In 1989 the Dutch government started the project ''Safety in Inland Wa
terway Transport'' to establish a minimum safety lever and to develop
a model to assess the effect and effectiveness of new safety measures.
This model is called the Risk Effect Model, and calculates the integr
al impacts of safety measures for the entire waterway system including
the risks of transporting dangerous goods. The final result of the pr
oject is a framework for evaluation, which supports cost-benefit analy
sis by weighing negative economical effects against achievements in sa
fety, for different measures. In this paper we will present the method
s which have been used to calculate the probability of an accident, us
ing casuistry. In this project the probability of an accident is model
led per elementary traffic situation (a combination of several ships c
arrying out a ship motion produces a traffic situation), The number of
accidents can be estimated by the number of elementary traffic situat
ions multiplied by the probability of an accident per elementary traff
ic situation. In the paper we describe fitting procedures in order to
obtain the model that ''forecasts'' the probability of accidents as fu
nction of waterway attributes and circumstances. We have used Generali
zed Linear Models (GLM), which do not need the assumption that the acc
ident probability is normally distributed. We have used the binomial a
pproach in the GLM models. We present the results of the fitting proce
dures for one group of elementary traffic situations (through-going tr
affic mutually). The primary governing variables appear to be visibili
ty, wind speed, the ratio of the navigable width and the necessary wid
th for an elementary traffic situation, and the bend radius of the wat
erway. The circumstances (visibility and wind speed) are more explanat
ory with respect to the probability of accidents than the waterway cha
racteristics are.