Research 38 (2007) 493–499www.nsc.org

☆ This study was funded by the Robgrant 051809 to Alexander C. Wagenand conclusions are solely the authors'JD for helping collect and code the legcontributions to the statistical methods⁎ Corresponding author. Tel.: +1 352 2E-mail address: wagenaar@ufl.edu

www.elsevier.com/locate/jsr

0022-4375/$ - see front matter © 200doi:10.1016/j.jsr.2007.06.001

Journal of Safety

Effects of Legal BAC Limits on Fatal Crash Involvement: Analyses of 28States from 1976 through 2002☆

Alexander C. Wagenaar ⁎, Mildred M. Maldonado-Molina,Linan Ma, Amy L. Tobler, Kelli A. Komro

University of Florida, College of Medicine, Department of Epidemiology & Health Policy Research and Institute for Child Health Policy,1329 SW 16th St. Rm 5130, Box 100177, Gainesville, FL 32610–0177 USA

Received 23 March 2007; accepted 13 June 2007Available online 29 September 2007

Abstract

Problem: Hundreds of laws have been implemented in the United States over the past few decades designed to reduce alcohol-impaireddriving and the crashes that often result. One approach has been to lower the legally allowable alcohol concentration for drivers. Weexamined the effects of changes in legal BAC limit in 28 U.S. states from January, 1976 to December, 2002. Method: An interrupted time-series quasi-experimental design was used, incorporating non-alcohol-related crashes as comparisons. Four outcome measures of alcohol-related crash involvement were examined: single-vehicle nighttime, BAC=0.01–0.07, BAC=0.08–0.14, and BAC≥0.15. Missing BAC testresult data were handled by using multiple imputations. Analyses involved estimation of state-specific ARIMA models, controlling for otherfactors affecting overall crash rates and other major DUI policy changes. Inverse variance weighting methods were used to pool results acrossstates for the most precise underlying estimate of effect of legal BAC limits. Results: Considerable state by state variability in estimatedeffects was observed, but results from the pooled analyses were clear and consistent. Changes in legal BAC limits significantly affectedalcohol-related fatal crash involvement for both the SVN and BAC test result measures, and the laws affected drivers at all drinking levels.Summary: An estimated 360 deaths are prevented each year in the United States as a result of the move from a 0.10 to 0.08 legal limit inrecent years, and an additional 538 lives could be saved each year if the United States reduced the limit to 0.05, consistent with limits in mostcountries worldwide. Impact on Industry: Given the significant effects of lower legal BAC limits on fatal crash involvement, businessesshould support implementation of laws that further reduce the legal BAC limit for all drivers. Furthermore, all companies should set higherstandards for employees, such as a zero allowable BAC limit for driving during work time.© 2007 National Safety Council and Elsevier Ltd. All rights reserved.

Keywords: Alcohol; Automobile crashes; BAC limits; ARIMA time series; Policy evaluation

1. Problem

Driving after drinking or driving under the influence ofalcohol (DUI) is known to be a major contributor to trafficcrashes and fatalities. While the United States has made

ert Wood Johnson Foundation underaar, Principal Investigator. Findings. The authors thank Cindy Silianoff,al data and Darin Erickson, PhD for.65 0111x85683; fax: +1 352 265 8047.(A.C. Wagenaar).

7 National Safety Council and Elsevier Ltd

substantial progress in reducing alcohol-related traffic fatali-ties in recent decades, approximately 40% of traffic fatalitiesinvolve alcohol, resulting in approximately 17,000 lost liveseach year (National Highway Traffic Safety Administration[NHTSA], 2006b). Of all alcohol-related traffic deaths, 86%were killed in crashes where at least one driver or non-occupant had a blood alcohol concentration (BAC) of .08 g/dlor higher (NHTSA, 2006b). Given the high social costs ofdriving after drinking, hundreds of statutory changes across the50 U.S. states to deter the general population from drivingwhile impaired by alcohol have been implemented, of whichreductions in BAC limits for driving are one type.

. All rights reserved.

