The two and three nodes networks technically analyzed here have been sufficient to prove
the main point, which is that network expansion should be studied through scope rather
than scale economies. As stated earlier, this approach was not intended to find optimal
service structures for transport firms, but to understand key issues relating cost functions,
transport networks and industrial organization. Nevertheless, it is worth asking whether
further insight could be gained by expanding this type of detailed analysis to larger
networks when feasible. We believe that it might help finding aggregate specifications for
transport product that are adequate for the empirical estimation of cost functions. It could
help as well to examine how optimal frequencies and route structures affect aggregates that
are used in the literature, as average distance or average load. It should be stressed, though,
that finding analytical cost functions for actual transport firms serving many OD pairs over
complex networks is simply unfeasible, which is precisely the reason why it is done
econometrically. The point is, however, that statistical feasibility should be no excuse to
forget neither the true product nor the type of decisions made by a transport firm on a
network when using this empirical cost functions for industrial organization analysis.