A STOCHASTIC MODEL FOR AUTOMOBILE ACCIDENT EXPERIENCE

A STOCHASTIC MOTOR VEHICLE ACCIDENT MODEL IS DEVELOPED. IT PERTAINS TO AN INDIVIDUAL DRIVER AND TO GROUPS OF DRIVERS AND FEATURES TWO ASPECTS OF ACCIDENTS: INCIDENTS AND SEVERITY. USING DATA ON 148,000 CALIFORNIA DRIVERS OVER A THREE-YEAR PERIOD, CANDIDATES FOR THE ACCIDET INVOLVEMENT DISTRIBUTION ARE EXAMINED WITH PARTICULAR REFERENCE TO THE HISTORICAL CONCEPT OF "ACCIDENT PRONENESS". IN THE CASE OF INDIVIDUALS, THE ASSUMPTION OF A POISSON PROCESS IS FOUND TO BE TENABLE. THE POISSONS PARAMETER REFERRED TO AS ACCIDENT RATE POTENTIAL, IS ESTIMATED USING AT LEAST SQUARES MULTIPLE POISSON REGRESSION PROCEDURE WHICH SIMULTANEOUSLY PROVIDES AN ESTIMATE OF THE VARIANCE FOR OUR ESTIMATE. THE METHOD YIELDS A SOLUTION COMPATIBLE WITH THE MAXIMUM LIKELIHOOD SYSTEM OF EQUATIONS. BASED ON MOTOR VEHICLE ACCIDENT COST STUDIES, A FUNCTION RELATIVE TO AN INDIVIDUAL, IS ASSUMED TO BE THE DISTRIBUTION FUNCTION OF AN EXPONENTIAL RANDOM VARIABLE. BECAUSE OF THE HIGH VARIABILITY IN COST DATA, A PROCEDURE ANALOGOUS TO THAT USED BY THE CASUALTY INSURANCE INDUSTRY IS PROPOSED. SOME OF THE CONSEQUENCES OF THE VARIABILITYT CHARACTERISTIC OF ACCIDENT DISTRIBUTIONS IS NOTED ALONG WITH SUGGESTIONS FOR FUTURE RESEARCH. /AUTHOR/

• Supplemental Notes:
  • 138 Pp, 6 FIG, 25 TAB, 45 REF
• Corporate Authors:

  North Carolina State University, Raleigh

  Raleigh, NC  United States  27695
• Authors:
  • Weber, D C
• Publication Date: 1970

Subject/Index Terms

• TRT Terms: Costs; Crash severity; High risk drivers; Mathematical models; Stochastic processes; Traffic crashes
• Uncontrolled Terms: Motor vehicle accidents
• Subject Areas: Data and Information Technology; Finance; Highways; Safety and Human Factors;

Filing Info

• Accession Number: 00220785
• Record Type: Publication
• Source Agency: Nat Safety Council Safety Res Info Serv
• Files: TRIS
• Created Date: Aug 26 1970 12:00AM