the population consumption model, alcohol control practices, and alcohol-related traffic fatalities
TRANSCRIPT
Preventive Medicine 34, 187–197 (2001)doi:10.1006/pmed.2001.0970, available online at http://www.idealibrary.com on
The Population Consumption Model, Alcohol Control Practices,and Alcohol-Related Traffic Fatalities1
Deborah A. Cohen, M.D., M.P.H.,2 Karen Mason, M.Sc., and Richard Scribner, M.D., M.P.H.
b
J
Louisiana State University Health Sciences Center, Department of Pu
Published online
Background. More than 40% of urban traffic fatalitiesare alcohol related and the rate of such fatalities variesmore than 10-fold across U.S. cities. These variationsmight be explained by differences in local alcohol con-trol policies and practices.
Methods. We conducted a cross-sectional survey ofstate Alcohol Beverage Control agencies and local citypolice departments in 107 cities that participate in theNational Highway and Traffic Safety Administration’sFatality Analysis Reporting System. We examined theassociation of alcohol control practices in 1997 andalcohol-related traffic fatalities per daily vehicle milestraveled, 1995–1997.
Results. Ninety-seven (91%) cities participated. Regu-lations related to alcohol accessibility, licensure ofalcohol outlets, disciplinary procedures of alcohol out-lets, and enforcement of blood alcohol concentrationlaws were associated with lower rates of fatalities.Cities with 9 or fewer of the 20 regulations had 1.46-fold greater alcohol-related traffic fatality rates thancities with 15 or more of these regulations, represent-ing 392 excess deaths annually. Beer consumption wasfound to be a potential mediator of the effect of regula-tions on traffic fatalities.
Conclusions. Alcohol beverage regulations are asso-ciated with alcohol-related traffic fatalities. Localitiesshould consider greater restrictions on alcohol accessi-bility, stricter disciplinary measures for violations, andstricter licensure requirements as a potential means to
reduce alcohol-related traffic fatalities. q 2002 AmericanHealth Foundation and Elsevier Science (USA)
Key Words: alcohol policy; traffic fatalities; alcoholaccessibility.
1 This study was supported by Grant 031603 from the Robert WoodJohnson Substance Abuse Policy Research Program.
2 To whom correspondence and reprint requests should be ad-dressed at the RAND Corporation, 1700 Main Street, Los Angeles,CA 90405. E-mail: [email protected].
187
lic Health and Preventive Medicine, New Orleans, Louisiana 70112
anuary 9, 2002
INTRODUCTION
Alcohol-related problems are typically viewed as hav-ing a genetic basis in which environmental factors con-tribute to the expression of the underlying disease [1].This represents a medical disease model. People whodrink alcohol excessively and suffer negative conse-quences of alcohol use are believed to be suffering froma disease related to heritable genetic characteristicstriggered by environmental or psychological factors [2].Most physicians manage alcoholism by prescribingcounseling or other intensive, individualized treatmentregimens. Accordingly, if this is the correct manage-ment model, the incidence of alcohol-related problemsshould decrease if alcoholics are matched to the treat-ment regimen appropriate for their level of disease.
An epidemiological perspective on alcohol use, the“population consumption model” [3], also called the sin-gle distribution theory [4–6], has gained a considerabledegree of acceptance in explaining alcohol use in popu-lations [7,8]. This model states that alcohol consump-tion by any individual is, in part, a function of theoverall distribution of consumption of the communityand leads to the conclusion that the magnitude ofalcohol-related health problems in a population is di-rectly related to per-capita consumption [3,9–11]. Indi-vidual consumption in turn is associated with variousfactors affecting the physical and social availability ofthe product within the community in which individualsreside [12].
Despite the evidence that community-level consump-tion of alcohol drives alcohol-related morbidity and mor-tality, control of alcohol-related problems through com-munity interventions has not advanced in the sameway control of tobacco has [13,14]. While price controls
in the form of taxes are likely to be the most potentcontrol of alcohol consumption [15–17], taxes on alcoholare less than 1% of all taxes, significantly lower thanat the turn of the century, when alcohol taxes comprised0091-7435/02 $35.00q 2002 American Health Foundation and Elsevier Science (USA)
All rights reserved.
188 COHEN, MASON
TABLE 1
Alcohol-Related Traffic Fatality Rates and Beer Tax by City(Sorted by Traffic Fatality Rate, Column 1)
Average Averagealcohol- alcohol- Beer taxrelated related in centstraffic traffic per drink
fatality rates fatality rates (12-ozCity per 100,000a per DVMTb glass)
Lincoln, NE 0.86 0.046 2.15Syracuse, NY 1.31 0.053 1.50New York, NY 1.67 0.111 1.50Madison, WI 1.74 0.083 0.60Fremont, CA 1.87 0.084 1.87Honolulu CDP, HI 2.01 0.124 5.06San Francisco, CA 2.07 0.099 1.87Boston, MA 2.16 0.109 1.00Jersey City, NJ 2.29 0.150 1.12Cincinnati, OH 2.36 0.089 1.69San Jose, CA 2.47 0.111 1.87Indianapolis, IN 2.58 0.087 1.08Spokane, WA 2.63 0.135 2.44Wichita, KS 2.64 0.132 1.69Virginia Beach, VA 2.71 0.127 3.35Aurora, CO 2.72 0.128 0.75Minneapolis, MN 2.80 0.121 1.39Columbus, OH 2.91 0.110 1.69Seattle, WA 2.91 0.117 2.44Raleigh, NC 2.91 0.091 4.98Baltimore, MD 2.92 0.146 0.84Akron, OH 3.01 0.126 1.69Bakersfield, CA 3.05 0.176 1.87Grand Rapids, MI 3.09 0.136 1.91St. Paul, MN 3.15 0.136 1.39Buffalo, NY 3.18 0.173 1.50Milwaukee, WI 3.18 0.130 0.60Stockton, CA 3.20 0.177 1.87Toledo, OH 3.29 0.140 1.69San Diego, CA 3.33 0.155 1.87Richmond, VA 3.41 0.130 3.35St. Petersburg, FL 3.42 0.161 4.50Rochester, NY 3.43 0.146 1.50Miami, FL 3.44 0.189 4.50Fort Wayne, IN 3.45 0.170 1.08Colorado Springs, CO 3.49 0.196 0.75Pittsburgh, PA 3.54 0.177 0.99Long Beach, CA 3.55 0.163 1.87Modesto, CA 3.55 0.202 1.87Santa Ana, CA 3.56 0.164 1.87Omaha, NE 3.57 0.173 2.15Los Angeles, CA 3.58 0.164 1.87Philadelphia, PA 3.59 0.220 0.99Des Moines, IA 3.68 0.152 1.78Lexington–Fayette, KY 3.68 0.130 4.69Tulsa, OK 3.91 0.176 3.78Anaheim, CA 3.93 0.181 1.87Mesa, AZ 3.96 0.187 1.50Chicago, IL 4.00 0.208 1.73Anchorage, AK 4.01 0.253 3.28Newport News, VA 4.03 0.190 3.35
