To sum up, we have analyzed the results of the theoretical model using alternative esti-
mation approaches. We find a negative effect of investments on own termination rates
and no significant effect of investments on incoming traffic which means that the first
hypothesis cannot be rejected. Moreover, the estimations support the assumption in
the theoretical model that investments in mobile infrastructure are mainly cost-related.
With regard to the second hypothesis that investments positively affect competitors’
termination rates and also increase competitors’ incoming traffic the outcome is am-
biguous: While competitors reduce termination rates due to other MNPs’ investments
their incoming traffic rises. In the following extension the empirical findings will be
replaced into the wholesale profit functions and I will observe how investments affect
the investor’s and the competitors’ wholesale profits. In doing so I particularly want to
check whether competitors’ termination rate reduction is profit-increasing even in the
short-run.
Extension: Calculation of the Investment Effect on Prof-
its
By adopting the estimation results to the theoretical model I calculate the effect of in-
vestments on wholesale profits. I do this exercise only for the 3SLS results due to the
restrictions of the OLS approach. As the per-unit costs of call termination are very
small I fix it to zero. The change in profit is independent of the underlying retail market
pricing scheme meaning that the traffic dependent change in profits is identical whether
we consider linear retail pricing or non-linear retail pricing.
The investor’s absolute change in profits from incoming calls due to investments is given
by equation (5). Rewriting this equation yields:
∂πji i ∂(ti - ci) ∂ Pj qj,i
-kk~ = ~∂kΓ j+ + (ti- ci) ~∂~
(16)
(17)
Dividing (16) by profits per investment yields the relative profit change:
dπji ki = ∂(ti - Ci) ki + d ∑j qj,i ki
∂ki πj,i ∂ki ti ∂ki Pj qj,i
In table 4 both coefficients βiqn1v and βtq1 are not significantly different from zero in the
traffic equation whereas βiqn1v is weakly significant in table 5. Note that the quantity
26