2 - Method


Our evaluation was composed of two components, a process evaluation and an impact evaluation. The process evaluation sought to determine the effect of the State’s lower BAC law on the State’s DWI control system. This included the nature and extent of the system’s support to enforcement of the selected State’s law. The impact evaluation sought to determine the deterrent effect of the law on the traffic crash involvement of convicted DWIs. Deterrent effect was measured in terms of general deterrence (reducing DWI among convicted DWIs in general), and specific deterrence (reducing DWI among convicted DWIs who have been caught and convicted for another DWI violation). The flow of the evaluation was as follows:

  1. Select a case-study State.

  2. Perform a process evaluation of the effect of the selected State’s lower BAC law on its DWI control system.

  3. Analyze the general deterrent effect of the selected State’s law.

  4. Analyze the specific deterrent effect of the selected State’s law.

  5. Synthesize the results of steps 2 through 4 into an overall evaluation of the selected State’s law.

In this approach, we were looking for a case-study State that, among other things, had a law that seemed to have had a general deterrent effect. We wanted to find out, if possible, why a law worked rather than why it didn’t work. Thus, in step 1 above we did a preliminary general deterrence analysis in States with lower BAC laws. In step 2, we performed a more detailed impact evaluation of the lower BAC law in the selected case-study State, Maine.

State Selection

This activity involved identifying candidate case-study States and then selecting a case-study State from these candidate States.

Candidate States

To identify States with lower BAC limits for DWI offenders, we first examined NHTSA’s most recent digest of State laws (U.S. Department of Transportation NHTSA, 2001). We then searched State statutes in all of the contiguous United States to identify any pertinent laws that might not have been captured in the NHTSA compendium. As result, five States were found with lower BAC laws, viz.:


Maine was the first State to pass a lower BAC law for convicted OUI offenders. The law lowered the per se limit for convicted offenders in two steps, the first step from .10 to .05 (in 1988) and the second step from .05 to .00 (in 1995). Adjudication and sanctioning are performed through an administrative process. Maine has a .08 per se law for OUI and an administrative per se law for drivers with a BAC of .08 or higher.

Conviction of a first-offense OUI results in an administrative license suspension under the condition the offender not drive with a BAC of .05+ (first step) or .01+ (second step) for one year after license reinstatement. The minimum period of license suspension is 90 days for a first offense, but increases to 18 months for a second offense within 10 years, and to four years for a third and subsequent offense within 10 years. In the second step, the period of suspension for violating this condition was raised from two months to one year and to two years for test refusal if there were probable cause the person was driving at the lower limit. Reinstatement of the driver license after suspension under the lower BAC law is in the form of a conditional license issued by the Bureau of Motor Vehicles (BMV), which stipulates the violator must have satisfactorily completed an alcohol educational program and, when required, has satisfactorily completed an approved alcohol treatment or rehabilitation program. The text of the statute is contained in the Appendix to this report.


Tennessee has a .10 per se law. For drivers with no prior alcohol-related offenses, its presumptive limit is .10 for the offense of DUI and .08 for the lesser offense of driving while impaired (DWI). For drivers with a prior DUI or DWI, its presumptive limit for DUI is .08.


Since 1989, Wisconsin’s illegal per se law has had a BAC limit of .10 for persons with one or fewer prior DWIs, a limit of .08 for persons with two priors, and a limit of .02 for persons with three or more priors. A prior study of the effectiveness of alternative sanctions for repeat DWI offenders (Jones, Wiliszowski, and Lacey, 1996) found only a small percentage of Wisconsin repeat offenders had three or more priors.


Utah’s illegal per se law has a BAC limit of .08 for first offenders and, since 1998, a lower BAC limit for repeat offenders. The new law establishes a zero limit lasting for two years for repeat offenders with one prior and for six years for repeat offenders with two or more priors. However, our contacts in Utah tell us their law has not been enforced and only one case involving the law has arisen.


When we were selecting a study site, Connecticut’s per se limit was .10 for drivers with no prior alcohol-related offenses2 . The State’s lower BAC law for convicted DWIs reduced the per se limit from .10 to .07 for repeat DWI offenders.

Selecting a Study State

In accordance with the above discussion, our major State-selection criterion was that the law in the case-study State likely had a general deterrent effect. To estimate the extent of any general deterrent effect, we performed preliminary interrupted time-series analyses of fatal crashes in candidate States to determine (1) whether drivers had a DWI conviction prior to their crash, and (2) whether drivers with a prior conviction experienced a post-law reduction in fatal crash involvement. Data from NHTSA’s Fatality Analysis Re-porting System (FARS) were used in the analysis because it contained a field indicating whether a crash driver had a DWI conviction in the past three years. The measure of effectiveness was number of crash-involved drivers with such a prior conviction as a percentage of all crash-involved drivers. Data covering the period 1975 through 2000 were used in the analyses.

In Maine, we found a near-significant reduction (p=0.059) of 82% in the Maine series at the 1988 intervention point and continuing on through the remainder of the series but no statistically significant reduction at the 1995 intervention point3. We also performed an analysis that included a comparison series, which we constructed from pooled data from two similar adjoining States, New Hampshire and Vermont. This analysis con-firmed the finding of a positive effect of the 1988 law and little or no effect of the 1995 law. However, note Maine’s implementation of its standard .08 per se law for the general public (and other repeat offender provisions) also occurred in 1988 and these could have been at least partially responsible for the reduction noted above.

Our analysis of similar time series in the four other States with lower BAC limits for repeat offenders (Tennessee, Wisconsin, Utah, and Connecticut) followed the same approach as the Maine analysis. We found both Tennessee and Wisconsin had large reductions in this measure after their laws became effective. The reductions amounted to 37% for Tennessee (p=0.080) and 27% for Wisconsin (p=0.089). (These reductions were not statistically significant at p=0.050 level, the relatively high p levels for such large effects being due to the large variances in the data.) No reductions were found in Utah or Connecticut.