494 A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

Early U.S. BAC legislation (late 1960s–early 1980s)established limits for which an individual was presumed tobe intoxicated. These presumptive and prima facie lawsgenerally set the BAC limit to 0.10–0.15 g/dl and alloweddrivers to rebut their BAC levels with other evidence,potentially allowing drivers with BAC levels over the legallimit to escape criminal liability. It was not until 1983 thatUtah passed the first 0.08 per se law, which established a0.08 g/dl BAC limit for a violation, without regard to otherevidence of intoxication. By limiting the use of evidence bydefendants, these more stringent per se laws increased thelikelihood for conviction (Cowan & Joffie, 2002). Otherstates soon followed, with many additional states loweringthe legal BAC limit after passage of federal legislation inOctober of 2000, which created strong financial incentivesfor states to lower their legal limit to 0.08 g/dl (23 U.S.C.§163). At present, all states have enacted 0.08 per se laws(Alcohol Policy Information System, 2006).

Fell and Voas (2006) recently reviewed the literature on theeffects of legal BAC limits for drivers, with a particular focuson 14 studies of the experience in the United States. Previousreviews have included studies across many developedcountries (Chamberlain & Solomon, 2002;Mann,Macdonald,Stoduto, Bondy, & Shaikh, 1998). The clear majority ofstudies show reductions in alcohol-related traffic crashesassociated with reduced BAC limits, but the magnitude ofestimated effects vary considerably. For example, reducing thelimit from 0.10 to 0.08 g/dl in the United States is associatedwith crash rate reductions ranging from 3% (Eisenberg, 2001)to 40% (Johnson & Fell, 1995). Part of the variability inestimated effects may be due to differing jurisdictions includedin studies, but studies by different investigators within thesame state have also produced varying results. Rogers (1995),for example, did not find a significant reduction in alcohol-related traffic fatalities following California's implementationof a 0.08 BAC limit, whereas Research and EvaluationAssociates (1991) found a 12% reduction. Similarly, in NorthCarolina, one study (Foss, Stewart, & Reinfurt, 1998) did notshow a significant reduction in alcohol-related fatal trafficcrashes following the implementation of the law, while Apsler,Char, Harding and Klein (1999) did.

Tippetts, Voas, Fell, and Nichols (2005) recentlypublished a quality study that represents the most compre-hensive evaluation of the effects of lowering the legallyallowable BAC for drivers to .08 in the United States from1982 to 2000. Examining 19 jurisdictions (18 states plus theDistrict of Columbia) using state-specific ARIMA interven-tion models, they found significant reductions in drinkingdrivers in fatal crashes in 9 of the 19 jurisdictions, and effectsin the hypothesized direction in an additional 7 (although notstatistically significant). Pooling across the 16 states wherereductions were observed, they estimate a 15% averagereduction following a change in legally allowable BAC from.10 to .08 g/dl.

Our study reported here is similar to Tippetts et al. (2005)in many ways, but also has some important differences. We

examined the effects of legal BAC limit changes insignificantly more states (n=28; every statutory change inBAC limit across the 48 contiguous U.S. states betweenJanuary, 1976 through December, 2002, excluding Massa-chusetts, Oregon, Utah, Washington and West Virginia,where BAC limit changes occurred simultaneously withother major DUI control policies, making it impossible todistinguish effects of the BAC limit change alone).Estimated effects of state traffic safety laws are consistentlyfound to vary from state to state. Such variance may be dueto a large number of measured and unmeasured differencesacross states, may be due to the high degree of stochasticresidual error when models are based on the low fatalitycounts in many states, or may be deemed the result of aninherent state-level random effect, with each state treated asone case of implementation of a specific law. As largernumbers of states are included in a single study using aconsistent analytic method, the precision of policy-effectestimates improves.

Another difference from Tippetts et al. (2005) is inoutcome measures. They based their results on one outcomemeasure—counts of drivers with BACN0 in fatal crashes,while we stratified our effect analyses by BAC level andconducted confirming analyses using single-vehicle night-time crashes, an indicator not requiring imputation methodsfor the substantial number of cases where BAC test resultsare missing.

