It is straightforward to see that ›ln Cijt=›uijt and ›ln Cijt=
›yijt are the same in the two specifications. Since both
Sweden and Finland are fairly small countries with factor
prices that are harmonized at large (marginal) prices are
assumed to be equal across track units.
The data section suggested that there are two
different ways of classifying a track unit in Sweden;
main and secondary lines. In order to understand whether
this distinction has something to say about differences
in maintenance techniques, the parameters for track
length and utilization level are estimated separately for
the respective type of railway lines. In Finland, where
this classification is not used, the information about
whether the line is electrified or not is used as a proxy.
We can then test for whether the track length and
utilization level parameters are the same for these two
classes of railway lines. Furthermore, we also test for
whether our data should be pooled over the 3 years, i.e.
whether we can set the restriction that bz; byy; buu; and
byu are constant over the 3 years. For the Swedish data,
we allow for constant district dummies aj; j ¼ 1;…; 20:
Altogether, this gives us