stations in similar formats. For example, the “Adult Contemporary” format category
contains formats such as Adult Contemporary, Soft AC, Lite AC, Lite Rock, and Soft
Rock. The identification strategy involves looking at how the similarity of station play-
lists, measured in various ways, changes when pairs, or small groups, of stations become
commonly owned or cease to be commonly owned.
46
Consistent with the spirit of Berry
and Waldfogel’s results, in the sense that they view the driving force behind increased
variety as a desire to avoid audience cannibalization, Sweeting finds that common owners
tend to differentiate their stations. He also finds that the merging stations tend to signif-
icantly increase their combined audience.
However, Sweeting also shows that, at the same time, common owners make at least
some of their stations more similar to stations owned by other firms, and that the listen-
ership of these stations tends to fall, by about as much as the merging stations gain, so that
when one looks at format listening as a whole, ownership consolidation is not associated
with significant changes, consistent with
Berry and Waldfogel’s 1999b listenership result.
To capture the intuition for what seems to happen, suppose that there are three indepen-
dent stations, A, B, and C, arranged in a two-dimensional product space, and that initially
they are arranged symmetrically (say, at the vertices of an equilateral triangle). Following a
merger between the owners of stations A and B, suppose that the new common owner
differentiates them by moving B further away from A. Potentially it could do so by mak-
ing the station more differentiated from C as well, but the data suggests that it actually
makes at least one of its stations more similar to C than it was pre-merger, to try to take
listeners from C, as might happen in a spatial model where price competition is relatively
limited, and, as seems plausible for radio, there is limited scope to increase total radio or
format listenership by introducing completely new programming.
47
Sweeting’s results come with the caveat that there is no quasi-experimental variation
in ownership at the station level. Reassuringly, however, the patterns in the data are quite
similar looking at both changes in local market structure that result from very large
national transactions involving many hundreds of stations in different markets and differ-
ent formats, such as the 2000 AMFM–Clear Channel transaction, and local transactions
involving trades of stations in an individual market. For large transactions, the claim that
46
One approach defines different artists as different dimensions of the product space, and then uses a station’s
playlist to identify a location in this high-dimensional space. The difference between two playlists can be
measured by the angle between the location vectors at the origin. Alternative approaches include simply
looking at the proportion of playtime devoted to artists who are not played at all on other stations, and, for
small groups of stations, the total number of different artists played. In a working paper (
Sweeting, 2004),
Sweeting also projected the main artists in a format category into a two-dimensional space and then placed
the stations in this product space based on the artists appearing on their playlists. All of these measures
produce qualitatively similar results.
47
Studies by Borenstein (1986) and Rogers and Woodbury (1996), using data from prior to ownership
deregulation, support the contention that there is significant cannibalization both within programming
formats and at the aggregate level.
366
Handbook of Media Economics