Must carry rules

18

System-level

variables

First, contrary to the main hypothesis, vertical

integration had a significantly negative effect (

β

= -.056, *p *< .01). That is, cable systems

whose owners had more program interests in cable networks were less likely to deny

carrying local stations. Specifically, a system owner that is vertically integrated with one

more cable network increased the probability and odds of a station being carried by its

system by 0.003 and 5.4% respectively, holding other variables constant.

26

This may be

because firms with cable program interests, particularly those larger cable firms, are often

partners of other media companies that have stakes in broadcast networks, TV stations

and other broadcast interests. They thus have less incentive to refuse carrying broadcast

stations on their systems.

The national size of a system’s owner, *LNSUB*, had a significantly positive effect

on *DROP* (

β

= .096, *p *< .05). That is, larger cable MSOs were more likely to drop a

broadcast station from their systems. On the average, a 1% increase in the number of a

cable MSO’s subscribers increased the probability of a station being dropped by 0.006,

holding other variables constant. In terms of odds ratio, that would increase the odds of a

station being denied carriage by 10%. Larger cable MSOs have been shown to have

lower marginal costs with respect to program supply and, as a result, tend to supply more

of a unit change in an independent variable; that is,

*i*

*i*

*p*

*x*

∂

∂

. For the logit model,

(1

)

*i*

*i*

*i*

*i*

*i*

*p*

*p*

*p*

*x*

β

∂

=

−

∂

. In the

current sample, the probability of *DROP=1* is 0.065. Therefore, the marginal effect of

*i*

*x*

is calculated as

* 0.065(1 0.065)

0.061*

*i*

*i*

*i*

*i*

*p*

*x*

β

β

∂

=

−

=

∂

.The odds ratio estimates measure, for a unit change in

*i*

*x*

, how

the odds of the dependent variable occurring change. They are calculated as

*i*

*e*

β

. Note that odds ratio is a

multiplicative measure. So an odds ratio greater than one means a positive effect, while a ratio between 0

and 1 implies a negative effect (Long, 1996, p. 73).

26

The results were similar when *VI* or *VI_TOT* were used instead of *VI_N*. Because *VI_N* and the

interaction term (*LNSUB*VI_N*) have a nearly perfect correlation, this result implies that larger MSOs with

program interest in more cable program networks were less likely to drop broadcast stations from their

systems, although the impact was minuscule.