4.2 Data Description and Data Adjustments
Data from multiple sources is employed including information about the fully vertically
integrated MNPs in the EU 15 as well as Norway and Switzerland, data about termi-
nation regulation and the termination rates, information about individual transmission
technology per provider and data about population concentration on a quarterly basis
between 2001 and 2007.13 Unfortunately, due to lags in the data structure we have to
ignore Greece and Luxembourg in the estimations. As not all information is provided in
the form required for the analysis I adjust the data.14 The most-relevant adjustment is
the calculation of incoming calling minutes. Among other sources data from the Merrill
Lynch European Wireless Matrix is used which provides only aggregated data on calling
minutes and does not distinguish on-net traffic from off-net traffic. Furthermore, it is
stated in the data description that due to double-counting of incoming and outgoing
calls total calling minutes are upward-distorted.15 I correct the data using standard ap-
proaches known from the theoretical literature (e.g. A-LRT). Usually, separating on-net
from off-net traffic is done on the basis of market shares assuming a balanced calling
scheme. The total off-net traffic originating from MNP j is calculated as one minus the
own market share times the total minutes of usage:
qj,l,z = (1 - sj,l,z)M OUj,l,z
where M OUj,l,z are the total minutes of usage of MNP j. The higher the individual
market share the more outgoing calls are terminated on the own network.
From these adjusted values I calculate the quantity of incoming calls to any network i
as
jj Qj,i,,z = j jι,z
(15)
= si,l,z Pj MOUj,l,z
As in Dewenter and Haucap (2005) information on costs of call termination is proxied
with variables like population concentration and a dummy for GSM 1800 technology.
The dummy takes the value 1 if the provider offers mobile services only in the frequency
band around 1800 MHz. If the provider also offers mobile services in the 900 MHz fre-
quency band or only in this lower frequency band the dummy takes the value of 0.
13 I appreciate the comprehensive support with data access to the Deutsche Telekom AG which en-
abled the access to non-publicly available information on traffic, company-specific information and
termination rates.
14 The description of raw data adjustments is provided in appendix A.4 and A.5.
15 See p. 18 of the Merrill Lynch European Wireless Matrix 2007.
17