Marginal cost pricing is an important keystone in Swedish transport policy. The infrastructure charge made by the Swedish Transport Administration to the train operators includes a component for rail-road level crossing accidents that should be based on the marginal cost principle.
The objective of the study is to examine the relationship between train traffic and the accident risk for road users at level crossings.
The accident charge today in Sweden due to level crossing accidents is set to 0.34 – 0.80 SEK/train km (Trafikanalys, 2014) based on an earlier study using accident records for 2000-2008 (Jonsson, 2011). The accident charge per km in the present study was estimated to SEK 0.78. In the earlier study, the average marginal cost per train passage was estimated to a value of SEK 1.13 in 2008, compared to SEK 1.28 in 2012 in the present study.
In the present paper we have enlarged the data set, which now also includes the years 2009 to 2012. We also base the estimation on other data regarding traffic volume, which now includes station areas. The values for the injuries are also updated according to the official Swedish values which can explain the, relatively small, difference in average marginal cost between the two studies.
The size of the marginal cost is not only influenced by the variables included in the model, there are also a lot of estimation specifications influencing the size of the cost. For example, in this paper we preferred an accident probability model based on logarithmic train traffic volume, which results in other marginal effects/costs compared to a quadratic model or a linear model.
We also preferred to calculate weighted average marginal effects and marginal costs. In that way, crossings with larger traffic volume have a larger impact on the marginal effects/costs than crossings with smaller traffic which we see as an advantage. The reason why we chose to calculate the weights over train passages year 2012 was that this is the year in our data set that consists of most updated information.
We also preferred road types as proxy variable for road traffic. On an aggregated level the marginal cost does not differ much from calculations based on road administrator instead of road type. However, because the road type variable is not updated, it is probably better to use the variable road administrator if accident charges are going to be differentiated on track sections.