Internet radio services, including Pandora’s music streaming service, also provide a threat,
although traditional broadcasters have tried to develop online services to exploit the
medium. For example, Clear Channel’s platform, iHeart Radio, carries both terrestrial
stations and offers services that allow users to create their own playlists that are designed
to compete with Pandora. In 2013, 45% of the population reported listening to online
radio in the last month, and these people listened for an average of 12 h per week. The
threat that Pandora offers to traditional radio’s revenues, as well as its listenership, has also
increased as the platform now sells local advertising in large markets.
20
At the same time,
however, broadcast stations have exploited new technology to proliferate their channels,
both by offering channels in HD and using translator and booster stations to offer more
programming to their broadcast stations.
21
Finally, while syndicated political talk radio, usually of a conservative bent, was one of
the big growth areas of radio programming in the 1990s and early 2000s, especially for
AM stations, it may now be in retreat.
22
While personalities such as Rush Limbaugh and
Sean Hannity attract widespread media attention, pressure from social media has also
made some advertisers reluctant to advertise on their programs, causing a shift back to
less controversial News or music programming.
23
8.3. DATA
To study any industry in detail requires reliable data, and this is especially important when
trying to study and interpret changes over time as much of the empirical literature on radio
has tried to do. In this section, I identify some of the sources that have been used most
frequently by empirical researchers, and try to indicate some of the issues that need to be
considered when using them.
Unfortunately for researchers, the FCC stopped collecting information on advertising
revenues and programming content in the early 1980s. In its recent work on the industry,
the FCC has relied primarily on the radio version of BIA/Kelsey’s
Media Access Pro
20
http://www.crainsnewyork.com/article/20140421/MEDIA_ENTERTAINMENT/304209988/
pandora-unleashes-sales-force-on-local-market (accessed February 14, 2015).
21
For example, AM stations use FM translators to put AM programming onto the FM band, while boosters
are sometimes used to carry FM signals into places where FM signals cannot easily be received because of
terrain. However, these services can also be used to carry programming that the station is carrying on its
HD channels rather than on its primary AM or FM channel, and they are not included in the station counts
used to determine whether a station satisfies ownership caps (conversation with Mark Fratrik of BIA/
Kelsey, February 13, 2015).
22
The number of stations carrying political talk radio increased from 400 in 1990 to 1400 in 2006 (http://
www.stateofthemedia.org/2007/radio-intro/talk-radio/, accessed February 15, 2015).
23
http://www.wsj.com/articles/talk-radios-advertising-problem-1423011395 (article February 6, 2015,
accessed February 15, 2015).
352 Handbook of Media Economics
database (BIA hereafter).
24
This data has also been used by academic researchers, includ-
ing
Jeziorski (2014a,b) and Sweeting (2009, 2013), who have focused on the post-1996
period. This database contains Arbitron (now Nielsen)
25
station ratings data for at least
the Spring and Fall reporting periods each year, including some measures for demo-
graphic sub-groups, detailed programming format classifications, technical information
(e.g., licensed transmitter power, signal coverage, including contour maps), information
on station personnel, and BIA’s estimates of annual station, as well as market, revenues. As
one would expect, these revenue estimates are closely related to audience size, market
share and format, but it is unclear whether this relationship reflects a very close relation-
ship that exists in reality or just the relationship that BIA assumes when making its esti-
mates. The database also contains a detailed ownership and transaction history for each
station, dating back to well before 1996. For some transactions, a price is recorded, but
when groups of stations, possibly from different markets, are traded, it may be very dif-
ficult to impute a price for each station. Recent additions to the BIA database include
information on HD and multicast programming, and market-level estimates of online
revenues.
Arbitron estimates commercial station ratings in over 270 geographic markets (the
exact set of smaller markets has changed over time due to changes in market population).
Arbitron markets are smaller than the designated marketing areas used to analyze local
television markets, and many of them coincide almost exactly with the Metropolitan Sta-
tistical Areas used by the US Census (
Arbitron Company, 2009). Listening in rural areas
outside of these urban markets is unmeasured. As part of its surveys, Arbitron also collects
data on non-commercial listening. Examples of this data are available through the Radio
Research Consortium (
http://www.rrconline.org). While simple market share data is
reported on a number of radio-related websites, detailed data for listening by specific
demographics, which is valuable to advertisers who want to target specific population
groups, is available through BIA subscriptions and through Arbitron.
The traditional way that Arbitron has measured audiences is by recruiting a sample of
the population and getting them to complete diaries that record stations that they listened
to for at least 5 min during quarter-hour periods. An obvious concern is that diary data
may contain systematic misreporting, especially when listeners change stations frequently
and may not always be aware of exactly which station they are listening to. In response
to this concern, as well as a desire to get data to advertisers and stations more quickly,
Arbitron began to switch to using electronic Portable People Meters (PPMs) in larger
24
http://www.biakelsey.com/Broadcast-Media/Media-Access-Pro (accessed March 1, 2015).
25
Nielsen purchased Arbitron in 2013, rebranding Arbitron as Nielsen Audio (http://www.nielsen.com/
us/en/press-room/2013/nielsen-acquires-arbitron.html, accessed February 15, 2015). As all of the work
discussed in this chapter used data prior to 2013, I will refer to “Arbitron” data.
353Radio
urban markets in 2010.
