more than one countermeasure, one cannot assume that their total impact would be the sum of their individual impacts. For programs involving many countermeasures, this could lead to the absurd conclusion that the program would eliminate more than 100% of the alcohol-crash problem.

Another limitation is that most of the evaluations were of countermeasures that were implemented in just one or a few jurisdictions. Thus, it cannot be assumed that such a countermeasure would be effective in every jurisdiction, regardless of local conditions. Similarly, a finding of no effect in a one or two jurisdictions does not necessarily mean that the countermeasure would be ineffective in every jurisdiction. And of course, the lack of any evaluation at all also does not necessarily indicate that a countermeasure is ineffective.

  1. As used here, the term "treatment and rehabilitation" includes DWI / alcohol education programs requiring attendance by drivers convicted of DWI. 
  2. Meta analysis is the use of statistical methods in literature reviews to compare and synthesize the findings of studies.
  3. The study was also published as a NHTSA report: DeYoung, DJ. (1997). An evaluation of the specific deterrent effect of vehicle impoundment on suspended, revoked and unlicensed drivers in California. Washington, DC: National Highway Traffic Safety Administration.
  4. Rogers and Schoenig (1994) considered the formation of MADD (not separate MADD initiatives) as an intervention in their evaluation of California's 1982 DWI legislation (See page 130).