In the selection of a market cost function we looked for a relation between the total network replacement costs and the number of lines in the service area.
SPSS for Windows was used to perform a regression line fitting to the replacement cost data from our 14 study service areas. We selected the Power curve which obtained the best fit.
The Power function has the following form.
The estimated power fuction has a 
and a
.
An F value was computed to test how well the
regression model fits the data.
This was F=142.58 with significance level .00001.
Although not a statistically comparative measure,
the 
values shows how the observed data points
fit to the Power curve that was estimated.
This was =.92237 with standard error =.18767.
In short, this is the network cost function for the incumbent
LEC, in our study Provider 0.
To estimate the network cost function for the new entrant we make
use of the Reed study.[11] He reported on new construction
costs of Fiber-to-the-Neighborhood Networks, with cable to the
home. Penetration rates in his study were based on reasonable
levels if a cable company were to compete against a telephone
company. The lowest average cost per subscriber was estimated
for the case of an Integrated Network for Telephone and Distributed
Video Services with penetration rates of 20 
and 40
respectively. For this case, the average cost per subscriber
was 1420 U.S. dollars.
If the Total Network Costs from our Table 1a. and 1b. are summed and divided by the total number of households we compute an average cost per subscriber of 902.36 U.S. dollars for our combined study region. We take the ratio of these two average costs, which is 1.57, and use that to differentiate costs of two providers in our interactive game model. We assume that the incumbent Provider 0 and the new entrant Provider 1 have cost functions that are the same shape. But we multiply the costs of the new entrant by a factor of 1.57.