Hialeah, FL 4.05 0.223 4.50Louisville, KY 4.21 0.145 4.69Tacoma, WA 4.32 0.198 2.44Sacramento, CA 4.45 0.208 1.87Providence, RI 4.53 0.231 0.91, AND SCRIBNER
TABLE 1—Continued
Average Averagealcohol- alcohol- Beer taxrelated related in centstraffic traffic per drink
fatality rates fatality rates (12-ozCity per 100,000a per DVMTb glass)
Greensboro, NC 4.55 0.131 4.98Oakland, CA 4.55 0.217 1.87Norfolk, VA 4.58 0.215 3.35Cleveland, OH 4.59 0.211 1.69Charlotte, NC 4.62 0.175 4.98Portland, OR 4.79 0.226 0.79Orlando, FL 4.84 0.184 4.50Winston–Salem, NC 4.98 0.163 4.98Dayton, OH 4.99 0.183 1.69Arlington, TX 5.06 0.154 1.81Fresno, CA 5.07 0.281 1.87Riverside, CA 5.08 0.230 1.87Huntsville, AL 5.23 0.187 9.87Jacksonville, FL 5.31 0.198 4.50Salt Lake City, UT 5.38 0.249 3.32Little Rock, AK 5.45 0.212 2.27Corpus Christi, TX 5.66 0.242 1.81Birmingham, AL 5.67 0.170 9.87Las Vegas, NV 5.68 0.309 0.84Washington, DC 5.91 0.254 0.84Mobile, AL 5.92 0.227 9.87El Paso, TX 5.98 0.329 1.81Amarillo, TX 6.17 0.242 1.81Tucson, AZ 6.19 0.337 1.50Lubbock, TX 6.30 0.266 1.81Knoxville, TN 6.30 0.201 1.18Denver, CO 6.32 0.296 0.75Montgomery, AL 6.33 0.250 9.87Jackson, MS 6.44 0.256 4.00Baton Rouge, LA 6.52 0.282 3.02Shreveport, LA 6.58 0.276 3.02Columbus, GA 6.62 0.318 4.50Oklahoma City, OK 6.94 0.283 3.78Houston, TX 7.03 0.235 1.81Austin, TX 7.11 0.244 1.81New Orleans, LA 7.15 0.523 3.02Atlanta, GA 7.23 0.200 4.50Fort Worth, TX 7.48 0.228 1.81Memphis, TN 7.54 0.335 1.18San Antonio, TX 7.54 0.291 1.81Tampa, FL 7.55 0.356 4.50Newark, NJ 7.55 0.501 1.12Chattanooga, TN 7.92 0.263 1.18St. Louis, MO 7.92 0.279 0.56Phoenix, AZ 8.14 0.387 1.50Detroit, MI 8.22 0.359 1.91Nashville–Davidson, TN 8.47 0.251 1.18Albuquerque, NM 8.62 0.330 3.84Kansas City, MO 10.10 0.358 0.56Dallas, TX 10.23 0.312 1.81Mean 4.75 0.203 2.64Range 0.86–10.23 0.046–0.523 0.56–9.87
a Averaged for 1995, 1996, and 1997.
b Daily vehicle miles traveled.Note. Pearson correlation (P value): fatalities per 100,000 popula-tion and fatalities per DVMT, 0.837 (0.000); fatalities per 100,000population and beer tax, 0.103 (0.294); fatalities per DVMT and beertax, 20.009 (0.930).
T
beverage laws. Licensure refers to how rigorous the procedure is forgiving an outlet a license to sell alcohol. Enforcement is a measure
ALCOHOL-RELATED
40% of all taxes [18]. Alcoholic beverages sometimescost less per ounce than bottled water, orange juice, ormilk, yet most state and federal legislators have beenunwilling to increase taxes on alcohol.
Alcohol-related traffic fatalities accounted for 50,562deaths nationally during the 3 years from 1995 through1997 [19]. In general, alcohol-related traffic fatalitieshave been declining over the past 2 decades, which canbe partly explained by declines in alcohol consumption,enactment of the minimum drinking age on a nationallevel, and widespread adoption of deterrence strategies[20]. Specific policies that appear to reduce alcohol-related traffic fatalities include the adoption of lawsgoverning the minimum allowable blood alcohol concen-tration (BAC) [21], dram shop liability [22,23], servertraining programs [24], alcohol excise taxes [17,18], lim-iting the density of alcohol outlets [25–27], and a varietyof individually targeted deterrence practices such aslicense suspension and sobriety checkpoints [28,29].However, in a national analysis, Sloan et al. found nobenefit of mandatory penalties for driving while intoxi-cated [22]. In many cases evaluations of alcohol controlpractices have been limited to only one policy [21,23,17,18,25], have examined interventions that were im-plemented over a short period of time [21,28], or wererestricted to specific geographic areas [21,23,24,27,28].Only a handful of national studies have examined alimited number of alcohol control practices and theirpossible impact on alcohol-related traffic fatalities, andthese were at the state level [4,14–17].
The U.S. Constitution left the states the power toregulate matters related to health, welfare, and safety.Thus, significant state and local infrastructures havebeen developed over the past 225 years to enforce alco-hol control laws and regulate alcohol outlets. Becausecities usually have the ability to enact even more strin-gent measures than the state, there are significant dif-ferences at the local level in how alcohol is regulatedacross the United States. Not only do laws, licensingpolicies, and taxes vary, but so does enforcement of lawsthat might otherwise be similar among cities. Indeed,there is a 12-fold variation in the rate of alcohol-relatedfatalities per capita among the largest U.S. cities [19].Considering that alcohol-related fatality rates differacross cities with similar alcohol taxes (Table 1), otherregulatory practices are likely to protect againstalcohol-related crashes, while at the same time remainpolitically acceptable to the American public. Such reg-ulations would aim to control the time and place ofalcohol consumption (Fig. 1).
In order to identify protective alcohol beverage con-
trol strategies that appear to be effective across theUnited States, we conducted a national study of theagencies that are charged with the enforcement of alco-hol beverage laws. While other studies have examineddifferences in state level practices, our approach isof how active a locality is in monitoring compliance with alcoholbeverage laws. Price is the cost of alcohol, including taxes. Promotionrefers to the effort and use of educational tools and media to promoteor constrain the consumption of alcohol.
