Appendix
A.1 The Indirect Investment Effect
The own investment effect and the investment externality are the investment effects usu-
ally analyzed in the literature. For example Vareda (2007) compares quality-increasing
and cost-reducing investments if unbundling is mandatory and if it is not. Valletti
and Cambini (2005) analyze investments with two competing networks and show that
cost-based regulation reduces the incentive to invest. As in the literature on network in-
vestments mainly two competing networks are considered, the indirect investment effect
is ignored.20 An indirect investment effect stems from the investment of one MNP on
the traffic between two other MNPs. By reducing its per-unit costs the outgoing traffic
from the investor is increased. As termination rates (at least partially) depend on the
average incoming calls from all MNPs, i’s investments affect all competitors’ termination
rates for incoming calls. With prices dependent on termination rates outgoing calls are
affected by any provider’s investments. In less concentrated markets a lower indirect
investment effect should be observed. In contrast, with one large MNP and a number
of smaller providers we expect a strong indirect investment effect from the larger one
to the others whereas the indirect investment effect should be lower for investments of
smaller MNPs.
With termination regulation the indirect investment effect disappears as all forms of
price regulation ignore traffic and, thus, the influences from investments.
The profit stemming from j’s incoming calls from any MNP -j 6= i is given by P-j6=i π-j j,j
= (tj - cj)sj -j6=i s-j (a - b-j p-j,j). Although i is not involved in the interconnection
between j and any competitor -j 6= i, as we know from (6), i’s investment in cost-
reduction affects j ’s termination rate . The increase in wholesale prices is passed on
to -j’s off-net prices altering -j’s outgoing traffic. Deriving j’s wholesale profit with
respect to ki yields:
-j 6=i
πj
j,j
∂ki
c0i(ki)
sibi
4 -j s-jb-j
sjXs-j a
b-j
-j
cj
c-j + —
. j 2
Σ-j s-j (a - b-jc-j)
4 P-j s-jb-j
+ci (ki)(tj -cj)sj s-j b-j
-j
4 -js-jb-j
(21)
= -c0i (ki)
ibi
4 -j6=is-jb-j
s-j
(a-b-jc-j) -b-j
-j
Σ-j s-j (a - b-jc-j)
P-j s-jb-j
20We discuss this effect only in the theoretical part for reasons of completeness as we cannot single it
out with our data set.
33