II. Data and Terminology

A. Data on Fatal Crashes

A state-by-state database was created for the study that includes motor vehicle fatality information, demographic and economic data, and impaired driving laws and enforcement. This section describes the specific items in the database and their sources.

Drivers in fatal crashes. Data on fatal crashes came from NHTSA's Fatality Analysis Reporting System (FARS) and its supplemental Alcohol Imputation files. FARS is an enumeration of virtually all fatal motor vehicle crashes that occur in the U.S. The Alcohol Imputation files are based on available BAC data or use statistical methods to estimate (or impute) whether or not a crash involved alcohol if there is no direct evidence of alcohol presence or absence.

The Alcohol Imputation files are made up of two parts. The part used in this study contains probability values that each driver and nonoccupant in a fatal crash had a BAC of 0.00, a BAC of 0.01-0.09, and a BAC of 0.10 or more. The second part contains probability values at the crash level. That is, based on the driver and nonoccupant probabilities, a probability value is estimated that the crash involved drivers and nonoccupants with only 0.00 BAC, that the crash involved at least one driver or nonoccupant with a BAC in the 0.01-0.09 range but none higher, and that the crash involved at least one driver or nonoccupant with a BAC of 0.10 or higher. For both parts, when a driver or nonoccupant has a known BAC (zero or positive), that BAC value is used. When the BAC is unknown, a discriminant function model estimates the probabilities that the person's BAC was in the three categories. The variables entered into the discriminant function are those from the FARS database (vehicle type, police reported alcohol, time, day of week, driver age and sex, etc.). The imputation methodology is described by Klein (1986a) and a guide for using the imputation files can be found in Klein (1986b).

FARS has used the Alcohol Imputation process since 1982. For this study, drivers in fatal crashes have been examined over the 17-year period 1982-1998. The data used in the analyses for 1998 are from the preliminary file, as the final 1998 version had not been released when these analyses were completed. The preliminary data for 1998 indicated that there were 8,116 drivers under the age of 21 involved in fatal crashes, with 1,704 of these determined to have been drinking. The final FARS version for 1998 shows that there were 8,128 drivers under the age of 21 involved in fatal crashes, with 1,714 of these determined to have been drinking.

Some states do not use these estimated data to report alcohol-related traffic fatalities. Instead, they count as alcohol-related only those crashes and drivers for which there is positive evidence from a BAC test, a police report, or another official source. These states will report a lower number of alcohol-related traffic fatalities and alcohol-related drivers in fatal crashes than the FARS estimates. In addition, some states use a different definition of alcohol-related, such as by including only crashes in which a driver had a BAC of 0.10 or greater. The FARS data were used in this study to assure that all states were compared on a common basis. While some states report different alcohol-related traffic fatality totals than FARS, in general the trends in FARS and state data are very similar. Trends also are similar regardless of what definition is used (total fatalities or drivers above 0.10 BAC, for example).

Data on Canadian fatal crashes were graciously provided by the Traffic Injury Research Foundation (TIRF), Ottawa (Beirness, 2000). The TIRF Fatality Database is supported financially by the Canadian Council of Motor Transport Administrators and Transport Canada.

B. Economic and Demographic Factors, and Laws

Population Data. The Bureau of the Census produces various annual estimates of the U.S. population. For this study, the most useful was the annual Estimates of the Population of the U.S. and States by Single Year of Age and Sex. Data on population by age in each state for the period 1982-1998 were compiled.

Economic Data. Annual state-by-state data were obtained from the Bureau of Labor Statistics on the number of persons employed, the number not employed, and the number not in the labor force for 1982-1997.

State Laws. NHTSA provided the dates when different traffic and alcohol laws became effective in each of the states.

Alcohol Consumption. Published data were obtained from the National Institute of Alcohol Abuse and Alcoholism (NIAAA) and the Beer Institute of America.

DWI and Liquor Law Violations. Data were obtained from the Federal Bureau of Investigation (FBI) Uniform Crime Reporting System.

C. Data on Drinking and Drinking-related Behavior

Monitoring the Future. The best data on youth drinking and drinking-related behavior come from the ongoing Monitoring the Future study. Approximately 17,000 high school seniors, from a nationally representative sample of about 135 schools, have provided confidential self-reported information on alcohol and drug use and related behavior each year since 1975. Results are reported annually (most recently in Johnston, O'Malley, and Bachman, 1999). Questions on drinking and driving have been included since 1984 for one-sixth of the sample. O'Malley and Johnston (1999) summarize and discuss these data and trends. Data for this study were obtained from Johnston, O'Malley, and Bachman (1999) and O'Malley (2000).

College Age Youth Surveys. While there are no ongoing surveys of college students comparable to Monitoring the Future, two studies provide information on college students' drinking and drinking-related problems, including driving after drinking, at several points in the past 20 years. Hanson and Engs (1992) surveyed approximately 4,000 students at 65 representative 4-year colleges in 1982, 1985, 1988, and 1991. Wechsler, Davenport, Dowdall et al. surveyed 17,552 students at 140 nationally-representative 4-year colleges in 1993.

General Public Surveys. Three other surveys provide useful information on drinking or drinking and driving for both young and other persons since 1980. Balmforth (1998) reports results from nationally representative surveys of approximately 4,000 persons age 16 and older in each of the years 1991, 1993, 1995, and 1997. Balmforth aggregates results into the age groups 16-20 and 21-29. Caetano and Clark (1997) surveyed nationally-representative samples of 1,947 black, 1,453 Hispanic, and 1,777 white adults (age 18 and older) in 1984 and 1,582 blacks, 1,585 Hispanics, and 1,636 whites in 1995. Midanik and Clark (1994) surveyed 2,058 adults in a national household probability sample in 1990 and compared their results with data from 5,221 respondents in a 1984 national alcohol survey. The latter two studies aggregate results for the age group 18-29.

D. Terminology

The following terms and abbreviations are used throughout the report.

BAC - alcohol content in the body, originally measured using alcohol in blood (BAC stood for Blood Alcohol Content, expressed in milligrams of alcohol per 100 milliliters of blood), now typically measured using alcohol in breath.

Drinking Driver - a driver involved in a traffic fatality who is estimated to have a positive BAC (as reported by FARS).

DWI - the offense of driving while impaired by alcohol. The formal offense differs from state to state (Driving While Impaired, Driving While Intoxicated, and Driving Under the Influence are common). Here, DWI will be used to describe each state's standard impaired driving offense.

Youth (or young driver) - a person (or driver) under 21 years of age.