RAFFIC FATALITIES 189
FIG. 1. Theoretical model of how laws and policies influence alco-hol-related traffic fatalities [10]. Accessibility refers to how easy itis to obtain and use alcohol in public settings. Discipline refers tothe severity of penalties given to outlets that violate alcohol control
unique in that our focus was on regulations and agencyenforcement practices at the city level. We examinedthe association between regulatory policies and the inci-dence of alcohol-related traffic fatalities.
METHODS
In 1998 questionnaires with items that covered sixregulatory domains were sent to all city offices of thestate Alcohol Beverage Control (ABC) agencies and tothe local city police departments in all 107 citieswith populations over 150,000 that participate in theNational Highway and Traffic Safety Administration’sFatality Analysis Reporting System (FARS) [19]. Thequestions referred to actual practices in these domainsin the calendar year 1997.
The 107 cities represented 38 states. Some states hadonly one city represented, but others had as many as15 (California) or 10 (Texas). The next highest were 5in Ohio and 6 in Florida. Questions asked of the stateABC agencies were nearly identical to those asked ofthe local police; however we asked them to provide addi-tional information on licensing of alcohol beverage out-lets (ABOs), penalties to owners of ABOs, density limits,restrictions on alcohol sales, and server training.
The seven domains we studied include (1) laws gov-erning accessibility of alcohol [34–36]; (2) licensurerequirements for alcohol outlets; (3) disciplinary proce-
dures for outlets that violate existing laws; (4) enforce-ment practices (the type, frequency, and quality of en-forcement of existing laws and whether these practiceshad changed in the previous year) [37]; (5) policies relat-ing to driving under the influence (DUI); (6) resources
,
190 COHEN, MASONavailable to enforcement officers, including rewards,benefits, salaries, training, and position on the enforce-ment hierarchy; and (7) public relations/education un-dertaken by the regulatory agencies targeted either to-ward the public or toward the outlets.
Indices were created for each of five alcohol-controldomains to reduce the number of variables while im-proving the predictive power of the measure. There wasso little variability in the latter two domains that noreliable index could be created. The selection of vari-ables for each group was based on a priori hypothesesof their importance toward the control of alcohol. Allvariables were equally weighted by normalizing vari-ables with three or more categories and by log-transfor-mation of skewed variables. In addition, variables werecoded to reflect the same direction of control, that is,from looser to stricter policies. A reliability analysiswas performed for each index (Tables 2 through 6). Afteranalyzing the indices separately, we grouped regula-tions from the first four indices together and found thatremoval of items in the enforcement index significantlyincreased the reliability of a combined scale (Cronbach’sa 5 0.85) [38]. We dichotomized each of the measuresto reflect any policy implementation versus none (forregulatory practices with ordinal responses) or to reflectabove and below the median (for continuous responses).We then examined the association of the combined scaleof 20 regulatory practices and our outcome measure.
Combining ABC and Local Data
Information collected from both the ABC agenciesand the local police departments was combined beforeindices were created. The accessibility variables werereadily combined as they tended to measure well-established laws—such as open containers in vehicles,public drinking laws, and drive-through liquor stores.For this reason the agreement between ABC and policewas quite good. On the other hand, it was difficult toget complete data on DUI enforcement, as this wasoften regulated by differing agencies. In some casesstate police or Departments of Public Safety were re-sponsible for DUI enforcement, rather than local policeor ABC agencies. Combining enforcement variablesposed several problems because practices tended to dif-fer between ABC and police; in some cases their dutiesoverlapped. We addressed this by taking the answerthat showed the greatest level of enforcement, thusgiving the best possible scenario for enforcement.Where only the ABC or the police completed the survey,responses may not accurately reflect enforcement prac-
tices for that city. We also inquired as to how enforce-ment practices had changed since the previous year.We examined the relationship of the direction of changein enforcement with the change in alcohol-related traf-fic fatalities both together and separately for responsesAND SCRIBNER
by police and ABC agencies. Discipline and licensureinformation was provided only by the ABC agencies.Public awareness activities, routine procedures, andnon-alcohol-related duties also differed between ABCagencies and police. The same procedure used to com-bine enforcement variables was applied here. Resourcevariables were kept separate since most informationwas on the qualifications of agents.
Missing Data
Like many studies in the social sciences dealing withquestionnaire data, missing data are a common prob-lem, which must be handled in a statistically robustmanner. In our study, missing data are a particularproblem because of the sampling frame. The analysismethod used full information maximum likelihood toestimate the regression parameters with all data overpartially complete data. These analyses were conductedusing the AMOS (analysis of moment structures) tech-nique [39]. Although we present results on only theobserved data, results were subsequently confirmed,and in most cases the statistical significance of associa-tions was strengthened, on a more complete data filewith analysis of complete and partially complete datausing the AMOS procedure.
Outcome Measures
We chose the FARS database as a relatively unbiasedmeasure of alcohol-related mortality. The FARS data-base is collected by the National Highway and TrafficSafety Administration and documents all traffic fatalit-ies annually, including alcohol-positive pedestrians andcyclists killed by a sober driver. Fatalities per 100,000population for which the police accident report explic-itly states or implies that alcohol was involved wereadjusted for the proportion of fatalities with alcoholinvolvement. The proportion of urban crashes that werealcohol-involved from 1995 through 1997 nationallywas greater than 44%. Because the risk of a trafficcrash is highly associated with the number of milesdriven [40,41], we divided each city’s annual alcohol-related traffic fatalities per 100,000 by its per capitadaily vehicle miles traveled for the corresponding year,making our outcome measure alcohol-related trafficfatalities per daily vehicle miles traveled (DVMT) [42].We then averaged the results from 1995 through 1997.Table 1 shows how alcohol-related traffic fatalities vary
across cities when adjusted for population, comparedwith how they vary when adjusted for DVMT. The corre-lation between the two measures is high (r 5 0.84) andmost cities have similar relative positions in bothrankings.
T
ALCOHOL-RELATEDAdditional Variables
Other factors known to affect traffic crashes werebased upon literature review [21–34,41,43] and col-lected from various sources. It is important to note thatsome of these were city level and others state level.The variables of interest were minimum age for a statedriver’s license in 1997 [44,45], vehicle inspection prac-tices [46], percentage total city earnings from hotelsand lodges in 1997 (tourism) [47], and city demographicinformation, including age (% population over 65 years),gender distribution (% males ages 15–24), race, educa-tion (% population over 25 with a high school diplomaand % population over 25 with a college degree), andpercentage of the population in poverty [48]. All censusdata were obtained from the 1990 U.S. Census, withpopulation adjustments to 1995 [48,49].
