consistent with the fact that the NYT increases the number of stories about
Democratic topics during the presidential campaign only if the incumbent
president is a Republican. When the incumbent president is a Democrat,
the aggregate count of stories about Democratic topics is already high out of
the presidential campaign, as synthetised by the positive coefficient on the
incumbent president dummy.
Taking into account definitions 2 and 3, one cannot reject the hypothe-
sis that the New York Times, with respect to Democratic topics, acts as an
electoral watchdog towards the incumbent president, in that during the pres-
idential campaign it differentially gives more emphasis to topics over which
the incumbent is weak.
Finally, it turns out that the larger is the newspaper (as proxied by the
total number of stories being sampled each month), the higher is the count of
stories about Democratic topics: this is a scale effect, which is very precisely
estimated and is consistently positive.
6.1.2
Stories about Republican topics
In table 5 I summarise the Poisson regression results about stories concerning
Republican topics, namely Law & Crime and Defense stories. The structure
of the table exactly parallels the one of table 4.
First of all, regarding the effect of the incumbent president being a Demo-
crat, the general finding is that across the different subcategories of news this
effect is statistically not distinguishable from zero, apart from the case of
news with no public policy content, where the effect is positive at 10% con-
fidence level. In five cases out of nine, the sign of this effect is positive, and
is negative in the remaining ones. Again, the coefficient is not significantly
different from zero in eight out of nine cases.
Secondly, the effect of the presidential campaign dummy is not signifi-
cantly different from zero in all but two cases, namely for TriState and NYC
stories. For TriState stories, there are 64% more stories about Republican
topics during the presidential campaign than out of it, when the president
is a Republican: this effect is different from zero at 5% confidence level. On
the contrary, there are almost 43% less NYC stories about Republican top-
ics during the presidential campaign than out of it, at 5% confidence level.
Across different categories of stories, in seven cases out of nine there are less
Republican stories during the presidential campaign than out of it.
Finally, the effect of the interaction term is positive and significantly
29
Convention