THE DISTRIBUTION AND PREDICTION OF DRIVER ACCIDENT FREQUENCIES

A LARGE, RANDOM SAMPLE OF DRIVERS WAS IDENTIFIED FROM CALIFORNIA DMV FILES AND BIOGRAPHICAL AND DRIVER RECORD INFORMATION WAS EXTRACTED FOR ANALYSIS. ANALYSIS INDICATED THAT THE ACCIDENT AND TRAFFIC CONVICTION FREQUENCIES WERE NOT POISSON-DISTRIBUTED AND THAT THE FREQUENCIES IN DIFFERENT TIME PERIODS WERE SIGNIFICANTLY CORRELATED. THE BETWEEN-INTERVAL CORRELATIONS AVERAGED 0.05 FOR ACCIDENTS AND 0.23 FOR CONVICTIONS. ANALYSIS OF THE BIVARIATE FREQUENCY DISTRIBUTIONS INDICATED THAT 87 PERCENT OF THE ACCIDENT IN 1 YEAR (1963) INVOLVED DRIVERS WHO HAD BEEN ACCIDENT-FREE IN THE PRECEDING 2 YEARS (1961-62). THE CONCURRENT PREDICTION OF ACCIDENTS FROM BIO-DRIVER RECORD DATA YEILDED MULTIPLE CORRELATIONS (R'S) RANGING FROM 0.20-0.28 FOR MALES AND 0.15-0.24 FOR FEMALES. THE NONCONCURRENT REGRESSION ANALYSES (1963 ACCIDENT FREQUENCY PREDICTED FROM 1961-62 DRIVER RECORD DATA) RESULTED IN SOMEWHAT LOWER MULTIPLE R'S. THE TOTAL NUMBER OF MOVING TRAFFIC CONVICTIONS WAS BY FAR THE BEST UNIQUE ACCIDENT PREDICTOR FOR BOTH SEXES, WHEREAS THE SPECIFIC VIOLATION TYPES AND PRIOR-TO-ACCIDENT HISTORY MADE ONLY SLIGHT PREDICTIVE CONTRIBUTIONS BEYOND TOTAL CONVICTION FREQUENCY. THIS FINDING WAS INTERPRETED AS EVIDENCE AGAINST COMPLEX NEGLIGENT OPERATOR POINT SYSTEMS IN WHICH THE VARIOUS VIOLATION AND ACCIDENT TYPES ARE WEIGHTED DIFFERENTLY. QUESTIONNAIRE INFORMATION OBTAINED FROM A SUB-SAMPLE OF DRIVERS INCREASED VALIDITY COEFFICIENTS (R SQUARED) BY APPROXIMATELY 30 PERCENT BEYOND BASIC DRIVER RECORD DATA. BY FAR THE MOST IMPORTANT SINGLE QUESTIONNAIRE VARIABLE WAS ANNUAL MILEAGE. THE STUDY FINDINGS SUPPORTED THE HYPOTHESIS THAT DRIVERS DIFFER IN ACCIDENT LIABILITY OR PROENESS. HOWEVER THE EXPLAINED VARIABILITY OF ACCIDENT IS SMALL RELATIVE TO RANDOM AND NON-PERSON CENTERED COMPONENTS. ALTHOUGH INCREASED PREDICTIVE VALIDITY COULD PROBABLY BE ACHIEVED BY THE ADDITION OF PSYCHOMETRIC AND BEHAVIORAL MEASURES TO THE EQUATION, HIGH-VALIDY COEFFICIENTS WOULD SEEM PRECLUDED BY THE TEMPORAL INSTABILITY OF ACCIDENT FREQUENCIES AND THE INSENSITIVE CRITERION SCALE PROVIDED BY ACCIDENT RECORDS. /AA&P/

• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/00014575
• Supplemental Notes:
  • Vol 2, No 4, PP 243-299, 3 FIG, 44 TAB, REFS, APP
• Authors:
  • Peck, R C
  • McBride, R S
  • COPPIN, R S
• Publication Date: 1971-3

Media Info

• Serial:
  • Accident Analysis & Prevention
  • Volume: 4
  • Issue Number: 1
  • Publisher: Elsevier
  • ISSN: 0001-4575
  • Serial URL: http://www.sciencedirect.com/science/journal/00014575

Subject/Index Terms

• TRT Terms: Correlation analysis; Drivers; Forecasting; Frequency distributions; High risk drivers; History; Regression analysis; Statistical analysis; Traffic conviction; Traffic crashes
• Uncontrolled Terms: Bivariate analysis; Violations
• Old TRIS Terms: Multiple correlation
• Subject Areas: Data and Information Technology; Highways; History; Safety and Human Factors;

Filing Info

• Accession Number: 00223550
• Record Type: Publication
• Report/Paper Numbers: HS-009861
• Files: TRIS
• Created Date: Aug 1 1971 12:00AM