Alcohol consumption data were available only at thestate level [50]. We tested the relationship of total alco-hol consumption including wine and beer consumptionas a full mediated model, recognizing that the measureof alcohol consumption per capita was calculated basedon statewide consumption, rather than the city.
Analysis
We analyzed data using SPSS [51]. Potential con-founding factors associated with alcohol-related fatal-ities were controlled for in a multivariate analysis usingbackward elimination, a method in the ordinary least-squares regression procedure [52]. Separate models us-ing alcohol-related traffic fatalities as the dependentvariable were examined for all four indices due to collin-earity and sampling problems (see Appendix for correla-tions). The other independent variables were enteredwith each index. We also ran the models with and with-out a state-level variable as to whether BAC testing ismandatory in traffic fatalities. The presence of manda-tory BAC testing did not alter our findings, suggesting alack of assessment bias. A summary index of protectivealcohol-related policies was developed after checkingfor reliability using Cronbach’s a [38]. For each of theindices that were associated with lower rates of alcohol-related traffic fatalities, we dichotomized the regulationas either present or absent or at the midpoint of its
value (e.g., number of ways to inform public about a new license application, #3 vs $4). We then calculatedthe number of excess deaths by comparing the numberand rate of deaths in cities with few regulations with thedeaths in cities with most of the 20 regulatory practices.RESULTS
Of 107 cities surveyed responses were available for97 (91%). This included responses from 61 local andstate agencies serving the same locality, only ABC(n 5 25), and only local (n 5 11). By city, the number
RAFFIC FATALITIES 191
of alcohol-related traffic fatalities varied from 0.8/100,000 to 10/100,000.
Tables 2–6 describe the characteristics of the accessi-bility, licensure, discipline, enforcement, and DUI indi-ces and their reliabilities. No composite index was iden-tified for the resources available to enforcement agentsand for public relations/education. However, the itemsthat we investigated were relatively similar throughoutthe United States. The average starting salary for anenforcement officer was $27,378, and in 78/97 (80%)cities new agents earn #$30,000 per year. New Jerseywas an outlier, paying $65,000 because officers have toattain the level of detective before working in the alco-hol unit. Salary alone was related to traffic fatalities,but it appeared to be only a proxy for socioeconomicstatus as the relationship disappeared once it was ad-justed for per-capita income. There were no significantdifferences between ABC agents and police officers inthe length of stay in the job, even though 47% of stateABC agencies claimed that benefits for state agentswere worse than for the local police. Public relationsefforts were minimal throughout the country and nooffice had a dedicated annual budget for this. Sixty-three percent of states had a dedicated public informa-tion officer, but funds for public information and materi-als were available only on a limited basis. Publicinformation campaigns were usually paid for by otherdepartments or programs or were not under the alcoholenforcement agency infrastructure. No items in thesetwo domains were correlated with alcohol-related traf-fic fatalities.
The accessibility index includes measures of how dif-ficult it is to obtain alcohol. Of note is that among the89 cities for which data are available, the number ofalcohol outlets per 100,000 population varied between77 and 1,533. Eighty-five percent of cities prohibitedpublic drinking, 70% prohibited anyone from drinkingin a car, and 62% prohibited drive-through alcohol out-lets. Only 31% required keg registration. Twenty-fivepercent had no food service laws associated with on-premise alcohol sales, 64% had no restrictions aboutalcohol consumption at sports events, and 73% allowedyouth under 21 to enter bars. Cities with more restric-tions on the accessibility of alcohol had significantlylower traffic fatalities (r 5 20.331; P , 0.002) (seeTable 2).
All states require businesses that sell alcohol to havea state license. The licensure index contains items thatare requirements of, or restrictions on, alcohol licenses.The average initial cost of an alcohol license in a restau-rant serving alcohol was $3,702 and varied from $30
to $20,000. Few cities had limits on beer and wine outletalcohol licenses; 49% restricted the number of outletsselling hard liquor. More than half the cities did atleast three different things to inform the public priorto approving alcohol licenses, while 10% of cities had
2 5 all types) 97 0.95 (0.68) 20.207 (0.042)
Number alcohol restrictions in sports events (0 to 3)Drinking prohibited in cars (0 5 no laws, 1 5 passengers can drink,2 5 no one can drink)Number of outlets per 100,000 population (reverse coded)
no mechanism to inform the public at all. The stricterthe license requirement, the lower the alcohol-relatedtraffic fatalities (r 5 0.293, P , 0.012) (see Table 3).
The discipline index focused on serious penalties thatoutlets could receive if convicted of a violation, includ-ing fines and suspension of their license to sell alcohol.Forty-five percent of cities said they never waivedsecond penalties, 50% rarely waived penalties, and 5%of cities reported that they sometimes did. The moredefinite and serious the penalty, the lower the trafficfatalities in the city (r 5 20.379, P , 0.000) (seeTable 4).
When examining the scope of enforcement it wasclear that many jurisdictions do not monitor compliancewith alcohol beverage laws on a routine basis. Twelvecities had no dedicated alcohol enforcement agents.Among those that did, the average number of outletsper enforcement agent was 296 and varied from 1.5to 1,638. Nationally, of the 114,837 outlets for whichenforcement information is available from this survey,20% are never monitored with underage compliancechecks. An additional 32% were from cities where less
than 5% of all outlets were visited in 1997. The enforce-Initial cost of a state license for a restaurantLimited number of licenses (0 5 no, 1 5 yes)Are both state and local authorities notified in licensure? (0 5 no, 1 5Number of things done to inform public before approving license (0 toNumber of things done to inspect location before licensing (0 to 3)
97 0.43 (0.64) 20.172 (0.091)
97 1.61 (0.65) 20.160 (0.117)89 327 (246) 20.016 (0.878)
When we analyzed the item asking how enforcementpractices had changed from the year prior to 1997, 43/76 (56.6%) state ABC agencies had increased their en-forcement efforts. Between 1995 and 1996, 27/76 (36%)cities had experienced an increase in alcohol-relatedtraffic fatalities of more than a 0.5 standard deviation.Of these 27 cities, 20 (74%) subsequently increasedtheir enforcement efforts in 1997 compared to 7/27(26%) that did not increase enforcement efforts [RR3.23 (95% CI 1.04, 10.31; P , 0.02)].
The DUI index captured the state legal limit for BAC,frequency of sobriety checkpoints, and whether thesewere random or nonrandom. The index was negativelycorrelated with alcohol-related fatality rates. Amongstates with a BAC of ,0.08, 68% set up random check-points and only 3% had no checkpoints. In contrast,49% of cities with BAC at 0.10 did not set up check-points, and only 21% did random checks.
