A CALCULATION OF THE EFFECTIVENESS OF CAR SEAT BELTS BY A COMPARISON OF UTILIZATION BY DAYTIME CAR DRIVER FATALITIES WITH UTILIZATION BY THE PARENT POPULATION.

June 2005

FOREWORD

The following changes have been introduced since the last (2000) study.

  1. Professor Elliott Levine of  the Philosophy Department, University of Winnipeg has assisted me by making keystone suggestions in the production of this paper over the past two years.   It was his crucial suggestion that since NOPUS parent data was obtained during daylight, then the derivative FARS data (of fatalities) should also be limited to daylight, rather than presuming (as I had) that the daylight wearing trend would extend to 24 hours.  He was also responsible for drawing my attention to a discrepancy which led to my discovery of the document in heading (2).
  2. The earlier study had to be revised following publication of a document explaining details on the use of the FARS database.
  3. The police reported "unknown" and police "not reported" data for alcohol affected fatalities have been allocated.(Table 3 Notes)
  4. It is argued that ignoring the belt status "Unknown" data in FARS provides a conservative result.
I have not extended the study beyond 1994 - 1996 because air bags became a possibly significant influence on data collected after that date.

INTRODUCTION.

The novel procedure employed in this study is to compare the belt utilization of live car drivers (the parent population) with the belt utilization of car driver fatalities.  This procedure avoids problems found in time-step and pair comparison methods.  The time-step procedure is susceptible to chance disturbances caused by such factors as engineering improvements (such as collapsing steering columns), economic influences (such as petrol prices) legislative action (such as police blitzes on alcohol usage) or even the weather.   The pair comparison procedure relies upon self reporting.  Evidence indicates that even in the absence of penalties, drivers will report that they were wearing a belt when they were not.

Belt utilization by the parent population, defined as "daytime car drivers on urban or rural roads" was obtained from NOPUS (National Occupant Protection Use Survey) studies.  The NHTSA commissioned NOPUS to estimate the proportion of car occupants that wear belts.  NOPUS has stated that "NOPUS are not inconsistent with the national use rate estimates as calculated from state surveys."

The data on belt utilization by daytime car driver fatalities was extracted from FARS (Fatality Analysis Reporting System) which is an online database of all road fatalities in the USA provided by NHTSA.  This link leads to documentation of the data mining procedure from FARS.

NOPUS TABLE

Table 1 was constructed using three 1994 and two 1996 NOPUS (NHTSA) research notes.

  1. a "moving traffic" study in 1994 where observers determined the shoulder and lap & shoulder belt use of 167,000 car drivers by region.
  2. a "controlled intersection study" in 1994, where observers determined the shoulder and lap & shoulder belt use at intersections of 50,000 car drivers according to age, race and urbanization in 1994.
  3. a "shopping center study" where observers were able to determine ratios of use of lap, shoulder and lap & shoulder belt use of 5,500 car drivers as they alighted from vehicles in a shopping center.  Lap belt alone was found to be used by 1% of total drivers in 1994.
  4. a second "moving traffic" study in 1996 where observers determined the shoulder and lap & shoulder belt 176,651 car drivers by region.
  5. a "controlled intersection" study 1996 where observers determined the shoulder and lap & shoulder belt use at intersections of 49,387 car drivers according to age, race and urbanization in 1996.
TABLE 1 - NOPUS ESTIMATES.
YEAR
1994
1996
Moving Traffic 64.2%(3.6%)
65.1%(4.2%)
Intersctn City 60.6%(8.2%) 65.6%(4.0%)
Intersctn Suburbs 67.3%(2.6%) 69.1%(3.4%)
Intersctn Rural 60.6%(6.5%) 64.3%(9.0%)
Intersctn (C+S+R) 63.2%(3.9%) 66.2%(3.6%)
Shopping Center
68.0%
N/A
Urban Adjusted 66.7%(5.4%) 68.5%(3.7%)
Rural Adjusted 61.6%(6.5%) 65.3%(9.0%)

TABLE 1 NOTES

  1. The data given in parentheses (nn%) are NOPUS-calculated two standard deviations.
  2. The first six rows of data (columns 1994 & 1996) were obtained from the five NOPUS papers listed above.
  3. The "shopping center" study established that about 1% of drivers used a lap belt.  Only shoulder or shoulder & lap belts were counted in the other four studies.
  4. "Urban Adjusted" was calculated by combining city & suburban intersection wearing rates, (using linear interpolation weighted by SD), and increased by 1% to account for lap belts.
  5. "Rural Adjusted" is from rural intersection increased by 1% to account for lap belts.
  6. NOPUS studies were carried out during daylight hours (8am - 6pm).
DAY TABLES

