With Karl Kim at the University of Hawaii, we developed a methodology for geocoding motor vehicle crashes using Honolulu Police Department motor vehicle accident records. Typically, motor vehicle crashes are identified on police report by a main street in which the crash occured and by a reference point, usually the nearest cross street. Thus, most crashes are assigned to the nearest intersection. However, we found that many of the geocoding routines in desktop GIS packages will create considerable spatial errors unless both the attribute and geographic databases are corrected. These errors can approach 50% in the assignment of geographic locations to crashes. To increase accuracy, the motor vehicle records must be cleaned prior to geocoding. In addition, the geographic files must be modified to incorporate intersections that don't actually exist on the TIGER files (e.g., the intersection of a freeway and an access ramp)


  • Ned Levine and Karl E. Kim, "The spatial location of motor vehicle accidents: A methodology for geocoding intersections", Computers, Environment, and Urban Systems. 1999, 22(6):557-576.

  • Karl E. Kim and Ned Levine, "Using GIS to improve highway safety", Computers, Environment, and Urban Systems. 1996, 20 (4/5): 289-302.
Map of Errors Produced in Geocoding 
Motor Vehicle Crashes