Alcohol taxes, including beer taxes, were not associ-ated with alcohol fatality rates in our sample. In amultivariate model using backward regression control-ling for the minimum age for a driver’s license, the
192 COHEN, MASON, AND SCRIBNER
TABLE 2
Description of Accessibility Index and Its Component Variables
Bivariate correlationwith traffic fatality
Variable N Mean (SD) rates r (P value)
Accessibility index (Cronbach’s a 5 0.6113) 89 2.19 (3.23) 20.331 (0.002)Drive through liquor stores prohibited (0 5 no, 1 5 yes) 97 0.62 (0.49) 20.303 (0.003)Keg registration required (0 5 no, 1 5 yes) 97 0.31 (0.46) 2.288 (0.004)Drinking prohibited in public (0 5 no, 1 5 yes) 97 0.85 (0.36) 20.276 (0.006)Youth ,21 prohibited in bars (0 5 no, 1 5 yes) 97 .27 (0.45) 20.218 (0.032)Food service laws for outlets (0 5 no laws, 1 5 restaurant,
percentage of households in poverty, vehicle inspection
practices, percentage of population over 65, percentagement index was positively correlated with alcohol-related fatalities (r 5 0.284, P , 0.008), indicating a of population male between the ages of 15 and 24,percentage of population over 25 with a high schoolgreater number of traffic fatalities in cities whose en-
forcement practices are stricter (P 5 0.021, Table 5). diploma or a college degree, and level of tourism, alcohol
TABLE 3
Description of Licensure Index and Its Component Variables
Bivariate correlation with trafficVariable N Mean (SD) fatality rates r (P value)
Licensure index (Cronbach’s a 5 0.7407) 73 0.98 (2.52) 20.293 (0.012)
83 $3,702 ($4,828) 20.215 (0.051)76 0.49 (0.50) 20.176 (0.129)yes) 84 0.48 (0.50) 20.156 (0.156)6) 84 2.89 (1.83) 20.127 (0.249)
84 2.01 (0.94) 20.106 (0.337)
3 5 not negotiable)Penalty to outlet for 2nd violation (1 5 warning, 2 5 fine only, 84 3.06 (0.95) 20.197 (0.072)
3 5 fine and/or suspension, 4 5 suspension only, 5 5revocation)
Proportion 2nd penalty reduced (0 5 none, 1 5 rarely, 2 5 some)
accessibility remained a predictor of traffic-related fa-talities (P 5 0.046; see Table 6). In separate models,the licensure and discipline indices were likewise asso-ciated with traffic fatalities (P 5 0.059 and P 5 0.058,respectively). Total alcohol consumption was not signifi-cantly associated with alcohol-related traffic fatalitiesin any of the multivariate models (not shown). However,when we added beer consumption separately as a medi-ating variable, the effect of the regulatory indices wasreduced, indicating that the beer consumption variablemay be mediating the association between regulatoryactivities and alcohol-related traffic fatalities (Tables 7and 8). Wine and spirits consumption, in contrast, didnot have any mediating affect on alcohol-related trafficfatalities, separately or together.
The accessibility, licensure, and discipline indiceswere correlated with each other (r 5 0.46–0.56), whileenforcement did not correlate with any of the other
three (Appendix). The DUI index was correlated withat least weekly)How often are drink specials enforced? (1 5 few per year to never, 2 5
week to once/month, 3 5 at least few times/week)Outlets per agent (reverse for index—0 given maximum)Percentage outlets that had underage compliance checks
83 0.78 (0.59) 20.170 (0.125)
traffic fatality rates than cities with 15 or more of the20 practices (P , 0.002). This represents an estimated392 excess deaths annually. Similarly, cities with analcohol fatality rate below 21.00 standard deviationfrom the normalized mean had an average of 14.2 regu-lations compared to an average of 8.9 regulations for
ALCOHOL-RELATED TRAFFIC FATALITIES 193
TABLE 4
Description of Discipline Index and Its Component Variables
Bivariate correlation with trafficVariable N Mean (SD) fatality rates r (P value)
Discipline index (a low, variables selected a priori) 82 0.38 (2.10) 20.379 (0.000)Are violations published? (0 5 no, 1 5 yes) 84 0.38 (0.49) 20.346 (0.001)Second penalty negotiable? (1 5 yes, always, 2 5 yes, sometimes, 83 2.36 (0.73) 20.234 (0.033)
cities more than 1.00 standard deviation above the nor-malized mean (P , 0.019). Figure 2 depicts a linearrelationship between the number of fatalities and thenumber of regulations constraining the context of alco-hol consumption (r 5 20.456).
DISCUSSION
The results indicate that alcohol regulatory practicesexplain many differences in alcohol-related traffic fatal-ities across U.S. cities. The items most strongly associ-ated with lower rates of alcohol-related traffic fatalities
are those that include random sobriety checkpoints and those that influence access to alcohol, such as laws re-the accessibility index but not with any of the otherindices. Cities with 9 or fewer of the 20 regulatory prac- stricting the places where it is acceptable to drink andlaws limiting youth access to alcohol. Where outletstices in a combined index of accessibility, licensure, and
disciplinary items had 1.46 times the alcohol-related are given a certain penalty such as a fine or license
TABLE 5
Description of Enforcement Index and Its Component Variables
Bivariate correlation with trafficVariable N Mean (SD) fatality rates r (P value)
Enforcement index (Cronbach’s a 5 0.6083) 87 0.32 (4.11) 0.284 (0.008)How often are raids done? (1 5 few per year to never, 2 5 once/week to once/ 97 1.81 (0.83) 0.262 (0.010)
month, 3 5 at least few times/week)Number of things done to enforce food service laws (0 to 3) 97 1.51 (0.98) 0.181 (0.075)How often are covert operations done? (1 5 few per year to never, 2 5 once/ 96 2.35 (0.81) 0.151 (0.141)
week to once/month, 3 5 at least few times/week)Percentage outlets did Cops-in-Shops 96 1.73 (4.20) 0.128 (0.212)How often is compliance with food service laws verified? (1 5 never or don’t 97 2.97 (1.35) 0.119 (0.247)
do, 2 5 annually, 3 5 few times/year, 4 5 once to few times/month, 5 5
once/ 97 2.04 (0.79) 0.113 (0.269)
92 257 (347) 20.079 (0.454)88 8.90 (12.21) 0.048 (0.660)
How often are sobriety checkpoints done? (0 5 never, 81 0.91 (0.71) 20.222 (0.046)
1 5 few times/once per year, 2 5 at least once per month)Are sobriety checks done in random places? (0 5 none,1 5 not random, 2 5 random)
Legal BAC limit (0 5 0.10, 1 5 0.08 or less)
suspension (as opposed to only warnings or waivers)for alcohol-related violations, traffic fatalities werealso lower.