2. Method

We used a quasi-experimental design, examining 324repeated monthly counts of the outcome measures, in mostcases for at least several years before and several years afterimplementation of a change in legal BAC level. Within eachstate, a comparison series of non-alcohol-related fatalcrashes was included in estimating models to efficientlycontrol for the effects of many other factors affecting theoverall level of fatal crash involvement in a state (e.g., trafficlevels, changing vehicle mix, auto safety standards, safetybelt laws, speed limit changes, and many others). Strengthsof the design include a long series of observations, simul-taneous control of effects of many other factors affectingcrash frequencies, and multiple replications across 28 states.The result is a design of high internal validity, reducedchance observed effects are due to idiosyncratic patterns inone state or during a single short time period, and reasonablyhigh confidence in causal interpretation of observed effects(Shadish, Cook, & Campbell, 2001).

Previous research has shown relatively high rates of errorsin commonly used published lists of alcohol laws, mostnotably errors in effective dates or in specific dimensions ofsuch laws (La Fond, Toomey, Rothstein, Manning, &Wagenaar, 2000). To minimize measurement error, weused original legal research to identify, locate, and reviewthe original text of relevant statutes or session laws in the 50U.S. states changing legal BAC limit policies from 1976 to

Fig. 1. Effects of BAC laws per 0.01 g/dl increase in legal limit on single-vehicle nighttime crashes. Note: Asterisk denotes significant at pb .05, two-tailed; figure shows 95% confidence intervals.

Table 1Changes in Legal BAC Limits for Drivers by State

State Year Month BAC Limit

AL 1995 8 0.1 to 0.08AR 2001 8 0.1 to 0.08AZ 2001 9 0.1 to 0.08CA 1990 1 0.1 to 0.08CT 2002 7 0.1 to 0.08FL 1994 1 0.1 to 0.08GA 1983 9 0.1 to 0.12

1991 7 0.12 to 0.12001 7 0.1 to 0.08

ID 1984 3 0.08 to 0.11997 7 0.1 to 0.08

IL 1997 7 0.1 to 0.08IN 2001 7 0.1 to 0.08KS 1993 7 0.1 to 0.08KY 2000 10 0.1 to 0.08MD 1981 7 0.15 to 0.13

1988 7 0.13 to 0.12001 10 0.1 to 0.08

ME 1988 8 0.1 to 0.08MO 2001 10 0.1 to 0.08MS 1981 7 0.15 to 0.1

2002 7 0.1 to 0.08NC 1993 10 0.1 to 0.08NE 2001 9 0.1 to 0.08NH 1981 6 0.15 to 0.1

1994 1 0.1 to 0.08NM 1994 1 0.1 to 0.08OK 2001 7 0.1 to 0.08RI 2000 7 0.1 to 0.08SD 2002 7 0.1 to 0.08TX 1999 9 0.1 to 0.08VA 1994 7 0.1 to 0.08VT 1991 7 0.1 to 0.08WI 1982 5 0.15 to 0.1WY 2002 7 0.1 to 0.08

495A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

2002. We began by reviewing each state's laws regardingDUI that were in effect as of January 1, 1976. Afterestablishing the 1976 baseline for each state, laws weretracked forward in time to determine all changes in eachstate. For recent years, electronic legal data bases were used(e.g., Westlaw and the Alcohol Policy Information System),while law library hardcopy archives of statute publicationswere used for earlier years (Table 1).

Previous research has found fairly consistent significanteffects of changes in license suspension policies (Chaloupka,Saffer, & Grossman, 1993; Voas, Tippetts, & Fell, 2000;Wagenaar & Maldonado-Molina, 2007), and varying effectsof legal changes in other DUI penalties from state to state(Voas et al., 2000; Wagenaar et al., 2007; Wagenaar, Zobeck,& Williams, 1995). Therefore, we controlled for the possibleeffects of changes in such other state-specific DUI counter-measures when estimating effects of statutory changes inallowable BAC for drivers by including parameters for suchother law changes in each state-specific model.

Outcome data on alcohol-related traffic crashes are fromthe Fatality Analysis Reporting System (FARS) maintainedby the National Highway Traffic Safety Administration

(2006a). FARS collects information on every traffic crash inthe United States that results in at least one fatality within 30days of the crash. Because the current study focuses onstatutory provisions that may affect driver (not passenger)behavior, all outcomes are counts of drivers involved in fatalcrashes (as opposed to counts of fatalities or counts of fatalcrashes).

