Evaluating road safety and safety effects using Empirical Bayesian method

The safety evaluation of a given road is frequently based only on accidents happened during recent years. However, due to random variation and rareness of accidents, accident history provides an unreliable estimate of the expected number of accidents in the future. This is even more true regarding evaluation of fatalities. Consequently, if the current safety situation is inadequately evaluated, one can nothing but fail in evaluating the effects of road safety improvements. Similar problems concern so-called black spots with relatively small number of accidents during few last years. The most reliable estimate of the current road safety situation on a given road can be received using Empirical Bayesian method. In Finland an evaluation programme entitled TARVA using that method was created in 1993 and it has been in constant use ever since. First, TARVA combines information about accident history with the accident model information to evaluate current safety situation. Secondly, effect coefficients are used to evaluate the safety impacts due to one or several simultaneous improvements. Thirdly, the severity of accidents is taken into consideration to evaluate the effects on fatalities. The evaluations can be done easily on the whole national road network