The finding that beer consumption may mediate alco-hol-related traffic fatalities supports the population-consumption model. Beer is the most common type ofalcoholic beverage sold [50], and it is widely availablecold, in single servings, at convenience stores and gasstations. It is heavily promoted with advertisementsthat often show people drinking beer in outdoor set-tings, such as beaches, parks, fairs—all places onewould need to drive to in a car. Wine, in contrast, isoften portrayed as being consumed with meals, at home,and on special occasions. Spirits are not as heavily ad-vertised and in many cities their access is under statecontrol. Although the alcohol consumption measure isflawed, as it was measured at the state level ratherthan the city, one might expect that a city level measurewould be more precise and would possibly yield an evenstronger mediating effect than the current data indi-cate [53].
The positive association of enforcement with traffic
Percentage all private earnings from hotels 0.160 (0.080) 0and lodges (tourism)
(Constant) 0.458N 87R2 0.378
81 1.09 (0.83) 20.196 (0.083)
92 0.45 (0.50) 20.125 (0.235)
regularly applied over time. In this study most cities’agencies reported that their enforcement practices wereinitiated primarily in response to complaints, ratherthan implemented on a regular basis. Interviews withenforcement officers revealed that enforcement opera-tions like raids and underage compliance checks varywidely from year to year and may be implemented inresponse to political concerns or when extra grant fundsare available [54]. In contrast, the accessibility andlicensure indices reflect policies that are more stableover time. Disciplinary practices, however, appear toreflect the political leadership, as the alcohol commis-sioners are frequently political appointees who maychange with every state election. The existence of lawsfor BAC ,0.08 was associated with random sobrietycheckpoints and is likely to indicate an overall greaterlocal commitment to this aspect of alcohol control.
We think the reason we did not find associations be-tween resources available to alcohol enforcement offi-cers and alcohol-related traffic fatalities is that there
194 COHEN, MASON, AND SCRIBNER
TABLE 6
Description of DUI Index and Its Component Variables
Bivariate correlation with trafficVariable N Mean (SD) fatality rates r (P value)
DUI index (Cronbach’s a 5 0.7960) 81 1.48 (1.08) 20.191 (0.087)
is little variation in this area across the U.S. cities.
There are few incentives for aggressive job performancefatalities was unexpected. Given the increase in en-in alcohol enforcement. In contrast to illegal drug en-forcement in the majority of cities after they experi-forcement, in which a proportion of seizures (includingenced an increase in alcohol-related traffic fatalities, itappears that enforcement is reactive, rather than being cash and vehicles) becomes the property of the police
TABLE 7
Regression Coefficients for Indices and Control Variables—Alcohol-Related Traffic Fatalities per DVMT Are Dependent Variable [b(P Values)]
Variable Model 1 Model 2 Model 3 Model 4 Model 5
Accessibility index 20.188 (0.046)Licensure index 20.213 (0.059)Discipline 20.196 (0.058)Enforcement index 0.210 (0.021)DUI index 20.279 (0.004)Minimum age for driver’s license without 20.264 (0.006) 20.218 (0.055) 20.229 (0.020) 20.331 (0.000) 20.286 (0.003)
education classPercentage below poverty 0.388 (0.000) 0.390 (0.000) 0.433 (0.000) 0.391 (0.000) 0.450 (0.000)
.190 (0.069) 0.121 (0.218) 0.117 (0.194) 0.144 (0.137)
0.385 0.384 0.542 0.50370 79 86 780.355 0.398 0.403 0.390
Percentage all private earnings from hotels 0.092 (0.305) 0.144 (0.154) 0.118 (0.214) 0.079 (0.360) 0.087 (0.374)
and lodges (tourism)(Constant) 0.323N 87R2 0.441
department, most localities send alcohol-related finesto the general fund of the state.
There is every reason to believe that public educationcampaigns would increase compliance with alcohol bev-erage laws, as has been shown with tobacco preventioncampaigns [55,56]; however, no localities had any sig-nificant budgets to mount any media campaigns.
Although we did not receive complete data from allstates and cities, using a robust statistical procedureto estimate missing data did not change our resultsand, in fact, strengthened their statistical significance.Another limitation is that the data are based upon the
self-report of state and city agencies. However, in mostFIG. 2. Total number of regulations by alcohol-related trafficfatalities per DVMT (r 5 20.456, P 5 0.000).
0.250 0.268 0.371 0.29870 79 86 780.420 0.440 0.464 0.427
measure of what is actually happening in the field. Theindices on DUI and enforcement should be viewed withcaution. The DUI data were not as complete, as re-sponses on these practices were received from only 81/97 (84%) cities included in the study. The use of state-level data in the city-level models violates the assump-tion of independence; therefore, the models includingstate-level beer consumption data as a mediator shouldbe viewed with caution. In addition, it is possible thatbeer consumption is a confounder having an indepen-dent effect on alcohol-related traffic fatalities, ratherthan serving as a mediator. While we have theoreticalreasons to suspect beer mediates the relationship[5,6,57], our analyses cannot differentiate between con-
ALCOHOL-RELATED TRAFFIC FATALITIES 195
TABLE 8
Regression Coefficients for Indices and Control Variables with Beer Consumption Added—Alcohol-Related Traffic Fatalities perDVMT Are Dependent Variable [b (P Values)]
Variable Model 1 Model 2 Model 3 Model 4 Model 5
Accessibility index 20.027 (0.798)Licensure index 20.147 (0.179)Discipline 20.082 (0.459)Enforcement index 0.155 (0.078)DUI index 20.136 (0.231)Beer consumption 0.304 (0.003) 0.275 (0.009) 0.240 (0.021) 0.264 (0.003) 0.246 (0.035)Minimum age for driver’s license without 20.289 (0.002) 20.217 (0.046) 20.237 (0.013) 20.307 (0.000) 20.249 (0.010)
education classPercentage below poverty 0.370 (0.000) 0.337 (0.001) 0.399 (0.000) 0.361 (0.000) 0.412 (0.000)
cases in which we had both state- and city-level reports,there was good agreement on the accessibility items.The highest disagreement was on enforcement prac-tices, which indicates that that index may not be a good
founding and mediation. Longitudinal studies areneeded. One final caveat is that our study populationdata are based upon city residence, while crashes mayreflect movement of people to and from suburbs andcentral cities.