Two measures of involvement in alcohol-related fatalcrashes were created. The first is the number of driversinvolved in single-vehicle nighttime (SVN) crashes for eachmonth for each state. A crash was defined as single vehiclenighttime if it occurred between the hours of 9:00 p.m. and5:59 a.m. and included only one moving vehicle and nopedestrians or bicyclists. Previous research has establishedthat the majority of SVN crashes involve alcohol (Sub-ramanian, 2003), and number of vehicles involved and timeof day have been reliably and consistently recorded overmany decades. Counts of multiple-vehicle daytime (MVD)crashes for each month for each state were also created, as anindex of crashes least likely to involve alcohol.

Beginning in 1982, breath or blood alcohol concentration(BAC) measures collected from drivers in vehicles involvedin fatal crashes were included in FARS. We created countsfor each month for each state for drivers in four distinctcategories of BAC levels: zero BAC, BAC between 0.01 and

Fig. 2. Effects of BAC laws per 0.01 g/dl increase in legal limit on low BACcrashes (.01-.07 g/dl). Note: Asterisk denotes significant at pb .05, two-tailed; figure shows 95% confidence intervals.

Fig. 3. Effects of BAC laws per 0.01 g/dl increase in legal limit on mediumBAC crashes (.08-.14 g/dl). Note: Asterisk denotes significant at pb .05,two-tailed; figure shows 95% confidence intervals.

496 A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

0.07 g/dl (representing moderate levels of alcohol impair-ment below current legal limits for driving a motor vehicleby adults), BAC between 0.08 and 0.14 (clearly impairinglevels of alcohol above current legal limits for driving), andBAC greater than or equal to 0.15 (very high levels ofalcohol consumption before the crash). As with MVD, thenumber of drivers with zero BAC was used as a covariate inall statistical models using BAC outcome measures tocontrol for important variability in fatal car crashes unrelatedto alcohol.

In addition to being available only after 1982, some statesand years have a substantial proportion of cases for whichBAC test results are missing (Hedlund, Ulmer, & Northrup,2004). To avoid biases due to missing data, we analyzed 10multiple imputations of BAC levels for drivers with missingdata (Schafer, 1997). Imputations for BAC level of driversfor each case missing a BAC test result were previouslycreated using a two-stage General Linear Location Model(Subramanian, 2002), where a logit model distinguishedBAC=0 and BACN0, and a regression model predictedexact BAC for cases where BAC was predicted to be greaterthan 0. Thus, we have 10 repeated counts for each state foreach month for the four BAC measures.

Interrupted time series models (i.e., Box-Jenkins ARIMAmodels) were used to estimate effects of state-specific legal

BAC limit. All models simultaneously controlled for all non-alcohol-related factors affecting fatal crash rates (using zeroBAC and multiple-vehicle-daytime crash indicators), andcontrolled for other major DUI policy changes that previousresearch indicates may significantly affect alcohol-relatedcrash rates. All analyses were limited to drivers age 21 andover, to avoid confounding by the known effects of manypolicy changes in recent decades specifically affectingdrivers under age 21 (e.g., legal drinking age, separate zeroBAC limits for youth). Exact model error structures variedslightly from state to state, but most included seasonal (lag12) differencing and moving average factors, and first-order(lag 1) differencing and moving average factors. Each modelwithin each state was developed independently, using theconventional identification, estimation, and diagnosticstrategy of Box and Jenkins (1976). We manually developedand analyzed the adequacy of every model; no automatedsoftware-driven selection procedure was used. All finalresults reported here are based on combined ARIMA-transfer function models with white noise residuals, pro-ducing unbiased error estimates.

Final models were estimated separately for each of the 28states for each of 4 outcome measures, plus each state's modelfor each of the three outcomes based on BAC test results was

Fig. 4. Effects of BAC laws per 0.01 g/dl increase in legal limit on high BACcrashes (.15+ g/dl). Note: Asterisk denotes significant at pb .05, two-tailed;figure shows 95% confidence intervals.

497A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

estimated 10 times, each repeated model estimation using oneimputation. Thus, 868 ARIMA-transfer function models wereestimated. The models repeatedly estimated due to multipleimputation of missing BAC test results were combined usingSAS Proc MIANALYZE (SAS Institute, 2002) to obtainparameter estimates and standard errors that take into accounteffects of missing data and their imputation. In addition toreporting state-specific results, we combined estimates acrossstates by weighting each state's estimate by the inverse of thevariance (Lipsey & Wilson, 2001). This common meta-analytic technique directly weights each estimate by itsprecision, and produces an optimum overall single estimateof effect (and its standard error) across all the states.