26
This technology is able to record the programming that the
wearer is able to hear, but of course not all people actively take in the programming that
they are exposed to. For example, many people may encounter radio programs when
they are at a dentist’s office, for example, without actively listening to it.
One result of this difference in what is being recorded is that PPMs have lead to higher
estimates of station “cume” ratings (the proportion of the population who listen to a sta-
tion for at least 5 min during a particular daypart), but station shares, the average propor-
tion of the population listening to a station at any given time, remained pretty much the
same as diarists actually reported the stations that they listened to a lot quite accurately.
The relationship between shares and cume is potentially important because, as will be
discussed below, listeners who only listen to a single station (single-homing) may be
much more valuable to stations than listeners who multi-home. There has been no formal
research to date into whether this change in measurement technology has led to changes
in station programming, even though, as will be discussed in
Section 8.5.2, its introduc-
tion was especially controversial with minority-oriented broadcasters whose estimated
ratings fell. In
Section 8.7, I will also cite some evidence that suggests that PPM data
may, for example, change traditional perceptions about how listeners respond to com-
mercials. PPMs may also have made it more attractive for stations to have “special
programming” segments (for example, a high-profile sports event), as the more accurate
PPM data allows them to make a more convincing case to advertisers that people actually
listen to it.
27
To look at quarters before 1996 researchers cannot make use of BIA, but widespread
use has been made of Duncan’s American Radio quarterly Reports and annual Market
Guides. Duncan ceased publication in 2002, but most of the reports back through the
1970s are available in PDF format, together with a wealth of information from other pub-
lications, on the website
http://www.americanradiohistory.com. The Duncan reports
contain Arbitron ratings and estimates of station revenues (see Duncan (2002) for a dis-
cussion of the sources used for these estimates). Some of Duncan’s revenue numbers actu-
ally come from stations that simply passed on the revenue numbers calculated by the
auditing firms that provide revenue information to licensing organizations such as
ASCAP and BMI. Duncan also contains station ownership information and information
on recent station sales.
The Duncan reports also contain station format information. One thing to be aware
of when using format information in Duncan (a caveat which may also be relevant for
BIA) is that Duncan’s format classification became finer over time. For example,
26
Unless otherwise noted, this discussion is based on “Arbitron PPM President Pierre Bouvard: It’s Radio’s
Turn to Eat at the Adult’s Table”, Radio Ink, August 1, 2009, article by Reed Bunzel.
http://www.
radioink.com/listingsEntry.asp?ID¼370205 (accessed February 18, 2014).
27
Conversation with Mark Fratrik of BIA/Kelsey, February 13, 2015.
354
Handbook of Media Economics
Duncan’s American Radio market report publications had 10 format categories in 1977,
12 in 1986, 21 in 1995, and 37 in 1998. In part, this is to capture changes in radio station
programming, as music formats did become increasingly specialized, partly because reg-
ulatory changes meant that stations did not have to broadcast a certain amount of news
programming, but it also may reflect the fact that over time it became easier for Duncan
to collect and record more detailed information on what stations were doing.
While the BIA or Duncan revenue estimates are useful, they do not facilitate estima-
tion of fundamental demand and supply relationships in advertising markets, as they do
not distinguish between the prices at which advertisers purchased commercials and the
quantities of advertising that are sold. A common source of price data is “SQAD” esti-
mates of CPMs (the cost of reaching a thousand listeners) and CPPs (the cost of reaching
one ratings point of listeners), which are based on actual transactions reported by media
buyers.
28
This data is available at the market level, broken down by demographics and
daypart. One limitation of this data is that it is not station-specific, so estimating the num-
ber of commercials by simply dividing estimated station revenues by prices (which
researchers do in many contexts to estimate quantities) requires the strong assumption
that different stations charge the same prices per listener. Such an assumption is likely
incorrect, given an older literature noting that per-listener prices tend to increase with
market share (
Fisher et al., 1980) and the ability of station groups to potentially extract
more surplus by bundling commercials. In addition, using average prices across stations
will tend to make it harder to identify the effects of particular station mergers, especially
when the number of stations is large.
Data on the quantity of commercials aired is more difficult to find, and typically relies
on creating or identifying detailed airplay data. Knowing whether the quantity of com-
mercials has increased or decreased is potentially important for understanding how con-
solidation has affected the welfare of advertisers. For example,
Duncan (2004, p. 5) argues
that most of the growth in radio industry revenues in the mid and late 1990s, when rapid
consolidation happened, was “caused by expanded inventory of radio spots. The healthy
economy used up the increased inventory but only at compressed prices.” As part of its
2006 media ownership review, the FCC commissioned a detailed analysis of airplay data
that recorded information on many different types of content, including the amount of
time spent on commercials. This data on 1014 stations has been used by
Chipty (2007)
and Mooney (2010a). The limitations are that it contains information only on a single
cross-section and that because the material was collected by a team of listeners, there
is only 2 h of programming for each station. As a result there may be considerable noise
in inferring the average behavior of each station even in the quarter that the cross-section
was taken.
Sweeting (2009, 2010) uses panel data from 1998 to 2001 that was collected by
28
Details of this data is available at http://sqad.com/products/#spot-radio. SQAD data also appears in the
Standard Rate and Data Service reports, http://next.srds.com/for-media/premium-data.
355Radio
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.