CONCLUSION
In experimental studies, policies like the drinkingage and regulations on blood alcohol concentration havebeen shown to reduce alcohol-related traffic fatalities.It is likely that other policies regulating alcohol usewould have a similar effect. Given the cross-sectionalnature of the study, a causal relationship between thepolicies we have identified as being associated withalcohol-related traffic fatalities are not proven, but therelationships we have documented make good commonsense. Our study suggests that to reduce alcohol-relatedtraffic fatalities localities should consider greater re-
strictions on alcohol accessibility, disciplinary measuresfor alcohol outlets that violate alcohol beverage lawsshould not be waived, alcohol licensure requirementsshould be stricter, and BAC regulations ,0.8 with ran-dom sobriety checkpoints should be implemented. More
,
196 COHEN, MASONresearch is needed to determine the relative contribu-tion of beer versus other alcohol products to alcohol-related fatality rates. If beer is implicated as beingdisproportionately responsible, then additional effortsto regulate beer should be considered.
Because traffic fatalities are only one outcome ofalcohol use, alcohol regulations may have positive im-pacts on other health-related outcomes as well [58],including crime and violence [59,60], homicide and sui-cide [61–63], and even sexually transmitted diseases[64]. Alcohol was estimated to cost the United States$184.6 billion per year in direct and indirect costs ofmorbidity, mortality, and lost wages and productivityin 1998 [65]. This figure is second only to the impact
of tobacco and is 1.5 times greater than the cost ofillegal drugs to society [66]. If states or localities makesuch changes in their alcohol control practices, then itmay be possible to prospectively assess the full scopeof alcohol control policy benefits on health.APPENDIX: CORRELATION MATRIX OF INDICES
Accessibility Discipline Licensure Enforcement
Accessibility — — — —
Discipline 0.561 (0.000) (78) — — —
Licensure 0.551 (0.000) (70) 0.457 (0.000) (71) — —
Enforcement 0.074 (0.493) (87) 0.028 (0.808) (78) 0.031 (0.797) (70) —
DUI 0.315 (0.006) (75) 0.201 (0.106) (66) 0.110 (0.404) (60) 20.031 (0.793) (73)
Note. Pearson correlation (P value) (N ).
ACKNOWLEDGMENTS
We thank all the participating staff from state Alcohol BeverageControl agencies and the local police departments for their coopera-tion and for making this study possible. We acknowledge Dr. DavidMacKinnon, Arizona State University, for his help on handling miss-ing data.
REFERENCES
1. Recent progress in the genetics of alcoholism: tenth special reportto the U.S. Congress on Alcohol and Health. Washington: U.S.Department of Health and Human Services, 2000:169–81.
2. Psychosocial factors in alcohol use and alcoholism: tenth specialreport to the U.S. Congress on Alcohol and Health. Washington:U.S. Department of Health and Human Services, 2000:181–96.
3. Ashley MJ, Rankin JG. A public health approach to the preven-tion of alcohol-related health problems. Annu Rev PublicHealth 1988;9:233–71.
4. Ledermann S. Alcohol, alcoholism, alcoholisation. Vol. 1. Paris:Presses Univ. France, 1956.
5. Ledermann S. Alcohol, alcoholism, alcoholisation. Vol. 2. Paris:Presses Univ. France, 1964.
6. Skog O-J. The collectivity of drinking cultures: A theory of thedistribution of alcohol consumption. Br J Addict 1985;80:83–99.
7. Colhoun H, Ben-Shlomo Y, Dong W, Bost L, Marmot M. Ecologicalanalysis of collectivity of alcohol consumption in England: impor-tance of average drinker. BMJ 1997;314:1164–6.
8. Rose G. The strategy of preventive medicine. Oxford: OxfordUniv. Press, 1992.
AND SCRIBNER
9. Terris M. Epidemiology of cirrhosis of the liver: national mortalitydata. Am J Public Health 1967;57:2076–88.
10. Rush BR, Gliksman L, Brook R. Alcohol availability, alcoholconsumption and alcohol related damage. I. The distribution ofconsumption model. J Stud Alcohol 1986;47:1–10.
11. Kreitman N. Alcohol consumption and the preventive paradox.Br J Addict 1986;81:353–63.
12. Scribner RA, Cohen DA, Fisher W. Evidence of a structural effectfor alcohol outlet density: a multilevel analysis. Alcohol Clin ExpRes 2000;24:188–95.
13. Centers for Disease Control and Prevention (U.S.) NCCDP, Officeof Smoking and Health. Preventing tobacco use among youngpeople: a report of the Surgeon General. Atlanta (GA): Depart-ment of Health and Human Services, 1994.
14. Meier KJ, Licari MJ. The effect of cigarette taxes on cigaretteconsumption, 1955 through 1994. Am J Public Health 1997;87:1126–30.
15. Center for Science in the Public Interest. State alcohol taxes andhealth: a citizen’s action guide. Washington: Center for Sciencein the Public Interest, 1996.
16. Manning WG, Blumberg L, Moulton LH The demand for alcohol:the differential response to price. J Health Educ 1995;14:123–48.
17. Cook PJ, Tauchen G. The effect of liquor taxes on heavy drinking.Bell J Econ 1982;13:379–90.
18. Mosher JF, Beauchamp DE. Justifying alcohol taxes to publicofficials. J Public Health Policy 1983;4:422–39.
19. National Highway Traffic Safety Administration. Fatality analy-sis reporting system: 1995–97. Washington: U.S. Department ofTransportation, 1998.
20. Reducing alcohol impaired driving: tenth special report to theU.S. Congress on Alcohol and Health. Washington: U.S. Depart-ment of Health and Human Services, 2000:375–96.
21. Hingson R, Heeren T, Winter M. Lowering state legal bloodalcohol limits to 0.08%: the effect on fatal motor vehicle crashes.Am J Public Health 1996;86:1297–9.
22. Sloan FA, Reilly BA, Schenzler C. Effects of prices, civil andcriminal sanctions, and law enforcement on alcohol relatedmortality. J. Stud Alcohol 1994;55:454–65.
23. Wagenaar AC, Holder HD. Effects of alcoholic beverage serverliability on traffic crash injuries. Alcohol Clin Exp Res 1991;15:942–7.
24. Holder HD, Wagenaar AC. Mandated server training and reducedalcohol-involved traffic crashes: a time series analysis of theOregon experience. Accid Anal Prev 1994;26:89–97.
25. Scribner RA, MacKinnon DP, Dwyer JH. Alcohol outlet densityand motor vehicle crashes in Los Angeles County cities. J StudAlcohol 1994;55:447–53.
26. Reynolds RI, Holder HD, Gruenewald PJ. Community preventionand alcohol retail access. Addiction 1997;92:S261–72.
27. Gruenewald PJ, Ponicki WR. The relationship of the retail avail-ability of alcohol and alcohol sales to alcohol related trafficcrashes. Accid Anal Prev 1995;27:249–59.
