There has been written quite some literature on costs of congestion. However, real data usually lacks in these papers and studies. Costs of congestion are calculated based on congestion functions, speed-flow functions, or other relations that give the speed or travel times as a function of the traffic flows or transport volumes. Some of these models are called bottleneck model, with an exponential function or BPR curve, with a 4th (or other) power function. These functions are assumed to be correct, or in more elaborate cases, calibrated to a few points. Usually this is the free flow speed (at night time) and the traffic volumes during a peak hour. In some cases, a speed-flow function is calibrated on a large set of data point, based on the traffic flow on a certain link. A calibration to reflect to correct speeds (or travel times) is rare. A calibration of the actual shape of the function is even quasi non-existing. However, it is actually the shape of the speed-flow function that matters when calculating the exact costs of congestion. A proper impact assessment of any transport policy should take into account the costs of congestion. In fact, it usually even does, if only when arguing the case for a certain policy. Large transport infrastructure projects or road pricing projects are sold to the public with arguments as relieving the congestion pressure and providing a better accessibility. But most projects lack sufficient data to prove this. Even worse, in the case of road pricing, it is the derivate of the speed-flow function that matters. A full social marginal cost pricing is only achieved when the user price reflects the marginal cost of (amongst others) congestion. The marginal costs of congestion are highly dependable on the derivate of the speed-flow function, and therefore, the shape of it must be calculated with greatest care. In the case of the Belgian motorways, we estimated the speed-flow function for all available data points in 2005. This is 1800 traffic measurement detectors, where speed and flow is monitored for every minute of 2005. This data has been aggregated to vehicle-kilometres and time losses per motorway link (e.g. Antwerp-Brussels), which then resulted in a fully calibrated speed-flow function for each motorway in Belgium. The actual goal of the study was the calculation of the marginal costs of congestion for each motorway. A large set of assumptions on the shape of the function, and the aggregation level of the data have been checked. First results show that aggregate speed-flow functions are much flatter than usually assumed. As a consequence, the marginal costs of congestion are very modest. The shape of the function has a large influence on the marginal costs of congestion, though it never reached the levels sometimes seen it literature. The bottleneck model, and the BPR curve can be suitable for urban traffic, but certainly not for modelling congestion on motorways. Another and less surprising result is that the peak and off-peak figures differ. Again, even the peak hour results remain pretty low, which can be explain by the large amount of uncongested traffic in peak hours, e.g. outbound morning peak traffic. The off-peak hours show a congestion level only slightly below the peak hour traffic, explained by traffic jams due to incidents and to tourism and leasure peaks. For the covering abstract see ITRD E137145.
Samenvatting