Thus, only Maine, Tennessee, and Wisconsin met our requirement for providing evidence of a positive effect on crashes involving drivers with a conviction of DWI. We chose Maine as the study State for this project, mainly because it would give us the opportunity to establish whether a limit of .05 or .00 can reduce fatal crashes among convicted offenders. Studying the effect of the relatively weak laws in Tennessee or Wisconsin would only provide an additional study of .08 (except in Wisconsin, which has an .02 limit for offenders with three or more priors), which has already been shown to be effective for DWIs in general. Finally, contacts in Maine were agreeable to providing us the data we need for the study, particularly driver records data needed for the analysis of specific deterrence.

Study State

Process Analysis

The process evaluation was tied to a functional description of the Maine implementation of the law. The description indicated the functions performed in implementing the law and the procedures followed in performing those functions. The sequencing and interdependencies of the functions were indicated through narrative descriptions and a flow chart. Because the Maine law is administrative in nature (see Appendix A), primary emphasis was placed on the enforcement function and BMV-related functions.

Changes to the operational environment of the law were also documented as a part of the system description, for example:

Most of the information required for the program description came from discussions with people in the several agencies having major involvement in implementing the lower BAC law, specifically:

Quantification of the activities described in the functional description followed. This involved stating to the extent possible the levels of performance obtained for each major function. For enforcement, measures of activity sought included:

To quantify these measures, detailed arrest and BAC data for both first and repeat of-fenders for a period both before and after initiation of the new BAC limits were sought. Information of interest included age of offender, prior record of offender, BAC (or refusal) at time of arrest, and charge filed.

The data elements to create the process data set for performing these analyses were obtained primarily from the BMV, which maintains a consistent driver history file that includes information on OUI offenders and administrative driver license actions state-wide.

In addition to collecting and analyzing the objective data described above, we queried law enforcement personnel about such issues as:

In general, law enforcement officers and supervisors were asked whether problems have arisen in detecting and charging violators for the new offense, and if so, whether any remedies were identified; what new policies were promulgated; what new training was implemented; and what other solutions were implemented or are needed.

Another issue we addressed in these informal discussions was the extent to which patrol officers are aware of the law and the procedures in place for its implementation and whether they support the concept.

The analysis of BMV-related functions addressed the extent to which the lower limits law has affected BMV operations and what actions the BMV has taken to respond to these demands. Issues of concern included:

The impact of the law on sanction providers such as treatment providers, probation departments, and jail administrators was also investigated. In general, we were interested in identifying changes in policies, procedures, and additional burdens may have resulted from the lower limit law, for example, additional reporting requirements to the BMV, and determining whether the lower limit violation violated the conditions of probation for a prior offense and reporting such a probation violation to the sentencing judge.

Finally, we examined the extent to which OUI offenders were aware of the law, along with the nature of any related public information and education (PI&E) programs involving hard news coverage, public service announcements (PSAs), and other media. With respect to awareness among OUI offenders, alcohol assessment staff from DEEP contractors queried OUI offenders during the intake process. In addition to data on law aware-ness, the assessment staff sought data on offender characteristics such as age, sex, and number of prior convictions, and we used the resulting data to determine awareness as a function of these variables.

Impact Analysis

As indicated above, the impact analysis was composed of two parts, an analysis of the law’s general deterrent effect and an analysis of the law’s specific deterrent effect. Time series analysis of FARS data going back to 1975 was the primary technique used in the general deterrence analyses. The main criterion variable used was percentage of convicted-offender drivers in fatal crashes. Fatal crashes were used because FARS includes a field indicating whether the driver had a prior OUI in the previous three years.

Regression to the mean is a well-known threat to the validity of the before-and-after experimental design. A random variable with a number of values above the mean before some point will eventually have a number of values below the mean after that point due to chance alone. The effect can be amplified when a trend is present. To overcome this problem, we used a time-series design. Our outcome measure was number of fatal -crash-involved drivers who were convicted offenders as a percentage of all fatal-crash-involved drivers. For a comparison series, we used pooled data from two similar adjoining States, New Hampshire and Vermont. Because of the small numbers of convicted-offender drivers in fatal crashes, and the need for enough points to perform a valid time-series analysis, we settled on a series of semi-annual data, giving us 50 data points for each series.

Additional analyses were also conducted in Maine using the Generalized Linear Model (GLM) method to determine the impact of the law on of percentage of convicted-offender drivers with a given BAC (.01+ and .10+) in fatal crashes. Three time periods were used in the analyses: 1982-1987 (no lower BAC law), 1988-1995 (a .05 limit for convicted offenders), and 1996-2000 (a .00 limit for convicted offenders). The GLM method was used instead of time series because of the small number of this type of crash in Maine.

Our analysis of the specific deterrent effect of the lower BAC limit was based on a special form of the before-and-after experimental design. The recidivism of five cohorts of drivers with prior OUIs was studied to see if the rates changed significantly after the implementation of the latest lower BAC limit in 1995. Each cohort consisted of all drivers convicted of an OUI-related offense in a given year, two cohorts before 1995 (1993 and 1994), one cohort during 1995, and two cohorts after 1995 (1996 and 1997). The analysis controlled analytically for period-to-period changes in the composition of the cohorts with respect to subject age, sex, and number of prior OUIs. Data for the analyses were extracted from driver records provided by the BMV. Information for identifying individual drivers was removed from the records by the BMV prior to delivery of the data to Mid-America.

2The per se limit in Connecticut was reduced to .08 in July 2002.
3The details of the time series analyses for Maine are provided in Chapter 3.