What is the logic of the results found above, i.e. what do
we learn about the structure of track maintenance costs? A
first observation is of technical nature; the Translog
specification of the functional relationship between costs
and explanatory variables seems to provide a good basis
for understanding the pattern of spending on track
maintenance. But although costs do not vary linearly
with variations in traffic and track length, non-linearities
are not very strong.
A second observation is that the explanatory power of
our model is high; information about traffic levels, track
length, number of switches and quality provides a good
intuition for understanding the levels of spending on track
maintenance. When information about these variables is
available, the classification of tracks to one administrative
class or another does not seem to add to our understanding
of maintenance patterns.
Third, similarities between the two countries are
considerable but not complete. In particular, the impact of
traffic levels on costs is less distinct in Finland than in
Sweden. The mean elasticity of costs relative to traffic is,
however, estimated to be strikingly similar. Is it then
logically consistent that marginal costs for track use are
almost 90% higher in Finland than in Sweden? The answer
must be affirmative. Each country has its own features of the
network. The two countries here, for instance, differ with
respect to track gauge and winter conditions in Finland are
generally harsher than in Sweden. In particular, Finnish but
not Swedish observations also include costs that are
common for the organization. The way in which these
common costs have been allocated across track units may,
per se, have an effect on our estimates. While the pattern of
spending—the logic of allocating resources for the purpose
of maintaining tracks—seems to be common, it is not