Marginal railway infrastructure cost estimates in the presence of unobserved effects

New railway legislation in Sweden has increased the need for transparent access charges on the Swedish railway network. The purpose of this paper is to estimate cost functions for infrastructure operation, maintenance and renewal in the Swedish national railway network, using unobserved effects models. The estimated cost functions are used to calculate the marginal cost for railway infrastructure wear and tear and give the Swedish National Rail Administration (Banverket) support in designing a marginal cost based pricing scheme in line with Swedish transport policy. A panel of 185 track sections is observed during 1999-2002. Collected data covers infrastructure, traffic and cost data. Despite having a rich set of information, unobserved effects are assumed to exist, which will bias estimates by ordinary least squares (OLS) due to omitted variables. We therefore use random effects specifications to estimate our models. We find evidence of unobserved effects at a track district level for infrastructure operation, maintenance and renewal costs. We estimate the marginal infrastructure operation cost to SEK 0.50 per train kilometre and the marginal maintenance cost to SEK 0.0029 per gross tonne kilometre. Combining maintenance and renewal cost increases the estimate to SEK 0.0065 per gross tonne kilometre. For all models, elasticities and marginal cost estimates are in line with previous work using pooled ordinary least squares (Andersson, In press).