There is an abundance of literature on cost functions and economies of scale in the rail sector. Results from such studies are used in decisions on privatization, regulation and mergers. Different studies may, however, come to very different conclusions. In this paper we examine the results of studies on rail cost functions, and try to explain the difference in the results in a statistical analysis. Some conclusions are that cost elasticities with respect to output(s) reported for Europe are significantly higher than the elasticities reported for North America. Elasticities reported for recent years are significantly lower than elasticities for earlier years.There are also indications that the elasticities are lower in studies using generalized Box-Cox or generalized Leontieff specifications of the cost function.