A linked database composed of police crash reports and hospital records was used to compare seat belt and alcohol use. For these linked cases, the police reported a belt use rate of 88.1 percent; hospital records, however, indicated a use rate of 59.9 percent. Although police overreported the seat belt use rate, they underreported the alcohol involvement rate. Hospital records indicated alcohol involvement in 25.5 percent of the cases; police reporting captured only a fraction of the alcohol cases (8.4 percent of total cases). In addition to reporting the general patterns of inconsistency between police and hospital reporting across driver, vehicle, and crash characteristics, it was shown that there are significant problems associ- ated with reporting of seat belt and alcohol use. The analysis is based on comprehensive information from Hawaii of about 369 drivers with linked crash and hospital records in 1990. Some recommendations for improving the quality and accuracy of reporting are provided. The saying of Mark Twain that there are “lies, damn lies, and sta- tistics” serves as an important reminder of the need for accurate reporting in traffic safety. The “lie factor” in traffic safety is often mentioned with regard to studies that use police crash reports. Many suspect that people may falsely claim to have been belted or deny consumption of alcohol to avoid penalties from police. The growth in societal acceptance of both seat belt use and refraining from alco- hol consumption while driving may encourage crash-involved motorists to lie. The purpose of this research is to estimate the lie factor for seat belt and alcohol use among crash-involved drivers in Hawaii. In addition to reporting the magnitude of differences between police and hospital reporting for 369 linked cases (matched individual records), the relationship between driver, vehicle, and crash char- acteristics and the lie factor was analyzed. Situations in which motorists are more likely to lie were identified both as a starting point for improvement of police reporting and as a way to adjust and improve the efficacy of traffic safety programs. BACKGROUND This study was motivated by both a general curiosity about the lie factor as well as the availability of certain technologies and data- bases that would enable one to carry out a detailed investigation. At several national meetings, the issues of accuracy and reliability about both seat belt and alcohol use have been raised. O’Day ( 1) provides an excellent summary of many of the problems with using police crash reports and a useful discussion about the importance of data quality. There have been various initiatives undertaken over the years to enhance the quality of police-collected data involving agen- cies from the federal government as well as organizations such as the National Safety Council’s Traffic Records Committee. There has also been increased interest in and use of police crash data—to investigate traffic safety problems at the state and local level of gov- ernment, to help design new community-based countermeasures, and to evaluate various safety programs. The use of local data to support studies and analyses is often considered more desirable than either national studies or research from other places. The use of local data, however, has raised serious concerns about the quality and accuracy of police reports. The purpose of this research is to focus on two areas of concern: the reporting of seat belt and alcohol use among crash-involved drivers. Increasing seat belt use and decreas- ing alcohol involvement have long been priority areas in traffic safety. Many people, however, have suspected that police reporting of belt use is overstated and reporting of alcohol impairment is underreported. Given that a majority of states have adopted manda- tory seat belt use laws and increased penalties for driving under the influence (DUI), it is not unreasonable to suspect that motorists may not be truthful in reporting these activities to the police. DATA AND METHODS This study begins with an important assumption. Although there may be an incentive to lie to the police about seat belt and alcohol use, there may be disincentives for motorists to lie to physicians and other med- ical personnel. Moreover, alcohol use in Hawaii is established by blood testing, instead of by merely asking the patient whether he or she has consumed alcohol. As one emergency room doctor stated, “We routinely test for alcohol because we’re very concerned about the pos- sibilities of drug interactions with alcohol.” Similarly, medical per- sonnel have reported that there are patterns of injury, particularly in severe collisions, that are correlated with belted and unbelted victims. For these reasons, the reporting of hospitals with regard to belt and alcohol use is considered more accurate and reliable than information derived solely from police crash reports. To carry out this study, two different databases were combined. Comprehensive police crash reports were obtained from the state of Hawaii Department of Transportation, and records for all hospital- izations for motor vehicle collisions over the same period (1990) were obtained. The hospital data were collected as part of a larger study entitled “Injury in Hawaii,” funded by the Centers for Disease Control and Prevention. Both databases were edited and reformatted to allow for data processing. The intent was to understand general characteristics of the two populations (police crash reports and hos- pitalizations) and to provide sufficient inputs to carry out the proba- bilistic linkage of individual records across both databases. By using the Automatch software, police and hospital records were linked, so Lie Factor in Traffic Safety Comparison of Police and Hospital Reporting of Seat Belt and Alcohol Use in Hawaii KARL KIM University of Hawaii at Manoa, Department of Urban and Regional Planning, 2424 Maile Way, No. 107, Honolulu, HI 96822. TRANSPORTATION RESEARCH RECORD 1665 Paper No. 99-0386 141