Following is an extract from the NHTSA web site:

The Fatality Analysis Reporting System (FARS) contains data on a census of fatal traffic crashes within the 50 States, the District of Columbia, and Puerto Rico. To be included in FARS, a crash must involve a motor vehicle traveling on a trafficway customarily open to the public and result in the death of a person (occupant of a vehicle or a non-occupant) within 30 days of the crash. FARS has been operational since 1975 and has collected information on over 989,451 motor vehicle fatalities and collects information on over 100 different coded data elements that characterizes the crash, the vehicle, and the people involved.
To obtain as near a match as possible with NOPUS, the FARS data was extracted for daytime (8am - 6pm) fatalities for drivers of motor cars on urban and rural roads.

Urban results for 1994 - 1996 are shown in Table 2a, rural results are on Table 2b.  Later years were not considered because of the increasing use of airbags.


TABLE 2a DAYTIME URBAN CAR DRIVER FATALITIES
VARIABLE 1994 1996
Unbelted fatalities 507
533
Belted fatalities 639
715
Status Unknown 126
125
Belted fatalities % 55.8%
57.3%
NOPUS 66.7%
68.5%
Effectiveness "E"
37%
38%

TABLE 2b DAYTIME RURAL CAR DRIVER FATALITIES
VARIABLE 1994 1996
Unbelted fatalities 990
927
Belted fatalities 1128
1129
Status Unknown 98
136
Belted fatalities % 53.3%
54.9%
NOPUS 61.6%
65.3%
Effectiveness "E"
29%
35%

TABLE 2 NOTES.

  1. In approximately 12% of urban fatalities and 6% of rural fatalities the police reported driver belt wearing status is marked as "unknown".  This failure to obtain data might have been due to intervention by a third party who removed the safety belt (perhaps in an attempt to administer first aid to the victim.)  The high (12%) urban value supports this possibility.  Another suggestion is that the quality of the accident might have been such that it was not possible to determine with any accuracy whether the driver was originally belted.
  2. "Belted fatalities %" was calculated by dividing the "Belted" column value for each year by the sum of "Belted" and "Unbelted" column values.   The Unknown row was ignored.
  3. Effectiveness factor of E was determined by the formula  E = 1 (Nb/Nu) x (1 - Nopus)/Nopus where Nb & Nu are the count of belted & unbelted fatalities, and Nopus is the relevant observed belt wearing ratio.
ALCOHOL EFFECT TABLES.

According to the NHTSA, drivers under the influence of alcohol are much less likely to wear safety belts.

The proportion of drivers in the parent population who are alcohol affected is estimated as less than 1%.

However it is variously estimated that over 20% of drivers involved in fatal accidents are alcohol affected.

It follows that a small (< 1%) proportion of drivers who rarely wear belts produces a large proportion (>20%) of fatalities.  This imbalance generates an error, known to statisticians as Simpson's Paradox.   To correct for this error, driver fatalities that were reported by the police to be alcohol affected were isolated from the FARS data collected.

CONCLUSIONS

The calculated result is shown in Tables 2a and 2b.  If the "failure to report" entries in the FARS data are random, then seat belts provide an effectiveness of around 35%.  This is somewhat lower than the figure of 45% suggested by the NHTSA, but establishes the efficacy of safety belts.

REFERENCES

BOOKS
  1. Adams - John Adams' "Risk" (UCL 1995)
  2. Evans - Leonard Evans' book "Traffic Safety and the Driver" (Van Nostrand Reinhold 1991)
  3. Wilde - Gerald J S Wilde's "Target Risk" (PDE 1994) also available online
NHTSA

(NOTE February 2012. The links below are now dead.  However enquiries at NOPUS NHTSA might - or might not- be productive.)
  1. FARS.   Fatality Analysis Reporting System
  2. NHTSA Estimating Lives saved by Restraint Use in Potentially Fatal Crashes. 1995
  3. NOPUS 1994 - Controlled Intersection Study ~ US Department of Transportation.  (I have placed a copy here)
  4. NOPUS 1995 - Shopping Center Study~ US Department of Transportation.
  5. NOPUS 1995 - Observed Safety Belt Use in 1994 ~ Moving Traffic Study US Department of Transportation
  6. NOPUS 1996 - Controlled Intersection Study ~ US Department of Transportation.
  7. NOPUS 1997 - Observed Safety Belt Use in 1996" ~ research Note US Department of Transportation
  8. NHTSA Alcohol Traffic Safety Facts 1996
Back to seatbelt index page