3. Results

Estimates from the state-specific analyses vary widely, andthe majority are not statistically significant (Figs. 1 2 3 and 4).Many estimates are in the hypothesized direction, but othersare not. Based on such a pattern, some might be quick toconclude that BAC limit laws have little effect. However,pooled results across states show quite clearly the significantand substantively important effects of these laws. The varianceweighted pooled estimate of the effect of a change of 0.01 g/dl

in the legal BAC limit is 0.228 (SE=0.018) for the SVNoutcome measure, 0.047 (SE=0.020) for crash involvementamong drivers with BAC=0.01−0.07, 0.049 (SE=0.026) forfatal crash involvement among drivers with BAC=0.08 –0.14, and 0.153 (SE=0.042) for fatal crash involvementamong drivers with BAC≥0.15. Estimate units are frequencyof fatal crash involvement per month per state. All fouroutcome measures are in the hypothesized direction—anincrease in crash involvement for each unit (i.e., 0.01) changein legal BAC limit. Three of the four are clearly statisticallysignificant (pb .05) and one approaches significance (p=0.07).The magnitude of the effect estimate using the surrogate SVNoutcome indicator is very similar to the magnitude of theoverall effect using BAC test results (0.228 for SVN vs. 0.249for the sum of the three BAC categories). Finally, themagnitude of the estimated effects for the BAC measuresincreases monotonically by BAC level, consistent withexpectations based on the epidemiology of alcohol-relatedfatal crash risks.

4. Discussion

The coherent and consistent pattern of results for thepooled results across states confirms that changes in legallyallowable BAC limits significantly affect the overall burdenof fatal car crash involvement. Moreover, our results showclearly that the legal BAC limit affects drivers at all levels ofdrinking, from severely impaired drivers at 0.15 g/dl or over,to drivers with more modest levels of alcohol impairmentreflected in BACs of 0.01 to 0.07 g/dl (under the currentlegal limit). All levels of alcohol consumption carryincreased risk for crash involvement (Compton et al.,2002; Zador, Krawchuk, & Voas, 2000), and a policy thatreduces risk not only among the heaviest drinkers but alsolighter drinkers who nevertheless remain at increased risk offatal crash involvement provides a large aggregate publichealth benefit. In terms of a practical metric for themagnitude of these effects, our results indicate that therecent move from 0.10 to 0.08 legal limits in the UnitedStates prevents 360 deaths per year. A further reduction inthe legally allowable driver BAC from 0.08 to 0.05 wouldprevent an additional 538 deaths per year in the UnitedStates. Most countries worldwide have legal limits fordriving of 0.05 or lower.

Results show considerable state-level variability in theestimated effects of traffic safety laws such as the BAC limit.Such variability is consistent with the literature, whereestimated effects vary considerably from study to study,including a few studies that report no effects. The studyreported here, using all available study states, confirms ahigh degree of variability in estimated effects across states.Our own inclination is to be very cautious in interpretingdifferences between individual states, because it appears thatmany of the differences may reflect the inherent stochasticvariability in fatal crash rates across time and across statesdue to what are, statistically speaking, very small counts in

498 A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

most states. Nevertheless, there are potentially fruitfulavenues for further research on such variability in policyeffects across states.

The most obvious plausible reason for state-by-statedifferences in effects are differences in implementation—aconcept that here includes many possible dimensions ofspecific legal provisions, enforcement and administrativeprocedures, and resource allocations. Second, effects of agiven law such as BAC limit may vary based on the extantconfiguration of a series of other state-specific laws,regulations, and practices that may interact with the BAClimit policy to shape alcohol-impaired driving behaviors of astate's residents. Evaluating the effects of the large set ofpossible combinations of alcohol-impaired driving counter-measures is complex and subject to the risk of capitalizing onchance. However, specific theory-based combinations ofclosely related policies are worth evaluating (such as thecombination of administrative license suspension policieswith lower BAC limits). Because there are so manydimensions of DUI control policies that vary from state tostate, further research involving empirical assessments ofunderlying profiles or structures of state DUI policies usinglatent variable and latent class analytic approaches (Lanza,Collins, & Lemmon, in press) might be fruitful. Suchmethods may well find that the dozens or hundreds of cross-state and cross-time differences in DUI control policiescluster into a much smaller set of policy profiles. States couldthen be grouped by such profiles and differential alcohol-related crash rates across states and over time examined, tofurther our understanding of the optimum policy profiles orcombinations for prevention of alcohol-impaired driving.