28. Wagenaar AC, Zobeck TS, Williams GD, Hingson R. Methodsused in studies of drink–drive control efforts: a meta-analysis ofthe literature from 1960 to 1991. Accid Anal Prev 1995;27:307–16.
29. Legge, JS, Jr, Park J. Policies to reduce alcohol-impaired driving:evaluating elements of deterrence. Soc Sci Q 1994;75:594.
30. GAO report: drinking age laws: an evaluation synthesis of theimpact on highway safety. Washington: Govt. Accounting Office,1987 Mar.
31. Hingson R, Heeren T, Winter M. Lowering legal blood alcohollimits for young drivers. Public Health Rep 1994;109:738–44.
ALCOHOL-RELATED TRAFFIC FATALITIES 197
32. Voas RB, Tippetts AS. The relationship of alcohol safety laws to 49. U.S. Bureau of the Census. 1990 census lookup tables; STF3A.U.S. Department of Commerce. http://venus.census.giv/cdrom/drinking drivers in fatal crashes. Washington: National Highway
Traffic Safety Admin., 1999.33. Zador PL, Lund AK, Fields M, Weinberg K. Fatal crash involve-
ment and laws against alcohol impaired driving. J Public HealthPolicy 1989;10:467–85.
34. MacKinnon DP, Scribner R, Taft KA. Development and applica-tions of a city-level alcohol availability and alcohol problemsdatabase. Stat Med 1995;14:591–604.
35. Holder HD. Changes in access to and availability of alcohol inthe United States: research and policy implications. Addiction1993;88:67S–74S.
36. Wittman FD, Hilton ME. Use of planning and zoning ordinancesto regulate alcohol outlets in California cities. Adv Subst Abuse1987;Suppl 1:337–66.
37. Wagenaar AC, Wolfson M. Enforcement of the legal minimumdrinking age in the U.S. J Public Health Policy 1994;15:37–53.
38. Measures of reliability in scale problems. In: SPSS base 9.0 appli-cations guide. Chap. 17. Chicago: SPSS, Inc., 1999:362.
39. Arbuckle. AMOS, version 4. Chicago: SmallWaters Corp., 1999.40. Hakim S, Shefer D, Kakkert AS, Hocherman I. A critical review
of macro models for road accidents. Accid Anal Prev 1991;23:379–400.
41. Zlatoper TJ. Models explaining motor vehicle death rates in theUnited States. Accid Anal Prev 1989;21:125–54.
42. Federal Highway Administration. Highway statistics series(1995–97). Sect. 5. Roadway extent characteristics and perfor-mance. Washington: U.S. Department of Transportation, 1998.[Table HM-72 under urbanized area summaries]
43. Keeler TE. Highway safety, economic behavior and driving envi-ronment. Am Econ Rev 1994;84:684–93.
44. Federal Highway Administration, Office of Highway Manage-ment. Drivers license administration requirements and fees.Washington: Department of Transportation, U.S. National DataManagement and Dissemination Division, 1996. [HPM 40; Publi-cation No. FHWA-PL-96-011]
45. Williams AF, Karpf RS, Zador PL. Variations in minimum licens-ing age and fatal motor vehicle crashes. Am J Public Health1983;73:1401–3.
46. American Association of Motor Vehicle Administrators. The fasttrack to vehicle services facts: a motor vehicle regulations andprocedures information guide. 1997–1998 ed., Chap. 8. Dubuque(IA): Kendall/Hunt, 1999:143–80.
47. Bureau of Economic Analysis. REIS: regional economic informa-tion system. U.S. Department of Commerce Geostat Website:http://fisher.lib.virginia.edu/cgi-local/reisbin.
48. U.S. Bureau of the Census. State and metropolitan area databook 1997–98. 5th ed. Washington: U.S. Department of Com-merce, 1999.
lookup/967653930.50. Apparent per capita alcohol consumption: national, state and
regional trends 1977–96. Washington: National Institute of Alco-hol Abuse and Alcoholism, 1998 Dec. [Surveillance Rep. 47]
51. SPSS statistical data analysis software for Windows, version10.0.5. Chicago: SPSS, Inc., 1999 Nov.
52. Simple and multiple linear regression. In: SPSS base 9.0 applica-tions guide. Chap. 12. Chicago: SPSS, Inc., 1999.
53. Gorman DM, Labouvie EW. Using social indicators to informcommunity drug and alcohol prevention policy. J Public HealthPolicy 2000;21:304–22.
54. Cohen DA, Scribner R, Kaplan S. Case studies of alcohol enforce-ment practices. [Technical report]
55. Pierce JP, Gilpin EA, Emery SL, White MM, Rosbrook B, BerryCC, Farkas AJ. Has the California tobacco control program re-duced smoking? JAMA 1998;280:893–9.
56. Elder JP, Edwards CC, Conway TL, Kenney E, Johnson CA,Bennett ED. Independent evaluation of the California TobaccoEducation Program. Public Health Rep 1996;111:353–8.
57. Kreitman N. Alcohol consumption and the preventive paradox.Br J Addict 1986;81:353–63.
58. Thakker KD. An overview of health risks and benefits of alcoholconsumption. Alcohol Clin Exp Res 1998;22:285S–98S.
59. Scribner RA, MacKinnon DP, Dwyer JH. The risk of assaultiveviolence and alcohol availability in Los Angeles County. Am JPublic Health 1995;85:335–40.
60. Speer PW, Gorman DM, Labouvie EW, Ontkush MJ. Violent crimeand alcohol availability: relationships in an urban community.J Public Health Policy 1998;19:303–18.
61. Scribner RA, Cohen DA, Kaplan S, Allen S. Alcohol availabilityand homicide in New Orleans: conceptual considerations or smallarea analysis of alcohol availability and related outcomes. J StudAlcohol 1999;60:310–6.
62. Gruenewald PJ, Ponicki WR, Mitchell PR. Suicide rates andalcohol consumption in the United States. Addiction 1995;90:1063–75.
63. Smith GS, Branas CC, Miller TR. Fatal nontraffic injuries involv-ing alcohol: a meta-analysis. Ann Emerg Med 1999;33:659–68.
64. Scribner RA, Cohen DA, Farley TA. Geographic relation betweenalcohol availability and gonorrhea rates. Sex Transm Dis1998;25:544–8.
65. The economic costs of alcohol abuse: tenth special report to theU.S. Congress on Alcohol and Health. Washington: U.S. Depart-ment of Health and Human Services, 2000:364–72.
66. The economic costs of alcohol and drug abuse in the UnitedStates—1992. National Institute on Drug Abuse, National Insti-tute on Alcohol Abuse and Alcoholism. http://www.nida.nih.gov/EconomicCosts/Chapter1.html#1.10.