5 Estimation Approaches, Results and Discussion
I estimate the equation systems provided in section 4.1 using two alternative estimation
methods as only very little experience in estimating mobile competition models is cur-
rently available from the literature. In doing so I try to provide some more insight into
alternative methods which fit the structure typically assumed in theoretical models.
The first approach is standard generally least squares with heteroscedasticity robust
standard errors. I estimate incoming traffic and termination rates separately thus ignor-
ing any endogeneity of termination rates on incoming traffic. The alternative method
is a simultaneous estimation approach (3SLS) where the term log (termination rate) is
assumed to be endogenous. In doing so this term is explained with the variables of the
termination rates equation.17
The results of the alternative estimation approaches for own investment effects are given
in table 4. First, I separately estimate the investment effects on own termination rates
and MOU (columns (1) and (2)), and on competitors’ termination rates and MOU
(columns (3) and (4)) and afterwards combine them in one equation system (columns
(5) and (6)).
By comparing the investment coefficients of the OLS estimations with those of the 3SLS
estimations lower investment coefficients in the termination rate equations and higher
coefficients in the traffic equations are found. The deviation of the OLS coefficients from
the 3SLS coefficients for investments is driven by ignoring the endogeneity of termination
rates in the traffic equation. With higher investments, first, the investor’s own per-unit
costs and, second, also the competitors’ termination rates are affected. Ignoring the
(positive) indirect effect of investments on traffic leads to a larger capex coefficient and
a higher termination rate coefficient for the OLS approach.
17 For a more technical description of the implementation of 3SLS estimation approaches see Cameron
and Trivedi (2006), Cameron and Trivedi (2009).
21