Finally, the extent of state-level variability in estimatedpolicy effects found in the current study demonstrates theutility of a large number of replications across states andlong-term follow-ups of policy effects. Cumulating theevidence across dozens of states and many years producesprecise estimates of effects. Single-jurisdiction studiesprovide major contributions to the literature when evaluatinginnovative policies that have not yet spread to many states(or countries), but they are best confirmed with multiplereplications and longer follow-ups to ensure the experiencein the initial states was not idiosyncratic and to increaseprecision of policy effect estimates.

5. Summary

State statutes that reduce the legally allowable BAC limitfor drivers result in decreased fatal traffic crash involve-ment. The laws reduce crash involvement among drivers atall levels of drinking—very heavy drinking driverssubstantially over the legal limit, those just over the legallimit, and drinking drivers under the legal limit. Reductionsin the legal limit in the United States over the past decadeare saving hundreds of lives per year, and further mortalityreductions are likely if the United States reduces the legallimit further, in line with the lower limits in most other

countries. Conclusions are strengthened by analyses of alarge number of individual states that changed the legallimit, use of long data series including hundreds of repeatedobservations before and after a legal change within eachstate, controls for other factors affecting crash rates, andpooling of estimates across states using meta-analyticmethods to obtain more precise estimates of effect. Furtherresearch on patterns and causes of state-level variability inpolicy effects is warranted.

References

Alcohol Policy Information System. (2006). Blood alcohol concentrationlimits: Adult operators of noncommercial motor vehicles. RetrievedOctober 30, 2006, from www.alcoholpolicy.niaaa.nih.gov

Apsler, R., Char, A. R., Harding, W. M., & Klein, T. M. (1999). The effectsof .08 BAC laws. Washington, DC: U.S. Department of Transportation,National Highway Traffic Safety Administration.

Box, G. E. P., & Jenkins, G. (1976). Time Series Analysis: Forecasting andControl. San Francisco: Holden-Day.

Chaloupka, F. J., Saffer, H., & Grossman, M. (1993). Alcohol-controlpolicies and motor-vehicle fatalities. Legal Studies, 22, 161−186.

Chamberlain, E., & Solomon, R. (2002). The case for a 0.05% criminal lawblood-alcohol concentration limit for driving. Injury Prevention, 8(Suppl. III), iii1−iii7.

Compton, R. P., Bloomberg, R. D., Moskowitz, H., Burns, M., Peck, R. C., &Fiorentino, D. (2002). Crash risk of alcohol impaired driving. In D. R.Mayhew & C. Dussault (Eds.), Proceedings of alcohol, drugs and trafficsafety- T 2002: 16th international conference on alcohol, drugs andtraffic safety, August 4–9, 2002, Vol. 1 (pp. 39–44).

Cowan, J., & Joffie, S. (2002). Proof and disproof of alcohol-induceddriving impairment through blood alcohol testing. American Jurispru-dence Proof of Facts, Vol. 3d (pp. 229).

Eisenberg, D. (2001). Evaluating the effectiveness of a 0.08% BAC limit andother policies related to drunk driving. Stanford, CA: Stanford Institutefor Economic Policy Research, Stanford University.

Fell, J. C., & Voas, R. B. (2006). The effectiveness of reducing illegal bloodalcohol concentration (BAC) limits for driving: evidence for loweringthe limit to .05 BAC. Journal of Safety Research, 37(3), 233−243.

Foss, R. D., Stewart, J. R., & Reinfurt, D.W. (1998). Evaluation of the effectsof North Carolina's 0.08% BAC law.Washington, DC: U.S. Departmentof Transportation, National Highway Traffic Safety Administration.

