4: The TRULS-1 Model for Norway

The TRULS-1 model for Norway was developed at the Institute of Transport Economics (Transportøkonomisk institutt – TØI) in Oslo, as part of the author’s PhD thesis submitted to the University of Oslo in April 1999.

Ancestor studies of the TRULS-1 model include the generalized Poisson regression models estimated by Fridstrøm and Ingebrigtsen (1991) and by Fridstrøm et al. (1995). The latter study is a four-country analysis for Denmark, Finland, Norway and Sweden, focusing on the role of random variation, exposure, weather and daylight conditions in explaining casualty counts. It shows that, when account is taken of the inevitable (objective) random variation present in accident counts, there is a limit to the obtainable goodness-of-fit. When the data set consists of many small accident counts, this limit could well drop below 50 per cent, as judged by the ordinary R² measure. In the Norwegian data by province and month, randomness accounts for a full 80 per cent of the total variation in fatal accidents, yielding a maximal obtainable R² of about 0.20.