Applicability cost-benefit analysis

Do all road safety measures lend themselves equally well to cost-benefit analysis? No, such analyses are more readily done for some measures than for others.

In the Handbook of Road Safety Measures [16], the following main groups of road safety measures are identified:

General purpose policy instruments is a heterogeneous bag of measures that includes, among other things, motor vehicle taxation, regulation of commercial transport, urban and regional planning and access to medical services. Most of the general purpose policy instruments are quite complex and their effects on road safety are indirect and for some of the measures poorly known. Due to their great complexity and the comparatively poor state of knowledge regarding their effects, these measures do not lend themselves very well to cost-benefit analysis. This is not to say that it is impossible to do cost-benefit analyses of some of these measures. There have, for example, been several cost-benefit analyses of road pricing.

In general, to be amenable to cost-benefit analysis, a road safety measure should satisfy the following criteria:

It should be known what category of accidents the measure affects (all accidents, accidents involving young drivers, accidents in the dark, etc), preferably so that the number of "target" accidents can estimated numerically.

The effects of the measure on target accidents should known, i.e. numerical estimates of these effects should be available. If possible, these estimates should state the severity of accidents or injuries they apply to.

In short, cost-benefit analysis requires quite extensive knowledge of the impacts of a measure. This knowledge will not be available for all road safety measures.

In a recent road safety impact assessment for Norway [13], a survey was made of 139 road safety measures. Only 45 of them were included in a cost-benefit analysis. A total of 94 measures were omitted. Reasons for omitting measures included:

Some of the measures that were included have so far not been used extensively, but were included because there is reason to believe they could improve road safety. This applies to ISA (Intelligent Speed Adaptation), for example, which favourably influences driving speed, a known risk factor for accidents and injuries.

To give a short example, consider the conversion of three leg junctions to roundabouts. From the Norwegian road data bank, it was determined that 120 junctions with a mean daily traffic of 12,000 are candidates for conversion to roundabouts. Thus, the effect on fatalities can be estimated as follows:

120 x 12,000 x 365 x 0.091 x 10^-6 x 0.018 x 0.49 = 0.42

The first three terms (120, 12,000, 365) denote the total traffic volume in the 120 junctions during one year. This is the traffic that will be exposed to the conversion. The next term (0.091 x 10^-6) is the mean risk of injury per million vehicles entering a three leg junction. A little less than 2 % of the injuries (0.018) are fatal. The rest are serious or slight. Thus, the overall injury rate is decomposed into a rate of fatal injury, a rate of serious injury and a rate of slight injury. Finally, roundabouts reduce the number of fatalities by 49 % (0.49). Hence, in the 120 junctions, 0.42 fatalities will be prevented.

The fatalities prevented can be converted to monetary terms as follows:

0.42 x 26.5 x 14.828 = 165 million NOK

Here, 0.42 is the number of fatalities prevented, 26.5 is the value, in million NOK, of preventing a fatality, and 14.828 the accumulated present value factor for a 25-year time horizon using a discount rate of 4.5 % per year. In general the present value of a benefit (or cost) is estimated as:

Present value = Present value =

In this formula, B denotes benefit in year i and r is the discount rate. The summation is from year 0 to year n, the end of the time horizon considered. Thus, if the benefit in year 0 is 100, in year 3 it will be:

100/(1.045³) = 100/1.1412 = 87.6

As the years pass, the present value of a constant stream of benefits gradually becomes smaller.