Hedlund, J. H., Ulmer, R. G., & Northrup, V. S. (2004). State laws andpractices for BAC testing and reporting drivers involved in fatal crashes.Washington, DC: National Highway Traffic Safety Administration, U.S.Department of Transportation.

Johnson, D., & Fell, J. C. (1995). The impact of lowering the illegal BAC limit to.08 in five states in theU.S.Paper presented at the 39th annual proceedings forthe Association for the Advancement of Automotive Medicine, Chicago, IL.

La Fond, C., Toomey, T. L., Rothstein, C., Manning, W., & Wagenaar, A. C.(2000). Policy evaluation research: measuring the independent variables.Evaluation Review, 24(1), 92−101.

Lanza, S. T., Collins, L. M., & Lemmon, D. R. (in press). PROC LCA: ASAS procedure for latent class analysis. Structural Equation Modeling.

Lipsey, M. W., & Wilson, D. B. (2001). Practical Meta-Analysis. ThousandOaks, CA: Sage Publications.

Mann, R. E., Macdonald, S., Stoduto, G., Bondy, S., & Shaikh, A. (1998).Assessing the potential impact of lowering the legal blood alcohol limitto 50 mg% in Canada. Ottawa, Canada: Transport Canada.

National Highway Traffic Safety Administration [NHTSA]. (2006). FatalityAnalysis Reporting System Data Files 1976–2002. Washington, DC:National Center for Statistics and Analysis.

National Highway Traffic Safety Administration [NHTSA]. (2006). TrafficSafety Facts: 2005 Data.Washington, DC: National Center for Statisticsand Analysis.

499A.C. Wagenaar et al. / Journal of Safety Research 38 (2007) 493–499

Research and Evaluation Associates. (1991). The effects following the implemen-tation of an 0.08 BAC limit and an administrative per se law in California.Washington, DC: National Highway Traffic Safety Administration.

Rogers, P. (1995). The general deterrent impact of California's 0.08% bloodalcohol concentration limit and administrative per se license suspensionlaws. Sacramento, CA: California Department of Motor Vehicles.

SAS Institute. (2002). SAS/STAT User's Guide, Version 8. Cary, NC: Author.Schafer, J. L. (1997). Analysis of Incomplete Multivariate Data. London:

Chapman and Hall.Shadish, W. R., Cook, T. D., & Campbell, D. T. (2001). Experimental and

Quasi-experimental Designs forGeneralized Causal Inference.NewYork:Houghton Mifflin.

Subramanian, R. (2002). Transitioning to multiple imputation – A newmethod to estimate missing blood alcohol concentration (BAC) values inFARS. Washington, DC: U.S. Department of Transportation.

Subramanian, R. (2003). Alcohol involvement in fatal crashes 2001.Washington, DC: U.S. Department of Transportation.

Tippetts, A. S., Voas, R. B., Fell, J. C., & Nichols, J. L. (2005). A meta-analysis of .08 BAC laws in 19 jurisdictions in the United States. Ac-cident Analysis and Prevention, 37(1), 149−161.

Voas, R. B., Tippetts, A. S., & Fell, J. (2000). The relationship of alcoholsafety laws to drinking drivers in fatal crashes. Accident Analysis andPrevention, 32(4), 483−492.

Wagenaar, A. C., & Maldonado-Molina, M. M. (2007). Effects of drivers'license suspension policies on alcohol-related crash involvement: Long-term follow-up in 46 states. Alcoholism: Clinical and ExperimentalResearch, 31(8), 1399−1406.

Wagenaar, A. C., Maldonado-Molina, M. M., Erickson, D. J., Ma, L.,Tobler, A. L., & Komro, K. A. (2007). General deterrence effects of U.S.statutory DUI fine and jail penalties: Long-term follow-up in 32 states.Accident Analysis and Prevention, 39(5), 982−994.

Wagenaar, A. C., Zobeck, T. S., & Williams, G. D. (1995). Methods used instudies of drink–drive control efforts: a meta-analysis of the literaturefrom 1960 to 1991. Accident Analysis and Prevention, 27(3), 307−316.

Zador, P. L., Krawchuk, S. A., & Voas, R. B. (2000). Alcohol-related relativerisk of driver fatalities and driver involvement in fatal crashes in relationto driver age and gender: An update using 1996 data. Journal of Studieson Alcohol, 61(3), 387−395.

