Örjan Hallberg
Hallberg Independent Research
e-mail: oerjan.hallberg@swipnet.se
Web: www.hir.nu
Updated 2009-01-07
Abstract
A new method involving time dependent modeling has been used in order to analyze
traffic accident statistics in the Nordic countries. The method makes it possible
to predict traffic accident levels based on an estimate of the future car population.
Measured reductions in accident rates relate to the implementation of major
traffic safety regulations. Although the number of killed persons does not seem
to diminish the outcome indicates a continuous improvement in relative traffic
safety.
It has been demonstrated /1/ that the traffic accidental risk is a function of the age and experience of the driver. This risk can be characterized by a time dependent function. The total amount of accidents can then be seen as a function of both the number of new active drivers added each year to the population and this time dependency.
The object of this study is to perform an analysis of traffic accident statistics by the use of a recently developed software program, EriView 2000 /2/, that initially was developed for the field reliability analysis of electronics. This program automatically finds a life distribution that gives the best fit of the calculated to the measured failure response to the total installed population over time. This software has been used to analyze the field performance of many types of printed circuit board assemblies, always giving good correlation to other more time- and data consuming methods /3/.
As traffic statistics involves large numbers and long observation periods it was an attempting task to apply the method for research within this area. No model that includes time dependent risks for the drivers is known by the author. Ref. 5-6 are dealing with modeling of fatal rates but they do not account for the accumulation of driver experience over time.
The reported net number of cars added to the traffic each year in Sweden and the reported number of killed people in the traffic was used as the input to an EriView 2000 analysis, assuming an active driving time of 25 years. In order to characterize the traffic safety of today the software was used to find the life distribution that gives the best fit to reported number of killed people between 1975-1985. Figure 1 shows the number of added cars and the reported and calculated number of killed people each year since 1935. As can be seen the fit is excellent for the past 20 years. The prediction after 2003 is based on an estimated future car population /4/. Note that the prediction after 1985 is slightly better than predicted, thus indicating that an improvment of the general traffic safety has occurred as compared with the previous period 1977-1985. The improvement up to December 2007 can be quantified as 606 saved lives since 1986, most of these were saved since 1997 (Figure 2).
Figure 1. Annual car population growth, actually and estimated killed persons using the optimum parameters for the period of 1977-1985: sigma=2,09; t0,1%=0,48 years.
The good correlation between predicted and measured fatal rates shows that the net increase of the car population is a main factor to consider when predicting the number of people that will be killed in future traffic. The fact that only two parameters (sigma and t0,1%) and one variable (net increase of car population) are used in the model underscores this conclusion. Other models /5/ use up to 7 variables. A large number of variables does not necessarily make them more useful for prediction purposes. U. Brüde /7/ predicted the fatalities based on regression analysis between 1977-1989 also with good correlation with the outcome up to 1993 by using three parameters and one variable (traffic index). After 1993 the fatalities became lower than predicted but merged again in 1997.
Figure 2 shows the relation between reported and calculated number of killed people using the optimum life distribution parameters of 1977-1985. It should be pointed out that the safety belts were introduced gradually before 1975 and this is likely to be one important reason to the improvement seen between 1969-1977. There has been an improvement in the traffic safety in Sweden in terms of actual vs. predicted fatalities corresponding to 20% since 1997. One of the major contributors is thought to be the introduction of 2+1 safety racks on many highways.
Figure 2. Traffic fatal rate in Sweden relative to that in 1977-1985. An improvement is noticed in the period 1986-2001 as compared with the period 1975-1985.
One conclusion from this analysis is that we should try to stop the increase of the active car population. This is the only way to come closer to a reduced number of traffic killed persons. I think that 3000 000 cars are quite enough for this country, especially with regard to the problems we all have with CO2 outlet. Obviousely, we should increase the scrap bonus for scrapping old cars and support a smaller, modern car population. This would surely pay off.
In order to see the effect of a reduction in the increase of the car park, the future number of traffic killed people was calculated assuming a net increase of only 10 000 cars per year. The result is shown in Figure 3. At least the first 10 years would follow this prediction quite closely. The total effect would be 2383 saved lives up to 2015 that is equal to a 30% improvement compared to the non-restricted, wild car population that we have today.
In a shorter perspective the likely outcome in 2007 is 90% (from Figure 2) out of the 528 calculated cases (from Figure 1) equal to 475 cases.
Figure 3. The expected effect of a controlled car population expansion to max 10000 cars per year.
This would be a very efficient corrective action to a low cost. The costs would be related to extended scrapping of old cars and less profit for Volvo and Saab. Assume we will pay 10000 SEK per scrapped car and in average phase out 50 000 cars per year the total scrapping cost for the 15 years will be 7,5 BSEK. The price per saved live is then 3,14 MSEK which is very cost effective.
Ref 8 is a strategic analysis of cost effectiveness of different actions to improve traffic safety. This report does not consider any control of the total car population but mentions increased scrapping on its page 9. As an exampple a value of 124 MSEK/saved live is considered reasonable in this report.
1. S Alexandersson, "Trafikskador", SCB annual statistics, 1995 e.g.
2. P Oscarsson and Ö Hallberg, "EriView 2000 - A Tool
For The Analysis Of Field Statistics", Proc. ESREL 97, Lisbon, June 1997,
ISBN 0-08-042835-5.
3. Ö Hallberg, "Facts and Fiction about the Reliability of Electronics", Proc
ESREF’95, Bordeaux, Oct. 1995
4. Bilismen i Sverige, 1997, AB Bilstatistik, Stockholm
5. L Fridström et. al. Accid. Anal. And Prev., Vol 27, No 1, pp 1-20, 1995
6. U Brüde, Anal. And Prev., Vol 27, No 3, pp 405-410, 1995
7. U Brüde, "The development in the number of traffic fatalities",
KFB & VTI research 28A, 1999. ISSN 1101-2986
8. A Johansson and B Svensson, "Strategisk analys - underlagsrapport till
samplan om trafiksäkerhetsåtgärder", Vägverket/SIKA
August 1999.