Alexander C. Wagenaar, is Professor in the Department of Epidemiologyand Health Policy Research and in the Institute for Child Health Policy, atthe College of Medicine, University of Florida. He has been at theUniversity of Florida since September, 2004. From 1979–1989 Dr.Wagenaar was at the University of Michigan, with appointments in theTransportation Research Institute and the School of Public Health, and from1990–2004 was Professor and Director of the Alcohol EpidemiologyProgram at the University of Minnesota. He is an epidemiologist focusingon evaluation of public policy changes and community-level interventions,using both randomized trial and time-series research designs and statisticalmethods. He is currently principal investigator on four major researchprojects, serves as a scientific reviewer for two dozen journals, is a memberof the Editorial Boards of Prevention Science and the Journal of SafetyResearch, and is an Assistant Editor of the journal Addiction. Dr. Wagenaarhas published over 150 papers on social epidemiology, health policy,community intervention trials, alcohol studies, violence prevention, trafficsafety, and injury control. He holds the designation as a “Highly CitedResearcher” by the Institute for Scientific Information, an honor limited tothe top one-half of one percent of most cited scientists worldwide.

Mildred M. Maldonado-Molina, is Assistant Professor in the College ofMedicine, Department of Epidemiology and Health Policy Research at theUniversity of Florida. Dr. Maldonado-Molina is a methodologist withinterests in longitudinal methods and the prevention of substance useamong adolescents. Her interests include latent class and latent transitionanalysis, latent growth modeling, and multi-level models. Her substantiveresearch interests include the study of patterns of substance use onsetamong adolescents, prevention of alcohol use among youth, and evaluationof alcohol control policies. She has a Bachelor's degree in Psychology fromthe University of Puerto Rico and Master's and Doctoral degrees in HumanDevelopment and Family Studies from Pennsylvania State University.

Linan Ma, is the Statistical Coordinator for the Social & CommunityEpidemiology research group in the College of Medicine, Department ofEpidemiology and Health Policy Research at the University of Florida.His expertise is in multivariate statistical analyses, particularly of longi-tudinal data, and management of large, complex data sets. He received aBachelor's degree in Medical Science from the Capital University ofMedical Sciences in Beijing and a Master's degree in Public Health, witha major in Biostatistics, from the University of South Florida.

Amy L. Tobler, is the Research Coordinator for the Social & CommunityEpidemiology research group in the College of Medicine Department ofEpidemiology and Health Policy Research and Coordinator for the Institutefor Child Health Policy at the University of Florida. She has been with thegroup since January, 2005. Ms. Tobler has previously been employed as aDepartment Chair and Instructor at City College (2004–2005) and as aResearch Assistant in the College of Health and Human PerformanceDepartment of Health Education and Behavior (2003–2004). Ms. Toblerhas a Bachelor's degree in Health Science and a Master's degree in PublicHealth from the University of Florida. She is also a doctoral student in theCollege of Health and Human Performance Department of HealthEducation and Behavior.

Kelli A. Komro, is Associate Professor in the College of MedicineDepartment of Epidemiology and Health Policy Research and AssociateDirector of the Institute for Child Health Policy at the University of Florida.Dr. Komro has been at the University of Florida since September, 2004. Shewas previously a faculty member at the University of Minnesota in theDivision of Epidemiology of the School of Public Health (1995–2004). Sheis an epidemiologist specializing in the social determinants of health amongchildren and adolescents. Her research focuses on designing and evaluatingcommunity-wide strategies to promote health among children andadolescents. Dr. Komro has developed a successful research program inyouth health promotion with an emphasis on randomized controlledcommunity trials to prevent tobacco, alcohol, drugs, violence, and HIVamong adolescents. She currently serves as the Principal Investigator of agrant entitled, “An Adaptation of Project Northland for Urban Youth” (alsoknown as Project Northland Chicago), which is the largest alcohol useprevention trial ever funded by the National Institute on Alcohol Abuse andAlcoholism. She received her Bachelor's degree in Psychology from theUniversity of Wisconsin, a Master's degree in Psychology from Connecti-cut College, and Master's and Doctoral degrees from the University ofMinnesota in epidemiology.